From 80f72e687cfb3ffcee1547d4dc67ec4b59dad871 Mon Sep 17 00:00:00 2001
From: "K. S. Ernest (iFire) Lee" <ernest.lee@chibifire.com>
Date: Thu, 9 May 2024 05:09:25 -0700
Subject: [PATCH] Add csg boolean operators using elalish/manifold.

---
 modules/csg/SCsub                             |   59 +
 modules/csg/csg.cpp                           | 1547 ----
 modules/csg/csg.h                             |  204 -
 modules/csg/csg_shape.cpp                     |  320 +-
 modules/csg/csg_shape.h                       |    3 +-
 thirdparty/manifold/.gitrepo                  |   12 +
 thirdparty/manifold/AUTHORS                   |   10 +
 thirdparty/manifold/LICENSE                   |  201 +
 .../manifold/src/collider/include/collider.h  |   46 +
 .../manifold/src/collider/src/collider.cpp    |  395 +
 .../src/cross_section/include/cross_section.h |  176 +
 .../src/cross_section/src/cross_section.cpp   |  795 ++
 .../manifold/src/manifold/include/manifold.h  |  311 +
 .../manifold/src/manifold/src/boolean3.cpp    |  602 ++
 .../manifold/src/manifold/src/boolean3.h      |   60 +
 .../src/manifold/src/boolean_result.cpp       |  831 ++
 .../src/manifold/src/constructors.cpp         |  518 ++
 .../manifold/src/manifold/src/csg_tree.cpp    |  651 ++
 .../manifold/src/manifold/src/csg_tree.h      |  109 +
 .../manifold/src/manifold/src/edge_op.cpp     |  678 ++
 .../manifold/src/manifold/src/face_op.cpp     |  321 +
 thirdparty/manifold/src/manifold/src/impl.cpp |  956 +++
 thirdparty/manifold/src/manifold/src/impl.h   |  190 +
 .../manifold/src/manifold/src/manifold.cpp    |  970 +++
 .../manifold/src/manifold/src/mesh_fixes.h    |   57 +
 .../manifold/src/manifold/src/properties.cpp  |  418 +
 thirdparty/manifold/src/manifold/src/shared.h |  232 +
 .../manifold/src/manifold/src/smoothing.cpp   |  847 +++
 thirdparty/manifold/src/manifold/src/sort.cpp |  616 ++
 .../manifold/src/manifold/src/subdivision.cpp |  792 ++
 .../manifold/src/polygon/include/polygon.h    |   51 +
 .../manifold/src/polygon/src/polygon.cpp      |  877 +++
 thirdparty/manifold/src/sdf/include/sdf.h     |   24 +
 thirdparty/manifold/src/sdf/src/sdf.cpp       |  362 +
 .../src/third_party/quickhull/.gitignore      |    1 +
 .../src/third_party/quickhull/.gitrepo        |   12 +
 .../src/third_party/quickhull/ConvexHull.hpp  |  182 +
 .../third_party/quickhull/HalfEdgeMesh.hpp    |   77 +
 .../src/third_party/quickhull/MathUtils.hpp   |   46 +
 .../src/third_party/quickhull/QuickHull.cpp   |  503 ++
 .../src/third_party/quickhull/QuickHull.hpp   |  223 +
 .../src/third_party/quickhull/README.md       |   31 +
 .../third_party/quickhull/Structs/Mesh.hpp    |  255 +
 .../third_party/quickhull/Structs/Plane.hpp   |   36 +
 .../third_party/quickhull/Structs/Pool.hpp    |   35 +
 .../src/third_party/quickhull/Structs/Ray.hpp |   21 +
 .../third_party/quickhull/Structs/Vector3.hpp |  140 +
 .../quickhull/Structs/VertexDataSource.hpp    |   48 +
 .../src/utilities/include/hashtable.h         |  174 +
 .../src/utilities/include/optional_assert.h   |   38 +
 .../manifold/src/utilities/include/par.h      |  195 +
 .../manifold/src/utilities/include/public.h   |  696 ++
 .../manifold/src/utilities/include/sparse.h   |  201 +
 .../manifold/src/utilities/include/svd.h      |  310 +
 .../manifold/src/utilities/include/tri_dist.h |  225 +
 .../manifold/src/utilities/include/utils.h    |  242 +
 .../manifold/src/utilities/include/vec.h      |  244 +
 .../manifold/src/utilities/include/vec_view.h |  116 +
 thirdparty/manifold/thirdparty/glm/.gitrepo   |   12 +
 .../manifold/thirdparty/glm/copying.txt       |   54 +
 .../thirdparty/glm/glm/CMakeLists.txt         |   69 +
 .../manifold/thirdparty/glm/glm/common.hpp    |  539 ++
 .../thirdparty/glm/glm/detail/_features.hpp   |  394 +
 .../thirdparty/glm/glm/detail/_fixes.hpp      |   27 +
 .../thirdparty/glm/glm/detail/_noise.hpp      |   81 +
 .../thirdparty/glm/glm/detail/_swizzle.hpp    |  804 ++
 .../glm/glm/detail/_swizzle_func.hpp          |  682 ++
 .../thirdparty/glm/glm/detail/_vectorize.hpp  |  162 +
 .../glm/glm/detail/compute_common.hpp         |   50 +
 .../glm/detail/compute_vector_relational.hpp  |   30 +
 .../thirdparty/glm/glm/detail/func_common.inl |  792 ++
 .../glm/glm/detail/func_common_simd.inl       |  231 +
 .../glm/glm/detail/func_exponential.inl       |  152 +
 .../glm/glm/detail/func_exponential_simd.inl  |   37 +
 .../glm/glm/detail/func_geometric.inl         |  243 +
 .../glm/glm/detail/func_geometric_simd.inl    |  163 +
 .../glm/glm/detail/func_integer.inl           |  369 +
 .../glm/glm/detail/func_integer_simd.inl      |   65 +
 .../thirdparty/glm/glm/detail/func_matrix.inl |  443 ++
 .../glm/glm/detail/func_matrix_simd.inl       |  252 +
 .../glm/glm/detail/func_packing.inl           |  189 +
 .../glm/glm/detail/func_packing_simd.inl      |    6 +
 .../glm/glm/detail/func_trigonometric.inl     |  197 +
 .../glm/detail/func_trigonometric_simd.inl    |    0
 .../glm/glm/detail/func_vector_relational.inl |   87 +
 .../detail/func_vector_relational_simd.inl    |    6 +
 .../thirdparty/glm/glm/detail/glm.cpp         |  263 +
 .../thirdparty/glm/glm/detail/qualifier.hpp   |  229 +
 .../thirdparty/glm/glm/detail/setup.hpp       | 1167 +++
 .../thirdparty/glm/glm/detail/type_float.hpp  |   68 +
 .../thirdparty/glm/glm/detail/type_half.hpp   |   16 +
 .../thirdparty/glm/glm/detail/type_half.inl   |  241 +
 .../thirdparty/glm/glm/detail/type_mat2x2.hpp |  177 +
 .../thirdparty/glm/glm/detail/type_mat2x2.inl |  536 ++
 .../thirdparty/glm/glm/detail/type_mat2x3.hpp |  159 +
 .../thirdparty/glm/glm/detail/type_mat2x3.inl |  510 ++
 .../thirdparty/glm/glm/detail/type_mat2x4.hpp |  161 +
 .../thirdparty/glm/glm/detail/type_mat2x4.inl |  520 ++
 .../thirdparty/glm/glm/detail/type_mat3x2.hpp |  167 +
 .../thirdparty/glm/glm/detail/type_mat3x2.inl |  532 ++
 .../thirdparty/glm/glm/detail/type_mat3x3.hpp |  184 +
 .../thirdparty/glm/glm/detail/type_mat3x3.inl |  601 ++
 .../thirdparty/glm/glm/detail/type_mat3x4.hpp |  166 +
 .../thirdparty/glm/glm/detail/type_mat3x4.inl |  578 ++
 .../thirdparty/glm/glm/detail/type_mat4x2.hpp |  171 +
 .../thirdparty/glm/glm/detail/type_mat4x2.inl |  574 ++
 .../thirdparty/glm/glm/detail/type_mat4x3.hpp |  171 +
 .../thirdparty/glm/glm/detail/type_mat4x3.inl |  598 ++
 .../thirdparty/glm/glm/detail/type_mat4x4.hpp |  189 +
 .../thirdparty/glm/glm/detail/type_mat4x4.inl |  706 ++
 .../glm/glm/detail/type_mat4x4_simd.inl       |    6 +
 .../thirdparty/glm/glm/detail/type_quat.hpp   |  193 +
 .../thirdparty/glm/glm/detail/type_quat.inl   |  424 ++
 .../glm/glm/detail/type_quat_simd.inl         |  208 +
 .../thirdparty/glm/glm/detail/type_vec1.hpp   |  308 +
 .../thirdparty/glm/glm/detail/type_vec1.inl   |  553 ++
 .../thirdparty/glm/glm/detail/type_vec2.hpp   |  402 +
 .../thirdparty/glm/glm/detail/type_vec2.inl   |  915 +++
 .../thirdparty/glm/glm/detail/type_vec3.hpp   |  435 ++
 .../thirdparty/glm/glm/detail/type_vec3.inl   | 1070 +++
 .../thirdparty/glm/glm/detail/type_vec4.hpp   |  508 ++
 .../thirdparty/glm/glm/detail/type_vec4.inl   | 1142 +++
 .../glm/glm/detail/type_vec4_simd.inl         |  788 ++
 .../thirdparty/glm/glm/exponential.hpp        |  110 +
 .../manifold/thirdparty/glm/glm/ext.hpp       |  266 +
 .../glm/glm/ext/_matrix_vectorize.hpp         |  128 +
 .../glm/glm/ext/matrix_clip_space.hpp         |  522 ++
 .../glm/glm/ext/matrix_clip_space.inl         |  595 ++
 .../thirdparty/glm/glm/ext/matrix_common.hpp  |   39 +
 .../thirdparty/glm/glm/ext/matrix_common.inl  |   34 +
 .../glm/glm/ext/matrix_double2x2.hpp          |   23 +
 .../glm/ext/matrix_double2x2_precision.hpp    |   49 +
 .../glm/glm/ext/matrix_double2x3.hpp          |   18 +
 .../glm/ext/matrix_double2x3_precision.hpp    |   31 +
 .../glm/glm/ext/matrix_double2x4.hpp          |   18 +
 .../glm/ext/matrix_double2x4_precision.hpp    |   31 +
 .../glm/glm/ext/matrix_double3x2.hpp          |   18 +
 .../glm/ext/matrix_double3x2_precision.hpp    |   31 +
 .../glm/glm/ext/matrix_double3x3.hpp          |   23 +
 .../glm/ext/matrix_double3x3_precision.hpp    |   49 +
 .../glm/glm/ext/matrix_double3x4.hpp          |   18 +
 .../glm/ext/matrix_double3x4_precision.hpp    |   31 +
 .../glm/glm/ext/matrix_double4x2.hpp          |   18 +
 .../glm/ext/matrix_double4x2_precision.hpp    |   31 +
 .../glm/glm/ext/matrix_double4x3.hpp          |   18 +
 .../glm/ext/matrix_double4x3_precision.hpp    |   31 +
 .../glm/glm/ext/matrix_double4x4.hpp          |   23 +
 .../glm/ext/matrix_double4x4_precision.hpp    |   49 +
 .../glm/glm/ext/matrix_float2x2.hpp           |   23 +
 .../glm/glm/ext/matrix_float2x2_precision.hpp |   49 +
 .../glm/glm/ext/matrix_float2x3.hpp           |   18 +
 .../glm/glm/ext/matrix_float2x3_precision.hpp |   31 +
 .../glm/glm/ext/matrix_float2x4.hpp           |   18 +
 .../glm/glm/ext/matrix_float2x4_precision.hpp |   31 +
 .../glm/glm/ext/matrix_float3x2.hpp           |   18 +
 .../glm/glm/ext/matrix_float3x2_precision.hpp |   31 +
 .../glm/glm/ext/matrix_float3x3.hpp           |   23 +
 .../glm/glm/ext/matrix_float3x3_precision.hpp |   49 +
 .../glm/glm/ext/matrix_float3x4.hpp           |   18 +
 .../glm/glm/ext/matrix_float3x4_precision.hpp |   31 +
 .../glm/glm/ext/matrix_float4x2.hpp           |   18 +
 .../glm/glm/ext/matrix_float4x2_precision.hpp |   31 +
 .../glm/glm/ext/matrix_float4x3.hpp           |   18 +
 .../glm/glm/ext/matrix_float4x3_precision.hpp |   31 +
 .../glm/glm/ext/matrix_float4x4.hpp           |   23 +
 .../glm/glm/ext/matrix_float4x4_precision.hpp |   49 +
 .../thirdparty/glm/glm/ext/matrix_int2x2.hpp  |   38 +
 .../glm/glm/ext/matrix_int2x2_sized.hpp       |   70 +
 .../thirdparty/glm/glm/ext/matrix_int2x3.hpp  |   33 +
 .../glm/glm/ext/matrix_int2x3_sized.hpp       |   49 +
 .../thirdparty/glm/glm/ext/matrix_int2x4.hpp  |   33 +
 .../glm/glm/ext/matrix_int2x4_sized.hpp       |   49 +
 .../thirdparty/glm/glm/ext/matrix_int3x2.hpp  |   33 +
 .../glm/glm/ext/matrix_int3x2_sized.hpp       |   49 +
 .../thirdparty/glm/glm/ext/matrix_int3x3.hpp  |   38 +
 .../glm/glm/ext/matrix_int3x3_sized.hpp       |   70 +
 .../thirdparty/glm/glm/ext/matrix_int3x4.hpp  |   33 +
 .../glm/glm/ext/matrix_int3x4_sized.hpp       |   49 +
 .../thirdparty/glm/glm/ext/matrix_int4x2.hpp  |   33 +
 .../glm/glm/ext/matrix_int4x2_sized.hpp       |   49 +
 .../thirdparty/glm/glm/ext/matrix_int4x3.hpp  |   33 +
 .../glm/glm/ext/matrix_int4x3_sized.hpp       |   49 +
 .../thirdparty/glm/glm/ext/matrix_int4x4.hpp  |   38 +
 .../glm/glm/ext/matrix_int4x4_sized.hpp       |   70 +
 .../thirdparty/glm/glm/ext/matrix_integer.hpp |   91 +
 .../thirdparty/glm/glm/ext/matrix_integer.inl |   38 +
 .../glm/glm/ext/matrix_projection.hpp         |  149 +
 .../glm/glm/ext/matrix_projection.inl         |  106 +
 .../glm/glm/ext/matrix_relational.hpp         |  132 +
 .../glm/glm/ext/matrix_relational.inl         |   88 +
 .../glm/glm/ext/matrix_transform.hpp          |  171 +
 .../glm/glm/ext/matrix_transform.inl          |  207 +
 .../thirdparty/glm/glm/ext/matrix_uint2x2.hpp |   38 +
 .../glm/glm/ext/matrix_uint2x2_sized.hpp      |   70 +
 .../thirdparty/glm/glm/ext/matrix_uint2x3.hpp |   33 +
 .../glm/glm/ext/matrix_uint2x3_sized.hpp      |   49 +
 .../thirdparty/glm/glm/ext/matrix_uint2x4.hpp |   33 +
 .../glm/glm/ext/matrix_uint2x4_sized.hpp      |   49 +
 .../thirdparty/glm/glm/ext/matrix_uint3x2.hpp |   33 +
 .../glm/glm/ext/matrix_uint3x2_sized.hpp      |   49 +
 .../thirdparty/glm/glm/ext/matrix_uint3x3.hpp |   38 +
 .../glm/glm/ext/matrix_uint3x3_sized.hpp      |   70 +
 .../thirdparty/glm/glm/ext/matrix_uint3x4.hpp |   33 +
 .../glm/glm/ext/matrix_uint3x4_sized.hpp      |   49 +
 .../thirdparty/glm/glm/ext/matrix_uint4x2.hpp |   33 +
 .../glm/glm/ext/matrix_uint4x2_sized.hpp      |   49 +
 .../thirdparty/glm/glm/ext/matrix_uint4x3.hpp |   33 +
 .../glm/glm/ext/matrix_uint4x3_sized.hpp      |   49 +
 .../thirdparty/glm/glm/ext/matrix_uint4x4.hpp |   38 +
 .../glm/glm/ext/matrix_uint4x4_sized.hpp      |   70 +
 .../glm/glm/ext/quaternion_common.hpp         |  135 +
 .../glm/glm/ext/quaternion_common.inl         |  144 +
 .../glm/glm/ext/quaternion_common_simd.inl    |   18 +
 .../glm/glm/ext/quaternion_double.hpp         |   39 +
 .../glm/ext/quaternion_double_precision.hpp   |   42 +
 .../glm/glm/ext/quaternion_exponential.hpp    |   63 +
 .../glm/glm/ext/quaternion_exponential.inl    |   89 +
 .../glm/glm/ext/quaternion_float.hpp          |   39 +
 .../glm/ext/quaternion_float_precision.hpp    |   36 +
 .../glm/glm/ext/quaternion_geometric.hpp      |   70 +
 .../glm/glm/ext/quaternion_geometric.inl      |   36 +
 .../glm/glm/ext/quaternion_relational.hpp     |   62 +
 .../glm/glm/ext/quaternion_relational.inl     |   35 +
 .../glm/glm/ext/quaternion_transform.hpp      |   47 +
 .../glm/glm/ext/quaternion_transform.inl      |   24 +
 .../glm/glm/ext/quaternion_trigonometric.hpp  |   65 +
 .../glm/glm/ext/quaternion_trigonometric.inl  |   37 +
 .../thirdparty/glm/glm/ext/scalar_common.hpp  |  181 +
 .../thirdparty/glm/glm/ext/scalar_common.inl  |  170 +
 .../glm/glm/ext/scalar_constants.hpp          |   40 +
 .../glm/glm/ext/scalar_constants.inl          |   24 +
 .../glm/glm/ext/scalar_int_sized.hpp          |   70 +
 .../thirdparty/glm/glm/ext/scalar_integer.hpp |   92 +
 .../thirdparty/glm/glm/ext/scalar_integer.inl |  243 +
 .../thirdparty/glm/glm/ext/scalar_packing.hpp |   32 +
 .../thirdparty/glm/glm/ext/scalar_packing.inl |    0
 .../glm/glm/ext/scalar_reciprocal.hpp         |  135 +
 .../glm/glm/ext/scalar_reciprocal.inl         |  107 +
 .../glm/glm/ext/scalar_relational.hpp         |   68 +
 .../glm/glm/ext/scalar_relational.inl         |   40 +
 .../glm/glm/ext/scalar_uint_sized.hpp         |   70 +
 .../thirdparty/glm/glm/ext/scalar_ulp.hpp     |   77 +
 .../thirdparty/glm/glm/ext/scalar_ulp.inl     |  284 +
 .../thirdparty/glm/glm/ext/vector_bool1.hpp   |   30 +
 .../glm/glm/ext/vector_bool1_precision.hpp    |   34 +
 .../thirdparty/glm/glm/ext/vector_bool2.hpp   |   18 +
 .../glm/glm/ext/vector_bool2_precision.hpp    |   31 +
 .../thirdparty/glm/glm/ext/vector_bool3.hpp   |   18 +
 .../glm/glm/ext/vector_bool3_precision.hpp    |   31 +
 .../thirdparty/glm/glm/ext/vector_bool4.hpp   |   18 +
 .../glm/glm/ext/vector_bool4_precision.hpp    |   31 +
 .../thirdparty/glm/glm/ext/vector_common.hpp  |  228 +
 .../thirdparty/glm/glm/ext/vector_common.inl  |  147 +
 .../thirdparty/glm/glm/ext/vector_double1.hpp |   31 +
 .../glm/glm/ext/vector_double1_precision.hpp  |   36 +
 .../thirdparty/glm/glm/ext/vector_double2.hpp |   18 +
 .../glm/glm/ext/vector_double2_precision.hpp  |   31 +
 .../thirdparty/glm/glm/ext/vector_double3.hpp |   18 +
 .../glm/glm/ext/vector_double3_precision.hpp  |   34 +
 .../thirdparty/glm/glm/ext/vector_double4.hpp |   18 +
 .../glm/glm/ext/vector_double4_precision.hpp  |   35 +
 .../thirdparty/glm/glm/ext/vector_float1.hpp  |   31 +
 .../glm/glm/ext/vector_float1_precision.hpp   |   36 +
 .../thirdparty/glm/glm/ext/vector_float2.hpp  |   18 +
 .../glm/glm/ext/vector_float2_precision.hpp   |   31 +
 .../thirdparty/glm/glm/ext/vector_float3.hpp  |   18 +
 .../glm/glm/ext/vector_float3_precision.hpp   |   31 +
 .../thirdparty/glm/glm/ext/vector_float4.hpp  |   18 +
 .../glm/glm/ext/vector_float4_precision.hpp   |   31 +
 .../thirdparty/glm/glm/ext/vector_int1.hpp    |   32 +
 .../glm/glm/ext/vector_int1_sized.hpp         |   49 +
 .../thirdparty/glm/glm/ext/vector_int2.hpp    |   18 +
 .../glm/glm/ext/vector_int2_sized.hpp         |   49 +
 .../thirdparty/glm/glm/ext/vector_int3.hpp    |   18 +
 .../glm/glm/ext/vector_int3_sized.hpp         |   49 +
 .../thirdparty/glm/glm/ext/vector_int4.hpp    |   18 +
 .../glm/glm/ext/vector_int4_sized.hpp         |   49 +
 .../thirdparty/glm/glm/ext/vector_integer.hpp |  149 +
 .../thirdparty/glm/glm/ext/vector_integer.inl |   85 +
 .../thirdparty/glm/glm/ext/vector_packing.hpp |   32 +
 .../thirdparty/glm/glm/ext/vector_packing.inl |    0
 .../glm/glm/ext/vector_reciprocal.hpp         |  135 +
 .../glm/glm/ext/vector_reciprocal.inl         |  105 +
 .../glm/glm/ext/vector_relational.hpp         |  107 +
 .../glm/glm/ext/vector_relational.inl         |   75 +
 .../thirdparty/glm/glm/ext/vector_uint1.hpp   |   32 +
 .../glm/glm/ext/vector_uint1_sized.hpp        |   49 +
 .../thirdparty/glm/glm/ext/vector_uint2.hpp   |   18 +
 .../glm/glm/ext/vector_uint2_sized.hpp        |   49 +
 .../thirdparty/glm/glm/ext/vector_uint3.hpp   |   18 +
 .../glm/glm/ext/vector_uint3_sized.hpp        |   49 +
 .../thirdparty/glm/glm/ext/vector_uint4.hpp   |   18 +
 .../glm/glm/ext/vector_uint4_sized.hpp        |   49 +
 .../thirdparty/glm/glm/ext/vector_ulp.hpp     |  112 +
 .../thirdparty/glm/glm/ext/vector_ulp.inl     |   74 +
 .../manifold/thirdparty/glm/glm/fwd.hpp       | 1233 +++
 .../manifold/thirdparty/glm/glm/geometric.hpp |  116 +
 .../manifold/thirdparty/glm/glm/glm.cppm      | 2675 +++++++
 .../manifold/thirdparty/glm/glm/glm.hpp       |  137 +
 .../thirdparty/glm/glm/gtc/bitfield.hpp       |  266 +
 .../thirdparty/glm/glm/gtc/bitfield.inl       |  626 ++
 .../thirdparty/glm/glm/gtc/color_space.hpp    |   56 +
 .../thirdparty/glm/glm/gtc/color_space.inl    |   84 +
 .../thirdparty/glm/glm/gtc/constants.hpp      |  170 +
 .../thirdparty/glm/glm/gtc/constants.inl      |  173 +
 .../thirdparty/glm/glm/gtc/epsilon.hpp        |   60 +
 .../thirdparty/glm/glm/gtc/epsilon.inl        |   80 +
 .../thirdparty/glm/glm/gtc/integer.hpp        |   43 +
 .../thirdparty/glm/glm/gtc/integer.inl        |   33 +
 .../thirdparty/glm/glm/gtc/matrix_access.hpp  |   60 +
 .../thirdparty/glm/glm/gtc/matrix_access.inl  |   62 +
 .../thirdparty/glm/glm/gtc/matrix_integer.hpp |  433 ++
 .../thirdparty/glm/glm/gtc/matrix_inverse.hpp |   50 +
 .../thirdparty/glm/glm/gtc/matrix_inverse.inl |  118 +
 .../glm/glm/gtc/matrix_transform.hpp          |   36 +
 .../glm/glm/gtc/matrix_transform.inl          |    3 +
 .../manifold/thirdparty/glm/glm/gtc/noise.hpp |   61 +
 .../manifold/thirdparty/glm/glm/gtc/noise.inl |  807 ++
 .../thirdparty/glm/glm/gtc/packing.hpp        |  728 ++
 .../thirdparty/glm/glm/gtc/packing.inl        |  938 +++
 .../thirdparty/glm/glm/gtc/quaternion.hpp     |  173 +
 .../thirdparty/glm/glm/gtc/quaternion.inl     |  208 +
 .../glm/glm/gtc/quaternion_simd.inl           |    0
 .../thirdparty/glm/glm/gtc/random.hpp         |   82 +
 .../thirdparty/glm/glm/gtc/random.inl         |  303 +
 .../thirdparty/glm/glm/gtc/reciprocal.hpp     |   24 +
 .../manifold/thirdparty/glm/glm/gtc/round.hpp |  160 +
 .../manifold/thirdparty/glm/glm/gtc/round.inl |  155 +
 .../thirdparty/glm/glm/gtc/type_aligned.hpp   | 1315 ++++
 .../thirdparty/glm/glm/gtc/type_precision.hpp | 2094 +++++
 .../thirdparty/glm/glm/gtc/type_precision.inl |    6 +
 .../thirdparty/glm/glm/gtc/type_ptr.hpp       |  230 +
 .../thirdparty/glm/glm/gtc/type_ptr.inl       |  386 +
 .../manifold/thirdparty/glm/glm/gtc/ulp.hpp   |  155 +
 .../manifold/thirdparty/glm/glm/gtc/ulp.inl   |  173 +
 .../manifold/thirdparty/glm/glm/gtc/vec1.hpp  |   30 +
 .../glm/glm/gtx/associated_min_max.hpp        |  207 +
 .../glm/glm/gtx/associated_min_max.inl        |  354 +
 .../manifold/thirdparty/glm/glm/gtx/bit.hpp   |   98 +
 .../manifold/thirdparty/glm/glm/gtx/bit.inl   |   92 +
 .../thirdparty/glm/glm/gtx/closest_point.hpp  |   49 +
 .../thirdparty/glm/glm/gtx/closest_point.inl  |   45 +
 .../thirdparty/glm/glm/gtx/color_encoding.hpp |   54 +
 .../thirdparty/glm/glm/gtx/color_encoding.inl |   45 +
 .../thirdparty/glm/glm/gtx/color_space.hpp    |   72 +
 .../thirdparty/glm/glm/gtx/color_space.inl    |  141 +
 .../glm/glm/gtx/color_space_YCoCg.hpp         |   60 +
 .../glm/glm/gtx/color_space_YCoCg.inl         |  107 +
 .../thirdparty/glm/glm/gtx/common.hpp         |   76 +
 .../thirdparty/glm/glm/gtx/common.inl         |  125 +
 .../thirdparty/glm/glm/gtx/compatibility.hpp  |  133 +
 .../thirdparty/glm/glm/gtx/compatibility.inl  |   62 +
 .../thirdparty/glm/glm/gtx/component_wise.hpp |   69 +
 .../thirdparty/glm/glm/gtx/component_wise.inl |  127 +
 .../glm/glm/gtx/dual_quaternion.hpp           |  274 +
 .../glm/glm/gtx/dual_quaternion.inl           |  352 +
 .../thirdparty/glm/glm/gtx/easing.hpp         |  219 +
 .../thirdparty/glm/glm/gtx/easing.inl         |  436 ++
 .../thirdparty/glm/glm/gtx/euler_angles.hpp   |  335 +
 .../thirdparty/glm/glm/gtx/euler_angles.inl   |  899 +++
 .../thirdparty/glm/glm/gtx/extend.hpp         |   42 +
 .../thirdparty/glm/glm/gtx/extend.inl         |   48 +
 .../glm/glm/gtx/extended_min_max.hpp          |  137 +
 .../glm/glm/gtx/extended_min_max.inl          |  138 +
 .../glm/glm/gtx/exterior_product.hpp          |   45 +
 .../glm/glm/gtx/exterior_product.inl          |   26 +
 .../glm/glm/gtx/fast_exponential.hpp          |   95 +
 .../glm/glm/gtx/fast_exponential.inl          |  136 +
 .../glm/glm/gtx/fast_square_root.hpp          |   98 +
 .../glm/glm/gtx/fast_square_root.inl          |   75 +
 .../glm/glm/gtx/fast_trigonometry.hpp         |   79 +
 .../glm/glm/gtx/fast_trigonometry.inl         |  142 +
 .../glm/glm/gtx/float_notmalize.inl           |   13 +
 .../thirdparty/glm/glm/gtx/functions.hpp      |   56 +
 .../thirdparty/glm/glm/gtx/functions.inl      |   30 +
 .../thirdparty/glm/glm/gtx/gradient_paint.hpp |   53 +
 .../thirdparty/glm/glm/gtx/gradient_paint.inl |   36 +
 .../glm/glm/gtx/handed_coordinate_space.hpp   |   50 +
 .../glm/glm/gtx/handed_coordinate_space.inl   |   26 +
 .../manifold/thirdparty/glm/glm/gtx/hash.hpp  |  146 +
 .../manifold/thirdparty/glm/glm/gtx/hash.inl  |  175 +
 .../thirdparty/glm/glm/gtx/integer.hpp        |   76 +
 .../thirdparty/glm/glm/gtx/integer.inl        |  185 +
 .../thirdparty/glm/glm/gtx/intersect.hpp      |   92 +
 .../thirdparty/glm/glm/gtx/intersect.inl      |  200 +
 .../manifold/thirdparty/glm/glm/gtx/io.hpp    |  201 +
 .../manifold/thirdparty/glm/glm/gtx/io.inl    |  440 ++
 .../thirdparty/glm/glm/gtx/log_base.hpp       |   48 +
 .../thirdparty/glm/glm/gtx/log_base.inl       |   16 +
 .../glm/glm/gtx/matrix_cross_product.hpp      |   47 +
 .../glm/glm/gtx/matrix_cross_product.inl      |   37 +
 .../glm/glm/gtx/matrix_decompose.hpp          |   52 +
 .../glm/glm/gtx/matrix_decompose.inl          |  234 +
 .../glm/glm/gtx/matrix_factorisation.hpp      |   69 +
 .../glm/glm/gtx/matrix_factorisation.inl      |   84 +
 .../glm/glm/gtx/matrix_interpolation.hpp      |   60 +
 .../glm/glm/gtx/matrix_interpolation.inl      |  146 +
 .../glm/glm/gtx/matrix_major_storage.hpp      |  119 +
 .../glm/glm/gtx/matrix_major_storage.inl      |  166 +
 .../glm/glm/gtx/matrix_operation.hpp          |  103 +
 .../glm/glm/gtx/matrix_operation.inl          |  176 +
 .../thirdparty/glm/glm/gtx/matrix_query.hpp   |   77 +
 .../thirdparty/glm/glm/gtx/matrix_query.inl   |  119 +
 .../glm/glm/gtx/matrix_transform_2d.hpp       |   81 +
 .../glm/glm/gtx/matrix_transform_2d.inl       |   68 +
 .../thirdparty/glm/glm/gtx/mixed_product.hpp  |   41 +
 .../thirdparty/glm/glm/gtx/mixed_product.inl  |   15 +
 .../manifold/thirdparty/glm/glm/gtx/norm.hpp  |   88 +
 .../manifold/thirdparty/glm/glm/gtx/norm.inl  |   95 +
 .../thirdparty/glm/glm/gtx/normal.hpp         |   41 +
 .../thirdparty/glm/glm/gtx/normal.inl         |   15 +
 .../thirdparty/glm/glm/gtx/normalize_dot.hpp  |   49 +
 .../thirdparty/glm/glm/gtx/normalize_dot.inl  |   16 +
 .../glm/glm/gtx/number_precision.hpp          |   47 +
 .../thirdparty/glm/glm/gtx/optimum_pow.hpp    |   52 +
 .../thirdparty/glm/glm/gtx/optimum_pow.inl    |   22 +
 .../thirdparty/glm/glm/gtx/orthonormalize.hpp |   49 +
 .../thirdparty/glm/glm/gtx/orthonormalize.inl |   29 +
 .../manifold/thirdparty/glm/glm/gtx/pca.hpp   |  115 +
 .../manifold/thirdparty/glm/glm/gtx/pca.inl   |  343 +
 .../thirdparty/glm/glm/gtx/perpendicular.hpp  |   41 +
 .../thirdparty/glm/glm/gtx/perpendicular.inl  |   10 +
 .../glm/glm/gtx/polar_coordinates.hpp         |   48 +
 .../glm/glm/gtx/polar_coordinates.inl         |   36 +
 .../thirdparty/glm/glm/gtx/projection.hpp     |   43 +
 .../thirdparty/glm/glm/gtx/projection.inl     |   10 +
 .../thirdparty/glm/glm/gtx/quaternion.hpp     |  174 +
 .../thirdparty/glm/glm/gtx/quaternion.inl     |  159 +
 .../manifold/thirdparty/glm/glm/gtx/range.hpp |   98 +
 .../thirdparty/glm/glm/gtx/raw_data.hpp       |   51 +
 .../thirdparty/glm/glm/gtx/raw_data.inl       |    2 +
 .../glm/glm/gtx/rotate_normalized_axis.hpp    |   68 +
 .../glm/glm/gtx/rotate_normalized_axis.inl    |   58 +
 .../thirdparty/glm/glm/gtx/rotate_vector.hpp  |  123 +
 .../thirdparty/glm/glm/gtx/rotate_vector.inl  |  187 +
 .../glm/glm/gtx/scalar_multiplication.hpp     |   82 +
 .../glm/glm/gtx/scalar_relational.hpp         |   36 +
 .../glm/glm/gtx/scalar_relational.inl         |   88 +
 .../thirdparty/glm/glm/gtx/spline.hpp         |   65 +
 .../thirdparty/glm/glm/gtx/spline.inl         |   60 +
 .../thirdparty/glm/glm/gtx/std_based_type.hpp |   68 +
 .../thirdparty/glm/glm/gtx/std_based_type.inl |    6 +
 .../thirdparty/glm/glm/gtx/string_cast.hpp    |   46 +
 .../thirdparty/glm/glm/gtx/string_cast.inl    |  492 ++
 .../thirdparty/glm/glm/gtx/texture.hpp        |   46 +
 .../thirdparty/glm/glm/gtx/texture.inl        |   17 +
 .../thirdparty/glm/glm/gtx/transform.hpp      |   60 +
 .../thirdparty/glm/glm/gtx/transform.inl      |   23 +
 .../thirdparty/glm/glm/gtx/transform2.hpp     |   89 +
 .../thirdparty/glm/glm/gtx/transform2.inl     |  125 +
 .../thirdparty/glm/glm/gtx/type_aligned.hpp   |  982 +++
 .../thirdparty/glm/glm/gtx/type_aligned.inl   |    6 +
 .../thirdparty/glm/glm/gtx/type_trait.hpp     |   85 +
 .../thirdparty/glm/glm/gtx/type_trait.inl     |   61 +
 .../thirdparty/glm/glm/gtx/vec_swizzle.hpp    | 2786 +++++++
 .../thirdparty/glm/glm/gtx/vector_angle.hpp   |   57 +
 .../thirdparty/glm/glm/gtx/vector_angle.inl   |   45 +
 .../thirdparty/glm/glm/gtx/vector_query.hpp   |   66 +
 .../thirdparty/glm/glm/gtx/vector_query.inl   |  154 +
 .../manifold/thirdparty/glm/glm/gtx/wrap.hpp  |   37 +
 .../manifold/thirdparty/glm/glm/gtx/wrap.inl  |    6 +
 .../manifold/thirdparty/glm/glm/integer.hpp   |  212 +
 .../manifold/thirdparty/glm/glm/mat2x2.hpp    |    9 +
 .../manifold/thirdparty/glm/glm/mat2x3.hpp    |    9 +
 .../manifold/thirdparty/glm/glm/mat2x4.hpp    |    9 +
 .../manifold/thirdparty/glm/glm/mat3x2.hpp    |    9 +
 .../manifold/thirdparty/glm/glm/mat3x3.hpp    |    8 +
 .../manifold/thirdparty/glm/glm/mat3x4.hpp    |    8 +
 .../manifold/thirdparty/glm/glm/mat4x2.hpp    |    9 +
 .../manifold/thirdparty/glm/glm/mat4x3.hpp    |    8 +
 .../manifold/thirdparty/glm/glm/mat4x4.hpp    |    9 +
 .../manifold/thirdparty/glm/glm/matrix.hpp    |  161 +
 .../manifold/thirdparty/glm/glm/packing.hpp   |  173 +
 .../manifold/thirdparty/glm/glm/simd/common.h |  240 +
 .../thirdparty/glm/glm/simd/exponential.h     |   20 +
 .../thirdparty/glm/glm/simd/geometric.h       |  124 +
 .../thirdparty/glm/glm/simd/integer.h         |  115 +
 .../manifold/thirdparty/glm/glm/simd/matrix.h | 1028 +++
 .../manifold/thirdparty/glm/glm/simd/neon.h   |  155 +
 .../thirdparty/glm/glm/simd/packing.h         |    8 +
 .../thirdparty/glm/glm/simd/platform.h        |  469 ++
 .../thirdparty/glm/glm/simd/trigonometric.h   |    9 +
 .../glm/glm/simd/vector_relational.h          |    8 +
 .../thirdparty/glm/glm/trigonometric.hpp      |  210 +
 .../manifold/thirdparty/glm/glm/vec2.hpp      |   14 +
 .../manifold/thirdparty/glm/glm/vec3.hpp      |   14 +
 .../manifold/thirdparty/glm/glm/vec4.hpp      |   15 +
 .../thirdparty/glm/glm/vector_relational.hpp  |  121 +
 .../manifold/thirdparty/thrust/.gitrepo       |   12 +
 thirdparty/manifold/thirdparty/thrust/LICENSE |  249 +
 .../thrust/dependencies/libcudacxx/.gitrepo   |   12 +
 .../dependencies/libcudacxx/LICENSE.TXT       |  368 +
 .../libcudacxx/include/cuda/annotated_ptr     |  337 +
 .../libcudacxx/include/cuda/atomic            |   10 +
 .../libcudacxx/include/cuda/barrier           |   10 +
 .../libcudacxx/include/cuda/latch             |   10 +
 .../libcudacxx/include/cuda/pipeline          |  635 ++
 .../libcudacxx/include/cuda/semaphore         |   10 +
 .../libcudacxx/include/cuda/std/array         |   34 +
 .../libcudacxx/include/cuda/std/atomic        |  296 +
 .../libcudacxx/include/cuda/std/barrier       |  468 ++
 .../libcudacxx/include/cuda/std/bit           |   24 +
 .../libcudacxx/include/cuda/std/cassert       |   28 +
 .../libcudacxx/include/cuda/std/ccomplex      |   14 +
 .../libcudacxx/include/cuda/std/cfloat        |   26 +
 .../libcudacxx/include/cuda/std/chrono        |   83 +
 .../libcudacxx/include/cuda/std/climits       |  106 +
 .../libcudacxx/include/cuda/std/cmath         |   27 +
 .../libcudacxx/include/cuda/std/complex       |   27 +
 .../libcudacxx/include/cuda/std/cstddef       |   40 +
 .../libcudacxx/include/cuda/std/cstdint       |   90 +
 .../libcudacxx/include/cuda/std/ctime         |   27 +
 .../include/cuda/std/detail/__access_property |  323 +
 .../include/cuda/std/detail/__annotated_ptr   |  227 +
 .../include/cuda/std/detail/__config          |  196 +
 .../include/cuda/std/detail/__functional_base |   25 +
 .../include/cuda/std/detail/__pragma_pop      |   11 +
 .../include/cuda/std/detail/__pragma_push     |   12 +
 .../cuda/std/detail/__threading_support       |   26 +
 .../std/detail/libcxx/include/CMakeLists.txt  |  277 +
 .../std/detail/libcxx/include/__bit_reference | 1289 ++++
 .../libcxx/include/__bsd_locale_defaults.h    |   36 +
 .../libcxx/include/__bsd_locale_fallbacks.h   |  139 +
 .../cuda/std/detail/libcxx/include/__config   | 1943 +++++
 .../detail/libcxx/include/__config_site.in    |   36 +
 .../cuda/std/detail/libcxx/include/__debug    |  284 +
 .../cuda/std/detail/libcxx/include/__errc     |  217 +
 .../std/detail/libcxx/include/__functional_03 | 1595 ++++
 .../detail/libcxx/include/__functional_base   |  673 ++
 .../libcxx/include/__functional_base_03       |  223 +
 .../std/detail/libcxx/include/__hash_table    | 2914 +++++++
 .../detail/libcxx/include/__libcpp_version    |    1 +
 .../cuda/std/detail/libcxx/include/__locale   | 1553 ++++
 .../std/detail/libcxx/include/__mutex_base    |  541 ++
 .../std/detail/libcxx/include/__node_handle   |  208 +
 .../cuda/std/detail/libcxx/include/__nullptr  |   61 +
 .../std/detail/libcxx/include/__pragma_pop    |   16 +
 .../std/detail/libcxx/include/__pragma_push   |   25 +
 .../std/detail/libcxx/include/__split_buffer  |  644 ++
 .../std/detail/libcxx/include/__sso_allocator |   76 +
 .../std/detail/libcxx/include/__std_stream    |  361 +
 .../cuda/std/detail/libcxx/include/__string   |  985 +++
 .../detail/libcxx/include/__threading_support |  787 ++
 .../cuda/std/detail/libcxx/include/__tree     | 2844 +++++++
 .../cuda/std/detail/libcxx/include/__tuple    |  567 ++
 .../std/detail/libcxx/include/__undef_macros  |   33 +
 .../cuda/std/detail/libcxx/include/algorithm  | 5769 ++++++++++++++
 .../cuda/std/detail/libcxx/include/any        |  671 ++
 .../cuda/std/detail/libcxx/include/array      |  486 ++
 .../cuda/std/detail/libcxx/include/atomic     | 2763 +++++++
 .../cuda/std/detail/libcxx/include/barrier    |  384 +
 .../cuda/std/detail/libcxx/include/bit        |  895 +++
 .../cuda/std/detail/libcxx/include/bitset     | 1109 +++
 .../cuda/std/detail/libcxx/include/cassert    |   25 +
 .../cuda/std/detail/libcxx/include/ccomplex   |   28 +
 .../cuda/std/detail/libcxx/include/cctype     |  120 +
 .../cuda/std/detail/libcxx/include/cerrno     |   32 +
 .../cuda/std/detail/libcxx/include/cfenv      |   81 +
 .../cuda/std/detail/libcxx/include/cfloat     |   79 +
 .../cuda/std/detail/libcxx/include/charconv   |  616 ++
 .../cuda/std/detail/libcxx/include/chrono     | 3348 ++++++++
 .../cuda/std/detail/libcxx/include/cinttypes  |  257 +
 .../cuda/std/detail/libcxx/include/ciso646    |   24 +
 .../cuda/std/detail/libcxx/include/climits    |   62 +
 .../cuda/std/detail/libcxx/include/clocale    |   54 +
 .../cuda/std/detail/libcxx/include/cmath      |  725 ++
 .../cuda/std/detail/libcxx/include/codecvt    |  549 ++
 .../cuda/std/detail/libcxx/include/compare    |  678 ++
 .../cuda/std/detail/libcxx/include/complex    | 1590 ++++
 .../cuda/std/detail/libcxx/include/complex.h  |   36 +
 .../detail/libcxx/include/condition_variable  |  268 +
 .../cuda/std/detail/libcxx/include/csetjmp    |   47 +
 .../cuda/std/detail/libcxx/include/csignal    |   57 +
 .../cuda/std/detail/libcxx/include/cstdarg    |   54 +
 .../cuda/std/detail/libcxx/include/cstdbool   |   38 +
 .../cuda/std/detail/libcxx/include/cstddef    |  143 +
 .../cuda/std/detail/libcxx/include/cstdint    |  197 +
 .../cuda/std/detail/libcxx/include/cstdio     |  171 +
 .../cuda/std/detail/libcxx/include/cstdlib    |  182 +
 .../cuda/std/detail/libcxx/include/cstring    |  103 +
 .../cuda/std/detail/libcxx/include/ctgmath    |   28 +
 .../cuda/std/detail/libcxx/include/ctime      |   94 +
 .../cuda/std/detail/libcxx/include/ctype.h    |   59 +
 .../cuda/std/detail/libcxx/include/cwchar     |  192 +
 .../cuda/std/detail/libcxx/include/cwctype    |   86 +
 .../cuda/std/detail/libcxx/include/deque      | 3039 ++++++++
 .../cuda/std/detail/libcxx/include/errno.h    |  397 +
 .../cuda/std/detail/libcxx/include/exception  |  329 +
 .../cuda/std/detail/libcxx/include/execution  |   19 +
 .../libcxx/include/experimental/__config      |   79 +
 .../libcxx/include/experimental/__memory      |   89 +
 .../libcxx/include/experimental/algorithm     |   59 +
 .../libcxx/include/experimental/coroutine     |  334 +
 .../detail/libcxx/include/experimental/deque  |   46 +
 .../libcxx/include/experimental/filesystem    |  256 +
 .../libcxx/include/experimental/forward_list  |   46 +
 .../libcxx/include/experimental/functional    |  458 ++
 .../libcxx/include/experimental/iterator      |  113 +
 .../detail/libcxx/include/experimental/list   |   46 +
 .../detail/libcxx/include/experimental/map    |   56 +
 .../include/experimental/memory_resource      |  426 ++
 .../include/experimental/propagate_const      |  578 ++
 .../detail/libcxx/include/experimental/regex  |   61 +
 .../detail/libcxx/include/experimental/set    |   56 +
 .../detail/libcxx/include/experimental/simd   | 1569 ++++
 .../detail/libcxx/include/experimental/string |   61 +
 .../libcxx/include/experimental/type_traits   |  154 +
 .../libcxx/include/experimental/unordered_map |   64 +
 .../libcxx/include/experimental/unordered_set |   58 +
 .../libcxx/include/experimental/utility       |   46 +
 .../detail/libcxx/include/experimental/vector |   46 +
 .../cuda/std/detail/libcxx/include/ext/__hash |  133 +
 .../std/detail/libcxx/include/ext/hash_map    |  984 +++
 .../std/detail/libcxx/include/ext/hash_set    |  659 ++
 .../cuda/std/detail/libcxx/include/fenv.h     |  116 +
 .../cuda/std/detail/libcxx/include/filesystem | 2644 +++++++
 .../cuda/std/detail/libcxx/include/float.h    |   97 +
 .../std/detail/libcxx/include/forward_list    | 1781 +++++
 .../cuda/std/detail/libcxx/include/fstream    | 1763 +++++
 .../cuda/std/detail/libcxx/include/functional | 3126 ++++++++
 .../cuda/std/detail/libcxx/include/future     | 2608 +++++++
 .../detail/libcxx/include/initializer_list    |  117 +
 .../cuda/std/detail/libcxx/include/inttypes.h |  262 +
 .../cuda/std/detail/libcxx/include/iomanip    |  670 ++
 .../cuda/std/detail/libcxx/include/ios        | 1066 +++
 .../cuda/std/detail/libcxx/include/iosfwd     |  229 +
 .../cuda/std/detail/libcxx/include/iostream   |   63 +
 .../cuda/std/detail/libcxx/include/istream    | 1651 ++++
 .../cuda/std/detail/libcxx/include/iterator   | 1943 +++++
 .../cuda/std/detail/libcxx/include/latch      |  129 +
 .../cuda/std/detail/libcxx/include/limits     |  837 ++
 .../cuda/std/detail/libcxx/include/limits.h   |   68 +
 .../cuda/std/detail/libcxx/include/list       | 2488 ++++++
 .../cuda/std/detail/libcxx/include/locale     | 4353 +++++++++++
 .../cuda/std/detail/libcxx/include/locale.h   |   44 +
 .../cuda/std/detail/libcxx/include/map        | 2246 ++++++
 .../cuda/std/detail/libcxx/include/math.h     | 1574 ++++
 .../cuda/std/detail/libcxx/include/memory     | 5443 +++++++++++++
 .../detail/libcxx/include/module.modulemap    |  612 ++
 .../cuda/std/detail/libcxx/include/mutex      |  711 ++
 .../cuda/std/detail/libcxx/include/new        |  378 +
 .../cuda/std/detail/libcxx/include/numeric    |  593 ++
 .../cuda/std/detail/libcxx/include/optional   | 1420 ++++
 .../cuda/std/detail/libcxx/include/ostream    | 1107 +++
 .../cuda/std/detail/libcxx/include/queue      |  803 ++
 .../cuda/std/detail/libcxx/include/random     | 6745 +++++++++++++++++
 .../cuda/std/detail/libcxx/include/ratio      |  533 ++
 .../cuda/std/detail/libcxx/include/regex      | 6662 ++++++++++++++++
 .../detail/libcxx/include/scoped_allocator    |  683 ++
 .../cuda/std/detail/libcxx/include/semaphore  |  447 ++
 .../cuda/std/detail/libcxx/include/set        | 1493 ++++
 .../cuda/std/detail/libcxx/include/setjmp.h   |   44 +
 .../std/detail/libcxx/include/shared_mutex    |  508 ++
 .../cuda/std/detail/libcxx/include/span       |  591 ++
 .../cuda/std/detail/libcxx/include/sstream    |  985 +++
 .../cuda/std/detail/libcxx/include/stack      |  321 +
 .../cuda/std/detail/libcxx/include/stdbool.h  |   38 +
 .../cuda/std/detail/libcxx/include/stddef.h   |   66 +
 .../cuda/std/detail/libcxx/include/stdexcept  |  321 +
 .../cuda/std/detail/libcxx/include/stdint.h   |  129 +
 .../cuda/std/detail/libcxx/include/stdio.h    |  119 +
 .../cuda/std/detail/libcxx/include/stdlib.h   |  103 +
 .../cuda/std/detail/libcxx/include/streambuf  |  500 ++
 .../cuda/std/detail/libcxx/include/string     | 4365 +++++++++++
 .../cuda/std/detail/libcxx/include/string.h   |  109 +
 .../std/detail/libcxx/include/string_view     |  840 ++
 .../cuda/std/detail/libcxx/include/strstream  |  399 +
 .../include/support/android/locale_bionic.h   |   67 +
 .../include/support/atomic/atomic_base.h      |  192 +
 .../include/support/atomic/atomic_c11.h       |  183 +
 .../include/support/atomic/atomic_cuda.h      |  482 ++
 .../support/atomic/atomic_cuda_derived.h      |  143 +
 .../support/atomic/atomic_cuda_generated.h    | 2644 +++++++
 .../include/support/atomic/atomic_gcc.h       |   16 +
 .../include/support/atomic/atomic_msvc.h      |  449 ++
 .../include/support/atomic/atomic_nvrtc.h     |   16 +
 .../include/support/atomic/atomic_scopes.h    |   46 +
 .../include/support/atomic/cxx_atomic.h       |  160 +
 .../libcxx/include/support/fuchsia/xlocale.h  |   22 +
 .../libcxx/include/support/ibm/limits.h       |   98 +
 .../include/support/ibm/locale_mgmt_aix.h     |   84 +
 .../libcxx/include/support/ibm/support.h      |   53 +
 .../libcxx/include/support/ibm/xlocale.h      |  270 +
 .../libcxx/include/support/musl/xlocale.h     |   57 +
 .../libcxx/include/support/newlib/xlocale.h   |   27 +
 .../include/support/solaris/floatingpoint.h   |   13 +
 .../libcxx/include/support/solaris/wchar.h    |   46 +
 .../libcxx/include/support/solaris/xlocale.h  |   76 +
 .../support/xlocale/__nop_locale_mgmt.h       |   51 +
 .../support/xlocale/__posix_l_fallback.h      |  164 +
 .../support/xlocale/__strtonum_fallback.h     |   66 +
 .../std/detail/libcxx/include/system_error    |  486 ++
 .../cuda/std/detail/libcxx/include/tgmath.h   |   36 +
 .../cuda/std/detail/libcxx/include/thread     |  420 +
 .../cuda/std/detail/libcxx/include/tuple      | 1547 ++++
 .../std/detail/libcxx/include/type_traits     | 4898 ++++++++++++
 .../cuda/std/detail/libcxx/include/typeindex  |  102 +
 .../cuda/std/detail/libcxx/include/typeinfo   |  350 +
 .../std/detail/libcxx/include/unordered_map   | 2445 ++++++
 .../std/detail/libcxx/include/unordered_set   | 1680 ++++
 .../cuda/std/detail/libcxx/include/utility    | 1650 ++++
 .../cuda/std/detail/libcxx/include/valarray   | 4930 ++++++++++++
 .../cuda/std/detail/libcxx/include/variant    | 1668 ++++
 .../cuda/std/detail/libcxx/include/vector     | 3406 +++++++++
 .../cuda/std/detail/libcxx/include/version    |  242 +
 .../cuda/std/detail/libcxx/include/wchar.h    |  181 +
 .../cuda/std/detail/libcxx/include/wctype.h   |   78 +
 .../libcudacxx/include/cuda/std/functional    |   32 +
 .../include/cuda/std/initializer_list         |   25 +
 .../libcudacxx/include/cuda/std/iterator      |   29 +
 .../libcudacxx/include/cuda/std/latch         |   40 +
 .../libcudacxx/include/cuda/std/limits        |   29 +
 .../libcudacxx/include/cuda/std/ratio         |   27 +
 .../libcudacxx/include/cuda/std/semaphore     |   47 +
 .../libcudacxx/include/cuda/std/tuple         |   33 +
 .../libcudacxx/include/cuda/std/type_traits   |   28 +
 .../libcudacxx/include/cuda/std/utility       |   27 +
 .../libcudacxx/include/cuda/std/version       |   21 +
 .../include/nv/detail/__preprocessor          |  117 +
 .../include/nv/detail/__target_macros         |  493 ++
 .../dependencies/libcudacxx/include/nv/target |  207 +
 .../thirdparty/thrust/thrust/addressof.h      |   31 +
 .../thrust/thrust/adjacent_difference.h       |  244 +
 .../thirdparty/thrust/thrust/advance.h        |  140 +
 .../thrust/thrust/allocate_unique.h           |  443 ++
 .../thirdparty/thrust/thrust/async/copy.h     |  154 +
 .../thirdparty/thrust/thrust/async/for_each.h |  123 +
 .../thirdparty/thrust/thrust/async/reduce.h   |  446 ++
 .../thirdparty/thrust/thrust/async/scan.h     |  344 +
 .../thirdparty/thrust/thrust/async/sort.h     |  280 +
 .../thrust/thrust/async/transform.h           |  138 +
 .../thirdparty/thrust/thrust/binary_search.h  | 1899 +++++
 .../thirdparty/thrust/thrust/complex.h        | 1047 +++
 .../manifold/thirdparty/thrust/thrust/copy.h  |  512 ++
 .../manifold/thirdparty/thrust/thrust/count.h |  231 +
 .../thrust/detail/adjacent_difference.inl     |   86 +
 .../thrust/thrust/detail/advance.inl          |   74 +
 .../thrust/thrust/detail/algorithm_wrapper.h  |   27 +
 .../thrust/thrust/detail/alignment.h          |  230 +
 .../detail/allocator/allocator_traits.h       |  442 ++
 .../detail/allocator/allocator_traits.inl     |  465 ++
 .../detail/allocator/copy_construct_range.h   |   46 +
 .../detail/allocator/copy_construct_range.inl |  310 +
 .../allocator/default_construct_range.h       |   36 +
 .../allocator/default_construct_range.inl     |  112 +
 .../thrust/detail/allocator/destroy_range.h   |   33 +
 .../thrust/detail/allocator/destroy_range.inl |  167 +
 .../detail/allocator/fill_construct_range.h   |   35 +
 .../detail/allocator/fill_construct_range.inl |  114 +
 .../detail/allocator/malloc_allocator.h       |   51 +
 .../detail/allocator/malloc_allocator.inl     |   65 +
 .../detail/allocator/no_throw_allocator.h     |   72 +
 .../detail/allocator/tagged_allocator.h       |  100 +
 .../detail/allocator/tagged_allocator.inl     |  104 +
 .../detail/allocator/temporary_allocator.h    |   84 +
 .../detail/allocator/temporary_allocator.inl  |   79 +
 .../detail/allocator_aware_execution_policy.h |  101 +
 .../thrust/thrust/detail/binary_search.inl    |  480 ++
 .../thrust/thrust/detail/caching_allocator.h  |   47 +
 .../thrust/thrust/detail/complex/arithmetic.h |  303 +
 .../thrust/thrust/detail/complex/c99math.h    |  197 +
 .../thrust/thrust/detail/complex/catrig.h     |  785 ++
 .../thrust/thrust/detail/complex/catrigf.h    |  500 ++
 .../thrust/thrust/detail/complex/ccosh.h      |  215 +
 .../thrust/thrust/detail/complex/ccoshf.h     |  143 +
 .../thrust/thrust/detail/complex/cexp.h       |  185 +
 .../thrust/thrust/detail/complex/cexpf.h      |  163 +
 .../thrust/thrust/detail/complex/clog.h       |  214 +
 .../thrust/thrust/detail/complex/clogf.h      |  200 +
 .../thrust/thrust/detail/complex/complex.inl  |  355 +
 .../thrust/thrust/detail/complex/cpow.h       |   57 +
 .../thrust/thrust/detail/complex/cproj.h      |   72 +
 .../thrust/thrust/detail/complex/csinh.h      |  207 +
 .../thrust/thrust/detail/complex/csinhf.h     |  144 +
 .../thrust/thrust/detail/complex/csqrt.h      |  154 +
 .../thrust/thrust/detail/complex/csqrtf.h     |  149 +
 .../thrust/thrust/detail/complex/ctanh.h      |  202 +
 .../thrust/thrust/detail/complex/ctanhf.h     |  126 +
 .../thrust/detail/complex/math_private.h      |  136 +
 .../thrust/thrust/detail/complex/stream.h     |   74 +
 .../thirdparty/thrust/thrust/detail/config.h  |   24 +
 .../thrust/thrust/detail/config/compiler.h    |  189 +
 .../thrust/detail/config/compiler_fence.h     |   62 +
 .../thrust/thrust/detail/config/config.h      |   40 +
 .../thrust/detail/config/cpp_compatibility.h  |  101 +
 .../thrust/thrust/detail/config/cpp_dialect.h |  140 +
 .../thrust/thrust/detail/config/debug.h       |   32 +
 .../thrust/thrust/detail/config/deprecated.h  |   42 +
 .../thrust/detail/config/device_system.h      |   44 +
 .../thrust/detail/config/exec_check_disable.h |   43 +
 .../thrust/thrust/detail/config/forceinline.h |   36 +
 .../thrust/detail/config/global_workarounds.h |   27 +
 .../thrust/thrust/detail/config/host_device.h |   44 +
 .../thrust/thrust/detail/config/host_system.h |   41 +
 .../thrust/detail/config/memory_resource.h    |   35 +
 .../thrust/thrust/detail/config/namespace.h   |  120 +
 .../thrust/detail/config/simple_defines.h     |   30 +
 .../thrust/thrust/detail/contiguous_storage.h |  235 +
 .../thrust/detail/contiguous_storage.inl      |  550 ++
 .../thirdparty/thrust/thrust/detail/copy.h    |   90 +
 .../thirdparty/thrust/thrust/detail/copy.inl  |  129 +
 .../thirdparty/thrust/thrust/detail/copy_if.h |   71 +
 .../thrust/thrust/detail/copy_if.inl          |  107 +
 .../thirdparty/thrust/thrust/detail/count.h   |   60 +
 .../thirdparty/thrust/thrust/detail/count.inl |   77 +
 .../thrust/thrust/detail/cpp11_required.h     |   26 +
 .../thrust/thrust/detail/cpp14_required.h     |   26 +
 .../thirdparty/thrust/thrust/detail/cstdint.h |   83 +
 .../dependencies_aware_execution_policy.h     |  106 +
 .../thrust/thrust/detail/device_delete.inl    |   44 +
 .../thrust/thrust/detail/device_free.inl      |   39 +
 .../thrust/thrust/detail/device_malloc.inl    |   53 +
 .../thrust/thrust/detail/device_new.inl       |   56 +
 .../thrust/thrust/detail/device_ptr.inl       |   64 +
 .../thrust/thrust/detail/distance.inl         |   35 +
 .../thirdparty/thrust/thrust/detail/equal.inl |   80 +
 .../thrust/thrust/detail/event_error.h        |  164 +
 .../thrust/detail/execute_with_allocator.h    |  149 +
 .../detail/execute_with_allocator_fwd.h       |  106 +
 .../thrust/detail/execute_with_dependencies.h |  267 +
 .../thrust/thrust/detail/execution_policy.h   |   76 +
 .../thrust/thrust/detail/extrema.inl          |  169 +
 .../thirdparty/thrust/thrust/detail/fill.inl  |   86 +
 .../thirdparty/thrust/thrust/detail/find.inl  |  108 +
 .../thrust/thrust/detail/for_each.inl         |   86 +
 .../thrust/thrust/detail/function.h           |  161 +
 .../thrust/thrust/detail/functional.inl       |  126 +
 .../thrust/thrust/detail/functional/actor.h   |  155 +
 .../thrust/thrust/detail/functional/actor.inl |  114 +
 .../thrust/detail/functional/argument.h       |   74 +
 .../thrust/detail/functional/composite.h      |  164 +
 .../thrust/detail/functional/operators.h      |   25 +
 .../operators/arithmetic_operators.h          |  436 ++
 .../operators/assignment_operator.h           |   79 +
 .../functional/operators/bitwise_operators.h  |  338 +
 .../operators/compound_assignment_operators.h |  512 ++
 .../functional/operators/logical_operators.h  |  143 +
 .../functional/operators/operator_adaptors.h  |  136 +
 .../operators/relational_operators.h          |  322 +
 .../thrust/detail/functional/placeholder.h    |   38 +
 .../thrust/thrust/detail/functional/value.h   |   79 +
 .../thrust/thrust/detail/gather.inl           |  161 +
 .../thrust/thrust/detail/generate.inl         |   92 +
 .../thrust/thrust/detail/get_iterator_value.h |   55 +
 .../thrust/thrust/detail/inner_product.inl    |  101 +
 .../thrust/thrust/detail/integer_math.h       |  152 +
 .../thrust/thrust/detail/integer_traits.h     |  130 +
 .../thrust/detail/internal_functional.h       |  558 ++
 .../thrust/thrust/detail/logical.inl          |   95 +
 .../thrust/thrust/detail/malloc_and_free.h    |   83 +
 .../thrust/thrust/detail/memory_algorithms.h  |  237 +
 .../thrust/thrust/detail/memory_wrapper.h     |   30 +
 .../thirdparty/thrust/thrust/detail/merge.inl |  221 +
 .../thirdparty/thrust/thrust/detail/minmax.h  |   51 +
 .../thrust/thrust/detail/mismatch.inl         |   89 +
 .../thrust/detail/modern_gcc_required.h       |   26 +
 .../thrust/thrust/detail/mpl/math.h           |  175 +
 .../thrust/thrust/detail/numeric_traits.h     |  129 +
 .../thrust/thrust/detail/numeric_wrapper.h    |   27 +
 .../thrust/thrust/detail/overlapped_copy.h    |  131 +
 .../thirdparty/thrust/thrust/detail/pair.inl  |  231 +
 .../thrust/thrust/detail/partition.inl        |  411 +
 .../thirdparty/thrust/thrust/detail/pointer.h |  255 +
 .../thrust/thrust/detail/pointer.inl          |  209 +
 .../thrust/thrust/detail/preprocessor.h       | 1182 +++
 .../thrust/thrust/detail/range/head_flags.h   |  229 +
 .../thrust/thrust/detail/range/tail_flags.h   |  133 +
 .../thrust/thrust/detail/raw_pointer_cast.h   |   50 +
 .../thrust/thrust/detail/raw_reference_cast.h |  332 +
 .../thrust/thrust/detail/reduce.inl           |  276 +
 .../thrust/thrust/detail/reference.h          |  518 ++
 .../detail/reference_forward_declaration.h    |   28 +
 .../thrust/thrust/detail/remove.inl           |  246 +
 .../thrust/thrust/detail/replace.inl          |  218 +
 .../thrust/thrust/detail/reverse.inl          |   87 +
 .../thirdparty/thrust/thrust/detail/scan.inl  |  522 ++
 .../thrust/thrust/detail/scatter.inl          |  163 +
 .../thrust/thrust/detail/select_system.h      |   84 +
 .../thirdparty/thrust/thrust/detail/seq.h     |   52 +
 .../thrust/thrust/detail/sequence.inl         |  114 +
 .../thrust/thrust/detail/set_operations.inl   |  865 +++
 .../thrust/thrust/detail/shuffle.inl          |   83 +
 .../thirdparty/thrust/thrust/detail/sort.inl  |  404 +
 .../thrust/thrust/detail/static_assert.h      |   91 +
 .../thrust/thrust/detail/static_map.h         |  169 +
 .../thirdparty/thrust/thrust/detail/swap.h    |   35 +
 .../thirdparty/thrust/thrust/detail/swap.inl  |   22 +
 .../thrust/thrust/detail/swap_ranges.inl      |   64 +
 .../thrust/thrust/detail/tabulate.inl         |   58 +
 .../thrust/thrust/detail/temporary_array.h    |  181 +
 .../thrust/thrust/detail/temporary_array.inl  |  168 +
 .../thrust/thrust/detail/temporary_buffer.h   |   75 +
 .../thrust/thrust/detail/transform.inl        |  247 +
 .../thrust/thrust/detail/transform_reduce.inl |   68 +
 .../thrust/thrust/detail/transform_scan.inl   |  117 +
 .../thrust/thrust/detail/trivial_sequence.h   |   96 +
 .../thirdparty/thrust/thrust/detail/tuple.inl | 1004 +++
 .../thrust/thrust/detail/tuple_algorithms.h   |  110 +
 .../thrust/detail/tuple_meta_transform.h      |   58 +
 .../thrust/thrust/detail/tuple_transform.h    |   84 +
 .../thrust/thrust/detail/type_deduction.h     |   90 +
 .../thrust/thrust/detail/type_traits.h        |  717 ++
 .../detail/type_traits/function_traits.h      |   97 +
 .../detail/type_traits/has_member_function.h  |   38 +
 .../detail/type_traits/has_nested_type.h      |   32 +
 .../detail/type_traits/has_trivial_assign.h   |   47 +
 .../detail/type_traits/is_call_possible.h     |  162 +
 .../type_traits/is_metafunction_defined.h     |   42 +
 .../iterator/is_discard_iterator.h            |   39 +
 .../type_traits/iterator/is_output_iterator.h |   65 +
 .../thrust/detail/type_traits/minimum_type.h  |  163 +
 .../detail/type_traits/pointer_traits.h       |  395 +
 .../result_of_adaptable_function.h            |   59 +
 .../thrust/detail/uninitialized_copy.inl      |   95 +
 .../thrust/detail/uninitialized_fill.inl      |   90 +
 .../thrust/thrust/detail/unique.inl           |  393 +
 .../thrust/thrust/detail/use_default.h        |   26 +
 .../thrust/thrust/detail/util/align.h         |   60 +
 .../thrust/thrust/detail/vector_base.h        |  603 ++
 .../thrust/thrust/detail/vector_base.inl      | 1345 ++++
 .../thrust/thrust/device_allocator.h          |  140 +
 .../thirdparty/thrust/thrust/device_delete.h  |   53 +
 .../thirdparty/thrust/thrust/device_free.h    |   65 +
 .../thrust/thrust/device_make_unique.h        |   60 +
 .../thirdparty/thrust/thrust/device_malloc.h  |  100 +
 .../thrust/thrust/device_malloc_allocator.h   |  180 +
 .../thirdparty/thrust/thrust/device_new.h     |   86 +
 .../thrust/thrust/device_new_allocator.h      |  172 +
 .../thirdparty/thrust/thrust/device_ptr.h     |  211 +
 .../thrust/thrust/device_reference.h          |  987 +++
 .../thirdparty/thrust/thrust/device_vector.h  |  511 ++
 .../thirdparty/thrust/thrust/distance.h       |   74 +
 .../manifold/thirdparty/thrust/thrust/equal.h |  235 +
 .../manifold/thirdparty/thrust/thrust/event.h |   26 +
 .../thrust/thrust/execution_policy.h          |  392 +
 .../thirdparty/thrust/thrust/extrema.h        |  801 ++
 .../manifold/thirdparty/thrust/thrust/fill.h  |  206 +
 .../manifold/thirdparty/thrust/thrust/find.h  |  381 +
 .../thirdparty/thrust/thrust/for_each.h       |  278 +
 .../thirdparty/thrust/thrust/functional.h     | 1719 +++++
 .../thirdparty/thrust/thrust/future.h         |  176 +
 .../thirdparty/thrust/thrust/gather.h         |  445 ++
 .../thirdparty/thrust/thrust/generate.h       |  211 +
 .../thirdparty/thrust/thrust/host_vector.h    |  534 ++
 .../thirdparty/thrust/thrust/inner_product.h  |  262 +
 .../thrust/iterator/constant_iterator.h       |  250 +
 .../thrust/iterator/counting_iterator.h       |  246 +
 .../thrust/iterator/detail/any_assign.h       |   54 +
 .../thrust/iterator/detail/any_system_tag.h   |   33 +
 .../iterator/detail/constant_iterator_base.h  |   71 +
 .../iterator/detail/counting_iterator.inl     |  142 +
 .../iterator/detail/device_system_tag.h       |   30 +
 .../iterator/detail/discard_iterator_base.h   |   64 +
 .../iterator/detail/distance_from_result.h    |   41 +
 .../thrust/iterator/detail/host_system_tag.h  |   30 +
 .../iterator/detail/is_iterator_category.h    |   59 +
 .../iterator/detail/iterator_adaptor_base.h   |  112 +
 .../detail/iterator_category_to_system.h      |   79 +
 .../detail/iterator_category_to_traversal.h   |  130 +
 ...rator_category_with_system_and_traversal.h |   56 +
 .../detail/iterator_facade_category.h         |  252 +
 .../iterator/detail/iterator_traits.inl       |  136 +
 .../iterator/detail/iterator_traversal_tags.h |   42 +
 .../thrust/iterator/detail/join_iterator.h    |  133 +
 .../thrust/iterator/detail/minimum_category.h |   53 +
 .../thrust/iterator/detail/minimum_system.h   |   81 +
 .../thrust/iterator/detail/normal_iterator.h  |   79 +
 .../detail/permutation_iterator_base.h        |   54 +
 .../thrust/thrust/iterator/detail/retag.h     |  147 +
 .../iterator/detail/reverse_iterator.inl      |  118 +
 .../iterator/detail/reverse_iterator_base.h   |   43 +
 .../thrust/iterator/detail/tagged_iterator.h  |   85 +
 .../transform_input_output_iterator.inl       |  107 +
 .../iterator/detail/transform_iterator.inl    |   75 +
 .../detail/transform_output_iterator.inl      |   81 +
 .../detail/tuple_of_iterator_references.h     |  146 +
 .../iterator/detail/universal_categories.h    |   86 +
 .../thrust/iterator/detail/zip_iterator.inl   |  152 +
 .../iterator/detail/zip_iterator_base.h       |  353 +
 .../thrust/thrust/iterator/discard_iterator.h |  174 +
 .../thrust/thrust/iterator/iterator_adaptor.h |  242 +
 .../thrust/iterator/iterator_categories.h     |  223 +
 .../thrust/thrust/iterator/iterator_facade.h  |  542 ++
 .../thrust/thrust/iterator/iterator_traits.h  |   65 +
 .../thrust/iterator/permutation_iterator.h    |  216 +
 .../thirdparty/thrust/thrust/iterator/retag.h |   69 +
 .../thrust/thrust/iterator/reverse_iterator.h |  237 +
 .../transform_input_output_iterator.h         |  164 +
 .../thrust/iterator/transform_iterator.h      |  355 +
 .../iterator/transform_output_iterator.h      |  164 +
 .../thrust/thrust/iterator/zip_iterator.h     |  258 +
 .../thirdparty/thrust/thrust/limits.h         |   18 +
 .../thirdparty/thrust/thrust/logical.h        |  275 +
 .../thirdparty/thrust/thrust/memory.h         |  396 +
 .../manifold/thirdparty/thrust/thrust/merge.h |  677 ++
 .../thirdparty/thrust/thrust/mismatch.h       |  257 +
 .../thirdparty/thrust/thrust/mr/allocator.h   |  253 +
 .../thrust/thrust/mr/device_memory_resource.h |   38 +
 .../thrust/thrust/mr/disjoint_pool.h          |  489 ++
 .../thrust/thrust/mr/disjoint_sync_pool.h     |  117 +
 .../thrust/thrust/mr/disjoint_tls_pool.h      |   69 +
 .../thrust/thrust/mr/fancy_pointer_resource.h |   61 +
 .../thrust/thrust/mr/host_memory_resource.h   |   32 +
 .../thrust/thrust/mr/memory_resource.h        |  217 +
 .../thirdparty/thrust/thrust/mr/new.h         |   89 +
 .../thrust/thrust/mr/polymorphic_adaptor.h    |   57 +
 .../thirdparty/thrust/thrust/mr/pool.h        |  507 ++
 .../thrust/thrust/mr/pool_options.h           |  128 +
 .../thirdparty/thrust/thrust/mr/sync_pool.h   |  114 +
 .../thirdparty/thrust/thrust/mr/tls_pool.h    |   64 +
 .../thrust/mr/universal_memory_resource.h     |   22 +
 .../thirdparty/thrust/thrust/mr/validator.h   |   51 +
 .../thirdparty/thrust/thrust/optional.h       | 2876 +++++++
 .../manifold/thirdparty/thrust/thrust/pair.h  |  281 +
 .../thirdparty/thrust/thrust/partition.h      | 1436 ++++
 .../thrust/thrust/per_device_resource.h       |  102 +
 .../thirdparty/thrust/thrust/random.h         |  117 +
 .../thirdparty/thrust/thrust/reduce.h         |  781 ++
 .../thirdparty/thrust/thrust/remove.h         |  802 ++
 .../thirdparty/thrust/thrust/replace.h        |  819 ++
 .../thirdparty/thrust/thrust/reverse.h        |  211 +
 .../manifold/thirdparty/thrust/thrust/scan.h  | 1656 ++++
 .../thirdparty/thrust/thrust/scatter.h        |  419 +
 .../thirdparty/thrust/thrust/sequence.h       |  293 +
 .../thirdparty/thrust/thrust/set_operations.h | 2959 ++++++++
 .../thirdparty/thrust/thrust/shuffle.h        |  179 +
 .../manifold/thirdparty/thrust/thrust/sort.h  | 1358 ++++
 .../manifold/thirdparty/thrust/thrust/swap.h  |  184 +
 .../system/cpp/detail/adjacent_difference.h   |   23 +
 .../thrust/system/cpp/detail/assign_value.h   |   23 +
 .../thrust/system/cpp/detail/binary_search.h  |   23 +
 .../thrust/thrust/system/cpp/detail/copy.h    |   23 +
 .../thrust/thrust/system/cpp/detail/copy_if.h |   23 +
 .../thrust/thrust/system/cpp/detail/count.h   |   22 +
 .../thrust/thrust/system/cpp/detail/equal.h   |   22 +
 .../system/cpp/detail/execution_policy.h      |   80 +
 .../thrust/thrust/system/cpp/detail/extrema.h |   23 +
 .../thrust/thrust/system/cpp/detail/fill.h    |   22 +
 .../thrust/thrust/system/cpp/detail/find.h    |   23 +
 .../thrust/system/cpp/detail/for_each.h       |   23 +
 .../thrust/thrust/system/cpp/detail/gather.h  |   22 +
 .../thrust/system/cpp/detail/generate.h       |   22 +
 .../thrust/system/cpp/detail/get_value.h      |   23 +
 .../thrust/system/cpp/detail/inner_product.h  |   22 +
 .../thrust/system/cpp/detail/iter_swap.h      |   23 +
 .../thrust/thrust/system/cpp/detail/logical.h |   22 +
 .../system/cpp/detail/malloc_and_free.h       |   23 +
 .../thrust/system/cpp/detail/memory.inl       |   52 +
 .../thrust/thrust/system/cpp/detail/merge.h   |   23 +
 .../thrust/system/cpp/detail/mismatch.h       |   22 +
 .../thrust/thrust/system/cpp/detail/par.h     |   61 +
 .../thrust/system/cpp/detail/partition.h      |   23 +
 .../system/cpp/detail/per_device_resource.h   |   22 +
 .../thrust/thrust/system/cpp/detail/reduce.h  |   23 +
 .../thrust/system/cpp/detail/reduce_by_key.h  |   23 +
 .../thrust/thrust/system/cpp/detail/remove.h  |   23 +
 .../thrust/thrust/system/cpp/detail/replace.h |   22 +
 .../thrust/thrust/system/cpp/detail/reverse.h |   22 +
 .../thrust/thrust/system/cpp/detail/scan.h    |   23 +
 .../thrust/system/cpp/detail/scan_by_key.h    |   23 +
 .../thrust/thrust/system/cpp/detail/scatter.h |   22 +
 .../thrust/system/cpp/detail/sequence.h       |   22 +
 .../thrust/system/cpp/detail/set_operations.h |   23 +
 .../thrust/thrust/system/cpp/detail/sort.h    |   23 +
 .../thrust/system/cpp/detail/swap_ranges.h    |   22 +
 .../thrust/system/cpp/detail/tabulate.h       |   22 +
 .../system/cpp/detail/temporary_buffer.h      |   22 +
 .../thrust/system/cpp/detail/transform.h      |   22 +
 .../system/cpp/detail/transform_reduce.h      |   22 +
 .../thrust/system/cpp/detail/transform_scan.h |   22 +
 .../system/cpp/detail/uninitialized_copy.h    |   22 +
 .../system/cpp/detail/uninitialized_fill.h    |   22 +
 .../thrust/thrust/system/cpp/detail/unique.h  |   23 +
 .../thrust/system/cpp/detail/unique_by_key.h  |   23 +
 .../thrust/system/cpp/detail/vector.inl       |  146 +
 .../thrust/system/cpp/execution_policy.h      |  156 +
 .../thrust/thrust/system/cpp/memory.h         |  100 +
 .../thrust/system/cpp/memory_resource.h       |   71 +
 .../thrust/thrust/system/cpp/pointer.h        |  117 +
 .../thrust/thrust/system/cpp/vector.h         |   82 +
 .../system/detail/adl/adjacent_difference.h   |   44 +
 .../thrust/system/detail/adl/assign_value.h   |   44 +
 .../thrust/system/detail/adl/async/copy.h     |   34 +
 .../thrust/system/detail/adl/async/for_each.h |   34 +
 .../thrust/system/detail/adl/async/reduce.h   |   34 +
 .../thrust/system/detail/adl/async/scan.h     |   34 +
 .../thrust/system/detail/adl/async/sort.h     |   34 +
 .../system/detail/adl/async/transform.h       |   34 +
 .../thrust/system/detail/adl/binary_search.h  |   44 +
 .../thrust/thrust/system/detail/adl/copy.h    |   44 +
 .../thrust/thrust/system/detail/adl/copy_if.h |   44 +
 .../thrust/thrust/system/detail/adl/count.h   |   44 +
 .../thrust/thrust/system/detail/adl/equal.h   |   44 +
 .../thrust/thrust/system/detail/adl/extrema.h |   44 +
 .../thrust/thrust/system/detail/adl/fill.h    |   44 +
 .../thrust/thrust/system/detail/adl/find.h    |   44 +
 .../thrust/system/detail/adl/for_each.h       |   44 +
 .../thrust/thrust/system/detail/adl/gather.h  |   44 +
 .../thrust/system/detail/adl/generate.h       |   44 +
 .../thrust/system/detail/adl/get_value.h      |   44 +
 .../thrust/system/detail/adl/inner_product.h  |   44 +
 .../thrust/system/detail/adl/iter_swap.h      |   44 +
 .../thrust/thrust/system/detail/adl/logical.h |   44 +
 .../system/detail/adl/malloc_and_free.h       |   44 +
 .../thrust/thrust/system/detail/adl/merge.h   |   44 +
 .../thrust/system/detail/adl/mismatch.h       |   44 +
 .../thrust/system/detail/adl/partition.h      |   44 +
 .../system/detail/adl/per_device_resource.h   |   41 +
 .../thrust/thrust/system/detail/adl/reduce.h  |   44 +
 .../thrust/system/detail/adl/reduce_by_key.h  |   44 +
 .../thrust/thrust/system/detail/adl/remove.h  |   44 +
 .../thrust/thrust/system/detail/adl/replace.h |   44 +
 .../thrust/thrust/system/detail/adl/reverse.h |   44 +
 .../thrust/thrust/system/detail/adl/scan.h    |   44 +
 .../thrust/system/detail/adl/scan_by_key.h    |   44 +
 .../thrust/thrust/system/detail/adl/scatter.h |   44 +
 .../thrust/system/detail/adl/sequence.h       |   44 +
 .../thrust/system/detail/adl/set_operations.h |   44 +
 .../thrust/thrust/system/detail/adl/sort.h    |   44 +
 .../thrust/system/detail/adl/swap_ranges.h    |   44 +
 .../thrust/system/detail/adl/tabulate.h       |   44 +
 .../system/detail/adl/temporary_buffer.h      |   44 +
 .../thrust/system/detail/adl/transform.h      |   44 +
 .../system/detail/adl/transform_reduce.h      |   44 +
 .../thrust/system/detail/adl/transform_scan.h |   44 +
 .../system/detail/adl/uninitialized_copy.h    |   44 +
 .../system/detail/adl/uninitialized_fill.h    |   44 +
 .../thrust/thrust/system/detail/adl/unique.h  |   44 +
 .../thrust/system/detail/adl/unique_by_key.h  |   44 +
 .../thrust/thrust/system/detail/bad_alloc.h   |   58 +
 .../thrust/thrust/system/detail/errno.h       |  119 +
 .../thrust/system/detail/error_category.inl   |  237 +
 .../thrust/system/detail/error_code.inl       |  198 +
 .../thrust/system/detail/error_condition.inl  |  134 +
 .../detail/generic/adjacent_difference.h      |   57 +
 .../detail/generic/adjacent_difference.inl    |   82 +
 .../thrust/system/detail/generic/advance.h    |   40 +
 .../thrust/system/detail/generic/advance.inl  |   68 +
 .../system/detail/generic/binary_search.h     |  173 +
 .../system/detail/generic/binary_search.inl   |  396 +
 .../thrust/system/detail/generic/copy.h       |   58 +
 .../thrust/system/detail/generic/copy.inl     |   80 +
 .../thrust/system/detail/generic/copy_if.h    |   63 +
 .../thrust/system/detail/generic/copy_if.inl  |  160 +
 .../thrust/system/detail/generic/count.h      |   50 +
 .../thrust/system/detail/generic/count.inl    |   82 +
 .../thrust/system/detail/generic/distance.h   |   42 +
 .../thrust/system/detail/generic/distance.inl |   80 +
 .../thrust/system/detail/generic/equal.h      |   47 +
 .../thrust/system/detail/generic/equal.inl    |   59 +
 .../thrust/system/detail/generic/extrema.h    |   88 +
 .../thrust/system/detail/generic/extrema.inl  |  262 +
 .../thrust/system/detail/generic/fill.h       |   61 +
 .../thrust/system/detail/generic/find.h       |   62 +
 .../thrust/system/detail/generic/find.inl     |  151 +
 .../thrust/system/detail/generic/for_each.h   |   77 +
 .../thrust/system/detail/generic/gather.h     |   80 +
 .../thrust/system/detail/generic/gather.inl   |  108 +
 .../thrust/system/detail/generic/generate.h   |   56 +
 .../thrust/system/detail/generic/generate.inl |  100 +
 .../system/detail/generic/inner_product.h     |   58 +
 .../system/detail/generic/inner_product.inl   |   73 +
 .../thrust/system/detail/generic/logical.h    |   62 +
 .../thrust/system/detail/generic/memory.h     |   70 +
 .../thrust/system/detail/generic/memory.inl   |  105 +
 .../thrust/system/detail/generic/merge.h      |   90 +
 .../thrust/system/detail/generic/merge.inl    |  130 +
 .../thrust/system/detail/generic/mismatch.h   |   57 +
 .../thrust/system/detail/generic/mismatch.inl |   75 +
 .../thrust/system/detail/generic/partition.h  |  169 +
 .../system/detail/generic/partition.inl       |  249 +
 .../detail/generic/per_device_resource.h      |   46 +
 .../thrust/system/detail/generic/reduce.h     |   58 +
 .../thrust/system/detail/generic/reduce.inl   |   80 +
 .../system/detail/generic/reduce_by_key.h     |   88 +
 .../system/detail/generic/reduce_by_key.inl   |  193 +
 .../thrust/system/detail/generic/remove.h     |  112 +
 .../thrust/system/detail/generic/remove.inl   |  146 +
 .../thrust/system/detail/generic/replace.h    |   97 +
 .../thrust/system/detail/generic/replace.inl  |  179 +
 .../thrust/system/detail/generic/reverse.h    |   55 +
 .../thrust/system/detail/generic/reverse.inl  |   76 +
 .../detail/generic/scalar/binary_search.h     |   84 +
 .../detail/generic/scalar/binary_search.inl   |  157 +
 .../thrust/system/detail/generic/scan.h       |   98 +
 .../thrust/system/detail/generic/scan.inl     |  127 +
 .../system/detail/generic/scan_by_key.h       |  143 +
 .../system/detail/generic/scan_by_key.inl     |  245 +
 .../thrust/system/detail/generic/scatter.h    |   80 +
 .../thrust/system/detail/generic/scatter.inl  |   97 +
 .../system/detail/generic/select_system.h     |  124 +
 .../system/detail/generic/select_system.inl   |  178 +
 .../detail/generic/select_system_exists.h     |  167 +
 .../thrust/system/detail/generic/sequence.h   |   63 +
 .../thrust/system/detail/generic/sequence.inl |  106 +
 .../system/detail/generic/set_operations.h    |  318 +
 .../system/detail/generic/set_operations.inl  |  476 ++
 .../thrust/system/detail/generic/shuffle.h    |   54 +
 .../thrust/system/detail/generic/shuffle.inl  |  190 +
 .../thrust/system/detail/generic/sort.h       |  153 +
 .../thrust/system/detail/generic/sort.inl     |  219 +
 .../system/detail/generic/swap_ranges.h       |   46 +
 .../system/detail/generic/swap_ranges.inl     |   79 +
 .../thrust/system/detail/generic/tabulate.h   |   48 +
 .../thrust/system/detail/generic/tabulate.inl |   61 +
 .../thrust/thrust/system/detail/generic/tag.h |   47 +
 .../system/detail/generic/temporary_buffer.h  |   57 +
 .../detail/generic/temporary_buffer.inl       |   86 +
 .../thrust/system/detail/generic/transform.h  |  105 +
 .../system/detail/generic/transform.inl       |  191 +
 .../system/detail/generic/transform_reduce.h  |   52 +
 .../detail/generic/transform_reduce.inl       |   57 +
 .../system/detail/generic/transform_scan.h    |   67 +
 .../system/detail/generic/transform_scan.inl  |   91 +
 .../detail/generic/uninitialized_copy.h       |   56 +
 .../detail/generic/uninitialized_copy.inl     |  194 +
 .../detail/generic/uninitialized_fill.h       |   56 +
 .../detail/generic/uninitialized_fill.inl     |  135 +
 .../thrust/system/detail/generic/unique.h     |   97 +
 .../thrust/system/detail/generic/unique.inl   |  139 +
 .../system/detail/generic/unique_by_key.h     |   94 +
 .../system/detail/generic/unique_by_key.inl   |  139 +
 .../thrust/system/detail/internal/decompose.h |  113 +
 .../detail/sequential/adjacent_difference.h   |   73 +
 .../system/detail/sequential/assign_value.h   |   42 +
 .../system/detail/sequential/binary_search.h  |  158 +
 .../thrust/system/detail/sequential/copy.h    |   62 +
 .../thrust/system/detail/sequential/copy.inl  |  146 +
 .../system/detail/sequential/copy_backward.h  |   53 +
 .../thrust/system/detail/sequential/copy_if.h |   72 +
 .../thrust/system/detail/sequential/count.h   |   22 +
 .../thrust/system/detail/sequential/equal.h   |   22 +
 .../detail/sequential/execution_policy.h      |   75 +
 .../thrust/system/detail/sequential/extrema.h |  138 +
 .../thrust/system/detail/sequential/fill.h    |   22 +
 .../thrust/system/detail/sequential/find.h    |   70 +
 .../system/detail/sequential/for_each.h       |   94 +
 .../thrust/system/detail/sequential/gather.h  |   22 +
 .../system/detail/sequential/general_copy.h   |  146 +
 .../system/detail/sequential/generate.h       |   22 +
 .../system/detail/sequential/get_value.h      |   45 +
 .../system/detail/sequential/inner_product.h  |   22 +
 .../system/detail/sequential/insertion_sort.h |  152 +
 .../system/detail/sequential/iter_swap.h      |   46 +
 .../thrust/system/detail/sequential/logical.h |   22 +
 .../detail/sequential/malloc_and_free.h       |   53 +
 .../thrust/system/detail/sequential/merge.h   |   79 +
 .../thrust/system/detail/sequential/merge.inl |  154 +
 .../system/detail/sequential/mismatch.h       |   22 +
 .../system/detail/sequential/partition.h      |  340 +
 .../detail/sequential/per_device_resource.h   |   22 +
 .../thrust/system/detail/sequential/reduce.h  |   72 +
 .../system/detail/sequential/reduce_by_key.h  |  102 +
 .../thrust/system/detail/sequential/remove.h  |  201 +
 .../thrust/system/detail/sequential/replace.h |   22 +
 .../thrust/system/detail/sequential/reverse.h |   22 +
 .../thrust/system/detail/sequential/scan.h    |  121 +
 .../system/detail/sequential/scan_by_key.h    |  149 +
 .../thrust/system/detail/sequential/scatter.h |   22 +
 .../system/detail/sequential/sequence.h       |   22 +
 .../system/detail/sequential/set_operations.h |  223 +
 .../thrust/system/detail/sequential/sort.h    |   63 +
 .../thrust/system/detail/sequential/sort.inl  |  208 +
 .../detail/sequential/stable_merge_sort.h     |   59 +
 .../detail/sequential/stable_merge_sort.inl   |  393 +
 .../detail/sequential/stable_primitive_sort.h |   55 +
 .../sequential/stable_primitive_sort.inl      |  160 +
 .../detail/sequential/stable_radix_sort.h     |   55 +
 .../detail/sequential/stable_radix_sort.inl   |  597 ++
 .../system/detail/sequential/swap_ranges.h    |   22 +
 .../system/detail/sequential/tabulate.h       |   22 +
 .../detail/sequential/temporary_buffer.h      |   22 +
 .../system/detail/sequential/transform.h      |   22 +
 .../detail/sequential/transform_reduce.h      |   22 +
 .../system/detail/sequential/transform_scan.h |   22 +
 .../system/detail/sequential/trivial_copy.h   |   61 +
 .../detail/sequential/uninitialized_copy.h    |   22 +
 .../detail/sequential/uninitialized_fill.h    |   22 +
 .../thrust/system/detail/sequential/unique.h  |  130 +
 .../system/detail/sequential/unique_by_key.h  |  115 +
 .../thrust/system/detail/system_error.inl     |  112 +
 .../thrust/thrust/system/error_code.h         |  522 ++
 .../thrust/thrust/system/system_error.h       |  178 +
 .../thirdparty/thrust/thrust/system_error.h   |   49 +
 .../thirdparty/thrust/thrust/tabulate.h       |  125 +
 .../thirdparty/thrust/thrust/transform.h      |  721 ++
 .../thrust/thrust/transform_reduce.h          |  194 +
 .../thirdparty/thrust/thrust/transform_scan.h |  320 +
 .../manifold/thirdparty/thrust/thrust/tuple.h |  575 ++
 .../thrust/type_traits/integer_sequence.h     |  381 +
 .../type_traits/is_contiguous_iterator.h      |  295 +
 .../thrust/type_traits/is_execution_policy.h  |   69 +
 ...operator_less_or_greater_function_object.h |  208 +
 .../is_operator_plus_function_object.h        |  118 +
 .../type_traits/is_trivially_relocatable.h    |  366 +
 .../type_traits/logical_metafunctions.h       |  287 +
 .../thrust/thrust/type_traits/remove_cvref.h  |   97 +
 .../thrust/thrust/type_traits/void_t.h        |   77 +
 .../thrust/thrust/uninitialized_copy.h        |  299 +
 .../thrust/thrust/uninitialized_fill.h        |  272 +
 .../thirdparty/thrust/thrust/unique.h         | 1140 +++
 .../thrust/thrust/universal_allocator.h       |   77 +
 .../thirdparty/thrust/thrust/universal_ptr.h  |   26 +
 .../thrust/thrust/universal_vector.h          |   55 +
 .../thirdparty/thrust/thrust/version.h        |   75 +
 .../thirdparty/thrust/thrust/zip_function.h   |  212 +
 1306 files changed, 332430 insertions(+), 1761 deletions(-)
 delete mode 100644 modules/csg/csg.cpp
 delete mode 100644 modules/csg/csg.h
 create mode 100644 thirdparty/manifold/.gitrepo
 create mode 100644 thirdparty/manifold/AUTHORS
 create mode 100644 thirdparty/manifold/LICENSE
 create mode 100644 thirdparty/manifold/src/collider/include/collider.h
 create mode 100644 thirdparty/manifold/src/collider/src/collider.cpp
 create mode 100644 thirdparty/manifold/src/cross_section/include/cross_section.h
 create mode 100644 thirdparty/manifold/src/cross_section/src/cross_section.cpp
 create mode 100644 thirdparty/manifold/src/manifold/include/manifold.h
 create mode 100644 thirdparty/manifold/src/manifold/src/boolean3.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/boolean3.h
 create mode 100644 thirdparty/manifold/src/manifold/src/boolean_result.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/constructors.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/csg_tree.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/csg_tree.h
 create mode 100644 thirdparty/manifold/src/manifold/src/edge_op.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/face_op.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/impl.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/impl.h
 create mode 100644 thirdparty/manifold/src/manifold/src/manifold.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/mesh_fixes.h
 create mode 100644 thirdparty/manifold/src/manifold/src/properties.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/shared.h
 create mode 100644 thirdparty/manifold/src/manifold/src/smoothing.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/sort.cpp
 create mode 100644 thirdparty/manifold/src/manifold/src/subdivision.cpp
 create mode 100644 thirdparty/manifold/src/polygon/include/polygon.h
 create mode 100644 thirdparty/manifold/src/polygon/src/polygon.cpp
 create mode 100644 thirdparty/manifold/src/sdf/include/sdf.h
 create mode 100644 thirdparty/manifold/src/sdf/src/sdf.cpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/.gitignore
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/.gitrepo
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/ConvexHull.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/HalfEdgeMesh.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/MathUtils.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/QuickHull.cpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/QuickHull.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/README.md
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/Structs/Mesh.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/Structs/Plane.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/Structs/Pool.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/Structs/Ray.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/Structs/Vector3.hpp
 create mode 100644 thirdparty/manifold/src/third_party/quickhull/Structs/VertexDataSource.hpp
 create mode 100644 thirdparty/manifold/src/utilities/include/hashtable.h
 create mode 100644 thirdparty/manifold/src/utilities/include/optional_assert.h
 create mode 100644 thirdparty/manifold/src/utilities/include/par.h
 create mode 100644 thirdparty/manifold/src/utilities/include/public.h
 create mode 100644 thirdparty/manifold/src/utilities/include/sparse.h
 create mode 100644 thirdparty/manifold/src/utilities/include/svd.h
 create mode 100644 thirdparty/manifold/src/utilities/include/tri_dist.h
 create mode 100644 thirdparty/manifold/src/utilities/include/utils.h
 create mode 100644 thirdparty/manifold/src/utilities/include/vec.h
 create mode 100644 thirdparty/manifold/src/utilities/include/vec_view.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/.gitrepo
 create mode 100644 thirdparty/manifold/thirdparty/glm/copying.txt
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/CMakeLists.txt
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/_features.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/_fixes.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/_noise.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle_func.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/_vectorize.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/compute_common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/compute_vector_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_common.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_common_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_integer.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_integer_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_packing.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_packing_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/glm.cpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/qualifier.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/setup.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_float.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_half.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_half.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_quat_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/exponential.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/_matrix_vectorize.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_int_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_uint_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4_sized.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/fwd.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/geometric.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/glm.cppm
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/glm.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/constants.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/constants.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/integer.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/noise.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/noise.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/packing.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/packing.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion_simd.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/random.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/random.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/reciprocal.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/round.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/round.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/type_aligned.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtc/vec1.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/bit.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/bit.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/common.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/common.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/easing.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/easing.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/extend.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/extend.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/float_notmalize.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/functions.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/functions.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/hash.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/hash.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/integer.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/io.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/io.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/norm.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/norm.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/normal.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/normal.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/number_precision.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/pca.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/pca.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/projection.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/projection.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/range.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_multiplication.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/spline.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/spline.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/texture.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/texture.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/transform.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/transform.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/vec_swizzle.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.inl
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/integer.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat2x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat2x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat2x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat3x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat3x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat3x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat4x2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat4x3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/mat4x4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/matrix.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/packing.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/common.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/exponential.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/geometric.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/integer.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/matrix.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/neon.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/packing.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/platform.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/trigonometric.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/simd/vector_relational.h
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/trigonometric.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/vec2.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/vec3.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/vec4.hpp
 create mode 100644 thirdparty/manifold/thirdparty/glm/glm/vector_relational.hpp
 create mode 100644 thirdparty/manifold/thirdparty/thrust/.gitrepo
 create mode 100644 thirdparty/manifold/thirdparty/thrust/LICENSE
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/.gitrepo
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/LICENSE.TXT
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/annotated_ptr
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/atomic
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/barrier
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/latch
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/pipeline
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/semaphore
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/array
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/atomic
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/barrier
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/bit
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cassert
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ccomplex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cfloat
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/chrono
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/climits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cmath
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/complex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstddef
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstdint
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ctime
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__access_property
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__annotated_ptr
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__config
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__functional_base
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_pop
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_push
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__threading_support
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config_site.in
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__debug
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__errc
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_03
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base_03
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__libcpp_version
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__locale
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_pop
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__string
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tree
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__undef_macros
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/any
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/array
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/atomic
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/barrier
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bit
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bitset
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cassert
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cctype
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfloat
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/charconv
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/chrono
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ciso646
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/climits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/clocale
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cmath
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/compare
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csignal
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstring
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctime
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/deque
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/exception
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/execution
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__memory
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/iterator
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/__hash
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_set
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/float.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fstream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/functional
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/future
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ios
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iostream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/istream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iterator
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/latch
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/list
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/map
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/math.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/memory
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/module.modulemap
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/mutex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/new
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/numeric
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/optional
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ostream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/queue
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/random
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ratio
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/regex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/set
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/span
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/sstream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stack
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string_view
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/strstream
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/android/locale_bionic.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_c11.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_derived.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_generated.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_gcc.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_msvc.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_nvrtc.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_scopes.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/cxx_atomic.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/fuchsia/xlocale.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/limits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/locale_mgmt_aix.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/support.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/xlocale.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/musl/xlocale.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/newlib/xlocale.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/floatingpoint.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/wchar.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/xlocale.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__nop_locale_mgmt.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__posix_l_fallback.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__strtonum_fallback.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/system_error
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/thread
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tuple
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/utility
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/valarray
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/variant
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/vector
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/version
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/functional
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/initializer_list
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/iterator
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/latch
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/limits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ratio
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/semaphore
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/tuple
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/type_traits
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/utility
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/version
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__preprocessor
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__target_macros
 create mode 100644 thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/target
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/addressof.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/adjacent_difference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/advance.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/allocate_unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/async/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/async/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/async/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/async/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/async/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/async/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/binary_search.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/complex.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/count.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/adjacent_difference.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/advance.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/algorithm_wrapper.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/alignment.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/no_throw_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator_aware_execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/binary_search.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/caching_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/arithmetic.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/c99math.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrig.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrigf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccosh.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccoshf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexp.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexpf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clog.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clogf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/complex.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cpow.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cproj.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinh.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinhf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrt.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrtf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanh.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanhf.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/math_private.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/stream.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler_fence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/config.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_compatibility.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_dialect.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/debug.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/deprecated.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/device_system.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/exec_check_disable.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/forceinline.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/global_workarounds.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_device.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_system.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/memory_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/namespace.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/config/simple_defines.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/count.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/count.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp11_required.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp14_required.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/cstdint.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/dependencies_aware_execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/device_delete.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/device_free.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/device_malloc.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/device_new.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/device_ptr.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/distance.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/equal.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/event_error.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator_fwd.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_dependencies.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/extrema.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/fill.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/find.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/for_each.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/function.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/argument.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/composite.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/arithmetic_operators.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/assignment_operator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/bitwise_operators.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/compound_assignment_operators.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/logical_operators.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/operator_adaptors.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/relational_operators.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/placeholder.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/gather.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/generate.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/get_iterator_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/inner_product.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_math.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/internal_functional.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/logical.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/malloc_and_free.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_algorithms.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_wrapper.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/merge.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/minmax.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/mismatch.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/modern_gcc_required.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/mpl/math.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_wrapper.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/overlapped_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/pair.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/partition.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/preprocessor.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/range/head_flags.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/range/tail_flags.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_pointer_cast.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_reference_cast.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/reduce.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/reference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/reference_forward_declaration.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/remove.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/replace.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/reverse.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/scan.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/scatter.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/select_system.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/seq.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/sequence.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/set_operations.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/shuffle.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/sort.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/static_assert.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/static_map.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/swap_ranges.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/tabulate.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_buffer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/transform.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_reduce.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_scan.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/trivial_sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_algorithms.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_meta_transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_deduction.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/function_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_member_function.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_nested_type.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_trivial_assign.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_call_possible.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_metafunction_defined.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/minimum_type.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/pointer_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/result_of_adaptable_function.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_copy.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_fill.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/unique.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/use_default.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/util/align.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_delete.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_free.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_make_unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_malloc.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_malloc_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_new.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_new_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_ptr.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_reference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/device_vector.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/distance.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/equal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/event.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/extrema.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/find.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/functional.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/future.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/gather.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/generate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/host_vector.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/inner_product.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/constant_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/counting_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_assign.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_system_tag.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/constant_iterator_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/counting_iterator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/device_system_tag.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/discard_iterator_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/distance_from_result.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/host_system_tag.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/is_iterator_category.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_adaptor_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_system.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_traversal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_facade_category.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traits.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traversal_tags.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/join_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_category.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_system.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/normal_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/permutation_iterator_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/retag.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tagged_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_input_output_iterator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_iterator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_output_iterator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tuple_of_iterator_references.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/universal_categories.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator_base.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/discard_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_adaptor.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_categories.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_facade.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_traits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/permutation_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/retag.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/reverse_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_input_output_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_output_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/iterator/zip_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/limits.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/logical.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/memory.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/merge.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mismatch.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/device_memory_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_pool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_sync_pool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_tls_pool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/fancy_pointer_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/host_memory_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/memory_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/new.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/polymorphic_adaptor.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/pool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/pool_options.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/sync_pool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/tls_pool.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/universal_memory_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/mr/validator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/optional.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/pair.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/partition.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/per_device_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/random.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/remove.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/replace.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/reverse.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/scatter.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/set_operations.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/shuffle.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/swap.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/adjacent_difference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/assign_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/binary_search.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy_if.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/count.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/equal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/extrema.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/find.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/gather.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/generate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/get_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/inner_product.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/iter_swap.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/logical.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/malloc_and_free.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/memory.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/merge.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/mismatch.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/par.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/partition.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/per_device_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/remove.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/replace.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reverse.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scatter.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/set_operations.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/swap_ranges.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/tabulate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/temporary_buffer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/vector.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/pointer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/vector.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/adjacent_difference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/assign_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/binary_search.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy_if.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/count.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/equal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/extrema.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/find.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/gather.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/generate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/get_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/inner_product.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/iter_swap.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/logical.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/malloc_and_free.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/merge.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/mismatch.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/partition.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/per_device_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/remove.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/replace.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reverse.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scatter.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/set_operations.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/swap_ranges.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/tabulate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/temporary_buffer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/bad_alloc.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/errno.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_category.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_code.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_condition.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/logical.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/per_device_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system_exists.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tag.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/internal/decompose.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/adjacent_difference.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/assign_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/binary_search.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_backward.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_if.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/count.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/equal.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/extrema.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/find.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/for_each.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/gather.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/general_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/generate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/get_value.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/inner_product.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/insertion_sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/iter_swap.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/logical.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/malloc_and_free.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/mismatch.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/partition.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/per_device_resource.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/remove.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/replace.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reverse.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scatter.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/set_operations.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/swap_ranges.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/tabulate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/temporary_buffer.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/trivial_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique_by_key.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/detail/system_error.inl
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/error_code.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system/system_error.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/system_error.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/tabulate.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/transform.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/transform_reduce.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/transform_scan.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/tuple.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/integer_sequence.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_contiguous_iterator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_execution_policy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_plus_function_object.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_trivially_relocatable.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/logical_metafunctions.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/remove_cvref.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/type_traits/void_t.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_copy.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_fill.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/unique.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/universal_allocator.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/universal_ptr.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/universal_vector.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/version.h
 create mode 100644 thirdparty/manifold/thirdparty/thrust/thrust/zip_function.h

diff --git a/modules/csg/SCsub b/modules/csg/SCsub
index 1cf9974fc138..87ffb8e3fa8b 100644
--- a/modules/csg/SCsub
+++ b/modules/csg/SCsub
@@ -6,6 +6,65 @@ Import("env_modules")
 env_csg = env_modules.Clone()
 
 # Godot source files
+
+if env_csg["disable_exceptions"]:
+    # Enable exceptions for only manifold.
+    if env_csg.msvc:
+        env_csg.Append(CPPDEFINES=[("_HAS_EXCEPTIONS", 1)], CCFLAGS=["/EHsc"])
+    else:
+        env_csg.Append(CXXFLAGS=["-fexceptions"])
+
+
+thirdparty_dir = "#thirdparty/manifold/"
+thirdparty_sources = [
+    thirdparty_dir + file
+    for file in [
+        "src/polygon/src/polygon.cpp",
+        "src/manifold/src/constructors.cpp",
+        "src/manifold/src/edge_op.cpp",
+        "src/manifold/src/face_op.cpp",
+        "src/manifold/src/impl.cpp",
+        "src/manifold/src/boolean_result.cpp",
+        "src/manifold/src/boolean3.cpp",
+        "src/manifold/src/manifold.cpp",
+        "src/manifold/src/properties.cpp",
+        "src/manifold/src/smoothing.cpp",
+        "src/manifold/src/sort.cpp",
+        "src/collider/src/collider.cpp",
+        "src/manifold/src/subdivision.cpp",
+        "src/manifold/src/csg_tree.cpp",
+        "src/cross_section/src/cross_section.cpp",
+        "src/third_party/quickhull/QuickHull.cpp",
+    ]
+]
+
+env_csg.Append(CPPDEFINES=["THRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_CPP"])
+if env["platform"] == "web":
+    env_csg.Append(CPPDEFINES=["_LIBCUDACXX_HAS_THREAD_API_EXTERNAL", "_LIBCUDACXX_HAS_THREAD_API_CUDA"])
+    env_csg.Append(CCFLAGS=["-sDISABLE_EXCEPTION_CATCHING=0"])
+    env_csg.Append(LINKFLAGS=["-sERROR_ON_UNDEFINED_SYMBOLS=0"])
+
+env_csg.Prepend(
+    CPPPATH=[
+        thirdparty_dir + path
+        for path in [
+            "src/third_party/quickhull",
+            "src/manifold/include",
+            "src/utilities/include",
+            "src/cross_section/include",
+            "src/polygon/include",
+            "src/collider/include",
+            "thirdparty/thrust/dependencies/libcudacxx/include",
+            "thirdparty/thrust",
+            "thirdparty/glm",
+        ]
+    ]
+)
+
+env_thirdparty = env_csg.Clone()
+env_thirdparty.disable_warnings()
+env_thirdparty.add_source_files(env.modules_sources, thirdparty_sources)
+
 env_csg.add_source_files(env.modules_sources, "*.cpp")
 if env.editor_build:
     env_csg.add_source_files(env.modules_sources, "editor/*.cpp")
diff --git a/modules/csg/csg.cpp b/modules/csg/csg.cpp
deleted file mode 100644
index a4a3c768e94f..000000000000
--- a/modules/csg/csg.cpp
+++ /dev/null
@@ -1,1547 +0,0 @@
-/**************************************************************************/
-/*  csg.cpp                                                               */
-/**************************************************************************/
-/*                         This file is part of:                          */
-/*                             GODOT ENGINE                               */
-/*                        https://godotengine.org                         */
-/**************************************************************************/
-/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */
-/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur.                  */
-/*                                                                        */
-/* Permission is hereby granted, free of charge, to any person obtaining  */
-/* a copy of this software and associated documentation files (the        */
-/* "Software"), to deal in the Software without restriction, including    */
-/* without limitation the rights to use, copy, modify, merge, publish,    */
-/* distribute, sublicense, and/or sell copies of the Software, and to     */
-/* permit persons to whom the Software is furnished to do so, subject to  */
-/* the following conditions:                                              */
-/*                                                                        */
-/* The above copyright notice and this permission notice shall be         */
-/* included in all copies or substantial portions of the Software.        */
-/*                                                                        */
-/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,        */
-/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF     */
-/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */
-/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY   */
-/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,   */
-/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE      */
-/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                 */
-/**************************************************************************/
-
-#include "csg.h"
-
-#include "core/math/geometry_2d.h"
-#include "core/math/math_funcs.h"
-#include "core/templates/sort_array.h"
-
-// Static helper functions.
-
-inline static bool is_snapable(const Vector3 &p_point1, const Vector3 &p_point2, real_t p_distance) {
-	return p_point2.distance_squared_to(p_point1) < p_distance * p_distance;
-}
-
-inline static Vector2 interpolate_segment_uv(const Vector2 p_segment_points[2], const Vector2 p_uvs[2], const Vector2 &p_interpolation_point) {
-	if (p_segment_points[0].is_equal_approx(p_segment_points[1])) {
-		return p_uvs[0];
-	}
-
-	float segment_length = p_segment_points[0].distance_to(p_segment_points[1]);
-	float distance = p_segment_points[0].distance_to(p_interpolation_point);
-	float fraction = distance / segment_length;
-
-	return p_uvs[0].lerp(p_uvs[1], fraction);
-}
-
-inline static Vector2 interpolate_triangle_uv(const Vector2 p_vertices[3], const Vector2 p_uvs[3], const Vector2 &p_interpolation_point) {
-	if (p_interpolation_point.is_equal_approx(p_vertices[0])) {
-		return p_uvs[0];
-	}
-	if (p_interpolation_point.is_equal_approx(p_vertices[1])) {
-		return p_uvs[1];
-	}
-	if (p_interpolation_point.is_equal_approx(p_vertices[2])) {
-		return p_uvs[2];
-	}
-
-	Vector2 edge1 = p_vertices[1] - p_vertices[0];
-	Vector2 edge2 = p_vertices[2] - p_vertices[0];
-	Vector2 interpolation = p_interpolation_point - p_vertices[0];
-
-	float edge1_on_edge1 = edge1.dot(edge1);
-	float edge1_on_edge2 = edge1.dot(edge2);
-	float edge2_on_edge2 = edge2.dot(edge2);
-	float inter_on_edge1 = interpolation.dot(edge1);
-	float inter_on_edge2 = interpolation.dot(edge2);
-	float scale = (edge1_on_edge1 * edge2_on_edge2 - edge1_on_edge2 * edge1_on_edge2);
-	if (scale == 0) {
-		return p_uvs[0];
-	}
-
-	float v = (edge2_on_edge2 * inter_on_edge1 - edge1_on_edge2 * inter_on_edge2) / scale;
-	float w = (edge1_on_edge1 * inter_on_edge2 - edge1_on_edge2 * inter_on_edge1) / scale;
-	float u = 1.0f - v - w;
-
-	return p_uvs[0] * u + p_uvs[1] * v + p_uvs[2] * w;
-}
-
-static inline bool ray_intersects_triangle(const Vector3 &p_from, const Vector3 &p_dir, const Vector3 p_vertices[3], float p_tolerance, Vector3 &r_intersection_point) {
-	Vector3 edge1 = p_vertices[1] - p_vertices[0];
-	Vector3 edge2 = p_vertices[2] - p_vertices[0];
-	Vector3 h = p_dir.cross(edge2);
-	real_t a = edge1.dot(h);
-	// Check if ray is parallel to triangle.
-	if (Math::is_zero_approx(a)) {
-		return false;
-	}
-	real_t f = 1.0 / a;
-
-	Vector3 s = p_from - p_vertices[0];
-	real_t u = f * s.dot(h);
-	if (u < 0.0 - p_tolerance || u > 1.0 + p_tolerance) {
-		return false;
-	}
-
-	Vector3 q = s.cross(edge1);
-	real_t v = f * p_dir.dot(q);
-	if (v < 0.0 - p_tolerance || u + v > 1.0 + p_tolerance) {
-		return false;
-	}
-
-	// Ray intersects triangle.
-	// Calculate distance.
-	real_t t = f * edge2.dot(q);
-	// Confirm triangle is in front of ray.
-	if (t >= p_tolerance) {
-		r_intersection_point = p_from + p_dir * t;
-		return true;
-	} else {
-		return false;
-	}
-}
-
-inline bool is_point_in_triangle(const Vector3 &p_point, const Vector3 p_vertices[3], int p_shifted = 0) {
-	real_t det = p_vertices[0].dot(p_vertices[1].cross(p_vertices[2]));
-
-	// If determinant is, zero try shift the triangle and the point.
-	if (Math::is_zero_approx(det)) {
-		if (p_shifted > 2) {
-			// Triangle appears degenerate, so ignore it.
-			return false;
-		}
-		Vector3 shift_by;
-		shift_by[p_shifted] = 1;
-		Vector3 shifted_point = p_point + shift_by;
-		Vector3 shifted_vertices[3] = { p_vertices[0] + shift_by, p_vertices[1] + shift_by, p_vertices[2] + shift_by };
-		return is_point_in_triangle(shifted_point, shifted_vertices, p_shifted + 1);
-	}
-
-	// Find the barycentric coordinates of the point with respect to the vertices.
-	real_t lambda[3];
-	lambda[0] = p_vertices[1].cross(p_vertices[2]).dot(p_point) / det;
-	lambda[1] = p_vertices[2].cross(p_vertices[0]).dot(p_point) / det;
-	lambda[2] = p_vertices[0].cross(p_vertices[1]).dot(p_point) / det;
-
-	// Point is in the plane if all lambdas sum to 1.
-	if (!Math::is_equal_approx(lambda[0] + lambda[1] + lambda[2], 1)) {
-		return false;
-	}
-
-	// Point is inside the triangle if all lambdas are positive.
-	if (lambda[0] < 0 || lambda[1] < 0 || lambda[2] < 0) {
-		return false;
-	}
-
-	return true;
-}
-
-inline static bool is_triangle_degenerate(const Vector2 p_vertices[3], real_t p_vertex_snap2) {
-	real_t det = p_vertices[0].x * p_vertices[1].y - p_vertices[0].x * p_vertices[2].y +
-			p_vertices[0].y * p_vertices[2].x - p_vertices[0].y * p_vertices[1].x +
-			p_vertices[1].x * p_vertices[2].y - p_vertices[1].y * p_vertices[2].x;
-
-	return det < p_vertex_snap2;
-}
-
-inline static bool are_segments_parallel(const Vector2 p_segment1_points[2], const Vector2 p_segment2_points[2], float p_vertex_snap2) {
-	Vector2 segment1 = p_segment1_points[1] - p_segment1_points[0];
-	Vector2 segment2 = p_segment2_points[1] - p_segment2_points[0];
-	real_t segment1_length2 = segment1.dot(segment1);
-	real_t segment2_length2 = segment2.dot(segment2);
-	real_t segment_onto_segment = segment2.dot(segment1);
-
-	if (segment1_length2 < p_vertex_snap2 || segment2_length2 < p_vertex_snap2) {
-		return true;
-	}
-
-	real_t max_separation2;
-	if (segment1_length2 > segment2_length2) {
-		max_separation2 = segment2_length2 - segment_onto_segment * segment_onto_segment / segment1_length2;
-	} else {
-		max_separation2 = segment1_length2 - segment_onto_segment * segment_onto_segment / segment2_length2;
-	}
-
-	return max_separation2 < p_vertex_snap2;
-}
-
-// CSGBrush
-
-void CSGBrush::_regen_face_aabbs() {
-	for (int i = 0; i < faces.size(); i++) {
-		faces.write[i].aabb = AABB();
-		faces.write[i].aabb.position = faces[i].vertices[0];
-		faces.write[i].aabb.expand_to(faces[i].vertices[1]);
-		faces.write[i].aabb.expand_to(faces[i].vertices[2]);
-	}
-}
-
-void CSGBrush::build_from_faces(const Vector<Vector3> &p_vertices, const Vector<Vector2> &p_uvs, const Vector<bool> &p_smooth, const Vector<Ref<Material>> &p_materials, const Vector<bool> &p_flip_faces) {
-	faces.clear();
-
-	int vc = p_vertices.size();
-
-	ERR_FAIL_COND((vc % 3) != 0);
-
-	const Vector3 *rv = p_vertices.ptr();
-	int uvc = p_uvs.size();
-	const Vector2 *ruv = p_uvs.ptr();
-	int sc = p_smooth.size();
-	const bool *rs = p_smooth.ptr();
-	int mc = p_materials.size();
-	const Ref<Material> *rm = p_materials.ptr();
-	int ic = p_flip_faces.size();
-	const bool *ri = p_flip_faces.ptr();
-
-	HashMap<Ref<Material>, int> material_map;
-
-	faces.resize(p_vertices.size() / 3);
-
-	for (int i = 0; i < faces.size(); i++) {
-		Face &f = faces.write[i];
-		f.vertices[0] = rv[i * 3 + 0];
-		f.vertices[1] = rv[i * 3 + 1];
-		f.vertices[2] = rv[i * 3 + 2];
-
-		if (uvc == vc) {
-			f.uvs[0] = ruv[i * 3 + 0];
-			f.uvs[1] = ruv[i * 3 + 1];
-			f.uvs[2] = ruv[i * 3 + 2];
-		}
-
-		if (sc == vc / 3) {
-			f.smooth = rs[i];
-		} else {
-			f.smooth = false;
-		}
-
-		if (ic == vc / 3) {
-			f.invert = ri[i];
-		} else {
-			f.invert = false;
-		}
-
-		if (mc == vc / 3) {
-			Ref<Material> mat = rm[i];
-			if (mat.is_valid()) {
-				HashMap<Ref<Material>, int>::ConstIterator E = material_map.find(mat);
-
-				if (E) {
-					f.material = E->value;
-				} else {
-					f.material = material_map.size();
-					material_map[mat] = f.material;
-				}
-
-			} else {
-				f.material = -1;
-			}
-		}
-	}
-
-	materials.resize(material_map.size());
-	for (const KeyValue<Ref<Material>, int> &E : material_map) {
-		materials.write[E.value] = E.key;
-	}
-
-	_regen_face_aabbs();
-}
-
-void CSGBrush::copy_from(const CSGBrush &p_brush, const Transform3D &p_xform) {
-	faces = p_brush.faces;
-	materials = p_brush.materials;
-
-	for (int i = 0; i < faces.size(); i++) {
-		for (int j = 0; j < 3; j++) {
-			faces.write[i].vertices[j] = p_xform.xform(p_brush.faces[i].vertices[j]);
-		}
-	}
-
-	_regen_face_aabbs();
-}
-
-// CSGBrushOperation
-
-void CSGBrushOperation::merge_brushes(Operation p_operation, const CSGBrush &p_brush_a, const CSGBrush &p_brush_b, CSGBrush &r_merged_brush, float p_vertex_snap) {
-	// Check for face collisions and add necessary faces.
-	Build2DFaceCollection build2DFaceCollection;
-	for (int i = 0; i < p_brush_a.faces.size(); i++) {
-		for (int j = 0; j < p_brush_b.faces.size(); j++) {
-			if (p_brush_a.faces[i].aabb.intersects_inclusive(p_brush_b.faces[j].aabb)) {
-				update_faces(p_brush_a, i, p_brush_b, j, build2DFaceCollection, p_vertex_snap);
-			}
-		}
-	}
-
-	// Add faces to MeshMerge.
-	MeshMerge mesh_merge;
-	mesh_merge.vertex_snap = p_vertex_snap;
-
-	for (int i = 0; i < p_brush_a.faces.size(); i++) {
-		Ref<Material> material;
-		if (p_brush_a.faces[i].material != -1) {
-			material = p_brush_a.materials[p_brush_a.faces[i].material];
-		}
-
-		if (build2DFaceCollection.build2DFacesA.has(i)) {
-			build2DFaceCollection.build2DFacesA[i].addFacesToMesh(mesh_merge, p_brush_a.faces[i].smooth, p_brush_a.faces[i].invert, material, false);
-		} else {
-			Vector3 points[3];
-			Vector2 uvs[3];
-			for (int j = 0; j < 3; j++) {
-				points[j] = p_brush_a.faces[i].vertices[j];
-				uvs[j] = p_brush_a.faces[i].uvs[j];
-			}
-			mesh_merge.add_face(points, uvs, p_brush_a.faces[i].smooth, p_brush_a.faces[i].invert, material, false);
-		}
-	}
-
-	for (int i = 0; i < p_brush_b.faces.size(); i++) {
-		Ref<Material> material;
-		if (p_brush_b.faces[i].material != -1) {
-			material = p_brush_b.materials[p_brush_b.faces[i].material];
-		}
-
-		if (build2DFaceCollection.build2DFacesB.has(i)) {
-			build2DFaceCollection.build2DFacesB[i].addFacesToMesh(mesh_merge, p_brush_b.faces[i].smooth, p_brush_b.faces[i].invert, material, true);
-		} else {
-			Vector3 points[3];
-			Vector2 uvs[3];
-			for (int j = 0; j < 3; j++) {
-				points[j] = p_brush_b.faces[i].vertices[j];
-				uvs[j] = p_brush_b.faces[i].uvs[j];
-			}
-			mesh_merge.add_face(points, uvs, p_brush_b.faces[i].smooth, p_brush_b.faces[i].invert, material, true);
-		}
-	}
-
-	// Mark faces that ended up inside the intersection.
-	mesh_merge.mark_inside_faces();
-
-	// Create new brush and fill with new faces.
-	r_merged_brush.faces.clear();
-
-	switch (p_operation) {
-		case OPERATION_UNION: {
-			int outside_count = 0;
-
-			for (int i = 0; i < mesh_merge.faces.size(); i++) {
-				if (mesh_merge.faces[i].inside) {
-					continue;
-				}
-				outside_count++;
-			}
-
-			r_merged_brush.faces.resize(outside_count);
-
-			outside_count = 0;
-
-			for (int i = 0; i < mesh_merge.faces.size(); i++) {
-				if (mesh_merge.faces[i].inside) {
-					continue;
-				}
-
-				for (int j = 0; j < 3; j++) {
-					r_merged_brush.faces.write[outside_count].vertices[j] = mesh_merge.points[mesh_merge.faces[i].points[j]];
-					r_merged_brush.faces.write[outside_count].uvs[j] = mesh_merge.faces[i].uvs[j];
-				}
-
-				r_merged_brush.faces.write[outside_count].smooth = mesh_merge.faces[i].smooth;
-				r_merged_brush.faces.write[outside_count].invert = mesh_merge.faces[i].invert;
-				r_merged_brush.faces.write[outside_count].material = mesh_merge.faces[i].material_idx;
-				outside_count++;
-			}
-
-			r_merged_brush._regen_face_aabbs();
-
-		} break;
-
-		case OPERATION_INTERSECTION: {
-			int inside_count = 0;
-
-			for (int i = 0; i < mesh_merge.faces.size(); i++) {
-				if (!mesh_merge.faces[i].inside) {
-					continue;
-				}
-				inside_count++;
-			}
-
-			r_merged_brush.faces.resize(inside_count);
-
-			inside_count = 0;
-
-			for (int i = 0; i < mesh_merge.faces.size(); i++) {
-				if (!mesh_merge.faces[i].inside) {
-					continue;
-				}
-
-				for (int j = 0; j < 3; j++) {
-					r_merged_brush.faces.write[inside_count].vertices[j] = mesh_merge.points[mesh_merge.faces[i].points[j]];
-					r_merged_brush.faces.write[inside_count].uvs[j] = mesh_merge.faces[i].uvs[j];
-				}
-
-				r_merged_brush.faces.write[inside_count].smooth = mesh_merge.faces[i].smooth;
-				r_merged_brush.faces.write[inside_count].invert = mesh_merge.faces[i].invert;
-				r_merged_brush.faces.write[inside_count].material = mesh_merge.faces[i].material_idx;
-				inside_count++;
-			}
-
-			r_merged_brush._regen_face_aabbs();
-
-		} break;
-
-		case OPERATION_SUBTRACTION: {
-			int face_count = 0;
-
-			for (int i = 0; i < mesh_merge.faces.size(); i++) {
-				if (mesh_merge.faces[i].from_b && !mesh_merge.faces[i].inside) {
-					continue;
-				}
-				if (!mesh_merge.faces[i].from_b && mesh_merge.faces[i].inside) {
-					continue;
-				}
-				face_count++;
-			}
-
-			r_merged_brush.faces.resize(face_count);
-
-			face_count = 0;
-
-			for (int i = 0; i < mesh_merge.faces.size(); i++) {
-				if (mesh_merge.faces[i].from_b && !mesh_merge.faces[i].inside) {
-					continue;
-				}
-				if (!mesh_merge.faces[i].from_b && mesh_merge.faces[i].inside) {
-					continue;
-				}
-
-				for (int j = 0; j < 3; j++) {
-					r_merged_brush.faces.write[face_count].vertices[j] = mesh_merge.points[mesh_merge.faces[i].points[j]];
-					r_merged_brush.faces.write[face_count].uvs[j] = mesh_merge.faces[i].uvs[j];
-				}
-
-				if (mesh_merge.faces[i].from_b) {
-					//invert facing of insides of B
-					SWAP(r_merged_brush.faces.write[face_count].vertices[1], r_merged_brush.faces.write[face_count].vertices[2]);
-					SWAP(r_merged_brush.faces.write[face_count].uvs[1], r_merged_brush.faces.write[face_count].uvs[2]);
-				}
-
-				r_merged_brush.faces.write[face_count].smooth = mesh_merge.faces[i].smooth;
-				r_merged_brush.faces.write[face_count].invert = mesh_merge.faces[i].invert;
-				r_merged_brush.faces.write[face_count].material = mesh_merge.faces[i].material_idx;
-				face_count++;
-			}
-
-			r_merged_brush._regen_face_aabbs();
-
-		} break;
-	}
-
-	// Update the list of materials.
-	r_merged_brush.materials.resize(mesh_merge.materials.size());
-	for (const KeyValue<Ref<Material>, int> &E : mesh_merge.materials) {
-		r_merged_brush.materials.write[E.value] = E.key;
-	}
-}
-
-// CSGBrushOperation::MeshMerge
-
-// Use a limit to speed up bvh and limit the depth.
-#define BVH_LIMIT 8
-
-int CSGBrushOperation::MeshMerge::_create_bvh(FaceBVH *r_facebvhptr, FaceBVH **r_facebvhptrptr, int p_from, int p_size, int p_depth, int &r_max_depth, int &r_max_alloc) {
-	if (p_depth > r_max_depth) {
-		r_max_depth = p_depth;
-	}
-
-	if (p_size == 0) {
-		return -1;
-	}
-
-	if (p_size <= BVH_LIMIT) {
-		for (int i = 0; i < p_size - 1; i++) {
-			r_facebvhptrptr[p_from + i]->next = r_facebvhptrptr[p_from + i + 1] - r_facebvhptr;
-		}
-		return r_facebvhptrptr[p_from] - r_facebvhptr;
-	}
-
-	AABB aabb;
-	aabb = r_facebvhptrptr[p_from]->aabb;
-	for (int i = 1; i < p_size; i++) {
-		aabb.merge_with(r_facebvhptrptr[p_from + i]->aabb);
-	}
-
-	int li = aabb.get_longest_axis_index();
-
-	switch (li) {
-		case Vector3::AXIS_X: {
-			SortArray<FaceBVH *, FaceBVHCmpX> sort_x;
-			sort_x.nth_element(0, p_size, p_size / 2, &r_facebvhptrptr[p_from]);
-			//sort_x.sort(&p_bb[p_from],p_size);
-		} break;
-
-		case Vector3::AXIS_Y: {
-			SortArray<FaceBVH *, FaceBVHCmpY> sort_y;
-			sort_y.nth_element(0, p_size, p_size / 2, &r_facebvhptrptr[p_from]);
-			//sort_y.sort(&p_bb[p_from],p_size);
-		} break;
-
-		case Vector3::AXIS_Z: {
-			SortArray<FaceBVH *, FaceBVHCmpZ> sort_z;
-			sort_z.nth_element(0, p_size, p_size / 2, &r_facebvhptrptr[p_from]);
-			//sort_z.sort(&p_bb[p_from],p_size);
-		} break;
-	}
-
-	int left = _create_bvh(r_facebvhptr, r_facebvhptrptr, p_from, p_size / 2, p_depth + 1, r_max_depth, r_max_alloc);
-	int right = _create_bvh(r_facebvhptr, r_facebvhptrptr, p_from + p_size / 2, p_size - p_size / 2, p_depth + 1, r_max_depth, r_max_alloc);
-
-	int index = r_max_alloc++;
-	FaceBVH *_new = &r_facebvhptr[index];
-	_new->aabb = aabb;
-	_new->center = aabb.get_center();
-	_new->face = -1;
-	_new->left = left;
-	_new->right = right;
-	_new->next = -1;
-
-	return index;
-}
-
-void CSGBrushOperation::MeshMerge::_add_distance(List<IntersectionDistance> &r_intersectionsA, List<IntersectionDistance> &r_intersectionsB, bool p_from_B, real_t p_distance_squared, bool p_is_conormal) const {
-	List<IntersectionDistance> &intersections = p_from_B ? r_intersectionsB : r_intersectionsA;
-
-	// Check if distance exists.
-	for (const IntersectionDistance E : intersections) {
-		if (E.is_conormal == p_is_conormal && Math::is_equal_approx(E.distance_squared, p_distance_squared)) {
-			return;
-		}
-	}
-	IntersectionDistance distance;
-	distance.is_conormal = p_is_conormal;
-	distance.distance_squared = p_distance_squared;
-	intersections.push_back(distance);
-}
-
-bool CSGBrushOperation::MeshMerge::_bvh_inside(FaceBVH *r_facebvhptr, int p_max_depth, int p_bvh_first, int p_face_idx) const {
-	Face face = faces[p_face_idx];
-	Vector3 face_points[3] = {
-		points[face.points[0]],
-		points[face.points[1]],
-		points[face.points[2]]
-	};
-	Vector3 face_center = (face_points[0] + face_points[1] + face_points[2]) / 3.0;
-	Vector3 face_normal = Plane(face_points[0], face_points[1], face_points[2]).normal;
-
-	uint32_t *stack = (uint32_t *)alloca(sizeof(int) * p_max_depth);
-
-	enum {
-		TEST_AABB_BIT = 0,
-		VISIT_LEFT_BIT = 1,
-		VISIT_RIGHT_BIT = 2,
-		VISIT_DONE_BIT = 3,
-		VISITED_BIT_SHIFT = 29,
-		NODE_IDX_MASK = (1 << VISITED_BIT_SHIFT) - 1,
-		VISITED_BIT_MASK = ~NODE_IDX_MASK
-	};
-
-	List<IntersectionDistance> intersectionsA;
-	List<IntersectionDistance> intersectionsB;
-
-	Intersection closest_intersection;
-	closest_intersection.found = false;
-
-	int level = 0;
-	int pos = p_bvh_first;
-	stack[0] = pos;
-
-	while (true) {
-		uint32_t node = stack[level] & NODE_IDX_MASK;
-		const FaceBVH *current_facebvhptr = &(r_facebvhptr[node]);
-		bool done = false;
-
-		switch (stack[level] >> VISITED_BIT_SHIFT) {
-			case TEST_AABB_BIT: {
-				if (current_facebvhptr->face >= 0) {
-					while (current_facebvhptr) {
-						if (p_face_idx != current_facebvhptr->face &&
-								current_facebvhptr->aabb.intersects_ray(face_center, face_normal)) {
-							const Face &current_face = faces[current_facebvhptr->face];
-							Vector3 current_points[3] = {
-								points[current_face.points[0]],
-								points[current_face.points[1]],
-								points[current_face.points[2]]
-							};
-							Vector3 current_normal = Plane(current_points[0], current_points[1], current_points[2]).normal;
-							Vector3 intersection_point;
-							// Check if faces are co-planar.
-							if (current_normal.is_equal_approx(face_normal) &&
-									is_point_in_triangle(face_center, current_points)) {
-								// Only add an intersection if not a B face.
-								if (!face.from_b) {
-									_add_distance(intersectionsA, intersectionsB, current_face.from_b, 0, true);
-								}
-							} else if (ray_intersects_triangle(face_center, face_normal, current_points, CMP_EPSILON, intersection_point)) {
-								real_t distance_squared = face_center.distance_squared_to(intersection_point);
-								real_t inner = current_normal.dot(face_normal);
-								// If the faces are perpendicular, ignore this face.
-								// The triangles on the side should be intersected and result in the correct behavior.
-								if (!Math::is_zero_approx(inner)) {
-									_add_distance(intersectionsA, intersectionsB, current_face.from_b, distance_squared, inner > 0.0f);
-								}
-							}
-
-							if (face.from_b != current_face.from_b) {
-								if (current_normal.is_equal_approx(face_normal) &&
-										is_point_in_triangle(face_center, current_points)) {
-									// Only add an intersection if not a B face.
-									if (!face.from_b) {
-										closest_intersection.found = true;
-										closest_intersection.conormal = 1.0f;
-										closest_intersection.distance_squared = 0.0f;
-										closest_intersection.origin_angle = -FLT_MAX;
-									}
-								} else if (ray_intersects_triangle(face_center, face_normal, current_points, CMP_EPSILON, intersection_point)) {
-									Intersection potential_intersection;
-									potential_intersection.found = true;
-									potential_intersection.conormal = face_normal.dot(current_normal);
-									potential_intersection.distance_squared = face_center.distance_squared_to(intersection_point);
-									potential_intersection.origin_angle = Math::abs(potential_intersection.conormal);
-									real_t intersection_dist_from_face = face_normal.dot(intersection_point - face_center);
-									for (int i = 0; i < 3; i++) {
-										real_t point_dist_from_face = face_normal.dot(current_points[i] - face_center);
-										if (!Math::is_equal_approx(point_dist_from_face, intersection_dist_from_face) &&
-												point_dist_from_face < intersection_dist_from_face) {
-											potential_intersection.origin_angle = -potential_intersection.origin_angle;
-											break;
-										}
-									}
-									if (potential_intersection.conormal != 0.0f) {
-										if (!closest_intersection.found) {
-											closest_intersection = potential_intersection;
-										} else if (!Math::is_equal_approx(potential_intersection.distance_squared, closest_intersection.distance_squared) &&
-												potential_intersection.distance_squared < closest_intersection.distance_squared) {
-											closest_intersection = potential_intersection;
-										} else if (Math::is_equal_approx(potential_intersection.distance_squared, closest_intersection.distance_squared)) {
-											if (potential_intersection.origin_angle < closest_intersection.origin_angle) {
-												closest_intersection = potential_intersection;
-											}
-										}
-									}
-								}
-							}
-						}
-
-						if (current_facebvhptr->next != -1) {
-							current_facebvhptr = &r_facebvhptr[current_facebvhptr->next];
-						} else {
-							current_facebvhptr = nullptr;
-						}
-					}
-
-					stack[level] = (VISIT_DONE_BIT << VISITED_BIT_SHIFT) | node;
-
-				} else {
-					bool valid = current_facebvhptr->aabb.intersects_ray(face_center, face_normal);
-
-					if (!valid) {
-						stack[level] = (VISIT_DONE_BIT << VISITED_BIT_SHIFT) | node;
-					} else {
-						stack[level] = (VISIT_LEFT_BIT << VISITED_BIT_SHIFT) | node;
-					}
-				}
-				continue;
-			}
-
-			case VISIT_LEFT_BIT: {
-				stack[level] = (VISIT_RIGHT_BIT << VISITED_BIT_SHIFT) | node;
-				stack[level + 1] = current_facebvhptr->left | TEST_AABB_BIT;
-				level++;
-				continue;
-			}
-
-			case VISIT_RIGHT_BIT: {
-				stack[level] = (VISIT_DONE_BIT << VISITED_BIT_SHIFT) | node;
-				stack[level + 1] = current_facebvhptr->right | TEST_AABB_BIT;
-				level++;
-				continue;
-			}
-
-			case VISIT_DONE_BIT: {
-				if (level == 0) {
-					done = true;
-					break;
-				} else {
-					level--;
-				}
-				continue;
-			}
-		}
-
-		if (done) {
-			break;
-		}
-	}
-
-	if (!closest_intersection.found) {
-		return false;
-	} else {
-		return closest_intersection.conormal > 0.0f;
-	}
-}
-
-void CSGBrushOperation::MeshMerge::mark_inside_faces() {
-	// Mark faces that are inside. This helps later do the boolean ops when merging.
-	// This approach is very brute force with a bunch of optimizations,
-	// such as BVH and pre AABB intersection test.
-
-	Vector<FaceBVH> bvhvec;
-	bvhvec.resize(faces.size() * 3); // Will never be larger than this (TODO: Make better)
-	FaceBVH *facebvh = bvhvec.ptrw();
-
-	AABB aabb_a;
-	AABB aabb_b;
-
-	bool first_a = true;
-	bool first_b = true;
-
-	for (int i = 0; i < faces.size(); i++) {
-		facebvh[i].left = -1;
-		facebvh[i].right = -1;
-		facebvh[i].face = i;
-		facebvh[i].aabb.position = points[faces[i].points[0]];
-		facebvh[i].aabb.expand_to(points[faces[i].points[1]]);
-		facebvh[i].aabb.expand_to(points[faces[i].points[2]]);
-		facebvh[i].center = facebvh[i].aabb.get_center();
-		facebvh[i].aabb.grow_by(vertex_snap);
-		facebvh[i].next = -1;
-
-		if (faces[i].from_b) {
-			if (first_b) {
-				aabb_b = facebvh[i].aabb;
-				first_b = false;
-			} else {
-				aabb_b.merge_with(facebvh[i].aabb);
-			}
-		} else {
-			if (first_a) {
-				aabb_a = facebvh[i].aabb;
-				first_a = false;
-			} else {
-				aabb_a.merge_with(facebvh[i].aabb);
-			}
-		}
-	}
-
-	AABB intersection_aabb = aabb_a.intersection(aabb_b);
-
-	// Check if shape AABBs intersect.
-	if (intersection_aabb.size == Vector3()) {
-		return;
-	}
-
-	Vector<FaceBVH *> bvhtrvec;
-	bvhtrvec.resize(faces.size());
-	FaceBVH **bvhptr = bvhtrvec.ptrw();
-	for (int i = 0; i < faces.size(); i++) {
-		bvhptr[i] = &facebvh[i];
-	}
-
-	int max_depth = 0;
-	int max_alloc = faces.size();
-	_create_bvh(facebvh, bvhptr, 0, faces.size(), 1, max_depth, max_alloc);
-
-	for (int i = 0; i < faces.size(); i++) {
-		// Check if face AABB intersects the intersection AABB.
-		if (!intersection_aabb.intersects_inclusive(facebvh[i].aabb)) {
-			continue;
-		}
-
-		if (_bvh_inside(facebvh, max_depth, max_alloc - 1, i)) {
-			faces.write[i].inside = true;
-		}
-	}
-}
-
-void CSGBrushOperation::MeshMerge::add_face(const Vector3 p_points[3], const Vector2 p_uvs[3], bool p_smooth, bool p_invert, const Ref<Material> &p_material, bool p_from_b) {
-	int indices[3];
-	for (int i = 0; i < 3; i++) {
-		VertexKey vk;
-		vk.x = int((double(p_points[i].x) + double(vertex_snap) * 0.31234) / double(vertex_snap));
-		vk.y = int((double(p_points[i].y) + double(vertex_snap) * 0.31234) / double(vertex_snap));
-		vk.z = int((double(p_points[i].z) + double(vertex_snap) * 0.31234) / double(vertex_snap));
-
-		int res;
-		if (snap_cache.lookup(vk, res)) {
-			indices[i] = res;
-		} else {
-			indices[i] = points.size();
-			points.push_back(p_points[i]);
-			snap_cache.set(vk, indices[i]);
-		}
-	}
-
-	// Don't add degenerate faces.
-	if (indices[0] == indices[2] || indices[0] == indices[1] || indices[1] == indices[2]) {
-		return;
-	}
-
-	MeshMerge::Face face;
-	face.from_b = p_from_b;
-	face.inside = false;
-	face.smooth = p_smooth;
-	face.invert = p_invert;
-
-	if (p_material.is_valid()) {
-		if (!materials.has(p_material)) {
-			face.material_idx = materials.size();
-			materials[p_material] = face.material_idx;
-		} else {
-			face.material_idx = materials[p_material];
-		}
-	} else {
-		face.material_idx = -1;
-	}
-
-	for (int k = 0; k < 3; k++) {
-		face.points[k] = indices[k];
-		face.uvs[k] = p_uvs[k];
-	}
-
-	faces.push_back(face);
-}
-
-// CSGBrushOperation::Build2DFaces
-
-int CSGBrushOperation::Build2DFaces::_get_point_idx(const Vector2 &p_point) {
-	for (int vertex_idx = 0; vertex_idx < vertices.size(); ++vertex_idx) {
-		if (vertices[vertex_idx].point.distance_squared_to(p_point) < vertex_snap2) {
-			return vertex_idx;
-		}
-	}
-	return -1;
-}
-
-int CSGBrushOperation::Build2DFaces::_add_vertex(const Vertex2D &p_vertex) {
-	// Check if vertex exists.
-	int vertex_id = _get_point_idx(p_vertex.point);
-	if (vertex_id != -1) {
-		return vertex_id;
-	}
-
-	vertices.push_back(p_vertex);
-	return vertices.size() - 1;
-}
-
-void CSGBrushOperation::Build2DFaces::_add_vertex_idx_sorted(Vector<int> &r_vertex_indices, int p_new_vertex_index) {
-	if (p_new_vertex_index >= 0 && !r_vertex_indices.has(p_new_vertex_index)) {
-		ERR_FAIL_COND_MSG(p_new_vertex_index >= vertices.size(), "Invalid vertex index.");
-
-		// The first vertex.
-		if (r_vertex_indices.size() == 0) {
-			// Simply add it.
-			r_vertex_indices.push_back(p_new_vertex_index);
-			return;
-		}
-
-		// The second vertex.
-		if (r_vertex_indices.size() == 1) {
-			Vector2 first_point = vertices[r_vertex_indices[0]].point;
-			Vector2 new_point = vertices[p_new_vertex_index].point;
-
-			// Sort along the axis with the greatest difference.
-			int axis = 0;
-			if (Math::abs(new_point.x - first_point.x) < Math::abs(new_point.y - first_point.y)) {
-				axis = 1;
-			}
-
-			// Add it to the beginning or the end appropriately.
-			if (new_point[axis] < first_point[axis]) {
-				r_vertex_indices.insert(0, p_new_vertex_index);
-			} else {
-				r_vertex_indices.push_back(p_new_vertex_index);
-			}
-
-			return;
-		}
-
-		// Third or later vertices.
-		Vector2 first_point = vertices[r_vertex_indices[0]].point;
-		Vector2 last_point = vertices[r_vertex_indices[r_vertex_indices.size() - 1]].point;
-		Vector2 new_point = vertices[p_new_vertex_index].point;
-
-		// Determine axis being sorted against i.e. the axis with the greatest difference.
-		int axis = 0;
-		if (Math::abs(last_point.x - first_point.x) < Math::abs(last_point.y - first_point.y)) {
-			axis = 1;
-		}
-
-		// Insert the point at the appropriate index.
-		for (int insert_idx = 0; insert_idx < r_vertex_indices.size(); ++insert_idx) {
-			Vector2 insert_point = vertices[r_vertex_indices[insert_idx]].point;
-			if (new_point[axis] < insert_point[axis]) {
-				r_vertex_indices.insert(insert_idx, p_new_vertex_index);
-				return;
-			}
-		}
-
-		// New largest, add it to the end.
-		r_vertex_indices.push_back(p_new_vertex_index);
-	}
-}
-
-void CSGBrushOperation::Build2DFaces::_merge_faces(const Vector<int> &p_segment_indices) {
-	int segments = p_segment_indices.size() - 1;
-	if (segments < 2) {
-		return;
-	}
-
-	// Faces around an inner vertex are merged by moving the inner vertex to the first vertex.
-	for (int sorted_idx = 1; sorted_idx < segments; ++sorted_idx) {
-		int closest_idx = 0;
-		int inner_idx = p_segment_indices[sorted_idx];
-
-		if (sorted_idx > segments / 2) {
-			// Merge to other segment end.
-			closest_idx = segments;
-			// Reverse the merge order.
-			inner_idx = p_segment_indices[segments + segments / 2 - sorted_idx];
-		}
-
-		// Find the mergeable faces.
-		Vector<int> merge_faces_idx;
-		Vector<Face2D> merge_faces;
-		Vector<int> merge_faces_inner_vertex_idx;
-		for (int face_idx = 0; face_idx < faces.size(); ++face_idx) {
-			for (int face_vertex_idx = 0; face_vertex_idx < 3; ++face_vertex_idx) {
-				if (faces[face_idx].vertex_idx[face_vertex_idx] == inner_idx) {
-					merge_faces_idx.push_back(face_idx);
-					merge_faces.push_back(faces[face_idx]);
-					merge_faces_inner_vertex_idx.push_back(face_vertex_idx);
-				}
-			}
-		}
-
-		Vector<int> degenerate_points;
-
-		// Create the new faces.
-		for (int merge_idx = 0; merge_idx < merge_faces.size(); ++merge_idx) {
-			int outer_edge_idx[2];
-			outer_edge_idx[0] = merge_faces[merge_idx].vertex_idx[(merge_faces_inner_vertex_idx[merge_idx] + 1) % 3];
-			outer_edge_idx[1] = merge_faces[merge_idx].vertex_idx[(merge_faces_inner_vertex_idx[merge_idx] + 2) % 3];
-
-			// Skip flattened faces.
-			if (outer_edge_idx[0] == p_segment_indices[closest_idx] ||
-					outer_edge_idx[1] == p_segment_indices[closest_idx]) {
-				continue;
-			}
-
-			//Don't create degenerate triangles.
-			Vector2 edge1[2] = {
-				vertices[outer_edge_idx[0]].point,
-				vertices[p_segment_indices[closest_idx]].point
-			};
-			Vector2 edge2[2] = {
-				vertices[outer_edge_idx[1]].point,
-				vertices[p_segment_indices[closest_idx]].point
-			};
-			if (are_segments_parallel(edge1, edge2, vertex_snap2)) {
-				if (!degenerate_points.find(outer_edge_idx[0])) {
-					degenerate_points.push_back(outer_edge_idx[0]);
-				}
-				if (!degenerate_points.find(outer_edge_idx[1])) {
-					degenerate_points.push_back(outer_edge_idx[1]);
-				}
-				continue;
-			}
-
-			// Create new faces.
-			Face2D new_face;
-			new_face.vertex_idx[0] = p_segment_indices[closest_idx];
-			new_face.vertex_idx[1] = outer_edge_idx[0];
-			new_face.vertex_idx[2] = outer_edge_idx[1];
-			faces.push_back(new_face);
-		}
-
-		// Delete the old faces in reverse index order.
-		merge_faces_idx.sort();
-		merge_faces_idx.reverse();
-		for (int i = 0; i < merge_faces_idx.size(); ++i) {
-			faces.remove_at(merge_faces_idx[i]);
-		}
-
-		if (degenerate_points.size() == 0) {
-			continue;
-		}
-
-		// Split faces using degenerate points.
-		for (int face_idx = 0; face_idx < faces.size(); ++face_idx) {
-			Face2D face = faces[face_idx];
-			Vertex2D face_vertices[3] = {
-				vertices[face.vertex_idx[0]],
-				vertices[face.vertex_idx[1]],
-				vertices[face.vertex_idx[2]]
-			};
-			Vector2 face_points[3] = {
-				face_vertices[0].point,
-				face_vertices[1].point,
-				face_vertices[2].point
-			};
-
-			for (int point_idx = 0; point_idx < degenerate_points.size(); ++point_idx) {
-				int degenerate_idx = degenerate_points[point_idx];
-				Vector2 point_2D = vertices[degenerate_idx].point;
-
-				// Check if point is existing face vertex.
-				bool existing = false;
-				for (int i = 0; i < 3; ++i) {
-					if (face_vertices[i].point.distance_squared_to(point_2D) < vertex_snap2) {
-						existing = true;
-						break;
-					}
-				}
-				if (existing) {
-					continue;
-				}
-
-				// Check if point is on each edge.
-				for (int face_edge_idx = 0; face_edge_idx < 3; ++face_edge_idx) {
-					Vector2 edge_points[2] = {
-						face_points[face_edge_idx],
-						face_points[(face_edge_idx + 1) % 3]
-					};
-					Vector2 closest_point = Geometry2D::get_closest_point_to_segment(point_2D, edge_points);
-
-					if (point_2D.distance_squared_to(closest_point) < vertex_snap2) {
-						int opposite_vertex_idx = face.vertex_idx[(face_edge_idx + 2) % 3];
-
-						// If new vertex snaps to degenerate vertex, just delete this face.
-						if (degenerate_idx == opposite_vertex_idx) {
-							faces.remove_at(face_idx);
-							// Update index.
-							--face_idx;
-							break;
-						}
-
-						// Create two new faces around the new edge and remove this face.
-						// The new edge is the last edge.
-						Face2D left_face;
-						left_face.vertex_idx[0] = degenerate_idx;
-						left_face.vertex_idx[1] = face.vertex_idx[(face_edge_idx + 1) % 3];
-						left_face.vertex_idx[2] = opposite_vertex_idx;
-						Face2D right_face;
-						right_face.vertex_idx[0] = opposite_vertex_idx;
-						right_face.vertex_idx[1] = face.vertex_idx[face_edge_idx];
-						right_face.vertex_idx[2] = degenerate_idx;
-						faces.remove_at(face_idx);
-						faces.insert(face_idx, right_face);
-						faces.insert(face_idx, left_face);
-
-						// Don't check against the new faces.
-						++face_idx;
-
-						// No need to check other edges.
-						break;
-					}
-				}
-			}
-		}
-	}
-}
-
-void CSGBrushOperation::Build2DFaces::_find_edge_intersections(const Vector2 p_segment_points[2], Vector<int> &r_segment_indices) {
-	LocalVector<Vector<Vector2>> processed_edges;
-
-	// For each face.
-	for (int face_idx = 0; face_idx < faces.size(); ++face_idx) {
-		Face2D face = faces[face_idx];
-		Vertex2D face_vertices[3] = {
-			vertices[face.vertex_idx[0]],
-			vertices[face.vertex_idx[1]],
-			vertices[face.vertex_idx[2]]
-		};
-
-		// Check each edge.
-		for (int face_edge_idx = 0; face_edge_idx < 3; ++face_edge_idx) {
-			Vector<Vector2> edge_points_and_uvs = {
-				face_vertices[face_edge_idx].point,
-				face_vertices[(face_edge_idx + 1) % 3].point,
-				face_vertices[face_edge_idx].uv,
-				face_vertices[(face_edge_idx + 1) % 3].uv
-			};
-
-			Vector2 edge_points[2] = {
-				edge_points_and_uvs[0],
-				edge_points_and_uvs[1],
-			};
-			Vector2 edge_uvs[2] = {
-				edge_points_and_uvs[2],
-				edge_points_and_uvs[3],
-			};
-
-			// Check if edge has already been processed.
-			if (processed_edges.has(edge_points_and_uvs)) {
-				continue;
-			}
-
-			processed_edges.push_back(edge_points_and_uvs);
-
-			// First check if the ends of the segment are on the edge.
-			Vector2 intersection_point;
-
-			bool on_edge = false;
-			for (int edge_point_idx = 0; edge_point_idx < 2; ++edge_point_idx) {
-				intersection_point = Geometry2D::get_closest_point_to_segment(p_segment_points[edge_point_idx], edge_points);
-				if (p_segment_points[edge_point_idx].distance_squared_to(intersection_point) < vertex_snap2) {
-					on_edge = true;
-					break;
-				}
-			}
-
-			// Else check if the segment intersects the edge.
-			if (on_edge || Geometry2D::segment_intersects_segment(p_segment_points[0], p_segment_points[1], edge_points[0], edge_points[1], &intersection_point)) {
-				// Check if intersection point is an edge point.
-				if ((edge_points[0].distance_squared_to(intersection_point) < vertex_snap2) ||
-						(edge_points[1].distance_squared_to(intersection_point) < vertex_snap2)) {
-					continue;
-				}
-
-				// Check if edge exists, by checking if the intersecting segment is parallel to the edge.
-				if (are_segments_parallel(p_segment_points, edge_points, vertex_snap2)) {
-					continue;
-				}
-
-				// Add the intersection point as a new vertex.
-				Vertex2D new_vertex;
-				new_vertex.point = intersection_point;
-				new_vertex.uv = interpolate_segment_uv(edge_points, edge_uvs, intersection_point);
-				int new_vertex_idx = _add_vertex(new_vertex);
-				int opposite_vertex_idx = face.vertex_idx[(face_edge_idx + 2) % 3];
-				_add_vertex_idx_sorted(r_segment_indices, new_vertex_idx);
-
-				// If new vertex snaps to opposite vertex, just delete this face.
-				if (new_vertex_idx == opposite_vertex_idx) {
-					faces.remove_at(face_idx);
-					// Update index.
-					--face_idx;
-					break;
-				}
-
-				// If opposite point is on the segment, add its index to segment indices too.
-				Vector2 closest_point = Geometry2D::get_closest_point_to_segment(vertices[opposite_vertex_idx].point, p_segment_points);
-				if (vertices[opposite_vertex_idx].point.distance_squared_to(closest_point) < vertex_snap2) {
-					_add_vertex_idx_sorted(r_segment_indices, opposite_vertex_idx);
-				}
-
-				// Create two new faces around the new edge and remove this face.
-				// The new edge is the last edge.
-				Face2D left_face;
-				left_face.vertex_idx[0] = new_vertex_idx;
-				left_face.vertex_idx[1] = face.vertex_idx[(face_edge_idx + 1) % 3];
-				left_face.vertex_idx[2] = opposite_vertex_idx;
-				Face2D right_face;
-				right_face.vertex_idx[0] = opposite_vertex_idx;
-				right_face.vertex_idx[1] = face.vertex_idx[face_edge_idx];
-				right_face.vertex_idx[2] = new_vertex_idx;
-				faces.remove_at(face_idx);
-				faces.insert(face_idx, right_face);
-				faces.insert(face_idx, left_face);
-
-				// Check against the new faces.
-				--face_idx;
-				break;
-			}
-		}
-	}
-}
-
-int CSGBrushOperation::Build2DFaces::_insert_point(const Vector2 &p_point) {
-	int new_vertex_idx = -1;
-
-	for (int face_idx = 0; face_idx < faces.size(); ++face_idx) {
-		Face2D face = faces[face_idx];
-		Vertex2D face_vertices[3] = {
-			vertices[face.vertex_idx[0]],
-			vertices[face.vertex_idx[1]],
-			vertices[face.vertex_idx[2]]
-		};
-		Vector2 points[3] = {
-			face_vertices[0].point,
-			face_vertices[1].point,
-			face_vertices[2].point
-		};
-		Vector2 uvs[3] = {
-			face_vertices[0].uv,
-			face_vertices[1].uv,
-			face_vertices[2].uv
-		};
-
-		// Skip degenerate triangles.
-		if (is_triangle_degenerate(points, vertex_snap2)) {
-			continue;
-		}
-
-		// Check if point is existing face vertex.
-		for (int i = 0; i < 3; ++i) {
-			if (face_vertices[i].point.distance_squared_to(p_point) < vertex_snap2) {
-				return face.vertex_idx[i];
-			}
-		}
-
-		// Check if point is on each edge.
-		bool on_edge = false;
-		for (int face_edge_idx = 0; face_edge_idx < 3; ++face_edge_idx) {
-			Vector2 edge_points[2] = {
-				points[face_edge_idx],
-				points[(face_edge_idx + 1) % 3]
-			};
-			Vector2 edge_uvs[2] = {
-				uvs[face_edge_idx],
-				uvs[(face_edge_idx + 1) % 3]
-			};
-
-			Vector2 closest_point = Geometry2D::get_closest_point_to_segment(p_point, edge_points);
-			if (p_point.distance_squared_to(closest_point) < vertex_snap2) {
-				on_edge = true;
-
-				// Add the point as a new vertex.
-				Vertex2D new_vertex;
-				new_vertex.point = p_point;
-				new_vertex.uv = interpolate_segment_uv(edge_points, edge_uvs, p_point);
-				new_vertex_idx = _add_vertex(new_vertex);
-				int opposite_vertex_idx = face.vertex_idx[(face_edge_idx + 2) % 3];
-
-				// If new vertex snaps to opposite vertex, just delete this face.
-				if (new_vertex_idx == opposite_vertex_idx) {
-					faces.remove_at(face_idx);
-					// Update index.
-					--face_idx;
-					break;
-				}
-
-				// Don't create degenerate triangles.
-				Vector2 split_edge1[2] = { vertices[new_vertex_idx].point, edge_points[0] };
-				Vector2 split_edge2[2] = { vertices[new_vertex_idx].point, edge_points[1] };
-				Vector2 new_edge[2] = { vertices[new_vertex_idx].point, vertices[opposite_vertex_idx].point };
-				if (are_segments_parallel(split_edge1, new_edge, vertex_snap2) &&
-						are_segments_parallel(split_edge2, new_edge, vertex_snap2)) {
-					break;
-				}
-
-				// Create two new faces around the new edge and remove this face.
-				// The new edge is the last edge.
-				Face2D left_face;
-				left_face.vertex_idx[0] = new_vertex_idx;
-				left_face.vertex_idx[1] = face.vertex_idx[(face_edge_idx + 1) % 3];
-				left_face.vertex_idx[2] = opposite_vertex_idx;
-				Face2D right_face;
-				right_face.vertex_idx[0] = opposite_vertex_idx;
-				right_face.vertex_idx[1] = face.vertex_idx[face_edge_idx];
-				right_face.vertex_idx[2] = new_vertex_idx;
-				faces.remove_at(face_idx);
-				faces.insert(face_idx, right_face);
-				faces.insert(face_idx, left_face);
-
-				// Don't check against the new faces.
-				++face_idx;
-
-				// No need to check other edges.
-				break;
-			}
-		}
-
-		// If not on an edge, check if the point is inside the face.
-		if (!on_edge && Geometry2D::is_point_in_triangle(p_point, face_vertices[0].point, face_vertices[1].point, face_vertices[2].point)) {
-			// Add the point as a new vertex.
-			Vertex2D new_vertex;
-			new_vertex.point = p_point;
-			new_vertex.uv = interpolate_triangle_uv(points, uvs, p_point);
-			new_vertex_idx = _add_vertex(new_vertex);
-
-			// Create three new faces around this point and remove this face.
-			// The new vertex is the last vertex.
-			for (int i = 0; i < 3; ++i) {
-				// Don't create degenerate triangles.
-				Vector2 new_points[3] = { points[i], points[(i + 1) % 3], vertices[new_vertex_idx].point };
-				if (is_triangle_degenerate(new_points, vertex_snap2)) {
-					continue;
-				}
-
-				Face2D new_face;
-				new_face.vertex_idx[0] = face.vertex_idx[i];
-				new_face.vertex_idx[1] = face.vertex_idx[(i + 1) % 3];
-				new_face.vertex_idx[2] = new_vertex_idx;
-				faces.push_back(new_face);
-			}
-			faces.remove_at(face_idx);
-
-			// No need to check other faces.
-			break;
-		}
-	}
-
-	return new_vertex_idx;
-}
-
-void CSGBrushOperation::Build2DFaces::insert(const CSGBrush &p_brush, int p_face_idx) {
-	// Find edge points that cross the plane and face points that are in the plane.
-	// Map those points to 2D.
-	// Create new faces from those points.
-
-	Vector2 points_2D[3];
-	int points_count = 0;
-
-	for (int i = 0; i < 3; i++) {
-		Vector3 point_3D = p_brush.faces[p_face_idx].vertices[i];
-
-		if (plane.has_point(point_3D)) {
-			// Point is in the plane, add it.
-			Vector3 point_2D = plane.project(point_3D);
-			point_2D = to_2D.xform(point_2D);
-			points_2D[points_count++] = Vector2(point_2D.x, point_2D.y);
-
-		} else {
-			Vector3 next_point_3D = p_brush.faces[p_face_idx].vertices[(i + 1) % 3];
-
-			if (plane.has_point(next_point_3D)) {
-				continue; // Next point is in plane, it will be added separately.
-			}
-			if (plane.is_point_over(point_3D) == plane.is_point_over(next_point_3D)) {
-				continue; // Both points on the same side of the plane, ignore.
-			}
-
-			// Edge crosses the plane, find and add the intersection point.
-			Vector3 point_2D;
-			if (plane.intersects_segment(point_3D, next_point_3D, &point_2D)) {
-				point_2D = to_2D.xform(point_2D);
-				points_2D[points_count++] = Vector2(point_2D.x, point_2D.y);
-			}
-		}
-	}
-
-	Vector<int> segment_indices;
-	Vector2 segment[2];
-	int inserted_index[3] = { -1, -1, -1 };
-
-	// Insert points.
-	for (int i = 0; i < points_count; ++i) {
-		inserted_index[i] = _insert_point(points_2D[i]);
-	}
-
-	if (points_count == 2) {
-		// Insert a single segment.
-		segment[0] = points_2D[0];
-		segment[1] = points_2D[1];
-		_find_edge_intersections(segment, segment_indices);
-		for (int i = 0; i < 2; ++i) {
-			_add_vertex_idx_sorted(segment_indices, inserted_index[i]);
-		}
-		_merge_faces(segment_indices);
-	}
-
-	if (points_count == 3) {
-		// Insert three segments.
-		for (int edge_idx = 0; edge_idx < 3; ++edge_idx) {
-			segment[0] = points_2D[edge_idx];
-			segment[1] = points_2D[(edge_idx + 1) % 3];
-			_find_edge_intersections(segment, segment_indices);
-			for (int i = 0; i < 2; ++i) {
-				_add_vertex_idx_sorted(segment_indices, inserted_index[(edge_idx + i) % 3]);
-			}
-			_merge_faces(segment_indices);
-			segment_indices.clear();
-		}
-	}
-}
-
-void CSGBrushOperation::Build2DFaces::addFacesToMesh(MeshMerge &r_mesh_merge, bool p_smooth, bool p_invert, const Ref<Material> &p_material, bool p_from_b) {
-	for (int face_idx = 0; face_idx < faces.size(); ++face_idx) {
-		Face2D face = faces[face_idx];
-		Vertex2D fv[3] = {
-			vertices[face.vertex_idx[0]],
-			vertices[face.vertex_idx[1]],
-			vertices[face.vertex_idx[2]]
-		};
-
-		// Convert 2D vertex points to 3D.
-		Vector3 points_3D[3];
-		Vector2 uvs[3];
-		for (int i = 0; i < 3; ++i) {
-			Vector3 point_2D(fv[i].point.x, fv[i].point.y, 0);
-			points_3D[i] = to_3D.xform(point_2D);
-			uvs[i] = fv[i].uv;
-		}
-
-		r_mesh_merge.add_face(points_3D, uvs, p_smooth, p_invert, p_material, p_from_b);
-	}
-}
-
-CSGBrushOperation::Build2DFaces::Build2DFaces(const CSGBrush &p_brush, int p_face_idx, float p_vertex_snap2) :
-		vertex_snap2(p_vertex_snap2 * p_vertex_snap2) {
-	// Convert 3D vertex points to 2D.
-	Vector3 points_3D[3] = {
-		p_brush.faces[p_face_idx].vertices[0],
-		p_brush.faces[p_face_idx].vertices[1],
-		p_brush.faces[p_face_idx].vertices[2],
-	};
-
-	plane = Plane(points_3D[0], points_3D[1], points_3D[2]);
-	to_3D.origin = points_3D[0];
-	to_3D.basis.set_column(2, plane.normal);
-	to_3D.basis.set_column(0, (points_3D[1] - points_3D[2]).normalized());
-	to_3D.basis.set_column(1, to_3D.basis.get_column(0).cross(to_3D.basis.get_column(2)).normalized());
-	to_2D = to_3D.affine_inverse();
-
-	Face2D face;
-	for (int i = 0; i < 3; i++) {
-		Vertex2D vertex;
-		Vector3 point_2D = to_2D.xform(points_3D[i]);
-		vertex.point.x = point_2D.x;
-		vertex.point.y = point_2D.y;
-		vertex.uv = p_brush.faces[p_face_idx].uvs[i];
-		vertices.push_back(vertex);
-		face.vertex_idx[i] = i;
-	}
-	faces.push_back(face);
-}
-
-void CSGBrushOperation::update_faces(const CSGBrush &p_brush_a, const int p_face_idx_a, const CSGBrush &p_brush_b, const int p_face_idx_b, Build2DFaceCollection &p_collection, float p_vertex_snap) {
-	Vector3 vertices_a[3] = {
-		p_brush_a.faces[p_face_idx_a].vertices[0],
-		p_brush_a.faces[p_face_idx_a].vertices[1],
-		p_brush_a.faces[p_face_idx_a].vertices[2],
-	};
-
-	Vector3 vertices_b[3] = {
-		p_brush_b.faces[p_face_idx_b].vertices[0],
-		p_brush_b.faces[p_face_idx_b].vertices[1],
-		p_brush_b.faces[p_face_idx_b].vertices[2],
-	};
-
-	// Don't use degenerate faces.
-	bool has_degenerate = false;
-	if (is_snapable(vertices_a[0], vertices_a[1], p_vertex_snap) ||
-			is_snapable(vertices_a[0], vertices_a[2], p_vertex_snap) ||
-			is_snapable(vertices_a[1], vertices_a[2], p_vertex_snap)) {
-		p_collection.build2DFacesA[p_face_idx_a] = Build2DFaces();
-		has_degenerate = true;
-	}
-
-	if (is_snapable(vertices_b[0], vertices_b[1], p_vertex_snap) ||
-			is_snapable(vertices_b[0], vertices_b[2], p_vertex_snap) ||
-			is_snapable(vertices_b[1], vertices_b[2], p_vertex_snap)) {
-		p_collection.build2DFacesB[p_face_idx_b] = Build2DFaces();
-		has_degenerate = true;
-	}
-	if (has_degenerate) {
-		return;
-	}
-
-	// Ensure B has points either side of or in the plane of A.
-	int over_count = 0, under_count = 0;
-	Plane plane_a(vertices_a[0], vertices_a[1], vertices_a[2]);
-	ERR_FAIL_COND_MSG(plane_a.normal == Vector3(), "Couldn't form plane from Brush A face.");
-
-	for (int i = 0; i < 3; i++) {
-		if (plane_a.has_point(vertices_b[i])) {
-			// In plane.
-		} else if (plane_a.is_point_over(vertices_b[i])) {
-			over_count++;
-		} else {
-			under_count++;
-		}
-	}
-	// If all points under or over the plane, there is no intersection.
-	if (over_count == 3 || under_count == 3) {
-		return;
-	}
-
-	// Ensure A has points either side of or in the plane of B.
-	over_count = 0;
-	under_count = 0;
-	Plane plane_b(vertices_b[0], vertices_b[1], vertices_b[2]);
-	ERR_FAIL_COND_MSG(plane_b.normal == Vector3(), "Couldn't form plane from Brush B face.");
-
-	for (int i = 0; i < 3; i++) {
-		if (plane_b.has_point(vertices_a[i])) {
-			// In plane.
-		} else if (plane_b.is_point_over(vertices_a[i])) {
-			over_count++;
-		} else {
-			under_count++;
-		}
-	}
-	// If all points under or over the plane, there is no intersection.
-	if (over_count == 3 || under_count == 3) {
-		return;
-	}
-
-	// Check for intersection using the SAT theorem.
-	{
-		// Edge pair cross product combinations.
-		for (int i = 0; i < 3; i++) {
-			Vector3 axis_a = (vertices_a[i] - vertices_a[(i + 1) % 3]).normalized();
-
-			for (int j = 0; j < 3; j++) {
-				Vector3 axis_b = (vertices_b[j] - vertices_b[(j + 1) % 3]).normalized();
-
-				Vector3 sep_axis = axis_a.cross(axis_b);
-				if (sep_axis == Vector3()) {
-					continue; //colineal
-				}
-				sep_axis.normalize();
-
-				real_t min_a = 1e20, max_a = -1e20;
-				real_t min_b = 1e20, max_b = -1e20;
-
-				for (int k = 0; k < 3; k++) {
-					real_t d = sep_axis.dot(vertices_a[k]);
-					min_a = MIN(min_a, d);
-					max_a = MAX(max_a, d);
-					d = sep_axis.dot(vertices_b[k]);
-					min_b = MIN(min_b, d);
-					max_b = MAX(max_b, d);
-				}
-
-				min_b -= (max_a - min_a) * 0.5;
-				max_b += (max_a - min_a) * 0.5;
-
-				real_t dmin = min_b - (min_a + max_a) * 0.5;
-				real_t dmax = max_b - (min_a + max_a) * 0.5;
-
-				if (dmin > CMP_EPSILON || dmax < -CMP_EPSILON) {
-					return; // Does not contain zero, so they don't overlap.
-				}
-			}
-		}
-	}
-
-	// If we're still here, the faces probably intersect, so add new faces.
-	if (!p_collection.build2DFacesA.has(p_face_idx_a)) {
-		p_collection.build2DFacesA[p_face_idx_a] = Build2DFaces(p_brush_a, p_face_idx_a, p_vertex_snap);
-	}
-	p_collection.build2DFacesA[p_face_idx_a].insert(p_brush_b, p_face_idx_b);
-
-	if (!p_collection.build2DFacesB.has(p_face_idx_b)) {
-		p_collection.build2DFacesB[p_face_idx_b] = Build2DFaces(p_brush_b, p_face_idx_b, p_vertex_snap);
-	}
-	p_collection.build2DFacesB[p_face_idx_b].insert(p_brush_a, p_face_idx_a);
-}
diff --git a/modules/csg/csg.h b/modules/csg/csg.h
deleted file mode 100644
index 2a0831e1ce18..000000000000
--- a/modules/csg/csg.h
+++ /dev/null
@@ -1,204 +0,0 @@
-/**************************************************************************/
-/*  csg.h                                                                 */
-/**************************************************************************/
-/*                         This file is part of:                          */
-/*                             GODOT ENGINE                               */
-/*                        https://godotengine.org                         */
-/**************************************************************************/
-/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */
-/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur.                  */
-/*                                                                        */
-/* Permission is hereby granted, free of charge, to any person obtaining  */
-/* a copy of this software and associated documentation files (the        */
-/* "Software"), to deal in the Software without restriction, including    */
-/* without limitation the rights to use, copy, modify, merge, publish,    */
-/* distribute, sublicense, and/or sell copies of the Software, and to     */
-/* permit persons to whom the Software is furnished to do so, subject to  */
-/* the following conditions:                                              */
-/*                                                                        */
-/* The above copyright notice and this permission notice shall be         */
-/* included in all copies or substantial portions of the Software.        */
-/*                                                                        */
-/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,        */
-/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF     */
-/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */
-/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY   */
-/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,   */
-/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE      */
-/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                 */
-/**************************************************************************/
-
-#ifndef CSG_H
-#define CSG_H
-
-#include "core/math/aabb.h"
-#include "core/math/plane.h"
-#include "core/math/transform_3d.h"
-#include "core/math/vector2.h"
-#include "core/math/vector3.h"
-#include "core/object/ref_counted.h"
-#include "core/templates/list.h"
-#include "core/templates/oa_hash_map.h"
-#include "core/templates/vector.h"
-#include "scene/resources/material.h"
-
-struct CSGBrush {
-	struct Face {
-		Vector3 vertices[3];
-		Vector2 uvs[3];
-		AABB aabb;
-		bool smooth = false;
-		bool invert = false;
-		int material = 0;
-	};
-
-	Vector<Face> faces;
-	Vector<Ref<Material>> materials;
-
-	inline void _regen_face_aabbs();
-
-	// Create a brush from faces.
-	void build_from_faces(const Vector<Vector3> &p_vertices, const Vector<Vector2> &p_uvs, const Vector<bool> &p_smooth, const Vector<Ref<Material>> &p_materials, const Vector<bool> &p_invert_faces);
-	void copy_from(const CSGBrush &p_brush, const Transform3D &p_xform);
-};
-
-struct CSGBrushOperation {
-	enum Operation {
-		OPERATION_UNION,
-		OPERATION_INTERSECTION,
-		OPERATION_SUBTRACTION,
-	};
-
-	void merge_brushes(Operation p_operation, const CSGBrush &p_brush_a, const CSGBrush &p_brush_b, CSGBrush &r_merged_brush, float p_vertex_snap);
-
-	struct MeshMerge {
-		struct Face {
-			bool from_b = false;
-			bool inside = false;
-			int points[3] = {};
-			Vector2 uvs[3];
-			bool smooth = false;
-			bool invert = false;
-			int material_idx = 0;
-		};
-
-		struct FaceBVH {
-			int face = 0;
-			int left = 0;
-			int right = 0;
-			int next = 0;
-			Vector3 center;
-			AABB aabb;
-		};
-
-		struct FaceBVHCmpX {
-			_FORCE_INLINE_ bool operator()(const FaceBVH *p_left, const FaceBVH *p_right) const {
-				return p_left->center.x < p_right->center.x;
-			}
-		};
-
-		struct FaceBVHCmpY {
-			_FORCE_INLINE_ bool operator()(const FaceBVH *p_left, const FaceBVH *p_right) const {
-				return p_left->center.y < p_right->center.y;
-			}
-		};
-		struct FaceBVHCmpZ {
-			_FORCE_INLINE_ bool operator()(const FaceBVH *p_left, const FaceBVH *p_right) const {
-				return p_left->center.z < p_right->center.z;
-			}
-		};
-
-		struct VertexKey {
-			int32_t x, y, z;
-			_FORCE_INLINE_ bool operator<(const VertexKey &p_key) const {
-				if (x == p_key.x) {
-					if (y == p_key.y) {
-						return z < p_key.z;
-					} else {
-						return y < p_key.y;
-					}
-				} else {
-					return x < p_key.x;
-				}
-			}
-
-			_FORCE_INLINE_ bool operator==(const VertexKey &p_key) const {
-				return (x == p_key.x && y == p_key.y && z == p_key.z);
-			}
-		};
-
-		struct VertexKeyHash {
-			static _FORCE_INLINE_ uint32_t hash(const VertexKey &p_vk) {
-				uint32_t h = hash_murmur3_one_32(p_vk.x);
-				h = hash_murmur3_one_32(p_vk.y, h);
-				h = hash_murmur3_one_32(p_vk.z, h);
-				return h;
-			}
-		};
-		struct Intersection {
-			bool found = false;
-			real_t conormal = FLT_MAX;
-			real_t distance_squared = FLT_MAX;
-			real_t origin_angle = FLT_MAX;
-		};
-
-		struct IntersectionDistance {
-			bool is_conormal;
-			real_t distance_squared;
-		};
-
-		Vector<Vector3> points;
-		Vector<Face> faces;
-		HashMap<Ref<Material>, int> materials;
-		HashMap<Vector3, int> vertex_map;
-		OAHashMap<VertexKey, int, VertexKeyHash> snap_cache;
-		float vertex_snap = 0.0;
-
-		inline void _add_distance(List<IntersectionDistance> &r_intersectionsA, List<IntersectionDistance> &r_intersectionsB, bool p_from_B, real_t p_distance, bool p_is_conormal) const;
-		inline bool _bvh_inside(FaceBVH *r_facebvhptr, int p_max_depth, int p_bvh_first, int p_face_idx) const;
-		inline int _create_bvh(FaceBVH *r_facebvhptr, FaceBVH **r_facebvhptrptr, int p_from, int p_size, int p_depth, int &r_max_depth, int &r_max_alloc);
-
-		void add_face(const Vector3 p_points[3], const Vector2 p_uvs[3], bool p_smooth, bool p_invert, const Ref<Material> &p_material, bool p_from_b);
-		void mark_inside_faces();
-	};
-
-	struct Build2DFaces {
-		struct Vertex2D {
-			Vector2 point;
-			Vector2 uv;
-		};
-
-		struct Face2D {
-			int vertex_idx[3] = {};
-		};
-
-		Vector<Vertex2D> vertices;
-		Vector<Face2D> faces;
-		Plane plane;
-		Transform3D to_2D;
-		Transform3D to_3D;
-		float vertex_snap2 = 0.0;
-
-		inline int _get_point_idx(const Vector2 &p_point);
-		inline int _add_vertex(const Vertex2D &p_vertex);
-		inline void _add_vertex_idx_sorted(Vector<int> &r_vertex_indices, int p_new_vertex_index);
-		inline void _merge_faces(const Vector<int> &p_segment_indices);
-		inline void _find_edge_intersections(const Vector2 p_segment_points[2], Vector<int> &r_segment_indices);
-		inline int _insert_point(const Vector2 &p_point);
-
-		void insert(const CSGBrush &p_brush, int p_brush_face);
-		void addFacesToMesh(MeshMerge &r_mesh_merge, bool p_smooth, bool p_invert, const Ref<Material> &p_material, bool p_from_b);
-
-		Build2DFaces() {}
-		Build2DFaces(const CSGBrush &p_brush, int p_brush_face, float p_vertex_snap2);
-	};
-
-	struct Build2DFaceCollection {
-		HashMap<int, Build2DFaces> build2DFacesA;
-		HashMap<int, Build2DFaces> build2DFacesB;
-	};
-
-	void update_faces(const CSGBrush &p_brush_a, const int p_face_idx_a, const CSGBrush &p_brush_b, const int p_face_idx_b, Build2DFaceCollection &p_collection, float p_vertex_snap);
-};
-
-#endif // CSG_H
diff --git a/modules/csg/csg_shape.cpp b/modules/csg/csg_shape.cpp
index 7c93fbf081b2..2ad7aa290fed 100644
--- a/modules/csg/csg_shape.cpp
+++ b/modules/csg/csg_shape.cpp
@@ -32,6 +32,306 @@
 
 #include "core/math/geometry_2d.h"
 
+#include "thirdparty/manifold/src/manifold/include/manifold.h"
+#include "thirdparty/manifold/src/utilities/include/public.h"
+
+// Static helper functions.
+
+struct CSGBrush {
+	struct Face {
+		Vector3 vertices[3];
+		Vector2 uvs[3];
+		AABB aabb;
+		bool smooth = false;
+		bool invert = false;
+		int material = 0;
+	};
+
+	Vector<Face> faces;
+	Vector<Ref<Material>> materials;
+
+	manifold::Manifold manifold;
+	enum {
+		MANIFOLD_PROPERTY_POS_X = 0,
+		MANIFOLD_PROPERTY_POS_Y,
+		MANIFOLD_PROPERTY_POS_Z,
+		MANIFOLD_PROPERTY_UV_X,
+		MANIFOLD_PROPERTY_UV_Y,
+		MANIFOLD_PROPERTY_MATERIAL,
+		MANIFOLD_PROPERTY_SMOOTH,
+		MANIFOLD_PROPERTY_INVERT,
+		MANIFOLD_MAX
+	};
+	static constexpr int MANIFOLD_TRIANGLE = 3;
+
+	inline void _regen_face_aabbs() {
+		for (int i = 0; i < faces.size(); i++) {
+			faces.write[i].aabb = AABB();
+			faces.write[i].aabb.position = faces[i].vertices[0];
+			faces.write[i].aabb.expand_to(faces[i].vertices[1]);
+			faces.write[i].aabb.expand_to(faces[i].vertices[2]);
+		}
+	}
+
+	void pack_manifold(const float p_snap, HashMap<uint32_t, Ref<Material>> &r_materials) {
+		if (faces.is_empty()) {
+			return;
+		}
+		faces.sort_custom<CSGBrush::MaterialIndexComparator>();
+
+		HashSet<int32_t> unique_materials;
+		for (CSGBrush::Face &face : faces) {
+			unique_materials.insert(face.material);
+		}
+
+		LocalVector<CSGBrush> split_brushes;
+		CowData<CSGBrush::Face>::Size unique_face_i = 0;
+		for (int32_t material_id : unique_materials) {
+			uint32_t reserved_id = manifold.ReserveIDs(1);
+			CSGBrush current_brush;
+			Ref<Material> material;
+			if (material_id >= 0 && material_id < materials.size()) {
+				material = materials[material_id];
+			}
+			CSGBrush::Face starting_face = faces[unique_face_i];
+			while (unique_face_i < faces.size()) {
+				CSGBrush::Face face = faces[unique_face_i];
+				if (starting_face.material != face.material) {
+					break;
+				}
+				face.material = reserved_id;
+				current_brush.faces.push_back(face);
+				unique_face_i++;
+			}
+			manifold::MeshGL mesh;
+			mesh.runOriginalID.push_back(reserved_id);
+			mesh.triVerts.resize(current_brush.faces.size() * CSGBrush::MANIFOLD_TRIANGLE, 0);
+			mesh.vertProperties.resize(current_brush.faces.size() * CSGBrush::MANIFOLD_TRIANGLE * CSGBrush::MANIFOLD_MAX, std::numeric_limits<float>::quiet_NaN());
+			mesh.numProp = CSGBrush::MANIFOLD_MAX;
+			constexpr size_t order[CSGBrush::MANIFOLD_TRIANGLE] = { 0, 2, 1 };
+			for (CowData<CSGBrush::Face>::Size face_i = 0; face_i < current_brush.faces.size(); face_i++) {
+				const CSGBrush::Face &face = current_brush.faces[face_i];
+				for (size_t vertex_i = 0; vertex_i < CSGBrush::MANIFOLD_TRIANGLE; vertex_i++) {
+					int32_t index = face_i * CSGBrush::MANIFOLD_TRIANGLE + vertex_i;
+					mesh.triVerts[face_i * CSGBrush::MANIFOLD_TRIANGLE + order[vertex_i]] = index;
+					Vector3 pos = face.vertices[vertex_i];
+					Vector2 uv = face.uvs[vertex_i];
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_POS_X] = pos.x;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_POS_Y] = pos.y;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_POS_Z] = pos.z;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_UV_X] = uv.x;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_UV_Y] = uv.y;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_MATERIAL] = reserved_id;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_SMOOTH] = face.smooth ? 1.0f : 0.0f;
+					mesh.vertProperties[index * CSGBrush::MANIFOLD_MAX + CSGBrush::MANIFOLD_PROPERTY_INVERT] = face.invert ? 1.0f : 0.0f;
+				}
+			}
+			mesh.precision = p_snap;
+			mesh.Merge();
+			r_materials.insert(reserved_id, material);
+			manifold = manifold.Boolean(manifold::Manifold(mesh), manifold::OpType::Add);
+			split_brushes.push_back(current_brush);
+		}
+	}
+
+	void unpack_manifold(const HashMap<uint32_t, Ref<Material>> &p_materials) {
+		Ref<StandardMaterial3D> default_material;
+		default_material.instantiate();
+		manifold::MeshGL mesh = manifold.GetMeshGL();
+		LocalVector<Vector3> manifold_positions;
+		manifold_positions.resize(mesh.vertProperties.size());
+		LocalVector<Vector2> manifold_uvs;
+		manifold_uvs.resize(mesh.vertProperties.size());
+		LocalVector<uint64_t> manifold_materials;
+		manifold_materials.resize(mesh.vertProperties.size());
+		LocalVector<bool> manifold_smooths;
+		manifold_smooths.resize(mesh.vertProperties.size());
+		LocalVector<bool> manifold_inverts;
+		manifold_inverts.resize(mesh.vertProperties.size());
+		ERR_FAIL_COND_MSG(mesh.vertProperties.size() % mesh.numProp != 0, "Invalid vertex properties size");
+		size_t position_index = 0;
+		size_t uv_index = 0;
+		size_t material_index = 0;
+		size_t smooth_index = 0;
+		size_t invert_index = 0;
+		for (size_t property_i = 0; property_i < mesh.vertProperties.size(); property_i += CSGBrush::MANIFOLD_MAX) {
+			manifold_positions[position_index++] = Vector3(mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_POS_X], mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_POS_Y], mesh.vertProperties[property_i + MANIFOLD_PROPERTY_POS_Z]);
+			manifold_uvs[uv_index++] = Vector2(mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_UV_X], mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_UV_Y]);
+			manifold_materials[material_index++] = static_cast<int>(Math::round(mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_MATERIAL]));
+			manifold_smooths[smooth_index++] = mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_SMOOTH] > 0.5f;
+			manifold_inverts[invert_index++] = mesh.vertProperties[property_i + CSGBrush::MANIFOLD_PROPERTY_INVERT] > 0.5f;
+		}
+		faces.resize(mesh.triVerts.size() / CSGBrush::MANIFOLD_TRIANGLE);
+		constexpr int32_t order[CSGBrush::MANIFOLD_TRIANGLE] = { 0, 2, 1 };
+		for (size_t triangle_i = 0; triangle_i < mesh.triVerts.size() / CSGBrush::MANIFOLD_TRIANGLE; triangle_i++) {
+			CSGBrush::Face &face = faces.write[triangle_i];
+			for (int32_t vertex_i = 0; vertex_i < 3; vertex_i++) {
+				int32_t index = mesh.triVerts[triangle_i * CSGBrush::MANIFOLD_TRIANGLE + order[vertex_i]];
+				Vector3 position = manifold_positions[index];
+				Vector2 uv = manifold_uvs[index];
+				face.vertices[vertex_i] = Vector3(position.x, position.y, position.z);
+				face.uvs[vertex_i] = Vector2(uv.x, uv.y);
+				face.smooth = manifold_smooths[index];
+				face.invert = manifold_inverts[index];
+				uint32_t rid_id = manifold_materials[index];
+				if (p_materials.has(rid_id)) {
+					Ref<Material> material = p_materials[rid_id];
+					int32_t material_id = materials.find(material);
+					if (material_id != -1) {
+						face.material = material_id;
+					} else {
+						face.material = materials.size();
+						materials.push_back(material);
+					}
+					continue;
+				}
+				int32_t material_id = materials.find(default_material);
+				if (material_id != -1) {
+					face.material = material_id;
+				} else {
+					face.material = manifold_materials.size();
+					materials.push_back(default_material);
+				}
+			}
+		}
+
+		_regen_face_aabbs();
+	}
+
+	class MaterialIndexComparator {
+	public:
+		bool operator()(const CSGBrush::Face &a, const CSGBrush::Face &b) const {
+			return a.material < b.material;
+		}
+	};
+
+	// Create a brush from faces.
+	void build_from_faces(const Vector<Vector3> &p_vertices, const Vector<Vector2> &p_uvs, const Vector<bool> &p_smooth, const Vector<Ref<Material>> &p_materials, const Vector<bool> &p_invert_faces);
+	void copy_from(const CSGBrush &p_brush, const Transform3D &p_xform, float p_snap, HashMap<uint32_t, Ref<Material>> &r_mesh_materials);
+};
+
+inline bool is_point_in_triangle(const Vector3 &p_point, const Vector3 p_vertices[3], int p_shifted = 0) {
+	real_t det = p_vertices[0].dot(p_vertices[1].cross(p_vertices[2]));
+
+	// If determinant is, zero try shift the triangle and the point.
+	if (Math::is_zero_approx(det)) {
+		if (p_shifted > 2) {
+			// Triangle appears degenerate, so ignore it.
+			return false;
+		}
+		Vector3 shift_by;
+		shift_by[p_shifted] = 1;
+		Vector3 shifted_point = p_point + shift_by;
+		Vector3 shifted_vertices[3] = { p_vertices[0] + shift_by, p_vertices[1] + shift_by, p_vertices[2] + shift_by };
+		return is_point_in_triangle(shifted_point, shifted_vertices, p_shifted + 1);
+	}
+
+	// Find the barycentric coordinates of the point with respect to the vertices.
+	real_t lambda[3];
+	lambda[0] = p_vertices[1].cross(p_vertices[2]).dot(p_point) / det;
+	lambda[1] = p_vertices[2].cross(p_vertices[0]).dot(p_point) / det;
+	lambda[2] = p_vertices[0].cross(p_vertices[1]).dot(p_point) / det;
+
+	// Point is in the plane if all lambdas sum to 1.
+	if (!Math::is_equal_approx(lambda[0] + lambda[1] + lambda[2], 1)) {
+		return false;
+	}
+
+	// Point is inside the triangle if all lambdas are positive.
+	if (lambda[0] < 0 || lambda[1] < 0 || lambda[2] < 0) {
+		return false;
+	}
+
+	return true;
+}
+
+// CSGBrush
+
+void CSGBrush::build_from_faces(const Vector<Vector3> &p_vertices, const Vector<Vector2> &p_uvs, const Vector<bool> &p_smooth, const Vector<Ref<Material>> &p_materials, const Vector<bool> &p_flip_faces) {
+	faces.clear();
+
+	int vc = p_vertices.size();
+
+	ERR_FAIL_COND((vc % 3) != 0);
+
+	const Vector3 *rv = p_vertices.ptr();
+	int uvc = p_uvs.size();
+	const Vector2 *ruv = p_uvs.ptr();
+	int sc = p_smooth.size();
+	const bool *rs = p_smooth.ptr();
+	int mc = p_materials.size();
+	const Ref<Material> *rm = p_materials.ptr();
+	int ic = p_flip_faces.size();
+	const bool *ri = p_flip_faces.ptr();
+
+	HashMap<Ref<Material>, int> material_map;
+
+	faces.resize(p_vertices.size() / 3);
+
+	for (int i = 0; i < faces.size(); i++) {
+		Face &f = faces.write[i];
+		f.vertices[0] = rv[i * 3 + 0];
+		f.vertices[1] = rv[i * 3 + 1];
+		f.vertices[2] = rv[i * 3 + 2];
+
+		if (uvc == vc) {
+			f.uvs[0] = ruv[i * 3 + 0];
+			f.uvs[1] = ruv[i * 3 + 1];
+			f.uvs[2] = ruv[i * 3 + 2];
+		}
+
+		if (sc == vc / 3) {
+			f.smooth = rs[i];
+		} else {
+			f.smooth = false;
+		}
+
+		if (ic == vc / 3) {
+			f.invert = ri[i];
+		} else {
+			f.invert = false;
+		}
+
+		if (mc == vc / 3) {
+			Ref<Material> mat = rm[i];
+			if (mat.is_valid()) {
+				HashMap<Ref<Material>, int>::ConstIterator E = material_map.find(mat);
+
+				if (E) {
+					f.material = E->value;
+				} else {
+					f.material = material_map.size();
+					material_map[mat] = f.material;
+				}
+
+			} else {
+				f.material = -1;
+			}
+		}
+	}
+
+	materials.resize(material_map.size());
+	for (const KeyValue<Ref<Material>, int> &E : material_map) {
+		materials.write[E.value] = E.key;
+	}
+
+	_regen_face_aabbs();
+}
+
+void CSGBrush::copy_from(const CSGBrush &p_brush, const Transform3D &p_xform, float p_snap, HashMap<uint32_t, Ref<Material>> &r_mesh_materials) {
+	faces = p_brush.faces;
+	materials = p_brush.materials;
+
+	for (int i = 0; i < faces.size(); i++) {
+		for (int j = 0; j < 3; j++) {
+			faces.write[i].vertices[j] = p_xform.xform(p_brush.faces[i].vertices[j]);
+		}
+	}
+
+	_regen_face_aabbs();
+	pack_manifold(p_snap, r_mesh_materials);
+}
+
 void CSGShape3D::set_use_collision(bool p_enable) {
 	if (use_collision == p_enable) {
 		return;
@@ -170,6 +470,7 @@ CSGBrush *CSGShape3D::_get_brush() {
 		brush = nullptr;
 
 		CSGBrush *n = _build_brush();
+		HashMap<uint32_t, Ref<Material>> mesh_materials;
 
 		for (int i = 0; i < get_child_count(); i++) {
 			CSGShape3D *child = Object::cast_to<CSGShape3D>(get_child(i));
@@ -186,29 +487,31 @@ CSGBrush *CSGShape3D::_get_brush() {
 			}
 			if (!n) {
 				n = memnew(CSGBrush);
+				n->manifold = manifold::Manifold();
+				mesh_materials.clear();
 
-				n->copy_from(*n2, child->get_transform());
+				n->copy_from(*n2, child->get_transform(), snap, mesh_materials);
 
 			} else {
+				n->pack_manifold(snap, mesh_materials);
 				CSGBrush *nn = memnew(CSGBrush);
 				CSGBrush *nn2 = memnew(CSGBrush);
-				nn2->copy_from(*n2, child->get_transform());
-
-				CSGBrushOperation bop;
-
+				manifold::Manifold manifold_nn2;
+				nn2->copy_from(*n2, child->get_transform(), snap, mesh_materials);
 				switch (child->get_operation()) {
 					case CSGShape3D::OPERATION_UNION:
-						bop.merge_brushes(CSGBrushOperation::OPERATION_UNION, *n, *nn2, *nn, snap);
+						nn->manifold = n->manifold.Boolean(nn2->manifold, manifold::OpType::Add);
 						break;
 					case CSGShape3D::OPERATION_INTERSECTION:
-						bop.merge_brushes(CSGBrushOperation::OPERATION_INTERSECTION, *n, *nn2, *nn, snap);
+						nn->manifold = n->manifold.Boolean(nn2->manifold, manifold::OpType::Intersect);
 						break;
 					case CSGShape3D::OPERATION_SUBTRACTION:
-						bop.merge_brushes(CSGBrushOperation::OPERATION_SUBTRACTION, *n, *nn2, *nn, snap);
+						nn->manifold = n->manifold.Boolean(nn2->manifold, manifold::OpType::Subtract);
 						break;
 				}
 				memdelete(n);
 				memdelete(nn2);
+				nn->unpack_manifold(mesh_materials);
 				n = nn;
 			}
 		}
@@ -230,6 +533,7 @@ CSGBrush *CSGShape3D::_get_brush() {
 		}
 
 		brush = n;
+		brush->pack_manifold(snap, mesh_materials);
 
 		dirty = false;
 	}
diff --git a/modules/csg/csg_shape.h b/modules/csg/csg_shape.h
index bb7c8be43155..501c42eaba7f 100644
--- a/modules/csg/csg_shape.h
+++ b/modules/csg/csg_shape.h
@@ -31,14 +31,13 @@
 #ifndef CSG_SHAPE_H
 #define CSG_SHAPE_H
 
-#include "csg.h"
-
 #include "scene/3d/path_3d.h"
 #include "scene/3d/visual_instance_3d.h"
 #include "scene/resources/3d/concave_polygon_shape_3d.h"
 
 #include "thirdparty/misc/mikktspace.h"
 
+struct CSGBrush;
 class CSGShape3D : public GeometryInstance3D {
 	GDCLASS(CSGShape3D, GeometryInstance3D);
 
diff --git a/thirdparty/manifold/.gitrepo b/thirdparty/manifold/.gitrepo
new file mode 100644
index 000000000000..1b57ab9f6248
--- /dev/null
+++ b/thirdparty/manifold/.gitrepo
@@ -0,0 +1,12 @@
+; DO NOT EDIT (unless you know what you are doing)
+;
+; This subdirectory is a git "subrepo", and this file is maintained by the
+; git-subrepo command. See https://github.com/ingydotnet/git-subrepo#readme
+;
+[subrepo]
+	remote = https://github.com/elalish/manifold.git
+	branch = master
+	commit = 53015422549ef9f2d8bca24d4fc47d6c3e80ece0
+	parent = 1f00d4751e9b895b191a08a830c436dc4f7b2372
+	method = merge
+	cmdver = 0.4.6
diff --git a/thirdparty/manifold/AUTHORS b/thirdparty/manifold/AUTHORS
new file mode 100644
index 000000000000..a3e3c587f416
--- /dev/null
+++ b/thirdparty/manifold/AUTHORS
@@ -0,0 +1,10 @@
+# This is the list of Manifold's significant contributors.
+#
+# This does not necessarily list everyone who has contributed code,
+# especially since many employees of one corporation may be contributing.
+# To see the full list of contributors, see the revision history in
+# source control.
+Emmett Lalish <elalish>
+Chun Kit LAM <pca006132>
+Geoff deRosenroll <geoffder>
+Google LLC
diff --git a/thirdparty/manifold/LICENSE b/thirdparty/manifold/LICENSE
new file mode 100644
index 000000000000..261eeb9e9f8b
--- /dev/null
+++ b/thirdparty/manifold/LICENSE
@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
diff --git a/thirdparty/manifold/src/collider/include/collider.h b/thirdparty/manifold/src/collider/include/collider.h
new file mode 100644
index 000000000000..8d750815c95e
--- /dev/null
+++ b/thirdparty/manifold/src/collider/include/collider.h
@@ -0,0 +1,46 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include "public.h"
+#include "sparse.h"
+#include "vec.h"
+
+namespace manifold {
+
+/** @ingroup Private */
+class Collider {
+ public:
+  Collider() {}
+  Collider(const VecView<const Box>& leafBB,
+           const VecView<const uint32_t>& leafMorton);
+  bool Transform(glm::mat4x3);
+  void UpdateBoxes(const VecView<const Box>& leafBB);
+  template <const bool selfCollision = false, const bool inverted = false,
+            typename T>
+  SparseIndices Collisions(const VecView<const T>& queriesIn) const;
+
+ private:
+  Vec<Box> nodeBBox_;
+  Vec<int> nodeParent_;
+  // even nodes are leaves, odd nodes are internal, root is 1
+  Vec<thrust::pair<int, int>> internalChildren_;
+
+  int NumInternal() const { return internalChildren_.size(); };
+  int NumLeaves() const {
+    return internalChildren_.empty() ? 0 : (NumInternal() + 1);
+  };
+};
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/collider/src/collider.cpp b/thirdparty/manifold/src/collider/src/collider.cpp
new file mode 100644
index 000000000000..ca32f653d94f
--- /dev/null
+++ b/thirdparty/manifold/src/collider/src/collider.cpp
@@ -0,0 +1,395 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "collider.h"
+
+#include "par.h"
+#include "utils.h"
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#endif
+
+// Adjustable parameters
+constexpr int kInitialLength = 128;
+constexpr int kLengthMultiple = 4;
+constexpr int kSequentialThreshold = 512;
+// Fundamental constants
+constexpr int kRoot = 1;
+
+#ifdef _MSC_VER
+
+#ifndef _WINDEF_
+typedef unsigned long DWORD;
+#endif
+
+uint32_t __inline ctz(uint32_t value) {
+  DWORD trailing_zero = 0;
+
+  if (_BitScanForward(&trailing_zero, value)) {
+    return trailing_zero;
+  } else {
+    // This is undefined, I better choose 32 than 0
+    return 32;
+  }
+}
+
+uint32_t __inline clz(uint32_t value) {
+  DWORD leading_zero = 0;
+
+  if (_BitScanReverse(&leading_zero, value)) {
+    return 31 - leading_zero;
+  } else {
+    // Same remarks as above
+    return 32;
+  }
+}
+#endif
+
+namespace {
+using namespace manifold;
+
+bool IsLeaf(int node) { return node % 2 == 0; }
+bool IsInternal(int node) { return node % 2 == 1; }
+int Node2Internal(int node) { return (node - 1) / 2; }
+int Internal2Node(int internal) { return internal * 2 + 1; }
+int Node2Leaf(int node) { return node / 2; }
+int Leaf2Node(int leaf) { return leaf * 2; }
+
+struct CreateRadixTree {
+  VecView<int> nodeParent_;
+  VecView<thrust::pair<int, int>> internalChildren_;
+  const VecView<const uint32_t> leafMorton_;
+
+  int PrefixLength(uint32_t a, uint32_t b) const {
+// count-leading-zeros is used to find the number of identical highest-order
+// bits
+#ifdef _MSC_VER
+    // return __lzcnt(a ^ b);
+    return clz(a ^ b);
+#else
+    return __builtin_clz(a ^ b);
+#endif
+  }
+
+  int PrefixLength(int i, int j) const {
+    if (j < 0 || j >= leafMorton_.size()) {
+      return -1;
+    } else {
+      int out;
+      if (leafMorton_[i] == leafMorton_[j])
+        // use index to disambiguate
+        out = 32 +
+              PrefixLength(static_cast<uint32_t>(i), static_cast<uint32_t>(j));
+      else
+        out = PrefixLength(leafMorton_[i], leafMorton_[j]);
+      return out;
+    }
+  }
+
+  int RangeEnd(int i) const {
+    // Determine direction of range (+1 or -1)
+    int dir = PrefixLength(i, i + 1) - PrefixLength(i, i - 1);
+    dir = (dir > 0) - (dir < 0);
+    // Compute conservative range length with exponential increase
+    int commonPrefix = PrefixLength(i, i - dir);
+    int max_length = kInitialLength;
+    while (PrefixLength(i, i + dir * max_length) > commonPrefix)
+      max_length *= kLengthMultiple;
+    // Compute precise range length with binary search
+    int length = 0;
+    for (int step = max_length / 2; step > 0; step /= 2) {
+      if (PrefixLength(i, i + dir * (length + step)) > commonPrefix)
+        length += step;
+    }
+    return i + dir * length;
+  }
+
+  int FindSplit(int first, int last) const {
+    int commonPrefix = PrefixLength(first, last);
+    // Find the furthest object that shares more than commonPrefix bits with the
+    // first one, using binary search.
+    int split = first;
+    int step = last - first;
+    do {
+      step = (step + 1) >> 1;  // divide by 2, rounding up
+      int newSplit = split + step;
+      if (newSplit < last) {
+        int splitPrefix = PrefixLength(first, newSplit);
+        if (splitPrefix > commonPrefix) split = newSplit;
+      }
+    } while (step > 1);
+    return split;
+  }
+
+  void operator()(int internal) {
+    int first = internal;
+    // Find the range of objects with a common prefix
+    int last = RangeEnd(first);
+    if (first > last) thrust::swap(first, last);
+    // Determine where the next-highest difference occurs
+    int split = FindSplit(first, last);
+    int child1 = split == first ? Leaf2Node(split) : Internal2Node(split);
+    ++split;
+    int child2 = split == last ? Leaf2Node(split) : Internal2Node(split);
+    // Record parent_child relationships.
+    internalChildren_[internal].first = child1;
+    internalChildren_[internal].second = child2;
+    int node = Internal2Node(internal);
+    nodeParent_[child1] = node;
+    nodeParent_[child2] = node;
+  }
+};
+
+template <typename T, const bool selfCollision, typename Recorder>
+struct FindCollisions {
+  VecView<const Box> nodeBBox_;
+  VecView<const thrust::pair<int, int>> internalChildren_;
+  Recorder recorder;
+
+  int RecordCollision(int node, thrust::tuple<T, int>& query) {
+    const T& queryObj = thrust::get<0>(query);
+    const int queryIdx = thrust::get<1>(query);
+
+    bool overlaps = nodeBBox_[node].DoesOverlap(queryObj);
+    if (overlaps && IsLeaf(node)) {
+      int leafIdx = Node2Leaf(node);
+      if (!selfCollision || leafIdx != queryIdx) {
+        recorder.record(queryIdx, leafIdx);
+      }
+    }
+    return overlaps && IsInternal(node);  // Should traverse into node
+  }
+
+  void operator()(thrust::tuple<T, int> query) {
+    // stack cannot overflow because radix tree has max depth 30 (Morton code) +
+    // 32 (index).
+    int stack[64];
+    int top = -1;
+    // Depth-first search
+    int node = kRoot;
+    const int queryIdx = thrust::get<1>(query);
+    // same implies that this query do not have any collision
+    if (recorder.earlyexit(queryIdx)) return;
+    while (1) {
+      int internal = Node2Internal(node);
+      int child1 = internalChildren_[internal].first;
+      int child2 = internalChildren_[internal].second;
+
+      int traverse1 = RecordCollision(child1, query);
+      int traverse2 = RecordCollision(child2, query);
+
+      if (!traverse1 && !traverse2) {
+        if (top < 0) break;   // done
+        node = stack[top--];  // get a saved node
+      } else {
+        node = traverse1 ? child1 : child2;  // go here next
+        if (traverse1 && traverse2) {
+          stack[++top] = child2;  // save the other for later
+        }
+      }
+    }
+    recorder.end(queryIdx);
+  }
+};
+
+struct CountCollisions {
+  VecView<int> counts;
+  VecView<char> empty;
+  void record(int queryIdx, int _leafIdx) { counts[queryIdx]++; }
+  bool earlyexit(int _queryIdx) { return false; }
+  void end(int queryIdx) {
+    if (counts[queryIdx] == 0) empty[queryIdx] = 1;
+  }
+};
+
+template <const bool inverted>
+struct SeqCollisionRecorder {
+  SparseIndices& queryTri_;
+  void record(int queryIdx, int leafIdx) const {
+    if (inverted)
+      queryTri_.Add(leafIdx, queryIdx);
+    else
+      queryTri_.Add(queryIdx, leafIdx);
+  }
+  bool earlyexit(int queryIdx) const { return false; }
+  void end(int queryIdx) const {}
+};
+
+template <const bool inverted>
+struct ParCollisionRecorder {
+  SparseIndices& queryTri;
+  VecView<int> counts;
+  VecView<char> empty;
+  void record(int queryIdx, int leafIdx) {
+    int pos = counts[queryIdx]++;
+    if (inverted)
+      queryTri.Set(pos, leafIdx, queryIdx);
+    else
+      queryTri.Set(pos, queryIdx, leafIdx);
+  }
+  bool earlyexit(int queryIdx) const { return empty[queryIdx] == 1; }
+  void end(int queryIdx) const {}
+};
+
+struct BuildInternalBoxes {
+  VecView<Box> nodeBBox_;
+  VecView<int> counter_;
+  const VecView<int> nodeParent_;
+  const VecView<thrust::pair<int, int>> internalChildren_;
+
+  void operator()(int leaf) {
+    int node = Leaf2Node(leaf);
+    do {
+      node = nodeParent_[node];
+      int internal = Node2Internal(node);
+      if (AtomicAdd(counter_[internal], 1) == 0) return;
+      nodeBBox_[node] = nodeBBox_[internalChildren_[internal].first].Union(
+          nodeBBox_[internalChildren_[internal].second]);
+    } while (node != kRoot);
+  }
+};
+
+struct TransformBox {
+  const glm::mat4x3 transform;
+  void operator()(Box& box) { box = box.Transform(transform); }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Creates a Bounding Volume Hierarchy (BVH) from an input set of axis-aligned
+ * bounding boxes and corresponding Morton codes. It is assumed these vectors
+ * are already sorted by increasing Morton code.
+ */
+Collider::Collider(const VecView<const Box>& leafBB,
+                   const VecView<const uint32_t>& leafMorton) {
+  ZoneScoped;
+  ASSERT(leafBB.size() == leafMorton.size(), userErr,
+         "vectors must be the same length");
+  int num_nodes = 2 * leafBB.size() - 1;
+  // assign and allocate members
+  nodeBBox_.resize(num_nodes);
+  nodeParent_.resize(num_nodes, -1);
+  internalChildren_.resize(leafBB.size() - 1, thrust::make_pair(-1, -1));
+  // organize tree
+  for_each_n(autoPolicy(NumInternal()), countAt(0), NumInternal(),
+             CreateRadixTree({nodeParent_, internalChildren_, leafMorton}));
+  UpdateBoxes(leafBB);
+}
+
+/**
+ * For a vector of query objects, this returns a sparse array of overlaps
+ * between the queries and the bounding boxes of the collider. Queries are
+ * normally axis-aligned bounding boxes. Points can also be used, and this case
+ * overlaps are defined as lying in the XY projection of the bounding box. If
+ * the query vector is the leaf vector, set selfCollision to true, which will
+ * then not report any collisions between an index and itself.
+ */
+template <const bool selfCollision, const bool inverted, typename T>
+SparseIndices Collider::Collisions(const VecView<const T>& queriesIn) const {
+  ZoneScoped;
+  // note that the length is 1 larger than the number of queries so the last
+  // element can store the sum when using exclusive scan
+  if (queriesIn.size() < kSequentialThreshold) {
+    SparseIndices queryTri;
+    for_each_n(ExecutionPolicy::Seq, zip(queriesIn.cbegin(), countAt(0)),
+               queriesIn.size(),
+               FindCollisions<T, selfCollision, SeqCollisionRecorder<inverted>>{
+                   nodeBBox_, internalChildren_, {queryTri}});
+    return queryTri;
+  } else {
+    // compute the number of collisions to determine the size for allocation and
+    // offset, this avoids the need for atomic
+    Vec<int> counts(queriesIn.size() + 1, 0);
+    Vec<char> empty(queriesIn.size(), 0);
+    for_each_n(ExecutionPolicy::Par, zip(queriesIn.cbegin(), countAt(0)),
+               queriesIn.size(),
+               FindCollisions<T, selfCollision, CountCollisions>{
+                   nodeBBox_, internalChildren_, {counts, empty}});
+    // compute start index for each query and total count
+    exclusive_scan(ExecutionPolicy::Par, counts.begin(), counts.end(),
+                   counts.begin(), 0, std::plus<int>());
+    if (counts.back() == 0) return SparseIndices(0);
+    SparseIndices queryTri(counts.back());
+    // actually recording collisions
+    for_each_n(ExecutionPolicy::Par, zip(queriesIn.cbegin(), countAt(0)),
+               queriesIn.size(),
+               FindCollisions<T, selfCollision, ParCollisionRecorder<inverted>>{
+                   nodeBBox_, internalChildren_, {queryTri, counts, empty}});
+    return queryTri;
+  }
+}
+
+/**
+ * Recalculate the collider's internal bounding boxes without changing the
+ * hierarchy.
+ */
+void Collider::UpdateBoxes(const VecView<const Box>& leafBB) {
+  ZoneScoped;
+  ASSERT(leafBB.size() == NumLeaves(), userErr,
+         "must have the same number of updated boxes as original");
+  // copy in leaf node Boxes
+  strided_range<Vec<Box>::Iter> leaves(nodeBBox_.begin(), nodeBBox_.end(), 2);
+  auto policy = autoPolicy(NumInternal());
+  copy(policy, leafBB.cbegin(), leafBB.cend(), leaves.begin());
+  // create global counters
+  Vec<int> counter(NumInternal(), 0);
+  // kernel over leaves to save internal Boxes
+  for_each_n(
+      policy, countAt(0), NumLeaves(),
+      BuildInternalBoxes({nodeBBox_, counter, nodeParent_, internalChildren_}));
+}
+
+/**
+ * Apply axis-aligned transform to all bounding boxes. If transform is not
+ * axis-aligned, abort and return false to indicate recalculation is necessary.
+ */
+bool Collider::Transform(glm::mat4x3 transform) {
+  ZoneScoped;
+  bool axisAligned = true;
+  for (int row : {0, 1, 2}) {
+    int count = 0;
+    for (int col : {0, 1, 2}) {
+      if (transform[col][row] == 0.0f) ++count;
+    }
+    if (count != 2) axisAligned = false;
+  }
+  if (axisAligned) {
+    for_each(autoPolicy(nodeBBox_.size()), nodeBBox_.begin(), nodeBBox_.end(),
+             TransformBox({transform}));
+  }
+  return axisAligned;
+}
+
+template SparseIndices Collider::Collisions<true, false, Box>(
+    const VecView<const Box>&) const;
+
+template SparseIndices Collider::Collisions<false, false, Box>(
+    const VecView<const Box>&) const;
+
+template SparseIndices Collider::Collisions<false, false, glm::vec3>(
+    const VecView<const glm::vec3>&) const;
+
+template SparseIndices Collider::Collisions<true, true, Box>(
+    const VecView<const Box>&) const;
+
+template SparseIndices Collider::Collisions<false, true, Box>(
+    const VecView<const Box>&) const;
+
+template SparseIndices Collider::Collisions<false, true, glm::vec3>(
+    const VecView<const glm::vec3>&) const;
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/cross_section/include/cross_section.h b/thirdparty/manifold/src/cross_section/include/cross_section.h
new file mode 100644
index 000000000000..9e15bb5297f8
--- /dev/null
+++ b/thirdparty/manifold/src/cross_section/include/cross_section.h
@@ -0,0 +1,176 @@
+// Copyright 2023 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <functional>
+#include <memory>
+#include <vector>
+
+#include "glm/ext/matrix_float3x2.hpp"
+#include "glm/ext/vector_float2.hpp"
+#include "public.h"
+#include "vec_view.h"
+
+namespace manifold {
+
+/** @addtogroup Core
+ *  @{
+ */
+
+struct PathImpl;
+
+/**
+ * Two-dimensional cross sections guaranteed to be without self-intersections,
+ * or overlaps between polygons (from construction onwards). This class makes
+ * use of the [Clipper2](http://www.angusj.com/clipper2/Docs/Overview.htm)
+ * library for polygon clipping (boolean) and offsetting operations.
+ */
+class CrossSection {
+ public:
+  /** @name Creation
+   *  Constructors
+   */
+  ///@{
+
+  CrossSection();
+  ~CrossSection();
+
+  CrossSection(const CrossSection& other);
+  CrossSection& operator=(const CrossSection& other);
+  CrossSection(CrossSection&&) noexcept;
+  CrossSection& operator=(CrossSection&&) noexcept;
+
+  // Adapted from Clipper2 docs:
+  // http://www.angusj.com/clipper2/Docs/Units/Clipper/Types/FillRule.htm
+  // (Copyright © 2010-2023 Angus Johnson)
+  /**
+   * Filling rules defining which polygon sub-regions are considered to be
+   * inside a given polygon, and which sub-regions will not (based on winding
+   * numbers). See the [Clipper2
+   * docs](http://www.angusj.com/clipper2/Docs/Units/Clipper/Types/FillRule.htm)
+   * for a detailed explaination with illusrations.
+   */
+  enum class FillRule {
+    EvenOdd,   ///< Only odd numbered sub-regions are filled.
+    NonZero,   ///< Only non-zero sub-regions are filled.
+    Positive,  ///< Only sub-regions with winding counts > 0 are filled.
+    Negative   ///< Only sub-regions with winding counts < 0 are filled.
+  };
+
+  CrossSection(const SimplePolygon& contour,
+               FillRule fillrule = FillRule::Positive);
+  CrossSection(const Polygons& contours,
+               FillRule fillrule = FillRule::Positive);
+  CrossSection(const Rect& rect);
+  static CrossSection Square(const glm::vec2 dims, bool center = false);
+  static CrossSection Circle(float radius, int circularSegments = 0);
+  ///@}
+
+  /** @name Information
+   *  Details of the cross-section
+   */
+  ///@{
+  double Area() const;
+  int NumVert() const;
+  int NumContour() const;
+  bool IsEmpty() const;
+  Rect Bounds() const;
+  ///@}
+
+  /** @name Modification
+   */
+  ///@{
+  CrossSection Translate(const glm::vec2 v) const;
+  CrossSection Rotate(float degrees) const;
+  CrossSection Scale(const glm::vec2 s) const;
+  CrossSection Mirror(const glm::vec2 ax) const;
+  CrossSection Transform(const glm::mat3x2& m) const;
+  CrossSection Warp(std::function<void(glm::vec2&)> warpFunc) const;
+  CrossSection WarpBatch(
+      std::function<void(VecView<glm::vec2>)> warpFunc) const;
+  CrossSection Simplify(double epsilon = 1e-6) const;
+
+  // Adapted from Clipper2 docs:
+  // http://www.angusj.com/clipper2/Docs/Units/Clipper/Types/JoinType.htm
+  // (Copyright © 2010-2023 Angus Johnson)
+  /**
+   * Specifies the treatment of path/contour joins (corners) when offseting
+   * CrossSections. See the [Clipper2
+   * doc](http://www.angusj.com/clipper2/Docs/Units/Clipper/Types/JoinType.htm)
+   * for illustrations.
+   */
+  enum class JoinType {
+    Square, /*!< Squaring is applied uniformly at all joins where the internal
+              join angle is less that 90 degrees. The squared edge will be at
+              exactly the offset distance from the join vertex. */
+    Round,  /*!< Rounding is applied to all joins that have convex external
+             angles, and it maintains the exact offset distance from the join
+             vertex. */
+    Miter   /*!< There's a necessary limit to mitered joins (to avoid narrow
+             angled joins producing excessively long and narrow
+             [spikes](http://www.angusj.com/clipper2/Docs/Units/Clipper.Offset/Classes/ClipperOffset/Properties/MiterLimit.htm)).
+             So where mitered joins would exceed a given maximum miter distance
+             (relative to the offset distance), these are 'squared' instead. */
+  };
+
+  CrossSection Offset(double delta, JoinType jt, double miter_limit = 2.0,
+                      int circularSegments = 0) const;
+  ///@}
+
+  /** @name Boolean
+   *  Combine two manifolds
+   */
+  ///@{
+  CrossSection Boolean(const CrossSection& second, OpType op) const;
+  static CrossSection BatchBoolean(
+      const std::vector<CrossSection>& crossSections, OpType op);
+  CrossSection operator+(const CrossSection&) const;
+  CrossSection& operator+=(const CrossSection&);
+  CrossSection operator-(const CrossSection&) const;
+  CrossSection& operator-=(const CrossSection&);
+  CrossSection operator^(const CrossSection&) const;
+  CrossSection& operator^=(const CrossSection&);
+  ///@}
+
+  /** @name Topological
+   */
+  ///@{
+  static CrossSection Compose(std::vector<CrossSection>&);
+  std::vector<CrossSection> Decompose() const;
+  ///@}
+
+  /** @name Convex Hulling
+   */
+  ///@{
+  CrossSection Hull() const;
+  static CrossSection Hull(const std::vector<CrossSection>& crossSections);
+  static CrossSection Hull(const SimplePolygon poly);
+  static CrossSection Hull(const Polygons polys);
+  ///@}
+  ///
+  /** @name Conversion
+   */
+  ///@{
+  Polygons ToPolygons() const;
+  ///@}
+
+ private:
+  mutable std::shared_ptr<const PathImpl> paths_;
+  mutable glm::mat3x2 transform_ = glm::mat3x2(1.0f);
+  CrossSection(std::shared_ptr<const PathImpl> paths);
+  std::shared_ptr<const PathImpl> GetPaths() const;
+};
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/cross_section/src/cross_section.cpp b/thirdparty/manifold/src/cross_section/src/cross_section.cpp
new file mode 100644
index 000000000000..a7bb15c29d59
--- /dev/null
+++ b/thirdparty/manifold/src/cross_section/src/cross_section.cpp
@@ -0,0 +1,795 @@
+// Copyright 2023 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "cross_section.h"
+
+#include "clipper2/clipper.core.h"
+#include "clipper2/clipper.h"
+#include "clipper2/clipper.offset.h"
+
+namespace C2 = Clipper2Lib;
+
+using namespace manifold;
+
+namespace manifold {
+struct PathImpl {
+  PathImpl(const C2::PathsD paths_) : paths_(paths_) {}
+  operator const C2::PathsD&() const { return paths_; }
+  const C2::PathsD paths_;
+};
+}  // namespace manifold
+
+namespace {
+const int precision_ = 8;
+
+C2::ClipType cliptype_of_op(OpType op) {
+  C2::ClipType ct = C2::ClipType::Union;
+  switch (op) {
+    case OpType::Add:
+      break;
+    case OpType::Subtract:
+      ct = C2::ClipType::Difference;
+      break;
+    case OpType::Intersect:
+      ct = C2::ClipType::Intersection;
+      break;
+  };
+  return ct;
+}
+
+C2::FillRule fr(CrossSection::FillRule fillrule) {
+  C2::FillRule fr = C2::FillRule::EvenOdd;
+  switch (fillrule) {
+    case CrossSection::FillRule::EvenOdd:
+      break;
+    case CrossSection::FillRule::NonZero:
+      fr = C2::FillRule::NonZero;
+      break;
+    case CrossSection::FillRule::Positive:
+      fr = C2::FillRule::Positive;
+      break;
+    case CrossSection::FillRule::Negative:
+      fr = C2::FillRule::Negative;
+      break;
+  };
+  return fr;
+}
+
+C2::JoinType jt(CrossSection::JoinType jointype) {
+  C2::JoinType jt = C2::JoinType::Square;
+  switch (jointype) {
+    case CrossSection::JoinType::Square:
+      break;
+    case CrossSection::JoinType::Round:
+      jt = C2::JoinType::Round;
+      break;
+    case CrossSection::JoinType::Miter:
+      jt = C2::JoinType::Miter;
+      break;
+  };
+  return jt;
+}
+
+glm::vec2 v2_of_pd(const C2::PointD p) { return {p.x, p.y}; }
+
+C2::PointD v2_to_pd(const glm::vec2 v) { return C2::PointD(v.x, v.y); }
+
+C2::PathD pathd_of_contour(const SimplePolygon& ctr) {
+  auto p = C2::PathD();
+  p.reserve(ctr.size());
+  for (auto v : ctr) {
+    p.push_back(v2_to_pd(v));
+  }
+  return p;
+}
+
+C2::PathsD transform(const C2::PathsD ps, const glm::mat3x2 m) {
+  const bool invert = glm::determinant(glm::mat2(m)) < 0;
+  auto transformed = C2::PathsD();
+  transformed.reserve(ps.size());
+  for (auto path : ps) {
+    auto sz = path.size();
+    auto s = C2::PathD(sz);
+    for (int i = 0; i < sz; ++i) {
+      auto idx = invert ? sz - 1 - i : i;
+      s[idx] = v2_to_pd(m * glm::vec3(path[i].x, path[i].y, 1));
+    }
+    transformed.push_back(s);
+  }
+  return transformed;
+}
+
+std::shared_ptr<const PathImpl> shared_paths(const C2::PathsD& ps) {
+  return std::make_shared<const PathImpl>(ps);
+}
+
+// forward declaration for mutual recursion
+void decompose_hole(const C2::PolyTreeD* outline,
+                    std::vector<C2::PathsD>& polys, C2::PathsD& poly,
+                    int n_holes, int j);
+
+void decompose_outline(const C2::PolyTreeD* tree,
+                       std::vector<C2::PathsD>& polys, int i) {
+  auto n_outlines = tree->Count();
+  if (i < n_outlines) {
+    auto outline = tree->Child(i);
+    auto n_holes = outline->Count();
+    auto poly = C2::PathsD(n_holes + 1);
+    poly[0] = outline->Polygon();
+    decompose_hole(outline, polys, poly, n_holes, 0);
+    polys.push_back(poly);
+    if (i < n_outlines - 1) {
+      decompose_outline(tree, polys, i + 1);
+    }
+  }
+}
+
+void decompose_hole(const C2::PolyTreeD* outline,
+                    std::vector<C2::PathsD>& polys, C2::PathsD& poly,
+                    int n_holes, int j) {
+  if (j < n_holes) {
+    auto child = outline->Child(j);
+    decompose_outline(child, polys, 0);
+    poly[j + 1] = child->Polygon();
+    decompose_hole(outline, polys, poly, n_holes, j + 1);
+  }
+}
+
+void flatten(const C2::PolyTreeD* tree, C2::PathsD& polys, int i) {
+  auto n_outlines = tree->Count();
+  if (i < n_outlines) {
+    auto outline = tree->Child(i);
+    flatten(outline, polys, 0);
+    polys.push_back(outline->Polygon());
+    if (i < n_outlines - 1) {
+      flatten(tree, polys, i + 1);
+    }
+  }
+}
+
+bool V2Lesser(glm::vec2 a, glm::vec2 b) {
+  if (a.x == b.x) return a.y < b.y;
+  return a.x < b.x;
+}
+
+void HullBacktrack(const glm::vec2& pt, std::vector<glm::vec2>& stack) {
+  auto sz = stack.size();
+  while (sz >= 2 && CCW(stack[sz - 2], stack[sz - 1], pt, 0.0f) <= 0.0f) {
+    stack.pop_back();
+    sz = stack.size();
+  }
+}
+
+// Based on method described here:
+// https://www.hackerearth.com/practice/math/geometry/line-sweep-technique/tutorial/
+// Changed to follow:
+// https://en.wikibooks.org/wiki/Algorithm_Implementation/Geometry/Convex_hull/Monotone_chain
+// This is the same algorithm (Andrew, also called Montone Chain).
+C2::PathD HullImpl(SimplePolygon& pts) {
+  int len = pts.size();
+  if (len < 3) return C2::PathD();  // not enough points to create a polygon
+  std::sort(pts.begin(), pts.end(), V2Lesser);
+
+  auto lower = std::vector<glm::vec2>{};
+  for (int i = 0; i < len; i++) {
+    HullBacktrack(pts[i], lower);
+    lower.push_back(pts[i]);
+  }
+  auto upper = std::vector<glm::vec2>{};
+  for (int i = len - 1; i >= 0; i--) {
+    HullBacktrack(pts[i], upper);
+    upper.push_back(pts[i]);
+  }
+
+  upper.pop_back();
+  lower.pop_back();
+
+  auto path = C2::PathD(lower.size() + upper.size());
+  for (int i = 0; i < lower.size(); i++) {
+    path[i] = v2_to_pd(lower[i]);
+  }
+  auto llen = lower.size();
+  int sz = upper.size();  // "fix" -Waggressive-loop-optimizations warning.
+  for (int i = 0; i < sz; i++) {
+    path[i + llen] = v2_to_pd(upper[i]);
+  }
+  return path;
+}
+}  // namespace
+
+namespace manifold {
+
+/**
+ * The default constructor is an empty cross-section (containing no contours).
+ */
+CrossSection::CrossSection() {
+  paths_ = std::make_shared<const PathImpl>(C2::PathsD());
+}
+
+CrossSection::~CrossSection() = default;
+CrossSection::CrossSection(CrossSection&&) noexcept = default;
+CrossSection& CrossSection::operator=(CrossSection&&) noexcept = default;
+
+/**
+ * The copy constructor avoids copying the underlying paths vector (sharing
+ * with its parent via shared_ptr), however subsequent transformations, and
+ * their application will not be shared. It is generally recommended to avoid
+ * this, opting instead to simply create CrossSections with the available
+ * const methods.
+ */
+CrossSection::CrossSection(const CrossSection& other) {
+  paths_ = other.paths_;
+  transform_ = other.transform_;
+}
+
+CrossSection& CrossSection::operator=(const CrossSection& other) {
+  if (this != &other) {
+    paths_ = other.paths_;
+    transform_ = other.transform_;
+  }
+  return *this;
+};
+
+// Private, skips unioning.
+CrossSection::CrossSection(std::shared_ptr<const PathImpl> ps) { paths_ = ps; }
+
+/**
+ * Create a 2d cross-section from a single contour. A boolean union operation
+ * (with Positive filling rule by default) is performed to ensure the
+ * resulting CrossSection is free of self-intersections.
+ *
+ * @param contour A closed path outlining the desired cross-section.
+ * @param fillrule The filling rule used to interpret polygon sub-regions
+ * created by self-intersections in contour.
+ */
+CrossSection::CrossSection(const SimplePolygon& contour, FillRule fillrule) {
+  auto ps = C2::PathsD{(pathd_of_contour(contour))};
+  paths_ = shared_paths(C2::Union(ps, fr(fillrule), precision_));
+}
+
+/**
+ * Create a 2d cross-section from a set of contours (complex polygons). A
+ * boolean union operation (with Positive filling rule by default) is
+ * performed to combine overlapping polygons and ensure the resulting
+ * CrossSection is free of intersections.
+ *
+ * @param contours A set of closed paths describing zero or more complex
+ * polygons.
+ * @param fillrule The filling rule used to interpret polygon sub-regions in
+ * contours.
+ */
+CrossSection::CrossSection(const Polygons& contours, FillRule fillrule) {
+  auto ps = C2::PathsD();
+  ps.reserve(contours.size());
+  for (auto ctr : contours) {
+    ps.push_back(pathd_of_contour(ctr));
+  }
+  paths_ = shared_paths(C2::Union(ps, fr(fillrule), precision_));
+}
+
+/**
+ * Create a 2d cross-section from an axis-aligned rectangle (bounding box).
+ *
+ * @param rect An axis-aligned rectangular bounding box.
+ */
+CrossSection::CrossSection(const Rect& rect) {
+  C2::PathD p(4);
+  p[0] = C2::PointD(rect.min.x, rect.min.y);
+  p[1] = C2::PointD(rect.max.x, rect.min.y);
+  p[2] = C2::PointD(rect.max.x, rect.max.y);
+  p[3] = C2::PointD(rect.min.x, rect.max.y);
+  paths_ = shared_paths(C2::PathsD{p});
+}
+
+// Private
+// All access to paths_ should be done through the GetPaths() method, which
+// applies the accumulated transform_
+std::shared_ptr<const PathImpl> CrossSection::GetPaths() const {
+  if (transform_ == glm::mat3x2(1.0f)) {
+    return paths_;
+  }
+  paths_ = shared_paths(::transform(paths_->paths_, transform_));
+  transform_ = glm::mat3x2(1.0f);
+  return paths_;
+}
+
+/**
+ * Constructs a square with the given XY dimensions. By default it is
+ * positioned in the first quadrant, touching the origin. If any dimensions in
+ * size are negative, or if all are zero, an empty Manifold will be returned.
+ *
+ * @param size The X, and Y dimensions of the square.
+ * @param center Set to true to shift the center to the origin.
+ */
+CrossSection CrossSection::Square(const glm::vec2 size, bool center) {
+  if (size.x < 0.0f || size.y < 0.0f || glm::length(size) == 0.0f) {
+    return CrossSection();
+  }
+
+  auto p = C2::PathD(4);
+  if (center) {
+    const auto w = size.x / 2;
+    const auto h = size.y / 2;
+    p[0] = C2::PointD(w, h);
+    p[1] = C2::PointD(-w, h);
+    p[2] = C2::PointD(-w, -h);
+    p[3] = C2::PointD(w, -h);
+  } else {
+    const double x = size.x;
+    const double y = size.y;
+    p[0] = C2::PointD(0.0, 0.0);
+    p[1] = C2::PointD(x, 0.0);
+    p[2] = C2::PointD(x, y);
+    p[3] = C2::PointD(0.0, y);
+  }
+  return CrossSection(shared_paths(C2::PathsD{p}));
+}
+
+/**
+ * Constructs a circle of a given radius.
+ *
+ * @param radius Radius of the circle. Must be positive.
+ * @param circularSegments Number of segments along its diameter. Default is
+ * calculated by the static Quality defaults according to the radius.
+ */
+CrossSection CrossSection::Circle(float radius, int circularSegments) {
+  if (radius <= 0.0f) {
+    return CrossSection();
+  }
+  int n = circularSegments > 2 ? circularSegments
+                               : Quality::GetCircularSegments(radius);
+  float dPhi = 360.0f / n;
+  auto circle = C2::PathD(n);
+  for (int i = 0; i < n; ++i) {
+    circle[i] = C2::PointD(radius * cosd(dPhi * i), radius * sind(dPhi * i));
+  }
+  return CrossSection(shared_paths(C2::PathsD{circle}));
+}
+
+/**
+ * Perform the given boolean operation between this and another CrossSection.
+ */
+CrossSection CrossSection::Boolean(const CrossSection& second,
+                                   OpType op) const {
+  auto ct = cliptype_of_op(op);
+  auto res = C2::BooleanOp(ct, C2::FillRule::Positive, GetPaths()->paths_,
+                           second.GetPaths()->paths_, precision_);
+  return CrossSection(shared_paths(res));
+}
+
+/**
+ * Perform the given boolean operation on a list of CrossSections. In case of
+ * Subtract, all CrossSections in the tail are differenced from the head.
+ */
+CrossSection CrossSection::BatchBoolean(
+    const std::vector<CrossSection>& crossSections, OpType op) {
+  if (crossSections.size() == 0)
+    return CrossSection();
+  else if (crossSections.size() == 1)
+    return crossSections[0];
+
+  auto subjs = crossSections[0].GetPaths();
+  int n_clips = 0;
+  for (int i = 1; i < crossSections.size(); ++i) {
+    n_clips += crossSections[i].GetPaths()->paths_.size();
+  }
+  auto clips = C2::PathsD();
+  clips.reserve(n_clips);
+  for (int i = 1; i < crossSections.size(); ++i) {
+    auto ps = crossSections[i].GetPaths();
+    clips.insert(clips.end(), ps->paths_.begin(), ps->paths_.end());
+  }
+
+  auto ct = cliptype_of_op(op);
+  auto res = C2::BooleanOp(ct, C2::FillRule::Positive, subjs->paths_, clips,
+                           precision_);
+  return CrossSection(shared_paths(res));
+}
+
+/**
+ * Compute the boolean union between two cross-sections.
+ */
+CrossSection CrossSection::operator+(const CrossSection& Q) const {
+  return Boolean(Q, OpType::Add);
+}
+
+/**
+ * Compute the boolean union between two cross-sections, assigning the result
+ * to the first.
+ */
+CrossSection& CrossSection::operator+=(const CrossSection& Q) {
+  *this = *this + Q;
+  return *this;
+}
+
+/**
+ * Compute the boolean difference of a (clip) cross-section from another
+ * (subject).
+ */
+CrossSection CrossSection::operator-(const CrossSection& Q) const {
+  return Boolean(Q, OpType::Subtract);
+}
+
+/**
+ * Compute the boolean difference of a (clip) cross-section from a another
+ * (subject), assigning the result to the subject.
+ */
+CrossSection& CrossSection::operator-=(const CrossSection& Q) {
+  *this = *this - Q;
+  return *this;
+}
+
+/**
+ * Compute the boolean intersection between two cross-sections.
+ */
+CrossSection CrossSection::operator^(const CrossSection& Q) const {
+  return Boolean(Q, OpType::Intersect);
+}
+
+/**
+ * Compute the boolean intersection between two cross-sections, assigning the
+ * result to the first.
+ */
+CrossSection& CrossSection::operator^=(const CrossSection& Q) {
+  *this = *this ^ Q;
+  return *this;
+}
+
+/**
+ * Construct a CrossSection from a vector of other CrossSections (batch
+ * boolean union).
+ */
+CrossSection CrossSection::Compose(std::vector<CrossSection>& crossSections) {
+  return BatchBoolean(crossSections, OpType::Add);
+}
+
+/**
+ * This operation returns a vector of CrossSections that are topologically
+ * disconnected, each containing one outline contour with zero or more
+ * holes.
+ */
+std::vector<CrossSection> CrossSection::Decompose() const {
+  if (NumContour() < 2) {
+    return std::vector<CrossSection>{CrossSection(*this)};
+  }
+
+  C2::PolyTreeD tree;
+  C2::BooleanOp(C2::ClipType::Union, C2::FillRule::Positive, GetPaths()->paths_,
+                C2::PathsD(), tree, precision_);
+
+  auto polys = std::vector<C2::PathsD>();
+  decompose_outline(&tree, polys, 0);
+
+  auto n_polys = polys.size();
+  auto comps = std::vector<CrossSection>(n_polys);
+  // reverse the stack while wrapping
+  for (int i = 0; i < n_polys; ++i) {
+    comps[n_polys - i - 1] = CrossSection(shared_paths(polys[i]));
+  }
+
+  return comps;
+}
+
+/**
+ * Move this CrossSection in space. This operation can be chained. Transforms
+ * are combined and applied lazily.
+ *
+ * @param v The vector to add to every vertex.
+ */
+CrossSection CrossSection::Translate(const glm::vec2 v) const {
+  glm::mat3x2 m(1.0f, 0.0f,  //
+                0.0f, 1.0f,  //
+                v.x, v.y);
+  return Transform(m);
+}
+
+/**
+ * Applies a (Z-axis) rotation to the CrossSection, in degrees. This operation
+ * can be chained. Transforms are combined and applied lazily.
+ *
+ * @param degrees degrees about the Z-axis to rotate.
+ */
+CrossSection CrossSection::Rotate(float degrees) const {
+  auto s = sind(degrees);
+  auto c = cosd(degrees);
+  glm::mat3x2 m(c, s,   //
+                -s, c,  //
+                0.0f, 0.0f);
+  return Transform(m);
+}
+
+/**
+ * Scale this CrossSection in space. This operation can be chained. Transforms
+ * are combined and applied lazily.
+ *
+ * @param v The vector to multiply every vertex by per component.
+ */
+CrossSection CrossSection::Scale(const glm::vec2 scale) const {
+  glm::mat3x2 m(scale.x, 0.0f,  //
+                0.0f, scale.y,  //
+                0.0f, 0.0f);
+  return Transform(m);
+}
+
+/**
+ * Mirror this CrossSection over the arbitrary axis described by the unit form
+ * of the given vector. If the length of the vector is zero, an empty
+ * CrossSection is returned. This operation can be chained. Transforms are
+ * combined and applied lazily.
+ *
+ * @param ax the axis to be mirrored over
+ */
+CrossSection CrossSection::Mirror(const glm::vec2 ax) const {
+  if (glm::length(ax) == 0.) {
+    return CrossSection();
+  }
+  auto n = glm::normalize(glm::abs(ax));
+  auto m = glm::mat3x2(glm::mat2(1.0f) - 2.0f * glm::outerProduct(n, n));
+  return Transform(m);
+}
+
+/**
+ * Transform this CrossSection in space. The first two columns form a 2x2
+ * matrix transform and the last is a translation vector. This operation can
+ * be chained. Transforms are combined and applied lazily.
+ *
+ * @param m The affine transform matrix to apply to all the vertices.
+ */
+CrossSection CrossSection::Transform(const glm::mat3x2& m) const {
+  auto transformed = CrossSection();
+  transformed.transform_ = m * glm::mat3(transform_);
+  transformed.paths_ = paths_;
+  return transformed;
+}
+
+/**
+ * Move the vertices of this CrossSection (creating a new one) according to
+ * any arbitrary input function, followed by a union operation (with a
+ * Positive fill rule) that ensures any introduced intersections are not
+ * included in the result.
+ *
+ * @param warpFunc A function that modifies a given vertex position.
+ */
+CrossSection CrossSection::Warp(
+    std::function<void(glm::vec2&)> warpFunc) const {
+  return WarpBatch([&warpFunc](VecView<glm::vec2> vecs) {
+    for (glm::vec2& p : vecs) {
+      warpFunc(p);
+    }
+  });
+}
+
+/**
+ * Same as CrossSection::Warp but calls warpFunc with
+ * a VecView which is roughly equivalent to std::span
+ * pointing to all vec2 elements to be modified in-place
+ *
+ * @param warpFunc A function that modifies multiple vertex positions.
+ */
+CrossSection CrossSection::WarpBatch(
+    std::function<void(VecView<glm::vec2>)> warpFunc) const {
+  std::vector<glm::vec2> tmp_verts;
+  C2::PathsD paths = GetPaths()->paths_;  // deep copy
+  for (C2::PathD const& path : paths) {
+    for (C2::PointD const& p : path) {
+      tmp_verts.push_back(v2_of_pd(p));
+    }
+  }
+
+  warpFunc(VecView<glm::vec2>(tmp_verts.data(), tmp_verts.size()));
+
+  auto cursor = tmp_verts.begin();
+  for (C2::PathD& path : paths) {
+    for (C2::PointD& p : path) {
+      p = v2_to_pd(*cursor);
+      ++cursor;
+    }
+  }
+
+  return CrossSection(
+      shared_paths(C2::Union(paths, C2::FillRule::Positive, precision_)));
+}
+
+/**
+ * Remove vertices from the contours in this CrossSection that are less than
+ * the specified distance epsilon from an imaginary line that passes through
+ * its two adjacent vertices. Near duplicate vertices and collinear points
+ * will be removed at lower epsilons, with elimination of line segments
+ * becoming increasingly aggressive with larger epsilons.
+ *
+ * It is recommended to apply this function following Offset, in order to
+ * clean up any spurious tiny line segments introduced that do not improve
+ * quality in any meaningful way. This is particularly important if further
+ * offseting operations are to be performed, which would compound the issue.
+ */
+CrossSection CrossSection::Simplify(double epsilon) const {
+  C2::PolyTreeD tree;
+  C2::BooleanOp(C2::ClipType::Union, C2::FillRule::Positive, GetPaths()->paths_,
+                C2::PathsD(), tree, precision_);
+
+  C2::PathsD polys;
+  flatten(&tree, polys, 0);
+
+  // Filter out contours less than epsilon wide.
+  C2::PathsD filtered;
+  for (C2::PathD poly : polys) {
+    auto area = C2::Area(poly);
+    Rect box;
+    for (auto vert : poly) {
+      box.Union(glm::vec2(vert.x, vert.y));
+    }
+    glm::vec2 size = box.Size();
+    if (glm::abs(area) > glm::max(size.x, size.y) * epsilon) {
+      filtered.push_back(poly);
+    }
+  }
+
+  auto ps = SimplifyPaths(filtered, epsilon, true);
+  return CrossSection(shared_paths(ps));
+}
+
+/**
+ * Inflate the contours in CrossSection by the specified delta, handling
+ * corners according to the given JoinType.
+ *
+ * @param delta Positive deltas will cause the expansion of outlining contours
+ * to expand, and retraction of inner (hole) contours. Negative deltas will
+ * have the opposite effect.
+ * @param jt The join type specifying the treatment of contour joins
+ * (corners).
+ * @param miter_limit The maximum distance in multiples of delta that vertices
+ * can be offset from their original positions with before squaring is
+ * applied, <B>when the join type is Miter</B> (default is 2, which is the
+ * minimum allowed). See the [Clipper2
+ * MiterLimit](http://www.angusj.com/clipper2/Docs/Units/Clipper.Offset/Classes/ClipperOffset/Properties/MiterLimit.htm)
+ * page for a visual example.
+ * @param circularSegments Number of segments per 360 degrees of
+ * <B>JoinType::Round</B> corners (roughly, the number of vertices that
+ * will be added to each contour). Default is calculated by the static Quality
+ * defaults according to the radius.
+ */
+CrossSection CrossSection::Offset(double delta, JoinType jointype,
+                                  double miter_limit,
+                                  int circularSegments) const {
+  double arc_tol = 0.;
+  if (jointype == JoinType::Round) {
+    int n = circularSegments > 2 ? circularSegments
+                                 : Quality::GetCircularSegments(delta);
+    // This calculates tolerance as a function of circular segments and delta
+    // (radius) in order to get back the same number of segments in Clipper2:
+    // steps_per_360 = PI / acos(1 - arc_tol / abs_delta)
+    const double abs_delta = std::fabs(delta);
+    const double scaled_delta = abs_delta * std::pow(10, precision_);
+    arc_tol = (std::cos(Clipper2Lib::PI / n) - 1) * -scaled_delta;
+  }
+  auto ps =
+      C2::InflatePaths(GetPaths()->paths_, delta, jt(jointype),
+                       C2::EndType::Polygon, miter_limit, precision_, arc_tol);
+  return CrossSection(shared_paths(ps));
+}
+
+/**
+ * Compute the convex hull enveloping a set of cross-sections.
+ *
+ * @param crossSections A vector of cross-sections over which to compute a
+ * convex hull.
+ */
+CrossSection CrossSection::Hull(
+    const std::vector<CrossSection>& crossSections) {
+  int n = 0;
+  for (auto cs : crossSections) n += cs.NumVert();
+  SimplePolygon pts;
+  pts.reserve(n);
+  for (auto cs : crossSections) {
+    auto paths = cs.GetPaths()->paths_;
+    for (auto path : paths) {
+      for (auto p : path) {
+        pts.push_back(v2_of_pd(p));
+      }
+    }
+  }
+  return CrossSection(shared_paths(C2::PathsD{HullImpl(pts)}));
+}
+
+/**
+ * Compute the convex hull of this cross-section.
+ */
+CrossSection CrossSection::Hull() const {
+  return Hull(std::vector<CrossSection>{*this});
+}
+
+/**
+ * Compute the convex hull of a set of points. If the given points are fewer
+ * than 3, an empty CrossSection will be returned.
+ *
+ * @param pts A vector of 2-dimensional points over which to compute a convex
+ * hull.
+ */
+CrossSection CrossSection::Hull(SimplePolygon pts) {
+  return CrossSection(shared_paths(C2::PathsD{HullImpl(pts)}));
+}
+
+/**
+ * Compute the convex hull of a set of points/polygons. If the given points are
+ * fewer than 3, an empty CrossSection will be returned.
+ *
+ * @param pts A vector of vectors of 2-dimensional points over which to compute
+ * a convex hull.
+ */
+CrossSection CrossSection::Hull(const Polygons polys) {
+  SimplePolygon pts;
+  for (auto poly : polys) {
+    for (auto p : poly) {
+      pts.push_back(p);
+    }
+  }
+  return Hull(pts);
+}
+
+/**
+ * Return the total area covered by complex polygons making up the
+ * CrossSection.
+ */
+double CrossSection::Area() const { return C2::Area(GetPaths()->paths_); }
+
+/**
+ * Return the number of vertices in the CrossSection.
+ */
+int CrossSection::NumVert() const {
+  int n = 0;
+  auto paths = GetPaths()->paths_;
+  for (auto p : paths) {
+    n += p.size();
+  }
+  return n;
+}
+
+/**
+ * Return the number of contours (both outer and inner paths) in the
+ * CrossSection.
+ */
+int CrossSection::NumContour() const { return GetPaths()->paths_.size(); }
+
+/**
+ * Does the CrossSection contain any contours?
+ */
+bool CrossSection::IsEmpty() const { return GetPaths()->paths_.empty(); }
+
+/**
+ * Returns the axis-aligned bounding rectangle of all the CrossSections'
+ * vertices.
+ */
+Rect CrossSection::Bounds() const {
+  auto r = C2::GetBounds(GetPaths()->paths_);
+  return Rect({r.left, r.bottom}, {r.right, r.top});
+}
+
+/**
+ * Return the contours of this CrossSection as a Polygons.
+ */
+Polygons CrossSection::ToPolygons() const {
+  auto polys = Polygons();
+  auto paths = GetPaths()->paths_;
+  polys.reserve(paths.size());
+  for (auto p : paths) {
+    auto sp = SimplePolygon();
+    sp.reserve(p.size());
+    for (auto v : p) {
+      sp.push_back({v.x, v.y});
+    }
+    polys.push_back(sp);
+  }
+  return polys;
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/include/manifold.h b/thirdparty/manifold/src/manifold/include/manifold.h
new file mode 100644
index 000000000000..6d8772bc9fd9
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/include/manifold.h
@@ -0,0 +1,311 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <functional>
+#include <memory>
+
+#include "cross_section.h"
+#include "public.h"
+#include "vec_view.h"
+
+namespace manifold {
+
+/**
+ * @ingroup Debug
+ *
+ * Allows modification of the assertions checked in MANIFOLD_DEBUG mode.
+ *
+ * @return ExecutionParams&
+ */
+ExecutionParams& ManifoldParams();
+
+class CsgNode;
+class CsgLeafNode;
+
+/** @ingroup Connections
+ *  @{
+ */
+
+/**
+ * An alternative to Mesh for output suitable for pushing into graphics
+ * libraries directly. This may not be manifold since the verts are duplicated
+ * along property boundaries that do not match. The additional merge vectors
+ * store this missing information, allowing the manifold to be reconstructed.
+ */
+struct MeshGL {
+  /// Number of property vertices
+  uint32_t NumVert() const {
+    if (vertProperties.size() / numProp >= static_cast<std::vector<float>::size_type>(std::numeric_limits<int>::max()))
+      throw std::out_of_range("mesh too large");
+    return vertProperties.size() / numProp;
+  };
+  /// Number of triangles
+  uint32_t NumTri() const {
+    if (vertProperties.size() / numProp >= static_cast<std::vector<float>::size_type>(std::numeric_limits<int>::max()))
+      throw std::out_of_range("mesh too large");
+    return triVerts.size() / 3;
+  };
+
+  /// Number of properties per vertex, always >= 3.
+  uint32_t numProp = 3;
+  /// Flat, GL-style interleaved list of all vertex properties: propVal =
+  /// vertProperties[vert * numProp + propIdx]. The first three properties are
+  /// always the position x, y, z.
+  std::vector<float> vertProperties;
+  /// The vertex indices of the three triangle corners in CCW (from the outside)
+  /// order, for each triangle.
+  std::vector<uint32_t> triVerts;
+  /// Optional: A list of only the vertex indicies that need to be merged to
+  /// reconstruct the manifold.
+  std::vector<uint32_t> mergeFromVert;
+  /// Optional: The same length as mergeFromVert, and the corresponding value
+  /// contains the vertex to merge with. It will have an identical position, but
+  /// the other properties may differ.
+  std::vector<uint32_t> mergeToVert;
+  /// Optional: Indicates runs of triangles that correspond to a particular
+  /// input mesh instance. The runs encompass all of triVerts and are sorted
+  /// by runOriginalID. Run i begins at triVerts[runIndex[i]] and ends at
+  /// triVerts[runIndex[i+1]]. All runIndex values are divisible by 3.
+  std::vector<uint32_t> runIndex;
+  /// Optional: The OriginalID of the mesh this triangle run came from. This ID
+  /// is ideal for reapplying materials to the output mesh. Multiple runs may
+  /// have the same ID, e.g. representing different copies of the same input
+  /// mesh. If you create an input MeshGL that you want to be able to reference
+  /// as one or more originals, be sure to set unique values from ReserveIDs().
+  std::vector<uint32_t> runOriginalID;
+  /// Optional: For each run, a 3x4 transform is stored representing how the
+  /// corresponding original mesh was transformed to create this triangle run.
+  /// This matrix is stored in column-major order and the length of the overall
+  /// vector is 12 * runOriginalID.size().
+  std::vector<float> runTransform;
+  /// Optional: Length NumTri, contains an ID of the source face this triangle
+  /// comes from. When auto-generated, this ID will be a triangle index into the
+  /// original mesh. All neighboring coplanar triangles from that input mesh
+  /// will refer to a single triangle of that group as the faceID. When
+  /// supplying faceIDs, ensure that triangles with the same ID are in fact
+  /// coplanar and have consistent properties (within some tolerance) or the
+  /// output will be surprising.
+  std::vector<uint32_t> faceID;
+  /// Optional: The X-Y-Z-W weighted tangent vectors for smooth Refine(). If
+  /// non-empty, must be exactly four times as long as Mesh.triVerts. Indexed
+  /// as 4 * (3 * tri + i) + j, i < 3, j < 4, representing the tangent value
+  /// Mesh.triVerts[tri][i] along the CCW edge. If empty, mesh is faceted.
+  std::vector<float> halfedgeTangent;
+  /// The absolute precision of the vertex positions, based on accrued rounding
+  /// errors. When creating a Manifold, the precision used will be the maximum
+  /// of this and a baseline precision from the size of the bounding box. Any
+  /// edge shorter than precision may be collapsed.
+  float precision = 0;
+
+  MeshGL() = default;
+  MeshGL(const Mesh& mesh);
+
+  bool Merge();
+};
+/** @} */
+
+/** @defgroup Core
+ *  @brief The central classes of the library
+ *  @{
+ */
+
+/**
+ * This library's internal representation of an oriented, 2-manifold, triangle
+ * mesh - a simple boundary-representation of a solid object. Use this class to
+ * store and operate on solids, and use MeshGL for input and output, or
+ * potentially Mesh if only basic geometry is required.
+ *
+ * In addition to storing geometric data, a Manifold can also store an arbitrary
+ * number of vertex properties. These could be anything, e.g. normals, UV
+ * coordinates, colors, etc, but this library is completely agnostic. All
+ * properties are merely float values indexed by channel number. It is up to the
+ * user to associate channel numbers with meaning.
+ *
+ * Manifold allows vertex properties to be shared for efficient storage, or to
+ * have multiple property verts associated with a single geometric vertex,
+ * allowing sudden property changes, e.g. at Boolean intersections, without
+ * sacrificing manifoldness.
+ *
+ * Manifolds also keep track of their relationships to their inputs, via
+ * OriginalIDs and the faceIDs and transforms accessible through MeshGL. This
+ * allows object-level properties to be re-associated with the output after many
+ * operations, particularly useful for materials. Since separate object's
+ * properties are not mixed, there is no requirement that channels have
+ * consistent meaning between different inputs.
+ */
+class Manifold {
+ public:
+  /** @name Creation
+   *  Constructors
+   */
+  ///@{
+  Manifold();
+  ~Manifold();
+  Manifold(const Manifold& other);
+  Manifold& operator=(const Manifold& other);
+  Manifold(Manifold&&) noexcept;
+  Manifold& operator=(Manifold&&) noexcept;
+
+  Manifold(const MeshGL&, const std::vector<float>& propertyTolerance = {});
+  Manifold(const Mesh&);
+
+  static Manifold Smooth(const MeshGL&,
+                         const std::vector<Smoothness>& sharpenedEdges = {});
+  static Manifold Smooth(const Mesh&,
+                         const std::vector<Smoothness>& sharpenedEdges = {});
+  static Manifold Tetrahedron();
+  static Manifold Cube(glm::vec3 size = glm::vec3(1.0f), bool center = false);
+  static Manifold Cylinder(float height, float radiusLow,
+                           float radiusHigh = -1.0f, int circularSegments = 0,
+                           bool center = false);
+  static Manifold Sphere(float radius, int circularSegments = 0);
+  static Manifold Extrude(const CrossSection& crossSection, float height,
+                          int nDivisions = 0, float twistDegrees = 0.0f,
+                          glm::vec2 scaleTop = glm::vec2(1.0f));
+  static Manifold Revolve(const CrossSection& crossSection,
+                          int circularSegments = 0,
+                          float revolveDegrees = 360.0f);
+  ///@}
+
+  /** @name Topological
+   *  No geometric calculations.
+   */
+  ///@{
+  static Manifold Compose(const std::vector<Manifold>&);
+  std::vector<Manifold> Decompose() const;
+  ///@}
+
+  /** @name Information
+   *  Details of the manifold
+   */
+  ///@{
+  Mesh GetMesh() const;
+  MeshGL GetMeshGL(glm::ivec3 normalIdx = glm::ivec3(0)) const;
+  bool IsEmpty() const;
+  enum class Error {
+    NoError,
+    NonFiniteVertex,
+    NotManifold,
+    VertexOutOfBounds,
+    PropertiesWrongLength,
+    MissingPositionProperties,
+    MergeVectorsDifferentLengths,
+    MergeIndexOutOfBounds,
+    TransformWrongLength,
+    RunIndexWrongLength,
+    FaceIDWrongLength,
+    InvalidConstruction,
+  };
+  Error Status() const;
+  int NumVert() const;
+  int NumEdge() const;
+  int NumTri() const;
+  int NumProp() const;
+  int NumPropVert() const;
+  Box BoundingBox() const;
+  float Precision() const;
+  int Genus() const;
+  Properties GetProperties() const;
+  float MinGap(const Manifold& other, float searchLength) const;
+  ///@}
+
+  /** @name Mesh ID
+   *  Details of the manifold's relation to its input meshes, for the purposes
+   * of reapplying mesh properties.
+   */
+  ///@{
+  int OriginalID() const;
+  Manifold AsOriginal() const;
+  static uint32_t ReserveIDs(uint32_t);
+  ///@}
+
+  /** @name Modification
+   */
+  ///@{
+  Manifold Translate(glm::vec3) const;
+  Manifold Scale(glm::vec3) const;
+  Manifold Rotate(float xDegrees, float yDegrees = 0.0f,
+                  float zDegrees = 0.0f) const;
+  Manifold Transform(const glm::mat4x3&) const;
+  Manifold Mirror(glm::vec3) const;
+  Manifold Warp(std::function<void(glm::vec3&)>) const;
+  Manifold WarpBatch(std::function<void(VecView<glm::vec3>)>) const;
+  Manifold SetProperties(
+      int, std::function<void(float*, glm::vec3, const float*)>) const;
+  Manifold CalculateCurvature(int gaussianIdx, int meanIdx) const;
+  Manifold CalculateNormals(int normalIdx, float minSharpAngle = 60) const;
+  Manifold SmoothByNormals(int normalIdx) const;
+  Manifold SmoothOut(float minSharpAngle = 60, float minSmoothness = 0) const;
+  Manifold Refine(int) const;
+  Manifold RefineToLength(float) const;
+  // Manifold RefineToPrecision(float);
+  ///@}
+
+  /** @name Boolean
+   *  Combine two manifolds
+   */
+  ///@{
+  Manifold Boolean(const Manifold& second, OpType op) const;
+  static Manifold BatchBoolean(const std::vector<Manifold>& manifolds,
+                               OpType op);
+  // Boolean operation shorthand
+  Manifold operator+(const Manifold&) const;  // Add (Union)
+  Manifold& operator+=(const Manifold&);
+  Manifold operator-(const Manifold&) const;  // Subtract (Difference)
+  Manifold& operator-=(const Manifold&);
+  Manifold operator^(const Manifold&) const;  // Intersect
+  Manifold& operator^=(const Manifold&);
+  std::pair<Manifold, Manifold> Split(const Manifold&) const;
+  std::pair<Manifold, Manifold> SplitByPlane(glm::vec3 normal,
+                                             float originOffset) const;
+  Manifold TrimByPlane(glm::vec3 normal, float originOffset) const;
+  ///@}
+
+  /** @name 2D from 3D
+   */
+  ///@{
+  CrossSection Slice(float height = 0) const;
+  CrossSection Project() const;
+  ///@}
+
+  /** @name Convex hull
+   */
+  ///@{
+  Manifold Hull() const;
+  static Manifold Hull(const std::vector<Manifold>& manifolds);
+  static Manifold Hull(const std::vector<glm::vec3>& pts);
+  ///@}
+
+  /** @name Testing hooks
+   *  These are just for internal testing.
+   */
+  ///@{
+  bool MatchesTriNormals() const;
+  int NumDegenerateTris() const;
+  int NumOverlaps(const Manifold& second) const;
+  ///@}
+
+  struct Impl;
+
+ private:
+  Manifold(std::shared_ptr<CsgNode> pNode_);
+  Manifold(std::shared_ptr<Impl> pImpl_);
+  static Manifold Invalid();
+  mutable std::shared_ptr<CsgNode> pNode_;
+
+  CsgLeafNode& GetCsgLeafNode() const;
+};
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/boolean3.cpp b/thirdparty/manifold/src/manifold/src/boolean3.cpp
new file mode 100644
index 000000000000..5ed8660d30f0
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/boolean3.cpp
@@ -0,0 +1,602 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "boolean3.h"
+
+#include <limits>
+
+#include "par.h"
+
+using namespace manifold;
+
+namespace {
+
+// These two functions (Interpolate and Intersect) are the only places where
+// floating-point operations take place in the whole Boolean function. These are
+// carefully designed to minimize rounding error and to eliminate it at edge
+// cases to ensure consistency.
+
+glm::vec2 Interpolate(glm::vec3 pL, glm::vec3 pR, float x) {
+  const float dxL = x - pL.x;
+  const float dxR = x - pR.x;
+  ASSERT(dxL * dxR <= 0, logicErr, "Boolean manifold error: not in domain");
+  const bool useL = fabs(dxL) < fabs(dxR);
+  const glm::vec3 dLR = pR - pL;
+  const float lambda = (useL ? dxL : dxR) / dLR.x;
+  if (!isfinite(lambda) || !isfinite(dLR.y) || !isfinite(dLR.z))
+    return glm::vec2(pL.y, pL.z);
+  glm::vec2 yz;
+  yz[0] = (useL ? pL.y : pR.y) + lambda * dLR.y;
+  yz[1] = (useL ? pL.z : pR.z) + lambda * dLR.z;
+  return yz;
+}
+
+glm::vec4 Intersect(const glm::vec3 &pL, const glm::vec3 &pR,
+                    const glm::vec3 &qL, const glm::vec3 &qR) {
+  const float dyL = qL.y - pL.y;
+  const float dyR = qR.y - pR.y;
+  ASSERT(dyL * dyR <= 0, logicErr, "Boolean manifold error: no intersection");
+  const bool useL = fabs(dyL) < fabs(dyR);
+  const float dx = pR.x - pL.x;
+  float lambda = (useL ? dyL : dyR) / (dyL - dyR);
+  if (!isfinite(lambda)) lambda = 0.0f;
+  glm::vec4 xyzz;
+  xyzz.x = (useL ? pL.x : pR.x) + lambda * dx;
+  const float pDy = pR.y - pL.y;
+  const float qDy = qR.y - qL.y;
+  const bool useP = fabs(pDy) < fabs(qDy);
+  xyzz.y = (useL ? (useP ? pL.y : qL.y) : (useP ? pR.y : qR.y)) +
+           lambda * (useP ? pDy : qDy);
+  xyzz.z = (useL ? pL.z : pR.z) + lambda * (pR.z - pL.z);
+  xyzz.w = (useL ? qL.z : qR.z) + lambda * (qR.z - qL.z);
+  return xyzz;
+}
+
+template <const bool inverted>
+struct CopyFaceEdges {
+  const SparseIndices &p1q1;
+  // const int *p1q1;
+  // x can be either vert or edge (0 or 1).
+  SparseIndices &pXq1;
+  VecView<const Halfedge> halfedgesQ;
+
+  void operator()(thrust::tuple<size_t, size_t> in) {
+    int idx = 3 * thrust::get<0>(in);
+    size_t i = thrust::get<1>(in);
+    int pX = p1q1.Get(i, inverted);
+    int q2 = p1q1.Get(i, !inverted);
+
+    for (const int j : {0, 1, 2}) {
+      const int q1 = 3 * q2 + j;
+      const Halfedge edge = halfedgesQ[q1];
+      int a = pX;
+      int b = edge.IsForward() ? q1 : edge.pairedHalfedge;
+      if (inverted) std::swap(a, b);
+      pXq1.Set(idx + static_cast<size_t>(j), a, b);
+    }
+  }
+};
+
+SparseIndices Filter11(const Manifold::Impl &inP, const Manifold::Impl &inQ,
+                       const SparseIndices &p1q2, const SparseIndices &p2q1) {
+  ZoneScoped;
+  SparseIndices p1q1(3 * p1q2.size() + 3 * p2q1.size());
+  for_each_n(autoPolicy(p1q2.size()), zip(countAt(0_z), countAt(0_z)),
+             p1q2.size(), CopyFaceEdges<false>({p1q2, p1q1, inQ.halfedge_}));
+  for_each_n(autoPolicy(p2q1.size()), zip(countAt(p1q2.size()), countAt(0_z)),
+             p2q1.size(), CopyFaceEdges<true>({p2q1, p1q1, inP.halfedge_}));
+  p1q1.Unique();
+  return p1q1;
+}
+
+inline bool Shadows(float p, float q, float dir) {
+  return p == q ? dir < 0 : p < q;
+}
+
+inline thrust::pair<int, glm::vec2> Shadow01(
+    const int p0, const int q1, VecView<const glm::vec3> vertPosP,
+    VecView<const glm::vec3> vertPosQ, VecView<const Halfedge> halfedgeQ,
+    const float expandP, VecView<const glm::vec3> normalP, const bool reverse) {
+  const int q1s = halfedgeQ[q1].startVert;
+  const int q1e = halfedgeQ[q1].endVert;
+  const float p0x = vertPosP[p0].x;
+  const float q1sx = vertPosQ[q1s].x;
+  const float q1ex = vertPosQ[q1e].x;
+  int s01 = reverse ? Shadows(q1sx, p0x, expandP * normalP[q1s].x) -
+                          Shadows(q1ex, p0x, expandP * normalP[q1e].x)
+                    : Shadows(p0x, q1ex, expandP * normalP[p0].x) -
+                          Shadows(p0x, q1sx, expandP * normalP[p0].x);
+  glm::vec2 yz01(NAN);
+
+  if (s01 != 0) {
+    yz01 = Interpolate(vertPosQ[q1s], vertPosQ[q1e], vertPosP[p0].x);
+    if (reverse) {
+      glm::vec3 diff = vertPosQ[q1s] - vertPosP[p0];
+      const float start2 = glm::dot(diff, diff);
+      diff = vertPosQ[q1e] - vertPosP[p0];
+      const float end2 = glm::dot(diff, diff);
+      const float dir = start2 < end2 ? normalP[q1s].y : normalP[q1e].y;
+      if (!Shadows(yz01[0], vertPosP[p0].y, expandP * dir)) s01 = 0;
+    } else {
+      if (!Shadows(vertPosP[p0].y, yz01[0], expandP * normalP[p0].y)) s01 = 0;
+    }
+  }
+  return thrust::make_pair(s01, yz01);
+}
+
+// https://github.com/scandum/binary_search/blob/master/README.md
+// much faster than standard binary search on large arrays
+size_t monobound_quaternary_search(VecView<const int64_t> array, int64_t key) {
+  if (array.size() == 0) {
+    return -1;
+  }
+  size_t bot = 0;
+  size_t top = array.size();
+  while (top >= 65536) {
+    size_t mid = top / 4;
+    top -= mid * 3;
+    if (key < array[bot + mid * 2]) {
+      if (key >= array[bot + mid]) {
+        bot += mid;
+      }
+    } else {
+      bot += mid * 2;
+      if (key >= array[bot + mid]) {
+        bot += mid;
+      }
+    }
+  }
+
+  while (top > 3) {
+    size_t mid = top / 2;
+    if (key >= array[bot + mid]) {
+      bot += mid;
+    }
+    top -= mid;
+  }
+
+  while (top--) {
+    if (key == array[bot + top]) {
+      return bot + top;
+    }
+  }
+  return -1;
+}
+
+struct Kernel11 {
+  VecView<const glm::vec3> vertPosP;
+  VecView<const glm::vec3> vertPosQ;
+  VecView<const Halfedge> halfedgeP;
+  VecView<const Halfedge> halfedgeQ;
+  float expandP;
+  VecView<const glm::vec3> normalP;
+  const SparseIndices &p1q1;
+
+  void operator()(thrust::tuple<size_t, glm::vec4 &, int &> inout) {
+    const int p1 = p1q1.Get(thrust::get<0>(inout), false);
+    const int q1 = p1q1.Get(thrust::get<0>(inout), true);
+    glm::vec4 &xyzz11 = thrust::get<1>(inout);
+    int &s11 = thrust::get<2>(inout);
+
+    // For pRL[k], qRL[k], k==0 is the left and k==1 is the right.
+    int k = 0;
+    glm::vec3 pRL[2], qRL[2];
+    // Either the left or right must shadow, but not both. This ensures the
+    // intersection is between the left and right.
+    bool shadows = false;
+    s11 = 0;
+
+    const int p0[2] = {halfedgeP[p1].startVert, halfedgeP[p1].endVert};
+    for (int i : {0, 1}) {
+      const auto syz01 = Shadow01(p0[i], q1, vertPosP, vertPosQ, halfedgeQ,
+                                  expandP, normalP, false);
+      const int s01 = syz01.first;
+      const glm::vec2 yz01 = syz01.second;
+      // If the value is NaN, then these do not overlap.
+      if (isfinite(yz01[0])) {
+        s11 += s01 * (i == 0 ? -1 : 1);
+        if (k < 2 && (k == 0 || (s01 != 0) != shadows)) {
+          shadows = s01 != 0;
+          pRL[k] = vertPosP[p0[i]];
+          qRL[k] = glm::vec3(pRL[k].x, yz01);
+          ++k;
+        }
+      }
+    }
+
+    const int q0[2] = {halfedgeQ[q1].startVert, halfedgeQ[q1].endVert};
+    for (int i : {0, 1}) {
+      const auto syz10 = Shadow01(q0[i], p1, vertPosQ, vertPosP, halfedgeP,
+                                  expandP, normalP, true);
+      const int s10 = syz10.first;
+      const glm::vec2 yz10 = syz10.second;
+      // If the value is NaN, then these do not overlap.
+      if (isfinite(yz10[0])) {
+        s11 += s10 * (i == 0 ? -1 : 1);
+        if (k < 2 && (k == 0 || (s10 != 0) != shadows)) {
+          shadows = s10 != 0;
+          qRL[k] = vertPosQ[q0[i]];
+          pRL[k] = glm::vec3(qRL[k].x, yz10);
+          ++k;
+        }
+      }
+    }
+
+    if (s11 == 0) {  // No intersection
+      xyzz11 = glm::vec4(NAN);
+    } else {
+      ASSERT(k == 2, logicErr, "Boolean manifold error: s11");
+      xyzz11 = Intersect(pRL[0], pRL[1], qRL[0], qRL[1]);
+
+      const int p1s = halfedgeP[p1].startVert;
+      const int p1e = halfedgeP[p1].endVert;
+      glm::vec3 diff = vertPosP[p1s] - glm::vec3(xyzz11);
+      const float start2 = glm::dot(diff, diff);
+      diff = vertPosP[p1e] - glm::vec3(xyzz11);
+      const float end2 = glm::dot(diff, diff);
+      const float dir = start2 < end2 ? normalP[p1s].z : normalP[p1e].z;
+
+      if (!Shadows(xyzz11.z, xyzz11.w, expandP * dir)) s11 = 0;
+    }
+  }
+};
+
+std::tuple<Vec<int>, Vec<glm::vec4>> Shadow11(SparseIndices &p1q1,
+                                              const Manifold::Impl &inP,
+                                              const Manifold::Impl &inQ,
+                                              float expandP) {
+  ZoneScoped;
+  Vec<int> s11(p1q1.size());
+  Vec<glm::vec4> xyzz11(p1q1.size());
+
+  for_each_n(autoPolicy(p1q1.size()),
+             zip(countAt(0_z), xyzz11.begin(), s11.begin()), p1q1.size(),
+             Kernel11({inP.vertPos_, inQ.vertPos_, inP.halfedge_, inQ.halfedge_,
+                       expandP, inP.vertNormal_, p1q1}));
+
+  p1q1.KeepFinite(xyzz11, s11);
+
+  return std::make_tuple(s11, xyzz11);
+};
+
+struct Kernel02 {
+  VecView<const glm::vec3> vertPosP;
+  VecView<const Halfedge> halfedgeQ;
+  VecView<const glm::vec3> vertPosQ;
+  const float expandP;
+  VecView<const glm::vec3> vertNormalP;
+  const SparseIndices &p0q2;
+  const bool forward;
+
+  void operator()(thrust::tuple<size_t, int &, float &> inout) {
+    const int p0 = p0q2.Get(thrust::get<0>(inout), !forward);
+    const int q2 = p0q2.Get(thrust::get<0>(inout), forward);
+    int &s02 = thrust::get<1>(inout);
+    float &z02 = thrust::get<2>(inout);
+
+    // For yzzLR[k], k==0 is the left and k==1 is the right.
+    int k = 0;
+    glm::vec3 yzzRL[2];
+    // Either the left or right must shadow, but not both. This ensures the
+    // intersection is between the left and right.
+    bool shadows = false;
+    int closestVert = -1;
+    float minMetric = std::numeric_limits<float>::infinity();
+    s02 = 0;
+
+    const glm::vec3 posP = vertPosP[p0];
+    for (const int i : {0, 1, 2}) {
+      const int q1 = 3 * q2 + i;
+      const Halfedge edge = halfedgeQ[q1];
+      const int q1F = edge.IsForward() ? q1 : edge.pairedHalfedge;
+
+      if (!forward) {
+        const int qVert = halfedgeQ[q1F].startVert;
+        const glm::vec3 diff = posP - vertPosQ[qVert];
+        const float metric = glm::dot(diff, diff);
+        if (metric < minMetric) {
+          minMetric = metric;
+          closestVert = qVert;
+        }
+      }
+
+      const auto syz01 = Shadow01(p0, q1F, vertPosP, vertPosQ, halfedgeQ,
+                                  expandP, vertNormalP, !forward);
+      const int s01 = syz01.first;
+      const glm::vec2 yz01 = syz01.second;
+      // If the value is NaN, then these do not overlap.
+      if (isfinite(yz01[0])) {
+        s02 += s01 * (forward == edge.IsForward() ? -1 : 1);
+        if (k < 2 && (k == 0 || (s01 != 0) != shadows)) {
+          shadows = s01 != 0;
+          yzzRL[k++] = glm::vec3(yz01[0], yz01[1], yz01[1]);
+        }
+      }
+    }
+
+    if (s02 == 0) {  // No intersection
+      z02 = NAN;
+    } else {
+      ASSERT(k == 2, logicErr, "Boolean manifold error: s02");
+      glm::vec3 vertPos = vertPosP[p0];
+      z02 = Interpolate(yzzRL[0], yzzRL[1], vertPos.y)[1];
+      if (forward) {
+        if (!Shadows(vertPos.z, z02, expandP * vertNormalP[p0].z)) s02 = 0;
+      } else {
+        // ASSERT(closestVert != -1, topologyErr, "No closest vert");
+        if (!Shadows(z02, vertPos.z, expandP * vertNormalP[closestVert].z))
+          s02 = 0;
+      }
+    }
+  }
+};
+
+std::tuple<Vec<int>, Vec<float>> Shadow02(const Manifold::Impl &inP,
+                                          const Manifold::Impl &inQ,
+                                          SparseIndices &p0q2, bool forward,
+                                          float expandP) {
+  ZoneScoped;
+  Vec<int> s02(p0q2.size());
+  Vec<float> z02(p0q2.size());
+
+  auto vertNormalP = forward ? inP.vertNormal_ : inQ.vertNormal_;
+  for_each_n(autoPolicy(p0q2.size()),
+             zip(countAt(0_z), s02.begin(), z02.begin()), p0q2.size(),
+             Kernel02({inP.vertPos_, inQ.halfedge_, inQ.vertPos_, expandP,
+                       vertNormalP, p0q2, forward}));
+
+  p0q2.KeepFinite(z02, s02);
+
+  return std::make_tuple(s02, z02);
+};
+
+struct Kernel12 {
+  VecView<const int64_t> p0q2;
+  VecView<const int> s02;
+  VecView<const float> z02;
+  VecView<const int64_t> p1q1;
+  VecView<const int> s11;
+  VecView<const glm::vec4> xyzz11;
+  VecView<const Halfedge> halfedgesP;
+  VecView<const Halfedge> halfedgesQ;
+  VecView<const glm::vec3> vertPosP;
+  const bool forward;
+  const SparseIndices &p1q2;
+
+  void operator()(thrust::tuple<size_t, int &, glm::vec3 &> inout) {
+    int p1 = p1q2.Get(thrust::get<0>(inout), !forward);
+    int q2 = p1q2.Get(thrust::get<0>(inout), forward);
+    int &x12 = thrust::get<1>(inout);
+    glm::vec3 &v12 = thrust::get<2>(inout);
+
+    // For xzyLR-[k], k==0 is the left and k==1 is the right.
+    int k = 0;
+    glm::vec3 xzyLR0[2];
+    glm::vec3 xzyLR1[2];
+    // Either the left or right must shadow, but not both. This ensures the
+    // intersection is between the left and right.
+    bool shadows = false;
+    x12 = 0;
+
+    const Halfedge edge = halfedgesP[p1];
+
+    for (int vert : {edge.startVert, edge.endVert}) {
+      const int64_t key = forward ? SparseIndices::EncodePQ(vert, q2)
+                                  : SparseIndices::EncodePQ(q2, vert);
+      const size_t idx = monobound_quaternary_search(p0q2, key);
+      if (idx != -1) {
+        const int s = s02[idx];
+        x12 += s * ((vert == edge.startVert) == forward ? 1 : -1);
+        if (k < 2 && (k == 0 || (s != 0) != shadows)) {
+          shadows = s != 0;
+          xzyLR0[k] = vertPosP[vert];
+          thrust::swap(xzyLR0[k].y, xzyLR0[k].z);
+          xzyLR1[k] = xzyLR0[k];
+          xzyLR1[k][1] = z02[idx];
+          k++;
+        }
+      }
+    }
+
+    for (const int i : {0, 1, 2}) {
+      const int q1 = 3 * q2 + i;
+      const Halfedge edge = halfedgesQ[q1];
+      const int q1F = edge.IsForward() ? q1 : edge.pairedHalfedge;
+      const int64_t key = forward ? SparseIndices::EncodePQ(p1, q1F)
+                                  : SparseIndices::EncodePQ(q1F, p1);
+      const size_t idx = monobound_quaternary_search(p1q1, key);
+      if (idx != -1) {  // s is implicitly zero for anything not found
+        const int s = s11[idx];
+        x12 -= s * (edge.IsForward() ? 1 : -1);
+        if (k < 2 && (k == 0 || (s != 0) != shadows)) {
+          shadows = s != 0;
+          const glm::vec4 xyzz = xyzz11[idx];
+          xzyLR0[k][0] = xyzz.x;
+          xzyLR0[k][1] = xyzz.z;
+          xzyLR0[k][2] = xyzz.y;
+          xzyLR1[k] = xzyLR0[k];
+          xzyLR1[k][1] = xyzz.w;
+          if (!forward) thrust::swap(xzyLR0[k][1], xzyLR1[k][1]);
+          k++;
+        }
+      }
+    }
+
+    if (x12 == 0) {  // No intersection
+      v12 = glm::vec3(NAN);
+    } else {
+      ASSERT(k == 2, logicErr, "Boolean manifold error: v12");
+      const glm::vec4 xzyy =
+          Intersect(xzyLR0[0], xzyLR0[1], xzyLR1[0], xzyLR1[1]);
+      v12.x = xzyy[0];
+      v12.y = xzyy[2];
+      v12.z = xzyy[1];
+    }
+  }
+};
+
+std::tuple<Vec<int>, Vec<glm::vec3>> Intersect12(
+    const Manifold::Impl &inP, const Manifold::Impl &inQ, const Vec<int> &s02,
+    const SparseIndices &p0q2, const Vec<int> &s11, const SparseIndices &p1q1,
+    const Vec<float> &z02, const Vec<glm::vec4> &xyzz11, SparseIndices &p1q2,
+    bool forward) {
+  ZoneScoped;
+  Vec<int> x12(p1q2.size());
+  Vec<glm::vec3> v12(p1q2.size());
+
+  for_each_n(
+      autoPolicy(p1q2.size()), zip(countAt(0_z), x12.begin(), v12.begin()),
+      p1q2.size(),
+      Kernel12({p0q2.AsVec64(), s02, z02, p1q1.AsVec64(), s11, xyzz11,
+                inP.halfedge_, inQ.halfedge_, inP.vertPos_, forward, p1q2}));
+
+  p1q2.KeepFinite(v12, x12);
+
+  return std::make_tuple(x12, v12);
+};
+
+Vec<int> Winding03(const Manifold::Impl &inP, Vec<int> &vertices, Vec<int> &s02,
+                   bool reverse) {
+  ZoneScoped;
+  // verts that are not shadowed (not in p0q2) have winding number zero.
+  Vec<int> w03(inP.NumVert(), 0);
+  // checking is slow, so just sort and reduce
+  auto policy = autoPolicy(vertices.size());
+  stable_sort(
+      policy, zip(vertices.begin(), s02.begin()),
+      zip(vertices.end(), s02.end()),
+      [](const thrust::tuple<int, int> &a, const thrust::tuple<int, int> &b) {
+        return thrust::get<0>(a) < thrust::get<0>(b);
+      });
+  Vec<int> w03val(w03.size());
+  Vec<int> w03vert(w03.size());
+  // sum known s02 values into w03 (winding number)
+  auto endPair = reduce_by_key<
+      thrust::pair<decltype(w03val.begin()), decltype(w03val.begin())>>(
+      policy, vertices.begin(), vertices.end(), s02.begin(), w03vert.begin(),
+      w03val.begin());
+  scatter(policy, w03val.begin(), endPair.second, w03vert.begin(), w03.begin());
+
+  if (reverse)
+    transform(policy, w03.begin(), w03.end(), w03.begin(),
+              thrust::negate<int>());
+  return w03;
+};
+}  // namespace
+
+namespace manifold {
+Boolean3::Boolean3(const Manifold::Impl &inP, const Manifold::Impl &inQ,
+                   OpType op)
+    : inP_(inP), inQ_(inQ), expandP_(op == OpType::Add ? 1.0 : -1.0) {
+  // Symbolic perturbation:
+  // Union -> expand inP
+  // Difference, Intersection -> contract inP
+
+#ifdef MANIFOLD_DEBUG
+  Timer broad;
+  broad.Start();
+#endif
+
+  if (inP.IsEmpty() || inQ.IsEmpty() || !inP.bBox_.DoesOverlap(inQ.bBox_)) {
+    PRINT("No overlap, early out");
+    w03_.resize(inP.NumVert(), 0);
+    w30_.resize(inQ.NumVert(), 0);
+    return;
+  }
+
+  // Level 3
+  // Find edge-triangle overlaps (broad phase)
+  p1q2_ = inQ_.EdgeCollisions(inP_);
+  p2q1_ = inP_.EdgeCollisions(inQ_, true);  // inverted
+
+  p1q2_.Sort();
+  PRINT("p1q2 size = " << p1q2_.size());
+
+  p2q1_.Sort();
+  PRINT("p2q1 size = " << p2q1_.size());
+
+  // Level 2
+  // Find vertices that overlap faces in XY-projection
+  SparseIndices p0q2 = inQ.VertexCollisionsZ(inP.vertPos_);
+  p0q2.Sort();
+  PRINT("p0q2 size = " << p0q2.size());
+
+  SparseIndices p2q0 = inP.VertexCollisionsZ(inQ.vertPos_, true);  // inverted
+  p2q0.Sort();
+  PRINT("p2q0 size = " << p2q0.size());
+
+  // Find involved edge pairs from Level 3
+  SparseIndices p1q1 = Filter11(inP_, inQ_, p1q2_, p2q1_);
+  PRINT("p1q1 size = " << p1q1.size());
+
+#ifdef MANIFOLD_DEBUG
+  broad.Stop();
+  Timer intersections;
+  intersections.Start();
+#endif
+
+  // Level 2
+  // Build up XY-projection intersection of two edges, including the z-value for
+  // each edge, keeping only those whose intersection exists.
+  Vec<int> s11;
+  Vec<glm::vec4> xyzz11;
+  std::tie(s11, xyzz11) = Shadow11(p1q1, inP, inQ, expandP_);
+  PRINT("s11 size = " << s11.size());
+
+  // Build up Z-projection of vertices onto triangles, keeping only those that
+  // fall inside the triangle.
+  Vec<int> s02;
+  Vec<float> z02;
+  std::tie(s02, z02) = Shadow02(inP, inQ, p0q2, true, expandP_);
+  PRINT("s02 size = " << s02.size());
+
+  Vec<int> s20;
+  Vec<float> z20;
+  std::tie(s20, z20) = Shadow02(inQ, inP, p2q0, false, expandP_);
+  PRINT("s20 size = " << s20.size());
+
+  // Level 3
+  // Build up the intersection of the edges and triangles, keeping only those
+  // that intersect, and record the direction the edge is passing through the
+  // triangle.
+  std::tie(x12_, v12_) =
+      Intersect12(inP, inQ, s02, p0q2, s11, p1q1, z02, xyzz11, p1q2_, true);
+  PRINT("x12 size = " << x12_.size());
+
+  std::tie(x21_, v21_) =
+      Intersect12(inQ, inP, s20, p2q0, s11, p1q1, z20, xyzz11, p2q1_, false);
+  PRINT("x21 size = " << x21_.size());
+
+  if (x12_.size() + x21_.size() >= std::numeric_limits<int>::max())
+    throw std::out_of_range("mesh too large");
+
+  Vec<int> p0 = p0q2.Copy(false);
+  p0q2.Resize(0);
+  Vec<int> q0 = p2q0.Copy(true);
+  p2q0.Resize(0);
+  // Sum up the winding numbers of all vertices.
+  w03_ = Winding03(inP, p0, s02, false);
+
+  w30_ = Winding03(inQ, q0, s20, true);
+
+#ifdef MANIFOLD_DEBUG
+  intersections.Stop();
+
+  if (ManifoldParams().verbose) {
+    broad.Print("Broad phase");
+    intersections.Print("Intersections");
+  }
+#endif
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/boolean3.h b/thirdparty/manifold/src/manifold/src/boolean3.h
new file mode 100644
index 000000000000..9491084c9cfe
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/boolean3.h
@@ -0,0 +1,60 @@
+// Copyright 2020 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include "impl.h"
+
+#ifdef MANIFOLD_DEBUG
+#define PRINT(msg) \
+  if (ManifoldParams().verbose) std::cout << msg << std::endl;
+#else
+#define PRINT(msg)
+#endif
+
+/**
+ * The notation in these files is abbreviated due to the complexity of the
+ * functions involved. The key is that the input manifolds are P and Q, while
+ * the output is R, and these letters in both upper and lower case refer to
+ * these objects. Operations are based on dimensionality: vert: 0, edge: 1,
+ * face: 2, solid: 3. X denotes a winding-number type quantity from the source
+ * paper of this algorithm, while S is closely related but includes only the
+ * subset of X values which "shadow" (are on the correct side of).
+ *
+ * Nearly everything here are sparse arrays, where for instance each pair in
+ * p2q1 refers to a face index of P interacting with a halfedge index of Q.
+ * Adjacent arrays like x21 refer to the values of X corresponding to each
+ * sparse index pair.
+ *
+ * Note many functions are designed to work symmetrically, for instance for both
+ * p2q1 and p1q2. Inside of these functions P and Q are marked as though the
+ * function is forwards, but it may include a Boolean "reverse" that indicates P
+ * and Q have been swapped.
+ */
+
+namespace manifold {
+
+/** @ingroup Private */
+class Boolean3 {
+ public:
+  Boolean3(const Manifold::Impl& inP, const Manifold::Impl& inQ, OpType op);
+  Manifold::Impl Result(OpType op) const;
+
+ private:
+  const Manifold::Impl &inP_, &inQ_;
+  const float expandP_;
+  SparseIndices p1q2_, p2q1_;
+  Vec<int> x12_, x21_, w03_, w30_;
+  Vec<glm::vec3> v12_, v21_;
+};
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/boolean_result.cpp b/thirdparty/manifold/src/manifold/src/boolean_result.cpp
new file mode 100644
index 000000000000..c6750aca201f
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/boolean_result.cpp
@@ -0,0 +1,831 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <algorithm>
+#include <array>
+#include <map>
+
+#if MANIFOLD_PAR == 'T' && __has_include(<tbb/concurrent_map.h>)
+#define TBB_PREVIEW_CONCURRENT_ORDERED_CONTAINERS 1
+#include <tbb/concurrent_map.h>
+#include <tbb/parallel_for.h>
+
+template <typename K, typename V>
+using concurrent_map = tbb::concurrent_map<K, V>;
+#else
+template <typename K, typename V>
+// not really concurrent when tbb is disabled
+using concurrent_map = std::map<K, V>;
+#endif
+#include "boolean3.h"
+#include "par.h"
+#include "polygon.h"
+
+using namespace manifold;
+using namespace thrust::placeholders;
+
+template <>
+struct std::hash<std::pair<int, int>> {
+  size_t operator()(const std::pair<int, int> &p) const {
+    return std::hash<int>()(p.first) ^ std::hash<int>()(p.second);
+  }
+};
+
+namespace {
+
+constexpr int kParallelThreshold = 128;
+
+struct AbsSum : public thrust::binary_function<int, int, int> {
+  int operator()(int a, int b) { return abs(a) + abs(b); }
+};
+
+struct DuplicateVerts {
+  VecView<glm::vec3> vertPosR;
+
+  void operator()(thrust::tuple<int, int, glm::vec3> in) {
+    int inclusion = abs(thrust::get<0>(in));
+    int vertR = thrust::get<1>(in);
+    glm::vec3 vertPosP = thrust::get<2>(in);
+
+    for (int i = 0; i < inclusion; ++i) {
+      vertPosR[vertR + i] = vertPosP;
+    }
+  }
+};
+
+struct CountVerts {
+  VecView<int> count;
+  VecView<const int> inclusion;
+
+  void operator()(const Halfedge &edge) {
+    AtomicAdd(count[edge.face], glm::abs(inclusion[edge.startVert]));
+  }
+};
+
+template <const bool inverted>
+struct CountNewVerts {
+  VecView<int> countP;
+  VecView<int> countQ;
+  const SparseIndices &pq;
+  VecView<const Halfedge> halfedges;
+
+  void operator()(thrust::tuple<int, int> in) {
+    int edgeP = pq.Get(thrust::get<0>(in), inverted);
+    int faceQ = pq.Get(thrust::get<0>(in), !inverted);
+    int inclusion = glm::abs(thrust::get<1>(in));
+
+    AtomicAdd(countQ[faceQ], inclusion);
+    const Halfedge half = halfedges[edgeP];
+    AtomicAdd(countP[half.face], inclusion);
+    AtomicAdd(countP[halfedges[half.pairedHalfedge].face], inclusion);
+  }
+};
+
+struct NotZero : public thrust::unary_function<int, int> {
+  int operator()(int x) const { return x > 0 ? 1 : 0; }
+};
+
+std::tuple<Vec<int>, Vec<int>> SizeOutput(
+    Manifold::Impl &outR, const Manifold::Impl &inP, const Manifold::Impl &inQ,
+    const Vec<int> &i03, const Vec<int> &i30, const Vec<int> &i12,
+    const Vec<int> &i21, const SparseIndices &p1q2, const SparseIndices &p2q1,
+    bool invertQ) {
+  ZoneScoped;
+  Vec<int> sidesPerFacePQ(inP.NumTri() + inQ.NumTri(), 0);
+  // note: numFaceR <= facePQ2R.size() = sidesPerFacePQ.size() + 1
+  if (sidesPerFacePQ.size() + 1 >= std::numeric_limits<int>::max())
+    throw std::out_of_range("boolean result too large");
+
+  auto sidesPerFaceP = sidesPerFacePQ.view(0, inP.NumTri());
+  auto sidesPerFaceQ = sidesPerFacePQ.view(inP.NumTri(), inQ.NumTri());
+
+  for_each(autoPolicy(inP.halfedge_.size()), inP.halfedge_.begin(),
+           inP.halfedge_.end(), CountVerts({sidesPerFaceP, i03}));
+  for_each(autoPolicy(inP.halfedge_.size()), inQ.halfedge_.begin(),
+           inQ.halfedge_.end(), CountVerts({sidesPerFaceQ, i30}));
+  for_each_n(autoPolicy(i12.size()), zip(countAt(0), i12.begin()), i12.size(),
+             CountNewVerts<false>(
+                 {sidesPerFaceP, sidesPerFaceQ, p1q2, inP.halfedge_}));
+  for_each_n(
+      autoPolicy(i21.size()), zip(countAt(0), i21.begin()), i21.size(),
+      CountNewVerts<true>({sidesPerFaceQ, sidesPerFaceP, p2q1, inQ.halfedge_}));
+
+  Vec<int> facePQ2R(inP.NumTri() + inQ.NumTri() + 1, 0);
+  auto keepFace =
+      thrust::make_transform_iterator(sidesPerFacePQ.begin(), NotZero());
+  inclusive_scan(autoPolicy(sidesPerFacePQ.size()), keepFace,
+                 keepFace + sidesPerFacePQ.size(), facePQ2R.begin() + 1);
+  int numFaceR = facePQ2R.back();
+  facePQ2R.resize(inP.NumTri() + inQ.NumTri());
+
+  outR.faceNormal_.resize(numFaceR);
+  auto next = copy_if<decltype(outR.faceNormal_.begin())>(
+      autoPolicy(inP.faceNormal_.size()), inP.faceNormal_.begin(),
+      inP.faceNormal_.end(), keepFace, outR.faceNormal_.begin(),
+      thrust::identity<bool>());
+  if (invertQ) {
+    auto start = thrust::make_transform_iterator(inQ.faceNormal_.begin(),
+                                                 thrust::negate<glm::vec3>());
+    auto end = thrust::make_transform_iterator(inQ.faceNormal_.end(),
+                                               thrust::negate<glm::vec3>());
+    copy_if<decltype(inQ.faceNormal_.begin())>(
+        autoPolicy(inQ.faceNormal_.size()), start, end, keepFace + inP.NumTri(),
+        next, thrust::identity<bool>());
+  } else {
+    copy_if<decltype(inQ.faceNormal_.begin())>(
+        autoPolicy(inQ.faceNormal_.size()), inQ.faceNormal_.begin(),
+        inQ.faceNormal_.end(), keepFace + inP.NumTri(), next,
+        thrust::identity<bool>());
+  }
+
+  auto newEnd = remove<decltype(sidesPerFacePQ.begin())>(
+      autoPolicy(sidesPerFacePQ.size()), sidesPerFacePQ.begin(),
+      sidesPerFacePQ.end(), 0);
+  Vec<int> faceEdge(newEnd - sidesPerFacePQ.begin() + 1, 0);
+  inclusive_scan(autoPolicy(std::distance(sidesPerFacePQ.begin(), newEnd)),
+                 sidesPerFacePQ.begin(), newEnd, faceEdge.begin() + 1);
+  outR.halfedge_.resize(faceEdge.back());
+
+  return std::make_tuple(faceEdge, facePQ2R);
+}
+
+struct EdgePos {
+  int vert;
+  float edgePos;
+  bool isStart;
+};
+
+void AddNewEdgeVerts(
+    // we need concurrent_map because we will be adding things concurrently
+    concurrent_map<int, std::vector<EdgePos>> &edgesP,
+    concurrent_map<std::pair<int, int>, std::vector<EdgePos>> &edgesNew,
+    const SparseIndices &p1q2, const Vec<int> &i12, const Vec<int> &v12R,
+    const Vec<Halfedge> &halfedgeP, bool forward) {
+  ZoneScoped;
+  // For each edge of P that intersects a face of Q (p1q2), add this vertex to
+  // P's corresponding edge vector and to the two new edges, which are
+  // intersections between the face of Q and the two faces of P attached to the
+  // edge. The direction and duplicity are given by i12, while v12R remaps to
+  // the output vert index. When forward is false, all is reversed.
+  auto process = [&](std::function<void(size_t)> lock,
+                     std::function<void(size_t)> unlock, size_t i) {
+    const int edgeP = p1q2.Get(i, !forward);
+    const int faceQ = p1q2.Get(i, forward);
+    const int vert = v12R[i];
+    const int inclusion = i12[i];
+
+    Halfedge halfedge = halfedgeP[edgeP];
+    std::pair<int, int> keyRight = {halfedgeP[halfedge.pairedHalfedge].face,
+                                    faceQ};
+    if (!forward) std::swap(keyRight.first, keyRight.second);
+
+    std::pair<int, int> keyLeft = {halfedge.face, faceQ};
+    if (!forward) std::swap(keyLeft.first, keyLeft.second);
+
+    bool direction = inclusion < 0;
+    std::hash<std::pair<int, int>> pairHasher;
+    std::array<std::tuple<bool, size_t, std::vector<EdgePos> *>, 3> edges = {
+        std::make_tuple(direction, std::hash<int>{}(edgeP), &edgesP[edgeP]),
+        std::make_tuple(direction ^ !forward,  // revert if not forward
+                        pairHasher(keyRight), &edgesNew[keyRight]),
+        std::make_tuple(direction ^ forward,  // revert if forward
+                        pairHasher(keyLeft), &edgesNew[keyLeft])};
+    for (const auto &tuple : edges) {
+      lock(std::get<1>(tuple));
+      for (int j = 0; j < glm::abs(inclusion); ++j)
+        std::get<2>(tuple)->push_back({vert + j, 0.0f, std::get<0>(tuple)});
+      unlock(std::get<1>(tuple));
+      direction = !direction;
+    }
+  };
+#if MANIFOLD_PAR == 'T' && __has_include(<tbb/tbb.h>)
+  // parallelize operations, requires concurrent_map so we can only enable this
+  // with tbb
+  if (!ManifoldParams().deterministic && p1q2.size() > kParallelThreshold) {
+    // ideally we should have 1 mutex per key, but kParallelThreshold is enough
+    // to avoid contention for most of the cases
+    std::array<std::mutex, kParallelThreshold> mutexes;
+    static tbb::affinity_partitioner ap;
+    auto processFun = std::bind(
+        process, [&](size_t hash) { mutexes[hash % mutexes.size()].lock(); },
+        [&](size_t hash) { mutexes[hash % mutexes.size()].unlock(); },
+        std::placeholders::_1);
+    tbb::parallel_for(
+        tbb::blocked_range<size_t>(0_z, p1q2.size(), 32),
+        [&](const tbb::blocked_range<size_t> &range) {
+          for (size_t i = range.begin(); i != range.end(); i++) processFun(i);
+        },
+        ap);
+    return;
+  }
+#endif
+  auto processFun = std::bind(
+      process, [](size_t _) {}, [](size_t _) {}, std::placeholders::_1);
+  for (size_t i = 0; i < p1q2.size(); ++i) processFun(i);
+}
+
+std::vector<Halfedge> PairUp(std::vector<EdgePos> &edgePos) {
+  // Pair start vertices with end vertices to form edges. The choice of pairing
+  // is arbitrary for the manifoldness guarantee, but must be ordered to be
+  // geometrically valid. If the order does not go start-end-start-end... then
+  // the input and output are not geometrically valid and this algorithm becomes
+  // a heuristic.
+  ASSERT(edgePos.size() % 2 == 0, topologyErr,
+         "Non-manifold edge! Not an even number of points.");
+  size_t nEdges = edgePos.size() / 2;
+  auto middle = std::partition(edgePos.begin(), edgePos.end(),
+                               [](EdgePos x) { return x.isStart; });
+  ASSERT(middle - edgePos.begin() == nEdges, topologyErr, "Non-manifold edge!");
+  auto cmp = [](EdgePos a, EdgePos b) { return a.edgePos < b.edgePos; };
+  std::stable_sort(edgePos.begin(), middle, cmp);
+  std::stable_sort(middle, edgePos.end(), cmp);
+  std::vector<Halfedge> edges;
+  for (size_t i = 0; i < nEdges; ++i)
+    edges.push_back({edgePos[i].vert, edgePos[i + nEdges].vert, -1, -1});
+  return edges;
+}
+
+void AppendPartialEdges(Manifold::Impl &outR, Vec<char> &wholeHalfedgeP,
+                        Vec<int> &facePtrR,
+                        concurrent_map<int, std::vector<EdgePos>> &edgesP,
+                        Vec<TriRef> &halfedgeRef, const Manifold::Impl &inP,
+                        const Vec<int> &i03, const Vec<int> &vP2R,
+                        const Vec<int>::IterC faceP2R, bool forward) {
+  ZoneScoped;
+  // Each edge in the map is partially retained; for each of these, look up
+  // their original verts and include them based on their winding number (i03),
+  // while remapping them to the output using vP2R. Use the verts position
+  // projected along the edge vector to pair them up, then distribute these
+  // edges to their faces.
+  Vec<Halfedge> &halfedgeR = outR.halfedge_;
+  const Vec<glm::vec3> &vertPosP = inP.vertPos_;
+  const Vec<Halfedge> &halfedgeP = inP.halfedge_;
+
+  for (auto &value : edgesP) {
+    const int edgeP = value.first;
+    std::vector<EdgePos> &edgePosP = value.second;
+
+    const Halfedge &halfedge = halfedgeP[edgeP];
+    wholeHalfedgeP[edgeP] = false;
+    wholeHalfedgeP[halfedge.pairedHalfedge] = false;
+
+    const int vStart = halfedge.startVert;
+    const int vEnd = halfedge.endVert;
+    const glm::vec3 edgeVec = vertPosP[vEnd] - vertPosP[vStart];
+    // Fill in the edge positions of the old points.
+    for (EdgePos &edge : edgePosP) {
+      edge.edgePos = glm::dot(outR.vertPos_[edge.vert], edgeVec);
+    }
+
+    int inclusion = i03[vStart];
+    EdgePos edgePos = {vP2R[vStart],
+                       glm::dot(outR.vertPos_[vP2R[vStart]], edgeVec),
+                       inclusion > 0};
+    for (int j = 0; j < glm::abs(inclusion); ++j) {
+      edgePosP.push_back(edgePos);
+      ++edgePos.vert;
+    }
+
+    inclusion = i03[vEnd];
+    edgePos = {vP2R[vEnd], glm::dot(outR.vertPos_[vP2R[vEnd]], edgeVec),
+               inclusion < 0};
+    for (int j = 0; j < glm::abs(inclusion); ++j) {
+      edgePosP.push_back(edgePos);
+      ++edgePos.vert;
+    }
+
+    // sort edges into start/end pairs along length
+    std::vector<Halfedge> edges = PairUp(edgePosP);
+
+    // add halfedges to result
+    const int faceLeftP = halfedge.face;
+    const int faceLeft = faceP2R[faceLeftP];
+    const int faceRightP = halfedgeP[halfedge.pairedHalfedge].face;
+    const int faceRight = faceP2R[faceRightP];
+    // Negative inclusion means the halfedges are reversed, which means our
+    // reference is now to the endVert instead of the startVert, which is one
+    // position advanced CCW. This is only valid if this is a retained vert; it
+    // will be ignored later if the vert is new.
+    const TriRef forwardRef = {forward ? 0 : 1, -1, faceLeftP};
+    const TriRef backwardRef = {forward ? 0 : 1, -1, faceRightP};
+
+    for (Halfedge e : edges) {
+      const int forwardEdge = facePtrR[faceLeft]++;
+      const int backwardEdge = facePtrR[faceRight]++;
+
+      e.face = faceLeft;
+      e.pairedHalfedge = backwardEdge;
+      halfedgeR[forwardEdge] = e;
+      halfedgeRef[forwardEdge] = forwardRef;
+
+      std::swap(e.startVert, e.endVert);
+      e.face = faceRight;
+      e.pairedHalfedge = forwardEdge;
+      halfedgeR[backwardEdge] = e;
+      halfedgeRef[backwardEdge] = backwardRef;
+    }
+  }
+}
+
+void AppendNewEdges(
+    Manifold::Impl &outR, Vec<int> &facePtrR,
+    concurrent_map<std::pair<int, int>, std::vector<EdgePos>> &edgesNew,
+    Vec<TriRef> &halfedgeRef, const Vec<int> &facePQ2R, const int numFaceP) {
+  ZoneScoped;
+  // Pair up each edge's verts and distribute to faces based on indices in key.
+  Vec<Halfedge> &halfedgeR = outR.halfedge_;
+  Vec<glm::vec3> &vertPosR = outR.vertPos_;
+
+  for (auto &value : edgesNew) {
+    const int faceP = value.first.first;
+    const int faceQ = value.first.second;
+    std::vector<EdgePos> &edgePos = value.second;
+
+    Box bbox;
+    for (auto edge : edgePos) {
+      bbox.Union(vertPosR[edge.vert]);
+    }
+    const glm::vec3 size = bbox.Size();
+    // Order the points along their longest dimension.
+    const int i = (size.x > size.y && size.x > size.z) ? 0
+                  : size.y > size.z                    ? 1
+                                                       : 2;
+    for (auto &edge : edgePos) {
+      edge.edgePos = vertPosR[edge.vert][i];
+    }
+
+    // sort edges into start/end pairs along length.
+    std::vector<Halfedge> edges = PairUp(edgePos);
+
+    // add halfedges to result
+    const int faceLeft = facePQ2R[faceP];
+    const int faceRight = facePQ2R[numFaceP + faceQ];
+    const TriRef forwardRef = {0, -1, faceP};
+    const TriRef backwardRef = {1, -1, faceQ};
+    for (Halfedge e : edges) {
+      const int forwardEdge = facePtrR[faceLeft]++;
+      const int backwardEdge = facePtrR[faceRight]++;
+
+      e.face = faceLeft;
+      e.pairedHalfedge = backwardEdge;
+      halfedgeR[forwardEdge] = e;
+      halfedgeRef[forwardEdge] = forwardRef;
+
+      std::swap(e.startVert, e.endVert);
+      e.face = faceRight;
+      e.pairedHalfedge = forwardEdge;
+      halfedgeR[backwardEdge] = e;
+      halfedgeRef[backwardEdge] = backwardRef;
+    }
+  }
+}
+
+struct DuplicateHalfedges {
+  VecView<Halfedge> halfedgesR;
+  VecView<TriRef> halfedgeRef;
+  VecView<int> facePtr;
+  VecView<const Halfedge> halfedgesP;
+  VecView<const int> i03;
+  VecView<const int> vP2R;
+  VecView<const int> faceP2R;
+  const bool forward;
+
+  void operator()(thrust::tuple<bool, Halfedge, int> in) {
+    if (!thrust::get<0>(in)) return;
+    Halfedge halfedge = thrust::get<1>(in);
+    if (!halfedge.IsForward()) return;
+
+    const int inclusion = i03[halfedge.startVert];
+    if (inclusion == 0) return;
+    if (inclusion < 0) {  // reverse
+      int tmp = halfedge.startVert;
+      halfedge.startVert = halfedge.endVert;
+      halfedge.endVert = tmp;
+    }
+    halfedge.startVert = vP2R[halfedge.startVert];
+    halfedge.endVert = vP2R[halfedge.endVert];
+    const int faceLeftP = halfedge.face;
+    halfedge.face = faceP2R[faceLeftP];
+    const int faceRightP = halfedgesP[halfedge.pairedHalfedge].face;
+    const int faceRight = faceP2R[faceRightP];
+    // Negative inclusion means the halfedges are reversed, which means our
+    // reference is now to the endVert instead of the startVert, which is one
+    // position advanced CCW.
+    const TriRef forwardRef = {forward ? 0 : 1, -1, faceLeftP};
+    const TriRef backwardRef = {forward ? 0 : 1, -1, faceRightP};
+
+    for (int i = 0; i < glm::abs(inclusion); ++i) {
+      int forwardEdge = AtomicAdd(facePtr[halfedge.face], 1);
+      int backwardEdge = AtomicAdd(facePtr[faceRight], 1);
+      halfedge.pairedHalfedge = backwardEdge;
+
+      halfedgesR[forwardEdge] = halfedge;
+      halfedgesR[backwardEdge] = {halfedge.endVert, halfedge.startVert,
+                                  forwardEdge, faceRight};
+      halfedgeRef[forwardEdge] = forwardRef;
+      halfedgeRef[backwardEdge] = backwardRef;
+
+      ++halfedge.startVert;
+      ++halfedge.endVert;
+    }
+  }
+};
+
+void AppendWholeEdges(Manifold::Impl &outR, Vec<int> &facePtrR,
+                      Vec<TriRef> &halfedgeRef, const Manifold::Impl &inP,
+                      const Vec<char> wholeHalfedgeP, const Vec<int> &i03,
+                      const Vec<int> &vP2R, VecView<const int> faceP2R,
+                      bool forward) {
+  ZoneScoped;
+  for_each_n(ManifoldParams().deterministic ? ExecutionPolicy::Seq
+                                            : autoPolicy(inP.halfedge_.size()),
+             zip(wholeHalfedgeP.begin(), inP.halfedge_.begin(), countAt(0)),
+             inP.halfedge_.size(),
+             DuplicateHalfedges({outR.halfedge_, halfedgeRef, facePtrR,
+                                 inP.halfedge_, i03, vP2R, faceP2R, forward}));
+}
+
+struct MapTriRef {
+  VecView<const TriRef> triRefP;
+  VecView<const TriRef> triRefQ;
+  const int offsetQ;
+
+  void operator()(TriRef &triRef) {
+    const int tri = triRef.tri;
+    const bool PQ = triRef.meshID == 0;
+    triRef = PQ ? triRefP[tri] : triRefQ[tri];
+    if (!PQ) triRef.meshID += offsetQ;
+  }
+};
+
+void UpdateReference(Manifold::Impl &outR, const Manifold::Impl &inP,
+                     const Manifold::Impl &inQ, bool invertQ) {
+  const int offsetQ = Manifold::Impl::meshIDCounter_;
+  for_each_n(
+      autoPolicy(outR.NumTri()), outR.meshRelation_.triRef.begin(),
+      outR.NumTri(),
+      MapTriRef({inP.meshRelation_.triRef, inQ.meshRelation_.triRef, offsetQ}));
+
+  for (const auto &pair : inP.meshRelation_.meshIDtransform) {
+    outR.meshRelation_.meshIDtransform[pair.first] = pair.second;
+  }
+  for (const auto &pair : inQ.meshRelation_.meshIDtransform) {
+    outR.meshRelation_.meshIDtransform[pair.first + offsetQ] = pair.second;
+    outR.meshRelation_.meshIDtransform[pair.first + offsetQ].backSide ^=
+        invertQ;
+  }
+}
+
+struct Barycentric {
+  VecView<glm::vec3> uvw;
+  VecView<const glm::vec3> vertPosP;
+  VecView<const glm::vec3> vertPosQ;
+  VecView<const glm::vec3> vertPosR;
+  VecView<const Halfedge> halfedgeP;
+  VecView<const Halfedge> halfedgeQ;
+  VecView<const Halfedge> halfedgeR;
+  const float precision;
+
+  void operator()(thrust::tuple<int, TriRef> in) {
+    const int tri = thrust::get<0>(in);
+    const TriRef refPQ = thrust::get<1>(in);
+    if (halfedgeR[3 * tri].startVert < 0) return;
+
+    const int triPQ = refPQ.tri;
+    const bool PQ = refPQ.meshID == 0;
+    const auto &vertPos = PQ ? vertPosP : vertPosQ;
+    const auto &halfedge = PQ ? halfedgeP : halfedgeQ;
+
+    glm::mat3 triPos;
+    for (const int j : {0, 1, 2})
+      triPos[j] = vertPos[halfedge[3 * triPQ + j].startVert];
+
+    for (const int i : {0, 1, 2}) {
+      const int vert = halfedgeR[3 * tri + i].startVert;
+      uvw[3 * tri + i] = GetBarycentric(vertPosR[vert], triPos, precision);
+    }
+  }
+};
+
+void CreateProperties(Manifold::Impl &outR, const Manifold::Impl &inP,
+                      const Manifold::Impl &inQ) {
+  ZoneScoped;
+  const int numPropP = inP.NumProp();
+  const int numPropQ = inQ.NumProp();
+  const int numProp = glm::max(numPropP, numPropQ);
+  outR.meshRelation_.numProp = numProp;
+  if (numProp == 0) return;
+
+  const int numTri = outR.NumTri();
+  outR.meshRelation_.triProperties.resize(numTri);
+
+  Vec<glm::vec3> bary(outR.halfedge_.size());
+  for_each_n(autoPolicy(numTri),
+             zip(countAt(0), outR.meshRelation_.triRef.cbegin()), numTri,
+             Barycentric({bary, inP.vertPos_, inQ.vertPos_, outR.vertPos_,
+                          inP.halfedge_, inQ.halfedge_, outR.halfedge_,
+                          outR.precision_}));
+
+  using Entry = std::pair<glm::ivec3, int>;
+  int idMissProp = outR.NumVert();
+  std::vector<std::vector<Entry>> propIdx(outR.NumVert() + 1);
+  std::vector<int> propMissIdx[2];
+  propMissIdx[0].resize(inQ.NumPropVert(), -1);
+  propMissIdx[1].resize(inP.NumPropVert(), -1);
+
+  if (static_cast<size_t>(outR.NumVert()) * static_cast<size_t>(numProp) >=
+      std::numeric_limits<int>::max())
+    throw std::out_of_range("too many vertices");
+
+  outR.meshRelation_.properties.reserve(outR.NumVert() * numProp);
+  int idx = 0;
+
+  for (int tri = 0; tri < numTri; ++tri) {
+    // Skip collapsed triangles
+    if (outR.halfedge_[3 * tri].startVert < 0) continue;
+
+    const TriRef ref = outR.meshRelation_.triRef[tri];
+    const bool PQ = ref.meshID == 0;
+    const int oldNumProp = PQ ? numPropP : numPropQ;
+    const auto &properties =
+        PQ ? inP.meshRelation_.properties : inQ.meshRelation_.properties;
+    const glm::ivec3 &triProp = oldNumProp == 0 ? glm::ivec3(-1)
+                                : PQ ? inP.meshRelation_.triProperties[ref.tri]
+                                     : inQ.meshRelation_.triProperties[ref.tri];
+
+    for (const int i : {0, 1, 2}) {
+      const int vert = outR.halfedge_[3 * tri + i].startVert;
+      const glm::vec3 &uvw = bary[3 * tri + i];
+
+      glm::ivec4 key(PQ, idMissProp, -1, -1);
+      if (oldNumProp > 0) {
+        int edge = -2;
+        for (const int j : {0, 1, 2}) {
+          if (uvw[j] == 1) {
+            // On a retained vert, the propVert must also match
+            key[2] = triProp[j];
+            edge = -1;
+            break;
+          }
+          if (uvw[j] == 0) edge = j;
+        }
+        if (edge >= 0) {
+          // On an edge, both propVerts must match
+          const int p0 = triProp[Next3(edge)];
+          const int p1 = triProp[Prev3(edge)];
+          key[1] = vert;
+          key[2] = glm::min(p0, p1);
+          key[3] = glm::max(p0, p1);
+        } else if (edge == -2) {
+          key[1] = vert;
+        }
+      }
+
+      if (key.y == idMissProp && key.z >= 0) {
+        // only key.x/key.z matters
+        auto &entry = propMissIdx[key.x][key.z];
+        if (entry >= 0) {
+          outR.meshRelation_.triProperties[tri][i] = entry;
+          continue;
+        }
+        entry = idx;
+      } else {
+        auto &bin = propIdx[key.y];
+        bool bFound = false;
+        for (int k = 0; k < bin.size(); ++k) {
+          if (bin[k].first == glm::ivec3(key.x, key.z, key.w)) {
+            bFound = true;
+            outR.meshRelation_.triProperties[tri][i] = bin[k].second;
+            break;
+          }
+        }
+        if (bFound) continue;
+        bin.push_back(std::make_pair(glm::ivec3(key.x, key.z, key.w), idx));
+      }
+
+      outR.meshRelation_.triProperties[tri][i] = idx++;
+      for (int p = 0; p < numProp; ++p) {
+        if (p < oldNumProp) {
+          glm::vec3 oldProps;
+          for (const int j : {0, 1, 2})
+            oldProps[j] = properties[oldNumProp * triProp[j] + p];
+          outR.meshRelation_.properties.push_back(glm::dot(uvw, oldProps));
+        } else {
+          outR.meshRelation_.properties.push_back(0);
+        }
+      }
+    }
+  }
+}
+}  // namespace
+
+namespace manifold {
+
+Manifold::Impl Boolean3::Result(OpType op) const {
+#ifdef MANIFOLD_DEBUG
+  Timer assemble;
+  assemble.Start();
+#endif
+
+  ASSERT((expandP_ > 0) == (op == OpType::Add), logicErr,
+         "Result op type not compatible with constructor op type.");
+  const int c1 = op == OpType::Intersect ? 0 : 1;
+  const int c2 = op == OpType::Add ? 1 : 0;
+  const int c3 = op == OpType::Intersect ? 1 : -1;
+
+  if (inP_.IsEmpty()) {
+    if (!inQ_.IsEmpty() && op == OpType::Add) {
+      return inQ_;
+    }
+    return Manifold::Impl();
+  } else if (inQ_.IsEmpty()) {
+    if (op == OpType::Intersect) {
+      return Manifold::Impl();
+    }
+    return inP_;
+  }
+
+  const bool invertQ = op == OpType::Subtract;
+
+  // Convert winding numbers to inclusion values based on operation type.
+  Vec<int> i12(x12_.size());
+  Vec<int> i21(x21_.size());
+  Vec<int> i03(w03_.size());
+  Vec<int> i30(w30_.size());
+
+  transform(autoPolicy(x12_.size()), x12_.begin(), x12_.end(), i12.begin(),
+            c3 * _1);
+  transform(autoPolicy(x21_.size()), x21_.begin(), x21_.end(), i21.begin(),
+            c3 * _1);
+  transform(autoPolicy(w03_.size()), w03_.begin(), w03_.end(), i03.begin(),
+            c1 + c3 * _1);
+  transform(autoPolicy(w30_.size()), w30_.begin(), w30_.end(), i30.begin(),
+            c2 + c3 * _1);
+
+  Vec<int> vP2R(inP_.NumVert());
+  exclusive_scan(autoPolicy(i03.size()), i03.begin(), i03.end(), vP2R.begin(),
+                 0, AbsSum());
+  int numVertR = AbsSum()(vP2R.back(), i03.back());
+  const int nPv = numVertR;
+
+  Vec<int> vQ2R(inQ_.NumVert());
+  exclusive_scan(autoPolicy(i30.size()), i30.begin(), i30.end(), vQ2R.begin(),
+                 numVertR, AbsSum());
+  numVertR = AbsSum()(vQ2R.back(), i30.back());
+  const int nQv = numVertR - nPv;
+
+  Vec<int> v12R(v12_.size());
+  if (v12_.size() > 0) {
+    exclusive_scan(autoPolicy(i12.size()), i12.begin(), i12.end(), v12R.begin(),
+                   numVertR, AbsSum());
+    numVertR = AbsSum()(v12R.back(), i12.back());
+  }
+  const int n12 = numVertR - nPv - nQv;
+
+  Vec<int> v21R(v21_.size());
+  if (v21_.size() > 0) {
+    exclusive_scan(autoPolicy(i21.size()), i21.begin(), i21.end(), v21R.begin(),
+                   numVertR, AbsSum());
+    numVertR = AbsSum()(v21R.back(), i21.back());
+  }
+  const int n21 = numVertR - nPv - nQv - n12;
+
+  // Create the output Manifold
+  Manifold::Impl outR;
+
+  if (numVertR == 0) return outR;
+
+  outR.precision_ = glm::max(inP_.precision_, inQ_.precision_);
+
+  outR.vertPos_.resize(numVertR);
+  // Add vertices, duplicating for inclusion numbers not in [-1, 1].
+  // Retained vertices from P and Q:
+  for_each_n(autoPolicy(inP_.NumVert()),
+             zip(i03.begin(), vP2R.begin(), inP_.vertPos_.begin()),
+             inP_.NumVert(), DuplicateVerts({outR.vertPos_}));
+  for_each_n(autoPolicy(inQ_.NumVert()),
+             zip(i30.begin(), vQ2R.begin(), inQ_.vertPos_.begin()),
+             inQ_.NumVert(), DuplicateVerts({outR.vertPos_}));
+  // New vertices created from intersections:
+  for_each_n(autoPolicy(i12.size()),
+             zip(i12.begin(), v12R.begin(), v12_.begin()), i12.size(),
+             DuplicateVerts({outR.vertPos_}));
+  for_each_n(autoPolicy(i21.size()),
+             zip(i21.begin(), v21R.begin(), v21_.begin()), i21.size(),
+             DuplicateVerts({outR.vertPos_}));
+
+  PRINT(nPv << " verts from inP");
+  PRINT(nQv << " verts from inQ");
+  PRINT(n12 << " new verts from edgesP -> facesQ");
+  PRINT(n21 << " new verts from facesP -> edgesQ");
+
+  // Build up new polygonal faces from triangle intersections. At this point the
+  // calculation switches from parallel to serial.
+
+  // Level 3
+
+  // This key is the forward halfedge index of P or Q. Only includes intersected
+  // edges.
+  concurrent_map<int, std::vector<EdgePos>> edgesP, edgesQ;
+  // This key is the face index of <P, Q>
+  concurrent_map<std::pair<int, int>, std::vector<EdgePos>> edgesNew;
+
+  AddNewEdgeVerts(edgesP, edgesNew, p1q2_, i12, v12R, inP_.halfedge_, true);
+  AddNewEdgeVerts(edgesQ, edgesNew, p2q1_, i21, v21R, inQ_.halfedge_, false);
+
+  // Level 4
+  Vec<int> faceEdge;
+  Vec<int> facePQ2R;
+  std::tie(faceEdge, facePQ2R) =
+      SizeOutput(outR, inP_, inQ_, i03, i30, i12, i21, p1q2_, p2q1_, invertQ);
+
+  // This gets incremented for each halfedge that's added to a face so that the
+  // next one knows where to slot in.
+  Vec<int> facePtrR = faceEdge;
+  // Intersected halfedges are marked false.
+  Vec<char> wholeHalfedgeP(inP_.halfedge_.size(), true);
+  Vec<char> wholeHalfedgeQ(inQ_.halfedge_.size(), true);
+  // The halfedgeRef contains the data that will become triRef once the faces
+  // are triangulated.
+  Vec<TriRef> halfedgeRef(2 * outR.NumEdge());
+
+  AppendPartialEdges(outR, wholeHalfedgeP, facePtrR, edgesP, halfedgeRef, inP_,
+                     i03, vP2R, facePQ2R.begin(), true);
+  AppendPartialEdges(outR, wholeHalfedgeQ, facePtrR, edgesQ, halfedgeRef, inQ_,
+                     i30, vQ2R, facePQ2R.begin() + inP_.NumTri(), false);
+
+  AppendNewEdges(outR, facePtrR, edgesNew, halfedgeRef, facePQ2R,
+                 inP_.NumTri());
+
+  AppendWholeEdges(outR, facePtrR, halfedgeRef, inP_, wholeHalfedgeP, i03, vP2R,
+                   facePQ2R.cview(0, inP_.NumTri()), true);
+  AppendWholeEdges(outR, facePtrR, halfedgeRef, inQ_, wholeHalfedgeQ, i30, vQ2R,
+                   facePQ2R.cview(inP_.NumTri(), inQ_.NumTri()), false);
+
+#ifdef MANIFOLD_DEBUG
+  assemble.Stop();
+  Timer triangulate;
+  triangulate.Start();
+#endif
+
+  // Level 6
+
+  if (ManifoldParams().intermediateChecks)
+    ASSERT(outR.IsManifold(), logicErr, "polygon mesh is not manifold!");
+
+  outR.Face2Tri(faceEdge, halfedgeRef);
+
+#ifdef MANIFOLD_DEBUG
+  triangulate.Stop();
+  Timer simplify;
+  simplify.Start();
+#endif
+
+  if (ManifoldParams().intermediateChecks)
+    ASSERT(outR.IsManifold(), logicErr, "triangulated mesh is not manifold!");
+
+  CreateProperties(outR, inP_, inQ_);
+
+  UpdateReference(outR, inP_, inQ_, invertQ);
+
+  outR.SimplifyTopology();
+
+  if (ManifoldParams().intermediateChecks)
+    ASSERT(outR.Is2Manifold(), logicErr, "simplified mesh is not 2-manifold!");
+
+#ifdef MANIFOLD_DEBUG
+  simplify.Stop();
+  Timer sort;
+  sort.Start();
+#endif
+
+  outR.Finish();
+  outR.IncrementMeshIDs();
+
+#ifdef MANIFOLD_DEBUG
+  sort.Stop();
+  if (ManifoldParams().verbose) {
+    assemble.Print("Assembly");
+    triangulate.Print("Triangulation");
+    simplify.Print("Simplification");
+    sort.Print("Sorting");
+    std::cout << outR.NumVert() << " verts and " << outR.NumTri() << " tris"
+              << std::endl;
+  }
+#endif
+
+  return outR;
+}
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/constructors.cpp b/thirdparty/manifold/src/manifold/src/constructors.cpp
new file mode 100644
index 000000000000..0350201661e9
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/constructors.cpp
@@ -0,0 +1,518 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <thrust/sequence.h>
+
+#include "cross_section.h"
+#include "csg_tree.h"
+#include "impl.h"
+#include "par.h"
+#include "polygon.h"
+
+namespace {
+using namespace manifold;
+using namespace thrust::placeholders;
+
+struct ToSphere {
+  float length;
+  void operator()(glm::vec3& v) {
+    v = glm::cos(glm::half_pi<float>() * (1.0f - v));
+    v = length * glm::normalize(v);
+    if (isnan(v.x)) v = glm::vec3(0.0);
+  }
+};
+
+struct Equals {
+  int val;
+  bool operator()(int x) { return x == val; }
+};
+
+struct RemoveFace {
+  VecView<const Halfedge> halfedge;
+  VecView<const int> vertLabel;
+  const int keepLabel;
+
+  bool operator()(int face) {
+    return vertLabel[halfedge[3 * face].startVert] != keepLabel;
+  }
+};
+}  // namespace
+
+namespace manifold {
+/**
+ * Constructs a smooth version of the input mesh by creating tangents; this
+ * method will throw if you have supplied tangents with your mesh already. The
+ * actual triangle resolution is unchanged; use the Refine() method to
+ * interpolate to a higher-resolution curve.
+ *
+ * By default, every edge is calculated for maximum smoothness (very much
+ * approximately), attempting to minimize the maximum mean Curvature magnitude.
+ * No higher-order derivatives are considered, as the interpolation is
+ * independent per triangle, only sharing constraints on their boundaries.
+ *
+ * @param meshGL input MeshGL.
+ * @param sharpenedEdges If desired, you can supply a vector of sharpened
+ * halfedges, which should in general be a small subset of all halfedges. Order
+ * of entries doesn't matter, as each one specifies the desired smoothness
+ * (between zero and one, with one the default for all unspecified halfedges)
+ * and the halfedge index (3 * triangle index + [0,1,2] where 0 is the edge
+ * between triVert 0 and 1, etc).
+ *
+ * At a smoothness value of zero, a sharp crease is made. The smoothness is
+ * interpolated along each edge, so the specified value should be thought of as
+ * an average. Where exactly two sharpened edges meet at a vertex, their
+ * tangents are rotated to be colinear so that the sharpened edge can be
+ * continuous. Vertices with only one sharpened edge are completely smooth,
+ * allowing sharpened edges to smoothly vanish at termination. A single vertex
+ * can be sharpened by sharping all edges that are incident on it, allowing
+ * cones to be formed.
+ */
+Manifold Manifold::Smooth(const MeshGL& meshGL,
+                          const std::vector<Smoothness>& sharpenedEdges) {
+  ASSERT(meshGL.halfedgeTangent.empty(), std::runtime_error,
+         "when supplying tangents, the normal constructor should be used "
+         "rather than Smooth().");
+
+  // Don't allow any triangle merging.
+  std::vector<float> propertyTolerance(meshGL.numProp - 3, -1);
+  std::shared_ptr<Impl> impl =
+      std::make_shared<Impl>(meshGL, propertyTolerance);
+  impl->CreateTangents(impl->UpdateSharpenedEdges(sharpenedEdges));
+  return Manifold(impl);
+}
+
+/**
+ * Constructs a smooth version of the input mesh by creating tangents; this
+ * method will throw if you have supplied tangents with your mesh already. The
+ * actual triangle resolution is unchanged; use the Refine() method to
+ * interpolate to a higher-resolution curve.
+ *
+ * By default, every edge is calculated for maximum smoothness (very much
+ * approximately), attempting to minimize the maximum mean Curvature magnitude.
+ * No higher-order derivatives are considered, as the interpolation is
+ * independent per triangle, only sharing constraints on their boundaries.
+ *
+ * @param mesh input Mesh.
+ * @param sharpenedEdges If desired, you can supply a vector of sharpened
+ * halfedges, which should in general be a small subset of all halfedges. Order
+ * of entries doesn't matter, as each one specifies the desired smoothness
+ * (between zero and one, with one the default for all unspecified halfedges)
+ * and the halfedge index (3 * triangle index + [0,1,2] where 0 is the edge
+ * between triVert 0 and 1, etc).
+ *
+ * At a smoothness value of zero, a sharp crease is made. The smoothness is
+ * interpolated along each edge, so the specified value should be thought of as
+ * an average. Where exactly two sharpened edges meet at a vertex, their
+ * tangents are rotated to be colinear so that the sharpened edge can be
+ * continuous. Vertices with only one sharpened edge are completely smooth,
+ * allowing sharpened edges to smoothly vanish at termination. A single vertex
+ * can be sharpened by sharping all edges that are incident on it, allowing
+ * cones to be formed.
+ */
+Manifold Manifold::Smooth(const Mesh& mesh,
+                          const std::vector<Smoothness>& sharpenedEdges) {
+  ASSERT(mesh.halfedgeTangent.empty(), std::runtime_error,
+         "when supplying tangents, the normal constructor should be used "
+         "rather than Smooth().");
+
+  Impl::MeshRelationD relation = {(int)ReserveIDs(1)};
+  std::shared_ptr<Impl> impl = std::make_shared<Impl>(mesh, relation);
+  impl->CreateTangents(impl->UpdateSharpenedEdges(sharpenedEdges));
+  return Manifold(impl);
+}
+
+/**
+ * Constructs a tetrahedron centered at the origin with one vertex at (1,1,1)
+ * and the rest at similarly symmetric points.
+ */
+Manifold Manifold::Tetrahedron() {
+  return Manifold(std::make_shared<Impl>(Impl::Shape::Tetrahedron));
+}
+
+/**
+ * Constructs a unit cube (edge lengths all one), by default in the first
+ * octant, touching the origin. If any dimensions in size are negative, or if
+ * all are zero, an empty Manifold will be returned.
+ *
+ * @param size The X, Y, and Z dimensions of the box.
+ * @param center Set to true to shift the center to the origin.
+ */
+Manifold Manifold::Cube(glm::vec3 size, bool center) {
+  if (size.x < 0.0f || size.y < 0.0f || size.z < 0.0f ||
+      glm::length(size) == 0.) {
+    return Invalid();
+  }
+  glm::mat4x3 m =
+      glm::translate(center ? (-size / 2.0f) : glm::vec3(0)) * glm::scale(size);
+  return Manifold(std::make_shared<Impl>(Manifold::Impl::Shape::Cube, m));
+}
+
+/**
+ * A convenience constructor for the common case of extruding a circle. Can also
+ * form cones if both radii are specified.
+ *
+ * @param height Z-extent
+ * @param radiusLow Radius of bottom circle. Must be positive.
+ * @param radiusHigh Radius of top circle. Can equal zero. Default is equal to
+ * radiusLow.
+ * @param circularSegments How many line segments to use around the circle.
+ * Default is calculated by the static Defaults.
+ * @param center Set to true to shift the center to the origin. Default is
+ * origin at the bottom.
+ */
+Manifold Manifold::Cylinder(float height, float radiusLow, float radiusHigh,
+                            int circularSegments, bool center) {
+  if (height <= 0.0f || radiusLow <= 0.0f) {
+    return Invalid();
+  }
+  float scale = radiusHigh >= 0.0f ? radiusHigh / radiusLow : 1.0f;
+  float radius = fmax(radiusLow, radiusHigh);
+  int n = circularSegments > 2 ? circularSegments
+                               : Quality::GetCircularSegments(radius);
+
+  CrossSection circle = CrossSection::Circle(radiusLow, n);
+  Manifold cylinder =
+      Manifold::Extrude(circle, height, 0, 0.0f, glm::vec2(scale));
+  if (center)
+    cylinder =
+        cylinder.Translate(glm::vec3(0.0f, 0.0f, -height / 2.0f)).AsOriginal();
+  return cylinder;
+}
+
+/**
+ * Constructs a geodesic sphere of a given radius.
+ *
+ * @param radius Radius of the sphere. Must be positive.
+ * @param circularSegments Number of segments along its
+ * diameter. This number will always be rounded up to the nearest factor of
+ * four, as this sphere is constructed by refining an octahedron. This means
+ * there are a circle of vertices on all three of the axis planes. Default is
+ * calculated by the static Defaults.
+ */
+Manifold Manifold::Sphere(float radius, int circularSegments) {
+  if (radius <= 0.0f) {
+    return Invalid();
+  }
+  int n = circularSegments > 0 ? (circularSegments + 3) / 4
+                               : Quality::GetCircularSegments(radius) / 4;
+  auto pImpl_ = std::make_shared<Impl>(Impl::Shape::Octahedron);
+  pImpl_->Subdivide([n](glm::vec3 edge) { return n - 1; });
+  for_each_n(autoPolicy(pImpl_->NumVert()), pImpl_->vertPos_.begin(),
+             pImpl_->NumVert(), ToSphere({radius}));
+  pImpl_->Finish();
+  // Ignore preceding octahedron.
+  pImpl_->InitializeOriginal();
+  return Manifold(pImpl_);
+}
+
+/**
+ * Constructs a manifold from a set of polygons by extruding them along the
+ * Z-axis.
+ * Note that high twistDegrees with small nDivisions may cause
+ * self-intersection. This is not checked here and it is up to the user to
+ * choose the correct parameters.
+ *
+ * @param crossSection A set of non-overlapping polygons to extrude.
+ * @param height Z-extent of extrusion.
+ * @param nDivisions Number of extra copies of the crossSection to insert into
+ * the shape vertically; especially useful in combination with twistDegrees to
+ * avoid interpolation artifacts. Default is none.
+ * @param twistDegrees Amount to twist the top crossSection relative to the
+ * bottom, interpolated linearly for the divisions in between.
+ * @param scaleTop Amount to scale the top (independently in X and Y). If the
+ * scale is {0, 0}, a pure cone is formed with only a single vertex at the top.
+ * Note that scale is applied after twist.
+ * Default {1, 1}.
+ */
+Manifold Manifold::Extrude(const CrossSection& crossSection, float height,
+                           int nDivisions, float twistDegrees,
+                           glm::vec2 scaleTop) {
+  ZoneScoped;
+  auto polygons = crossSection.ToPolygons();
+  if (polygons.size() == 0 || height <= 0.0f) {
+    return Invalid();
+  }
+
+  scaleTop.x = glm::max(scaleTop.x, 0.0f);
+  scaleTop.y = glm::max(scaleTop.y, 0.0f);
+
+  auto pImpl_ = std::make_shared<Impl>();
+  ++nDivisions;
+  auto& vertPos = pImpl_->vertPos_;
+  Vec<glm::ivec3> triVertsDH;
+  auto& triVerts = triVertsDH;
+  int nCrossSection = 0;
+  bool isCone = scaleTop.x == 0.0 && scaleTop.y == 0.0;
+  int idx = 0;
+  PolygonsIdx polygonsIndexed;
+  for (auto& poly : polygons) {
+    nCrossSection += poly.size();
+    SimplePolygonIdx simpleIndexed;
+    for (const glm::vec2& polyVert : poly) {
+      vertPos.push_back({polyVert.x, polyVert.y, 0.0f});
+      simpleIndexed.push_back({polyVert, idx++});
+    }
+    polygonsIndexed.push_back(simpleIndexed);
+  }
+  for (int i = 1; i < nDivisions + 1; ++i) {
+    float alpha = i / float(nDivisions);
+    float phi = alpha * twistDegrees;
+    glm::vec2 scale = glm::mix(glm::vec2(1.0f), scaleTop, alpha);
+    glm::mat2 rotation(cosd(phi), sind(phi), -sind(phi), cosd(phi));
+    glm::mat2 transform = glm::mat2(scale.x, 0.0f, 0.0f, scale.y) * rotation;
+    int j = 0;
+    int idx = 0;
+    for (const auto& poly : polygons) {
+      for (int vert = 0; vert < poly.size(); ++vert) {
+        int offset = idx + nCrossSection * i;
+        int thisVert = vert + offset;
+        int lastVert = (vert == 0 ? poly.size() : vert) - 1 + offset;
+        if (i == nDivisions && isCone) {
+          triVerts.push_back({nCrossSection * i + j, lastVert - nCrossSection,
+                              thisVert - nCrossSection});
+        } else {
+          glm::vec2 pos = transform * poly[vert];
+          vertPos.push_back({pos.x, pos.y, height * alpha});
+          triVerts.push_back({thisVert, lastVert, thisVert - nCrossSection});
+          triVerts.push_back(
+              {lastVert, lastVert - nCrossSection, thisVert - nCrossSection});
+        }
+      }
+      ++j;
+      idx += poly.size();
+    }
+  }
+  if (isCone)
+    for (int j = 0; j < polygons.size(); ++j)  // Duplicate vertex for Genus
+      vertPos.push_back({0.0f, 0.0f, height});
+  std::vector<glm::ivec3> top = TriangulateIdx(polygonsIndexed);
+  for (const glm::ivec3& tri : top) {
+    triVerts.push_back({tri[0], tri[2], tri[1]});
+    if (!isCone) triVerts.push_back(tri + nCrossSection * nDivisions);
+  }
+
+  pImpl_->CreateHalfedges(triVertsDH);
+  pImpl_->Finish();
+  pImpl_->meshRelation_.originalID = ReserveIDs(1);
+  pImpl_->InitializeOriginal();
+  pImpl_->CreateFaces();
+  return Manifold(pImpl_);
+}
+
+/**
+ * Constructs a manifold from a set of polygons by revolving this cross-section
+ * around its Y-axis and then setting this as the Z-axis of the resulting
+ * manifold. If the polygons cross the Y-axis, only the part on the positive X
+ * side is used. Geometrically valid input will result in geometrically valid
+ * output.
+ *
+ * @param crossSection A set of non-overlapping polygons to revolve.
+ * @param circularSegments Number of segments along its diameter. Default is
+ * calculated by the static Defaults.
+ * @param revolveDegrees Number of degrees to revolve. Default is 360 degrees.
+ */
+Manifold Manifold::Revolve(const CrossSection& crossSection,
+                           int circularSegments, float revolveDegrees) {
+  ZoneScoped;
+  Polygons polygons = crossSection.ToPolygons();
+
+  if (polygons.size() == 0) {
+    return Invalid();
+  }
+
+  const Rect bounds = crossSection.Bounds();
+  const float radius = bounds.max.x;
+
+  if (radius <= 0) {
+    return Invalid();
+  } else if (bounds.min.x < 0) {
+    // Take the x>=0 slice.
+    glm::vec2 min = bounds.min;
+    glm::vec2 max = bounds.max;
+    CrossSection posBoundingBox = CrossSection(
+        {{0.0, min.y}, {max.x, min.y}, {max.x, max.y}, {0.0, max.y}});
+
+    polygons = (crossSection ^ posBoundingBox).ToPolygons();
+  }
+
+  if (revolveDegrees > 360.0f) {
+    revolveDegrees = 360.0f;
+  }
+  const bool isFullRevolution = revolveDegrees == 360.0f;
+
+  const int nDivisions =
+      circularSegments > 2
+          ? circularSegments
+          : Quality::GetCircularSegments(radius) * revolveDegrees / 360;
+
+  auto pImpl_ = std::make_shared<Impl>();
+  auto& vertPos = pImpl_->vertPos_;
+  Vec<glm::ivec3> triVertsDH;
+  auto& triVerts = triVertsDH;
+
+  std::vector<int> startPoses;
+  std::vector<int> endPoses;
+
+  const float dPhi = revolveDegrees / nDivisions;
+  // first and last slice are distinguished if not a full revolution.
+  const int nSlices = isFullRevolution ? nDivisions : nDivisions + 1;
+
+  for (const auto& poly : polygons) {
+    std::size_t nPosVerts = 0;
+    std::size_t nRevolveAxisVerts = 0;
+    for (auto& pt : poly) {
+      if (pt.x > 0) {
+        nPosVerts++;
+      } else {
+        nRevolveAxisVerts++;
+      }
+    }
+
+    for (int polyVert = 0; polyVert < poly.size(); ++polyVert) {
+      const int startPosIndex = vertPos.size();
+
+      if (!isFullRevolution) startPoses.push_back(startPosIndex);
+
+      const glm::vec2 currPolyVertex = poly[polyVert];
+      const glm::vec2 prevPolyVertex =
+          poly[polyVert == 0 ? poly.size() - 1 : polyVert - 1];
+
+      const int prevStartPosIndex =
+          startPosIndex +
+          (polyVert == 0 ? nRevolveAxisVerts + (nSlices * nPosVerts) : 0) +
+          (prevPolyVertex.x == 0.0 ? -1 : -nSlices);
+
+      for (int slice = 0; slice < nSlices; ++slice) {
+        const float phi = slice * dPhi;
+        if (slice == 0 || currPolyVertex.x > 0) {
+          vertPos.push_back({currPolyVertex.x * cosd(phi),
+                             currPolyVertex.x * sind(phi), currPolyVertex.y});
+        }
+
+        if (isFullRevolution || slice > 0) {
+          const int lastSlice = (slice == 0 ? nDivisions : slice) - 1;
+          if (currPolyVertex.x > 0.0) {
+            triVerts.push_back(
+                {startPosIndex + slice, startPosIndex + lastSlice,
+                 // "Reuse" vertex of first slice if it lies on the revolve axis
+                 (prevPolyVertex.x == 0.0 ? prevStartPosIndex
+                                          : prevStartPosIndex + lastSlice)});
+          }
+
+          if (prevPolyVertex.x > 0.0) {
+            triVerts.push_back(
+                {prevStartPosIndex + lastSlice, prevStartPosIndex + slice,
+                 (currPolyVertex.x == 0.0 ? startPosIndex
+                                          : startPosIndex + slice)});
+          }
+        }
+      }
+      if (!isFullRevolution) endPoses.push_back(vertPos.size() - 1);
+    }
+  }
+
+  // Add front and back triangles if not a full revolution.
+  if (!isFullRevolution) {
+    std::vector<glm::ivec3> frontTriangles =
+        Triangulate(polygons, pImpl_->precision_);
+    for (auto& t : frontTriangles) {
+      triVerts.push_back({startPoses[t.x], startPoses[t.y], startPoses[t.z]});
+    }
+
+    for (auto& t : frontTriangles) {
+      triVerts.push_back({endPoses[t.z], endPoses[t.y], endPoses[t.x]});
+    }
+  }
+
+  pImpl_->CreateHalfedges(triVertsDH);
+  pImpl_->Finish();
+  pImpl_->meshRelation_.originalID = ReserveIDs(1);
+  pImpl_->InitializeOriginal();
+  pImpl_->CreateFaces();
+  return Manifold(pImpl_);
+}
+
+/**
+ * Constructs a new manifold from a vector of other manifolds. This is a purely
+ * topological operation, so care should be taken to avoid creating
+ * overlapping results. It is the inverse operation of Decompose().
+ *
+ * @param manifolds A vector of Manifolds to lazy-union together.
+ */
+Manifold Manifold::Compose(const std::vector<Manifold>& manifolds) {
+  std::vector<std::shared_ptr<CsgLeafNode>> children;
+  for (const auto& manifold : manifolds) {
+    children.push_back(manifold.pNode_->ToLeafNode());
+  }
+  return Manifold(std::make_shared<Impl>(CsgLeafNode::Compose(children)));
+}
+
+/**
+ * This operation returns a vector of Manifolds that are topologically
+ * disconnected. If everything is connected, the vector is length one,
+ * containing a copy of the original. It is the inverse operation of Compose().
+ */
+std::vector<Manifold> Manifold::Decompose() const {
+  ZoneScoped;
+  UnionFind<> uf(NumVert());
+  // Graph graph;
+  auto pImpl_ = GetCsgLeafNode().GetImpl();
+  for (const Halfedge& halfedge : pImpl_->halfedge_) {
+    if (halfedge.IsForward()) uf.unionXY(halfedge.startVert, halfedge.endVert);
+  }
+  std::vector<int> componentIndices;
+  const int numComponents = uf.connectedComponents(componentIndices);
+
+  if (numComponents == 1) {
+    std::vector<Manifold> meshes(1);
+    meshes[0] = *this;
+    return meshes;
+  }
+  Vec<int> vertLabel(componentIndices);
+
+  std::vector<Manifold> meshes;
+  for (int i = 0; i < numComponents; ++i) {
+    auto impl = std::make_shared<Impl>();
+    // inherit original object's precision
+    impl->precision_ = pImpl_->precision_;
+    impl->vertPos_.resize(NumVert());
+    Vec<int> vertNew2Old(NumVert());
+    auto policy = autoPolicy(NumVert());
+    auto start = zip(impl->vertPos_.begin(), vertNew2Old.begin());
+    int nVert =
+        copy_if<decltype(start)>(
+            policy, zip(pImpl_->vertPos_.begin(), countAt(0)),
+            zip(pImpl_->vertPos_.end(), countAt(NumVert())), vertLabel.begin(),
+            zip(impl->vertPos_.begin(), vertNew2Old.begin()), Equals({i})) -
+        start;
+    impl->vertPos_.resize(nVert);
+
+    Vec<int> faceNew2Old(NumTri());
+    sequence(policy, faceNew2Old.begin(), faceNew2Old.end());
+
+    int nFace = remove_if<decltype(faceNew2Old.begin())>(
+                    policy, faceNew2Old.begin(), faceNew2Old.end(),
+                    RemoveFace({pImpl_->halfedge_, vertLabel, i})) -
+                faceNew2Old.begin();
+    faceNew2Old.resize(nFace);
+
+    impl->GatherFaces(*pImpl_, faceNew2Old);
+    impl->ReindexVerts(vertNew2Old, pImpl_->NumVert());
+    impl->Finish();
+
+    meshes.push_back(Manifold(impl));
+  }
+  return meshes;
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/csg_tree.cpp b/thirdparty/manifold/src/manifold/src/csg_tree.cpp
new file mode 100644
index 000000000000..f6c994472dd3
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/csg_tree.cpp
@@ -0,0 +1,651 @@
+// Copyright 2022 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#if MANIFOLD_PAR == 'T' && __has_include(<tbb/concurrent_priority_queue.h>)
+#include <tbb/tbb.h>
+#define TBB_PREVIEW_CONCURRENT_ORDERED_CONTAINERS 1
+#include <tbb/concurrent_priority_queue.h>
+#endif
+
+#include <algorithm>
+#include <variant>
+
+#include "boolean3.h"
+#include "csg_tree.h"
+#include "impl.h"
+#include "mesh_fixes.h"
+#include "par.h"
+
+constexpr int kParallelThreshold = 4096;
+
+namespace {
+using namespace manifold;
+struct Transform4x3 {
+  const glm::mat4x3 transform;
+
+  glm::vec3 operator()(glm::vec3 position) {
+    return transform * glm::vec4(position, 1.0f);
+  }
+};
+
+struct UpdateHalfedge {
+  const int nextVert;
+  const int nextEdge;
+  const int nextFace;
+
+  Halfedge operator()(Halfedge edge) {
+    edge.startVert += nextVert;
+    edge.endVert += nextVert;
+    edge.pairedHalfedge += nextEdge;
+    edge.face += nextFace;
+    return edge;
+  }
+};
+
+struct UpdateTriProp {
+  const int nextProp;
+
+  glm::ivec3 operator()(glm::ivec3 tri) {
+    tri += nextProp;
+    return tri;
+  }
+};
+
+struct UpdateMeshIDs {
+  const int offset;
+
+  TriRef operator()(TriRef ref) {
+    ref.meshID += offset;
+    return ref;
+  }
+};
+
+struct CheckOverlap {
+  VecView<const Box> boxes;
+  const size_t i;
+  bool operator()(int j) { return boxes[i].DoesOverlap(boxes[j]); }
+};
+
+using SharedImpl = std::variant<std::shared_ptr<const Manifold::Impl>,
+                                std::shared_ptr<Manifold::Impl>>;
+struct GetImplPtr {
+  const Manifold::Impl *operator()(const SharedImpl &p) {
+    if (std::holds_alternative<std::shared_ptr<const Manifold::Impl>>(p)) {
+      return std::get_if<std::shared_ptr<const Manifold::Impl>>(&p)->get();
+    } else {
+      return std::get_if<std::shared_ptr<Manifold::Impl>>(&p)->get();
+    }
+  };
+};
+
+struct MeshCompare {
+  bool operator()(const SharedImpl &a, const SharedImpl &b) {
+    return GetImplPtr()(a)->NumVert() < GetImplPtr()(b)->NumVert();
+  }
+};
+
+}  // namespace
+namespace manifold {
+
+std::shared_ptr<CsgNode> CsgNode::Boolean(
+    const std::shared_ptr<CsgNode> &second, OpType op) {
+  if (auto opNode = std::dynamic_pointer_cast<CsgOpNode>(second)) {
+    // "this" is not a CsgOpNode (which overrides Boolean), but if "second" is
+    // and the operation is commutative, we let it built the tree.
+    if ((op == OpType::Add || op == OpType::Intersect)) {
+      return opNode->Boolean(shared_from_this(), op);
+    }
+  }
+  std::vector<std::shared_ptr<CsgNode>> children({shared_from_this(), second});
+  return std::make_shared<CsgOpNode>(children, op);
+}
+
+std::shared_ptr<CsgNode> CsgNode::Translate(const glm::vec3 &t) const {
+  glm::mat4x3 transform(1.0f);
+  transform[3] += t;
+  return Transform(transform);
+}
+
+std::shared_ptr<CsgNode> CsgNode::Scale(const glm::vec3 &v) const {
+  glm::mat4x3 transform(1.0f);
+  for (int i : {0, 1, 2}) transform[i] *= v;
+  return Transform(transform);
+}
+
+std::shared_ptr<CsgNode> CsgNode::Rotate(float xDegrees, float yDegrees,
+                                         float zDegrees) const {
+  glm::mat3 rX(1.0f, 0.0f, 0.0f,                      //
+               0.0f, cosd(xDegrees), sind(xDegrees),  //
+               0.0f, -sind(xDegrees), cosd(xDegrees));
+  glm::mat3 rY(cosd(yDegrees), 0.0f, -sind(yDegrees),  //
+               0.0f, 1.0f, 0.0f,                       //
+               sind(yDegrees), 0.0f, cosd(yDegrees));
+  glm::mat3 rZ(cosd(zDegrees), sind(zDegrees), 0.0f,   //
+               -sind(zDegrees), cosd(zDegrees), 0.0f,  //
+               0.0f, 0.0f, 1.0f);
+  glm::mat3 result = rZ * rY * rX;
+  glm::mat4x3 transform(result[0], result[1], result[2], glm::vec3(0.0f, 0.0f, 0.0f));
+  return Transform(transform);
+}
+
+CsgLeafNode::CsgLeafNode() : pImpl_(std::make_shared<Manifold::Impl>()) {}
+
+CsgLeafNode::CsgLeafNode(std::shared_ptr<const Manifold::Impl> pImpl_)
+    : pImpl_(pImpl_) {}
+
+CsgLeafNode::CsgLeafNode(std::shared_ptr<const Manifold::Impl> pImpl_,
+                         glm::mat4x3 transform_)
+    : pImpl_(pImpl_), transform_(transform_) {}
+
+std::shared_ptr<const Manifold::Impl> CsgLeafNode::GetImpl() const {
+  if (transform_ == glm::mat4x3(1.0f)) return pImpl_;
+  pImpl_ =
+      std::make_shared<const Manifold::Impl>(pImpl_->Transform(transform_));
+  transform_ = glm::mat4x3(1.0f);
+  return pImpl_;
+}
+
+glm::mat4x3 CsgLeafNode::GetTransform() const { return transform_; }
+
+std::shared_ptr<CsgLeafNode> CsgLeafNode::ToLeafNode() const {
+  return std::make_shared<CsgLeafNode>(*this);
+}
+
+std::shared_ptr<CsgNode> CsgLeafNode::Transform(const glm::mat4x3 &m) const {
+  return std::make_shared<CsgLeafNode>(pImpl_, m * glm::mat4(transform_));
+}
+
+CsgNodeType CsgLeafNode::GetNodeType() const { return CsgNodeType::Leaf; }
+
+/**
+ * Efficient union of a set of pairwise disjoint meshes.
+ */
+Manifold::Impl CsgLeafNode::Compose(
+    const std::vector<std::shared_ptr<CsgLeafNode>> &nodes) {
+  ZoneScoped;
+  float precision = -1;
+  int numVert = 0;
+  int numEdge = 0;
+  int numTri = 0;
+  int numPropVert = 0;
+  std::vector<int> vertIndices;
+  std::vector<int> edgeIndices;
+  std::vector<int> triIndices;
+  std::vector<int> propVertIndices;
+  int numPropOut = 0;
+  for (auto &node : nodes) {
+    float nodeOldScale = node->pImpl_->bBox_.Scale();
+    float nodeNewScale =
+        node->pImpl_->bBox_.Transform(node->transform_).Scale();
+    float nodePrecision = node->pImpl_->precision_;
+    nodePrecision *= glm::max(1.0f, nodeNewScale / nodeOldScale);
+    nodePrecision = glm::max(nodePrecision, kTolerance * nodeNewScale);
+    if (!glm::isfinite(nodePrecision)) nodePrecision = -1;
+    precision = glm::max(precision, nodePrecision);
+
+    vertIndices.push_back(numVert);
+    edgeIndices.push_back(numEdge * 2);
+    triIndices.push_back(numTri);
+    propVertIndices.push_back(numPropVert);
+    numVert += node->pImpl_->NumVert();
+    numEdge += node->pImpl_->NumEdge();
+    numTri += node->pImpl_->NumTri();
+    const int numProp = node->pImpl_->NumProp();
+    numPropOut = glm::max(numPropOut, numProp);
+    numPropVert +=
+        numProp == 0 ? 1
+                     : node->pImpl_->meshRelation_.properties.size() / numProp;
+  }
+
+  Manifold::Impl combined;
+  combined.precision_ = precision;
+  combined.vertPos_.resize(numVert);
+  combined.halfedge_.resize(2 * numEdge);
+  combined.faceNormal_.resize(numTri);
+  combined.halfedgeTangent_.resize(2 * numEdge);
+  combined.meshRelation_.triRef.resize(numTri);
+  if (numPropOut > 0) {
+    combined.meshRelation_.numProp = numPropOut;
+    combined.meshRelation_.properties.resize(numPropOut * numPropVert, 0);
+    combined.meshRelation_.triProperties.resize(numTri);
+  }
+  auto policy = autoPolicy(numTri);
+
+  // if we are already parallelizing for each node, do not perform multithreaded
+  // copying as it will slightly hurt performance
+  if (nodes.size() > 1 && policy == ExecutionPolicy::Par)
+    policy = ExecutionPolicy::Seq;
+
+  for_each_n(
+      nodes.size() > 1 ? ExecutionPolicy::Par : ExecutionPolicy::Seq,
+      countAt(0), nodes.size(),
+      [&nodes, &vertIndices, &edgeIndices, &triIndices, &propVertIndices,
+       numPropOut, &combined, policy](int i) {
+        auto &node = nodes[i];
+        copy(policy, node->pImpl_->halfedgeTangent_.begin(),
+             node->pImpl_->halfedgeTangent_.end(),
+             combined.halfedgeTangent_.begin() + edgeIndices[i]);
+        transform(
+            policy, node->pImpl_->halfedge_.begin(),
+            node->pImpl_->halfedge_.end(),
+            combined.halfedge_.begin() + edgeIndices[i],
+            UpdateHalfedge({vertIndices[i], edgeIndices[i], triIndices[i]}));
+
+        if (numPropOut > 0) {
+          auto start =
+              combined.meshRelation_.triProperties.begin() + triIndices[i];
+          if (node->pImpl_->NumProp() > 0) {
+            auto &triProp = node->pImpl_->meshRelation_.triProperties;
+            transform(policy, triProp.begin(), triProp.end(), start,
+                      UpdateTriProp({propVertIndices[i]}));
+
+            const int numProp = node->pImpl_->NumProp();
+            auto &oldProp = node->pImpl_->meshRelation_.properties;
+            auto &newProp = combined.meshRelation_.properties;
+            for (int p = 0; p < numProp; ++p) {
+              strided_range<Vec<float>::IterC> oldRange(oldProp.begin() + p,
+                                                        oldProp.end(), numProp);
+              strided_range<Vec<float>::Iter> newRange(
+                  newProp.begin() + numPropOut * propVertIndices[i] + p,
+                  newProp.end(), numPropOut);
+              copy(policy, oldRange.begin(), oldRange.end(), newRange.begin());
+            }
+          } else {
+            // point all triangles at single new property of zeros.
+            fill(policy, start, start + node->pImpl_->NumTri(),
+                 glm::ivec3(propVertIndices[i]));
+          }
+        }
+
+        if (node->transform_ == glm::mat4x3(1.0f)) {
+          copy(policy, node->pImpl_->vertPos_.begin(),
+               node->pImpl_->vertPos_.end(),
+               combined.vertPos_.begin() + vertIndices[i]);
+          copy(policy, node->pImpl_->faceNormal_.begin(),
+               node->pImpl_->faceNormal_.end(),
+               combined.faceNormal_.begin() + triIndices[i]);
+        } else {
+          // no need to apply the transform to the node, just copy the vertices
+          // and face normals and apply transform on the fly
+          auto vertPosBegin = thrust::make_transform_iterator(
+              node->pImpl_->vertPos_.begin(), Transform4x3({node->transform_}));
+          glm::mat3 normalTransform =
+              glm::inverse(glm::transpose(glm::mat3(node->transform_)));
+          auto faceNormalBegin = thrust::make_transform_iterator(
+              node->pImpl_->faceNormal_.begin(),
+              TransformNormals({normalTransform}));
+          copy_n(policy, vertPosBegin, node->pImpl_->vertPos_.size(),
+                 combined.vertPos_.begin() + vertIndices[i]);
+          copy_n(policy, faceNormalBegin, node->pImpl_->faceNormal_.size(),
+                 combined.faceNormal_.begin() + triIndices[i]);
+
+          const bool invert = glm::determinant(glm::mat3(node->transform_)) < 0;
+          for_each_n(policy,
+                     zip(combined.halfedgeTangent_.begin() + edgeIndices[i],
+                         countAt(0)),
+                     node->pImpl_->halfedgeTangent_.size(),
+                     TransformTangents{glm::mat3(node->transform_), invert,
+                                       node->pImpl_->halfedgeTangent_,
+                                       node->pImpl_->halfedge_});
+          if (invert)
+            for_each_n(policy,
+                       zip(combined.meshRelation_.triRef.begin(),
+                           countAt(triIndices[i])),
+                       node->pImpl_->NumTri(), FlipTris({combined.halfedge_}));
+        }
+        // Since the nodes may be copies containing the same meshIDs, it is
+        // important to add an offset so that each node instance gets
+        // unique meshIDs.
+        const int offset = i * Manifold::Impl::meshIDCounter_;
+        transform(policy, node->pImpl_->meshRelation_.triRef.begin(),
+                  node->pImpl_->meshRelation_.triRef.end(),
+                  combined.meshRelation_.triRef.begin() + triIndices[i],
+                  UpdateMeshIDs({offset}));
+      });
+
+  for (int i = 0; i < nodes.size(); i++) {
+    auto &node = nodes[i];
+    const int offset = i * Manifold::Impl::meshIDCounter_;
+
+    for (const auto pair : node->pImpl_->meshRelation_.meshIDtransform) {
+      combined.meshRelation_.meshIDtransform[pair.first + offset] = pair.second;
+    }
+  }
+
+  // required to remove parts that are smaller than the precision
+  combined.SimplifyTopology();
+  combined.Finish();
+  combined.IncrementMeshIDs();
+  return combined;
+}
+
+CsgOpNode::CsgOpNode() {}
+
+CsgOpNode::CsgOpNode(const std::vector<std::shared_ptr<CsgNode>> &children,
+                     OpType op)
+    : impl_(Impl{}) {
+  auto impl = impl_.GetGuard();
+  impl->children_ = children;
+  SetOp(op);
+}
+
+CsgOpNode::CsgOpNode(std::vector<std::shared_ptr<CsgNode>> &&children,
+                     OpType op)
+    : impl_(Impl{}) {
+  auto impl = impl_.GetGuard();
+  impl->children_ = children;
+  SetOp(op);
+}
+
+std::shared_ptr<CsgNode> CsgOpNode::Boolean(
+    const std::shared_ptr<CsgNode> &second, OpType op) {
+  std::vector<std::shared_ptr<CsgNode>> children;
+
+  auto isReused = [](const auto &node) { return node->impl_.UseCount() > 1; };
+
+  auto copyChildren = [&](const auto &list, const glm::mat4x3 &transform) {
+    for (const auto &child : list) {
+      children.push_back(child->Transform(transform));
+    }
+  };
+
+  auto self = std::dynamic_pointer_cast<CsgOpNode>(shared_from_this());
+  assert(self);
+  if (IsOp(op) && !isReused(self)) {
+    auto impl = impl_.GetGuard();
+    copyChildren(impl->children_, transform_);
+  } else {
+    children.push_back(self);
+  }
+
+  auto secondOp = std::dynamic_pointer_cast<CsgOpNode>(second);
+  auto canInlineSecondOp = [&]() {
+    switch (op) {
+      case OpType::Add:
+      case OpType::Intersect:
+        return secondOp->IsOp(op);
+      case OpType::Subtract:
+        return secondOp->IsOp(OpType::Add);
+      default:
+        return false;
+    }
+  };
+
+  if (secondOp && canInlineSecondOp() && !isReused(secondOp)) {
+    auto secondImpl = secondOp->impl_.GetGuard();
+    copyChildren(secondImpl->children_, secondOp->transform_);
+  } else {
+    children.push_back(second);
+  }
+
+  return std::make_shared<CsgOpNode>(children, op);
+}
+
+std::shared_ptr<CsgNode> CsgOpNode::Transform(const glm::mat4x3 &m) const {
+  auto node = std::make_shared<CsgOpNode>();
+  node->impl_ = impl_;
+  node->transform_ = m * glm::mat4(transform_);
+  node->op_ = op_;
+  return node;
+}
+
+std::shared_ptr<CsgLeafNode> CsgOpNode::ToLeafNode() const {
+  if (cache_ != nullptr) return cache_;
+  // turn the children into leaf nodes
+  GetChildren();
+  auto impl = impl_.GetGuard();
+  auto &children_ = impl->children_;
+  if (children_.size() > 1) {
+    switch (op_) {
+      case CsgNodeType::Union:
+        BatchUnion();
+        break;
+      case CsgNodeType::Intersection: {
+        std::vector<std::shared_ptr<const Manifold::Impl>> impls;
+        for (auto &child : children_) {
+          impls.push_back(
+              std::dynamic_pointer_cast<CsgLeafNode>(child)->GetImpl());
+        }
+        children_.clear();
+        children_.push_back(std::make_shared<CsgLeafNode>(
+            BatchBoolean(OpType::Intersect, impls)));
+        break;
+      };
+      case CsgNodeType::Difference: {
+        // take the lhs out and treat the remaining nodes as the rhs, perform
+        // union optimization for them
+        auto lhs = std::dynamic_pointer_cast<CsgLeafNode>(children_.front());
+        children_.erase(children_.begin());
+        BatchUnion();
+        auto rhs = std::dynamic_pointer_cast<CsgLeafNode>(children_.front());
+        children_.clear();
+        Boolean3 boolean(*lhs->GetImpl(), *rhs->GetImpl(), OpType::Subtract);
+        children_.push_back(
+            std::make_shared<CsgLeafNode>(std::make_shared<Manifold::Impl>(
+                boolean.Result(OpType::Subtract))));
+      };
+      case CsgNodeType::Leaf:
+        // unreachable
+        break;
+    }
+  } else if (children_.size() == 0) {
+    return nullptr;
+  }
+  // children_ must contain only one CsgLeafNode now, and its Transform will
+  // give CsgLeafNode as well
+  cache_ = std::dynamic_pointer_cast<CsgLeafNode>(
+      children_.front()->Transform(transform_));
+  return cache_;
+}
+
+/**
+ * Efficient boolean operation on a set of nodes utilizing commutativity of the
+ * operation. Only supports union and intersection.
+ */
+std::shared_ptr<Manifold::Impl> CsgOpNode::BatchBoolean(
+    OpType operation,
+    std::vector<std::shared_ptr<const Manifold::Impl>> &results) {
+  ZoneScoped;
+  auto getImplPtr = GetImplPtr();
+  ASSERT(operation != OpType::Subtract, logicErr,
+         "BatchBoolean doesn't support Difference.");
+  // common cases
+  if (results.size() == 0) return std::make_shared<Manifold::Impl>();
+  if (results.size() == 1)
+    return std::make_shared<Manifold::Impl>(*results.front());
+  if (results.size() == 2) {
+    Boolean3 boolean(*results[0], *results[1], operation);
+    return std::make_shared<Manifold::Impl>(boolean.Result(operation));
+  }
+#if MANIFOLD_PAR == 'T' && __has_include(<tbb/tbb.h>)
+  if (!ManifoldParams().deterministic) {
+    tbb::task_group group;
+    tbb::concurrent_priority_queue<SharedImpl, MeshCompare> queue(
+        results.size());
+    for (auto result : results) {
+      queue.emplace(result);
+    }
+    results.clear();
+    std::function<void()> process = [&]() {
+      while (queue.size() > 1) {
+        SharedImpl a, b;
+        if (!queue.try_pop(a)) continue;
+        if (!queue.try_pop(b)) {
+          queue.push(a);
+          continue;
+        }
+        group.run([&, a, b]() {
+          const Manifold::Impl *aImpl;
+          const Manifold::Impl *bImpl;
+          Boolean3 boolean(*getImplPtr(a), *getImplPtr(b), operation);
+          queue.emplace(
+              std::make_shared<Manifold::Impl>(boolean.Result(operation)));
+          return group.run(process);
+        });
+      }
+    };
+    group.run_and_wait(process);
+    SharedImpl r;
+    queue.try_pop(r);
+    return *std::get_if<std::shared_ptr<Manifold::Impl>>(&r);
+  }
+#endif
+  // apply boolean operations starting from smaller meshes
+  // the assumption is that boolean operations on smaller meshes is faster,
+  // due to less data being copied and processed
+  auto cmpFn = MeshCompare();
+  std::make_heap(results.begin(), results.end(), cmpFn);
+  while (results.size() > 1) {
+    std::pop_heap(results.begin(), results.end(), cmpFn);
+    auto a = std::move(results.back());
+    results.pop_back();
+    std::pop_heap(results.begin(), results.end(), cmpFn);
+    auto b = std::move(results.back());
+    results.pop_back();
+    // boolean operation
+    Boolean3 boolean(*a, *b, operation);
+    if (results.size() == 0) {
+      return std::make_shared<Manifold::Impl>(boolean.Result(operation));
+    }
+    results.push_back(
+        std::make_shared<const Manifold::Impl>(boolean.Result(operation)));
+    std::push_heap(results.begin(), results.end(), cmpFn);
+  }
+  return std::make_shared<Manifold::Impl>(*results.front());
+}
+
+/**
+ * Efficient union operation on a set of nodes by doing Compose as much as
+ * possible.
+ * Note: Due to some unknown issues with `Compose`, we are now doing
+ * `BatchBoolean` instead of using `Compose` for non-intersecting manifolds.
+ */
+void CsgOpNode::BatchUnion() const {
+  ZoneScoped;
+  // INVARIANT: children_ is a vector of leaf nodes
+  // this kMaxUnionSize is a heuristic to avoid the pairwise disjoint check
+  // with O(n^2) complexity to take too long.
+  // If the number of children exceeded this limit, we will operate on chunks
+  // with size kMaxUnionSize.
+  constexpr int kMaxUnionSize = 1000;
+  auto impl = impl_.GetGuard();
+  auto &children_ = impl->children_;
+  while (children_.size() > 1) {
+    const int start = (children_.size() > kMaxUnionSize)
+                          ? (children_.size() - kMaxUnionSize)
+                          : 0;
+    Vec<Box> boxes;
+    boxes.reserve(children_.size() - start);
+    for (int i = start; i < children_.size(); i++) {
+      boxes.push_back(std::dynamic_pointer_cast<CsgLeafNode>(children_[i])
+                          ->GetImpl()
+                          ->bBox_);
+    }
+    // partition the children into a set of disjoint sets
+    // each set contains a set of children that are pairwise disjoint
+    std::vector<Vec<size_t>> disjointSets;
+    for (size_t i = 0; i < boxes.size(); i++) {
+      auto lambda = [&boxes, i](const Vec<size_t> &set) {
+        return std::find_if(set.begin(), set.end(), CheckOverlap({boxes, i})) ==
+               set.end();
+      };
+      auto it = std::find_if(disjointSets.begin(), disjointSets.end(), lambda);
+      if (it == disjointSets.end()) {
+        disjointSets.push_back(std::vector<size_t>{i});
+      } else {
+        it->push_back(i);
+      }
+    }
+    // compose each set of disjoint children
+    std::vector<std::shared_ptr<const Manifold::Impl>> impls;
+    for (auto &set : disjointSets) {
+      if (set.size() == 1) {
+        impls.push_back(
+            std::dynamic_pointer_cast<CsgLeafNode>(children_[start + set[0]])
+                ->GetImpl());
+      } else {
+        std::vector<std::shared_ptr<CsgLeafNode>> tmp;
+        for (size_t j : set) {
+          tmp.push_back(
+              std::dynamic_pointer_cast<CsgLeafNode>(children_[start + j]));
+        }
+        impls.push_back(
+            std::make_shared<const Manifold::Impl>(CsgLeafNode::Compose(tmp)));
+      }
+    }
+
+    children_.erase(children_.begin() + start, children_.end());
+    children_.push_back(
+        std::make_shared<CsgLeafNode>(BatchBoolean(OpType::Add, impls)));
+    // move it to the front as we process from the back, and the newly added
+    // child should be quite complicated
+    std::swap(children_.front(), children_.back());
+  }
+}
+
+/**
+ * Flatten the children to a list of leaf nodes and return them.
+ * If forceToLeafNodes is true, the list will be guaranteed to be a list of leaf
+ * nodes (i.e. no ops). Otherwise, the list may contain ops. Note that this
+ * function will not apply the transform to children, as they may be shared with
+ * other nodes.
+ */
+std::vector<std::shared_ptr<CsgNode>> &CsgOpNode::GetChildren(
+    bool forceToLeafNodes) const {
+  auto impl = impl_.GetGuard();
+
+  if (forceToLeafNodes && !impl->forcedToLeafNodes_) {
+    impl->forcedToLeafNodes_ = true;
+    for_each(impl->children_.size() > 1 && !ManifoldParams().deterministic
+                 ? ExecutionPolicy::Par
+                 : ExecutionPolicy::Seq,
+             impl->children_.begin(), impl->children_.end(), [](auto &child) {
+               if (child->GetNodeType() != CsgNodeType::Leaf) {
+                 child = child->ToLeafNode();
+               }
+             });
+  }
+  return impl->children_;
+}
+
+void CsgOpNode::SetOp(OpType op) {
+  switch (op) {
+    case OpType::Add:
+      op_ = CsgNodeType::Union;
+      break;
+    case OpType::Subtract:
+      op_ = CsgNodeType::Difference;
+      break;
+    case OpType::Intersect:
+      op_ = CsgNodeType::Intersection;
+      break;
+  }
+}
+
+bool CsgOpNode::IsOp(OpType op) {
+  switch (op) {
+    case OpType::Add:
+      return op_ == CsgNodeType::Union;
+    case OpType::Subtract:
+      return op_ == CsgNodeType::Difference;
+    case OpType::Intersect:
+      return op_ == CsgNodeType::Intersection;
+    default:
+      return false;
+  }
+}
+
+glm::mat4x3 CsgOpNode::GetTransform() const { return transform_; }
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/csg_tree.h b/thirdparty/manifold/src/manifold/src/csg_tree.h
new file mode 100644
index 000000000000..5acbd25fa297
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/csg_tree.h
@@ -0,0 +1,109 @@
+// Copyright 2022 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include "manifold.h"
+#include "utils.h"
+
+namespace manifold {
+
+enum class CsgNodeType { Union, Intersection, Difference, Leaf };
+
+class CsgLeafNode;
+
+class CsgNode : public std::enable_shared_from_this<CsgNode> {
+ public:
+  virtual std::shared_ptr<CsgLeafNode> ToLeafNode() const = 0;
+  virtual std::shared_ptr<CsgNode> Transform(const glm::mat4x3 &m) const = 0;
+  virtual CsgNodeType GetNodeType() const = 0;
+  virtual glm::mat4x3 GetTransform() const = 0;
+
+  virtual std::shared_ptr<CsgNode> Boolean(
+      const std::shared_ptr<CsgNode> &second, OpType op);
+
+  std::shared_ptr<CsgNode> Translate(const glm::vec3 &t) const;
+  std::shared_ptr<CsgNode> Scale(const glm::vec3 &s) const;
+  std::shared_ptr<CsgNode> Rotate(float xDegrees = 0, float yDegrees = 0,
+                                  float zDegrees = 0) const;
+};
+
+class CsgLeafNode final : public CsgNode {
+ public:
+  CsgLeafNode();
+  CsgLeafNode(std::shared_ptr<const Manifold::Impl> pImpl_);
+  CsgLeafNode(std::shared_ptr<const Manifold::Impl> pImpl_,
+              glm::mat4x3 transform_);
+
+  std::shared_ptr<const Manifold::Impl> GetImpl() const;
+
+  std::shared_ptr<CsgLeafNode> ToLeafNode() const override;
+
+  std::shared_ptr<CsgNode> Transform(const glm::mat4x3 &m) const override;
+
+  CsgNodeType GetNodeType() const override;
+
+  glm::mat4x3 GetTransform() const override;
+
+  static Manifold::Impl Compose(
+      const std::vector<std::shared_ptr<CsgLeafNode>> &nodes);
+
+ private:
+  mutable std::shared_ptr<const Manifold::Impl> pImpl_;
+  mutable glm::mat4x3 transform_ = glm::mat4x3(1.0f);
+};
+
+class CsgOpNode final : public CsgNode {
+ public:
+  CsgOpNode();
+
+  CsgOpNode(const std::vector<std::shared_ptr<CsgNode>> &children, OpType op);
+
+  CsgOpNode(std::vector<std::shared_ptr<CsgNode>> &&children, OpType op);
+
+  std::shared_ptr<CsgNode> Boolean(const std::shared_ptr<CsgNode> &second,
+                                   OpType op) override;
+
+  std::shared_ptr<CsgNode> Transform(const glm::mat4x3 &m) const override;
+
+  std::shared_ptr<CsgLeafNode> ToLeafNode() const override;
+
+  CsgNodeType GetNodeType() const override { return op_; }
+
+  glm::mat4x3 GetTransform() const override;
+
+ private:
+  struct Impl {
+    std::vector<std::shared_ptr<CsgNode>> children_;
+    bool forcedToLeafNodes_ = false;
+  };
+  mutable ConcurrentSharedPtr<Impl> impl_ = ConcurrentSharedPtr<Impl>(Impl{});
+  CsgNodeType op_;
+  glm::mat4x3 transform_ = glm::mat4x3(1.0f);
+  // the following fields are for lazy evaluation, so they are mutable
+  mutable std::shared_ptr<CsgLeafNode> cache_ = nullptr;
+
+  void SetOp(OpType);
+  bool IsOp(OpType op);
+
+  static std::shared_ptr<Manifold::Impl> BatchBoolean(
+      OpType operation,
+      std::vector<std::shared_ptr<const Manifold::Impl>> &results);
+
+  void BatchUnion() const;
+
+  std::vector<std::shared_ptr<CsgNode>> &GetChildren(
+      bool forceToLeafNodes = true) const;
+};
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/edge_op.cpp b/thirdparty/manifold/src/manifold/src/edge_op.cpp
new file mode 100644
index 000000000000..51460b911451
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/edge_op.cpp
@@ -0,0 +1,678 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "impl.h"
+#include "par.h"
+
+namespace {
+using namespace manifold;
+
+glm::ivec3 TriOf(int edge) {
+  glm::ivec3 triEdge;
+  triEdge[0] = edge;
+  triEdge[1] = NextHalfedge(triEdge[0]);
+  triEdge[2] = NextHalfedge(triEdge[1]);
+  return triEdge;
+}
+
+bool Is01Longest(glm::vec2 v0, glm::vec2 v1, glm::vec2 v2) {
+  const glm::vec2 e[3] = {v1 - v0, v2 - v1, v0 - v2};
+  float l[3];
+  for (int i : {0, 1, 2}) l[i] = glm::dot(e[i], e[i]);
+  return l[0] > l[1] && l[0] > l[2];
+}
+
+struct DuplicateEdge {
+  const Halfedge* sortedHalfedge;
+
+  bool operator()(int edge) {
+    const Halfedge& halfedge = sortedHalfedge[edge];
+    const Halfedge& nextHalfedge = sortedHalfedge[edge + 1];
+    return halfedge.startVert == nextHalfedge.startVert &&
+           halfedge.endVert == nextHalfedge.endVert;
+  }
+};
+
+struct ShortEdge {
+  VecView<const Halfedge> halfedge;
+  VecView<const glm::vec3> vertPos;
+  const float precision;
+
+  bool operator()(int edge) const {
+    if (halfedge[edge].pairedHalfedge < 0) return false;
+    // Flag short edges
+    const glm::vec3 delta =
+        vertPos[halfedge[edge].endVert] - vertPos[halfedge[edge].startVert];
+    return glm::dot(delta, delta) < precision * precision;
+  }
+};
+
+struct FlagEdge {
+  VecView<const Halfedge> halfedge;
+  VecView<const TriRef> triRef;
+
+  bool operator()(int edge) const {
+    if (halfedge[edge].pairedHalfedge < 0) return false;
+    // Flag redundant edges - those where the startVert is surrounded by only
+    // two original triangles.
+    const TriRef ref0 = triRef[edge / 3];
+    int current = NextHalfedge(halfedge[edge].pairedHalfedge);
+    const TriRef ref1 = triRef[current / 3];
+    while (current != edge) {
+      current = NextHalfedge(halfedge[current].pairedHalfedge);
+      int tri = current / 3;
+      const TriRef ref = triRef[tri];
+      if (!ref.SameFace(ref0) && !ref.SameFace(ref1)) return false;
+    }
+    return true;
+  }
+};
+
+struct SwappableEdge {
+  VecView<const Halfedge> halfedge;
+  VecView<const glm::vec3> vertPos;
+  VecView<const glm::vec3> triNormal;
+  const float precision;
+
+  bool operator()(int edge) const {
+    if (halfedge[edge].pairedHalfedge < 0) return false;
+
+    int tri = halfedge[edge].face;
+    glm::ivec3 triedge = TriOf(edge);
+    glm::mat3x2 projection = GetAxisAlignedProjection(triNormal[tri]);
+    glm::vec2 v[3];
+    for (int i : {0, 1, 2})
+      v[i] = projection * vertPos[halfedge[triedge[i]].startVert];
+    if (CCW(v[0], v[1], v[2], precision) > 0 || !Is01Longest(v[0], v[1], v[2]))
+      return false;
+
+    // Switch to neighbor's projection.
+    edge = halfedge[edge].pairedHalfedge;
+    tri = halfedge[edge].face;
+    triedge = TriOf(edge);
+    projection = GetAxisAlignedProjection(triNormal[tri]);
+    for (int i : {0, 1, 2})
+      v[i] = projection * vertPos[halfedge[triedge[i]].startVert];
+    return CCW(v[0], v[1], v[2], precision) > 0 ||
+           Is01Longest(v[0], v[1], v[2]);
+  }
+};
+
+struct SortEntry {
+  int start;
+  int end;
+  int index;
+  inline bool operator<(const SortEntry& other) const {
+    return start == other.start ? end < other.end : start < other.start;
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Collapses degenerate triangles by removing edges shorter than precision_ and
+ * any edge that is preceeded by an edge that joins the same two face relations.
+ * It also performs edge swaps on the long edges of degenerate triangles, though
+ * there are some configurations of degenerates that cannot be removed this way.
+ *
+ * Before collapsing edges, the mesh is checked for duplicate edges (more than
+ * one pair of triangles sharing the same edge), which are removed by
+ * duplicating one vert and adding two triangles. These degenerate triangles are
+ * likely to be collapsed again in the subsequent simplification.
+ *
+ * Note when an edge collapse would result in something non-manifold, the
+ * vertices are duplicated in such a way as to remove handles or separate
+ * meshes, thus decreasing the Genus(). It only increases when meshes that have
+ * collapsed to just a pair of triangles are removed entirely.
+ *
+ * Rather than actually removing the edges, this step merely marks them for
+ * removal, by setting vertPos to NaN and halfedge to {-1, -1, -1, -1}.
+ */
+void Manifold::Impl::SimplifyTopology() {
+  if (!halfedge_.size()) return;
+
+  const size_t nbEdges = halfedge_.size();
+  auto policy = autoPolicy(nbEdges);
+  size_t numFlagged = 0;
+  Vec<uint8_t> bflags(nbEdges);
+
+  // In the case of a very bad triangulation, it is possible to create pinched
+  // verts. They must be removed before edge collapse.
+  SplitPinchedVerts();
+
+  {
+    ZoneScopedN("DedupeEdge");
+    Vec<SortEntry> entries;
+    entries.reserve(nbEdges / 2);
+    for (int i = 0; i < nbEdges; ++i) {
+      if (halfedge_[i].IsForward()) {
+        entries.push_back({halfedge_[i].startVert, halfedge_[i].endVert, i});
+      }
+    }
+
+    stable_sort(policy, entries.begin(), entries.end());
+    for (int i = 0; i < entries.size() - 1; ++i) {
+      if (entries[i].start == entries[i + 1].start &&
+          entries[i].end == entries[i + 1].end) {
+        DedupeEdge(entries[i].index);
+        numFlagged++;
+      }
+    }
+  }
+
+#ifdef MANIFOLD_DEBUG
+  if (ManifoldParams().verbose && numFlagged > 0) {
+    std::cout << "found " << numFlagged << " duplicate edges to split"
+              << std::endl;
+  }
+#endif
+
+  if (!ManifoldParams().cleanupTriangles) {
+    return;
+  }
+
+  std::vector<int> scratchBuffer;
+  scratchBuffer.reserve(10);
+  {
+    ZoneScopedN("CollapseShortEdge");
+    numFlagged = 0;
+    ShortEdge se{halfedge_, vertPos_, precision_};
+    for_each_n(policy, countAt(0_z), nbEdges,
+               [&](size_t i) { bflags[i] = se(i); });
+    for (size_t i = 0; i < nbEdges; ++i) {
+      if (bflags[i]) {
+        CollapseEdge(i, scratchBuffer);
+        scratchBuffer.resize(0);
+        numFlagged++;
+      }
+    }
+  }
+
+#ifdef MANIFOLD_DEBUG
+  if (ManifoldParams().verbose && numFlagged > 0) {
+    std::cout << "found " << numFlagged << " short edges to collapse"
+              << std::endl;
+  }
+#endif
+
+  {
+    ZoneScopedN("CollapseFlaggedEdge");
+    numFlagged = 0;
+    FlagEdge se{halfedge_, meshRelation_.triRef};
+    for_each_n(policy, countAt(0_z), nbEdges,
+               [&](size_t i) { bflags[i] = se(i); });
+    for (size_t i = 0; i < nbEdges; ++i) {
+      if (bflags[i]) {
+        CollapseEdge(i, scratchBuffer);
+        scratchBuffer.resize(0);
+        numFlagged++;
+      }
+    }
+  }
+
+#ifdef MANIFOLD_DEBUG
+  if (ManifoldParams().verbose && numFlagged > 0) {
+    std::cout << "found " << numFlagged << " colinear edges to collapse"
+              << std::endl;
+  }
+#endif
+
+  {
+    ZoneScopedN("RecursiveEdgeSwap");
+    numFlagged = 0;
+    SwappableEdge se{halfedge_, vertPos_, faceNormal_, precision_};
+    for_each_n(policy, countAt(0_z), nbEdges,
+               [&](size_t i) { bflags[i] = se(i); });
+    std::vector<int> edgeSwapStack;
+    std::vector<int> visited(halfedge_.size(), -1);
+    int tag = 0;
+    for (size_t i = 0; i < nbEdges; ++i) {
+      if (bflags[i]) {
+        numFlagged++;
+        tag++;
+        RecursiveEdgeSwap(i, tag, visited, edgeSwapStack, scratchBuffer);
+        while (!edgeSwapStack.empty()) {
+          int last = edgeSwapStack.back();
+          edgeSwapStack.pop_back();
+          RecursiveEdgeSwap(last, tag, visited, edgeSwapStack, scratchBuffer);
+        }
+      }
+    }
+  }
+
+#ifdef MANIFOLD_DEBUG
+  if (ManifoldParams().verbose && numFlagged > 0) {
+    std::cout << "found " << numFlagged << " edges to swap" << std::endl;
+  }
+#endif
+}
+
+// Deduplicate the given 4-manifold edge by duplicating endVert, thus making the
+// edges distinct. Also duplicates startVert if it becomes pinched.
+void Manifold::Impl::DedupeEdge(const int edge) {
+  // Orbit endVert
+  const int startVert = halfedge_[edge].startVert;
+  const int endVert = halfedge_[edge].endVert;
+  int current = halfedge_[NextHalfedge(edge)].pairedHalfedge;
+  while (current != edge) {
+    const int vert = halfedge_[current].startVert;
+    if (vert == startVert) {
+      // Single topological unit needs 2 faces added to be split
+      const int newVert = vertPos_.size();
+      vertPos_.push_back(vertPos_[endVert]);
+      if (vertNormal_.size() > 0) vertNormal_.push_back(vertNormal_[endVert]);
+      current = halfedge_[NextHalfedge(current)].pairedHalfedge;
+      const int opposite = halfedge_[NextHalfedge(edge)].pairedHalfedge;
+
+      UpdateVert(newVert, current, opposite);
+
+      int newHalfedge = halfedge_.size();
+      int newFace = newHalfedge / 3;
+      int oldFace = halfedge_[current].face;
+      int outsideVert = halfedge_[current].startVert;
+      halfedge_.push_back({endVert, newVert, -1, newFace});
+      halfedge_.push_back({newVert, outsideVert, -1, newFace});
+      halfedge_.push_back({outsideVert, endVert, -1, newFace});
+      PairUp(newHalfedge + 2, halfedge_[current].pairedHalfedge);
+      PairUp(newHalfedge + 1, current);
+      if (meshRelation_.triRef.size() > 0)
+        meshRelation_.triRef.push_back(meshRelation_.triRef[oldFace]);
+      if (meshRelation_.triProperties.size() > 0)
+        meshRelation_.triProperties.push_back(
+            meshRelation_.triProperties[oldFace]);
+      if (faceNormal_.size() > 0) faceNormal_.push_back(faceNormal_[oldFace]);
+
+      newHalfedge += 3;
+      ++newFace;
+      oldFace = halfedge_[opposite].face;
+      outsideVert = halfedge_[opposite].startVert;
+      halfedge_.push_back({newVert, endVert, -1, newFace});
+      halfedge_.push_back({endVert, outsideVert, -1, newFace});
+      halfedge_.push_back({outsideVert, newVert, -1, newFace});
+      PairUp(newHalfedge + 2, halfedge_[opposite].pairedHalfedge);
+      PairUp(newHalfedge + 1, opposite);
+      PairUp(newHalfedge, newHalfedge - 3);
+      if (meshRelation_.triRef.size() > 0)
+        meshRelation_.triRef.push_back(meshRelation_.triRef[oldFace]);
+      if (meshRelation_.triProperties.size() > 0)
+        meshRelation_.triProperties.push_back(
+            meshRelation_.triProperties[oldFace]);
+      if (faceNormal_.size() > 0) faceNormal_.push_back(faceNormal_[oldFace]);
+
+      break;
+    }
+
+    current = halfedge_[NextHalfedge(current)].pairedHalfedge;
+  }
+
+  if (current == edge) {
+    // Separate topological unit needs no new faces to be split
+    const int newVert = vertPos_.size();
+    vertPos_.push_back(vertPos_[endVert]);
+    if (vertNormal_.size() > 0) vertNormal_.push_back(vertNormal_[endVert]);
+
+    ForVert(NextHalfedge(current), [this, newVert](int e) {
+      halfedge_[e].startVert = newVert;
+      halfedge_[halfedge_[e].pairedHalfedge].endVert = newVert;
+    });
+  }
+
+  // Orbit startVert
+  const int pair = halfedge_[edge].pairedHalfedge;
+  current = halfedge_[NextHalfedge(pair)].pairedHalfedge;
+  while (current != pair) {
+    const int vert = halfedge_[current].startVert;
+    if (vert == endVert) {
+      break;  // Connected: not a pinched vert
+    }
+    current = halfedge_[NextHalfedge(current)].pairedHalfedge;
+  }
+
+  if (current == pair) {
+    // Split the pinched vert the previous split created.
+    const int newVert = vertPos_.size();
+    vertPos_.push_back(vertPos_[endVert]);
+    if (vertNormal_.size() > 0) vertNormal_.push_back(vertNormal_[endVert]);
+
+    ForVert(NextHalfedge(current), [this, newVert](int e) {
+      halfedge_[e].startVert = newVert;
+      halfedge_[halfedge_[e].pairedHalfedge].endVert = newVert;
+    });
+  }
+}
+
+void Manifold::Impl::PairUp(int edge0, int edge1) {
+  halfedge_[edge0].pairedHalfedge = edge1;
+  halfedge_[edge1].pairedHalfedge = edge0;
+}
+
+// Traverses CW around startEdge.endVert from startEdge to endEdge
+// (edgeEdge.endVert must == startEdge.endVert), updating each edge to point
+// to vert instead.
+void Manifold::Impl::UpdateVert(int vert, int startEdge, int endEdge) {
+  int current = startEdge;
+  while (current != endEdge) {
+    halfedge_[current].endVert = vert;
+    current = NextHalfedge(current);
+    halfedge_[current].startVert = vert;
+    current = halfedge_[current].pairedHalfedge;
+    ASSERT(current != startEdge, logicErr, "infinite loop in decimator!");
+  }
+}
+
+// In the event that the edge collapse would create a non-manifold edge,
+// instead we duplicate the two verts and attach the manifolds the other way
+// across this edge.
+void Manifold::Impl::FormLoop(int current, int end) {
+  int startVert = vertPos_.size();
+  vertPos_.push_back(vertPos_[halfedge_[current].startVert]);
+  int endVert = vertPos_.size();
+  vertPos_.push_back(vertPos_[halfedge_[current].endVert]);
+
+  int oldMatch = halfedge_[current].pairedHalfedge;
+  int newMatch = halfedge_[end].pairedHalfedge;
+
+  UpdateVert(startVert, oldMatch, newMatch);
+  UpdateVert(endVert, end, current);
+
+  halfedge_[current].pairedHalfedge = newMatch;
+  halfedge_[newMatch].pairedHalfedge = current;
+  halfedge_[end].pairedHalfedge = oldMatch;
+  halfedge_[oldMatch].pairedHalfedge = end;
+
+  RemoveIfFolded(end);
+}
+
+void Manifold::Impl::CollapseTri(const glm::ivec3& triEdge) {
+  if (halfedge_[triEdge[1]].pairedHalfedge == -1) return;
+  int pair1 = halfedge_[triEdge[1]].pairedHalfedge;
+  int pair2 = halfedge_[triEdge[2]].pairedHalfedge;
+  halfedge_[pair1].pairedHalfedge = pair2;
+  halfedge_[pair2].pairedHalfedge = pair1;
+  for (int i : {0, 1, 2}) {
+    halfedge_[triEdge[i]] = {-1, -1, -1, -1};
+  }
+}
+
+void Manifold::Impl::RemoveIfFolded(int edge) {
+  const glm::ivec3 tri0edge = TriOf(edge);
+  const glm::ivec3 tri1edge = TriOf(halfedge_[edge].pairedHalfedge);
+  if (halfedge_[tri0edge[1]].pairedHalfedge == -1) return;
+  if (halfedge_[tri0edge[1]].endVert == halfedge_[tri1edge[1]].endVert) {
+    if (halfedge_[tri0edge[1]].pairedHalfedge == tri1edge[2]) {
+      if (halfedge_[tri0edge[2]].pairedHalfedge == tri1edge[1]) {
+        for (int i : {0, 1, 2})
+          vertPos_[halfedge_[tri0edge[i]].startVert] = glm::vec3(NAN);
+      } else {
+        vertPos_[halfedge_[tri0edge[1]].startVert] = glm::vec3(NAN);
+      }
+    } else {
+      if (halfedge_[tri0edge[2]].pairedHalfedge == tri1edge[1]) {
+        vertPos_[halfedge_[tri1edge[1]].startVert] = glm::vec3(NAN);
+      }
+    }
+    PairUp(halfedge_[tri0edge[1]].pairedHalfedge,
+           halfedge_[tri1edge[2]].pairedHalfedge);
+    PairUp(halfedge_[tri0edge[2]].pairedHalfedge,
+           halfedge_[tri1edge[1]].pairedHalfedge);
+    for (int i : {0, 1, 2}) {
+      halfedge_[tri0edge[i]] = {-1, -1, -1, -1};
+      halfedge_[tri1edge[i]] = {-1, -1, -1, -1};
+    }
+  }
+}
+
+// Collapses the given edge by removing startVert. May split the mesh
+// topologically if the collapse would have resulted in a 4-manifold edge. Do
+// not collapse an edge if startVert is pinched - the vert will be marked NaN,
+// but other edges may still be pointing to it.
+void Manifold::Impl::CollapseEdge(const int edge, std::vector<int>& edges) {
+  Vec<TriRef>& triRef = meshRelation_.triRef;
+  Vec<glm::ivec3>& triProp = meshRelation_.triProperties;
+
+  const Halfedge toRemove = halfedge_[edge];
+  if (toRemove.pairedHalfedge < 0) return;
+
+  const int endVert = toRemove.endVert;
+  const glm::ivec3 tri0edge = TriOf(edge);
+  const glm::ivec3 tri1edge = TriOf(toRemove.pairedHalfedge);
+
+  const glm::vec3 pNew = vertPos_[endVert];
+  const glm::vec3 pOld = vertPos_[toRemove.startVert];
+  const glm::vec3 delta = pNew - pOld;
+  const bool shortEdge = glm::dot(delta, delta) < precision_ * precision_;
+
+  // Orbit endVert
+  int current = halfedge_[tri0edge[1]].pairedHalfedge;
+  while (current != tri1edge[2]) {
+    current = NextHalfedge(current);
+    edges.push_back(current);
+    current = halfedge_[current].pairedHalfedge;
+  }
+
+  // Orbit startVert
+  int start = halfedge_[tri1edge[1]].pairedHalfedge;
+  if (!shortEdge) {
+    current = start;
+    TriRef refCheck = triRef[toRemove.pairedHalfedge / 3];
+    glm::vec3 pLast = vertPos_[halfedge_[tri1edge[1]].endVert];
+    while (current != tri0edge[2]) {
+      current = NextHalfedge(current);
+      glm::vec3 pNext = vertPos_[halfedge_[current].endVert];
+      const int tri = current / 3;
+      const TriRef ref = triRef[tri];
+      const glm::mat3x2 projection = GetAxisAlignedProjection(faceNormal_[tri]);
+      // Don't collapse if the edge is not redundant (this may have changed due
+      // to the collapse of neighbors).
+      if (!ref.SameFace(refCheck)) {
+        refCheck = triRef[edge / 3];
+        if (!ref.SameFace(refCheck)) {
+          return;
+        } else {
+          // Don't collapse if the edges separating the faces are not colinear
+          // (can happen when the two faces are coplanar).
+          if (CCW(projection * pOld, projection * pLast, projection * pNew,
+                  precision_) != 0)
+            return;
+        }
+      }
+
+      // Don't collapse edge if it would cause a triangle to invert.
+      if (CCW(projection * pNext, projection * pLast, projection * pNew,
+              precision_) < 0)
+        return;
+
+      pLast = pNext;
+      current = halfedge_[current].pairedHalfedge;
+    }
+  }
+
+  // Remove toRemove.startVert and replace with endVert.
+  vertPos_[toRemove.startVert] = glm::vec3(NAN);
+  CollapseTri(tri1edge);
+
+  // Orbit startVert
+  const int tri0 = edge / 3;
+  const int tri1 = toRemove.pairedHalfedge / 3;
+  const int triVert0 = (edge + 1) % 3;
+  const int triVert1 = toRemove.pairedHalfedge % 3;
+  current = start;
+  while (current != tri0edge[2]) {
+    current = NextHalfedge(current);
+
+    if (triProp.size() > 0) {
+      // Update the shifted triangles to the vertBary of endVert
+      const int tri = current / 3;
+      const int vIdx = current - 3 * tri;
+      if (triRef[tri].SameFace(triRef[tri0])) {
+        triProp[tri][vIdx] = triProp[tri0][triVert0];
+      } else if (triRef[tri].SameFace(triRef[tri1])) {
+        triProp[tri][vIdx] = triProp[tri1][triVert1];
+      }
+    }
+
+    const int vert = halfedge_[current].endVert;
+    const int next = halfedge_[current].pairedHalfedge;
+    for (int i = 0; i < edges.size(); ++i) {
+      if (vert == halfedge_[edges[i]].endVert) {
+        FormLoop(edges[i], current);
+        start = next;
+        edges.resize(i);
+        break;
+      }
+    }
+    current = next;
+  }
+
+  UpdateVert(endVert, start, tri0edge[2]);
+  CollapseTri(tri0edge);
+  RemoveIfFolded(start);
+}
+
+void Manifold::Impl::RecursiveEdgeSwap(const int edge, int& tag,
+                                       std::vector<int>& visited,
+                                       std::vector<int>& edgeSwapStack,
+                                       std::vector<int>& edges) {
+  Vec<TriRef>& triRef = meshRelation_.triRef;
+
+  if (edge < 0) return;
+  const int pair = halfedge_[edge].pairedHalfedge;
+  if (pair < 0) return;
+
+  // avoid infinite recursion
+  if (visited[edge] == tag && visited[pair] == tag) return;
+
+  const glm::ivec3 tri0edge = TriOf(edge);
+  const glm::ivec3 tri1edge = TriOf(pair);
+  const glm::ivec3 perm0 = TriOf(edge % 3);
+  const glm::ivec3 perm1 = TriOf(pair % 3);
+
+  glm::mat3x2 projection = GetAxisAlignedProjection(faceNormal_[edge / 3]);
+  glm::vec2 v[4];
+  for (int i : {0, 1, 2})
+    v[i] = projection * vertPos_[halfedge_[tri0edge[i]].startVert];
+  // Only operate on the long edge of a degenerate triangle.
+  if (CCW(v[0], v[1], v[2], precision_) > 0 || !Is01Longest(v[0], v[1], v[2]))
+    return;
+
+  // Switch to neighbor's projection.
+  projection = GetAxisAlignedProjection(faceNormal_[halfedge_[pair].face]);
+  for (int i : {0, 1, 2})
+    v[i] = projection * vertPos_[halfedge_[tri0edge[i]].startVert];
+  v[3] = projection * vertPos_[halfedge_[tri1edge[2]].startVert];
+
+  auto SwapEdge = [&]() {
+    // The 0-verts are swapped to the opposite 2-verts.
+    const int v0 = halfedge_[tri0edge[2]].startVert;
+    const int v1 = halfedge_[tri1edge[2]].startVert;
+    halfedge_[tri0edge[0]].startVert = v1;
+    halfedge_[tri0edge[2]].endVert = v1;
+    halfedge_[tri1edge[0]].startVert = v0;
+    halfedge_[tri1edge[2]].endVert = v0;
+    PairUp(tri0edge[0], halfedge_[tri1edge[2]].pairedHalfedge);
+    PairUp(tri1edge[0], halfedge_[tri0edge[2]].pairedHalfedge);
+    PairUp(tri0edge[2], tri1edge[2]);
+    // Both triangles are now subsets of the neighboring triangle.
+    const int tri0 = halfedge_[tri0edge[0]].face;
+    const int tri1 = halfedge_[tri1edge[0]].face;
+    faceNormal_[tri0] = faceNormal_[tri1];
+    triRef[tri0] = triRef[tri1];
+    const float l01 = glm::length(v[1] - v[0]);
+    const float l02 = glm::length(v[2] - v[0]);
+    const float a = glm::max(0.0f, glm::min(1.0f, l02 / l01));
+    // Update properties if applicable
+    if (meshRelation_.properties.size() > 0) {
+      Vec<glm::ivec3>& triProp = meshRelation_.triProperties;
+      Vec<float>& prop = meshRelation_.properties;
+      triProp[tri0] = triProp[tri1];
+      triProp[tri0][perm0[1]] = triProp[tri1][perm1[0]];
+      triProp[tri0][perm0[0]] = triProp[tri1][perm1[2]];
+      const int numProp = NumProp();
+      const int newProp = prop.size() / numProp;
+      const int propIdx0 = triProp[tri1][perm1[0]];
+      const int propIdx1 = triProp[tri1][perm1[1]];
+      for (int p = 0; p < numProp; ++p) {
+        prop.push_back(a * prop[numProp * propIdx0 + p] +
+                       (1 - a) * prop[numProp * propIdx1 + p]);
+      }
+      triProp[tri1][perm1[0]] = newProp;
+      triProp[tri0][perm0[2]] = newProp;
+    }
+
+    // if the new edge already exists, duplicate the verts and split the mesh.
+    int current = halfedge_[tri1edge[0]].pairedHalfedge;
+    const int endVert = halfedge_[tri1edge[1]].endVert;
+    while (current != tri0edge[1]) {
+      current = NextHalfedge(current);
+      if (halfedge_[current].endVert == endVert) {
+        FormLoop(tri0edge[2], current);
+        RemoveIfFolded(tri0edge[2]);
+        return;
+      }
+      current = halfedge_[current].pairedHalfedge;
+    }
+  };
+
+  // Only operate if the other triangles are not degenerate.
+  if (CCW(v[1], v[0], v[3], precision_) <= 0) {
+    if (!Is01Longest(v[1], v[0], v[3])) return;
+    // Two facing, long-edge degenerates can swap.
+    SwapEdge();
+    const glm::vec2 e23 = v[3] - v[2];
+    if (glm::dot(e23, e23) < precision_ * precision_) {
+      tag++;
+      CollapseEdge(tri0edge[2], edges);
+      edges.resize(0);
+    } else {
+      visited[edge] = tag;
+      visited[pair] = tag;
+      edgeSwapStack.insert(edgeSwapStack.end(), {tri1edge[1], tri1edge[0],
+                                                 tri0edge[1], tri0edge[0]});
+    }
+    return;
+  } else if (CCW(v[0], v[3], v[2], precision_) <= 0 ||
+             CCW(v[1], v[2], v[3], precision_) <= 0) {
+    return;
+  }
+  // Normal path
+  SwapEdge();
+  visited[edge] = tag;
+  visited[pair] = tag;
+  edgeSwapStack.insert(edgeSwapStack.end(),
+                       {halfedge_[tri1edge[0]].pairedHalfedge,
+                        halfedge_[tri0edge[1]].pairedHalfedge});
+}
+
+void Manifold::Impl::SplitPinchedVerts() {
+  ZoneScoped;
+  std::vector<bool> vertProcessed(NumVert(), false);
+  std::vector<bool> halfedgeProcessed(halfedge_.size(), false);
+  for (int i = 0; i < halfedge_.size(); ++i) {
+    if (halfedgeProcessed[i]) continue;
+    int vert = halfedge_[i].startVert;
+    if (vertProcessed[vert]) {
+      vertPos_.push_back(vertPos_[vert]);
+      vert = NumVert() - 1;
+    } else {
+      vertProcessed[vert] = true;
+    }
+    ForVert(i, [this, &halfedgeProcessed, vert](int current) {
+      halfedgeProcessed[current] = true;
+      halfedge_[current].startVert = vert;
+      halfedge_[halfedge_[current].pairedHalfedge].endVert = vert;
+    });
+  }
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/face_op.cpp b/thirdparty/manifold/src/manifold/src/face_op.cpp
new file mode 100644
index 000000000000..29bc90488bf1
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/face_op.cpp
@@ -0,0 +1,321 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#if MANIFOLD_PAR == 'T' && __has_include(<tbb/concurrent_map.h>)
+#include <tbb/tbb.h>
+#define TBB_PREVIEW_CONCURRENT_ORDERED_CONTAINERS 1
+#include <tbb/concurrent_map.h>
+#endif
+#include <unordered_set>
+
+#include "impl.h"
+#include "polygon.h"
+
+namespace manifold {
+
+using GeneralTriangulation = std::function<std::vector<glm::ivec3>(int)>;
+using AddTriangle = std::function<void(int, glm::ivec3, glm::vec3, TriRef)>;
+
+/**
+ * Triangulates the faces. In this case, the halfedge_ vector is not yet a set
+ * of triangles as required by this data structure, but is instead a set of
+ * general faces with the input faceEdge vector having length of the number of
+ * faces + 1. The values are indicies into the halfedge_ vector for the first
+ * edge of each face, with the final value being the length of the halfedge_
+ * vector itself. Upon return, halfedge_ has been lengthened and properly
+ * represents the mesh as a set of triangles as usual. In this process the
+ * faceNormal_ values are retained, repeated as necessary.
+ */
+void Manifold::Impl::Face2Tri(const Vec<int>& faceEdge,
+                              const Vec<TriRef>& halfedgeRef) {
+  ZoneScoped;
+  Vec<glm::ivec3> triVerts;
+  Vec<glm::vec3> triNormal;
+  Vec<TriRef>& triRef = meshRelation_.triRef;
+  triRef.resize(0);
+  auto processFace = [&](GeneralTriangulation general, AddTriangle addTri,
+                         int face) {
+    const int firstEdge = faceEdge[face];
+    const int lastEdge = faceEdge[face + 1];
+    const int numEdge = lastEdge - firstEdge;
+    ASSERT(numEdge >= 3, topologyErr, "face has less than three edges.");
+    const glm::vec3 normal = faceNormal_[face];
+
+    if (numEdge == 3) {  // Single triangle
+      int mapping[3] = {halfedge_[firstEdge].startVert,
+                        halfedge_[firstEdge + 1].startVert,
+                        halfedge_[firstEdge + 2].startVert};
+      glm::ivec3 tri(halfedge_[firstEdge].startVert,
+                     halfedge_[firstEdge + 1].startVert,
+                     halfedge_[firstEdge + 2].startVert);
+      glm::ivec3 ends(halfedge_[firstEdge].endVert,
+                      halfedge_[firstEdge + 1].endVert,
+                      halfedge_[firstEdge + 2].endVert);
+      if (ends[0] == tri[2]) {
+        std::swap(tri[1], tri[2]);
+        std::swap(ends[1], ends[2]);
+      }
+      ASSERT(ends[0] == tri[1] && ends[1] == tri[2] && ends[2] == tri[0],
+             topologyErr, "These 3 edges do not form a triangle!");
+
+      addTri(face, tri, normal, halfedgeRef[firstEdge]);
+    } else if (numEdge == 4) {  // Pair of triangles
+      int mapping[4] = {halfedge_[firstEdge].startVert,
+                        halfedge_[firstEdge + 1].startVert,
+                        halfedge_[firstEdge + 2].startVert,
+                        halfedge_[firstEdge + 3].startVert};
+      const glm::mat3x2 projection = GetAxisAlignedProjection(normal);
+      auto triCCW = [&projection, this](const glm::ivec3 tri) {
+        return CCW(projection * this->vertPos_[tri[0]],
+                   projection * this->vertPos_[tri[1]],
+                   projection * this->vertPos_[tri[2]], precision_) >= 0;
+      };
+
+      glm::ivec3 tri0(halfedge_[firstEdge].startVert,
+                      halfedge_[firstEdge].endVert, -1);
+      glm::ivec3 tri1(-1, -1, tri0[0]);
+      for (const int i : {1, 2, 3}) {
+        if (halfedge_[firstEdge + i].startVert == tri0[1]) {
+          tri0[2] = halfedge_[firstEdge + i].endVert;
+          tri1[0] = tri0[2];
+        }
+        if (halfedge_[firstEdge + i].endVert == tri0[0]) {
+          tri1[1] = halfedge_[firstEdge + i].startVert;
+        }
+      }
+      ASSERT(glm::all(glm::greaterThanEqual(tri0, glm::ivec3(0))) &&
+                 glm::all(glm::greaterThanEqual(tri1, glm::ivec3(0))),
+             topologyErr, "non-manifold quad!");
+      bool firstValid = triCCW(tri0) && triCCW(tri1);
+      tri0[2] = tri1[1];
+      tri1[2] = tri0[1];
+      bool secondValid = triCCW(tri0) && triCCW(tri1);
+
+      if (!secondValid) {
+        tri0[2] = tri1[0];
+        tri1[2] = tri0[0];
+      } else if (firstValid) {
+        glm::vec3 firstCross = vertPos_[tri0[0]] - vertPos_[tri1[0]];
+        glm::vec3 secondCross = vertPos_[tri0[1]] - vertPos_[tri1[1]];
+        if (glm::dot(firstCross, firstCross) <
+            glm::dot(secondCross, secondCross)) {
+          tri0[2] = tri1[0];
+          tri1[2] = tri0[0];
+        }
+      }
+
+      for (const auto& tri : {tri0, tri1}) {
+        addTri(face, tri, normal, halfedgeRef[firstEdge]);
+      }
+    } else {  // General triangulation
+      for (const auto& tri : general(face)) {
+        addTri(face, tri, normal, halfedgeRef[firstEdge]);
+      }
+    }
+  };
+  auto generalTriangulation = [&](int face) {
+    const glm::vec3 normal = faceNormal_[face];
+    const glm::mat3x2 projection = GetAxisAlignedProjection(normal);
+    const PolygonsIdx polys =
+        Face2Polygons(halfedge_.cbegin() + faceEdge[face],
+                      halfedge_.cbegin() + faceEdge[face + 1], projection);
+    return TriangulateIdx(polys, precision_);
+  };
+#if MANIFOLD_PAR == 'T' && __has_include(<tbb/tbb.h>)
+  tbb::task_group group;
+  // map from face to triangle
+  tbb::concurrent_unordered_map<int, std::vector<glm::ivec3>> results;
+  Vec<size_t> triCount(faceEdge.size());
+  triCount.back() = 0;
+  // precompute number of triangles per face, and launch async tasks to
+  // triangulate complex faces
+  for_each(autoPolicy(faceEdge.size()), countAt(0_z),
+           countAt(faceEdge.size() - 1), [&](size_t face) {
+             triCount[face] = faceEdge[face + 1] - faceEdge[face] - 2;
+             ASSERT(triCount[face] >= 1, topologyErr,
+                    "face has less than three edges.");
+             if (triCount[face] > 2)
+               group.run([&, face] {
+                 std::vector<glm::ivec3> newTris = generalTriangulation(face);
+                 triCount[face] = newTris.size();
+                 results[face] = std::move(newTris);
+               });
+           });
+  group.wait();
+  // prefix sum computation (assign unique index to each face) and preallocation
+  exclusive_scan(autoPolicy(triCount.size()), triCount.begin(), triCount.end(),
+                 triCount.begin(), 0_z);
+  if (triCount.back() >= std::numeric_limits<int>::max())
+    throw std::out_of_range("too many triangles");
+  triVerts.resize(triCount.back());
+  triNormal.resize(triCount.back());
+  triRef.resize(triCount.back());
+
+  auto processFace2 = std::bind(
+      processFace, [&](size_t face) { return std::move(results[face]); },
+      [&](int face, glm::ivec3 tri, glm::vec3 normal, TriRef r) {
+        triVerts[triCount[face]] = tri;
+        triNormal[triCount[face]] = normal;
+        triRef[triCount[face]] = r;
+        triCount[face]++;
+      },
+      std::placeholders::_1);
+  // set triangles in parallel
+  for_each(autoPolicy(faceEdge.size()), countAt(0_z),
+           countAt(faceEdge.size() - 1), processFace2);
+#else
+  triVerts.reserve(faceEdge.size());
+  triNormal.reserve(faceEdge.size());
+  triRef.reserve(faceEdge.size());
+  auto processFace2 = std::bind(
+      processFace, generalTriangulation,
+      [&](int _face, glm::ivec3 tri, glm::vec3 normal, TriRef r) {
+        triVerts.push_back(tri);
+        triNormal.push_back(normal);
+        triRef.push_back(r);
+      },
+      std::placeholders::_1);
+  for (int face = 0; face < faceEdge.size() - 1; ++face) {
+    processFace2(face);
+  }
+#endif
+
+  faceNormal_ = std::move(triNormal);
+  CreateHalfedges(triVerts);
+}
+
+/**
+ * Returns a set of 2D polygons formed by the input projection of the vertices
+ * of the list of Halfedges, which must be an even-manifold, meaning each vert
+ * must be referenced the same number of times as a startVert and endVert.
+ */
+PolygonsIdx Manifold::Impl::Face2Polygons(VecView<Halfedge>::IterC start,
+                                          VecView<Halfedge>::IterC end,
+                                          glm::mat3x2 projection) const {
+  std::multimap<int, int> vert_edge;
+  for (auto edge = start; edge != end; ++edge) {
+    vert_edge.emplace(std::make_pair(edge->startVert, edge - start));
+  }
+
+  PolygonsIdx polys;
+  int startEdge = 0;
+  int thisEdge = startEdge;
+  while (1) {
+    if (thisEdge == startEdge) {
+      if (vert_edge.empty()) break;
+      startEdge = vert_edge.begin()->second;
+      thisEdge = startEdge;
+      polys.push_back({});
+    }
+    int vert = (start + thisEdge)->startVert;
+    polys.back().push_back({projection * vertPos_[vert], vert});
+    const auto result = vert_edge.find((start + thisEdge)->endVert);
+    ASSERT(result != vert_edge.end(), topologyErr, "non-manifold edge");
+    thisEdge = result->second;
+    vert_edge.erase(result);
+  }
+  return polys;
+}
+
+CrossSection Manifold::Impl::Slice(float height) const {
+  Box plane = bBox_;
+  plane.min.z = plane.max.z = height;
+  Vec<Box> query;
+  query.push_back(plane);
+  const SparseIndices collisions =
+      collider_.Collisions<false, false>(query.cview());
+
+  std::unordered_set<int> tris;
+  for (int i = 0; i < collisions.size(); ++i) {
+    const int tri = collisions.Get(i, 1);
+    float min = std::numeric_limits<float>::infinity();
+    float max = -std::numeric_limits<float>::infinity();
+    for (const int j : {0, 1, 2}) {
+      const float z = vertPos_[halfedge_[3 * tri + j].startVert].z;
+      min = glm::min(min, z);
+      max = glm::max(max, z);
+    }
+
+    if (min <= height && max > height) {
+      tris.insert(tri);
+    }
+  }
+
+  Polygons polys;
+  while (!tris.empty()) {
+    const int startTri = *tris.begin();
+    SimplePolygon poly;
+
+    int k = 0;
+    for (const int j : {0, 1, 2}) {
+      if (vertPos_[halfedge_[3 * startTri + j].startVert].z > height &&
+          vertPos_[halfedge_[3 * startTri + Next3(j)].startVert].z <= height) {
+        k = Next3(j);
+        break;
+      }
+    }
+
+    int tri = startTri;
+    do {
+      tris.erase(tris.find(tri));
+      if (vertPos_[halfedge_[3 * tri + k].endVert].z <= height) {
+        k = Next3(k);
+      }
+
+      Halfedge up = halfedge_[3 * tri + k];
+      const glm::vec3 below = vertPos_[up.startVert];
+      const glm::vec3 above = vertPos_[up.endVert];
+      const float a = (height - below.z) / (above.z - below.z);
+      poly.push_back(glm::vec2(glm::mix(below, above, a)));
+
+      const int pair = up.pairedHalfedge;
+      tri = pair / 3;
+      k = Next3(pair % 3);
+    } while (tri != startTri);
+
+    polys.push_back(poly);
+  }
+
+  return CrossSection(polys);
+}
+
+CrossSection Manifold::Impl::Project() const {
+  const glm::mat3x2 projection = GetAxisAlignedProjection({0, 0, 1});
+  auto policy = autoPolicy(halfedge_.size());
+
+  Vec<Halfedge> cusps(NumEdge());
+  cusps.resize(copy_if<decltype(cusps.begin())>(
+                   policy, halfedge_.cbegin(), halfedge_.cend(), cusps.begin(),
+                   [&](Halfedge edge) {
+                     return faceNormal_[edge.face].z >= 0 &&
+                            faceNormal_[halfedge_[edge.pairedHalfedge].face].z <
+                                0;
+                   }) -
+               cusps.begin());
+
+  PolygonsIdx polysIndexed =
+      Face2Polygons(cusps.cbegin(), cusps.cend(), projection);
+
+  Polygons polys;
+  for (const auto& poly : polysIndexed) {
+    SimplePolygon simple;
+    for (const PolyVert& polyVert : poly) {
+      simple.push_back(polyVert.pos);
+    }
+    polys.push_back(simple);
+  }
+
+  return CrossSection(polys).Simplify(precision_);
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/impl.cpp b/thirdparty/manifold/src/manifold/src/impl.cpp
new file mode 100644
index 000000000000..ed8373ce1158
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/impl.cpp
@@ -0,0 +1,956 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "impl.h"
+
+#include <algorithm>
+#include <atomic>
+#include <map>
+#include <numeric>
+
+#include "hashtable.h"
+#include "mesh_fixes.h"
+#include "par.h"
+#include "svd.h"
+#include "tri_dist.h"
+
+namespace {
+using namespace manifold;
+
+constexpr uint64_t kRemove = std::numeric_limits<uint64_t>::max();
+
+void AtomicAddVec3(glm::vec3& target, const glm::vec3& add) {
+  for (int i : {0, 1, 2}) {
+    std::atomic<float>& tar = reinterpret_cast<std::atomic<float>&>(target[i]);
+    float old_val = tar.load(std::memory_order_relaxed);
+    while (!tar.compare_exchange_weak(old_val, old_val + add[i],
+                                      std::memory_order_relaxed))
+      ;
+  }
+}
+
+struct Normalize {
+  void operator()(glm::vec3& v) { v = SafeNormalize(v); }
+};
+
+struct Transform4x3 {
+  const glm::mat4x3 transform;
+
+  glm::vec3 operator()(glm::vec3 position) {
+    return transform * glm::vec4(position, 1.0f);
+  }
+};
+
+struct AssignNormals {
+  VecView<glm::vec3> vertNormal;
+  VecView<const glm::vec3> vertPos;
+  VecView<const Halfedge> halfedges;
+  const float precision;
+  const bool calculateTriNormal;
+
+  void operator()(thrust::tuple<glm::vec3&, int> in) {
+    glm::vec3& triNormal = thrust::get<0>(in);
+    const int face = thrust::get<1>(in);
+
+    glm::ivec3 triVerts;
+    for (int i : {0, 1, 2}) triVerts[i] = halfedges[3 * face + i].startVert;
+
+    glm::vec3 edge[3];
+    for (int i : {0, 1, 2}) {
+      const int j = (i + 1) % 3;
+      edge[i] = glm::normalize(vertPos[triVerts[j]] - vertPos[triVerts[i]]);
+    }
+
+    if (calculateTriNormal) {
+      triNormal = glm::normalize(glm::cross(edge[0], edge[1]));
+      if (isnan(triNormal.x)) triNormal = glm::vec3(0, 0, 1);
+    }
+
+    // corner angles
+    glm::vec3 phi;
+    float dot = -glm::dot(edge[2], edge[0]);
+    phi[0] = dot >= 1 ? 0 : (dot <= -1 ? glm::pi<float>() : glm::acos(dot));
+    dot = -glm::dot(edge[0], edge[1]);
+    phi[1] = dot >= 1 ? 0 : (dot <= -1 ? glm::pi<float>() : glm::acos(dot));
+    phi[2] = glm::pi<float>() - phi[0] - phi[1];
+
+    // assign weighted sum
+    for (int i : {0, 1, 2}) {
+      AtomicAddVec3(vertNormal[triVerts[i]], phi[i] * triNormal);
+    }
+  }
+};
+
+struct Tri2Halfedges {
+  VecView<Halfedge> halfedges;
+  VecView<glm::uint64_t> edges;
+
+  void operator()(thrust::tuple<int, const glm::ivec3&> in) {
+    const int tri = thrust::get<0>(in);
+    const glm::ivec3& triVerts = thrust::get<1>(in);
+    for (const int i : {0, 1, 2}) {
+      const int j = (i + 1) % 3;
+      const int edge = 3 * tri + i;
+      halfedges[edge] = {triVerts[i], triVerts[j], -1, tri};
+      // Sort the forward halfedges in front of the backward ones by setting the
+      // highest-order bit.
+      edges[edge] = glm::uint64_t(triVerts[i] < triVerts[j] ? 1 : 0) << 63 |
+                    ((glm::uint64_t)glm::min(triVerts[i], triVerts[j])) << 32 |
+                    glm::max(triVerts[i], triVerts[j]);
+    }
+  }
+};
+
+struct LinkHalfedges {
+  VecView<Halfedge> halfedges;
+  VecView<const int> ids;
+  const int numEdge;
+
+  void operator()(int i) {
+    const int pair0 = ids[i];
+    const int pair1 = ids[i + numEdge];
+    halfedges[pair0].pairedHalfedge = pair1;
+    halfedges[pair1].pairedHalfedge = pair0;
+  }
+};
+
+struct MarkVerts {
+  VecView<int> vert;
+
+  void operator()(glm::ivec3 triVerts) {
+    for (int i : {0, 1, 2}) {
+      reinterpret_cast<std::atomic<int>*>(&vert[triVerts[i]])
+          ->store(1, std::memory_order_relaxed);
+    }
+  }
+};
+
+struct ReindexTriVerts {
+  VecView<const int> old2new;
+
+  void operator()(glm::ivec3& triVerts) {
+    for (int i : {0, 1, 2}) {
+      triVerts[i] = old2new[triVerts[i]];
+    }
+  }
+};
+
+struct InitializeTriRef {
+  const int meshID;
+  VecView<const Halfedge> halfedge;
+
+  void operator()(thrust::tuple<TriRef&, int> inOut) {
+    TriRef& baryRef = thrust::get<0>(inOut);
+    int tri = thrust::get<1>(inOut);
+
+    baryRef.meshID = meshID;
+    baryRef.originalID = meshID;
+    baryRef.tri = tri;
+  }
+};
+
+struct UpdateMeshID {
+  const HashTableD<uint32_t> meshIDold2new;
+
+  void operator()(TriRef& ref) { ref.meshID = meshIDold2new[ref.meshID]; }
+};
+
+struct CoplanarEdge {
+  VecView<float> triArea;
+  VecView<const Halfedge> halfedge;
+  VecView<const glm::vec3> vertPos;
+  VecView<const TriRef> triRef;
+  VecView<const glm::ivec3> triProp;
+  VecView<const float> prop;
+  VecView<const float> propTol;
+  const int numProp;
+  const float precision;
+
+  // FIXME: race condition
+  void operator()(
+      thrust::tuple<thrust::pair<int, int>&, thrust::pair<int, int>&, int>
+          inOut) {
+    thrust::pair<int, int>& face2face = thrust::get<0>(inOut);
+    thrust::pair<int, int>& vert2vert = thrust::get<1>(inOut);
+    const int edgeIdx = thrust::get<2>(inOut);
+
+    const Halfedge edge = halfedge[edgeIdx];
+    const Halfedge pair = halfedge[edge.pairedHalfedge];
+
+    if (triRef[edge.face].meshID != triRef[pair.face].meshID) return;
+
+    const glm::vec3 base = vertPos[edge.startVert];
+    const int baseNum = edgeIdx - 3 * edge.face;
+    const int jointNum = edge.pairedHalfedge - 3 * pair.face;
+
+    if (numProp > 0) {
+      const int prop0 = triProp[edge.face][baseNum];
+      const int prop1 = triProp[pair.face][jointNum == 2 ? 0 : jointNum + 1];
+      bool propEqual = true;
+      for (int p = 0; p < numProp; ++p) {
+        if (glm::abs(prop[numProp * prop0 + p] - prop[numProp * prop1 + p]) >
+            propTol[p]) {
+          propEqual = false;
+          break;
+        }
+      }
+      if (propEqual) {
+        vert2vert.first = prop0;
+        vert2vert.second = prop1;
+      }
+    }
+
+    if (!edge.IsForward()) return;
+
+    const int edgeNum = baseNum == 0 ? 2 : baseNum - 1;
+    const int pairNum = jointNum == 0 ? 2 : jointNum - 1;
+    const glm::vec3 jointVec = vertPos[pair.startVert] - base;
+    const glm::vec3 edgeVec =
+        vertPos[halfedge[3 * edge.face + edgeNum].startVert] - base;
+    const glm::vec3 pairVec =
+        vertPos[halfedge[3 * pair.face + pairNum].startVert] - base;
+
+    const float length = glm::max(glm::length(jointVec), glm::length(edgeVec));
+    const float lengthPair =
+        glm::max(glm::length(jointVec), glm::length(pairVec));
+    glm::vec3 normal = glm::cross(jointVec, edgeVec);
+    const float area = glm::length(normal);
+    const float areaPair = glm::length(glm::cross(pairVec, jointVec));
+    triArea[edge.face] = area;
+    triArea[pair.face] = areaPair;
+    // Don't link degenerate triangles
+    if (area < length * precision || areaPair < lengthPair * precision) return;
+
+    const float volume = glm::abs(glm::dot(normal, pairVec));
+    // Only operate on coplanar triangles
+    if (volume > glm::max(area, areaPair) * precision) return;
+
+    // Check property linearity
+    if (area > 0) {
+      normal /= area;
+      for (int i = 0; i < numProp; ++i) {
+        const float scale = precision / propTol[i];
+
+        const float baseProp = prop[numProp * triProp[edge.face][baseNum] + i];
+        const float jointProp =
+            prop[numProp * triProp[pair.face][jointNum] + i];
+        const float edgeProp = prop[numProp * triProp[edge.face][edgeNum] + i];
+        const float pairProp = prop[numProp * triProp[pair.face][pairNum] + i];
+
+        const glm::vec3 iJointVec =
+            jointVec + normal * scale * (jointProp - baseProp);
+        const glm::vec3 iEdgeVec =
+            edgeVec + normal * scale * (edgeProp - baseProp);
+        const glm::vec3 iPairVec =
+            pairVec + normal * scale * (pairProp - baseProp);
+
+        glm::vec3 cross = glm::cross(iJointVec, iEdgeVec);
+        const float areaP = glm::max(
+            glm::length(cross), glm::length(glm::cross(iPairVec, iJointVec)));
+        const float volumeP = glm::abs(glm::dot(cross, iPairVec));
+        // Only operate on consistent triangles
+        if (volumeP > areaP * precision) return;
+      }
+    }
+
+    face2face.first = edge.face;
+    face2face.second = pair.face;
+  }
+};
+
+struct CheckCoplanarity {
+  VecView<int> comp2tri;
+  VecView<const Halfedge> halfedge;
+  VecView<const glm::vec3> vertPos;
+  std::vector<int>* components;
+  const float precision;
+
+  void operator()(int tri) {
+    const int component = (*components)[tri];
+    const int referenceTri = comp2tri[component];
+    if (referenceTri < 0 || referenceTri == tri) return;
+
+    const glm::vec3 origin = vertPos[halfedge[3 * referenceTri].startVert];
+    const glm::vec3 normal = glm::normalize(
+        glm::cross(vertPos[halfedge[3 * referenceTri + 1].startVert] - origin,
+                   vertPos[halfedge[3 * referenceTri + 2].startVert] - origin));
+
+    for (const int i : {0, 1, 2}) {
+      const glm::vec3 vert = vertPos[halfedge[3 * tri + i].startVert];
+      // If any component vertex is not coplanar with the component's reference
+      // triangle, unmark the entire component so that none of its triangles are
+      // marked coplanar.
+      if (glm::abs(glm::dot(normal, vert - origin)) > precision) {
+        reinterpret_cast<std::atomic<int>*>(&comp2tri[component])
+            ->store(-1, std::memory_order_relaxed);
+        break;
+      }
+    }
+  }
+};
+
+struct EdgeBox {
+  VecView<const glm::vec3> vertPos;
+
+  void operator()(thrust::tuple<Box&, const TmpEdge&> inout) {
+    const TmpEdge& edge = thrust::get<1>(inout);
+    thrust::get<0>(inout) = Box(vertPos[edge.first], vertPos[edge.second]);
+  }
+};
+
+int GetLabels(std::vector<int>& components,
+              const Vec<thrust::pair<int, int>>& edges, int numNodes) {
+  UnionFind<> uf(numNodes);
+  for (auto edge : edges) {
+    if (edge.first == -1 || edge.second == -1) continue;
+    if (edge.first >= numNodes || edge.second >= numNodes) continue;
+    uf.unionXY(edge.first, edge.second);
+  }
+
+  return uf.connectedComponents(components);
+}
+
+void DedupePropVerts(manifold::Vec<glm::ivec3>& triProp,
+                     const Vec<thrust::pair<int, int>>& vert2vert) {
+  ZoneScoped;
+  std::vector<int> vertLabels(vert2vert.size());
+  const int numLabels = GetLabels(vertLabels, vert2vert, vert2vert.size());
+
+  std::vector<int> label2vert(numLabels);
+  for (int v = 0; v < vert2vert.size(); ++v) {
+    if (vertLabels[v] < numLabels) {
+      label2vert[vertLabels[v]] = v;
+    }
+  }
+  for (int tri = 0; tri < triProp.size(); ++tri) {
+    for (int i : {0, 1, 2}) {
+      if (triProp[tri][i] < vertLabels.size()) {
+        triProp[tri][i] = label2vert[vertLabels[triProp[tri][i]]];
+      }
+    }
+  }
+}
+}  // namespace
+
+namespace manifold {
+
+std::atomic<uint32_t> Manifold::Impl::meshIDCounter_(1);
+
+uint32_t Manifold::Impl::ReserveIDs(uint32_t n) {
+  return Manifold::Impl::meshIDCounter_.fetch_add(n, std::memory_order_relaxed);
+}
+
+Manifold::Impl::Impl(const MeshGL& meshGL,
+                     std::vector<float> propertyTolerance) {
+  Mesh mesh;
+  mesh.precision = meshGL.precision;
+  const int numVert = meshGL.NumVert();
+  const int numTri = meshGL.NumTri();
+
+  if (meshGL.numProp > 3 &&
+      static_cast<size_t>(numVert) * static_cast<size_t>(meshGL.numProp - 3) >=
+          std::numeric_limits<int>::max())
+    throw std::out_of_range("mesh too large");
+
+  if (meshGL.numProp < 3) {
+    MarkFailure(Error::MissingPositionProperties);
+    return;
+  }
+
+  mesh.triVerts.resize(numTri);
+  if (meshGL.mergeFromVert.size() != meshGL.mergeToVert.size()) {
+    MarkFailure(Error::MergeVectorsDifferentLengths);
+    return;
+  }
+
+  if (!meshGL.runTransform.empty() &&
+      12 * meshGL.runOriginalID.size() != meshGL.runTransform.size()) {
+    MarkFailure(Error::TransformWrongLength);
+    return;
+  }
+
+  if (!meshGL.runOriginalID.empty() && !meshGL.runIndex.empty() &&
+      meshGL.runOriginalID.size() + 1 != meshGL.runIndex.size()) {
+    MarkFailure(Error::RunIndexWrongLength);
+    return;
+  }
+
+  if (!meshGL.faceID.empty() && meshGL.faceID.size() != meshGL.NumTri()) {
+    MarkFailure(Error::FaceIDWrongLength);
+    return;
+  }
+
+  std::vector<int> prop2vert(numVert);
+  std::iota(prop2vert.begin(), prop2vert.end(), 0);
+  for (size_t i = 0; i < meshGL.mergeFromVert.size(); ++i) {
+    const int from = meshGL.mergeFromVert[i];
+    const int to = meshGL.mergeToVert[i];
+    if (from >= numVert || to >= numVert) {
+      MarkFailure(Error::MergeIndexOutOfBounds);
+      return;
+    }
+    prop2vert[from] = to;
+  }
+  for (size_t i = 0; i < numTri; ++i) {
+    for (const size_t j : {0, 1, 2}) {
+      const int vert = meshGL.triVerts[3 * i + j];
+      if (vert < 0 || vert >= numVert) {
+        MarkFailure(Error::VertexOutOfBounds);
+        return;
+      }
+      mesh.triVerts[i][j] = prop2vert[vert];
+    }
+  }
+
+  MeshRelationD relation;
+
+  if (meshGL.numProp > 3) {
+    relation.triProperties.resize(numTri);
+    for (size_t i = 0; i < numTri; ++i) {
+      for (const size_t j : {0, 1, 2}) {
+        relation.triProperties[i][j] = meshGL.triVerts[3 * i + j];
+      }
+    }
+  }
+
+  const int numProp = meshGL.numProp - 3;
+  relation.numProp = numProp;
+  relation.properties.resize(meshGL.NumVert() * numProp);
+  // This will have unreferenced duplicate positions that will be removed by
+  // Impl::RemoveUnreferencedVerts().
+  mesh.vertPos.resize(meshGL.NumVert());
+
+  for (int i = 0; i < meshGL.NumVert(); ++i) {
+    for (const int j : {0, 1, 2})
+      mesh.vertPos[i][j] = meshGL.vertProperties[meshGL.numProp * i + j];
+    for (int j = 0; j < numProp; ++j)
+      relation.properties[i * numProp + j] =
+          meshGL.vertProperties[meshGL.numProp * i + 3 + j];
+  }
+
+  mesh.halfedgeTangent.resize(meshGL.halfedgeTangent.size() / 4);
+  for (int i = 0; i < mesh.halfedgeTangent.size(); ++i) {
+    for (const int j : {0, 1, 2, 3})
+      mesh.halfedgeTangent[i][j] = meshGL.halfedgeTangent[4 * i + j];
+  }
+
+  if (meshGL.runOriginalID.empty()) {
+    relation.originalID = Impl::ReserveIDs(1);
+  } else {
+    std::vector<uint32_t> runIndex = meshGL.runIndex;
+    if (runIndex.empty()) {
+      runIndex = {0, 3 * meshGL.NumTri()};
+    }
+    relation.triRef.resize(meshGL.NumTri());
+    const int startID = Impl::ReserveIDs(meshGL.runOriginalID.size());
+    for (int i = 0; i < meshGL.runOriginalID.size(); ++i) {
+      const int meshID = startID + i;
+      const int originalID = meshGL.runOriginalID[i];
+      for (int tri = runIndex[i] / 3; tri < runIndex[i + 1] / 3; ++tri) {
+        TriRef& ref = relation.triRef[tri];
+        ref.meshID = meshID;
+        ref.originalID = originalID;
+        ref.tri = meshGL.faceID.empty() ? tri : meshGL.faceID[tri];
+      }
+
+      if (meshGL.runTransform.empty()) {
+        relation.meshIDtransform[meshID] = {originalID};
+      } else {
+        const float* m = meshGL.runTransform.data() + 12 * i;
+        relation.meshIDtransform[meshID] = {
+            originalID,
+            {m[0], m[1], m[2], m[3], m[4], m[5], m[6], m[7], m[8], m[9], m[10],
+             m[11]}};
+      }
+    }
+  }
+
+  *this = Impl(mesh, relation, propertyTolerance, !meshGL.faceID.empty());
+
+  // A Manifold created from an input mesh is never an original - the input is
+  // the original.
+  meshRelation_.originalID = -1;
+}
+
+/**
+ * Create a manifold from an input triangle Mesh. Will return an empty Manifold
+ * and set an Error Status if the Mesh is not manifold or otherwise invalid.
+ * TODO: update halfedgeTangent during SimplifyTopology.
+ */
+Manifold::Impl::Impl(const Mesh& mesh, const MeshRelationD& relation,
+                     const std::vector<float>& propertyTolerance,
+                     bool hasFaceIDs)
+    : vertPos_(mesh.vertPos), halfedgeTangent_(mesh.halfedgeTangent) {
+  if (mesh.triVerts.size() >= std::numeric_limits<int>::max())
+    throw std::out_of_range("mesh too large");
+  meshRelation_ = {relation.originalID, relation.numProp, relation.properties,
+                   relation.meshIDtransform};
+
+  Vec<glm::ivec3> triVerts;
+  for (size_t i = 0; i < mesh.triVerts.size(); ++i) {
+    const glm::ivec3 tri = mesh.triVerts[i];
+    // Remove topological degenerates
+    if (tri[0] != tri[1] && tri[1] != tri[2] && tri[2] != tri[0]) {
+      triVerts.push_back(tri);
+      if (relation.triRef.size() > 0) {
+        meshRelation_.triRef.push_back(relation.triRef[i]);
+      }
+      if (relation.triProperties.size() > 0) {
+        meshRelation_.triProperties.push_back(relation.triProperties[i]);
+      }
+    }
+  }
+
+  if (!IsIndexInBounds(triVerts)) {
+    MarkFailure(Error::VertexOutOfBounds);
+    return;
+  }
+  RemoveUnreferencedVerts(triVerts);
+
+  CalculateBBox();
+  if (!IsFinite()) {
+    MarkFailure(Error::NonFiniteVertex);
+    return;
+  }
+  SetPrecision(mesh.precision);
+
+  CreateHalfedges(triVerts);
+  if (!IsManifold()) {
+    MarkFailure(Error::NotManifold);
+    return;
+  }
+
+  SplitPinchedVerts();
+
+  CalculateNormals();
+
+  InitializeOriginal();
+  if (!hasFaceIDs) {
+    CreateFaces(propertyTolerance);
+  }
+
+  SimplifyTopology();
+  Finish();
+}
+
+/**
+ * Create either a unit tetrahedron, cube or octahedron. The cube is in the
+ * first octant, while the others are symmetric about the origin.
+ */
+Manifold::Impl::Impl(Shape shape, const glm::mat4x3 m) {
+  std::vector<glm::vec3> vertPos;
+  std::vector<glm::ivec3> triVerts;
+  switch (shape) {
+    case Shape::Tetrahedron:
+      vertPos = {{-1.0f, -1.0f, 1.0f},
+                 {-1.0f, 1.0f, -1.0f},
+                 {1.0f, -1.0f, -1.0f},
+                 {1.0f, 1.0f, 1.0f}};
+      triVerts = {{2, 0, 1}, {0, 3, 1}, {2, 3, 0}, {3, 2, 1}};
+      break;
+    case Shape::Cube:
+      vertPos = {{0.0f, 0.0f, 0.0f},  //
+                 {0.0f, 0.0f, 1.0f},  //
+                 {0.0f, 1.0f, 0.0f},  //
+                 {0.0f, 1.0f, 1.0f},  //
+                 {1.0f, 0.0f, 0.0f},  //
+                 {1.0f, 0.0f, 1.0f},  //
+                 {1.0f, 1.0f, 0.0f},  //
+                 {1.0f, 1.0f, 1.0f}};
+      triVerts = {{1, 0, 4}, {2, 4, 0},  //
+                  {1, 3, 0}, {3, 1, 5},  //
+                  {3, 2, 0}, {3, 7, 2},  //
+                  {5, 4, 6}, {5, 1, 4},  //
+                  {6, 4, 2}, {7, 6, 2},  //
+                  {7, 3, 5}, {7, 5, 6}};
+      break;
+    case Shape::Octahedron:
+      vertPos = {{1.0f, 0.0f, 0.0f},   //
+                 {-1.0f, 0.0f, 0.0f},  //
+                 {0.0f, 1.0f, 0.0f},   //
+                 {0.0f, -1.0f, 0.0f},  //
+                 {0.0f, 0.0f, 1.0f},   //
+                 {0.0f, 0.0f, -1.0f}};
+      triVerts = {{0, 2, 4}, {1, 5, 3},  //
+                  {2, 1, 4}, {3, 5, 0},  //
+                  {1, 3, 4}, {0, 5, 2},  //
+                  {3, 0, 4}, {2, 5, 1}};
+      break;
+  }
+  vertPos_ = vertPos;
+  for (auto& v : vertPos_) v = m * glm::vec4(v, 1.0f);
+  CreateHalfedges(triVerts);
+  Finish();
+  meshRelation_.originalID = ReserveIDs(1);
+  InitializeOriginal();
+  CreateFaces();
+}
+
+void Manifold::Impl::RemoveUnreferencedVerts(Vec<glm::ivec3>& triVerts) {
+  ZoneScoped;
+  Vec<int> vertOld2New(NumVert() + 1, 0);
+  auto policy = autoPolicy(NumVert());
+  for_each(policy, triVerts.cbegin(), triVerts.cend(),
+           MarkVerts({vertOld2New.view(1)}));
+
+  const Vec<glm::vec3> oldVertPos = vertPos_;
+  vertPos_.resize(copy_if<decltype(vertPos_.begin())>(
+                      policy, oldVertPos.cbegin(), oldVertPos.cend(),
+                      vertOld2New.cbegin() + 1, vertPos_.begin(),
+                      thrust::identity<int>()) -
+                  vertPos_.begin());
+
+  inclusive_scan(policy, vertOld2New.begin() + 1, vertOld2New.end(),
+                 vertOld2New.begin() + 1);
+
+  for_each(policy, triVerts.begin(), triVerts.end(),
+           ReindexTriVerts({vertOld2New}));
+}
+
+void Manifold::Impl::InitializeOriginal() {
+  const int meshID = meshRelation_.originalID;
+  // Don't initialize if it's not an original
+  if (meshID < 0) return;
+  meshRelation_.triRef.resize(NumTri());
+  for_each_n(autoPolicy(NumTri()),
+             zip(meshRelation_.triRef.begin(), countAt(0)), NumTri(),
+             InitializeTriRef({meshID, halfedge_}));
+  meshRelation_.meshIDtransform.clear();
+  meshRelation_.meshIDtransform[meshID] = {meshID};
+}
+
+void Manifold::Impl::CreateFaces(const std::vector<float>& propertyTolerance) {
+  ZoneScoped;
+  Vec<float> propertyToleranceD =
+      propertyTolerance.empty() ? Vec<float>(meshRelation_.numProp, kTolerance)
+                                : propertyTolerance;
+
+  Vec<thrust::pair<int, int>> face2face(halfedge_.size(), {-1, -1});
+  Vec<thrust::pair<int, int>> vert2vert(halfedge_.size(), {-1, -1});
+  Vec<float> triArea(NumTri());
+  for_each_n(
+      autoPolicy(halfedge_.size()),
+      zip(face2face.begin(), vert2vert.begin(), countAt(0)), halfedge_.size(),
+      CoplanarEdge({triArea, halfedge_, vertPos_, meshRelation_.triRef,
+                    meshRelation_.triProperties, meshRelation_.properties,
+                    propertyToleranceD, meshRelation_.numProp, precision_}));
+
+  if (meshRelation_.triProperties.size() > 0) {
+    DedupePropVerts(meshRelation_.triProperties, vert2vert);
+  }
+
+  std::vector<int> components;
+  const int numComponent = GetLabels(components, face2face, NumTri());
+
+  Vec<int> comp2tri(numComponent, -1);
+  for (int tri = 0; tri < NumTri(); ++tri) {
+    const int comp = components[tri];
+    const int current = comp2tri[comp];
+    if (current < 0 || triArea[tri] > triArea[current]) {
+      comp2tri[comp] = tri;
+      triArea[comp] = triArea[tri];
+    }
+  }
+
+  for_each_n(autoPolicy(halfedge_.size()), countAt(0), NumTri(),
+             CheckCoplanarity(
+                 {comp2tri, halfedge_, vertPos_, &components, precision_}));
+
+  Vec<TriRef>& triRef = meshRelation_.triRef;
+  for (int tri = 0; tri < NumTri(); ++tri) {
+    const int referenceTri = comp2tri[components[tri]];
+    if (referenceTri >= 0) {
+      triRef[tri].tri = referenceTri;
+    }
+  }
+}
+
+/**
+ * Create the halfedge_ data structure from an input triVerts array like Mesh.
+ */
+void Manifold::Impl::CreateHalfedges(const Vec<glm::ivec3>& triVerts) {
+  ZoneScoped;
+  const int numTri = triVerts.size();
+  const int numHalfedge = 3 * numTri;
+  // drop the old value first to avoid copy
+  halfedge_.resize(0);
+  halfedge_.resize(numHalfedge);
+  Vec<uint64_t> edge(numHalfedge);
+  Vec<int> ids(numHalfedge);
+  auto policy = autoPolicy(numTri);
+  sequence(policy, ids.begin(), ids.end());
+  for_each_n(policy, zip(countAt(0), triVerts.begin()), numTri,
+             Tri2Halfedges({halfedge_, edge}));
+  // Stable sort is required here so that halfedges from the same face are
+  // paired together (the triangles were created in face order). In some
+  // degenerate situations the triangulator can add the same internal edge in
+  // two different faces, causing this edge to not be 2-manifold. These are
+  // fixed by duplicating verts in SimplifyTopology.
+  stable_sort(policy, zip(edge.begin(), ids.begin()),
+              zip(edge.end(), ids.end()),
+              [](const thrust::tuple<uint64_t, int>& a,
+                 const thrust::tuple<uint64_t, int>& b) {
+                return thrust::get<0>(a) < thrust::get<0>(b);
+              });
+  // Once sorted, the first half of the range is the forward halfedges, which
+  // correspond to their backward pair at the same offset in the second half
+  // of the range.
+  for_each_n(policy, countAt(0), numHalfedge / 2,
+             LinkHalfedges({halfedge_, ids, numHalfedge / 2}));
+}
+
+/**
+ * Does a full recalculation of the face bounding boxes, including updating
+ * the collider, but does not resort the faces.
+ */
+void Manifold::Impl::Update() {
+  CalculateBBox();
+  Vec<Box> faceBox;
+  Vec<uint32_t> faceMorton;
+  GetFaceBoxMorton(faceBox, faceMorton);
+  collider_.UpdateBoxes(faceBox);
+}
+
+void Manifold::Impl::MarkFailure(Error status) {
+  bBox_ = Box();
+  vertPos_.resize(0);
+  halfedge_.resize(0);
+  vertNormal_.resize(0);
+  faceNormal_.resize(0);
+  halfedgeTangent_.resize(0);
+  meshRelation_ = MeshRelationD();
+  status_ = status;
+}
+
+void Manifold::Impl::Warp(std::function<void(glm::vec3&)> warpFunc) {
+  WarpBatch([&warpFunc](VecView<glm::vec3> vecs) {
+    thrust::for_each(thrust::host, vecs.begin(), vecs.end(), warpFunc);
+  });
+}
+
+void Manifold::Impl::WarpBatch(
+    std::function<void(VecView<glm::vec3>)> warpFunc) {
+  warpFunc(vertPos_.view());
+  CalculateBBox();
+  if (!IsFinite()) {
+    MarkFailure(Error::NonFiniteVertex);
+    return;
+  }
+  Update();
+  faceNormal_.resize(0);  // force recalculation of triNormal
+  CalculateNormals();
+  SetPrecision();
+  CreateFaces();
+  Finish();
+}
+
+Manifold::Impl Manifold::Impl::Transform(const glm::mat4x3& transform_) const {
+  ZoneScoped;
+  if (transform_ == glm::mat4x3(1.0f)) return *this;
+  auto policy = autoPolicy(NumVert());
+  Impl result;
+  result.collider_ = collider_;
+  result.meshRelation_ = meshRelation_;
+  result.precision_ = precision_;
+  result.bBox_ = bBox_;
+  result.halfedge_ = halfedge_;
+  result.halfedgeTangent_.resize(halfedgeTangent_.size());
+
+  result.meshRelation_.originalID = -1;
+  for (auto& m : result.meshRelation_.meshIDtransform) {
+    m.second.transform = transform_ * glm::mat4(m.second.transform);
+  }
+
+  result.vertPos_.resize(NumVert());
+  result.faceNormal_.resize(faceNormal_.size());
+  result.vertNormal_.resize(vertNormal_.size());
+  transform(policy, vertPos_.begin(), vertPos_.end(), result.vertPos_.begin(),
+            Transform4x3({transform_}));
+
+  glm::mat3 normalTransform = NormalTransform(transform_);
+  transform(policy, faceNormal_.begin(), faceNormal_.end(),
+            result.faceNormal_.begin(), TransformNormals({normalTransform}));
+  transform(policy, vertNormal_.begin(), vertNormal_.end(),
+            result.vertNormal_.begin(), TransformNormals({normalTransform}));
+
+  const bool invert = glm::determinant(glm::mat3(transform_)) < 0;
+
+  if (halfedgeTangent_.size() > 0) {
+    for_each_n(policy, zip(result.halfedgeTangent_.begin(), countAt(0)),
+               halfedgeTangent_.size(),
+               TransformTangents({glm::mat3(transform_), invert,
+                                  halfedgeTangent_, halfedge_}));
+  }
+
+  if (invert) {
+    for_each_n(policy, zip(result.meshRelation_.triRef.begin(), countAt(0)),
+               result.NumTri(), FlipTris({result.halfedge_}));
+  }
+
+  // This optimization does a cheap collider update if the transform is
+  // axis-aligned.
+  if (!result.collider_.Transform(transform_)) result.Update();
+
+  result.CalculateBBox();
+  // Scale the precision by the norm of the 3x3 portion of the transform.
+  result.precision_ *= SpectralNorm(glm::mat3(transform_));
+  // Maximum of inherited precision loss and translational precision loss.
+  result.SetPrecision(result.precision_);
+  return result;
+}
+
+/**
+ * Sets the precision based on the bounding box, and limits its minimum value
+ * by the optional input.
+ */
+void Manifold::Impl::SetPrecision(float minPrecision) {
+  precision_ = MaxPrecision(minPrecision, bBox_);
+}
+
+/**
+ * If face normals are already present, this function uses them to compute
+ * vertex normals (angle-weighted pseudo-normals); otherwise it also computes
+ * the face normals. Face normals are only calculated when needed because
+ * nearly degenerate faces will accrue rounding error, while the Boolean can
+ * retain their original normal, which is more accurate and can help with
+ * merging coplanar faces.
+ *
+ * If the face normals have been invalidated by an operation like Warp(),
+ * ensure you do faceNormal_.resize(0) before calling this function to force
+ * recalculation.
+ */
+void Manifold::Impl::CalculateNormals() {
+  ZoneScoped;
+  vertNormal_.resize(NumVert());
+  auto policy = autoPolicy(NumTri());
+  fill(policy, vertNormal_.begin(), vertNormal_.end(), glm::vec3(0));
+  bool calculateTriNormal = false;
+  if (faceNormal_.size() != NumTri()) {
+    faceNormal_.resize(NumTri());
+    calculateTriNormal = true;
+  }
+  for_each_n(policy, zip(faceNormal_.begin(), countAt(0)), NumTri(),
+             AssignNormals({vertNormal_, vertPos_, halfedge_, precision_,
+                            calculateTriNormal}));
+  for_each(policy, vertNormal_.begin(), vertNormal_.end(), Normalize());
+}
+
+/**
+ * Remaps all the contained meshIDs to new unique values to represent new
+ * instances of these meshes.
+ */
+void Manifold::Impl::IncrementMeshIDs() {
+  HashTable<uint32_t> meshIDold2new(meshRelation_.meshIDtransform.size() * 2);
+  // Update keys of the transform map
+  std::map<int, Relation> oldTransforms;
+  std::swap(meshRelation_.meshIDtransform, oldTransforms);
+  const int numMeshIDs = oldTransforms.size();
+  int nextMeshID = ReserveIDs(numMeshIDs);
+  for (const auto& pair : oldTransforms) {
+    meshIDold2new.D().Insert(pair.first, nextMeshID);
+    meshRelation_.meshIDtransform[nextMeshID++] = pair.second;
+  }
+
+  const int numTri = NumTri();
+  for_each_n(autoPolicy(numTri), meshRelation_.triRef.begin(), numTri,
+             UpdateMeshID({meshIDold2new.D()}));
+}
+
+/**
+ * Returns a sparse array of the bounding box overlaps between the edges of
+ * the input manifold, Q and the faces of this manifold. Returned indices only
+ * point to forward halfedges.
+ */
+SparseIndices Manifold::Impl::EdgeCollisions(const Impl& Q,
+                                             bool inverted) const {
+  ZoneScoped;
+  Vec<TmpEdge> edges = CreateTmpEdges(Q.halfedge_);
+  const int numEdge = edges.size();
+  Vec<Box> QedgeBB(numEdge);
+  auto policy = autoPolicy(numEdge);
+  for_each_n(policy, zip(QedgeBB.begin(), edges.cbegin()), numEdge,
+             EdgeBox({Q.vertPos_}));
+
+  SparseIndices q1p2(0);
+  if (inverted)
+    q1p2 = collider_.Collisions<false, true>(QedgeBB.cview());
+  else
+    q1p2 = collider_.Collisions<false, false>(QedgeBB.cview());
+
+  if (inverted)
+    for_each(policy, countAt(0_z), countAt(q1p2.size()),
+             ReindexEdge<true>({edges, q1p2}));
+  else
+    for_each(policy, countAt(0_z), countAt(q1p2.size()),
+             ReindexEdge<false>({edges, q1p2}));
+  return q1p2;
+}
+
+/**
+ * Returns a sparse array of the input vertices that project inside the XY
+ * bounding boxes of the faces of this manifold.
+ */
+SparseIndices Manifold::Impl::VertexCollisionsZ(
+    VecView<const glm::vec3> vertsIn, bool inverted) const {
+  ZoneScoped;
+  if (inverted)
+    return collider_.Collisions<false, true>(vertsIn);
+  else
+    return collider_.Collisions<false, false>(vertsIn);
+}
+
+/*
+ * Returns the minimum gap between two manifolds. Returns a float between
+ * 0 and searchLength.
+ */
+float Manifold::Impl::MinGap(const Manifold::Impl& other,
+                             float searchLength) const {
+  ZoneScoped;
+  Vec<Box> faceBoxOther;
+  Vec<uint32_t> faceMortonOther;
+
+  other.GetFaceBoxMorton(faceBoxOther, faceMortonOther);
+
+  transform(autoPolicy(faceBoxOther.size()), faceBoxOther.begin(),
+            faceBoxOther.end(), faceBoxOther.begin(),
+            [searchLength](const Box& box) {
+              return Box(box.min - glm::vec3(searchLength),
+                         box.max + glm::vec3(searchLength));
+            });
+
+  SparseIndices collisions = collider_.Collisions(faceBoxOther.cview());
+
+  float minDistanceSquared = transform_reduce<float>(
+      autoPolicy(collisions.size()), thrust::counting_iterator<int>(0),
+      thrust::counting_iterator<int>(collisions.size()),
+      [&collisions, this, &other](int i) {
+        const int tri = collisions.Get(i, 1);
+        const int triOther = collisions.Get(i, 0);
+
+        std::array<glm::vec3, 3> p;
+        std::array<glm::vec3, 3> q;
+
+        for (const int j : {0, 1, 2}) {
+          p[j] = vertPos_[halfedge_[3 * tri + j].startVert];
+          q[j] = other.vertPos_[other.halfedge_[3 * triOther + j].startVert];
+        }
+
+        return DistanceTriangleTriangleSquared(p, q);
+      },
+      searchLength * searchLength, thrust::minimum<float>());
+
+  return sqrt(minDistanceSquared);
+};
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/impl.h b/thirdparty/manifold/src/manifold/src/impl.h
new file mode 100644
index 000000000000..859e14905dc4
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/impl.h
@@ -0,0 +1,190 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <map>
+
+#include "collider.h"
+#include "manifold.h"
+#include "optional_assert.h"
+#include "polygon.h"
+#include "shared.h"
+#include "sparse.h"
+#include "utils.h"
+#include "vec.h"
+
+namespace manifold {
+
+/** @ingroup Private */
+struct Manifold::Impl {
+  struct Relation {
+    int originalID = -1;
+    glm::mat4x3 transform = glm::mat4x3(1);
+    bool backSide = false;
+  };
+  struct MeshRelationD {
+    /// The originalID of this Manifold if it is an original; -1 otherwise.
+    int originalID = -1;
+    int numProp = 0;
+    Vec<float> properties;
+    std::map<int, Relation> meshIDtransform;
+    Vec<TriRef> triRef;
+    Vec<glm::ivec3> triProperties;
+  };
+  struct BaryIndices {
+    int tri, start4, end4;
+  };
+
+  Box bBox_;
+  float precision_ = -1;
+  Error status_ = Error::NoError;
+  Vec<glm::vec3> vertPos_;
+  Vec<Halfedge> halfedge_;
+  Vec<glm::vec3> vertNormal_;
+  Vec<glm::vec3> faceNormal_;
+  Vec<glm::vec4> halfedgeTangent_;
+  MeshRelationD meshRelation_;
+  Collider collider_;
+
+  static std::atomic<uint32_t> meshIDCounter_;
+  static uint32_t ReserveIDs(uint32_t);
+
+  Impl() {}
+  enum class Shape { Tetrahedron, Cube, Octahedron };
+  Impl(Shape, const glm::mat4x3 = glm::mat4x3(1));
+
+  Impl(const MeshGL&, std::vector<float> propertyTolerance = {});
+  Impl(const Mesh&, const MeshRelationD& relation,
+       const std::vector<float>& propertyTolerance = {},
+       bool hasFaceIDs = false);
+
+  inline void ForVert(int halfedge, std::function<void(int halfedge)> func) {
+    int current = halfedge;
+    do {
+      current = NextHalfedge(halfedge_[current].pairedHalfedge);
+      func(current);
+    } while (current != halfedge);
+  }
+
+  template <typename T>
+  void ForVert(int halfedge, std::function<T(int halfedge)> transform,
+               std::function<void(int halfedge, const T& here, const T& next)>
+                   binaryOp) {
+    T here = transform(halfedge);
+    int current = halfedge;
+    do {
+      const int nextHalfedge = NextHalfedge(halfedge_[current].pairedHalfedge);
+      const T next = transform(nextHalfedge);
+      binaryOp(current, here, next);
+      here = next;
+      current = nextHalfedge;
+    } while (current != halfedge);
+  }
+
+  void CreateFaces(const std::vector<float>& propertyTolerance = {});
+  void RemoveUnreferencedVerts(Vec<glm::ivec3>& triVerts);
+  void InitializeOriginal();
+  void CreateHalfedges(const Vec<glm::ivec3>& triVerts);
+  void CalculateNormals();
+  void IncrementMeshIDs();
+
+  void Update();
+  void MarkFailure(Error status);
+  void Warp(std::function<void(glm::vec3&)> warpFunc);
+  void WarpBatch(std::function<void(VecView<glm::vec3>)> warpFunc);
+  Impl Transform(const glm::mat4x3& transform) const;
+  SparseIndices EdgeCollisions(const Impl& B, bool inverted = false) const;
+  SparseIndices VertexCollisionsZ(VecView<const glm::vec3> vertsIn,
+                                  bool inverted = false) const;
+  float MinGap(const Impl& other, float searchLength) const;
+
+  bool IsEmpty() const { return NumVert() == 0; }
+  int NumVert() const { return vertPos_.size(); }
+  int NumEdge() const { return halfedge_.size() / 2; }
+  int NumTri() const { return halfedge_.size() / 3; }
+  int NumProp() const { return meshRelation_.numProp; }
+  int NumPropVert() const {
+    return NumProp() == 0 ? NumVert()
+                          : meshRelation_.properties.size() / NumProp();
+  }
+
+  // properties.cu
+  Properties GetProperties() const;
+  void CalculateCurvature(int gaussianIdx, int meanIdx);
+  void CalculateBBox();
+  bool IsFinite() const;
+  bool IsIndexInBounds(VecView<const glm::ivec3> triVerts) const;
+  void SetPrecision(float minPrecision = -1);
+  bool IsManifold() const;
+  bool Is2Manifold() const;
+  bool MatchesTriNormals() const;
+  int NumDegenerateTris() const;
+
+  // sort.cu
+  void Finish();
+  void SortVerts();
+  void ReindexVerts(const Vec<int>& vertNew2Old, int numOldVert);
+  void CompactProps();
+  void GetFaceBoxMorton(Vec<Box>& faceBox, Vec<uint32_t>& faceMorton) const;
+  void SortFaces(Vec<Box>& faceBox, Vec<uint32_t>& faceMorton);
+  void GatherFaces(const Vec<int>& faceNew2Old);
+  void GatherFaces(const Impl& old, const Vec<int>& faceNew2Old);
+
+  // face_op.cu
+  void Face2Tri(const Vec<int>& faceEdge, const Vec<TriRef>& halfedgeRef);
+  PolygonsIdx Face2Polygons(VecView<Halfedge>::IterC start,
+                            VecView<Halfedge>::IterC end,
+                            glm::mat3x2 projection) const;
+  CrossSection Slice(float height) const;
+  CrossSection Project() const;
+
+  // edge_op.cu
+  void SimplifyTopology();
+  void DedupeEdge(int edge);
+  void CollapseEdge(int edge, std::vector<int>& edges);
+  void RecursiveEdgeSwap(int edge, int& tag, std::vector<int>& visited,
+                         std::vector<int>& edgeSwapStack,
+                         std::vector<int>& edges);
+  void RemoveIfFolded(int edge);
+  void PairUp(int edge0, int edge1);
+  void UpdateVert(int vert, int startEdge, int endEdge);
+  void FormLoop(int current, int end);
+  void CollapseTri(const glm::ivec3& triEdge);
+  void SplitPinchedVerts();
+
+  // subdivision.cpp
+  int GetNeighbor(int tri) const;
+  glm::ivec4 GetHalfedges(int tri) const;
+  BaryIndices GetIndices(int halfedge) const;
+  void FillRetainedVerts(Vec<Barycentric>& vertBary) const;
+  Vec<Barycentric> Subdivide(std::function<int(glm::vec3)>);
+
+  // smoothing.cpp
+  bool IsInsideQuad(int halfedge) const;
+  bool IsMarkedInsideQuad(int halfedge) const;
+  glm::vec3 GetNormal(int halfedge, int normalIdx) const;
+  std::vector<Smoothness> UpdateSharpenedEdges(
+      const std::vector<Smoothness>&) const;
+  Vec<bool> FlatFaces() const;
+  Vec<int> VertFlatFace(const Vec<bool>&) const;
+  std::vector<Smoothness> SharpenEdges(float minSharpAngle,
+                                       float minSmoothness) const;
+  void SharpenTangent(int halfedge, float smoothness);
+  void SetNormals(int normalIdx, float minSharpAngle);
+  void LinearizeFlatTangents();
+  void CreateTangents(int normalIdx);
+  void CreateTangents(std::vector<Smoothness>);
+  void Refine(std::function<int(glm::vec3)>);
+};
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/manifold.cpp b/thirdparty/manifold/src/manifold/src/manifold.cpp
new file mode 100644
index 000000000000..4b9c46a2f1c1
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/manifold.cpp
@@ -0,0 +1,970 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <algorithm>
+#include <map>
+#include <numeric>
+
+#include "QuickHull.hpp"
+#include "boolean3.h"
+#include "csg_tree.h"
+#include "impl.h"
+#include "par.h"
+#include "tri_dist.h"
+
+namespace {
+using namespace manifold;
+using namespace thrust::placeholders;
+
+ExecutionParams manifoldParams;
+
+struct MakeTri {
+  VecView<const Halfedge> halfedges;
+
+  void operator()(thrust::tuple<glm::ivec3&, int> inOut) {
+    glm::ivec3& tri = thrust::get<0>(inOut);
+    const int face = 3 * thrust::get<1>(inOut);
+
+    for (int i : {0, 1, 2}) {
+      tri[i] = halfedges[face + i].startVert;
+    }
+  }
+};
+
+struct UpdateProperties {
+  float* properties;
+  const int numProp;
+  const float* oldProperties;
+  const int numOldProp;
+  const glm::vec3* vertPos;
+  const glm::ivec3* triProperties;
+  const Halfedge* halfedges;
+  std::function<void(float*, glm::vec3, const float*)> propFunc;
+
+  void operator()(int tri) {
+    for (int i : {0, 1, 2}) {
+      const int vert = halfedges[3 * tri + i].startVert;
+      const int propVert = triProperties[tri][i];
+      propFunc(properties + numProp * propVert, vertPos[vert],
+               oldProperties + numOldProp * propVert);
+    }
+  }
+};
+
+Manifold Halfspace(Box bBox, glm::vec3 normal, float originOffset) {
+  normal = glm::normalize(normal);
+  Manifold cutter =
+      Manifold::Cube(glm::vec3(2.0f), true).Translate({1.0f, 0.0f, 0.0f});
+  float size = glm::length(bBox.Center() - normal * originOffset) +
+               0.5f * glm::length(bBox.Size());
+  cutter = cutter.Scale(glm::vec3(size)).Translate({originOffset, 0.0f, 0.0f});
+  float yDeg = glm::degrees(-glm::asin(normal.z));
+  float zDeg = glm::degrees(glm::atan(normal.y, normal.x));
+  return cutter.Rotate(0.0f, yDeg, zDeg);
+}
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Construct an empty Manifold.
+ *
+ */
+Manifold::Manifold() : pNode_{std::make_shared<CsgLeafNode>()} {}
+Manifold::~Manifold() = default;
+Manifold::Manifold(Manifold&&) noexcept = default;
+Manifold& Manifold::operator=(Manifold&&) noexcept = default;
+
+Manifold::Manifold(const Manifold& other) : pNode_(other.pNode_) {}
+
+Manifold::Manifold(std::shared_ptr<CsgNode> pNode) : pNode_(pNode) {}
+
+Manifold::Manifold(std::shared_ptr<Impl> pImpl_)
+    : pNode_(std::make_shared<CsgLeafNode>(pImpl_)) {}
+
+Manifold Manifold::Invalid() {
+  auto pImpl_ = std::make_shared<Impl>();
+  pImpl_->status_ = Error::InvalidConstruction;
+  return Manifold(pImpl_);
+}
+
+Manifold& Manifold::operator=(const Manifold& other) {
+  if (this != &other) {
+    pNode_ = other.pNode_;
+  }
+  return *this;
+}
+
+CsgLeafNode& Manifold::GetCsgLeafNode() const {
+  if (pNode_->GetNodeType() != CsgNodeType::Leaf) {
+    pNode_ = pNode_->ToLeafNode();
+  }
+  return *std::static_pointer_cast<CsgLeafNode>(pNode_);
+}
+
+/**
+ * Convert a MeshGL into a Manifold, retaining its properties and merging only
+ * the positions according to the merge vectors. Will return an empty Manifold
+ * and set an Error Status if the result is not an oriented 2-manifold. Will
+ * collapse degenerate triangles and unnecessary vertices.
+ *
+ * All fields are read, making this structure suitable for a lossless round-trip
+ * of data from GetMeshGL. For multi-material input, use ReserveIDs to set a
+ * unique originalID for each material, and sort the materials into triangle
+ * runs.
+ *
+ * @param meshGL The input MeshGL.
+ * @param propertyTolerance A vector of precision values for each property
+ * beyond position. If specified, the propertyTolerance vector must have size =
+ * numProp - 3. This is the amount of interpolation error allowed before two
+ * neighboring triangles are considered to be on a property boundary edge.
+ * Property boundary edges will be retained across operations even if the
+ * triangles are coplanar. Defaults to 1e-5, which works well for most
+ * properties in the [-1, 1] range.
+ */
+Manifold::Manifold(const MeshGL& meshGL,
+                   const std::vector<float>& propertyTolerance)
+    : pNode_(std::make_shared<CsgLeafNode>(
+          std::make_shared<Impl>(meshGL, propertyTolerance))) {}
+
+/**
+ * Convert a Mesh into a Manifold. Will return an empty Manifold
+ * and set an Error Status if the Mesh is not an oriented 2-manifold. Will
+ * collapse degenerate triangles and unnecessary vertices.
+ *
+ * @param mesh The input Mesh.
+ */
+Manifold::Manifold(const Mesh& mesh) {
+  Impl::MeshRelationD relation = {(int)ReserveIDs(1)};
+  pNode_ =
+      std::make_shared<CsgLeafNode>(std::make_shared<Impl>(mesh, relation));
+}
+
+/**
+ * This returns a Mesh of simple vectors of vertices and triangles suitable for
+ * saving or other operations outside of the context of this library.
+ */
+Mesh Manifold::GetMesh() const {
+  ZoneScoped;
+  const Impl& impl = *GetCsgLeafNode().GetImpl();
+
+  Mesh result;
+  result.precision = Precision();
+  result.vertPos.insert(result.vertPos.end(), impl.vertPos_.begin(),
+                        impl.vertPos_.end());
+  result.vertNormal.insert(result.vertNormal.end(), impl.vertNormal_.begin(),
+                           impl.vertNormal_.end());
+  result.halfedgeTangent.insert(result.halfedgeTangent.end(),
+                                impl.halfedgeTangent_.begin(),
+                                impl.halfedgeTangent_.end());
+
+  result.triVerts.resize(NumTri());
+  // note that `triVerts` is `std::vector`, so we cannot use thrust::device
+  thrust::for_each_n(thrust::host, zip(result.triVerts.begin(), countAt(0)),
+                     NumTri(), MakeTri({impl.halfedge_}));
+
+  return result;
+}
+
+/**
+ * The most complete output of this library, returning a MeshGL that is designed
+ * to easily push into a renderer, including all interleaved vertex properties
+ * that may have been input. It also includes relations to all the input meshes
+ * that form a part of this result and the transforms applied to each.
+ *
+ * @param normalIdx If the original MeshGL inputs that formed this manifold had
+ * properties corresponding to normal vectors, you can specify which property
+ * channels these are (x, y, z), which will cause this output MeshGL to
+ * automatically update these normals according to the applied transforms and
+ * front/back side. Each channel must be >= 3 and < numProp, and all original
+ * MeshGLs must use the same channels for their normals.
+ */
+MeshGL Manifold::GetMeshGL(glm::ivec3 normalIdx) const {
+  ZoneScoped;
+  const Impl& impl = *GetCsgLeafNode().GetImpl();
+
+  const int numProp = NumProp();
+  const int numVert = NumPropVert();
+  const int numTri = NumTri();
+
+  const bool isOriginal = impl.meshRelation_.originalID >= 0;
+  const bool updateNormals =
+      !isOriginal && glm::all(glm::greaterThan(normalIdx, glm::ivec3(2)));
+
+  MeshGL out;
+  out.precision = Precision();
+  out.numProp = 3 + numProp;
+  out.triVerts.resize(3 * numTri);
+
+  const int numHalfedge = impl.halfedgeTangent_.size();
+  out.halfedgeTangent.resize(4 * numHalfedge);
+  for (int i = 0; i < numHalfedge; ++i) {
+    const glm::vec4 t = impl.halfedgeTangent_[i];
+    out.halfedgeTangent[4 * i] = t.x;
+    out.halfedgeTangent[4 * i + 1] = t.y;
+    out.halfedgeTangent[4 * i + 2] = t.z;
+    out.halfedgeTangent[4 * i + 3] = t.w;
+  }
+
+  // Sort the triangles into runs
+  out.faceID.resize(numTri);
+  std::vector<int> triNew2Old(numTri);
+  std::iota(triNew2Old.begin(), triNew2Old.end(), 0);
+  VecView<const TriRef> triRef = impl.meshRelation_.triRef;
+  // Don't sort originals - keep them in order
+  if (!isOriginal) {
+    std::sort(triNew2Old.begin(), triNew2Old.end(), [triRef](int a, int b) {
+      return triRef[a].originalID == triRef[b].originalID
+                 ? triRef[a].meshID < triRef[b].meshID
+                 : triRef[a].originalID < triRef[b].originalID;
+    });
+  }
+
+  std::vector<glm::mat3> runNormalTransform;
+
+  auto addRun = [updateNormals, isOriginal](
+                    MeshGL& out, std::vector<glm::mat3>& runNormalTransform,
+                    int tri, const Impl::Relation& rel) {
+    out.runIndex.push_back(3 * tri);
+    out.runOriginalID.push_back(rel.originalID);
+    if (updateNormals) {
+      runNormalTransform.push_back(NormalTransform(rel.transform) *
+                                   (rel.backSide ? -1.0f : 1.0f));
+    }
+    if (!isOriginal) {
+      for (const int col : {0, 1, 2, 3}) {
+        for (const int row : {0, 1, 2}) {
+          out.runTransform.push_back(rel.transform[col][row]);
+        }
+      }
+    }
+  };
+
+  auto meshIDtransform = impl.meshRelation_.meshIDtransform;
+  int lastID = -1;
+  for (int tri = 0; tri < numTri; ++tri) {
+    const int oldTri = triNew2Old[tri];
+    const auto ref = triRef[oldTri];
+    const int meshID = ref.meshID;
+
+    out.faceID[tri] = ref.tri;
+    for (const int i : {0, 1, 2})
+      out.triVerts[3 * tri + i] = impl.halfedge_[3 * oldTri + i].startVert;
+
+    if (meshID != lastID) {
+      Impl::Relation rel;
+      auto it = meshIDtransform.find(meshID);
+      if (it != meshIDtransform.end()) rel = it->second;
+      addRun(out, runNormalTransform, tri, rel);
+      meshIDtransform.erase(meshID);
+      lastID = meshID;
+    }
+  }
+  // Add runs for originals that did not contribute any faces to the output
+  for (const auto& pair : meshIDtransform) {
+    addRun(out, runNormalTransform, numTri, pair.second);
+  }
+  out.runIndex.push_back(3 * numTri);
+
+  // Early return for no props
+  if (numProp == 0) {
+    out.vertProperties.resize(3 * numVert);
+    for (int i = 0; i < numVert; ++i) {
+      const glm::vec3 v = impl.vertPos_[i];
+      out.vertProperties[3 * i] = v.x;
+      out.vertProperties[3 * i + 1] = v.y;
+      out.vertProperties[3 * i + 2] = v.z;
+    }
+    return out;
+  }
+
+  // Duplicate verts with different props
+  std::vector<int> vert2idx(impl.NumVert(), -1);
+  std::vector<std::vector<glm::ivec2>> vertPropPair(impl.NumVert());
+  out.vertProperties.reserve(numVert * static_cast<size_t>(out.numProp));
+
+  for (int run = 0; run < out.runOriginalID.size(); ++run) {
+    for (int tri = out.runIndex[run] / 3; tri < out.runIndex[run + 1] / 3;
+         ++tri) {
+      const glm::ivec3 triProp =
+          impl.meshRelation_.triProperties[triNew2Old[tri]];
+      for (const int i : {0, 1, 2}) {
+        const int prop = triProp[i];
+        const int vert = out.triVerts[3 * tri + i];
+
+        auto& bin = vertPropPair[vert];
+        bool bFound = false;
+        for (int k = 0; k < bin.size(); ++k) {
+          if (bin[k].x == prop) {
+            bFound = true;
+            out.triVerts[3 * tri + i] = bin[k].y;
+            break;
+          }
+        }
+        if (bFound) continue;
+        const int idx = out.vertProperties.size() / out.numProp;
+        out.triVerts[3 * tri + i] = idx;
+        bin.push_back({prop, idx});
+
+        for (int p : {0, 1, 2}) {
+          out.vertProperties.push_back(impl.vertPos_[vert][p]);
+        }
+        for (int p = 0; p < numProp; ++p) {
+          out.vertProperties.push_back(
+              impl.meshRelation_.properties[prop * numProp + p]);
+        }
+
+        if (updateNormals) {
+          glm::vec3 normal;
+          const int start = out.vertProperties.size() - out.numProp;
+          for (int i : {0, 1, 2}) {
+            normal[i] = out.vertProperties[start + normalIdx[i]];
+          }
+          normal = glm::normalize(runNormalTransform[run] * normal);
+          for (int i : {0, 1, 2}) {
+            out.vertProperties[start + normalIdx[i]] = normal[i];
+          }
+        }
+
+        if (vert2idx[vert] == -1) {
+          vert2idx[vert] = idx;
+        } else {
+          out.mergeFromVert.push_back(idx);
+          out.mergeToVert.push_back(vert2idx[vert]);
+        }
+      }
+    }
+  }
+  return out;
+}
+
+/**
+ * Does the Manifold have any triangles?
+ */
+bool Manifold::IsEmpty() const { return GetCsgLeafNode().GetImpl()->IsEmpty(); }
+/**
+ * Returns the reason for an input Mesh producing an empty Manifold. This Status
+ * only applies to Manifolds newly-created from an input Mesh - once they are
+ * combined into a new Manifold via operations, the status reverts to NoError,
+ * simply processing the problem mesh as empty. Likewise, empty meshes may still
+ * show NoError, for instance if they are small enough relative to their
+ * precision to be collapsed to nothing.
+ */
+Manifold::Error Manifold::Status() const {
+  return GetCsgLeafNode().GetImpl()->status_;
+}
+/**
+ * The number of vertices in the Manifold.
+ */
+int Manifold::NumVert() const { return GetCsgLeafNode().GetImpl()->NumVert(); }
+/**
+ * The number of edges in the Manifold.
+ */
+int Manifold::NumEdge() const { return GetCsgLeafNode().GetImpl()->NumEdge(); }
+/**
+ * The number of triangles in the Manifold.
+ */
+int Manifold::NumTri() const { return GetCsgLeafNode().GetImpl()->NumTri(); }
+/**
+ * The number of properties per vertex in the Manifold.
+ */
+int Manifold::NumProp() const { return GetCsgLeafNode().GetImpl()->NumProp(); }
+/**
+ * The number of property vertices in the Manifold. This will always be >=
+ * NumVert, as some physical vertices may be duplicated to account for different
+ * properties on different neighboring triangles.
+ */
+int Manifold::NumPropVert() const {
+  return GetCsgLeafNode().GetImpl()->NumPropVert();
+}
+
+/**
+ * Returns the axis-aligned bounding box of all the Manifold's vertices.
+ */
+Box Manifold::BoundingBox() const { return GetCsgLeafNode().GetImpl()->bBox_; }
+
+/**
+ * Returns the precision of this Manifold's vertices, which tracks the
+ * approximate rounding error over all the transforms and operations that have
+ * led to this state. Any triangles that are colinear within this precision are
+ * considered degenerate and removed. This is the value of &epsilon; defining
+ * [&epsilon;-valid](https://github.com/elalish/manifold/wiki/Manifold-Library#definition-of-%CE%B5-valid).
+ */
+float Manifold::Precision() const {
+  return GetCsgLeafNode().GetImpl()->precision_;
+}
+
+/**
+ * The genus is a topological property of the manifold, representing the number
+ * of "handles". A sphere is 0, torus 1, etc. It is only meaningful for a single
+ * mesh, so it is best to call Decompose() first.
+ */
+int Manifold::Genus() const {
+  int chi = NumVert() - NumEdge() + NumTri();
+  return 1 - chi / 2;
+}
+
+/**
+ * Returns the surface area and volume of the manifold.
+ */
+Properties Manifold::GetProperties() const {
+  return GetCsgLeafNode().GetImpl()->GetProperties();
+}
+
+/**
+ * If this mesh is an original, this returns its meshID that can be referenced
+ * by product manifolds' MeshRelation. If this manifold is a product, this
+ * returns -1.
+ */
+int Manifold::OriginalID() const {
+  return GetCsgLeafNode().GetImpl()->meshRelation_.originalID;
+}
+
+/**
+ * This function condenses all coplanar faces in the relation, and
+ * collapses those edges. In the process the relation to ancestor meshes is lost
+ * and this new Manifold is marked an original. Properties are preserved, so if
+ * they do not match across an edge, that edge will be kept.
+ */
+Manifold Manifold::AsOriginal() const {
+  auto newImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  newImpl->meshRelation_.originalID = ReserveIDs(1);
+  newImpl->InitializeOriginal();
+  newImpl->CreateFaces();
+  newImpl->SimplifyTopology();
+  newImpl->Finish();
+  return Manifold(std::make_shared<CsgLeafNode>(newImpl));
+}
+
+/**
+ * Returns the first of n sequential new unique mesh IDs for marking sets of
+ * triangles that can be looked up after further operations. Assign to
+ * MeshGL.runOriginalID vector.
+ */
+uint32_t Manifold::ReserveIDs(uint32_t n) {
+  return Manifold::Impl::ReserveIDs(n);
+}
+
+/**
+ * The triangle normal vectors are saved over the course of operations rather
+ * than recalculated to avoid rounding error. This checks that triangles still
+ * match their normal vectors within Precision().
+ */
+bool Manifold::MatchesTriNormals() const {
+  return GetCsgLeafNode().GetImpl()->MatchesTriNormals();
+}
+
+/**
+ * The number of triangles that are colinear within Precision(). This library
+ * attempts to remove all of these, but it cannot always remove all of them
+ * without changing the mesh by too much.
+ */
+int Manifold::NumDegenerateTris() const {
+  return GetCsgLeafNode().GetImpl()->NumDegenerateTris();
+}
+
+/**
+ * This is a checksum-style verification of the collider, simply returning the
+ * total number of edge-face bounding box overlaps between this and other.
+ *
+ * @param other A Manifold to overlap with.
+ */
+int Manifold::NumOverlaps(const Manifold& other) const {
+  SparseIndices overlaps = GetCsgLeafNode().GetImpl()->EdgeCollisions(
+      *other.GetCsgLeafNode().GetImpl());
+  int num_overlaps = overlaps.size();
+
+  overlaps = other.GetCsgLeafNode().GetImpl()->EdgeCollisions(
+      *GetCsgLeafNode().GetImpl());
+  return num_overlaps + overlaps.size();
+}
+
+/**
+ * Move this Manifold in space. This operation can be chained. Transforms are
+ * combined and applied lazily.
+ *
+ * @param v The vector to add to every vertex.
+ */
+Manifold Manifold::Translate(glm::vec3 v) const {
+  return Manifold(pNode_->Translate(v));
+}
+
+/**
+ * Scale this Manifold in space. This operation can be chained. Transforms are
+ * combined and applied lazily.
+ *
+ * @param v The vector to multiply every vertex by per component.
+ */
+Manifold Manifold::Scale(glm::vec3 v) const {
+  return Manifold(pNode_->Scale(v));
+}
+
+/**
+ * Applies an Euler angle rotation to the manifold, first about the X axis, then
+ * Y, then Z, in degrees. We use degrees so that we can minimize rounding error,
+ * and eliminate it completely for any multiples of 90 degrees. Additionally,
+ * more efficient code paths are used to update the manifold when the transforms
+ * only rotate by multiples of 90 degrees. This operation can be chained.
+ * Transforms are combined and applied lazily.
+ *
+ * @param xDegrees First rotation, degrees about the X-axis.
+ * @param yDegrees Second rotation, degrees about the Y-axis.
+ * @param zDegrees Third rotation, degrees about the Z-axis.
+ */
+Manifold Manifold::Rotate(float xDegrees, float yDegrees,
+                          float zDegrees) const {
+  return Manifold(pNode_->Rotate(xDegrees, yDegrees, zDegrees));
+}
+
+/**
+ * Transform this Manifold in space. The first three columns form a 3x3 matrix
+ * transform and the last is a translation vector. This operation can be
+ * chained. Transforms are combined and applied lazily.
+ *
+ * @param m The affine transform matrix to apply to all the vertices.
+ */
+Manifold Manifold::Transform(const glm::mat4x3& m) const {
+  return Manifold(pNode_->Transform(m));
+}
+
+/**
+ * Mirror this Manifold over the plane described by the unit form of the given
+ * normal vector. If the length of the normal is zero, an empty Manifold is
+ * returned. This operation can be chained. Transforms are combined and applied
+ * lazily.
+ *
+ * @param normal The normal vector of the plane to be mirrored over
+ */
+Manifold Manifold::Mirror(glm::vec3 normal) const {
+  if (glm::length(normal) == 0.) {
+    return Manifold();
+  }
+  auto n = glm::normalize(normal);
+  auto m = glm::mat4x3(glm::mat3(1.0f) - 2.0f * glm::outerProduct(n, n));
+  return Manifold(pNode_->Transform(m));
+}
+
+/**
+ * This function does not change the topology, but allows the vertices to be
+ * moved according to any arbitrary input function. It is easy to create a
+ * function that warps a geometrically valid object into one which overlaps, but
+ * that is not checked here, so it is up to the user to choose their function
+ * with discretion.
+ *
+ * @param warpFunc A function that modifies a given vertex position.
+ */
+Manifold Manifold::Warp(std::function<void(glm::vec3&)> warpFunc) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  pImpl->Warp(warpFunc);
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Same as Manifold::Warp but calls warpFunc with with
+ * a VecView which is roughly equivalent to std::span
+ * pointing to all vec3 elements to be modified in-place
+ *
+ * @param warpFunc A function that modifies multiple vertex positions.
+ */
+Manifold Manifold::WarpBatch(
+    std::function<void(VecView<glm::vec3>)> warpFunc) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  pImpl->WarpBatch(warpFunc);
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Create a new copy of this manifold with updated vertex properties by
+ * supplying a function that takes the existing position and properties as
+ * input. You may specify any number of output properties, allowing creation and
+ * removal of channels. Note: undefined behavior will result if you read past
+ * the number of input properties or write past the number of output properties.
+ *
+ * @param numProp The new number of properties per vertex.
+ * @param propFunc A function that modifies the properties of a given vertex.
+ */
+Manifold Manifold::SetProperties(
+    int numProp, std::function<void(float* newProp, glm::vec3 position,
+                                    const float* oldProp)>
+                     propFunc) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  const int oldNumProp = NumProp();
+  const Vec<float> oldProperties = pImpl->meshRelation_.properties;
+
+  auto& triProperties = pImpl->meshRelation_.triProperties;
+  if (numProp == 0) {
+    triProperties.resize(0);
+    pImpl->meshRelation_.properties.resize(0);
+  } else {
+    if (triProperties.size() == 0) {
+      const int numTri = NumTri();
+      triProperties.resize(numTri);
+      int idx = 0;
+      for (int i = 0; i < numTri; ++i) {
+        for (const int j : {0, 1, 2}) {
+          triProperties[i][j] = idx++;
+        }
+      }
+      pImpl->meshRelation_.properties = Vec<float>(numProp * idx, 0);
+    } else {
+      pImpl->meshRelation_.properties = Vec<float>(numProp * NumPropVert(), 0);
+    }
+    thrust::for_each_n(
+        thrust::host, countAt(0), NumTri(),
+        UpdateProperties({pImpl->meshRelation_.properties.data(), numProp,
+                          oldProperties.data(), oldNumProp,
+                          pImpl->vertPos_.data(), triProperties.data(),
+                          pImpl->halfedge_.data(), propFunc}));
+  }
+
+  pImpl->meshRelation_.numProp = numProp;
+  pImpl->CreateFaces();
+  pImpl->Finish();
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Curvature is the inverse of the radius of curvature, and signed such that
+ * positive is convex and negative is concave. There are two orthogonal
+ * principal curvatures at any point on a manifold, with one maximum and the
+ * other minimum. Gaussian curvature is their product, while mean
+ * curvature is their sum. This approximates them for every vertex and assigns
+ * them as vertex properties on the given channels.
+ *
+ * @param gaussianIdx The property channel index in which to store the Gaussian
+ * curvature. An index < 0 will be ignored (stores nothing). The property set
+ * will be automatically expanded to include the channel index specified.
+ *
+ * @param meanIdx The property channel index in which to store the mean
+ * curvature. An index < 0 will be ignored (stores nothing). The property set
+ * will be automatically expanded to include the channel index specified.
+ */
+Manifold Manifold::CalculateCurvature(int gaussianIdx, int meanIdx) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  pImpl->CalculateCurvature(gaussianIdx, meanIdx);
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Fills in vertex properties for normal vectors, calculated from the mesh
+ * geometry. Flat faces composed of three or more triangles will remain flat.
+ *
+ * @param normalIdx The property channel in which to store the X
+ * values of the normals. The X, Y, and Z channels will be sequential. The
+ * property set will be automatically expanded such that NumProp will be at
+ * least normalIdx + 3.
+ *
+ * @param minSharpAngle Any edges with angles greater than this value will
+ * remain sharp, getting different normal vector properties on each side of the
+ * edge. By default, no edges are sharp and all normals are shared. With a value
+ * of zero, the model is faceted and all normals match their triangle normals,
+ * but in this case it would be better not to calculate normals at all.
+ */
+Manifold Manifold::CalculateNormals(int normalIdx, float minSharpAngle) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  pImpl->SetNormals(normalIdx, minSharpAngle);
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Smooths out the Manifold by filling in the halfedgeTangent vectors. The
+ * geometry will remain unchanged until Refine or RefineToLength is called to
+ * interpolate the surface. This version uses the supplied vertex normal
+ * properties to define the tangent vectors. Faces of two coplanar triangles
+ * will be marked as quads, while faces with three or more will be flat.
+ *
+ * @param normalIdx The first property channel of the normals. NumProp must be
+ * at least normalIdx + 3. Any vertex where multiple normals exist and don't
+ * agree will result in a sharp edge.
+ */
+Manifold Manifold::SmoothByNormals(int normalIdx) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  if (!IsEmpty()) {
+    pImpl->CreateTangents(normalIdx);
+  }
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Smooths out the Manifold by filling in the halfedgeTangent vectors. The
+ * geometry will remain unchanged until Refine or RefineToLength is called to
+ * interpolate the surface. This version uses the geometry of the triangles and
+ * pseudo-normals to define the tangent vectors. Faces of two coplanar triangles
+ * will be marked as quads.
+ *
+ * @param minSharpAngle degrees, default 60. Any edges with angles greater than
+ * this value will remain sharp. The rest will be smoothed to G1 continuity,
+ * with the caveat that flat faces of three or more triangles will always remain
+ * flat. With a value of zero, the model is faceted, but in this case there is
+ * no point in smoothing.
+ *
+ * @param minSmoothness range: 0 - 1, default 0. The smoothness applied to sharp
+ * angles. The default gives a hard edge, while values > 0 will give a small
+ * fillet on these sharp edges. A value of 1 is equivalent to a minSharpAngle of
+ * 180 - all edges will be smooth.
+ */
+Manifold Manifold::SmoothOut(float minSharpAngle, float minSmoothness) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  if (!IsEmpty()) {
+    pImpl->CreateTangents(pImpl->SharpenEdges(minSharpAngle, minSmoothness));
+  }
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Increase the density of the mesh by splitting every edge into n pieces. For
+ * instance, with n = 2, each triangle will be split into 4 triangles. Quads
+ * will ignore their interior triangle bisector. These will all be coplanar (and
+ * will not be immediately collapsed) unless the Mesh/Manifold has
+ * halfedgeTangents specified (e.g. from the Smooth() constructor), in which
+ * case the new vertices will be moved to the interpolated surface according to
+ * their barycentric coordinates.
+ *
+ * @param n The number of pieces to split every edge into. Must be > 1.
+ */
+Manifold Manifold::Refine(int n) const {
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  if (n > 1) {
+    pImpl->Refine([n](glm::vec3 edge) { return n - 1; });
+  }
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * Increase the density of the mesh by splitting each edge into pieces of
+ * roughly the input length. Interior verts are added to keep the rest of the
+ * triangulation edges also of roughly the same length. If halfedgeTangents are
+ * present (e.g. from the Smooth() constructor), the new vertices will be moved
+ * to the interpolated surface according to their barycentric coordinates. Quads
+ * will ignore their interior triangle bisector.
+ *
+ * @param length The length that edges will be broken down to.
+ */
+Manifold Manifold::RefineToLength(float length) const {
+  length = glm::abs(length);
+  auto pImpl = std::make_shared<Impl>(*GetCsgLeafNode().GetImpl());
+  pImpl->Refine(
+      [length](glm::vec3 edge) { return glm::length(edge) / length; });
+  return Manifold(std::make_shared<CsgLeafNode>(pImpl));
+}
+
+/**
+ * The central operation of this library: the Boolean combines two manifolds
+ * into another by calculating their intersections and removing the unused
+ * portions.
+ * [&epsilon;-valid](https://github.com/elalish/manifold/wiki/Manifold-Library#definition-of-%CE%B5-valid)
+ * inputs will produce &epsilon;-valid output. &epsilon;-invalid input may fail
+ * triangulation.
+ *
+ * These operations are optimized to produce nearly-instant results if either
+ * input is empty or their bounding boxes do not overlap.
+ *
+ * @param second The other Manifold.
+ * @param op The type of operation to perform.
+ */
+Manifold Manifold::Boolean(const Manifold& second, OpType op) const {
+  return Manifold(pNode_->Boolean(second.pNode_, op));
+}
+
+/**
+ * Perform the given boolean operation on a list of Manifolds. In case of
+ * Subtract, all Manifolds in the tail are differenced from the head.
+ */
+Manifold Manifold::BatchBoolean(const std::vector<Manifold>& manifolds,
+                                OpType op) {
+  if (manifolds.size() == 0)
+    return Manifold();
+  else if (manifolds.size() == 1)
+    return manifolds[0];
+  std::vector<std::shared_ptr<CsgNode>> children;
+  children.reserve(manifolds.size());
+  for (const auto& m : manifolds) children.push_back(m.pNode_);
+  return Manifold(std::make_shared<CsgOpNode>(children, op));
+}
+
+/**
+ * Shorthand for Boolean Union.
+ */
+Manifold Manifold::operator+(const Manifold& Q) const {
+  return Boolean(Q, OpType::Add);
+}
+
+/**
+ * Shorthand for Boolean Union assignment.
+ */
+Manifold& Manifold::operator+=(const Manifold& Q) {
+  *this = *this + Q;
+  return *this;
+}
+
+/**
+ * Shorthand for Boolean Difference.
+ */
+Manifold Manifold::operator-(const Manifold& Q) const {
+  return Boolean(Q, OpType::Subtract);
+}
+
+/**
+ * Shorthand for Boolean Difference assignment.
+ */
+Manifold& Manifold::operator-=(const Manifold& Q) {
+  *this = *this - Q;
+  return *this;
+}
+
+/**
+ * Shorthand for Boolean Intersection.
+ */
+Manifold Manifold::operator^(const Manifold& Q) const {
+  return Boolean(Q, OpType::Intersect);
+}
+
+/**
+ * Shorthand for Boolean Intersection assignment.
+ */
+Manifold& Manifold::operator^=(const Manifold& Q) {
+  *this = *this ^ Q;
+  return *this;
+}
+
+/**
+ * Split cuts this manifold in two using the cutter manifold. The first result
+ * is the intersection, second is the difference. This is more efficient than
+ * doing them separately.
+ *
+ * @param cutter
+ */
+std::pair<Manifold, Manifold> Manifold::Split(const Manifold& cutter) const {
+  auto impl1 = GetCsgLeafNode().GetImpl();
+  auto impl2 = cutter.GetCsgLeafNode().GetImpl();
+
+  Boolean3 boolean(*impl1, *impl2, OpType::Subtract);
+  auto result1 = std::make_shared<CsgLeafNode>(
+      std::make_unique<Impl>(boolean.Result(OpType::Intersect)));
+  auto result2 = std::make_shared<CsgLeafNode>(
+      std::make_unique<Impl>(boolean.Result(OpType::Subtract)));
+  return std::make_pair(Manifold(result1), Manifold(result2));
+}
+
+/**
+ * Convenient version of Split() for a half-space.
+ *
+ * @param normal This vector is normal to the cutting plane and its length does
+ * not matter. The first result is in the direction of this vector, the second
+ * result is on the opposite side.
+ * @param originOffset The distance of the plane from the origin in the
+ * direction of the normal vector.
+ */
+std::pair<Manifold, Manifold> Manifold::SplitByPlane(glm::vec3 normal,
+                                                     float originOffset) const {
+  return Split(Halfspace(BoundingBox(), normal, originOffset));
+}
+
+/**
+ * Identical to SplitByPlane(), but calculating and returning only the first
+ * result.
+ *
+ * @param normal This vector is normal to the cutting plane and its length does
+ * not matter. The result is in the direction of this vector from the plane.
+ * @param originOffset The distance of the plane from the origin in the
+ * direction of the normal vector.
+ */
+Manifold Manifold::TrimByPlane(glm::vec3 normal, float originOffset) const {
+  return *this ^ Halfspace(BoundingBox(), normal, originOffset);
+}
+
+/**
+ * Returns the cross section of this object parallel to the X-Y plane at the
+ * specified Z height, defaulting to zero. Using a height equal to the bottom of
+ * the bounding box will return the bottom faces, while using a height equal to
+ * the top of the bounding box will return empty.
+ */
+CrossSection Manifold::Slice(float height) const {
+  return GetCsgLeafNode().GetImpl()->Slice(height);
+}
+
+/**
+ * Returns a cross section representing the projected outline of this object
+ * onto the X-Y plane.
+ */
+CrossSection Manifold::Project() const {
+  return GetCsgLeafNode().GetImpl()->Project();
+}
+
+ExecutionParams& ManifoldParams() { return manifoldParams; }
+
+/**
+ * Compute the convex hull of a set of points. If the given points are fewer
+ * than 4, or they are all coplanar, an empty Manifold will be returned.
+ *
+ * @param pts A vector of 3-dimensional points over which to compute a convex
+ * hull.
+ */
+Manifold Manifold::Hull(const std::vector<glm::vec3>& pts) {
+  ZoneScoped;
+  const int numVert = pts.size();
+  if (numVert < 4) return Manifold();
+
+  std::vector<quickhull::Vector3<double>> vertices(numVert);
+  for (int i = 0; i < numVert; i++) {
+    vertices[i] = {pts[i].x, pts[i].y, pts[i].z};
+  }
+
+  quickhull::QuickHull<double> qh;
+  // bools: correct triangle winding, and use original indices
+  auto hull = qh.getConvexHull(vertices, false, true);
+  const auto& triangles = hull.getIndexBuffer();
+  const int numTris = triangles.size() / 3;
+
+  Mesh mesh;
+  mesh.vertPos = pts;
+  mesh.triVerts.reserve(numTris);
+  for (int i = 0; i < numTris; i++) {
+    const int j = i * 3;
+    mesh.triVerts.push_back({triangles[j], triangles[j + 1], triangles[j + 2]});
+  }
+  return Manifold(mesh);
+}
+
+/**
+ * Compute the convex hull of this manifold.
+ */
+Manifold Manifold::Hull() const { return Hull(GetMesh().vertPos); }
+
+/**
+ * Compute the convex hull enveloping a set of manifolds.
+ *
+ * @param manifolds A vector of manifolds over which to compute a convex hull.
+ */
+Manifold Manifold::Hull(const std::vector<Manifold>& manifolds) {
+  return Compose(manifolds).Hull();
+}
+
+/**
+ * Returns the minimum gap between two manifolds. Returns a float between
+ * 0 and searchLength.
+ *
+ * @param other The other manifold to compute the minimum gap to.
+ * @param searchLength The maximum distance to search for a minimum gap.
+ */
+float Manifold::MinGap(const Manifold& other, float searchLength) const {
+  auto intersect = *this ^ other;
+  auto prop = intersect.GetProperties();
+
+  if (prop.volume != 0) return 0.0f;
+
+  return GetCsgLeafNode().GetImpl()->MinGap(*other.GetCsgLeafNode().GetImpl(),
+                                            searchLength);
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/mesh_fixes.h b/thirdparty/manifold/src/manifold/src/mesh_fixes.h
new file mode 100644
index 000000000000..d8cb4de892d7
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/mesh_fixes.h
@@ -0,0 +1,57 @@
+#include "impl.h"
+
+namespace {
+using namespace manifold;
+
+inline int FlipHalfedge(int halfedge) {
+  const int tri = halfedge / 3;
+  const int vert = 2 - (halfedge - 3 * tri);
+  return 3 * tri + vert;
+}
+
+struct TransformNormals {
+  const glm::mat3 transform;
+
+  glm::vec3 operator()(glm::vec3 normal) {
+    normal = glm::normalize(transform * normal);
+    if (isnan(normal.x)) normal = glm::vec3(0.0f);
+    return normal;
+  }
+};
+
+struct TransformTangents {
+  const glm::mat3 transform;
+  const bool invert;
+  VecView<const glm::vec4> oldTangents;
+  VecView<const Halfedge> halfedge;
+
+  void operator()(thrust::tuple<glm::vec4&, int> inOut) {
+    glm::vec4& tangent = thrust::get<0>(inOut);
+    int edge = thrust::get<1>(inOut);
+    if (invert) {
+      edge = halfedge[FlipHalfedge(edge)].pairedHalfedge;
+    }
+
+    tangent = glm::vec4(transform * glm::vec3(oldTangents[edge]),
+                        oldTangents[edge].w);
+  }
+};
+
+struct FlipTris {
+  VecView<Halfedge> halfedge;
+
+  void operator()(thrust::tuple<TriRef&, int> inOut) {
+    TriRef& bary = thrust::get<0>(inOut);
+    const int tri = thrust::get<1>(inOut);
+
+    thrust::swap(halfedge[3 * tri], halfedge[3 * tri + 2]);
+
+    for (const int i : {0, 1, 2}) {
+      thrust::swap(halfedge[3 * tri + i].startVert,
+                   halfedge[3 * tri + i].endVert);
+      halfedge[3 * tri + i].pairedHalfedge =
+          FlipHalfedge(halfedge[3 * tri + i].pairedHalfedge);
+    }
+  }
+};
+}  // namespace
diff --git a/thirdparty/manifold/src/manifold/src/properties.cpp b/thirdparty/manifold/src/manifold/src/properties.cpp
new file mode 100644
index 000000000000..d85945106b6a
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/properties.cpp
@@ -0,0 +1,418 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <limits>
+
+#include "impl.h"
+#include "par.h"
+
+namespace {
+using namespace manifold;
+
+struct FaceAreaVolume {
+  VecView<const Halfedge> halfedges;
+  VecView<const glm::vec3> vertPos;
+  const float precision;
+
+  thrust::pair<float, float> operator()(int face) {
+    float perimeter = 0;
+    glm::vec3 edge[3];
+    for (int i : {0, 1, 2}) {
+      const int j = (i + 1) % 3;
+      edge[i] = vertPos[halfedges[3 * face + j].startVert] -
+                vertPos[halfedges[3 * face + i].startVert];
+      perimeter += glm::length(edge[i]);
+    }
+    glm::vec3 crossP = glm::cross(edge[0], edge[1]);
+
+    float area = glm::length(crossP);
+    float volume = glm::dot(crossP, vertPos[halfedges[3 * face].startVert]);
+
+    return thrust::make_pair(area / 2.0f, volume / 6.0f);
+  }
+};
+
+struct PosMin
+    : public thrust::binary_function<glm::vec3, glm::vec3, glm::vec3> {
+  glm::vec3 operator()(glm::vec3 a, glm::vec3 b) {
+    if (isnan(a.x)) return b;
+    if (isnan(b.x)) return a;
+    return glm::min(a, b);
+  }
+};
+
+struct PosMax
+    : public thrust::binary_function<glm::vec3, glm::vec3, glm::vec3> {
+  glm::vec3 operator()(glm::vec3 a, glm::vec3 b) {
+    if (isnan(a.x)) return b;
+    if (isnan(b.x)) return a;
+    return glm::max(a, b);
+  }
+};
+
+struct FiniteVert {
+  bool operator()(glm::vec3 v) { return glm::all(glm::isfinite(v)); }
+};
+
+struct MakeMinMax {
+  glm::ivec2 operator()(glm::ivec3 tri) {
+    return glm::ivec2(glm::min(tri[0], glm::min(tri[1], tri[2])),
+                      glm::max(tri[0], glm::max(tri[1], tri[2])));
+  }
+};
+
+struct MinMax
+    : public thrust::binary_function<glm::ivec2, glm::ivec2, glm::ivec2> {
+  glm::ivec2 operator()(glm::ivec2 a, glm::ivec2 b) {
+    a[0] = glm::min(a[0], b[0]);
+    a[1] = glm::max(a[1], b[1]);
+    return a;
+  }
+};
+
+struct SumPair : public thrust::binary_function<thrust::pair<float, float>,
+                                                thrust::pair<float, float>,
+                                                thrust::pair<float, float>> {
+  thrust::pair<float, float> operator()(thrust::pair<float, float> a,
+                                        thrust::pair<float, float> b) {
+    a.first += b.first;
+    a.second += b.second;
+    return a;
+  }
+};
+
+struct CurvatureAngles {
+  VecView<float> meanCurvature;
+  VecView<float> gaussianCurvature;
+  VecView<float> area;
+  VecView<float> degree;
+  VecView<const Halfedge> halfedge;
+  VecView<const glm::vec3> vertPos;
+  VecView<const glm::vec3> triNormal;
+
+  void operator()(int tri) {
+    glm::vec3 edge[3];
+    glm::vec3 edgeLength(0.0);
+    for (int i : {0, 1, 2}) {
+      const int startVert = halfedge[3 * tri + i].startVert;
+      const int endVert = halfedge[3 * tri + i].endVert;
+      edge[i] = vertPos[endVert] - vertPos[startVert];
+      edgeLength[i] = glm::length(edge[i]);
+      edge[i] /= edgeLength[i];
+      const int neighborTri = halfedge[3 * tri + i].pairedHalfedge / 3;
+      const float dihedral =
+          0.25 * edgeLength[i] *
+          glm::asin(glm::dot(glm::cross(triNormal[tri], triNormal[neighborTri]),
+                             edge[i]));
+      AtomicAdd(meanCurvature[startVert], dihedral);
+      AtomicAdd(meanCurvature[endVert], dihedral);
+      AtomicAdd(degree[startVert], 1.0f);
+    }
+
+    glm::vec3 phi;
+    phi[0] = glm::acos(-glm::dot(edge[2], edge[0]));
+    phi[1] = glm::acos(-glm::dot(edge[0], edge[1]));
+    phi[2] = glm::pi<float>() - phi[0] - phi[1];
+    const float area3 = edgeLength[0] * edgeLength[1] *
+                        glm::length(glm::cross(edge[0], edge[1])) / 6;
+
+    for (int i : {0, 1, 2}) {
+      const int vert = halfedge[3 * tri + i].startVert;
+      AtomicAdd(gaussianCurvature[vert], -phi[i]);
+      AtomicAdd(area[vert], area3);
+    }
+  }
+};
+
+struct NormalizeCurvature {
+  void operator()(thrust::tuple<float&, float&, float, float> inOut) {
+    float& meanCurvature = thrust::get<0>(inOut);
+    float& gaussianCurvature = thrust::get<1>(inOut);
+    float area = thrust::get<2>(inOut);
+    float degree = thrust::get<3>(inOut);
+    float factor = degree / (6 * area);
+    meanCurvature *= factor;
+    gaussianCurvature *= factor;
+  }
+};
+
+struct UpdateProperties {
+  VecView<float> properties;
+
+  VecView<const float> oldProperties;
+  VecView<const Halfedge> halfedge;
+  VecView<const float> meanCurvature;
+  VecView<const float> gaussianCurvature;
+  const int oldNumProp;
+  const int numProp;
+  const int gaussianIdx;
+  const int meanIdx;
+
+  // FIXME: race condition
+  void operator()(thrust::tuple<glm::ivec3&, int> inOut) {
+    glm::ivec3& triProp = thrust::get<0>(inOut);
+    const int tri = thrust::get<1>(inOut);
+
+    for (const int i : {0, 1, 2}) {
+      const int vert = halfedge[3 * tri + i].startVert;
+      if (oldNumProp == 0) {
+        triProp[i] = vert;
+      }
+      const int propVert = triProp[i];
+
+      for (int p = 0; p < oldNumProp; ++p) {
+        properties[numProp * propVert + p] =
+            oldProperties[oldNumProp * propVert + p];
+      }
+
+      if (gaussianIdx >= 0) {
+        properties[numProp * propVert + gaussianIdx] = gaussianCurvature[vert];
+      }
+      if (meanIdx >= 0) {
+        properties[numProp * propVert + meanIdx] = meanCurvature[vert];
+      }
+    }
+  }
+};
+
+struct CheckHalfedges {
+  VecView<const Halfedge> halfedges;
+  VecView<const glm::vec3> vertPos;
+
+  bool operator()(size_t edge) {
+    const Halfedge halfedge = halfedges[edge];
+    if (halfedge.startVert == -1 || halfedge.endVert == -1) return true;
+    if (halfedge.pairedHalfedge == -1) return false;
+
+    if (!isfinite(vertPos[halfedge.startVert][0])) return false;
+    if (!isfinite(vertPos[halfedge.endVert][0])) return false;
+
+    const Halfedge paired = halfedges[halfedge.pairedHalfedge];
+    bool good = true;
+    good &= paired.pairedHalfedge == edge;
+    good &= halfedge.startVert != halfedge.endVert;
+    good &= halfedge.startVert == paired.endVert;
+    good &= halfedge.endVert == paired.startVert;
+    return good;
+  }
+};
+
+struct NoDuplicates {
+  VecView<const Halfedge> halfedges;
+
+  bool operator()(int edge) {
+    const Halfedge halfedge = halfedges[edge];
+    if (halfedge.startVert == -1 && halfedge.endVert == -1 &&
+        halfedge.pairedHalfedge == -1)
+      return true;
+    return halfedge.startVert != halfedges[edge + 1].startVert ||
+           halfedge.endVert != halfedges[edge + 1].endVert;
+  }
+};
+
+struct CheckCCW {
+  VecView<const Halfedge> halfedges;
+  VecView<const glm::vec3> vertPos;
+  VecView<const glm::vec3> triNormal;
+  const float tol;
+
+  bool operator()(int face) {
+    if (halfedges[3 * face].pairedHalfedge < 0) return true;
+
+    const glm::mat3x2 projection = GetAxisAlignedProjection(triNormal[face]);
+    glm::vec2 v[3];
+    for (int i : {0, 1, 2})
+      v[i] = projection * vertPos[halfedges[3 * face + i].startVert];
+
+    int ccw = CCW(v[0], v[1], v[2], glm::abs(tol));
+    bool check = tol > 0 ? ccw >= 0 : ccw == 0;
+
+#ifdef MANIFOLD_DEBUG
+    if (tol > 0 && !check) {
+      glm::vec2 v1 = v[1] - v[0];
+      glm::vec2 v2 = v[2] - v[0];
+      float area = v1.x * v2.y - v1.y * v2.x;
+      float base2 = glm::max(glm::dot(v1, v1), glm::dot(v2, v2));
+      float base = glm::sqrt(base2);
+      glm::vec3 V0 = vertPos[halfedges[3 * face].startVert];
+      glm::vec3 V1 = vertPos[halfedges[3 * face + 1].startVert];
+      glm::vec3 V2 = vertPos[halfedges[3 * face + 2].startVert];
+      glm::vec3 norm = glm::cross(V1 - V0, V2 - V0);
+      printf(
+          "Tri %d does not match normal, approx height = %g, base = %g\n"
+          "tol = %g, area2 = %g, base2*tol2 = %g\n"
+          "normal = %g, %g, %g\n"
+          "norm = %g, %g, %g\nverts: %d, %d, %d\n",
+          face, area / base, base, tol, area * area, base2 * tol * tol,
+          triNormal[face].x, triNormal[face].y, triNormal[face].z, norm.x,
+          norm.y, norm.z, halfedges[3 * face].startVert,
+          halfedges[3 * face + 1].startVert, halfedges[3 * face + 2].startVert);
+    }
+#endif
+    return check;
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Returns true if this manifold is in fact an oriented even manifold and all of
+ * the data structures are consistent.
+ */
+bool Manifold::Impl::IsManifold() const {
+  if (halfedge_.size() == 0) return true;
+  auto policy = autoPolicy(halfedge_.size());
+
+  return all_of(policy, countAt(0_z), countAt(halfedge_.size()),
+                CheckHalfedges({halfedge_, vertPos_}));
+}
+
+/**
+ * Returns true if this manifold is in fact an oriented 2-manifold and all of
+ * the data structures are consistent.
+ */
+bool Manifold::Impl::Is2Manifold() const {
+  if (halfedge_.size() == 0) return true;
+  auto policy = autoPolicy(halfedge_.size());
+
+  if (!IsManifold()) return false;
+
+  Vec<Halfedge> halfedge(halfedge_);
+  stable_sort(policy, halfedge.begin(), halfedge.end());
+
+  return all_of(policy, countAt(0), countAt(2 * NumEdge() - 1),
+                NoDuplicates({halfedge}));
+}
+
+/**
+ * Returns true if all triangles are CCW relative to their triNormals_.
+ */
+bool Manifold::Impl::MatchesTriNormals() const {
+  if (halfedge_.size() == 0 || faceNormal_.size() != NumTri()) return true;
+  return all_of(autoPolicy(NumTri()), countAt(0), countAt(NumTri()),
+                CheckCCW({halfedge_, vertPos_, faceNormal_, 2 * precision_}));
+}
+
+/**
+ * Returns the number of triangles that are colinear within precision_.
+ */
+int Manifold::Impl::NumDegenerateTris() const {
+  if (halfedge_.size() == 0 || faceNormal_.size() != NumTri()) return true;
+  return count_if(
+      autoPolicy(NumTri()), countAt(0), countAt(NumTri()),
+      CheckCCW({halfedge_, vertPos_, faceNormal_, -1 * precision_ / 2}));
+}
+
+Properties Manifold::Impl::GetProperties() const {
+  ZoneScoped;
+  if (IsEmpty()) return {0, 0};
+  // Kahan summation
+  float area = 0;
+  float volume = 0;
+  float areaCompensation = 0;
+  float volumeCompensation = 0;
+  for (int i = 0; i < NumTri(); ++i) {
+    auto [area1, volume1] =
+        FaceAreaVolume({halfedge_, vertPos_, precision_})(i);
+    const float t1 = area + area1;
+    const float t2 = volume + volume1;
+    areaCompensation += (area - t1) + area1;
+    volumeCompensation += (volume - t2) + volume1;
+    area = t1;
+    volume = t2;
+  }
+  area += areaCompensation;
+  volume += volumeCompensation;
+
+  return {area, volume};
+}
+
+void Manifold::Impl::CalculateCurvature(int gaussianIdx, int meanIdx) {
+  ZoneScoped;
+  if (IsEmpty()) return;
+  if (gaussianIdx < 0 && meanIdx < 0) return;
+  Vec<float> vertMeanCurvature(NumVert(), 0);
+  Vec<float> vertGaussianCurvature(NumVert(), glm::two_pi<float>());
+  Vec<float> vertArea(NumVert(), 0);
+  Vec<float> degree(NumVert(), 0);
+  auto policy = autoPolicy(NumTri());
+  for_each(policy, countAt(0), countAt(NumTri()),
+           CurvatureAngles({vertMeanCurvature, vertGaussianCurvature, vertArea,
+                            degree, halfedge_, vertPos_, faceNormal_}));
+  for_each_n(policy,
+             zip(vertMeanCurvature.begin(), vertGaussianCurvature.begin(),
+                 vertArea.begin(), degree.begin()),
+             NumVert(), NormalizeCurvature());
+
+  const int oldNumProp = NumProp();
+  const int numProp = glm::max(oldNumProp, glm::max(gaussianIdx, meanIdx) + 1);
+  const Vec<float> oldProperties = meshRelation_.properties;
+  meshRelation_.properties = Vec<float>(numProp * NumPropVert(), 0);
+  meshRelation_.numProp = numProp;
+  if (meshRelation_.triProperties.size() == 0) {
+    meshRelation_.triProperties.resize(NumTri());
+  }
+
+  for_each_n(
+      policy, zip(meshRelation_.triProperties.begin(), countAt(0)), NumTri(),
+      UpdateProperties({meshRelation_.properties, oldProperties, halfedge_,
+                        vertMeanCurvature, vertGaussianCurvature, oldNumProp,
+                        numProp, gaussianIdx, meanIdx}));
+
+  CreateFaces();
+  Finish();
+}
+
+/**
+ * Calculates the bounding box of the entire manifold, which is stored
+ * internally to short-cut Boolean operations and to serve as the precision
+ * range for Morton code calculation. Ignores NaNs.
+ */
+void Manifold::Impl::CalculateBBox() {
+  auto policy = autoPolicy(NumVert());
+  bBox_.min = reduce<glm::vec3>(
+      policy, vertPos_.begin(), vertPos_.end(),
+      glm::vec3(std::numeric_limits<float>::infinity()), PosMin());
+  bBox_.max = reduce<glm::vec3>(
+      policy, vertPos_.begin(), vertPos_.end(),
+      glm::vec3(-std::numeric_limits<float>::infinity()), PosMax());
+}
+
+/**
+ * Determines if all verts are finite. Checking just the bounding box dimensions
+ * is insufficient as it ignores NaNs.
+ */
+bool Manifold::Impl::IsFinite() const {
+  auto policy = autoPolicy(NumVert());
+  return transform_reduce<bool>(policy, vertPos_.begin(), vertPos_.end(),
+                                FiniteVert(), true,
+                                thrust::logical_and<bool>());
+}
+
+/**
+ * Checks that the input triVerts array has all indices inside bounds of the
+ * vertPos_ array.
+ */
+bool Manifold::Impl::IsIndexInBounds(VecView<const glm::ivec3> triVerts) const {
+  auto policy = autoPolicy(triVerts.size());
+  glm::ivec2 minmax = transform_reduce<glm::ivec2>(
+      policy, triVerts.begin(), triVerts.end(), MakeMinMax(),
+      glm::ivec2(std::numeric_limits<int>::max(),
+                 std::numeric_limits<int>::min()),
+      MinMax());
+
+  return minmax[0] >= 0 && minmax[1] < NumVert();
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/shared.h b/thirdparty/manifold/src/manifold/src/shared.h
new file mode 100644
index 000000000000..ed457dad5e32
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/shared.h
@@ -0,0 +1,232 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "par.h"
+#include "sparse.h"
+#include "utils.h"
+#include "vec.h"
+
+namespace manifold {
+
+/** @addtogroup Private
+ *  @{
+ */
+inline glm::vec3 SafeNormalize(glm::vec3 v) {
+  v = glm::normalize(v);
+  return glm::isfinite(v.x) ? v : glm::vec3(0);
+}
+
+inline float MaxPrecision(float minPrecision, const Box& bBox) {
+  float precision = glm::max(minPrecision, kTolerance * bBox.Scale());
+  return glm::isfinite(precision) ? precision : -1;
+}
+
+inline int NextHalfedge(int current) {
+  ++current;
+  if (current % 3 == 0) current -= 3;
+  return current;
+}
+
+inline glm::mat3 NormalTransform(const glm::mat4x3& transform) {
+  return glm::inverse(glm::transpose(glm::mat3(transform)));
+}
+
+/**
+ * By using the closest axis-aligned projection to the normal instead of a
+ * projection along the normal, we avoid introducing any rounding error.
+ */
+inline glm::mat3x2 GetAxisAlignedProjection(glm::vec3 normal) {
+  glm::vec3 absNormal = glm::abs(normal);
+  float xyzMax;
+  glm::mat2x3 projection;
+  if (absNormal.z > absNormal.x && absNormal.z > absNormal.y) {
+    projection = glm::mat2x3(1.0f, 0.0f, 0.0f,  //
+                             0.0f, 1.0f, 0.0f);
+    xyzMax = normal.z;
+  } else if (absNormal.y > absNormal.x) {
+    projection = glm::mat2x3(0.0f, 0.0f, 1.0f,  //
+                             1.0f, 0.0f, 0.0f);
+    xyzMax = normal.y;
+  } else {
+    projection = glm::mat2x3(0.0f, 1.0f, 0.0f,  //
+                             0.0f, 0.0f, 1.0f);
+    xyzMax = normal.x;
+  }
+  if (xyzMax < 0) projection[0] *= -1.0f;
+  return glm::transpose(projection);
+}
+
+inline glm::vec3 GetBarycentric(const glm::vec3& v, const glm::mat3& triPos,
+                                float precision) {
+  const glm::mat3 edges(triPos[2] - triPos[1], triPos[0] - triPos[2],
+                        triPos[1] - triPos[0]);
+  const glm::vec3 d2(glm::dot(edges[0], edges[0]), glm::dot(edges[1], edges[1]),
+                     glm::dot(edges[2], edges[2]));
+  const int longSide = d2[0] > d2[1] && d2[0] > d2[2] ? 0
+                       : d2[1] > d2[2]                ? 1
+                                                      : 2;
+  const glm::vec3 crossP = glm::cross(edges[0], edges[1]);
+  const float area2 = glm::dot(crossP, crossP);
+  const float tol2 = precision * precision;
+
+  glm::vec3 uvw(0);
+  for (const int i : {0, 1, 2}) {
+    const glm::vec3 dv = v - triPos[i];
+    if (glm::dot(dv, dv) < tol2) {
+      // Return exactly equal if within tolerance of vert.
+      uvw[i] = 1;
+      return uvw;
+    }
+  }
+
+  if (d2[longSide] < tol2) {  // point
+    return glm::vec3(1, 0, 0);
+  } else if (area2 > d2[longSide] * tol2) {  // triangle
+    for (const int i : {0, 1, 2}) {
+      const int j = Next3(i);
+      const glm::vec3 crossPv = glm::cross(edges[i], v - triPos[j]);
+      const float area2v = glm::dot(crossPv, crossPv);
+      // Return exactly equal if within tolerance of edge.
+      uvw[i] = area2v < d2[i] * tol2 ? 0 : glm::dot(crossPv, crossP);
+    }
+    uvw /= (uvw[0] + uvw[1] + uvw[2]);
+    return uvw;
+  } else {  // line
+    const int nextV = Next3(longSide);
+    const float alpha =
+        glm::dot(v - triPos[nextV], edges[longSide]) / d2[longSide];
+    uvw[longSide] = 0;
+    uvw[nextV] = 1 - alpha;
+    const int lastV = Next3(nextV);
+    uvw[lastV] = alpha;
+    return uvw;
+  }
+}
+
+/**
+ * The fundamental component of the halfedge data structure used for storing and
+ * operating on the Manifold.
+ */
+struct Halfedge {
+  int startVert, endVert;
+  int pairedHalfedge;
+  int face;
+  bool IsForward() const { return startVert < endVert; }
+  bool operator<(const Halfedge& other) const {
+    return startVert == other.startVert ? endVert < other.endVert
+                                        : startVert < other.startVert;
+  }
+};
+
+struct Barycentric {
+  int tri;
+  glm::vec4 uvw;
+};
+
+struct TriRef {
+  /// The unique ID of the mesh instance of this triangle. If .meshID and .tri
+  /// match for two triangles, then they are coplanar and came from the same
+  /// face.
+  int meshID;
+  /// The OriginalID of the mesh this triangle came from. This ID is ideal for
+  /// reapplying properties like UV coordinates to the output mesh.
+  int originalID;
+  /// The triangle index of the original triangle this was part of:
+  /// Mesh.triVerts[tri].
+  int tri;
+
+  bool SameFace(const TriRef& other) const {
+    return meshID == other.meshID && tri == other.tri;
+  }
+};
+
+/**
+ * This is a temporary edge structure which only stores edges forward and
+ * references the halfedge it was created from.
+ */
+struct TmpEdge {
+  int first, second, halfedgeIdx;
+
+  TmpEdge() {}
+  TmpEdge(int start, int end, int idx) {
+    first = glm::min(start, end);
+    second = glm::max(start, end);
+    halfedgeIdx = idx;
+  }
+
+  bool operator<(const TmpEdge& other) const {
+    return first == other.first ? second < other.second : first < other.first;
+  }
+};
+/** @} */
+
+struct Halfedge2Tmp {
+  void operator()(thrust::tuple<TmpEdge&, const Halfedge&, int> inout) {
+    const Halfedge& halfedge = thrust::get<1>(inout);
+    int idx = thrust::get<2>(inout);
+    if (!halfedge.IsForward()) idx = -1;
+
+    thrust::get<0>(inout) = TmpEdge(halfedge.startVert, halfedge.endVert, idx);
+  }
+};
+
+struct TmpInvalid {
+  bool operator()(const TmpEdge& edge) { return edge.halfedgeIdx < 0; }
+};
+
+Vec<TmpEdge> inline CreateTmpEdges(const Vec<Halfedge>& halfedge) {
+  Vec<TmpEdge> edges(halfedge.size());
+  for_each_n(autoPolicy(edges.size()),
+             zip(edges.begin(), halfedge.begin(), countAt(0)), edges.size(),
+             Halfedge2Tmp());
+  int numEdge =
+      remove_if<decltype(edges.begin())>(
+          autoPolicy(edges.size()), edges.begin(), edges.end(), TmpInvalid()) -
+      edges.begin();
+  ASSERT(numEdge == halfedge.size() / 2, topologyErr, "Not oriented!");
+  edges.resize(numEdge);
+  return edges;
+}
+
+template <const bool inverted>
+struct ReindexEdge {
+  VecView<const TmpEdge> edges;
+  SparseIndices& indices;
+
+  void operator()(size_t i) {
+    int& edge = indices.Get(i, inverted);
+    edge = edges[edge].halfedgeIdx;
+  }
+};
+
+#ifdef MANIFOLD_DEBUG
+inline std::ostream& operator<<(std::ostream& stream, const Halfedge& edge) {
+  return stream << "startVert = " << edge.startVert
+                << ", endVert = " << edge.endVert
+                << ", pairedHalfedge = " << edge.pairedHalfedge
+                << ", face = " << edge.face;
+}
+
+inline std::ostream& operator<<(std::ostream& stream, const Barycentric& bary) {
+  return stream << "tri = " << bary.tri << ", uvw = " << bary.uvw;
+}
+
+inline std::ostream& operator<<(std::ostream& stream, const TriRef& ref) {
+  return stream << "meshID: " << ref.meshID
+                << ", originalID: " << ref.originalID << ", tri: " << ref.tri;
+}
+#endif
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/smoothing.cpp b/thirdparty/manifold/src/manifold/src/smoothing.cpp
new file mode 100644
index 000000000000..294f2cdf5a74
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/smoothing.cpp
@@ -0,0 +1,847 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <glm/gtx/quaternion.hpp>
+
+#include "impl.h"
+#include "par.h"
+
+namespace {
+using namespace manifold;
+
+glm::vec3 OrthogonalTo(glm::vec3 in, glm::vec3 ref) {
+  in -= glm::dot(in, ref) * ref;
+  return in;
+}
+
+// Calculate a tangent vector in the form of a weighted cubic Bezier taking as
+// input the desired tangent direction (length doesn't matter) and the edge
+// vector to the neighboring vertex. In a symmetric situation where the tangents
+// at each end are mirror images of each other, this will result in a circular
+// arc.
+glm::vec4 CircularTangent(const glm::vec3& tangent, const glm::vec3& edgeVec) {
+  const glm::vec3 dir = SafeNormalize(tangent);
+
+  float weight = glm::abs(glm::dot(dir, SafeNormalize(edgeVec)));
+  if (weight == 0) {
+    weight = 1;
+  }
+  // Quadratic weighted bezier for circular interpolation
+  const glm::vec4 bz2 =
+      weight * glm::vec4(dir * glm::length(edgeVec) / (2 * weight), 1);
+  // Equivalent cubic weighted bezier
+  const glm::vec4 bz3 = glm::mix(glm::vec4(0, 0, 0, 1), bz2, 2 / 3.0f);
+  // Convert from homogeneous form to geometric form
+  return glm::vec4(glm::vec3(bz3) / bz3.w, bz3.w);
+}
+
+struct SmoothBezier {
+  const Manifold::Impl* impl;
+  VecView<const glm::vec3> vertNormal;
+
+  void operator()(thrust::tuple<glm::vec4&, Halfedge, int> inOut) {
+    glm::vec4& tangent = thrust::get<0>(inOut);
+    const Halfedge edge = thrust::get<1>(inOut);
+    const int edgeIdx = thrust::get<2>(inOut);
+
+    if (impl->IsInsideQuad(edgeIdx)) {
+      tangent = glm::vec4(0, 0, 0, -1);
+      return;
+    }
+
+    const glm::vec3 edgeVec =
+        impl->vertPos_[edge.endVert] - impl->vertPos_[edge.startVert];
+    const glm::vec3 edgeNormal =
+        impl->faceNormal_[edge.face] +
+        impl->faceNormal_[impl->halfedge_[edge.pairedHalfedge].face];
+    glm::vec3 dir =
+        glm::cross(glm::cross(edgeNormal, edgeVec), vertNormal[edge.startVert]);
+    tangent = CircularTangent(dir, edgeVec);
+  }
+};
+
+struct InterpTri {
+  const Manifold::Impl* impl;
+
+  glm::vec4 Homogeneous(glm::vec4 v) const {
+    v.x *= v.w;
+    v.y *= v.w;
+    v.z *= v.w;
+    return v;
+  }
+
+  glm::vec4 Homogeneous(glm::vec3 v) const { return glm::vec4(v, 1.0f); }
+
+  glm::vec3 HNormalize(glm::vec4 v) const {
+    return v.w == 0 ? v : (glm::vec3(v) / v.w);
+  }
+
+  glm::vec4 Scale(glm::vec4 v, float scale) const {
+    return glm::vec4(scale * glm::vec3(v), v.w);
+  }
+
+  glm::vec4 Bezier(glm::vec3 point, glm::vec4 tangent) const {
+    return Homogeneous(glm::vec4(point, 0) + tangent);
+  }
+
+  glm::mat2x4 CubicBezier2Linear(glm::vec4 p0, glm::vec4 p1, glm::vec4 p2,
+                                 glm::vec4 p3, float x) const {
+    glm::mat2x4 out;
+    glm::vec4 p12 = glm::mix(p1, p2, x);
+    out[0] = glm::mix(glm::mix(p0, p1, x), p12, x);
+    out[1] = glm::mix(p12, glm::mix(p2, p3, x), x);
+    return out;
+  }
+
+  glm::vec3 BezierPoint(glm::mat2x4 points, float x) const {
+    return HNormalize(glm::mix(points[0], points[1], x));
+  }
+
+  glm::vec3 BezierTangent(glm::mat2x4 points) const {
+    return SafeNormalize(HNormalize(points[1]) - HNormalize(points[0]));
+  }
+
+  glm::vec3 RotateFromTo(glm::vec3 v, glm::quat start, glm::quat end) const {
+    return end * glm::conjugate(start) * v;
+  }
+
+  glm::mat2x4 Bezier2Bezier(const glm::mat2x3& corners,
+                            const glm::mat2x4& tangentsX,
+                            const glm::mat2x4& tangentsY, float x,
+                            const glm::bvec2& pointedEnds,
+                            const glm::vec3& anchor) const {
+    const glm::mat2x4 bez = CubicBezier2Linear(
+        Homogeneous(corners[0]), Bezier(corners[0], tangentsX[0]),
+        Bezier(corners[1], tangentsX[1]), Homogeneous(corners[1]), x);
+    const glm::vec3 end = BezierPoint(bez, x);
+    const glm::vec3 tangent = BezierTangent(bez);
+
+    const glm::mat2x3 nTangentsX(SafeNormalize(glm::vec3(tangentsX[0])),
+                                 -SafeNormalize(glm::vec3(tangentsX[1])));
+    const glm::mat2x3 biTangents = {
+        SafeNormalize(OrthogonalTo(
+            glm::vec3(tangentsY[0]) + kTolerance * (anchor - corners[0]),
+            nTangentsX[0])),
+        SafeNormalize(OrthogonalTo(
+            glm::vec3(tangentsY[1]) + kTolerance * (anchor - corners[1]),
+            nTangentsX[1]))};
+
+    const glm::quat q0 =
+        glm::quat_cast(glm::mat3(nTangentsX[0], biTangents[0],
+                                 glm::cross(nTangentsX[0], biTangents[0])));
+    const glm::quat q1 =
+        glm::quat_cast(glm::mat3(nTangentsX[1], biTangents[1],
+                                 glm::cross(nTangentsX[1], biTangents[1])));
+    const glm::quat qTmp = glm::slerp(q0, q1, x);
+    const glm::quat q =
+        glm::rotation(qTmp * glm::vec3(1, 0, 0), tangent) * qTmp;
+
+    const glm::vec3 end0 = pointedEnds[0]
+                               ? glm::vec3(0)
+                               : RotateFromTo(glm::vec3(tangentsY[0]), q0, q);
+    const glm::vec3 end1 = pointedEnds[1]
+                               ? glm::vec3(0)
+                               : RotateFromTo(glm::vec3(tangentsY[1]), q1, q);
+    const glm::vec3 delta = glm::mix(end0, end1, x);
+    const float deltaW = glm::mix(tangentsY[0].w, tangentsY[1].w, x);
+
+    return {Homogeneous(end), glm::vec4(delta, deltaW)};
+  }
+
+  glm::vec3 Bezier2D(const glm::mat4x3& corners, const glm::mat4& tangentsX,
+                     const glm::mat4& tangentsY, float x, float y,
+                     const glm::vec3& centroid, bool isTriangle) const {
+    glm::mat2x4 bez0 = Bezier2Bezier(
+        {corners[0], corners[1]}, {tangentsX[0], tangentsX[1]},
+        {tangentsY[0], tangentsY[1]}, x, {isTriangle, false}, centroid);
+    glm::mat2x4 bez1 = Bezier2Bezier(
+        {corners[2], corners[3]}, {tangentsX[2], tangentsX[3]},
+        {tangentsY[2], tangentsY[3]}, 1 - x, {false, isTriangle}, centroid);
+
+    const float flatLen =
+        isTriangle ? x * glm::length(corners[1] - corners[2])
+                   : glm::length(glm::mix(corners[0], corners[1], x) -
+                                 glm::mix(corners[3], corners[2], x));
+    const float scale = glm::length(glm::vec3(bez0[0] - bez1[0])) / flatLen;
+
+    const glm::mat2x4 bez = CubicBezier2Linear(
+        bez0[0], Bezier(glm::vec3(bez0[0]), Scale(bez0[1], scale)),
+        Bezier(glm::vec3(bez1[0]), Scale(bez1[1], scale)), bez1[0], y);
+    return BezierPoint(bez, y);
+  }
+
+  void operator()(thrust::tuple<glm::vec3&, Barycentric> inOut) {
+    glm::vec3& pos = thrust::get<0>(inOut);
+    const int tri = thrust::get<1>(inOut).tri;
+    const glm::vec4 uvw = thrust::get<1>(inOut).uvw;
+
+    const glm::ivec4 halfedges = impl->GetHalfedges(tri);
+    const glm::mat4x3 corners = {
+        impl->vertPos_[impl->halfedge_[halfedges[0]].startVert],
+        impl->vertPos_[impl->halfedge_[halfedges[1]].startVert],
+        impl->vertPos_[impl->halfedge_[halfedges[2]].startVert],
+        halfedges[3] < 0
+            ? glm::vec3(0)
+            : impl->vertPos_[impl->halfedge_[halfedges[3]].startVert]};
+
+    for (const int i : {0, 1, 2, 3}) {
+      if (uvw[i] == 1) {
+        pos = corners[i];
+        return;
+      }
+    }
+
+    glm::vec4 posH(0);
+
+    if (halfedges[3] < 0) {  // tri
+      const glm::mat3x4 tangentR = {impl->halfedgeTangent_[halfedges[0]],
+                                    impl->halfedgeTangent_[halfedges[1]],
+                                    impl->halfedgeTangent_[halfedges[2]]};
+      const glm::mat3x4 tangentL = {
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[2]].pairedHalfedge],
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[0]].pairedHalfedge],
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[1]].pairedHalfedge]};
+      const glm::vec3 centroid = glm::mat3(corners) * glm::vec3(1.0f / 3);
+
+      for (const int i : {0, 1, 2}) {
+        const int j = Next3(i);
+        const int k = Prev3(i);
+        const float x = uvw[k] / (1 - uvw[i]);
+        const glm::vec3 p =
+            Bezier2D({corners[i], corners[j], corners[k], corners[i]},
+                     {tangentR[i], tangentL[j], tangentR[k], tangentL[i]},
+                     {tangentL[i], tangentR[j], tangentL[k], tangentR[i]},
+                     1 - uvw[i], x, centroid, true);
+        posH += Homogeneous(glm::vec4(p, uvw[i]));
+      }
+    } else {  // quad
+      const glm::mat4 tangentsX = {
+          impl->halfedgeTangent_[halfedges[0]],
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[0]].pairedHalfedge],
+          impl->halfedgeTangent_[halfedges[2]],
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[2]].pairedHalfedge]};
+      const glm::mat4 tangentsY = {
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[3]].pairedHalfedge],
+          impl->halfedgeTangent_[halfedges[1]],
+          impl->halfedgeTangent_[impl->halfedge_[halfedges[1]].pairedHalfedge],
+          impl->halfedgeTangent_[halfedges[3]]};
+      const glm::vec3 centroid = corners * glm::vec4(0.25);
+      const float x = uvw[1] + uvw[2];
+      const float y = uvw[2] + uvw[3];
+      const glm::vec3 pX =
+          Bezier2D(corners, tangentsX, tangentsY, x, y, centroid, false);
+      const glm::vec3 pY =
+          Bezier2D({corners[1], corners[2], corners[3], corners[0]},
+                   {tangentsY[1], tangentsY[2], tangentsY[3], tangentsY[0]},
+                   {tangentsX[1], tangentsX[2], tangentsX[3], tangentsX[0]}, y,
+                   1 - x, centroid, false);
+      posH += Homogeneous(glm::vec4(pX, x * (1 - x)));
+      posH += Homogeneous(glm::vec4(pY, y * (1 - y)));
+    }
+    pos = HNormalize(posH);
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Get the property normal associated with the startVert of this halfedge, where
+ * normalIdx shows the beginning of where normals are stored in the properties.
+ */
+glm::vec3 Manifold::Impl::GetNormal(int halfedge, int normalIdx) const {
+  const int tri = halfedge / 3;
+  const int j = halfedge % 3;
+  const int prop = meshRelation_.triProperties[tri][j];
+  glm::vec3 normal;
+  for (const int i : {0, 1, 2}) {
+    normal[i] =
+        meshRelation_.properties[prop * meshRelation_.numProp + normalIdx + i];
+  }
+  return normal;
+}
+
+/**
+ * Returns true if this halfedge should be marked as the interior of a quad, as
+ * defined by its two triangles referring to the same face, and those triangles
+ * having no further face neighbors beyond.
+ */
+bool Manifold::Impl::IsInsideQuad(int halfedge) const {
+  if (halfedgeTangent_.size() > 0) {
+    return halfedgeTangent_[halfedge].w < 0;
+  }
+  const int tri = halfedge_[halfedge].face;
+  const TriRef ref = meshRelation_.triRef[tri];
+  const int pair = halfedge_[halfedge].pairedHalfedge;
+  const int pairTri = halfedge_[pair].face;
+  const TriRef pairRef = meshRelation_.triRef[pairTri];
+  if (!ref.SameFace(pairRef)) return false;
+
+  auto SameFace = [this](int halfedge, const TriRef& ref) {
+    return ref.SameFace(
+        meshRelation_.triRef[halfedge_[halfedge].pairedHalfedge / 3]);
+  };
+
+  int neighbor = NextHalfedge(halfedge);
+  if (SameFace(neighbor, ref)) return false;
+  neighbor = NextHalfedge(neighbor);
+  if (SameFace(neighbor, ref)) return false;
+  neighbor = NextHalfedge(pair);
+  if (SameFace(neighbor, pairRef)) return false;
+  neighbor = NextHalfedge(neighbor);
+  if (SameFace(neighbor, pairRef)) return false;
+  return true;
+}
+
+/**
+ * Returns true if this halfedge is an interior of a quad, as defined by its
+ * halfedge tangent having negative weight.
+ */
+bool Manifold::Impl::IsMarkedInsideQuad(int halfedge) const {
+  return halfedgeTangent_.size() > 0 && halfedgeTangent_[halfedge].w < 0;
+}
+
+// sharpenedEdges are referenced to the input Mesh, but the triangles have
+// been sorted in creating the Manifold, so the indices are converted using
+// meshRelation_.
+std::vector<Smoothness> Manifold::Impl::UpdateSharpenedEdges(
+    const std::vector<Smoothness>& sharpenedEdges) const {
+  std::unordered_map<int, int> oldHalfedge2New;
+  for (int tri = 0; tri < NumTri(); ++tri) {
+    int oldTri = meshRelation_.triRef[tri].tri;
+    for (int i : {0, 1, 2}) oldHalfedge2New[3 * oldTri + i] = 3 * tri + i;
+  }
+  std::vector<Smoothness> newSharp = sharpenedEdges;
+  for (Smoothness& edge : newSharp) {
+    edge.halfedge = oldHalfedge2New[edge.halfedge];
+  }
+  return newSharp;
+}
+
+// Find faces containing at least 3 triangles - these will not have
+// interpolated normals - all their vert normals must match their face normal.
+Vec<bool> Manifold::Impl::FlatFaces() const {
+  const int numTri = NumTri();
+  Vec<bool> triIsFlatFace(numTri, false);
+  for_each_n(autoPolicy(numTri), countAt(0), numTri,
+             [this, &triIsFlatFace](const int tri) {
+               const TriRef& ref = meshRelation_.triRef[tri];
+               int faceNeighbors = 0;
+               glm::ivec3 faceTris = {-1, -1, -1};
+               for (const int j : {0, 1, 2}) {
+                 const int neighborTri =
+                     halfedge_[halfedge_[3 * tri + j].pairedHalfedge].face;
+                 const TriRef& jRef = meshRelation_.triRef[neighborTri];
+                 if (jRef.SameFace(ref)) {
+                   ++faceNeighbors;
+                   faceTris[j] = neighborTri;
+                 }
+               }
+               if (faceNeighbors > 1) {
+                 triIsFlatFace[tri] = true;
+                 for (const int j : {0, 1, 2}) {
+                   if (faceTris[j] >= 0) {
+                     triIsFlatFace[faceTris[j]] = true;
+                   }
+                 }
+               }
+             });
+  return triIsFlatFace;
+}
+
+// Returns a vector of length numVert that has a tri that is part of a
+// neighboring flat face if there is only one flat face. If there are none it
+// gets -1, and if there are more than one it gets -2.
+Vec<int> Manifold::Impl::VertFlatFace(const Vec<bool>& flatFaces) const {
+  Vec<int> vertFlatFace(NumVert(), -1);
+  Vec<TriRef> vertRef(NumVert(), {-1, -1, -1});
+  for (int tri = 0; tri < NumTri(); ++tri) {
+    if (flatFaces[tri]) {
+      for (const int j : {0, 1, 2}) {
+        const int vert = halfedge_[3 * tri + j].startVert;
+        if (vertRef[vert].SameFace(meshRelation_.triRef[tri])) continue;
+        vertRef[vert] = meshRelation_.triRef[tri];
+        vertFlatFace[vert] = vertFlatFace[vert] == -1 ? tri : -2;
+      }
+    }
+  }
+  return vertFlatFace;
+}
+
+std::vector<Smoothness> Manifold::Impl::SharpenEdges(
+    float minSharpAngle, float minSmoothness) const {
+  std::vector<Smoothness> sharpenedEdges;
+  const float minRadians = glm::radians(minSharpAngle);
+  for (int e = 0; e < halfedge_.size(); ++e) {
+    if (!halfedge_[e].IsForward()) continue;
+    const int pair = halfedge_[e].pairedHalfedge;
+    const float dihedral =
+        glm::acos(glm::dot(faceNormal_[e / 3], faceNormal_[pair / 3]));
+    if (dihedral > minRadians) {
+      sharpenedEdges.push_back({e, minSmoothness});
+      sharpenedEdges.push_back({pair, minSmoothness});
+    }
+  }
+  return sharpenedEdges;
+}
+
+/**
+ * Sharpen tangents that intersect an edge to sharpen that edge. The weight is
+ * unchanged, as this has a squared effect on radius of curvature, except
+ * in the case of zero radius, which is marked with weight = 0.
+ */
+void Manifold::Impl::SharpenTangent(int halfedge, float smoothness) {
+  halfedgeTangent_[halfedge] =
+      glm::vec4(smoothness * glm::vec3(halfedgeTangent_[halfedge]),
+                smoothness == 0 ? 0 : halfedgeTangent_[halfedge].w);
+}
+
+/**
+ * Instead of calculating the internal shared normals like CalculateNormals
+ * does, this method fills in vertex properties, unshared across edges that
+ * are bent more than minSharpAngle.
+ */
+void Manifold::Impl::SetNormals(int normalIdx, float minSharpAngle) {
+  if (IsEmpty()) return;
+  if (normalIdx < 0) return;
+
+  const int oldNumProp = NumProp();
+  const int numTri = NumTri();
+
+  Vec<bool> triIsFlatFace = FlatFaces();
+  Vec<int> vertFlatFace = VertFlatFace(triIsFlatFace);
+  Vec<int> vertNumSharp(NumVert(), 0);
+  for (int e = 0; e < halfedge_.size(); ++e) {
+    if (!halfedge_[e].IsForward()) continue;
+    const int pair = halfedge_[e].pairedHalfedge;
+    const int tri1 = e / 3;
+    const int tri2 = pair / 3;
+    const float dihedral =
+        glm::degrees(glm::acos(glm::dot(faceNormal_[tri1], faceNormal_[tri2])));
+    if (dihedral > minSharpAngle) {
+      ++vertNumSharp[halfedge_[e].startVert];
+      ++vertNumSharp[halfedge_[e].endVert];
+    } else {
+      const bool faceSplit =
+          triIsFlatFace[tri1] != triIsFlatFace[tri2] ||
+          (triIsFlatFace[tri1] && triIsFlatFace[tri2] &&
+           !meshRelation_.triRef[tri1].SameFace(meshRelation_.triRef[tri2]));
+      if (vertFlatFace[halfedge_[e].startVert] == -2 && faceSplit) {
+        ++vertNumSharp[halfedge_[e].startVert];
+      }
+      if (vertFlatFace[halfedge_[e].endVert] == -2 && faceSplit) {
+        ++vertNumSharp[halfedge_[e].endVert];
+      }
+    }
+  }
+
+  const int numProp = glm::max(oldNumProp, normalIdx + 3);
+  Vec<float> oldProperties(numProp * NumPropVert(), 0);
+  meshRelation_.properties.swap(oldProperties);
+  meshRelation_.numProp = numProp;
+  if (meshRelation_.triProperties.size() == 0) {
+    meshRelation_.triProperties.resize(numTri);
+    for_each_n(autoPolicy(numTri), countAt(0), numTri, [this](int tri) {
+      for (const int j : {0, 1, 2})
+        meshRelation_.triProperties[tri][j] = halfedge_[3 * tri + j].startVert;
+    });
+  }
+  Vec<glm::ivec3> oldTriProp(numTri, {-1, -1, -1});
+  meshRelation_.triProperties.swap(oldTriProp);
+
+  for (int tri = 0; tri < numTri; ++tri) {
+    for (const int i : {0, 1, 2}) {
+      if (meshRelation_.triProperties[tri][i] >= 0) continue;
+      int startEdge = 3 * tri + i;
+      const int vert = halfedge_[startEdge].startVert;
+
+      if (vertNumSharp[vert] < 2) {
+        const glm::vec3 normal = vertFlatFace[vert] >= 0
+                                     ? faceNormal_[vertFlatFace[vert]]
+                                     : vertNormal_[vert];
+        int lastProp = -1;
+        ForVert(startEdge, [&](int current) {
+          const int thisTri = current / 3;
+          const int j = current - 3 * thisTri;
+          const int prop = oldTriProp[thisTri][j];
+          meshRelation_.triProperties[thisTri][j] = prop;
+          if (prop == lastProp) return;
+          lastProp = prop;
+          auto start = oldProperties.begin() + prop * oldNumProp;
+          std::copy(start, start + oldNumProp,
+                    meshRelation_.properties.begin() + prop * numProp);
+          for (const int i : {0, 1, 2})
+            meshRelation_.properties[prop * numProp + normalIdx + i] =
+                normal[i];
+        });
+      } else {
+        const glm::vec3 centerPos = vertPos_[vert];
+        // Length degree
+        std::vector<int> group;
+        // Length number of normals
+        std::vector<glm::vec3> normals;
+        int current = startEdge;
+        int prevFace = halfedge_[current].face;
+
+        do {
+          int next = NextHalfedge(halfedge_[current].pairedHalfedge);
+          const int face = halfedge_[next].face;
+
+          const float dihedral = glm::degrees(
+              glm::acos(glm::dot(faceNormal_[face], faceNormal_[prevFace])));
+          if (dihedral > minSharpAngle ||
+              triIsFlatFace[face] != triIsFlatFace[prevFace] ||
+              (triIsFlatFace[face] && triIsFlatFace[prevFace] &&
+               !meshRelation_.triRef[face].SameFace(
+                   meshRelation_.triRef[prevFace]))) {
+            break;
+          }
+          current = next;
+          prevFace = face;
+        } while (current != startEdge);
+
+        const int endEdge = current;
+
+        struct FaceEdge {
+          int face;
+          glm::vec3 edgeVec;
+        };
+
+        ForVert<FaceEdge>(
+            endEdge,
+            [this, centerPos](int current) {
+              return FaceEdge(
+                  {halfedge_[current].face,
+                   glm::normalize(vertPos_[halfedge_[current].endVert] -
+                                  centerPos)});
+            },
+            [this, &triIsFlatFace, &normals, &group, minSharpAngle](
+                int current, const FaceEdge& here, const FaceEdge& next) {
+              const float dihedral = glm::degrees(glm::acos(
+                  glm::dot(faceNormal_[here.face], faceNormal_[next.face])));
+              if (dihedral > minSharpAngle ||
+                  triIsFlatFace[here.face] != triIsFlatFace[next.face] ||
+                  (triIsFlatFace[here.face] && triIsFlatFace[next.face] &&
+                   !meshRelation_.triRef[here.face].SameFace(
+                       meshRelation_.triRef[next.face]))) {
+                normals.push_back(glm::vec3(0));
+              }
+              group.push_back(normals.size() - 1);
+              float dot = glm::dot(here.edgeVec, next.edgeVec);
+              const float phi =
+                  dot >= 1 ? kTolerance
+                           : (dot <= -1 ? glm::pi<float>() : glm::acos(dot));
+              normals.back() += faceNormal_[next.face] * phi;
+            });
+
+        for (auto& normal : normals) {
+          normal = glm::normalize(normal);
+        }
+
+        int lastGroup = 0;
+        int lastProp = -1;
+        int newProp = -1;
+        int idx = 0;
+        ForVert(endEdge, [&](int current1) {
+          const int thisTri = current1 / 3;
+          const int j = current1 - 3 * thisTri;
+          const int prop = oldTriProp[thisTri][j];
+          auto start = oldProperties.begin() + prop * oldNumProp;
+
+          if (group[idx] != lastGroup && group[idx] != 0 && prop == lastProp) {
+            lastGroup = group[idx];
+            newProp = NumPropVert();
+            meshRelation_.properties.resize(meshRelation_.properties.size() +
+                                            numProp);
+            std::copy(start, start + oldNumProp,
+                      meshRelation_.properties.begin() + newProp * numProp);
+            for (const int i : {0, 1, 2}) {
+              meshRelation_.properties[newProp * numProp + normalIdx + i] =
+                  normals[group[idx]][i];
+            }
+          } else if (prop != lastProp) {
+            lastProp = prop;
+            newProp = prop;
+            std::copy(start, start + oldNumProp,
+                      meshRelation_.properties.begin() + prop * numProp);
+            for (const int i : {0, 1, 2})
+              meshRelation_.properties[prop * numProp + normalIdx + i] =
+                  normals[group[idx]][i];
+          }
+
+          meshRelation_.triProperties[thisTri][j] = newProp;
+          ++idx;
+        });
+      }
+    }
+  }
+}
+
+/**
+ * Tangents get flattened to create sharp edges by setting their weight to zero.
+ * This is the natural limit of reducing the weight to increase the sharpness
+ * smoothly. This limit gives a decent shape, but it causes the parameterization
+ * to be stretched and compresses it near the edges, which is good for resolving
+ * tight curvature, but bad for property interpolation. This function fixes the
+ * parameter stretch at the limit for sharp edges, since there is no curvature
+ * to resolve. Note this also changes the overall shape - making it more evenly
+ * curved.
+ */
+void Manifold::Impl::LinearizeFlatTangents() {
+  const int n = halfedgeTangent_.size();
+  for_each_n(
+      autoPolicy(n), zip(halfedgeTangent_.begin(), countAt(0)), n,
+      [this](thrust::tuple<glm::vec4&, int> inOut) {
+        glm::vec4& tangent = thrust::get<0>(inOut);
+        const int halfedge = thrust::get<1>(inOut);
+        glm::vec4& otherTangent =
+            halfedgeTangent_[halfedge_[halfedge].pairedHalfedge];
+
+        const glm::bvec2 flat(tangent.w == 0, otherTangent.w == 0);
+        if (!halfedge_[halfedge].IsForward() || (!flat[0] && !flat[1])) {
+          return;
+        }
+
+        const glm::vec3 edgeVec = vertPos_[halfedge_[halfedge].endVert] -
+                                  vertPos_[halfedge_[halfedge].startVert];
+
+        if (flat[0] && flat[1]) {
+          tangent = glm::vec4(edgeVec / 3.0f, 1);
+          otherTangent = glm::vec4(-edgeVec / 3.0f, 1);
+        } else if (flat[0]) {
+          tangent = glm::vec4((edgeVec + glm::vec3(otherTangent)) / 2.0f, 1);
+        } else {
+          otherTangent = glm::vec4((-edgeVec + glm::vec3(tangent)) / 2.0f, 1);
+        }
+      });
+}
+
+/**
+ * Calculates halfedgeTangent_, allowing the manifold to be refined and
+ * smoothed. The tangents form weighted cubic Beziers along each edge. This
+ * function creates circular arcs where possible (minimizing maximum curvature),
+ * constrained to the indicated property normals. Across edges that form
+ * discontinuities in the normals, the tangent vectors are zero-length, allowing
+ * the shape to form a sharp corner with minimal oscillation.
+ */
+void Manifold::Impl::CreateTangents(int normalIdx) {
+  ZoneScoped;
+  const int numVert = NumVert();
+  const int numHalfedge = halfedge_.size();
+  halfedgeTangent_.resize(0);
+  Vec<glm::vec4> tangent(numHalfedge);
+
+  Vec<glm::vec3> vertNormal(numVert);
+  Vec<glm::ivec2> vertSharpHalfedge(numVert, glm::ivec2(-1));
+  for (int e = 0; e < numHalfedge; ++e) {
+    const int vert = halfedge_[e].startVert;
+    auto& sharpHalfedge = vertSharpHalfedge[vert];
+    if (sharpHalfedge[0] >= 0 && sharpHalfedge[1] >= 0) continue;
+
+    int idx = 0;
+    // Only used when there is only one.
+    glm::vec3& lastNormal = vertNormal[vert];
+
+    ForVert<glm::vec3>(
+        e,
+        [normalIdx, this](int halfedge) {
+          return GetNormal(halfedge, normalIdx);
+        },
+        [&sharpHalfedge, &idx, &lastNormal](int halfedge,
+                                            const glm::vec3& normal,
+                                            const glm::vec3& nextNormal) {
+          const glm::vec3 diff = nextNormal - normal;
+          if (glm::dot(diff, diff) > kTolerance * kTolerance) {
+            if (idx < 2) {
+              sharpHalfedge[idx++] = halfedge;
+            } else {
+              sharpHalfedge[0] = -1;  // marks more than 2 sharp edges
+            }
+          }
+          lastNormal = normal;
+        });
+  }
+
+  for_each_n(autoPolicy(numHalfedge),
+             zip(tangent.begin(), halfedge_.cbegin(), countAt(0)), numHalfedge,
+             SmoothBezier({this, vertNormal}));
+
+  halfedgeTangent_.swap(tangent);
+
+  for (int vert = 0; vert < numVert; ++vert) {
+    const int first = vertSharpHalfedge[vert][0];
+    const int second = vertSharpHalfedge[vert][1];
+    if (second == -1) continue;
+    if (first != -1) {  // Make continuous edge
+      const glm::vec3 newTangent = glm::normalize(glm::cross(
+          GetNormal(first, normalIdx), GetNormal(second, normalIdx)));
+      if (!isfinite(newTangent[0])) continue;
+
+      halfedgeTangent_[first] = CircularTangent(
+          newTangent, vertPos_[halfedge_[first].endVert] - vertPos_[vert]);
+      halfedgeTangent_[second] = CircularTangent(
+          -newTangent, vertPos_[halfedge_[second].endVert] - vertPos_[vert]);
+
+      ForVert(first, [this, first, second](int current) {
+        if (current != first && current != second &&
+            !IsMarkedInsideQuad(current)) {
+          SharpenTangent(current, 0);
+        }
+      });
+    } else {  // Sharpen vertex uniformly
+      ForVert(second, [this](int current) {
+        if (!IsMarkedInsideQuad(current)) {
+          SharpenTangent(current, 0);
+        }
+      });
+    }
+  }
+  LinearizeFlatTangents();
+}
+
+/**
+ * Calculates halfedgeTangent_, allowing the manifold to be refined and
+ * smoothed. The tangents form weighted cubic Beziers along each edge. This
+ * function creates circular arcs where possible (minimizing maximum curvature),
+ * constrained to the vertex normals. Where sharpenedEdges are specified, the
+ * tangents are shortened that intersect the sharpened edge, concentrating the
+ * curvature there, while the tangents of the sharp edges themselves are aligned
+ * for continuity.
+ */
+void Manifold::Impl::CreateTangents(std::vector<Smoothness> sharpenedEdges) {
+  ZoneScoped;
+  const int numHalfedge = halfedge_.size();
+  halfedgeTangent_.resize(0);
+  Vec<glm::vec4> tangent(numHalfedge);
+
+  Vec<bool> triIsFlatFace = FlatFaces();
+  Vec<int> vertFlatFace = VertFlatFace(triIsFlatFace);
+  Vec<glm::vec3> vertNormal = vertNormal_;
+  for (int v = 0; v < NumVert(); ++v) {
+    if (vertFlatFace[v] >= 0) {
+      vertNormal[v] = faceNormal_[vertFlatFace[v]];
+    }
+  }
+
+  for_each_n(autoPolicy(numHalfedge),
+             zip(tangent.begin(), halfedge_.cbegin(), countAt(0)), numHalfedge,
+             SmoothBezier({this, vertNormal}));
+
+  halfedgeTangent_.swap(tangent);
+
+  // Add sharpened edges around faces, just on the face side.
+  for (int tri = 0; tri < NumTri(); ++tri) {
+    if (!triIsFlatFace[tri]) continue;
+    for (const int j : {0, 1, 2}) {
+      const int tri2 = halfedge_[3 * tri + j].pairedHalfedge / 3;
+      if (!triIsFlatFace[tri2] ||
+          !meshRelation_.triRef[tri].SameFace(meshRelation_.triRef[tri2])) {
+        sharpenedEdges.push_back({3 * tri + j, 0});
+      }
+    }
+  }
+
+  if (sharpenedEdges.empty()) return;
+
+  using Pair = std::pair<Smoothness, Smoothness>;
+  // Fill in missing pairs with default smoothness = 1.
+  std::map<int, Pair> edges;
+  for (Smoothness edge : sharpenedEdges) {
+    if (edge.smoothness >= 1) continue;
+    const bool forward = halfedge_[edge.halfedge].IsForward();
+    const int pair = halfedge_[edge.halfedge].pairedHalfedge;
+    const int idx = forward ? edge.halfedge : pair;
+    if (edges.find(idx) == edges.end()) {
+      edges[idx] = {edge, {pair, 1}};
+      if (!forward) std::swap(edges[idx].first, edges[idx].second);
+    } else {
+      Smoothness& e = forward ? edges[idx].first : edges[idx].second;
+      e.smoothness = glm::min(edge.smoothness, e.smoothness);
+    }
+  }
+
+  std::map<int, std::vector<Pair>> vertTangents;
+  for (const auto& value : edges) {
+    const Pair edge = value.second;
+    vertTangents[halfedge_[edge.first.halfedge].startVert].push_back(edge);
+    vertTangents[halfedge_[edge.second.halfedge].startVert].push_back(
+        {edge.second, edge.first});
+  }
+
+  for (const auto& value : vertTangents) {
+    const std::vector<Pair>& vert = value.second;
+    // Sharp edges that end are smooth at their terminal vert.
+    if (vert.size() == 1) continue;
+    if (vert.size() == 2) {  // Make continuous edge
+      const int first = vert[0].first.halfedge;
+      const int second = vert[1].first.halfedge;
+      const glm::vec3 newTangent =
+          glm::normalize(glm::vec3(halfedgeTangent_[first]) -
+                         glm::vec3(halfedgeTangent_[second]));
+
+      const glm::vec3 pos = vertPos_[halfedge_[first].startVert];
+      halfedgeTangent_[first] =
+          CircularTangent(newTangent, vertPos_[halfedge_[first].endVert] - pos);
+      halfedgeTangent_[second] = CircularTangent(
+          -newTangent, vertPos_[halfedge_[second].endVert] - pos);
+
+      float smoothness =
+          (vert[0].second.smoothness + vert[1].first.smoothness) / 2;
+      ForVert(first, [this, &smoothness, &vert, first, second](int current) {
+        if (current == second) {
+          smoothness =
+              (vert[1].second.smoothness + vert[0].first.smoothness) / 2;
+        } else if (current != first && !IsMarkedInsideQuad(current)) {
+          SharpenTangent(current, smoothness);
+        }
+      });
+    } else {  // Sharpen vertex uniformly
+      float smoothness = 0;
+      for (const Pair& pair : vert) {
+        smoothness += pair.first.smoothness;
+        smoothness += pair.second.smoothness;
+      }
+      smoothness /= 2 * vert.size();
+
+      ForVert(vert[0].first.halfedge, [this, smoothness](int current) {
+        if (!IsMarkedInsideQuad(current)) {
+          SharpenTangent(current, smoothness);
+        }
+      });
+    }
+  }
+  LinearizeFlatTangents();
+}
+
+void Manifold::Impl::Refine(std::function<int(glm::vec3)> edgeDivisions) {
+  if (IsEmpty()) return;
+  Manifold::Impl old = *this;
+  Vec<Barycentric> vertBary = Subdivide(edgeDivisions);
+  if (vertBary.size() == 0) return;
+
+  if (old.halfedgeTangent_.size() == old.halfedge_.size()) {
+    for_each_n(autoPolicy(NumTri()), zip(vertPos_.begin(), vertBary.begin()),
+               NumVert(), InterpTri({&old}));
+    // Make original since the subdivided faces have been warped into
+    // being non-coplanar, and hence not being related to the original faces.
+    meshRelation_.originalID = ReserveIDs(1);
+    InitializeOriginal();
+    CreateFaces();
+  }
+
+  halfedgeTangent_.resize(0);
+  Finish();
+}
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/sort.cpp b/thirdparty/manifold/src/manifold/src/sort.cpp
new file mode 100644
index 000000000000..a5dba9b3dcd0
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/sort.cpp
@@ -0,0 +1,616 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <atomic>
+#include <numeric>
+#include <set>
+
+#include "impl.h"
+#include "par.h"
+
+namespace {
+using namespace manifold;
+
+constexpr uint32_t kNoCode = 0xFFFFFFFFu;
+
+struct Extrema : public thrust::binary_function<Halfedge, Halfedge, Halfedge> {
+  void MakeForward(Halfedge& a) {
+    if (!a.IsForward()) {
+      int tmp = a.startVert;
+      a.startVert = a.endVert;
+      a.endVert = tmp;
+    }
+  }
+
+  int MaxOrMinus(int a, int b) {
+    return glm::min(a, b) < 0 ? -1 : glm::max(a, b);
+  }
+
+  Halfedge operator()(Halfedge a, Halfedge b) {
+    MakeForward(a);
+    MakeForward(b);
+    a.startVert = glm::min(a.startVert, b.startVert);
+    a.endVert = glm::max(a.endVert, b.endVert);
+    a.face = MaxOrMinus(a.face, b.face);
+    a.pairedHalfedge = MaxOrMinus(a.pairedHalfedge, b.pairedHalfedge);
+    return a;
+  }
+};
+
+uint32_t SpreadBits3(uint32_t v) {
+  v = 0xFF0000FFu & (v * 0x00010001u);
+  v = 0x0F00F00Fu & (v * 0x00000101u);
+  v = 0xC30C30C3u & (v * 0x00000011u);
+  v = 0x49249249u & (v * 0x00000005u);
+  return v;
+}
+
+uint32_t MortonCode(glm::vec3 position, Box bBox) {
+  // Unreferenced vertices are marked NaN, and this will sort them to the end
+  // (the Morton code only uses the first 30 of 32 bits).
+  if (isnan(position.x)) return kNoCode;
+
+  glm::vec3 xyz = (position - bBox.min) / (bBox.max - bBox.min);
+  xyz = glm::min(glm::vec3(1023.0f), glm::max(glm::vec3(0.0f), 1024.0f * xyz));
+  uint32_t x = SpreadBits3(static_cast<uint32_t>(xyz.x));
+  uint32_t y = SpreadBits3(static_cast<uint32_t>(xyz.y));
+  uint32_t z = SpreadBits3(static_cast<uint32_t>(xyz.z));
+  return x * 4 + y * 2 + z;
+}
+
+struct Morton {
+  const Box bBox;
+
+  void operator()(thrust::tuple<uint32_t&, const glm::vec3&> inout) {
+    glm::vec3 position = thrust::get<1>(inout);
+    thrust::get<0>(inout) = MortonCode(position, bBox);
+  }
+};
+
+struct FaceMortonBox {
+  VecView<const Halfedge> halfedge;
+  VecView<const glm::vec3> vertPos;
+  const Box bBox;
+
+  void operator()(thrust::tuple<uint32_t&, Box&, int> inout) {
+    uint32_t& mortonCode = thrust::get<0>(inout);
+    Box& faceBox = thrust::get<1>(inout);
+    int face = thrust::get<2>(inout);
+
+    // Removed tris are marked by all halfedges having pairedHalfedge = -1, and
+    // this will sort them to the end (the Morton code only uses the first 30 of
+    // 32 bits).
+    if (halfedge[3 * face].pairedHalfedge < 0) {
+      mortonCode = kNoCode;
+      return;
+    }
+
+    glm::vec3 center(0.0f);
+
+    for (const int i : {0, 1, 2}) {
+      const glm::vec3 pos = vertPos[halfedge[3 * face + i].startVert];
+      center += pos;
+      faceBox.Union(pos);
+    }
+    center /= 3;
+
+    mortonCode = MortonCode(center, bBox);
+  }
+};
+
+struct Reindex {
+  VecView<const int> indexInv;
+
+  void operator()(Halfedge& edge) {
+    if (edge.startVert < 0) return;
+    edge.startVert = indexInv[edge.startVert];
+    edge.endVert = indexInv[edge.endVert];
+  }
+};
+
+struct MarkProp {
+  VecView<int> keep;
+
+  void operator()(glm::ivec3 triProp) {
+    for (const int i : {0, 1, 2}) {
+      reinterpret_cast<std::atomic<int>*>(&keep[triProp[i]])
+          ->store(1, std::memory_order_relaxed);
+    }
+  }
+};
+
+struct GatherProps {
+  VecView<float> properties;
+  VecView<const float> oldProperties;
+  const int numProp;
+
+  void operator()(thrust::tuple<int, int, int> in) {
+    const int oldIdx = thrust::get<0>(in);
+    const int newIdx = thrust::get<1>(in);
+    const int keep = thrust::get<2>(in);
+    if (keep == 0) return;
+    for (int p = 0; p < numProp; ++p) {
+      properties[newIdx * numProp + p] = oldProperties[oldIdx * numProp + p];
+    }
+  }
+};
+
+struct ReindexProps {
+  VecView<const int> old2new;
+
+  void operator()(glm::ivec3& triProp) {
+    for (const int i : {0, 1, 2}) {
+      triProp[i] = old2new[triProp[i]];
+    }
+  }
+};
+
+template <typename T>
+void Permute(Vec<T>& inOut, const Vec<int>& new2Old) {
+  Vec<T> tmp(std::move(inOut));
+  inOut.resize(new2Old.size());
+  gather(autoPolicy(new2Old.size()), new2Old.begin(), new2Old.end(),
+         tmp.begin(), inOut.begin());
+}
+
+template void Permute<TriRef>(Vec<TriRef>&, const Vec<int>&);
+template void Permute<glm::vec3>(Vec<glm::vec3>&, const Vec<int>&);
+
+struct ReindexFace {
+  VecView<Halfedge> halfedge;
+  VecView<glm::vec4> halfedgeTangent;
+  VecView<const Halfedge> oldHalfedge;
+  VecView<const glm::vec4> oldHalfedgeTangent;
+  VecView<const int> faceNew2Old;
+  VecView<const int> faceOld2New;
+
+  void operator()(int newFace) {
+    const int oldFace = faceNew2Old[newFace];
+    for (const int i : {0, 1, 2}) {
+      const int oldEdge = 3 * oldFace + i;
+      Halfedge edge = oldHalfedge[oldEdge];
+      edge.face = newFace;
+      const int pairedFace = edge.pairedHalfedge / 3;
+      const int offset = edge.pairedHalfedge - 3 * pairedFace;
+      edge.pairedHalfedge = 3 * faceOld2New[pairedFace] + offset;
+      const int newEdge = 3 * newFace + i;
+      halfedge[newEdge] = edge;
+      if (!oldHalfedgeTangent.empty()) {
+        halfedgeTangent[newEdge] = oldHalfedgeTangent[oldEdge];
+      }
+    }
+  }
+};
+
+struct VertMortonBox {
+  VecView<const float> vertProperties;
+  const uint32_t numProp;
+  const float tol;
+  const Box bBox;
+
+  void operator()(thrust::tuple<uint32_t&, Box&, int> inout) {
+    uint32_t& mortonCode = thrust::get<0>(inout);
+    Box& vertBox = thrust::get<1>(inout);
+    int vert = thrust::get<2>(inout);
+
+    const glm::vec3 center(vertProperties[numProp * vert],
+                           vertProperties[numProp * vert + 1],
+                           vertProperties[numProp * vert + 2]);
+
+    vertBox.min = center - tol / 2;
+    vertBox.max = center + tol / 2;
+
+    mortonCode = MortonCode(center, bBox);
+  }
+};
+
+struct Duplicate {
+  thrust::pair<float, float> operator()(float x) {
+    return thrust::make_pair(x, x);
+  }
+};
+
+struct MinMax : public thrust::binary_function<thrust::pair<float, float>,
+                                               thrust::pair<float, float>,
+                                               thrust::pair<float, float>> {
+  thrust::pair<float, float> operator()(thrust::pair<float, float> a,
+                                        thrust::pair<float, float> b) {
+    return thrust::make_pair(glm::min(a.first, b.first),
+                             glm::max(a.second, b.second));
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Once halfedge_ has been filled in, this function can be called to create the
+ * rest of the internal data structures. This function also removes the verts
+ * and halfedges flagged for removal (NaN verts and -1 halfedges).
+ */
+void Manifold::Impl::Finish() {
+  if (halfedge_.size() == 0) return;
+
+  CalculateBBox();
+  SetPrecision(precision_);
+  if (!bBox_.IsFinite()) {
+    // Decimated out of existence - early out.
+    MarkFailure(Error::NoError);
+    return;
+  }
+
+  SortVerts();
+  Vec<Box> faceBox;
+  Vec<uint32_t> faceMorton;
+  GetFaceBoxMorton(faceBox, faceMorton);
+  SortFaces(faceBox, faceMorton);
+  if (halfedge_.size() == 0) return;
+  CompactProps();
+
+  ASSERT(halfedge_.size() % 6 == 0, topologyErr,
+         "Not an even number of faces after sorting faces!");
+
+#ifdef MANIFOLD_DEBUG
+  Halfedge extrema = {0, 0, 0, 0};
+  extrema = reduce<Halfedge>(autoPolicy(halfedge_.size()), halfedge_.begin(),
+                             halfedge_.end(), extrema, Extrema());
+#endif
+
+  ASSERT(extrema.startVert >= 0, topologyErr, "Vertex index is negative!");
+  ASSERT(extrema.endVert < NumVert(), topologyErr,
+         "Vertex index exceeds number of verts!");
+  ASSERT(extrema.face >= 0, topologyErr, "Face index is negative!");
+  ASSERT(extrema.face < NumTri(), topologyErr,
+         "Face index exceeds number of faces!");
+  ASSERT(extrema.pairedHalfedge >= 0, topologyErr,
+         "Halfedge index is negative!");
+  ASSERT(extrema.pairedHalfedge < 2 * NumEdge(), topologyErr,
+         "Halfedge index exceeds number of halfedges!");
+  ASSERT(meshRelation_.triRef.size() == NumTri() ||
+             meshRelation_.triRef.size() == 0,
+         logicErr, "Mesh Relation doesn't fit!");
+  ASSERT(faceNormal_.size() == NumTri() || faceNormal_.size() == 0, logicErr,
+         "faceNormal size = " + std::to_string(faceNormal_.size()) +
+             ", NumTri = " + std::to_string(NumTri()));
+  // TODO: figure out why this has a flaky failure and then enable reading
+  // vertNormals from a Mesh.
+  // ASSERT(vertNormal_.size() == NumVert() || vertNormal_.size() == 0,
+  // logicErr,
+  //        "vertNormal size = " + std::to_string(vertNormal_.size()) +
+  //            ", NumVert = " + std::to_string(NumVert()));
+
+  CalculateNormals();
+  collider_ = Collider(faceBox, faceMorton);
+
+  ASSERT(Is2Manifold(), logicErr, "mesh is not 2-manifold!");
+}
+
+/**
+ * Sorts the vertices according to their Morton code.
+ */
+void Manifold::Impl::SortVerts() {
+  ZoneScoped;
+  const int numVert = NumVert();
+  Vec<uint32_t> vertMorton(numVert);
+  auto policy = autoPolicy(numVert);
+  for_each_n(policy, zip(vertMorton.begin(), vertPos_.cbegin()), numVert,
+             Morton({bBox_}));
+
+  Vec<int> vertNew2Old(numVert);
+  sequence(policy, vertNew2Old.begin(), vertNew2Old.end());
+
+  stable_sort(policy, zip(vertMorton.begin(), vertNew2Old.begin()),
+              zip(vertMorton.end(), vertNew2Old.end()),
+              [](const thrust::tuple<uint32_t, int>& a,
+                 const thrust::tuple<uint32_t, int>& b) {
+                return thrust::get<0>(a) < thrust::get<0>(b);
+              });
+
+  ReindexVerts(vertNew2Old, numVert);
+
+  // Verts were flagged for removal with NaNs and assigned kNoCode to sort
+  // them to the end, which allows them to be removed.
+  const int newNumVert =
+      std::lower_bound(vertMorton.begin(), vertMorton.end(), kNoCode) -
+      vertMorton.begin();
+
+  vertNew2Old.resize(newNumVert);
+  Permute(vertPos_, vertNew2Old);
+
+  if (vertNormal_.size() == numVert) {
+    Permute(vertNormal_, vertNew2Old);
+  }
+}
+
+/**
+ * Updates the halfedges to point to new vert indices based on a mapping,
+ * vertNew2Old. This may be a subset, so the total number of original verts is
+ * also given.
+ */
+void Manifold::Impl::ReindexVerts(const Vec<int>& vertNew2Old, int oldNumVert) {
+  ZoneScoped;
+  Vec<int> vertOld2New(oldNumVert);
+  scatter(autoPolicy(oldNumVert), countAt(0), countAt(NumVert()),
+          vertNew2Old.begin(), vertOld2New.begin());
+  for_each(autoPolicy(oldNumVert), halfedge_.begin(), halfedge_.end(),
+           Reindex({vertOld2New}));
+}
+
+/**
+ * Removes unreferenced property verts and reindexes triProperties.
+ */
+void Manifold::Impl::CompactProps() {
+  ZoneScoped;
+  if (meshRelation_.numProp == 0) return;
+
+  const int numVerts = meshRelation_.properties.size() / meshRelation_.numProp;
+  Vec<int> keep(numVerts, 0);
+  auto policy = autoPolicy(numVerts);
+
+  for_each(policy, meshRelation_.triProperties.cbegin(),
+           meshRelation_.triProperties.cend(), MarkProp({keep}));
+  Vec<int> propOld2New(numVerts + 1, 0);
+  inclusive_scan(policy, keep.begin(), keep.end(), propOld2New.begin() + 1);
+
+  Vec<float> oldProp = meshRelation_.properties;
+  const int numVertsNew = propOld2New[numVerts];
+  meshRelation_.properties.resize(meshRelation_.numProp * numVertsNew);
+  for_each_n(
+      policy, zip(countAt(0), propOld2New.cbegin(), keep.cbegin()), numVerts,
+      GatherProps({meshRelation_.properties, oldProp, meshRelation_.numProp}));
+  for_each_n(policy, meshRelation_.triProperties.begin(), NumTri(),
+             ReindexProps({propOld2New}));
+}
+
+/**
+ * Fills the faceBox and faceMorton input with the bounding boxes and Morton
+ * codes of the faces, respectively. The Morton code is based on the center of
+ * the bounding box.
+ */
+void Manifold::Impl::GetFaceBoxMorton(Vec<Box>& faceBox,
+                                      Vec<uint32_t>& faceMorton) const {
+  ZoneScoped;
+  faceBox.resize(NumTri());
+  faceMorton.resize(NumTri());
+  for_each_n(autoPolicy(NumTri()),
+             zip(faceMorton.begin(), faceBox.begin(), countAt(0)), NumTri(),
+             FaceMortonBox({halfedge_, vertPos_, bBox_}));
+}
+
+/**
+ * Sorts the faces of this manifold according to their input Morton code. The
+ * bounding box and Morton code arrays are also sorted accordingly.
+ */
+void Manifold::Impl::SortFaces(Vec<Box>& faceBox, Vec<uint32_t>& faceMorton) {
+  ZoneScoped;
+  Vec<int> faceNew2Old(NumTri());
+  auto policy = autoPolicy(faceNew2Old.size());
+  sequence(policy, faceNew2Old.begin(), faceNew2Old.end());
+
+  stable_sort(policy, zip(faceMorton.begin(), faceNew2Old.begin()),
+              zip(faceMorton.end(), faceNew2Old.end()),
+              [](const thrust::tuple<uint32_t, int>& a,
+                 const thrust::tuple<uint32_t, int>& b) {
+                return thrust::get<0>(a) < thrust::get<0>(b);
+              });
+
+  // Tris were flagged for removal with pairedHalfedge = -1 and assigned kNoCode
+  // to sort them to the end, which allows them to be removed.
+  const int newNumTri =
+      find<decltype(faceMorton.begin())>(policy, faceMorton.begin(),
+                                         faceMorton.end(), kNoCode) -
+      faceMorton.begin();
+  faceMorton.resize(newNumTri);
+  faceNew2Old.resize(newNumTri);
+
+  Permute(faceBox, faceNew2Old);
+  GatherFaces(faceNew2Old);
+}
+
+/**
+ * Creates the halfedge_ vector for this manifold by copying a set of faces from
+ * another manifold, given by oldHalfedge. Input faceNew2Old defines the old
+ * faces to gather into this.
+ */
+void Manifold::Impl::GatherFaces(const Vec<int>& faceNew2Old) {
+  ZoneScoped;
+  const int numTri = faceNew2Old.size();
+  if (meshRelation_.triRef.size() == NumTri())
+    Permute(meshRelation_.triRef, faceNew2Old);
+  if (meshRelation_.triProperties.size() == NumTri())
+    Permute(meshRelation_.triProperties, faceNew2Old);
+  if (faceNormal_.size() == NumTri()) Permute(faceNormal_, faceNew2Old);
+
+  Vec<Halfedge> oldHalfedge(std::move(halfedge_));
+  Vec<glm::vec4> oldHalfedgeTangent(std::move(halfedgeTangent_));
+  Vec<int> faceOld2New(oldHalfedge.size() / 3);
+  auto policy = autoPolicy(numTri);
+  scatter(policy, countAt(0), countAt(numTri), faceNew2Old.begin(),
+          faceOld2New.begin());
+
+  halfedge_.resize(3 * numTri);
+  if (oldHalfedgeTangent.size() != 0) halfedgeTangent_.resize(3 * numTri);
+  for_each_n(policy, countAt(0), numTri,
+             ReindexFace({halfedge_, halfedgeTangent_, oldHalfedge,
+                          oldHalfedgeTangent, faceNew2Old, faceOld2New}));
+}
+
+void Manifold::Impl::GatherFaces(const Impl& old, const Vec<int>& faceNew2Old) {
+  ZoneScoped;
+  const int numTri = faceNew2Old.size();
+  auto policy = autoPolicy(numTri);
+
+  meshRelation_.triRef.resize(numTri);
+  gather(policy, faceNew2Old.begin(), faceNew2Old.end(),
+         old.meshRelation_.triRef.begin(), meshRelation_.triRef.begin());
+
+  for (const auto& pair : old.meshRelation_.meshIDtransform) {
+    meshRelation_.meshIDtransform[pair.first] = pair.second;
+  }
+
+  if (old.meshRelation_.triProperties.size() > 0) {
+    meshRelation_.triProperties.resize(numTri);
+    gather(policy, faceNew2Old.begin(), faceNew2Old.end(),
+           old.meshRelation_.triProperties.begin(),
+           meshRelation_.triProperties.begin());
+    meshRelation_.numProp = old.meshRelation_.numProp;
+    meshRelation_.properties = old.meshRelation_.properties;
+  }
+
+  if (old.faceNormal_.size() == old.NumTri()) {
+    faceNormal_.resize(numTri);
+    gather(policy, faceNew2Old.begin(), faceNew2Old.end(),
+           old.faceNormal_.begin(), faceNormal_.begin());
+  }
+
+  Vec<int> faceOld2New(old.NumTri());
+  scatter(policy, countAt(0), countAt(numTri), faceNew2Old.begin(),
+          faceOld2New.begin());
+
+  halfedge_.resize(3 * numTri);
+  if (old.halfedgeTangent_.size() != 0) halfedgeTangent_.resize(3 * numTri);
+  for_each_n(policy, countAt(0), numTri,
+             ReindexFace({halfedge_, halfedgeTangent_, old.halfedge_,
+                          old.halfedgeTangent_, faceNew2Old, faceOld2New}));
+}
+
+/// Constructs a position-only MeshGL from the input Mesh.
+MeshGL::MeshGL(const Mesh& mesh) {
+  numProp = 3;
+  precision = mesh.precision;
+  vertProperties.resize(numProp * mesh.vertPos.size());
+  for (int i = 0; i < mesh.vertPos.size(); ++i) {
+    for (int j : {0, 1, 2}) vertProperties[3 * i + j] = mesh.vertPos[i][j];
+  }
+  triVerts.resize(3 * mesh.triVerts.size());
+  for (int i = 0; i < mesh.triVerts.size(); ++i) {
+    for (int j : {0, 1, 2}) triVerts[3 * i + j] = mesh.triVerts[i][j];
+  }
+  halfedgeTangent.resize(4 * mesh.halfedgeTangent.size());
+  for (int i = 0; i < mesh.halfedgeTangent.size(); ++i) {
+    for (int j : {0, 1, 2, 3})
+      halfedgeTangent[4 * i + j] = mesh.halfedgeTangent[i][j];
+  }
+}
+
+/**
+ * Updates the mergeFromVert and mergeToVert vectors in order to create a
+ * manifold solid. If the MeshGL is already manifold, no change will occur and
+ * the function will return false. Otherwise, this will merge verts along open
+ * edges within precision (the maximum of the MeshGL precision and the baseline
+ * bounding-box precision), keeping any from the existing merge vectors.
+ *
+ * There is no guarantee the result will be manifold - this is a best-effort
+ * helper function designed primarily to aid in the case where a manifold
+ * multi-material MeshGL was produced, but its merge vectors were lost due to a
+ * round-trip through a file format. Constructing a Manifold from the result
+ * will report a Status if it is not manifold.
+ */
+bool MeshGL::Merge() {
+  ZoneScoped;
+  std::multiset<std::pair<int, int>> openEdges;
+
+  std::vector<int> merge(NumVert());
+  std::iota(merge.begin(), merge.end(), 0);
+  for (int i = 0; i < mergeFromVert.size(); ++i) {
+    merge[mergeFromVert[i]] = mergeToVert[i];
+  }
+
+  const int numVert = NumVert();
+  const int numTri = NumTri();
+  const int next[3] = {1, 2, 0};
+  for (int tri = 0; tri < numTri; ++tri) {
+    for (int i : {0, 1, 2}) {
+      auto edge = std::make_pair(merge[triVerts[3 * tri + next[i]]],
+                                 merge[triVerts[3 * tri + i]]);
+      auto it = openEdges.find(edge);
+      if (it == openEdges.end()) {
+        std::swap(edge.first, edge.second);
+        openEdges.insert(edge);
+      } else {
+        openEdges.erase(it);
+      }
+    }
+  }
+
+  if (openEdges.empty()) {
+    return false;
+  }
+
+  const int numOpenVert = openEdges.size();
+  Vec<int> openVerts(numOpenVert);
+  int i = 0;
+  for (const auto& edge : openEdges) {
+    const int vert = edge.first;
+    openVerts[i++] = vert;
+  }
+
+  Vec<float> vertPropD(vertProperties);
+  Box bBox;
+  for (const int i : {0, 1, 2}) {
+    strided_range<Vec<float>::Iter> iPos(vertPropD.begin() + i, vertPropD.end(),
+                                         numProp);
+    auto minMax = transform_reduce<thrust::pair<float, float>>(
+        autoPolicy(numVert), iPos.begin(), iPos.end(), Duplicate(),
+        thrust::make_pair(std::numeric_limits<float>::infinity(),
+                          -std::numeric_limits<float>::infinity()),
+        MinMax());
+    bBox.min[i] = minMax.first;
+    bBox.max[i] = minMax.second;
+  }
+  precision = MaxPrecision(precision, bBox);
+  if (precision < 0) return false;
+
+  auto policy = autoPolicy(numOpenVert);
+  Vec<Box> vertBox(numOpenVert);
+  Vec<uint32_t> vertMorton(numOpenVert);
+
+  for_each_n(policy,
+             zip(vertMorton.begin(), vertBox.begin(), openVerts.cbegin()),
+             numOpenVert, VertMortonBox({vertPropD, numProp, precision, bBox}));
+
+  stable_sort(policy,
+              zip(vertMorton.begin(), vertBox.begin(), openVerts.begin()),
+              zip(vertMorton.end(), vertBox.end(), openVerts.end()),
+              [](const thrust::tuple<uint32_t, Box, int>& a,
+                 const thrust::tuple<uint32_t, Box, int>& b) {
+                return thrust::get<0>(a) < thrust::get<0>(b);
+              });
+
+  Collider collider(vertBox, vertMorton);
+  SparseIndices toMerge = collider.Collisions<true>(vertBox.cview());
+
+  UnionFind<> uf(numVert);
+  for (int i = 0; i < mergeFromVert.size(); ++i) {
+    uf.unionXY(static_cast<int>(mergeFromVert[i]),
+               static_cast<int>(mergeToVert[i]));
+  }
+  for (int i = 0; i < toMerge.size(); ++i) {
+    uf.unionXY(openVerts[toMerge.Get(i, false)],
+               openVerts[toMerge.Get(i, true)]);
+  }
+
+  mergeToVert.clear();
+  mergeFromVert.clear();
+  for (int v = 0; v < numVert; ++v) {
+    const int mergeTo = uf.find(v);
+    if (mergeTo != v) {
+      mergeFromVert.push_back(v);
+      mergeToVert.push_back(mergeTo);
+    }
+  }
+
+  return true;
+}
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/manifold/src/subdivision.cpp b/thirdparty/manifold/src/manifold/src/subdivision.cpp
new file mode 100644
index 000000000000..3c5995bb516a
--- /dev/null
+++ b/thirdparty/manifold/src/manifold/src/subdivision.cpp
@@ -0,0 +1,792 @@
+// Copyright 2024 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "impl.h"
+#include "par.h"
+
+template <>
+struct std::hash<glm::ivec4> {
+  size_t operator()(const glm::ivec4& p) const {
+    return std::hash<int>()(p.x) ^ std::hash<int>()(p.y) ^
+           std::hash<int>()(p.z) ^ std::hash<int>()(p.w);
+  }
+};
+
+namespace {
+using namespace manifold;
+
+struct ReindexHalfedge {
+  VecView<int> half2Edge;
+  const VecView<Halfedge> halfedges;
+
+  void operator()(thrust::tuple<int, TmpEdge> in) {
+    const int edge = thrust::get<0>(in);
+    const int halfedge = thrust::get<1>(in).halfedgeIdx;
+
+    half2Edge[halfedge] = edge;
+    half2Edge[halfedges[halfedge].pairedHalfedge] = edge;
+  }
+};
+
+class Partition {
+ public:
+  // The cached partitions don't have idx - it's added to the copy returned
+  // from GetPartition that contains the mapping of the input divisions into the
+  // sorted divisions that are uniquely cached.
+  glm::ivec4 idx;
+  glm::ivec4 sortedDivisions;
+  Vec<glm::vec4> vertBary;
+  Vec<glm::ivec3> triVert;
+
+  int InteriorOffset() const {
+    return sortedDivisions[0] + sortedDivisions[1] + sortedDivisions[2] +
+           sortedDivisions[3];
+  }
+
+  int NumInterior() const { return vertBary.size() - InteriorOffset(); }
+
+  static Partition GetPartition(glm::ivec4 divisions) {
+    if (divisions[0] == 0) return Partition();  // skip wrong side of quad
+
+    glm::ivec4 sortedDiv = divisions;
+    glm::ivec4 triIdx = {0, 1, 2, 3};
+    if (divisions[3] == 0) {  // triangle
+      if (sortedDiv[2] > sortedDiv[1]) {
+        std::swap(sortedDiv[2], sortedDiv[1]);
+        std::swap(triIdx[2], triIdx[1]);
+      }
+      if (sortedDiv[1] > sortedDiv[0]) {
+        std::swap(sortedDiv[1], sortedDiv[0]);
+        std::swap(triIdx[1], triIdx[0]);
+        if (sortedDiv[2] > sortedDiv[1]) {
+          std::swap(sortedDiv[2], sortedDiv[1]);
+          std::swap(triIdx[2], triIdx[1]);
+        }
+      }
+    } else {  // quad
+      int minIdx = 0;
+      int min = divisions[minIdx];
+      int next = divisions[1];
+      for (const int i : {1, 2, 3}) {
+        const int n = divisions[(i + 1) % 4];
+        if (divisions[i] < min || (divisions[i] == min && n < next)) {
+          minIdx = i;
+          min = divisions[i];
+          next = n;
+        }
+      }
+      // Backwards (mirrored) quads get a separate cache key for now for
+      // simplicity, so there is no reversal necessary for quads when
+      // re-indexing.
+      glm::ivec4 tmp = sortedDiv;
+      for (const int i : {0, 1, 2, 3}) {
+        triIdx[i] = (i + minIdx) % 4;
+        sortedDiv[i] = tmp[triIdx[i]];
+      }
+    }
+
+    Partition partition = GetCachedPartition(sortedDiv);
+    partition.idx = triIdx;
+
+    return partition;
+  }
+
+  Vec<glm::ivec3> Reindex(glm::ivec4 tri, glm::ivec4 edgeOffsets,
+                          glm::bvec4 edgeFwd, int interiorOffset) const {
+    Vec<int> newVerts;
+    newVerts.reserve(vertBary.size());
+    glm::ivec4 triIdx = idx;
+    glm::ivec4 outTri = {0, 1, 2, 3};
+    if (tri[3] < 0 && idx[1] != Next3(idx[0])) {
+      triIdx = {idx[2], idx[0], idx[1], idx[3]};
+      edgeFwd = glm::not_(edgeFwd);
+      std::swap(outTri[0], outTri[1]);
+    }
+    for (const int i : {0, 1, 2, 3}) {
+      if (tri[triIdx[i]] >= 0) newVerts.push_back(tri[triIdx[i]]);
+    }
+    for (const int i : {0, 1, 2, 3}) {
+      const int n = sortedDivisions[i] - 1;
+      int offset = edgeOffsets[idx[i]] + (edgeFwd[idx[i]] ? 0 : n - 1);
+      for (int j = 0; j < n; ++j) {
+        newVerts.push_back(offset);
+        offset += edgeFwd[idx[i]] ? 1 : -1;
+      }
+    }
+    const int offset = interiorOffset - newVerts.size();
+    for (int i = newVerts.size(); i < vertBary.size(); ++i) {
+      newVerts.push_back(i + offset);
+    }
+
+    const int numTri = triVert.size();
+    Vec<glm::ivec3> newTriVert(numTri);
+    for_each_n(
+        autoPolicy(numTri), zip(newTriVert.begin(), triVert.begin()), numTri,
+        [&outTri, &newVerts](thrust::tuple<glm::ivec3&, glm::ivec3> inOut) {
+          for (const int j : {0, 1, 2}) {
+            thrust::get<0>(inOut)[outTri[j]] =
+                newVerts[thrust::get<1>(inOut)[j]];
+          }
+        });
+    return newTriVert;
+  }
+
+ private:
+  static inline auto cacheLock = std::mutex();
+  static inline auto cache =
+      std::unordered_map<glm::ivec4, std::unique_ptr<Partition>>();
+
+  // This triangulation is purely topological - it depends only on the number of
+  // divisions of the three sides of the triangle. This allows them to be cached
+  // and reused for similar triangles. The shape of the final surface is defined
+  // by the tangents and the barycentric coordinates of the new verts. For
+  // triangles, the input must be sorted: n[0] >= n[1] >= n[2] > 0.
+  static Partition GetCachedPartition(glm::ivec4 n) {
+    {
+      auto lockGuard = std::lock_guard<std::mutex>(cacheLock);
+      auto cached = cache.find(n);
+      if (cached != cache.end()) {
+        return *cached->second;
+      }
+    }
+    Partition partition;
+    partition.sortedDivisions = n;
+    if (n[3] > 0) {  // quad
+      partition.vertBary.push_back({1, 0, 0, 0});
+      partition.vertBary.push_back({0, 1, 0, 0});
+      partition.vertBary.push_back({0, 0, 1, 0});
+      partition.vertBary.push_back({0, 0, 0, 1});
+      glm::ivec4 edgeOffsets;
+      edgeOffsets[0] = 4;
+      for (const int i : {0, 1, 2, 3}) {
+        if (i > 0) {
+          edgeOffsets[i] = edgeOffsets[i - 1] + n[i - 1] - 1;
+        }
+        const glm::vec4 nextBary = partition.vertBary[(i + 1) % 4];
+        for (int j = 1; j < n[i]; ++j) {
+          partition.vertBary.push_back(
+              glm::mix(partition.vertBary[i], nextBary, (float)j / n[i]));
+        }
+      }
+      PartitionQuad(partition.triVert, partition.vertBary, {0, 1, 2, 3},
+                    edgeOffsets, n - 1, {true, true, true, true});
+    } else {  // tri
+      partition.vertBary.push_back({1, 0, 0, 0});
+      partition.vertBary.push_back({0, 1, 0, 0});
+      partition.vertBary.push_back({0, 0, 1, 0});
+      for (const int i : {0, 1, 2}) {
+        const glm::vec4 nextBary = partition.vertBary[(i + 1) % 3];
+        for (int j = 1; j < n[i]; ++j) {
+          partition.vertBary.push_back(
+              glm::mix(partition.vertBary[i], nextBary, (float)j / n[i]));
+        }
+      }
+      const glm::ivec3 edgeOffsets = {3, 3 + n[0] - 1, 3 + n[0] - 1 + n[1] - 1};
+
+      const float f = n[2] * n[2] + n[0] * n[0];
+      if (n[1] == 1) {
+        if (n[0] == 1) {
+          partition.triVert.push_back({0, 1, 2});
+        } else {
+          PartitionFan(partition.triVert, {0, 1, 2}, n[0] - 1, edgeOffsets[0]);
+        }
+      } else if (n[1] * n[1] >
+                 f - glm::sqrt(2.0f) * n[0] * n[2]) {  // acute-ish
+        partition.triVert.push_back({edgeOffsets[1] - 1, 1, edgeOffsets[1]});
+        PartitionQuad(partition.triVert, partition.vertBary,
+                      {edgeOffsets[1] - 1, edgeOffsets[1], 2, 0},
+                      {-1, edgeOffsets[1] + 1, edgeOffsets[2], edgeOffsets[0]},
+                      {0, n[1] - 2, n[2] - 1, n[0] - 2},
+                      {true, true, true, true});
+      } else {  // obtuse -> spit into two acute
+        // portion of n[0] under n[2]
+        const int ns =
+            glm::min(n[0] - 2, (int)glm::round((f - n[1] * n[1]) / (2 * n[0])));
+        // height from n[0]: nh <= n[2]
+        const int nh =
+            glm::max(1., glm::round(glm::sqrt(n[2] * n[2] - ns * ns)));
+
+        const int hOffset = partition.vertBary.size();
+        const glm::vec4 middleBary =
+            partition.vertBary[edgeOffsets[0] + ns - 1];
+        for (int j = 1; j < nh; ++j) {
+          partition.vertBary.push_back(
+              glm::mix(partition.vertBary[2], middleBary, (float)j / nh));
+        }
+
+        partition.triVert.push_back({edgeOffsets[1] - 1, 1, edgeOffsets[1]});
+        PartitionQuad(
+            partition.triVert, partition.vertBary,
+            {edgeOffsets[1] - 1, edgeOffsets[1], 2, edgeOffsets[0] + ns - 1},
+            {-1, edgeOffsets[1] + 1, hOffset, edgeOffsets[0] + ns},
+            {0, n[1] - 2, nh - 1, n[0] - ns - 2}, {true, true, true, true});
+
+        if (n[2] == 1) {
+          PartitionFan(partition.triVert, {0, edgeOffsets[0] + ns - 1, 2},
+                       ns - 1, edgeOffsets[0]);
+        } else {
+          if (ns == 1) {
+            partition.triVert.push_back({hOffset, 2, edgeOffsets[2]});
+            PartitionQuad(partition.triVert, partition.vertBary,
+                          {hOffset, edgeOffsets[2], 0, edgeOffsets[0]},
+                          {-1, edgeOffsets[2] + 1, -1, hOffset + nh - 2},
+                          {0, n[2] - 2, ns - 1, nh - 2},
+                          {true, true, true, false});
+          } else {
+            partition.triVert.push_back({hOffset - 1, 0, edgeOffsets[0]});
+            PartitionQuad(
+                partition.triVert, partition.vertBary,
+                {hOffset - 1, edgeOffsets[0], edgeOffsets[0] + ns - 1, 2},
+                {-1, edgeOffsets[0] + 1, hOffset + nh - 2, edgeOffsets[2]},
+                {0, ns - 2, nh - 1, n[2] - 2}, {true, true, false, true});
+          }
+        }
+      }
+    }
+
+    auto lockGuard = std::lock_guard<std::mutex>(cacheLock);
+    cache.insert({n, std::make_unique<Partition>(partition)});
+    return partition;
+  }
+
+  // Side 0 has added edges while sides 1 and 2 do not. Fan spreads from vert 2.
+  static void PartitionFan(Vec<glm::ivec3>& triVert, glm::ivec3 cornerVerts,
+                           int added, int edgeOffset) {
+    int last = cornerVerts[0];
+    for (int i = 0; i < added; ++i) {
+      const int next = edgeOffset + i;
+      triVert.push_back({last, next, cornerVerts[2]});
+      last = next;
+    }
+    triVert.push_back({last, cornerVerts[1], cornerVerts[2]});
+  }
+
+  // Partitions are parallel to the first edge unless two consecutive edgeAdded
+  // are zero, in which case a terminal triangulation is performed.
+  static void PartitionQuad(Vec<glm::ivec3>& triVert, Vec<glm::vec4>& vertBary,
+                            glm::ivec4 cornerVerts, glm::ivec4 edgeOffsets,
+                            glm::ivec4 edgeAdded, glm::bvec4 edgeFwd) {
+    auto GetEdgeVert = [&](int edge, int idx) {
+      return edgeOffsets[edge] + (edgeFwd[edge] ? 1 : -1) * idx;
+    };
+
+    ASSERT(glm::all(glm::greaterThanEqual(edgeAdded, glm::ivec4(0))), logicErr,
+           "negative divisions!");
+
+    int corner = -1;
+    int last = 3;
+    int maxEdge = -1;
+    for (const int i : {0, 1, 2, 3}) {
+      if (corner == -1 && edgeAdded[i] == 0 && edgeAdded[last] == 0) {
+        corner = i;
+      }
+      if (edgeAdded[i] > 0) {
+        maxEdge = maxEdge == -1 ? i : -2;
+      }
+      last = i;
+    }
+    if (corner >= 0) {  // terminate
+      if (maxEdge >= 0) {
+        glm::ivec4 edge = (glm::ivec4(0, 1, 2, 3) + maxEdge) % 4;
+        const int middle = edgeAdded[maxEdge] / 2;
+        triVert.push_back({cornerVerts[edge[2]], cornerVerts[edge[3]],
+                           GetEdgeVert(maxEdge, middle)});
+        int last = cornerVerts[edge[0]];
+        for (int i = 0; i <= middle; ++i) {
+          const int next = GetEdgeVert(maxEdge, i);
+          triVert.push_back({cornerVerts[edge[3]], last, next});
+          last = next;
+        }
+        last = cornerVerts[edge[1]];
+        for (int i = edgeAdded[maxEdge] - 1; i >= middle; --i) {
+          const int next = GetEdgeVert(maxEdge, i);
+          triVert.push_back({cornerVerts[edge[2]], next, last});
+          last = next;
+        }
+      } else {
+        int sideVert = cornerVerts[0];  // initial value is unused
+        for (const int j : {1, 2}) {
+          const int side = (corner + j) % 4;
+          if (j == 2 && edgeAdded[side] > 0) {
+            triVert.push_back(
+                {cornerVerts[side], GetEdgeVert(side, 0), sideVert});
+          } else {
+            sideVert = cornerVerts[side];
+          }
+          for (int i = 0; i < edgeAdded[side]; ++i) {
+            const int nextVert = GetEdgeVert(side, i);
+            triVert.push_back({cornerVerts[corner], sideVert, nextVert});
+            sideVert = nextVert;
+          }
+          if (j == 2 || edgeAdded[side] == 0) {
+            triVert.push_back({cornerVerts[corner], sideVert,
+                               cornerVerts[(corner + j + 1) % 4]});
+          }
+        }
+      }
+      return;
+    }
+    // recursively partition
+    const int partitions = 1 + glm::min(edgeAdded[1], edgeAdded[3]);
+    glm::ivec4 newCornerVerts = {cornerVerts[1], -1, -1, cornerVerts[0]};
+    glm::ivec4 newEdgeOffsets = {
+        edgeOffsets[1], -1, GetEdgeVert(3, edgeAdded[3] + 1), edgeOffsets[0]};
+    glm::ivec4 newEdgeAdded = {0, -1, 0, edgeAdded[0]};
+    glm::bvec4 newEdgeFwd = {edgeFwd[1], true, edgeFwd[3], edgeFwd[0]};
+
+    for (int i = 1; i < partitions; ++i) {
+      const int cornerOffset1 = (edgeAdded[1] * i) / partitions;
+      const int cornerOffset3 =
+          edgeAdded[3] - 1 - (edgeAdded[3] * i) / partitions;
+      const int nextOffset1 = GetEdgeVert(1, cornerOffset1 + 1);
+      const int nextOffset3 = GetEdgeVert(3, cornerOffset3 + 1);
+      const int added = glm::round(glm::mix(
+          (float)edgeAdded[0], (float)edgeAdded[2], (float)i / partitions));
+
+      newCornerVerts[1] = GetEdgeVert(1, cornerOffset1);
+      newCornerVerts[2] = GetEdgeVert(3, cornerOffset3);
+      newEdgeAdded[0] = std::abs(nextOffset1 - newEdgeOffsets[0]) - 1;
+      newEdgeAdded[1] = added;
+      newEdgeAdded[2] = std::abs(nextOffset3 - newEdgeOffsets[2]) - 1;
+      newEdgeOffsets[1] = vertBary.size();
+      newEdgeOffsets[2] = nextOffset3;
+
+      for (int j = 0; j < added; ++j) {
+        vertBary.push_back(glm::mix(vertBary[newCornerVerts[1]],
+                                    vertBary[newCornerVerts[2]],
+                                    (j + 1.0f) / (added + 1.0f)));
+      }
+
+      PartitionQuad(triVert, vertBary, newCornerVerts, newEdgeOffsets,
+                    newEdgeAdded, newEdgeFwd);
+
+      newCornerVerts[0] = newCornerVerts[1];
+      newCornerVerts[3] = newCornerVerts[2];
+      newEdgeAdded[3] = newEdgeAdded[1];
+      newEdgeOffsets[0] = nextOffset1;
+      newEdgeOffsets[3] = newEdgeOffsets[1] + newEdgeAdded[1] - 1;
+      newEdgeFwd[3] = false;
+    }
+
+    newCornerVerts[1] = cornerVerts[2];
+    newCornerVerts[2] = cornerVerts[3];
+    newEdgeOffsets[1] = edgeOffsets[2];
+    newEdgeAdded[0] =
+        edgeAdded[1] - std::abs(newEdgeOffsets[0] - edgeOffsets[1]);
+    newEdgeAdded[1] = edgeAdded[2];
+    newEdgeAdded[2] = std::abs(newEdgeOffsets[2] - edgeOffsets[3]) - 1;
+    newEdgeOffsets[2] = edgeOffsets[3];
+    newEdgeFwd[1] = edgeFwd[2];
+
+    PartitionQuad(triVert, vertBary, newCornerVerts, newEdgeOffsets,
+                  newEdgeAdded, newEdgeFwd);
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * Returns the tri side index (0-2) connected to the other side of this quad if
+ * this tri is part of a quad, or -1 otherwise.
+ */
+int Manifold::Impl::GetNeighbor(int tri) const {
+  int neighbor = -1;
+  for (const int i : {0, 1, 2}) {
+    if (IsMarkedInsideQuad(3 * tri + i)) {
+      neighbor = neighbor == -1 ? i : -2;
+    }
+  }
+  return neighbor;
+}
+
+/**
+ * For the given triangle index, returns either the three halfedge indices of
+ * that triangle and halfedges[3] = -1, or if the triangle is part of a quad, it
+ * returns those four indices. If the triangle is part of a quad and is not the
+ * lower of the two triangle indices, it returns all -1s.
+ */
+glm::ivec4 Manifold::Impl::GetHalfedges(int tri) const {
+  glm::ivec4 halfedges(-1);
+  for (const int i : {0, 1, 2}) {
+    halfedges[i] = 3 * tri + i;
+  }
+  const int neighbor = GetNeighbor(tri);
+  if (neighbor >= 0) {  // quad
+    const int pair = halfedge_[3 * tri + neighbor].pairedHalfedge;
+    if (pair / 3 < tri) {
+      return glm::ivec4(-1);  // only process lower tri index
+    }
+    glm::ivec2 otherHalf;
+    otherHalf[0] = NextHalfedge(pair);
+    otherHalf[1] = NextHalfedge(otherHalf[0]);
+    halfedges[neighbor] = otherHalf[0];
+    if (neighbor == 2) {
+      halfedges[3] = otherHalf[1];
+    } else if (neighbor == 1) {
+      halfedges[3] = halfedges[2];
+      halfedges[2] = otherHalf[1];
+    } else {
+      halfedges[3] = halfedges[2];
+      halfedges[2] = halfedges[1];
+      halfedges[1] = otherHalf[1];
+    }
+  }
+  return halfedges;
+}
+
+/**
+ * Returns the BaryIndices, which gives the tri and indices (0-3), such that
+ * GetHalfedges(val.tri)[val.start4] points back to this halfedge, and val.end4
+ * will point to the next one. This function handles this for both triangles and
+ * quads. Returns {-1, -1, -1} if the edge is the interior of a quad.
+ */
+Manifold::Impl::BaryIndices Manifold::Impl::GetIndices(int halfedge) const {
+  int tri = halfedge / 3;
+  int idx = halfedge % 3;
+  const int neighbor = GetNeighbor(tri);
+  if (idx == neighbor) {
+    return {-1, -1, -1};
+  }
+
+  if (neighbor < 0) {  // tri
+    return {tri, idx, Next3(idx)};
+  } else {  // quad
+    const int pair = halfedge_[3 * tri + neighbor].pairedHalfedge;
+    if (pair / 3 < tri) {
+      tri = pair / 3;
+      const int j = pair % 3;
+      idx = Next3(neighbor) == idx ? j : (j + 1) % 4;
+    } else if (idx > neighbor) {
+      ++idx;
+    }
+    return {tri, idx, (idx + 1) % 4};
+  }
+}
+
+/**
+ * Retained verts are part of several triangles, and it doesn't matter which one
+ * the vertBary refers to. Here, whichever is last will win and it's done on the
+ * CPU for simplicity for now. Using AtomicCAS on .tri should work for a GPU
+ * version if desired.
+ */
+void Manifold::Impl::FillRetainedVerts(Vec<Barycentric>& vertBary) const {
+  const int numTri = halfedge_.size() / 3;
+  for (int tri = 0; tri < numTri; ++tri) {
+    for (const int i : {0, 1, 2}) {
+      const BaryIndices indices = GetIndices(3 * tri + i);
+      if (indices.start4 < 0) continue;  // skip quad interiors
+      glm::vec4 uvw(0);
+      uvw[indices.start4] = 1;
+      vertBary[halfedge_[3 * tri + i].startVert] = {indices.tri, uvw};
+    }
+  }
+}
+
+/**
+ * Split each edge into n pieces as defined by calling the edgeDivisions
+ * function, and sub-triangulate each triangle accordingly. This function
+ * doesn't run Finish(), as that is expensive and it'll need to be run after
+ * the new vertices have moved, which is a likely scenario after refinement
+ * (smoothing).
+ */
+Vec<Barycentric> Manifold::Impl::Subdivide(
+    std::function<int(glm::vec3)> edgeDivisions) {
+  Vec<TmpEdge> edges = CreateTmpEdges(halfedge_);
+  const int numVert = NumVert();
+  const int numEdge = edges.size();
+  const int numTri = NumTri();
+  Vec<int> half2Edge(2 * numEdge);
+  auto policy = autoPolicy(numEdge);
+  for_each_n(policy, zip(countAt(0), edges.begin()), numEdge,
+             ReindexHalfedge({half2Edge, halfedge_}));
+
+  Vec<glm::ivec4> faceHalfedges(numTri);
+  for_each_n(policy, zip(faceHalfedges.begin(), countAt(0)), numTri,
+             [this](thrust::tuple<glm::ivec4&, int> inOut) {
+               glm::ivec4& halfedges = thrust::get<0>(inOut);
+               const int tri = thrust::get<1>(inOut);
+               halfedges = GetHalfedges(tri);
+             });
+
+  Vec<int> edgeAdded(numEdge);
+  for_each_n(policy, zip(edgeAdded.begin(), edges.cbegin()), numEdge,
+             [edgeDivisions, this](thrust::tuple<int&, TmpEdge> inOut) {
+               int& divisions = thrust::get<0>(inOut);
+               const TmpEdge edge = thrust::get<1>(inOut);
+               if (IsMarkedInsideQuad(edge.halfedgeIdx)) {
+                 divisions = 0;
+                 return;
+               }
+               const glm::vec3 vec =
+                   vertPos_[edge.first] - vertPos_[edge.second];
+               divisions = edgeDivisions(vec);
+             });
+
+  Vec<int> edgeOffset(numEdge);
+  exclusive_scan(policy, edgeAdded.begin(), edgeAdded.end(), edgeOffset.begin(),
+                 numVert);
+
+  Vec<Barycentric> vertBary(edgeOffset.back() + edgeAdded.back());
+  const int totalEdgeAdded = vertBary.size() - numVert;
+  FillRetainedVerts(vertBary);
+  for_each_n(policy, zip(edges.begin(), edgeAdded.begin(), edgeOffset.begin()),
+             numEdge, [this, &vertBary](thrust::tuple<TmpEdge, int, int> in) {
+               const TmpEdge edge = thrust::get<0>(in);
+               const int n = thrust::get<1>(in);
+               const int offset = thrust::get<2>(in);
+
+               const BaryIndices indices = GetIndices(edge.halfedgeIdx);
+               if (indices.tri < 0) {
+                 return;  // inside quad
+               }
+               const float frac = 1.0f / (n + 1);
+
+               for (int i = 0; i < n; ++i) {
+                 glm::vec4 uvw(0);
+                 uvw[indices.end4] = (i + 1) * frac;
+                 uvw[indices.start4] = 1 - uvw[indices.end4];
+                 vertBary[offset + i].uvw = uvw;
+                 vertBary[offset + i].tri = indices.tri;
+               }
+             });
+
+  std::vector<Partition> subTris(numTri);
+  for_each_n(policy, countAt(0), numTri,
+             [this, &subTris, &half2Edge, &edgeAdded, &faceHalfedges](int tri) {
+               const glm::ivec4 halfedges = faceHalfedges[tri];
+               glm::ivec4 divisions(0);
+               for (const int i : {0, 1, 2, 3}) {
+                 if (halfedges[i] >= 0) {
+                   divisions[i] = edgeAdded[half2Edge[halfedges[i]]] + 1;
+                 }
+               }
+               subTris[tri] = Partition::GetPartition(divisions);
+             });
+
+  Vec<int> triOffset(numTri);
+  auto numSubTris = thrust::make_transform_iterator(
+      subTris.begin(),
+      [](const Partition& part) { return part.triVert.size(); });
+  exclusive_scan(policy, numSubTris, numSubTris + numTri, triOffset.begin(), 0);
+
+  Vec<int> interiorOffset(numTri);
+  auto numInterior = thrust::make_transform_iterator(
+      subTris.begin(),
+      [](const Partition& part) { return part.NumInterior(); });
+  exclusive_scan(policy, numInterior, numInterior + numTri,
+                 interiorOffset.begin(), vertBary.size());
+
+  Vec<glm::ivec3> triVerts(triOffset.back() + subTris.back().triVert.size());
+  vertBary.resize(interiorOffset.back() + subTris.back().NumInterior());
+  Vec<TriRef> triRef(triVerts.size());
+  for_each_n(
+      policy, countAt(0), numTri,
+      [this, &triVerts, &triRef, &vertBary, &subTris, &edgeOffset, &half2Edge,
+       &triOffset, &interiorOffset, &faceHalfedges](int tri) {
+        const glm::ivec4 halfedges = faceHalfedges[tri];
+        if (halfedges[0] < 0) return;
+        glm::ivec4 tri3;
+        glm::ivec4 edgeOffsets;
+        glm::bvec4 edgeFwd;
+        for (const int i : {0, 1, 2, 3}) {
+          if (halfedges[i] < 0) {
+            tri3[i] = -1;
+            continue;
+          }
+          const Halfedge& halfedge = halfedge_[halfedges[i]];
+          tri3[i] = halfedge.startVert;
+          edgeOffsets[i] = edgeOffset[half2Edge[halfedges[i]]];
+          edgeFwd[i] = halfedge.IsForward();
+        }
+
+        Vec<glm::ivec3> newTris = subTris[tri].Reindex(
+            tri3, edgeOffsets, edgeFwd, interiorOffset[tri]);
+        copy(ExecutionPolicy::Seq, newTris.begin(), newTris.end(),
+             triVerts.begin() + triOffset[tri]);
+        auto start = triRef.begin() + triOffset[tri];
+        fill(ExecutionPolicy::Seq, start, start + newTris.size(),
+             meshRelation_.triRef[tri]);
+
+        const glm::ivec4 idx = subTris[tri].idx;
+        const glm::ivec4 vIdx =
+            halfedges[3] >= 0 || idx[1] == Next3(idx[0])
+                ? idx
+                : glm::ivec4(idx[2], idx[0], idx[1], idx[3]);
+        glm::ivec4 rIdx;
+        for (const int i : {0, 1, 2, 3}) {
+          rIdx[vIdx[i]] = i;
+        }
+
+        const auto& subBary = subTris[tri].vertBary;
+        transform(ExecutionPolicy::Seq,
+                  subBary.begin() + subTris[tri].InteriorOffset(),
+                  subBary.end(), vertBary.begin() + interiorOffset[tri],
+                  [tri, rIdx](glm::vec4 bary) {
+                    return Barycentric({tri,
+                                        {bary[rIdx[0]], bary[rIdx[1]],
+                                         bary[rIdx[2]], bary[rIdx[3]]}});
+                  });
+      });
+  meshRelation_.triRef = triRef;
+
+  Vec<glm::vec3> newVertPos(vertBary.size());
+  for_each_n(
+      policy, zip(newVertPos.begin(), vertBary.begin()), vertBary.size(),
+      [this, &faceHalfedges](thrust::tuple<glm::vec3&, Barycentric> inOut) {
+        glm::vec3& pos = thrust::get<0>(inOut);
+        const Barycentric bary = thrust::get<1>(inOut);
+
+        const glm::ivec4 halfedges = faceHalfedges[bary.tri];
+        if (halfedges[3] < 0) {
+          glm::mat3 triPos;
+          for (const int i : {0, 1, 2}) {
+            triPos[i] = vertPos_[halfedge_[halfedges[i]].startVert];
+          }
+          pos = triPos * glm::vec3(bary.uvw);
+        } else {
+          glm::mat4x3 quadPos;
+          for (const int i : {0, 1, 2, 3}) {
+            quadPos[i] = vertPos_[halfedge_[halfedges[i]].startVert];
+          }
+          pos = quadPos * bary.uvw;
+        }
+      });
+  vertPos_ = newVertPos;
+
+  faceNormal_.resize(0);
+
+  if (meshRelation_.numProp > 0) {
+    const int numPropVert = NumPropVert();
+    const int addedVerts = NumVert() - numVert;
+    const int propOffset = numPropVert - numVert;
+    Vec<float> prop(meshRelation_.numProp *
+                    (numPropVert + addedVerts + totalEdgeAdded));
+
+    // copy retained prop verts
+    copy(policy, meshRelation_.properties.begin(),
+         meshRelation_.properties.end(), prop.begin());
+
+    // copy interior prop verts and forward edge prop verts
+    for_each_n(
+        policy, zip(countAt(numPropVert), vertBary.begin() + numVert),
+        addedVerts,
+        [this, &prop, &faceHalfedges](thrust::tuple<int, Barycentric> in) {
+          const int vert = thrust::get<0>(in);
+          const Barycentric bary = thrust::get<1>(in);
+          const glm::ivec4 halfedges = faceHalfedges[bary.tri];
+          auto& rel = meshRelation_;
+
+          for (int p = 0; p < rel.numProp; ++p) {
+            if (halfedges[3] < 0) {
+              glm::vec3 triProp;
+              for (const int i : {0, 1, 2}) {
+                triProp[i] = rel.properties[rel.triProperties[bary.tri][i] *
+                                                rel.numProp +
+                                            p];
+              }
+              prop[vert * rel.numProp + p] =
+                  glm::dot(triProp, glm::vec3(bary.uvw));
+            } else {
+              glm::vec4 quadProp;
+              for (const int i : {0, 1, 2, 3}) {
+                const int tri = halfedges[i] / 3;
+                const int j = halfedges[i] % 3;
+                quadProp[i] =
+                    rel.properties[rel.triProperties[tri][j] * rel.numProp + p];
+              }
+              prop[vert * rel.numProp + p] = glm::dot(quadProp, bary.uvw);
+            }
+          }
+        });
+
+    // copy backward edge prop verts
+    for_each_n(
+        policy, zip(edges.begin(), edgeAdded.begin(), edgeOffset.begin()),
+        numEdge,
+        [this, &prop, propOffset,
+         addedVerts](thrust::tuple<TmpEdge, int, int> in) {
+          const TmpEdge edge = thrust::get<0>(in);
+          const int n = thrust::get<1>(in);
+          const int offset = thrust::get<2>(in) + propOffset + addedVerts;
+          auto& rel = meshRelation_;
+
+          const float frac = 1.0f / (n + 1);
+          const int halfedgeIdx = halfedge_[edge.halfedgeIdx].pairedHalfedge;
+          const int v0 = halfedgeIdx % 3;
+          const int tri = halfedgeIdx / 3;
+          const int prop0 = rel.triProperties[tri][v0];
+          const int prop1 = rel.triProperties[tri][Next3(v0)];
+          for (int i = 0; i < n; ++i) {
+            for (int p = 0; p < rel.numProp; ++p) {
+              prop[(offset + i) * rel.numProp + p] = glm::mix(
+                  rel.properties[prop0 * rel.numProp + p],
+                  rel.properties[prop1 * rel.numProp + p], (i + 1) * frac);
+            }
+          }
+        });
+
+    Vec<glm::ivec3> triProp(triVerts.size());
+    for_each_n(policy, countAt(0), numTri,
+               [this, &triProp, &subTris, &edgeOffset, &half2Edge, &triOffset,
+                &interiorOffset, &faceHalfedges, propOffset,
+                addedVerts](const int tri) {
+                 const glm::ivec4 halfedges = faceHalfedges[tri];
+                 if (halfedges[0] < 0) return;
+
+                 auto& rel = meshRelation_;
+                 glm::ivec4 tri3;
+                 glm::ivec4 edgeOffsets;
+                 glm::bvec4 edgeFwd(true);
+                 for (const int i : {0, 1, 2, 3}) {
+                   if (halfedges[i] < 0) {
+                     tri3[i] = -1;
+                     continue;
+                   }
+                   const int thisTri = halfedges[i] / 3;
+                   const int j = halfedges[i] % 3;
+                   const Halfedge& halfedge = halfedge_[halfedges[i]];
+                   tri3[i] = rel.triProperties[thisTri][j];
+                   edgeOffsets[i] = edgeOffset[half2Edge[halfedges[i]]];
+                   if (!halfedge.IsForward()) {
+                     const int pairTri = halfedge.pairedHalfedge / 3;
+                     const int k = halfedge.pairedHalfedge % 3;
+                     if (rel.triProperties[pairTri][k] !=
+                             rel.triProperties[thisTri][Next3(j)] ||
+                         rel.triProperties[pairTri][Next3(k)] !=
+                             rel.triProperties[thisTri][j]) {
+                       edgeOffsets[i] += addedVerts;
+                     } else {
+                       edgeFwd[i] = false;
+                     }
+                   }
+                 }
+
+                 Vec<glm::ivec3> newTris = subTris[tri].Reindex(
+                     tri3, edgeOffsets + propOffset, edgeFwd,
+                     interiorOffset[tri] + propOffset);
+                 copy(ExecutionPolicy::Seq, newTris.begin(), newTris.end(),
+                      triProp.begin() + triOffset[tri]);
+               });
+
+    meshRelation_.properties = prop;
+    meshRelation_.triProperties = triProp;
+  }
+
+  CreateHalfedges(triVerts);
+
+  return vertBary;
+}
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/polygon/include/polygon.h b/thirdparty/manifold/src/polygon/include/polygon.h
new file mode 100644
index 000000000000..8e5f11f0bd0c
--- /dev/null
+++ b/thirdparty/manifold/src/polygon/include/polygon.h
@@ -0,0 +1,51 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <functional>
+
+#include "public.h"
+
+namespace manifold {
+
+/** @addtogroup Private
+ *  @{
+ */
+
+/**
+ * Polygon vertex.
+ */
+struct PolyVert {
+  /// X-Y position
+  glm::vec2 pos;
+  /// ID or index into another vertex vector
+  int idx;
+};
+
+using SimplePolygonIdx = std::vector<PolyVert>;
+using PolygonsIdx = std::vector<SimplePolygonIdx>;
+
+std::vector<glm::ivec3> TriangulateIdx(const PolygonsIdx &polys,
+                                       float precision = -1);
+/** @} */
+
+/** @ingroup Connections
+ *  @{
+ */
+std::vector<glm::ivec3> Triangulate(const Polygons &polygons,
+                                    float precision = -1);
+
+ExecutionParams &PolygonParams();
+/** @} */
+}  // namespace manifold
\ No newline at end of file
diff --git a/thirdparty/manifold/src/polygon/src/polygon.cpp b/thirdparty/manifold/src/polygon/src/polygon.cpp
new file mode 100644
index 000000000000..a8df49bf419c
--- /dev/null
+++ b/thirdparty/manifold/src/polygon/src/polygon.cpp
@@ -0,0 +1,877 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "polygon.h"
+
+#include <algorithm>
+#include <map>
+#include <set>
+
+#include "optional_assert.h"
+#include "utils.h"
+
+namespace {
+using namespace manifold;
+
+static ExecutionParams params;
+
+constexpr float kBest = -std::numeric_limits<float>::infinity();
+
+// it seems that MSVC cannot optimize glm::determinant(glm::mat2(a, b))
+constexpr float determinant2x2(glm::vec2 a, glm::vec2 b) {
+  return a.x * b.y - a.y * b.x;
+}
+
+#ifdef MANIFOLD_DEBUG
+struct PolyEdge {
+  int startVert, endVert;
+};
+
+std::vector<PolyEdge> Polygons2Edges(const PolygonsIdx &polys) {
+  std::vector<PolyEdge> halfedges;
+  for (const auto &poly : polys) {
+    for (int i = 1; i < poly.size(); ++i) {
+      halfedges.push_back({poly[i - 1].idx, poly[i].idx});
+    }
+    halfedges.push_back({poly.back().idx, poly[0].idx});
+  }
+  return halfedges;
+}
+
+std::vector<PolyEdge> Triangles2Edges(
+    const std::vector<glm::ivec3> &triangles) {
+  std::vector<PolyEdge> halfedges;
+  halfedges.reserve(triangles.size() * 3);
+  for (const glm::ivec3 &tri : triangles) {
+    halfedges.push_back({tri[0], tri[1]});
+    halfedges.push_back({tri[1], tri[2]});
+    halfedges.push_back({tri[2], tri[0]});
+  }
+  return halfedges;
+}
+
+void CheckTopology(const std::vector<PolyEdge> &halfedges) {
+  ASSERT(halfedges.size() % 2 == 0, topologyErr, "Odd number of halfedges.");
+  size_t n_edges = halfedges.size() / 2;
+  std::vector<PolyEdge> forward(halfedges.size()), backward(halfedges.size());
+
+  auto end = std::copy_if(halfedges.begin(), halfedges.end(), forward.begin(),
+                          [](PolyEdge e) { return e.endVert > e.startVert; });
+  ASSERT(std::distance(forward.begin(), end) == n_edges, topologyErr,
+         "Half of halfedges should be forward.");
+  forward.resize(n_edges);
+
+  end = std::copy_if(halfedges.begin(), halfedges.end(), backward.begin(),
+                     [](PolyEdge e) { return e.endVert < e.startVert; });
+  ASSERT(std::distance(backward.begin(), end) == n_edges, topologyErr,
+         "Half of halfedges should be backward.");
+  backward.resize(n_edges);
+
+  std::for_each(backward.begin(), backward.end(),
+                [](PolyEdge &e) { std::swap(e.startVert, e.endVert); });
+  auto cmp = [](const PolyEdge &a, const PolyEdge &b) {
+    return a.startVert < b.startVert ||
+           (a.startVert == b.startVert && a.endVert < b.endVert);
+  };
+  std::stable_sort(forward.begin(), forward.end(), cmp);
+  std::stable_sort(backward.begin(), backward.end(), cmp);
+  for (int i = 0; i < n_edges; ++i) {
+    ASSERT(forward[i].startVert == backward[i].startVert &&
+               forward[i].endVert == backward[i].endVert,
+           topologyErr, "Not manifold.");
+  }
+}
+
+void CheckTopology(const std::vector<glm::ivec3> &triangles,
+                   const PolygonsIdx &polys) {
+  std::vector<PolyEdge> halfedges = Triangles2Edges(triangles);
+  std::vector<PolyEdge> openEdges = Polygons2Edges(polys);
+  for (PolyEdge e : openEdges) {
+    halfedges.push_back({e.endVert, e.startVert});
+  }
+  CheckTopology(halfedges);
+}
+
+void CheckGeometry(const std::vector<glm::ivec3> &triangles,
+                   const PolygonsIdx &polys, float precision) {
+  std::unordered_map<int, glm::vec2> vertPos;
+  for (const auto &poly : polys) {
+    for (int i = 0; i < poly.size(); ++i) {
+      vertPos[poly[i].idx] = poly[i].pos;
+    }
+  }
+  ASSERT(std::all_of(triangles.begin(), triangles.end(),
+                     [&vertPos, precision](const glm::ivec3 &tri) {
+                       return CCW(vertPos[tri[0]], vertPos[tri[1]],
+                                  vertPos[tri[2]], precision) >= 0;
+                     }),
+         geometryErr, "triangulation is not entirely CCW!");
+}
+
+void Dump(const PolygonsIdx &polys) {
+  for (auto poly : polys) {
+    std::cout << "polys.push_back({" << std::setprecision(9) << std::endl;
+    for (auto v : poly) {
+      std::cout << "    {" << v.pos.x << ", " << v.pos.y << "},  //"
+                << std::endl;
+    }
+    std::cout << "});" << std::endl;
+  }
+  for (auto poly : polys) {
+    std::cout << "show(array([" << std::endl;
+    for (auto v : poly) {
+      std::cout << "  [" << v.pos.x << ", " << v.pos.y << "]," << std::endl;
+    }
+    std::cout << "]))" << std::endl;
+  }
+}
+
+void PrintFailure(const std::exception &e, const PolygonsIdx &polys,
+                  std::vector<glm::ivec3> &triangles, float precision) {
+  std::cout << "-----------------------------------" << std::endl;
+  std::cout << "Triangulation failed! Precision = " << precision << std::endl;
+  std::cout << e.what() << std::endl;
+  Dump(polys);
+  std::cout << "produced this triangulation:" << std::endl;
+  for (int j = 0; j < triangles.size(); ++j) {
+    std::cout << triangles[j][0] << ", " << triangles[j][1] << ", "
+              << triangles[j][2] << std::endl;
+  }
+}
+
+#define PRINT(msg) \
+  if (params.verbose) std::cout << msg << std::endl;
+#else
+#define PRINT(msg)
+#endif
+
+/**
+ * Ear-clipping triangulator based on David Eberly's approach from Geometric
+ * Tools, but adjusted to handle epsilon-valid polygons, and including a
+ * fallback that ensures a manifold triangulation even for overlapping polygons.
+ * This is an O(n^2) algorithm, but hopefully this is not a big problem as the
+ * number of edges in a given polygon is generally much less than the number of
+ * triangles in a mesh, and relatively few faces even need triangulation.
+ *
+ * The main adjustments for robustness involve clipping the sharpest ears first
+ * (a known technique to get higher triangle quality), and doing an exhaustive
+ * search to determine ear convexity exactly if the first geometric result is
+ * within precision.
+ */
+
+class EarClip {
+ public:
+  EarClip(const PolygonsIdx &polys, float precision) : precision_(precision) {
+    ZoneScoped;
+
+    int numVert = 0;
+    for (const SimplePolygonIdx &poly : polys) {
+      numVert += poly.size();
+    }
+    polygon_.reserve(numVert + 2 * polys.size());
+
+    std::vector<VertItr> starts = Initialize(polys);
+
+    for (VertItr v = polygon_.begin(); v != polygon_.end(); ++v) {
+      ClipIfDegenerate(v);
+    }
+
+    for (const VertItr first : starts) {
+      FindStart(first);
+    }
+  }
+
+  std::vector<glm::ivec3> Triangulate() {
+    ZoneScoped;
+
+    for (const VertItr start : holes_) {
+      CutKeyhole(start);
+    }
+
+    for (const VertItr start : simples_) {
+      TriangulatePoly(start);
+    }
+
+    return triangles_;
+  }
+
+  float GetPrecision() const { return precision_; }
+
+ private:
+  struct Vert;
+  typedef std::vector<Vert>::iterator VertItr;
+  struct MaxX {
+    bool operator()(const VertItr &a, const VertItr &b) const {
+      return a->pos.x > b->pos.x;
+    }
+  };
+  struct MinCost {
+    bool operator()(const VertItr &a, const VertItr &b) const {
+      return a->cost < b->cost;
+    }
+  };
+  typedef std::set<VertItr, MinCost>::iterator qItr;
+
+  // The flat list where all the Verts are stored. Not used much for traversal.
+  std::vector<Vert> polygon_;
+  // The set of right-most starting points, one for each negative-area contour.
+  std::multiset<VertItr, MaxX> holes_;
+  // The set of starting points, one for each positive-area contour.
+  std::vector<VertItr> outers_;
+  // The set of starting points, one for each simple polygon.
+  std::vector<VertItr> simples_;
+  // Maps each hole (by way of starting point) to its bounding box.
+  std::map<VertItr, Rect> hole2BBox_;
+  // A priority queue of valid ears - the multiset allows them to be updated.
+  std::multiset<VertItr, MinCost> earsQueue_;
+  // The output triangulation.
+  std::vector<glm::ivec3> triangles_;
+  // Working precision: max of float error and input value.
+  float precision_;
+
+  // A circularly-linked list representing the polygon(s) that still need to be
+  // triangulated. This gets smaller as ears are clipped until it degenerates to
+  // two points and terminates.
+  struct Vert {
+    int mesh_idx;
+    float cost;
+    qItr ear;
+    glm::vec2 pos, rightDir;
+    VertItr left, right;
+
+    // Shorter than half of precision, to be conservative so that it doesn't
+    // cause CW triangles that exceed precision due to rounding error.
+    bool IsShort(float precision) const {
+      const glm::vec2 edge = right->pos - pos;
+      return glm::dot(edge, edge) * 4 < precision * precision;
+    }
+
+    // Like CCW, returns 1 if v is on the inside of the angle formed at this
+    // vert, -1 on the outside, and 0 if it's within precision of the boundary.
+    // Ensure v is more than precision from pos, as case this will not return 0.
+    int Interior(glm::vec2 v, float precision) const {
+      const glm::vec2 diff = v - pos;
+      if (glm::dot(diff, diff) < precision * precision) {
+        return 0;
+      }
+      return CCW(pos, left->pos, right->pos, precision) +
+             CCW(pos, right->pos, v, precision) +
+             CCW(pos, v, left->pos, precision);
+    }
+
+    // Returns true if Vert is on the inside of the edge that goes from tail to
+    // tail->right. This will walk the edges if necessary until a clear answer
+    // is found (beyond precision). If toLeft is true, this Vert will walk its
+    // edges to the left. This should be chosen so that the edges walk in the
+    // same general direction - tail always walks to the right.
+    bool InsideEdge(VertItr tail, float precision, bool toLeft) const {
+      const float p2 = precision * precision;
+      VertItr nextL = left->right;
+      VertItr nextR = tail->right;
+      VertItr center = tail;
+      VertItr last = center;
+
+      while (nextL != nextR && tail != nextR &&
+             nextL != (toLeft ? right : left)) {
+        const glm::vec2 edgeL = nextL->pos - center->pos;
+        const float l2 = glm::dot(edgeL, edgeL);
+        if (l2 <= p2) {
+          nextL = toLeft ? nextL->left : nextL->right;
+          continue;
+        }
+
+        const glm::vec2 edgeR = nextR->pos - center->pos;
+        const float r2 = glm::dot(edgeR, edgeR);
+        if (r2 <= p2) {
+          nextR = nextR->right;
+          continue;
+        }
+
+        const glm::vec2 vecLR = nextR->pos - nextL->pos;
+        const float lr2 = glm::dot(vecLR, vecLR);
+        if (lr2 <= p2) {
+          last = center;
+          center = nextL;
+          nextL = toLeft ? nextL->left : nextL->right;
+          if (nextL == nextR) break;
+          nextR = nextR->right;
+          continue;
+        }
+
+        int convexity = CCW(nextL->pos, center->pos, nextR->pos, precision);
+        if (center != last) {
+          convexity += CCW(last->pos, center->pos, nextL->pos, precision) +
+                       CCW(nextR->pos, center->pos, last->pos, precision);
+        }
+        if (convexity != 0) return convexity > 0;
+
+        if (l2 < r2) {
+          center = nextL;
+          nextL = toLeft ? nextL->left : nextL->right;
+        } else {
+          center = nextR;
+          nextR = nextR->right;
+        }
+        last = center;
+      }
+      // The whole polygon is degenerate - consider this to be convex.
+      return true;
+    }
+
+    // A major key to robustness is to only clip convex ears, but this is
+    // difficult to determine when an edge is folded back on itself. This
+    // function walks down the kinks in a degenerate portion of a polygon until
+    // it finds a clear geometric result. In the vast majority of cases the loop
+    // will only need one or two iterations.
+    bool IsConvex(float precision) const {
+      return left->InsideEdge(left->right, precision, true);
+    }
+
+    // This function is the core of finding a proper place to keyhole. It runs
+    // on this Vert, which represents the edge from this to right. It returns
+    // an iterator to the vert to connect to (either this or right) and a bool
+    // denoting if the edge is a valid option for a keyhole (must be upwards and
+    // cross the start.y-value).
+    //
+    // If the edge terminates within the precision band, it checks the next edge
+    // to ensure validity. No while loop is necessary because short edges have
+    // already been removed. The onTop value is 1 if the start.y-value is at the
+    // top of the polygon's bounding box, -1 if it's at the bottom, and 0
+    // otherwise. This allows proper handling of horizontal edges.
+    std::pair<VertItr, bool> InterpY2X(glm::vec2 start, int onTop,
+                                       float precision) const {
+      const auto none = std::make_pair(left, false);
+      if (pos.y < start.y && right->pos.y >= start.y) {
+        return std::make_pair(left->right, true);
+      } else if (pos.x > start.x - precision && pos.y > start.y - precision &&
+                 pos.y < start.y + precision &&
+                 Interior(start, precision) >= 0) {
+        if (onTop > 0 && left->pos.x < pos.x &&
+            left->pos.y > start.y - precision) {
+          return none;
+        }
+        if (onTop < 0 && right->pos.x < pos.x &&
+            right->pos.y < start.y + precision) {
+          return none;
+        }
+        const VertItr p = pos.x < right->pos.x ? right : left->right;
+        return std::make_pair(p, true);
+      }
+      // Edge does not cross start.y going up
+      return none;
+    }
+
+    // This finds the cost of this vert relative to one of the two closed sides
+    // of the ear. Points are valid even when they touch, so long as their edge
+    // goes to the outside. No need to check the other side, since all verts are
+    // processed in the EarCost loop.
+    float SignedDist(VertItr v, glm::vec2 unit, float precision) const {
+      float d = determinant2x2(unit, v->pos - pos);
+      if (std::abs(d) < precision) {
+        d = glm::max(d, determinant2x2(unit, v->right->pos - pos));
+        d = glm::max(d, determinant2x2(unit, v->left->pos - pos));
+      }
+      return d;
+    }
+
+    // Find the cost of Vert v within this ear, where openSide is the unit
+    // vector from Verts right to left - passed in for reuse.
+    float Cost(VertItr v, glm::vec2 openSide, float precision) const {
+      float cost = glm::min(SignedDist(v, rightDir, precision),
+                            SignedDist(v, left->rightDir, precision));
+
+      const float openCost = determinant2x2(openSide, v->pos - right->pos);
+      return glm::min(cost, openCost);
+    }
+
+    // For verts outside the ear, apply a cost based on the Delaunay condition
+    // to aid in prioritization and produce cleaner triangulations. This doesn't
+    // affect robustness, but may be adjusted to improve output.
+    float DelaunayCost(glm::vec2 diff, float scale, float precision) const {
+      return -precision - scale * glm::dot(diff, diff);
+    }
+
+    // This is the O(n^2) part of the algorithm, checking this ear against every
+    // Vert to ensure none are inside. It may be possible to improve performance
+    // by using the Collider to get it down to nlogn or doing some
+    // parallelization, but that may be more trouble than it's worth.
+    //
+    // Think of a cost as vaguely a distance metric - 0 is right on the edge of
+    // being invalid. cost > precision is definitely invalid. Cost < -precision
+    // is definitely valid, so all improvement costs are designed to always give
+    // values < -precision so they will never affect validity. The first
+    // totalCost is designed to give priority to sharper angles. Any cost < (-1
+    // - precision) has satisfied the Delaunay condition.
+    float EarCost(float precision) const {
+      glm::vec2 openSide = left->pos - right->pos;
+      const glm::vec2 center = 0.5f * (left->pos + right->pos);
+      const float scale = 4 / glm::dot(openSide, openSide);
+      openSide = glm::normalize(openSide);
+
+      float totalCost = glm::dot(left->rightDir, rightDir) - 1 - precision;
+      if (CCW(pos, left->pos, right->pos, precision) == 0) {
+        // Clip folded ears first
+        return totalCost < -1 ? kBest : 0;
+      }
+      VertItr test = right->right;
+      auto lid = left->mesh_idx;
+      auto rid = right->mesh_idx;
+      while (test != left) {
+        if (test->mesh_idx != mesh_idx && test->mesh_idx != lid &&
+            test->mesh_idx != rid) {  // Skip duplicated verts
+          float cost = Cost(test, openSide, precision);
+          if (cost < -precision) {
+            cost = DelaunayCost(test->pos - center, scale, precision);
+          }
+          totalCost = glm::max(totalCost, cost);
+        }
+
+        test = test->right;
+      }
+      return totalCost;
+    }
+
+    void PrintVert() const {
+#ifdef MANIFOLD_DEBUG
+      if (!params.verbose) return;
+      std::cout << "vert: " << mesh_idx << ", left: " << left->mesh_idx
+                << ", right: " << right->mesh_idx << ", cost: " << cost
+                << std::endl;
+#endif
+    }
+  };
+
+  glm::vec2 SafeNormalize(glm::vec2 v) const {
+    glm::vec2 n = glm::normalize(v);
+    return glm::isfinite(n.x) ? n : glm::vec2(0, 0);
+  }
+
+  // This function and JoinPolygons are the only functions that affect the
+  // circular list data structure. This helps ensure it remains circular.
+  void Link(VertItr left, VertItr right) const {
+    left->right = right;
+    right->left = left;
+    left->rightDir = SafeNormalize(right->pos - left->pos);
+  }
+
+  // When an ear vert is clipped, its neighbors get linked, so they get unlinked
+  // from it, but it is still linked to them.
+  bool Clipped(VertItr v) const { return v->right->left != v; }
+
+  // Apply func to each un-clipped vert in a polygon and return an un-clipped
+  // vert.
+  VertItr Loop(VertItr first, std::function<void(VertItr)> func) {
+    VertItr v = first;
+    do {
+      if (Clipped(v)) {
+        // Update first to an un-clipped vert so we will return to it instead
+        // of infinite-looping.
+        first = v->right->left;
+        if (!Clipped(first)) {
+          v = first;
+          if (v->right == v->left) {
+            return polygon_.end();
+          }
+          func(v);
+        }
+      } else {
+        if (v->right == v->left) {
+          return polygon_.end();
+        }
+        func(v);
+      }
+      v = v->right;
+    } while (v != first);
+    return v;
+  }
+
+  // Remove this vert from the circular list and output a corresponding
+  // triangle.
+  void ClipEar(VertItr ear) {
+    Link(ear->left, ear->right);
+    if (ear->left->mesh_idx != ear->mesh_idx &&
+        ear->mesh_idx != ear->right->mesh_idx &&
+        ear->right->mesh_idx != ear->left->mesh_idx) {
+      // Filter out topological degenerates, which can form in bad
+      // triangulations of polygons with holes, due to vert duplication.
+      triangles_.push_back(
+          {ear->left->mesh_idx, ear->mesh_idx, ear->right->mesh_idx});
+#ifdef MANIFOLD_DEBUG
+      if (params.verbose) {
+        std::cout << "output tri: " << ear->mesh_idx << ", "
+                  << ear->right->mesh_idx << ", " << ear->left->mesh_idx
+                  << std::endl;
+      }
+#endif
+    } else {
+      PRINT("Topological degenerate!");
+    }
+  }
+
+  // If an ear will make a degenerate triangle, clip it early to avoid
+  // difficulty in key-holing. This function is recursive, as the process of
+  // clipping may cause the neighbors to degenerate. Reflex degenerates *must
+  // not* be clipped, unless they have a short edge.
+  void ClipIfDegenerate(VertItr ear) {
+    if (Clipped(ear)) {
+      return;
+    }
+    if (ear->left == ear->right) {
+      return;
+    }
+    if (ear->IsShort(precision_) ||
+        (CCW(ear->left->pos, ear->pos, ear->right->pos, precision_) == 0 &&
+         glm::dot(ear->left->pos - ear->pos, ear->right->pos - ear->pos) > 0 &&
+         ear->IsConvex(precision_))) {
+      ClipEar(ear);
+      ClipIfDegenerate(ear->left);
+      ClipIfDegenerate(ear->right);
+    }
+  }
+
+  // Build the circular list polygon structures.
+  std::vector<VertItr> Initialize(const PolygonsIdx &polys) {
+    std::vector<VertItr> starts;
+    float bound = 0;
+    for (const SimplePolygonIdx &poly : polys) {
+      auto vert = poly.begin();
+      polygon_.push_back({vert->idx, 0.0f, earsQueue_.end(), vert->pos});
+      const VertItr first = std::prev(polygon_.end());
+
+      bound = glm::max(
+          bound, glm::max(std::abs(first->pos.x), std::abs(first->pos.y)));
+      VertItr last = first;
+      // This is not the real rightmost start, but just an arbitrary vert for
+      // now to identify each polygon.
+      starts.push_back(first);
+
+      for (++vert; vert != poly.end(); ++vert) {
+        bound = glm::max(
+            bound, glm::max(std::abs(vert->pos.x), std::abs(vert->pos.y)));
+
+        polygon_.push_back({vert->idx, 0.0f, earsQueue_.end(), vert->pos});
+        VertItr next = std::prev(polygon_.end());
+
+        Link(last, next);
+        last = next;
+      }
+      Link(last, first);
+    }
+
+    if (precision_ < 0) precision_ = bound * kTolerance;
+
+    // Slightly more than enough, since each hole can cause two extra triangles.
+    triangles_.reserve(polygon_.size() + 2 * starts.size());
+    return starts;
+  }
+
+  // Find the actual rightmost starts after degenerate removal. Also calculate
+  // the polygon bounding boxes.
+  void FindStart(VertItr first) {
+    const glm::vec2 origin = first->pos;
+
+    VertItr start = first;
+    float maxX = -std::numeric_limits<float>::infinity();
+    Rect bBox;
+    // Kahan summation
+    double area = 0;
+    double areaCompensation = 0;
+
+    auto AddPoint = [&](VertItr v) {
+      bBox.Union(v->pos);
+      const double area1 =
+          determinant2x2(v->pos - origin, v->right->pos - origin);
+      const double t1 = area + area1;
+      areaCompensation += (area - t1) + area1;
+      area = t1;
+
+      if (!v->IsConvex(precision_) && v->pos.x > maxX) {
+        maxX = v->pos.x;
+        start = v;
+      }
+    };
+
+    if (Loop(first, AddPoint) == polygon_.end()) {
+      // No polygon left if all ears were degenerate and already clipped.
+      return;
+    }
+
+    area += areaCompensation;
+    const glm::vec2 size = bBox.Size();
+    const float minArea = precision_ * glm::max(size.x, size.y);
+
+    if (glm::isfinite(maxX) && area < -minArea) {
+      holes_.insert(start);
+      hole2BBox_.insert({start, bBox});
+    } else {
+      simples_.push_back(start);
+      if (area > minArea) {
+        outers_.push_back(start);
+      }
+    }
+  }
+
+  // All holes must be key-holed (attached to an outer polygon) before ear
+  // clipping can commence. Instead of relying on sorting, which may be
+  // incorrect due to precision, we check for polygon edges both ahead and
+  // behind to ensure all valid options are found.
+  void CutKeyhole(const VertItr start) {
+    const Rect bBox = hole2BBox_[start];
+    const int onTop = start->pos.y >= bBox.max.y - precision_   ? 1
+                      : start->pos.y <= bBox.min.y + precision_ ? -1
+                                                                : 0;
+    VertItr connector = polygon_.end();
+
+    auto CheckEdge = [&](VertItr edge) {
+      const std::pair<VertItr, bool> pair =
+          edge->InterpY2X(start->pos, onTop, precision_);
+      if (pair.second && start->InsideEdge(pair.first, precision_, true) &&
+          (connector == polygon_.end() ||
+           (connector->pos.y < pair.first->pos.y
+                ? pair.first->InsideEdge(connector, precision_, false)
+                : !connector->InsideEdge(pair.first, precision_, false)))) {
+        connector = pair.first;
+      }
+    };
+
+    for (const VertItr first : outers_) {
+      Loop(first, CheckEdge);
+    }
+
+    if (connector == polygon_.end()) {
+      PRINT("hole did not find an outer contour!");
+      simples_.push_back(start);
+      return;
+    }
+
+    connector = FindCloserBridge(start, connector, onTop);
+
+    JoinPolygons(start, connector);
+
+#ifdef MANIFOLD_DEBUG
+    if (params.verbose) {
+      std::cout << "connected " << start->mesh_idx << " to "
+                << connector->mesh_idx << std::endl;
+    }
+#endif
+  }
+
+  // This converts the initial guess for the keyhole location into the final one
+  // and returns it. It does so by finding any reflex verts inside the triangle
+  // containing the best connection and the initial horizontal line.
+  VertItr FindCloserBridge(VertItr start, VertItr edge, int onTop) {
+    VertItr best = edge->pos.x > edge->right->pos.x ? edge : edge->right;
+    const float maxX = best->pos.x;
+    const float above = best->pos.y > start->pos.y ? 1 : -1;
+
+    auto CheckVert = [&](VertItr vert) {
+      const float inside = above * CCW(start->pos, vert->pos, best->pos, 0);
+      if (vert->pos.x > start->pos.x - precision_ &&
+          vert->pos.x < maxX + precision_ &&
+          vert->pos.y * above > start->pos.y * above - precision_ &&
+          (inside > 0 || (inside == 0 && vert->pos.x < best->pos.x)) &&
+          vert->InsideEdge(edge, precision_, true) &&
+          !vert->IsConvex(precision_)) {
+        if (vert->pos.y > start->pos.y - precision_ &&
+            vert->pos.y < start->pos.y + precision_) {
+          if (onTop > 0 && vert->left->pos.x < vert->pos.x &&
+              vert->left->pos.y > start->pos.y - precision_) {
+            return;
+          }
+          if (onTop < 0 && vert->right->pos.x < vert->pos.x &&
+              vert->right->pos.y < start->pos.y + precision_) {
+            return;
+          }
+        }
+        best = vert;
+      }
+    };
+
+    for (const VertItr first : outers_) {
+      Loop(first, CheckVert);
+    }
+
+    return best;
+  }
+
+  // Creates a keyhole between the start vert of a hole and the connector vert
+  // of an outer polygon. To do this, both verts are duplicated and reattached.
+  // This process may create degenerate ears, so these are clipped if necessary
+  // to keep from confusing subsequent key-holing operations.
+  void JoinPolygons(VertItr start, VertItr connector) {
+    polygon_.push_back(*start);
+    const VertItr newStart = std::prev(polygon_.end());
+    polygon_.push_back(*connector);
+    const VertItr newConnector = std::prev(polygon_.end());
+
+    start->right->left = newStart;
+    connector->left->right = newConnector;
+    Link(start, connector);
+    Link(newConnector, newStart);
+
+    ClipIfDegenerate(start);
+    ClipIfDegenerate(newStart);
+    ClipIfDegenerate(connector);
+    ClipIfDegenerate(newConnector);
+  }
+
+  // Recalculate the cost of the Vert v ear, updating it in the queue by
+  // removing and reinserting it.
+  void ProcessEar(VertItr v) {
+    if (v->ear != earsQueue_.end()) {
+      earsQueue_.erase(v->ear);
+      v->ear = earsQueue_.end();
+    }
+    if (v->IsShort(precision_)) {
+      v->cost = kBest;
+      v->ear = earsQueue_.insert(v);
+    } else if (v->IsConvex(precision_)) {
+      v->cost = v->EarCost(precision_);
+      v->ear = earsQueue_.insert(v);
+    }
+  }
+
+  // The main ear-clipping loop. This is called once for each simple polygon -
+  // all holes have already been key-holed and joined to an outer polygon.
+  void TriangulatePoly(VertItr start) {
+    ZoneScoped;
+
+    // A simple polygon always creates two fewer triangles than it has verts.
+    int numTri = -2;
+    earsQueue_.clear();
+
+    auto QueueVert = [&](VertItr v) {
+      ProcessEar(v);
+      ++numTri;
+      v->PrintVert();
+    };
+
+    VertItr v = Loop(start, QueueVert);
+    if (v == polygon_.end()) return;
+    Dump(v);
+
+    while (numTri > 0) {
+      const qItr ear = earsQueue_.begin();
+      if (ear != earsQueue_.end()) {
+        v = *ear;
+        // Cost should always be negative, generally < -precision.
+        v->PrintVert();
+        earsQueue_.erase(ear);
+      } else {
+        PRINT("No ear found!");
+      }
+
+      ClipEar(v);
+      --numTri;
+
+      ProcessEar(v->left);
+      ProcessEar(v->right);
+      // This is a backup vert that is used if the queue is empty (geometrically
+      // invalid polygon), to ensure manifoldness.
+      v = v->right;
+    }
+
+    ASSERT(v->right == v->left, logicErr, "Triangulator error!");
+    PRINT("Finished poly");
+  }
+
+  void Dump(VertItr start) const {
+#ifdef MANIFOLD_DEBUG
+    if (!params.verbose) return;
+    VertItr v = start;
+    std::cout << "show(array([" << std::endl;
+    do {
+      std::cout << "  [" << v->pos.x << ", " << v->pos.y << "],# "
+                << v->mesh_idx << ", cost: " << v->cost << std::endl;
+      v = v->right;
+    } while (v != start);
+    std::cout << "  [" << v->pos.x << ", " << v->pos.y << "],# " << v->mesh_idx
+              << std::endl;
+    std::cout << "]))" << std::endl;
+#endif
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/**
+ * @brief Triangulates a set of &epsilon;-valid polygons. If the input is not
+ * &epsilon;-valid, the triangulation may overlap, but will always return a
+ * manifold result that matches the input edge directions.
+ *
+ * @param polys The set of polygons, wound CCW and representing multiple
+ * polygons and/or holes. These have 2D-projected positions as well as
+ * references back to the original vertices.
+ * @param precision The value of &epsilon;, bounding the uncertainty of the
+ * input.
+ * @return std::vector<glm::ivec3> The triangles, referencing the original
+ * vertex indicies.
+ */
+std::vector<glm::ivec3> TriangulateIdx(const PolygonsIdx &polys,
+                                       float precision) {
+  std::vector<glm::ivec3> triangles;
+  try {
+    EarClip triangulator(polys, precision);
+    triangles = triangulator.Triangulate();
+#ifdef MANIFOLD_DEBUG
+    if (params.intermediateChecks) {
+      CheckTopology(triangles, polys);
+      if (!params.processOverlaps) {
+        CheckGeometry(triangles, polys, 2 * triangulator.GetPrecision());
+      }
+    }
+  } catch (const geometryErr &e) {
+    if (!params.suppressErrors) {
+      PrintFailure(e, polys, triangles, precision);
+    }
+    throw;
+  } catch (const std::exception &e) {
+    PrintFailure(e, polys, triangles, precision);
+    throw;
+#else
+  } catch (const std::exception &e) {
+#endif
+  }
+  return triangles;
+}
+
+/**
+ * @brief Triangulates a set of &epsilon;-valid polygons. If the input is not
+ * &epsilon;-valid, the triangulation may overlap, but will always return a
+ * manifold result that matches the input edge directions.
+ *
+ * @param polygons The set of polygons, wound CCW and representing multiple
+ * polygons and/or holes.
+ * @param precision The value of &epsilon;, bounding the uncertainty of the
+ * input.
+ * @return std::vector<glm::ivec3> The triangles, referencing the original
+ * polygon points in order.
+ */
+std::vector<glm::ivec3> Triangulate(const Polygons &polygons, float precision) {
+  int idx = 0;
+  PolygonsIdx polygonsIndexed;
+  for (const auto &poly : polygons) {
+    SimplePolygonIdx simpleIndexed;
+    for (const glm::vec2 &polyVert : poly) {
+      simpleIndexed.push_back({polyVert, idx++});
+    }
+    polygonsIndexed.push_back(simpleIndexed);
+  }
+  return TriangulateIdx(polygonsIndexed, precision);
+}
+
+ExecutionParams &PolygonParams() { return params; }
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/sdf/include/sdf.h b/thirdparty/manifold/src/sdf/include/sdf.h
new file mode 100644
index 000000000000..d65afe803dac
--- /dev/null
+++ b/thirdparty/manifold/src/sdf/include/sdf.h
@@ -0,0 +1,24 @@
+// Copyright 2023 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <functional>
+
+#include "public.h"
+
+namespace manifold {
+Mesh LevelSet(std::function<float(glm::vec3)> sdf, Box bounds, float edgeLength,
+              float level = 0, bool canParallel = true);
+}
diff --git a/thirdparty/manifold/src/sdf/src/sdf.cpp b/thirdparty/manifold/src/sdf/src/sdf.cpp
new file mode 100644
index 000000000000..eaabe1f90a09
--- /dev/null
+++ b/thirdparty/manifold/src/sdf/src/sdf.cpp
@@ -0,0 +1,362 @@
+// Copyright 2023 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "sdf.h"
+
+#include "hashtable.h"
+#include "par.h"
+#include "utils.h"
+#include "vec.h"
+
+namespace {
+using namespace manifold;
+Uint64 identity(Uint64 x) { return x; }
+
+glm::ivec3 TetTri0(int i) {
+  constexpr glm::ivec3 tetTri0[16] = {{-1, -1, -1},  //
+                                      {0, 3, 4},     //
+                                      {0, 1, 5},     //
+                                      {1, 5, 3},     //
+                                      {1, 4, 2},     //
+                                      {1, 0, 3},     //
+                                      {2, 5, 0},     //
+                                      {5, 3, 2},     //
+                                      {2, 3, 5},     //
+                                      {0, 5, 2},     //
+                                      {3, 0, 1},     //
+                                      {2, 4, 1},     //
+                                      {3, 5, 1},     //
+                                      {5, 1, 0},     //
+                                      {4, 3, 0},     //
+                                      {-1, -1, -1}};
+  return tetTri0[i];
+}
+
+glm::ivec3 TetTri1(int i) {
+  constexpr glm::ivec3 tetTri1[16] = {{-1, -1, -1},  //
+                                      {-1, -1, -1},  //
+                                      {-1, -1, -1},  //
+                                      {3, 4, 1},     //
+                                      {-1, -1, -1},  //
+                                      {3, 2, 1},     //
+                                      {0, 4, 2},     //
+                                      {-1, -1, -1},  //
+                                      {-1, -1, -1},  //
+                                      {2, 4, 0},     //
+                                      {1, 2, 3},     //
+                                      {-1, -1, -1},  //
+                                      {1, 4, 3},     //
+                                      {-1, -1, -1},  //
+                                      {-1, -1, -1},  //
+                                      {-1, -1, -1}};
+  return tetTri1[i];
+}
+
+glm::ivec4 Neighbors(int i) {
+  constexpr glm::ivec4 neighbors[7] = {{0, 0, 0, 1},   //
+                                       {1, 0, 0, 0},   //
+                                       {0, 1, 0, 0},   //
+                                       {0, 0, 1, 0},   //
+                                       {-1, 0, 0, 1},  //
+                                       {0, -1, 0, 1},  //
+                                       {0, 0, -1, 1}};
+  return neighbors[i];
+}
+
+Uint64 SpreadBits3(Uint64 v) {
+  v = v & 0x1fffff;
+  v = (v | v << 32) & 0x1f00000000ffff;
+  v = (v | v << 16) & 0x1f0000ff0000ff;
+  v = (v | v << 8) & 0x100f00f00f00f00f;
+  v = (v | v << 4) & 0x10c30c30c30c30c3;
+  v = (v | v << 2) & 0x1249249249249249;
+  return v;
+}
+
+Uint64 SqueezeBits3(Uint64 v) {
+  v = v & 0x1249249249249249;
+  v = (v ^ v >> 2) & 0x10c30c30c30c30c3;
+  v = (v ^ v >> 4) & 0x100f00f00f00f00f;
+  v = (v ^ v >> 8) & 0x1f0000ff0000ff;
+  v = (v ^ v >> 16) & 0x1f00000000ffff;
+  v = (v ^ v >> 32) & 0x1fffff;
+  return v;
+}
+
+// This is a modified 3D MortonCode, where the xyz code is shifted by one bit
+// and the w bit is added as the least significant. This allows 21 bits per x,
+// y, and z channel and 1 for w, filling the 64 bit total.
+Uint64 MortonCode(const glm::ivec4& index) {
+  return static_cast<Uint64>(index.w) | (SpreadBits3(index.x) << 1) |
+         (SpreadBits3(index.y) << 2) | (SpreadBits3(index.z) << 3);
+}
+
+glm::ivec4 DecodeMorton(Uint64 code) {
+  glm::ivec4 index;
+  index.x = SqueezeBits3(code >> 1);
+  index.y = SqueezeBits3(code >> 2);
+  index.z = SqueezeBits3(code >> 3);
+  index.w = code & 0x1u;
+  return index;
+}
+
+struct GridVert {
+  float distance = NAN;
+  int edgeVerts[7] = {-1, -1, -1, -1, -1, -1, -1};
+
+  int Inside() const { return distance > 0 ? 1 : -1; }
+
+  int NeighborInside(int i) const {
+    return Inside() * (edgeVerts[i] < 0 ? 1 : -1);
+  }
+};
+
+struct ComputeVerts {
+  VecView<glm::vec3> vertPos;
+  VecView<int> vertIndex;
+  HashTableD<GridVert, identity> gridVerts;
+  const std::function<float(glm::vec3)> sdf;
+  const glm::vec3 origin;
+  const glm::ivec3 gridSize;
+  const glm::vec3 spacing;
+  const float level;
+
+  inline glm::vec3 Position(glm::ivec4 gridIndex) const {
+    return origin +
+           spacing * (glm::vec3(gridIndex) + (gridIndex.w == 1 ? 0.0f : -0.5f));
+  }
+
+  inline float BoundedSDF(glm::ivec4 gridIndex) const {
+    const float d = sdf(Position(gridIndex)) - level;
+
+    const glm::ivec3 xyz(gridIndex);
+    const bool onLowerBound = glm::any(glm::lessThanEqual(xyz, glm::ivec3(0)));
+    const bool onUpperBound = glm::any(glm::greaterThanEqual(xyz, gridSize));
+    const bool onHalfBound =
+        gridIndex.w == 1 && glm::any(glm::greaterThanEqual(xyz, gridSize - 1));
+    if (onLowerBound || onUpperBound || onHalfBound) return glm::min(d, 0.0f);
+
+    return d;
+  }
+
+  inline void operator()(Uint64 mortonCode) {
+    ZoneScoped;
+    if (gridVerts.Full()) return;
+
+    const glm::ivec4 gridIndex = DecodeMorton(mortonCode);
+
+    if (glm::any(glm::greaterThan(glm::ivec3(gridIndex), gridSize))) return;
+
+    const glm::vec3 position = Position(gridIndex);
+
+    GridVert gridVert;
+    gridVert.distance = BoundedSDF(gridIndex);
+
+    bool keep = false;
+    // These seven edges are uniquely owned by this gridVert; any of them
+    // which intersect the surface create a vert.
+    for (int i = 0; i < 7; ++i) {
+      glm::ivec4 neighborIndex = gridIndex + Neighbors(i);
+      if (neighborIndex.w == 2) {
+        neighborIndex += 1;
+        neighborIndex.w = 0;
+      }
+      const float val = BoundedSDF(neighborIndex);
+      if ((val > 0) == (gridVert.distance > 0)) continue;
+      keep = true;
+
+      const int idx = AtomicAdd(vertIndex[0], 1);
+      vertPos[idx] =
+          (val * position - gridVert.distance * Position(neighborIndex)) /
+          (val - gridVert.distance);
+      gridVert.edgeVerts[i] = idx;
+    }
+
+    if (keep) gridVerts.Insert(mortonCode, gridVert);
+  }
+};
+
+struct BuildTris {
+  VecView<glm::ivec3> triVerts;
+  VecView<int> triIndex;
+  const HashTableD<GridVert, identity> gridVerts;
+
+  void CreateTri(const glm::ivec3& tri, const int edges[6]) {
+    if (tri[0] < 0) return;
+    int idx = AtomicAdd(triIndex[0], 1);
+    triVerts[idx] = {edges[tri[0]], edges[tri[1]], edges[tri[2]]};
+  }
+
+  void CreateTris(const glm::ivec4& tet, const int edges[6]) {
+    const int i = (tet[0] > 0 ? 1 : 0) + (tet[1] > 0 ? 2 : 0) +
+                  (tet[2] > 0 ? 4 : 0) + (tet[3] > 0 ? 8 : 0);
+    CreateTri(TetTri0(i), edges);
+    CreateTri(TetTri1(i), edges);
+  }
+
+  void operator()(int idx) {
+    ZoneScoped;
+    Uint64 basekey = gridVerts.KeyAt(idx);
+    if (basekey == kOpen) return;
+
+    const GridVert& base = gridVerts.At(idx);
+    const glm::ivec4 baseIndex = DecodeMorton(basekey);
+
+    glm::ivec4 leadIndex = baseIndex;
+    if (leadIndex.w == 0)
+      leadIndex.w = 1;
+    else {
+      leadIndex += 1;
+      leadIndex.w = 0;
+    }
+
+    // This GridVert is in charge of the 6 tetrahedra surrounding its edge in
+    // the (1,1,1) direction (edge 0).
+    glm::ivec4 tet(base.NeighborInside(0), base.Inside(), -2, -2);
+    glm::ivec4 thisIndex = baseIndex;
+    thisIndex.x += 1;
+
+    GridVert thisVert = gridVerts[MortonCode(thisIndex)];
+
+    tet[2] = base.NeighborInside(1);
+    for (const int i : {0, 1, 2}) {
+      thisIndex = leadIndex;
+      --thisIndex[Prev3(i)];
+      // MortonCodes take unsigned input, so check for negatives, given the
+      // decrement.
+      GridVert nextVert = thisIndex[Prev3(i)] < 0
+                              ? GridVert()
+                              : gridVerts[MortonCode(thisIndex)];
+      tet[3] = base.NeighborInside(Prev3(i) + 4);
+
+      const int edges1[6] = {base.edgeVerts[0],
+                             base.edgeVerts[i + 1],
+                             nextVert.edgeVerts[Next3(i) + 4],
+                             nextVert.edgeVerts[Prev3(i) + 1],
+                             thisVert.edgeVerts[i + 4],
+                             base.edgeVerts[Prev3(i) + 4]};
+      thisVert = nextVert;
+      CreateTris(tet, edges1);
+
+      thisIndex = baseIndex;
+      ++thisIndex[Next3(i)];
+      nextVert = gridVerts[MortonCode(thisIndex)];
+      tet[2] = tet[3];
+      tet[3] = base.NeighborInside(Next3(i) + 1);
+
+      const int edges2[6] = {base.edgeVerts[0],
+                             edges1[5],
+                             thisVert.edgeVerts[i + 4],
+                             nextVert.edgeVerts[Next3(i) + 4],
+                             edges1[3],
+                             base.edgeVerts[Next3(i) + 1]};
+      thisVert = nextVert;
+      CreateTris(tet, edges2);
+
+      tet[2] = tet[3];
+    }
+  }
+};
+}  // namespace
+
+namespace manifold {
+
+/** @addtogroup Core
+ *  @{
+ */
+
+/**
+ * Constructs a level-set Mesh from the input Signed-Distance Function (SDF).
+ * This uses a form of Marching Tetrahedra (akin to Marching Cubes, but better
+ * for manifoldness). Instead of using a cubic grid, it uses a body-centered
+ * cubic grid (two shifted cubic grids). This means if your function's interior
+ * exceeds the given bounds, you will see a kind of egg-crate shape closing off
+ * the manifold, which is due to the underlying grid.
+ *
+ * @param sdf The signed-distance functor, containing this function signature:
+ * `float operator()(glm::vec3 point)`, which returns the
+ * signed distance of a given point in R^3. Positive values are inside,
+ * negative outside.
+ * @param bounds An axis-aligned box that defines the extent of the grid.
+ * @param edgeLength Approximate maximum edge length of the triangles in the
+ * final result. This affects grid spacing, and hence has a strong effect on
+ * performance.
+ * @param level You can inset your Mesh by using a positive value, or outset
+ * it with a negative value.
+ * @param canParallel Parallel policies violate will crash language runtimes
+ * with runtime locks that expect to not be called back by unregistered threads.
+ * This allows bindings use LevelSet despite being compiled with MANIFOLD_PAR
+ * active.
+ * @return Mesh This class does not depend on Manifold, so it just returns a
+ * Mesh, but it is guaranteed to be manifold and so can always be used as
+ * input to the Manifold constructor for further operations.
+ */
+Mesh LevelSet(std::function<float(glm::vec3)> sdf, Box bounds, float edgeLength,
+              float level, bool canParallel) {
+  Mesh out;
+
+  const glm::vec3 dim = bounds.Size();
+  const glm::ivec3 gridSize(dim / edgeLength);
+  const glm::vec3 spacing = dim / (glm::vec3(gridSize));
+
+  const Uint64 maxMorton = MortonCode(glm::ivec4(gridSize + 1, 1));
+
+  // Parallel policies violate will crash language runtimes with runtime locks
+  // that expect to not be called back by unregistered threads. This allows
+  // bindings use LevelSet despite being compiled with MANIFOLD_PAR
+  // active.
+  const auto pol = canParallel ? autoPolicy(maxMorton) : ExecutionPolicy::Seq;
+
+  size_t tableSize = glm::min(
+      2 * maxMorton, static_cast<Uint64>(10 * glm::pow(maxMorton, 0.667)));
+  HashTable<GridVert, identity> gridVerts(tableSize);
+  Vec<glm::vec3> vertPos(gridVerts.Size() * 7);
+
+  while (1) {
+    Vec<int> index(1, 0);
+    for_each_n(pol, countAt(0_z), maxMorton + 1,
+               ComputeVerts({vertPos, index, gridVerts.D(), sdf, bounds.min,
+                             gridSize + 1, spacing, level}));
+
+    if (gridVerts.Full()) {  // Resize HashTable
+      const glm::vec3 lastVert = vertPos[index[0] - 1];
+      const Uint64 lastMorton =
+          MortonCode(glm::ivec4((lastVert - bounds.min) / spacing, 1));
+      const float ratio = static_cast<float>(maxMorton) / lastMorton;
+
+      if (ratio > 1000)  // do not trust the ratio if it is too large
+        tableSize *= 2;
+      else
+        tableSize *= ratio;
+      gridVerts = HashTable<GridVert, identity>(tableSize);
+      vertPos = Vec<glm::vec3>(gridVerts.Size() * 7);
+    } else {  // Success
+      vertPos.resize(index[0]);
+      break;
+    }
+  }
+
+  Vec<glm::ivec3> triVerts(gridVerts.Entries() * 12);  // worst case
+
+  Vec<int> index(1, 0);
+  for_each_n(pol, countAt(0_z), gridVerts.Size(),
+             BuildTris({triVerts, index, gridVerts.D()}));
+  triVerts.resize(index[0]);
+
+  out.vertPos.insert(out.vertPos.end(), vertPos.begin(), vertPos.end());
+  out.triVerts.insert(out.triVerts.end(), triVerts.begin(), triVerts.end());
+  return out;
+}
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/third_party/quickhull/.gitignore b/thirdparty/manifold/src/third_party/quickhull/.gitignore
new file mode 100644
index 000000000000..496ee2ca6a2f
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/.gitignore
@@ -0,0 +1 @@
+.DS_Store
\ No newline at end of file
diff --git a/thirdparty/manifold/src/third_party/quickhull/.gitrepo b/thirdparty/manifold/src/third_party/quickhull/.gitrepo
new file mode 100644
index 000000000000..d54da571cc74
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/.gitrepo
@@ -0,0 +1,12 @@
+; DO NOT EDIT (unless you know what you are doing)
+;
+; This subdirectory is a git "subrepo", and this file is maintained by the
+; git-subrepo command. See https://github.com/ingydotnet/git-subrepo#readme
+;
+[subrepo]
+	remote = https://github.com/akuukka/quickhull.git
+	branch = 1ffbc6f884ea1da89e104a5996cf8a726db673d5
+	commit = 1ffbc6f884ea1da89e104a5996cf8a726db673d5
+	parent = b8b715dff0615a5e85dc956d514bcf680efa61fa
+	method = merge
+	cmdver = 0.4.6
diff --git a/thirdparty/manifold/src/third_party/quickhull/ConvexHull.hpp b/thirdparty/manifold/src/third_party/quickhull/ConvexHull.hpp
new file mode 100644
index 000000000000..71b2ee1114f3
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/ConvexHull.hpp
@@ -0,0 +1,182 @@
+#ifndef CONVEXHULL_HPP_
+#define CONVEXHULL_HPP_
+
+#include "Structs/Vector3.hpp"
+#include "Structs/Mesh.hpp"
+#include "Structs/VertexDataSource.hpp"
+#include <vector>
+#include <unordered_map>
+#include <fstream>
+#include <memory>
+
+namespace quickhull {
+
+	template<typename T>
+	class ConvexHull {
+		std::unique_ptr<std::vector<Vector3<T>>> m_optimizedVertexBuffer;
+		VertexDataSource<T> m_vertices;
+		std::vector<size_t> m_indices;
+	public:
+		ConvexHull() {}
+		
+		// Copy constructor
+		ConvexHull(const ConvexHull& o) {
+			m_indices = o.m_indices;
+			if (o.m_optimizedVertexBuffer) {
+				m_optimizedVertexBuffer.reset(new std::vector<Vector3<T>>(*o.m_optimizedVertexBuffer));
+				m_vertices = VertexDataSource<T>(*m_optimizedVertexBuffer);
+			}
+			else {
+				m_vertices = o.m_vertices;
+			}
+		}
+		
+		ConvexHull& operator=(const ConvexHull& o) {
+			if (&o == this) {
+				return *this;
+			}
+			m_indices = o.m_indices;
+			if (o.m_optimizedVertexBuffer) {
+				m_optimizedVertexBuffer.reset(new std::vector<Vector3<T>>(*o.m_optimizedVertexBuffer));
+				m_vertices = VertexDataSource<T>(*m_optimizedVertexBuffer);
+			}
+			else {
+				m_vertices = o.m_vertices;
+			}
+			return *this;
+		}
+		
+		ConvexHull(ConvexHull&& o) {
+			m_indices = std::move(o.m_indices);
+			if (o.m_optimizedVertexBuffer) {
+				m_optimizedVertexBuffer = std::move(o.m_optimizedVertexBuffer);
+				o.m_vertices = VertexDataSource<T>();
+				m_vertices = VertexDataSource<T>(*m_optimizedVertexBuffer);
+			}
+			else {
+				m_vertices = o.m_vertices;
+			}
+		}
+		
+		ConvexHull& operator=(ConvexHull&& o) {
+			if (&o == this) {
+				return *this;
+			}
+			m_indices = std::move(o.m_indices);
+			if (o.m_optimizedVertexBuffer) {
+				m_optimizedVertexBuffer = std::move(o.m_optimizedVertexBuffer);
+				o.m_vertices = VertexDataSource<T>();
+				m_vertices = VertexDataSource<T>(*m_optimizedVertexBuffer);
+			}
+			else {
+				m_vertices = o.m_vertices;
+			}
+			return *this;
+		}
+		
+		// Construct vertex and index buffers from half edge mesh and pointcloud
+		ConvexHull(const MeshBuilder<T>& mesh, const VertexDataSource<T>& pointCloud, bool CCW, bool useOriginalIndices) {
+			if (!useOriginalIndices) {
+				m_optimizedVertexBuffer.reset(new std::vector<Vector3<T>>());
+			}
+			
+			std::vector<bool> faceProcessed(mesh.m_faces.size(),false);
+			std::vector<size_t> faceStack;
+			std::unordered_map<size_t,size_t> vertexIndexMapping; // Map vertex indices from original point cloud to the new mesh vertex indices
+			for (size_t i = 0;i<mesh.m_faces.size();i++) {
+				if (!mesh.m_faces[i].isDisabled()) {
+					faceStack.push_back(i);
+					break;
+				}
+			}
+			if (faceStack.size()==0) {
+				return;
+			}
+
+			const size_t iCCW = CCW ? 1 : 0;
+			const size_t finalMeshFaceCount = mesh.m_faces.size() - mesh.m_disabledFaces.size();
+			m_indices.reserve(finalMeshFaceCount*3);
+
+			while (faceStack.size()) {
+				auto it = faceStack.end()-1;
+				size_t top = *it;
+				assert(!mesh.m_faces[top].isDisabled());
+				faceStack.erase(it);
+				if (faceProcessed[top]) {
+					continue;
+				}
+				else {
+					faceProcessed[top]=true;
+					auto halfEdges = mesh.getHalfEdgeIndicesOfFace(mesh.m_faces[top]);
+					size_t adjacent[] = {mesh.m_halfEdges[mesh.m_halfEdges[halfEdges[0]].m_opp].m_face,mesh.m_halfEdges[mesh.m_halfEdges[halfEdges[1]].m_opp].m_face,mesh.m_halfEdges[mesh.m_halfEdges[halfEdges[2]].m_opp].m_face};
+					for (auto a : adjacent) {
+						if (!faceProcessed[a] && !mesh.m_faces[a].isDisabled()) {
+							faceStack.push_back(a);
+						}
+					}
+					auto vertices = mesh.getVertexIndicesOfFace(mesh.m_faces[top]);
+					if (!useOriginalIndices) {
+						for (auto& v : vertices) {
+							auto itV = vertexIndexMapping.find(v);
+							if (itV == vertexIndexMapping.end()) {
+								m_optimizedVertexBuffer->push_back(pointCloud[v]);
+								vertexIndexMapping[v] = m_optimizedVertexBuffer->size()-1;
+								v = m_optimizedVertexBuffer->size()-1;
+							}
+							else {
+								v = itV->second;
+							}
+						}
+					}
+					m_indices.push_back(vertices[0]);
+					m_indices.push_back(vertices[1 + iCCW]);
+					m_indices.push_back(vertices[2 - iCCW]);
+				}
+			}
+			
+			if (!useOriginalIndices) {
+				m_vertices = VertexDataSource<T>(*m_optimizedVertexBuffer);
+			}
+			else {
+				m_vertices = pointCloud;
+			}
+		}
+
+		std::vector<size_t>& getIndexBuffer() {
+			return m_indices;
+		}
+
+		const std::vector<size_t>& getIndexBuffer() const {
+			return m_indices;
+		}
+
+		VertexDataSource<T>& getVertexBuffer() {
+			return m_vertices;
+		}
+		
+		const VertexDataSource<T>& getVertexBuffer() const {
+			return m_vertices;
+		}
+		
+		// Export the mesh to a Waveform OBJ file
+		void writeWaveformOBJ(const std::string& filename, const std::string& objectName = "quickhull") const
+		{
+			std::ofstream objFile;
+			objFile.open (filename);
+			objFile << "o " << objectName << "\n";
+			for (const auto& v : getVertexBuffer()) {
+				objFile << "v " << v.x << " " << v.y << " " << v.z << "\n";
+			}
+			const auto& indBuf = getIndexBuffer();
+			size_t triangleCount = indBuf.size()/3;
+			for (size_t i=0;i<triangleCount;i++) {
+				objFile << "f " << indBuf[i*3]+1 << " " << indBuf[i*3+1]+1 << " " << indBuf[i*3+2]+1 << "\n";
+			}
+			objFile.close();
+		}
+
+	};
+
+}
+
+#endif /* CONVEXHULL_HPP_ */
diff --git a/thirdparty/manifold/src/third_party/quickhull/HalfEdgeMesh.hpp b/thirdparty/manifold/src/third_party/quickhull/HalfEdgeMesh.hpp
new file mode 100644
index 000000000000..5800a3542727
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/HalfEdgeMesh.hpp
@@ -0,0 +1,77 @@
+#include "Structs/Mesh.hpp"
+
+#ifndef HalfEdgeMesh_h
+#define HalfEdgeMesh_h
+
+namespace quickhull {
+	
+	template<typename FloatType, typename IndexType>
+	class HalfEdgeMesh {
+	public:
+		
+		struct HalfEdge {
+			IndexType m_endVertex;
+			IndexType m_opp;
+			IndexType m_face;
+			IndexType m_next;
+		};
+		
+		struct Face {
+			IndexType m_halfEdgeIndex; // Index of one of the half edges of this face
+		};
+		
+		std::vector<Vector3<FloatType>> m_vertices;
+		std::vector<Face> m_faces;
+		std::vector<HalfEdge> m_halfEdges;
+		
+		HalfEdgeMesh(const MeshBuilder<FloatType>& builderObject, const VertexDataSource<FloatType>& vertexData )
+		{
+			std::unordered_map<IndexType,IndexType> faceMapping;
+			std::unordered_map<IndexType,IndexType> halfEdgeMapping;
+			std::unordered_map<IndexType, IndexType> vertexMapping;
+			
+			size_t i=0;
+			for (const auto& face : builderObject.m_faces) {
+				if (!face.isDisabled()) {
+					m_faces.push_back({static_cast<IndexType>(face.m_he)});
+					faceMapping[i] = m_faces.size()-1;
+					
+					const auto heIndices = builderObject.getHalfEdgeIndicesOfFace(face);
+					for (const auto heIndex : heIndices) {
+						const IndexType vertexIndex = builderObject.m_halfEdges[heIndex].m_endVertex;
+						if (vertexMapping.count(vertexIndex)==0) {
+							m_vertices.push_back(vertexData[vertexIndex]);
+							vertexMapping[vertexIndex] = m_vertices.size()-1;
+						}
+					}
+				}
+				i++;
+			}
+			
+			i=0;
+			for (const auto& halfEdge : builderObject.m_halfEdges) {
+				if (!halfEdge.isDisabled()) {
+					m_halfEdges.push_back({static_cast<IndexType>(halfEdge.m_endVertex),static_cast<IndexType>(halfEdge.m_opp),static_cast<IndexType>(halfEdge.m_face),static_cast<IndexType>(halfEdge.m_next)});
+					halfEdgeMapping[i] = m_halfEdges.size()-1;
+				}
+				i++;
+			}
+			
+			for (auto& face : m_faces) {
+				assert(halfEdgeMapping.count(face.m_halfEdgeIndex) == 1);
+				face.m_halfEdgeIndex = halfEdgeMapping[face.m_halfEdgeIndex];
+			}
+			
+			for (auto& he : m_halfEdges) {
+				he.m_face = faceMapping[he.m_face];
+				he.m_opp = halfEdgeMapping[he.m_opp];
+				he.m_next = halfEdgeMapping[he.m_next];
+				he.m_endVertex = vertexMapping[he.m_endVertex];
+			}
+		}
+		
+	};
+}
+
+
+#endif /* HalfEdgeMesh_h */
diff --git a/thirdparty/manifold/src/third_party/quickhull/MathUtils.hpp b/thirdparty/manifold/src/third_party/quickhull/MathUtils.hpp
new file mode 100644
index 000000000000..b87b3b934097
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/MathUtils.hpp
@@ -0,0 +1,46 @@
+
+#ifndef QuickHull_MathUtils_hpp
+#define QuickHull_MathUtils_hpp
+
+#include "Structs/Vector3.hpp"
+#include "Structs/Ray.hpp"
+
+namespace quickhull {
+	
+	namespace mathutils {
+		
+		template <typename T>
+		inline T getSquaredDistanceBetweenPointAndRay(const Vector3<T>& p, const Ray<T>& r) {
+			const Vector3<T> s = p-r.m_S;
+			T t = s.dotProduct(r.m_V);
+			return s.getLengthSquared() - t*t*r.m_VInvLengthSquared;
+		}
+		
+		// Note that the unit of distance returned is relative to plane's normal's length (divide by N.getNormalized() if needed to get the "real" distance).
+		template <typename T>
+		inline T getSignedDistanceToPlane(const Vector3<T>& v, const Plane<T>& p) {
+			return p.m_N.dotProduct(v) + p.m_D;
+		}
+		
+		template <typename T>
+		inline Vector3<T> getTriangleNormal(const Vector3<T>& a,const Vector3<T>& b,const Vector3<T>& c) {
+			// We want to get (a-c).crossProduct(b-c) without constructing temp vectors
+			T x = a.x - c.x;
+			T y = a.y - c.y;
+			T z = a.z - c.z;
+			T rhsx = b.x - c.x;
+			T rhsy = b.y - c.y;
+			T rhsz = b.z - c.z;
+			T px = y * rhsz - z * rhsy ;
+			T py = z * rhsx - x * rhsz ;
+			T pz = x * rhsy - y * rhsx ;
+			return Vector3<T>(px,py,pz);
+		}
+		
+		
+	}
+	
+}
+
+
+#endif
diff --git a/thirdparty/manifold/src/third_party/quickhull/QuickHull.cpp b/thirdparty/manifold/src/third_party/quickhull/QuickHull.cpp
new file mode 100644
index 000000000000..fc02f3d00423
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/QuickHull.cpp
@@ -0,0 +1,503 @@
+#include "QuickHull.hpp"
+#include "MathUtils.hpp"
+#include <cmath>
+#include <cassert>
+#include <iostream>
+#include <algorithm>
+#include <limits>
+#include "Structs/Mesh.hpp"
+
+namespace quickhull {
+
+	template<>
+	float defaultEps() { 
+		return 0.0001f;
+	}
+	
+	template<>
+	double defaultEps() {
+		return 0.0000001; 
+	}
+	
+	/*
+	 * Implementation of the algorithm
+	 */
+	
+	template<typename T>
+	ConvexHull<T> QuickHull<T>::getConvexHull(const std::vector<Vector3<T>>& pointCloud, bool CCW, bool useOriginalIndices, T epsilon) {
+		VertexDataSource<T> vertexDataSource(pointCloud);
+		return getConvexHull(vertexDataSource,CCW,useOriginalIndices,epsilon);
+	}
+	
+	template<typename T>
+	ConvexHull<T> QuickHull<T>::getConvexHull(const Vector3<T>* vertexData, size_t vertexCount, bool CCW, bool useOriginalIndices, T epsilon) {
+		VertexDataSource<T> vertexDataSource(vertexData,vertexCount);
+		return getConvexHull(vertexDataSource,CCW,useOriginalIndices,epsilon);
+	}
+	
+	template<typename T>
+	ConvexHull<T> QuickHull<T>::getConvexHull(const T* vertexData, size_t vertexCount, bool CCW, bool useOriginalIndices, T epsilon) {
+		VertexDataSource<T> vertexDataSource((const vec3*)vertexData,vertexCount);
+		return getConvexHull(vertexDataSource,CCW,useOriginalIndices,epsilon);
+	}
+	
+	template<typename FloatType>
+	HalfEdgeMesh<FloatType, size_t> QuickHull<FloatType>::getConvexHullAsMesh(const FloatType* vertexData, size_t vertexCount, bool CCW, FloatType epsilon) {
+		VertexDataSource<FloatType> vertexDataSource((const vec3*)vertexData,vertexCount);
+		buildMesh(vertexDataSource, CCW, false, epsilon);
+		return HalfEdgeMesh<FloatType, size_t>(m_mesh, m_vertexData);
+	}
+	
+	template<typename T>
+	void QuickHull<T>::buildMesh(const VertexDataSource<T>& pointCloud, bool CCW, bool useOriginalIndices, T epsilon) {
+		// CCW is unused for now
+		(void)CCW;
+		// useOriginalIndices is unused for now
+		(void)useOriginalIndices;
+
+		if (pointCloud.size()==0) {
+			m_mesh = MeshBuilder<T>();
+			return;
+		}
+		m_vertexData = pointCloud;
+		
+		// Very first: find extreme values and use them to compute the scale of the point cloud.
+		m_extremeValues = getExtremeValues();
+		m_scale = getScale(m_extremeValues);
+		
+		// Epsilon we use depends on the scale
+		m_epsilon = epsilon*m_scale;
+		m_epsilonSquared = m_epsilon*m_epsilon;
+		
+		// Reset diagnostics
+		m_diagnostics = DiagnosticsData();
+		
+		m_planar = false; // The planar case happens when all the points appear to lie on a two dimensional subspace of R^3.
+		createConvexHalfEdgeMesh();
+		if (m_planar) {
+			const size_t extraPointIndex = m_planarPointCloudTemp.size()-1;
+			for (auto& he : m_mesh.m_halfEdges) {
+				if (he.m_endVertex == extraPointIndex) {
+					he.m_endVertex = 0;
+				}
+			}
+			m_vertexData = pointCloud;
+			m_planarPointCloudTemp.clear();
+		}
+	}
+
+	template<typename T>
+	ConvexHull<T> QuickHull<T>::getConvexHull(const VertexDataSource<T>& pointCloud, bool CCW, bool useOriginalIndices, T epsilon) {
+		buildMesh(pointCloud,CCW,useOriginalIndices,epsilon);
+		return ConvexHull<T>(m_mesh,m_vertexData, CCW, useOriginalIndices);
+	}
+
+	template<typename T>
+	void QuickHull<T>::createConvexHalfEdgeMesh() {
+		m_visibleFaces.clear();
+		m_horizonEdges.clear();
+		m_possiblyVisibleFaces.clear();
+		
+		// Compute base tetrahedron
+		setupInitialTetrahedron();
+		assert(m_mesh.m_faces.size()==4);
+
+		// Init face stack with those faces that have points assigned to them
+		m_faceList.clear();
+		for (size_t i=0;i < 4;i++) {
+			auto& f = m_mesh.m_faces[i];
+			if (f.m_pointsOnPositiveSide && f.m_pointsOnPositiveSide->size()>0) {
+				m_faceList.push_back(i);
+				f.m_inFaceStack = 1;
+			}
+		}
+
+		// Process faces until the face list is empty.
+		size_t iter = 0;
+		while (!m_faceList.empty()) {
+			iter++;
+			if (iter == std::numeric_limits<size_t>::max()) {
+				// Visible face traversal marks visited faces with iteration counter (to mark that the face has been visited on this iteration) and the max value represents unvisited faces. At this point we have to reset iteration counter. This shouldn't be an
+				// issue on 64 bit machines.
+				iter = 0;
+			}
+			
+			const size_t topFaceIndex = m_faceList.front();
+			m_faceList.pop_front();
+			
+			auto& tf = m_mesh.m_faces[topFaceIndex];
+			tf.m_inFaceStack = 0;
+
+			assert(!tf.m_pointsOnPositiveSide || tf.m_pointsOnPositiveSide->size()>0);
+			if (!tf.m_pointsOnPositiveSide || tf.isDisabled()) {
+				continue;
+			}
+			
+			// Pick the most distant point to this triangle plane as the point to which we extrude
+			const vec3& activePoint = m_vertexData[tf.m_mostDistantPoint];
+			const size_t activePointIndex = tf.m_mostDistantPoint;
+
+			// Find out the faces that have our active point on their positive side (these are the "visible faces"). The face on top of the stack of course is one of them. At the same time, we create a list of horizon edges.
+			m_horizonEdges.clear();
+			m_possiblyVisibleFaces.clear();
+			m_visibleFaces.clear();
+			m_possiblyVisibleFaces.emplace_back(topFaceIndex,std::numeric_limits<size_t>::max());
+			while (m_possiblyVisibleFaces.size()) {
+				const auto faceData = m_possiblyVisibleFaces.back();
+				m_possiblyVisibleFaces.pop_back();
+				auto& pvf = m_mesh.m_faces[faceData.m_faceIndex];
+				assert(!pvf.isDisabled());
+				
+				if (pvf.m_visibilityCheckedOnIteration == iter) {
+					if (pvf.m_isVisibleFaceOnCurrentIteration) {
+						continue;
+					}
+				}
+				else {
+					const Plane<T>& P = pvf.m_P;
+					pvf.m_visibilityCheckedOnIteration = iter;
+					const T d = P.m_N.dotProduct(activePoint)+P.m_D;
+					if (d>0) {
+						pvf.m_isVisibleFaceOnCurrentIteration = 1;
+						pvf.m_horizonEdgesOnCurrentIteration = 0;
+						m_visibleFaces.push_back(faceData.m_faceIndex);
+						for (auto heIndex : m_mesh.getHalfEdgeIndicesOfFace(pvf)) {
+							if (m_mesh.m_halfEdges[heIndex].m_opp != faceData.m_enteredFromHalfEdge) {
+								m_possiblyVisibleFaces.emplace_back( m_mesh.m_halfEdges[m_mesh.m_halfEdges[heIndex].m_opp].m_face,heIndex );
+							}
+						}
+						continue;
+					}
+					assert(faceData.m_faceIndex != topFaceIndex);
+				}
+
+				// The face is not visible. Therefore, the halfedge we came from is part of the horizon edge.
+				pvf.m_isVisibleFaceOnCurrentIteration = 0;
+				m_horizonEdges.push_back(faceData.m_enteredFromHalfEdge);
+				// Store which half edge is the horizon edge. The other half edges of the face will not be part of the final mesh so their data slots can by recycled.
+				const auto halfEdges = m_mesh.getHalfEdgeIndicesOfFace(m_mesh.m_faces[m_mesh.m_halfEdges[faceData.m_enteredFromHalfEdge].m_face]);
+				const std::int8_t ind = (halfEdges[0]==faceData.m_enteredFromHalfEdge) ? 0 : (halfEdges[1]==faceData.m_enteredFromHalfEdge ? 1 : 2);
+				m_mesh.m_faces[m_mesh.m_halfEdges[faceData.m_enteredFromHalfEdge].m_face].m_horizonEdgesOnCurrentIteration |= (1<<ind);
+			}
+			const size_t horizonEdgeCount = m_horizonEdges.size();
+
+			// Order horizon edges so that they form a loop. This may fail due to numerical instability in which case we give up trying to solve horizon edge for this point and accept a minor degeneration in the convex hull.
+			if (!reorderHorizonEdges(m_horizonEdges)) {
+				m_diagnostics.m_failedHorizonEdges++;
+				std::cerr << "Failed to solve horizon edge." << std::endl;
+				auto it = std::find(tf.m_pointsOnPositiveSide->begin(),tf.m_pointsOnPositiveSide->end(),activePointIndex);
+				tf.m_pointsOnPositiveSide->erase(it);
+				if (tf.m_pointsOnPositiveSide->size()==0) {
+					reclaimToIndexVectorPool(tf.m_pointsOnPositiveSide);
+				}
+				continue;
+			}
+
+			// Except for the horizon edges, all half edges of the visible faces can be marked as disabled. Their data slots will be reused.
+			// The faces will be disabled as well, but we need to remember the points that were on the positive side of them - therefore
+			// we save pointers to them.
+			m_newFaceIndices.clear();
+			m_newHalfEdgeIndices.clear();
+			m_disabledFacePointVectors.clear();
+			size_t disableCounter = 0;
+			for (auto faceIndex : m_visibleFaces) {
+				auto& disabledFace = m_mesh.m_faces[faceIndex];
+				auto halfEdges = m_mesh.getHalfEdgeIndicesOfFace(disabledFace);
+				for (size_t j=0;j<3;j++) {
+					if ((disabledFace.m_horizonEdgesOnCurrentIteration & (1<<j)) == 0) {
+						if (disableCounter < horizonEdgeCount*2) {
+							// Use on this iteration
+							m_newHalfEdgeIndices.push_back(halfEdges[j]);
+							disableCounter++;
+						}
+						else {
+							// Mark for reusal on later iteration step
+							m_mesh.disableHalfEdge(halfEdges[j]);
+						}
+					}
+				}
+				// Disable the face, but retain pointer to the points that were on the positive side of it. We need to assign those points
+				// to the new faces we create shortly.
+				auto t = m_mesh.disableFace(faceIndex);
+				if (t) {
+					assert(t->size()); // Because we should not assign point vectors to faces unless needed...
+					m_disabledFacePointVectors.push_back(std::move(t));
+				}
+			}
+			if (disableCounter < horizonEdgeCount*2) {
+				const size_t newHalfEdgesNeeded = horizonEdgeCount*2-disableCounter;
+				for (size_t i=0;i<newHalfEdgesNeeded;i++) {
+					m_newHalfEdgeIndices.push_back(m_mesh.addHalfEdge());
+				}
+			}
+			
+			// Create new faces using the edgeloop
+			for (size_t i = 0; i < horizonEdgeCount; i++) {
+				const size_t AB = m_horizonEdges[i];
+
+				auto horizonEdgeVertexIndices = m_mesh.getVertexIndicesOfHalfEdge(m_mesh.m_halfEdges[AB]);
+				size_t A,B,C;
+				A = horizonEdgeVertexIndices[0];
+				B = horizonEdgeVertexIndices[1];
+				C = activePointIndex;
+
+				const size_t newFaceIndex = m_mesh.addFace();
+				m_newFaceIndices.push_back(newFaceIndex);
+
+				const size_t CA = m_newHalfEdgeIndices[2*i+0];
+				const size_t BC = m_newHalfEdgeIndices[2*i+1];
+
+				m_mesh.m_halfEdges[AB].m_next = BC;
+				m_mesh.m_halfEdges[BC].m_next = CA;
+				m_mesh.m_halfEdges[CA].m_next = AB;
+
+				m_mesh.m_halfEdges[BC].m_face = newFaceIndex;
+				m_mesh.m_halfEdges[CA].m_face = newFaceIndex;
+				m_mesh.m_halfEdges[AB].m_face = newFaceIndex;
+
+				m_mesh.m_halfEdges[CA].m_endVertex = A;
+				m_mesh.m_halfEdges[BC].m_endVertex = C;
+
+				auto& newFace = m_mesh.m_faces[newFaceIndex];
+
+				const Vector3<T> planeNormal = mathutils::getTriangleNormal(m_vertexData[A],m_vertexData[B],activePoint);
+				newFace.m_P = Plane<T>(planeNormal,activePoint);
+				newFace.m_he = AB;
+
+				m_mesh.m_halfEdges[CA].m_opp = m_newHalfEdgeIndices[i>0 ? i*2-1 : 2*horizonEdgeCount-1];
+				m_mesh.m_halfEdges[BC].m_opp = m_newHalfEdgeIndices[((i+1)*2) % (horizonEdgeCount*2)];
+			}
+
+			// Assign points that were on the positive side of the disabled faces to the new faces.
+			for (auto& disabledPoints : m_disabledFacePointVectors) {
+				assert(disabledPoints);
+				for (const auto& point : *(disabledPoints)) {
+					if (point == activePointIndex) {
+						continue;
+					}
+					for (size_t j=0;j<horizonEdgeCount;j++) {
+						if (addPointToFace(m_mesh.m_faces[m_newFaceIndices[j]], point)) {
+							break;
+						}
+					}
+				}
+				// The points are no longer needed: we can move them to the vector pool for reuse.
+				reclaimToIndexVectorPool(disabledPoints);
+			}
+
+			// Increase face stack size if needed
+			for (const auto newFaceIndex : m_newFaceIndices) {
+				auto& newFace = m_mesh.m_faces[newFaceIndex];
+				if (newFace.m_pointsOnPositiveSide) {
+					assert(newFace.m_pointsOnPositiveSide->size()>0);
+					if (!newFace.m_inFaceStack) {
+						m_faceList.push_back(newFaceIndex);
+						newFace.m_inFaceStack = 1;
+					}
+				}
+			}
+		}
+		
+		// Cleanup
+		m_indexVectorPool.clear();
+	}
+	
+	/*
+	 * Private helper functions
+	 */
+
+	template <typename T>
+	std::array<size_t,6> QuickHull<T>::getExtremeValues() {
+		std::array<size_t,6> outIndices{0,0,0,0,0,0};
+		T extremeVals[6] = {m_vertexData[0].x,m_vertexData[0].x,m_vertexData[0].y,m_vertexData[0].y,m_vertexData[0].z,m_vertexData[0].z};
+		const size_t vCount = m_vertexData.size();
+		for (size_t i=1;i<vCount;i++) {
+			const Vector3<T>& pos = m_vertexData[i];
+			if (pos.x>extremeVals[0]) {
+				extremeVals[0]=pos.x;
+				outIndices[0]=i;
+			}
+			else if (pos.x<extremeVals[1]) {
+				extremeVals[1]=pos.x;
+				outIndices[1]=i;
+			}
+			if (pos.y>extremeVals[2]) {
+				extremeVals[2]=pos.y;
+				outIndices[2]=i;
+			}
+			else if (pos.y<extremeVals[3]) {
+				extremeVals[3]=pos.y;
+				outIndices[3]=i;
+			}
+			if (pos.z>extremeVals[4]) {
+				extremeVals[4]=pos.z;
+				outIndices[4]=i;
+			}
+			else if (pos.z<extremeVals[5]) {
+				extremeVals[5]=pos.z;
+				outIndices[5]=i;
+			}
+		}
+		return outIndices;
+	}
+
+	template<typename T>
+	bool QuickHull<T>::reorderHorizonEdges(std::vector<size_t>& horizonEdges) {
+		const size_t horizonEdgeCount = horizonEdges.size();
+		for (size_t i=0;i<horizonEdgeCount-1;i++) {
+			const size_t endVertex = m_mesh.m_halfEdges[ horizonEdges[i] ].m_endVertex;
+			bool foundNext = false;
+			for (size_t j=i+1;j<horizonEdgeCount;j++) {
+				const size_t beginVertex = m_mesh.m_halfEdges[ m_mesh.m_halfEdges[horizonEdges[j]].m_opp ].m_endVertex;
+				if (beginVertex == endVertex) {
+					std::swap(horizonEdges[i+1],horizonEdges[j]);
+					foundNext = true;
+					break;
+				}
+			}
+			if (!foundNext) {
+				return false;
+			}
+		}
+		assert(m_mesh.m_halfEdges[ horizonEdges[horizonEdges.size()-1] ].m_endVertex == m_mesh.m_halfEdges[ m_mesh.m_halfEdges[horizonEdges[0]].m_opp ].m_endVertex);
+		return true;
+	}
+	
+	template <typename T>
+	T QuickHull<T>::getScale(const std::array<size_t,6>& extremeValues) {
+		T s = 0;
+		for (size_t i=0;i<6;i++) {
+			const T* v = (const T*)(&m_vertexData[extremeValues[i]]);
+			v += i/2;
+			auto a = std::abs(*v);
+			if (a>s) {
+				s = a;
+			}
+		}
+		return s;
+	}
+
+	template<typename T>
+	void QuickHull<T>::setupInitialTetrahedron() {
+		const size_t vertexCount = m_vertexData.size();
+		
+		// If we have at most 4 points, just return a degenerate tetrahedron:
+		if (vertexCount <= 4) {
+			size_t v[4] = {0,std::min((size_t)1,vertexCount-1),std::min((size_t)2,vertexCount-1),std::min((size_t)3,vertexCount-1)};
+			const Vector3<T> N = mathutils::getTriangleNormal(m_vertexData[v[0]],m_vertexData[v[1]],m_vertexData[v[2]]);
+			const Plane<T> trianglePlane(N,m_vertexData[v[0]]);
+			if (trianglePlane.isPointOnPositiveSide(m_vertexData[v[3]])) {
+				std::swap(v[0],v[1]);
+			}
+			return m_mesh.setup(v[0],v[1],v[2],v[3]);
+		}
+		
+		// Find two most distant extreme points.
+		T maxD = m_epsilonSquared;
+		std::pair<size_t,size_t> selectedPoints;
+		for (size_t i=0;i<6;i++) {
+			for (size_t j=i+1;j<6;j++) {
+				const T d = m_vertexData[ m_extremeValues[i] ].getSquaredDistanceTo( m_vertexData[ m_extremeValues[j] ] );
+				if (d > maxD) {
+					maxD=d;
+					selectedPoints={m_extremeValues[i],m_extremeValues[j]};
+				}
+			}
+		}
+		if (maxD == m_epsilonSquared) {
+			// A degenerate case: the point cloud seems to consists of a single point
+			return m_mesh.setup(0,std::min((size_t)1,vertexCount-1),std::min((size_t)2,vertexCount-1),std::min((size_t)3,vertexCount-1));
+		}
+		assert(selectedPoints.first != selectedPoints.second);
+		
+		// Find the most distant point to the line between the two chosen extreme points.
+		const Ray<T> r(m_vertexData[selectedPoints.first], (m_vertexData[selectedPoints.second] - m_vertexData[selectedPoints.first]));
+		maxD = m_epsilonSquared;
+		size_t maxI=std::numeric_limits<size_t>::max();
+		const size_t vCount = m_vertexData.size();
+		for (size_t i=0;i<vCount;i++) {
+			const T distToRay = mathutils::getSquaredDistanceBetweenPointAndRay(m_vertexData[i],r);
+			if (distToRay > maxD) {
+				maxD=distToRay;
+				maxI=i;
+			}
+		}
+		if (maxD == m_epsilonSquared) {
+			// It appears that the point cloud belongs to a 1 dimensional subspace of R^3: convex hull has no volume => return a thin triangle
+			// Pick any point other than selectedPoints.first and selectedPoints.second as the third point of the triangle
+			auto it = std::find_if(m_vertexData.begin(),m_vertexData.end(),[&](const vec3& ve) {
+				return ve != m_vertexData[selectedPoints.first] && ve != m_vertexData[selectedPoints.second];
+			});
+			const size_t thirdPoint = (it == m_vertexData.end()) ? selectedPoints.first : std::distance(m_vertexData.begin(),it);
+			it = std::find_if(m_vertexData.begin(),m_vertexData.end(),[&](const vec3& ve) {
+				return ve != m_vertexData[selectedPoints.first] && ve != m_vertexData[selectedPoints.second] && ve != m_vertexData[thirdPoint];
+			});
+			const size_t fourthPoint = (it == m_vertexData.end()) ? selectedPoints.first : std::distance(m_vertexData.begin(),it);
+			return m_mesh.setup(selectedPoints.first,selectedPoints.second,thirdPoint,fourthPoint);
+		}
+
+		// These three points form the base triangle for our tetrahedron.
+		assert(selectedPoints.first != maxI && selectedPoints.second != maxI);
+		std::array<size_t,3> baseTriangle{selectedPoints.first, selectedPoints.second, maxI};
+		const Vector3<T> baseTriangleVertices[]={ m_vertexData[baseTriangle[0]], m_vertexData[baseTriangle[1]],  m_vertexData[baseTriangle[2]] };
+		
+		// Next step is to find the 4th vertex of the tetrahedron. We naturally choose the point farthest away from the triangle plane.
+		maxD=m_epsilon;
+		maxI=0;
+		const Vector3<T> N = mathutils::getTriangleNormal(baseTriangleVertices[0],baseTriangleVertices[1],baseTriangleVertices[2]);
+		Plane<T> trianglePlane(N,baseTriangleVertices[0]);
+		for (size_t i=0;i<vCount;i++) {
+			const T d = std::abs(mathutils::getSignedDistanceToPlane(m_vertexData[i],trianglePlane));
+			if (d>maxD) {
+				maxD=d;
+				maxI=i;
+			}
+		}
+		if (maxD == m_epsilon) {
+			// All the points seem to lie on a 2D subspace of R^3. How to handle this? Well, let's add one extra point to the point cloud so that the convex hull will have volume.
+			m_planar = true;
+			const vec3 N1 = mathutils::getTriangleNormal(baseTriangleVertices[1],baseTriangleVertices[2],baseTriangleVertices[0]);
+			m_planarPointCloudTemp.clear();
+			m_planarPointCloudTemp.insert(m_planarPointCloudTemp.begin(),m_vertexData.begin(),m_vertexData.end());
+			const vec3 extraPoint = N1 + m_vertexData[0];
+			m_planarPointCloudTemp.push_back(extraPoint);
+			maxI = m_planarPointCloudTemp.size()-1;
+			m_vertexData = VertexDataSource<T>(m_planarPointCloudTemp);
+		}
+
+		// Enforce CCW orientation (if user prefers clockwise orientation, swap two vertices in each triangle when final mesh is created)
+		const Plane<T> triPlane(N,baseTriangleVertices[0]);
+		if (triPlane.isPointOnPositiveSide(m_vertexData[maxI])) {
+			std::swap(baseTriangle[0],baseTriangle[1]);
+		}
+
+		// Create a tetrahedron half edge mesh and compute planes defined by each triangle
+		m_mesh.setup(baseTriangle[0],baseTriangle[1],baseTriangle[2],maxI);
+		for (auto& f : m_mesh.m_faces) {
+			auto v = m_mesh.getVertexIndicesOfFace(f);
+			const Vector3<T>& va = m_vertexData[v[0]];
+			const Vector3<T>& vb = m_vertexData[v[1]];
+			const Vector3<T>& vc = m_vertexData[v[2]];
+			const Vector3<T> N1 = mathutils::getTriangleNormal(va, vb, vc);
+			const Plane<T> plane(N1,va);
+			f.m_P = plane;
+		}
+
+		// Finally we assign a face for each vertex outside the tetrahedron (vertices inside the tetrahedron have no role anymore)
+		for (size_t i=0;i<vCount;i++) {
+			for (auto& face : m_mesh.m_faces) {
+				if (addPointToFace(face, i)) {
+					break;
+				}
+			}
+		}
+	}
+	
+	/*
+	 * Explicit template specifications for float and double
+	 */
+
+	template class QuickHull<float>;
+	template class QuickHull<double>;
+}
+
diff --git a/thirdparty/manifold/src/third_party/quickhull/QuickHull.hpp b/thirdparty/manifold/src/third_party/quickhull/QuickHull.hpp
new file mode 100644
index 000000000000..436c4d9f165c
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/QuickHull.hpp
@@ -0,0 +1,223 @@
+#ifndef QUICKHULL_HPP_
+#define QUICKHULL_HPP_
+#include <deque>
+#include <vector>
+#include <array>
+#include <limits>
+#include "Structs/Vector3.hpp"
+#include "Structs/Plane.hpp"
+#include "Structs/Pool.hpp"
+#include "Structs/Mesh.hpp"
+#include "ConvexHull.hpp"
+#include "HalfEdgeMesh.hpp"
+#include "MathUtils.hpp"
+
+/*
+ * Implementation of the 3D QuickHull algorithm by Antti Kuukka
+ *
+ * No copyrights. What follows is 100% Public Domain.
+ *
+ *
+ *
+ * INPUT:  a list of points in 3D space (for example, vertices of a 3D mesh)
+ *
+ * OUTPUT: a ConvexHull object which provides vertex and index buffers of the generated convex hull as a triangle mesh.
+ *
+ *
+ *
+ * The implementation is thread-safe if each thread is using its own QuickHull object.
+ *
+ *
+ * SUMMARY OF THE ALGORITHM:
+ *         - Create initial simplex (tetrahedron) using extreme points. We have four faces now and they form a convex mesh M.
+ *         - For each point, assign them to the first face for which they are on the positive side of (so each point is assigned to at most
+ *           one face). Points inside the initial tetrahedron are left behind now and no longer affect the calculations.
+ *         - Add all faces that have points assigned to them to Face Stack.
+ *         - Iterate until Face Stack is empty:
+ *              - Pop topmost face F from the stack
+ *              - From the points assigned to F, pick the point P that is farthest away from the plane defined by F.
+ *              - Find all faces of M that have P on their positive side. Let us call these the "visible faces".
+ *              - Because of the way M is constructed, these faces are connected. Solve their horizon edge loop.
+ *				- "Extrude to P": Create new faces by connecting P with the points belonging to the horizon edge. Add the new faces to M and remove the visible
+ *                faces from M.
+ *              - Each point that was assigned to visible faces is now assigned to at most one of the newly created faces.
+ *              - Those new faces that have points assigned to them are added to the top of Face Stack.
+ *          - M is now the convex hull.
+ *
+ * TO DO:
+ *  - Implement a proper 2D QuickHull and use that to solve the degenerate 2D case (when all the points lie on the same plane in 3D space).
+ * */
+
+namespace quickhull {
+	
+	struct DiagnosticsData {
+		size_t m_failedHorizonEdges; // How many times QuickHull failed to solve the horizon edge. Failures lead to degenerated convex hulls.
+		
+		DiagnosticsData() : m_failedHorizonEdges(0) { }
+	};
+
+	template<typename FloatType>
+	FloatType defaultEps();
+
+	template<typename FloatType>
+	class QuickHull {
+		using vec3 = Vector3<FloatType>;
+
+		FloatType m_epsilon, m_epsilonSquared, m_scale;
+		bool m_planar;
+		std::vector<vec3> m_planarPointCloudTemp;
+		VertexDataSource<FloatType> m_vertexData;
+		MeshBuilder<FloatType> m_mesh;
+		std::array<size_t,6> m_extremeValues;
+		DiagnosticsData m_diagnostics;
+
+		// Temporary variables used during iteration process
+		std::vector<size_t> m_newFaceIndices;
+		std::vector<size_t> m_newHalfEdgeIndices;
+		std::vector< std::unique_ptr<std::vector<size_t>> > m_disabledFacePointVectors;
+		std::vector<size_t> m_visibleFaces;
+		std::vector<size_t> m_horizonEdges;
+		struct FaceData {
+			size_t m_faceIndex;
+			size_t m_enteredFromHalfEdge; // If the face turns out not to be visible, this half edge will be marked as horizon edge
+			FaceData(size_t fi, size_t he) : m_faceIndex(fi),m_enteredFromHalfEdge(he) {}
+		};
+		std::vector<FaceData> m_possiblyVisibleFaces;
+		std::deque<size_t> m_faceList;
+
+		// Create a half edge mesh representing the base tetrahedron from which the QuickHull iteration proceeds. m_extremeValues must be properly set up when this is called.
+		void setupInitialTetrahedron();
+
+		// Given a list of half edges, try to rearrange them so that they form a loop. Return true on success.
+		bool reorderHorizonEdges(std::vector<size_t>& horizonEdges);
+		
+		// Find indices of extreme values (max x, min x, max y, min y, max z, min z) for the given point cloud
+		std::array<size_t,6> getExtremeValues();
+		
+		// Compute scale of the vertex data.
+		FloatType getScale(const std::array<size_t,6>& extremeValues);
+		
+		// Each face contains a unique pointer to a vector of indices. However, many - often most - faces do not have any points on the positive
+		// side of them especially at the the end of the iteration. When a face is removed from the mesh, its associated point vector, if such
+		// exists, is moved to the index vector pool, and when we need to add new faces with points on the positive side to the mesh,
+		// we reuse these vectors. This reduces the amount of std::vectors we have to deal with, and impact on performance is remarkable.
+		Pool<std::vector<size_t>> m_indexVectorPool;
+		inline std::unique_ptr<std::vector<size_t>> getIndexVectorFromPool();
+		inline void reclaimToIndexVectorPool(std::unique_ptr<std::vector<size_t>>& ptr);
+		
+		// Associates a point with a face if the point resides on the positive side of the plane. Returns true if the points was on the positive side.
+		inline bool addPointToFace(typename MeshBuilder<FloatType>::Face& f, size_t pointIndex);
+		
+		// This will update m_mesh from which we create the ConvexHull object that getConvexHull function returns
+		void createConvexHalfEdgeMesh();
+		
+		// Constructs the convex hull into a MeshBuilder object which can be converted to a ConvexHull or Mesh object
+		void buildMesh(const VertexDataSource<FloatType>& pointCloud, bool CCW, bool useOriginalIndices, FloatType eps);
+		
+		// The public getConvexHull functions will setup a VertexDataSource object and call this
+		ConvexHull<FloatType> getConvexHull(const VertexDataSource<FloatType>& pointCloud, bool CCW, bool useOriginalIndices, FloatType eps);
+	public:
+		// Computes convex hull for a given point cloud.
+		// Params:
+		//   pointCloud: a vector of of 3D points
+		//   CCW: whether the output mesh triangles should have CCW orientation
+		//   useOriginalIndices: should the output mesh use same vertex indices as the original point cloud. If this is false,
+		//      then we generate a new vertex buffer which contains only the vertices that are part of the convex hull.
+		//   eps: minimum distance to a plane to consider a point being on positive of it (for a point cloud with scale 1)
+		ConvexHull<FloatType> getConvexHull(const std::vector<Vector3<FloatType>>& pointCloud,
+											bool CCW,
+											bool useOriginalIndices,
+											FloatType eps = defaultEps<FloatType>());
+		
+		// Computes convex hull for a given point cloud.
+		// Params:
+		//   vertexData: pointer to the first 3D point of the point cloud
+		//   vertexCount: number of vertices in the point cloud
+		//   CCW: whether the output mesh triangles should have CCW orientation
+		//   useOriginalIndices: should the output mesh use same vertex indices as the original point cloud. If this is false,
+		//      then we generate a new vertex buffer which contains only the vertices that are part of the convex hull.
+		//   eps: minimum distance to a plane to consider a point being on positive side of it (for a point cloud with scale 1)
+		ConvexHull<FloatType> getConvexHull(const Vector3<FloatType>* vertexData,
+											size_t vertexCount,
+											bool CCW,
+											bool useOriginalIndices,
+											FloatType eps = defaultEps<FloatType>());
+		
+		// Computes convex hull for a given point cloud. This function assumes that the vertex data resides in memory
+		// in the following format: x_0,y_0,z_0,x_1,y_1,z_1,...
+		// Params:
+		//   vertexData: pointer to the X component of the first point of the point cloud.
+		//   vertexCount: number of vertices in the point cloud
+		//   CCW: whether the output mesh triangles should have CCW orientation
+		//   useOriginalIndices: should the output mesh use same vertex indices as the original point cloud. If this is false,
+		//      then we generate a new vertex buffer which contains only the vertices that are part of the convex hull.
+		//   eps: minimum distance to a plane to consider a point being on positive side of it (for a point cloud with scale 1)
+		ConvexHull<FloatType> getConvexHull(const FloatType* vertexData,
+											size_t vertexCount,
+											bool CCW,
+											bool useOriginalIndices,
+											FloatType eps = defaultEps<FloatType>());
+		
+		// Computes convex hull for a given point cloud. This function assumes that the vertex data resides in memory
+		// in the following format: x_0,y_0,z_0,x_1,y_1,z_1,...
+		// Params:
+		//   vertexData: pointer to the X component of the first point of the point cloud.
+		//   vertexCount: number of vertices in the point cloud
+		//   CCW: whether the output mesh triangles should have CCW orientation
+		//   eps: minimum distance to a plane to consider a point being on positive side of it (for a point cloud with scale 1)
+		// Returns:
+		//   Convex hull of the point cloud as a mesh object with half edge structure.
+		HalfEdgeMesh<FloatType, size_t> getConvexHullAsMesh(const FloatType* vertexData,
+															size_t vertexCount,
+															bool CCW,
+															FloatType eps = defaultEps<FloatType>());
+		
+		// Get diagnostics about last generated convex hull
+		const DiagnosticsData& getDiagnostics() {
+			return m_diagnostics;
+		}
+	};
+	
+	/*
+	 * Inline function definitions
+	 */
+	
+	template<typename T>
+	std::unique_ptr<std::vector<size_t>> QuickHull<T>::getIndexVectorFromPool() {
+		auto r = m_indexVectorPool.get();
+		r->clear();
+		return r;
+	}
+	
+	template<typename T>
+	void QuickHull<T>::reclaimToIndexVectorPool(std::unique_ptr<std::vector<size_t>>& ptr) {
+		const size_t oldSize = ptr->size();
+		if ((oldSize+1)*128 < ptr->capacity()) {
+			// Reduce memory usage! Huge vectors are needed at the beginning of iteration when faces have many points on their positive side. Later on, smaller vectors will suffice.
+			ptr.reset(nullptr);
+			return;
+		}
+		m_indexVectorPool.reclaim(ptr);
+	}
+
+	template<typename T>
+	bool QuickHull<T>::addPointToFace(typename MeshBuilder<T>::Face& f, size_t pointIndex) {
+		const T D = mathutils::getSignedDistanceToPlane(m_vertexData[ pointIndex ],f.m_P);
+		if (D>0 && D*D > m_epsilonSquared*f.m_P.m_sqrNLength) {
+			if (!f.m_pointsOnPositiveSide) {
+				f.m_pointsOnPositiveSide = std::move(getIndexVectorFromPool());
+			}
+			f.m_pointsOnPositiveSide->push_back( pointIndex );
+			if (D > f.m_mostDistantPointDist) {
+				f.m_mostDistantPointDist = D;
+				f.m_mostDistantPoint = pointIndex;
+			}
+			return true;
+		}
+		return false;
+	}
+
+}
+
+
+#endif /* QUICKHULL_HPP_ */
diff --git a/thirdparty/manifold/src/third_party/quickhull/README.md b/thirdparty/manifold/src/third_party/quickhull/README.md
new file mode 100644
index 000000000000..a36c081b7855
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/README.md
@@ -0,0 +1,31 @@
+This implementation is 100% Public Domain.
+
+Feel free to use.
+ 
+C++11 is needed to compile it.
+
+Basic usage:
+
+	#include "quickhull/quickhull.hpp"
+
+	using namespace quickhull;
+	QuickHull<float> qh; // Could be double as well
+	std::vector<Vector3<float>> pointCloud;
+	// Add points to point cloud
+	...
+	auto hull = qh.getConvexHull(pointCloud, true, false);
+	const auto& indexBuffer = hull.getIndexBuffer();
+	const auto& vertexBuffer = hull.getVertexBuffer();
+	// Do what you want with the convex triangle mesh
+
+Vertex data can be passed as a pointer to float/double as long as the data is in X_0,Y_0,Z_0,X_1,Y_1,Z_1,...,X_N,Y_N_Z_N format:
+
+	auto hull = qh.getConvexHull(&pointCloud[0].x, pointCloud.size(), true, false);
+
+The first boolean parameter of getConvexHull specifies whether the resulting mesh should have its triangles in CCW orientation.
+
+The second boolean parameter specifies whether the mesh should use vertex indices of the original point cloud. If it is false, a new vertex buffer is generated which consists only of those vertices that are part of the convex hull. In this case, the new vertex buffer is owned by the returned ConvexHull object. Otherwise, the original point cloud is used as vertex buffer and since the vertices are not copied, make sure you don't call ConvexHull::getVertexBuffer after releasing the memory that contains the original point cloud data.
+
+This implementation is fast, because the convex hull is internally built using a half edge mesh representation which provides quick access to adjacent faces. It is also possible to get the output convex hull as a half edge mesh:
+
+	auto mesh = qh.getConvexHullAsMesh(&pointCloud[0].x, pointCloud.size(), true);
diff --git a/thirdparty/manifold/src/third_party/quickhull/Structs/Mesh.hpp b/thirdparty/manifold/src/third_party/quickhull/Structs/Mesh.hpp
new file mode 100644
index 000000000000..f58c5afcdfb7
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/Structs/Mesh.hpp
@@ -0,0 +1,255 @@
+#ifndef MESH_HPP_
+#define MESH_HPP_
+
+#include <vector>
+#include "Vector3.hpp"
+#include "Plane.hpp"
+#include "Pool.hpp"
+#include <array>
+#include <cassert>
+#include <limits>
+#include <memory>
+#include "VertexDataSource.hpp"
+#include <unordered_map>
+#include <cinttypes>
+
+namespace quickhull {
+
+	template <typename T>
+	class MeshBuilder {
+	public:
+		struct HalfEdge {
+			size_t m_endVertex;
+			size_t m_opp;
+			size_t m_face;
+			size_t m_next;
+			
+			void disable() {
+				m_endVertex = std::numeric_limits<size_t>::max();
+			}
+			
+			bool isDisabled() const {
+				return m_endVertex == std::numeric_limits<size_t>::max();
+			}
+		};
+
+		struct Face {
+			size_t m_he;
+			Plane<T> m_P{};
+			T m_mostDistantPointDist;
+			size_t m_mostDistantPoint;
+			size_t m_visibilityCheckedOnIteration;
+			std::uint8_t m_isVisibleFaceOnCurrentIteration : 1;
+			std::uint8_t m_inFaceStack : 1;
+			std::uint8_t m_horizonEdgesOnCurrentIteration : 3; // Bit for each half edge assigned to this face, each being 0 or 1 depending on whether the edge belongs to horizon edge
+			std::unique_ptr<std::vector<size_t>> m_pointsOnPositiveSide;
+
+			Face() : m_he(std::numeric_limits<size_t>::max()),
+					 m_mostDistantPointDist(0),
+					 m_mostDistantPoint(0),
+					 m_visibilityCheckedOnIteration(0),
+					 m_isVisibleFaceOnCurrentIteration(0),
+					 m_inFaceStack(0),
+					 m_horizonEdgesOnCurrentIteration(0)
+			{
+
+			}
+			
+			void disable() {
+				m_he = std::numeric_limits<size_t>::max();
+			}
+
+			bool isDisabled() const {
+				return m_he == std::numeric_limits<size_t>::max();
+			}
+		};
+
+		// Mesh data
+		std::vector<Face> m_faces;
+		std::vector<HalfEdge> m_halfEdges;
+		
+		// When the mesh is modified and faces and half edges are removed from it, we do not actually remove them from the container vectors.
+		// Insted, they are marked as disabled which means that the indices can be reused when we need to add new faces and half edges to the mesh.
+		// We store the free indices in the following vectors.
+		std::vector<size_t> m_disabledFaces,m_disabledHalfEdges;
+		
+		size_t addFace() {
+			if (m_disabledFaces.size()) {
+				size_t index = m_disabledFaces.back();
+				auto& f = m_faces[index];
+				assert(f.isDisabled());
+				assert(!f.m_pointsOnPositiveSide);
+				f.m_mostDistantPointDist = 0;
+				m_disabledFaces.pop_back();
+				return index;
+			}
+			m_faces.emplace_back();
+			return m_faces.size()-1;
+		}
+
+		size_t addHalfEdge()	{
+			if (m_disabledHalfEdges.size()) {
+				const size_t index = m_disabledHalfEdges.back();
+				m_disabledHalfEdges.pop_back();
+				return index;
+			}
+			m_halfEdges.emplace_back();
+			return m_halfEdges.size()-1;
+		}
+
+		// Mark a face as disabled and return a pointer to the points that were on the positive of it.
+		std::unique_ptr<std::vector<size_t>> disableFace(size_t faceIndex) {
+			auto& f = m_faces[faceIndex];
+			f.disable();
+			m_disabledFaces.push_back(faceIndex);
+			return std::move(f.m_pointsOnPositiveSide);
+		}
+
+		void disableHalfEdge(size_t heIndex) {
+			auto& he = m_halfEdges[heIndex];
+			he.disable();
+			m_disabledHalfEdges.push_back(heIndex);
+		}
+
+		MeshBuilder() = default;
+		
+		// Create a mesh with initial tetrahedron ABCD. Dot product of AB with the normal of triangle ABC should be negative.
+		void setup(size_t a, size_t b, size_t c, size_t d) {
+			m_faces.clear();
+			m_halfEdges.clear();
+			m_disabledFaces.clear();
+			m_disabledHalfEdges.clear();
+			
+			m_faces.reserve(4);
+			m_halfEdges.reserve(12);
+			
+			// Create halfedges
+			HalfEdge AB;
+			AB.m_endVertex = b;
+			AB.m_opp = 6;
+			AB.m_face = 0;
+			AB.m_next = 1;
+			m_halfEdges.push_back(AB);
+
+			HalfEdge BC;
+			BC.m_endVertex = c;
+			BC.m_opp = 9;
+			BC.m_face = 0;
+			BC.m_next = 2;
+			m_halfEdges.push_back(BC);
+
+			HalfEdge CA;
+			CA.m_endVertex = a;
+			CA.m_opp = 3;
+			CA.m_face = 0;
+			CA.m_next = 0;
+			m_halfEdges.push_back(CA);
+
+			HalfEdge AC;
+			AC.m_endVertex = c;
+			AC.m_opp = 2;
+			AC.m_face = 1;
+			AC.m_next = 4;
+			m_halfEdges.push_back(AC);
+
+			HalfEdge CD;
+			CD.m_endVertex = d;
+			CD.m_opp = 11;
+			CD.m_face = 1;
+			CD.m_next = 5;
+			m_halfEdges.push_back(CD);
+
+			HalfEdge DA;
+			DA.m_endVertex = a;
+			DA.m_opp = 7;
+			DA.m_face = 1;
+			DA.m_next = 3;
+			m_halfEdges.push_back(DA);
+
+			HalfEdge BA;
+			BA.m_endVertex = a;
+			BA.m_opp = 0;
+			BA.m_face = 2;
+			BA.m_next = 7;
+			m_halfEdges.push_back(BA);
+
+			HalfEdge AD;
+			AD.m_endVertex = d;
+			AD.m_opp = 5;
+			AD.m_face = 2;
+			AD.m_next = 8;
+			m_halfEdges.push_back(AD);
+
+			HalfEdge DB;
+			DB.m_endVertex = b;
+			DB.m_opp = 10;
+			DB.m_face = 2;
+			DB.m_next = 6;
+			m_halfEdges.push_back(DB);
+
+			HalfEdge CB;
+			CB.m_endVertex = b;
+			CB.m_opp = 1;
+			CB.m_face = 3;
+			CB.m_next = 10;
+			m_halfEdges.push_back(CB);
+
+			HalfEdge BD;
+			BD.m_endVertex = d;
+			BD.m_opp = 8;
+			BD.m_face = 3;
+			BD.m_next = 11;
+			m_halfEdges.push_back(BD);
+
+			HalfEdge DC;
+			DC.m_endVertex = c;
+			DC.m_opp = 4;
+			DC.m_face = 3;
+			DC.m_next = 9;
+			m_halfEdges.push_back(DC);
+
+			// Create faces
+			Face ABC;
+			ABC.m_he = 0;
+			m_faces.push_back(std::move(ABC));
+
+			Face ACD;
+			ACD.m_he = 3;
+			m_faces.push_back(std::move(ACD));
+
+			Face BAD;
+			BAD.m_he = 6;
+			m_faces.push_back(std::move(BAD));
+
+			Face CBD;
+			CBD.m_he = 9;
+			m_faces.push_back(std::move(CBD));
+		}
+
+		std::array<size_t,3> getVertexIndicesOfFace(const Face& f) const {
+			std::array<size_t,3> v;
+			const HalfEdge* he = &m_halfEdges[f.m_he];
+			v[0] = he->m_endVertex;
+			he = &m_halfEdges[he->m_next];
+			v[1] = he->m_endVertex;
+			he = &m_halfEdges[he->m_next];
+			v[2] = he->m_endVertex;
+			return v;
+		}
+
+		std::array<size_t,2> getVertexIndicesOfHalfEdge(const HalfEdge& he) const {
+			return {m_halfEdges[he.m_opp].m_endVertex,he.m_endVertex};
+		}
+
+		std::array<size_t,3> getHalfEdgeIndicesOfFace(const Face& f) const {
+			return {f.m_he,m_halfEdges[f.m_he].m_next,m_halfEdges[m_halfEdges[f.m_he].m_next].m_next};
+		}
+	};
+	
+
+
+}
+
+
+
+#endif 
diff --git a/thirdparty/manifold/src/third_party/quickhull/Structs/Plane.hpp b/thirdparty/manifold/src/third_party/quickhull/Structs/Plane.hpp
new file mode 100644
index 000000000000..903a7efe3630
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/Structs/Plane.hpp
@@ -0,0 +1,36 @@
+#ifndef QHPLANE_HPP_
+#define QHPLANE_HPP_
+
+#include "Vector3.hpp"
+
+namespace quickhull {
+
+	template<typename T>
+	class Plane {
+	public:
+		Vector3<T> m_N;
+		
+		// Signed distance (if normal is of length 1) to the plane from origin
+		T m_D;
+		
+		// Normal length squared
+		T m_sqrNLength;
+
+		bool isPointOnPositiveSide(const Vector3<T>& Q) const {
+			T d = m_N.dotProduct(Q)+m_D;
+			if (d>=0) return true;
+			return false;
+		}
+
+		Plane() = default;
+
+		// Construct a plane using normal N and any point P on the plane
+		Plane(const Vector3<T>& N, const Vector3<T>& P) : m_N(N), m_D(-N.dotProduct(P)), m_sqrNLength(m_N.x*m_N.x+m_N.y*m_N.y+m_N.z*m_N.z) {
+			
+		}
+	};
+
+}
+
+
+#endif /* PLANE_HPP_ */
diff --git a/thirdparty/manifold/src/third_party/quickhull/Structs/Pool.hpp b/thirdparty/manifold/src/third_party/quickhull/Structs/Pool.hpp
new file mode 100644
index 000000000000..4216b4a15339
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/Structs/Pool.hpp
@@ -0,0 +1,35 @@
+#ifndef Pool_h
+#define Pool_h
+
+#include <vector>
+#include <memory>
+
+namespace quickhull {
+	
+	template<typename T>
+	class Pool {
+		std::vector<std::unique_ptr<T>> m_data;
+	public:
+		void clear() {
+			m_data.clear();
+		}
+		
+		void reclaim(std::unique_ptr<T>& ptr) {
+			m_data.push_back(std::move(ptr));
+		}
+		
+		std::unique_ptr<T> get() {
+			if (m_data.size()==0) {
+				return std::unique_ptr<T>(new T());
+			}
+			auto it = m_data.end()-1;
+			std::unique_ptr<T> r = std::move(*it);
+			m_data.erase(it);
+			return r;
+		}
+		
+	};
+	
+}
+
+#endif /* Pool_h */
diff --git a/thirdparty/manifold/src/third_party/quickhull/Structs/Ray.hpp b/thirdparty/manifold/src/third_party/quickhull/Structs/Ray.hpp
new file mode 100644
index 000000000000..19b60728e5b1
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/Structs/Ray.hpp
@@ -0,0 +1,21 @@
+#ifndef QuickHull_Ray_hpp
+#define QuickHull_Ray_hpp
+
+#include "Vector3.hpp"
+
+namespace quickhull {
+
+	template <typename T>
+	struct Ray {
+		const Vector3<T> m_S;
+		const Vector3<T> m_V;
+		const T m_VInvLengthSquared;
+		
+		Ray(const Vector3<T>& S,const Vector3<T>& V) : m_S(S), m_V(V), m_VInvLengthSquared(1/m_V.getLengthSquared()) {
+		}
+	};
+	
+}
+
+
+#endif
diff --git a/thirdparty/manifold/src/third_party/quickhull/Structs/Vector3.hpp b/thirdparty/manifold/src/third_party/quickhull/Structs/Vector3.hpp
new file mode 100644
index 000000000000..47a5f74617dc
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/Structs/Vector3.hpp
@@ -0,0 +1,140 @@
+#ifndef QuickHull_Vector3_hpp
+#define QuickHull_Vector3_hpp
+
+#include <cmath>
+#include <iostream>
+
+namespace quickhull {
+	
+	template <typename T>
+	class Vector3
+	{
+	public:
+		Vector3() = default;
+		
+		Vector3(T x, T y, T z) : x(x), y(y), z(z) {
+			
+		}
+		
+		T x,y,z;
+		
+		T dotProduct(const Vector3& other) const {
+			return x*other.x+y*other.y+z*other.z;
+		}
+		
+		void normalize() {
+			const T len = getLength();
+			x/=len;
+			y/=len;
+			z/=len;
+		}
+		
+		Vector3 getNormalized() const {
+			const T len = getLength();
+			return Vector3(x/len,y/len,z/len);
+		}
+		
+		T getLength() const {
+			return std::sqrt(x*x+y*y+z*z);
+		}
+		
+		Vector3 operator-(const Vector3& other) const {
+			return Vector3(x-other.x,y-other.y,z-other.z);
+		}
+		
+		Vector3 operator+(const Vector3& other) const {
+			return Vector3(x+other.x,y+other.y,z+other.z);
+		}
+		
+		Vector3& operator+=(const Vector3& other) {
+			x+=other.x;
+			y+=other.y;
+			z+=other.z;
+			return *this;
+		}
+		Vector3& operator-=(const Vector3& other) {
+			x-=other.x;
+			y-=other.y;
+			z-=other.z;
+			return *this;
+		}
+		Vector3& operator*=(T c) {
+			x*=c;
+			y*=c;
+			z*=c;
+			return *this;
+		}
+		
+		Vector3& operator/=(T c) {
+			x/=c;
+			y/=c;
+			z/=c;
+			return *this;
+		}
+		
+		Vector3 operator-() const {
+			return Vector3(-x,-y,-z);
+		}
+
+		template<typename S>
+		Vector3 operator*(S c) const {
+			return Vector3(x*c,y*c,z*c);
+		}
+		
+		template<typename S>
+		Vector3 operator/(S c) const {
+			return Vector3(x/c,y/c,z/c);
+		}
+		
+		T getLengthSquared() const {
+			return x*x + y*y + z*z;
+		}
+		
+		bool operator!=(const Vector3& o) const {
+			return x != o.x || y != o.y || z != o.z;
+		}
+		
+		// Projection onto another vector
+		Vector3 projection(const Vector3& o) const {
+			T C = dotProduct(o)/o.getLengthSquared();
+			return o*C;
+		}
+		
+		Vector3 crossProduct (const Vector3& rhs ) {
+			T a = y * rhs.z - z * rhs.y ;
+			T b = z * rhs.x - x * rhs.z ;
+			T c = x * rhs.y - y * rhs.x ;
+			Vector3 product( a , b , c ) ;
+			return product ;
+		}
+		
+		T getDistanceTo(const Vector3& other) const {
+			Vector3 diff = *this - other;
+			return diff.getLength();
+		}
+		
+		T getSquaredDistanceTo(const Vector3& other) const {
+			const T dx = x-other.x;
+			const T dy = y-other.y;
+			const T dz = z-other.z;
+			return dx*dx+dy*dy+dz*dz;
+		}
+		
+	};
+	
+	// Overload also << operator for easy printing of debug data
+	template <typename T>
+	inline std::ostream& operator<<(std::ostream& os, const Vector3<T>& vec) {
+		os << "(" << vec.x << "," << vec.y << "," << vec.z << ")";
+		return os;
+	}
+	
+	template <typename T>
+	inline Vector3<T> operator*(T c, const Vector3<T>& v) {
+		return Vector3<T>(v.x*c,v.y*c,v.z*c);
+	}
+	
+}
+
+
+#endif
diff --git a/thirdparty/manifold/src/third_party/quickhull/Structs/VertexDataSource.hpp b/thirdparty/manifold/src/third_party/quickhull/Structs/VertexDataSource.hpp
new file mode 100644
index 000000000000..162b643bd096
--- /dev/null
+++ b/thirdparty/manifold/src/third_party/quickhull/Structs/VertexDataSource.hpp
@@ -0,0 +1,48 @@
+#ifndef VertexDataSource_h
+#define VertexDataSource_h
+
+#include "Vector3.hpp"
+
+namespace quickhull {
+	
+	template<typename T>
+	class VertexDataSource {
+		const Vector3<T>* m_ptr;
+		size_t m_count;
+	
+	public:
+		VertexDataSource(const Vector3<T>* ptr, size_t count) : m_ptr(ptr), m_count(count) {
+			
+		}
+		
+		VertexDataSource(const std::vector<Vector3<T>>& vec) : m_ptr(&vec[0]), m_count(vec.size()) {
+			
+		}
+		
+		VertexDataSource() : m_ptr(nullptr), m_count(0) {
+			
+		}
+		
+		VertexDataSource& operator=(const VertexDataSource& other) = default;
+		
+		size_t size() const {
+			return m_count;
+		}
+		
+		const Vector3<T>& operator[](size_t index) const {
+			return m_ptr[index];
+		}
+		
+		const Vector3<T>* begin() const {
+			return m_ptr;
+		}
+		
+		const Vector3<T>* end() const {
+			return m_ptr + m_count;
+		}
+	};
+	
+}
+
+
+#endif /* VertexDataSource_h */
diff --git a/thirdparty/manifold/src/utilities/include/hashtable.h b/thirdparty/manifold/src/utilities/include/hashtable.h
new file mode 100644
index 000000000000..ed0f5538d2bc
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/hashtable.h
@@ -0,0 +1,174 @@
+// Copyright 2022 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+#pragma once
+#include <stdint.h>
+
+#include <atomic>
+
+#include "public.h"
+#include "utils.h"
+#include "vec.h"
+
+namespace {
+typedef unsigned long long int Uint64;
+typedef Uint64 (*hash_fun_t)(Uint64);
+inline constexpr Uint64 kOpen = std::numeric_limits<Uint64>::max();
+
+template <typename T>
+T AtomicCAS(T& target, T compare, T val) {
+  std::atomic<T>& tar = reinterpret_cast<std::atomic<T>&>(target);
+  tar.compare_exchange_strong(compare, val, std::memory_order_acq_rel);
+  return compare;
+}
+
+template <typename T>
+void AtomicStore(T& target, T val) {
+  std::atomic<T>& tar = reinterpret_cast<std::atomic<T>&>(target);
+  // release is good enough, although not really something general
+  tar.store(val, std::memory_order_release);
+}
+
+template <typename T>
+T AtomicLoad(const T& target) {
+  const std::atomic<T>& tar = reinterpret_cast<const std::atomic<T>&>(target);
+  // acquire is good enough, although not general
+  return tar.load(std::memory_order_acquire);
+}
+
+// https://stackoverflow.com/questions/664014/what-integer-hash-function-are-good-that-accepts-an-integer-hash-key
+inline Uint64 hash64bit(Uint64 x) {
+  x = (x ^ (x >> 30)) * 0xbf58476d1ce4e5b9ull;
+  x = (x ^ (x >> 27)) * 0x94d049bb133111ebull;
+  x = x ^ (x >> 31);
+  return x;
+}
+}  // namespace
+
+namespace manifold {
+/** @addtogroup Private
+ *  @{
+ */
+
+template <typename V, hash_fun_t H = hash64bit>
+class HashTableD {
+ public:
+  HashTableD(Vec<Uint64>& keys, Vec<V>& values, std::atomic<size_t>& used,
+             uint32_t step = 1)
+      : step_{step}, keys_{keys}, values_{values}, used_{used} {}
+
+  int Size() const { return keys_.size(); }
+
+  bool Full() const {
+    return used_.load(std::memory_order_relaxed) * 2 > Size();
+  }
+
+  void Insert(Uint64 key, const V& val) {
+    uint32_t idx = H(key) & (Size() - 1);
+    while (1) {
+      if (Full()) return;
+      Uint64& k = keys_[idx];
+      const Uint64 found = AtomicCAS(k, kOpen, key);
+      if (found == kOpen) {
+        used_.fetch_add(1, std::memory_order_relaxed);
+        values_[idx] = val;
+        return;
+      }
+      if (found == key) return;
+      idx = (idx + step_) & (Size() - 1);
+    }
+  }
+
+  V& operator[](Uint64 key) {
+    uint32_t idx = H(key) & (Size() - 1);
+    while (1) {
+      const Uint64 k = AtomicLoad(keys_[idx]);
+      if (k == key || k == kOpen) {
+        return values_[idx];
+      }
+      idx = (idx + step_) & (Size() - 1);
+    }
+  }
+
+  const V& operator[](Uint64 key) const {
+    uint32_t idx = H(key) & (Size() - 1);
+    while (1) {
+      const Uint64 k = AtomicLoad(keys_[idx]);
+      if (k == key || k == kOpen) {
+        return values_[idx];
+      }
+      idx = (idx + step_) & (Size() - 1);
+    }
+  }
+
+  Uint64 KeyAt(int idx) const { return AtomicLoad(keys_[idx]); }
+  V& At(int idx) { return values_[idx]; }
+  const V& At(int idx) const { return values_[idx]; }
+
+ private:
+  uint32_t step_;
+  VecView<Uint64> keys_;
+  VecView<V> values_;
+  std::atomic<size_t>& used_;
+};
+
+template <typename V, hash_fun_t H = hash64bit>
+class HashTable {
+ public:
+  HashTable(size_t size, uint32_t step = 1)
+      : keys_{size == 0 ? 0 : 1_z << (int)ceil(log2(size)), kOpen},
+        values_{size == 0 ? 0 : 1_z << (int)ceil(log2(size)), {}},
+        step_(step) {}
+
+  HashTable(const HashTable& other)
+      : keys_(other.keys_),
+        values_(other.values_),
+        used_(other.used_),
+        step_(other.step_) {}
+
+  HashTable& operator=(const HashTable& other) {
+    if (this == &other) return *this;
+    keys_ = other.keys_;
+    values_ = other.values_;
+    used_.store(other.used_.load());
+    step_ = other.step_;
+    return *this;
+  }
+
+  HashTableD<V, H> D() { return {keys_, values_, used_, step_}; }
+
+  int Entries() const { return used_.load(std::memory_order_relaxed); }
+
+  size_t Size() const { return keys_.size(); }
+
+  bool Full() const {
+    return used_.load(std::memory_order_relaxed) * 2 > Size();
+  }
+
+  float FilledFraction() const {
+    return static_cast<float>(used_.load(std::memory_order_relaxed)) / Size();
+  }
+
+  Vec<V>& GetValueStore() { return values_; }
+
+  static Uint64 Open() { return kOpen; }
+
+ private:
+  Vec<Uint64> keys_;
+  Vec<V> values_;
+  std::atomic<size_t> used_ = 0;
+  uint32_t step_;
+};
+
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/optional_assert.h b/thirdparty/manifold/src/utilities/include/optional_assert.h
new file mode 100644
index 000000000000..75f32684a12d
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/optional_assert.h
@@ -0,0 +1,38 @@
+// Copyright 2022 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "public.h"
+
+#ifdef MANIFOLD_DEBUG
+#include <iostream>
+#include <sstream>
+#include <string>
+
+template <typename Ex>
+void Assert(bool condition, const char* file, int line, const std::string& cond,
+            const std::string& msg) {
+  if (!condition) {
+    std::ostringstream output;
+    output << "Error in file: " << file << " (" << line << "): \'" << cond
+           << "\' is false: " << msg;
+    throw Ex(output.str());
+  }
+}
+#define ASSERT(condition, EX, msg) \
+  Assert<EX>(condition, __FILE__, __LINE__, #condition, msg);
+#else
+#define ASSERT(condition, EX, msg)
+#endif
\ No newline at end of file
diff --git a/thirdparty/manifold/src/utilities/include/par.h b/thirdparty/manifold/src/utilities/include/par.h
new file mode 100644
index 000000000000..2049a5011d70
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/par.h
@@ -0,0 +1,195 @@
+// Copyright 2022 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <thrust/binary_search.h>
+#include <thrust/count.h>
+#include <thrust/execution_policy.h>
+#include <thrust/gather.h>
+#include <thrust/logical.h>
+#include <thrust/remove.h>
+#include <thrust/sequence.h>
+#include <thrust/sort.h>
+#include <thrust/system/cpp/execution_policy.h>
+#include <thrust/uninitialized_copy.h>
+
+#include <algorithm>
+#include <numeric>
+#if MANIFOLD_PAR == 'T'
+#include <thrust/system/tbb/execution_policy.h>
+
+#if MANIFOLD_PAR == 'T' && TBB_INTERFACE_VERSION >= 10000 && \
+    __has_include(<pstl/glue_execution_defs.h>)
+#include <execution>
+#endif
+
+#include "tbb/tbb.h"
+#define MANIFOLD_PAR_NS tbb
+#else
+#define MANIFOLD_PAR_NS cpp
+#endif
+
+namespace manifold {
+
+enum class ExecutionPolicy {
+  Par,
+  Seq,
+};
+
+// ExecutionPolicy:
+// - Sequential for small workload,
+// - Parallel (CPU) for medium workload,
+// - GPU for large workload if available.
+inline constexpr ExecutionPolicy autoPolicy(size_t size) {
+  // some random numbers
+  if (size <= (1 << 12)) {
+    return ExecutionPolicy::Seq;
+  }
+  return ExecutionPolicy::Par;
+}
+
+#define THRUST_DYNAMIC_BACKEND_VOID(NAME)                    \
+  template <typename... Args>                                \
+  void NAME(ExecutionPolicy policy, Args... args) {          \
+    switch (policy) {                                        \
+      case ExecutionPolicy::Par:                             \
+        thrust::NAME(thrust::MANIFOLD_PAR_NS::par, args...); \
+        break;                                               \
+      case ExecutionPolicy::Seq:                             \
+        thrust::NAME(thrust::cpp::par, args...);             \
+        break;                                               \
+    }                                                        \
+  }
+
+#define THRUST_DYNAMIC_BACKEND(NAME, RET)                           \
+  template <typename Ret = RET, typename... Args>                   \
+  Ret NAME(ExecutionPolicy policy, Args... args) {                  \
+    switch (policy) {                                               \
+      case ExecutionPolicy::Par:                                    \
+        return thrust::NAME(thrust::MANIFOLD_PAR_NS::par, args...); \
+      case ExecutionPolicy::Seq:                                    \
+        break;                                                      \
+    }                                                               \
+    return thrust::NAME(thrust::cpp::par, args...);                 \
+  }
+
+#if MANIFOLD_PAR != 'T' || \
+    (TBB_INTERFACE_VERSION >= 10000 && __has_include(<pstl/glue_execution_defs.h>))
+#if MANIFOLD_PAR == 'T'
+#define STL_DYNAMIC_BACKEND(NAME, RET)                        \
+  template <typename Ret = RET, typename... Args>             \
+  Ret NAME(ExecutionPolicy policy, Args... args) {            \
+    switch (policy) {                                         \
+      case ExecutionPolicy::Par:                              \
+        return std::NAME(std::execution::par_unseq, args...); \
+      case ExecutionPolicy::Seq:                              \
+        break;                                                \
+    }                                                         \
+    return std::NAME(args...);                                \
+  }
+#define STL_DYNAMIC_BACKEND_VOID(NAME)                 \
+  template <typename... Args>                          \
+  void NAME(ExecutionPolicy policy, Args... args) {    \
+    switch (policy) {                                  \
+      case ExecutionPolicy::Par:                       \
+        std::NAME(std::execution::par_unseq, args...); \
+        break;                                         \
+      case ExecutionPolicy::Seq:                       \
+        std::NAME(args...);                            \
+        break;                                         \
+    }                                                  \
+  }
+#else
+#define STL_DYNAMIC_BACKEND(NAME, RET)             \
+  template <typename Ret = RET, typename... Args>  \
+  Ret NAME(ExecutionPolicy policy, Args... args) { \
+    return std::NAME(args...);                     \
+  }
+#define STL_DYNAMIC_BACKEND_VOID(NAME)              \
+  template <typename... Args>                       \
+  void NAME(ExecutionPolicy policy, Args... args) { \
+    std::NAME(args...);                             \
+  }
+#endif
+
+template <typename... Args>
+void exclusive_scan(ExecutionPolicy policy, Args... args) {
+  // https://github.com/llvm/llvm-project/issues/59810
+  std::exclusive_scan(args...);
+}
+template <typename DerivedPolicy, typename InputIterator1,
+          typename InputIterator2, typename OutputIterator, typename Predicate>
+OutputIterator copy_if(ExecutionPolicy policy, InputIterator1 first,
+                       InputIterator1 last, InputIterator2 stencil,
+                       OutputIterator result, Predicate pred) {
+  if (policy == ExecutionPolicy::Seq)
+    return thrust::copy_if(thrust::cpp::par, first, last, stencil, result,
+                           pred);
+  else
+    // note: this is not a typo, see
+    // https://github.com/NVIDIA/thrust/issues/1977
+    return thrust::copy_if(first, last, stencil, result, pred);
+}
+template <typename DerivedPolicy, typename InputIterator1,
+          typename OutputIterator, typename Predicate>
+OutputIterator copy_if(ExecutionPolicy policy, InputIterator1 first,
+                       InputIterator1 last, OutputIterator result,
+                       Predicate pred) {
+#if MANIFOLD_PAR == 'T'
+  if (policy == ExecutionPolicy::Seq)
+    return std::copy_if(first, last, result, pred);
+  else
+    return std::copy_if(std::execution::par_unseq, first, last, result, pred);
+#else
+  return std::copy_if(first, last, result, pred);
+#endif
+}
+
+#else
+#define STL_DYNAMIC_BACKEND(NAME, RET) THRUST_DYNAMIC_BACKEND(NAME, RET)
+#define STL_DYNAMIC_BACKEND_VOID(NAME) THRUST_DYNAMIC_BACKEND_VOID(NAME)
+
+THRUST_DYNAMIC_BACKEND_VOID(exclusive_scan)
+THRUST_DYNAMIC_BACKEND(copy_if, void)
+#endif
+
+THRUST_DYNAMIC_BACKEND_VOID(gather)
+THRUST_DYNAMIC_BACKEND_VOID(scatter)
+THRUST_DYNAMIC_BACKEND_VOID(for_each)
+THRUST_DYNAMIC_BACKEND_VOID(for_each_n)
+THRUST_DYNAMIC_BACKEND_VOID(sequence)
+STL_DYNAMIC_BACKEND_VOID(transform)
+STL_DYNAMIC_BACKEND_VOID(uninitialized_fill)
+STL_DYNAMIC_BACKEND_VOID(uninitialized_copy)
+STL_DYNAMIC_BACKEND_VOID(stable_sort)
+STL_DYNAMIC_BACKEND_VOID(fill)
+STL_DYNAMIC_BACKEND_VOID(copy)
+STL_DYNAMIC_BACKEND_VOID(inclusive_scan)
+STL_DYNAMIC_BACKEND_VOID(copy_n)
+
+// void implies that the user have to specify the return type in the template
+// argument, as we are unable to deduce it
+THRUST_DYNAMIC_BACKEND(transform_reduce, void)
+THRUST_DYNAMIC_BACKEND(gather_if, void)
+THRUST_DYNAMIC_BACKEND(reduce_by_key, void)
+STL_DYNAMIC_BACKEND(remove, void)
+STL_DYNAMIC_BACKEND(find, void)
+STL_DYNAMIC_BACKEND(find_if, void)
+STL_DYNAMIC_BACKEND(all_of, bool)
+STL_DYNAMIC_BACKEND(is_sorted, bool)
+STL_DYNAMIC_BACKEND(reduce, void)
+STL_DYNAMIC_BACKEND(count_if, int)
+STL_DYNAMIC_BACKEND(remove_if, void)
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/public.h b/thirdparty/manifold/src/utilities/include/public.h
new file mode 100644
index 000000000000..b9dfa9d283af
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/public.h
@@ -0,0 +1,696 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#define GLM_ENABLE_EXPERIMENTAL  // needed for glm/gtx/compatibility.hpp
+#define GLM_FORCE_EXPLICIT_CTOR
+#include <glm/ext/matrix_transform.hpp>
+#include <glm/glm.hpp>
+#include <glm/gtc/constants.hpp>
+#include <glm/gtx/compatibility.hpp>
+#include <glm/gtx/rotate_vector.hpp>
+#include <limits>
+#include <memory>
+#include <unordered_map>
+#include <vector>
+
+#ifdef MANIFOLD_DEBUG
+#include <iomanip>
+#include <iostream>
+#include <sstream>
+#endif
+
+constexpr std::size_t operator""_z(unsigned long long n) { return n; }
+
+namespace manifold {
+
+constexpr float kTolerance = 1e-5;
+
+/** @defgroup Connections
+ *  @brief Move data in and out of the Manifold class.
+ *  @{
+ */
+
+/**
+ * Sine function where multiples of 90 degrees come out exact.
+ *
+ * @param x Angle in degrees.
+ */
+inline float sind(float x) {
+  if (!std::isfinite(x)) return sin(x);
+  if (x < 0.0f) return -sind(-x);
+  int quo;
+  x = remquo(fabs(x), 90.0f, &quo);
+  switch (quo % 4) {
+    case 0:
+      return sin(glm::radians(x));
+    case 1:
+      return cos(glm::radians(x));
+    case 2:
+      return -sin(glm::radians(x));
+    case 3:
+      return -cos(glm::radians(x));
+  }
+  return 0.0f;
+}
+
+/**
+ * Cosine function where multiples of 90 degrees come out exact.
+ *
+ * @param x Angle in degrees.
+ */
+inline float cosd(float x) { return sind(x + 90.0f); }
+
+/**
+ * This 4x3 matrix can be used as an input to Manifold.Transform() to turn an
+ * object. Turns along the shortest path from given up-vector to (0, 0, 1).
+ *
+ * @param up The vector to be turned to point upwards. Length does not matter.
+ */
+inline glm::mat4x3 RotateUp(glm::vec3 up) {
+  up = glm::normalize(up);
+  glm::vec3 axis = glm::cross(up, {0, 0, 1});
+  float angle = glm::asin(glm::length(axis));
+  if (glm::dot(up, {0, 0, 1}) < 0) angle = glm::pi<float>() - angle;
+  return glm::mat4x3(glm::rotate(glm::mat4(1), angle, axis));
+}
+
+/**
+ * Determines if the three points are wound counter-clockwise, clockwise, or
+ * colinear within the specified tolerance.
+ *
+ * @param p0 First point
+ * @param p1 Second point
+ * @param p2 Third point
+ * @param tol Tolerance value for colinearity
+ * @return int, like Signum, this returns 1 for CCW, -1 for CW, and 0 if within
+ * tol of colinear.
+ */
+inline int CCW(glm::vec2 p0, glm::vec2 p1, glm::vec2 p2, float tol) {
+  glm::vec2 v1 = p1 - p0;
+  glm::vec2 v2 = p2 - p0;
+  float area = v1.x * v2.y - v1.y * v2.x;
+  float base2 = glm::max(glm::dot(v1, v1), glm::dot(v2, v2));
+  if (area * area * 4 <= base2 * tol * tol)
+    return 0;
+  else
+    return area > 0 ? 1 : -1;
+}
+
+/**
+ * Single polygon contour, wound CCW. First and last point are implicitly
+ * connected. Should ensure all input is
+ * [&epsilon;-valid](https://github.com/elalish/manifold/wiki/Manifold-Library#definition-of-%CE%B5-valid).
+ */
+using SimplePolygon = std::vector<glm::vec2>;
+
+/**
+ * Set of polygons with holes. Order of contours is arbitrary. Can contain any
+ * depth of nested holes and any number of separate polygons. Should ensure all
+ * input is
+ * [&epsilon;-valid](https://github.com/elalish/manifold/wiki/Manifold-Library#definition-of-%CE%B5-valid).
+ */
+using Polygons = std::vector<SimplePolygon>;
+
+/**
+ * The triangle-mesh input and output of this library.
+ */
+struct Mesh {
+  /// Required: The X-Y-Z positions of all vertices.
+  std::vector<glm::vec3> vertPos;
+  /// Required: The vertex indices of the three triangle corners in CCW (from
+  /// the outside) order, for each triangle.
+  std::vector<glm::ivec3> triVerts;
+  /// Optional: The X-Y-Z normal vectors of each vertex. If non-empty, must have
+  /// the same length as vertPos. If empty, these will be calculated
+  /// automatically.
+  std::vector<glm::vec3> vertNormal;
+  /// Optional: The X-Y-Z-W weighted tangent vectors for smooth Refine(). If
+  /// non-empty, must be exactly three times as long as Mesh.triVerts. Indexed
+  /// as 3 * tri + i, representing the tangent from Mesh.triVerts[tri][i] along
+  /// the CCW edge. If empty, mesh is faceted.
+  std::vector<glm::vec4> halfedgeTangent;
+  /// The absolute precision of the vertex positions, based on accrued rounding
+  /// errors. When creating a Manifold, the precision used will be the maximum
+  /// of this and a baseline precision from the size of the bounding box. Any
+  /// edge shorter than precision may be collapsed.
+  float precision = 0;
+};
+
+/**
+ * Defines which edges to sharpen and how much for the Manifold.Smooth()
+ * constructor.
+ */
+struct Smoothness {
+  /// The halfedge index = 3 * tri + i, referring to Mesh.triVerts[tri][i].
+  int halfedge;
+  /// A value between 0 and 1, where 0 is sharp and 1 is the default and the
+  /// curvature is interpolated between these values. The two paired halfedges
+  /// can have different values while maintaining C-1 continuity (except for 0).
+  float smoothness;
+};
+
+/**
+ * Geometric properties of the manifold, created with Manifold.GetProperties().
+ */
+struct Properties {
+  float surfaceArea, volume;
+};
+
+struct Box {
+  glm::vec3 min = glm::vec3(std::numeric_limits<float>::infinity());
+  glm::vec3 max = glm::vec3(-std::numeric_limits<float>::infinity());
+
+  /**
+   * Default constructor is an infinite box that contains all space.
+   */
+  Box() {}
+
+  /**
+   * Creates a box that contains the two given points.
+   */
+  Box(const glm::vec3 p1, const glm::vec3 p2) {
+    min = glm::min(p1, p2);
+    max = glm::max(p1, p2);
+  }
+
+  /**
+   * Returns the dimensions of the Box.
+   */
+  glm::vec3 Size() const { return max - min; }
+
+  /**
+   * Returns the center point of the Box.
+   */
+  glm::vec3 Center() const { return 0.5f * (max + min); }
+
+  /**
+   * Returns the absolute-largest coordinate value of any contained
+   * point.
+   */
+  float Scale() const {
+    glm::vec3 absMax = glm::max(glm::abs(min), glm::abs(max));
+    return glm::max(absMax.x, glm::max(absMax.y, absMax.z));
+  }
+
+  /**
+   * Does this box contain (includes equal) the given point?
+   */
+  bool Contains(const glm::vec3& p) const {
+    return glm::all(glm::greaterThanEqual(p, min)) &&
+           glm::all(glm::greaterThanEqual(max, p));
+  }
+
+  /**
+   * Does this box contain (includes equal) the given box?
+   */
+  bool Contains(const Box& box) const {
+    return glm::all(glm::greaterThanEqual(box.min, min)) &&
+           glm::all(glm::greaterThanEqual(max, box.max));
+  }
+
+  /**
+   * Expand this box to include the given point.
+   */
+  void Union(const glm::vec3 p) {
+    min = glm::min(min, p);
+    max = glm::max(max, p);
+  }
+
+  /**
+   * Expand this box to include the given box.
+   */
+  Box Union(const Box& box) const {
+    Box out;
+    out.min = glm::min(min, box.min);
+    out.max = glm::max(max, box.max);
+    return out;
+  }
+
+  /**
+   * Transform the given box by the given axis-aligned affine transform.
+   *
+   * Ensure the transform passed in is axis-aligned (rotations are all
+   * multiples of 90 degrees), or else the resulting bounding box will no longer
+   * bound properly.
+   */
+  Box Transform(const glm::mat4x3& transform) const {
+    Box out;
+    glm::vec3 minT = transform * glm::vec4(min, 1.0f);
+    glm::vec3 maxT = transform * glm::vec4(max, 1.0f);
+    out.min = glm::min(minT, maxT);
+    out.max = glm::max(minT, maxT);
+    return out;
+  }
+
+  /**
+   * Shift this box by the given vector.
+   */
+  Box operator+(glm::vec3 shift) const {
+    Box out;
+    out.min = min + shift;
+    out.max = max + shift;
+    return out;
+  }
+
+  /**
+   * Shift this box in-place by the given vector.
+   */
+  Box& operator+=(glm::vec3 shift) {
+    min += shift;
+    max += shift;
+    return *this;
+  }
+
+  /**
+   * Scale this box by the given vector.
+   */
+  Box operator*(glm::vec3 scale) const {
+    Box out;
+    out.min = min * scale;
+    out.max = max * scale;
+    return out;
+  }
+
+  /**
+   * Scale this box in-place by the given vector.
+   */
+  Box& operator*=(glm::vec3 scale) {
+    min *= scale;
+    max *= scale;
+    return *this;
+  }
+
+  /**
+   * Does this box overlap the one given (including equality)?
+   */
+  inline bool DoesOverlap(const Box& box) const {
+    return min.x <= box.max.x && min.y <= box.max.y && min.z <= box.max.z &&
+           max.x >= box.min.x && max.y >= box.min.y && max.z >= box.min.z;
+  }
+
+  /**
+   * Does the given point project within the XY extent of this box
+   * (including equality)?
+   */
+  inline bool DoesOverlap(glm::vec3 p) const {  // projected in z
+    return p.x <= max.x && p.x >= min.x && p.y <= max.y && p.y >= min.y;
+  }
+
+  /**
+   * Does this box have finite bounds?
+   */
+  bool IsFinite() const {
+    return glm::all(glm::isfinite(min)) && glm::all(glm::isfinite(max));
+  }
+};
+
+/**
+ * Axis-aligned rectangular bounds.
+ */
+struct Rect {
+  glm::vec2 min = glm::vec2(std::numeric_limits<float>::infinity());
+  glm::vec2 max = glm::vec2(-std::numeric_limits<float>::infinity());
+
+  /**
+   * Default constructor is an empty rectangle..
+   */
+  Rect() {}
+
+  /**
+   * Create a rectangle that contains the two given points.
+   */
+  Rect(const glm::vec2 a, const glm::vec2 b) {
+    min = glm::min(a, b);
+    max = glm::max(a, b);
+  }
+
+  /** @name Information
+   *  Details of the rectangle
+   */
+  ///@{
+
+  /**
+   * Return the dimensions of the rectangle.
+   */
+  glm::vec2 Size() const { return max - min; }
+
+  /**
+   * Return the area of the rectangle.
+   */
+  float Area() const {
+    auto sz = Size();
+    return sz.x * sz.y;
+  }
+
+  /**
+   * Returns the absolute-largest coordinate value of any contained
+   * point.
+   */
+  float Scale() const {
+    glm::vec2 absMax = glm::max(glm::abs(min), glm::abs(max));
+    return glm::max(absMax.x, absMax.y);
+  }
+
+  /**
+   * Returns the center point of the rectangle.
+   */
+  glm::vec2 Center() const { return 0.5f * (max + min); }
+
+  /**
+   * Does this rectangle contain (includes on border) the given point?
+   */
+  bool Contains(const glm::vec2& p) const {
+    return glm::all(glm::greaterThanEqual(p, min)) &&
+           glm::all(glm::greaterThanEqual(max, p));
+  }
+
+  /**
+   * Does this rectangle contain (includes equal) the given rectangle?
+   */
+  bool Contains(const Rect& rect) const {
+    return glm::all(glm::greaterThanEqual(rect.min, min)) &&
+           glm::all(glm::greaterThanEqual(max, rect.max));
+  }
+
+  /**
+   * Does this rectangle overlap the one given (including equality)?
+   */
+  bool DoesOverlap(const Rect& rect) const {
+    return min.x <= rect.max.x && min.y <= rect.max.y && max.x >= rect.min.x &&
+           max.y >= rect.min.y;
+  }
+
+  /**
+   * Is the rectangle empty (containing no space)?
+   */
+  bool IsEmpty() const { return max.y <= min.y || max.x <= min.x; };
+
+  /**
+   * Does this recangle have finite bounds?
+   */
+  bool IsFinite() const {
+    return glm::all(glm::isfinite(min)) && glm::all(glm::isfinite(max));
+  }
+
+  ///@}
+
+  /** @name Modification
+   */
+  ///@{
+
+  /**
+   * Expand this rectangle (in place) to include the given point.
+   */
+  void Union(const glm::vec2 p) {
+    min = glm::min(min, p);
+    max = glm::max(max, p);
+  }
+
+  /**
+   * Expand this rectangle to include the given Rect.
+   */
+  Rect Union(const Rect& rect) const {
+    Rect out;
+    out.min = glm::min(min, rect.min);
+    out.max = glm::max(max, rect.max);
+    return out;
+  }
+
+  /**
+   * Shift this rectangle by the given vector.
+   */
+  Rect operator+(const glm::vec2 shift) const {
+    Rect out;
+    out.min = min + shift;
+    out.max = max + shift;
+    return out;
+  }
+
+  /**
+   * Shift this rectangle in-place by the given vector.
+   */
+  Rect& operator+=(const glm::vec2 shift) {
+    min += shift;
+    max += shift;
+    return *this;
+  }
+
+  /**
+   * Scale this rectangle by the given vector.
+   */
+  Rect operator*(const glm::vec2 scale) const {
+    Rect out;
+    out.min = min * scale;
+    out.max = max * scale;
+    return out;
+  }
+
+  /**
+   * Scale this rectangle in-place by the given vector.
+   */
+  Rect& operator*=(const glm::vec2 scale) {
+    min *= scale;
+    max *= scale;
+    return *this;
+  }
+
+  /**
+   * Transform the rectangle by the given axis-aligned affine transform.
+   *
+   * Ensure the transform passed in is axis-aligned (rotations are all
+   * multiples of 90 degrees), or else the resulting rectangle will no longer
+   * bound properly.
+   */
+  Rect Transform(const glm::mat3x2& m) const {
+    Rect rect;
+    rect.min = m * glm::vec3(min, 1);
+    rect.max = m * glm::vec3(max, 1);
+    return rect;
+  }
+  ///@}
+};
+/** @} */
+
+/** @addtogroup Core
+ *  @{
+ */
+
+/**
+ * Boolean operation type: Add (Union), Subtract (Difference), and Intersect.
+ */
+enum class OpType { Add, Subtract, Intersect };
+
+/**
+ * These static properties control how circular shapes are quantized by
+ * default on construction. If circularSegments is specified, it takes
+ * precedence. If it is zero, then instead the minimum is used of the segments
+ * calculated based on edge length and angle, rounded up to the nearest
+ * multiple of four. To get numbers not divisible by four, circularSegments
+ * must be specified.
+ */
+class Quality {
+ private:
+  inline static int circularSegments_ = 0;
+  inline static float circularAngle_ = 10.0f;
+  inline static float circularEdgeLength_ = 1.0f;
+
+ public:
+  /**
+   * Sets an angle constraint the default number of circular segments for the
+   * CrossSection::Circle(), Manifold::Cylinder(), Manifold::Sphere(), and
+   * Manifold::Revolve() constructors. The number of segments will be rounded up
+   * to the nearest factor of four.
+   *
+   * @param angle The minimum angle in degrees between consecutive segments. The
+   * angle will increase if the the segments hit the minimum edge length.
+   * Default is 10 degrees.
+   */
+  static void SetMinCircularAngle(float angle) {
+    if (angle <= 0) return;
+    circularAngle_ = angle;
+  }
+
+  /**
+   * Sets a length constraint the default number of circular segments for the
+   * CrossSection::Circle(), Manifold::Cylinder(), Manifold::Sphere(), and
+   * Manifold::Revolve() constructors. The number of segments will be rounded up
+   * to the nearest factor of four.
+   *
+   * @param length The minimum length of segments. The length will
+   * increase if the the segments hit the minimum angle. Default is 1.0.
+   */
+  static void SetMinCircularEdgeLength(float length) {
+    if (length <= 0) return;
+    circularEdgeLength_ = length;
+  }
+
+  /**
+   * Sets the default number of circular segments for the
+   * CrossSection::Circle(), Manifold::Cylinder(), Manifold::Sphere(), and
+   * Manifold::Revolve() constructors. Overrides the edge length and angle
+   * constraints and sets the number of segments to exactly this value.
+   *
+   * @param number Number of circular segments. Default is 0, meaning no
+   * constraint is applied.
+   */
+  static void SetCircularSegments(int number) {
+    if (number < 3 && number != 0) return;
+    circularSegments_ = number;
+  }
+
+  /**
+   * Determine the result of the SetMinCircularAngle(),
+   * SetMinCircularEdgeLength(), and SetCircularSegments() defaults.
+   *
+   * @param radius For a given radius of circle, determine how many default
+   * segments there will be.
+   */
+  static int GetCircularSegments(float radius) {
+    if (circularSegments_ > 0) return circularSegments_;
+    int nSegA = 360.0f / circularAngle_;
+    int nSegL = 2.0f * radius * glm::pi<float>() / circularEdgeLength_;
+    int nSeg = fmin(nSegA, nSegL) + 3;
+    nSeg -= nSeg % 4;
+    return std::max(nSeg, 3);
+  }
+};
+/** @} */
+
+/** @defgroup Debug
+ *  @brief Debugging features
+ *
+ * The features require compiler flags to be enabled. Assertions are enabled
+ * with the MANIFOLD_DEBUG flag and then controlled with ExecutionParams.
+ * Exceptions are only thrown if the MANIFOLD_EXCEPTIONS flag is set. Import and
+ * Export of 3D models is only supported with the MANIFOLD_EXPORT flag, which
+ * also requires linking in the Assimp dependency.
+ *  @{
+ */
+
+/** @defgroup Exceptions
+ *  @brief Custom Exceptions
+ * @{
+ */
+#ifdef MANIFOLD_DEBUG
+struct userErr : public virtual std::runtime_error {
+  using std::runtime_error::runtime_error;
+};
+struct topologyErr : public virtual std::runtime_error {
+  using std::runtime_error::runtime_error;
+};
+struct geometryErr : public virtual std::runtime_error {
+  using std::runtime_error::runtime_error;
+};
+using logicErr = std::logic_error;
+#endif
+/** @} */
+
+/**
+ * Global parameters that control debugging output. Only has an
+ * effect when compiled with the MANIFOLD_DEBUG flag.
+ */
+struct ExecutionParams {
+  /// Perform extra sanity checks and assertions on the intermediate data
+  /// structures.
+  bool intermediateChecks = false;
+  /// Verbose output primarily of the Boolean, including timing info and vector
+  /// sizes.
+  bool verbose = false;
+  /// If processOverlaps is false, a geometric check will be performed to assert
+  /// all triangles are CCW.
+  bool processOverlaps = true;
+  /// Suppresses printed errors regarding CW triangles. Has no effect if
+  /// processOverlaps is true.
+  bool suppressErrors = false;
+  /// Deterministic outputs. Will disable some parallel optimizations.
+  bool deterministic = false;
+  /// Perform optional but recommended triangle cleanups in SimplifyTopology()
+  bool cleanupTriangles = true;
+};
+
+#ifdef MANIFOLD_DEBUG
+
+template <typename T>
+inline std::ostream& operator<<(std::ostream& stream, const glm::tvec2<T>& v) {
+  return stream << "x = " << v.x << ", y = " << v.y;
+}
+
+template <typename T>
+inline std::ostream& operator<<(std::ostream& stream, const glm::tvec3<T>& v) {
+  return stream << "x = " << v.x << ", y = " << v.y << ", z = " << v.z;
+}
+
+template <typename T>
+inline std::ostream& operator<<(std::ostream& stream, const glm::tvec4<T>& v) {
+  return stream << "x = " << v.x << ", y = " << v.y << ", z = " << v.z
+                << ", w = " << v.w;
+}
+
+inline std::ostream& operator<<(std::ostream& stream, const glm::mat3& mat) {
+  glm::mat3 tam = glm::transpose(mat);
+  return stream << tam[0] << std::endl
+                << tam[1] << std::endl
+                << tam[2] << std::endl;
+}
+
+inline std::ostream& operator<<(std::ostream& stream, const glm::mat4x3& mat) {
+  glm::mat3x4 tam = glm::transpose(mat);
+  return stream << tam[0] << std::endl
+                << tam[1] << std::endl
+                << tam[2] << std::endl;
+}
+
+inline std::ostream& operator<<(std::ostream& stream, const Box& box) {
+  return stream << "min: " << box.min << ", "
+                << "max: " << box.max;
+}
+
+inline std::ostream& operator<<(std::ostream& stream, const Rect& box) {
+  return stream << "min: " << box.min << ", "
+                << "max: " << box.max;
+}
+
+/**
+ * Print the contents of this vector to standard output. Only exists if compiled
+ * with MANIFOLD_DEBUG flag.
+ */
+template <typename T>
+void Dump(const std::vector<T>& vec) {
+  std::cout << "Vec = " << std::endl;
+  for (int i = 0; i < vec.size(); ++i) {
+    std::cout << i << ", " << vec[i] << ", " << std::endl;
+  }
+  std::cout << std::endl;
+}
+
+template <typename T>
+void Diff(const std::vector<T>& a, const std::vector<T>& b) {
+  std::cout << "Diff = " << std::endl;
+  if (a.size() != b.size()) {
+    std::cout << "a and b must have the same length, aborting Diff"
+              << std::endl;
+    return;
+  }
+  for (int i = 0; i < a.size(); ++i) {
+    if (a[i] != b[i])
+      std::cout << i << ": " << a[i] << ", " << b[i] << std::endl;
+  }
+  std::cout << std::endl;
+}
+/** @} */
+#endif
+}  // namespace manifold
+
+#undef HOST_DEVICE
diff --git a/thirdparty/manifold/src/utilities/include/sparse.h b/thirdparty/manifold/src/utilities/include/sparse.h
new file mode 100644
index 000000000000..05a711c44e6a
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/sparse.h
@@ -0,0 +1,201 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <math.h>
+
+#include "optional_assert.h"
+#include "par.h"
+#include "public.h"
+#include "utils.h"
+#include "vec.h"
+
+namespace manifold {
+
+/** @ingroup Private */
+class SparseIndices {
+  // sparse indices where {p1: q1, p2: q2, ...} are laid out as
+  // p1 q1 p2 q2 or q1 p1 q2 p2, depending on endianness
+  // such that the indices are sorted by (p << 32) | q
+ public:
+#if defined(__BYTE_ORDER) && __BYTE_ORDER == __BIG_ENDIAN ||                 \
+    defined(__BIG_ENDIAN__) || defined(__ARMEB__) || defined(__THUMBEB__) || \
+    defined(__AARCH64EB__) || defined(_MIBSEB) || defined(__MIBSEB) ||       \
+    defined(__MIBSEB__)
+  static constexpr size_t pOffset = 0;
+#elif defined(__BYTE_ORDER) && __BYTE_ORDER == __LITTLE_ENDIAN ||          \
+    defined(__LITTLE_ENDIAN__) || defined(__ARMEL__) ||                    \
+    defined(__THUMBEL__) || defined(__AARCH64EL__) || defined(_MIPSEL) ||  \
+    defined(__MIPSEL) || defined(__MIPSEL__) || defined(__EMSCRIPTEN__) || \
+    defined(_WIN32)
+  static constexpr size_t pOffset = 1;
+#else
+#error "unknown architecture"
+#endif
+  static constexpr int64_t EncodePQ(int p, int q) {
+    return (int64_t(p) << 32) | q;
+  }
+
+  SparseIndices() = default;
+  SparseIndices(size_t size) { data_ = Vec<char>(size * sizeof(int64_t)); }
+
+  size_t size() const { return data_.size() / sizeof(int64_t); }
+
+  Vec<int> Copy(bool use_q) const {
+    Vec<int> out(size());
+    size_t offset = pOffset;
+    if (use_q) offset = 1 - offset;
+    const int* p = ptr();
+    for_each(autoPolicy(out.size()), countAt(0_z), countAt(out.size()),
+             [&](size_t i) { out[i] = p[i * 2 + offset]; });
+    return out;
+  }
+
+  void Sort() {
+    VecView<int64_t> view = AsVec64();
+    stable_sort(autoPolicy(size()), view.begin(), view.end());
+  }
+
+  void Resize(size_t size) { data_.resize(size * sizeof(int64_t), -1); }
+
+  inline int& Get(size_t i, bool use_q) {
+    if (use_q)
+      return ptr()[2 * i + 1 - pOffset];
+    else
+      return ptr()[2 * i + pOffset];
+  }
+
+  inline int Get(size_t i, bool use_q) const {
+    if (use_q)
+      return ptr()[2 * i + 1 - pOffset];
+    else
+      return ptr()[2 * i + pOffset];
+  }
+
+  inline int64_t GetPQ(size_t i) const {
+    VecView<const int64_t> view = AsVec64();
+    return view[i];
+  }
+
+  inline void Set(size_t i, int p, int q) {
+    VecView<int64_t> view = AsVec64();
+    view[i] = EncodePQ(p, q);
+  }
+
+  inline void SetPQ(size_t i, int64_t pq) {
+    VecView<int64_t> view = AsVec64();
+    view[i] = pq;
+  }
+
+  VecView<int64_t> AsVec64() {
+    return VecView<int64_t>(reinterpret_cast<int64_t*>(data_.data()),
+                            data_.size() / sizeof(int64_t));
+  }
+
+  VecView<const int64_t> AsVec64() const {
+    return VecView<const int64_t>(
+        reinterpret_cast<const int64_t*>(data_.data()),
+        data_.size() / sizeof(int64_t));
+  }
+
+  VecView<int32_t> AsVec32() {
+    return VecView<int32_t>(reinterpret_cast<int32_t*>(data_.data()),
+                            data_.size() / sizeof(int32_t));
+  }
+
+  VecView<const int32_t> AsVec32() const {
+    return VecView<const int32_t>(
+        reinterpret_cast<const int32_t*>(data_.data()),
+        data_.size() / sizeof(int32_t));
+  }
+
+  inline void Add(int p, int q) {
+    for (int i = 0; i < sizeof(int64_t); ++i) data_.push_back(-1);
+    Set(size() - 1, p, q);
+  }
+
+  void Unique() {
+    Sort();
+    VecView<int64_t> view = AsVec64();
+    size_t newSize = std::unique(view.begin(), view.end()) - view.begin();
+    Resize(newSize);
+  }
+
+  size_t RemoveZeros(Vec<int>& S) {
+    ASSERT(S.size() == size(), userErr,
+           "Different number of values than indicies!");
+    VecView<int64_t> view = AsVec64();
+    auto zBegin = zip(S.begin(), view.begin());
+    auto zEnd = zip(S.end(), view.end());
+    size_t size =
+        remove_if<decltype(zBegin)>(autoPolicy(S.size()), zBegin, zEnd,
+                                    [](thrust::tuple<int, int64_t> x) {
+                                      return thrust::get<0>(x) == 0;
+                                    }) -
+        zBegin;
+    S.resize(size, -1);
+    Resize(size);
+    return size;
+  }
+
+  template <typename T>
+  struct firstNonFinite {
+    bool NotFinite(float v) const { return !isfinite(v); }
+    bool NotFinite(glm::vec2 v) const { return !isfinite(v[0]); }
+    bool NotFinite(glm::vec3 v) const { return !isfinite(v[0]); }
+    bool NotFinite(glm::vec4 v) const { return !isfinite(v[0]); }
+
+    bool operator()(thrust::tuple<T, int, int64_t> x) const {
+      bool result = NotFinite(thrust::get<0>(x));
+      return result;
+    }
+  };
+
+  template <typename T>
+  size_t KeepFinite(Vec<T>& v, Vec<int>& x) {
+    ASSERT(x.size() == size(), userErr,
+           "Different number of values than indicies!");
+    VecView<int64_t> view = AsVec64();
+    auto zBegin = zip(v.begin(), x.begin(), view.begin());
+    auto zEnd = zip(v.end(), x.end(), view.end());
+    size_t size = remove_if<decltype(zBegin)>(autoPolicy(v.size()), zBegin,
+                                              zEnd, firstNonFinite<T>()) -
+                  zBegin;
+    v.resize(size);
+    x.resize(size);
+    Resize(size);
+    return size;
+  }
+
+#ifdef MANIFOLD_DEBUG
+  void Dump() const {
+    std::cout << "SparseIndices = " << std::endl;
+    const int* p = ptr();
+    for (size_t i = 0; i < size(); ++i) {
+      std::cout << i << ", p = " << Get(i, false) << ", q = " << Get(i, true)
+                << std::endl;
+    }
+    std::cout << std::endl;
+  }
+#endif
+
+ private:
+  Vec<char> data_;
+  inline int* ptr() { return reinterpret_cast<int32_t*>(data_.data()); }
+  inline const int* ptr() const {
+    return reinterpret_cast<const int32_t*>(data_.data());
+  }
+};
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/svd.h b/thirdparty/manifold/src/utilities/include/svd.h
new file mode 100644
index 000000000000..243d898bc9cb
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/svd.h
@@ -0,0 +1,310 @@
+// MIT License
+
+// Copyright (c) 2019 wi-re
+// Copyright 2023 The Manifold Authors.
+
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+
+// Modified from https://github.com/wi-re/tbtSVD, removing CUDA dependence and
+// approximate inverse square roots.
+
+#include <algorithm>
+#include <cmath>
+#include <glm/glm.hpp>
+#include <tuple>
+#include <utility>
+
+namespace {
+// Constants used for calculation of Givens quaternions
+inline constexpr float _gamma = 5.828427124f;   // sqrt(8)+3;
+inline constexpr float _cStar = 0.923879532f;   // cos(pi/8)
+inline constexpr float _sStar = 0.3826834323f;  // sin(pi/8)
+// Threshold value
+inline constexpr float _SVD_EPSILON = 1e-6f;
+// Iteration counts for Jacobi Eigen Analysis, influences precision
+inline constexpr int JACOBI_STEPS = 12;
+
+// Helper function used to swap X with Y and Y with  X if c == true
+inline void CondSwap(bool c, float& X, float& Y) {
+  float Z = X;
+  X = c ? Y : X;
+  Y = c ? Z : Y;
+}
+// Helper function used to swap X with Y and Y with -X if c == true
+inline void CondNegSwap(bool c, float& X, float& Y) {
+  float Z = -X;
+  X = c ? Y : X;
+  Y = c ? Z : Y;
+}
+// A simple symmetric 3x3 Matrix class (contains no storage for (0, 1) (0, 2)
+// and (1, 2)
+struct Symmetric3x3 {
+  float m_00 = 1.f;
+  float m_10 = 0.f, m_11 = 1.f;
+  float m_20 = 0.f, m_21 = 0.f, m_22 = 1.f;
+
+  Symmetric3x3(float a11 = 1.f, float a21 = 0.f, float a22 = 1.f,
+               float a31 = 0.f, float a32 = 0.f, float a33 = 1.f)
+      : m_00(a11), m_10(a21), m_11(a22), m_20(a31), m_21(a32), m_22(a33) {}
+  Symmetric3x3(glm::mat3 o)
+      : m_00(o[0][0]),
+        m_10(o[0][1]),
+        m_11(o[1][1]),
+        m_20(o[0][2]),
+        m_21(o[1][2]),
+        m_22(o[2][2]) {}
+};
+// Helper struct to store 2 floats to avoid OUT parameters on functions
+struct Givens {
+  float ch = _cStar;
+  float sh = _sStar;
+};
+// Helper struct to store 2 Matrices to avoid OUT parameters on functions
+struct QR {
+  glm::mat3 Q, R;
+};
+// Calculates the squared norm of the vector.
+inline float Dist2(glm::vec3 v) { return glm::dot(v, v); }
+// For an explanation of the math see
+// http://pages.cs.wisc.edu/~sifakis/papers/SVD_TR1690.pdf Computing the
+// Singular Value Decomposition of 3 x 3 matrices with minimal branching and
+// elementary floating point operations See Algorithm 2 in reference. Given a
+// matrix A this function returns the Givens quaternion (x and w component, y
+// and z are 0)
+inline Givens ApproximateGivensQuaternion(Symmetric3x3& A) {
+  Givens g{2.f * (A.m_00 - A.m_11), A.m_10};
+  bool b = _gamma * g.sh * g.sh < g.ch * g.ch;
+  float w = 1.f / sqrt(fmaf(g.ch, g.ch, g.sh * g.sh));
+  if (w != w) b = 0;
+  return Givens{b ? w * g.ch : (float)_cStar, b ? w * g.sh : (float)_sStar};
+}
+// Function used to apply a Givens rotation S. Calculates the weights and
+// updates the quaternion to contain the cumulative rotation
+inline void JacobiConjugation(const int32_t x, const int32_t y, const int32_t z,
+                              Symmetric3x3& S, glm::vec4& q) {
+  auto g = ApproximateGivensQuaternion(S);
+  float scale = 1.f / fmaf(g.ch, g.ch, g.sh * g.sh);
+  float a = fmaf(g.ch, g.ch, -g.sh * g.sh) * scale;
+  float b = 2.f * g.sh * g.ch * scale;
+  Symmetric3x3 _S = S;
+  // perform conjugation S = Q'*S*Q
+  S.m_00 = fmaf(a, fmaf(a, _S.m_00, b * _S.m_10),
+                b * (fmaf(a, _S.m_10, b * _S.m_11)));
+  S.m_10 = fmaf(a, fmaf(-b, _S.m_00, a * _S.m_10),
+                b * (fmaf(-b, _S.m_10, a * _S.m_11)));
+  S.m_11 = fmaf(-b, fmaf(-b, _S.m_00, a * _S.m_10),
+                a * (fmaf(-b, _S.m_10, a * _S.m_11)));
+  S.m_20 = fmaf(a, _S.m_20, b * _S.m_21);
+  S.m_21 = fmaf(-b, _S.m_20, a * _S.m_21);
+  S.m_22 = _S.m_22;
+  // update cumulative rotation qV
+  glm::vec3 tmp = g.sh * glm::vec3(q);
+  g.sh *= q[3];
+  // (x,y,z) corresponds to ((0,1,2),(1,2,0),(2,0,1)) for (p,q) =
+  // ((0,1),(1,2),(0,2))
+  q[z] = fmaf(q[z], g.ch, g.sh);
+  q[3] = fmaf(q[3], g.ch, -tmp[z]);  // w
+  q[x] = fmaf(q[x], g.ch, tmp[y]);
+  q[y] = fmaf(q[y], g.ch, -tmp[x]);
+  // re-arrange matrix for next iteration
+  _S.m_00 = S.m_11;
+  _S.m_10 = S.m_21;
+  _S.m_11 = S.m_22;
+  _S.m_20 = S.m_10;
+  _S.m_21 = S.m_20;
+  _S.m_22 = S.m_00;
+  S.m_00 = _S.m_00;
+  S.m_10 = _S.m_10;
+  S.m_11 = _S.m_11;
+  S.m_20 = _S.m_20;
+  S.m_21 = _S.m_21;
+  S.m_22 = _S.m_22;
+}
+// Function used to contain the Givens permutations and the loop of the jacobi
+// steps controlled by JACOBI_STEPS Returns the quaternion q containing the
+// cumulative result used to reconstruct S
+inline glm::mat3 JacobiEigenAnalysis(Symmetric3x3 S) {
+  glm::vec4 q(0, 0, 0, 1);
+  for (int32_t i = 0; i < JACOBI_STEPS; i++) {
+    JacobiConjugation(0, 1, 2, S, q);
+    JacobiConjugation(1, 2, 0, S, q);
+    JacobiConjugation(2, 0, 1, S, q);
+  }
+  return glm::mat3(1.f - 2.f * (fmaf(q.y, q.y, q.z * q.z)),  //
+                   2.f * fmaf(q.x, q.y, +q.w * q.z),         //
+                   2.f * fmaf(q.x, q.z, -q.w * q.y),         //
+                   2 * fmaf(q.x, q.y, -q.w * q.z),           //
+                   1 - 2 * fmaf(q.x, q.x, q.z * q.z),        //
+                   2 * fmaf(q.y, q.z, q.w * q.x),            //
+                   2 * fmaf(q.x, q.z, q.w * q.y),            //
+                   2 * fmaf(q.y, q.z, -q.w * q.x),           //
+                   1 - 2 * fmaf(q.x, q.x, q.y * q.y));
+}
+// Implementation of Algorithm 3
+inline void SortSingularValues(glm::mat3& B, glm::mat3& V) {
+  float rho1 = Dist2(B[0]);
+  float rho2 = Dist2(B[1]);
+  float rho3 = Dist2(B[2]);
+  bool c;
+  c = rho1 < rho2;
+  CondNegSwap(c, B[0][0], B[1][0]);
+  CondNegSwap(c, V[0][0], V[1][0]);
+  CondNegSwap(c, B[0][1], B[1][1]);
+  CondNegSwap(c, V[0][1], V[1][1]);
+  CondNegSwap(c, B[0][2], B[1][2]);
+  CondNegSwap(c, V[0][2], V[1][2]);
+  CondSwap(c, rho1, rho2);
+  c = rho1 < rho3;
+  CondNegSwap(c, B[0][0], B[2][0]);
+  CondNegSwap(c, V[0][0], V[2][0]);
+  CondNegSwap(c, B[0][1], B[2][1]);
+  CondNegSwap(c, V[0][1], V[2][1]);
+  CondNegSwap(c, B[0][2], B[2][2]);
+  CondNegSwap(c, V[0][2], V[2][2]);
+  CondSwap(c, rho1, rho3);
+  c = rho2 < rho3;
+  CondNegSwap(c, B[1][0], B[2][0]);
+  CondNegSwap(c, V[1][0], V[2][0]);
+  CondNegSwap(c, B[1][1], B[2][1]);
+  CondNegSwap(c, V[1][1], V[2][1]);
+  CondNegSwap(c, B[1][2], B[2][2]);
+  CondNegSwap(c, V[1][2], V[2][2]);
+}
+// Implementation of Algorithm 4
+inline Givens QRGivensQuaternion(float a1, float a2) {
+  // a1 = pivot point on diagonal
+  // a2 = lower triangular entry we want to annihilate
+  float epsilon = (float)_SVD_EPSILON;
+  float rho = sqrt(fmaf(a1, a1, +a2 * a2));
+  Givens g{fabsf(a1) + fmaxf(rho, epsilon), rho > epsilon ? a2 : 0};
+  bool b = a1 < 0.f;
+  CondSwap(b, g.sh, g.ch);
+  float w = 1.f / sqrt(fmaf(g.ch, g.ch, g.sh * g.sh));
+  g.ch *= w;
+  g.sh *= w;
+  return g;
+}
+// Implements a QR decomposition of a Matrix, see Sec 4.2
+inline QR QRDecomposition(glm::mat3& B) {
+  glm::mat3 Q, R;
+  // first Givens rotation (ch,0,0,sh)
+  auto g1 = QRGivensQuaternion(B[0][0], B[0][1]);
+  auto a = fmaf(-2.f, g1.sh * g1.sh, 1.f);
+  auto b = 2.f * g1.ch * g1.sh;
+  // apply B = Q' * B
+  R[0][0] = fmaf(a, B[0][0], b * B[0][1]);
+  R[1][0] = fmaf(a, B[1][0], b * B[1][1]);
+  R[2][0] = fmaf(a, B[2][0], b * B[2][1]);
+  R[0][1] = fmaf(-b, B[0][0], a * B[0][1]);
+  R[1][1] = fmaf(-b, B[1][0], a * B[1][1]);
+  R[2][1] = fmaf(-b, B[2][0], a * B[2][1]);
+  R[0][2] = B[0][2];
+  R[1][2] = B[1][2];
+  R[2][2] = B[2][2];
+  // second Givens rotation (ch,0,-sh,0)
+  auto g2 = QRGivensQuaternion(R[0][0], R[0][2]);
+  a = fmaf(-2.f, g2.sh * g2.sh, 1.f);
+  b = 2.f * g2.ch * g2.sh;
+  // apply B = Q' * B;
+  B[0][0] = fmaf(a, R[0][0], b * R[0][2]);
+  B[1][0] = fmaf(a, R[1][0], b * R[1][2]);
+  B[2][0] = fmaf(a, R[2][0], b * R[2][2]);
+  B[0][1] = R[0][1];
+  B[1][1] = R[1][1];
+  B[2][1] = R[2][1];
+  B[0][2] = fmaf(-b, R[0][0], a * R[0][2]);
+  B[1][2] = fmaf(-b, R[1][0], a * R[1][2]);
+  B[2][2] = fmaf(-b, R[2][0], a * R[2][2]);
+  // third Givens rotation (ch,sh,0,0)
+  auto g3 = QRGivensQuaternion(B[1][1], B[1][2]);
+  a = fmaf(-2.f, g3.sh * g3.sh, 1.f);
+  b = 2.f * g3.ch * g3.sh;
+  // R is now set to desired value
+  R[0][0] = B[0][0];
+  R[1][0] = B[1][0];
+  R[2][0] = B[2][0];
+  R[0][1] = fmaf(a, B[0][1], b * B[0][2]);
+  R[1][1] = fmaf(a, B[1][1], b * B[1][2]);
+  R[2][1] = fmaf(a, B[2][1], b * B[2][2]);
+  R[0][2] = fmaf(-b, B[0][1], a * B[0][2]);
+  R[1][2] = fmaf(-b, B[1][1], a * B[1][2]);
+  R[2][2] = fmaf(-b, B[2][1], a * B[2][2]);
+  // construct the cumulative rotation Q=Q1 * Q2 * Q3
+  // the number of floating point operations for three quaternion
+  // multiplications is more or less comparable to the explicit form of the
+  // joined matrix. certainly more memory-efficient!
+  auto sh12 = 2.f * fmaf(g1.sh, g1.sh, -0.5f);
+  auto sh22 = 2.f * fmaf(g2.sh, g2.sh, -0.5f);
+  auto sh32 = 2.f * fmaf(g3.sh, g3.sh, -0.5f);
+  Q[0][0] = sh12 * sh22;
+  Q[1][0] = fmaf(4.f * g2.ch * g3.ch, sh12 * g2.sh * g3.sh,
+                 2.f * g1.ch * g1.sh * sh32);
+  Q[2][0] = fmaf(4.f * g1.ch * g3.ch, g1.sh * g3.sh,
+                 -2.f * g2.ch * sh12 * g2.sh * sh32);
+
+  Q[0][1] = -2.f * g1.ch * g1.sh * sh22;
+  Q[1][1] =
+      fmaf(-8.f * g1.ch * g2.ch * g3.ch, g1.sh * g2.sh * g3.sh, sh12 * sh32);
+  Q[2][1] = fmaf(
+      -2.f * g3.ch, g3.sh,
+      4.f * g1.sh * fmaf(g3.ch * g1.sh, g3.sh, g1.ch * g2.ch * g2.sh * sh32));
+
+  Q[0][2] = 2.f * g2.ch * g2.sh;
+  Q[1][2] = -2.f * g3.ch * sh22 * g3.sh;
+  Q[2][2] = sh22 * sh32;
+  return QR{Q, R};
+}
+}  // namespace
+
+namespace manifold {
+/** @addtogroup Connections
+ *  @{
+ */
+
+/**
+ * The three matrices of a Singular Value Decomposition.
+ */
+struct SVDSet {
+  glm::mat3 U, S, V;
+};
+
+/**
+ * Returns the Singular Value Decomposition of A: A = U * S * glm::transpose(V).
+ *
+ * @param A The matrix to decompose.
+ */
+inline SVDSet SVD(glm::mat3 A) {
+  glm::mat3 V = JacobiEigenAnalysis(glm::transpose(A) * A);
+  auto B = A * V;
+  SortSingularValues(B, V);
+  QR qr = QRDecomposition(B);
+  return SVDSet{qr.Q, qr.R, V};
+}
+
+/**
+ * Returns the largest singular value of A.
+ *
+ * @param A The matrix to measure.
+ */
+inline float SpectralNorm(glm::mat3 A) {
+  SVDSet usv = SVD(A);
+  return usv.S[0][0];
+}
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/tri_dist.h b/thirdparty/manifold/src/utilities/include/tri_dist.h
new file mode 100644
index 000000000000..f768f7f90cf3
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/tri_dist.h
@@ -0,0 +1,225 @@
+// Copyright 2024 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <array>
+#include <glm/glm.hpp>
+
+namespace manifold {
+
+// From NVIDIA-Omniverse PhysX - BSD 3-Clause "New" or "Revised" License
+// https://github.com/NVIDIA-Omniverse/PhysX/blob/main/LICENSE.md
+// https://github.com/NVIDIA-Omniverse/PhysX/blob/main/physx/source/geomutils/src/sweep/GuSweepCapsuleCapsule.cpp
+// With minor modifications
+
+/**
+ * Returns the distance between two line segments.
+ *
+ * @param[out] x Closest point on line segment pa.
+ * @param[out] y Closest point on line segment qb.
+ * @param[in]  p  One endpoint of the first line segment.
+ * @param[in]  a  Other endpoint of the first line segment.
+ * @param[in]  p  One endpoint of the second line segment.
+ * @param[in]  b  Other endpoint of the second line segment.
+ */
+inline void EdgeEdgeDist(glm::vec3& x, glm::vec3& y,  // closest points
+                         const glm::vec3& p,
+                         const glm::vec3& a,  // seg 1 origin, vector
+                         const glm::vec3& q,
+                         const glm::vec3& b)  // seg 2 origin, vector
+{
+  const glm::vec3 T = q - p;
+  const float ADotA = glm::dot(a, a);
+  const float BDotB = glm::dot(b, b);
+  const float ADotB = glm::dot(a, b);
+  const float ADotT = glm::dot(a, T);
+  const float BDotT = glm::dot(b, T);
+
+  // t parameterizes ray (p, a)
+  // u parameterizes ray (q, b)
+
+  // Compute t for the closest point on ray (p, a) to ray (q, b)
+  const float Denom = ADotA * BDotB - ADotB * ADotB;
+
+  float t;  // We will clamp result so t is on the segment (p, a)
+  t = Denom != 0.0f
+          ? glm::clamp((ADotT * BDotB - BDotT * ADotB) / Denom, 0.0f, 1.0f)
+          : 0.0f;
+
+  // find u for point on ray (q, b) closest to point at t
+  float u;
+  if (BDotB != 0.0f) {
+    u = (t * ADotB - BDotT) / BDotB;
+
+    // if u is on segment (q, b), t and u correspond to closest points,
+    // otherwise, clamp u, recompute and clamp t
+    if (u < 0.0f) {
+      u = 0.0f;
+      t = ADotA != 0.0f ? glm::clamp(ADotT / ADotA, 0.0f, 1.0f) : 0.0f;
+    } else if (u > 1.0f) {
+      u = 1.0f;
+      t = ADotA != 0.0f ? glm::clamp((ADotB + ADotT) / ADotA, 0.0f, 1.0f)
+                        : 0.0f;
+    }
+  } else {
+    u = 0.0f;
+    t = ADotA != 0.0f ? glm::clamp(ADotT / ADotA, 0.0f, 1.0f) : 0.0f;
+  }
+  x = p + a * t;
+  y = q + b * u;
+}
+
+// From NVIDIA-Omniverse PhysX - BSD 3-Clause "New" or "Revised" License
+// https://github.com/NVIDIA-Omniverse/PhysX/blob/main/LICENSE.md
+// https://github.com/NVIDIA-Omniverse/PhysX/blob/main/physx/source/geomutils/src/distance/GuDistanceTriangleTriangle.cpp
+// With minor modifications
+
+/**
+ * Returns the minimum squared distance between two triangles.
+ *
+ * @param  p  First  triangle.
+ * @param  q  Second triangle.
+ */
+inline float DistanceTriangleTriangleSquared(
+    const std::array<glm::vec3, 3>& p, const std::array<glm::vec3, 3>& q) {
+  std::array<glm::vec3, 3> Sv;
+  Sv[0] = p[1] - p[0];
+  Sv[1] = p[2] - p[1];
+  Sv[2] = p[0] - p[2];
+
+  std::array<glm::vec3, 3> Tv;
+  Tv[0] = q[1] - q[0];
+  Tv[1] = q[2] - q[1];
+  Tv[2] = q[0] - q[2];
+
+  bool shown_disjoint = false;
+
+  float mindd = std::numeric_limits<float>::max();
+
+  for (uint32_t i = 0; i < 3; i++) {
+    for (uint32_t j = 0; j < 3; j++) {
+      glm::vec3 cp;
+      glm::vec3 cq;
+      EdgeEdgeDist(cp, cq, p[i], Sv[i], q[j], Tv[j]);
+      const glm::vec3 V = cq - cp;
+      const float dd = glm::dot(V, V);
+
+      if (dd <= mindd) {
+        mindd = dd;
+
+        uint32_t id = i + 2;
+        if (id >= 3) id -= 3;
+        glm::vec3 Z = p[id] - cp;
+        float a = glm::dot(Z, V);
+        id = j + 2;
+        if (id >= 3) id -= 3;
+        Z = q[id] - cq;
+        float b = glm::dot(Z, V);
+
+        if ((a <= 0.0f) && (b >= 0.0f)) {
+          return glm::dot(V, V);
+        };
+
+        if (a <= 0.0f)
+          a = 0.0f;
+        else if (b > 0.0f)
+          b = 0.0f;
+
+        if ((mindd - a + b) > 0.0f) shown_disjoint = true;
+      }
+    }
+  }
+
+  glm::vec3 Sn = glm::cross(Sv[0], Sv[1]);
+  float Snl = glm::dot(Sn, Sn);
+
+  if (Snl > 1e-15f) {
+    const glm::vec3 Tp(glm::dot(p[0] - q[0], Sn), glm::dot(p[0] - q[1], Sn),
+                       glm::dot(p[0] - q[2], Sn));
+
+    int index = -1;
+    if ((Tp[0] > 0.0f) && (Tp[1] > 0.0f) && (Tp[2] > 0.0f)) {
+      index = Tp[0] < Tp[1] ? 0 : 1;
+      if (Tp[2] < Tp[index]) index = 2;
+    } else if ((Tp[0] < 0.0f) && (Tp[1] < 0.0f) && (Tp[2] < 0.0f)) {
+      index = Tp[0] > Tp[1] ? 0 : 1;
+      if (Tp[2] > Tp[index]) index = 2;
+    }
+
+    if (index >= 0) {
+      shown_disjoint = true;
+
+      const glm::vec3& qIndex = q[index];
+
+      glm::vec3 V = qIndex - p[0];
+      glm::vec3 Z = glm::cross(Sn, Sv[0]);
+      if (glm::dot(V, Z) > 0.0f) {
+        V = qIndex - p[1];
+        Z = glm::cross(Sn, Sv[1]);
+        if (glm::dot(V, Z) > 0.0f) {
+          V = qIndex - p[2];
+          Z = glm::cross(Sn, Sv[2]);
+          if (glm::dot(V, Z) > 0.0f) {
+            glm::vec3 cp = qIndex + Sn * Tp[index] / Snl;
+            glm::vec3 cq = qIndex;
+            return glm::dot(cp - cq, cp - cq);
+          }
+        }
+      }
+    }
+  }
+
+  glm::vec3 Tn = glm::cross(Tv[0], Tv[1]);
+  float Tnl = glm::dot(Tn, Tn);
+
+  if (Tnl > 1e-15f) {
+    const glm::vec3 Sp(glm::dot(q[0] - p[0], Tn), glm::dot(q[0] - p[1], Tn),
+                       glm::dot(q[0] - p[2], Tn));
+
+    int index = -1;
+    if ((Sp[0] > 0.0f) && (Sp[1] > 0.0f) && (Sp[2] > 0.0f)) {
+      index = Sp[0] < Sp[1] ? 0 : 1;
+      if (Sp[2] < Sp[index]) index = 2;
+    } else if ((Sp[0] < 0.0f) && (Sp[1] < 0.0f) && (Sp[2] < 0.0f)) {
+      index = Sp[0] > Sp[1] ? 0 : 1;
+      if (Sp[2] > Sp[index]) index = 2;
+    }
+
+    if (index >= 0) {
+      shown_disjoint = true;
+
+      const glm::vec3& pIndex = p[index];
+
+      glm::vec3 V = pIndex - q[0];
+      glm::vec3 Z = glm::cross(Tn, Tv[0]);
+      if (glm::dot(V, Z) > 0.0f) {
+        V = pIndex - q[1];
+        Z = glm::cross(Tn, Tv[1]);
+        if (glm::dot(V, Z) > 0.0f) {
+          V = pIndex - q[2];
+          Z = glm::cross(Tn, Tv[2]);
+          if (glm::dot(V, Z) > 0.0f) {
+            glm::vec3 cp = pIndex;
+            glm::vec3 cq = pIndex + Tn * Sp[index] / Tnl;
+            return glm::dot(cp - cq, cp - cq);
+          }
+        }
+      }
+    }
+  }
+
+  return shown_disjoint ? mindd : 0.0f;
+};
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/utils.h b/thirdparty/manifold/src/utilities/include/utils.h
new file mode 100644
index 000000000000..1c13cd0d367a
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/utils.h
@@ -0,0 +1,242 @@
+// Copyright 2020 The Manifold Authors, Jared Hoberock and Nathan Bell of
+// NVIDIA Research
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <atomic>
+#include <mutex>
+#include <unordered_map>
+
+#ifdef MANIFOLD_DEBUG
+#include <chrono>
+#include <iostream>
+#endif
+
+#include "par.h"
+#include "vec.h"
+
+#if __has_include(<tracy/Tracy.hpp>)
+#include <tracy/Tracy.hpp>
+#else
+#define FrameMarkStart(x)
+#define FrameMarkEnd(x)
+// putting ZoneScoped in a function will instrument the function execution when
+// TRACY_ENABLE is set, which allows the profiler to record more accurate
+// timing.
+#define ZoneScoped
+#define ZoneScopedN(name)
+#endif
+
+namespace manifold {
+
+/** @defgroup Private
+ *  @brief Internal classes of the library; not currently part of the public API
+ *  @{
+ */
+#ifdef MANIFOLD_DEBUG
+struct Timer {
+  std::chrono::high_resolution_clock::time_point start, end;
+
+  void Start() { start = std::chrono::high_resolution_clock::now(); }
+
+  void Stop() { end = std::chrono::high_resolution_clock::now(); }
+
+  float Elapsed() {
+    return std::chrono::duration_cast<std::chrono::milliseconds>(end - start)
+        .count();
+  }
+  void Print(std::string message) {
+    std::cout << "----------- " << std::round(Elapsed()) << " ms for "
+              << message << std::endl;
+  }
+};
+#endif
+
+template <typename... Iters>
+thrust::zip_iterator<thrust::tuple<Iters...>> zip(Iters... iters) {
+  return thrust::make_zip_iterator(thrust::make_tuple(iters...));
+}
+
+template <typename A, typename B>
+thrust::permutation_iterator<A, B> perm(A a, B b) {
+  return thrust::make_permutation_iterator(a, b);
+}
+
+template <typename T>
+thrust::counting_iterator<T> countAt(T i) {
+  return thrust::make_counting_iterator(i);
+}
+
+inline int Next3(int i) {
+  constexpr glm::ivec3 next3(1, 2, 0);
+  return next3[i];
+}
+
+inline int Prev3(int i) {
+  constexpr glm::ivec3 prev3(2, 0, 1);
+  return prev3[i];
+}
+
+template <typename T>
+T AtomicAdd(T& target, T add) {
+  std::atomic<T>& tar = reinterpret_cast<std::atomic<T>&>(target);
+  T old_val = tar.load();
+  while (!tar.compare_exchange_weak(old_val, old_val + add,
+                                    std::memory_order_seq_cst))
+    ;
+  return old_val;
+}
+
+template <>
+inline int AtomicAdd(int& target, int add) {
+  std::atomic<int>& tar = reinterpret_cast<std::atomic<int>&>(target);
+  int old_val = tar.fetch_add(add, std::memory_order_seq_cst);
+  return old_val;
+}
+
+// Copied from
+// https://github.com/thrust/thrust/blob/master/examples/strided_range.cu
+template <typename Iterator>
+class strided_range {
+ public:
+  typedef typename thrust::iterator_difference<Iterator>::type difference_type;
+
+  struct stride_functor
+      : public thrust::unary_function<difference_type, difference_type> {
+    difference_type stride;
+
+    stride_functor(difference_type stride) : stride(stride) {}
+
+    difference_type operator()(const difference_type& i) const {
+      return stride * i;
+    }
+  };
+
+  typedef typename thrust::counting_iterator<difference_type> CountingIterator;
+  typedef typename thrust::transform_iterator<stride_functor, CountingIterator>
+      TransformIterator;
+  typedef typename thrust::permutation_iterator<Iterator, TransformIterator>
+      PermutationIterator;
+
+  // type of the strided_range iterator
+  typedef PermutationIterator iterator;
+
+  // construct strided_range for the range [first,last)
+  strided_range(Iterator first, Iterator last, difference_type stride)
+      : first(first), last(last), stride(stride) {}
+  strided_range() {}
+
+  iterator begin(void) const {
+    return PermutationIterator(
+        first, TransformIterator(CountingIterator(0), stride_functor(stride)));
+  }
+
+  iterator end(void) const {
+    return begin() + ((last - first) + (stride - 1)) / stride;
+  }
+
+ protected:
+  Iterator first;
+  Iterator last;
+  difference_type stride;
+};
+
+template <typename T>
+class ConcurrentSharedPtr {
+ public:
+  ConcurrentSharedPtr(T value) : impl(std::make_shared<T>(value)) {}
+  ConcurrentSharedPtr(const ConcurrentSharedPtr<T>& other)
+      : impl(other.impl), mutex(other.mutex) {}
+  class SharedPtrGuard {
+   public:
+    SharedPtrGuard(std::recursive_mutex* mutex, T* content)
+        : mutex(mutex), content(content) {
+      mutex->lock();
+    }
+    ~SharedPtrGuard() { mutex->unlock(); }
+
+    T& operator*() { return *content; }
+    T* operator->() { return content; }
+
+   private:
+    std::recursive_mutex* mutex;
+    T* content;
+  };
+  SharedPtrGuard GetGuard() { return SharedPtrGuard(mutex.get(), impl.get()); };
+  unsigned int UseCount() { return impl.use_count(); };
+
+ private:
+  std::shared_ptr<T> impl;
+  std::shared_ptr<std::recursive_mutex> mutex =
+      std::make_shared<std::recursive_mutex>();
+};
+
+template <typename I = int, typename R = unsigned char>
+struct UnionFind {
+  Vec<I> parents;
+  // we do union by rank
+  // note that we shift rank by 1, rank 0 means it is not connected to anything
+  // else
+  Vec<R> ranks;
+
+  UnionFind(I numNodes) : parents(numNodes), ranks(numNodes, 0) {
+    sequence(autoPolicy(numNodes), parents.begin(), parents.end());
+  }
+
+  I find(I x) {
+    while (parents[x] != x) {
+      parents[x] = parents[parents[x]];
+      x = parents[x];
+    }
+    return x;
+  }
+
+  void unionXY(I x, I y) {
+    if (x == y) return;
+    if (ranks[x] == 0) ranks[x] = 1;
+    if (ranks[y] == 0) ranks[y] = 1;
+    x = find(x);
+    y = find(y);
+    if (x == y) return;
+    if (ranks[x] < ranks[y]) std::swap(x, y);
+    if (ranks[x] == ranks[y]) ranks[x]++;
+    parents[y] = x;
+  }
+
+  I connectedComponents(std::vector<I>& components) {
+    components.resize(parents.size());
+    I lonelyNodes = 0;
+    std::unordered_map<I, I> toLabel;
+    for (size_t i = 0; i < parents.size(); ++i) {
+      // we optimize for connected component of size 1
+      // no need to put them into the hashmap
+      if (ranks[i] == 0) {
+        components[i] = static_cast<I>(toLabel.size()) + lonelyNodes++;
+        continue;
+      }
+      parents[i] = find(i);
+      auto iter = toLabel.find(parents[i]);
+      if (iter == toLabel.end()) {
+        I s = static_cast<I>(toLabel.size()) + lonelyNodes;
+        toLabel.insert(std::make_pair(parents[i], s));
+        components[i] = s;
+      } else {
+        components[i] = iter->second;
+      }
+    }
+    return toLabel.size() + lonelyNodes;
+  }
+};
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/vec.h b/thirdparty/manifold/src/utilities/include/vec.h
new file mode 100644
index 000000000000..5409843d94a9
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/vec.h
@@ -0,0 +1,244 @@
+// Copyright 2021 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+#include <exception>
+#if TRACY_ENABLE && TRACY_MEMORY_USAGE
+#include "tracy/Tracy.hpp"
+#else
+#define TracyAllocS(ptr, size, n) (void)0
+#define TracyFreeS(ptr, n) (void)0
+#endif
+
+#include "par.h"
+#include "public.h"
+#include "vec_view.h"
+
+namespace manifold {
+
+/** @addtogroup Private
+ *  @{
+ */
+template <typename T>
+class Vec;
+
+/*
+ * Specialized vector implementation with multithreaded fill and uninitialized
+ * memory optimizations.
+ * Note that the constructor and resize function will not perform initialization
+ * if the parameter val is not set. Also, this implementation is a toy
+ * implementation that did not consider things like non-trivial
+ * constructor/destructor, please keep T trivial.
+ */
+template <typename T>
+class Vec : public VecView<T> {
+ public:
+  Vec() {}
+
+  // Note that the vector constructed with this constructor will contain
+  // uninitialized memory. Please specify `val` if you need to make sure that
+  // the data is initialized.
+  Vec(size_t size) {
+    reserve(size);
+    this->size_ = size;
+  }
+
+  Vec(size_t size, T val) { resize(size, val); }
+
+  Vec(const Vec<T> &vec) {
+    this->size_ = vec.size();
+    this->capacity_ = this->size_;
+    auto policy = autoPolicy(this->size_);
+    if (this->size_ != 0) {
+      this->ptr_ = reinterpret_cast<T *>(malloc(this->size_ * sizeof(T)));
+      if (this->ptr_ == nullptr) throw std::bad_alloc();
+      TracyAllocS(this->ptr_, this->size_ * sizeof(T), 3);
+      uninitialized_copy(policy, vec.begin(), vec.end(), this->ptr_);
+    }
+  }
+
+  Vec(const std::vector<T> &vec) {
+    this->size_ = vec.size();
+    this->capacity_ = this->size_;
+    auto policy = autoPolicy(this->size_);
+    if (this->size_ != 0) {
+      this->ptr_ = reinterpret_cast<T *>(malloc(this->size_ * sizeof(T)));
+      if (this->ptr_ == nullptr) throw std::bad_alloc();
+      TracyAllocS(this->ptr_, this->size_ * sizeof(T), 3);
+      uninitialized_copy(policy, vec.begin(), vec.end(), this->ptr_);
+    }
+  }
+
+  Vec(Vec<T> &&vec) {
+    this->ptr_ = vec.ptr_;
+    this->size_ = vec.size_;
+    capacity_ = vec.capacity_;
+    vec.ptr_ = nullptr;
+    vec.size_ = 0;
+    vec.capacity_ = 0;
+  }
+
+  operator VecView<T>() { return {this->ptr_, this->size_}; }
+  operator VecView<const T>() const { return {this->ptr_, this->size_}; }
+
+  ~Vec() {
+    if (this->ptr_ != nullptr) {
+      TracyFreeS(this->ptr_, 3);
+      free(this->ptr_);
+    }
+    this->ptr_ = nullptr;
+    this->size_ = 0;
+    capacity_ = 0;
+  }
+
+  Vec<T> &operator=(const Vec<T> &other) {
+    if (&other == this) return *this;
+    if (this->ptr_ != nullptr) {
+      TracyFreeS(this->ptr_, 3);
+      free(this->ptr_);
+    }
+    this->size_ = other.size_;
+    capacity_ = other.size_;
+    auto policy = autoPolicy(this->size_);
+    if (this->size_ != 0) {
+      this->ptr_ = reinterpret_cast<T *>(malloc(this->size_ * sizeof(T)));
+      if (this->ptr_ == nullptr) throw std::bad_alloc();
+      TracyAllocS(this->ptr_, this->size_ * sizeof(T), 3);
+      uninitialized_copy(policy, other.begin(), other.end(), this->ptr_);
+    }
+    return *this;
+  }
+
+  Vec<T> &operator=(Vec<T> &&other) {
+    if (&other == this) return *this;
+    if (this->ptr_ != nullptr) {
+      TracyFreeS(this->ptr_, 3);
+      free(this->ptr_);
+    }
+    this->size_ = other.size_;
+    capacity_ = other.capacity_;
+    this->ptr_ = other.ptr_;
+    other.ptr_ = nullptr;
+    other.size_ = 0;
+    other.capacity_ = 0;
+    return *this;
+  }
+
+  operator VecView<T>() const { return {this->ptr_, this->size_}; }
+
+  void swap(Vec<T> &other) {
+    std::swap(this->ptr_, other.ptr_);
+    std::swap(this->size_, other.size_);
+    std::swap(capacity_, other.capacity_);
+  }
+
+  inline void push_back(const T &val) {
+    if (this->size_ >= capacity_) {
+      // avoid dangling pointer in case val is a reference of our array
+      T val_copy = val;
+      reserve(capacity_ == 0 ? 128 : capacity_ * 2);
+      this->ptr_[this->size_++] = val_copy;
+      return;
+    }
+    this->ptr_[this->size_++] = val;
+  }
+
+  void reserve(size_t n) {
+    if (n > capacity_) {
+      T *newBuffer = reinterpret_cast<T *>(malloc(n * sizeof(T)));
+      if (newBuffer == nullptr) throw std::bad_alloc();
+      TracyAllocS(newBuffer, n * sizeof(T), 3);
+      if (this->size_ > 0)
+        uninitialized_copy(autoPolicy(this->size_), this->ptr_,
+                           this->ptr_ + this->size_, newBuffer);
+      if (this->ptr_ != nullptr) {
+        TracyFreeS(this->ptr_, 3);
+        free(this->ptr_);
+      }
+      this->ptr_ = newBuffer;
+      capacity_ = n;
+    }
+  }
+
+  void resize(size_t newSize, T val = T()) {
+    bool shrink = this->size_ > 2 * newSize;
+    reserve(newSize);
+    if (this->size_ < newSize) {
+      uninitialized_fill(autoPolicy(newSize - this->size_),
+                         this->ptr_ + this->size_, this->ptr_ + newSize, val);
+    }
+    this->size_ = newSize;
+    if (shrink) shrink_to_fit();
+  }
+
+  void shrink_to_fit() {
+    T *newBuffer = nullptr;
+    if (this->size_ > 0) {
+      newBuffer = reinterpret_cast<T *>(malloc(this->size_ * sizeof(T)));
+      if (newBuffer == nullptr) throw std::bad_alloc();
+      TracyAllocS(newBuffer, this->size_ * sizeof(T), 3);
+      uninitialized_copy(autoPolicy(this->size_), this->ptr_,
+                         this->ptr_ + this->size_, newBuffer);
+    }
+    if (this->ptr_ != nullptr) {
+      TracyFreeS(this->ptr_, 3);
+      free(this->ptr_);
+    }
+    this->ptr_ = newBuffer;
+    capacity_ = this->size_;
+  }
+
+  VecView<T> view(size_t offset = 0,
+                  size_t length = std::numeric_limits<size_t>::max()) {
+    if (length == std::numeric_limits<size_t>::max()) {
+      length = this->size_ - offset;
+      if (length < 0) throw std::out_of_range("Vec::view out of range");
+    } else if (offset + length > this->size_ || offset < 0) {
+      throw std::out_of_range("Vec::view out of range");
+    } else if (length < 0) {
+      throw std::out_of_range("Vec::view negative length is not allowed");
+    }
+    return VecView<T>(this->ptr_ + offset, length);
+  }
+
+  VecView<const T> cview(
+      size_t offset = 0,
+      size_t length = std::numeric_limits<size_t>::max()) const {
+    if (length == std::numeric_limits<size_t>::max()) {
+      length = this->size_ - offset;
+      if (length < 0) throw std::out_of_range("Vec::cview out of range");
+    } else if (offset + length > this->size_ || offset < 0) {
+      throw std::out_of_range("Vec::cview out of range");
+    } else if (length < 0) {
+      throw std::out_of_range("Vec::cview negative length is not allowed");
+    }
+    return VecView<const T>(this->ptr_ + offset, length);
+  }
+
+  VecView<const T> view(
+      size_t offset = 0,
+      size_t length = std::numeric_limits<size_t>::max()) const {
+    return cview(offset, length);
+  }
+
+  T *data() { return this->ptr_; }
+  const T *data() const { return this->ptr_; }
+
+ private:
+  size_t capacity_ = 0;
+
+  static_assert(std::is_trivially_destructible<T>::value);
+};
+/** @} */
+}  // namespace manifold
diff --git a/thirdparty/manifold/src/utilities/include/vec_view.h b/thirdparty/manifold/src/utilities/include/vec_view.h
new file mode 100644
index 000000000000..5945f20c8106
--- /dev/null
+++ b/thirdparty/manifold/src/utilities/include/vec_view.h
@@ -0,0 +1,116 @@
+// Copyright 2023 The Manifold Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <stdexcept>
+
+namespace manifold {
+
+/**
+ * View for Vec, can perform offset operation.
+ * This will be invalidated when the original vector is dropped or changes
+ * length. Roughly equivalent to std::span<T> from c++20
+ */
+template <typename T>
+class VecView {
+ public:
+  using Iter = T *;
+  using IterC = const T *;
+
+  VecView(T *ptr, size_t size) : ptr_(ptr), size_(size) {}
+
+  VecView(const VecView &other) {
+    ptr_ = other.ptr_;
+    size_ = other.size_;
+  }
+
+  VecView &operator=(const VecView &other) {
+    ptr_ = other.ptr_;
+    size_ = other.size_;
+    return *this;
+  }
+
+  // allows conversion to a const VecView
+  operator VecView<const T>() const { return {ptr_, size_}; }
+
+  inline const T &operator[](size_t i) const {
+    if (i >= size_) {
+      throw std::out_of_range("Vec out of range");
+    }
+    return ptr_[i];
+  }
+
+  inline T &operator[](size_t i) {
+    if (i >= size_) {
+      throw std::out_of_range("Vec out of range");
+    }
+    return ptr_[i];
+  }
+
+  IterC cbegin() const { return ptr_; }
+  IterC cend() const { return ptr_ + size_; }
+
+  IterC begin() const { return cbegin(); }
+  IterC end() const { return cend(); }
+
+  Iter begin() { return ptr_; }
+  Iter end() { return ptr_ + size_; }
+
+  const T &front() const {
+    if (size_ == 0)
+      throw std::out_of_range("attempt to take the front of an empty vector");
+    return ptr_[0];
+  }
+
+  const T &back() const {
+    if (size_ == 0)
+      throw std::out_of_range("attempt to take the back of an empty vector");
+    return ptr_[size_ - 1];
+  }
+
+  T &front() {
+    if (size_ == 0)
+      throw std::out_of_range("attempt to take the front of an empty vector");
+    return ptr_[0];
+  }
+
+  T &back() {
+    if (size_ == 0)
+      throw std::out_of_range("attempt to take the back of an empty vector");
+    return ptr_[size_ - 1];
+  }
+
+  size_t size() const { return size_; }
+
+  bool empty() const { return size_ == 0; }
+
+#ifdef MANIFOLD_DEBUG
+  void Dump() const {
+    std::cout << "Vec = " << std::endl;
+    for (size_t i = 0; i < size(); ++i) {
+      std::cout << i << ", " << ptr_[i] << ", " << std::endl;
+    }
+    std::cout << std::endl;
+  }
+#endif
+
+ protected:
+  T *ptr_ = nullptr;
+  size_t size_ = 0;
+
+  VecView() = default;
+};
+
+}  // namespace manifold
diff --git a/thirdparty/manifold/thirdparty/glm/.gitrepo b/thirdparty/manifold/thirdparty/glm/.gitrepo
new file mode 100644
index 000000000000..49bb6ea957e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/.gitrepo
@@ -0,0 +1,12 @@
+; DO NOT EDIT (unless you know what you are doing)
+;
+; This subdirectory is a git "subrepo", and this file is maintained by the
+; git-subrepo command. See https://github.com/ingydotnet/git-subrepo#readme
+;
+[subrepo]
+	remote = https://github.com/g-truc/glm.git
+	branch = b06b775c1c80af51a1183c0e167f9de3b2351a79
+	commit = b06b775c1c80af51a1183c0e167f9de3b2351a79
+	parent = e416591e8048b25becc0a3717b106c815dcb060b
+	method = merge
+	cmdver = 0.4.6
diff --git a/thirdparty/manifold/thirdparty/glm/copying.txt b/thirdparty/manifold/thirdparty/glm/copying.txt
new file mode 100644
index 000000000000..779c32fb9afe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/copying.txt
@@ -0,0 +1,54 @@
+================================================================================
+OpenGL Mathematics (GLM)
+--------------------------------------------------------------------------------
+GLM is licensed under The Happy Bunny License or MIT License
+
+================================================================================
+The Happy Bunny License (Modified MIT License)
+--------------------------------------------------------------------------------
+Copyright (c) 2005 - G-Truc Creation
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+Restrictions:
+ By making use of the Software for military purposes, you choose to make a
+ Bunny unhappy.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+
+================================================================================
+The MIT License
+--------------------------------------------------------------------------------
+Copyright (c) 2005 - G-Truc Creation
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
diff --git a/thirdparty/manifold/thirdparty/glm/glm/CMakeLists.txt b/thirdparty/manifold/thirdparty/glm/glm/CMakeLists.txt
new file mode 100644
index 000000000000..178d23abdfbc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/CMakeLists.txt
@@ -0,0 +1,69 @@
+file(GLOB ROOT_SOURCE *.cpp)
+file(GLOB ROOT_INLINE *.inl)
+file(GLOB ROOT_HEADER *.hpp)
+file(GLOB ROOT_TEXT ../*.txt)
+file(GLOB ROOT_MD ../*.md)
+file(GLOB ROOT_NAT ../util/glm.natvis)
+
+file(GLOB_RECURSE CORE_SOURCE ./detail/*.cpp)
+file(GLOB_RECURSE CORE_INLINE ./detail/*.inl)
+file(GLOB_RECURSE CORE_HEADER ./detail/*.hpp)
+
+file(GLOB_RECURSE EXT_SOURCE ./ext/*.cpp)
+file(GLOB_RECURSE EXT_INLINE ./ext/*.inl)
+file(GLOB_RECURSE EXT_HEADER ./ext/*.hpp)
+
+file(GLOB_RECURSE GTC_SOURCE ./gtc/*.cpp)
+file(GLOB_RECURSE GTC_INLINE ./gtc/*.inl)
+file(GLOB_RECURSE GTC_HEADER ./gtc/*.hpp)
+
+file(GLOB_RECURSE GTX_SOURCE ./gtx/*.cpp)
+file(GLOB_RECURSE GTX_INLINE ./gtx/*.inl)
+file(GLOB_RECURSE GTX_HEADER ./gtx/*.hpp)
+
+file(GLOB_RECURSE SIMD_SOURCE ./simd/*.cpp)
+file(GLOB_RECURSE SIMD_INLINE ./simd/*.inl)
+file(GLOB_RECURSE SIMD_HEADER ./simd/*.h)
+
+source_group("Text Files" FILES ${ROOT_TEXT} ${ROOT_MD})
+source_group("Core Files" FILES ${CORE_SOURCE})
+source_group("Core Files" FILES ${CORE_INLINE})
+source_group("Core Files" FILES ${CORE_HEADER})
+source_group("EXT Files" FILES ${EXT_SOURCE})
+source_group("EXT Files" FILES ${EXT_INLINE})
+source_group("EXT Files" FILES ${EXT_HEADER})
+source_group("GTC Files" FILES ${GTC_SOURCE})
+source_group("GTC Files" FILES ${GTC_INLINE})
+source_group("GTC Files" FILES ${GTC_HEADER})
+source_group("GTX Files" FILES ${GTX_SOURCE})
+source_group("GTX Files" FILES ${GTX_INLINE})
+source_group("GTX Files" FILES ${GTX_HEADER})
+source_group("SIMD Files" FILES ${SIMD_SOURCE})
+source_group("SIMD Files" FILES ${SIMD_INLINE})
+source_group("SIMD Files" FILES ${SIMD_HEADER})
+
+add_library(glm-header-only INTERFACE)
+add_library(glm::glm-header-only ALIAS glm-header-only)
+
+target_include_directories(glm-header-only INTERFACE
+	"$<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}>"
+	"$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>"
+)
+
+if (GLM_BUILD_LIBRARY)
+	add_library(glm
+		${ROOT_TEXT}      ${ROOT_MD}        ${ROOT_NAT}
+		${ROOT_SOURCE}    ${ROOT_INLINE}    ${ROOT_HEADER}
+		${CORE_SOURCE}    ${CORE_INLINE}    ${CORE_HEADER}
+		${EXT_SOURCE}     ${EXT_INLINE}     ${EXT_HEADER}
+		${GTC_SOURCE}     ${GTC_INLINE}     ${GTC_HEADER}
+		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER}
+		${SIMD_SOURCE}    ${SIMD_INLINE}    ${SIMD_HEADER}
+	)
+	add_library(glm::glm ALIAS glm)
+	target_link_libraries(glm PUBLIC glm-header-only)
+else()
+	add_library(glm INTERFACE)
+	add_library(glm::glm ALIAS glm)
+	target_link_libraries(glm INTERFACE glm-header-only)
+endif()
diff --git a/thirdparty/manifold/thirdparty/glm/glm/common.hpp b/thirdparty/manifold/thirdparty/glm/glm/common.hpp
new file mode 100644
index 000000000000..b59657d549f2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/common.hpp
@@ -0,0 +1,539 @@
+/// @ref core
+/// @file glm/common.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+///
+/// @defgroup core_func_common Common functions
+/// @ingroup core
+///
+/// Provides GLSL common functions
+///
+/// These all operate component-wise. The description is per component.
+///
+/// Include <glm/common.hpp> to use these core features.
+
+#pragma once
+
+#include "detail/qualifier.hpp"
+#include "detail/_fixes.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_func_common
+	/// @{
+
+	/// Returns x if x >= 0; otherwise, it returns -x.
+	///
+	/// @tparam genType floating-point or signed integer; scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/abs.xml">GLSL abs man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType abs(genType x);
+
+	/// Returns x if x >= 0; otherwise, it returns -x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or signed integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/abs.xml">GLSL abs man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> abs(vec<L, T, Q> const& x);
+
+	/// Returns 1.0 if x > 0, 0.0 if x == 0, or -1.0 if x < 0.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/sign.xml">GLSL sign man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> sign(vec<L, T, Q> const& x);
+
+	/// Returns a value equal to the nearest integer that is less then or equal to x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/floor.xml">GLSL floor man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> floor(vec<L, T, Q> const& x);
+
+	/// Returns a value equal to the nearest integer to x
+	/// whose absolute value is not larger than the absolute value of x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/trunc.xml">GLSL trunc man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> trunc(vec<L, T, Q> const& x);
+
+	/// Returns a value equal to the nearest integer to x.
+	/// The fraction 0.5 will round in a direction chosen by the
+	/// implementation, presumably the direction that is fastest.
+	/// This includes the possibility that round(x) returns the
+	/// same value as roundEven(x) for all values of x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/round.xml">GLSL round man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> round(vec<L, T, Q> const& x);
+
+	/// Returns a value equal to the nearest integer to x.
+	/// A fractional part of 0.5 will round toward the nearest even
+	/// integer. (Both 3.5 and 4.5 for x will return 4.0.)
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/roundEven.xml">GLSL roundEven man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	/// @see <a href="http://developer.amd.com/documentation/articles/pages/New-Round-to-Even-Technique.aspx">New round to even technique</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> roundEven(vec<L, T, Q> const& x);
+
+	/// Returns a value equal to the nearest integer
+	/// that is greater than or equal to x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/ceil.xml">GLSL ceil man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> ceil(vec<L, T, Q> const& x);
+
+	/// Return x - floor(x).
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/fract.xml">GLSL fract man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType fract(genType x);
+
+	/// Return x - floor(x).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/fract.xml">GLSL fract man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fract(vec<L, T, Q> const& x);
+
+	template<typename genType>
+	GLM_FUNC_DECL genType mod(genType x, genType y);
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> mod(vec<L, T, Q> const& x, T y);
+
+	/// Modulus. Returns x - y * floor(x / y)
+	/// for each component in x using the floating point value y.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types, include glm/gtc/integer for integer scalar types support
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/mod.xml">GLSL mod man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> mod(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the fractional part of x and sets i to the integer
+	/// part (as a whole number floating point value). Both the
+	/// return value and the output parameter will have the same
+	/// sign as x.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/modf.xml">GLSL modf man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType modf(genType x, genType& i);
+
+	/// Returns y if y < x; otherwise, it returns x.
+	///
+	/// @tparam genType Floating-point or integer; scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/min.xml">GLSL min man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType min(genType x, genType y);
+
+	/// Returns y if y < x; otherwise, it returns x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/min.xml">GLSL min man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& x, T y);
+
+	/// Returns y if y < x; otherwise, it returns x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/min.xml">GLSL min man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns y if x < y; otherwise, it returns x.
+	///
+	/// @tparam genType Floating-point or integer; scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/max.xml">GLSL max man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType max(genType x, genType y);
+
+	/// Returns y if x < y; otherwise, it returns x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/max.xml">GLSL max man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& x, T y);
+
+	/// Returns y if x < y; otherwise, it returns x.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/max.xml">GLSL max man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns min(max(x, minVal), maxVal) for each component in x
+	/// using the floating-point values minVal and maxVal.
+	///
+	/// @tparam genType Floating-point or integer; scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/clamp.xml">GLSL clamp man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType clamp(genType x, genType minVal, genType maxVal);
+
+	/// Returns min(max(x, minVal), maxVal) for each component in x
+	/// using the floating-point values minVal and maxVal.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/clamp.xml">GLSL clamp man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> clamp(vec<L, T, Q> const& x, T minVal, T maxVal);
+
+	/// Returns min(max(x, minVal), maxVal) for each component in x
+	/// using the floating-point values minVal and maxVal.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/clamp.xml">GLSL clamp man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> clamp(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal);
+
+	/// If genTypeU is a floating scalar or vector:
+	/// Returns x * (1.0 - a) + y * a, i.e., the linear blend of
+	/// x and y using the floating-point value a.
+	/// The value for a is not restricted to the range [0, 1].
+	///
+	/// If genTypeU is a boolean scalar or vector:
+	/// Selects which vector each returned component comes
+	/// from. For a component of 'a' that is false, the
+	/// corresponding component of 'x' is returned. For a
+	/// component of 'a' that is true, the corresponding
+	/// component of 'y' is returned. Components of 'x' and 'y' that
+	/// are not selected are allowed to be invalid floating point
+	/// values and will have no effect on the results. Thus, this
+	/// provides different functionality than
+	/// genType mix(genType x, genType y, genType(a))
+	/// where a is a Boolean vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/mix.xml">GLSL mix man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	///
+	/// @param[in]  x Value to interpolate.
+	/// @param[in]  y Value to interpolate.
+	/// @param[in]  a Interpolant.
+	///
+	/// @tparam	genTypeT Floating point scalar or vector.
+	/// @tparam genTypeU Floating point or boolean scalar or vector. It can't be a vector if it is the length of genTypeT.
+	///
+	/// @code
+	/// #include <glm/glm.hpp>
+	/// ...
+	/// float a;
+	/// bool b;
+	/// glm::dvec3 e;
+	/// glm::dvec3 f;
+	/// glm::vec4 g;
+	/// glm::vec4 h;
+	/// ...
+	/// glm::vec4 r = glm::mix(g, h, a); // Interpolate with a floating-point scalar two vectors.
+	/// glm::vec4 s = glm::mix(g, h, b); // Returns g or h;
+	/// glm::dvec3 t = glm::mix(e, f, a); // Types of the third parameter is not required to match with the first and the second.
+	/// glm::vec4 u = glm::mix(g, h, r); // Interpolations can be perform per component with a vector for the last parameter.
+	/// @endcode
+	template<typename genTypeT, typename genTypeU>
+	GLM_FUNC_DECL GLM_CONSTEXPR genTypeT mix(genTypeT x, genTypeT y, genTypeU a);
+
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> mix(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, U, Q> const& a);
+
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> mix(vec<L, T, Q> const& x, vec<L, T, Q> const& y, U a);
+
+	/// Returns 0.0 if x < edge, otherwise it returns 1.0 for each component of a genType.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/step.xml">GLSL step man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType step(genType edge, genType x);
+
+	/// Returns 0.0 if x < edge, otherwise it returns 1.0.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/step.xml">GLSL step man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> step(T edge, vec<L, T, Q> const& x);
+
+	/// Returns 0.0 if x < edge, otherwise it returns 1.0.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/step.xml">GLSL step man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> step(vec<L, T, Q> const& edge, vec<L, T, Q> const& x);
+
+	/// Returns 0.0 if x <= edge0 and 1.0 if x >= edge1 and
+	/// performs smooth Hermite interpolation between 0 and 1
+	/// when edge0 < x < edge1. This is useful in cases where
+	/// you would want a threshold function with a smooth
+	/// transition. This is equivalent to:
+	/// genType t;
+	/// t = clamp ((x - edge0) / (edge1 - edge0), 0, 1);
+	/// return t * t * (3 - 2 * t);
+	/// Results are undefined if edge0 >= edge1.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/smoothstep.xml">GLSL smoothstep man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType smoothstep(genType edge0, genType edge1, genType x);
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> smoothstep(T edge0, T edge1, vec<L, T, Q> const& x);
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> smoothstep(vec<L, T, Q> const& edge0, vec<L, T, Q> const& edge1, vec<L, T, Q> const& x);
+
+	/// Returns true if x holds a NaN (not a number)
+	/// representation in the underlying implementation's set of
+	/// floating point representations. Returns false otherwise,
+	/// including for implementations with no NaN
+	/// representations.
+	///
+	/// /!\ When using compiler fast math, this function may fail.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/isnan.xml">GLSL isnan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> isnan(vec<L, T, Q> const& x);
+
+	/// Returns true if x holds a positive infinity or negative
+	/// infinity representation in the underlying implementation's
+	/// set of floating point representations. Returns false
+	/// otherwise, including for implementations with no infinity
+	/// representations.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/isinf.xml">GLSL isinf man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> isinf(vec<L, T, Q> const& x);
+
+	/// Returns a signed integer value representing
+	/// the encoding of a floating-point value. The floating-point
+	/// value's bit-level representation is preserved.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/floatBitsToInt.xml">GLSL floatBitsToInt man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	GLM_FUNC_DECL int floatBitsToInt(float v);
+
+	/// Returns a signed integer value representing
+	/// the encoding of a floating-point value. The floatingpoint
+	/// value's bit-level representation is preserved.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/floatBitsToInt.xml">GLSL floatBitsToInt man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> floatBitsToInt(vec<L, float, Q> const& v);
+
+	/// Returns a unsigned integer value representing
+	/// the encoding of a floating-point value. The floatingpoint
+	/// value's bit-level representation is preserved.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/floatBitsToUint.xml">GLSL floatBitsToUint man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	GLM_FUNC_DECL uint floatBitsToUint(float v);
+
+	/// Returns a unsigned integer value representing
+	/// the encoding of a floating-point value. The floatingpoint
+	/// value's bit-level representation is preserved.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/floatBitsToUint.xml">GLSL floatBitsToUint man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, uint, Q> floatBitsToUint(vec<L, float, Q> const& v);
+
+	/// Returns a floating-point value corresponding to a signed
+	/// integer encoding of a floating-point value.
+	/// If an inf or NaN is passed in, it will not signal, and the
+	/// resulting floating point value is unspecified. Otherwise,
+	/// the bit-level representation is preserved.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/intBitsToFloat.xml">GLSL intBitsToFloat man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	GLM_FUNC_DECL float intBitsToFloat(int v);
+
+	/// Returns a floating-point value corresponding to a signed
+	/// integer encoding of a floating-point value.
+	/// If an inf or NaN is passed in, it will not signal, and the
+	/// resulting floating point value is unspecified. Otherwise,
+	/// the bit-level representation is preserved.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/intBitsToFloat.xml">GLSL intBitsToFloat man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, float, Q> intBitsToFloat(vec<L, int, Q> const& v);
+
+	/// Returns a floating-point value corresponding to a
+	/// unsigned integer encoding of a floating-point value.
+	/// If an inf or NaN is passed in, it will not signal, and the
+	/// resulting floating point value is unspecified. Otherwise,
+	/// the bit-level representation is preserved.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/uintBitsToFloat.xml">GLSL uintBitsToFloat man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	GLM_FUNC_DECL float uintBitsToFloat(uint v);
+
+	/// Returns a floating-point value corresponding to a
+	/// unsigned integer encoding of a floating-point value.
+	/// If an inf or NaN is passed in, it will not signal, and the
+	/// resulting floating point value is unspecified. Otherwise,
+	/// the bit-level representation is preserved.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/uintBitsToFloat.xml">GLSL uintBitsToFloat man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, float, Q> uintBitsToFloat(vec<L, uint, Q> const& v);
+
+	/// Computes and returns a * b + c.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/fma.xml">GLSL fma man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType fma(genType const& a, genType const& b, genType const& c);
+
+	/// Splits x into a floating-point significand in the range
+	/// [0.5, 1.0) and an integral exponent of two, such that:
+	/// x = significand * exp(2, exponent)
+	///
+	/// The significand is returned by the function and the
+	/// exponent is returned in the parameter exp. For a
+	/// floating-point value of zero, the significant and exponent
+	/// are both zero. For a floating-point value that is an
+	/// infinity or is not a number, the results are undefined.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/frexp.xml">GLSL frexp man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType frexp(genType x, int& exp);
+	
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> frexp(vec<L, T, Q> const& v, vec<L, int, Q>& exp);
+
+	/// Builds a floating-point number from x and the
+	/// corresponding integral exponent of two in exp, returning:
+	/// significand * exp(2, exponent)
+	///
+	/// If this product is too large to be represented in the
+	/// floating-point type, the result is undefined.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/ldexp.xml">GLSL ldexp man page</a>;
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL genType ldexp(genType const& x, int const& exp);
+	
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> ldexp(vec<L, T, Q> const& v, vec<L, int, Q> const& exp);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_common.inl"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/_features.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/_features.hpp
new file mode 100644
index 000000000000..b0cbe9ff02cf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/_features.hpp
@@ -0,0 +1,394 @@
+#pragma once
+
+// #define GLM_CXX98_EXCEPTIONS
+// #define GLM_CXX98_RTTI
+
+// #define GLM_CXX11_RVALUE_REFERENCES
+// Rvalue references - GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
+
+// GLM_CXX11_TRAILING_RETURN
+// Rvalue references for *this - GCC not supported
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
+
+// GLM_CXX11_NONSTATIC_MEMBER_INIT
+// Initialization of class objects by rvalues - GCC any
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1610.html
+
+// GLM_CXX11_NONSTATIC_MEMBER_INIT
+// Non-static data member initializers - GCC 4.7
+// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2008/n2756.htm
+
+// #define GLM_CXX11_VARIADIC_TEMPLATE
+// Variadic templates - GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2242.pdf
+
+//
+// Extending variadic template template parameters - GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2555.pdf
+
+// #define GLM_CXX11_GENERALIZED_INITIALIZERS
+// Initializer lists - GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2672.htm
+
+// #define GLM_CXX11_STATIC_ASSERT
+// Static assertions - GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
+
+// #define GLM_CXX11_AUTO_TYPE
+// auto-typed variables - GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
+
+// #define GLM_CXX11_AUTO_TYPE
+// Multi-declarator auto - GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
+
+// #define GLM_CXX11_AUTO_TYPE
+// Removal of auto as a storage-class specifier - GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2546.htm
+
+// #define GLM_CXX11_AUTO_TYPE
+// New function declarator syntax - GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
+
+// #define GLM_CXX11_LAMBDAS
+// New wording for C++0x lambdas - GCC 4.5
+// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2009/n2927.pdf
+
+// #define GLM_CXX11_DECLTYPE
+// Declared type of an expression - GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
+
+//
+// Right angle brackets - GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
+
+//
+// Default template arguments for function templates	DR226	GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#226
+
+//
+// Solving the SFINAE problem for expressions	DR339	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2634.html
+
+// #define GLM_CXX11_ALIAS_TEMPLATE
+// Template aliases	N2258	GCC 4.7
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2258.pdf
+
+//
+// Extern templates	N1987	Yes
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
+
+// #define GLM_CXX11_NULLPTR
+// Null pointer constant	N2431	GCC 4.6
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
+
+// #define GLM_CXX11_STRONG_ENUMS
+// Strongly-typed enums	N2347	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
+
+//
+// Forward declarations for enums	N2764	GCC 4.6
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
+
+//
+// Generalized attributes	N2761	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2761.pdf
+
+//
+// Generalized constant expressions	N2235	GCC 4.6
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2235.pdf
+
+//
+// Alignment support	N2341	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2341.pdf
+
+// #define GLM_CXX11_DELEGATING_CONSTRUCTORS
+// Delegating constructors	N1986	GCC 4.7
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1986.pdf
+
+//
+// Inheriting constructors	N2540	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2540.htm
+
+// #define GLM_CXX11_EXPLICIT_CONVERSIONS
+// Explicit conversion operators	N2437	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2437.pdf
+
+//
+// New character types	N2249	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2249.html
+
+//
+// Unicode string literals	N2442	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2442.htm
+
+//
+// Raw string literals	N2442	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2442.htm
+
+//
+// Universal character name literals	N2170	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2170.html
+
+// #define GLM_CXX11_USER_LITERALS
+// User-defined literals		N2765	GCC 4.7
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2765.pdf
+
+//
+// Standard Layout Types	N2342	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2342.htm
+
+// #define GLM_CXX11_DEFAULTED_FUNCTIONS
+// #define GLM_CXX11_DELETED_FUNCTIONS
+// Defaulted and deleted functions	N2346	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2346.htm
+
+//
+// Extended friend declarations	N1791	GCC 4.7
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1791.pdf
+
+//
+// Extending sizeof	N2253	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2253.html
+
+// #define GLM_CXX11_INLINE_NAMESPACES
+// Inline namespaces	N2535	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2535.htm
+
+// #define GLM_CXX11_UNRESTRICTED_UNIONS
+// Unrestricted unions	N2544	GCC 4.6
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2544.pdf
+
+// #define GLM_CXX11_LOCAL_TYPE_TEMPLATE_ARGS
+// Local and unnamed types as template arguments	N2657	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
+
+// #define GLM_CXX11_RANGE_FOR
+// Range-based for	N2930	GCC 4.6
+// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2009/n2930.html
+
+// #define GLM_CXX11_OVERRIDE_CONTROL
+// Explicit virtual overrides	N2928 N3206 N3272	GCC 4.7
+// http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2009/n2928.htm
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
+
+//
+// Minimal support for garbage collection and reachability-based leak detection	N2670	No
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2670.htm
+
+// #define GLM_CXX11_NOEXCEPT
+// Allowing move constructors to throw [noexcept]	N3050	GCC 4.6 (core language only)
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3050.html
+
+//
+// Defining move special member functions	N3053	GCC 4.6
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3053.html
+
+//
+// Sequence points	N2239	Yes
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2239.html
+
+//
+// Atomic operations	N2427	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2239.html
+
+//
+// Strong Compare and Exchange	N2748	GCC 4.5
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2427.html
+
+//
+// Bidirectional Fences	N2752	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2752.htm
+
+//
+// Memory model	N2429	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
+
+//
+// Data-dependency ordering: atomics and memory model	N2664	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2664.htm
+
+//
+// Propagating exceptions	N2179	GCC 4.4
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2179.html
+
+//
+// Abandoning a process and at_quick_exit	N2440	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2440.htm
+
+//
+// Allow atomics use in signal handlers	N2547	Yes
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2547.htm
+
+//
+// Thread-local storage	N2659	GCC 4.8
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2659.htm
+
+//
+// Dynamic initialization and destruction with concurrency	N2660	GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2660.htm
+
+//
+// __func__ predefined identifier	N2340	GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2340.htm
+
+//
+// C99 preprocessor	N1653	GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1653.htm
+
+//
+// long long	N1811	GCC 4.3
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1811.pdf
+
+//
+// Extended integral types	N1988	Yes
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1988.pdf
+
+#if(GLM_COMPILER & GLM_COMPILER_GCC)
+
+#	define GLM_CXX11_STATIC_ASSERT
+
+#elif(GLM_COMPILER & GLM_COMPILER_CLANG)
+#	if(__has_feature(cxx_exceptions))
+#		define GLM_CXX98_EXCEPTIONS
+#	endif
+
+#	if(__has_feature(cxx_rtti))
+#		define GLM_CXX98_RTTI
+#	endif
+
+#	if(__has_feature(cxx_access_control_sfinae))
+#		define GLM_CXX11_ACCESS_CONTROL_SFINAE
+#	endif
+
+#	if(__has_feature(cxx_alias_templates))
+#		define GLM_CXX11_ALIAS_TEMPLATE
+#	endif
+
+#	if(__has_feature(cxx_alignas))
+#		define GLM_CXX11_ALIGNAS
+#	endif
+
+#	if(__has_feature(cxx_attributes))
+#		define GLM_CXX11_ATTRIBUTES
+#	endif
+
+#	if(__has_feature(cxx_constexpr))
+#		define GLM_CXX11_CONSTEXPR
+#	endif
+
+#	if(__has_feature(cxx_decltype))
+#		define GLM_CXX11_DECLTYPE
+#	endif
+
+#	if(__has_feature(cxx_default_function_template_args))
+#		define GLM_CXX11_DEFAULT_FUNCTION_TEMPLATE_ARGS
+#	endif
+
+#	if(__has_feature(cxx_defaulted_functions))
+#		define GLM_CXX11_DEFAULTED_FUNCTIONS
+#	endif
+
+#	if(__has_feature(cxx_delegating_constructors))
+#		define GLM_CXX11_DELEGATING_CONSTRUCTORS
+#	endif
+
+#	if(__has_feature(cxx_deleted_functions))
+#		define GLM_CXX11_DELETED_FUNCTIONS
+#	endif
+
+#	if(__has_feature(cxx_explicit_conversions))
+#		define GLM_CXX11_EXPLICIT_CONVERSIONS
+#	endif
+
+#	if(__has_feature(cxx_generalized_initializers))
+#		define GLM_CXX11_GENERALIZED_INITIALIZERS
+#	endif
+
+#	if(__has_feature(cxx_implicit_moves))
+#		define GLM_CXX11_IMPLICIT_MOVES
+#	endif
+
+#	if(__has_feature(cxx_inheriting_constructors))
+#		define GLM_CXX11_INHERITING_CONSTRUCTORS
+#	endif
+
+#	if(__has_feature(cxx_inline_namespaces))
+#		define GLM_CXX11_INLINE_NAMESPACES
+#	endif
+
+#	if(__has_feature(cxx_lambdas))
+#		define GLM_CXX11_LAMBDAS
+#	endif
+
+#	if(__has_feature(cxx_local_type_template_args))
+#		define GLM_CXX11_LOCAL_TYPE_TEMPLATE_ARGS
+#	endif
+
+#	if(__has_feature(cxx_noexcept))
+#		define GLM_CXX11_NOEXCEPT
+#	endif
+
+#	if(__has_feature(cxx_nonstatic_member_init))
+#		define GLM_CXX11_NONSTATIC_MEMBER_INIT
+#	endif
+
+#	if(__has_feature(cxx_nullptr))
+#		define GLM_CXX11_NULLPTR
+#	endif
+
+#	if(__has_feature(cxx_override_control))
+#		define GLM_CXX11_OVERRIDE_CONTROL
+#	endif
+
+#	if(__has_feature(cxx_reference_qualified_functions))
+#		define GLM_CXX11_REFERENCE_QUALIFIED_FUNCTIONS
+#	endif
+
+#	if(__has_feature(cxx_range_for))
+#		define GLM_CXX11_RANGE_FOR
+#	endif
+
+#	if(__has_feature(cxx_raw_string_literals))
+#		define GLM_CXX11_RAW_STRING_LITERALS
+#	endif
+
+#	if(__has_feature(cxx_rvalue_references))
+#		define GLM_CXX11_RVALUE_REFERENCES
+#	endif
+
+#	if(__has_feature(cxx_static_assert))
+#		define GLM_CXX11_STATIC_ASSERT
+#	endif
+
+#	if(__has_feature(cxx_auto_type))
+#		define GLM_CXX11_AUTO_TYPE
+#	endif
+
+#	if(__has_feature(cxx_strong_enums))
+#		define GLM_CXX11_STRONG_ENUMS
+#	endif
+
+#	if(__has_feature(cxx_trailing_return))
+#		define GLM_CXX11_TRAILING_RETURN
+#	endif
+
+#	if(__has_feature(cxx_unicode_literals))
+#		define GLM_CXX11_UNICODE_LITERALS
+#	endif
+
+#	if(__has_feature(cxx_unrestricted_unions))
+#		define GLM_CXX11_UNRESTRICTED_UNIONS
+#	endif
+
+#	if(__has_feature(cxx_user_literals))
+#		define GLM_CXX11_USER_LITERALS
+#	endif
+
+#	if(__has_feature(cxx_variadic_templates))
+#		define GLM_CXX11_VARIADIC_TEMPLATES
+#	endif
+
+#endif//(GLM_COMPILER & GLM_COMPILER_CLANG)
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/_fixes.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/_fixes.hpp
new file mode 100644
index 000000000000..a503c7c0d041
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/_fixes.hpp
@@ -0,0 +1,27 @@
+#include <cmath>
+
+//! Workaround for compatibility with other libraries
+#ifdef max
+#undef max
+#endif
+
+//! Workaround for compatibility with other libraries
+#ifdef min
+#undef min
+#endif
+
+//! Workaround for Android
+#ifdef isnan
+#undef isnan
+#endif
+
+//! Workaround for Android
+#ifdef isinf
+#undef isinf
+#endif
+
+//! Workaround for Chrone Native Client
+#ifdef log2
+#undef log2
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/_noise.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/_noise.hpp
new file mode 100644
index 000000000000..5a874a02221f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/_noise.hpp
@@ -0,0 +1,81 @@
+#pragma once
+
+#include "../common.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER T mod289(T const& x)
+	{
+		return x - floor(x * (static_cast<T>(1.0) / static_cast<T>(289.0))) * static_cast<T>(289.0);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T permute(T const& x)
+	{
+		return mod289(((x * static_cast<T>(34)) + static_cast<T>(1)) * x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> permute(vec<2, T, Q> const& x)
+	{
+		return mod289(((x * static_cast<T>(34)) + static_cast<T>(1)) * x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> permute(vec<3, T, Q> const& x)
+	{
+		return mod289(((x * static_cast<T>(34)) + static_cast<T>(1)) * x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> permute(vec<4, T, Q> const& x)
+	{
+		return mod289(((x * static_cast<T>(34)) + static_cast<T>(1)) * x);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T taylorInvSqrt(T const& r)
+	{
+		return static_cast<T>(1.79284291400159) - static_cast<T>(0.85373472095314) * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> taylorInvSqrt(vec<2, T, Q> const& r)
+	{
+		return static_cast<T>(1.79284291400159) - static_cast<T>(0.85373472095314) * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> taylorInvSqrt(vec<3, T, Q> const& r)
+	{
+		return static_cast<T>(1.79284291400159) - static_cast<T>(0.85373472095314) * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> taylorInvSqrt(vec<4, T, Q> const& r)
+	{
+		return static_cast<T>(1.79284291400159) - static_cast<T>(0.85373472095314) * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> fade(vec<2, T, Q> const& t)
+	{
+		return (t * t * t) * (t * (t * static_cast<T>(6) - static_cast<T>(15)) + static_cast<T>(10));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> fade(vec<3, T, Q> const& t)
+	{
+		return (t * t * t) * (t * (t * static_cast<T>(6) - static_cast<T>(15)) + static_cast<T>(10));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> fade(vec<4, T, Q> const& t)
+	{
+		return (t * t * t) * (t * (t * static_cast<T>(6) - static_cast<T>(15)) + static_cast<T>(10));
+	}
+}//namespace detail
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle.hpp
new file mode 100644
index 000000000000..87896ef4f6f2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle.hpp
@@ -0,0 +1,804 @@
+#pragma once
+
+namespace glm{
+namespace detail
+{
+	// Internal class for implementing swizzle operators
+	template<typename T, int N>
+	struct _swizzle_base0
+	{
+	protected:
+		GLM_FUNC_QUALIFIER T& elem(size_t i){ return (reinterpret_cast<T*>(_buffer))[i]; }
+		GLM_FUNC_QUALIFIER T const& elem(size_t i) const{ return (reinterpret_cast<const T*>(_buffer))[i]; }
+
+		// Use an opaque buffer to *ensure* the compiler doesn't call a constructor.
+		// The size 1 buffer is assumed to aligned to the actual members so that the
+		// elem()
+		char    _buffer[1];
+	};
+
+	template<int N, typename T, qualifier Q, int E0, int E1, int E2, int E3, bool Aligned>
+	struct _swizzle_base1 : public _swizzle_base0<T, N>
+	{
+	};
+
+	template<typename T, qualifier Q, int E0, int E1, bool Aligned>
+	struct _swizzle_base1<2, T, Q, E0,E1,-1,-2, Aligned> : public _swizzle_base0<T, 2>
+	{
+		GLM_FUNC_QUALIFIER vec<2, T, Q> operator ()()  const { return vec<2, T, Q>(this->elem(E0), this->elem(E1)); }
+	};
+
+	template<typename T, qualifier Q, int E0, int E1, int E2, bool Aligned>
+	struct _swizzle_base1<3, T, Q, E0,E1,E2,-1, Aligned> : public _swizzle_base0<T, 3>
+	{
+		GLM_FUNC_QUALIFIER vec<3, T, Q> operator ()()  const { return vec<3, T, Q>(this->elem(E0), this->elem(E1), this->elem(E2)); }
+	};
+
+	template<typename T, qualifier Q, int E0, int E1, int E2, int E3, bool Aligned>
+	struct _swizzle_base1<4, T, Q, E0,E1,E2,E3, Aligned> : public _swizzle_base0<T, 4>
+	{
+		GLM_FUNC_QUALIFIER vec<4, T, Q> operator ()()  const { return vec<4, T, Q>(this->elem(E0), this->elem(E1), this->elem(E2), this->elem(E3)); }
+	};
+
+	// Internal class for implementing swizzle operators
+	/*
+		Template parameters:
+
+		T			= type of scalar values (e.g. float, double)
+		N			= number of components in the vector (e.g. 3)
+		E0...3		= what index the n-th element of this swizzle refers to in the unswizzled vec
+
+		DUPLICATE_ELEMENTS = 1 if there is a repeated element, 0 otherwise (used to specialize swizzles
+			containing duplicate elements so that they cannot be used as r-values).
+	*/
+	template<int N, typename T, qualifier Q, int E0, int E1, int E2, int E3, int DUPLICATE_ELEMENTS>
+	struct _swizzle_base2 : public _swizzle_base1<N, T, Q, E0,E1,E2,E3, detail::is_aligned<Q>::value>
+	{
+		struct op_equal
+		{
+			GLM_FUNC_QUALIFIER void operator() (T& e, T& t) const{ e = t; }
+		};
+
+		struct op_minus
+		{
+			GLM_FUNC_QUALIFIER void operator() (T& e, T& t) const{ e -= t; }
+		};
+
+		struct op_plus
+		{
+			GLM_FUNC_QUALIFIER void operator() (T& e, T& t) const{ e += t; }
+		};
+
+		struct op_mul
+		{
+			GLM_FUNC_QUALIFIER void operator() (T& e, T& t) const{ e *= t; }
+		};
+
+		struct op_div
+		{
+			GLM_FUNC_QUALIFIER void operator() (T& e, T& t) const{ e /= t; }
+		};
+
+	public:
+		GLM_FUNC_QUALIFIER _swizzle_base2& operator= (const T& t)
+		{
+			for (int i = 0; i < N; ++i)
+				(*this)[i] = t;
+			return *this;
+		}
+
+		GLM_FUNC_QUALIFIER _swizzle_base2& operator= (vec<N, T, Q> const& that)
+		{
+			_apply_op(that, op_equal());
+			return *this;
+		}
+
+		GLM_FUNC_QUALIFIER void operator -= (vec<N, T, Q> const& that)
+		{
+			_apply_op(that, op_minus());
+		}
+
+		GLM_FUNC_QUALIFIER void operator += (vec<N, T, Q> const& that)
+		{
+			_apply_op(that, op_plus());
+		}
+
+		GLM_FUNC_QUALIFIER void operator *= (vec<N, T, Q> const& that)
+		{
+			_apply_op(that, op_mul());
+		}
+
+		GLM_FUNC_QUALIFIER void operator /= (vec<N, T, Q> const& that)
+		{
+			_apply_op(that, op_div());
+		}
+
+		GLM_FUNC_QUALIFIER T& operator[](size_t i)
+		{
+			const int offset_dst[4] = { E0, E1, E2, E3 };
+			return this->elem(offset_dst[i]);
+		}
+		GLM_FUNC_QUALIFIER T operator[](size_t i) const
+		{
+			const int offset_dst[4] = { E0, E1, E2, E3 };
+			return this->elem(offset_dst[i]);
+		}
+
+	protected:
+		template<typename U>
+		GLM_FUNC_QUALIFIER void _apply_op(vec<N, T, Q> const& that, const U& op)
+		{
+			// Make a copy of the data in this == &that.
+			// The copier should optimize out the copy in cases where the function is
+			// properly inlined and the copy is not necessary.
+			T t[N];
+			for (int i = 0; i < N; ++i)
+				t[i] = that[i];
+			for (int i = 0; i < N; ++i)
+				op( (*this)[i], t[i] );
+		}
+	};
+
+	// Specialization for swizzles containing duplicate elements.  These cannot be modified.
+	template<int N, typename T, qualifier Q, int E0, int E1, int E2, int E3>
+	struct _swizzle_base2<N, T, Q, E0,E1,E2,E3, 1> : public _swizzle_base1<N, T, Q, E0,E1,E2,E3, detail::is_aligned<Q>::value>
+	{
+		struct Stub {};
+
+		GLM_FUNC_QUALIFIER _swizzle_base2& operator= (Stub const&) { return *this; }
+
+		GLM_FUNC_QUALIFIER T operator[]  (size_t i) const
+		{
+			const int offset_dst[4] = { E0, E1, E2, E3 };
+			return this->elem(offset_dst[i]);
+		}
+	};
+
+	template<int N, typename T, qualifier Q, int E0, int E1, int E2, int E3>
+	struct _swizzle : public _swizzle_base2<N, T, Q, E0, E1, E2, E3, (E0 == E1 || E0 == E2 || E0 == E3 || E1 == E2 || E1 == E3 || E2 == E3)>
+	{
+		typedef _swizzle_base2<N, T, Q, E0, E1, E2, E3, (E0 == E1 || E0 == E2 || E0 == E3 || E1 == E2 || E1 == E3 || E2 == E3)> base_type;
+
+		using base_type::operator=;
+
+		GLM_FUNC_QUALIFIER operator vec<N, T, Q> () const { return (*this)(); }
+	};
+
+//
+// To prevent the C++ syntax from getting entirely overwhelming, define some alias macros
+//
+#define GLM_SWIZZLE_TEMPLATE1   template<int N, typename T, qualifier Q, int E0, int E1, int E2, int E3>
+#define GLM_SWIZZLE_TEMPLATE2   template<int N, typename T, qualifier Q, int E0, int E1, int E2, int E3, int F0, int F1, int F2, int F3>
+#define GLM_SWIZZLE_TYPE1       _swizzle<N, T, Q, E0, E1, E2, E3>
+#define GLM_SWIZZLE_TYPE2       _swizzle<N, T, Q, F0, F1, F2, F3>
+
+//
+// Wrapper for a binary operator (e.g. u.yy + v.zy)
+//
+#define GLM_SWIZZLE_VECTOR_BINARY_OPERATOR_IMPLEMENTATION(OPERAND)                 \
+	GLM_SWIZZLE_TEMPLATE2                                                          \
+	GLM_FUNC_QUALIFIER vec<N, T, Q> operator OPERAND ( const GLM_SWIZZLE_TYPE1& a, const GLM_SWIZZLE_TYPE2& b)  \
+	{                                                                               \
+		return a() OPERAND b();                                                     \
+	}                                                                               \
+	GLM_SWIZZLE_TEMPLATE1                                                          \
+	GLM_FUNC_QUALIFIER vec<N, T, Q> operator OPERAND ( const GLM_SWIZZLE_TYPE1& a, const vec<N, T, Q>& b)                   \
+	{                                                                               \
+		return a() OPERAND b;                                                       \
+	}                                                                               \
+	GLM_SWIZZLE_TEMPLATE1                                                          \
+	GLM_FUNC_QUALIFIER vec<N, T, Q> operator OPERAND ( const vec<N, T, Q>& a, const GLM_SWIZZLE_TYPE1& b)                   \
+	{                                                                               \
+		return a OPERAND b();                                                       \
+	}
+
+//
+// Wrapper for a operand between a swizzle and a binary (e.g. 1.0f - u.xyz)
+//
+#define GLM_SWIZZLE_SCALAR_BINARY_OPERATOR_IMPLEMENTATION(OPERAND)								\
+	GLM_SWIZZLE_TEMPLATE1																		\
+	GLM_FUNC_QUALIFIER vec<N, T, Q> operator OPERAND ( const GLM_SWIZZLE_TYPE1& a, const T& b)	\
+	{																							\
+		return a() OPERAND b;																	\
+	}																							\
+	GLM_SWIZZLE_TEMPLATE1																		\
+	GLM_FUNC_QUALIFIER vec<N, T, Q> operator OPERAND ( const T& a, const GLM_SWIZZLE_TYPE1& b)	\
+	{																							\
+		return a OPERAND b();																	\
+	}
+
+//
+// Macro for wrapping a function taking one argument (e.g. abs())
+//
+#define GLM_SWIZZLE_FUNCTION_1_ARGS(RETURN_TYPE,FUNCTION)												\
+	GLM_SWIZZLE_TEMPLATE1																				\
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a)		\
+	{																									\
+		return FUNCTION(a());																			\
+	}
+
+//
+// Macro for wrapping a function taking two vector arguments (e.g. dot()).
+//
+#define GLM_SWIZZLE_FUNCTION_2_ARGS(RETURN_TYPE,FUNCTION)                                                       \
+	GLM_SWIZZLE_TEMPLATE2                                                                                       \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a, const GLM_SWIZZLE_TYPE2& b) \
+	{                                                                                                           \
+		return FUNCTION(a(), b());                                                                              \
+	}                                                                                                           \
+	GLM_SWIZZLE_TEMPLATE1                                                                                       \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a, const GLM_SWIZZLE_TYPE1& b) \
+	{                                                                                                           \
+		return FUNCTION(a(), b());                                                                              \
+	}                                                                                                           \
+	GLM_SWIZZLE_TEMPLATE1                                                                                       \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a, const typename V& b)         \
+	{                                                                                                           \
+		return FUNCTION(a(), b);                                                                                \
+	}                                                                                                           \
+	GLM_SWIZZLE_TEMPLATE1                                                                                       \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const V& a, const GLM_SWIZZLE_TYPE1& b)                  \
+	{                                                                                                           \
+		return FUNCTION(a, b());                                                                                \
+	}
+
+//
+// Macro for wrapping a function take 2 vec arguments followed by a scalar (e.g. mix()).
+//
+#define GLM_SWIZZLE_FUNCTION_2_ARGS_SCALAR(RETURN_TYPE,FUNCTION)                                                             \
+	GLM_SWIZZLE_TEMPLATE2                                                                                                    \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a, const GLM_SWIZZLE_TYPE2& b, const T& c)   \
+	{                                                                                                                         \
+		return FUNCTION(a(), b(), c);                                                                                         \
+	}                                                                                                                         \
+	GLM_SWIZZLE_TEMPLATE1                                                                                                    \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a, const GLM_SWIZZLE_TYPE1& b, const T& c)   \
+	{                                                                                                                         \
+		return FUNCTION(a(), b(), c);                                                                                         \
+	}                                                                                                                         \
+	GLM_SWIZZLE_TEMPLATE1                                                                                                    \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const GLM_SWIZZLE_TYPE1& a, const typename S0::vec_type& b, const T& c)\
+	{                                                                                                                         \
+		return FUNCTION(a(), b, c);                                                                                           \
+	}                                                                                                                         \
+	GLM_SWIZZLE_TEMPLATE1                                                                                                    \
+	GLM_FUNC_QUALIFIER typename GLM_SWIZZLE_TYPE1::RETURN_TYPE FUNCTION(const typename V& a, const GLM_SWIZZLE_TYPE1& b, const T& c)           \
+	{                                                                                                                         \
+		return FUNCTION(a, b(), c);                                                                                           \
+	}
+
+}//namespace detail
+}//namespace glm
+
+namespace glm
+{
+	namespace detail
+	{
+		GLM_SWIZZLE_SCALAR_BINARY_OPERATOR_IMPLEMENTATION(-)
+		GLM_SWIZZLE_SCALAR_BINARY_OPERATOR_IMPLEMENTATION(*)
+		GLM_SWIZZLE_VECTOR_BINARY_OPERATOR_IMPLEMENTATION(+)
+		GLM_SWIZZLE_VECTOR_BINARY_OPERATOR_IMPLEMENTATION(-)
+		GLM_SWIZZLE_VECTOR_BINARY_OPERATOR_IMPLEMENTATION(*)
+		GLM_SWIZZLE_VECTOR_BINARY_OPERATOR_IMPLEMENTATION(/)
+	}
+
+	//
+	// Swizzles are distinct types from the unswizzled type.  The below macros will
+	// provide template specializations for the swizzle types for the given functions
+	// so that the compiler does not have any ambiguity to choosing how to handle
+	// the function.
+	//
+	// The alternative is to use the operator()() when calling the function in order
+	// to explicitly convert the swizzled type to the unswizzled type.
+	//
+
+	//GLM_SWIZZLE_FUNCTION_1_ARGS(vec_type,    abs);
+	//GLM_SWIZZLE_FUNCTION_1_ARGS(vec_type,    acos);
+	//GLM_SWIZZLE_FUNCTION_1_ARGS(vec_type,    acosh);
+	//GLM_SWIZZLE_FUNCTION_1_ARGS(vec_type,    all);
+	//GLM_SWIZZLE_FUNCTION_1_ARGS(vec_type,    any);
+
+	//GLM_SWIZZLE_FUNCTION_2_ARGS(value_type,  dot);
+	//GLM_SWIZZLE_FUNCTION_2_ARGS(vec_type,    cross);
+	//GLM_SWIZZLE_FUNCTION_2_ARGS(vec_type,    step);
+	//GLM_SWIZZLE_FUNCTION_2_ARGS_SCALAR(vec_type, mix);
+}
+
+#define GLM_SWIZZLE2_2_MEMBERS(T, Q, E0,E1) \
+	struct { detail::_swizzle<2, T, Q, 0,0,-1,-2> E0 ## E0; }; \
+	struct { detail::_swizzle<2, T, Q, 0,1,-1,-2> E0 ## E1; }; \
+	struct { detail::_swizzle<2, T, Q, 1,0,-1,-2> E1 ## E0; }; \
+	struct { detail::_swizzle<2, T, Q, 1,1,-1,-2> E1 ## E1; };
+
+#define GLM_SWIZZLE2_3_MEMBERS(T, Q, E0,E1) \
+	struct { detail::_swizzle<3,T, Q, 0,0,0,-1> E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3,T, Q, 0,0,1,-1> E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3,T, Q, 0,1,0,-1> E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3,T, Q, 0,1,1,-1> E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3,T, Q, 1,0,0,-1> E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3,T, Q, 1,0,1,-1> E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3,T, Q, 1,1,0,-1> E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3,T, Q, 1,1,1,-1> E1 ## E1 ## E1; };
+
+#define GLM_SWIZZLE2_4_MEMBERS(T, Q, E0,E1) \
+	struct { detail::_swizzle<4,T, Q, 0,0,0,0> E0 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,0,1> E0 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,1,0> E0 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,1,1> E0 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,0,0> E0 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,0,1> E0 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,1,0> E0 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,1,1> E0 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,0,0> E1 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,0,1> E1 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,1,0> E1 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,1,1> E1 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,0,0> E1 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,0,1> E1 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,1,0> E1 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,1,1> E1 ## E1 ## E1 ## E1; };
+
+#define GLM_SWIZZLE3_2_MEMBERS(T, Q, E0,E1,E2) \
+	struct { detail::_swizzle<2,T, Q, 0,0,-1,-2> E0 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 0,1,-1,-2> E0 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 0,2,-1,-2> E0 ## E2; }; \
+	struct { detail::_swizzle<2,T, Q, 1,0,-1,-2> E1 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 1,1,-1,-2> E1 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 1,2,-1,-2> E1 ## E2; }; \
+	struct { detail::_swizzle<2,T, Q, 2,0,-1,-2> E2 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 2,1,-1,-2> E2 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 2,2,-1,-2> E2 ## E2; };
+
+#define GLM_SWIZZLE3_3_MEMBERS(T, Q ,E0,E1,E2) \
+	struct { detail::_swizzle<3, T, Q, 0,0,0,-1> E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,0,1,-1> E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,0,2,-1> E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,0,-1> E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,1,-1> E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,2,-1> E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,0,-1> E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,1,-1> E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,2,-1> E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,0,-1> E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,1,-1> E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,2,-1> E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,0,-1> E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,1,-1> E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,2,-1> E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,0,-1> E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,1,-1> E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,2,-1> E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,0,-1> E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,1,-1> E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,2,-1> E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,0,-1> E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,1,-1> E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,2,-1> E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,0,-1> E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,1,-1> E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,2,-1> E2 ## E2 ## E2; };
+
+#define GLM_SWIZZLE3_4_MEMBERS(T, Q, E0,E1,E2) \
+	struct { detail::_swizzle<4,T, Q, 0,0,0,0> E0 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,0,1> E0 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,0,2> E0 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,1,0> E0 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,1,1> E0 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,1,2> E0 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,2,0> E0 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,2,1> E0 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,0,2,2> E0 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,0,0> E0 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,0,1> E0 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,0,2> E0 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,1,0> E0 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,1,1> E0 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,1,2> E0 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,2,0> E0 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,2,1> E0 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,1,2,2> E0 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,0,0> E0 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,0,1> E0 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,0,2> E0 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,1,0> E0 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,1,1> E0 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,1,2> E0 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,2,0> E0 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,2,1> E0 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 0,2,2,2> E0 ## E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,0,0> E1 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,0,1> E1 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,0,2> E1 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,1,0> E1 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,1,1> E1 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,1,2> E1 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,2,0> E1 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,2,1> E1 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,0,2,2> E1 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,0,0> E1 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,0,1> E1 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,0,2> E1 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,1,0> E1 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,1,1> E1 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,1,2> E1 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,2,0> E1 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,2,1> E1 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,1,2,2> E1 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,0,0> E1 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,0,1> E1 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,0,2> E1 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,1,0> E1 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,1,1> E1 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,1,2> E1 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,2,0> E1 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,2,1> E1 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 1,2,2,2> E1 ## E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,0,0> E2 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,0,1> E2 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,0,2> E2 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,1,0> E2 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,1,1> E2 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,1,2> E2 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,2,0> E2 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,2,1> E2 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,0,2,2> E2 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,0,0> E2 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,0,1> E2 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,0,2> E2 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,1,0> E2 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,1,1> E2 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,1,2> E2 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,2,0> E2 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,2,1> E2 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,1,2,2> E2 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,0,0> E2 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,0,1> E2 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,0,2> E2 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,1,0> E2 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,1,1> E2 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,1,2> E2 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,2,0> E2 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,2,1> E2 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4,T, Q, 2,2,2,2> E2 ## E2 ## E2 ## E2; };
+
+#define GLM_SWIZZLE4_2_MEMBERS(T, Q, E0,E1,E2,E3) \
+	struct { detail::_swizzle<2,T, Q, 0,0,-1,-2> E0 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 0,1,-1,-2> E0 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 0,2,-1,-2> E0 ## E2; }; \
+	struct { detail::_swizzle<2,T, Q, 0,3,-1,-2> E0 ## E3; }; \
+	struct { detail::_swizzle<2,T, Q, 1,0,-1,-2> E1 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 1,1,-1,-2> E1 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 1,2,-1,-2> E1 ## E2; }; \
+	struct { detail::_swizzle<2,T, Q, 1,3,-1,-2> E1 ## E3; }; \
+	struct { detail::_swizzle<2,T, Q, 2,0,-1,-2> E2 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 2,1,-1,-2> E2 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 2,2,-1,-2> E2 ## E2; }; \
+	struct { detail::_swizzle<2,T, Q, 2,3,-1,-2> E2 ## E3; }; \
+	struct { detail::_swizzle<2,T, Q, 3,0,-1,-2> E3 ## E0; }; \
+	struct { detail::_swizzle<2,T, Q, 3,1,-1,-2> E3 ## E1; }; \
+	struct { detail::_swizzle<2,T, Q, 3,2,-1,-2> E3 ## E2; }; \
+	struct { detail::_swizzle<2,T, Q, 3,3,-1,-2> E3 ## E3; };
+
+#define GLM_SWIZZLE4_3_MEMBERS(T, Q, E0,E1,E2,E3) \
+	struct { detail::_swizzle<3, T, Q, 0,0,0,-1> E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,0,1,-1> E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,0,2,-1> E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 0,0,3,-1> E0 ## E0 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,0,-1> E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,1,-1> E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,2,-1> E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 0,1,3,-1> E0 ## E1 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,0,-1> E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,1,-1> E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,2,-1> E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 0,2,3,-1> E0 ## E2 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 0,3,0,-1> E0 ## E3 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 0,3,1,-1> E0 ## E3 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 0,3,2,-1> E0 ## E3 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 0,3,3,-1> E0 ## E3 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,0,-1> E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,1,-1> E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,2,-1> E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,0,3,-1> E1 ## E0 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,0,-1> E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,1,-1> E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,2,-1> E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,1,3,-1> E1 ## E1 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,0,-1> E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,1,-1> E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,2,-1> E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,2,3,-1> E1 ## E2 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 1,3,0,-1> E1 ## E3 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 1,3,1,-1> E1 ## E3 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 1,3,2,-1> E1 ## E3 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 1,3,3,-1> E1 ## E3 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,0,-1> E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,1,-1> E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,2,-1> E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,0,3,-1> E2 ## E0 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,0,-1> E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,1,-1> E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,2,-1> E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,1,3,-1> E2 ## E1 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,0,-1> E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,1,-1> E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,2,-1> E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,2,3,-1> E2 ## E2 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 2,3,0,-1> E2 ## E3 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 2,3,1,-1> E2 ## E3 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 2,3,2,-1> E2 ## E3 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 2,3,3,-1> E2 ## E3 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 3,0,0,-1> E3 ## E0 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 3,0,1,-1> E3 ## E0 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 3,0,2,-1> E3 ## E0 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 3,0,3,-1> E3 ## E0 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 3,1,0,-1> E3 ## E1 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 3,1,1,-1> E3 ## E1 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 3,1,2,-1> E3 ## E1 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 3,1,3,-1> E3 ## E1 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 3,2,0,-1> E3 ## E2 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 3,2,1,-1> E3 ## E2 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 3,2,2,-1> E3 ## E2 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 3,2,3,-1> E3 ## E2 ## E3; }; \
+	struct { detail::_swizzle<3, T, Q, 3,3,0,-1> E3 ## E3 ## E0; }; \
+	struct { detail::_swizzle<3, T, Q, 3,3,1,-1> E3 ## E3 ## E1; }; \
+	struct { detail::_swizzle<3, T, Q, 3,3,2,-1> E3 ## E3 ## E2; }; \
+	struct { detail::_swizzle<3, T, Q, 3,3,3,-1> E3 ## E3 ## E3; };
+
+#define GLM_SWIZZLE4_4_MEMBERS(T, Q, E0,E1,E2,E3) \
+	struct { detail::_swizzle<4, T, Q, 0,0,0,0> E0 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,0,1> E0 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,0,2> E0 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,0,3> E0 ## E0 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,1,0> E0 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,1,1> E0 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,1,2> E0 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,1,3> E0 ## E0 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,2,0> E0 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,2,1> E0 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,2,2> E0 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,2,3> E0 ## E0 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,3,0> E0 ## E0 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,3,1> E0 ## E0 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,3,2> E0 ## E0 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,0,3,3> E0 ## E0 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,0,0> E0 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,0,1> E0 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,0,2> E0 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,0,3> E0 ## E1 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,1,0> E0 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,1,1> E0 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,1,2> E0 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,1,3> E0 ## E1 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,2,0> E0 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,2,1> E0 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,2,2> E0 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,2,3> E0 ## E1 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,3,0> E0 ## E1 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,3,1> E0 ## E1 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,3,2> E0 ## E1 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,1,3,3> E0 ## E1 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,0,0> E0 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,0,1> E0 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,0,2> E0 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,0,3> E0 ## E2 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,1,0> E0 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,1,1> E0 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,1,2> E0 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,1,3> E0 ## E2 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,2,0> E0 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,2,1> E0 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,2,2> E0 ## E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,2,3> E0 ## E2 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,3,0> E0 ## E2 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,3,1> E0 ## E2 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,3,2> E0 ## E2 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,2,3,3> E0 ## E2 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,0,0> E0 ## E3 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,0,1> E0 ## E3 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,0,2> E0 ## E3 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,0,3> E0 ## E3 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,1,0> E0 ## E3 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,1,1> E0 ## E3 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,1,2> E0 ## E3 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,1,3> E0 ## E3 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,2,0> E0 ## E3 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,2,1> E0 ## E3 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,2,2> E0 ## E3 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,2,3> E0 ## E3 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,3,0> E0 ## E3 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,3,1> E0 ## E3 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,3,2> E0 ## E3 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 0,3,3,3> E0 ## E3 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,0,0> E1 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,0,1> E1 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,0,2> E1 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,0,3> E1 ## E0 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,1,0> E1 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,1,1> E1 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,1,2> E1 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,1,3> E1 ## E0 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,2,0> E1 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,2,1> E1 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,2,2> E1 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,2,3> E1 ## E0 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,3,0> E1 ## E0 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,3,1> E1 ## E0 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,3,2> E1 ## E0 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,0,3,3> E1 ## E0 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,0,0> E1 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,0,1> E1 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,0,2> E1 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,0,3> E1 ## E1 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,1,0> E1 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,1,1> E1 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,1,2> E1 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,1,3> E1 ## E1 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,2,0> E1 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,2,1> E1 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,2,2> E1 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,2,3> E1 ## E1 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,3,0> E1 ## E1 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,3,1> E1 ## E1 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,3,2> E1 ## E1 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,1,3,3> E1 ## E1 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,0,0> E1 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,0,1> E1 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,0,2> E1 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,0,3> E1 ## E2 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,1,0> E1 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,1,1> E1 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,1,2> E1 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,1,3> E1 ## E2 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,2,0> E1 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,2,1> E1 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,2,2> E1 ## E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,2,3> E1 ## E2 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,3,0> E1 ## E2 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,3,1> E1 ## E2 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,3,2> E1 ## E2 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,2,3,3> E1 ## E2 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,0,0> E1 ## E3 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,0,1> E1 ## E3 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,0,2> E1 ## E3 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,0,3> E1 ## E3 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,1,0> E1 ## E3 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,1,1> E1 ## E3 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,1,2> E1 ## E3 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,1,3> E1 ## E3 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,2,0> E1 ## E3 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,2,1> E1 ## E3 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,2,2> E1 ## E3 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,2,3> E1 ## E3 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,3,0> E1 ## E3 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,3,1> E1 ## E3 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,3,2> E1 ## E3 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 1,3,3,3> E1 ## E3 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,0,0> E2 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,0,1> E2 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,0,2> E2 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,0,3> E2 ## E0 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,1,0> E2 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,1,1> E2 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,1,2> E2 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,1,3> E2 ## E0 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,2,0> E2 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,2,1> E2 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,2,2> E2 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,2,3> E2 ## E0 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,3,0> E2 ## E0 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,3,1> E2 ## E0 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,3,2> E2 ## E0 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,0,3,3> E2 ## E0 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,0,0> E2 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,0,1> E2 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,0,2> E2 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,0,3> E2 ## E1 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,1,0> E2 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,1,1> E2 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,1,2> E2 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,1,3> E2 ## E1 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,2,0> E2 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,2,1> E2 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,2,2> E2 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,2,3> E2 ## E1 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,3,0> E2 ## E1 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,3,1> E2 ## E1 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,3,2> E2 ## E1 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,1,3,3> E2 ## E1 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,0,0> E2 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,0,1> E2 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,0,2> E2 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,0,3> E2 ## E2 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,1,0> E2 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,1,1> E2 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,1,2> E2 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,1,3> E2 ## E2 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,2,0> E2 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,2,1> E2 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,2,2> E2 ## E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,2,3> E2 ## E2 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,3,0> E2 ## E2 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,3,1> E2 ## E2 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,3,2> E2 ## E2 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,2,3,3> E2 ## E2 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,0,0> E2 ## E3 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,0,1> E2 ## E3 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,0,2> E2 ## E3 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,0,3> E2 ## E3 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,1,0> E2 ## E3 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,1,1> E2 ## E3 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,1,2> E2 ## E3 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,1,3> E2 ## E3 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,2,0> E2 ## E3 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,2,1> E2 ## E3 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,2,2> E2 ## E3 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,2,3> E2 ## E3 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,3,0> E2 ## E3 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,3,1> E2 ## E3 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,3,2> E2 ## E3 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 2,3,3,3> E2 ## E3 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,0,0> E3 ## E0 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,0,1> E3 ## E0 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,0,2> E3 ## E0 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,0,3> E3 ## E0 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,1,0> E3 ## E0 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,1,1> E3 ## E0 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,1,2> E3 ## E0 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,1,3> E3 ## E0 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,2,0> E3 ## E0 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,2,1> E3 ## E0 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,2,2> E3 ## E0 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,2,3> E3 ## E0 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,3,0> E3 ## E0 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,3,1> E3 ## E0 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,3,2> E3 ## E0 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,0,3,3> E3 ## E0 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,0,0> E3 ## E1 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,0,1> E3 ## E1 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,0,2> E3 ## E1 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,0,3> E3 ## E1 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,1,0> E3 ## E1 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,1,1> E3 ## E1 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,1,2> E3 ## E1 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,1,3> E3 ## E1 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,2,0> E3 ## E1 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,2,1> E3 ## E1 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,2,2> E3 ## E1 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,2,3> E3 ## E1 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,3,0> E3 ## E1 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,3,1> E3 ## E1 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,3,2> E3 ## E1 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,1,3,3> E3 ## E1 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,0,0> E3 ## E2 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,0,1> E3 ## E2 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,0,2> E3 ## E2 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,0,3> E3 ## E2 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,1,0> E3 ## E2 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,1,1> E3 ## E2 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,1,2> E3 ## E2 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,1,3> E3 ## E2 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,2,0> E3 ## E2 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,2,1> E3 ## E2 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,2,2> E3 ## E2 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,2,3> E3 ## E2 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,3,0> E3 ## E2 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,3,1> E3 ## E2 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,3,2> E3 ## E2 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,2,3,3> E3 ## E2 ## E3 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,0,0> E3 ## E3 ## E0 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,0,1> E3 ## E3 ## E0 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,0,2> E3 ## E3 ## E0 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,0,3> E3 ## E3 ## E0 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,1,0> E3 ## E3 ## E1 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,1,1> E3 ## E3 ## E1 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,1,2> E3 ## E3 ## E1 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,1,3> E3 ## E3 ## E1 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,2,0> E3 ## E3 ## E2 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,2,1> E3 ## E3 ## E2 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,2,2> E3 ## E3 ## E2 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,2,3> E3 ## E3 ## E2 ## E3; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,3,0> E3 ## E3 ## E3 ## E0; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,3,1> E3 ## E3 ## E3 ## E1; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,3,2> E3 ## E3 ## E3 ## E2; }; \
+	struct { detail::_swizzle<4, T, Q, 3,3,3,3> E3 ## E3 ## E3 ## E3; };
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle_func.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle_func.hpp
new file mode 100644
index 000000000000..a264ae9eff3a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/_swizzle_func.hpp
@@ -0,0 +1,682 @@
+#pragma once
+
+#define GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, CONST, A, B)	\
+	GLM_FUNC_QUALIFIER vec<2, T, Q> A ## B() CONST							\
+	{													\
+		return vec<2, T, Q>(this->A, this->B);			\
+	}
+
+#define GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, CONST, A, B, C)		\
+	GLM_FUNC_QUALIFIER vec<3, T, Q> A ## B ## C() CONST							\
+	{															\
+		return vec<3, T, Q>(this->A, this->B, this->C);			\
+	}
+
+#define GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, CONST, A, B, C, D)					\
+	GLM_FUNC_QUALIFIER vec<4, T, Q> A ## B ## C ## D() CONST									\
+	{																		\
+		return vec<4, T, Q>(this->A, this->B, this->C, this->D);			\
+	}
+
+#define GLM_SWIZZLE_GEN_VEC2_ENTRY_DEF(T, P, L, CONST, A, B)	\
+	template<typename T>										\
+	GLM_FUNC_QUALIFIER vec<L, T, Q> vec<L, T, Q>::A ## B() CONST					\
+	{															\
+		return vec<2, T, Q>(this->A, this->B);					\
+	}
+
+#define GLM_SWIZZLE_GEN_VEC3_ENTRY_DEF(T, P, L, CONST, A, B, C)		\
+	template<typename T>											\
+	GLM_FUNC_QUALIFIER vec<3, T, Q> vec<L, T, Q>::A ## B ## C() CONST					\
+	{																\
+		return vec<3, T, Q>(this->A, this->B, this->C);				\
+	}
+
+#define GLM_SWIZZLE_GEN_VEC4_ENTRY_DEF(T, P, L, CONST, A, B, C, D)		\
+	template<typename T>												\
+	GLM_FUNC_QUALIFIER vec<4, T, Q> vec<L, T, Q>::A ## B ## C ## D() CONST					\
+	{																	\
+		return vec<4, T, Q>(this->A, this->B, this->C, this->D);		\
+	}
+
+#define GLM_MUTABLE
+
+#define GLM_SWIZZLE_GEN_REF2_FROM_VEC2_SWIZZLE(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, 2, GLM_MUTABLE, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, 2, GLM_MUTABLE, B, A)
+
+#define GLM_SWIZZLE_GEN_REF_FROM_VEC2(T, P) \
+	GLM_SWIZZLE_GEN_REF2_FROM_VEC2_SWIZZLE(T, P, x, y) \
+	GLM_SWIZZLE_GEN_REF2_FROM_VEC2_SWIZZLE(T, P, r, g) \
+	GLM_SWIZZLE_GEN_REF2_FROM_VEC2_SWIZZLE(T, P, s, t)
+
+#define GLM_SWIZZLE_GEN_REF2_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, A, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, B, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, C, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, C, B)
+
+#define GLM_SWIZZLE_GEN_REF3_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, GLM_MUTABLE, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, GLM_MUTABLE, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, GLM_MUTABLE, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, GLM_MUTABLE, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, GLM_MUTABLE, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, GLM_MUTABLE, C, B, A)
+
+#define GLM_SWIZZLE_GEN_REF_FROM_VEC3_COMP(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_REF3_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_REF2_FROM_VEC3_SWIZZLE(T, P, A, B, C)
+
+#define GLM_SWIZZLE_GEN_REF_FROM_VEC3(T, P) \
+	GLM_SWIZZLE_GEN_REF_FROM_VEC3_COMP(T, P, x, y, z) \
+	GLM_SWIZZLE_GEN_REF_FROM_VEC3_COMP(T, P, r, g, b) \
+	GLM_SWIZZLE_GEN_REF_FROM_VEC3_COMP(T, P, s, t, p)
+
+#define GLM_SWIZZLE_GEN_REF2_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, A, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, A, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, B, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, B, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, C, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, C, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, C, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, D, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, D, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, GLM_MUTABLE, D, C)
+
+#define GLM_SWIZZLE_GEN_REF3_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , A, B, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , A, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , A, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , A, D, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , A, D, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , B, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , B, A, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , B, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , B, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , B, D, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , B, D, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , C, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , C, A, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , C, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , C, B, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , C, D, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , C, D, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , D, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , D, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , D, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , D, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , D, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, , D, C, B)
+
+#define GLM_SWIZZLE_GEN_REF4_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , A, C, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , A, C, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , A, D, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , A, D, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , A, B, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , B, C, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , B, C, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , B, D, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , B, D, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , B, A, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , B, A, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , C, B, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , C, B, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , C, D, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , C, D, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , C, A, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , C, A, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , D, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , D, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , D, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , D, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , D, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, , D, B, C, A)
+
+#define GLM_SWIZZLE_GEN_REF_FROM_VEC4_COMP(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_REF2_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_REF3_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_REF4_FROM_VEC4_SWIZZLE(T, P, A, B, C, D)
+
+#define GLM_SWIZZLE_GEN_REF_FROM_VEC4(T, P) \
+	GLM_SWIZZLE_GEN_REF_FROM_VEC4_COMP(T, P, x, y, z, w) \
+	GLM_SWIZZLE_GEN_REF_FROM_VEC4_COMP(T, P, r, g, b, a) \
+	GLM_SWIZZLE_GEN_REF_FROM_VEC4_COMP(T, P, s, t, p, q)
+
+#define GLM_SWIZZLE_GEN_VEC2_FROM_VEC2_SWIZZLE(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, B)
+
+#define GLM_SWIZZLE_GEN_VEC3_FROM_VEC2_SWIZZLE(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, B)
+
+#define GLM_SWIZZLE_GEN_VEC4_FROM_VEC2_SWIZZLE(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, B)
+
+#define GLM_SWIZZLE_GEN_VEC_FROM_VEC2_COMP(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_FROM_VEC2_SWIZZLE(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_FROM_VEC2_SWIZZLE(T, P, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_FROM_VEC2_SWIZZLE(T, P, A, B)
+
+#define GLM_SWIZZLE_GEN_VEC_FROM_VEC2(T, P)			\
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC2_COMP(T, P, x, y)	\
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC2_COMP(T, P, r, g)	\
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC2_COMP(T, P, s, t)
+
+#define GLM_SWIZZLE_GEN_VEC2_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, C)
+
+#define GLM_SWIZZLE_GEN_VEC3_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, C)
+
+#define GLM_SWIZZLE_GEN_VEC4_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, C)
+
+#define GLM_SWIZZLE_GEN_VEC_FROM_VEC3_COMP(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_FROM_VEC3_SWIZZLE(T, P, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_FROM_VEC3_SWIZZLE(T, P, A, B, C)
+
+#define GLM_SWIZZLE_GEN_VEC_FROM_VEC3(T, P) \
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC3_COMP(T, P, x, y, z) \
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC3_COMP(T, P, r, g, b) \
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC3_COMP(T, P, s, t, p)
+
+#define GLM_SWIZZLE_GEN_VEC2_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, A, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, B, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, C, D) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, D, A) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, D, B) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, D, C) \
+	GLM_SWIZZLE_GEN_VEC2_ENTRY(T, P, const, D, D)
+
+#define GLM_SWIZZLE_GEN_VEC3_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, A, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, B, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, D, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, D, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, D, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, A, D, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, A, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, B, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, D, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, D, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, D, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, B, D, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, A, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, B, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, D, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, D, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, D, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, C, D, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, A, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, A, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, A, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, A, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, B, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, B, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, B, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, B, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, C, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, C, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, C, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, D, A) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, D, B) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, D, C) \
+	GLM_SWIZZLE_GEN_VEC3_ENTRY(T, P, const, D, D, D)
+
+#define GLM_SWIZZLE_GEN_VEC4_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, A, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, B, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, C, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, A, D, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, A, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, B, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, C, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, B, D, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, A, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, B, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, C, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, C, D, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, A, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, B, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, C, D, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, A, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, A, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, A, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, A, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, B, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, B, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, B, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, B, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, C, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, C, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, C, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, D, A) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, D, B) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, D, C) \
+	GLM_SWIZZLE_GEN_VEC4_ENTRY(T, P, const, D, D, D, D)
+
+#define GLM_SWIZZLE_GEN_VEC_FROM_VEC4_COMP(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC2_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC3_FROM_VEC4_SWIZZLE(T, P, A, B, C, D) \
+	GLM_SWIZZLE_GEN_VEC4_FROM_VEC4_SWIZZLE(T, P, A, B, C, D)
+
+#define GLM_SWIZZLE_GEN_VEC_FROM_VEC4(T, P) \
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC4_COMP(T, P, x, y, z, w) \
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC4_COMP(T, P, r, g, b, a) \
+	GLM_SWIZZLE_GEN_VEC_FROM_VEC4_COMP(T, P, s, t, p, q)
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/_vectorize.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/_vectorize.hpp
new file mode 100644
index 000000000000..1fcaec315284
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/_vectorize.hpp
@@ -0,0 +1,162 @@
+#pragma once
+
+namespace glm{
+namespace detail
+{
+	template<template<length_t L, typename T, qualifier Q> class vec, length_t L, typename R, typename T, qualifier Q>
+	struct functor1{};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename R, typename T, qualifier Q>
+	struct functor1<vec, 1, R, T, Q>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<1, R, Q> call(R (*Func) (T x), vec<1, T, Q> const& v)
+		{
+			return vec<1, R, Q>(Func(v.x));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename R, typename T, qualifier Q>
+	struct functor1<vec, 2, R, T, Q>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<2, R, Q> call(R (*Func) (T x), vec<2, T, Q> const& v)
+		{
+			return vec<2, R, Q>(Func(v.x), Func(v.y));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename R, typename T, qualifier Q>
+	struct functor1<vec, 3, R, T, Q>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<3, R, Q> call(R (*Func) (T x), vec<3, T, Q> const& v)
+		{
+			return vec<3, R, Q>(Func(v.x), Func(v.y), Func(v.z));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename R, typename T, qualifier Q>
+	struct functor1<vec, 4, R, T, Q>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, R, Q> call(R (*Func) (T x), vec<4, T, Q> const& v)
+		{
+			return vec<4, R, Q>(Func(v.x), Func(v.y), Func(v.z), Func(v.w));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, length_t L, typename T, qualifier Q>
+	struct functor2{};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2<vec, 1, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<1, T, Q> call(T (*Func) (T x, T y), vec<1, T, Q> const& a, vec<1, T, Q> const& b)
+		{
+			return vec<1, T, Q>(Func(a.x, b.x));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2<vec, 2, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<2, T, Q> call(T (*Func) (T x, T y), vec<2, T, Q> const& a, vec<2, T, Q> const& b)
+		{
+			return vec<2, T, Q>(Func(a.x, b.x), Func(a.y, b.y));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2<vec, 3, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, T, Q> call(T (*Func) (T x, T y), vec<3, T, Q> const& a, vec<3, T, Q> const& b)
+		{
+			return vec<3, T, Q>(Func(a.x, b.x), Func(a.y, b.y), Func(a.z, b.z));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2<vec, 4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, T, Q> call(T (*Func) (T x, T y), vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(Func(a.x, b.x), Func(a.y, b.y), Func(a.z, b.z), Func(a.w, b.w));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, length_t L, typename T, qualifier Q>
+	struct functor2_vec_sca{};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2_vec_sca<vec, 1, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<1, T, Q> call(T (*Func) (T x, T y), vec<1, T, Q> const& a, T b)
+		{
+			return vec<1, T, Q>(Func(a.x, b));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2_vec_sca<vec, 2, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<2, T, Q> call(T (*Func) (T x, T y), vec<2, T, Q> const& a, T b)
+		{
+			return vec<2, T, Q>(Func(a.x, b), Func(a.y, b));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2_vec_sca<vec, 3, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, T, Q> call(T (*Func) (T x, T y), vec<3, T, Q> const& a, T b)
+		{
+			return vec<3, T, Q>(Func(a.x, b), Func(a.y, b), Func(a.z, b));
+		}
+	};
+
+	template<template<length_t L, typename T, qualifier Q> class vec, typename T, qualifier Q>
+	struct functor2_vec_sca<vec, 4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, T, Q> call(T (*Func) (T x, T y), vec<4, T, Q> const& a, T b)
+		{
+			return vec<4, T, Q>(Func(a.x, b), Func(a.y, b), Func(a.z, b), Func(a.w, b));
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct functor2_vec_int {};
+
+	template<typename T, qualifier Q>
+	struct functor2_vec_int<1, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<1, int, Q> call(int (*Func) (T x, int y), vec<1, T, Q> const& a, vec<1, int, Q> const& b)
+		{
+			return vec<1, int, Q>(Func(a.x, b.x));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct functor2_vec_int<2, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<2, int, Q> call(int (*Func) (T x, int y), vec<2, T, Q> const& a, vec<2, int, Q> const& b)
+		{
+			return vec<2, int, Q>(Func(a.x, b.x), Func(a.y, b.y));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct functor2_vec_int<3, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, int, Q> call(int (*Func) (T x, int y), vec<3, T, Q> const& a, vec<3, int, Q> const& b)
+		{
+			return vec<3, int, Q>(Func(a.x, b.x), Func(a.y, b.y), Func(a.z, b.z));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct functor2_vec_int<4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, int, Q> call(int (*Func) (T x, int y), vec<4, T, Q> const& a, vec<4, int, Q> const& b)
+		{
+			return vec<4, int, Q>(Func(a.x, b.x), Func(a.y, b.y), Func(a.z, b.z), Func(a.w, b.w));
+		}
+	};
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/compute_common.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/compute_common.hpp
new file mode 100644
index 000000000000..83362bc7b388
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/compute_common.hpp
@@ -0,0 +1,50 @@
+#pragma once
+
+#include "setup.hpp"
+#include <limits>
+
+namespace glm{
+namespace detail
+{
+	template<typename genFIType, bool /*signed*/>
+	struct compute_abs
+	{};
+
+	template<typename genFIType>
+	struct compute_abs<genFIType, true>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static genFIType call(genFIType x)
+		{
+			GLM_STATIC_ASSERT(
+				std::numeric_limits<genFIType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<genFIType>::is_signed,
+				"'abs' only accept floating-point and integer scalar or vector inputs");
+
+			return x >= genFIType(0) ? x : -x;
+			// TODO, perf comp with: *(((int *) &x) + 1) &= 0x7fffffff;
+		}
+	};
+
+#if (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+	template<>
+	struct compute_abs<float, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(float x)
+		{
+			return fabsf(x);
+		}
+	};
+#endif
+
+	template<typename genFIType>
+	struct compute_abs<genFIType, false>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static genFIType call(genFIType x)
+		{
+			GLM_STATIC_ASSERT(
+				(!std::numeric_limits<genFIType>::is_signed && std::numeric_limits<genFIType>::is_integer),
+				"'abs' only accept floating-point and integer scalar or vector inputs");
+			return x;
+		}
+	};
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/compute_vector_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/compute_vector_relational.hpp
new file mode 100644
index 000000000000..167b6345dd39
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/compute_vector_relational.hpp
@@ -0,0 +1,30 @@
+#pragma once
+
+//#include "compute_common.hpp"
+#include "setup.hpp"
+#include <limits>
+
+namespace glm{
+namespace detail
+{
+	template <typename T, bool isFloat>
+	struct compute_equal
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static bool call(T a, T b)
+		{
+			return a == b;
+		}
+	};
+/*
+	template <typename T>
+	struct compute_equal<T, true>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static bool call(T a, T b)
+		{
+			return detail::compute_abs<T, std::numeric_limits<T>::is_signed>::call(b - a) <= static_cast<T>(0);
+			//return std::memcmp(&a, &b, sizeof(T)) == 0;
+		}
+	};
+*/
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_common.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_common.inl
new file mode 100644
index 000000000000..c90ba227ae65
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_common.inl
@@ -0,0 +1,792 @@
+/// @ref core
+/// @file glm/detail/func_common.inl
+
+#include "../vector_relational.hpp"
+#include "compute_common.hpp"
+#include "type_vec1.hpp"
+#include "type_vec2.hpp"
+#include "type_vec3.hpp"
+#include "type_vec4.hpp"
+#include "_vectorize.hpp"
+#include <limits>
+
+namespace glm
+{
+	// min
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType min(genType x, genType y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<genType>::is_integer, "'min' only accept floating-point or integer inputs");
+		return (y < x) ? y : x;
+	}
+
+	// max
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType max(genType x, genType y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<genType>::is_integer, "'max' only accept floating-point or integer inputs");
+
+		return (x < y) ? y : x;
+	}
+
+	// abs
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR int abs(int x)
+	{
+		int const y = x >> (sizeof(int) * 8 - 1);
+		return (x ^ y) - y;
+	}
+
+	// round
+#	if GLM_HAS_CXX11_STL
+		using ::std::round;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER genType round(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'round' only accept floating-point inputs");
+
+			return x < static_cast<genType>(0) ? static_cast<genType>(int(x - static_cast<genType>(0.5))) : static_cast<genType>(int(x + static_cast<genType>(0.5)));
+		}
+#	endif
+
+	// trunc
+#	if GLM_HAS_CXX11_STL
+		using ::std::trunc;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER genType trunc(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'trunc' only accept floating-point inputs");
+
+			return x < static_cast<genType>(0) ? -std::floor(-x) : std::floor(x);
+		}
+#	endif
+
+}//namespace glm
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_abs_vector
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return detail::functor1<vec, L, T, T, Q>::call(abs, x);
+		}
+	};
+
+	template<length_t L, typename T, typename U, qualifier Q, bool Aligned>
+	struct compute_mix_vector
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, U, Q> const& a)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<U>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || GLM_CONFIG_UNRESTRICTED_GENTYPE, "'mix' only accept floating-point inputs for the interpolator a");
+
+			return vec<L, T, Q>(vec<L, U, Q>(x) * (static_cast<U>(1) - a) + vec<L, U, Q>(y) * a);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_mix_vector<L, T, bool, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, bool, Q> const& a)
+		{
+			vec<L, T, Q> Result;
+			for(length_t i = 0; i < x.length(); ++i)
+				Result[i] = a[i] ? y[i] : x[i];
+			return Result;
+		}
+	};
+
+	template<length_t L, typename T, typename U, qualifier Q, bool Aligned>
+	struct compute_mix_scalar
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, U const& a)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<U>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || GLM_CONFIG_UNRESTRICTED_GENTYPE, "'mix' only accept floating-point inputs for the interpolator a");
+
+			return vec<L, T, Q>(vec<L, U, Q>(x) * (static_cast<U>(1) - a) + vec<L, U, Q>(y) * a);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_mix_scalar<L, T, bool, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y, bool const& a)
+		{
+			return a ? y : x;
+		}
+	};
+
+	template<typename T, typename U>
+	struct compute_mix
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(T const& x, T const& y, U const& a)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<U>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || GLM_CONFIG_UNRESTRICTED_GENTYPE, "'mix' only accept floating-point inputs for the interpolator a");
+
+			return static_cast<T>(static_cast<U>(x) * (static_cast<U>(1) - a) + static_cast<U>(y) * a);
+		}
+	};
+
+	template<typename T>
+	struct compute_mix<T, bool>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(T const& x, T const& y, bool const& a)
+		{
+			return a ? y : x;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool isFloat, bool Aligned>
+	struct compute_sign
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return vec<L, T, Q>(glm::lessThan(vec<L, T, Q>(0), x)) - vec<L, T, Q>(glm::lessThan(x, vec<L, T, Q>(0)));
+		}
+	};
+
+#	if GLM_ARCH == GLM_ARCH_X86
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_sign<L, T, Q, false, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			T const Shift(static_cast<T>(sizeof(T) * 8 - 1));
+			vec<L, T, Q> const y(vec<L, typename detail::make_unsigned<T>::type, Q>(-x) >> typename detail::make_unsigned<T>::type(Shift));
+
+			return (x >> Shift) | y;
+		}
+	};
+#	endif
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_floor
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return detail::functor1<vec, L, T, T, Q>::call(std::floor, x);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_ceil
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return detail::functor1<vec, L, T, T, Q>::call(std::ceil, x);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_fract
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return x - floor(x);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_trunc
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return detail::functor1<vec, L, T, T, Q>::call(trunc, x);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_round
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return detail::functor1<vec, L, T, T, Q>::call(round, x);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_mod
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'mod' only accept floating-point inputs. Include <glm/gtc/integer.hpp> for integer inputs.");
+			return a - b * floor(a / b);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_min_vector
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+		{
+			return detail::functor2<vec, L, T, Q>::call(min, x, y);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_max_vector
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+		{
+			return detail::functor2<vec, L, T, Q>::call(max, x, y);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_clamp_vector
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal)
+		{
+			return min(max(x, minVal), maxVal);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_step_vector
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& edge, vec<L, T, Q> const& x)
+		{
+			return mix(vec<L, T, Q>(1), vec<L, T, Q>(0), glm::lessThan(x, edge));
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_smoothstep_vector
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& edge0, vec<L, T, Q> const& edge1, vec<L, T, Q> const& x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || GLM_CONFIG_UNRESTRICTED_GENTYPE, "'smoothstep' only accept floating-point inputs");
+			vec<L, T, Q> const tmp(clamp((x - edge0) / (edge1 - edge0), static_cast<T>(0), static_cast<T>(1)));
+			return tmp * tmp * (static_cast<T>(3) - static_cast<T>(2) * tmp);
+		}
+	};
+}//namespace detail
+
+	template<typename genFIType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genFIType abs(genFIType x)
+	{
+		return detail::compute_abs<genFIType, std::numeric_limits<genFIType>::is_signed>::call(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> abs(vec<L, T, Q> const& x)
+	{
+		return detail::compute_abs_vector<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// sign
+	// fast and works for any type
+	template<typename genFIType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genFIType sign(genFIType x)
+	{
+		GLM_STATIC_ASSERT(
+			std::numeric_limits<genFIType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || (std::numeric_limits<genFIType>::is_signed && std::numeric_limits<genFIType>::is_integer),
+			"'sign' only accept signed inputs");
+
+		return detail::compute_sign<1, genFIType, defaultp,
+                                    std::numeric_limits<genFIType>::is_iec559, detail::is_aligned<highp>::value>::call(vec<1, genFIType>(x)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> sign(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(
+			std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || (std::numeric_limits<T>::is_signed && std::numeric_limits<T>::is_integer),
+			"'sign' only accept signed inputs");
+
+		return detail::compute_sign<L, T, Q, std::numeric_limits<T>::is_iec559, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// floor
+	using ::std::floor;
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> floor(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'floor' only accept floating-point inputs.");
+		return detail::compute_floor<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> trunc(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'trunc' only accept floating-point inputs");
+		return detail::compute_trunc<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> round(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'round' only accept floating-point inputs");
+		return detail::compute_round<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+/*
+	// roundEven
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType roundEven(genType const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'roundEven' only accept floating-point inputs");
+
+		return genType(int(x + genType(int(x) % 2)));
+	}
+*/
+
+	// roundEven
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType roundEven(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'roundEven' only accept floating-point inputs");
+
+		int Integer = static_cast<int>(x);
+		genType IntegerPart = static_cast<genType>(Integer);
+		genType FractionalPart = fract(x);
+
+		if(FractionalPart > static_cast<genType>(0.5) || FractionalPart < static_cast<genType>(0.5))
+		{
+			return round(x);
+		}
+		else if((Integer % 2) == 0)
+		{
+			return IntegerPart;
+		}
+		else if(x <= static_cast<genType>(0)) // Work around...
+		{
+			return IntegerPart - static_cast<genType>(1);
+		}
+		else
+		{
+			return IntegerPart + static_cast<genType>(1);
+		}
+		//else // Bug on MinGW 4.5.2
+		//{
+		//	return mix(IntegerPart + genType(-1), IntegerPart + genType(1), x <= genType(0));
+		//}
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> roundEven(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'roundEven' only accept floating-point inputs");
+		return detail::functor1<vec, L, T, T, Q>::call(roundEven, x);
+	}
+
+	// ceil
+	using ::std::ceil;
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> ceil(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'ceil' only accept floating-point inputs");
+		return detail::compute_ceil<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// fract
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fract(genType x)
+	{
+		return fract(vec<1, genType>(x)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fract(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fract' only accept floating-point inputs");
+		return detail::compute_fract<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// mod
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType mod(genType x, genType y)
+	{
+#		if (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+			// Another Cuda compiler bug https://github.com/g-truc/glm/issues/530
+			vec<1, genType, defaultp> Result(mod(vec<1, genType, defaultp>(x), y));
+			return Result.x;
+#		else
+			return mod(vec<1, genType, defaultp>(x), y).x;
+#		endif
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> mod(vec<L, T, Q> const& x, T y)
+	{
+		return detail::compute_mod<L, T, Q, detail::is_aligned<Q>::value>::call(x, vec<L, T, Q>(y));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> mod(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		return detail::compute_mod<L, T, Q, detail::is_aligned<Q>::value>::call(x, y);
+	}
+
+	// modf
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType modf(genType x, genType & i)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'modf' only accept floating-point inputs");
+		return std::modf(x, &i);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<1, T, Q> modf(vec<1, T, Q> const& x, vec<1, T, Q> & i)
+	{
+		return vec<1, T, Q>(
+			modf(x.x, i.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> modf(vec<2, T, Q> const& x, vec<2, T, Q> & i)
+	{
+		return vec<2, T, Q>(
+			modf(x.x, i.x),
+			modf(x.y, i.y));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> modf(vec<3, T, Q> const& x, vec<3, T, Q> & i)
+	{
+		return vec<3, T, Q>(
+			modf(x.x, i.x),
+			modf(x.y, i.y),
+			modf(x.z, i.z));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> modf(vec<4, T, Q> const& x, vec<4, T, Q> & i)
+	{
+		return vec<4, T, Q>(
+			modf(x.x, i.x),
+			modf(x.y, i.y),
+			modf(x.z, i.z),
+			modf(x.w, i.w));
+	}
+
+	//// Only valid if (INT_MIN <= x-y <= INT_MAX)
+	//// min(x,y)
+	//r = y + ((x - y) & ((x - y) >> (sizeof(int) *
+	//CHAR_BIT - 1)));
+	//// max(x,y)
+	//r = x - ((x - y) & ((x - y) >> (sizeof(int) *
+	//CHAR_BIT - 1)));
+
+	// min
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& a, T b)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<T>::is_integer, "'min' only accept floating-point or integer inputs");
+		return detail::compute_min_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, vec<L, T, Q>(b));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+	{
+		return detail::compute_min_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, b);
+	}
+
+	// max
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& a, T b)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<T>::is_integer, "'max' only accept floating-point or integer inputs");
+		return detail::compute_max_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, vec<L, T, Q>(b));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+	{
+		return detail::compute_max_vector<L, T, Q, detail::is_aligned<Q>::value>::call(a, b);
+	}
+
+	// clamp
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType clamp(genType x, genType minVal, genType maxVal)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<genType>::is_integer, "'clamp' only accept floating-point or integer inputs");
+		return min(max(x, minVal), maxVal);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> clamp(vec<L, T, Q> const& x, T minVal, T maxVal)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<T>::is_integer, "'clamp' only accept floating-point or integer inputs");
+		return detail::compute_clamp_vector<L, T, Q, detail::is_aligned<Q>::value>::call(x, vec<L, T, Q>(minVal), vec<L, T, Q>(maxVal));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> clamp(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || std::numeric_limits<T>::is_integer, "'clamp' only accept floating-point or integer inputs");
+		return detail::compute_clamp_vector<L, T, Q, detail::is_aligned<Q>::value>::call(x, minVal, maxVal);
+	}
+
+	template<typename genTypeT, typename genTypeU>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genTypeT mix(genTypeT x, genTypeT y, genTypeU a)
+	{
+		return detail::compute_mix<genTypeT, genTypeU>::call(x, y, a);
+	}
+
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> mix(vec<L, T, Q> const& x, vec<L, T, Q> const& y, U a)
+	{
+		return detail::compute_mix_scalar<L, T, U, Q, detail::is_aligned<Q>::value>::call(x, y, a);
+	}
+
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> mix(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, U, Q> const& a)
+	{
+		return detail::compute_mix_vector<L, T, U, Q, detail::is_aligned<Q>::value>::call(x, y, a);
+	}
+
+	// step
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType step(genType edge, genType x)
+	{
+		return mix(static_cast<genType>(1), static_cast<genType>(0), x < edge);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> step(T edge, vec<L, T, Q> const& x)
+	{
+		return detail::compute_step_vector<L, T, Q, detail::is_aligned<Q>::value>::call(vec<L, T, Q>(edge), x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> step(vec<L, T, Q> const& edge, vec<L, T, Q> const& x)
+	{
+		return detail::compute_step_vector<L, T, Q, detail::is_aligned<Q>::value>::call(edge, x);
+	}
+
+	// smoothstep
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType smoothstep(genType edge0, genType edge1, genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT || GLM_CONFIG_UNRESTRICTED_GENTYPE, "'smoothstep' only accept floating-point inputs");
+
+		genType const tmp(clamp((x - edge0) / (edge1 - edge0), genType(0), genType(1)));
+		return tmp * tmp * (genType(3) - genType(2) * tmp);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> smoothstep(T edge0, T edge1, vec<L, T, Q> const& x)
+	{
+		return detail::compute_smoothstep_vector<L, T, Q, detail::is_aligned<Q>::value>::call(vec<L, T, Q>(edge0), vec<L, T, Q>(edge1), x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> smoothstep(vec<L, T, Q> const& edge0, vec<L, T, Q> const& edge1, vec<L, T, Q> const& x)
+	{
+		return detail::compute_smoothstep_vector<L, T, Q, detail::is_aligned<Q>::value>::call(edge0, edge1, x);
+	}
+
+#	if GLM_HAS_CXX11_STL
+		using std::isnan;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER bool isnan(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isnan' only accept floating-point inputs");
+
+#			if GLM_HAS_CXX11_STL
+				return std::isnan(x);
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+				return _isnan(x) != 0;
+#			elif GLM_COMPILER & GLM_COMPILER_INTEL
+#				if GLM_PLATFORM & GLM_PLATFORM_WINDOWS
+					return _isnan(x) != 0;
+#				else
+					return ::isnan(x) != 0;
+#				endif
+#			elif (GLM_COMPILER & (GLM_COMPILER_GCC | GLM_COMPILER_CLANG)) && (GLM_PLATFORM & GLM_PLATFORM_ANDROID) && __cplusplus < 201103L
+				return _isnan(x) != 0;
+#			elif (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+				return ::isnan(x) != 0;
+#			else
+				return std::isnan(x);
+#			endif
+		}
+#	endif
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> isnan(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isnan' only accept floating-point inputs");
+
+		vec<L, bool, Q> Result;
+		for (length_t l = 0; l < v.length(); ++l)
+			Result[l] = glm::isnan(v[l]);
+		return Result;
+	}
+
+#	if GLM_HAS_CXX11_STL
+		using std::isinf;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER bool isinf(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isinf' only accept floating-point inputs");
+
+#			if GLM_HAS_CXX11_STL
+				return std::isinf(x);
+#			elif GLM_COMPILER & (GLM_COMPILER_INTEL | GLM_COMPILER_VC)
+#				if(GLM_PLATFORM & GLM_PLATFORM_WINDOWS)
+					return _fpclass(x) == _FPCLASS_NINF || _fpclass(x) == _FPCLASS_PINF;
+#				else
+					return ::isinf(x);
+#				endif
+#			elif GLM_COMPILER & (GLM_COMPILER_GCC | GLM_COMPILER_CLANG)
+#				if(GLM_PLATFORM & GLM_PLATFORM_ANDROID && __cplusplus < 201103L)
+					return _isinf(x) != 0;
+#				else
+					return std::isinf(x);
+#				endif
+#			elif (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+				// http://developer.download.nvidia.com/compute/cuda/4_2/rel/toolkit/docs/online/group__CUDA__MATH__DOUBLE_g13431dd2b40b51f9139cbb7f50c18fab.html#g13431dd2b40b51f9139cbb7f50c18fab
+				return ::isinf(double(x)) != 0;
+#			else
+				return std::isinf(x);
+#			endif
+	}
+#	endif
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> isinf(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isinf' only accept floating-point inputs");
+
+		vec<L, bool, Q> Result;
+		for (length_t l = 0; l < v.length(); ++l)
+			Result[l] = glm::isinf(v[l]);
+		return Result;
+	}
+
+	GLM_FUNC_QUALIFIER int floatBitsToInt(float v)
+	{
+		union
+		{
+			float in;
+			int out;
+		} u;
+
+		u.in = v;
+
+		return u.out;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> floatBitsToInt(vec<L, float, Q> const& v)
+	{
+		return detail::functor1<vec, L, int, float, Q>::call(floatBitsToInt, v);
+	}
+
+	GLM_FUNC_QUALIFIER uint floatBitsToUint(float v)
+	{
+		union
+		{
+			float in;
+			uint out;
+		} u;
+
+		u.in = v;
+
+		return u.out;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, uint, Q> floatBitsToUint(vec<L, float, Q> const& v)
+	{
+		return detail::functor1<vec, L, uint, float, Q>::call(floatBitsToUint, v);
+	}
+
+	GLM_FUNC_QUALIFIER float intBitsToFloat(int v)
+	{
+		union
+		{
+			int in;
+			float out;
+		} u;
+
+		u.in = v;
+
+		return u.out;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, float, Q> intBitsToFloat(vec<L, int, Q> const& v)
+	{
+		return detail::functor1<vec, L, float, int, Q>::call(intBitsToFloat, v);
+	}
+
+	GLM_FUNC_QUALIFIER float uintBitsToFloat(uint v)
+	{
+		union
+		{
+			uint in;
+			float out;
+		} u;
+
+		u.in = v;
+
+		return u.out;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, float, Q> uintBitsToFloat(vec<L, uint, Q> const& v)
+	{
+		return reinterpret_cast<vec<L, float, Q>&>(const_cast<vec<L, uint, Q>&>(v));
+	}
+
+#	if GLM_HAS_CXX11_STL
+		using std::fma;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER genType fma(genType const& a, genType const& b, genType const& c)
+		{
+			return a * b + c;
+		}
+#	endif
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType frexp(genType x, int& exp)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'frexp' only accept floating-point inputs");
+
+		return std::frexp(x, &exp);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> frexp(vec<L, T, Q> const& v, vec<L, int, Q>& exp)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'frexp' only accept floating-point inputs");
+
+		vec<L, T, Q> Result;
+		for (length_t l = 0; l < v.length(); ++l)
+			Result[l] = std::frexp(v[l], &exp[l]);
+		return Result;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType ldexp(genType const& x, int const& exp)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'ldexp' only accept floating-point inputs");
+
+		return std::ldexp(x, exp);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> ldexp(vec<L, T, Q> const& v, vec<L, int, Q> const& exp)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'ldexp' only accept floating-point inputs");
+
+		vec<L, T, Q> Result;
+		for (length_t l = 0; l < v.length(); ++l)
+			Result[l] = std::ldexp(v[l], exp[l]);
+		return Result;
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_common_simd.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_common_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_common_simd.inl
new file mode 100644
index 000000000000..ce0032d33fef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_common_simd.inl
@@ -0,0 +1,231 @@
+/// @ref core
+/// @file glm/detail/func_common_simd.inl
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+#include "../simd/common.h"
+
+#include <immintrin.h>
+
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_abs_vector<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> result;
+			result.data = glm_vec4_abs(v.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_abs_vector<4, int, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v)
+		{
+			vec<4, int, Q> result;
+			result.data = glm_ivec4_abs(v.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_floor<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> result;
+			result.data = glm_vec4_floor(v.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_ceil<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> result;
+			result.data = glm_vec4_ceil(v.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_fract<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> result;
+			result.data = glm_vec4_fract(v.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_round<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> result;
+			result.data = glm_vec4_round(v.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_mod<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& x, vec<4, float, Q> const& y)
+		{
+			vec<4, float, Q> result;
+			result.data = glm_vec4_mod(x.data, y.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_min_vector<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
+		{
+			vec<4, float, Q> result;
+			result.data = _mm_min_ps(v1.data, v2.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_min_vector<4, int, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
+		{
+			vec<4, int, Q> result;
+			result.data = _mm_min_epi32(v1.data, v2.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_min_vector<4, uint, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v1, vec<4, uint, Q> const& v2)
+		{
+			vec<4, uint, Q> result;
+			result.data = _mm_min_epu32(v1.data, v2.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_max_vector<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
+		{
+			vec<4, float, Q> result;
+			result.data = _mm_max_ps(v1.data, v2.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_max_vector<4, int, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
+		{
+			vec<4, int, Q> result;
+			result.data = _mm_max_epi32(v1.data, v2.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_max_vector<4, uint, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v1, vec<4, uint, Q> const& v2)
+		{
+			vec<4, uint, Q> result;
+			result.data = _mm_max_epu32(v1.data, v2.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_clamp_vector<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& x, vec<4, float, Q> const& minVal, vec<4, float, Q> const& maxVal)
+		{
+			vec<4, float, Q> result;
+			result.data = _mm_min_ps(_mm_max_ps(x.data, minVal.data), maxVal.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_clamp_vector<4, int, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& x, vec<4, int, Q> const& minVal, vec<4, int, Q> const& maxVal)
+		{
+			vec<4, int, Q> result;
+			result.data = _mm_min_epi32(_mm_max_epi32(x.data, minVal.data), maxVal.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_clamp_vector<4, uint, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& x, vec<4, uint, Q> const& minVal, vec<4, uint, Q> const& maxVal)
+		{
+			vec<4, uint, Q> result;
+			result.data = _mm_min_epu32(_mm_max_epu32(x.data, minVal.data), maxVal.data);
+			return result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_mix_vector<4, float, bool, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& x, vec<4, float, Q> const& y, vec<4, bool, Q> const& a)
+		{
+			__m128i const Load = _mm_set_epi32(-static_cast<int>(a.w), -static_cast<int>(a.z), -static_cast<int>(a.y), -static_cast<int>(a.x));
+			__m128 const Mask = _mm_castsi128_ps(Load);
+
+			vec<4, float, Q> Result;
+#			if 0 && GLM_ARCH & GLM_ARCH_AVX
+				Result.data = _mm_blendv_ps(x.data, y.data, Mask);
+#			else
+				Result.data = _mm_or_ps(_mm_and_ps(Mask, y.data), _mm_andnot_ps(Mask, x.data));
+#			endif
+			return Result;
+		}
+	};
+/* FIXME
+	template<qualifier Q>
+	struct compute_step_vector<float, Q, tvec4>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& edge, vec<4, float, Q> const& x)
+		{
+			vec<4, float, Q> Result;
+			result.data = glm_vec4_step(edge.data, x.data);
+			return result;
+		}
+	};
+*/
+	template<qualifier Q>
+	struct compute_smoothstep_vector<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& edge0, vec<4, float, Q> const& edge1, vec<4, float, Q> const& x)
+		{
+			vec<4, float, Q> Result;
+			Result.data = glm_vec4_smoothstep(edge0.data, edge1.data, x.data);
+			return Result;
+		}
+	};
+}//namespace detail
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential.inl
new file mode 100644
index 000000000000..2efcdc60adc6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential.inl
@@ -0,0 +1,152 @@
+/// @ref core
+/// @file glm/detail/func_exponential.inl
+
+#include "../vector_relational.hpp"
+#include "_vectorize.hpp"
+#include <limits>
+#include <cmath>
+#include <cassert>
+
+namespace glm{
+namespace detail
+{
+#	if GLM_HAS_CXX11_STL
+		using std::log2;
+#	else
+		template<typename genType>
+		genType log2(genType Value)
+		{
+			return std::log(Value) * static_cast<genType>(1.4426950408889634073599246810019);
+		}
+#	endif
+
+	template<length_t L, typename T, qualifier Q, bool isFloat, bool Aligned>
+	struct compute_log2
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'log2' only accept floating-point inputs. Include <glm/gtc/integer.hpp> for integer inputs.");
+
+			return detail::functor1<vec, L, T, T, Q>::call(log2, v);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_sqrt
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return detail::functor1<vec, L, T, T, Q>::call(std::sqrt, x);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_inversesqrt
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			return static_cast<T>(1) / sqrt(x);
+		}
+	};
+
+	template<length_t L, bool Aligned>
+	struct compute_inversesqrt<L, float, lowp, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, float, lowp> call(vec<L, float, lowp> const& x)
+		{
+			vec<L, float, lowp> tmp(x);
+			vec<L, float, lowp> xhalf(tmp * 0.5f);
+			vec<L, uint, lowp>* p = reinterpret_cast<vec<L, uint, lowp>*>(const_cast<vec<L, float, lowp>*>(&x));
+			vec<L, uint, lowp> i = vec<L, uint, lowp>(0x5f375a86) - (*p >> vec<L, uint, lowp>(1));
+			vec<L, float, lowp>* ptmp = reinterpret_cast<vec<L, float, lowp>*>(&i);
+			tmp = *ptmp;
+			tmp = tmp * (1.5f - xhalf * tmp * tmp);
+			return tmp;
+		}
+	};
+}//namespace detail
+
+	// pow
+	using std::pow;
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> pow(vec<L, T, Q> const& base, vec<L, T, Q> const& exponent)
+	{
+		return detail::functor2<vec, L, T, Q>::call(pow, base, exponent);
+	}
+
+	// exp
+	using std::exp;
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> exp(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(exp, x);
+	}
+
+	// log
+	using std::log;
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> log(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(log, x);
+	}
+
+#   if GLM_HAS_CXX11_STL
+    using std::exp2;
+#   else
+	//exp2, ln2 = 0.69314718055994530941723212145818f
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType exp2(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'exp2' only accept floating-point inputs");
+
+		return std::exp(static_cast<genType>(0.69314718055994530941723212145818) * x);
+	}
+#   endif
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> exp2(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(exp2, x);
+	}
+
+	// log2, ln2 = 0.69314718055994530941723212145818f
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType log2(genType x)
+	{
+		return log2(vec<1, genType>(x)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> log2(vec<L, T, Q> const& x)
+	{
+		return detail::compute_log2<L, T, Q, std::numeric_limits<T>::is_iec559, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// sqrt
+	using std::sqrt;
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> sqrt(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'sqrt' only accept floating-point inputs");
+		return detail::compute_sqrt<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// inversesqrt
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType inversesqrt(genType x)
+	{
+		return static_cast<genType>(1) / sqrt(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> inversesqrt(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'inversesqrt' only accept floating-point inputs");
+		return detail::compute_inversesqrt<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_exponential_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential_simd.inl
new file mode 100644
index 000000000000..fb78951727f1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_exponential_simd.inl
@@ -0,0 +1,37 @@
+/// @ref core
+/// @file glm/detail/func_exponential_simd.inl
+
+#include "../simd/exponential.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_sqrt<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> Result;
+			Result.data = _mm_sqrt_ps(v.data);
+			return Result;
+		}
+	};
+
+#	if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+	template<>
+	struct compute_sqrt<4, float, aligned_lowp, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, aligned_lowp> call(vec<4, float, aligned_lowp> const& v)
+		{
+			vec<4, float, aligned_lowp> Result;
+			Result.data = glm_vec4_sqrt_lowp(v.data);
+			return Result;
+		}
+	};
+#	endif
+}//namespace detail
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric.inl
new file mode 100644
index 000000000000..404c99056ab4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric.inl
@@ -0,0 +1,243 @@
+#include "../exponential.hpp"
+#include "../common.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_length
+	{
+		GLM_FUNC_QUALIFIER static T call(vec<L, T, Q> const& v)
+		{
+			return sqrt(dot(v, v));
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_distance
+	{
+		GLM_FUNC_QUALIFIER static T call(vec<L, T, Q> const& p0, vec<L, T, Q> const& p1)
+		{
+			return length(p1 - p0);
+		}
+	};
+
+	template<typename V, typename T, bool Aligned>
+	struct compute_dot{};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_dot<vec<1, T, Q>, T, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(vec<1, T, Q> const& a, vec<1, T, Q> const& b)
+		{
+			return a.x * b.x;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_dot<vec<2, T, Q>, T, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(vec<2, T, Q> const& a, vec<2, T, Q> const& b)
+		{
+			vec<2, T, Q> tmp(a * b);
+			return tmp.x + tmp.y;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_dot<vec<3, T, Q>, T, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(vec<3, T, Q> const& a, vec<3, T, Q> const& b)
+		{
+			vec<3, T, Q> tmp(a * b);
+			return tmp.x + tmp.y + tmp.z;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_dot<vec<4, T, Q>, T, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			vec<4, T, Q> tmp(a * b);
+			return (tmp.x + tmp.y) + (tmp.z + tmp.w);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_cross
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<3, T, Q> call(vec<3, T, Q> const& x, vec<3, T, Q> const& y)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'cross' accepts only floating-point inputs");
+
+			return vec<3, T, Q>(
+				x.y * y.z - y.y * x.z,
+				x.z * y.x - y.z * x.x,
+				x.x * y.y - y.x * x.y);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_normalize
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'normalize' accepts only floating-point inputs");
+
+			return v * inversesqrt(dot(v, v));
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_faceforward
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& N, vec<L, T, Q> const& I, vec<L, T, Q> const& Nref)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'normalize' accepts only floating-point inputs");
+
+			return dot(Nref, I) < static_cast<T>(0) ? N : -N;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_reflect
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& I, vec<L, T, Q> const& N)
+		{
+			return I - N * dot(N, I) * static_cast<T>(2);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_refract
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& I, vec<L, T, Q> const& N, T eta)
+		{
+			T const dotValue(dot(N, I));
+			T const k(static_cast<T>(1) - eta * eta * (static_cast<T>(1) - dotValue * dotValue));
+			vec<L, T, Q> const Result =
+                (k >= static_cast<T>(0)) ? (eta * I - (eta * dotValue + std::sqrt(k)) * N) : vec<L, T, Q>(0);
+			return Result;
+		}
+	};
+}//namespace detail
+
+	// length
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType length(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'length' accepts only floating-point inputs");
+
+		return abs(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T length(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'length' accepts only floating-point inputs");
+
+		return detail::compute_length<L, T, Q, detail::is_aligned<Q>::value>::call(v);
+	}
+
+	// distance
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType distance(genType const& p0, genType const& p1)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'distance' accepts only floating-point inputs");
+
+		return length(p1 - p0);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T distance(vec<L, T, Q> const& p0, vec<L, T, Q> const& p1)
+	{
+		return detail::compute_distance<L, T, Q, detail::is_aligned<Q>::value>::call(p0, p1);
+	}
+
+	// dot
+	template<typename T>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T dot(T x, T y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'dot' accepts only floating-point inputs");
+		return x * y;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T dot(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'dot' accepts only floating-point inputs");
+		return detail::compute_dot<vec<L, T, Q>, T, detail::is_aligned<Q>::value>::call(x, y);
+	}
+
+	// cross
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> cross(vec<3, T, Q> const& x, vec<3, T, Q> const& y)
+	{
+		return detail::compute_cross<T, Q, detail::is_aligned<Q>::value>::call(x, y);
+	}
+/*
+	// normalize
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType normalize(genType const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'normalize' accepts only floating-point inputs");
+
+		return x < genType(0) ? genType(-1) : genType(1);
+	}
+*/
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> normalize(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'normalize' accepts only floating-point inputs");
+
+		return detail::compute_normalize<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// faceforward
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType faceforward(genType const& N, genType const& I, genType const& Nref)
+	{
+		return dot(Nref, I) < static_cast<genType>(0) ? N : -N;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> faceforward(vec<L, T, Q> const& N, vec<L, T, Q> const& I, vec<L, T, Q> const& Nref)
+	{
+		return detail::compute_faceforward<L, T, Q, detail::is_aligned<Q>::value>::call(N, I, Nref);
+	}
+
+	// reflect
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType reflect(genType const& I, genType const& N)
+	{
+		return I - N * dot(N, I) * genType(2);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> reflect(vec<L, T, Q> const& I, vec<L, T, Q> const& N)
+	{
+		return detail::compute_reflect<L, T, Q, detail::is_aligned<Q>::value>::call(I, N);
+	}
+
+	// refract
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType refract(genType const& I, genType const& N, genType eta)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'refract' accepts only floating-point inputs");
+		genType const dotValue(dot(N, I));
+		genType const k(static_cast<genType>(1) - eta * eta * (static_cast<genType>(1) - dotValue * dotValue));
+		return (eta * I - (eta * dotValue + sqrt(k)) * N) * static_cast<genType>(k >= static_cast<genType>(0));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> refract(vec<L, T, Q> const& I, vec<L, T, Q> const& N, T eta)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'refract' accepts only floating-point inputs");
+		return detail::compute_refract<L, T, Q, detail::is_aligned<Q>::value>::call(I, N, eta);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_geometric_simd.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric_simd.inl
new file mode 100644
index 000000000000..2076dae055c3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_geometric_simd.inl
@@ -0,0 +1,163 @@
+/// @ref core
+/// @file glm/detail/func_geometric_simd.inl
+
+#include "../simd/geometric.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_length<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& v)
+		{
+			return _mm_cvtss_f32(glm_vec4_length(v.data));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_distance<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& p0, vec<4, float, Q> const& p1)
+		{
+			return _mm_cvtss_f32(glm_vec4_distance(p0.data, p1.data));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_dot<vec<4, float, Q>, float, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& x, vec<4, float, Q> const& y)
+		{
+			return _mm_cvtss_f32(glm_vec1_dot(x.data, y.data));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_cross<float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, float, Q> call(vec<3, float, Q> const& a, vec<3, float, Q> const& b)
+		{
+			__m128 const set0 = _mm_set_ps(0.0f, a.z, a.y, a.x);
+			__m128 const set1 = _mm_set_ps(0.0f, b.z, b.y, b.x);
+			__m128 const xpd0 = glm_vec4_cross(set0, set1);
+
+			vec<4, float, Q> Result;
+			Result.data = xpd0;
+			return vec<3, float, Q>(Result);
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_normalize<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			vec<4, float, Q> Result;
+			Result.data = glm_vec4_normalize(v.data);
+			return Result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_faceforward<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& N, vec<4, float, Q> const& I, vec<4, float, Q> const& Nref)
+		{
+			vec<4, float, Q> Result;
+			Result.data = glm_vec4_faceforward(N.data, I.data, Nref.data);
+			return Result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_reflect<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& I, vec<4, float, Q> const& N)
+		{
+			vec<4, float, Q> Result;
+			Result.data = glm_vec4_reflect(I.data, N.data);
+			return Result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_refract<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& I, vec<4, float, Q> const& N, float eta)
+		{
+			vec<4, float, Q> Result;
+			Result.data = glm_vec4_refract(I.data, N.data, _mm_set1_ps(eta));
+			return Result;
+		}
+	};
+}//namespace detail
+}//namespace glm
+
+#elif GLM_ARCH & GLM_ARCH_NEON_BIT
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_length<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& v)
+		{
+			return sqrt(compute_dot<vec<4, float, Q>, float, true>::call(v, v));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_distance<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& p0, vec<4, float, Q> const& p1)
+		{
+			return compute_length<4, float, Q, true>::call(p1 - p0);
+		}
+	};
+
+
+	template<qualifier Q>
+	struct compute_dot<vec<4, float, Q>, float, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(vec<4, float, Q> const& x, vec<4, float, Q> const& y)
+		{
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+			float32x4_t v = vmulq_f32(x.data, y.data);
+			return vaddvq_f32(v);
+#else  // Armv7a with Neon
+			float32x4_t p = vmulq_f32(x.data, y.data);
+			float32x2_t v = vpadd_f32(vget_low_f32(p), vget_high_f32(p));
+			v = vpadd_f32(v, v);
+			return vget_lane_f32(v, 0);
+#endif
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_normalize<4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v)
+		{
+			float32x4_t p = vmulq_f32(v.data, v.data);
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+			p = vpaddq_f32(p, p);
+			p = vpaddq_f32(p, p);
+#else
+			float32x2_t t = vpadd_f32(vget_low_f32(p), vget_high_f32(p));
+			t = vpadd_f32(t, t);
+			p = vcombine_f32(t, t);
+#endif
+
+			float32x4_t vd = vrsqrteq_f32(p);
+			vec<4, float, Q> Result;
+			Result.data = vmulq_f32(v.data, vd);
+			return Result;
+		}
+	};
+}//namespace detail
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_integer.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_integer.inl
new file mode 100644
index 000000000000..68b92e99fc61
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_integer.inl
@@ -0,0 +1,369 @@
+/// @ref core
+
+#include "_vectorize.hpp"
+#if(GLM_ARCH & GLM_ARCH_X86 && GLM_COMPILER & GLM_COMPILER_VC)
+#	include <intrin.h>
+#	pragma intrinsic(_BitScanReverse)
+#endif//(GLM_ARCH & GLM_ARCH_X86 && GLM_COMPILER & GLM_COMPILER_VC)
+#include <limits>
+
+#if !GLM_HAS_EXTENDED_INTEGER_TYPE
+#	if GLM_COMPILER & GLM_COMPILER_GCC
+#		pragma GCC diagnostic ignored "-Wlong-long"
+#	endif
+#	if (GLM_COMPILER & GLM_COMPILER_CLANG)
+#		pragma clang diagnostic ignored "-Wc++11-long-long"
+#	endif
+#endif
+
+namespace glm{
+namespace detail
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER T mask(T Bits)
+	{
+		return Bits >= static_cast<T>(sizeof(T) * 8) ? ~static_cast<T>(0) : (static_cast<T>(1) << Bits) - static_cast<T>(1);
+	}
+
+	template<length_t L, typename T, qualifier Q, bool Aligned, bool EXEC>
+	struct compute_bitfieldReverseStep
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T, T)
+		{
+			return v;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_bitfieldReverseStep<L, T, Q, Aligned, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T Mask, T Shift)
+		{
+			return (v & Mask) << Shift | (v & (~Mask)) >> Shift;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned, bool EXEC>
+	struct compute_bitfieldBitCountStep
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T, T)
+		{
+			return v;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_bitfieldBitCountStep<L, T, Q, Aligned, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T Mask, T Shift)
+		{
+			return (v & Mask) + ((v >> Shift) & Mask);
+		}
+	};
+
+	template<typename genIUType, size_t Bits>
+	struct compute_findLSB
+	{
+		GLM_FUNC_QUALIFIER static int call(genIUType Value)
+		{
+			if(Value == 0)
+				return -1;
+
+			return glm::bitCount(~Value & (Value - static_cast<genIUType>(1)));
+		}
+	};
+
+#	if GLM_HAS_BITSCAN_WINDOWS
+		template<typename genIUType>
+		struct compute_findLSB<genIUType, 32>
+		{
+			GLM_FUNC_QUALIFIER static int call(genIUType Value)
+			{
+				unsigned long Result(0);
+				unsigned char IsNotNull = _BitScanForward(&Result, *reinterpret_cast<unsigned long*>(&Value));
+				return IsNotNull ? int(Result) : -1;
+			}
+		};
+
+#		if !((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_MODEL == GLM_MODEL_32))
+		template<typename genIUType>
+		struct compute_findLSB<genIUType, 64>
+		{
+			GLM_FUNC_QUALIFIER static int call(genIUType Value)
+			{
+				unsigned long Result(0);
+				unsigned char IsNotNull = _BitScanForward64(&Result, *reinterpret_cast<unsigned __int64*>(&Value));
+				return IsNotNull ? int(Result) : -1;
+			}
+		};
+#		endif
+#	endif//GLM_HAS_BITSCAN_WINDOWS
+
+	template<length_t L, typename T, qualifier Q, bool EXEC = true>
+	struct compute_findMSB_step_vec
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, T Shift)
+		{
+			return x | (x >> Shift);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct compute_findMSB_step_vec<L, T, Q, false>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, T)
+		{
+			return x;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, int>
+	struct compute_findMSB_vec
+	{
+		GLM_FUNC_QUALIFIER static vec<L, int, Q> call(vec<L, T, Q> const& v)
+		{
+			vec<L, T, Q> x(v);
+			x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >=  8>::call(x, static_cast<T>( 1));
+			x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >=  8>::call(x, static_cast<T>( 2));
+			x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >=  8>::call(x, static_cast<T>( 4));
+			x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 16>::call(x, static_cast<T>( 8));
+			x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 32>::call(x, static_cast<T>(16));
+			x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 64>::call(x, static_cast<T>(32));
+			return vec<L, int, Q>(sizeof(T) * 8 - 1) - glm::bitCount(~x);
+		}
+	};
+
+#	if GLM_HAS_BITSCAN_WINDOWS
+		template<typename genIUType>
+		GLM_FUNC_QUALIFIER int compute_findMSB_32(genIUType Value)
+		{
+			unsigned long Result(0);
+			unsigned char IsNotNull = _BitScanReverse(&Result, *reinterpret_cast<unsigned long*>(&Value));
+			return IsNotNull ? int(Result) : -1;
+		}
+
+		template<length_t L, typename T, qualifier Q>
+		struct compute_findMSB_vec<L, T, Q, 32>
+		{
+			GLM_FUNC_QUALIFIER static vec<L, int, Q> call(vec<L, T, Q> const& x)
+			{
+				return detail::functor1<vec, L, int, T, Q>::call(compute_findMSB_32, x);
+			}
+		};
+
+#		if !((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_MODEL == GLM_MODEL_32))
+		template<typename genIUType>
+		GLM_FUNC_QUALIFIER int compute_findMSB_64(genIUType Value)
+		{
+			unsigned long Result(0);
+			unsigned char IsNotNull = _BitScanReverse64(&Result, *reinterpret_cast<unsigned __int64*>(&Value));
+			return IsNotNull ? int(Result) : -1;
+		}
+
+		template<length_t L, typename T, qualifier Q>
+		struct compute_findMSB_vec<L, T, Q, 64>
+		{
+			GLM_FUNC_QUALIFIER static vec<L, int, Q> call(vec<L, T, Q> const& x)
+			{
+				return detail::functor1<vec, L, int, T, Q>::call(compute_findMSB_64, x);
+			}
+		};
+#		endif
+#	endif//GLM_HAS_BITSCAN_WINDOWS
+}//namespace detail
+
+	// uaddCarry
+	GLM_FUNC_QUALIFIER uint uaddCarry(uint const& x, uint const& y, uint & Carry)
+	{
+		detail::uint64 const Value64(static_cast<detail::uint64>(x) + static_cast<detail::uint64>(y));
+		detail::uint64 const Max32((static_cast<detail::uint64>(1) << static_cast<detail::uint64>(32)) - static_cast<detail::uint64>(1));
+		Carry = Value64 > Max32 ? 1u : 0u;
+		return static_cast<uint>(Value64 % (Max32 + static_cast<detail::uint64>(1)));
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, uint, Q> uaddCarry(vec<L, uint, Q> const& x, vec<L, uint, Q> const& y, vec<L, uint, Q>& Carry)
+	{
+		vec<L, detail::uint64, Q> Value64(vec<L, detail::uint64, Q>(x) + vec<L, detail::uint64, Q>(y));
+		vec<L, detail::uint64, Q> Max32((static_cast<detail::uint64>(1) << static_cast<detail::uint64>(32)) - static_cast<detail::uint64>(1));
+		Carry = mix(vec<L, uint, Q>(0), vec<L, uint, Q>(1), greaterThan(Value64, Max32));
+		return vec<L, uint, Q>(Value64 % (Max32 + static_cast<detail::uint64>(1)));
+	}
+	
+	// usubBorrow
+	GLM_FUNC_QUALIFIER uint usubBorrow(uint const& x, uint const& y, uint & Borrow)
+	{
+		Borrow = x >= y ? static_cast<uint>(0) : static_cast<uint>(1);
+		return y - x;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, uint, Q> usubBorrow(vec<L, uint, Q> const& x, vec<L, uint, Q> const& y, vec<L, uint, Q>& Borrow)
+	{
+		Borrow = mix(vec<L, uint, Q>(1), vec<L, uint, Q>(0), greaterThanEqual(x, y));
+		vec<L, uint, Q> const YgeX(y - x);
+		vec<L, uint, Q> const XgeY(vec<L, uint, Q>((static_cast<detail::int64>(1) << static_cast<detail::int64>(32)) + (vec<L, detail::int64, Q>(y) - vec<L, detail::int64, Q>(x))));
+		return mix(XgeY, YgeX, greaterThanEqual(y, x));
+	}
+
+	// umulExtended
+	GLM_FUNC_QUALIFIER void umulExtended(uint const& x, uint const& y, uint & msb, uint & lsb)
+	{
+		detail::uint64 Value64 = static_cast<detail::uint64>(x) * static_cast<detail::uint64>(y);
+		msb = static_cast<uint>(Value64 >> static_cast<detail::uint64>(32));
+		lsb = static_cast<uint>(Value64);
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER void umulExtended(vec<L, uint, Q> const& x, vec<L, uint, Q> const& y, vec<L, uint, Q>& msb, vec<L, uint, Q>& lsb)
+	{
+		vec<L, detail::uint64, Q> Value64(vec<L, detail::uint64, Q>(x) * vec<L, detail::uint64, Q>(y));
+		msb = vec<L, uint, Q>(Value64 >> static_cast<detail::uint64>(32));
+		lsb = vec<L, uint, Q>(Value64);
+	}
+
+	// imulExtended
+	GLM_FUNC_QUALIFIER void imulExtended(int x, int y, int& msb, int& lsb)
+	{
+		detail::int64 Value64 = static_cast<detail::int64>(x) * static_cast<detail::int64>(y);
+		msb = static_cast<int>(Value64 >> static_cast<detail::int64>(32));
+		lsb = static_cast<int>(Value64);
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER void imulExtended(vec<L, int, Q> const& x, vec<L, int, Q> const& y, vec<L, int, Q>& msb, vec<L, int, Q>& lsb)
+	{
+		vec<L, detail::int64, Q> Value64(vec<L, detail::int64, Q>(x) * vec<L, detail::int64, Q>(y));
+		lsb = vec<L, int, Q>(Value64 & static_cast<detail::int64>(0xFFFFFFFF));
+		msb = vec<L, int, Q>((Value64 >> static_cast<detail::int64>(32)) & static_cast<detail::int64>(0xFFFFFFFF));
+	}
+
+	// bitfieldExtract
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType bitfieldExtract(genIUType Value, int Offset, int Bits)
+	{
+		return bitfieldExtract(vec<1, genIUType>(Value), Offset, Bits).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldExtract(vec<L, T, Q> const& Value, int Offset, int Bits)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldExtract' only accept integer inputs");
+
+		return (Value >> static_cast<T>(Offset)) & static_cast<T>(detail::mask(Bits));
+	}
+
+	// bitfieldInsert
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType bitfieldInsert(genIUType const& Base, genIUType const& Insert, int Offset, int Bits)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'bitfieldInsert' only accept integer values");
+
+		return bitfieldInsert(vec<1, genIUType>(Base), vec<1, genIUType>(Insert), Offset, Bits).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldInsert(vec<L, T, Q> const& Base, vec<L, T, Q> const& Insert, int Offset, int Bits)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldInsert' only accept integer values");
+
+		T const Mask = detail::mask(static_cast<T>(Bits)) << Offset;
+		return (Base & ~Mask) | ((Insert << static_cast<T>(Offset)) & Mask);
+	}
+
+	// bitfieldReverse
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType bitfieldReverse(genIUType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'bitfieldReverse' only accept integer values");
+
+		return bitfieldReverse(glm::vec<1, genIUType, glm::defaultp>(x)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldReverse(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldReverse' only accept integer values");
+
+		vec<L, T, Q> x(v);
+		x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>=  2>::call(x, static_cast<T>(0x5555555555555555ull), static_cast<T>( 1));
+		x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>=  4>::call(x, static_cast<T>(0x3333333333333333ull), static_cast<T>( 2));
+		x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>=  8>::call(x, static_cast<T>(0x0F0F0F0F0F0F0F0Full), static_cast<T>( 4));
+		x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 16>::call(x, static_cast<T>(0x00FF00FF00FF00FFull), static_cast<T>( 8));
+		x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 32>::call(x, static_cast<T>(0x0000FFFF0000FFFFull), static_cast<T>(16));
+		x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 64>::call(x, static_cast<T>(0x00000000FFFFFFFFull), static_cast<T>(32));
+		return x;
+	}
+
+	// bitCount
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER int bitCount(genIUType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'bitCount' only accept integer values");
+
+		return bitCount(glm::vec<1, genIUType, glm::defaultp>(x)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> bitCount(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitCount' only accept integer values");
+
+#		if GLM_COMPILER & GLM_COMPILER_VC
+#			pragma warning(push)
+#			pragma warning(disable : 4310) //cast truncates constant value
+#		endif
+
+		vec<L, typename detail::make_unsigned<T>::type, Q> x(v);
+		x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>=  2>::call(x, typename detail::make_unsigned<T>::type(0x5555555555555555ull), typename detail::make_unsigned<T>::type( 1));
+		x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>=  4>::call(x, typename detail::make_unsigned<T>::type(0x3333333333333333ull), typename detail::make_unsigned<T>::type( 2));
+		x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>=  8>::call(x, typename detail::make_unsigned<T>::type(0x0F0F0F0F0F0F0F0Full), typename detail::make_unsigned<T>::type( 4));
+		x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 16>::call(x, typename detail::make_unsigned<T>::type(0x00FF00FF00FF00FFull), typename detail::make_unsigned<T>::type( 8));
+		x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 32>::call(x, typename detail::make_unsigned<T>::type(0x0000FFFF0000FFFFull), typename detail::make_unsigned<T>::type(16));
+		x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 64>::call(x, typename detail::make_unsigned<T>::type(0x00000000FFFFFFFFull), typename detail::make_unsigned<T>::type(32));
+		return vec<L, int, Q>(x);
+
+#		if GLM_COMPILER & GLM_COMPILER_VC
+#			pragma warning(pop)
+#		endif
+	}
+
+	// findLSB
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER int findLSB(genIUType Value)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'findLSB' only accept integer values");
+
+		return detail::compute_findLSB<genIUType, sizeof(genIUType) * 8>::call(Value);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> findLSB(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'findLSB' only accept integer values");
+
+		return detail::functor1<vec, L, int, T, Q>::call(findLSB, x);
+	}
+
+	// findMSB
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER int findMSB(genIUType v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'findMSB' only accept integer values");
+
+		return findMSB(vec<1, genIUType>(v)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> findMSB(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'findMSB' only accept integer values");
+
+		return detail::compute_findMSB_vec<L, T, Q, static_cast<int>(sizeof(T) * 8)>::call(v);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_integer_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_integer_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_integer_simd.inl
new file mode 100644
index 000000000000..8be6c9ce4dc1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_integer_simd.inl
@@ -0,0 +1,65 @@
+#include "../simd/integer.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_bitfieldReverseStep<4, uint, Q, true, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v, uint Mask, uint Shift)
+		{
+			__m128i const set0 = v.data;
+
+			__m128i const set1 = _mm_set1_epi32(static_cast<int>(Mask));
+			__m128i const and1 = _mm_and_si128(set0, set1);
+			__m128i const sft1 = _mm_slli_epi32(and1, Shift);
+
+			__m128i const set2 = _mm_andnot_si128(set0, _mm_set1_epi32(-1));
+			__m128i const and2 = _mm_and_si128(set0, set2);
+			__m128i const sft2 = _mm_srai_epi32(and2, Shift);
+
+			__m128i const or0 = _mm_or_si128(sft1, sft2);
+
+			return or0;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_bitfieldBitCountStep<4, uint, Q, true, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v, uint Mask, uint Shift)
+		{
+			__m128i const set0 = v.data;
+
+			__m128i const set1 = _mm_set1_epi32(static_cast<int>(Mask));
+			__m128i const and0 = _mm_and_si128(set0, set1);
+			__m128i const sft0 = _mm_slli_epi32(set0, Shift);
+			__m128i const and1 = _mm_and_si128(sft0, set1);
+			__m128i const add0 = _mm_add_epi32(and0, and1);
+
+			return add0;
+		}
+	};
+}//namespace detail
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+	template<>
+	GLM_FUNC_QUALIFIER int bitCount(uint x)
+	{
+		return _mm_popcnt_u32(x);
+	}
+
+#	if(GLM_MODEL == GLM_MODEL_64)
+	template<>
+	GLM_FUNC_QUALIFIER int bitCount(detail::uint64 x)
+	{
+		return static_cast<int>(_mm_popcnt_u64(x));
+	}
+#	endif//GLM_MODEL
+#	endif//GLM_ARCH
+
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix.inl
new file mode 100644
index 000000000000..c2d568ff471d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix.inl
@@ -0,0 +1,443 @@
+#include "../geometric.hpp"
+#include <limits>
+
+namespace glm{
+namespace detail
+{
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_matrixCompMult
+	{
+		GLM_FUNC_QUALIFIER static mat<C, R, T, Q> call(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y)
+		{
+			mat<C, R, T, Q> Result;
+			for(length_t i = 0; i < Result.length(); ++i)
+				Result[i] = x[i] * y[i];
+			return Result;
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool IsFloat, bool Aligned>
+	struct compute_matrixCompMult_type {
+		GLM_FUNC_QUALIFIER static mat<C, R, T, Q> call(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_GENTYPE, 
+				"'matrixCompMult' only accept floating-point inputs, include <glm/ext/matrix_integer.hpp> to discard this restriction.");
+			return detail::compute_matrixCompMult<C, R, T, Q, detail::is_aligned<Q>::value>::call(x, y);
+		}
+	};
+
+	template<length_t DA, length_t DB, typename T, qualifier Q>
+	struct compute_outerProduct {
+		GLM_FUNC_QUALIFIER static typename detail::outerProduct_trait<DA, DB, T, Q>::type call(vec<DA, T, Q> const& c, vec<DB, T, Q> const& r)
+		{
+			typename detail::outerProduct_trait<DA, DB, T, Q>::type m;
+			for(length_t i = 0; i < m.length(); ++i)
+				m[i] = c * r[i];
+			return m;
+		}
+	};
+
+	template<length_t DA, length_t DB, typename T, qualifier Q, bool IsFloat>
+	struct compute_outerProduct_type {
+		GLM_FUNC_QUALIFIER static typename detail::outerProduct_trait<DA, DB, T, Q>::type call(vec<DA, T, Q> const& c, vec<DB, T, Q> const& r)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_GENTYPE,
+				"'outerProduct' only accept floating-point inputs, include <glm/ext/matrix_integer.hpp> to discard this restriction.");
+
+			return detail::compute_outerProduct<DA, DB, T, Q>::call(c, r);
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_transpose{};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<2, 2, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<2, 2, T, Q> call(mat<2, 2, T, Q> const& m)
+		{
+			mat<2, 2, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<2, 3, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<3, 2, T, Q> call(mat<2, 3, T, Q> const& m)
+		{
+			mat<3,2, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[2][0] = m[0][2];
+			Result[2][1] = m[1][2];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<2, 4, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 2, T, Q> call(mat<2, 4, T, Q> const& m)
+		{
+			mat<4, 2, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[2][0] = m[0][2];
+			Result[2][1] = m[1][2];
+			Result[3][0] = m[0][3];
+			Result[3][1] = m[1][3];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<3, 2, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<2, 3, T, Q> call(mat<3, 2, T, Q> const& m)
+		{
+			mat<2, 3, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[0][2] = m[2][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[1][2] = m[2][1];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<3, 3, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<3, 3, T, Q> call(mat<3, 3, T, Q> const& m)
+		{
+			mat<3, 3, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[0][2] = m[2][0];
+
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[1][2] = m[2][1];
+
+			Result[2][0] = m[0][2];
+			Result[2][1] = m[1][2];
+			Result[2][2] = m[2][2];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<3, 4, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 3, T, Q> call(mat<3, 4, T, Q> const& m)
+		{
+			mat<4, 3, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[0][2] = m[2][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[1][2] = m[2][1];
+			Result[2][0] = m[0][2];
+			Result[2][1] = m[1][2];
+			Result[2][2] = m[2][2];
+			Result[3][0] = m[0][3];
+			Result[3][1] = m[1][3];
+			Result[3][2] = m[2][3];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<4, 2, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<2, 4, T, Q> call(mat<4, 2, T, Q> const& m)
+		{
+			mat<2, 4, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[0][2] = m[2][0];
+			Result[0][3] = m[3][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[1][2] = m[2][1];
+			Result[1][3] = m[3][1];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<4, 3, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<3, 4, T, Q> call(mat<4, 3, T, Q> const& m)
+		{
+			mat<3, 4, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[0][2] = m[2][0];
+			Result[0][3] = m[3][0];
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[1][2] = m[2][1];
+			Result[1][3] = m[3][1];
+			Result[2][0] = m[0][2];
+			Result[2][1] = m[1][2];
+			Result[2][2] = m[2][2];
+			Result[2][3] = m[3][2];
+			return Result;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_transpose<4, 4, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 4, T, Q> call(mat<4, 4, T, Q> const& m)
+		{
+			mat<4, 4, T, Q> Result;
+			Result[0][0] = m[0][0];
+			Result[0][1] = m[1][0];
+			Result[0][2] = m[2][0];
+			Result[0][3] = m[3][0];
+
+			Result[1][0] = m[0][1];
+			Result[1][1] = m[1][1];
+			Result[1][2] = m[2][1];
+			Result[1][3] = m[3][1];
+
+			Result[2][0] = m[0][2];
+			Result[2][1] = m[1][2];
+			Result[2][2] = m[2][2];
+			Result[2][3] = m[3][2];
+
+			Result[3][0] = m[0][3];
+			Result[3][1] = m[1][3];
+			Result[3][2] = m[2][3];
+			Result[3][3] = m[3][3];
+			return Result;
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool IsFloat, bool Aligned>
+	struct compute_transpose_type {
+		GLM_FUNC_QUALIFIER static mat<R, C, T, Q> call(mat<C, R, T, Q> const& m)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_GENTYPE, 
+				"'transpose' only accept floating-point inputs, include <glm/ext/matrix_integer.hpp> to discard this restriction.");
+			return detail::compute_transpose<C, R, T, Q, detail::is_aligned<Q>::value>::call(m);
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_determinant{};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_determinant<2, 2, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static T call(mat<2, 2, T, Q> const& m)
+		{
+			return m[0][0] * m[1][1] - m[1][0] * m[0][1];
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_determinant<3, 3, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static T call(mat<3, 3, T, Q> const& m)
+		{
+			return
+				+ m[0][0] * (m[1][1] * m[2][2] - m[2][1] * m[1][2])
+				- m[1][0] * (m[0][1] * m[2][2] - m[2][1] * m[0][2])
+				+ m[2][0] * (m[0][1] * m[1][2] - m[1][1] * m[0][2]);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_determinant<4, 4, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static T call(mat<4, 4, T, Q> const& m)
+		{
+			T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+			T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+			T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+			T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+			T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+			T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+
+			vec<4, T, Q> DetCof(
+				+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
+				- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
+				+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
+				- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
+
+			return
+				m[0][0] * DetCof[0] + m[0][1] * DetCof[1] +
+				m[0][2] * DetCof[2] + m[0][3] * DetCof[3];
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool IsFloat, bool Aligned>
+	struct compute_determinant_type{
+	
+		GLM_FUNC_QUALIFIER static T call(mat<C, R, T, Q> const& m)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_GENTYPE, 
+				"'determinant' only accept floating-point inputs, include <glm/ext/matrix_integer.hpp> to discard this restriction.");
+			return detail::compute_determinant<C, R, T, Q, detail::is_aligned<Q>::value>::call(m);
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_inverse{};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_inverse<2, 2, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<2, 2, T, Q> call(mat<2, 2, T, Q> const& m)
+		{
+			T OneOverDeterminant = static_cast<T>(1) / (
+				+ m[0][0] * m[1][1]
+				- m[1][0] * m[0][1]);
+
+			mat<2, 2, T, Q> Inverse(
+				+ m[1][1] * OneOverDeterminant,
+				- m[0][1] * OneOverDeterminant,
+				- m[1][0] * OneOverDeterminant,
+				+ m[0][0] * OneOverDeterminant);
+
+			return Inverse;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_inverse<3, 3, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<3, 3, T, Q> call(mat<3, 3, T, Q> const& m)
+		{
+			T OneOverDeterminant = static_cast<T>(1) / (
+				+ m[0][0] * (m[1][1] * m[2][2] - m[2][1] * m[1][2])
+				- m[1][0] * (m[0][1] * m[2][2] - m[2][1] * m[0][2])
+				+ m[2][0] * (m[0][1] * m[1][2] - m[1][1] * m[0][2]));
+
+			mat<3, 3, T, Q> Inverse;
+			Inverse[0][0] = + (m[1][1] * m[2][2] - m[2][1] * m[1][2]) * OneOverDeterminant;
+			Inverse[1][0] = - (m[1][0] * m[2][2] - m[2][0] * m[1][2]) * OneOverDeterminant;
+			Inverse[2][0] = + (m[1][0] * m[2][1] - m[2][0] * m[1][1]) * OneOverDeterminant;
+			Inverse[0][1] = - (m[0][1] * m[2][2] - m[2][1] * m[0][2]) * OneOverDeterminant;
+			Inverse[1][1] = + (m[0][0] * m[2][2] - m[2][0] * m[0][2]) * OneOverDeterminant;
+			Inverse[2][1] = - (m[0][0] * m[2][1] - m[2][0] * m[0][1]) * OneOverDeterminant;
+			Inverse[0][2] = + (m[0][1] * m[1][2] - m[1][1] * m[0][2]) * OneOverDeterminant;
+			Inverse[1][2] = - (m[0][0] * m[1][2] - m[1][0] * m[0][2]) * OneOverDeterminant;
+			Inverse[2][2] = + (m[0][0] * m[1][1] - m[1][0] * m[0][1]) * OneOverDeterminant;
+
+			return Inverse;
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_inverse<4, 4, T, Q, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 4, T, Q> call(mat<4, 4, T, Q> const& m)
+		{
+			T Coef00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+			T Coef02 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+			T Coef03 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+			T Coef04 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+			T Coef06 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+			T Coef07 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+			T Coef08 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+			T Coef10 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+			T Coef11 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+			T Coef12 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+			T Coef14 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+			T Coef15 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+			T Coef16 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+			T Coef18 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+			T Coef19 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+			T Coef20 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+			T Coef22 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+			T Coef23 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+			vec<4, T, Q> Fac0(Coef00, Coef00, Coef02, Coef03);
+			vec<4, T, Q> Fac1(Coef04, Coef04, Coef06, Coef07);
+			vec<4, T, Q> Fac2(Coef08, Coef08, Coef10, Coef11);
+			vec<4, T, Q> Fac3(Coef12, Coef12, Coef14, Coef15);
+			vec<4, T, Q> Fac4(Coef16, Coef16, Coef18, Coef19);
+			vec<4, T, Q> Fac5(Coef20, Coef20, Coef22, Coef23);
+
+			vec<4, T, Q> Vec0(m[1][0], m[0][0], m[0][0], m[0][0]);
+			vec<4, T, Q> Vec1(m[1][1], m[0][1], m[0][1], m[0][1]);
+			vec<4, T, Q> Vec2(m[1][2], m[0][2], m[0][2], m[0][2]);
+			vec<4, T, Q> Vec3(m[1][3], m[0][3], m[0][3], m[0][3]);
+
+			vec<4, T, Q> Inv0(Vec1 * Fac0 - Vec2 * Fac1 + Vec3 * Fac2);
+			vec<4, T, Q> Inv1(Vec0 * Fac0 - Vec2 * Fac3 + Vec3 * Fac4);
+			vec<4, T, Q> Inv2(Vec0 * Fac1 - Vec1 * Fac3 + Vec3 * Fac5);
+			vec<4, T, Q> Inv3(Vec0 * Fac2 - Vec1 * Fac4 + Vec2 * Fac5);
+
+			vec<4, T, Q> SignA(+1, -1, +1, -1);
+			vec<4, T, Q> SignB(-1, +1, -1, +1);
+			mat<4, 4, T, Q> Inverse(Inv0 * SignA, Inv1 * SignB, Inv2 * SignA, Inv3 * SignB);
+
+			vec<4, T, Q> Row0(Inverse[0][0], Inverse[1][0], Inverse[2][0], Inverse[3][0]);
+
+			vec<4, T, Q> Dot0(m[0] * Row0);
+			T Dot1 = (Dot0.x + Dot0.y) + (Dot0.z + Dot0.w);
+
+			T OneOverDeterminant = static_cast<T>(1) / Dot1;
+
+			return Inverse * OneOverDeterminant;
+		}
+	};
+}//namespace detail
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<C, R, T, Q> matrixCompMult(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y)
+	{
+		return detail::compute_matrixCompMult_type<C, R, T, Q, std::numeric_limits<T>::is_iec559, detail::is_aligned<Q>::value>::call(x, y);
+	}
+
+	template<length_t DA, length_t DB, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename detail::outerProduct_trait<DA, DB, T, Q>::type outerProduct(vec<DA, T, Q> const& c, vec<DB, T, Q> const& r)
+	{
+		return detail::compute_outerProduct_type<DA, DB, T, Q, std::numeric_limits<T>::is_iec559>::call(c, r);
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename mat<C, R, T, Q>::transpose_type transpose(mat<C, R, T, Q> const& m)
+	{
+		return detail::compute_transpose_type<C, R, T, Q, std::numeric_limits<T>::is_iec559, detail::is_aligned<Q>::value>::call(m);
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T determinant(mat<C, R, T, Q> const& m)
+	{
+		return detail::compute_determinant_type<C, R, T, Q, std::numeric_limits<T>::is_iec559, detail::is_aligned<Q>::value>::call(m);
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<C, R, T, Q> inverse(mat<C, R, T, Q> const& m)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_GENTYPE, "'inverse' only accept floating-point inputs");
+		return detail::compute_inverse<C, R, T, Q, detail::is_aligned<Q>::value>::call(m);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_matrix_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix_simd.inl
new file mode 100644
index 000000000000..b9bb4615de67
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_matrix_simd.inl
@@ -0,0 +1,252 @@
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+#include "type_mat4x4.hpp"
+#include "../geometric.hpp"
+#include "../simd/matrix.h"
+#include <cstring>
+
+namespace glm{
+namespace detail
+{
+#	if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+	template<qualifier Q>
+	struct compute_matrixCompMult<4, 4, float, Q, true>
+	{
+		GLM_STATIC_ASSERT(detail::is_aligned<Q>::value, "Specialization requires aligned");
+
+		GLM_FUNC_QUALIFIER static mat<4, 4, float, Q> call(mat<4, 4, float, Q> const& x, mat<4, 4, float, Q> const& y)
+		{
+			mat<4, 4, float, Q> Result;
+			glm_mat4_matrixCompMult(
+			        &x[0].data,
+			        &y[0].data,
+			        &Result[0].data);
+			return Result;
+		}
+	};
+#	endif
+
+	template<qualifier Q>
+	struct compute_transpose<4, 4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 4, float, Q> call(mat<4, 4, float, Q> const& m)
+		{
+			mat<4, 4, float, Q> Result;
+			glm_mat4_transpose(&m[0].data, &Result[0].data);
+			return Result;
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_determinant<4, 4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static float call(mat<4, 4, float, Q> const& m)
+		{
+			return _mm_cvtss_f32(glm_mat4_determinant(&m[0].data));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_inverse<4, 4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 4, float, Q> call(mat<4, 4, float, Q> const& m)
+		{
+			mat<4, 4, float, Q> Result;
+			glm_mat4_inverse(&m[0].data, &Result[0].data);
+			return Result;
+		}
+	};
+}//namespace detail
+
+#	if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+	template<>
+	GLM_FUNC_QUALIFIER mat<4, 4, float, aligned_lowp> outerProduct<4, 4, float, aligned_lowp>(vec<4, float, aligned_lowp> const& c, vec<4, float, aligned_lowp> const& r)
+	{
+		__m128 NativeResult[4];
+		glm_mat4_outerProduct(c.data, r.data, NativeResult);
+		mat<4, 4, float, aligned_lowp> Result;
+		std::memcpy(&Result[0], &NativeResult[0], sizeof(Result));
+		return Result;
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER mat<4, 4, float, aligned_mediump> outerProduct<4, 4, float, aligned_mediump>(vec<4, float, aligned_mediump> const& c, vec<4, float, aligned_mediump> const& r)
+	{
+		__m128 NativeResult[4];
+		glm_mat4_outerProduct(c.data, r.data, NativeResult);
+		mat<4, 4, float, aligned_mediump> Result;
+		std::memcpy(&Result[0], &NativeResult[0], sizeof(Result));
+		return Result;
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER mat<4, 4, float, aligned_highp> outerProduct<4, 4, float, aligned_highp>(vec<4, float, aligned_highp> const& c, vec<4, float, aligned_highp> const& r)
+	{
+		__m128 NativeResult[4];
+		glm_mat4_outerProduct(c.data, r.data, NativeResult);
+		mat<4, 4, float, aligned_highp> Result;
+		std::memcpy(&Result[0], &NativeResult[0], sizeof(Result));
+		return Result;
+	}
+#	endif
+}//namespace glm
+
+#elif GLM_ARCH & GLM_ARCH_NEON_BIT
+
+namespace glm {
+#if GLM_LANG & GLM_LANG_CXX11_FLAG
+	template <qualifier Q>
+	GLM_FUNC_QUALIFIER
+	typename std::enable_if<detail::is_aligned<Q>::value, mat<4, 4, float, Q>>::type
+	operator*(mat<4, 4, float, Q> const & m1, mat<4, 4, float, Q> const & m2)
+	{
+		auto MulRow = [&](int l) {
+			float32x4_t const SrcA = m2[l].data;
+
+			float32x4_t r = neon::mul_lane(m1[0].data, SrcA, 0);
+			r = neon::madd_lane(r, m1[1].data, SrcA, 1);
+			r = neon::madd_lane(r, m1[2].data, SrcA, 2);
+			r = neon::madd_lane(r, m1[3].data, SrcA, 3);
+
+			return r;
+		};
+
+		mat<4, 4, float, aligned_highp> Result;
+		Result[0].data = MulRow(0);
+		Result[1].data = MulRow(1);
+		Result[2].data = MulRow(2);
+		Result[3].data = MulRow(3);
+
+		return Result;
+	}
+#endif // CXX11
+
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_inverse<4, 4, float, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static mat<4, 4, float, Q> call(mat<4, 4, float, Q> const& m)
+		{
+			float32x4_t const& m0 = m[0].data;
+			float32x4_t const& m1 = m[1].data;
+			float32x4_t const& m2 = m[2].data;
+			float32x4_t const& m3 = m[3].data;
+
+			// m[2][2] * m[3][3] - m[3][2] * m[2][3];
+			// m[2][2] * m[3][3] - m[3][2] * m[2][3];
+			// m[1][2] * m[3][3] - m[3][2] * m[1][3];
+			// m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+			float32x4_t Fac0;
+			{
+				float32x4_t w0 = vcombine_f32(neon::dup_lane(m2, 2), neon::dup_lane(m1, 2));
+				float32x4_t w1 = neon::copy_lane(neon::dupq_lane(m3, 3), 3, m2, 3);
+				float32x4_t w2 = neon::copy_lane(neon::dupq_lane(m3, 2), 3, m2, 2);
+				float32x4_t w3 = vcombine_f32(neon::dup_lane(m2, 3), neon::dup_lane(m1, 3));
+				Fac0 = w0 * w1 -  w2 * w3;
+			}
+
+			// m[2][1] * m[3][3] - m[3][1] * m[2][3];
+			// m[2][1] * m[3][3] - m[3][1] * m[2][3];
+			// m[1][1] * m[3][3] - m[3][1] * m[1][3];
+			// m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+			float32x4_t Fac1;
+			{
+				float32x4_t w0 = vcombine_f32(neon::dup_lane(m2, 1), neon::dup_lane(m1, 1));
+				float32x4_t w1 = neon::copy_lane(neon::dupq_lane(m3, 3), 3, m2, 3);
+				float32x4_t w2 = neon::copy_lane(neon::dupq_lane(m3, 1), 3, m2, 1);
+				float32x4_t w3 = vcombine_f32(neon::dup_lane(m2, 3), neon::dup_lane(m1, 3));
+				Fac1 = w0 * w1 - w2 * w3;
+			}
+
+			// m[2][1] * m[3][2] - m[3][1] * m[2][2];
+			// m[2][1] * m[3][2] - m[3][1] * m[2][2];
+			// m[1][1] * m[3][2] - m[3][1] * m[1][2];
+			// m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+			float32x4_t Fac2;
+			{
+				float32x4_t w0 = vcombine_f32(neon::dup_lane(m2, 1), neon::dup_lane(m1, 1));
+				float32x4_t w1 = neon::copy_lane(neon::dupq_lane(m3, 2), 3, m2, 2);
+				float32x4_t w2 = neon::copy_lane(neon::dupq_lane(m3, 1), 3, m2, 1);
+				float32x4_t w3 = vcombine_f32(neon::dup_lane(m2, 2), neon::dup_lane(m1, 2));
+				Fac2 = w0 * w1 - w2 * w3;
+			}
+
+			// m[2][0] * m[3][3] - m[3][0] * m[2][3];
+			// m[2][0] * m[3][3] - m[3][0] * m[2][3];
+			// m[1][0] * m[3][3] - m[3][0] * m[1][3];
+			// m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+			float32x4_t Fac3;
+			{
+				float32x4_t w0 = vcombine_f32(neon::dup_lane(m2, 0), neon::dup_lane(m1, 0));
+				float32x4_t w1 = neon::copy_lane(neon::dupq_lane(m3, 3), 3, m2, 3);
+				float32x4_t w2 = neon::copy_lane(neon::dupq_lane(m3, 0), 3, m2, 0);
+				float32x4_t w3 = vcombine_f32(neon::dup_lane(m2, 3), neon::dup_lane(m1, 3));
+				Fac3 = w0 * w1 - w2 * w3;
+			}
+
+			// m[2][0] * m[3][2] - m[3][0] * m[2][2];
+			// m[2][0] * m[3][2] - m[3][0] * m[2][2];
+			// m[1][0] * m[3][2] - m[3][0] * m[1][2];
+			// m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+			float32x4_t Fac4;
+			{
+				float32x4_t w0 = vcombine_f32(neon::dup_lane(m2, 0), neon::dup_lane(m1, 0));
+				float32x4_t w1 = neon::copy_lane(neon::dupq_lane(m3, 2), 3, m2, 2);
+				float32x4_t w2 = neon::copy_lane(neon::dupq_lane(m3, 0), 3, m2, 0);
+				float32x4_t w3 = vcombine_f32(neon::dup_lane(m2, 2), neon::dup_lane(m1, 2));
+				Fac4 = w0 * w1 - w2 * w3;
+			}
+
+			// m[2][0] * m[3][1] - m[3][0] * m[2][1];
+			// m[2][0] * m[3][1] - m[3][0] * m[2][1];
+			// m[1][0] * m[3][1] - m[3][0] * m[1][1];
+			// m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+			float32x4_t Fac5;
+			{
+				float32x4_t w0 = vcombine_f32(neon::dup_lane(m2, 0), neon::dup_lane(m1, 0));
+				float32x4_t w1 = neon::copy_lane(neon::dupq_lane(m3, 1), 3, m2, 1);
+				float32x4_t w2 = neon::copy_lane(neon::dupq_lane(m3, 0), 3, m2, 0);
+				float32x4_t w3 = vcombine_f32(neon::dup_lane(m2, 1), neon::dup_lane(m1, 1));
+				Fac5 = w0 * w1 - w2 * w3;
+			}
+
+			float32x4_t Vec0 = neon::copy_lane(neon::dupq_lane(m0, 0), 0, m1, 0); // (m[1][0], m[0][0], m[0][0], m[0][0]);
+			float32x4_t Vec1 = neon::copy_lane(neon::dupq_lane(m0, 1), 0, m1, 1); // (m[1][1], m[0][1], m[0][1], m[0][1]);
+			float32x4_t Vec2 = neon::copy_lane(neon::dupq_lane(m0, 2), 0, m1, 2); // (m[1][2], m[0][2], m[0][2], m[0][2]);
+			float32x4_t Vec3 = neon::copy_lane(neon::dupq_lane(m0, 3), 0, m1, 3); // (m[1][3], m[0][3], m[0][3], m[0][3]);
+
+			float32x4_t Inv0 = Vec1 * Fac0 - Vec2 * Fac1 + Vec3 * Fac2;
+			float32x4_t Inv1 = Vec0 * Fac0 - Vec2 * Fac3 + Vec3 * Fac4;
+			float32x4_t Inv2 = Vec0 * Fac1 - Vec1 * Fac3 + Vec3 * Fac5;
+			float32x4_t Inv3 = Vec0 * Fac2 - Vec1 * Fac4 + Vec2 * Fac5;
+
+			float32x4_t r0 = float32x4_t{-1, +1, -1, +1} * Inv0;
+			float32x4_t r1 = float32x4_t{+1, -1, +1, -1} * Inv1;
+			float32x4_t r2 = float32x4_t{-1, +1, -1, +1} * Inv2;
+			float32x4_t r3 = float32x4_t{+1, -1, +1, -1} * Inv3;
+
+			float32x4_t det = neon::mul_lane(r0, m0, 0);
+			det = neon::madd_lane(det, r1, m0, 1);
+			det = neon::madd_lane(det, r2, m0, 2);
+			det = neon::madd_lane(det, r3, m0, 3);
+
+			float32x4_t rdet = vdupq_n_f32(1 / vgetq_lane_f32(det, 0));
+
+			mat<4, 4, float, Q> r;
+			r[0].data = vmulq_f32(r0, rdet);
+			r[1].data = vmulq_f32(r1, rdet);
+			r[2].data = vmulq_f32(r2, rdet);
+			r[3].data = vmulq_f32(r3, rdet);
+			return r;
+		}
+	};
+}//namespace detail
+}//namespace glm
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_packing.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_packing.inl
new file mode 100644
index 000000000000..234b093c081c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_packing.inl
@@ -0,0 +1,189 @@
+/// @ref core
+/// @file glm/detail/func_packing.inl
+
+#include "../common.hpp"
+#include "type_half.hpp"
+
+namespace glm
+{
+	GLM_FUNC_QUALIFIER uint packUnorm2x16(vec2 const& v)
+	{
+		union
+		{
+			unsigned short in[2];
+			uint out;
+		} u;
+
+		vec<2, unsigned short, defaultp> result(round(clamp(v, 0.0f, 1.0f) * 65535.0f));
+
+		u.in[0] = result[0];
+		u.in[1] = result[1];
+
+		return u.out;
+	}
+
+	GLM_FUNC_QUALIFIER vec2 unpackUnorm2x16(uint p)
+	{
+		union
+		{
+			uint in;
+			unsigned short out[2];
+		} u;
+
+		u.in = p;
+
+		return vec2(u.out[0], u.out[1]) * 1.5259021896696421759365224689097e-5f;
+	}
+
+	GLM_FUNC_QUALIFIER uint packSnorm2x16(vec2 const& v)
+	{
+		union
+		{
+			signed short in[2];
+			uint out;
+		} u;
+ 
+		vec<2, short, defaultp> result(round(clamp(v, -1.0f, 1.0f) * 32767.0f));
+
+		u.in[0] = result[0];
+		u.in[1] = result[1];
+
+		return u.out;
+	}
+
+	GLM_FUNC_QUALIFIER vec2 unpackSnorm2x16(uint p)
+	{
+		union
+		{
+			uint in;
+			signed short out[2];
+		} u;
+
+		u.in = p;
+
+		return clamp(vec2(u.out[0], u.out[1]) * 3.0518509475997192297128208258309e-5f, -1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER uint packUnorm4x8(vec4 const& v)
+	{
+		union
+		{
+			unsigned char in[4];
+			uint out;
+		} u;
+
+		vec<4, unsigned char, defaultp> result(round(clamp(v, 0.0f, 1.0f) * 255.0f));
+
+		u.in[0] = result[0];
+		u.in[1] = result[1];
+		u.in[2] = result[2];
+		u.in[3] = result[3];
+
+		return u.out;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackUnorm4x8(uint p)
+	{
+		union
+		{
+			uint in;
+			unsigned char out[4];
+		} u;
+
+		u.in = p;
+
+		return vec4(u.out[0], u.out[1], u.out[2], u.out[3]) * 0.0039215686274509803921568627451f;
+	}
+
+	GLM_FUNC_QUALIFIER uint packSnorm4x8(vec4 const& v)
+	{
+		union
+		{
+			signed char in[4];
+			uint out;
+		} u;
+
+		vec<4, signed char, defaultp> result(round(clamp(v, -1.0f, 1.0f) * 127.0f));
+
+		u.in[0] = result[0];
+		u.in[1] = result[1];
+		u.in[2] = result[2];
+		u.in[3] = result[3];
+
+		return u.out;
+	}
+
+	GLM_FUNC_QUALIFIER glm::vec4 unpackSnorm4x8(uint p)
+	{
+		union
+		{
+			uint in;
+			signed char out[4];
+		} u;
+
+		u.in = p;
+
+		return clamp(vec4(u.out[0], u.out[1], u.out[2], u.out[3]) * 0.0078740157480315f, -1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER double packDouble2x32(uvec2 const& v)
+	{
+		union
+		{
+			uint   in[2];
+			double out;
+		} u;
+
+		u.in[0] = v[0];
+		u.in[1] = v[1];
+
+		return u.out;
+	}
+
+	GLM_FUNC_QUALIFIER uvec2 unpackDouble2x32(double v)
+	{
+		union
+		{
+			double in;
+			uint   out[2];
+		} u;
+
+		u.in = v;
+
+		return uvec2(u.out[0], u.out[1]);
+	}
+
+	GLM_FUNC_QUALIFIER uint packHalf2x16(vec2 const& v)
+	{
+		union
+		{
+			signed short in[2];
+			uint out;
+		} u;
+
+		u.in[0] = detail::toFloat16(v.x);
+		u.in[1] = detail::toFloat16(v.y);
+
+		return u.out;
+	}
+
+	GLM_FUNC_QUALIFIER vec2 unpackHalf2x16(uint v)
+	{
+		union
+		{
+			uint in;
+			signed short out[2];
+		} u;
+
+		u.in = v;
+
+		return vec2(
+			detail::toFloat32(u.out[0]),
+			detail::toFloat32(u.out[1]));
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_packing_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_packing_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_packing_simd.inl
new file mode 100644
index 000000000000..fd0fe8b7d9b4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_packing_simd.inl
@@ -0,0 +1,6 @@
+namespace glm{
+namespace detail
+{
+
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric.inl
new file mode 100644
index 000000000000..9e6d9cfb1ce7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric.inl
@@ -0,0 +1,197 @@
+#include "_vectorize.hpp"
+#include <cmath>
+#include <limits>
+
+namespace glm
+{
+	// radians
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType radians(genType degrees)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'radians' only accept floating-point input");
+
+		return degrees * static_cast<genType>(0.01745329251994329576923690768489);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> radians(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(radians, v);
+	}
+
+	// degrees
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType degrees(genType radians)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'degrees' only accept floating-point input");
+
+		return radians * static_cast<genType>(57.295779513082320876798154814105);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> degrees(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(degrees, v);
+	}
+
+	// sin
+	using ::std::sin;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> sin(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(sin, v);
+	}
+
+	// cos
+	using std::cos;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> cos(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(cos, v);
+	}
+
+	// tan
+	using std::tan;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> tan(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(tan, v);
+	}
+
+	// asin
+	using std::asin;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> asin(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(asin, v);
+	}
+
+	// acos
+	using std::acos;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> acos(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(acos, v);
+	}
+
+	// atan
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType atan(genType y, genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'atan' only accept floating-point input");
+
+		return ::std::atan2(y, x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> atan(vec<L, T, Q> const& y, vec<L, T, Q> const& x)
+	{
+		return detail::functor2<vec, L, T, Q>::call(::std::atan2, y, x);
+	}
+
+	using std::atan;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> atan(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(atan, v);
+	}
+
+	// sinh
+	using std::sinh;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> sinh(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(sinh, v);
+	}
+
+	// cosh
+	using std::cosh;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> cosh(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(cosh, v);
+	}
+
+	// tanh
+	using std::tanh;
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> tanh(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(tanh, v);
+	}
+
+	// asinh
+#	if GLM_HAS_CXX11_STL
+		using std::asinh;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER genType asinh(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'asinh' only accept floating-point input");
+
+			return (x < static_cast<genType>(0) ? static_cast<genType>(-1) : (x > static_cast<genType>(0) ? static_cast<genType>(1) : static_cast<genType>(0))) * log(std::abs(x) + sqrt(static_cast<genType>(1) + x * x));
+		}
+#	endif
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> asinh(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(asinh, v);
+	}
+
+	// acosh
+#	if GLM_HAS_CXX11_STL
+		using std::acosh;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER genType acosh(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acosh' only accept floating-point input");
+
+			if(x < static_cast<genType>(1))
+				return static_cast<genType>(0);
+			return log(x + sqrt(x * x - static_cast<genType>(1)));
+		}
+#	endif
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> acosh(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(acosh, v);
+	}
+
+	// atanh
+#	if GLM_HAS_CXX11_STL
+		using std::atanh;
+#	else
+		template<typename genType>
+		GLM_FUNC_QUALIFIER genType atanh(genType x)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'atanh' only accept floating-point input");
+
+			if(std::abs(x) >= static_cast<genType>(1))
+				return 0;
+			return static_cast<genType>(0.5) * log((static_cast<genType>(1) + x) / (static_cast<genType>(1) - x));
+		}
+#	endif
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> atanh(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(atanh, v);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_trigonometric_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_trigonometric_simd.inl
new file mode 100644
index 000000000000..e69de29bb2d1
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational.inl
new file mode 100644
index 000000000000..80c9e87fcb97
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational.inl
@@ -0,0 +1,87 @@
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> lessThan(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = x[i] < y[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> lessThanEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = x[i] <= y[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> greaterThan(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = x[i] > y[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> greaterThanEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = x[i] >= y[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = x[i] == y[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = x[i] != y[i];
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool any(vec<L, bool, Q> const& v)
+	{
+		bool Result = false;
+		for(length_t i = 0; i < L; ++i)
+			Result = Result || v[i];
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool all(vec<L, bool, Q> const& v)
+	{
+		bool Result = true;
+		for(length_t i = 0; i < L; ++i)
+			Result = Result && v[i];
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> not_(vec<L, bool, Q> const& v)
+	{
+		vec<L, bool, Q> Result(true);
+		for(length_t i = 0; i < L; ++i)
+			Result[i] = !v[i];
+		return Result;
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "func_vector_relational_simd.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational_simd.inl
new file mode 100644
index 000000000000..fd0fe8b7d9b4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/func_vector_relational_simd.inl
@@ -0,0 +1,6 @@
+namespace glm{
+namespace detail
+{
+
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/glm.cpp b/thirdparty/manifold/thirdparty/glm/glm/detail/glm.cpp
new file mode 100644
index 000000000000..e0755bd65d46
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/glm.cpp
@@ -0,0 +1,263 @@
+/// @ref core
+/// @file glm/glm.cpp
+
+#ifndef GLM_ENABLE_EXPERIMENTAL
+#define GLM_ENABLE_EXPERIMENTAL
+#endif
+#include <glm/gtx/dual_quaternion.hpp>
+#include <glm/gtc/vec1.hpp>
+#include <glm/gtc/quaternion.hpp>
+#include <glm/ext/scalar_int_sized.hpp>
+#include <glm/ext/scalar_uint_sized.hpp>
+#include <glm/glm.hpp>
+
+namespace glm
+{
+// tvec1 type explicit instantiation
+template struct vec<1, uint8, lowp>;
+template struct vec<1, uint16, lowp>;
+template struct vec<1, uint32, lowp>;
+template struct vec<1, uint64, lowp>;
+template struct vec<1, int8, lowp>;
+template struct vec<1, int16, lowp>;
+template struct vec<1, int32, lowp>;
+template struct vec<1, int64, lowp>;
+template struct vec<1, float32, lowp>;
+template struct vec<1, float64, lowp>;
+
+template struct vec<1, uint8, mediump>;
+template struct vec<1, uint16, mediump>;
+template struct vec<1, uint32, mediump>;
+template struct vec<1, uint64, mediump>;
+template struct vec<1, int8, mediump>;
+template struct vec<1, int16, mediump>;
+template struct vec<1, int32, mediump>;
+template struct vec<1, int64, mediump>;
+template struct vec<1, float32, mediump>;
+template struct vec<1, float64, mediump>;
+
+template struct vec<1, uint8, highp>;
+template struct vec<1, uint16, highp>;
+template struct vec<1, uint32, highp>;
+template struct vec<1, uint64, highp>;
+template struct vec<1, int8, highp>;
+template struct vec<1, int16, highp>;
+template struct vec<1, int32, highp>;
+template struct vec<1, int64, highp>;
+template struct vec<1, float32, highp>;
+template struct vec<1, float64, highp>;
+
+// tvec2 type explicit instantiation
+template struct vec<2, uint8, lowp>;
+template struct vec<2, uint16, lowp>;
+template struct vec<2, uint32, lowp>;
+template struct vec<2, uint64, lowp>;
+template struct vec<2, int8, lowp>;
+template struct vec<2, int16, lowp>;
+template struct vec<2, int32, lowp>;
+template struct vec<2, int64, lowp>;
+template struct vec<2, float32, lowp>;
+template struct vec<2, float64, lowp>;
+
+template struct vec<2, uint8, mediump>;
+template struct vec<2, uint16, mediump>;
+template struct vec<2, uint32, mediump>;
+template struct vec<2, uint64, mediump>;
+template struct vec<2, int8, mediump>;
+template struct vec<2, int16, mediump>;
+template struct vec<2, int32, mediump>;
+template struct vec<2, int64, mediump>;
+template struct vec<2, float32, mediump>;
+template struct vec<2, float64, mediump>;
+
+template struct vec<2, uint8, highp>;
+template struct vec<2, uint16, highp>;
+template struct vec<2, uint32, highp>;
+template struct vec<2, uint64, highp>;
+template struct vec<2, int8, highp>;
+template struct vec<2, int16, highp>;
+template struct vec<2, int32, highp>;
+template struct vec<2, int64, highp>;
+template struct vec<2, float32, highp>;
+template struct vec<2, float64, highp>;
+
+// tvec3 type explicit instantiation
+template struct vec<3, uint8, lowp>;
+template struct vec<3, uint16, lowp>;
+template struct vec<3, uint32, lowp>;
+template struct vec<3, uint64, lowp>;
+template struct vec<3, int8, lowp>;
+template struct vec<3, int16, lowp>;
+template struct vec<3, int32, lowp>;
+template struct vec<3, int64, lowp>;
+template struct vec<3, float32, lowp>;
+template struct vec<3, float64, lowp>;
+
+template struct vec<3, uint8, mediump>;
+template struct vec<3, uint16, mediump>;
+template struct vec<3, uint32, mediump>;
+template struct vec<3, uint64, mediump>;
+template struct vec<3, int8, mediump>;
+template struct vec<3, int16, mediump>;
+template struct vec<3, int32, mediump>;
+template struct vec<3, int64, mediump>;
+template struct vec<3, float32, mediump>;
+template struct vec<3, float64, mediump>;
+
+template struct vec<3, uint8, highp>;
+template struct vec<3, uint16, highp>;
+template struct vec<3, uint32, highp>;
+template struct vec<3, uint64, highp>;
+template struct vec<3, int8, highp>;
+template struct vec<3, int16, highp>;
+template struct vec<3, int32, highp>;
+template struct vec<3, int64, highp>;
+template struct vec<3, float32, highp>;
+template struct vec<3, float64, highp>;
+
+// tvec4 type explicit instantiation
+template struct vec<4, uint8, lowp>;
+template struct vec<4, uint16, lowp>;
+template struct vec<4, uint32, lowp>;
+template struct vec<4, uint64, lowp>;
+template struct vec<4, int8, lowp>;
+template struct vec<4, int16, lowp>;
+template struct vec<4, int32, lowp>;
+template struct vec<4, int64, lowp>;
+template struct vec<4, float32, lowp>;
+template struct vec<4, float64, lowp>;
+
+template struct vec<4, uint8, mediump>;
+template struct vec<4, uint16, mediump>;
+template struct vec<4, uint32, mediump>;
+template struct vec<4, uint64, mediump>;
+template struct vec<4, int8, mediump>;
+template struct vec<4, int16, mediump>;
+template struct vec<4, int32, mediump>;
+template struct vec<4, int64, mediump>;
+template struct vec<4, float32, mediump>;
+template struct vec<4, float64, mediump>;
+
+template struct vec<4, uint8, highp>;
+template struct vec<4, uint16, highp>;
+template struct vec<4, uint32, highp>;
+template struct vec<4, uint64, highp>;
+template struct vec<4, int8, highp>;
+template struct vec<4, int16, highp>;
+template struct vec<4, int32, highp>;
+template struct vec<4, int64, highp>;
+template struct vec<4, float32, highp>;
+template struct vec<4, float64, highp>;
+
+// tmat2x2 type explicit instantiation
+template struct mat<2, 2, float32, lowp>;
+template struct mat<2, 2, float64, lowp>;
+
+template struct mat<2, 2, float32, mediump>;
+template struct mat<2, 2, float64, mediump>;
+
+template struct mat<2, 2, float32, highp>;
+template struct mat<2, 2, float64, highp>;
+
+// tmat2x3 type explicit instantiation
+template struct mat<2, 3, float32, lowp>;
+template struct mat<2, 3, float64, lowp>;
+
+template struct mat<2, 3, float32, mediump>;
+template struct mat<2, 3, float64, mediump>;
+
+template struct mat<2, 3, float32, highp>;
+template struct mat<2, 3, float64, highp>;
+
+// tmat2x4 type explicit instantiation
+template struct mat<2, 4, float32, lowp>;
+template struct mat<2, 4, float64, lowp>;
+
+template struct mat<2, 4, float32, mediump>;
+template struct mat<2, 4, float64, mediump>;
+
+template struct mat<2, 4, float32, highp>;
+template struct mat<2, 4, float64, highp>;
+
+// tmat3x2 type explicit instantiation
+template struct mat<3, 2, float32, lowp>;
+template struct mat<3, 2, float64, lowp>;
+
+template struct mat<3, 2, float32, mediump>;
+template struct mat<3, 2, float64, mediump>;
+
+template struct mat<3, 2, float32, highp>;
+template struct mat<3, 2, float64, highp>;
+
+// tmat3x3 type explicit instantiation
+template struct mat<3, 3, float32, lowp>;
+template struct mat<3, 3, float64, lowp>;
+
+template struct mat<3, 3, float32, mediump>;
+template struct mat<3, 3, float64, mediump>;
+
+template struct mat<3, 3, float32, highp>;
+template struct mat<3, 3, float64, highp>;
+
+// tmat3x4 type explicit instantiation
+template struct mat<3, 4, float32, lowp>;
+template struct mat<3, 4, float64, lowp>;
+
+template struct mat<3, 4, float32, mediump>;
+template struct mat<3, 4, float64, mediump>;
+
+template struct mat<3, 4, float32, highp>;
+template struct mat<3, 4, float64, highp>;
+
+// tmat4x2 type explicit instantiation
+template struct mat<4, 2, float32, lowp>;
+template struct mat<4, 2, float64, lowp>;
+
+template struct mat<4, 2, float32, mediump>;
+template struct mat<4, 2, float64, mediump>;
+
+template struct mat<4, 2, float32, highp>;
+template struct mat<4, 2, float64, highp>;
+
+// tmat4x3 type explicit instantiation
+template struct mat<4, 3, float32, lowp>;
+template struct mat<4, 3, float64, lowp>;
+
+template struct mat<4, 3, float32, mediump>;
+template struct mat<4, 3, float64, mediump>;
+
+template struct mat<4, 3, float32, highp>;
+template struct mat<4, 3, float64, highp>;
+
+// tmat4x4 type explicit instantiation
+template struct mat<4, 4, float32, lowp>;
+template struct mat<4, 4, float64, lowp>;
+
+template struct mat<4, 4, float32, mediump>;
+template struct mat<4, 4, float64, mediump>;
+
+template struct mat<4, 4, float32, highp>;
+template struct mat<4, 4, float64, highp>;
+
+// tquat type explicit instantiation
+template struct qua<float32, lowp>;
+template struct qua<float64, lowp>;
+
+template struct qua<float32, mediump>;
+template struct qua<float64, mediump>;
+
+template struct qua<float32, highp>;
+template struct qua<float64, highp>;
+
+//tdualquat type explicit instantiation
+template struct tdualquat<float32, lowp>;
+template struct tdualquat<float64, lowp>;
+
+template struct tdualquat<float32, mediump>;
+template struct tdualquat<float64, mediump>;
+
+template struct tdualquat<float32, highp>;
+template struct tdualquat<float64, highp>;
+
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/qualifier.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/qualifier.hpp
new file mode 100644
index 000000000000..a6c96cca5c2e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/qualifier.hpp
@@ -0,0 +1,229 @@
+#pragma once
+
+#include "setup.hpp"
+
+namespace glm
+{
+	/// Qualify GLM types in term of alignment (packed, aligned) and precision in term of ULPs (lowp, mediump, highp)
+	enum qualifier
+	{
+		packed_highp, ///< Typed data is tightly packed in memory and operations are executed with high precision in term of ULPs
+		packed_mediump, ///< Typed data is tightly packed in memory  and operations are executed with medium precision in term of ULPs for higher performance
+		packed_lowp, ///< Typed data is tightly packed in memory  and operations are executed with low precision in term of ULPs to maximize performance
+
+#		if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+			aligned_highp, ///< Typed data is aligned in memory allowing SIMD optimizations and operations are executed with high precision in term of ULPs
+			aligned_mediump, ///< Typed data is aligned in memory allowing SIMD optimizations and operations are executed with high precision in term of ULPs for higher performance
+			aligned_lowp, // ///< Typed data is aligned in memory allowing SIMD optimizations and operations are executed with high precision in term of ULPs to maximize performance
+			aligned = aligned_highp, ///< By default aligned qualifier is also high precision
+#		endif
+
+		highp = packed_highp, ///< By default highp qualifier is also packed
+		mediump = packed_mediump, ///< By default mediump qualifier is also packed
+		lowp = packed_lowp, ///< By default lowp qualifier is also packed
+		packed = packed_highp, ///< By default packed qualifier is also high precision
+
+#		if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE && defined(GLM_FORCE_DEFAULT_ALIGNED_GENTYPES)
+			defaultp = aligned_highp
+#		else
+			defaultp = highp
+#		endif
+	};
+
+	typedef qualifier precision;
+
+	template<length_t L, typename T, qualifier Q = defaultp> struct vec;
+	template<length_t C, length_t R, typename T, qualifier Q = defaultp> struct mat;
+	template<typename T, qualifier Q = defaultp> struct qua;
+
+#	if GLM_HAS_TEMPLATE_ALIASES
+		template <typename T, qualifier Q = defaultp> using tvec1 = vec<1, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tvec2 = vec<2, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tvec3 = vec<3, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tvec4 = vec<4, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat2x2 = mat<2, 2, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat2x3 = mat<2, 3, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat2x4 = mat<2, 4, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat3x2 = mat<3, 2, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat3x3 = mat<3, 3, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat3x4 = mat<3, 4, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat4x2 = mat<4, 2, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat4x3 = mat<4, 3, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tmat4x4 = mat<4, 4, T, Q>;
+		template <typename T, qualifier Q = defaultp> using tquat = qua<T, Q>;
+#	endif
+
+namespace detail
+{
+	template<glm::qualifier P>
+	struct is_aligned
+	{
+		static const bool value = false;
+	};
+
+#	if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+		template<>
+		struct is_aligned<glm::aligned_lowp>
+		{
+			static const bool value = true;
+		};
+
+		template<>
+		struct is_aligned<glm::aligned_mediump>
+		{
+			static const bool value = true;
+		};
+
+		template<>
+		struct is_aligned<glm::aligned_highp>
+		{
+			static const bool value = true;
+		};
+#	endif
+
+	template<length_t L, typename T, bool is_aligned>
+	struct storage
+	{
+		typedef struct type {
+			T data[L];
+		} type;
+	};
+
+#	if GLM_HAS_ALIGNOF
+		template<length_t L, typename T>
+		struct storage<L, T, true>
+		{
+			typedef struct alignas(L * sizeof(T)) type {
+				T data[L];
+			} type;
+		};
+
+		template<typename T>
+		struct storage<3, T, true>
+		{
+			typedef struct alignas(4 * sizeof(T)) type {
+				T data[4];
+			} type;
+		};
+#	endif
+
+#	if GLM_ARCH & GLM_ARCH_SSE2_BIT
+	template<>
+	struct storage<4, float, true>
+	{
+		typedef glm_f32vec4 type;
+	};
+
+	template<>
+	struct storage<4, int, true>
+	{
+		typedef glm_i32vec4 type;
+	};
+
+	template<>
+	struct storage<4, unsigned int, true>
+	{
+		typedef glm_u32vec4 type;
+	};
+
+	template<>
+	struct storage<2, double, true>
+	{
+		typedef glm_f64vec2 type;
+	};
+
+	template<>
+	struct storage<2, detail::int64, true>
+	{
+		typedef glm_i64vec2 type;
+	};
+
+	template<>
+	struct storage<2, detail::uint64, true>
+	{
+		typedef glm_u64vec2 type;
+	};
+#	endif
+#	if (GLM_ARCH & GLM_ARCH_AVX_BIT)
+	template<>
+	struct storage<4, double, true>
+	{
+		typedef glm_f64vec4 type;
+	};
+#	endif
+
+#	if (GLM_ARCH & GLM_ARCH_AVX2_BIT)
+	template<>
+	struct storage<4, detail::int64, true>
+	{
+		typedef glm_i64vec4 type;
+	};
+
+	template<>
+	struct storage<4, detail::uint64, true>
+	{
+		typedef glm_u64vec4 type;
+	};
+#	endif
+
+#	if GLM_ARCH & GLM_ARCH_NEON_BIT
+	template<>
+	struct storage<4, float, true>
+	{
+		typedef glm_f32vec4 type;
+	};
+
+	template<>
+	struct storage<4, int, true>
+	{
+		typedef glm_i32vec4 type;
+	};
+
+	template<>
+	struct storage<4, unsigned int, true>
+	{
+		typedef glm_u32vec4 type;
+	};
+#	endif
+
+	enum genTypeEnum
+	{
+		GENTYPE_VEC,
+		GENTYPE_MAT,
+		GENTYPE_QUAT
+	};
+
+	template <typename genType>
+	struct genTypeTrait
+	{};
+
+	template <length_t C, length_t R, typename T>
+	struct genTypeTrait<mat<C, R, T> >
+	{
+		static const genTypeEnum GENTYPE = GENTYPE_MAT;
+	};
+
+	template<typename genType, genTypeEnum type>
+	struct init_gentype
+	{
+	};
+
+	template<typename genType>
+	struct init_gentype<genType, GENTYPE_QUAT>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static genType identity()
+		{
+			return genType(1, 0, 0, 0);
+		}
+	};
+
+	template<typename genType>
+	struct init_gentype<genType, GENTYPE_MAT>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static genType identity()
+		{
+			return genType(1);
+		}
+	};
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/setup.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/setup.hpp
new file mode 100644
index 000000000000..19953bcb4e57
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/setup.hpp
@@ -0,0 +1,1167 @@
+#ifndef GLM_SETUP_INCLUDED
+
+#include <cassert>
+#include <cstddef>
+
+#define GLM_VERSION_MAJOR 0
+#define GLM_VERSION_MINOR 9
+#define GLM_VERSION_PATCH 9
+#define GLM_VERSION_REVISION 9
+#define GLM_VERSION 999
+#define GLM_VERSION_MESSAGE "GLM: version 0.9.9.9"
+
+#define GLM_SETUP_INCLUDED GLM_VERSION
+
+///////////////////////////////////////////////////////////////////////////////////
+// Active states
+
+#define GLM_DISABLE		0
+#define GLM_ENABLE		1
+
+///////////////////////////////////////////////////////////////////////////////////
+// Messages
+
+#if defined(GLM_FORCE_MESSAGES)
+#	define GLM_MESSAGES GLM_ENABLE
+#else
+#	define GLM_MESSAGES GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Detect the platform
+
+#include "../simd/platform.h"
+
+///////////////////////////////////////////////////////////////////////////////////
+// Build model
+
+#if defined(_M_ARM64) || defined(__LP64__) || defined(_M_X64) || defined(__ppc64__) || defined(__x86_64__)
+#	define GLM_MODEL	GLM_MODEL_64
+#elif defined(__i386__) || defined(__ppc__) || defined(__ILP32__) || defined(_M_ARM)
+#	define GLM_MODEL	GLM_MODEL_32
+#else
+#	define GLM_MODEL	GLM_MODEL_32
+#endif//
+
+#if !defined(GLM_MODEL) && GLM_COMPILER != 0
+#	error "GLM_MODEL undefined, your compiler may not be supported by GLM. Add #define GLM_MODEL 0 to ignore this message."
+#endif//GLM_MODEL
+
+///////////////////////////////////////////////////////////////////////////////////
+// C++ Version
+
+// User defines: GLM_FORCE_CXX98, GLM_FORCE_CXX03, GLM_FORCE_CXX11, GLM_FORCE_CXX14, GLM_FORCE_CXX17, GLM_FORCE_CXX2A
+
+#define GLM_LANG_CXX98_FLAG			(1 << 1)
+#define GLM_LANG_CXX03_FLAG			(1 << 2)
+#define GLM_LANG_CXX0X_FLAG			(1 << 3)
+#define GLM_LANG_CXX11_FLAG			(1 << 4)
+#define GLM_LANG_CXX14_FLAG			(1 << 5)
+#define GLM_LANG_CXX17_FLAG			(1 << 6)
+#define GLM_LANG_CXX20_FLAG			(1 << 7)
+#define GLM_LANG_CXXMS_FLAG			(1 << 8)
+#define GLM_LANG_CXXGNU_FLAG		(1 << 9)
+
+#define GLM_LANG_CXX98			GLM_LANG_CXX98_FLAG
+#define GLM_LANG_CXX03			(GLM_LANG_CXX98 | GLM_LANG_CXX03_FLAG)
+#define GLM_LANG_CXX0X			(GLM_LANG_CXX03 | GLM_LANG_CXX0X_FLAG)
+#define GLM_LANG_CXX11			(GLM_LANG_CXX0X | GLM_LANG_CXX11_FLAG)
+#define GLM_LANG_CXX14			(GLM_LANG_CXX11 | GLM_LANG_CXX14_FLAG)
+#define GLM_LANG_CXX17			(GLM_LANG_CXX14 | GLM_LANG_CXX17_FLAG)
+#define GLM_LANG_CXX20			(GLM_LANG_CXX17 | GLM_LANG_CXX20_FLAG)
+#define GLM_LANG_CXXMS			GLM_LANG_CXXMS_FLAG
+#define GLM_LANG_CXXGNU			GLM_LANG_CXXGNU_FLAG
+
+#if (defined(_MSC_EXTENSIONS))
+#	define GLM_LANG_EXT GLM_LANG_CXXMS_FLAG
+#elif ((GLM_COMPILER & (GLM_COMPILER_CLANG | GLM_COMPILER_GCC)) && (GLM_ARCH & GLM_ARCH_SIMD_BIT))
+#	define GLM_LANG_EXT GLM_LANG_CXXMS_FLAG
+#else
+#	define GLM_LANG_EXT 0
+#endif
+
+#if (defined(GLM_FORCE_CXX_UNKNOWN))
+#	define GLM_LANG 0
+#elif defined(GLM_FORCE_CXX20)
+#	define GLM_LANG (GLM_LANG_CXX20 | GLM_LANG_EXT)
+#	define GLM_LANG_STL11_FORCED
+#elif defined(GLM_FORCE_CXX17)
+#	define GLM_LANG (GLM_LANG_CXX17 | GLM_LANG_EXT)
+#	define GLM_LANG_STL11_FORCED
+#elif defined(GLM_FORCE_CXX14)
+#	define GLM_LANG (GLM_LANG_CXX14 | GLM_LANG_EXT)
+#	define GLM_LANG_STL11_FORCED
+#elif defined(GLM_FORCE_CXX11)
+#	define GLM_LANG (GLM_LANG_CXX11 | GLM_LANG_EXT)
+#	define GLM_LANG_STL11_FORCED
+#elif defined(GLM_FORCE_CXX03)
+#	define GLM_LANG (GLM_LANG_CXX03 | GLM_LANG_EXT)
+#elif defined(GLM_FORCE_CXX98)
+#	define GLM_LANG (GLM_LANG_CXX98 | GLM_LANG_EXT)
+#else
+#	if GLM_COMPILER & GLM_COMPILER_VC && defined(_MSVC_LANG)
+#		if GLM_COMPILER >= GLM_COMPILER_VC15_7
+#			define GLM_LANG_PLATFORM _MSVC_LANG
+#		elif GLM_COMPILER >= GLM_COMPILER_VC15
+#			if _MSVC_LANG > 201402L
+#				define GLM_LANG_PLATFORM 201402L
+#			else
+#				define GLM_LANG_PLATFORM _MSVC_LANG
+#			endif
+#		else
+#			define GLM_LANG_PLATFORM 0
+#		endif
+#	else
+#		define GLM_LANG_PLATFORM 0
+#	endif
+
+#	if __cplusplus > 201703L || GLM_LANG_PLATFORM > 201703L
+#		define GLM_LANG (GLM_LANG_CXX20 | GLM_LANG_EXT)
+#	elif __cplusplus == 201703L || GLM_LANG_PLATFORM == 201703L
+#		define GLM_LANG (GLM_LANG_CXX17 | GLM_LANG_EXT)
+#	elif __cplusplus == 201402L || __cplusplus == 201406L || __cplusplus == 201500L || GLM_LANG_PLATFORM == 201402L
+#		define GLM_LANG (GLM_LANG_CXX14 | GLM_LANG_EXT)
+#	elif __cplusplus == 201103L || GLM_LANG_PLATFORM == 201103L
+#		define GLM_LANG (GLM_LANG_CXX11 | GLM_LANG_EXT)
+#	elif defined(__INTEL_CXX11_MODE__) || defined(_MSC_VER) || defined(__GXX_EXPERIMENTAL_CXX0X__)
+#		define GLM_LANG (GLM_LANG_CXX0X | GLM_LANG_EXT)
+#	elif __cplusplus == 199711L
+#		define GLM_LANG (GLM_LANG_CXX98 | GLM_LANG_EXT)
+#	else
+#		define GLM_LANG (0 | GLM_LANG_EXT)
+#	endif
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Has of C++ features
+
+// http://clang.llvm.org/cxx_status.html
+// http://gcc.gnu.org/projects/cxx0x.html
+// http://msdn.microsoft.com/en-us/library/vstudio/hh567368(v=vs.120).aspx
+
+// Android has multiple STLs but C++11 STL detection doesn't always work #284 #564
+#if GLM_PLATFORM == GLM_PLATFORM_ANDROID && !defined(GLM_LANG_STL11_FORCED)
+#	define GLM_HAS_CXX11_STL 0
+#elif (GLM_COMPILER & GLM_COMPILER_CUDA_RTC) == GLM_COMPILER_CUDA_RTC
+#	define GLM_HAS_CXX11_STL 0
+#elif (GLM_COMPILER & GLM_COMPILER_HIP)
+#	define GLM_HAS_CXX11_STL 0
+#elif GLM_COMPILER & GLM_COMPILER_CLANG
+#	if (defined(_LIBCPP_VERSION) || (GLM_LANG & GLM_LANG_CXX11_FLAG) || defined(GLM_LANG_STL11_FORCED))
+#		define GLM_HAS_CXX11_STL 1
+#	else
+#		define GLM_HAS_CXX11_STL 0
+#	endif
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_CXX11_STL 1
+#else
+#	define GLM_HAS_CXX11_STL ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_GCC) && (GLM_COMPILER >= GLM_COMPILER_GCC48)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC12)) || \
+		((GLM_PLATFORM != GLM_PLATFORM_WINDOWS) && (GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_COMPILER >= GLM_COMPILER_INTEL15))))
+#endif
+
+// N1720
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_STATIC_ASSERT __has_feature(cxx_static_assert)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_STATIC_ASSERT 1
+#else
+#	define GLM_HAS_STATIC_ASSERT ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N1988
+#if GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_EXTENDED_INTEGER_TYPE 1
+#else
+#	define GLM_HAS_EXTENDED_INTEGER_TYPE (\
+		((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (GLM_COMPILER & GLM_COMPILER_VC)) || \
+		((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (GLM_COMPILER & GLM_COMPILER_CLANG)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP)))
+#endif
+
+// N2672 Initializer lists http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2672.htm
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_INITIALIZER_LISTS __has_feature(cxx_generalized_initializers)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_INITIALIZER_LISTS 1
+#else
+#	define GLM_HAS_INITIALIZER_LISTS ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC15)) || \
+		((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_COMPILER >= GLM_COMPILER_INTEL14)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N2544 Unrestricted unions http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2544.pdf
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_UNRESTRICTED_UNIONS __has_feature(cxx_unrestricted_unions)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_UNRESTRICTED_UNIONS 1
+#else
+#	define GLM_HAS_UNRESTRICTED_UNIONS (GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		(GLM_COMPILER & GLM_COMPILER_VC) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP)))
+#endif
+
+// N2346
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_DEFAULTED_FUNCTIONS __has_feature(cxx_defaulted_functions)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_DEFAULTED_FUNCTIONS 1
+#else
+#	define GLM_HAS_DEFAULTED_FUNCTIONS ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC12)) || \
+		((GLM_COMPILER & GLM_COMPILER_INTEL)) || \
+		(GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP)))
+#endif
+
+// N2118
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_RVALUE_REFERENCES __has_feature(cxx_rvalue_references)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_RVALUE_REFERENCES 1
+#else
+#	define GLM_HAS_RVALUE_REFERENCES ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_VC)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N2437 http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2437.pdf
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_EXPLICIT_CONVERSION_OPERATORS __has_feature(cxx_explicit_conversions)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_EXPLICIT_CONVERSION_OPERATORS 1
+#else
+#	define GLM_HAS_EXPLICIT_CONVERSION_OPERATORS ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_COMPILER >= GLM_COMPILER_INTEL14)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC12)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N2258 http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2258.pdf
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_TEMPLATE_ALIASES __has_feature(cxx_alias_templates)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_TEMPLATE_ALIASES 1
+#else
+#	define GLM_HAS_TEMPLATE_ALIASES ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_INTEL)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC12)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N2930 http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_RANGE_FOR __has_feature(cxx_range_for)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_RANGE_FOR 1
+#else
+#	define GLM_HAS_RANGE_FOR ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_INTEL)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N2341 http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2341.pdf
+#if GLM_COMPILER & GLM_COMPILER_CLANG
+#	define GLM_HAS_ALIGNOF __has_feature(cxx_alignas)
+#elif GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_ALIGNOF 1
+#else
+#	define GLM_HAS_ALIGNOF ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_COMPILER >= GLM_COMPILER_INTEL15)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC14)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+// N2235 Generalized Constant Expressions http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2235.pdf
+// N3652 Extended Constant Expressions http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3652.html
+#if (GLM_ARCH & GLM_ARCH_SIMD_BIT) // Compiler SIMD intrinsics don't support constexpr...
+#	define GLM_HAS_CONSTEXPR 0
+#elif (GLM_COMPILER & GLM_COMPILER_CLANG)
+#	define GLM_HAS_CONSTEXPR __has_feature(cxx_relaxed_constexpr)
+#elif (GLM_LANG & GLM_LANG_CXX14_FLAG)
+#	define GLM_HAS_CONSTEXPR 1
+#else
+#	define GLM_HAS_CONSTEXPR ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && GLM_HAS_INITIALIZER_LISTS && (\
+		((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_COMPILER >= GLM_COMPILER_INTEL17)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC15))))
+#endif
+
+#if GLM_HAS_CONSTEXPR
+#	define GLM_CONSTEXPR constexpr
+#else
+#	define GLM_CONSTEXPR
+#endif
+
+//
+#if GLM_HAS_CONSTEXPR
+# if (GLM_COMPILER & GLM_COMPILER_CLANG)
+#	if __has_feature(cxx_if_constexpr)
+#		define GLM_HAS_IF_CONSTEXPR 1
+#	else
+# 		define GLM_HAS_IF_CONSTEXPR 0
+#	endif
+# elif (GLM_LANG & GLM_LANG_CXX17_FLAG)
+# 	define GLM_HAS_IF_CONSTEXPR 1
+# else
+# 	define GLM_HAS_IF_CONSTEXPR 0
+# endif
+#else
+#	define GLM_HAS_IF_CONSTEXPR 0
+#endif
+
+#if GLM_HAS_IF_CONSTEXPR
+# 	define GLM_IF_CONSTEXPR if constexpr
+#else
+#	define GLM_IF_CONSTEXPR if
+#endif
+
+//
+#if GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_ASSIGNABLE 1
+#else
+#	define GLM_HAS_ASSIGNABLE ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC15)) || \
+		((GLM_COMPILER & GLM_COMPILER_GCC) && (GLM_COMPILER >= GLM_COMPILER_GCC49))))
+#endif
+
+//
+#define GLM_HAS_TRIVIAL_QUERIES 0
+
+//
+#if GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_MAKE_SIGNED 1
+#else
+#	define GLM_HAS_MAKE_SIGNED ((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC12)) || \
+		((GLM_COMPILER & GLM_COMPILER_CUDA)) || \
+		((GLM_COMPILER & GLM_COMPILER_HIP))))
+#endif
+
+//
+#if defined(GLM_FORCE_INTRINSICS)
+#	define GLM_HAS_BITSCAN_WINDOWS ((GLM_PLATFORM & GLM_PLATFORM_WINDOWS) && (\
+		((GLM_COMPILER & GLM_COMPILER_INTEL)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC14) && (GLM_ARCH & GLM_ARCH_X86_BIT))))
+#else
+#	define GLM_HAS_BITSCAN_WINDOWS 0
+#endif
+
+#if GLM_LANG & GLM_LANG_CXX11_FLAG
+#	define GLM_HAS_NOEXCEPT 1
+#else
+#	define GLM_HAS_NOEXCEPT 0
+#endif
+
+#if GLM_HAS_NOEXCEPT
+#	define GLM_NOEXCEPT noexcept
+#else
+#	define GLM_NOEXCEPT
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// OpenMP
+#ifdef _OPENMP
+#	if GLM_COMPILER & GLM_COMPILER_GCC
+#		if GLM_COMPILER >= GLM_COMPILER_GCC61
+#			define GLM_HAS_OPENMP 45
+#		elif GLM_COMPILER >= GLM_COMPILER_GCC49
+#			define GLM_HAS_OPENMP 40
+#		elif GLM_COMPILER >= GLM_COMPILER_GCC47
+#			define GLM_HAS_OPENMP 31
+#		else
+#			define GLM_HAS_OPENMP 0
+#		endif
+#	elif GLM_COMPILER & GLM_COMPILER_CLANG
+#		if GLM_COMPILER >= GLM_COMPILER_CLANG38
+#			define GLM_HAS_OPENMP 31
+#		else
+#			define GLM_HAS_OPENMP 0
+#		endif
+#	elif GLM_COMPILER & GLM_COMPILER_VC
+#		define GLM_HAS_OPENMP 20
+#	elif GLM_COMPILER & GLM_COMPILER_INTEL
+#		if GLM_COMPILER >= GLM_COMPILER_INTEL16
+#			define GLM_HAS_OPENMP 40
+#		else
+#			define GLM_HAS_OPENMP 0
+#		endif
+#	else
+#		define GLM_HAS_OPENMP 0
+#	endif
+#else
+#	define GLM_HAS_OPENMP 0
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// nullptr
+
+#if GLM_LANG & GLM_LANG_CXX0X_FLAG
+#	define GLM_CONFIG_NULLPTR GLM_ENABLE
+#else
+#	define GLM_CONFIG_NULLPTR GLM_DISABLE
+#endif
+
+#if GLM_CONFIG_NULLPTR == GLM_ENABLE
+#	define GLM_NULLPTR nullptr
+#else
+#	define GLM_NULLPTR 0
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Static assert
+
+#if GLM_HAS_STATIC_ASSERT
+#	define GLM_STATIC_ASSERT(x, message) static_assert(x, message)
+#elif GLM_COMPILER & GLM_COMPILER_VC
+#	define GLM_STATIC_ASSERT(x, message) typedef char __CASSERT__##__LINE__[(x) ? 1 : -1]
+#else
+#	define GLM_STATIC_ASSERT(x, message) assert(x)
+#endif//GLM_LANG
+
+///////////////////////////////////////////////////////////////////////////////////
+// Qualifiers
+
+// User defines: GLM_CUDA_FORCE_DEVICE_FUNC, GLM_CUDA_FORCE_HOST_FUNC
+
+#if (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+#	if defined(GLM_CUDA_FORCE_DEVICE_FUNC) && defined(GLM_CUDA_FORCE_HOST_FUNC)
+#		error "GLM error: GLM_CUDA_FORCE_DEVICE_FUNC and GLM_CUDA_FORCE_HOST_FUNC should not be defined at the same time, GLM by default generates both device and host code for CUDA compiler."
+#	endif//defined(GLM_CUDA_FORCE_DEVICE_FUNC) && defined(GLM_CUDA_FORCE_HOST_FUNC)
+
+#	if defined(GLM_CUDA_FORCE_DEVICE_FUNC)
+#		define GLM_CUDA_FUNC_DEF __device__
+#		define GLM_CUDA_FUNC_DECL __device__
+#	elif defined(GLM_CUDA_FORCE_HOST_FUNC)
+#		define GLM_CUDA_FUNC_DEF __host__
+#		define GLM_CUDA_FUNC_DECL __host__
+#	else
+#		define GLM_CUDA_FUNC_DEF __device__ __host__
+#		define GLM_CUDA_FUNC_DECL __device__ __host__
+#	endif//defined(GLM_CUDA_FORCE_XXXX_FUNC)
+#else
+#	define GLM_CUDA_FUNC_DEF
+#	define GLM_CUDA_FUNC_DECL
+#endif
+
+#if defined(GLM_FORCE_INLINE)
+#	if GLM_COMPILER & GLM_COMPILER_VC
+#		define GLM_INLINE __forceinline
+#		define GLM_NEVER_INLINE __declspec(noinline)
+#	elif GLM_COMPILER & (GLM_COMPILER_GCC | GLM_COMPILER_CLANG)
+#		define GLM_INLINE inline __attribute__((__always_inline__))
+#		define GLM_NEVER_INLINE __attribute__((__noinline__))
+#	elif (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+#		define GLM_INLINE __forceinline__
+#		define GLM_NEVER_INLINE __noinline__
+#	else
+#		define GLM_INLINE inline
+#		define GLM_NEVER_INLINE
+#	endif//GLM_COMPILER
+#else
+#	define GLM_INLINE inline
+#	define GLM_NEVER_INLINE
+#endif//defined(GLM_FORCE_INLINE)
+
+#define GLM_FUNC_DECL GLM_CUDA_FUNC_DECL
+#define GLM_FUNC_QUALIFIER GLM_CUDA_FUNC_DEF GLM_INLINE
+
+// Do not use CUDA function qualifiers on CUDA compiler when functions are made default
+#if GLM_HAS_DEFAULTED_FUNCTIONS
+#	define GLM_DEFAULTED_FUNC_DECL
+#	define GLM_DEFAULTED_FUNC_QUALIFIER GLM_INLINE
+#else
+#	define GLM_DEFAULTED_FUNC_DECL GLM_FUNC_DECL
+#	define GLM_DEFAULTED_FUNC_QUALIFIER GLM_FUNC_QUALIFIER
+#endif//GLM_HAS_DEFAULTED_FUNCTIONS
+#if !defined(GLM_FORCE_CTOR_INIT)
+#	define GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_DEFAULTED_FUNC_DECL
+#	define GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_DEFAULTED_FUNC_QUALIFIER
+#else
+#	define GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_FUNC_DECL
+#	define GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_FUNC_QUALIFIER
+#endif//GLM_FORCE_CTOR_INIT
+
+///////////////////////////////////////////////////////////////////////////////////
+// Swizzle operators
+
+// User defines: GLM_FORCE_SWIZZLE
+
+#define GLM_SWIZZLE_DISABLED		0
+#define GLM_SWIZZLE_OPERATOR		1
+#define GLM_SWIZZLE_FUNCTION		2
+
+#if defined(GLM_SWIZZLE)
+#	pragma message("GLM: GLM_SWIZZLE is deprecated, use GLM_FORCE_SWIZZLE instead.")
+#	define GLM_FORCE_SWIZZLE
+#endif
+
+#if defined(GLM_FORCE_SWIZZLE) && (GLM_LANG & GLM_LANG_CXXMS_FLAG) && !defined(GLM_FORCE_XYZW_ONLY)
+#	define GLM_CONFIG_SWIZZLE GLM_SWIZZLE_OPERATOR
+#elif defined(GLM_FORCE_SWIZZLE)
+#	define GLM_CONFIG_SWIZZLE GLM_SWIZZLE_FUNCTION
+#else
+#	define GLM_CONFIG_SWIZZLE GLM_SWIZZLE_DISABLED
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Allows using not basic types as genType
+
+// #define GLM_FORCE_UNRESTRICTED_GENTYPE
+
+#ifdef GLM_FORCE_UNRESTRICTED_GENTYPE
+#	define GLM_CONFIG_UNRESTRICTED_GENTYPE GLM_ENABLE
+#else
+#	define GLM_CONFIG_UNRESTRICTED_GENTYPE GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Allows using any scaler as float
+
+// #define GLM_FORCE_UNRESTRICTED_FLOAT
+
+#ifdef GLM_FORCE_UNRESTRICTED_FLOAT
+#	define GLM_CONFIG_UNRESTRICTED_FLOAT GLM_ENABLE
+#else
+#	define GLM_CONFIG_UNRESTRICTED_FLOAT GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Clip control, define GLM_FORCE_DEPTH_ZERO_TO_ONE before including GLM
+// to use a clip space between 0 to 1.
+// Coordinate system, define GLM_FORCE_LEFT_HANDED before including GLM
+// to use left handed coordinate system by default.
+
+#define GLM_CLIP_CONTROL_ZO_BIT		(1 << 0) // ZERO_TO_ONE
+#define GLM_CLIP_CONTROL_NO_BIT		(1 << 1) // NEGATIVE_ONE_TO_ONE
+#define GLM_CLIP_CONTROL_LH_BIT		(1 << 2) // LEFT_HANDED, For DirectX, Metal, Vulkan
+#define GLM_CLIP_CONTROL_RH_BIT		(1 << 3) // RIGHT_HANDED, For OpenGL, default in GLM
+
+#define GLM_CLIP_CONTROL_LH_ZO (GLM_CLIP_CONTROL_LH_BIT | GLM_CLIP_CONTROL_ZO_BIT)
+#define GLM_CLIP_CONTROL_LH_NO (GLM_CLIP_CONTROL_LH_BIT | GLM_CLIP_CONTROL_NO_BIT)
+#define GLM_CLIP_CONTROL_RH_ZO (GLM_CLIP_CONTROL_RH_BIT | GLM_CLIP_CONTROL_ZO_BIT)
+#define GLM_CLIP_CONTROL_RH_NO (GLM_CLIP_CONTROL_RH_BIT | GLM_CLIP_CONTROL_NO_BIT)
+
+#ifdef GLM_FORCE_DEPTH_ZERO_TO_ONE
+#	ifdef GLM_FORCE_LEFT_HANDED
+#		define GLM_CONFIG_CLIP_CONTROL GLM_CLIP_CONTROL_LH_ZO
+#	else
+#		define GLM_CONFIG_CLIP_CONTROL GLM_CLIP_CONTROL_RH_ZO
+#	endif
+#else
+#	ifdef GLM_FORCE_LEFT_HANDED
+#		define GLM_CONFIG_CLIP_CONTROL GLM_CLIP_CONTROL_LH_NO
+#	else
+#		define GLM_CONFIG_CLIP_CONTROL GLM_CLIP_CONTROL_RH_NO
+#	endif
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Qualifiers
+
+#if (GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS))
+#	define GLM_DEPRECATED __declspec(deprecated)
+#	define GLM_ALIGNED_TYPEDEF(type, name, alignment) typedef __declspec(align(alignment)) type name
+#elif GLM_COMPILER & (GLM_COMPILER_GCC | GLM_COMPILER_CLANG | GLM_COMPILER_INTEL)
+#	define GLM_DEPRECATED __attribute__((__deprecated__))
+#	define GLM_ALIGNED_TYPEDEF(type, name, alignment) typedef type name __attribute__((aligned(alignment)))
+#elif (GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP)
+#	define GLM_DEPRECATED
+#	define GLM_ALIGNED_TYPEDEF(type, name, alignment) typedef type name __align__(x)
+#else
+#	define GLM_DEPRECATED
+#	define GLM_ALIGNED_TYPEDEF(type, name, alignment) typedef type name
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+
+#ifdef GLM_FORCE_EXPLICIT_CTOR
+#	define GLM_EXPLICIT explicit
+#else
+#	define GLM_EXPLICIT
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Length type: all length functions returns a length_t type.
+// When GLM_FORCE_SIZE_T_LENGTH is defined, length_t is a typedef of size_t otherwise
+// length_t is a typedef of int like GLSL defines it.
+
+#define GLM_LENGTH_INT		1
+#define GLM_LENGTH_SIZE_T	2
+
+#ifdef GLM_FORCE_SIZE_T_LENGTH
+#	define GLM_CONFIG_LENGTH_TYPE		GLM_LENGTH_SIZE_T
+#else
+#	define GLM_CONFIG_LENGTH_TYPE		GLM_LENGTH_INT
+#endif
+
+namespace glm
+{
+	using std::size_t;
+#	if GLM_CONFIG_LENGTH_TYPE == GLM_LENGTH_SIZE_T
+		typedef size_t length_t;
+#	else
+		typedef int length_t;
+#	endif
+}//namespace glm
+
+///////////////////////////////////////////////////////////////////////////////////
+// constexpr
+
+#if GLM_HAS_CONSTEXPR
+#	define GLM_CONFIG_CONSTEXP GLM_ENABLE
+
+	namespace glm
+	{
+		template<typename T, std::size_t N>
+		constexpr std::size_t countof(T const (&)[N])
+		{
+			return N;
+		}
+	}//namespace glm
+#	define GLM_COUNTOF(arr) glm::countof(arr)
+#elif defined(_MSC_VER)
+#	define GLM_CONFIG_CONSTEXP GLM_DISABLE
+
+#	define GLM_COUNTOF(arr) _countof(arr)
+#else
+#	define GLM_CONFIG_CONSTEXP GLM_DISABLE
+
+#	define GLM_COUNTOF(arr) sizeof(arr) / sizeof(arr[0])
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// uint
+
+namespace glm{
+namespace detail
+{
+	template<typename T>
+	struct is_int
+	{
+		enum test {value = 0};
+	};
+
+	template<>
+	struct is_int<unsigned int>
+	{
+		enum test {value = ~0};
+	};
+
+	template<>
+	struct is_int<signed int>
+	{
+		enum test {value = ~0};
+	};
+}//namespace detail
+
+	typedef unsigned int	uint;
+}//namespace glm
+
+///////////////////////////////////////////////////////////////////////////////////
+// 64-bit int
+
+#if GLM_HAS_EXTENDED_INTEGER_TYPE
+#	include <cstdint>
+#endif
+
+namespace glm{
+namespace detail
+{
+#	if GLM_HAS_EXTENDED_INTEGER_TYPE
+		typedef std::uint64_t						uint64;
+		typedef std::int64_t						int64;
+#	elif (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) // C99 detected, 64 bit types available
+		typedef uint64_t							uint64;
+		typedef int64_t								int64;
+#	elif GLM_COMPILER & GLM_COMPILER_VC
+		typedef unsigned __int64					uint64;
+		typedef signed __int64						int64;
+#	elif GLM_COMPILER & GLM_COMPILER_GCC
+#		pragma GCC diagnostic ignored "-Wlong-long"
+		__extension__ typedef unsigned long long	uint64;
+		__extension__ typedef signed long long		int64;
+#	elif (GLM_COMPILER & GLM_COMPILER_CLANG)
+#		pragma clang diagnostic ignored "-Wc++11-long-long"
+		typedef unsigned long long					uint64;
+		typedef signed long long					int64;
+#	else//unknown compiler
+		typedef unsigned long long					uint64;
+		typedef signed long long					int64;
+#	endif
+}//namespace detail
+}//namespace glm
+
+///////////////////////////////////////////////////////////////////////////////////
+// make_unsigned
+
+#if GLM_HAS_MAKE_SIGNED
+#	include <type_traits>
+
+namespace glm{
+namespace detail
+{
+	using std::make_unsigned;
+}//namespace detail
+}//namespace glm
+
+#else
+
+namespace glm{
+namespace detail
+{
+	template<typename genType>
+	struct make_unsigned
+	{};
+
+	template<>
+	struct make_unsigned<char>
+	{
+		typedef unsigned char type;
+	};
+
+	template<>
+	struct make_unsigned<signed char>
+	{
+		typedef unsigned char type;
+	};
+
+	template<>
+	struct make_unsigned<short>
+	{
+		typedef unsigned short type;
+	};
+
+	template<>
+	struct make_unsigned<int>
+	{
+		typedef unsigned int type;
+	};
+
+	template<>
+	struct make_unsigned<long>
+	{
+		typedef unsigned long type;
+	};
+
+	template<>
+	struct make_unsigned<int64>
+	{
+		typedef uint64 type;
+	};
+
+	template<>
+	struct make_unsigned<unsigned char>
+	{
+		typedef unsigned char type;
+	};
+
+	template<>
+	struct make_unsigned<unsigned short>
+	{
+		typedef unsigned short type;
+	};
+
+	template<>
+	struct make_unsigned<unsigned int>
+	{
+		typedef unsigned int type;
+	};
+
+	template<>
+	struct make_unsigned<unsigned long>
+	{
+		typedef unsigned long type;
+	};
+
+	template<>
+	struct make_unsigned<uint64>
+	{
+		typedef uint64 type;
+	};
+}//namespace detail
+}//namespace glm
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Only use x, y, z, w as vector type components
+
+#ifdef GLM_FORCE_XYZW_ONLY
+#	define GLM_CONFIG_XYZW_ONLY GLM_ENABLE
+#else
+#	define GLM_CONFIG_XYZW_ONLY GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Configure the use of defaulted initialized types
+
+#define GLM_CTOR_INIT_DISABLE		0
+#define GLM_CTOR_INITIALIZER_LIST	1
+#define GLM_CTOR_INITIALISATION		2
+
+#if defined(GLM_FORCE_CTOR_INIT) && GLM_HAS_INITIALIZER_LISTS
+#	define GLM_CONFIG_CTOR_INIT GLM_CTOR_INITIALIZER_LIST
+#elif defined(GLM_FORCE_CTOR_INIT) && !GLM_HAS_INITIALIZER_LISTS
+#	define GLM_CONFIG_CTOR_INIT GLM_CTOR_INITIALISATION
+#else
+#	define GLM_CONFIG_CTOR_INIT GLM_CTOR_INIT_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Use SIMD instruction sets
+
+#if GLM_HAS_ALIGNOF && (GLM_LANG & GLM_LANG_CXXMS_FLAG) && (GLM_ARCH & GLM_ARCH_SIMD_BIT)
+#	define GLM_CONFIG_SIMD GLM_ENABLE
+#else
+#	define GLM_CONFIG_SIMD GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Configure the use of defaulted function
+
+#if GLM_HAS_DEFAULTED_FUNCTIONS
+#	define GLM_CONFIG_DEFAULTED_FUNCTIONS GLM_ENABLE
+#	define GLM_DEFAULT = default
+#else
+#	define GLM_CONFIG_DEFAULTED_FUNCTIONS GLM_DISABLE
+#	define GLM_DEFAULT
+#endif
+
+#if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INIT_DISABLE && GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_ENABLE
+#	define GLM_CONFIG_DEFAULTED_DEFAULT_CTOR GLM_ENABLE
+#	define GLM_DEFAULT_CTOR GLM_DEFAULT
+#else
+#	define GLM_CONFIG_DEFAULTED_DEFAULT_CTOR GLM_DISABLE
+#	define GLM_DEFAULT_CTOR
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Configure the use of aligned gentypes
+
+#ifdef GLM_FORCE_ALIGNED // Legacy define
+#	define GLM_FORCE_DEFAULT_ALIGNED_GENTYPES
+#endif
+
+#ifdef GLM_FORCE_DEFAULT_ALIGNED_GENTYPES
+#	define GLM_FORCE_ALIGNED_GENTYPES
+#endif
+
+#if GLM_HAS_ALIGNOF && (GLM_LANG & GLM_LANG_CXXMS_FLAG) && (defined(GLM_FORCE_ALIGNED_GENTYPES) || (GLM_CONFIG_SIMD == GLM_ENABLE))
+#	define GLM_CONFIG_ALIGNED_GENTYPES GLM_ENABLE
+#else
+#	define GLM_CONFIG_ALIGNED_GENTYPES GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Configure the use of anonymous structure as implementation detail
+
+#if ((GLM_CONFIG_SIMD == GLM_ENABLE) || (GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR) || (GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE))
+#	define GLM_CONFIG_ANONYMOUS_STRUCT GLM_ENABLE
+#else
+#	define GLM_CONFIG_ANONYMOUS_STRUCT GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Silent warnings
+
+#ifdef GLM_FORCE_SILENT_WARNINGS
+#	define GLM_SILENT_WARNINGS GLM_ENABLE
+#else
+#	define GLM_SILENT_WARNINGS GLM_DISABLE
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Precision
+
+#define GLM_HIGHP		1
+#define GLM_MEDIUMP		2
+#define GLM_LOWP		3
+
+#if defined(GLM_FORCE_PRECISION_HIGHP_BOOL) || defined(GLM_PRECISION_HIGHP_BOOL)
+#	define GLM_CONFIG_PRECISION_BOOL		GLM_HIGHP
+#elif defined(GLM_FORCE_PRECISION_MEDIUMP_BOOL) || defined(GLM_PRECISION_MEDIUMP_BOOL)
+#	define GLM_CONFIG_PRECISION_BOOL		GLM_MEDIUMP
+#elif defined(GLM_FORCE_PRECISION_LOWP_BOOL) || defined(GLM_PRECISION_LOWP_BOOL)
+#	define GLM_CONFIG_PRECISION_BOOL		GLM_LOWP
+#else
+#	define GLM_CONFIG_PRECISION_BOOL		GLM_HIGHP
+#endif
+
+#if defined(GLM_FORCE_PRECISION_HIGHP_INT) || defined(GLM_PRECISION_HIGHP_INT)
+#	define GLM_CONFIG_PRECISION_INT			GLM_HIGHP
+#elif defined(GLM_FORCE_PRECISION_MEDIUMP_INT) || defined(GLM_PRECISION_MEDIUMP_INT)
+#	define GLM_CONFIG_PRECISION_INT			GLM_MEDIUMP
+#elif defined(GLM_FORCE_PRECISION_LOWP_INT) || defined(GLM_PRECISION_LOWP_INT)
+#	define GLM_CONFIG_PRECISION_INT			GLM_LOWP
+#else
+#	define GLM_CONFIG_PRECISION_INT			GLM_HIGHP
+#endif
+
+#if defined(GLM_FORCE_PRECISION_HIGHP_UINT) || defined(GLM_PRECISION_HIGHP_UINT)
+#	define GLM_CONFIG_PRECISION_UINT		GLM_HIGHP
+#elif defined(GLM_FORCE_PRECISION_MEDIUMP_UINT) || defined(GLM_PRECISION_MEDIUMP_UINT)
+#	define GLM_CONFIG_PRECISION_UINT		GLM_MEDIUMP
+#elif defined(GLM_FORCE_PRECISION_LOWP_UINT) || defined(GLM_PRECISION_LOWP_UINT)
+#	define GLM_CONFIG_PRECISION_UINT		GLM_LOWP
+#else
+#	define GLM_CONFIG_PRECISION_UINT		GLM_HIGHP
+#endif
+
+#if defined(GLM_FORCE_PRECISION_HIGHP_FLOAT) || defined(GLM_PRECISION_HIGHP_FLOAT)
+#	define GLM_CONFIG_PRECISION_FLOAT		GLM_HIGHP
+#elif defined(GLM_FORCE_PRECISION_MEDIUMP_FLOAT) || defined(GLM_PRECISION_MEDIUMP_FLOAT)
+#	define GLM_CONFIG_PRECISION_FLOAT		GLM_MEDIUMP
+#elif defined(GLM_FORCE_PRECISION_LOWP_FLOAT) || defined(GLM_PRECISION_LOWP_FLOAT)
+#	define GLM_CONFIG_PRECISION_FLOAT		GLM_LOWP
+#else
+#	define GLM_CONFIG_PRECISION_FLOAT		GLM_HIGHP
+#endif
+
+#if defined(GLM_FORCE_PRECISION_HIGHP_DOUBLE) || defined(GLM_PRECISION_HIGHP_DOUBLE)
+#	define GLM_CONFIG_PRECISION_DOUBLE		GLM_HIGHP
+#elif defined(GLM_FORCE_PRECISION_MEDIUMP_DOUBLE) || defined(GLM_PRECISION_MEDIUMP_DOUBLE)
+#	define GLM_CONFIG_PRECISION_DOUBLE		GLM_MEDIUMP
+#elif defined(GLM_FORCE_PRECISION_LOWP_DOUBLE) || defined(GLM_PRECISION_LOWP_DOUBLE)
+#	define GLM_CONFIG_PRECISION_DOUBLE		GLM_LOWP
+#else
+#	define GLM_CONFIG_PRECISION_DOUBLE		GLM_HIGHP
+#endif
+
+///////////////////////////////////////////////////////////////////////////////////
+// Check inclusions of different versions of GLM
+
+#elif ((GLM_SETUP_INCLUDED != GLM_VERSION) && !defined(GLM_FORCE_IGNORE_VERSION))
+#	error "GLM error: A different version of GLM is already included. Define GLM_FORCE_IGNORE_VERSION before including GLM headers to ignore this error."
+#elif GLM_SETUP_INCLUDED == GLM_VERSION
+
+///////////////////////////////////////////////////////////////////////////////////
+// Messages
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_MESSAGE_DISPLAYED)
+#	define GLM_MESSAGE_DISPLAYED
+#		define GLM_STR_HELPER(x) #x
+#		define GLM_STR(x) GLM_STR_HELPER(x)
+
+	// Report GLM version
+#		pragma message (GLM_STR(GLM_VERSION_MESSAGE))
+
+	// Report C++ language
+#	if (GLM_LANG & GLM_LANG_CXX20_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 20 with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX20_FLAG)
+#		pragma message("GLM: C++ 2A")
+#	elif (GLM_LANG & GLM_LANG_CXX17_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 17 with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX17_FLAG)
+#		pragma message("GLM: C++ 17")
+#	elif (GLM_LANG & GLM_LANG_CXX14_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 14 with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX14_FLAG)
+#		pragma message("GLM: C++ 14")
+#	elif (GLM_LANG & GLM_LANG_CXX11_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 11 with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX11_FLAG)
+#		pragma message("GLM: C++ 11")
+#	elif (GLM_LANG & GLM_LANG_CXX0X_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 0x with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX0X_FLAG)
+#		pragma message("GLM: C++ 0x")
+#	elif (GLM_LANG & GLM_LANG_CXX03_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 03 with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX03_FLAG)
+#		pragma message("GLM: C++ 03")
+#	elif (GLM_LANG & GLM_LANG_CXX98_FLAG) && (GLM_LANG & GLM_LANG_EXT)
+#		pragma message("GLM: C++ 98 with extensions")
+#	elif (GLM_LANG & GLM_LANG_CXX98_FLAG)
+#		pragma message("GLM: C++ 98")
+#	else
+#		pragma message("GLM: C++ language undetected")
+#	endif//GLM_LANG
+
+	// Report compiler detection
+#	if GLM_COMPILER & GLM_COMPILER_CUDA
+#		pragma message("GLM: CUDA compiler detected")
+#	elif GLM_COMPILER & GLM_COMPILER_HIP
+#		pragma message("GLM: HIP compiler detected")
+#	elif GLM_COMPILER & GLM_COMPILER_VC
+#		pragma message("GLM: Visual C++ compiler detected")
+#	elif GLM_COMPILER & GLM_COMPILER_CLANG
+#		pragma message("GLM: Clang compiler detected")
+#	elif GLM_COMPILER & GLM_COMPILER_INTEL
+#		pragma message("GLM: Intel Compiler detected")
+#	elif GLM_COMPILER & GLM_COMPILER_GCC
+#		pragma message("GLM: GCC compiler detected")
+#	else
+#		pragma message("GLM: Compiler not detected")
+#	endif
+
+	// Report build target
+#	if (GLM_ARCH & GLM_ARCH_AVX2_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with AVX2 instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_AVX2_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with AVX2 instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_AVX_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with AVX instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_AVX_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with AVX instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_SSE42_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with SSE4.2 instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_SSE42_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with SSE4.2 instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_SSE41_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with SSE4.1 instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_SSE41_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with SSE4.1 instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_SSSE3_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with SSSE3 instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_SSSE3_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with SSSE3 instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_SSE3_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with SSE3 instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_SSE3_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with SSE3 instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_SSE2_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits with SSE2 instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_SSE2_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits with SSE2 instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_X86_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: x86 64 bits build target")
+#	elif (GLM_ARCH & GLM_ARCH_X86_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: x86 32 bits build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_NEON_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: ARM 64 bits with Neon instruction set build target")
+#	elif (GLM_ARCH & GLM_ARCH_NEON_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: ARM 32 bits with Neon instruction set build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_ARM_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: ARM 64 bits build target")
+#	elif (GLM_ARCH & GLM_ARCH_ARM_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: ARM 32 bits build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_MIPS_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: MIPS 64 bits build target")
+#	elif (GLM_ARCH & GLM_ARCH_MIPS_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: MIPS 32 bits build target")
+
+#	elif (GLM_ARCH & GLM_ARCH_PPC_BIT) && (GLM_MODEL == GLM_MODEL_64)
+#		pragma message("GLM: PowerPC 64 bits build target")
+#	elif (GLM_ARCH & GLM_ARCH_PPC_BIT) && (GLM_MODEL == GLM_MODEL_32)
+#		pragma message("GLM: PowerPC 32 bits build target")
+#	else
+#		pragma message("GLM: Unknown build target")
+#	endif//GLM_ARCH
+
+	// Report platform name
+#	if(GLM_PLATFORM & GLM_PLATFORM_QNXNTO)
+#		pragma message("GLM: QNX platform detected")
+//#	elif(GLM_PLATFORM & GLM_PLATFORM_IOS)
+//#		pragma message("GLM: iOS platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_APPLE)
+#		pragma message("GLM: Apple platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_WINCE)
+#		pragma message("GLM: WinCE platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_WINDOWS)
+#		pragma message("GLM: Windows platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_CHROME_NACL)
+#		pragma message("GLM: Native Client detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+#		pragma message("GLM: Android platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_LINUX)
+#		pragma message("GLM: Linux platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_UNIX)
+#		pragma message("GLM: UNIX platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_UNKNOWN)
+#		pragma message("GLM: platform unknown")
+#	else
+#		pragma message("GLM: platform not detected")
+#	endif
+
+	// Report whether only xyzw component are used
+#	if defined GLM_FORCE_XYZW_ONLY
+#		pragma message("GLM: GLM_FORCE_XYZW_ONLY is defined. Only x, y, z and w component are available in vector type. This define disables swizzle operators and SIMD instruction sets.")
+#	endif
+
+	// Report swizzle operator support
+#	if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+#		pragma message("GLM: GLM_FORCE_SWIZZLE is defined, swizzling operators enabled.")
+#	elif GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+#		pragma message("GLM: GLM_FORCE_SWIZZLE is defined, swizzling functions enabled. Enable compiler C++ language extensions to enable swizzle operators.")
+#	else
+#		pragma message("GLM: GLM_FORCE_SWIZZLE is undefined. swizzling functions or operators are disabled.")
+#	endif
+
+	// Report .length() type
+#	if GLM_CONFIG_LENGTH_TYPE == GLM_LENGTH_SIZE_T
+#		pragma message("GLM: GLM_FORCE_SIZE_T_LENGTH is defined. .length() returns a glm::length_t, a typedef of std::size_t.")
+#	else
+#		pragma message("GLM: GLM_FORCE_SIZE_T_LENGTH is undefined. .length() returns a glm::length_t, a typedef of int following GLSL.")
+#	endif
+
+#	if GLM_CONFIG_UNRESTRICTED_GENTYPE == GLM_ENABLE
+#		pragma message("GLM: GLM_FORCE_UNRESTRICTED_GENTYPE is defined. Removes GLSL restrictions on valid function genTypes.")
+#	else
+#		pragma message("GLM: GLM_FORCE_UNRESTRICTED_GENTYPE is undefined. Follows strictly GLSL on valid function genTypes.")
+#	endif
+
+#	if GLM_SILENT_WARNINGS == GLM_ENABLE
+#		pragma message("GLM: GLM_FORCE_SILENT_WARNINGS is defined. Ignores C++ warnings from using C++ language extensions.")
+#	else
+#		pragma message("GLM: GLM_FORCE_SILENT_WARNINGS is undefined. Shows C++ warnings from using C++ language extensions.")
+#	endif
+
+#	ifdef GLM_FORCE_SINGLE_ONLY
+#		pragma message("GLM: GLM_FORCE_SINGLE_ONLY is defined. Using only single precision floating-point types.")
+#	endif
+
+#	if defined(GLM_FORCE_ALIGNED_GENTYPES) && (GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE)
+#		undef GLM_FORCE_ALIGNED_GENTYPES
+#		pragma message("GLM: GLM_FORCE_ALIGNED_GENTYPES is defined, allowing aligned types. This prevents the use of C++ constexpr.")
+#	elif defined(GLM_FORCE_ALIGNED_GENTYPES) && (GLM_CONFIG_ALIGNED_GENTYPES == GLM_DISABLE)
+#		undef GLM_FORCE_ALIGNED_GENTYPES
+#		pragma message("GLM: GLM_FORCE_ALIGNED_GENTYPES is defined but is disabled. It requires C++11 and language extensions.")
+#	endif
+
+#	if defined(GLM_FORCE_DEFAULT_ALIGNED_GENTYPES)
+#		if GLM_CONFIG_ALIGNED_GENTYPES == GLM_DISABLE
+#			undef GLM_FORCE_DEFAULT_ALIGNED_GENTYPES
+#			pragma message("GLM: GLM_FORCE_DEFAULT_ALIGNED_GENTYPES is defined but is disabled. It requires C++11 and language extensions.")
+#		elif GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+#			pragma message("GLM: GLM_FORCE_DEFAULT_ALIGNED_GENTYPES is defined. All gentypes (e.g. vec3) will be aligned and padded by default.")
+#		endif
+#	endif
+
+#	if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+#		pragma message("GLM: GLM_FORCE_DEPTH_ZERO_TO_ONE is defined. Using zero to one depth clip space.")
+#	else
+#		pragma message("GLM: GLM_FORCE_DEPTH_ZERO_TO_ONE is undefined. Using negative one to one depth clip space.")
+#	endif
+
+#	if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+#		pragma message("GLM: GLM_FORCE_LEFT_HANDED is defined. Using left handed coordinate system.")
+#	else
+#		pragma message("GLM: GLM_FORCE_LEFT_HANDED is undefined. Using right handed coordinate system.")
+#	endif
+#endif//GLM_MESSAGES
+
+#endif//GLM_SETUP_INCLUDED
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_float.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_float.hpp
new file mode 100644
index 000000000000..c8037ebd7aa2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_float.hpp
@@ -0,0 +1,68 @@
+#pragma once
+
+#include "setup.hpp"
+
+#if GLM_COMPILER == GLM_COMPILER_VC12
+#	pragma warning(push)
+#	pragma warning(disable: 4512) // assignment operator could not be generated
+#endif
+
+namespace glm{
+namespace detail
+{
+	template <typename T>
+	union float_t
+	{};
+
+	// https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
+	template <>
+	union float_t<float>
+	{
+		typedef int int_type;
+		typedef float float_type;
+
+		GLM_CONSTEXPR float_t(float_type Num = 0.0f) : f(Num) {}
+
+		GLM_CONSTEXPR float_t& operator=(float_t const& x)
+		{
+			f = x.f;
+			return *this;
+		}
+
+		// Portable extraction of components.
+		GLM_CONSTEXPR bool negative() const { return i < 0; }
+		GLM_CONSTEXPR int_type mantissa() const { return i & ((1 << 23) - 1); }
+		GLM_CONSTEXPR int_type exponent() const { return (i >> 23) & ((1 << 8) - 1); }
+
+		int_type i;
+		float_type f;
+	};
+
+	template <>
+	union float_t<double>
+	{
+		typedef detail::int64 int_type;
+		typedef double float_type;
+
+		GLM_CONSTEXPR float_t(float_type Num = static_cast<float_type>(0)) : f(Num) {}
+
+		GLM_CONSTEXPR float_t& operator=(float_t const& x)
+		{
+			f = x.f;
+			return *this;
+		}
+
+		// Portable extraction of components.
+		GLM_CONSTEXPR bool negative() const { return i < 0; }
+		GLM_CONSTEXPR int_type mantissa() const { return i & ((int_type(1) << 52) - 1); }
+		GLM_CONSTEXPR int_type exponent() const { return (i >> 52) & ((int_type(1) << 11) - 1); }
+
+		int_type i;
+		float_type f;
+	};
+}//namespace detail
+}//namespace glm
+
+#if GLM_COMPILER == GLM_COMPILER_VC12
+#	pragma warning(pop)
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_half.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_half.hpp
new file mode 100644
index 000000000000..40b8bec00d34
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_half.hpp
@@ -0,0 +1,16 @@
+#pragma once
+
+#include "setup.hpp"
+
+namespace glm{
+namespace detail
+{
+	typedef short hdata;
+
+	GLM_FUNC_DECL float toFloat32(hdata value);
+	GLM_FUNC_DECL hdata toFloat16(float const& value);
+
+}//namespace detail
+}//namespace glm
+
+#include "type_half.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_half.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_half.inl
new file mode 100644
index 000000000000..5d239cf22c25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_half.inl
@@ -0,0 +1,241 @@
+namespace glm{
+namespace detail
+{
+	GLM_FUNC_QUALIFIER float overflow()
+	{
+		volatile float f = 1e10;
+
+		for(int i = 0; i < 10; ++i)
+			f = f * f; // this will overflow before the for loop terminates
+		return f;
+	}
+
+	union uif32
+	{
+		GLM_FUNC_QUALIFIER uif32() :
+			i(0)
+		{}
+
+		GLM_FUNC_QUALIFIER uif32(float f_) :
+			f(f_)
+		{}
+
+		GLM_FUNC_QUALIFIER uif32(unsigned int i_) :
+			i(i_)
+		{}
+
+		float f;
+		unsigned int i;
+	};
+
+	GLM_FUNC_QUALIFIER float toFloat32(hdata value)
+	{
+		int s = (value >> 15) & 0x00000001;
+		int e = (value >> 10) & 0x0000001f;
+		int m =  value        & 0x000003ff;
+
+		if(e == 0)
+		{
+			if(m == 0)
+			{
+				//
+				// Plus or minus zero
+				//
+
+				detail::uif32 result;
+				result.i = static_cast<unsigned int>(s << 31);
+				return result.f;
+			}
+			else
+			{
+				//
+				// Denormalized number -- renormalize it
+				//
+
+				while(!(m & 0x00000400))
+				{
+					m <<= 1;
+					e -=  1;
+				}
+
+				e += 1;
+				m &= ~0x00000400;
+			}
+		}
+		else if(e == 31)
+		{
+			if(m == 0)
+			{
+				//
+				// Positive or negative infinity
+				//
+
+				uif32 result;
+				result.i = static_cast<unsigned int>((s << 31) | 0x7f800000);
+				return result.f;
+			}
+			else
+			{
+				//
+				// Nan -- preserve sign and significand bits
+				//
+
+				uif32 result;
+				result.i = static_cast<unsigned int>((s << 31) | 0x7f800000 | (m << 13));
+				return result.f;
+			}
+		}
+
+		//
+		// Normalized number
+		//
+
+		e = e + (127 - 15);
+		m = m << 13;
+
+		//
+		// Assemble s, e and m.
+		//
+
+		uif32 Result;
+		Result.i = static_cast<unsigned int>((s << 31) | (e << 23) | m);
+		return Result.f;
+	}
+
+	GLM_FUNC_QUALIFIER hdata toFloat16(float const& f)
+	{
+		uif32 Entry;
+		Entry.f = f;
+		int i = static_cast<int>(Entry.i);
+
+		//
+		// Our floating point number, f, is represented by the bit
+		// pattern in integer i.  Disassemble that bit pattern into
+		// the sign, s, the exponent, e, and the significand, m.
+		// Shift s into the position where it will go in the
+		// resulting half number.
+		// Adjust e, accounting for the different exponent bias
+		// of float and half (127 versus 15).
+		//
+
+		int s =  (i >> 16) & 0x00008000;
+		int e = ((i >> 23) & 0x000000ff) - (127 - 15);
+		int m =   i        & 0x007fffff;
+
+		//
+		// Now reassemble s, e and m into a half:
+		//
+
+		if(e <= 0)
+		{
+			if(e < -10)
+			{
+				//
+				// E is less than -10.  The absolute value of f is
+				// less than half_MIN (f may be a small normalized
+				// float, a denormalized float or a zero).
+				//
+				// We convert f to a half zero.
+				//
+
+				return hdata(s);
+			}
+
+			//
+			// E is between -10 and 0.  F is a normalized float,
+			// whose magnitude is less than __half_NRM_MIN.
+			//
+			// We convert f to a denormalized half.
+			//
+
+			m = (m | 0x00800000) >> (1 - e);
+
+			//
+			// Round to nearest, round "0.5" up.
+			//
+			// Rounding may cause the significand to overflow and make
+			// our number normalized.  Because of the way a half's bits
+			// are laid out, we don't have to treat this case separately;
+			// the code below will handle it correctly.
+			//
+
+			if(m & 0x00001000)
+				m += 0x00002000;
+
+			//
+			// Assemble the half from s, e (zero) and m.
+			//
+
+			return hdata(s | (m >> 13));
+		}
+		else if(e == 0xff - (127 - 15))
+		{
+			if(m == 0)
+			{
+				//
+				// F is an infinity; convert f to a half
+				// infinity with the same sign as f.
+				//
+
+				return hdata(s | 0x7c00);
+			}
+			else
+			{
+				//
+				// F is a NAN; we produce a half NAN that preserves
+				// the sign bit and the 10 leftmost bits of the
+				// significand of f, with one exception: If the 10
+				// leftmost bits are all zero, the NAN would turn
+				// into an infinity, so we have to set at least one
+				// bit in the significand.
+				//
+
+				m >>= 13;
+
+				return hdata(s | 0x7c00 | m | (m == 0));
+			}
+		}
+		else
+		{
+			//
+			// E is greater than zero.  F is a normalized float.
+			// We try to convert f to a normalized half.
+			//
+
+			//
+			// Round to nearest, round "0.5" up
+			//
+
+			if(m &  0x00001000)
+			{
+				m += 0x00002000;
+
+				if(m & 0x00800000)
+				{
+					m =  0;     // overflow in significand,
+					e += 1;     // adjust exponent
+				}
+			}
+
+			//
+			// Handle exponent overflow
+			//
+
+			if (e > 30)
+			{
+				overflow();        // Cause a hardware floating point overflow;
+
+				return hdata(s | 0x7c00);
+				// if this returns, the half becomes an
+			}   // infinity with the same sign as f.
+
+			//
+			// Assemble the half from s, e and m.
+			//
+
+			return hdata(s | (e << 10) | (m >> 13));
+		}
+	}
+
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.hpp
new file mode 100644
index 000000000000..a61bded4b7ae
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.hpp
@@ -0,0 +1,177 @@
+/// @ref core
+/// @file glm/detail/type_mat2x2.hpp
+
+#pragma once
+
+#include "type_vec2.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<2, 2, T, Q>
+	{
+		typedef vec<2, T, Q> col_type;
+		typedef vec<2, T, Q> row_type;
+		typedef mat<2, 2, T, Q> type;
+		typedef mat<2, 2, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[2];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 2; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<2, 2, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T const& x1, T const& y1,
+			T const& x2, T const& y2);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v1,
+			col_type const& v2);
+
+		// -- Conversions --
+
+		template<typename U, typename V, typename M, typename N>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			U const& x1, V const& y1,
+			M const& x2, N const& y2);
+
+		template<typename U, typename V>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<2, U, Q> const& v1,
+			vec<2, V, Q> const& v2);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator=(mat<2, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator+=(mat<2, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator-=(mat<2, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator*=(mat<2, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator/=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator/=(mat<2, 2, U, Q> const& m);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator++ ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> & operator-- ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator+(mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator-(mat<2, 2, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator+(mat<2, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator+(T scalar, mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator+(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator-(mat<2, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator-(T scalar, mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator-(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<2, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator*(T scalar, mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 2, T, Q>::col_type operator*(mat<2, 2, T, Q> const& m, typename mat<2, 2, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 2, T, Q>::row_type operator*(typename mat<2, 2, T, Q>::col_type const& v, mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<2, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<2, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator/(mat<2, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator/(T scalar, mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 2, T, Q>::col_type operator/(mat<2, 2, T, Q> const& m, typename mat<2, 2, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 2, T, Q>::row_type operator/(typename mat<2, 2, T, Q>::col_type const& v, mat<2, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator/(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+} //namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat2x2.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.inl
new file mode 100644
index 000000000000..88eca20b4040
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x2.inl
@@ -0,0 +1,536 @@
+#include "../matrix.hpp"
+
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0), col_type(0, 1)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0);
+				this->value[1] = col_type(0, 1);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<2, 2, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(T scalar)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(scalar, 0), col_type(0, scalar)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(scalar, 0);
+			this->value[1] = col_type(0, scalar);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat
+	(
+		T const& x0, T const& y0,
+		T const& x1, T const& y1
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0), col_type(x1, y1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0);
+			this->value[1] = col_type(x1, y1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(col_type const& v0, col_type const& v1)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{v0, v1}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<typename X1, typename Y1, typename X2, typename Y2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat
+	(
+		X1 const& x1, Y1 const& y1,
+		X2 const& x2, Y2 const& y2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(static_cast<T>(x1), value_type(y1)), col_type(static_cast<T>(x2), value_type(y2)) }
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(static_cast<T>(x1), value_type(y1));
+			this->value[1] = col_type(static_cast<T>(x2), value_type(y2));
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(vec<2, V1, Q> const& v1, vec<2, V2, Q> const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v1);
+			this->value[1] = col_type(v2);
+#		endif
+	}
+
+	// -- mat2x2 matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<2, 2, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 2, T, Q>::col_type& mat<2, 2, T, Q>::operator[](typename mat<2, 2, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 2, T, Q>::col_type const& mat<2, 2, T, Q>::operator[](typename mat<2, 2, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator=(mat<2, 2, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator+=(U scalar)
+	{
+		this->value[0] += scalar;
+		this->value[1] += scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator+=(mat<2, 2, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator-=(U scalar)
+	{
+		this->value[0] -= scalar;
+		this->value[1] -= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator-=(mat<2, 2, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator*=(U scalar)
+	{
+		this->value[0] *= scalar;
+		this->value[1] *= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator*=(mat<2, 2, U, Q> const& m)
+	{
+		return (*this = *this * m);
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator/=(U scalar)
+	{
+		this->value[0] /= scalar;
+		this->value[1] /= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator/=(mat<2, 2, U, Q> const& m)
+	{
+		return *this *= inverse(m);
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q>& mat<2, 2, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> mat<2, 2, T, Q>::operator++(int)
+	{
+		mat<2, 2, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> mat<2, 2, T, Q>::operator--(int)
+	{
+		mat<2, 2, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator+(mat<2, 2, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator-(mat<2, 2, T, Q> const& m)
+	{
+		return mat<2, 2, T, Q>(
+			-m[0],
+			-m[1]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator+(mat<2, 2, T, Q> const& m, T scalar)
+	{
+		return mat<2, 2, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator+(T scalar, mat<2, 2, T, Q> const& m)
+	{
+		return mat<2, 2, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator+(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return mat<2, 2, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator-(mat<2, 2, T, Q> const& m, T scalar)
+	{
+		return mat<2, 2, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator-(T scalar, mat<2, 2, T, Q> const& m)
+	{
+		return mat<2, 2, T, Q>(
+			scalar - m[0],
+			scalar - m[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator-(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return mat<2, 2, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<2, 2, T, Q> const& m, T scalar)
+	{
+		return mat<2, 2, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator*(T scalar, mat<2, 2, T, Q> const& m)
+	{
+		return mat<2, 2, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 2, T, Q>::col_type operator*
+	(
+		mat<2, 2, T, Q> const& m,
+		typename mat<2, 2, T, Q>::row_type const& v
+	)
+	{
+		return vec<2, T, Q>(
+			m[0][0] * v.x + m[1][0] * v.y,
+			m[0][1] * v.x + m[1][1] * v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 2, T, Q>::row_type operator*
+	(
+		typename mat<2, 2, T, Q>::col_type const& v,
+		mat<2, 2, T, Q> const& m
+	)
+	{
+		return vec<2, T, Q>(
+			v.x * m[0][0] + v.y * m[0][1],
+			v.x * m[1][0] + v.y * m[1][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return mat<2, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<2, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		return mat<3, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<2, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		return mat<4, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1],
+			m1[0][0] * m2[3][0] + m1[1][0] * m2[3][1],
+			m1[0][1] * m2[3][0] + m1[1][1] * m2[3][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator/(mat<2, 2, T, Q> const& m, T scalar)
+	{
+		return mat<2, 2, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator/(T scalar, mat<2, 2, T, Q> const& m)
+	{
+		return mat<2, 2, T, Q>(
+			scalar / m[0],
+			scalar / m[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 2, T, Q>::col_type operator/(mat<2, 2, T, Q> const& m, typename mat<2, 2, T, Q>::row_type const& v)
+	{
+		return inverse(m) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 2, T, Q>::row_type operator/(typename mat<2, 2, T, Q>::col_type const& v, mat<2, 2, T, Q> const& m)
+	{
+		return v *  inverse(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator/(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		mat<2, 2, T, Q> m1_copy(m1);
+		return m1_copy /= m2;
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<2, 2, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.hpp
new file mode 100644
index 000000000000..21015f485d15
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.hpp
@@ -0,0 +1,159 @@
+/// @ref core
+/// @file glm/detail/type_mat2x3.hpp
+
+#pragma once
+
+#include "type_vec2.hpp"
+#include "type_vec3.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<2, 3, T, Q>
+	{
+		typedef vec<3, T, Q> col_type;
+		typedef vec<2, T, Q> row_type;
+		typedef mat<2, 3, T, Q> type;
+		typedef mat<3, 2, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[2];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 2; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<2, 3, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T x0, T y0, T z0,
+			T x1, T y1, T z1);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1);
+
+		// -- Conversions --
+
+		template<typename X1, typename Y1, typename Z1, typename X2, typename Y2, typename Z2>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 x1, Y1 y1, Z1 z1,
+			X2 x2, Y2 y2, Z2 z2);
+
+		template<typename U, typename V>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<3, U, Q> const& v1,
+			vec<3, V, Q> const& v2);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator=(mat<2, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator+=(mat<2, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator-=(mat<2, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator/=(U s);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator++ ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> & operator-- ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator+(mat<2, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator-(mat<2, 3, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator+(mat<2, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator+(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator-(mat<2, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator-(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<2, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator*(T scalar, mat<2, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 3, T, Q>::col_type operator*(mat<2, 3, T, Q> const& m, typename mat<2, 3, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 3, T, Q>::row_type operator*(typename mat<2, 3, T, Q>::col_type const& v, mat<2, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<2, 3, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<2, 3, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<2, 3, T, Q> const& m1, mat<4, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator/(mat<2, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator/(T scalar, mat<2, 3, T, Q> const& m);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat2x3.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.inl
new file mode 100644
index 000000000000..88077518f9f0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x3.inl
@@ -0,0 +1,510 @@
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0, 0), col_type(0, 1, 0)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0, 0);
+				this->value[1] = col_type(0, 1, 0);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<2, 3, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m.value[0];
+			this->value[1] = m.value[1];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(T scalar)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(scalar, 0, 0), col_type(0, scalar, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(scalar, 0, 0);
+			this->value[1] = col_type(0, scalar, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat
+	(
+		T x0, T y0, T z0,
+		T x1, T y1, T z1
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0, z0), col_type(x1, y1, z1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0);
+			this->value[1] = col_type(x1, y1, z1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(col_type const& v0, col_type const& v1)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v0);
+			this->value[1] = col_type(v1);
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X1, typename Y1, typename Z1,
+		typename X2, typename Y2, typename Z2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat
+	(
+		X1 x1, Y1 y1, Z1 z1,
+		X2 x2, Y2 y2, Z2 z2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x1, y1, z1), col_type(x2, y2, z2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x1, y1, z1);
+			this->value[1] = col_type(x2, y2, z2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(vec<3, V1, Q> const& v1, vec<3, V2, Q> const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v1);
+			this->value[1] = col_type(v2);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<2, 3, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR  mat<2, 3, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+		: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 3, T, Q>::col_type & mat<2, 3, T, Q>::operator[](typename mat<2, 3, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 3, T, Q>::col_type const& mat<2, 3, T, Q>::operator[](typename mat<2, 3, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>& mat<2, 3, T, Q>::operator=(mat<2, 3, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> & mat<2, 3, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>& mat<2, 3, T, Q>::operator+=(mat<2, 3, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>& mat<2, 3, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>& mat<2, 3, T, Q>::operator-=(mat<2, 3, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q>& mat<2, 3, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> & mat<2, 3, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> & mat<2, 3, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> & mat<2, 3, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> mat<2, 3, T, Q>::operator++(int)
+	{
+		mat<2, 3, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> mat<2, 3, T, Q>::operator--(int)
+	{
+		mat<2, 3, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator+(mat<2, 3, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator-(mat<2, 3, T, Q> const& m)
+	{
+		return mat<2, 3, T, Q>(
+			-m[0],
+			-m[1]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator+(mat<2, 3, T, Q> const& m, T scalar)
+	{
+		return mat<2, 3, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator+(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		return mat<2, 3, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator-(mat<2, 3, T, Q> const& m, T scalar)
+	{
+		return mat<2, 3, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator-(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		return mat<2, 3, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<2, 3, T, Q> const& m, T scalar)
+	{
+		return mat<2, 3, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator*(T scalar, mat<2, 3, T, Q> const& m)
+	{
+		return mat<2, 3, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 3, T, Q>::col_type operator*
+	(
+		mat<2, 3, T, Q> const& m,
+		typename mat<2, 3, T, Q>::row_type const& v)
+	{
+		return typename mat<2, 3, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y,
+			m[0][1] * v.x + m[1][1] * v.y,
+			m[0][2] * v.x + m[1][2] * v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 3, T, Q>::row_type operator*
+	(
+		typename mat<2, 3, T, Q>::col_type const& v,
+		mat<2, 3, T, Q> const& m)
+	{
+		return typename mat<2, 3, T, Q>::row_type(
+			v.x * m[0][0] + v.y * m[0][1] + v.z * m[0][2],
+			v.x * m[1][0] + v.y * m[1][1] + v.z * m[1][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<2, 3, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return mat<2, 3, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<2, 3, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		T SrcA00 = m1[0][0];
+		T SrcA01 = m1[0][1];
+		T SrcA02 = m1[0][2];
+		T SrcA10 = m1[1][0];
+		T SrcA11 = m1[1][1];
+		T SrcA12 = m1[1][2];
+
+		T SrcB00 = m2[0][0];
+		T SrcB01 = m2[0][1];
+		T SrcB10 = m2[1][0];
+		T SrcB11 = m2[1][1];
+		T SrcB20 = m2[2][0];
+		T SrcB21 = m2[2][1];
+
+		mat<3, 3, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01;
+		Result[0][2] = SrcA02 * SrcB00 + SrcA12 * SrcB01;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11;
+		Result[1][2] = SrcA02 * SrcB10 + SrcA12 * SrcB11;
+		Result[2][0] = SrcA00 * SrcB20 + SrcA10 * SrcB21;
+		Result[2][1] = SrcA01 * SrcB20 + SrcA11 * SrcB21;
+		Result[2][2] = SrcA02 * SrcB20 + SrcA12 * SrcB21;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<2, 3, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		return mat<4, 3, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1],
+			m1[0][2] * m2[2][0] + m1[1][2] * m2[2][1],
+			m1[0][0] * m2[3][0] + m1[1][0] * m2[3][1],
+			m1[0][1] * m2[3][0] + m1[1][1] * m2[3][1],
+			m1[0][2] * m2[3][0] + m1[1][2] * m2[3][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator/(mat<2, 3, T, Q> const& m, T scalar)
+	{
+		return mat<2, 3, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator/(T scalar, mat<2, 3, T, Q> const& m)
+	{
+		return mat<2, 3, T, Q>(
+			scalar / m[0],
+			scalar / m[1]);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<2, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.hpp
new file mode 100644
index 000000000000..ee6dc68574e5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.hpp
@@ -0,0 +1,161 @@
+/// @ref core
+/// @file glm/detail/type_mat2x4.hpp
+
+#pragma once
+
+#include "type_vec2.hpp"
+#include "type_vec4.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<2, 4, T, Q>
+	{
+		typedef vec<4, T, Q> col_type;
+		typedef vec<2, T, Q> row_type;
+		typedef mat<2, 4, T, Q> type;
+		typedef mat<4, 2, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[2];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 2; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<2, 4, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T x0, T y0, T z0, T w0,
+			T x1, T y1, T z1, T w1);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1);
+
+		// -- Conversions --
+
+		template<
+			typename X1, typename Y1, typename Z1, typename W1,
+			typename X2, typename Y2, typename Z2, typename W2>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 x1, Y1 y1, Z1 z1, W1 w1,
+			X2 x2, Y2 y2, Z2 z2, W2 w2);
+
+		template<typename U, typename V>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<4, U, Q> const& v1,
+			vec<4, V, Q> const& v2);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator=(mat<2, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator+=(mat<2, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator-=(mat<2, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator/=(U s);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator++ ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> & operator-- ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator+(mat<2, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator-(mat<2, 4, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator+(mat<2, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator+(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator-(mat<2, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator-(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<2, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator*(T scalar, mat<2, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 4, T, Q>::col_type operator*(mat<2, 4, T, Q> const& m, typename mat<2, 4, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<2, 4, T, Q>::row_type operator*(typename mat<2, 4, T, Q>::col_type const& v, mat<2, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<2, 4, T, Q> const& m1, mat<4, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<2, 4, T, Q> const& m1, mat<2, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<2, 4, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator/(mat<2, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator/(T scalar, mat<2, 4, T, Q> const& m);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat2x4.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.inl
new file mode 100644
index 000000000000..e70753c06a3a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat2x4.inl
@@ -0,0 +1,520 @@
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0, 0, 0), col_type(0, 1, 0, 0)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0, 0, 0);
+				this->value[1] = col_type(0, 1, 0, 0);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<2, 4, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0, 0, 0), col_type(0, s, 0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0, 0, 0);
+			this->value[1] = col_type(0, s, 0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat
+	(
+		T x0, T y0, T z0, T w0,
+		T x1, T y1, T z1, T w1
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0, z0, w0), col_type(x1, y1, z1, w1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0, w0);
+			this->value[1] = col_type(x1, y1, z1, w1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(col_type const& v0, col_type const& v1)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X1, typename Y1, typename Z1, typename W1,
+		typename X2, typename Y2, typename Z2, typename W2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat
+	(
+		X1 x1, Y1 y1, Z1 z1, W1 w1,
+		X2 x2, Y2 y2, Z2 z2, W2 w2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{
+				col_type(x1, y1, z1, w1),
+				col_type(x2, y2, z2, w2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x1, y1, z1, w1);
+			this->value[1] = col_type(x2, y2, z2, w2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(vec<4, V1, Q> const& v1, vec<4, V2, Q> const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v1);
+			this->value[1] = col_type(v2);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<2, 4, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 4, T, Q>::col_type & mat<2, 4, T, Q>::operator[](typename mat<2, 4, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 4, T, Q>::col_type const& mat<2, 4, T, Q>::operator[](typename mat<2, 4, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator=(mat<2, 4, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator+=(mat<2, 4, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator-=(mat<2, 4, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> & mat<2, 4, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q>& mat<2, 4, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> mat<2, 4, T, Q>::operator++(int)
+	{
+		mat<2, 4, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> mat<2, 4, T, Q>::operator--(int)
+	{
+		mat<2, 4, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator+(mat<2, 4, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator-(mat<2, 4, T, Q> const& m)
+	{
+		return mat<2, 4, T, Q>(
+			-m[0],
+			-m[1]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator+(mat<2, 4, T, Q> const& m, T scalar)
+	{
+		return mat<2, 4, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator+(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		return mat<2, 4, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator-(mat<2, 4, T, Q> const& m, T scalar)
+	{
+		return mat<2, 4, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator-(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		return mat<2, 4, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<2, 4, T, Q> const& m, T scalar)
+	{
+		return mat<2, 4, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator*(T scalar, mat<2, 4, T, Q> const& m)
+	{
+		return mat<2, 4, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 4, T, Q>::col_type operator*(mat<2, 4, T, Q> const& m, typename mat<2, 4, T, Q>::row_type const& v)
+	{
+		return typename mat<2, 4, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y,
+			m[0][1] * v.x + m[1][1] * v.y,
+			m[0][2] * v.x + m[1][2] * v.y,
+			m[0][3] * v.x + m[1][3] * v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<2, 4, T, Q>::row_type operator*(typename mat<2, 4, T, Q>::col_type const& v, mat<2, 4, T, Q> const& m)
+	{
+		return typename mat<2, 4, T, Q>::row_type(
+			v.x * m[0][0] + v.y * m[0][1] + v.z * m[0][2] + v.w * m[0][3],
+			v.x * m[1][0] + v.y * m[1][1] + v.z * m[1][2] + v.w * m[1][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<2, 4, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		T SrcA00 = m1[0][0];
+		T SrcA01 = m1[0][1];
+		T SrcA02 = m1[0][2];
+		T SrcA03 = m1[0][3];
+		T SrcA10 = m1[1][0];
+		T SrcA11 = m1[1][1];
+		T SrcA12 = m1[1][2];
+		T SrcA13 = m1[1][3];
+
+		T SrcB00 = m2[0][0];
+		T SrcB01 = m2[0][1];
+		T SrcB10 = m2[1][0];
+		T SrcB11 = m2[1][1];
+		T SrcB20 = m2[2][0];
+		T SrcB21 = m2[2][1];
+		T SrcB30 = m2[3][0];
+		T SrcB31 = m2[3][1];
+
+		mat<4, 4, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01;
+		Result[0][2] = SrcA02 * SrcB00 + SrcA12 * SrcB01;
+		Result[0][3] = SrcA03 * SrcB00 + SrcA13 * SrcB01;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11;
+		Result[1][2] = SrcA02 * SrcB10 + SrcA12 * SrcB11;
+		Result[1][3] = SrcA03 * SrcB10 + SrcA13 * SrcB11;
+		Result[2][0] = SrcA00 * SrcB20 + SrcA10 * SrcB21;
+		Result[2][1] = SrcA01 * SrcB20 + SrcA11 * SrcB21;
+		Result[2][2] = SrcA02 * SrcB20 + SrcA12 * SrcB21;
+		Result[2][3] = SrcA03 * SrcB20 + SrcA13 * SrcB21;
+		Result[3][0] = SrcA00 * SrcB30 + SrcA10 * SrcB31;
+		Result[3][1] = SrcA01 * SrcB30 + SrcA11 * SrcB31;
+		Result[3][2] = SrcA02 * SrcB30 + SrcA12 * SrcB31;
+		Result[3][3] = SrcA03 * SrcB30 + SrcA13 * SrcB31;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<2, 4, T, Q> const& m1, mat<2, 2, T, Q> const& m2)
+	{
+		return mat<2, 4, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1],
+			m1[0][3] * m2[0][0] + m1[1][3] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1],
+			m1[0][3] * m2[1][0] + m1[1][3] * m2[1][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<2, 4, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		return mat<3, 4, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1],
+			m1[0][3] * m2[0][0] + m1[1][3] * m2[0][1],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1],
+			m1[0][3] * m2[1][0] + m1[1][3] * m2[1][1],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1],
+			m1[0][2] * m2[2][0] + m1[1][2] * m2[2][1],
+			m1[0][3] * m2[2][0] + m1[1][3] * m2[2][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator/(mat<2, 4, T, Q> const& m, T scalar)
+	{
+		return mat<2, 4, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator/(T scalar, mat<2, 4, T, Q> const& m)
+	{
+		return mat<2, 4, T, Q>(
+			scalar / m[0],
+			scalar / m[1]);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<2, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.hpp
new file mode 100644
index 000000000000..4fe521aa89a0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.hpp
@@ -0,0 +1,167 @@
+/// @ref core
+/// @file glm/detail/type_mat3x2.hpp
+
+#pragma once
+
+#include "type_vec2.hpp"
+#include "type_vec3.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<3, 2, T, Q>
+	{
+		typedef vec<2, T, Q> col_type;
+		typedef vec<3, T, Q> row_type;
+		typedef mat<3, 2, T, Q> type;
+		typedef mat<2, 3, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[3];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 3; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<3, 2, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T x0, T y0,
+			T x1, T y1,
+			T x2, T y2);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1,
+			col_type const& v2);
+
+		// -- Conversions --
+
+		template<
+			typename X1, typename Y1,
+			typename X2, typename Y2,
+			typename X3, typename Y3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 x1, Y1 y1,
+			X2 x2, Y2 y2,
+			X3 x3, Y3 y3);
+
+		template<typename V1, typename V2, typename V3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<2, V1, Q> const& v1,
+			vec<2, V2, Q> const& v2,
+			vec<2, V3, Q> const& v3);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator=(mat<3, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator+=(mat<3, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator-=(mat<3, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator/=(U s);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator++ ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> & operator-- ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator+(mat<3, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator-(mat<3, 2, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator+(mat<3, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator+(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator-(mat<3, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator-(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<3, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator*(T scalar, mat<3, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 2, T, Q>::col_type operator*(mat<3, 2, T, Q> const& m, typename mat<3, 2, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 2, T, Q>::row_type operator*(typename mat<3, 2, T, Q>::col_type const& v, mat<3, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<3, 2, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<3, 2, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<3, 2, T, Q> const& m1, mat<4, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator/(mat<3, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator/(T scalar, mat<3, 2, T, Q> const& m);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2);
+
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat3x2.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.inl
new file mode 100644
index 000000000000..5a1b4c010a04
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x2.inl
@@ -0,0 +1,532 @@
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0), col_type(0, 1), col_type(0, 0)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0);
+				this->value[1] = col_type(0, 1);
+				this->value[2] = col_type(0, 0);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<3, 2, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0), col_type(0, s), col_type(0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0);
+			this->value[1] = col_type(0, s);
+			this->value[2] = col_type(0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat
+	(
+		T x0, T y0,
+		T x1, T y1,
+		T x2, T y2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0), col_type(x1, y1), col_type(x2, y2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0);
+			this->value[1] = col_type(x1, y1);
+			this->value[2] = col_type(x2, y2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(col_type const& v0, col_type const& v1, col_type const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+			this->value[2] = v2;
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X0, typename Y0,
+		typename X1, typename Y1,
+		typename X2, typename Y2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat
+	(
+		X0 x0, Y0 y0,
+		X1 x1, Y1 y1,
+		X2 x2, Y2 y2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0), col_type(x1, y1), col_type(x2, y2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0);
+			this->value[1] = col_type(x1, y1);
+			this->value[2] = col_type(x2, y2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V0, typename V1, typename V2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(vec<2, V0, Q> const& v0, vec<2, V1, Q> const& v1, vec<2, V2, Q> const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v0);
+			this->value[1] = col_type(v1);
+			this->value[2] = col_type(v2);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<3, 2, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 2, T, Q>::col_type & mat<3, 2, T, Q>::operator[](typename mat<3, 2, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 2, T, Q>::col_type const& mat<3, 2, T, Q>::operator[](typename mat<3, 2, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator=(mat<3, 2, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		this->value[2] = m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		this->value[2] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator+=(mat<3, 2, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		this->value[2] += m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		this->value[2] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator-=(mat<3, 2, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		this->value[2] -= m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		this->value[2] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> & mat<3, 2, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		this->value[2] /= s;
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		++this->value[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q>& mat<3, 2, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		--this->value[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> mat<3, 2, T, Q>::operator++(int)
+	{
+		mat<3, 2, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> mat<3, 2, T, Q>::operator--(int)
+	{
+		mat<3, 2, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator+(mat<3, 2, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator-(mat<3, 2, T, Q> const& m)
+	{
+		return mat<3, 2, T, Q>(
+			-m[0],
+			-m[1],
+			-m[2]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator+(mat<3, 2, T, Q> const& m, T scalar)
+	{
+		return mat<3, 2, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator+(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		return mat<3, 2, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1],
+			m1[2] + m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator-(mat<3, 2, T, Q> const& m, T scalar)
+	{
+		return mat<3, 2, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar,
+			m[2] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator-(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		return mat<3, 2, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1],
+			m1[2] - m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<3, 2, T, Q> const& m, T scalar)
+	{
+		return mat<3, 2, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator*(T scalar, mat<3, 2, T, Q> const& m)
+	{
+		return mat<3, 2, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 2, T, Q>::col_type operator*(mat<3, 2, T, Q> const& m, typename mat<3, 2, T, Q>::row_type const& v)
+	{
+		return typename mat<3, 2, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z,
+			m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 2, T, Q>::row_type operator*(typename mat<3, 2, T, Q>::col_type const& v, mat<3, 2, T, Q> const& m)
+	{
+		return typename mat<3, 2, T, Q>::row_type(
+			v.x * m[0][0] + v.y * m[0][1],
+			v.x * m[1][0] + v.y * m[1][1],
+			v.x * m[2][0] + v.y * m[2][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<3, 2, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		const T SrcA00 = m1[0][0];
+		const T SrcA01 = m1[0][1];
+		const T SrcA10 = m1[1][0];
+		const T SrcA11 = m1[1][1];
+		const T SrcA20 = m1[2][0];
+		const T SrcA21 = m1[2][1];
+
+		const T SrcB00 = m2[0][0];
+		const T SrcB01 = m2[0][1];
+		const T SrcB02 = m2[0][2];
+		const T SrcB10 = m2[1][0];
+		const T SrcB11 = m2[1][1];
+		const T SrcB12 = m2[1][2];
+
+		mat<2, 2, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01 + SrcA20 * SrcB02;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01 + SrcA21 * SrcB02;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11 + SrcA20 * SrcB12;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11 + SrcA21 * SrcB12;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<3, 2, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		return mat<3, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<3, 2, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		return mat<4, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2],
+			m1[0][0] * m2[3][0] + m1[1][0] * m2[3][1] + m1[2][0] * m2[3][2],
+			m1[0][1] * m2[3][0] + m1[1][1] * m2[3][1] + m1[2][1] * m2[3][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator/(mat<3, 2, T, Q> const& m, T scalar)
+	{
+		return mat<3, 2, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar,
+			m[2] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator/(T scalar, mat<3, 2, T, Q> const& m)
+	{
+		return mat<3, 2, T, Q>(
+			scalar / m[0],
+			scalar / m[1],
+			scalar / m[2]);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]) && (m1[2] == m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<3, 2, T, Q> const& m1, mat<3, 2, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]) || (m1[2] != m2[2]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.hpp
new file mode 100644
index 000000000000..fc7df9824a0f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.hpp
@@ -0,0 +1,184 @@
+/// @ref core
+/// @file glm/detail/type_mat3x3.hpp
+
+#pragma once
+
+#include "type_vec3.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<3, 3, T, Q>
+	{
+		typedef vec<3, T, Q> col_type;
+		typedef vec<3, T, Q> row_type;
+		typedef mat<3, 3, T, Q> type;
+		typedef mat<3, 3, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[3];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 3; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<3, 3, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T x0, T y0, T z0,
+			T x1, T y1, T z1,
+			T x2, T y2, T z2);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1,
+			col_type const& v2);
+
+		// -- Conversions --
+
+		template<
+			typename X1, typename Y1, typename Z1,
+			typename X2, typename Y2, typename Z2,
+			typename X3, typename Y3, typename Z3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 x1, Y1 y1, Z1 z1,
+			X2 x2, Y2 y2, Z2 z2,
+			X3 x3, Y3 y3, Z3 z3);
+
+		template<typename V1, typename V2, typename V3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<3, V1, Q> const& v1,
+			vec<3, V2, Q> const& v2,
+			vec<3, V3, Q> const& v3);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator=(mat<3, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator+=(mat<3, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator-=(mat<3, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator*=(mat<3, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator/=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator/=(mat<3, 3, U, Q> const& m);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator+(mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator-(mat<3, 3, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator+(mat<3, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator+(T scalar, mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator+(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator-(mat<3, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator-(T scalar, mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator-(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<3, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator*(T scalar, mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 3, T, Q>::col_type operator*(mat<3, 3, T, Q> const& m, typename mat<3, 3, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 3, T, Q>::row_type operator*(typename mat<3, 3, T, Q>::col_type const& v, mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<3, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<3, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator/(mat<3, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator/(T scalar, mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 3, T, Q>::col_type operator/(mat<3, 3, T, Q> const& m, typename mat<3, 3, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 3, T, Q>::row_type operator/(typename mat<3, 3, T, Q>::col_type const& v, mat<3, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator/(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat3x3.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.inl
new file mode 100644
index 000000000000..ddb00d46bd23
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x3.inl
@@ -0,0 +1,601 @@
+#include "../matrix.hpp"
+
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0, 0), col_type(0, 1, 0), col_type(0, 0, 1)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+			this->value[0] = col_type(1, 0, 0);
+				this->value[1] = col_type(0, 1, 0);
+				this->value[2] = col_type(0, 0, 1);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<3, 3, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0, 0), col_type(0, s, 0), col_type(0, 0, s)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0, 0);
+			this->value[1] = col_type(0, s, 0);
+			this->value[2] = col_type(0, 0, s);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat
+	(
+		T x0, T y0, T z0,
+		T x1, T y1, T z1,
+		T x2, T y2, T z2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0, z0), col_type(x1, y1, z1), col_type(x2, y2, z2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0);
+			this->value[1] = col_type(x1, y1, z1);
+			this->value[2] = col_type(x2, y2, z2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(col_type const& v0, col_type const& v1, col_type const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v0);
+			this->value[1] = col_type(v1);
+			this->value[2] = col_type(v2);
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X1, typename Y1, typename Z1,
+		typename X2, typename Y2, typename Z2,
+		typename X3, typename Y3, typename Z3>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat
+	(
+		X1 x1, Y1 y1, Z1 z1,
+		X2 x2, Y2 y2, Z2 z2,
+		X3 x3, Y3 y3, Z3 z3
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x1, y1, z1), col_type(x2, y2, z2), col_type(x3, y3, z3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x1, y1, z1);
+			this->value[1] = col_type(x2, y2, z2);
+			this->value[2] = col_type(x3, y3, z3);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2, typename V3>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(vec<3, V1, Q> const& v1, vec<3, V2, Q> const& v2, vec<3, V3, Q> const& v3)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v1), col_type(v2), col_type(v3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v1);
+			this->value[1] = col_type(v2);
+			this->value[2] = col_type(v3);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<3, 3, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 3, T, Q>::col_type & mat<3, 3, T, Q>::operator[](typename mat<3, 3, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 3, T, Q>::col_type const& mat<3, 3, T, Q>::operator[](typename mat<3, 3, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator=(mat<3, 3, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		this->value[2] = m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		this->value[2] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator+=(mat<3, 3, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		this->value[2] += m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		this->value[2] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator-=(mat<3, 3, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		this->value[2] -= m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		this->value[2] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator*=(mat<3, 3, U, Q> const& m)
+	{
+		return (*this = *this * m);
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		this->value[2] /= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator/=(mat<3, 3, U, Q> const& m)
+	{
+		return *this *= inverse(m);
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		++this->value[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> & mat<3, 3, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		--this->value[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> mat<3, 3, T, Q>::operator++(int)
+	{
+		mat<3, 3, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> mat<3, 3, T, Q>::operator--(int)
+	{
+		mat<3, 3, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator+(mat<3, 3, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator-(mat<3, 3, T, Q> const& m)
+	{
+		return mat<3, 3, T, Q>(
+			-m[0],
+			-m[1],
+			-m[2]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator+(mat<3, 3, T, Q> const& m, T scalar)
+	{
+		return mat<3, 3, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator+(T scalar, mat<3, 3, T, Q> const& m)
+	{
+		return mat<3, 3, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator+(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		return mat<3, 3, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1],
+			m1[2] + m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator-(mat<3, 3, T, Q> const& m, T scalar)
+	{
+		return mat<3, 3, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar,
+			m[2] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator-(T scalar, mat<3, 3, T, Q> const& m)
+	{
+		return mat<3, 3, T, Q>(
+			scalar - m[0],
+			scalar - m[1],
+			scalar - m[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator-(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		return mat<3, 3, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1],
+			m1[2] - m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<3, 3, T, Q> const& m, T scalar)
+	{
+		return mat<3, 3, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator*(T scalar, mat<3, 3, T, Q> const& m)
+	{
+		return mat<3, 3, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 3, T, Q>::col_type operator*(mat<3, 3, T, Q> const& m, typename mat<3, 3, T, Q>::row_type const& v)
+	{
+		return typename mat<3, 3, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z,
+			m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z,
+			m[0][2] * v.x + m[1][2] * v.y + m[2][2] * v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 3, T, Q>::row_type operator*(typename mat<3, 3, T, Q>::col_type const& v, mat<3, 3, T, Q> const& m)
+	{
+		return typename mat<3, 3, T, Q>::row_type(
+			m[0][0] * v.x + m[0][1] * v.y + m[0][2] * v.z,
+			m[1][0] * v.x + m[1][1] * v.y + m[1][2] * v.z,
+			m[2][0] * v.x + m[2][1] * v.y + m[2][2] * v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		T const SrcA00 = m1[0][0];
+		T const SrcA01 = m1[0][1];
+		T const SrcA02 = m1[0][2];
+		T const SrcA10 = m1[1][0];
+		T const SrcA11 = m1[1][1];
+		T const SrcA12 = m1[1][2];
+		T const SrcA20 = m1[2][0];
+		T const SrcA21 = m1[2][1];
+		T const SrcA22 = m1[2][2];
+
+		T const SrcB00 = m2[0][0];
+		T const SrcB01 = m2[0][1];
+		T const SrcB02 = m2[0][2];
+		T const SrcB10 = m2[1][0];
+		T const SrcB11 = m2[1][1];
+		T const SrcB12 = m2[1][2];
+		T const SrcB20 = m2[2][0];
+		T const SrcB21 = m2[2][1];
+		T const SrcB22 = m2[2][2];
+
+		mat<3, 3, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01 + SrcA20 * SrcB02;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01 + SrcA21 * SrcB02;
+		Result[0][2] = SrcA02 * SrcB00 + SrcA12 * SrcB01 + SrcA22 * SrcB02;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11 + SrcA20 * SrcB12;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11 + SrcA21 * SrcB12;
+		Result[1][2] = SrcA02 * SrcB10 + SrcA12 * SrcB11 + SrcA22 * SrcB12;
+		Result[2][0] = SrcA00 * SrcB20 + SrcA10 * SrcB21 + SrcA20 * SrcB22;
+		Result[2][1] = SrcA01 * SrcB20 + SrcA11 * SrcB21 + SrcA21 * SrcB22;
+		Result[2][2] = SrcA02 * SrcB20 + SrcA12 * SrcB21 + SrcA22 * SrcB22;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<3, 3, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		return mat<2, 3, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<3, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		return mat<4, 3, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2],
+			m1[0][2] * m2[2][0] + m1[1][2] * m2[2][1] + m1[2][2] * m2[2][2],
+			m1[0][0] * m2[3][0] + m1[1][0] * m2[3][1] + m1[2][0] * m2[3][2],
+			m1[0][1] * m2[3][0] + m1[1][1] * m2[3][1] + m1[2][1] * m2[3][2],
+			m1[0][2] * m2[3][0] + m1[1][2] * m2[3][1] + m1[2][2] * m2[3][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator/(mat<3, 3, T, Q> const& m,	T scalar)
+	{
+		return mat<3, 3, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar,
+			m[2] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator/(T scalar, mat<3, 3, T, Q> const& m)
+	{
+		return mat<3, 3, T, Q>(
+			scalar / m[0],
+			scalar / m[1],
+			scalar / m[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 3, T, Q>::col_type operator/(mat<3, 3, T, Q> const& m, typename mat<3, 3, T, Q>::row_type const& v)
+	{
+		return  inverse(m) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 3, T, Q>::row_type operator/(typename mat<3, 3, T, Q>::col_type const& v, mat<3, 3, T, Q> const& m)
+	{
+		return v * inverse(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator/(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		mat<3, 3, T, Q> m1_copy(m1);
+		return m1_copy /= m2;
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]) && (m1[2] == m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<3, 3, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]) || (m1[2] != m2[2]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.hpp
new file mode 100644
index 000000000000..e2617a32952b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.hpp
@@ -0,0 +1,166 @@
+/// @ref core
+/// @file glm/detail/type_mat3x4.hpp
+
+#pragma once
+
+#include "type_vec3.hpp"
+#include "type_vec4.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<3, 4, T, Q>
+	{
+		typedef vec<4, T, Q> col_type;
+		typedef vec<3, T, Q> row_type;
+		typedef mat<3, 4, T, Q> type;
+		typedef mat<4, 3, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[3];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 3; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<3, 4, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T x0, T y0, T z0, T w0,
+			T x1, T y1, T z1, T w1,
+			T x2, T y2, T z2, T w2);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1,
+			col_type const& v2);
+
+		// -- Conversions --
+
+		template<
+			typename X1, typename Y1, typename Z1, typename W1,
+			typename X2, typename Y2, typename Z2, typename W2,
+			typename X3, typename Y3, typename Z3, typename W3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 x1, Y1 y1, Z1 z1, W1 w1,
+			X2 x2, Y2 y2, Z2 z2, W2 w2,
+			X3 x3, Y3 y3, Z3 z3, W3 w3);
+
+		template<typename V1, typename V2, typename V3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<4, V1, Q> const& v1,
+			vec<4, V2, Q> const& v2,
+			vec<4, V3, Q> const& v3);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator=(mat<3, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator+=(mat<3, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator-=(mat<3, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator/=(U s);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator+(mat<3, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator-(mat<3, 4, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator+(mat<3, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator+(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator-(mat<3, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator-(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<3, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator*(T scalar, mat<3, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 4, T, Q>::col_type operator*(mat<3, 4, T, Q> const& m, typename mat<3, 4, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<3, 4, T, Q>::row_type operator*(typename mat<3, 4, T, Q>::col_type const& v, mat<3, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<3, 4, T, Q> const& m1,	mat<4, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<3, 4, T, Q> const& m1, mat<2, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<3, 4, T, Q> const& m1,	mat<3, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator/(mat<3, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator/(T scalar, mat<3, 4, T, Q> const& m);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat3x4.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.inl
new file mode 100644
index 000000000000..5e2eccd85d9b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat3x4.inl
@@ -0,0 +1,578 @@
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0, 0, 0), col_type(0, 1, 0, 0), col_type(0, 0, 1, 0)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0, 0, 0);
+				this->value[1] = col_type(0, 1, 0, 0);
+				this->value[2] = col_type(0, 0, 1, 0);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<3, 4, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0, 0, 0), col_type(0, s, 0, 0), col_type(0, 0, s, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0, 0, 0);
+			this->value[1] = col_type(0, s, 0, 0);
+			this->value[2] = col_type(0, 0, s, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat
+	(
+		T x0, T y0, T z0, T w0,
+		T x1, T y1, T z1, T w1,
+		T x2, T y2, T z2, T w2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{
+				col_type(x0, y0, z0, w0),
+				col_type(x1, y1, z1, w1),
+				col_type(x2, y2, z2, w2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0, w0);
+			this->value[1] = col_type(x1, y1, z1, w1);
+			this->value[2] = col_type(x2, y2, z2, w2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(col_type const& v0, col_type const& v1, col_type const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+			this->value[2] = v2;
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X0, typename Y0, typename Z0, typename W0,
+		typename X1, typename Y1, typename Z1, typename W1,
+		typename X2, typename Y2, typename Z2, typename W2>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat
+	(
+		X0 x0, Y0 y0, Z0 z0, W0 w0,
+		X1 x1, Y1 y1, Z1 z1, W1 w1,
+		X2 x2, Y2 y2, Z2 z2, W2 w2
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{
+				col_type(x0, y0, z0, w0),
+				col_type(x1, y1, z1, w1),
+				col_type(x2, y2, z2, w2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0, w0);
+			this->value[1] = col_type(x1, y1, z1, w1);
+			this->value[2] = col_type(x2, y2, z2, w2);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2, typename V3>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(vec<4, V1, Q> const& v0, vec<4, V2, Q> const& v1, vec<4, V3, Q> const& v2)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v0);
+			this->value[1] = col_type(v1);
+			this->value[2] = col_type(v2);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<3, 4, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0), col_type(0, 0, 1, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+			this->value[2] = col_type(0, 0, 1, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(0, 0, 1, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(0, 0, 1, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0), col_type(m[2], 1, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+			this->value[2] = col_type(m[2], 1, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0, 0, 1, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0, 0, 1, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0), col_type(m[2], 1, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+			this->value[2] = col_type(m[2], 1, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 0);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 4, T, Q>::col_type & mat<3, 4, T, Q>::operator[](typename mat<3, 4, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 4, T, Q>::col_type const& mat<3, 4, T, Q>::operator[](typename mat<3, 4, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator=(mat<3, 4, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		this->value[2] = m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		this->value[2] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator+=(mat<3, 4, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		this->value[2] += m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		this->value[2] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator-=(mat<3, 4, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		this->value[2] -= m[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		this->value[2] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> & mat<3, 4, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		this->value[2] /= s;
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		++this->value[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q>& mat<3, 4, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		--this->value[2];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> mat<3, 4, T, Q>::operator++(int)
+	{
+		mat<3, 4, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> mat<3, 4, T, Q>::operator--(int)
+	{
+		mat<3, 4, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator+(mat<3, 4, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator-(mat<3, 4, T, Q> const& m)
+	{
+		return mat<3, 4, T, Q>(
+			-m[0],
+			-m[1],
+			-m[2]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator+(mat<3, 4, T, Q> const& m, T scalar)
+	{
+		return mat<3, 4, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator+(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		return mat<3, 4, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1],
+			m1[2] + m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator-(mat<3, 4, T, Q> const& m,	T scalar)
+	{
+		return mat<3, 4, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar,
+			m[2] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator-(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		return mat<3, 4, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1],
+			m1[2] - m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<3, 4, T, Q> const& m, T scalar)
+	{
+		return mat<3, 4, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator*(T scalar, mat<3, 4, T, Q> const& m)
+	{
+		return mat<3, 4, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 4, T, Q>::col_type operator*
+	(
+		mat<3, 4, T, Q> const& m,
+		typename mat<3, 4, T, Q>::row_type const& v
+	)
+	{
+		return typename mat<3, 4, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z,
+			m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z,
+			m[0][2] * v.x + m[1][2] * v.y + m[2][2] * v.z,
+			m[0][3] * v.x + m[1][3] * v.y + m[2][3] * v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<3, 4, T, Q>::row_type operator*
+	(
+		typename mat<3, 4, T, Q>::col_type const& v,
+		mat<3, 4, T, Q> const& m
+	)
+	{
+		return typename mat<3, 4, T, Q>::row_type(
+			v.x * m[0][0] + v.y * m[0][1] + v.z * m[0][2] + v.w * m[0][3],
+			v.x * m[1][0] + v.y * m[1][1] + v.z * m[1][2] + v.w * m[1][3],
+			v.x * m[2][0] + v.y * m[2][1] + v.z * m[2][2] + v.w * m[2][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<3, 4, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		const T SrcA00 = m1[0][0];
+		const T SrcA01 = m1[0][1];
+		const T SrcA02 = m1[0][2];
+		const T SrcA03 = m1[0][3];
+		const T SrcA10 = m1[1][0];
+		const T SrcA11 = m1[1][1];
+		const T SrcA12 = m1[1][2];
+		const T SrcA13 = m1[1][3];
+		const T SrcA20 = m1[2][0];
+		const T SrcA21 = m1[2][1];
+		const T SrcA22 = m1[2][2];
+		const T SrcA23 = m1[2][3];
+
+		const T SrcB00 = m2[0][0];
+		const T SrcB01 = m2[0][1];
+		const T SrcB02 = m2[0][2];
+		const T SrcB10 = m2[1][0];
+		const T SrcB11 = m2[1][1];
+		const T SrcB12 = m2[1][2];
+		const T SrcB20 = m2[2][0];
+		const T SrcB21 = m2[2][1];
+		const T SrcB22 = m2[2][2];
+		const T SrcB30 = m2[3][0];
+		const T SrcB31 = m2[3][1];
+		const T SrcB32 = m2[3][2];
+
+		mat<4, 4, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01 + SrcA20 * SrcB02;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01 + SrcA21 * SrcB02;
+		Result[0][2] = SrcA02 * SrcB00 + SrcA12 * SrcB01 + SrcA22 * SrcB02;
+		Result[0][3] = SrcA03 * SrcB00 + SrcA13 * SrcB01 + SrcA23 * SrcB02;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11 + SrcA20 * SrcB12;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11 + SrcA21 * SrcB12;
+		Result[1][2] = SrcA02 * SrcB10 + SrcA12 * SrcB11 + SrcA22 * SrcB12;
+		Result[1][3] = SrcA03 * SrcB10 + SrcA13 * SrcB11 + SrcA23 * SrcB12;
+		Result[2][0] = SrcA00 * SrcB20 + SrcA10 * SrcB21 + SrcA20 * SrcB22;
+		Result[2][1] = SrcA01 * SrcB20 + SrcA11 * SrcB21 + SrcA21 * SrcB22;
+		Result[2][2] = SrcA02 * SrcB20 + SrcA12 * SrcB21 + SrcA22 * SrcB22;
+		Result[2][3] = SrcA03 * SrcB20 + SrcA13 * SrcB21 + SrcA23 * SrcB22;
+		Result[3][0] = SrcA00 * SrcB30 + SrcA10 * SrcB31 + SrcA20 * SrcB32;
+		Result[3][1] = SrcA01 * SrcB30 + SrcA11 * SrcB31 + SrcA21 * SrcB32;
+		Result[3][2] = SrcA02 * SrcB30 + SrcA12 * SrcB31 + SrcA22 * SrcB32;
+		Result[3][3] = SrcA03 * SrcB30 + SrcA13 * SrcB31 + SrcA23 * SrcB32;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<3, 4, T, Q> const& m1, mat<2, 3, T, Q> const& m2)
+	{
+		return mat<2, 4, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2],
+			m1[0][3] * m2[0][0] + m1[1][3] * m2[0][1] + m1[2][3] * m2[0][2],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2],
+			m1[0][3] * m2[1][0] + m1[1][3] * m2[1][1] + m1[2][3] * m2[1][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<3, 4, T, Q> const& m1, mat<3, 3, T, Q> const& m2)
+	{
+		return mat<3, 4, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2],
+			m1[0][3] * m2[0][0] + m1[1][3] * m2[0][1] + m1[2][3] * m2[0][2],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2],
+			m1[0][3] * m2[1][0] + m1[1][3] * m2[1][1] + m1[2][3] * m2[1][2],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2],
+			m1[0][2] * m2[2][0] + m1[1][2] * m2[2][1] + m1[2][2] * m2[2][2],
+			m1[0][3] * m2[2][0] + m1[1][3] * m2[2][1] + m1[2][3] * m2[2][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator/(mat<3, 4, T, Q> const& m,	T scalar)
+	{
+		return mat<3, 4, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar,
+			m[2] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator/(T scalar, mat<3, 4, T, Q> const& m)
+	{
+		return mat<3, 4, T, Q>(
+			scalar / m[0],
+			scalar / m[1],
+			scalar / m[2]);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]) && (m1[2] == m2[2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<3, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]) || (m1[2] != m2[2]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.hpp
new file mode 100644
index 000000000000..5eb31184c8fd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.hpp
@@ -0,0 +1,171 @@
+/// @ref core
+/// @file glm/detail/type_mat4x2.hpp
+
+#pragma once
+
+#include "type_vec2.hpp"
+#include "type_vec4.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<4, 2, T, Q>
+	{
+		typedef vec<2, T, Q> col_type;
+		typedef vec<4, T, Q> row_type;
+		typedef mat<4, 2, T, Q> type;
+		typedef mat<2, 4, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[4];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 4; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<4, 2, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T x0, T y0,
+			T x1, T y1,
+			T x2, T y2,
+			T x3, T y3);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1,
+			col_type const& v2,
+			col_type const& v3);
+
+		// -- Conversions --
+
+		template<
+			typename X0, typename Y0,
+			typename X1, typename Y1,
+			typename X2, typename Y2,
+			typename X3, typename Y3>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X0 x0, Y0 y0,
+			X1 x1, Y1 y1,
+			X2 x2, Y2 y2,
+			X3 x3, Y3 y3);
+
+		template<typename V1, typename V2, typename V3, typename V4>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<2, V1, Q> const& v1,
+			vec<2, V2, Q> const& v2,
+			vec<2, V3, Q> const& v3,
+			vec<2, V4, Q> const& v4);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator=(mat<4, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator+=(mat<4, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator-=(mat<4, 2, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator/=(U s);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator++ ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> & operator-- ();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator+(mat<4, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator-(mat<4, 2, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator+(mat<4, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator+(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator-(mat<4, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator-(mat<4, 2, T, Q> const& m1,	mat<4, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<4, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator*(T scalar, mat<4, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 2, T, Q>::col_type operator*(mat<4, 2, T, Q> const& m, typename mat<4, 2, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 2, T, Q>::row_type operator*(typename mat<4, 2, T, Q>::col_type const& v, mat<4, 2, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<4, 2, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<4, 2, T, Q> const& m1, mat<3, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<4, 2, T, Q> const& m1, mat<4, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator/(mat<4, 2, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 2, T, Q> operator/(T scalar, mat<4, 2, T, Q> const& m);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat4x2.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.inl
new file mode 100644
index 000000000000..1a4f3f168d01
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x2.inl
@@ -0,0 +1,574 @@
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0), col_type(0, 1), col_type(0, 0), col_type(0, 0)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0);
+				this->value[1] = col_type(0, 1);
+				this->value[2] = col_type(0, 0);
+				this->value[3] = col_type(0, 0);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<4, 2, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+			this->value[3] = m[3];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0), col_type(0, s), col_type(0, 0), col_type(0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0);
+			this->value[1] = col_type(0, s);
+			this->value[2] = col_type(0, 0);
+			this->value[3] = col_type(0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat
+	(
+		T x0, T y0,
+		T x1, T y1,
+		T x2, T y2,
+		T x3, T y3
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0), col_type(x1, y1), col_type(x2, y2), col_type(x3, y3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0);
+			this->value[1] = col_type(x1, y1);
+			this->value[2] = col_type(x2, y2);
+			this->value[3] = col_type(x3, y3);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(col_type const& v0, col_type const& v1, col_type const& v2, col_type const& v3)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2), col_type(v3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+			this->value[2] = v2;
+			this->value[3] = v3;
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X0, typename Y0,
+		typename X1, typename Y1,
+		typename X2, typename Y2,
+		typename X3, typename Y3>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat
+	(
+		X0 x0, Y0 y0,
+		X1 x1, Y1 y1,
+		X2 x2, Y2 y2,
+		X3 x3, Y3 y3
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0), col_type(x1, y1), col_type(x2, y2), col_type(x3, y3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0);
+			this->value[1] = col_type(x1, y1);
+			this->value[2] = col_type(x2, y2);
+			this->value[3] = col_type(x3, y3);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V0, typename V1, typename V2, typename V3>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(vec<2, V0, Q> const& v0, vec<2, V1, Q> const& v1, vec<2, V2, Q> const& v2, vec<2, V3, Q> const& v3)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2), col_type(v3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v0);
+			this->value[1] = col_type(v1);
+			this->value[2] = col_type(v2);
+			this->value[3] = col_type(v3);
+#		endif
+	}
+
+	// -- Conversion --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<4, 2, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(m[3]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(m[3]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+				this->value[0] = col_type(m[0]);
+				this->value[1] = col_type(m[1]);
+				this->value[2] = col_type(m[2]);
+				this->value[3] = col_type(m[3]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 2, T, Q>::col_type & mat<4, 2, T, Q>::operator[](typename mat<4, 2, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 2, T, Q>::col_type const& mat<4, 2, T, Q>::operator[](typename mat<4, 2, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q>& mat<4, 2, T, Q>::operator=(mat<4, 2, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		this->value[2] = m[2];
+		this->value[3] = m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		this->value[2] += s;
+		this->value[3] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator+=(mat<4, 2, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		this->value[2] += m[2];
+		this->value[3] += m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		this->value[2] -= s;
+		this->value[3] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator-=(mat<4, 2, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		this->value[2] -= m[2];
+		this->value[3] -= m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		this->value[2] *= s;
+		this->value[3] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		this->value[2] /= s;
+		this->value[3] /= s;
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		++this->value[2];
+		++this->value[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> & mat<4, 2, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		--this->value[2];
+		--this->value[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> mat<4, 2, T, Q>::operator++(int)
+	{
+		mat<4, 2, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> mat<4, 2, T, Q>::operator--(int)
+	{
+		mat<4, 2, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator+(mat<4, 2, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator-(mat<4, 2, T, Q> const& m)
+	{
+		return mat<4, 2, T, Q>(
+			-m[0],
+			-m[1],
+			-m[2],
+			-m[3]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator+(mat<4, 2, T, Q> const& m, T scalar)
+	{
+		return mat<4, 2, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar,
+			m[3] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator+(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		return mat<4, 2, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1],
+			m1[2] + m2[2],
+			m1[3] + m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator-(mat<4, 2, T, Q> const& m, T scalar)
+	{
+		return mat<4, 2, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar,
+			m[2] - scalar,
+			m[3] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator-(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		return mat<4, 2, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1],
+			m1[2] - m2[2],
+			m1[3] - m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<4, 2, T, Q> const& m, T scalar)
+	{
+		return mat<4, 2, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar,
+			m[3] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator*(T scalar, mat<4, 2, T, Q> const& m)
+	{
+		return mat<4, 2, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar,
+			m[3] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 2, T, Q>::col_type operator*(mat<4, 2, T, Q> const& m, typename mat<4, 2, T, Q>::row_type const& v)
+	{
+		return typename mat<4, 2, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z + m[3][0] * v.w,
+			m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z + m[3][1] * v.w);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 2, T, Q>::row_type operator*(typename mat<4, 2, T, Q>::col_type const& v, mat<4, 2, T, Q> const& m)
+	{
+		return typename mat<4, 2, T, Q>::row_type(
+			v.x * m[0][0] + v.y * m[0][1],
+			v.x * m[1][0] + v.y * m[1][1],
+			v.x * m[2][0] + v.y * m[2][1],
+			v.x * m[3][0] + v.y * m[3][1]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 2, T, Q> operator*(mat<4, 2, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		T const SrcA00 = m1[0][0];
+		T const SrcA01 = m1[0][1];
+		T const SrcA10 = m1[1][0];
+		T const SrcA11 = m1[1][1];
+		T const SrcA20 = m1[2][0];
+		T const SrcA21 = m1[2][1];
+		T const SrcA30 = m1[3][0];
+		T const SrcA31 = m1[3][1];
+
+		T const SrcB00 = m2[0][0];
+		T const SrcB01 = m2[0][1];
+		T const SrcB02 = m2[0][2];
+		T const SrcB03 = m2[0][3];
+		T const SrcB10 = m2[1][0];
+		T const SrcB11 = m2[1][1];
+		T const SrcB12 = m2[1][2];
+		T const SrcB13 = m2[1][3];
+
+		mat<2, 2, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01 + SrcA20 * SrcB02 + SrcA30 * SrcB03;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01 + SrcA21 * SrcB02 + SrcA31 * SrcB03;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11 + SrcA20 * SrcB12 + SrcA30 * SrcB13;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11 + SrcA21 * SrcB12 + SrcA31 * SrcB13;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 2, T, Q> operator*(mat<4, 2, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		return mat<3, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2] + m1[3][0] * m2[0][3],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2] + m1[3][1] * m2[0][3],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2] + m1[3][0] * m2[1][3],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2] + m1[3][1] * m2[1][3],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2] + m1[3][0] * m2[2][3],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2] + m1[3][1] * m2[2][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator*(mat<4, 2, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		return mat<4, 2, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2] + m1[3][0] * m2[0][3],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2] + m1[3][1] * m2[0][3],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2] + m1[3][0] * m2[1][3],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2] + m1[3][1] * m2[1][3],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2] + m1[3][0] * m2[2][3],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2] + m1[3][1] * m2[2][3],
+			m1[0][0] * m2[3][0] + m1[1][0] * m2[3][1] + m1[2][0] * m2[3][2] + m1[3][0] * m2[3][3],
+			m1[0][1] * m2[3][0] + m1[1][1] * m2[3][1] + m1[2][1] * m2[3][2] + m1[3][1] * m2[3][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator/(mat<4, 2, T, Q> const& m, T scalar)
+	{
+		return mat<4, 2, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar,
+			m[2] / scalar,
+			m[3] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 2, T, Q> operator/(T scalar, mat<4, 2, T, Q> const& m)
+	{
+		return mat<4, 2, T, Q>(
+			scalar / m[0],
+			scalar / m[1],
+			scalar / m[2],
+			scalar / m[3]);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]) && (m1[2] == m2[2]) && (m1[3] == m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<4, 2, T, Q> const& m1, mat<4, 2, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]) || (m1[2] != m2[2]) || (m1[3] != m2[3]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.hpp
new file mode 100644
index 000000000000..9650546a6305
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.hpp
@@ -0,0 +1,171 @@
+/// @ref core
+/// @file glm/detail/type_mat4x3.hpp
+
+#pragma once
+
+#include "type_vec3.hpp"
+#include "type_vec4.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<4, 3, T, Q>
+	{
+		typedef vec<3, T, Q> col_type;
+		typedef vec<4, T, Q> row_type;
+		typedef mat<4, 3, T, Q> type;
+		typedef mat<3, 4, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[4];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length() { return 4; }
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<4, 3, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T s);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T const& x0, T const& y0, T const& z0,
+			T const& x1, T const& y1, T const& z1,
+			T const& x2, T const& y2, T const& z2,
+			T const& x3, T const& y3, T const& z3);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1,
+			col_type const& v2,
+			col_type const& v3);
+
+		// -- Conversions --
+
+		template<
+			typename X1, typename Y1, typename Z1,
+			typename X2, typename Y2, typename Z2,
+			typename X3, typename Y3, typename Z3,
+			typename X4, typename Y4, typename Z4>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 const& x1, Y1 const& y1, Z1 const& z1,
+			X2 const& x2, Y2 const& y2, Z2 const& z2,
+			X3 const& x3, Y3 const& y3, Z3 const& z3,
+			X4 const& x4, Y4 const& y4, Z4 const& z4);
+
+		template<typename V1, typename V2, typename V3, typename V4>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<3, V1, Q> const& v1,
+			vec<3, V2, Q> const& v2,
+			vec<3, V3, Q> const& v3,
+			vec<3, V4, Q> const& v4);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator=(mat<4, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator+=(mat<4, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator-=(mat<4, 3, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> & operator/=(U s);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q>& operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q>& operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator+(mat<4, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator-(mat<4, 3, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator+(mat<4, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator+(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator-(mat<4, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator-(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<4, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator*(T scalar, mat<4, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 3, T, Q>::col_type operator*(mat<4, 3, T, Q> const& m, typename mat<4, 3, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 3, T, Q>::row_type operator*(typename mat<4, 3, T, Q>::col_type const& v, mat<4, 3, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<4, 3, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<4, 3, T, Q> const& m1,	mat<3, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<4, 3, T, Q> const& m1, mat<4, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator/(mat<4, 3, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 3, T, Q> operator/(T scalar, mat<4, 3, T, Q> const& m);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat4x3.inl"
+#endif //GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.inl
new file mode 100644
index 000000000000..c2fe3a44532d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x3.inl
@@ -0,0 +1,598 @@
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0, 0), col_type(0, 1, 0), col_type(0, 0, 1), col_type(0, 0, 0)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0, 0);
+				this->value[1] = col_type(0, 1, 0);
+				this->value[2] = col_type(0, 0, 1);
+				this->value[3] = col_type(0, 0, 0);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<4, 3, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+			this->value[3] = m[3];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0, 0), col_type(0, s, 0), col_type(0, 0, s), col_type(0, 0, 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0, 0);
+			this->value[1] = col_type(0, s, 0);
+			this->value[2] = col_type(0, 0, s);
+			this->value[3] = col_type(0, 0, 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat
+	(
+		T const& x0, T const& y0, T const& z0,
+		T const& x1, T const& y1, T const& z1,
+		T const& x2, T const& y2, T const& z2,
+		T const& x3, T const& y3, T const& z3
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0, z0), col_type(x1, y1, z1), col_type(x2, y2, z2), col_type(x3, y3, z3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0);
+			this->value[1] = col_type(x1, y1, z1);
+			this->value[2] = col_type(x2, y2, z2);
+			this->value[3] = col_type(x3, y3, z3);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(col_type const& v0, col_type const& v1, col_type const& v2, col_type const& v3)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2), col_type(v3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+			this->value[2] = v2;
+			this->value[3] = v3;
+#		endif
+	}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X0, typename Y0, typename Z0,
+		typename X1, typename Y1, typename Z1,
+		typename X2, typename Y2, typename Z2,
+		typename X3, typename Y3, typename Z3>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat
+	(
+		X0 const& x0, Y0 const& y0, Z0 const& z0,
+		X1 const& x1, Y1 const& y1, Z1 const& z1,
+		X2 const& x2, Y2 const& y2, Z2 const& z2,
+		X3 const& x3, Y3 const& y3, Z3 const& z3
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x0, y0, z0), col_type(x1, y1, z1), col_type(x2, y2, z2), col_type(x3, y3, z3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0);
+			this->value[1] = col_type(x1, y1, z1);
+			this->value[2] = col_type(x2, y2, z2);
+			this->value[3] = col_type(x3, y3, z3);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2, typename V3, typename V4>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(vec<3, V1, Q> const& v1, vec<3, V2, Q> const& v2, vec<3, V3, Q> const& v3, vec<3, V4, Q> const& v4)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v1), col_type(v2), col_type(v3), col_type(v4)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v1);
+			this->value[1] = col_type(v2);
+			this->value[2] = col_type(v3);
+			this->value[3] = col_type(v4);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<4, 3, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(m[3]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(0, 0, 1), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(0, 0, 1);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<4, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(m[3]);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0, 0, 1), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0, 0, 1);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 1), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 1);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0, 0, 1), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(0, 0, 1);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 1), col_type(m[3], 0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 1);
+			this->value[3] = col_type(m[3], 0);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(0)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(0);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 3, T, Q>::col_type & mat<4, 3, T, Q>::operator[](typename mat<4, 3, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 3, T, Q>::col_type const& mat<4, 3, T, Q>::operator[](typename mat<4, 3, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary updatable operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q>& mat<4, 3, T, Q>::operator=(mat<4, 3, U, Q> const& m)
+	{
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		this->value[2] = m[2];
+		this->value[3] = m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		this->value[2] += s;
+		this->value[3] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator+=(mat<4, 3, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		this->value[2] += m[2];
+		this->value[3] += m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		this->value[2] -= s;
+		this->value[3] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator-=(mat<4, 3, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		this->value[2] -= m[2];
+		this->value[3] -= m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		this->value[2] *= s;
+		this->value[3] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		this->value[2] /= s;
+		this->value[3] /= s;
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		++this->value[2];
+		++this->value[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> & mat<4, 3, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		--this->value[2];
+		--this->value[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> mat<4, 3, T, Q>::operator++(int)
+	{
+		mat<4, 3, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> mat<4, 3, T, Q>::operator--(int)
+	{
+		mat<4, 3, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator+(mat<4, 3, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator-(mat<4, 3, T, Q> const& m)
+	{
+		return mat<4, 3, T, Q>(
+			-m[0],
+			-m[1],
+			-m[2],
+			-m[3]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator+(mat<4, 3, T, Q> const& m, T scalar)
+	{
+		return mat<4, 3, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar,
+			m[3] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator+(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		return mat<4, 3, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1],
+			m1[2] + m2[2],
+			m1[3] + m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator-(mat<4, 3, T, Q> const& m, T scalar)
+	{
+		return mat<4, 3, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar,
+			m[2] - scalar,
+			m[3] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator-(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		return mat<4, 3, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1],
+			m1[2] - m2[2],
+			m1[3] - m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<4, 3, T, Q> const& m, T scalar)
+	{
+		return mat<4, 3, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar,
+			m[3] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator*(T scalar, mat<4, 3, T, Q> const& m)
+	{
+		return mat<4, 3, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar,
+			m[3] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 3, T, Q>::col_type operator*
+	(
+		mat<4, 3, T, Q> const& m,
+		typename mat<4, 3, T, Q>::row_type const& v)
+	{
+		return typename mat<4, 3, T, Q>::col_type(
+			m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z + m[3][0] * v.w,
+			m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z + m[3][1] * v.w,
+			m[0][2] * v.x + m[1][2] * v.y + m[2][2] * v.z + m[3][2] * v.w);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 3, T, Q>::row_type operator*
+	(
+		typename mat<4, 3, T, Q>::col_type const& v,
+		mat<4, 3, T, Q> const& m)
+	{
+		return typename mat<4, 3, T, Q>::row_type(
+			v.x * m[0][0] + v.y * m[0][1] + v.z * m[0][2],
+			v.x * m[1][0] + v.y * m[1][1] + v.z * m[1][2],
+			v.x * m[2][0] + v.y * m[2][1] + v.z * m[2][2],
+			v.x * m[3][0] + v.y * m[3][1] + v.z * m[3][2]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 3, T, Q> operator*(mat<4, 3, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		return mat<2, 3, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2] + m1[3][0] * m2[0][3],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2] + m1[3][1] * m2[0][3],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2] + m1[3][2] * m2[0][3],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2] + m1[3][0] * m2[1][3],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2] + m1[3][1] * m2[1][3],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2] + m1[3][2] * m2[1][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 3, T, Q> operator*(mat<4, 3, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		T const SrcA00 = m1[0][0];
+		T const SrcA01 = m1[0][1];
+		T const SrcA02 = m1[0][2];
+		T const SrcA10 = m1[1][0];
+		T const SrcA11 = m1[1][1];
+		T const SrcA12 = m1[1][2];
+		T const SrcA20 = m1[2][0];
+		T const SrcA21 = m1[2][1];
+		T const SrcA22 = m1[2][2];
+		T const SrcA30 = m1[3][0];
+		T const SrcA31 = m1[3][1];
+		T const SrcA32 = m1[3][2];
+
+		T const SrcB00 = m2[0][0];
+		T const SrcB01 = m2[0][1];
+		T const SrcB02 = m2[0][2];
+		T const SrcB03 = m2[0][3];
+		T const SrcB10 = m2[1][0];
+		T const SrcB11 = m2[1][1];
+		T const SrcB12 = m2[1][2];
+		T const SrcB13 = m2[1][3];
+		T const SrcB20 = m2[2][0];
+		T const SrcB21 = m2[2][1];
+		T const SrcB22 = m2[2][2];
+		T const SrcB23 = m2[2][3];
+
+		mat<3, 3, T, Q> Result;
+		Result[0][0] = SrcA00 * SrcB00 + SrcA10 * SrcB01 + SrcA20 * SrcB02 + SrcA30 * SrcB03;
+		Result[0][1] = SrcA01 * SrcB00 + SrcA11 * SrcB01 + SrcA21 * SrcB02 + SrcA31 * SrcB03;
+		Result[0][2] = SrcA02 * SrcB00 + SrcA12 * SrcB01 + SrcA22 * SrcB02 + SrcA32 * SrcB03;
+		Result[1][0] = SrcA00 * SrcB10 + SrcA10 * SrcB11 + SrcA20 * SrcB12 + SrcA30 * SrcB13;
+		Result[1][1] = SrcA01 * SrcB10 + SrcA11 * SrcB11 + SrcA21 * SrcB12 + SrcA31 * SrcB13;
+		Result[1][2] = SrcA02 * SrcB10 + SrcA12 * SrcB11 + SrcA22 * SrcB12 + SrcA32 * SrcB13;
+		Result[2][0] = SrcA00 * SrcB20 + SrcA10 * SrcB21 + SrcA20 * SrcB22 + SrcA30 * SrcB23;
+		Result[2][1] = SrcA01 * SrcB20 + SrcA11 * SrcB21 + SrcA21 * SrcB22 + SrcA31 * SrcB23;
+		Result[2][2] = SrcA02 * SrcB20 + SrcA12 * SrcB21 + SrcA22 * SrcB22 + SrcA32 * SrcB23;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator*(mat<4, 3, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		return mat<4, 3, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2] + m1[3][0] * m2[0][3],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2] + m1[3][1] * m2[0][3],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2] + m1[3][2] * m2[0][3],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2] + m1[3][0] * m2[1][3],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2] + m1[3][1] * m2[1][3],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2] + m1[3][2] * m2[1][3],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2] + m1[3][0] * m2[2][3],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2] + m1[3][1] * m2[2][3],
+			m1[0][2] * m2[2][0] + m1[1][2] * m2[2][1] + m1[2][2] * m2[2][2] + m1[3][2] * m2[2][3],
+			m1[0][0] * m2[3][0] + m1[1][0] * m2[3][1] + m1[2][0] * m2[3][2] + m1[3][0] * m2[3][3],
+			m1[0][1] * m2[3][0] + m1[1][1] * m2[3][1] + m1[2][1] * m2[3][2] + m1[3][1] * m2[3][3],
+			m1[0][2] * m2[3][0] + m1[1][2] * m2[3][1] + m1[2][2] * m2[3][2] + m1[3][2] * m2[3][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator/(mat<4, 3, T, Q> const& m, T scalar)
+	{
+		return mat<4, 3, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar,
+			m[2] / scalar,
+			m[3] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 3, T, Q> operator/(T scalar, mat<4, 3, T, Q> const& m)
+	{
+		return mat<4, 3, T, Q>(
+			scalar / m[0],
+			scalar / m[1],
+			scalar / m[2],
+			scalar / m[3]);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]) && (m1[2] == m2[2]) && (m1[3] == m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<4, 3, T, Q> const& m1, mat<4, 3, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]) || (m1[2] != m2[2]) || (m1[3] != m2[3]);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.hpp
new file mode 100644
index 000000000000..14387e2f9b33
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.hpp
@@ -0,0 +1,189 @@
+/// @ref core
+/// @file glm/detail/type_mat4x4.hpp
+
+#pragma once
+
+#include "type_vec4.hpp"
+#include <limits>
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct mat<4, 4, T, Q>
+	{
+		typedef vec<4, T, Q> col_type;
+		typedef vec<4, T, Q> row_type;
+		typedef mat<4, 4, T, Q> type;
+		typedef mat<4, 4, T, Q> transpose_type;
+		typedef T value_type;
+
+	private:
+		col_type value[4];
+
+	public:
+		// -- Accesses --
+
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 4;}
+
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type & operator[](length_type i) GLM_NOEXCEPT;
+		GLM_FUNC_DECL GLM_CONSTEXPR col_type const& operator[](length_type i) const GLM_NOEXCEPT;
+
+		// -- Constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR mat() GLM_DEFAULT_CTOR;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(mat<4, 4, T, P> const& m);
+
+		GLM_FUNC_DECL explicit GLM_CONSTEXPR mat(T s);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			T const& x0, T const& y0, T const& z0, T const& w0,
+			T const& x1, T const& y1, T const& z1, T const& w1,
+			T const& x2, T const& y2, T const& z2, T const& w2,
+			T const& x3, T const& y3, T const& z3, T const& w3);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			col_type const& v0,
+			col_type const& v1,
+			col_type const& v2,
+			col_type const& v3);
+
+		// -- Conversions --
+
+		template<
+			typename X1, typename Y1, typename Z1, typename W1,
+			typename X2, typename Y2, typename Z2, typename W2,
+			typename X3, typename Y3, typename Z3, typename W3,
+			typename X4, typename Y4, typename Z4, typename W4>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			X1 const& x1, Y1 const& y1, Z1 const& z1, W1 const& w1,
+			X2 const& x2, Y2 const& y2, Z2 const& z2, W2 const& w2,
+			X3 const& x3, Y3 const& y3, Z3 const& z3, W3 const& w3,
+			X4 const& x4, Y4 const& y4, Z4 const& z4, W4 const& w4);
+
+		template<typename V1, typename V2, typename V3, typename V4>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat(
+			vec<4, V1, Q> const& v1,
+			vec<4, V2, Q> const& v2,
+			vec<4, V3, Q> const& v3,
+			vec<4, V4, Q> const& v4);
+
+		// -- Matrix conversions --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 4, U, P> const& m);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 3, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<2, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 2, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<3, 4, T, Q> const& x);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR mat(mat<4, 3, T, Q> const& x);
+
+		// -- Unary arithmetic operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator=(mat<4, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator+=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator+=(mat<4, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator-=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator-=(mat<4, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator*=(mat<4, 4, U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator/=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator/=(mat<4, 4, U, Q> const& m);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator--(int);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator+(mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator-(mat<4, 4, T, Q> const& m);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator+(mat<4, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator+(T scalar, mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator+(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator-(mat<4, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator-(T scalar, mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator-(mat<4, 4, T, Q> const& m1,	mat<4, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<4, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator*(T scalar, mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 4, T, Q>::col_type operator*(mat<4, 4, T, Q> const& m, typename mat<4, 4, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 4, T, Q>::row_type operator*(typename mat<4, 4, T, Q>::col_type const& v, mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator/(mat<4, 4, T, Q> const& m, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator/(T scalar, mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 4, T, Q>::col_type operator/(mat<4, 4, T, Q> const& m, typename mat<4, 4, T, Q>::row_type const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR typename mat<4, 4, T, Q>::row_type operator/(typename mat<4, 4, T, Q>::col_type const& v, mat<4, 4, T, Q> const& m);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> operator/(mat<4, 4, T, Q> const& m1,	mat<4, 4, T, Q> const& m2);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_mat4x4.inl"
+#endif//GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.inl
new file mode 100644
index 000000000000..db77673163a7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4.inl
@@ -0,0 +1,706 @@
+#include "../matrix.hpp"
+
+namespace glm
+{
+	// -- Constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat()
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALIZER_LIST
+				: value{col_type(1, 0, 0, 0), col_type(0, 1, 0, 0), col_type(0, 0, 1, 0), col_type(0, 0, 0, 1)}
+#			endif
+		{
+#			if GLM_CONFIG_CTOR_INIT == GLM_CTOR_INITIALISATION
+				this->value[0] = col_type(1, 0, 0, 0);
+				this->value[1] = col_type(0, 1, 0, 0);
+				this->value[2] = col_type(0, 0, 1, 0);
+				this->value[3] = col_type(0, 0, 0, 1);
+#			endif
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<4, 4, T, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+			this->value[3] = m[3];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(T s)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(s, 0, 0, 0), col_type(0, s, 0, 0), col_type(0, 0, s, 0), col_type(0, 0, 0, s)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(s, 0, 0, 0);
+			this->value[1] = col_type(0, s, 0, 0);
+			this->value[2] = col_type(0, 0, s, 0);
+			this->value[3] = col_type(0, 0, 0, s);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat
+	(
+		T const& x0, T const& y0, T const& z0, T const& w0,
+		T const& x1, T const& y1, T const& z1, T const& w1,
+		T const& x2, T const& y2, T const& z2, T const& w2,
+		T const& x3, T const& y3, T const& z3, T const& w3
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{
+				col_type(x0, y0, z0, w0),
+				col_type(x1, y1, z1, w1),
+				col_type(x2, y2, z2, w2),
+				col_type(x3, y3, z3, w3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x0, y0, z0, w0);
+			this->value[1] = col_type(x1, y1, z1, w1);
+			this->value[2] = col_type(x2, y2, z2, w2);
+			this->value[3] = col_type(x3, y3, z3, w3);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(col_type const& v0, col_type const& v1, col_type const& v2, col_type const& v3)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v0), col_type(v1), col_type(v2), col_type(v3)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = v0;
+			this->value[1] = v1;
+			this->value[2] = v2;
+			this->value[3] = v3;
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<4, 4, U, P> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(m[3])}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0]);
+			this->value[1] = col_type(m[1]);
+			this->value[2] = col_type(m[2]);
+			this->value[3] = col_type(m[3]);
+#		endif
+	}
+
+	// -- Conversions --
+
+	template<typename T, qualifier Q>
+	template<
+		typename X1, typename Y1, typename Z1, typename W1,
+		typename X2, typename Y2, typename Z2, typename W2,
+		typename X3, typename Y3, typename Z3, typename W3,
+		typename X4, typename Y4, typename Z4, typename W4>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat
+	(
+		X1 const& x1, Y1 const& y1, Z1 const& z1, W1 const& w1,
+		X2 const& x2, Y2 const& y2, Z2 const& z2, W2 const& w2,
+		X3 const& x3, Y3 const& y3, Z3 const& z3, W3 const& w3,
+		X4 const& x4, Y4 const& y4, Z4 const& z4, W4 const& w4
+	)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(x1, y1, z1, w1), col_type(x2, y2, z2, w2), col_type(x3, y3, z3, w3), col_type(x4, y4, z4, w4)}
+#		endif
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<X1>::is_iec559 || std::numeric_limits<X1>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 1st parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Y1>::is_iec559 || std::numeric_limits<Y1>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 2nd parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Z1>::is_iec559 || std::numeric_limits<Z1>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 3rd parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<W1>::is_iec559 || std::numeric_limits<W1>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 4th parameter type invalid.");
+
+		GLM_STATIC_ASSERT(std::numeric_limits<X2>::is_iec559 || std::numeric_limits<X2>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 5th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Y2>::is_iec559 || std::numeric_limits<Y2>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 6th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Z2>::is_iec559 || std::numeric_limits<Z2>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 7th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<W2>::is_iec559 || std::numeric_limits<W2>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 8th parameter type invalid.");
+
+		GLM_STATIC_ASSERT(std::numeric_limits<X3>::is_iec559 || std::numeric_limits<X3>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 9th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Y3>::is_iec559 || std::numeric_limits<Y3>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 10th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Z3>::is_iec559 || std::numeric_limits<Z3>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 11th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<W3>::is_iec559 || std::numeric_limits<W3>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 12th parameter type invalid.");
+
+		GLM_STATIC_ASSERT(std::numeric_limits<X4>::is_iec559 || std::numeric_limits<X4>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 13th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Y4>::is_iec559 || std::numeric_limits<Y4>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 14th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<Z4>::is_iec559 || std::numeric_limits<Z4>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 15th parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<W4>::is_iec559 || std::numeric_limits<W4>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 16th parameter type invalid.");
+
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(x1, y1, z1, w1);
+			this->value[1] = col_type(x2, y2, z2, w2);
+			this->value[2] = col_type(x3, y3, z3, w3);
+			this->value[3] = col_type(x4, y4, z4, w4);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	template<typename V1, typename V2, typename V3, typename V4>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(vec<4, V1, Q> const& v1, vec<4, V2, Q> const& v2, vec<4, V3, Q> const& v3, vec<4, V4, Q> const& v4)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(v1), col_type(v2), col_type(v3), col_type(v4)}
+#		endif
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<V1>::is_iec559 || std::numeric_limits<V1>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 1st parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<V2>::is_iec559 || std::numeric_limits<V2>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 2nd parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<V3>::is_iec559 || std::numeric_limits<V3>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 3rd parameter type invalid.");
+		GLM_STATIC_ASSERT(std::numeric_limits<V4>::is_iec559 || std::numeric_limits<V4>::is_integer || GLM_CONFIG_UNRESTRICTED_GENTYPE, "*mat4x4 constructor only takes float and integer types, 4th parameter type invalid.");
+
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(v1);
+			this->value[1] = col_type(v2);
+			this->value[2] = col_type(v3);
+			this->value[3] = col_type(v4);
+#		endif
+	}
+
+	// -- Matrix conversions --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<2, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0), col_type(0, 0, 1, 0), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+			this->value[2] = col_type(0, 0, 1, 0);
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<3, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 0), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 0);
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<2, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(0, 0, 1, 0), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(0, 0, 1, 0);
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<3, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0), col_type(m[2], 1, 0), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+			this->value[2] = col_type(m[2], 1, 0);
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<2, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(0, 0, 1, 0), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = col_type(0, 0, 1, 0);
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<4, 2, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0, 0), col_type(m[1], 0, 0), col_type(0, 0, 1, 0), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0, 0);
+			this->value[1] = col_type(m[1], 0, 0);
+			this->value[2] = col_type(0, 0, 1, 0);
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<3, 4, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0]), col_type(m[1]), col_type(m[2]), col_type(0, 0, 0, 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = m[0];
+			this->value[1] = m[1];
+			this->value[2] = m[2];
+			this->value[3] = col_type(0, 0, 0, 1);
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>::mat(mat<4, 3, T, Q> const& m)
+#		if GLM_HAS_INITIALIZER_LISTS
+			: value{col_type(m[0], 0), col_type(m[1], 0), col_type(m[2], 0), col_type(m[3], 1)}
+#		endif
+	{
+#		if !GLM_HAS_INITIALIZER_LISTS
+			this->value[0] = col_type(m[0], 0);
+			this->value[1] = col_type(m[1], 0);
+			this->value[2] = col_type(m[2], 0);
+			this->value[3] = col_type(m[3], 1);
+#		endif
+	}
+
+	// -- Accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 4, T, Q>::col_type & mat<4, 4, T, Q>::operator[](typename mat<4, 4, T, Q>::length_type i) GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 4, T, Q>::col_type const& mat<4, 4, T, Q>::operator[](typename mat<4, 4, T, Q>::length_type i) const GLM_NOEXCEPT
+	{
+		assert(i < this->length());
+		return this->value[i];
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>& mat<4, 4, T, Q>::operator=(mat<4, 4, U, Q> const& m)
+	{
+		//memcpy could be faster
+		//memcpy(&this->value, &m.value, 16 * sizeof(valType));
+		this->value[0] = m[0];
+		this->value[1] = m[1];
+		this->value[2] = m[2];
+		this->value[3] = m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>& mat<4, 4, T, Q>::operator+=(U s)
+	{
+		this->value[0] += s;
+		this->value[1] += s;
+		this->value[2] += s;
+		this->value[3] += s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q>& mat<4, 4, T, Q>::operator+=(mat<4, 4, U, Q> const& m)
+	{
+		this->value[0] += m[0];
+		this->value[1] += m[1];
+		this->value[2] += m[2];
+		this->value[3] += m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator-=(U s)
+	{
+		this->value[0] -= s;
+		this->value[1] -= s;
+		this->value[2] -= s;
+		this->value[3] -= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator-=(mat<4, 4, U, Q> const& m)
+	{
+		this->value[0] -= m[0];
+		this->value[1] -= m[1];
+		this->value[2] -= m[2];
+		this->value[3] -= m[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator*=(U s)
+	{
+		this->value[0] *= s;
+		this->value[1] *= s;
+		this->value[2] *= s;
+		this->value[3] *= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator*=(mat<4, 4, U, Q> const& m)
+	{
+		return (*this = *this * m);
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator/=(U s)
+	{
+		this->value[0] /= s;
+		this->value[1] /= s;
+		this->value[2] /= s;
+		this->value[3] /= s;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator/=(mat<4, 4, U, Q> const& m)
+	{
+		return *this *= inverse(m);
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator++()
+	{
+		++this->value[0];
+		++this->value[1];
+		++this->value[2];
+		++this->value[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> & mat<4, 4, T, Q>::operator--()
+	{
+		--this->value[0];
+		--this->value[1];
+		--this->value[2];
+		--this->value[3];
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> mat<4, 4, T, Q>::operator++(int)
+	{
+		mat<4, 4, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> mat<4, 4, T, Q>::operator--(int)
+	{
+		mat<4, 4, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary constant operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator+(mat<4, 4, T, Q> const& m)
+	{
+		return m;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator-(mat<4, 4, T, Q> const& m)
+	{
+		return mat<4, 4, T, Q>(
+			-m[0],
+			-m[1],
+			-m[2],
+			-m[3]);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator+(mat<4, 4, T, Q> const& m, T scalar)
+	{
+		return mat<4, 4, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar,
+			m[3] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator+(T scalar, mat<4, 4, T, Q> const& m)
+	{
+		return mat<4, 4, T, Q>(
+			m[0] + scalar,
+			m[1] + scalar,
+			m[2] + scalar,
+			m[3] + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator+(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		return mat<4, 4, T, Q>(
+			m1[0] + m2[0],
+			m1[1] + m2[1],
+			m1[2] + m2[2],
+			m1[3] + m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator-(mat<4, 4, T, Q> const& m, T scalar)
+	{
+		return mat<4, 4, T, Q>(
+			m[0] - scalar,
+			m[1] - scalar,
+			m[2] - scalar,
+			m[3] - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator-(T scalar, mat<4, 4, T, Q> const& m)
+	{
+		return mat<4, 4, T, Q>(
+			scalar - m[0],
+			scalar - m[1],
+			scalar - m[2],
+			scalar - m[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator-(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		return mat<4, 4, T, Q>(
+			m1[0] - m2[0],
+			m1[1] - m2[1],
+			m1[2] - m2[2],
+			m1[3] - m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<4, 4, T, Q> const& m, T scalar)
+	{
+		return mat<4, 4, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar,
+			m[3] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator*(T scalar, mat<4, 4, T, Q> const& m)
+	{
+		return mat<4, 4, T, Q>(
+			m[0] * scalar,
+			m[1] * scalar,
+			m[2] * scalar,
+			m[3] * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 4, T, Q>::col_type operator*
+	(
+		mat<4, 4, T, Q> const& m,
+		typename mat<4, 4, T, Q>::row_type const& v
+	)
+	{
+/*
+		__m128 v0 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 v1 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 v2 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 v3 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(m[0].data, v0);
+		__m128 m1 = _mm_mul_ps(m[1].data, v1);
+		__m128 a0 = _mm_add_ps(m0, m1);
+
+		__m128 m2 = _mm_mul_ps(m[2].data, v2);
+		__m128 m3 = _mm_mul_ps(m[3].data, v3);
+		__m128 a1 = _mm_add_ps(m2, m3);
+
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		return typename mat<4, 4, T, Q>::col_type(a2);
+*/
+
+		typename mat<4, 4, T, Q>::col_type const Mov0(v[0]);
+		typename mat<4, 4, T, Q>::col_type const Mov1(v[1]);
+		typename mat<4, 4, T, Q>::col_type const Mul0 = m[0] * Mov0;
+		typename mat<4, 4, T, Q>::col_type const Mul1 = m[1] * Mov1;
+		typename mat<4, 4, T, Q>::col_type const Add0 = Mul0 + Mul1;
+		typename mat<4, 4, T, Q>::col_type const Mov2(v[2]);
+		typename mat<4, 4, T, Q>::col_type const Mov3(v[3]);
+		typename mat<4, 4, T, Q>::col_type const Mul2 = m[2] * Mov2;
+		typename mat<4, 4, T, Q>::col_type const Mul3 = m[3] * Mov3;
+		typename mat<4, 4, T, Q>::col_type const Add1 = Mul2 + Mul3;
+		typename mat<4, 4, T, Q>::col_type const Add2 = Add0 + Add1;
+		return Add2;
+
+/*
+		return typename mat<4, 4, T, Q>::col_type(
+			m[0][0] * v[0] + m[1][0] * v[1] + m[2][0] * v[2] + m[3][0] * v[3],
+			m[0][1] * v[0] + m[1][1] * v[1] + m[2][1] * v[2] + m[3][1] * v[3],
+			m[0][2] * v[0] + m[1][2] * v[1] + m[2][2] * v[2] + m[3][2] * v[3],
+			m[0][3] * v[0] + m[1][3] * v[1] + m[2][3] * v[2] + m[3][3] * v[3]);
+*/
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 4, T, Q>::row_type operator*
+	(
+		typename mat<4, 4, T, Q>::col_type const& v,
+		mat<4, 4, T, Q> const& m
+	)
+	{
+		return typename mat<4, 4, T, Q>::row_type(
+			m[0][0] * v[0] + m[0][1] * v[1] + m[0][2] * v[2] + m[0][3] * v[3],
+			m[1][0] * v[0] + m[1][1] * v[1] + m[1][2] * v[2] + m[1][3] * v[3],
+			m[2][0] * v[0] + m[2][1] * v[1] + m[2][2] * v[2] + m[2][3] * v[3],
+			m[3][0] * v[0] + m[3][1] * v[1] + m[3][2] * v[2] + m[3][3] * v[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<2, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<2, 4, T, Q> const& m2)
+	{
+		return mat<2, 4, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2] + m1[3][0] * m2[0][3],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2] + m1[3][1] * m2[0][3],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2] + m1[3][2] * m2[0][3],
+			m1[0][3] * m2[0][0] + m1[1][3] * m2[0][1] + m1[2][3] * m2[0][2] + m1[3][3] * m2[0][3],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2] + m1[3][0] * m2[1][3],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2] + m1[3][1] * m2[1][3],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2] + m1[3][2] * m2[1][3],
+			m1[0][3] * m2[1][0] + m1[1][3] * m2[1][1] + m1[2][3] * m2[1][2] + m1[3][3] * m2[1][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<3, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<3, 4, T, Q> const& m2)
+	{
+		return mat<3, 4, T, Q>(
+			m1[0][0] * m2[0][0] + m1[1][0] * m2[0][1] + m1[2][0] * m2[0][2] + m1[3][0] * m2[0][3],
+			m1[0][1] * m2[0][0] + m1[1][1] * m2[0][1] + m1[2][1] * m2[0][2] + m1[3][1] * m2[0][3],
+			m1[0][2] * m2[0][0] + m1[1][2] * m2[0][1] + m1[2][2] * m2[0][2] + m1[3][2] * m2[0][3],
+			m1[0][3] * m2[0][0] + m1[1][3] * m2[0][1] + m1[2][3] * m2[0][2] + m1[3][3] * m2[0][3],
+			m1[0][0] * m2[1][0] + m1[1][0] * m2[1][1] + m1[2][0] * m2[1][2] + m1[3][0] * m2[1][3],
+			m1[0][1] * m2[1][0] + m1[1][1] * m2[1][1] + m1[2][1] * m2[1][2] + m1[3][1] * m2[1][3],
+			m1[0][2] * m2[1][0] + m1[1][2] * m2[1][1] + m1[2][2] * m2[1][2] + m1[3][2] * m2[1][3],
+			m1[0][3] * m2[1][0] + m1[1][3] * m2[1][1] + m1[2][3] * m2[1][2] + m1[3][3] * m2[1][3],
+			m1[0][0] * m2[2][0] + m1[1][0] * m2[2][1] + m1[2][0] * m2[2][2] + m1[3][0] * m2[2][3],
+			m1[0][1] * m2[2][0] + m1[1][1] * m2[2][1] + m1[2][1] * m2[2][2] + m1[3][1] * m2[2][3],
+			m1[0][2] * m2[2][0] + m1[1][2] * m2[2][1] + m1[2][2] * m2[2][2] + m1[3][2] * m2[2][3],
+			m1[0][3] * m2[2][0] + m1[1][3] * m2[2][1] + m1[2][3] * m2[2][2] + m1[3][3] * m2[2][3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator*(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		typename mat<4, 4, T, Q>::col_type const SrcA0 = m1[0];
+		typename mat<4, 4, T, Q>::col_type const SrcA1 = m1[1];
+		typename mat<4, 4, T, Q>::col_type const SrcA2 = m1[2];
+		typename mat<4, 4, T, Q>::col_type const SrcA3 = m1[3];
+
+		typename mat<4, 4, T, Q>::col_type const SrcB0 = m2[0];
+		typename mat<4, 4, T, Q>::col_type const SrcB1 = m2[1];
+		typename mat<4, 4, T, Q>::col_type const SrcB2 = m2[2];
+		typename mat<4, 4, T, Q>::col_type const SrcB3 = m2[3];
+
+		mat<4, 4, T, Q> Result;
+		Result[0] = SrcA0 * SrcB0[0] + SrcA1 * SrcB0[1] + SrcA2 * SrcB0[2] + SrcA3 * SrcB0[3];
+		Result[1] = SrcA0 * SrcB1[0] + SrcA1 * SrcB1[1] + SrcA2 * SrcB1[2] + SrcA3 * SrcB1[3];
+		Result[2] = SrcA0 * SrcB2[0] + SrcA1 * SrcB2[1] + SrcA2 * SrcB2[2] + SrcA3 * SrcB2[3];
+		Result[3] = SrcA0 * SrcB3[0] + SrcA1 * SrcB3[1] + SrcA2 * SrcB3[2] + SrcA3 * SrcB3[3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator/(mat<4, 4, T, Q> const& m, T scalar)
+	{
+		return mat<4, 4, T, Q>(
+			m[0] / scalar,
+			m[1] / scalar,
+			m[2] / scalar,
+			m[3] / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator/(T scalar,	mat<4, 4, T, Q> const& m)
+	{
+		return mat<4, 4, T, Q>(
+			scalar / m[0],
+			scalar / m[1],
+			scalar / m[2],
+			scalar / m[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 4, T, Q>::col_type operator/(mat<4, 4, T, Q> const& m, typename mat<4, 4, T, Q>::row_type const& v)
+	{
+		return inverse(m) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR typename mat<4, 4, T, Q>::row_type operator/(typename mat<4, 4, T, Q>::col_type const& v, mat<4, 4, T, Q> const& m)
+	{
+		return v * inverse(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> operator/(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		mat<4, 4, T, Q> m1_copy(m1);
+		return m1_copy /= m2;
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		return (m1[0] == m2[0]) && (m1[1] == m2[1]) && (m1[2] == m2[2]) && (m1[3] == m2[3]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2)
+	{
+		return (m1[0] != m2[0]) || (m1[1] != m2[1]) || (m1[2] != m2[2]) || (m1[3] != m2[3]);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "type_mat4x4_simd.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4_simd.inl
new file mode 100644
index 000000000000..fb3a16f06290
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_mat4x4_simd.inl
@@ -0,0 +1,6 @@
+/// @ref core
+
+namespace glm
+{
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.hpp
new file mode 100644
index 000000000000..d489e0a233af
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.hpp
@@ -0,0 +1,193 @@
+/// @ref core
+/// @file glm/detail/type_quat.hpp
+
+#pragma once
+
+// Dependency:
+#include "../detail/type_mat3x3.hpp"
+#include "../detail/type_mat4x4.hpp"
+#include "../detail/type_vec3.hpp"
+#include "../detail/type_vec4.hpp"
+#include "../ext/vector_relational.hpp"
+#include "../ext/quaternion_relational.hpp"
+#include "../gtc/constants.hpp"
+#include "../gtc/matrix_transform.hpp"
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct qua
+	{
+		// -- Implementation detail --
+
+		typedef qua<T, Q> type;
+		typedef T value_type;
+
+		// -- Data --
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic push
+#				pragma GCC diagnostic ignored "-Wpedantic"
+#			elif GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic push
+#				pragma clang diagnostic ignored "-Wgnu-anonymous-struct"
+#				pragma clang diagnostic ignored "-Wnested-anon-types"
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(push)
+#				pragma warning(disable: 4201)  // nonstandard extension used : nameless struct/union
+#			endif
+#		endif
+
+#		if GLM_LANG & GLM_LANG_CXXMS_FLAG
+			union
+			{
+#				ifdef GLM_FORCE_QUAT_DATA_WXYZ
+					struct { T w, x, y, z; };
+#				else
+					struct { T x, y, z, w; };
+#				endif
+
+				typename detail::storage<4, T, detail::is_aligned<Q>::value>::type data;
+			};
+#		else
+#			ifdef GLM_FORCE_QUAT_DATA_WXYZ
+				T w, x, y, z;
+#			else
+				T x, y, z, w;
+#			endif
+#		endif
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(pop)
+#			endif
+#		endif
+
+		// -- Component accesses --
+
+		typedef length_t length_type;
+
+		/// Return the count of components of a quaternion
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 4;}
+
+		GLM_FUNC_DECL GLM_CONSTEXPR T & operator[](length_type i);
+		GLM_FUNC_DECL GLM_CONSTEXPR T const& operator[](length_type i) const;
+
+		// -- Implicit basic constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR qua() GLM_DEFAULT_CTOR;
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR qua(qua<T, Q> const& q) GLM_DEFAULT;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua(qua<T, P> const& q);
+
+		// -- Explicit basic constructors --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR qua(T s, vec<3, T, Q> const& v);
+
+#		ifdef GLM_FORCE_QUAT_CTOR_XYZW
+		GLM_FUNC_DECL GLM_CONSTEXPR qua(T x, T y, T z, T w);
+#		else
+		GLM_FUNC_DECL GLM_CONSTEXPR qua(T w, T x, T y, T z);
+#		endif
+
+		GLM_FUNC_DECL static GLM_CONSTEXPR qua<T, Q> wxyz(T w, T x, T y, T z);
+
+		// -- Conversion constructors --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT qua(qua<U, P> const& q);
+
+		/// Explicit conversion operators
+#		if GLM_HAS_EXPLICIT_CONVERSION_OPERATORS
+			GLM_FUNC_DECL explicit operator mat<3, 3, T, Q>() const;
+			GLM_FUNC_DECL explicit operator mat<4, 4, T, Q>() const;
+#		endif
+
+		/// Create a quaternion from two normalized axis
+		///
+		/// @param u A first normalized axis
+		/// @param v A second normalized axis
+		/// @see gtc_quaternion
+		/// @see http://lolengine.net/blog/2013/09/18/beautiful-maths-quaternion-from-vectors
+		GLM_FUNC_DECL qua(vec<3, T, Q> const& u, vec<3, T, Q> const& v);
+
+		/// Build a quaternion from euler angles (pitch, yaw, roll), in radians.
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT qua(vec<3, T, Q> const& eulerAngles);
+		GLM_FUNC_DECL GLM_EXPLICIT qua(mat<3, 3, T, Q> const& q);
+		GLM_FUNC_DECL GLM_EXPLICIT qua(mat<4, 4, T, Q> const& q);
+
+		// -- Unary arithmetic operators --
+
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator=(qua<T, Q> const& q) GLM_DEFAULT;
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator=(qua<U, Q> const& q);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator+=(qua<U, Q> const& q);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator-=(qua<U, Q> const& q);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator*=(qua<U, Q> const& q);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q>& operator/=(U s);
+	};
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator+(qua<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator-(qua<T, Q> const& q);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator+(qua<T, Q> const& q, qua<T, Q> const& p);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator-(qua<T, Q> const& q, qua<T, Q> const& p);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator*(qua<T, Q> const& q, qua<T, Q> const& p);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(qua<T, Q> const& q, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v, qua<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(qua<T, Q> const& q, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v, qua<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator*(qua<T, Q> const& q, T const& s);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator*(T const& s, qua<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> operator/(qua<T, Q> const& q, T const& s);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(qua<T, Q> const& q1, qua<T, Q> const& q2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(qua<T, Q> const& q1, qua<T, Q> const& q2);
+} //namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_quat.inl"
+#endif//GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.inl
new file mode 100644
index 000000000000..3213ea6aacc9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat.inl
@@ -0,0 +1,424 @@
+#include "../trigonometric.hpp"
+#include "../exponential.hpp"
+#include "../ext/quaternion_common.hpp"
+#include "../ext/quaternion_geometric.hpp"
+#include <limits>
+
+namespace glm{
+namespace detail
+{
+	template <typename T>
+	struct genTypeTrait<qua<T> >
+	{
+		static const genTypeEnum GENTYPE = GENTYPE_QUAT;
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_dot<qua<T, Q>, T, Aligned>
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(qua<T, Q> const& a, qua<T, Q> const& b)
+		{
+			vec<4, T, Q> tmp(a.w * b.w, a.x * b.x, a.y * b.y, a.z * b.z);
+			return (tmp.x + tmp.y) + (tmp.z + tmp.w);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_quat_add
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static qua<T, Q> call(qua<T, Q> const& q, qua<T, Q> const& p)
+		{
+			return qua<T, Q>::wxyz(q.w + p.w, q.x + p.x, q.y + p.y, q.z + p.z);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_quat_sub
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static qua<T, Q> call(qua<T, Q> const& q, qua<T, Q> const& p)
+		{
+			return qua<T, Q>::wxyz(q.w - p.w, q.x - p.x, q.y - p.y, q.z - p.z);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_quat_mul_scalar
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static qua<T, Q> call(qua<T, Q> const& q, T s)
+		{
+			return qua<T, Q>::wxyz(q.w * s, q.x * s, q.y * s, q.z * s);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_quat_div_scalar
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static qua<T, Q> call(qua<T, Q> const& q, T s)
+		{
+			return qua<T, Q>::wxyz(q.w / s, q.x / s, q.y / s, q.z / s);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_quat_mul_vec4
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(qua<T, Q> const& q, vec<4, T, Q> const& v)
+		{
+			return vec<4, T, Q>(q * vec<3, T, Q>(v), v.w);
+		}
+	};
+}//namespace detail
+
+	// -- Component accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T & qua<T, Q>::operator[](typename qua<T, Q>::length_type i)
+	{
+		assert(i >= 0 && i < this->length());
+#		ifdef GLM_FORCE_QUAT_DATA_WXYZ
+			return (&w)[i];
+#		else
+			return (&x)[i];
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T const& qua<T, Q>::operator[](typename qua<T, Q>::length_type i) const
+	{
+		assert(i >= 0 && i < this->length());
+#		ifdef GLM_FORCE_QUAT_DATA_WXYZ
+			return (&w)[i];
+#		else
+			return (&x)[i];
+#		endif
+	}
+
+	// -- Implicit basic constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua()
+#			if GLM_CONFIG_CTOR_INIT != GLM_CTOR_INIT_DISABLE
+#				ifdef GLM_FORCE_QUAT_DATA_WXYZ
+					: w(1), x(0), y(0), z(0)
+#				else
+					: x(0), y(0), z(0), w(1)
+#				endif
+#			endif
+		{}
+#	endif
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(qua<T, Q> const& q)
+#			ifdef GLM_FORCE_QUAT_DATA_WXYZ
+				: w(q.w), x(q.x), y(q.y), z(q.z)
+#			else
+				: x(q.x), y(q.y), z(q.z), w(q.w)
+#			endif
+		{}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(qua<T, P> const& q)
+#		ifdef GLM_FORCE_QUAT_DATA_WXYZ
+			: w(q.w), x(q.x), y(q.y), z(q.z)
+#		else
+			: x(q.x), y(q.y), z(q.z), w(q.w)
+#		endif
+	{}
+
+	// -- Explicit basic constructors --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(T s, vec<3, T, Q> const& v)
+#		ifdef GLM_FORCE_QUAT_DATA_WXYZ
+			: w(s), x(v.x), y(v.y), z(v.z)
+#		else
+			: x(v.x), y(v.y), z(v.z), w(s)
+#		endif
+	{}
+
+	template <typename T, qualifier Q>
+#		ifdef GLM_FORCE_QUAT_CTOR_XYZW
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(T _x, T _y, T _z, T _w)
+#		else
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(T _w, T _x, T _y, T _z)
+#		endif
+#		ifdef GLM_FORCE_QUAT_DATA_WXYZ
+			: w(_w), x(_x), y(_y), z(_z)
+#		else
+			: x(_x), y(_y), z(_z), w(_w)
+#		endif
+	{}
+
+	template <typename T, qualifier Q>
+	GLM_CONSTEXPR qua<T, Q> qua<T, Q>::wxyz(T w, T x, T y, T z) {
+#	ifdef GLM_FORCE_QUAT_DATA_XYZW
+		return qua<T, Q>(x, y, z, w);
+#	else
+		return qua<T, Q>(w, x, y, z);
+#	endif
+	}
+	
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(qua<U, P> const& q)
+#		ifdef GLM_FORCE_QUAT_DATA_WXYZ
+			: w(static_cast<T>(q.w)), x(static_cast<T>(q.x)), y(static_cast<T>(q.y)), z(static_cast<T>(q.z))
+#		else
+			: x(static_cast<T>(q.x)), y(static_cast<T>(q.y)), z(static_cast<T>(q.z)), w(static_cast<T>(q.w))
+#		endif
+	{}
+
+	//template<typename valType>
+	//GLM_FUNC_QUALIFIER qua<valType>::qua
+	//(
+	//	valType const& pitch,
+	//	valType const& yaw,
+	//	valType const& roll
+	//)
+	//{
+	//	vec<3, valType> eulerAngle(pitch * valType(0.5), yaw * valType(0.5), roll * valType(0.5));
+	//	vec<3, valType> c = glm::cos(eulerAngle * valType(0.5));
+	//	vec<3, valType> s = glm::sin(eulerAngle * valType(0.5));
+	//
+	//	this->w = c.x * c.y * c.z + s.x * s.y * s.z;
+	//	this->x = s.x * c.y * c.z - c.x * s.y * s.z;
+	//	this->y = c.x * s.y * c.z + s.x * c.y * s.z;
+	//	this->z = c.x * c.y * s.z - s.x * s.y * c.z;
+	//}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q>::qua(vec<3, T, Q> const& u, vec<3, T, Q> const& v)
+	{
+		T norm_u_norm_v = sqrt(dot(u, u) * dot(v, v));
+		T real_part = norm_u_norm_v + dot(u, v);
+		vec<3, T, Q> t;
+
+		if(real_part < static_cast<T>(1.e-6f) * norm_u_norm_v)
+		{
+			// If u and v are exactly opposite, rotate 180 degrees
+			// around an arbitrary orthogonal axis. Axis normalisation
+			// can happen later, when we normalise the quaternion.
+			real_part = static_cast<T>(0);
+			t = abs(u.x) > abs(u.z) ? vec<3, T, Q>(-u.y, u.x, static_cast<T>(0)) : vec<3, T, Q>(static_cast<T>(0), -u.z, u.y);
+		}
+		else
+		{
+			// Otherwise, build quaternion the standard way.
+			t = cross(u, v);
+		}
+
+		*this = normalize(qua<T, Q>::wxyz(real_part, t.x, t.y, t.z));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q>::qua(vec<3, T, Q> const& eulerAngle)
+	{
+		vec<3, T, Q> c = glm::cos(eulerAngle * T(0.5));
+		vec<3, T, Q> s = glm::sin(eulerAngle * T(0.5));
+
+		this->w = c.x * c.y * c.z + s.x * s.y * s.z;
+		this->x = s.x * c.y * c.z - c.x * s.y * s.z;
+		this->y = c.x * s.y * c.z + s.x * c.y * s.z;
+		this->z = c.x * c.y * s.z - s.x * s.y * c.z;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q>::qua(mat<3, 3, T, Q> const& m)
+	{
+		*this = quat_cast(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q>::qua(mat<4, 4, T, Q> const& m)
+	{
+		*this = quat_cast(m);
+	}
+
+#	if GLM_HAS_EXPLICIT_CONVERSION_OPERATORS
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q>::operator mat<3, 3, T, Q>() const
+	{
+		return mat3_cast(*this);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q>::operator mat<4, 4, T, Q>() const
+	{
+		return mat4_cast(*this);
+	}
+#	endif//GLM_HAS_EXPLICIT_CONVERSION_OPERATORS
+
+	// -- Unary arithmetic operators --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator=(qua<T, Q> const& q)
+		{
+			this->w = q.w;
+			this->x = q.x;
+			this->y = q.y;
+			this->z = q.z;
+			return *this;
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator=(qua<U, Q> const& q)
+	{
+		this->w = static_cast<T>(q.w);
+		this->x = static_cast<T>(q.x);
+		this->y = static_cast<T>(q.y);
+		this->z = static_cast<T>(q.z);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator+=(qua<U, Q> const& q)
+	{
+		return (*this = detail::compute_quat_add<T, Q, detail::is_aligned<Q>::value>::call(*this, qua<T, Q>(q)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator-=(qua<U, Q> const& q)
+	{
+		return (*this = detail::compute_quat_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, qua<T, Q>(q)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator*=(qua<U, Q> const& r)
+	{
+		qua<T, Q> const p(*this);
+		qua<T, Q> const q(r);
+
+		this->w = p.w * q.w - p.x * q.x - p.y * q.y - p.z * q.z;
+		this->x = p.w * q.x + p.x * q.w + p.y * q.z - p.z * q.y;
+		this->y = p.w * q.y + p.y * q.w + p.z * q.x - p.x * q.z;
+		this->z = p.w * q.z + p.z * q.w + p.x * q.y - p.y * q.x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator*=(U s)
+	{
+		return (*this = detail::compute_quat_mul_scalar<T, Q, detail::is_aligned<Q>::value>::call(*this, static_cast<U>(s)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> & qua<T, Q>::operator/=(U s)
+	{
+		return (*this = detail::compute_quat_div_scalar<T, Q, detail::is_aligned<Q>::value>::call(*this, static_cast<U>(s)));
+	}
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator+(qua<T, Q> const& q)
+	{
+		return q;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator-(qua<T, Q> const& q)
+	{
+		return qua<T, Q>::wxyz(-q.w, -q.x, -q.y, -q.z);
+	}
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator+(qua<T, Q> const& q, qua<T, Q> const& p)
+	{
+		return qua<T, Q>(q) += p;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator-(qua<T, Q> const& q, qua<T, Q> const& p)
+	{
+		return qua<T, Q>(q) -= p;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator*(qua<T, Q> const& q, qua<T, Q> const& p)
+	{
+		return qua<T, Q>(q) *= p;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(qua<T, Q> const& q, vec<3, T, Q> const& v)
+	{
+		vec<3, T, Q> const QuatVector(q.x, q.y, q.z);
+		vec<3, T, Q> const uv(glm::cross(QuatVector, v));
+		vec<3, T, Q> const uuv(glm::cross(QuatVector, uv));
+
+		return v + ((uv * q.w) + uuv) * static_cast<T>(2);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v, qua<T, Q> const& q)
+	{
+		return glm::inverse(q) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(qua<T, Q> const& q, vec<4, T, Q> const& v)
+	{
+		return detail::compute_quat_mul_vec4<T, Q, detail::is_aligned<Q>::value>::call(q, v);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v, qua<T, Q> const& q)
+	{
+		return glm::inverse(q) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator*(qua<T, Q> const& q, T const& s)
+	{
+		return qua<T, Q>::wxyz(
+			q.w * s, q.x * s, q.y * s, q.z * s);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator*(T const& s, qua<T, Q> const& q)
+	{
+		return q * s;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> operator/(qua<T, Q> const& q, T const& s)
+	{
+		return qua<T, Q>::wxyz(
+			q.w / s, q.x / s, q.y / s, q.z / s);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(qua<T, Q> const& q1, qua<T, Q> const& q2)
+	{
+		return q1.x == q2.x && q1.y == q2.y && q1.z == q2.z && q1.w == q2.w;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(qua<T, Q> const& q1, qua<T, Q> const& q2)
+	{
+		return q1.x != q2.x || q1.y != q2.y || q1.z != q2.z || q1.w != q2.w;
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "type_quat_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat_simd.inl
new file mode 100644
index 000000000000..fa6da198326c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_quat_simd.inl
@@ -0,0 +1,208 @@
+/// @ref core
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+namespace glm{
+namespace detail
+{
+/*
+	template<qualifier Q>
+	struct compute_quat_mul<float, Q, true>
+	{
+		static qua<float, Q> call(qua<float, Q> const& q1, qua<float, Q> const& q2)
+		{
+			// SSE2 STATS: 11 shuffle, 8 mul, 8 add
+			// SSE4 STATS: 3 shuffle, 4 mul, 4 dpps
+
+			__m128 const mul0 = _mm_mul_ps(q1.data, _mm_shuffle_ps(q2.data, q2.data, _MM_SHUFFLE(0, 1, 2, 3)));
+			__m128 const mul1 = _mm_mul_ps(q1.data, _mm_shuffle_ps(q2.data, q2.data, _MM_SHUFFLE(1, 0, 3, 2)));
+			__m128 const mul2 = _mm_mul_ps(q1.data, _mm_shuffle_ps(q2.data, q2.data, _MM_SHUFFLE(2, 3, 0, 1)));
+			__m128 const mul3 = _mm_mul_ps(q1.data, q2.data);
+
+#			if GLM_ARCH & GLM_ARCH_SSE41_BIT
+				__m128 const add0 = _mm_dp_ps(mul0, _mm_set_ps(1.0f, -1.0f,  1.0f,  1.0f), 0xff);
+				__m128 const add1 = _mm_dp_ps(mul1, _mm_set_ps(1.0f,  1.0f,  1.0f, -1.0f), 0xff);
+				__m128 const add2 = _mm_dp_ps(mul2, _mm_set_ps(1.0f,  1.0f, -1.0f,  1.0f), 0xff);
+				__m128 const add3 = _mm_dp_ps(mul3, _mm_set_ps(1.0f, -1.0f, -1.0f, -1.0f), 0xff);
+#			else
+				__m128 const mul4 = _mm_mul_ps(mul0, _mm_set_ps(1.0f, -1.0f,  1.0f,  1.0f));
+				__m128 const add0 = _mm_add_ps(mul0, _mm_movehl_ps(mul4, mul4));
+				__m128 const add4 = _mm_add_ss(add0, _mm_shuffle_ps(add0, add0, 1));
+
+				__m128 const mul5 = _mm_mul_ps(mul1, _mm_set_ps(1.0f,  1.0f,  1.0f, -1.0f));
+				__m128 const add1 = _mm_add_ps(mul1, _mm_movehl_ps(mul5, mul5));
+				__m128 const add5 = _mm_add_ss(add1, _mm_shuffle_ps(add1, add1, 1));
+
+				__m128 const mul6 = _mm_mul_ps(mul2, _mm_set_ps(1.0f,  1.0f, -1.0f,  1.0f));
+				__m128 const add2 = _mm_add_ps(mul6, _mm_movehl_ps(mul6, mul6));
+				__m128 const add6 = _mm_add_ss(add2, _mm_shuffle_ps(add2, add2, 1));
+
+				__m128 const mul7 = _mm_mul_ps(mul3, _mm_set_ps(1.0f, -1.0f, -1.0f, -1.0f));
+				__m128 const add3 = _mm_add_ps(mul3, _mm_movehl_ps(mul7, mul7));
+				__m128 const add7 = _mm_add_ss(add3, _mm_shuffle_ps(add3, add3, 1));
+		#endif
+
+			// This SIMD code is a politically correct way of doing this, but in every test I've tried it has been slower than
+			// the final code below. I'll keep this here for reference - maybe somebody else can do something better...
+			//
+			//__m128 xxyy = _mm_shuffle_ps(add4, add5, _MM_SHUFFLE(0, 0, 0, 0));
+			//__m128 zzww = _mm_shuffle_ps(add6, add7, _MM_SHUFFLE(0, 0, 0, 0));
+			//
+			//return _mm_shuffle_ps(xxyy, zzww, _MM_SHUFFLE(2, 0, 2, 0));
+
+			qua<float, Q> Result;
+			_mm_store_ss(&Result.x, add4);
+			_mm_store_ss(&Result.y, add5);
+			_mm_store_ss(&Result.z, add6);
+			_mm_store_ss(&Result.w, add7);
+			return Result;
+		}
+	};
+*/
+
+	template<qualifier Q>
+	struct compute_quat_add<float, Q, true>
+	{
+		static qua<float, Q> call(qua<float, Q> const& q, qua<float, Q> const& p)
+		{
+			qua<float, Q> Result;
+			Result.data = _mm_add_ps(q.data, p.data);
+			return Result;
+		}
+	};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+	template<qualifier Q>
+	struct compute_quat_add<double, Q, true>
+	{
+		static qua<double, Q> call(qua<double, Q> const& a, qua<double, Q> const& b)
+		{
+			qua<double, Q> Result;
+			Result.data = _mm256_add_pd(a.data, b.data);
+			return Result;
+		}
+	};
+#	endif
+
+	template<qualifier Q>
+	struct compute_quat_sub<float, Q, true>
+	{
+		static qua<float, Q> call(qua<float, Q> const& q, qua<float, Q> const& p)
+		{
+			qua<float, Q> Result;
+			Result.data = _mm_sub_ps(q.data, p.data);
+			return Result;
+		}
+	};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+	template<qualifier Q>
+	struct compute_quat_sub<double, Q, true>
+	{
+		static qua<double, Q> call(qua<double, Q> const& a, qua<double, Q> const& b)
+		{
+			qua<double, Q> Result;
+			Result.data = _mm256_sub_pd(a.data, b.data);
+			return Result;
+		}
+	};
+#	endif
+
+	template<qualifier Q>
+	struct compute_quat_mul_scalar<float, Q, true>
+	{
+		static qua<float, Q> call(qua<float, Q> const& q, float s)
+		{
+			vec<4, float, Q> Result;
+			Result.data = _mm_mul_ps(q.data, _mm_set_ps1(s));
+			return Result;
+		}
+	};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+	template<qualifier Q>
+	struct compute_quat_mul_scalar<double, Q, true>
+	{
+		static qua<double, Q> call(qua<double, Q> const& q, double s)
+		{
+			qua<double, Q> Result;
+			Result.data = _mm256_mul_pd(q.data, _mm_set_ps1(s));
+			return Result;
+		}
+	};
+#	endif
+
+	template<qualifier Q>
+	struct compute_quat_div_scalar<float, Q, true>
+	{
+		static qua<float, Q> call(qua<float, Q> const& q, float s)
+		{
+			vec<4, float, Q> Result;
+			Result.data = _mm_div_ps(q.data, _mm_set_ps1(s));
+			return Result;
+		}
+	};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+	template<qualifier Q>
+	struct compute_quat_div_scalar<double, Q, true>
+	{
+		static qua<double, Q> call(qua<double, Q> const& q, double s)
+		{
+			qua<double, Q> Result;
+			Result.data = _mm256_div_pd(q.data, _mm_set_ps1(s));
+			return Result;
+		}
+	};
+#	endif
+
+	template<qualifier Q>
+	struct compute_quat_mul_vec4<float, Q, true>
+	{
+		static vec<4, float, Q> call(qua<float, Q> const& q, vec<4, float, Q> const& v)
+		{
+#			ifdef GLM_FORCE_QUAT_DATA_WXYZ
+				__m128 const q_wwww = _mm_shuffle_ps(q.data, q.data, _MM_SHUFFLE(0, 0, 0, 0));
+				__m128 const q_swp0 = _mm_shuffle_ps(q.data, q.data, _MM_SHUFFLE(0, 1, 3, 2));
+				__m128 const q_swp1 = _mm_shuffle_ps(q.data, q.data, _MM_SHUFFLE(0, 2, 1, 3));
+				__m128 const v_swp0 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(3, 0, 2, 1));
+				__m128 const v_swp1 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(3, 1, 0, 2));
+
+				__m128 uv = _mm_sub_ps(_mm_mul_ps(q_swp0, v_swp1), _mm_mul_ps(q_swp1, v_swp0));
+				__m128 uv_swp0 = _mm_shuffle_ps(uv, uv, _MM_SHUFFLE(3, 0, 2, 1));
+				__m128 uv_swp1 = _mm_shuffle_ps(uv, uv, _MM_SHUFFLE(3, 1, 0, 2));
+				__m128 uuv = _mm_sub_ps(_mm_mul_ps(q_swp0, uv_swp1), _mm_mul_ps(q_swp1, uv_swp0));
+
+				__m128 const two = _mm_set1_ps(2.0f);
+				uv = _mm_mul_ps(uv, _mm_mul_ps(q_wwww, two));
+				uuv = _mm_mul_ps(uuv, two);
+
+				vec<4, float, Q> Result;
+				Result.data = _mm_add_ps(v.data, _mm_add_ps(uv, uuv));
+				return Result;
+#			else
+				__m128 const q_wwww = _mm_shuffle_ps(q.data, q.data, _MM_SHUFFLE(3, 3, 3, 3));
+				__m128 const q_swp0 = _mm_shuffle_ps(q.data, q.data, _MM_SHUFFLE(3, 0, 2, 1));
+				__m128 const q_swp1 = _mm_shuffle_ps(q.data, q.data, _MM_SHUFFLE(3, 1, 0, 2));
+				__m128 const v_swp0 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(3, 0, 2, 1));
+				__m128 const v_swp1 = _mm_shuffle_ps(v.data, v.data, _MM_SHUFFLE(3, 1, 0, 2));
+
+				__m128 uv      = _mm_sub_ps(_mm_mul_ps(q_swp0, v_swp1), _mm_mul_ps(q_swp1, v_swp0));
+				__m128 uv_swp0 = _mm_shuffle_ps(uv, uv, _MM_SHUFFLE(3, 0, 2, 1));
+				__m128 uv_swp1 = _mm_shuffle_ps(uv, uv, _MM_SHUFFLE(3, 1, 0, 2));
+				__m128 uuv     = _mm_sub_ps(_mm_mul_ps(q_swp0, uv_swp1), _mm_mul_ps(q_swp1, uv_swp0));
+
+				__m128 const two = _mm_set1_ps(2.0f);
+				uv  = _mm_mul_ps(uv, _mm_mul_ps(q_wwww, two));
+				uuv = _mm_mul_ps(uuv, two);
+
+				vec<4, float, Q> Result;
+				Result.data = _mm_add_ps(v.data, _mm_add_ps(uv, uuv));
+				return Result;
+#			endif
+		}
+	};
+}//namespace detail
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.hpp
new file mode 100644
index 000000000000..91232f9cbabe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.hpp
@@ -0,0 +1,308 @@
+/// @ref core
+/// @file glm/detail/type_vec1.hpp
+
+#pragma once
+
+#include "qualifier.hpp"
+#if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+#	include "_swizzle.hpp"
+#elif GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+#	include "_swizzle_func.hpp"
+#endif
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct vec<1, T, Q>
+	{
+		// -- Implementation detail --
+
+		typedef T value_type;
+		typedef vec<1, T, Q> type;
+		typedef vec<1, bool, Q> bool_type;
+
+		// -- Data --
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic push
+#				pragma GCC diagnostic ignored "-Wpedantic"
+#			elif GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic push
+#				pragma clang diagnostic ignored "-Wgnu-anonymous-struct"
+#				pragma clang diagnostic ignored "-Wnested-anon-types"
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(push)
+#				pragma warning(disable: 4201)  // nonstandard extension used : nameless struct/union
+#			endif
+#		endif
+
+#		if GLM_CONFIG_XYZW_ONLY
+			T x;
+#		elif GLM_CONFIG_ANONYMOUS_STRUCT == GLM_ENABLE
+			union
+			{
+				T x;
+				T r;
+				T s;
+
+				typename detail::storage<1, T, detail::is_aligned<Q>::value>::type data;
+/*
+#				if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+					_GLM_SWIZZLE1_2_MEMBERS(T, Q, x)
+					_GLM_SWIZZLE1_2_MEMBERS(T, Q, r)
+					_GLM_SWIZZLE1_2_MEMBERS(T, Q, s)
+					_GLM_SWIZZLE1_3_MEMBERS(T, Q, x)
+					_GLM_SWIZZLE1_3_MEMBERS(T, Q, r)
+					_GLM_SWIZZLE1_3_MEMBERS(T, Q, s)
+					_GLM_SWIZZLE1_4_MEMBERS(T, Q, x)
+					_GLM_SWIZZLE1_4_MEMBERS(T, Q, r)
+					_GLM_SWIZZLE1_4_MEMBERS(T, Q, s)
+#				endif
+*/
+			};
+#		else
+			union {T x, r, s;};
+/*
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+				GLM_SWIZZLE_GEN_VEC_FROM_VEC1(T, Q)
+#			endif
+*/
+#		endif
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(pop)
+#			endif
+#		endif
+
+		// -- Component accesses --
+
+		/// Return the count of components of the vector
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 1;}
+
+		GLM_FUNC_DECL GLM_CONSTEXPR T & operator[](length_type i);
+		GLM_FUNC_DECL GLM_CONSTEXPR T const& operator[](length_type i) const;
+
+		// -- Implicit basic constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR vec() GLM_DEFAULT_CTOR;
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec(vec const& v) GLM_DEFAULT;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, T, P> const& v);
+
+		// -- Explicit basic constructors --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(T scalar);
+
+		// -- Conversion vector constructors --
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<2, U, P> const& v);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<3, U, P> const& v);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<4, U, P> const& v);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<1, U, P> const& v);
+
+		// -- Swizzle constructors --
+/*
+#		if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+			template<int E0>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<1, T, Q, E0, -1,-2,-3> const& that)
+			{
+				*this = that();
+			}
+#		endif//GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+*/
+		// -- Unary arithmetic operators --
+
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator=(vec const& v) GLM_DEFAULT;
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator+=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator+=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator-=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator-=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator*=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator*=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator/=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator/=(vec<1, U, Q> const& v);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator--(int);
+
+		// -- Unary bit operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator%=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator%=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator&=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator&=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator|=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator|=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator^=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator^=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator<<=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator<<=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator>>=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> & operator>>=(vec<1, U, Q> const& v);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator+(vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator-(vec<1, T, Q> const& v);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator+(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator+(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator+(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator-(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator-(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator-(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator*(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator*(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator*(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator/(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator/(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator/(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator%(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator%(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator%(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator&(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator&(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator&(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator|(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator|(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator|(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator^(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator^(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator^(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator<<(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator<<(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator<<(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator>>(vec<1, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator>>(T scalar, vec<1, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator>>(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, T, Q> operator~(vec<1, T, Q> const& v);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, bool, Q> operator&&(vec<1, bool, Q> const& v1, vec<1, bool, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<1, bool, Q> operator||(vec<1, bool, Q> const& v1, vec<1, bool, Q> const& v2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_vec1.inl"
+#endif//GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.inl
new file mode 100644
index 000000000000..18411e7f76c5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec1.inl
@@ -0,0 +1,553 @@
+/// @ref core
+
+#include "./compute_vector_relational.hpp"
+
+namespace glm
+{
+	// -- Implicit basic constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec()
+#			if GLM_CONFIG_CTOR_INIT != GLM_CTOR_INIT_DISABLE
+				: x(0)
+#			endif
+		{}
+#	endif
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(vec<1, T, Q> const& v)
+			: x(v.x)
+		{}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(vec<1, T, P> const& v)
+		: x(v.x)
+	{}
+
+	// -- Explicit basic constructors --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(T scalar)
+		: x(scalar)
+	{}
+
+	// -- Conversion vector constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(vec<1, U, P> const& v)
+		: x(static_cast<T>(v.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(vec<2, U, P> const& v)
+		: x(static_cast<T>(v.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(vec<3, U, P> const& v)
+		: x(static_cast<T>(v.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q>::vec(vec<4, U, P> const& v)
+		: x(static_cast<T>(v.x))
+	{}
+
+	// -- Component accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T & vec<1, T, Q>::operator[](typename vec<1, T, Q>::length_type)
+	{
+		return x;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T const& vec<1, T, Q>::operator[](typename vec<1, T, Q>::length_type) const
+	{
+		return x;
+	}
+
+	// -- Unary arithmetic operators --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator=(vec<1, T, Q> const& v)
+		{
+			this->x = v.x;
+			return *this;
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator=(vec<1, U, Q> const& v)
+	{
+		this->x = static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator+=(U scalar)
+	{
+		this->x += static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator+=(vec<1, U, Q> const& v)
+	{
+		this->x += static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator-=(U scalar)
+	{
+		this->x -= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator-=(vec<1, U, Q> const& v)
+	{
+		this->x -= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator*=(U scalar)
+	{
+		this->x *= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator*=(vec<1, U, Q> const& v)
+	{
+		this->x *= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator/=(U scalar)
+	{
+		this->x /= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator/=(vec<1, U, Q> const& v)
+	{
+		this->x /= static_cast<T>(v.x);
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator++()
+	{
+		++this->x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator--()
+	{
+		--this->x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> vec<1, T, Q>::operator++(int)
+	{
+		vec<1, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> vec<1, T, Q>::operator--(int)
+	{
+		vec<1, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator%=(U scalar)
+	{
+		this->x %= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator%=(vec<1, U, Q> const& v)
+	{
+		this->x %= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator&=(U scalar)
+	{
+		this->x &= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator&=(vec<1, U, Q> const& v)
+	{
+		this->x &= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator|=(U scalar)
+	{
+		this->x |= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator|=(vec<1, U, Q> const& v)
+	{
+		this->x |= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator^=(U scalar)
+	{
+		this->x ^= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator^=(vec<1, U, Q> const& v)
+	{
+		this->x ^= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator<<=(U scalar)
+	{
+		this->x <<= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator<<=(vec<1, U, Q> const& v)
+	{
+		this->x <<= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator>>=(U scalar)
+	{
+		this->x >>= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> & vec<1, T, Q>::operator>>=(vec<1, U, Q> const& v)
+	{
+		this->x >>= static_cast<T>(v.x);
+		return *this;
+	}
+
+	// -- Unary constant operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator+(vec<1, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator-(vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			-v.x);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator+(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator+(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar + v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator+(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x + v2.x);
+	}
+
+	//operator-
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator-(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator-(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar - v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator-(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x - v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator*(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator*(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar * v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator*(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x * v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator/(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator/(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar / v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator/(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x / v2.x);
+	}
+
+	// -- Binary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator%(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x % scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator%(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar % v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator%(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x % v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator&(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x & scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator&(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar & v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator&(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x & v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator|(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x | scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator|(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar | v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator|(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x | v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator^(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			v.x ^ scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator^(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			scalar ^ v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator^(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			v1.x ^ v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator<<(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			static_cast<T>(v.x << scalar));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator<<(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			static_cast<T>(scalar << v.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator<<(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			static_cast<T>(v1.x << v2.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator>>(vec<1, T, Q> const& v, T scalar)
+	{
+		return vec<1, T, Q>(
+			static_cast<T>(v.x >> scalar));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator>>(T scalar, vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			static_cast<T>(scalar >> v.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator>>(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<1, T, Q>(
+			static_cast<T>(v1.x >> v2.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, T, Q> operator~(vec<1, T, Q> const& v)
+	{
+		return vec<1, T, Q>(
+			~v.x);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.x, v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(vec<1, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return !(v1 == v2);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, bool, Q> operator&&(vec<1, bool, Q> const& v1, vec<1, bool, Q> const& v2)
+	{
+		return vec<1, bool, Q>(v1.x && v2.x);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<1, bool, Q> operator||(vec<1, bool, Q> const& v1, vec<1, bool, Q> const& v2)
+	{
+		return vec<1, bool, Q>(v1.x || v2.x);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.hpp
new file mode 100644
index 000000000000..424868f555c5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.hpp
@@ -0,0 +1,402 @@
+/// @ref core
+/// @file glm/detail/type_vec2.hpp
+
+#pragma once
+
+#include "qualifier.hpp"
+#if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+#	include "_swizzle.hpp"
+#elif GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+#	include "_swizzle_func.hpp"
+#endif
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct vec<2, T, Q>
+	{
+		// -- Implementation detail --
+
+		typedef T value_type;
+		typedef vec<2, T, Q> type;
+		typedef vec<2, bool, Q> bool_type;
+
+		// -- Data --
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic push
+#				pragma GCC diagnostic ignored "-Wpedantic"
+#			elif GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic push
+#				pragma clang diagnostic ignored "-Wgnu-anonymous-struct"
+#				pragma clang diagnostic ignored "-Wnested-anon-types"
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(push)
+#				pragma warning(disable: 4201)  // nonstandard extension used : nameless struct/union
+#			endif
+#		endif
+
+#		if GLM_CONFIG_XYZW_ONLY
+			T x, y;
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+			GLM_SWIZZLE_GEN_VEC_FROM_VEC2_COMP(T, Q, x, y)
+#			endif//GLM_CONFIG_SWIZZLE
+#		elif GLM_CONFIG_ANONYMOUS_STRUCT == GLM_ENABLE
+			union
+			{
+				struct{ T x, y; };
+				struct{ T r, g; };
+				struct{ T s, t; };
+
+				typename detail::storage<2, T, detail::is_aligned<Q>::value>::type data;
+
+#				if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+					GLM_SWIZZLE2_2_MEMBERS(T, Q, x, y)
+					GLM_SWIZZLE2_2_MEMBERS(T, Q, r, g)
+					GLM_SWIZZLE2_2_MEMBERS(T, Q, s, t)
+					GLM_SWIZZLE2_3_MEMBERS(T, Q, x, y)
+					GLM_SWIZZLE2_3_MEMBERS(T, Q, r, g)
+					GLM_SWIZZLE2_3_MEMBERS(T, Q, s, t)
+					GLM_SWIZZLE2_4_MEMBERS(T, Q, x, y)
+					GLM_SWIZZLE2_4_MEMBERS(T, Q, r, g)
+					GLM_SWIZZLE2_4_MEMBERS(T, Q, s, t)
+#				endif
+			};
+#		else
+			union {T x, r, s;};
+			union {T y, g, t;};
+
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+				GLM_SWIZZLE_GEN_VEC_FROM_VEC2(T, Q)
+#			endif//GLM_CONFIG_SWIZZLE
+#		endif
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(pop)
+#			endif
+#		endif
+
+		// -- Component accesses --
+
+		/// Return the count of components of the vector
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 2;}
+
+		GLM_FUNC_DECL GLM_CONSTEXPR T& operator[](length_type i);
+		GLM_FUNC_DECL GLM_CONSTEXPR T const& operator[](length_type i) const;
+
+		// -- Implicit basic constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR vec() GLM_DEFAULT_CTOR;
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec(vec const& v) GLM_DEFAULT;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, T, P> const& v);
+
+		// -- Explicit basic constructors --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(T x, T y);
+
+		// -- Conversion constructors --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(vec<1, U, P> const& v);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A x, B y);
+		template<typename A, typename B>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, Q> const& x, B y);
+		template<typename A, typename B>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A x, vec<1, B, Q> const& y);
+		template<typename A, typename B>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, Q> const& x, vec<1, B, Q> const& y);
+
+		// -- Conversion vector constructors --
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<3, U, P> const& v);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<4, U, P> const& v);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<2, U, P> const& v);
+
+		// -- Swizzle constructors --
+#		if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+			template<int E0, int E1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<2, T, Q, E0, E1,-1,-2> const& that)
+			{
+				*this = that();
+			}
+#		endif//GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+
+		// -- Unary arithmetic operators --
+
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator=(vec const& v) GLM_DEFAULT;
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator+=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator+=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator+=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator-=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator-=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator-=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator*=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator*=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator*=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator/=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator/=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator/=(vec<2, U, Q> const& v);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator--(int);
+
+		// -- Unary bit operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator%=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator%=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator%=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator&=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator&=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator&=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator|=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator|=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator|=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator^=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator^=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator^=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator<<=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator<<=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator<<=(vec<2, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator>>=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator>>=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> & operator>>=(vec<2, U, Q> const& v);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator+(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator+(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator-(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator-(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator*(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator*(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator*(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator*(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator*(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator/(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator/(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator/(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator/(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator/(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator%(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator%(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator%(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator%(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator%(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator&(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator&(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator&(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator&(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator&(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator|(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator|(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator|(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator|(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator|(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator^(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator^(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator^(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator^(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator^(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator<<(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<2, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator>>(T scalar, vec<2, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, T, Q> operator~(vec<2, T, Q> const& v);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, bool, Q> operator&&(vec<2, bool, Q> const& v1, vec<2, bool, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<2, bool, Q> operator||(vec<2, bool, Q> const& v1, vec<2, bool, Q> const& v2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_vec2.inl"
+#endif//GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.inl
new file mode 100644
index 000000000000..24850965b510
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec2.inl
@@ -0,0 +1,915 @@
+/// @ref core
+
+#include "./compute_vector_relational.hpp"
+
+namespace glm
+{
+	// -- Implicit basic constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec()
+#			if GLM_CONFIG_CTOR_INIT != GLM_CTOR_INIT_DISABLE
+				: x(0), y(0)
+#			endif
+		{}
+#	endif
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<2, T, Q> const& v)
+			: x(v.x), y(v.y)
+		{}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<2, T, P> const& v)
+		: x(v.x), y(v.y)
+	{}
+
+	// -- Explicit basic constructors --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(T scalar)
+		: x(scalar), y(scalar)
+	{}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(T _x, T _y)
+		: x(_x), y(_y)
+	{}
+
+	// -- Conversion scalar constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<1, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(A _x, B _y)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<1, A, Q> const& _x, B _y)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(A _x, vec<1, B, Q> const& _y)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<1, A, Q> const& _x, vec<1, B, Q> const& _y)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+	{}
+
+	// -- Conversion vector constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<2, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<3, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q>::vec(vec<4, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.y))
+	{}
+
+	// -- Component accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T & vec<2, T, Q>::operator[](typename vec<2, T, Q>::length_type i)
+	{
+		assert(i >= 0 && i < this->length());
+		switch(i)
+		{
+		default:
+		case 0:
+			return x;
+		case 1:
+			return y;
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T const& vec<2, T, Q>::operator[](typename vec<2, T, Q>::length_type i) const
+	{
+		assert(i >= 0 && i < this->length());
+		switch(i)
+		{
+		default:
+		case 0:
+			return x;
+		case 1:
+			return y;
+		}
+	}
+
+	// -- Unary arithmetic operators --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator=(vec<2, T, Q> const& v)
+		{
+			this->x = v.x;
+			this->y = v.y;
+			return *this;
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator=(vec<2, U, Q> const& v)
+	{
+		this->x = static_cast<T>(v.x);
+		this->y = static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator+=(U scalar)
+	{
+		this->x += static_cast<T>(scalar);
+		this->y += static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator+=(vec<1, U, Q> const& v)
+	{
+		this->x += static_cast<T>(v.x);
+		this->y += static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator+=(vec<2, U, Q> const& v)
+	{
+		this->x += static_cast<T>(v.x);
+		this->y += static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator-=(U scalar)
+	{
+		this->x -= static_cast<T>(scalar);
+		this->y -= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator-=(vec<1, U, Q> const& v)
+	{
+		this->x -= static_cast<T>(v.x);
+		this->y -= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator-=(vec<2, U, Q> const& v)
+	{
+		this->x -= static_cast<T>(v.x);
+		this->y -= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator*=(U scalar)
+	{
+		this->x *= static_cast<T>(scalar);
+		this->y *= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator*=(vec<1, U, Q> const& v)
+	{
+		this->x *= static_cast<T>(v.x);
+		this->y *= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator*=(vec<2, U, Q> const& v)
+	{
+		this->x *= static_cast<T>(v.x);
+		this->y *= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator/=(U scalar)
+	{
+		this->x /= static_cast<T>(scalar);
+		this->y /= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator/=(vec<1, U, Q> const& v)
+	{
+		this->x /= static_cast<T>(v.x);
+		this->y /= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator/=(vec<2, U, Q> const& v)
+	{
+		this->x /= static_cast<T>(v.x);
+		this->y /= static_cast<T>(v.y);
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator++()
+	{
+		++this->x;
+		++this->y;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator--()
+	{
+		--this->x;
+		--this->y;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> vec<2, T, Q>::operator++(int)
+	{
+		vec<2, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> vec<2, T, Q>::operator--(int)
+	{
+		vec<2, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator%=(U scalar)
+	{
+		this->x %= static_cast<T>(scalar);
+		this->y %= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator%=(vec<1, U, Q> const& v)
+	{
+		this->x %= static_cast<T>(v.x);
+		this->y %= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator%=(vec<2, U, Q> const& v)
+	{
+		this->x %= static_cast<T>(v.x);
+		this->y %= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator&=(U scalar)
+	{
+		this->x &= static_cast<T>(scalar);
+		this->y &= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator&=(vec<1, U, Q> const& v)
+	{
+		this->x &= static_cast<T>(v.x);
+		this->y &= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator&=(vec<2, U, Q> const& v)
+	{
+		this->x &= static_cast<T>(v.x);
+		this->y &= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator|=(U scalar)
+	{
+		this->x |= static_cast<T>(scalar);
+		this->y |= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator|=(vec<1, U, Q> const& v)
+	{
+		this->x |= static_cast<T>(v.x);
+		this->y |= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator|=(vec<2, U, Q> const& v)
+	{
+		this->x |= static_cast<T>(v.x);
+		this->y |= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator^=(U scalar)
+	{
+		this->x ^= static_cast<T>(scalar);
+		this->y ^= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator^=(vec<1, U, Q> const& v)
+	{
+		this->x ^= static_cast<T>(v.x);
+		this->y ^= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator^=(vec<2, U, Q> const& v)
+	{
+		this->x ^= static_cast<T>(v.x);
+		this->y ^= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator<<=(U scalar)
+	{
+		this->x <<= static_cast<T>(scalar);
+		this->y <<= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator<<=(vec<1, U, Q> const& v)
+	{
+		this->x <<= static_cast<T>(v.x);
+		this->y <<= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator<<=(vec<2, U, Q> const& v)
+	{
+		this->x <<= static_cast<T>(v.x);
+		this->y <<= static_cast<T>(v.y);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator>>=(U scalar)
+	{
+		this->x >>= static_cast<T>(scalar);
+		this->y >>= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator>>=(vec<1, U, Q> const& v)
+	{
+		this->x >>= static_cast<T>(v.x);
+		this->y >>= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> & vec<2, T, Q>::operator>>=(vec<2, U, Q> const& v)
+	{
+		this->x >>= static_cast<T>(v.x);
+		this->y >>= static_cast<T>(v.y);
+		return *this;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			-v.x,
+			-v.y);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x + scalar,
+			v.y + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x + v2.x,
+			v1.y + v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator+(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar + v.x,
+			scalar + v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator+(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x + v2.x,
+			v1.x + v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator+(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x + v2.x,
+			v1.y + v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x - scalar,
+			v.y - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x - v2.x,
+			v1.y - v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator-(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar - v.x,
+			scalar - v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator-(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x - v2.x,
+			v1.x - v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator-(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x - v2.x,
+			v1.y - v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator*(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x * scalar,
+			v.y * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator*(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x * v2.x,
+			v1.y * v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator*(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar * v.x,
+			scalar * v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator*(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x * v2.x,
+			v1.x * v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator*(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x * v2.x,
+			v1.y * v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator/(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x / scalar,
+			v.y / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator/(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x / v2.x,
+			v1.y / v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator/(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar / v.x,
+			scalar / v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator/(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x / v2.x,
+			v1.x / v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator/(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x / v2.x,
+			v1.y / v2.y);
+	}
+
+	// -- Binary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator%(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x % scalar,
+			v.y % scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator%(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x % v2.x,
+			v1.y % v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator%(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar % v.x,
+			scalar % v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator%(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x % v2.x,
+			v1.x % v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator%(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x % v2.x,
+			v1.y % v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator&(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x & scalar,
+			v.y & scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator&(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x & v2.x,
+			v1.y & v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator&(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar & v.x,
+			scalar & v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator&(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x & v2.x,
+			v1.x & v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator&(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x & v2.x,
+			v1.y & v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator|(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x | scalar,
+			v.y | scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator|(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x | v2.x,
+			v1.y | v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator|(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar | v.x,
+			scalar | v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator|(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x | v2.x,
+			v1.x | v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator|(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x | v2.x,
+			v1.y | v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator^(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x ^ scalar,
+			v.y ^ scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator^(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x ^ v2.x,
+			v1.y ^ v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator^(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar ^ v.x,
+			scalar ^ v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator^(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x ^ v2.x,
+			v1.x ^ v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator^(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x ^ v2.x,
+			v1.y ^ v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x << scalar,
+			v.y << scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x << v2.x,
+			v1.y << v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator<<(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar << v.x,
+			scalar << v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x << v2.x,
+			v1.x << v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator<<(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x << v2.x,
+			v1.y << v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<2, T, Q> const& v, T scalar)
+	{
+		return vec<2, T, Q>(
+			v.x >> scalar,
+			v.y >> scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<2, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x >> v2.x,
+			v1.y >> v2.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator>>(T scalar, vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			scalar >> v.x,
+			scalar >> v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<1, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x >> v2.x,
+			v1.x >> v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator>>(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return vec<2, T, Q>(
+			v1.x >> v2.x,
+			v1.y >> v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, T, Q> operator~(vec<2, T, Q> const& v)
+	{
+		return vec<2, T, Q>(
+			~v.x,
+			~v.y);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return
+			detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.x, v2.x) &&
+			detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.y, v2.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(vec<2, T, Q> const& v1, vec<2, T, Q> const& v2)
+	{
+		return !(v1 == v2);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, bool, Q> operator&&(vec<2, bool, Q> const& v1, vec<2, bool, Q> const& v2)
+	{
+		return vec<2, bool, Q>(v1.x && v2.x, v1.y && v2.y);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<2, bool, Q> operator||(vec<2, bool, Q> const& v1, vec<2, bool, Q> const& v2)
+	{
+		return vec<2, bool, Q>(v1.x || v2.x, v1.y || v2.y);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.hpp
new file mode 100644
index 000000000000..67104800cfd9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.hpp
@@ -0,0 +1,435 @@
+/// @ref core
+/// @file glm/detail/type_vec3.hpp
+
+#pragma once
+
+#include "qualifier.hpp"
+#if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+#	include "_swizzle.hpp"
+#elif GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+#	include "_swizzle_func.hpp"
+#endif
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct vec<3, T, Q>
+	{
+		// -- Implementation detail --
+
+		typedef T value_type;
+		typedef vec<3, T, Q> type;
+		typedef vec<3, bool, Q> bool_type;
+
+		// -- Data --
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic push
+#				pragma GCC diagnostic ignored "-Wpedantic"
+#			elif GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic push
+#				pragma clang diagnostic ignored "-Wgnu-anonymous-struct"
+#				pragma clang diagnostic ignored "-Wnested-anon-types"
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(push)
+#				pragma warning(disable: 4201)  // nonstandard extension used : nameless struct/union
+#				if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+#					pragma warning(disable: 4324)  // structure was padded due to alignment specifier
+#				endif
+#			endif
+#		endif
+
+#		if GLM_CONFIG_XYZW_ONLY
+			T x, y, z;
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+			GLM_SWIZZLE_GEN_VEC_FROM_VEC3_COMP(T, Q, x, y, z)
+#			endif//GLM_CONFIG_SWIZZLE
+#		elif GLM_CONFIG_ANONYMOUS_STRUCT == GLM_ENABLE
+			union
+			{
+				struct{ T x, y, z; };
+				struct{ T r, g, b; };
+				struct{ T s, t, p; };
+
+				typename detail::storage<3, T, detail::is_aligned<Q>::value>::type data;
+
+#				if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+					GLM_SWIZZLE3_2_MEMBERS(T, Q, x, y, z)
+					GLM_SWIZZLE3_2_MEMBERS(T, Q, r, g, b)
+					GLM_SWIZZLE3_2_MEMBERS(T, Q, s, t, p)
+					GLM_SWIZZLE3_3_MEMBERS(T, Q, x, y, z)
+					GLM_SWIZZLE3_3_MEMBERS(T, Q, r, g, b)
+					GLM_SWIZZLE3_3_MEMBERS(T, Q, s, t, p)
+					GLM_SWIZZLE3_4_MEMBERS(T, Q, x, y, z)
+					GLM_SWIZZLE3_4_MEMBERS(T, Q, r, g, b)
+					GLM_SWIZZLE3_4_MEMBERS(T, Q, s, t, p)
+#				endif
+			};
+#		else
+			union { T x, r, s; };
+			union { T y, g, t; };
+			union { T z, b, p; };
+
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+				GLM_SWIZZLE_GEN_VEC_FROM_VEC3(T, Q)
+#			endif//GLM_CONFIG_SWIZZLE
+#		endif//GLM_LANG
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(pop)
+#			endif
+#		endif
+
+		// -- Component accesses --
+
+		/// Return the count of components of the vector
+		typedef length_t length_type;
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 3;}
+
+		GLM_FUNC_DECL GLM_CONSTEXPR T & operator[](length_type i);
+		GLM_FUNC_DECL GLM_CONSTEXPR T const& operator[](length_type i) const;
+
+		// -- Implicit basic constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR vec() GLM_DEFAULT_CTOR;
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec(vec const& v) GLM_DEFAULT;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<3, T, P> const& v);
+
+		// -- Explicit basic constructors --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(T a, T b, T c);
+
+		// -- Conversion scalar constructors --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(vec<1, U, P> const& v);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X x, Y y, Z z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, Y _y, Z _z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, vec<1, Y, Q> const& _y, Z _z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, Z _z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, Y _y, vec<1, Z, Q> const& _z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, Y _y, vec<1, Z, Q> const& _z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z);
+		template<typename X, typename Y, typename Z>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z);
+
+		// -- Conversion vector constructors --
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, B _z);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, vec<1, B, P> const& _z);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A _x, vec<2, B, P> const& _yz);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, P> const& _x, vec<2, B, P> const& _yz);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<4, U, P> const& v);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<3, U, P> const& v);
+
+		// -- Swizzle constructors --
+#		if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+			template<int E0, int E1, int E2>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<3, T, Q, E0, E1, E2, -1> const& that)
+			{
+				*this = that();
+			}
+
+			template<int E0, int E1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<2, T, Q, E0, E1, -1, -2> const& v, T const& scalar)
+			{
+				*this = vec(v(), scalar);
+			}
+
+			template<int E0, int E1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(T const& scalar, detail::_swizzle<2, T, Q, E0, E1, -1, -2> const& v)
+			{
+				*this = vec(scalar, v());
+			}
+#		endif//GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+
+		// -- Unary arithmetic operators --
+
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q>& operator=(vec<3, T, Q> const& v) GLM_DEFAULT;
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator+=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator+=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator+=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator-=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator-=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator-=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator*=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator*=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator*=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator/=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator/=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator/=(vec<3, U, Q> const& v);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator--(int);
+
+		// -- Unary bit operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator%=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator%=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator%=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator&=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator&=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator&=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator|=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator|=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator|=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator^=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator^=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator^=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator<<=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator<<=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator<<=(vec<3, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator>>=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator>>=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> & operator>>=(vec<3, U, Q> const& v);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator+(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator+(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator-(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator-(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator/(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator/(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator/(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator/(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator/(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator%(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator%(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator%(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator%(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator%(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator&(vec<3, T, Q> const& v1, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator&(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator&(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator&(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator&(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator|(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator|(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator|(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator|(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator|(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator^(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator^(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator^(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator^(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator^(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator<<(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<3, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<3, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator>>(T scalar, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<1, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> operator~(vec<3, T, Q> const& v);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, bool, Q> operator&&(vec<3, bool, Q> const& v1, vec<3, bool, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, bool, Q> operator||(vec<3, bool, Q> const& v1, vec<3, bool, Q> const& v2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_vec3.inl"
+#endif//GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.inl
new file mode 100644
index 000000000000..5a258d1f9b88
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec3.inl
@@ -0,0 +1,1070 @@
+/// @ref core
+
+#include "compute_vector_relational.hpp"
+
+namespace glm
+{
+	// -- Implicit basic constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec()
+#			if GLM_CONFIG_CTOR_INIT != GLM_CTOR_INIT_DISABLE
+				: x(0), y(0), z(0)
+#			endif
+		{}
+#	endif
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<3, T, Q> const& v)
+			: x(v.x), y(v.y), z(v.z)
+		{}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<3, T, P> const& v)
+		: x(v.x), y(v.y), z(v.z)
+	{}
+
+	// -- Explicit basic constructors --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(T scalar)
+		: x(scalar), y(scalar), z(scalar)
+	{}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(T _x, T _y, T _z)
+		: x(_x), y(_y), z(_z)
+	{}
+
+	// -- Conversion scalar constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<1, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.x))
+		, z(static_cast<T>(v.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(X _x, Y _y, Z _z)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<1, X, Q> const& _x, Y _y, Z _z)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(X _x, vec<1, Y, Q> const& _y, Z _z)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, Z _z)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(X _x, Y _y, vec<1, Z, Q> const& _z)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<1, X, Q> const& _x, Y _y, vec<1, Z, Q> const& _z)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(X _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z.x))
+	{}
+
+	// -- Conversion vector constructors --
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<2, A, P> const& _xy, B _z)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<2, A, P> const& _xy, vec<1, B, P> const& _z)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_z.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(A _x, vec<2, B, P> const& _yz)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_yz.x))
+		, z(static_cast<T>(_yz.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<1, A, P> const& _x, vec<2, B, P> const& _yz)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_yz.x))
+		, z(static_cast<T>(_yz.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<3, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.y))
+		, z(static_cast<T>(v.z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>::vec(vec<4, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.y))
+		, z(static_cast<T>(v.z))
+	{}
+
+	// -- Component accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T & vec<3, T, Q>::operator[](typename vec<3, T, Q>::length_type i)
+	{
+		assert(i >= 0 && i < this->length());
+		switch(i)
+		{
+		default:
+			case 0:
+		return x;
+			case 1:
+		return y;
+			case 2:
+		return z;
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T const& vec<3, T, Q>::operator[](typename vec<3, T, Q>::length_type i) const
+	{
+		assert(i >= 0 && i < this->length());
+		switch(i)
+		{
+		default:
+		case 0:
+			return x;
+		case 1:
+			return y;
+		case 2:
+			return z;
+		}
+	}
+
+	// -- Unary arithmetic operators --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>& vec<3, T, Q>::operator=(vec<3, T, Q> const& v)
+		{
+			this->x = v.x;
+			this->y = v.y;
+			this->z = v.z;
+			return *this;
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q>& vec<3, T, Q>::operator=(vec<3, U, Q> const& v)
+	{
+		this->x = static_cast<T>(v.x);
+		this->y = static_cast<T>(v.y);
+		this->z = static_cast<T>(v.z);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator+=(U scalar)
+	{
+		this->x += static_cast<T>(scalar);
+		this->y += static_cast<T>(scalar);
+		this->z += static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator+=(vec<1, U, Q> const& v)
+	{
+		this->x += static_cast<T>(v.x);
+		this->y += static_cast<T>(v.x);
+		this->z += static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator+=(vec<3, U, Q> const& v)
+	{
+		this->x += static_cast<T>(v.x);
+		this->y += static_cast<T>(v.y);
+		this->z += static_cast<T>(v.z);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator-=(U scalar)
+	{
+		this->x -= static_cast<T>(scalar);
+		this->y -= static_cast<T>(scalar);
+		this->z -= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator-=(vec<1, U, Q> const& v)
+	{
+		this->x -= static_cast<T>(v.x);
+		this->y -= static_cast<T>(v.x);
+		this->z -= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator-=(vec<3, U, Q> const& v)
+	{
+		this->x -= static_cast<T>(v.x);
+		this->y -= static_cast<T>(v.y);
+		this->z -= static_cast<T>(v.z);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator*=(U scalar)
+	{
+		this->x *= static_cast<T>(scalar);
+		this->y *= static_cast<T>(scalar);
+		this->z *= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator*=(vec<1, U, Q> const& v)
+	{
+		this->x *= static_cast<T>(v.x);
+		this->y *= static_cast<T>(v.x);
+		this->z *= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator*=(vec<3, U, Q> const& v)
+	{
+		this->x *= static_cast<T>(v.x);
+		this->y *= static_cast<T>(v.y);
+		this->z *= static_cast<T>(v.z);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator/=(U v)
+	{
+		this->x /= static_cast<T>(v);
+		this->y /= static_cast<T>(v);
+		this->z /= static_cast<T>(v);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator/=(vec<1, U, Q> const& v)
+	{
+		this->x /= static_cast<T>(v.x);
+		this->y /= static_cast<T>(v.x);
+		this->z /= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator/=(vec<3, U, Q> const& v)
+	{
+		this->x /= static_cast<T>(v.x);
+		this->y /= static_cast<T>(v.y);
+		this->z /= static_cast<T>(v.z);
+		return *this;
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator++()
+	{
+		++this->x;
+		++this->y;
+		++this->z;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator--()
+	{
+		--this->x;
+		--this->y;
+		--this->z;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> vec<3, T, Q>::operator++(int)
+	{
+		vec<3, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> vec<3, T, Q>::operator--(int)
+	{
+		vec<3, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator%=(U scalar)
+	{
+		this->x %= scalar;
+		this->y %= scalar;
+		this->z %= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator%=(vec<1, U, Q> const& v)
+	{
+		this->x %= v.x;
+		this->y %= v.x;
+		this->z %= v.x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator%=(vec<3, U, Q> const& v)
+	{
+		this->x %= v.x;
+		this->y %= v.y;
+		this->z %= v.z;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator&=(U scalar)
+	{
+		this->x &= scalar;
+		this->y &= scalar;
+		this->z &= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator&=(vec<1, U, Q> const& v)
+	{
+		this->x &= v.x;
+		this->y &= v.x;
+		this->z &= v.x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator&=(vec<3, U, Q> const& v)
+	{
+		this->x &= v.x;
+		this->y &= v.y;
+		this->z &= v.z;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator|=(U scalar)
+	{
+		this->x |= scalar;
+		this->y |= scalar;
+		this->z |= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator|=(vec<1, U, Q> const& v)
+	{
+		this->x |= v.x;
+		this->y |= v.x;
+		this->z |= v.x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator|=(vec<3, U, Q> const& v)
+	{
+		this->x |= v.x;
+		this->y |= v.y;
+		this->z |= v.z;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator^=(U scalar)
+	{
+		this->x ^= scalar;
+		this->y ^= scalar;
+		this->z ^= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator^=(vec<1, U, Q> const& v)
+	{
+		this->x ^= v.x;
+		this->y ^= v.x;
+		this->z ^= v.x;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator^=(vec<3, U, Q> const& v)
+	{
+		this->x ^= v.x;
+		this->y ^= v.y;
+		this->z ^= v.z;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator<<=(U scalar)
+	{
+		this->x <<= scalar;
+		this->y <<= scalar;
+		this->z <<= scalar;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator<<=(vec<1, U, Q> const& v)
+	{
+		this->x <<= static_cast<T>(v.x);
+		this->y <<= static_cast<T>(v.x);
+		this->z <<= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator<<=(vec<3, U, Q> const& v)
+	{
+		this->x <<= static_cast<T>(v.x);
+		this->y <<= static_cast<T>(v.y);
+		this->z <<= static_cast<T>(v.z);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator>>=(U scalar)
+	{
+		this->x >>= static_cast<T>(scalar);
+		this->y >>= static_cast<T>(scalar);
+		this->z >>= static_cast<T>(scalar);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator>>=(vec<1, U, Q> const& v)
+	{
+		this->x >>= static_cast<T>(v.x);
+		this->y >>= static_cast<T>(v.x);
+		this->z >>= static_cast<T>(v.x);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> & vec<3, T, Q>::operator>>=(vec<3, U, Q> const& v)
+	{
+		this->x >>= static_cast<T>(v.x);
+		this->y >>= static_cast<T>(v.y);
+		this->z >>= static_cast<T>(v.z);
+		return *this;
+	}
+
+	// -- Unary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			-v.x,
+			-v.y,
+			-v.z);
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x + scalar,
+			v.y + scalar,
+			v.z + scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x + scalar.x,
+			v.y + scalar.x,
+			v.z + scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator+(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar + v.x,
+			scalar + v.y,
+			scalar + v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator+(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x + v.x,
+			scalar.x + v.y,
+			scalar.x + v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator+(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x + v2.x,
+			v1.y + v2.y,
+			v1.z + v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x - scalar,
+			v.y - scalar,
+			v.z - scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x - scalar.x,
+			v.y - scalar.x,
+			v.z - scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator-(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar - v.x,
+			scalar - v.y,
+			scalar - v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator-(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x - v.x,
+			scalar.x - v.y,
+			scalar.x - v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator-(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x - v2.x,
+			v1.y - v2.y,
+			v1.z - v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x * scalar,
+			v.y * scalar,
+			v.z * scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x * scalar.x,
+			v.y * scalar.x,
+			v.z * scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar * v.x,
+			scalar * v.y,
+			scalar * v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x * v.x,
+			scalar.x * v.y,
+			scalar.x * v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator*(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x * v2.x,
+			v1.y * v2.y,
+			v1.z * v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator/(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x / scalar,
+			v.y / scalar,
+			v.z / scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator/(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x / scalar.x,
+			v.y / scalar.x,
+			v.z / scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator/(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar / v.x,
+			scalar / v.y,
+			scalar / v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator/(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x / v.x,
+			scalar.x / v.y,
+			scalar.x / v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator/(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x / v2.x,
+			v1.y / v2.y,
+			v1.z / v2.z);
+	}
+
+	// -- Binary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator%(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x % scalar,
+			v.y % scalar,
+			v.z % scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator%(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x % scalar.x,
+			v.y % scalar.x,
+			v.z % scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator%(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar % v.x,
+			scalar % v.y,
+			scalar % v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator%(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x % v.x,
+			scalar.x % v.y,
+			scalar.x % v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator%(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x % v2.x,
+			v1.y % v2.y,
+			v1.z % v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator&(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x & scalar,
+			v.y & scalar,
+			v.z & scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator&(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x & scalar.x,
+			v.y & scalar.x,
+			v.z & scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator&(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar & v.x,
+			scalar & v.y,
+			scalar & v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator&(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x & v.x,
+			scalar.x & v.y,
+			scalar.x & v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator&(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x & v2.x,
+			v1.y & v2.y,
+			v1.z & v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator|(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x | scalar,
+			v.y | scalar,
+			v.z | scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator|(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x | scalar.x,
+			v.y | scalar.x,
+			v.z | scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator|(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar | v.x,
+			scalar | v.y,
+			scalar | v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator|(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x | v.x,
+			scalar.x | v.y,
+			scalar.x | v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator|(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x | v2.x,
+			v1.y | v2.y,
+			v1.z | v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator^(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x ^ scalar,
+			v.y ^ scalar,
+			v.z ^ scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator^(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x ^ scalar.x,
+			v.y ^ scalar.x,
+			v.z ^ scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator^(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar ^ v.x,
+			scalar ^ v.y,
+			scalar ^ v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator^(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x ^ v.x,
+			scalar.x ^ v.y,
+			scalar.x ^ v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator^(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x ^ v2.x,
+			v1.y ^ v2.y,
+			v1.z ^ v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x << scalar,
+			v.y << scalar,
+			v.z << scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x << scalar.x,
+			v.y << scalar.x,
+			v.z << scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator<<(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar << v.x,
+			scalar << v.y,
+			scalar << v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x << v.x,
+			scalar.x << v.y,
+			scalar.x << v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator<<(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x << v2.x,
+			v1.y << v2.y,
+			v1.z << v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<3, T, Q> const& v, T scalar)
+	{
+		return vec<3, T, Q>(
+			v.x >> scalar,
+			v.y >> scalar,
+			v.z >> scalar);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<3, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<3, T, Q>(
+			v.x >> scalar.x,
+			v.y >> scalar.x,
+			v.z >> scalar.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator>>(T scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar >> v.x,
+			scalar >> v.y,
+			scalar >> v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<1, T, Q> const& scalar, vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			scalar.x >> v.x,
+			scalar.x >> v.y,
+			scalar.x >> v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator>>(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return vec<3, T, Q>(
+			v1.x >> v2.x,
+			v1.y >> v2.y,
+			v1.z >> v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> operator~(vec<3, T, Q> const& v)
+	{
+		return vec<3, T, Q>(
+			~v.x,
+			~v.y,
+			~v.z);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return
+			detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.x, v2.x) &&
+			detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.y, v2.y) &&
+			detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.z, v2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(vec<3, T, Q> const& v1, vec<3, T, Q> const& v2)
+	{
+		return !(v1 == v2);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, bool, Q> operator&&(vec<3, bool, Q> const& v1, vec<3, bool, Q> const& v2)
+	{
+		return vec<3, bool, Q>(v1.x && v2.x, v1.y && v2.y, v1.z && v2.z);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, bool, Q> operator||(vec<3, bool, Q> const& v1, vec<3, bool, Q> const& v2)
+	{
+		return vec<3, bool, Q>(v1.x || v2.x, v1.y || v2.y, v1.z || v2.z);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.hpp b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.hpp
new file mode 100644
index 000000000000..601256c37529
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.hpp
@@ -0,0 +1,508 @@
+/// @ref core
+/// @file glm/detail/type_vec4.hpp
+
+#pragma once
+
+#include "qualifier.hpp"
+#if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+#	include "_swizzle.hpp"
+#elif GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+#	include "_swizzle_func.hpp"
+#endif
+#include <cstddef>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	struct vec<4, T, Q>
+	{
+		// -- Implementation detail --
+
+		typedef T value_type;
+		typedef vec<4, T, Q> type;
+		typedef vec<4, bool, Q> bool_type;
+
+		// -- Data --
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic push
+#				pragma GCC diagnostic ignored "-Wpedantic"
+#			elif GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic push
+#				pragma clang diagnostic ignored "-Wgnu-anonymous-struct"
+#				pragma clang diagnostic ignored "-Wnested-anon-types"
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(push)
+#				pragma warning(disable: 4201)  // nonstandard extension used : nameless struct/union
+#			endif
+#		endif
+
+#		if GLM_CONFIG_XYZW_ONLY
+			T x, y, z, w;
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+			GLM_SWIZZLE_GEN_VEC_FROM_VEC4_COMP(T, Q, x, y, z, w)
+#			endif//GLM_CONFIG_SWIZZLE
+#		elif GLM_CONFIG_ANONYMOUS_STRUCT == GLM_ENABLE
+			union
+			{
+				struct { T x, y, z, w; };
+				struct { T r, g, b, a; };
+				struct { T s, t, p, q; };
+
+				typename detail::storage<4, T, detail::is_aligned<Q>::value>::type data;
+
+#				if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+					GLM_SWIZZLE4_2_MEMBERS(T, Q, x, y, z, w)
+					GLM_SWIZZLE4_2_MEMBERS(T, Q, r, g, b, a)
+					GLM_SWIZZLE4_2_MEMBERS(T, Q, s, t, p, q)
+					GLM_SWIZZLE4_3_MEMBERS(T, Q, x, y, z, w)
+					GLM_SWIZZLE4_3_MEMBERS(T, Q, r, g, b, a)
+					GLM_SWIZZLE4_3_MEMBERS(T, Q, s, t, p, q)
+					GLM_SWIZZLE4_4_MEMBERS(T, Q, x, y, z, w)
+					GLM_SWIZZLE4_4_MEMBERS(T, Q, r, g, b, a)
+					GLM_SWIZZLE4_4_MEMBERS(T, Q, s, t, p, q)
+#				endif
+			};
+#		else
+			union { T x, r, s; };
+			union { T y, g, t; };
+			union { T z, b, p; };
+			union { T w, a, q; };
+
+#			if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_FUNCTION
+				GLM_SWIZZLE_GEN_VEC_FROM_VEC4(T, Q)
+#			endif
+#		endif
+
+#		if GLM_SILENT_WARNINGS == GLM_ENABLE
+#			if GLM_COMPILER & GLM_COMPILER_CLANG
+#				pragma clang diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_GCC
+#				pragma GCC diagnostic pop
+#			elif GLM_COMPILER & GLM_COMPILER_VC
+#				pragma warning(pop)
+#			endif
+#		endif
+
+		// -- Component accesses --
+
+		typedef length_t length_type;
+
+		/// Return the count of components of the vector
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 4;}
+
+		GLM_FUNC_DECL GLM_CONSTEXPR T & operator[](length_type i);
+		GLM_FUNC_DECL GLM_CONSTEXPR T const& operator[](length_type i) const;
+
+		// -- Implicit basic constructors --
+
+		GLM_DEFAULTED_DEFAULT_CTOR_DECL GLM_CONSTEXPR vec() GLM_DEFAULT_CTOR;
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec(vec<4, T, Q> const& v) GLM_DEFAULT;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<4, T, P> const& v);
+
+		// -- Explicit basic constructors --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(T scalar);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(T x, T y, T z, T w);
+
+		// -- Conversion scalar constructors --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR explicit vec(vec<1, U, P> const& v);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, Y _y, Z _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, Y _y, Z _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, vec<1, Y, Q> const& _y, Z _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, Z _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, Y _y, vec<1, Z, Q> const& _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, Y _y, vec<1, Z, Q> const& _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, W _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, Y _y, Z _z, vec<1, W, Q> const& _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, vec<1, Y, Q> const& _y, Z _z, vec<1, W, Q> const& _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, Z _z, vec<1, W, Q> const& _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, Y _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, Y _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(X _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w);
+		template<typename X, typename Y, typename Z, typename W>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w);
+
+		// -- Conversion vector constructors --
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, B _z, C _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, vec<1, B, P> const& _z, C _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, B _z, vec<1, C, P> const& _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, vec<1, B, P> const& _z, vec<1, C, P> const& _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A _x, vec<2, B, P> const& _yz, C _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, P> const& _x, vec<2, B, P> const& _yz, C _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A _x, vec<2, B, P> const& _yz, vec<1, C, P> const& _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, P> const& _x, vec<2, B, P> const& _yz, vec<1, C, P> const& _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A _x, B _y, vec<2, C, P> const& _zw);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, P> const& _x, B _y, vec<2, C, P> const& _zw);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A _x, vec<1, B, P> const& _y, vec<2, C, P> const& _zw);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, typename C, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, P> const& _x, vec<1, B, P> const& _y, vec<2, C, P> const& _zw);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<3, A, P> const& _xyz, B _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<3, A, P> const& _xyz, vec<1, B, P> const& _w);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(A _x, vec<3, B, P> const& _yzw);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<1, A, P> const& _x, vec<3, B, P> const& _yzw);
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename A, typename B, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec(vec<2, A, P> const& _xy, vec<2, B, P> const& _zw);
+
+		/// Explicit conversions (From section 5.4.1 Conversion and scalar constructors of GLSL 1.30.08 specification)
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT vec(vec<4, U, P> const& v);
+
+		// -- Swizzle constructors --
+#		if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+			template<int E0, int E1, int E2, int E3>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<4, T, Q, E0, E1, E2, E3> const& that)
+			{
+				*this = that();
+			}
+
+			template<int E0, int E1, int F0, int F1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<2, T, Q, E0, E1, -1, -2> const& v, detail::_swizzle<2, T, Q, F0, F1, -1, -2> const& u)
+			{
+				*this = vec<4, T, Q>(v(), u());
+			}
+
+			template<int E0, int E1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(T const& x, T const& y, detail::_swizzle<2, T, Q, E0, E1, -1, -2> const& v)
+			{
+				*this = vec<4, T, Q>(x, y, v());
+			}
+
+			template<int E0, int E1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(T const& x, detail::_swizzle<2, T, Q, E0, E1, -1, -2> const& v, T const& w)
+			{
+				*this = vec<4, T, Q>(x, v(), w);
+			}
+
+			template<int E0, int E1>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<2, T, Q, E0, E1, -1, -2> const& v, T const& z, T const& w)
+			{
+				*this = vec<4, T, Q>(v(), z, w);
+			}
+
+			template<int E0, int E1, int E2>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(detail::_swizzle<3, T, Q, E0, E1, E2, -1> const& v, T const& w)
+			{
+				*this = vec<4, T, Q>(v(), w);
+			}
+
+			template<int E0, int E1, int E2>
+			GLM_FUNC_DECL GLM_CONSTEXPR vec(T const& x, detail::_swizzle<3, T, Q, E0, E1, E2, -1> const& v)
+			{
+				*this = vec<4, T, Q>(x, v());
+			}
+#		endif//GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+
+		// -- Unary arithmetic operators --
+
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator=(vec<4, T, Q> const& v) GLM_DEFAULT;
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator+=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator+=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator+=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator-=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator-=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator-=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator*=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator*=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator*=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator/=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator/=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q>& operator/=(vec<4, U, Q> const& v);
+
+		// -- Increment and decrement operators --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator++();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator--();
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator++(int);
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator--(int);
+
+		// -- Unary bit operators --
+
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator%=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator%=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator%=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator&=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator&=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator&=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator|=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator|=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator|=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator^=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator^=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator^=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator<<=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator<<=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator<<=(vec<4, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator>>=(U scalar);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator>>=(vec<1, U, Q> const& v);
+		template<typename U>
+		GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> & operator>>=(vec<4, U, Q> const& v);
+	};
+
+	// -- Unary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator+(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator+(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator-(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator-(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator/(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator/(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator/(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator/(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator/(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator%(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator%(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator%(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator%(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator%(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator&(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator&(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator&(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator&(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator&(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator|(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator|(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator|(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator|(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator|(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator^(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator^(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator^(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator^(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator^(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator<<(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<4, T, Q> const& v, T scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator>>(T scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, T, Q> operator~(vec<4, T, Q> const& v);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator==(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool operator!=(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, bool, Q> operator&&(vec<4, bool, Q> const& v1, vec<4, bool, Q> const& v2);
+
+	template<qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, bool, Q> operator||(vec<4, bool, Q> const& v1, vec<4, bool, Q> const& v2);
+}//namespace glm
+
+#ifndef GLM_EXTERNAL_TEMPLATE
+#include "type_vec4.inl"
+#endif//GLM_EXTERNAL_TEMPLATE
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.inl
new file mode 100644
index 000000000000..440de5fcb7ea
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4.inl
@@ -0,0 +1,1142 @@
+/// @ref core
+
+#include "compute_vector_relational.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_vec4_add
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x + b.x, a.y + b.y, a.z + b.z, a.w + b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_vec4_sub
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x - b.x, a.y - b.y, a.z - b.z, a.w - b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_vec4_mul
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x * b.x, a.y * b.y, a.z * b.z, a.w * b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_vec4_div
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x / b.x, a.y / b.y, a.z / b.z, a.w / b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_vec4_mod
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x % b.x, a.y % b.y, a.z % b.z, a.w % b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_and
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x & b.x, a.y & b.y, a.z & b.z, a.w & b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_or
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x | b.x, a.y | b.y, a.z | b.z, a.w | b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_xor
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_shift_left
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x << b.x, a.y << b.y, a.z << b.z, a.w << b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_shift_right
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+		{
+			return vec<4, T, Q>(a.x >> b.x, a.y >> b.y, a.z >> b.z, a.w >> b.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_equal
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static bool call(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+		{
+			return
+				detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.x, v2.x) &&
+				detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.y, v2.y) &&
+				detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.z, v2.z) &&
+				detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(v1.w, v2.w);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_nequal
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static bool call(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+		{
+			return !compute_vec4_equal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v1, v2);
+		}
+	};
+
+	template<typename T, qualifier Q, int IsInt, std::size_t Size, bool Aligned>
+	struct compute_vec4_bitwise_not
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static vec<4, T, Q> call(vec<4, T, Q> const& v)
+		{
+			return vec<4, T, Q>(~v.x, ~v.y, ~v.z, ~v.w);
+		}
+	};
+}//namespace detail
+
+	// -- Implicit basic constructors --
+
+#	if GLM_CONFIG_DEFAULTED_DEFAULT_CTOR == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_DEFAULT_CTOR_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec()
+#			if GLM_CONFIG_CTOR_INIT != GLM_CTOR_INIT_DISABLE
+				: x(0), y(0), z(0), w(0)
+#			endif
+		{}
+#	endif
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<4, T, Q> const& v)
+			: x(v.x), y(v.y), z(v.z), w(v.w)
+		{}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<4, T, P> const& v)
+		: x(v.x), y(v.y), z(v.z), w(v.w)
+	{}
+
+	// -- Explicit basic constructors --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(T scalar)
+		: x(scalar), y(scalar), z(scalar), w(scalar)
+	{}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(T _x, T _y, T _z, T _w)
+		: x(_x), y(_y), z(_z), w(_w)
+	{}
+
+	// -- Conversion scalar constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.x))
+		, z(static_cast<T>(v.x))
+		, w(static_cast<T>(v.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, Y _y, Z _z, W _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, Y _y, Z _z, W _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, vec<1, Y, Q> const& _y, Z _z, W _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, Z _z, W _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, Y _y, vec<1, Z, Q> const& _z, W _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, Y _y, vec<1, Z, Q> const& _z, W _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, W _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, W _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, Y _y, Z _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, vec<1, Y, Q> const& _y, Z _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, Z _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, Y _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, Y _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(X _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename X, typename Y, typename Z, typename W>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, X, Q> const& _x, vec<1, Y, Q> const& _y, vec<1, Z, Q> const& _z, vec<1, W, Q> const& _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	// -- Conversion vector constructors --
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<2, A, P> const& _xy, B _z, C _w)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<2, A, P> const& _xy, vec<1, B, P> const& _z, C _w)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<2, A, P> const& _xy, B _z, vec<1, C, P> const& _w)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_z))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<2, A, P> const& _xy, vec<1, B, P> const& _z, vec<1, C, P> const& _w)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_z.x))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(A _x, vec<2, B, P> const& _yz, C _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_yz.x))
+		, z(static_cast<T>(_yz.y))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, A, P> const& _x, vec<2, B, P> const& _yz, C _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_yz.x))
+		, z(static_cast<T>(_yz.y))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(A _x, vec<2, B, P> const& _yz, vec<1, C, P> const& _w)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_yz.x))
+		, z(static_cast<T>(_yz.y))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, A, P> const& _x, vec<2, B, P> const& _yz, vec<1, C, P> const& _w)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_yz.x))
+		, z(static_cast<T>(_yz.y))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(A _x, B _y, vec<2, C, P> const& _zw)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_zw.x))
+		, w(static_cast<T>(_zw.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, A, P> const& _x, B _y, vec<2, C, P> const& _zw)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y))
+		, z(static_cast<T>(_zw.x))
+		, w(static_cast<T>(_zw.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(A _x, vec<1, B, P> const& _y, vec<2, C, P> const& _zw)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_zw.x))
+		, w(static_cast<T>(_zw.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, typename C, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, A, P> const& _x, vec<1, B, P> const& _y, vec<2, C, P> const& _zw)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_y.x))
+		, z(static_cast<T>(_zw.x))
+		, w(static_cast<T>(_zw.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<3, A, P> const& _xyz, B _w)
+		: x(static_cast<T>(_xyz.x))
+		, y(static_cast<T>(_xyz.y))
+		, z(static_cast<T>(_xyz.z))
+		, w(static_cast<T>(_w))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<3, A, P> const& _xyz, vec<1, B, P> const& _w)
+		: x(static_cast<T>(_xyz.x))
+		, y(static_cast<T>(_xyz.y))
+		, z(static_cast<T>(_xyz.z))
+		, w(static_cast<T>(_w.x))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(A _x, vec<3, B, P> const& _yzw)
+		: x(static_cast<T>(_x))
+		, y(static_cast<T>(_yzw.x))
+		, z(static_cast<T>(_yzw.y))
+		, w(static_cast<T>(_yzw.z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<1, A, P> const& _x, vec<3, B, P> const& _yzw)
+		: x(static_cast<T>(_x.x))
+		, y(static_cast<T>(_yzw.x))
+		, z(static_cast<T>(_yzw.y))
+		, w(static_cast<T>(_yzw.z))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename A, typename B, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<2, A, P> const& _xy, vec<2, B, P> const& _zw)
+		: x(static_cast<T>(_xy.x))
+		, y(static_cast<T>(_xy.y))
+		, z(static_cast<T>(_zw.x))
+		, w(static_cast<T>(_zw.y))
+	{}
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>::vec(vec<4, U, P> const& v)
+		: x(static_cast<T>(v.x))
+		, y(static_cast<T>(v.y))
+		, z(static_cast<T>(v.z))
+		, w(static_cast<T>(v.w))
+	{}
+
+	// -- Component accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T& vec<4, T, Q>::operator[](typename vec<4, T, Q>::length_type i)
+	{
+		assert(i >= 0 && i < this->length());
+		switch(i)
+		{
+		default:
+		case 0:
+			return x;
+		case 1:
+			return y;
+		case 2:
+			return z;
+		case 3:
+			return w;
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T const& vec<4, T, Q>::operator[](typename vec<4, T, Q>::length_type i) const
+	{
+		assert(i >= 0 && i < this->length());
+		switch(i)
+		{
+		default:
+		case 0:
+			return x;
+		case 1:
+			return y;
+		case 2:
+			return z;
+		case 3:
+			return w;
+		}
+	}
+
+	// -- Unary arithmetic operators --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>& vec<4, T, Q>::operator=(vec<4, T, Q> const& v)
+		{
+			this->x = v.x;
+			this->y = v.y;
+			this->z = v.z;
+			this->w = v.w;
+			return *this;
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q>& vec<4, T, Q>::operator=(vec<4, U, Q> const& v)
+	{
+		this->x = static_cast<T>(v.x);
+		this->y = static_cast<T>(v.y);
+		this->z = static_cast<T>(v.z);
+		this->w = static_cast<T>(v.w);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator+=(U scalar)
+	{
+		return (*this = detail::compute_vec4_add<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator+=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_add<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator+=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_add<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator-=(U scalar)
+	{
+		return (*this = detail::compute_vec4_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator-=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator-=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_sub<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator*=(U scalar)
+	{
+		return (*this = detail::compute_vec4_mul<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator*=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_mul<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator*=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_mul<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator/=(U scalar)
+	{
+		return (*this = detail::compute_vec4_div<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator/=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_div<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v.x)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator/=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_div<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	// -- Increment and decrement operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator++()
+	{
+		++this->x;
+		++this->y;
+		++this->z;
+		++this->w;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator--()
+	{
+		--this->x;
+		--this->y;
+		--this->z;
+		--this->w;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> vec<4, T, Q>::operator++(int)
+	{
+		vec<4, T, Q> Result(*this);
+		++*this;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> vec<4, T, Q>::operator--(int)
+	{
+		vec<4, T, Q> Result(*this);
+		--*this;
+		return Result;
+	}
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator%=(U scalar)
+	{
+		return (*this = detail::compute_vec4_mod<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator%=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_mod<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator%=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_mod<T, Q, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator&=(U scalar)
+	{
+		return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator&=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator&=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_and<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator|=(U scalar)
+	{
+		return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator|=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator|=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_or<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator^=(U scalar)
+	{
+		return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator^=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator^=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_xor<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator<<=(U scalar)
+	{
+		return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator<<=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator<<=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_shift_left<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator>>=(U scalar)
+	{
+		return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(scalar)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator>>=(vec<1, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> & vec<4, T, Q>::operator>>=(vec<4, U, Q> const& v)
+	{
+		return (*this = detail::compute_vec4_shift_right<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(*this, vec<4, T, Q>(v)));
+	}
+
+	// -- Unary constant operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(0) -= v;
+	}
+
+	// -- Binary arithmetic operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) += scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) += v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator+(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(v) += scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator+(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v2) += v1;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator+(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) += v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) -= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) -= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator-(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) -= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator-(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) -= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator-(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) -= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) *= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) *= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(v) *= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v2) *= v1;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator*(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) *= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator/(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) /= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator/(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) /= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator/(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) /= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator/(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) /= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator/(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) /= v2;
+	}
+
+	// -- Binary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator%(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) %= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator%(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) %= v2.x;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator%(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) %= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator%(vec<1, T, Q> const& scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar.x) %= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator%(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) %= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator&(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) &= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator&(vec<4, T, Q> const& v, vec<1, T, Q> const& scalar)
+	{
+		return vec<4, T, Q>(v) &= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator&(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) &= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator&(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) &= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator&(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) &= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator|(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) |= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator|(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) |= v2.x;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator|(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) |= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator|(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) |= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator|(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) |= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator^(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) ^= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator^(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) ^= v2.x;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator^(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) ^= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator^(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) ^= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator^(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) ^= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) <<= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) <<= v2.x;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator<<(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) <<= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) <<= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator<<(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) <<= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<4, T, Q> const& v, T scalar)
+	{
+		return vec<4, T, Q>(v) >>= scalar;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<4, T, Q> const& v1, vec<1, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) >>= v2.x;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator>>(T scalar, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(scalar) >>= v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<1, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1.x) >>= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator>>(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return vec<4, T, Q>(v1) >>= v2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, T, Q> operator~(vec<4, T, Q> const& v)
+	{
+		return detail::compute_vec4_bitwise_not<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator==(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return detail::compute_vec4_equal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v1, v2);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool operator!=(vec<4, T, Q> const& v1, vec<4, T, Q> const& v2)
+	{
+		return detail::compute_vec4_nequal<T, Q, detail::is_int<T>::value, sizeof(T) * 8, detail::is_aligned<Q>::value>::call(v1, v2);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, bool, Q> operator&&(vec<4, bool, Q> const& v1, vec<4, bool, Q> const& v2)
+	{
+		return vec<4, bool, Q>(v1.x && v2.x, v1.y && v2.y, v1.z && v2.z, v1.w && v2.w);
+	}
+
+	template<qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, bool, Q> operator||(vec<4, bool, Q> const& v1, vec<4, bool, Q> const& v2)
+	{
+		return vec<4, bool, Q>(v1.x || v2.x, v1.y || v2.y, v1.z || v2.z, v1.w || v2.w);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "type_vec4_simd.inl"
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4_simd.inl
new file mode 100644
index 000000000000..816ef45bf0d6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/detail/type_vec4_simd.inl
@@ -0,0 +1,788 @@
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+namespace glm {
+	namespace detail
+	{
+#	if GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+		template<qualifier Q, int E0, int E1, int E2, int E3>
+		struct _swizzle_base1<4, float, Q, E0, E1, E2, E3, true> : public _swizzle_base0<float, 4>
+		{
+			GLM_FUNC_QUALIFIER vec<4, float, Q> operator ()()  const
+			{
+				__m128 data = *reinterpret_cast<__m128 const*>(&this->_buffer);
+
+				vec<4, float, Q> Result;
+#			if GLM_ARCH & GLM_ARCH_AVX_BIT
+				Result.data = _mm_permute_ps(data, _MM_SHUFFLE(E3, E2, E1, E0));
+#			else
+				Result.data = _mm_shuffle_ps(data, data, _MM_SHUFFLE(E3, E2, E1, E0));
+#			endif
+				return Result;
+			}
+		};
+
+		template<qualifier Q, int E0, int E1, int E2, int E3>
+		struct _swizzle_base1<4, int, Q, E0, E1, E2, E3, true> : public _swizzle_base0<int, 4>
+		{
+			GLM_FUNC_QUALIFIER vec<4, int, Q> operator ()()  const
+			{
+				__m128i data = *reinterpret_cast<__m128i const*>(&this->_buffer);
+
+				vec<4, int, Q> Result;
+				Result.data = _mm_shuffle_epi32(data, _MM_SHUFFLE(E3, E2, E1, E0));
+				return Result;
+			}
+		};
+
+		template<qualifier Q, int E0, int E1, int E2, int E3>
+		struct _swizzle_base1<4, uint, Q, E0, E1, E2, E3, true> : public _swizzle_base0<uint, 4>
+		{
+			GLM_FUNC_QUALIFIER vec<4, uint, Q> operator ()()  const
+			{
+				__m128i data = *reinterpret_cast<__m128i const*>(&this->_buffer);
+
+				vec<4, uint, Q> Result;
+				Result.data = _mm_shuffle_epi32(data, _MM_SHUFFLE(E3, E2, E1, E0));
+				return Result;
+			}
+		};
+#	endif// GLM_CONFIG_SWIZZLE == GLM_SWIZZLE_OPERATOR
+
+		template<qualifier Q>
+		struct compute_vec4_add<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = _mm_add_ps(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+		template<qualifier Q>
+		struct compute_vec4_add<double, Q, true>
+		{
+			static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
+			{
+				vec<4, double, Q> Result;
+				Result.data = _mm256_add_pd(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<qualifier Q>
+		struct compute_vec4_sub<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = _mm_sub_ps(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+		template<qualifier Q>
+		struct compute_vec4_sub<double, Q, true>
+		{
+			static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
+			{
+				vec<4, double, Q> Result;
+				Result.data = _mm256_sub_pd(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<qualifier Q>
+		struct compute_vec4_mul<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = _mm_mul_ps(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+		template<qualifier Q>
+		struct compute_vec4_mul<double, Q, true>
+		{
+			static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
+			{
+				vec<4, double, Q> Result;
+				Result.data = _mm256_mul_pd(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<qualifier Q>
+		struct compute_vec4_div<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = _mm_div_ps(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+		template<qualifier Q>
+		struct compute_vec4_div<double, Q, true>
+		{
+			static vec<4, double, Q> call(vec<4, double, Q> const& a, vec<4, double, Q> const& b)
+			{
+				vec<4, double, Q> Result;
+				Result.data = _mm256_div_pd(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<>
+		struct compute_vec4_div<float, aligned_lowp, true>
+		{
+			static vec<4, float, aligned_lowp> call(vec<4, float, aligned_lowp> const& a, vec<4, float, aligned_lowp> const& b)
+			{
+				vec<4, float, aligned_lowp> Result;
+				Result.data = _mm_mul_ps(a.data, _mm_rcp_ps(b.data));
+				return Result;
+			}
+		};
+
+		template<typename T, qualifier Q>
+		struct compute_vec4_and<T, Q, true, 32, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm_and_si128(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		template<typename T, qualifier Q>
+		struct compute_vec4_and<T, Q, true, 64, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm256_and_si256(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<typename T, qualifier Q>
+		struct compute_vec4_or<T, Q, true, 32, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm_or_si128(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		template<typename T, qualifier Q>
+		struct compute_vec4_or<T, Q, true, 64, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm256_or_si256(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<typename T, qualifier Q>
+		struct compute_vec4_xor<T, Q, true, 32, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm_xor_si128(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		template<typename T, qualifier Q>
+		struct compute_vec4_xor<T, Q, true, 64, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm256_xor_si256(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<typename T, qualifier Q>
+		struct compute_vec4_shift_left<T, Q, true, 32, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm_sll_epi32(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		template<typename T, qualifier Q>
+		struct compute_vec4_shift_left<T, Q, true, 64, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm256_sll_epi64(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<typename T, qualifier Q>
+		struct compute_vec4_shift_right<T, Q, true, 32, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm_srl_epi32(a.data, b.data);
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		template<typename T, qualifier Q>
+		struct compute_vec4_shift_right<T, Q, true, 64, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& a, vec<4, T, Q> const& b)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm256_srl_epi64(a.data, b.data);
+				return Result;
+			}
+		};
+#	endif
+
+		template<typename T, qualifier Q>
+		struct compute_vec4_bitwise_not<T, Q, true, 32, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& v)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm_xor_si128(v.data, _mm_set1_epi32(-1));
+				return Result;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		template<typename T, qualifier Q>
+		struct compute_vec4_bitwise_not<T, Q, true, 64, true>
+		{
+			static vec<4, T, Q> call(vec<4, T, Q> const& v)
+			{
+				vec<4, T, Q> Result;
+				Result.data = _mm256_xor_si256(v.data, _mm_set1_epi32(-1));
+				return Result;
+			}
+		};
+#	endif
+
+		template<qualifier Q>
+		struct compute_vec4_equal<float, Q, false, 32, true>
+		{
+			static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
+			{
+				return _mm_movemask_ps(_mm_cmpneq_ps(v1.data, v2.data)) == 0;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_SSE41_BIT
+		template<qualifier Q>
+		struct compute_vec4_equal<int, Q, true, 32, true>
+		{
+			static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
+			{
+				//return _mm_movemask_epi8(_mm_cmpeq_epi32(v1.data, v2.data)) != 0;
+				__m128i neq = _mm_xor_si128(v1.data, v2.data);
+				return _mm_test_all_zeros(neq, neq) == 0;
+			}
+		};
+#	endif
+
+		template<qualifier Q>
+		struct compute_vec4_nequal<float, Q, false, 32, true>
+		{
+			static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
+			{
+				return _mm_movemask_ps(_mm_cmpneq_ps(v1.data, v2.data)) != 0;
+			}
+		};
+
+#	if GLM_ARCH & GLM_ARCH_SSE41_BIT
+		template<qualifier Q>
+		struct compute_vec4_nequal<int, Q, true, 32, true>
+		{
+			static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
+			{
+				//return _mm_movemask_epi8(_mm_cmpneq_epi32(v1.data, v2.data)) != 0;
+				__m128i neq = _mm_xor_si128(v1.data, v2.data);
+				return _mm_test_all_zeros(neq, neq) != 0;
+			}
+		};
+#	endif
+	}//namespace detail
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_lowp>::vec(float _s) :
+		data(_mm_set1_ps(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_mediump>::vec(float _s) :
+		data(_mm_set1_ps(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(float _s) :
+		data(_mm_set1_ps(_s))
+	{}
+
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, double, aligned_lowp>::vec(double _s) :
+		data(_mm256_set1_pd(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, double, aligned_mediump>::vec(double _s) :
+		data(_mm256_set1_pd(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, double, aligned_highp>::vec(double _s) :
+		data(_mm256_set1_pd(_s))
+	{}
+#	endif
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_lowp>::vec(int _s) :
+		data(_mm_set1_epi32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_mediump>::vec(int _s) :
+		data(_mm_set1_epi32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_highp>::vec(int _s) :
+		data(_mm_set1_epi32(_s))
+	{}
+
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, detail::int64, aligned_lowp>::vec(detail::int64 _s) :
+		data(_mm256_set1_epi64x(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, detail::int64, aligned_mediump>::vec(detail::int64 _s) :
+		data(_mm256_set1_epi64x(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, detail::int64, aligned_highp>::vec(detail::int64 _s) :
+		data(_mm256_set1_epi64x(_s))
+	{}
+#	endif
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_lowp>::vec(float _x, float _y, float _z, float _w) :
+		data(_mm_set_ps(_w, _z, _y, _x))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_mediump>::vec(float _x, float _y, float _z, float _w) :
+		data(_mm_set_ps(_w, _z, _y, _x))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(float _x, float _y, float _z, float _w) :
+		data(_mm_set_ps(_w, _z, _y, _x))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_lowp>::vec(int _x, int _y, int _z, int _w) :
+		data(_mm_set_epi32(_w, _z, _y, _x))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_mediump>::vec(int _x, int _y, int _z, int _w) :
+		data(_mm_set_epi32(_w, _z, _y, _x))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_highp>::vec(int _x, int _y, int _z, int _w) :
+		data(_mm_set_epi32(_w, _z, _y, _x))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_lowp>::vec(int _x, int _y, int _z, int _w) :
+		data(_mm_cvtepi32_ps(_mm_set_epi32(_w, _z, _y, _x)))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_mediump>::vec(int _x, int _y, int _z, int _w) :
+		data(_mm_cvtepi32_ps(_mm_set_epi32(_w, _z, _y, _x)))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(int _x, int _y, int _z, int _w) :
+		data(_mm_cvtepi32_ps(_mm_set_epi32(_w, _z, _y, _x)))
+	{}
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+#if GLM_ARCH & GLM_ARCH_NEON_BIT
+namespace glm {
+	namespace detail {
+
+		template<qualifier Q>
+		struct compute_vec4_add<float, Q, true>
+		{
+			static
+				vec<4, float, Q>
+				call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = vaddq_f32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_add<uint, Q, true>
+		{
+			static
+				vec<4, uint, Q>
+				call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
+			{
+				vec<4, uint, Q> Result;
+				Result.data = vaddq_u32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_add<int, Q, true>
+		{
+			static
+				vec<4, int, Q>
+				call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
+			{
+				vec<4, int, Q> Result;
+				Result.data = vaddq_s32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_sub<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = vsubq_f32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_sub<uint, Q, true>
+		{
+			static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
+			{
+				vec<4, uint, Q> Result;
+				Result.data = vsubq_u32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_sub<int, Q, true>
+		{
+			static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
+			{
+				vec<4, int, Q> Result;
+				Result.data = vsubq_s32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_mul<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+				Result.data = vmulq_f32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_mul<uint, Q, true>
+		{
+			static vec<4, uint, Q> call(vec<4, uint, Q> const& a, vec<4, uint, Q> const& b)
+			{
+				vec<4, uint, Q> Result;
+				Result.data = vmulq_u32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_mul<int, Q, true>
+		{
+			static vec<4, int, Q> call(vec<4, int, Q> const& a, vec<4, int, Q> const& b)
+			{
+				vec<4, int, Q> Result;
+				Result.data = vmulq_s32(a.data, b.data);
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_div<float, Q, true>
+		{
+			static vec<4, float, Q> call(vec<4, float, Q> const& a, vec<4, float, Q> const& b)
+			{
+				vec<4, float, Q> Result;
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+				Result.data = vdivq_f32(a.data, b.data);
+#else
+				/* Arm assembler reference:
+				 *
+				 * The Newton-Raphson iteration: x[n+1] = x[n] * (2 - d * x[n])
+				 * converges to (1/d) if x0 is the result of VRECPE applied to d.
+				 *
+				 * Note: The precision usually improves with two interactions, but more than two iterations are not helpful. */
+				float32x4_t x = vrecpeq_f32(b.data);
+				x = vmulq_f32(vrecpsq_f32(b.data, x), x);
+				x = vmulq_f32(vrecpsq_f32(b.data, x), x);
+				Result.data = vmulq_f32(a.data, x);
+#endif
+				return Result;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_equal<float, Q, false, 32, true>
+		{
+			static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
+			{
+				uint32x4_t cmp = vceqq_f32(v1.data, v2.data);
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+				cmp = vpminq_u32(cmp, cmp);
+				cmp = vpminq_u32(cmp, cmp);
+				uint32_t r = cmp[0];
+#else
+				uint32x2_t cmpx2 = vpmin_u32(vget_low_u32(cmp), vget_high_u32(cmp));
+				cmpx2 = vpmin_u32(cmpx2, cmpx2);
+				uint32_t r = cmpx2[0];
+#endif
+				return r == ~0u;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_equal<uint, Q, false, 32, true>
+		{
+			static bool call(vec<4, uint, Q> const& v1, vec<4, uint, Q> const& v2)
+			{
+				uint32x4_t cmp = vceqq_u32(v1.data, v2.data);
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+				cmp = vpminq_u32(cmp, cmp);
+				cmp = vpminq_u32(cmp, cmp);
+				uint32_t r = cmp[0];
+#else
+				uint32x2_t cmpx2 = vpmin_u32(vget_low_u32(cmp), vget_high_u32(cmp));
+				cmpx2 = vpmin_u32(cmpx2, cmpx2);
+				uint32_t r = cmpx2[0];
+#endif
+				return r == ~0u;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_equal<int, Q, false, 32, true>
+		{
+			static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
+			{
+				uint32x4_t cmp = vceqq_s32(v1.data, v2.data);
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+				cmp = vpminq_u32(cmp, cmp);
+				cmp = vpminq_u32(cmp, cmp);
+				uint32_t r = cmp[0];
+#else
+				uint32x2_t cmpx2 = vpmin_u32(vget_low_u32(cmp), vget_high_u32(cmp));
+				cmpx2 = vpmin_u32(cmpx2, cmpx2);
+				uint32_t r = cmpx2[0];
+#endif
+				return r == ~0u;
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_nequal<float, Q, false, 32, true>
+		{
+			static bool call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2)
+			{
+				return !compute_vec4_equal<float, Q, false, 32, true>::call(v1, v2);
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_nequal<uint, Q, false, 32, true>
+		{
+			static bool call(vec<4, uint, Q> const& v1, vec<4, uint, Q> const& v2)
+			{
+				return !compute_vec4_equal<uint, Q, false, 32, true>::call(v1, v2);
+			}
+		};
+
+		template<qualifier Q>
+		struct compute_vec4_nequal<int, Q, false, 32, true>
+		{
+			static bool call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2)
+			{
+				return !compute_vec4_equal<int, Q, false, 32, true>::call(v1, v2);
+			}
+		};
+
+	}//namespace detail
+
+#if !GLM_CONFIG_XYZW_ONLY
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_lowp>::vec(float _s) :
+		data(vdupq_n_f32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_mediump>::vec(float _s) :
+		data(vdupq_n_f32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(float _s) :
+		data(vdupq_n_f32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_lowp>::vec(int _s) :
+		data(vdupq_n_s32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_mediump>::vec(int _s) :
+		data(vdupq_n_s32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, int, aligned_highp>::vec(int _s) :
+		data(vdupq_n_s32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, uint, aligned_lowp>::vec(uint _s) :
+		data(vdupq_n_u32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, uint, aligned_mediump>::vec(uint _s) :
+		data(vdupq_n_u32(_s))
+	{}
+
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, uint, aligned_highp>::vec(uint _s) :
+		data(vdupq_n_u32(_s))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(const vec<4, float, aligned_highp>& rhs) :
+		data(rhs.data)
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(const vec<4, int, aligned_highp>& rhs) :
+		data(vcvtq_f32_s32(rhs.data))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(const vec<4, uint, aligned_highp>& rhs) :
+		data(vcvtq_f32_u32(rhs.data))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_lowp>::vec(int _x, int _y, int _z, int _w) :
+		data(vcvtq_f32_s32(vec<4, int, aligned_lowp>(_x, _y, _z, _w).data))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_mediump>::vec(int _x, int _y, int _z, int _w) :
+		data(vcvtq_f32_s32(vec<4, int, aligned_mediump>(_x, _y, _z, _w).data))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(int _x, int _y, int _z, int _w) :
+		data(vcvtq_f32_s32(vec<4, int, aligned_highp>(_x, _y, _z, _w).data))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_lowp>::vec(uint _x, uint _y, uint _z, uint _w) :
+		data(vcvtq_f32_u32(vec<4, uint, aligned_lowp>(_x, _y, _z, _w).data))
+	{}
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_mediump>::vec(uint _x, uint _y, uint _z, uint _w) :
+		data(vcvtq_f32_u32(vec<4, uint, aligned_mediump>(_x, _y, _z, _w).data))
+	{}
+
+
+	template<>
+	template<>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, float, aligned_highp>::vec(uint _x, uint _y, uint _z, uint _w) :
+		data(vcvtq_f32_u32(vec<4, uint, aligned_highp>(_x, _y, _z, _w).data))
+	{}
+
+#endif
+}//namespace glm
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/exponential.hpp b/thirdparty/manifold/thirdparty/glm/glm/exponential.hpp
new file mode 100644
index 000000000000..1614f7695dbd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/exponential.hpp
@@ -0,0 +1,110 @@
+/// @ref core
+/// @file glm/exponential.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+///
+/// @defgroup core_func_exponential Exponential functions
+/// @ingroup core
+///
+/// Provides GLSL exponential functions
+///
+/// These all operate component-wise. The description is per component.
+///
+/// Include <glm/exponential.hpp> to use these core features.
+
+#pragma once
+
+#include "detail/type_vec1.hpp"
+#include "detail/type_vec2.hpp"
+#include "detail/type_vec3.hpp"
+#include "detail/type_vec4.hpp"
+#include <cmath>
+
+namespace glm
+{
+	/// @addtogroup core_func_exponential
+	/// @{
+
+	/// Returns 'base' raised to the power 'exponent'.
+	///
+	/// @param base Floating point value. pow function is defined for input values of 'base' defined in the range (inf-, inf+) in the limit of the type qualifier.
+	/// @param exponent Floating point value representing the 'exponent'.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/pow.xml">GLSL pow man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> pow(vec<L, T, Q> const& base, vec<L, T, Q> const& exponent);
+
+	/// Returns the natural exponentiation of v, i.e., e^v.
+	///
+	/// @param v exp function is defined for input values of v defined in the range (inf-, inf+) in the limit of the type qualifier.
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/exp.xml">GLSL exp man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> exp(vec<L, T, Q> const& v);
+
+	/// Returns the natural logarithm of v, i.e.,
+	/// returns the value y which satisfies the equation x = e^y.
+	/// Results are undefined if v <= 0.
+	///
+	/// @param v log function is defined for input values of v defined in the range (0, inf+) in the limit of the type qualifier.
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/log.xml">GLSL log man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> log(vec<L, T, Q> const& v);
+
+	/// Returns 2 raised to the v power.
+	///
+	/// @param v exp2 function is defined for input values of v defined in the range (inf-, inf+) in the limit of the type qualifier.
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/exp2.xml">GLSL exp2 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> exp2(vec<L, T, Q> const& v);
+
+	/// Returns the base 2 log of x, i.e., returns the value y,
+	/// which satisfies the equation x = 2 ^ y.
+	///
+	/// @param v log2 function is defined for input values of v defined in the range (0, inf+) in the limit of the type qualifier.
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/log2.xml">GLSL log2 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> log2(vec<L, T, Q> const& v);
+
+	/// Returns the positive square root of v.
+	///
+	/// @param v sqrt function is defined for input values of v defined in the range [0, inf+) in the limit of the type qualifier.
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/sqrt.xml">GLSL sqrt man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> sqrt(vec<L, T, Q> const& v);
+
+	/// Returns the reciprocal of the positive square root of v.
+	///
+	/// @param v inversesqrt function is defined for input values of v defined in the range [0, inf+) in the limit of the type qualifier.
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/inversesqrt.xml">GLSL inversesqrt man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.2 Exponential Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> inversesqrt(vec<L, T, Q> const& v);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_exponential.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext.hpp
new file mode 100644
index 000000000000..164dc2a18cc4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext.hpp
@@ -0,0 +1,266 @@
+/// @file glm/ext.hpp
+///
+/// @ref core (Dependence)
+
+#include "detail/setup.hpp"
+
+#pragma once
+
+#include "glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_MESSAGE_EXT_INCLUDED_DISPLAYED)
+#	define GLM_MESSAGE_EXT_INCLUDED_DISPLAYED
+#	pragma message("GLM: All extensions included (not recommended)")
+#endif//GLM_MESSAGES
+
+#include "./ext/matrix_clip_space.hpp"
+#include "./ext/matrix_common.hpp"
+
+#include "./ext/matrix_double2x2.hpp"
+#include "./ext/matrix_double2x2_precision.hpp"
+#include "./ext/matrix_double2x3.hpp"
+#include "./ext/matrix_double2x3_precision.hpp"
+#include "./ext/matrix_double2x4.hpp"
+#include "./ext/matrix_double2x4_precision.hpp"
+#include "./ext/matrix_double3x2.hpp"
+#include "./ext/matrix_double3x2_precision.hpp"
+#include "./ext/matrix_double3x3.hpp"
+#include "./ext/matrix_double3x3_precision.hpp"
+#include "./ext/matrix_double3x4.hpp"
+#include "./ext/matrix_double3x4_precision.hpp"
+#include "./ext/matrix_double4x2.hpp"
+#include "./ext/matrix_double4x2_precision.hpp"
+#include "./ext/matrix_double4x3.hpp"
+#include "./ext/matrix_double4x3_precision.hpp"
+#include "./ext/matrix_double4x4.hpp"
+#include "./ext/matrix_double4x4_precision.hpp"
+
+#include "./ext/matrix_float2x2.hpp"
+#include "./ext/matrix_float2x2_precision.hpp"
+#include "./ext/matrix_float2x3.hpp"
+#include "./ext/matrix_float2x3_precision.hpp"
+#include "./ext/matrix_float2x4.hpp"
+#include "./ext/matrix_float2x4_precision.hpp"
+#include "./ext/matrix_float3x2.hpp"
+#include "./ext/matrix_float3x2_precision.hpp"
+#include "./ext/matrix_float3x3.hpp"
+#include "./ext/matrix_float3x3_precision.hpp"
+#include "./ext/matrix_float3x4.hpp"
+#include "./ext/matrix_float3x4_precision.hpp"
+#include "./ext/matrix_float4x2.hpp"
+#include "./ext/matrix_float4x2_precision.hpp"
+#include "./ext/matrix_float4x3.hpp"
+#include "./ext/matrix_float4x3_precision.hpp"
+#include "./ext/matrix_float4x4.hpp"
+#include "./ext/matrix_float4x4_precision.hpp"
+
+#include "./ext/matrix_int2x2.hpp"
+#include "./ext/matrix_int2x2_sized.hpp"
+#include "./ext/matrix_int2x3.hpp"
+#include "./ext/matrix_int2x3_sized.hpp"
+#include "./ext/matrix_int2x4.hpp"
+#include "./ext/matrix_int2x4_sized.hpp"
+#include "./ext/matrix_int3x2.hpp"
+#include "./ext/matrix_int3x2_sized.hpp"
+#include "./ext/matrix_int3x3.hpp"
+#include "./ext/matrix_int3x3_sized.hpp"
+#include "./ext/matrix_int3x4.hpp"
+#include "./ext/matrix_int3x4_sized.hpp"
+#include "./ext/matrix_int4x2.hpp"
+#include "./ext/matrix_int4x2_sized.hpp"
+#include "./ext/matrix_int4x3.hpp"
+#include "./ext/matrix_int4x3_sized.hpp"
+#include "./ext/matrix_int4x4.hpp"
+#include "./ext/matrix_int4x4_sized.hpp"
+
+#include "./ext/matrix_uint2x2.hpp"
+#include "./ext/matrix_uint2x2_sized.hpp"
+#include "./ext/matrix_uint2x3.hpp"
+#include "./ext/matrix_uint2x3_sized.hpp"
+#include "./ext/matrix_uint2x4.hpp"
+#include "./ext/matrix_uint2x4_sized.hpp"
+#include "./ext/matrix_uint3x2.hpp"
+#include "./ext/matrix_uint3x2_sized.hpp"
+#include "./ext/matrix_uint3x3.hpp"
+#include "./ext/matrix_uint3x3_sized.hpp"
+#include "./ext/matrix_uint3x4.hpp"
+#include "./ext/matrix_uint3x4_sized.hpp"
+#include "./ext/matrix_uint4x2.hpp"
+#include "./ext/matrix_uint4x2_sized.hpp"
+#include "./ext/matrix_uint4x3.hpp"
+#include "./ext/matrix_uint4x3_sized.hpp"
+#include "./ext/matrix_uint4x4.hpp"
+#include "./ext/matrix_uint4x4_sized.hpp"
+
+#include "./ext/matrix_projection.hpp"
+#include "./ext/matrix_relational.hpp"
+#include "./ext/matrix_transform.hpp"
+
+#include "./ext/quaternion_common.hpp"
+#include "./ext/quaternion_double.hpp"
+#include "./ext/quaternion_double_precision.hpp"
+#include "./ext/quaternion_float.hpp"
+#include "./ext/quaternion_float_precision.hpp"
+#include "./ext/quaternion_exponential.hpp"
+#include "./ext/quaternion_geometric.hpp"
+#include "./ext/quaternion_relational.hpp"
+#include "./ext/quaternion_transform.hpp"
+#include "./ext/quaternion_trigonometric.hpp"
+
+#include "./ext/scalar_common.hpp"
+#include "./ext/scalar_constants.hpp"
+#include "./ext/scalar_integer.hpp"
+#include "./ext/scalar_packing.hpp"
+#include "./ext/scalar_reciprocal.hpp"
+#include "./ext/scalar_relational.hpp"
+#include "./ext/scalar_ulp.hpp"
+
+#include "./ext/scalar_int_sized.hpp"
+#include "./ext/scalar_uint_sized.hpp"
+
+#include "./ext/vector_common.hpp"
+#include "./ext/vector_integer.hpp"
+#include "./ext/vector_packing.hpp"
+#include "./ext/vector_reciprocal.hpp"
+#include "./ext/vector_relational.hpp"
+#include "./ext/vector_ulp.hpp"
+
+#include "./ext/vector_bool1.hpp"
+#include "./ext/vector_bool1_precision.hpp"
+#include "./ext/vector_bool2.hpp"
+#include "./ext/vector_bool2_precision.hpp"
+#include "./ext/vector_bool3.hpp"
+#include "./ext/vector_bool3_precision.hpp"
+#include "./ext/vector_bool4.hpp"
+#include "./ext/vector_bool4_precision.hpp"
+
+#include "./ext/vector_double1.hpp"
+#include "./ext/vector_double1_precision.hpp"
+#include "./ext/vector_double2.hpp"
+#include "./ext/vector_double2_precision.hpp"
+#include "./ext/vector_double3.hpp"
+#include "./ext/vector_double3_precision.hpp"
+#include "./ext/vector_double4.hpp"
+#include "./ext/vector_double4_precision.hpp"
+
+#include "./ext/vector_float1.hpp"
+#include "./ext/vector_float1_precision.hpp"
+#include "./ext/vector_float2.hpp"
+#include "./ext/vector_float2_precision.hpp"
+#include "./ext/vector_float3.hpp"
+#include "./ext/vector_float3_precision.hpp"
+#include "./ext/vector_float4.hpp"
+#include "./ext/vector_float4_precision.hpp"
+
+#include "./ext/vector_int1.hpp"
+#include "./ext/vector_int1_sized.hpp"
+#include "./ext/vector_int2.hpp"
+#include "./ext/vector_int2_sized.hpp"
+#include "./ext/vector_int3.hpp"
+#include "./ext/vector_int3_sized.hpp"
+#include "./ext/vector_int4.hpp"
+#include "./ext/vector_int4_sized.hpp"
+
+#include "./ext/vector_uint1.hpp"
+#include "./ext/vector_uint1_sized.hpp"
+#include "./ext/vector_uint2.hpp"
+#include "./ext/vector_uint2_sized.hpp"
+#include "./ext/vector_uint3.hpp"
+#include "./ext/vector_uint3_sized.hpp"
+#include "./ext/vector_uint4.hpp"
+#include "./ext/vector_uint4_sized.hpp"
+
+#include "./gtc/bitfield.hpp"
+#include "./gtc/color_space.hpp"
+#include "./gtc/constants.hpp"
+#include "./gtc/epsilon.hpp"
+#include "./gtc/integer.hpp"
+#include "./gtc/matrix_access.hpp"
+#include "./gtc/matrix_integer.hpp"
+#include "./gtc/matrix_inverse.hpp"
+#include "./gtc/matrix_transform.hpp"
+#include "./gtc/noise.hpp"
+#include "./gtc/packing.hpp"
+#include "./gtc/quaternion.hpp"
+#include "./gtc/random.hpp"
+#include "./gtc/reciprocal.hpp"
+#include "./gtc/round.hpp"
+#include "./gtc/type_precision.hpp"
+#include "./gtc/type_ptr.hpp"
+#include "./gtc/ulp.hpp"
+#include "./gtc/vec1.hpp"
+#if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+#	include "./gtc/type_aligned.hpp"
+#endif
+
+#ifdef GLM_ENABLE_EXPERIMENTAL
+#include "./gtx/associated_min_max.hpp"
+#include "./gtx/bit.hpp"
+#include "./gtx/closest_point.hpp"
+#include "./gtx/color_encoding.hpp"
+#include "./gtx/color_space.hpp"
+#include "./gtx/color_space_YCoCg.hpp"
+#include "./gtx/common.hpp"
+#include "./gtx/compatibility.hpp"
+#include "./gtx/component_wise.hpp"
+#include "./gtx/dual_quaternion.hpp"
+#include "./gtx/easing.hpp"
+#include "./gtx/euler_angles.hpp"
+#include "./gtx/extend.hpp"
+#include "./gtx/extended_min_max.hpp"
+#include "./gtx/fast_exponential.hpp"
+#include "./gtx/fast_square_root.hpp"
+#include "./gtx/fast_trigonometry.hpp"
+#include "./gtx/functions.hpp"
+#include "./gtx/gradient_paint.hpp"
+#include "./gtx/handed_coordinate_space.hpp"
+
+#if __cplusplus >= 201103L
+#include "./gtx/hash.hpp"
+#endif
+
+#include "./gtx/integer.hpp"
+#include "./gtx/intersect.hpp"
+#include "./gtx/io.hpp"
+#include "./gtx/log_base.hpp"
+#include "./gtx/matrix_cross_product.hpp"
+#include "./gtx/matrix_decompose.hpp"
+#include "./gtx/matrix_factorisation.hpp"
+#include "./gtx/matrix_interpolation.hpp"
+#include "./gtx/matrix_major_storage.hpp"
+#include "./gtx/matrix_operation.hpp"
+#include "./gtx/matrix_query.hpp"
+#include "./gtx/mixed_product.hpp"
+#include "./gtx/norm.hpp"
+#include "./gtx/normal.hpp"
+#include "./gtx/normalize_dot.hpp"
+#include "./gtx/optimum_pow.hpp"
+#include "./gtx/orthonormalize.hpp"
+#include "./gtx/pca.hpp"
+#include "./gtx/perpendicular.hpp"
+#include "./gtx/polar_coordinates.hpp"
+#include "./gtx/projection.hpp"
+#include "./gtx/quaternion.hpp"
+#include "./gtx/raw_data.hpp"
+#include "./gtx/rotate_normalized_axis.hpp"
+#include "./gtx/rotate_vector.hpp"
+#include "./gtx/spline.hpp"
+#include "./gtx/std_based_type.hpp"
+#if !((GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP))
+#	include "./gtx/string_cast.hpp"
+#endif
+#include "./gtx/transform.hpp"
+#include "./gtx/transform2.hpp"
+#include "./gtx/vec_swizzle.hpp"
+#include "./gtx/vector_angle.hpp"
+#include "./gtx/vector_query.hpp"
+#include "./gtx/wrap.hpp"
+
+#if GLM_HAS_TEMPLATE_ALIASES
+#	include "./gtx/scalar_multiplication.hpp"
+#endif
+
+#if GLM_HAS_RANGE_FOR
+#	include "./gtx/range.hpp"
+#endif
+#endif//GLM_ENABLE_EXPERIMENTAL
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/_matrix_vectorize.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/_matrix_vectorize.hpp
new file mode 100644
index 000000000000..0d08117ed1d1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/_matrix_vectorize.hpp
@@ -0,0 +1,128 @@
+#pragma once
+
+namespace glm {
+
+	namespace detail {
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, length_t C, length_t R, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1 {
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 2, 2, Ret, T, Q> {
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<2, 2, T, Q> call(Ret (*Func)(T x), mat<2, 2, T, Q> const &x) {
+				return mat<2, 2, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]),
+					Func(x[1][0]), Func(x[1][1])
+				);
+			}
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 2, 3, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<2, 3, T, Q> call(Ret (*Func)(T x), mat<2, 3, T, Q> const &x) {
+				return mat<2, 3, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]), Func(x[0][2]),
+					Func(x[1][0]), Func(x[1][1]), Func(x[1][2])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 2, 4, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<2, 4, T, Q> call(Ret (*Func)(T x), mat<2, 4, T, Q> const &x) {
+				return mat<2, 4, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]), Func(x[0][2]), Func(x[0][3]),
+					Func(x[1][0]), Func(x[1][1]), Func(x[1][2]), Func(x[1][3])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 3, 2, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<3, 2, T, Q> call(Ret (*Func)(T x), mat<3, 2, T, Q> const &x) {
+				return mat<3, 2, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]),
+					Func(x[1][0]), Func(x[1][1]),
+					Func(x[2][0]), Func(x[2][1])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 3, 3, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<3, 3, T, Q> call(Ret (*Func)(T x), mat<3, 3, T, Q> const &x) {
+				return mat<3, 3, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]), Func(x[0][2]),
+					Func(x[1][0]), Func(x[1][1]), Func(x[1][2]),
+					Func(x[2][0]), Func(x[2][1]), Func(x[2][2])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 3, 4, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<3, 4, T, Q> call(Ret (*Func)(T x), mat<3, 4, T, Q> const &x) {
+				return mat<3, 4, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]), Func(x[0][2]), Func(x[0][3]),
+					Func(x[1][0]), Func(x[1][1]), Func(x[1][2]), Func(x[1][3]),
+					Func(x[2][0]), Func(x[2][1]), Func(x[2][2]), Func(x[2][3])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 4, 2, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<4, 2, T, Q> call(Ret (*Func)(T x), mat<4, 2, T, Q> const &x) {
+				return mat<4, 2, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]),
+					Func(x[1][0]), Func(x[1][1]),
+					Func(x[2][0]), Func(x[2][1]),
+					Func(x[3][0]), Func(x[3][1])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 4, 3, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<4, 3, T, Q> call(Ret (*Func)(T x), mat<4, 3, T, Q> const &x) {
+				return mat<4, 3, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]), Func(x[0][2]),
+					Func(x[1][0]), Func(x[1][1]), Func(x[1][2]),
+					Func(x[2][0]), Func(x[2][1]), Func(x[2][2]),
+					Func(x[3][0]), Func(x[3][1]), Func(x[3][2])
+				);
+			}
+
+		};
+
+		template<template<length_t C, length_t R, typename T, qualifier Q> class mat, typename Ret, typename T, qualifier Q>
+		struct matrix_functor_1<mat, 4, 4, Ret, T, Q> {
+
+			GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<4, 4, T, Q> call(Ret (*Func)(T x), mat<4, 4, T, Q> const &x) {
+				return mat<4, 4, Ret, Q>(
+					Func(x[0][0]), Func(x[0][1]), Func(x[0][2]), Func(x[0][3]),
+					Func(x[1][0]), Func(x[1][1]), Func(x[1][2]), Func(x[1][3]),
+					Func(x[2][0]), Func(x[2][1]), Func(x[2][2]), Func(x[2][3]),
+					Func(x[3][0]), Func(x[3][1]), Func(x[3][2]), Func(x[3][3])
+				);
+			}
+
+		};
+
+	}
+
+}// namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.hpp
new file mode 100644
index 000000000000..43579b8eee6c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.hpp
@@ -0,0 +1,522 @@
+/// @ref ext_matrix_clip_space
+/// @file glm/ext/matrix_clip_space.hpp
+///
+/// @defgroup ext_matrix_clip_space GLM_EXT_matrix_clip_space
+/// @ingroup ext
+///
+/// Defines functions that generate clip space transformation matrices.
+///
+/// The matrices generated by this extension use standard OpenGL fixed-function
+/// conventions. For example, the lookAt function generates a transform from world
+/// space into the specific eye space that the projective matrix functions
+/// (perspective, ortho, etc) are designed to expect. The OpenGL compatibility
+/// specifications defines the particular layout of this eye space.
+///
+/// Include <glm/ext/matrix_clip_space.hpp> to use the features of this extension.
+///
+/// @see ext_matrix_transform
+/// @see ext_matrix_projection
+
+#pragma once
+
+// Dependencies
+#include "../ext/scalar_constants.hpp"
+#include "../geometric.hpp"
+#include "../trigonometric.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_clip_space extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_clip_space
+	/// @{
+
+	/// Creates a matrix for projecting two-dimensional coordinates onto the screen.
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top, T const& zNear, T const& zFar)
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluOrtho2D.xml">gluOrtho2D man page</a>
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> ortho(
+		T left, T right, T bottom, T top);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using left-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoLH_ZO(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume using left-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoLH_NO(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using right-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoRH_ZO(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using right-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoRH_NO(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using left-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoZO(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoNO(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using left-handed coordinates.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoLH(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using right-handed coordinates.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> orthoRH(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a matrix for an orthographic parallel viewing volume, using the default handedness and default near and far clip planes definition.
+	/// To change default handedness use GLM_FORCE_LEFT_HANDED. To change default near and far clip planes definition use GLM_FORCE_DEPTH_ZERO_TO_ONE.
+	///
+	/// @tparam T A floating-point scalar type
+	///
+	/// @see - glm::ortho(T const& left, T const& right, T const& bottom, T const& top)
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glOrtho.xml">glOrtho man page</a>
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> ortho(
+		T left, T right, T bottom, T top, T zNear, T zFar);
+
+	/// Creates a left-handed frustum matrix.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumLH_ZO(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a left-handed frustum matrix.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumLH_NO(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a right-handed frustum matrix.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumRH_ZO(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a right-handed frustum matrix.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumRH_NO(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a frustum matrix using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumZO(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a frustum matrix using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumNO(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a left-handed frustum matrix.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumLH(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a right-handed frustum matrix.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustumRH(
+		T left, T right, T bottom, T top, T near, T far);
+
+	/// Creates a frustum matrix with default handedness, using the default handedness and default near and far clip planes definition.
+	/// To change default handedness use GLM_FORCE_LEFT_HANDED. To change default near and far clip planes definition use GLM_FORCE_DEPTH_ZERO_TO_ONE.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glFrustum.xml">glFrustum man page</a>
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> frustum(
+		T left, T right, T bottom, T top, T near, T far);
+
+
+	/// Creates a matrix for a right-handed, symmetric perspective-view frustum.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveRH_ZO(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a right-handed, symmetric perspective-view frustum.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveRH_NO(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a left-handed, symmetric perspective-view frustum.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveLH_ZO(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a left-handed, symmetric perspective-view frustum.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveLH_NO(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a symmetric perspective-view frustum using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveZO(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a symmetric perspective-view frustum using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveNO(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a right-handed, symmetric perspective-view frustum.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveRH(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a left-handed, symmetric perspective-view frustum.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveLH(
+		T fovy, T aspect, T near, T far);
+
+	/// Creates a matrix for a symmetric perspective-view frustum based on the default handedness and default near and far clip planes definition.
+	/// To change default handedness use GLM_FORCE_LEFT_HANDED. To change default near and far clip planes definition use GLM_FORCE_DEPTH_ZERO_TO_ONE.
+	///
+	/// @param fovy Specifies the field of view angle in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluPerspective.xml">gluPerspective man page</a>
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspective(
+		T fovy, T aspect, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view using right-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovRH_ZO(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view using right-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovRH_NO(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view using left-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovLH_ZO(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view using left-handed coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovLH_NO(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovZO(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view using left-handed coordinates if GLM_FORCE_LEFT_HANDED if defined or right-handed coordinates otherwise.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovNO(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a right-handed perspective projection matrix based on a field of view.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovRH(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a left-handed perspective projection matrix based on a field of view.
+	/// If GLM_FORCE_DEPTH_ZERO_TO_ONE is defined, the near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	/// Otherwise, the near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFovLH(
+		T fov, T width, T height, T near, T far);
+
+	/// Builds a perspective projection matrix based on a field of view and the default handedness and default near and far clip planes definition.
+	/// To change default handedness use GLM_FORCE_LEFT_HANDED. To change default near and far clip planes definition use GLM_FORCE_DEPTH_ZERO_TO_ONE.
+	///
+	/// @param fov Expressed in radians.
+	/// @param width Width of the viewport
+	/// @param height Height of the viewport
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param far Specifies the distance from the viewer to the far clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> perspectiveFov(
+		T fov, T width, T height, T near, T far);
+
+	/// Creates a matrix for a left-handed, symmetric perspective-view frustum with far plane at infinite.
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> infinitePerspectiveLH(
+		T fovy, T aspect, T near);
+
+	/// Creates a matrix for a right-handed, symmetric perspective-view frustum with far plane at infinite.
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> infinitePerspectiveRH(
+		T fovy, T aspect, T near);
+
+	/// Creates a matrix for a symmetric perspective-view frustum with far plane at infinite with default handedness.
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> infinitePerspective(
+		T fovy, T aspect, T near);
+
+	/// Creates a matrix for a symmetric perspective-view frustum with far plane at infinite for graphics hardware that doesn't support depth clamping.
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> tweakedInfinitePerspective(
+		T fovy, T aspect, T near);
+
+	/// Creates a matrix for a symmetric perspective-view frustum with far plane at infinite for graphics hardware that doesn't support depth clamping.
+	///
+	/// @param fovy Specifies the field of view angle, in degrees, in the y direction. Expressed in radians.
+	/// @param aspect Specifies the aspect ratio that determines the field of view in the x direction. The aspect ratio is the ratio of x (width) to y (height).
+	/// @param near Specifies the distance from the viewer to the near clipping plane (always positive).
+	/// @param ep Epsilon
+	///
+	/// @tparam T A floating-point scalar type
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> tweakedInfinitePerspective(
+		T fovy, T aspect, T near, T ep);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_clip_space.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.inl
new file mode 100644
index 000000000000..27fb6a13f75d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_clip_space.inl
@@ -0,0 +1,595 @@
+namespace glm
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> ortho(T left, T right, T bottom, T top)
+	{
+		mat<4, 4, T, defaultp> Result(static_cast<T>(1));
+		Result[0][0] = static_cast<T>(2) / (right - left);
+		Result[1][1] = static_cast<T>(2) / (top - bottom);
+		Result[2][2] = - static_cast<T>(1);
+		Result[3][0] = - (right + left) / (right - left);
+		Result[3][1] = - (top + bottom) / (top - bottom);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoLH_ZO(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+		mat<4, 4, T, defaultp> Result(1);
+		Result[0][0] = static_cast<T>(2) / (right - left);
+		Result[1][1] = static_cast<T>(2) / (top - bottom);
+		Result[2][2] = static_cast<T>(1) / (zFar - zNear);
+		Result[3][0] = - (right + left) / (right - left);
+		Result[3][1] = - (top + bottom) / (top - bottom);
+		Result[3][2] = - zNear / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoLH_NO(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+		mat<4, 4, T, defaultp> Result(1);
+		Result[0][0] = static_cast<T>(2) / (right - left);
+		Result[1][1] = static_cast<T>(2) / (top - bottom);
+		Result[2][2] = static_cast<T>(2) / (zFar - zNear);
+		Result[3][0] = - (right + left) / (right - left);
+		Result[3][1] = - (top + bottom) / (top - bottom);
+		Result[3][2] = - (zFar + zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoRH_ZO(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+		mat<4, 4, T, defaultp> Result(1);
+		Result[0][0] = static_cast<T>(2) / (right - left);
+		Result[1][1] = static_cast<T>(2) / (top - bottom);
+		Result[2][2] = - static_cast<T>(1) / (zFar - zNear);
+		Result[3][0] = - (right + left) / (right - left);
+		Result[3][1] = - (top + bottom) / (top - bottom);
+		Result[3][2] = - zNear / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoRH_NO(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+		mat<4, 4, T, defaultp> Result(1);
+		Result[0][0] = static_cast<T>(2) / (right - left);
+		Result[1][1] = static_cast<T>(2) / (top - bottom);
+		Result[2][2] = - static_cast<T>(2) / (zFar - zNear);
+		Result[3][0] = - (right + left) / (right - left);
+		Result[3][1] = - (top + bottom) / (top - bottom);
+		Result[3][2] = - (zFar + zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoZO(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return orthoLH_ZO(left, right, bottom, top, zNear, zFar);
+#		else
+			return orthoRH_ZO(left, right, bottom, top, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoNO(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return orthoLH_NO(left, right, bottom, top, zNear, zFar);
+#		else
+			return orthoRH_NO(left, right, bottom, top, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoLH(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return orthoLH_ZO(left, right, bottom, top, zNear, zFar);
+#		else
+			return orthoLH_NO(left, right, bottom, top, zNear, zFar);
+#		endif
+
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> orthoRH(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return orthoRH_ZO(left, right, bottom, top, zNear, zFar);
+#		else
+			return orthoRH_NO(left, right, bottom, top, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> ortho(T left, T right, T bottom, T top, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_ZO
+			return orthoLH_ZO(left, right, bottom, top, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_NO
+			return orthoLH_NO(left, right, bottom, top, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_ZO
+			return orthoRH_ZO(left, right, bottom, top, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_NO
+			return orthoRH_NO(left, right, bottom, top, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumLH_ZO(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+		mat<4, 4, T, defaultp> Result(0);
+		Result[0][0] = (static_cast<T>(2) * nearVal) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * nearVal) / (top - bottom);
+		Result[2][0] = -(right + left) / (right - left);
+		Result[2][1] = -(top + bottom) / (top - bottom);
+		Result[2][2] = farVal / (farVal - nearVal);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = -(farVal * nearVal) / (farVal - nearVal);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumLH_NO(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+		mat<4, 4, T, defaultp> Result(0);
+		Result[0][0] = (static_cast<T>(2) * nearVal) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * nearVal) / (top - bottom);
+		Result[2][0] = -(right + left) / (right - left);
+		Result[2][1] = -(top + bottom) / (top - bottom);
+		Result[2][2] = (farVal + nearVal) / (farVal - nearVal);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = - (static_cast<T>(2) * farVal * nearVal) / (farVal - nearVal);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumRH_ZO(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+		mat<4, 4, T, defaultp> Result(0);
+		Result[0][0] = (static_cast<T>(2) * nearVal) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * nearVal) / (top - bottom);
+		Result[2][0] = (right + left) / (right - left);
+		Result[2][1] = (top + bottom) / (top - bottom);
+		Result[2][2] = farVal / (nearVal - farVal);
+		Result[2][3] = static_cast<T>(-1);
+		Result[3][2] = -(farVal * nearVal) / (farVal - nearVal);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumRH_NO(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+		mat<4, 4, T, defaultp> Result(0);
+		Result[0][0] = (static_cast<T>(2) * nearVal) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * nearVal) / (top - bottom);
+		Result[2][0] = (right + left) / (right - left);
+		Result[2][1] = (top + bottom) / (top - bottom);
+		Result[2][2] = - (farVal + nearVal) / (farVal - nearVal);
+		Result[2][3] = static_cast<T>(-1);
+		Result[3][2] = - (static_cast<T>(2) * farVal * nearVal) / (farVal - nearVal);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumZO(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return frustumLH_ZO(left, right, bottom, top, nearVal, farVal);
+#		else
+			return frustumRH_ZO(left, right, bottom, top, nearVal, farVal);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumNO(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return frustumLH_NO(left, right, bottom, top, nearVal, farVal);
+#		else
+			return frustumRH_NO(left, right, bottom, top, nearVal, farVal);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumLH(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return frustumLH_ZO(left, right, bottom, top, nearVal, farVal);
+#		else
+			return frustumLH_NO(left, right, bottom, top, nearVal, farVal);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustumRH(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return frustumRH_ZO(left, right, bottom, top, nearVal, farVal);
+#		else
+			return frustumRH_NO(left, right, bottom, top, nearVal, farVal);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> frustum(T left, T right, T bottom, T top, T nearVal, T farVal)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_ZO
+			return frustumLH_ZO(left, right, bottom, top, nearVal, farVal);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_NO
+			return frustumLH_NO(left, right, bottom, top, nearVal, farVal);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_ZO
+			return frustumRH_ZO(left, right, bottom, top, nearVal, farVal);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_NO
+			return frustumRH_NO(left, right, bottom, top, nearVal, farVal);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveRH_ZO(T fovy, T aspect, T zNear, T zFar)
+	{
+		assert(abs(aspect - std::numeric_limits<T>::epsilon()) > static_cast<T>(0));
+
+		T const tanHalfFovy = tan(fovy / static_cast<T>(2));
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = static_cast<T>(1) / (aspect * tanHalfFovy);
+		Result[1][1] = static_cast<T>(1) / (tanHalfFovy);
+		Result[2][2] = zFar / (zNear - zFar);
+		Result[2][3] = - static_cast<T>(1);
+		Result[3][2] = -(zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveRH_NO(T fovy, T aspect, T zNear, T zFar)
+	{
+		assert(abs(aspect - std::numeric_limits<T>::epsilon()) > static_cast<T>(0));
+
+		T const tanHalfFovy = tan(fovy / static_cast<T>(2));
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = static_cast<T>(1) / (aspect * tanHalfFovy);
+		Result[1][1] = static_cast<T>(1) / (tanHalfFovy);
+		Result[2][2] = - (zFar + zNear) / (zFar - zNear);
+		Result[2][3] = - static_cast<T>(1);
+		Result[3][2] = - (static_cast<T>(2) * zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveLH_ZO(T fovy, T aspect, T zNear, T zFar)
+	{
+		assert(abs(aspect - std::numeric_limits<T>::epsilon()) > static_cast<T>(0));
+
+		T const tanHalfFovy = tan(fovy / static_cast<T>(2));
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = static_cast<T>(1) / (aspect * tanHalfFovy);
+		Result[1][1] = static_cast<T>(1) / (tanHalfFovy);
+		Result[2][2] = zFar / (zFar - zNear);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = -(zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveLH_NO(T fovy, T aspect, T zNear, T zFar)
+	{
+		assert(abs(aspect - std::numeric_limits<T>::epsilon()) > static_cast<T>(0));
+
+		T const tanHalfFovy = tan(fovy / static_cast<T>(2));
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = static_cast<T>(1) / (aspect * tanHalfFovy);
+		Result[1][1] = static_cast<T>(1) / (tanHalfFovy);
+		Result[2][2] = (zFar + zNear) / (zFar - zNear);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = - (static_cast<T>(2) * zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveZO(T fovy, T aspect, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return perspectiveLH_ZO(fovy, aspect, zNear, zFar);
+#		else
+			return perspectiveRH_ZO(fovy, aspect, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveNO(T fovy, T aspect, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return perspectiveLH_NO(fovy, aspect, zNear, zFar);
+#		else
+			return perspectiveRH_NO(fovy, aspect, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveLH(T fovy, T aspect, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return perspectiveLH_ZO(fovy, aspect, zNear, zFar);
+#		else
+			return perspectiveLH_NO(fovy, aspect, zNear, zFar);
+#		endif
+
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveRH(T fovy, T aspect, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return perspectiveRH_ZO(fovy, aspect, zNear, zFar);
+#		else
+			return perspectiveRH_NO(fovy, aspect, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspective(T fovy, T aspect, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_ZO
+			return perspectiveLH_ZO(fovy, aspect, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_NO
+			return perspectiveLH_NO(fovy, aspect, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_ZO
+			return perspectiveRH_ZO(fovy, aspect, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_NO
+			return perspectiveRH_NO(fovy, aspect, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovRH_ZO(T fov, T width, T height, T zNear, T zFar)
+	{
+		assert(width > static_cast<T>(0));
+		assert(height > static_cast<T>(0));
+		assert(fov > static_cast<T>(0));
+
+		T const rad = fov;
+		T const h = glm::cos(static_cast<T>(0.5) * rad) / glm::sin(static_cast<T>(0.5) * rad);
+		T const w = h * height / width; ///todo max(width , Height) / min(width , Height)?
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = w;
+		Result[1][1] = h;
+		Result[2][2] = zFar / (zNear - zFar);
+		Result[2][3] = - static_cast<T>(1);
+		Result[3][2] = -(zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovRH_NO(T fov, T width, T height, T zNear, T zFar)
+	{
+		assert(width > static_cast<T>(0));
+		assert(height > static_cast<T>(0));
+		assert(fov > static_cast<T>(0));
+
+		T const rad = fov;
+		T const h = glm::cos(static_cast<T>(0.5) * rad) / glm::sin(static_cast<T>(0.5) * rad);
+		T const w = h * height / width; ///todo max(width , Height) / min(width , Height)?
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = w;
+		Result[1][1] = h;
+		Result[2][2] = - (zFar + zNear) / (zFar - zNear);
+		Result[2][3] = - static_cast<T>(1);
+		Result[3][2] = - (static_cast<T>(2) * zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovLH_ZO(T fov, T width, T height, T zNear, T zFar)
+	{
+		assert(width > static_cast<T>(0));
+		assert(height > static_cast<T>(0));
+		assert(fov > static_cast<T>(0));
+
+		T const rad = fov;
+		T const h = glm::cos(static_cast<T>(0.5) * rad) / glm::sin(static_cast<T>(0.5) * rad);
+		T const w = h * height / width; ///todo max(width , Height) / min(width , Height)?
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = w;
+		Result[1][1] = h;
+		Result[2][2] = zFar / (zFar - zNear);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = -(zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovLH_NO(T fov, T width, T height, T zNear, T zFar)
+	{
+		assert(width > static_cast<T>(0));
+		assert(height > static_cast<T>(0));
+		assert(fov > static_cast<T>(0));
+
+		T const rad = fov;
+		T const h = glm::cos(static_cast<T>(0.5) * rad) / glm::sin(static_cast<T>(0.5) * rad);
+		T const w = h * height / width; ///todo max(width , Height) / min(width , Height)?
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = w;
+		Result[1][1] = h;
+		Result[2][2] = (zFar + zNear) / (zFar - zNear);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = - (static_cast<T>(2) * zFar * zNear) / (zFar - zNear);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovZO(T fov, T width, T height, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return perspectiveFovLH_ZO(fov, width, height, zNear, zFar);
+#		else
+			return perspectiveFovRH_ZO(fov, width, height, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovNO(T fov, T width, T height, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return perspectiveFovLH_NO(fov, width, height, zNear, zFar);
+#		else
+			return perspectiveFovRH_NO(fov, width, height, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovLH(T fov, T width, T height, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return perspectiveFovLH_ZO(fov, width, height, zNear, zFar);
+#		else
+			return perspectiveFovLH_NO(fov, width, height, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFovRH(T fov, T width, T height, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return perspectiveFovRH_ZO(fov, width, height, zNear, zFar);
+#		else
+			return perspectiveFovRH_NO(fov, width, height, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> perspectiveFov(T fov, T width, T height, T zNear, T zFar)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_ZO
+			return perspectiveFovLH_ZO(fov, width, height, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_NO
+			return perspectiveFovLH_NO(fov, width, height, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_ZO
+			return perspectiveFovRH_ZO(fov, width, height, zNear, zFar);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_NO
+			return perspectiveFovRH_NO(fov, width, height, zNear, zFar);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> infinitePerspectiveRH_NO(T fovy, T aspect, T zNear)
+	{
+		T const range = tan(fovy / static_cast<T>(2)) * zNear;
+		T const left = -range * aspect;
+		T const right = range * aspect;
+		T const bottom = -range;
+		T const top = range;
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = (static_cast<T>(2) * zNear) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * zNear) / (top - bottom);
+		Result[2][2] = - static_cast<T>(1);
+		Result[2][3] = - static_cast<T>(1);
+		Result[3][2] = - static_cast<T>(2) * zNear;
+		return Result;
+	}
+	
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> infinitePerspectiveRH_ZO(T fovy, T aspect, T zNear)
+	{
+		T const range = tan(fovy / static_cast<T>(2)) * zNear;
+		T const left = -range * aspect;
+		T const right = range * aspect;
+		T const bottom = -range;
+		T const top = range;
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = (static_cast<T>(2) * zNear) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * zNear) / (top - bottom);
+		Result[2][2] = - static_cast<T>(1);
+		Result[2][3] = - static_cast<T>(1);
+		Result[3][2] = - zNear;
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> infinitePerspectiveLH_NO(T fovy, T aspect, T zNear)
+	{
+		T const range = tan(fovy / static_cast<T>(2)) * zNear;
+		T const left = -range * aspect;
+		T const right = range * aspect;
+		T const bottom = -range;
+		T const top = range;
+
+		mat<4, 4, T, defaultp> Result(T(0));
+		Result[0][0] = (static_cast<T>(2) * zNear) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * zNear) / (top - bottom);
+		Result[2][2] = static_cast<T>(1);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = - static_cast<T>(2) * zNear;
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> infinitePerspectiveLH_ZO(T fovy, T aspect, T zNear)
+	{
+		T const range = tan(fovy / static_cast<T>(2)) * zNear;
+		T const left = -range * aspect;
+		T const right = range * aspect;
+		T const bottom = -range;
+		T const top = range;
+
+		mat<4, 4, T, defaultp> Result(T(0));
+		Result[0][0] = (static_cast<T>(2) * zNear) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * zNear) / (top - bottom);
+		Result[2][2] = static_cast<T>(1);
+		Result[2][3] = static_cast<T>(1);
+		Result[3][2] = - zNear;
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> infinitePerspective(T fovy, T aspect, T zNear)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_ZO
+			return infinitePerspectiveLH_ZO(fovy, aspect, zNear);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_LH_NO
+			return infinitePerspectiveLH_NO(fovy, aspect, zNear);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_ZO
+			return infinitePerspectiveRH_ZO(fovy, aspect, zNear);
+#		elif GLM_CONFIG_CLIP_CONTROL == GLM_CLIP_CONTROL_RH_NO
+			return infinitePerspectiveRH_NO(fovy, aspect, zNear);
+#		endif
+	}
+
+	// Infinite projection matrix: http://www.terathon.com/gdc07_lengyel.pdf
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> tweakedInfinitePerspective(T fovy, T aspect, T zNear, T ep)
+	{
+		T const range = tan(fovy / static_cast<T>(2)) * zNear;
+		T const left = -range * aspect;
+		T const right = range * aspect;
+		T const bottom = -range;
+		T const top = range;
+
+		mat<4, 4, T, defaultp> Result(static_cast<T>(0));
+		Result[0][0] = (static_cast<T>(2) * zNear) / (right - left);
+		Result[1][1] = (static_cast<T>(2) * zNear) / (top - bottom);
+		Result[2][2] = ep - static_cast<T>(1);
+		Result[2][3] = static_cast<T>(-1);
+		Result[3][2] = (ep - static_cast<T>(2)) * zNear;
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> tweakedInfinitePerspective(T fovy, T aspect, T zNear)
+	{
+		return tweakedInfinitePerspective(fovy, aspect, zNear, epsilon<T>());
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.hpp
new file mode 100644
index 000000000000..6bb3d06e7ece
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.hpp
@@ -0,0 +1,39 @@
+/// @ref ext_matrix_common
+/// @file glm/ext/matrix_common.hpp
+///
+/// @defgroup ext_matrix_common GLM_EXT_matrix_common
+/// @ingroup ext
+///
+/// Defines functions for common matrix operations.
+///
+/// Include <glm/ext/matrix_common.hpp> to use the features of this extension.
+///
+/// @see ext_matrix_common
+
+#pragma once
+
+#include "../detail/qualifier.hpp"
+#include "../detail/_fixes.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_common extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_common
+	/// @{
+
+	template<length_t C, length_t R, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> mix(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, mat<C, R, U, Q> const& a);
+
+	template<length_t C, length_t R, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> mix(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, U a);
+
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<C, R, T, Q> abs(mat<C, R, T, Q> const& x);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_common.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.inl
new file mode 100644
index 000000000000..1be422202edb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_common.inl
@@ -0,0 +1,34 @@
+#include "../matrix.hpp"
+
+#include "_matrix_vectorize.hpp"
+
+namespace glm
+{
+	template<length_t C, length_t R, typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<C, R, T, Q> mix(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, U a)
+	{
+		return mat<C, R, U, Q>(x) * (static_cast<U>(1) - a) + mat<C, R, U, Q>(y) * a;
+	}
+
+	template<length_t C, length_t R, typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<C, R, T, Q> mix(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, mat<C, R, U, Q> const& a)
+	{
+		return matrixCompMult(mat<C, R, U, Q>(x), static_cast<U>(1) - a) + matrixCompMult(mat<C, R, U, Q>(y), a);
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_abs_matrix
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static mat<C, R, T, Q> call(mat<C, R, T, Q> const& x)
+		{
+			return detail::matrix_functor_1<mat, C, R, T, T, Q>::call(abs, x);
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<C, R, T, Q> abs(mat<C, R, T, Q> const& x)
+	{
+		return compute_abs_matrix<C, R, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2.hpp
new file mode 100644
index 000000000000..94dca54b59bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2.hpp
@@ -0,0 +1,23 @@
+/// @ref core
+/// @file glm/ext/matrix_double2x2.hpp
+
+#pragma once
+#include "../detail/type_mat2x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 2, double, defaultp>		dmat2x2;
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 2, double, defaultp>		dmat2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2_precision.hpp
new file mode 100644
index 000000000000..9e2c174e43be
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x2_precision.hpp
@@ -0,0 +1,49 @@
+/// @ref core
+/// @file glm/ext/matrix_double2x2_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, double, lowp>		lowp_dmat2;
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, double, mediump>	mediump_dmat2;
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, double, highp>	highp_dmat2;
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, double, lowp>		lowp_dmat2x2;
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, double, mediump>	mediump_dmat2x2;
+
+	/// 2 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, double, highp>	highp_dmat2x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3.hpp
new file mode 100644
index 000000000000..bfef87a666c1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_double2x3.hpp
+
+#pragma once
+#include "../detail/type_mat2x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 2 columns of 3 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 3, double, defaultp>		dmat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3_precision.hpp
new file mode 100644
index 000000000000..098fb6046e88
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x3_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_double2x3_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 2 columns of 3 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 3, double, lowp>		lowp_dmat2x3;
+
+	/// 2 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 3, double, mediump>	mediump_dmat2x3;
+
+	/// 2 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 3, double, highp>	highp_dmat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4.hpp
new file mode 100644
index 000000000000..499284bce161
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_double2x4.hpp
+
+#pragma once
+#include "../detail/type_mat2x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 2 columns of 4 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 4, double, defaultp>		dmat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4_precision.hpp
new file mode 100644
index 000000000000..9b61ebcee1ef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double2x4_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_double2x4_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 2 columns of 4 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 4, double, lowp>		lowp_dmat2x4;
+
+	/// 2 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 4, double, mediump>	mediump_dmat2x4;
+
+	/// 2 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 4, double, highp>	highp_dmat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2.hpp
new file mode 100644
index 000000000000..dd23f36cdbb2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_double3x2.hpp
+
+#pragma once
+#include "../detail/type_mat3x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 3 columns of 2 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 2, double, defaultp>		dmat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2_precision.hpp
new file mode 100644
index 000000000000..068d9e911721
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_double3x2_precision.hpp
+
+#pragma once
+#include "../detail/type_mat3x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 3 columns of 2 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 2, double, lowp>		lowp_dmat3x2;
+
+	/// 3 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 2, double, mediump>	mediump_dmat3x2;
+
+	/// 3 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 2, double, highp>	highp_dmat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3.hpp
new file mode 100644
index 000000000000..53572b735626
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3.hpp
@@ -0,0 +1,23 @@
+/// @ref core
+/// @file glm/ext/matrix_double3x3.hpp
+
+#pragma once
+#include "../detail/type_mat3x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 3, double, defaultp>		dmat3x3;
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 3, double, defaultp>		dmat3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3_precision.hpp
new file mode 100644
index 000000000000..8691e7808dc9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x3_precision.hpp
@@ -0,0 +1,49 @@
+/// @ref core
+/// @file glm/ext/matrix_double3x3_precision.hpp
+
+#pragma once
+#include "../detail/type_mat3x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, double, lowp>		lowp_dmat3;
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, double, mediump>	mediump_dmat3;
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, double, highp>	highp_dmat3;
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, double, lowp>		lowp_dmat3x3;
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, double, mediump>	mediump_dmat3x3;
+
+	/// 3 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, double, highp>	highp_dmat3x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4.hpp
new file mode 100644
index 000000000000..c572d637cd2d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_double3x4.hpp
+
+#pragma once
+#include "../detail/type_mat3x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 3 columns of 4 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 4, double, defaultp>		dmat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4_precision.hpp
new file mode 100644
index 000000000000..f040217e748a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double3x4_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_double3x4_precision.hpp
+
+#pragma once
+#include "../detail/type_mat3x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 3 columns of 4 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 4, double, lowp>		lowp_dmat3x4;
+
+	/// 3 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 4, double, mediump>	mediump_dmat3x4;
+
+	/// 3 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 4, double, highp>	highp_dmat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2.hpp
new file mode 100644
index 000000000000..9b229f471e8d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_double4x2.hpp
+
+#pragma once
+#include "../detail/type_mat4x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 4 columns of 2 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 2, double, defaultp>		dmat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2_precision.hpp
new file mode 100644
index 000000000000..6ad18ba9e65e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_double4x2_precision.hpp
+
+#pragma once
+#include "../detail/type_mat4x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 4 columns of 2 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 2, double, lowp>		lowp_dmat4x2;
+
+	/// 4 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 2, double, mediump>	mediump_dmat4x2;
+
+	/// 4 columns of 2 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 2, double, highp>	highp_dmat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3.hpp
new file mode 100644
index 000000000000..dca4cf956f91
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_double4x3.hpp
+
+#pragma once
+#include "../detail/type_mat4x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 4 columns of 3 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 3, double, defaultp>		dmat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3_precision.hpp
new file mode 100644
index 000000000000..f7371de84942
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x3_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_double4x3_precision.hpp
+
+#pragma once
+#include "../detail/type_mat4x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 4 columns of 3 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 3, double, lowp>		lowp_dmat4x3;
+
+	/// 4 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 3, double, mediump>	mediump_dmat4x3;
+
+	/// 4 columns of 3 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 3, double, highp>	highp_dmat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4.hpp
new file mode 100644
index 000000000000..81e1bf65cb5c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4.hpp
@@ -0,0 +1,23 @@
+/// @ref core
+/// @file glm/ext/matrix_double4x4.hpp
+
+#pragma once
+#include "../detail/type_mat4x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 4, double, defaultp>		dmat4x4;
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 4, double, defaultp>		dmat4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4_precision.hpp
new file mode 100644
index 000000000000..4c36a8486c72
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_double4x4_precision.hpp
@@ -0,0 +1,49 @@
+/// @ref core
+/// @file glm/ext/matrix_double4x4_precision.hpp
+
+#pragma once
+#include "../detail/type_mat4x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, double, lowp>		lowp_dmat4;
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, double, mediump>	mediump_dmat4;
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, double, highp>	highp_dmat4;
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, double, lowp>		lowp_dmat4x4;
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, double, mediump>	mediump_dmat4x4;
+
+	/// 4 columns of 4 components matrix of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, double, highp>	highp_dmat4x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2.hpp
new file mode 100644
index 000000000000..53df921fe216
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2.hpp
@@ -0,0 +1,23 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x2.hpp
+
+#pragma once
+#include "../detail/type_mat2x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 2, float, defaultp>		mat2x2;
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 2, float, defaultp>		mat2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2_precision.hpp
new file mode 100644
index 000000000000..898b6db71408
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x2_precision.hpp
@@ -0,0 +1,49 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x2_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, float, lowp>		lowp_mat2;
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, float, mediump>	mediump_mat2;
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, float, highp>		highp_mat2;
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, float, lowp>		lowp_mat2x2;
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, float, mediump>	mediump_mat2x2;
+
+	/// 2 columns of 2 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 2, float, highp>		highp_mat2x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3.hpp
new file mode 100644
index 000000000000..6f68822dbf1e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x3.hpp
+
+#pragma once
+#include "../detail/type_mat2x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 2 columns of 3 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 3, float, defaultp>		mat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3_precision.hpp
new file mode 100644
index 000000000000..50c103245c3a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x3_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x3_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 2 columns of 3 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 3, float, lowp>		lowp_mat2x3;
+
+	/// 2 columns of 3 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 3, float, mediump>	mediump_mat2x3;
+
+	/// 2 columns of 3 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 3, float, highp>		highp_mat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4.hpp
new file mode 100644
index 000000000000..30f30de3cbd4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x4.hpp
+
+#pragma once
+#include "../detail/type_mat2x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 2 columns of 4 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<2, 4, float, defaultp>		mat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4_precision.hpp
new file mode 100644
index 000000000000..079d63828631
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float2x4_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x4_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 2 columns of 4 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 4, float, lowp>		lowp_mat2x4;
+
+	/// 2 columns of 4 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 4, float, mediump>	mediump_mat2x4;
+
+	/// 2 columns of 4 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<2, 4, float, highp>		highp_mat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2.hpp
new file mode 100644
index 000000000000..280d0a3e974b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_float3x2.hpp
+
+#pragma once
+#include "../detail/type_mat3x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 3 columns of 2 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 2, float, defaultp>			mat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2_precision.hpp
new file mode 100644
index 000000000000..8572c2a1b20e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_float3x2_precision.hpp
+
+#pragma once
+#include "../detail/type_mat3x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 3 columns of 2 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 2, float, lowp>		lowp_mat3x2;
+
+	/// 3 columns of 2 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 2, float, mediump>	mediump_mat3x2;
+
+	/// 3 columns of 2 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 2, float, highp>		highp_mat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3.hpp
new file mode 100644
index 000000000000..177d809ff9f4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3.hpp
@@ -0,0 +1,23 @@
+/// @ref core
+/// @file glm/ext/matrix_float3x3.hpp
+
+#pragma once
+#include "../detail/type_mat3x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 3, float, defaultp>			mat3x3;
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 3, float, defaultp>			mat3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3_precision.hpp
new file mode 100644
index 000000000000..8a900c164200
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x3_precision.hpp
@@ -0,0 +1,49 @@
+/// @ref core
+/// @file glm/ext/matrix_float3x3_precision.hpp
+
+#pragma once
+#include "../detail/type_mat3x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, float, lowp>		lowp_mat3;
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, float, mediump>	mediump_mat3;
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, float, highp>		highp_mat3;
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, float, lowp>		lowp_mat3x3;
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, float, mediump>	mediump_mat3x3;
+
+	/// 3 columns of 3 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 3, float, highp>		highp_mat3x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4.hpp
new file mode 100644
index 000000000000..64b8459dcdd2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_float3x4.hpp
+
+#pragma once
+#include "../detail/type_mat3x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 3 columns of 4 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<3, 4, float, defaultp>			mat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4_precision.hpp
new file mode 100644
index 000000000000..bc36bf13a1e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float3x4_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_float3x4_precision.hpp
+
+#pragma once
+#include "../detail/type_mat3x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 3 columns of 4 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 4, float, lowp>		lowp_mat3x4;
+
+	/// 3 columns of 4 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 4, float, mediump>	mediump_mat3x4;
+
+	/// 3 columns of 4 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<3, 4, float, highp>		highp_mat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2.hpp
new file mode 100644
index 000000000000..1ed5227bf580
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_float4x2.hpp
+
+#pragma once
+#include "../detail/type_mat4x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 4 columns of 2 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 2, float, defaultp>			mat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2_precision.hpp
new file mode 100644
index 000000000000..88fd069630a8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_float2x2_precision.hpp
+
+#pragma once
+#include "../detail/type_mat2x2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 4 columns of 2 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 2, float, lowp>		lowp_mat4x2;
+
+	/// 4 columns of 2 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 2, float, mediump>	mediump_mat4x2;
+
+	/// 4 columns of 2 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 2, float, highp>		highp_mat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3.hpp
new file mode 100644
index 000000000000..5dbe7657043f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/matrix_float4x3.hpp
+
+#pragma once
+#include "../detail/type_mat4x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix
+	/// @{
+
+	/// 4 columns of 3 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 3, float, defaultp>			mat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3_precision.hpp
new file mode 100644
index 000000000000..846ed4fc8d9c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x3_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/matrix_float4x3_precision.hpp
+
+#pragma once
+#include "../detail/type_mat4x3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 4 columns of 3 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 3, float, lowp>		lowp_mat4x3;
+
+	/// 4 columns of 3 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 3, float, mediump>	mediump_mat4x3;
+
+	/// 4 columns of 3 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 3, float, highp>		highp_mat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4.hpp
new file mode 100644
index 000000000000..5ba111de0481
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4.hpp
@@ -0,0 +1,23 @@
+/// @ref core
+/// @file glm/ext/matrix_float4x4.hpp
+
+#pragma once
+#include "../detail/type_mat4x4.hpp"
+
+namespace glm
+{
+	/// @ingroup core_matrix
+	/// @{
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 4, float, defaultp>			mat4x4;
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	typedef mat<4, 4, float, defaultp>			mat4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4_precision.hpp
new file mode 100644
index 000000000000..597149bcf90f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_float4x4_precision.hpp
@@ -0,0 +1,49 @@
+/// @ref core
+/// @file glm/ext/matrix_float4x4_precision.hpp
+
+#pragma once
+#include "../detail/type_mat4x4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_matrix_precision
+	/// @{
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, float, lowp>		lowp_mat4;
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, float, mediump>	mediump_mat4;
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, float, highp>		highp_mat4;
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, float, lowp>		lowp_mat4x4;
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, float, mediump>	mediump_mat4x4;
+
+	/// 4 columns of 4 components matrix of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.6 Matrices</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef mat<4, 4, float, highp>		highp_mat4x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2.hpp
new file mode 100644
index 000000000000..c6aa0686ae78
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2.hpp
@@ -0,0 +1,38 @@
+/// @ref ext_matrix_int2x2
+/// @file glm/ext/matrix_int2x2.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int2x2 GLM_EXT_matrix_int2x2
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int2x2.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int2x2 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int2x2
+	/// @{
+
+	/// Signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2
+	typedef mat<2, 2, int, defaultp>	imat2x2;
+
+	/// Signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2
+	typedef mat<2, 2, int, defaultp>	imat2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2_sized.hpp
new file mode 100644
index 000000000000..70c0c2106acd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x2_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_matrix_int2x2_sized
+/// @file glm/ext/matrix_int2x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int2x2_sized GLM_EXT_matrix_int2x2_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int2x2_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x2.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int2x2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int2x2_sized
+	/// @{
+
+	/// 8 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int8, defaultp>				i8mat2x2;
+
+	/// 16 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int16, defaultp>				i16mat2x2;
+
+	/// 32 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int32, defaultp>				i32mat2x2;
+
+	/// 64 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int64, defaultp>				i64mat2x2;
+
+
+	/// 8 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int8, defaultp>				i8mat2;
+
+	/// 16 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int16, defaultp>				i16mat2;
+
+	/// 32 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int32, defaultp>				i32mat2;
+
+	/// 64 bit signed integer 2x2 matrix.
+	///
+	/// @see ext_matrix_int2x2_sized
+	typedef mat<2, 2, int64, defaultp>				i64mat2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3.hpp
new file mode 100644
index 000000000000..aee415caa6e1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_int2x3
+/// @file glm/ext/matrix_int2x3.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int2x3 GLM_EXT_matrix_int2x3
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int2x3.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x3.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int2x3 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int2x3
+	/// @{
+
+	/// Signed integer 2x3 matrix.
+	///
+	/// @see ext_matrix_int2x3
+	typedef mat<2, 3, int, defaultp>	imat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3_sized.hpp
new file mode 100644
index 000000000000..b5526fe55143
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x3_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_int2x3_sized
+/// @file glm/ext/matrix_int2x3_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int2x3_sized GLM_EXT_matrix_int2x3_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int2x3_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x3.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int2x3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int2x3_sized
+	/// @{
+
+	/// 8 bit signed integer 2x3 matrix.
+	///
+	/// @see ext_matrix_int2x3_sized
+	typedef mat<2, 3, int8, defaultp>				i8mat2x3;
+
+	/// 16 bit signed integer 2x3 matrix.
+	///
+	/// @see ext_matrix_int2x3_sized
+	typedef mat<2, 3, int16, defaultp>				i16mat2x3;
+
+	/// 32 bit signed integer 2x3 matrix.
+	///
+	/// @see ext_matrix_int2x3_sized
+	typedef mat<2, 3, int32, defaultp>				i32mat2x3;
+
+	/// 64 bit signed integer 2x3 matrix.
+	///
+	/// @see ext_matrix_int2x3_sized
+	typedef mat<2, 3, int64, defaultp>				i64mat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4.hpp
new file mode 100644
index 000000000000..4f36331d6602
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_int2x4
+/// @file glm/ext/matrix_int2x4.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int2x4 GLM_EXT_matrix_int2x4
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int2x4.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int2x4 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int2x4
+	/// @{
+
+	/// Signed integer 2x4 matrix.
+	///
+	/// @see ext_matrix_int2x4
+	typedef mat<2, 4, int, defaultp>	imat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4_sized.hpp
new file mode 100644
index 000000000000..a66a5e726881
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int2x4_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_int2x4_sized
+/// @file glm/ext/matrix_int2x4_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int2x4_sized GLM_EXT_matrix_int2x4_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int2x4_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x4.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int2x4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int2x4_sized
+	/// @{
+
+	/// 8 bit signed integer 2x4 matrix.
+	///
+	/// @see ext_matrix_int2x4_sized
+	typedef mat<2, 4, int8, defaultp>				i8mat2x4;
+
+	/// 16 bit signed integer 2x4 matrix.
+	///
+	/// @see ext_matrix_int2x4_sized
+	typedef mat<2, 4, int16, defaultp>				i16mat2x4;
+
+	/// 32 bit signed integer 2x4 matrix.
+	///
+	/// @see ext_matrix_int2x4_sized
+	typedef mat<2, 4, int32, defaultp>				i32mat2x4;
+
+	/// 64 bit signed integer 2x4 matrix.
+	///
+	/// @see ext_matrix_int2x4_sized
+	typedef mat<2, 4, int64, defaultp>				i64mat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2.hpp
new file mode 100644
index 000000000000..3bd563b7de9c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_int3x2
+/// @file glm/ext/matrix_int3x2.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int3x2 GLM_EXT_matrix_int3x2
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int3x2.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int3x2 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int3x2
+	/// @{
+
+	/// Signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_int3x2
+	typedef mat<3, 2, int, defaultp>	imat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2_sized.hpp
new file mode 100644
index 000000000000..7e34c5240f40
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x2_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_int3x2_sized
+/// @file glm/ext/matrix_int3x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int3x2_sized GLM_EXT_matrix_int3x2_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int3x2_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x2.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int3x2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int3x2_sized
+	/// @{
+
+	/// 8 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_int3x2_sized
+	typedef mat<3, 2, int8, defaultp>				i8mat3x2;
+
+	/// 16 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_int3x2_sized
+	typedef mat<3, 2, int16, defaultp>				i16mat3x2;
+
+	/// 32 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_int3x2_sized
+	typedef mat<3, 2, int32, defaultp>				i32mat3x2;
+
+	/// 64 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_int3x2_sized
+	typedef mat<3, 2, int64, defaultp>				i64mat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3.hpp
new file mode 100644
index 000000000000..287488da0343
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3.hpp
@@ -0,0 +1,38 @@
+/// @ref ext_matrix_int3x3
+/// @file glm/ext/matrix_int3x3.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int3x3 GLM_EXT_matrix_int3x3
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int3x3.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x3.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int3x3 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int3x3
+	/// @{
+
+	/// Signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3
+	typedef mat<3, 3, int, defaultp>	imat3x3;
+
+	/// Signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3
+	typedef mat<3, 3, int, defaultp>	imat3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3_sized.hpp
new file mode 100644
index 000000000000..577e305aa7f7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x3_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_matrix_int3x3_sized
+/// @file glm/ext/matrix_int3x3_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int3x3_sized GLM_EXT_matrix_int3x3_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int3x3_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x3.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int3x3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int3x3_sized
+	/// @{
+
+	/// 8 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int8, defaultp>				i8mat3x3;
+
+	/// 16 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int16, defaultp>				i16mat3x3;
+
+	/// 32 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int32, defaultp>				i32mat3x3;
+
+	/// 64 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int64, defaultp>				i64mat3x3;
+
+
+	/// 8 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int8, defaultp>				i8mat3;
+
+	/// 16 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int16, defaultp>				i16mat3;
+
+	/// 32 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int32, defaultp>				i32mat3;
+
+	/// 64 bit signed integer 3x3 matrix.
+	///
+	/// @see ext_matrix_int3x3_sized
+	typedef mat<3, 3, int64, defaultp>				i64mat3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4.hpp
new file mode 100644
index 000000000000..08e534d9c4d5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_int3x4
+/// @file glm/ext/matrix_int3x4.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int3x4 GLM_EXT_matrix_int3x4
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int3x4.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int3x4 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int3x4
+	/// @{
+
+	/// Signed integer 3x4 matrix.
+	///
+	/// @see ext_matrix_int3x4
+	typedef mat<3, 4, int, defaultp>	imat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4_sized.hpp
new file mode 100644
index 000000000000..692c48c439e7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int3x4_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_int3x4_sized
+/// @file glm/ext/matrix_int3x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int3x4_sized GLM_EXT_matrix_int3x4_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int3x4_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x4.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int3x4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int3x4_sized
+	/// @{
+
+	/// 8 bit signed integer 3x4 matrix.
+	///
+	/// @see ext_matrix_int3x4_sized
+	typedef mat<3, 4, int8, defaultp>				i8mat3x4;
+
+	/// 16 bit signed integer 3x4 matrix.
+	///
+	/// @see ext_matrix_int3x4_sized
+	typedef mat<3, 4, int16, defaultp>				i16mat3x4;
+
+	/// 32 bit signed integer 3x4 matrix.
+	///
+	/// @see ext_matrix_int3x4_sized
+	typedef mat<3, 4, int32, defaultp>				i32mat3x4;
+
+	/// 64 bit signed integer 3x4 matrix.
+	///
+	/// @see ext_matrix_int3x4_sized
+	typedef mat<3, 4, int64, defaultp>				i64mat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2.hpp
new file mode 100644
index 000000000000..f756ef2804c8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_int4x2
+/// @file glm/ext/matrix_int4x2.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int4x2 GLM_EXT_matrix_int4x2
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int4x2.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int4x2 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int4x2
+	/// @{
+
+	/// Signed integer 4x2 matrix.
+	///
+	/// @see ext_matrix_int4x2
+	typedef mat<4, 2, int, defaultp>	imat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2_sized.hpp
new file mode 100644
index 000000000000..63a99d604dc5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x2_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_int4x2_sized
+/// @file glm/ext/matrix_int4x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int4x2_sized GLM_EXT_matrix_int4x2_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int4x2_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x2.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int4x2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int4x2_sized
+	/// @{
+
+	/// 8 bit signed integer 4x2 matrix.
+	///
+	/// @see ext_matrix_int4x2_sized
+	typedef mat<4, 2, int8, defaultp>				i8mat4x2;
+
+	/// 16 bit signed integer 4x2 matrix.
+	///
+	/// @see ext_matrix_int4x2_sized
+	typedef mat<4, 2, int16, defaultp>				i16mat4x2;
+
+	/// 32 bit signed integer 4x2 matrix.
+	///
+	/// @see ext_matrix_int4x2_sized
+	typedef mat<4, 2, int32, defaultp>				i32mat4x2;
+
+	/// 64 bit signed integer 4x2 matrix.
+	///
+	/// @see ext_matrix_int4x2_sized
+	typedef mat<4, 2, int64, defaultp>				i64mat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3.hpp
new file mode 100644
index 000000000000..d5d97a7a37cf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_int4x3
+/// @file glm/ext/matrix_int4x3.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int4x3 GLM_EXT_matrix_int4x3
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int4x3.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x3.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int4x3 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int4x3
+	/// @{
+
+	/// Signed integer 4x3 matrix.
+	///
+	/// @see ext_matrix_int4x3
+	typedef mat<4, 3, int, defaultp>	imat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3_sized.hpp
new file mode 100644
index 000000000000..55078fadc60f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x3_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_int4x3_sized
+/// @file glm/ext/matrix_int4x3_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int4x3_sized GLM_EXT_matrix_int4x3_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int4x3_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x3.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int4x3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int4x3_sized
+	/// @{
+
+	/// 8 bit signed integer 4x3 matrix.
+	///
+	/// @see ext_matrix_int4x3_sized
+	typedef mat<4, 3, int8, defaultp>				i8mat4x3;
+
+	/// 16 bit signed integer 4x3 matrix.
+	///
+	/// @see ext_matrix_int4x3_sized
+	typedef mat<4, 3, int16, defaultp>				i16mat4x3;
+
+	/// 32 bit signed integer 4x3 matrix.
+	///
+	/// @see ext_matrix_int4x3_sized
+	typedef mat<4, 3, int32, defaultp>				i32mat4x3;
+
+	/// 64 bit signed integer 4x3 matrix.
+	///
+	/// @see ext_matrix_int4x3_sized
+	typedef mat<4, 3, int64, defaultp>				i64mat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4.hpp
new file mode 100644
index 000000000000..e17cff17f9fb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4.hpp
@@ -0,0 +1,38 @@
+/// @ref ext_matrix_int4x4
+/// @file glm/ext/matrix_int4x4.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int4x4 GLM_EXT_matrix_int4x4
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int4x4.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int4x4 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int4x4
+	/// @{
+
+	/// Signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4
+	typedef mat<4, 4, int, defaultp>	imat4x4;
+
+	/// Signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4
+	typedef mat<4, 4, int, defaultp>	imat4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4_sized.hpp
new file mode 100644
index 000000000000..4a11203eb25b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_int4x4_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_matrix_int4x4_sized
+/// @file glm/ext/matrix_int4x4_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_int4x4_sized GLM_EXT_matrix_int4x4_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_int4x4_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x4.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_int4x4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_int4x4_sized
+	/// @{
+
+	/// 8 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int8, defaultp>				i8mat4x4;
+
+	/// 16 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int16, defaultp>				i16mat4x4;
+
+	/// 32 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int32, defaultp>				i32mat4x4;
+
+	/// 64 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int64, defaultp>				i64mat4x4;
+
+
+	/// 8 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int8, defaultp>				i8mat4;
+
+	/// 16 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int16, defaultp>				i16mat4;
+
+	/// 32 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int32, defaultp>				i32mat4;
+
+	/// 64 bit signed integer 4x4 matrix.
+	///
+	/// @see ext_matrix_int4x4_sized
+	typedef mat<4, 4, int64, defaultp>				i64mat4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.hpp
new file mode 100644
index 000000000000..7d7dfc5a5b08
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.hpp
@@ -0,0 +1,91 @@
+/// @ref ext_matrix_integer
+/// @file glm/ext/matrix_integer.hpp
+///
+/// @defgroup ext_matrix_integer GLM_EXT_matrix_integer
+/// @ingroup ext
+///
+/// Defines functions that generate common transformation matrices.
+///
+/// The matrices generated by this extension use standard OpenGL fixed-function
+/// conventions. For example, the lookAt function generates a transform from world
+/// space into the specific eye space that the projective matrix functions
+/// (perspective, ortho, etc) are designed to expect. The OpenGL compatibility
+/// specifications defines the particular layout of this eye space.
+///
+/// Include <glm/ext/matrix_integer.hpp> to use the features of this extension.
+///
+/// @see ext_matrix_projection
+/// @see ext_matrix_clip_space
+
+#pragma once
+
+// Dependencies
+#include "../gtc/constants.hpp"
+#include "../geometric.hpp"
+#include "../trigonometric.hpp"
+#include "../matrix.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_integer extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_integer
+	/// @{
+
+	 /// Multiply matrix x by matrix y component-wise, i.e.,
+	 /// result[i][j] is the scalar product of x[i][j] and y[i][j].
+	 ///
+	 /// @tparam C Integer between 1 and 4 included that qualify the number a column
+	 /// @tparam R Integer between 1 and 4 included that qualify the number a row
+	 /// @tparam T Floating-point or signed integer scalar types
+	 /// @tparam Q Value from qualifier enum
+	 ///
+	 /// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/matrixCompMult.xml">GLSL matrixCompMult man page</a>
+	 /// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> matrixCompMult(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y);
+
+	/// Treats the first parameter c as a column vector
+	/// and the second parameter r as a row vector
+	/// and does a linear algebraic matrix multiply c * r.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point or signed integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/outerProduct.xml">GLSL outerProduct man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL typename detail::outerProduct_trait<C, R, T, Q>::type outerProduct(vec<C, T, Q> const& c, vec<R, T, Q> const& r);
+
+	/// Returns the transposed matrix of x
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point or signed integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/transpose.xml">GLSL transpose man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL typename mat<C, R, T, Q>::transpose_type transpose(mat<C, R, T, Q> const& x);
+
+	/// Return the determinant of a squared matrix.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point or signed integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/determinant.xml">GLSL determinant man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL T determinant(mat<C, R, T, Q> const& m);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_integer.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.inl
new file mode 100644
index 000000000000..8b377ce2a8e6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_integer.inl
@@ -0,0 +1,38 @@
+namespace glm{
+namespace detail
+{
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_matrixCompMult_type<C, R, T, Q, false, Aligned> {
+		GLM_FUNC_QUALIFIER static mat<C, R, T, Q> call(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y)
+		{
+			return detail::compute_matrixCompMult<C, R, T, Q, detail::is_aligned<Q>::value>::call(x, y);
+		}
+	};
+
+	template<length_t DA, length_t DB, typename T, qualifier Q>
+	struct compute_outerProduct_type<DA, DB, T, Q, false> {
+		GLM_FUNC_QUALIFIER static typename detail::outerProduct_trait<DA, DB, T, Q>::type call(vec<DA, T, Q> const& c, vec<DB, T, Q> const& r)
+		{
+			return detail::compute_outerProduct<DA, DB, T, Q>::call(c, r);
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_transpose_type<C, R, T, Q, false, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static mat<R, C, T, Q> call(mat<C, R, T, Q> const& m)
+		{
+			return detail::compute_transpose<C, R, T, Q, detail::is_aligned<Q>::value>::call(m);
+		}
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q, bool Aligned>
+	struct compute_determinant_type<C, R, T, Q, false, Aligned>{
+	
+		GLM_FUNC_QUALIFIER static T call(mat<C, R, T, Q> const& m)
+		{
+			return detail::compute_determinant<C, R, T, Q, detail::is_aligned<Q>::value>::call(m);
+		}
+	};
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.hpp
new file mode 100644
index 000000000000..51fd01bd8ee7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.hpp
@@ -0,0 +1,149 @@
+/// @ref ext_matrix_projection
+/// @file glm/ext/matrix_projection.hpp
+///
+/// @defgroup ext_matrix_projection GLM_EXT_matrix_projection
+/// @ingroup ext
+///
+/// Functions that generate common projection transformation matrices.
+///
+/// The matrices generated by this extension use standard OpenGL fixed-function
+/// conventions. For example, the lookAt function generates a transform from world
+/// space into the specific eye space that the projective matrix functions
+/// (perspective, ortho, etc) are designed to expect. The OpenGL compatibility
+/// specifications defines the particular layout of this eye space.
+///
+/// Include <glm/ext/matrix_projection.hpp> to use the features of this extension.
+///
+/// @see ext_matrix_transform
+/// @see ext_matrix_clip_space
+
+#pragma once
+
+// Dependencies
+#include "../gtc/constants.hpp"
+#include "../geometric.hpp"
+#include "../trigonometric.hpp"
+#include "../matrix.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_projection extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_projection
+	/// @{
+
+	/// Map the specified object coordinates (obj.x, obj.y, obj.z) into window coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param obj Specify the object coordinates.
+	/// @param model Specifies the current modelview matrix
+	/// @param proj Specifies the current projection matrix
+	/// @param viewport Specifies the current viewport
+	/// @return Return the computed window coordinates.
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluProject.xml">gluProject man page</a>
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> projectZO(
+		vec<3, T, Q> const& obj, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport);
+
+	/// Map the specified object coordinates (obj.x, obj.y, obj.z) into window coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param obj Specify the object coordinates.
+	/// @param model Specifies the current modelview matrix
+	/// @param proj Specifies the current projection matrix
+	/// @param viewport Specifies the current viewport
+	/// @return Return the computed window coordinates.
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluProject.xml">gluProject man page</a>
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> projectNO(
+		vec<3, T, Q> const& obj, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport);
+
+	/// Map the specified object coordinates (obj.x, obj.y, obj.z) into window coordinates using default near and far clip planes definition.
+	/// To change default near and far clip planes definition use GLM_FORCE_DEPTH_ZERO_TO_ONE.
+	///
+	/// @param obj Specify the object coordinates.
+	/// @param model Specifies the current modelview matrix
+	/// @param proj Specifies the current projection matrix
+	/// @param viewport Specifies the current viewport
+	/// @return Return the computed window coordinates.
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluProject.xml">gluProject man page</a>
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> project(
+		vec<3, T, Q> const& obj, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport);
+
+	/// Map the specified window coordinates (win.x, win.y, win.z) into object coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of 0 and +1 respectively. (Direct3D clip volume definition)
+	///
+	/// @param win Specify the window coordinates to be mapped.
+	/// @param model Specifies the modelview matrix
+	/// @param proj Specifies the projection matrix
+	/// @param viewport Specifies the viewport
+	/// @return Returns the computed object coordinates.
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluUnProject.xml">gluUnProject man page</a>
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> unProjectZO(
+		vec<3, T, Q> const& win, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport);
+
+	/// Map the specified window coordinates (win.x, win.y, win.z) into object coordinates.
+	/// The near and far clip planes correspond to z normalized device coordinates of -1 and +1 respectively. (OpenGL clip volume definition)
+	///
+	/// @param win Specify the window coordinates to be mapped.
+	/// @param model Specifies the modelview matrix
+	/// @param proj Specifies the projection matrix
+	/// @param viewport Specifies the viewport
+	/// @return Returns the computed object coordinates.
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluUnProject.xml">gluUnProject man page</a>
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> unProjectNO(
+		vec<3, T, Q> const& win, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport);
+
+	/// Map the specified window coordinates (win.x, win.y, win.z) into object coordinates using default near and far clip planes definition.
+	/// To change default near and far clip planes definition use GLM_FORCE_DEPTH_ZERO_TO_ONE.
+	///
+	/// @param win Specify the window coordinates to be mapped.
+	/// @param model Specifies the modelview matrix
+	/// @param proj Specifies the projection matrix
+	/// @param viewport Specifies the viewport
+	/// @return Returns the computed object coordinates.
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluUnProject.xml">gluUnProject man page</a>
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> unProject(
+		vec<3, T, Q> const& win, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport);
+
+	/// Define a picking region
+	///
+	/// @param center Specify the center of a picking region in window coordinates.
+	/// @param delta Specify the width and height, respectively, of the picking region in window coordinates.
+	/// @param viewport Rendering viewport
+	/// @tparam T Native type used for the computation. Currently supported: half (not recommended), float or double.
+	/// @tparam U Currently supported: Floating-point types and integer types.
+	///
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluPickMatrix.xml">gluPickMatrix man page</a>
+	template<typename T, qualifier Q, typename U>
+	GLM_FUNC_DECL mat<4, 4, T, Q> pickMatrix(
+		vec<2, T, Q> const& center, vec<2, T, Q> const& delta, vec<4, U, Q> const& viewport);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_projection.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.inl
new file mode 100644
index 000000000000..2f2c196aac5f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_projection.inl
@@ -0,0 +1,106 @@
+namespace glm
+{
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> projectZO(vec<3, T, Q> const& obj, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport)
+	{
+		vec<4, T, Q> tmp = vec<4, T, Q>(obj, static_cast<T>(1));
+		tmp = model * tmp;
+		tmp = proj * tmp;
+
+		tmp /= tmp.w;
+		tmp.x = tmp.x * static_cast<T>(0.5) + static_cast<T>(0.5);
+		tmp.y = tmp.y * static_cast<T>(0.5) + static_cast<T>(0.5);
+
+		tmp[0] = tmp[0] * T(viewport[2]) + T(viewport[0]);
+		tmp[1] = tmp[1] * T(viewport[3]) + T(viewport[1]);
+
+		return vec<3, T, Q>(tmp);
+	}
+
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> projectNO(vec<3, T, Q> const& obj, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport)
+	{
+		vec<4, T, Q> tmp = vec<4, T, Q>(obj, static_cast<T>(1));
+		tmp = model * tmp;
+		tmp = proj * tmp;
+
+		tmp /= tmp.w;
+		tmp = tmp * static_cast<T>(0.5) + static_cast<T>(0.5);
+		tmp[0] = tmp[0] * T(viewport[2]) + T(viewport[0]);
+		tmp[1] = tmp[1] * T(viewport[3]) + T(viewport[1]);
+
+		return vec<3, T, Q>(tmp);
+	}
+
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> project(vec<3, T, Q> const& obj, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return projectZO(obj, model, proj, viewport);
+#		else
+			return projectNO(obj, model, proj, viewport);
+#		endif
+	}
+
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> unProjectZO(vec<3, T, Q> const& win, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport)
+	{
+		mat<4, 4, T, Q> Inverse = inverse(proj * model);
+
+		vec<4, T, Q> tmp = vec<4, T, Q>(win, T(1));
+		tmp.x = (tmp.x - T(viewport[0])) / T(viewport[2]);
+		tmp.y = (tmp.y - T(viewport[1])) / T(viewport[3]);
+		tmp.x = tmp.x * static_cast<T>(2) - static_cast<T>(1);
+		tmp.y = tmp.y * static_cast<T>(2) - static_cast<T>(1);
+
+		vec<4, T, Q> obj = Inverse * tmp;
+		obj /= obj.w;
+
+		return vec<3, T, Q>(obj);
+	}
+
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> unProjectNO(vec<3, T, Q> const& win, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport)
+	{
+		mat<4, 4, T, Q> Inverse = inverse(proj * model);
+
+		vec<4, T, Q> tmp = vec<4, T, Q>(win, T(1));
+		tmp.x = (tmp.x - T(viewport[0])) / T(viewport[2]);
+		tmp.y = (tmp.y - T(viewport[1])) / T(viewport[3]);
+		tmp = tmp * static_cast<T>(2) - static_cast<T>(1);
+
+		vec<4, T, Q> obj = Inverse * tmp;
+		obj /= obj.w;
+
+		return vec<3, T, Q>(obj);
+	}
+
+	template<typename T, typename U, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> unProject(vec<3, T, Q> const& win, mat<4, 4, T, Q> const& model, mat<4, 4, T, Q> const& proj, vec<4, U, Q> const& viewport)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_ZO_BIT
+			return unProjectZO(win, model, proj, viewport);
+#		else
+			return unProjectNO(win, model, proj, viewport);
+#		endif
+	}
+
+	template<typename T, qualifier Q, typename U>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> pickMatrix(vec<2, T, Q> const& center, vec<2, T, Q> const& delta, vec<4, U, Q> const& viewport)
+	{
+		assert(delta.x > static_cast<T>(0) && delta.y > static_cast<T>(0));
+		mat<4, 4, T, Q> Result(static_cast<T>(1));
+
+		if(!(delta.x > static_cast<T>(0) && delta.y > static_cast<T>(0)))
+			return Result; // Error
+
+		vec<3, T, Q> Temp(
+			(static_cast<T>(viewport[2]) - static_cast<T>(2) * (center.x - static_cast<T>(viewport[0]))) / delta.x,
+			(static_cast<T>(viewport[3]) - static_cast<T>(2) * (center.y - static_cast<T>(viewport[1]))) / delta.y,
+			static_cast<T>(0));
+
+		// Translate and scale the picked region to the entire window
+		Result = translate(Result, Temp);
+		return scale(Result, vec<3, T, Q>(static_cast<T>(viewport[2]) / delta.x, static_cast<T>(viewport[3]) / delta.y, static_cast<T>(1)));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.hpp
new file mode 100644
index 000000000000..20023ad89a0c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.hpp
@@ -0,0 +1,132 @@
+/// @ref ext_matrix_relational
+/// @file glm/ext/matrix_relational.hpp
+///
+/// @defgroup ext_matrix_relational GLM_EXT_matrix_relational
+/// @ingroup ext
+///
+/// Exposes comparison functions for matrix types that take a user defined epsilon values.
+///
+/// Include <glm/ext/matrix_relational.hpp> to use the features of this extension.
+///
+/// @see ext_vector_relational
+/// @see ext_scalar_relational
+/// @see ext_quaternion_relational
+
+#pragma once
+
+// Dependencies
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_relational extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_relational
+	/// @{
+
+	/// Perform a component-wise equal-to comparison of two matrices.
+	/// Return a boolean vector which components value is True if this expression is satisfied per column of the matrices.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y);
+
+	/// Perform a component-wise not-equal-to comparison of two matrices.
+	/// Return a boolean vector which components value is True if this expression is satisfied per column of the matrices.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, T epsilon);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, vec<C, T, Q> const& epsilon);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, T epsilon);
+
+	/// Returns the component-wise comparison of |x - y| >= epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, vec<C, T, Q> const& epsilon);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, int ULPs);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, vec<C, int, Q> const& ULPs);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, int ULPs);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number of columns of the matrix
+	/// @tparam R Integer between 1 and 4 included that qualify the number of rows of the matrix
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y, vec<C, int, Q> const& ULPs);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_relational.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.inl
new file mode 100644
index 000000000000..9cd42b772b32
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_relational.inl
@@ -0,0 +1,88 @@
+/// @ref ext_vector_relational
+/// @file glm/ext/vector_relational.inl
+
+// Dependency:
+#include "../ext/vector_relational.hpp"
+#include "../common.hpp"
+
+namespace glm
+{
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b)
+	{
+		vec<C, bool, Q> Result(true);
+		for(length_t i = 0; i < C; ++i)
+			Result[i] = all(equal(a[i], b[i]));
+		return Result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, T Epsilon)
+	{
+		return equal(a, b, vec<C, T, Q>(Epsilon));
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, vec<C, T, Q> const& Epsilon)
+	{
+		vec<C, bool, Q> Result(true);
+		for(length_t i = 0; i < C; ++i)
+			Result[i] = all(equal(a[i], b[i], Epsilon[i]));
+		return Result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b)
+	{
+		vec<C, bool, Q> Result(true);
+		for(length_t i = 0; i < C; ++i)
+			Result[i] = any(notEqual(a[i], b[i]));
+		return Result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, T Epsilon)
+	{
+		return notEqual(a, b, vec<C, T, Q>(Epsilon));
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, vec<C, T, Q> const& Epsilon)
+	{
+		vec<C, bool, Q> Result(true);
+		for(length_t i = 0; i < C; ++i)
+			Result[i] = any(notEqual(a[i], b[i], Epsilon[i]));
+		return Result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, int MaxULPs)
+	{
+		return equal(a, b, vec<C, int, Q>(MaxULPs));
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> equal(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, vec<C, int, Q> const& MaxULPs)
+	{
+		vec<C, bool, Q> Result(true);
+		for(length_t i = 0; i < C; ++i)
+			Result[i] = all(equal(a[i], b[i], MaxULPs[i]));
+		return Result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, int MaxULPs)
+	{
+		return notEqual(a, b, vec<C, int, Q>(MaxULPs));
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<C, bool, Q> notEqual(mat<C, R, T, Q> const& a, mat<C, R, T, Q> const& b, vec<C, int, Q> const& MaxULPs)
+	{
+		vec<C, bool, Q> Result(true);
+		for(length_t i = 0; i < C; ++i)
+			Result[i] = any(notEqual(a[i], b[i], MaxULPs[i]));
+		return Result;
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.hpp
new file mode 100644
index 000000000000..52695b8bf49d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.hpp
@@ -0,0 +1,171 @@
+/// @ref ext_matrix_transform
+/// @file glm/ext/matrix_transform.hpp
+///
+/// @defgroup ext_matrix_transform GLM_EXT_matrix_transform
+/// @ingroup ext
+///
+/// Defines functions that generate common transformation matrices.
+///
+/// The matrices generated by this extension use standard OpenGL fixed-function
+/// conventions. For example, the lookAt function generates a transform from world
+/// space into the specific eye space that the projective matrix functions
+/// (perspective, ortho, etc) are designed to expect. The OpenGL compatibility
+/// specifications defines the particular layout of this eye space.
+///
+/// Include <glm/ext/matrix_transform.hpp> to use the features of this extension.
+///
+/// @see ext_matrix_projection
+/// @see ext_matrix_clip_space
+
+#pragma once
+
+// Dependencies
+#include "../gtc/constants.hpp"
+#include "../geometric.hpp"
+#include "../trigonometric.hpp"
+#include "../matrix.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_transform extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_transform
+	/// @{
+
+	/// Builds an identity matrix.
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType identity();
+
+	/// Builds a translation 4 * 4 matrix created from a vector of 3 components.
+	///
+	/// @param m Input matrix multiplied by this translation matrix.
+	/// @param v Coordinates of a translation vector.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @code
+	/// #include <glm/glm.hpp>
+	/// #include <glm/gtc/matrix_transform.hpp>
+	/// ...
+	/// glm::mat4 m = glm::translate(glm::mat4(1.0f), glm::vec3(1.0f));
+	/// // m[0][0] == 1.0f, m[0][1] == 0.0f, m[0][2] == 0.0f, m[0][3] == 0.0f
+	/// // m[1][0] == 0.0f, m[1][1] == 1.0f, m[1][2] == 0.0f, m[1][3] == 0.0f
+	/// // m[2][0] == 0.0f, m[2][1] == 0.0f, m[2][2] == 1.0f, m[2][3] == 0.0f
+	/// // m[3][0] == 1.0f, m[3][1] == 1.0f, m[3][2] == 1.0f, m[3][3] == 1.0f
+	/// @endcode
+	///
+	/// @see - translate(mat<4, 4, T, Q> const& m, T x, T y, T z)
+	/// @see - translate(vec<3, T, Q> const& v)
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glTranslate.xml">glTranslate man page</a>
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR mat<4, 4, T, Q> translate(
+		mat<4, 4, T, Q> const& m, vec<3, T, Q> const& v);
+
+	/// Builds a rotation 4 * 4 matrix created from an axis vector and an angle.
+	///
+	/// @param m Input matrix multiplied by this rotation matrix.
+	/// @param angle Rotation angle expressed in radians.
+	/// @param axis Rotation axis, recommended to be normalized.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @see - rotate(mat<4, 4, T, Q> const& m, T angle, T x, T y, T z)
+	/// @see - rotate(T angle, vec<3, T, Q> const& v)
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glRotate.xml">glRotate man page</a>
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> rotate(
+		mat<4, 4, T, Q> const& m, T angle, vec<3, T, Q> const& axis);
+
+	/// Builds a scale 4 * 4 matrix created from 3 scalars.
+	///
+	/// @param m Input matrix multiplied by this scale matrix.
+	/// @param v Ratio of scaling for each axis.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @see - scale(mat<4, 4, T, Q> const& m, T x, T y, T z)
+	/// @see - scale(vec<3, T, Q> const& v)
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glScale.xml">glScale man page</a>
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> scale(
+		mat<4, 4, T, Q> const& m, vec<3, T, Q> const& v);
+
+    /// Builds a scale 4 * 4 matrix created from point referent 3 shearers.
+    ///
+    /// @param m Input matrix multiplied by this shear matrix.
+    /// @param p Point of shearing as reference.
+    /// @param l_x Ratio of matrix.x projection in YZ plane relative to the y-axis/z-axis.
+    /// @param l_y Ratio of matrix.y projection in XZ plane relative to the x-axis/z-axis.
+    /// @param l_z Ratio of matrix.z projection in XY plane relative to the x-axis/y-axis.
+    ///
+    /// as example:
+    ///                                         [1   , l_xy, l_xz, -(l_xy+l_xz) * p_x] [x]  T
+    ///   [x`, y`, z`, w`] = [x`, y`, z`, w`] * [l_yx, 1   , l_yz, -(l_yx+l_yz) * p_y] [y]
+    ///                                         [l_zx, l_zy, 1   , -(l_zx+l_zy) * p_z] [z]
+    ///                                         [0   , 0   , 0   , 1                 ] [w]
+    ///
+    /// @tparam T A floating-point shear type
+    /// @tparam Q A value from qualifier enum
+    ///
+    /// @see - shear(mat<4, 4, T, Q> const& m, T x, T y, T z)
+    /// @see - shear(vec<3, T, Q> const& p)
+    /// @see - shear(vec<2, T, Q> const& l_x)
+    /// @see - shear(vec<2, T, Q> const& l_y)
+    /// @see - shear(vec<2, T, Q> const& l_z)
+    /// @see no resource...
+    template <typename T, qualifier Q>
+    GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shear(
+        mat<4, 4, T, Q> const &m, vec<3, T, Q> const& p, vec<2, T, Q> const &l_x, vec<2, T, Q> const &l_y, vec<2, T, Q> const &l_z);
+
+    /// Build a right handed look at view matrix.
+	///
+	/// @param eye Position of the camera
+	/// @param center Position where the camera is looking at
+	/// @param up Normalized up vector, how the camera is oriented. Typically (0, 0, 1)
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @see - frustum(T const& left, T const& right, T const& bottom, T const& top, T const& nearVal, T const& farVal) frustum(T const& left, T const& right, T const& bottom, T const& top, T const& nearVal, T const& farVal)
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> lookAtRH(
+		vec<3, T, Q> const& eye, vec<3, T, Q> const& center, vec<3, T, Q> const& up);
+
+	/// Build a left handed look at view matrix.
+	///
+	/// @param eye Position of the camera
+	/// @param center Position where the camera is looking at
+	/// @param up Normalized up vector, how the camera is oriented. Typically (0, 0, 1)
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @see - frustum(T const& left, T const& right, T const& bottom, T const& top, T const& nearVal, T const& farVal) frustum(T const& left, T const& right, T const& bottom, T const& top, T const& nearVal, T const& farVal)
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> lookAtLH(
+		vec<3, T, Q> const& eye, vec<3, T, Q> const& center, vec<3, T, Q> const& up);
+
+	/// Build a look at view matrix based on the default handedness.
+	///
+	/// @param eye Position of the camera
+	/// @param center Position where the camera is looking at
+	/// @param up Normalized up vector, how the camera is oriented. Typically (0, 0, 1)
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @see - frustum(T const& left, T const& right, T const& bottom, T const& top, T const& nearVal, T const& farVal) frustum(T const& left, T const& right, T const& bottom, T const& top, T const& nearVal, T const& farVal)
+	/// @see <a href="https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluLookAt.xml">gluLookAt man page</a>
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> lookAt(
+		vec<3, T, Q> const& eye, vec<3, T, Q> const& center, vec<3, T, Q> const& up);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_transform.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.inl
new file mode 100644
index 000000000000..40459bbb71bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_transform.inl
@@ -0,0 +1,207 @@
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType identity()
+	{
+		return detail::init_gentype<genType, detail::genTypeTrait<genType>::GENTYPE>::identity();
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR mat<4, 4, T, Q> translate(mat<4, 4, T, Q> const& m, vec<3, T, Q> const& v)
+	{
+		mat<4, 4, T, Q> Result(m);
+		Result[3] = m[0] * v[0] + m[1] * v[1] + m[2] * v[2] + m[3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> rotate(mat<4, 4, T, Q> const& m, T angle, vec<3, T, Q> const& v)
+	{
+		T const a = angle;
+		T const c = cos(a);
+		T const s = sin(a);
+
+		vec<3, T, Q> axis(normalize(v));
+		vec<3, T, Q> temp((T(1) - c) * axis);
+
+		mat<4, 4, T, Q> Rotate;
+		Rotate[0][0] = c + temp[0] * axis[0];
+		Rotate[0][1] = temp[0] * axis[1] + s * axis[2];
+		Rotate[0][2] = temp[0] * axis[2] - s * axis[1];
+
+		Rotate[1][0] = temp[1] * axis[0] - s * axis[2];
+		Rotate[1][1] = c + temp[1] * axis[1];
+		Rotate[1][2] = temp[1] * axis[2] + s * axis[0];
+
+		Rotate[2][0] = temp[2] * axis[0] + s * axis[1];
+		Rotate[2][1] = temp[2] * axis[1] - s * axis[0];
+		Rotate[2][2] = c + temp[2] * axis[2];
+
+		mat<4, 4, T, Q> Result;
+		Result[0] = m[0] * Rotate[0][0] + m[1] * Rotate[0][1] + m[2] * Rotate[0][2];
+		Result[1] = m[0] * Rotate[1][0] + m[1] * Rotate[1][1] + m[2] * Rotate[1][2];
+		Result[2] = m[0] * Rotate[2][0] + m[1] * Rotate[2][1] + m[2] * Rotate[2][2];
+		Result[3] = m[3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> rotate_slow(mat<4, 4, T, Q> const& m, T angle, vec<3, T, Q> const& v)
+	{
+		T const a = angle;
+		T const c = cos(a);
+		T const s = sin(a);
+		mat<4, 4, T, Q> Result;
+
+		vec<3, T, Q> axis = normalize(v);
+
+		Result[0][0] = c + (static_cast<T>(1) - c)      * axis.x     * axis.x;
+		Result[0][1] = (static_cast<T>(1) - c) * axis.x * axis.y + s * axis.z;
+		Result[0][2] = (static_cast<T>(1) - c) * axis.x * axis.z - s * axis.y;
+		Result[0][3] = static_cast<T>(0);
+
+		Result[1][0] = (static_cast<T>(1) - c) * axis.y * axis.x - s * axis.z;
+		Result[1][1] = c + (static_cast<T>(1) - c) * axis.y * axis.y;
+		Result[1][2] = (static_cast<T>(1) - c) * axis.y * axis.z + s * axis.x;
+		Result[1][3] = static_cast<T>(0);
+
+		Result[2][0] = (static_cast<T>(1) - c) * axis.z * axis.x + s * axis.y;
+		Result[2][1] = (static_cast<T>(1) - c) * axis.z * axis.y - s * axis.x;
+		Result[2][2] = c + (static_cast<T>(1) - c) * axis.z * axis.z;
+		Result[2][3] = static_cast<T>(0);
+
+		Result[3] = vec<4, T, Q>(0, 0, 0, 1);
+		return m * Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> scale(mat<4, 4, T, Q> const& m, vec<3, T, Q> const& v)
+	{
+		mat<4, 4, T, Q> Result;
+		Result[0] = m[0] * v[0];
+		Result[1] = m[1] * v[1];
+		Result[2] = m[2] * v[2];
+		Result[3] = m[3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> scale_slow(mat<4, 4, T, Q> const& m, vec<3, T, Q> const& v)
+	{
+		mat<4, 4, T, Q> Result(T(1));
+		Result[0][0] = v.x;
+		Result[1][1] = v.y;
+		Result[2][2] = v.z;
+		return m * Result;
+	}
+
+    template <typename T, qualifier Q>
+    GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shear(mat<4, 4, T, Q> const &m, vec<3, T, Q> const& p, vec<2, T, Q> const &l_x, vec<2, T, Q> const &l_y, vec<2, T, Q> const &l_z)
+    {
+        T const lambda_xy = l_x[0];
+        T const lambda_xz = l_x[1];
+        T const lambda_yx = l_y[0];
+        T const lambda_yz = l_y[1];
+        T const lambda_zx = l_z[0];
+        T const lambda_zy = l_z[1];
+
+        vec<3, T, Q> point_lambda = vec<3, T, Q>(
+            (lambda_xy + lambda_xz), (lambda_yx + lambda_yz), (lambda_zx + lambda_zy)
+        );
+
+        mat<4, 4, T, Q> Shear = mat<4, 4, T, Q>(
+            1                      , lambda_yx              , lambda_zx              , 0,
+            lambda_xy              , 1                      , lambda_zy              , 0,
+            lambda_xz              , lambda_yz              , 1                      , 0,
+            -point_lambda[0] * p[0], -point_lambda[1] * p[1], -point_lambda[2] * p[2], 1
+        );
+
+        mat<4, 4, T, Q> Result;
+        Result[0] = Shear[0] * m[0][0] + Shear[1] * m[0][1] + Shear[2] * m[0][2] + Shear[3] * m[0][3];
+        Result[1] = Shear[0] * m[1][0] + Shear[1] * m[1][1] + Shear[2] * m[1][2] + Shear[3] * m[1][3];
+        Result[2] = Shear[0] * m[2][0] + Shear[1] * m[2][1] + Shear[2] * m[2][2] + Shear[3] * m[2][3];
+        Result[3] = Shear[0] * m[3][0] + Shear[1] * m[3][1] + Shear[2] * m[3][2] + Shear[3] * m[3][3];
+        return Result;
+    }
+
+    template <typename T, qualifier Q>
+    GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shear_slow(mat<4, 4, T, Q> const &m, vec<3, T, Q> const& p, vec<2, T, Q> const &l_x, vec<2, T, Q> const &l_y, vec<2, T, Q> const &l_z)
+    {
+        T const lambda_xy = static_cast<T>(l_x[0]);
+        T const lambda_xz = static_cast<T>(l_x[1]);
+        T const lambda_yx = static_cast<T>(l_y[0]);
+        T const lambda_yz = static_cast<T>(l_y[1]);
+        T const lambda_zx = static_cast<T>(l_z[0]);
+        T const lambda_zy = static_cast<T>(l_z[1]);
+
+        vec<3, T, Q> point_lambda = vec<3, T, Q>(
+            static_cast<T>(lambda_xy + lambda_xz),
+            static_cast<T>(lambda_yx + lambda_yz),
+            static_cast<T>(lambda_zx + lambda_zy)
+        );
+
+        mat<4, 4, T, Q> Shear = mat<4, 4, T, Q>(
+            1                      , lambda_yx              , lambda_zx              , 0,
+            lambda_xy              , 1                      , lambda_zy              , 0,
+            lambda_xz              , lambda_yz              , 1                      , 0,
+            -point_lambda[0] * p[0], -point_lambda[1] * p[1], -point_lambda[2] * p[2], 1
+        );
+        return m * Shear;
+    }
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> lookAtRH(vec<3, T, Q> const& eye, vec<3, T, Q> const& center, vec<3, T, Q> const& up)
+	{
+		vec<3, T, Q> const f(normalize(center - eye));
+		vec<3, T, Q> const s(normalize(cross(f, up)));
+		vec<3, T, Q> const u(cross(s, f));
+
+		mat<4, 4, T, Q> Result(1);
+		Result[0][0] = s.x;
+		Result[1][0] = s.y;
+		Result[2][0] = s.z;
+		Result[0][1] = u.x;
+		Result[1][1] = u.y;
+		Result[2][1] = u.z;
+		Result[0][2] =-f.x;
+		Result[1][2] =-f.y;
+		Result[2][2] =-f.z;
+		Result[3][0] =-dot(s, eye);
+		Result[3][1] =-dot(u, eye);
+		Result[3][2] = dot(f, eye);
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> lookAtLH(vec<3, T, Q> const& eye, vec<3, T, Q> const& center, vec<3, T, Q> const& up)
+	{
+		vec<3, T, Q> const f(normalize(center - eye));
+		vec<3, T, Q> const s(normalize(cross(up, f)));
+		vec<3, T, Q> const u(cross(f, s));
+
+		mat<4, 4, T, Q> Result(1);
+		Result[0][0] = s.x;
+		Result[1][0] = s.y;
+		Result[2][0] = s.z;
+		Result[0][1] = u.x;
+		Result[1][1] = u.y;
+		Result[2][1] = u.z;
+		Result[0][2] = f.x;
+		Result[1][2] = f.y;
+		Result[2][2] = f.z;
+		Result[3][0] = -dot(s, eye);
+		Result[3][1] = -dot(u, eye);
+		Result[3][2] = -dot(f, eye);
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> lookAt(vec<3, T, Q> const& eye, vec<3, T, Q> const& center, vec<3, T, Q> const& up)
+	{
+#       if (GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT)
+            return lookAtLH(eye, center, up);
+#       else
+            return lookAtRH(eye, center, up);
+#       endif
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2.hpp
new file mode 100644
index 000000000000..034771ae5225
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2.hpp
@@ -0,0 +1,38 @@
+/// @ref ext_matrix_uint2x2
+/// @file glm/ext/matrix_uint2x2.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint2x2 GLM_EXT_matrix_uint2x2
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint2x2.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint2x2 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint2x2
+	/// @{
+
+	/// Unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2
+	typedef mat<2, 2, uint, defaultp>	umat2x2;
+
+	/// Unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2
+	typedef mat<2, 2, uint, defaultp>	umat2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2_sized.hpp
new file mode 100644
index 000000000000..4555324d2b52
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x2_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_matrix_uint2x2_sized
+/// @file glm/ext/matrix_uint2x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint2x2_sized GLM_EXT_matrix_uint2x2_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint2x2_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x2.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint2x2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint2x2_sized
+	/// @{
+
+	/// 8 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint8, defaultp>				u8mat2x2;
+
+	/// 16 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint16, defaultp>				u16mat2x2;
+
+	/// 32 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint32, defaultp>				u32mat2x2;
+
+	/// 64 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint64, defaultp>				u64mat2x2;
+
+
+	/// 8 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint8, defaultp>				u8mat2;
+
+	/// 16 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint16, defaultp>				u16mat2;
+
+	/// 32 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint32, defaultp>				u32mat2;
+
+	/// 64 bit unsigned integer 2x2 matrix.
+	///
+	/// @see ext_matrix_uint2x2_sized
+	typedef mat<2, 2, uint64, defaultp>				u64mat2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3.hpp
new file mode 100644
index 000000000000..f496c531a0b8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_uint2x3
+/// @file glm/ext/matrix_uint2x3.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint2x3 GLM_EXT_matrix_uint2x3
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint2x3.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x3.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint2x3 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint2x3
+	/// @{
+
+	/// Unsigned integer 2x3 matrix.
+	///
+	/// @see ext_matrix_uint2x3
+	typedef mat<2, 3, uint, defaultp>	umat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3_sized.hpp
new file mode 100644
index 000000000000..db7939c94658
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x3_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_uint2x3_sized
+/// @file glm/ext/matrix_uint2x3_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint2x3_sized GLM_EXT_matrix_uint2x3_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint2x3_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x3.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint2x3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint2x3_sized
+	/// @{
+
+	/// 8 bit unsigned integer 2x3 matrix.
+	///
+	/// @see ext_matrix_uint2x3_sized
+	typedef mat<2, 3, uint8, defaultp>				u8mat2x3;
+
+	/// 16 bit unsigned integer 2x3 matrix.
+	///
+	/// @see ext_matrix_uint2x3_sized
+	typedef mat<2, 3, uint16, defaultp>				u16mat2x3;
+
+	/// 32 bit unsigned integer 2x3 matrix.
+	///
+	/// @see ext_matrix_uint2x3_sized
+	typedef mat<2, 3, uint32, defaultp>				u32mat2x3;
+
+	/// 64 bit unsigned integer 2x3 matrix.
+	///
+	/// @see ext_matrix_uint2x3_sized
+	typedef mat<2, 3, uint64, defaultp>				u64mat2x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4.hpp
new file mode 100644
index 000000000000..0f993509c23b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_uint2x4
+/// @file glm/ext/matrix_uint2x4.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint2x4 GLM_EXT_matrix_int2x4
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint2x4.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint2x4 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint2x4
+	/// @{
+
+	/// Unsigned integer 2x4 matrix.
+	///
+	/// @see ext_matrix_uint2x4
+	typedef mat<2, 4, uint, defaultp>	umat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4_sized.hpp
new file mode 100644
index 000000000000..5c55547ff76a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint2x4_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_uint2x4_sized
+/// @file glm/ext/matrix_uint2x4_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint2x4_sized GLM_EXT_matrix_uint2x4_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint2x4_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x4.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint2x4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint2x4_sized
+	/// @{
+
+	/// 8 bit unsigned integer 2x4 matrix.
+	///
+	/// @see ext_matrix_uint2x4_sized
+	typedef mat<2, 4, uint8, defaultp>				u8mat2x4;
+
+	/// 16 bit unsigned integer 2x4 matrix.
+	///
+	/// @see ext_matrix_uint2x4_sized
+	typedef mat<2, 4, uint16, defaultp>				u16mat2x4;
+
+	/// 32 bit unsigned integer 2x4 matrix.
+	///
+	/// @see ext_matrix_uint2x4_sized
+	typedef mat<2, 4, uint32, defaultp>				u32mat2x4;
+
+	/// 64 bit unsigned integer 2x4 matrix.
+	///
+	/// @see ext_matrix_uint2x4_sized
+	typedef mat<2, 4, uint64, defaultp>				u64mat2x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2.hpp
new file mode 100644
index 000000000000..55a9bed688e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_uint3x2
+/// @file glm/ext/matrix_uint3x2.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint3x2 GLM_EXT_matrix_uint3x2
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint3x2.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint3x2 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint3x2
+	/// @{
+
+	/// Unsigned integer 3x2 matrix.
+	///
+	/// @see ext_matrix_uint3x2
+	typedef mat<3, 2, uint, defaultp>	umat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2_sized.hpp
new file mode 100644
index 000000000000..c81af8f968d6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x2_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_uint3x2_sized
+/// @file glm/ext/matrix_uint3x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint3x2_sized GLM_EXT_matrix_uint3x2_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint3x2_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x2.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint3x2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint3x2_sized
+	/// @{
+
+	/// 8 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_uint3x2_sized
+	typedef mat<3, 2, uint8, defaultp>				u8mat3x2;
+
+	/// 16 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_uint3x2_sized
+	typedef mat<3, 2, uint16, defaultp>				u16mat3x2;
+
+	/// 32 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_uint3x2_sized
+	typedef mat<3, 2, uint32, defaultp>				u32mat3x2;
+
+	/// 64 bit signed integer 3x2 matrix.
+	///
+	/// @see ext_matrix_uint3x2_sized
+	typedef mat<3, 2, uint64, defaultp>				u64mat3x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3.hpp
new file mode 100644
index 000000000000..1004c0d2d541
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3.hpp
@@ -0,0 +1,38 @@
+/// @ref ext_matrix_uint3x3
+/// @file glm/ext/matrix_uint3x3.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint3x3 GLM_EXT_matrix_uint3x3
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint3x3.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x3.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint3x3 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint3x3
+	/// @{
+
+	/// Unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3
+	typedef mat<3, 3, uint, defaultp>	umat3x3;
+
+	/// Unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3
+	typedef mat<3, 3, uint, defaultp>	umat3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3_sized.hpp
new file mode 100644
index 000000000000..41a8be748660
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x3_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_matrix_uint3x3_sized
+/// @file glm/ext/matrix_uint3x3_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint3x3_sized GLM_EXT_matrix_uint3x3_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint3x3_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x3.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint3x3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint3x3_sized
+	/// @{
+
+	/// 8 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint8, defaultp>				u8mat3x3;
+
+	/// 16 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint16, defaultp>				u16mat3x3;
+
+	/// 32 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint32, defaultp>				u32mat3x3;
+
+	/// 64 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint64, defaultp>				u64mat3x3;
+
+
+	/// 8 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint8, defaultp>				u8mat3;
+
+	/// 16 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint16, defaultp>				u16mat3;
+
+	/// 32 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint32, defaultp>				u32mat3;
+
+	/// 64 bit unsigned integer 3x3 matrix.
+	///
+	/// @see ext_matrix_uint3x3_sized
+	typedef mat<3, 3, uint64, defaultp>				u64mat3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4.hpp
new file mode 100644
index 000000000000..c6dd78c4a053
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_uint3x4
+/// @file glm/ext/matrix_uint3x4.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint3x4 GLM_EXT_matrix_uint3x4
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint3x4.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint3x4 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint3x4
+	/// @{
+
+	/// Signed integer 3x4 matrix.
+	///
+	/// @see ext_matrix_uint3x4
+	typedef mat<3, 4, uint, defaultp>	umat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4_sized.hpp
new file mode 100644
index 000000000000..2ce28ad816fc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint3x4_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_uint3x4_sized
+/// @file glm/ext/matrix_uint3x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint3x4_sized GLM_EXT_matrix_uint3x4_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint3x4_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x4.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint3x4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint3x4_sized
+	/// @{
+
+	/// 8 bit unsigned integer 3x4 matrix.
+	///
+	/// @see ext_matrix_uint3x4_sized
+	typedef mat<3, 4, uint8, defaultp>				u8mat3x4;
+
+	/// 16 bit unsigned integer 3x4 matrix.
+	///
+	/// @see ext_matrix_uint3x4_sized
+	typedef mat<3, 4, uint16, defaultp>				u16mat3x4;
+
+	/// 32 bit unsigned integer 3x4 matrix.
+	///
+	/// @see ext_matrix_uint3x4_sized
+	typedef mat<3, 4, uint32, defaultp>				u32mat3x4;
+
+	/// 64 bit unsigned integer 3x4 matrix.
+	///
+	/// @see ext_matrix_uint3x4_sized
+	typedef mat<3, 4, uint64, defaultp>				u64mat3x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2.hpp
new file mode 100644
index 000000000000..0446f5745bde
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_uint4x2
+/// @file glm/ext/matrix_uint4x2.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint4x2 GLM_EXT_matrix_uint4x2
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint4x2.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint4x2 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint4x2
+	/// @{
+
+	/// Unsigned integer 4x2 matrix.
+	///
+	/// @see ext_matrix_uint4x2
+	typedef mat<4, 2, uint, defaultp>	umat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2_sized.hpp
new file mode 100644
index 000000000000..57a66bf9b8b0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x2_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_uint4x2_sized
+/// @file glm/ext/matrix_uint4x2_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint4x2_sized GLM_EXT_matrix_uint4x2_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint4x2_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x2.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint4x2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint4x2_sized
+	/// @{
+
+	/// 8 bit unsigned integer 4x2 matrix.
+	///
+	/// @see ext_matrix_uint4x2_sized
+	typedef mat<4, 2, uint8, defaultp>				u8mat4x2;
+
+	/// 16 bit unsigned integer 4x2 matrix.
+	///
+	/// @see ext_matrix_uint4x2_sized
+	typedef mat<4, 2, uint16, defaultp>				u16mat4x2;
+
+	/// 32 bit unsigned integer 4x2 matrix.
+	///
+	/// @see ext_matrix_uint4x2_sized
+	typedef mat<4, 2, uint32, defaultp>				u32mat4x2;
+
+	/// 64 bit unsigned integer 4x2 matrix.
+	///
+	/// @see ext_matrix_uint4x2_sized
+	typedef mat<4, 2, uint64, defaultp>				u64mat4x2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3.hpp
new file mode 100644
index 000000000000..54c24e4e50b1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3.hpp
@@ -0,0 +1,33 @@
+/// @ref ext_matrix_uint4x3
+/// @file glm/ext/matrix_uint4x3.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint4x3 GLM_EXT_matrix_uint4x3
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint4x3.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x3.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint4x3 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint4x3
+	/// @{
+
+	/// Unsigned integer 4x3 matrix.
+	///
+	/// @see ext_matrix_uint4x3
+	typedef mat<4, 3, uint, defaultp>	umat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3_sized.hpp
new file mode 100644
index 000000000000..2e61124d63b4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x3_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_matrix_uint4x3_sized
+/// @file glm/ext/matrix_uint4x3_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint4x3_sized GLM_EXT_matrix_uint4x3_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint4x3_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x3.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint4x3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint4x3_sized
+	/// @{
+
+	/// 8 bit unsigned integer 4x3 matrix.
+	///
+	/// @see ext_matrix_uint4x3_sized
+	typedef mat<4, 3, uint8, defaultp>				u8mat4x3;
+
+	/// 16 bit unsigned integer 4x3 matrix.
+	///
+	/// @see ext_matrix_uint4x3_sized
+	typedef mat<4, 3, uint16, defaultp>				u16mat4x3;
+
+	/// 32 bit unsigned integer 4x3 matrix.
+	///
+	/// @see ext_matrix_uint4x3_sized
+	typedef mat<4, 3, uint32, defaultp>				u32mat4x3;
+
+	/// 64 bit unsigned integer 4x3 matrix.
+	///
+	/// @see ext_matrix_uint4x3_sized
+	typedef mat<4, 3, uint64, defaultp>				u64mat4x3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4.hpp
new file mode 100644
index 000000000000..5cc84553d936
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4.hpp
@@ -0,0 +1,38 @@
+/// @ref ext_matrix_uint4x4
+/// @file glm/ext/matrix_uint4x4.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint4x4 GLM_EXT_matrix_uint4x4
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint4x4.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint4x4 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint4x4
+	/// @{
+
+	/// Unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4
+	typedef mat<4, 4, uint, defaultp>	umat4x4;
+
+	/// Unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4
+	typedef mat<4, 4, uint, defaultp>	umat4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4_sized.hpp
new file mode 100644
index 000000000000..bb10bd2b77d1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/matrix_uint4x4_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_matrix_uint4x4_sized
+/// @file glm/ext/matrix_uint4x4_sized.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_matrix_uint4x4_sized GLM_EXT_matrix_uint4x4_sized
+/// @ingroup ext
+///
+/// Include <glm/ext/matrix_uint4x4_sized.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat4x4.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_matrix_uint4x4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_matrix_uint4x4_sized
+	/// @{
+
+	/// 8 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint8, defaultp>				u8mat4x4;
+
+	/// 16 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint16, defaultp>				u16mat4x4;
+
+	/// 32 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint32, defaultp>				u32mat4x4;
+
+	/// 64 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint64, defaultp>				u64mat4x4;
+
+
+	/// 8 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint8, defaultp>				u8mat4;
+
+	/// 16 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint16, defaultp>				u16mat4;
+
+	/// 32 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint32, defaultp>				u32mat4;
+
+	/// 64 bit unsigned integer 4x4 matrix.
+	///
+	/// @see ext_matrix_uint4x4_sized
+	typedef mat<4, 4, uint64, defaultp>				u64mat4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.hpp
new file mode 100644
index 000000000000..f738692a472e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.hpp
@@ -0,0 +1,135 @@
+/// @ref ext_quaternion_common
+/// @file glm/ext/quaternion_common.hpp
+///
+/// @defgroup ext_quaternion_common GLM_EXT_quaternion_common
+/// @ingroup ext
+///
+/// Provides common functions for quaternion types
+///
+/// Include <glm/ext/quaternion_common.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_common
+/// @see ext_vector_common
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double
+/// @see ext_quaternion_exponential
+/// @see ext_quaternion_geometric
+/// @see ext_quaternion_relational
+/// @see ext_quaternion_trigonometric
+/// @see ext_quaternion_transform
+
+#pragma once
+
+// Dependency:
+#include "../ext/scalar_constants.hpp"
+#include "../ext/quaternion_geometric.hpp"
+#include "../common.hpp"
+#include "../trigonometric.hpp"
+#include "../exponential.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_common extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_common
+	/// @{
+
+	/// Spherical linear interpolation of two quaternions.
+	/// The interpolation is oriented and the rotation is performed at constant speed.
+	/// For short path spherical linear interpolation, use the slerp function.
+	///
+	/// @param x A quaternion
+	/// @param y A quaternion
+	/// @param a Interpolation factor. The interpolation is defined beyond the range [0, 1].
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	///
+	/// @see - slerp(qua<T, Q> const& x, qua<T, Q> const& y, T const& a)
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> mix(qua<T, Q> const& x, qua<T, Q> const& y, T a);
+
+	/// Linear interpolation of two quaternions.
+	/// The interpolation is oriented.
+	///
+	/// @param x A quaternion
+	/// @param y A quaternion
+	/// @param a Interpolation factor. The interpolation is defined in the range [0, 1].
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> lerp(qua<T, Q> const& x, qua<T, Q> const& y, T a);
+
+	/// Spherical linear interpolation of two quaternions.
+	/// The interpolation always take the short path and the rotation is performed at constant speed.
+	///
+	/// @param x A quaternion
+	/// @param y A quaternion
+	/// @param a Interpolation factor. The interpolation is defined beyond the range [0, 1].
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> slerp(qua<T, Q> const& x, qua<T, Q> const& y, T a);
+
+    /// Spherical linear interpolation of two quaternions with multiple spins over rotation axis.
+    /// The interpolation always take the short path when the spin count is positive and long path
+    /// when count is negative. Rotation is performed at constant speed.
+    ///
+    /// @param x A quaternion
+    /// @param y A quaternion
+    /// @param a Interpolation factor. The interpolation is defined beyond the range [0, 1].
+    /// @param k Additional spin count. If Value is negative interpolation will be on "long" path.
+    ///
+    /// @tparam T A floating-point scalar type
+    /// @tparam S An integer scalar type
+    /// @tparam Q A value from qualifier enum
+    template<typename T, typename S, qualifier Q>
+    GLM_FUNC_DECL qua<T, Q> slerp(qua<T, Q> const& x, qua<T, Q> const& y, T a, S k);
+
+	/// Returns the q conjugate.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> conjugate(qua<T, Q> const& q);
+
+	/// Returns the q inverse.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> inverse(qua<T, Q> const& q);
+
+	/// Returns true if x holds a NaN (not a number)
+	/// representation in the underlying implementation's set of
+	/// floating point representations. Returns false otherwise,
+	/// including for implementations with no NaN
+	/// representations.
+	///
+	/// /!\ When using compiler fast math, this function may fail.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, bool, Q> isnan(qua<T, Q> const& x);
+
+	/// Returns true if x holds a positive infinity or negative
+	/// infinity representation in the underlying implementation's
+	/// set of floating point representations. Returns false
+	/// otherwise, including for implementations with no infinity
+	/// representations.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, bool, Q> isinf(qua<T, Q> const& x);
+
+	/// @}
+} //namespace glm
+
+#include "quaternion_common.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.inl
new file mode 100644
index 000000000000..ad171f9d4bdf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common.inl
@@ -0,0 +1,144 @@
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> mix(qua<T, Q> const& x, qua<T, Q> const& y, T a)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'mix' only accept floating-point inputs");
+
+		T const cosTheta = dot(x, y);
+
+		// Perform a linear interpolation when cosTheta is close to 1 to avoid side effect of sin(angle) becoming a zero denominator
+		if(cosTheta > static_cast<T>(1) - epsilon<T>())
+		{
+			// Linear interpolation
+			return qua<T, Q>::wxyz(
+				mix(x.w, y.w, a),
+				mix(x.x, y.x, a),
+				mix(x.y, y.y, a),
+				mix(x.z, y.z, a));
+		}
+		else
+		{
+			// Essential Mathematics, page 467
+			T angle = acos(cosTheta);
+			return (sin((static_cast<T>(1) - a) * angle) * x + sin(a * angle) * y) / sin(angle);
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> lerp(qua<T, Q> const& x, qua<T, Q> const& y, T a)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'lerp' only accept floating-point inputs");
+
+		// Lerp is only defined in [0, 1]
+		assert(a >= static_cast<T>(0));
+		assert(a <= static_cast<T>(1));
+
+		return x * (static_cast<T>(1) - a) + (y * a);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> slerp(qua<T, Q> const& x, qua<T, Q> const& y, T a)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'slerp' only accept floating-point inputs");
+
+		qua<T, Q> z = y;
+
+		T cosTheta = dot(x, y);
+
+		// If cosTheta < 0, the interpolation will take the long way around the sphere.
+		// To fix this, one quat must be negated.
+		if(cosTheta < static_cast<T>(0))
+		{
+			z = -y;
+			cosTheta = -cosTheta;
+		}
+
+		// Perform a linear interpolation when cosTheta is close to 1 to avoid side effect of sin(angle) becoming a zero denominator
+		if(cosTheta > static_cast<T>(1) - epsilon<T>())
+		{
+			// Linear interpolation
+			return qua<T, Q>::wxyz(
+				mix(x.w, z.w, a),
+				mix(x.x, z.x, a),
+				mix(x.y, z.y, a),
+				mix(x.z, z.z, a));
+		}
+		else
+		{
+			// Essential Mathematics, page 467
+			T angle = acos(cosTheta);
+			return (sin((static_cast<T>(1) - a) * angle) * x + sin(a * angle) * z) / sin(angle);
+		}
+	}
+
+    template<typename T, typename S, qualifier Q>
+    GLM_FUNC_QUALIFIER qua<T, Q> slerp(qua<T, Q> const& x, qua<T, Q> const& y, T a, S k)
+    {
+        GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'slerp' only accept floating-point inputs");
+        GLM_STATIC_ASSERT(std::numeric_limits<S>::is_integer, "'slerp' only accept integer for spin count");
+
+        qua<T, Q> z = y;
+
+        T cosTheta = dot(x, y);
+
+        // If cosTheta < 0, the interpolation will take the long way around the sphere.
+        // To fix this, one quat must be negated.
+        if (cosTheta < static_cast<T>(0))
+        {
+            z = -y;
+            cosTheta = -cosTheta;
+        }
+
+        // Perform a linear interpolation when cosTheta is close to 1 to avoid side effect of sin(angle) becoming a zero denominator
+        if (cosTheta > static_cast<T>(1) - epsilon<T>())
+        {
+            // Linear interpolation
+            return qua<T, Q>::wxyz(
+                mix(x.w, z.w, a),
+                mix(x.x, z.x, a),
+                mix(x.y, z.y, a),
+                mix(x.z, z.z, a));
+        }
+        else
+        {
+            // Graphics Gems III, page 96
+            T angle = acos(cosTheta);
+            T phi = angle + static_cast<T>(k) * glm::pi<T>();
+            return (sin(angle - a * phi)* x + sin(a * phi) * z) / sin(angle);
+        }
+    }
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> conjugate(qua<T, Q> const& q)
+	{
+		return qua<T, Q>::wxyz(q.w, -q.x, -q.y, -q.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> inverse(qua<T, Q> const& q)
+	{
+		return conjugate(q) / dot(q, q);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> isnan(qua<T, Q> const& q)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isnan' only accept floating-point inputs");
+
+		return vec<4, bool, Q>(isnan(q.x), isnan(q.y), isnan(q.z), isnan(q.w));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> isinf(qua<T, Q> const& q)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isinf' only accept floating-point inputs");
+
+		return vec<4, bool, Q>(isinf(q.x), isinf(q.y), isinf(q.z), isinf(q.w));
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "quaternion_common_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common_simd.inl
new file mode 100644
index 000000000000..ddfc8a44f6a3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_common_simd.inl
@@ -0,0 +1,18 @@
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+namespace glm{
+namespace detail
+{
+	template<qualifier Q>
+	struct compute_dot<qua<float, Q>, float, true>
+	{
+		static GLM_FUNC_QUALIFIER float call(qua<float, Q> const& x, qua<float, Q> const& y)
+		{
+			return _mm_cvtss_f32(glm_vec1_dot(x.data, y.data));
+		}
+	};
+}//namespace detail
+}//namespace glm
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double.hpp
new file mode 100644
index 000000000000..63b24de4d52a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double.hpp
@@ -0,0 +1,39 @@
+/// @ref ext_quaternion_double
+/// @file glm/ext/quaternion_double.hpp
+///
+/// @defgroup ext_quaternion_double GLM_EXT_quaternion_double
+/// @ingroup ext
+///
+/// Exposes double-precision floating point quaternion type.
+///
+/// Include <glm/ext/quaternion_double.hpp> to use the features of this extension.
+///
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double_precision
+/// @see ext_quaternion_common
+/// @see ext_quaternion_exponential
+/// @see ext_quaternion_geometric
+/// @see ext_quaternion_relational
+/// @see ext_quaternion_transform
+/// @see ext_quaternion_trigonometric
+
+#pragma once
+
+// Dependency:
+#include "../detail/type_quat.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_double extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_double
+	/// @{
+
+	/// Quaternion of double-precision floating-point numbers.
+	typedef qua<double, defaultp>		dquat;
+
+	/// @}
+} //namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double_precision.hpp
new file mode 100644
index 000000000000..8aa24a17752d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_double_precision.hpp
@@ -0,0 +1,42 @@
+/// @ref ext_quaternion_double_precision
+/// @file glm/ext/quaternion_double_precision.hpp
+///
+/// @defgroup ext_quaternion_double_precision GLM_EXT_quaternion_double_precision
+/// @ingroup ext
+///
+/// Exposes double-precision floating point quaternion type with various precision in term of ULPs.
+///
+/// Include <glm/ext/quaternion_double_precision.hpp> to use the features of this extension.
+
+#pragma once
+
+// Dependency:
+#include "../detail/type_quat.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_double_precision extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_double_precision
+	/// @{
+
+	/// Quaternion of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see ext_quaternion_double_precision
+	typedef qua<double, lowp>		lowp_dquat;
+
+	/// Quaternion of medium double-qualifier floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see ext_quaternion_double_precision
+	typedef qua<double, mediump>	mediump_dquat;
+
+	/// Quaternion of high double-qualifier floating-point numbers using high precision arithmetic in term of ULPs.
+	///
+	/// @see ext_quaternion_double_precision
+	typedef qua<double, highp>		highp_dquat;
+
+	/// @}
+} //namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.hpp
new file mode 100644
index 000000000000..affe2979aad5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.hpp
@@ -0,0 +1,63 @@
+/// @ref ext_quaternion_exponential
+/// @file glm/ext/quaternion_exponential.hpp
+///
+/// @defgroup ext_quaternion_exponential GLM_EXT_quaternion_exponential
+/// @ingroup ext
+///
+/// Provides exponential functions for quaternion types
+///
+/// Include <glm/ext/quaternion_exponential.hpp> to use the features of this extension.
+///
+/// @see core_exponential
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double
+
+#pragma once
+
+// Dependency:
+#include "../common.hpp"
+#include "../trigonometric.hpp"
+#include "../geometric.hpp"
+#include "../ext/scalar_constants.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_exponential extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_transform
+	/// @{
+
+	/// Returns a exponential of a quaternion.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> exp(qua<T, Q> const& q);
+
+	/// Returns a logarithm of a quaternion
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> log(qua<T, Q> const& q);
+
+	/// Returns a quaternion raised to a power.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> pow(qua<T, Q> const& q, T y);
+
+	/// Returns the square root of a quaternion
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> sqrt(qua<T, Q> const& q);
+
+	/// @}
+} //namespace glm
+
+#include "quaternion_exponential.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.inl
new file mode 100644
index 000000000000..8a9d774b396a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_exponential.inl
@@ -0,0 +1,89 @@
+#include "scalar_constants.hpp"
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> exp(qua<T, Q> const& q)
+	{
+		vec<3, T, Q> u(q.x, q.y, q.z);
+		T const Angle = glm::length(u);
+		if (Angle < epsilon<T>())
+			return qua<T, Q>();
+
+		vec<3, T, Q> const v(u / Angle);
+		return qua<T, Q>(cos(Angle), sin(Angle) * v);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> log(qua<T, Q> const& q)
+	{
+		vec<3, T, Q> u(q.x, q.y, q.z);
+		T Vec3Len = length(u);
+
+		if (Vec3Len < epsilon<T>())
+		{
+			if(q.w > static_cast<T>(0))
+				return qua<T, Q>::wxyz(log(q.w), static_cast<T>(0), static_cast<T>(0), static_cast<T>(0));
+			else if(q.w < static_cast<T>(0))
+				return qua<T, Q>::wxyz(log(-q.w), pi<T>(), static_cast<T>(0), static_cast<T>(0));
+			else
+				return qua<T, Q>::wxyz(std::numeric_limits<T>::infinity(), std::numeric_limits<T>::infinity(), std::numeric_limits<T>::infinity(), std::numeric_limits<T>::infinity());
+		}
+		else
+		{
+			T t = atan(Vec3Len, T(q.w)) / Vec3Len;
+			T QuatLen2 = Vec3Len * Vec3Len + q.w * q.w;
+			return qua<T, Q>::wxyz(static_cast<T>(0.5) * log(QuatLen2), t * q.x, t * q.y, t * q.z);
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> pow(qua<T, Q> const& x, T y)
+	{
+		//Raising to the power of 0 should yield 1
+		//Needed to prevent a division by 0 error later on
+		if(y > -epsilon<T>() && y < epsilon<T>())
+			return qua<T, Q>::wxyz(1,0,0,0);
+
+		//To deal with non-unit quaternions
+		T magnitude = sqrt(x.x * x.x + x.y * x.y + x.z * x.z + x.w *x.w);
+
+		T Angle;
+		if(abs(x.w / magnitude) > cos_one_over_two<T>())
+		{
+			//Scalar component is close to 1; using it to recover angle would lose precision
+			//Instead, we use the non-scalar components since sin() is accurate around 0
+
+			//Prevent a division by 0 error later on
+			T VectorMagnitude = x.x * x.x + x.y * x.y + x.z * x.z;
+			//Despite the compiler might say, we actually want to compare
+			//VectorMagnitude to 0. here; we could use denorm_int() compiling a
+			//project with unsafe maths optimizations might make the comparison
+			//always false, even when VectorMagnitude is 0.
+			if (VectorMagnitude < std::numeric_limits<T>::min()) {
+				//Equivalent to raising a real number to a power
+				return qua<T, Q>::wxyz(pow(x.w, y), 0, 0, 0);
+			}
+
+			Angle = asin(sqrt(VectorMagnitude) / magnitude);
+		}
+		else
+		{
+			//Scalar component is small, shouldn't cause loss of precision
+			Angle = acos(x.w / magnitude);
+		}
+
+		T NewAngle = Angle * y;
+		T Div = sin(NewAngle) / sin(Angle);
+		T Mag = pow(magnitude, y - static_cast<T>(1));
+		return qua<T, Q>::wxyz(cos(NewAngle) * magnitude * Mag, x.x * Div * Mag, x.y * Div * Mag, x.z * Div * Mag);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> sqrt(qua<T, Q> const& x)
+	{
+		return pow(x, static_cast<T>(0.5));
+	}
+}//namespace glm
+
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float.hpp
new file mode 100644
index 000000000000..ca42a60597f4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float.hpp
@@ -0,0 +1,39 @@
+/// @ref ext_quaternion_float
+/// @file glm/ext/quaternion_float.hpp
+///
+/// @defgroup ext_quaternion_float GLM_EXT_quaternion_float
+/// @ingroup ext
+///
+/// Exposes single-precision floating point quaternion type.
+///
+/// Include <glm/ext/quaternion_float.hpp> to use the features of this extension.
+///
+/// @see ext_quaternion_double
+/// @see ext_quaternion_float_precision
+/// @see ext_quaternion_common
+/// @see ext_quaternion_exponential
+/// @see ext_quaternion_geometric
+/// @see ext_quaternion_relational
+/// @see ext_quaternion_transform
+/// @see ext_quaternion_trigonometric
+
+#pragma once
+
+// Dependency:
+#include "../detail/type_quat.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_float extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_float
+	/// @{
+
+	/// Quaternion of single-precision floating-point numbers.
+	typedef qua<float, defaultp>		quat;
+
+	/// @}
+} //namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float_precision.hpp
new file mode 100644
index 000000000000..f9e4f5c21d90
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_float_precision.hpp
@@ -0,0 +1,36 @@
+/// @ref ext_quaternion_float_precision
+/// @file glm/ext/quaternion_float_precision.hpp
+///
+/// @defgroup ext_quaternion_float_precision GLM_EXT_quaternion_float_precision
+/// @ingroup ext
+///
+/// Exposes single-precision floating point quaternion type with various precision in term of ULPs.
+///
+/// Include <glm/ext/quaternion_float_precision.hpp> to use the features of this extension.
+
+#pragma once
+
+// Dependency:
+#include "../detail/type_quat.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_float_precision extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_float_precision
+	/// @{
+
+	/// Quaternion of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef qua<float, lowp>		lowp_quat;
+
+	/// Quaternion of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef qua<float, mediump>		mediump_quat;
+
+	/// Quaternion of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef qua<float, highp>		highp_quat;
+
+	/// @}
+} //namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.hpp
new file mode 100644
index 000000000000..6a2403fd2818
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_quaternion_geometric
+/// @file glm/ext/quaternion_geometric.hpp
+///
+/// @defgroup ext_quaternion_geometric GLM_EXT_quaternion_geometric
+/// @ingroup ext
+///
+/// Provides geometric functions for quaternion types
+///
+/// Include <glm/ext/quaternion_geometric.hpp> to use the features of this extension.
+///
+/// @see core_func_geometric
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double
+
+#pragma once
+
+// Dependency:
+#include "../geometric.hpp"
+#include "../exponential.hpp"
+#include "../ext/vector_relational.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_geometric extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_geometric
+	/// @{
+
+	/// Returns the norm of a quaternions
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_geometric
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T length(qua<T, Q> const& q);
+
+	/// Returns the normalized quaternion.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_geometric
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> normalize(qua<T, Q> const& q);
+
+	/// Returns dot product of q1 and q2, i.e., q1[0] * q2[0] + q1[1] * q2[1] + ...
+	///
+	/// @tparam T Floating-point scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_geometric
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR T dot(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Compute a cross product.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_geometric
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> cross(qua<T, Q> const& q1, qua<T, Q> const& q2);
+
+	/// @}
+} //namespace glm
+
+#include "quaternion_geometric.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.inl
new file mode 100644
index 000000000000..88dc4d63de18
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_geometric.inl
@@ -0,0 +1,36 @@
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T dot(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'dot' accepts only floating-point inputs");
+		return detail::compute_dot<qua<T, Q>, T, detail::is_aligned<Q>::value>::call(x, y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T length(qua<T, Q> const& q)
+	{
+		return glm::sqrt(dot(q, q));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> normalize(qua<T, Q> const& q)
+	{
+		T len = length(q);
+		if(len <= static_cast<T>(0)) // Problem
+			return qua<T, Q>::wxyz(static_cast<T>(1), static_cast<T>(0), static_cast<T>(0), static_cast<T>(0));
+		T oneOverLen = static_cast<T>(1) / len;
+		return qua<T, Q>::wxyz(q.w * oneOverLen, q.x * oneOverLen, q.y * oneOverLen, q.z * oneOverLen);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> cross(qua<T, Q> const& q1, qua<T, Q> const& q2)
+	{
+		return qua<T, Q>::wxyz(
+			q1.w * q2.w - q1.x * q2.x - q1.y * q2.y - q1.z * q2.z,
+			q1.w * q2.x + q1.x * q2.w + q1.y * q2.z - q1.z * q2.y,
+			q1.w * q2.y + q1.y * q2.w + q1.z * q2.x - q1.x * q2.z,
+			q1.w * q2.z + q1.z * q2.w + q1.x * q2.y - q1.y * q2.x);
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.hpp
new file mode 100644
index 000000000000..7aa121da0a15
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.hpp
@@ -0,0 +1,62 @@
+/// @ref ext_quaternion_relational
+/// @file glm/ext/quaternion_relational.hpp
+///
+/// @defgroup ext_quaternion_relational GLM_EXT_quaternion_relational
+/// @ingroup ext
+///
+/// Exposes comparison functions for quaternion types that take a user defined epsilon values.
+///
+/// Include <glm/ext/quaternion_relational.hpp> to use the features of this extension.
+///
+/// @see core_vector_relational
+/// @see ext_vector_relational
+/// @see ext_matrix_relational
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double
+
+#pragma once
+
+// Dependency:
+#include "../vector_relational.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_relational extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_relational
+	/// @{
+
+	/// Returns the component-wise comparison of result x == y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, bool, Q> equal(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, bool, Q> equal(qua<T, Q> const& x, qua<T, Q> const& y, T epsilon);
+
+	/// Returns the component-wise comparison of result x != y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, bool, Q> notEqual(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Returns the component-wise comparison of |x - y| >= epsilon.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, bool, Q> notEqual(qua<T, Q> const& x, qua<T, Q> const& y, T epsilon);
+
+	/// @}
+} //namespace glm
+
+#include "quaternion_relational.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.inl
new file mode 100644
index 000000000000..b1713e95c6c5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_relational.inl
@@ -0,0 +1,35 @@
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> equal(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		vec<4, bool, Q> Result;
+		for(length_t i = 0; i < x.length(); ++i)
+			Result[i] = x[i] == y[i];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> equal(qua<T, Q> const& x, qua<T, Q> const& y, T epsilon)
+	{
+		vec<4, T, Q> v(x.x - y.x, x.y - y.y, x.z - y.z, x.w - y.w);
+		return lessThan(abs(v), vec<4, T, Q>(epsilon));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> notEqual(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		vec<4, bool, Q> Result;
+		for(length_t i = 0; i < x.length(); ++i)
+			Result[i] = x[i] != y[i];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> notEqual(qua<T, Q> const& x, qua<T, Q> const& y, T epsilon)
+	{
+		vec<4, T, Q> v(x.x - y.x, x.y - y.y, x.z - y.z, x.w - y.w);
+		return greaterThanEqual(abs(v), vec<4, T, Q>(epsilon));
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.hpp
new file mode 100644
index 000000000000..a9cc5c2b59ff
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.hpp
@@ -0,0 +1,47 @@
+/// @ref ext_quaternion_transform
+/// @file glm/ext/quaternion_transform.hpp
+///
+/// @defgroup ext_quaternion_transform GLM_EXT_quaternion_transform
+/// @ingroup ext
+///
+/// Provides transformation functions for quaternion types
+///
+/// Include <glm/ext/quaternion_transform.hpp> to use the features of this extension.
+///
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double
+/// @see ext_quaternion_exponential
+/// @see ext_quaternion_geometric
+/// @see ext_quaternion_relational
+/// @see ext_quaternion_trigonometric
+
+#pragma once
+
+// Dependency:
+#include "../common.hpp"
+#include "../trigonometric.hpp"
+#include "../geometric.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_transform extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_transform
+	/// @{
+
+	/// Rotates a quaternion from a vector of 3 components axis and an angle.
+	///
+	/// @param q Source orientation
+	/// @param angle Angle expressed in radians.
+	/// @param axis Axis of the rotation
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> rotate(qua<T, Q> const& q, T const& angle, vec<3, T, Q> const& axis);
+	/// @}
+} //namespace glm
+
+#include "quaternion_transform.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.inl
new file mode 100644
index 000000000000..7e773fbd264d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_transform.inl
@@ -0,0 +1,24 @@
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> rotate(qua<T, Q> const& q, T const& angle, vec<3, T, Q> const& v)
+	{
+		vec<3, T, Q> Tmp = v;
+
+		// Axis of rotation must be normalised
+		T len = glm::length(Tmp);
+		if(abs(len - static_cast<T>(1)) > static_cast<T>(0.001))
+		{
+			T oneOverLen = static_cast<T>(1) / len;
+			Tmp.x *= oneOverLen;
+			Tmp.y *= oneOverLen;
+			Tmp.z *= oneOverLen;
+		}
+
+		T const AngleRad(angle);
+		T const Sin = sin(AngleRad * static_cast<T>(0.5));
+
+		return q * qua<T, Q>::wxyz(cos(AngleRad * static_cast<T>(0.5)), Tmp.x * Sin, Tmp.y * Sin, Tmp.z * Sin);
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.hpp
new file mode 100644
index 000000000000..574a70479576
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.hpp
@@ -0,0 +1,65 @@
+/// @ref ext_quaternion_trigonometric
+/// @file glm/ext/quaternion_trigonometric.hpp
+///
+/// @defgroup ext_quaternion_trigonometric GLM_EXT_quaternion_trigonometric
+/// @ingroup ext
+///
+/// Provides trigonometric functions for quaternion types
+///
+/// Include <glm/ext/quaternion_trigonometric.hpp> to use the features of this extension.
+///
+/// @see ext_quaternion_float
+/// @see ext_quaternion_double
+/// @see ext_quaternion_exponential
+/// @see ext_quaternion_geometric
+/// @see ext_quaternion_relational
+/// @see ext_quaternion_transform
+
+#pragma once
+
+// Dependency:
+#include "../trigonometric.hpp"
+#include "../exponential.hpp"
+#include "scalar_constants.hpp"
+#include "vector_relational.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_quaternion_trigonometric extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_quaternion_trigonometric
+	/// @{
+
+	/// Returns the quaternion rotation angle.
+	///
+	/// @param x A normalized quaternion.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T angle(qua<T, Q> const& x);
+
+	/// Returns the q rotation axis.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> axis(qua<T, Q> const& x);
+
+	/// Build a quaternion from an angle and a normalized axis.
+	///
+	/// @param angle Angle expressed in radians.
+	/// @param axis Axis of the quaternion, must be normalized.
+	///
+	/// @tparam T A floating-point scalar type
+	/// @tparam Q A value from qualifier enum
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> angleAxis(T const& angle, vec<3, T, Q> const& axis);
+
+	/// @}
+} //namespace glm
+
+#include "quaternion_trigonometric.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.inl
new file mode 100644
index 000000000000..896449aa6e1f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/quaternion_trigonometric.inl
@@ -0,0 +1,37 @@
+#include "scalar_constants.hpp"
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T angle(qua<T, Q> const& x)
+	{
+		if (abs(x.w) > cos_one_over_two<T>())
+		{
+			T const a = asin(sqrt(x.x * x.x + x.y * x.y + x.z * x.z)) * static_cast<T>(2);
+			if(x.w < static_cast<T>(0))
+				return pi<T>() * static_cast<T>(2) - a;
+			return a;
+		}
+
+		return acos(x.w) * static_cast<T>(2);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> axis(qua<T, Q> const& x)
+	{
+		T const tmp1 = static_cast<T>(1) - x.w * x.w;
+		if(tmp1 <= static_cast<T>(0))
+			return vec<3, T, Q>(0, 0, 1);
+		T const tmp2 = static_cast<T>(1) / sqrt(tmp1);
+		return vec<3, T, Q>(x.x * tmp2, x.y * tmp2, x.z * tmp2);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> angleAxis(T const& angle, vec<3, T, Q> const& v)
+	{
+		T const a(angle);
+		T const s = glm::sin(a * static_cast<T>(0.5));
+
+		return qua<T, Q>(glm::cos(a * static_cast<T>(0.5)), v * s);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.hpp
new file mode 100644
index 000000000000..df04b6b809a9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.hpp
@@ -0,0 +1,181 @@
+/// @ref ext_scalar_common
+/// @file glm/ext/scalar_common.hpp
+///
+/// @defgroup ext_scalar_common GLM_EXT_scalar_common
+/// @ingroup ext
+///
+/// Exposes min and max functions for 3 to 4 scalar parameters.
+///
+/// Include <glm/ext/scalar_common.hpp> to use the features of this extension.
+///
+/// @see core_func_common
+/// @see ext_vector_common
+
+#pragma once
+
+// Dependency:
+#include "../common.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_common extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_common
+	/// @{
+
+	/// Returns the minimum component-wise values of 3 inputs
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T min(T a, T b, T c);
+
+	/// Returns the minimum component-wise values of 4 inputs
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T min(T a, T b, T c, T d);
+
+	/// Returns the maximum component-wise values of 3 inputs
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T max(T a, T b, T c);
+
+	/// Returns the maximum component-wise values of 4 inputs
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T max(T a, T b, T c, T d);
+
+	/// Returns the minimum component-wise values of 2 inputs. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T fmin(T a, T b);
+
+	/// Returns the minimum component-wise values of 3 inputs. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T fmin(T a, T b, T c);
+
+	/// Returns the minimum component-wise values of 4 inputs. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T fmin(T a, T b, T c, T d);
+
+	/// Returns the maximum component-wise values of 2 inputs. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T fmax(T a, T b);
+
+	/// Returns the maximum component-wise values of 3 inputs. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T fmax(T a, T b, T C);
+
+	/// Returns the maximum component-wise values of 4 inputs. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam T A floating-point scalar type.
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	/// @see ext_scalar_common
+	template<typename T>
+	GLM_FUNC_DECL T fmax(T a, T b, T C, T D);
+
+	/// Returns min(max(x, minVal), maxVal) for each component in x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam genType Floating-point scalar types.
+	///
+	/// @see ext_scalar_common
+	template<typename genType>
+	GLM_FUNC_DECL genType fclamp(genType x, genType minVal, genType maxVal);
+
+	/// Simulate GL_CLAMP OpenGL wrap mode
+	///
+	/// @tparam genType Floating-point scalar types.
+	///
+	/// @see ext_scalar_common extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType clamp(genType const& Texcoord);
+
+	/// Simulate GL_REPEAT OpenGL wrap mode
+	///
+	/// @tparam genType Floating-point scalar types.
+	///
+	/// @see ext_scalar_common extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType repeat(genType const& Texcoord);
+
+	/// Simulate GL_MIRRORED_REPEAT OpenGL wrap mode
+	///
+	/// @tparam genType Floating-point scalar types.
+	///
+	/// @see ext_scalar_common extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType mirrorClamp(genType const& Texcoord);
+
+	/// Simulate GL_MIRROR_REPEAT OpenGL wrap mode
+	///
+	/// @tparam genType Floating-point scalar types.
+	///
+	/// @see ext_scalar_common extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType mirrorRepeat(genType const& Texcoord);
+
+	/// Returns a value equal to the nearest integer to x.
+	/// The fraction 0.5 will round in a direction chosen by the
+	/// implementation, presumably the direction that is fastest.
+	///
+	/// @param x The values of the argument must be greater or equal to zero.
+	/// @tparam genType floating point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/round.xml">GLSL round man page</a>
+	/// @see ext_scalar_common extension.
+	template<typename genType>
+	GLM_FUNC_DECL int iround(genType const& x);
+
+	/// Returns a value equal to the nearest integer to x.
+	/// The fraction 0.5 will round in a direction chosen by the
+	/// implementation, presumably the direction that is fastest.
+	///
+	/// @param x The values of the argument must be greater or equal to zero.
+	/// @tparam genType floating point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/round.xml">GLSL round man page</a>
+	/// @see ext_scalar_common extension.
+	template<typename genType>
+	GLM_FUNC_DECL uint uround(genType const& x);
+
+	/// @}
+}//namespace glm
+
+#include "scalar_common.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.inl
new file mode 100644
index 000000000000..3d09fef0cb4f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_common.inl
@@ -0,0 +1,170 @@
+namespace glm
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER T min(T a, T b, T c)
+	{
+		return glm::min(glm::min(a, b), c);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T min(T a, T b, T c, T d)
+	{
+		return glm::min(glm::min(a, b), glm::min(c, d));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T max(T a, T b, T c)
+	{
+		return glm::max(glm::max(a, b), c);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T max(T a, T b, T c, T d)
+	{
+		return glm::max(glm::max(a, b), glm::max(c, d));
+	}
+
+#	if GLM_HAS_CXX11_STL
+		using std::fmin;
+#	else
+		template<typename T>
+		GLM_FUNC_QUALIFIER T fmin(T a, T b)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fmin' only accept floating-point input");
+
+			if (isnan(a))
+				return b;
+			return min(a, b);
+		}
+#	endif
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fmin(T a, T b, T c)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fmin' only accept floating-point input");
+
+		if (isnan(a))
+			return fmin(b, c);
+		if (isnan(b))
+			return fmin(a, c);
+		if (isnan(c))
+			return min(a, b);
+		return min(a, b, c);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fmin(T a, T b, T c, T d)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fmin' only accept floating-point input");
+
+		if (isnan(a))
+			return fmin(b, c, d);
+		if (isnan(b))
+			return min(a, fmin(c, d));
+		if (isnan(c))
+			return fmin(min(a, b), d);
+		if (isnan(d))
+			return min(a, b, c);
+		return min(a, b, c, d);
+	}
+
+
+#	if GLM_HAS_CXX11_STL
+		using std::fmax;
+#	else
+		template<typename T>
+		GLM_FUNC_QUALIFIER T fmax(T a, T b)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fmax' only accept floating-point input");
+
+			if (isnan(a))
+				return b;
+			return max(a, b);
+		}
+#	endif
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fmax(T a, T b, T c)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fmax' only accept floating-point input");
+
+		if (isnan(a))
+			return fmax(b, c);
+		if (isnan(b))
+			return fmax(a, c);
+		if (isnan(c))
+			return max(a, b);
+		return max(a, b, c);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fmax(T a, T b, T c, T d)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fmax' only accept floating-point input");
+
+		if (isnan(a))
+			return fmax(b, c, d);
+		if (isnan(b))
+			return max(a, fmax(c, d));
+		if (isnan(c))
+			return fmax(max(a, b), d);
+		if (isnan(d))
+			return max(a, b, c);
+		return max(a, b, c, d);
+	}
+
+	// fclamp
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fclamp(genType x, genType minVal, genType maxVal)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fclamp' only accept floating-point or integer inputs");
+		return fmin(fmax(x, minVal), maxVal);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType clamp(genType const& Texcoord)
+	{
+		return glm::clamp(Texcoord, static_cast<genType>(0), static_cast<genType>(1));
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType repeat(genType const& Texcoord)
+	{
+		return glm::fract(Texcoord);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType mirrorClamp(genType const& Texcoord)
+	{
+		return glm::fract(glm::abs(Texcoord));
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType mirrorRepeat(genType const& Texcoord)
+	{
+		genType const Abs = glm::abs(Texcoord);
+		genType const Clamp = glm::mod(glm::floor(Abs), static_cast<genType>(2));
+		genType const Floor = glm::floor(Abs);
+		genType const Rest = Abs - Floor;
+		genType const Mirror = Clamp + Rest;
+		return mix(Rest, static_cast<genType>(1) - Rest, Mirror >= static_cast<genType>(1));
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER int iround(genType const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'iround' only accept floating-point inputs");
+		assert(static_cast<genType>(0.0) <= x);
+
+		return static_cast<int>(x + static_cast<genType>(0.5));
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER uint uround(genType const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'uround' only accept floating-point inputs");
+		assert(static_cast<genType>(0.0) <= x);
+
+		return static_cast<uint>(x + static_cast<genType>(0.5));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.hpp
new file mode 100644
index 000000000000..74e210d9c09e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.hpp
@@ -0,0 +1,40 @@
+/// @ref ext_scalar_constants
+/// @file glm/ext/scalar_constants.hpp
+///
+/// @defgroup ext_scalar_constants GLM_EXT_scalar_constants
+/// @ingroup ext
+///
+/// Provides a list of constants and precomputed useful values.
+///
+/// Include <glm/ext/scalar_constants.hpp> to use the features of this extension.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_constants extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_constants
+	/// @{
+
+	/// Return the epsilon constant for floating point types.
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType epsilon();
+
+	/// Return the pi constant for floating point types.
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType pi();
+
+	/// Return the value of cos(1 / 2) for floating point types.
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType cos_one_over_two();
+
+	/// @}
+} //namespace glm
+
+#include "scalar_constants.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.inl
new file mode 100644
index 000000000000..b928e5118bff
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_constants.inl
@@ -0,0 +1,24 @@
+#include <limits>
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType epsilon()
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'epsilon' only accepts floating-point inputs");
+		return std::numeric_limits<genType>::epsilon();
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType pi()
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'pi' only accepts floating-point inputs");
+		return static_cast<genType>(3.14159265358979323846264338327950288);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType cos_one_over_two()
+	{
+		return genType(0.877582561890372716130286068203503191);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_int_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_int_sized.hpp
new file mode 100644
index 000000000000..8e9c511c9cb5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_int_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_scalar_int_sized
+/// @file glm/ext/scalar_int_sized.hpp
+///
+/// @defgroup ext_scalar_int_sized GLM_EXT_scalar_int_sized
+/// @ingroup ext
+///
+/// Exposes sized signed integer scalar types.
+///
+/// Include <glm/ext/scalar_int_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_uint_sized
+
+#pragma once
+
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_int_sized extension included")
+#endif
+
+namespace glm{
+namespace detail
+{
+#	if GLM_HAS_EXTENDED_INTEGER_TYPE
+		typedef std::int8_t			int8;
+		typedef std::int16_t		int16;
+		typedef std::int32_t		int32;
+#	else
+		typedef signed char			int8;
+		typedef signed short		int16;
+		typedef signed int			int32;
+#endif//
+
+	template<>
+	struct is_int<int8>
+	{
+		enum test {value = ~0};
+	};
+
+	template<>
+	struct is_int<int16>
+	{
+		enum test {value = ~0};
+	};
+
+	template<>
+	struct is_int<int64>
+	{
+		enum test {value = ~0};
+	};
+}//namespace detail
+
+
+	/// @addtogroup ext_scalar_int_sized
+	/// @{
+
+	/// 8 bit signed integer type.
+	typedef detail::int8		int8;
+
+	/// 16 bit signed integer type.
+	typedef detail::int16		int16;
+
+	/// 32 bit signed integer type.
+	typedef detail::int32		int32;
+
+	/// 64 bit signed integer type.
+	typedef detail::int64		int64;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.hpp
new file mode 100644
index 000000000000..a2ca8a2ae37c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.hpp
@@ -0,0 +1,92 @@
+/// @ref ext_scalar_integer
+/// @file glm/ext/scalar_integer.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_scalar_integer GLM_EXT_scalar_integer
+/// @ingroup ext
+///
+/// Include <glm/ext/scalar_integer.hpp> to use the features of this extension.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/_vectorize.hpp"
+#include "../detail/type_float.hpp"
+#include "../vector_relational.hpp"
+#include "../common.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_integer extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_integer
+	/// @{
+
+	/// Return true if the value is a power of two number.
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL bool isPowerOfTwo(genIUType v);
+
+	/// Return the power of two number which value is just higher the input value,
+	/// round up to a power of two.
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType nextPowerOfTwo(genIUType v);
+
+	/// Return the power of two number which value is just lower the input value,
+	/// round down to a power of two.
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType prevPowerOfTwo(genIUType v);
+
+	/// Return true if the 'Value' is a multiple of 'Multiple'.
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL bool isMultiple(genIUType v, genIUType Multiple);
+
+	/// Higher multiple number of Source.
+	///
+	/// @tparam genIUType Integer scalar or vector types.
+	///
+	/// @param v Source value to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType nextMultiple(genIUType v, genIUType Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam genIUType Integer scalar or vector types.
+	///
+	/// @param v Source value to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType prevMultiple(genIUType v, genIUType Multiple);
+
+	/// Returns the bit number of the Nth significant bit set to
+	/// 1 in the binary representation of value.
+	/// If value bitcount is less than the Nth significant bit, -1 will be returned.
+	///
+	/// @tparam genIUType Signed or unsigned integer scalar types.
+	///
+	/// @see ext_scalar_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL int findNSB(genIUType x, int significantBitCount);
+
+	/// @}
+} //namespace glm
+
+#include "scalar_integer.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.inl
new file mode 100644
index 000000000000..d416197e8667
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_integer.inl
@@ -0,0 +1,243 @@
+#include "../integer.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q, bool compute = false>
+	struct compute_ceilShift
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T)
+		{
+			return v;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct compute_ceilShift<L, T, Q, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T Shift)
+		{
+			return v | (v >> Shift);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q, bool isSigned = true>
+	struct compute_ceilPowerOfTwo
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			GLM_STATIC_ASSERT(!std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'ceilPowerOfTwo' only accept integer scalar or vector inputs");
+
+			vec<L, T, Q> const Sign(sign(x));
+
+			vec<L, T, Q> v(abs(x));
+
+			v = v - static_cast<T>(1);
+			v = v | (v >> static_cast<T>(1));
+			v = v | (v >> static_cast<T>(2));
+			v = v | (v >> static_cast<T>(4));
+			v = compute_ceilShift<L, T, Q, sizeof(T) >= 2>::call(v, 8);
+			v = compute_ceilShift<L, T, Q, sizeof(T) >= 4>::call(v, 16);
+			v = compute_ceilShift<L, T, Q, sizeof(T) >= 8>::call(v, 32);
+			return (v + static_cast<T>(1)) * Sign;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct compute_ceilPowerOfTwo<L, T, Q, false>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x)
+		{
+			GLM_STATIC_ASSERT(!std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'ceilPowerOfTwo' only accept integer scalar or vector inputs");
+
+			vec<L, T, Q> v(x);
+
+			v = v - static_cast<T>(1);
+			v = v | (v >> static_cast<T>(1));
+			v = v | (v >> static_cast<T>(2));
+			v = v | (v >> static_cast<T>(4));
+			v = compute_ceilShift<L, T, Q, sizeof(T) >= 2>::call(v, 8);
+			v = compute_ceilShift<L, T, Q, sizeof(T) >= 4>::call(v, 16);
+			v = compute_ceilShift<L, T, Q, sizeof(T) >= 8>::call(v, 32);
+			return v + static_cast<T>(1);
+		}
+	};
+
+	template<bool is_float, bool is_signed>
+	struct compute_ceilMultiple{};
+
+	template<>
+	struct compute_ceilMultiple<true, true>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if(Source > genType(0))
+				return Source + (Multiple - std::fmod(Source, Multiple));
+			else
+				return Source + std::fmod(-Source, Multiple);
+		}
+	};
+
+	template<>
+	struct compute_ceilMultiple<false, false>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			genType Tmp = Source - genType(1);
+			return Tmp + (Multiple - (Tmp % Multiple));
+		}
+	};
+
+	template<>
+	struct compute_ceilMultiple<false, true>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			assert(Multiple > genType(0));
+			if(Source > genType(0))
+			{
+				genType Tmp = Source - genType(1);
+				return Tmp + (Multiple - (Tmp % Multiple));
+			}
+			else
+				return Source + (-Source % Multiple);
+		}
+	};
+
+	template<bool is_float, bool is_signed>
+	struct compute_floorMultiple{};
+
+	template<>
+	struct compute_floorMultiple<true, true>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if(Source >= genType(0))
+				return Source - std::fmod(Source, Multiple);
+			else
+				return Source - std::fmod(Source, Multiple) - Multiple;
+		}
+	};
+
+	template<>
+	struct compute_floorMultiple<false, false>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if(Source >= genType(0))
+				return Source - Source % Multiple;
+			else
+			{
+				genType Tmp = Source + genType(1);
+				return Tmp - Tmp % Multiple - Multiple;
+			}
+		}
+	};
+
+	template<>
+	struct compute_floorMultiple<false, true>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if(Source >= genType(0))
+				return Source - Source % Multiple;
+			else
+			{
+				genType Tmp = Source + genType(1);
+				return Tmp - Tmp % Multiple - Multiple;
+			}
+		}
+	};
+}//namespace detail
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER bool isPowerOfTwo(genIUType Value)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'isPowerOfTwo' only accept integer inputs");
+
+		genIUType const Result = glm::abs(Value);
+		return !(Result & (Result - 1));
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType nextPowerOfTwo(genIUType value)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'nextPowerOfTwo' only accept integer inputs");
+
+		return detail::compute_ceilPowerOfTwo<1, genIUType, defaultp, std::numeric_limits<genIUType>::is_signed>::call(vec<1, genIUType, defaultp>(value)).x;
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType prevPowerOfTwo(genIUType value)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'prevPowerOfTwo' only accept integer inputs");
+
+		return isPowerOfTwo(value) ? value : static_cast<genIUType>(static_cast<genIUType>(1) << static_cast<genIUType>(findMSB(value)));
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER bool isMultiple(genIUType Value, genIUType Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'isMultiple' only accept integer inputs");
+
+		return isMultiple(vec<1, genIUType>(Value), vec<1, genIUType>(Multiple)).x;
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType nextMultiple(genIUType Source, genIUType Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'nextMultiple' only accept integer inputs");
+
+		return detail::compute_ceilMultiple<std::numeric_limits<genIUType>::is_iec559, std::numeric_limits<genIUType>::is_signed>::call(Source, Multiple);
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType prevMultiple(genIUType Source, genIUType Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'prevMultiple' only accept integer inputs");
+
+		return detail::compute_floorMultiple<std::numeric_limits<genIUType>::is_iec559, std::numeric_limits<genIUType>::is_signed>::call(Source, Multiple);
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER int findNSB(genIUType x, int significantBitCount)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'findNSB' only accept integer inputs");
+
+		if(bitCount(x) < significantBitCount)
+			return -1;
+
+		genIUType const One = static_cast<genIUType>(1);
+		int bitPos = 0;
+
+		genIUType key = x;
+		int nBitCount = significantBitCount;
+		int Step = sizeof(x) * 8 / 2;
+		while (key > One)
+		{
+			genIUType Mask = static_cast<genIUType>((One << Step) - One);
+			genIUType currentKey = key & Mask;
+			int currentBitCount = bitCount(currentKey);
+			if (nBitCount > currentBitCount)
+			{
+				nBitCount -= currentBitCount;
+				bitPos += Step;
+				key >>= static_cast<genIUType>(Step);
+			}
+			else
+			{
+				key = key & Mask;
+			}
+
+			Step >>= 1;
+		}
+
+		return static_cast<int>(bitPos);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.hpp
new file mode 100644
index 000000000000..18b85b72a404
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.hpp
@@ -0,0 +1,32 @@
+/// @ref ext_scalar_packing
+/// @file glm/ext/scalar_packing.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_scalar_packing GLM_EXT_scalar_packing
+/// @ingroup ext
+///
+/// Include <glm/ext/scalar_packing.hpp> to use the features of this extension.
+///
+/// This extension provides a set of function to convert scalar values to packed
+/// formats.
+
+#pragma once
+
+// Dependency:
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_packing extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_packing
+	/// @{
+
+
+	/// @}
+}// namespace glm
+
+#include "scalar_packing.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_packing.inl
new file mode 100644
index 000000000000..e69de29bb2d1
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.hpp
new file mode 100644
index 000000000000..1c7b81dde09a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.hpp
@@ -0,0 +1,135 @@
+/// @ref ext_scalar_reciprocal
+/// @file glm/ext/scalar_reciprocal.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_scalar_reciprocal GLM_EXT_scalar_reciprocal
+/// @ingroup ext
+///
+/// Include <glm/ext/scalar_reciprocal.hpp> to use the features of this extension.
+///
+/// Define secant, cosecant and cotangent functions.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_reciprocal extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_reciprocal
+	/// @{
+
+	/// Secant function.
+	/// hypotenuse / adjacent or 1 / cos(x)
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType sec(genType angle);
+
+	/// Cosecant function.
+	/// hypotenuse / opposite or 1 / sin(x)
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType csc(genType angle);
+
+	/// Cotangent function.
+	/// adjacent / opposite or 1 / tan(x)
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType cot(genType angle);
+
+	/// Inverse secant function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType asec(genType x);
+
+	/// Inverse cosecant function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acsc(genType x);
+
+	/// Inverse cotangent function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acot(genType x);
+
+	/// Secant hyperbolic function.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType sech(genType angle);
+
+	/// Cosecant hyperbolic function.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType csch(genType angle);
+
+	/// Cotangent hyperbolic function.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType coth(genType angle);
+
+	/// Inverse secant hyperbolic function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType asech(genType x);
+
+	/// Inverse cosecant hyperbolic function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acsch(genType x);
+
+	/// Inverse cotangent hyperbolic function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_scalar_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acoth(genType x);
+
+	/// @}
+}//namespace glm
+
+#include "scalar_reciprocal.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.inl
new file mode 100644
index 000000000000..0cd5f87b4e48
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_reciprocal.inl
@@ -0,0 +1,107 @@
+/// @ref ext_scalar_reciprocal
+
+#include "../trigonometric.hpp"
+#include <limits>
+
+namespace glm
+{
+	// sec
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType sec(genType angle)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'sec' only accept floating-point values");
+		return genType(1) / glm::cos(angle);
+	}
+
+	// csc
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType csc(genType angle)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'csc' only accept floating-point values");
+		return genType(1) / glm::sin(angle);
+	}
+
+	// cot
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType cot(genType angle)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'cot' only accept floating-point values");
+
+		genType const pi_over_2 = genType(3.1415926535897932384626433832795 / 2.0);
+		return glm::tan(pi_over_2 - angle);
+	}
+
+	// asec
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType asec(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'asec' only accept floating-point values");
+		return acos(genType(1) / x);
+	}
+
+	// acsc
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType acsc(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acsc' only accept floating-point values");
+		return asin(genType(1) / x);
+	}
+
+	// acot
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType acot(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acot' only accept floating-point values");
+
+		genType const pi_over_2 = genType(3.1415926535897932384626433832795 / 2.0);
+		return pi_over_2 - atan(x);
+	}
+
+	// sech
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType sech(genType angle)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'sech' only accept floating-point values");
+		return genType(1) / glm::cosh(angle);
+	}
+
+	// csch
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType csch(genType angle)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'csch' only accept floating-point values");
+		return genType(1) / glm::sinh(angle);
+	}
+
+	// coth
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType coth(genType angle)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'coth' only accept floating-point values");
+		return glm::cosh(angle) / glm::sinh(angle);
+	}
+
+	// asech
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType asech(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'asech' only accept floating-point values");
+		return acosh(genType(1) / x);
+	}
+
+	// acsch
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType acsch(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acsch' only accept floating-point values");
+		return asinh(genType(1) / x);
+	}
+
+	// acoth
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType acoth(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acoth' only accept floating-point values");
+		return atanh(genType(1) / x);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.hpp
new file mode 100644
index 000000000000..e84df1786824
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.hpp
@@ -0,0 +1,68 @@
+/// @ref ext_scalar_relational
+/// @file glm/ext/scalar_relational.hpp
+///
+/// @defgroup ext_scalar_relational GLM_EXT_scalar_relational
+/// @ingroup ext
+///
+/// Exposes comparison functions for scalar types that take a user defined epsilon values.
+///
+/// Include <glm/ext/scalar_relational.hpp> to use the features of this extension.
+///
+/// @see core_vector_relational
+/// @see ext_vector_relational
+/// @see ext_matrix_relational
+
+#pragma once
+
+// Dependencies
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_relational extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_relational
+	/// @{
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam genType Floating-point or integer scalar types
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool equal(genType const& x, genType const& y, genType const& epsilon);
+
+	/// Returns the component-wise comparison of |x - y| >= epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam genType Floating-point or integer scalar types
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool notEqual(genType const& x, genType const& y, genType const& epsilon);
+
+	/// Returns the component-wise comparison between two scalars in term of ULPs.
+	/// True if this expression is satisfied.
+	///
+	/// @param x First operand.
+	/// @param y Second operand.
+	/// @param ULPs Maximum difference in ULPs between the two operators to consider them equal.
+	///
+	/// @tparam genType Floating-point or integer scalar types
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool equal(genType const& x, genType const& y, int ULPs);
+
+	/// Returns the component-wise comparison between two scalars in term of ULPs.
+	/// True if this expression is not satisfied.
+	///
+	/// @param x First operand.
+	/// @param y Second operand.
+	/// @param ULPs Maximum difference in ULPs between the two operators to consider them not equal.
+	///
+	/// @tparam genType Floating-point or integer scalar types
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool notEqual(genType const& x, genType const& y, int ULPs);
+
+	/// @}
+}//namespace glm
+
+#include "scalar_relational.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.inl
new file mode 100644
index 000000000000..c85583ef5bbd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_relational.inl
@@ -0,0 +1,40 @@
+#include "../common.hpp"
+#include "../ext/scalar_int_sized.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+#include "../detail/type_float.hpp"
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool equal(genType const& x, genType const& y, genType const& epsilon)
+	{
+		return abs(x - y) <= epsilon;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool notEqual(genType const& x, genType const& y, genType const& epsilon)
+	{
+		return abs(x - y) > epsilon;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool equal(genType const& x, genType const& y, int MaxULPs)
+	{
+		detail::float_t<genType> const a(x);
+		detail::float_t<genType> const b(y);
+
+		// Different signs means they do not match.
+		if(a.negative() != b.negative())
+			return false;
+
+		// Find the difference in ULPs.
+		typename detail::float_t<genType>::int_type const DiffULPs = abs(a.i - b.i);
+		return DiffULPs <= MaxULPs;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR bool notEqual(genType const& x, genType const& y, int ULPs)
+	{
+		return !equal(x, y, ULPs);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_uint_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_uint_sized.hpp
new file mode 100644
index 000000000000..fd5267fad7c1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_uint_sized.hpp
@@ -0,0 +1,70 @@
+/// @ref ext_scalar_uint_sized
+/// @file glm/ext/scalar_uint_sized.hpp
+///
+/// @defgroup ext_scalar_uint_sized GLM_EXT_scalar_uint_sized
+/// @ingroup ext
+///
+/// Exposes sized unsigned integer scalar types.
+///
+/// Include <glm/ext/scalar_uint_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_int_sized
+
+#pragma once
+
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_uint_sized extension included")
+#endif
+
+namespace glm{
+namespace detail
+{
+#	if GLM_HAS_EXTENDED_INTEGER_TYPE
+		typedef std::uint8_t		uint8;
+		typedef std::uint16_t		uint16;
+		typedef std::uint32_t		uint32;
+#	else
+		typedef unsigned char		uint8;
+		typedef unsigned short		uint16;
+		typedef unsigned int		uint32;
+#endif
+
+	template<>
+	struct is_int<uint8>
+	{
+		enum test {value = ~0};
+	};
+
+	template<>
+	struct is_int<uint16>
+	{
+		enum test {value = ~0};
+	};
+
+	template<>
+	struct is_int<uint64>
+	{
+		enum test {value = ~0};
+	};
+}//namespace detail
+
+
+	/// @addtogroup ext_scalar_uint_sized
+	/// @{
+
+	/// 8 bit unsigned integer type.
+	typedef detail::uint8		uint8;
+
+	/// 16 bit unsigned integer type.
+	typedef detail::uint16		uint16;
+
+	/// 32 bit unsigned integer type.
+	typedef detail::uint32		uint32;
+
+	/// 64 bit unsigned integer type.
+	typedef detail::uint64		uint64;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.hpp
new file mode 100644
index 000000000000..6344d95bf23c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.hpp
@@ -0,0 +1,77 @@
+/// @ref ext_scalar_ulp
+/// @file glm/ext/scalar_ulp.hpp
+///
+/// @defgroup ext_scalar_ulp GLM_EXT_scalar_ulp
+/// @ingroup ext
+///
+/// Allow the measurement of the accuracy of a function against a reference
+/// implementation. This extension works on floating-point data and provide results
+/// in ULP.
+///
+/// Include <glm/ext/scalar_ulp.hpp> to use the features of this extension.
+///
+/// @see ext_vector_ulp
+/// @see ext_scalar_relational
+
+#pragma once
+
+// Dependencies
+#include "../ext/scalar_int_sized.hpp"
+#include "../common.hpp"
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_scalar_ulp extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_scalar_ulp
+	/// @{
+
+	/// Return the next ULP value(s) after the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see ext_scalar_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType nextFloat(genType x);
+
+	/// Return the previous ULP value(s) before the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see ext_scalar_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType prevFloat(genType x);
+
+	/// Return the value(s) ULP distance after the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see ext_scalar_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType nextFloat(genType x, int ULPs);
+
+	/// Return the value(s) ULP distance before the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see ext_scalar_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType prevFloat(genType x, int ULPs);
+
+	/// Return the distance in the number of ULP between 2 single-precision floating-point scalars.
+	///
+	/// @see ext_scalar_ulp
+	GLM_FUNC_DECL int floatDistance(float x, float y);
+
+	/// Return the distance in the number of ULP between 2 double-precision floating-point scalars.
+	///
+	/// @see ext_scalar_ulp
+	GLM_FUNC_DECL int64 floatDistance(double x, double y);
+
+	/// @}
+}//namespace glm
+
+#include "scalar_ulp.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.inl
new file mode 100644
index 000000000000..919403fd6ad3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/scalar_ulp.inl
@@ -0,0 +1,284 @@
+/// Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+///
+/// Developed at SunPro, a Sun Microsystems, Inc. business.
+/// Permission to use, copy, modify, and distribute this
+/// software is freely granted, provided that this notice
+/// is preserved.
+
+#include "../detail/type_float.hpp"
+#include "../ext/scalar_constants.hpp"
+#include <cmath>
+#include <cfloat>
+
+#if(GLM_COMPILER & GLM_COMPILER_VC)
+#	pragma warning(push)
+#	pragma warning(disable : 4127)
+#endif
+
+typedef union
+{
+	float value;
+	/* FIXME: Assumes 32 bit int.  */
+	unsigned int word;
+} ieee_float_shape_type;
+
+typedef union
+{
+	double value;
+	struct
+	{
+		int lsw;
+		int msw;
+	} parts;
+} ieee_double_shape_type;
+
+#define GLM_EXTRACT_WORDS(ix0,ix1,d)		\
+	do {									\
+		ieee_double_shape_type ew_u;		\
+		ew_u.value = (d);					\
+		(ix0) = ew_u.parts.msw;				\
+		(ix1) = ew_u.parts.lsw;				\
+	} while (0)
+
+#define GLM_GET_FLOAT_WORD(i,d)				\
+	do {									\
+		ieee_float_shape_type gf_u;			\
+		gf_u.value = (d);					\
+		(i) = gf_u.word;					\
+	} while (0)
+
+#define GLM_SET_FLOAT_WORD(d,i)				\
+	do {									\
+		ieee_float_shape_type sf_u;			\
+		sf_u.word = (i);					\
+		(d) = sf_u.value;					\
+	} while (0)
+
+#define GLM_INSERT_WORDS(d,ix0,ix1)			\
+	do {									\
+		ieee_double_shape_type iw_u;		\
+		iw_u.parts.msw = (ix0);				\
+		iw_u.parts.lsw = (ix1);				\
+		(d) = iw_u.value;					\
+	} while (0)
+
+namespace glm{
+namespace detail
+{
+	GLM_FUNC_QUALIFIER float nextafterf(float x, float y)
+	{
+		volatile float t;
+		int hx, hy, ix, iy;
+
+		GLM_GET_FLOAT_WORD(hx, x);
+		GLM_GET_FLOAT_WORD(hy, y);
+		ix = hx & 0x7fffffff;		// |x|
+		iy = hy & 0x7fffffff;		// |y|
+
+		if((ix > 0x7f800000) ||	// x is nan
+			(iy > 0x7f800000))	// y is nan
+			return x + y;
+		if(abs(y - x) <= epsilon<float>())
+			return y;		// x=y, return y
+		if(ix == 0)
+		{				// x == 0
+			GLM_SET_FLOAT_WORD(x, (hy & 0x80000000) | 1);// return +-minsubnormal
+			t = x * x;
+			if(abs(t - x) <= epsilon<float>())
+				return t;
+			else
+				return x;	// raise underflow flag
+		}
+		if(hx >= 0)
+		{						// x > 0
+			if(hx > hy)			// x > y, x -= ulp
+				hx -= 1;
+			else				// x < y, x += ulp
+				hx += 1;
+		}
+		else
+		{						// x < 0
+			if(hy >= 0 || hx > hy)	// x < y, x -= ulp
+				hx -= 1;
+			else				// x > y, x += ulp
+				hx += 1;
+		}
+		hy = hx & 0x7f800000;
+		if(hy >= 0x7f800000)
+			return x + x;  		// overflow
+		if(hy < 0x00800000)		// underflow
+		{
+			t = x * x;
+			if(abs(t - x) > epsilon<float>())
+			{					// raise underflow flag
+				GLM_SET_FLOAT_WORD(y, hx);
+				return y;
+			}
+		}
+		GLM_SET_FLOAT_WORD(x, hx);
+		return x;
+	}
+
+	GLM_FUNC_QUALIFIER double nextafter(double x, double y)
+	{
+		volatile double t;
+		int hx, hy, ix, iy;
+		unsigned int lx, ly;
+
+		GLM_EXTRACT_WORDS(hx, lx, x);
+		GLM_EXTRACT_WORDS(hy, ly, y);
+		ix = hx & 0x7fffffff;								// |x|
+		iy = hy & 0x7fffffff;								// |y|
+
+		if(((ix >= 0x7ff00000) && ((ix - 0x7ff00000) | lx) != 0) ||	// x is nan
+			((iy >= 0x7ff00000) && ((iy - 0x7ff00000) | ly) != 0))	// y is nan
+			return x + y;
+		if(abs(y - x) <= epsilon<double>())
+			return y;									// x=y, return y
+		if((ix | lx) == 0)
+		{													// x == 0
+			GLM_INSERT_WORDS(x, hy & 0x80000000, 1);		// return +-minsubnormal
+			t = x * x;
+			if(abs(t - x) <= epsilon<double>())
+				return t;
+			else
+				return x;   // raise underflow flag
+		}
+		if(hx >= 0) {                             // x > 0
+			if(hx > hy || ((hx == hy) && (lx > ly))) {    // x > y, x -= ulp
+				if(lx == 0) hx -= 1;
+				lx -= 1;
+			}
+			else {                            // x < y, x += ulp
+				lx += 1;
+				if(lx == 0) hx += 1;
+			}
+		}
+		else {                                // x < 0
+			if(hy >= 0 || hx > hy || ((hx == hy) && (lx > ly))){// x < y, x -= ulp
+				if(lx == 0) hx -= 1;
+				lx -= 1;
+			}
+			else {                            // x > y, x += ulp
+				lx += 1;
+				if(lx == 0) hx += 1;
+			}
+		}
+		hy = hx & 0x7ff00000;
+		if(hy >= 0x7ff00000)
+			return x + x;			// overflow
+		if(hy < 0x00100000)
+		{						// underflow
+			t = x * x;
+			if(abs(t - x) > epsilon<double>())
+			{					// raise underflow flag
+				GLM_INSERT_WORDS(y, hx, lx);
+				return y;
+			}
+		}
+		GLM_INSERT_WORDS(x, hx, lx);
+		return x;
+	}
+}//namespace detail
+}//namespace glm
+
+#if(GLM_COMPILER & GLM_COMPILER_VC)
+#	pragma warning(pop)
+#endif
+
+namespace glm
+{
+	template<>
+	GLM_FUNC_QUALIFIER float nextFloat(float x)
+	{
+#		if GLM_HAS_CXX11_STL
+			return std::nextafter(x, std::numeric_limits<float>::max());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+			return detail::nextafterf(x, FLT_MAX);
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+			return __builtin_nextafterf(x, FLT_MAX);
+#		else
+			return nextafterf(x, FLT_MAX);
+#		endif
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER double nextFloat(double x)
+	{
+#		if GLM_HAS_CXX11_STL
+			return std::nextafter(x, std::numeric_limits<double>::max());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+			return detail::nextafter(x, std::numeric_limits<double>::max());
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+			return __builtin_nextafter(x, DBL_MAX);
+#		else
+			return nextafter(x, DBL_MAX);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T nextFloat(T x, int ULPs)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'next_float' only accept floating-point input");
+		assert(ULPs >= 0);
+
+		T temp = x;
+		for(int i = 0; i < ULPs; ++i)
+			temp = nextFloat(temp);
+		return temp;
+	}
+
+	GLM_FUNC_QUALIFIER float prevFloat(float x)
+	{
+#		if GLM_HAS_CXX11_STL
+			return std::nextafter(x, std::numeric_limits<float>::min());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+			return detail::nextafterf(x, FLT_MIN);
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+			return __builtin_nextafterf(x, FLT_MIN);
+#		else
+			return nextafterf(x, FLT_MIN);
+#		endif
+	}
+
+	GLM_FUNC_QUALIFIER double prevFloat(double x)
+	{
+#		if GLM_HAS_CXX11_STL
+			return std::nextafter(x, std::numeric_limits<double>::min());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+			return _nextafter(x, DBL_MIN);
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+			return __builtin_nextafter(x, DBL_MIN);
+#		else
+			return nextafter(x, DBL_MIN);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T prevFloat(T x, int ULPs)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'prev_float' only accept floating-point input");
+		assert(ULPs >= 0);
+
+		T temp = x;
+		for(int i = 0; i < ULPs; ++i)
+			temp = prevFloat(temp);
+		return temp;
+	}
+
+	GLM_FUNC_QUALIFIER int floatDistance(float x, float y)
+	{
+		detail::float_t<float> const a(x);
+		detail::float_t<float> const b(y);
+
+		return abs(a.i - b.i);
+	}
+
+	GLM_FUNC_QUALIFIER int64 floatDistance(double x, double y)
+	{
+		detail::float_t<double> const a(x);
+		detail::float_t<double> const b(y);
+
+		return abs(a.i - b.i);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1.hpp
new file mode 100644
index 000000000000..002c3202adf0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1.hpp
@@ -0,0 +1,30 @@
+/// @ref ext_vector_bool1
+/// @file glm/ext/vector_bool1.hpp
+///
+/// @defgroup ext_vector_bool1 GLM_EXT_vector_bool1
+/// @ingroup ext
+///
+/// Exposes bvec1 vector type.
+///
+/// Include <glm/ext/vector_bool1.hpp> to use the features of this extension.
+///
+/// @see ext_vector_bool1_precision extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_bool1 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_bool1
+	/// @{
+
+	/// 1 components vector of boolean.
+	typedef vec<1, bool, defaultp>		bvec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1_precision.hpp
new file mode 100644
index 000000000000..e62d3cfb5fd4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool1_precision.hpp
@@ -0,0 +1,34 @@
+/// @ref ext_vector_bool1_precision
+/// @file glm/ext/vector_bool1_precision.hpp
+///
+/// @defgroup ext_vector_bool1_precision GLM_EXT_vector_bool1_precision
+/// @ingroup ext
+///
+/// Exposes highp_bvec1, mediump_bvec1 and lowp_bvec1 types.
+///
+/// Include <glm/ext/vector_bool1_precision.hpp> to use the features of this extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_bool1_precision extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_bool1_precision
+	/// @{
+
+	/// 1 component vector of bool values.
+	typedef vec<1, bool, highp>			highp_bvec1;
+
+	/// 1 component vector of bool values.
+	typedef vec<1, bool, mediump>		mediump_bvec1;
+
+	/// 1 component vector of bool values.
+	typedef vec<1, bool, lowp>			lowp_bvec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2.hpp
new file mode 100644
index 000000000000..52288b75c697
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_bool2.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 2 components vector of boolean.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<2, bool, defaultp>		bvec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2_precision.hpp
new file mode 100644
index 000000000000..43709332c629
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_bool2_precision.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 2 components vector of high qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, bool, highp>		highp_bvec2;
+
+	/// 2 components vector of medium qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, bool, mediump>	mediump_bvec2;
+
+	/// 2 components vector of low qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, bool, lowp>		lowp_bvec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3.hpp
new file mode 100644
index 000000000000..90a0b7ea5ac0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_bool3.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 3 components vector of boolean.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<3, bool, defaultp>		bvec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3_precision.hpp
new file mode 100644
index 000000000000..89cd2d3207a1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool3_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_bool3_precision.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 3 components vector of high qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, bool, highp>		highp_bvec3;
+
+	/// 3 components vector of medium qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, bool, mediump>	mediump_bvec3;
+
+	/// 3 components vector of low qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, bool, lowp>		lowp_bvec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4.hpp
new file mode 100644
index 000000000000..18aa71bd0f49
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_bool4.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 4 components vector of boolean.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<4, bool, defaultp>		bvec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4_precision.hpp
new file mode 100644
index 000000000000..79786e54206b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_bool4_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_bool4_precision.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 4 components vector of high qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, bool, highp>		highp_bvec4;
+
+	/// 4 components vector of medium qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, bool, mediump>	mediump_bvec4;
+
+	/// 4 components vector of low qualifier bool numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, bool, lowp>		lowp_bvec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.hpp
new file mode 100644
index 000000000000..c0a2858cc288
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.hpp
@@ -0,0 +1,228 @@
+/// @ref ext_vector_common
+/// @file glm/ext/vector_common.hpp
+///
+/// @defgroup ext_vector_common GLM_EXT_vector_common
+/// @ingroup ext
+///
+/// Exposes min and max functions for 3 to 4 vector parameters.
+///
+/// Include <glm/ext/vector_common.hpp> to use the features of this extension.
+///
+/// @see core_common
+/// @see ext_scalar_common
+
+#pragma once
+
+// Dependency:
+#include "../ext/scalar_common.hpp"
+#include "../common.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_common extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_common
+	/// @{
+
+	/// Return the minimum component-wise values of 3 inputs
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c);
+
+	/// Return the minimum component-wise values of 4 inputs
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c, vec<L, T, Q> const& d);
+
+	/// Return the maximum component-wise values of 3 inputs
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& z);
+
+	/// Return the maximum component-wise values of 4 inputs
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> max( vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& z, vec<L, T, Q> const& w);
+
+	/// Returns y if y < x; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmin(vec<L, T, Q> const& x, T y);
+
+	/// Returns y if y < x; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmin(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns y if y < x; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmin(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c);
+
+	/// Returns y if y < x; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmin">std::fmin documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmin(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c, vec<L, T, Q> const& d);
+
+	/// Returns y if x < y; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmax(vec<L, T, Q> const& a, T b);
+
+	/// Returns y if x < y; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmax(vec<L, T, Q> const& a, vec<L, T, Q> const& b);
+
+	/// Returns y if x < y; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmax(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c);
+
+	/// Returns y if x < y; otherwise, it returns x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://en.cppreference.com/w/cpp/numeric/math/fmax">std::fmax documentation</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmax(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c, vec<L, T, Q> const& d);
+
+	/// Returns min(max(x, minVal), maxVal) for each component in x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_common
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fclamp(vec<L, T, Q> const& x, T minVal, T maxVal);
+
+	/// Returns min(max(x, minVal), maxVal) for each component in x. If one of the two arguments is NaN, the value of the other argument is returned.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_common
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fclamp(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal);
+
+	/// Simulate GL_CLAMP OpenGL wrap mode
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_common extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> clamp(vec<L, T, Q> const& Texcoord);
+
+	/// Simulate GL_REPEAT OpenGL wrap mode
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_common extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> repeat(vec<L, T, Q> const& Texcoord);
+
+	/// Simulate GL_MIRRORED_REPEAT OpenGL wrap mode
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_common extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> mirrorClamp(vec<L, T, Q> const& Texcoord);
+
+	/// Simulate GL_MIRROR_REPEAT OpenGL wrap mode
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_common extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> mirrorRepeat(vec<L, T, Q> const& Texcoord);
+
+	/// Returns a value equal to the nearest integer to x.
+	/// The fraction 0.5 will round in a direction chosen by the
+	/// implementation, presumably the direction that is fastest.
+	///
+	/// @param x The values of the argument must be greater or equal to zero.
+	/// @tparam T floating point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/round.xml">GLSL round man page</a>
+	/// @see ext_vector_common extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> iround(vec<L, T, Q> const& x);
+
+	/// Returns a value equal to the nearest integer to x.
+	/// The fraction 0.5 will round in a direction chosen by the
+	/// implementation, presumably the direction that is fastest.
+	///
+	/// @param x The values of the argument must be greater or equal to zero.
+	/// @tparam T floating point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/round.xml">GLSL round man page</a>
+	/// @see ext_vector_common extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, uint, Q> uround(vec<L, T, Q> const& x);
+
+	/// @}
+}//namespace glm
+
+#include "vector_common.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.inl
new file mode 100644
index 000000000000..67817fc554b0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_common.inl
@@ -0,0 +1,147 @@
+#include "../detail/_vectorize.hpp"
+
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& z)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer, "'min' only accept floating-point or integer inputs");
+		return glm::min(glm::min(x, y), z);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> min(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& z, vec<L, T, Q> const& w)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer, "'min' only accept floating-point or integer inputs");
+		return glm::min(glm::min(x, y), glm::min(z, w));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& z)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer, "'max' only accept floating-point or integer inputs");
+		return glm::max(glm::max(x, y), z);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, T, Q> max(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& z, vec<L, T, Q> const& w)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || std::numeric_limits<T>::is_integer, "'max' only accept floating-point or integer inputs");
+		return glm::max(glm::max(x, y), glm::max(z, w));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmin(vec<L, T, Q> const& a, T b)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmin' only accept floating-point inputs");
+		return detail::functor2<vec, L, T, Q>::call(fmin, a, vec<L, T, Q>(b));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmin(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmin' only accept floating-point inputs");
+		return detail::functor2<vec, L, T, Q>::call(fmin, a, b);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmin(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmin' only accept floating-point inputs");
+		return fmin(fmin(a, b), c);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmin(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c, vec<L, T, Q> const& d)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmin' only accept floating-point inputs");
+		return fmin(fmin(a, b), fmin(c, d));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmax(vec<L, T, Q> const& a, T b)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmax' only accept floating-point inputs");
+		return detail::functor2<vec, L, T, Q>::call(fmax, a, vec<L, T, Q>(b));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmax(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmax' only accept floating-point inputs");
+		return detail::functor2<vec, L, T, Q>::call(fmax, a, b);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmax(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmax' only accept floating-point inputs");
+		return fmax(fmax(a, b), c);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmax(vec<L, T, Q> const& a, vec<L, T, Q> const& b, vec<L, T, Q> const& c, vec<L, T, Q> const& d)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'fmax' only accept floating-point inputs");
+		return fmax(fmax(a, b), fmax(c, d));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fclamp(vec<L, T, Q> const& x, T minVal, T maxVal)
+	{
+		return fmin(fmax(x, vec<L, T, Q>(minVal)), vec<L, T, Q>(maxVal));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fclamp(vec<L, T, Q> const& x, vec<L, T, Q> const& minVal, vec<L, T, Q> const& maxVal)
+	{
+		return fmin(fmax(x, minVal), maxVal);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> clamp(vec<L, T, Q> const& Texcoord)
+	{
+		return glm::clamp(Texcoord, vec<L, T, Q>(0), vec<L, T, Q>(1));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> repeat(vec<L, T, Q> const& Texcoord)
+	{
+		return glm::fract(Texcoord);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> mirrorClamp(vec<L, T, Q> const& Texcoord)
+	{
+		return glm::fract(glm::abs(Texcoord));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> mirrorRepeat(vec<L, T, Q> const& Texcoord)
+	{
+		vec<L, T, Q> const Abs = glm::abs(Texcoord);
+		vec<L, T, Q> const Clamp = glm::mod(glm::floor(Abs), vec<L, T, Q>(2));
+		vec<L, T, Q> const Floor = glm::floor(Abs);
+		vec<L, T, Q> const Rest = Abs - Floor;
+		vec<L, T, Q> const Mirror = Clamp + Rest;
+		return mix(Rest, vec<L, T, Q>(1) - Rest, glm::greaterThanEqual(Mirror, vec<L, T, Q>(1)));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> iround(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'iround' only accept floating-point inputs");
+		assert(all(lessThanEqual(vec<L, T, Q>(0), x)));
+
+		return vec<L, int, Q>(x + static_cast<T>(0.5));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, uint, Q> uround(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'uround' only accept floating-point inputs");
+		assert(all(lessThanEqual(vec<L, T, Q>(0), x)));
+
+		return vec<L, uint, Q>(x + static_cast<T>(0.5));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1.hpp
new file mode 100644
index 000000000000..388266774d8b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1.hpp
@@ -0,0 +1,31 @@
+/// @ref ext_vector_double1
+/// @file glm/ext/vector_double1.hpp
+///
+/// @defgroup ext_vector_double1 GLM_EXT_vector_double1
+/// @ingroup ext
+///
+/// Exposes double-precision floating point vector type with one component.
+///
+/// Include <glm/ext/vector_double1.hpp> to use the features of this extension.
+///
+/// @see ext_vector_double1_precision extension.
+/// @see ext_vector_float1 extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_double1 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_double1
+	/// @{
+
+	/// 1 components vector of double-precision floating-point numbers.
+	typedef vec<1, double, defaultp>		dvec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1_precision.hpp
new file mode 100644
index 000000000000..1d4719595481
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double1_precision.hpp
@@ -0,0 +1,36 @@
+/// @ref ext_vector_double1_precision
+/// @file glm/ext/vector_double1_precision.hpp
+///
+/// @defgroup ext_vector_double1_precision GLM_EXT_vector_double1_precision
+/// @ingroup ext
+///
+/// Exposes highp_dvec1, mediump_dvec1 and lowp_dvec1 types.
+///
+/// Include <glm/ext/vector_double1_precision.hpp> to use the features of this extension.
+///
+/// @see ext_vector_double1
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_double1_precision extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_double1_precision
+	/// @{
+
+	/// 1 component vector of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<1, double, highp>		highp_dvec1;
+
+	/// 1 component vector of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<1, double, mediump>		mediump_dvec1;
+
+	/// 1 component vector of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<1, double, lowp>		lowp_dvec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2.hpp
new file mode 100644
index 000000000000..60e357750b67
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_double2.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 2 components vector of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<2, double, defaultp>		dvec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2_precision.hpp
new file mode 100644
index 000000000000..fa53940f6bbe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_double2_precision.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 2 components vector of high double-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, double, highp>		highp_dvec2;
+
+	/// 2 components vector of medium double-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, double, mediump>		mediump_dvec2;
+
+	/// 2 components vector of low double-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, double, lowp>		lowp_dvec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3.hpp
new file mode 100644
index 000000000000..6dfe4c675b5d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_double3.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 3 components vector of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<3, double, defaultp>		dvec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3_precision.hpp
new file mode 100644
index 000000000000..a8cfa37a8c78
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double3_precision.hpp
@@ -0,0 +1,34 @@
+/// @ref core
+/// @file glm/ext/vector_double3_precision.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 3 components vector of high double-qualifier floating-point numbers.
+	/// There is no guarantee on the actual qualifier.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, double, highp>		highp_dvec3;
+
+	/// 3 components vector of medium double-qualifier floating-point numbers.
+	/// There is no guarantee on the actual qualifier.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, double, mediump>		mediump_dvec3;
+
+	/// 3 components vector of low double-qualifier floating-point numbers.
+	/// There is no guarantee on the actual qualifier.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, double, lowp>		lowp_dvec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4.hpp
new file mode 100644
index 000000000000..87f225f64d4a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_double4.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 4 components vector of double-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<4, double, defaultp>		dvec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4_precision.hpp
new file mode 100644
index 000000000000..09cafa1ebafd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_double4_precision.hpp
@@ -0,0 +1,35 @@
+/// @ref core
+/// @file glm/ext/vector_double4_precision.hpp
+
+#pragma once
+#include "../detail/setup.hpp"
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 4 components vector of high double-qualifier floating-point numbers.
+	/// There is no guarantee on the actual qualifier.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, double, highp>		highp_dvec4;
+
+	/// 4 components vector of medium double-qualifier floating-point numbers.
+	/// There is no guarantee on the actual qualifier.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, double, mediump>		mediump_dvec4;
+
+	/// 4 components vector of low double-qualifier floating-point numbers.
+	/// There is no guarantee on the actual qualifier.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, double, lowp>		lowp_dvec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1.hpp
new file mode 100644
index 000000000000..28acc2c9ca59
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1.hpp
@@ -0,0 +1,31 @@
+/// @ref ext_vector_float1
+/// @file glm/ext/vector_float1.hpp
+///
+/// @defgroup ext_vector_float1 GLM_EXT_vector_float1
+/// @ingroup ext
+///
+/// Exposes single-precision floating point vector type with one component.
+///
+/// Include <glm/ext/vector_float1.hpp> to use the features of this extension.
+///
+/// @see ext_vector_float1_precision extension.
+/// @see ext_vector_double1 extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_float1 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_float1
+	/// @{
+
+	/// 1 components vector of single-precision floating-point numbers.
+	typedef vec<1, float, defaultp>		vec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1_precision.hpp
new file mode 100644
index 000000000000..6e8dad8d17c8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float1_precision.hpp
@@ -0,0 +1,36 @@
+/// @ref ext_vector_float1_precision
+/// @file glm/ext/vector_float1_precision.hpp
+///
+/// @defgroup ext_vector_float1_precision GLM_EXT_vector_float1_precision
+/// @ingroup ext
+///
+/// Exposes highp_vec1, mediump_vec1 and lowp_vec1 types.
+///
+/// Include <glm/ext/vector_float1_precision.hpp> to use the features of this extension.
+///
+/// @see ext_vector_float1 extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_float1_precision extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_float1_precision
+	/// @{
+
+	/// 1 component vector of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<1, float, highp>		highp_vec1;
+
+	/// 1 component vector of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<1, float, mediump>		mediump_vec1;
+
+	/// 1 component vector of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<1, float, lowp>			lowp_vec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2.hpp
new file mode 100644
index 000000000000..d31545dcc966
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_float2.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 2 components vector of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<2, float, defaultp>	vec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2_precision.hpp
new file mode 100644
index 000000000000..23c0820d0ae8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float2_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_float2_precision.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 2 components vector of high single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, float, highp>		highp_vec2;
+
+	/// 2 components vector of medium single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, float, mediump>		mediump_vec2;
+
+	/// 2 components vector of low single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<2, float, lowp>			lowp_vec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3.hpp
new file mode 100644
index 000000000000..cd79a62004e4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_float3.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 3 components vector of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<3, float, defaultp>		vec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3_precision.hpp
new file mode 100644
index 000000000000..be640b531683
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float3_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_float3_precision.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 3 components vector of high single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, float, highp>		highp_vec3;
+
+	/// 3 components vector of medium single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, float, mediump>		mediump_vec3;
+
+	/// 3 components vector of low single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<3, float, lowp>			lowp_vec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4.hpp
new file mode 100644
index 000000000000..d84adcc22fd2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_float4.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 4 components vector of single-precision floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<4, float, defaultp>		vec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4_precision.hpp
new file mode 100644
index 000000000000..aede83882e55
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_float4_precision.hpp
@@ -0,0 +1,31 @@
+/// @ref core
+/// @file glm/ext/vector_float4_precision.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector_precision
+	/// @{
+
+	/// 4 components vector of high single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, float, highp>		highp_vec4;
+
+	/// 4 components vector of medium single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, float, mediump>		mediump_vec4;
+
+	/// 4 components vector of low single-qualifier floating-point numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.7.2 Precision Qualifier</a>
+	typedef vec<4, float, lowp>			lowp_vec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1.hpp
new file mode 100644
index 000000000000..dc8603891a99
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1.hpp
@@ -0,0 +1,32 @@
+/// @ref ext_vector_int1
+/// @file glm/ext/vector_int1.hpp
+///
+/// @defgroup ext_vector_int1 GLM_EXT_vector_int1
+/// @ingroup ext
+///
+/// Exposes ivec1 vector type.
+///
+/// Include <glm/ext/vector_int1.hpp> to use the features of this extension.
+///
+/// @see ext_vector_uint1 extension.
+/// @see ext_vector_int1_precision extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_int1 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_int1
+	/// @{
+
+	/// 1 component vector of signed integer numbers.
+	typedef vec<1, int, defaultp>			ivec1;
+
+	/// @}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1_sized.hpp
new file mode 100644
index 000000000000..de0d4cf82e63
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int1_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_int1_sized
+/// @file glm/ext/vector_int1_sized.hpp
+///
+/// @defgroup ext_vector_int1_sized GLM_EXT_vector_int1_sized
+/// @ingroup ext
+///
+/// Exposes sized signed integer vector types.
+///
+/// Include <glm/ext/vector_int1_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_int_sized
+/// @see ext_vector_uint1_sized
+
+#pragma once
+
+#include "../ext/vector_int1.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_int1_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_int1_sized
+	/// @{
+
+	/// 8 bit signed integer vector of 1 component type.
+	///
+	/// @see ext_vector_int1_sized
+	typedef vec<1, int8, defaultp>	i8vec1;
+
+	/// 16 bit signed integer vector of 1 component type.
+	///
+	/// @see ext_vector_int1_sized
+	typedef vec<1, int16, defaultp>	i16vec1;
+
+	/// 32 bit signed integer vector of 1 component type.
+	///
+	/// @see ext_vector_int1_sized
+	typedef vec<1, int32, defaultp>	i32vec1;
+
+	/// 64 bit signed integer vector of 1 component type.
+	///
+	/// @see ext_vector_int1_sized
+	typedef vec<1, int64, defaultp>	i64vec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2.hpp
new file mode 100644
index 000000000000..aef803e91b73
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_int2.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 2 components vector of signed integer numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<2, int, defaultp>		ivec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2_sized.hpp
new file mode 100644
index 000000000000..1fd57eef310d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int2_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_int2_sized
+/// @file glm/ext/vector_int2_sized.hpp
+///
+/// @defgroup ext_vector_int2_sized GLM_EXT_vector_int2_sized
+/// @ingroup ext
+///
+/// Exposes sized signed integer vector of 2 components type.
+///
+/// Include <glm/ext/vector_int2_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_int_sized
+/// @see ext_vector_uint2_sized
+
+#pragma once
+
+#include "../ext/vector_int2.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_int2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_int2_sized
+	/// @{
+
+	/// 8 bit signed integer vector of 2 components type.
+	///
+	/// @see ext_vector_int2_sized
+	typedef vec<2, int8, defaultp>		i8vec2;
+
+	/// 16 bit signed integer vector of 2 components type.
+	///
+	/// @see ext_vector_int2_sized
+	typedef vec<2, int16, defaultp>		i16vec2;
+
+	/// 32 bit signed integer vector of 2 components type.
+	///
+	/// @see ext_vector_int2_sized
+	typedef vec<2, int32, defaultp>		i32vec2;
+
+	/// 64 bit signed integer vector of 2 components type.
+	///
+	/// @see ext_vector_int2_sized
+	typedef vec<2, int64, defaultp>		i64vec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3.hpp
new file mode 100644
index 000000000000..4767e61e88c0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_int3.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 3 components vector of signed integer numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<3, int, defaultp>		ivec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3_sized.hpp
new file mode 100644
index 000000000000..085a3febbffd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int3_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_int3_sized
+/// @file glm/ext/vector_int3_sized.hpp
+///
+/// @defgroup ext_vector_int3_sized GLM_EXT_vector_int3_sized
+/// @ingroup ext
+///
+/// Exposes sized signed integer vector of 3 components type.
+///
+/// Include <glm/ext/vector_int3_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_int_sized
+/// @see ext_vector_uint3_sized
+
+#pragma once
+
+#include "../ext/vector_int3.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_int3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_int3_sized
+	/// @{
+
+	/// 8 bit signed integer vector of 3 components type.
+	///
+	/// @see ext_vector_int3_sized
+	typedef vec<3, int8, defaultp>		i8vec3;
+
+	/// 16 bit signed integer vector of 3 components type.
+	///
+	/// @see ext_vector_int3_sized
+	typedef vec<3, int16, defaultp>		i16vec3;
+
+	/// 32 bit signed integer vector of 3 components type.
+	///
+	/// @see ext_vector_int3_sized
+	typedef vec<3, int32, defaultp>		i32vec3;
+
+	/// 64 bit signed integer vector of 3 components type.
+	///
+	/// @see ext_vector_int3_sized
+	typedef vec<3, int64, defaultp>		i64vec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4.hpp
new file mode 100644
index 000000000000..bb23adf706c2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_int4.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 4 components vector of signed integer numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<4, int, defaultp>		ivec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4_sized.hpp
new file mode 100644
index 000000000000..c63d46540b3e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_int4_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_int4_sized
+/// @file glm/ext/vector_int4_sized.hpp
+///
+/// @defgroup ext_vector_int4_sized GLM_EXT_vector_int4_sized
+/// @ingroup ext
+///
+/// Exposes sized signed integer vector of 4 components type.
+///
+/// Include <glm/ext/vector_int4_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_int_sized
+/// @see ext_vector_uint4_sized
+
+#pragma once
+
+#include "../ext/vector_int4.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_int4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_int4_sized
+	/// @{
+
+	/// 8 bit signed integer vector of 4 components type.
+	///
+	/// @see ext_vector_int4_sized
+	typedef vec<4, int8, defaultp>		i8vec4;
+
+	/// 16 bit signed integer vector of 4 components type.
+	///
+	/// @see ext_vector_int4_sized
+	typedef vec<4, int16, defaultp>		i16vec4;
+
+	/// 32 bit signed integer vector of 4 components type.
+	///
+	/// @see ext_vector_int4_sized
+	typedef vec<4, int32, defaultp>		i32vec4;
+
+	/// 64 bit signed integer vector of 4 components type.
+	///
+	/// @see ext_vector_int4_sized
+	typedef vec<4, int64, defaultp>		i64vec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.hpp
new file mode 100644
index 000000000000..1304dd8d660f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.hpp
@@ -0,0 +1,149 @@
+/// @ref ext_vector_integer
+/// @file glm/ext/vector_integer.hpp
+///
+/// @see core (dependence)
+/// @see ext_vector_integer (dependence)
+///
+/// @defgroup ext_vector_integer GLM_EXT_vector_integer
+/// @ingroup ext
+///
+/// Include <glm/ext/vector_integer.hpp> to use the features of this extension.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/_vectorize.hpp"
+#include "../vector_relational.hpp"
+#include "../common.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_integer extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_integer
+	/// @{
+
+	/// Return true if the value is a power of two number.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> isPowerOfTwo(vec<L, T, Q> const& v);
+
+	/// Return the power of two number which value is just higher the input value,
+	/// round up to a power of two.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> nextPowerOfTwo(vec<L, T, Q> const& v);
+
+	/// Return the power of two number which value is just lower the input value,
+	/// round down to a power of two.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prevPowerOfTwo(vec<L, T, Q> const& v);
+
+	/// Return true if the 'Value' is a multiple of 'Multiple'.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> isMultiple(vec<L, T, Q> const& v, T Multiple);
+
+	/// Return true if the 'Value' is a multiple of 'Multiple'.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> isMultiple(vec<L, T, Q> const& v, vec<L, T, Q> const& Multiple);
+
+	/// Higher multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> nextMultiple(vec<L, T, Q> const& v, T Multiple);
+
+	/// Higher multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> nextMultiple(vec<L, T, Q> const& v, vec<L, T, Q> const& Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prevMultiple(vec<L, T, Q> const& v, T Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed or unsigned integer scalar types.
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prevMultiple(vec<L, T, Q> const& v, vec<L, T, Q> const& Multiple);
+
+	/// Returns the bit number of the Nth significant bit set to
+	/// 1 in the binary representation of value.
+	/// If value bitcount is less than the Nth significant bit, -1 will be returned.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar types.
+	///
+	/// @see ext_vector_integer
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> findNSB(vec<L, T, Q> const& Source, vec<L, int, Q> SignificantBitCount);
+
+	/// @}
+} //namespace glm
+
+#include "vector_integer.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.inl
new file mode 100644
index 000000000000..cefb132e237a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_integer.inl
@@ -0,0 +1,85 @@
+#include "scalar_integer.hpp"
+
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> isPowerOfTwo(vec<L, T, Q> const& Value)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'isPowerOfTwo' only accept integer inputs");
+
+		vec<L, T, Q> const Result(abs(Value));
+		return equal(Result & (Result - vec<L, T, Q>(1)), vec<L, T, Q>(0));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> nextPowerOfTwo(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'nextPowerOfTwo' only accept integer inputs");
+
+		return detail::compute_ceilPowerOfTwo<L, T, Q, std::numeric_limits<T>::is_signed>::call(v);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prevPowerOfTwo(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'prevPowerOfTwo' only accept integer inputs");
+
+		return detail::functor1<vec, L, T, T, Q>::call(prevPowerOfTwo, v);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> isMultiple(vec<L, T, Q> const& Value, T Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'isMultiple' only accept integer inputs");
+
+		return equal(Value % Multiple, vec<L, T, Q>(0));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> isMultiple(vec<L, T, Q> const& Value, vec<L, T, Q> const& Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'isMultiple' only accept integer inputs");
+
+		return equal(Value % Multiple, vec<L, T, Q>(0));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> nextMultiple(vec<L, T, Q> const& Source, T Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'nextMultiple' only accept integer inputs");
+
+		return detail::functor2<vec, L, T, Q>::call(nextMultiple, Source, vec<L, T, Q>(Multiple));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> nextMultiple(vec<L, T, Q> const& Source, vec<L, T, Q> const& Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'nextMultiple' only accept integer inputs");
+
+		return detail::functor2<vec, L, T, Q>::call(nextMultiple, Source, Multiple);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prevMultiple(vec<L, T, Q> const& Source, T Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'prevMultiple' only accept integer inputs");
+
+		return detail::functor2<vec, L, T, Q>::call(prevMultiple, Source, vec<L, T, Q>(Multiple));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prevMultiple(vec<L, T, Q> const& Source, vec<L, T, Q> const& Multiple)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'prevMultiple' only accept integer inputs");
+
+		return detail::functor2<vec, L, T, Q>::call(prevMultiple, Source, Multiple);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> findNSB(vec<L, T, Q> const& Source, vec<L, int, Q> SignificantBitCount)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'findNSB' only accept integer inputs");
+
+		return detail::functor2_vec_int<L, T, Q>::call(findNSB, Source, SignificantBitCount);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.hpp
new file mode 100644
index 000000000000..76e5d0cc6c2f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.hpp
@@ -0,0 +1,32 @@
+/// @ref ext_vector_packing
+/// @file glm/ext/vector_packing.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_vector_packing GLM_EXT_vector_packing
+/// @ingroup ext
+///
+/// Include <glm/ext/vector_packing.hpp> to use the features of this extension.
+///
+/// This extension provides a set of function to convert vectors to packed
+/// formats.
+
+#pragma once
+
+// Dependency:
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_packing extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_packing
+	/// @{
+
+
+	/// @}
+}// namespace glm
+
+#include "vector_packing.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_packing.inl
new file mode 100644
index 000000000000..e69de29bb2d1
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.hpp
new file mode 100644
index 000000000000..84d67662f26e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.hpp
@@ -0,0 +1,135 @@
+/// @ref ext_vector_reciprocal
+/// @file glm/ext/vector_reciprocal.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup ext_vector_reciprocal GLM_EXT_vector_reciprocal
+/// @ingroup ext
+///
+/// Include <glm/ext/vector_reciprocal.hpp> to use the features of this extension.
+///
+/// Define secant, cosecant and cotangent functions.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_reciprocal extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_reciprocal
+	/// @{
+
+	/// Secant function.
+	/// hypotenuse / adjacent or 1 / cos(x)
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType sec(genType angle);
+
+	/// Cosecant function.
+	/// hypotenuse / opposite or 1 / sin(x)
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType csc(genType angle);
+
+	/// Cotangent function.
+	/// adjacent / opposite or 1 / tan(x)
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType cot(genType angle);
+
+	/// Inverse secant function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType asec(genType x);
+
+	/// Inverse cosecant function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acsc(genType x);
+
+	/// Inverse cotangent function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acot(genType x);
+
+	/// Secant hyperbolic function.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType sech(genType angle);
+
+	/// Cosecant hyperbolic function.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType csch(genType angle);
+
+	/// Cotangent hyperbolic function.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType coth(genType angle);
+
+	/// Inverse secant hyperbolic function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType asech(genType x);
+
+	/// Inverse cosecant hyperbolic function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acsch(genType x);
+
+	/// Inverse cotangent hyperbolic function.
+	///
+	/// @return Return an angle expressed in radians.
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see ext_vector_reciprocal
+	template<typename genType>
+	GLM_FUNC_DECL genType acoth(genType x);
+
+	/// @}
+}//namespace glm
+
+#include "vector_reciprocal.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.inl
new file mode 100644
index 000000000000..b85102a2f578
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_reciprocal.inl
@@ -0,0 +1,105 @@
+/// @ref ext_vector_reciprocal
+
+#include "../trigonometric.hpp"
+#include <limits>
+
+namespace glm
+{
+	// sec
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> sec(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'sec' only accept floating-point inputs");
+		return static_cast<T>(1) / detail::functor1<vec, L, T, T, Q>::call(cos, x);
+	}
+
+	// csc
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> csc(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'csc' only accept floating-point inputs");
+		return static_cast<T>(1) / detail::functor1<vec, L, T, T, Q>::call(sin, x);
+	}
+
+	// cot
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> cot(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'cot' only accept floating-point inputs");
+		T const pi_over_2 = static_cast<T>(3.1415926535897932384626433832795 / 2.0);
+		return detail::functor1<vec, L, T, T, Q>::call(tan, pi_over_2 - x);
+	}
+
+	// asec
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> asec(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'asec' only accept floating-point inputs");
+		return detail::functor1<vec, L, T, T, Q>::call(acos, static_cast<T>(1) / x);
+	}
+
+	// acsc
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> acsc(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acsc' only accept floating-point inputs");
+		return detail::functor1<vec, L, T, T, Q>::call(asin, static_cast<T>(1) / x);
+	}
+
+	// acot
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> acot(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acot' only accept floating-point inputs");
+		T const pi_over_2 = static_cast<T>(3.1415926535897932384626433832795 / 2.0);
+		return pi_over_2 - detail::functor1<vec, L, T, T, Q>::call(atan, x);
+	}
+
+	// sech
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> sech(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'sech' only accept floating-point inputs");
+		return static_cast<T>(1) / detail::functor1<vec, L, T, T, Q>::call(cosh, x);
+	}
+
+	// csch
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> csch(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'csch' only accept floating-point inputs");
+		return static_cast<T>(1) / detail::functor1<vec, L, T, T, Q>::call(sinh, x);
+	}
+
+	// coth
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> coth(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'coth' only accept floating-point inputs");
+		return glm::cosh(x) / glm::sinh(x);
+	}
+
+	// asech
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> asech(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'asech' only accept floating-point inputs");
+		return detail::functor1<vec, L, T, T, Q>::call(acosh, static_cast<T>(1) / x);
+	}
+
+	// acsch
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> acsch(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acsch' only accept floating-point inputs");
+		return detail::functor1<vec, L, T, T, Q>::call(asinh, static_cast<T>(1) / x);
+	}
+
+	// acoth
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> acoth(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'acoth' only accept floating-point inputs");
+		return detail::functor1<vec, L, T, T, Q>::call(atanh, static_cast<T>(1) / x);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.hpp
new file mode 100644
index 000000000000..1c2367dc0231
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.hpp
@@ -0,0 +1,107 @@
+/// @ref ext_vector_relational
+/// @file glm/ext/vector_relational.hpp
+///
+/// @see core (dependence)
+/// @see ext_scalar_integer (dependence)
+///
+/// @defgroup ext_vector_relational GLM_EXT_vector_relational
+/// @ingroup ext
+///
+/// Exposes comparison functions for vector types that take a user defined epsilon values.
+///
+/// Include <glm/ext/vector_relational.hpp> to use the features of this extension.
+///
+/// @see core_vector_relational
+/// @see ext_scalar_relational
+/// @see ext_matrix_relational
+
+#pragma once
+
+// Dependencies
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_relational extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_relational
+	/// @{
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T epsilon);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& epsilon);
+
+	/// Returns the component-wise comparison of |x - y| >= epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T epsilon);
+
+	/// Returns the component-wise comparison of |x - y| >= epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& epsilon);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, int ULPs);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, int, Q> const& ULPs);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, int ULPs);
+
+	/// Returns the component-wise comparison between two vectors in term of ULPs.
+	/// True if this expression is not satisfied.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, int, Q> const& ULPs);
+
+	/// @}
+}//namespace glm
+
+#include "vector_relational.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.inl
new file mode 100644
index 000000000000..7a39ab50897e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_relational.inl
@@ -0,0 +1,75 @@
+#include "../vector_relational.hpp"
+#include "../common.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/type_float.hpp"
+
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T Epsilon)
+	{
+		return equal(x, y, vec<L, T, Q>(Epsilon));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& Epsilon)
+	{
+		return lessThanEqual(abs(x - y), Epsilon);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T Epsilon)
+	{
+		return notEqual(x, y, vec<L, T, Q>(Epsilon));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& Epsilon)
+	{
+		return greaterThan(abs(x - y), Epsilon);
+	}
+
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, int MaxULPs)
+	{
+		return equal(x, y, vec<L, int, Q>(MaxULPs));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, int, Q> const& MaxULPs)
+	{
+		vec<L, bool, Q> Result(false);
+		for(length_t i = 0; i < L; ++i)
+		{
+			detail::float_t<T> const a(x[i]);
+			detail::float_t<T> const b(y[i]);
+
+			// Different signs means they do not match.
+			if(a.negative() != b.negative())
+			{
+				// Check for equality to make sure +0==-0
+				Result[i] = a.mantissa() == b.mantissa() && a.exponent() == b.exponent();
+			}
+			else
+			{
+				// Find the difference in ULPs.
+				typename detail::float_t<T>::int_type const DiffULPs = abs(a.i - b.i);
+				Result[i] = DiffULPs <= MaxULPs[i];
+			}
+		}
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, int MaxULPs)
+	{
+		return notEqual(x, y, vec<L, int, Q>(MaxULPs));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, int, Q> const& MaxULPs)
+	{
+		return not_(equal(x, y, MaxULPs));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1.hpp
new file mode 100644
index 000000000000..eb8a7049761f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1.hpp
@@ -0,0 +1,32 @@
+/// @ref ext_vector_uint1
+/// @file glm/ext/vector_uint1.hpp
+///
+/// @defgroup ext_vector_uint1 GLM_EXT_vector_uint1
+/// @ingroup ext
+///
+/// Exposes uvec1 vector type.
+///
+/// Include <glm/ext/vector_uvec1.hpp> to use the features of this extension.
+///
+/// @see ext_vector_int1 extension.
+/// @see ext_vector_uint1_precision extension.
+
+#pragma once
+
+#include "../detail/type_vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_uint1 extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_uint1
+	/// @{
+
+	/// 1 component vector of unsigned integer numbers.
+	typedef vec<1, unsigned int, defaultp>			uvec1;
+
+	/// @}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1_sized.hpp
new file mode 100644
index 000000000000..2a938bbaf616
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint1_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_uint1_sized
+/// @file glm/ext/vector_uint1_sized.hpp
+///
+/// @defgroup ext_vector_uint1_sized GLM_EXT_vector_uint1_sized
+/// @ingroup ext
+///
+/// Exposes sized unsigned integer vector types.
+///
+/// Include <glm/ext/vector_uint1_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_uint_sized
+/// @see ext_vector_int1_sized
+
+#pragma once
+
+#include "../ext/vector_uint1.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_uint1_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_uint1_sized
+	/// @{
+
+	/// 8 bit unsigned integer vector of 1 component type.
+	///
+	/// @see ext_vector_uint1_sized
+	typedef vec<1, uint8, defaultp>		u8vec1;
+
+	/// 16 bit unsigned integer vector of 1 component type.
+	///
+	/// @see ext_vector_uint1_sized
+	typedef vec<1, uint16, defaultp>	u16vec1;
+
+	/// 32 bit unsigned integer vector of 1 component type.
+	///
+	/// @see ext_vector_uint1_sized
+	typedef vec<1, uint32, defaultp>	u32vec1;
+
+	/// 64 bit unsigned integer vector of 1 component type.
+	///
+	/// @see ext_vector_uint1_sized
+	typedef vec<1, uint64, defaultp>	u64vec1;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2.hpp
new file mode 100644
index 000000000000..03c00f5ff584
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_uint2.hpp
+
+#pragma once
+#include "../detail/type_vec2.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 2 components vector of unsigned integer numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<2, unsigned int, defaultp>		uvec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2_sized.hpp
new file mode 100644
index 000000000000..620fdc6ece39
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint2_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_uint2_sized
+/// @file glm/ext/vector_uint2_sized.hpp
+///
+/// @defgroup ext_vector_uint2_sized GLM_EXT_vector_uint2_sized
+/// @ingroup ext
+///
+/// Exposes sized unsigned integer vector of 2 components type.
+///
+/// Include <glm/ext/vector_uint2_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_uint_sized
+/// @see ext_vector_int2_sized
+
+#pragma once
+
+#include "../ext/vector_uint2.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_uint2_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_uint2_sized
+	/// @{
+
+	/// 8 bit unsigned integer vector of 2 components type.
+	///
+	/// @see ext_vector_uint2_sized
+	typedef vec<2, uint8, defaultp>		u8vec2;
+
+	/// 16 bit unsigned integer vector of 2 components type.
+	///
+	/// @see ext_vector_uint2_sized
+	typedef vec<2, uint16, defaultp>	u16vec2;
+
+	/// 32 bit unsigned integer vector of 2 components type.
+	///
+	/// @see ext_vector_uint2_sized
+	typedef vec<2, uint32, defaultp>	u32vec2;
+
+	/// 64 bit unsigned integer vector of 2 components type.
+	///
+	/// @see ext_vector_uint2_sized
+	typedef vec<2, uint64, defaultp>	u64vec2;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3.hpp
new file mode 100644
index 000000000000..f5b41c40882a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_uint3.hpp
+
+#pragma once
+#include "../detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 3 components vector of unsigned integer numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<3, unsigned int, defaultp>		uvec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3_sized.hpp
new file mode 100644
index 000000000000..6f96b98e276f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint3_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_uint3_sized
+/// @file glm/ext/vector_uint3_sized.hpp
+///
+/// @defgroup ext_vector_uint3_sized GLM_EXT_vector_uint3_sized
+/// @ingroup ext
+///
+/// Exposes sized unsigned integer vector of 3 components type.
+///
+/// Include <glm/ext/vector_uint3_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_uint_sized
+/// @see ext_vector_int3_sized
+
+#pragma once
+
+#include "../ext/vector_uint3.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_uint3_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_uint3_sized
+	/// @{
+
+	/// 8 bit unsigned integer vector of 3 components type.
+	///
+	/// @see ext_vector_uint3_sized
+	typedef vec<3, uint8, defaultp>		u8vec3;
+
+	/// 16 bit unsigned integer vector of 3 components type.
+	///
+	/// @see ext_vector_uint3_sized
+	typedef vec<3, uint16, defaultp>	u16vec3;
+
+	/// 32 bit unsigned integer vector of 3 components type.
+	///
+	/// @see ext_vector_uint3_sized
+	typedef vec<3, uint32, defaultp>	u32vec3;
+
+	/// 64 bit unsigned integer vector of 3 components type.
+	///
+	/// @see ext_vector_uint3_sized
+	typedef vec<3, uint64, defaultp>	u64vec3;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4.hpp
new file mode 100644
index 000000000000..32ced58a8f03
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4.hpp
@@ -0,0 +1,18 @@
+/// @ref core
+/// @file glm/ext/vector_uint4.hpp
+
+#pragma once
+#include "../detail/type_vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_vector
+	/// @{
+
+	/// 4 components vector of unsigned integer numbers.
+	///
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 4.1.5 Vectors</a>
+	typedef vec<4, unsigned int, defaultp>		uvec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4_sized.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4_sized.hpp
new file mode 100644
index 000000000000..da992ea2da86
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_uint4_sized.hpp
@@ -0,0 +1,49 @@
+/// @ref ext_vector_uint4_sized
+/// @file glm/ext/vector_uint4_sized.hpp
+///
+/// @defgroup ext_vector_uint4_sized GLM_EXT_vector_uint4_sized
+/// @ingroup ext
+///
+/// Exposes sized unsigned integer vector of 4 components type.
+///
+/// Include <glm/ext/vector_uint4_sized.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_uint_sized
+/// @see ext_vector_int4_sized
+
+#pragma once
+
+#include "../ext/vector_uint4.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_uint4_sized extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_uint4_sized
+	/// @{
+
+	/// 8 bit unsigned integer vector of 4 components type.
+	///
+	/// @see ext_vector_uint4_sized
+	typedef vec<4, uint8, defaultp>		u8vec4;
+
+	/// 16 bit unsigned integer vector of 4 components type.
+	///
+	/// @see ext_vector_uint4_sized
+	typedef vec<4, uint16, defaultp>	u16vec4;
+
+	/// 32 bit unsigned integer vector of 4 components type.
+	///
+	/// @see ext_vector_uint4_sized
+	typedef vec<4, uint32, defaultp>	u32vec4;
+
+	/// 64 bit unsigned integer vector of 4 components type.
+	///
+	/// @see ext_vector_uint4_sized
+	typedef vec<4, uint64, defaultp>	u64vec4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.hpp b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.hpp
new file mode 100644
index 000000000000..7c539bbf8866
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.hpp
@@ -0,0 +1,112 @@
+/// @ref ext_vector_ulp
+/// @file glm/ext/vector_ulp.hpp
+///
+/// @defgroup ext_vector_ulp GLM_EXT_vector_ulp
+/// @ingroup ext
+///
+/// Allow the measurement of the accuracy of a function against a reference
+/// implementation. This extension works on floating-point data and provide results
+/// in ULP.
+///
+/// Include <glm/ext/vector_ulp.hpp> to use the features of this extension.
+///
+/// @see ext_scalar_ulp
+/// @see ext_scalar_relational
+/// @see ext_vector_relational
+
+#pragma once
+
+// Dependencies
+#include "../ext/scalar_ulp.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_EXT_vector_ulp extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup ext_vector_ulp
+	/// @{
+
+	/// Return the next ULP value(s) after the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> nextFloat(vec<L, T, Q> const& x);
+
+	/// Return the value(s) ULP distance after the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> nextFloat(vec<L, T, Q> const& x, int ULPs);
+
+	/// Return the value(s) ULP distance after the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> nextFloat(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs);
+
+	/// Return the previous ULP value(s) before the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prevFloat(vec<L, T, Q> const& x);
+
+	/// Return the value(s) ULP distance before the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prevFloat(vec<L, T, Q> const& x, int ULPs);
+
+	/// Return the value(s) ULP distance before the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prevFloat(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs);
+
+	/// Return the distance in the number of ULP between 2 single-precision floating-point scalars.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> floatDistance(vec<L, float, Q> const& x, vec<L, float, Q> const& y);
+
+	/// Return the distance in the number of ULP between 2 double-precision floating-point scalars.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_scalar_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int64, Q> floatDistance(vec<L, double, Q> const& x, vec<L, double, Q> const& y);
+
+	/// @}
+}//namespace glm
+
+#include "vector_ulp.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.inl b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.inl
new file mode 100644
index 000000000000..91565ce51074
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/ext/vector_ulp.inl
@@ -0,0 +1,74 @@
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> nextFloat(vec<L, T, Q> const& x)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = nextFloat(x[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> nextFloat(vec<L, T, Q> const& x, int ULPs)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = nextFloat(x[i], ULPs);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> nextFloat(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = nextFloat(x[i], ULPs[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prevFloat(vec<L, T, Q> const& x)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = prevFloat(x[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prevFloat(vec<L, T, Q> const& x, int ULPs)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = prevFloat(x[i], ULPs);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prevFloat(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = prevFloat(x[i], ULPs[i]);
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> floatDistance(vec<L, float, Q> const& x, vec<L, float, Q> const& y)
+	{
+		vec<L, int, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = floatDistance(x[i], y[i]);
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int64, Q> floatDistance(vec<L, double, Q> const& x, vec<L, double, Q> const& y)
+	{
+		vec<L, int64, Q> Result;
+		for(length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = floatDistance(x[i], y[i]);
+		return Result;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/fwd.hpp b/thirdparty/manifold/thirdparty/glm/glm/fwd.hpp
new file mode 100644
index 000000000000..9c2e5eafa4ea
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/fwd.hpp
@@ -0,0 +1,1233 @@
+#pragma once
+
+#include "detail/qualifier.hpp"
+
+namespace glm
+{
+#if GLM_HAS_EXTENDED_INTEGER_TYPE
+	typedef std::int8_t				int8;
+	typedef std::int16_t			int16;
+	typedef std::int32_t			int32;
+	typedef std::int64_t			int64;
+
+	typedef std::uint8_t			uint8;
+	typedef std::uint16_t			uint16;
+	typedef std::uint32_t			uint32;
+	typedef std::uint64_t			uint64;
+#else
+	typedef signed char				int8;
+	typedef signed short			int16;
+	typedef signed int				int32;
+	typedef detail::int64			int64;
+
+	typedef unsigned char			uint8;
+	typedef unsigned short			uint16;
+	typedef unsigned int			uint32;
+	typedef detail::uint64			uint64;
+#endif
+
+	// Scalar int
+
+	typedef int8					lowp_i8;
+	typedef int8					mediump_i8;
+	typedef int8					highp_i8;
+	typedef int8					i8;
+
+	typedef int8					lowp_int8;
+	typedef int8					mediump_int8;
+	typedef int8					highp_int8;
+
+	typedef int8					lowp_int8_t;
+	typedef int8					mediump_int8_t;
+	typedef int8					highp_int8_t;
+	typedef int8					int8_t;
+
+	typedef int16					lowp_i16;
+	typedef int16					mediump_i16;
+	typedef int16					highp_i16;
+	typedef int16					i16;
+
+	typedef int16					lowp_int16;
+	typedef int16					mediump_int16;
+	typedef int16					highp_int16;
+
+	typedef int16					lowp_int16_t;
+	typedef int16					mediump_int16_t;
+	typedef int16					highp_int16_t;
+	typedef int16					int16_t;
+
+	typedef int32					lowp_i32;
+	typedef int32					mediump_i32;
+	typedef int32					highp_i32;
+	typedef int32					i32;
+
+	typedef int32					lowp_int32;
+	typedef int32					mediump_int32;
+	typedef int32					highp_int32;
+
+	typedef int32					lowp_int32_t;
+	typedef int32					mediump_int32_t;
+	typedef int32					highp_int32_t;
+	typedef int32					int32_t;
+
+	typedef int64					lowp_i64;
+	typedef int64					mediump_i64;
+	typedef int64					highp_i64;
+	typedef int64					i64;
+
+	typedef int64					lowp_int64;
+	typedef int64					mediump_int64;
+	typedef int64					highp_int64;
+
+	typedef int64					lowp_int64_t;
+	typedef int64					mediump_int64_t;
+	typedef int64					highp_int64_t;
+	typedef int64					int64_t;
+
+	// Scalar uint
+
+	typedef unsigned int			uint;
+
+	typedef uint8					lowp_u8;
+	typedef uint8					mediump_u8;
+	typedef uint8					highp_u8;
+	typedef uint8					u8;
+
+	typedef uint8					lowp_uint8;
+	typedef uint8					mediump_uint8;
+	typedef uint8					highp_uint8;
+
+	typedef uint8					lowp_uint8_t;
+	typedef uint8					mediump_uint8_t;
+	typedef uint8					highp_uint8_t;
+	typedef uint8					uint8_t;
+
+	typedef uint16					lowp_u16;
+	typedef uint16					mediump_u16;
+	typedef uint16					highp_u16;
+	typedef uint16					u16;
+
+	typedef uint16					lowp_uint16;
+	typedef uint16					mediump_uint16;
+	typedef uint16					highp_uint16;
+
+	typedef uint16					lowp_uint16_t;
+	typedef uint16					mediump_uint16_t;
+	typedef uint16					highp_uint16_t;
+	typedef uint16					uint16_t;
+
+	typedef uint32					lowp_u32;
+	typedef uint32					mediump_u32;
+	typedef uint32					highp_u32;
+	typedef uint32					u32;
+
+	typedef uint32					lowp_uint32;
+	typedef uint32					mediump_uint32;
+	typedef uint32					highp_uint32;
+
+	typedef uint32					lowp_uint32_t;
+	typedef uint32					mediump_uint32_t;
+	typedef uint32					highp_uint32_t;
+	typedef uint32					uint32_t;
+
+	typedef uint64					lowp_u64;
+	typedef uint64					mediump_u64;
+	typedef uint64					highp_u64;
+	typedef uint64					u64;
+
+	typedef uint64					lowp_uint64;
+	typedef uint64					mediump_uint64;
+	typedef uint64					highp_uint64;
+
+	typedef uint64					lowp_uint64_t;
+	typedef uint64					mediump_uint64_t;
+	typedef uint64					highp_uint64_t;
+	typedef uint64					uint64_t;
+
+	// Scalar float
+
+	typedef float					lowp_f32;
+	typedef float					mediump_f32;
+	typedef float					highp_f32;
+	typedef float					f32;
+
+	typedef float					lowp_float32;
+	typedef float					mediump_float32;
+	typedef float					highp_float32;
+	typedef float					float32;
+
+	typedef float					lowp_float32_t;
+	typedef float					mediump_float32_t;
+	typedef float					highp_float32_t;
+	typedef float					float32_t;
+
+
+	typedef double					lowp_f64;
+	typedef double					mediump_f64;
+	typedef double					highp_f64;
+	typedef double					f64;
+
+	typedef double					lowp_float64;
+	typedef double					mediump_float64;
+	typedef double					highp_float64;
+	typedef double					float64;
+
+	typedef double					lowp_float64_t;
+	typedef double					mediump_float64_t;
+	typedef double					highp_float64_t;
+	typedef double					float64_t;
+
+	// Vector bool
+
+	typedef vec<1, bool, lowp>		lowp_bvec1;
+	typedef vec<2, bool, lowp>		lowp_bvec2;
+	typedef vec<3, bool, lowp>		lowp_bvec3;
+	typedef vec<4, bool, lowp>		lowp_bvec4;
+
+	typedef vec<1, bool, mediump>	mediump_bvec1;
+	typedef vec<2, bool, mediump>	mediump_bvec2;
+	typedef vec<3, bool, mediump>	mediump_bvec3;
+	typedef vec<4, bool, mediump>	mediump_bvec4;
+
+	typedef vec<1, bool, highp>		highp_bvec1;
+	typedef vec<2, bool, highp>		highp_bvec2;
+	typedef vec<3, bool, highp>		highp_bvec3;
+	typedef vec<4, bool, highp>		highp_bvec4;
+
+	typedef vec<1, bool, defaultp>	bvec1;
+	typedef vec<2, bool, defaultp>	bvec2;
+	typedef vec<3, bool, defaultp>	bvec3;
+	typedef vec<4, bool, defaultp>	bvec4;
+
+	// Vector int
+
+	typedef vec<1, int, lowp>		lowp_ivec1;
+	typedef vec<2, int, lowp>		lowp_ivec2;
+	typedef vec<3, int, lowp>		lowp_ivec3;
+	typedef vec<4, int, lowp>		lowp_ivec4;
+
+	typedef vec<1, int, mediump>	mediump_ivec1;
+	typedef vec<2, int, mediump>	mediump_ivec2;
+	typedef vec<3, int, mediump>	mediump_ivec3;
+	typedef vec<4, int, mediump>	mediump_ivec4;
+
+	typedef vec<1, int, highp>		highp_ivec1;
+	typedef vec<2, int, highp>		highp_ivec2;
+	typedef vec<3, int, highp>		highp_ivec3;
+	typedef vec<4, int, highp>		highp_ivec4;
+
+	typedef vec<1, int, defaultp>	ivec1;
+	typedef vec<2, int, defaultp>	ivec2;
+	typedef vec<3, int, defaultp>	ivec3;
+	typedef vec<4, int, defaultp>	ivec4;
+
+	typedef vec<1, i8, lowp>		lowp_i8vec1;
+	typedef vec<2, i8, lowp>		lowp_i8vec2;
+	typedef vec<3, i8, lowp>		lowp_i8vec3;
+	typedef vec<4, i8, lowp>		lowp_i8vec4;
+
+	typedef vec<1, i8, mediump>		mediump_i8vec1;
+	typedef vec<2, i8, mediump>		mediump_i8vec2;
+	typedef vec<3, i8, mediump>		mediump_i8vec3;
+	typedef vec<4, i8, mediump>		mediump_i8vec4;
+
+	typedef vec<1, i8, highp>		highp_i8vec1;
+	typedef vec<2, i8, highp>		highp_i8vec2;
+	typedef vec<3, i8, highp>		highp_i8vec3;
+	typedef vec<4, i8, highp>		highp_i8vec4;
+
+	typedef vec<1, i8, defaultp>	i8vec1;
+	typedef vec<2, i8, defaultp>	i8vec2;
+	typedef vec<3, i8, defaultp>	i8vec3;
+	typedef vec<4, i8, defaultp>	i8vec4;
+
+	typedef vec<1, i16, lowp>		lowp_i16vec1;
+	typedef vec<2, i16, lowp>		lowp_i16vec2;
+	typedef vec<3, i16, lowp>		lowp_i16vec3;
+	typedef vec<4, i16, lowp>		lowp_i16vec4;
+
+	typedef vec<1, i16, mediump>	mediump_i16vec1;
+	typedef vec<2, i16, mediump>	mediump_i16vec2;
+	typedef vec<3, i16, mediump>	mediump_i16vec3;
+	typedef vec<4, i16, mediump>	mediump_i16vec4;
+
+	typedef vec<1, i16, highp>		highp_i16vec1;
+	typedef vec<2, i16, highp>		highp_i16vec2;
+	typedef vec<3, i16, highp>		highp_i16vec3;
+	typedef vec<4, i16, highp>		highp_i16vec4;
+
+	typedef vec<1, i16, defaultp>	i16vec1;
+	typedef vec<2, i16, defaultp>	i16vec2;
+	typedef vec<3, i16, defaultp>	i16vec3;
+	typedef vec<4, i16, defaultp>	i16vec4;
+
+	typedef vec<1, i32, lowp>		lowp_i32vec1;
+	typedef vec<2, i32, lowp>		lowp_i32vec2;
+	typedef vec<3, i32, lowp>		lowp_i32vec3;
+	typedef vec<4, i32, lowp>		lowp_i32vec4;
+
+	typedef vec<1, i32, mediump>	mediump_i32vec1;
+	typedef vec<2, i32, mediump>	mediump_i32vec2;
+	typedef vec<3, i32, mediump>	mediump_i32vec3;
+	typedef vec<4, i32, mediump>	mediump_i32vec4;
+
+	typedef vec<1, i32, highp>		highp_i32vec1;
+	typedef vec<2, i32, highp>		highp_i32vec2;
+	typedef vec<3, i32, highp>		highp_i32vec3;
+	typedef vec<4, i32, highp>		highp_i32vec4;
+
+	typedef vec<1, i32, defaultp>	i32vec1;
+	typedef vec<2, i32, defaultp>	i32vec2;
+	typedef vec<3, i32, defaultp>	i32vec3;
+	typedef vec<4, i32, defaultp>	i32vec4;
+
+	typedef vec<1, i64, lowp>		lowp_i64vec1;
+	typedef vec<2, i64, lowp>		lowp_i64vec2;
+	typedef vec<3, i64, lowp>		lowp_i64vec3;
+	typedef vec<4, i64, lowp>		lowp_i64vec4;
+
+	typedef vec<1, i64, mediump>	mediump_i64vec1;
+	typedef vec<2, i64, mediump>	mediump_i64vec2;
+	typedef vec<3, i64, mediump>	mediump_i64vec3;
+	typedef vec<4, i64, mediump>	mediump_i64vec4;
+
+	typedef vec<1, i64, highp>		highp_i64vec1;
+	typedef vec<2, i64, highp>		highp_i64vec2;
+	typedef vec<3, i64, highp>		highp_i64vec3;
+	typedef vec<4, i64, highp>		highp_i64vec4;
+
+	typedef vec<1, i64, defaultp>	i64vec1;
+	typedef vec<2, i64, defaultp>	i64vec2;
+	typedef vec<3, i64, defaultp>	i64vec3;
+	typedef vec<4, i64, defaultp>	i64vec4;
+
+	// Vector uint
+
+	typedef vec<1, uint, lowp>		lowp_uvec1;
+	typedef vec<2, uint, lowp>		lowp_uvec2;
+	typedef vec<3, uint, lowp>		lowp_uvec3;
+	typedef vec<4, uint, lowp>		lowp_uvec4;
+
+	typedef vec<1, uint, mediump>	mediump_uvec1;
+	typedef vec<2, uint, mediump>	mediump_uvec2;
+	typedef vec<3, uint, mediump>	mediump_uvec3;
+	typedef vec<4, uint, mediump>	mediump_uvec4;
+
+	typedef vec<1, uint, highp>		highp_uvec1;
+	typedef vec<2, uint, highp>		highp_uvec2;
+	typedef vec<3, uint, highp>		highp_uvec3;
+	typedef vec<4, uint, highp>		highp_uvec4;
+
+	typedef vec<1, uint, defaultp>	uvec1;
+	typedef vec<2, uint, defaultp>	uvec2;
+	typedef vec<3, uint, defaultp>	uvec3;
+	typedef vec<4, uint, defaultp>	uvec4;
+
+	typedef vec<1, u8, lowp>		lowp_u8vec1;
+	typedef vec<2, u8, lowp>		lowp_u8vec2;
+	typedef vec<3, u8, lowp>		lowp_u8vec3;
+	typedef vec<4, u8, lowp>		lowp_u8vec4;
+
+	typedef vec<1, u8, mediump>		mediump_u8vec1;
+	typedef vec<2, u8, mediump>		mediump_u8vec2;
+	typedef vec<3, u8, mediump>		mediump_u8vec3;
+	typedef vec<4, u8, mediump>		mediump_u8vec4;
+
+	typedef vec<1, u8, highp>		highp_u8vec1;
+	typedef vec<2, u8, highp>		highp_u8vec2;
+	typedef vec<3, u8, highp>		highp_u8vec3;
+	typedef vec<4, u8, highp>		highp_u8vec4;
+
+	typedef vec<1, u8, defaultp>	u8vec1;
+	typedef vec<2, u8, defaultp>	u8vec2;
+	typedef vec<3, u8, defaultp>	u8vec3;
+	typedef vec<4, u8, defaultp>	u8vec4;
+
+	typedef vec<1, u16, lowp>		lowp_u16vec1;
+	typedef vec<2, u16, lowp>		lowp_u16vec2;
+	typedef vec<3, u16, lowp>		lowp_u16vec3;
+	typedef vec<4, u16, lowp>		lowp_u16vec4;
+
+	typedef vec<1, u16, mediump>	mediump_u16vec1;
+	typedef vec<2, u16, mediump>	mediump_u16vec2;
+	typedef vec<3, u16, mediump>	mediump_u16vec3;
+	typedef vec<4, u16, mediump>	mediump_u16vec4;
+
+	typedef vec<1, u16, highp>		highp_u16vec1;
+	typedef vec<2, u16, highp>		highp_u16vec2;
+	typedef vec<3, u16, highp>		highp_u16vec3;
+	typedef vec<4, u16, highp>		highp_u16vec4;
+
+	typedef vec<1, u16, defaultp>	u16vec1;
+	typedef vec<2, u16, defaultp>	u16vec2;
+	typedef vec<3, u16, defaultp>	u16vec3;
+	typedef vec<4, u16, defaultp>	u16vec4;
+
+	typedef vec<1, u32, lowp>		lowp_u32vec1;
+	typedef vec<2, u32, lowp>		lowp_u32vec2;
+	typedef vec<3, u32, lowp>		lowp_u32vec3;
+	typedef vec<4, u32, lowp>		lowp_u32vec4;
+
+	typedef vec<1, u32, mediump>	mediump_u32vec1;
+	typedef vec<2, u32, mediump>	mediump_u32vec2;
+	typedef vec<3, u32, mediump>	mediump_u32vec3;
+	typedef vec<4, u32, mediump>	mediump_u32vec4;
+
+	typedef vec<1, u32, highp>		highp_u32vec1;
+	typedef vec<2, u32, highp>		highp_u32vec2;
+	typedef vec<3, u32, highp>		highp_u32vec3;
+	typedef vec<4, u32, highp>		highp_u32vec4;
+
+	typedef vec<1, u32, defaultp>	u32vec1;
+	typedef vec<2, u32, defaultp>	u32vec2;
+	typedef vec<3, u32, defaultp>	u32vec3;
+	typedef vec<4, u32, defaultp>	u32vec4;
+
+	typedef vec<1, u64, lowp>		lowp_u64vec1;
+	typedef vec<2, u64, lowp>		lowp_u64vec2;
+	typedef vec<3, u64, lowp>		lowp_u64vec3;
+	typedef vec<4, u64, lowp>		lowp_u64vec4;
+
+	typedef vec<1, u64, mediump>	mediump_u64vec1;
+	typedef vec<2, u64, mediump>	mediump_u64vec2;
+	typedef vec<3, u64, mediump>	mediump_u64vec3;
+	typedef vec<4, u64, mediump>	mediump_u64vec4;
+
+	typedef vec<1, u64, highp>		highp_u64vec1;
+	typedef vec<2, u64, highp>		highp_u64vec2;
+	typedef vec<3, u64, highp>		highp_u64vec3;
+	typedef vec<4, u64, highp>		highp_u64vec4;
+
+	typedef vec<1, u64, defaultp>	u64vec1;
+	typedef vec<2, u64, defaultp>	u64vec2;
+	typedef vec<3, u64, defaultp>	u64vec3;
+	typedef vec<4, u64, defaultp>	u64vec4;
+
+	// Vector float
+
+	typedef vec<1, float, lowp>			lowp_vec1;
+	typedef vec<2, float, lowp>			lowp_vec2;
+	typedef vec<3, float, lowp>			lowp_vec3;
+	typedef vec<4, float, lowp>			lowp_vec4;
+
+	typedef vec<1, float, mediump>		mediump_vec1;
+	typedef vec<2, float, mediump>		mediump_vec2;
+	typedef vec<3, float, mediump>		mediump_vec3;
+	typedef vec<4, float, mediump>		mediump_vec4;
+
+	typedef vec<1, float, highp>		highp_vec1;
+	typedef vec<2, float, highp>		highp_vec2;
+	typedef vec<3, float, highp>		highp_vec3;
+	typedef vec<4, float, highp>		highp_vec4;
+
+	typedef vec<1, float, defaultp>		vec1;
+	typedef vec<2, float, defaultp>		vec2;
+	typedef vec<3, float, defaultp>		vec3;
+	typedef vec<4, float, defaultp>		vec4;
+
+	typedef vec<1, float, lowp>			lowp_fvec1;
+	typedef vec<2, float, lowp>			lowp_fvec2;
+	typedef vec<3, float, lowp>			lowp_fvec3;
+	typedef vec<4, float, lowp>			lowp_fvec4;
+
+	typedef vec<1, float, mediump>		mediump_fvec1;
+	typedef vec<2, float, mediump>		mediump_fvec2;
+	typedef vec<3, float, mediump>		mediump_fvec3;
+	typedef vec<4, float, mediump>		mediump_fvec4;
+
+	typedef vec<1, float, highp>		highp_fvec1;
+	typedef vec<2, float, highp>		highp_fvec2;
+	typedef vec<3, float, highp>		highp_fvec3;
+	typedef vec<4, float, highp>		highp_fvec4;
+
+	typedef vec<1, f32, defaultp>		fvec1;
+	typedef vec<2, f32, defaultp>		fvec2;
+	typedef vec<3, f32, defaultp>		fvec3;
+	typedef vec<4, f32, defaultp>		fvec4;
+
+	typedef vec<1, f32, lowp>			lowp_f32vec1;
+	typedef vec<2, f32, lowp>			lowp_f32vec2;
+	typedef vec<3, f32, lowp>			lowp_f32vec3;
+	typedef vec<4, f32, lowp>			lowp_f32vec4;
+
+	typedef vec<1, f32, mediump>		mediump_f32vec1;
+	typedef vec<2, f32, mediump>		mediump_f32vec2;
+	typedef vec<3, f32, mediump>		mediump_f32vec3;
+	typedef vec<4, f32, mediump>		mediump_f32vec4;
+
+	typedef vec<1, f32, highp>			highp_f32vec1;
+	typedef vec<2, f32, highp>			highp_f32vec2;
+	typedef vec<3, f32, highp>			highp_f32vec3;
+	typedef vec<4, f32, highp>			highp_f32vec4;
+
+	typedef vec<1, f32, defaultp>		f32vec1;
+	typedef vec<2, f32, defaultp>		f32vec2;
+	typedef vec<3, f32, defaultp>		f32vec3;
+	typedef vec<4, f32, defaultp>		f32vec4;
+
+	typedef vec<1, f64, lowp>			lowp_dvec1;
+	typedef vec<2, f64, lowp>			lowp_dvec2;
+	typedef vec<3, f64, lowp>			lowp_dvec3;
+	typedef vec<4, f64, lowp>			lowp_dvec4;
+
+	typedef vec<1, f64, mediump>		mediump_dvec1;
+	typedef vec<2, f64, mediump>		mediump_dvec2;
+	typedef vec<3, f64, mediump>		mediump_dvec3;
+	typedef vec<4, f64, mediump>		mediump_dvec4;
+
+	typedef vec<1, f64, highp>			highp_dvec1;
+	typedef vec<2, f64, highp>			highp_dvec2;
+	typedef vec<3, f64, highp>			highp_dvec3;
+	typedef vec<4, f64, highp>			highp_dvec4;
+
+	typedef vec<1, f64, defaultp>		dvec1;
+	typedef vec<2, f64, defaultp>		dvec2;
+	typedef vec<3, f64, defaultp>		dvec3;
+	typedef vec<4, f64, defaultp>		dvec4;
+
+	typedef vec<1, f64, lowp>			lowp_f64vec1;
+	typedef vec<2, f64, lowp>			lowp_f64vec2;
+	typedef vec<3, f64, lowp>			lowp_f64vec3;
+	typedef vec<4, f64, lowp>			lowp_f64vec4;
+
+	typedef vec<1, f64, mediump>		mediump_f64vec1;
+	typedef vec<2, f64, mediump>		mediump_f64vec2;
+	typedef vec<3, f64, mediump>		mediump_f64vec3;
+	typedef vec<4, f64, mediump>		mediump_f64vec4;
+
+	typedef vec<1, f64, highp>			highp_f64vec1;
+	typedef vec<2, f64, highp>			highp_f64vec2;
+	typedef vec<3, f64, highp>			highp_f64vec3;
+	typedef vec<4, f64, highp>			highp_f64vec4;
+
+	typedef vec<1, f64, defaultp>		f64vec1;
+	typedef vec<2, f64, defaultp>		f64vec2;
+	typedef vec<3, f64, defaultp>		f64vec3;
+	typedef vec<4, f64, defaultp>		f64vec4;
+
+	// Matrix NxN
+
+	typedef mat<2, 2, f32, lowp>		lowp_mat2;
+	typedef mat<3, 3, f32, lowp>		lowp_mat3;
+	typedef mat<4, 4, f32, lowp>		lowp_mat4;
+
+	typedef mat<2, 2, f32, mediump>		mediump_mat2;
+	typedef mat<3, 3, f32, mediump>		mediump_mat3;
+	typedef mat<4, 4, f32, mediump>		mediump_mat4;
+
+	typedef mat<2, 2, f32, highp>		highp_mat2;
+	typedef mat<3, 3, f32, highp>		highp_mat3;
+	typedef mat<4, 4, f32, highp>		highp_mat4;
+
+	typedef mat<2, 2, f32, defaultp>	mat2;
+	typedef mat<3, 3, f32, defaultp>	mat3;
+	typedef mat<4, 4, f32, defaultp>	mat4;
+
+	typedef mat<2, 2, f32, lowp>		lowp_fmat2;
+	typedef mat<3, 3, f32, lowp>		lowp_fmat3;
+	typedef mat<4, 4, f32, lowp>		lowp_fmat4;
+
+	typedef mat<2, 2, f32, mediump>		mediump_fmat2;
+	typedef mat<3, 3, f32, mediump>		mediump_fmat3;
+	typedef mat<4, 4, f32, mediump>		mediump_fmat4;
+
+	typedef mat<2, 2, f32, highp>		highp_fmat2;
+	typedef mat<3, 3, f32, highp>		highp_fmat3;
+	typedef mat<4, 4, f32, highp>		highp_fmat4;
+
+	typedef mat<2, 2, f32, defaultp>	fmat2;
+	typedef mat<3, 3, f32, defaultp>	fmat3;
+	typedef mat<4, 4, f32, defaultp>	fmat4;
+
+	typedef mat<2, 2, f32, lowp>		lowp_f32mat2;
+	typedef mat<3, 3, f32, lowp>		lowp_f32mat3;
+	typedef mat<4, 4, f32, lowp>		lowp_f32mat4;
+
+	typedef mat<2, 2, f32, mediump>		mediump_f32mat2;
+	typedef mat<3, 3, f32, mediump>		mediump_f32mat3;
+	typedef mat<4, 4, f32, mediump>		mediump_f32mat4;
+
+	typedef mat<2, 2, f32, highp>		highp_f32mat2;
+	typedef mat<3, 3, f32, highp>		highp_f32mat3;
+	typedef mat<4, 4, f32, highp>		highp_f32mat4;
+
+	typedef mat<2, 2, f32, defaultp>	f32mat2;
+	typedef mat<3, 3, f32, defaultp>	f32mat3;
+	typedef mat<4, 4, f32, defaultp>	f32mat4;
+
+	typedef mat<2, 2, f64, lowp>		lowp_dmat2;
+	typedef mat<3, 3, f64, lowp>		lowp_dmat3;
+	typedef mat<4, 4, f64, lowp>		lowp_dmat4;
+
+	typedef mat<2, 2, f64, mediump>		mediump_dmat2;
+	typedef mat<3, 3, f64, mediump>		mediump_dmat3;
+	typedef mat<4, 4, f64, mediump>		mediump_dmat4;
+
+	typedef mat<2, 2, f64, highp>		highp_dmat2;
+	typedef mat<3, 3, f64, highp>		highp_dmat3;
+	typedef mat<4, 4, f64, highp>		highp_dmat4;
+
+	typedef mat<2, 2, f64, defaultp>	dmat2;
+	typedef mat<3, 3, f64, defaultp>	dmat3;
+	typedef mat<4, 4, f64, defaultp>	dmat4;
+
+	typedef mat<2, 2, f64, lowp>		lowp_f64mat2;
+	typedef mat<3, 3, f64, lowp>		lowp_f64mat3;
+	typedef mat<4, 4, f64, lowp>		lowp_f64mat4;
+
+	typedef mat<2, 2, f64, mediump>		mediump_f64mat2;
+	typedef mat<3, 3, f64, mediump>		mediump_f64mat3;
+	typedef mat<4, 4, f64, mediump>		mediump_f64mat4;
+
+	typedef mat<2, 2, f64, highp>		highp_f64mat2;
+	typedef mat<3, 3, f64, highp>		highp_f64mat3;
+	typedef mat<4, 4, f64, highp>		highp_f64mat4;
+
+	typedef mat<2, 2, f64, defaultp>	f64mat2;
+	typedef mat<3, 3, f64, defaultp>	f64mat3;
+	typedef mat<4, 4, f64, defaultp>	f64mat4;
+
+	// Matrix MxN
+
+	typedef mat<2, 2, f32, lowp>		lowp_mat2x2;
+	typedef mat<2, 3, f32, lowp>		lowp_mat2x3;
+	typedef mat<2, 4, f32, lowp>		lowp_mat2x4;
+	typedef mat<3, 2, f32, lowp>		lowp_mat3x2;
+	typedef mat<3, 3, f32, lowp>		lowp_mat3x3;
+	typedef mat<3, 4, f32, lowp>		lowp_mat3x4;
+	typedef mat<4, 2, f32, lowp>		lowp_mat4x2;
+	typedef mat<4, 3, f32, lowp>		lowp_mat4x3;
+	typedef mat<4, 4, f32, lowp>		lowp_mat4x4;
+
+	typedef mat<2, 2, f32, mediump>		mediump_mat2x2;
+	typedef mat<2, 3, f32, mediump>		mediump_mat2x3;
+	typedef mat<2, 4, f32, mediump>		mediump_mat2x4;
+	typedef mat<3, 2, f32, mediump>		mediump_mat3x2;
+	typedef mat<3, 3, f32, mediump>		mediump_mat3x3;
+	typedef mat<3, 4, f32, mediump>		mediump_mat3x4;
+	typedef mat<4, 2, f32, mediump>		mediump_mat4x2;
+	typedef mat<4, 3, f32, mediump>		mediump_mat4x3;
+	typedef mat<4, 4, f32, mediump>		mediump_mat4x4;
+
+	typedef mat<2, 2, f32, highp>		highp_mat2x2;
+	typedef mat<2, 3, f32, highp>		highp_mat2x3;
+	typedef mat<2, 4, f32, highp>		highp_mat2x4;
+	typedef mat<3, 2, f32, highp>		highp_mat3x2;
+	typedef mat<3, 3, f32, highp>		highp_mat3x3;
+	typedef mat<3, 4, f32, highp>		highp_mat3x4;
+	typedef mat<4, 2, f32, highp>		highp_mat4x2;
+	typedef mat<4, 3, f32, highp>		highp_mat4x3;
+	typedef mat<4, 4, f32, highp>		highp_mat4x4;
+
+	typedef mat<2, 2, f32, defaultp>	mat2x2;
+	typedef mat<2, 3, f32, defaultp>	mat2x3;
+	typedef mat<2, 4, f32, defaultp>	mat2x4;
+	typedef mat<3, 2, f32, defaultp>	mat3x2;
+	typedef mat<3, 3, f32, defaultp>	mat3x3;
+	typedef mat<3, 4, f32, defaultp>	mat3x4;
+	typedef mat<4, 2, f32, defaultp>	mat4x2;
+	typedef mat<4, 3, f32, defaultp>	mat4x3;
+	typedef mat<4, 4, f32, defaultp>	mat4x4;
+
+	typedef mat<2, 2, f32, lowp>		lowp_fmat2x2;
+	typedef mat<2, 3, f32, lowp>		lowp_fmat2x3;
+	typedef mat<2, 4, f32, lowp>		lowp_fmat2x4;
+	typedef mat<3, 2, f32, lowp>		lowp_fmat3x2;
+	typedef mat<3, 3, f32, lowp>		lowp_fmat3x3;
+	typedef mat<3, 4, f32, lowp>		lowp_fmat3x4;
+	typedef mat<4, 2, f32, lowp>		lowp_fmat4x2;
+	typedef mat<4, 3, f32, lowp>		lowp_fmat4x3;
+	typedef mat<4, 4, f32, lowp>		lowp_fmat4x4;
+
+	typedef mat<2, 2, f32, mediump>		mediump_fmat2x2;
+	typedef mat<2, 3, f32, mediump>		mediump_fmat2x3;
+	typedef mat<2, 4, f32, mediump>		mediump_fmat2x4;
+	typedef mat<3, 2, f32, mediump>		mediump_fmat3x2;
+	typedef mat<3, 3, f32, mediump>		mediump_fmat3x3;
+	typedef mat<3, 4, f32, mediump>		mediump_fmat3x4;
+	typedef mat<4, 2, f32, mediump>		mediump_fmat4x2;
+	typedef mat<4, 3, f32, mediump>		mediump_fmat4x3;
+	typedef mat<4, 4, f32, mediump>		mediump_fmat4x4;
+
+	typedef mat<2, 2, f32, highp>		highp_fmat2x2;
+	typedef mat<2, 3, f32, highp>		highp_fmat2x3;
+	typedef mat<2, 4, f32, highp>		highp_fmat2x4;
+	typedef mat<3, 2, f32, highp>		highp_fmat3x2;
+	typedef mat<3, 3, f32, highp>		highp_fmat3x3;
+	typedef mat<3, 4, f32, highp>		highp_fmat3x4;
+	typedef mat<4, 2, f32, highp>		highp_fmat4x2;
+	typedef mat<4, 3, f32, highp>		highp_fmat4x3;
+	typedef mat<4, 4, f32, highp>		highp_fmat4x4;
+
+	typedef mat<2, 2, f32, defaultp>	fmat2x2;
+	typedef mat<2, 3, f32, defaultp>	fmat2x3;
+	typedef mat<2, 4, f32, defaultp>	fmat2x4;
+	typedef mat<3, 2, f32, defaultp>	fmat3x2;
+	typedef mat<3, 3, f32, defaultp>	fmat3x3;
+	typedef mat<3, 4, f32, defaultp>	fmat3x4;
+	typedef mat<4, 2, f32, defaultp>	fmat4x2;
+	typedef mat<4, 3, f32, defaultp>	fmat4x3;
+	typedef mat<4, 4, f32, defaultp>	fmat4x4;
+
+	typedef mat<2, 2, f32, lowp>		lowp_f32mat2x2;
+	typedef mat<2, 3, f32, lowp>		lowp_f32mat2x3;
+	typedef mat<2, 4, f32, lowp>		lowp_f32mat2x4;
+	typedef mat<3, 2, f32, lowp>		lowp_f32mat3x2;
+	typedef mat<3, 3, f32, lowp>		lowp_f32mat3x3;
+	typedef mat<3, 4, f32, lowp>		lowp_f32mat3x4;
+	typedef mat<4, 2, f32, lowp>		lowp_f32mat4x2;
+	typedef mat<4, 3, f32, lowp>		lowp_f32mat4x3;
+	typedef mat<4, 4, f32, lowp>		lowp_f32mat4x4;
+	
+	typedef mat<2, 2, f32, mediump>		mediump_f32mat2x2;
+	typedef mat<2, 3, f32, mediump>		mediump_f32mat2x3;
+	typedef mat<2, 4, f32, mediump>		mediump_f32mat2x4;
+	typedef mat<3, 2, f32, mediump>		mediump_f32mat3x2;
+	typedef mat<3, 3, f32, mediump>		mediump_f32mat3x3;
+	typedef mat<3, 4, f32, mediump>		mediump_f32mat3x4;
+	typedef mat<4, 2, f32, mediump>		mediump_f32mat4x2;
+	typedef mat<4, 3, f32, mediump>		mediump_f32mat4x3;
+	typedef mat<4, 4, f32, mediump>		mediump_f32mat4x4;
+
+	typedef mat<2, 2, f32, highp>		highp_f32mat2x2;
+	typedef mat<2, 3, f32, highp>		highp_f32mat2x3;
+	typedef mat<2, 4, f32, highp>		highp_f32mat2x4;
+	typedef mat<3, 2, f32, highp>		highp_f32mat3x2;
+	typedef mat<3, 3, f32, highp>		highp_f32mat3x3;
+	typedef mat<3, 4, f32, highp>		highp_f32mat3x4;
+	typedef mat<4, 2, f32, highp>		highp_f32mat4x2;
+	typedef mat<4, 3, f32, highp>		highp_f32mat4x3;
+	typedef mat<4, 4, f32, highp>		highp_f32mat4x4;
+
+	typedef mat<2, 2, f32, defaultp>	f32mat2x2;
+	typedef mat<2, 3, f32, defaultp>	f32mat2x3;
+	typedef mat<2, 4, f32, defaultp>	f32mat2x4;
+	typedef mat<3, 2, f32, defaultp>	f32mat3x2;
+	typedef mat<3, 3, f32, defaultp>	f32mat3x3;
+	typedef mat<3, 4, f32, defaultp>	f32mat3x4;
+	typedef mat<4, 2, f32, defaultp>	f32mat4x2;
+	typedef mat<4, 3, f32, defaultp>	f32mat4x3;
+	typedef mat<4, 4, f32, defaultp>	f32mat4x4;
+
+	typedef mat<2, 2, double, lowp>		lowp_dmat2x2;
+	typedef mat<2, 3, double, lowp>		lowp_dmat2x3;
+	typedef mat<2, 4, double, lowp>		lowp_dmat2x4;
+	typedef mat<3, 2, double, lowp>		lowp_dmat3x2;
+	typedef mat<3, 3, double, lowp>		lowp_dmat3x3;
+	typedef mat<3, 4, double, lowp>		lowp_dmat3x4;
+	typedef mat<4, 2, double, lowp>		lowp_dmat4x2;
+	typedef mat<4, 3, double, lowp>		lowp_dmat4x3;
+	typedef mat<4, 4, double, lowp>		lowp_dmat4x4;
+
+	typedef mat<2, 2, double, mediump>	mediump_dmat2x2;
+	typedef mat<2, 3, double, mediump>	mediump_dmat2x3;
+	typedef mat<2, 4, double, mediump>	mediump_dmat2x4;
+	typedef mat<3, 2, double, mediump>	mediump_dmat3x2;
+	typedef mat<3, 3, double, mediump>	mediump_dmat3x3;
+	typedef mat<3, 4, double, mediump>	mediump_dmat3x4;
+	typedef mat<4, 2, double, mediump>	mediump_dmat4x2;
+	typedef mat<4, 3, double, mediump>	mediump_dmat4x3;
+	typedef mat<4, 4, double, mediump>	mediump_dmat4x4;
+
+	typedef mat<2, 2, double, highp>	highp_dmat2x2;
+	typedef mat<2, 3, double, highp>	highp_dmat2x3;
+	typedef mat<2, 4, double, highp>	highp_dmat2x4;
+	typedef mat<3, 2, double, highp>	highp_dmat3x2;
+	typedef mat<3, 3, double, highp>	highp_dmat3x3;
+	typedef mat<3, 4, double, highp>	highp_dmat3x4;
+	typedef mat<4, 2, double, highp>	highp_dmat4x2;
+	typedef mat<4, 3, double, highp>	highp_dmat4x3;
+	typedef mat<4, 4, double, highp>	highp_dmat4x4;
+
+	typedef mat<2, 2, double, defaultp>	dmat2x2;
+	typedef mat<2, 3, double, defaultp>	dmat2x3;
+	typedef mat<2, 4, double, defaultp>	dmat2x4;
+	typedef mat<3, 2, double, defaultp>	dmat3x2;
+	typedef mat<3, 3, double, defaultp>	dmat3x3;
+	typedef mat<3, 4, double, defaultp>	dmat3x4;
+	typedef mat<4, 2, double, defaultp>	dmat4x2;
+	typedef mat<4, 3, double, defaultp>	dmat4x3;
+	typedef mat<4, 4, double, defaultp>	dmat4x4;
+
+	typedef mat<2, 2, f64, lowp>		lowp_f64mat2x2;
+	typedef mat<2, 3, f64, lowp>		lowp_f64mat2x3;
+	typedef mat<2, 4, f64, lowp>		lowp_f64mat2x4;
+	typedef mat<3, 2, f64, lowp>		lowp_f64mat3x2;
+	typedef mat<3, 3, f64, lowp>		lowp_f64mat3x3;
+	typedef mat<3, 4, f64, lowp>		lowp_f64mat3x4;
+	typedef mat<4, 2, f64, lowp>		lowp_f64mat4x2;
+	typedef mat<4, 3, f64, lowp>		lowp_f64mat4x3;
+	typedef mat<4, 4, f64, lowp>		lowp_f64mat4x4;
+
+	typedef mat<2, 2, f64, mediump>		mediump_f64mat2x2;
+	typedef mat<2, 3, f64, mediump>		mediump_f64mat2x3;
+	typedef mat<2, 4, f64, mediump>		mediump_f64mat2x4;
+	typedef mat<3, 2, f64, mediump>		mediump_f64mat3x2;
+	typedef mat<3, 3, f64, mediump>		mediump_f64mat3x3;
+	typedef mat<3, 4, f64, mediump>		mediump_f64mat3x4;
+	typedef mat<4, 2, f64, mediump>		mediump_f64mat4x2;
+	typedef mat<4, 3, f64, mediump>		mediump_f64mat4x3;
+	typedef mat<4, 4, f64, mediump>		mediump_f64mat4x4;
+
+	typedef mat<2, 2, f64, highp>		highp_f64mat2x2;
+	typedef mat<2, 3, f64, highp>		highp_f64mat2x3;
+	typedef mat<2, 4, f64, highp>		highp_f64mat2x4;
+	typedef mat<3, 2, f64, highp>		highp_f64mat3x2;
+	typedef mat<3, 3, f64, highp>		highp_f64mat3x3;
+	typedef mat<3, 4, f64, highp>		highp_f64mat3x4;
+	typedef mat<4, 2, f64, highp>		highp_f64mat4x2;
+	typedef mat<4, 3, f64, highp>		highp_f64mat4x3;
+	typedef mat<4, 4, f64, highp>		highp_f64mat4x4;
+
+	typedef mat<2, 2, f64, defaultp>	f64mat2x2;
+	typedef mat<2, 3, f64, defaultp>	f64mat2x3;
+	typedef mat<2, 4, f64, defaultp>	f64mat2x4;
+	typedef mat<3, 2, f64, defaultp>	f64mat3x2;
+	typedef mat<3, 3, f64, defaultp>	f64mat3x3;
+	typedef mat<3, 4, f64, defaultp>	f64mat3x4;
+	typedef mat<4, 2, f64, defaultp>	f64mat4x2;
+	typedef mat<4, 3, f64, defaultp>	f64mat4x3;
+	typedef mat<4, 4, f64, defaultp>	f64mat4x4;
+
+	// Signed integer matrix MxN
+
+	typedef mat<2, 2, int, lowp>		lowp_imat2x2;
+	typedef mat<2, 3, int, lowp>		lowp_imat2x3;
+	typedef mat<2, 4, int, lowp>		lowp_imat2x4;
+	typedef mat<3, 2, int, lowp>		lowp_imat3x2;
+	typedef mat<3, 3, int, lowp>		lowp_imat3x3;
+	typedef mat<3, 4, int, lowp>		lowp_imat3x4;
+	typedef mat<4, 2, int, lowp>		lowp_imat4x2;
+	typedef mat<4, 3, int, lowp>		lowp_imat4x3;
+	typedef mat<4, 4, int, lowp>		lowp_imat4x4;
+
+	typedef mat<2, 2, int, mediump>		mediump_imat2x2;
+	typedef mat<2, 3, int, mediump>		mediump_imat2x3;
+	typedef mat<2, 4, int, mediump>		mediump_imat2x4;
+	typedef mat<3, 2, int, mediump>		mediump_imat3x2;
+	typedef mat<3, 3, int, mediump>		mediump_imat3x3;
+	typedef mat<3, 4, int, mediump>		mediump_imat3x4;
+	typedef mat<4, 2, int, mediump>		mediump_imat4x2;
+	typedef mat<4, 3, int, mediump>		mediump_imat4x3;
+	typedef mat<4, 4, int, mediump>		mediump_imat4x4;
+
+	typedef mat<2, 2, int, highp>		highp_imat2x2;
+	typedef mat<2, 3, int, highp>		highp_imat2x3;
+	typedef mat<2, 4, int, highp>		highp_imat2x4;
+	typedef mat<3, 2, int, highp>		highp_imat3x2;
+	typedef mat<3, 3, int, highp>		highp_imat3x3;
+	typedef mat<3, 4, int, highp>		highp_imat3x4;
+	typedef mat<4, 2, int, highp>		highp_imat4x2;
+	typedef mat<4, 3, int, highp>		highp_imat4x3;
+	typedef mat<4, 4, int, highp>		highp_imat4x4;
+
+	typedef mat<2, 2, int, defaultp>	imat2x2;
+	typedef mat<2, 3, int, defaultp>	imat2x3;
+	typedef mat<2, 4, int, defaultp>	imat2x4;
+	typedef mat<3, 2, int, defaultp>	imat3x2;
+	typedef mat<3, 3, int, defaultp>	imat3x3;
+	typedef mat<3, 4, int, defaultp>	imat3x4;
+	typedef mat<4, 2, int, defaultp>	imat4x2;
+	typedef mat<4, 3, int, defaultp>	imat4x3;
+	typedef mat<4, 4, int, defaultp>	imat4x4;
+
+
+	typedef mat<2, 2, int8, lowp>		lowp_i8mat2x2;
+	typedef mat<2, 3, int8, lowp>		lowp_i8mat2x3;
+	typedef mat<2, 4, int8, lowp>		lowp_i8mat2x4;
+	typedef mat<3, 2, int8, lowp>		lowp_i8mat3x2;
+	typedef mat<3, 3, int8, lowp>		lowp_i8mat3x3;
+	typedef mat<3, 4, int8, lowp>		lowp_i8mat3x4;
+	typedef mat<4, 2, int8, lowp>		lowp_i8mat4x2;
+	typedef mat<4, 3, int8, lowp>		lowp_i8mat4x3;
+	typedef mat<4, 4, int8, lowp>		lowp_i8mat4x4;
+
+	typedef mat<2, 2, int8, mediump>	mediump_i8mat2x2;
+	typedef mat<2, 3, int8, mediump>	mediump_i8mat2x3;
+	typedef mat<2, 4, int8, mediump>	mediump_i8mat2x4;
+	typedef mat<3, 2, int8, mediump>	mediump_i8mat3x2;
+	typedef mat<3, 3, int8, mediump>	mediump_i8mat3x3;
+	typedef mat<3, 4, int8, mediump>	mediump_i8mat3x4;
+	typedef mat<4, 2, int8, mediump>	mediump_i8mat4x2;
+	typedef mat<4, 3, int8, mediump>	mediump_i8mat4x3;
+	typedef mat<4, 4, int8, mediump>	mediump_i8mat4x4;
+
+	typedef mat<2, 2, int8, highp>		highp_i8mat2x2;
+	typedef mat<2, 3, int8, highp>		highp_i8mat2x3;
+	typedef mat<2, 4, int8, highp>		highp_i8mat2x4;
+	typedef mat<3, 2, int8, highp>		highp_i8mat3x2;
+	typedef mat<3, 3, int8, highp>		highp_i8mat3x3;
+	typedef mat<3, 4, int8, highp>		highp_i8mat3x4;
+	typedef mat<4, 2, int8, highp>		highp_i8mat4x2;
+	typedef mat<4, 3, int8, highp>		highp_i8mat4x3;
+	typedef mat<4, 4, int8, highp>		highp_i8mat4x4;
+
+	typedef mat<2, 2, int8, defaultp>	i8mat2x2;
+	typedef mat<2, 3, int8, defaultp>	i8mat2x3;
+	typedef mat<2, 4, int8, defaultp>	i8mat2x4;
+	typedef mat<3, 2, int8, defaultp>	i8mat3x2;
+	typedef mat<3, 3, int8, defaultp>	i8mat3x3;
+	typedef mat<3, 4, int8, defaultp>	i8mat3x4;
+	typedef mat<4, 2, int8, defaultp>	i8mat4x2;
+	typedef mat<4, 3, int8, defaultp>	i8mat4x3;
+	typedef mat<4, 4, int8, defaultp>	i8mat4x4;
+
+
+	typedef mat<2, 2, int16, lowp>		lowp_i16mat2x2;
+	typedef mat<2, 3, int16, lowp>		lowp_i16mat2x3;
+	typedef mat<2, 4, int16, lowp>		lowp_i16mat2x4;
+	typedef mat<3, 2, int16, lowp>		lowp_i16mat3x2;
+	typedef mat<3, 3, int16, lowp>		lowp_i16mat3x3;
+	typedef mat<3, 4, int16, lowp>		lowp_i16mat3x4;
+	typedef mat<4, 2, int16, lowp>		lowp_i16mat4x2;
+	typedef mat<4, 3, int16, lowp>		lowp_i16mat4x3;
+	typedef mat<4, 4, int16, lowp>		lowp_i16mat4x4;
+
+	typedef mat<2, 2, int16, mediump>	mediump_i16mat2x2;
+	typedef mat<2, 3, int16, mediump>	mediump_i16mat2x3;
+	typedef mat<2, 4, int16, mediump>	mediump_i16mat2x4;
+	typedef mat<3, 2, int16, mediump>	mediump_i16mat3x2;
+	typedef mat<3, 3, int16, mediump>	mediump_i16mat3x3;
+	typedef mat<3, 4, int16, mediump>	mediump_i16mat3x4;
+	typedef mat<4, 2, int16, mediump>	mediump_i16mat4x2;
+	typedef mat<4, 3, int16, mediump>	mediump_i16mat4x3;
+	typedef mat<4, 4, int16, mediump>	mediump_i16mat4x4;
+
+	typedef mat<2, 2, int16, highp>		highp_i16mat2x2;
+	typedef mat<2, 3, int16, highp>		highp_i16mat2x3;
+	typedef mat<2, 4, int16, highp>		highp_i16mat2x4;
+	typedef mat<3, 2, int16, highp>		highp_i16mat3x2;
+	typedef mat<3, 3, int16, highp>		highp_i16mat3x3;
+	typedef mat<3, 4, int16, highp>		highp_i16mat3x4;
+	typedef mat<4, 2, int16, highp>		highp_i16mat4x2;
+	typedef mat<4, 3, int16, highp>		highp_i16mat4x3;
+	typedef mat<4, 4, int16, highp>		highp_i16mat4x4;
+
+	typedef mat<2, 2, int16, defaultp>	i16mat2x2;
+	typedef mat<2, 3, int16, defaultp>	i16mat2x3;
+	typedef mat<2, 4, int16, defaultp>	i16mat2x4;
+	typedef mat<3, 2, int16, defaultp>	i16mat3x2;
+	typedef mat<3, 3, int16, defaultp>	i16mat3x3;
+	typedef mat<3, 4, int16, defaultp>	i16mat3x4;
+	typedef mat<4, 2, int16, defaultp>	i16mat4x2;
+	typedef mat<4, 3, int16, defaultp>	i16mat4x3;
+	typedef mat<4, 4, int16, defaultp>	i16mat4x4;
+
+
+	typedef mat<2, 2, int32, lowp>		lowp_i32mat2x2;
+	typedef mat<2, 3, int32, lowp>		lowp_i32mat2x3;
+	typedef mat<2, 4, int32, lowp>		lowp_i32mat2x4;
+	typedef mat<3, 2, int32, lowp>		lowp_i32mat3x2;
+	typedef mat<3, 3, int32, lowp>		lowp_i32mat3x3;
+	typedef mat<3, 4, int32, lowp>		lowp_i32mat3x4;
+	typedef mat<4, 2, int32, lowp>		lowp_i32mat4x2;
+	typedef mat<4, 3, int32, lowp>		lowp_i32mat4x3;
+	typedef mat<4, 4, int32, lowp>		lowp_i32mat4x4;
+
+	typedef mat<2, 2, int32, mediump>	mediump_i32mat2x2;
+	typedef mat<2, 3, int32, mediump>	mediump_i32mat2x3;
+	typedef mat<2, 4, int32, mediump>	mediump_i32mat2x4;
+	typedef mat<3, 2, int32, mediump>	mediump_i32mat3x2;
+	typedef mat<3, 3, int32, mediump>	mediump_i32mat3x3;
+	typedef mat<3, 4, int32, mediump>	mediump_i32mat3x4;
+	typedef mat<4, 2, int32, mediump>	mediump_i32mat4x2;
+	typedef mat<4, 3, int32, mediump>	mediump_i32mat4x3;
+	typedef mat<4, 4, int32, mediump>	mediump_i32mat4x4;
+
+	typedef mat<2, 2, int32, highp>		highp_i32mat2x2;
+	typedef mat<2, 3, int32, highp>		highp_i32mat2x3;
+	typedef mat<2, 4, int32, highp>		highp_i32mat2x4;
+	typedef mat<3, 2, int32, highp>		highp_i32mat3x2;
+	typedef mat<3, 3, int32, highp>		highp_i32mat3x3;
+	typedef mat<3, 4, int32, highp>		highp_i32mat3x4;
+	typedef mat<4, 2, int32, highp>		highp_i32mat4x2;
+	typedef mat<4, 3, int32, highp>		highp_i32mat4x3;
+	typedef mat<4, 4, int32, highp>		highp_i32mat4x4;
+
+	typedef mat<2, 2, int32, defaultp>	i32mat2x2;
+	typedef mat<2, 3, int32, defaultp>	i32mat2x3;
+	typedef mat<2, 4, int32, defaultp>	i32mat2x4;
+	typedef mat<3, 2, int32, defaultp>	i32mat3x2;
+	typedef mat<3, 3, int32, defaultp>	i32mat3x3;
+	typedef mat<3, 4, int32, defaultp>	i32mat3x4;
+	typedef mat<4, 2, int32, defaultp>	i32mat4x2;
+	typedef mat<4, 3, int32, defaultp>	i32mat4x3;
+	typedef mat<4, 4, int32, defaultp>	i32mat4x4;
+
+
+	typedef mat<2, 2, int64, lowp>		lowp_i64mat2x2;
+	typedef mat<2, 3, int64, lowp>		lowp_i64mat2x3;
+	typedef mat<2, 4, int64, lowp>		lowp_i64mat2x4;
+	typedef mat<3, 2, int64, lowp>		lowp_i64mat3x2;
+	typedef mat<3, 3, int64, lowp>		lowp_i64mat3x3;
+	typedef mat<3, 4, int64, lowp>		lowp_i64mat3x4;
+	typedef mat<4, 2, int64, lowp>		lowp_i64mat4x2;
+	typedef mat<4, 3, int64, lowp>		lowp_i64mat4x3;
+	typedef mat<4, 4, int64, lowp>		lowp_i64mat4x4;
+
+	typedef mat<2, 2, int64, mediump>	mediump_i64mat2x2;
+	typedef mat<2, 3, int64, mediump>	mediump_i64mat2x3;
+	typedef mat<2, 4, int64, mediump>	mediump_i64mat2x4;
+	typedef mat<3, 2, int64, mediump>	mediump_i64mat3x2;
+	typedef mat<3, 3, int64, mediump>	mediump_i64mat3x3;
+	typedef mat<3, 4, int64, mediump>	mediump_i64mat3x4;
+	typedef mat<4, 2, int64, mediump>	mediump_i64mat4x2;
+	typedef mat<4, 3, int64, mediump>	mediump_i64mat4x3;
+	typedef mat<4, 4, int64, mediump>	mediump_i64mat4x4;
+
+	typedef mat<2, 2, int64, highp>		highp_i64mat2x2;
+	typedef mat<2, 3, int64, highp>		highp_i64mat2x3;
+	typedef mat<2, 4, int64, highp>		highp_i64mat2x4;
+	typedef mat<3, 2, int64, highp>		highp_i64mat3x2;
+	typedef mat<3, 3, int64, highp>		highp_i64mat3x3;
+	typedef mat<3, 4, int64, highp>		highp_i64mat3x4;
+	typedef mat<4, 2, int64, highp>		highp_i64mat4x2;
+	typedef mat<4, 3, int64, highp>		highp_i64mat4x3;
+	typedef mat<4, 4, int64, highp>		highp_i64mat4x4;
+
+	typedef mat<2, 2, int64, defaultp>	i64mat2x2;
+	typedef mat<2, 3, int64, defaultp>	i64mat2x3;
+	typedef mat<2, 4, int64, defaultp>	i64mat2x4;
+	typedef mat<3, 2, int64, defaultp>	i64mat3x2;
+	typedef mat<3, 3, int64, defaultp>	i64mat3x3;
+	typedef mat<3, 4, int64, defaultp>	i64mat3x4;
+	typedef mat<4, 2, int64, defaultp>	i64mat4x2;
+	typedef mat<4, 3, int64, defaultp>	i64mat4x3;
+	typedef mat<4, 4, int64, defaultp>	i64mat4x4;
+
+
+	// Unsigned integer matrix MxN
+
+	typedef mat<2, 2, uint, lowp>		lowp_umat2x2;
+	typedef mat<2, 3, uint, lowp>		lowp_umat2x3;
+	typedef mat<2, 4, uint, lowp>		lowp_umat2x4;
+	typedef mat<3, 2, uint, lowp>		lowp_umat3x2;
+	typedef mat<3, 3, uint, lowp>		lowp_umat3x3;
+	typedef mat<3, 4, uint, lowp>		lowp_umat3x4;
+	typedef mat<4, 2, uint, lowp>		lowp_umat4x2;
+	typedef mat<4, 3, uint, lowp>		lowp_umat4x3;
+	typedef mat<4, 4, uint, lowp>		lowp_umat4x4;
+
+	typedef mat<2, 2, uint, mediump>	mediump_umat2x2;
+	typedef mat<2, 3, uint, mediump>	mediump_umat2x3;
+	typedef mat<2, 4, uint, mediump>	mediump_umat2x4;
+	typedef mat<3, 2, uint, mediump>	mediump_umat3x2;
+	typedef mat<3, 3, uint, mediump>	mediump_umat3x3;
+	typedef mat<3, 4, uint, mediump>	mediump_umat3x4;
+	typedef mat<4, 2, uint, mediump>	mediump_umat4x2;
+	typedef mat<4, 3, uint, mediump>	mediump_umat4x3;
+	typedef mat<4, 4, uint, mediump>	mediump_umat4x4;
+
+	typedef mat<2, 2, uint, highp>		highp_umat2x2;
+	typedef mat<2, 3, uint, highp>		highp_umat2x3;
+	typedef mat<2, 4, uint, highp>		highp_umat2x4;
+	typedef mat<3, 2, uint, highp>		highp_umat3x2;
+	typedef mat<3, 3, uint, highp>		highp_umat3x3;
+	typedef mat<3, 4, uint, highp>		highp_umat3x4;
+	typedef mat<4, 2, uint, highp>		highp_umat4x2;
+	typedef mat<4, 3, uint, highp>		highp_umat4x3;
+	typedef mat<4, 4, uint, highp>		highp_umat4x4;
+
+	typedef mat<2, 2, uint, defaultp>	umat2x2;
+	typedef mat<2, 3, uint, defaultp>	umat2x3;
+	typedef mat<2, 4, uint, defaultp>	umat2x4;
+	typedef mat<3, 2, uint, defaultp>	umat3x2;
+	typedef mat<3, 3, uint, defaultp>	umat3x3;
+	typedef mat<3, 4, uint, defaultp>	umat3x4;
+	typedef mat<4, 2, uint, defaultp>	umat4x2;
+	typedef mat<4, 3, uint, defaultp>	umat4x3;
+	typedef mat<4, 4, uint, defaultp>	umat4x4;
+
+
+	typedef mat<2, 2, uint8, lowp>		lowp_u8mat2x2;
+	typedef mat<2, 3, uint8, lowp>		lowp_u8mat2x3;
+	typedef mat<2, 4, uint8, lowp>		lowp_u8mat2x4;
+	typedef mat<3, 2, uint8, lowp>		lowp_u8mat3x2;
+	typedef mat<3, 3, uint8, lowp>		lowp_u8mat3x3;
+	typedef mat<3, 4, uint8, lowp>		lowp_u8mat3x4;
+	typedef mat<4, 2, uint8, lowp>		lowp_u8mat4x2;
+	typedef mat<4, 3, uint8, lowp>		lowp_u8mat4x3;
+	typedef mat<4, 4, uint8, lowp>		lowp_u8mat4x4;
+
+	typedef mat<2, 2, uint8, mediump>	mediump_u8mat2x2;
+	typedef mat<2, 3, uint8, mediump>	mediump_u8mat2x3;
+	typedef mat<2, 4, uint8, mediump>	mediump_u8mat2x4;
+	typedef mat<3, 2, uint8, mediump>	mediump_u8mat3x2;
+	typedef mat<3, 3, uint8, mediump>	mediump_u8mat3x3;
+	typedef mat<3, 4, uint8, mediump>	mediump_u8mat3x4;
+	typedef mat<4, 2, uint8, mediump>	mediump_u8mat4x2;
+	typedef mat<4, 3, uint8, mediump>	mediump_u8mat4x3;
+	typedef mat<4, 4, uint8, mediump>	mediump_u8mat4x4;
+
+	typedef mat<2, 2, uint8, highp>		highp_u8mat2x2;
+	typedef mat<2, 3, uint8, highp>		highp_u8mat2x3;
+	typedef mat<2, 4, uint8, highp>		highp_u8mat2x4;
+	typedef mat<3, 2, uint8, highp>		highp_u8mat3x2;
+	typedef mat<3, 3, uint8, highp>		highp_u8mat3x3;
+	typedef mat<3, 4, uint8, highp>		highp_u8mat3x4;
+	typedef mat<4, 2, uint8, highp>		highp_u8mat4x2;
+	typedef mat<4, 3, uint8, highp>		highp_u8mat4x3;
+	typedef mat<4, 4, uint8, highp>		highp_u8mat4x4;
+
+	typedef mat<2, 2, uint8, defaultp>	u8mat2x2;
+	typedef mat<2, 3, uint8, defaultp>	u8mat2x3;
+	typedef mat<2, 4, uint8, defaultp>	u8mat2x4;
+	typedef mat<3, 2, uint8, defaultp>	u8mat3x2;
+	typedef mat<3, 3, uint8, defaultp>	u8mat3x3;
+	typedef mat<3, 4, uint8, defaultp>	u8mat3x4;
+	typedef mat<4, 2, uint8, defaultp>	u8mat4x2;
+	typedef mat<4, 3, uint8, defaultp>	u8mat4x3;
+	typedef mat<4, 4, uint8, defaultp>	u8mat4x4;
+
+
+	typedef mat<2, 2, uint16, lowp>		lowp_u16mat2x2;
+	typedef mat<2, 3, uint16, lowp>		lowp_u16mat2x3;
+	typedef mat<2, 4, uint16, lowp>		lowp_u16mat2x4;
+	typedef mat<3, 2, uint16, lowp>		lowp_u16mat3x2;
+	typedef mat<3, 3, uint16, lowp>		lowp_u16mat3x3;
+	typedef mat<3, 4, uint16, lowp>		lowp_u16mat3x4;
+	typedef mat<4, 2, uint16, lowp>		lowp_u16mat4x2;
+	typedef mat<4, 3, uint16, lowp>		lowp_u16mat4x3;
+	typedef mat<4, 4, uint16, lowp>		lowp_u16mat4x4;
+
+	typedef mat<2, 2, uint16, mediump>	mediump_u16mat2x2;
+	typedef mat<2, 3, uint16, mediump>	mediump_u16mat2x3;
+	typedef mat<2, 4, uint16, mediump>	mediump_u16mat2x4;
+	typedef mat<3, 2, uint16, mediump>	mediump_u16mat3x2;
+	typedef mat<3, 3, uint16, mediump>	mediump_u16mat3x3;
+	typedef mat<3, 4, uint16, mediump>	mediump_u16mat3x4;
+	typedef mat<4, 2, uint16, mediump>	mediump_u16mat4x2;
+	typedef mat<4, 3, uint16, mediump>	mediump_u16mat4x3;
+	typedef mat<4, 4, uint16, mediump>	mediump_u16mat4x4;
+
+	typedef mat<2, 2, uint16, highp>	highp_u16mat2x2;
+	typedef mat<2, 3, uint16, highp>	highp_u16mat2x3;
+	typedef mat<2, 4, uint16, highp>	highp_u16mat2x4;
+	typedef mat<3, 2, uint16, highp>	highp_u16mat3x2;
+	typedef mat<3, 3, uint16, highp>	highp_u16mat3x3;
+	typedef mat<3, 4, uint16, highp>	highp_u16mat3x4;
+	typedef mat<4, 2, uint16, highp>	highp_u16mat4x2;
+	typedef mat<4, 3, uint16, highp>	highp_u16mat4x3;
+	typedef mat<4, 4, uint16, highp>	highp_u16mat4x4;
+
+	typedef mat<2, 2, uint16, defaultp>	u16mat2x2;
+	typedef mat<2, 3, uint16, defaultp>	u16mat2x3;
+	typedef mat<2, 4, uint16, defaultp>	u16mat2x4;
+	typedef mat<3, 2, uint16, defaultp>	u16mat3x2;
+	typedef mat<3, 3, uint16, defaultp>	u16mat3x3;
+	typedef mat<3, 4, uint16, defaultp>	u16mat3x4;
+	typedef mat<4, 2, uint16, defaultp>	u16mat4x2;
+	typedef mat<4, 3, uint16, defaultp>	u16mat4x3;
+	typedef mat<4, 4, uint16, defaultp>	u16mat4x4;
+
+
+	typedef mat<2, 2, uint32, lowp>		lowp_u32mat2x2;
+	typedef mat<2, 3, uint32, lowp>		lowp_u32mat2x3;
+	typedef mat<2, 4, uint32, lowp>		lowp_u32mat2x4;
+	typedef mat<3, 2, uint32, lowp>		lowp_u32mat3x2;
+	typedef mat<3, 3, uint32, lowp>		lowp_u32mat3x3;
+	typedef mat<3, 4, uint32, lowp>		lowp_u32mat3x4;
+	typedef mat<4, 2, uint32, lowp>		lowp_u32mat4x2;
+	typedef mat<4, 3, uint32, lowp>		lowp_u32mat4x3;
+	typedef mat<4, 4, uint32, lowp>		lowp_u32mat4x4;
+
+	typedef mat<2, 2, uint32, mediump>	mediump_u32mat2x2;
+	typedef mat<2, 3, uint32, mediump>	mediump_u32mat2x3;
+	typedef mat<2, 4, uint32, mediump>	mediump_u32mat2x4;
+	typedef mat<3, 2, uint32, mediump>	mediump_u32mat3x2;
+	typedef mat<3, 3, uint32, mediump>	mediump_u32mat3x3;
+	typedef mat<3, 4, uint32, mediump>	mediump_u32mat3x4;
+	typedef mat<4, 2, uint32, mediump>	mediump_u32mat4x2;
+	typedef mat<4, 3, uint32, mediump>	mediump_u32mat4x3;
+	typedef mat<4, 4, uint32, mediump>	mediump_u32mat4x4;
+
+	typedef mat<2, 2, uint32, highp>	highp_u32mat2x2;
+	typedef mat<2, 3, uint32, highp>	highp_u32mat2x3;
+	typedef mat<2, 4, uint32, highp>	highp_u32mat2x4;
+	typedef mat<3, 2, uint32, highp>	highp_u32mat3x2;
+	typedef mat<3, 3, uint32, highp>	highp_u32mat3x3;
+	typedef mat<3, 4, uint32, highp>	highp_u32mat3x4;
+	typedef mat<4, 2, uint32, highp>	highp_u32mat4x2;
+	typedef mat<4, 3, uint32, highp>	highp_u32mat4x3;
+	typedef mat<4, 4, uint32, highp>	highp_u32mat4x4;
+
+	typedef mat<2, 2, uint32, defaultp>	u32mat2x2;
+	typedef mat<2, 3, uint32, defaultp>	u32mat2x3;
+	typedef mat<2, 4, uint32, defaultp>	u32mat2x4;
+	typedef mat<3, 2, uint32, defaultp>	u32mat3x2;
+	typedef mat<3, 3, uint32, defaultp>	u32mat3x3;
+	typedef mat<3, 4, uint32, defaultp>	u32mat3x4;
+	typedef mat<4, 2, uint32, defaultp>	u32mat4x2;
+	typedef mat<4, 3, uint32, defaultp>	u32mat4x3;
+	typedef mat<4, 4, uint32, defaultp>	u32mat4x4;
+
+
+	typedef mat<2, 2, uint64, lowp>		lowp_u64mat2x2;
+	typedef mat<2, 3, uint64, lowp>		lowp_u64mat2x3;
+	typedef mat<2, 4, uint64, lowp>		lowp_u64mat2x4;
+	typedef mat<3, 2, uint64, lowp>		lowp_u64mat3x2;
+	typedef mat<3, 3, uint64, lowp>		lowp_u64mat3x3;
+	typedef mat<3, 4, uint64, lowp>		lowp_u64mat3x4;
+	typedef mat<4, 2, uint64, lowp>		lowp_u64mat4x2;
+	typedef mat<4, 3, uint64, lowp>		lowp_u64mat4x3;
+	typedef mat<4, 4, uint64, lowp>		lowp_u64mat4x4;
+
+	typedef mat<2, 2, uint64, mediump>	mediump_u64mat2x2;
+	typedef mat<2, 3, uint64, mediump>	mediump_u64mat2x3;
+	typedef mat<2, 4, uint64, mediump>	mediump_u64mat2x4;
+	typedef mat<3, 2, uint64, mediump>	mediump_u64mat3x2;
+	typedef mat<3, 3, uint64, mediump>	mediump_u64mat3x3;
+	typedef mat<3, 4, uint64, mediump>	mediump_u64mat3x4;
+	typedef mat<4, 2, uint64, mediump>	mediump_u64mat4x2;
+	typedef mat<4, 3, uint64, mediump>	mediump_u64mat4x3;
+	typedef mat<4, 4, uint64, mediump>	mediump_u64mat4x4;
+
+	typedef mat<2, 2, uint64, highp>	highp_u64mat2x2;
+	typedef mat<2, 3, uint64, highp>	highp_u64mat2x3;
+	typedef mat<2, 4, uint64, highp>	highp_u64mat2x4;
+	typedef mat<3, 2, uint64, highp>	highp_u64mat3x2;
+	typedef mat<3, 3, uint64, highp>	highp_u64mat3x3;
+	typedef mat<3, 4, uint64, highp>	highp_u64mat3x4;
+	typedef mat<4, 2, uint64, highp>	highp_u64mat4x2;
+	typedef mat<4, 3, uint64, highp>	highp_u64mat4x3;
+	typedef mat<4, 4, uint64, highp>	highp_u64mat4x4;
+
+	typedef mat<2, 2, uint64, defaultp>	u64mat2x2;
+	typedef mat<2, 3, uint64, defaultp>	u64mat2x3;
+	typedef mat<2, 4, uint64, defaultp>	u64mat2x4;
+	typedef mat<3, 2, uint64, defaultp>	u64mat3x2;
+	typedef mat<3, 3, uint64, defaultp>	u64mat3x3;
+	typedef mat<3, 4, uint64, defaultp>	u64mat3x4;
+	typedef mat<4, 2, uint64, defaultp>	u64mat4x2;
+	typedef mat<4, 3, uint64, defaultp>	u64mat4x3;
+	typedef mat<4, 4, uint64, defaultp>	u64mat4x4;
+
+	// Quaternion
+
+	typedef qua<float, lowp>			lowp_quat;
+	typedef qua<float, mediump>			mediump_quat;
+	typedef qua<float, highp>			highp_quat;
+	typedef qua<float, defaultp>		quat;
+
+	typedef qua<float, lowp>			lowp_fquat;
+	typedef qua<float, mediump>			mediump_fquat;
+	typedef qua<float, highp>			highp_fquat;
+	typedef qua<float, defaultp>		fquat;
+
+	typedef qua<f32, lowp>				lowp_f32quat;
+	typedef qua<f32, mediump>			mediump_f32quat;
+	typedef qua<f32, highp>				highp_f32quat;
+	typedef qua<f32, defaultp>			f32quat;
+
+	typedef qua<double, lowp>			lowp_dquat;
+	typedef qua<double, mediump>		mediump_dquat;
+	typedef qua<double, highp>			highp_dquat;
+	typedef qua<double, defaultp>		dquat;
+
+	typedef qua<f64, lowp>				lowp_f64quat;
+	typedef qua<f64, mediump>			mediump_f64quat;
+	typedef qua<f64, highp>				highp_f64quat;
+	typedef qua<f64, defaultp>			f64quat;
+}//namespace glm
+
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/geometric.hpp b/thirdparty/manifold/thirdparty/glm/glm/geometric.hpp
new file mode 100644
index 000000000000..ac857e69d4cf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/geometric.hpp
@@ -0,0 +1,116 @@
+/// @ref core
+/// @file glm/geometric.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+///
+/// @defgroup core_func_geometric Geometric functions
+/// @ingroup core
+///
+/// These operate on vectors as vectors, not component-wise.
+///
+/// Include <glm/geometric.hpp> to use these core features.
+
+#pragma once
+
+#include "detail/type_vec3.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_func_geometric
+	/// @{
+
+	/// Returns the length of x, i.e., sqrt(x * x).
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/length.xml">GLSL length man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T length(vec<L, T, Q> const& x);
+
+	/// Returns the distance between p0 and p1, i.e., length(p0 - p1).
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/distance.xml">GLSL distance man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T distance(vec<L, T, Q> const& p0, vec<L, T, Q> const& p1);
+
+	/// Returns the dot product of x and y, i.e., result = x * y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/dot.xml">GLSL dot man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR T dot(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the cross product of x and y.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/cross.xml">GLSL cross man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> cross(vec<3, T, Q> const& x, vec<3, T, Q> const& y);
+
+	/// Returns a vector in the same direction as x but with length of 1.
+	/// According to issue 10 GLSL 1.10 specification, if length(x) == 0 then result is undefined and generate an error.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/normalize.xml">GLSL normalize man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> normalize(vec<L, T, Q> const& x);
+
+	/// If dot(Nref, I) < 0.0, return N, otherwise, return -N.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/faceforward.xml">GLSL faceforward man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> faceforward(
+		vec<L, T, Q> const& N,
+		vec<L, T, Q> const& I,
+		vec<L, T, Q> const& Nref);
+
+	/// For the incident vector I and surface orientation N,
+	/// returns the reflection direction : result = I - 2.0 * dot(N, I) * N.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/reflect.xml">GLSL reflect man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> reflect(
+		vec<L, T, Q> const& I,
+		vec<L, T, Q> const& N);
+
+	/// For the incident vector I and surface normal N,
+	/// and the ratio of indices of refraction eta,
+	/// return the refraction vector.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/refract.xml">GLSL refract man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.5 Geometric Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> refract(
+		vec<L, T, Q> const& I,
+		vec<L, T, Q> const& N,
+		T eta);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_geometric.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/glm.cppm b/thirdparty/manifold/thirdparty/glm/glm/glm.cppm
new file mode 100644
index 000000000000..85e946e0f83c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/glm.cppm
@@ -0,0 +1,2675 @@
+module;
+
+// #define GLM_GTC_INLINE_NAMESPACE to inline glm::gtc into glm
+// #define GLM_EXT_INLINE_NAMESPACE to inline glm::ext into glm
+// #define GLM_GTX_INLINE_NAMESPACE to inline glm::gtx into glm
+
+#include <glm/glm.hpp>
+#include <glm/ext.hpp>
+
+export module glm;
+
+export namespace glm {
+	// Base types
+	using glm::qualifier;
+	using glm::precision;
+	using glm::vec;
+	using glm::mat;
+	using glm::qua;
+#	if GLM_HAS_TEMPLATE_ALIASES
+	using glm::tvec1;
+	using glm::tvec2;
+	using glm::tvec3;
+	using glm::tvec4;
+	using glm::tmat2x2;
+	using glm::tmat2x3;
+	using glm::tmat2x4;
+	using glm::tmat3x2;
+	using glm::tmat3x3;
+	using glm::tmat3x4;
+	using glm::tmat4x2;
+	using glm::tmat4x3;
+	using glm::tmat4x4;
+	using glm::tquat;
+#	endif
+
+	using glm::int8;
+	using glm::int16;
+	using glm::int32;
+	using glm::int64;
+	using glm::uint8;
+	using glm::uint16;
+	using glm::uint32;
+	using glm::uint64;
+	using glm::lowp_i8;
+	using glm::mediump_i8;
+	using glm::highp_i8;
+	using glm::i8;
+	using glm::lowp_int8;
+	using glm::mediump_int8;
+	using glm::highp_int8;
+	using glm::lowp_int8_t;
+	using glm::mediump_int8_t;
+	using glm::highp_int8_t;
+	using glm::int8_t;
+	using glm::lowp_i16;
+	using glm::mediump_i16;
+	using glm::highp_i16;
+	using glm::i16;
+	using glm::lowp_int16;
+	using glm::mediump_int16;
+	using glm::highp_int16;
+	using glm::lowp_int16_t;
+	using glm::mediump_int16_t;
+	using glm::highp_int16_t;
+	using glm::int16_t;
+	using glm::lowp_i32;
+	using glm::mediump_i32;
+	using glm::highp_i32;
+	using glm::i32;
+	using glm::lowp_int32;
+	using glm::mediump_int32;
+	using glm::highp_int32;
+	using glm::lowp_int32_t;
+	using glm::mediump_int32_t;
+	using glm::highp_int32_t;
+	using glm::int32_t;
+	using glm::lowp_i64;
+	using glm::mediump_i64;
+	using glm::highp_i64;
+	using glm::i64;
+	using glm::lowp_int64;
+	using glm::mediump_int64;
+	using glm::highp_int64;
+	using glm::lowp_int64_t;
+	using glm::mediump_int64_t;
+	using glm::highp_int64_t;
+	using glm::int64_t;
+	using glm::uint;
+	using glm::lowp_u8;
+	using glm::mediump_u8;
+	using glm::highp_u8;
+	using glm::u8;
+	using glm::lowp_uint8;
+	using glm::mediump_uint8;
+	using glm::highp_uint8;
+	using glm::lowp_uint8_t;
+	using glm::mediump_uint8_t;
+	using glm::highp_uint8_t;
+	using glm::uint8_t;
+	using glm::lowp_u16;
+	using glm::mediump_u16;
+	using glm::highp_u16;
+	using glm::u16;
+	using glm::lowp_uint16;
+	using glm::mediump_uint16;
+	using glm::highp_uint16;
+	using glm::lowp_uint16_t;
+	using glm::mediump_uint16_t;
+	using glm::highp_uint16_t;
+	using glm::uint16_t;
+	using glm::lowp_u32;
+	using glm::mediump_u32;
+	using glm::highp_u32;
+	using glm::u32;
+	using glm::lowp_uint32;
+	using glm::mediump_uint32;
+	using glm::highp_uint32;
+	using glm::lowp_uint32_t;
+	using glm::mediump_uint32_t;
+	using glm::highp_uint32_t;
+	using glm::uint32_t;
+	using glm::lowp_u64;
+	using glm::mediump_u64;
+	using glm::highp_u64;
+	using glm::u64;
+	using glm::lowp_uint64;
+	using glm::mediump_uint64;
+	using glm::highp_uint64;
+	using glm::lowp_uint64_t;
+	using glm::mediump_uint64_t;
+	using glm::highp_uint64_t;
+	using glm::uint64_t;
+	using glm::lowp_f32;
+	using glm::mediump_f32;
+	using glm::highp_f32;
+	using glm::f32;
+	using glm::lowp_float32;
+	using glm::mediump_float32;
+	using glm::highp_float32;
+	using glm::float32;
+	using glm::lowp_float32_t;
+	using glm::mediump_float32_t;
+	using glm::highp_float32_t;
+	using glm::float32_t;
+	using glm::lowp_f64;
+	using glm::mediump_f64;
+	using glm::highp_f64;
+	using glm::f64;
+	using glm::lowp_float64;
+	using glm::mediump_float64;
+	using glm::highp_float64;
+	using glm::float64;
+	using glm::lowp_float64_t;
+	using glm::mediump_float64_t;
+	using glm::highp_float64_t;
+	using glm::float64_t;
+	using glm::lowp_bvec1;
+	using glm::lowp_bvec2;
+	using glm::lowp_bvec3;
+	using glm::lowp_bvec4;
+	using glm::mediump_bvec1;
+	using glm::mediump_bvec2;
+	using glm::mediump_bvec3;
+	using glm::mediump_bvec4;
+	using glm::highp_bvec1;
+	using glm::highp_bvec2;
+	using glm::highp_bvec3;
+	using glm::highp_bvec4;
+	using glm::bvec1;
+	using glm::bvec2;
+	using glm::bvec3;
+	using glm::bvec4;
+	using glm::lowp_ivec1;
+	using glm::lowp_ivec2;
+	using glm::lowp_ivec3;
+	using glm::lowp_ivec4;
+	using glm::mediump_ivec1;
+	using glm::mediump_ivec2;
+	using glm::mediump_ivec3;
+	using glm::mediump_ivec4;
+	using glm::highp_ivec1;
+	using glm::highp_ivec2;
+	using glm::highp_ivec3;
+	using glm::highp_ivec4;
+	using glm::ivec1;
+	using glm::ivec2;
+	using glm::ivec3;
+	using glm::ivec4;
+	using glm::lowp_i8vec1;
+	using glm::lowp_i8vec2;
+	using glm::lowp_i8vec3;
+	using glm::lowp_i8vec4;
+	using glm::mediump_i8vec1;
+	using glm::mediump_i8vec2;
+	using glm::mediump_i8vec3;
+	using glm::mediump_i8vec4;
+	using glm::highp_i8vec1;
+	using glm::highp_i8vec2;
+	using glm::highp_i8vec3;
+	using glm::highp_i8vec4;
+	using glm::i8vec1;
+	using glm::i8vec2;
+	using glm::i8vec3;
+	using glm::i8vec4;
+	using glm::lowp_i16vec1;
+	using glm::lowp_i16vec2;
+	using glm::lowp_i16vec3;
+	using glm::lowp_i16vec4;
+	using glm::mediump_i16vec1;
+	using glm::mediump_i16vec2;
+	using glm::mediump_i16vec3;
+	using glm::mediump_i16vec4;
+	using glm::highp_i16vec1;
+	using glm::highp_i16vec2;
+	using glm::highp_i16vec3;
+	using glm::highp_i16vec4;
+	using glm::i16vec1;
+	using glm::i16vec2;
+	using glm::i16vec3;
+	using glm::i16vec4;
+	using glm::lowp_i32vec1;
+	using glm::lowp_i32vec2;
+	using glm::lowp_i32vec3;
+	using glm::lowp_i32vec4;
+	using glm::mediump_i32vec1;
+	using glm::mediump_i32vec2;
+	using glm::mediump_i32vec3;
+	using glm::mediump_i32vec4;
+	using glm::highp_i32vec1;
+	using glm::highp_i32vec2;
+	using glm::highp_i32vec3;
+	using glm::highp_i32vec4;
+	using glm::i32vec1;
+	using glm::i32vec2;
+	using glm::i32vec3;
+	using glm::i32vec4;
+	using glm::lowp_i64vec1;
+	using glm::lowp_i64vec2;
+	using glm::lowp_i64vec3;
+	using glm::lowp_i64vec4;
+	using glm::mediump_i64vec1;
+	using glm::mediump_i64vec2;
+	using glm::mediump_i64vec3;
+	using glm::mediump_i64vec4;
+	using glm::highp_i64vec1;
+	using glm::highp_i64vec2;
+	using glm::highp_i64vec3;
+	using glm::highp_i64vec4;
+	using glm::i64vec1;
+	using glm::i64vec2;
+	using glm::i64vec3;
+	using glm::i64vec4;
+	using glm::lowp_uvec1;
+	using glm::lowp_uvec2;
+	using glm::lowp_uvec3;
+	using glm::lowp_uvec4;
+	using glm::mediump_uvec1;
+	using glm::mediump_uvec2;
+	using glm::mediump_uvec3;
+	using glm::mediump_uvec4;
+	using glm::highp_uvec1;
+	using glm::highp_uvec2;
+	using glm::highp_uvec3;
+	using glm::highp_uvec4;
+	using glm::uvec1;
+	using glm::uvec2;
+	using glm::uvec3;
+	using glm::uvec4;
+	using glm::lowp_u8vec1;
+	using glm::lowp_u8vec2;
+	using glm::lowp_u8vec3;
+	using glm::lowp_u8vec4;
+	using glm::mediump_u8vec1;
+	using glm::mediump_u8vec2;
+	using glm::mediump_u8vec3;
+	using glm::mediump_u8vec4;
+	using glm::highp_u8vec1;
+	using glm::highp_u8vec2;
+	using glm::highp_u8vec3;
+	using glm::highp_u8vec4;
+	using glm::u8vec1;
+	using glm::u8vec2;
+	using glm::u8vec3;
+	using glm::u8vec4;
+	using glm::lowp_u16vec1;
+	using glm::lowp_u16vec2;
+	using glm::lowp_u16vec3;
+	using glm::lowp_u16vec4;
+	using glm::mediump_u16vec1;
+	using glm::mediump_u16vec2;
+	using glm::mediump_u16vec3;
+	using glm::mediump_u16vec4;
+	using glm::highp_u16vec1;
+	using glm::highp_u16vec2;
+	using glm::highp_u16vec3;
+	using glm::highp_u16vec4;
+	using glm::u16vec1;
+	using glm::u16vec2;
+	using glm::u16vec3;
+	using glm::u16vec4;
+	using glm::lowp_u32vec1;
+	using glm::lowp_u32vec2;
+	using glm::lowp_u32vec3;
+	using glm::lowp_u32vec4;
+	using glm::mediump_u32vec1;
+	using glm::mediump_u32vec2;
+	using glm::mediump_u32vec3;
+	using glm::mediump_u32vec4;
+	using glm::highp_u32vec1;
+	using glm::highp_u32vec2;
+	using glm::highp_u32vec3;
+	using glm::highp_u32vec4;
+	using glm::u32vec1;
+	using glm::u32vec2;
+	using glm::u32vec3;
+	using glm::u32vec4;
+	using glm::lowp_u64vec1;
+	using glm::lowp_u64vec2;
+	using glm::lowp_u64vec3;
+	using glm::lowp_u64vec4;
+	using glm::mediump_u64vec1;
+	using glm::mediump_u64vec2;
+	using glm::mediump_u64vec3;
+	using glm::mediump_u64vec4;
+	using glm::highp_u64vec1;
+	using glm::highp_u64vec2;
+	using glm::highp_u64vec3;
+	using glm::highp_u64vec4;
+	using glm::u64vec1;
+	using glm::u64vec2;
+	using glm::u64vec3;
+	using glm::u64vec4;
+	using glm::lowp_vec1;
+	using glm::lowp_vec2;
+	using glm::lowp_vec3;
+	using glm::lowp_vec4;
+	using glm::mediump_vec1;
+	using glm::mediump_vec2;
+	using glm::mediump_vec3;
+	using glm::mediump_vec4;
+	using glm::highp_vec1;
+	using glm::highp_vec2;
+	using glm::highp_vec3;
+	using glm::highp_vec4;
+	using glm::vec1;
+	using glm::vec2;
+	using glm::vec3;
+	using glm::vec4;
+	using glm::lowp_fvec1;
+	using glm::lowp_fvec2;
+	using glm::lowp_fvec3;
+	using glm::lowp_fvec4;
+	using glm::mediump_fvec1;
+	using glm::mediump_fvec2;
+	using glm::mediump_fvec3;
+	using glm::mediump_fvec4;
+	using glm::highp_fvec1;
+	using glm::highp_fvec2;
+	using glm::highp_fvec3;
+	using glm::highp_fvec4;
+	using glm::fvec1;
+	using glm::fvec2;
+	using glm::fvec3;
+	using glm::fvec4;
+	using glm::lowp_f32vec1;
+	using glm::lowp_f32vec2;
+	using glm::lowp_f32vec3;
+	using glm::lowp_f32vec4;
+	using glm::mediump_f32vec1;
+	using glm::mediump_f32vec2;
+	using glm::mediump_f32vec3;
+	using glm::mediump_f32vec4;
+	using glm::highp_f32vec1;
+	using glm::highp_f32vec2;
+	using glm::highp_f32vec3;
+	using glm::highp_f32vec4;
+	using glm::f32vec1;
+	using glm::f32vec2;
+	using glm::f32vec3;
+	using glm::f32vec4;
+	using glm::lowp_dvec1;
+	using glm::lowp_dvec2;
+	using glm::lowp_dvec3;
+	using glm::lowp_dvec4;
+	using glm::mediump_dvec1;
+	using glm::mediump_dvec2;
+	using glm::mediump_dvec3;
+	using glm::mediump_dvec4;
+	using glm::highp_dvec1;
+	using glm::highp_dvec2;
+	using glm::highp_dvec3;
+	using glm::highp_dvec4;
+	using glm::dvec1;
+	using glm::dvec2;
+	using glm::dvec3;
+	using glm::dvec4;
+	using glm::lowp_f64vec1;
+	using glm::lowp_f64vec2;
+	using glm::lowp_f64vec3;
+	using glm::lowp_f64vec4;
+	using glm::mediump_f64vec1;
+	using glm::mediump_f64vec2;
+	using glm::mediump_f64vec3;
+	using glm::mediump_f64vec4;
+	using glm::highp_f64vec1;
+	using glm::highp_f64vec2;
+	using glm::highp_f64vec3;
+	using glm::highp_f64vec4;
+	using glm::f64vec1;
+	using glm::f64vec2;
+	using glm::f64vec3;
+	using glm::f64vec4;
+	using glm::lowp_mat2;
+	using glm::lowp_mat3;
+	using glm::lowp_mat4;
+	using glm::mediump_mat2;
+	using glm::mediump_mat3;
+	using glm::mediump_mat4;
+	using glm::highp_mat2;
+	using glm::highp_mat3;
+	using glm::highp_mat4;
+	using glm::mat2;
+	using glm::mat3;
+	using glm::mat4;
+	using glm::lowp_fmat2;
+	using glm::lowp_fmat3;
+	using glm::lowp_fmat4;
+	using glm::mediump_fmat2;
+	using glm::mediump_fmat3;
+	using glm::mediump_fmat4;
+	using glm::highp_fmat2;
+	using glm::highp_fmat3;
+	using glm::highp_fmat4;
+	using glm::fmat2;
+	using glm::fmat3;
+	using glm::fmat4;
+	using glm::lowp_f32mat2;
+	using glm::lowp_f32mat3;
+	using glm::lowp_f32mat4;
+	using glm::mediump_f32mat2;
+	using glm::mediump_f32mat3;
+	using glm::mediump_f32mat4;
+	using glm::highp_f32mat2;
+	using glm::highp_f32mat3;
+	using glm::highp_f32mat4;
+	using glm::f32mat2;
+	using glm::f32mat3;
+	using glm::f32mat4;
+	using glm::lowp_dmat2;
+	using glm::lowp_dmat3;
+	using glm::lowp_dmat4;
+	using glm::mediump_dmat2;
+	using glm::mediump_dmat3;
+	using glm::mediump_dmat4;
+	using glm::highp_dmat2;
+	using glm::highp_dmat3;
+	using glm::highp_dmat4;
+	using glm::dmat2;
+	using glm::dmat3;
+	using glm::dmat4;
+	using glm::lowp_f64mat2;
+	using glm::lowp_f64mat3;
+	using glm::lowp_f64mat4;
+	using glm::mediump_f64mat2;
+	using glm::mediump_f64mat3;
+	using glm::mediump_f64mat4;
+	using glm::highp_f64mat2;
+	using glm::highp_f64mat3;
+	using glm::highp_f64mat4;
+	using glm::f64mat2;
+	using glm::f64mat3;
+	using glm::f64mat4;
+	using glm::lowp_mat2x2;
+	using glm::lowp_mat2x3;
+	using glm::lowp_mat2x4;
+	using glm::lowp_mat3x2;
+	using glm::lowp_mat3x3;
+	using glm::lowp_mat3x4;
+	using glm::lowp_mat4x2;
+	using glm::lowp_mat4x3;
+	using glm::lowp_mat4x4;
+	using glm::mediump_mat2x2;
+	using glm::mediump_mat2x3;
+	using glm::mediump_mat2x4;
+	using glm::mediump_mat3x2;
+	using glm::mediump_mat3x3;
+	using glm::mediump_mat3x4;
+	using glm::mediump_mat4x2;
+	using glm::mediump_mat4x3;
+	using glm::mediump_mat4x4;
+	using glm::highp_mat2x2;
+	using glm::highp_mat2x3;
+	using glm::highp_mat2x4;
+	using glm::highp_mat3x2;
+	using glm::highp_mat3x3;
+	using glm::highp_mat3x4;
+	using glm::highp_mat4x2;
+	using glm::highp_mat4x3;
+	using glm::highp_mat4x4;
+	using glm::mat2x2;
+	using glm::mat2x3;
+	using glm::mat2x4;
+	using glm::mat3x2;
+	using glm::mat3x3;
+	using glm::mat3x4;
+	using glm::mat4x2;
+	using glm::mat4x3;
+	using glm::mat4x4;
+	using glm::lowp_fmat2x2;
+	using glm::lowp_fmat2x3;
+	using glm::lowp_fmat2x4;
+	using glm::lowp_fmat3x2;
+	using glm::lowp_fmat3x3;
+	using glm::lowp_fmat3x4;
+	using glm::lowp_fmat4x2;
+	using glm::lowp_fmat4x3;
+	using glm::lowp_fmat4x4;
+	using glm::mediump_fmat2x2;
+	using glm::mediump_fmat2x3;
+	using glm::mediump_fmat2x4;
+	using glm::mediump_fmat3x2;
+	using glm::mediump_fmat3x3;
+	using glm::mediump_fmat3x4;
+	using glm::mediump_fmat4x2;
+	using glm::mediump_fmat4x3;
+	using glm::mediump_fmat4x4;
+	using glm::highp_fmat2x2;
+	using glm::highp_fmat2x3;
+	using glm::highp_fmat2x4;
+	using glm::highp_fmat3x2;
+	using glm::highp_fmat3x3;
+	using glm::highp_fmat3x4;
+	using glm::highp_fmat4x2;
+	using glm::highp_fmat4x3;
+	using glm::highp_fmat4x4;
+	using glm::fmat2x2;
+	using glm::fmat2x3;
+	using glm::fmat2x4;
+	using glm::fmat3x2;
+	using glm::fmat3x3;
+	using glm::fmat3x4;
+	using glm::fmat4x2;
+	using glm::fmat4x3;
+	using glm::fmat4x4;
+	using glm::lowp_f32mat2x2;
+	using glm::lowp_f32mat2x3;
+	using glm::lowp_f32mat2x4;
+	using glm::lowp_f32mat3x2;
+	using glm::lowp_f32mat3x3;
+	using glm::lowp_f32mat3x4;
+	using glm::lowp_f32mat4x2;
+	using glm::lowp_f32mat4x3;
+	using glm::lowp_f32mat4x4;
+
+	using glm::mediump_f32mat2x2;
+	using glm::mediump_f32mat2x3;
+	using glm::mediump_f32mat2x4;
+	using glm::mediump_f32mat3x2;
+	using glm::mediump_f32mat3x3;
+	using glm::mediump_f32mat3x4;
+	using glm::mediump_f32mat4x2;
+	using glm::mediump_f32mat4x3;
+	using glm::mediump_f32mat4x4;
+	using glm::highp_f32mat2x2;
+	using glm::highp_f32mat2x3;
+	using glm::highp_f32mat2x4;
+	using glm::highp_f32mat3x2;
+	using glm::highp_f32mat3x3;
+	using glm::highp_f32mat3x4;
+	using glm::highp_f32mat4x2;
+	using glm::highp_f32mat4x3;
+	using glm::highp_f32mat4x4;
+	using glm::f32mat2x2;
+	using glm::f32mat2x3;
+	using glm::f32mat2x4;
+	using glm::f32mat3x2;
+	using glm::f32mat3x3;
+	using glm::f32mat3x4;
+	using glm::f32mat4x2;
+	using glm::f32mat4x3;
+	using glm::f32mat4x4;
+	using glm::lowp_dmat2x2;
+	using glm::lowp_dmat2x3;
+	using glm::lowp_dmat2x4;
+	using glm::lowp_dmat3x2;
+	using glm::lowp_dmat3x3;
+	using glm::lowp_dmat3x4;
+	using glm::lowp_dmat4x2;
+	using glm::lowp_dmat4x3;
+	using glm::lowp_dmat4x4;
+	using glm::mediump_dmat2x2;
+	using glm::mediump_dmat2x3;
+	using glm::mediump_dmat2x4;
+	using glm::mediump_dmat3x2;
+	using glm::mediump_dmat3x3;
+	using glm::mediump_dmat3x4;
+	using glm::mediump_dmat4x2;
+	using glm::mediump_dmat4x3;
+	using glm::mediump_dmat4x4;
+	using glm::highp_dmat2x2;
+	using glm::highp_dmat2x3;
+	using glm::highp_dmat2x4;
+	using glm::highp_dmat3x2;
+	using glm::highp_dmat3x3;
+	using glm::highp_dmat3x4;
+	using glm::highp_dmat4x2;
+	using glm::highp_dmat4x3;
+	using glm::highp_dmat4x4;
+	using glm::dmat2x2;
+	using glm::dmat2x3;
+	using glm::dmat2x4;
+	using glm::dmat3x2;
+	using glm::dmat3x3;
+	using glm::dmat3x4;
+	using glm::dmat4x2;
+	using glm::dmat4x3;
+	using glm::dmat4x4;
+	using glm::lowp_f64mat2x2;
+	using glm::lowp_f64mat2x3;
+	using glm::lowp_f64mat2x4;
+	using glm::lowp_f64mat3x2;
+	using glm::lowp_f64mat3x3;
+	using glm::lowp_f64mat3x4;
+	using glm::lowp_f64mat4x2;
+	using glm::lowp_f64mat4x3;
+	using glm::lowp_f64mat4x4;
+	using glm::mediump_f64mat2x2;
+	using glm::mediump_f64mat2x3;
+	using glm::mediump_f64mat2x4;
+	using glm::mediump_f64mat3x2;
+	using glm::mediump_f64mat3x3;
+	using glm::mediump_f64mat3x4;
+	using glm::mediump_f64mat4x2;
+	using glm::mediump_f64mat4x3;
+	using glm::mediump_f64mat4x4;
+	using glm::highp_f64mat2x2;
+	using glm::highp_f64mat2x3;
+	using glm::highp_f64mat2x4;
+	using glm::highp_f64mat3x2;
+	using glm::highp_f64mat3x3;
+	using glm::highp_f64mat3x4;
+	using glm::highp_f64mat4x2;
+	using glm::highp_f64mat4x3;
+	using glm::highp_f64mat4x4;
+	using glm::f64mat2x2;
+	using glm::f64mat2x3;
+	using glm::f64mat2x4;
+	using glm::f64mat3x2;
+	using glm::f64mat3x3;
+	using glm::f64mat3x4;
+	using glm::f64mat4x2;
+	using glm::f64mat4x3;
+	using glm::f64mat4x4;
+	using glm::lowp_imat2x2;
+	using glm::lowp_imat2x3;
+	using glm::lowp_imat2x4;
+	using glm::lowp_imat3x2;
+	using glm::lowp_imat3x3;
+	using glm::lowp_imat3x4;
+	using glm::lowp_imat4x2;
+	using glm::lowp_imat4x3;
+	using glm::lowp_imat4x4;
+	using glm::mediump_imat2x2;
+	using glm::mediump_imat2x3;
+	using glm::mediump_imat2x4;
+	using glm::mediump_imat3x2;
+	using glm::mediump_imat3x3;
+	using glm::mediump_imat3x4;
+	using glm::mediump_imat4x2;
+	using glm::mediump_imat4x3;
+	using glm::mediump_imat4x4;
+	using glm::highp_imat2x2;
+	using glm::highp_imat2x3;
+	using glm::highp_imat2x4;
+	using glm::highp_imat3x2;
+	using glm::highp_imat3x3;
+	using glm::highp_imat3x4;
+	using glm::highp_imat4x2;
+	using glm::highp_imat4x3;
+	using glm::highp_imat4x4;
+	using glm::imat2x2;
+	using glm::imat2x3;
+	using glm::imat2x4;
+	using glm::imat3x2;
+	using glm::imat3x3;
+	using glm::imat3x4;
+	using glm::imat4x2;
+	using glm::imat4x3;
+	using glm::imat4x4;
+	using glm::lowp_i8mat2x2;
+	using glm::lowp_i8mat2x3;
+	using glm::lowp_i8mat2x4;
+	using glm::lowp_i8mat3x2;
+	using glm::lowp_i8mat3x3;
+	using glm::lowp_i8mat3x4;
+	using glm::lowp_i8mat4x2;
+	using glm::lowp_i8mat4x3;
+	using glm::lowp_i8mat4x4;
+	using glm::mediump_i8mat2x2;
+	using glm::mediump_i8mat2x3;
+	using glm::mediump_i8mat2x4;
+	using glm::mediump_i8mat3x2;
+	using glm::mediump_i8mat3x3;
+	using glm::mediump_i8mat3x4;
+	using glm::mediump_i8mat4x2;
+	using glm::mediump_i8mat4x3;
+	using glm::mediump_i8mat4x4;
+	using glm::highp_i8mat2x2;
+	using glm::highp_i8mat2x3;
+	using glm::highp_i8mat2x4;
+	using glm::highp_i8mat3x2;
+	using glm::highp_i8mat3x3;
+	using glm::highp_i8mat3x4;
+	using glm::highp_i8mat4x2;
+	using glm::highp_i8mat4x3;
+	using glm::highp_i8mat4x4;
+	using glm::i8mat2x2;
+	using glm::i8mat2x3;
+	using glm::i8mat2x4;
+	using glm::i8mat3x2;
+	using glm::i8mat3x3;
+	using glm::i8mat3x4;
+	using glm::i8mat4x2;
+	using glm::i8mat4x3;
+	using glm::i8mat4x4;
+	using glm::lowp_i16mat2x2;
+	using glm::lowp_i16mat2x3;
+	using glm::lowp_i16mat2x4;
+	using glm::lowp_i16mat3x2;
+	using glm::lowp_i16mat3x3;
+	using glm::lowp_i16mat3x4;
+	using glm::lowp_i16mat4x2;
+	using glm::lowp_i16mat4x3;
+	using glm::lowp_i16mat4x4;
+	using glm::mediump_i16mat2x2;
+	using glm::mediump_i16mat2x3;
+	using glm::mediump_i16mat2x4;
+	using glm::mediump_i16mat3x2;
+	using glm::mediump_i16mat3x3;
+	using glm::mediump_i16mat3x4;
+	using glm::mediump_i16mat4x2;
+	using glm::mediump_i16mat4x3;
+	using glm::mediump_i16mat4x4;
+	using glm::highp_i16mat2x2;
+	using glm::highp_i16mat2x3;
+	using glm::highp_i16mat2x4;
+	using glm::highp_i16mat3x2;
+	using glm::highp_i16mat3x3;
+	using glm::highp_i16mat3x4;
+	using glm::highp_i16mat4x2;
+	using glm::highp_i16mat4x3;
+	using glm::highp_i16mat4x4;
+	using glm::i16mat2x2;
+	using glm::i16mat2x3;
+	using glm::i16mat2x4;
+	using glm::i16mat3x2;
+	using glm::i16mat3x3;
+	using glm::i16mat3x4;
+	using glm::i16mat4x2;
+	using glm::i16mat4x3;
+	using glm::i16mat4x4;
+	using glm::lowp_i32mat2x2;
+	using glm::lowp_i32mat2x3;
+	using glm::lowp_i32mat2x4;
+	using glm::lowp_i32mat3x2;
+	using glm::lowp_i32mat3x3;
+	using glm::lowp_i32mat3x4;
+	using glm::lowp_i32mat4x2;
+	using glm::lowp_i32mat4x3;
+	using glm::lowp_i32mat4x4;
+	using glm::mediump_i32mat2x2;
+	using glm::mediump_i32mat2x3;
+	using glm::mediump_i32mat2x4;
+	using glm::mediump_i32mat3x2;
+	using glm::mediump_i32mat3x3;
+	using glm::mediump_i32mat3x4;
+	using glm::mediump_i32mat4x2;
+	using glm::mediump_i32mat4x3;
+	using glm::mediump_i32mat4x4;
+	using glm::highp_i32mat2x2;
+	using glm::highp_i32mat2x3;
+	using glm::highp_i32mat2x4;
+	using glm::highp_i32mat3x2;
+	using glm::highp_i32mat3x3;
+	using glm::highp_i32mat3x4;
+	using glm::highp_i32mat4x2;
+	using glm::highp_i32mat4x3;
+	using glm::highp_i32mat4x4;
+	using glm::i32mat2x2;
+	using glm::i32mat2x3;
+	using glm::i32mat2x4;
+	using glm::i32mat3x2;
+	using glm::i32mat3x3;
+	using glm::i32mat3x4;
+	using glm::i32mat4x2;
+	using glm::i32mat4x3;
+	using glm::i32mat4x4;
+	using glm::lowp_i64mat2x2;
+	using glm::lowp_i64mat2x3;
+	using glm::lowp_i64mat2x4;
+	using glm::lowp_i64mat3x2;
+	using glm::lowp_i64mat3x3;
+	using glm::lowp_i64mat3x4;
+	using glm::lowp_i64mat4x2;
+	using glm::lowp_i64mat4x3;
+	using glm::lowp_i64mat4x4;
+	using glm::mediump_i64mat2x2;
+	using glm::mediump_i64mat2x3;
+	using glm::mediump_i64mat2x4;
+	using glm::mediump_i64mat3x2;
+	using glm::mediump_i64mat3x3;
+	using glm::mediump_i64mat3x4;
+	using glm::mediump_i64mat4x2;
+	using glm::mediump_i64mat4x3;
+	using glm::mediump_i64mat4x4;
+	using glm::highp_i64mat2x2;
+	using glm::highp_i64mat2x3;
+	using glm::highp_i64mat2x4;
+	using glm::highp_i64mat3x2;
+	using glm::highp_i64mat3x3;
+	using glm::highp_i64mat3x4;
+	using glm::highp_i64mat4x2;
+	using glm::highp_i64mat4x3;
+	using glm::highp_i64mat4x4;
+	using glm::i64mat2x2;
+	using glm::i64mat2x3;
+	using glm::i64mat2x4;
+	using glm::i64mat3x2;
+	using glm::i64mat3x3;
+	using glm::i64mat3x4;
+	using glm::i64mat4x2;
+	using glm::i64mat4x3;
+	using glm::i64mat4x4;
+	using glm::lowp_umat2x2;
+	using glm::lowp_umat2x3;
+	using glm::lowp_umat2x4;
+	using glm::lowp_umat3x2;
+	using glm::lowp_umat3x3;
+	using glm::lowp_umat3x4;
+	using glm::lowp_umat4x2;
+	using glm::lowp_umat4x3;
+	using glm::lowp_umat4x4;
+	using glm::mediump_umat2x2;
+	using glm::mediump_umat2x3;
+	using glm::mediump_umat2x4;
+	using glm::mediump_umat3x2;
+	using glm::mediump_umat3x3;
+	using glm::mediump_umat3x4;
+	using glm::mediump_umat4x2;
+	using glm::mediump_umat4x3;
+	using glm::mediump_umat4x4;
+	using glm::highp_umat2x2;
+	using glm::highp_umat2x3;
+	using glm::highp_umat2x4;
+	using glm::highp_umat3x2;
+	using glm::highp_umat3x3;
+	using glm::highp_umat3x4;
+	using glm::highp_umat4x2;
+	using glm::highp_umat4x3;
+	using glm::highp_umat4x4;
+	using glm::umat2x2;
+	using glm::umat2x3;
+	using glm::umat2x4;
+	using glm::umat3x2;
+	using glm::umat3x3;
+	using glm::umat3x4;
+	using glm::umat4x2;
+	using glm::umat4x3;
+	using glm::umat4x4;
+	using glm::lowp_u8mat2x2;
+	using glm::lowp_u8mat2x3;
+	using glm::lowp_u8mat2x4;
+	using glm::lowp_u8mat3x2;
+	using glm::lowp_u8mat3x3;
+	using glm::lowp_u8mat3x4;
+	using glm::lowp_u8mat4x2;
+	using glm::lowp_u8mat4x3;
+	using glm::lowp_u8mat4x4;
+	using glm::mediump_u8mat2x2;
+	using glm::mediump_u8mat2x3;
+	using glm::mediump_u8mat2x4;
+	using glm::mediump_u8mat3x2;
+	using glm::mediump_u8mat3x3;
+	using glm::mediump_u8mat3x4;
+	using glm::mediump_u8mat4x2;
+	using glm::mediump_u8mat4x3;
+	using glm::mediump_u8mat4x4;
+	using glm::highp_u8mat2x2;
+	using glm::highp_u8mat2x3;
+	using glm::highp_u8mat2x4;
+	using glm::highp_u8mat3x2;
+	using glm::highp_u8mat3x3;
+	using glm::highp_u8mat3x4;
+	using glm::highp_u8mat4x2;
+	using glm::highp_u8mat4x3;
+	using glm::highp_u8mat4x4;
+	using glm::u8mat2x2;
+	using glm::u8mat2x3;
+	using glm::u8mat2x4;
+	using glm::u8mat3x2;
+	using glm::u8mat3x3;
+	using glm::u8mat3x4;
+	using glm::u8mat4x2;
+	using glm::u8mat4x3;
+	using glm::u8mat4x4;
+	using glm::lowp_u16mat2x2;
+	using glm::lowp_u16mat2x3;
+	using glm::lowp_u16mat2x4;
+	using glm::lowp_u16mat3x2;
+	using glm::lowp_u16mat3x3;
+	using glm::lowp_u16mat3x4;
+	using glm::lowp_u16mat4x2;
+	using glm::lowp_u16mat4x3;
+	using glm::lowp_u16mat4x4;
+	using glm::mediump_u16mat2x2;
+	using glm::mediump_u16mat2x3;
+	using glm::mediump_u16mat2x4;
+	using glm::mediump_u16mat3x2;
+	using glm::mediump_u16mat3x3;
+	using glm::mediump_u16mat3x4;
+	using glm::mediump_u16mat4x2;
+	using glm::mediump_u16mat4x3;
+	using glm::mediump_u16mat4x4;
+	using glm::highp_u16mat2x2;
+	using glm::highp_u16mat2x3;
+	using glm::highp_u16mat2x4;
+	using glm::highp_u16mat3x2;
+	using glm::highp_u16mat3x3;
+	using glm::highp_u16mat3x4;
+	using glm::highp_u16mat4x2;
+	using glm::highp_u16mat4x3;
+	using glm::highp_u16mat4x4;
+	using glm::u16mat2x2;
+	using glm::u16mat2x3;
+	using glm::u16mat2x4;
+	using glm::u16mat3x2;
+	using glm::u16mat3x3;
+	using glm::u16mat3x4;
+	using glm::u16mat4x2;
+	using glm::u16mat4x3;
+	using glm::u16mat4x4;
+	using glm::lowp_u32mat2x2;
+	using glm::lowp_u32mat2x3;
+	using glm::lowp_u32mat2x4;
+	using glm::lowp_u32mat3x2;
+	using glm::lowp_u32mat3x3;
+	using glm::lowp_u32mat3x4;
+	using glm::lowp_u32mat4x2;
+	using glm::lowp_u32mat4x3;
+	using glm::lowp_u32mat4x4;
+	using glm::mediump_u32mat2x2;
+	using glm::mediump_u32mat2x3;
+	using glm::mediump_u32mat2x4;
+	using glm::mediump_u32mat3x2;
+	using glm::mediump_u32mat3x3;
+	using glm::mediump_u32mat3x4;
+	using glm::mediump_u32mat4x2;
+	using glm::mediump_u32mat4x3;
+	using glm::mediump_u32mat4x4;
+	using glm::highp_u32mat2x2;
+	using glm::highp_u32mat2x3;
+	using glm::highp_u32mat2x4;
+	using glm::highp_u32mat3x2;
+	using glm::highp_u32mat3x3;
+	using glm::highp_u32mat3x4;
+	using glm::highp_u32mat4x2;
+	using glm::highp_u32mat4x3;
+	using glm::highp_u32mat4x4;
+	using glm::u32mat2x2;
+	using glm::u32mat2x3;
+	using glm::u32mat2x4;
+	using glm::u32mat3x2;
+	using glm::u32mat3x3;
+	using glm::u32mat3x4;
+	using glm::u32mat4x2;
+	using glm::u32mat4x3;
+	using glm::u32mat4x4;
+	using glm::lowp_u64mat2x2;
+	using glm::lowp_u64mat2x3;
+	using glm::lowp_u64mat2x4;
+	using glm::lowp_u64mat3x2;
+	using glm::lowp_u64mat3x3;
+	using glm::lowp_u64mat3x4;
+	using glm::lowp_u64mat4x2;
+	using glm::lowp_u64mat4x3;
+	using glm::lowp_u64mat4x4;
+	using glm::mediump_u64mat2x2;
+	using glm::mediump_u64mat2x3;
+	using glm::mediump_u64mat2x4;
+	using glm::mediump_u64mat3x2;
+	using glm::mediump_u64mat3x3;
+	using glm::mediump_u64mat3x4;
+	using glm::mediump_u64mat4x2;
+	using glm::mediump_u64mat4x3;
+	using glm::mediump_u64mat4x4;
+	using glm::highp_u64mat2x2;
+	using glm::highp_u64mat2x3;
+	using glm::highp_u64mat2x4;
+	using glm::highp_u64mat3x2;
+	using glm::highp_u64mat3x3;
+	using glm::highp_u64mat3x4;
+	using glm::highp_u64mat4x2;
+	using glm::highp_u64mat4x3;
+	using glm::highp_u64mat4x4;
+	using glm::u64mat2x2;
+	using glm::u64mat2x3;
+	using glm::u64mat2x4;
+	using glm::u64mat3x2;
+	using glm::u64mat3x3;
+	using glm::u64mat3x4;
+	using glm::u64mat4x2;
+	using glm::u64mat4x3;
+	using glm::u64mat4x4;
+	using glm::lowp_quat;
+	using glm::mediump_quat;
+	using glm::highp_quat;
+	using glm::quat;
+	using glm::lowp_fquat;
+	using glm::mediump_fquat;
+	using glm::highp_fquat;
+	using glm::fquat;
+	using glm::lowp_f32quat;
+	using glm::mediump_f32quat;
+	using glm::highp_f32quat;
+	using glm::f32quat;
+	using glm::lowp_dquat;
+	using glm::mediump_dquat;
+	using glm::highp_dquat;
+	using glm::dquat;
+	using glm::lowp_f64quat;
+	using glm::mediump_f64quat;
+	using glm::highp_f64quat;
+	using glm::f64quat;
+
+	// Operators
+	using glm::operator+;
+	using glm::operator-;
+	using glm::operator*;
+	using glm::operator/;
+	using glm::operator%;
+	using glm::operator^;
+	using glm::operator&;
+	using glm::operator|;
+	using glm::operator~;
+	using glm::operator<<;
+	using glm::operator>>;
+	using glm::operator==;
+	using glm::operator!=;
+	using glm::operator&&;
+	using glm::operator||;
+
+	// Core functions
+	using glm::abs;
+	using glm::acos;
+	using glm::acosh;
+	using glm::all;
+	using glm::any;
+	using glm::asin;
+	using glm::asinh;
+	using glm::atan;
+	using glm::atanh;
+	using glm::bitCount;
+	using glm::bitfieldExtract;
+	using glm::bitfieldInsert;
+	using glm::bitfieldReverse;
+	using glm::ceil;
+	using glm::clamp;
+	using glm::cos;
+	using glm::cosh;
+	using glm::cross;
+	using glm::degrees;
+	using glm::determinant;
+	using glm::distance;
+	using glm::dot;
+	using glm::equal;
+	using glm::exp;
+	using glm::exp2;
+	using glm::faceforward;
+	using glm::findLSB;
+	using glm::findMSB;
+	using glm::floatBitsToInt;
+	using glm::floatBitsToUint;
+	using glm::floor;
+	using glm::fma;
+	using glm::fract;
+	using glm::frexp;
+	using glm::greaterThan;
+	using glm::greaterThanEqual;
+	using glm::imulExtended;
+	using glm::intBitsToFloat;
+	using glm::inverse;
+	using glm::inversesqrt;
+	using glm::isinf;
+	using glm::isnan;
+	using glm::ldexp;
+	using glm::length;
+	using glm::lessThan;
+	using glm::lessThanEqual;
+	using glm::log;
+	using glm::log2;
+	using glm::matrixCompMult;
+	using glm::max;
+	using glm::min;
+	using glm::mix;
+	using glm::mod;
+	using glm::modf;
+	using glm::normalize;
+	using glm::notEqual;
+	using glm::not_;
+	using glm::outerProduct;
+	using glm::packDouble2x32;
+	using glm::packHalf2x16;
+	using glm::packSnorm2x16;
+	using glm::packSnorm4x8;
+	using glm::packUnorm2x16;
+	using glm::packUnorm4x8;
+	using glm::pow;
+	using glm::radians;
+	using glm::reflect;
+	using glm::refract;
+	using glm::round;
+	using glm::roundEven;
+	using glm::sign;
+	using glm::sin;
+	using glm::sinh;
+	using glm::smoothstep;
+	using glm::sqrt;
+	using glm::step;
+	using glm::tan;
+	using glm::tanh;
+	using glm::transpose;
+	using glm::trunc;
+	using glm::uaddCarry;
+	using glm::uintBitsToFloat;
+	using glm::umulExtended;
+	using glm::unpackDouble2x32;
+	using glm::unpackHalf2x16;
+	using glm::unpackSnorm2x16;
+	using glm::unpackSnorm4x8;
+	using glm::unpackUnorm2x16;
+	using glm::unpackUnorm4x8;
+	using glm::usubBorrow;
+
+#   ifdef GLM_GTC_INLINE_NAMESPACE
+	inline
+#   endif
+	namespace gtc {
+#       if GLM_CONFIG_ALIGNED_GENTYPES == GLM_ENABLE
+		using glm::aligned_highp_vec1;
+		using glm::aligned_mediump_vec1;
+		using glm::aligned_lowp_vec1;
+		using glm::aligned_highp_dvec1;
+		using glm::aligned_mediump_dvec1;
+		using glm::aligned_lowp_dvec1;
+		using glm::aligned_highp_ivec1;
+		using glm::aligned_mediump_ivec1;
+		using glm::aligned_lowp_ivec1;
+		using glm::aligned_highp_uvec1;
+		using glm::aligned_mediump_uvec1;
+		using glm::aligned_lowp_uvec1;
+		using glm::aligned_highp_bvec1;
+		using glm::aligned_mediump_bvec1;
+		using glm::aligned_lowp_bvec1;
+		using glm::packed_highp_vec1;
+		using glm::packed_mediump_vec1;
+		using glm::packed_lowp_vec1;
+		using glm::packed_highp_dvec1;
+		using glm::packed_mediump_dvec1;
+		using glm::packed_lowp_dvec1;
+		using glm::packed_highp_ivec1;
+		using glm::packed_mediump_ivec1;
+		using glm::packed_lowp_ivec1;
+		using glm::packed_highp_uvec1;
+		using glm::packed_mediump_uvec1;
+		using glm::packed_lowp_uvec1;
+		using glm::packed_highp_bvec1;
+		using glm::packed_mediump_bvec1;
+		using glm::packed_lowp_bvec1;
+		using glm::aligned_highp_vec2;
+		using glm::aligned_mediump_vec2;
+		using glm::aligned_lowp_vec2;
+		using glm::aligned_highp_dvec2;
+		using glm::aligned_mediump_dvec2;
+		using glm::aligned_lowp_dvec2;
+		using glm::aligned_highp_ivec2;
+		using glm::aligned_mediump_ivec2;
+		using glm::aligned_lowp_ivec2;
+		using glm::aligned_highp_uvec2;
+		using glm::aligned_mediump_uvec2;
+		using glm::aligned_lowp_uvec2;
+		using glm::aligned_highp_bvec2;
+		using glm::aligned_mediump_bvec2;
+		using glm::aligned_lowp_bvec2;
+		using glm::packed_highp_vec2;
+		using glm::packed_mediump_vec2;
+		using glm::packed_lowp_vec2;
+		using glm::packed_highp_dvec2;
+		using glm::packed_mediump_dvec2;
+		using glm::packed_lowp_dvec2;
+		using glm::packed_highp_ivec2;
+		using glm::packed_mediump_ivec2;
+		using glm::packed_lowp_ivec2;
+		using glm::packed_highp_uvec2;
+		using glm::packed_mediump_uvec2;
+		using glm::packed_lowp_uvec2;
+		using glm::packed_highp_bvec2;
+		using glm::packed_mediump_bvec2;
+		using glm::packed_lowp_bvec2;
+		using glm::aligned_highp_vec3;
+		using glm::aligned_mediump_vec3;
+		using glm::aligned_lowp_vec3;
+		using glm::aligned_highp_dvec3;
+		using glm::aligned_mediump_dvec3;
+		using glm::aligned_lowp_dvec3;
+		using glm::aligned_highp_ivec3;
+		using glm::aligned_mediump_ivec3;
+		using glm::aligned_lowp_ivec3;
+		using glm::aligned_highp_uvec3;
+		using glm::aligned_mediump_uvec3;
+		using glm::aligned_lowp_uvec3;
+		using glm::aligned_highp_bvec3;
+		using glm::aligned_mediump_bvec3;
+		using glm::aligned_lowp_bvec3;
+		using glm::packed_highp_vec3;
+		using glm::packed_mediump_vec3;
+		using glm::packed_lowp_vec3;
+		using glm::packed_highp_dvec3;
+		using glm::packed_mediump_dvec3;
+		using glm::packed_lowp_dvec3;
+		using glm::packed_highp_ivec3;
+		using glm::packed_mediump_ivec3;
+		using glm::packed_lowp_ivec3;
+		using glm::packed_highp_uvec3;
+		using glm::packed_mediump_uvec3;
+		using glm::packed_lowp_uvec3;
+		using glm::packed_highp_bvec3;
+		using glm::packed_mediump_bvec3;
+		using glm::packed_lowp_bvec3;
+		using glm::aligned_highp_vec4;
+		using glm::aligned_mediump_vec4;
+		using glm::aligned_lowp_vec4;
+		using glm::aligned_highp_dvec4;
+		using glm::aligned_mediump_dvec4;
+		using glm::aligned_lowp_dvec4;
+		using glm::aligned_highp_ivec4;
+		using glm::aligned_mediump_ivec4;
+		using glm::aligned_lowp_ivec4;
+		using glm::aligned_highp_uvec4;
+		using glm::aligned_mediump_uvec4;
+		using glm::aligned_lowp_uvec4;
+		using glm::aligned_highp_bvec4;
+		using glm::aligned_mediump_bvec4;
+		using glm::aligned_lowp_bvec4;
+		using glm::packed_highp_vec4;
+		using glm::packed_mediump_vec4;
+		using glm::packed_lowp_vec4;
+		using glm::packed_highp_dvec4;
+		using glm::packed_mediump_dvec4;
+		using glm::packed_lowp_dvec4;
+		using glm::packed_highp_ivec4;
+		using glm::packed_mediump_ivec4;
+		using glm::packed_lowp_ivec4;
+		using glm::packed_highp_uvec4;
+		using glm::packed_mediump_uvec4;
+		using glm::packed_lowp_uvec4;
+		using glm::packed_highp_bvec4;
+		using glm::packed_mediump_bvec4;
+		using glm::packed_lowp_bvec4;
+		using glm::aligned_highp_mat2;
+		using glm::aligned_mediump_mat2;
+		using glm::aligned_lowp_mat2;
+		using glm::aligned_highp_dmat2;
+		using glm::aligned_mediump_dmat2;
+		using glm::aligned_lowp_dmat2;
+		using glm::packed_highp_mat2;
+		using glm::packed_mediump_mat2;
+		using glm::packed_lowp_mat2;
+		using glm::packed_highp_dmat2;
+		using glm::packed_mediump_dmat2;
+		using glm::packed_lowp_dmat2;
+		using glm::aligned_highp_mat3;
+		using glm::aligned_mediump_mat3;
+		using glm::aligned_lowp_mat3;
+		using glm::aligned_highp_dmat3;
+		using glm::aligned_mediump_dmat3;
+		using glm::aligned_lowp_dmat3;
+		using glm::packed_highp_mat3;
+		using glm::packed_mediump_mat3;
+		using glm::packed_lowp_mat3;
+		using glm::packed_highp_dmat3;
+		using glm::packed_mediump_dmat3;
+		using glm::packed_lowp_dmat3;
+		using glm::aligned_highp_mat4;
+		using glm::aligned_mediump_mat4;
+		using glm::aligned_lowp_mat4;
+		using glm::aligned_highp_dmat4;
+		using glm::aligned_mediump_dmat4;
+		using glm::aligned_lowp_dmat4;
+		using glm::packed_highp_mat4;
+		using glm::packed_mediump_mat4;
+		using glm::packed_lowp_mat4;
+		using glm::packed_highp_dmat4;
+		using glm::packed_mediump_dmat4;
+		using glm::packed_lowp_dmat4;
+		using glm::aligned_highp_mat2x2;
+		using glm::aligned_mediump_mat2x2;
+		using glm::aligned_lowp_mat2x2;
+		using glm::aligned_highp_dmat2x2;
+		using glm::aligned_mediump_dmat2x2;
+		using glm::aligned_lowp_dmat2x2;
+		using glm::packed_highp_mat2x2;
+		using glm::packed_mediump_mat2x2;
+		using glm::packed_lowp_mat2x2;
+		using glm::packed_highp_dmat2x2;
+		using glm::packed_mediump_dmat2x2;
+		using glm::packed_lowp_dmat2x2;
+		using glm::aligned_highp_mat2x3;
+		using glm::aligned_mediump_mat2x3;
+		using glm::aligned_lowp_mat2x3;
+		using glm::aligned_highp_dmat2x3;
+		using glm::aligned_mediump_dmat2x3;
+		using glm::aligned_lowp_dmat2x3;
+		using glm::packed_highp_mat2x3;
+		using glm::packed_mediump_mat2x3;
+		using glm::packed_lowp_mat2x3;
+		using glm::packed_highp_dmat2x3;
+		using glm::packed_mediump_dmat2x3;
+		using glm::packed_lowp_dmat2x3;
+		using glm::aligned_highp_mat2x4;
+		using glm::aligned_mediump_mat2x4;
+		using glm::aligned_lowp_mat2x4;
+		using glm::aligned_highp_dmat2x4;
+		using glm::aligned_mediump_dmat2x4;
+		using glm::aligned_lowp_dmat2x4;
+		using glm::packed_highp_mat2x4;
+		using glm::packed_mediump_mat2x4;
+		using glm::packed_lowp_mat2x4;
+		using glm::packed_highp_dmat2x4;
+		using glm::packed_mediump_dmat2x4;
+		using glm::packed_lowp_dmat2x4;
+		using glm::aligned_highp_mat3x2;
+		using glm::aligned_mediump_mat3x2;
+		using glm::aligned_lowp_mat3x2;
+		using glm::aligned_highp_dmat3x2;
+		using glm::aligned_mediump_dmat3x2;
+		using glm::aligned_lowp_dmat3x2;
+		using glm::packed_highp_mat3x2;
+		using glm::packed_mediump_mat3x2;
+		using glm::packed_lowp_mat3x2;
+		using glm::packed_highp_dmat3x2;
+		using glm::packed_mediump_dmat3x2;
+		using glm::packed_lowp_dmat3x2;
+		using glm::aligned_highp_mat3x3;
+		using glm::aligned_mediump_mat3x3;
+		using glm::aligned_lowp_mat3x3;
+		using glm::aligned_highp_dmat3x3;
+		using glm::aligned_mediump_dmat3x3;
+		using glm::aligned_lowp_dmat3x3;
+		using glm::packed_highp_mat3x3;
+		using glm::packed_mediump_mat3x3;
+		using glm::packed_lowp_mat3x3;
+		using glm::packed_highp_dmat3x3;
+		using glm::packed_mediump_dmat3x3;
+		using glm::packed_lowp_dmat3x3;
+		using glm::aligned_highp_mat3x4;
+		using glm::aligned_mediump_mat3x4;
+		using glm::aligned_lowp_mat3x4;
+		using glm::aligned_highp_dmat3x4;
+		using glm::aligned_mediump_dmat3x4;
+		using glm::aligned_lowp_dmat3x4;
+		using glm::packed_highp_mat3x4;
+		using glm::packed_mediump_mat3x4;
+		using glm::packed_lowp_mat3x4;
+		using glm::packed_highp_dmat3x4;
+		using glm::packed_mediump_dmat3x4;
+		using glm::packed_lowp_dmat3x4;
+		using glm::aligned_highp_mat4x2;
+		using glm::aligned_mediump_mat4x2;
+		using glm::aligned_lowp_mat4x2;
+		using glm::aligned_highp_dmat4x2;
+		using glm::aligned_mediump_dmat4x2;
+		using glm::aligned_lowp_dmat4x2;
+		using glm::packed_highp_mat4x2;
+		using glm::packed_mediump_mat4x2;
+		using glm::packed_lowp_mat4x2;
+		using glm::packed_highp_dmat4x2;
+		using glm::packed_mediump_dmat4x2;
+		using glm::packed_lowp_dmat4x2;
+		using glm::aligned_highp_mat4x3;
+		using glm::aligned_mediump_mat4x3;
+		using glm::aligned_lowp_mat4x3;
+		using glm::aligned_highp_dmat4x3;
+		using glm::aligned_mediump_dmat4x3;
+		using glm::aligned_lowp_dmat4x3;
+		using glm::packed_highp_mat4x3;
+		using glm::packed_mediump_mat4x3;
+		using glm::packed_lowp_mat4x3;
+		using glm::packed_highp_dmat4x3;
+		using glm::packed_mediump_dmat4x3;
+		using glm::packed_lowp_dmat4x3;
+		using glm::aligned_highp_mat4x4;
+		using glm::aligned_mediump_mat4x4;
+		using glm::aligned_lowp_mat4x4;
+		using glm::aligned_highp_dmat4x4;
+		using glm::aligned_mediump_dmat4x4;
+		using glm::aligned_lowp_dmat4x4;
+		using glm::packed_highp_mat4x4;
+		using glm::packed_mediump_mat4x4;
+		using glm::packed_lowp_mat4x4;
+		using glm::packed_highp_dmat4x4;
+		using glm::packed_mediump_dmat4x4;
+		using glm::packed_lowp_dmat4x4;
+#       if(defined(GLM_PRECISION_LOWP_FLOAT))
+		using glm::aligned_vec1;
+		using glm::aligned_vec2;
+		using glm::aligned_vec3;
+		using glm::aligned_vec4;
+		using glm::packed_vec1;
+		using glm::packed_vec2;
+		using glm::packed_vec3;
+		using glm::packed_vec4;
+		using glm::aligned_mat2;
+		using glm::aligned_mat3;
+		using glm::aligned_mat4;
+		using glm::packed_mat2;
+		using glm::packed_mat3;
+		using glm::packed_mat4;
+		using glm::aligned_mat2x2;
+		using glm::aligned_mat2x3;
+		using glm::aligned_mat2x4;
+		using glm::aligned_mat3x2;
+		using glm::aligned_mat3x3;
+		using glm::aligned_mat3x4;
+		using glm::aligned_mat4x2;
+		using glm::aligned_mat4x3;
+		using glm::aligned_mat4x4;
+		using glm::packed_mat2x2;
+		using glm::packed_mat2x3;
+		using glm::packed_mat2x4;
+		using glm::packed_mat3x2;
+		using glm::packed_mat3x3;
+		using glm::packed_mat3x4;
+		using glm::packed_mat4x2;
+		using glm::packed_mat4x3;
+		using glm::packed_mat4x4;
+#       elif(defined(GLM_PRECISION_MEDIUMP_FLOAT))
+		using glm::aligned_vec1;
+		using glm::aligned_vec2;
+		using glm::aligned_vec3;
+		using glm::aligned_vec4;
+		using glm::packed_vec1;
+		using glm::packed_vec2;
+		using glm::packed_vec3;
+		using glm::packed_vec4;
+		using glm::aligned_mat2;
+		using glm::aligned_mat3;
+		using glm::aligned_mat4;
+		using glm::packed_mat2;
+		using glm::packed_mat3;
+		using glm::packed_mat4;
+		using glm::aligned_mat2x2;
+		using glm::aligned_mat2x3;
+		using glm::aligned_mat2x4;
+		using glm::aligned_mat3x2;
+		using glm::aligned_mat3x3;
+		using glm::aligned_mat3x4;
+		using glm::aligned_mat4x2;
+		using glm::aligned_mat4x3;
+		using glm::aligned_mat4x4;
+		using glm::packed_mat2x2;
+		using glm::packed_mat2x3;
+		using glm::packed_mat2x4;
+		using glm::packed_mat3x2;
+		using glm::packed_mat3x3;
+		using glm::packed_mat3x4;
+		using glm::packed_mat4x2;
+		using glm::packed_mat4x3;
+		using glm::packed_mat4x4;
+#       else //defined(GLM_PRECISION_HIGHP_FLOAT)
+		using glm::aligned_vec1;
+		using glm::aligned_vec2;
+		using glm::aligned_vec3;
+		using glm::aligned_vec4;
+		using glm::packed_vec1;
+		using glm::packed_vec2;
+		using glm::packed_vec3;
+		using glm::packed_vec4;
+		using glm::aligned_mat2;
+		using glm::aligned_mat3;
+		using glm::aligned_mat4;
+		using glm::packed_mat2;
+		using glm::packed_mat3;
+		using glm::packed_mat4;
+		using glm::aligned_mat2x2;
+		using glm::aligned_mat2x3;
+		using glm::aligned_mat2x4;
+		using glm::aligned_mat3x2;
+		using glm::aligned_mat3x3;
+		using glm::aligned_mat3x4;
+		using glm::aligned_mat4x2;
+		using glm::aligned_mat4x3;
+		using glm::aligned_mat4x4;
+		using glm::packed_mat2x2;
+		using glm::packed_mat2x3;
+		using glm::packed_mat2x4;
+		using glm::packed_mat3x2;
+		using glm::packed_mat3x3;
+		using glm::packed_mat3x4;
+		using glm::packed_mat4x2;
+		using glm::packed_mat4x3;
+		using glm::packed_mat4x4;
+#       endif//GLM_PRECISION
+#       if(defined(GLM_PRECISION_LOWP_DOUBLE))
+		using glm::aligned_dvec1;
+		using glm::aligned_dvec2;
+		using glm::aligned_dvec3;
+		using glm::aligned_dvec4;
+		using glm::packed_dvec1;
+		using glm::packed_dvec2;
+		using glm::packed_dvec3;
+		using glm::packed_dvec4;
+		using glm::aligned_dmat2;
+		using glm::aligned_dmat3;
+		using glm::aligned_dmat4;
+		using glm::packed_dmat2;
+		using glm::packed_dmat3;
+		using glm::packed_dmat4;
+		using glm::aligned_dmat2x2;
+		using glm::aligned_dmat2x3;
+		using glm::aligned_dmat2x4;
+		using glm::aligned_dmat3x2;
+		using glm::aligned_dmat3x3;
+		using glm::aligned_dmat3x4;
+		using glm::aligned_dmat4x2;
+		using glm::aligned_dmat4x3;
+		using glm::aligned_dmat4x4;
+		using glm::packed_dmat2x2;
+		using glm::packed_dmat2x3;
+		using glm::packed_dmat2x4;
+		using glm::packed_dmat3x2;
+		using glm::packed_dmat3x3;
+		using glm::packed_dmat3x4;
+		using glm::packed_dmat4x2;
+		using glm::packed_dmat4x3;
+		using glm::packed_dmat4x4;
+#       elif(defined(GLM_PRECISION_MEDIUMP_DOUBLE))
+		using glm::aligned_dvec1;
+		using glm::aligned_dvec2;
+		using glm::aligned_dvec3;
+		using glm::aligned_dvec4;
+		using glm::packed_dvec1;
+		using glm::packed_dvec2;
+		using glm::packed_dvec3;
+		using glm::packed_dvec4;
+		using glm::aligned_dmat2;
+		using glm::aligned_dmat3;
+		using glm::aligned_dmat4;
+		using glm::packed_dmat2;
+		using glm::packed_dmat3;
+		using glm::packed_dmat4;
+		using glm::aligned_dmat2x2;
+		using glm::aligned_dmat2x3;
+		using glm::aligned_dmat2x4;
+		using glm::aligned_dmat3x2;
+		using glm::aligned_dmat3x3;
+		using glm::aligned_dmat3x4;
+		using glm::aligned_dmat4x2;
+		using glm::aligned_dmat4x3;
+		using glm::aligned_dmat4x4;
+		using glm::packed_dmat2x2;
+		using glm::packed_dmat2x3;
+		using glm::packed_dmat2x4;
+		using glm::packed_dmat3x2;
+		using glm::packed_dmat3x3;
+		using glm::packed_dmat3x4;
+		using glm::packed_dmat4x2;
+		using glm::packed_dmat4x3;
+		using glm::packed_dmat4x4;
+#       else //defined(GLM_PRECISION_HIGHP_DOUBLE)
+		using glm::aligned_dvec1;
+		using glm::aligned_dvec2;
+		using glm::aligned_dvec3;
+		using glm::aligned_dvec4;
+		using glm::packed_dvec1;
+		using glm::packed_dvec2;
+		using glm::packed_dvec3;
+		using glm::packed_dvec4;
+		using glm::aligned_dmat2;
+		using glm::aligned_dmat3;
+		using glm::aligned_dmat4;
+		using glm::packed_dmat2;
+		using glm::packed_dmat3;
+		using glm::packed_dmat4;
+		using glm::aligned_dmat2x2;
+		using glm::aligned_dmat2x3;
+		using glm::aligned_dmat2x4;
+		using glm::aligned_dmat3x2;
+		using glm::aligned_dmat3x3;
+		using glm::aligned_dmat3x4;
+		using glm::aligned_dmat4x2;
+		using glm::aligned_dmat4x3;
+		using glm::aligned_dmat4x4;
+		using glm::packed_dmat2x2;
+		using glm::packed_dmat2x3;
+		using glm::packed_dmat2x4;
+		using glm::packed_dmat3x2;
+		using glm::packed_dmat3x3;
+		using glm::packed_dmat3x4;
+		using glm::packed_dmat4x2;
+		using glm::packed_dmat4x3;
+		using glm::packed_dmat4x4;
+#       endif//GLM_PRECISION
+#       if(defined(GLM_PRECISION_LOWP_INT))
+		using glm::aligned_ivec1;
+		using glm::aligned_ivec2;
+		using glm::aligned_ivec3;
+		using glm::aligned_ivec4;
+#       elif(defined(GLM_PRECISION_MEDIUMP_INT))
+		using glm::aligned_ivec1;
+		using glm::aligned_ivec2;
+		using glm::aligned_ivec3;
+		using glm::aligned_ivec4;
+#       else //defined(GLM_PRECISION_HIGHP_INT)
+		using glm::aligned_ivec1;
+		using glm::aligned_ivec2;
+		using glm::aligned_ivec3;
+		using glm::aligned_ivec4;
+		using glm::packed_ivec1;
+		using glm::packed_ivec2;
+		using glm::packed_ivec3;
+		using glm::packed_ivec4;
+#       endif//GLM_PRECISION
+#       if(defined(GLM_PRECISION_LOWP_UINT))
+		using glm::aligned_uvec1;
+		using glm::aligned_uvec2;
+		using glm::aligned_uvec3;
+		using glm::aligned_uvec4;
+#       elif(defined(GLM_PRECISION_MEDIUMP_UINT))
+		using glm::aligned_uvec1;
+		using glm::aligned_uvec2;
+		using glm::aligned_uvec3;
+		using glm::aligned_uvec4;
+#       else //defined(GLM_PRECISION_HIGHP_UINT)
+		using glm::aligned_uvec1;
+		using glm::aligned_uvec2;
+		using glm::aligned_uvec3;
+		using glm::aligned_uvec4;
+		using glm::packed_uvec1;
+		using glm::packed_uvec2;
+		using glm::packed_uvec3;
+		using glm::packed_uvec4;
+#       endif//GLM_PRECISION
+#       if(defined(GLM_PRECISION_LOWP_BOOL))
+		using glm::aligned_bvec1;
+		using glm::aligned_bvec2;
+		using glm::aligned_bvec3;
+		using glm::aligned_bvec4;
+#       elif(defined(GLM_PRECISION_MEDIUMP_BOOL))
+		using glm::aligned_bvec1;
+		using glm::aligned_bvec2;
+		using glm::aligned_bvec3;
+		using glm::aligned_bvec4;
+#       else //defined(GLM_PRECISION_HIGHP_BOOL)
+		using glm::aligned_bvec1;
+		using glm::aligned_bvec2;
+		using glm::aligned_bvec3;
+		using glm::aligned_bvec4;
+		using glm::packed_bvec1;
+		using glm::packed_bvec2;
+		using glm::packed_bvec3;
+		using glm::packed_bvec4;
+#       endif//GLM_PRECISION
+#       endif
+
+
+		using glm::abs;
+		using glm::acos;
+		using glm::acosh;
+		using glm::acot;
+		using glm::acoth;
+		using glm::acsc;
+		using glm::acsch;
+		using glm::affineInverse;
+		using glm::all;
+		using glm::angle;
+		using glm::angleAxis;
+		using glm::any;
+		using glm::asec;
+		using glm::asech;
+		using glm::asin;
+		using glm::asinh;
+		using glm::atan;
+		using glm::atanh;
+		using glm::axis;
+		using glm::ballRand;
+		using glm::bitCount;
+		using glm::bitfieldDeinterleave;
+		using glm::bitfieldExtract;
+		using glm::bitfieldFillOne;
+		using glm::bitfieldFillZero;
+		using glm::bitfieldInsert;
+		using glm::bitfieldInterleave;
+		using glm::bitfieldReverse;
+		using glm::bitfieldRotateLeft;
+		using glm::bitfieldRotateRight;
+		using glm::ceil;
+		using glm::ceilMultiple;
+		using glm::ceilPowerOfTwo;
+		using glm::circularRand;
+		using glm::clamp;
+		using glm::column;
+		using glm::conjugate;
+		using glm::convertLinearToSRGB;
+		using glm::convertSRGBToLinear;
+		using glm::cos;
+		using glm::cos_one_over_two;
+		using glm::cosh;
+		using glm::cot;
+		using glm::coth;
+		using glm::cross;
+		using glm::csc;
+		using glm::csch;
+		using glm::degrees;
+		using glm::determinant;
+		using glm::diskRand;
+		using glm::distance;
+		using glm::dot;
+		using glm::e;
+		using glm::epsilon;
+		using glm::epsilonEqual;
+		using glm::epsilonNotEqual;
+		using glm::equal;
+		using glm::euler;
+		using glm::eulerAngles;
+		using glm::exp;
+		using glm::exp2;
+		using glm::faceforward;
+		using glm::fclamp;
+		using glm::findLSB;
+		using glm::findMSB;
+		using glm::floatBitsToInt;
+		using glm::floatBitsToUint;
+		using glm::float_distance;
+		using glm::floor;
+		using glm::floorMultiple;
+		using glm::floorPowerOfTwo;
+		using glm::fma;
+		using glm::fmax;
+		using glm::fmin;
+		using glm::four_over_pi;
+		using glm::fract;
+		using glm::frexp;
+		using glm::frustum;
+		using glm::frustumLH;
+		using glm::frustumLH_NO;
+		using glm::frustumLH_ZO;
+		using glm::frustumNO;
+		using glm::frustumRH;
+		using glm::frustumRH_NO;
+		using glm::frustumRH_ZO;
+		using glm::frustumZO;
+		using glm::gaussRand;
+		using glm::golden_ratio;
+		using glm::greaterThan;
+		using glm::greaterThanEqual;
+		using glm::half_pi;
+		using glm::identity;
+		using glm::imulExtended;
+		using glm::infinitePerspective;
+		using glm::infinitePerspectiveLH;
+		using glm::infinitePerspectiveRH;
+		using glm::intBitsToFloat;
+		using glm::inverse;
+		using glm::inverseTranspose;
+		using glm::inversesqrt;
+		using glm::iround;
+		using glm::isinf;
+		using glm::isnan;
+		using glm::ldexp;
+		using glm::length;
+		using glm::lerp;
+		using glm::lessThan;
+		using glm::lessThanEqual;
+		using glm::linearRand;
+		using glm::ln_ln_two;
+		using glm::ln_ten;
+		using glm::ln_two;
+		using glm::log;
+		using glm::log2;
+		using glm::lookAt;
+		using glm::lookAtLH;
+		using glm::lookAtRH;
+		using glm::make_mat2;
+		using glm::make_mat2x2;
+		using glm::make_mat2x3;
+		using glm::make_mat2x4;
+		using glm::make_mat3;
+		using glm::make_mat3x2;
+		using glm::make_mat3x3;
+		using glm::make_mat3x4;
+		using glm::make_mat4;
+		using glm::make_mat4x2;
+		using glm::make_mat4x3;
+		using glm::make_mat4x4;
+		using glm::make_quat;
+		using glm::make_vec1;
+		using glm::make_vec2;
+		using glm::make_vec3;
+		using glm::make_vec4;
+		using glm::mask;
+		using glm::mat3_cast;
+		using glm::mat4_cast;
+		using glm::matrixCompMult;
+		using glm::max;
+		using glm::min;
+		using glm::mirrorClamp;
+		using glm::mirrorRepeat;
+		using glm::mix;
+		using glm::mod;
+		using glm::modf;
+		using glm::next_float;
+		using glm::normalize;
+		using glm::notEqual;
+		using glm::not_;
+		using glm::one;
+		using glm::one_over_pi;
+		using glm::one_over_root_two;
+		using glm::one_over_two_pi;
+		using glm::ortho;
+		using glm::orthoLH;
+		using glm::orthoLH_NO;
+		using glm::orthoLH_ZO;
+		using glm::orthoNO;
+		using glm::orthoRH;
+		using glm::orthoRH_NO;
+		using glm::orthoRH_ZO;
+		using glm::orthoZO;
+		using glm::outerProduct;
+		using glm::packF2x11_1x10;
+		using glm::packF3x9_E1x5;
+		using glm::packHalf;
+		using glm::packHalf1x16;
+		using glm::packHalf4x16;
+		using glm::packI3x10_1x2;
+		using glm::packInt2x16;
+		using glm::packInt2x32;
+		using glm::packInt2x8;
+		using glm::packInt4x16;
+		using glm::packInt4x8;
+		using glm::packRGBM;
+		using glm::packSnorm;
+		using glm::packSnorm1x16;
+		using glm::packSnorm1x8;
+		using glm::packSnorm2x8;
+		using glm::packSnorm3x10_1x2;
+		using glm::packSnorm4x16;
+		using glm::packU3x10_1x2;
+		using glm::packUint2x16;
+		using glm::packUint2x32;
+		using glm::packUint2x8;
+		using glm::packUint4x16;
+		using glm::packUint4x8;
+		using glm::packUnorm;
+		using glm::packUnorm1x16;
+		using glm::packUnorm1x5_1x6_1x5;
+		using glm::packUnorm1x8;
+		using glm::packUnorm2x3_1x2;
+		using glm::packUnorm2x4;
+		using glm::packUnorm2x8;
+		using glm::packUnorm3x10_1x2;
+		using glm::packUnorm3x5_1x1;
+		using glm::packUnorm4x16;
+		using glm::packUnorm4x4;
+		using glm::perlin;
+		using glm::perspective;
+		using glm::perspectiveFov;
+		using glm::perspectiveFovLH;
+		using glm::perspectiveFovLH_NO;
+		using glm::perspectiveFovLH_ZO;
+		using glm::perspectiveFovNO;
+		using glm::perspectiveFovRH;
+		using glm::perspectiveFovRH_NO;
+		using glm::perspectiveFovRH_ZO;
+		using glm::perspectiveFovZO;
+		using glm::perspectiveLH;
+		using glm::perspectiveLH_NO;
+		using glm::perspectiveLH_ZO;
+		using glm::perspectiveNO;
+		using glm::perspectiveRH;
+		using glm::perspectiveRH_NO;
+		using glm::perspectiveRH_ZO;
+		using glm::perspectiveZO;
+		using glm::pi;
+		using glm::pickMatrix;
+		using glm::pitch;
+		using glm::pow;
+		using glm::prev_float;
+		using glm::project;
+		using glm::projectNO;
+		using glm::projectZO;
+		using glm::quarter_pi;
+		using glm::quatLookAt;
+		using glm::quatLookAtLH;
+		using glm::quatLookAtRH;
+		using glm::quat_cast;
+		using glm::radians;
+		using glm::reflect;
+		using glm::refract;
+		using glm::repeat;
+		using glm::roll;
+		using glm::root_five;
+		using glm::root_half_pi;
+		using glm::root_ln_four;
+		using glm::root_pi;
+		using glm::root_three;
+		using glm::root_two;
+		using glm::root_two_pi;
+		using glm::rotate;
+		using glm::round;
+		using glm::roundEven;
+		using glm::roundMultiple;
+		using glm::roundPowerOfTwo;
+		using glm::row;
+		using glm::scale;
+		using glm::sec;
+		using glm::sech;
+		using glm::sign;
+		using glm::simplex;
+		using glm::sin;
+		using glm::sinh;
+		using glm::slerp;
+		using glm::smoothstep;
+		using glm::sphericalRand;
+		using glm::sqrt;
+		using glm::step;
+		using glm::tan;
+		using glm::tanh;
+		using glm::third;
+		using glm::three_over_two_pi;
+		using glm::translate;
+		using glm::transpose;
+		using glm::trunc;
+		using glm::tweakedInfinitePerspective;
+		using glm::two_over_pi;
+		using glm::two_over_root_pi;
+		using glm::two_pi;
+		using glm::two_thirds;
+		using glm::uaddCarry;
+		using glm::uintBitsToFloat;
+		using glm::umulExtended;
+		using glm::unProject;
+		using glm::unProjectNO;
+		using glm::unProjectZO;
+		using glm::unpackF2x11_1x10;
+		using glm::unpackF3x9_E1x5;
+		using glm::unpackHalf;
+		using glm::unpackHalf1x16;
+		using glm::unpackHalf4x16;
+		using glm::unpackI3x10_1x2;
+		using glm::unpackInt2x16;
+		using glm::unpackInt2x32;
+		using glm::unpackInt2x8;
+		using glm::unpackInt4x16;
+		using glm::unpackInt4x8;
+		using glm::unpackRGBM;
+		using glm::unpackSnorm;
+		using glm::unpackSnorm1x16;
+		using glm::unpackSnorm1x8;
+		using glm::unpackSnorm2x8;
+		using glm::unpackSnorm3x10_1x2;
+		using glm::unpackSnorm4x16;
+		using glm::unpackU3x10_1x2;
+		using glm::unpackUint2x16;
+		using glm::unpackUint2x32;
+		using glm::unpackUint2x8;
+		using glm::unpackUint4x16;
+		using glm::unpackUint4x8;
+		using glm::unpackUnorm;
+		using glm::unpackUnorm1x16;
+		using glm::unpackUnorm1x5_1x6_1x5;
+		using glm::unpackUnorm1x8;
+		using glm::unpackUnorm2x3_1x2;
+		using glm::unpackUnorm2x4;
+		using glm::unpackUnorm2x8;
+		using glm::unpackUnorm3x10_1x2;
+		using glm::unpackUnorm3x5_1x1;
+		using glm::unpackUnorm4x16;
+		using glm::unpackUnorm4x4;
+		using glm::uround;
+		using glm::usubBorrow;
+		using glm::value_ptr;
+		using glm::yaw;
+		using glm::zero;
+	}
+
+#   ifdef GLM_EXT_INLINE_NAMESPACE
+	inline
+#   endif
+	namespace ext {
+		using glm::abs;
+		using glm::acos;
+		using glm::acosh;
+		using glm::acot;
+		using glm::acoth;
+		using glm::acsc;
+		using glm::acsch;
+		using glm::all;
+		using glm::angle;
+		using glm::angleAxis;
+		using glm::any;
+		using glm::asec;
+		using glm::asech;
+		using glm::asin;
+		using glm::asinh;
+		using glm::atan;
+		using glm::atanh;
+		using glm::axis;
+		using glm::ceil;
+		using glm::clamp;
+		using glm::conjugate;
+		using glm::cos;
+		using glm::cos_one_over_two;
+		using glm::cosh;
+		using glm::cot;
+		using glm::coth;
+		using glm::cross;
+		using glm::csc;
+		using glm::csch;
+		using glm::degrees;
+		using glm::determinant;
+		using glm::distance;
+		using glm::dot;
+		using glm::e;
+		using glm::epsilon;
+		using glm::equal;
+		using glm::euler;
+		using glm::exp;
+		using glm::exp2;
+		using glm::faceforward;
+		using glm::fclamp;
+		using glm::findNSB;
+		using glm::floatBitsToInt;
+		using glm::floatBitsToUint;
+		using glm::floatDistance;
+		using glm::floor;
+		using glm::fma;
+		using glm::fmax;
+		using glm::fmin;
+		using glm::four_over_pi;
+		using glm::fract;
+		using glm::frexp;
+		using glm::frustum;
+		using glm::frustumLH;
+		using glm::frustumLH_NO;
+		using glm::frustumLH_ZO;
+		using glm::frustumNO;
+		using glm::frustumRH;
+		using glm::frustumRH_NO;
+		using glm::frustumRH_ZO;
+		using glm::frustumZO;
+		using glm::golden_ratio;
+		using glm::greaterThan;
+		using glm::greaterThanEqual;
+		using glm::half_pi;
+		using glm::identity;
+		using glm::infinitePerspective;
+		using glm::infinitePerspectiveLH;
+		using glm::infinitePerspectiveRH;
+		using glm::intBitsToFloat;
+		using glm::inverse;
+		using glm::inversesqrt;
+		using glm::iround;
+		using glm::isMultiple;
+		using glm::isPowerOfTwo;
+		using glm::isinf;
+		using glm::isnan;
+		using glm::ldexp;
+		using glm::length;
+		using glm::lerp;
+		using glm::lessThan;
+		using glm::lessThanEqual;
+		using glm::ln_ln_two;
+		using glm::ln_ten;
+		using glm::ln_two;
+		using glm::log;
+		using glm::log2;
+		using glm::lookAt;
+		using glm::lookAtLH;
+		using glm::lookAtRH;
+		using glm::matrixCompMult;
+		using glm::max;
+		using glm::min;
+		using glm::mirrorClamp;
+		using glm::mirrorRepeat;
+		using glm::mix;
+		using glm::mod;
+		using glm::modf;
+		using glm::nextFloat;
+		using glm::nextMultiple;
+		using glm::nextPowerOfTwo;
+		using glm::normalize;
+		using glm::notEqual;
+		using glm::not_;
+		using glm::one;
+		using glm::one_over_pi;
+		using glm::one_over_root_two;
+		using glm::one_over_two_pi;
+		using glm::ortho;
+		using glm::orthoLH;
+		using glm::orthoLH_NO;
+		using glm::orthoLH_ZO;
+		using glm::orthoNO;
+		using glm::orthoRH;
+		using glm::orthoRH_NO;
+		using glm::orthoRH_ZO;
+		using glm::orthoZO;
+		using glm::outerProduct;
+		using glm::perspective;
+		using glm::perspectiveFov;
+		using glm::perspectiveFovLH;
+		using glm::perspectiveFovLH_NO;
+		using glm::perspectiveFovLH_ZO;
+		using glm::perspectiveFovNO;
+		using glm::perspectiveFovRH;
+		using glm::perspectiveFovRH_NO;
+		using glm::perspectiveFovRH_ZO;
+		using glm::perspectiveFovZO;
+		using glm::perspectiveLH;
+		using glm::perspectiveLH_NO;
+		using glm::perspectiveLH_ZO;
+		using glm::perspectiveNO;
+		using glm::perspectiveRH;
+		using glm::perspectiveRH_NO;
+		using glm::perspectiveRH_ZO;
+		using glm::perspectiveZO;
+		using glm::pi;
+		using glm::pickMatrix;
+		using glm::pow;
+		using glm::prevFloat;
+		using glm::prevMultiple;
+		using glm::prevPowerOfTwo;
+		using glm::project;
+		using glm::projectNO;
+		using glm::projectZO;
+		using glm::quarter_pi;
+		using glm::radians;
+		using glm::reflect;
+		using glm::refract;
+		using glm::repeat;
+		using glm::root_five;
+		using glm::root_half_pi;
+		using glm::root_ln_four;
+		using glm::root_pi;
+		using glm::root_three;
+		using glm::root_two;
+		using glm::root_two_pi;
+		using glm::rotate;
+		using glm::round;
+		using glm::roundEven;
+		using glm::scale;
+		using glm::sec;
+		using glm::sech;
+		using glm::sign;
+		using glm::sin;
+		using glm::sinh;
+		using glm::slerp;
+		using glm::smoothstep;
+		using glm::sqrt;
+		using glm::step;
+		using glm::tan;
+		using glm::tanh;
+		using glm::third;
+		using glm::three_over_two_pi;
+		using glm::translate;
+		using glm::transpose;
+		using glm::trunc;
+		using glm::tweakedInfinitePerspective;
+		using glm::two_over_pi;
+		using glm::two_over_root_pi;
+		using glm::two_pi;
+		using glm::two_thirds;
+		using glm::uintBitsToFloat;
+		using glm::unProject;
+		using glm::unProjectNO;
+		using glm::unProjectZO;
+		using glm::uround;
+		using glm::zero;
+	}
+
+#   ifdef GLM_ENABLE_EXPERIMENTAL
+#   ifdef GLM_GTX_INLINE_NAMESPACE
+	inline
+#   endif
+	namespace gtx {
+		using glm::io::order_type;
+		using glm::io::format_punct;
+		using glm::io::basic_state_saver;
+		using glm::io::basic_format_saver;
+		using glm::io::precision;
+		using glm::io::width;
+		using glm::io::delimeter;
+		using glm::io::order;
+		using glm::io::get_facet;
+		using glm::io::formatted;
+		using glm::io::unformatted;
+		using glm::io::operator<<;
+		using glm::operator<<;
+		using glm::tdualquat;
+
+#       if !((GLM_COMPILER & GLM_COMPILER_CUDA) || (GLM_COMPILER & GLM_COMPILER_HIP))
+		using glm::to_string;
+#       endif
+#       if GLM_HAS_TEMPLATE_ALIASES
+		using glm::operator*;
+		using glm::operator/;
+#       endif
+#       if GLM_HAS_RANGE_FOR
+		using glm::components;
+		using glm::begin;
+		using glm::end;
+#       endif
+
+		using glm::abs;
+		using glm::acos;
+		using glm::acosh;
+		using glm::adjugate;
+		using glm::all;
+		using glm::angle;
+		using glm::angleAxis;
+		using glm::any;
+		using glm::areCollinear;
+		using glm::areOrthogonal;
+		using glm::areOrthonormal;
+		using glm::asin;
+		using glm::asinh;
+		using glm::associatedMax;
+		using glm::associatedMin;
+		using glm::atan;
+		using glm::atanh;
+		using glm::axis;
+		using glm::axisAngle;
+		using glm::axisAngleMatrix;
+		using glm::backEaseIn;
+		using glm::backEaseInOut;
+		using glm::backEaseOut;
+		using glm::bitCount;
+		using glm::bitfieldDeinterleave;
+		using glm::bitfieldExtract;
+		using glm::bitfieldFillOne;
+		using glm::bitfieldFillZero;
+		using glm::bitfieldInsert;
+		using glm::bitfieldInterleave;
+		using glm::bitfieldReverse;
+		using glm::bitfieldRotateLeft;
+		using glm::bitfieldRotateRight;
+		using glm::bounceEaseIn;
+		using glm::bounceEaseInOut;
+		using glm::bounceEaseOut;
+		using glm::catmullRom;
+		using glm::ceil;
+		using glm::circularEaseIn;
+		using glm::circularEaseInOut;
+		using glm::circularEaseOut;
+		using glm::clamp;
+		using glm::closeBounded;
+		using glm::closestPointOnLine;
+		using glm::colMajor2;
+		using glm::colMajor3;
+		using glm::colMajor4;
+		using glm::compAdd;
+		using glm::compMax;
+		using glm::compMin;
+		using glm::compMul;
+		using glm::compNormalize;
+		using glm::compScale;
+		using glm::computeCovarianceMatrix;
+		using glm::conjugate;
+		using glm::convertD65XYZToD50XYZ;
+		using glm::convertD65XYZToLinearSRGB;
+		using glm::convertLinearSRGBToD50XYZ;
+		using glm::convertLinearSRGBToD65XYZ;
+		using glm::cos;
+		using glm::cos_one_over_two;
+		using glm::cosh;
+		using glm::cross;
+		using glm::cubic;
+		using glm::cubicEaseIn;
+		using glm::cubicEaseInOut;
+		using glm::cubicEaseOut;
+		using glm::decompose;
+		using glm::degrees;
+		using glm::derivedEulerAngleX;
+		using glm::derivedEulerAngleY;
+		using glm::derivedEulerAngleZ;
+		using glm::determinant;
+		using glm::diagonal2x2;
+		using glm::diagonal2x3;
+		using glm::diagonal2x4;
+		using glm::diagonal3x2;
+		using glm::diagonal3x3;
+		using glm::diagonal3x4;
+		using glm::diagonal4x2;
+		using glm::diagonal4x3;
+		using glm::diagonal4x4;
+		using glm::distance;
+		using glm::distance2;
+		using glm::dot;
+		using glm::dual_quat_identity;
+		using glm::dualquat_cast;
+		using glm::e;
+		using glm::elasticEaseIn;
+		using glm::elasticEaseInOut;
+		using glm::elasticEaseOut;
+		using glm::epsilon;
+		using glm::epsilonEqual;
+		using glm::epsilonNotEqual;
+		using glm::equal;
+		using glm::euclidean;
+		using glm::euler;
+		using glm::eulerAngleX;
+		using glm::eulerAngleXY;
+		using glm::eulerAngleXYX;
+		using glm::eulerAngleXYZ;
+		using glm::eulerAngleXZ;
+		using glm::eulerAngleXZX;
+		using glm::eulerAngleXZY;
+		using glm::eulerAngleY;
+		using glm::eulerAngleYX;
+		using glm::eulerAngleYXY;
+		using glm::eulerAngleYXZ;
+		using glm::eulerAngleYZ;
+		using glm::eulerAngleYZX;
+		using glm::eulerAngleYZY;
+		using glm::eulerAngleZ;
+		using glm::eulerAngleZX;
+		using glm::eulerAngleZXY;
+		using glm::eulerAngleZXZ;
+		using glm::eulerAngleZY;
+		using glm::eulerAngleZYX;
+		using glm::eulerAngleZYZ;
+		using glm::eulerAngles;
+		using glm::exp;
+		using glm::exp2;
+		using glm::exponentialEaseIn;
+		using glm::exponentialEaseInOut;
+		using glm::exponentialEaseOut;
+		using glm::extend;
+		using glm::extractEulerAngleXYX;
+		using glm::extractEulerAngleXYZ;
+		using glm::extractEulerAngleXZX;
+		using glm::extractEulerAngleXZY;
+		using glm::extractEulerAngleYXY;
+		using glm::extractEulerAngleYXZ;
+		using glm::extractEulerAngleYZX;
+		using glm::extractEulerAngleYZY;
+		using glm::extractEulerAngleZXY;
+		using glm::extractEulerAngleZXZ;
+		using glm::extractEulerAngleZYX;
+		using glm::extractEulerAngleZYZ;
+		using glm::extractMatrixRotation;
+		using glm::extractRealComponent;
+		using glm::faceforward;
+		using glm::factorial;
+		using glm::fastAcos;
+		using glm::fastAsin;
+		using glm::fastAtan;
+		using glm::fastCos;
+		using glm::fastDistance;
+		using glm::fastExp;
+		using glm::fastExp2;
+		using glm::fastInverseSqrt;
+		using glm::fastLength;
+		using glm::fastLog;
+		using glm::fastLog2;
+		using glm::fastMix;
+		using glm::fastNormalize;
+		using glm::fastNormalizeDot;
+		using glm::fastPow;
+		using glm::fastSin;
+		using glm::fastSqrt;
+		using glm::fastTan;
+		using glm::fclamp;
+		using glm::findLSB;
+		using glm::findMSB;
+		using glm::fliplr;
+		using glm::flipud;
+		using glm::floatBitsToInt;
+		using glm::floatBitsToUint;
+		using glm::floor;
+		using glm::floor_log2;
+		using glm::fma;
+		using glm::fmax;
+		using glm::fmin;
+		using glm::fmod;
+		using glm::four_over_pi;
+		using glm::fract;
+		using glm::frexp;
+		using glm::frustum;
+		using glm::frustumLH;
+		using glm::frustumLH_NO;
+		using glm::frustumLH_ZO;
+		using glm::frustumNO;
+		using glm::frustumRH;
+		using glm::frustumRH_NO;
+		using glm::frustumRH_ZO;
+		using glm::frustumZO;
+		using glm::gauss;
+		using glm::golden_ratio;
+		using glm::greaterThan;
+		using glm::greaterThanEqual;
+		using glm::half_pi;
+		using glm::hermite;
+		using glm::highestBitValue;
+		using glm::hsvColor;
+		using glm::identity;
+		using glm::imulExtended;
+		using glm::infinitePerspective;
+		using glm::infinitePerspectiveLH;
+		using glm::infinitePerspectiveRH;
+		using glm::intBitsToFloat;
+		using glm::intermediate;
+		using glm::interpolate;
+		using glm::intersectLineSphere;
+		using glm::intersectLineTriangle;
+		using glm::intersectRayPlane;
+		using glm::intersectRaySphere;
+		using glm::intersectRayTriangle;
+		using glm::inverse;
+		using glm::inversesqrt;
+		using glm::iround;
+		using glm::isCompNull;
+		using glm::isIdentity;
+		using glm::isNormalized;
+		using glm::isNull;
+		using glm::isOrthogonal;
+		using glm::isdenormal;
+		using glm::isfinite;
+		using glm::isinf;
+		using glm::isnan;
+		using glm::l1Norm;
+		using glm::l2Norm;
+		using glm::lMaxNorm;
+		using glm::ldexp;
+		using glm::leftHanded;
+		using glm::length;
+		using glm::length2;
+		using glm::lerp;
+		using glm::lessThan;
+		using glm::lessThanEqual;
+		using glm::linearGradient;
+		using glm::linearInterpolation;
+		using glm::ln_ln_two;
+		using glm::ln_ten;
+		using glm::ln_two;
+		using glm::log;
+		using glm::log2;
+		using glm::lookAt;
+		using glm::lookAtLH;
+		using glm::lookAtRH;
+		using glm::lowestBitValue;
+		using glm::luminosity;
+		using glm::lxNorm;
+		using glm::make_mat2;
+		using glm::make_mat2x2;
+		using glm::make_mat2x3;
+		using glm::make_mat2x4;
+		using glm::make_mat3;
+		using glm::make_mat3x2;
+		using glm::make_mat3x3;
+		using glm::make_mat3x4;
+		using glm::make_mat4;
+		using glm::make_mat4x2;
+		using glm::make_mat4x3;
+		using glm::make_mat4x4;
+		using glm::make_quat;
+		using glm::make_vec1;
+		using glm::make_vec2;
+		using glm::make_vec3;
+		using glm::make_vec4;
+		using glm::mask;
+		using glm::mat2x4_cast;
+		using glm::mat3_cast;
+		using glm::mat3x4_cast;
+		using glm::mat4_cast;
+		using glm::matrixCompMult;
+		using glm::matrixCross3;
+		using glm::matrixCross4;
+		using glm::max;
+		using glm::min;
+		using glm::mirrorClamp;
+		using glm::mirrorRepeat;
+		using glm::mix;
+		using glm::mixedProduct;
+		using glm::mod;
+		using glm::modf;
+		using glm::nlz;
+		using glm::normalize;
+		using glm::normalizeDot;
+		using glm::notEqual;
+		using glm::not_;
+		using glm::YCoCg2rgb;
+		using glm::YCoCgR2rgb;
+		using glm::one;
+		using glm::one_over_pi;
+		using glm::one_over_root_two;
+		using glm::one_over_two_pi;
+		using glm::openBounded;
+		using glm::orientate2;
+		using glm::orientate3;
+		using glm::orientate4;
+		using glm::orientation;
+		using glm::orientedAngle;
+		using glm::ortho;
+		using glm::orthoLH;
+		using glm::orthoLH_NO;
+		using glm::orthoLH_ZO;
+		using glm::orthoNO;
+		using glm::orthoRH;
+		using glm::orthoRH_NO;
+		using glm::orthoRH_ZO;
+		using glm::orthoZO;
+		using glm::orthonormalize;
+		using glm::outerProduct;
+		using glm::packDouble2x32;
+		using glm::packHalf2x16;
+		using glm::packSnorm2x16;
+		using glm::packSnorm4x8;
+		using glm::packUnorm2x16;
+		using glm::packUnorm4x8;
+		using glm::perp;
+		using glm::perspective;
+		using glm::perspectiveFov;
+		using glm::perspectiveFovLH;
+		using glm::perspectiveFovLH_NO;
+		using glm::perspectiveFovLH_ZO;
+		using glm::perspectiveFovNO;
+		using glm::perspectiveFovRH;
+		using glm::perspectiveFovRH_NO;
+		using glm::perspectiveFovRH_ZO;
+		using glm::perspectiveFovZO;
+		using glm::perspectiveLH;
+		using glm::perspectiveLH_NO;
+		using glm::perspectiveLH_ZO;
+		using glm::perspectiveNO;
+		using glm::perspectiveRH;
+		using glm::perspectiveRH_NO;
+		using glm::perspectiveRH_ZO;
+		using glm::perspectiveZO;
+		using glm::pi;
+		using glm::pickMatrix;
+		using glm::pitch;
+		using glm::polar;
+		using glm::pow;
+		using glm::pow2;
+		using glm::pow3;
+		using glm::pow4;
+		using glm::powerOfTwoAbove;
+		using glm::powerOfTwoBelow;
+		using glm::powerOfTwoNearest;
+		using glm::proj;
+		using glm::proj2D;
+		using glm::proj3D;
+		using glm::project;
+		using glm::projectNO;
+		using glm::projectZO;
+		using glm::qr_decompose;
+		using glm::quadraticEaseIn;
+		using glm::quadraticEaseInOut;
+		using glm::quadraticEaseOut;
+		using glm::quarter_pi;
+		using glm::quarticEaseIn;
+		using glm::quarticEaseInOut;
+		using glm::quarticEaseOut;
+		using glm::quatLookAt;
+		using glm::quatLookAtLH;
+		using glm::quatLookAtRH;
+		using glm::quat_cast;
+		using glm::quat_identity;
+		using glm::quinticEaseIn;
+		using glm::quinticEaseInOut;
+		using glm::quinticEaseOut;
+		using glm::radialGradient;
+		using glm::radians;
+		using glm::recompose;
+		using glm::reflect;
+		using glm::refract;
+		using glm::repeat;
+		using glm::rgb2YCoCg;
+		using glm::rgb2YCoCgR;
+		using glm::rgbColor;
+		using glm::rightHanded;
+		using glm::roll;
+		using glm::root_five;
+		using glm::root_half_pi;
+		using glm::root_ln_four;
+		using glm::root_pi;
+		using glm::root_three;
+		using glm::root_two;
+		using glm::root_two_pi;
+		using glm::rotate;
+		using glm::rotateNormalizedAxis;
+		using glm::rotateX;
+		using glm::rotateY;
+		using glm::rotateZ;
+		using glm::rotation;
+		using glm::round;
+		using glm::roundEven;
+		using glm::rowMajor2;
+		using glm::rowMajor3;
+		using glm::rowMajor4;
+		using glm::rq_decompose;
+		using glm::saturation;
+		using glm::scale;
+		using glm::scaleBias;
+		using glm::shearX2D;
+		using glm::shearX3D;
+		using glm::shearY2D;
+		using glm::shearY3D;
+		using glm::shearZ3D;
+		using glm::shortMix;
+		using glm::sign;
+		using glm::sin;
+		using glm::sineEaseIn;
+		using glm::sineEaseInOut;
+		using glm::sineEaseOut;
+		using glm::sinh;
+		using glm::slerp;
+		using glm::smoothstep;
+		using glm::sortEigenvalues;
+		using glm::sqrt;
+		using glm::squad;
+		using glm::step;
+		using glm::tan;
+		using glm::tanh;
+		using glm::third;
+		using glm::three_over_two_pi;
+		using glm::translate;
+		using glm::transpose;
+		using glm::triangleNormal;
+		using glm::trunc;
+		using glm::tweakedInfinitePerspective;
+		using glm::two_over_pi;
+		using glm::two_over_root_pi;
+		using glm::two_pi;
+		using glm::two_thirds;
+		using glm::uaddCarry;
+		using glm::uintBitsToFloat;
+		using glm::umulExtended;
+		using glm::unProject;
+		using glm::unProjectNO;
+		using glm::unProjectZO;
+		using glm::unpackDouble2x32;
+		using glm::unpackHalf2x16;
+		using glm::unpackSnorm2x16;
+		using glm::unpackSnorm4x8;
+		using glm::unpackUnorm2x16;
+		using glm::unpackUnorm4x8;
+		using glm::uround;
+		using glm::usubBorrow;
+		using glm::value_ptr;
+		using glm::wrapAngle;
+		using glm::wxyz;
+		using glm::yaw;
+		using glm::yawPitchRoll;
+		using glm::zero;
+	}
+#   endif
+}
+
+#if defined(_MSC_VER) // Workaround
+// Partial template specialization doesn't need to be exported explicitly, but this may not work otherwise on MSVC.
+export namespace std {
+	using std::hash; // See GLM_GTX_hash
+}
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/glm.hpp b/thirdparty/manifold/thirdparty/glm/glm/glm.hpp
new file mode 100644
index 000000000000..8b375459a784
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/glm.hpp
@@ -0,0 +1,137 @@
+/// @ref core
+/// @file glm/glm.hpp
+///
+/// @mainpage OpenGL Mathematics (GLM)
+/// - Website: <a href="https://glm.g-truc.net">glm.g-truc.net</a>
+/// - <a href="modules.html">GLM API documentation</a>
+/// - <a href="https://github.com/g-truc/glm/blob/master/manual.md">GLM Manual</a>
+///
+/// @defgroup core Core features
+///
+/// @brief Features that implement in C++ the GLSL specification as closely as possible.
+///
+/// The GLM core consists of C++ types that mirror GLSL types and
+/// C++ functions that mirror the GLSL functions.
+///
+/// The best documentation for GLM Core is the current GLSL specification,
+/// <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.clean.pdf">version 4.2
+/// (pdf file)</a>.
+///
+/// GLM core functionalities require <glm/glm.hpp> to be included to be used.
+///
+///
+/// @defgroup core_vector Vector types
+///
+/// Vector types of two to four components with an exhaustive set of operators.
+///
+/// @ingroup core
+///
+///
+/// @defgroup core_vector_precision Vector types with precision qualifiers
+///
+/// @brief Vector types with precision qualifiers which may result in various precision in term of ULPs
+///
+/// GLSL allows defining qualifiers for particular variables.
+/// With OpenGL's GLSL, these qualifiers have no effect; they are there for compatibility,
+/// with OpenGL ES's GLSL, these qualifiers do have an effect.
+///
+/// C++ has no language equivalent to qualifier qualifiers. So GLM provides the next-best thing:
+/// a number of typedefs that use a particular qualifier.
+///
+/// None of these types make any guarantees about the actual qualifier used.
+///
+/// @ingroup core
+///
+///
+/// @defgroup core_matrix Matrix types
+///
+/// Matrix types of with C columns and R rows where C and R are values between 2 to 4 included.
+/// These types have exhaustive sets of operators.
+///
+/// @ingroup core
+///
+///
+/// @defgroup core_matrix_precision Matrix types with precision qualifiers
+///
+/// @brief Matrix types with precision qualifiers which may result in various precision in term of ULPs
+///
+/// GLSL allows defining qualifiers for particular variables.
+/// With OpenGL's GLSL, these qualifiers have no effect; they are there for compatibility,
+/// with OpenGL ES's GLSL, these qualifiers do have an effect.
+///
+/// C++ has no language equivalent to qualifier qualifiers. So GLM provides the next-best thing:
+/// a number of typedefs that use a particular qualifier.
+///
+/// None of these types make any guarantees about the actual qualifier used.
+///
+/// @ingroup core
+///
+///
+/// @defgroup ext Stable extensions
+///
+/// @brief Additional features not specified by GLSL specification.
+///
+/// EXT extensions are fully tested and documented.
+///
+/// Even if it's highly unrecommended, it's possible to include all the extensions at once by
+/// including <glm/ext.hpp>. Otherwise, each extension needs to be included  a specific file.
+///
+///
+/// @defgroup gtc Recommended extensions
+///
+/// @brief Additional features not specified by GLSL specification.
+///
+/// GTC extensions aim to be stable with tests and documentation.
+///
+/// Even if it's highly unrecommended, it's possible to include all the extensions at once by
+/// including <glm/ext.hpp>. Otherwise, each extension needs to be included  a specific file.
+///
+///
+/// @defgroup gtx Experimental extensions
+///
+/// @brief Experimental features not specified by GLSL specification.
+///
+/// Experimental extensions are useful functions and types, but the development of
+/// their API and functionality is not necessarily stable. They can change
+/// substantially between versions. Backwards compatibility is not much of an issue
+/// for them.
+///
+/// Even if it's highly unrecommended, it's possible to include all the extensions
+/// at once by including <glm/ext.hpp>. Otherwise, each extension needs to be
+/// included  a specific file.
+///
+
+#include "detail/_fixes.hpp"
+
+#include "detail/setup.hpp"
+
+#pragma once
+
+#include <cmath>
+#include <climits>
+#include <cfloat>
+#include <limits>
+#include <cassert>
+#include "fwd.hpp"
+
+#include "vec2.hpp"
+#include "vec3.hpp"
+#include "vec4.hpp"
+#include "mat2x2.hpp"
+#include "mat2x3.hpp"
+#include "mat2x4.hpp"
+#include "mat3x2.hpp"
+#include "mat3x3.hpp"
+#include "mat3x4.hpp"
+#include "mat4x2.hpp"
+#include "mat4x3.hpp"
+#include "mat4x4.hpp"
+
+#include "trigonometric.hpp"
+#include "exponential.hpp"
+#include "common.hpp"
+#include "packing.hpp"
+#include "geometric.hpp"
+#include "matrix.hpp"
+#include "vector_relational.hpp"
+#include "integer.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.hpp
new file mode 100644
index 000000000000..084fbe75ff14
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.hpp
@@ -0,0 +1,266 @@
+/// @ref gtc_bitfield
+/// @file glm/gtc/bitfield.hpp
+///
+/// @see core (dependence)
+/// @see gtc_bitfield (dependence)
+///
+/// @defgroup gtc_bitfield GLM_GTC_bitfield
+/// @ingroup gtc
+///
+/// Include <glm/gtc/bitfield.hpp> to use the features of this extension.
+///
+/// Allow to perform bit operations on integer values
+
+#include "../detail/setup.hpp"
+
+#pragma once
+
+// Dependencies
+#include "../ext/scalar_int_sized.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/_vectorize.hpp"
+#include "type_precision.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_bitfield extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_bitfield
+	/// @{
+
+	/// Build a mask of 'count' bits
+	///
+	/// @see gtc_bitfield
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType mask(genIUType Bits);
+
+	/// Build a mask of 'count' bits
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed and unsigned integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_bitfield
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> mask(vec<L, T, Q> const& v);
+
+	/// Rotate all bits to the right. All the bits dropped in the right side are inserted back on the left side.
+	///
+	/// @see gtc_bitfield
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType bitfieldRotateRight(genIUType In, int Shift);
+
+	/// Rotate all bits to the right. All the bits dropped in the right side are inserted back on the left side.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed and unsigned integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_bitfield
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldRotateRight(vec<L, T, Q> const& In, int Shift);
+
+	/// Rotate all bits to the left. All the bits dropped in the left side are inserted back on the right side.
+	///
+	/// @see gtc_bitfield
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType bitfieldRotateLeft(genIUType In, int Shift);
+
+	/// Rotate all bits to the left. All the bits dropped in the left side are inserted back on the right side.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed and unsigned integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_bitfield
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldRotateLeft(vec<L, T, Q> const& In, int Shift);
+
+	/// Set to 1 a range of bits.
+	///
+	/// @see gtc_bitfield
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType bitfieldFillOne(genIUType Value, int FirstBit, int BitCount);
+
+	/// Set to 1 a range of bits.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed and unsigned integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_bitfield
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldFillOne(vec<L, T, Q> const& Value, int FirstBit, int BitCount);
+
+	/// Set to 0 a range of bits.
+	///
+	/// @see gtc_bitfield
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType bitfieldFillZero(genIUType Value, int FirstBit, int BitCount);
+
+	/// Set to 0 a range of bits.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Signed and unsigned integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_bitfield
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldFillZero(vec<L, T, Q> const& Value, int FirstBit, int BitCount);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of x followed by the first bit of y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int16 bitfieldInterleave(int8 x, int8 y);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of x followed by the first bit of y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint16 bitfieldInterleave(uint8 x, uint8 y);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of v.x followed by the first bit of v.y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint16 bitfieldInterleave(u8vec2 const& v);
+
+	/// Deinterleaves the bits of x.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL glm::u8vec2 bitfieldDeinterleave(glm::uint16 x);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of x followed by the first bit of y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int32 bitfieldInterleave(int16 x, int16 y);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of x followed by the first bit of y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint32 bitfieldInterleave(uint16 x, uint16 y);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of v.x followed by the first bit of v.y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint32 bitfieldInterleave(u16vec2 const& v);
+
+	/// Deinterleaves the bits of x.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL glm::u16vec2 bitfieldDeinterleave(glm::uint32 x);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of x followed by the first bit of y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int64 bitfieldInterleave(int32 x, int32 y);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of x followed by the first bit of y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint64 bitfieldInterleave(uint32 x, uint32 y);
+
+	/// Interleaves the bits of x and y.
+	/// The first bit is the first bit of v.x followed by the first bit of v.y.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint64 bitfieldInterleave(u32vec2 const& v);
+
+	/// Deinterleaves the bits of x.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL glm::u32vec2 bitfieldDeinterleave(glm::uint64 x);
+
+	/// Interleaves the bits of x, y and z.
+	/// The first bit is the first bit of x followed by the first bit of y and the first bit of z.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int32 bitfieldInterleave(int8 x, int8 y, int8 z);
+
+	/// Interleaves the bits of x, y and z.
+	/// The first bit is the first bit of x followed by the first bit of y and the first bit of z.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint32 bitfieldInterleave(uint8 x, uint8 y, uint8 z);
+
+	/// Interleaves the bits of x, y and z.
+	/// The first bit is the first bit of x followed by the first bit of y and the first bit of z.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int64 bitfieldInterleave(int16 x, int16 y, int16 z);
+
+	/// Interleaves the bits of x, y and z.
+	/// The first bit is the first bit of x followed by the first bit of y and the first bit of z.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint64 bitfieldInterleave(uint16 x, uint16 y, uint16 z);
+
+	/// Interleaves the bits of x, y and z.
+	/// The first bit is the first bit of x followed by the first bit of y and the first bit of z.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int64 bitfieldInterleave(int32 x, int32 y, int32 z);
+
+	/// Interleaves the bits of x, y and z.
+	/// The first bit is the first bit of x followed by the first bit of y and the first bit of z.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint64 bitfieldInterleave(uint32 x, uint32 y, uint32 z);
+
+	/// Interleaves the bits of x, y, z and w.
+	/// The first bit is the first bit of x followed by the first bit of y, the first bit of z and finally the first bit of w.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int32 bitfieldInterleave(int8 x, int8 y, int8 z, int8 w);
+
+	/// Interleaves the bits of x, y, z and w.
+	/// The first bit is the first bit of x followed by the first bit of y, the first bit of z and finally the first bit of w.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint32 bitfieldInterleave(uint8 x, uint8 y, uint8 z, uint8 w);
+
+	/// Interleaves the bits of x, y, z and w.
+	/// The first bit is the first bit of x followed by the first bit of y, the first bit of z and finally the first bit of w.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL int64 bitfieldInterleave(int16 x, int16 y, int16 z, int16 w);
+
+	/// Interleaves the bits of x, y, z and w.
+	/// The first bit is the first bit of x followed by the first bit of y, the first bit of z and finally the first bit of w.
+	/// The other bits are interleaved following the previous sequence.
+	///
+	/// @see gtc_bitfield
+	GLM_FUNC_DECL uint64 bitfieldInterleave(uint16 x, uint16 y, uint16 z, uint16 w);
+
+	/// @}
+} //namespace glm
+
+#include "bitfield.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.inl
new file mode 100644
index 000000000000..06cf1889cd40
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/bitfield.inl
@@ -0,0 +1,626 @@
+/// @ref gtc_bitfield
+
+#include "../simd/integer.h"
+
+namespace glm{
+namespace detail
+{
+	template<typename PARAM, typename RET>
+	GLM_FUNC_DECL RET bitfieldInterleave(PARAM x, PARAM y);
+
+	template<typename PARAM, typename RET>
+	GLM_FUNC_DECL RET bitfieldInterleave(PARAM x, PARAM y, PARAM z);
+
+	template<typename PARAM, typename RET>
+	GLM_FUNC_DECL RET bitfieldInterleave(PARAM x, PARAM y, PARAM z, PARAM w);
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint16 bitfieldInterleave(glm::uint8 x, glm::uint8 y)
+	{
+		glm::uint16 REG1(x);
+		glm::uint16 REG2(y);
+
+		REG1 = ((REG1 <<  4) | REG1) & static_cast<glm::uint16>(0x0F0F);
+		REG2 = ((REG2 <<  4) | REG2) & static_cast<glm::uint16>(0x0F0F);
+
+		REG1 = ((REG1 <<  2) | REG1) & static_cast<glm::uint16>(0x3333);
+		REG2 = ((REG2 <<  2) | REG2) & static_cast<glm::uint16>(0x3333);
+
+		REG1 = ((REG1 <<  1) | REG1) & static_cast<glm::uint16>(0x5555);
+		REG2 = ((REG2 <<  1) | REG2) & static_cast<glm::uint16>(0x5555);
+
+		return REG1 | static_cast<glm::uint16>(REG2 << 1);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint32 bitfieldInterleave(glm::uint16 x, glm::uint16 y)
+	{
+		glm::uint32 REG1(x);
+		glm::uint32 REG2(y);
+
+		REG1 = ((REG1 <<  8) | REG1) & static_cast<glm::uint32>(0x00FF00FF);
+		REG2 = ((REG2 <<  8) | REG2) & static_cast<glm::uint32>(0x00FF00FF);
+
+		REG1 = ((REG1 <<  4) | REG1) & static_cast<glm::uint32>(0x0F0F0F0F);
+		REG2 = ((REG2 <<  4) | REG2) & static_cast<glm::uint32>(0x0F0F0F0F);
+
+		REG1 = ((REG1 <<  2) | REG1) & static_cast<glm::uint32>(0x33333333);
+		REG2 = ((REG2 <<  2) | REG2) & static_cast<glm::uint32>(0x33333333);
+
+		REG1 = ((REG1 <<  1) | REG1) & static_cast<glm::uint32>(0x55555555);
+		REG2 = ((REG2 <<  1) | REG2) & static_cast<glm::uint32>(0x55555555);
+
+		return REG1 | (REG2 << 1);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(glm::uint32 x, glm::uint32 y)
+	{
+		glm::uint64 REG1(x);
+		glm::uint64 REG2(y);
+
+		REG1 = ((REG1 << 16) | REG1) & static_cast<glm::uint64>(0x0000FFFF0000FFFFull);
+		REG2 = ((REG2 << 16) | REG2) & static_cast<glm::uint64>(0x0000FFFF0000FFFFull);
+
+		REG1 = ((REG1 <<  8) | REG1) & static_cast<glm::uint64>(0x00FF00FF00FF00FFull);
+		REG2 = ((REG2 <<  8) | REG2) & static_cast<glm::uint64>(0x00FF00FF00FF00FFull);
+
+		REG1 = ((REG1 <<  4) | REG1) & static_cast<glm::uint64>(0x0F0F0F0F0F0F0F0Full);
+		REG2 = ((REG2 <<  4) | REG2) & static_cast<glm::uint64>(0x0F0F0F0F0F0F0F0Full);
+
+		REG1 = ((REG1 <<  2) | REG1) & static_cast<glm::uint64>(0x3333333333333333ull);
+		REG2 = ((REG2 <<  2) | REG2) & static_cast<glm::uint64>(0x3333333333333333ull);
+
+		REG1 = ((REG1 <<  1) | REG1) & static_cast<glm::uint64>(0x5555555555555555ull);
+		REG2 = ((REG2 <<  1) | REG2) & static_cast<glm::uint64>(0x5555555555555555ull);
+
+		return REG1 | (REG2 << 1);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint32 bitfieldInterleave(glm::uint8 x, glm::uint8 y, glm::uint8 z)
+	{
+		glm::uint32 REG1(x);
+		glm::uint32 REG2(y);
+		glm::uint32 REG3(z);
+
+		REG1 = ((REG1 << 16) | REG1) & static_cast<glm::uint32>(0xFF0000FFu);
+		REG2 = ((REG2 << 16) | REG2) & static_cast<glm::uint32>(0xFF0000FFu);
+		REG3 = ((REG3 << 16) | REG3) & static_cast<glm::uint32>(0xFF0000FFu);
+
+		REG1 = ((REG1 <<  8) | REG1) & static_cast<glm::uint32>(0x0F00F00Fu);
+		REG2 = ((REG2 <<  8) | REG2) & static_cast<glm::uint32>(0x0F00F00Fu);
+		REG3 = ((REG3 <<  8) | REG3) & static_cast<glm::uint32>(0x0F00F00Fu);
+
+		REG1 = ((REG1 <<  4) | REG1) & static_cast<glm::uint32>(0xC30C30C3u);
+		REG2 = ((REG2 <<  4) | REG2) & static_cast<glm::uint32>(0xC30C30C3u);
+		REG3 = ((REG3 <<  4) | REG3) & static_cast<glm::uint32>(0xC30C30C3u);
+
+		REG1 = ((REG1 <<  2) | REG1) & static_cast<glm::uint32>(0x49249249u);
+		REG2 = ((REG2 <<  2) | REG2) & static_cast<glm::uint32>(0x49249249u);
+		REG3 = ((REG3 <<  2) | REG3) & static_cast<glm::uint32>(0x49249249u);
+
+		return REG1 | (REG2 << 1) | (REG3 << 2);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(glm::uint16 x, glm::uint16 y, glm::uint16 z)
+	{
+		glm::uint64 REG1(x);
+		glm::uint64 REG2(y);
+		glm::uint64 REG3(z);
+
+		REG1 = ((REG1 << 32) | REG1) & static_cast<glm::uint64>(0xFFFF00000000FFFFull);
+		REG2 = ((REG2 << 32) | REG2) & static_cast<glm::uint64>(0xFFFF00000000FFFFull);
+		REG3 = ((REG3 << 32) | REG3) & static_cast<glm::uint64>(0xFFFF00000000FFFFull);
+
+		REG1 = ((REG1 << 16) | REG1) & static_cast<glm::uint64>(0x00FF0000FF0000FFull);
+		REG2 = ((REG2 << 16) | REG2) & static_cast<glm::uint64>(0x00FF0000FF0000FFull);
+		REG3 = ((REG3 << 16) | REG3) & static_cast<glm::uint64>(0x00FF0000FF0000FFull);
+
+		REG1 = ((REG1 <<  8) | REG1) & static_cast<glm::uint64>(0xF00F00F00F00F00Full);
+		REG2 = ((REG2 <<  8) | REG2) & static_cast<glm::uint64>(0xF00F00F00F00F00Full);
+		REG3 = ((REG3 <<  8) | REG3) & static_cast<glm::uint64>(0xF00F00F00F00F00Full);
+
+		REG1 = ((REG1 <<  4) | REG1) & static_cast<glm::uint64>(0x30C30C30C30C30C3ull);
+		REG2 = ((REG2 <<  4) | REG2) & static_cast<glm::uint64>(0x30C30C30C30C30C3ull);
+		REG3 = ((REG3 <<  4) | REG3) & static_cast<glm::uint64>(0x30C30C30C30C30C3ull);
+
+		REG1 = ((REG1 <<  2) | REG1) & static_cast<glm::uint64>(0x9249249249249249ull);
+		REG2 = ((REG2 <<  2) | REG2) & static_cast<glm::uint64>(0x9249249249249249ull);
+		REG3 = ((REG3 <<  2) | REG3) & static_cast<glm::uint64>(0x9249249249249249ull);
+
+		return REG1 | (REG2 << 1) | (REG3 << 2);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(glm::uint32 x, glm::uint32 y, glm::uint32 z)
+	{
+		glm::uint64 REG1(x);
+		glm::uint64 REG2(y);
+		glm::uint64 REG3(z);
+
+		REG1 = ((REG1 << 32) | REG1) & static_cast<glm::uint64>(0xFFFF00000000FFFFull);
+		REG2 = ((REG2 << 32) | REG2) & static_cast<glm::uint64>(0xFFFF00000000FFFFull);
+		REG3 = ((REG3 << 32) | REG3) & static_cast<glm::uint64>(0xFFFF00000000FFFFull);
+
+		REG1 = ((REG1 << 16) | REG1) & static_cast<glm::uint64>(0x00FF0000FF0000FFull);
+		REG2 = ((REG2 << 16) | REG2) & static_cast<glm::uint64>(0x00FF0000FF0000FFull);
+		REG3 = ((REG3 << 16) | REG3) & static_cast<glm::uint64>(0x00FF0000FF0000FFull);
+
+		REG1 = ((REG1 <<  8) | REG1) & static_cast<glm::uint64>(0xF00F00F00F00F00Full);
+		REG2 = ((REG2 <<  8) | REG2) & static_cast<glm::uint64>(0xF00F00F00F00F00Full);
+		REG3 = ((REG3 <<  8) | REG3) & static_cast<glm::uint64>(0xF00F00F00F00F00Full);
+
+		REG1 = ((REG1 <<  4) | REG1) & static_cast<glm::uint64>(0x30C30C30C30C30C3ull);
+		REG2 = ((REG2 <<  4) | REG2) & static_cast<glm::uint64>(0x30C30C30C30C30C3ull);
+		REG3 = ((REG3 <<  4) | REG3) & static_cast<glm::uint64>(0x30C30C30C30C30C3ull);
+
+		REG1 = ((REG1 <<  2) | REG1) & static_cast<glm::uint64>(0x9249249249249249ull);
+		REG2 = ((REG2 <<  2) | REG2) & static_cast<glm::uint64>(0x9249249249249249ull);
+		REG3 = ((REG3 <<  2) | REG3) & static_cast<glm::uint64>(0x9249249249249249ull);
+
+		return REG1 | (REG2 << 1) | (REG3 << 2);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint32 bitfieldInterleave(glm::uint8 x, glm::uint8 y, glm::uint8 z, glm::uint8 w)
+	{
+		glm::uint32 REG1(x);
+		glm::uint32 REG2(y);
+		glm::uint32 REG3(z);
+		glm::uint32 REG4(w);
+
+		REG1 = ((REG1 << 12) | REG1) & static_cast<glm::uint32>(0x000F000Fu);
+		REG2 = ((REG2 << 12) | REG2) & static_cast<glm::uint32>(0x000F000Fu);
+		REG3 = ((REG3 << 12) | REG3) & static_cast<glm::uint32>(0x000F000Fu);
+		REG4 = ((REG4 << 12) | REG4) & static_cast<glm::uint32>(0x000F000Fu);
+
+		REG1 = ((REG1 <<  6) | REG1) & static_cast<glm::uint32>(0x03030303u);
+		REG2 = ((REG2 <<  6) | REG2) & static_cast<glm::uint32>(0x03030303u);
+		REG3 = ((REG3 <<  6) | REG3) & static_cast<glm::uint32>(0x03030303u);
+		REG4 = ((REG4 <<  6) | REG4) & static_cast<glm::uint32>(0x03030303u);
+
+		REG1 = ((REG1 <<  3) | REG1) & static_cast<glm::uint32>(0x11111111u);
+		REG2 = ((REG2 <<  3) | REG2) & static_cast<glm::uint32>(0x11111111u);
+		REG3 = ((REG3 <<  3) | REG3) & static_cast<glm::uint32>(0x11111111u);
+		REG4 = ((REG4 <<  3) | REG4) & static_cast<glm::uint32>(0x11111111u);
+
+		return REG1 | (REG2 << 1) | (REG3 << 2) | (REG4 << 3);
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(glm::uint16 x, glm::uint16 y, glm::uint16 z, glm::uint16 w)
+	{
+		glm::uint64 REG1(x);
+		glm::uint64 REG2(y);
+		glm::uint64 REG3(z);
+		glm::uint64 REG4(w);
+
+		REG1 = ((REG1 << 24) | REG1) & static_cast<glm::uint64>(0x000000FF000000FFull);
+		REG2 = ((REG2 << 24) | REG2) & static_cast<glm::uint64>(0x000000FF000000FFull);
+		REG3 = ((REG3 << 24) | REG3) & static_cast<glm::uint64>(0x000000FF000000FFull);
+		REG4 = ((REG4 << 24) | REG4) & static_cast<glm::uint64>(0x000000FF000000FFull);
+
+		REG1 = ((REG1 << 12) | REG1) & static_cast<glm::uint64>(0x000F000F000F000Full);
+		REG2 = ((REG2 << 12) | REG2) & static_cast<glm::uint64>(0x000F000F000F000Full);
+		REG3 = ((REG3 << 12) | REG3) & static_cast<glm::uint64>(0x000F000F000F000Full);
+		REG4 = ((REG4 << 12) | REG4) & static_cast<glm::uint64>(0x000F000F000F000Full);
+
+		REG1 = ((REG1 <<  6) | REG1) & static_cast<glm::uint64>(0x0303030303030303ull);
+		REG2 = ((REG2 <<  6) | REG2) & static_cast<glm::uint64>(0x0303030303030303ull);
+		REG3 = ((REG3 <<  6) | REG3) & static_cast<glm::uint64>(0x0303030303030303ull);
+		REG4 = ((REG4 <<  6) | REG4) & static_cast<glm::uint64>(0x0303030303030303ull);
+
+		REG1 = ((REG1 <<  3) | REG1) & static_cast<glm::uint64>(0x1111111111111111ull);
+		REG2 = ((REG2 <<  3) | REG2) & static_cast<glm::uint64>(0x1111111111111111ull);
+		REG3 = ((REG3 <<  3) | REG3) & static_cast<glm::uint64>(0x1111111111111111ull);
+		REG4 = ((REG4 <<  3) | REG4) & static_cast<glm::uint64>(0x1111111111111111ull);
+
+		return REG1 | (REG2 << 1) | (REG3 << 2) | (REG4 << 3);
+	}
+}//namespace detail
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType mask(genIUType Bits)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'mask' accepts only integer values");
+
+		return Bits >= sizeof(genIUType) * 8 ? ~static_cast<genIUType>(0) : (static_cast<genIUType>(1) << Bits) - static_cast<genIUType>(1);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> mask(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'mask' accepts only integer values");
+
+		return detail::functor1<vec, L, T, T, Q>::call(mask, v);
+	}
+
+	template<typename genIType>
+	GLM_FUNC_QUALIFIER genIType bitfieldRotateRight(genIType In, int Shift)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'bitfieldRotateRight' accepts only integer values");
+
+		int const BitSize = static_cast<genIType>(sizeof(genIType) * 8);
+		return (In << static_cast<genIType>(Shift)) | (In >> static_cast<genIType>(BitSize - Shift));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldRotateRight(vec<L, T, Q> const& In, int Shift)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldRotateRight' accepts only integer values");
+
+		int const BitSize = static_cast<int>(sizeof(T) * 8);
+		return (In << static_cast<T>(Shift)) | (In >> static_cast<T>(BitSize - Shift));
+	}
+
+	template<typename genIType>
+	GLM_FUNC_QUALIFIER genIType bitfieldRotateLeft(genIType In, int Shift)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIType>::is_integer, "'bitfieldRotateLeft' accepts only integer values");
+
+		int const BitSize = static_cast<genIType>(sizeof(genIType) * 8);
+		return (In >> static_cast<genIType>(Shift)) | (In << static_cast<genIType>(BitSize - Shift));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldRotateLeft(vec<L, T, Q> const& In, int Shift)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldRotateLeft' accepts only integer values");
+
+		int const BitSize = static_cast<int>(sizeof(T) * 8);
+		return (In >> static_cast<T>(Shift)) | (In << static_cast<T>(BitSize - Shift));
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType bitfieldFillOne(genIUType Value, int FirstBit, int BitCount)
+	{
+		return Value | static_cast<genIUType>(mask(BitCount) << FirstBit);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldFillOne(vec<L, T, Q> const& Value, int FirstBit, int BitCount)
+	{
+		return Value | static_cast<T>(mask(BitCount) << FirstBit);
+	}
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType bitfieldFillZero(genIUType Value, int FirstBit, int BitCount)
+	{
+		return Value & static_cast<genIUType>(~(mask(BitCount) << FirstBit));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldFillZero(vec<L, T, Q> const& Value, int FirstBit, int BitCount)
+	{
+		return Value & static_cast<T>(~(mask(BitCount) << FirstBit));
+	}
+
+	GLM_FUNC_QUALIFIER int16 bitfieldInterleave(int8 x, int8 y)
+	{
+		union sign8
+		{
+			int8 i;
+			uint8 u;
+		} sign_x, sign_y;
+
+		union sign16
+		{
+			int16 i;
+			uint16 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint16 bitfieldInterleave(uint8 x, uint8 y)
+	{
+		return detail::bitfieldInterleave<uint8, uint16>(x, y);
+	}
+
+	GLM_FUNC_QUALIFIER uint16 bitfieldInterleave(u8vec2 const& v)
+	{
+		return detail::bitfieldInterleave<uint8, uint16>(v.x, v.y);
+	}
+
+	GLM_FUNC_QUALIFIER u8vec2 bitfieldDeinterleave(glm::uint16 x)
+	{
+		uint16 REG1(x);
+		uint16 REG2(x >>= 1);
+
+		REG1 = REG1 & static_cast<uint16>(0x5555);
+		REG2 = REG2 & static_cast<uint16>(0x5555);
+
+		REG1 = ((REG1 >> 1) | REG1) & static_cast<uint16>(0x3333);
+		REG2 = ((REG2 >> 1) | REG2) & static_cast<uint16>(0x3333);
+
+		REG1 = ((REG1 >> 2) | REG1) & static_cast<uint16>(0x0F0F);
+		REG2 = ((REG2 >> 2) | REG2) & static_cast<uint16>(0x0F0F);
+
+		REG1 = ((REG1 >> 4) | REG1) & static_cast<uint16>(0x00FF);
+		REG2 = ((REG2 >> 4) | REG2) & static_cast<uint16>(0x00FF);
+
+		REG1 = ((REG1 >> 8) | REG1) & static_cast<uint16>(0xFFFF);
+		REG2 = ((REG2 >> 8) | REG2) & static_cast<uint16>(0xFFFF);
+
+		return glm::u8vec2(REG1, REG2);
+	}
+
+	GLM_FUNC_QUALIFIER int32 bitfieldInterleave(int16 x, int16 y)
+	{
+		union sign16
+		{
+			int16 i;
+			uint16 u;
+		} sign_x, sign_y;
+
+		union sign32
+		{
+			int32 i;
+			uint32 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint32 bitfieldInterleave(uint16 x, uint16 y)
+	{
+		return detail::bitfieldInterleave<uint16, uint32>(x, y);
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint32 bitfieldInterleave(u16vec2 const& v)
+	{
+		return detail::bitfieldInterleave<uint16, uint32>(v.x, v.y);
+	}
+
+	GLM_FUNC_QUALIFIER glm::u16vec2 bitfieldDeinterleave(glm::uint32 x)
+	{
+		glm::uint32 REG1(x);
+		glm::uint32 REG2(x >>= 1);
+
+		REG1 = REG1 & static_cast<glm::uint32>(0x55555555);
+		REG2 = REG2 & static_cast<glm::uint32>(0x55555555);
+
+		REG1 = ((REG1 >> 1) | REG1) & static_cast<glm::uint32>(0x33333333);
+		REG2 = ((REG2 >> 1) | REG2) & static_cast<glm::uint32>(0x33333333);
+
+		REG1 = ((REG1 >> 2) | REG1) & static_cast<glm::uint32>(0x0F0F0F0F);
+		REG2 = ((REG2 >> 2) | REG2) & static_cast<glm::uint32>(0x0F0F0F0F);
+
+		REG1 = ((REG1 >> 4) | REG1) & static_cast<glm::uint32>(0x00FF00FF);
+		REG2 = ((REG2 >> 4) | REG2) & static_cast<glm::uint32>(0x00FF00FF);
+
+		REG1 = ((REG1 >> 8) | REG1) & static_cast<glm::uint32>(0x0000FFFF);
+		REG2 = ((REG2 >> 8) | REG2) & static_cast<glm::uint32>(0x0000FFFF);
+
+		return glm::u16vec2(REG1, REG2);
+	}
+
+	GLM_FUNC_QUALIFIER int64 bitfieldInterleave(int32 x, int32 y)
+	{
+		union sign32
+		{
+			int32 i;
+			uint32 u;
+		} sign_x, sign_y;
+
+		union sign64
+		{
+			int64 i;
+			uint64 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(uint32 x, uint32 y)
+	{
+		return detail::bitfieldInterleave<uint32, uint64>(x, y);
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint64 bitfieldInterleave(u32vec2 const& v)
+	{
+		return detail::bitfieldInterleave<uint32, uint64>(v.x, v.y);
+	}
+
+	GLM_FUNC_QUALIFIER glm::u32vec2 bitfieldDeinterleave(glm::uint64 x)
+	{
+		glm::uint64 REG1(x);
+		glm::uint64 REG2(x >>= 1);
+
+		REG1 = REG1 & static_cast<glm::uint64>(0x5555555555555555ull);
+		REG2 = REG2 & static_cast<glm::uint64>(0x5555555555555555ull);
+
+		REG1 = ((REG1 >> 1) | REG1) & static_cast<glm::uint64>(0x3333333333333333ull);
+		REG2 = ((REG2 >> 1) | REG2) & static_cast<glm::uint64>(0x3333333333333333ull);
+
+		REG1 = ((REG1 >> 2) | REG1) & static_cast<glm::uint64>(0x0F0F0F0F0F0F0F0Full);
+		REG2 = ((REG2 >> 2) | REG2) & static_cast<glm::uint64>(0x0F0F0F0F0F0F0F0Full);
+
+		REG1 = ((REG1 >> 4) | REG1) & static_cast<glm::uint64>(0x00FF00FF00FF00FFull);
+		REG2 = ((REG2 >> 4) | REG2) & static_cast<glm::uint64>(0x00FF00FF00FF00FFull);
+
+		REG1 = ((REG1 >> 8) | REG1) & static_cast<glm::uint64>(0x0000FFFF0000FFFFull);
+		REG2 = ((REG2 >> 8) | REG2) & static_cast<glm::uint64>(0x0000FFFF0000FFFFull);
+
+		REG1 = ((REG1 >> 16) | REG1) & static_cast<glm::uint64>(0x00000000FFFFFFFFull);
+		REG2 = ((REG2 >> 16) | REG2) & static_cast<glm::uint64>(0x00000000FFFFFFFFull);
+
+		return glm::u32vec2(REG1, REG2);
+	}
+
+	GLM_FUNC_QUALIFIER int32 bitfieldInterleave(int8 x, int8 y, int8 z)
+	{
+		union sign8
+		{
+			int8 i;
+			uint8 u;
+		} sign_x, sign_y, sign_z;
+
+		union sign32
+		{
+			int32 i;
+			uint32 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		sign_z.i = z;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u, sign_z.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint32 bitfieldInterleave(uint8 x, uint8 y, uint8 z)
+	{
+		return detail::bitfieldInterleave<uint8, uint32>(x, y, z);
+	}
+
+	GLM_FUNC_QUALIFIER uint32 bitfieldInterleave(u8vec3 const& v)
+	{
+		return detail::bitfieldInterleave<uint8, uint32>(v.x, v.y, v.z);
+	}
+
+	GLM_FUNC_QUALIFIER int64 bitfieldInterleave(int16 x, int16 y, int16 z)
+	{
+		union sign16
+		{
+			int16 i;
+			uint16 u;
+		} sign_x, sign_y, sign_z;
+
+		union sign64
+		{
+			int64 i;
+			uint64 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		sign_z.i = z;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u, sign_z.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(uint16 x, uint16 y, uint16 z)
+	{
+		return detail::bitfieldInterleave<uint32, uint64>(x, y, z);
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(u16vec3 const& v)
+	{
+		return detail::bitfieldInterleave<uint32, uint64>(v.x, v.y, v.z);
+	}
+
+	GLM_FUNC_QUALIFIER int64 bitfieldInterleave(int32 x, int32 y, int32 z)
+	{
+		union sign16
+		{
+			int32 i;
+			uint32 u;
+		} sign_x, sign_y, sign_z;
+
+		union sign64
+		{
+			int64 i;
+			uint64 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		sign_z.i = z;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u, sign_z.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(uint32 x, uint32 y, uint32 z)
+	{
+		return detail::bitfieldInterleave<uint32, uint64>(x, y, z);
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(u32vec3 const& v)
+	{
+		return detail::bitfieldInterleave<uint32, uint64>(v.x, v.y, v.z);
+	}
+
+	GLM_FUNC_QUALIFIER int32 bitfieldInterleave(int8 x, int8 y, int8 z, int8 w)
+	{
+		union sign8
+		{
+			int8 i;
+			uint8 u;
+		} sign_x, sign_y, sign_z, sign_w;
+
+		union sign32
+		{
+			int32 i;
+			uint32 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		sign_z.i = z;
+		sign_w.i = w;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u, sign_z.u, sign_w.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint32 bitfieldInterleave(uint8 x, uint8 y, uint8 z, uint8 w)
+	{
+		return detail::bitfieldInterleave<uint8, uint32>(x, y, z, w);
+	}
+
+	GLM_FUNC_QUALIFIER uint32 bitfieldInterleave(u8vec4 const& v)
+	{
+		return detail::bitfieldInterleave<uint8, uint32>(v.x, v.y, v.z, v.w);
+	}
+
+	GLM_FUNC_QUALIFIER int64 bitfieldInterleave(int16 x, int16 y, int16 z, int16 w)
+	{
+		union sign16
+		{
+			int16 i;
+			uint16 u;
+		} sign_x, sign_y, sign_z, sign_w;
+
+		union sign64
+		{
+			int64 i;
+			uint64 u;
+		} result;
+
+		sign_x.i = x;
+		sign_y.i = y;
+		sign_z.i = z;
+		sign_w.i = w;
+		result.u = bitfieldInterleave(sign_x.u, sign_y.u, sign_z.u, sign_w.u);
+
+		return result.i;
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(uint16 x, uint16 y, uint16 z, uint16 w)
+	{
+		return detail::bitfieldInterleave<uint16, uint64>(x, y, z, w);
+	}
+
+	GLM_FUNC_QUALIFIER uint64 bitfieldInterleave(u16vec4 const& v)
+	{
+		return detail::bitfieldInterleave<uint16, uint64>(v.x, v.y, v.z, v.w);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.hpp
new file mode 100644
index 000000000000..cffd9f093fb9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.hpp
@@ -0,0 +1,56 @@
+/// @ref gtc_color_space
+/// @file glm/gtc/color_space.hpp
+///
+/// @see core (dependence)
+/// @see gtc_color_space (dependence)
+///
+/// @defgroup gtc_color_space GLM_GTC_color_space
+/// @ingroup gtc
+///
+/// Include <glm/gtc/color_space.hpp> to use the features of this extension.
+///
+/// Allow to perform bit operations on integer values
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../exponential.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_color_space extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_color_space
+	/// @{
+
+	/// Convert a linear color to sRGB color using a standard gamma correction.
+	/// IEC 61966-2-1:1999 / Rec. 709 specification https://www.w3.org/Graphics/Color/srgb
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> convertLinearToSRGB(vec<L, T, Q> const& ColorLinear);
+
+	/// Convert a linear color to sRGB color using a custom gamma correction.
+	/// IEC 61966-2-1:1999 / Rec. 709 specification https://www.w3.org/Graphics/Color/srgb
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> convertLinearToSRGB(vec<L, T, Q> const& ColorLinear, T Gamma);
+
+	/// Convert a sRGB color to linear color using a standard gamma correction.
+	/// IEC 61966-2-1:1999 / Rec. 709 specification https://www.w3.org/Graphics/Color/srgb
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> convertSRGBToLinear(vec<L, T, Q> const& ColorSRGB);
+
+	/// Convert a sRGB color to linear color using a custom gamma correction.
+	// IEC 61966-2-1:1999 / Rec. 709 specification https://www.w3.org/Graphics/Color/srgb
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> convertSRGBToLinear(vec<L, T, Q> const& ColorSRGB, T Gamma);
+
+	/// @}
+} //namespace glm
+
+#include "color_space.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.inl
new file mode 100644
index 000000000000..2a900044e99b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/color_space.inl
@@ -0,0 +1,84 @@
+/// @ref gtc_color_space
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q>
+	struct compute_rgbToSrgb
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& ColorRGB, T GammaCorrection)
+		{
+			vec<L, T, Q> const ClampedColor(clamp(ColorRGB, static_cast<T>(0), static_cast<T>(1)));
+
+			return mix(
+				pow(ClampedColor, vec<L, T, Q>(GammaCorrection)) * static_cast<T>(1.055) - static_cast<T>(0.055),
+				ClampedColor * static_cast<T>(12.92),
+				lessThan(ClampedColor, vec<L, T, Q>(static_cast<T>(0.0031308))));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_rgbToSrgb<4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, T, Q> call(vec<4, T, Q> const& ColorRGB, T GammaCorrection)
+		{
+			return vec<4, T, Q>(compute_rgbToSrgb<3, T, Q>::call(vec<3, T, Q>(ColorRGB), GammaCorrection), ColorRGB.w);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct compute_srgbToRgb
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& ColorSRGB, T Gamma)
+		{
+			return mix(
+				pow((ColorSRGB + static_cast<T>(0.055)) * static_cast<T>(0.94786729857819905213270142180095), vec<L, T, Q>(Gamma)),
+				ColorSRGB * static_cast<T>(0.07739938080495356037151702786378),
+				lessThanEqual(ColorSRGB, vec<L, T, Q>(static_cast<T>(0.04045))));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_srgbToRgb<4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, T, Q> call(vec<4, T, Q> const& ColorSRGB, T Gamma)
+		{
+			return vec<4, T, Q>(compute_srgbToRgb<3, T, Q>::call(vec<3, T, Q>(ColorSRGB), Gamma), ColorSRGB.w);
+		}
+	};
+}//namespace detail
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> convertLinearToSRGB(vec<L, T, Q> const& ColorLinear)
+	{
+		return detail::compute_rgbToSrgb<L, T, Q>::call(ColorLinear, static_cast<T>(0.41666));
+	}
+
+	// Based on Ian Taylor http://chilliant.blogspot.fr/2012/08/srgb-approximations-for-hlsl.html
+	template<>
+	GLM_FUNC_QUALIFIER vec<3, float, lowp> convertLinearToSRGB(vec<3, float, lowp> const& ColorLinear)
+	{
+		vec<3, float, lowp> S1 = sqrt(ColorLinear);
+		vec<3, float, lowp> S2 = sqrt(S1);
+		vec<3, float, lowp> S3 = sqrt(S2);
+		return 0.662002687f * S1 + 0.684122060f * S2 - 0.323583601f * S3 - 0.0225411470f * ColorLinear;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> convertLinearToSRGB(vec<L, T, Q> const& ColorLinear, T Gamma)
+	{
+		return detail::compute_rgbToSrgb<L, T, Q>::call(ColorLinear, static_cast<T>(1) / Gamma);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> convertSRGBToLinear(vec<L, T, Q> const& ColorSRGB)
+	{
+		return detail::compute_srgbToRgb<L, T, Q>::call(ColorSRGB, static_cast<T>(2.4));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> convertSRGBToLinear(vec<L, T, Q> const& ColorSRGB, T Gamma)
+	{
+		return detail::compute_srgbToRgb<L, T, Q>::call(ColorSRGB, Gamma);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/constants.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/constants.hpp
new file mode 100644
index 000000000000..6a1f37d30f95
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/constants.hpp
@@ -0,0 +1,170 @@
+/// @ref gtc_constants
+/// @file glm/gtc/constants.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_constants GLM_GTC_constants
+/// @ingroup gtc
+///
+/// Include <glm/gtc/constants.hpp> to use the features of this extension.
+///
+/// Provide a list of constants and precomputed useful values.
+
+#pragma once
+
+// Dependencies
+#include "../ext/scalar_constants.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_constants extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_constants
+	/// @{
+
+	/// Return 0.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType zero();
+
+	/// Return 1.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType one();
+
+	/// Return pi * 2.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType two_pi();
+
+	/// Return unit-circle circumference, or pi * 2.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType tau();
+
+	/// Return square root of pi.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_pi();
+
+	/// Return pi / 2.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType half_pi();
+
+	/// Return pi / 2 * 3.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType three_over_two_pi();
+
+	/// Return pi / 4.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType quarter_pi();
+
+	/// Return 1 / pi.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType one_over_pi();
+
+	/// Return 1 / (pi * 2).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType one_over_two_pi();
+
+	/// Return 2 / pi.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType two_over_pi();
+
+	/// Return 4 / pi.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType four_over_pi();
+
+	/// Return 2 / sqrt(pi).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType two_over_root_pi();
+
+	/// Return 1 / sqrt(2).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType one_over_root_two();
+
+	/// Return sqrt(pi / 2).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_half_pi();
+
+	/// Return sqrt(2 * pi).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_two_pi();
+
+	/// Return sqrt(ln(4)).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_ln_four();
+
+	/// Return e constant.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType e();
+
+	/// Return Euler's constant.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType euler();
+
+	/// Return sqrt(2).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_two();
+
+	/// Return sqrt(3).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_three();
+
+	/// Return sqrt(5).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType root_five();
+
+	/// Return ln(2).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType ln_two();
+
+	/// Return ln(10).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType ln_ten();
+
+	/// Return ln(ln(2)).
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType ln_ln_two();
+
+	/// Return 1 / 3.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType third();
+
+	/// Return 2 / 3.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType two_thirds();
+
+	/// Return the golden ratio constant.
+	/// @see gtc_constants
+	template<typename genType>
+	GLM_FUNC_DECL GLM_CONSTEXPR genType golden_ratio();
+
+	/// @}
+} //namespace glm
+
+#include "constants.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/constants.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/constants.inl
new file mode 100644
index 000000000000..e9d3776148d0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/constants.inl
@@ -0,0 +1,173 @@
+/// @ref gtc_constants
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType zero()
+	{
+		return genType(0);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType one()
+	{
+		return genType(1);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType two_pi()
+	{
+		return genType(6.28318530717958647692528676655900576);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType tau()
+	{
+		return two_pi<genType>();
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_pi()
+	{
+		return genType(1.772453850905516027);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType half_pi()
+	{
+		return genType(1.57079632679489661923132169163975144);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType three_over_two_pi()
+	{
+		return genType(4.71238898038468985769396507491925432);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType quarter_pi()
+	{
+		return genType(0.785398163397448309615660845819875721);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType one_over_pi()
+	{
+		return genType(0.318309886183790671537767526745028724);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType one_over_two_pi()
+	{
+		return genType(0.159154943091895335768883763372514362);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType two_over_pi()
+	{
+		return genType(0.636619772367581343075535053490057448);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType four_over_pi()
+	{
+		return genType(1.273239544735162686151070106980114898);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType two_over_root_pi()
+	{
+		return genType(1.12837916709551257389615890312154517);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType one_over_root_two()
+	{
+		return genType(0.707106781186547524400844362104849039);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_half_pi()
+	{
+		return genType(1.253314137315500251);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_two_pi()
+	{
+		return genType(2.506628274631000502);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_ln_four()
+	{
+		return genType(1.17741002251547469);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType e()
+	{
+		return genType(2.71828182845904523536);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType euler()
+	{
+		return genType(0.577215664901532860606);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_two()
+	{
+		return genType(1.41421356237309504880168872420969808);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_three()
+	{
+		return genType(1.73205080756887729352744634150587236);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType root_five()
+	{
+		return genType(2.23606797749978969640917366873127623);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType ln_two()
+	{
+		return genType(0.693147180559945309417232121458176568);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType ln_ten()
+	{
+		return genType(2.30258509299404568401799145468436421);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType ln_ln_two()
+	{
+		return genType(-0.3665129205816643);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType third()
+	{
+		return genType(0.3333333333333333333333333333333333333333);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType two_thirds()
+	{
+		return genType(0.666666666666666666666666666666666666667);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR genType golden_ratio()
+	{
+		return genType(1.61803398874989484820458683436563811);
+	}
+
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.hpp
new file mode 100644
index 000000000000..640439b11c36
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.hpp
@@ -0,0 +1,60 @@
+/// @ref gtc_epsilon
+/// @file glm/gtc/epsilon.hpp
+///
+/// @see core (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtc_epsilon GLM_GTC_epsilon
+/// @ingroup gtc
+///
+/// Include <glm/gtc/epsilon.hpp> to use the features of this extension.
+///
+/// Comparison functions for a user defined epsilon values.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_epsilon extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_epsilon
+	/// @{
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @see gtc_epsilon
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> epsilonEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T const& epsilon);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is satisfied.
+	///
+	/// @see gtc_epsilon
+	template<typename genType>
+	GLM_FUNC_DECL bool epsilonEqual(genType const& x, genType const& y, genType const& epsilon);
+
+	/// Returns the component-wise comparison of |x - y| < epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @see gtc_epsilon
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> epsilonNotEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T const& epsilon);
+
+	/// Returns the component-wise comparison of |x - y| >= epsilon.
+	/// True if this expression is not satisfied.
+	///
+	/// @see gtc_epsilon
+	template<typename genType>
+	GLM_FUNC_DECL bool epsilonNotEqual(genType const& x, genType const& y, genType const& epsilon);
+
+	/// @}
+}//namespace glm
+
+#include "epsilon.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.inl
new file mode 100644
index 000000000000..508b9f8966fe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/epsilon.inl
@@ -0,0 +1,80 @@
+/// @ref gtc_epsilon
+
+// Dependency:
+#include "../vector_relational.hpp"
+#include "../common.hpp"
+
+namespace glm
+{
+	template<>
+	GLM_FUNC_QUALIFIER bool epsilonEqual
+	(
+		float const& x,
+		float const& y,
+		float const& epsilon
+	)
+	{
+		return abs(x - y) < epsilon;
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER bool epsilonEqual
+	(
+		double const& x,
+		double const& y,
+		double const& epsilon
+	)
+	{
+		return abs(x - y) < epsilon;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> epsilonEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T const& epsilon)
+	{
+		return lessThan(abs(x - y), vec<L, T, Q>(epsilon));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> epsilonEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& epsilon)
+	{
+		return lessThan(abs(x - y), vec<L, T, Q>(epsilon));
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER bool epsilonNotEqual(float const& x, float const& y, float const& epsilon)
+	{
+		return abs(x - y) >= epsilon;
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER bool epsilonNotEqual(double const& x, double const& y, double const& epsilon)
+	{
+		return abs(x - y) >= epsilon;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> epsilonNotEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, T const& epsilon)
+	{
+		return greaterThanEqual(abs(x - y), vec<L, T, Q>(epsilon));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> epsilonNotEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y, vec<L, T, Q> const& epsilon)
+	{
+		return greaterThanEqual(abs(x - y), vec<L, T, Q>(epsilon));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> epsilonEqual(qua<T, Q> const& x, qua<T, Q> const& y, T const& epsilon)
+	{
+		vec<4, T, Q> v(x.x - y.x, x.y - y.y, x.z - y.z, x.w - y.w);
+		return lessThan(abs(v), vec<4, T, Q>(epsilon));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> epsilonNotEqual(qua<T, Q> const& x, qua<T, Q> const& y, T const& epsilon)
+	{
+		vec<4, T, Q> v(x.x - y.x, x.y - y.y, x.z - y.z, x.w - y.w);
+		return greaterThanEqual(abs(v), vec<4, T, Q>(epsilon));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/integer.hpp
new file mode 100644
index 000000000000..ee52e0e0c154
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/integer.hpp
@@ -0,0 +1,43 @@
+/// @ref gtc_integer
+/// @file glm/gtc/integer.hpp
+///
+/// @see core (dependence)
+/// @see gtc_integer (dependence)
+///
+/// @defgroup gtc_integer GLM_GTC_integer
+/// @ingroup gtc
+///
+/// Include <glm/gtc/integer.hpp> to use the features of this extension.
+///
+/// @brief Allow to perform bit operations on integer values
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../common.hpp"
+#include "../integer.hpp"
+#include "../exponential.hpp"
+#include "../ext/scalar_common.hpp"
+#include "../ext/vector_common.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_integer extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_integer
+	/// @{
+
+	/// Returns the log2 of x for integer values. Useful to compute mipmap count from the texture size.
+	/// @see gtc_integer
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType log2(genIUType x);
+
+	/// @}
+} //namespace glm
+
+#include "integer.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/integer.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/integer.inl
new file mode 100644
index 000000000000..5f66dfe2c096
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/integer.inl
@@ -0,0 +1,33 @@
+/// @ref gtc_integer
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_log2<L, T, Q, false, Aligned>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v)
+		{
+			//Equivalent to return findMSB(vec); but save one function call in ASM with VC
+			//return findMSB(vec);
+			return vec<L, T, Q>(detail::compute_findMSB_vec<L, T, Q, sizeof(T) * 8>::call(v));
+		}
+	};
+
+#	if GLM_HAS_BITSCAN_WINDOWS
+		template<qualifier Q, bool Aligned>
+		struct compute_log2<4, int, Q, false, Aligned>
+		{
+			GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v)
+			{
+				vec<4, int, Q> Result;
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.x), v.x);
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.y), v.y);
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.z), v.z);
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.w), v.w);
+				return Result;
+			}
+		};
+#	endif//GLM_HAS_BITSCAN_WINDOWS
+}//namespace detail
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.hpp
new file mode 100644
index 000000000000..4935ba755dd5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.hpp
@@ -0,0 +1,60 @@
+/// @ref gtc_matrix_access
+/// @file glm/gtc/matrix_access.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_matrix_access GLM_GTC_matrix_access
+/// @ingroup gtc
+///
+/// Include <glm/gtc/matrix_access.hpp> to use the features of this extension.
+///
+/// Defines functions to access rows or columns of a matrix easily.
+
+#pragma once
+
+// Dependency:
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_matrix_access extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_matrix_access
+	/// @{
+
+	/// Get a specific row of a matrix.
+	/// @see gtc_matrix_access
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::row_type row(
+		genType const& m,
+		length_t index);
+
+	/// Set a specific row to a matrix.
+	/// @see gtc_matrix_access
+	template<typename genType>
+	GLM_FUNC_DECL genType row(
+		genType const& m,
+		length_t index,
+		typename genType::row_type const& x);
+
+	/// Get a specific column of a matrix.
+	/// @see gtc_matrix_access
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::col_type column(
+		genType const& m,
+		length_t index);
+
+	/// Set a specific column to a matrix.
+	/// @see gtc_matrix_access
+	template<typename genType>
+	GLM_FUNC_DECL genType column(
+		genType const& m,
+		length_t index,
+		typename genType::col_type const& x);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_access.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.inl
new file mode 100644
index 000000000000..09fcc10d3d7e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_access.inl
@@ -0,0 +1,62 @@
+/// @ref gtc_matrix_access
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType row
+	(
+		genType const& m,
+		length_t index,
+		typename genType::row_type const& x
+	)
+	{
+		assert(index >= 0 && index < m[0].length());
+
+		genType Result = m;
+		for(length_t i = 0; i < m.length(); ++i)
+			Result[i][index] = x[i];
+		return Result;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER typename genType::row_type row
+	(
+		genType const& m,
+		length_t index
+	)
+	{
+		assert(index >= 0 && index < m[0].length());
+
+		typename genType::row_type Result(0);
+		for(length_t i = 0; i < m.length(); ++i)
+			Result[i] = m[i][index];
+		return Result;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType column
+	(
+		genType const& m,
+		length_t index,
+		typename genType::col_type const& x
+	)
+	{
+		assert(index >= 0 && index < m.length());
+
+		genType Result = m;
+		Result[index] = x;
+		return Result;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER typename genType::col_type column
+	(
+		genType const& m,
+		length_t index
+	)
+	{
+		assert(index >= 0 && index < m.length());
+
+		return m[index];
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_integer.hpp
new file mode 100644
index 000000000000..d7ebdc719221
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_integer.hpp
@@ -0,0 +1,433 @@
+/// @ref gtc_matrix_integer
+/// @file glm/gtc/matrix_integer.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_matrix_integer GLM_GTC_matrix_integer
+/// @ingroup gtc
+///
+/// Include <glm/gtc/matrix_integer.hpp> to use the features of this extension.
+///
+/// Defines a number of matrices with integer types.
+
+#pragma once
+
+// Dependency:
+#include "../mat2x2.hpp"
+#include "../mat2x3.hpp"
+#include "../mat2x4.hpp"
+#include "../mat3x2.hpp"
+#include "../mat3x3.hpp"
+#include "../mat3x4.hpp"
+#include "../mat4x2.hpp"
+#include "../mat4x3.hpp"
+#include "../mat4x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_matrix_integer extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_matrix_integer
+	/// @{
+
+	/// High-qualifier signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, highp>				highp_imat2;
+
+	/// High-qualifier signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, highp>				highp_imat3;
+
+	/// High-qualifier signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, highp>				highp_imat4;
+
+	/// High-qualifier signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, highp>				highp_imat2x2;
+
+	/// High-qualifier signed integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, int, highp>				highp_imat2x3;
+
+	/// High-qualifier signed integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, int, highp>				highp_imat2x4;
+
+	/// High-qualifier signed integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, int, highp>				highp_imat3x2;
+
+	/// High-qualifier signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, highp>				highp_imat3x3;
+
+	/// High-qualifier signed integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, int, highp>				highp_imat3x4;
+
+	/// High-qualifier signed integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, int, highp>				highp_imat4x2;
+
+	/// High-qualifier signed integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, int, highp>				highp_imat4x3;
+
+	/// High-qualifier signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, highp>				highp_imat4x4;
+
+
+	/// Medium-qualifier signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, mediump>			mediump_imat2;
+
+	/// Medium-qualifier signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, mediump>			mediump_imat3;
+
+	/// Medium-qualifier signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, mediump>			mediump_imat4;
+
+
+	/// Medium-qualifier signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, mediump>			mediump_imat2x2;
+
+	/// Medium-qualifier signed integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, int, mediump>			mediump_imat2x3;
+
+	/// Medium-qualifier signed integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, int, mediump>			mediump_imat2x4;
+
+	/// Medium-qualifier signed integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, int, mediump>			mediump_imat3x2;
+
+	/// Medium-qualifier signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, mediump>			mediump_imat3x3;
+
+	/// Medium-qualifier signed integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, int, mediump>			mediump_imat3x4;
+
+	/// Medium-qualifier signed integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, int, mediump>			mediump_imat4x2;
+
+	/// Medium-qualifier signed integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, int, mediump>			mediump_imat4x3;
+
+	/// Medium-qualifier signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, mediump>			mediump_imat4x4;
+
+
+	/// Low-qualifier signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, lowp>				lowp_imat2;
+
+	/// Low-qualifier signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, lowp>				lowp_imat3;
+
+	/// Low-qualifier signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, lowp>				lowp_imat4;
+
+
+	/// Low-qualifier signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, lowp>				lowp_imat2x2;
+
+	/// Low-qualifier signed integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, int, lowp>				lowp_imat2x3;
+
+	/// Low-qualifier signed integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, int, lowp>				lowp_imat2x4;
+
+	/// Low-qualifier signed integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, int, lowp>				lowp_imat3x2;
+
+	/// Low-qualifier signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, lowp>				lowp_imat3x3;
+
+	/// Low-qualifier signed integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, int, lowp>				lowp_imat3x4;
+
+	/// Low-qualifier signed integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, int, lowp>				lowp_imat4x2;
+
+	/// Low-qualifier signed integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, int, lowp>				lowp_imat4x3;
+
+	/// Low-qualifier signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, lowp>				lowp_imat4x4;
+
+
+	/// High-qualifier unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, highp>				highp_umat2;
+
+	/// High-qualifier unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, highp>				highp_umat3;
+
+	/// High-qualifier unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, highp>				highp_umat4;
+
+	/// High-qualifier unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, highp>				highp_umat2x2;
+
+	/// High-qualifier unsigned integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, uint, highp>				highp_umat2x3;
+
+	/// High-qualifier unsigned integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, uint, highp>				highp_umat2x4;
+
+	/// High-qualifier unsigned integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, uint, highp>				highp_umat3x2;
+
+	/// High-qualifier unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, highp>				highp_umat3x3;
+
+	/// High-qualifier unsigned integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, uint, highp>				highp_umat3x4;
+
+	/// High-qualifier unsigned integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, uint, highp>				highp_umat4x2;
+
+	/// High-qualifier unsigned integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, uint, highp>				highp_umat4x3;
+
+	/// High-qualifier unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, highp>				highp_umat4x4;
+
+
+	/// Medium-qualifier unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, mediump>			mediump_umat2;
+
+	/// Medium-qualifier unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, mediump>			mediump_umat3;
+
+	/// Medium-qualifier unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, mediump>			mediump_umat4;
+
+
+	/// Medium-qualifier unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, mediump>			mediump_umat2x2;
+
+	/// Medium-qualifier unsigned integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, uint, mediump>			mediump_umat2x3;
+
+	/// Medium-qualifier unsigned integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, uint, mediump>			mediump_umat2x4;
+
+	/// Medium-qualifier unsigned integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, uint, mediump>			mediump_umat3x2;
+
+	/// Medium-qualifier unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, mediump>			mediump_umat3x3;
+
+	/// Medium-qualifier unsigned integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, uint, mediump>			mediump_umat3x4;
+
+	/// Medium-qualifier unsigned integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, uint, mediump>			mediump_umat4x2;
+
+	/// Medium-qualifier unsigned integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, uint, mediump>			mediump_umat4x3;
+
+	/// Medium-qualifier unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, mediump>			mediump_umat4x4;
+
+
+	/// Low-qualifier unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, lowp>				lowp_umat2;
+
+	/// Low-qualifier unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, lowp>				lowp_umat3;
+
+	/// Low-qualifier unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, lowp>				lowp_umat4;
+
+
+	/// Low-qualifier unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, lowp>				lowp_umat2x2;
+
+	/// Low-qualifier unsigned integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, uint, lowp>				lowp_umat2x3;
+
+	/// Low-qualifier unsigned integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, uint, lowp>				lowp_umat2x4;
+
+	/// Low-qualifier unsigned integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, uint, lowp>				lowp_umat3x2;
+
+	/// Low-qualifier unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, lowp>				lowp_umat3x3;
+
+	/// Low-qualifier unsigned integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, uint, lowp>				lowp_umat3x4;
+
+	/// Low-qualifier unsigned integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, uint, lowp>				lowp_umat4x2;
+
+	/// Low-qualifier unsigned integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, uint, lowp>				lowp_umat4x3;
+
+	/// Low-qualifier unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, lowp>				lowp_umat4x4;
+
+
+
+	/// Signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, defaultp>				imat2;
+
+	/// Signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, defaultp>				imat3;
+
+	/// Signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, defaultp>				imat4;
+
+	/// Signed integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, int, defaultp>				imat2x2;
+
+	/// Signed integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, int, defaultp>				imat2x3;
+
+	/// Signed integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, int, defaultp>				imat2x4;
+
+	/// Signed integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, int, defaultp>				imat3x2;
+
+	/// Signed integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, int, defaultp>				imat3x3;
+
+	/// Signed integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, int, defaultp>				imat3x4;
+
+	/// Signed integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, int, defaultp>				imat4x2;
+
+	/// Signed integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, int, defaultp>				imat4x3;
+
+	/// Signed integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, int, defaultp>				imat4x4;
+
+
+
+	/// Unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, defaultp>				umat2;
+
+	/// Unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, defaultp>				umat3;
+
+	/// Unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, defaultp>				umat4;
+
+	/// Unsigned integer 2x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 2, uint, defaultp>				umat2x2;
+
+	/// Unsigned integer 2x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 3, uint, defaultp>				umat2x3;
+
+	/// Unsigned integer 2x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<2, 4, uint, defaultp>				umat2x4;
+
+	/// Unsigned integer 3x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 2, uint, defaultp>				umat3x2;
+
+	/// Unsigned integer 3x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 3, uint, defaultp>				umat3x3;
+
+	/// Unsigned integer 3x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<3, 4, uint, defaultp>				umat3x4;
+
+	/// Unsigned integer 4x2 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 2, uint, defaultp>				umat4x2;
+
+	/// Unsigned integer 4x3 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 3, uint, defaultp>				umat4x3;
+
+	/// Unsigned integer 4x4 matrix.
+	/// @see gtc_matrix_integer
+	typedef mat<4, 4, uint, defaultp>				umat4x4;
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.hpp
new file mode 100644
index 000000000000..75d53f20234d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.hpp
@@ -0,0 +1,50 @@
+/// @ref gtc_matrix_inverse
+/// @file glm/gtc/matrix_inverse.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_matrix_inverse GLM_GTC_matrix_inverse
+/// @ingroup gtc
+///
+/// Include <glm/gtc/matrix_inverse.hpp> to use the features of this extension.
+///
+/// Defines additional matrix inverting functions.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../matrix.hpp"
+#include "../mat2x2.hpp"
+#include "../mat3x3.hpp"
+#include "../mat4x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_matrix_inverse extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_matrix_inverse
+	/// @{
+
+	/// Fast matrix inverse for affine matrix.
+	///
+	/// @param m Input matrix to invert.
+	/// @tparam genType Squared floating-point matrix: half, float or double. Inverse of matrix based of half-qualifier floating point value is highly inaccurate.
+	/// @see gtc_matrix_inverse
+	template<typename genType>
+	GLM_FUNC_DECL genType affineInverse(genType const& m);
+
+	/// Compute the inverse transpose of a matrix.
+	///
+	/// @param m Input matrix to invert transpose.
+	/// @tparam genType Squared floating-point matrix: half, float or double. Inverse of matrix based of half-qualifier floating point value is highly inaccurate.
+	/// @see gtc_matrix_inverse
+	template<typename genType>
+	GLM_FUNC_DECL genType inverseTranspose(genType const& m);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_inverse.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.inl
new file mode 100644
index 000000000000..c004b9e14670
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_inverse.inl
@@ -0,0 +1,118 @@
+/// @ref gtc_matrix_inverse
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> affineInverse(mat<3, 3, T, Q> const& m)
+	{
+		mat<2, 2, T, Q> const Inv(inverse(mat<2, 2, T, Q>(m)));
+
+		return mat<3, 3, T, Q>(
+			vec<3, T, Q>(Inv[0], static_cast<T>(0)),
+			vec<3, T, Q>(Inv[1], static_cast<T>(0)),
+			vec<3, T, Q>(-Inv * vec<2, T, Q>(m[2]), static_cast<T>(1)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> affineInverse(mat<4, 4, T, Q> const& m)
+	{
+		mat<3, 3, T, Q> const Inv(inverse(mat<3, 3, T, Q>(m)));
+
+		return mat<4, 4, T, Q>(
+			vec<4, T, Q>(Inv[0], static_cast<T>(0)),
+			vec<4, T, Q>(Inv[1], static_cast<T>(0)),
+			vec<4, T, Q>(Inv[2], static_cast<T>(0)),
+			vec<4, T, Q>(-Inv * vec<3, T, Q>(m[3]), static_cast<T>(1)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> inverseTranspose(mat<2, 2, T, Q> const& m)
+	{
+		T Determinant = m[0][0] * m[1][1] - m[1][0] * m[0][1];
+
+		mat<2, 2, T, Q> Inverse(
+			+ m[1][1] / Determinant,
+			- m[0][1] / Determinant,
+			- m[1][0] / Determinant,
+			+ m[0][0] / Determinant);
+
+		return Inverse;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> inverseTranspose(mat<3, 3, T, Q> const& m)
+	{
+		T Determinant =
+			+ m[0][0] * (m[1][1] * m[2][2] - m[1][2] * m[2][1])
+			- m[0][1] * (m[1][0] * m[2][2] - m[1][2] * m[2][0])
+			+ m[0][2] * (m[1][0] * m[2][1] - m[1][1] * m[2][0]);
+
+		mat<3, 3, T, Q> Inverse;
+		Inverse[0][0] = + (m[1][1] * m[2][2] - m[2][1] * m[1][2]);
+		Inverse[0][1] = - (m[1][0] * m[2][2] - m[2][0] * m[1][2]);
+		Inverse[0][2] = + (m[1][0] * m[2][1] - m[2][0] * m[1][1]);
+		Inverse[1][0] = - (m[0][1] * m[2][2] - m[2][1] * m[0][2]);
+		Inverse[1][1] = + (m[0][0] * m[2][2] - m[2][0] * m[0][2]);
+		Inverse[1][2] = - (m[0][0] * m[2][1] - m[2][0] * m[0][1]);
+		Inverse[2][0] = + (m[0][1] * m[1][2] - m[1][1] * m[0][2]);
+		Inverse[2][1] = - (m[0][0] * m[1][2] - m[1][0] * m[0][2]);
+		Inverse[2][2] = + (m[0][0] * m[1][1] - m[1][0] * m[0][1]);
+		Inverse /= Determinant;
+
+		return Inverse;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> inverseTranspose(mat<4, 4, T, Q> const& m)
+	{
+		T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		T SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		T SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		T SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		T SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		T SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		T SubFactor11 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		T SubFactor12 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+		T SubFactor13 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+		T SubFactor14 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+		T SubFactor15 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+		T SubFactor16 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+		T SubFactor17 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		mat<4, 4, T, Q> Inverse;
+		Inverse[0][0] = + (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02);
+		Inverse[0][1] = - (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04);
+		Inverse[0][2] = + (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05);
+		Inverse[0][3] = - (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05);
+
+		Inverse[1][0] = - (m[0][1] * SubFactor00 - m[0][2] * SubFactor01 + m[0][3] * SubFactor02);
+		Inverse[1][1] = + (m[0][0] * SubFactor00 - m[0][2] * SubFactor03 + m[0][3] * SubFactor04);
+		Inverse[1][2] = - (m[0][0] * SubFactor01 - m[0][1] * SubFactor03 + m[0][3] * SubFactor05);
+		Inverse[1][3] = + (m[0][0] * SubFactor02 - m[0][1] * SubFactor04 + m[0][2] * SubFactor05);
+
+		Inverse[2][0] = + (m[0][1] * SubFactor06 - m[0][2] * SubFactor07 + m[0][3] * SubFactor08);
+		Inverse[2][1] = - (m[0][0] * SubFactor06 - m[0][2] * SubFactor09 + m[0][3] * SubFactor10);
+		Inverse[2][2] = + (m[0][0] * SubFactor07 - m[0][1] * SubFactor09 + m[0][3] * SubFactor11);
+		Inverse[2][3] = - (m[0][0] * SubFactor08 - m[0][1] * SubFactor10 + m[0][2] * SubFactor11);
+
+		Inverse[3][0] = - (m[0][1] * SubFactor12 - m[0][2] * SubFactor13 + m[0][3] * SubFactor14);
+		Inverse[3][1] = + (m[0][0] * SubFactor12 - m[0][2] * SubFactor15 + m[0][3] * SubFactor16);
+		Inverse[3][2] = - (m[0][0] * SubFactor13 - m[0][1] * SubFactor15 + m[0][3] * SubFactor17);
+		Inverse[3][3] = + (m[0][0] * SubFactor14 - m[0][1] * SubFactor16 + m[0][2] * SubFactor17);
+
+		T Determinant =
+			+ m[0][0] * Inverse[0][0]
+			+ m[0][1] * Inverse[0][1]
+			+ m[0][2] * Inverse[0][2]
+			+ m[0][3] * Inverse[0][3];
+
+		Inverse /= Determinant;
+
+		return Inverse;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.hpp
new file mode 100644
index 000000000000..612418fa51c4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.hpp
@@ -0,0 +1,36 @@
+/// @ref gtc_matrix_transform
+/// @file glm/gtc/matrix_transform.hpp
+///
+/// @see core (dependence)
+/// @see gtx_transform
+/// @see gtx_transform2
+///
+/// @defgroup gtc_matrix_transform GLM_GTC_matrix_transform
+/// @ingroup gtc
+///
+/// Include <glm/gtc/matrix_transform.hpp> to use the features of this extension.
+///
+/// Defines functions that generate common transformation matrices.
+///
+/// The matrices generated by this extension use standard OpenGL fixed-function
+/// conventions. For example, the lookAt function generates a transform from world
+/// space into the specific eye space that the projective matrix functions
+/// (perspective, ortho, etc) are designed to expect. The OpenGL compatibility
+/// specifications defines the particular layout of this eye space.
+
+#pragma once
+
+// Dependencies
+#include "../mat4x4.hpp"
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../ext/matrix_projection.hpp"
+#include "../ext/matrix_clip_space.hpp"
+#include "../ext/matrix_transform.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_matrix_transform extension included")
+#endif
+
+#include "matrix_transform.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.inl
new file mode 100644
index 000000000000..15b46bc9db61
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/matrix_transform.inl
@@ -0,0 +1,3 @@
+#include "../geometric.hpp"
+#include "../trigonometric.hpp"
+#include "../matrix.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/noise.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/noise.hpp
new file mode 100644
index 000000000000..ab1772e78125
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/noise.hpp
@@ -0,0 +1,61 @@
+/// @ref gtc_noise
+/// @file glm/gtc/noise.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_noise GLM_GTC_noise
+/// @ingroup gtc
+///
+/// Include <glm/gtc/noise.hpp> to use the features of this extension.
+///
+/// Defines 2D, 3D and 4D procedural noise functions
+/// Based on the work of Stefan Gustavson and Ashima Arts on "webgl-noise":
+/// https://github.com/ashima/webgl-noise
+/// Following Stefan Gustavson's paper "Simplex noise demystified":
+/// http://www.itn.liu.se/~stegu/simplexnoise/simplexnoise.pdf
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/_noise.hpp"
+#include "../geometric.hpp"
+#include "../common.hpp"
+#include "../vector_relational.hpp"
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_noise extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_noise
+	/// @{
+
+	/// Classic perlin noise.
+	/// @see gtc_noise
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T perlin(
+		vec<L, T, Q> const& p);
+
+	/// Periodic perlin noise.
+	/// @see gtc_noise
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T perlin(
+		vec<L, T, Q> const& p,
+		vec<L, T, Q> const& rep);
+
+	/// Simplex noise.
+	/// @see gtc_noise
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T simplex(
+		vec<L, T, Q> const& p);
+
+	/// @}
+}//namespace glm
+
+#include "noise.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/noise.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/noise.inl
new file mode 100644
index 000000000000..a1cf399d38d4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/noise.inl
@@ -0,0 +1,807 @@
+/// @ref gtc_noise
+///
+// Based on the work of Stefan Gustavson and Ashima Arts on "webgl-noise":
+// https://github.com/ashima/webgl-noise
+// Following Stefan Gustavson's paper "Simplex noise demystified":
+// http://www.itn.liu.se/~stegu/simplexnoise/simplexnoise.pdf
+
+namespace glm{
+namespace detail
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> grad4(T const& j, vec<4, T, Q> const& ip)
+	{
+		vec<3, T, Q> pXYZ = floor(fract(vec<3, T, Q>(j) * vec<3, T, Q>(ip)) * T(7)) * ip[2] - T(1);
+		T pW = static_cast<T>(1.5) - dot(abs(pXYZ), vec<3, T, Q>(1));
+		vec<4, T, Q> s = vec<4, T, Q>(lessThan(vec<4, T, Q>(pXYZ, pW), vec<4, T, Q>(0.0)));
+		pXYZ = pXYZ + (vec<3, T, Q>(s) * T(2) - T(1)) * s.w;
+		return vec<4, T, Q>(pXYZ, pW);
+	}
+}//namespace detail
+
+	// Classic Perlin noise
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<2, T, Q> const& Position)
+	{
+		vec<4, T, Q> Pi = glm::floor(vec<4, T, Q>(Position.x, Position.y, Position.x, Position.y)) + vec<4, T, Q>(0.0, 0.0, 1.0, 1.0);
+		vec<4, T, Q> Pf = glm::fract(vec<4, T, Q>(Position.x, Position.y, Position.x, Position.y)) - vec<4, T, Q>(0.0, 0.0, 1.0, 1.0);
+		Pi = mod(Pi, vec<4, T, Q>(289)); // To avoid truncation effects in permutation
+		vec<4, T, Q> ix(Pi.x, Pi.z, Pi.x, Pi.z);
+		vec<4, T, Q> iy(Pi.y, Pi.y, Pi.w, Pi.w);
+		vec<4, T, Q> fx(Pf.x, Pf.z, Pf.x, Pf.z);
+		vec<4, T, Q> fy(Pf.y, Pf.y, Pf.w, Pf.w);
+
+		vec<4, T, Q> i = detail::permute(detail::permute(ix) + iy);
+
+		vec<4, T, Q> gx = static_cast<T>(2) * glm::fract(i / T(41)) - T(1);
+		vec<4, T, Q> gy = glm::abs(gx) - T(0.5);
+		vec<4, T, Q> tx = glm::floor(gx + T(0.5));
+		gx = gx - tx;
+
+		vec<2, T, Q> g00(gx.x, gy.x);
+		vec<2, T, Q> g10(gx.y, gy.y);
+		vec<2, T, Q> g01(gx.z, gy.z);
+		vec<2, T, Q> g11(gx.w, gy.w);
+
+		vec<4, T, Q> norm = detail::taylorInvSqrt(vec<4, T, Q>(dot(g00, g00), dot(g01, g01), dot(g10, g10), dot(g11, g11)));
+		g00 *= norm.x;
+		g01 *= norm.y;
+		g10 *= norm.z;
+		g11 *= norm.w;
+
+		T n00 = dot(g00, vec<2, T, Q>(fx.x, fy.x));
+		T n10 = dot(g10, vec<2, T, Q>(fx.y, fy.y));
+		T n01 = dot(g01, vec<2, T, Q>(fx.z, fy.z));
+		T n11 = dot(g11, vec<2, T, Q>(fx.w, fy.w));
+
+		vec<2, T, Q> fade_xy = detail::fade(vec<2, T, Q>(Pf.x, Pf.y));
+		vec<2, T, Q> n_x = mix(vec<2, T, Q>(n00, n01), vec<2, T, Q>(n10, n11), fade_xy.x);
+		T n_xy = mix(n_x.x, n_x.y, fade_xy.y);
+		return T(2.3) * n_xy;
+	}
+
+	// Classic Perlin noise
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<3, T, Q> const& Position)
+	{
+		vec<3, T, Q> Pi0 = floor(Position); // Integer part for indexing
+		vec<3, T, Q> Pi1 = Pi0 + T(1); // Integer part + 1
+		Pi0 = detail::mod289(Pi0);
+		Pi1 = detail::mod289(Pi1);
+		vec<3, T, Q> Pf0 = fract(Position); // Fractional part for interpolation
+		vec<3, T, Q> Pf1 = Pf0 - T(1); // Fractional part - 1.0
+		vec<4, T, Q> ix(Pi0.x, Pi1.x, Pi0.x, Pi1.x);
+		vec<4, T, Q> iy = vec<4, T, Q>(vec<2, T, Q>(Pi0.y), vec<2, T, Q>(Pi1.y));
+		vec<4, T, Q> iz0(Pi0.z);
+		vec<4, T, Q> iz1(Pi1.z);
+
+		vec<4, T, Q> ixy = detail::permute(detail::permute(ix) + iy);
+		vec<4, T, Q> ixy0 = detail::permute(ixy + iz0);
+		vec<4, T, Q> ixy1 = detail::permute(ixy + iz1);
+
+		vec<4, T, Q> gx0 = ixy0 * T(1.0 / 7.0);
+		vec<4, T, Q> gy0 = fract(floor(gx0) * T(1.0 / 7.0)) - T(0.5);
+		gx0 = fract(gx0);
+		vec<4, T, Q> gz0 = vec<4, T, Q>(0.5) - abs(gx0) - abs(gy0);
+		vec<4, T, Q> sz0 = step(gz0, vec<4, T, Q>(0.0));
+		gx0 -= sz0 * (step(T(0), gx0) - T(0.5));
+		gy0 -= sz0 * (step(T(0), gy0) - T(0.5));
+
+		vec<4, T, Q> gx1 = ixy1 * T(1.0 / 7.0);
+		vec<4, T, Q> gy1 = fract(floor(gx1) * T(1.0 / 7.0)) - T(0.5);
+		gx1 = fract(gx1);
+		vec<4, T, Q> gz1 = vec<4, T, Q>(0.5) - abs(gx1) - abs(gy1);
+		vec<4, T, Q> sz1 = step(gz1, vec<4, T, Q>(0.0));
+		gx1 -= sz1 * (step(T(0), gx1) - T(0.5));
+		gy1 -= sz1 * (step(T(0), gy1) - T(0.5));
+
+		vec<3, T, Q> g000(gx0.x, gy0.x, gz0.x);
+		vec<3, T, Q> g100(gx0.y, gy0.y, gz0.y);
+		vec<3, T, Q> g010(gx0.z, gy0.z, gz0.z);
+		vec<3, T, Q> g110(gx0.w, gy0.w, gz0.w);
+		vec<3, T, Q> g001(gx1.x, gy1.x, gz1.x);
+		vec<3, T, Q> g101(gx1.y, gy1.y, gz1.y);
+		vec<3, T, Q> g011(gx1.z, gy1.z, gz1.z);
+		vec<3, T, Q> g111(gx1.w, gy1.w, gz1.w);
+
+		vec<4, T, Q> norm0 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g000, g000), dot(g010, g010), dot(g100, g100), dot(g110, g110)));
+		g000 *= norm0.x;
+		g010 *= norm0.y;
+		g100 *= norm0.z;
+		g110 *= norm0.w;
+		vec<4, T, Q> norm1 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g001, g001), dot(g011, g011), dot(g101, g101), dot(g111, g111)));
+		g001 *= norm1.x;
+		g011 *= norm1.y;
+		g101 *= norm1.z;
+		g111 *= norm1.w;
+
+		T n000 = dot(g000, Pf0);
+		T n100 = dot(g100, vec<3, T, Q>(Pf1.x, Pf0.y, Pf0.z));
+		T n010 = dot(g010, vec<3, T, Q>(Pf0.x, Pf1.y, Pf0.z));
+		T n110 = dot(g110, vec<3, T, Q>(Pf1.x, Pf1.y, Pf0.z));
+		T n001 = dot(g001, vec<3, T, Q>(Pf0.x, Pf0.y, Pf1.z));
+		T n101 = dot(g101, vec<3, T, Q>(Pf1.x, Pf0.y, Pf1.z));
+		T n011 = dot(g011, vec<3, T, Q>(Pf0.x, Pf1.y, Pf1.z));
+		T n111 = dot(g111, Pf1);
+
+		vec<3, T, Q> fade_xyz = detail::fade(Pf0);
+		vec<4, T, Q> n_z = mix(vec<4, T, Q>(n000, n100, n010, n110), vec<4, T, Q>(n001, n101, n011, n111), fade_xyz.z);
+		vec<2, T, Q> n_yz = mix(vec<2, T, Q>(n_z.x, n_z.y), vec<2, T, Q>(n_z.z, n_z.w), fade_xyz.y);
+		T n_xyz = mix(n_yz.x, n_yz.y, fade_xyz.x);
+		return T(2.2) * n_xyz;
+	}
+	/*
+	// Classic Perlin noise
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<3, T, Q> const& P)
+	{
+		vec<3, T, Q> Pi0 = floor(P); // Integer part for indexing
+		vec<3, T, Q> Pi1 = Pi0 + T(1); // Integer part + 1
+		Pi0 = mod(Pi0, T(289));
+		Pi1 = mod(Pi1, T(289));
+		vec<3, T, Q> Pf0 = fract(P); // Fractional part for interpolation
+		vec<3, T, Q> Pf1 = Pf0 - T(1); // Fractional part - 1.0
+		vec<4, T, Q> ix(Pi0.x, Pi1.x, Pi0.x, Pi1.x);
+		vec<4, T, Q> iy(Pi0.y, Pi0.y, Pi1.y, Pi1.y);
+		vec<4, T, Q> iz0(Pi0.z);
+		vec<4, T, Q> iz1(Pi1.z);
+
+		vec<4, T, Q> ixy = permute(permute(ix) + iy);
+		vec<4, T, Q> ixy0 = permute(ixy + iz0);
+		vec<4, T, Q> ixy1 = permute(ixy + iz1);
+
+		vec<4, T, Q> gx0 = ixy0 / T(7);
+		vec<4, T, Q> gy0 = fract(floor(gx0) / T(7)) - T(0.5);
+		gx0 = fract(gx0);
+		vec<4, T, Q> gz0 = vec<4, T, Q>(0.5) - abs(gx0) - abs(gy0);
+		vec<4, T, Q> sz0 = step(gz0, vec<4, T, Q>(0.0));
+		gx0 -= sz0 * (step(0.0, gx0) - T(0.5));
+		gy0 -= sz0 * (step(0.0, gy0) - T(0.5));
+
+		vec<4, T, Q> gx1 = ixy1 / T(7);
+		vec<4, T, Q> gy1 = fract(floor(gx1) / T(7)) - T(0.5);
+		gx1 = fract(gx1);
+		vec<4, T, Q> gz1 = vec<4, T, Q>(0.5) - abs(gx1) - abs(gy1);
+		vec<4, T, Q> sz1 = step(gz1, vec<4, T, Q>(0.0));
+		gx1 -= sz1 * (step(T(0), gx1) - T(0.5));
+		gy1 -= sz1 * (step(T(0), gy1) - T(0.5));
+
+		vec<3, T, Q> g000(gx0.x, gy0.x, gz0.x);
+		vec<3, T, Q> g100(gx0.y, gy0.y, gz0.y);
+		vec<3, T, Q> g010(gx0.z, gy0.z, gz0.z);
+		vec<3, T, Q> g110(gx0.w, gy0.w, gz0.w);
+		vec<3, T, Q> g001(gx1.x, gy1.x, gz1.x);
+		vec<3, T, Q> g101(gx1.y, gy1.y, gz1.y);
+		vec<3, T, Q> g011(gx1.z, gy1.z, gz1.z);
+		vec<3, T, Q> g111(gx1.w, gy1.w, gz1.w);
+
+		vec<4, T, Q> norm0 = taylorInvSqrt(vec<4, T, Q>(dot(g000, g000), dot(g010, g010), dot(g100, g100), dot(g110, g110)));
+		g000 *= norm0.x;
+		g010 *= norm0.y;
+		g100 *= norm0.z;
+		g110 *= norm0.w;
+		vec<4, T, Q> norm1 = taylorInvSqrt(vec<4, T, Q>(dot(g001, g001), dot(g011, g011), dot(g101, g101), dot(g111, g111)));
+		g001 *= norm1.x;
+		g011 *= norm1.y;
+		g101 *= norm1.z;
+		g111 *= norm1.w;
+
+		T n000 = dot(g000, Pf0);
+		T n100 = dot(g100, vec<3, T, Q>(Pf1.x, Pf0.y, Pf0.z));
+		T n010 = dot(g010, vec<3, T, Q>(Pf0.x, Pf1.y, Pf0.z));
+		T n110 = dot(g110, vec<3, T, Q>(Pf1.x, Pf1.y, Pf0.z));
+		T n001 = dot(g001, vec<3, T, Q>(Pf0.x, Pf0.y, Pf1.z));
+		T n101 = dot(g101, vec<3, T, Q>(Pf1.x, Pf0.y, Pf1.z));
+		T n011 = dot(g011, vec<3, T, Q>(Pf0.x, Pf1.y, Pf1.z));
+		T n111 = dot(g111, Pf1);
+
+		vec<3, T, Q> fade_xyz = fade(Pf0);
+		vec<4, T, Q> n_z = mix(vec<4, T, Q>(n000, n100, n010, n110), vec<4, T, Q>(n001, n101, n011, n111), fade_xyz.z);
+		vec<2, T, Q> n_yz = mix(
+			vec<2, T, Q>(n_z.x, n_z.y),
+			vec<2, T, Q>(n_z.z, n_z.w), fade_xyz.y);
+		T n_xyz = mix(n_yz.x, n_yz.y, fade_xyz.x);
+		return T(2.2) * n_xyz;
+	}
+	*/
+	// Classic Perlin noise
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<4, T, Q> const& Position)
+	{
+		vec<4, T, Q> Pi0 = floor(Position);	// Integer part for indexing
+		vec<4, T, Q> Pi1 = Pi0 + T(1);		// Integer part + 1
+		Pi0 = mod(Pi0, vec<4, T, Q>(289));
+		Pi1 = mod(Pi1, vec<4, T, Q>(289));
+		vec<4, T, Q> Pf0 = fract(Position);	// Fractional part for interpolation
+		vec<4, T, Q> Pf1 = Pf0 - T(1);		// Fractional part - 1.0
+		vec<4, T, Q> ix(Pi0.x, Pi1.x, Pi0.x, Pi1.x);
+		vec<4, T, Q> iy(Pi0.y, Pi0.y, Pi1.y, Pi1.y);
+		vec<4, T, Q> iz0(Pi0.z);
+		vec<4, T, Q> iz1(Pi1.z);
+		vec<4, T, Q> iw0(Pi0.w);
+		vec<4, T, Q> iw1(Pi1.w);
+
+		vec<4, T, Q> ixy = detail::permute(detail::permute(ix) + iy);
+		vec<4, T, Q> ixy0 = detail::permute(ixy + iz0);
+		vec<4, T, Q> ixy1 = detail::permute(ixy + iz1);
+		vec<4, T, Q> ixy00 = detail::permute(ixy0 + iw0);
+		vec<4, T, Q> ixy01 = detail::permute(ixy0 + iw1);
+		vec<4, T, Q> ixy10 = detail::permute(ixy1 + iw0);
+		vec<4, T, Q> ixy11 = detail::permute(ixy1 + iw1);
+
+		vec<4, T, Q> gx00 = ixy00 / T(7);
+		vec<4, T, Q> gy00 = floor(gx00) / T(7);
+		vec<4, T, Q> gz00 = floor(gy00) / T(6);
+		gx00 = fract(gx00) - T(0.5);
+		gy00 = fract(gy00) - T(0.5);
+		gz00 = fract(gz00) - T(0.5);
+		vec<4, T, Q> gw00 = vec<4, T, Q>(0.75) - abs(gx00) - abs(gy00) - abs(gz00);
+		vec<4, T, Q> sw00 = step(gw00, vec<4, T, Q>(0.0));
+		gx00 -= sw00 * (step(T(0), gx00) - T(0.5));
+		gy00 -= sw00 * (step(T(0), gy00) - T(0.5));
+
+		vec<4, T, Q> gx01 = ixy01 / T(7);
+		vec<4, T, Q> gy01 = floor(gx01) / T(7);
+		vec<4, T, Q> gz01 = floor(gy01) / T(6);
+		gx01 = fract(gx01) - T(0.5);
+		gy01 = fract(gy01) - T(0.5);
+		gz01 = fract(gz01) - T(0.5);
+		vec<4, T, Q> gw01 = vec<4, T, Q>(0.75) - abs(gx01) - abs(gy01) - abs(gz01);
+		vec<4, T, Q> sw01 = step(gw01, vec<4, T, Q>(0.0));
+		gx01 -= sw01 * (step(T(0), gx01) - T(0.5));
+		gy01 -= sw01 * (step(T(0), gy01) - T(0.5));
+
+		vec<4, T, Q> gx10 = ixy10 / T(7);
+		vec<4, T, Q> gy10 = floor(gx10) / T(7);
+		vec<4, T, Q> gz10 = floor(gy10) / T(6);
+		gx10 = fract(gx10) - T(0.5);
+		gy10 = fract(gy10) - T(0.5);
+		gz10 = fract(gz10) - T(0.5);
+		vec<4, T, Q> gw10 = vec<4, T, Q>(0.75) - abs(gx10) - abs(gy10) - abs(gz10);
+		vec<4, T, Q> sw10 = step(gw10, vec<4, T, Q>(0));
+		gx10 -= sw10 * (step(T(0), gx10) - T(0.5));
+		gy10 -= sw10 * (step(T(0), gy10) - T(0.5));
+
+		vec<4, T, Q> gx11 = ixy11 / T(7);
+		vec<4, T, Q> gy11 = floor(gx11) / T(7);
+		vec<4, T, Q> gz11 = floor(gy11) / T(6);
+		gx11 = fract(gx11) - T(0.5);
+		gy11 = fract(gy11) - T(0.5);
+		gz11 = fract(gz11) - T(0.5);
+		vec<4, T, Q> gw11 = vec<4, T, Q>(0.75) - abs(gx11) - abs(gy11) - abs(gz11);
+		vec<4, T, Q> sw11 = step(gw11, vec<4, T, Q>(0.0));
+		gx11 -= sw11 * (step(T(0), gx11) - T(0.5));
+		gy11 -= sw11 * (step(T(0), gy11) - T(0.5));
+
+		vec<4, T, Q> g0000(gx00.x, gy00.x, gz00.x, gw00.x);
+		vec<4, T, Q> g1000(gx00.y, gy00.y, gz00.y, gw00.y);
+		vec<4, T, Q> g0100(gx00.z, gy00.z, gz00.z, gw00.z);
+		vec<4, T, Q> g1100(gx00.w, gy00.w, gz00.w, gw00.w);
+		vec<4, T, Q> g0010(gx10.x, gy10.x, gz10.x, gw10.x);
+		vec<4, T, Q> g1010(gx10.y, gy10.y, gz10.y, gw10.y);
+		vec<4, T, Q> g0110(gx10.z, gy10.z, gz10.z, gw10.z);
+		vec<4, T, Q> g1110(gx10.w, gy10.w, gz10.w, gw10.w);
+		vec<4, T, Q> g0001(gx01.x, gy01.x, gz01.x, gw01.x);
+		vec<4, T, Q> g1001(gx01.y, gy01.y, gz01.y, gw01.y);
+		vec<4, T, Q> g0101(gx01.z, gy01.z, gz01.z, gw01.z);
+		vec<4, T, Q> g1101(gx01.w, gy01.w, gz01.w, gw01.w);
+		vec<4, T, Q> g0011(gx11.x, gy11.x, gz11.x, gw11.x);
+		vec<4, T, Q> g1011(gx11.y, gy11.y, gz11.y, gw11.y);
+		vec<4, T, Q> g0111(gx11.z, gy11.z, gz11.z, gw11.z);
+		vec<4, T, Q> g1111(gx11.w, gy11.w, gz11.w, gw11.w);
+
+		vec<4, T, Q> norm00 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0000, g0000), dot(g0100, g0100), dot(g1000, g1000), dot(g1100, g1100)));
+		g0000 *= norm00.x;
+		g0100 *= norm00.y;
+		g1000 *= norm00.z;
+		g1100 *= norm00.w;
+
+		vec<4, T, Q> norm01 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0001, g0001), dot(g0101, g0101), dot(g1001, g1001), dot(g1101, g1101)));
+		g0001 *= norm01.x;
+		g0101 *= norm01.y;
+		g1001 *= norm01.z;
+		g1101 *= norm01.w;
+
+		vec<4, T, Q> norm10 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0010, g0010), dot(g0110, g0110), dot(g1010, g1010), dot(g1110, g1110)));
+		g0010 *= norm10.x;
+		g0110 *= norm10.y;
+		g1010 *= norm10.z;
+		g1110 *= norm10.w;
+
+		vec<4, T, Q> norm11 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0011, g0011), dot(g0111, g0111), dot(g1011, g1011), dot(g1111, g1111)));
+		g0011 *= norm11.x;
+		g0111 *= norm11.y;
+		g1011 *= norm11.z;
+		g1111 *= norm11.w;
+
+		T n0000 = dot(g0000, Pf0);
+		T n1000 = dot(g1000, vec<4, T, Q>(Pf1.x, Pf0.y, Pf0.z, Pf0.w));
+		T n0100 = dot(g0100, vec<4, T, Q>(Pf0.x, Pf1.y, Pf0.z, Pf0.w));
+		T n1100 = dot(g1100, vec<4, T, Q>(Pf1.x, Pf1.y, Pf0.z, Pf0.w));
+		T n0010 = dot(g0010, vec<4, T, Q>(Pf0.x, Pf0.y, Pf1.z, Pf0.w));
+		T n1010 = dot(g1010, vec<4, T, Q>(Pf1.x, Pf0.y, Pf1.z, Pf0.w));
+		T n0110 = dot(g0110, vec<4, T, Q>(Pf0.x, Pf1.y, Pf1.z, Pf0.w));
+		T n1110 = dot(g1110, vec<4, T, Q>(Pf1.x, Pf1.y, Pf1.z, Pf0.w));
+		T n0001 = dot(g0001, vec<4, T, Q>(Pf0.x, Pf0.y, Pf0.z, Pf1.w));
+		T n1001 = dot(g1001, vec<4, T, Q>(Pf1.x, Pf0.y, Pf0.z, Pf1.w));
+		T n0101 = dot(g0101, vec<4, T, Q>(Pf0.x, Pf1.y, Pf0.z, Pf1.w));
+		T n1101 = dot(g1101, vec<4, T, Q>(Pf1.x, Pf1.y, Pf0.z, Pf1.w));
+		T n0011 = dot(g0011, vec<4, T, Q>(Pf0.x, Pf0.y, Pf1.z, Pf1.w));
+		T n1011 = dot(g1011, vec<4, T, Q>(Pf1.x, Pf0.y, Pf1.z, Pf1.w));
+		T n0111 = dot(g0111, vec<4, T, Q>(Pf0.x, Pf1.y, Pf1.z, Pf1.w));
+		T n1111 = dot(g1111, Pf1);
+
+		vec<4, T, Q> fade_xyzw = detail::fade(Pf0);
+		vec<4, T, Q> n_0w = mix(vec<4, T, Q>(n0000, n1000, n0100, n1100), vec<4, T, Q>(n0001, n1001, n0101, n1101), fade_xyzw.w);
+		vec<4, T, Q> n_1w = mix(vec<4, T, Q>(n0010, n1010, n0110, n1110), vec<4, T, Q>(n0011, n1011, n0111, n1111), fade_xyzw.w);
+		vec<4, T, Q> n_zw = mix(n_0w, n_1w, fade_xyzw.z);
+		vec<2, T, Q> n_yzw = mix(vec<2, T, Q>(n_zw.x, n_zw.y), vec<2, T, Q>(n_zw.z, n_zw.w), fade_xyzw.y);
+		T n_xyzw = mix(n_yzw.x, n_yzw.y, fade_xyzw.x);
+		return T(2.2) * n_xyzw;
+	}
+
+	// Classic Perlin noise, periodic variant
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<2, T, Q> const& Position, vec<2, T, Q> const& rep)
+	{
+		vec<4, T, Q> Pi = floor(vec<4, T, Q>(Position.x, Position.y, Position.x, Position.y)) + vec<4, T, Q>(0.0, 0.0, 1.0, 1.0);
+		vec<4, T, Q> Pf = fract(vec<4, T, Q>(Position.x, Position.y, Position.x, Position.y)) - vec<4, T, Q>(0.0, 0.0, 1.0, 1.0);
+		Pi = mod(Pi, vec<4, T, Q>(rep.x, rep.y, rep.x, rep.y)); // To create noise with explicit period
+		Pi = mod(Pi, vec<4, T, Q>(289)); // To avoid truncation effects in permutation
+		vec<4, T, Q> ix(Pi.x, Pi.z, Pi.x, Pi.z);
+		vec<4, T, Q> iy(Pi.y, Pi.y, Pi.w, Pi.w);
+		vec<4, T, Q> fx(Pf.x, Pf.z, Pf.x, Pf.z);
+		vec<4, T, Q> fy(Pf.y, Pf.y, Pf.w, Pf.w);
+
+		vec<4, T, Q> i = detail::permute(detail::permute(ix) + iy);
+
+		vec<4, T, Q> gx = static_cast<T>(2) * fract(i / T(41)) - T(1);
+		vec<4, T, Q> gy = abs(gx) - T(0.5);
+		vec<4, T, Q> tx = floor(gx + T(0.5));
+		gx = gx - tx;
+
+		vec<2, T, Q> g00(gx.x, gy.x);
+		vec<2, T, Q> g10(gx.y, gy.y);
+		vec<2, T, Q> g01(gx.z, gy.z);
+		vec<2, T, Q> g11(gx.w, gy.w);
+
+		vec<4, T, Q> norm = detail::taylorInvSqrt(vec<4, T, Q>(dot(g00, g00), dot(g01, g01), dot(g10, g10), dot(g11, g11)));
+		g00 *= norm.x;
+		g01 *= norm.y;
+		g10 *= norm.z;
+		g11 *= norm.w;
+
+		T n00 = dot(g00, vec<2, T, Q>(fx.x, fy.x));
+		T n10 = dot(g10, vec<2, T, Q>(fx.y, fy.y));
+		T n01 = dot(g01, vec<2, T, Q>(fx.z, fy.z));
+		T n11 = dot(g11, vec<2, T, Q>(fx.w, fy.w));
+
+		vec<2, T, Q> fade_xy = detail::fade(vec<2, T, Q>(Pf.x, Pf.y));
+		vec<2, T, Q> n_x = mix(vec<2, T, Q>(n00, n01), vec<2, T, Q>(n10, n11), fade_xy.x);
+		T n_xy = mix(n_x.x, n_x.y, fade_xy.y);
+		return T(2.3) * n_xy;
+	}
+
+	// Classic Perlin noise, periodic variant
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<3, T, Q> const& Position, vec<3, T, Q> const& rep)
+	{
+		vec<3, T, Q> Pi0 = mod(floor(Position), rep); // Integer part, modulo period
+		vec<3, T, Q> Pi1 = mod(Pi0 + vec<3, T, Q>(T(1)), rep); // Integer part + 1, mod period
+		Pi0 = mod(Pi0, vec<3, T, Q>(289));
+		Pi1 = mod(Pi1, vec<3, T, Q>(289));
+		vec<3, T, Q> Pf0 = fract(Position); // Fractional part for interpolation
+		vec<3, T, Q> Pf1 = Pf0 - vec<3, T, Q>(T(1)); // Fractional part - 1.0
+		vec<4, T, Q> ix = vec<4, T, Q>(Pi0.x, Pi1.x, Pi0.x, Pi1.x);
+		vec<4, T, Q> iy = vec<4, T, Q>(Pi0.y, Pi0.y, Pi1.y, Pi1.y);
+		vec<4, T, Q> iz0(Pi0.z);
+		vec<4, T, Q> iz1(Pi1.z);
+
+		vec<4, T, Q> ixy = detail::permute(detail::permute(ix) + iy);
+		vec<4, T, Q> ixy0 = detail::permute(ixy + iz0);
+		vec<4, T, Q> ixy1 = detail::permute(ixy + iz1);
+
+		vec<4, T, Q> gx0 = ixy0 / T(7);
+		vec<4, T, Q> gy0 = fract(floor(gx0) / T(7)) - T(0.5);
+		gx0 = fract(gx0);
+		vec<4, T, Q> gz0 = vec<4, T, Q>(0.5) - abs(gx0) - abs(gy0);
+		vec<4, T, Q> sz0 = step(gz0, vec<4, T, Q>(0));
+		gx0 -= sz0 * (step(T(0), gx0) - T(0.5));
+		gy0 -= sz0 * (step(T(0), gy0) - T(0.5));
+
+		vec<4, T, Q> gx1 = ixy1 / T(7);
+		vec<4, T, Q> gy1 = fract(floor(gx1) / T(7)) - T(0.5);
+		gx1 = fract(gx1);
+		vec<4, T, Q> gz1 = vec<4, T, Q>(0.5) - abs(gx1) - abs(gy1);
+		vec<4, T, Q> sz1 = step(gz1, vec<4, T, Q>(T(0)));
+		gx1 -= sz1 * (step(T(0), gx1) - T(0.5));
+		gy1 -= sz1 * (step(T(0), gy1) - T(0.5));
+
+		vec<3, T, Q> g000 = vec<3, T, Q>(gx0.x, gy0.x, gz0.x);
+		vec<3, T, Q> g100 = vec<3, T, Q>(gx0.y, gy0.y, gz0.y);
+		vec<3, T, Q> g010 = vec<3, T, Q>(gx0.z, gy0.z, gz0.z);
+		vec<3, T, Q> g110 = vec<3, T, Q>(gx0.w, gy0.w, gz0.w);
+		vec<3, T, Q> g001 = vec<3, T, Q>(gx1.x, gy1.x, gz1.x);
+		vec<3, T, Q> g101 = vec<3, T, Q>(gx1.y, gy1.y, gz1.y);
+		vec<3, T, Q> g011 = vec<3, T, Q>(gx1.z, gy1.z, gz1.z);
+		vec<3, T, Q> g111 = vec<3, T, Q>(gx1.w, gy1.w, gz1.w);
+
+		vec<4, T, Q> norm0 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g000, g000), dot(g010, g010), dot(g100, g100), dot(g110, g110)));
+		g000 *= norm0.x;
+		g010 *= norm0.y;
+		g100 *= norm0.z;
+		g110 *= norm0.w;
+		vec<4, T, Q> norm1 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g001, g001), dot(g011, g011), dot(g101, g101), dot(g111, g111)));
+		g001 *= norm1.x;
+		g011 *= norm1.y;
+		g101 *= norm1.z;
+		g111 *= norm1.w;
+
+		T n000 = dot(g000, Pf0);
+		T n100 = dot(g100, vec<3, T, Q>(Pf1.x, Pf0.y, Pf0.z));
+		T n010 = dot(g010, vec<3, T, Q>(Pf0.x, Pf1.y, Pf0.z));
+		T n110 = dot(g110, vec<3, T, Q>(Pf1.x, Pf1.y, Pf0.z));
+		T n001 = dot(g001, vec<3, T, Q>(Pf0.x, Pf0.y, Pf1.z));
+		T n101 = dot(g101, vec<3, T, Q>(Pf1.x, Pf0.y, Pf1.z));
+		T n011 = dot(g011, vec<3, T, Q>(Pf0.x, Pf1.y, Pf1.z));
+		T n111 = dot(g111, Pf1);
+
+		vec<3, T, Q> fade_xyz = detail::fade(Pf0);
+		vec<4, T, Q> n_z = mix(vec<4, T, Q>(n000, n100, n010, n110), vec<4, T, Q>(n001, n101, n011, n111), fade_xyz.z);
+		vec<2, T, Q> n_yz = mix(vec<2, T, Q>(n_z.x, n_z.y), vec<2, T, Q>(n_z.z, n_z.w), fade_xyz.y);
+		T n_xyz = mix(n_yz.x, n_yz.y, fade_xyz.x);
+		return T(2.2) * n_xyz;
+	}
+
+	// Classic Perlin noise, periodic version
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T perlin(vec<4, T, Q> const& Position, vec<4, T, Q> const& rep)
+	{
+		vec<4, T, Q> Pi0 = mod(floor(Position), rep); // Integer part modulo rep
+		vec<4, T, Q> Pi1 = mod(Pi0 + T(1), rep); // Integer part + 1 mod rep
+		vec<4, T, Q> Pf0 = fract(Position); // Fractional part for interpolation
+		vec<4, T, Q> Pf1 = Pf0 - T(1); // Fractional part - 1.0
+		vec<4, T, Q> ix = vec<4, T, Q>(Pi0.x, Pi1.x, Pi0.x, Pi1.x);
+		vec<4, T, Q> iy = vec<4, T, Q>(Pi0.y, Pi0.y, Pi1.y, Pi1.y);
+		vec<4, T, Q> iz0(Pi0.z);
+		vec<4, T, Q> iz1(Pi1.z);
+		vec<4, T, Q> iw0(Pi0.w);
+		vec<4, T, Q> iw1(Pi1.w);
+
+		vec<4, T, Q> ixy = detail::permute(detail::permute(ix) + iy);
+		vec<4, T, Q> ixy0 = detail::permute(ixy + iz0);
+		vec<4, T, Q> ixy1 = detail::permute(ixy + iz1);
+		vec<4, T, Q> ixy00 = detail::permute(ixy0 + iw0);
+		vec<4, T, Q> ixy01 = detail::permute(ixy0 + iw1);
+		vec<4, T, Q> ixy10 = detail::permute(ixy1 + iw0);
+		vec<4, T, Q> ixy11 = detail::permute(ixy1 + iw1);
+
+		vec<4, T, Q> gx00 = ixy00 / T(7);
+		vec<4, T, Q> gy00 = floor(gx00) / T(7);
+		vec<4, T, Q> gz00 = floor(gy00) / T(6);
+		gx00 = fract(gx00) - T(0.5);
+		gy00 = fract(gy00) - T(0.5);
+		gz00 = fract(gz00) - T(0.5);
+		vec<4, T, Q> gw00 = vec<4, T, Q>(0.75) - abs(gx00) - abs(gy00) - abs(gz00);
+		vec<4, T, Q> sw00 = step(gw00, vec<4, T, Q>(0));
+		gx00 -= sw00 * (step(T(0), gx00) - T(0.5));
+		gy00 -= sw00 * (step(T(0), gy00) - T(0.5));
+
+		vec<4, T, Q> gx01 = ixy01 / T(7);
+		vec<4, T, Q> gy01 = floor(gx01) / T(7);
+		vec<4, T, Q> gz01 = floor(gy01) / T(6);
+		gx01 = fract(gx01) - T(0.5);
+		gy01 = fract(gy01) - T(0.5);
+		gz01 = fract(gz01) - T(0.5);
+		vec<4, T, Q> gw01 = vec<4, T, Q>(0.75) - abs(gx01) - abs(gy01) - abs(gz01);
+		vec<4, T, Q> sw01 = step(gw01, vec<4, T, Q>(0.0));
+		gx01 -= sw01 * (step(T(0), gx01) - T(0.5));
+		gy01 -= sw01 * (step(T(0), gy01) - T(0.5));
+
+		vec<4, T, Q> gx10 = ixy10 / T(7);
+		vec<4, T, Q> gy10 = floor(gx10) / T(7);
+		vec<4, T, Q> gz10 = floor(gy10) / T(6);
+		gx10 = fract(gx10) - T(0.5);
+		gy10 = fract(gy10) - T(0.5);
+		gz10 = fract(gz10) - T(0.5);
+		vec<4, T, Q> gw10 = vec<4, T, Q>(0.75) - abs(gx10) - abs(gy10) - abs(gz10);
+		vec<4, T, Q> sw10 = step(gw10, vec<4, T, Q>(0.0));
+		gx10 -= sw10 * (step(T(0), gx10) - T(0.5));
+		gy10 -= sw10 * (step(T(0), gy10) - T(0.5));
+
+		vec<4, T, Q> gx11 = ixy11 / T(7);
+		vec<4, T, Q> gy11 = floor(gx11) / T(7);
+		vec<4, T, Q> gz11 = floor(gy11) / T(6);
+		gx11 = fract(gx11) - T(0.5);
+		gy11 = fract(gy11) - T(0.5);
+		gz11 = fract(gz11) - T(0.5);
+		vec<4, T, Q> gw11 = vec<4, T, Q>(0.75) - abs(gx11) - abs(gy11) - abs(gz11);
+		vec<4, T, Q> sw11 = step(gw11, vec<4, T, Q>(T(0)));
+		gx11 -= sw11 * (step(T(0), gx11) - T(0.5));
+		gy11 -= sw11 * (step(T(0), gy11) - T(0.5));
+
+		vec<4, T, Q> g0000(gx00.x, gy00.x, gz00.x, gw00.x);
+		vec<4, T, Q> g1000(gx00.y, gy00.y, gz00.y, gw00.y);
+		vec<4, T, Q> g0100(gx00.z, gy00.z, gz00.z, gw00.z);
+		vec<4, T, Q> g1100(gx00.w, gy00.w, gz00.w, gw00.w);
+		vec<4, T, Q> g0010(gx10.x, gy10.x, gz10.x, gw10.x);
+		vec<4, T, Q> g1010(gx10.y, gy10.y, gz10.y, gw10.y);
+		vec<4, T, Q> g0110(gx10.z, gy10.z, gz10.z, gw10.z);
+		vec<4, T, Q> g1110(gx10.w, gy10.w, gz10.w, gw10.w);
+		vec<4, T, Q> g0001(gx01.x, gy01.x, gz01.x, gw01.x);
+		vec<4, T, Q> g1001(gx01.y, gy01.y, gz01.y, gw01.y);
+		vec<4, T, Q> g0101(gx01.z, gy01.z, gz01.z, gw01.z);
+		vec<4, T, Q> g1101(gx01.w, gy01.w, gz01.w, gw01.w);
+		vec<4, T, Q> g0011(gx11.x, gy11.x, gz11.x, gw11.x);
+		vec<4, T, Q> g1011(gx11.y, gy11.y, gz11.y, gw11.y);
+		vec<4, T, Q> g0111(gx11.z, gy11.z, gz11.z, gw11.z);
+		vec<4, T, Q> g1111(gx11.w, gy11.w, gz11.w, gw11.w);
+
+		vec<4, T, Q> norm00 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0000, g0000), dot(g0100, g0100), dot(g1000, g1000), dot(g1100, g1100)));
+		g0000 *= norm00.x;
+		g0100 *= norm00.y;
+		g1000 *= norm00.z;
+		g1100 *= norm00.w;
+
+		vec<4, T, Q> norm01 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0001, g0001), dot(g0101, g0101), dot(g1001, g1001), dot(g1101, g1101)));
+		g0001 *= norm01.x;
+		g0101 *= norm01.y;
+		g1001 *= norm01.z;
+		g1101 *= norm01.w;
+
+		vec<4, T, Q> norm10 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0010, g0010), dot(g0110, g0110), dot(g1010, g1010), dot(g1110, g1110)));
+		g0010 *= norm10.x;
+		g0110 *= norm10.y;
+		g1010 *= norm10.z;
+		g1110 *= norm10.w;
+
+		vec<4, T, Q> norm11 = detail::taylorInvSqrt(vec<4, T, Q>(dot(g0011, g0011), dot(g0111, g0111), dot(g1011, g1011), dot(g1111, g1111)));
+		g0011 *= norm11.x;
+		g0111 *= norm11.y;
+		g1011 *= norm11.z;
+		g1111 *= norm11.w;
+
+		T n0000 = dot(g0000, Pf0);
+		T n1000 = dot(g1000, vec<4, T, Q>(Pf1.x, Pf0.y, Pf0.z, Pf0.w));
+		T n0100 = dot(g0100, vec<4, T, Q>(Pf0.x, Pf1.y, Pf0.z, Pf0.w));
+		T n1100 = dot(g1100, vec<4, T, Q>(Pf1.x, Pf1.y, Pf0.z, Pf0.w));
+		T n0010 = dot(g0010, vec<4, T, Q>(Pf0.x, Pf0.y, Pf1.z, Pf0.w));
+		T n1010 = dot(g1010, vec<4, T, Q>(Pf1.x, Pf0.y, Pf1.z, Pf0.w));
+		T n0110 = dot(g0110, vec<4, T, Q>(Pf0.x, Pf1.y, Pf1.z, Pf0.w));
+		T n1110 = dot(g1110, vec<4, T, Q>(Pf1.x, Pf1.y, Pf1.z, Pf0.w));
+		T n0001 = dot(g0001, vec<4, T, Q>(Pf0.x, Pf0.y, Pf0.z, Pf1.w));
+		T n1001 = dot(g1001, vec<4, T, Q>(Pf1.x, Pf0.y, Pf0.z, Pf1.w));
+		T n0101 = dot(g0101, vec<4, T, Q>(Pf0.x, Pf1.y, Pf0.z, Pf1.w));
+		T n1101 = dot(g1101, vec<4, T, Q>(Pf1.x, Pf1.y, Pf0.z, Pf1.w));
+		T n0011 = dot(g0011, vec<4, T, Q>(Pf0.x, Pf0.y, Pf1.z, Pf1.w));
+		T n1011 = dot(g1011, vec<4, T, Q>(Pf1.x, Pf0.y, Pf1.z, Pf1.w));
+		T n0111 = dot(g0111, vec<4, T, Q>(Pf0.x, Pf1.y, Pf1.z, Pf1.w));
+		T n1111 = dot(g1111, Pf1);
+
+		vec<4, T, Q> fade_xyzw = detail::fade(Pf0);
+		vec<4, T, Q> n_0w = mix(vec<4, T, Q>(n0000, n1000, n0100, n1100), vec<4, T, Q>(n0001, n1001, n0101, n1101), fade_xyzw.w);
+		vec<4, T, Q> n_1w = mix(vec<4, T, Q>(n0010, n1010, n0110, n1110), vec<4, T, Q>(n0011, n1011, n0111, n1111), fade_xyzw.w);
+		vec<4, T, Q> n_zw = mix(n_0w, n_1w, fade_xyzw.z);
+		vec<2, T, Q> n_yzw = mix(vec<2, T, Q>(n_zw.x, n_zw.y), vec<2, T, Q>(n_zw.z, n_zw.w), fade_xyzw.y);
+		T n_xyzw = mix(n_yzw.x, n_yzw.y, fade_xyzw.x);
+		return T(2.2) * n_xyzw;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T simplex(glm::vec<2, T, Q> const& v)
+	{
+		vec<4, T, Q> const C = vec<4, T, Q>(
+			T( 0.211324865405187),  // (3.0 -  sqrt(3.0)) / 6.0
+			T( 0.366025403784439),  //  0.5 * (sqrt(3.0)  - 1.0)
+			T(-0.577350269189626),	// -1.0 + 2.0 * C.x
+			T( 0.024390243902439)); //  1.0 / 41.0
+
+		// First corner
+		vec<2, T, Q> i  = floor(v + dot(v, vec<2, T, Q>(C[1])));
+		vec<2, T, Q> x0 = v -   i + dot(i, vec<2, T, Q>(C[0]));
+
+		// Other corners
+		//i1.x = step( x0.y, x0.x ); // x0.x > x0.y ? 1.0 : 0.0
+		//i1.y = 1.0 - i1.x;
+		vec<2, T, Q> i1 = (x0.x > x0.y) ? vec<2, T, Q>(1, 0) : vec<2, T, Q>(0, 1);
+		// x0 = x0 - 0.0 + 0.0 * C.xx ;
+		// x1 = x0 - i1 + 1.0 * C.xx ;
+		// x2 = x0 - 1.0 + 2.0 * C.xx ;
+		vec<4, T, Q> x12 = vec<4, T, Q>(x0.x, x0.y, x0.x, x0.y) + vec<4, T, Q>(C.x, C.x, C.z, C.z);
+		x12 = vec<4, T, Q>(vec<2, T, Q>(x12) - i1, x12.z, x12.w);
+
+		// Permutations
+		i = mod(i, vec<2, T, Q>(289)); // Avoid truncation effects in permutation
+		vec<3, T, Q> p = detail::permute(
+			detail::permute(i.y + vec<3, T, Q>(T(0), i1.y, T(1)))
+			+ i.x + vec<3, T, Q>(T(0), i1.x, T(1)));
+
+		vec<3, T, Q> m = max(vec<3, T, Q>(0.5) - vec<3, T, Q>(
+			dot(x0, x0),
+			dot(vec<2, T, Q>(x12.x, x12.y), vec<2, T, Q>(x12.x, x12.y)),
+			dot(vec<2, T, Q>(x12.z, x12.w), vec<2, T, Q>(x12.z, x12.w))), vec<3, T, Q>(0));
+		m = m * m ;
+		m = m * m ;
+
+		// Gradients: 41 points uniformly over a line, mapped onto a diamond.
+		// The ring size 17*17 = 289 is close to a multiple of 41 (41*7 = 287)
+
+		vec<3, T, Q> x = static_cast<T>(2) * fract(p * C.w) - T(1);
+		vec<3, T, Q> h = abs(x) - T(0.5);
+		vec<3, T, Q> ox = floor(x + T(0.5));
+		vec<3, T, Q> a0 = x - ox;
+
+		// Normalise gradients implicitly by scaling m
+		// Inlined for speed: m *= taylorInvSqrt( a0*a0 + h*h );
+		m *= static_cast<T>(1.79284291400159) - T(0.85373472095314) * (a0 * a0 + h * h);
+
+		// Compute final noise value at P
+		vec<3, T, Q> g;
+		g.x  = a0.x  * x0.x  + h.x  * x0.y;
+		//g.yz = a0.yz * x12.xz + h.yz * x12.yw;
+		g.y = a0.y * x12.x + h.y * x12.y;
+		g.z = a0.z * x12.z + h.z * x12.w;
+		return T(130) * dot(m, g);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T simplex(vec<3, T, Q> const& v)
+	{
+		vec<2, T, Q> const C(1.0 / 6.0, 1.0 / 3.0);
+		vec<4, T, Q> const D(0.0, 0.5, 1.0, 2.0);
+
+		// First corner
+		vec<3, T, Q> i(floor(v + dot(v, vec<3, T, Q>(C.y))));
+		vec<3, T, Q> x0(v - i + dot(i, vec<3, T, Q>(C.x)));
+
+		// Other corners
+		vec<3, T, Q> g(step(vec<3, T, Q>(x0.y, x0.z, x0.x), x0));
+		vec<3, T, Q> l(T(1) - g);
+		vec<3, T, Q> i1(min(g, vec<3, T, Q>(l.z, l.x, l.y)));
+		vec<3, T, Q> i2(max(g, vec<3, T, Q>(l.z, l.x, l.y)));
+
+		//   x0 = x0 - 0.0 + 0.0 * C.xxx;
+		//   x1 = x0 - i1  + 1.0 * C.xxx;
+		//   x2 = x0 - i2  + 2.0 * C.xxx;
+		//   x3 = x0 - 1.0 + 3.0 * C.xxx;
+		vec<3, T, Q> x1(x0 - i1 + C.x);
+		vec<3, T, Q> x2(x0 - i2 + C.y); // 2.0*C.x = 1/3 = C.y
+		vec<3, T, Q> x3(x0 - D.y);      // -1.0+3.0*C.x = -0.5 = -D.y
+
+		// Permutations
+		i = detail::mod289(i);
+		vec<4, T, Q> p(detail::permute(detail::permute(detail::permute(
+			i.z + vec<4, T, Q>(T(0), i1.z, i2.z, T(1))) +
+			i.y + vec<4, T, Q>(T(0), i1.y, i2.y, T(1))) +
+			i.x + vec<4, T, Q>(T(0), i1.x, i2.x, T(1))));
+
+		// Gradients: 7x7 points over a square, mapped onto an octahedron.
+		// The ring size 17*17 = 289 is close to a multiple of 49 (49*6 = 294)
+		T n_ = static_cast<T>(0.142857142857); // 1.0/7.0
+		vec<3, T, Q> ns(n_ * vec<3, T, Q>(D.w, D.y, D.z) - vec<3, T, Q>(D.x, D.z, D.x));
+
+		vec<4, T, Q> j(p - T(49) * floor(p * ns.z * ns.z));  //  mod(p,7*7)
+
+		vec<4, T, Q> x_(floor(j * ns.z));
+		vec<4, T, Q> y_(floor(j - T(7) * x_));    // mod(j,N)
+
+		vec<4, T, Q> x(x_ * ns.x + ns.y);
+		vec<4, T, Q> y(y_ * ns.x + ns.y);
+		vec<4, T, Q> h(T(1) - abs(x) - abs(y));
+
+		vec<4, T, Q> b0(x.x, x.y, y.x, y.y);
+		vec<4, T, Q> b1(x.z, x.w, y.z, y.w);
+
+		// vec4 s0 = vec4(lessThan(b0,0.0))*2.0 - 1.0;
+		// vec4 s1 = vec4(lessThan(b1,0.0))*2.0 - 1.0;
+		vec<4, T, Q> s0(floor(b0) * T(2) + T(1));
+		vec<4, T, Q> s1(floor(b1) * T(2) + T(1));
+		vec<4, T, Q> sh(-step(h, vec<4, T, Q>(0.0)));
+
+		vec<4, T, Q> a0 = vec<4, T, Q>(b0.x, b0.z, b0.y, b0.w) + vec<4, T, Q>(s0.x, s0.z, s0.y, s0.w) * vec<4, T, Q>(sh.x, sh.x, sh.y, sh.y);
+		vec<4, T, Q> a1 = vec<4, T, Q>(b1.x, b1.z, b1.y, b1.w) + vec<4, T, Q>(s1.x, s1.z, s1.y, s1.w) * vec<4, T, Q>(sh.z, sh.z, sh.w, sh.w);
+
+		vec<3, T, Q> p0(a0.x, a0.y, h.x);
+		vec<3, T, Q> p1(a0.z, a0.w, h.y);
+		vec<3, T, Q> p2(a1.x, a1.y, h.z);
+		vec<3, T, Q> p3(a1.z, a1.w, h.w);
+
+		// Normalise gradients
+		vec<4, T, Q> norm = detail::taylorInvSqrt(vec<4, T, Q>(dot(p0, p0), dot(p1, p1), dot(p2, p2), dot(p3, p3)));
+		p0 *= norm.x;
+		p1 *= norm.y;
+		p2 *= norm.z;
+		p3 *= norm.w;
+
+		// Mix final noise value
+		vec<4, T, Q> m = max(T(0.6) - vec<4, T, Q>(dot(x0, x0), dot(x1, x1), dot(x2, x2), dot(x3, x3)), vec<4, T, Q>(0));
+		m = m * m;
+		return T(42) * dot(m * m, vec<4, T, Q>(dot(p0, x0), dot(p1, x1), dot(p2, x2), dot(p3, x3)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T simplex(vec<4, T, Q> const& v)
+	{
+		vec<4, T, Q> const C(
+			0.138196601125011,  // (5 - sqrt(5))/20  G4
+			0.276393202250021,  // 2 * G4
+			0.414589803375032,  // 3 * G4
+			-0.447213595499958); // -1 + 4 * G4
+
+		// (sqrt(5) - 1)/4 = F4, used once below
+		T const F4 = static_cast<T>(0.309016994374947451);
+
+		// First corner
+		vec<4, T, Q> i  = floor(v + dot(v, vec<4, T, Q>(F4)));
+		vec<4, T, Q> x0 = v -   i + dot(i, vec<4, T, Q>(C.x));
+
+		// Other corners
+
+		// Rank sorting originally contributed by Bill Licea-Kane, AMD (formerly ATI)
+		vec<4, T, Q> i0;
+		vec<3, T, Q> isX = step(vec<3, T, Q>(x0.y, x0.z, x0.w), vec<3, T, Q>(x0.x));
+		vec<3, T, Q> isYZ = step(vec<3, T, Q>(x0.z, x0.w, x0.w), vec<3, T, Q>(x0.y, x0.y, x0.z));
+		//  i0.x = dot(isX, vec3(1.0));
+		//i0.x = isX.x + isX.y + isX.z;
+		//i0.yzw = static_cast<T>(1) - isX;
+		i0 = vec<4, T, Q>(isX.x + isX.y + isX.z, T(1) - isX);
+		//  i0.y += dot(isYZ.xy, vec2(1.0));
+		i0.y += isYZ.x + isYZ.y;
+		//i0.zw += 1.0 - vec<2, T, Q>(isYZ.x, isYZ.y);
+		i0.z += static_cast<T>(1) - isYZ.x;
+		i0.w += static_cast<T>(1) - isYZ.y;
+		i0.z += isYZ.z;
+		i0.w += static_cast<T>(1) - isYZ.z;
+
+		// i0 now contains the unique values 0,1,2,3 in each channel
+		vec<4, T, Q> i3 = clamp(i0, T(0), T(1));
+		vec<4, T, Q> i2 = clamp(i0 - T(1), T(0), T(1));
+		vec<4, T, Q> i1 = clamp(i0 - T(2), T(0), T(1));
+
+		//  x0 = x0 - 0.0 + 0.0 * C.xxxx
+		//  x1 = x0 - i1  + 0.0 * C.xxxx
+		//  x2 = x0 - i2  + 0.0 * C.xxxx
+		//  x3 = x0 - i3  + 0.0 * C.xxxx
+		//  x4 = x0 - 1.0 + 4.0 * C.xxxx
+		vec<4, T, Q> x1 = x0 - i1 + C.x;
+		vec<4, T, Q> x2 = x0 - i2 + C.y;
+		vec<4, T, Q> x3 = x0 - i3 + C.z;
+		vec<4, T, Q> x4 = x0 + C.w;
+
+		// Permutations
+		i = mod(i, vec<4, T, Q>(289));
+		T j0 = detail::permute(detail::permute(detail::permute(detail::permute(i.w) + i.z) + i.y) + i.x);
+		vec<4, T, Q> j1 = detail::permute(detail::permute(detail::permute(detail::permute(
+			i.w + vec<4, T, Q>(i1.w, i2.w, i3.w, T(1))) +
+			i.z + vec<4, T, Q>(i1.z, i2.z, i3.z, T(1))) +
+			i.y + vec<4, T, Q>(i1.y, i2.y, i3.y, T(1))) +
+			i.x + vec<4, T, Q>(i1.x, i2.x, i3.x, T(1)));
+
+		// Gradients: 7x7x6 points over a cube, mapped onto a 4-cross polytope
+		// 7*7*6 = 294, which is close to the ring size 17*17 = 289.
+		vec<4, T, Q> ip = vec<4, T, Q>(T(1) / T(294), T(1) / T(49), T(1) / T(7), T(0));
+
+		vec<4, T, Q> p0 = detail::grad4(j0,   ip);
+		vec<4, T, Q> p1 = detail::grad4(j1.x, ip);
+		vec<4, T, Q> p2 = detail::grad4(j1.y, ip);
+		vec<4, T, Q> p3 = detail::grad4(j1.z, ip);
+		vec<4, T, Q> p4 = detail::grad4(j1.w, ip);
+
+		// Normalise gradients
+		vec<4, T, Q> norm = detail::taylorInvSqrt(vec<4, T, Q>(dot(p0, p0), dot(p1, p1), dot(p2, p2), dot(p3, p3)));
+		p0 *= norm.x;
+		p1 *= norm.y;
+		p2 *= norm.z;
+		p3 *= norm.w;
+		p4 *= detail::taylorInvSqrt(dot(p4, p4));
+
+		// Mix contributions from the five corners
+		vec<3, T, Q> m0 = max(T(0.6) - vec<3, T, Q>(dot(x0, x0), dot(x1, x1), dot(x2, x2)), vec<3, T, Q>(0));
+		vec<2, T, Q> m1 = max(T(0.6) - vec<2, T, Q>(dot(x3, x3), dot(x4, x4)             ), vec<2, T, Q>(0));
+		m0 = m0 * m0;
+		m1 = m1 * m1;
+		return T(49) *
+			(dot(m0 * m0, vec<3, T, Q>(dot(p0, x0), dot(p1, x1), dot(p2, x2))) +
+			dot(m1 * m1, vec<2, T, Q>(dot(p3, x3), dot(p4, x4))));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/packing.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/packing.hpp
new file mode 100644
index 000000000000..8e416b3fe1b5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/packing.hpp
@@ -0,0 +1,728 @@
+/// @ref gtc_packing
+/// @file glm/gtc/packing.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_packing GLM_GTC_packing
+/// @ingroup gtc
+///
+/// Include <glm/gtc/packing.hpp> to use the features of this extension.
+///
+/// This extension provides a set of function to convert vertors to packed
+/// formats.
+
+#pragma once
+
+// Dependency:
+#include "type_precision.hpp"
+#include "../ext/vector_packing.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_packing extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_packing
+	/// @{
+
+	/// First, converts the normalized floating-point value v into a 8-bit integer value.
+	/// Then, the results are packed into the returned 8-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm1x8:	round(clamp(c, 0, +1) * 255.0)
+	///
+	/// @see gtc_packing
+	/// @see uint16 packUnorm2x8(vec2 const& v)
+	/// @see uint32 packUnorm4x8(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packUnorm4x8.xml">GLSL packUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint8 packUnorm1x8(float v);
+
+	/// Convert a single 8-bit integer to a normalized floating-point value.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackUnorm4x8: f / 255.0
+	///
+	/// @see gtc_packing
+	/// @see vec2 unpackUnorm2x8(uint16 p)
+	/// @see vec4 unpackUnorm4x8(uint32 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackUnorm4x8.xml">GLSL unpackUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL float unpackUnorm1x8(uint8 p);
+
+	/// First, converts each component of the normalized floating-point value v into 8-bit integer values.
+	/// Then, the results are packed into the returned 16-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm2x8:	round(clamp(c, 0, +1) * 255.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint8 packUnorm1x8(float const& v)
+	/// @see uint32 packUnorm4x8(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packUnorm4x8.xml">GLSL packUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint16 packUnorm2x8(vec2 const& v);
+
+	/// First, unpacks a single 16-bit unsigned integer p into a pair of 8-bit unsigned integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned two-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackUnorm4x8: f / 255.0
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see float unpackUnorm1x8(uint8 v)
+	/// @see vec4 unpackUnorm4x8(uint32 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackUnorm4x8.xml">GLSL unpackUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec2 unpackUnorm2x8(uint16 p);
+
+	/// First, converts the normalized floating-point value v into 8-bit integer value.
+	/// Then, the results are packed into the returned 8-bit unsigned integer.
+	///
+	/// The conversion to fixed point is done as follows:
+	/// packSnorm1x8:	round(clamp(s, -1, +1) * 127.0)
+	///
+	/// @see gtc_packing
+	/// @see uint16 packSnorm2x8(vec2 const& v)
+	/// @see uint32 packSnorm4x8(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packSnorm4x8.xml">GLSL packSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint8 packSnorm1x8(float s);
+
+	/// First, unpacks a single 8-bit unsigned integer p into a single 8-bit signed integers.
+	/// Then, the value is converted to a normalized floating-point value to generate the returned scalar.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm1x8: clamp(f / 127.0, -1, +1)
+	///
+	/// @see gtc_packing
+	/// @see vec2 unpackSnorm2x8(uint16 p)
+	/// @see vec4 unpackSnorm4x8(uint32 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackSnorm4x8.xml">GLSL unpackSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL float unpackSnorm1x8(uint8 p);
+
+	/// First, converts each component of the normalized floating-point value v into 8-bit integer values.
+	/// Then, the results are packed into the returned 16-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packSnorm2x8:	round(clamp(c, -1, +1) * 127.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint8 packSnorm1x8(float const& v)
+	/// @see uint32 packSnorm4x8(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packSnorm4x8.xml">GLSL packSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint16 packSnorm2x8(vec2 const& v);
+
+	/// First, unpacks a single 16-bit unsigned integer p into a pair of 8-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned two-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm2x8: clamp(f / 127.0, -1, +1)
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see float unpackSnorm1x8(uint8 p)
+	/// @see vec4 unpackSnorm4x8(uint32 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackSnorm4x8.xml">GLSL unpackSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec2 unpackSnorm2x8(uint16 p);
+
+	/// First, converts the normalized floating-point value v into a 16-bit integer value.
+	/// Then, the results are packed into the returned 16-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm1x16:	round(clamp(c, 0, +1) * 65535.0)
+	///
+	/// @see gtc_packing
+	/// @see uint16 packSnorm1x16(float const& v)
+	/// @see uint64 packSnorm4x16(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packUnorm4x8.xml">GLSL packUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint16 packUnorm1x16(float v);
+
+	/// First, unpacks a single 16-bit unsigned integer p into a of 16-bit unsigned integers.
+	/// Then, the value is converted to a normalized floating-point value to generate the returned scalar.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackUnorm1x16: f / 65535.0
+	///
+	/// @see gtc_packing
+	/// @see vec2 unpackUnorm2x16(uint32 p)
+	/// @see vec4 unpackUnorm4x16(uint64 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackUnorm2x16.xml">GLSL unpackUnorm2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL float unpackUnorm1x16(uint16 p);
+
+	/// First, converts each component of the normalized floating-point value v into 16-bit integer values.
+	/// Then, the results are packed into the returned 64-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm4x16:	round(clamp(c, 0, +1) * 65535.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packUnorm1x16(float const& v)
+	/// @see uint32 packUnorm2x16(vec2 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packUnorm4x8.xml">GLSL packUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint64 packUnorm4x16(vec4 const& v);
+
+	/// First, unpacks a single 64-bit unsigned integer p into four 16-bit unsigned integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackUnormx4x16: f / 65535.0
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see float unpackUnorm1x16(uint16 p)
+	/// @see vec2 unpackUnorm2x16(uint32 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackUnorm2x16.xml">GLSL unpackUnorm2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec4 unpackUnorm4x16(uint64 p);
+
+	/// First, converts the normalized floating-point value v into 16-bit integer value.
+	/// Then, the results are packed into the returned 16-bit unsigned integer.
+	///
+	/// The conversion to fixed point is done as follows:
+	/// packSnorm1x8:	round(clamp(s, -1, +1) * 32767.0)
+	///
+	/// @see gtc_packing
+	/// @see uint32 packSnorm2x16(vec2 const& v)
+	/// @see uint64 packSnorm4x16(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packSnorm4x8.xml">GLSL packSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint16 packSnorm1x16(float v);
+
+	/// First, unpacks a single 16-bit unsigned integer p into a single 16-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned scalar.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm1x16: clamp(f / 32767.0, -1, +1)
+	///
+	/// @see gtc_packing
+	/// @see vec2 unpackSnorm2x16(uint32 p)
+	/// @see vec4 unpackSnorm4x16(uint64 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackSnorm1x16.xml">GLSL unpackSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL float unpackSnorm1x16(uint16 p);
+
+	/// First, converts each component of the normalized floating-point value v into 16-bit integer values.
+	/// Then, the results are packed into the returned 64-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packSnorm2x8:	round(clamp(c, -1, +1) * 32767.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packSnorm1x16(float const& v)
+	/// @see uint32 packSnorm2x16(vec2 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packSnorm4x8.xml">GLSL packSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint64 packSnorm4x16(vec4 const& v);
+
+	/// First, unpacks a single 64-bit unsigned integer p into four 16-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm4x16: clamp(f / 32767.0, -1, +1)
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see float unpackSnorm1x16(uint16 p)
+	/// @see vec2 unpackSnorm2x16(uint32 p)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackSnorm2x16.xml">GLSL unpackSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec4 unpackSnorm4x16(uint64 p);
+
+	/// Returns an unsigned integer obtained by converting the components of a floating-point scalar
+	/// to the 16-bit floating-point representation found in the OpenGL Specification,
+	/// and then packing this 16-bit value into a 16-bit unsigned integer.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packHalf2x16(vec2 const& v)
+	/// @see uint64 packHalf4x16(vec4 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packHalf2x16.xml">GLSL packHalf2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint16 packHalf1x16(float v);
+
+	/// Returns a floating-point scalar with components obtained by unpacking a 16-bit unsigned integer into a 16-bit value,
+	/// interpreted as a 16-bit floating-point number according to the OpenGL Specification,
+	/// and converting it to 32-bit floating-point values.
+	///
+	/// @see gtc_packing
+	/// @see vec2 unpackHalf2x16(uint32 const& v)
+	/// @see vec4 unpackHalf4x16(uint64 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackHalf2x16.xml">GLSL unpackHalf2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL float unpackHalf1x16(uint16 v);
+
+	/// Returns an unsigned integer obtained by converting the components of a four-component floating-point vector
+	/// to the 16-bit floating-point representation found in the OpenGL Specification,
+	/// and then packing these four 16-bit values into a 64-bit unsigned integer.
+	/// The first vector component specifies the 16 least-significant bits of the result;
+	/// the forth component specifies the 16 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packHalf1x16(float const& v)
+	/// @see uint32 packHalf2x16(vec2 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packHalf2x16.xml">GLSL packHalf2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint64 packHalf4x16(vec4 const& v);
+
+	/// Returns a four-component floating-point vector with components obtained by unpacking a 64-bit unsigned integer into four 16-bit values,
+	/// interpreting those values as 16-bit floating-point numbers according to the OpenGL Specification,
+	/// and converting them to 32-bit floating-point values.
+	/// The first component of the vector is obtained from the 16 least-significant bits of v;
+	/// the forth component is obtained from the 16 most-significant bits of v.
+	///
+	/// @see gtc_packing
+	/// @see float unpackHalf1x16(uint16 const& v)
+	/// @see vec2 unpackHalf2x16(uint32 const& v)
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackHalf2x16.xml">GLSL unpackHalf2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec4 unpackHalf4x16(uint64 p);
+
+	/// Returns an unsigned integer obtained by converting the components of a four-component signed integer vector
+	/// to the 10-10-10-2-bit signed integer representation found in the OpenGL Specification,
+	/// and then packing these four values into a 32-bit unsigned integer.
+	/// The first vector component specifies the 10 least-significant bits of the result;
+	/// the forth component specifies the 2 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packI3x10_1x2(uvec4 const& v)
+	/// @see uint32 packSnorm3x10_1x2(vec4 const& v)
+	/// @see uint32 packUnorm3x10_1x2(vec4 const& v)
+	/// @see ivec4 unpackI3x10_1x2(uint32 const& p)
+	GLM_FUNC_DECL uint32 packI3x10_1x2(ivec4 const& v);
+
+	/// Unpacks a single 32-bit unsigned integer p into three 10-bit and one 2-bit signed integers.
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packU3x10_1x2(uvec4 const& v)
+	/// @see vec4 unpackSnorm3x10_1x2(uint32 const& p);
+	/// @see uvec4 unpackI3x10_1x2(uint32 const& p);
+	GLM_FUNC_DECL ivec4 unpackI3x10_1x2(uint32 p);
+
+	/// Returns an unsigned integer obtained by converting the components of a four-component unsigned integer vector
+	/// to the 10-10-10-2-bit unsigned integer representation found in the OpenGL Specification,
+	/// and then packing these four values into a 32-bit unsigned integer.
+	/// The first vector component specifies the 10 least-significant bits of the result;
+	/// the forth component specifies the 2 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packI3x10_1x2(ivec4 const& v)
+	/// @see uint32 packSnorm3x10_1x2(vec4 const& v)
+	/// @see uint32 packUnorm3x10_1x2(vec4 const& v)
+	/// @see ivec4 unpackU3x10_1x2(uint32 const& p)
+	GLM_FUNC_DECL uint32 packU3x10_1x2(uvec4 const& v);
+
+	/// Unpacks a single 32-bit unsigned integer p into three 10-bit and one 2-bit unsigned integers.
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packU3x10_1x2(uvec4 const& v)
+	/// @see vec4 unpackSnorm3x10_1x2(uint32 const& p);
+	/// @see uvec4 unpackI3x10_1x2(uint32 const& p);
+	GLM_FUNC_DECL uvec4 unpackU3x10_1x2(uint32 p);
+
+	/// First, converts the first three components of the normalized floating-point value v into 10-bit signed integer values.
+	/// Then, converts the forth component of the normalized floating-point value v into 2-bit signed integer values.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packSnorm3x10_1x2(xyz):	round(clamp(c, -1, +1) * 511.0)
+	/// packSnorm3x10_1x2(w):	round(clamp(c, -1, +1) * 1.0)
+	///
+	/// The first vector component specifies the 10 least-significant bits of the result;
+	/// the forth component specifies the 2 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see vec4 unpackSnorm3x10_1x2(uint32 const& p)
+	/// @see uint32 packUnorm3x10_1x2(vec4 const& v)
+	/// @see uint32 packU3x10_1x2(uvec4 const& v)
+	/// @see uint32 packI3x10_1x2(ivec4 const& v)
+	GLM_FUNC_DECL uint32 packSnorm3x10_1x2(vec4 const& v);
+
+	/// First, unpacks a single 32-bit unsigned integer p into four 16-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm3x10_1x2(xyz): clamp(f / 511.0, -1, +1)
+	/// unpackSnorm3x10_1x2(w): clamp(f / 511.0, -1, +1)
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packSnorm3x10_1x2(vec4 const& v)
+	/// @see vec4 unpackUnorm3x10_1x2(uint32 const& p))
+	/// @see uvec4 unpackI3x10_1x2(uint32 const& p)
+	/// @see uvec4 unpackU3x10_1x2(uint32 const& p)
+	GLM_FUNC_DECL vec4 unpackSnorm3x10_1x2(uint32 p);
+
+	/// First, converts the first three components of the normalized floating-point value v into 10-bit unsigned integer values.
+	/// Then, converts the forth component of the normalized floating-point value v into 2-bit signed uninteger values.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm3x10_1x2(xyz):	round(clamp(c, 0, +1) * 1023.0)
+	/// packUnorm3x10_1x2(w):	round(clamp(c, 0, +1) * 3.0)
+	///
+	/// The first vector component specifies the 10 least-significant bits of the result;
+	/// the forth component specifies the 2 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see vec4 unpackUnorm3x10_1x2(uint32 const& p)
+	/// @see uint32 packUnorm3x10_1x2(vec4 const& v)
+	/// @see uint32 packU3x10_1x2(uvec4 const& v)
+	/// @see uint32 packI3x10_1x2(ivec4 const& v)
+	GLM_FUNC_DECL uint32 packUnorm3x10_1x2(vec4 const& v);
+
+	/// First, unpacks a single 32-bit unsigned integer p into four 16-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm3x10_1x2(xyz): clamp(f / 1023.0, 0, +1)
+	/// unpackSnorm3x10_1x2(w): clamp(f / 3.0, 0, +1)
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packSnorm3x10_1x2(vec4 const& v)
+	/// @see vec4 unpackInorm3x10_1x2(uint32 const& p))
+	/// @see uvec4 unpackI3x10_1x2(uint32 const& p)
+	/// @see uvec4 unpackU3x10_1x2(uint32 const& p)
+	GLM_FUNC_DECL vec4 unpackUnorm3x10_1x2(uint32 p);
+
+	/// First, converts the first two components of the normalized floating-point value v into 11-bit signless floating-point values.
+	/// Then, converts the third component of the normalized floating-point value v into a 10-bit signless floating-point value.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The first vector component specifies the 11 least-significant bits of the result;
+	/// the last component specifies the 10 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see vec3 unpackF2x11_1x10(uint32 const& p)
+	GLM_FUNC_DECL uint32 packF2x11_1x10(vec3 const& v);
+
+	/// First, unpacks a single 32-bit unsigned integer p into two 11-bit signless floating-point values and one 10-bit signless floating-point value .
+	/// Then, each component is converted to a normalized floating-point value to generate the returned three-component vector.
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packF2x11_1x10(vec3 const& v)
+	GLM_FUNC_DECL vec3 unpackF2x11_1x10(uint32 p);
+
+
+	/// First, converts the first two components of the normalized floating-point value v into 11-bit signless floating-point values.
+	/// Then, converts the third component of the normalized floating-point value v into a 10-bit signless floating-point value.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The first vector component specifies the 11 least-significant bits of the result;
+	/// the last component specifies the 10 most-significant bits.
+	///
+	/// packF3x9_E1x5 allows encoding into RGBE / RGB9E5 format
+	///
+	/// @see gtc_packing
+	/// @see vec3 unpackF3x9_E1x5(uint32 const& p)
+	GLM_FUNC_DECL uint32 packF3x9_E1x5(vec3 const& v);
+
+	/// First, unpacks a single 32-bit unsigned integer p into two 11-bit signless floating-point values and one 10-bit signless floating-point value .
+	/// Then, each component is converted to a normalized floating-point value to generate the returned three-component vector.
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// unpackF3x9_E1x5 allows decoding RGBE / RGB9E5 data
+	///
+	/// @see gtc_packing
+	/// @see uint32 packF3x9_E1x5(vec3 const& v)
+	GLM_FUNC_DECL vec3 unpackF3x9_E1x5(uint32 p);
+
+	/// Returns an unsigned integer vector obtained by converting the components of a floating-point vector
+	/// to the 16-bit floating-point representation found in the OpenGL Specification.
+	/// The first vector component specifies the 16 least-significant bits of the result;
+	/// the forth component specifies the 16 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see vec<3, T, Q> unpackRGBM(vec<4, T, Q> const& p)
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> packRGBM(vec<3, T, Q> const& rgb);
+
+	/// Returns a floating-point vector with components obtained by reinterpreting an integer vector as 16-bit floating-point numbers and converting them to 32-bit floating-point values.
+	/// The first component of the vector is obtained from the 16 least-significant bits of v;
+	/// the forth component is obtained from the 16 most-significant bits of v.
+	///
+	/// @see gtc_packing
+	/// @see vec<4, T, Q> packRGBM(vec<3, float, Q> const& v)
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> unpackRGBM(vec<4, T, Q> const& rgbm);
+
+	/// Returns an unsigned integer vector obtained by converting the components of a floating-point vector
+	/// to the 16-bit floating-point representation found in the OpenGL Specification.
+	/// The first vector component specifies the 16 least-significant bits of the result;
+	/// the forth component specifies the 16 most-significant bits.
+	///
+	/// @see gtc_packing
+	/// @see vec<L, float, Q> unpackHalf(vec<L, uint16, Q> const& p)
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, uint16, Q> packHalf(vec<L, float, Q> const& v);
+
+	/// Returns a floating-point vector with components obtained by reinterpreting an integer vector as 16-bit floating-point numbers and converting them to 32-bit floating-point values.
+	/// The first component of the vector is obtained from the 16 least-significant bits of v;
+	/// the forth component is obtained from the 16 most-significant bits of v.
+	///
+	/// @see gtc_packing
+	/// @see vec<L, uint16, Q> packHalf(vec<L, float, Q> const& v)
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, float, Q> unpackHalf(vec<L, uint16, Q> const& p);
+
+	/// Convert each component of the normalized floating-point vector into unsigned integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec<L, floatType, Q> unpackUnorm(vec<L, intType, Q> const& p);
+	template<typename uintType, length_t L, typename floatType, qualifier Q>
+	GLM_FUNC_DECL vec<L, uintType, Q> packUnorm(vec<L, floatType, Q> const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see vec<L, intType, Q> packUnorm(vec<L, floatType, Q> const& v)
+	template<typename floatType, length_t L, typename uintType, qualifier Q>
+	GLM_FUNC_DECL vec<L, floatType, Q> unpackUnorm(vec<L, uintType, Q> const& v);
+
+	/// Convert each component of the normalized floating-point vector into signed integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec<L, floatType, Q> unpackSnorm(vec<L, intType, Q> const& p);
+	template<typename intType, length_t L, typename floatType, qualifier Q>
+	GLM_FUNC_DECL vec<L, intType, Q> packSnorm(vec<L, floatType, Q> const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see vec<L, intType, Q> packSnorm(vec<L, floatType, Q> const& v)
+	template<typename floatType, length_t L, typename intType, qualifier Q>
+	GLM_FUNC_DECL vec<L, floatType, Q> unpackSnorm(vec<L, intType, Q> const& v);
+
+	/// Convert each component of the normalized floating-point vector into unsigned integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec2 unpackUnorm2x4(uint8 p)
+	GLM_FUNC_DECL uint8 packUnorm2x4(vec2 const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see uint8 packUnorm2x4(vec2 const& v)
+	GLM_FUNC_DECL vec2 unpackUnorm2x4(uint8 p);
+
+	/// Convert each component of the normalized floating-point vector into unsigned integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec4 unpackUnorm4x4(uint16 p)
+	GLM_FUNC_DECL uint16 packUnorm4x4(vec4 const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packUnorm4x4(vec4 const& v)
+	GLM_FUNC_DECL vec4 unpackUnorm4x4(uint16 p);
+
+	/// Convert each component of the normalized floating-point vector into unsigned integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec3 unpackUnorm1x5_1x6_1x5(uint16 p)
+	GLM_FUNC_DECL uint16 packUnorm1x5_1x6_1x5(vec3 const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packUnorm1x5_1x6_1x5(vec3 const& v)
+	GLM_FUNC_DECL vec3 unpackUnorm1x5_1x6_1x5(uint16 p);
+
+	/// Convert each component of the normalized floating-point vector into unsigned integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec4 unpackUnorm3x5_1x1(uint16 p)
+	GLM_FUNC_DECL uint16 packUnorm3x5_1x1(vec4 const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packUnorm3x5_1x1(vec4 const& v)
+	GLM_FUNC_DECL vec4 unpackUnorm3x5_1x1(uint16 p);
+
+	/// Convert each component of the normalized floating-point vector into unsigned integer values.
+	///
+	/// @see gtc_packing
+	/// @see vec3 unpackUnorm2x3_1x2(uint8 p)
+	GLM_FUNC_DECL uint8 packUnorm2x3_1x2(vec3 const& v);
+
+	/// Convert a packed integer to a normalized floating-point vector.
+	///
+	/// @see gtc_packing
+	/// @see uint8 packUnorm2x3_1x2(vec3 const& v)
+	GLM_FUNC_DECL vec3 unpackUnorm2x3_1x2(uint8 p);
+
+
+
+	/// Convert each component from an integer vector into a packed integer.
+	///
+	/// @see gtc_packing
+	/// @see i8vec2 unpackInt2x8(int16 p)
+	GLM_FUNC_DECL int16 packInt2x8(i8vec2 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see int16 packInt2x8(i8vec2 const& v)
+	GLM_FUNC_DECL i8vec2 unpackInt2x8(int16 p);
+
+	/// Convert each component from an integer vector into a packed unsigned integer.
+	///
+	/// @see gtc_packing
+	/// @see u8vec2 unpackInt2x8(uint16 p)
+	GLM_FUNC_DECL uint16 packUint2x8(u8vec2 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see uint16 packInt2x8(u8vec2 const& v)
+	GLM_FUNC_DECL u8vec2 unpackUint2x8(uint16 p);
+
+	/// Convert each component from an integer vector into a packed integer.
+	///
+	/// @see gtc_packing
+	/// @see i8vec4 unpackInt4x8(int32 p)
+	GLM_FUNC_DECL int32 packInt4x8(i8vec4 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see int32 packInt2x8(i8vec4 const& v)
+	GLM_FUNC_DECL i8vec4 unpackInt4x8(int32 p);
+
+	/// Convert each component from an integer vector into a packed unsigned integer.
+	///
+	/// @see gtc_packing
+	/// @see u8vec4 unpackUint4x8(uint32 p)
+	GLM_FUNC_DECL uint32 packUint4x8(u8vec4 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see uint32 packUint4x8(u8vec2 const& v)
+	GLM_FUNC_DECL u8vec4 unpackUint4x8(uint32 p);
+
+	/// Convert each component from an integer vector into a packed integer.
+	///
+	/// @see gtc_packing
+	/// @see i16vec2 unpackInt2x16(int p)
+	GLM_FUNC_DECL int packInt2x16(i16vec2 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see int packInt2x16(i16vec2 const& v)
+	GLM_FUNC_DECL i16vec2 unpackInt2x16(int p);
+
+	/// Convert each component from an integer vector into a packed integer.
+	///
+	/// @see gtc_packing
+	/// @see i16vec4 unpackInt4x16(int64 p)
+	GLM_FUNC_DECL int64 packInt4x16(i16vec4 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see int64 packInt4x16(i16vec4 const& v)
+	GLM_FUNC_DECL i16vec4 unpackInt4x16(int64 p);
+
+	/// Convert each component from an integer vector into a packed unsigned integer.
+	///
+	/// @see gtc_packing
+	/// @see u16vec2 unpackUint2x16(uint p)
+	GLM_FUNC_DECL uint packUint2x16(u16vec2 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see uint packUint2x16(u16vec2 const& v)
+	GLM_FUNC_DECL u16vec2 unpackUint2x16(uint p);
+
+	/// Convert each component from an integer vector into a packed unsigned integer.
+	///
+	/// @see gtc_packing
+	/// @see u16vec4 unpackUint4x16(uint64 p)
+	GLM_FUNC_DECL uint64 packUint4x16(u16vec4 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see uint64 packUint4x16(u16vec4 const& v)
+	GLM_FUNC_DECL u16vec4 unpackUint4x16(uint64 p);
+
+	/// Convert each component from an integer vector into a packed integer.
+	///
+	/// @see gtc_packing
+	/// @see i32vec2 unpackInt2x32(int p)
+	GLM_FUNC_DECL int64 packInt2x32(i32vec2 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see int packInt2x16(i32vec2 const& v)
+	GLM_FUNC_DECL i32vec2 unpackInt2x32(int64 p);
+
+	/// Convert each component from an integer vector into a packed unsigned integer.
+	///
+	/// @see gtc_packing
+	/// @see u32vec2 unpackUint2x32(int p)
+	GLM_FUNC_DECL uint64 packUint2x32(u32vec2 const& v);
+
+	/// Convert a packed integer into an integer vector.
+	///
+	/// @see gtc_packing
+	/// @see int packUint2x16(u32vec2 const& v)
+	GLM_FUNC_DECL u32vec2 unpackUint2x32(uint64 p);
+
+	/// @}
+}// namespace glm
+
+#include "packing.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/packing.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/packing.inl
new file mode 100644
index 000000000000..84ad60c70f85
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/packing.inl
@@ -0,0 +1,938 @@
+/// @ref gtc_packing
+
+#include "../ext/scalar_relational.hpp"
+#include "../ext/vector_relational.hpp"
+#include "../common.hpp"
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../detail/type_half.hpp"
+#include <cstring>
+#include <limits>
+
+namespace glm{
+namespace detail
+{
+	GLM_FUNC_QUALIFIER glm::uint16 float2half(glm::uint32 f)
+	{
+		// 10 bits    =>                         EE EEEFFFFF
+		// 11 bits    =>                        EEE EEFFFFFF
+		// Half bits  =>                   SEEEEEFF FFFFFFFF
+		// Float bits => SEEEEEEE EFFFFFFF FFFFFFFF FFFFFFFF
+
+		// 0x00007c00 => 00000000 00000000 01111100 00000000
+		// 0x000003ff => 00000000 00000000 00000011 11111111
+		// 0x38000000 => 00111000 00000000 00000000 00000000
+		// 0x7f800000 => 01111111 10000000 00000000 00000000
+		// 0x00008000 => 00000000 00000000 10000000 00000000
+		return
+			((f >> 16) & 0x8000) | // sign
+			((((f & 0x7f800000) - 0x38000000) >> 13) & 0x7c00) | // exponential
+			((f >> 13) & 0x03ff); // Mantissa
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint32 float2packed11(glm::uint32 f)
+	{
+		// 10 bits    =>                         EE EEEFFFFF
+		// 11 bits    =>                        EEE EEFFFFFF
+		// Half bits  =>                   SEEEEEFF FFFFFFFF
+		// Float bits => SEEEEEEE EFFFFFFF FFFFFFFF FFFFFFFF
+
+		// 0x000007c0 => 00000000 00000000 00000111 11000000
+		// 0x00007c00 => 00000000 00000000 01111100 00000000
+		// 0x000003ff => 00000000 00000000 00000011 11111111
+		// 0x38000000 => 00111000 00000000 00000000 00000000
+		// 0x7f800000 => 01111111 10000000 00000000 00000000
+		// 0x00008000 => 00000000 00000000 10000000 00000000
+		return
+			((((f & 0x7f800000) - 0x38000000) >> 17) & 0x07c0) | // exponential
+			((f >> 17) & 0x003f); // Mantissa
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint32 packed11ToFloat(glm::uint32 p)
+	{
+		// 10 bits    =>                         EE EEEFFFFF
+		// 11 bits    =>                        EEE EEFFFFFF
+		// Half bits  =>                   SEEEEEFF FFFFFFFF
+		// Float bits => SEEEEEEE EFFFFFFF FFFFFFFF FFFFFFFF
+
+		// 0x000007c0 => 00000000 00000000 00000111 11000000
+		// 0x00007c00 => 00000000 00000000 01111100 00000000
+		// 0x000003ff => 00000000 00000000 00000011 11111111
+		// 0x38000000 => 00111000 00000000 00000000 00000000
+		// 0x7f800000 => 01111111 10000000 00000000 00000000
+		// 0x00008000 => 00000000 00000000 10000000 00000000
+		return
+			((((p & 0x07c0) << 17) + 0x38000000) & 0x7f800000) | // exponential
+			((p & 0x003f) << 17); // Mantissa
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint32 float2packed10(glm::uint32 f)
+	{
+		// 10 bits    =>                         EE EEEFFFFF
+		// 11 bits    =>                        EEE EEFFFFFF
+		// Half bits  =>                   SEEEEEFF FFFFFFFF
+		// Float bits => SEEEEEEE EFFFFFFF FFFFFFFF FFFFFFFF
+
+		// 0x0000001F => 00000000 00000000 00000000 00011111
+		// 0x0000003F => 00000000 00000000 00000000 00111111
+		// 0x000003E0 => 00000000 00000000 00000011 11100000
+		// 0x000007C0 => 00000000 00000000 00000111 11000000
+		// 0x00007C00 => 00000000 00000000 01111100 00000000
+		// 0x000003FF => 00000000 00000000 00000011 11111111
+		// 0x38000000 => 00111000 00000000 00000000 00000000
+		// 0x7f800000 => 01111111 10000000 00000000 00000000
+		// 0x00008000 => 00000000 00000000 10000000 00000000
+		return
+			((((f & 0x7f800000) - 0x38000000) >> 18) & 0x03E0) | // exponential
+			((f >> 18) & 0x001f); // Mantissa
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint32 packed10ToFloat(glm::uint32 p)
+	{
+		// 10 bits    =>                         EE EEEFFFFF
+		// 11 bits    =>                        EEE EEFFFFFF
+		// Half bits  =>                   SEEEEEFF FFFFFFFF
+		// Float bits => SEEEEEEE EFFFFFFF FFFFFFFF FFFFFFFF
+
+		// 0x0000001F => 00000000 00000000 00000000 00011111
+		// 0x0000003F => 00000000 00000000 00000000 00111111
+		// 0x000003E0 => 00000000 00000000 00000011 11100000
+		// 0x000007C0 => 00000000 00000000 00000111 11000000
+		// 0x00007C00 => 00000000 00000000 01111100 00000000
+		// 0x000003FF => 00000000 00000000 00000011 11111111
+		// 0x38000000 => 00111000 00000000 00000000 00000000
+		// 0x7f800000 => 01111111 10000000 00000000 00000000
+		// 0x00008000 => 00000000 00000000 10000000 00000000
+		return
+			((((p & 0x03E0) << 18) + 0x38000000) & 0x7f800000) | // exponential
+			((p & 0x001f) << 18); // Mantissa
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint half2float(glm::uint h)
+	{
+		return ((h & 0x8000) << 16) | ((( h & 0x7c00) + 0x1C000) << 13) | ((h & 0x03FF) << 13);
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint floatTo11bit(float x)
+	{
+		if(x == 0.0f)
+			return 0u;
+		else if(glm::isnan(x))
+			return ~0u;
+		else if(glm::isinf(x))
+			return 0x1Fu << 6u;
+
+		uint Pack = 0u;
+		memcpy(&Pack, &x, sizeof(Pack));
+		return float2packed11(Pack);
+	}
+
+	GLM_FUNC_QUALIFIER float packed11bitToFloat(glm::uint x)
+	{
+		if(x == 0)
+			return 0.0f;
+		else if(x == ((1 << 11) - 1))
+			return ~0;//NaN
+		else if(x == (0x1f << 6))
+			return ~0;//Inf
+
+		uint Result = packed11ToFloat(x);
+
+		float Temp = 0;
+		memcpy(&Temp, &Result, sizeof(Temp));
+		return Temp;
+	}
+
+	GLM_FUNC_QUALIFIER glm::uint floatTo10bit(float x)
+	{
+		if(x == 0.0f)
+			return 0u;
+		else if(glm::isnan(x))
+			return ~0u;
+		else if(glm::isinf(x))
+			return 0x1Fu << 5u;
+
+		uint Pack = 0;
+		memcpy(&Pack, &x, sizeof(Pack));
+		return float2packed10(Pack);
+	}
+
+	GLM_FUNC_QUALIFIER float packed10bitToFloat(glm::uint x)
+	{
+		if(x == 0)
+			return 0.0f;
+		else if(x == ((1 << 10) - 1))
+			return ~0;//NaN
+		else if(x == (0x1f << 5))
+			return ~0;//Inf
+
+		uint Result = packed10ToFloat(x);
+
+		float Temp = 0;
+		memcpy(&Temp, &Result, sizeof(Temp));
+		return Temp;
+	}
+
+//	GLM_FUNC_QUALIFIER glm::uint f11_f11_f10(float x, float y, float z)
+//	{
+//		return ((floatTo11bit(x) & ((1 << 11) - 1)) << 0) |  ((floatTo11bit(y) & ((1 << 11) - 1)) << 11) | ((floatTo10bit(z) & ((1 << 10) - 1)) << 22);
+//	}
+
+	union u3u3u2
+	{
+		struct
+		{
+			uint x : 3;
+			uint y : 3;
+			uint z : 2;
+		} data;
+		uint8 pack;
+	};
+
+	union u4u4
+	{
+		struct
+		{
+			uint x : 4;
+			uint y : 4;
+		} data;
+		uint8 pack;
+	};
+
+	union u4u4u4u4
+	{
+		struct
+		{
+			uint x : 4;
+			uint y : 4;
+			uint z : 4;
+			uint w : 4;
+		} data;
+		uint16 pack;
+	};
+
+	union u5u6u5
+	{
+		struct
+		{
+			uint x : 5;
+			uint y : 6;
+			uint z : 5;
+		} data;
+		uint16 pack;
+	};
+
+	union u5u5u5u1
+	{
+		struct
+		{
+			uint x : 5;
+			uint y : 5;
+			uint z : 5;
+			uint w : 1;
+		} data;
+		uint16 pack;
+	};
+
+	union u10u10u10u2
+	{
+		struct
+		{
+			uint x : 10;
+			uint y : 10;
+			uint z : 10;
+			uint w : 2;
+		} data;
+		uint32 pack;
+	};
+
+	union i10i10i10i2
+	{
+		struct
+		{
+			int x : 10;
+			int y : 10;
+			int z : 10;
+			int w : 2;
+		} data;
+		uint32 pack;
+	};
+
+	union u9u9u9e5
+	{
+		struct
+		{
+			uint x : 9;
+			uint y : 9;
+			uint z : 9;
+			uint w : 5;
+		} data;
+		uint32 pack;
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_half
+	{};
+
+	template<qualifier Q>
+	struct compute_half<1, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<1, uint16, Q> pack(vec<1, float, Q> const& v)
+		{
+			int16 const Unpack(detail::toFloat16(v.x));
+			u16vec1 Packed;
+			memcpy(&Packed, &Unpack, sizeof(Packed));
+			return Packed;
+		}
+
+		GLM_FUNC_QUALIFIER static vec<1, float, Q> unpack(vec<1, uint16, Q> const& v)
+		{
+			i16vec1 Unpack;
+			memcpy(&Unpack, &v, sizeof(Unpack));
+			return vec<1, float, Q>(detail::toFloat32(v.x));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_half<2, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<2, uint16, Q> pack(vec<2, float, Q> const& v)
+		{
+			vec<2, int16, Q> const Unpack(detail::toFloat16(v.x), detail::toFloat16(v.y));
+			u16vec2 Packed;
+			memcpy(&Packed, &Unpack, sizeof(Packed));
+			return Packed;
+		}
+
+		GLM_FUNC_QUALIFIER static vec<2, float, Q> unpack(vec<2, uint16, Q> const& v)
+		{
+			i16vec2 Unpack;
+			memcpy(&Unpack, &v, sizeof(Unpack));
+			return vec<2, float, Q>(detail::toFloat32(v.x), detail::toFloat32(v.y));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_half<3, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, uint16, Q> pack(vec<3, float, Q> const& v)
+		{
+			vec<3, int16, Q> const Unpack(detail::toFloat16(v.x), detail::toFloat16(v.y), detail::toFloat16(v.z));
+			u16vec3 Packed;
+			memcpy(&Packed, &Unpack, sizeof(Packed));
+			return Packed;
+		}
+
+		GLM_FUNC_QUALIFIER static vec<3, float, Q> unpack(vec<3, uint16, Q> const& v)
+		{
+			i16vec3 Unpack;
+			memcpy(&Unpack, &v, sizeof(Unpack));
+			return vec<3, float, Q>(detail::toFloat32(v.x), detail::toFloat32(v.y), detail::toFloat32(v.z));
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_half<4, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint16, Q> pack(vec<4, float, Q> const& v)
+		{
+			vec<4, int16, Q> const Unpack(detail::toFloat16(v.x), detail::toFloat16(v.y), detail::toFloat16(v.z), detail::toFloat16(v.w));
+			u16vec4 Packed;
+			memcpy(&Packed, &Unpack, sizeof(Packed));
+			return Packed;
+		}
+
+		GLM_FUNC_QUALIFIER static vec<4, float, Q> unpack(vec<4, uint16, Q> const& v)
+		{
+			i16vec4 Unpack;
+			memcpy(&Unpack, &v, sizeof(Unpack));
+			return vec<4, float, Q>(detail::toFloat32(v.x), detail::toFloat32(v.y), detail::toFloat32(v.z), detail::toFloat32(v.w));
+		}
+	};
+}//namespace detail
+
+	GLM_FUNC_QUALIFIER uint8 packUnorm1x8(float v)
+	{
+		return static_cast<uint8>(round(clamp(v, 0.0f, 1.0f) * 255.0f));
+	}
+
+	GLM_FUNC_QUALIFIER float unpackUnorm1x8(uint8 p)
+	{
+		float const Unpack(p);
+		return Unpack * static_cast<float>(0.0039215686274509803921568627451); // 1 / 255
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packUnorm2x8(vec2 const& v)
+	{
+		u8vec2 const Topack(round(clamp(v, 0.0f, 1.0f) * 255.0f));
+
+		uint16 Unpack = 0;
+		memcpy(&Unpack, &Topack, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER vec2 unpackUnorm2x8(uint16 p)
+	{
+		u8vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return vec2(Unpack) * float(0.0039215686274509803921568627451); // 1 / 255
+	}
+
+	GLM_FUNC_QUALIFIER uint8 packSnorm1x8(float v)
+	{
+		int8 const Topack(static_cast<int8>(round(clamp(v ,-1.0f, 1.0f) * 127.0f)));
+		uint8 Packed = 0;
+		memcpy(&Packed, &Topack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER float unpackSnorm1x8(uint8 p)
+	{
+		int8 Unpack = 0;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return clamp(
+			static_cast<float>(Unpack) * 0.00787401574803149606299212598425f, // 1.0f / 127.0f
+			-1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packSnorm2x8(vec2 const& v)
+	{
+		i8vec2 const Topack(round(clamp(v, -1.0f, 1.0f) * 127.0f));
+		uint16 Packed = 0;
+		memcpy(&Packed, &Topack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER vec2 unpackSnorm2x8(uint16 p)
+	{
+		i8vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return clamp(
+			vec2(Unpack) * 0.00787401574803149606299212598425f, // 1.0f / 127.0f
+			-1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packUnorm1x16(float s)
+	{
+		return static_cast<uint16>(round(clamp(s, 0.0f, 1.0f) * 65535.0f));
+	}
+
+	GLM_FUNC_QUALIFIER float unpackUnorm1x16(uint16 p)
+	{
+		float const Unpack(p);
+		return Unpack * 1.5259021896696421759365224689097e-5f; // 1.0 / 65535.0
+	}
+
+	GLM_FUNC_QUALIFIER uint64 packUnorm4x16(vec4 const& v)
+	{
+		u16vec4 const Topack(round(clamp(v , 0.0f, 1.0f) * 65535.0f));
+		uint64 Packed = 0;
+		memcpy(&Packed, &Topack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackUnorm4x16(uint64 p)
+	{
+		u16vec4 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return vec4(Unpack) * 1.5259021896696421759365224689097e-5f; // 1.0 / 65535.0
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packSnorm1x16(float v)
+	{
+		int16 const Topack = static_cast<int16>(round(clamp(v ,-1.0f, 1.0f) * 32767.0f));
+		uint16 Packed = 0;
+		memcpy(&Packed, &Topack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER float unpackSnorm1x16(uint16 p)
+	{
+		int16 Unpack = 0;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return clamp(
+			static_cast<float>(Unpack) * 3.0518509475997192297128208258309e-5f, //1.0f / 32767.0f,
+			-1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER uint64 packSnorm4x16(vec4 const& v)
+	{
+		i16vec4 const Topack(round(clamp(v ,-1.0f, 1.0f) * 32767.0f));
+		uint64 Packed = 0;
+		memcpy(&Packed, &Topack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackSnorm4x16(uint64 p)
+	{
+		i16vec4 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return clamp(
+			vec4(Unpack) * 3.0518509475997192297128208258309e-5f, //1.0f / 32767.0f,
+			-1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packHalf1x16(float v)
+	{
+		int16 const Topack(detail::toFloat16(v));
+		uint16 Packed = 0;
+		memcpy(&Packed, &Topack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER float unpackHalf1x16(uint16 v)
+	{
+		int16 Unpack = 0;
+		memcpy(&Unpack, &v, sizeof(Unpack));
+		return detail::toFloat32(Unpack);
+	}
+
+	GLM_FUNC_QUALIFIER uint64 packHalf4x16(glm::vec4 const& v)
+	{
+		i16vec4 const Unpack(
+			detail::toFloat16(v.x),
+			detail::toFloat16(v.y),
+			detail::toFloat16(v.z),
+			detail::toFloat16(v.w));
+		uint64 Packed = 0;
+		memcpy(&Packed, &Unpack, sizeof(Packed));
+		return Packed;
+	}
+
+	GLM_FUNC_QUALIFIER glm::vec4 unpackHalf4x16(uint64 v)
+	{
+		i16vec4 Unpack;
+		memcpy(&Unpack, &v, sizeof(Unpack));
+		return vec4(
+			detail::toFloat32(Unpack.x),
+			detail::toFloat32(Unpack.y),
+			detail::toFloat32(Unpack.z),
+			detail::toFloat32(Unpack.w));
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packI3x10_1x2(ivec4 const& v)
+	{
+		detail::i10i10i10i2 Result;
+		Result.data.x = v.x;
+		Result.data.y = v.y;
+		Result.data.z = v.z;
+		Result.data.w = v.w;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER ivec4 unpackI3x10_1x2(uint32 v)
+	{
+		detail::i10i10i10i2 Unpack;
+		Unpack.pack = v;
+		return ivec4(
+			Unpack.data.x,
+			Unpack.data.y,
+			Unpack.data.z,
+			Unpack.data.w);
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packU3x10_1x2(uvec4 const& v)
+	{
+		detail::u10u10u10u2 Result;
+		Result.data.x = v.x;
+		Result.data.y = v.y;
+		Result.data.z = v.z;
+		Result.data.w = v.w;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER uvec4 unpackU3x10_1x2(uint32 v)
+	{
+		detail::u10u10u10u2 Unpack;
+		Unpack.pack = v;
+		return uvec4(
+			Unpack.data.x,
+			Unpack.data.y,
+			Unpack.data.z,
+			Unpack.data.w);
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packSnorm3x10_1x2(vec4 const& v)
+	{
+		ivec4 const Pack(round(clamp(v,-1.0f, 1.0f) * vec4(511.f, 511.f, 511.f, 1.f)));
+
+		detail::i10i10i10i2 Result;
+		Result.data.x = Pack.x;
+		Result.data.y = Pack.y;
+		Result.data.z = Pack.z;
+		Result.data.w = Pack.w;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackSnorm3x10_1x2(uint32 v)
+	{
+		detail::i10i10i10i2 Unpack;
+		Unpack.pack = v;
+
+		vec4 const Result(Unpack.data.x, Unpack.data.y, Unpack.data.z, Unpack.data.w);
+
+		return clamp(Result * vec4(1.f / 511.f, 1.f / 511.f, 1.f / 511.f, 1.f), -1.0f, 1.0f);
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packUnorm3x10_1x2(vec4 const& v)
+	{
+		uvec4 const Unpack(round(clamp(v, 0.0f, 1.0f) * vec4(1023.f, 1023.f, 1023.f, 3.f)));
+
+		detail::u10u10u10u2 Result;
+		Result.data.x = Unpack.x;
+		Result.data.y = Unpack.y;
+		Result.data.z = Unpack.z;
+		Result.data.w = Unpack.w;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackUnorm3x10_1x2(uint32 v)
+	{
+		vec4 const ScaleFactors(1.0f / 1023.f, 1.0f / 1023.f, 1.0f / 1023.f, 1.0f / 3.f);
+
+		detail::u10u10u10u2 Unpack;
+		Unpack.pack = v;
+		return vec4(Unpack.data.x, Unpack.data.y, Unpack.data.z, Unpack.data.w) * ScaleFactors;
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packF2x11_1x10(vec3 const& v)
+	{
+		return
+			((detail::floatTo11bit(v.x) & ((1 << 11) - 1)) <<  0) |
+			((detail::floatTo11bit(v.y) & ((1 << 11) - 1)) << 11) |
+			((detail::floatTo10bit(v.z) & ((1 << 10) - 1)) << 22);
+	}
+
+	GLM_FUNC_QUALIFIER vec3 unpackF2x11_1x10(uint32 v)
+	{
+		return vec3(
+			detail::packed11bitToFloat(v >> 0),
+			detail::packed11bitToFloat(v >> 11),
+			detail::packed10bitToFloat(v >> 22));
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packF3x9_E1x5(vec3 const& v)
+	{
+		float const SharedExpMax = (pow(2.0f, 9.0f - 1.0f) / pow(2.0f, 9.0f)) * pow(2.0f, 31.f - 15.f);
+		vec3 const Color = clamp(v, 0.0f, SharedExpMax);
+		float const MaxColor = max(Color.x, max(Color.y, Color.z));
+
+		float const ExpSharedP = max(-15.f - 1.f, floor(log2(MaxColor))) + 1.0f + 15.f;
+		float const MaxShared = floor(MaxColor / pow(2.0f, (ExpSharedP - 15.f - 9.f)) + 0.5f);
+		float const ExpShared = equal(MaxShared, pow(2.0f, 9.0f), epsilon<float>()) ? ExpSharedP + 1.0f : ExpSharedP;
+
+		uvec3 const ColorComp(floor(Color / pow(2.f, (ExpShared - 15.f - 9.f)) + 0.5f));
+
+		detail::u9u9u9e5 Unpack;
+		Unpack.data.x = ColorComp.x;
+		Unpack.data.y = ColorComp.y;
+		Unpack.data.z = ColorComp.z;
+		Unpack.data.w = uint(ExpShared);
+		return Unpack.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec3 unpackF3x9_E1x5(uint32 v)
+	{
+		detail::u9u9u9e5 Unpack;
+		Unpack.pack = v;
+
+		return vec3(Unpack.data.x, Unpack.data.y, Unpack.data.z) * pow(2.0f, Unpack.data.w - 15.f - 9.f);
+	}
+
+	// Based on Brian Karis http://graphicrants.blogspot.fr/2009/04/rgbm-color-encoding.html
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> packRGBM(vec<3, T, Q> const& rgb)
+	{
+		vec<3, T, Q> const Color(rgb * static_cast<T>(1.0 / 6.0));
+		T Alpha = clamp(max(max(Color.x, Color.y), max(Color.z, static_cast<T>(1e-6))), static_cast<T>(0), static_cast<T>(1));
+		Alpha = ceil(Alpha * static_cast<T>(255.0)) / static_cast<T>(255.0);
+		return vec<4, T, Q>(Color / Alpha, Alpha);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> unpackRGBM(vec<4, T, Q> const& rgbm)
+	{
+		return vec<3, T, Q>(rgbm.x, rgbm.y, rgbm.z) * rgbm.w * static_cast<T>(6);
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, uint16, Q> packHalf(vec<L, float, Q> const& v)
+	{
+		return detail::compute_half<L, Q>::pack(v);
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, float, Q> unpackHalf(vec<L, uint16, Q> const& v)
+	{
+		return detail::compute_half<L, Q>::unpack(v);
+	}
+
+	template<typename uintType, length_t L, typename floatType, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, uintType, Q> packUnorm(vec<L, floatType, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<uintType>::is_integer, "uintType must be an integer type");
+		GLM_STATIC_ASSERT(std::numeric_limits<floatType>::is_iec559, "floatType must be a floating point type");
+
+		return vec<L, uintType, Q>(round(clamp(v, static_cast<floatType>(0), static_cast<floatType>(1)) * static_cast<floatType>(std::numeric_limits<uintType>::max())));
+	}
+
+	template<typename floatType, length_t L, typename uintType, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, floatType, Q> unpackUnorm(vec<L, uintType, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<uintType>::is_integer, "uintType must be an integer type");
+		GLM_STATIC_ASSERT(std::numeric_limits<floatType>::is_iec559, "floatType must be a floating point type");
+
+		return vec<L, floatType, Q>(v) * (static_cast<floatType>(1) / static_cast<floatType>(std::numeric_limits<uintType>::max()));
+	}
+
+	template<typename intType, length_t L, typename floatType, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, intType, Q> packSnorm(vec<L, floatType, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<intType>::is_integer, "uintType must be an integer type");
+		GLM_STATIC_ASSERT(std::numeric_limits<floatType>::is_iec559, "floatType must be a floating point type");
+
+		return vec<L, intType, Q>(round(clamp(v , static_cast<floatType>(-1), static_cast<floatType>(1)) * static_cast<floatType>(std::numeric_limits<intType>::max())));
+	}
+
+	template<typename floatType, length_t L, typename intType, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, floatType, Q> unpackSnorm(vec<L, intType, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<intType>::is_integer, "uintType must be an integer type");
+		GLM_STATIC_ASSERT(std::numeric_limits<floatType>::is_iec559, "floatType must be a floating point type");
+
+		return clamp(vec<L, floatType, Q>(v) * (static_cast<floatType>(1) / static_cast<floatType>(std::numeric_limits<intType>::max())), static_cast<floatType>(-1), static_cast<floatType>(1));
+	}
+
+	GLM_FUNC_QUALIFIER uint8 packUnorm2x4(vec2 const& v)
+	{
+		u32vec2 const Unpack(round(clamp(v, 0.0f, 1.0f) * 15.0f));
+		detail::u4u4 Result;
+		Result.data.x = Unpack.x;
+		Result.data.y = Unpack.y;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec2 unpackUnorm2x4(uint8 v)
+	{
+		float const ScaleFactor(1.f / 15.f);
+		detail::u4u4 Unpack;
+		Unpack.pack = v;
+		return vec2(Unpack.data.x, Unpack.data.y) * ScaleFactor;
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packUnorm4x4(vec4 const& v)
+	{
+		u32vec4 const Unpack(round(clamp(v, 0.0f, 1.0f) * 15.0f));
+		detail::u4u4u4u4 Result;
+		Result.data.x = Unpack.x;
+		Result.data.y = Unpack.y;
+		Result.data.z = Unpack.z;
+		Result.data.w = Unpack.w;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackUnorm4x4(uint16 v)
+	{
+		float const ScaleFactor(1.f / 15.f);
+		detail::u4u4u4u4 Unpack;
+		Unpack.pack = v;
+		return vec4(Unpack.data.x, Unpack.data.y, Unpack.data.z, Unpack.data.w) * ScaleFactor;
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packUnorm1x5_1x6_1x5(vec3 const& v)
+	{
+		u32vec3 const Unpack(round(clamp(v, 0.0f, 1.0f) * vec3(31.f, 63.f, 31.f)));
+		detail::u5u6u5 Result;
+		Result.data.x = Unpack.x;
+		Result.data.y = Unpack.y;
+		Result.data.z = Unpack.z;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec3 unpackUnorm1x5_1x6_1x5(uint16 v)
+	{
+		vec3 const ScaleFactor(1.f / 31.f, 1.f / 63.f, 1.f / 31.f);
+		detail::u5u6u5 Unpack;
+		Unpack.pack = v;
+		return vec3(Unpack.data.x, Unpack.data.y, Unpack.data.z) * ScaleFactor;
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packUnorm3x5_1x1(vec4 const& v)
+	{
+		u32vec4 const Unpack(round(clamp(v, 0.0f, 1.0f) * vec4(31.f, 31.f, 31.f, 1.f)));
+		detail::u5u5u5u1 Result;
+		Result.data.x = Unpack.x;
+		Result.data.y = Unpack.y;
+		Result.data.z = Unpack.z;
+		Result.data.w = Unpack.w;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec4 unpackUnorm3x5_1x1(uint16 v)
+	{
+		vec4 const ScaleFactor(1.f / 31.f, 1.f / 31.f, 1.f / 31.f, 1.f);
+		detail::u5u5u5u1 Unpack;
+		Unpack.pack = v;
+		return vec4(Unpack.data.x, Unpack.data.y, Unpack.data.z, Unpack.data.w) * ScaleFactor;
+	}
+
+	GLM_FUNC_QUALIFIER uint8 packUnorm2x3_1x2(vec3 const& v)
+	{
+		u32vec3 const Unpack(round(clamp(v, 0.0f, 1.0f) * vec3(7.f, 7.f, 3.f)));
+		detail::u3u3u2 Result;
+		Result.data.x = Unpack.x;
+		Result.data.y = Unpack.y;
+		Result.data.z = Unpack.z;
+		return Result.pack;
+	}
+
+	GLM_FUNC_QUALIFIER vec3 unpackUnorm2x3_1x2(uint8 v)
+	{
+		vec3 const ScaleFactor(1.f / 7.f, 1.f / 7.f, 1.f / 3.f);
+		detail::u3u3u2 Unpack;
+		Unpack.pack = v;
+		return vec3(Unpack.data.x, Unpack.data.y, Unpack.data.z) * ScaleFactor;
+	}
+
+	GLM_FUNC_QUALIFIER int16 packInt2x8(i8vec2 const& v)
+	{
+		int16 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER i8vec2 unpackInt2x8(int16 p)
+	{
+		i8vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER uint16 packUint2x8(u8vec2 const& v)
+	{
+		uint16 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER u8vec2 unpackUint2x8(uint16 p)
+	{
+		u8vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER int32 packInt4x8(i8vec4 const& v)
+	{
+		int32 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER i8vec4 unpackInt4x8(int32 p)
+	{
+		i8vec4 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER uint32 packUint4x8(u8vec4 const& v)
+	{
+		uint32 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER u8vec4 unpackUint4x8(uint32 p)
+	{
+		u8vec4 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER int packInt2x16(i16vec2 const& v)
+	{
+		int Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER i16vec2 unpackInt2x16(int p)
+	{
+		i16vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER int64 packInt4x16(i16vec4 const& v)
+	{
+		int64 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER i16vec4 unpackInt4x16(int64 p)
+	{
+		i16vec4 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER uint packUint2x16(u16vec2 const& v)
+	{
+		uint Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER u16vec2 unpackUint2x16(uint p)
+	{
+		u16vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER uint64 packUint4x16(u16vec4 const& v)
+	{
+		uint64 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER u16vec4 unpackUint4x16(uint64 p)
+	{
+		u16vec4 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER int64 packInt2x32(i32vec2 const& v)
+	{
+		int64 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER i32vec2 unpackInt2x32(int64 p)
+	{
+		i32vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+
+	GLM_FUNC_QUALIFIER uint64 packUint2x32(u32vec2 const& v)
+	{
+		uint64 Pack = 0;
+		memcpy(&Pack, &v, sizeof(Pack));
+		return Pack;
+	}
+
+	GLM_FUNC_QUALIFIER u32vec2 unpackUint2x32(uint64 p)
+	{
+		u32vec2 Unpack;
+		memcpy(&Unpack, &p, sizeof(Unpack));
+		return Unpack;
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.hpp
new file mode 100644
index 000000000000..314449ebd97e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.hpp
@@ -0,0 +1,173 @@
+/// @ref gtc_quaternion
+/// @file glm/gtc/quaternion.hpp
+///
+/// @see core (dependence)
+/// @see gtc_constants (dependence)
+///
+/// @defgroup gtc_quaternion GLM_GTC_quaternion
+/// @ingroup gtc
+///
+/// Include <glm/gtc/quaternion.hpp> to use the features of this extension.
+///
+/// Defines a templated quaternion type and several quaternion operations.
+
+#pragma once
+
+// Dependency:
+#include "../gtc/constants.hpp"
+#include "../gtc/matrix_transform.hpp"
+#include "../ext/vector_relational.hpp"
+#include "../ext/quaternion_common.hpp"
+#include "../ext/quaternion_float.hpp"
+#include "../ext/quaternion_float_precision.hpp"
+#include "../ext/quaternion_double.hpp"
+#include "../ext/quaternion_double_precision.hpp"
+#include "../ext/quaternion_relational.hpp"
+#include "../ext/quaternion_geometric.hpp"
+#include "../ext/quaternion_trigonometric.hpp"
+#include "../ext/quaternion_transform.hpp"
+#include "../detail/type_mat3x3.hpp"
+#include "../detail/type_mat4x4.hpp"
+#include "../detail/type_vec3.hpp"
+#include "../detail/type_vec4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_quaternion extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_quaternion
+	/// @{
+
+	/// Returns euler angles, pitch as x, yaw as y, roll as z.
+	/// The result is expressed in radians.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> eulerAngles(qua<T, Q> const& x);
+
+	/// Returns roll value of euler angles expressed in radians.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T roll(qua<T, Q> const& x);
+
+	/// Returns pitch value of euler angles expressed in radians.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T pitch(qua<T, Q> const& x);
+
+	/// Returns yaw value of euler angles expressed in radians.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T yaw(qua<T, Q> const& x);
+
+	/// Converts a quaternion to a 3 * 3 matrix.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> mat3_cast(qua<T, Q> const& x);
+
+	/// Converts a quaternion to a 4 * 4 matrix.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> mat4_cast(qua<T, Q> const& x);
+
+	/// Converts a pure rotation 3 * 3 matrix to a quaternion.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> quat_cast(mat<3, 3, T, Q> const& x);
+
+	/// Converts a pure rotation 4 * 4 matrix to a quaternion.
+	///
+	/// @tparam T Floating-point scalar types.
+	///
+	/// @see gtc_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> quat_cast(mat<4, 4, T, Q> const& x);
+
+	/// Returns the component-wise comparison result of x < y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_relational
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, bool, Q> lessThan(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x <= y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_relational
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, bool, Q> lessThanEqual(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x > y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_relational
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, bool, Q> greaterThan(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x >= y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_quaternion_relational
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<4, bool, Q> greaterThanEqual(qua<T, Q> const& x, qua<T, Q> const& y);
+
+	/// Build a look at quaternion based on the default handedness.
+	///
+	/// @param direction Desired forward direction. Needs to be normalized.
+	/// @param up Up vector, how the camera is oriented. Typically (0, 1, 0).
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> quatLookAt(
+		vec<3, T, Q> const& direction,
+		vec<3, T, Q> const& up);
+
+	/// Build a right-handed look at quaternion.
+	///
+	/// @param direction Desired forward direction onto which the -z-axis gets mapped. Needs to be normalized.
+	/// @param up Up vector, how the camera is oriented. Typically (0, 1, 0).
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> quatLookAtRH(
+		vec<3, T, Q> const& direction,
+		vec<3, T, Q> const& up);
+
+	/// Build a left-handed look at quaternion.
+	///
+	/// @param direction Desired forward direction onto which the +z-axis gets mapped. Needs to be normalized.
+	/// @param up Up vector, how the camera is oriented. Typically (0, 1, 0).
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> quatLookAtLH(
+		vec<3, T, Q> const& direction,
+		vec<3, T, Q> const& up);
+	/// @}
+} //namespace glm
+
+#include "quaternion.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.inl
new file mode 100644
index 000000000000..ea159f298819
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion.inl
@@ -0,0 +1,208 @@
+#include "../trigonometric.hpp"
+#include "../geometric.hpp"
+#include "../exponential.hpp"
+#include "epsilon.hpp"
+#include <limits>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> eulerAngles(qua<T, Q> const& x)
+	{
+		return vec<3, T, Q>(pitch(x), yaw(x), roll(x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T roll(qua<T, Q> const& q)
+	{
+		T const y = static_cast<T>(2) * (q.x * q.y + q.w * q.z);
+		T const x = q.w * q.w + q.x * q.x - q.y * q.y - q.z * q.z;
+
+		if(all(equal(vec<2, T, Q>(x, y), vec<2, T, Q>(0), epsilon<T>()))) //avoid atan2(0,0) - handle singularity - Matiis
+			return static_cast<T>(0);
+
+		return static_cast<T>(atan(y, x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T pitch(qua<T, Q> const& q)
+	{
+		//return T(atan(T(2) * (q.y * q.z + q.w * q.x), q.w * q.w - q.x * q.x - q.y * q.y + q.z * q.z));
+		T const y = static_cast<T>(2) * (q.y * q.z + q.w * q.x);
+		T const x = q.w * q.w - q.x * q.x - q.y * q.y + q.z * q.z;
+
+		if(all(equal(vec<2, T, Q>(x, y), vec<2, T, Q>(0), epsilon<T>()))) //avoid atan2(0,0) - handle singularity - Matiis
+			return static_cast<T>(static_cast<T>(2) * atan(q.x, q.w));
+
+		return static_cast<T>(atan(y, x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T yaw(qua<T, Q> const& q)
+	{
+		return asin(clamp(static_cast<T>(-2) * (q.x * q.z - q.w * q.y), static_cast<T>(-1), static_cast<T>(1)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> mat3_cast(qua<T, Q> const& q)
+	{
+		mat<3, 3, T, Q> Result(T(1));
+		T qxx(q.x * q.x);
+		T qyy(q.y * q.y);
+		T qzz(q.z * q.z);
+		T qxz(q.x * q.z);
+		T qxy(q.x * q.y);
+		T qyz(q.y * q.z);
+		T qwx(q.w * q.x);
+		T qwy(q.w * q.y);
+		T qwz(q.w * q.z);
+
+		Result[0][0] = T(1) - T(2) * (qyy +  qzz);
+		Result[0][1] = T(2) * (qxy + qwz);
+		Result[0][2] = T(2) * (qxz - qwy);
+
+		Result[1][0] = T(2) * (qxy - qwz);
+		Result[1][1] = T(1) - T(2) * (qxx +  qzz);
+		Result[1][2] = T(2) * (qyz + qwx);
+
+		Result[2][0] = T(2) * (qxz + qwy);
+		Result[2][1] = T(2) * (qyz - qwx);
+		Result[2][2] = T(1) - T(2) * (qxx +  qyy);
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> mat4_cast(qua<T, Q> const& q)
+	{
+		return mat<4, 4, T, Q>(mat3_cast(q));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> quat_cast(mat<3, 3, T, Q> const& m)
+	{
+		T fourXSquaredMinus1 = m[0][0] - m[1][1] - m[2][2];
+		T fourYSquaredMinus1 = m[1][1] - m[0][0] - m[2][2];
+		T fourZSquaredMinus1 = m[2][2] - m[0][0] - m[1][1];
+		T fourWSquaredMinus1 = m[0][0] + m[1][1] + m[2][2];
+
+		int biggestIndex = 0;
+		T fourBiggestSquaredMinus1 = fourWSquaredMinus1;
+		if(fourXSquaredMinus1 > fourBiggestSquaredMinus1)
+		{
+			fourBiggestSquaredMinus1 = fourXSquaredMinus1;
+			biggestIndex = 1;
+		}
+		if(fourYSquaredMinus1 > fourBiggestSquaredMinus1)
+		{
+			fourBiggestSquaredMinus1 = fourYSquaredMinus1;
+			biggestIndex = 2;
+		}
+		if(fourZSquaredMinus1 > fourBiggestSquaredMinus1)
+		{
+			fourBiggestSquaredMinus1 = fourZSquaredMinus1;
+			biggestIndex = 3;
+		}
+
+		T biggestVal = sqrt(fourBiggestSquaredMinus1 + static_cast<T>(1)) * static_cast<T>(0.5);
+		T mult = static_cast<T>(0.25) / biggestVal;
+
+		switch(biggestIndex)
+		{
+		case 0:
+			return qua<T, Q>::wxyz(biggestVal, (m[1][2] - m[2][1]) * mult, (m[2][0] - m[0][2]) * mult, (m[0][1] - m[1][0]) * mult);
+		case 1:
+			return qua<T, Q>::wxyz((m[1][2] - m[2][1]) * mult, biggestVal, (m[0][1] + m[1][0]) * mult, (m[2][0] + m[0][2]) * mult);
+		case 2:
+			return qua<T, Q>::wxyz((m[2][0] - m[0][2]) * mult, (m[0][1] + m[1][0]) * mult, biggestVal, (m[1][2] + m[2][1]) * mult);
+		case 3:
+			return qua<T, Q>::wxyz((m[0][1] - m[1][0]) * mult, (m[2][0] + m[0][2]) * mult, (m[1][2] + m[2][1]) * mult, biggestVal);
+		default: // Silence a -Wswitch-default warning in GCC. Should never actually get here. Assert is just for sanity.
+			assert(false);
+			return qua<T, Q>::wxyz(1, 0, 0, 0);
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> quat_cast(mat<4, 4, T, Q> const& m4)
+	{
+		return quat_cast(mat<3, 3, T, Q>(m4));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, bool, Q> lessThan(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		vec<4, bool, Q> Result(false, false, false, false);
+		for(length_t i = 0; i < x.length(); ++i)
+			Result[i] = x[i] < y[i];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, bool, Q> lessThanEqual(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		vec<4, bool, Q> Result(false, false, false, false);
+		for(length_t i = 0; i < x.length(); ++i)
+			Result[i] = x[i] <= y[i];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, bool, Q> greaterThan(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		vec<4, bool, Q> Result(false, false, false, false);
+		for(length_t i = 0; i < x.length(); ++i)
+			Result[i] = x[i] > y[i];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<4, bool, Q> greaterThanEqual(qua<T, Q> const& x, qua<T, Q> const& y)
+	{
+		vec<4, bool, Q> Result(false, false, false, false);
+		for(length_t i = 0; i < x.length(); ++i)
+			Result[i] = x[i] >= y[i];
+		return Result;
+	}
+
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> quatLookAt(vec<3, T, Q> const& direction, vec<3, T, Q> const& up)
+	{
+#		if GLM_CONFIG_CLIP_CONTROL & GLM_CLIP_CONTROL_LH_BIT
+			return quatLookAtLH(direction, up);
+#		else
+			return quatLookAtRH(direction, up);
+# 		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> quatLookAtRH(vec<3, T, Q> const& direction, vec<3, T, Q> const& up)
+	{
+		mat<3, 3, T, Q> Result;
+
+		Result[2] = -direction;
+		vec<3, T, Q> const& Right = cross(up, Result[2]);
+		Result[0] = Right * inversesqrt(max(static_cast<T>(0.00001), dot(Right, Right)));
+		Result[1] = cross(Result[2], Result[0]);
+
+		return quat_cast(Result);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> quatLookAtLH(vec<3, T, Q> const& direction, vec<3, T, Q> const& up)
+	{
+		mat<3, 3, T, Q> Result;
+
+		Result[2] = direction;
+		vec<3, T, Q> const& Right = cross(up, Result[2]);
+		Result[0] = Right * inversesqrt(max(static_cast<T>(0.00001), dot(Right, Right)));
+		Result[1] = cross(Result[2], Result[0]);
+
+		return quat_cast(Result);
+	}
+}//namespace glm
+
+#if GLM_CONFIG_SIMD == GLM_ENABLE
+#	include "quaternion_simd.inl"
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion_simd.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/quaternion_simd.inl
new file mode 100644
index 000000000000..e69de29bb2d1
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/random.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/random.hpp
new file mode 100644
index 000000000000..c6485bf1da38
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/random.hpp
@@ -0,0 +1,82 @@
+/// @ref gtc_random
+/// @file glm/gtc/random.hpp
+///
+/// @see core (dependence)
+/// @see gtx_random (extended)
+///
+/// @defgroup gtc_random GLM_GTC_random
+/// @ingroup gtc
+///
+/// Include <glm/gtc/random.hpp> to use the features of this extension.
+///
+/// Generate random number from various distribution methods.
+
+#pragma once
+
+// Dependency:
+#include "../ext/scalar_int_sized.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_random extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_random
+	/// @{
+
+	/// Generate random numbers in the interval [Min, Max], according a linear distribution
+	///
+	/// @param Min Minimum value included in the sampling
+	/// @param Max Maximum value included in the sampling
+	/// @tparam genType Value type. Currently supported: float or double scalars.
+	/// @see gtc_random
+	template<typename genType>
+	GLM_FUNC_DECL genType linearRand(genType Min, genType Max);
+
+	/// Generate random numbers in the interval [Min, Max], according a linear distribution
+	///
+	/// @param Min Minimum value included in the sampling
+	/// @param Max Maximum value included in the sampling
+	/// @tparam T Value type. Currently supported: float or double.
+	///
+	/// @see gtc_random
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> linearRand(vec<L, T, Q> const& Min, vec<L, T, Q> const& Max);
+
+	/// Generate random numbers in the interval [Min, Max], according a gaussian distribution
+	///
+	/// @see gtc_random
+	template<typename genType>
+	GLM_FUNC_DECL genType gaussRand(genType Mean, genType Deviation);
+
+	/// Generate a random 2D vector which coordinates are regularly distributed on a circle of a given radius
+	///
+	/// @see gtc_random
+	template<typename T>
+	GLM_FUNC_DECL vec<2, T, defaultp> circularRand(T Radius);
+
+	/// Generate a random 3D vector which coordinates are regularly distributed on a sphere of a given radius
+	///
+	/// @see gtc_random
+	template<typename T>
+	GLM_FUNC_DECL vec<3, T, defaultp> sphericalRand(T Radius);
+
+	/// Generate a random 2D vector which coordinates are regularly distributed within the area of a disk of a given radius
+	///
+	/// @see gtc_random
+	template<typename T>
+	GLM_FUNC_DECL vec<2, T, defaultp> diskRand(T Radius);
+
+	/// Generate a random 3D vector which coordinates are regularly distributed within the volume of a ball of a given radius
+	///
+	/// @see gtc_random
+	template<typename T>
+	GLM_FUNC_DECL vec<3, T, defaultp> ballRand(T Radius);
+
+	/// @}
+}//namespace glm
+
+#include "random.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/random.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/random.inl
new file mode 100644
index 000000000000..249ec9f92b48
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/random.inl
@@ -0,0 +1,303 @@
+#include "../geometric.hpp"
+#include "../exponential.hpp"
+#include "../trigonometric.hpp"
+#include "../detail/type_vec1.hpp"
+#include <cstdlib>
+#include <ctime>
+#include <cassert>
+#include <cmath>
+
+namespace glm{
+namespace detail
+{
+	template <length_t L, typename T, qualifier Q>
+	struct compute_rand
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call();
+	};
+
+	template <qualifier P>
+	struct compute_rand<1, uint8, P>
+	{
+		GLM_FUNC_QUALIFIER static vec<1, uint8, P> call()
+		{
+			return vec<1, uint8, P>(
+				static_cast<uint8>(std::rand() % std::numeric_limits<uint8>::max()));
+		}
+	};
+
+	template <qualifier P>
+	struct compute_rand<2, uint8, P>
+	{
+		GLM_FUNC_QUALIFIER static vec<2, uint8, P> call()
+		{
+			return vec<2, uint8, P>(
+				std::rand() % std::numeric_limits<uint8>::max(),
+				std::rand() % std::numeric_limits<uint8>::max());
+		}
+	};
+
+	template <qualifier P>
+	struct compute_rand<3, uint8, P>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, uint8, P> call()
+		{
+			return vec<3, uint8, P>(
+				std::rand() % std::numeric_limits<uint8>::max(),
+				std::rand() % std::numeric_limits<uint8>::max(),
+				std::rand() % std::numeric_limits<uint8>::max());
+		}
+	};
+
+	template <qualifier P>
+	struct compute_rand<4, uint8, P>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, uint8, P> call()
+		{
+			return vec<4, uint8, P>(
+				std::rand() % std::numeric_limits<uint8>::max(),
+				std::rand() % std::numeric_limits<uint8>::max(),
+				std::rand() % std::numeric_limits<uint8>::max(),
+				std::rand() % std::numeric_limits<uint8>::max());
+		}
+	};
+
+	template <length_t L, qualifier Q>
+	struct compute_rand<L, uint16, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint16, Q> call()
+		{
+			return
+				(vec<L, uint16, Q>(compute_rand<L, uint8, Q>::call()) << static_cast<uint16>(8)) |
+				(vec<L, uint16, Q>(compute_rand<L, uint8, Q>::call()) << static_cast<uint16>(0));
+		}
+	};
+
+	template <length_t L, qualifier Q>
+	struct compute_rand<L, uint32, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint32, Q> call()
+		{
+			return
+				(vec<L, uint32, Q>(compute_rand<L, uint16, Q>::call()) << static_cast<uint32>(16)) |
+				(vec<L, uint32, Q>(compute_rand<L, uint16, Q>::call()) << static_cast<uint32>(0));
+		}
+	};
+
+	template <length_t L, qualifier Q>
+	struct compute_rand<L, uint64, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint64, Q> call()
+		{
+			return
+				(vec<L, uint64, Q>(compute_rand<L, uint32, Q>::call()) << static_cast<uint64>(32)) |
+				(vec<L, uint64, Q>(compute_rand<L, uint32, Q>::call()) << static_cast<uint64>(0));
+		}
+	};
+
+	template <length_t L, typename T, qualifier Q>
+	struct compute_linearRand
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& Min, vec<L, T, Q> const& Max);
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, int8, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, int8, Q> call(vec<L, int8, Q> const& Min, vec<L, int8, Q> const& Max)
+		{
+			return (vec<L, int8, Q>(compute_rand<L, uint8, Q>::call() % vec<L, uint8, Q>(Max + static_cast<int8>(1) - Min))) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, uint8, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint8, Q> call(vec<L, uint8, Q> const& Min, vec<L, uint8, Q> const& Max)
+		{
+			return (compute_rand<L, uint8, Q>::call() % (Max + static_cast<uint8>(1) - Min)) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, int16, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, int16, Q> call(vec<L, int16, Q> const& Min, vec<L, int16, Q> const& Max)
+		{
+			return (vec<L, int16, Q>(compute_rand<L, uint16, Q>::call() % vec<L, uint16, Q>(Max + static_cast<int16>(1) - Min))) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, uint16, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint16, Q> call(vec<L, uint16, Q> const& Min, vec<L, uint16, Q> const& Max)
+		{
+			return (compute_rand<L, uint16, Q>::call() % (Max + static_cast<uint16>(1) - Min)) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, int32, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, int32, Q> call(vec<L, int32, Q> const& Min, vec<L, int32, Q> const& Max)
+		{
+			return (vec<L, int32, Q>(compute_rand<L, uint32, Q>::call() % vec<L, uint32, Q>(Max + static_cast<int32>(1) - Min))) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, uint32, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint32, Q> call(vec<L, uint32, Q> const& Min, vec<L, uint32, Q> const& Max)
+		{
+			return (compute_rand<L, uint32, Q>::call() % (Max + static_cast<uint32>(1) - Min)) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, int64, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, int64, Q> call(vec<L, int64, Q> const& Min, vec<L, int64, Q> const& Max)
+		{
+			return (vec<L, int64, Q>(compute_rand<L, uint64, Q>::call() % vec<L, uint64, Q>(Max + static_cast<int64>(1) - Min))) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, uint64, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, uint64, Q> call(vec<L, uint64, Q> const& Min, vec<L, uint64, Q> const& Max)
+		{
+			return (compute_rand<L, uint64, Q>::call() % (Max + static_cast<uint64>(1) - Min)) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, float, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, float, Q> call(vec<L, float, Q> const& Min, vec<L, float, Q> const& Max)
+		{
+			return vec<L, float, Q>(compute_rand<L, uint32, Q>::call()) / static_cast<float>(std::numeric_limits<uint32>::max()) * (Max - Min) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, double, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, double, Q> call(vec<L, double, Q> const& Min, vec<L, double, Q> const& Max)
+		{
+			return vec<L, double, Q>(compute_rand<L, uint64, Q>::call()) / static_cast<double>(std::numeric_limits<uint64>::max()) * (Max - Min) + Min;
+		}
+	};
+
+	template<length_t L, qualifier Q>
+	struct compute_linearRand<L, long double, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, long double, Q> call(vec<L, long double, Q> const& Min, vec<L, long double, Q> const& Max)
+		{
+			return vec<L, long double, Q>(compute_rand<L, uint64, Q>::call()) / static_cast<long double>(std::numeric_limits<uint64>::max()) * (Max - Min) + Min;
+		}
+	};
+}//namespace detail
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType linearRand(genType Min, genType Max)
+	{
+		return detail::compute_linearRand<1, genType, highp>::call(
+			vec<1, genType, highp>(Min),
+			vec<1, genType, highp>(Max)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> linearRand(vec<L, T, Q> const& Min, vec<L, T, Q> const& Max)
+	{
+		return detail::compute_linearRand<L, T, Q>::call(Min, Max);
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType gaussRand(genType Mean, genType Deviation)
+	{
+		genType w, x1, x2;
+
+		do
+		{
+			x1 = linearRand(genType(-1), genType(1));
+			x2 = linearRand(genType(-1), genType(1));
+
+			w = x1 * x1 + x2 * x2;
+		} while(w > genType(1));
+
+		return static_cast<genType>(x2 * Deviation * Deviation * sqrt((genType(-2) * log(w)) / w) + Mean);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> gaussRand(vec<L, T, Q> const& Mean, vec<L, T, Q> const& Deviation)
+	{
+		return detail::functor2<vec, L, T, Q>::call(gaussRand, Mean, Deviation);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<2, T, defaultp> diskRand(T Radius)
+	{
+		assert(Radius > static_cast<T>(0));
+
+		vec<2, T, defaultp> Result(T(0));
+		T LenRadius(T(0));
+
+		do
+		{
+			Result = linearRand(
+				vec<2, T, defaultp>(-Radius),
+				vec<2, T, defaultp>(Radius));
+			LenRadius = length(Result);
+		}
+		while(LenRadius > Radius);
+
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<3, T, defaultp> ballRand(T Radius)
+	{
+		assert(Radius > static_cast<T>(0));
+
+		vec<3, T, defaultp> Result(T(0));
+		T LenRadius(T(0));
+
+		do
+		{
+			Result = linearRand(
+				vec<3, T, defaultp>(-Radius),
+				vec<3, T, defaultp>(Radius));
+			LenRadius = length(Result);
+		}
+		while(LenRadius > Radius);
+
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<2, T, defaultp> circularRand(T Radius)
+	{
+		assert(Radius > static_cast<T>(0));
+
+		T a = linearRand(T(0), static_cast<T>(6.283185307179586476925286766559));
+		return vec<2, T, defaultp>(glm::cos(a), glm::sin(a)) * Radius;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<3, T, defaultp> sphericalRand(T Radius)
+	{
+		assert(Radius > static_cast<T>(0));
+
+		T theta = linearRand(T(0), T(6.283185307179586476925286766559f));
+		T phi = std::acos(linearRand(T(-1.0f), T(1.0f)));
+
+		T x = std::sin(phi) * std::cos(theta);
+		T y = std::sin(phi) * std::sin(theta);
+		T z = std::cos(phi);
+
+		return vec<3, T, defaultp>(x, y, z) * Radius;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/reciprocal.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/reciprocal.hpp
new file mode 100644
index 000000000000..4d0fc91ca657
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/reciprocal.hpp
@@ -0,0 +1,24 @@
+/// @ref gtc_reciprocal
+/// @file glm/gtc/reciprocal.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_reciprocal GLM_GTC_reciprocal
+/// @ingroup gtc
+///
+/// Include <glm/gtc/reciprocal.hpp> to use the features of this extension.
+///
+/// Define secant, cosecant and cotangent functions.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_reciprocal extension included")
+#endif
+
+#include "../ext/scalar_reciprocal.hpp"
+#include "../ext/vector_reciprocal.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/round.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/round.hpp
new file mode 100644
index 000000000000..56edbbca30b5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/round.hpp
@@ -0,0 +1,160 @@
+/// @ref gtc_round
+/// @file glm/gtc/round.hpp
+///
+/// @see core (dependence)
+/// @see gtc_round (dependence)
+///
+/// @defgroup gtc_round GLM_GTC_round
+/// @ingroup gtc
+///
+/// Include <glm/gtc/round.hpp> to use the features of this extension.
+///
+/// Rounding value to specific boundings
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/_vectorize.hpp"
+#include "../vector_relational.hpp"
+#include "../common.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_round extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_round
+	/// @{
+
+	/// Return the power of two number which value is just higher the input value,
+	/// round up to a power of two.
+	///
+	/// @see gtc_round
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType ceilPowerOfTwo(genIUType v);
+
+	/// Return the power of two number which value is just higher the input value,
+	/// round up to a power of two.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_round
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> ceilPowerOfTwo(vec<L, T, Q> const& v);
+
+	/// Return the power of two number which value is just lower the input value,
+	/// round down to a power of two.
+	///
+	/// @see gtc_round
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType floorPowerOfTwo(genIUType v);
+
+	/// Return the power of two number which value is just lower the input value,
+	/// round down to a power of two.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_round
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> floorPowerOfTwo(vec<L, T, Q> const& v);
+
+	/// Return the power of two number which value is the closet to the input value.
+	///
+	/// @see gtc_round
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType roundPowerOfTwo(genIUType v);
+
+	/// Return the power of two number which value is the closet to the input value.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_round
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> roundPowerOfTwo(vec<L, T, Q> const& v);
+
+	/// Higher multiple number of Source.
+	///
+	/// @tparam genType Floating-point or integer scalar or vector types.
+	///
+	/// @param v Source value to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see gtc_round
+	template<typename genType>
+	GLM_FUNC_DECL genType ceilMultiple(genType v, genType Multiple);
+
+	/// Higher multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see gtc_round
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> ceilMultiple(vec<L, T, Q> const& v, vec<L, T, Q> const& Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam genType Floating-point or integer scalar or vector types.
+	///
+	/// @param v Source value to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see gtc_round
+	template<typename genType>
+	GLM_FUNC_DECL genType floorMultiple(genType v, genType Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see gtc_round
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> floorMultiple(vec<L, T, Q> const& v, vec<L, T, Q> const& Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam genType Floating-point or integer scalar or vector types.
+	///
+	/// @param v Source value to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see gtc_round
+	template<typename genType>
+	GLM_FUNC_DECL genType roundMultiple(genType v, genType Multiple);
+
+	/// Lower multiple number of Source.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @param v Source values to which is applied the function
+	/// @param Multiple Must be a null or positive value
+	///
+	/// @see gtc_round
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> roundMultiple(vec<L, T, Q> const& v, vec<L, T, Q> const& Multiple);
+
+	/// @}
+} //namespace glm
+
+#include "round.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/round.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/round.inl
new file mode 100644
index 000000000000..48411e41dc34
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/round.inl
@@ -0,0 +1,155 @@
+/// @ref gtc_round
+
+#include "../integer.hpp"
+#include "../ext/vector_integer.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<bool is_float, bool is_signed>
+	struct compute_roundMultiple {};
+
+	template<>
+	struct compute_roundMultiple<true, true>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if (Source >= genType(0))
+				return Source - std::fmod(Source, Multiple);
+			else
+			{
+				genType Tmp = Source + genType(1);
+				return Tmp - std::fmod(Tmp, Multiple) - Multiple;
+			}
+		}
+	};
+
+	template<>
+	struct compute_roundMultiple<false, false>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if (Source >= genType(0))
+				return Source - Source % Multiple;
+			else
+			{
+				genType Tmp = Source + genType(1);
+				return Tmp - Tmp % Multiple - Multiple;
+			}
+		}
+	};
+
+	template<>
+	struct compute_roundMultiple<false, true>
+	{
+		template<typename genType>
+		GLM_FUNC_QUALIFIER static genType call(genType Source, genType Multiple)
+		{
+			if (Source >= genType(0))
+				return Source - Source % Multiple;
+			else
+			{
+				genType Tmp = Source + genType(1);
+				return Tmp - Tmp % Multiple - Multiple;
+			}
+		}
+	};
+}//namespace detail
+
+	//////////////////
+	// ceilPowerOfTwo
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType ceilPowerOfTwo(genType value)
+	{
+		return detail::compute_ceilPowerOfTwo<1, genType, defaultp, std::numeric_limits<genType>::is_signed>::call(vec<1, genType, defaultp>(value)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> ceilPowerOfTwo(vec<L, T, Q> const& v)
+	{
+		return detail::compute_ceilPowerOfTwo<L, T, Q, std::numeric_limits<T>::is_signed>::call(v);
+	}
+
+	///////////////////
+	// floorPowerOfTwo
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType floorPowerOfTwo(genType value)
+	{
+		return isPowerOfTwo(value) ? value : static_cast<genType>(1) << findMSB(value);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> floorPowerOfTwo(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(floorPowerOfTwo, v);
+	}
+
+	///////////////////
+	// roundPowerOfTwo
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType roundPowerOfTwo(genIUType value)
+	{
+		if(isPowerOfTwo(value))
+			return value;
+
+		genIUType const prev = static_cast<genIUType>(1) << findMSB(value);
+		genIUType const next = prev << static_cast<genIUType>(1);
+		return (next - value) < (value - prev) ? next : prev;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> roundPowerOfTwo(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(roundPowerOfTwo, v);
+	}
+
+	//////////////////////
+	// ceilMultiple
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType ceilMultiple(genType Source, genType Multiple)
+	{
+		return detail::compute_ceilMultiple<std::numeric_limits<genType>::is_iec559, std::numeric_limits<genType>::is_signed>::call(Source, Multiple);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> ceilMultiple(vec<L, T, Q> const& Source, vec<L, T, Q> const& Multiple)
+	{
+		return detail::functor2<vec, L, T, Q>::call(ceilMultiple, Source, Multiple);
+	}
+
+	//////////////////////
+	// floorMultiple
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType floorMultiple(genType Source, genType Multiple)
+	{
+		return detail::compute_floorMultiple<std::numeric_limits<genType>::is_iec559, std::numeric_limits<genType>::is_signed>::call(Source, Multiple);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> floorMultiple(vec<L, T, Q> const& Source, vec<L, T, Q> const& Multiple)
+	{
+		return detail::functor2<vec, L, T, Q>::call(floorMultiple, Source, Multiple);
+	}
+
+	//////////////////////
+	// roundMultiple
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType roundMultiple(genType Source, genType Multiple)
+	{
+		return detail::compute_roundMultiple<std::numeric_limits<genType>::is_iec559, std::numeric_limits<genType>::is_signed>::call(Source, Multiple);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> roundMultiple(vec<L, T, Q> const& Source, vec<L, T, Q> const& Multiple)
+	{
+		return detail::functor2<vec, L, T, Q>::call(roundMultiple, Source, Multiple);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/type_aligned.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_aligned.hpp
new file mode 100644
index 000000000000..5403abf67525
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_aligned.hpp
@@ -0,0 +1,1315 @@
+/// @ref gtc_type_aligned
+/// @file glm/gtc/type_aligned.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_type_aligned GLM_GTC_type_aligned
+/// @ingroup gtc
+///
+/// Include <glm/gtc/type_aligned.hpp> to use the features of this extension.
+///
+/// Aligned types allowing SIMD optimizations of vectors and matrices types
+
+#pragma once
+
+#if (GLM_CONFIG_ALIGNED_GENTYPES == GLM_DISABLE)
+#	error "GLM: Aligned gentypes require to enable C++ language extensions. Define GLM_FORCE_ALIGNED_GENTYPES before including GLM headers to use aligned types."
+#endif
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+# pragma message("GLM: GLM_GTC_type_aligned extension included")
+#endif
+
+#include "../mat4x4.hpp"
+#include "../mat4x3.hpp"
+#include "../mat4x2.hpp"
+#include "../mat3x4.hpp"
+#include "../mat3x3.hpp"
+#include "../mat3x2.hpp"
+#include "../mat2x4.hpp"
+#include "../mat2x3.hpp"
+#include "../mat2x2.hpp"
+#include "../gtc/vec1.hpp"
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+
+namespace glm
+{
+	/// @addtogroup gtc_type_aligned
+	/// @{
+
+	// -- *vec1 --
+
+	/// 1 component vector aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<1, float, aligned_highp>	aligned_highp_vec1;
+
+	/// 1 component vector aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<1, float, aligned_mediump>	aligned_mediump_vec1;
+
+	/// 1 component vector aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<1, float, aligned_lowp>		aligned_lowp_vec1;
+
+	/// 1 component vector aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<1, double, aligned_highp>	aligned_highp_dvec1;
+
+	/// 1 component vector aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<1, double, aligned_mediump>	aligned_mediump_dvec1;
+
+	/// 1 component vector aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<1, double, aligned_lowp>	aligned_lowp_dvec1;
+
+	/// 1 component vector aligned in memory of signed integer numbers.
+	typedef vec<1, int, aligned_highp>		aligned_highp_ivec1;
+
+	/// 1 component vector aligned in memory of signed integer numbers.
+	typedef vec<1, int, aligned_mediump>	aligned_mediump_ivec1;
+
+	/// 1 component vector aligned in memory of signed integer numbers.
+	typedef vec<1, int, aligned_lowp>		aligned_lowp_ivec1;
+
+	/// 1 component vector aligned in memory of unsigned integer numbers.
+	typedef vec<1, uint, aligned_highp>		aligned_highp_uvec1;
+
+	/// 1 component vector aligned in memory of unsigned integer numbers.
+	typedef vec<1, uint, aligned_mediump>	aligned_mediump_uvec1;
+
+	/// 1 component vector aligned in memory of unsigned integer numbers.
+	typedef vec<1, uint, aligned_lowp>		aligned_lowp_uvec1;
+
+	/// 1 component vector aligned in memory of bool values.
+	typedef vec<1, bool, aligned_highp>		aligned_highp_bvec1;
+
+	/// 1 component vector aligned in memory of bool values.
+	typedef vec<1, bool, aligned_mediump>	aligned_mediump_bvec1;
+
+	/// 1 component vector aligned in memory of bool values.
+	typedef vec<1, bool, aligned_lowp>		aligned_lowp_bvec1;
+
+	/// 1 component vector tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<1, float, packed_highp>		packed_highp_vec1;
+
+	/// 1 component vector tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<1, float, packed_mediump>	packed_mediump_vec1;
+
+	/// 1 component vector tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<1, float, packed_lowp>		packed_lowp_vec1;
+
+	/// 1 component vector tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<1, double, packed_highp>	packed_highp_dvec1;
+
+	/// 1 component vector tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<1, double, packed_mediump>	packed_mediump_dvec1;
+
+	/// 1 component vector tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<1, double, packed_lowp>		packed_lowp_dvec1;
+
+	/// 1 component vector tightly packed in memory of signed integer numbers.
+	typedef vec<1, int, packed_highp>		packed_highp_ivec1;
+
+	/// 1 component vector tightly packed in memory of signed integer numbers.
+	typedef vec<1, int, packed_mediump>		packed_mediump_ivec1;
+
+	/// 1 component vector tightly packed in memory of signed integer numbers.
+	typedef vec<1, int, packed_lowp>		packed_lowp_ivec1;
+
+	/// 1 component vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<1, uint, packed_highp>		packed_highp_uvec1;
+
+	/// 1 component vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<1, uint, packed_mediump>	packed_mediump_uvec1;
+
+	/// 1 component vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<1, uint, packed_lowp>		packed_lowp_uvec1;
+
+	/// 1 component vector tightly packed in memory of bool values.
+	typedef vec<1, bool, packed_highp>		packed_highp_bvec1;
+
+	/// 1 component vector tightly packed in memory of bool values.
+	typedef vec<1, bool, packed_mediump>	packed_mediump_bvec1;
+
+	/// 1 component vector tightly packed in memory of bool values.
+	typedef vec<1, bool, packed_lowp>		packed_lowp_bvec1;
+
+	// -- *vec2 --
+
+	/// 2 components vector aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<2, float, aligned_highp>	aligned_highp_vec2;
+
+	/// 2 components vector aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<2, float, aligned_mediump>	aligned_mediump_vec2;
+
+	/// 2 components vector aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<2, float, aligned_lowp>		aligned_lowp_vec2;
+
+	/// 2 components vector aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<2, double, aligned_highp>	aligned_highp_dvec2;
+
+	/// 2 components vector aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<2, double, aligned_mediump>	aligned_mediump_dvec2;
+
+	/// 2 components vector aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<2, double, aligned_lowp>	aligned_lowp_dvec2;
+
+	/// 2 components vector aligned in memory of signed integer numbers.
+	typedef vec<2, int, aligned_highp>		aligned_highp_ivec2;
+
+	/// 2 components vector aligned in memory of signed integer numbers.
+	typedef vec<2, int, aligned_mediump>	aligned_mediump_ivec2;
+
+	/// 2 components vector aligned in memory of signed integer numbers.
+	typedef vec<2, int, aligned_lowp>		aligned_lowp_ivec2;
+
+	/// 2 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<2, uint, aligned_highp>		aligned_highp_uvec2;
+
+	/// 2 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<2, uint, aligned_mediump>	aligned_mediump_uvec2;
+
+	/// 2 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<2, uint, aligned_lowp>		aligned_lowp_uvec2;
+
+	/// 2 components vector aligned in memory of bool values.
+	typedef vec<2, bool, aligned_highp>		aligned_highp_bvec2;
+
+	/// 2 components vector aligned in memory of bool values.
+	typedef vec<2, bool, aligned_mediump>	aligned_mediump_bvec2;
+
+	/// 2 components vector aligned in memory of bool values.
+	typedef vec<2, bool, aligned_lowp>		aligned_lowp_bvec2;
+
+	/// 2 components vector tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<2, float, packed_highp>		packed_highp_vec2;
+
+	/// 2 components vector tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<2, float, packed_mediump>	packed_mediump_vec2;
+
+	/// 2 components vector tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<2, float, packed_lowp>		packed_lowp_vec2;
+
+	/// 2 components vector tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<2, double, packed_highp>	packed_highp_dvec2;
+
+	/// 2 components vector tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<2, double, packed_mediump>	packed_mediump_dvec2;
+
+	/// 2 components vector tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<2, double, packed_lowp>		packed_lowp_dvec2;
+
+	/// 2 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<2, int, packed_highp>		packed_highp_ivec2;
+
+	/// 2 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<2, int, packed_mediump>		packed_mediump_ivec2;
+
+	/// 2 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<2, int, packed_lowp>		packed_lowp_ivec2;
+
+	/// 2 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<2, uint, packed_highp>		packed_highp_uvec2;
+
+	/// 2 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<2, uint, packed_mediump>	packed_mediump_uvec2;
+
+	/// 2 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<2, uint, packed_lowp>		packed_lowp_uvec2;
+
+	/// 2 components vector tightly packed in memory of bool values.
+	typedef vec<2, bool, packed_highp>		packed_highp_bvec2;
+
+	/// 2 components vector tightly packed in memory of bool values.
+	typedef vec<2, bool, packed_mediump>	packed_mediump_bvec2;
+
+	/// 2 components vector tightly packed in memory of bool values.
+	typedef vec<2, bool, packed_lowp>		packed_lowp_bvec2;
+
+	// -- *vec3 --
+
+	/// 3 components vector aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<3, float, aligned_highp>	aligned_highp_vec3;
+
+	/// 3 components vector aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<3, float, aligned_mediump>	aligned_mediump_vec3;
+
+	/// 3 components vector aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<3, float, aligned_lowp>		aligned_lowp_vec3;
+
+	/// 3 components vector aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<3, double, aligned_highp>	aligned_highp_dvec3;
+
+	/// 3 components vector aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<3, double, aligned_mediump>	aligned_mediump_dvec3;
+
+	/// 3 components vector aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<3, double, aligned_lowp>	aligned_lowp_dvec3;
+
+	/// 3 components vector aligned in memory of signed integer numbers.
+	typedef vec<3, int, aligned_highp>		aligned_highp_ivec3;
+
+	/// 3 components vector aligned in memory of signed integer numbers.
+	typedef vec<3, int, aligned_mediump>	aligned_mediump_ivec3;
+
+	/// 3 components vector aligned in memory of signed integer numbers.
+	typedef vec<3, int, aligned_lowp>		aligned_lowp_ivec3;
+
+	/// 3 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<3, uint, aligned_highp>		aligned_highp_uvec3;
+
+	/// 3 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<3, uint, aligned_mediump>	aligned_mediump_uvec3;
+
+	/// 3 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<3, uint, aligned_lowp>		aligned_lowp_uvec3;
+
+	/// 3 components vector aligned in memory of bool values.
+	typedef vec<3, bool, aligned_highp>		aligned_highp_bvec3;
+
+	/// 3 components vector aligned in memory of bool values.
+	typedef vec<3, bool, aligned_mediump>	aligned_mediump_bvec3;
+
+	/// 3 components vector aligned in memory of bool values.
+	typedef vec<3, bool, aligned_lowp>		aligned_lowp_bvec3;
+
+	/// 3 components vector tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<3, float, packed_highp>		packed_highp_vec3;
+
+	/// 3 components vector tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<3, float, packed_mediump>	packed_mediump_vec3;
+
+	/// 3 components vector tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<3, float, packed_lowp>		packed_lowp_vec3;
+
+	/// 3 components vector tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<3, double, packed_highp>	packed_highp_dvec3;
+
+	/// 3 components vector tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<3, double, packed_mediump>	packed_mediump_dvec3;
+
+	/// 3 components vector tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<3, double, packed_lowp>		packed_lowp_dvec3;
+
+	/// 3 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<3, int, packed_highp>		packed_highp_ivec3;
+
+	/// 3 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<3, int, packed_mediump>		packed_mediump_ivec3;
+
+	/// 3 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<3, int, packed_lowp>		packed_lowp_ivec3;
+
+	/// 3 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<3, uint, packed_highp>		packed_highp_uvec3;
+
+	/// 3 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<3, uint, packed_mediump>	packed_mediump_uvec3;
+
+	/// 3 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<3, uint, packed_lowp>		packed_lowp_uvec3;
+
+	/// 3 components vector tightly packed in memory of bool values.
+	typedef vec<3, bool, packed_highp>		packed_highp_bvec3;
+
+	/// 3 components vector tightly packed in memory of bool values.
+	typedef vec<3, bool, packed_mediump>	packed_mediump_bvec3;
+
+	/// 3 components vector tightly packed in memory of bool values.
+	typedef vec<3, bool, packed_lowp>		packed_lowp_bvec3;
+
+	// -- *vec4 --
+
+	/// 4 components vector aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<4, float, aligned_highp>	aligned_highp_vec4;
+
+	/// 4 components vector aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<4, float, aligned_mediump>	aligned_mediump_vec4;
+
+	/// 4 components vector aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<4, float, aligned_lowp>		aligned_lowp_vec4;
+
+	/// 4 components vector aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<4, double, aligned_highp>	aligned_highp_dvec4;
+
+	/// 4 components vector aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<4, double, aligned_mediump>	aligned_mediump_dvec4;
+
+	/// 4 components vector aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<4, double, aligned_lowp>	aligned_lowp_dvec4;
+
+	/// 4 components vector aligned in memory of signed integer numbers.
+	typedef vec<4, int, aligned_highp>		aligned_highp_ivec4;
+
+	/// 4 components vector aligned in memory of signed integer numbers.
+	typedef vec<4, int, aligned_mediump>	aligned_mediump_ivec4;
+
+	/// 4 components vector aligned in memory of signed integer numbers.
+	typedef vec<4, int, aligned_lowp>		aligned_lowp_ivec4;
+
+	/// 4 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<4, uint, aligned_highp>		aligned_highp_uvec4;
+
+	/// 4 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<4, uint, aligned_mediump>	aligned_mediump_uvec4;
+
+	/// 4 components vector aligned in memory of unsigned integer numbers.
+	typedef vec<4, uint, aligned_lowp>		aligned_lowp_uvec4;
+
+	/// 4 components vector aligned in memory of bool values.
+	typedef vec<4, bool, aligned_highp>		aligned_highp_bvec4;
+
+	/// 4 components vector aligned in memory of bool values.
+	typedef vec<4, bool, aligned_mediump>	aligned_mediump_bvec4;
+
+	/// 4 components vector aligned in memory of bool values.
+	typedef vec<4, bool, aligned_lowp>		aligned_lowp_bvec4;
+
+	/// 4 components vector tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<4, float, packed_highp>		packed_highp_vec4;
+
+	/// 4 components vector tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<4, float, packed_mediump>	packed_mediump_vec4;
+
+	/// 4 components vector tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<4, float, packed_lowp>		packed_lowp_vec4;
+
+	/// 4 components vector tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef vec<4, double, packed_highp>	packed_highp_dvec4;
+
+	/// 4 components vector tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef vec<4, double, packed_mediump>	packed_mediump_dvec4;
+
+	/// 4 components vector tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef vec<4, double, packed_lowp>		packed_lowp_dvec4;
+
+	/// 4 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<4, int, packed_highp>		packed_highp_ivec4;
+
+	/// 4 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<4, int, packed_mediump>		packed_mediump_ivec4;
+
+	/// 4 components vector tightly packed in memory of signed integer numbers.
+	typedef vec<4, int, packed_lowp>		packed_lowp_ivec4;
+
+	/// 4 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<4, uint, packed_highp>		packed_highp_uvec4;
+
+	/// 4 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<4, uint, packed_mediump>	packed_mediump_uvec4;
+
+	/// 4 components vector tightly packed in memory of unsigned integer numbers.
+	typedef vec<4, uint, packed_lowp>		packed_lowp_uvec4;
+
+	/// 4 components vector tightly packed in memory of bool values.
+	typedef vec<4, bool, packed_highp>		packed_highp_bvec4;
+
+	/// 4 components vector tightly packed in memory of bool values.
+	typedef vec<4, bool, packed_mediump>	packed_mediump_bvec4;
+
+	/// 4 components vector tightly packed in memory of bool values.
+	typedef vec<4, bool, packed_lowp>		packed_lowp_bvec4;
+
+	// -- *mat2 --
+
+	/// 2 by 2 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, aligned_highp>		aligned_highp_mat2;
+
+	/// 2 by 2 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, aligned_mediump>	aligned_mediump_mat2;
+
+	/// 2 by 2 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, aligned_lowp>		aligned_lowp_mat2;
+
+	/// 2 by 2 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, aligned_highp>	aligned_highp_dmat2;
+
+	/// 2 by 2 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, aligned_mediump>	aligned_mediump_dmat2;
+
+	/// 2 by 2 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, aligned_lowp>		aligned_lowp_dmat2;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, packed_highp>		packed_highp_mat2;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, packed_mediump>	packed_mediump_mat2;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, packed_lowp>		packed_lowp_mat2;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, packed_highp>		packed_highp_dmat2;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, packed_mediump>	packed_mediump_dmat2;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, packed_lowp>		packed_lowp_dmat2;
+
+	// -- *mat3 --
+
+	/// 3 by 3 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, aligned_highp>		aligned_highp_mat3;
+
+	/// 3 by 3 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, aligned_mediump>	aligned_mediump_mat3;
+
+	/// 3 by 3 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, aligned_lowp>		aligned_lowp_mat3;
+
+	/// 3 by 3 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, aligned_highp>	aligned_highp_dmat3;
+
+	/// 3 by 3 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, aligned_mediump>	aligned_mediump_dmat3;
+
+	/// 3 by 3 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, aligned_lowp>		aligned_lowp_dmat3;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, packed_highp>		packed_highp_mat3;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, packed_mediump>	packed_mediump_mat3;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, packed_lowp>		packed_lowp_mat3;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, packed_highp>		packed_highp_dmat3;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, packed_mediump>	packed_mediump_dmat3;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, packed_lowp>		packed_lowp_dmat3;
+
+	// -- *mat4 --
+
+	/// 4 by 4 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, aligned_highp>		aligned_highp_mat4;
+
+	/// 4 by 4 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, aligned_mediump>	aligned_mediump_mat4;
+
+	/// 4 by 4 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, aligned_lowp>		aligned_lowp_mat4;
+
+	/// 4 by 4 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, aligned_highp>	aligned_highp_dmat4;
+
+	/// 4 by 4 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, aligned_mediump>	aligned_mediump_dmat4;
+
+	/// 4 by 4 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, aligned_lowp>		aligned_lowp_dmat4;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, packed_highp>		packed_highp_mat4;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, packed_mediump>	packed_mediump_mat4;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, packed_lowp>		packed_lowp_mat4;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, packed_highp>		packed_highp_dmat4;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, packed_mediump>	packed_mediump_dmat4;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, packed_lowp>		packed_lowp_dmat4;
+
+	// -- *mat2x2 --
+
+	/// 2 by 2 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, aligned_highp>		aligned_highp_mat2x2;
+
+	/// 2 by 2 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, aligned_mediump>	aligned_mediump_mat2x2;
+
+	/// 2 by 2 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, aligned_lowp>		aligned_lowp_mat2x2;
+
+	/// 2 by 2 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, aligned_highp>	aligned_highp_dmat2x2;
+
+	/// 2 by 2 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, aligned_mediump>	aligned_mediump_dmat2x2;
+
+	/// 2 by 2 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, aligned_lowp>		aligned_lowp_dmat2x2;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, packed_highp>		packed_highp_mat2x2;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, packed_mediump>	packed_mediump_mat2x2;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, float, packed_lowp>		packed_lowp_mat2x2;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, packed_highp>		packed_highp_dmat2x2;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, packed_mediump>	packed_mediump_dmat2x2;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 2, double, packed_lowp>		packed_lowp_dmat2x2;
+
+	// -- *mat2x3 --
+
+	/// 2 by 3 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 3, float, aligned_highp>		aligned_highp_mat2x3;
+
+	/// 2 by 3 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 3, float, aligned_mediump>	aligned_mediump_mat2x3;
+
+	/// 2 by 3 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 3, float, aligned_lowp>		aligned_lowp_mat2x3;
+
+	/// 2 by 3 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 3, double, aligned_highp>	aligned_highp_dmat2x3;
+
+	/// 2 by 3 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 3, double, aligned_mediump>	aligned_mediump_dmat2x3;
+
+	/// 2 by 3 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 3, double, aligned_lowp>		aligned_lowp_dmat2x3;
+
+	/// 2 by 3 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 3, float, packed_highp>		packed_highp_mat2x3;
+
+	/// 2 by 3 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 3, float, packed_mediump>	packed_mediump_mat2x3;
+
+	/// 2 by 3 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 3, float, packed_lowp>		packed_lowp_mat2x3;
+
+	/// 2 by 3 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 3, double, packed_highp>		packed_highp_dmat2x3;
+
+	/// 2 by 3 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 3, double, packed_mediump>	packed_mediump_dmat2x3;
+
+	/// 2 by 3 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 3, double, packed_lowp>		packed_lowp_dmat2x3;
+
+	// -- *mat2x4 --
+
+	/// 2 by 4 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 4, float, aligned_highp>		aligned_highp_mat2x4;
+
+	/// 2 by 4 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 4, float, aligned_mediump>	aligned_mediump_mat2x4;
+
+	/// 2 by 4 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 4, float, aligned_lowp>		aligned_lowp_mat2x4;
+
+	/// 2 by 4 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 4, double, aligned_highp>	aligned_highp_dmat2x4;
+
+	/// 2 by 4 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 4, double, aligned_mediump>	aligned_mediump_dmat2x4;
+
+	/// 2 by 4 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 4, double, aligned_lowp>		aligned_lowp_dmat2x4;
+
+	/// 2 by 4 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 4, float, packed_highp>		packed_highp_mat2x4;
+
+	/// 2 by 4 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 4, float, packed_mediump>	packed_mediump_mat2x4;
+
+	/// 2 by 4 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 4, float, packed_lowp>		packed_lowp_mat2x4;
+
+	/// 2 by 4 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<2, 4, double, packed_highp>		packed_highp_dmat2x4;
+
+	/// 2 by 4 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<2, 4, double, packed_mediump>	packed_mediump_dmat2x4;
+
+	/// 2 by 4 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<2, 4, double, packed_lowp>		packed_lowp_dmat2x4;
+
+	// -- *mat3x2 --
+
+	/// 3 by 2 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 2, float, aligned_highp>		aligned_highp_mat3x2;
+
+	/// 3 by 2 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 2, float, aligned_mediump>	aligned_mediump_mat3x2;
+
+	/// 3 by 2 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 2, float, aligned_lowp>		aligned_lowp_mat3x2;
+
+	/// 3 by 2 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 2, double, aligned_highp>	aligned_highp_dmat3x2;
+
+	/// 3 by 2 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 2, double, aligned_mediump>	aligned_mediump_dmat3x2;
+
+	/// 3 by 2 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 2, double, aligned_lowp>		aligned_lowp_dmat3x2;
+
+	/// 3 by 2 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 2, float, packed_highp>		packed_highp_mat3x2;
+
+	/// 3 by 2 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 2, float, packed_mediump>	packed_mediump_mat3x2;
+
+	/// 3 by 2 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 2, float, packed_lowp>		packed_lowp_mat3x2;
+
+	/// 3 by 2 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 2, double, packed_highp>		packed_highp_dmat3x2;
+
+	/// 3 by 2 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 2, double, packed_mediump>	packed_mediump_dmat3x2;
+
+	/// 3 by 2 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 2, double, packed_lowp>		packed_lowp_dmat3x2;
+
+	// -- *mat3x3 --
+
+	/// 3 by 3 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, aligned_highp>		aligned_highp_mat3x3;
+
+	/// 3 by 3 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, aligned_mediump>	aligned_mediump_mat3x3;
+
+	/// 3 by 3 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, aligned_lowp>		aligned_lowp_mat3x3;
+
+	/// 3 by 3 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, aligned_highp>	aligned_highp_dmat3x3;
+
+	/// 3 by 3 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, aligned_mediump>	aligned_mediump_dmat3x3;
+
+	/// 3 by 3 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, aligned_lowp>		aligned_lowp_dmat3x3;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, packed_highp>		packed_highp_mat3x3;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, packed_mediump>	packed_mediump_mat3x3;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, float, packed_lowp>		packed_lowp_mat3x3;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, packed_highp>		packed_highp_dmat3x3;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, packed_mediump>	packed_mediump_dmat3x3;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 3, double, packed_lowp>		packed_lowp_dmat3x3;
+
+	// -- *mat3x4 --
+
+	/// 3 by 4 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 4, float, aligned_highp>		aligned_highp_mat3x4;
+
+	/// 3 by 4 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 4, float, aligned_mediump>	aligned_mediump_mat3x4;
+
+	/// 3 by 4 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 4, float, aligned_lowp>		aligned_lowp_mat3x4;
+
+	/// 3 by 4 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 4, double, aligned_highp>	aligned_highp_dmat3x4;
+
+	/// 3 by 4 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 4, double, aligned_mediump>	aligned_mediump_dmat3x4;
+
+	/// 3 by 4 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 4, double, aligned_lowp>		aligned_lowp_dmat3x4;
+
+	/// 3 by 4 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 4, float, packed_highp>		packed_highp_mat3x4;
+
+	/// 3 by 4 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 4, float, packed_mediump>	packed_mediump_mat3x4;
+
+	/// 3 by 4 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 4, float, packed_lowp>		packed_lowp_mat3x4;
+
+	/// 3 by 4 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<3, 4, double, packed_highp>		packed_highp_dmat3x4;
+
+	/// 3 by 4 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<3, 4, double, packed_mediump>	packed_mediump_dmat3x4;
+
+	/// 3 by 4 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<3, 4, double, packed_lowp>		packed_lowp_dmat3x4;
+
+	// -- *mat4x2 --
+
+	/// 4 by 2 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 2, float, aligned_highp>		aligned_highp_mat4x2;
+
+	/// 4 by 2 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 2, float, aligned_mediump>	aligned_mediump_mat4x2;
+
+	/// 4 by 2 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 2, float, aligned_lowp>		aligned_lowp_mat4x2;
+
+	/// 4 by 2 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 2, double, aligned_highp>	aligned_highp_dmat4x2;
+
+	/// 4 by 2 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 2, double, aligned_mediump>	aligned_mediump_dmat4x2;
+
+	/// 4 by 2 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 2, double, aligned_lowp>		aligned_lowp_dmat4x2;
+
+	/// 4 by 2 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 2, float, packed_highp>		packed_highp_mat4x2;
+
+	/// 4 by 2 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 2, float, packed_mediump>	packed_mediump_mat4x2;
+
+	/// 4 by 2 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 2, float, packed_lowp>		packed_lowp_mat4x2;
+
+	/// 4 by 2 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 2, double, packed_highp>		packed_highp_dmat4x2;
+
+	/// 4 by 2 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 2, double, packed_mediump>	packed_mediump_dmat4x2;
+
+	/// 4 by 2 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 2, double, packed_lowp>		packed_lowp_dmat4x2;
+
+	// -- *mat4x3 --
+
+	/// 4 by 3 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 3, float, aligned_highp>		aligned_highp_mat4x3;
+
+	/// 4 by 3 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 3, float, aligned_mediump>	aligned_mediump_mat4x3;
+
+	/// 4 by 3 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 3, float, aligned_lowp>		aligned_lowp_mat4x3;
+
+	/// 4 by 3 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 3, double, aligned_highp>	aligned_highp_dmat4x3;
+
+	/// 4 by 3 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 3, double, aligned_mediump>	aligned_mediump_dmat4x3;
+
+	/// 4 by 3 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 3, double, aligned_lowp>		aligned_lowp_dmat4x3;
+
+	/// 4 by 3 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 3, float, packed_highp>		packed_highp_mat4x3;
+
+	/// 4 by 3 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 3, float, packed_mediump>	packed_mediump_mat4x3;
+
+	/// 4 by 3 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 3, float, packed_lowp>		packed_lowp_mat4x3;
+
+	/// 4 by 3 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 3, double, packed_highp>		packed_highp_dmat4x3;
+
+	/// 4 by 3 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 3, double, packed_mediump>	packed_mediump_dmat4x3;
+
+	/// 4 by 3 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 3, double, packed_lowp>		packed_lowp_dmat4x3;
+
+	// -- *mat4x4 --
+
+	/// 4 by 4 matrix aligned in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, aligned_highp>		aligned_highp_mat4x4;
+
+	/// 4 by 4 matrix aligned in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, aligned_mediump>	aligned_mediump_mat4x4;
+
+	/// 4 by 4 matrix aligned in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, aligned_lowp>		aligned_lowp_mat4x4;
+
+	/// 4 by 4 matrix aligned in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, aligned_highp>	aligned_highp_dmat4x4;
+
+	/// 4 by 4 matrix aligned in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, aligned_mediump>	aligned_mediump_dmat4x4;
+
+	/// 4 by 4 matrix aligned in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, aligned_lowp>		aligned_lowp_dmat4x4;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, packed_highp>		packed_highp_mat4x4;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, packed_mediump>	packed_mediump_mat4x4;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, float, packed_lowp>		packed_lowp_mat4x4;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers using high precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, packed_highp>		packed_highp_dmat4x4;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers using medium precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, packed_mediump>	packed_mediump_dmat4x4;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers using low precision arithmetic in term of ULPs.
+	typedef mat<4, 4, double, packed_lowp>		packed_lowp_dmat4x4;
+
+	// -- default --
+
+#if(defined(GLM_PRECISION_LOWP_FLOAT))
+	typedef aligned_lowp_vec1			aligned_vec1;
+	typedef aligned_lowp_vec2			aligned_vec2;
+	typedef aligned_lowp_vec3			aligned_vec3;
+	typedef aligned_lowp_vec4			aligned_vec4;
+	typedef packed_lowp_vec1			packed_vec1;
+	typedef packed_lowp_vec2			packed_vec2;
+	typedef packed_lowp_vec3			packed_vec3;
+	typedef packed_lowp_vec4			packed_vec4;
+
+	typedef aligned_lowp_mat2			aligned_mat2;
+	typedef aligned_lowp_mat3			aligned_mat3;
+	typedef aligned_lowp_mat4			aligned_mat4;
+	typedef packed_lowp_mat2			packed_mat2;
+	typedef packed_lowp_mat3			packed_mat3;
+	typedef packed_lowp_mat4			packed_mat4;
+
+	typedef aligned_lowp_mat2x2			aligned_mat2x2;
+	typedef aligned_lowp_mat2x3			aligned_mat2x3;
+	typedef aligned_lowp_mat2x4			aligned_mat2x4;
+	typedef aligned_lowp_mat3x2			aligned_mat3x2;
+	typedef aligned_lowp_mat3x3			aligned_mat3x3;
+	typedef aligned_lowp_mat3x4			aligned_mat3x4;
+	typedef aligned_lowp_mat4x2			aligned_mat4x2;
+	typedef aligned_lowp_mat4x3			aligned_mat4x3;
+	typedef aligned_lowp_mat4x4			aligned_mat4x4;
+	typedef packed_lowp_mat2x2			packed_mat2x2;
+	typedef packed_lowp_mat2x3			packed_mat2x3;
+	typedef packed_lowp_mat2x4			packed_mat2x4;
+	typedef packed_lowp_mat3x2			packed_mat3x2;
+	typedef packed_lowp_mat3x3			packed_mat3x3;
+	typedef packed_lowp_mat3x4			packed_mat3x4;
+	typedef packed_lowp_mat4x2			packed_mat4x2;
+	typedef packed_lowp_mat4x3			packed_mat4x3;
+	typedef packed_lowp_mat4x4			packed_mat4x4;
+#elif(defined(GLM_PRECISION_MEDIUMP_FLOAT))
+	typedef aligned_mediump_vec1		aligned_vec1;
+	typedef aligned_mediump_vec2		aligned_vec2;
+	typedef aligned_mediump_vec3		aligned_vec3;
+	typedef aligned_mediump_vec4		aligned_vec4;
+	typedef packed_mediump_vec1			packed_vec1;
+	typedef packed_mediump_vec2			packed_vec2;
+	typedef packed_mediump_vec3			packed_vec3;
+	typedef packed_mediump_vec4			packed_vec4;
+
+	typedef aligned_mediump_mat2		aligned_mat2;
+	typedef aligned_mediump_mat3		aligned_mat3;
+	typedef aligned_mediump_mat4		aligned_mat4;
+	typedef packed_mediump_mat2			packed_mat2;
+	typedef packed_mediump_mat3			packed_mat3;
+	typedef packed_mediump_mat4			packed_mat4;
+
+	typedef aligned_mediump_mat2x2		aligned_mat2x2;
+	typedef aligned_mediump_mat2x3		aligned_mat2x3;
+	typedef aligned_mediump_mat2x4		aligned_mat2x4;
+	typedef aligned_mediump_mat3x2		aligned_mat3x2;
+	typedef aligned_mediump_mat3x3		aligned_mat3x3;
+	typedef aligned_mediump_mat3x4		aligned_mat3x4;
+	typedef aligned_mediump_mat4x2		aligned_mat4x2;
+	typedef aligned_mediump_mat4x3		aligned_mat4x3;
+	typedef aligned_mediump_mat4x4		aligned_mat4x4;
+	typedef packed_mediump_mat2x2		packed_mat2x2;
+	typedef packed_mediump_mat2x3		packed_mat2x3;
+	typedef packed_mediump_mat2x4		packed_mat2x4;
+	typedef packed_mediump_mat3x2		packed_mat3x2;
+	typedef packed_mediump_mat3x3		packed_mat3x3;
+	typedef packed_mediump_mat3x4		packed_mat3x4;
+	typedef packed_mediump_mat4x2		packed_mat4x2;
+	typedef packed_mediump_mat4x3		packed_mat4x3;
+	typedef packed_mediump_mat4x4		packed_mat4x4;
+#else //defined(GLM_PRECISION_HIGHP_FLOAT)
+	/// 1 component vector aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_vec1			aligned_vec1;
+
+	/// 2 components vector aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_vec2			aligned_vec2;
+
+	/// 3 components vector aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_vec3			aligned_vec3;
+
+	/// 4 components vector aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_vec4 			aligned_vec4;
+
+	/// 1 component vector tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_vec1			packed_vec1;
+
+	/// 2 components vector tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_vec2			packed_vec2;
+
+	/// 3 components vector tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_vec3			packed_vec3;
+
+	/// 4 components vector tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_vec4			packed_vec4;
+
+	/// 2 by 2 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat2			aligned_mat2;
+
+	/// 3 by 3 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat3			aligned_mat3;
+
+	/// 4 by 4 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat4			aligned_mat4;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat2			packed_mat2;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat3			packed_mat3;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat4			packed_mat4;
+
+	/// 2 by 2 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat2x2		aligned_mat2x2;
+
+	/// 2 by 3 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat2x3		aligned_mat2x3;
+
+	/// 2 by 4 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat2x4		aligned_mat2x4;
+
+	/// 3 by 2 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat3x2		aligned_mat3x2;
+
+	/// 3 by 3 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat3x3		aligned_mat3x3;
+
+	/// 3 by 4 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat3x4		aligned_mat3x4;
+
+	/// 4 by 2 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat4x2		aligned_mat4x2;
+
+	/// 4 by 3 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat4x3		aligned_mat4x3;
+
+	/// 4 by 4 matrix tightly aligned in memory of single-precision floating-point numbers.
+	typedef aligned_highp_mat4x4		aligned_mat4x4;
+
+	/// 2 by 2 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat2x2			packed_mat2x2;
+
+	/// 2 by 3 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat2x3			packed_mat2x3;
+
+	/// 2 by 4 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat2x4			packed_mat2x4;
+
+	/// 3 by 2 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat3x2			packed_mat3x2;
+
+	/// 3 by 3 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat3x3			packed_mat3x3;
+
+	/// 3 by 4 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat3x4			packed_mat3x4;
+
+	/// 4 by 2 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat4x2			packed_mat4x2;
+
+	/// 4 by 3 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat4x3			packed_mat4x3;
+
+	/// 4 by 4 matrix tightly packed in memory of single-precision floating-point numbers.
+	typedef packed_highp_mat4x4			packed_mat4x4;
+#endif//GLM_PRECISION
+
+#if(defined(GLM_PRECISION_LOWP_DOUBLE))
+	typedef aligned_lowp_dvec1			aligned_dvec1;
+	typedef aligned_lowp_dvec2			aligned_dvec2;
+	typedef aligned_lowp_dvec3			aligned_dvec3;
+	typedef aligned_lowp_dvec4			aligned_dvec4;
+	typedef packed_lowp_dvec1			packed_dvec1;
+	typedef packed_lowp_dvec2			packed_dvec2;
+	typedef packed_lowp_dvec3			packed_dvec3;
+	typedef packed_lowp_dvec4			packed_dvec4;
+
+	typedef aligned_lowp_dmat2			aligned_dmat2;
+	typedef aligned_lowp_dmat3			aligned_dmat3;
+	typedef aligned_lowp_dmat4			aligned_dmat4;
+	typedef packed_lowp_dmat2			packed_dmat2;
+	typedef packed_lowp_dmat3			packed_dmat3;
+	typedef packed_lowp_dmat4			packed_dmat4;
+
+	typedef aligned_lowp_dmat2x2		aligned_dmat2x2;
+	typedef aligned_lowp_dmat2x3		aligned_dmat2x3;
+	typedef aligned_lowp_dmat2x4		aligned_dmat2x4;
+	typedef aligned_lowp_dmat3x2		aligned_dmat3x2;
+	typedef aligned_lowp_dmat3x3		aligned_dmat3x3;
+	typedef aligned_lowp_dmat3x4		aligned_dmat3x4;
+	typedef aligned_lowp_dmat4x2		aligned_dmat4x2;
+	typedef aligned_lowp_dmat4x3		aligned_dmat4x3;
+	typedef aligned_lowp_dmat4x4		aligned_dmat4x4;
+	typedef packed_lowp_dmat2x2			packed_dmat2x2;
+	typedef packed_lowp_dmat2x3			packed_dmat2x3;
+	typedef packed_lowp_dmat2x4			packed_dmat2x4;
+	typedef packed_lowp_dmat3x2			packed_dmat3x2;
+	typedef packed_lowp_dmat3x3			packed_dmat3x3;
+	typedef packed_lowp_dmat3x4			packed_dmat3x4;
+	typedef packed_lowp_dmat4x2			packed_dmat4x2;
+	typedef packed_lowp_dmat4x3			packed_dmat4x3;
+	typedef packed_lowp_dmat4x4			packed_dmat4x4;
+#elif(defined(GLM_PRECISION_MEDIUMP_DOUBLE))
+	typedef aligned_mediump_dvec1		aligned_dvec1;
+	typedef aligned_mediump_dvec2		aligned_dvec2;
+	typedef aligned_mediump_dvec3		aligned_dvec3;
+	typedef aligned_mediump_dvec4		aligned_dvec4;
+	typedef packed_mediump_dvec1		packed_dvec1;
+	typedef packed_mediump_dvec2		packed_dvec2;
+	typedef packed_mediump_dvec3		packed_dvec3;
+	typedef packed_mediump_dvec4		packed_dvec4;
+
+	typedef aligned_mediump_dmat2		aligned_dmat2;
+	typedef aligned_mediump_dmat3		aligned_dmat3;
+	typedef aligned_mediump_dmat4		aligned_dmat4;
+	typedef packed_mediump_dmat2		packed_dmat2;
+	typedef packed_mediump_dmat3		packed_dmat3;
+	typedef packed_mediump_dmat4		packed_dmat4;
+
+	typedef aligned_mediump_dmat2x2		aligned_dmat2x2;
+	typedef aligned_mediump_dmat2x3		aligned_dmat2x3;
+	typedef aligned_mediump_dmat2x4		aligned_dmat2x4;
+	typedef aligned_mediump_dmat3x2		aligned_dmat3x2;
+	typedef aligned_mediump_dmat3x3		aligned_dmat3x3;
+	typedef aligned_mediump_dmat3x4		aligned_dmat3x4;
+	typedef aligned_mediump_dmat4x2		aligned_dmat4x2;
+	typedef aligned_mediump_dmat4x3		aligned_dmat4x3;
+	typedef aligned_mediump_dmat4x4		aligned_dmat4x4;
+	typedef packed_mediump_dmat2x2		packed_dmat2x2;
+	typedef packed_mediump_dmat2x3		packed_dmat2x3;
+	typedef packed_mediump_dmat2x4		packed_dmat2x4;
+	typedef packed_mediump_dmat3x2		packed_dmat3x2;
+	typedef packed_mediump_dmat3x3		packed_dmat3x3;
+	typedef packed_mediump_dmat3x4		packed_dmat3x4;
+	typedef packed_mediump_dmat4x2		packed_dmat4x2;
+	typedef packed_mediump_dmat4x3		packed_dmat4x3;
+	typedef packed_mediump_dmat4x4		packed_dmat4x4;
+#else //defined(GLM_PRECISION_HIGHP_DOUBLE)
+	/// 1 component vector aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dvec1			aligned_dvec1;
+
+	/// 2 components vector aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dvec2			aligned_dvec2;
+
+	/// 3 components vector aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dvec3			aligned_dvec3;
+
+	/// 4 components vector aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dvec4			aligned_dvec4;
+
+	/// 1 component vector tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dvec1			packed_dvec1;
+
+	/// 2 components vector tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dvec2			packed_dvec2;
+
+	/// 3 components vector tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dvec3			packed_dvec3;
+
+	/// 4 components vector tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dvec4			packed_dvec4;
+
+	/// 2 by 2 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat2			aligned_dmat2;
+
+	/// 3 by 3 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat3			aligned_dmat3;
+
+	/// 4 by 4 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat4			aligned_dmat4;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat2			packed_dmat2;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat3			packed_dmat3;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat4			packed_dmat4;
+
+	/// 2 by 2 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat2x2		aligned_dmat2x2;
+
+	/// 2 by 3 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat2x3		aligned_dmat2x3;
+
+	/// 2 by 4 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat2x4		aligned_dmat2x4;
+
+	/// 3 by 2 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat3x2		aligned_dmat3x2;
+
+	/// 3 by 3 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat3x3		aligned_dmat3x3;
+
+	/// 3 by 4 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat3x4		aligned_dmat3x4;
+
+	/// 4 by 2 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat4x2		aligned_dmat4x2;
+
+	/// 4 by 3 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat4x3		aligned_dmat4x3;
+
+	/// 4 by 4 matrix tightly aligned in memory of double-precision floating-point numbers.
+	typedef aligned_highp_dmat4x4		aligned_dmat4x4;
+
+	/// 2 by 2 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat2x2		packed_dmat2x2;
+
+	/// 2 by 3 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat2x3		packed_dmat2x3;
+
+	/// 2 by 4 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat2x4		packed_dmat2x4;
+
+	/// 3 by 2 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat3x2		packed_dmat3x2;
+
+	/// 3 by 3 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat3x3		packed_dmat3x3;
+
+	/// 3 by 4 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat3x4		packed_dmat3x4;
+
+	/// 4 by 2 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat4x2		packed_dmat4x2;
+
+	/// 4 by 3 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat4x3		packed_dmat4x3;
+
+	/// 4 by 4 matrix tightly packed in memory of double-precision floating-point numbers.
+	typedef packed_highp_dmat4x4		packed_dmat4x4;
+#endif//GLM_PRECISION
+
+#if(defined(GLM_PRECISION_LOWP_INT))
+	typedef aligned_lowp_ivec1			aligned_ivec1;
+	typedef aligned_lowp_ivec2			aligned_ivec2;
+	typedef aligned_lowp_ivec3			aligned_ivec3;
+	typedef aligned_lowp_ivec4			aligned_ivec4;
+#elif(defined(GLM_PRECISION_MEDIUMP_INT))
+	typedef aligned_mediump_ivec1		aligned_ivec1;
+	typedef aligned_mediump_ivec2		aligned_ivec2;
+	typedef aligned_mediump_ivec3		aligned_ivec3;
+	typedef aligned_mediump_ivec4		aligned_ivec4;
+#else //defined(GLM_PRECISION_HIGHP_INT)
+	/// 1 component vector aligned in memory of signed integer numbers.
+	typedef aligned_highp_ivec1			aligned_ivec1;
+
+	/// 2 components vector aligned in memory of signed integer numbers.
+	typedef aligned_highp_ivec2			aligned_ivec2;
+
+	/// 3 components vector aligned in memory of signed integer numbers.
+	typedef aligned_highp_ivec3			aligned_ivec3;
+
+	/// 4 components vector aligned in memory of signed integer numbers.
+	typedef aligned_highp_ivec4			aligned_ivec4;
+
+	/// 1 component vector tightly packed in memory of signed integer numbers.
+	typedef packed_highp_ivec1			packed_ivec1;
+
+	/// 2 components vector tightly packed in memory of signed integer numbers.
+	typedef packed_highp_ivec2			packed_ivec2;
+
+	/// 3 components vector tightly packed in memory of signed integer numbers.
+	typedef packed_highp_ivec3			packed_ivec3;
+
+	/// 4 components vector tightly packed in memory of signed integer numbers.
+	typedef packed_highp_ivec4			packed_ivec4;
+#endif//GLM_PRECISION
+
+	// -- Unsigned integer definition --
+
+#if(defined(GLM_PRECISION_LOWP_UINT))
+	typedef aligned_lowp_uvec1			aligned_uvec1;
+	typedef aligned_lowp_uvec2			aligned_uvec2;
+	typedef aligned_lowp_uvec3			aligned_uvec3;
+	typedef aligned_lowp_uvec4			aligned_uvec4;
+#elif(defined(GLM_PRECISION_MEDIUMP_UINT))
+	typedef aligned_mediump_uvec1		aligned_uvec1;
+	typedef aligned_mediump_uvec2		aligned_uvec2;
+	typedef aligned_mediump_uvec3		aligned_uvec3;
+	typedef aligned_mediump_uvec4		aligned_uvec4;
+#else //defined(GLM_PRECISION_HIGHP_UINT)
+	/// 1 component vector aligned in memory of unsigned integer numbers.
+	typedef aligned_highp_uvec1			aligned_uvec1;
+
+	/// 2 components vector aligned in memory of unsigned integer numbers.
+	typedef aligned_highp_uvec2			aligned_uvec2;
+
+	/// 3 components vector aligned in memory of unsigned integer numbers.
+	typedef aligned_highp_uvec3			aligned_uvec3;
+
+	/// 4 components vector aligned in memory of unsigned integer numbers.
+	typedef aligned_highp_uvec4			aligned_uvec4;
+
+	/// 1 component vector tightly packed in memory of unsigned integer numbers.
+	typedef packed_highp_uvec1			packed_uvec1;
+
+	/// 2 components vector tightly packed in memory of unsigned integer numbers.
+	typedef packed_highp_uvec2			packed_uvec2;
+
+	/// 3 components vector tightly packed in memory of unsigned integer numbers.
+	typedef packed_highp_uvec3			packed_uvec3;
+
+	/// 4 components vector tightly packed in memory of unsigned integer numbers.
+	typedef packed_highp_uvec4			packed_uvec4;
+#endif//GLM_PRECISION
+
+#if(defined(GLM_PRECISION_LOWP_BOOL))
+	typedef aligned_lowp_bvec1			aligned_bvec1;
+	typedef aligned_lowp_bvec2			aligned_bvec2;
+	typedef aligned_lowp_bvec3			aligned_bvec3;
+	typedef aligned_lowp_bvec4			aligned_bvec4;
+#elif(defined(GLM_PRECISION_MEDIUMP_BOOL))
+	typedef aligned_mediump_bvec1		aligned_bvec1;
+	typedef aligned_mediump_bvec2		aligned_bvec2;
+	typedef aligned_mediump_bvec3		aligned_bvec3;
+	typedef aligned_mediump_bvec4		aligned_bvec4;
+#else //defined(GLM_PRECISION_HIGHP_BOOL)
+	/// 1 component vector aligned in memory of bool values.
+	typedef aligned_highp_bvec1			aligned_bvec1;
+
+	/// 2 components vector aligned in memory of bool values.
+	typedef aligned_highp_bvec2			aligned_bvec2;
+
+	/// 3 components vector aligned in memory of bool values.
+	typedef aligned_highp_bvec3			aligned_bvec3;
+
+	/// 4 components vector aligned in memory of bool values.
+	typedef aligned_highp_bvec4			aligned_bvec4;
+
+	/// 1 components vector tightly packed in memory of bool values.
+	typedef packed_highp_bvec1			packed_bvec1;
+
+	/// 2 components vector tightly packed in memory of bool values.
+	typedef packed_highp_bvec2			packed_bvec2;
+
+	/// 3 components vector tightly packed in memory of bool values.
+	typedef packed_highp_bvec3			packed_bvec3;
+
+	/// 4 components vector tightly packed in memory of bool values.
+	typedef packed_highp_bvec4			packed_bvec4;
+#endif//GLM_PRECISION
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.hpp
new file mode 100644
index 000000000000..775e2f484d73
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.hpp
@@ -0,0 +1,2094 @@
+/// @ref gtc_type_precision
+/// @file glm/gtc/type_precision.hpp
+///
+/// @see core (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtc_type_precision GLM_GTC_type_precision
+/// @ingroup gtc
+///
+/// Include <glm/gtc/type_precision.hpp> to use the features of this extension.
+///
+/// Defines specific C++-based qualifier types.
+
+#pragma once
+
+// Dependency:
+#include "../gtc/quaternion.hpp"
+#include "../gtc/vec1.hpp"
+#include "../ext/vector_int1_sized.hpp"
+#include "../ext/vector_int2_sized.hpp"
+#include "../ext/vector_int3_sized.hpp"
+#include "../ext/vector_int4_sized.hpp"
+#include "../ext/scalar_int_sized.hpp"
+#include "../ext/vector_uint1_sized.hpp"
+#include "../ext/vector_uint2_sized.hpp"
+#include "../ext/vector_uint3_sized.hpp"
+#include "../ext/vector_uint4_sized.hpp"
+#include "../ext/scalar_uint_sized.hpp"
+#include "../detail/type_vec2.hpp"
+#include "../detail/type_vec3.hpp"
+#include "../detail/type_vec4.hpp"
+#include "../detail/type_mat2x2.hpp"
+#include "../detail/type_mat2x3.hpp"
+#include "../detail/type_mat2x4.hpp"
+#include "../detail/type_mat3x2.hpp"
+#include "../detail/type_mat3x3.hpp"
+#include "../detail/type_mat3x4.hpp"
+#include "../detail/type_mat4x2.hpp"
+#include "../detail/type_mat4x3.hpp"
+#include "../detail/type_mat4x4.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_type_precision extension included")
+#endif
+
+namespace glm
+{
+	///////////////////////////
+	// Signed int vector types
+
+	/// @addtogroup gtc_type_precision
+	/// @{
+
+	/// Low qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 lowp_int8;
+
+	/// Low qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 lowp_int16;
+
+	/// Low qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 lowp_int32;
+
+	/// Low qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 lowp_int64;
+
+	/// Low qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 lowp_int8_t;
+
+	/// Low qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 lowp_int16_t;
+
+	/// Low qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 lowp_int32_t;
+
+	/// Low qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 lowp_int64_t;
+
+	/// Low qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 lowp_i8;
+
+	/// Low qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 lowp_i16;
+
+	/// Low qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 lowp_i32;
+
+	/// Low qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 lowp_i64;
+
+	/// Medium qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 mediump_int8;
+
+	/// Medium qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 mediump_int16;
+
+	/// Medium qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 mediump_int32;
+
+	/// Medium qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 mediump_int64;
+
+	/// Medium qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 mediump_int8_t;
+
+	/// Medium qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 mediump_int16_t;
+
+	/// Medium qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 mediump_int32_t;
+
+	/// Medium qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 mediump_int64_t;
+
+	/// Medium qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 mediump_i8;
+
+	/// Medium qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 mediump_i16;
+
+	/// Medium qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 mediump_i32;
+
+	/// Medium qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 mediump_i64;
+
+	/// High qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 highp_int8;
+
+	/// High qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 highp_int16;
+
+	/// High qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 highp_int32;
+
+	/// High qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 highp_int64;
+
+	/// High qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 highp_int8_t;
+
+	/// High qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 highp_int16_t;
+
+	/// 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 highp_int32_t;
+
+	/// High qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 highp_int64_t;
+
+	/// High qualifier 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 highp_i8;
+
+	/// High qualifier 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 highp_i16;
+
+	/// High qualifier 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 highp_i32;
+
+	/// High qualifier 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 highp_i64;
+
+
+#if GLM_HAS_EXTENDED_INTEGER_TYPE
+	using std::int8_t;
+	using std::int16_t;
+	using std::int32_t;
+	using std::int64_t;
+#else
+	/// 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 int8_t;
+
+	/// 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 int16_t;
+
+	/// 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 int32_t;
+
+	/// 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 int64_t;
+#endif
+
+	/// 8 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int8 i8;
+
+	/// 16 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int16 i16;
+
+	/// 32 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int32 i32;
+
+	/// 64 bit signed integer type.
+	/// @see gtc_type_precision
+	typedef detail::int64 i64;
+
+	/////////////////////////////
+	// Unsigned int vector types
+
+	/// Low qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 lowp_uint8;
+
+	/// Low qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 lowp_uint16;
+
+	/// Low qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 lowp_uint32;
+
+	/// Low qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 lowp_uint64;
+
+	/// Low qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 lowp_uint8_t;
+
+	/// Low qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 lowp_uint16_t;
+
+	/// Low qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 lowp_uint32_t;
+
+	/// Low qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 lowp_uint64_t;
+
+	/// Low qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 lowp_u8;
+
+	/// Low qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 lowp_u16;
+
+	/// Low qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 lowp_u32;
+
+	/// Low qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 lowp_u64;
+
+	/// Medium qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 mediump_uint8;
+
+	/// Medium qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 mediump_uint16;
+
+	/// Medium qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 mediump_uint32;
+
+	/// Medium qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 mediump_uint64;
+
+	/// Medium qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 mediump_uint8_t;
+
+	/// Medium qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 mediump_uint16_t;
+
+	/// Medium qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 mediump_uint32_t;
+
+	/// Medium qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 mediump_uint64_t;
+
+	/// Medium qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 mediump_u8;
+
+	/// Medium qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 mediump_u16;
+
+	/// Medium qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 mediump_u32;
+
+	/// Medium qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 mediump_u64;
+
+	/// High qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 highp_uint8;
+
+	/// High qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 highp_uint16;
+
+	/// High qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 highp_uint32;
+
+	/// High qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 highp_uint64;
+
+	/// High qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 highp_uint8_t;
+
+	/// High qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 highp_uint16_t;
+
+	/// High qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 highp_uint32_t;
+
+	/// High qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 highp_uint64_t;
+
+	/// High qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 highp_u8;
+
+	/// High qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 highp_u16;
+
+	/// High qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 highp_u32;
+
+	/// High qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 highp_u64;
+
+#if GLM_HAS_EXTENDED_INTEGER_TYPE
+	using std::uint8_t;
+	using std::uint16_t;
+	using std::uint32_t;
+	using std::uint64_t;
+#else
+	/// Default qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 uint8_t;
+
+	/// Default qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 uint16_t;
+
+	/// Default qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 uint32_t;
+
+	/// Default qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 uint64_t;
+#endif
+
+	/// Default qualifier 8 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint8 u8;
+
+	/// Default qualifier 16 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint16 u16;
+
+	/// Default qualifier 32 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint32 u32;
+
+	/// Default qualifier 64 bit unsigned integer type.
+	/// @see gtc_type_precision
+	typedef detail::uint64 u64;
+
+
+
+
+
+	//////////////////////
+	// Float vector types
+
+	/// Single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float float32;
+
+	/// Double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef double float64;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_float32;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_float64;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_float32_t;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_float64_t;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_f32;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_f64;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_float32;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_float64;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_float32_t;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_float64_t;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_f32;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_f64;
+
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_float32;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_float64;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_float32_t;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_float64_t;
+
+	/// Low 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 lowp_f32;
+
+	/// Low 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 lowp_f64;
+
+
+	/// Medium 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 mediump_float32;
+
+	/// Medium 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 mediump_float64;
+
+	/// Medium 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 mediump_float32_t;
+
+	/// Medium 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 mediump_float64_t;
+
+	/// Medium 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 mediump_f32;
+
+	/// Medium 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 mediump_f64;
+
+
+	/// High 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 highp_float32;
+
+	/// High 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 highp_float64;
+
+	/// High 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 highp_float32_t;
+
+	/// High 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 highp_float64_t;
+
+	/// High 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 highp_f32;
+
+	/// High 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float64 highp_f64;
+
+
+#if(defined(GLM_PRECISION_LOWP_FLOAT))
+	/// Default 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef lowp_float32_t float32_t;
+
+	/// Default 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef lowp_float64_t float64_t;
+
+	/// Default 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef lowp_f32 f32;
+
+	/// Default 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef lowp_f64 f64;
+
+#elif(defined(GLM_PRECISION_MEDIUMP_FLOAT))
+	/// Default 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef mediump_float32 float32_t;
+
+	/// Default 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef mediump_float64 float64_t;
+
+	/// Default 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef mediump_float32 f32;
+
+	/// Default 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef mediump_float64 f64;
+
+#else//(defined(GLM_PRECISION_HIGHP_FLOAT))
+
+	/// Default 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef highp_float32_t float32_t;
+
+	/// Default 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef highp_float64_t float64_t;
+
+	/// Default 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef highp_float32_t f32;
+
+	/// Default 64 bit double-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef highp_float64_t f64;
+#endif
+
+
+	/// Low single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, float, lowp> lowp_fvec1;
+
+	/// Low single-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, float, lowp> lowp_fvec2;
+
+	/// Low single-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, float, lowp> lowp_fvec3;
+
+	/// Low single-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, float, lowp> lowp_fvec4;
+
+
+	/// Medium single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, float, mediump> mediump_fvec1;
+
+	/// Medium Single-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, float, mediump> mediump_fvec2;
+
+	/// Medium Single-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, float, mediump> mediump_fvec3;
+
+	/// Medium Single-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, float, mediump> mediump_fvec4;
+
+
+	/// High single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, float, highp> highp_fvec1;
+
+	/// High Single-qualifier floating-point vector of 2 components.
+	/// @see core_precision
+	typedef vec<2, float, highp> highp_fvec2;
+
+	/// High Single-qualifier floating-point vector of 3 components.
+	/// @see core_precision
+	typedef vec<3, float, highp> highp_fvec3;
+
+	/// High Single-qualifier floating-point vector of 4 components.
+	/// @see core_precision
+	typedef vec<4, float, highp> highp_fvec4;
+
+
+	/// Low single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f32, lowp> lowp_f32vec1;
+
+	/// Low single-qualifier floating-point vector of 2 components.
+	/// @see core_precision
+	typedef vec<2, f32, lowp> lowp_f32vec2;
+
+	/// Low single-qualifier floating-point vector of 3 components.
+	/// @see core_precision
+	typedef vec<3, f32, lowp> lowp_f32vec3;
+
+	/// Low single-qualifier floating-point vector of 4 components.
+	/// @see core_precision
+	typedef vec<4, f32, lowp> lowp_f32vec4;
+
+	/// Medium single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f32, mediump> mediump_f32vec1;
+
+	/// Medium single-qualifier floating-point vector of 2 components.
+	/// @see core_precision
+	typedef vec<2, f32, mediump> mediump_f32vec2;
+
+	/// Medium single-qualifier floating-point vector of 3 components.
+	/// @see core_precision
+	typedef vec<3, f32, mediump> mediump_f32vec3;
+
+	/// Medium single-qualifier floating-point vector of 4 components.
+	/// @see core_precision
+	typedef vec<4, f32, mediump> mediump_f32vec4;
+
+	/// High single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f32, highp> highp_f32vec1;
+
+	/// High single-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, f32, highp> highp_f32vec2;
+
+	/// High single-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, f32, highp> highp_f32vec3;
+
+	/// High single-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, f32, highp> highp_f32vec4;
+
+
+	/// Low double-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f64, lowp> lowp_f64vec1;
+
+	/// Low double-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, f64, lowp> lowp_f64vec2;
+
+	/// Low double-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, f64, lowp> lowp_f64vec3;
+
+	/// Low double-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, f64, lowp> lowp_f64vec4;
+
+	/// Medium double-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f64, mediump> mediump_f64vec1;
+
+	/// Medium double-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, f64, mediump> mediump_f64vec2;
+
+	/// Medium double-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, f64, mediump> mediump_f64vec3;
+
+	/// Medium double-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, f64, mediump> mediump_f64vec4;
+
+	/// High double-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f64, highp> highp_f64vec1;
+
+	/// High double-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, f64, highp> highp_f64vec2;
+
+	/// High double-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, f64, highp> highp_f64vec3;
+
+	/// High double-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, f64, highp> highp_f64vec4;
+
+
+
+	//////////////////////
+	// Float matrix types
+
+	/// Low single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef lowp_f32 lowp_fmat1x1;
+
+	/// Low single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, lowp> lowp_fmat2x2;
+
+	/// Low single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, lowp> lowp_fmat2x3;
+
+	/// Low single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, lowp> lowp_fmat2x4;
+
+	/// Low single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, lowp> lowp_fmat3x2;
+
+	/// Low single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, lowp> lowp_fmat3x3;
+
+	/// Low single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, lowp> lowp_fmat3x4;
+
+	/// Low single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, lowp> lowp_fmat4x2;
+
+	/// Low single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, lowp> lowp_fmat4x3;
+
+	/// Low single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, lowp> lowp_fmat4x4;
+
+	/// Low single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef lowp_fmat1x1 lowp_fmat1;
+
+	/// Low single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_fmat2x2 lowp_fmat2;
+
+	/// Low single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_fmat3x3 lowp_fmat3;
+
+	/// Low single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_fmat4x4 lowp_fmat4;
+
+
+	/// Medium single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef mediump_f32 mediump_fmat1x1;
+
+	/// Medium single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, mediump> mediump_fmat2x2;
+
+	/// Medium single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, mediump> mediump_fmat2x3;
+
+	/// Medium single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, mediump> mediump_fmat2x4;
+
+	/// Medium single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, mediump> mediump_fmat3x2;
+
+	/// Medium single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, mediump> mediump_fmat3x3;
+
+	/// Medium single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, mediump> mediump_fmat3x4;
+
+	/// Medium single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, mediump> mediump_fmat4x2;
+
+	/// Medium single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, mediump> mediump_fmat4x3;
+
+	/// Medium single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, mediump> mediump_fmat4x4;
+
+	/// Medium single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef mediump_fmat1x1 mediump_fmat1;
+
+	/// Medium single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_fmat2x2 mediump_fmat2;
+
+	/// Medium single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_fmat3x3 mediump_fmat3;
+
+	/// Medium single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_fmat4x4 mediump_fmat4;
+
+
+	/// High single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef highp_f32 highp_fmat1x1;
+
+	/// High single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, highp> highp_fmat2x2;
+
+	/// High single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, highp> highp_fmat2x3;
+
+	/// High single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, highp> highp_fmat2x4;
+
+	/// High single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, highp> highp_fmat3x2;
+
+	/// High single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, highp> highp_fmat3x3;
+
+	/// High single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, highp> highp_fmat3x4;
+
+	/// High single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, highp> highp_fmat4x2;
+
+	/// High single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, highp> highp_fmat4x3;
+
+	/// High single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, highp> highp_fmat4x4;
+
+	/// High single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef highp_fmat1x1 highp_fmat1;
+
+	/// High single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef highp_fmat2x2 highp_fmat2;
+
+	/// High single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef highp_fmat3x3 highp_fmat3;
+
+	/// High single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef highp_fmat4x4 highp_fmat4;
+
+
+	/// Low single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f32 lowp_f32mat1x1;
+
+	/// Low single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, lowp> lowp_f32mat2x2;
+
+	/// Low single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, lowp> lowp_f32mat2x3;
+
+	/// Low single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, lowp> lowp_f32mat2x4;
+
+	/// Low single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, lowp> lowp_f32mat3x2;
+
+	/// Low single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, lowp> lowp_f32mat3x3;
+
+	/// Low single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, lowp> lowp_f32mat3x4;
+
+	/// Low single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, lowp> lowp_f32mat4x2;
+
+	/// Low single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, lowp> lowp_f32mat4x3;
+
+	/// Low single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, lowp> lowp_f32mat4x4;
+
+	/// Low single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef detail::tmat1x1<f32, lowp> lowp_f32mat1;
+
+	/// Low single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_f32mat2x2 lowp_f32mat2;
+
+	/// Low single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_f32mat3x3 lowp_f32mat3;
+
+	/// Low single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_f32mat4x4 lowp_f32mat4;
+
+
+	/// High single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f32 mediump_f32mat1x1;
+
+	/// Low single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, mediump> mediump_f32mat2x2;
+
+	/// Medium single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, mediump> mediump_f32mat2x3;
+
+	/// Medium single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, mediump> mediump_f32mat2x4;
+
+	/// Medium single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, mediump> mediump_f32mat3x2;
+
+	/// Medium single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, mediump> mediump_f32mat3x3;
+
+	/// Medium single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, mediump> mediump_f32mat3x4;
+
+	/// Medium single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, mediump> mediump_f32mat4x2;
+
+	/// Medium single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, mediump> mediump_f32mat4x3;
+
+	/// Medium single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, mediump> mediump_f32mat4x4;
+
+	/// Medium single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef detail::tmat1x1<f32, mediump> f32mat1;
+
+	/// Medium single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_f32mat2x2 mediump_f32mat2;
+
+	/// Medium single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_f32mat3x3 mediump_f32mat3;
+
+	/// Medium single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_f32mat4x4 mediump_f32mat4;
+
+
+	/// High single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f32 highp_f32mat1x1;
+
+	/// High single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, highp> highp_f32mat2x2;
+
+	/// High single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, highp> highp_f32mat2x3;
+
+	/// High single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, highp> highp_f32mat2x4;
+
+	/// High single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, highp> highp_f32mat3x2;
+
+	/// High single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, highp> highp_f32mat3x3;
+
+	/// High single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, highp> highp_f32mat3x4;
+
+	/// High single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, highp> highp_f32mat4x2;
+
+	/// High single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, highp> highp_f32mat4x3;
+
+	/// High single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, highp> highp_f32mat4x4;
+
+	/// High single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef detail::tmat1x1<f32, highp> f32mat1;
+
+	/// High single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef highp_f32mat2x2 highp_f32mat2;
+
+	/// High single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef highp_f32mat3x3 highp_f32mat3;
+
+	/// High single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef highp_f32mat4x4 highp_f32mat4;
+
+
+	/// Low double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f64 lowp_f64mat1x1;
+
+	/// Low double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f64, lowp> lowp_f64mat2x2;
+
+	/// Low double-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f64, lowp> lowp_f64mat2x3;
+
+	/// Low double-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f64, lowp> lowp_f64mat2x4;
+
+	/// Low double-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f64, lowp> lowp_f64mat3x2;
+
+	/// Low double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f64, lowp> lowp_f64mat3x3;
+
+	/// Low double-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f64, lowp> lowp_f64mat3x4;
+
+	/// Low double-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f64, lowp> lowp_f64mat4x2;
+
+	/// Low double-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f64, lowp> lowp_f64mat4x3;
+
+	/// Low double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f64, lowp> lowp_f64mat4x4;
+
+	/// Low double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef lowp_f64mat1x1 lowp_f64mat1;
+
+	/// Low double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_f64mat2x2 lowp_f64mat2;
+
+	/// Low double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_f64mat3x3 lowp_f64mat3;
+
+	/// Low double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef lowp_f64mat4x4 lowp_f64mat4;
+
+
+	/// Medium double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f64 Highp_f64mat1x1;
+
+	/// Medium double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f64, mediump> mediump_f64mat2x2;
+
+	/// Medium double-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f64, mediump> mediump_f64mat2x3;
+
+	/// Medium double-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f64, mediump> mediump_f64mat2x4;
+
+	/// Medium double-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f64, mediump> mediump_f64mat3x2;
+
+	/// Medium double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f64, mediump> mediump_f64mat3x3;
+
+	/// Medium double-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f64, mediump> mediump_f64mat3x4;
+
+	/// Medium double-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f64, mediump> mediump_f64mat4x2;
+
+	/// Medium double-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f64, mediump> mediump_f64mat4x3;
+
+	/// Medium double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f64, mediump> mediump_f64mat4x4;
+
+	/// Medium double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef mediump_f64mat1x1 mediump_f64mat1;
+
+	/// Medium double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_f64mat2x2 mediump_f64mat2;
+
+	/// Medium double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_f64mat3x3 mediump_f64mat3;
+
+	/// Medium double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mediump_f64mat4x4 mediump_f64mat4;
+
+	/// High double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f64 highp_f64mat1x1;
+
+	/// High double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f64, highp> highp_f64mat2x2;
+
+	/// High double-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f64, highp> highp_f64mat2x3;
+
+	/// High double-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f64, highp> highp_f64mat2x4;
+
+	/// High double-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f64, highp> highp_f64mat3x2;
+
+	/// High double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f64, highp> highp_f64mat3x3;
+
+	/// High double-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f64, highp> highp_f64mat3x4;
+
+	/// High double-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f64, highp> highp_f64mat4x2;
+
+	/// High double-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f64, highp> highp_f64mat4x3;
+
+	/// High double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f64, highp> highp_f64mat4x4;
+
+	/// High double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef highp_f64mat1x1 highp_f64mat1;
+
+	/// High double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef highp_f64mat2x2 highp_f64mat2;
+
+	/// High double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef highp_f64mat3x3 highp_f64mat3;
+
+	/// High double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef highp_f64mat4x4 highp_f64mat4;
+
+
+	/////////////////////////////
+	// Signed int vector types
+
+	/// Low qualifier signed integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, int, lowp>		lowp_ivec1;
+
+	/// Low qualifier signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, int, lowp>		lowp_ivec2;
+
+	/// Low qualifier signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, int, lowp>		lowp_ivec3;
+
+	/// Low qualifier signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, int, lowp>		lowp_ivec4;
+
+
+	/// Medium qualifier signed integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, int, mediump>	mediump_ivec1;
+
+	/// Medium qualifier signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, int, mediump>	mediump_ivec2;
+
+	/// Medium qualifier signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, int, mediump>	mediump_ivec3;
+
+	/// Medium qualifier signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, int, mediump>	mediump_ivec4;
+
+
+	/// High qualifier signed integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, int, highp>		highp_ivec1;
+
+	/// High qualifier signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, int, highp>		highp_ivec2;
+
+	/// High qualifier signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, int, highp>		highp_ivec3;
+
+	/// High qualifier signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, int, highp>		highp_ivec4;
+
+
+	/// Low qualifier 8 bit signed integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, i8, lowp>		lowp_i8vec1;
+
+	/// Low qualifier 8 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i8, lowp>		lowp_i8vec2;
+
+	/// Low qualifier 8 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i8, lowp>		lowp_i8vec3;
+
+	/// Low qualifier 8 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i8, lowp>		lowp_i8vec4;
+
+
+	/// Medium qualifier 8 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i8, mediump>		mediump_i8vec1;
+
+	/// Medium qualifier 8 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i8, mediump>		mediump_i8vec2;
+
+	/// Medium qualifier 8 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i8, mediump>		mediump_i8vec3;
+
+	/// Medium qualifier 8 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i8, mediump>		mediump_i8vec4;
+
+
+	/// High qualifier 8 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i8, highp>		highp_i8vec1;
+
+	/// High qualifier 8 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i8, highp>		highp_i8vec2;
+
+	/// High qualifier 8 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i8, highp>		highp_i8vec3;
+
+	/// High qualifier 8 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i8, highp>		highp_i8vec4;
+
+
+	/// Low qualifier 16 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i16, lowp>		lowp_i16vec1;
+
+	/// Low qualifier 16 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i16, lowp>		lowp_i16vec2;
+
+	/// Low qualifier 16 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i16, lowp>		lowp_i16vec3;
+
+	/// Low qualifier 16 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i16, lowp>		lowp_i16vec4;
+
+
+	/// Medium qualifier 16 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i16, mediump>	mediump_i16vec1;
+
+	/// Medium qualifier 16 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i16, mediump>	mediump_i16vec2;
+
+	/// Medium qualifier 16 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i16, mediump>	mediump_i16vec3;
+
+	/// Medium qualifier 16 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i16, mediump>	mediump_i16vec4;
+
+
+	/// High qualifier 16 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i16, highp>		highp_i16vec1;
+
+	/// High qualifier 16 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i16, highp>		highp_i16vec2;
+
+	/// High qualifier 16 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i16, highp>		highp_i16vec3;
+
+	/// High qualifier 16 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i16, highp>		highp_i16vec4;
+
+
+	/// Low qualifier 32 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i32, lowp>		lowp_i32vec1;
+
+	/// Low qualifier 32 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i32, lowp>		lowp_i32vec2;
+
+	/// Low qualifier 32 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i32, lowp>		lowp_i32vec3;
+
+	/// Low qualifier 32 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i32, lowp>		lowp_i32vec4;
+
+
+	/// Medium qualifier 32 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i32, mediump>	mediump_i32vec1;
+
+	/// Medium qualifier 32 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i32, mediump>	mediump_i32vec2;
+
+	/// Medium qualifier 32 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i32, mediump>	mediump_i32vec3;
+
+	/// Medium qualifier 32 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i32, mediump>	mediump_i32vec4;
+
+
+	/// High qualifier 32 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i32, highp>		highp_i32vec1;
+
+	/// High qualifier 32 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i32, highp>		highp_i32vec2;
+
+	/// High qualifier 32 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i32, highp>		highp_i32vec3;
+
+	/// High qualifier 32 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i32, highp>		highp_i32vec4;
+
+
+	/// Low qualifier 64 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i64, lowp>		lowp_i64vec1;
+
+	/// Low qualifier 64 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i64, lowp>		lowp_i64vec2;
+
+	/// Low qualifier 64 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i64, lowp>		lowp_i64vec3;
+
+	/// Low qualifier 64 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i64, lowp>		lowp_i64vec4;
+
+
+	/// Medium qualifier 64 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i64, mediump>	mediump_i64vec1;
+
+	/// Medium qualifier 64 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i64, mediump>	mediump_i64vec2;
+
+	/// Medium qualifier 64 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i64, mediump>	mediump_i64vec3;
+
+	/// Medium qualifier 64 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i64, mediump>	mediump_i64vec4;
+
+
+	/// High qualifier 64 bit signed integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, i64, highp>		highp_i64vec1;
+
+	/// High qualifier 64 bit signed integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, i64, highp>		highp_i64vec2;
+
+	/// High qualifier 64 bit signed integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, i64, highp>		highp_i64vec3;
+
+	/// High qualifier 64 bit signed integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, i64, highp>		highp_i64vec4;
+
+
+	/////////////////////////////
+	// Unsigned int vector types
+
+	/// Low qualifier unsigned integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, uint, lowp>		lowp_uvec1;
+
+	/// Low qualifier unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, uint, lowp>		lowp_uvec2;
+
+	/// Low qualifier unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, uint, lowp>		lowp_uvec3;
+
+	/// Low qualifier unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, uint, lowp>		lowp_uvec4;
+
+
+	/// Medium qualifier unsigned integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, uint, mediump>	mediump_uvec1;
+
+	/// Medium qualifier unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, uint, mediump>	mediump_uvec2;
+
+	/// Medium qualifier unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, uint, mediump>	mediump_uvec3;
+
+	/// Medium qualifier unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, uint, mediump>	mediump_uvec4;
+
+
+	/// High qualifier unsigned integer vector of 1 component type.
+	/// @see gtc_type_precision
+	typedef vec<1, uint, highp>		highp_uvec1;
+
+	/// High qualifier unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, uint, highp>		highp_uvec2;
+
+	/// High qualifier unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, uint, highp>		highp_uvec3;
+
+	/// High qualifier unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, uint, highp>		highp_uvec4;
+
+
+	/// Low qualifier 8 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u8, lowp>		lowp_u8vec1;
+
+	/// Low qualifier 8 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u8, lowp>		lowp_u8vec2;
+
+	/// Low qualifier 8 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u8, lowp>		lowp_u8vec3;
+
+	/// Low qualifier 8 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u8, lowp>		lowp_u8vec4;
+
+
+	/// Medium qualifier 8 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u8, mediump>		mediump_u8vec1;
+
+	/// Medium qualifier 8 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u8, mediump>		mediump_u8vec2;
+
+	/// Medium qualifier 8 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u8, mediump>		mediump_u8vec3;
+
+	/// Medium qualifier 8 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u8, mediump>		mediump_u8vec4;
+
+
+	/// High qualifier 8 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u8, highp>		highp_u8vec1;
+
+	/// High qualifier 8 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u8, highp>		highp_u8vec2;
+
+	/// High qualifier 8 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u8, highp>		highp_u8vec3;
+
+	/// High qualifier 8 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u8, highp>		highp_u8vec4;
+
+
+	/// Low qualifier 16 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u16, lowp>		lowp_u16vec1;
+
+	/// Low qualifier 16 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u16, lowp>		lowp_u16vec2;
+
+	/// Low qualifier 16 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u16, lowp>		lowp_u16vec3;
+
+	/// Low qualifier 16 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u16, lowp>		lowp_u16vec4;
+
+
+	/// Medium qualifier 16 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u16, mediump>	mediump_u16vec1;
+
+	/// Medium qualifier 16 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u16, mediump>	mediump_u16vec2;
+
+	/// Medium qualifier 16 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u16, mediump>	mediump_u16vec3;
+
+	/// Medium qualifier 16 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u16, mediump>	mediump_u16vec4;
+
+
+	/// High qualifier 16 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u16, highp>		highp_u16vec1;
+
+	/// High qualifier 16 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u16, highp>		highp_u16vec2;
+
+	/// High qualifier 16 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u16, highp>		highp_u16vec3;
+
+	/// High qualifier 16 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u16, highp>		highp_u16vec4;
+
+
+	/// Low qualifier 32 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u32, lowp>		lowp_u32vec1;
+
+	/// Low qualifier 32 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u32, lowp>		lowp_u32vec2;
+
+	/// Low qualifier 32 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u32, lowp>		lowp_u32vec3;
+
+	/// Low qualifier 32 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u32, lowp>		lowp_u32vec4;
+
+
+	/// Medium qualifier 32 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u32, mediump>	mediump_u32vec1;
+
+	/// Medium qualifier 32 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u32, mediump>	mediump_u32vec2;
+
+	/// Medium qualifier 32 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u32, mediump>	mediump_u32vec3;
+
+	/// Medium qualifier 32 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u32, mediump>	mediump_u32vec4;
+
+
+	/// High qualifier 32 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u32, highp>		highp_u32vec1;
+
+	/// High qualifier 32 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u32, highp>		highp_u32vec2;
+
+	/// High qualifier 32 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u32, highp>		highp_u32vec3;
+
+	/// High qualifier 32 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u32, highp>		highp_u32vec4;
+
+
+	/// Low qualifier 64 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u64, lowp>		lowp_u64vec1;
+
+	/// Low qualifier 64 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u64, lowp>		lowp_u64vec2;
+
+	/// Low qualifier 64 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u64, lowp>		lowp_u64vec3;
+
+	/// Low qualifier 64 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u64, lowp>		lowp_u64vec4;
+
+
+	/// Medium qualifier 64 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u64, mediump>	mediump_u64vec1;
+
+	/// Medium qualifier 64 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u64, mediump>	mediump_u64vec2;
+
+	/// Medium qualifier 64 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u64, mediump>	mediump_u64vec3;
+
+	/// Medium qualifier 64 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u64, mediump>	mediump_u64vec4;
+
+
+	/// High qualifier 64 bit unsigned integer scalar type.
+	/// @see gtc_type_precision
+	typedef vec<1, u64, highp>		highp_u64vec1;
+
+	/// High qualifier 64 bit unsigned integer vector of 2 components type.
+	/// @see gtc_type_precision
+	typedef vec<2, u64, highp>		highp_u64vec2;
+
+	/// High qualifier 64 bit unsigned integer vector of 3 components type.
+	/// @see gtc_type_precision
+	typedef vec<3, u64, highp>		highp_u64vec3;
+
+	/// High qualifier 64 bit unsigned integer vector of 4 components type.
+	/// @see gtc_type_precision
+	typedef vec<4, u64, highp>		highp_u64vec4;
+
+
+	//////////////////////
+	// Float vector types
+
+	/// 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 float32_t;
+
+	/// 32 bit single-qualifier floating-point scalar.
+	/// @see gtc_type_precision
+	typedef float32 f32;
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+		/// 64 bit double-qualifier floating-point scalar.
+		/// @see gtc_type_precision
+		typedef float64 float64_t;
+
+		/// 64 bit double-qualifier floating-point scalar.
+		/// @see gtc_type_precision
+		typedef float64 f64;
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+	/// Single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, float, defaultp> fvec1;
+
+	/// Single-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, float, defaultp> fvec2;
+
+	/// Single-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, float, defaultp> fvec3;
+
+	/// Single-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, float, defaultp> fvec4;
+
+
+	/// Single-qualifier floating-point vector of 1 component.
+	/// @see gtc_type_precision
+	typedef vec<1, f32, defaultp> f32vec1;
+
+	/// Single-qualifier floating-point vector of 2 components.
+	/// @see gtc_type_precision
+	typedef vec<2, f32, defaultp> f32vec2;
+
+	/// Single-qualifier floating-point vector of 3 components.
+	/// @see gtc_type_precision
+	typedef vec<3, f32, defaultp> f32vec3;
+
+	/// Single-qualifier floating-point vector of 4 components.
+	/// @see gtc_type_precision
+	typedef vec<4, f32, defaultp> f32vec4;
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+		/// Double-qualifier floating-point vector of 1 component.
+		/// @see gtc_type_precision
+		typedef vec<1, f64, defaultp> f64vec1;
+
+		/// Double-qualifier floating-point vector of 2 components.
+		/// @see gtc_type_precision
+		typedef vec<2, f64, defaultp> f64vec2;
+
+		/// Double-qualifier floating-point vector of 3 components.
+		/// @see gtc_type_precision
+		typedef vec<3, f64, defaultp> f64vec3;
+
+		/// Double-qualifier floating-point vector of 4 components.
+		/// @see gtc_type_precision
+		typedef vec<4, f64, defaultp> f64vec4;
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+
+	//////////////////////
+	// Float matrix types
+
+	/// Single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef detail::tmat1x1<f32> fmat1;
+
+	/// Single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, defaultp> fmat2;
+
+	/// Single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, defaultp> fmat3;
+
+	/// Single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, defaultp> fmat4;
+
+
+	/// Single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f32 fmat1x1;
+
+	/// Single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, defaultp> fmat2x2;
+
+	/// Single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, defaultp> fmat2x3;
+
+	/// Single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, defaultp> fmat2x4;
+
+	/// Single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, defaultp> fmat3x2;
+
+	/// Single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, defaultp> fmat3x3;
+
+	/// Single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, defaultp> fmat3x4;
+
+	/// Single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, defaultp> fmat4x2;
+
+	/// Single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, defaultp> fmat4x3;
+
+	/// Single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, defaultp> fmat4x4;
+
+
+	/// Single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef detail::tmat1x1<f32, defaultp> f32mat1;
+
+	/// Single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, defaultp> f32mat2;
+
+	/// Single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, defaultp> f32mat3;
+
+	/// Single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, defaultp> f32mat4;
+
+
+	/// Single-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f32 f32mat1x1;
+
+	/// Single-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f32, defaultp> f32mat2x2;
+
+	/// Single-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f32, defaultp> f32mat2x3;
+
+	/// Single-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f32, defaultp> f32mat2x4;
+
+	/// Single-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f32, defaultp> f32mat3x2;
+
+	/// Single-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f32, defaultp> f32mat3x3;
+
+	/// Single-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f32, defaultp> f32mat3x4;
+
+	/// Single-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f32, defaultp> f32mat4x2;
+
+	/// Single-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f32, defaultp> f32mat4x3;
+
+	/// Single-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f32, defaultp> f32mat4x4;
+
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+	/// Double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef detail::tmat1x1<f64, defaultp> f64mat1;
+
+	/// Double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f64, defaultp> f64mat2;
+
+	/// Double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f64, defaultp> f64mat3;
+
+	/// Double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f64, defaultp> f64mat4;
+
+
+	/// Double-qualifier floating-point 1x1 matrix.
+	/// @see gtc_type_precision
+	//typedef f64 f64mat1x1;
+
+	/// Double-qualifier floating-point 2x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 2, f64, defaultp> f64mat2x2;
+
+	/// Double-qualifier floating-point 2x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 3, f64, defaultp> f64mat2x3;
+
+	/// Double-qualifier floating-point 2x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<2, 4, f64, defaultp> f64mat2x4;
+
+	/// Double-qualifier floating-point 3x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 2, f64, defaultp> f64mat3x2;
+
+	/// Double-qualifier floating-point 3x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 3, f64, defaultp> f64mat3x3;
+
+	/// Double-qualifier floating-point 3x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<3, 4, f64, defaultp> f64mat3x4;
+
+	/// Double-qualifier floating-point 4x2 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 2, f64, defaultp> f64mat4x2;
+
+	/// Double-qualifier floating-point 4x3 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 3, f64, defaultp> f64mat4x3;
+
+	/// Double-qualifier floating-point 4x4 matrix.
+	/// @see gtc_type_precision
+	typedef mat<4, 4, f64, defaultp> f64mat4x4;
+
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+	//////////////////////////
+	// Quaternion types
+
+	/// Single-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f32, defaultp> f32quat;
+
+	/// Low single-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f32, lowp> lowp_f32quat;
+
+	/// Low double-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f64, lowp> lowp_f64quat;
+
+	/// Medium single-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f32, mediump> mediump_f32quat;
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+	/// Medium double-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f64, mediump> mediump_f64quat;
+
+	/// High single-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f32, highp> highp_f32quat;
+
+	/// High double-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f64, highp> highp_f64quat;
+
+	/// Double-qualifier floating-point quaternion.
+	/// @see gtc_type_precision
+	typedef qua<f64, defaultp> f64quat;
+
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+	/// @}
+}//namespace glm
+
+#include "type_precision.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.inl
new file mode 100644
index 000000000000..ae8091206bd4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_precision.inl
@@ -0,0 +1,6 @@
+/// @ref gtc_precision
+
+namespace glm
+{
+
+}
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.hpp
new file mode 100644
index 000000000000..d7e625aa5917
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.hpp
@@ -0,0 +1,230 @@
+/// @ref gtc_type_ptr
+/// @file glm/gtc/type_ptr.hpp
+///
+/// @see core (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtc_type_ptr GLM_GTC_type_ptr
+/// @ingroup gtc
+///
+/// Include <glm/gtc/type_ptr.hpp> to use the features of this extension.
+///
+/// Handles the interaction between pointers and vector, matrix types.
+///
+/// This extension defines an overloaded function, glm::value_ptr. It returns
+/// a pointer to the memory layout of the object. Matrix types store their values
+/// in column-major order.
+///
+/// This is useful for uploading data to matrices or copying data to buffer objects.
+///
+/// Example:
+/// @code
+/// #include <glm/glm.hpp>
+/// #include <glm/gtc/type_ptr.hpp>
+///
+/// glm::vec3 aVector(3);
+/// glm::mat4 someMatrix(1.0);
+///
+/// glUniform3fv(uniformLoc, 1, glm::value_ptr(aVector));
+/// glUniformMatrix4fv(uniformMatrixLoc, 1, GL_FALSE, glm::value_ptr(someMatrix));
+/// @endcode
+///
+/// <glm/gtc/type_ptr.hpp> need to be included to use the features of this extension.
+
+#pragma once
+
+// Dependency:
+#include "../gtc/quaternion.hpp"
+#include "../gtc/vec1.hpp"
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../mat2x2.hpp"
+#include "../mat2x3.hpp"
+#include "../mat2x4.hpp"
+#include "../mat3x2.hpp"
+#include "../mat3x3.hpp"
+#include "../mat3x4.hpp"
+#include "../mat4x2.hpp"
+#include "../mat4x3.hpp"
+#include "../mat4x4.hpp"
+#include <cstring>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_type_ptr extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_type_ptr
+	/// @{
+
+	/// Return the constant address to the data of the input parameter.
+	/// @see gtc_type_ptr
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::value_type const * value_ptr(genType const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<1, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<2, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<3, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<4, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<1, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<2, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<3, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<4, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<1, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<2, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<3, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<4, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<1, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<2, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<3, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<4, T, Q> const& v);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL vec<2, T, defaultp> make_vec2(T const * const ptr);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL vec<3, T, defaultp> make_vec3(T const * const ptr);
+
+	/// Build a vector from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL vec<4, T, defaultp> make_vec4(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<2, 2, T, defaultp> make_mat2x2(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<2, 3, T, defaultp> make_mat2x3(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<2, 4, T, defaultp> make_mat2x4(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<3, 2, T, defaultp> make_mat3x2(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<3, 3, T, defaultp> make_mat3x3(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<3, 4, T, defaultp> make_mat3x4(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 2, T, defaultp> make_mat4x2(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 3, T, defaultp> make_mat4x3(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> make_mat4x4(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<2, 2, T, defaultp> make_mat2(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<3, 3, T, defaultp> make_mat3(T const * const ptr);
+
+	/// Build a matrix from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> make_mat4(T const * const ptr);
+
+	/// Build a quaternion from a pointer.
+	/// @see gtc_type_ptr
+	template<typename T>
+	GLM_FUNC_DECL qua<T, defaultp> make_quat(T const * const ptr);
+
+	/// @}
+}//namespace glm
+
+#include "type_ptr.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.inl
new file mode 100644
index 000000000000..26b20b52e9a9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/type_ptr.inl
@@ -0,0 +1,386 @@
+/// @ref gtc_type_ptr
+
+#include <cstring>
+
+namespace glm
+{
+	/// @addtogroup gtc_type_ptr
+	/// @{
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(vec<2, T, Q> const& v)
+	{
+		return &(v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(vec<2, T, Q>& v)
+	{
+		return &(v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const * value_ptr(vec<3, T, Q> const& v)
+	{
+		return &(v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(vec<3, T, Q>& v)
+	{
+		return &(v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(vec<4, T, Q> const& v)
+	{
+		return &(v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(vec<4, T, Q>& v)
+	{
+		return &(v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<2, 2, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<2, 2, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<3, 3, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<3, 3, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<4, 4, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<4, 4, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<2, 3, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<2, 3, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<3, 2, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<3, 2, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<2, 4, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<2, 4, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<4, 2, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<4, 2, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<3, 4, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(mat<3, 4, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const* value_ptr(mat<4, 3, T, Q> const& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T * value_ptr(mat<4, 3, T, Q>& m)
+	{
+		return &(m[0].x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T const * value_ptr(qua<T, Q> const& q)
+	{
+		return &(q[0]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T* value_ptr(qua<T, Q>& q)
+	{
+		return &(q[0]);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<1, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<2, T, Q> const& v)
+	{
+		return vec<1, T, Q>(v);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<3, T, Q> const& v)
+	{
+		return vec<1, T, Q>(v);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<1, T, Q> make_vec1(vec<4, T, Q> const& v)
+	{
+		return vec<1, T, Q>(v);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<1, T, Q> const& v)
+	{
+		return vec<2, T, Q>(v.x, static_cast<T>(0));
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<2, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<3, T, Q> const& v)
+	{
+		return vec<2, T, Q>(v);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> make_vec2(vec<4, T, Q> const& v)
+	{
+		return vec<2, T, Q>(v);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<1, T, Q> const& v)
+	{
+		return vec<3, T, Q>(v.x, static_cast<T>(0), static_cast<T>(0));
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<2, T, Q> const& v)
+	{
+		return vec<3, T, Q>(v.x, v.y, static_cast<T>(0));
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<3, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> make_vec3(vec<4, T, Q> const& v)
+	{
+		return vec<3, T, Q>(v);
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<1, T, Q> const& v)
+	{
+		return vec<4, T, Q>(v.x, static_cast<T>(0), static_cast<T>(0), static_cast<T>(1));
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<2, T, Q> const& v)
+	{
+		return vec<4, T, Q>(v.x, v.y, static_cast<T>(0), static_cast<T>(1));
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<3, T, Q> const& v)
+	{
+		return vec<4, T, Q>(v.x, v.y, v.z, static_cast<T>(1));
+	}
+
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> make_vec4(vec<4, T, Q> const& v)
+	{
+		return v;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<2, T, defaultp> make_vec2(T const *const ptr)
+	{
+		vec<2, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(vec<2, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<3, T, defaultp> make_vec3(T const *const ptr)
+	{
+		vec<3, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(vec<3, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER vec<4, T, defaultp> make_vec4(T const *const ptr)
+	{
+		vec<4, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(vec<4, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, defaultp> make_mat2x2(T const *const ptr)
+	{
+		mat<2, 2, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<2, 2, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<2, 3, T, defaultp> make_mat2x3(T const *const ptr)
+	{
+		mat<2, 3, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<2, 3, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<2, 4, T, defaultp> make_mat2x4(T const *const ptr)
+	{
+		mat<2, 4, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<2, 4, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<3, 2, T, defaultp> make_mat3x2(T const *const ptr)
+	{
+		mat<3, 2, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<3, 2, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, defaultp> make_mat3x3(T const *const ptr)
+	{
+		mat<3, 3, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<3, 3, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<3, 4, T, defaultp> make_mat3x4(T const *const ptr)
+	{
+		mat<3, 4, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<3, 4, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 2, T, defaultp> make_mat4x2(T const *const ptr)
+	{
+		mat<4, 2, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<4, 2, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 3, T, defaultp> make_mat4x3(T const *const ptr)
+	{
+		mat<4, 3, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<4, 3, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> make_mat4x4(T const *const ptr)
+	{
+		mat<4, 4, T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(mat<4, 4, T, defaultp>));
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, defaultp> make_mat2(T const *const ptr)
+	{
+		return make_mat2x2(ptr);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, defaultp> make_mat3(T const *const ptr)
+	{
+		return make_mat3x3(ptr);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> make_mat4(T const *const ptr)
+	{
+		return make_mat4x4(ptr);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER qua<T, defaultp> make_quat(T const *const ptr)
+	{
+		qua<T, defaultp> Result;
+		memcpy(value_ptr(Result), ptr, sizeof(qua<T, defaultp>));
+		return Result;
+	}
+
+	/// @}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.hpp
new file mode 100644
index 000000000000..7b918f0fa764
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.hpp
@@ -0,0 +1,155 @@
+/// @ref gtc_ulp
+/// @file glm/gtc/ulp.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_ulp GLM_GTC_ulp
+/// @ingroup gtc
+///
+/// Include <glm/gtc/ulp.hpp> to use the features of this extension.
+///
+/// Allow the measurement of the accuracy of a function against a reference
+/// implementation. This extension works on floating-point data and provide results
+/// in ULP.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/_vectorize.hpp"
+#include "../ext/scalar_int_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_ulp extension included")
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtc_ulp
+	/// @{
+
+	/// Return the next ULP value(s) after the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see gtc_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType next_float(genType x);
+
+	/// Return the previous ULP value(s) before the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see gtc_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType prev_float(genType x);
+
+	/// Return the value(s) ULP distance after the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see gtc_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType next_float(genType x, int ULPs);
+
+	/// Return the value(s) ULP distance before the input value(s).
+	///
+	/// @tparam genType A floating-point scalar type.
+	///
+	/// @see gtc_ulp
+	template<typename genType>
+	GLM_FUNC_DECL genType prev_float(genType x, int ULPs);
+
+	/// Return the distance in the number of ULP between 2 single-precision floating-point scalars.
+	///
+	/// @see gtc_ulp
+	GLM_FUNC_DECL int float_distance(float x, float y);
+
+	/// Return the distance in the number of ULP between 2 double-precision floating-point scalars.
+	///
+	/// @see gtc_ulp
+	GLM_FUNC_DECL int64 float_distance(double x, double y);
+
+	/// Return the next ULP value(s) after the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> next_float(vec<L, T, Q> const& x);
+
+	/// Return the value(s) ULP distance after the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> next_float(vec<L, T, Q> const& x, int ULPs);
+
+	/// Return the value(s) ULP distance after the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> next_float(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs);
+
+	/// Return the previous ULP value(s) before the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prev_float(vec<L, T, Q> const& x);
+
+	/// Return the value(s) ULP distance before the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prev_float(vec<L, T, Q> const& x, int ULPs);
+
+	/// Return the value(s) ULP distance before the input value(s).
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> prev_float(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs);
+
+	/// Return the distance in the number of ULP between 2 single-precision floating-point scalars.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> float_distance(vec<L, float, Q> const& x, vec<L, float, Q> const& y);
+
+	/// Return the distance in the number of ULP between 2 double-precision floating-point scalars.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see gtc_ulp
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int64, Q> float_distance(vec<L, double, Q> const& x, vec<L, double, Q> const& y);
+
+	/// @}
+}//namespace glm
+
+#include "ulp.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.inl b/thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.inl
new file mode 100644
index 000000000000..836c84b4a250
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/ulp.inl
@@ -0,0 +1,173 @@
+/// @ref gtc_ulp
+
+#include "../ext/scalar_ulp.hpp"
+
+namespace glm
+{
+	template<>
+	GLM_FUNC_QUALIFIER float next_float(float x)
+	{
+#		if GLM_HAS_CXX11_STL
+		return std::nextafter(x, std::numeric_limits<float>::max());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+		return detail::nextafterf(x, FLT_MAX);
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+		return __builtin_nextafterf(x, FLT_MAX);
+#		else
+		return nextafterf(x, FLT_MAX);
+#		endif
+	}
+
+	template<>
+	GLM_FUNC_QUALIFIER double next_float(double x)
+	{
+#		if GLM_HAS_CXX11_STL
+		return std::nextafter(x, std::numeric_limits<double>::max());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+		return detail::nextafter(x, std::numeric_limits<double>::max());
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+		return __builtin_nextafter(x, DBL_MAX);
+#		else
+		return nextafter(x, DBL_MAX);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T next_float(T x, int ULPs)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'next_float' only accept floating-point input");
+		assert(ULPs >= 0);
+
+		T temp = x;
+		for (int i = 0; i < ULPs; ++i)
+			temp = next_float(temp);
+		return temp;
+	}
+
+	GLM_FUNC_QUALIFIER float prev_float(float x)
+	{
+#		if GLM_HAS_CXX11_STL
+		return std::nextafter(x, std::numeric_limits<float>::min());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+		return detail::nextafterf(x, FLT_MIN);
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+		return __builtin_nextafterf(x, FLT_MIN);
+#		else
+		return nextafterf(x, FLT_MIN);
+#		endif
+	}
+
+	GLM_FUNC_QUALIFIER double prev_float(double x)
+	{
+#		if GLM_HAS_CXX11_STL
+		return std::nextafter(x, std::numeric_limits<double>::min());
+#		elif((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & GLM_COMPILER_INTEL) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+		return _nextafter(x, DBL_MIN);
+#		elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+		return __builtin_nextafter(x, DBL_MIN);
+#		else
+		return nextafter(x, DBL_MIN);
+#		endif
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T prev_float(T x, int ULPs)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'prev_float' only accept floating-point input");
+		assert(ULPs >= 0);
+
+		T temp = x;
+		for (int i = 0; i < ULPs; ++i)
+			temp = prev_float(temp);
+		return temp;
+	}
+
+	GLM_FUNC_QUALIFIER int float_distance(float x, float y)
+	{
+		detail::float_t<float> const a(x);
+		detail::float_t<float> const b(y);
+
+		return abs(a.i - b.i);
+	}
+
+	GLM_FUNC_QUALIFIER int64 float_distance(double x, double y)
+	{
+		detail::float_t<double> const a(x);
+		detail::float_t<double> const b(y);
+
+		return abs(a.i - b.i);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> next_float(vec<L, T, Q> const& x)
+	{
+		vec<L, T, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = next_float(x[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> next_float(vec<L, T, Q> const& x, int ULPs)
+	{
+		vec<L, T, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = next_float(x[i], ULPs);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> next_float(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs)
+	{
+		vec<L, T, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = next_float(x[i], ULPs[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prev_float(vec<L, T, Q> const& x)
+	{
+		vec<L, T, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = prev_float(x[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prev_float(vec<L, T, Q> const& x, int ULPs)
+	{
+		vec<L, T, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = prev_float(x[i], ULPs);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> prev_float(vec<L, T, Q> const& x, vec<L, int, Q> const& ULPs)
+	{
+		vec<L, T, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = prev_float(x[i], ULPs[i]);
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int, Q> float_distance(vec<L, float, Q> const& x, vec<L, float, Q> const& y)
+	{
+		vec<L, int, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = float_distance(x[i], y[i]);
+		return Result;
+	}
+
+	template<length_t L, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, int64, Q> float_distance(vec<L, double, Q> const& x, vec<L, double, Q> const& y)
+	{
+		vec<L, int64, Q> Result;
+		for (length_t i = 0, n = Result.length(); i < n; ++i)
+			Result[i] = float_distance(x[i], y[i]);
+		return Result;
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtc/vec1.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtc/vec1.hpp
new file mode 100644
index 000000000000..63697a215750
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtc/vec1.hpp
@@ -0,0 +1,30 @@
+/// @ref gtc_vec1
+/// @file glm/gtc/vec1.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtc_vec1 GLM_GTC_vec1
+/// @ingroup gtc
+///
+/// Include <glm/gtc/vec1.hpp> to use the features of this extension.
+///
+/// Add vec1, ivec1, uvec1 and bvec1 types.
+
+#pragma once
+
+// Dependency:
+#include "../ext/vector_bool1.hpp"
+#include "../ext/vector_bool1_precision.hpp"
+#include "../ext/vector_float1.hpp"
+#include "../ext/vector_float1_precision.hpp"
+#include "../ext/vector_double1.hpp"
+#include "../ext/vector_double1_precision.hpp"
+#include "../ext/vector_int1.hpp"
+#include "../ext/vector_int1_sized.hpp"
+#include "../ext/vector_uint1.hpp"
+#include "../ext/vector_uint1_sized.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	pragma message("GLM: GLM_GTC_vec1 extension included")
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.hpp
new file mode 100644
index 000000000000..4c036add2186
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.hpp
@@ -0,0 +1,207 @@
+/// @ref gtx_associated_min_max
+/// @file glm/gtx/associated_min_max.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_associated_min_max GLM_GTX_associated_min_max
+/// @ingroup gtx
+///
+/// Include <glm/gtx/associated_min_max.hpp> to use the features of this extension.
+///
+/// @brief Min and max functions that return associated values not the compared ones.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_associated_min_max is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_associated_min_max extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_associated_min_max
+	/// @{
+
+	/// Minimum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<typename T, typename U>
+	GLM_FUNC_DECL U associatedMin(T x, U a, T y, U b);
+
+	/// Minimum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+		vec<L, T, Q> const& y, vec<L, U, Q> const& b);
+
+	/// Minimum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		T x, const vec<L, U, Q>& a,
+		T y, const vec<L, U, Q>& b);
+
+	/// Minimum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		vec<L, T, Q> const& x, U a,
+		vec<L, T, Q> const& y, U b);
+
+	/// Minimum comparison between 3 variables and returns 3 associated variable values
+	/// @see gtx_associated_min_max
+	template<typename T, typename U>
+	GLM_FUNC_DECL U associatedMin(
+		T x, U a,
+		T y, U b,
+		T z, U c);
+
+	/// Minimum comparison between 3 variables and returns 3 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+		vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+		vec<L, T, Q> const& z, vec<L, U, Q> const& c);
+
+	/// Minimum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<typename T, typename U>
+	GLM_FUNC_DECL U associatedMin(
+		T x, U a,
+		T y, U b,
+		T z, U c,
+		T w, U d);
+
+	/// Minimum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+		vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+		vec<L, T, Q> const& z, vec<L, U, Q> const& c,
+		vec<L, T, Q> const& w, vec<L, U, Q> const& d);
+
+	/// Minimum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		T x, vec<L, U, Q> const& a,
+		T y, vec<L, U, Q> const& b,
+		T z, vec<L, U, Q> const& c,
+		T w, vec<L, U, Q> const& d);
+
+	/// Minimum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMin(
+		vec<L, T, Q> const& x, U a,
+		vec<L, T, Q> const& y, U b,
+		vec<L, T, Q> const& z, U c,
+		vec<L, T, Q> const& w, U d);
+
+	/// Maximum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<typename T, typename U>
+	GLM_FUNC_DECL U associatedMax(T x, U a, T y, U b);
+
+	/// Maximum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+		vec<L, T, Q> const& y, vec<L, U, Q> const& b);
+
+	/// Maximum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> associatedMax(
+		T x, vec<L, U, Q> const& a,
+		T y, vec<L, U, Q> const& b);
+
+	/// Maximum comparison between 2 variables and returns 2 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		vec<L, T, Q> const& x, U a,
+		vec<L, T, Q> const& y, U b);
+
+	/// Maximum comparison between 3 variables and returns 3 associated variable values
+	/// @see gtx_associated_min_max
+	template<typename T, typename U>
+	GLM_FUNC_DECL U associatedMax(
+		T x, U a,
+		T y, U b,
+		T z, U c);
+
+	/// Maximum comparison between 3 variables and returns 3 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+		vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+		vec<L, T, Q> const& z, vec<L, U, Q> const& c);
+
+	/// Maximum comparison between 3 variables and returns 3 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> associatedMax(
+		T x, vec<L, U, Q> const& a,
+		T y, vec<L, U, Q> const& b,
+		T z, vec<L, U, Q> const& c);
+
+	/// Maximum comparison between 3 variables and returns 3 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		vec<L, T, Q> const& x, U a,
+		vec<L, T, Q> const& y, U b,
+		vec<L, T, Q> const& z, U c);
+
+	/// Maximum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<typename T, typename U>
+	GLM_FUNC_DECL U associatedMax(
+		T x, U a,
+		T y, U b,
+		T z, U c,
+		T w, U d);
+
+	/// Maximum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+		vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+		vec<L, T, Q> const& z, vec<L, U, Q> const& c,
+		vec<L, T, Q> const& w, vec<L, U, Q> const& d);
+
+	/// Maximum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		T x, vec<L, U, Q> const& a,
+		T y, vec<L, U, Q> const& b,
+		T z, vec<L, U, Q> const& c,
+		T w, vec<L, U, Q> const& d);
+
+	/// Maximum comparison between 4 variables and returns 4 associated variable values
+	/// @see gtx_associated_min_max
+	template<length_t L, typename T, typename U, qualifier Q>
+	GLM_FUNC_DECL vec<L, U, Q> associatedMax(
+		vec<L, T, Q> const& x, U a,
+		vec<L, T, Q> const& y, U b,
+		vec<L, T, Q> const& z, U c,
+		vec<L, T, Q> const& w, U d);
+
+	/// @}
+} //namespace glm
+
+#include "associated_min_max.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.inl
new file mode 100644
index 000000000000..f09f5bb74c25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/associated_min_max.inl
@@ -0,0 +1,354 @@
+/// @ref gtx_associated_min_max
+
+namespace glm{
+
+// Min comparison between 2 variables
+template<typename T, typename U>
+GLM_FUNC_QUALIFIER U associatedMin(T x, U a, T y, U b)
+{
+	return x < y ? a : b;
+}
+
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+	vec<L, T, Q> const& y, vec<L, U, Q> const& b
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] < y[i] ? a[i] : b[i];
+	return Result;
+}
+
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	T x, const vec<L, U, Q>& a,
+	T y, const vec<L, U, Q>& b
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x < y ? a[i] : b[i];
+	return Result;
+}
+
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	vec<L, T, Q> const& x, U a,
+	vec<L, T, Q> const& y, U b
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] < y[i] ? a : b;
+	return Result;
+}
+
+// Min comparison between 3 variables
+template<typename T, typename U>
+GLM_FUNC_QUALIFIER U associatedMin
+(
+	T x, U a,
+	T y, U b,
+	T z, U c
+)
+{
+	U Result = x < y ? (x < z ? a : c) : (y < z ? b : c);
+	return Result;
+}
+
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+	vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+	vec<L, T, Q> const& z, vec<L, U, Q> const& c
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] < y[i] ? (x[i] < z[i] ? a[i] : c[i]) : (y[i] < z[i] ? b[i] : c[i]);
+	return Result;
+}
+
+// Min comparison between 4 variables
+template<typename T, typename U>
+GLM_FUNC_QUALIFIER U associatedMin
+(
+	T x, U a,
+	T y, U b,
+	T z, U c,
+	T w, U d
+)
+{
+	T Test1 = min(x, y);
+	T Test2 = min(z, w);
+	U Result1 = x < y ? a : b;
+	U Result2 = z < w ? c : d;
+	U Result = Test1 < Test2 ? Result1 : Result2;
+	return Result;
+}
+
+// Min comparison between 4 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+	vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+	vec<L, T, Q> const& z, vec<L, U, Q> const& c,
+	vec<L, T, Q> const& w, vec<L, U, Q> const& d
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+	{
+		T Test1 = min(x[i], y[i]);
+		T Test2 = min(z[i], w[i]);
+		U Result1 = x[i] < y[i] ? a[i] : b[i];
+		U Result2 = z[i] < w[i] ? c[i] : d[i];
+		Result[i] = Test1 < Test2 ? Result1 : Result2;
+	}
+	return Result;
+}
+
+// Min comparison between 4 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	T x, vec<L, U, Q> const& a,
+	T y, vec<L, U, Q> const& b,
+	T z, vec<L, U, Q> const& c,
+	T w, vec<L, U, Q> const& d
+)
+{
+	T Test1 = min(x, y);
+	T Test2 = min(z, w);
+
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+	{
+		U Result1 = x < y ? a[i] : b[i];
+		U Result2 = z < w ? c[i] : d[i];
+		Result[i] = Test1 < Test2 ? Result1 : Result2;
+	}
+	return Result;
+}
+
+// Min comparison between 4 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMin
+(
+	vec<L, T, Q> const& x, U a,
+	vec<L, T, Q> const& y, U b,
+	vec<L, T, Q> const& z, U c,
+	vec<L, T, Q> const& w, U d
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+	{
+		T Test1 = min(x[i], y[i]);
+		T Test2 = min(z[i], w[i]);
+		U Result1 = x[i] < y[i] ? a : b;
+		U Result2 = z[i] < w[i] ? c : d;
+		Result[i] = Test1 < Test2 ? Result1 : Result2;
+	}
+	return Result;
+}
+
+// Max comparison between 2 variables
+template<typename T, typename U>
+GLM_FUNC_QUALIFIER U associatedMax(T x, U a, T y, U b)
+{
+	return x > y ? a : b;
+}
+
+// Max comparison between 2 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+	vec<L, T, Q> const& y, vec<L, U, Q> const& b
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] > y[i] ? a[i] : b[i];
+	return Result;
+}
+
+// Max comparison between 2 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, T, Q> associatedMax
+(
+	T x, vec<L, U, Q> const& a,
+	T y, vec<L, U, Q> const& b
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x > y ? a[i] : b[i];
+	return Result;
+}
+
+// Max comparison between 2 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	vec<L, T, Q> const& x, U a,
+	vec<L, T, Q> const& y, U b
+)
+{
+	vec<L, T, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] > y[i] ? a : b;
+	return Result;
+}
+
+// Max comparison between 3 variables
+template<typename T, typename U>
+GLM_FUNC_QUALIFIER U associatedMax
+(
+	T x, U a,
+	T y, U b,
+	T z, U c
+)
+{
+	U Result = x > y ? (x > z ? a : c) : (y > z ? b : c);
+	return Result;
+}
+
+// Max comparison between 3 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+	vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+	vec<L, T, Q> const& z, vec<L, U, Q> const& c
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] > y[i] ? (x[i] > z[i] ? a[i] : c[i]) : (y[i] > z[i] ? b[i] : c[i]);
+	return Result;
+}
+
+// Max comparison between 3 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, T, Q> associatedMax
+(
+	T x, vec<L, U, Q> const& a,
+	T y, vec<L, U, Q> const& b,
+	T z, vec<L, U, Q> const& c
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x > y ? (x > z ? a[i] : c[i]) : (y > z ? b[i] : c[i]);
+	return Result;
+}
+
+// Max comparison between 3 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	vec<L, T, Q> const& x, U a,
+	vec<L, T, Q> const& y, U b,
+	vec<L, T, Q> const& z, U c
+)
+{
+	vec<L, T, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+		Result[i] = x[i] > y[i] ? (x[i] > z[i] ? a : c) : (y[i] > z[i] ? b : c);
+	return Result;
+}
+
+// Max comparison between 4 variables
+template<typename T, typename U>
+GLM_FUNC_QUALIFIER U associatedMax
+(
+	T x, U a,
+	T y, U b,
+	T z, U c,
+	T w, U d
+)
+{
+	T Test1 = max(x, y);
+	T Test2 = max(z, w);
+	U Result1 = x > y ? a : b;
+	U Result2 = z > w ? c : d;
+	U Result = Test1 > Test2 ? Result1 : Result2;
+	return Result;
+}
+
+// Max comparison between 4 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	vec<L, T, Q> const& x, vec<L, U, Q> const& a,
+	vec<L, T, Q> const& y, vec<L, U, Q> const& b,
+	vec<L, T, Q> const& z, vec<L, U, Q> const& c,
+	vec<L, T, Q> const& w, vec<L, U, Q> const& d
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+	{
+		T Test1 = max(x[i], y[i]);
+		T Test2 = max(z[i], w[i]);
+		U Result1 = x[i] > y[i] ? a[i] : b[i];
+		U Result2 = z[i] > w[i] ? c[i] : d[i];
+		Result[i] = Test1 > Test2 ? Result1 : Result2;
+	}
+	return Result;
+}
+
+// Max comparison between 4 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	T x, vec<L, U, Q> const& a,
+	T y, vec<L, U, Q> const& b,
+	T z, vec<L, U, Q> const& c,
+	T w, vec<L, U, Q> const& d
+)
+{
+	T Test1 = max(x, y);
+	T Test2 = max(z, w);
+
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+	{
+		U Result1 = x > y ? a[i] : b[i];
+		U Result2 = z > w ? c[i] : d[i];
+		Result[i] = Test1 > Test2 ? Result1 : Result2;
+	}
+	return Result;
+}
+
+// Max comparison between 4 variables
+template<length_t L, typename T, typename U, qualifier Q>
+GLM_FUNC_QUALIFIER vec<L, U, Q> associatedMax
+(
+	vec<L, T, Q> const& x, U a,
+	vec<L, T, Q> const& y, U b,
+	vec<L, T, Q> const& z, U c,
+	vec<L, T, Q> const& w, U d
+)
+{
+	vec<L, U, Q> Result;
+	for(length_t i = 0, n = Result.length(); i < n; ++i)
+	{
+		T Test1 = max(x[i], y[i]);
+		T Test2 = max(z[i], w[i]);
+		U Result1 = x[i] > y[i] ? a : b;
+		U Result2 = z[i] > w[i] ? c : d;
+		Result[i] = Test1 > Test2 ? Result1 : Result2;
+	}
+	return Result;
+}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/bit.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/bit.hpp
new file mode 100644
index 000000000000..60a7aef1b463
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/bit.hpp
@@ -0,0 +1,98 @@
+/// @ref gtx_bit
+/// @file glm/gtx/bit.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_bit GLM_GTX_bit
+/// @ingroup gtx
+///
+/// Include <glm/gtx/bit.hpp> to use the features of this extension.
+///
+/// Allow to perform bit operations on integer values
+
+#pragma once
+
+// Dependencies
+#include "../gtc/bitfield.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_bit is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_bit extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_bit
+	/// @{
+
+	/// @see gtx_bit
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType highestBitValue(genIUType Value);
+
+	/// @see gtx_bit
+	template<typename genIUType>
+	GLM_FUNC_DECL genIUType lowestBitValue(genIUType Value);
+
+	/// Find the highest bit set to 1 in a integer variable and return its value.
+	///
+	/// @see gtx_bit
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> highestBitValue(vec<L, T, Q> const& value);
+
+	/// Return the power of two number which value is just higher the input value.
+	/// Deprecated, use ceilPowerOfTwo from GTC_round instead
+	///
+	/// @see gtc_round
+	/// @see gtx_bit
+	template<typename genIUType>
+	GLM_DEPRECATED GLM_FUNC_DECL genIUType powerOfTwoAbove(genIUType Value);
+
+	/// Return the power of two number which value is just higher the input value.
+	/// Deprecated, use ceilPowerOfTwo from GTC_round instead
+	///
+	/// @see gtc_round
+	/// @see gtx_bit
+	template<length_t L, typename T, qualifier Q>
+	GLM_DEPRECATED GLM_FUNC_DECL vec<L, T, Q> powerOfTwoAbove(vec<L, T, Q> const& value);
+
+	/// Return the power of two number which value is just lower the input value.
+	/// Deprecated, use floorPowerOfTwo from GTC_round instead
+	///
+	/// @see gtc_round
+	/// @see gtx_bit
+	template<typename genIUType>
+	GLM_DEPRECATED GLM_FUNC_DECL genIUType powerOfTwoBelow(genIUType Value);
+
+	/// Return the power of two number which value is just lower the input value.
+	/// Deprecated, use floorPowerOfTwo from GTC_round instead
+	///
+	/// @see gtc_round
+	/// @see gtx_bit
+	template<length_t L, typename T, qualifier Q>
+	GLM_DEPRECATED GLM_FUNC_DECL vec<L, T, Q> powerOfTwoBelow(vec<L, T, Q> const& value);
+
+	/// Return the power of two number which value is the closet to the input value.
+	/// Deprecated, use roundPowerOfTwo from GTC_round instead
+	///
+	/// @see gtc_round
+	/// @see gtx_bit
+	template<typename genIUType>
+	GLM_DEPRECATED GLM_FUNC_DECL genIUType powerOfTwoNearest(genIUType Value);
+
+	/// Return the power of two number which value is the closet to the input value.
+	/// Deprecated, use roundPowerOfTwo from GTC_round instead
+	///
+	/// @see gtc_round
+	/// @see gtx_bit
+	template<length_t L, typename T, qualifier Q>
+	GLM_DEPRECATED GLM_FUNC_DECL vec<L, T, Q> powerOfTwoNearest(vec<L, T, Q> const& value);
+
+	/// @}
+} //namespace glm
+
+
+#include "bit.inl"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/bit.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/bit.inl
new file mode 100644
index 000000000000..621b6262406d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/bit.inl
@@ -0,0 +1,92 @@
+/// @ref gtx_bit
+
+namespace glm
+{
+	///////////////////
+	// highestBitValue
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType highestBitValue(genIUType Value)
+	{
+		genIUType tmp = Value;
+		genIUType result = genIUType(0);
+		while(tmp)
+		{
+			result = (tmp & (~tmp + 1)); // grab lowest bit
+			tmp &= ~result; // clear lowest bit
+		}
+		return result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> highestBitValue(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(highestBitValue, v);
+	}
+
+	///////////////////
+	// lowestBitValue
+
+	template<typename genIUType>
+	GLM_FUNC_QUALIFIER genIUType lowestBitValue(genIUType Value)
+	{
+		return (Value & (~Value + 1));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> lowestBitValue(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(lowestBitValue, v);
+	}
+
+	///////////////////
+	// powerOfTwoAbove
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType powerOfTwoAbove(genType value)
+	{
+		return isPowerOfTwo(value) ? value : highestBitValue(value) << 1;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> powerOfTwoAbove(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(powerOfTwoAbove, v);
+	}
+
+	///////////////////
+	// powerOfTwoBelow
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType powerOfTwoBelow(genType value)
+	{
+		return isPowerOfTwo(value) ? value : highestBitValue(value);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> powerOfTwoBelow(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(powerOfTwoBelow, v);
+	}
+
+	/////////////////////
+	// powerOfTwoNearest
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType powerOfTwoNearest(genType value)
+	{
+		if(isPowerOfTwo(value))
+			return value;
+
+		genType const prev = highestBitValue(value);
+		genType const next = prev << 1;
+		return (next - value) < (value - prev) ? next : prev;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> powerOfTwoNearest(vec<L, T, Q> const& v)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(powerOfTwoNearest, v);
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.hpp
new file mode 100644
index 000000000000..de6dbbff9447
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.hpp
@@ -0,0 +1,49 @@
+/// @ref gtx_closest_point
+/// @file glm/gtx/closest_point.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_closest_point GLM_GTX_closest_point
+/// @ingroup gtx
+///
+/// Include <glm/gtx/closest_point.hpp> to use the features of this extension.
+///
+/// Find the point on a straight line which is the closet of a point.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_closest_point is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_closest_point extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_closest_point
+	/// @{
+
+	/// Find the point on a straight line which is the closet of a point.
+	/// @see gtx_closest_point
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> closestPointOnLine(
+		vec<3, T, Q> const& point,
+		vec<3, T, Q> const& a,
+		vec<3, T, Q> const& b);
+
+	/// 2d lines work as well
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> closestPointOnLine(
+		vec<2, T, Q> const& point,
+		vec<2, T, Q> const& a,
+		vec<2, T, Q> const& b);
+
+	/// @}
+}// namespace glm
+
+#include "closest_point.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.inl
new file mode 100644
index 000000000000..0a39b042b88c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/closest_point.inl
@@ -0,0 +1,45 @@
+/// @ref gtx_closest_point
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> closestPointOnLine
+	(
+		vec<3, T, Q> const& point,
+		vec<3, T, Q> const& a,
+		vec<3, T, Q> const& b
+	)
+	{
+		T LineLength = distance(a, b);
+		vec<3, T, Q> Vector = point - a;
+		vec<3, T, Q> LineDirection = (b - a) / LineLength;
+
+		// Project Vector to LineDirection to get the distance of point from a
+		T Distance = dot(Vector, LineDirection);
+
+		if(Distance <= T(0)) return a;
+		if(Distance >= LineLength) return b;
+		return a + LineDirection * Distance;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> closestPointOnLine
+	(
+		vec<2, T, Q> const& point,
+		vec<2, T, Q> const& a,
+		vec<2, T, Q> const& b
+	)
+	{
+		T LineLength = distance(a, b);
+		vec<2, T, Q> Vector = point - a;
+		vec<2, T, Q> LineDirection = (b - a) / LineLength;
+
+		// Project Vector to LineDirection to get the distance of point from a
+		T Distance = dot(Vector, LineDirection);
+
+		if(Distance <= T(0)) return a;
+		if(Distance >= LineLength) return b;
+		return a + LineDirection * Distance;
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.hpp
new file mode 100644
index 000000000000..96ded2a2770f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.hpp
@@ -0,0 +1,54 @@
+/// @ref gtx_color_encoding
+/// @file glm/gtx/color_encoding.hpp
+///
+/// @see core (dependence)
+/// @see gtx_color_encoding (dependence)
+///
+/// @defgroup gtx_color_encoding GLM_GTX_color_encoding
+/// @ingroup gtx
+///
+/// Include <glm/gtx/color_encoding.hpp> to use the features of this extension.
+///
+/// @brief Allow to perform bit operations on integer values
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../vec3.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTC_color_encoding is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTC_color_encoding extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_color_encoding
+	/// @{
+
+	/// Convert a linear sRGB color to D65 YUV.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> convertLinearSRGBToD65XYZ(vec<3, T, Q> const& ColorLinearSRGB);
+
+	/// Convert a linear sRGB color to D50 YUV.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> convertLinearSRGBToD50XYZ(vec<3, T, Q> const& ColorLinearSRGB);
+
+	/// Convert a D65 YUV color to linear sRGB.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> convertD65XYZToLinearSRGB(vec<3, T, Q> const& ColorD65XYZ);
+
+	/// Convert a D65 YUV color to D50 YUV.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> convertD65XYZToD50XYZ(vec<3, T, Q> const& ColorD65XYZ);
+
+	/// @}
+} //namespace glm
+
+#include "color_encoding.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.inl
new file mode 100644
index 000000000000..e50fa3efa42c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_encoding.inl
@@ -0,0 +1,45 @@
+/// @ref gtx_color_encoding
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> convertLinearSRGBToD65XYZ(vec<3, T, Q> const& ColorLinearSRGB)
+	{
+		vec<3, T, Q> const M(0.490f, 0.17697f, 0.2f);
+		vec<3, T, Q> const N(0.31f,  0.8124f, 0.01063f);
+		vec<3, T, Q> const O(0.490f, 0.01f, 0.99f);
+
+		return (M * ColorLinearSRGB + N * ColorLinearSRGB + O * ColorLinearSRGB) * static_cast<T>(5.650675255693055f);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> convertLinearSRGBToD50XYZ(vec<3, T, Q> const& ColorLinearSRGB)
+	{
+		vec<3, T, Q> const M(0.436030342570117f, 0.222438466210245f, 0.013897440074263f);
+		vec<3, T, Q> const N(0.385101860087134f, 0.716942745571917f, 0.097076381494207f);
+		vec<3, T, Q> const O(0.143067806654203f, 0.060618777416563f, 0.713926257896652f);
+
+		return M * ColorLinearSRGB + N * ColorLinearSRGB + O * ColorLinearSRGB;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> convertD65XYZToLinearSRGB(vec<3, T, Q> const& ColorD65XYZ)
+	{
+		vec<3, T, Q> const M(0.41847f, -0.091169f, 0.0009209f);
+		vec<3, T, Q> const N(-0.15866f, 0.25243f, 0.015708f);
+		vec<3, T, Q> const O(0.0009209f, -0.0025498f, 0.1786f);
+
+		return M * ColorD65XYZ + N * ColorD65XYZ + O * ColorD65XYZ;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> convertD65XYZToD50XYZ(vec<3, T, Q> const& ColorD65XYZ)
+	{
+		vec<3, T, Q> const M(+1.047844353856414f, +0.029549007606644f, -0.009250984365223f);
+		vec<3, T, Q> const N(+0.022898981050086f, +0.990508028941971f, +0.015072338237051f);
+		vec<3, T, Q> const O(-0.050206647741605f, -0.017074711360960f, +0.751717835079977f);
+
+		return M * ColorD65XYZ + N * ColorD65XYZ + O * ColorD65XYZ;
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.hpp
new file mode 100644
index 000000000000..a63439214908
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.hpp
@@ -0,0 +1,72 @@
+/// @ref gtx_color_space
+/// @file glm/gtx/color_space.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_color_space GLM_GTX_color_space
+/// @ingroup gtx
+///
+/// Include <glm/gtx/color_space.hpp> to use the features of this extension.
+///
+/// Related to RGB to HSV conversions and operations.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_color_space is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_color_space extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_color_space
+	/// @{
+
+	/// Converts a color from HSV color space to its color in RGB color space.
+	/// @see gtx_color_space
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rgbColor(
+		vec<3, T, Q> const& hsvValue);
+
+	/// Converts a color from RGB color space to its color in HSV color space.
+	/// @see gtx_color_space
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> hsvColor(
+		vec<3, T, Q> const& rgbValue);
+
+	/// Build a saturation matrix.
+	/// @see gtx_color_space
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> saturation(
+		T const s);
+
+	/// Modify the saturation of a color.
+	/// @see gtx_color_space
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> saturation(
+		T const s,
+		vec<3, T, Q> const& color);
+
+	/// Modify the saturation of a color.
+	/// @see gtx_color_space
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> saturation(
+		T const s,
+		vec<4, T, Q> const& color);
+
+	/// Compute color luminosity associating ratios (0.33, 0.59, 0.11) to RGB canals.
+	/// @see gtx_color_space
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T luminosity(
+		vec<3, T, Q> const& color);
+
+	/// @}
+}//namespace glm
+
+#include "color_space.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.inl
new file mode 100644
index 000000000000..0a7059fabb7b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space.inl
@@ -0,0 +1,141 @@
+/// @ref gtx_color_space
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rgbColor(const vec<3, T, Q>& hsvColor)
+	{
+		vec<3, T, Q> hsv = hsvColor;
+		vec<3, T, Q> rgbColor;
+
+		if(hsv.y == static_cast<T>(0))
+			// achromatic (grey)
+			rgbColor = vec<3, T, Q>(hsv.z);
+		else
+		{
+			T sector = floor(hsv.x * (T(1) / T(60)));
+			T frac = (hsv.x * (T(1) / T(60))) - sector;
+			// factorial part of h
+			T o = hsv.z * (T(1) - hsv.y);
+			T p = hsv.z * (T(1) - hsv.y * frac);
+			T q = hsv.z * (T(1) - hsv.y * (T(1) - frac));
+
+			switch(int(sector))
+			{
+			default:
+			case 0:
+				rgbColor.r = hsv.z;
+				rgbColor.g = q;
+				rgbColor.b = o;
+				break;
+			case 1:
+				rgbColor.r = p;
+				rgbColor.g = hsv.z;
+				rgbColor.b = o;
+				break;
+			case 2:
+				rgbColor.r = o;
+				rgbColor.g = hsv.z;
+				rgbColor.b = q;
+				break;
+			case 3:
+				rgbColor.r = o;
+				rgbColor.g = p;
+				rgbColor.b = hsv.z;
+				break;
+			case 4:
+				rgbColor.r = q;
+				rgbColor.g = o;
+				rgbColor.b = hsv.z;
+				break;
+			case 5:
+				rgbColor.r = hsv.z;
+				rgbColor.g = o;
+				rgbColor.b = p;
+				break;
+			}
+		}
+
+		return rgbColor;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> hsvColor(const vec<3, T, Q>& rgbColor)
+	{
+		vec<3, T, Q> hsv = rgbColor;
+		T Min   = min(min(rgbColor.r, rgbColor.g), rgbColor.b);
+		T Max   = max(max(rgbColor.r, rgbColor.g), rgbColor.b);
+		T Delta = Max - Min;
+
+		hsv.z = Max;
+
+		if(Max != static_cast<T>(0))
+		{
+			hsv.y = Delta / hsv.z;
+			T h = static_cast<T>(0);
+
+			if(rgbColor.r == Max)
+				// between yellow & magenta
+				h = static_cast<T>(0) + T(60) * (rgbColor.g - rgbColor.b) / Delta;
+			else if(rgbColor.g == Max)
+				// between cyan & yellow
+				h = static_cast<T>(120) + T(60) * (rgbColor.b - rgbColor.r) / Delta;
+			else
+				// between magenta & cyan
+				h = static_cast<T>(240) + T(60) * (rgbColor.r - rgbColor.g) / Delta;
+
+			if(h < T(0))
+				hsv.x = h + T(360);
+			else
+				hsv.x = h;
+		}
+		else
+		{
+			// If r = g = b = 0 then s = 0, h is undefined
+			hsv.y = static_cast<T>(0);
+			hsv.x = static_cast<T>(0);
+		}
+
+		return hsv;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> saturation(T const s)
+	{
+		vec<3, T, defaultp> rgbw = vec<3, T, defaultp>(T(0.2126), T(0.7152), T(0.0722));
+
+		vec<3, T, defaultp> const col((T(1) - s) * rgbw);
+
+		mat<4, 4, T, defaultp> result(T(1));
+		result[0][0] = col.x + s;
+		result[0][1] = col.x;
+		result[0][2] = col.x;
+		result[1][0] = col.y;
+		result[1][1] = col.y + s;
+		result[1][2] = col.y;
+		result[2][0] = col.z;
+		result[2][1] = col.z;
+		result[2][2] = col.z + s;
+
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> saturation(const T s, const vec<3, T, Q>& color)
+	{
+		return vec<3, T, Q>(saturation(s) * vec<4, T, Q>(color, T(0)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> saturation(const T s, const vec<4, T, Q>& color)
+	{
+		return saturation(s) * color;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T luminosity(const vec<3, T, Q>& color)
+	{
+		const vec<3, T, Q> tmp = vec<3, T, Q>(0.33, 0.59, 0.11);
+		return dot(color, tmp);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.hpp
new file mode 100644
index 000000000000..dd2b771693f6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.hpp
@@ -0,0 +1,60 @@
+/// @ref gtx_color_space_YCoCg
+/// @file glm/gtx/color_space_YCoCg.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_color_space_YCoCg GLM_GTX_color_space_YCoCg
+/// @ingroup gtx
+///
+/// Include <glm/gtx/color_space_YCoCg.hpp> to use the features of this extension.
+///
+/// RGB to YCoCg conversions and operations
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_color_space_YCoCg is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_color_space_YCoCg extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_color_space_YCoCg
+	/// @{
+
+	/// Convert a color from RGB color space to YCoCg color space.
+	/// @see gtx_color_space_YCoCg
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rgb2YCoCg(
+		vec<3, T, Q> const& rgbColor);
+
+	/// Convert a color from YCoCg color space to RGB color space.
+	/// @see gtx_color_space_YCoCg
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> YCoCg2rgb(
+		vec<3, T, Q> const& YCoCgColor);
+
+	/// Convert a color from RGB color space to YCoCgR color space.
+	/// @see "YCoCg-R: A Color Space with RGB Reversibility and Low Dynamic Range"
+	/// @see gtx_color_space_YCoCg
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rgb2YCoCgR(
+		vec<3, T, Q> const& rgbColor);
+
+	/// Convert a color from YCoCgR color space to RGB color space.
+	/// @see "YCoCg-R: A Color Space with RGB Reversibility and Low Dynamic Range"
+	/// @see gtx_color_space_YCoCg
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> YCoCgR2rgb(
+		vec<3, T, Q> const& YCoCgColor);
+
+	/// @}
+}//namespace glm
+
+#include "color_space_YCoCg.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.inl
new file mode 100644
index 000000000000..83ba857c08bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/color_space_YCoCg.inl
@@ -0,0 +1,107 @@
+/// @ref gtx_color_space_YCoCg
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rgb2YCoCg
+	(
+		vec<3, T, Q> const& rgbColor
+	)
+	{
+		vec<3, T, Q> result;
+		result.x/*Y */ =   rgbColor.r / T(4) + rgbColor.g / T(2) + rgbColor.b / T(4);
+		result.y/*Co*/ =   rgbColor.r / T(2) + rgbColor.g * T(0) - rgbColor.b / T(2);
+		result.z/*Cg*/ = - rgbColor.r / T(4) + rgbColor.g / T(2) - rgbColor.b / T(4);
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> YCoCg2rgb
+	(
+		vec<3, T, Q> const& YCoCgColor
+	)
+	{
+		vec<3, T, Q> result;
+		result.r = YCoCgColor.x + YCoCgColor.y - YCoCgColor.z;
+		result.g = YCoCgColor.x				   + YCoCgColor.z;
+		result.b = YCoCgColor.x - YCoCgColor.y - YCoCgColor.z;
+		return result;
+	}
+
+	template<typename T, qualifier Q, bool isInteger>
+	class compute_YCoCgR {
+	public:
+		static GLM_FUNC_QUALIFIER vec<3, T, Q> rgb2YCoCgR
+		(
+			vec<3, T, Q> const& rgbColor
+		)
+		{
+			vec<3, T, Q> result;
+			result.x/*Y */ = rgbColor.g * static_cast<T>(0.5) + (rgbColor.r + rgbColor.b) * static_cast<T>(0.25);
+			result.y/*Co*/ = rgbColor.r - rgbColor.b;
+			result.z/*Cg*/ = rgbColor.g - (rgbColor.r + rgbColor.b) * static_cast<T>(0.5);
+			return result;
+		}
+
+		static GLM_FUNC_QUALIFIER vec<3, T, Q> YCoCgR2rgb
+		(
+			vec<3, T, Q> const& YCoCgRColor
+		)
+		{
+			vec<3, T, Q> result;
+			T tmp = YCoCgRColor.x - (YCoCgRColor.z * static_cast<T>(0.5));
+			result.g = YCoCgRColor.z + tmp;
+			result.b = tmp - (YCoCgRColor.y * static_cast<T>(0.5));
+			result.r = result.b + YCoCgRColor.y;
+			return result;
+		}
+	};
+
+	template<typename T, qualifier Q>
+	class compute_YCoCgR<T, Q, true> {
+	public:
+		static GLM_FUNC_QUALIFIER vec<3, T, Q> rgb2YCoCgR
+		(
+			vec<3, T, Q> const& rgbColor
+		)
+		{
+			vec<3, T, Q> result;
+			result.y/*Co*/ = rgbColor.r - rgbColor.b;
+			T tmp = rgbColor.b + (result.y >> 1);
+			result.z/*Cg*/ = rgbColor.g - tmp;
+			result.x/*Y */ = tmp + (result.z >> 1);
+			return result;
+		}
+
+		static GLM_FUNC_QUALIFIER vec<3, T, Q> YCoCgR2rgb
+		(
+			vec<3, T, Q> const& YCoCgRColor
+		)
+		{
+			vec<3, T, Q> result;
+			T tmp = YCoCgRColor.x - (YCoCgRColor.z >> 1);
+			result.g = YCoCgRColor.z + tmp;
+			result.b = tmp - (YCoCgRColor.y >> 1);
+			result.r = result.b + YCoCgRColor.y;
+			return result;
+		}
+	};
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rgb2YCoCgR
+	(
+		vec<3, T, Q> const& rgbColor
+	)
+	{
+		return compute_YCoCgR<T, Q, std::numeric_limits<T>::is_integer>::rgb2YCoCgR(rgbColor);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> YCoCgR2rgb
+	(
+		vec<3, T, Q> const& YCoCgRColor
+	)
+	{
+		return compute_YCoCgR<T, Q, std::numeric_limits<T>::is_integer>::YCoCgR2rgb(YCoCgRColor);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/common.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/common.hpp
new file mode 100644
index 000000000000..254ada2d7695
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/common.hpp
@@ -0,0 +1,76 @@
+/// @ref gtx_common
+/// @file glm/gtx/common.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_common GLM_GTX_common
+/// @ingroup gtx
+///
+/// Include <glm/gtx/common.hpp> to use the features of this extension.
+///
+/// @brief Provide functions to increase the compatibility with Cg and HLSL languages
+
+#pragma once
+
+// Dependencies:
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../gtc/vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_common is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_common extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_common
+	/// @{
+
+	/// Returns true if x is a denormalized number
+	/// Numbers whose absolute value is too small to be represented in the normal format are represented in an alternate, denormalized format.
+	/// This format is less precise but can represent values closer to zero.
+	///
+	/// @tparam genType Floating-point scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/isnan.xml">GLSL isnan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::bool_type isdenormal(genType const& x);
+
+	/// Similar to 'mod' but with a different rounding and integer support.
+	/// Returns 'x - y * trunc(x/y)' instead of 'x - y * floor(x/y)'
+	///
+	/// @see <a href="http://stackoverflow.com/questions/7610631/glsl-mod-vs-hlsl-fmod">GLSL mod vs HLSL fmod</a>
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/mod.xml">GLSL mod man page</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fmod(vec<L, T, Q> const& v);
+
+	/// Returns whether vector components values are within an interval. A open interval excludes its endpoints, and is denoted with square brackets.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_relational
+	template <length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> openBounded(vec<L, T, Q> const& Value, vec<L, T, Q> const& Min, vec<L, T, Q> const& Max);
+
+	/// Returns whether vector components values are within an interval. A closed interval includes its endpoints, and is denoted with square brackets.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or integer scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see ext_vector_relational
+	template <length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> closeBounded(vec<L, T, Q> const& Value, vec<L, T, Q> const& Min, vec<L, T, Q> const& Max);
+
+	/// @}
+}//namespace glm
+
+#include "common.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/common.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/common.inl
new file mode 100644
index 000000000000..4575b20752f8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/common.inl
@@ -0,0 +1,125 @@
+/// @ref gtx_common
+
+#include <cmath>
+#include "../gtc/epsilon.hpp"
+#include "../gtc/constants.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q, bool isFloat = true>
+	struct compute_fmod
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+		{
+			return detail::functor2<vec, L, T, Q>::call(std::fmod, a, b);
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct compute_fmod<L, T, Q, false>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& a, vec<L, T, Q> const& b)
+		{
+			return a % b;
+		}
+	};
+}//namespace detail
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool isdenormal(T const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isdenormal' only accept floating-point inputs");
+
+#		if GLM_HAS_CXX11_STL
+			return std::fpclassify(x) == FP_SUBNORMAL;
+#		else
+			return epsilonNotEqual(x, static_cast<T>(0), epsilon<T>()) && std::fabs(x) < std::numeric_limits<T>::min();
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename vec<1, T, Q>::bool_type isdenormal
+	(
+		vec<1, T, Q> const& x
+	)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isdenormal' only accept floating-point inputs");
+
+		return typename vec<1, T, Q>::bool_type(
+			isdenormal(x.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename vec<2, T, Q>::bool_type isdenormal
+	(
+		vec<2, T, Q> const& x
+	)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isdenormal' only accept floating-point inputs");
+
+		return typename vec<2, T, Q>::bool_type(
+			isdenormal(x.x),
+			isdenormal(x.y));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename vec<3, T, Q>::bool_type isdenormal
+	(
+		vec<3, T, Q> const& x
+	)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isdenormal' only accept floating-point inputs");
+
+		return typename vec<3, T, Q>::bool_type(
+			isdenormal(x.x),
+			isdenormal(x.y),
+			isdenormal(x.z));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename vec<4, T, Q>::bool_type isdenormal
+	(
+		vec<4, T, Q> const& x
+	)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'isdenormal' only accept floating-point inputs");
+
+		return typename vec<4, T, Q>::bool_type(
+			isdenormal(x.x),
+			isdenormal(x.y),
+			isdenormal(x.z),
+			isdenormal(x.w));
+	}
+
+	// fmod
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fmod(genType x, genType y)
+	{
+		return fmod(vec<1, genType>(x), y).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmod(vec<L, T, Q> const& x, T y)
+	{
+		return detail::compute_fmod<L, T, Q, std::numeric_limits<T>::is_iec559>::call(x, vec<L, T, Q>(y));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fmod(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		return detail::compute_fmod<L, T, Q, std::numeric_limits<T>::is_iec559>::call(x, y);
+	}
+
+	template <length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> openBounded(vec<L, T, Q> const& Value, vec<L, T, Q> const& Min, vec<L, T, Q> const& Max)
+	{
+		return greaterThan(Value, Min) && lessThan(Value, Max);
+	}
+
+	template <length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> closeBounded(vec<L, T, Q> const& Value, vec<L, T, Q> const& Min, vec<L, T, Q> const& Max)
+	{
+		return greaterThanEqual(Value, Min) && lessThanEqual(Value, Max);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.hpp
new file mode 100644
index 000000000000..0af75583b96a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.hpp
@@ -0,0 +1,133 @@
+/// @ref gtx_compatibility
+/// @file glm/gtx/compatibility.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_compatibility GLM_GTX_compatibility
+/// @ingroup gtx
+///
+/// Include <glm/gtx/compatibility.hpp> to use the features of this extension.
+///
+/// Provide functions to increase the compatibility with Cg and HLSL languages
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/quaternion.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_compatibility is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_compatibility extension included")
+#	endif
+#endif
+
+#if GLM_COMPILER & GLM_COMPILER_VC
+#	include <cfloat>
+#elif GLM_COMPILER & GLM_COMPILER_GCC
+#	include <cmath>
+#	if(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+#		undef isfinite
+#	endif
+#endif//GLM_COMPILER
+
+namespace glm
+{
+	/// @addtogroup gtx_compatibility
+	/// @{
+
+	template<typename T> GLM_FUNC_QUALIFIER T lerp(T x, T y, T a){return mix(x, y, a);}																					//!< \brief Returns x * (1.0 - a) + y * a, i.e., the linear blend of x and y using the floating-point value a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<2, T, Q> lerp(const vec<2, T, Q>& x, const vec<2, T, Q>& y, T a){return mix(x, y, a);}							//!< \brief Returns x * (1.0 - a) + y * a, i.e., the linear blend of x and y using the floating-point value a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<3, T, Q> lerp(const vec<3, T, Q>& x, const vec<3, T, Q>& y, T a){return mix(x, y, a);}							//!< \brief Returns x * (1.0 - a) + y * a, i.e., the linear blend of x and y using the floating-point value a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<4, T, Q> lerp(const vec<4, T, Q>& x, const vec<4, T, Q>& y, T a){return mix(x, y, a);}							//!< \brief Returns x * (1.0 - a) + y * a, i.e., the linear blend of x and y using the floating-point value a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<2, T, Q> lerp(const vec<2, T, Q>& x, const vec<2, T, Q>& y, const vec<2, T, Q>& a){return mix(x, y, a);}	//!< \brief Returns the component-wise result of x * (1.0 - a) + y * a, i.e., the linear blend of x and y using vector a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<3, T, Q> lerp(const vec<3, T, Q>& x, const vec<3, T, Q>& y, const vec<3, T, Q>& a){return mix(x, y, a);}	//!< \brief Returns the component-wise result of x * (1.0 - a) + y * a, i.e., the linear blend of x and y using vector a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<4, T, Q> lerp(const vec<4, T, Q>& x, const vec<4, T, Q>& y, const vec<4, T, Q>& a){return mix(x, y, a);}	//!< \brief Returns the component-wise result of x * (1.0 - a) + y * a, i.e., the linear blend of x and y using vector a. The value for a is not restricted to the range [0, 1]. (From GLM_GTX_compatibility)
+
+	template<typename T> GLM_FUNC_QUALIFIER T saturate(T x){return clamp(x, T(0), T(1));}														//!< \brief Returns clamp(x, 0, 1) for each component in x. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<2, T, Q> saturate(const vec<2, T, Q>& x){return clamp(x, T(0), T(1));}					//!< \brief Returns clamp(x, 0, 1) for each component in x. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<3, T, Q> saturate(const vec<3, T, Q>& x){return clamp(x, T(0), T(1));}					//!< \brief Returns clamp(x, 0, 1) for each component in x. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<4, T, Q> saturate(const vec<4, T, Q>& x){return clamp(x, T(0), T(1));}					//!< \brief Returns clamp(x, 0, 1) for each component in x. (From GLM_GTX_compatibility)
+
+	template<typename T> GLM_FUNC_QUALIFIER T atan2(T y, T x){return atan(y, x);}																//!< \brief Arc tangent. Returns an angle whose tangent is y/x. The signs of x and y are used to determine what quadrant the angle is in. The range of values returned by this function is [-PI, PI]. Results are undefined if x and y are both 0. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<2, T, Q> atan2(const vec<2, T, Q>& y, const vec<2, T, Q>& x){return atan(y, x);}	//!< \brief Arc tangent. Returns an angle whose tangent is y/x. The signs of x and y are used to determine what quadrant the angle is in. The range of values returned by this function is [-PI, PI]. Results are undefined if x and y are both 0. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<3, T, Q> atan2(const vec<3, T, Q>& y, const vec<3, T, Q>& x){return atan(y, x);}	//!< \brief Arc tangent. Returns an angle whose tangent is y/x. The signs of x and y are used to determine what quadrant the angle is in. The range of values returned by this function is [-PI, PI]. Results are undefined if x and y are both 0. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_QUALIFIER vec<4, T, Q> atan2(const vec<4, T, Q>& y, const vec<4, T, Q>& x){return atan(y, x);}	//!< \brief Arc tangent. Returns an angle whose tangent is y/x. The signs of x and y are used to determine what quadrant the angle is in. The range of values returned by this function is [-PI, PI]. Results are undefined if x and y are both 0. (From GLM_GTX_compatibility)
+
+	template<typename genType> GLM_FUNC_DECL bool isfinite(genType const& x);											//!< \brief Test whether or not a scalar or each vector component is a finite value. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_DECL vec<1, bool, Q> isfinite(const vec<1, T, Q>& x);				//!< \brief Test whether or not a scalar or each vector component is a finite value. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_DECL vec<2, bool, Q> isfinite(const vec<2, T, Q>& x);				//!< \brief Test whether or not a scalar or each vector component is a finite value. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_DECL vec<3, bool, Q> isfinite(const vec<3, T, Q>& x);				//!< \brief Test whether or not a scalar or each vector component is a finite value. (From GLM_GTX_compatibility)
+	template<typename T, qualifier Q> GLM_FUNC_DECL vec<4, bool, Q> isfinite(const vec<4, T, Q>& x);				//!< \brief Test whether or not a scalar or each vector component is a finite value. (From GLM_GTX_compatibility)
+
+	typedef bool						bool1;			//!< \brief boolean type with 1 component. (From GLM_GTX_compatibility extension)
+	typedef vec<2, bool, highp>			bool2;			//!< \brief boolean type with 2 components. (From GLM_GTX_compatibility extension)
+	typedef vec<3, bool, highp>			bool3;			//!< \brief boolean type with 3 components. (From GLM_GTX_compatibility extension)
+	typedef vec<4, bool, highp>			bool4;			//!< \brief boolean type with 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef bool						bool1x1;		//!< \brief boolean matrix with 1 x 1 component. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 2, bool, highp>		bool2x2;		//!< \brief boolean matrix with 2 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 3, bool, highp>		bool2x3;		//!< \brief boolean matrix with 2 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 4, bool, highp>		bool2x4;		//!< \brief boolean matrix with 2 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 2, bool, highp>		bool3x2;		//!< \brief boolean matrix with 3 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 3, bool, highp>		bool3x3;		//!< \brief boolean matrix with 3 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 4, bool, highp>		bool3x4;		//!< \brief boolean matrix with 3 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 2, bool, highp>		bool4x2;		//!< \brief boolean matrix with 4 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 3, bool, highp>		bool4x3;		//!< \brief boolean matrix with 4 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 4, bool, highp>		bool4x4;		//!< \brief boolean matrix with 4 x 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef int							int1;			//!< \brief integer vector with 1 component. (From GLM_GTX_compatibility extension)
+	typedef vec<2, int, highp>			int2;			//!< \brief integer vector with 2 components. (From GLM_GTX_compatibility extension)
+	typedef vec<3, int, highp>			int3;			//!< \brief integer vector with 3 components. (From GLM_GTX_compatibility extension)
+	typedef vec<4, int, highp>			int4;			//!< \brief integer vector with 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef int							int1x1;			//!< \brief integer matrix with 1 component. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 2, int, highp>		int2x2;			//!< \brief integer matrix with 2 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 3, int, highp>		int2x3;			//!< \brief integer matrix with 2 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 4, int, highp>		int2x4;			//!< \brief integer matrix with 2 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 2, int, highp>		int3x2;			//!< \brief integer matrix with 3 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 3, int, highp>		int3x3;			//!< \brief integer matrix with 3 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 4, int, highp>		int3x4;			//!< \brief integer matrix with 3 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 2, int, highp>		int4x2;			//!< \brief integer matrix with 4 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 3, int, highp>		int4x3;			//!< \brief integer matrix with 4 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 4, int, highp>		int4x4;			//!< \brief integer matrix with 4 x 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef float						float1;			//!< \brief single-qualifier floating-point vector with 1 component. (From GLM_GTX_compatibility extension)
+	typedef vec<2, float, highp>		float2;			//!< \brief single-qualifier floating-point vector with 2 components. (From GLM_GTX_compatibility extension)
+	typedef vec<3, float, highp>		float3;			//!< \brief single-qualifier floating-point vector with 3 components. (From GLM_GTX_compatibility extension)
+	typedef vec<4, float, highp>		float4;			//!< \brief single-qualifier floating-point vector with 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef float						float1x1;		//!< \brief single-qualifier floating-point matrix with 1 component. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 2, float, highp>		float2x2;		//!< \brief single-qualifier floating-point matrix with 2 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 3, float, highp>		float2x3;		//!< \brief single-qualifier floating-point matrix with 2 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 4, float, highp>		float2x4;		//!< \brief single-qualifier floating-point matrix with 2 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 2, float, highp>		float3x2;		//!< \brief single-qualifier floating-point matrix with 3 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 3, float, highp>		float3x3;		//!< \brief single-qualifier floating-point matrix with 3 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 4, float, highp>		float3x4;		//!< \brief single-qualifier floating-point matrix with 3 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 2, float, highp>		float4x2;		//!< \brief single-qualifier floating-point matrix with 4 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 3, float, highp>		float4x3;		//!< \brief single-qualifier floating-point matrix with 4 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 4, float, highp>		float4x4;		//!< \brief single-qualifier floating-point matrix with 4 x 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef double						double1;		//!< \brief double-qualifier floating-point vector with 1 component. (From GLM_GTX_compatibility extension)
+	typedef vec<2, double, highp>		double2;		//!< \brief double-qualifier floating-point vector with 2 components. (From GLM_GTX_compatibility extension)
+	typedef vec<3, double, highp>		double3;		//!< \brief double-qualifier floating-point vector with 3 components. (From GLM_GTX_compatibility extension)
+	typedef vec<4, double, highp>		double4;		//!< \brief double-qualifier floating-point vector with 4 components. (From GLM_GTX_compatibility extension)
+
+	typedef double						double1x1;		//!< \brief double-qualifier floating-point matrix with 1 component. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 2, double, highp>		double2x2;		//!< \brief double-qualifier floating-point matrix with 2 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 3, double, highp>		double2x3;		//!< \brief double-qualifier floating-point matrix with 2 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<2, 4, double, highp>		double2x4;		//!< \brief double-qualifier floating-point matrix with 2 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 2, double, highp>		double3x2;		//!< \brief double-qualifier floating-point matrix with 3 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 3, double, highp>		double3x3;		//!< \brief double-qualifier floating-point matrix with 3 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<3, 4, double, highp>		double3x4;		//!< \brief double-qualifier floating-point matrix with 3 x 4 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 2, double, highp>		double4x2;		//!< \brief double-qualifier floating-point matrix with 4 x 2 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 3, double, highp>		double4x3;		//!< \brief double-qualifier floating-point matrix with 4 x 3 components. (From GLM_GTX_compatibility extension)
+	typedef mat<4, 4, double, highp>		double4x4;		//!< \brief double-qualifier floating-point matrix with 4 x 4 components. (From GLM_GTX_compatibility extension)
+
+	/// @}
+}//namespace glm
+
+#include "compatibility.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.inl
new file mode 100644
index 000000000000..1d49496b6c6e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/compatibility.inl
@@ -0,0 +1,62 @@
+#include <limits>
+
+namespace glm
+{
+	// isfinite
+	template<typename genType>
+	GLM_FUNC_QUALIFIER bool isfinite(
+		genType const& x)
+	{
+#		if GLM_HAS_CXX11_STL
+			return std::isfinite(x) != 0;
+#		elif GLM_COMPILER & GLM_COMPILER_VC
+			return _finite(x) != 0;
+#		elif GLM_COMPILER & GLM_COMPILER_GCC && GLM_PLATFORM & GLM_PLATFORM_ANDROID
+			return _isfinite(x) != 0;
+#		else
+			if (std::numeric_limits<genType>::is_integer || std::denorm_absent == std::numeric_limits<genType>::has_denorm)
+				return std::numeric_limits<genType>::min() <= x && std::numeric_limits<genType>::max() >= x;
+			else
+				return -std::numeric_limits<genType>::max() <= x && std::numeric_limits<genType>::max() >= x;
+#		endif
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<1, bool, Q> isfinite(
+		vec<1, T, Q> const& x)
+	{
+		return vec<1, bool, Q>(
+			isfinite(x.x));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, bool, Q> isfinite(
+		vec<2, T, Q> const& x)
+	{
+		return vec<2, bool, Q>(
+			isfinite(x.x),
+			isfinite(x.y));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, bool, Q> isfinite(
+		vec<3, T, Q> const& x)
+	{
+		return vec<3, bool, Q>(
+			isfinite(x.x),
+			isfinite(x.y),
+			isfinite(x.z));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> isfinite(
+		vec<4, T, Q> const& x)
+	{
+		return vec<4, bool, Q>(
+			isfinite(x.x),
+			isfinite(x.y),
+			isfinite(x.z),
+			isfinite(x.w));
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.hpp
new file mode 100644
index 000000000000..34a2b0a37517
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.hpp
@@ -0,0 +1,69 @@
+/// @ref gtx_component_wise
+/// @file glm/gtx/component_wise.hpp
+/// @date 2007-05-21 / 2011-06-07
+/// @author Christophe Riccio
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_component_wise GLM_GTX_component_wise
+/// @ingroup gtx
+///
+/// Include <glm/gtx/component_wise.hpp> to use the features of this extension.
+///
+/// Operations between components of a type
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_component_wise is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_component_wise extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_component_wise
+	/// @{
+
+	/// Convert an integer vector to a normalized float vector.
+	/// If the parameter value type is already a floating qualifier type, the value is passed through.
+	/// @see gtx_component_wise
+	template<typename floatType, length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, floatType, Q> compNormalize(vec<L, T, Q> const& v);
+
+	/// Convert a normalized float vector to an integer vector.
+	/// If the parameter value type is already a floating qualifier type, the value is passed through.
+	/// @see gtx_component_wise
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> compScale(vec<L, floatType, Q> const& v);
+
+	/// Add all vector components together.
+	/// @see gtx_component_wise
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::value_type compAdd(genType const& v);
+
+	/// Multiply all vector components together.
+	/// @see gtx_component_wise
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::value_type compMul(genType const& v);
+
+	/// Find the minimum value between single vector components.
+	/// @see gtx_component_wise
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::value_type compMin(genType const& v);
+
+	/// Find the maximum value between single vector components.
+	/// @see gtx_component_wise
+	template<typename genType>
+	GLM_FUNC_DECL typename genType::value_type compMax(genType const& v);
+
+	/// @}
+}//namespace glm
+
+#include "component_wise.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.inl
new file mode 100644
index 000000000000..cbbc7d41ec00
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/component_wise.inl
@@ -0,0 +1,127 @@
+/// @ref gtx_component_wise
+
+#include <limits>
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, typename floatType, qualifier Q, bool isInteger, bool signedType>
+	struct compute_compNormalize
+	{};
+
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	struct compute_compNormalize<L, T, floatType, Q, true, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, floatType, Q> call(vec<L, T, Q> const& v)
+		{
+			floatType const Min = static_cast<floatType>(std::numeric_limits<T>::min());
+			floatType const Max = static_cast<floatType>(std::numeric_limits<T>::max());
+			return (vec<L, floatType, Q>(v) - Min) / (Max - Min) * static_cast<floatType>(2) - static_cast<floatType>(1);
+		}
+	};
+
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	struct compute_compNormalize<L, T, floatType, Q, true, false>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, floatType, Q> call(vec<L, T, Q> const& v)
+		{
+			return vec<L, floatType, Q>(v) / static_cast<floatType>(std::numeric_limits<T>::max());
+		}
+	};
+
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	struct compute_compNormalize<L, T, floatType, Q, false, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, floatType, Q> call(vec<L, T, Q> const& v)
+		{
+			return v;
+		}
+	};
+
+	template<length_t L, typename T, typename floatType, qualifier Q, bool isInteger, bool signedType>
+	struct compute_compScale
+	{};
+
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	struct compute_compScale<L, T, floatType, Q, true, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, floatType, Q> const& v)
+		{
+			floatType const Max = static_cast<floatType>(std::numeric_limits<T>::max()) + static_cast<floatType>(0.5);
+			vec<L, floatType, Q> const Scaled(v * Max);
+			vec<L, T, Q> const Result(Scaled - static_cast<floatType>(0.5));
+			return Result;
+		}
+	};
+
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	struct compute_compScale<L, T, floatType, Q, true, false>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, floatType, Q> const& v)
+		{
+			return vec<L, T, Q>(vec<L, floatType, Q>(v) * static_cast<floatType>(std::numeric_limits<T>::max()));
+		}
+	};
+
+	template<length_t L, typename T, typename floatType, qualifier Q>
+	struct compute_compScale<L, T, floatType, Q, false, true>
+	{
+		GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, floatType, Q> const& v)
+		{
+			return v;
+		}
+	};
+}//namespace detail
+
+	template<typename floatType, length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, floatType, Q> compNormalize(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<floatType>::is_iec559, "'compNormalize' accepts only floating-point types for 'floatType' template parameter");
+
+		return detail::compute_compNormalize<L, T, floatType, Q, std::numeric_limits<T>::is_integer, std::numeric_limits<T>::is_signed>::call(v);
+	}
+
+	template<typename T, length_t L, typename floatType, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> compScale(vec<L, floatType, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<floatType>::is_iec559, "'compScale' accepts only floating-point types for 'floatType' template parameter");
+
+		return detail::compute_compScale<L, T, floatType, Q, std::numeric_limits<T>::is_integer, std::numeric_limits<T>::is_signed>::call(v);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T compAdd(vec<L, T, Q> const& v)
+	{
+		T Result(0);
+		for(length_t i = 0, n = v.length(); i < n; ++i)
+			Result += v[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T compMul(vec<L, T, Q> const& v)
+	{
+		T Result(1);
+		for(length_t i = 0, n = v.length(); i < n; ++i)
+			Result *= v[i];
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T compMin(vec<L, T, Q> const& v)
+	{
+		T Result(v[0]);
+		for(length_t i = 1, n = v.length(); i < n; ++i)
+			Result = min(Result, v[i]);
+		return Result;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T compMax(vec<L, T, Q> const& v)
+	{
+		T Result(v[0]);
+		for(length_t i = 1, n = v.length(); i < n; ++i)
+			Result = max(Result, v[i]);
+		return Result;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.hpp
new file mode 100644
index 000000000000..a6f57613db1b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.hpp
@@ -0,0 +1,274 @@
+/// @ref gtx_dual_quaternion
+/// @file glm/gtx/dual_quaternion.hpp
+/// @author Maksim Vorobiev (msomeone@gmail.com)
+///
+/// @see core (dependence)
+/// @see gtc_constants (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtx_dual_quaternion GLM_GTX_dual_quaternion
+/// @ingroup gtx
+///
+/// Include <glm/gtx/dual_quaternion.hpp> to use the features of this extension.
+///
+/// Defines a templated dual-quaternion type and several dual-quaternion operations.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/constants.hpp"
+#include "../gtc/quaternion.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_dual_quaternion is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_dual_quaternion extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_dual_quaternion
+	/// @{
+
+	template<typename T, qualifier Q = defaultp>
+	struct tdualquat
+	{
+		// -- Implementation detail --
+
+		typedef T value_type;
+		typedef qua<T, Q> part_type;
+
+		// -- Data --
+
+		qua<T, Q> real, dual;
+
+		// -- Component accesses --
+
+		typedef length_t length_type;
+		/// Return the count of components of a dual quaternion
+		GLM_FUNC_DECL static GLM_CONSTEXPR length_type length(){return 2;}
+
+		GLM_FUNC_DECL part_type & operator[](length_type i);
+		GLM_FUNC_DECL part_type const& operator[](length_type i) const;
+
+		// -- Implicit basic constructors --
+
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR tdualquat() GLM_DEFAULT;
+		GLM_DEFAULTED_FUNC_DECL GLM_CONSTEXPR tdualquat(tdualquat<T, Q> const& d) GLM_DEFAULT;
+		template<qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR tdualquat(tdualquat<T, P> const& d);
+
+		// -- Explicit basic constructors --
+
+		GLM_FUNC_DECL GLM_CONSTEXPR tdualquat(qua<T, Q> const& real);
+		GLM_FUNC_DECL GLM_CONSTEXPR tdualquat(qua<T, Q> const& orientation, vec<3, T, Q> const& translation);
+		GLM_FUNC_DECL GLM_CONSTEXPR tdualquat(qua<T, Q> const& real, qua<T, Q> const& dual);
+
+		// -- Conversion constructors --
+
+		template<typename U, qualifier P>
+		GLM_FUNC_DECL GLM_CONSTEXPR GLM_EXPLICIT tdualquat(tdualquat<U, P> const& q);
+
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR tdualquat(mat<2, 4, T, Q> const& holder_mat);
+		GLM_FUNC_DECL GLM_EXPLICIT GLM_CONSTEXPR tdualquat(mat<3, 4, T, Q> const& aug_mat);
+
+		// -- Unary arithmetic operators --
+
+		GLM_DEFAULTED_FUNC_DECL tdualquat<T, Q> & operator=(tdualquat<T, Q> const& m) GLM_DEFAULT;
+
+		template<typename U>
+		GLM_FUNC_DECL tdualquat<T, Q> & operator=(tdualquat<U, Q> const& m);
+		template<typename U>
+		GLM_FUNC_DECL tdualquat<T, Q> & operator*=(U s);
+		template<typename U>
+		GLM_FUNC_DECL tdualquat<T, Q> & operator/=(U s);
+	};
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator+(tdualquat<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator-(tdualquat<T, Q> const& q);
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator+(tdualquat<T, Q> const& q, tdualquat<T, Q> const& p);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator*(tdualquat<T, Q> const& q, tdualquat<T, Q> const& p);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> operator*(tdualquat<T, Q> const& q, vec<3, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> operator*(vec<3, T, Q> const& v, tdualquat<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> operator*(tdualquat<T, Q> const& q, vec<4, T, Q> const& v);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> operator*(vec<4, T, Q> const& v, tdualquat<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator*(tdualquat<T, Q> const& q, T const& s);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator*(T const& s, tdualquat<T, Q> const& q);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> operator/(tdualquat<T, Q> const& q, T const& s);
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool operator==(tdualquat<T, Q> const& q1, tdualquat<T, Q> const& q2);
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool operator!=(tdualquat<T, Q> const& q1, tdualquat<T, Q> const& q2);
+
+	/// Creates an identity dual quaternion.
+	///
+	/// @see gtx_dual_quaternion
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> dual_quat_identity();
+
+	/// Returns the normalized quaternion.
+	///
+	/// @see gtx_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> normalize(tdualquat<T, Q> const& q);
+
+	/// Returns the linear interpolation of two dual quaternion.
+	///
+	/// @see gtc_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> lerp(tdualquat<T, Q> const& x, tdualquat<T, Q> const& y, T const& a);
+
+	/// Returns the q inverse.
+	///
+	/// @see gtx_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> inverse(tdualquat<T, Q> const& q);
+
+	/// Converts a quaternion to a 2 * 4 matrix.
+	///
+	/// @see gtx_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 4, T, Q> mat2x4_cast(tdualquat<T, Q> const& x);
+
+	/// Converts a quaternion to a 3 * 4 matrix.
+	///
+	/// @see gtx_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 4, T, Q> mat3x4_cast(tdualquat<T, Q> const& x);
+
+	/// Converts a 2 * 4 matrix (matrix which holds real and dual parts) to a quaternion.
+	///
+	/// @see gtx_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> dualquat_cast(mat<2, 4, T, Q> const& x);
+
+	/// Converts a 3 * 4 matrix (augmented matrix rotation + translation) to a quaternion.
+	///
+	/// @see gtx_dual_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL tdualquat<T, Q> dualquat_cast(mat<3, 4, T, Q> const& x);
+
+
+	/// Dual-quaternion of low single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<float, lowp>		lowp_dualquat;
+
+	/// Dual-quaternion of medium single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<float, mediump>	mediump_dualquat;
+
+	/// Dual-quaternion of high single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<float, highp>		highp_dualquat;
+
+
+	/// Dual-quaternion of low single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<float, lowp>		lowp_fdualquat;
+
+	/// Dual-quaternion of medium single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<float, mediump>	mediump_fdualquat;
+
+	/// Dual-quaternion of high single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<float, highp>		highp_fdualquat;
+
+
+	/// Dual-quaternion of low double-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<double, lowp>		lowp_ddualquat;
+
+	/// Dual-quaternion of medium double-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<double, mediump>	mediump_ddualquat;
+
+	/// Dual-quaternion of high double-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef tdualquat<double, highp>	highp_ddualquat;
+
+
+#if(!defined(GLM_PRECISION_HIGHP_FLOAT) && !defined(GLM_PRECISION_MEDIUMP_FLOAT) && !defined(GLM_PRECISION_LOWP_FLOAT))
+	/// Dual-quaternion of floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef highp_fdualquat			dualquat;
+
+	/// Dual-quaternion of single-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef highp_fdualquat			fdualquat;
+#elif(defined(GLM_PRECISION_HIGHP_FLOAT) && !defined(GLM_PRECISION_MEDIUMP_FLOAT) && !defined(GLM_PRECISION_LOWP_FLOAT))
+	typedef highp_fdualquat			dualquat;
+	typedef highp_fdualquat			fdualquat;
+#elif(!defined(GLM_PRECISION_HIGHP_FLOAT) && defined(GLM_PRECISION_MEDIUMP_FLOAT) && !defined(GLM_PRECISION_LOWP_FLOAT))
+	typedef mediump_fdualquat		dualquat;
+	typedef mediump_fdualquat		fdualquat;
+#elif(!defined(GLM_PRECISION_HIGHP_FLOAT) && !defined(GLM_PRECISION_MEDIUMP_FLOAT) && defined(GLM_PRECISION_LOWP_FLOAT))
+	typedef lowp_fdualquat			dualquat;
+	typedef lowp_fdualquat			fdualquat;
+#else
+#	error "GLM error: multiple default precision requested for single-precision floating-point types"
+#endif
+
+
+#if(!defined(GLM_PRECISION_HIGHP_DOUBLE) && !defined(GLM_PRECISION_MEDIUMP_DOUBLE) && !defined(GLM_PRECISION_LOWP_DOUBLE))
+	/// Dual-quaternion of default double-qualifier floating-point numbers.
+	///
+	/// @see gtx_dual_quaternion
+	typedef highp_ddualquat			ddualquat;
+#elif(defined(GLM_PRECISION_HIGHP_DOUBLE) && !defined(GLM_PRECISION_MEDIUMP_DOUBLE) && !defined(GLM_PRECISION_LOWP_DOUBLE))
+	typedef highp_ddualquat			ddualquat;
+#elif(!defined(GLM_PRECISION_HIGHP_DOUBLE) && defined(GLM_PRECISION_MEDIUMP_DOUBLE) && !defined(GLM_PRECISION_LOWP_DOUBLE))
+	typedef mediump_ddualquat		ddualquat;
+#elif(!defined(GLM_PRECISION_HIGHP_DOUBLE) && !defined(GLM_PRECISION_MEDIUMP_DOUBLE) && defined(GLM_PRECISION_LOWP_DOUBLE))
+	typedef lowp_ddualquat			ddualquat;
+#else
+#	error "GLM error: Multiple default precision requested for double-precision floating-point types"
+#endif
+
+	/// @}
+} //namespace glm
+
+#include "dual_quaternion.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.inl
new file mode 100644
index 000000000000..3a04160e3a16
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/dual_quaternion.inl
@@ -0,0 +1,352 @@
+/// @ref gtx_dual_quaternion
+
+#include "../geometric.hpp"
+#include <limits>
+
+namespace glm
+{
+	// -- Component accesses --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename tdualquat<T, Q>::part_type & tdualquat<T, Q>::operator[](typename tdualquat<T, Q>::length_type i)
+	{
+		assert(i >= 0 && i < this->length());
+		return (&real)[i];
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER typename tdualquat<T, Q>::part_type const& tdualquat<T, Q>::operator[](typename tdualquat<T, Q>::length_type i) const
+	{
+		assert(i >= 0 && i < this->length());
+		return (&real)[i];
+	}
+
+	// -- Implicit basic constructors --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat()
+#			if GLM_CONFIG_DEFAULTED_FUNCTIONS != GLM_DISABLE
+			: real(qua<T, Q>())
+			, dual(qua<T, Q>::wxyz(0, 0, 0, 0))
+#			endif
+		{}
+
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(tdualquat<T, Q> const& d)
+			: real(d.real)
+			, dual(d.dual)
+		{}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(tdualquat<T, P> const& d)
+		: real(d.real)
+		, dual(d.dual)
+	{}
+
+	// -- Explicit basic constructors --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(qua<T, Q> const& r)
+		: real(r), dual(qua<T, Q>::wxyz(0, 0, 0, 0))
+	{}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(qua<T, Q> const& q, vec<3, T, Q> const& p)
+		: real(q), dual(qua<T, Q>::wxyz(
+			T(-0.5) * ( p.x*q.x + p.y*q.y + p.z*q.z),
+			T(+0.5) * ( p.x*q.w + p.y*q.z - p.z*q.y),
+			T(+0.5) * (-p.x*q.z + p.y*q.w + p.z*q.x),
+			T(+0.5) * ( p.x*q.y - p.y*q.x + p.z*q.w)))
+	{}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(qua<T, Q> const& r, qua<T, Q> const& d)
+		: real(r), dual(d)
+	{}
+
+	// -- Conversion constructors --
+
+	template<typename T, qualifier Q>
+	template<typename U, qualifier P>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(tdualquat<U, P> const& q)
+		: real(q.real)
+		, dual(q.dual)
+	{}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(mat<2, 4, T, Q> const& m)
+	{
+		*this = dualquat_cast(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR tdualquat<T, Q>::tdualquat(mat<3, 4, T, Q> const& m)
+	{
+		*this = dualquat_cast(m);
+	}
+
+	// -- Unary arithmetic operators --
+
+#	if GLM_CONFIG_DEFAULTED_FUNCTIONS == GLM_DISABLE
+		template<typename T, qualifier Q>
+		GLM_DEFAULTED_FUNC_QUALIFIER tdualquat<T, Q> & tdualquat<T, Q>::operator=(tdualquat<T, Q> const& q)
+		{
+			this->real = q.real;
+			this->dual = q.dual;
+			return *this;
+		}
+#	endif
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> & tdualquat<T, Q>::operator=(tdualquat<U, Q> const& q)
+	{
+		this->real = q.real;
+		this->dual = q.dual;
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> & tdualquat<T, Q>::operator*=(U s)
+	{
+		this->real *= static_cast<T>(s);
+		this->dual *= static_cast<T>(s);
+		return *this;
+	}
+
+	template<typename T, qualifier Q>
+	template<typename U>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> & tdualquat<T, Q>::operator/=(U s)
+	{
+		this->real /= static_cast<T>(s);
+		this->dual /= static_cast<T>(s);
+		return *this;
+	}
+
+	// -- Unary bit operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator+(tdualquat<T, Q> const& q)
+	{
+		return q;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator-(tdualquat<T, Q> const& q)
+	{
+		return tdualquat<T, Q>(-q.real, -q.dual);
+	}
+
+	// -- Binary operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator+(tdualquat<T, Q> const& q, tdualquat<T, Q> const& p)
+	{
+		return tdualquat<T, Q>(q.real + p.real,q.dual + p.dual);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator*(tdualquat<T, Q> const& p, tdualquat<T, Q> const& o)
+	{
+		return tdualquat<T, Q>(p.real * o.real,p.real * o.dual + p.dual * o.real);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> operator*(tdualquat<T, Q> const& q, vec<3, T, Q> const& v)
+	{
+		vec<3, T, Q> const real_v3(q.real.x,q.real.y,q.real.z);
+		vec<3, T, Q> const dual_v3(q.dual.x,q.dual.y,q.dual.z);
+		return (cross(real_v3, cross(real_v3,v) + v * q.real.w + dual_v3) + dual_v3 * q.real.w - real_v3 * q.dual.w) * T(2) + v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> operator*(vec<3, T, Q> const& v,	tdualquat<T, Q> const& q)
+	{
+		return glm::inverse(q) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> operator*(tdualquat<T, Q> const& q, vec<4, T, Q> const& v)
+	{
+		return vec<4, T, Q>(q * vec<3, T, Q>(v), v.w);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> operator*(vec<4, T, Q> const& v,	tdualquat<T, Q> const& q)
+	{
+		return glm::inverse(q) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator*(tdualquat<T, Q> const& q, T const& s)
+	{
+		return tdualquat<T, Q>(q.real * s, q.dual * s);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator*(T const& s, tdualquat<T, Q> const& q)
+	{
+		return q * s;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> operator/(tdualquat<T, Q> const& q,	T const& s)
+	{
+		return tdualquat<T, Q>(q.real / s, q.dual / s);
+	}
+
+	// -- Boolean operators --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool operator==(tdualquat<T, Q> const& q1, tdualquat<T, Q> const& q2)
+	{
+		return (q1.real == q2.real) && (q1.dual == q2.dual);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool operator!=(tdualquat<T, Q> const& q1, tdualquat<T, Q> const& q2)
+	{
+		return (q1.real != q2.real) || (q1.dual != q2.dual);
+	}
+
+	// -- Operations --
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> dual_quat_identity()
+	{
+		return tdualquat<T, Q>(
+			qua<T, Q>::wxyz(static_cast<T>(1), static_cast<T>(0), static_cast<T>(0), static_cast<T>(0)),
+			qua<T, Q>::wxyz(static_cast<T>(0), static_cast<T>(0), static_cast<T>(0), static_cast<T>(0)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> normalize(tdualquat<T, Q> const& q)
+	{
+		return q / length(q.real);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> lerp(tdualquat<T, Q> const& x, tdualquat<T, Q> const& y, T const& a)
+	{
+		// Dual Quaternion Linear blend aka DLB:
+		// Lerp is only defined in [0, 1]
+		assert(a >= static_cast<T>(0));
+		assert(a <= static_cast<T>(1));
+		T const k = dot(x.real,y.real) < static_cast<T>(0) ? -a : a;
+		T const one(1);
+		return tdualquat<T, Q>(x * (one - a) + y * k);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> inverse(tdualquat<T, Q> const& q)
+	{
+		const glm::qua<T, Q> real = conjugate(q.real);
+		const glm::qua<T, Q> dual = conjugate(q.dual);
+		return tdualquat<T, Q>(real, dual + (real * (-2.0f * dot(real,dual))));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 4, T, Q> mat2x4_cast(tdualquat<T, Q> const& x)
+	{
+		return mat<2, 4, T, Q>( x[0].x, x[0].y, x[0].z, x[0].w, x[1].x, x[1].y, x[1].z, x[1].w );
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 4, T, Q> mat3x4_cast(tdualquat<T, Q> const& x)
+	{
+		qua<T, Q> r = x.real / length2(x.real);
+
+		qua<T, Q> const rr(r.w * x.real.w, r.x * x.real.x, r.y * x.real.y, r.z * x.real.z);
+		r *= static_cast<T>(2);
+
+		T const xy = r.x * x.real.y;
+		T const xz = r.x * x.real.z;
+		T const yz = r.y * x.real.z;
+		T const wx = r.w * x.real.x;
+		T const wy = r.w * x.real.y;
+		T const wz = r.w * x.real.z;
+
+		vec<4, T, Q> const a(
+			rr.w + rr.x - rr.y - rr.z,
+			xy - wz,
+			xz + wy,
+			-(x.dual.w * r.x - x.dual.x * r.w + x.dual.y * r.z - x.dual.z * r.y));
+
+		vec<4, T, Q> const b(
+			xy + wz,
+			rr.w + rr.y - rr.x - rr.z,
+			yz - wx,
+			-(x.dual.w * r.y - x.dual.x * r.z - x.dual.y * r.w + x.dual.z * r.x));
+
+		vec<4, T, Q> const c(
+			xz - wy,
+			yz + wx,
+			rr.w + rr.z - rr.x - rr.y,
+			-(x.dual.w * r.z + x.dual.x * r.y - x.dual.y * r.x - x.dual.z * r.w));
+
+		return mat<3, 4, T, Q>(a, b, c);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> dualquat_cast(mat<2, 4, T, Q> const& x)
+	{
+		return tdualquat<T, Q>(
+			qua<T, Q>::wxyz( x[0].w, x[0].x, x[0].y, x[0].z ),
+			qua<T, Q>::wxyz( x[1].w, x[1].x, x[1].y, x[1].z ));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER tdualquat<T, Q> dualquat_cast(mat<3, 4, T, Q> const& x)
+	{
+		qua<T, Q> real;
+
+		T const trace = x[0].x + x[1].y + x[2].z;
+		if(trace > static_cast<T>(0))
+		{
+			T const r = sqrt(T(1) + trace);
+			T const invr = static_cast<T>(0.5) / r;
+			real.w = static_cast<T>(0.5) * r;
+			real.x = (x[2].y - x[1].z) * invr;
+			real.y = (x[0].z - x[2].x) * invr;
+			real.z = (x[1].x - x[0].y) * invr;
+		}
+		else if(x[0].x > x[1].y && x[0].x > x[2].z)
+		{
+			T const r = sqrt(T(1) + x[0].x - x[1].y - x[2].z);
+			T const invr = static_cast<T>(0.5) / r;
+			real.x = static_cast<T>(0.5)*r;
+			real.y = (x[1].x + x[0].y) * invr;
+			real.z = (x[0].z + x[2].x) * invr;
+			real.w = (x[2].y - x[1].z) * invr;
+		}
+		else if(x[1].y > x[2].z)
+		{
+			T const r = sqrt(T(1) + x[1].y - x[0].x - x[2].z);
+			T const invr = static_cast<T>(0.5) / r;
+			real.x = (x[1].x + x[0].y) * invr;
+			real.y = static_cast<T>(0.5) * r;
+			real.z = (x[2].y + x[1].z) * invr;
+			real.w = (x[0].z - x[2].x) * invr;
+		}
+		else
+		{
+			T const r = sqrt(T(1) + x[2].z - x[0].x - x[1].y);
+			T const invr = static_cast<T>(0.5) / r;
+			real.x = (x[0].z + x[2].x) * invr;
+			real.y = (x[2].y + x[1].z) * invr;
+			real.z = static_cast<T>(0.5) * r;
+			real.w = (x[1].x - x[0].y) * invr;
+		}
+
+		qua<T, Q> dual;
+		dual.x =  static_cast<T>(0.5) * ( x[0].w * real.w + x[1].w * real.z - x[2].w * real.y);
+		dual.y =  static_cast<T>(0.5) * (-x[0].w * real.z + x[1].w * real.w + x[2].w * real.x);
+		dual.z =  static_cast<T>(0.5) * ( x[0].w * real.y - x[1].w * real.x + x[2].w * real.w);
+		dual.w = -static_cast<T>(0.5) * ( x[0].w * real.x + x[1].w * real.y + x[2].w * real.z);
+		return tdualquat<T, Q>(real, dual);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/easing.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/easing.hpp
new file mode 100644
index 000000000000..da89826a90eb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/easing.hpp
@@ -0,0 +1,219 @@
+/// @ref gtx_easing
+/// @file glm/gtx/easing.hpp
+/// @author Robert Chisholm
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_easing GLM_GTX_easing
+/// @ingroup gtx
+///
+/// Include <glm/gtx/easing.hpp> to use the features of this extension.
+///
+/// Easing functions for animations and transitions
+/// All functions take a parameter x in the range [0.0,1.0]
+///
+/// Based on the AHEasing project of Warren Moore (https://github.com/warrenm/AHEasing)
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/constants.hpp"
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_easing is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_easing extension included")
+#	endif
+#endif
+
+namespace glm{
+	/// @addtogroup gtx_easing
+	/// @{
+
+	/// Modelled after the line y = x
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType linearInterpolation(genType const & a);
+
+	/// Modelled after the parabola y = x^2
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quadraticEaseIn(genType const & a);
+
+	/// Modelled after the parabola y = -x^2 + 2x
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quadraticEaseOut(genType const & a);
+
+	/// Modelled after the piecewise quadratic
+	/// y = (1/2)((2x)^2)				; [0, 0.5)
+	/// y = -(1/2)((2x-1)*(2x-3) - 1)	; [0.5, 1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quadraticEaseInOut(genType const & a);
+
+	/// Modelled after the cubic y = x^3
+	template <typename genType>
+	GLM_FUNC_DECL genType cubicEaseIn(genType const & a);
+
+	/// Modelled after the cubic y = (x - 1)^3 + 1
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType cubicEaseOut(genType const & a);
+
+	/// Modelled after the piecewise cubic
+	/// y = (1/2)((2x)^3)		; [0, 0.5)
+	/// y = (1/2)((2x-2)^3 + 2)	; [0.5, 1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType cubicEaseInOut(genType const & a);
+
+	/// Modelled after the quartic x^4
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quarticEaseIn(genType const & a);
+
+	/// Modelled after the quartic y = 1 - (x - 1)^4
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quarticEaseOut(genType const & a);
+
+	/// Modelled after the piecewise quartic
+	/// y = (1/2)((2x)^4)			; [0, 0.5)
+	/// y = -(1/2)((2x-2)^4 - 2)	; [0.5, 1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quarticEaseInOut(genType const & a);
+
+	/// Modelled after the quintic y = x^5
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quinticEaseIn(genType const & a);
+
+	/// Modelled after the quintic y = (x - 1)^5 + 1
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quinticEaseOut(genType const & a);
+
+	/// Modelled after the piecewise quintic
+	/// y = (1/2)((2x)^5)		; [0, 0.5)
+	/// y = (1/2)((2x-2)^5 + 2) ; [0.5, 1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType quinticEaseInOut(genType const & a);
+
+	/// Modelled after quarter-cycle of sine wave
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType sineEaseIn(genType const & a);
+
+	/// Modelled after quarter-cycle of sine wave (different phase)
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType sineEaseOut(genType const & a);
+
+	/// Modelled after half sine wave
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType sineEaseInOut(genType const & a);
+
+	/// Modelled after shifted quadrant IV of unit circle
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType circularEaseIn(genType const & a);
+
+	/// Modelled after shifted quadrant II of unit circle
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType circularEaseOut(genType const & a);
+
+	/// Modelled after the piecewise circular function
+	/// y = (1/2)(1 - sqrt(1 - 4x^2))			; [0, 0.5)
+	/// y = (1/2)(sqrt(-(2x - 3)*(2x - 1)) + 1) ; [0.5, 1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType circularEaseInOut(genType const & a);
+
+	/// Modelled after the exponential function y = 2^(10(x - 1))
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType exponentialEaseIn(genType const & a);
+
+	/// Modelled after the exponential function y = -2^(-10x) + 1
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType exponentialEaseOut(genType const & a);
+
+	/// Modelled after the piecewise exponential
+	/// y = (1/2)2^(10(2x - 1))			; [0,0.5)
+	/// y = -(1/2)*2^(-10(2x - 1))) + 1 ; [0.5,1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType exponentialEaseInOut(genType const & a);
+
+	/// Modelled after the damped sine wave y = sin(13pi/2*x)*pow(2, 10 * (x - 1))
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType elasticEaseIn(genType const & a);
+
+	/// Modelled after the damped sine wave y = sin(-13pi/2*(x + 1))*pow(2, -10x) + 1
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType elasticEaseOut(genType const & a);
+
+	/// Modelled after the piecewise exponentially-damped sine wave:
+	/// y = (1/2)*sin(13pi/2*(2*x))*pow(2, 10 * ((2*x) - 1))		; [0,0.5)
+	/// y = (1/2)*(sin(-13pi/2*((2x-1)+1))*pow(2,-10(2*x-1)) + 2)	; [0.5, 1]
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType elasticEaseInOut(genType const & a);
+
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType backEaseIn(genType const& a);
+
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType backEaseOut(genType const& a);
+
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType backEaseInOut(genType const& a);
+
+	/// @param a parameter
+	/// @param o Optional overshoot modifier
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType backEaseIn(genType const& a, genType const& o);
+
+	/// @param a parameter
+	/// @param o Optional overshoot modifier
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType backEaseOut(genType const& a, genType const& o);
+
+	/// @param a parameter
+	/// @param o Optional overshoot modifier
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType backEaseInOut(genType const& a, genType const& o);
+
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType bounceEaseIn(genType const& a);
+
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType bounceEaseOut(genType const& a);
+
+	/// @see gtx_easing
+	template <typename genType>
+	GLM_FUNC_DECL genType bounceEaseInOut(genType const& a);
+
+	/// @}
+}//namespace glm
+
+#include "easing.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/easing.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/easing.inl
new file mode 100644
index 000000000000..b599c30664af
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/easing.inl
@@ -0,0 +1,436 @@
+/// @ref gtx_easing
+
+#include <cmath>
+
+namespace glm{
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType linearInterpolation(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return a;
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quadraticEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return a * a;
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quadraticEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return -(a * (a - static_cast<genType>(2)));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quadraticEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+		{
+			return static_cast<genType>(2) * a * a;
+		}
+		else
+		{
+			return (-static_cast<genType>(2) * a * a) + (4 * a) - one<genType>();
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType cubicEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return a * a * a;
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType cubicEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		genType const f = a - one<genType>();
+		return f * f * f + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType cubicEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if (a < static_cast<genType>(0.5))
+		{
+			return static_cast<genType>(4) * a * a * a;
+		}
+		else
+		{
+			genType const f = ((static_cast<genType>(2) * a) - static_cast<genType>(2));
+			return static_cast<genType>(0.5) * f * f * f + one<genType>();
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quarticEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return a * a * a * a;
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quarticEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		genType const f = (a - one<genType>());
+		return f * f * f * (one<genType>() - a) + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quarticEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+		{
+			return static_cast<genType>(8) * a * a * a * a;
+		}
+		else
+		{
+			genType const f = (a - one<genType>());
+			return -static_cast<genType>(8) * f * f * f * f + one<genType>();
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quinticEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return a * a * a * a * a;
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quinticEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		genType const f = (a - one<genType>());
+		return f * f * f * f * f + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType quinticEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+		{
+			return static_cast<genType>(16) * a * a * a * a * a;
+		}
+		else
+		{
+			genType const f = ((static_cast<genType>(2) * a) - static_cast<genType>(2));
+			return static_cast<genType>(0.5) * f * f * f * f * f + one<genType>();
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType sineEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return sin((a - one<genType>()) * half_pi<genType>()) + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType sineEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return sin(a * half_pi<genType>());
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType sineEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return static_cast<genType>(0.5) * (one<genType>() - cos(a * pi<genType>()));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType circularEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return one<genType>() - sqrt(one<genType>() - (a * a));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType circularEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return sqrt((static_cast<genType>(2) - a) * a);
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType circularEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+		{
+			return static_cast<genType>(0.5) * (one<genType>() - std::sqrt(one<genType>() - static_cast<genType>(4) * (a * a)));
+		}
+		else
+		{
+			return static_cast<genType>(0.5) * (std::sqrt(-((static_cast<genType>(2) * a) - static_cast<genType>(3)) * ((static_cast<genType>(2) * a) - one<genType>())) + one<genType>());
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType exponentialEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a <= zero<genType>())
+			return a;
+		else
+		{
+			genType const Complementary = a - one<genType>();
+			genType const Two = static_cast<genType>(2);
+			
+			return glm::pow(Two, Complementary * static_cast<genType>(10));
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType exponentialEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a >= one<genType>())
+			return a;
+		else
+		{
+			return one<genType>() - glm::pow(static_cast<genType>(2), -static_cast<genType>(10) * a);
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType exponentialEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+			return static_cast<genType>(0.5) * glm::pow(static_cast<genType>(2), (static_cast<genType>(20) * a) - static_cast<genType>(10));
+		else
+			return -static_cast<genType>(0.5) * glm::pow(static_cast<genType>(2), (-static_cast<genType>(20) * a) + static_cast<genType>(10)) + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType elasticEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return std::sin(static_cast<genType>(13) * half_pi<genType>() * a) * glm::pow(static_cast<genType>(2), static_cast<genType>(10) * (a - one<genType>()));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType elasticEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return std::sin(-static_cast<genType>(13) * half_pi<genType>() * (a + one<genType>())) * glm::pow(static_cast<genType>(2), -static_cast<genType>(10) * a) + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType elasticEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+			return static_cast<genType>(0.5) * std::sin(static_cast<genType>(13) * half_pi<genType>() * (static_cast<genType>(2) * a)) * glm::pow(static_cast<genType>(2), static_cast<genType>(10) * ((static_cast<genType>(2) * a) - one<genType>()));
+		else
+			return static_cast<genType>(0.5) * (std::sin(-static_cast<genType>(13) * half_pi<genType>() * ((static_cast<genType>(2) * a - one<genType>()) + one<genType>())) * glm::pow(static_cast<genType>(2), -static_cast<genType>(10) * (static_cast<genType>(2) * a - one<genType>())) + static_cast<genType>(2));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType backEaseIn(genType const& a, genType const& o)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		genType z = ((o + one<genType>()) * a) - o;
+		return (a * a * z);
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType backEaseOut(genType const& a, genType const& o)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		genType n = a - one<genType>();
+		genType z = ((o + one<genType>()) * n) + o;
+		return (n * n * z) + one<genType>();
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType backEaseInOut(genType const& a, genType const& o)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		genType s = o * static_cast<genType>(1.525);
+		genType x = static_cast<genType>(0.5);
+		genType n = a / static_cast<genType>(0.5);
+
+		if (n < static_cast<genType>(1))
+		{
+			genType z = ((s + static_cast<genType>(1)) * n) - s;
+			genType m = n * n * z;
+			return x * m;
+		}
+		else 
+		{
+			n -= static_cast<genType>(2);
+			genType z = ((s + static_cast<genType>(1)) * n) + s;
+			genType m = (n*n*z) + static_cast<genType>(2);
+			return x * m;
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType backEaseIn(genType const& a)
+	{
+		return backEaseIn(a, static_cast<genType>(1.70158));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType backEaseOut(genType const& a)
+	{
+		return backEaseOut(a, static_cast<genType>(1.70158));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType backEaseInOut(genType const& a)
+	{
+		return backEaseInOut(a, static_cast<genType>(1.70158));
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType bounceEaseOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(4.0 / 11.0))
+		{
+			return (static_cast<genType>(121) * a * a) / static_cast<genType>(16);
+		}
+		else if(a < static_cast<genType>(8.0 / 11.0))
+		{
+			return (static_cast<genType>(363.0 / 40.0) * a * a) - (static_cast<genType>(99.0 / 10.0) * a) + static_cast<genType>(17.0 / 5.0);
+		}
+		else if(a < static_cast<genType>(9.0 / 10.0))
+		{
+			return (static_cast<genType>(4356.0 / 361.0) * a * a) - (static_cast<genType>(35442.0 / 1805.0) * a) + static_cast<genType>(16061.0 / 1805.0);
+		}
+		else
+		{
+			return (static_cast<genType>(54.0 / 5.0) * a * a) - (static_cast<genType>(513.0 / 25.0) * a) + static_cast<genType>(268.0 / 25.0);
+		}
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType bounceEaseIn(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		return one<genType>() - bounceEaseOut(one<genType>() - a);
+	}
+
+	template <typename genType>
+	GLM_FUNC_QUALIFIER genType bounceEaseInOut(genType const& a)
+	{
+		// Only defined in [0, 1]
+		assert(a >= zero<genType>());
+		assert(a <= one<genType>());
+
+		if(a < static_cast<genType>(0.5))
+		{
+			return static_cast<genType>(0.5) * (one<genType>() - bounceEaseOut(one<genType>() - a * static_cast<genType>(2)));
+		}
+		else
+		{
+			return static_cast<genType>(0.5) * bounceEaseOut(a * static_cast<genType>(2) - one<genType>()) + static_cast<genType>(0.5);
+		}
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.hpp
new file mode 100644
index 000000000000..27236973af60
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.hpp
@@ -0,0 +1,335 @@
+/// @ref gtx_euler_angles
+/// @file glm/gtx/euler_angles.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_euler_angles GLM_GTX_euler_angles
+/// @ingroup gtx
+///
+/// Include <glm/gtx/euler_angles.hpp> to use the features of this extension.
+///
+/// Build matrices from Euler angles.
+///
+/// Extraction of Euler angles from rotation matrix.
+/// Based on the original paper 2014 Mike Day - Extracting Euler Angles from a Rotation Matrix.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_euler_angles is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_euler_angles extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_euler_angles
+	/// @{
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from an euler angle X.
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleX(
+		T const& angleX);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from an euler angle Y.
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleY(
+		T const& angleY);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from an euler angle Z.
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZ(
+		T const& angleZ);
+
+	/// Creates a 3D 4 * 4 homogeneous derived matrix from the rotation matrix about X-axis.
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> derivedEulerAngleX(
+		T const & angleX, T const & angularVelocityX);
+
+	/// Creates a 3D 4 * 4 homogeneous derived matrix from the rotation matrix about Y-axis.
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> derivedEulerAngleY(
+		T const & angleY, T const & angularVelocityY);
+
+	/// Creates a 3D 4 * 4 homogeneous derived matrix from the rotation matrix about Z-axis.
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> derivedEulerAngleZ(
+		T const & angleZ, T const & angularVelocityZ);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (X * Y).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleXY(
+		T const& angleX,
+		T const& angleY);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * X).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleYX(
+		T const& angleY,
+		T const& angleX);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (X * Z).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleXZ(
+		T const& angleX,
+		T const& angleZ);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Z * X).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZX(
+		T const& angle,
+		T const& angleX);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * Z).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleYZ(
+		T const& angleY,
+		T const& angleZ);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Z * Y).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZY(
+		T const& angleZ,
+		T const& angleY);
+
+    /// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (X * Y * Z).
+    /// @see gtx_euler_angles
+    template<typename T>
+    GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleXYZ(
+        T const& t1,
+        T const& t2,
+        T const& t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * X * Z).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleYXZ(
+		T const& yaw,
+		T const& pitch,
+		T const& roll);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (X * Z * X).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleXZX(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (X * Y * X).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleXYX(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * X * Y).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleYXY(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * Z * Y).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleYZY(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Z * Y * Z).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZYZ(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Z * X * Z).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZXZ(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (X * Z * Y).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleXZY(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * Z * X).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleYZX(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Z * Y * X).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZYX(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Z * X * Y).
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> eulerAngleZXY(
+		T const & t1,
+		T const & t2,
+		T const & t3);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * X * Z).
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<4, 4, T, defaultp> yawPitchRoll(
+		T const& yaw,
+		T const& pitch,
+		T const& roll);
+
+	/// Creates a 2D 2 * 2 rotation matrix from an euler angle.
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<2, 2, T, defaultp> orientate2(T const& angle);
+
+	/// Creates a 2D 4 * 4 homogeneous rotation matrix from an euler angle.
+	/// @see gtx_euler_angles
+	template<typename T>
+	GLM_FUNC_DECL mat<3, 3, T, defaultp> orientate3(T const& angle);
+
+	/// Creates a 3D 3 * 3 rotation matrix from euler angles (Y * X * Z).
+	/// @see gtx_euler_angles
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> orientate3(vec<3, T, Q> const& angles);
+
+	/// Creates a 3D 4 * 4 homogeneous rotation matrix from euler angles (Y * X * Z).
+	/// @see gtx_euler_angles
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> orientate4(vec<3, T, Q> const& angles);
+
+    /// Extracts the (X * Y * Z) Euler angles from the rotation matrix M
+    /// @see gtx_euler_angles
+    template<typename T>
+    GLM_FUNC_DECL void extractEulerAngleXYZ(mat<4, 4, T, defaultp> const& M,
+                                            T & t1,
+                                            T & t2,
+                                            T & t3);
+
+	/// Extracts the (Y * X * Z) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleYXZ(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (X * Z * X) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleXZX(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (X * Y * X) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleXYX(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Y * X * Y) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleYXY(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Y * Z * Y) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleYZY(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Z * Y * Z) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleZYZ(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Z * X * Z) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleZXZ(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (X * Z * Y) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleXZY(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Y * Z * X) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleYZX(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Z * Y * X) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleZYX(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// Extracts the (Z * X * Y) Euler angles from the rotation matrix M
+	/// @see gtx_euler_angles
+	template <typename T>
+	GLM_FUNC_DECL void extractEulerAngleZXY(mat<4, 4, T, defaultp> const & M,
+											T & t1,
+											T & t2,
+											T & t3);
+
+	/// @}
+}//namespace glm
+
+#include "euler_angles.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.inl
new file mode 100644
index 000000000000..134d499d91fc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/euler_angles.inl
@@ -0,0 +1,899 @@
+/// @ref gtx_euler_angles
+
+#include "compatibility.hpp" // glm::atan2
+
+namespace glm
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleX
+	(
+		T const& angleX
+	)
+	{
+		T cosX = glm::cos(angleX);
+		T sinX = glm::sin(angleX);
+
+		return mat<4, 4, T, defaultp>(
+			T(1), T(0), T(0), T(0),
+			T(0), cosX, sinX, T(0),
+			T(0),-sinX, cosX, T(0),
+			T(0), T(0), T(0), T(1));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleY
+	(
+		T const& angleY
+	)
+	{
+		T cosY = glm::cos(angleY);
+		T sinY = glm::sin(angleY);
+
+		return mat<4, 4, T, defaultp>(
+			cosY,	T(0),	-sinY,	T(0),
+			T(0),	T(1),	T(0),	T(0),
+			sinY,	T(0),	cosY,	T(0),
+			T(0),	T(0),	T(0),	T(1));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZ
+	(
+		T const& angleZ
+	)
+	{
+		T cosZ = glm::cos(angleZ);
+		T sinZ = glm::sin(angleZ);
+
+		return mat<4, 4, T, defaultp>(
+			cosZ,	sinZ,	T(0), T(0),
+			-sinZ,	cosZ,	T(0), T(0),
+			T(0),	T(0),	T(1), T(0),
+			T(0),	T(0),	T(0), T(1));
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> derivedEulerAngleX
+	(
+		T const & angleX,
+		T const & angularVelocityX
+	)
+	{
+		T cosX = glm::cos(angleX) * angularVelocityX;
+		T sinX = glm::sin(angleX) * angularVelocityX;
+
+		return mat<4, 4, T, defaultp>(
+			T(0), T(0), T(0), T(0),
+			T(0),-sinX, cosX, T(0),
+			T(0),-cosX,-sinX, T(0),
+			T(0), T(0), T(0), T(0));
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> derivedEulerAngleY
+	(
+		T const & angleY,
+		T const & angularVelocityY
+	)
+	{
+		T cosY = glm::cos(angleY) * angularVelocityY;
+		T sinY = glm::sin(angleY) * angularVelocityY;
+
+		return mat<4, 4, T, defaultp>(
+			-sinY, T(0), -cosY, T(0),
+			 T(0), T(0),  T(0), T(0),
+			 cosY, T(0), -sinY, T(0),
+			 T(0), T(0),  T(0), T(0));
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> derivedEulerAngleZ
+	(
+		T const & angleZ,
+		T const & angularVelocityZ
+	)
+	{
+		T cosZ = glm::cos(angleZ) * angularVelocityZ;
+		T sinZ = glm::sin(angleZ) * angularVelocityZ;
+
+		return mat<4, 4, T, defaultp>(
+			-sinZ,  cosZ, T(0), T(0),
+			-cosZ, -sinZ, T(0), T(0),
+			 T(0),  T(0), T(0), T(0),
+			 T(0),  T(0), T(0), T(0));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleXY
+	(
+		T const& angleX,
+		T const& angleY
+	)
+	{
+		T cosX = glm::cos(angleX);
+		T sinX = glm::sin(angleX);
+		T cosY = glm::cos(angleY);
+		T sinY = glm::sin(angleY);
+
+		return mat<4, 4, T, defaultp>(
+			cosY,   -sinX * -sinY,  cosX * -sinY,   T(0),
+			T(0),   cosX,           sinX,           T(0),
+			sinY,   -sinX * cosY,   cosX * cosY,    T(0),
+			T(0),   T(0),           T(0),           T(1));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleYX
+	(
+		T const& angleY,
+		T const& angleX
+	)
+	{
+		T cosX = glm::cos(angleX);
+		T sinX = glm::sin(angleX);
+		T cosY = glm::cos(angleY);
+		T sinY = glm::sin(angleY);
+
+		return mat<4, 4, T, defaultp>(
+			cosY,          0,      -sinY,    T(0),
+			sinY * sinX,  cosX, cosY * sinX, T(0),
+			sinY * cosX, -sinX, cosY * cosX, T(0),
+			T(0),         T(0),     T(0),    T(1));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleXZ
+	(
+		T const& angleX,
+		T const& angleZ
+	)
+	{
+		return eulerAngleX(angleX) * eulerAngleZ(angleZ);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZX
+	(
+		T const& angleZ,
+		T const& angleX
+	)
+	{
+		return eulerAngleZ(angleZ) * eulerAngleX(angleX);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleYZ
+	(
+		T const& angleY,
+		T const& angleZ
+	)
+	{
+		return eulerAngleY(angleY) * eulerAngleZ(angleZ);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZY
+	(
+		T const& angleZ,
+		T const& angleY
+	)
+	{
+		return eulerAngleZ(angleZ) * eulerAngleY(angleY);
+	}
+
+    template<typename T>
+    GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleXYZ
+    (
+     T const& t1,
+     T const& t2,
+     T const& t3
+     )
+    {
+        T c1 = glm::cos(-t1);
+        T c2 = glm::cos(-t2);
+        T c3 = glm::cos(-t3);
+        T s1 = glm::sin(-t1);
+        T s2 = glm::sin(-t2);
+        T s3 = glm::sin(-t3);
+
+        mat<4, 4, T, defaultp> Result;
+        Result[0][0] = c2 * c3;
+        Result[0][1] =-c1 * s3 + s1 * s2 * c3;
+        Result[0][2] = s1 * s3 + c1 * s2 * c3;
+        Result[0][3] = static_cast<T>(0);
+        Result[1][0] = c2 * s3;
+        Result[1][1] = c1 * c3 + s1 * s2 * s3;
+        Result[1][2] =-s1 * c3 + c1 * s2 * s3;
+        Result[1][3] = static_cast<T>(0);
+        Result[2][0] =-s2;
+        Result[2][1] = s1 * c2;
+        Result[2][2] = c1 * c2;
+        Result[2][3] = static_cast<T>(0);
+        Result[3][0] = static_cast<T>(0);
+        Result[3][1] = static_cast<T>(0);
+        Result[3][2] = static_cast<T>(0);
+        Result[3][3] = static_cast<T>(1);
+        return Result;
+    }
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleYXZ
+	(
+		T const& yaw,
+		T const& pitch,
+		T const& roll
+	)
+	{
+		T tmp_ch = glm::cos(yaw);
+		T tmp_sh = glm::sin(yaw);
+		T tmp_cp = glm::cos(pitch);
+		T tmp_sp = glm::sin(pitch);
+		T tmp_cb = glm::cos(roll);
+		T tmp_sb = glm::sin(roll);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = tmp_ch * tmp_cb + tmp_sh * tmp_sp * tmp_sb;
+		Result[0][1] = tmp_sb * tmp_cp;
+		Result[0][2] = -tmp_sh * tmp_cb + tmp_ch * tmp_sp * tmp_sb;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] = -tmp_ch * tmp_sb + tmp_sh * tmp_sp * tmp_cb;
+		Result[1][1] = tmp_cb * tmp_cp;
+		Result[1][2] = tmp_sb * tmp_sh + tmp_ch * tmp_sp * tmp_cb;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = tmp_sh * tmp_cp;
+		Result[2][1] = -tmp_sp;
+		Result[2][2] = tmp_ch * tmp_cp;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleXZX
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c2;
+		Result[0][1] = c1 * s2;
+		Result[0][2] = s1 * s2;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] =-c3 * s2;
+		Result[1][1] = c1 * c2 * c3 - s1 * s3;
+		Result[1][2] = c1 * s3 + c2 * c3 * s1;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = s2 * s3;
+		Result[2][1] =-c3 * s1 - c1 * c2 * s3;
+		Result[2][2] = c1 * c3 - c2 * s1 * s3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleXYX
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c2;
+		Result[0][1] = s1 * s2;
+		Result[0][2] =-c1 * s2;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] = s2 * s3;
+		Result[1][1] = c1 * c3 - c2 * s1 * s3;
+		Result[1][2] = c3 * s1 + c1 * c2 * s3;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c3 * s2;
+		Result[2][1] =-c1 * s3 - c2 * c3 * s1;
+		Result[2][2] = c1 * c2 * c3 - s1 * s3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleYXY
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c3 - c2 * s1 * s3;
+		Result[0][1] = s2* s3;
+		Result[0][2] =-c3 * s1 - c1 * c2 * s3;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] = s1 * s2;
+		Result[1][1] = c2;
+		Result[1][2] = c1 * s2;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c1 * s3 + c2 * c3 * s1;
+		Result[2][1] =-c3 * s2;
+		Result[2][2] = c1 * c2 * c3 - s1 * s3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleYZY
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c2 * c3 - s1 * s3;
+		Result[0][1] = c3 * s2;
+		Result[0][2] =-c1 * s3 - c2 * c3 * s1;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] =-c1 * s2;
+		Result[1][1] = c2;
+		Result[1][2] = s1 * s2;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c3 * s1 + c1 * c2 * s3;
+		Result[2][1] = s2 * s3;
+		Result[2][2] = c1 * c3 - c2 * s1 * s3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZYZ
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c2 * c3 - s1 * s3;
+		Result[0][1] = c1 * s3 + c2 * c3 * s1;
+		Result[0][2] =-c3 * s2;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] =-c3 * s1 - c1 * c2 * s3;
+		Result[1][1] = c1 * c3 - c2 * s1 * s3;
+		Result[1][2] = s2 * s3;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c1 * s2;
+		Result[2][1] = s1 * s2;
+		Result[2][2] = c2;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZXZ
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c3 - c2 * s1 * s3;
+		Result[0][1] = c3 * s1 + c1 * c2 * s3;
+		Result[0][2] = s2 *s3;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] =-c1 * s3 - c2 * c3 * s1;
+		Result[1][1] = c1 * c2 * c3 - s1 * s3;
+		Result[1][2] = c3 * s2;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = s1 * s2;
+		Result[2][1] =-c1 * s2;
+		Result[2][2] = c2;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleXZY
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c2 * c3;
+		Result[0][1] = s1 * s3 + c1 * c3 * s2;
+		Result[0][2] = c3 * s1 * s2 - c1 * s3;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] =-s2;
+		Result[1][1] = c1 * c2;
+		Result[1][2] = c2 * s1;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c2 * s3;
+		Result[2][1] = c1 * s2 * s3 - c3 * s1;
+		Result[2][2] = c1 * c3 + s1 * s2 *s3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleYZX
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c2;
+		Result[0][1] = s2;
+		Result[0][2] =-c2 * s1;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] = s1 * s3 - c1 * c3 * s2;
+		Result[1][1] = c2 * c3;
+		Result[1][2] = c1 * s3 + c3 * s1 * s2;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c3 * s1 + c1 * s2 * s3;
+		Result[2][1] =-c2 * s3;
+		Result[2][2] = c1 * c3 - s1 * s2 * s3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZYX
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c2;
+		Result[0][1] = c2 * s1;
+		Result[0][2] =-s2;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] = c1 * s2 * s3 - c3 * s1;
+		Result[1][1] = c1 * c3 + s1 * s2 * s3;
+		Result[1][2] = c2 * s3;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = s1 * s3 + c1 * c3 * s2;
+		Result[2][1] = c3 * s1 * s2 - c1 * s3;
+		Result[2][2] = c2 * c3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> eulerAngleZXY
+	(
+		T const & t1,
+		T const & t2,
+		T const & t3
+	)
+	{
+		T c1 = glm::cos(t1);
+		T s1 = glm::sin(t1);
+		T c2 = glm::cos(t2);
+		T s2 = glm::sin(t2);
+		T c3 = glm::cos(t3);
+		T s3 = glm::sin(t3);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = c1 * c3 - s1 * s2 * s3;
+		Result[0][1] = c3 * s1 + c1 * s2 * s3;
+		Result[0][2] =-c2 * s3;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] =-c2 * s1;
+		Result[1][1] = c1 * c2;
+		Result[1][2] = s2;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = c1 * s3 + c3 * s1 * s2;
+		Result[2][1] = s1 * s3 - c1 * c3 * s2;
+		Result[2][2] = c2 * c3;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, defaultp> yawPitchRoll
+	(
+		T const& yaw,
+		T const& pitch,
+		T const& roll
+	)
+	{
+		T tmp_ch = glm::cos(yaw);
+		T tmp_sh = glm::sin(yaw);
+		T tmp_cp = glm::cos(pitch);
+		T tmp_sp = glm::sin(pitch);
+		T tmp_cb = glm::cos(roll);
+		T tmp_sb = glm::sin(roll);
+
+		mat<4, 4, T, defaultp> Result;
+		Result[0][0] = tmp_ch * tmp_cb + tmp_sh * tmp_sp * tmp_sb;
+		Result[0][1] = tmp_sb * tmp_cp;
+		Result[0][2] = -tmp_sh * tmp_cb + tmp_ch * tmp_sp * tmp_sb;
+		Result[0][3] = static_cast<T>(0);
+		Result[1][0] = -tmp_ch * tmp_sb + tmp_sh * tmp_sp * tmp_cb;
+		Result[1][1] = tmp_cb * tmp_cp;
+		Result[1][2] = tmp_sb * tmp_sh + tmp_ch * tmp_sp * tmp_cb;
+		Result[1][3] = static_cast<T>(0);
+		Result[2][0] = tmp_sh * tmp_cp;
+		Result[2][1] = -tmp_sp;
+		Result[2][2] = tmp_ch * tmp_cp;
+		Result[2][3] = static_cast<T>(0);
+		Result[3][0] = static_cast<T>(0);
+		Result[3][1] = static_cast<T>(0);
+		Result[3][2] = static_cast<T>(0);
+		Result[3][3] = static_cast<T>(1);
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, defaultp> orientate2
+	(
+		T const& angle
+	)
+	{
+		T c = glm::cos(angle);
+		T s = glm::sin(angle);
+
+		mat<2, 2, T, defaultp> Result;
+		Result[0][0] = c;
+		Result[0][1] = s;
+		Result[1][0] = -s;
+		Result[1][1] = c;
+		return Result;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, defaultp> orientate3
+	(
+		T const& angle
+	)
+	{
+		T c = glm::cos(angle);
+		T s = glm::sin(angle);
+
+		mat<3, 3, T, defaultp> Result;
+		Result[0][0] = c;
+		Result[0][1] = s;
+		Result[0][2] = T(0.0);
+		Result[1][0] = -s;
+		Result[1][1] = c;
+		Result[1][2] = T(0.0);
+		Result[2][0] = T(0.0);
+		Result[2][1] = T(0.0);
+		Result[2][2] = T(1.0);
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> orientate3
+	(
+		vec<3, T, Q> const& angles
+	)
+	{
+		return mat<3, 3, T, Q>(yawPitchRoll(angles.z, angles.x, angles.y));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> orientate4
+	(
+		vec<3, T, Q> const& angles
+	)
+	{
+		return yawPitchRoll(angles.z, angles.x, angles.y);
+	}
+
+    template<typename T>
+    GLM_FUNC_DECL void extractEulerAngleXYZ(mat<4, 4, T, defaultp> const& M,
+                                            T & t1,
+                                            T & t2,
+                                            T & t3)
+    {
+        T T1 = glm::atan2(M[2][1], M[2][2]);
+        T C2 = glm::sqrt(M[0][0]*M[0][0] + M[1][0]*M[1][0]);
+        T T2 = glm::atan2(-M[2][0], C2);
+        T S1 = glm::sin(T1);
+        T C1 = glm::cos(T1);
+        T T3 = glm::atan2(S1*M[0][2] - C1*M[0][1], C1*M[1][1] - S1*M[1][2  ]);
+        t1 = -T1;
+        t2 = -T2;
+        t3 = -T3;
+    }
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleYXZ(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[2][0], M[2][2]);
+		T C2 = glm::sqrt(M[0][1]*M[0][1] + M[1][1]*M[1][1]);
+		T T2 = glm::atan2(-M[2][1], C2);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(S1*M[1][2] - C1*M[1][0], C1*M[0][0] - S1*M[0][2]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleXZX(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[0][2], M[0][1]);
+		T S2 = glm::sqrt(M[1][0]*M[1][0] + M[2][0]*M[2][0]);
+		T T2 = glm::atan2(S2, M[0][0]);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(C1*M[1][2] - S1*M[1][1], C1*M[2][2] - S1*M[2][1]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleXYX(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[0][1], -M[0][2]);
+		T S2 = glm::sqrt(M[1][0]*M[1][0] + M[2][0]*M[2][0]);
+		T T2 = glm::atan2(S2, M[0][0]);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(-C1*M[2][1] - S1*M[2][2], C1*M[1][1] + S1*M[1][2]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleYXY(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[1][0], M[1][2]);
+		T S2 = glm::sqrt(M[0][1]*M[0][1] + M[2][1]*M[2][1]);
+		T T2 = glm::atan2(S2, M[1][1]);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(C1*M[2][0] - S1*M[2][2], C1*M[0][0] - S1*M[0][2]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleYZY(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[1][2], -M[1][0]);
+		T S2 = glm::sqrt(M[0][1]*M[0][1] + M[2][1]*M[2][1]);
+		T T2 = glm::atan2(S2, M[1][1]);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(-S1*M[0][0] - C1*M[0][2], S1*M[2][0] + C1*M[2][2]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleZYZ(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[2][1], M[2][0]);
+		T S2 = glm::sqrt(M[0][2]*M[0][2] + M[1][2]*M[1][2]);
+		T T2 = glm::atan2(S2, M[2][2]);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(C1*M[0][1] - S1*M[0][0], C1*M[1][1] - S1*M[1][0]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleZXZ(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[2][0], -M[2][1]);
+		T S2 = glm::sqrt(M[0][2]*M[0][2] + M[1][2]*M[1][2]);
+		T T2 = glm::atan2(S2, M[2][2]);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(-C1*M[1][0] - S1*M[1][1], C1*M[0][0] + S1*M[0][1]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleXZY(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[1][2], M[1][1]);
+		T C2 = glm::sqrt(M[0][0]*M[0][0] + M[2][0]*M[2][0]);
+		T T2 = glm::atan2(-M[1][0], C2);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(S1*M[0][1] - C1*M[0][2], C1*M[2][2] - S1*M[2][1]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleYZX(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(-M[0][2], M[0][0]);
+		T C2 = glm::sqrt(M[1][1]*M[1][1] + M[2][1]*M[2][1]);
+		T T2 = glm::atan2(M[0][1], C2);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(S1*M[1][0] + C1*M[1][2], S1*M[2][0] + C1*M[2][2]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleZYX(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(M[0][1], M[0][0]);
+		T C2 = glm::sqrt(M[1][2]*M[1][2] + M[2][2]*M[2][2]);
+		T T2 = glm::atan2(-M[0][2], C2);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(S1*M[2][0] - C1*M[2][1], C1*M[1][1] - S1*M[1][0]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+
+	template <typename T>
+	GLM_FUNC_QUALIFIER void extractEulerAngleZXY(mat<4, 4, T, defaultp> const & M,
+												 T & t1,
+												 T & t2,
+												 T & t3)
+	{
+		T T1 = glm::atan2(-M[1][0], M[1][1]);
+		T C2 = glm::sqrt(M[0][2]*M[0][2] + M[2][2]*M[2][2]);
+		T T2 = glm::atan2(M[1][2], C2);
+		T S1 = glm::sin(T1);
+		T C1 = glm::cos(T1);
+		T T3 = glm::atan2(C1*M[2][0] + S1*M[2][1], C1*M[0][0] + S1*M[0][1]);
+		t1 = T1;
+		t2 = T2;
+		t3 = T3;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/extend.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/extend.hpp
new file mode 100644
index 000000000000..28b7c5c014a4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/extend.hpp
@@ -0,0 +1,42 @@
+/// @ref gtx_extend
+/// @file glm/gtx/extend.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_extend GLM_GTX_extend
+/// @ingroup gtx
+///
+/// Include <glm/gtx/extend.hpp> to use the features of this extension.
+///
+/// Extend a position from a source to a position at a defined length.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_extend is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_extend extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_extend
+	/// @{
+
+	/// Extends of Length the Origin position using the (Source - Origin) direction.
+	/// @see gtx_extend
+	template<typename genType>
+	GLM_FUNC_DECL genType extend(
+		genType const& Origin,
+		genType const& Source,
+		typename genType::value_type const Length);
+
+	/// @}
+}//namespace glm
+
+#include "extend.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/extend.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/extend.inl
new file mode 100644
index 000000000000..32128eb209ac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/extend.inl
@@ -0,0 +1,48 @@
+/// @ref gtx_extend
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType extend
+	(
+		genType const& Origin,
+		genType const& Source,
+		genType const& Distance
+	)
+	{
+		return Origin + (Source - Origin) * Distance;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> extend
+	(
+		vec<2, T, Q> const& Origin,
+		vec<2, T, Q> const& Source,
+		T const& Distance
+	)
+	{
+		return Origin + (Source - Origin) * Distance;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> extend
+	(
+		vec<3, T, Q> const& Origin,
+		vec<3, T, Q> const& Source,
+		T const& Distance
+	)
+	{
+		return Origin + (Source - Origin) * Distance;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> extend
+	(
+		vec<4, T, Q> const& Origin,
+		vec<4, T, Q> const& Source,
+		T const& Distance
+	)
+	{
+		return Origin + (Source - Origin) * Distance;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.hpp
new file mode 100644
index 000000000000..20cd89b0d519
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.hpp
@@ -0,0 +1,137 @@
+/// @ref gtx_extended_min_max
+/// @file glm/gtx/extended_min_max.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_extended_min_max GLM_GTX_extended_min_max
+/// @ingroup gtx
+///
+/// Include <glm/gtx/extended_min_max.hpp> to use the features of this extension.
+///
+/// Min and max functions for 3 to 4 parameters.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../ext/vector_common.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_extended_min_max is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_extended_min_max extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_extended_min_max
+	/// @{
+
+	/// Return the minimum component-wise values of 3 inputs
+	/// @see gtx_extented_min_max
+	template<typename T>
+	GLM_FUNC_DECL T min(
+		T const& x,
+		T const& y,
+		T const& z);
+
+	/// Return the minimum component-wise values of 3 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> min(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z);
+
+	/// Return the minimum component-wise values of 3 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> min(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z);
+
+	/// Return the minimum component-wise values of 4 inputs
+	/// @see gtx_extented_min_max
+	template<typename T>
+	GLM_FUNC_DECL T min(
+		T const& x,
+		T const& y,
+		T const& z,
+		T const& w);
+
+	/// Return the minimum component-wise values of 4 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> min(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z,
+		typename C<T>::T const& w);
+
+	/// Return the minimum component-wise values of 4 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> min(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z,
+		C<T> const& w);
+
+	/// Return the maximum component-wise values of 3 inputs
+	/// @see gtx_extented_min_max
+	template<typename T>
+	GLM_FUNC_DECL T max(
+		T const& x,
+		T const& y,
+		T const& z);
+
+	/// Return the maximum component-wise values of 3 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> max(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z);
+
+	/// Return the maximum component-wise values of 3 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> max(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z);
+
+	/// Return the maximum component-wise values of 4 inputs
+	/// @see gtx_extented_min_max
+	template<typename T>
+	GLM_FUNC_DECL T max(
+		T const& x,
+		T const& y,
+		T const& z,
+		T const& w);
+
+	/// Return the maximum component-wise values of 4 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> max(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z,
+		typename C<T>::T const& w);
+
+	/// Return the maximum component-wise values of 4 inputs
+	/// @see gtx_extented_min_max
+	template<typename T, template<typename> class C>
+	GLM_FUNC_DECL C<T> max(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z,
+		C<T> const& w);
+
+	/// @}
+}//namespace glm
+
+#include "extended_min_max.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.inl
new file mode 100644
index 000000000000..de5998fadd65
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/extended_min_max.inl
@@ -0,0 +1,138 @@
+/// @ref gtx_extended_min_max
+
+namespace glm
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER T min(
+		T const& x,
+		T const& y,
+		T const& z)
+	{
+		return glm::min(glm::min(x, y), z);
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> min
+	(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z
+	)
+	{
+		return glm::min(glm::min(x, y), z);
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> min
+	(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z
+	)
+	{
+		return glm::min(glm::min(x, y), z);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T min
+	(
+		T const& x,
+		T const& y,
+		T const& z,
+		T const& w
+	)
+	{
+		return glm::min(glm::min(x, y), glm::min(z, w));
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> min
+	(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z,
+		typename C<T>::T const& w
+	)
+	{
+		return glm::min(glm::min(x, y), glm::min(z, w));
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> min
+	(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z,
+		C<T> const& w
+	)
+	{
+		return glm::min(glm::min(x, y), glm::min(z, w));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T max(
+		T const& x,
+		T const& y,
+		T const& z)
+	{
+		return glm::max(glm::max(x, y), z);
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> max
+	(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z
+	)
+	{
+		return glm::max(glm::max(x, y), z);
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> max
+	(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z
+	)
+	{
+		return glm::max(glm::max(x, y), z);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T max
+	(
+		T const& x,
+		T const& y,
+		T const& z,
+		T const& w
+	)
+	{
+		return glm::max(glm::max(x, y), glm::max(z, w));
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> max
+	(
+		C<T> const& x,
+		typename C<T>::T const& y,
+		typename C<T>::T const& z,
+		typename C<T>::T const& w
+	)
+	{
+		return glm::max(glm::max(x, y), glm::max(z, w));
+	}
+
+	template<typename T, template<typename> class C>
+	GLM_FUNC_QUALIFIER C<T> max
+	(
+		C<T> const& x,
+		C<T> const& y,
+		C<T> const& z,
+		C<T> const& w
+	)
+	{
+		return glm::max(glm::max(x, y), glm::max(z, w));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.hpp
new file mode 100644
index 000000000000..7d6c2e19b112
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.hpp
@@ -0,0 +1,45 @@
+/// @ref gtx_exterior_product
+/// @file glm/gtx/exterior_product.hpp
+///
+/// @see core (dependence)
+/// @see gtx_exterior_product (dependence)
+///
+/// @defgroup gtx_exterior_product GLM_GTX_exterior_product
+/// @ingroup gtx
+///
+/// Include <glm/gtx/exterior_product.hpp> to use the features of this extension.
+///
+/// @brief Allow to perform bit operations on integer values
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_exterior_product is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_exterior_product extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_exterior_product
+	/// @{
+
+	/// Returns the cross product of x and y.
+	///
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="https://en.wikipedia.org/wiki/Exterior_algebra#Cross_and_triple_products">Exterior product</a>
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR T cross(vec<2, T, Q> const& v, vec<2, T, Q> const& u);
+
+	/// @}
+} //namespace glm
+
+#include "exterior_product.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.inl
new file mode 100644
index 000000000000..690085d6f669
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/exterior_product.inl
@@ -0,0 +1,26 @@
+/// @ref gtx_exterior_product
+
+#include <limits>
+
+namespace glm {
+namespace detail
+{
+	template<typename T, qualifier Q, bool Aligned>
+	struct compute_cross_vec2
+	{
+		GLM_FUNC_QUALIFIER GLM_CONSTEXPR static T call(vec<2, T, Q> const& v, vec<2, T, Q> const& u)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'cross' accepts only floating-point inputs");
+
+			return v.x * u.y - u.x * v.y;
+		}
+	};
+}//namespace detail
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T cross(vec<2, T, Q> const& x, vec<2, T, Q> const& y)
+	{
+		return detail::compute_cross_vec2<T, Q, detail::is_aligned<Q>::value>::call(x, y);
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.hpp
new file mode 100644
index 000000000000..6fb7286528cf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.hpp
@@ -0,0 +1,95 @@
+/// @ref gtx_fast_exponential
+/// @file glm/gtx/fast_exponential.hpp
+///
+/// @see core (dependence)
+/// @see gtx_half_float (dependence)
+///
+/// @defgroup gtx_fast_exponential GLM_GTX_fast_exponential
+/// @ingroup gtx
+///
+/// Include <glm/gtx/fast_exponential.hpp> to use the features of this extension.
+///
+/// Fast but less accurate implementations of exponential based functions.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_fast_exponential is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_fast_exponential extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_fast_exponential
+	/// @{
+
+	/// Faster than the common pow function but less accurate.
+	/// @see gtx_fast_exponential
+	template<typename genType>
+	GLM_FUNC_DECL genType fastPow(genType x, genType y);
+
+	/// Faster than the common pow function but less accurate.
+	/// @see gtx_fast_exponential
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastPow(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Faster than the common pow function but less accurate.
+	/// @see gtx_fast_exponential
+	template<typename genTypeT, typename genTypeU>
+	GLM_FUNC_DECL genTypeT fastPow(genTypeT x, genTypeU y);
+
+	/// Faster than the common pow function but less accurate.
+	/// @see gtx_fast_exponential
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastPow(vec<L, T, Q> const& x);
+
+	/// Faster than the common exp function but less accurate.
+	/// @see gtx_fast_exponential
+	template<typename T>
+	GLM_FUNC_DECL T fastExp(T x);
+
+	/// Faster than the common exp function but less accurate.
+	/// @see gtx_fast_exponential
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastExp(vec<L, T, Q> const& x);
+
+	/// Faster than the common log function but less accurate.
+	/// @see gtx_fast_exponential
+	template<typename T>
+	GLM_FUNC_DECL T fastLog(T x);
+
+	/// Faster than the common exp2 function but less accurate.
+	/// @see gtx_fast_exponential
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastLog(vec<L, T, Q> const& x);
+
+	/// Faster than the common exp2 function but less accurate.
+	/// @see gtx_fast_exponential
+	template<typename T>
+	GLM_FUNC_DECL T fastExp2(T x);
+
+	/// Faster than the common exp2 function but less accurate.
+	/// @see gtx_fast_exponential
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastExp2(vec<L, T, Q> const& x);
+
+	/// Faster than the common log2 function but less accurate.
+	/// @see gtx_fast_exponential
+	template<typename T>
+	GLM_FUNC_DECL T fastLog2(T x);
+
+	/// Faster than the common log2 function but less accurate.
+	/// @see gtx_fast_exponential
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastLog2(vec<L, T, Q> const& x);
+
+	/// @}
+}//namespace glm
+
+#include "fast_exponential.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.inl
new file mode 100644
index 000000000000..5b1174246bbc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_exponential.inl
@@ -0,0 +1,136 @@
+/// @ref gtx_fast_exponential
+
+namespace glm
+{
+	// fastPow:
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastPow(genType x, genType y)
+	{
+		return exp(y * log(x));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastPow(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		return exp(y * log(x));
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastPow(T x, int y)
+	{
+		T f = static_cast<T>(1);
+		for(int i = 0; i < y; ++i)
+			f *= x;
+		return f;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastPow(vec<L, T, Q> const& x, vec<L, int, Q> const& y)
+	{
+		vec<L, T, Q> Result;
+		for(length_t i = 0, n = x.length(); i < n; ++i)
+			Result[i] = fastPow(x[i], y[i]);
+		return Result;
+	}
+
+	// fastExp
+	// Note: This function provides accurate results only for value between -1 and 1, else avoid it.
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastExp(T x)
+	{
+		// This has a better looking and same performance in release mode than the following code. However, in debug mode it's slower.
+		// return 1.0f + x * (1.0f + x * 0.5f * (1.0f + x * 0.3333333333f * (1.0f + x * 0.25 * (1.0f + x * 0.2f))));
+		T x2 = x * x;
+		T x3 = x2 * x;
+		T x4 = x3 * x;
+		T x5 = x4 * x;
+		return T(1) + x + (x2 * T(0.5)) + (x3 * T(0.1666666667)) + (x4 * T(0.041666667)) + (x5 * T(0.008333333333));
+	}
+	/*  // Try to handle all values of float... but often shower than std::exp, glm::floor and the loop kill the performance
+	GLM_FUNC_QUALIFIER float fastExp(float x)
+	{
+		const float e = 2.718281828f;
+		const float IntegerPart = floor(x);
+		const float FloatPart = x - IntegerPart;
+		float z = 1.f;
+
+		for(int i = 0; i < int(IntegerPart); ++i)
+			z *= e;
+
+		const float x2 = FloatPart * FloatPart;
+		const float x3 = x2 * FloatPart;
+		const float x4 = x3 * FloatPart;
+		const float x5 = x4 * FloatPart;
+		return z * (1.0f + FloatPart + (x2 * 0.5f) + (x3 * 0.1666666667f) + (x4 * 0.041666667f) + (x5 * 0.008333333333f));
+	}
+
+	// Increase accuracy on number bigger that 1 and smaller than -1 but it's not enough for high and negative numbers
+	GLM_FUNC_QUALIFIER float fastExp(float x)
+	{
+		// This has a better looking and same performance in release mode than the following code. However, in debug mode it's slower.
+		// return 1.0f + x * (1.0f + x * 0.5f * (1.0f + x * 0.3333333333f * (1.0f + x * 0.25 * (1.0f + x * 0.2f))));
+		float x2 = x * x;
+		float x3 = x2 * x;
+		float x4 = x3 * x;
+		float x5 = x4 * x;
+		float x6 = x5 * x;
+		float x7 = x6 * x;
+		float x8 = x7 * x;
+		return 1.0f + x + (x2 * 0.5f) + (x3 * 0.1666666667f) + (x4 * 0.041666667f) + (x5 * 0.008333333333f)+ (x6 * 0.00138888888888f) + (x7 * 0.000198412698f) + (x8 * 0.0000248015873f);;
+	}
+	*/
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastExp(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastExp, x);
+	}
+
+	// fastLog
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastLog(genType x)
+	{
+		return std::log(x);
+	}
+
+	/* Slower than the VC7.1 function...
+	GLM_FUNC_QUALIFIER float fastLog(float x)
+	{
+		float y1 = (x - 1.0f) / (x + 1.0f);
+		float y2 = y1 * y1;
+		return 2.0f * y1 * (1.0f + y2 * (0.3333333333f + y2 * (0.2f + y2 * 0.1428571429f)));
+	}
+	*/
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastLog(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastLog, x);
+	}
+
+	//fastExp2, ln2 = 0.69314718055994530941723212145818f
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastExp2(genType x)
+	{
+		return fastExp(static_cast<genType>(0.69314718055994530941723212145818) * x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastExp2(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastExp2, x);
+	}
+
+	// fastLog2, ln2 = 0.69314718055994530941723212145818f
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastLog2(genType x)
+	{
+		return fastLog(x) / static_cast<genType>(0.69314718055994530941723212145818);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastLog2(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastLog2, x);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.hpp
new file mode 100644
index 000000000000..ac42a9c006cb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.hpp
@@ -0,0 +1,98 @@
+/// @ref gtx_fast_square_root
+/// @file glm/gtx/fast_square_root.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_fast_square_root GLM_GTX_fast_square_root
+/// @ingroup gtx
+///
+/// Include <glm/gtx/fast_square_root.hpp> to use the features of this extension.
+///
+/// Fast but less accurate implementations of square root based functions.
+/// - Sqrt optimisation based on Newton's method,
+/// www.gamedev.net/community/forums/topic.asp?topic id=139956
+
+#pragma once
+
+// Dependency:
+#include "../common.hpp"
+#include "../exponential.hpp"
+#include "../geometric.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_fast_square_root is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_fast_square_root extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_fast_square_root
+	/// @{
+
+	/// Faster than the common sqrt function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType fastSqrt(genType x);
+
+	/// Faster than the common sqrt function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastSqrt(vec<L, T, Q> const& x);
+
+	/// Faster than the common inversesqrt function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType fastInverseSqrt(genType x);
+
+	/// Faster than the common inversesqrt function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastInverseSqrt(vec<L, T, Q> const& x);
+
+	/// Faster than the common length function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType fastLength(genType x);
+
+	/// Faster than the common length function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T fastLength(vec<L, T, Q> const& x);
+
+	/// Faster than the common distance function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType fastDistance(genType x, genType y);
+
+	/// Faster than the common distance function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T fastDistance(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Faster than the common normalize function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType fastNormalize(genType x);
+
+	/// Faster than the common normalize function but less accurate.
+	///
+	/// @see gtx_fast_square_root extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> fastNormalize(vec<L, T, Q> const& x);
+
+	/// @}
+}// namespace glm
+
+#include "fast_square_root.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.inl
new file mode 100644
index 000000000000..60fdb7a5ac7c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_square_root.inl
@@ -0,0 +1,75 @@
+/// @ref gtx_fast_square_root
+
+namespace glm
+{
+	// fastSqrt
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastSqrt(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fastSqrt' only accept floating-point input");
+
+		return genType(1) / fastInverseSqrt(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastSqrt(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastSqrt, x);
+	}
+
+	// fastInversesqrt
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastInverseSqrt(genType x)
+	{
+		return detail::compute_inversesqrt<1, genType, lowp, detail::is_aligned<lowp>::value>::call(vec<1, genType, lowp>(x)).x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastInverseSqrt(vec<L, T, Q> const& x)
+	{
+		return detail::compute_inversesqrt<L, T, Q, detail::is_aligned<Q>::value>::call(x);
+	}
+
+	// fastLength
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastLength(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fastLength' only accept floating-point inputs");
+
+		return abs(x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T fastLength(vec<L, T, Q> const& x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'fastLength' only accept floating-point inputs");
+
+		return fastSqrt(dot(x, x));
+	}
+
+	// fastDistance
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastDistance(genType x, genType y)
+	{
+		return fastLength(y - x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T fastDistance(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		return fastLength(y - x);
+	}
+
+	// fastNormalize
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType fastNormalize(genType x)
+	{
+		return x > genType(0) ? genType(1) : -genType(1);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastNormalize(vec<L, T, Q> const& x)
+	{
+		return x * fastInverseSqrt(dot(x, x));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.hpp
new file mode 100644
index 000000000000..2650d6e4d6e3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.hpp
@@ -0,0 +1,79 @@
+/// @ref gtx_fast_trigonometry
+/// @file glm/gtx/fast_trigonometry.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_fast_trigonometry GLM_GTX_fast_trigonometry
+/// @ingroup gtx
+///
+/// Include <glm/gtx/fast_trigonometry.hpp> to use the features of this extension.
+///
+/// Fast but less accurate implementations of trigonometric functions.
+
+#pragma once
+
+// Dependency:
+#include "../gtc/constants.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_fast_trigonometry is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_fast_trigonometry extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_fast_trigonometry
+	/// @{
+
+	/// Wrap an angle to [0 2pi[
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T wrapAngle(T angle);
+
+	/// Faster than the common sin function but less accurate.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastSin(T angle);
+
+	/// Faster than the common cos function but less accurate.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastCos(T angle);
+
+	/// Faster than the common tan function but less accurate.
+	/// Defined between -2pi and 2pi.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastTan(T angle);
+
+	/// Faster than the common asin function but less accurate.
+	/// Defined between -2pi and 2pi.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastAsin(T angle);
+
+	/// Faster than the common acos function but less accurate.
+	/// Defined between -2pi and 2pi.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastAcos(T angle);
+
+	/// Faster than the common atan function but less accurate.
+	/// Defined between -2pi and 2pi.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastAtan(T y, T x);
+
+	/// Faster than the common atan function but less accurate.
+	/// Defined between -2pi and 2pi.
+	/// From GLM_GTX_fast_trigonometry extension.
+	template<typename T>
+	GLM_FUNC_DECL T fastAtan(T angle);
+
+	/// @}
+}//namespace glm
+
+#include "fast_trigonometry.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.inl
new file mode 100644
index 000000000000..1a710cbcd08d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/fast_trigonometry.inl
@@ -0,0 +1,142 @@
+/// @ref gtx_fast_trigonometry
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> taylorCos(vec<L, T, Q> const& x)
+	{
+		return static_cast<T>(1)
+			- (x * x) * (1.f / 2.f)
+			+ ((x * x) * (x * x)) * (1.f / 24.f)
+			- (((x * x) * (x * x)) * (x * x)) * (1.f / 720.f)
+			+ (((x * x) * (x * x)) * ((x * x) * (x * x))) * (1.f / 40320.f);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T cos_52s(T x)
+	{
+		T const xx(x * x);
+		return (T(0.9999932946) + xx * (T(-0.4999124376) + xx * (T(0.0414877472) + xx * T(-0.0012712095))));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> cos_52s(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(cos_52s, x);
+	}
+}//namespace detail
+
+	// wrapAngle
+	template<typename T>
+	GLM_FUNC_QUALIFIER T wrapAngle(T angle)
+	{
+		return abs<T>(mod<T>(angle, two_pi<T>()));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> wrapAngle(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(wrapAngle, x);
+	}
+
+	// cos
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastCos(T x)
+	{
+		T const angle(wrapAngle<T>(x));
+
+		if(angle < half_pi<T>())
+			return detail::cos_52s(angle);
+		if(angle < pi<T>())
+			return -detail::cos_52s(pi<T>() - angle);
+		if(angle < (T(3) * half_pi<T>()))
+			return -detail::cos_52s(angle - pi<T>());
+
+		return detail::cos_52s(two_pi<T>() - angle);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastCos(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastCos, x);
+	}
+
+	// sin
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastSin(T x)
+	{
+		return fastCos<T>(half_pi<T>() - x);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastSin(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastSin, x);
+	}
+
+	// tan
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastTan(T x)
+	{
+		return x + (x * x * x * T(0.3333333333)) + (x * x * x * x * x * T(0.1333333333333)) + (x * x * x * x * x * x * x * T(0.0539682539));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastTan(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastTan, x);
+	}
+
+	// asin
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastAsin(T x)
+	{
+		return x + (x * x * x * T(0.166666667)) + (x * x * x * x * x * T(0.075)) + (x * x * x * x * x * x * x * T(0.0446428571)) + (x * x * x * x * x * x * x * x * x * T(0.0303819444));// + (x * x * x * x * x * x * x * x * x * x * x * T(0.022372159));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastAsin(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastAsin, x);
+	}
+
+	// acos
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastAcos(T x)
+	{
+		return T(1.5707963267948966192313216916398) - fastAsin(x); //(PI / 2)
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastAcos(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastAcos, x);
+	}
+
+	// atan
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastAtan(T y, T x)
+	{
+		T sgn = sign(y) * sign(x);
+		return abs(fastAtan(y / x)) * sgn;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastAtan(vec<L, T, Q> const& y, vec<L, T, Q> const& x)
+	{
+		return detail::functor2<vec, L, T, Q>::call(fastAtan, y, x);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T fastAtan(T x)
+	{
+		return x - (x * x * x * T(0.333333333333)) + (x * x * x * x * x * T(0.2)) - (x * x * x * x * x * x * x * T(0.1428571429)) + (x * x * x * x * x * x * x * x * x * T(0.111111111111)) - (x * x * x * x * x * x * x * x * x * x * x * T(0.0909090909));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> fastAtan(vec<L, T, Q> const& x)
+	{
+		return detail::functor1<vec, L, T, T, Q>::call(fastAtan, x);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/float_notmalize.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/float_notmalize.inl
new file mode 100644
index 000000000000..8cdbc5aaa9c3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/float_notmalize.inl
@@ -0,0 +1,13 @@
+/// @ref gtx_float_normalize
+
+#include <limits>
+
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, float, Q> floatNormalize(vec<L, T, Q> const& v)
+	{
+		return vec<L, float, Q>(v) / static_cast<float>(std::numeric_limits<T>::max());
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/functions.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/functions.hpp
new file mode 100644
index 000000000000..9f4166c4c1c8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/functions.hpp
@@ -0,0 +1,56 @@
+/// @ref gtx_functions
+/// @file glm/gtx/functions.hpp
+///
+/// @see core (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtx_functions GLM_GTX_functions
+/// @ingroup gtx
+///
+/// Include <glm/gtx/functions.hpp> to use the features of this extension.
+///
+/// List of useful common functions.
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+#include "../detail/qualifier.hpp"
+#include "../detail/type_vec2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_functions is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_functions extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_functions
+	/// @{
+
+	/// 1D gauss function
+	///
+	/// @see gtc_epsilon
+	template<typename T>
+	GLM_FUNC_DECL T gauss(
+		T x,
+		T ExpectedValue,
+		T StandardDeviation);
+
+	/// 2D gauss function
+	///
+	/// @see gtc_epsilon
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T gauss(
+		vec<2, T, Q> const& Coord,
+		vec<2, T, Q> const& ExpectedValue,
+		vec<2, T, Q> const& StandardDeviation);
+
+	/// @}
+}//namespace glm
+
+#include "functions.inl"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/functions.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/functions.inl
new file mode 100644
index 000000000000..29cbb20b80fa
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/functions.inl
@@ -0,0 +1,30 @@
+/// @ref gtx_functions
+
+#include "../exponential.hpp"
+
+namespace glm
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER T gauss
+	(
+		T x,
+		T ExpectedValue,
+		T StandardDeviation
+	)
+	{
+		return exp(-((x - ExpectedValue) * (x - ExpectedValue)) / (static_cast<T>(2) * StandardDeviation * StandardDeviation)) / (StandardDeviation * sqrt(static_cast<T>(6.28318530717958647692528676655900576)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T gauss
+	(
+		vec<2, T, Q> const& Coord,
+		vec<2, T, Q> const& ExpectedValue,
+		vec<2, T, Q> const& StandardDeviation
+	)
+	{
+		vec<2, T, Q> const Squared = ((Coord - ExpectedValue) * (Coord - ExpectedValue)) / (static_cast<T>(2) * StandardDeviation * StandardDeviation);
+		return exp(-(Squared.x + Squared.y));
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.hpp
new file mode 100644
index 000000000000..6f85bf482d9f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.hpp
@@ -0,0 +1,53 @@
+/// @ref gtx_gradient_paint
+/// @file glm/gtx/gradient_paint.hpp
+///
+/// @see core (dependence)
+/// @see gtx_optimum_pow (dependence)
+///
+/// @defgroup gtx_gradient_paint GLM_GTX_gradient_paint
+/// @ingroup gtx
+///
+/// Include <glm/gtx/gradient_paint.hpp> to use the features of this extension.
+///
+/// Functions that return the color of procedural gradient for specific coordinates.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtx/optimum_pow.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_gradient_paint is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_gradient_paint extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_gradient_paint
+	/// @{
+
+	/// Return a color from a radial gradient.
+	/// @see - gtx_gradient_paint
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T radialGradient(
+		vec<2, T, Q> const& Center,
+		T const& Radius,
+		vec<2, T, Q> const& Focal,
+		vec<2, T, Q> const& Position);
+
+	/// Return a color from a linear gradient.
+	/// @see - gtx_gradient_paint
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T linearGradient(
+		vec<2, T, Q> const& Point0,
+		vec<2, T, Q> const& Point1,
+		vec<2, T, Q> const& Position);
+
+	/// @}
+}// namespace glm
+
+#include "gradient_paint.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.inl
new file mode 100644
index 000000000000..4c495e62cbff
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/gradient_paint.inl
@@ -0,0 +1,36 @@
+/// @ref gtx_gradient_paint
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T radialGradient
+	(
+		vec<2, T, Q> const& Center,
+		T const& Radius,
+		vec<2, T, Q> const& Focal,
+		vec<2, T, Q> const& Position
+	)
+	{
+		vec<2, T, Q> F = Focal - Center;
+		vec<2, T, Q> D = Position - Focal;
+		T Radius2 = pow2(Radius);
+		T Fx2 = pow2(F.x);
+		T Fy2 = pow2(F.y);
+
+		T Numerator = (D.x * F.x + D.y * F.y) + sqrt(Radius2 * (pow2(D.x) + pow2(D.y)) - pow2(D.x * F.y - D.y * F.x));
+		T Denominator = Radius2 - (Fx2 + Fy2);
+		return Numerator / Denominator;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T linearGradient
+	(
+		vec<2, T, Q> const& Point0,
+		vec<2, T, Q> const& Point1,
+		vec<2, T, Q> const& Position
+	)
+	{
+		vec<2, T, Q> Dist = Point1 - Point0;
+		return (Dist.x * (Position.x - Point0.x) + Dist.y * (Position.y - Point0.y)) / glm::dot(Dist, Dist);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.hpp
new file mode 100644
index 000000000000..e59e3e2ac273
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.hpp
@@ -0,0 +1,50 @@
+/// @ref gtx_handed_coordinate_space
+/// @file glm/gtx/handed_coordinate_space.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_handed_coordinate_space GLM_GTX_handed_coordinate_space
+/// @ingroup gtx
+///
+/// Include <glm/gtx/handed_coordinate_space.hpp> to use the features of this extension.
+///
+/// To know if a set of three basis vectors defines a right or left-handed coordinate system.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_handed_coordinate_space is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_handed_coordinate_space extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_handed_coordinate_space
+	/// @{
+
+	//! Return if a trihedron right handed or not.
+	//! From GLM_GTX_handed_coordinate_space extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool rightHanded(
+		vec<3, T, Q> const& tangent,
+		vec<3, T, Q> const& binormal,
+		vec<3, T, Q> const& normal);
+
+	//! Return if a trihedron left handed or not.
+	//! From GLM_GTX_handed_coordinate_space extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool leftHanded(
+		vec<3, T, Q> const& tangent,
+		vec<3, T, Q> const& binormal,
+		vec<3, T, Q> const& normal);
+
+	/// @}
+}// namespace glm
+
+#include "handed_coordinate_space.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.inl
new file mode 100644
index 000000000000..e43c17bd3120
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/handed_coordinate_space.inl
@@ -0,0 +1,26 @@
+/// @ref gtx_handed_coordinate_space
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool rightHanded
+	(
+		vec<3, T, Q> const& tangent,
+		vec<3, T, Q> const& binormal,
+		vec<3, T, Q> const& normal
+	)
+	{
+		return dot(cross(normal, tangent), binormal) > T(0);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool leftHanded
+	(
+		vec<3, T, Q> const& tangent,
+		vec<3, T, Q> const& binormal,
+		vec<3, T, Q> const& normal
+	)
+	{
+		return dot(cross(normal, tangent), binormal) < T(0);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/hash.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/hash.hpp
new file mode 100644
index 000000000000..ef89290b439f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/hash.hpp
@@ -0,0 +1,146 @@
+/// @ref gtx_hash
+/// @file glm/gtx/hash.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_hash GLM_GTX_hash
+/// @ingroup gtx
+///
+/// Include <glm/gtx/hash.hpp> to use the features of this extension.
+///
+/// Add std::hash support for glm types
+
+#pragma once
+
+#if defined(GLM_FORCE_MESSAGES) && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_hash is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_hash extension included")
+#	endif
+#endif
+
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../gtc/vec1.hpp"
+
+#include "../gtc/quaternion.hpp"
+#include "../gtx/dual_quaternion.hpp"
+
+#include "../mat2x2.hpp"
+#include "../mat2x3.hpp"
+#include "../mat2x4.hpp"
+
+#include "../mat3x2.hpp"
+#include "../mat3x3.hpp"
+#include "../mat3x4.hpp"
+
+#include "../mat4x2.hpp"
+#include "../mat4x3.hpp"
+#include "../mat4x4.hpp"
+
+#if __cplusplus < 201103L
+#pragma message("GLM_GTX_hash requires C++11 standard library support")
+#endif
+
+#if GLM_LANG & GLM_LANG_CXX11
+#define GLM_GTX_hash 1
+#include <functional>
+
+namespace std
+{
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::vec<1, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::vec<1, T, Q> const& v) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::vec<2, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::vec<2, T, Q> const& v) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::vec<3, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::vec<3, T, Q> const& v) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::vec<4, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::vec<4, T, Q> const& v) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::qua<T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::qua<T, Q> const& q) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::tdualquat<T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::tdualquat<T,Q> const& q) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<2, 2, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<2, 2, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<2, 3, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<2, 3, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<2, 4, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<2, 4, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<3, 2, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<3, 2, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<3, 3, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<3, 3, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<3, 4, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<3, 4, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<4, 2, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<4, 2, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<4, 3, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<4, 3, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+
+	template<typename T, glm::qualifier Q>
+	struct hash<glm::mat<4, 4, T, Q> >
+	{
+		GLM_FUNC_DECL size_t operator()(glm::mat<4, 4, T,Q> const& m) const GLM_NOEXCEPT;
+	};
+} // namespace std
+
+#include "hash.inl"
+
+#endif //GLM_LANG & GLM_LANG_CXX11
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/hash.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/hash.inl
new file mode 100644
index 000000000000..bcadfe53e558
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/hash.inl
@@ -0,0 +1,175 @@
+/// @ref gtx_hash
+
+namespace glm {
+namespace detail
+{
+	GLM_INLINE void hash_combine(size_t &seed, size_t hash)
+	{
+		hash += 0x9e3779b9 + (seed << 6) + (seed >> 2);
+		seed ^= hash;
+	}
+}}
+
+namespace std
+{
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::vec<1, T, Q> >::operator()(glm::vec<1, T, Q> const& v) const GLM_NOEXCEPT
+	{
+		hash<T> hasher;
+		return hasher(v.x);
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::vec<2, T, Q> >::operator()(glm::vec<2, T, Q> const& v) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<T> hasher;
+		glm::detail::hash_combine(seed, hasher(v.x));
+		glm::detail::hash_combine(seed, hasher(v.y));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::vec<3, T, Q> >::operator()(glm::vec<3, T, Q> const& v) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<T> hasher;
+		glm::detail::hash_combine(seed, hasher(v.x));
+		glm::detail::hash_combine(seed, hasher(v.y));
+		glm::detail::hash_combine(seed, hasher(v.z));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::vec<4, T, Q> >::operator()(glm::vec<4, T, Q> const& v) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<T> hasher;
+		glm::detail::hash_combine(seed, hasher(v.x));
+		glm::detail::hash_combine(seed, hasher(v.y));
+		glm::detail::hash_combine(seed, hasher(v.z));
+		glm::detail::hash_combine(seed, hasher(v.w));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::qua<T, Q> >::operator()(glm::qua<T,Q> const& q) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<T> hasher;
+		glm::detail::hash_combine(seed, hasher(q.x));
+		glm::detail::hash_combine(seed, hasher(q.y));
+		glm::detail::hash_combine(seed, hasher(q.z));
+		glm::detail::hash_combine(seed, hasher(q.w));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::tdualquat<T, Q> >::operator()(glm::tdualquat<T, Q> const& q) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::qua<T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(q.real));
+		glm::detail::hash_combine(seed, hasher(q.dual));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<2, 2, T, Q> >::operator()(glm::mat<2, 2, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<2, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<2, 3, T, Q> >::operator()(glm::mat<2, 3, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<3, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<2, 4, T, Q> >::operator()(glm::mat<2, 4, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<4, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<3, 2, T, Q> >::operator()(glm::mat<3, 2, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<2, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		glm::detail::hash_combine(seed, hasher(m[2]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<3, 3, T, Q> >::operator()(glm::mat<3, 3, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<3, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		glm::detail::hash_combine(seed, hasher(m[2]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<3, 4, T, Q> >::operator()(glm::mat<3, 4, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<4, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		glm::detail::hash_combine(seed, hasher(m[2]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<4, 2, T,Q> >::operator()(glm::mat<4, 2, T,Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<2, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		glm::detail::hash_combine(seed, hasher(m[2]));
+		glm::detail::hash_combine(seed, hasher(m[3]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<4, 3, T,Q> >::operator()(glm::mat<4, 3, T,Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<3, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		glm::detail::hash_combine(seed, hasher(m[2]));
+		glm::detail::hash_combine(seed, hasher(m[3]));
+		return seed;
+	}
+
+	template<typename T, glm::qualifier Q>
+	GLM_FUNC_QUALIFIER size_t hash<glm::mat<4, 4, T,Q> >::operator()(glm::mat<4, 4, T, Q> const& m) const GLM_NOEXCEPT
+	{
+		size_t seed = 0;
+		hash<glm::vec<4, T, Q> > hasher;
+		glm::detail::hash_combine(seed, hasher(m[0]));
+		glm::detail::hash_combine(seed, hasher(m[1]));
+		glm::detail::hash_combine(seed, hasher(m[2]));
+		glm::detail::hash_combine(seed, hasher(m[3]));
+		return seed;
+	}
+}
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/integer.hpp
new file mode 100644
index 000000000000..d0b4c61a3fd4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/integer.hpp
@@ -0,0 +1,76 @@
+/// @ref gtx_integer
+/// @file glm/gtx/integer.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_integer GLM_GTX_integer
+/// @ingroup gtx
+///
+/// Include <glm/gtx/integer.hpp> to use the features of this extension.
+///
+/// Add support for integer for core functions
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/integer.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_integer is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_integer extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_integer
+	/// @{
+
+	//! Returns x raised to the y power.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL int pow(int x, uint y);
+
+	//! Returns the positive square root of x.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL int sqrt(int x);
+
+	//! Returns the floor log2 of x.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL unsigned int floor_log2(unsigned int x);
+
+	//! Modulus. Returns x - y * floor(x / y) for each component in x using the floating point value y.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL int mod(int x, int y);
+
+	//! Return the factorial value of a number (!12 max, integer only)
+	//! From GLM_GTX_integer extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType factorial(genType const& x);
+
+	//! 32bit signed integer.
+	//! From GLM_GTX_integer extension.
+	typedef signed int					sint;
+
+	//! Returns x raised to the y power.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL uint pow(uint x, uint y);
+
+	//! Returns the positive square root of x.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL uint sqrt(uint x);
+
+	//! Modulus. Returns x - y * floor(x / y) for each component in x using the floating point value y.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL uint mod(uint x, uint y);
+
+	//! Returns the number of leading zeros.
+	//! From GLM_GTX_integer extension.
+	GLM_FUNC_DECL uint nlz(uint x);
+
+	/// @}
+}//namespace glm
+
+#include "integer.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/integer.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/integer.inl
new file mode 100644
index 000000000000..956366b250f8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/integer.inl
@@ -0,0 +1,185 @@
+/// @ref gtx_integer
+
+namespace glm
+{
+	// pow
+	GLM_FUNC_QUALIFIER int pow(int x, uint y)
+	{
+		if(y == 0)
+			return x >= 0 ? 1 : -1;
+
+		int result = x;
+		for(uint i = 1; i < y; ++i)
+			result *= x;
+		return result;
+	}
+
+	// sqrt: From Christopher J. Musial, An integer square root, Graphics Gems, 1990, page 387
+	GLM_FUNC_QUALIFIER int sqrt(int x)
+	{
+		if(x <= 1) return x;
+
+		int NextTrial = x >> 1;
+		int CurrentAnswer;
+
+		do
+		{
+			CurrentAnswer = NextTrial;
+			NextTrial = (NextTrial + x / NextTrial) >> 1;
+		} while(NextTrial < CurrentAnswer);
+
+		return CurrentAnswer;
+	}
+
+// Henry Gordon Dietz: http://aggregate.org/MAGIC/
+namespace detail
+{
+	GLM_FUNC_QUALIFIER unsigned int ones32(unsigned int x)
+	{
+		/* 32-bit recursive reduction using SWAR...
+		but first step is mapping 2-bit values
+		into sum of 2 1-bit values in sneaky way
+		*/
+		x -= ((x >> 1) & 0x55555555);
+		x = (((x >> 2) & 0x33333333) + (x & 0x33333333));
+		x = (((x >> 4) + x) & 0x0f0f0f0f);
+		x += (x >> 8);
+		x += (x >> 16);
+		return(x & 0x0000003f);
+	}
+}//namespace detail
+
+	// Henry Gordon Dietz: http://aggregate.org/MAGIC/
+/*
+	GLM_FUNC_QUALIFIER unsigned int floor_log2(unsigned int x)
+	{
+		x |= (x >> 1);
+		x |= (x >> 2);
+		x |= (x >> 4);
+		x |= (x >> 8);
+		x |= (x >> 16);
+
+		return _detail::ones32(x) >> 1;
+	}
+*/
+	// mod
+	GLM_FUNC_QUALIFIER int mod(int x, int y)
+	{
+		return ((x % y) + y) % y;
+	}
+
+	// factorial (!12 max, integer only)
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType factorial(genType const& x)
+	{
+		genType Temp = x;
+		genType Result;
+		for(Result = 1; Temp > 1; --Temp)
+			Result *= Temp;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> factorial(
+		vec<2, T, Q> const& x)
+	{
+		return vec<2, T, Q>(
+			factorial(x.x),
+			factorial(x.y));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> factorial(
+		vec<3, T, Q> const& x)
+	{
+		return vec<3, T, Q>(
+			factorial(x.x),
+			factorial(x.y),
+			factorial(x.z));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> factorial(
+		vec<4, T, Q> const& x)
+	{
+		return vec<4, T, Q>(
+			factorial(x.x),
+			factorial(x.y),
+			factorial(x.z),
+			factorial(x.w));
+	}
+
+	GLM_FUNC_QUALIFIER uint pow(uint x, uint y)
+	{
+		if (y == 0)
+			return 1u;
+
+		uint result = x;
+		for(uint i = 1; i < y; ++i)
+			result *= x;
+		return result;
+	}
+
+	GLM_FUNC_QUALIFIER uint sqrt(uint x)
+	{
+		if(x <= 1) return x;
+
+		uint NextTrial = x >> 1;
+		uint CurrentAnswer;
+
+		do
+		{
+			CurrentAnswer = NextTrial;
+			NextTrial = (NextTrial + x / NextTrial) >> 1;
+		} while(NextTrial < CurrentAnswer);
+
+		return CurrentAnswer;
+	}
+
+	GLM_FUNC_QUALIFIER uint mod(uint x, uint y)
+	{
+		return x - y * (x / y);
+	}
+
+#if(GLM_COMPILER & (GLM_COMPILER_VC | GLM_COMPILER_GCC))
+
+	GLM_FUNC_QUALIFIER unsigned int nlz(unsigned int x)
+	{
+		return 31u - findMSB(x);
+	}
+
+#else
+
+	// Hackers Delight: http://www.hackersdelight.org/HDcode/nlz.c.txt
+	GLM_FUNC_QUALIFIER unsigned int nlz(unsigned int x)
+	{
+		int y, m, n;
+
+		y = -int(x >> 16);      // If left half of x is 0,
+		m = (y >> 16) & 16;  // set n = 16.  If left half
+		n = 16 - m;          // is nonzero, set n = 0 and
+		x = x >> m;          // shift x right 16.
+							// Now x is of the form 0000xxxx.
+		y = x - 0x100;       // If positions 8-15 are 0,
+		m = (y >> 16) & 8;   // add 8 to n and shift x left 8.
+		n = n + m;
+		x = x << m;
+
+		y = x - 0x1000;      // If positions 12-15 are 0,
+		m = (y >> 16) & 4;   // add 4 to n and shift x left 4.
+		n = n + m;
+		x = x << m;
+
+		y = x - 0x4000;      // If positions 14-15 are 0,
+		m = (y >> 16) & 2;   // add 2 to n and shift x left 2.
+		n = n + m;
+		x = x << m;
+
+		y = x >> 14;         // Set y = 0, 1, 2, or 3.
+		m = y & ~(y >> 1);   // Set m = 0, 1, 2, or 2 resp.
+		return unsigned(n + 2 - m);
+	}
+
+#endif//(GLM_COMPILER)
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.hpp
new file mode 100644
index 000000000000..f5c0621deb1a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.hpp
@@ -0,0 +1,92 @@
+/// @ref gtx_intersect
+/// @file glm/gtx/intersect.hpp
+///
+/// @see core (dependence)
+/// @see gtx_closest_point (dependence)
+///
+/// @defgroup gtx_intersect GLM_GTX_intersect
+/// @ingroup gtx
+///
+/// Include <glm/gtx/intersect.hpp> to use the features of this extension.
+///
+/// Add intersection functions
+
+#pragma once
+
+// Dependency:
+#include <cfloat>
+#include <limits>
+#include "../glm.hpp"
+#include "../geometric.hpp"
+#include "../gtx/closest_point.hpp"
+#include "../gtx/vector_query.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_closest_point is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_closest_point extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_intersect
+	/// @{
+
+	//! Compute the intersection of a ray and a plane.
+	//! Ray direction and plane normal must be unit length.
+	//! From GLM_GTX_intersect extension.
+	template<typename genType>
+	GLM_FUNC_DECL bool intersectRayPlane(
+		genType const& orig, genType const& dir,
+		genType const& planeOrig, genType const& planeNormal,
+		typename genType::value_type & intersectionDistance);
+
+	//! Compute the intersection of a ray and a triangle.
+	/// Based om Tomas Möller implementation http://fileadmin.cs.lth.se/cs/Personal/Tomas_Akenine-Moller/raytri/
+	//! From GLM_GTX_intersect extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool intersectRayTriangle(
+		vec<3, T, Q> const& orig, vec<3, T, Q> const& dir,
+		vec<3, T, Q> const& v0, vec<3, T, Q> const& v1, vec<3, T, Q> const& v2,
+		vec<2, T, Q>& baryPosition, T& distance);
+
+	//! Compute the intersection of a line and a triangle.
+	//! From GLM_GTX_intersect extension.
+	template<typename genType>
+	GLM_FUNC_DECL bool intersectLineTriangle(
+		genType const& orig, genType const& dir,
+		genType const& vert0, genType const& vert1, genType const& vert2,
+		genType & position);
+
+	//! Compute the intersection distance of a ray and a sphere.
+	//! The ray direction vector is unit length.
+	//! From GLM_GTX_intersect extension.
+	template<typename genType>
+	GLM_FUNC_DECL bool intersectRaySphere(
+		genType const& rayStarting, genType const& rayNormalizedDirection,
+		genType const& sphereCenter, typename genType::value_type const sphereRadiusSquared,
+		typename genType::value_type & intersectionDistance);
+
+	//! Compute the intersection of a ray and a sphere.
+	//! From GLM_GTX_intersect extension.
+	template<typename genType>
+	GLM_FUNC_DECL bool intersectRaySphere(
+		genType const& rayStarting, genType const& rayNormalizedDirection,
+		genType const& sphereCenter, const typename genType::value_type sphereRadius,
+		genType & intersectionPosition, genType & intersectionNormal);
+
+	//! Compute the intersection of a line and a sphere.
+	//! From GLM_GTX_intersect extension
+	template<typename genType>
+	GLM_FUNC_DECL bool intersectLineSphere(
+		genType const& point0, genType const& point1,
+		genType const& sphereCenter, typename genType::value_type sphereRadius,
+		genType & intersectionPosition1, genType & intersectionNormal1,
+		genType & intersectionPosition2 = genType(), genType & intersectionNormal2 = genType());
+
+	/// @}
+}//namespace glm
+
+#include "intersect.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.inl
new file mode 100644
index 000000000000..925a903d4a02
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/intersect.inl
@@ -0,0 +1,200 @@
+/// @ref gtx_intersect
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER bool intersectRayPlane
+	(
+		genType const& orig, genType const& dir,
+		genType const& planeOrig, genType const& planeNormal,
+		typename genType::value_type & intersectionDistance
+	)
+	{
+		typename genType::value_type d = glm::dot(dir, planeNormal);
+		typename genType::value_type Epsilon = std::numeric_limits<typename genType::value_type>::epsilon();
+
+		if(glm::abs(d) > Epsilon)  // if dir and planeNormal are not perpendicular
+		{
+			typename genType::value_type const tmp_intersectionDistance = 	glm::dot(planeOrig - orig, planeNormal) / d;
+			if (tmp_intersectionDistance > static_cast<typename genType::value_type>(0)) { // allow only intersections
+				intersectionDistance = tmp_intersectionDistance;
+				return true;
+			}
+		}
+
+		return false;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool intersectRayTriangle
+	(
+		vec<3, T, Q> const& orig, vec<3, T, Q> const& dir,
+		vec<3, T, Q> const& vert0, vec<3, T, Q> const& vert1, vec<3, T, Q> const& vert2,
+		vec<2, T, Q>& baryPosition, T& distance
+	)
+	{
+		// find vectors for two edges sharing vert0
+		vec<3, T, Q> const edge1 = vert1 - vert0;
+		vec<3, T, Q> const edge2 = vert2 - vert0;
+
+		// begin calculating determinant - also used to calculate U parameter
+		vec<3, T, Q> const p = glm::cross(dir, edge2);
+
+		// if determinant is near zero, ray lies in plane of triangle
+		T const det = glm::dot(edge1, p);
+
+		vec<3, T, Q> Perpendicular(0);
+
+		if (det > static_cast<T>(0))
+		{
+			// calculate distance from vert0 to ray origin
+			vec<3, T, Q> const dist = orig - vert0;
+
+			// calculate U parameter and test bounds
+			baryPosition.x = glm::dot(dist, p);
+			if(baryPosition.x < static_cast<T>(0) || baryPosition.x > det)
+				return false;
+
+			// prepare to test V parameter
+			Perpendicular = glm::cross(dist, edge1);
+
+			// calculate V parameter and test bounds
+			baryPosition.y = glm::dot(dir, Perpendicular);
+			if((baryPosition.y < static_cast<T>(0)) || ((baryPosition.x + baryPosition.y) > det))
+				return false;
+		}
+		else if(det < static_cast<T>(0))
+		{
+			// calculate distance from vert0 to ray origin
+			vec<3, T, Q> const dist = orig - vert0;
+
+			// calculate U parameter and test bounds
+			baryPosition.x = glm::dot(dist, p);
+			if((baryPosition.x > static_cast<T>(0)) || (baryPosition.x < det))
+				return false;
+
+			// prepare to test V parameter
+			Perpendicular = glm::cross(dist, edge1);
+
+			// calculate V parameter and test bounds
+			baryPosition.y = glm::dot(dir, Perpendicular);
+			if((baryPosition.y > static_cast<T>(0)) || (baryPosition.x + baryPosition.y < det))
+				return false;
+		}
+		else
+			return false; // ray is parallel to the plane of the triangle
+
+		T inv_det = static_cast<T>(1) / det;
+
+		// calculate distance, ray intersects triangle
+		distance = glm::dot(edge2, Perpendicular) * inv_det;
+		baryPosition *= inv_det;
+
+		return true;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER bool intersectLineTriangle
+	(
+		genType const& orig, genType const& dir,
+		genType const& vert0, genType const& vert1, genType const& vert2,
+		genType & position
+	)
+	{
+		typename genType::value_type Epsilon = std::numeric_limits<typename genType::value_type>::epsilon();
+
+		genType edge1 = vert1 - vert0;
+		genType edge2 = vert2 - vert0;
+
+		genType Perpendicular = cross(dir, edge2);
+
+		typename genType::value_type det = dot(edge1, Perpendicular);
+
+		if (det > -Epsilon && det < Epsilon)
+			return false;
+		typename genType::value_type inv_det = typename genType::value_type(1) / det;
+
+		genType Tangent = orig - vert0;
+
+		position.y = dot(Tangent, Perpendicular) * inv_det;
+		if (position.y < typename genType::value_type(0) || position.y > typename genType::value_type(1))
+			return false;
+
+		genType Cotangent = cross(Tangent, edge1);
+
+		position.z = dot(dir, Cotangent) * inv_det;
+		if (position.z < typename genType::value_type(0) || position.y + position.z > typename genType::value_type(1))
+			return false;
+
+		position.x = dot(edge2, Cotangent) * inv_det;
+
+		return true;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER bool intersectRaySphere
+	(
+		genType const& rayStarting, genType const& rayNormalizedDirection,
+		genType const& sphereCenter, const typename genType::value_type sphereRadiusSquared,
+		typename genType::value_type & intersectionDistance
+	)
+	{
+		typename genType::value_type Epsilon = std::numeric_limits<typename genType::value_type>::epsilon();
+		genType diff = sphereCenter - rayStarting;
+		typename genType::value_type t0 = dot(diff, rayNormalizedDirection);
+		typename genType::value_type dSquared = dot(diff, diff) - t0 * t0;
+		if( dSquared > sphereRadiusSquared )
+		{
+			return false;
+		}
+		typename genType::value_type t1 = sqrt( sphereRadiusSquared - dSquared );
+		intersectionDistance = t0 > t1 + Epsilon ? t0 - t1 : t0 + t1;
+		return intersectionDistance > Epsilon;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER bool intersectRaySphere
+	(
+		genType const& rayStarting, genType const& rayNormalizedDirection,
+		genType const& sphereCenter, const typename genType::value_type sphereRadius,
+		genType & intersectionPosition, genType & intersectionNormal
+	)
+	{
+		typename genType::value_type distance;
+		if( intersectRaySphere( rayStarting, rayNormalizedDirection, sphereCenter, sphereRadius * sphereRadius, distance ) )
+		{
+			intersectionPosition = rayStarting + rayNormalizedDirection * distance;
+			intersectionNormal = (intersectionPosition - sphereCenter) / sphereRadius;
+			return true;
+		}
+		return false;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER bool intersectLineSphere
+	(
+		genType const& point0, genType const& point1,
+		genType const& sphereCenter, typename genType::value_type sphereRadius,
+		genType & intersectionPoint1, genType & intersectionNormal1,
+		genType & intersectionPoint2, genType & intersectionNormal2
+	)
+	{
+		typename genType::value_type Epsilon = std::numeric_limits<typename genType::value_type>::epsilon();
+		genType dir = normalize(point1 - point0);
+		genType diff = sphereCenter - point0;
+		typename genType::value_type t0 = dot(diff, dir);
+		typename genType::value_type dSquared = dot(diff, diff) - t0 * t0;
+		if( dSquared > sphereRadius * sphereRadius )
+		{
+			return false;
+		}
+		typename genType::value_type t1 = sqrt( sphereRadius * sphereRadius - dSquared );
+		if( t0 < t1 + Epsilon )
+			t1 = -t1;
+		intersectionPoint1 = point0 + dir * (t0 - t1);
+		intersectionNormal1 = (intersectionPoint1 - sphereCenter) / sphereRadius;
+		intersectionPoint2 = point0 + dir * (t0 + t1);
+		intersectionNormal2 = (intersectionPoint2 - sphereCenter) / sphereRadius;
+		return true;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/io.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/io.hpp
new file mode 100644
index 000000000000..68b5499bfbe7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/io.hpp
@@ -0,0 +1,201 @@
+/// @ref gtx_io
+/// @file glm/gtx/io.hpp
+/// @author Jan P Springer (regnirpsj@gmail.com)
+///
+/// @see core (dependence)
+/// @see gtc_matrix_access (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtx_io GLM_GTX_io
+/// @ingroup gtx
+///
+/// Include <glm/gtx/io.hpp> to use the features of this extension.
+///
+/// std::[w]ostream support for glm types
+///
+/// std::[w]ostream support for glm types + qualifier/width/etc. manipulators
+/// based on howard hinnant's std::chrono io proposal
+/// [http://home.roadrunner.com/~hinnant/bloomington/chrono_io.html]
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtx/quaternion.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_io is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_io extension included")
+#	endif
+#endif
+
+#include <iosfwd>  // std::basic_ostream<> (fwd)
+#include <locale>  // std::locale, std::locale::facet, std::locale::id
+#include <utility> // std::pair<>
+
+namespace glm
+{
+	/// @addtogroup gtx_io
+	/// @{
+
+	namespace io
+	{
+		enum order_type { column_major, row_major};
+
+		template<typename CTy>
+		class format_punct : public std::locale::facet
+		{
+			typedef CTy char_type;
+
+		public:
+
+			static std::locale::id id;
+
+			bool       formatted;
+			unsigned   precision;
+			unsigned   width;
+			char_type  separator;
+			char_type  delim_left;
+			char_type  delim_right;
+			char_type  space;
+			char_type  newline;
+			order_type order;
+
+			GLM_FUNC_DECL explicit format_punct(size_t a = 0);
+			GLM_FUNC_DECL explicit format_punct(format_punct const&);
+		};
+
+		template<typename CTy, typename CTr = std::char_traits<CTy> >
+		class basic_state_saver {
+
+		public:
+
+			GLM_FUNC_DECL explicit basic_state_saver(std::basic_ios<CTy,CTr>&);
+			GLM_FUNC_DECL ~basic_state_saver();
+
+		private:
+
+			typedef ::std::basic_ios<CTy,CTr>      state_type;
+			typedef typename state_type::char_type char_type;
+			typedef ::std::ios_base::fmtflags      flags_type;
+			typedef ::std::streamsize              streamsize_type;
+			typedef ::std::locale const            locale_type;
+
+			state_type&     state_;
+			flags_type      flags_;
+			streamsize_type precision_;
+			streamsize_type width_;
+			char_type       fill_;
+			locale_type     locale_;
+
+			GLM_FUNC_DECL basic_state_saver& operator=(basic_state_saver const&);
+		};
+
+		typedef basic_state_saver<char>     state_saver;
+		typedef basic_state_saver<wchar_t> wstate_saver;
+
+		template<typename CTy, typename CTr = std::char_traits<CTy> >
+		class basic_format_saver
+		{
+		public:
+
+			GLM_FUNC_DECL explicit basic_format_saver(std::basic_ios<CTy,CTr>&);
+			GLM_FUNC_DECL ~basic_format_saver();
+
+		private:
+
+			basic_state_saver<CTy> const bss_;
+
+			GLM_FUNC_DECL basic_format_saver& operator=(basic_format_saver const&);
+		};
+
+		typedef basic_format_saver<char>     format_saver;
+		typedef basic_format_saver<wchar_t> wformat_saver;
+
+		struct precision
+		{
+			unsigned value;
+
+			GLM_FUNC_DECL explicit precision(unsigned);
+		};
+
+		struct width
+		{
+			unsigned value;
+
+			GLM_FUNC_DECL explicit width(unsigned);
+		};
+
+		template<typename CTy>
+		struct delimeter
+		{
+			CTy value[3];
+
+			GLM_FUNC_DECL explicit delimeter(CTy /* left */, CTy /* right */, CTy /* separator */ = ',');
+		};
+
+		struct order
+		{
+			order_type value;
+
+			GLM_FUNC_DECL explicit order(order_type);
+		};
+
+		// functions, inlined (inline)
+
+		template<typename FTy, typename CTy, typename CTr>
+		FTy const& get_facet(std::basic_ios<CTy,CTr>&);
+		template<typename FTy, typename CTy, typename CTr>
+		std::basic_ios<CTy,CTr>& formatted(std::basic_ios<CTy,CTr>&);
+		template<typename FTy, typename CTy, typename CTr>
+		std::basic_ios<CTy,CTr>& unformatted(std::basic_ios<CTy,CTr>&);
+
+		template<typename CTy, typename CTr>
+		std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>&, precision const&);
+		template<typename CTy, typename CTr>
+		std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>&, width const&);
+		template<typename CTy, typename CTr>
+		std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>&, delimeter<CTy> const&);
+		template<typename CTy, typename CTr>
+		std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>&, order const&);
+	}//namespace io
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, qua<T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, vec<1, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, vec<2, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, vec<3, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, vec<4, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<2, 2, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<2, 3, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<2, 4, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<3, 2, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<3, 3, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<3, 4, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<4, 2, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<4, 3, T, Q> const&);
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>&, mat<4, 4, T, Q> const&);
+
+  template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_DECL std::basic_ostream<CTy,CTr> & operator<<(std::basic_ostream<CTy,CTr> &,
+                                                         std::pair<mat<4, 4, T, Q> const, mat<4, 4, T, Q> const> const&);
+
+	/// @}
+}//namespace glm
+
+#include "io.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/io.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/io.inl
new file mode 100644
index 000000000000..a3a1bb6c26b4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/io.inl
@@ -0,0 +1,440 @@
+/// @ref gtx_io
+/// @author Jan P Springer (regnirpsj@gmail.com)
+
+#include <iomanip>                  // std::fixed, std::setfill<>, std::setprecision, std::right, std::setw
+#include <ostream>                  // std::basic_ostream<>
+#include "../gtc/matrix_access.hpp" // glm::col, glm::row
+#include "../gtx/type_trait.hpp"    // glm::type<>
+
+namespace glm{
+namespace io
+{
+	template<typename CTy>
+	GLM_FUNC_QUALIFIER format_punct<CTy>::format_punct(size_t a)
+		: std::locale::facet(a)
+		, formatted(true)
+		, precision(3)
+		, width(1 + 4 + 1 + precision)
+		, separator(',')
+		, delim_left('[')
+		, delim_right(']')
+		, space(' ')
+		, newline('\n')
+		, order(column_major)
+	{}
+
+	template<typename CTy>
+	GLM_FUNC_QUALIFIER format_punct<CTy>::format_punct(format_punct const& a)
+		: std::locale::facet(0)
+		, formatted(a.formatted)
+		, precision(a.precision)
+		, width(a.width)
+		, separator(a.separator)
+		, delim_left(a.delim_left)
+		, delim_right(a.delim_right)
+		, space(a.space)
+		, newline(a.newline)
+		, order(a.order)
+	{}
+
+	template<typename CTy> std::locale::id format_punct<CTy>::id;
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER basic_state_saver<CTy, CTr>::basic_state_saver(std::basic_ios<CTy, CTr>& a)
+		: state_(a)
+		, flags_(a.flags())
+		, precision_(a.precision())
+		, width_(a.width())
+		, fill_(a.fill())
+		, locale_(a.getloc())
+	{}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER basic_state_saver<CTy, CTr>::~basic_state_saver()
+	{
+		state_.imbue(locale_);
+		state_.fill(fill_);
+		state_.width(width_);
+		state_.precision(precision_);
+		state_.flags(flags_);
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER basic_format_saver<CTy, CTr>::basic_format_saver(std::basic_ios<CTy, CTr>& a)
+		: bss_(a)
+	{
+		a.imbue(std::locale(a.getloc(), new format_punct<CTy>(get_facet<format_punct<CTy> >(a))));
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER
+	basic_format_saver<CTy, CTr>::~basic_format_saver()
+	{}
+
+	GLM_FUNC_QUALIFIER precision::precision(unsigned a)
+		: value(a)
+	{}
+
+	GLM_FUNC_QUALIFIER width::width(unsigned a)
+		: value(a)
+	{}
+
+	template<typename CTy>
+	GLM_FUNC_QUALIFIER delimeter<CTy>::delimeter(CTy a, CTy b, CTy c)
+		: value()
+	{
+		value[0] = a;
+		value[1] = b;
+		value[2] = c;
+	}
+
+	GLM_FUNC_QUALIFIER order::order(order_type a)
+		: value(a)
+	{}
+
+	template<typename FTy, typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER FTy const& get_facet(std::basic_ios<CTy, CTr>& ios)
+	{
+		if(!std::has_facet<FTy>(ios.getloc()))
+			ios.imbue(std::locale(ios.getloc(), new FTy));
+
+		return std::use_facet<FTy>(ios.getloc());
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER std::basic_ios<CTy, CTr>& formatted(std::basic_ios<CTy, CTr>& ios)
+	{
+		const_cast<format_punct<CTy>&>(get_facet<format_punct<CTy> >(ios)).formatted = true;
+		return ios;
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER std::basic_ios<CTy, CTr>& unformatted(std::basic_ios<CTy, CTr>& ios)
+	{
+		const_cast<format_punct<CTy>&>(get_facet<format_punct<CTy> >(ios)).formatted = false;
+		return ios;
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>& os, precision const& a)
+	{
+		const_cast<format_punct<CTy>&>(get_facet<format_punct<CTy> >(os)).precision = a.value;
+		return os;
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>& os, width const& a)
+	{
+		const_cast<format_punct<CTy>&>(get_facet<format_punct<CTy> >(os)).width = a.value;
+		return os;
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER  std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>& os, delimeter<CTy> const& a)
+	{
+		format_punct<CTy> & fmt(const_cast<format_punct<CTy>&>(get_facet<format_punct<CTy> >(os)));
+
+		fmt.delim_left  = a.value[0];
+		fmt.delim_right = a.value[1];
+		fmt.separator   = a.value[2];
+
+		return os;
+	}
+
+	template<typename CTy, typename CTr>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>& operator<<(std::basic_ostream<CTy, CTr>& os, order const& a)
+	{
+		const_cast<format_punct<CTy>&>(get_facet<format_punct<CTy> >(os)).order = a.value;
+		return os;
+	}
+} // namespace io
+
+namespace detail
+{
+	template<typename CTy, typename CTr, typename V>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>&
+	print_vector_on(std::basic_ostream<CTy, CTr>& os, V const& a)
+	{
+		typename std::basic_ostream<CTy, CTr>::sentry const cerberus(os);
+
+		if(cerberus)
+		{
+			io::format_punct<CTy> const& fmt(io::get_facet<io::format_punct<CTy> >(os));
+
+			length_t const& components(type<V>::components);
+
+			if(fmt.formatted)
+			{
+				io::basic_state_saver<CTy> const bss(os);
+
+				os << std::fixed << std::right << std::setprecision(fmt.precision) << std::setfill(fmt.space) << fmt.delim_left;
+
+				for(length_t i(0); i < components; ++i)
+				{
+					os << std::setw(fmt.width) << a[i];
+					if(components-1 != i)
+						os << fmt.separator;
+				}
+
+				os << fmt.delim_right;
+			}
+			else
+			{
+				for(length_t i(0); i < components; ++i)
+				{
+					os << a[i];
+
+					if(components-1 != i)
+						os << fmt.space;
+				}
+			}
+		}
+
+		return os;
+	}
+}//namespace detail
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, qua<T, Q> const& a)
+	{
+		return detail::print_vector_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, vec<1, T, Q> const& a)
+	{
+		return detail::print_vector_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, vec<2, T, Q> const& a)
+	{
+		return detail::print_vector_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, vec<3, T, Q> const& a)
+	{
+		return detail::print_vector_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, vec<4, T, Q> const& a)
+	{
+		return detail::print_vector_on(os, a);
+	}
+
+namespace detail
+{
+	template<typename CTy, typename CTr, template<length_t, length_t, typename, qualifier> class M, length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>& print_matrix_on(std::basic_ostream<CTy, CTr>& os, M<C, R, T, Q> const& a)
+	{
+		typename std::basic_ostream<CTy,CTr>::sentry const cerberus(os);
+
+		if(cerberus)
+		{
+			io::format_punct<CTy> const& fmt(io::get_facet<io::format_punct<CTy> >(os));
+
+			length_t const& cols(type<M<C, R, T, Q> >::cols);
+			length_t const& rows(type<M<C, R, T, Q> >::rows);
+
+			if(fmt.formatted)
+			{
+				os << fmt.newline << fmt.delim_left;
+
+				switch(fmt.order)
+				{
+					case io::column_major:
+					{
+						for(length_t i(0); i < rows; ++i)
+						{
+							if (0 != i)
+								os << fmt.space;
+
+							os << row(a, i);
+
+							if(rows-1 != i)
+								os << fmt.newline;
+						}
+					}
+					break;
+
+					case io::row_major:
+					{
+						for(length_t i(0); i < cols; ++i)
+						{
+							if(0 != i)
+								os << fmt.space;
+
+							os << column(a, i);
+
+							if(cols-1 != i)
+								os << fmt.newline;
+						}
+					}
+					break;
+				}
+
+				os << fmt.delim_right;
+			}
+			else
+			{
+				switch (fmt.order)
+				{
+					case io::column_major:
+					{
+						for(length_t i(0); i < cols; ++i)
+						{
+							os << column(a, i);
+
+							if(cols - 1 != i)
+								os << fmt.space;
+						}
+					}
+					break;
+
+					case io::row_major:
+					{
+						for (length_t i(0); i < rows; ++i)
+						{
+							os << row(a, i);
+
+							if (rows-1 != i)
+								os << fmt.space;
+						}
+					}
+					break;
+				}
+			}
+		}
+
+		return os;
+	}
+}//namespace detail
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, mat<2, 2, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, mat<2, 3, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, mat<2, 4, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, mat<3, 2, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr>& operator<<(std::basic_ostream<CTy,CTr>& os, mat<3, 3, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr> & operator<<(std::basic_ostream<CTy,CTr>& os, mat<3, 4, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr> & operator<<(std::basic_ostream<CTy,CTr>& os, mat<4, 2, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr> & operator<<(std::basic_ostream<CTy,CTr>& os, mat<4, 3, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy,CTr> & operator<<(std::basic_ostream<CTy,CTr>& os, mat<4, 4, T, Q> const& a)
+	{
+		return detail::print_matrix_on(os, a);
+	}
+
+namespace detail
+{
+	template<typename CTy, typename CTr, template<length_t, length_t, typename, qualifier> class M, length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>& print_matrix_pair_on(std::basic_ostream<CTy, CTr>& os, std::pair<M<C, R, T, Q> const, M<C, R, T, Q> const> const& a)
+	{
+		typename std::basic_ostream<CTy,CTr>::sentry const cerberus(os);
+
+		if(cerberus)
+		{
+			io::format_punct<CTy> const& fmt(io::get_facet<io::format_punct<CTy> >(os));
+			M<C, R, T, Q> const& ml(a.first);
+			M<C, R, T, Q> const& mr(a.second);
+			length_t const& cols(type<M<C, R, T, Q> >::cols);
+			length_t const& rows(type<M<C, R, T, Q> >::rows);
+
+			if(fmt.formatted)
+			{
+				os << fmt.newline << fmt.delim_left;
+
+				switch(fmt.order)
+				{
+					case io::column_major:
+					{
+						for(length_t i(0); i < rows; ++i)
+						{
+							if(0 != i)
+								os << fmt.space;
+
+							os << row(ml, i) << ((rows-1 != i) ? fmt.space : fmt.delim_right) << fmt.space << ((0 != i) ? fmt.space : fmt.delim_left) << row(mr, i);
+
+							if(rows-1 != i)
+								os << fmt.newline;
+						}
+					}
+					break;
+					case io::row_major:
+					{
+						for(length_t i(0); i < cols; ++i)
+						{
+							if(0 != i)
+								os << fmt.space;
+
+							os << column(ml, i) << ((cols-1 != i) ? fmt.space : fmt.delim_right) << fmt.space << ((0 != i) ? fmt.space : fmt.delim_left) << column(mr, i);
+
+							if(cols-1 != i)
+								os << fmt.newline;
+						}
+					}
+					break;
+				}
+
+				os << fmt.delim_right;
+			}
+			else
+			{
+				os << ml << fmt.space << mr;
+			}
+		}
+
+		return os;
+	}
+}//namespace detail
+
+	template<typename CTy, typename CTr, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER std::basic_ostream<CTy, CTr>& operator<<(
+		std::basic_ostream<CTy, CTr> & os,
+		std::pair<mat<4, 4, T, Q> const,
+		mat<4, 4, T, Q> const> const& a)
+	{
+		return detail::print_matrix_pair_on(os, a);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.hpp
new file mode 100644
index 000000000000..ba28c9d7bffc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.hpp
@@ -0,0 +1,48 @@
+/// @ref gtx_log_base
+/// @file glm/gtx/log_base.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_log_base GLM_GTX_log_base
+/// @ingroup gtx
+///
+/// Include <glm/gtx/log_base.hpp> to use the features of this extension.
+///
+/// Logarithm for any base. base can be a vector or a scalar.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_log_base is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_log_base extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_log_base
+	/// @{
+
+	/// Logarithm for any base.
+	/// From GLM_GTX_log_base.
+	template<typename genType>
+	GLM_FUNC_DECL genType log(
+		genType const& x,
+		genType const& base);
+
+	/// Logarithm for any base.
+	/// From GLM_GTX_log_base.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> sign(
+		vec<L, T, Q> const& x,
+		vec<L, T, Q> const& base);
+
+	/// @}
+}//namespace glm
+
+#include "log_base.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.inl
new file mode 100644
index 000000000000..4bbb8e895abb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/log_base.inl
@@ -0,0 +1,16 @@
+/// @ref gtx_log_base
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType log(genType const& x, genType const& base)
+	{
+		return glm::log(x) / glm::log(base);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, T, Q> log(vec<L, T, Q> const& x, vec<L, T, Q> const& base)
+	{
+		return glm::log(x) / glm::log(base);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.hpp
new file mode 100644
index 000000000000..1e585f9a4ffe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.hpp
@@ -0,0 +1,47 @@
+/// @ref gtx_matrix_cross_product
+/// @file glm/gtx/matrix_cross_product.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_matrix_cross_product GLM_GTX_matrix_cross_product
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_cross_product.hpp> to use the features of this extension.
+///
+/// Build cross product matrices
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_cross_product is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_cross_product extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_cross_product
+	/// @{
+
+	//! Build a cross product matrix.
+	//! From GLM_GTX_matrix_cross_product extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> matrixCross3(
+		vec<3, T, Q> const& x);
+
+	//! Build a cross product matrix.
+	//! From GLM_GTX_matrix_cross_product extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> matrixCross4(
+		vec<3, T, Q> const& x);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_cross_product.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.inl
new file mode 100644
index 000000000000..3a153977cf59
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_cross_product.inl
@@ -0,0 +1,37 @@
+/// @ref gtx_matrix_cross_product
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> matrixCross3
+	(
+		vec<3, T, Q> const& x
+	)
+	{
+		mat<3, 3, T, Q> Result(T(0));
+		Result[0][1] = x.z;
+		Result[1][0] = -x.z;
+		Result[0][2] = -x.y;
+		Result[2][0] = x.y;
+		Result[1][2] = x.x;
+		Result[2][1] = -x.x;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> matrixCross4
+	(
+		vec<3, T, Q> const& x
+	)
+	{
+		mat<4, 4, T, Q> Result(T(0));
+		Result[0][1] = x.z;
+		Result[1][0] = -x.z;
+		Result[0][2] = -x.y;
+		Result[2][0] = x.y;
+		Result[1][2] = x.x;
+		Result[2][1] = -x.x;
+		return Result;
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.hpp
new file mode 100644
index 000000000000..8ab38e6b36a5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.hpp
@@ -0,0 +1,52 @@
+/// @ref gtx_matrix_decompose
+/// @file glm/gtx/matrix_decompose.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_matrix_decompose GLM_GTX_matrix_decompose
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_decompose.hpp> to use the features of this extension.
+///
+/// Decomposes a model matrix to translations, rotation and scale components
+
+#pragma once
+
+// Dependencies
+#include "../mat4x4.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../geometric.hpp"
+#include "../gtc/quaternion.hpp"
+#include "../gtc/matrix_transform.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_decompose is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_decompose extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_decompose
+	/// @{
+
+	/// Decomposes a model matrix to translations, rotation and scale components
+	/// @see gtx_matrix_decompose
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool decompose(
+		mat<4, 4, T, Q> const& modelMatrix,
+		vec<3, T, Q> & scale, qua<T, Q> & orientation, vec<3, T, Q> & translation, vec<3, T, Q> & skew, vec<4, T, Q> & perspective);
+
+	// Recomposes a model matrix from a previously-decomposed matrix
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> recompose(
+		vec<3, T, Q> const& scale, qua<T, Q> const& orientation, vec<3, T, Q> const& translation,
+		vec<3, T, Q> const& skew, vec<4, T, Q> const& perspective);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_decompose.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.inl
new file mode 100644
index 000000000000..1b587e2a1b00
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_decompose.inl
@@ -0,0 +1,234 @@
+/// @ref gtx_matrix_decompose
+
+#include "../gtc/constants.hpp"
+#include "../gtc/epsilon.hpp"
+#include "../gtx/transform.hpp"
+
+namespace glm{
+namespace detail
+{
+	/// Make a linear combination of two vectors and return the result.
+	// result = (a * ascl) + (b * bscl)
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> combine(
+		vec<3, T, Q> const& a,
+		vec<3, T, Q> const& b,
+		T ascl, T bscl)
+	{
+		return (a * ascl) + (b * bscl);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> scale(vec<3, T, Q> const& v, T desiredLength)
+	{
+		return v * desiredLength / length(v);
+	}
+}//namespace detail
+
+	// Matrix decompose
+	// http://www.opensource.apple.com/source/WebCore/WebCore-514/platform/graphics/transforms/TransformationMatrix.cpp
+	// Decomposes the mode matrix to translations,rotation scale components
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool decompose(mat<4, 4, T, Q> const& ModelMatrix, vec<3, T, Q> & Scale, qua<T, Q> & Orientation, vec<3, T, Q> & Translation, vec<3, T, Q> & Skew, vec<4, T, Q> & Perspective)
+	{
+		mat<4, 4, T, Q> LocalMatrix(ModelMatrix);
+
+		// Normalize the matrix.
+		if(epsilonEqual(LocalMatrix[3][3], static_cast<T>(0), epsilon<T>()))
+			return false;
+
+		for(length_t i = 0; i < 4; ++i)
+		for(length_t j = 0; j < 4; ++j)
+			LocalMatrix[i][j] /= LocalMatrix[3][3];
+
+		// perspectiveMatrix is used to solve for perspective, but it also provides
+		// an easy way to test for singularity of the upper 3x3 component.
+		mat<4, 4, T, Q> PerspectiveMatrix(LocalMatrix);
+
+		for(length_t i = 0; i < 3; i++)
+			PerspectiveMatrix[i][3] = static_cast<T>(0);
+		PerspectiveMatrix[3][3] = static_cast<T>(1);
+
+		/// TODO: Fixme!
+		if(epsilonEqual(determinant(PerspectiveMatrix), static_cast<T>(0), epsilon<T>()))
+			return false;
+
+		// First, isolate perspective.  This is the messiest.
+		if(
+			epsilonNotEqual(LocalMatrix[0][3], static_cast<T>(0), epsilon<T>()) ||
+			epsilonNotEqual(LocalMatrix[1][3], static_cast<T>(0), epsilon<T>()) ||
+			epsilonNotEqual(LocalMatrix[2][3], static_cast<T>(0), epsilon<T>()))
+		{
+			// rightHandSide is the right hand side of the equation.
+			vec<4, T, Q> RightHandSide;
+			RightHandSide[0] = LocalMatrix[0][3];
+			RightHandSide[1] = LocalMatrix[1][3];
+			RightHandSide[2] = LocalMatrix[2][3];
+			RightHandSide[3] = LocalMatrix[3][3];
+
+			// Solve the equation by inverting PerspectiveMatrix and multiplying
+			// rightHandSide by the inverse.  (This is the easiest way, not
+			// necessarily the best.)
+			mat<4, 4, T, Q> InversePerspectiveMatrix = glm::inverse(PerspectiveMatrix);//   inverse(PerspectiveMatrix, inversePerspectiveMatrix);
+			mat<4, 4, T, Q> TransposedInversePerspectiveMatrix = glm::transpose(InversePerspectiveMatrix);//   transposeMatrix4(inversePerspectiveMatrix, transposedInversePerspectiveMatrix);
+
+			Perspective = TransposedInversePerspectiveMatrix * RightHandSide;
+			//  v4MulPointByMatrix(rightHandSide, transposedInversePerspectiveMatrix, perspectivePoint);
+
+			// Clear the perspective partition
+			LocalMatrix[0][3] = LocalMatrix[1][3] = LocalMatrix[2][3] = static_cast<T>(0);
+			LocalMatrix[3][3] = static_cast<T>(1);
+		}
+		else
+		{
+			// No perspective.
+			Perspective = vec<4, T, Q>(0, 0, 0, 1);
+		}
+
+		// Next take care of translation (easy).
+		Translation = vec<3, T, Q>(LocalMatrix[3]);
+		LocalMatrix[3] = vec<4, T, Q>(0, 0, 0, LocalMatrix[3].w);
+
+		vec<3, T, Q> Row[3], Pdum3;
+
+		// Now get scale and shear.
+		for(length_t i = 0; i < 3; ++i)
+		for(length_t j = 0; j < 3; ++j)
+			Row[i][j] = LocalMatrix[i][j];
+
+		// Compute X scale factor and normalize first row.
+		Scale.x = length(Row[0]);// v3Length(Row[0]);
+
+		Row[0] = detail::scale(Row[0], static_cast<T>(1));
+
+		// Compute XY shear factor and make 2nd row orthogonal to 1st.
+		Skew.z = dot(Row[0], Row[1]);
+		Row[1] = detail::combine(Row[1], Row[0], static_cast<T>(1), -Skew.z);
+
+		// Now, compute Y scale and normalize 2nd row.
+		Scale.y = length(Row[1]);
+		Row[1] = detail::scale(Row[1], static_cast<T>(1));
+		Skew.z /= Scale.y;
+
+		// Compute XZ and YZ shears, orthogonalize 3rd row.
+		Skew.y = glm::dot(Row[0], Row[2]);
+		Row[2] = detail::combine(Row[2], Row[0], static_cast<T>(1), -Skew.y);
+		Skew.x = glm::dot(Row[1], Row[2]);
+		Row[2] = detail::combine(Row[2], Row[1], static_cast<T>(1), -Skew.x);
+
+		// Next, get Z scale and normalize 3rd row.
+		Scale.z = length(Row[2]);
+		Row[2] = detail::scale(Row[2], static_cast<T>(1));
+		Skew.y /= Scale.z;
+		Skew.x /= Scale.z;
+
+		// At this point, the matrix (in rows[]) is orthonormal.
+		// Check for a coordinate system flip.  If the determinant
+		// is -1, then negate the matrix and the scaling factors.
+		Pdum3 = cross(Row[1], Row[2]); // v3Cross(row[1], row[2], Pdum3);
+		if(dot(Row[0], Pdum3) < 0)
+		{
+			for(length_t i = 0; i < 3; i++)
+			{
+				Scale[i] *= static_cast<T>(-1);
+				Row[i] *= static_cast<T>(-1);
+			}
+		}
+
+		// Now, get the rotations out, as described in the gem.
+
+		// FIXME - Add the ability to return either quaternions (which are
+		// easier to recompose with) or Euler angles (rx, ry, rz), which
+		// are easier for authors to deal with. The latter will only be useful
+		// when we fix https://bugs.webkit.org/show_bug.cgi?id=23799, so I
+		// will leave the Euler angle code here for now.
+
+		// ret.rotateY = asin(-Row[0][2]);
+		// if (cos(ret.rotateY) != 0) {
+		//     ret.rotateX = atan2(Row[1][2], Row[2][2]);
+		//     ret.rotateZ = atan2(Row[0][1], Row[0][0]);
+		// } else {
+		//     ret.rotateX = atan2(-Row[2][0], Row[1][1]);
+		//     ret.rotateZ = 0;
+		// }
+
+		int i, j, k = 0;
+		T root, trace = Row[0].x + Row[1].y + Row[2].z;
+		if(trace > static_cast<T>(0))
+		{
+			root = sqrt(trace + static_cast<T>(1.0));
+			Orientation.w = static_cast<T>(0.5) * root;
+			root = static_cast<T>(0.5) / root;
+			Orientation.x = root * (Row[1].z - Row[2].y);
+			Orientation.y = root * (Row[2].x - Row[0].z);
+			Orientation.z = root * (Row[0].y - Row[1].x);
+		} // End if > 0
+		else
+		{
+			static int Next[3] = {1, 2, 0};
+			i = 0;
+			if(Row[1].y > Row[0].x) i = 1;
+			if(Row[2].z > Row[i][i]) i = 2;
+			j = Next[i];
+			k = Next[j];
+
+#           ifdef GLM_FORCE_QUAT_DATA_WXYZ
+                int off = 1;
+#           else
+                int off = 0;
+#           endif
+
+			root = sqrt(Row[i][i] - Row[j][j] - Row[k][k] + static_cast<T>(1.0));
+
+			Orientation[i + off] = static_cast<T>(0.5) * root;
+			root = static_cast<T>(0.5) / root;
+			Orientation[j + off] = root * (Row[i][j] + Row[j][i]);
+			Orientation[k + off] = root * (Row[i][k] + Row[k][i]);
+			Orientation.w = root * (Row[j][k] - Row[k][j]);
+		} // End if <= 0
+
+		return true;
+	}
+
+	// Recomposes a model matrix from a previously-decomposed matrix
+	// http://www.opensource.apple.com/source/WebCore/WebCore-514/platform/graphics/transforms/TransformationMatrix.cpp
+	// https://stackoverflow.com/a/75573092/1047040
+	template <typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> recompose(
+		vec<3, T, Q> const& scale, qua<T, Q> const& orientation, vec<3, T, Q> const& translation,
+		vec<3, T, Q> const& skew, vec<4, T, Q> const& perspective)
+	{
+		glm::mat4 m = glm::mat4(1.f);
+
+		m[0][3] = perspective.x;
+		m[1][3] = perspective.y;
+		m[2][3] = perspective.z;
+		m[3][3] = perspective.w;
+
+		m *= glm::translate(translation);
+		m *= glm::mat4_cast(orientation);
+
+		if (abs(skew.x) > static_cast<T>(0)) {
+			glm::mat4 tmp(1.f);
+			tmp[2][1] = skew.x;
+			m *= tmp;
+		}
+
+		if (abs(skew.y) > static_cast<T>(0)) {
+			glm::mat4 tmp(1.f);
+			tmp[2][0] = skew.y;
+			m *= tmp;
+		}
+
+		if (abs(skew.z) > static_cast<T>(0)) {
+			glm::mat4 tmp(1.f);
+			tmp[1][0] = skew.z;
+			m *= tmp;
+		}
+
+		m *= glm::scale(scale);
+
+		return m;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.hpp
new file mode 100644
index 000000000000..5a975d60b6c2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.hpp
@@ -0,0 +1,69 @@
+/// @ref gtx_matrix_factorisation
+/// @file glm/gtx/matrix_factorisation.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_matrix_factorisation GLM_GTX_matrix_factorisation
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_factorisation.hpp> to use the features of this extension.
+///
+/// Functions to factor matrices in various forms
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_factorisation is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_factorisation extension included")
+#	endif
+#endif
+
+/*
+Suggestions:
+ - Move helper functions flipud and fliplr to another file: They may be helpful in more general circumstances.
+ - Implement other types of matrix factorisation, such as: QL and LQ, L(D)U, eigendecompositions, etc...
+*/
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_factorisation
+	/// @{
+
+	/// Flips the matrix rows up and down.
+	///
+	/// From GLM_GTX_matrix_factorisation extension.
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> flipud(mat<C, R, T, Q> const& in);
+
+	/// Flips the matrix columns right and left.
+	///
+	/// From GLM_GTX_matrix_factorisation extension.
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> fliplr(mat<C, R, T, Q> const& in);
+
+	/// Performs QR factorisation of a matrix.
+	/// Returns 2 matrices, q and r, such that the columns of q are orthonormal and span the same subspace than those of the input matrix, r is an upper triangular matrix, and q*r=in.
+	/// Given an n-by-m input matrix, q has dimensions min(n,m)-by-m, and r has dimensions n-by-min(n,m).
+	///
+	/// From GLM_GTX_matrix_factorisation extension.
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL void qr_decompose(mat<C, R, T, Q> const& in, mat<(C < R ? C : R), R, T, Q>& q, mat<C, (C < R ? C : R), T, Q>& r);
+
+	/// Performs RQ factorisation of a matrix.
+	/// Returns 2 matrices, r and q, such that r is an upper triangular matrix, the rows of q are orthonormal and span the same subspace than those of the input matrix, and r*q=in.
+	/// Note that in the context of RQ factorisation, the diagonal is seen as starting in the lower-right corner of the matrix, instead of the usual upper-left.
+	/// Given an n-by-m input matrix, r has dimensions min(n,m)-by-m, and q has dimensions n-by-min(n,m).
+	///
+	/// From GLM_GTX_matrix_factorisation extension.
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL void rq_decompose(mat<C, R, T, Q> const& in, mat<(C < R ? C : R), R, T, Q>& r, mat<C, (C < R ? C : R), T, Q>& q);
+
+	/// @}
+}
+
+#include "matrix_factorisation.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.inl
new file mode 100644
index 000000000000..6f1683c00780
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_factorisation.inl
@@ -0,0 +1,84 @@
+/// @ref gtx_matrix_factorisation
+
+namespace glm
+{
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<C, R, T, Q> flipud(mat<C, R, T, Q> const& in)
+	{
+		mat<R, C, T, Q> tin = transpose(in);
+		tin = fliplr(tin);
+		mat<C, R, T, Q> out = transpose(tin);
+
+		return out;
+	}
+
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<C, R, T, Q> fliplr(mat<C, R, T, Q> const& in)
+	{
+		mat<C, R, T, Q> out;
+		for (length_t i = 0; i < C; i++)
+		{
+			out[i] = in[(C - i) - 1];
+		}
+
+		return out;
+	}
+
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER void qr_decompose(mat<C, R, T, Q> const& in, mat<(C < R ? C : R), R, T, Q>& q, mat<C, (C < R ? C : R), T, Q>& r)
+	{
+		// Uses modified Gram-Schmidt method
+		// Source: https://en.wikipedia.org/wiki/Gram%E2%80%93Schmidt_process
+		// And https://en.wikipedia.org/wiki/QR_decomposition
+
+		//For all the linearly independs columns of the input...
+		// (there can be no more linearly independents columns than there are rows.)
+		for (length_t i = 0; i < (C < R ? C : R); i++)
+		{
+			//Copy in Q the input's i-th column.
+			q[i] = in[i];
+
+			//j = [0,i[
+			// Make that column orthogonal to all the previous ones by substracting to it the non-orthogonal projection of all the previous columns.
+			// Also: Fill the zero elements of R
+			for (length_t j = 0; j < i; j++)
+			{
+				q[i] -= dot(q[i], q[j])*q[j];
+				r[j][i] = 0;
+			}
+
+			//Now, Q i-th column is orthogonal to all the previous columns. Normalize it.
+			q[i] = normalize(q[i]);
+
+			//j = [i,C[
+			//Finally, compute the corresponding coefficients of R by computing the projection of the resulting column on the other columns of the input.
+			for (length_t j = i; j < C; j++)
+			{
+				r[j][i] = dot(in[j], q[i]);
+			}
+		}
+	}
+
+	template <length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER void rq_decompose(mat<C, R, T, Q> const& in, mat<(C < R ? C : R), R, T, Q>& r, mat<C, (C < R ? C : R), T, Q>& q)
+	{
+		// From https://en.wikipedia.org/wiki/QR_decomposition:
+		// The RQ decomposition transforms a matrix A into the product of an upper triangular matrix R (also known as right-triangular) and an orthogonal matrix Q. The only difference from QR decomposition is the order of these matrices.
+		// QR decomposition is Gram-Schmidt orthogonalization of columns of A, started from the first column.
+		// RQ decomposition is Gram-Schmidt orthogonalization of rows of A, started from the last row.
+
+		mat<R, C, T, Q> tin = transpose(in);
+		tin = fliplr(tin);
+
+		mat<R, (C < R ? C : R), T, Q> tr;
+		mat<(C < R ? C : R), C, T, Q> tq;
+		qr_decompose(tin, tq, tr);
+
+		tr = fliplr(tr);
+		r = transpose(tr);
+		r = fliplr(r);
+
+		tq = fliplr(tq);
+		q = transpose(tq);
+	}
+} //namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.hpp
new file mode 100644
index 000000000000..7d5ad4cd9ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.hpp
@@ -0,0 +1,60 @@
+/// @ref gtx_matrix_interpolation
+/// @file glm/gtx/matrix_interpolation.hpp
+/// @author Ghenadii Ursachi (the.asteroth@gmail.com)
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_matrix_interpolation GLM_GTX_matrix_interpolation
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_interpolation.hpp> to use the features of this extension.
+///
+/// Allows to directly interpolate two matrices.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_interpolation is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_interpolation extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_interpolation
+	/// @{
+
+	/// Get the axis and angle of the rotation from a matrix.
+	/// From GLM_GTX_matrix_interpolation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL void axisAngle(
+		mat<4, 4, T, Q> const& Mat, vec<3, T, Q> & Axis, T & Angle);
+
+	/// Build a matrix from axis and angle.
+	/// From GLM_GTX_matrix_interpolation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> axisAngleMatrix(
+		vec<3, T, Q> const& Axis, T const Angle);
+
+	/// Extracts the rotation part of a matrix.
+	/// From GLM_GTX_matrix_interpolation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> extractMatrixRotation(
+		mat<4, 4, T, Q> const& Mat);
+
+	/// Build a interpolation of 4 * 4 matrixes.
+	/// From GLM_GTX_matrix_interpolation extension.
+	/// Warning! works only with rotation and/or translation matrixes, scale will generate unexpected results.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> interpolate(
+		mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2, T const Delta);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_interpolation.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.inl
new file mode 100644
index 000000000000..f4ba3a6f3ac6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_interpolation.inl
@@ -0,0 +1,146 @@
+/// @ref gtx_matrix_interpolation
+
+#include "../ext/scalar_constants.hpp"
+
+#include <limits>
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER void axisAngle(mat<4, 4, T, Q> const& m, vec<3, T, Q>& axis, T& angle)
+	{
+		T const epsilon =
+		    std::numeric_limits<T>::epsilon() * static_cast<T>(1e2);
+
+        bool const nearSymmetrical =
+            abs(m[1][0] - m[0][1]) < epsilon &&
+            abs(m[2][0] - m[0][2]) < epsilon &&
+            abs(m[2][1] - m[1][2]) < epsilon;
+
+		if(nearSymmetrical)
+		{
+            bool const nearIdentity =
+                abs(m[1][0] + m[0][1]) < epsilon &&
+                abs(m[2][0] + m[0][2]) < epsilon &&
+                abs(m[2][1] + m[1][2]) < epsilon &&
+                abs(m[0][0] + m[1][1] + m[2][2] - T(3.0)) < epsilon;
+			if (nearIdentity)
+			{
+				angle = static_cast<T>(0.0);
+				axis = vec<3, T, Q>(
+				    static_cast<T>(1.0), static_cast<T>(0.0), static_cast<T>(0.0));
+				return;
+			}
+			angle = pi<T>();
+			T xx = (m[0][0] + static_cast<T>(1.0)) * static_cast<T>(0.5);
+			T yy = (m[1][1] + static_cast<T>(1.0)) * static_cast<T>(0.5);
+			T zz = (m[2][2] + static_cast<T>(1.0)) * static_cast<T>(0.5);
+			T xy = (m[1][0] + m[0][1]) * static_cast<T>(0.25);
+			T xz = (m[2][0] + m[0][2]) * static_cast<T>(0.25);
+			T yz = (m[2][1] + m[1][2]) * static_cast<T>(0.25);
+			if((xx > yy) && (xx > zz))
+			{
+				if(xx < epsilon)
+				{
+					axis.x = static_cast<T>(0.0);
+					axis.y = static_cast<T>(0.7071);
+					axis.z = static_cast<T>(0.7071);
+				}
+				else
+				{
+					axis.x = sqrt(xx);
+					axis.y = xy / axis.x;
+					axis.z = xz / axis.x;
+				}
+			}
+			else if (yy > zz)
+			{
+				if(yy < epsilon)
+				{
+					axis.x = static_cast<T>(0.7071);
+					axis.y = static_cast<T>(0.0);
+					axis.z = static_cast<T>(0.7071);
+				}
+				else
+				{
+					axis.y = sqrt(yy);
+					axis.x = xy / axis.y;
+					axis.z = yz / axis.y;
+				}
+			}
+			else
+			{
+				if (zz < epsilon)
+				{
+					axis.x = static_cast<T>(0.7071);
+					axis.y = static_cast<T>(0.7071);
+					axis.z = static_cast<T>(0.0);
+				}
+				else
+				{
+					axis.z = sqrt(zz);
+					axis.x = xz / axis.z;
+					axis.y = yz / axis.z;
+				}
+			}
+			return;
+		}
+
+		T const angleCos = (m[0][0] + m[1][1] + m[2][2] - static_cast<T>(1)) * static_cast<T>(0.5);
+		if(angleCos >= static_cast<T>(1.0))
+		{
+			angle = static_cast<T>(0.0);
+		}
+		else if (angleCos <= static_cast<T>(-1.0))
+		{
+			angle = pi<T>();
+		}
+		else
+		{
+			angle = acos(angleCos);
+		}
+
+        axis = glm::normalize(glm::vec<3, T, Q>(
+            m[1][2] - m[2][1], m[2][0] - m[0][2], m[0][1] - m[1][0]));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> axisAngleMatrix(vec<3, T, Q> const& axis, T const angle)
+	{
+		T c = cos(angle);
+		T s = sin(angle);
+		T t = static_cast<T>(1) - c;
+		vec<3, T, Q> n = normalize(axis);
+
+		return mat<4, 4, T, Q>(
+			t * n.x * n.x + c,          t * n.x * n.y + n.z * s,    t * n.x * n.z - n.y * s,    static_cast<T>(0.0),
+			t * n.x * n.y - n.z * s,    t * n.y * n.y + c,          t * n.y * n.z + n.x * s,    static_cast<T>(0.0),
+			t * n.x * n.z + n.y * s,    t * n.y * n.z - n.x * s,    t * n.z * n.z + c,          static_cast<T>(0.0),
+			static_cast<T>(0.0),        static_cast<T>(0.0),        static_cast<T>(0.0),        static_cast<T>(1.0));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> extractMatrixRotation(mat<4, 4, T, Q> const& m)
+	{
+		return mat<4, 4, T, Q>(
+			m[0][0], m[0][1], m[0][2], static_cast<T>(0.0),
+			m[1][0], m[1][1], m[1][2], static_cast<T>(0.0),
+			m[2][0], m[2][1], m[2][2], static_cast<T>(0.0),
+			static_cast<T>(0.0), static_cast<T>(0.0), static_cast<T>(0.0), static_cast<T>(1.0));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> interpolate(mat<4, 4, T, Q> const& m1, mat<4, 4, T, Q> const& m2, T const delta)
+	{
+		mat<4, 4, T, Q> m1rot = extractMatrixRotation(m1);
+		mat<4, 4, T, Q> dltRotation = m2 * transpose(m1rot);
+		vec<3, T, Q> dltAxis;
+		T dltAngle;
+		axisAngle(dltRotation, dltAxis, dltAngle);
+		mat<4, 4, T, Q> out = axisAngleMatrix(dltAxis, dltAngle * delta) * m1rot;
+		out[3][0] = m1[3][0] + delta * (m2[3][0] - m1[3][0]);
+		out[3][1] = m1[3][1] + delta * (m2[3][1] - m1[3][1]);
+		out[3][2] = m1[3][2] + delta * (m2[3][2] - m1[3][2]);
+		return out;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.hpp
new file mode 100644
index 000000000000..8c6bc22d14e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.hpp
@@ -0,0 +1,119 @@
+/// @ref gtx_matrix_major_storage
+/// @file glm/gtx/matrix_major_storage.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_matrix_major_storage GLM_GTX_matrix_major_storage
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_major_storage.hpp> to use the features of this extension.
+///
+/// Build matrices with specific matrix order, row or column
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_major_storage is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_major_storage extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_major_storage
+	/// @{
+
+	//! Build a row major matrix from row vectors.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 2, T, Q> rowMajor2(
+		vec<2, T, Q> const& v1,
+		vec<2, T, Q> const& v2);
+
+	//! Build a row major matrix from other matrix.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 2, T, Q> rowMajor2(
+		mat<2, 2, T, Q> const& m);
+
+	//! Build a row major matrix from row vectors.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> rowMajor3(
+		vec<3, T, Q> const& v1,
+		vec<3, T, Q> const& v2,
+		vec<3, T, Q> const& v3);
+
+	//! Build a row major matrix from other matrix.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> rowMajor3(
+		mat<3, 3, T, Q> const& m);
+
+	//! Build a row major matrix from row vectors.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> rowMajor4(
+		vec<4, T, Q> const& v1,
+		vec<4, T, Q> const& v2,
+		vec<4, T, Q> const& v3,
+		vec<4, T, Q> const& v4);
+
+	//! Build a row major matrix from other matrix.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> rowMajor4(
+		mat<4, 4, T, Q> const& m);
+
+	//! Build a column major matrix from column vectors.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 2, T, Q> colMajor2(
+		vec<2, T, Q> const& v1,
+		vec<2, T, Q> const& v2);
+
+	//! Build a column major matrix from other matrix.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 2, T, Q> colMajor2(
+		mat<2, 2, T, Q> const& m);
+
+	//! Build a column major matrix from column vectors.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> colMajor3(
+		vec<3, T, Q> const& v1,
+		vec<3, T, Q> const& v2,
+		vec<3, T, Q> const& v3);
+
+	//! Build a column major matrix from other matrix.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> colMajor3(
+		mat<3, 3, T, Q> const& m);
+
+	//! Build a column major matrix from column vectors.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> colMajor4(
+		vec<4, T, Q> const& v1,
+		vec<4, T, Q> const& v2,
+		vec<4, T, Q> const& v3,
+		vec<4, T, Q> const& v4);
+
+	//! Build a column major matrix from other matrix.
+	//! From GLM_GTX_matrix_major_storage extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> colMajor4(
+		mat<4, 4, T, Q> const& m);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_major_storage.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.inl
new file mode 100644
index 000000000000..279dd3433d0b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_major_storage.inl
@@ -0,0 +1,166 @@
+/// @ref gtx_matrix_major_storage
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> rowMajor2
+	(
+		vec<2, T, Q> const& v1,
+		vec<2, T, Q> const& v2
+	)
+	{
+		mat<2, 2, T, Q> Result;
+		Result[0][0] = v1.x;
+		Result[1][0] = v1.y;
+		Result[0][1] = v2.x;
+		Result[1][1] = v2.y;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> rowMajor2(
+		const mat<2, 2, T, Q>& m)
+	{
+		mat<2, 2, T, Q> Result;
+		Result[0][0] = m[0][0];
+		Result[0][1] = m[1][0];
+		Result[1][0] = m[0][1];
+		Result[1][1] = m[1][1];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> rowMajor3(
+		const vec<3, T, Q>& v1,
+		const vec<3, T, Q>& v2,
+		const vec<3, T, Q>& v3)
+	{
+		mat<3, 3, T, Q> Result;
+		Result[0][0] = v1.x;
+		Result[1][0] = v1.y;
+		Result[2][0] = v1.z;
+		Result[0][1] = v2.x;
+		Result[1][1] = v2.y;
+		Result[2][1] = v2.z;
+		Result[0][2] = v3.x;
+		Result[1][2] = v3.y;
+		Result[2][2] = v3.z;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> rowMajor3(
+		const mat<3, 3, T, Q>& m)
+	{
+		mat<3, 3, T, Q> Result;
+		Result[0][0] = m[0][0];
+		Result[0][1] = m[1][0];
+		Result[0][2] = m[2][0];
+		Result[1][0] = m[0][1];
+		Result[1][1] = m[1][1];
+		Result[1][2] = m[2][1];
+		Result[2][0] = m[0][2];
+		Result[2][1] = m[1][2];
+		Result[2][2] = m[2][2];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> rowMajor4(
+		const vec<4, T, Q>& v1,
+		const vec<4, T, Q>& v2,
+		const vec<4, T, Q>& v3,
+		const vec<4, T, Q>& v4)
+	{
+		mat<4, 4, T, Q> Result;
+		Result[0][0] = v1.x;
+		Result[1][0] = v1.y;
+		Result[2][0] = v1.z;
+		Result[3][0] = v1.w;
+		Result[0][1] = v2.x;
+		Result[1][1] = v2.y;
+		Result[2][1] = v2.z;
+		Result[3][1] = v2.w;
+		Result[0][2] = v3.x;
+		Result[1][2] = v3.y;
+		Result[2][2] = v3.z;
+		Result[3][2] = v3.w;
+		Result[0][3] = v4.x;
+		Result[1][3] = v4.y;
+		Result[2][3] = v4.z;
+		Result[3][3] = v4.w;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> rowMajor4(
+		const mat<4, 4, T, Q>& m)
+	{
+		mat<4, 4, T, Q> Result;
+		Result[0][0] = m[0][0];
+		Result[0][1] = m[1][0];
+		Result[0][2] = m[2][0];
+		Result[0][3] = m[3][0];
+		Result[1][0] = m[0][1];
+		Result[1][1] = m[1][1];
+		Result[1][2] = m[2][1];
+		Result[1][3] = m[3][1];
+		Result[2][0] = m[0][2];
+		Result[2][1] = m[1][2];
+		Result[2][2] = m[2][2];
+		Result[2][3] = m[3][2];
+		Result[3][0] = m[0][3];
+		Result[3][1] = m[1][3];
+		Result[3][2] = m[2][3];
+		Result[3][3] = m[3][3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> colMajor2(
+		const vec<2, T, Q>& v1,
+		const vec<2, T, Q>& v2)
+	{
+		return mat<2, 2, T, Q>(v1, v2);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> colMajor2(
+		const mat<2, 2, T, Q>& m)
+	{
+		return mat<2, 2, T, Q>(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> colMajor3(
+		const vec<3, T, Q>& v1,
+		const vec<3, T, Q>& v2,
+		const vec<3, T, Q>& v3)
+	{
+		return mat<3, 3, T, Q>(v1, v2, v3);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> colMajor3(
+		const mat<3, 3, T, Q>& m)
+	{
+		return mat<3, 3, T, Q>(m);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> colMajor4(
+		const vec<4, T, Q>& v1,
+		const vec<4, T, Q>& v2,
+		const vec<4, T, Q>& v3,
+		const vec<4, T, Q>& v4)
+	{
+		return mat<4, 4, T, Q>(v1, v2, v3, v4);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> colMajor4(
+		const mat<4, 4, T, Q>& m)
+	{
+		return mat<4, 4, T, Q>(m);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.hpp
new file mode 100644
index 000000000000..de6ff1f86f47
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.hpp
@@ -0,0 +1,103 @@
+/// @ref gtx_matrix_operation
+/// @file glm/gtx/matrix_operation.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_matrix_operation GLM_GTX_matrix_operation
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_operation.hpp> to use the features of this extension.
+///
+/// Build diagonal matrices from vectors.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_operation is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_operation extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_operation
+	/// @{
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 2, T, Q> diagonal2x2(
+		vec<2, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 3, T, Q> diagonal2x3(
+		vec<2, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 4, T, Q> diagonal2x4(
+		vec<2, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 2, T, Q> diagonal3x2(
+		vec<2, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> diagonal3x3(
+		vec<3, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 4, T, Q> diagonal3x4(
+		vec<3, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 2, T, Q> diagonal4x2(
+		vec<2, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 3, T, Q> diagonal4x3(
+		vec<3, T, Q> const& v);
+
+	//! Build a diagonal matrix.
+	//! From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> diagonal4x4(
+		vec<4, T, Q> const& v);
+
+	/// Build an adjugate  matrix.
+	/// From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<2, 2, T, Q> adjugate(mat<2, 2, T, Q> const& m);
+
+	/// Build an adjugate  matrix.
+	/// From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> adjugate(mat<3, 3, T, Q> const& m);
+
+	/// Build an adjugate  matrix.
+	/// From GLM_GTX_matrix_operation extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> adjugate(mat<4, 4, T, Q> const& m);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_operation.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.inl
new file mode 100644
index 000000000000..a4f4a850002f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_operation.inl
@@ -0,0 +1,176 @@
+/// @ref gtx_matrix_operation
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> diagonal2x2
+	(
+		vec<2, T, Q> const& v
+	)
+	{
+		mat<2, 2, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 3, T, Q> diagonal2x3
+	(
+		vec<2, T, Q> const& v
+	)
+	{
+		mat<2, 3, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 4, T, Q> diagonal2x4
+	(
+		vec<2, T, Q> const& v
+	)
+	{
+		mat<2, 4, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 2, T, Q> diagonal3x2
+	(
+		vec<2, T, Q> const& v
+	)
+	{
+		mat<3, 2, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> diagonal3x3
+	(
+		vec<3, T, Q> const& v
+	)
+	{
+		mat<3, 3, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		Result[2][2] = v[2];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 4, T, Q> diagonal3x4
+	(
+		vec<3, T, Q> const& v
+	)
+	{
+		mat<3, 4, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		Result[2][2] = v[2];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> diagonal4x4
+	(
+		vec<4, T, Q> const& v
+	)
+	{
+		mat<4, 4, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		Result[2][2] = v[2];
+		Result[3][3] = v[3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 3, T, Q> diagonal4x3
+	(
+		vec<3, T, Q> const& v
+	)
+	{
+		mat<4, 3, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		Result[2][2] = v[2];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 2, T, Q> diagonal4x2
+	(
+		vec<2, T, Q> const& v
+	)
+	{
+		mat<4, 2, T, Q> Result(static_cast<T>(1));
+		Result[0][0] = v[0];
+		Result[1][1] = v[1];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<2, 2, T, Q> adjugate(mat<2, 2, T, Q> const& m)
+	{
+		return mat<2, 2, T, Q>(
+			+m[1][1], -m[0][1],
+			-m[1][0], +m[0][0]);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> adjugate(mat<3, 3, T, Q> const& m)
+	{
+		T const m00 = determinant(mat<2, 2, T, Q>(m[1][1], m[2][1], m[1][2], m[2][2]));
+		T const m01 = determinant(mat<2, 2, T, Q>(m[0][1], m[2][1], m[0][2], m[2][2]));
+		T const m02 = determinant(mat<2, 2, T, Q>(m[0][1], m[1][1], m[0][2], m[1][2]));
+
+		T const m10 = determinant(mat<2, 2, T, Q>(m[1][0], m[2][0], m[1][2], m[2][2]));
+		T const m11 = determinant(mat<2, 2, T, Q>(m[0][0], m[2][0], m[0][2], m[2][2]));
+		T const m12 = determinant(mat<2, 2, T, Q>(m[0][0], m[1][0], m[0][2], m[1][2]));
+
+		T const m20 = determinant(mat<2, 2, T, Q>(m[1][0], m[2][0], m[1][1], m[2][1]));
+		T const m21 = determinant(mat<2, 2, T, Q>(m[0][0], m[2][0], m[0][1], m[2][1]));
+		T const m22 = determinant(mat<2, 2, T, Q>(m[0][0], m[1][0], m[0][1], m[1][1]));
+
+		return mat<3, 3, T, Q>(
+			+m00, -m01, +m02,
+			-m10, +m11, -m12,
+			+m20, -m21, +m22);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> adjugate(mat<4, 4, T, Q> const& m)
+	{
+		T const m00 = determinant(mat<3, 3, T, Q>(m[1][1], m[1][2], m[1][3], m[2][1], m[2][2], m[2][3], m[3][1], m[3][2], m[3][3]));
+		T const m01 = determinant(mat<3, 3, T, Q>(m[1][0], m[1][2], m[1][3], m[2][0], m[2][2], m[2][3], m[3][0], m[3][2], m[3][3]));
+		T const m02 = determinant(mat<3, 3, T, Q>(m[1][0], m[1][1], m[1][3], m[2][0], m[2][1], m[2][3], m[3][0], m[3][1], m[3][3]));
+		T const m03 = determinant(mat<3, 3, T, Q>(m[1][0], m[1][1], m[1][2], m[2][0], m[2][1], m[2][2], m[3][0], m[3][1], m[3][2]));
+
+		T const m10 = determinant(mat<3, 3, T, Q>(m[0][1], m[0][2], m[0][3], m[2][1], m[2][2], m[2][3], m[3][1], m[3][2], m[3][3]));
+		T const m11 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][2], m[0][3], m[2][0], m[2][2], m[2][3], m[3][0], m[3][2], m[3][3]));
+		T const m12 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][1], m[0][3], m[2][0], m[2][1], m[2][3], m[3][0], m[3][1], m[3][3]));
+		T const m13 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][1], m[0][2], m[2][0], m[2][1], m[2][2], m[3][0], m[3][1], m[3][2]));
+
+		T const m20 = determinant(mat<3, 3, T, Q>(m[0][1], m[0][2], m[0][3], m[1][1], m[1][2], m[1][3], m[3][1], m[3][2], m[3][3]));
+		T const m21 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][2], m[0][3], m[1][0], m[1][2], m[1][3], m[3][0], m[3][2], m[3][3]));
+		T const m22 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][1], m[0][3], m[1][0], m[1][1], m[1][3], m[3][0], m[3][1], m[3][3]));
+		T const m23 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][1], m[0][2], m[1][0], m[1][1], m[1][2], m[3][0], m[3][1], m[3][2]));
+
+		T const m30 = determinant(mat<3, 3, T, Q>(m[0][1], m[0][2], m[0][3], m[1][1], m[1][2], m[1][3], m[2][1], m[2][2], m[2][3]));
+		T const m31 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][2], m[0][3], m[1][0], m[1][2], m[1][3], m[2][0], m[2][2], m[2][3]));
+		T const m32 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][1], m[0][3], m[1][0], m[1][1], m[1][3], m[2][0], m[2][1], m[2][3]));
+		T const m33 = determinant(mat<3, 3, T, Q>(m[0][0], m[0][1], m[0][2], m[1][0], m[1][1], m[1][2], m[2][0], m[2][1], m[2][2]));
+
+		return mat<4, 4, T, Q>(
+			+m00, -m10, +m20, -m30,
+			-m01, +m11, -m21, +m31,
+			+m02, -m12, +m22, -m32,
+			-m03, +m13, -m23, +m33);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.hpp
new file mode 100644
index 000000000000..8011b2b1d469
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.hpp
@@ -0,0 +1,77 @@
+/// @ref gtx_matrix_query
+/// @file glm/gtx/matrix_query.hpp
+///
+/// @see core (dependence)
+/// @see gtx_vector_query (dependence)
+///
+/// @defgroup gtx_matrix_query GLM_GTX_matrix_query
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_query.hpp> to use the features of this extension.
+///
+/// Query to evaluate matrix properties
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtx/vector_query.hpp"
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_query is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_query extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_query
+	/// @{
+
+	/// Return whether a matrix a null matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNull(mat<2, 2, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix a null matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNull(mat<3, 3, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix is a null matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNull(mat<4, 4, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix is an identity matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<length_t C, length_t R, typename T, qualifier Q, template<length_t, length_t, typename, qualifier> class matType>
+	GLM_FUNC_DECL bool isIdentity(matType<C, R, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix is a normalized matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNormalized(mat<2, 2, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix is a normalized matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNormalized(mat<3, 3, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix is a normalized matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNormalized(mat<4, 4, T, Q> const& m, T const& epsilon);
+
+	/// Return whether a matrix is an orthonormalized matrix.
+	/// From GLM_GTX_matrix_query extension.
+	template<length_t C, length_t R, typename T, qualifier Q, template<length_t, length_t, typename, qualifier> class matType>
+	GLM_FUNC_DECL bool isOrthogonal(matType<C, R, T, Q> const& m, T const& epsilon);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_query.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.inl
new file mode 100644
index 000000000000..dc3ec8453b68
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_query.inl
@@ -0,0 +1,119 @@
+/// @ref gtx_matrix_query
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNull(mat<2, 2, T, Q> const& m, T const& epsilon)
+	{
+		bool result = true;
+		for(length_t i = 0; result && i < m.length() ; ++i)
+			result = isNull(m[i], epsilon);
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNull(mat<3, 3, T, Q> const& m, T const& epsilon)
+	{
+		bool result = true;
+		for(length_t i = 0; result && i < m.length() ; ++i)
+			result = isNull(m[i], epsilon);
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNull(mat<4, 4, T, Q> const& m, T const& epsilon)
+	{
+		bool result = true;
+		for(length_t i = 0; result && i < m.length() ; ++i)
+			result = isNull(m[i], epsilon);
+		return result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isIdentity(mat<C, R, T, Q> const& m, T const& epsilon)
+	{
+		bool result = true;
+		for(length_t i = 0; result && i < m.length(); ++i)
+		{
+			for(length_t j = 0; result && j < glm::min(i, m[0].length()); ++j)
+				result = abs(m[i][j]) <= epsilon;
+			if(result && i < m[0].length())
+				result = abs(m[i][i] - 1) <= epsilon;
+			for(length_t j = i + 1; result && j < m[0].length(); ++j)
+				result = abs(m[i][j]) <= epsilon;
+		}
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNormalized(mat<2, 2, T, Q> const& m, T const& epsilon)
+	{
+		bool result(true);
+		for(length_t i = 0; result && i < m.length(); ++i)
+			result = isNormalized(m[i], epsilon);
+		for(length_t i = 0; result && i < m.length(); ++i)
+		{
+			typename mat<2, 2, T, Q>::col_type v;
+			for(length_t j = 0; j < m.length(); ++j)
+				v[j] = m[j][i];
+			result = isNormalized(v, epsilon);
+		}
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNormalized(mat<3, 3, T, Q> const& m, T const& epsilon)
+	{
+		bool result(true);
+		for(length_t i = 0; result && i < m.length(); ++i)
+			result = isNormalized(m[i], epsilon);
+		for(length_t i = 0; result && i < m.length(); ++i)
+		{
+			typename mat<3, 3, T, Q>::col_type v;
+			for(length_t j = 0; j < m.length(); ++j)
+				v[j] = m[j][i];
+			result = isNormalized(v, epsilon);
+		}
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNormalized(mat<4, 4, T, Q> const& m, T const& epsilon)
+	{
+		bool result(true);
+		for(length_t i = 0; result && i < m.length(); ++i)
+			result = isNormalized(m[i], epsilon);
+		for(length_t i = 0; result && i < m.length(); ++i)
+		{
+			typename mat<4, 4, T, Q>::col_type v;
+			for(length_t j = 0; j < m.length(); ++j)
+				v[j] = m[j][i];
+			result = isNormalized(v, epsilon);
+		}
+		return result;
+	}
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isOrthogonal(mat<C, R, T, Q> const& m, T const& epsilon)
+	{
+		bool result = true;
+		for(length_t i(0); result && i < m.length(); ++i)
+		{
+			result = isNormalized(m[i], epsilon);
+			for(length_t j(i + 1); result && j < m.length(); ++j)
+				result = abs(dot(m[i], m[j])) <= epsilon;
+		}
+
+		if(result)
+		{
+			mat<C, R, T, Q> tmp = transpose(m);
+			for(length_t i(0); result && i < m.length(); ++i)
+			{
+				result = isNormalized(tmp[i], epsilon);
+				for(length_t j(i + 1); result && j < m.length(); ++j)
+					result = abs(dot(tmp[i], tmp[j])) <= epsilon;
+			}
+		}
+		return result;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.hpp
new file mode 100644
index 000000000000..5f9c54021851
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.hpp
@@ -0,0 +1,81 @@
+/// @ref gtx_matrix_transform_2d
+/// @file glm/gtx/matrix_transform_2d.hpp
+/// @author Miguel Ángel Pérez Martínez
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_matrix_transform_2d GLM_GTX_matrix_transform_2d
+/// @ingroup gtx
+///
+/// Include <glm/gtx/matrix_transform_2d.hpp> to use the features of this extension.
+///
+/// Defines functions that generate common 2d transformation matrices.
+
+#pragma once
+
+// Dependency:
+#include "../mat3x3.hpp"
+#include "../vec2.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_matrix_transform_2d is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_matrix_transform_2d extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_matrix_transform_2d
+	/// @{
+
+	/// Builds a translation 3 * 3 matrix created from a vector of 2 components.
+	///
+	/// @param m Input matrix multiplied by this translation matrix.
+	/// @param v Coordinates of a translation vector.
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> translate(
+		mat<3, 3, T, Q> const& m,
+		vec<2, T, Q> const& v);
+
+	/// Builds a rotation 3 * 3 matrix created from an angle.
+	///
+	/// @param m Input matrix multiplied by this translation matrix.
+	/// @param angle Rotation angle expressed in radians.
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> rotate(
+		mat<3, 3, T, Q> const& m,
+		T angle);
+
+	/// Builds a scale 3 * 3 matrix created from a vector of 2 components.
+	///
+	/// @param m Input matrix multiplied by this translation matrix.
+	/// @param v Coordinates of a scale vector.
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> scale(
+		mat<3, 3, T, Q> const& m,
+		vec<2, T, Q> const& v);
+
+	/// Builds an horizontal (parallel to the x axis) shear 3 * 3 matrix.
+	///
+	/// @param m Input matrix multiplied by this translation matrix.
+	/// @param y Shear factor.
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> shearX(
+		mat<3, 3, T, Q> const& m,
+		T y);
+
+	/// Builds a vertical (parallel to the y axis) shear 3 * 3 matrix.
+	///
+	/// @param m Input matrix multiplied by this translation matrix.
+	/// @param x Shear factor.
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> shearY(
+		mat<3, 3, T, Q> const& m,
+		T x);
+
+	/// @}
+}//namespace glm
+
+#include "matrix_transform_2d.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.inl
new file mode 100644
index 000000000000..a68d24dc9825
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/matrix_transform_2d.inl
@@ -0,0 +1,68 @@
+/// @ref gtx_matrix_transform_2d
+/// @author Miguel Ángel Pérez Martínez
+
+#include "../trigonometric.hpp"
+
+namespace glm
+{
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> translate(
+		mat<3, 3, T, Q> const& m,
+		vec<2, T, Q> const& v)
+	{
+		mat<3, 3, T, Q> Result(m);
+		Result[2] = m[0] * v[0] + m[1] * v[1] + m[2];
+		return Result;
+	}
+
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> rotate(
+		mat<3, 3, T, Q> const& m,
+		T angle)
+	{
+		T const a = angle;
+		T const c = cos(a);
+		T const s = sin(a);
+
+		mat<3, 3, T, Q> Result;
+		Result[0] = m[0] * c + m[1] * s;
+		Result[1] = m[0] * -s + m[1] * c;
+		Result[2] = m[2];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> scale(
+		mat<3, 3, T, Q> const& m,
+		vec<2, T, Q> const& v)
+	{
+		mat<3, 3, T, Q> Result;
+		Result[0] = m[0] * v[0];
+		Result[1] = m[1] * v[1];
+		Result[2] = m[2];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> shearX(
+		mat<3, 3, T, Q> const& m,
+		T y)
+	{
+		mat<3, 3, T, Q> Result(1);
+		Result[0][1] = y;
+		return m * Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> shearY(
+		mat<3, 3, T, Q> const& m,
+		T x)
+	{
+		mat<3, 3, T, Q> Result(1);
+		Result[1][0] = x;
+		return m * Result;
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.hpp
new file mode 100644
index 000000000000..b242e357e57a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.hpp
@@ -0,0 +1,41 @@
+/// @ref gtx_mixed_product
+/// @file glm/gtx/mixed_product.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_mixed_product GLM_GTX_mixed_producte
+/// @ingroup gtx
+///
+/// Include <glm/gtx/mixed_product.hpp> to use the features of this extension.
+///
+/// Mixed product of 3 vectors.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_mixed_product is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_mixed_product extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_mixed_product
+	/// @{
+
+	/// @brief Mixed product of 3 vectors (from GLM_GTX_mixed_product extension)
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T mixedProduct(
+		vec<3, T, Q> const& v1,
+		vec<3, T, Q> const& v2,
+		vec<3, T, Q> const& v3);
+
+	/// @}
+}// namespace glm
+
+#include "mixed_product.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.inl
new file mode 100644
index 000000000000..e5cdbdb49a2b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/mixed_product.inl
@@ -0,0 +1,15 @@
+/// @ref gtx_mixed_product
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T mixedProduct
+	(
+		vec<3, T, Q> const& v1,
+		vec<3, T, Q> const& v2,
+		vec<3, T, Q> const& v3
+	)
+	{
+		return dot(cross(v1, v2), v3);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/norm.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/norm.hpp
new file mode 100644
index 000000000000..dfaebb7a8be2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/norm.hpp
@@ -0,0 +1,88 @@
+/// @ref gtx_norm
+/// @file glm/gtx/norm.hpp
+///
+/// @see core (dependence)
+/// @see gtx_quaternion (dependence)
+/// @see gtx_component_wise (dependence)
+///
+/// @defgroup gtx_norm GLM_GTX_norm
+/// @ingroup gtx
+///
+/// Include <glm/gtx/norm.hpp> to use the features of this extension.
+///
+/// Various ways to compute vector norms.
+
+#pragma once
+
+// Dependency:
+#include "../geometric.hpp"
+#include "../gtx/quaternion.hpp"
+#include "../gtx/component_wise.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_norm is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_norm extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_norm
+	/// @{
+
+	/// Returns the squared length of x.
+	/// From GLM_GTX_norm extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T length2(vec<L, T, Q> const& x);
+
+	/// Returns the squared distance between p0 and p1, i.e., length2(p0 - p1).
+	/// From GLM_GTX_norm extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T distance2(vec<L, T, Q> const& p0, vec<L, T, Q> const& p1);
+
+	//! Returns the L1 norm between x and y.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T l1Norm(vec<3, T, Q> const& x, vec<3, T, Q> const& y);
+
+	//! Returns the L1 norm of v.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T l1Norm(vec<3, T, Q> const& v);
+
+	//! Returns the L2 norm between x and y.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T l2Norm(vec<3, T, Q> const& x, vec<3, T, Q> const& y);
+
+	//! Returns the L2 norm of v.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T l2Norm(vec<3, T, Q> const& x);
+
+	//! Returns the L norm between x and y.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T lxNorm(vec<3, T, Q> const& x, vec<3, T, Q> const& y, unsigned int Depth);
+
+	//! Returns the L norm of v.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T lxNorm(vec<3, T, Q> const& x, unsigned int Depth);
+
+	//! Returns the LMax norm between x and y.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T lMaxNorm(vec<3, T, Q> const& x, vec<3, T, Q> const& y);
+
+	//! Returns the LMax norm of v.
+	//! From GLM_GTX_norm extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T lMaxNorm(vec<3, T, Q> const& x);
+
+	/// @}
+}//namespace glm
+
+#include "norm.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/norm.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/norm.inl
new file mode 100644
index 000000000000..4a9f796451fd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/norm.inl
@@ -0,0 +1,95 @@
+/// @ref gtx_norm
+
+#include "../detail/qualifier.hpp"
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q, bool Aligned>
+	struct compute_length2
+	{
+		GLM_FUNC_QUALIFIER static T call(vec<L, T, Q> const& v)
+		{
+			return dot(v, v);
+		}
+	};
+}//namespace detail
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType length2(genType x)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'length2' accepts only floating-point inputs");
+		return x * x;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T length2(vec<L, T, Q> const& v)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'length2' accepts only floating-point inputs");
+		return detail::compute_length2<L, T, Q, detail::is_aligned<Q>::value>::call(v);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER T distance2(T p0, T p1)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'distance2' accepts only floating-point inputs");
+		return length2(p1 - p0);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T distance2(vec<L, T, Q> const& p0, vec<L, T, Q> const& p1)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'distance2' accepts only floating-point inputs");
+		return length2(p1 - p0);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T l1Norm(vec<3, T, Q> const& a, vec<3, T, Q> const& b)
+	{
+		return abs(b.x - a.x) + abs(b.y - a.y) + abs(b.z - a.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T l1Norm(vec<3, T, Q> const& v)
+	{
+		return abs(v.x) + abs(v.y) + abs(v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T l2Norm(vec<3, T, Q> const& a, vec<3, T, Q> const& b
+	)
+	{
+		return length(b - a);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T l2Norm(vec<3, T, Q> const& v)
+	{
+		return length(v);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T lxNorm(vec<3, T, Q> const& x, vec<3, T, Q> const& y, unsigned int Depth)
+	{
+		return pow(pow(abs(y.x - x.x), T(Depth)) + pow(abs(y.y - x.y), T(Depth)) + pow(abs(y.z - x.z), T(Depth)), T(1) / T(Depth));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T lxNorm(vec<3, T, Q> const& v, unsigned int Depth)
+	{
+		return pow(pow(abs(v.x), T(Depth)) + pow(abs(v.y), T(Depth)) + pow(abs(v.z), T(Depth)), T(1) / T(Depth));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T lMaxNorm(vec<3, T, Q> const& a, vec<3, T, Q> const& b)
+	{
+		return compMax(abs(b - a));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T lMaxNorm(vec<3, T, Q> const& v)
+	{
+		return compMax(abs(v));
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/normal.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/normal.hpp
new file mode 100644
index 000000000000..068682f75f2d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/normal.hpp
@@ -0,0 +1,41 @@
+/// @ref gtx_normal
+/// @file glm/gtx/normal.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_normal GLM_GTX_normal
+/// @ingroup gtx
+///
+/// Include <glm/gtx/normal.hpp> to use the features of this extension.
+///
+/// Compute the normal of a triangle.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_normal is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_normal extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_normal
+	/// @{
+
+	/// Computes triangle normal from triangle points.
+	///
+	/// @see gtx_normal
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> triangleNormal(vec<3, T, Q> const& p1, vec<3, T, Q> const& p2, vec<3, T, Q> const& p3);
+
+	/// @}
+}//namespace glm
+
+#include "normal.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/normal.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/normal.inl
new file mode 100644
index 000000000000..74f9fc994585
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/normal.inl
@@ -0,0 +1,15 @@
+/// @ref gtx_normal
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> triangleNormal
+	(
+		vec<3, T, Q> const& p1,
+		vec<3, T, Q> const& p2,
+		vec<3, T, Q> const& p3
+	)
+	{
+		return normalize(cross(p1 - p2, p1 - p3));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.hpp
new file mode 100644
index 000000000000..127aa1f65a85
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.hpp
@@ -0,0 +1,49 @@
+/// @ref gtx_normalize_dot
+/// @file glm/gtx/normalize_dot.hpp
+///
+/// @see core (dependence)
+/// @see gtx_fast_square_root (dependence)
+///
+/// @defgroup gtx_normalize_dot GLM_GTX_normalize_dot
+/// @ingroup gtx
+///
+/// Include <glm/gtx/normalize_dot.hpp> to use the features of this extension.
+///
+/// Dot product of vectors that need to be normalize with a single square root.
+
+#pragma once
+
+// Dependency:
+#include "../gtx/fast_square_root.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_normalize_dot is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_normalize_dot extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_normalize_dot
+	/// @{
+
+	/// Normalize parameters and returns the dot product of x and y.
+	/// It's faster that dot(normalize(x), normalize(y)).
+	///
+	/// @see gtx_normalize_dot extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T normalizeDot(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Normalize parameters and returns the dot product of x and y.
+	/// Faster that dot(fastNormalize(x), fastNormalize(y)).
+	///
+	/// @see gtx_normalize_dot extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T fastNormalizeDot(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// @}
+}//namespace glm
+
+#include "normalize_dot.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.inl
new file mode 100644
index 000000000000..7bcd9a534a8f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/normalize_dot.inl
@@ -0,0 +1,16 @@
+/// @ref gtx_normalize_dot
+
+namespace glm
+{
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T normalizeDot(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		return glm::dot(x, y) * glm::inversesqrt(glm::dot(x, x) * glm::dot(y, y));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T fastNormalizeDot(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		return glm::dot(x, y) * glm::fastInverseSqrt(glm::dot(x, x) * glm::dot(y, y));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/number_precision.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/number_precision.hpp
new file mode 100644
index 000000000000..3bf0ad68bdd0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/number_precision.hpp
@@ -0,0 +1,47 @@
+/// @ref gtx_number_precision
+/// @file glm/gtx/number_precision.hpp
+///
+/// @see core (dependence)
+/// @see gtc_type_precision (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtx_number_precision GLM_GTX_number_precision
+/// @ingroup gtx
+///
+/// Include <glm/gtx/number_precision.hpp> to use the features of this extension.
+///
+/// Defined size types.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/type_precision.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_number_precision is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_number_precision extension included")
+#	endif
+#endif
+
+namespace glm{
+	/////////////////////////////
+	// Unsigned int vector types
+
+	/// @addtogroup gtx_number_precision
+	/// @{
+
+	//////////////////////
+	// Float matrix types
+
+	typedef f32			f32mat1;	//!< \brief Single-qualifier floating-point scalar. (from GLM_GTX_number_precision extension)
+	typedef f32			f32mat1x1;	//!< \brief Single-qualifier floating-point scalar. (from GLM_GTX_number_precision extension)
+	typedef f64			f64mat1;	//!< \brief Double-qualifier floating-point scalar. (from GLM_GTX_number_precision extension)
+	typedef f64			f64mat1x1;	//!< \brief Double-qualifier floating-point scalar. (from GLM_GTX_number_precision extension)
+
+	/// @}
+}//namespace glm
+
+#include "number_precision.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.hpp
new file mode 100644
index 000000000000..a8ff6002f123
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.hpp
@@ -0,0 +1,52 @@
+/// @ref gtx_optimum_pow
+/// @file glm/gtx/optimum_pow.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_optimum_pow GLM_GTX_optimum_pow
+/// @ingroup gtx
+///
+/// Include <glm/gtx/optimum_pow.hpp> to use the features of this extension.
+///
+/// Integer exponentiation of power functions.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_optimum_pow is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_optimum_pow extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_optimum_pow
+	/// @{
+
+	/// Returns x raised to the power of 2.
+	///
+	/// @see gtx_optimum_pow
+	template<typename genType>
+	GLM_FUNC_DECL genType pow2(genType const& x);
+
+	/// Returns x raised to the power of 3.
+	///
+	/// @see gtx_optimum_pow
+	template<typename genType>
+	GLM_FUNC_DECL genType pow3(genType const& x);
+
+	/// Returns x raised to the power of 4.
+	///
+	/// @see gtx_optimum_pow
+	template<typename genType>
+	GLM_FUNC_DECL genType pow4(genType const& x);
+
+	/// @}
+}//namespace glm
+
+#include "optimum_pow.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.inl
new file mode 100644
index 000000000000..a26c19c18bfb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/optimum_pow.inl
@@ -0,0 +1,22 @@
+/// @ref gtx_optimum_pow
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType pow2(genType const& x)
+	{
+		return x * x;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType pow3(genType const& x)
+	{
+		return x * x * x;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType pow4(genType const& x)
+	{
+		return (x * x) * (x * x);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.hpp
new file mode 100644
index 000000000000..3e004fb06f9c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.hpp
@@ -0,0 +1,49 @@
+/// @ref gtx_orthonormalize
+/// @file glm/gtx/orthonormalize.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_orthonormalize GLM_GTX_orthonormalize
+/// @ingroup gtx
+///
+/// Include <glm/gtx/orthonormalize.hpp> to use the features of this extension.
+///
+/// Orthonormalize matrices.
+
+#pragma once
+
+// Dependency:
+#include "../vec3.hpp"
+#include "../mat3x3.hpp"
+#include "../geometric.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_orthonormalize is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_orthonormalize extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_orthonormalize
+	/// @{
+
+	/// Returns the orthonormalized matrix of m.
+	///
+	/// @see gtx_orthonormalize
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> orthonormalize(mat<3, 3, T, Q> const& m);
+
+	/// Orthonormalizes x according y.
+	///
+	/// @see gtx_orthonormalize
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> orthonormalize(vec<3, T, Q> const& x, vec<3, T, Q> const& y);
+
+	/// @}
+}//namespace glm
+
+#include "orthonormalize.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.inl
new file mode 100644
index 000000000000..cb553ba62157
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/orthonormalize.inl
@@ -0,0 +1,29 @@
+/// @ref gtx_orthonormalize
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> orthonormalize(mat<3, 3, T, Q> const& m)
+	{
+		mat<3, 3, T, Q> r = m;
+
+		r[0] = normalize(r[0]);
+
+		T d0 = dot(r[0], r[1]);
+		r[1] -= r[0] * d0;
+		r[1] = normalize(r[1]);
+
+		T d1 = dot(r[1], r[2]);
+		d0 = dot(r[0], r[2]);
+		r[2] -= r[0] * d0 + r[1] * d1;
+		r[2] = normalize(r[2]);
+
+		return r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> orthonormalize(vec<3, T, Q> const& x, vec<3, T, Q> const& y)
+	{
+		return normalize(x - y * dot(y, x));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/pca.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/pca.hpp
new file mode 100644
index 000000000000..d89c408e36ef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/pca.hpp
@@ -0,0 +1,115 @@
+/// @ref gtx_pca
+/// @file glm/gtx/pca.hpp
+///
+/// @see core (dependence)
+/// @see ext_scalar_relational (dependence)
+///
+/// @defgroup gtx_pca GLM_GTX_pca
+/// @ingroup gtx
+///
+/// Include <glm/gtx/pca.hpp> to use the features of this extension.
+///
+/// Implements functions required for fundamental 'princple component analysis' in 2D, 3D, and 4D:
+///   1) Computing a covariance matrics from a list of _relative_ position vectors
+///   2) Compute the eigenvalues and eigenvectors of the covariance matrics
+/// This is useful, e.g., to compute an object-aligned bounding box from vertices of an object.
+/// https://en.wikipedia.org/wiki/Principal_component_analysis
+///
+/// Example:
+/// ```
+/// std::vector<glm::dvec3> ptData;
+/// // ... fill ptData with some point data, e.g. vertices
+/// 
+/// glm::dvec3 center = computeCenter(ptData);
+/// 
+/// glm::dmat3 covarMat = glm::computeCovarianceMatrix(ptData.data(), ptData.size(), center);
+/// 
+/// glm::dvec3 evals;
+/// glm::dmat3 evecs;
+/// int evcnt = glm::findEigenvaluesSymReal(covarMat, evals, evecs);
+/// 
+/// if(evcnt != 3)
+///     // ... error handling
+/// 
+/// glm::sortEigenvalues(evals, evecs);
+/// 
+/// // ... now evecs[0] points in the direction (symmetric) of the largest spatial distribution within ptData
+/// ```
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../ext/scalar_relational.hpp"
+
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_pca is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_pca extension included")
+#	endif
+#endif
+
+namespace glm {
+	/// @addtogroup gtx_pca
+	/// @{
+
+	/// Compute a covariance matrix form an array of relative coordinates `v` (e.g., relative to the center of gravity of the object)
+	/// @param v Points to a memory holding `n` times vectors
+	/// @param n Number of points in v
+	template<length_t D, typename T, qualifier Q>
+	GLM_INLINE mat<D, D, T, Q> computeCovarianceMatrix(vec<D, T, Q> const* v, size_t n);
+
+	/// Compute a covariance matrix form an array of absolute coordinates `v` and a precomputed center of gravity `c`
+	/// @param v Points to a memory holding `n` times vectors
+	/// @param n Number of points in v
+	/// @param c Precomputed center of gravity
+	template<length_t D, typename T, qualifier Q>
+	GLM_INLINE mat<D, D, T, Q> computeCovarianceMatrix(vec<D, T, Q> const* v, size_t n, vec<D, T, Q> const& c);
+
+	/// Compute a covariance matrix form a pair of iterators `b` (begin) and `e` (end) of a container with relative coordinates (e.g., relative to the center of gravity of the object)
+	/// Dereferencing an iterator of type I must yield a `vec&lt;D, T, Q%gt;`
+	template<length_t D, typename T, qualifier Q, typename I>
+	GLM_FUNC_DECL mat<D, D, T, Q> computeCovarianceMatrix(I const& b, I const& e);
+
+	/// Compute a covariance matrix form a pair of iterators `b` (begin) and `e` (end) of a container with absolute coordinates and a precomputed center of gravity `c`
+	/// Dereferencing an iterator of type I must yield a `vec&lt;D, T, Q%gt;`
+	template<length_t D, typename T, qualifier Q, typename I>
+	GLM_FUNC_DECL mat<D, D, T, Q> computeCovarianceMatrix(I const& b, I const& e, vec<D, T, Q> const& c);
+
+	/// Assuming the provided covariance matrix `covarMat` is symmetric and real-valued, this function find the `D` Eigenvalues of the matrix, and also provides the corresponding Eigenvectors.
+	/// Note: the data in `outEigenvalues` and `outEigenvectors` are in matching order, i.e. `outEigenvector[i]` is the Eigenvector of the Eigenvalue `outEigenvalue[i]`.
+	/// This is a numeric implementation to find the Eigenvalues, using 'QL decomposition` (variant of QR decomposition: https://en.wikipedia.org/wiki/QR_decomposition).
+	///
+	/// @param[in] covarMat A symmetric, real-valued covariance matrix, e.g. computed from computeCovarianceMatrix
+	/// @param[out] outEigenvalues Vector to receive the found eigenvalues
+	/// @param[out] outEigenvectors Matrix to receive the found eigenvectors corresponding to the found eigenvalues, as column vectors
+	/// @return The number of eigenvalues found, usually D if the precondition of the covariance matrix is met.
+	template<length_t D, typename T, qualifier Q>
+	GLM_FUNC_DECL unsigned int findEigenvaluesSymReal
+	(
+		mat<D, D, T, Q> const& covarMat,
+		vec<D, T, Q>& outEigenvalues,
+		mat<D, D, T, Q>& outEigenvectors
+	);
+
+	/// Sorts a group of Eigenvalues&Eigenvectors, for largest Eigenvalue to smallest Eigenvalue.
+	/// The data in `outEigenvalues` and `outEigenvectors` are assumed to be matching order, i.e. `outEigenvector[i]` is the Eigenvector of the Eigenvalue `outEigenvalue[i]`.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL void sortEigenvalues(vec<2, T, Q>& eigenvalues, mat<2, 2, T, Q>& eigenvectors);
+
+	/// Sorts a group of Eigenvalues&Eigenvectors, for largest Eigenvalue to smallest Eigenvalue.
+	/// The data in `outEigenvalues` and `outEigenvectors` are assumed to be matching order, i.e. `outEigenvector[i]` is the Eigenvector of the Eigenvalue `outEigenvalue[i]`.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL void sortEigenvalues(vec<3, T, Q>& eigenvalues, mat<3, 3, T, Q>& eigenvectors);
+
+	/// Sorts a group of Eigenvalues&Eigenvectors, for largest Eigenvalue to smallest Eigenvalue.
+	/// The data in `outEigenvalues` and `outEigenvectors` are assumed to be matching order, i.e. `outEigenvector[i]` is the Eigenvector of the Eigenvalue `outEigenvalue[i]`.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL void sortEigenvalues(vec<4, T, Q>& eigenvalues, mat<4, 4, T, Q>& eigenvectors);
+
+	/// @}
+}//namespace glm
+
+#include "pca.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/pca.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/pca.inl
new file mode 100644
index 000000000000..94cae946e8a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/pca.inl
@@ -0,0 +1,343 @@
+/// @ref gtx_pca
+
+#ifndef GLM_HAS_CXX11_STL
+#include <algorithm>
+#else
+#include <utility>
+#endif
+
+namespace glm {
+
+
+	template<length_t D, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<D, D, T, Q> computeCovarianceMatrix(vec<D, T, Q> const* v, size_t n)
+	{
+		return computeCovarianceMatrix<D, T, Q, vec<D, T, Q> const*>(v, v + n);
+	}
+
+
+	template<length_t D, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<D, D, T, Q> computeCovarianceMatrix(vec<D, T, Q> const* v, size_t n, vec<D, T, Q> const& c)
+	{
+		return computeCovarianceMatrix<D, T, Q, vec<D, T, Q> const*>(v, v + n, c);
+	}
+
+
+	template<length_t D, typename T, qualifier Q, typename I>
+	GLM_FUNC_QUALIFIER mat<D, D, T, Q> computeCovarianceMatrix(I const& b, I const& e)
+	{
+		glm::mat<D, D, T, Q> m(0);
+
+		size_t cnt = 0;
+		for(I i = b; i != e; i++)
+		{
+			vec<D, T, Q> const& v = *i;
+			for(length_t x = 0; x < D; ++x)
+				for(length_t y = 0; y < D; ++y)
+					m[x][y] += static_cast<T>(v[x] * v[y]);
+			cnt++;
+		}
+		if(cnt > 0)
+			m /= static_cast<T>(cnt);
+
+		return m;
+	}
+
+
+	template<length_t D, typename T, qualifier Q, typename I>
+	GLM_FUNC_QUALIFIER mat<D, D, T, Q> computeCovarianceMatrix(I const& b, I const& e, vec<D, T, Q> const& c)
+	{
+		glm::mat<D, D, T, Q> m(0);
+		glm::vec<D, T, Q> v;
+
+		size_t cnt = 0;
+		for(I i = b; i != e; i++)
+		{
+			v = *i - c;
+			for(length_t x = 0; x < D; ++x)
+				for(length_t y = 0; y < D; ++y)
+					m[x][y] += static_cast<T>(v[x] * v[y]);
+			cnt++;
+		}
+		if(cnt > 0)
+			m /= static_cast<T>(cnt);
+
+		return m;
+	}
+
+	namespace _internal_
+	{
+
+		template<typename T>
+		GLM_FUNC_QUALIFIER static T transferSign(T const& v, T const& s)
+		{
+			return ((s) >= 0 ? glm::abs(v) : -glm::abs(v));
+		}
+
+		template<typename T>
+		GLM_FUNC_QUALIFIER static T pythag(T const& a, T const& b) {
+			static const T epsilon = static_cast<T>(0.0000001);
+			T absa = glm::abs(a);
+			T absb = glm::abs(b);
+			if(absa > absb) {
+				absb /= absa;
+				absb *= absb;
+				return absa * glm::sqrt(static_cast<T>(1) + absb);
+			}
+			if(glm::equal<T>(absb, 0, epsilon)) return static_cast<T>(0);
+			absa /= absb;
+			absa *= absa;
+			return absb * glm::sqrt(static_cast<T>(1) + absa);
+		}
+
+	}
+
+	template<length_t D, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER unsigned int findEigenvaluesSymReal
+	(
+		mat<D, D, T, Q> const& covarMat,
+		vec<D, T, Q>& outEigenvalues,
+		mat<D, D, T, Q>& outEigenvectors
+	)
+	{
+		using _internal_::transferSign;
+		using _internal_::pythag;
+
+		T a[D * D]; // matrix -- input and workspace for algorithm (will be changed inplace)
+		T d[D]; // diagonal elements
+		T e[D]; // off-diagonal elements
+
+		for(length_t r = 0; r < D; r++)
+			for(length_t c = 0; c < D; c++)
+				a[(r) * D + (c)] = covarMat[c][r];
+
+		// 1. Householder reduction.
+		length_t l, k, j, i;
+		T scale, hh, h, g, f;
+		static const T epsilon = static_cast<T>(0.0000001);
+
+		for(i = D; i >= 2; i--)
+		{
+			l = i - 1;
+			h = scale = 0;
+			if(l > 1)
+			{
+				for(k = 1; k <= l; k++)
+				{
+					scale += glm::abs(a[(i - 1) * D + (k - 1)]);
+				}
+				if(glm::equal<T>(scale, 0, epsilon))
+				{
+					e[i - 1] = a[(i - 1) * D + (l - 1)];
+				}
+				else
+				{
+					for(k = 1; k <= l; k++)
+					{
+						a[(i - 1) * D + (k - 1)] /= scale;
+						h += a[(i - 1) * D + (k - 1)] * a[(i - 1) * D + (k - 1)];
+					}
+					f = a[(i - 1) * D + (l - 1)];
+					g = ((f >= 0) ? -glm::sqrt(h) : glm::sqrt(h));
+					e[i - 1] = scale * g;
+					h -= f * g;
+					a[(i - 1) * D + (l - 1)] = f - g;
+					f = 0;
+					for(j = 1; j <= l; j++)
+					{
+						a[(j - 1) * D + (i - 1)] = a[(i - 1) * D + (j - 1)] / h;
+						g = 0;
+						for(k = 1; k <= j; k++)
+						{
+							g += a[(j - 1) * D + (k - 1)] * a[(i - 1) * D + (k - 1)];
+						}
+						for(k = j + 1; k <= l; k++)
+						{
+							g += a[(k - 1) * D + (j - 1)] * a[(i - 1) * D + (k - 1)];
+						}
+						e[j - 1] = g / h;
+						f += e[j - 1] * a[(i - 1) * D + (j - 1)];
+					}
+					hh = f / (h + h);
+					for(j = 1; j <= l; j++)
+					{
+						f = a[(i - 1) * D + (j - 1)];
+						e[j - 1] = g = e[j - 1] - hh * f;
+						for(k = 1; k <= j; k++)
+						{
+							a[(j - 1) * D + (k - 1)] -= (f * e[k - 1] + g * a[(i - 1) * D + (k - 1)]);
+						}
+					}
+				}
+			}
+			else
+			{
+				e[i - 1] = a[(i - 1) * D + (l - 1)];
+			}
+			d[i - 1] = h;
+		}
+		d[0] = 0;
+		e[0] = 0;
+		for(i = 1; i <= D; i++)
+		{
+			l = i - 1;
+			if(!glm::equal<T>(d[i - 1], 0, epsilon))
+			{
+				for(j = 1; j <= l; j++)
+				{
+					g = 0;
+					for(k = 1; k <= l; k++)
+					{
+						g += a[(i - 1) * D + (k - 1)] * a[(k - 1) * D + (j - 1)];
+					}
+					for(k = 1; k <= l; k++)
+					{
+						a[(k - 1) * D + (j - 1)] -= g * a[(k - 1) * D + (i - 1)];
+					}
+				}
+			}
+			d[i - 1] = a[(i - 1) * D + (i - 1)];
+			a[(i - 1) * D + (i - 1)] = 1;
+			for(j = 1; j <= l; j++)
+			{
+				a[(j - 1) * D + (i - 1)] = a[(i - 1) * D + (j - 1)] = 0;
+			}
+		}
+
+		// 2. Calculation of eigenvalues and eigenvectors (QL algorithm)
+		length_t m, iter;
+		T s, r, p, dd, c, b;
+		const length_t MAX_ITER = 30;
+
+		for(i = 2; i <= D; i++)
+		{
+			e[i - 2] = e[i - 1];
+		}
+		e[D - 1] = 0;
+
+		for(l = 1; l <= D; l++)
+		{
+			iter = 0;
+			do
+			{
+				for(m = l; m <= D - 1; m++)
+				{
+					dd = glm::abs(d[m - 1]) + glm::abs(d[m - 1 + 1]);
+					if(glm::equal<T>(glm::abs(e[m - 1]) + dd, dd, epsilon))
+						break;
+				}
+				if(m != l)
+				{
+					if(iter++ == MAX_ITER)
+					{
+						return 0; // Too many iterations in FindEigenvalues
+					}
+					g = (d[l - 1 + 1] - d[l - 1]) / (2 * e[l - 1]);
+					r = pythag<T>(g, 1);
+					g = d[m - 1] - d[l - 1] + e[l - 1] / (g + transferSign(r, g));
+					s = c = 1;
+					p = 0;
+					for(i = m - 1; i >= l; i--)
+					{
+						f = s * e[i - 1];
+						b = c * e[i - 1];
+						e[i - 1 + 1] = r = pythag(f, g);
+						if(glm::equal<T>(r, 0, epsilon))
+						{
+							d[i - 1 + 1] -= p;
+							e[m - 1] = 0;
+							break;
+						}
+						s = f / r;
+						c = g / r;
+						g = d[i - 1 + 1] - p;
+						r = (d[i - 1] - g) * s + 2 * c * b;
+						d[i - 1 + 1] = g + (p = s * r);
+						g = c * r - b;
+						for(k = 1; k <= D; k++)
+						{
+							f = a[(k - 1) * D + (i - 1 + 1)];
+							a[(k - 1) * D + (i - 1 + 1)] = s * a[(k - 1) * D + (i - 1)] + c * f;
+							a[(k - 1) * D + (i - 1)] = c * a[(k - 1) * D + (i - 1)] - s * f;
+						}
+					}
+					if(glm::equal<T>(r, 0, epsilon) && (i >= l))
+						continue;
+					d[l - 1] -= p;
+					e[l - 1] = g;
+					e[m - 1] = 0;
+				}
+			} while(m != l);
+		}
+
+		// 3. output
+		for(i = 0; i < D; i++)
+			outEigenvalues[i] = d[i];
+		for(i = 0; i < D; i++)
+			for(j = 0; j < D; j++)
+				outEigenvectors[i][j] = a[(j) * D + (i)];
+
+		return D;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER void sortEigenvalues(vec<2, T, Q>& eigenvalues, mat<2, 2, T, Q>& eigenvectors)
+	{
+		if (eigenvalues[0] < eigenvalues[1])
+		{
+			std::swap(eigenvalues[0], eigenvalues[1]);
+			std::swap(eigenvectors[0], eigenvectors[1]);
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER void sortEigenvalues(vec<3, T, Q>& eigenvalues, mat<3, 3, T, Q>& eigenvectors)
+	{
+		if (eigenvalues[0] < eigenvalues[1])
+		{
+			std::swap(eigenvalues[0], eigenvalues[1]);
+			std::swap(eigenvectors[0], eigenvectors[1]);
+		}
+		if (eigenvalues[0] < eigenvalues[2])
+		{
+			std::swap(eigenvalues[0], eigenvalues[2]);
+			std::swap(eigenvectors[0], eigenvectors[2]);
+		}
+		if (eigenvalues[1] < eigenvalues[2])
+		{
+			std::swap(eigenvalues[1], eigenvalues[2]);
+			std::swap(eigenvectors[1], eigenvectors[2]);
+		}
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER void sortEigenvalues(vec<4, T, Q>& eigenvalues, mat<4, 4, T, Q>& eigenvectors)
+	{
+		if (eigenvalues[0] < eigenvalues[2])
+		{
+			std::swap(eigenvalues[0], eigenvalues[2]);
+			std::swap(eigenvectors[0], eigenvectors[2]);
+		}
+		if (eigenvalues[1] < eigenvalues[3])
+		{
+			std::swap(eigenvalues[1], eigenvalues[3]);
+			std::swap(eigenvectors[1], eigenvectors[3]);
+		}
+		if (eigenvalues[0] < eigenvalues[1])
+		{
+			std::swap(eigenvalues[0], eigenvalues[1]);
+			std::swap(eigenvectors[0], eigenvectors[1]);
+		}
+		if (eigenvalues[2] < eigenvalues[3])
+		{
+			std::swap(eigenvalues[2], eigenvalues[3]);
+			std::swap(eigenvectors[2], eigenvectors[3]);
+		}
+		if (eigenvalues[1] < eigenvalues[2])
+		{
+			std::swap(eigenvalues[1], eigenvalues[2]);
+			std::swap(eigenvectors[1], eigenvectors[2]);
+		}
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.hpp
new file mode 100644
index 000000000000..72b77b6e2388
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.hpp
@@ -0,0 +1,41 @@
+/// @ref gtx_perpendicular
+/// @file glm/gtx/perpendicular.hpp
+///
+/// @see core (dependence)
+/// @see gtx_projection (dependence)
+///
+/// @defgroup gtx_perpendicular GLM_GTX_perpendicular
+/// @ingroup gtx
+///
+/// Include <glm/gtx/perpendicular.hpp> to use the features of this extension.
+///
+/// Perpendicular of a vector from other one
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtx/projection.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_perpendicular is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_perpendicular extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_perpendicular
+	/// @{
+
+	//! Projects x a perpendicular axis of Normal.
+	//! From GLM_GTX_perpendicular extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType perp(genType const& x, genType const& Normal);
+
+	/// @}
+}//namespace glm
+
+#include "perpendicular.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.inl
new file mode 100644
index 000000000000..1e72f334230d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/perpendicular.inl
@@ -0,0 +1,10 @@
+/// @ref gtx_perpendicular
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType perp(genType const& x, genType const& Normal)
+	{
+		return x - proj(x, Normal);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.hpp
new file mode 100644
index 000000000000..76beb82bd57c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.hpp
@@ -0,0 +1,48 @@
+/// @ref gtx_polar_coordinates
+/// @file glm/gtx/polar_coordinates.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_polar_coordinates GLM_GTX_polar_coordinates
+/// @ingroup gtx
+///
+/// Include <glm/gtx/polar_coordinates.hpp> to use the features of this extension.
+///
+/// Conversion from Euclidean space to polar space and revert.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_polar_coordinates is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_polar_coordinates extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_polar_coordinates
+	/// @{
+
+	/// Convert Euclidean to Polar coordinates, x is the latitude, y the longitude and z the xz distance.
+	///
+	/// @see gtx_polar_coordinates
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> polar(
+		vec<3, T, Q> const& euclidean);
+
+	/// Convert Polar to Euclidean coordinates.
+	///
+	/// @see gtx_polar_coordinates
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> euclidean(
+		vec<2, T, Q> const& polar);
+
+	/// @}
+}//namespace glm
+
+#include "polar_coordinates.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.inl
new file mode 100644
index 000000000000..371c8dddebd1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/polar_coordinates.inl
@@ -0,0 +1,36 @@
+/// @ref gtx_polar_coordinates
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> polar
+	(
+		vec<3, T, Q> const& euclidean
+	)
+	{
+		T const Length(length(euclidean));
+		vec<3, T, Q> const tmp(euclidean / Length);
+		T const xz_dist(sqrt(tmp.x * tmp.x + tmp.z * tmp.z));
+
+		return vec<3, T, Q>(
+			asin(tmp.y),	// latitude
+			atan(tmp.x, tmp.z),		// longitude
+			xz_dist);				// xz distance
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> euclidean
+	(
+		vec<2, T, Q> const& polar
+	)
+	{
+		T const latitude(polar.x);
+		T const longitude(polar.y);
+
+		return vec<3, T, Q>(
+			cos(latitude) * sin(longitude),
+			sin(latitude),
+			cos(latitude) * cos(longitude));
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/projection.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/projection.hpp
new file mode 100644
index 000000000000..678f3ad5a585
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/projection.hpp
@@ -0,0 +1,43 @@
+/// @ref gtx_projection
+/// @file glm/gtx/projection.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_projection GLM_GTX_projection
+/// @ingroup gtx
+///
+/// Include <glm/gtx/projection.hpp> to use the features of this extension.
+///
+/// Projection of a vector to other one
+
+#pragma once
+
+// Dependency:
+#include "../geometric.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_projection is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_projection extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_projection
+	/// @{
+
+	/// Projects x on Normal.
+	///
+	/// @param[in] x A vector to project
+	/// @param[in] Normal A normal that doesn't need to be of unit length.
+	///
+	/// @see gtx_projection
+	template<typename genType>
+	GLM_FUNC_DECL genType proj(genType const& x, genType const& Normal);
+
+	/// @}
+}//namespace glm
+
+#include "projection.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/projection.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/projection.inl
new file mode 100644
index 000000000000..f23f884fb93a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/projection.inl
@@ -0,0 +1,10 @@
+/// @ref gtx_projection
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType proj(genType const& x, genType const& Normal)
+	{
+		return glm::dot(x, Normal) / glm::dot(Normal, Normal) * Normal;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.hpp
new file mode 100644
index 000000000000..35c372b816e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.hpp
@@ -0,0 +1,174 @@
+/// @ref gtx_quaternion
+/// @file glm/gtx/quaternion.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_quaternion GLM_GTX_quaternion
+/// @ingroup gtx
+///
+/// Include <glm/gtx/quaternion.hpp> to use the features of this extension.
+///
+/// Extended quaternion types and functions
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/constants.hpp"
+#include "../gtc/quaternion.hpp"
+#include "../ext/quaternion_exponential.hpp"
+#include "../gtx/norm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_quaternion is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_quaternion extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_quaternion
+	/// @{
+
+	/// Create an identity quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR qua<T, Q> quat_identity();
+
+	/// Compute a cross product between a quaternion and a vector.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> cross(
+		qua<T, Q> const& q,
+		vec<3, T, Q> const& v);
+
+	//! Compute a cross product between a vector and a quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<3, T, Q> cross(
+		vec<3, T, Q> const& v,
+		qua<T, Q> const& q);
+
+	//! Compute a point on a path according squad equation.
+	//! q1 and q2 are control points; s1 and s2 are intermediate control points.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> squad(
+		qua<T, Q> const& q1,
+		qua<T, Q> const& q2,
+		qua<T, Q> const& s1,
+		qua<T, Q> const& s2,
+		T const& h);
+
+	//! Returns an intermediate control point for squad interpolation.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> intermediate(
+		qua<T, Q> const& prev,
+		qua<T, Q> const& curr,
+		qua<T, Q> const& next);
+
+	//! Returns quarternion square root.
+	///
+	/// @see gtx_quaternion
+	//template<typename T, qualifier Q>
+	//qua<T, Q> sqrt(
+	//	qua<T, Q> const& q);
+
+	//! Rotates a 3 components vector by a quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rotate(
+		qua<T, Q> const& q,
+		vec<3, T, Q> const& v);
+
+	/// Rotates a 4 components vector by a quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> rotate(
+		qua<T, Q> const& q,
+		vec<4, T, Q> const& v);
+
+	/// Extract the real component of a quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T extractRealComponent(
+		qua<T, Q> const& q);
+
+	/// Converts a quaternion to a 3 * 3 matrix.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> toMat3(
+		qua<T, Q> const& x){return mat3_cast(x);}
+
+	/// Converts a quaternion to a 4 * 4 matrix.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> toMat4(
+		qua<T, Q> const& x){return mat4_cast(x);}
+
+	/// Converts a 3 * 3 matrix to a quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> toQuat(
+		mat<3, 3, T, Q> const& x){return quat_cast(x);}
+
+	/// Converts a 4 * 4 matrix to a quaternion.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> toQuat(
+		mat<4, 4, T, Q> const& x){return quat_cast(x);}
+
+	/// Quaternion interpolation using the rotation short path.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> shortMix(
+		qua<T, Q> const& x,
+		qua<T, Q> const& y,
+		T const& a);
+
+	/// Quaternion normalized linear interpolation.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> fastMix(
+		qua<T, Q> const& x,
+		qua<T, Q> const& y,
+		T const& a);
+
+	/// Compute the rotation between two vectors.
+	/// @param orig vector, needs to be normalized
+	/// @param dest vector, needs to be normalized
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> rotation(
+		vec<3, T, Q> const& orig,
+		vec<3, T, Q> const& dest);
+
+	/// Returns the squared length of x.
+	///
+	/// @see gtx_quaternion
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR T length2(qua<T, Q> const& q);
+
+	/// @}
+}//namespace glm
+
+#include "quaternion.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.inl
new file mode 100644
index 000000000000..5e18899a714a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/quaternion.inl
@@ -0,0 +1,159 @@
+/// @ref gtx_quaternion
+
+#include <limits>
+#include "../gtc/constants.hpp"
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR qua<T, Q> quat_identity()
+	{
+		return qua<T, Q>::wxyz(static_cast<T>(1), static_cast<T>(0), static_cast<T>(0), static_cast<T>(0));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> cross(vec<3, T, Q> const& v, qua<T, Q> const& q)
+	{
+		return inverse(q) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR vec<3, T, Q> cross(qua<T, Q> const& q, vec<3, T, Q> const& v)
+	{
+		return q * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> squad
+	(
+		qua<T, Q> const& q1,
+		qua<T, Q> const& q2,
+		qua<T, Q> const& s1,
+		qua<T, Q> const& s2,
+		T const& h)
+	{
+		return mix(mix(q1, q2, h), mix(s1, s2, h), static_cast<T>(2) * (static_cast<T>(1) - h) * h);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> intermediate
+	(
+		qua<T, Q> const& prev,
+		qua<T, Q> const& curr,
+		qua<T, Q> const& next
+	)
+	{
+		qua<T, Q> invQuat = inverse(curr);
+		return exp((log(next * invQuat) + log(prev * invQuat)) / static_cast<T>(-4)) * curr;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rotate(qua<T, Q> const& q, vec<3, T, Q> const& v)
+	{
+		return q * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> rotate(qua<T, Q> const& q, vec<4, T, Q> const& v)
+	{
+		return q * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T extractRealComponent(qua<T, Q> const& q)
+	{
+		T w = static_cast<T>(1) - q.x * q.x - q.y * q.y - q.z * q.z;
+		if(w < T(0))
+			return T(0);
+		else
+			return -sqrt(w);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER GLM_CONSTEXPR T length2(qua<T, Q> const& q)
+	{
+		return q.x * q.x + q.y * q.y + q.z * q.z + q.w * q.w;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> shortMix(qua<T, Q> const& x, qua<T, Q> const& y, T const& a)
+	{
+		if(a <= static_cast<T>(0)) return x;
+		if(a >= static_cast<T>(1)) return y;
+
+		T fCos = dot(x, y);
+		qua<T, Q> y2(y); //BUG!!! qua<T> y2;
+		if(fCos < static_cast<T>(0))
+		{
+			y2 = -y;
+			fCos = -fCos;
+		}
+
+		//if(fCos > 1.0f) // problem
+		T k0, k1;
+		if(fCos > (static_cast<T>(1) - epsilon<T>()))
+		{
+			k0 = static_cast<T>(1) - a;
+			k1 = static_cast<T>(0) + a; //BUG!!! 1.0f + a;
+		}
+		else
+		{
+			T fSin = sqrt(T(1) - fCos * fCos);
+			T fAngle = atan(fSin, fCos);
+			T fOneOverSin = static_cast<T>(1) / fSin;
+			k0 = sin((static_cast<T>(1) - a) * fAngle) * fOneOverSin;
+			k1 = sin((static_cast<T>(0) + a) * fAngle) * fOneOverSin;
+		}
+
+		return qua<T, Q>::wxyz(
+			k0 * x.w + k1 * y2.w,
+			k0 * x.x + k1 * y2.x,
+			k0 * x.y + k1 * y2.y,
+			k0 * x.z + k1 * y2.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> fastMix(qua<T, Q> const& x, qua<T, Q> const& y, T const& a)
+	{
+		return glm::normalize(x * (static_cast<T>(1) - a) + (y * a));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> rotation(vec<3, T, Q> const& orig, vec<3, T, Q> const& dest)
+	{
+		T cosTheta = dot(orig, dest);
+		vec<3, T, Q> rotationAxis;
+
+		if(cosTheta >= static_cast<T>(1) - epsilon<T>()) {
+			// orig and dest point in the same direction
+			return quat_identity<T,Q>();
+		}
+
+		if(cosTheta < static_cast<T>(-1) + epsilon<T>())
+		{
+			// special case when vectors in opposite directions :
+			// there is no "ideal" rotation axis
+			// So guess one; any will do as long as it's perpendicular to start
+			// This implementation favors a rotation around the Up axis (Y),
+			// since it's often what you want to do.
+			rotationAxis = cross(vec<3, T, Q>(0, 0, 1), orig);
+			if(length2(rotationAxis) < epsilon<T>()) // bad luck, they were parallel, try again!
+				rotationAxis = cross(vec<3, T, Q>(1, 0, 0), orig);
+
+			rotationAxis = normalize(rotationAxis);
+			return angleAxis(pi<T>(), rotationAxis);
+		}
+
+		// Implementation from Stan Melax's Game Programming Gems 1 article
+		rotationAxis = cross(orig, dest);
+
+		T s = sqrt((T(1) + cosTheta) * static_cast<T>(2));
+		T invs = static_cast<T>(1) / s;
+
+		return qua<T, Q>::wxyz(
+			s * static_cast<T>(0.5f),
+			rotationAxis.x * invs,
+			rotationAxis.y * invs,
+			rotationAxis.z * invs);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/range.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/range.hpp
new file mode 100644
index 000000000000..93bcb9a65a0a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/range.hpp
@@ -0,0 +1,98 @@
+/// @ref gtx_range
+/// @file glm/gtx/range.hpp
+/// @author Joshua Moerman
+///
+/// @defgroup gtx_range GLM_GTX_range
+/// @ingroup gtx
+///
+/// Include <glm/gtx/range.hpp> to use the features of this extension.
+///
+/// Defines begin and end for vectors and matrices. Useful for range-based for loop.
+/// The range is defined over the elements, not over columns or rows (e.g. mat4 has 16 elements).
+
+#pragma once
+
+// Dependencies
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_range is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_range extension included")
+#	endif
+#endif
+
+#include "../gtc/type_ptr.hpp"
+#include "../gtc/vec1.hpp"
+
+namespace glm
+{
+	/// @addtogroup gtx_range
+	/// @{
+
+#	if GLM_COMPILER & GLM_COMPILER_VC
+#		pragma warning(push)
+#		pragma warning(disable : 4100) // unreferenced formal parameter
+#	endif
+
+	template<typename T, qualifier Q>
+	inline length_t components(vec<1, T, Q> const& v)
+	{
+		return v.length();
+	}
+
+	template<typename T, qualifier Q>
+	inline length_t components(vec<2, T, Q> const& v)
+	{
+		return v.length();
+	}
+
+	template<typename T, qualifier Q>
+	inline length_t components(vec<3, T, Q> const& v)
+	{
+		return v.length();
+	}
+
+	template<typename T, qualifier Q>
+	inline length_t components(vec<4, T, Q> const& v)
+	{
+		return v.length();
+	}
+
+	template<typename genType>
+	inline length_t components(genType const& m)
+	{
+		return m.length() * m[0].length();
+	}
+
+	template<typename genType>
+	inline typename genType::value_type const * begin(genType const& v)
+	{
+		return value_ptr(v);
+	}
+
+	template<typename genType>
+	inline typename genType::value_type const * end(genType const& v)
+	{
+		return begin(v) + components(v);
+	}
+
+	template<typename genType>
+	inline typename genType::value_type * begin(genType& v)
+	{
+		return value_ptr(v);
+	}
+
+	template<typename genType>
+	inline typename genType::value_type * end(genType& v)
+	{
+		return begin(v) + components(v);
+	}
+
+#	if GLM_COMPILER & GLM_COMPILER_VC
+#		pragma warning(pop)
+#	endif
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.hpp
new file mode 100644
index 000000000000..86cbe77d9ae5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.hpp
@@ -0,0 +1,51 @@
+/// @ref gtx_raw_data
+/// @file glm/gtx/raw_data.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_raw_data GLM_GTX_raw_data
+/// @ingroup gtx
+///
+/// Include <glm/gtx/raw_data.hpp> to use the features of this extension.
+///
+/// Projection of a vector to other one
+
+#pragma once
+
+// Dependencies
+#include "../ext/scalar_uint_sized.hpp"
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_raw_data is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_raw_data extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_raw_data
+	/// @{
+
+	//! Type for byte numbers.
+	//! From GLM_GTX_raw_data extension.
+	typedef detail::uint8		byte;
+
+	//! Type for word numbers.
+	//! From GLM_GTX_raw_data extension.
+	typedef detail::uint16		word;
+
+	//! Type for dword numbers.
+	//! From GLM_GTX_raw_data extension.
+	typedef detail::uint32		dword;
+
+	//! Type for qword numbers.
+	//! From GLM_GTX_raw_data extension.
+	typedef detail::uint64		qword;
+
+	/// @}
+}// namespace glm
+
+#include "raw_data.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.inl
new file mode 100644
index 000000000000..c740317d334e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/raw_data.inl
@@ -0,0 +1,2 @@
+/// @ref gtx_raw_data
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.hpp
new file mode 100644
index 000000000000..2103ca08f15e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.hpp
@@ -0,0 +1,68 @@
+/// @ref gtx_rotate_normalized_axis
+/// @file glm/gtx/rotate_normalized_axis.hpp
+///
+/// @see core (dependence)
+/// @see gtc_matrix_transform
+/// @see gtc_quaternion
+///
+/// @defgroup gtx_rotate_normalized_axis GLM_GTX_rotate_normalized_axis
+/// @ingroup gtx
+///
+/// Include <glm/gtx/rotate_normalized_axis.hpp> to use the features of this extension.
+///
+/// Quaternions and matrices rotations around normalized axis.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/epsilon.hpp"
+#include "../gtc/quaternion.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_rotate_normalized_axis is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_rotate_normalized_axis extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_rotate_normalized_axis
+	/// @{
+
+	/// Builds a rotation 4 * 4 matrix created from a normalized axis and an angle.
+	///
+	/// @param m Input matrix multiplied by this rotation matrix.
+	/// @param angle Rotation angle expressed in radians.
+	/// @param axis Rotation axis, must be normalized.
+	/// @tparam T Value type used to build the matrix. Currently supported: half (not recommended), float or double.
+	///
+	/// @see gtx_rotate_normalized_axis
+	/// @see - rotate(T angle, T x, T y, T z)
+	/// @see - rotate(mat<4, 4, T, Q> const& m, T angle, T x, T y, T z)
+	/// @see - rotate(T angle, vec<3, T, Q> const& v)
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> rotateNormalizedAxis(
+		mat<4, 4, T, Q> const& m,
+		T const& angle,
+		vec<3, T, Q> const& axis);
+
+	/// Rotates a quaternion from a vector of 3 components normalized axis and an angle.
+	///
+	/// @param q Source orientation
+	/// @param angle Angle expressed in radians.
+	/// @param axis Normalized axis of the rotation, must be normalized.
+	///
+	/// @see gtx_rotate_normalized_axis
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL qua<T, Q> rotateNormalizedAxis(
+		qua<T, Q> const& q,
+		T const& angle,
+		vec<3, T, Q> const& axis);
+
+	/// @}
+}//namespace glm
+
+#include "rotate_normalized_axis.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.inl
new file mode 100644
index 000000000000..352a56cb17ac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_normalized_axis.inl
@@ -0,0 +1,58 @@
+/// @ref gtx_rotate_normalized_axis
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> rotateNormalizedAxis
+	(
+		mat<4, 4, T, Q> const& m,
+		T const& angle,
+		vec<3, T, Q> const& v
+	)
+	{
+		T const a = angle;
+		T const c = cos(a);
+		T const s = sin(a);
+
+		vec<3, T, Q> const axis(v);
+
+		vec<3, T, Q> const temp((static_cast<T>(1) - c) * axis);
+
+		mat<4, 4, T, Q> Rotate;
+		Rotate[0][0] = c + temp[0] * axis[0];
+		Rotate[0][1] = 0 + temp[0] * axis[1] + s * axis[2];
+		Rotate[0][2] = 0 + temp[0] * axis[2] - s * axis[1];
+
+		Rotate[1][0] = 0 + temp[1] * axis[0] - s * axis[2];
+		Rotate[1][1] = c + temp[1] * axis[1];
+		Rotate[1][2] = 0 + temp[1] * axis[2] + s * axis[0];
+
+		Rotate[2][0] = 0 + temp[2] * axis[0] + s * axis[1];
+		Rotate[2][1] = 0 + temp[2] * axis[1] - s * axis[0];
+		Rotate[2][2] = c + temp[2] * axis[2];
+
+		mat<4, 4, T, Q> Result;
+		Result[0] = m[0] * Rotate[0][0] + m[1] * Rotate[0][1] + m[2] * Rotate[0][2];
+		Result[1] = m[0] * Rotate[1][0] + m[1] * Rotate[1][1] + m[2] * Rotate[1][2];
+		Result[2] = m[0] * Rotate[2][0] + m[1] * Rotate[2][1] + m[2] * Rotate[2][2];
+		Result[3] = m[3];
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER qua<T, Q> rotateNormalizedAxis
+	(
+		qua<T, Q> const& q,
+		T const& angle,
+		vec<3, T, Q> const& v
+	)
+	{
+		vec<3, T, Q> const Tmp(v);
+
+		T const AngleRad(angle);
+		T const Sin = sin(AngleRad * T(0.5));
+
+		return q * qua<T, Q>::wxyz(cos(AngleRad * static_cast<T>(0.5)), Tmp.x * Sin, Tmp.y * Sin, Tmp.z * Sin);
+		//return gtc::quaternion::cross(q, tquat<T, Q>(cos(AngleRad * T(0.5)), Tmp.x * fSin, Tmp.y * fSin, Tmp.z * fSin));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.hpp
new file mode 100644
index 000000000000..dcd5b95a6e5b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.hpp
@@ -0,0 +1,123 @@
+/// @ref gtx_rotate_vector
+/// @file glm/gtx/rotate_vector.hpp
+///
+/// @see core (dependence)
+/// @see gtx_transform (dependence)
+///
+/// @defgroup gtx_rotate_vector GLM_GTX_rotate_vector
+/// @ingroup gtx
+///
+/// Include <glm/gtx/rotate_vector.hpp> to use the features of this extension.
+///
+/// Function to directly rotate a vector
+
+#pragma once
+
+// Dependency:
+#include "../gtx/transform.hpp"
+#include "../gtc/epsilon.hpp"
+#include "../ext/vector_relational.hpp"
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_rotate_vector is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_rotate_vector extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_rotate_vector
+	/// @{
+
+	/// Returns Spherical interpolation between two vectors
+	///
+	/// @param x A first vector
+	/// @param y A second vector
+	/// @param a Interpolation factor. The interpolation is defined beyond the range [0, 1].
+	///
+	/// @see gtx_rotate_vector
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> slerp(
+		vec<3, T, Q> const& x,
+		vec<3, T, Q> const& y,
+		T const& a);
+
+	//! Rotate a two dimensional vector.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<2, T, Q> rotate(
+		vec<2, T, Q> const& v,
+		T const& angle);
+
+	//! Rotate a three dimensional vector around an axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rotate(
+		vec<3, T, Q> const& v,
+		T const& angle,
+		vec<3, T, Q> const& normal);
+
+	//! Rotate a four dimensional vector around an axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> rotate(
+		vec<4, T, Q> const& v,
+		T const& angle,
+		vec<3, T, Q> const& normal);
+
+	//! Rotate a three dimensional vector around the X axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rotateX(
+		vec<3, T, Q> const& v,
+		T const& angle);
+
+	//! Rotate a three dimensional vector around the Y axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rotateY(
+		vec<3, T, Q> const& v,
+		T const& angle);
+
+	//! Rotate a three dimensional vector around the Z axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<3, T, Q> rotateZ(
+		vec<3, T, Q> const& v,
+		T const& angle);
+
+	//! Rotate a four dimensional vector around the X axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> rotateX(
+		vec<4, T, Q> const& v,
+		T const& angle);
+
+	//! Rotate a four dimensional vector around the Y axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> rotateY(
+		vec<4, T, Q> const& v,
+		T const& angle);
+
+	//! Rotate a four dimensional vector around the Z axis.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL vec<4, T, Q> rotateZ(
+		vec<4, T, Q> const& v,
+		T const& angle);
+
+	//! Build a rotation matrix from a normal and a up vector.
+	//! From GLM_GTX_rotate_vector extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> orientation(
+		vec<3, T, Q> const& Normal,
+		vec<3, T, Q> const& Up);
+
+	/// @}
+}//namespace glm
+
+#include "rotate_vector.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.inl
new file mode 100644
index 000000000000..f8136e765e05
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/rotate_vector.inl
@@ -0,0 +1,187 @@
+/// @ref gtx_rotate_vector
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> slerp
+	(
+		vec<3, T, Q> const& x,
+		vec<3, T, Q> const& y,
+		T const& a
+	)
+	{
+		// get cosine of angle between vectors (-1 -> 1)
+		T CosAlpha = dot(x, y);
+		// get angle (0 -> pi)
+		T Alpha = acos(CosAlpha);
+		// get sine of angle between vectors (0 -> 1)
+		T SinAlpha = sin(Alpha);
+		// this breaks down when SinAlpha = 0, i.e. Alpha = 0 or pi
+		T t1 = sin((static_cast<T>(1) - a) * Alpha) / SinAlpha;
+		T t2 = sin(a * Alpha) / SinAlpha;
+
+		// interpolate src vectors
+		return x * t1 + y * t2;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, T, Q> rotate
+	(
+		vec<2, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<2, T, Q> Result;
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.x = v.x * Cos - v.y * Sin;
+		Result.y = v.x * Sin + v.y * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rotate
+	(
+		vec<3, T, Q> const& v,
+		T const& angle,
+		vec<3, T, Q> const& normal
+	)
+	{
+		return mat<3, 3, T, Q>(glm::rotate(angle, normal)) * v;
+	}
+	/*
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rotateGTX(
+		const vec<3, T, Q>& x,
+		T angle,
+		const vec<3, T, Q>& normal)
+	{
+		const T Cos = cos(radians(angle));
+		const T Sin = sin(radians(angle));
+		return x * Cos + ((x * normal) * (T(1) - Cos)) * normal + cross(x, normal) * Sin;
+	}
+	*/
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> rotate
+	(
+		vec<4, T, Q> const& v,
+		T const& angle,
+		vec<3, T, Q> const& normal
+	)
+	{
+		return rotate(angle, normal) * v;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rotateX
+	(
+		vec<3, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<3, T, Q> Result(v);
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.y = v.y * Cos - v.z * Sin;
+		Result.z = v.y * Sin + v.z * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rotateY
+	(
+		vec<3, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<3, T, Q> Result = v;
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.x =  v.x * Cos + v.z * Sin;
+		Result.z = -v.x * Sin + v.z * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, T, Q> rotateZ
+	(
+		vec<3, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<3, T, Q> Result = v;
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.x = v.x * Cos - v.y * Sin;
+		Result.y = v.x * Sin + v.y * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> rotateX
+	(
+		vec<4, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<4, T, Q> Result = v;
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.y = v.y * Cos - v.z * Sin;
+		Result.z = v.y * Sin + v.z * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> rotateY
+	(
+		vec<4, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<4, T, Q> Result = v;
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.x =  v.x * Cos + v.z * Sin;
+		Result.z = -v.x * Sin + v.z * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, T, Q> rotateZ
+	(
+		vec<4, T, Q> const& v,
+		T const& angle
+	)
+	{
+		vec<4, T, Q> Result = v;
+		T const Cos(cos(angle));
+		T const Sin(sin(angle));
+
+		Result.x = v.x * Cos - v.y * Sin;
+		Result.y = v.x * Sin + v.y * Cos;
+		return Result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> orientation
+	(
+		vec<3, T, Q> const& Normal,
+		vec<3, T, Q> const& Up
+	)
+	{
+		if(all(equal(Normal, Up, epsilon<T>())))
+			return mat<4, 4, T, Q>(static_cast<T>(1));
+
+		vec<3, T, Q> RotationAxis = cross(Up, Normal);
+		T Angle = acos(dot(Normal, Up));
+
+		return rotate(Angle, RotationAxis);
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_multiplication.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_multiplication.hpp
new file mode 100644
index 000000000000..9f9f2fb3edbc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_multiplication.hpp
@@ -0,0 +1,82 @@
+/// @ref gtx_scalar_multiplication
+/// @file glm/gtx/scalar_multiplication.hpp
+/// @author Joshua Moerman
+///
+/// @defgroup gtx_scalar_multiplication GLM_GTX_scalar_multiplication
+/// @ingroup gtx
+///
+/// Include <glm/gtx/scalar_multiplication.hpp> to use the features of this extension.
+///
+/// Enables scalar multiplication for all types
+///
+/// Since GLSL is very strict about types, the following (often used) combinations do not work:
+///    double * vec4
+///    int * vec4
+///    vec4 / int
+/// So we'll fix that! Of course "float * vec4" should remain the same (hence the enable_if magic)
+
+#pragma once
+
+#include "../detail/setup.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_scalar_multiplication is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_scalar_multiplication extension included")
+#	endif
+#endif
+
+#include "../vec2.hpp"
+#include "../vec3.hpp"
+#include "../vec4.hpp"
+#include "../mat2x2.hpp"
+#include <type_traits>
+
+namespace glm
+{
+	/// @addtogroup gtx_scalar_multiplication
+	/// @{
+
+	template<typename T, typename Vec>
+	using return_type_scalar_multiplication = typename std::enable_if<
+		!std::is_same<T, float>::value       // T may not be a float
+		&& std::is_arithmetic<T>::value, Vec // But it may be an int or double (no vec3 or mat3, ...)
+	>::type;
+
+#define GLM_IMPLEMENT_SCAL_MULT(Vec) \
+	template<typename T> \
+	return_type_scalar_multiplication<T, Vec> \
+	operator*(T const& s, Vec rh){ \
+		return rh *= static_cast<float>(s); \
+	} \
+	 \
+	template<typename T> \
+	return_type_scalar_multiplication<T, Vec> \
+	operator*(Vec lh, T const& s){ \
+		return lh *= static_cast<float>(s); \
+	} \
+	 \
+	template<typename T> \
+	return_type_scalar_multiplication<T, Vec> \
+	operator/(Vec lh, T const& s){ \
+		return lh *= 1.0f / static_cast<float>(s); \
+	}
+
+GLM_IMPLEMENT_SCAL_MULT(vec2)
+GLM_IMPLEMENT_SCAL_MULT(vec3)
+GLM_IMPLEMENT_SCAL_MULT(vec4)
+
+GLM_IMPLEMENT_SCAL_MULT(mat2)
+GLM_IMPLEMENT_SCAL_MULT(mat2x3)
+GLM_IMPLEMENT_SCAL_MULT(mat2x4)
+GLM_IMPLEMENT_SCAL_MULT(mat3x2)
+GLM_IMPLEMENT_SCAL_MULT(mat3)
+GLM_IMPLEMENT_SCAL_MULT(mat3x4)
+GLM_IMPLEMENT_SCAL_MULT(mat4x2)
+GLM_IMPLEMENT_SCAL_MULT(mat4x3)
+GLM_IMPLEMENT_SCAL_MULT(mat4)
+
+#undef GLM_IMPLEMENT_SCAL_MULT
+	/// @}
+} // namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.hpp
new file mode 100644
index 000000000000..8be9c57b8b35
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.hpp
@@ -0,0 +1,36 @@
+/// @ref gtx_scalar_relational
+/// @file glm/gtx/scalar_relational.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_scalar_relational GLM_GTX_scalar_relational
+/// @ingroup gtx
+///
+/// Include <glm/gtx/scalar_relational.hpp> to use the features of this extension.
+///
+/// Extend a position from a source to a position at a defined length.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_extend is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_extend extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_scalar_relational
+	/// @{
+
+
+
+	/// @}
+}//namespace glm
+
+#include "scalar_relational.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.inl
new file mode 100644
index 000000000000..c2a121cff977
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/scalar_relational.inl
@@ -0,0 +1,88 @@
+/// @ref gtx_scalar_relational
+
+namespace glm
+{
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool lessThan
+	(
+		T const& x,
+		T const& y
+	)
+	{
+		return x < y;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool lessThanEqual
+	(
+		T const& x,
+		T const& y
+	)
+	{
+		return x <= y;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool greaterThan
+	(
+		T const& x,
+		T const& y
+	)
+	{
+		return x > y;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool greaterThanEqual
+	(
+		T const& x,
+		T const& y
+	)
+	{
+		return x >= y;
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool equal
+	(
+		T const& x,
+		T const& y
+	)
+	{
+		return detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(x, y);
+	}
+
+	template<typename T>
+	GLM_FUNC_QUALIFIER bool notEqual
+	(
+		T const& x,
+		T const& y
+	)
+	{
+		return !detail::compute_equal<T, std::numeric_limits<T>::is_iec559>::call(x, y);
+	}
+
+	GLM_FUNC_QUALIFIER bool any
+	(
+		bool const& x
+	)
+	{
+		return x;
+	}
+
+	GLM_FUNC_QUALIFIER bool all
+	(
+		bool const& x
+	)
+	{
+		return x;
+	}
+
+	GLM_FUNC_QUALIFIER bool not_
+	(
+		bool const& x
+	)
+	{
+		return !x;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/spline.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/spline.hpp
new file mode 100644
index 000000000000..731c979e358a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/spline.hpp
@@ -0,0 +1,65 @@
+/// @ref gtx_spline
+/// @file glm/gtx/spline.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_spline GLM_GTX_spline
+/// @ingroup gtx
+///
+/// Include <glm/gtx/spline.hpp> to use the features of this extension.
+///
+/// Spline functions
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtx/optimum_pow.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_spline is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_spline extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_spline
+	/// @{
+
+	/// Return a point from a catmull rom curve.
+	/// @see gtx_spline extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType catmullRom(
+		genType const& v1,
+		genType const& v2,
+		genType const& v3,
+		genType const& v4,
+		typename genType::value_type const& s);
+
+	/// Return a point from a hermite curve.
+	/// @see gtx_spline extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType hermite(
+		genType const& v1,
+		genType const& t1,
+		genType const& v2,
+		genType const& t2,
+		typename genType::value_type const& s);
+
+	/// Return a point from a cubic curve.
+	/// @see gtx_spline extension.
+	template<typename genType>
+	GLM_FUNC_DECL genType cubic(
+		genType const& v1,
+		genType const& v2,
+		genType const& v3,
+		genType const& v4,
+		typename genType::value_type const& s);
+
+	/// @}
+}//namespace glm
+
+#include "spline.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/spline.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/spline.inl
new file mode 100644
index 000000000000..c3fd05656291
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/spline.inl
@@ -0,0 +1,60 @@
+/// @ref gtx_spline
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType catmullRom
+	(
+		genType const& v1,
+		genType const& v2,
+		genType const& v3,
+		genType const& v4,
+		typename genType::value_type const& s
+	)
+	{
+		typename genType::value_type s2 = pow2(s);
+		typename genType::value_type s3 = pow3(s);
+
+		typename genType::value_type f1 = -s3 + typename genType::value_type(2) * s2 - s;
+		typename genType::value_type f2 = typename genType::value_type(3) * s3 - typename genType::value_type(5) * s2 + typename genType::value_type(2);
+		typename genType::value_type f3 = typename genType::value_type(-3) * s3 + typename genType::value_type(4) * s2 + s;
+		typename genType::value_type f4 = s3 - s2;
+
+		return (f1 * v1 + f2 * v2 + f3 * v3 + f4 * v4) / typename genType::value_type(2);
+
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType hermite
+	(
+		genType const& v1,
+		genType const& t1,
+		genType const& v2,
+		genType const& t2,
+		typename genType::value_type const& s
+	)
+	{
+		typename genType::value_type s2 = pow2(s);
+		typename genType::value_type s3 = pow3(s);
+
+		typename genType::value_type f1 = typename genType::value_type(2) * s3 - typename genType::value_type(3) * s2 + typename genType::value_type(1);
+		typename genType::value_type f2 = typename genType::value_type(-2) * s3 + typename genType::value_type(3) * s2;
+		typename genType::value_type f3 = s3 - typename genType::value_type(2) * s2 + s;
+		typename genType::value_type f4 = s3 - s2;
+
+		return f1 * v1 + f2 * v2 + f3 * t1 + f4 * t2;
+	}
+
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType cubic
+	(
+		genType const& v1,
+		genType const& v2,
+		genType const& v3,
+		genType const& v4,
+		typename genType::value_type const& s
+	)
+	{
+		return ((v1 * s + v2) * s + v3) * s + v4;
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.hpp
new file mode 100644
index 000000000000..cd3be8cb7892
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.hpp
@@ -0,0 +1,68 @@
+/// @ref gtx_std_based_type
+/// @file glm/gtx/std_based_type.hpp
+///
+/// @see core (dependence)
+/// @see gtx_extented_min_max (dependence)
+///
+/// @defgroup gtx_std_based_type GLM_GTX_std_based_type
+/// @ingroup gtx
+///
+/// Include <glm/gtx/std_based_type.hpp> to use the features of this extension.
+///
+/// Adds vector types based on STL value types.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include <cstdlib>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_std_based_type is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_std_based_type extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_std_based_type
+	/// @{
+
+	/// Vector type based of one std::size_t component.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<1, std::size_t, defaultp>		size1;
+
+	/// Vector type based of two std::size_t components.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<2, std::size_t, defaultp>		size2;
+
+	/// Vector type based of three std::size_t components.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<3, std::size_t, defaultp>		size3;
+
+	/// Vector type based of four std::size_t components.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<4, std::size_t, defaultp>		size4;
+
+	/// Vector type based of one std::size_t component.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<1, std::size_t, defaultp>		size1_t;
+
+	/// Vector type based of two std::size_t components.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<2, std::size_t, defaultp>		size2_t;
+
+	/// Vector type based of three std::size_t components.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<3, std::size_t, defaultp>		size3_t;
+
+	/// Vector type based of four std::size_t components.
+	/// @see GLM_GTX_std_based_type
+	typedef vec<4, std::size_t, defaultp>		size4_t;
+
+	/// @}
+}//namespace glm
+
+#include "std_based_type.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.inl
new file mode 100644
index 000000000000..9c34bdb6e0f7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/std_based_type.inl
@@ -0,0 +1,6 @@
+/// @ref gtx_std_based_type
+
+namespace glm
+{
+
+}
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.hpp
new file mode 100644
index 000000000000..71f6ece4c9f1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.hpp
@@ -0,0 +1,46 @@
+/// @ref gtx_string_cast
+/// @file glm/gtx/string_cast.hpp
+///
+/// @see core (dependence)
+/// @see gtx_integer (dependence)
+/// @see gtx_quaternion (dependence)
+///
+/// @defgroup gtx_string_cast GLM_GTX_string_cast
+/// @ingroup gtx
+///
+/// Include <glm/gtx/string_cast.hpp> to use the features of this extension.
+///
+/// Setup strings for GLM type values
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/type_precision.hpp"
+#include "../gtc/quaternion.hpp"
+#include "../gtx/dual_quaternion.hpp"
+#include <string>
+#include <cmath>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_string_cast is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_string_cast extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_string_cast
+	/// @{
+
+	/// Create a string from a GLM vector or matrix typed variable.
+	/// @see gtx_string_cast extension.
+	template<typename genType>
+	GLM_FUNC_DECL std::string to_string(genType const& x);
+
+	/// @}
+}//namespace glm
+
+#include "string_cast.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.inl
new file mode 100644
index 000000000000..0b2d4b6aebc6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/string_cast.inl
@@ -0,0 +1,492 @@
+/// @ref gtx_string_cast
+
+#include <cstdarg>
+#include <cstdio>
+
+namespace glm{
+namespace detail
+{
+	template <typename T>
+	struct cast
+	{
+		typedef T value_type;
+	};
+
+	template <>
+	struct cast<float>
+	{
+		typedef double value_type;
+	};
+
+	GLM_FUNC_QUALIFIER std::string format(const char* msg, ...)
+	{
+		std::size_t const STRING_BUFFER(4096);
+		char text[STRING_BUFFER];
+		va_list list;
+
+		if(msg == GLM_NULLPTR)
+			return std::string();
+
+		va_start(list, msg);
+#		if (GLM_COMPILER & GLM_COMPILER_VC)
+			vsprintf_s(text, STRING_BUFFER, msg, list);
+#		else//
+			std::vsnprintf(text, STRING_BUFFER, msg, list);
+#		endif//
+		va_end(list);
+
+		return std::string(text);
+	}
+
+	static const char* LabelTrue = "true";
+	static const char* LabelFalse = "false";
+
+	template<typename T, bool isFloat = false>
+	struct literal
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "%d";}
+	};
+
+	template<typename T>
+	struct literal<T, true>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "%f";}
+	};
+
+#	if GLM_MODEL == GLM_MODEL_32 && GLM_COMPILER && GLM_COMPILER_VC
+	template<>
+	struct literal<uint64_t, false>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "%lld";}
+	};
+
+	template<>
+	struct literal<int64_t, false>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "%lld";}
+	};
+#	endif//GLM_MODEL == GLM_MODEL_32 && GLM_COMPILER && GLM_COMPILER_VC
+
+	template<typename T>
+	struct prefix{};
+
+	template<>
+	struct prefix<float>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "";}
+	};
+
+	template<>
+	struct prefix<double>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "d";}
+	};
+
+	template<>
+	struct prefix<bool>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "b";}
+	};
+
+	template<>
+	struct prefix<uint8_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "u8";}
+	};
+
+	template<>
+	struct prefix<int8_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "i8";}
+	};
+
+	template<>
+	struct prefix<uint16_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "u16";}
+	};
+
+	template<>
+	struct prefix<int16_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "i16";}
+	};
+
+	template<>
+	struct prefix<uint32_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "u";}
+	};
+
+	template<>
+	struct prefix<int32_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "i";}
+	};
+
+	template<>
+	struct prefix<uint64_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "u64";}
+	};
+
+	template<>
+	struct prefix<int64_t>
+	{
+		GLM_FUNC_QUALIFIER static char const * value() {return "i64";}
+	};
+
+	template<typename matType>
+	struct compute_to_string
+	{};
+
+	template<qualifier Q>
+	struct compute_to_string<vec<1, bool, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<1, bool, Q> const& x)
+		{
+			return detail::format("bvec1(%s)",
+				x[0] ? detail::LabelTrue : detail::LabelFalse);
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_to_string<vec<2, bool, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<2, bool, Q> const& x)
+		{
+			return detail::format("bvec2(%s, %s)",
+				x[0] ? detail::LabelTrue : detail::LabelFalse,
+				x[1] ? detail::LabelTrue : detail::LabelFalse);
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_to_string<vec<3, bool, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<3, bool, Q> const& x)
+		{
+			return detail::format("bvec3(%s, %s, %s)",
+				x[0] ? detail::LabelTrue : detail::LabelFalse,
+				x[1] ? detail::LabelTrue : detail::LabelFalse,
+				x[2] ? detail::LabelTrue : detail::LabelFalse);
+		}
+	};
+
+	template<qualifier Q>
+	struct compute_to_string<vec<4, bool, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<4, bool, Q> const& x)
+		{
+			return detail::format("bvec4(%s, %s, %s, %s)",
+				x[0] ? detail::LabelTrue : detail::LabelFalse,
+				x[1] ? detail::LabelTrue : detail::LabelFalse,
+				x[2] ? detail::LabelTrue : detail::LabelFalse,
+				x[3] ? detail::LabelTrue : detail::LabelFalse);
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<vec<1, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<1, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%svec1(%s)",
+				PrefixStr,
+				LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<vec<2, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<2, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%svec2(%s, %s)",
+				PrefixStr,
+				LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0]),
+				static_cast<typename cast<T>::value_type>(x[1]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<vec<3, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<3, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%svec3(%s, %s, %s)",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0]),
+				static_cast<typename cast<T>::value_type>(x[1]),
+				static_cast<typename cast<T>::value_type>(x[2]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<vec<4, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(vec<4, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%svec4(%s, %s, %s, %s)",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0]),
+				static_cast<typename cast<T>::value_type>(x[1]),
+				static_cast<typename cast<T>::value_type>(x[2]),
+				static_cast<typename cast<T>::value_type>(x[3]));
+		}
+	};
+
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<2, 2, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<2, 2, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat2x2((%s, %s), (%s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<2, 3, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<2, 3, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat2x3((%s, %s, %s), (%s, %s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]), static_cast<typename cast<T>::value_type>(x[0][2]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]), static_cast<typename cast<T>::value_type>(x[1][2]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<2, 4, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<2, 4, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat2x4((%s, %s, %s, %s), (%s, %s, %s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]), static_cast<typename cast<T>::value_type>(x[0][2]), static_cast<typename cast<T>::value_type>(x[0][3]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]), static_cast<typename cast<T>::value_type>(x[1][2]), static_cast<typename cast<T>::value_type>(x[1][3]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<3, 2, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<3, 2, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat3x2((%s, %s), (%s, %s), (%s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]),
+				static_cast<typename cast<T>::value_type>(x[2][0]), static_cast<typename cast<T>::value_type>(x[2][1]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<3, 3, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<3, 3, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat3x3((%s, %s, %s), (%s, %s, %s), (%s, %s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]), static_cast<typename cast<T>::value_type>(x[0][2]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]), static_cast<typename cast<T>::value_type>(x[1][2]),
+				static_cast<typename cast<T>::value_type>(x[2][0]), static_cast<typename cast<T>::value_type>(x[2][1]), static_cast<typename cast<T>::value_type>(x[2][2]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<3, 4, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<3, 4, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat3x4((%s, %s, %s, %s), (%s, %s, %s, %s), (%s, %s, %s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]), static_cast<typename cast<T>::value_type>(x[0][2]), static_cast<typename cast<T>::value_type>(x[0][3]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]), static_cast<typename cast<T>::value_type>(x[1][2]), static_cast<typename cast<T>::value_type>(x[1][3]),
+				static_cast<typename cast<T>::value_type>(x[2][0]), static_cast<typename cast<T>::value_type>(x[2][1]), static_cast<typename cast<T>::value_type>(x[2][2]), static_cast<typename cast<T>::value_type>(x[2][3]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<4, 2, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<4, 2, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat4x2((%s, %s), (%s, %s), (%s, %s), (%s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]),
+				static_cast<typename cast<T>::value_type>(x[2][0]), static_cast<typename cast<T>::value_type>(x[2][1]),
+				static_cast<typename cast<T>::value_type>(x[3][0]), static_cast<typename cast<T>::value_type>(x[3][1]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<4, 3, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<4, 3, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat4x3((%s, %s, %s), (%s, %s, %s), (%s, %s, %s), (%s, %s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]), static_cast<typename cast<T>::value_type>(x[0][2]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]), static_cast<typename cast<T>::value_type>(x[1][2]),
+				static_cast<typename cast<T>::value_type>(x[2][0]), static_cast<typename cast<T>::value_type>(x[2][1]), static_cast<typename cast<T>::value_type>(x[2][2]),
+				static_cast<typename cast<T>::value_type>(x[3][0]), static_cast<typename cast<T>::value_type>(x[3][1]), static_cast<typename cast<T>::value_type>(x[3][2]));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<mat<4, 4, T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(mat<4, 4, T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%smat4x4((%s, %s, %s, %s), (%s, %s, %s, %s), (%s, %s, %s, %s), (%s, %s, %s, %s))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x[0][0]), static_cast<typename cast<T>::value_type>(x[0][1]), static_cast<typename cast<T>::value_type>(x[0][2]), static_cast<typename cast<T>::value_type>(x[0][3]),
+				static_cast<typename cast<T>::value_type>(x[1][0]), static_cast<typename cast<T>::value_type>(x[1][1]), static_cast<typename cast<T>::value_type>(x[1][2]), static_cast<typename cast<T>::value_type>(x[1][3]),
+				static_cast<typename cast<T>::value_type>(x[2][0]), static_cast<typename cast<T>::value_type>(x[2][1]), static_cast<typename cast<T>::value_type>(x[2][2]), static_cast<typename cast<T>::value_type>(x[2][3]),
+				static_cast<typename cast<T>::value_type>(x[3][0]), static_cast<typename cast<T>::value_type>(x[3][1]), static_cast<typename cast<T>::value_type>(x[3][2]), static_cast<typename cast<T>::value_type>(x[3][3]));
+		}
+	};
+
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<qua<T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(qua<T, Q> const& q)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%squat(%s, {%s, %s, %s})",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(q.w),
+				static_cast<typename cast<T>::value_type>(q.x),
+				static_cast<typename cast<T>::value_type>(q.y),
+				static_cast<typename cast<T>::value_type>(q.z));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_to_string<tdualquat<T, Q> >
+	{
+		GLM_FUNC_QUALIFIER static std::string call(tdualquat<T, Q> const& x)
+		{
+			char const * PrefixStr = prefix<T>::value();
+			char const * LiteralStr = literal<T, std::numeric_limits<T>::is_iec559>::value();
+			std::string FormatStr(detail::format("%sdualquat((%s, {%s, %s, %s}), (%s, {%s, %s, %s}))",
+				PrefixStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr,
+				LiteralStr, LiteralStr, LiteralStr, LiteralStr));
+
+			return detail::format(FormatStr.c_str(),
+				static_cast<typename cast<T>::value_type>(x.real.w),
+				static_cast<typename cast<T>::value_type>(x.real.x),
+				static_cast<typename cast<T>::value_type>(x.real.y),
+				static_cast<typename cast<T>::value_type>(x.real.z),
+				static_cast<typename cast<T>::value_type>(x.dual.w),
+				static_cast<typename cast<T>::value_type>(x.dual.x),
+				static_cast<typename cast<T>::value_type>(x.dual.y),
+				static_cast<typename cast<T>::value_type>(x.dual.z));
+		}
+	};
+
+}//namespace detail
+
+template<class matType>
+GLM_FUNC_QUALIFIER std::string to_string(matType const& x)
+{
+	return detail::compute_to_string<matType>::call(x);
+}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/texture.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/texture.hpp
new file mode 100644
index 000000000000..20585e68ce11
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/texture.hpp
@@ -0,0 +1,46 @@
+/// @ref gtx_texture
+/// @file glm/gtx/texture.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_texture GLM_GTX_texture
+/// @ingroup gtx
+///
+/// Include <glm/gtx/texture.hpp> to use the features of this extension.
+///
+/// Wrapping mode of texture coordinates.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/integer.hpp"
+#include "../gtx/component_wise.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_texture is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_texture extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_texture
+	/// @{
+
+	/// Compute the number of mipmaps levels necessary to create a mipmap complete texture
+	///
+	/// @param Extent Extent of the texture base level mipmap
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point or signed integer scalar types
+	/// @tparam Q Value from qualifier enum
+	template <length_t L, typename T, qualifier Q>
+	T levels(vec<L, T, Q> const& Extent);
+
+	/// @}
+}// namespace glm
+
+#include "texture.inl"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/texture.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/texture.inl
new file mode 100644
index 000000000000..593c826141b0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/texture.inl
@@ -0,0 +1,17 @@
+/// @ref gtx_texture
+
+namespace glm
+{
+	template <length_t L, typename T, qualifier Q>
+	inline T levels(vec<L, T, Q> const& Extent)
+	{
+		return glm::log2(compMax(Extent)) + static_cast<T>(1);
+	}
+
+	template <typename T>
+	inline T levels(T Extent)
+	{
+		return vec<1, T, defaultp>(Extent).x;
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/transform.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform.hpp
new file mode 100644
index 000000000000..0279fc8bd329
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform.hpp
@@ -0,0 +1,60 @@
+/// @ref gtx_transform
+/// @file glm/gtx/transform.hpp
+///
+/// @see core (dependence)
+/// @see gtc_matrix_transform (dependence)
+/// @see gtx_transform
+/// @see gtx_transform2
+///
+/// @defgroup gtx_transform GLM_GTX_transform
+/// @ingroup gtx
+///
+/// Include <glm/gtx/transform.hpp> to use the features of this extension.
+///
+/// Add transformation matrices
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/matrix_transform.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_transform is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_transform extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_transform
+	/// @{
+
+	/// Transforms a matrix with a translation 4 * 4 matrix created from 3 scalars.
+	/// @see gtc_matrix_transform
+	/// @see gtx_transform
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> translate(
+		vec<3, T, Q> const& v);
+
+	/// Builds a rotation 4 * 4 matrix created from an axis of 3 scalars and an angle expressed in radians.
+	/// @see gtc_matrix_transform
+	/// @see gtx_transform
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> rotate(
+		T angle,
+		vec<3, T, Q> const& v);
+
+	/// Transforms a matrix with a scale 4 * 4 matrix created from a vector of 3 components.
+	/// @see gtc_matrix_transform
+	/// @see gtx_transform
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> scale(
+		vec<3, T, Q> const& v);
+
+	/// @}
+}// namespace glm
+
+#include "transform.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/transform.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform.inl
new file mode 100644
index 000000000000..48ee6801b651
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform.inl
@@ -0,0 +1,23 @@
+/// @ref gtx_transform
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> translate(vec<3, T, Q> const& v)
+	{
+		return translate(mat<4, 4, T, Q>(static_cast<T>(1)), v);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> rotate(T angle, vec<3, T, Q> const& v)
+	{
+		return rotate(mat<4, 4, T, Q>(static_cast<T>(1)), angle, v);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> scale(vec<3, T, Q> const& v)
+	{
+		return scale(mat<4, 4, T, Q>(static_cast<T>(1)), v);
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.hpp
new file mode 100644
index 000000000000..0d8ba9d90bc5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.hpp
@@ -0,0 +1,89 @@
+/// @ref gtx_transform2
+/// @file glm/gtx/transform2.hpp
+///
+/// @see core (dependence)
+/// @see gtx_transform (dependence)
+///
+/// @defgroup gtx_transform2 GLM_GTX_transform2
+/// @ingroup gtx
+///
+/// Include <glm/gtx/transform2.hpp> to use the features of this extension.
+///
+/// Add extra transformation matrices
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtx/transform.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_transform2 is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_transform2 extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_transform2
+	/// @{
+
+	//! Transforms a matrix with a shearing on X axis.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> shearX2D(mat<3, 3, T, Q> const& m, T y);
+
+	//! Transforms a matrix with a shearing on Y axis.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> shearY2D(mat<3, 3, T, Q> const& m, T x);
+
+	//! Transforms a matrix with a shearing on X axis
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> shearX3D(mat<4, 4, T, Q> const& m, T y, T z);
+
+	//! Transforms a matrix with a shearing on Y axis.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> shearY3D(mat<4, 4, T, Q> const& m, T x, T z);
+
+	//! Transforms a matrix with a shearing on Z axis.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> shearZ3D(mat<4, 4, T, Q> const& m, T x, T y);
+
+	//template<typename T> GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shear(const mat<4, 4, T, Q> & m, shearPlane, planePoint, angle)
+	// Identity + tan(angle) * cross(Normal, OnPlaneVector)     0
+	// - dot(PointOnPlane, normal) * OnPlaneVector              1
+
+	// Reflect functions seem to don't work
+	//template<typename T> mat<3, 3, T, Q> reflect2D(const mat<3, 3, T, Q> & m, const vec<3, T, Q>& normal){return reflect2DGTX(m, normal);}									//!< \brief Build a reflection matrix (from GLM_GTX_transform2 extension)
+	//template<typename T> mat<4, 4, T, Q> reflect3D(const mat<4, 4, T, Q> & m, const vec<3, T, Q>& normal){return reflect3DGTX(m, normal);}									//!< \brief Build a reflection matrix (from GLM_GTX_transform2 extension)
+
+	//! Build planar projection matrix along normal axis.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<3, 3, T, Q> proj2D(mat<3, 3, T, Q> const& m, vec<3, T, Q> const& normal);
+
+	//! Build planar projection matrix along normal axis.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> proj3D(mat<4, 4, T, Q> const & m, vec<3, T, Q> const& normal);
+
+	//! Build a scale bias matrix.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> scaleBias(T scale, T bias);
+
+	//! Build a scale bias matrix.
+	//! From GLM_GTX_transform2 extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL mat<4, 4, T, Q> scaleBias(mat<4, 4, T, Q> const& m, T scale, T bias);
+
+	/// @}
+}// namespace glm
+
+#include "transform2.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.inl
new file mode 100644
index 000000000000..0118ab09603a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/transform2.inl
@@ -0,0 +1,125 @@
+/// @ref gtx_transform2
+
+namespace glm
+{
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> shearX2D(mat<3, 3, T, Q> const& m, T s)
+	{
+		mat<3, 3, T, Q> r(1);
+		r[1][0] = s;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> shearY2D(mat<3, 3, T, Q> const& m, T s)
+	{
+		mat<3, 3, T, Q> r(1);
+		r[0][1] = s;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shearX3D(mat<4, 4, T, Q> const& m, T s, T t)
+	{
+		mat<4, 4, T, Q> r(1);
+		r[0][1] = s;
+		r[0][2] = t;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shearY3D(mat<4, 4, T, Q> const& m, T s, T t)
+	{
+		mat<4, 4, T, Q> r(1);
+		r[1][0] = s;
+		r[1][2] = t;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> shearZ3D(mat<4, 4, T, Q> const& m, T s, T t)
+	{
+		mat<4, 4, T, Q> r(1);
+		r[2][0] = s;
+		r[2][1] = t;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> reflect2D(mat<3, 3, T, Q> const& m, vec<3, T, Q> const& normal)
+	{
+		mat<3, 3, T, Q> r(static_cast<T>(1));
+		r[0][0] = static_cast<T>(1) - static_cast<T>(2) * normal.x * normal.x;
+		r[0][1] = -static_cast<T>(2) * normal.x * normal.y;
+		r[1][0] = -static_cast<T>(2) * normal.x * normal.y;
+		r[1][1] = static_cast<T>(1) - static_cast<T>(2) * normal.y * normal.y;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> reflect3D(mat<4, 4, T, Q> const& m, vec<3, T, Q> const& normal)
+	{
+		mat<4, 4, T, Q> r(static_cast<T>(1));
+		r[0][0] = static_cast<T>(1) - static_cast<T>(2) * normal.x * normal.x;
+		r[0][1] = -static_cast<T>(2) * normal.x * normal.y;
+		r[0][2] = -static_cast<T>(2) * normal.x * normal.z;
+
+		r[1][0] = -static_cast<T>(2) * normal.x * normal.y;
+		r[1][1] = static_cast<T>(1) - static_cast<T>(2) * normal.y * normal.y;
+		r[1][2] = -static_cast<T>(2) * normal.y * normal.z;
+
+		r[2][0] = -static_cast<T>(2) * normal.x * normal.z;
+		r[2][1] = -static_cast<T>(2) * normal.y * normal.z;
+		r[2][2] = static_cast<T>(1) - static_cast<T>(2) * normal.z * normal.z;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<3, 3, T, Q> proj2D(
+		const mat<3, 3, T, Q>& m,
+		const vec<3, T, Q>& normal)
+	{
+		mat<3, 3, T, Q> r(static_cast<T>(1));
+		r[0][0] = static_cast<T>(1) - normal.x * normal.x;
+		r[0][1] = - normal.x * normal.y;
+		r[1][0] = - normal.x * normal.y;
+		r[1][1] = static_cast<T>(1) - normal.y * normal.y;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> proj3D(
+		const mat<4, 4, T, Q>& m,
+		const vec<3, T, Q>& normal)
+	{
+		mat<4, 4, T, Q> r(static_cast<T>(1));
+		r[0][0] = static_cast<T>(1) - normal.x * normal.x;
+		r[0][1] = - normal.x * normal.y;
+		r[0][2] = - normal.x * normal.z;
+		r[1][0] = - normal.x * normal.y;
+		r[1][1] = static_cast<T>(1) - normal.y * normal.y;
+		r[1][2] = - normal.y * normal.z;
+		r[2][0] = - normal.x * normal.z;
+		r[2][1] = - normal.y * normal.z;
+		r[2][2] = static_cast<T>(1) - normal.z * normal.z;
+		return m * r;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> scaleBias(T scale, T bias)
+	{
+		mat<4, 4, T, Q> result;
+		result[3] = vec<4, T, Q>(vec<3, T, Q>(bias), static_cast<T>(1));
+		result[0][0] = scale;
+		result[1][1] = scale;
+		result[2][2] = scale;
+		return result;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER mat<4, 4, T, Q> scaleBias(mat<4, 4, T, Q> const& m, T scale, T bias)
+	{
+		return m * scaleBias<T, Q>(scale, bias);
+	}
+}//namespace glm
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.hpp
new file mode 100644
index 000000000000..2ae522c1fc7e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.hpp
@@ -0,0 +1,982 @@
+/// @ref gtx_type_aligned
+/// @file glm/gtx/type_aligned.hpp
+///
+/// @see core (dependence)
+/// @see gtc_quaternion (dependence)
+///
+/// @defgroup gtx_type_aligned GLM_GTX_type_aligned
+/// @ingroup gtx
+///
+/// Include <glm/gtx/type_aligned.hpp> to use the features of this extension.
+///
+/// Defines aligned types.
+
+#pragma once
+
+// Dependency:
+#include "../gtc/type_precision.hpp"
+#include "../gtc/quaternion.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_type_aligned is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_type_aligned extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	///////////////////////////
+	// Signed int vector types
+
+	/// @addtogroup gtx_type_aligned
+	/// @{
+
+	/// Low qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int8, aligned_lowp_int8, 1);
+
+	/// Low qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int16, aligned_lowp_int16, 2);
+
+	/// Low qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int32, aligned_lowp_int32, 4);
+
+	/// Low qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int64, aligned_lowp_int64, 8);
+
+
+	/// Low qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int8_t, aligned_lowp_int8_t, 1);
+
+	/// Low qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int16_t, aligned_lowp_int16_t, 2);
+
+	/// Low qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int32_t, aligned_lowp_int32_t, 4);
+
+	/// Low qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_int64_t, aligned_lowp_int64_t, 8);
+
+
+	/// Low qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_i8, aligned_lowp_i8, 1);
+
+	/// Low qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_i16, aligned_lowp_i16, 2);
+
+	/// Low qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_i32, aligned_lowp_i32, 4);
+
+	/// Low qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_i64, aligned_lowp_i64, 8);
+
+
+	/// Medium qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int8, aligned_mediump_int8, 1);
+
+	/// Medium qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int16, aligned_mediump_int16, 2);
+
+	/// Medium qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int32, aligned_mediump_int32, 4);
+
+	/// Medium qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int64, aligned_mediump_int64, 8);
+
+
+	/// Medium qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int8_t, aligned_mediump_int8_t, 1);
+
+	/// Medium qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int16_t, aligned_mediump_int16_t, 2);
+
+	/// Medium qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int32_t, aligned_mediump_int32_t, 4);
+
+	/// Medium qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_int64_t, aligned_mediump_int64_t, 8);
+
+
+	/// Medium qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_i8, aligned_mediump_i8, 1);
+
+	/// Medium qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_i16, aligned_mediump_i16, 2);
+
+	/// Medium qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_i32, aligned_mediump_i32, 4);
+
+	/// Medium qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_i64, aligned_mediump_i64, 8);
+
+
+	/// High qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int8, aligned_highp_int8, 1);
+
+	/// High qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int16, aligned_highp_int16, 2);
+
+	/// High qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int32, aligned_highp_int32, 4);
+
+	/// High qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int64, aligned_highp_int64, 8);
+
+
+	/// High qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int8_t, aligned_highp_int8_t, 1);
+
+	/// High qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int16_t, aligned_highp_int16_t, 2);
+
+	/// High qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int32_t, aligned_highp_int32_t, 4);
+
+	/// High qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_int64_t, aligned_highp_int64_t, 8);
+
+
+	/// High qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_i8, aligned_highp_i8, 1);
+
+	/// High qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_i16, aligned_highp_i16, 2);
+
+	/// High qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_i32, aligned_highp_i32, 4);
+
+	/// High qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_i64, aligned_highp_i64, 8);
+
+
+	/// Default qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int8, aligned_int8, 1);
+
+	/// Default qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int16, aligned_int16, 2);
+
+	/// Default qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int32, aligned_int32, 4);
+
+	/// Default qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int64, aligned_int64, 8);
+
+
+	/// Default qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int8_t, aligned_int8_t, 1);
+
+	/// Default qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int16_t, aligned_int16_t, 2);
+
+	/// Default qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int32_t, aligned_int32_t, 4);
+
+	/// Default qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(int64_t, aligned_int64_t, 8);
+
+
+	/// Default qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i8, aligned_i8, 1);
+
+	/// Default qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i16, aligned_i16, 2);
+
+	/// Default qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i32, aligned_i32, 4);
+
+	/// Default qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i64, aligned_i64, 8);
+
+
+	/// Default qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(ivec1, aligned_ivec1, 4);
+
+	/// Default qualifier 32 bit signed integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(ivec2, aligned_ivec2, 8);
+
+	/// Default qualifier 32 bit signed integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(ivec3, aligned_ivec3, 16);
+
+	/// Default qualifier 32 bit signed integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(ivec4, aligned_ivec4, 16);
+
+
+	/// Default qualifier 8 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i8vec1, aligned_i8vec1, 1);
+
+	/// Default qualifier 8 bit signed integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i8vec2, aligned_i8vec2, 2);
+
+	/// Default qualifier 8 bit signed integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i8vec3, aligned_i8vec3, 4);
+
+	/// Default qualifier 8 bit signed integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i8vec4, aligned_i8vec4, 4);
+
+
+	/// Default qualifier 16 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i16vec1, aligned_i16vec1, 2);
+
+	/// Default qualifier 16 bit signed integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i16vec2, aligned_i16vec2, 4);
+
+	/// Default qualifier 16 bit signed integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i16vec3, aligned_i16vec3, 8);
+
+	/// Default qualifier 16 bit signed integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i16vec4, aligned_i16vec4, 8);
+
+
+	/// Default qualifier 32 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i32vec1, aligned_i32vec1, 4);
+
+	/// Default qualifier 32 bit signed integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i32vec2, aligned_i32vec2, 8);
+
+	/// Default qualifier 32 bit signed integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i32vec3, aligned_i32vec3, 16);
+
+	/// Default qualifier 32 bit signed integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i32vec4, aligned_i32vec4, 16);
+
+
+	/// Default qualifier 64 bit signed integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i64vec1, aligned_i64vec1, 8);
+
+	/// Default qualifier 64 bit signed integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i64vec2, aligned_i64vec2, 16);
+
+	/// Default qualifier 64 bit signed integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i64vec3, aligned_i64vec3, 32);
+
+	/// Default qualifier 64 bit signed integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(i64vec4, aligned_i64vec4, 32);
+
+
+	/////////////////////////////
+	// Unsigned int vector types
+
+	/// Low qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint8, aligned_lowp_uint8, 1);
+
+	/// Low qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint16, aligned_lowp_uint16, 2);
+
+	/// Low qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint32, aligned_lowp_uint32, 4);
+
+	/// Low qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint64, aligned_lowp_uint64, 8);
+
+
+	/// Low qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint8_t, aligned_lowp_uint8_t, 1);
+
+	/// Low qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint16_t, aligned_lowp_uint16_t, 2);
+
+	/// Low qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint32_t, aligned_lowp_uint32_t, 4);
+
+	/// Low qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_uint64_t, aligned_lowp_uint64_t, 8);
+
+
+	/// Low qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_u8, aligned_lowp_u8, 1);
+
+	/// Low qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_u16, aligned_lowp_u16, 2);
+
+	/// Low qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_u32, aligned_lowp_u32, 4);
+
+	/// Low qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(lowp_u64, aligned_lowp_u64, 8);
+
+
+	/// Medium qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint8, aligned_mediump_uint8, 1);
+
+	/// Medium qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint16, aligned_mediump_uint16, 2);
+
+	/// Medium qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint32, aligned_mediump_uint32, 4);
+
+	/// Medium qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint64, aligned_mediump_uint64, 8);
+
+
+	/// Medium qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint8_t, aligned_mediump_uint8_t, 1);
+
+	/// Medium qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint16_t, aligned_mediump_uint16_t, 2);
+
+	/// Medium qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint32_t, aligned_mediump_uint32_t, 4);
+
+	/// Medium qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_uint64_t, aligned_mediump_uint64_t, 8);
+
+
+	/// Medium qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_u8, aligned_mediump_u8, 1);
+
+	/// Medium qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_u16, aligned_mediump_u16, 2);
+
+	/// Medium qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_u32, aligned_mediump_u32, 4);
+
+	/// Medium qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mediump_u64, aligned_mediump_u64, 8);
+
+
+	/// High qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint8, aligned_highp_uint8, 1);
+
+	/// High qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint16, aligned_highp_uint16, 2);
+
+	/// High qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint32, aligned_highp_uint32, 4);
+
+	/// High qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint64, aligned_highp_uint64, 8);
+
+
+	/// High qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint8_t, aligned_highp_uint8_t, 1);
+
+	/// High qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint16_t, aligned_highp_uint16_t, 2);
+
+	/// High qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint32_t, aligned_highp_uint32_t, 4);
+
+	/// High qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_uint64_t, aligned_highp_uint64_t, 8);
+
+
+	/// High qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_u8, aligned_highp_u8, 1);
+
+	/// High qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_u16, aligned_highp_u16, 2);
+
+	/// High qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_u32, aligned_highp_u32, 4);
+
+	/// High qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(highp_u64, aligned_highp_u64, 8);
+
+
+	/// Default qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint8, aligned_uint8, 1);
+
+	/// Default qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint16, aligned_uint16, 2);
+
+	/// Default qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint32, aligned_uint32, 4);
+
+	/// Default qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint64, aligned_uint64, 8);
+
+
+	/// Default qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint8_t, aligned_uint8_t, 1);
+
+	/// Default qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint16_t, aligned_uint16_t, 2);
+
+	/// Default qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint32_t, aligned_uint32_t, 4);
+
+	/// Default qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uint64_t, aligned_uint64_t, 8);
+
+
+	/// Default qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u8, aligned_u8, 1);
+
+	/// Default qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u16, aligned_u16, 2);
+
+	/// Default qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u32, aligned_u32, 4);
+
+	/// Default qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u64, aligned_u64, 8);
+
+
+	/// Default qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uvec1, aligned_uvec1, 4);
+
+	/// Default qualifier 32 bit unsigned integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uvec2, aligned_uvec2, 8);
+
+	/// Default qualifier 32 bit unsigned integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uvec3, aligned_uvec3, 16);
+
+	/// Default qualifier 32 bit unsigned integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(uvec4, aligned_uvec4, 16);
+
+
+	/// Default qualifier 8 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u8vec1, aligned_u8vec1, 1);
+
+	/// Default qualifier 8 bit unsigned integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u8vec2, aligned_u8vec2, 2);
+
+	/// Default qualifier 8 bit unsigned integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u8vec3, aligned_u8vec3, 4);
+
+	/// Default qualifier 8 bit unsigned integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u8vec4, aligned_u8vec4, 4);
+
+
+	/// Default qualifier 16 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u16vec1, aligned_u16vec1, 2);
+
+	/// Default qualifier 16 bit unsigned integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u16vec2, aligned_u16vec2, 4);
+
+	/// Default qualifier 16 bit unsigned integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u16vec3, aligned_u16vec3, 8);
+
+	/// Default qualifier 16 bit unsigned integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u16vec4, aligned_u16vec4, 8);
+
+
+	/// Default qualifier 32 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u32vec1, aligned_u32vec1, 4);
+
+	/// Default qualifier 32 bit unsigned integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u32vec2, aligned_u32vec2, 8);
+
+	/// Default qualifier 32 bit unsigned integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u32vec3, aligned_u32vec3, 16);
+
+	/// Default qualifier 32 bit unsigned integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u32vec4, aligned_u32vec4, 16);
+
+
+	/// Default qualifier 64 bit unsigned integer aligned scalar type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u64vec1, aligned_u64vec1, 8);
+
+	/// Default qualifier 64 bit unsigned integer aligned vector of 2 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u64vec2, aligned_u64vec2, 16);
+
+	/// Default qualifier 64 bit unsigned integer aligned vector of 3 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u64vec3, aligned_u64vec3, 32);
+
+	/// Default qualifier 64 bit unsigned integer aligned vector of 4 components type.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(u64vec4, aligned_u64vec4, 32);
+
+
+	//////////////////////
+	// Float vector types
+
+	/// 32 bit single-qualifier floating-point aligned scalar.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(float32, aligned_float32, 4);
+
+	/// 32 bit single-qualifier floating-point aligned scalar.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(float32_t, aligned_float32_t, 4);
+
+	/// 32 bit single-qualifier floating-point aligned scalar.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(float32, aligned_f32, 4);
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+	/// 64 bit double-qualifier floating-point aligned scalar.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(float64, aligned_float64, 8);
+
+	/// 64 bit double-qualifier floating-point aligned scalar.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(float64_t, aligned_float64_t, 8);
+
+	/// 64 bit double-qualifier floating-point aligned scalar.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(float64, aligned_f64, 8);
+
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+
+	/// Single-qualifier floating-point aligned vector of 1 component.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(vec1, aligned_vec1, 4);
+
+	/// Single-qualifier floating-point aligned vector of 2 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(vec2, aligned_vec2, 8);
+
+	/// Single-qualifier floating-point aligned vector of 3 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(vec3, aligned_vec3, 16);
+
+	/// Single-qualifier floating-point aligned vector of 4 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(vec4, aligned_vec4, 16);
+
+
+	/// Single-qualifier floating-point aligned vector of 1 component.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fvec1, aligned_fvec1, 4);
+
+	/// Single-qualifier floating-point aligned vector of 2 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fvec2, aligned_fvec2, 8);
+
+	/// Single-qualifier floating-point aligned vector of 3 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fvec3, aligned_fvec3, 16);
+
+	/// Single-qualifier floating-point aligned vector of 4 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fvec4, aligned_fvec4, 16);
+
+
+	/// Single-qualifier floating-point aligned vector of 1 component.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32vec1, aligned_f32vec1, 4);
+
+	/// Single-qualifier floating-point aligned vector of 2 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32vec2, aligned_f32vec2, 8);
+
+	/// Single-qualifier floating-point aligned vector of 3 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32vec3, aligned_f32vec3, 16);
+
+	/// Single-qualifier floating-point aligned vector of 4 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32vec4, aligned_f32vec4, 16);
+
+
+	/// Double-qualifier floating-point aligned vector of 1 component.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(dvec1, aligned_dvec1, 8);
+
+	/// Double-qualifier floating-point aligned vector of 2 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(dvec2, aligned_dvec2, 16);
+
+	/// Double-qualifier floating-point aligned vector of 3 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(dvec3, aligned_dvec3, 32);
+
+	/// Double-qualifier floating-point aligned vector of 4 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(dvec4, aligned_dvec4, 32);
+
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+	/// Double-qualifier floating-point aligned vector of 1 component.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64vec1, aligned_f64vec1, 8);
+
+	/// Double-qualifier floating-point aligned vector of 2 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64vec2, aligned_f64vec2, 16);
+
+	/// Double-qualifier floating-point aligned vector of 3 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64vec3, aligned_f64vec3, 32);
+
+	/// Double-qualifier floating-point aligned vector of 4 components.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64vec4, aligned_f64vec4, 32);
+
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+	//////////////////////
+	// Float matrix types
+
+	/// Single-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef detail::tmat1<f32> mat1;
+
+	/// Single-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mat2, aligned_mat2, 16);
+
+	/// Single-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mat3, aligned_mat3, 16);
+
+	/// Single-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mat4, aligned_mat4, 16);
+
+
+	/// Single-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef detail::tmat1x1<f32> mat1;
+
+	/// Single-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mat2x2, aligned_mat2x2, 16);
+
+	/// Single-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mat3x3, aligned_mat3x3, 16);
+
+	/// Single-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(mat4x4, aligned_mat4x4, 16);
+
+
+	/// Single-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef detail::tmat1x1<f32> fmat1;
+
+	/// Single-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat2x2, aligned_fmat2, 16);
+
+	/// Single-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat3x3, aligned_fmat3, 16);
+
+	/// Single-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat4x4, aligned_fmat4, 16);
+
+
+	/// Single-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef f32 fmat1x1;
+
+	/// Single-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat2x2, aligned_fmat2x2, 16);
+
+	/// Single-qualifier floating-point aligned 2x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat2x3, aligned_fmat2x3, 16);
+
+	/// Single-qualifier floating-point aligned 2x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat2x4, aligned_fmat2x4, 16);
+
+	/// Single-qualifier floating-point aligned 3x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat3x2, aligned_fmat3x2, 16);
+
+	/// Single-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat3x3, aligned_fmat3x3, 16);
+
+	/// Single-qualifier floating-point aligned 3x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat3x4, aligned_fmat3x4, 16);
+
+	/// Single-qualifier floating-point aligned 4x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat4x2, aligned_fmat4x2, 16);
+
+	/// Single-qualifier floating-point aligned 4x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat4x3, aligned_fmat4x3, 16);
+
+	/// Single-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(fmat4x4, aligned_fmat4x4, 16);
+
+
+	/// Single-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef detail::tmat1x1<f32, defaultp> f32mat1;
+
+	/// Single-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat2x2, aligned_f32mat2, 16);
+
+	/// Single-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat3x3, aligned_f32mat3, 16);
+
+	/// Single-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat4x4, aligned_f32mat4, 16);
+
+
+	/// Single-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef f32 f32mat1x1;
+
+	/// Single-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat2x2, aligned_f32mat2x2, 16);
+
+	/// Single-qualifier floating-point aligned 2x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat2x3, aligned_f32mat2x3, 16);
+
+	/// Single-qualifier floating-point aligned 2x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat2x4, aligned_f32mat2x4, 16);
+
+	/// Single-qualifier floating-point aligned 3x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat3x2, aligned_f32mat3x2, 16);
+
+	/// Single-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat3x3, aligned_f32mat3x3, 16);
+
+	/// Single-qualifier floating-point aligned 3x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat3x4, aligned_f32mat3x4, 16);
+
+	/// Single-qualifier floating-point aligned 4x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat4x2, aligned_f32mat4x2, 16);
+
+	/// Single-qualifier floating-point aligned 4x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat4x3, aligned_f32mat4x3, 16);
+
+	/// Single-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32mat4x4, aligned_f32mat4x4, 16);
+
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+	/// Double-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef detail::tmat1x1<f64, defaultp> f64mat1;
+
+	/// Double-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat2x2, aligned_f64mat2, 32);
+
+	/// Double-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat3x3, aligned_f64mat3, 32);
+
+	/// Double-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat4x4, aligned_f64mat4, 32);
+
+
+	/// Double-qualifier floating-point aligned 1x1 matrix.
+	/// @see gtx_type_aligned
+	//typedef f64 f64mat1x1;
+
+	/// Double-qualifier floating-point aligned 2x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat2x2, aligned_f64mat2x2, 32);
+
+	/// Double-qualifier floating-point aligned 2x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat2x3, aligned_f64mat2x3, 32);
+
+	/// Double-qualifier floating-point aligned 2x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat2x4, aligned_f64mat2x4, 32);
+
+	/// Double-qualifier floating-point aligned 3x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat3x2, aligned_f64mat3x2, 32);
+
+	/// Double-qualifier floating-point aligned 3x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat3x3, aligned_f64mat3x3, 32);
+
+	/// Double-qualifier floating-point aligned 3x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat3x4, aligned_f64mat3x4, 32);
+
+	/// Double-qualifier floating-point aligned 4x2 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat4x2, aligned_f64mat4x2, 32);
+
+	/// Double-qualifier floating-point aligned 4x3 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat4x3, aligned_f64mat4x3, 32);
+
+	/// Double-qualifier floating-point aligned 4x4 matrix.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64mat4x4, aligned_f64mat4x4, 32);
+
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+
+	//////////////////////////
+	// Quaternion types
+
+	/// Single-qualifier floating-point aligned quaternion.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(quat, aligned_quat, 16);
+
+	/// Single-qualifier floating-point aligned quaternion.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(quat, aligned_fquat, 16);
+
+	/// Double-qualifier floating-point aligned quaternion.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(dquat, aligned_dquat, 32);
+
+	/// Single-qualifier floating-point aligned quaternion.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f32quat, aligned_f32quat, 16);
+
+#	ifndef GLM_FORCE_SINGLE_ONLY
+
+	/// Double-qualifier floating-point aligned quaternion.
+	/// @see gtx_type_aligned
+	GLM_ALIGNED_TYPEDEF(f64quat, aligned_f64quat, 32);
+
+#	endif//GLM_FORCE_SINGLE_ONLY
+
+	/// @}
+}//namespace glm
+
+#include "type_aligned.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.inl
new file mode 100644
index 000000000000..54c1b818b64a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_aligned.inl
@@ -0,0 +1,6 @@
+/// @ref gtc_type_aligned
+
+namespace glm
+{
+
+}
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.hpp
new file mode 100644
index 000000000000..56685c8cb98c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.hpp
@@ -0,0 +1,85 @@
+/// @ref gtx_type_trait
+/// @file glm/gtx/type_trait.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_type_trait GLM_GTX_type_trait
+/// @ingroup gtx
+///
+/// Include <glm/gtx/type_trait.hpp> to use the features of this extension.
+///
+/// Defines traits for each type.
+
+#pragma once
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_type_trait is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_type_trait extension included")
+#	endif
+#endif
+
+// Dependency:
+#include "../detail/qualifier.hpp"
+#include "../gtc/quaternion.hpp"
+#include "../gtx/dual_quaternion.hpp"
+
+namespace glm
+{
+	/// @addtogroup gtx_type_trait
+	/// @{
+
+	template<typename T>
+	struct type
+	{
+		static bool const is_vec = false;
+		static bool const is_mat = false;
+		static bool const is_quat = false;
+		static length_t const components = 0;
+		static length_t const cols = 0;
+		static length_t const rows = 0;
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct type<vec<L, T, Q> >
+	{
+		static bool const is_vec = true;
+		static bool const is_mat = false;
+		static bool const is_quat = false;
+		static length_t const components = L;
+	};
+
+	template<length_t C, length_t R, typename T, qualifier Q>
+	struct type<mat<C, R, T, Q> >
+	{
+		static bool const is_vec = false;
+		static bool const is_mat = true;
+		static bool const is_quat = false;
+		static length_t const components = C;
+		static length_t const cols = C;
+		static length_t const rows = R;
+	};
+
+	template<typename T, qualifier Q>
+	struct type<qua<T, Q> >
+	{
+		static bool const is_vec = false;
+		static bool const is_mat = false;
+		static bool const is_quat = true;
+		static length_t const components = 4;
+	};
+
+	template<typename T, qualifier Q>
+	struct type<tdualquat<T, Q> >
+	{
+		static bool const is_vec = false;
+		static bool const is_mat = false;
+		static bool const is_quat = true;
+		static length_t const components = 8;
+	};
+
+	/// @}
+}//namespace glm
+
+#include "type_trait.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.inl
new file mode 100644
index 000000000000..045de959cc21
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/type_trait.inl
@@ -0,0 +1,61 @@
+/// @ref gtx_type_trait
+
+namespace glm
+{
+	template<typename T>
+	bool const type<T>::is_vec;
+	template<typename T>
+	bool const type<T>::is_mat;
+	template<typename T>
+	bool const type<T>::is_quat;
+	template<typename T>
+	length_t const type<T>::components;
+	template<typename T>
+	length_t const type<T>::cols;
+	template<typename T>
+	length_t const type<T>::rows;
+
+	// vec
+	template<length_t L, typename T, qualifier Q>
+	bool const type<vec<L, T, Q> >::is_vec;
+	template<length_t L, typename T, qualifier Q>
+	bool const type<vec<L, T, Q> >::is_mat;
+	template<length_t L, typename T, qualifier Q>
+	bool const type<vec<L, T, Q> >::is_quat;
+	template<length_t L, typename T, qualifier Q>
+	length_t const type<vec<L, T, Q> >::components;
+
+	// mat
+	template<length_t C, length_t R, typename T, qualifier Q>
+	bool const type<mat<C, R, T, Q> >::is_vec;
+	template<length_t C, length_t R, typename T, qualifier Q>
+	bool const type<mat<C, R, T, Q> >::is_mat;
+	template<length_t C, length_t R, typename T, qualifier Q>
+	bool const type<mat<C, R, T, Q> >::is_quat;
+	template<length_t C, length_t R, typename T, qualifier Q>
+	length_t const type<mat<C, R, T, Q> >::components;
+	template<length_t C, length_t R, typename T, qualifier Q>
+	length_t const type<mat<C, R, T, Q> >::cols;
+	template<length_t C, length_t R, typename T, qualifier Q>
+	length_t const type<mat<C, R, T, Q> >::rows;
+
+	// tquat
+	template<typename T, qualifier Q>
+	bool const type<qua<T, Q> >::is_vec;
+	template<typename T, qualifier Q>
+	bool const type<qua<T, Q> >::is_mat;
+	template<typename T, qualifier Q>
+	bool const type<qua<T, Q> >::is_quat;
+	template<typename T, qualifier Q>
+	length_t const type<qua<T, Q> >::components;
+
+	// tdualquat
+	template<typename T, qualifier Q>
+	bool const type<tdualquat<T, Q> >::is_vec;
+	template<typename T, qualifier Q>
+	bool const type<tdualquat<T, Q> >::is_mat;
+	template<typename T, qualifier Q>
+	bool const type<tdualquat<T, Q> >::is_quat;
+	template<typename T, qualifier Q>
+	length_t const type<tdualquat<T, Q> >::components;
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/vec_swizzle.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/vec_swizzle.hpp
new file mode 100644
index 000000000000..4440c9accd54
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/vec_swizzle.hpp
@@ -0,0 +1,2786 @@
+/// @ref gtx_vec_swizzle
+/// @file glm/gtx/vec_swizzle.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_vec_swizzle GLM_GTX_vec_swizzle
+/// @ingroup gtx
+///
+/// Include <glm/gtx/vec_swizzle.hpp> to use the features of this extension.
+///
+/// Functions to perform swizzle operation.
+
+#pragma once
+
+#include "../glm.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_vec_swizzle is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_vec_swizzle extension included")
+#	endif
+#endif
+
+namespace glm {
+	/// @addtogroup gtx_vec_swizzle
+	/// @{
+
+	// xx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xx(const glm::vec<1, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.x);
+	}
+
+	// xy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.y);
+	}
+
+	// xz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.z);
+	}
+
+	// xw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> xw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.x, v.w);
+	}
+
+	// yx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.x);
+	}
+
+	// yy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.y);
+	}
+
+	// yz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.z);
+	}
+
+	// yw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> yw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.y, v.w);
+	}
+
+	// zx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.x);
+	}
+
+	// zy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.y);
+	}
+
+	// zz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.z);
+	}
+
+	// zw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> zw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.z, v.w);
+	}
+
+	// wx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> wx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.w, v.x);
+	}
+
+	// wy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> wy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.w, v.y);
+	}
+
+	// wz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> wz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.w, v.z);
+	}
+
+	// ww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<2, T, Q> ww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<2, T, Q>(v.w, v.w);
+	}
+
+	// xxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxx(const glm::vec<1, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.x);
+	}
+
+	// xxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.y);
+	}
+
+	// xxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.z);
+	}
+
+	// xxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.x, v.w);
+	}
+
+	// xyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.x);
+	}
+
+	// xyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.y);
+	}
+
+	// xyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.z);
+	}
+
+	// xyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.y, v.w);
+	}
+
+	// xzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.x);
+	}
+
+	// xzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.y);
+	}
+
+	// xzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.z);
+	}
+
+	// xzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.z, v.w);
+	}
+
+	// xwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.w, v.x);
+	}
+
+	// xwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.w, v.y);
+	}
+
+	// xwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.w, v.z);
+	}
+
+	// xww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> xww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.x, v.w, v.w);
+	}
+
+	// yxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.x);
+	}
+
+	// yxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.y);
+	}
+
+	// yxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.z);
+	}
+
+	// yxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.x, v.w);
+	}
+
+	// yyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.x);
+	}
+
+	// yyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.y);
+	}
+
+	// yyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.z);
+	}
+
+	// yyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.y, v.w);
+	}
+
+	// yzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.x);
+	}
+
+	// yzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.y);
+	}
+
+	// yzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.z);
+	}
+
+	// yzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.z, v.w);
+	}
+
+	// ywx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> ywx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.w, v.x);
+	}
+
+	// ywy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> ywy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.w, v.y);
+	}
+
+	// ywz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> ywz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.w, v.z);
+	}
+
+	// yww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> yww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.y, v.w, v.w);
+	}
+
+	// zxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.x);
+	}
+
+	// zxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.y);
+	}
+
+	// zxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.z);
+	}
+
+	// zxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.x, v.w);
+	}
+
+	// zyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.x);
+	}
+
+	// zyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.y);
+	}
+
+	// zyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.z);
+	}
+
+	// zyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.y, v.w);
+	}
+
+	// zzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.x);
+	}
+
+	// zzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.y);
+	}
+
+	// zzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.z);
+	}
+
+	// zzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.z, v.w);
+	}
+
+	// zwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.w, v.x);
+	}
+
+	// zwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.w, v.y);
+	}
+
+	// zwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.w, v.z);
+	}
+
+	// zww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> zww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.z, v.w, v.w);
+	}
+
+	// wxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.x, v.x);
+	}
+
+	// wxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.x, v.y);
+	}
+
+	// wxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.x, v.z);
+	}
+
+	// wxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.x, v.w);
+	}
+
+	// wyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.y, v.x);
+	}
+
+	// wyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.y, v.y);
+	}
+
+	// wyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.y, v.z);
+	}
+
+	// wyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.y, v.w);
+	}
+
+	// wzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.z, v.x);
+	}
+
+	// wzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.z, v.y);
+	}
+
+	// wzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.z, v.z);
+	}
+
+	// wzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.z, v.w);
+	}
+
+	// wwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.w, v.x);
+	}
+
+	// wwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.w, v.y);
+	}
+
+	// wwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> wwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.w, v.z);
+	}
+
+	// www
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<3, T, Q> www(const glm::vec<4, T, Q> &v) {
+		return glm::vec<3, T, Q>(v.w, v.w, v.w);
+	}
+
+	// xxxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxx(const glm::vec<1, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.x);
+	}
+
+	// xxxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.y);
+	}
+
+	// xxxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.z);
+	}
+
+	// xxxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.x, v.w);
+	}
+
+	// xxyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.x);
+	}
+
+	// xxyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.y);
+	}
+
+	// xxyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.z);
+	}
+
+	// xxyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.y, v.w);
+	}
+
+	// xxzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.x);
+	}
+
+	// xxzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.y);
+	}
+
+	// xxzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.z);
+	}
+
+	// xxzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.z, v.w);
+	}
+
+	// xxwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.w, v.x);
+	}
+
+	// xxwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.w, v.y);
+	}
+
+	// xxwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.w, v.z);
+	}
+
+	// xxww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xxww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.x, v.w, v.w);
+	}
+
+	// xyxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.x);
+	}
+
+	// xyxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.y);
+	}
+
+	// xyxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.z);
+	}
+
+	// xyxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.x, v.w);
+	}
+
+	// xyyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.x);
+	}
+
+	// xyyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.y);
+	}
+
+	// xyyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.z);
+	}
+
+	// xyyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.y, v.w);
+	}
+
+	// xyzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.x);
+	}
+
+	// xyzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.y);
+	}
+
+	// xyzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.z);
+	}
+
+	// xyzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.z, v.w);
+	}
+
+	// xywx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xywx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.w, v.x);
+	}
+
+	// xywy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xywy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.w, v.y);
+	}
+
+	// xywz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xywz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.w, v.z);
+	}
+
+	// xyww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xyww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.y, v.w, v.w);
+	}
+
+	// xzxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.x);
+	}
+
+	// xzxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.y);
+	}
+
+	// xzxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.z);
+	}
+
+	// xzxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.x, v.w);
+	}
+
+	// xzyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.x);
+	}
+
+	// xzyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.y);
+	}
+
+	// xzyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.z);
+	}
+
+	// xzyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.y, v.w);
+	}
+
+	// xzzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.x);
+	}
+
+	// xzzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.y);
+	}
+
+	// xzzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.z);
+	}
+
+	// xzzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.z, v.w);
+	}
+
+	// xzwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.w, v.x);
+	}
+
+	// xzwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.w, v.y);
+	}
+
+	// xzwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.w, v.z);
+	}
+
+	// xzww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xzww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.z, v.w, v.w);
+	}
+
+	// xwxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.x, v.x);
+	}
+
+	// xwxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.x, v.y);
+	}
+
+	// xwxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.x, v.z);
+	}
+
+	// xwxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.x, v.w);
+	}
+
+	// xwyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.y, v.x);
+	}
+
+	// xwyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.y, v.y);
+	}
+
+	// xwyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.y, v.z);
+	}
+
+	// xwyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.y, v.w);
+	}
+
+	// xwzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.z, v.x);
+	}
+
+	// xwzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.z, v.y);
+	}
+
+	// xwzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.z, v.z);
+	}
+
+	// xwzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.z, v.w);
+	}
+
+	// xwwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.w, v.x);
+	}
+
+	// xwwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.w, v.y);
+	}
+
+	// xwwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.w, v.z);
+	}
+
+	// xwww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> xwww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.x, v.w, v.w, v.w);
+	}
+
+	// yxxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.x);
+	}
+
+	// yxxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.y);
+	}
+
+	// yxxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.z);
+	}
+
+	// yxxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.x, v.w);
+	}
+
+	// yxyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.x);
+	}
+
+	// yxyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.y);
+	}
+
+	// yxyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.z);
+	}
+
+	// yxyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.y, v.w);
+	}
+
+	// yxzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.x);
+	}
+
+	// yxzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.y);
+	}
+
+	// yxzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.z);
+	}
+
+	// yxzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.z, v.w);
+	}
+
+	// yxwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.w, v.x);
+	}
+
+	// yxwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.w, v.y);
+	}
+
+	// yxwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.w, v.z);
+	}
+
+	// yxww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yxww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.x, v.w, v.w);
+	}
+
+	// yyxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.x);
+	}
+
+	// yyxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.y);
+	}
+
+	// yyxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.z);
+	}
+
+	// yyxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.x, v.w);
+	}
+
+	// yyyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyx(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.x);
+	}
+
+	// yyyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyy(const glm::vec<2, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.y);
+	}
+
+	// yyyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.z);
+	}
+
+	// yyyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.y, v.w);
+	}
+
+	// yyzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.x);
+	}
+
+	// yyzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.y);
+	}
+
+	// yyzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.z);
+	}
+
+	// yyzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.z, v.w);
+	}
+
+	// yywx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yywx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.w, v.x);
+	}
+
+	// yywy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yywy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.w, v.y);
+	}
+
+	// yywz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yywz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.w, v.z);
+	}
+
+	// yyww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yyww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.y, v.w, v.w);
+	}
+
+	// yzxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.x);
+	}
+
+	// yzxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.y);
+	}
+
+	// yzxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.z);
+	}
+
+	// yzxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.x, v.w);
+	}
+
+	// yzyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.x);
+	}
+
+	// yzyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.y);
+	}
+
+	// yzyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.z);
+	}
+
+	// yzyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.y, v.w);
+	}
+
+	// yzzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.x);
+	}
+
+	// yzzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.y);
+	}
+
+	// yzzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.z);
+	}
+
+	// yzzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.z, v.w);
+	}
+
+	// yzwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.w, v.x);
+	}
+
+	// yzwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.w, v.y);
+	}
+
+	// yzwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.w, v.z);
+	}
+
+	// yzww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> yzww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.z, v.w, v.w);
+	}
+
+	// ywxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.x, v.x);
+	}
+
+	// ywxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.x, v.y);
+	}
+
+	// ywxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.x, v.z);
+	}
+
+	// ywxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.x, v.w);
+	}
+
+	// ywyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.y, v.x);
+	}
+
+	// ywyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.y, v.y);
+	}
+
+	// ywyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.y, v.z);
+	}
+
+	// ywyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.y, v.w);
+	}
+
+	// ywzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.z, v.x);
+	}
+
+	// ywzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.z, v.y);
+	}
+
+	// ywzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.z, v.z);
+	}
+
+	// ywzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.z, v.w);
+	}
+
+	// ywwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.w, v.x);
+	}
+
+	// ywwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.w, v.y);
+	}
+
+	// ywwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.w, v.z);
+	}
+
+	// ywww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> ywww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.y, v.w, v.w, v.w);
+	}
+
+	// zxxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.x);
+	}
+
+	// zxxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.y);
+	}
+
+	// zxxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.z);
+	}
+
+	// zxxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.x, v.w);
+	}
+
+	// zxyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.x);
+	}
+
+	// zxyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.y);
+	}
+
+	// zxyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.z);
+	}
+
+	// zxyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.y, v.w);
+	}
+
+	// zxzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.x);
+	}
+
+	// zxzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.y);
+	}
+
+	// zxzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.z);
+	}
+
+	// zxzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.z, v.w);
+	}
+
+	// zxwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.w, v.x);
+	}
+
+	// zxwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.w, v.y);
+	}
+
+	// zxwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.w, v.z);
+	}
+
+	// zxww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zxww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.x, v.w, v.w);
+	}
+
+	// zyxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.x);
+	}
+
+	// zyxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.y);
+	}
+
+	// zyxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.z);
+	}
+
+	// zyxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.x, v.w);
+	}
+
+	// zyyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.x);
+	}
+
+	// zyyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.y);
+	}
+
+	// zyyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.z);
+	}
+
+	// zyyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.y, v.w);
+	}
+
+	// zyzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.x);
+	}
+
+	// zyzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.y);
+	}
+
+	// zyzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.z);
+	}
+
+	// zyzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.z, v.w);
+	}
+
+	// zywx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zywx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.w, v.x);
+	}
+
+	// zywy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zywy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.w, v.y);
+	}
+
+	// zywz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zywz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.w, v.z);
+	}
+
+	// zyww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zyww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.y, v.w, v.w);
+	}
+
+	// zzxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.x);
+	}
+
+	// zzxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.y);
+	}
+
+	// zzxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.z);
+	}
+
+	// zzxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.x, v.w);
+	}
+
+	// zzyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.x);
+	}
+
+	// zzyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.y);
+	}
+
+	// zzyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.z);
+	}
+
+	// zzyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.y, v.w);
+	}
+
+	// zzzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzx(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.x);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.x);
+	}
+
+	// zzzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzy(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.y);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.y);
+	}
+
+	// zzzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzz(const glm::vec<3, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.z);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.z);
+	}
+
+	// zzzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.z, v.w);
+	}
+
+	// zzwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.w, v.x);
+	}
+
+	// zzwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.w, v.y);
+	}
+
+	// zzwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.w, v.z);
+	}
+
+	// zzww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zzww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.z, v.w, v.w);
+	}
+
+	// zwxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.x, v.x);
+	}
+
+	// zwxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.x, v.y);
+	}
+
+	// zwxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.x, v.z);
+	}
+
+	// zwxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.x, v.w);
+	}
+
+	// zwyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.y, v.x);
+	}
+
+	// zwyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.y, v.y);
+	}
+
+	// zwyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.y, v.z);
+	}
+
+	// zwyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.y, v.w);
+	}
+
+	// zwzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.z, v.x);
+	}
+
+	// zwzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.z, v.y);
+	}
+
+	// zwzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.z, v.z);
+	}
+
+	// zwzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.z, v.w);
+	}
+
+	// zwwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.w, v.x);
+	}
+
+	// zwwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.w, v.y);
+	}
+
+	// zwwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.w, v.z);
+	}
+
+	// zwww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> zwww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.z, v.w, v.w, v.w);
+	}
+
+	// wxxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.x, v.x);
+	}
+
+	// wxxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.x, v.y);
+	}
+
+	// wxxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.x, v.z);
+	}
+
+	// wxxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.x, v.w);
+	}
+
+	// wxyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.y, v.x);
+	}
+
+	// wxyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.y, v.y);
+	}
+
+	// wxyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.y, v.z);
+	}
+
+	// wxyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.y, v.w);
+	}
+
+	// wxzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.z, v.x);
+	}
+
+	// wxzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.z, v.y);
+	}
+
+	// wxzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.z, v.z);
+	}
+
+	// wxzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.z, v.w);
+	}
+
+	// wxwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.w, v.x);
+	}
+
+	// wxwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.w, v.y);
+	}
+
+	// wxwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.w, v.z);
+	}
+
+	// wxww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wxww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.x, v.w, v.w);
+	}
+
+	// wyxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.x, v.x);
+	}
+
+	// wyxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.x, v.y);
+	}
+
+	// wyxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.x, v.z);
+	}
+
+	// wyxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.x, v.w);
+	}
+
+	// wyyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.y, v.x);
+	}
+
+	// wyyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.y, v.y);
+	}
+
+	// wyyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.y, v.z);
+	}
+
+	// wyyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.y, v.w);
+	}
+
+	// wyzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.z, v.x);
+	}
+
+	// wyzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.z, v.y);
+	}
+
+	// wyzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.z, v.z);
+	}
+
+	// wyzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.z, v.w);
+	}
+
+	// wywx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wywx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.w, v.x);
+	}
+
+	// wywy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wywy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.w, v.y);
+	}
+
+	// wywz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wywz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.w, v.z);
+	}
+
+	// wyww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wyww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.y, v.w, v.w);
+	}
+
+	// wzxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.x, v.x);
+	}
+
+	// wzxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.x, v.y);
+	}
+
+	// wzxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.x, v.z);
+	}
+
+	// wzxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.x, v.w);
+	}
+
+	// wzyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.y, v.x);
+	}
+
+	// wzyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.y, v.y);
+	}
+
+	// wzyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.y, v.z);
+	}
+
+	// wzyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.y, v.w);
+	}
+
+	// wzzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.z, v.x);
+	}
+
+	// wzzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.z, v.y);
+	}
+
+	// wzzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.z, v.z);
+	}
+
+	// wzzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.z, v.w);
+	}
+
+	// wzwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.w, v.x);
+	}
+
+	// wzwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.w, v.y);
+	}
+
+	// wzwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.w, v.z);
+	}
+
+	// wzww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wzww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.z, v.w, v.w);
+	}
+
+	// wwxx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwxx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.x, v.x);
+	}
+
+	// wwxy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwxy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.x, v.y);
+	}
+
+	// wwxz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwxz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.x, v.z);
+	}
+
+	// wwxw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwxw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.x, v.w);
+	}
+
+	// wwyx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwyx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.y, v.x);
+	}
+
+	// wwyy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwyy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.y, v.y);
+	}
+
+	// wwyz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwyz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.y, v.z);
+	}
+
+	// wwyw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwyw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.y, v.w);
+	}
+
+	// wwzx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwzx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.z, v.x);
+	}
+
+	// wwzy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwzy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.z, v.y);
+	}
+
+	// wwzz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwzz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.z, v.z);
+	}
+
+	// wwzw
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwzw(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.z, v.w);
+	}
+
+	// wwwx
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwwx(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.w, v.x);
+	}
+
+	// wwwy
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwwy(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.w, v.y);
+	}
+
+	// wwwz
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwwz(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.w, v.z);
+	}
+
+	// wwww
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER glm::vec<4, T, Q> wwww(const glm::vec<4, T, Q> &v) {
+		return glm::vec<4, T, Q>(v.w, v.w, v.w, v.w);
+	}
+
+	/// @}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.hpp
new file mode 100644
index 000000000000..9ae437126b19
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.hpp
@@ -0,0 +1,57 @@
+/// @ref gtx_vector_angle
+/// @file glm/gtx/vector_angle.hpp
+///
+/// @see core (dependence)
+/// @see gtx_quaternion (dependence)
+/// @see gtx_epsilon (dependence)
+///
+/// @defgroup gtx_vector_angle GLM_GTX_vector_angle
+/// @ingroup gtx
+///
+/// Include <glm/gtx/vector_angle.hpp> to use the features of this extension.
+///
+/// Compute angle between vectors
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../gtc/epsilon.hpp"
+#include "../gtx/quaternion.hpp"
+#include "../gtx/rotate_vector.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_vector_angle is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_vector_angle extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_vector_angle
+	/// @{
+
+	//! Returns the absolute angle between two vectors.
+	//! Parameters need to be normalized.
+	/// @see gtx_vector_angle extension.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL T angle(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	//! Returns the oriented angle between two 2d vectors.
+	//! Parameters need to be normalized.
+	/// @see gtx_vector_angle extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T orientedAngle(vec<2, T, Q> const& x, vec<2, T, Q> const& y);
+
+	//! Returns the oriented angle between two 3d vectors based from a reference axis.
+	//! Parameters need to be normalized.
+	/// @see gtx_vector_angle extension.
+	template<typename T, qualifier Q>
+	GLM_FUNC_DECL T orientedAngle(vec<3, T, Q> const& x, vec<3, T, Q> const& y, vec<3, T, Q> const& ref);
+
+	/// @}
+}// namespace glm
+
+#include "vector_angle.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.inl
new file mode 100644
index 000000000000..11e1a218372e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_angle.inl
@@ -0,0 +1,45 @@
+/// @ref gtx_vector_angle
+
+namespace glm
+{
+	template<typename genType>
+	GLM_FUNC_QUALIFIER genType angle
+	(
+		genType const& x,
+		genType const& y
+	)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genType>::is_iec559, "'angle' only accept floating-point inputs");
+		return acos(clamp(dot(x, y), genType(-1), genType(1)));
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T angle(vec<L, T, Q> const& x, vec<L, T, Q> const& y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'angle' only accept floating-point inputs");
+		return acos(clamp(dot(x, y), T(-1), T(1)));
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T orientedAngle(vec<2, T, Q> const& x, vec<2, T, Q> const& y)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'orientedAngle' only accept floating-point inputs");
+		T const Angle(acos(clamp(dot(x, y), T(-1), T(1))));
+
+		T const partialCross = x.x * y.y - y.x * x.y;
+
+		if (partialCross > T(0))
+			return Angle;
+		else
+			return -Angle;
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER T orientedAngle(vec<3, T, Q> const& x, vec<3, T, Q> const& y, vec<3, T, Q> const& ref)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559 || GLM_CONFIG_UNRESTRICTED_FLOAT, "'orientedAngle' only accept floating-point inputs");
+
+		T const Angle(acos(clamp(dot(x, y), T(-1), T(1))));
+		return mix(Angle, -Angle, dot(ref, cross(x, y)) < T(0));
+	}
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.hpp
new file mode 100644
index 000000000000..af1f7b9bd584
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.hpp
@@ -0,0 +1,66 @@
+/// @ref gtx_vector_query
+/// @file glm/gtx/vector_query.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_vector_query GLM_GTX_vector_query
+/// @ingroup gtx
+///
+/// Include <glm/gtx/vector_query.hpp> to use the features of this extension.
+///
+/// Query information of vector types
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include <cfloat>
+#include <limits>
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_vector_query is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_vector_query extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_vector_query
+	/// @{
+
+	//! Check whether two vectors are collinears.
+	/// @see gtx_vector_query extensions.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL bool areCollinear(vec<L, T, Q> const& v0, vec<L, T, Q> const& v1, T const& epsilon);
+
+	//! Check whether two vectors are orthogonals.
+	/// @see gtx_vector_query extensions.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL bool areOrthogonal(vec<L, T, Q> const& v0, vec<L, T, Q> const& v1, T const& epsilon);
+
+	//! Check whether a vector is normalized.
+	/// @see gtx_vector_query extensions.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNormalized(vec<L, T, Q> const& v, T const& epsilon);
+
+	//! Check whether a vector is null.
+	/// @see gtx_vector_query extensions.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL bool isNull(vec<L, T, Q> const& v, T const& epsilon);
+
+	//! Check whether a each component of a vector is null.
+	/// @see gtx_vector_query extensions.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, bool, Q> isCompNull(vec<L, T, Q> const& v, T const& epsilon);
+
+	//! Check whether two vectors are orthonormal.
+	/// @see gtx_vector_query extensions.
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL bool areOrthonormal(vec<L, T, Q> const& v0, vec<L, T, Q> const& v1, T const& epsilon);
+
+	/// @}
+}// namespace glm
+
+#include "vector_query.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.inl
new file mode 100644
index 000000000000..d1a5c9be46b1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/vector_query.inl
@@ -0,0 +1,154 @@
+/// @ref gtx_vector_query
+
+#include <cassert>
+
+namespace glm{
+namespace detail
+{
+	template<length_t L, typename T, qualifier Q>
+	struct compute_areCollinear{};
+
+	template<typename T, qualifier Q>
+	struct compute_areCollinear<2, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static bool call(vec<2, T, Q> const& v0, vec<2, T, Q> const& v1, T const& epsilon)
+		{
+			return length(cross(vec<3, T, Q>(v0, static_cast<T>(0)), vec<3, T, Q>(v1, static_cast<T>(0)))) < epsilon;
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_areCollinear<3, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static bool call(vec<3, T, Q> const& v0, vec<3, T, Q> const& v1, T const& epsilon)
+		{
+			return length(cross(v0, v1)) < epsilon;
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_areCollinear<4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static bool call(vec<4, T, Q> const& v0, vec<4, T, Q> const& v1, T const& epsilon)
+		{
+			return length(cross(vec<3, T, Q>(v0), vec<3, T, Q>(v1))) < epsilon;
+		}
+	};
+
+	template<length_t L, typename T, qualifier Q>
+	struct compute_isCompNull{};
+
+	template<typename T, qualifier Q>
+	struct compute_isCompNull<2, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<2, bool, Q> call(vec<2, T, Q> const& v, T const& epsilon)
+		{
+			return vec<2, bool, Q>(
+				(abs(v.x) < epsilon),
+				(abs(v.y) < epsilon));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_isCompNull<3, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<3, bool, Q> call(vec<3, T, Q> const& v, T const& epsilon)
+		{
+			return vec<3, bool, Q>(
+				(abs(v.x) < epsilon),
+				(abs(v.y) < epsilon),
+				(abs(v.z) < epsilon));
+		}
+	};
+
+	template<typename T, qualifier Q>
+	struct compute_isCompNull<4, T, Q>
+	{
+		GLM_FUNC_QUALIFIER static vec<4, bool, Q> call(vec<4, T, Q> const& v, T const& epsilon)
+		{
+			return vec<4, bool, Q>(
+				(abs(v.x) < epsilon),
+				(abs(v.y) < epsilon),
+				(abs(v.z) < epsilon),
+				(abs(v.w) < epsilon));
+		}
+	};
+
+}//namespace detail
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool areCollinear(vec<L, T, Q> const& v0, vec<L, T, Q> const& v1, T const& epsilon)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'areCollinear' only accept floating-point inputs");
+
+		return detail::compute_areCollinear<L, T, Q>::call(v0, v1, epsilon);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool areOrthogonal(vec<L, T, Q> const& v0, vec<L, T, Q> const& v1, T const& epsilon)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'areOrthogonal' only accept floating-point inputs");
+
+		return abs(dot(v0, v1)) <= max(
+			static_cast<T>(1),
+			length(v0)) * max(static_cast<T>(1), length(v1)) * epsilon;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNormalized(vec<L, T, Q> const& v, T const& epsilon)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'isNormalized' only accept floating-point inputs");
+
+		return abs(length(v) - static_cast<T>(1)) <= static_cast<T>(2) * epsilon;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool isNull(vec<L, T, Q> const& v, T const& epsilon)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'isNull' only accept floating-point inputs");
+
+		return length(v) <= epsilon;
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<L, bool, Q> isCompNull(vec<L, T, Q> const& v, T const& epsilon)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'isCompNull' only accept floating-point inputs");
+
+		return detail::compute_isCompNull<L, T, Q>::call(v, epsilon);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<2, bool, Q> isCompNull(vec<2, T, Q> const& v, T const& epsilon)
+	{
+		return vec<2, bool, Q>(
+			abs(v.x) < epsilon,
+			abs(v.y) < epsilon);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<3, bool, Q> isCompNull(vec<3, T, Q> const& v, T const& epsilon)
+	{
+		return vec<3, bool, Q>(
+			abs(v.x) < epsilon,
+			abs(v.y) < epsilon,
+			abs(v.z) < epsilon);
+	}
+
+	template<typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER vec<4, bool, Q> isCompNull(vec<4, T, Q> const& v, T const& epsilon)
+	{
+		return vec<4, bool, Q>(
+			abs(v.x) < epsilon,
+			abs(v.y) < epsilon,
+			abs(v.z) < epsilon,
+			abs(v.w) < epsilon);
+	}
+
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_QUALIFIER bool areOrthonormal(vec<L, T, Q> const& v0, vec<L, T, Q> const& v1, T const& epsilon)
+	{
+		return isNormalized(v0, epsilon) && isNormalized(v1, epsilon) && (abs(dot(v0, v1)) <= epsilon);
+	}
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.hpp b/thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.hpp
new file mode 100644
index 000000000000..ad4eb3fca740
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.hpp
@@ -0,0 +1,37 @@
+/// @ref gtx_wrap
+/// @file glm/gtx/wrap.hpp
+///
+/// @see core (dependence)
+///
+/// @defgroup gtx_wrap GLM_GTX_wrap
+/// @ingroup gtx
+///
+/// Include <glm/gtx/wrap.hpp> to use the features of this extension.
+///
+/// Wrapping mode of texture coordinates.
+
+#pragma once
+
+// Dependency:
+#include "../glm.hpp"
+#include "../ext/scalar_common.hpp"
+#include "../ext/vector_common.hpp"
+#include "../gtc/vec1.hpp"
+
+#if GLM_MESSAGES == GLM_ENABLE && !defined(GLM_EXT_INCLUDED)
+#	ifndef GLM_ENABLE_EXPERIMENTAL
+#		pragma message("GLM: GLM_GTX_wrap is an experimental extension and may change in the future. Use #define GLM_ENABLE_EXPERIMENTAL before including it, if you really want to use it.")
+#	else
+#		pragma message("GLM: GLM_GTX_wrap extension included")
+#	endif
+#endif
+
+namespace glm
+{
+	/// @addtogroup gtx_wrap
+	/// @{
+
+	/// @}
+}// namespace glm
+
+#include "wrap.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.inl b/thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.inl
new file mode 100644
index 000000000000..4be3b4c38aee
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/gtx/wrap.inl
@@ -0,0 +1,6 @@
+/// @ref gtx_wrap
+
+namespace glm
+{
+
+}//namespace glm
diff --git a/thirdparty/manifold/thirdparty/glm/glm/integer.hpp b/thirdparty/manifold/thirdparty/glm/glm/integer.hpp
new file mode 100644
index 000000000000..8817db3f0a22
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/integer.hpp
@@ -0,0 +1,212 @@
+/// @ref core
+/// @file glm/integer.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+///
+/// @defgroup core_func_integer Integer functions
+/// @ingroup core
+///
+/// Provides GLSL functions on integer types
+///
+/// These all operate component-wise. The description is per component.
+/// The notation [a, b] means the set of bits from bit-number a through bit-number
+/// b, inclusive. The lowest-order bit is bit 0.
+///
+/// Include <glm/integer.hpp> to use these core features.
+
+#pragma once
+
+#include "detail/qualifier.hpp"
+#include "common.hpp"
+#include "vector_relational.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_func_integer
+	/// @{
+
+	/// Adds 32-bit unsigned integer x and y, returning the sum
+	/// modulo pow(2, 32). The value carry is set to 0 if the sum was
+	/// less than pow(2, 32), or to 1 otherwise.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/uaddCarry.xml">GLSL uaddCarry man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, uint, Q> uaddCarry(
+		vec<L, uint, Q> const& x,
+		vec<L, uint, Q> const& y,
+		vec<L, uint, Q> & carry);
+
+	/// Subtracts the 32-bit unsigned integer y from x, returning
+	/// the difference if non-negative, or pow(2, 32) plus the difference
+	/// otherwise. The value borrow is set to 0 if x >= y, or to 1 otherwise.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/usubBorrow.xml">GLSL usubBorrow man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL vec<L, uint, Q> usubBorrow(
+		vec<L, uint, Q> const& x,
+		vec<L, uint, Q> const& y,
+		vec<L, uint, Q> & borrow);
+
+	/// Multiplies 32-bit integers x and y, producing a 64-bit
+	/// result. The 32 least-significant bits are returned in lsb.
+	/// The 32 most-significant bits are returned in msb.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/umulExtended.xml">GLSL umulExtended man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL void umulExtended(
+		vec<L, uint, Q> const& x,
+		vec<L, uint, Q> const& y,
+		vec<L, uint, Q> & msb,
+		vec<L, uint, Q> & lsb);
+
+	/// Multiplies 32-bit integers x and y, producing a 64-bit
+	/// result. The 32 least-significant bits are returned in lsb.
+	/// The 32 most-significant bits are returned in msb.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/imulExtended.xml">GLSL imulExtended man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL void imulExtended(
+		vec<L, int, Q> const& x,
+		vec<L, int, Q> const& y,
+		vec<L, int, Q> & msb,
+		vec<L, int, Q> & lsb);
+
+	/// Extracts bits [offset, offset + bits - 1] from value,
+	/// returning them in the least significant bits of the result.
+	/// For unsigned data types, the most significant bits of the
+	/// result will be set to zero. For signed data types, the
+	/// most significant bits will be set to the value of bit offset + base - 1.
+	///
+	/// If bits is zero, the result will be zero. The result will be
+	/// undefined if offset or bits is negative, or if the sum of
+	/// offset and bits is greater than the number of bits used
+	/// to store the operand.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/bitfieldExtract.xml">GLSL bitfieldExtract man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldExtract(
+		vec<L, T, Q> const& Value,
+		int Offset,
+		int Bits);
+
+	/// Returns the insertion the bits least-significant bits of insert into base.
+	///
+	/// The result will have bits [offset, offset + bits - 1] taken
+	/// from bits [0, bits - 1] of insert, and all other bits taken
+	/// directly from the corresponding bits of base. If bits is
+	/// zero, the result will simply be base. The result will be
+	/// undefined if offset or bits is negative, or if the sum of
+	/// offset and bits is greater than the number of bits used to
+	/// store the operand.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/bitfieldInsert.xml">GLSL bitfieldInsert man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldInsert(
+		vec<L, T, Q> const& Base,
+		vec<L, T, Q> const& Insert,
+		int Offset,
+		int Bits);
+
+	/// Returns the reversal of the bits of value.
+	/// The bit numbered n of the result will be taken from bit (bits - 1) - n of value,
+	/// where bits is the total number of bits used to represent value.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/bitfieldReverse.xml">GLSL bitfieldReverse man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> bitfieldReverse(vec<L, T, Q> const& v);
+
+	/// Returns the number of bits set to 1 in the binary representation of value.
+	///
+	/// @tparam genType Signed or unsigned integer scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/bitCount.xml">GLSL bitCount man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<typename genType>
+	GLM_FUNC_DECL int bitCount(genType v);
+
+	/// Returns the number of bits set to 1 in the binary representation of value.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar or vector types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/bitCount.xml">GLSL bitCount man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> bitCount(vec<L, T, Q> const& v);
+
+	/// Returns the bit number of the least significant bit set to
+	/// 1 in the binary representation of value.
+	/// If value is zero, -1 will be returned.
+	///
+	/// @tparam genIUType Signed or unsigned integer scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/findLSB.xml">GLSL findLSB man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<typename genIUType>
+	GLM_FUNC_DECL int findLSB(genIUType x);
+
+	/// Returns the bit number of the least significant bit set to
+	/// 1 in the binary representation of value.
+	/// If value is zero, -1 will be returned.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/findLSB.xml">GLSL findLSB man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> findLSB(vec<L, T, Q> const& v);
+
+	/// Returns the bit number of the most significant bit in the binary representation of value.
+	/// For positive integers, the result will be the bit number of the most significant bit set to 1.
+	/// For negative integers, the result will be the bit number of the most significant
+	/// bit set to 0. For a value of zero or negative one, -1 will be returned.
+	///
+	/// @tparam genIUType Signed or unsigned integer scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/findMSB.xml">GLSL findMSB man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<typename genIUType>
+	GLM_FUNC_DECL int findMSB(genIUType x);
+
+	/// Returns the bit number of the most significant bit in the binary representation of value.
+	/// For positive integers, the result will be the bit number of the most significant bit set to 1.
+	/// For negative integers, the result will be the bit number of the most significant
+	/// bit set to 0. For a value of zero or negative one, -1 will be returned.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T Signed or unsigned integer scalar types.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/findMSB.xml">GLSL findMSB man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.8 Integer Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, int, Q> findMSB(vec<L, T, Q> const& v);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_integer.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat2x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat2x2.hpp
new file mode 100644
index 000000000000..96bec96b9a63
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat2x2.hpp
@@ -0,0 +1,9 @@
+/// @ref core
+/// @file glm/mat2x2.hpp
+
+#pragma once
+#include "./ext/matrix_double2x2.hpp"
+#include "./ext/matrix_double2x2_precision.hpp"
+#include "./ext/matrix_float2x2.hpp"
+#include "./ext/matrix_float2x2_precision.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat2x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat2x3.hpp
new file mode 100644
index 000000000000..d68dc25eda94
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat2x3.hpp
@@ -0,0 +1,9 @@
+/// @ref core
+/// @file glm/mat2x3.hpp
+
+#pragma once
+#include "./ext/matrix_double2x3.hpp"
+#include "./ext/matrix_double2x3_precision.hpp"
+#include "./ext/matrix_float2x3.hpp"
+#include "./ext/matrix_float2x3_precision.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat2x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat2x4.hpp
new file mode 100644
index 000000000000..b04b7387b1a3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat2x4.hpp
@@ -0,0 +1,9 @@
+/// @ref core
+/// @file glm/mat2x4.hpp
+
+#pragma once
+#include "./ext/matrix_double2x4.hpp"
+#include "./ext/matrix_double2x4_precision.hpp"
+#include "./ext/matrix_float2x4.hpp"
+#include "./ext/matrix_float2x4_precision.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat3x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat3x2.hpp
new file mode 100644
index 000000000000..c85315372dc0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat3x2.hpp
@@ -0,0 +1,9 @@
+/// @ref core
+/// @file glm/mat3x2.hpp
+
+#pragma once
+#include "./ext/matrix_double3x2.hpp"
+#include "./ext/matrix_double3x2_precision.hpp"
+#include "./ext/matrix_float3x2.hpp"
+#include "./ext/matrix_float3x2_precision.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat3x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat3x3.hpp
new file mode 100644
index 000000000000..fd4fa31cdee0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat3x3.hpp
@@ -0,0 +1,8 @@
+/// @ref core
+/// @file glm/mat3x3.hpp
+
+#pragma once
+#include "./ext/matrix_double3x3.hpp"
+#include "./ext/matrix_double3x3_precision.hpp"
+#include "./ext/matrix_float3x3.hpp"
+#include "./ext/matrix_float3x3_precision.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat3x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat3x4.hpp
new file mode 100644
index 000000000000..6342bf5b992d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat3x4.hpp
@@ -0,0 +1,8 @@
+/// @ref core
+/// @file glm/mat3x4.hpp
+
+#pragma once
+#include "./ext/matrix_double3x4.hpp"
+#include "./ext/matrix_double3x4_precision.hpp"
+#include "./ext/matrix_float3x4.hpp"
+#include "./ext/matrix_float3x4_precision.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat4x2.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat4x2.hpp
new file mode 100644
index 000000000000..e013e46b9c20
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat4x2.hpp
@@ -0,0 +1,9 @@
+/// @ref core
+/// @file glm/mat4x2.hpp
+
+#pragma once
+#include "./ext/matrix_double4x2.hpp"
+#include "./ext/matrix_double4x2_precision.hpp"
+#include "./ext/matrix_float4x2.hpp"
+#include "./ext/matrix_float4x2_precision.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat4x3.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat4x3.hpp
new file mode 100644
index 000000000000..205725abd25a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat4x3.hpp
@@ -0,0 +1,8 @@
+/// @ref core
+/// @file glm/mat4x3.hpp
+
+#pragma once
+#include "./ext/matrix_double4x3.hpp"
+#include "./ext/matrix_double4x3_precision.hpp"
+#include "./ext/matrix_float4x3.hpp"
+#include "./ext/matrix_float4x3_precision.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/mat4x4.hpp b/thirdparty/manifold/thirdparty/glm/glm/mat4x4.hpp
new file mode 100644
index 000000000000..3515f7f370bf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/mat4x4.hpp
@@ -0,0 +1,9 @@
+/// @ref core
+/// @file glm/mat4x4.hpp
+
+#pragma once
+#include "./ext/matrix_double4x4.hpp"
+#include "./ext/matrix_double4x4_precision.hpp"
+#include "./ext/matrix_float4x4.hpp"
+#include "./ext/matrix_float4x4_precision.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/matrix.hpp b/thirdparty/manifold/thirdparty/glm/glm/matrix.hpp
new file mode 100644
index 000000000000..4584c92c3c4e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/matrix.hpp
@@ -0,0 +1,161 @@
+/// @ref core
+/// @file glm/matrix.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+///
+/// @defgroup core_func_matrix Matrix functions
+/// @ingroup core
+///
+/// Provides GLSL matrix functions.
+///
+/// Include <glm/matrix.hpp> to use these core features.
+
+#pragma once
+
+// Dependencies
+#include "detail/qualifier.hpp"
+#include "detail/setup.hpp"
+#include "vec2.hpp"
+#include "vec3.hpp"
+#include "vec4.hpp"
+#include "mat2x2.hpp"
+#include "mat2x3.hpp"
+#include "mat2x4.hpp"
+#include "mat3x2.hpp"
+#include "mat3x3.hpp"
+#include "mat3x4.hpp"
+#include "mat4x2.hpp"
+#include "mat4x3.hpp"
+#include "mat4x4.hpp"
+
+namespace glm {
+namespace detail
+{
+	template<length_t C, length_t R, typename T, qualifier Q>
+	struct outerProduct_trait{};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<2, 2, T, Q>
+	{
+		typedef mat<2, 2, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<2, 3, T, Q>
+	{
+		typedef mat<3, 2, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<2, 4, T, Q>
+	{
+		typedef mat<4, 2, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<3, 2, T, Q>
+	{
+		typedef mat<2, 3, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<3, 3, T, Q>
+	{
+		typedef mat<3, 3, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<3, 4, T, Q>
+	{
+		typedef mat<4, 3, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<4, 2, T, Q>
+	{
+		typedef mat<2, 4, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<4, 3, T, Q>
+	{
+		typedef mat<3, 4, T, Q> type;
+	};
+
+	template<typename T, qualifier Q>
+	struct outerProduct_trait<4, 4, T, Q>
+	{
+		typedef mat<4, 4, T, Q> type;
+	};
+}//namespace detail
+
+	 /// @addtogroup core_func_matrix
+	 /// @{
+
+	 /// Multiply matrix x by matrix y component-wise, i.e.,
+	 /// result[i][j] is the scalar product of x[i][j] and y[i][j].
+	 ///
+	 /// @tparam C Integer between 1 and 4 included that qualify the number a column
+	 /// @tparam R Integer between 1 and 4 included that qualify the number a row
+	 /// @tparam T Floating-point scalar types
+	 /// @tparam Q Value from qualifier enum
+	 ///
+	 /// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/matrixCompMult.xml">GLSL matrixCompMult man page</a>
+	 /// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> matrixCompMult(mat<C, R, T, Q> const& x, mat<C, R, T, Q> const& y);
+
+	/// Treats the first parameter c as a column vector
+	/// and the second parameter r as a row vector
+	/// and does a linear algebraic matrix multiply c * r.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/outerProduct.xml">GLSL outerProduct man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL typename detail::outerProduct_trait<C, R, T, Q>::type outerProduct(vec<C, T, Q> const& c, vec<R, T, Q> const& r);
+
+	/// Returns the transposed matrix of x
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/transpose.xml">GLSL transpose man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL typename mat<C, R, T, Q>::transpose_type transpose(mat<C, R, T, Q> const& x);
+
+	/// Return the determinant of a squared matrix.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/determinant.xml">GLSL determinant man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL T determinant(mat<C, R, T, Q> const& m);
+
+	/// Return the inverse of a squared matrix.
+	///
+	/// @tparam C Integer between 1 and 4 included that qualify the number a column
+	/// @tparam R Integer between 1 and 4 included that qualify the number a row
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/inverse.xml">GLSL inverse man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.6 Matrix Functions</a>
+	template<length_t C, length_t R, typename T, qualifier Q>
+	GLM_FUNC_DECL mat<C, R, T, Q> inverse(mat<C, R, T, Q> const& m);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_matrix.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/packing.hpp b/thirdparty/manifold/thirdparty/glm/glm/packing.hpp
new file mode 100644
index 000000000000..ca83ac1dec96
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/packing.hpp
@@ -0,0 +1,173 @@
+/// @ref core
+/// @file glm/packing.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+/// @see gtc_packing
+///
+/// @defgroup core_func_packing Floating-Point Pack and Unpack Functions
+/// @ingroup core
+///
+/// Provides GLSL functions to pack and unpack half, single and double-precision floating point values into more compact integer types.
+///
+/// These functions do not operate component-wise, rather as described in each case.
+///
+/// Include <glm/packing.hpp> to use these core features.
+
+#pragma once
+
+#include "./ext/vector_uint2.hpp"
+#include "./ext/vector_float2.hpp"
+#include "./ext/vector_float4.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_func_packing
+	/// @{
+
+	/// First, converts each component of the normalized floating-point value v into 8- or 16-bit integer values.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm2x16: round(clamp(c, 0, +1) * 65535.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packUnorm2x16.xml">GLSL packUnorm2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint packUnorm2x16(vec2 const& v);
+
+	/// First, converts each component of the normalized floating-point value v into 8- or 16-bit integer values.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packSnorm2x16: round(clamp(v, -1, +1) * 32767.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packSnorm2x16.xml">GLSL packSnorm2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint packSnorm2x16(vec2 const& v);
+
+	/// First, converts each component of the normalized floating-point value v into 8- or 16-bit integer values.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packUnorm4x8:	round(clamp(c, 0, +1) * 255.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packUnorm4x8.xml">GLSL packUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint packUnorm4x8(vec4 const& v);
+
+	/// First, converts each component of the normalized floating-point value v into 8- or 16-bit integer values.
+	/// Then, the results are packed into the returned 32-bit unsigned integer.
+	///
+	/// The conversion for component c of v to fixed point is done as follows:
+	/// packSnorm4x8:	round(clamp(c, -1, +1) * 127.0)
+	///
+	/// The first component of the vector will be written to the least significant bits of the output;
+	/// the last component will be written to the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packSnorm4x8.xml">GLSL packSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint packSnorm4x8(vec4 const& v);
+
+	/// First, unpacks a single 32-bit unsigned integer p into a pair of 16-bit unsigned integers, four 8-bit unsigned integers, or four 8-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned two- or four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackUnorm2x16: f / 65535.0
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackUnorm2x16.xml">GLSL unpackUnorm2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec2 unpackUnorm2x16(uint p);
+
+	/// First, unpacks a single 32-bit unsigned integer p into a pair of 16-bit unsigned integers, four 8-bit unsigned integers, or four 8-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned two- or four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm2x16: clamp(f / 32767.0, -1, +1)
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackSnorm2x16.xml">GLSL unpackSnorm2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec2 unpackSnorm2x16(uint p);
+
+	/// First, unpacks a single 32-bit unsigned integer p into a pair of 16-bit unsigned integers, four 8-bit unsigned integers, or four 8-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned two- or four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackUnorm4x8: f / 255.0
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackUnorm4x8.xml">GLSL unpackUnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec4 unpackUnorm4x8(uint p);
+
+	/// First, unpacks a single 32-bit unsigned integer p into a pair of 16-bit unsigned integers, four 8-bit unsigned integers, or four 8-bit signed integers.
+	/// Then, each component is converted to a normalized floating-point value to generate the returned two- or four-component vector.
+	///
+	/// The conversion for unpacked fixed-point value f to floating point is done as follows:
+	/// unpackSnorm4x8: clamp(f / 127.0, -1, +1)
+	///
+	/// The first component of the returned vector will be extracted from the least significant bits of the input;
+	/// the last component will be extracted from the most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackSnorm4x8.xml">GLSL unpackSnorm4x8 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec4 unpackSnorm4x8(uint p);
+
+	/// Returns a double-qualifier value obtained by packing the components of v into a 64-bit value.
+	/// If an IEEE 754 Inf or NaN is created, it will not signal, and the resulting floating point value is unspecified.
+	/// Otherwise, the bit- level representation of v is preserved.
+	/// The first vector component specifies the 32 least significant bits;
+	/// the second component specifies the 32 most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packDouble2x32.xml">GLSL packDouble2x32 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL double packDouble2x32(uvec2 const& v);
+
+	/// Returns a two-component unsigned integer vector representation of v.
+	/// The bit-level representation of v is preserved.
+	/// The first component of the vector contains the 32 least significant bits of the double;
+	/// the second component consists the 32 most significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackDouble2x32.xml">GLSL unpackDouble2x32 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uvec2 unpackDouble2x32(double v);
+
+	/// Returns an unsigned integer obtained by converting the components of a two-component floating-point vector
+	/// to the 16-bit floating-point representation found in the OpenGL Specification,
+	/// and then packing these two 16- bit integers into a 32-bit unsigned integer.
+	/// The first vector component specifies the 16 least-significant bits of the result;
+	/// the second component specifies the 16 most-significant bits.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/packHalf2x16.xml">GLSL packHalf2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL uint packHalf2x16(vec2 const& v);
+
+	/// Returns a two-component floating-point vector with components obtained by unpacking a 32-bit unsigned integer into a pair of 16-bit values,
+	/// interpreting those values as 16-bit floating-point numbers according to the OpenGL Specification,
+	/// and converting them to 32-bit floating-point values.
+	/// The first component of the vector is obtained from the 16 least-significant bits of v;
+	/// the second component is obtained from the 16 most-significant bits of v.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/unpackHalf2x16.xml">GLSL unpackHalf2x16 man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.4 Floating-Point Pack and Unpack Functions</a>
+	GLM_FUNC_DECL vec2 unpackHalf2x16(uint v);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_packing.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/common.h b/thirdparty/manifold/thirdparty/glm/glm/simd/common.h
new file mode 100644
index 000000000000..9b017cb4256e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/common.h
@@ -0,0 +1,240 @@
+/// @ref simd
+/// @file glm/simd/common.h
+
+#pragma once
+
+#include "platform.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_add(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_add_ps(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec1_add(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_add_ss(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_sub(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_sub_ps(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec1_sub(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_sub_ss(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_mul(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_mul_ps(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec1_mul(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_mul_ss(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_div(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_div_ps(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec1_div(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return _mm_div_ss(a, b);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_div_lowp(glm_f32vec4 a, glm_f32vec4 b)
+{
+	return glm_vec4_mul(a, _mm_rcp_ps(b));
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_swizzle_xyzw(glm_f32vec4 a)
+{
+#	if GLM_ARCH & GLM_ARCH_AVX2_BIT
+		return _mm_permute_ps(a, _MM_SHUFFLE(3, 2, 1, 0));
+#	else
+		return _mm_shuffle_ps(a, a, _MM_SHUFFLE(3, 2, 1, 0));
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec1_fma(glm_f32vec4 a, glm_f32vec4 b, glm_f32vec4 c)
+{
+#	if (GLM_ARCH & GLM_ARCH_AVX2_BIT) && !(GLM_COMPILER & GLM_COMPILER_CLANG)
+		return _mm_fmadd_ss(a, b, c);
+#	else
+		return _mm_add_ss(_mm_mul_ss(a, b), c);
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_fma(glm_f32vec4 a, glm_f32vec4 b, glm_f32vec4 c)
+{
+#	if (GLM_ARCH & GLM_ARCH_AVX2_BIT) && !(GLM_COMPILER & GLM_COMPILER_CLANG)
+		return _mm_fmadd_ps(a, b, c);
+#	else
+		return glm_vec4_add(glm_vec4_mul(a, b), c);
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_abs(glm_f32vec4 x)
+{
+	return _mm_and_ps(x, _mm_castsi128_ps(_mm_set1_epi32(0x7FFFFFFF)));
+}
+
+GLM_FUNC_QUALIFIER glm_ivec4 glm_ivec4_abs(glm_ivec4 x)
+{
+#	if GLM_ARCH & GLM_ARCH_SSSE3_BIT
+		return _mm_sign_epi32(x, x);
+#	else
+		glm_ivec4 const sgn0 = _mm_srai_epi32(x, 31);
+		glm_ivec4 const inv0 = _mm_xor_si128(x, sgn0);
+		glm_ivec4 const sub0 = _mm_sub_epi32(inv0, sgn0);
+		return sub0;
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_sign(glm_vec4 x)
+{
+	glm_vec4 const zro0 = _mm_setzero_ps();
+	glm_vec4 const cmp0 = _mm_cmplt_ps(x, zro0);
+	glm_vec4 const cmp1 = _mm_cmpgt_ps(x, zro0);
+	glm_vec4 const and0 = _mm_and_ps(cmp0, _mm_set1_ps(-1.0f));
+	glm_vec4 const and1 = _mm_and_ps(cmp1, _mm_set1_ps(1.0f));
+	glm_vec4 const or0 = _mm_or_ps(and0, and1);
+	return or0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_round(glm_vec4 x)
+{
+#	if GLM_ARCH & GLM_ARCH_SSE41_BIT
+		return _mm_round_ps(x, _MM_FROUND_TO_NEAREST_INT);
+#	else
+		glm_vec4 const sgn0 = _mm_castsi128_ps(_mm_set1_epi32(int(0x80000000)));
+		glm_vec4 const and0 = _mm_and_ps(sgn0, x);
+		glm_vec4 const or0 = _mm_or_ps(and0, _mm_set_ps1(8388608.0f));
+		glm_vec4 const add0 = glm_vec4_add(x, or0);
+		glm_vec4 const sub0 = glm_vec4_sub(add0, or0);
+		return sub0;
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_floor(glm_vec4 x)
+{
+#	if GLM_ARCH & GLM_ARCH_SSE41_BIT
+		return _mm_floor_ps(x);
+#	else
+		glm_vec4 const rnd0 = glm_vec4_round(x);
+		glm_vec4 const cmp0 = _mm_cmplt_ps(x, rnd0);
+		glm_vec4 const and0 = _mm_and_ps(cmp0, _mm_set1_ps(1.0f));
+		glm_vec4 const sub0 = glm_vec4_sub(rnd0, and0);
+		return sub0;
+#	endif
+}
+
+/* trunc TODO
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_trunc(glm_vec4 x)
+{
+	return glm_vec4();
+}
+*/
+
+//roundEven
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_roundEven(glm_vec4 x)
+{
+	glm_vec4 const sgn0 = _mm_castsi128_ps(_mm_set1_epi32(int(0x80000000)));
+	glm_vec4 const and0 = _mm_and_ps(sgn0, x);
+	glm_vec4 const or0 = _mm_or_ps(and0, _mm_set_ps1(8388608.0f));
+	glm_vec4 const add0 = glm_vec4_add(x, or0);
+	glm_vec4 const sub0 = glm_vec4_sub(add0, or0);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_ceil(glm_vec4 x)
+{
+#	if GLM_ARCH & GLM_ARCH_SSE41_BIT
+		return _mm_ceil_ps(x);
+#	else
+		glm_vec4 const rnd0 = glm_vec4_round(x);
+		glm_vec4 const cmp0 = _mm_cmpgt_ps(x, rnd0);
+		glm_vec4 const and0 = _mm_and_ps(cmp0, _mm_set1_ps(1.0f));
+		glm_vec4 const add0 = glm_vec4_add(rnd0, and0);
+		return add0;
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_fract(glm_vec4 x)
+{
+	glm_vec4 const flr0 = glm_vec4_floor(x);
+	glm_vec4 const sub0 = glm_vec4_sub(x, flr0);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_mod(glm_vec4 x, glm_vec4 y)
+{
+	glm_vec4 const div0 = glm_vec4_div(x, y);
+	glm_vec4 const flr0 = glm_vec4_floor(div0);
+	glm_vec4 const mul0 = glm_vec4_mul(y, flr0);
+	glm_vec4 const sub0 = glm_vec4_sub(x, mul0);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_clamp(glm_vec4 v, glm_vec4 minVal, glm_vec4 maxVal)
+{
+	glm_vec4 const min0 = _mm_min_ps(v, maxVal);
+	glm_vec4 const max0 = _mm_max_ps(min0, minVal);
+	return max0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_mix(glm_vec4 v1, glm_vec4 v2, glm_vec4 a)
+{
+	glm_vec4 const sub0 = glm_vec4_sub(_mm_set1_ps(1.0f), a);
+	glm_vec4 const mul0 = glm_vec4_mul(v1, sub0);
+	glm_vec4 const mad0 = glm_vec4_fma(v2, a, mul0);
+	return mad0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_step(glm_vec4 edge, glm_vec4 x)
+{
+	glm_vec4 const cmp = _mm_cmple_ps(x, edge);
+	return _mm_movemask_ps(cmp) == 0 ? _mm_set1_ps(1.0f) : _mm_setzero_ps();
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_smoothstep(glm_vec4 edge0, glm_vec4 edge1, glm_vec4 x)
+{
+	glm_vec4 const sub0 = glm_vec4_sub(x, edge0);
+	glm_vec4 const sub1 = glm_vec4_sub(edge1, edge0);
+	glm_vec4 const div0 = glm_vec4_sub(sub0, sub1);
+	glm_vec4 const clp0 = glm_vec4_clamp(div0, _mm_setzero_ps(), _mm_set1_ps(1.0f));
+	glm_vec4 const mul0 = glm_vec4_mul(_mm_set1_ps(2.0f), clp0);
+	glm_vec4 const sub2 = glm_vec4_sub(_mm_set1_ps(3.0f), mul0);
+	glm_vec4 const mul1 = glm_vec4_mul(clp0, clp0);
+	glm_vec4 const mul2 = glm_vec4_mul(mul1, sub2);
+	return mul2;
+}
+
+// Agner Fog method
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_nan(glm_vec4 x)
+{
+	glm_ivec4 const t1 = _mm_castps_si128(x);						// reinterpret as 32-bit integer
+	glm_ivec4 const t2 = _mm_sll_epi32(t1, _mm_cvtsi32_si128(1));	// shift out sign bit
+	glm_ivec4 const t3 = _mm_set1_epi32(int(0xFF000000));				// exponent mask
+	glm_ivec4 const t4 = _mm_and_si128(t2, t3);						// exponent
+	glm_ivec4 const t5 = _mm_andnot_si128(t3, t2);					// fraction
+	glm_ivec4 const Equal = _mm_cmpeq_epi32(t3, t4);
+	glm_ivec4 const Nequal = _mm_cmpeq_epi32(t5, _mm_setzero_si128());
+	glm_ivec4 const And = _mm_and_si128(Equal, Nequal);
+	return _mm_castsi128_ps(And);									// exponent = all 1s and fraction != 0
+}
+
+// Agner Fog method
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_inf(glm_vec4 x)
+{
+	glm_ivec4 const t1 = _mm_castps_si128(x);										// reinterpret as 32-bit integer
+	glm_ivec4 const t2 = _mm_sll_epi32(t1, _mm_cvtsi32_si128(1));					// shift out sign bit
+	return _mm_castsi128_ps(_mm_cmpeq_epi32(t2, _mm_set1_epi32(int(0xFF000000))));		// exponent is all 1s, fraction is 0
+}
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/exponential.h b/thirdparty/manifold/thirdparty/glm/glm/simd/exponential.h
new file mode 100644
index 000000000000..bc351d0119b9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/exponential.h
@@ -0,0 +1,20 @@
+/// @ref simd
+/// @file glm/simd/experimental.h
+
+#pragma once
+
+#include "platform.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec1_sqrt_lowp(glm_f32vec4 x)
+{
+	return _mm_mul_ss(_mm_rsqrt_ss(x), x);
+}
+
+GLM_FUNC_QUALIFIER glm_f32vec4 glm_vec4_sqrt_lowp(glm_f32vec4 x)
+{
+	return _mm_mul_ps(_mm_rsqrt_ps(x), x);
+}
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/geometric.h b/thirdparty/manifold/thirdparty/glm/glm/simd/geometric.h
new file mode 100644
index 000000000000..07d7cbcc425f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/geometric.h
@@ -0,0 +1,124 @@
+/// @ref simd
+/// @file glm/simd/geometric.h
+
+#pragma once
+
+#include "common.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+GLM_FUNC_DECL glm_vec4 glm_vec4_dot(glm_vec4 v1, glm_vec4 v2);
+GLM_FUNC_DECL glm_vec4 glm_vec1_dot(glm_vec4 v1, glm_vec4 v2);
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_length(glm_vec4 x)
+{
+	glm_vec4 const dot0 = glm_vec4_dot(x, x);
+	glm_vec4 const sqt0 = _mm_sqrt_ps(dot0);
+	return sqt0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_distance(glm_vec4 p0, glm_vec4 p1)
+{
+	glm_vec4 const sub0 = _mm_sub_ps(p0, p1);
+	glm_vec4 const len0 = glm_vec4_length(sub0);
+	return len0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_dot(glm_vec4 v1, glm_vec4 v2)
+{
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+		return _mm_dp_ps(v1, v2, 0xff);
+#	elif GLM_ARCH & GLM_ARCH_SSE3_BIT
+		glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
+		glm_vec4 const hadd0 = _mm_hadd_ps(mul0, mul0);
+		glm_vec4 const hadd1 = _mm_hadd_ps(hadd0, hadd0);
+		return hadd1;
+#	else
+		glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
+		glm_vec4 const swp0 = _mm_shuffle_ps(mul0, mul0, _MM_SHUFFLE(2, 3, 0, 1));
+		glm_vec4 const add0 = _mm_add_ps(mul0, swp0);
+		glm_vec4 const swp1 = _mm_shuffle_ps(add0, add0, _MM_SHUFFLE(0, 1, 2, 3));
+		glm_vec4 const add1 = _mm_add_ps(add0, swp1);
+		return add1;
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec1_dot(glm_vec4 v1, glm_vec4 v2)
+{
+#	if GLM_ARCH & GLM_ARCH_AVX_BIT
+		return _mm_dp_ps(v1, v2, 0xff);
+#	elif GLM_ARCH & GLM_ARCH_SSE3_BIT
+		glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
+		glm_vec4 const had0 = _mm_hadd_ps(mul0, mul0);
+		glm_vec4 const had1 = _mm_hadd_ps(had0, had0);
+		return had1;
+#	else
+		glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
+		glm_vec4 const mov0 = _mm_movehl_ps(mul0, mul0);
+		glm_vec4 const add0 = _mm_add_ps(mov0, mul0);
+		glm_vec4 const swp1 = _mm_shuffle_ps(add0, add0, 1);
+		glm_vec4 const add1 = _mm_add_ss(add0, swp1);
+		return add1;
+#	endif
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_cross(glm_vec4 v1, glm_vec4 v2)
+{
+	glm_vec4 const swp0 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 0, 2, 1));
+	glm_vec4 const swp1 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 1, 0, 2));
+	glm_vec4 const swp2 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 0, 2, 1));
+	glm_vec4 const swp3 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 1, 0, 2));
+	glm_vec4 const mul0 = _mm_mul_ps(swp0, swp3);
+	glm_vec4 const mul1 = _mm_mul_ps(swp1, swp2);
+	glm_vec4 const sub0 = _mm_sub_ps(mul0, mul1);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_normalize(glm_vec4 v)
+{
+	glm_vec4 const dot0 = glm_vec4_dot(v, v);
+	glm_vec4 const isr0 = _mm_rsqrt_ps(dot0);
+	glm_vec4 const mul0 = _mm_mul_ps(v, isr0);
+	return mul0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_faceforward(glm_vec4 N, glm_vec4 I, glm_vec4 Nref)
+{
+	glm_vec4 const dot0 = glm_vec4_dot(Nref, I);
+	glm_vec4 const sgn0 = glm_vec4_sign(dot0);
+	glm_vec4 const mul0 = _mm_mul_ps(sgn0, _mm_set1_ps(-1.0f));
+	glm_vec4 const mul1 = _mm_mul_ps(N, mul0);
+	return mul1;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_reflect(glm_vec4 I, glm_vec4 N)
+{
+	glm_vec4 const dot0 = glm_vec4_dot(N, I);
+	glm_vec4 const mul0 = _mm_mul_ps(N, dot0);
+	glm_vec4 const mul1 = _mm_mul_ps(mul0, _mm_set1_ps(2.0f));
+	glm_vec4 const sub0 = _mm_sub_ps(I, mul1);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_vec4_refract(glm_vec4 I, glm_vec4 N, glm_vec4 eta)
+{
+	glm_vec4 const dot0 = glm_vec4_dot(N, I);
+	glm_vec4 const mul0 = _mm_mul_ps(eta, eta);
+	glm_vec4 const mul1 = _mm_mul_ps(dot0, dot0);
+	glm_vec4 const sub0 = _mm_sub_ps(_mm_set1_ps(1.0f), mul0);
+	glm_vec4 const sub1 = _mm_sub_ps(_mm_set1_ps(1.0f), mul1);
+	glm_vec4 const mul2 = _mm_mul_ps(sub0, sub1);
+
+	if(_mm_movemask_ps(_mm_cmplt_ss(mul2, _mm_set1_ps(0.0f))) == 0)
+		return _mm_set1_ps(0.0f);
+
+	glm_vec4 const sqt0 = _mm_sqrt_ps(mul2);
+	glm_vec4 const mad0 = glm_vec4_fma(eta, dot0, sqt0);
+	glm_vec4 const mul4 = _mm_mul_ps(mad0, N);
+	glm_vec4 const mul5 = _mm_mul_ps(eta, I);
+	glm_vec4 const sub2 = _mm_sub_ps(mul5, mul4);
+
+	return sub2;
+}
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/integer.h b/thirdparty/manifold/thirdparty/glm/glm/simd/integer.h
new file mode 100644
index 000000000000..93814183fe02
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/integer.h
@@ -0,0 +1,115 @@
+/// @ref simd
+/// @file glm/simd/integer.h
+
+#pragma once
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+GLM_FUNC_QUALIFIER glm_uvec4 glm_i128_interleave(glm_uvec4 x)
+{
+	glm_uvec4 const Mask4 = _mm_set1_epi32(0x0000FFFF);
+	glm_uvec4 const Mask3 = _mm_set1_epi32(0x00FF00FF);
+	glm_uvec4 const Mask2 = _mm_set1_epi32(0x0F0F0F0F);
+	glm_uvec4 const Mask1 = _mm_set1_epi32(0x33333333);
+	glm_uvec4 const Mask0 = _mm_set1_epi32(0x55555555);
+
+	glm_uvec4 Reg1;
+	glm_uvec4 Reg2;
+
+	// REG1 = x;
+	// REG2 = y;
+	//Reg1 = _mm_unpacklo_epi64(x, y);
+	Reg1 = x;
+
+	//REG1 = ((REG1 << 16) | REG1) & glm::uint64(0x0000FFFF0000FFFF);
+	//REG2 = ((REG2 << 16) | REG2) & glm::uint64(0x0000FFFF0000FFFF);
+	Reg2 = _mm_slli_si128(Reg1, 2);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask4);
+
+	//REG1 = ((REG1 <<  8) | REG1) & glm::uint64(0x00FF00FF00FF00FF);
+	//REG2 = ((REG2 <<  8) | REG2) & glm::uint64(0x00FF00FF00FF00FF);
+	Reg2 = _mm_slli_si128(Reg1, 1);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask3);
+
+	//REG1 = ((REG1 <<  4) | REG1) & glm::uint64(0x0F0F0F0F0F0F0F0F);
+	//REG2 = ((REG2 <<  4) | REG2) & glm::uint64(0x0F0F0F0F0F0F0F0F);
+	Reg2 = _mm_slli_epi32(Reg1, 4);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask2);
+
+	//REG1 = ((REG1 <<  2) | REG1) & glm::uint64(0x3333333333333333);
+	//REG2 = ((REG2 <<  2) | REG2) & glm::uint64(0x3333333333333333);
+	Reg2 = _mm_slli_epi32(Reg1, 2);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask1);
+
+	//REG1 = ((REG1 <<  1) | REG1) & glm::uint64(0x5555555555555555);
+	//REG2 = ((REG2 <<  1) | REG2) & glm::uint64(0x5555555555555555);
+	Reg2 = _mm_slli_epi32(Reg1, 1);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask0);
+
+	//return REG1 | (REG2 << 1);
+	Reg2 = _mm_slli_epi32(Reg1, 1);
+	Reg2 = _mm_srli_si128(Reg2, 8);
+	Reg1 = _mm_or_si128(Reg1, Reg2);
+
+	return Reg1;
+}
+
+GLM_FUNC_QUALIFIER glm_uvec4 glm_i128_interleave2(glm_uvec4 x, glm_uvec4 y)
+{
+	glm_uvec4 const Mask4 = _mm_set1_epi32(0x0000FFFF);
+	glm_uvec4 const Mask3 = _mm_set1_epi32(0x00FF00FF);
+	glm_uvec4 const Mask2 = _mm_set1_epi32(0x0F0F0F0F);
+	glm_uvec4 const Mask1 = _mm_set1_epi32(0x33333333);
+	glm_uvec4 const Mask0 = _mm_set1_epi32(0x55555555);
+
+	glm_uvec4 Reg1;
+	glm_uvec4 Reg2;
+
+	// REG1 = x;
+	// REG2 = y;
+	Reg1 = _mm_unpacklo_epi64(x, y);
+
+	//REG1 = ((REG1 << 16) | REG1) & glm::uint64(0x0000FFFF0000FFFF);
+	//REG2 = ((REG2 << 16) | REG2) & glm::uint64(0x0000FFFF0000FFFF);
+	Reg2 = _mm_slli_si128(Reg1, 2);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask4);
+
+	//REG1 = ((REG1 <<  8) | REG1) & glm::uint64(0x00FF00FF00FF00FF);
+	//REG2 = ((REG2 <<  8) | REG2) & glm::uint64(0x00FF00FF00FF00FF);
+	Reg2 = _mm_slli_si128(Reg1, 1);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask3);
+
+	//REG1 = ((REG1 <<  4) | REG1) & glm::uint64(0x0F0F0F0F0F0F0F0F);
+	//REG2 = ((REG2 <<  4) | REG2) & glm::uint64(0x0F0F0F0F0F0F0F0F);
+	Reg2 = _mm_slli_epi32(Reg1, 4);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask2);
+
+	//REG1 = ((REG1 <<  2) | REG1) & glm::uint64(0x3333333333333333);
+	//REG2 = ((REG2 <<  2) | REG2) & glm::uint64(0x3333333333333333);
+	Reg2 = _mm_slli_epi32(Reg1, 2);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask1);
+
+	//REG1 = ((REG1 <<  1) | REG1) & glm::uint64(0x5555555555555555);
+	//REG2 = ((REG2 <<  1) | REG2) & glm::uint64(0x5555555555555555);
+	Reg2 = _mm_slli_epi32(Reg1, 1);
+	Reg1 = _mm_or_si128(Reg2, Reg1);
+	Reg1 = _mm_and_si128(Reg1, Mask0);
+
+	//return REG1 | (REG2 << 1);
+	Reg2 = _mm_slli_epi32(Reg1, 1);
+	Reg2 = _mm_srli_si128(Reg2, 8);
+	Reg1 = _mm_or_si128(Reg1, Reg2);
+
+	return Reg1;
+}
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/matrix.h b/thirdparty/manifold/thirdparty/glm/glm/simd/matrix.h
new file mode 100644
index 000000000000..b6c42ea4c17c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/matrix.h
@@ -0,0 +1,1028 @@
+/// @ref simd
+/// @file glm/simd/matrix.h
+
+#pragma once
+
+#include "geometric.h"
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+GLM_FUNC_QUALIFIER void glm_mat4_matrixCompMult(glm_vec4 const in1[4], glm_vec4 const in2[4], glm_vec4 out[4])
+{
+	out[0] = _mm_mul_ps(in1[0], in2[0]);
+	out[1] = _mm_mul_ps(in1[1], in2[1]);
+	out[2] = _mm_mul_ps(in1[2], in2[2]);
+	out[3] = _mm_mul_ps(in1[3], in2[3]);
+}
+
+GLM_FUNC_QUALIFIER void glm_mat4_add(glm_vec4 const in1[4], glm_vec4 const in2[4], glm_vec4 out[4])
+{
+	out[0] = _mm_add_ps(in1[0], in2[0]);
+	out[1] = _mm_add_ps(in1[1], in2[1]);
+	out[2] = _mm_add_ps(in1[2], in2[2]);
+	out[3] = _mm_add_ps(in1[3], in2[3]);
+}
+
+GLM_FUNC_QUALIFIER void glm_mat4_sub(glm_vec4 const in1[4], glm_vec4 const in2[4], glm_vec4 out[4])
+{
+	out[0] = _mm_sub_ps(in1[0], in2[0]);
+	out[1] = _mm_sub_ps(in1[1], in2[1]);
+	out[2] = _mm_sub_ps(in1[2], in2[2]);
+	out[3] = _mm_sub_ps(in1[3], in2[3]);
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_mat4_mul_vec4(glm_vec4 const m[4], glm_vec4 v)
+{
+	__m128 v0 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 v1 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 v2 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 v3 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3));
+
+	__m128 m0 = _mm_mul_ps(m[0], v0);
+	__m128 m1 = _mm_mul_ps(m[1], v1);
+	__m128 m2 = _mm_mul_ps(m[2], v2);
+	__m128 m3 = _mm_mul_ps(m[3], v3);
+
+	__m128 a0 = _mm_add_ps(m0, m1);
+	__m128 a1 = _mm_add_ps(m2, m3);
+	__m128 a2 = _mm_add_ps(a0, a1);
+
+	return a2;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_vec4_mul_mat4(glm_vec4 v, glm_vec4 const m[4])
+{
+	__m128 i0 = m[0];
+	__m128 i1 = m[1];
+	__m128 i2 = m[2];
+	__m128 i3 = m[3];
+
+	__m128 m0 = _mm_mul_ps(v, i0);
+	__m128 m1 = _mm_mul_ps(v, i1);
+	__m128 m2 = _mm_mul_ps(v, i2);
+	__m128 m3 = _mm_mul_ps(v, i3);
+
+	__m128 u0 = _mm_unpacklo_ps(m0, m1);
+	__m128 u1 = _mm_unpackhi_ps(m0, m1);
+	__m128 a0 = _mm_add_ps(u0, u1);
+
+	__m128 u2 = _mm_unpacklo_ps(m2, m3);
+	__m128 u3 = _mm_unpackhi_ps(m2, m3);
+	__m128 a1 = _mm_add_ps(u2, u3);
+
+	__m128 f0 = _mm_movelh_ps(a0, a1);
+	__m128 f1 = _mm_movehl_ps(a1, a0);
+	__m128 f2 = _mm_add_ps(f0, f1);
+
+	return f2;
+}
+
+GLM_FUNC_QUALIFIER void glm_mat4_mul(glm_vec4 const in1[4], glm_vec4 const in2[4], glm_vec4 out[4])
+{
+	{
+		__m128 e0 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[0] = a2;
+	}
+
+	{
+		__m128 e0 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[1] = a2;
+	}
+
+	{
+		__m128 e0 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[2] = a2;
+	}
+
+	{
+		//(__m128&)_mm_shuffle_epi32(__m128i&)in2[0], _MM_SHUFFLE(3, 3, 3, 3))
+		__m128 e0 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[3] = a2;
+	}
+}
+
+GLM_FUNC_QUALIFIER void glm_mat4_transpose(glm_vec4 const in[4], glm_vec4 out[4])
+{
+	__m128 tmp0 = _mm_shuffle_ps(in[0], in[1], 0x44);
+	__m128 tmp2 = _mm_shuffle_ps(in[0], in[1], 0xEE);
+	__m128 tmp1 = _mm_shuffle_ps(in[2], in[3], 0x44);
+	__m128 tmp3 = _mm_shuffle_ps(in[2], in[3], 0xEE);
+
+	out[0] = _mm_shuffle_ps(tmp0, tmp1, 0x88);
+	out[1] = _mm_shuffle_ps(tmp0, tmp1, 0xDD);
+	out[2] = _mm_shuffle_ps(tmp2, tmp3, 0x88);
+	out[3] = _mm_shuffle_ps(tmp2, tmp3, 0xDD);
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_mat4_determinant_highp(glm_vec4 const in[4])
+{
+	__m128 Fac0;
+	{
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac0 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac1;
+	{
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac1 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+
+	__m128 Fac2;
+	{
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac2 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac3;
+	{
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac3 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac4;
+	{
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac4 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac5;
+	{
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac5 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
+	__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
+
+	// m[1][0]
+	// m[0][0]
+	// m[0][0]
+	// m[0][0]
+	__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][1]
+	// m[0][1]
+	// m[0][1]
+	// m[0][1]
+	__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][2]
+	// m[0][2]
+	// m[0][2]
+	// m[0][2]
+	__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][3]
+	// m[0][3]
+	// m[0][3]
+	// m[0][3]
+	__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
+	__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// col0
+	// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
+	// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
+	// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
+	// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
+	__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
+	__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
+	__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
+	__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
+	__m128 Add00 = _mm_add_ps(Sub00, Mul02);
+	__m128 Inv0 = _mm_mul_ps(SignB, Add00);
+
+	// col1
+	// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
+	// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
+	// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
+	// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
+	__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
+	__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
+	__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
+	__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
+	__m128 Add01 = _mm_add_ps(Sub01, Mul05);
+	__m128 Inv1 = _mm_mul_ps(SignA, Add01);
+
+	// col2
+	// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
+	// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
+	// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
+	// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
+	__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
+	__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
+	__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
+	__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
+	__m128 Add02 = _mm_add_ps(Sub02, Mul08);
+	__m128 Inv2 = _mm_mul_ps(SignB, Add02);
+
+	// col3
+	// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
+	// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
+	// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
+	// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
+	__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
+	__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
+	__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
+	__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
+	__m128 Add03 = _mm_add_ps(Sub03, Mul11);
+	__m128 Inv3 = _mm_mul_ps(SignA, Add03);
+
+	__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
+
+	//	valType Determinant = m[0][0] * Inverse[0][0]
+	//						+ m[0][1] * Inverse[1][0]
+	//						+ m[0][2] * Inverse[2][0]
+	//						+ m[0][3] * Inverse[3][0];
+	__m128 Det0 = glm_vec4_dot(in[0], Row2);
+	return Det0;
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_mat4_determinant_lowp(glm_vec4 const m[4])
+{
+	// _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(
+
+	//T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+	//T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+	//T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+	//T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+	//T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+	//T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+
+	// First 2 columns
+ 	__m128 Swp2A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 1, 1, 2)));
+ 	__m128 Swp3A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(3, 2, 3, 3)));
+	__m128 MulA = _mm_mul_ps(Swp2A, Swp3A);
+
+	// Second 2 columns
+	__m128 Swp2B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(3, 2, 3, 3)));
+	__m128 Swp3B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(0, 1, 1, 2)));
+	__m128 MulB = _mm_mul_ps(Swp2B, Swp3B);
+
+	// Columns subtraction
+	__m128 SubE = _mm_sub_ps(MulA, MulB);
+
+	// Last 2 rows
+	__m128 Swp2C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 0, 1, 2)));
+	__m128 Swp3C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(1, 2, 0, 0)));
+	__m128 MulC = _mm_mul_ps(Swp2C, Swp3C);
+	__m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC);
+
+	//vec<4, T, Q> DetCof(
+	//	+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
+	//	- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
+	//	+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
+	//	- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
+
+	__m128 SubFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubE), _MM_SHUFFLE(2, 1, 0, 0)));
+	__m128 SwpFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(0, 0, 0, 1)));
+	__m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA);
+
+	__m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1));
+	__m128 SubFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpB), _MM_SHUFFLE(3, 1, 1, 0)));//SubF[0], SubE[3], SubE[3], SubE[1];
+	__m128 SwpFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(1, 1, 2, 2)));
+	__m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB);
+
+	__m128 SubRes = _mm_sub_ps(MulFacA, MulFacB);
+
+	__m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2));
+	__m128 SubFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpC), _MM_SHUFFLE(3, 3, 2, 0)));
+	__m128 SwpFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(2, 3, 3, 3)));
+	__m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC);
+
+	__m128 AddRes = _mm_add_ps(SubRes, MulFacC);
+	__m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f));
+
+	//return m[0][0] * DetCof[0]
+	//	 + m[0][1] * DetCof[1]
+	//	 + m[0][2] * DetCof[2]
+	//	 + m[0][3] * DetCof[3];
+
+	return glm_vec4_dot(m[0], DetCof);
+}
+
+GLM_FUNC_QUALIFIER glm_vec4 glm_mat4_determinant(glm_vec4 const m[4])
+{
+	// _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(add)
+
+	//T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+	//T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+	//T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+	//T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+	//T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+	//T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+
+	// First 2 columns
+ 	__m128 Swp2A = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 1, 1, 2));
+ 	__m128 Swp3A = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(3, 2, 3, 3));
+	__m128 MulA = _mm_mul_ps(Swp2A, Swp3A);
+
+	// Second 2 columns
+	__m128 Swp2B = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(3, 2, 3, 3));
+	__m128 Swp3B = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(0, 1, 1, 2));
+	__m128 MulB = _mm_mul_ps(Swp2B, Swp3B);
+
+	// Columns subtraction
+	__m128 SubE = _mm_sub_ps(MulA, MulB);
+
+	// Last 2 rows
+	__m128 Swp2C = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 0, 1, 2));
+	__m128 Swp3C = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(1, 2, 0, 0));
+	__m128 MulC = _mm_mul_ps(Swp2C, Swp3C);
+	__m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC);
+
+	//vec<4, T, Q> DetCof(
+	//	+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
+	//	- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
+	//	+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
+	//	- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
+
+	__m128 SubFacA = _mm_shuffle_ps(SubE, SubE, _MM_SHUFFLE(2, 1, 0, 0));
+	__m128 SwpFacA = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(0, 0, 0, 1));
+	__m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA);
+
+	__m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1));
+	__m128 SubFacB = _mm_shuffle_ps(SubTmpB, SubTmpB, _MM_SHUFFLE(3, 1, 1, 0));//SubF[0], SubE[3], SubE[3], SubE[1];
+	__m128 SwpFacB = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(1, 1, 2, 2));
+	__m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB);
+
+	__m128 SubRes = _mm_sub_ps(MulFacA, MulFacB);
+
+	__m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2));
+	__m128 SubFacC = _mm_shuffle_ps(SubTmpC, SubTmpC, _MM_SHUFFLE(3, 3, 2, 0));
+	__m128 SwpFacC = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(2, 3, 3, 3));
+	__m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC);
+
+	__m128 AddRes = _mm_add_ps(SubRes, MulFacC);
+	__m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f));
+
+	//return m[0][0] * DetCof[0]
+	//	 + m[0][1] * DetCof[1]
+	//	 + m[0][2] * DetCof[2]
+	//	 + m[0][3] * DetCof[3];
+
+	return glm_vec4_dot(m[0], DetCof);
+}
+
+GLM_FUNC_QUALIFIER void glm_mat4_inverse(glm_vec4 const in[4], glm_vec4 out[4])
+{
+	__m128 Fac0;
+	{
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac0 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac1;
+	{
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac1 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+
+	__m128 Fac2;
+	{
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac2 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac3;
+	{
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac3 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac4;
+	{
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac4 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac5;
+	{
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac5 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
+	__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
+
+	// m[1][0]
+	// m[0][0]
+	// m[0][0]
+	// m[0][0]
+	__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][1]
+	// m[0][1]
+	// m[0][1]
+	// m[0][1]
+	__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][2]
+	// m[0][2]
+	// m[0][2]
+	// m[0][2]
+	__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][3]
+	// m[0][3]
+	// m[0][3]
+	// m[0][3]
+	__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
+	__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// col0
+	// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
+	// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
+	// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
+	// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
+	__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
+	__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
+	__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
+	__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
+	__m128 Add00 = _mm_add_ps(Sub00, Mul02);
+	__m128 Inv0 = _mm_mul_ps(SignB, Add00);
+
+	// col1
+	// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
+	// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
+	// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
+	// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
+	__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
+	__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
+	__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
+	__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
+	__m128 Add01 = _mm_add_ps(Sub01, Mul05);
+	__m128 Inv1 = _mm_mul_ps(SignA, Add01);
+
+	// col2
+	// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
+	// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
+	// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
+	// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
+	__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
+	__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
+	__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
+	__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
+	__m128 Add02 = _mm_add_ps(Sub02, Mul08);
+	__m128 Inv2 = _mm_mul_ps(SignB, Add02);
+
+	// col3
+	// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
+	// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
+	// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
+	// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
+	__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
+	__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
+	__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
+	__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
+	__m128 Add03 = _mm_add_ps(Sub03, Mul11);
+	__m128 Inv3 = _mm_mul_ps(SignA, Add03);
+
+	__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
+
+	//	valType Determinant = m[0][0] * Inverse[0][0]
+	//						+ m[0][1] * Inverse[1][0]
+	//						+ m[0][2] * Inverse[2][0]
+	//						+ m[0][3] * Inverse[3][0];
+	__m128 Det0 = glm_vec4_dot(in[0], Row2);
+	__m128 Rcp0 = _mm_div_ps(_mm_set1_ps(1.0f), Det0);
+	//__m128 Rcp0 = _mm_rcp_ps(Det0);
+
+	//	Inverse /= Determinant;
+	out[0] = _mm_mul_ps(Inv0, Rcp0);
+	out[1] = _mm_mul_ps(Inv1, Rcp0);
+	out[2] = _mm_mul_ps(Inv2, Rcp0);
+	out[3] = _mm_mul_ps(Inv3, Rcp0);
+}
+
+GLM_FUNC_QUALIFIER void glm_mat4_inverse_lowp(glm_vec4 const in[4], glm_vec4 out[4])
+{
+	__m128 Fac0;
+	{
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac0 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac1;
+	{
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac1 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+
+	__m128 Fac2;
+	{
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac2 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac3;
+	{
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac3 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac4;
+	{
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac4 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac5;
+	{
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac5 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
+	__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
+
+	// m[1][0]
+	// m[0][0]
+	// m[0][0]
+	// m[0][0]
+	__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][1]
+	// m[0][1]
+	// m[0][1]
+	// m[0][1]
+	__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][2]
+	// m[0][2]
+	// m[0][2]
+	// m[0][2]
+	__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][3]
+	// m[0][3]
+	// m[0][3]
+	// m[0][3]
+	__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
+	__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// col0
+	// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
+	// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
+	// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
+	// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
+	__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
+	__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
+	__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
+	__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
+	__m128 Add00 = _mm_add_ps(Sub00, Mul02);
+	__m128 Inv0 = _mm_mul_ps(SignB, Add00);
+
+	// col1
+	// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
+	// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
+	// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
+	// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
+	__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
+	__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
+	__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
+	__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
+	__m128 Add01 = _mm_add_ps(Sub01, Mul05);
+	__m128 Inv1 = _mm_mul_ps(SignA, Add01);
+
+	// col2
+	// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
+	// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
+	// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
+	// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
+	__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
+	__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
+	__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
+	__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
+	__m128 Add02 = _mm_add_ps(Sub02, Mul08);
+	__m128 Inv2 = _mm_mul_ps(SignB, Add02);
+
+	// col3
+	// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
+	// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
+	// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
+	// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
+	__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
+	__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
+	__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
+	__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
+	__m128 Add03 = _mm_add_ps(Sub03, Mul11);
+	__m128 Inv3 = _mm_mul_ps(SignA, Add03);
+
+	__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
+
+	//	valType Determinant = m[0][0] * Inverse[0][0]
+	//						+ m[0][1] * Inverse[1][0]
+	//						+ m[0][2] * Inverse[2][0]
+	//						+ m[0][3] * Inverse[3][0];
+	__m128 Det0 = glm_vec4_dot(in[0], Row2);
+	__m128 Rcp0 = _mm_rcp_ps(Det0);
+	//__m128 Rcp0 = _mm_div_ps(one, Det0);
+	//	Inverse /= Determinant;
+	out[0] = _mm_mul_ps(Inv0, Rcp0);
+	out[1] = _mm_mul_ps(Inv1, Rcp0);
+	out[2] = _mm_mul_ps(Inv2, Rcp0);
+	out[3] = _mm_mul_ps(Inv3, Rcp0);
+}
+/*
+GLM_FUNC_QUALIFIER void glm_mat4_rotate(__m128 const in[4], float Angle, float const v[3], __m128 out[4])
+{
+	float a = glm::radians(Angle);
+	float c = cos(a);
+	float s = sin(a);
+
+	glm::vec4 AxisA(v[0], v[1], v[2], float(0));
+	__m128 AxisB = _mm_set_ps(AxisA.w, AxisA.z, AxisA.y, AxisA.x);
+	__m128 AxisC = detail::sse_nrm_ps(AxisB);
+
+	__m128 Cos0 = _mm_set_ss(c);
+	__m128 CosA = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Sin0 = _mm_set_ss(s);
+	__m128 SinA = _mm_shuffle_ps(Sin0, Sin0, _MM_SHUFFLE(0, 0, 0, 0));
+
+	// vec<3, T, Q> temp = (valType(1) - c) * axis;
+	__m128 Temp0 = _mm_sub_ps(one, CosA);
+	__m128 Temp1 = _mm_mul_ps(Temp0, AxisC);
+
+	//Rotate[0][0] = c + temp[0] * axis[0];
+	//Rotate[0][1] = 0 + temp[0] * axis[1] + s * axis[2];
+	//Rotate[0][2] = 0 + temp[0] * axis[2] - s * axis[1];
+	__m128 Axis0 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 TmpA0 = _mm_mul_ps(Axis0, AxisC);
+	__m128 CosA0 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 1, 0));
+	__m128 TmpA1 = _mm_add_ps(CosA0, TmpA0);
+	__m128 SinA0 = SinA;//_mm_set_ps(0.0f, s, -s, 0.0f);
+	__m128 TmpA2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 1, 2, 3));
+	__m128 TmpA3 = _mm_mul_ps(SinA0, TmpA2);
+	__m128 TmpA4 = _mm_add_ps(TmpA1, TmpA3);
+
+	//Rotate[1][0] = 0 + temp[1] * axis[0] - s * axis[2];
+	//Rotate[1][1] = c + temp[1] * axis[1];
+	//Rotate[1][2] = 0 + temp[1] * axis[2] + s * axis[0];
+	__m128 Axis1 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 TmpB0 = _mm_mul_ps(Axis1, AxisC);
+	__m128 CosA1 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 0, 1));
+	__m128 TmpB1 = _mm_add_ps(CosA1, TmpB0);
+	__m128 SinB0 = SinA;//_mm_set_ps(-s, 0.0f, s, 0.0f);
+	__m128 TmpB2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 0, 3, 2));
+	__m128 TmpB3 = _mm_mul_ps(SinA0, TmpB2);
+	__m128 TmpB4 = _mm_add_ps(TmpB1, TmpB3);
+
+	//Rotate[2][0] = 0 + temp[2] * axis[0] + s * axis[1];
+	//Rotate[2][1] = 0 + temp[2] * axis[1] - s * axis[0];
+	//Rotate[2][2] = c + temp[2] * axis[2];
+	__m128 Axis2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 TmpC0 = _mm_mul_ps(Axis2, AxisC);
+	__m128 CosA2 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 0, 1, 1));
+	__m128 TmpC1 = _mm_add_ps(CosA2, TmpC0);
+	__m128 SinC0 = SinA;//_mm_set_ps(s, -s, 0.0f, 0.0f);
+	__m128 TmpC2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 3, 0, 1));
+	__m128 TmpC3 = _mm_mul_ps(SinA0, TmpC2);
+	__m128 TmpC4 = _mm_add_ps(TmpC1, TmpC3);
+
+	__m128 Result[4];
+	Result[0] = TmpA4;
+	Result[1] = TmpB4;
+	Result[2] = TmpC4;
+	Result[3] = _mm_set_ps(1, 0, 0, 0);
+
+	//mat<4, 4, valType> Result;
+	//Result[0] = m[0] * Rotate[0][0] + m[1] * Rotate[0][1] + m[2] * Rotate[0][2];
+	//Result[1] = m[0] * Rotate[1][0] + m[1] * Rotate[1][1] + m[2] * Rotate[1][2];
+	//Result[2] = m[0] * Rotate[2][0] + m[1] * Rotate[2][1] + m[2] * Rotate[2][2];
+	//Result[3] = m[3];
+	//return Result;
+	sse_mul_ps(in, Result, out);
+}
+*/
+GLM_FUNC_QUALIFIER void glm_mat4_outerProduct(__m128 const& c, __m128 const& r, __m128 out[4])
+{
+	out[0] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(0, 0, 0, 0)));
+	out[1] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(1, 1, 1, 1)));
+	out[2] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(2, 2, 2, 2)));
+	out[3] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(3, 3, 3, 3)));
+}
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/neon.h b/thirdparty/manifold/thirdparty/glm/glm/simd/neon.h
new file mode 100644
index 000000000000..f85947f5c19a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/neon.h
@@ -0,0 +1,155 @@
+/// @ref simd_neon
+/// @file glm/simd/neon.h
+
+#pragma once
+
+#if GLM_ARCH & GLM_ARCH_NEON_BIT
+#include <arm_neon.h>
+
+namespace glm {
+	namespace neon {
+		static inline float32x4_t dupq_lane(float32x4_t vsrc, int lane) {
+			switch(lane) {
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+				case 0: return vdupq_laneq_f32(vsrc, 0);
+				case 1: return vdupq_laneq_f32(vsrc, 1);
+				case 2: return vdupq_laneq_f32(vsrc, 2);
+				case 3: return vdupq_laneq_f32(vsrc, 3);
+#else
+				case 0: return vdupq_n_f32(vgetq_lane_f32(vsrc, 0));
+				case 1: return vdupq_n_f32(vgetq_lane_f32(vsrc, 1));
+				case 2: return vdupq_n_f32(vgetq_lane_f32(vsrc, 2));
+				case 3: return vdupq_n_f32(vgetq_lane_f32(vsrc, 3));
+#endif
+			}
+			assert(!"Unreachable code executed!");
+			return vdupq_n_f32(0.0f);
+		}
+
+		static inline float32x2_t dup_lane(float32x4_t vsrc, int lane) {
+			switch(lane) {
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+				case 0: return vdup_laneq_f32(vsrc, 0);
+				case 1: return vdup_laneq_f32(vsrc, 1);
+				case 2: return vdup_laneq_f32(vsrc, 2);
+				case 3: return vdup_laneq_f32(vsrc, 3);
+#else
+				case 0: return vdup_n_f32(vgetq_lane_f32(vsrc, 0));
+				case 1: return vdup_n_f32(vgetq_lane_f32(vsrc, 1));
+				case 2: return vdup_n_f32(vgetq_lane_f32(vsrc, 2));
+				case 3: return vdup_n_f32(vgetq_lane_f32(vsrc, 3));
+#endif
+			}
+			assert(!"Unreachable code executed!");
+			return vdup_n_f32(0.0f);
+		}
+
+		static inline float32x4_t copy_lane(float32x4_t vdst, int dlane, float32x4_t vsrc, int slane) {
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+			switch(dlane) {
+				case 0:
+					switch(slane) {
+						case 0: return vcopyq_laneq_f32(vdst, 0, vsrc, 0);
+						case 1: return vcopyq_laneq_f32(vdst, 0, vsrc, 1);
+						case 2: return vcopyq_laneq_f32(vdst, 0, vsrc, 2);
+						case 3: return vcopyq_laneq_f32(vdst, 0, vsrc, 3);
+					}
+					assert(!"Unreachable code executed!");
+				case 1:
+					switch(slane) {
+						case 0: return vcopyq_laneq_f32(vdst, 1, vsrc, 0);
+						case 1: return vcopyq_laneq_f32(vdst, 1, vsrc, 1);
+						case 2: return vcopyq_laneq_f32(vdst, 1, vsrc, 2);
+						case 3: return vcopyq_laneq_f32(vdst, 1, vsrc, 3);
+					}
+					assert(!"Unreachable code executed!");
+				case 2:
+					switch(slane) {
+						case 0: return vcopyq_laneq_f32(vdst, 2, vsrc, 0);
+						case 1: return vcopyq_laneq_f32(vdst, 2, vsrc, 1);
+						case 2: return vcopyq_laneq_f32(vdst, 2, vsrc, 2);
+						case 3: return vcopyq_laneq_f32(vdst, 2, vsrc, 3);
+					}
+					assert(!"Unreachable code executed!");
+				case 3:
+					switch(slane) {
+						case 0: return vcopyq_laneq_f32(vdst, 3, vsrc, 0);
+						case 1: return vcopyq_laneq_f32(vdst, 3, vsrc, 1);
+						case 2: return vcopyq_laneq_f32(vdst, 3, vsrc, 2);
+						case 3: return vcopyq_laneq_f32(vdst, 3, vsrc, 3);
+					}
+					assert(!"Unreachable code executed!");
+			}
+#else
+
+			float l;
+			switch(slane) {
+				case 0: l = vgetq_lane_f32(vsrc, 0); break;
+				case 1: l = vgetq_lane_f32(vsrc, 1); break;
+				case 2: l = vgetq_lane_f32(vsrc, 2); break;
+				case 3: l = vgetq_lane_f32(vsrc, 3); break;
+				default: 
+					assert(!"Unreachable code executed!");
+			}
+			switch(dlane) {
+				case 0: return vsetq_lane_f32(l, vdst, 0);
+				case 1: return vsetq_lane_f32(l, vdst, 1);
+				case 2: return vsetq_lane_f32(l, vdst, 2);
+				case 3: return vsetq_lane_f32(l, vdst, 3);
+			}
+#endif
+			assert(!"Unreachable code executed!");
+			return vdupq_n_f32(0.0f);
+		}
+
+		static inline float32x4_t mul_lane(float32x4_t v, float32x4_t vlane, int lane) {
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+			switch(lane) { 
+				case 0: return vmulq_laneq_f32(v, vlane, 0); break;
+				case 1: return vmulq_laneq_f32(v, vlane, 1); break;
+				case 2: return vmulq_laneq_f32(v, vlane, 2); break;
+				case 3: return vmulq_laneq_f32(v, vlane, 3); break;
+				default: 
+					assert(!"Unreachable code executed!");
+			}
+			assert(!"Unreachable code executed!");
+			return vdupq_n_f32(0.0f);
+#else
+			return vmulq_f32(v, dupq_lane(vlane, lane));
+#endif
+		}
+
+		static inline float32x4_t madd_lane(float32x4_t acc, float32x4_t v, float32x4_t vlane, int lane) {
+#if GLM_ARCH & GLM_ARCH_ARMV8_BIT
+#ifdef GLM_CONFIG_FORCE_FMA
+#	define FMADD_LANE(acc, x, y, L) do { asm volatile ("fmla %0.4s, %1.4s, %2.4s" : "+w"(acc) : "w"(x), "w"(dup_lane(y, L))); } while(0)
+#else
+#	define FMADD_LANE(acc, x, y, L) do { acc = vmlaq_laneq_f32(acc, x, y, L); } while(0)
+#endif
+
+			switch(lane) { 
+				case 0: 
+					FMADD_LANE(acc, v, vlane, 0);
+					return acc;
+				case 1:
+					FMADD_LANE(acc, v, vlane, 1);
+					return acc;
+				case 2:
+					FMADD_LANE(acc, v, vlane, 2);
+					return acc;
+				case 3:
+					FMADD_LANE(acc, v, vlane, 3);
+					return acc;
+				default: 
+					assert(!"Unreachable code executed!");
+			}
+			assert(!"Unreachable code executed!");
+			return vdupq_n_f32(0.0f);
+#	undef FMADD_LANE
+#else
+			return vaddq_f32(acc, vmulq_f32(v, dupq_lane(vlane, lane)));
+#endif
+		}
+	} //namespace neon
+} // namespace glm
+#endif // GLM_ARCH & GLM_ARCH_NEON_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/packing.h b/thirdparty/manifold/thirdparty/glm/glm/simd/packing.h
new file mode 100644
index 000000000000..609163eb0d77
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/packing.h
@@ -0,0 +1,8 @@
+/// @ref simd
+/// @file glm/simd/packing.h
+
+#pragma once
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/platform.h b/thirdparty/manifold/thirdparty/glm/glm/simd/platform.h
new file mode 100644
index 000000000000..a318b098f9cf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/platform.h
@@ -0,0 +1,469 @@
+#pragma once
+
+///////////////////////////////////////////////////////////////////////////////////
+// Platform
+
+#define GLM_PLATFORM_UNKNOWN		0x00000000
+#define GLM_PLATFORM_WINDOWS		0x00010000
+#define GLM_PLATFORM_LINUX			0x00020000
+#define GLM_PLATFORM_APPLE			0x00040000
+//#define GLM_PLATFORM_IOS			0x00080000
+#define GLM_PLATFORM_ANDROID		0x00100000
+#define GLM_PLATFORM_CHROME_NACL	0x00200000
+#define GLM_PLATFORM_UNIX			0x00400000
+#define GLM_PLATFORM_QNXNTO			0x00800000
+#define GLM_PLATFORM_WINCE			0x01000000
+#define GLM_PLATFORM_CYGWIN			0x02000000
+
+#ifdef GLM_FORCE_PLATFORM_UNKNOWN
+#	define GLM_PLATFORM GLM_PLATFORM_UNKNOWN
+#elif defined(__CYGWIN__)
+#	define GLM_PLATFORM GLM_PLATFORM_CYGWIN
+#elif defined(__QNXNTO__)
+#	define GLM_PLATFORM GLM_PLATFORM_QNXNTO
+#elif defined(__APPLE__)
+#	define GLM_PLATFORM GLM_PLATFORM_APPLE
+#elif defined(WINCE)
+#	define GLM_PLATFORM GLM_PLATFORM_WINCE
+#elif defined(_WIN32)
+#	define GLM_PLATFORM GLM_PLATFORM_WINDOWS
+#elif defined(__native_client__)
+#	define GLM_PLATFORM GLM_PLATFORM_CHROME_NACL
+#elif defined(__ANDROID__)
+#	define GLM_PLATFORM GLM_PLATFORM_ANDROID
+#elif defined(__linux)
+#	define GLM_PLATFORM GLM_PLATFORM_LINUX
+#elif defined(__unix)
+#	define GLM_PLATFORM GLM_PLATFORM_UNIX
+#else
+#	define GLM_PLATFORM GLM_PLATFORM_UNKNOWN
+#endif//
+
+///////////////////////////////////////////////////////////////////////////////////
+// Compiler
+
+#define GLM_COMPILER_UNKNOWN		0x00000000
+
+// Intel
+#define GLM_COMPILER_INTEL			0x00100000
+#define GLM_COMPILER_INTEL14		0x00100040
+#define GLM_COMPILER_INTEL15		0x00100050
+#define GLM_COMPILER_INTEL16		0x00100060
+#define GLM_COMPILER_INTEL17		0x00100070
+#define GLM_COMPILER_INTEL18		0x00100080
+#define GLM_COMPILER_INTEL19		0x00100090
+#define GLM_COMPILER_INTEL21		0x001000A0
+
+// Visual C++ defines
+#define GLM_COMPILER_VC				0x01000000
+#define GLM_COMPILER_VC12			0x01000001 // Visual Studio 2013
+#define GLM_COMPILER_VC14			0x01000002 // Visual Studio 2015
+#define GLM_COMPILER_VC15			0x01000003 // Visual Studio 2017
+#define GLM_COMPILER_VC15_3			0x01000004
+#define GLM_COMPILER_VC15_5			0x01000005
+#define GLM_COMPILER_VC15_6			0x01000006
+#define GLM_COMPILER_VC15_7			0x01000007
+#define GLM_COMPILER_VC15_8			0x01000008
+#define GLM_COMPILER_VC15_9			0x01000009
+#define GLM_COMPILER_VC16			0x0100000A // Visual Studio 2019
+#define GLM_COMPILER_VC17			0x0100000B // Visual Studio 2022
+
+// GCC defines
+#define GLM_COMPILER_GCC			0x02000000
+#define GLM_COMPILER_GCC46			0x020000D0
+#define GLM_COMPILER_GCC47			0x020000E0
+#define GLM_COMPILER_GCC48			0x020000F0
+#define GLM_COMPILER_GCC49			0x02000100
+#define GLM_COMPILER_GCC5			0x02000200
+#define GLM_COMPILER_GCC6			0x02000300
+#define GLM_COMPILER_GCC61			0x02000800
+#define GLM_COMPILER_GCC7			0x02000400
+#define GLM_COMPILER_GCC8			0x02000500
+#define GLM_COMPILER_GCC9			0x02000600
+#define GLM_COMPILER_GCC10			0x02000700
+#define GLM_COMPILER_GCC11			0x02000800
+#define GLM_COMPILER_GCC12			0x02000900
+#define GLM_COMPILER_GCC13			0x02000A00
+#define GLM_COMPILER_GCC14			0x02000B00
+
+// CUDA
+#define GLM_COMPILER_CUDA			0x10000000
+#define GLM_COMPILER_CUDA75			0x10000001
+#define GLM_COMPILER_CUDA80			0x10000002
+#define GLM_COMPILER_CUDA90			0x10000004
+#define GLM_COMPILER_CUDA_RTC		0x10000100
+
+// Clang
+#define GLM_COMPILER_CLANG			0x20000000
+#define GLM_COMPILER_CLANG34		0x20000050
+#define GLM_COMPILER_CLANG35		0x20000060
+#define GLM_COMPILER_CLANG36		0x20000070
+#define GLM_COMPILER_CLANG37		0x20000080
+#define GLM_COMPILER_CLANG38		0x20000090
+#define GLM_COMPILER_CLANG39		0x200000A0
+#define GLM_COMPILER_CLANG4			0x200000B0
+#define GLM_COMPILER_CLANG5			0x200000C0
+#define GLM_COMPILER_CLANG6			0x200000D0
+#define GLM_COMPILER_CLANG7			0x200000E0
+#define GLM_COMPILER_CLANG8			0x200000F0
+#define GLM_COMPILER_CLANG9			0x20000100
+#define GLM_COMPILER_CLANG10		0x20000200
+#define GLM_COMPILER_CLANG11		0x20000300
+#define GLM_COMPILER_CLANG12		0x20000400
+#define GLM_COMPILER_CLANG13		0x20000500
+#define GLM_COMPILER_CLANG14		0x20000600
+#define GLM_COMPILER_CLANG15		0x20000700
+#define GLM_COMPILER_CLANG16		0x20000800
+#define GLM_COMPILER_CLANG17		0x20000900
+#define GLM_COMPILER_CLANG18		0x20000A00
+#define GLM_COMPILER_CLANG19		0x20000B00
+
+// HIP
+#define GLM_COMPILER_HIP			0x40000000
+
+// Build model
+#define GLM_MODEL_32				0x00000010
+#define GLM_MODEL_64				0x00000020
+
+// Force generic C++ compiler
+#ifdef GLM_FORCE_COMPILER_UNKNOWN
+#	define GLM_COMPILER GLM_COMPILER_UNKNOWN
+
+#elif defined(__INTEL_COMPILER)
+#	if __INTEL_COMPILER >= 2021
+#		define GLM_COMPILER GLM_COMPILER_INTEL21
+#	elif __INTEL_COMPILER >= 1900
+#		define GLM_COMPILER GLM_COMPILER_INTEL19
+#	elif __INTEL_COMPILER >= 1800
+#		define GLM_COMPILER GLM_COMPILER_INTEL18
+#	elif __INTEL_COMPILER >= 1700
+#		define GLM_COMPILER GLM_COMPILER_INTEL17
+#	elif __INTEL_COMPILER >= 1600
+#		define GLM_COMPILER GLM_COMPILER_INTEL16
+#	elif __INTEL_COMPILER >= 1500
+#		define GLM_COMPILER GLM_COMPILER_INTEL15
+#	elif __INTEL_COMPILER >= 1400
+#		define GLM_COMPILER GLM_COMPILER_INTEL14
+#	elif __INTEL_COMPILER < 1400
+#		error "GLM requires ICC 2013 SP1 or newer"
+#	endif
+
+// CUDA
+#elif defined(__CUDACC__)
+#	if !defined(CUDA_VERSION) && !defined(GLM_FORCE_CUDA)
+#		include <cuda.h>  // make sure version is defined since nvcc does not define it itself!
+#	endif
+#	if defined(__CUDACC_RTC__)
+#		define GLM_COMPILER GLM_COMPILER_CUDA_RTC
+#	elif CUDA_VERSION >= 8000
+#		define GLM_COMPILER GLM_COMPILER_CUDA80
+#	elif CUDA_VERSION >= 7500
+#		define GLM_COMPILER GLM_COMPILER_CUDA75
+#	elif CUDA_VERSION >= 7000
+#		define GLM_COMPILER GLM_COMPILER_CUDA70
+#	elif CUDA_VERSION < 7000
+#		error "GLM requires CUDA 7.0 or higher"
+#	endif
+
+// HIP
+#elif defined(__HIP__)
+#	define GLM_COMPILER GLM_COMPILER_HIP
+
+// Clang
+#elif defined(__clang__)
+#	if defined(__apple_build_version__)
+#		if (__clang_major__ < 6)
+#			error "GLM requires Clang 3.4 / Apple Clang 6.0 or higher"
+#		elif __clang_major__ == 6 && __clang_minor__ == 0
+#			define GLM_COMPILER GLM_COMPILER_CLANG35
+#		elif __clang_major__ == 6 && __clang_minor__ >= 1
+#			define GLM_COMPILER GLM_COMPILER_CLANG36
+#		elif __clang_major__ >= 7
+#			define GLM_COMPILER GLM_COMPILER_CLANG37
+#		endif
+#	else
+#		if ((__clang_major__ == 3) && (__clang_minor__ < 4)) || (__clang_major__ < 3)
+#			error "GLM requires Clang 3.4 or higher"
+#		elif __clang_major__ == 3 && __clang_minor__ == 4
+#			define GLM_COMPILER GLM_COMPILER_CLANG34
+#		elif __clang_major__ == 3 && __clang_minor__ == 5
+#			define GLM_COMPILER GLM_COMPILER_CLANG35
+#		elif __clang_major__ == 3 && __clang_minor__ == 6
+#			define GLM_COMPILER GLM_COMPILER_CLANG36
+#		elif __clang_major__ == 3 && __clang_minor__ == 7
+#			define GLM_COMPILER GLM_COMPILER_CLANG37
+#		elif __clang_major__ == 3 && __clang_minor__ == 8
+#			define GLM_COMPILER GLM_COMPILER_CLANG38
+#		elif __clang_major__ == 3 && __clang_minor__ >= 9
+#			define GLM_COMPILER GLM_COMPILER_CLANG39
+#		elif __clang_major__ == 4 && __clang_minor__ == 0
+#			define GLM_COMPILER GLM_COMPILER_CLANG4
+#		elif __clang_major__ == 5
+#			define GLM_COMPILER GLM_COMPILER_CLANG5
+#		elif __clang_major__ == 6
+#			define GLM_COMPILER GLM_COMPILER_CLANG6
+#		elif __clang_major__ == 7
+#			define GLM_COMPILER GLM_COMPILER_CLANG7
+#		elif __clang_major__ == 8
+#			define GLM_COMPILER GLM_COMPILER_CLANG8
+#		elif __clang_major__ == 9
+#			define GLM_COMPILER GLM_COMPILER_CLANG9
+#		elif __clang_major__ == 10
+#			define GLM_COMPILER GLM_COMPILER_CLANG10
+#		elif __clang_major__ == 11
+#			define GLM_COMPILER GLM_COMPILER_CLANG11
+#		elif __clang_major__ == 12
+#			define GLM_COMPILER GLM_COMPILER_CLANG12
+#		elif __clang_major__ == 13
+#			define GLM_COMPILER GLM_COMPILER_CLANG13
+#		elif __clang_major__ == 14
+#			define GLM_COMPILER GLM_COMPILER_CLANG14
+#		elif __clang_major__ == 15
+#			define GLM_COMPILER GLM_COMPILER_CLANG15
+#		elif __clang_major__ == 16
+#			define GLM_COMPILER GLM_COMPILER_CLANG16
+#		elif __clang_major__ == 17
+#			define GLM_COMPILER GLM_COMPILER_CLANG17
+#		elif __clang_major__ == 18
+#			define GLM_COMPILER GLM_COMPILER_CLANG18
+#		elif __clang_major__ >= 19
+#			define GLM_COMPILER GLM_COMPILER_CLANG19
+#		endif
+#	endif
+
+// Visual C++
+#elif defined(_MSC_VER)
+#   if _MSC_VER >= 1930
+#		define GLM_COMPILER GLM_COMPILER_VC17
+#	elif _MSC_VER >= 1920
+#		define GLM_COMPILER GLM_COMPILER_VC16
+#	elif _MSC_VER >= 1916
+#		define GLM_COMPILER GLM_COMPILER_VC15_9
+#	elif _MSC_VER >= 1915
+#		define GLM_COMPILER GLM_COMPILER_VC15_8
+#	elif _MSC_VER >= 1914
+#		define GLM_COMPILER GLM_COMPILER_VC15_7
+#	elif _MSC_VER >= 1913
+#		define GLM_COMPILER GLM_COMPILER_VC15_6
+#	elif _MSC_VER >= 1912
+#		define GLM_COMPILER GLM_COMPILER_VC15_5
+#	elif _MSC_VER >= 1911
+#		define GLM_COMPILER GLM_COMPILER_VC15_3
+#	elif _MSC_VER >= 1910
+#		define GLM_COMPILER GLM_COMPILER_VC15
+#	elif _MSC_VER >= 1900
+#		define GLM_COMPILER GLM_COMPILER_VC14
+#	elif _MSC_VER >= 1800
+#		define GLM_COMPILER GLM_COMPILER_VC12
+#	elif _MSC_VER < 1800
+#		error "GLM requires Visual C++ 12 - 2013 or higher"
+#	endif//_MSC_VER
+
+// G++
+#elif defined(__GNUC__) || defined(__MINGW32__)
+#	if __GNUC__ >= 14
+#		define GLM_COMPILER GLM_COMPILER_GCC14
+#	elif __GNUC__ >= 13
+#		define GLM_COMPILER GLM_COMPILER_GCC13
+#	elif __GNUC__ >= 12
+#		define GLM_COMPILER GLM_COMPILER_GCC12
+#	elif __GNUC__ >= 11
+#		define GLM_COMPILER GLM_COMPILER_GCC11
+#	elif __GNUC__ >= 10
+#		define GLM_COMPILER GLM_COMPILER_GCC10
+#	elif __GNUC__ >= 9
+#		define GLM_COMPILER GLM_COMPILER_GCC9
+#	elif __GNUC__ >= 8
+#		define GLM_COMPILER GLM_COMPILER_GCC8
+#	elif __GNUC__ >= 7
+#		define GLM_COMPILER GLM_COMPILER_GCC7
+#	elif __GNUC__ >= 6
+#		define GLM_COMPILER GLM_COMPILER_GCC6
+#	elif __GNUC__ >= 5
+#		define GLM_COMPILER GLM_COMPILER_GCC5
+#	elif __GNUC__ == 4 && __GNUC_MINOR__ >= 9
+#		define GLM_COMPILER GLM_COMPILER_GCC49
+#	elif __GNUC__ == 4 && __GNUC_MINOR__ >= 8
+#		define GLM_COMPILER GLM_COMPILER_GCC48
+#	elif __GNUC__ == 4 && __GNUC_MINOR__ >= 7
+#		define GLM_COMPILER GLM_COMPILER_GCC47
+#	elif __GNUC__ == 4 && __GNUC_MINOR__ >= 6
+#		define GLM_COMPILER GLM_COMPILER_GCC46
+#	elif ((__GNUC__ == 4) && (__GNUC_MINOR__ < 6)) || (__GNUC__ < 4)
+#		error "GLM requires GCC 4.6 or higher"
+#	endif
+
+#else
+#	define GLM_COMPILER GLM_COMPILER_UNKNOWN
+#endif
+
+#ifndef GLM_COMPILER
+#	error "GLM_COMPILER undefined, your compiler may not be supported by GLM. Add #define GLM_COMPILER 0 to ignore this message."
+#endif//GLM_COMPILER
+
+///////////////////////////////////////////////////////////////////////////////////
+// Instruction sets
+
+// User defines: GLM_FORCE_PURE GLM_FORCE_INTRINSICS GLM_FORCE_SSE2 GLM_FORCE_SSE3 GLM_FORCE_AVX GLM_FORCE_AVX2 GLM_FORCE_AVX2
+
+#define GLM_ARCH_MIPS_BIT	  (0x10000000)
+#define GLM_ARCH_PPC_BIT	  (0x20000000)
+#define GLM_ARCH_ARM_BIT	  (0x40000000)
+#define GLM_ARCH_ARMV8_BIT  (0x01000000)
+#define GLM_ARCH_X86_BIT	  (0x80000000)
+
+#define GLM_ARCH_SIMD_BIT	(0x00001000)
+
+#define GLM_ARCH_NEON_BIT	(0x00000001)
+#define GLM_ARCH_SSE_BIT	(0x00000002)
+#define GLM_ARCH_SSE2_BIT	(0x00000004)
+#define GLM_ARCH_SSE3_BIT	(0x00000008)
+#define GLM_ARCH_SSSE3_BIT	(0x00000010)
+#define GLM_ARCH_SSE41_BIT	(0x00000020)
+#define GLM_ARCH_SSE42_BIT	(0x00000040)
+#define GLM_ARCH_AVX_BIT	(0x00000080)
+#define GLM_ARCH_AVX2_BIT	(0x00000100)
+
+#define GLM_ARCH_UNKNOWN	(0)
+#define GLM_ARCH_X86		(GLM_ARCH_X86_BIT)
+#define GLM_ARCH_SSE		(GLM_ARCH_SSE_BIT | GLM_ARCH_SIMD_BIT | GLM_ARCH_X86)
+#define GLM_ARCH_SSE2		(GLM_ARCH_SSE2_BIT | GLM_ARCH_SSE)
+#define GLM_ARCH_SSE3		(GLM_ARCH_SSE3_BIT | GLM_ARCH_SSE2)
+#define GLM_ARCH_SSSE3		(GLM_ARCH_SSSE3_BIT | GLM_ARCH_SSE3)
+#define GLM_ARCH_SSE41		(GLM_ARCH_SSE41_BIT | GLM_ARCH_SSSE3)
+#define GLM_ARCH_SSE42		(GLM_ARCH_SSE42_BIT | GLM_ARCH_SSE41)
+#define GLM_ARCH_AVX		(GLM_ARCH_AVX_BIT | GLM_ARCH_SSE42)
+#define GLM_ARCH_AVX2		(GLM_ARCH_AVX2_BIT | GLM_ARCH_AVX)
+#define GLM_ARCH_ARM		(GLM_ARCH_ARM_BIT)
+#define GLM_ARCH_ARMV8		(GLM_ARCH_NEON_BIT | GLM_ARCH_SIMD_BIT | GLM_ARCH_ARM | GLM_ARCH_ARMV8_BIT)
+#define GLM_ARCH_NEON		(GLM_ARCH_NEON_BIT | GLM_ARCH_SIMD_BIT | GLM_ARCH_ARM)
+#define GLM_ARCH_MIPS		(GLM_ARCH_MIPS_BIT)
+#define GLM_ARCH_PPC		(GLM_ARCH_PPC_BIT)
+
+#if defined(GLM_FORCE_ARCH_UNKNOWN) || defined(GLM_FORCE_PURE)
+#	define GLM_ARCH GLM_ARCH_UNKNOWN
+#elif defined(GLM_FORCE_NEON)
+#	if __ARM_ARCH >= 8
+#		define GLM_ARCH (GLM_ARCH_ARMV8)
+#	else
+#		define GLM_ARCH (GLM_ARCH_NEON)
+#	endif
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_AVX2)
+#	define GLM_ARCH (GLM_ARCH_AVX2)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_AVX)
+#	define GLM_ARCH (GLM_ARCH_AVX)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_SSE42)
+#	define GLM_ARCH (GLM_ARCH_SSE42)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_SSE41)
+#	define GLM_ARCH (GLM_ARCH_SSE41)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_SSSE3)
+#	define GLM_ARCH (GLM_ARCH_SSSE3)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_SSE3)
+#	define GLM_ARCH (GLM_ARCH_SSE3)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_SSE2)
+#	define GLM_ARCH (GLM_ARCH_SSE2)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_SSE)
+#	define GLM_ARCH (GLM_ARCH_SSE)
+#	define GLM_FORCE_INTRINSICS
+#elif defined(GLM_FORCE_INTRINSICS) && !defined(GLM_FORCE_XYZW_ONLY)
+#	if defined(__AVX2__)
+#		define GLM_ARCH (GLM_ARCH_AVX2)
+#	elif defined(__AVX__)
+#		define GLM_ARCH (GLM_ARCH_AVX)
+#	elif defined(__SSE4_2__)
+#		define GLM_ARCH (GLM_ARCH_SSE42)
+#	elif defined(__SSE4_1__)
+#		define GLM_ARCH (GLM_ARCH_SSE41)
+#	elif defined(__SSSE3__)
+#		define GLM_ARCH (GLM_ARCH_SSSE3)
+#	elif defined(__SSE3__)
+#		define GLM_ARCH (GLM_ARCH_SSE3)
+#	elif defined(__SSE2__) || defined(__x86_64__) || defined(_M_X64) || defined(_M_IX86_FP)
+#		define GLM_ARCH (GLM_ARCH_SSE2)
+#	elif defined(__i386__)
+#		define GLM_ARCH (GLM_ARCH_X86)
+#	elif defined(__ARM_ARCH) && (__ARM_ARCH >= 8)
+#		define GLM_ARCH (GLM_ARCH_ARMV8)
+#	elif defined(__ARM_NEON)
+#		define GLM_ARCH (GLM_ARCH_ARM | GLM_ARCH_NEON)
+#	elif defined(__arm__ ) || defined(_M_ARM)
+#		define GLM_ARCH (GLM_ARCH_ARM)
+#	elif defined(__mips__ )
+#		define GLM_ARCH (GLM_ARCH_MIPS)
+#	elif defined(__powerpc__ ) || defined(_M_PPC)
+#		define GLM_ARCH (GLM_ARCH_PPC)
+#	else
+#		define GLM_ARCH (GLM_ARCH_UNKNOWN)
+#	endif
+#else
+#	if defined(__x86_64__) || defined(_M_X64) || defined(_M_IX86) || defined(__i386__)
+#		define GLM_ARCH (GLM_ARCH_X86)
+#	elif defined(__arm__) || defined(_M_ARM)
+#		define GLM_ARCH (GLM_ARCH_ARM)
+#	elif defined(__powerpc__) || defined(_M_PPC)
+#		define GLM_ARCH (GLM_ARCH_PPC)
+#	elif defined(__mips__)
+#		define GLM_ARCH (GLM_ARCH_MIPS)
+#	else
+#		define GLM_ARCH (GLM_ARCH_UNKNOWN)
+#	endif
+#endif
+
+#if GLM_ARCH & GLM_ARCH_AVX2_BIT
+#	include <immintrin.h>
+#elif GLM_ARCH & GLM_ARCH_AVX_BIT
+#	include <immintrin.h>
+#elif GLM_ARCH & GLM_ARCH_SSE42_BIT
+#	if GLM_COMPILER & GLM_COMPILER_CLANG
+#		include <popcntintrin.h>
+#	endif
+#	include <nmmintrin.h>
+#elif GLM_ARCH & GLM_ARCH_SSE41_BIT
+#	include <smmintrin.h>
+#elif GLM_ARCH & GLM_ARCH_SSSE3_BIT
+#	include <tmmintrin.h>
+#elif GLM_ARCH & GLM_ARCH_SSE3_BIT
+#	include <pmmintrin.h>
+#elif GLM_ARCH & GLM_ARCH_SSE2_BIT
+#	include <emmintrin.h>
+#elif GLM_ARCH & GLM_ARCH_NEON_BIT
+#	include "neon.h"
+#endif//GLM_ARCH
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+	typedef __m128			glm_f32vec4;
+	typedef __m128i			glm_i32vec4;
+	typedef __m128i			glm_u32vec4;
+	typedef __m128d			glm_f64vec2;
+	typedef __m128i			glm_i64vec2;
+	typedef __m128i			glm_u64vec2;
+
+	typedef glm_f32vec4		glm_vec4;
+	typedef glm_i32vec4		glm_ivec4;
+	typedef glm_u32vec4		glm_uvec4;
+	typedef glm_f64vec2		glm_dvec2;
+#endif
+
+#if GLM_ARCH & GLM_ARCH_AVX_BIT
+	typedef __m256d			glm_f64vec4;
+	typedef glm_f64vec4		glm_dvec4;
+#endif
+
+#if GLM_ARCH & GLM_ARCH_AVX2_BIT
+	typedef __m256i			glm_i64vec4;
+	typedef __m256i			glm_u64vec4;
+#endif
+
+#if GLM_ARCH & GLM_ARCH_NEON_BIT
+	typedef float32x4_t			glm_f32vec4;
+	typedef int32x4_t			glm_i32vec4;
+	typedef uint32x4_t			glm_u32vec4;
+#endif
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/trigonometric.h b/thirdparty/manifold/thirdparty/glm/glm/simd/trigonometric.h
new file mode 100644
index 000000000000..739b796e7e45
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/trigonometric.h
@@ -0,0 +1,9 @@
+/// @ref simd
+/// @file glm/simd/trigonometric.h
+
+#pragma once
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/simd/vector_relational.h b/thirdparty/manifold/thirdparty/glm/glm/simd/vector_relational.h
new file mode 100644
index 000000000000..f7385e974736
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/simd/vector_relational.h
@@ -0,0 +1,8 @@
+/// @ref simd
+/// @file glm/simd/vector_relational.h
+
+#pragma once
+
+#if GLM_ARCH & GLM_ARCH_SSE2_BIT
+
+#endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
diff --git a/thirdparty/manifold/thirdparty/glm/glm/trigonometric.hpp b/thirdparty/manifold/thirdparty/glm/glm/trigonometric.hpp
new file mode 100644
index 000000000000..51d49c132bb5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/trigonometric.hpp
@@ -0,0 +1,210 @@
+/// @ref core
+/// @file glm/trigonometric.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+///
+/// @defgroup core_func_trigonometric Angle and Trigonometry Functions
+/// @ingroup core
+///
+/// Function parameters specified as angle are assumed to be in units of radians.
+/// In no case will any of these functions result in a divide by zero error. If
+/// the divisor of a ratio is 0, then results will be undefined.
+///
+/// These all operate component-wise. The description is per component.
+///
+/// Include <glm/trigonometric.hpp> to use these core features.
+///
+/// @see ext_vector_trigonometric
+
+#pragma once
+
+#include "detail/setup.hpp"
+#include "detail/qualifier.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_func_trigonometric
+	/// @{
+
+	/// Converts degrees to radians and returns the result.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/radians.xml">GLSL radians man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> radians(vec<L, T, Q> const& degrees);
+
+	/// Converts radians to degrees and returns the result.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/degrees.xml">GLSL degrees man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, T, Q> degrees(vec<L, T, Q> const& radians);
+
+	/// The standard trigonometric sine function.
+	/// The values returned by this function will range from [-1, 1].
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/sin.xml">GLSL sin man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> sin(vec<L, T, Q> const& angle);
+
+	/// The standard trigonometric cosine function.
+	/// The values returned by this function will range from [-1, 1].
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/cos.xml">GLSL cos man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> cos(vec<L, T, Q> const& angle);
+
+	/// The standard trigonometric tangent function.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/tan.xml">GLSL tan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> tan(vec<L, T, Q> const& angle);
+
+	/// Arc sine. Returns an angle whose sine is x.
+	/// The range of values returned by this function is [-PI/2, PI/2].
+	/// Results are undefined if |x| > 1.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/asin.xml">GLSL asin man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> asin(vec<L, T, Q> const& x);
+
+	/// Arc cosine. Returns an angle whose cosine is x.
+	/// The range of values returned by this function is [0, PI].
+	/// Results are undefined if |x| > 1.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/acos.xml">GLSL acos man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> acos(vec<L, T, Q> const& x);
+
+	/// Arc tangent. Returns an angle whose tangent is y/x.
+	/// The signs of x and y are used to determine what
+	/// quadrant the angle is in. The range of values returned
+	/// by this function is [-PI, PI]. Results are undefined
+	/// if x and y are both 0.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/atan.xml">GLSL atan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> atan(vec<L, T, Q> const& y, vec<L, T, Q> const& x);
+
+	/// Arc tangent. Returns an angle whose tangent is y_over_x.
+	/// The range of values returned by this function is [-PI/2, PI/2].
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/atan.xml">GLSL atan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> atan(vec<L, T, Q> const& y_over_x);
+
+	/// Returns the hyperbolic sine function, (exp(x) - exp(-x)) / 2
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/sinh.xml">GLSL sinh man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> sinh(vec<L, T, Q> const& angle);
+
+	/// Returns the hyperbolic cosine function, (exp(x) + exp(-x)) / 2
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/cosh.xml">GLSL cosh man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> cosh(vec<L, T, Q> const& angle);
+
+	/// Returns the hyperbolic tangent function, sinh(angle) / cosh(angle)
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/tanh.xml">GLSL tanh man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> tanh(vec<L, T, Q> const& angle);
+
+	/// Arc hyperbolic sine; returns the inverse of sinh.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/asinh.xml">GLSL asinh man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> asinh(vec<L, T, Q> const& x);
+
+	/// Arc hyperbolic cosine; returns the non-negative inverse
+	/// of cosh. Results are undefined if x < 1.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/acosh.xml">GLSL acosh man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> acosh(vec<L, T, Q> const& x);
+
+	/// Arc hyperbolic tangent; returns the inverse of tanh.
+	/// Results are undefined if abs(x) >= 1.
+	///
+	/// @tparam L Integer between 1 and 4 included that qualify the dimension of the vector
+	/// @tparam T Floating-point scalar types
+	/// @tparam Q Value from qualifier enum
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/atanh.xml">GLSL atanh man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.1 Angle and Trigonometry Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL vec<L, T, Q> atanh(vec<L, T, Q> const& x);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_trigonometric.inl"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/vec2.hpp b/thirdparty/manifold/thirdparty/glm/glm/vec2.hpp
new file mode 100644
index 000000000000..cd4e0708e109
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/vec2.hpp
@@ -0,0 +1,14 @@
+/// @ref core
+/// @file glm/vec2.hpp
+
+#pragma once
+#include "./ext/vector_bool2.hpp"
+#include "./ext/vector_bool2_precision.hpp"
+#include "./ext/vector_float2.hpp"
+#include "./ext/vector_float2_precision.hpp"
+#include "./ext/vector_double2.hpp"
+#include "./ext/vector_double2_precision.hpp"
+#include "./ext/vector_int2.hpp"
+#include "./ext/vector_int2_sized.hpp"
+#include "./ext/vector_uint2.hpp"
+#include "./ext/vector_uint2_sized.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/vec3.hpp b/thirdparty/manifold/thirdparty/glm/glm/vec3.hpp
new file mode 100644
index 000000000000..f5a927dbe4c5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/vec3.hpp
@@ -0,0 +1,14 @@
+/// @ref core
+/// @file glm/vec3.hpp
+
+#pragma once
+#include "./ext/vector_bool3.hpp"
+#include "./ext/vector_bool3_precision.hpp"
+#include "./ext/vector_float3.hpp"
+#include "./ext/vector_float3_precision.hpp"
+#include "./ext/vector_double3.hpp"
+#include "./ext/vector_double3_precision.hpp"
+#include "./ext/vector_int3.hpp"
+#include "./ext/vector_int3_sized.hpp"
+#include "./ext/vector_uint3.hpp"
+#include "./ext/vector_uint3_sized.hpp"
diff --git a/thirdparty/manifold/thirdparty/glm/glm/vec4.hpp b/thirdparty/manifold/thirdparty/glm/glm/vec4.hpp
new file mode 100644
index 000000000000..c6ea9f1ff4c5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/vec4.hpp
@@ -0,0 +1,15 @@
+/// @ref core
+/// @file glm/vec4.hpp
+
+#pragma once
+#include "./ext/vector_bool4.hpp"
+#include "./ext/vector_bool4_precision.hpp"
+#include "./ext/vector_float4.hpp"
+#include "./ext/vector_float4_precision.hpp"
+#include "./ext/vector_double4.hpp"
+#include "./ext/vector_double4_precision.hpp"
+#include "./ext/vector_int4.hpp"
+#include "./ext/vector_int4_sized.hpp"
+#include "./ext/vector_uint4.hpp"
+#include "./ext/vector_uint4_sized.hpp"
+
diff --git a/thirdparty/manifold/thirdparty/glm/glm/vector_relational.hpp b/thirdparty/manifold/thirdparty/glm/glm/vector_relational.hpp
new file mode 100644
index 000000000000..a0fe17eb707a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/glm/glm/vector_relational.hpp
@@ -0,0 +1,121 @@
+/// @ref core
+/// @file glm/vector_relational.hpp
+///
+/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+///
+/// @defgroup core_func_vector_relational Vector Relational Functions
+/// @ingroup core
+///
+/// Relational and equality operators (<, <=, >, >=, ==, !=) are defined to
+/// operate on scalars and produce scalar Boolean results. For vector results,
+/// use the following built-in functions.
+///
+/// In all cases, the sizes of all the input and return vectors for any particular
+/// call must match.
+///
+/// Include <glm/vector_relational.hpp> to use these core features.
+///
+/// @see ext_vector_relational
+
+#pragma once
+
+#include "detail/qualifier.hpp"
+#include "detail/setup.hpp"
+
+namespace glm
+{
+	/// @addtogroup core_func_vector_relational
+	/// @{
+
+	/// Returns the component-wise comparison result of x < y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T A floating-point or integer scalar type.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/lessThan.xml">GLSL lessThan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> lessThan(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x <= y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T A floating-point or integer scalar type.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/lessThanEqual.xml">GLSL lessThanEqual man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> lessThanEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x > y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T A floating-point or integer scalar type.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/greaterThan.xml">GLSL greaterThan man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> greaterThan(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x >= y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T A floating-point or integer scalar type.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/greaterThanEqual.xml">GLSL greaterThanEqual man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> greaterThanEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x == y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T A floating-point, integer or bool scalar type.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/equal.xml">GLSL equal man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> equal(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns the component-wise comparison of result x != y.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	/// @tparam T A floating-point, integer or bool scalar type.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/notEqual.xml">GLSL notEqual man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, typename T, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> notEqual(vec<L, T, Q> const& x, vec<L, T, Q> const& y);
+
+	/// Returns true if any component of x is true.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/any.xml">GLSL any man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool any(vec<L, bool, Q> const& v);
+
+	/// Returns true if all components of x are true.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/all.xml">GLSL all man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR bool all(vec<L, bool, Q> const& v);
+
+	/// Returns the component-wise logical complement of x.
+	/// /!\ Because of language incompatibilities between C++ and GLSL, GLM defines the function not but not_ instead.
+	///
+	/// @tparam L An integer between 1 and 4 included that qualify the dimension of the vector.
+	///
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/not.xml">GLSL not man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.7 Vector Relational Functions</a>
+	template<length_t L, qualifier Q>
+	GLM_FUNC_DECL GLM_CONSTEXPR vec<L, bool, Q> not_(vec<L, bool, Q> const& v);
+
+	/// @}
+}//namespace glm
+
+#include "detail/func_vector_relational.inl"
diff --git a/thirdparty/manifold/thirdparty/thrust/.gitrepo b/thirdparty/manifold/thirdparty/thrust/.gitrepo
new file mode 100644
index 000000000000..f317a7ec8608
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/.gitrepo
@@ -0,0 +1,12 @@
+; DO NOT EDIT (unless you know what you are doing)
+;
+; This subdirectory is a git "subrepo", and this file is maintained by the
+; git-subrepo command. See https://github.com/ingydotnet/git-subrepo#readme
+;
+[subrepo]
+	remote = https://github.com/NVIDIA/thrust.git
+	branch = 2.1.0
+	commit = 3cd56842c94de4926157f6ccdfbbf03ef7e5d5dc
+	parent = 2150401aec548d8d3ca01554d59b2e3e113db527
+	method = merge
+	cmdver = 0.4.6
diff --git a/thirdparty/manifold/thirdparty/thrust/LICENSE b/thirdparty/manifold/thirdparty/thrust/LICENSE
new file mode 100644
index 000000000000..c22c225631e0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/LICENSE
@@ -0,0 +1,249 @@
+Unless otherwise noted, Thrust's source code is released under the Apache
+License, Version 2.0:
+
+================================================================================
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+================================================================================
+
+Some portions of Thrust may be licensed under other compatible open-source
+licenses. Any divergence from the Apache 2 license will be noted in the source
+code where applicable.
+
+Portions under other terms include, but are not limited to:
+
+================================================================================
+
+Various C++ utility classes in Thrust are based on the Boost Iterator, Tuple,
+System, and Random Number libraries, which are provided under the Boost Software
+License:
+
+    Boost Software License - Version 1.0 - August 17th, 2003
+
+    Permission is hereby granted, free of charge, to any person or organization
+    obtaining a copy of the software and accompanying documentation covered by
+    this license (the "Software") to use, reproduce, display, distribute,
+    execute, and transmit the Software, and to prepare derivative works of the
+    Software, and to permit third-parties to whom the Software is furnished to
+    do so, all subject to the following:
+
+    The copyright notices in the Software and this entire statement, including
+    the above license grant, this restriction and the following disclaimer,
+    must be included in all copies of the Software, in whole or in part, and
+    all derivative works of the Software, unless such copies or derivative
+    works are solely in the form of machine-executable object code generated by
+    a source language processor.
+
+    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+    FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT
+    SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE
+    FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE,
+    ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+    DEALINGS IN THE SOFTWARE.
+
+================================================================================
+
+Portions of the thrust::complex implementation are derived from FreeBSD with the
+following terms:
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions
+    are met:
+
+    1. Redistributions of source code must retain the above copyright
+       notice[1] unmodified, this list of conditions, and the following
+       disclaimer.
+    2. Redistributions in binary form must reproduce the above copyright
+       notice, this list of conditions and the following disclaimer in the
+       documentation and/or other materials provided with the distribution.
+
+    THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+    IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+    OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+    IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+    INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+    NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+    THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+[1] Individual copyright notices from the original authors are included in
+    the relevant source files.
+
+================================================================================
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/.gitrepo b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/.gitrepo
new file mode 100644
index 000000000000..b3ac4946d915
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/.gitrepo
@@ -0,0 +1,12 @@
+; DO NOT EDIT (unless you know what you are doing)
+;
+; This subdirectory is a git "subrepo", and this file is maintained by the
+; git-subrepo command. See https://github.com/ingydotnet/git-subrepo#readme
+;
+[subrepo]
+	remote = https://github.com/NVIDIA/libcudacxx.git
+	branch = 55dd2c99346baa3a14949a0f7e9c41865e434eda
+	commit = 55dd2c99346baa3a14949a0f7e9c41865e434eda
+	parent = 2308846c6cc63451eee7932ca936b112aad00c4e
+	method = merge
+	cmdver = 0.4.6
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/LICENSE.TXT b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/LICENSE.TXT
new file mode 100644
index 000000000000..bef1adba4af2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/LICENSE.TXT
@@ -0,0 +1,368 @@
+==============================================================================
+libcu++ is under the Apache License v2.0 with LLVM Exceptions:
+==============================================================================
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+    TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+    1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+    2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+    3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+    4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+    5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+    6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+    7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+    8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+    9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+    END OF TERMS AND CONDITIONS
+
+    APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+    Copyright [yyyy] [name of copyright owner]
+
+    Licensed under the Apache License, Version 2.0 (the "License");
+    you may not use this file except in compliance with the License.
+    You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+    Unless required by applicable law or agreed to in writing, software
+    distributed under the License is distributed on an "AS IS" BASIS,
+    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+    See the License for the specific language governing permissions and
+    limitations under the License.
+
+
+---- LLVM Exceptions to the Apache 2.0 License ----
+
+As an exception, if, as a result of your compiling your source code, portions
+of this Software are embedded into an Object form of such source code, you
+may redistribute such embedded portions in such Object form without complying
+with the conditions of Sections 4(a), 4(b) and 4(d) of the License.
+
+In addition, if you combine or link compiled forms of this Software with
+software that is licensed under the GPLv2 ("Combined Software") and if a
+court of competent jurisdiction determines that the patent provision (Section
+3), the indemnity provision (Section 9) or other Section of the License
+conflicts with the conditions of the GPLv2, you may retroactively and
+prospectively choose to deem waived or otherwise exclude such Section(s) of
+the License, but only in their entirety and only with respect to the Combined
+Software.
+
+==============================================================================
+Software from third parties included in the LLVM Project:
+==============================================================================
+The LLVM Project contains third party software which is under different license
+terms. All such code will be identified clearly using at least one of two
+mechanisms:
+1) It will be in a separate directory tree with its own `LICENSE.txt` or
+   `LICENSE` file at the top containing the specific license and restrictions
+   which apply to that software, or
+2) It will contain specific license and restriction terms at the top of every
+   file.
+
+==============================================================================
+Legacy LLVM License (https://llvm.org/docs/DeveloperPolicy.html#legacy):
+==============================================================================
+
+The libc++ library is dual licensed under both the University of Illinois
+"BSD-Like" license and the MIT license.  As a user of this code you may choose
+to use it under either license.  As a contributor, you agree to allow your code
+to be used under both.
+
+Full text of the relevant licenses is included below.
+
+==============================================================================
+
+University of Illinois/NCSA
+Open Source License
+
+Copyright (c) 2009-2019 by the contributors listed in CREDITS.TXT
+
+All rights reserved.
+
+Developed by:
+
+    LLVM Team
+
+    University of Illinois at Urbana-Champaign
+
+    http://llvm.org
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal with
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+
+    * Redistributions of source code must retain the above copyright notice,
+      this list of conditions and the following disclaimers.
+
+    * Redistributions in binary form must reproduce the above copyright notice,
+      this list of conditions and the following disclaimers in the
+      documentation and/or other materials provided with the distribution.
+
+    * Neither the names of the LLVM Team, University of Illinois at
+      Urbana-Champaign, nor the names of its contributors may be used to
+      endorse or promote products derived from this Software without specific
+      prior written permission.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS WITH THE
+SOFTWARE.
+
+==============================================================================
+
+Copyright (c) 2009-2014 by the contributors listed in CREDITS.TXT
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+
+==============================================================================
+
+Some libcudacxx components are covered by the below license. Each source file
+indicates which license it is under.
+
+==============================================================================
+
+NVIDIA SOFTWARE LICENSE
+
+This license is a legal agreement between you and NVIDIA Corporation ("NVIDIA") and governs your use of the NVIDIA/CUDA C++ Library software and materials provided hereunder (“SOFTWARE”).
+
+This license can be accepted only by an adult of legal age of majority in the country in which the SOFTWARE is used. If you are under the legal age of majority, you must ask your parent or legal guardian to consent to this license. By taking delivery of the SOFTWARE, you affirm that you have reached the legal age of majority, you accept the terms of this license, and you take legal and financial responsibility for the actions of your permitted users.
+
+You agree to use the SOFTWARE only for purposes that are permitted by (a) this license, and (b) any applicable law, regulation or generally accepted practices or guidelines in the relevant jurisdictions.
+
+1. LICENSE. Subject to the terms of this license, NVIDIA grants you a non-exclusive limited license to: (a) install and use the SOFTWARE, and (b) distribute the SOFTWARE subject to the distribution requirements described in this license. NVIDIA reserves all rights, title and interest in and to the SOFTWARE not expressly granted to you under this license.
+
+2. DISTRIBUTION REQUIREMENTS. These are the distribution requirements for you to exercise the distribution grant:
+a.	The terms under which you distribute the SOFTWARE must be consistent with the terms of this license, including (without limitation) terms relating to the license grant and license restrictions and protection of NVIDIA’s intellectual property rights.
+b.	You agree to notify NVIDIA in writing of any known or suspected distribution or use of the SOFTWARE not in compliance with the requirements of this license, and to enforce the terms of your agreements with respect to distributed SOFTWARE.
+
+3. LIMITATIONS. Your license to use the SOFTWARE is restricted as follows:
+a.	The SOFTWARE is licensed for you to develop applications only for use in systems with NVIDIA GPUs.
+b.	You may not reverse engineer, decompile or disassemble, or remove copyright or other proprietary notices from any portion of the SOFTWARE or copies of the SOFTWARE.
+c.	You may not modify or create derivative works of any portion of the SOFTWARE.
+d.	You may not bypass, disable, or circumvent any technical measure, encryption, security, digital rights management or authentication mechanism in the SOFTWARE.
+e.	You may not use the SOFTWARE in any manner that would cause it to become subject to an open source software license. As examples, licenses that require as a condition of use, modification, and/or distribution that the SOFTWARE be (i) disclosed or distributed in source code form; (ii) licensed for the purpose of making derivative works; or (iii) redistributable at no charge.
+f.	Unless you have an agreement with NVIDIA for this purpose, you may not use the SOFTWARE with any system or application where the use or failure of the system or application can reasonably be expected to threaten or result in personal injury, death, or catastrophic loss. Examples include use in avionics, navigation, military, medical, life support or other life critical applications. NVIDIA does not design, test or manufacture the SOFTWARE for these critical uses and NVIDIA shall not be liable to you or any third party, in whole or in part, for any claims or damages arising from such uses.
+g.	You agree to defend, indemnify and hold harmless NVIDIA and its affiliates, and their respective employees, contractors, agents, officers and directors, from and against any and all claims, damages, obligations, losses, liabilities, costs or debt, fines, restitutions and expenses (including but not limited to attorney’s fees and costs incident to establishing the right of indemnification) arising out of or related to use of the SOFTWARE outside of the scope of this Agreement, or not in compliance with its terms.
+
+4. PRE-RELEASE. SOFTWARE versions identified as alpha, beta, preview, early access or otherwise as pre-release may not be fully functional, may contain errors or design flaws, and may have reduced or different security, privacy, availability, and reliability standards relative to commercial versions of NVIDIA software and materials. You may use a pre-release SOFTWARE version at your own risk, understanding that these versions are not intended for use in production or business-critical systems.
+
+5. OWNERSHIP. The SOFTWARE and the related intellectual property rights therein are and will remain the sole and exclusive property of NVIDIA or its licensors. The SOFTWARE is copyrighted and protected by the laws of the United States and other countries, and international treaty provisions. NVIDIA may make changes to the SOFTWARE, at any time without notice, but is not obligated to support or update the SOFTWARE.
+
+6. COMPONENTS UNDER OTHER LICENSES. The SOFTWARE may include NVIDIA or third-party components with separate legal notices or terms as may be described in proprietary notices accompanying the SOFTWARE. If and to the extent there is a conflict between the terms in this license and the license terms associated with a component, the license terms associated with the components control only to the extent necessary to resolve the conflict.
+
+7. FEEDBACK. You may, but don’t have to, provide to NVIDIA any Feedback. “Feedback” means any suggestions, bug fixes, enhancements, modifications, feature requests or other feedback regarding the SOFTWARE. For any Feedback that you voluntarily provide, you hereby grant NVIDIA and its affiliates a perpetual, non-exclusive, worldwide, irrevocable license to use, reproduce, modify, license, sublicense (through multiple tiers of sublicensees), and distribute (through multiple tiers of distributors) the Feedback without the payment of any royalties or fees to you. NVIDIA will use Feedback at its choice.
+
+8. NO WARRANTIES. THE SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY EXPRESS OR IMPLIED WARRANTY OF ANY KIND INCLUDING, BUT NOT LIMITED TO, WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, OR FITNESS FOR A PARTICULAR PURPOSE. NVIDIA DOES NOT WARRANT THAT THE SOFTWARE WILL MEET YOUR REQUIREMENTS OR THAT THE OPERATION THEREOF WILL BE UNINTERRUPTED OR ERROR-FREE, OR THAT ALL ERRORS WILL BE CORRECTED.
+
+9. LIMITATIONS OF LIABILITY. TO THE MAXIMUM EXTENT PERMITTED BY LAW, NVIDIA AND ITS AFFILIATES SHALL NOT BE LIABLE FOR ANY SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL DAMAGES, OR ANY LOST PROFITS, PROJECT DELAYS, LOSS OF USE, LOSS OF DATA OR LOSS OF GOODWILL, OR THE COSTS OF PROCURING SUBSTITUTE PRODUCTS, ARISING OUT OF OR IN CONNECTION WITH THIS LICENSE OR THE USE OR PERFORMANCE OF THE SOFTWARE, WHETHER SUCH LIABILITY ARISES FROM ANY CLAIM BASED UPON BREACH OF CONTRACT, BREACH OF WARRANTY, TORT (INCLUDING NEGLIGENCE), PRODUCT LIABILITY OR ANY OTHER CAUSE OF ACTION OR THEORY OF LIABILITY, EVEN IF NVIDIA HAS PREVIOUSLY BEEN ADVISED OF, OR COULD REASONABLY HAVE FORESEEN, THE POSSIBILITY OF SUCH DAMAGES. IN NO EVENT WILL NVIDIA’S AND ITS AFFILIATES TOTAL CUMULATIVE LIABILITY UNDER OR ARISING OUT OF THIS LICENSE EXCEED US$10.00. THE NATURE OF THE LIABILITY OR THE NUMBER OF CLAIMS OR SUITS SHALL NOT ENLARGE OR EXTEND THIS LIMIT.
+
+10. TERMINATION. Your rights under this license will terminate automatically without notice from NVIDIA if you fail to comply with any term and condition of this license or if you commence or participate in any legal proceeding against NVIDIA with respect to the SOFTWARE. NVIDIA may terminate this license with advance written notice to you if NVIDIA decides to no longer provide the SOFTWARE in a country or, in NVIDIA’s sole discretion, the continued use of it is no longer commercially viable. Upon any termination of this license, you agree to promptly discontinue use of the SOFTWARE and destroy all copies in your possession or control. Your prior distributions in accordance with this license are not affected by the termination of this license. All provisions of this license will survive termination, except for the license granted to you.
+
+11. APPLICABLE LAW. This license will be governed in all respects by the laws of the United States and of the State of Delaware as those laws are applied to contracts entered into and performed entirely within Delaware by Delaware residents, without regard to the conflicts of laws principles. The United Nations Convention on Contracts for the International Sale of Goods is specifically disclaimed. You agree to all terms of this Agreement in the English language. The state or federal courts residing in Santa Clara County, California shall have exclusive jurisdiction over any dispute or claim arising out of this license. Notwithstanding this, you agree that NVIDIA shall still be allowed to apply for injunctive remedies or an equivalent type of urgent legal relief in any jurisdiction.
+
+12. NO ASSIGNMENT. This license and your rights and obligations thereunder may not be assigned by you by any means or operation of law without NVIDIA’s permission. Any attempted assignment not approved by NVIDIA in writing shall be void and of no effect.
+
+13. EXPORT. The SOFTWARE is subject to United States export laws and regulations. You agree that you will not ship, transfer or export the SOFTWARE into any country, or use the SOFTWARE in any manner, prohibited by the United States Bureau of Industry and Security or economic sanctions regulations administered by the U.S. Department of Treasury’s Office of Foreign Assets Control (OFAC), or any applicable export laws, restrictions or regulations. These laws include restrictions on destinations, end users and end use. By accepting this license, you confirm that you are not a resident or citizen of any country currently embargoed by the U.S. and that you are not otherwise prohibited from receiving the SOFTWARE.
+
+14. GOVERNMENT USE. The SOFTWARE has been developed entirely at private expense and is “commercial items” consisting of “commercial computer software” and “commercial computer software documentation” provided with RESTRICTED RIGHTS. Use, duplication or disclosure by the U.S. Government or a U.S. Government subcontractor is subject to the restrictions in this license pursuant to DFARS 227.7202-3(a) or as set forth in subparagraphs (b)(1) and (2) of the Commercial Computer Software - Restricted Rights clause at FAR 52.227-19, as applicable. Contractor/manufacturer is NVIDIA, 2788 San Tomas Expressway, Santa Clara, CA 95051.
+
+15. ENTIRE AGREEMENT. This license is the final, complete and exclusive agreement between the parties relating to the subject matter of this license and supersedes all prior or contemporaneous understandings and agreements relating to this subject matter, whether oral or written. If any court of competent jurisdiction determines that any provision of this license is illegal, invalid or unenforceable, the remaining provisions will remain in full force and effect. This license may only be modified in a writing signed by an authorized representative of each party.
+
+(v. August 20, 2021)
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/annotated_ptr b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/annotated_ptr
new file mode 100644
index 000000000000..fcb6e20d71ac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/annotated_ptr
@@ -0,0 +1,337 @@
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2021 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ *
+ * NVIDIA SOFTWARE LICENSE
+ *
+ * This license is a legal agreement between you and NVIDIA Corporation ("NVIDIA") and governs your use of the NVIDIA/CUDA C++ Library software and materials provided hereunder (“SOFTWARE”).
+ *
+ * This license can be accepted only by an adult of legal age of majority in the country in which the SOFTWARE is used. If you are under the legal age of majority, you must ask your parent or legal guardian to consent to this license. By taking delivery of the SOFTWARE, you affirm that you have reached the legal age of majority, you accept the terms of this license, and you take legal and financial responsibility for the actions of your permitted users.
+ *
+ * You agree to use the SOFTWARE only for purposes that are permitted by (a) this license, and (b) any applicable law, regulation or generally accepted practices or guidelines in the relevant jurisdictions.
+ *
+ * 1. LICENSE. Subject to the terms of this license, NVIDIA grants you a non-exclusive limited license to: (a) install and use the SOFTWARE, and (b) distribute the SOFTWARE subject to the distribution requirements described in this license. NVIDIA reserves all rights, title and interest in and to the SOFTWARE not expressly granted to you under this license.
+ *
+ * 2. DISTRIBUTION REQUIREMENTS. These are the distribution requirements for you to exercise the distribution grant:
+ * a.      The terms under which you distribute the SOFTWARE must be consistent with the terms of this license, including (without limitation) terms relating to the license grant and license restrictions and protection of NVIDIA’s intellectual property rights.
+ * b.      You agree to notify NVIDIA in writing of any known or suspected distribution or use of the SOFTWARE not in compliance with the requirements of this license, and to enforce the terms of your agreements with respect to distributed SOFTWARE.
+ *
+ * 3. LIMITATIONS. Your license to use the SOFTWARE is restricted as follows:
+ * a.      The SOFTWARE is licensed for you to develop applications only for use in systems with NVIDIA GPUs.
+ * b.      You may not reverse engineer, decompile or disassemble, or remove copyright or other proprietary notices from any portion of the SOFTWARE or copies of the SOFTWARE.
+ * c.      You may not modify or create derivative works of any portion of the SOFTWARE.
+ * d.      You may not bypass, disable, or circumvent any technical measure, encryption, security, digital rights management or authentication mechanism in the SOFTWARE.
+ * e.      You may not use the SOFTWARE in any manner that would cause it to become subject to an open source software license. As examples, licenses that require as a condition of use, modification, and/or distribution that the SOFTWARE be (i) disclosed or distributed in source code form; (ii) licensed for the purpose of making derivative works; or (iii) redistributable at no charge.
+ * f.      Unless you have an agreement with NVIDIA for this purpose, you may not use the SOFTWARE with any system or application where the use or failure of the system or application can reasonably be expected to threaten or result in personal injury, death, or catastrophic loss. Examples include use in avionics, navigation, military, medical, life support or other life critical applications. NVIDIA does not design, test or manufacture the SOFTWARE for these critical uses and NVIDIA shall not be liable to you or any third party, in whole or in part, for any claims or damages arising from such uses.
+ * g.      You agree to defend, indemnify and hold harmless NVIDIA and its affiliates, and their respective employees, contractors, agents, officers and directors, from and against any and all claims, damages, obligations, losses, liabilities, costs or debt, fines, restitutions and expenses (including but not limited to attorney’s fees and costs incident to establishing the right of indemnification) arising out of or related to use of the SOFTWARE outside of the scope of this Agreement, or not in compliance with its terms.
+ *
+ * 4. PRE-RELEASE. SOFTWARE versions identified as alpha, beta, preview, early access or otherwise as pre-release may not be fully functional, may contain errors or design flaws, and may have reduced or different security, privacy, availability, and reliability standards relative to commercial versions of NVIDIA software and materials. You may use a pre-release SOFTWARE version at your own risk, understanding that these versions are not intended for use in production or business-critical systems.
+ *
+ * 5. OWNERSHIP. The SOFTWARE and the related intellectual property rights therein are and will remain the sole and exclusive property of NVIDIA or its licensors. The SOFTWARE is copyrighted and protected by the laws of the United States and other countries, and international treaty provisions. NVIDIA may make changes to the SOFTWARE, at any time without notice, but is not obligated to support or update the SOFTWARE.
+ *
+ * 6. COMPONENTS UNDER OTHER LICENSES. The SOFTWARE may include NVIDIA or third-party components with separate legal notices or terms as may be described in proprietary notices accompanying the SOFTWARE. If and to the extent there is a conflict between the terms in this license and the license terms associated with a component, the license terms associated with the components control only to the extent necessary to resolve the conflict.
+ *
+ * 7. FEEDBACK. You may, but don’t have to, provide to NVIDIA any Feedback. “Feedback” means any suggestions, bug fixes, enhancements, modifications, feature requests or other feedback regarding the SOFTWARE. For any Feedback that you voluntarily provide, you hereby grant NVIDIA and its affiliates a perpetual, non-exclusive, worldwide, irrevocable license to use, reproduce, modify, license, sublicense (through multiple tiers of sublicensees), and distribute (through multiple tiers of distributors) the Feedback without the payment of any royalties or fees to you. NVIDIA will use Feedback at its choice.
+ *
+ * 8. NO WARRANTIES. THE SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY EXPRESS OR IMPLIED WARRANTY OF ANY KIND INCLUDING, BUT NOT LIMITED TO, WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, OR FITNESS FOR A PARTICULAR PURPOSE. NVIDIA DOES NOT WARRANT THAT THE SOFTWARE WILL MEET YOUR REQUIREMENTS OR THAT THE OPERATION THEREOF WILL BE UNINTERRUPTED OR ERROR-FREE, OR THAT ALL ERRORS WILL BE CORRECTED.
+ *
+ * 9. LIMITATIONS OF LIABILITY. TO THE MAXIMUM EXTENT PERMITTED BY LAW, NVIDIA AND ITS AFFILIATES SHALL NOT BE LIABLE FOR ANY SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL DAMAGES, OR ANY LOST PROFITS, PROJECT DELAYS, LOSS OF USE, LOSS OF DATA OR LOSS OF GOODWILL, OR THE COSTS OF PROCURING SUBSTITUTE PRODUCTS, ARISING OUT OF OR IN CONNECTION WITH THIS LICENSE OR THE USE OR PERFORMANCE OF THE SOFTWARE, WHETHER SUCH LIABILITY ARISES FROM ANY CLAIM BASED UPON BREACH OF CONTRACT, BREACH OF WARRANTY, TORT (INCLUDING NEGLIGENCE), PRODUCT LIABILITY OR ANY OTHER CAUSE OF ACTION OR THEORY OF LIABILITY, EVEN IF NVIDIA HAS PREVIOUSLY BEEN ADVISED OF, OR COULD REASONABLY HAVE FORESEEN, THE POSSIBILITY OF SUCH DAMAGES. IN NO EVENT WILL NVIDIA’S AND ITS AFFILIATES TOTAL CUMULATIVE LIABILITY UNDER OR ARISING OUT OF THIS LICENSE EXCEED US$10.00. THE NATURE OF THE LIABILITY OR THE NUMBER OF CLAIMS OR SUITS SHALL NOT ENLARGE OR EXTEND THIS LIMIT.
+ *
+ * 10. TERMINATION. Your rights under this license will terminate automatically without notice from NVIDIA if you fail to comply with any term and condition of this license or if you commence or participate in any legal proceeding against NVIDIA with respect to the SOFTWARE. NVIDIA may terminate this license with advance written notice to you if NVIDIA decides to no longer provide the SOFTWARE in a country or, in NVIDIA’s sole discretion, the continued use of it is no longer commercially viable. Upon any termination of this license, you agree to promptly discontinue use of the SOFTWARE and destroy all copies in your possession or control. Your prior distributions in accordance with this license are not affected by the termination of this license. All provisions of this license will survive termination, except for the license granted to you.
+ *
+ * 11. APPLICABLE LAW. This license will be governed in all respects by the laws of the United States and of the State of Delaware as those laws are applied to contracts entered into and performed entirely within Delaware by Delaware residents, without regard to the conflicts of laws principles. The United Nations Convention on Contracts for the International Sale of Goods is specifically disclaimed. You agree to all terms of this Agreement in the English language. The state or federal courts residing in Santa Clara County, California shall have exclusive jurisdiction over any dispute or claim arising out of this license. Notwithstanding this, you agree that NVIDIA shall still be allowed to apply for injunctive remedies or an equivalent type of urgent legal relief in any jurisdiction.
+ *
+ * 12. NO ASSIGNMENT. This license and your rights and obligations thereunder may not be assigned by you by any means or operation of law without NVIDIA’s permission. Any attempted assignment not approved by NVIDIA in writing shall be void and of no effect.
+ *
+ * 13. EXPORT. The SOFTWARE is subject to United States export laws and regulations. You agree that you will not ship, transfer or export the SOFTWARE into any country, or use the SOFTWARE in any manner, prohibited by the United States Bureau of Industry and Security or economic sanctions regulations administered by the U.S. Department of Treasury’s Office of Foreign Assets Control (OFAC), or any applicable export laws, restrictions or regulations. These laws include restrictions on destinations, end users and end use. By accepting this license, you confirm that you are not a resident or citizen of any country currently embargoed by the U.S. and that you are not otherwise prohibited from receiving the SOFTWARE.
+ *
+ * 14. GOVERNMENT USE. The SOFTWARE has been developed entirely at private expense and is “commercial items” consisting of “commercial computer software” and “commercial computer software documentation” provided with RESTRICTED RIGHTS. Use, duplication or disclosure by the U.S. Government or a U.S. Government subcontractor is subject to the restrictions in this license pursuant to DFARS 227.7202-3(a) or as set forth in subparagraphs (b)(1) and (2) of the Commercial Computer Software - Restricted Rights clause at FAR 52.227-19, as applicable. Contractor/manufacturer is NVIDIA, 2788 San Tomas Expressway, Santa Clara, CA 95051.
+ *
+ * 15. ENTIRE AGREEMENT. This license is the final, complete and exclusive agreement between the parties relating to the subject matter of this license and supersedes all prior or contemporaneous understandings and agreements relating to this subject matter, whether oral or written. If any court of competent jurisdiction determines that any provision of this license is illegal, invalid or unenforceable, the remaining provisions will remain in full force and effect. This license may only be modified in a writing signed by an authorized representative of each party.
+ *
+ * (v. August 20, 2021)
+ */
+
+#ifndef _CUDA_ANNOTATED_PTR
+#define _CUDA_ANNOTATED_PTR
+
+#include <cuda/std/cstdint>
+#include <cuda/barrier>
+
+#include "std/detail/__access_property"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+class access_property {
+  private:
+    std::uint64_t __descriptor = 0;
+
+  public:
+    struct shared {};
+    struct global {};
+    struct persisting {
+      __host__ __device__ constexpr operator cudaAccessProperty() const noexcept {
+        return cudaAccessProperty::cudaAccessPropertyPersisting;
+      }
+    };
+    struct streaming {
+      __host__ __device__ constexpr operator cudaAccessProperty() const noexcept {
+        return cudaAccessProperty::cudaAccessPropertyStreaming;
+      }
+    };
+    struct normal {
+      __host__ __device__ constexpr operator cudaAccessProperty() const noexcept {
+        return cudaAccessProperty::cudaAccessPropertyNormal;
+      }
+    };
+
+    __host__ __device__ constexpr access_property(global) noexcept : __descriptor(__detail_ap::__sm_80::__interleave_normal()) {}
+    __host__ __device__ constexpr access_property() noexcept : __descriptor(__detail_ap::__sm_80::__interleave_normal()) {}
+    constexpr access_property(access_property const&) noexcept = default;
+    access_property& operator=(const access_property& other) noexcept = default;
+
+    __host__ __device__ constexpr access_property(normal, float __fraction) : __descriptor(__detail_ap::__interleave(normal{}, __fraction)) {}
+    __host__ __device__ constexpr access_property(streaming, float __fraction) : __descriptor(__detail_ap::__interleave(streaming{}, __fraction)) {}
+    __host__ __device__ constexpr access_property(persisting, float __fraction) : __descriptor(__detail_ap::__interleave(persisting{}, __fraction)) {}
+    __host__ __device__ constexpr access_property(normal, float __fraction, streaming) : __descriptor(__detail_ap::__interleave(normal{}, __fraction, streaming{})) {}
+    __host__ __device__ constexpr access_property(persisting, float __fraction, streaming) : __descriptor(__detail_ap::__interleave(persisting{}, __fraction, streaming{})) {}
+
+    __host__ __device__ constexpr access_property(normal) noexcept : access_property(normal{}, 1.0) {}
+    __host__ __device__ constexpr access_property(streaming) noexcept : access_property(streaming{}, 1.0) {}
+    __host__ __device__ constexpr access_property(persisting) noexcept : access_property(persisting{}, 1.0) {}
+
+    __host__ __device__ constexpr access_property(void* __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, normal)
+      : __descriptor(__detail_ap::__block(__ptr, __hit_bytes, __total_bytes, normal{})) {}
+    __host__ __device__ constexpr access_property(void* __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, streaming)
+      : __descriptor(__detail_ap::__block(__ptr, __hit_bytes, __total_bytes, streaming{})) {}
+    __host__ __device__ constexpr access_property(void* __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, persisting)
+      : __descriptor(__detail_ap::__block(__ptr, __hit_bytes, __total_bytes, persisting{})) {}
+    __host__ __device__ constexpr access_property(void* __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, normal, streaming)
+      : __descriptor(__detail_ap::__block(__ptr, __hit_bytes, __total_bytes, normal{}, streaming{})) {}
+    __host__ __device__ constexpr access_property(void* __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, persisting, streaming)
+      : __descriptor(__detail_ap::__block(__ptr, __hit_bytes, __total_bytes, persisting{}, streaming{})) {}
+
+    __host__ __device__ constexpr explicit operator std::uint64_t() const noexcept { return __descriptor; }
+};
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#include "std/detail/__annotated_ptr"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+template <class _Tp, class _Property>
+__host__ __device__
+_Tp* associate_access_property(_Tp* __ptr, _Property __prop) {
+  static_assert(
+    std::is_same<_Property, access_property>::value ||
+    std::is_same<_Property, access_property::persisting>::value ||
+    std::is_same<_Property, access_property::streaming>::value ||
+    std::is_same<_Property, access_property::normal>::value ||
+    std::is_same<_Property, access_property::global>::value ||
+    std::is_same<_Property, access_property::shared>::value
+      , "property is not convertible to cuda::access_property");
+  return __detail_ap::__associate(__ptr, __prop);
+}
+
+template <class _Shape>
+__host__ __device__
+void apply_access_property(const volatile void* __ptr, const _Shape __shape, access_property::persisting __prop) noexcept {
+#if __CUDA_ARCH__ >= 800
+  if (!__isGlobal((void*)__ptr)) return;
+
+  char* __p = reinterpret_cast<char*>(const_cast<void*>(__ptr));
+  static constexpr std::size_t _LINE_SIZE = 128;
+  std::size_t __nbytes = static_cast<std::size_t>(__shape);
+  std::size_t __end = ((std::uintptr_t)(__p + __nbytes) % _LINE_SIZE) ? __nbytes + _LINE_SIZE : __nbytes;
+  __end /= _LINE_SIZE;
+
+  //Apply to all 128 bytes aligned cache lines inclusive of __p
+  for (std::size_t __i = 0; __i < __end; __i += _LINE_SIZE) {
+    asm volatile ("prefetch.global.L2::evict_last [%0];" ::"l"(__p + (__i * _LINE_SIZE)) :);
+  }
+#endif
+}
+
+template <class _Shape>
+__host__ __device__
+void apply_access_property(const volatile void* __ptr, const _Shape __shape, access_property::normal __prop) noexcept {
+#if __CUDA_ARCH__ >= 800
+  if (!__isGlobal((void*)__ptr)) return;
+
+  char* __p = reinterpret_cast<char*>(const_cast<void*>(__ptr));
+  static constexpr std::size_t _LINE_SIZE = 128;
+  std::size_t __nbytes = static_cast<std::size_t>(__shape);
+  std::size_t __end = ((std::uintptr_t)(__p + __nbytes) % _LINE_SIZE) ? __nbytes + _LINE_SIZE : __nbytes;
+  __end /= _LINE_SIZE;
+
+  //Apply to all 128 bytes aligned cache lines inclusive of __p
+  for (std::size_t __i = 0; __i < __end; __i += _LINE_SIZE) {
+    asm volatile ("prefetch.global.L2::evict_normal [%0];" ::"l"(__p + (__i * _LINE_SIZE)) :);
+  }
+#endif
+}
+
+inline
+__host__ __device__
+void discard_memory(volatile void* __ptr, std::size_t __nbytes) noexcept {
+#if __CUDA_ARCH__ >= 800
+  if (!__isGlobal((void*)__ptr)) return;
+
+  char* __p = reinterpret_cast<char*>(const_cast<void*>(__ptr));
+  static constexpr std::size_t _LINE_SIZE = 128;
+  std::size_t __start = (reinterpret_cast<std::uintptr_t>(__p) % _LINE_SIZE) ? 1 : 0;
+  std::size_t __end = (reinterpret_cast<std::uintptr_t>(__p + __nbytes) % _LINE_SIZE) ? __nbytes - _LINE_SIZE : __nbytes;
+  __end /= _LINE_SIZE;
+
+  //Trim the first block and last block if they're not 128 bytes aligned
+  for (std::size_t __i = __start; __i < __end; __i += _LINE_SIZE) {
+    asm volatile ("discard.global.L2 [%0], 128;" ::"l"(__p + (__i * _LINE_SIZE)) :);
+  }
+#endif
+}
+
+template<class _Tp, class _Property>
+class annotated_ptr: public __detail_ap::__annotated_ptr_base<_Property> {
+  public:
+    using value_type = _Tp;
+    using size_type = std::size_t;
+    using reference = value_type&;
+    using pointer = value_type*;
+    using const_pointer = value_type const*;
+    using difference_type = std::ptrdiff_t;
+
+  private:
+    using __self = annotated_ptr<_Tp, _Property>;
+
+    // Converting from a 64-bit to 32-bit shared pointer and maybe back just for storage might or might not be profitable.
+    pointer __repr = (pointer)((size_type)nullptr);
+
+    __host__ __device__ pointer __get(bool __skip_prop = false, difference_type __n = 0) const {
+#ifdef __CUDA_ARCH__
+      if (!__skip_prop) {
+        return static_cast<pointer>(this->__apply_prop(const_cast<void*>(static_cast<const volatile void*>(__repr + __n))));
+      }
+#endif
+      return __repr + __n;
+    }
+    __host__ __device__ pointer __offset(difference_type __n, bool __skip_prop = false) const {
+      return __get(__skip_prop, __n);
+    }
+
+  public:
+    __host__ __device__ pointer operator->() const {
+      return __get();
+    }
+
+    __host__ __device__ reference operator*() const {
+      return *__get();
+    }
+
+    __host__ __device__ reference operator[](difference_type __n) const {
+      return *__offset(__n);
+    }
+
+    __host__ __device__ constexpr difference_type operator-(annotated_ptr o) const {
+      return __repr - o.__repr;
+    }
+
+    constexpr annotated_ptr() noexcept = default;
+    constexpr annotated_ptr(annotated_ptr const&) noexcept = default;
+    // No constexpr for c11 as the method can't be const
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 annotated_ptr& operator=(annotated_ptr const& other) noexcept = default;
+
+    __host__ __device__ explicit annotated_ptr(pointer __p)
+      : __repr(__p)
+      {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2 && defined(__CUDA_ARCH__)
+      _LIBCUDACXX_DEBUG_ASSERT(std::is_same<_Property, shared>::value && __isShared(__p) || __isGlobal(__p));
+#endif
+    }
+
+    template <typename _RuntimeProperty>
+    __host__ __device__ annotated_ptr(pointer __p, _RuntimeProperty __prop)
+      : __detail_ap::__annotated_ptr_base<_Property>(static_cast<std::uint64_t>(access_property(__prop))), __repr(__p)
+    {
+      static_assert(std::is_same<_Property, access_property>::value,
+		    "This method requires annotated_ptr<T, cuda::access_property>");
+      static_assert(std::is_same<_RuntimeProperty, access_property::global>::value ||
+                    std::is_same<_RuntimeProperty, access_property::normal>::value ||
+                    std::is_same<_RuntimeProperty, access_property::streaming>::value ||
+                    std::is_same<_RuntimeProperty, access_property::persisting>::value ||
+                    std::is_same<_RuntimeProperty, access_property>::value,
+                    "This method requires RuntimeProperty=global|normal|streaming|persisting|access_property");
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2 && defined(__CUDA_ARCH__)
+      _LIBCUDACXX_DEBUG_ASSERT(__isGlobal(__p) == true);
+#endif
+    }
+
+    template<class _TTp, class _Prop>
+    __host__ __device__ annotated_ptr(const annotated_ptr<_TTp,_Prop>& __other);
+
+    __host__ __device__ constexpr explicit operator bool() const noexcept {
+      return __repr != nullptr;
+    }
+
+    __host__ __device__ pointer get() const noexcept {
+      constexpr bool __is_shared = std::is_same<_Property, access_property::shared>::value;
+      return __is_shared ? __repr : &(*annotated_ptr<value_type, access_property::global>(__repr));
+    }
+
+    __host__ __device__ _Property __property() const noexcept {
+      return this->__get_property();
+    }
+};
+
+
+template<class _Tp, class _Property>
+template<class _TTp, class _Prop>
+__host__ __device__ annotated_ptr<_Tp, _Property>::annotated_ptr(const annotated_ptr<_TTp,_Prop>& __other)
+  : __detail_ap::__annotated_ptr_base<_Property>(__other.__property()), __repr(__other.get())
+{
+  static_assert(std::is_assignable<pointer&, _TTp*>::value, "pointer must be assignable from other pointer");
+  static_assert((std::is_same<_Property, access_property>::value && !std::is_same<_Prop, access_property::shared>::value) ||
+		std::is_same<_Property, _Prop>::value, "Property must be either access_property or other property, and both properties must have same address space");
+  // note: precondition "__other.__rep must be compatible with _Property" currently always holds
+}
+
+template<class _Dst, class _Src, class _SrcProperty, class _Shape, class _Sync>
+__host__ __device__
+void memcpy_async(_Dst* __dst,
+    annotated_ptr<_Src,_SrcProperty> __src,
+    _Shape __shape, _Sync & __sync) {
+  memcpy_async(__dst, &(*__src), __shape, __sync);
+}
+
+template<class _Dst, class _DstProperty, class _Src, class _SrcProperty,
+  class _Shape, class _Sync>
+__host__ __device__
+void memcpy_async(annotated_ptr<_Dst,_DstProperty> __dst,
+    annotated_ptr<_Src,_SrcProperty> __src,
+    _Shape __shape, _Sync & __sync){
+  memcpy_async(&(*__dst), &(*__src), __shape, __sync);
+}
+
+template<class _Group, class _Dst, class _Src, class _SrcProperty,
+  class _Shape, class _Sync>
+__host__ __device__
+void memcpy_async(const _Group & __group,
+    _Dst * __dst,
+    annotated_ptr<_Src,_SrcProperty> __src,
+    _Shape __shape, _Sync & __sync) {
+  memcpy_async(__group, __dst, &(*__src), __shape, __sync);
+}
+
+template<class _Group, class _Dst, class _DstProperty, class _Src, class _SrcProperty,
+  class _Shape, class _Sync>
+__host__ __device__
+void memcpy_async(const _Group & __group,
+    annotated_ptr<_Dst,_DstProperty> __dst,
+    annotated_ptr<_Src,_SrcProperty> __src,
+    _Shape __shape, _Sync & __sync) {
+  memcpy_async(__group, &(*__dst), &(*__src), __shape, __sync);
+}
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#endif // _CUDA_ANNOTATED_PTR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/atomic b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/atomic
new file mode 100644
index 000000000000..b19967087b3f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/atomic
@@ -0,0 +1,10 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "std/atomic"
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/barrier b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/barrier
new file mode 100644
index 000000000000..616a4013887c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/barrier
@@ -0,0 +1,10 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "std/barrier"
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/latch b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/latch
new file mode 100644
index 000000000000..7b19dce5e5e4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/latch
@@ -0,0 +1,10 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "std/latch"
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/pipeline b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/pipeline
new file mode 100644
index 000000000000..54cdcb1b3598
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/pipeline
@@ -0,0 +1,635 @@
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2021 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ *
+ * NVIDIA SOFTWARE LICENSE
+ *
+ * This license is a legal agreement between you and NVIDIA Corporation ("NVIDIA") and governs your use of the NVIDIA/CUDA C++ Library software and materials provided hereunder (“SOFTWARE”).
+ *
+ * This license can be accepted only by an adult of legal age of majority in the country in which the SOFTWARE is used. If you are under the legal age of majority, you must ask your parent or legal guardian to consent to this license. By taking delivery of the SOFTWARE, you affirm that you have reached the legal age of majority, you accept the terms of this license, and you take legal and financial responsibility for the actions of your permitted users.
+ *
+ * You agree to use the SOFTWARE only for purposes that are permitted by (a) this license, and (b) any applicable law, regulation or generally accepted practices or guidelines in the relevant jurisdictions.
+ *
+ * 1. LICENSE. Subject to the terms of this license, NVIDIA grants you a non-exclusive limited license to: (a) install and use the SOFTWARE, and (b) distribute the SOFTWARE subject to the distribution requirements described in this license. NVIDIA reserves all rights, title and interest in and to the SOFTWARE not expressly granted to you under this license.
+ *
+ * 2. DISTRIBUTION REQUIREMENTS. These are the distribution requirements for you to exercise the distribution grant:
+ * a.      The terms under which you distribute the SOFTWARE must be consistent with the terms of this license, including (without limitation) terms relating to the license grant and license restrictions and protection of NVIDIA’s intellectual property rights.
+ * b.      You agree to notify NVIDIA in writing of any known or suspected distribution or use of the SOFTWARE not in compliance with the requirements of this license, and to enforce the terms of your agreements with respect to distributed SOFTWARE.
+ *
+ * 3. LIMITATIONS. Your license to use the SOFTWARE is restricted as follows:
+ * a.      The SOFTWARE is licensed for you to develop applications only for use in systems with NVIDIA GPUs.
+ * b.      You may not reverse engineer, decompile or disassemble, or remove copyright or other proprietary notices from any portion of the SOFTWARE or copies of the SOFTWARE.
+ * c.      You may not modify or create derivative works of any portion of the SOFTWARE.
+ * d.      You may not bypass, disable, or circumvent any technical measure, encryption, security, digital rights management or authentication mechanism in the SOFTWARE.
+ * e.      You may not use the SOFTWARE in any manner that would cause it to become subject to an open source software license. As examples, licenses that require as a condition of use, modification, and/or distribution that the SOFTWARE be (i) disclosed or distributed in source code form; (ii) licensed for the purpose of making derivative works; or (iii) redistributable at no charge.
+ * f.      Unless you have an agreement with NVIDIA for this purpose, you may not use the SOFTWARE with any system or application where the use or failure of the system or application can reasonably be expected to threaten or result in personal injury, death, or catastrophic loss. Examples include use in avionics, navigation, military, medical, life support or other life critical applications. NVIDIA does not design, test or manufacture the SOFTWARE for these critical uses and NVIDIA shall not be liable to you or any third party, in whole or in part, for any claims or damages arising from such uses.
+ * g.      You agree to defend, indemnify and hold harmless NVIDIA and its affiliates, and their respective employees, contractors, agents, officers and directors, from and against any and all claims, damages, obligations, losses, liabilities, costs or debt, fines, restitutions and expenses (including but not limited to attorney’s fees and costs incident to establishing the right of indemnification) arising out of or related to use of the SOFTWARE outside of the scope of this Agreement, or not in compliance with its terms.
+ *
+ * 4. PRE-RELEASE. SOFTWARE versions identified as alpha, beta, preview, early access or otherwise as pre-release may not be fully functional, may contain errors or design flaws, and may have reduced or different security, privacy, availability, and reliability standards relative to commercial versions of NVIDIA software and materials. You may use a pre-release SOFTWARE version at your own risk, understanding that these versions are not intended for use in production or business-critical systems.
+ *
+ * 5. OWNERSHIP. The SOFTWARE and the related intellectual property rights therein are and will remain the sole and exclusive property of NVIDIA or its licensors. The SOFTWARE is copyrighted and protected by the laws of the United States and other countries, and international treaty provisions. NVIDIA may make changes to the SOFTWARE, at any time without notice, but is not obligated to support or update the SOFTWARE.
+ *
+ * 6. COMPONENTS UNDER OTHER LICENSES. The SOFTWARE may include NVIDIA or third-party components with separate legal notices or terms as may be described in proprietary notices accompanying the SOFTWARE. If and to the extent there is a conflict between the terms in this license and the license terms associated with a component, the license terms associated with the components control only to the extent necessary to resolve the conflict.
+ *
+ * 7. FEEDBACK. You may, but don’t have to, provide to NVIDIA any Feedback. “Feedback” means any suggestions, bug fixes, enhancements, modifications, feature requests or other feedback regarding the SOFTWARE. For any Feedback that you voluntarily provide, you hereby grant NVIDIA and its affiliates a perpetual, non-exclusive, worldwide, irrevocable license to use, reproduce, modify, license, sublicense (through multiple tiers of sublicensees), and distribute (through multiple tiers of distributors) the Feedback without the payment of any royalties or fees to you. NVIDIA will use Feedback at its choice.
+ *
+ * 8. NO WARRANTIES. THE SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY EXPRESS OR IMPLIED WARRANTY OF ANY KIND INCLUDING, BUT NOT LIMITED TO, WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, OR FITNESS FOR A PARTICULAR PURPOSE. NVIDIA DOES NOT WARRANT THAT THE SOFTWARE WILL MEET YOUR REQUIREMENTS OR THAT THE OPERATION THEREOF WILL BE UNINTERRUPTED OR ERROR-FREE, OR THAT ALL ERRORS WILL BE CORRECTED.
+ *
+ * 9. LIMITATIONS OF LIABILITY. TO THE MAXIMUM EXTENT PERMITTED BY LAW, NVIDIA AND ITS AFFILIATES SHALL NOT BE LIABLE FOR ANY SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL DAMAGES, OR ANY LOST PROFITS, PROJECT DELAYS, LOSS OF USE, LOSS OF DATA OR LOSS OF GOODWILL, OR THE COSTS OF PROCURING SUBSTITUTE PRODUCTS, ARISING OUT OF OR IN CONNECTION WITH THIS LICENSE OR THE USE OR PERFORMANCE OF THE SOFTWARE, WHETHER SUCH LIABILITY ARISES FROM ANY CLAIM BASED UPON BREACH OF CONTRACT, BREACH OF WARRANTY, TORT (INCLUDING NEGLIGENCE), PRODUCT LIABILITY OR ANY OTHER CAUSE OF ACTION OR THEORY OF LIABILITY, EVEN IF NVIDIA HAS PREVIOUSLY BEEN ADVISED OF, OR COULD REASONABLY HAVE FORESEEN, THE POSSIBILITY OF SUCH DAMAGES. IN NO EVENT WILL NVIDIA’S AND ITS AFFILIATES TOTAL CUMULATIVE LIABILITY UNDER OR ARISING OUT OF THIS LICENSE EXCEED US$10.00. THE NATURE OF THE LIABILITY OR THE NUMBER OF CLAIMS OR SUITS SHALL NOT ENLARGE OR EXTEND THIS LIMIT.
+ *
+ * 10. TERMINATION. Your rights under this license will terminate automatically without notice from NVIDIA if you fail to comply with any term and condition of this license or if you commence or participate in any legal proceeding against NVIDIA with respect to the SOFTWARE. NVIDIA may terminate this license with advance written notice to you if NVIDIA decides to no longer provide the SOFTWARE in a country or, in NVIDIA’s sole discretion, the continued use of it is no longer commercially viable. Upon any termination of this license, you agree to promptly discontinue use of the SOFTWARE and destroy all copies in your possession or control. Your prior distributions in accordance with this license are not affected by the termination of this license. All provisions of this license will survive termination, except for the license granted to you.
+ *
+ * 11. APPLICABLE LAW. This license will be governed in all respects by the laws of the United States and of the State of Delaware as those laws are applied to contracts entered into and performed entirely within Delaware by Delaware residents, without regard to the conflicts of laws principles. The United Nations Convention on Contracts for the International Sale of Goods is specifically disclaimed. You agree to all terms of this Agreement in the English language. The state or federal courts residing in Santa Clara County, California shall have exclusive jurisdiction over any dispute or claim arising out of this license. Notwithstanding this, you agree that NVIDIA shall still be allowed to apply for injunctive remedies or an equivalent type of urgent legal relief in any jurisdiction.
+ *
+ * 12. NO ASSIGNMENT. This license and your rights and obligations thereunder may not be assigned by you by any means or operation of law without NVIDIA’s permission. Any attempted assignment not approved by NVIDIA in writing shall be void and of no effect.
+ *
+ * 13. EXPORT. The SOFTWARE is subject to United States export laws and regulations. You agree that you will not ship, transfer or export the SOFTWARE into any country, or use the SOFTWARE in any manner, prohibited by the United States Bureau of Industry and Security or economic sanctions regulations administered by the U.S. Department of Treasury’s Office of Foreign Assets Control (OFAC), or any applicable export laws, restrictions or regulations. These laws include restrictions on destinations, end users and end use. By accepting this license, you confirm that you are not a resident or citizen of any country currently embargoed by the U.S. and that you are not otherwise prohibited from receiving the SOFTWARE.
+ *
+ * 14. GOVERNMENT USE. The SOFTWARE has been developed entirely at private expense and is “commercial items” consisting of “commercial computer software” and “commercial computer software documentation” provided with RESTRICTED RIGHTS. Use, duplication or disclosure by the U.S. Government or a U.S. Government subcontractor is subject to the restrictions in this license pursuant to DFARS 227.7202-3(a) or as set forth in subparagraphs (b)(1) and (2) of the Commercial Computer Software - Restricted Rights clause at FAR 52.227-19, as applicable. Contractor/manufacturer is NVIDIA, 2788 San Tomas Expressway, Santa Clara, CA 95051.
+ *
+ * 15. ENTIRE AGREEMENT. This license is the final, complete and exclusive agreement between the parties relating to the subject matter of this license and supersedes all prior or contemporaneous understandings and agreements relating to this subject matter, whether oral or written. If any court of competent jurisdiction determines that any provision of this license is illegal, invalid or unenforceable, the remaining provisions will remain in full force and effect. This license may only be modified in a writing signed by an authorized representative of each party.
+ *
+ * (v. August 20, 2021)
+ */
+#ifndef _CUDA_PIPELINE
+#define _CUDA_PIPELINE
+
+#include <nv/target>
+#include <cuda/barrier>
+#include <cuda/atomic>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+    template<thread_scope _Scope>
+    class pipeline;
+
+    enum class pipeline_role {
+        producer,
+        consumer
+    };
+
+    template<thread_scope _Scope>
+    struct __pipeline_stage {
+        barrier<_Scope> __produced;
+        barrier<_Scope> __consumed;
+    };
+
+    template<thread_scope _Scope, uint8_t _Stages_count>
+    class pipeline_shared_state {
+    public:
+        pipeline_shared_state() = default;
+        pipeline_shared_state(const pipeline_shared_state &) = delete;
+        pipeline_shared_state(pipeline_shared_state &&) = delete;
+        pipeline_shared_state & operator=(pipeline_shared_state &&) = delete;
+        pipeline_shared_state & operator=(const pipeline_shared_state &) =  delete;
+
+    private:
+        __pipeline_stage<_Scope> __stages[_Stages_count];
+        atomic<uint32_t, _Scope> __refcount;
+
+        template<thread_scope _Pipeline_scope>
+        friend class pipeline;
+
+        template<class _Group, thread_scope _Pipeline_scope, uint8_t _Pipeline_stages_count>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline<_Pipeline_scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Pipeline_scope, _Pipeline_stages_count> * __shared_state);
+
+        template<class _Group, thread_scope _Pipeline_scope, uint8_t _Pipeline_stages_count>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline<_Pipeline_scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Pipeline_scope, _Pipeline_stages_count> * __shared_state, size_t __producer_count);
+     
+        template<class _Group, thread_scope _Pipeline_scope, uint8_t _Pipeline_stages_count>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline<_Pipeline_scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Pipeline_scope, _Pipeline_stages_count> * __shared_state, pipeline_role __role);
+    };
+
+    struct __pipeline_asm_helper {
+        __device__
+        static inline uint32_t __lane_id()
+        {
+            uint32_t __lane_id;
+            asm volatile ("mov.u32 %0, %%laneid;" : "=r"(__lane_id));
+            return __lane_id;
+        }
+    };
+
+    template<thread_scope _Scope>
+    class pipeline {
+    public:
+        pipeline(pipeline &&) = default;
+        pipeline(const pipeline &) = delete;
+        pipeline & operator=(pipeline &&) = delete;
+        pipeline & operator=(const pipeline &) = delete;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        ~pipeline()
+        {
+            if (__active) {
+                (void)quit();
+            }
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool quit()
+        {
+            bool __elected;
+            uint32_t __sub_count;
+NV_IF_TARGET(NV_IS_DEVICE,
+            const uint32_t __match_mask = __match_any_sync(__activemask(), reinterpret_cast<uintptr_t>(__shared_state_get_refcount()));
+            const uint32_t __elected_id = __ffs(__match_mask) - 1;
+            __elected = (__pipeline_asm_helper::__lane_id() == __elected_id);
+            __sub_count = __popc(__match_mask);
+,
+            __elected = true;
+            __sub_count = 1;
+)
+            bool __released = false;
+            if (__elected) {
+                const uint32_t __old = __shared_state_get_refcount()->fetch_sub(__sub_count);
+                const bool __last = (__old == __sub_count);
+                if (__last) {
+                    for (uint8_t __stage = 0; __stage < __stages_count; ++__stage) {
+                        __shared_state_get_stage(__stage)->__produced.~barrier();
+                        if (__partitioned) {
+                            __shared_state_get_stage(__stage)->__consumed.~barrier();
+                        }
+                    }
+                    __released = true;
+                }
+            }
+            __active = false;
+            return __released;
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void producer_acquire()
+        {
+            if (__partitioned) {
+                barrier<_Scope> & __stage_barrier = __shared_state_get_stage(__head)->__consumed;
+                (void)_CUDA_VSTD::__libcpp_thread_poll_with_backoff(__poll_tester(__stage_barrier, __consumed_phase_parity));
+            }
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void producer_commit()
+        {
+            barrier<_Scope> & __stage_barrier = __shared_state_get_stage(__head)->__produced;
+            __memcpy_async_synchronize(__stage_barrier, true);
+            (void)__stage_barrier.arrive();
+            if (++__head == __stages_count) {
+                __head = 0;
+                if (__partitioned) {
+                    __consumed_phase_parity = !__consumed_phase_parity;
+                }
+            }
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void consumer_wait()
+        {
+            barrier<_Scope> & __stage_barrier = __shared_state_get_stage(__tail)->__produced;
+            (void)_CUDA_VSTD::__libcpp_thread_poll_with_backoff(__poll_tester(__stage_barrier, __produced_phase_parity));
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void consumer_release()
+        {
+            if (__partitioned) {
+                (void)__shared_state_get_stage(__tail)->__consumed.arrive();
+            }
+            if (++__tail == __stages_count) {
+                __tail = 0;
+                __produced_phase_parity = !__produced_phase_parity;
+            }
+        }
+
+        template<class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool consumer_wait_for(const _CUDA_VSTD::chrono::duration<_Rep, _Period> & __duration)
+        {
+            barrier<_Scope> & __stage_barrier = __shared_state_get_stage(__tail)->__produced;
+            return _CUDA_VSTD::__libcpp_thread_poll_with_backoff(
+                __poll_tester(__stage_barrier, __produced_phase_parity),
+                _CUDA_VSTD::chrono::duration_cast<_CUDA_VSTD::chrono::nanoseconds>(__duration)
+            );
+        }
+
+        template<class _Clock, class _Duration> 
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool consumer_wait_until(const _CUDA_VSTD::chrono::time_point<_Clock, _Duration> & __time_point)
+        {
+            return consumer_wait_for(__time_point - _Clock::now());
+        }
+
+    private:
+        uint8_t __head               : 8;
+        uint8_t __tail               : 8;
+        const uint8_t __stages_count : 8;
+        bool __consumed_phase_parity : 1;
+        bool __produced_phase_parity : 1;
+        bool __active                : 1;
+        const bool __partitioned     : 1;
+        char * const __shared_state;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline(char * __shared_state, uint8_t __stages_count, bool __partitioned)
+            : __head(0)
+            , __tail(0)
+            , __stages_count(__stages_count)
+            , __consumed_phase_parity(true)
+            , __produced_phase_parity(false)
+            , __active(true)
+            , __partitioned(__partitioned)
+            , __shared_state(__shared_state)
+        {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static bool __barrier_try_wait_parity_impl(barrier<_Scope> & __barrier, bool __phase_parity)
+        {
+            typename barrier<_Scope>::arrival_token __synthesized_token = (__phase_parity ? 1ull : 0ull) << 63;
+            return __barrier.__try_wait(_CUDA_VSTD::move(__synthesized_token));
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static bool __barrier_try_wait_parity(barrier<_Scope> & __barrier, bool __phase_parity)
+        {
+            return __barrier_try_wait_parity_impl(__barrier, __phase_parity);
+        }
+
+        struct __poll_tester {
+            barrier<_Scope> & __barrier;
+            bool __phase_parity;
+
+            _LIBCUDACXX_INLINE_VISIBILITY
+            __poll_tester(barrier<_Scope> & __barrier, bool __phase_parity)
+              : __barrier(__barrier)
+              , __phase_parity(__phase_parity)
+            {}
+
+            _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator()() const
+            {
+                return __barrier_try_wait_parity(__barrier, __phase_parity);
+            }
+        };
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __pipeline_stage<_Scope> * __shared_state_get_stage(uint8_t __stage)
+        {
+            ptrdiff_t __stage_offset = __stage * sizeof(__pipeline_stage<_Scope>);
+            return reinterpret_cast<__pipeline_stage<_Scope>*>(__shared_state + __stage_offset);
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        atomic<uint32_t, _Scope> * __shared_state_get_refcount()
+        {
+            ptrdiff_t __refcount_offset = __stages_count * sizeof(__pipeline_stage<_Scope>);
+            return reinterpret_cast<atomic<uint32_t, _Scope>*>(__shared_state + __refcount_offset);
+        }
+
+        template<class _Group, thread_scope _Pipeline_scope, uint8_t _Pipeline_stages_count>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline<_Pipeline_scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Pipeline_scope, _Pipeline_stages_count> * __shared_state);
+
+        template<class _Group, thread_scope _Pipeline_scope, uint8_t _Pipeline_stages_count>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline<_Pipeline_scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Pipeline_scope, _Pipeline_stages_count> * __shared_state, size_t __producer_count);
+     
+        template<class _Group, thread_scope _Pipeline_scope, uint8_t _Pipeline_stages_count>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline<_Pipeline_scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Pipeline_scope, _Pipeline_stages_count> * __shared_state, pipeline_role __role);
+    };
+
+    template<>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline bool pipeline<thread_scope_block>::__barrier_try_wait_parity(barrier<thread_scope_block> & __barrier, bool __phase_parity)
+    {
+NV_IF_TARGET(NV_PROVIDES_SM_80,
+        if (__isShared(&__barrier)) {
+            uint64_t * __mbarrier = device::barrier_native_handle(__barrier);
+            uint16_t __wait_complete;
+
+            asm volatile ("{"
+                          "     .reg .pred %p;"
+                          "     mbarrier.test_wait.parity.shared.b64 %p, [%1], %2;"
+                          "     selp.u16 %0, 1, 0, %p;"
+                          "}"
+                : "=h"(__wait_complete)
+                : "r"(static_cast<uint32_t>(__cvta_generic_to_shared(__mbarrier))), "r"(static_cast<uint32_t>(__phase_parity))
+                : "memory");
+
+            return bool(__wait_complete);
+        } else
+        {
+            return __barrier_try_wait_parity_impl(__barrier, __phase_parity);
+        }
+,
+        return __barrier_try_wait_parity_impl(__barrier, __phase_parity);
+)
+    }
+
+    template<class _Group, thread_scope _Scope, uint8_t _Stages_count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pipeline<_Scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Scope, _Stages_count> * __shared_state)
+    {
+        const uint32_t __group_size = static_cast<uint32_t>(__group.size());
+        const uint32_t __thread_rank = static_cast<uint32_t>(__group.thread_rank());
+
+        if (__thread_rank == 0) {
+            for (uint8_t __stage = 0; __stage < _Stages_count; ++__stage) {
+                init(&__shared_state->__stages[__stage].__produced, __group_size);
+            }
+            __shared_state->__refcount.store(__group_size, std::memory_order_relaxed);
+        }
+        __group.sync();
+
+        return pipeline<_Scope>(reinterpret_cast<char*>(__shared_state->__stages), _Stages_count, false);
+    }
+
+    template<class _Group, thread_scope _Scope, uint8_t _Stages_count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pipeline<_Scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Scope, _Stages_count> * __shared_state, size_t __producer_count)
+    {
+        const uint32_t __group_size = static_cast<uint32_t>(__group.size());
+        const uint32_t __thread_rank = static_cast<uint32_t>(__group.thread_rank());
+
+        if (__thread_rank == 0) {
+            const size_t __consumer_count = __group_size - __producer_count;
+            for (uint8_t __stage = 0; __stage < _Stages_count; ++__stage) {
+                init(&__shared_state->__stages[__stage].__consumed, __consumer_count);
+                init(&__shared_state->__stages[__stage].__produced, __producer_count);
+            }
+            __shared_state->__refcount.store(__group_size, std::memory_order_relaxed);
+        }
+        __group.sync();
+
+        return pipeline<_Scope>(reinterpret_cast<char*>(__shared_state->__stages), _Stages_count, true);
+    }
+ 
+    template<class _Group, thread_scope _Scope, uint8_t _Stages_count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pipeline<_Scope> make_pipeline(const _Group & __group, pipeline_shared_state<_Scope, _Stages_count> * __shared_state, pipeline_role __role)
+    {
+        const uint32_t __group_size = static_cast<uint32_t>(__group.size());
+        const uint32_t __thread_rank = static_cast<uint32_t>(__group.thread_rank());
+
+        if (__thread_rank == 0) {
+            __shared_state->__refcount.store(0, std::memory_order_relaxed);
+        }
+        __group.sync();
+
+        if (__role == pipeline_role::producer) {
+            bool __elected;
+            uint32_t __add_count;
+NV_IF_TARGET(NV_IS_DEVICE,
+            const uint32_t __match_mask = __match_any_sync(__activemask(), reinterpret_cast<uintptr_t>(&__shared_state->__refcount));
+            const uint32_t __elected_id = __ffs(__match_mask) - 1;
+            __elected = (__pipeline_asm_helper::__lane_id() == __elected_id);
+            __add_count = __popc(__match_mask);
+,
+            __elected = true;
+            __add_count = 1;
+)
+            if (__elected) {
+                (void)__shared_state->__refcount.fetch_add(__add_count, std::memory_order_relaxed);
+            }
+        }
+        __group.sync();
+
+        if (__thread_rank == 0) {
+            const uint32_t __producer_count = __shared_state->__refcount.load(std::memory_order_relaxed);
+            const uint32_t __consumer_count = __group_size - __producer_count;
+            for (uint8_t __stage = 0; __stage < _Stages_count; ++__stage) {
+                init(&__shared_state->__stages[__stage].__consumed, __consumer_count);
+                init(&__shared_state->__stages[__stage].__produced, __producer_count);
+            }
+            __shared_state->__refcount.store(__group_size, std::memory_order_relaxed);
+        }
+        __group.sync();
+
+        return pipeline<_Scope>(reinterpret_cast<char*>(__shared_state->__stages), _Stages_count, true);
+    }
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA_DEVICE
+
+    template<uint8_t _Prior>
+    __device__
+    void __pipeline_consumer_wait(pipeline<thread_scope_thread> & __pipeline);
+
+    __device__
+    inline void __pipeline_consumer_wait(pipeline<thread_scope_thread> & __pipeline, uint8_t __prior);
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_DEVICE
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+    template<>
+    class pipeline<thread_scope_thread> {
+    public:
+        pipeline(pipeline &&) = default;
+        pipeline(const pipeline &) = delete;
+        pipeline & operator=(pipeline &&) = delete;
+        pipeline & operator=(const pipeline &) = delete;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        ~pipeline() {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool quit()
+        {
+            return true;
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void producer_acquire() {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void producer_commit()
+        {
+NV_IF_TARGET(NV_PROVIDES_SM_80,
+            asm volatile ("cp.async.commit_group;");
+            ++__head;
+)
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void consumer_wait()
+        {
+NV_IF_TARGET(NV_PROVIDES_SM_80,
+            if (__head == __tail) {
+                return;
+            }
+
+            const uint8_t __prior = __head - __tail - 1;
+            device::__pipeline_consumer_wait(*this, __prior);
+            ++__tail;
+)
+        }
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void consumer_release() {}
+    
+        template<class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool consumer_wait_for(const _CUDA_VSTD::chrono::duration<_Rep, _Period> & __duration)
+        {
+            (void)__duration;
+            consumer_wait();
+            return true;
+        }
+
+        template<class _Clock, class _Duration> 
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool consumer_wait_until(const _CUDA_VSTD::chrono::time_point<_Clock, _Duration> & __time_point)
+        {
+            (void)__time_point;
+            consumer_wait();
+            return true;
+        }
+
+    private:
+        uint8_t __head;
+        uint8_t __tail;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pipeline()
+            : __head(0)
+            , __tail(0)
+        {}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY inline pipeline<thread_scope_thread> make_pipeline();
+
+        template<uint8_t _Prior>
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        void pipeline_consumer_wait_prior(pipeline<thread_scope_thread> & __pipeline);
+    };
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA_DEVICE
+
+    template<uint8_t _Prior>
+    __device__
+    void __pipeline_consumer_wait(pipeline<thread_scope_thread> & __pipeline)
+    {
+        (void)__pipeline;
+NV_IF_TARGET(NV_PROVIDES_SM_80,
+        constexpr uint8_t __max_prior = 8;
+
+        asm volatile ("cp.async.wait_group %0;"
+            :
+            : "n"(_Prior < __max_prior ? _Prior : __max_prior));
+)
+    }
+
+    __device__
+    inline void __pipeline_consumer_wait(pipeline<thread_scope_thread> & __pipeline, uint8_t __prior)
+    {
+        switch (__prior) {
+        case 0:  device::__pipeline_consumer_wait<0>(__pipeline); break;
+        case 1:  device::__pipeline_consumer_wait<1>(__pipeline); break;
+        case 2:  device::__pipeline_consumer_wait<2>(__pipeline); break;
+        case 3:  device::__pipeline_consumer_wait<3>(__pipeline); break;
+        case 4:  device::__pipeline_consumer_wait<4>(__pipeline); break;
+        case 5:  device::__pipeline_consumer_wait<5>(__pipeline); break;
+        case 6:  device::__pipeline_consumer_wait<6>(__pipeline); break;
+        case 7:  device::__pipeline_consumer_wait<7>(__pipeline); break;
+        default: device::__pipeline_consumer_wait<8>(__pipeline); break;
+        }
+    }
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_DEVICE
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline pipeline<thread_scope_thread> make_pipeline()
+    {
+        return pipeline<thread_scope_thread>();
+    }
+
+    template<uint8_t _Prior>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pipeline_consumer_wait_prior(pipeline<thread_scope_thread> & __pipeline)
+    {
+NV_IF_TARGET(NV_PROVIDES_SM_80,
+        device::__pipeline_consumer_wait<_Prior>(__pipeline);
+        __pipeline.__tail = __pipeline.__head - _Prior;
+)
+    }
+
+    template<thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pipeline_producer_commit(pipeline<thread_scope_thread> & __pipeline, barrier<_Scope> & __barrier)
+    {
+        (void)__pipeline;
+NV_IF_TARGET(NV_PROVIDES_SM_80,
+        __memcpy_async_synchronize(__barrier, true);
+)
+    }
+
+    template<thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __memcpy_async_synchronize(pipeline<_Scope> & __pipeline, bool __is_async) {
+        // memcpy_async submissions never synchronize on their own in the pipeline path.
+        (void)__pipeline;
+        (void)__is_async;
+    }
+
+    template<typename _Group, class _Type, thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void memcpy_async(_Group const & __group, _Type * __destination, _Type const * __source, std::size_t __size, pipeline<_Scope> & __pipeline)
+    {
+        // When compiling with NVCC and GCC 4.8, certain user defined types that _are_ trivially copyable are
+        // incorrectly classified as not trivially copyable. Remove this assertion to allow for their usage with
+        // memcpy_async when compiling with GCC 4.8.
+        // FIXME: remove the #if once GCC 4.8 is no longer supported.
+    #if !defined(_LIBCUDACXX_COMPILER_GCC) || _GNUC_VER > 408
+        static_assert(_CUDA_VSTD::is_trivially_copyable<_Type>::value, "memcpy_async requires a trivially copyable type");
+    #endif
+
+        __memcpy_async<alignof(_Type)>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __pipeline);
+    }
+
+    template<typename _Group, class _Type, std::size_t _Alignment, thread_scope _Scope, std::size_t _Larger_alignment = (alignof(_Type) > _Alignment) ? alignof(_Type) : _Alignment>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void memcpy_async(_Group const & __group, _Type * __destination, _Type const * __source, aligned_size_t<_Alignment> __size, pipeline<_Scope> & __pipeline) {
+        // When compiling with NVCC and GCC 4.8, certain user defined types that _are_ trivially copyable are
+        // incorrectly classified as not trivially copyable. Remove this assertion to allow for their usage with
+        // memcpy_async when compiling with GCC 4.8.
+        // FIXME: remove the #if once GCC 4.8 is no longer supported.
+    #if !defined(_LIBCUDACXX_COMPILER_GCC) || _GNUC_VER > 408
+        static_assert(_CUDA_VSTD::is_trivially_copyable<_Type>::value, "memcpy_async requires a trivially copyable type");
+    #endif
+
+        __memcpy_async<_Larger_alignment>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __pipeline);
+    }
+
+    template<class _Type, typename _Size, thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void memcpy_async(_Type * __destination, _Type const * __source, _Size __size, pipeline<_Scope> & __pipeline) {
+        memcpy_async(__single_thread_group{}, __destination, __source, __size, __pipeline);
+    }
+
+    template<typename _Group, thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void memcpy_async(_Group const & __group, void * __destination, void const * __source, std::size_t __size, pipeline<_Scope> & __pipeline) {
+        __memcpy_async<1>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __pipeline);
+    }
+
+    template<typename _Group, std::size_t _Alignment, thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void memcpy_async(_Group const & __group, void * __destination, void const * __source, aligned_size_t<_Alignment> __size, pipeline<_Scope> & __pipeline) {
+        __memcpy_async<_Alignment>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __pipeline);
+    }
+
+    template<typename _Size, thread_scope _Scope>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void memcpy_async(void * __destination, void const * __source, _Size __size, pipeline<_Scope> & __pipeline) {
+        memcpy_async(__single_thread_group{}, __destination, __source, __size, __pipeline);
+    }
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#endif //_CUDA_PIPELINE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/semaphore b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/semaphore
new file mode 100644
index 000000000000..eb76c73537f1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/semaphore
@@ -0,0 +1,10 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "std/semaphore"
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/array b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/array
new file mode 100644
index 000000000000..64240539bb47
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/array
@@ -0,0 +1,34 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_ARRAY
+#define _CUDA_ARRAY
+
+#include "cassert"
+#include "cstdint"
+#include "limits"
+#include "type_traits"
+#include "iterator"
+#include "utility"
+#include "initializer_list"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/algorithm"
+#include "detail/libcxx/include/__tuple"
+#include "detail/libcxx/include/cstdlib"
+#include "detail/libcxx/include/stdexcept"
+#include "detail/libcxx/include/array"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_ARRAY
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/atomic b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/atomic
new file mode 100644
index 000000000000..65315f30d515
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/atomic
@@ -0,0 +1,296 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_ATOMIC
+#define _CUDA_ATOMIC
+
+#ifndef __CUDACC_RTC__
+    #include <atomic>
+    static_assert(ATOMIC_BOOL_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_CHAR_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_CHAR16_T_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_CHAR32_T_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_WCHAR_T_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_SHORT_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_INT_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_LONG_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_LLONG_LOCK_FREE == 2, "");
+    static_assert(ATOMIC_POINTER_LOCK_FREE == 2, "");
+    #undef ATOMIC_BOOL_LOCK_FREE
+    #undef ATOMIC_BOOL_LOCK_FREE
+    #undef ATOMIC_CHAR_LOCK_FREE
+    #undef ATOMIC_CHAR16_T_LOCK_FREE
+    #undef ATOMIC_CHAR32_T_LOCK_FREE
+    #undef ATOMIC_WCHAR_T_LOCK_FREE
+    #undef ATOMIC_SHORT_LOCK_FREE
+    #undef ATOMIC_INT_LOCK_FREE
+    #undef ATOMIC_LONG_LOCK_FREE
+    #undef ATOMIC_LLONG_LOCK_FREE
+    #undef ATOMIC_POINTER_LOCK_FREE
+    #undef ATOMIC_FLAG_INIT
+    #undef ATOMIC_VAR_INIT
+#endif //__CUDACC_RTC__
+
+// pre-define lock free query for heterogeneous compatibility
+#ifndef _LIBCUDACXX_ATOMIC_IS_LOCK_FREE
+#define _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(__x) (__x <= 8)
+#endif
+
+#include "cassert"
+#include "cstddef"
+#include "cstdint"
+#include "type_traits"
+#include "version"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/__threading_support"
+
+#include "detail/libcxx/include/atomic"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+using std::__detail::thread_scope;
+using std::__detail::thread_scope_system;
+using std::__detail::thread_scope_device;
+using std::__detail::thread_scope_block;
+using std::__detail::thread_scope_thread;
+
+namespace __detail {
+using std::__detail::__thread_scope_block_tag;
+using std::__detail::__thread_scope_device_tag;
+using std::__detail::__thread_scope_system_tag;
+}
+
+using memory_order = std::memory_order;
+
+constexpr memory_order memory_order_relaxed = std::memory_order_relaxed;
+constexpr memory_order memory_order_consume = std::memory_order_consume;
+constexpr memory_order memory_order_acquire = std::memory_order_acquire;
+constexpr memory_order memory_order_release = std::memory_order_release;
+constexpr memory_order memory_order_acq_rel = std::memory_order_acq_rel;
+constexpr memory_order memory_order_seq_cst = std::memory_order_seq_cst;
+
+// atomic<T>
+
+template <class _Tp, thread_scope _Sco = thread_scope::thread_scope_system>
+struct atomic
+    : public std::__atomic_base<_Tp, _Sco>
+{
+    typedef std::__atomic_base<_Tp, _Sco> __base;
+
+    constexpr atomic() noexcept = default;
+    __host__ __device__
+    constexpr atomic(_Tp __d) noexcept : __base(__d) {}
+
+    __host__ __device__
+    _Tp operator=(_Tp __d) volatile noexcept
+        {__base::store(__d); return __d;}
+    __host__ __device__
+    _Tp operator=(_Tp __d) noexcept
+        {__base::store(__d); return __d;}
+
+    __host__ __device__
+    _Tp fetch_max(const _Tp & __op, memory_order __m = memory_order_seq_cst) volatile noexcept
+    {
+        return std::__detail::__cxx_atomic_fetch_max(&this->__a_, __op, __m);
+    }
+
+    __host__ __device__
+    _Tp fetch_min(const _Tp & __op, memory_order __m = memory_order_seq_cst) volatile noexcept
+    {
+        return std::__detail::__cxx_atomic_fetch_min(&this->__a_, __op, __m);
+    }
+};
+
+// atomic<T*>
+
+template <class _Tp, thread_scope _Sco>
+struct atomic<_Tp*, _Sco>
+    : public std::__atomic_base<_Tp*, _Sco>
+{
+    typedef std::__atomic_base<_Tp*, _Sco> __base;
+
+    constexpr atomic() noexcept = default;
+    __host__ __device__
+    constexpr atomic(_Tp* __d) noexcept : __base(__d) {}
+
+    __host__ __device__
+    _Tp* operator=(_Tp* __d) volatile noexcept
+        {__base::store(__d); return __d;}
+    __host__ __device__
+    _Tp* operator=(_Tp* __d) noexcept
+        {__base::store(__d); return __d;}
+
+    __host__ __device__
+    _Tp* fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        volatile noexcept
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    __host__ __device__
+    _Tp* fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst) noexcept
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    __host__ __device__
+    _Tp* fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        volatile noexcept
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    __host__ __device__
+    _Tp* fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst) noexcept
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+
+    __host__ __device__
+    _Tp* operator++(int) volatile noexcept            {return fetch_add(1);}
+    __host__ __device__
+    _Tp* operator++(int) noexcept                     {return fetch_add(1);}
+    __host__ __device__
+    _Tp* operator--(int) volatile noexcept            {return fetch_sub(1);}
+    __host__ __device__
+    _Tp* operator--(int) noexcept                     {return fetch_sub(1);}
+    __host__ __device__
+    _Tp* operator++() volatile noexcept               {return fetch_add(1) + 1;}
+    __host__ __device__
+    _Tp* operator++() noexcept                        {return fetch_add(1) + 1;}
+    __host__ __device__
+    _Tp* operator--() volatile noexcept               {return fetch_sub(1) - 1;}
+    __host__ __device__
+    _Tp* operator--() noexcept                        {return fetch_sub(1) - 1;}
+    __host__ __device__
+    _Tp* operator+=(ptrdiff_t __op) volatile noexcept {return fetch_add(__op) + __op;}
+    __host__ __device__
+    _Tp* operator+=(ptrdiff_t __op) noexcept          {return fetch_add(__op) + __op;}
+    __host__ __device__
+    _Tp* operator-=(ptrdiff_t __op) volatile noexcept {return fetch_sub(__op) - __op;}
+    __host__ __device__
+    _Tp* operator-=(ptrdiff_t __op) noexcept          {return fetch_sub(__op) - __op;}
+};
+
+// atomic_ref<T>
+
+template <class _Tp, thread_scope _Sco = thread_scope::thread_scope_system>
+struct atomic_ref
+    : public std::__atomic_base_ref<_Tp, _Sco>
+{
+    typedef std::__atomic_base_ref<_Tp, _Sco> __base;
+
+    __host__ __device__
+    constexpr atomic_ref(_Tp& __d) noexcept : __base(__d) {}
+
+    __host__ __device__
+    _Tp operator=(_Tp __d) const volatile noexcept
+        {__base::store(__d); return __d;}
+    __host__ __device__
+    _Tp operator=(_Tp __d) const noexcept
+        {__base::store(__d); return __d;}
+
+    __host__ __device__
+    _Tp fetch_max(const _Tp & __op, memory_order __m = memory_order_seq_cst) const volatile noexcept
+    {
+        return std::__detail::__cxx_atomic_fetch_max(&this->__a_, __op, __m);
+    }
+
+    __host__ __device__
+    _Tp fetch_min(const _Tp & __op, memory_order __m = memory_order_seq_cst) const volatile noexcept
+    {
+        return std::__detail::__cxx_atomic_fetch_min(&this->__a_, __op, __m);
+    }
+};
+
+// atomic_ref<T*>
+
+template <class _Tp, thread_scope _Sco>
+struct atomic_ref<_Tp*, _Sco>
+    : public std::__atomic_base_ref<_Tp*, _Sco>
+{
+    typedef std::__atomic_base_ref<_Tp*, _Sco> __base;
+
+    __host__ __device__
+    constexpr atomic_ref(_Tp*& __d) noexcept : __base(__d) {}
+
+    __host__ __device__
+    _Tp* operator=(_Tp* __d) const volatile noexcept
+        {__base::store(__d); return __d;}
+    __host__ __device__
+    _Tp* operator=(_Tp* __d) const noexcept
+        {__base::store(__d); return __d;}
+
+    __host__ __device__
+    _Tp* fetch_add(ptrdiff_t __op,
+                   memory_order __m = memory_order_seq_cst) const volatile noexcept
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    __host__ __device__
+    _Tp* fetch_add(ptrdiff_t __op,
+                   memory_order __m = memory_order_seq_cst) const noexcept
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    __host__ __device__
+    _Tp* fetch_sub(ptrdiff_t __op,
+                   memory_order __m = memory_order_seq_cst) const volatile noexcept
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    __host__ __device__
+    _Tp* fetch_sub(ptrdiff_t __op,
+                   memory_order __m = memory_order_seq_cst) const noexcept
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+
+    __host__ __device__
+    _Tp* operator++(int) const volatile noexcept            {return fetch_add(1);}
+    __host__ __device__
+    _Tp* operator++(int) const noexcept                     {return fetch_add(1);}
+    __host__ __device__
+    _Tp* operator--(int) const volatile noexcept            {return fetch_sub(1);}
+    __host__ __device__
+    _Tp* operator--(int) const noexcept                     {return fetch_sub(1);}
+    __host__ __device__
+    _Tp* operator++() const volatile noexcept               {return fetch_add(1) + 1;}
+    __host__ __device__
+    _Tp* operator++() const noexcept                        {return fetch_add(1) + 1;}
+    __host__ __device__
+    _Tp* operator--() const volatile noexcept               {return fetch_sub(1) - 1;}
+    __host__ __device__
+    _Tp* operator--() const noexcept                        {return fetch_sub(1) - 1;}
+    __host__ __device__
+    _Tp* operator+=(ptrdiff_t __op) const volatile noexcept {return fetch_add(__op) + __op;}
+    __host__ __device__
+    _Tp* operator+=(ptrdiff_t __op) const noexcept          {return fetch_add(__op) + __op;}
+    __host__ __device__
+    _Tp* operator-=(ptrdiff_t __op) const volatile noexcept {return fetch_sub(__op) - __op;}
+    __host__ __device__
+    _Tp* operator-=(ptrdiff_t __op) const noexcept          {return fetch_sub(__op) - __op;}
+};
+
+inline __host__ __device__ void atomic_thread_fence(memory_order __m, thread_scope _Scope = thread_scope::thread_scope_system) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            switch(_Scope) {
+            case thread_scope::thread_scope_system:
+                std::__detail::__atomic_thread_fence_cuda((int)__m, __detail::__thread_scope_system_tag());
+                break;
+            case thread_scope::thread_scope_device:
+                std::__detail::__atomic_thread_fence_cuda((int)__m, __detail::__thread_scope_device_tag());
+                break;
+            case thread_scope::thread_scope_block:
+                std::__detail::__atomic_thread_fence_cuda((int)__m, __detail::__thread_scope_block_tag());
+                break;
+            }
+        ),
+        NV_IS_HOST, (
+            (void) _Scope;
+            std::atomic_thread_fence(__m);
+        )
+    )
+}
+
+inline __host__ __device__ void atomic_signal_fence(memory_order __m) {
+    std::atomic_signal_fence(__m);
+}
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_ATOMIC
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/barrier b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/barrier
new file mode 100644
index 000000000000..69a8ecbbfbdb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/barrier
@@ -0,0 +1,468 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 700
+#  error "CUDA synchronization primitives are only supported for sm_70 and up."
+#endif
+
+#ifndef _CUDA_BARRIER
+#define _CUDA_BARRIER
+
+#include "atomic"
+#include "cstddef"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/barrier"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+template<std::size_t _Alignment>
+struct aligned_size_t {
+    static constexpr std::size_t align = _Alignment;
+    std::size_t value;
+    __host__ __device__
+    explicit aligned_size_t(size_t __s) : value(__s) { }
+    __host__ __device__
+    operator size_t() const { return value; }
+};
+
+template<thread_scope _Sco, class _CompletionF = std::__empty_completion>
+class barrier : public std::__barrier_base<_CompletionF, _Sco> {
+    template<thread_scope>
+    friend class pipeline;
+
+    using std::__barrier_base<_CompletionF, _Sco>::__try_wait;
+
+public:
+    barrier() = default;
+
+    barrier(const barrier &) = delete;
+    barrier & operator=(const barrier &) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    barrier(std::ptrdiff_t __expected, _CompletionF __completion = _CompletionF())
+        : std::__barrier_base<_CompletionF, _Sco>(__expected, __completion) {
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend void init(barrier * __b, std::ptrdiff_t __expected) {
+        new (__b) barrier(__expected);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend void init(barrier * __b, std::ptrdiff_t __expected, _CompletionF __completion) {
+        new (__b) barrier(__expected, __completion);
+    }
+};
+
+struct __block_scope_barrier_base {};
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA_DEVICE
+
+__device__
+inline std::uint64_t * barrier_native_handle(barrier<thread_scope_block> & b);
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_DEVICE
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+template<>
+class barrier<thread_scope_block, std::__empty_completion> : public __block_scope_barrier_base {
+    using __barrier_base = std::__barrier_base<std::__empty_completion, (int)thread_scope_block>;
+    __barrier_base __barrier;
+
+    __device__
+    friend inline std::uint64_t * device::_LIBCUDACXX_CUDA_ABI_NAMESPACE::barrier_native_handle(barrier<thread_scope_block> & b);
+
+public:
+    using arrival_token = typename __barrier_base::arrival_token;
+
+private:
+    struct __poll_tester {
+        barrier const* __this;
+        arrival_token __phase;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __poll_tester(barrier const* __this_, arrival_token&& __phase_)
+          : __this(__this_)
+          , __phase(_CUDA_VSTD::move(__phase_))
+        {}
+
+        inline _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator()() const
+        {
+            return __this->__try_wait(__phase);
+        }
+    };
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __try_wait(arrival_token __phase) const {
+#if __CUDA_ARCH__ >= 800
+        if (__isShared(&__barrier)) {
+            int __ready = 0;
+            asm volatile ("{\n\t"
+                    ".reg .pred p;\n\t"
+                    "mbarrier.test_wait.shared.b64 p, [%1], %2;\n\t"
+                    "selp.b32 %0, 1, 0, p;\n\t"
+                    "}"
+                : "=r"(__ready)
+                : "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__barrier))), "l"(__phase)
+                : "memory");
+            return __ready;
+        }
+        else
+#endif
+        {
+            return __barrier.__try_wait(std::move(__phase));
+        }
+    }
+
+    template<thread_scope>
+    friend class pipeline;
+
+public:
+    barrier() = default;
+
+    barrier(const barrier &) = delete;
+    barrier & operator=(const barrier &) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    barrier(std::ptrdiff_t __expected, std::__empty_completion __completion = std::__empty_completion()) {
+        static_assert(_LIBCUDACXX_OFFSET_IS_ZERO(barrier<thread_scope_block>, __barrier), "fatal error: bad barrier layout");
+        init(this, __expected, __completion);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~barrier() {
+#if __CUDA_ARCH__ >= 800
+        if (__isShared(&__barrier)) {
+            asm volatile ("mbarrier.inval.shared.b64 [%0];"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__barrier)))
+                : "memory");
+        }
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend void init(barrier * __b, std::ptrdiff_t __expected, std::__empty_completion __completion = std::__empty_completion()) {
+#if __CUDA_ARCH__ >= 800
+        if (__isShared(&__b->__barrier)) {
+            asm volatile ("mbarrier.init.shared.b64 [%0], %1;"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__b->__barrier))),
+                    "r"(static_cast<std::uint32_t>(__expected))
+                : "memory");
+        }
+        else
+#endif
+        {
+            new (&__b->__barrier) __barrier_base(__expected);
+        }
+    }
+
+    _LIBCUDACXX_NODISCARD_ATTRIBUTE _LIBCUDACXX_INLINE_VISIBILITY
+    arrival_token arrive(std::ptrdiff_t __update = 1)
+    {
+#if __CUDA_ARCH__
+        if (__isShared(&__barrier)) {
+            arrival_token __token;
+#if __CUDA_ARCH__ >= 800
+            if (__update > 1) {
+                asm volatile ("mbarrier.arrive.noComplete.shared.b64 %0, [%1], %2;"
+                    : "=l"(__token)
+                    : "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__barrier))),
+                        "r"(static_cast<std::uint32_t>(__update - 1))
+                    : "memory");
+            }
+            asm volatile ("mbarrier.arrive.shared.b64 %0, [%1];"
+                : "=l"(__token)
+                : "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__barrier)))
+                : "memory");
+#else
+            unsigned int __activeA = __match_any_sync(__activemask(), __update);
+            unsigned int __activeB = __match_any_sync(__activemask(), reinterpret_cast<std::uintptr_t>(&__barrier));
+            unsigned int __active = __activeA & __activeB;
+            int __inc = __popc(__active) * __update;
+
+            unsigned __laneid;
+            asm volatile ("mov.u32 %0, %laneid;" : "=r"(__laneid));
+            int __leader = __ffs(__active) - 1;
+
+            if(__leader == __laneid)
+            {
+                __token = __barrier.arrive(__inc);
+            }
+            __token = __shfl_sync(__active, __token, __leader);
+#endif
+            return __token;
+        }
+        else
+#endif
+        {
+            return __barrier.arrive(__update);
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait(arrival_token && __phase) const
+    {
+        _CUDA_VSTD::__libcpp_thread_poll_with_backoff(__poll_tester(this, _CUDA_VSTD::move(__phase)));
+    }
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_wait()
+    {
+        wait(arrive());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_drop()
+    {
+#if __CUDA_ARCH__ >= 800
+        if (__isShared(&__barrier)) {
+            asm volatile ("mbarrier.arrive_drop.shared.b64 _, [%0];"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__barrier)))
+                : "memory");
+        }
+        else
+#endif
+        {
+            __barrier.arrive_and_drop();
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr ptrdiff_t max() noexcept
+    {
+        return (1 << 20) - 1;
+    }
+};
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA_DEVICE
+
+__device__
+inline std::uint64_t * barrier_native_handle(barrier<thread_scope_block> & b) {
+    return reinterpret_cast<std::uint64_t *>(&b.__barrier);
+}
+
+_LIBCUDACXX_END_NAMESPACE_CUDA_DEVICE
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+template<>
+class barrier<thread_scope_thread, std::__empty_completion> : private barrier<thread_scope_block> {
+    using __base = barrier<thread_scope_block>;
+
+public:
+    using __base::__base;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend void init(barrier * __b, std::ptrdiff_t __expected, std::__empty_completion __completion = std::__empty_completion()) {
+        init(static_cast<__base *>(__b), __expected, __completion);
+    }
+
+    using __base::arrive;
+    using __base::wait;
+    using __base::arrive_and_wait;
+    using __base::arrive_and_drop;
+    using __base::max;
+};
+
+template<std::size_t _Alignment>
+_LIBCUDACXX_INLINE_VISIBILITY
+inline void __strided_memcpy(char * __destination, char const * __source, std::size_t __total_size, std::size_t __rank, std::size_t __stride = 1) {
+    if (__stride == 1) {
+        memcpy(__destination, __source, __total_size);
+    }
+    else {
+        for (std::size_t __offset = __rank * _Alignment; __offset < __total_size; __offset += __stride * _Alignment) {
+            memcpy(__destination + __offset, __source + __offset, _Alignment);
+        }
+    }
+}
+
+#if __CUDA_ARCH__ >= 800
+template<std::size_t _Alignment, bool _Large = (_Alignment > 16)>
+struct __memcpy_async_impl {
+    __device__ static inline bool __copy(char * __destination, char const * __source, std::size_t __total_size, std::size_t __rank, std::size_t __stride) {
+        __strided_memcpy<_Alignment>(__destination, __source, __total_size, __rank, __stride);
+        return false;
+    }
+};
+
+template<>
+struct __memcpy_async_impl<4, false> {
+    __device__ static inline bool __copy(char * __destination, char const * __source, std::size_t __total_size, std::size_t __rank, std::size_t __stride) {
+        for (std::size_t __offset = __rank * 4; __offset < __total_size; __offset += __stride * 4) {
+            asm volatile ("cp.async.ca.shared.global [%0], [%1], 4, 4;"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(__destination + __offset))),
+                    "l"(__source + __offset)
+                : "memory");
+        }
+        return true;
+    }
+};
+
+template<>
+struct __memcpy_async_impl<8, false> {
+    __device__ static inline bool __copy(char * __destination, char const * __source, std::size_t __total_size, std::size_t __rank, std::size_t __stride) {
+        for (std::size_t __offset = __rank * 8; __offset < __total_size; __offset += __stride * 8) {
+            asm volatile ("cp.async.ca.shared.global [%0], [%1], 8, 8;"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(__destination + __offset))),
+                    "l"(__source + __offset)
+                : "memory");
+        }
+        return true;
+    }
+};
+
+template<>
+struct __memcpy_async_impl<16, false> {
+    __device__ static inline bool __copy(char * __destination, char const * __source, std::size_t __total_size, std::size_t __rank, std::size_t __stride) {
+        for (std::size_t __offset = __rank * 16; __offset < __total_size; __offset += __stride * 16) {
+            asm volatile ("cp.async.cg.shared.global [%0], [%1], 16, 16;"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(__destination + __offset))),
+                    "l"(__source + __offset)
+                : "memory");
+        }
+        return true;
+    }
+};
+
+template<std::size_t _Alignment>
+struct __memcpy_async_impl<_Alignment, true> : public __memcpy_async_impl<16, false> { };
+#endif
+
+template<thread_scope _Sco, typename _CompF, bool _Is_mbarrier = (_Sco >= thread_scope_block) && std::is_same<_CompF, std::__empty_completion>::value>
+_LIBCUDACXX_INLINE_VISIBILITY
+inline void __memcpy_async_synchronize(barrier<_Sco, _CompF> & __barrier, bool __is_async) {
+#if __CUDA_ARCH__ >= 800
+    if (__is_async) {
+        if (_Is_mbarrier && __isShared(&__barrier)) {
+            asm volatile ("cp.async.mbarrier.arrive.shared.b64 [%0];"
+                :: "r"(static_cast<std::uint32_t>(__cvta_generic_to_shared(&__barrier)))
+                : "memory");
+        }
+        else {
+            asm volatile ("cp.async.wait_all;"
+                ::: "memory");
+        }
+    }
+#endif
+}
+
+template<std::size_t _Native_alignment, typename _Group, typename _Sync>
+_LIBCUDACXX_INLINE_VISIBILITY
+void inline __memcpy_async(_Group const & __group, char * __destination, char const * __source, std::size_t __size, _Sync & __sync) {
+    bool __is_async = false;
+
+#if __CUDA_ARCH__ >= 800
+    __is_async = __isShared(__destination) && __isGlobal(__source);
+
+    if (__is_async) {
+        if (_Native_alignment < 4) {
+            auto __source_address = reinterpret_cast<std::uintptr_t>(__source);
+            auto __destination_address = reinterpret_cast<std::uintptr_t>(__destination);
+
+            // Lowest bit set will tell us what the common alignment of the three values is.
+            auto _Alignment = __ffs(__source_address | __destination_address | __size);
+
+            switch (_Alignment) {
+                default: __is_async = __memcpy_async_impl<16>::__copy(__destination, __source, __size, __group.thread_rank(), __group.size()); break;
+                case 4: __is_async = __memcpy_async_impl<8>::__copy(__destination, __source, __size, __group.thread_rank(), __group.size()); break;
+                case 3: __is_async = __memcpy_async_impl<4>::__copy(__destination, __source, __size, __group.thread_rank(), __group.size()); break;
+                case 2: // fallthrough
+                case 1: __is_async = __memcpy_async_impl<1>::__copy(__destination, __source, __size, __group.thread_rank(), __group.size()); break;
+            }
+        }
+        else {
+            __is_async = __memcpy_async_impl<_Native_alignment>::__copy(__destination, __source, __size, __group.thread_rank(), __group.size());
+        }
+    }
+    else
+#endif
+    {
+        __strided_memcpy<_Native_alignment>(__destination, __source, __size, __group.thread_rank(), __group.size());
+    }
+
+    __memcpy_async_synchronize(__sync, __is_async);
+}
+
+struct __single_thread_group {
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void sync() const {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr std::size_t size() const { return 1; };
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr std::size_t thread_rank() const { return 0; };
+};
+
+template<typename _Group, class _Tp, thread_scope _Sco, typename _CompF>
+_LIBCUDACXX_INLINE_VISIBILITY
+void memcpy_async(_Group const & __group, _Tp * __destination, _Tp const * __source, std::size_t __size, barrier<_Sco, _CompF> & __barrier) {
+    // When compiling with NVCC and GCC 4.8, certain user defined types that _are_ trivially copyable are
+    // incorrectly classified as not trivially copyable. Remove this assertion to allow for their usage with
+    // memcpy_async when compiling with GCC 4.8.
+    // FIXME: remove the #if once GCC 4.8 is no longer supported.
+#if !defined(_LIBCUDACXX_COMPILER_GCC) || _GNUC_VER > 408
+    static_assert(std::is_trivially_copyable<_Tp>::value, "memcpy_async requires a trivially copyable type");
+#endif
+
+    __memcpy_async<alignof(_Tp)>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __barrier);
+}
+
+template<typename _Group, class _Tp, std::size_t _Alignment, thread_scope _Sco, typename _CompF, std::size_t _Larger_alignment = (alignof(_Tp) > _Alignment) ? alignof(_Tp) : _Alignment>
+_LIBCUDACXX_INLINE_VISIBILITY
+void memcpy_async(_Group const & __group, _Tp * __destination, _Tp const * __source, aligned_size_t<_Alignment> __size, barrier<_Sco, _CompF> & __barrier) {
+    // When compiling with NVCC and GCC 4.8, certain user defined types that _are_ trivially copyable are
+    // incorrectly classified as not trivially copyable. Remove this assertion to allow for their usage with
+    // memcpy_async when compiling with GCC 4.8.
+    // FIXME: remove the #if once GCC 4.8 is no longer supported.
+#if !defined(_LIBCUDACXX_COMPILER_GCC) || _GNUC_VER > 408
+    static_assert(std::is_trivially_copyable<_Tp>::value, "memcpy_async requires a trivially copyable type");
+#endif
+
+    __memcpy_async<_Larger_alignment>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __barrier);
+}
+
+template<class _Tp, typename _Size, thread_scope _Sco, typename _CompF>
+_LIBCUDACXX_INLINE_VISIBILITY
+void memcpy_async(_Tp * __destination, _Tp const * __source, _Size __size, barrier<_Sco, _CompF> & __barrier) {
+    memcpy_async(__single_thread_group{}, __destination, __source, __size, __barrier);
+}
+
+template<typename _Group, thread_scope _Sco, typename _CompF>
+_LIBCUDACXX_INLINE_VISIBILITY
+void memcpy_async(_Group const & __group, void * __destination, void const * __source, std::size_t __size, barrier<_Sco, _CompF> & __barrier) {
+    __memcpy_async<1>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __barrier);
+}
+
+template<typename _Group, std::size_t _Alignment, thread_scope _Sco, typename _CompF>
+_LIBCUDACXX_INLINE_VISIBILITY
+void memcpy_async(_Group const & __group, void * __destination, void const * __source, aligned_size_t<_Alignment> __size, barrier<_Sco, _CompF> & __barrier) {
+    __memcpy_async<_Alignment>(__group, reinterpret_cast<char *>(__destination), reinterpret_cast<char const *>(__source), __size, __barrier);
+}
+
+template<typename _Size, thread_scope _Sco, typename _CompF>
+_LIBCUDACXX_INLINE_VISIBILITY
+void memcpy_async(void * __destination, void const * __source, _Size __size, barrier<_Sco, _CompF> & __barrier) {
+    memcpy_async(__single_thread_group{}, __destination, __source, __size, __barrier);
+}
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_BARRIER
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/bit b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/bit
new file mode 100644
index 000000000000..af80d4832c12
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/bit
@@ -0,0 +1,24 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_BIT
+#define _CUDA_BIT
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "cstdint"
+#include "limits"
+#include "type_traits"
+#include "detail/libcxx/include/bit"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_BIT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cassert b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cassert
new file mode 100644
index 000000000000..96b5c6e8f679
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cassert
@@ -0,0 +1,28 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CASSERT
+#define _CUDA_CASSERT
+
+#ifndef __CUDACC_RTC__
+
+#include <cassert>
+#include <assert.h>
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/cassert"
+
+#include "detail/__pragma_pop"
+
+#endif //__CUDACC_RTC__
+
+#endif //_CUDA_CASSERT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ccomplex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ccomplex
new file mode 100644
index 000000000000..29de0d603972
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ccomplex
@@ -0,0 +1,14 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CCOMPLEX
+#define _CUDA_CCOMPLEX
+
+#include "complex"
+
+#endif //_CUDA_CCOMPLEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cfloat b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cfloat
new file mode 100644
index 000000000000..ab752a097142
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cfloat
@@ -0,0 +1,26 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CFLOAT
+#define _CUDA_CFLOAT
+
+#ifndef __CUDACC_RTC__
+    #include <cfloat>
+    #include <float.h>
+#endif //__CUDACC_RTC__
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/cfloat"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CFLOAT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/chrono b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/chrono
new file mode 100644
index 000000000000..a3abd9ce7cb5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/chrono
@@ -0,0 +1,83 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CHRONO
+#define _CUDA_CHRONO
+
+#ifndef __CUDACC_RTC__
+    #include <chrono>
+#endif //__CUDACC_RTC__
+
+#include "ctime"
+#include "type_traits"
+#include "ratio"
+#include "limits"
+#include "version"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+// Silence NVCC warnings `long double` arising from chrono floating pointer
+// user-defined literals which are defined in terms of `long double`.
+
+// FIXME: There is currently no way to disable this diagnostic in a fine-grained
+// fashion; if you include this header, the diagnostic will be suppressed
+// throughout the translation unit. The alternative is loosing (conforming)
+// chrono user-defined literals; this seems like the lesser of two evils, so...
+#if defined(_LIBCUDACXX_COMPILER_NVCC)
+#  if (CUDART_VERSION >= 11050)
+#    pragma nv_diag_suppress cuda_demote_unsupported_floating_point
+#  else
+#    pragma diag_suppress cuda_demote_unsupported_floating_point
+#  endif
+#endif
+
+#include "detail/libcxx/include/chrono"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+namespace chrono {
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+system_clock::time_point system_clock::now() _NOEXCEPT
+{
+#ifdef __CUDA_ARCH__
+    uint64_t __time;
+    asm volatile("mov.u64 %0, %globaltimer;":"=l"(__time)::);
+    return time_point(duration_cast<duration>(nanoseconds(__time)));
+#else
+    return time_point(duration_cast<duration>(nanoseconds(
+            ::std::chrono::duration_cast<::std::chrono::nanoseconds>(
+                ::std::chrono::system_clock::now().time_since_epoch()
+            ).count()
+           )));
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+time_t system_clock::to_time_t(const system_clock::time_point& __t) _NOEXCEPT
+{
+    return time_t(duration_cast<seconds>(__t.time_since_epoch()).count());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+system_clock::time_point system_clock::from_time_t(time_t __t) _NOEXCEPT
+{
+    return time_point(seconds(__t));;
+}
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CHRONO
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/climits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/climits
new file mode 100644
index 000000000000..7c52a91b5ae2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/climits
@@ -0,0 +1,106 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CLIMITS
+#define _CUDA_CLIMITS
+
+#ifndef __CUDACC_RTC__
+    #include <climits>
+    #include <limits.h>
+    #include <cstdint>
+#else
+    #define CHAR_BIT 8
+
+    #define SCHAR_MIN (-128)
+    #define SCHAR_MAX 127
+    #define UCHAR_MAX 255
+    #define __CHAR_UNSIGNED__ ('\xff' > 0) // CURSED
+    #if __CHAR_UNSIGNED__
+        #define CHAR_MIN 0
+        #define CHAR_MAX UCHAR_MAX
+    #else
+        #define CHAR_MIN SCHAR_MIN
+        #define CHAR_MAX SCHAR_MAX
+    #endif
+    #define SHRT_MIN (-SHRT_MAX - 1)
+    #define SHRT_MAX 0x7fff
+    #define USHRT_MAX 0xffff
+    #define INT_MIN (-INT_MAX - 1)
+    #define INT_MAX 0x7fffffff
+    #define UINT_MAX 0xffffffff
+    #define LONG_MIN (-LONG_MAX - 1)
+    #ifdef __LP64__
+        #define LONG_MAX LLONG_MAX
+        #define ULONG_MAX ULLONG_MAX
+    #else
+        #define LONG_MAX INT_MAX
+        #define ULONG_MAX UINT_MAX
+    #endif
+    #define LLONG_MIN (-LLONG_MAX - 1)
+    #define LLONG_MAX 0x7fffffffffffffff
+    #define ULLONG_MAX 0xffffffffffffffff
+
+    #define __FLT_RADIX__ 2
+    #define __FLT_MANT_DIG__ 24
+    #define __FLT_DIG__ 6
+    #define __FLT_MIN__ 1.17549435082228750796873653722224568e-38F
+    #define __FLT_MAX__ 3.40282346638528859811704183484516925e+38F
+    #define __FLT_EPSILON__ 1.19209289550781250000000000000000000e-7F
+    #define __FLT_MIN_EXP__ (-125)
+    #define __FLT_MIN_10_EXP__ (-37)
+    #define __FLT_MAX_EXP__ 128
+    #define __FLT_MAX_10_EXP__ 38
+    #define __FLT_DENORM_MIN__ 1.40129846432481707092372958328991613e-45F
+    #define __DBL_MANT_DIG__ 53
+    #define __DBL_DIG__ 15
+    #define __DBL_MIN__ 2.22507385850720138309023271733240406e-308
+    #define __DBL_MAX__ 1.79769313486231570814527423731704357e+308
+    #define __DBL_EPSILON__ 2.22044604925031308084726333618164062e-16
+    #define __DBL_MIN_EXP__ (-1021)
+    #define __DBL_MIN_10_EXP__ (-307)
+    #define __DBL_MAX_EXP__ 1024
+    #define __DBL_MAX_10_EXP__ 308
+    #define __DBL_DENORM_MIN__ 4.94065645841246544176568792868221372e-324
+
+    template<typename _To, typename _From>
+    static __device__ __forceinline__
+    _To __cowchild_cast(_From __from)
+    {
+        static_assert(sizeof(_From) == sizeof(_To), "");
+        union __cast { _From __from; _To __to; };
+        __cast __c;
+        __c.__from = __from;
+        return __c.__to;
+    }
+
+    #define __builtin_huge_valf() __cowchild_cast<float>(0x7f800000)
+    #define __builtin_nanf(__dummy) __cowchild_cast<float>(0x7fc00000)
+    #define __builtin_nansf(__dummy) __cowchild_cast<float>(0x7fa00000)
+    #define __builtin_huge_val() __cowchild_cast<double>(0x7ff0000000000000)
+    #define __builtin_nan(__dummy) __cowchild_cast<double>(0x7ff8000000000000)
+    #define __builtin_nans(__dummy) __cowchild_cast<double>(0x7ff4000000000000)
+#endif //__CUDACC_RTC__
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/climits"
+
+// ICC defines __CHAR_BIT__ by default
+// accept that, but assert it is what we expect
+#ifdef __CHAR_BIT__
+    static_assert(__CHAR_BIT__ == 8, "");
+#else
+    #define __CHAR_BIT__ 8
+#endif
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CLIMITS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cmath b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cmath
new file mode 100644
index 000000000000..a186b93ceebf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cmath
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CMATH
+#define _CUDA_CMATH
+
+#include "limits"
+#include "type_traits"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#ifndef _LIBCUDACXX_COMPILER_NVRTC
+#include <math.h>
+#endif
+#include "detail/libcxx/include/cmath"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CMATH
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/complex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/complex
new file mode 100644
index 000000000000..69ec036d2489
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/complex
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_COMPLEX
+#define _CUDA_COMPLEX
+
+#include "cmath"
+#include "cstdint"
+#include "climits"
+#include "type_traits"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/complex"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_COMPLEX
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstddef b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstddef
new file mode 100644
index 000000000000..63a9dada91ed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstddef
@@ -0,0 +1,40 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CSTDDEF
+#define _CUDA_CSTDDEF
+
+#ifndef __CUDACC_RTC__
+    #include <cstddef>
+    #include <stddef.h>
+#else
+    #define offsetof(type, member) (cuda::std::size_t)((char*)&(((type *)0)->member) - (char*)0)
+#endif //__CUDACC_RTC__
+
+#include "version"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+typedef decltype(nullptr) nullptr_t;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#include "detail/libcxx/include/cstddef"
+
+#if _LIBCUDACXX_STD_VER > 14
+    #include "type_traits"
+#endif
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CSTDDEF
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstdint b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstdint
new file mode 100644
index 000000000000..e95accd4a4e0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/cstdint
@@ -0,0 +1,90 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CSTDINT
+#define _CUDA_CSTDINT
+
+#ifndef __CUDACC_RTC__
+    #include <cstdint>
+    #include <stdint.h>
+#else
+    typedef signed char int8_t;
+    typedef unsigned char uint8_t;
+    typedef signed short int16_t;
+    typedef unsigned short uint16_t;
+    typedef signed int int32_t;
+    typedef unsigned int uint32_t;
+    typedef signed long long int64_t;
+    typedef unsigned long long uint64_t;
+
+#define _LIBCUDACXX_ADDITIONAL_INTS(N) \
+    typedef int##N##_t int_fast##N##_t; \
+    typedef uint##N##_t uint_fast##N##_t; \
+    typedef int##N##_t int_least##N##_t; \
+    typedef uint##N##_t uint_least##N##_t
+
+    _LIBCUDACXX_ADDITIONAL_INTS(8);
+    _LIBCUDACXX_ADDITIONAL_INTS(16);
+    _LIBCUDACXX_ADDITIONAL_INTS(32);
+    _LIBCUDACXX_ADDITIONAL_INTS(64);
+#undef _LIBCUDACXX_ADDITIONAL_INTS
+
+    typedef int64_t intptr_t;
+    typedef uint64_t uintptr_t;
+    typedef int64_t intmax_t;
+    typedef uint64_t uintmax_t;
+
+    #define INT8_MIN SCHAR_MIN
+    #define INT16_MIN SHRT_MIN
+    #define INT32_MIN INT_MIN
+    #define INT64_MIN LLONG_MIN
+    #define INT8_MAX SCHAR_MAX
+    #define INT16_MAX SHRT_MAX
+    #define INT32_MAX INT_MAX
+    #define INT64_MAX LLONG_MAX
+    #define UINT8_MAX UCHAR_MAX
+    #define UINT16_MAX USHRT_MAX
+    #define UINT32_MAX UINT_MAX
+    #define UINT64_MAX ULLONG_MAX
+    #define INT_FAST8_MIN SCHAR_MIN
+    #define INT_FAST16_MIN SHRT_MIN
+    #define INT_FAST32_MIN INT_MIN
+    #define INT_FAST64_MIN LLONG_MIN
+    #define INT_FAST8_MAX SCHAR_MAX
+    #define INT_FAST16_MAX SHRT_MAX
+    #define INT_FAST32_MAX INT_MAX
+    #define INT_FAST64_MAX LLONG_MAX
+    #define UINT_FAST8_MAX UCHAR_MAX
+    #define UINT_FAST16_MAX USHRT_MAX
+    #define UINT_FAST32_MAX UINT_MAX
+    #define UINT_FAST64_MAX ULLONG_MAX
+
+    #define INT8_C(X) ((int_least8_t)(X))
+    #define INT16_C(X) ((int_least16_t)(X))
+    #define INT32_C(X) ((int_least32_t)(X))
+    #define INT64_C(X) ((int_least64_t)(X))
+    #define UINT8_C(X) ((uint_least8_t)(X))
+    #define UINT16_C(X) ((uint_least16_t)(X))
+    #define UINT32_C(X) ((uint_least32_t)(X))
+    #define UINT64_C(X) ((uint_least64_t)(X))
+    #define INTMAX_C(X) ((intmax_t)(X))
+    #define UINTMAX_C(X) ((uintmax_t)(X))
+#endif //__CUDACC_RTC__
+
+#include "version"
+#include "climits"
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/cstdint"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CSTDINT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ctime b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ctime
new file mode 100644
index 000000000000..1825c373bbb3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ctime
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_CTIME
+#define _CUDA_CTIME
+
+#ifndef __CUDACC_RTC__
+    #include <time.h>
+#else
+    typedef long long int time_t;
+#endif
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/ctime"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_CTIME
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__access_property b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__access_property
new file mode 100644
index 000000000000..d07a4bab3721
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__access_property
@@ -0,0 +1,323 @@
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2021 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ *
+ * NVIDIA SOFTWARE LICENSE
+ *
+ * This license is a legal agreement between you and NVIDIA Corporation ("NVIDIA") and governs your use of the NVIDIA/CUDA C++ Library software and materials provided hereunder (“SOFTWARE”).
+ *
+ * This license can be accepted only by an adult of legal age of majority in the country in which the SOFTWARE is used. If you are under the legal age of majority, you must ask your parent or legal guardian to consent to this license. By taking delivery of the SOFTWARE, you affirm that you have reached the legal age of majority, you accept the terms of this license, and you take legal and financial responsibility for the actions of your permitted users.
+ *
+ * You agree to use the SOFTWARE only for purposes that are permitted by (a) this license, and (b) any applicable law, regulation or generally accepted practices or guidelines in the relevant jurisdictions.
+ *
+ * 1. LICENSE. Subject to the terms of this license, NVIDIA grants you a non-exclusive limited license to: (a) install and use the SOFTWARE, and (b) distribute the SOFTWARE subject to the distribution requirements described in this license. NVIDIA reserves all rights, title and interest in and to the SOFTWARE not expressly granted to you under this license.
+ *
+ * 2. DISTRIBUTION REQUIREMENTS. These are the distribution requirements for you to exercise the distribution grant:
+ * a.      The terms under which you distribute the SOFTWARE must be consistent with the terms of this license, including (without limitation) terms relating to the license grant and license restrictions and protection of NVIDIA’s intellectual property rights.
+ * b.      You agree to notify NVIDIA in writing of any known or suspected distribution or use of the SOFTWARE not in compliance with the requirements of this license, and to enforce the terms of your agreements with respect to distributed SOFTWARE.
+ *
+ * 3. LIMITATIONS. Your license to use the SOFTWARE is restricted as follows:
+ * a.      The SOFTWARE is licensed for you to develop applications only for use in systems with NVIDIA GPUs.
+ * b.      You may not reverse engineer, decompile or disassemble, or remove copyright or other proprietary notices from any portion of the SOFTWARE or copies of the SOFTWARE.
+ * c.      You may not modify or create derivative works of any portion of the SOFTWARE.
+ * d.      You may not bypass, disable, or circumvent any technical measure, encryption, security, digital rights management or authentication mechanism in the SOFTWARE.
+ * e.      You may not use the SOFTWARE in any manner that would cause it to become subject to an open source software license. As examples, licenses that require as a condition of use, modification, and/or distribution that the SOFTWARE be (i) disclosed or distributed in source code form; (ii) licensed for the purpose of making derivative works; or (iii) redistributable at no charge.
+ * f.      Unless you have an agreement with NVIDIA for this purpose, you may not use the SOFTWARE with any system or application where the use or failure of the system or application can reasonably be expected to threaten or result in personal injury, death, or catastrophic loss. Examples include use in avionics, navigation, military, medical, life support or other life critical applications. NVIDIA does not design, test or manufacture the SOFTWARE for these critical uses and NVIDIA shall not be liable to you or any third party, in whole or in part, for any claims or damages arising from such uses.
+ * g.      You agree to defend, indemnify and hold harmless NVIDIA and its affiliates, and their respective employees, contractors, agents, officers and directors, from and against any and all claims, damages, obligations, losses, liabilities, costs or debt, fines, restitutions and expenses (including but not limited to attorney’s fees and costs incident to establishing the right of indemnification) arising out of or related to use of the SOFTWARE outside of the scope of this Agreement, or not in compliance with its terms.
+ *
+ * 4. PRE-RELEASE. SOFTWARE versions identified as alpha, beta, preview, early access or otherwise as pre-release may not be fully functional, may contain errors or design flaws, and may have reduced or different security, privacy, availability, and reliability standards relative to commercial versions of NVIDIA software and materials. You may use a pre-release SOFTWARE version at your own risk, understanding that these versions are not intended for use in production or business-critical systems.
+ *
+ * 5. OWNERSHIP. The SOFTWARE and the related intellectual property rights therein are and will remain the sole and exclusive property of NVIDIA or its licensors. The SOFTWARE is copyrighted and protected by the laws of the United States and other countries, and international treaty provisions. NVIDIA may make changes to the SOFTWARE, at any time without notice, but is not obligated to support or update the SOFTWARE.
+ *
+ * 6. COMPONENTS UNDER OTHER LICENSES. The SOFTWARE may include NVIDIA or third-party components with separate legal notices or terms as may be described in proprietary notices accompanying the SOFTWARE. If and to the extent there is a conflict between the terms in this license and the license terms associated with a component, the license terms associated with the components control only to the extent necessary to resolve the conflict.
+ *
+ * 7. FEEDBACK. You may, but don’t have to, provide to NVIDIA any Feedback. “Feedback” means any suggestions, bug fixes, enhancements, modifications, feature requests or other feedback regarding the SOFTWARE. For any Feedback that you voluntarily provide, you hereby grant NVIDIA and its affiliates a perpetual, non-exclusive, worldwide, irrevocable license to use, reproduce, modify, license, sublicense (through multiple tiers of sublicensees), and distribute (through multiple tiers of distributors) the Feedback without the payment of any royalties or fees to you. NVIDIA will use Feedback at its choice.
+ *
+ * 8. NO WARRANTIES. THE SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY EXPRESS OR IMPLIED WARRANTY OF ANY KIND INCLUDING, BUT NOT LIMITED TO, WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, OR FITNESS FOR A PARTICULAR PURPOSE. NVIDIA DOES NOT WARRANT THAT THE SOFTWARE WILL MEET YOUR REQUIREMENTS OR THAT THE OPERATION THEREOF WILL BE UNINTERRUPTED OR ERROR-FREE, OR THAT ALL ERRORS WILL BE CORRECTED.
+ *
+ * 9. LIMITATIONS OF LIABILITY. TO THE MAXIMUM EXTENT PERMITTED BY LAW, NVIDIA AND ITS AFFILIATES SHALL NOT BE LIABLE FOR ANY SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL DAMAGES, OR ANY LOST PROFITS, PROJECT DELAYS, LOSS OF USE, LOSS OF DATA OR LOSS OF GOODWILL, OR THE COSTS OF PROCURING SUBSTITUTE PRODUCTS, ARISING OUT OF OR IN CONNECTION WITH THIS LICENSE OR THE USE OR PERFORMANCE OF THE SOFTWARE, WHETHER SUCH LIABILITY ARISES FROM ANY CLAIM BASED UPON BREACH OF CONTRACT, BREACH OF WARRANTY, TORT (INCLUDING NEGLIGENCE), PRODUCT LIABILITY OR ANY OTHER CAUSE OF ACTION OR THEORY OF LIABILITY, EVEN IF NVIDIA HAS PREVIOUSLY BEEN ADVISED OF, OR COULD REASONABLY HAVE FORESEEN, THE POSSIBILITY OF SUCH DAMAGES. IN NO EVENT WILL NVIDIA’S AND ITS AFFILIATES TOTAL CUMULATIVE LIABILITY UNDER OR ARISING OUT OF THIS LICENSE EXCEED US$10.00. THE NATURE OF THE LIABILITY OR THE NUMBER OF CLAIMS OR SUITS SHALL NOT ENLARGE OR EXTEND THIS LIMIT.
+ *
+ * 10. TERMINATION. Your rights under this license will terminate automatically without notice from NVIDIA if you fail to comply with any term and condition of this license or if you commence or participate in any legal proceeding against NVIDIA with respect to the SOFTWARE. NVIDIA may terminate this license with advance written notice to you if NVIDIA decides to no longer provide the SOFTWARE in a country or, in NVIDIA’s sole discretion, the continued use of it is no longer commercially viable. Upon any termination of this license, you agree to promptly discontinue use of the SOFTWARE and destroy all copies in your possession or control. Your prior distributions in accordance with this license are not affected by the termination of this license. All provisions of this license will survive termination, except for the license granted to you.
+ *
+ * 11. APPLICABLE LAW. This license will be governed in all respects by the laws of the United States and of the State of Delaware as those laws are applied to contracts entered into and performed entirely within Delaware by Delaware residents, without regard to the conflicts of laws principles. The United Nations Convention on Contracts for the International Sale of Goods is specifically disclaimed. You agree to all terms of this Agreement in the English language. The state or federal courts residing in Santa Clara County, California shall have exclusive jurisdiction over any dispute or claim arising out of this license. Notwithstanding this, you agree that NVIDIA shall still be allowed to apply for injunctive remedies or an equivalent type of urgent legal relief in any jurisdiction.
+ *
+ * 12. NO ASSIGNMENT. This license and your rights and obligations thereunder may not be assigned by you by any means or operation of law without NVIDIA’s permission. Any attempted assignment not approved by NVIDIA in writing shall be void and of no effect.
+ *
+ * 13. EXPORT. The SOFTWARE is subject to United States export laws and regulations. You agree that you will not ship, transfer or export the SOFTWARE into any country, or use the SOFTWARE in any manner, prohibited by the United States Bureau of Industry and Security or economic sanctions regulations administered by the U.S. Department of Treasury’s Office of Foreign Assets Control (OFAC), or any applicable export laws, restrictions or regulations. These laws include restrictions on destinations, end users and end use. By accepting this license, you confirm that you are not a resident or citizen of any country currently embargoed by the U.S. and that you are not otherwise prohibited from receiving the SOFTWARE.
+ *
+ * 14. GOVERNMENT USE. The SOFTWARE has been developed entirely at private expense and is “commercial items” consisting of “commercial computer software” and “commercial computer software documentation” provided with RESTRICTED RIGHTS. Use, duplication or disclosure by the U.S. Government or a U.S. Government subcontractor is subject to the restrictions in this license pursuant to DFARS 227.7202-3(a) or as set forth in subparagraphs (b)(1) and (2) of the Commercial Computer Software - Restricted Rights clause at FAR 52.227-19, as applicable. Contractor/manufacturer is NVIDIA, 2788 San Tomas Expressway, Santa Clara, CA 95051.
+ *
+ * 15. ENTIRE AGREEMENT. This license is the final, complete and exclusive agreement between the parties relating to the subject matter of this license and supersedes all prior or contemporaneous understandings and agreements relating to this subject matter, whether oral or written. If any court of competent jurisdiction determines that any provision of this license is illegal, invalid or unenforceable, the remaining provisions will remain in full force and effect. This license may only be modified in a writing signed by an authorized representative of each party.
+ *
+ * (v. August 20, 2021)
+ */
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+namespace __detail_ap {
+
+  __host__ __device__
+  constexpr uint32_t __ap_floor_log2(uint32_t __x) {
+    return (__x == 1 | __x == 0) ? 0 : 1 + __ap_floor_log2(__x >> 1);
+  }
+
+  __host__ __device__
+  constexpr uint32_t __ap_ceil_log2(uint32_t __x) {
+    return (__x == 1 | __x == 0) ? 0 : __ap_floor_log2(__x - 1) + 1;
+  }
+
+  __host__ __device__
+  constexpr uint32_t __ap_min(uint32_t __a, uint32_t __b) noexcept {
+    return (__a < __b) ? __a : __b;
+  }
+
+  __host__ __device__
+  constexpr uint32_t __ap_max(uint32_t __a, uint32_t __b) noexcept {
+    return (__a > __b) ? __a : __b;
+  }
+
+// GCC-7 and below do not properly assume the required number of bits for enumerations
+#if defined(__GNUC__) && __GNUC__ < 8
+# define _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION
+#else
+# define _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION : uint64_t
+#endif
+
+  namespace __sm_80 {
+    namespace __off {
+      enum __l2_cop_off_t _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION {
+        _L2_EVICT_NORMAL = 0,
+        _L2_EVICT_FIRST = 1,
+      };
+    } // namespace __off
+
+    namespace __on {
+      enum __l2_cop_on_t _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION {
+        _L2_EVICT_NORMAL = 0,
+        _L2_EVICT_FIRST = 1,
+        _L2_EVICT_LAST = 2,
+        _L2_EVICT_NORMAL_DEMOTE = 3,
+      };
+    } // namespace __on
+
+    enum __l2_descriptor_mode_t _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION {
+      _DESC_IMPLICIT = 0,
+      _DESC_INTERLEAVED = 2,
+      _DESC_BLOCK_TYPE = 3,
+    };
+
+    enum __l2_eviction_max_way_t _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION {
+      _CUDA_AMPERE_MAX_L2_WAYS = std::uint32_t{16},
+    };
+
+    enum __block_size_t _LIBCUDACXX_AP_ENUM_TYPE_ANNOTATION {
+      _BLOCKSIZE_4K   = 0,
+      _BLOCKSIZE_8K   = 1,
+      _BLOCKSIZE_16K  = 2,
+      _BLOCKSIZE_32K  = 3,
+      _BLOCKSIZE_64K  = 4,
+      _BLOCKSIZE_128K = 5,
+      _BLOCKSIZE_256K = 6,
+      _BLOCKSIZE_512K = 7,
+      _BLOCKSIZE_1M   = 8,
+      _BLOCKSIZE_2M   = 9,
+      _BLOCKSIZE_4M   = 10,
+      _BLOCKSIZE_8M   = 11,
+      _BLOCKSIZE_16M  = 12,
+      _BLOCKSIZE_32M  = 13,
+    };
+
+    struct __block_desc_t {
+      uint64_t __ap_reserved : 37;
+      uint64_t __block_count: 7;
+      uint64_t __block_start: 7;
+      uint64_t __ap_reserved2 : 1;
+      __block_size_t __block_size : 4;
+      __off::__l2_cop_off_t __l2_cop_off : 1;
+      __on::__l2_cop_on_t __l2_cop_on : 2;
+      __l2_descriptor_mode_t __l2_descriptor_mode : 2;
+      uint64_t __l1_inv_dont_allocate : 1;
+      uint64_t __l2_sector_promote_256B : 1;
+      uint64_t __ap_reserved3 : 1;
+
+      __host__ __device__
+      constexpr std::uint64_t __get_descriptor_cexpr() const noexcept {
+        return
+          std::uint64_t(__ap_reserved)            << 0  |
+          std::uint64_t(__block_count)            << 37 |
+          std::uint64_t(__block_start)            << 44 |
+          std::uint64_t(__ap_reserved2)           << 51 |
+          std::uint64_t(__block_size)             << 52 |
+          std::uint64_t(__l2_cop_off)             << 56 |
+          std::uint64_t(__l2_cop_on)              << 57 |
+          std::uint64_t(__l2_descriptor_mode)     << 59 |
+          std::uint64_t(__l1_inv_dont_allocate)   << 61 |
+          std::uint64_t(__l2_sector_promote_256B) << 62 |
+          std::uint64_t(__ap_reserved3)           << 63;
+      }
+
+      inline
+      __host__ __device__
+      std::uint64_t __get_descriptor_non_cexpr() const noexcept { return *reinterpret_cast<const std::uint64_t*>(this); }
+
+      __host__ __device__
+      constexpr std::uint64_t __get_descriptor() const noexcept {
+#if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED)
+        return cuda::std::is_constant_evaluated() ?
+                    __get_descriptor_cexpr() :
+                    __get_descriptor_non_cexpr();
+#else
+        return __get_descriptor_cexpr();
+#endif
+      }
+    };
+    static_assert(sizeof(__block_desc_t) == 8, "__block_desc_t should be 8 bytes");
+    static_assert(sizeof(__block_desc_t) == sizeof(std::uint64_t), "");
+    static_assert(
+      __block_desc_t{(uint64_t)1, (uint64_t)1, (uint64_t)1, (uint64_t)1, __block_size_t::_BLOCKSIZE_8K, __off::_L2_EVICT_FIRST, __on::_L2_EVICT_FIRST, __l2_descriptor_mode_t::_DESC_INTERLEAVED, (uint64_t)1, (uint64_t)1, (uint64_t)1}.__get_descriptor()
+         == 0xF318102000000001, "");
+
+    /* Factory like struct to build a __block_desc_t due to constexpr C++11
+     */
+    struct __block_descriptor_builder { //variable declaration order matters == usage order
+      std::uint32_t __offset;
+      __block_size_t __block_size;
+      std::uint32_t __block_start, __end_hit;
+      std::uint32_t __block_count;
+      __off::__l2_cop_off_t __l2_cop_off;
+      __on::__l2_cop_on_t __l2_cop_on;
+      __l2_descriptor_mode_t __l2_descriptor_mode;
+      bool __l1_inv_dont_allocate, __l2_sector_promote_256B;
+
+      __host__ __device__ static constexpr std::uint32_t __calc_offset(std::size_t __total_bytes) {
+          return __ap_max(std::uint32_t{12}, static_cast<std::uint32_t>(__ap_ceil_log2(static_cast<uint32_t>(__total_bytes))) - std::uint32_t{7});
+      }
+
+      __host__ __device__ static constexpr std::uint32_t __calc_block_start(std::uintptr_t __ptr, std::size_t __total_bytes) {
+          return static_cast<uint32_t>(__ptr >> __calc_offset(static_cast<uint32_t>(__total_bytes)));
+      }
+
+      __host__ __device__ static constexpr std::uint32_t __calc_end_hit(std::uintptr_t __ptr, std::size_t __hit_bytes, std::size_t __total_bytes) {
+          return static_cast<uint32_t>((__ptr + __hit_bytes + (std::uintptr_t{1} << (__calc_offset(static_cast<uint32_t>(__total_bytes)))) - 1) >> __calc_offset(static_cast<uint32_t>(__total_bytes)));
+      }
+
+      __host__ __device__ constexpr __block_descriptor_builder(std::uintptr_t __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, __on::__l2_cop_on_t __hit_prop, __off::__l2_cop_off_t __miss_prop)
+        : __offset(__calc_offset(__total_bytes))
+        , __block_size(static_cast<__block_size_t>(__calc_offset(__total_bytes) - std::uint32_t{12}))
+        , __block_start(__calc_block_start(__ptr, __total_bytes))
+        , __end_hit(__calc_end_hit(__ptr, __hit_bytes, __total_bytes))
+        , __block_count(__calc_end_hit(__ptr, __hit_bytes, __total_bytes) - __calc_block_start(__ptr, __total_bytes))
+        , __l2_cop_off(__miss_prop)
+        , __l2_cop_on(__hit_prop)
+        , __l2_descriptor_mode(_DESC_BLOCK_TYPE)
+        , __l1_inv_dont_allocate(false)
+        , __l2_sector_promote_256B(false)
+        {}
+
+      __host__ __device__
+      constexpr __block_desc_t __get_block() const noexcept {
+        return __block_desc_t { 0, __ap_min(std::uint32_t{0x7f}, __block_count), (__block_start & std::uint32_t{0x7f}), 0, __block_size, __l2_cop_off, __l2_cop_on, _DESC_BLOCK_TYPE, false, false, 0 };
+      }
+    };
+    static_assert(sizeof(std::uintptr_t) > 4, "std::uintptr_t needs at least 5 bytes for this code to work");
+
+    struct __interleave_descriptor_t {
+      uint64_t __ap_reserved : 52;
+      uint64_t __fraction : 4;
+      __off::__l2_cop_off_t __l2_cop_off : 1;
+      __on::__l2_cop_on_t __l2_cop_on : 2;
+      __l2_descriptor_mode_t __l2_descriptor_mode : 2;
+      uint64_t __l1_inv_dont_allocate : 1;
+      uint64_t __l2_sector_promote_256B : 1;
+      uint64_t __ap_reserved2 : 1;
+
+      __host__ __device__
+      constexpr __interleave_descriptor_t(
+                    __on::__l2_cop_on_t __hit_prop,
+                    std::uint32_t __hit_ratio,
+                    __off::__l2_cop_off_t __miss_prop) noexcept
+                        : __fraction(__hit_ratio),
+                          __l2_cop_off(__miss_prop),
+                          __l2_cop_on(__hit_prop),
+                          __l2_descriptor_mode(_DESC_INTERLEAVED),
+                          __l1_inv_dont_allocate(0x0),
+                          __l2_sector_promote_256B(0x0),
+                          __ap_reserved(0x0),
+                          __ap_reserved2(0x0) {}
+
+      __host__ __device__
+      constexpr std::uint64_t __get_descriptor_cexpr() const {
+        return
+          std::uint64_t(__ap_reserved)            << 0  |
+          std::uint64_t(__fraction)               << 52 |
+          std::uint64_t(__l2_cop_off)             << 56 |
+          std::uint64_t(__l2_cop_on)              << 57 |
+          std::uint64_t(__l2_descriptor_mode)     << 59 |
+          std::uint64_t(__l1_inv_dont_allocate)   << 61 |
+          std::uint64_t(__l2_sector_promote_256B) << 62 |
+          std::uint64_t(__ap_reserved2)           << 63;
+      }
+
+      inline
+      __host__ __device__
+      std::uint64_t __get_descriptor_non_cexpr() const noexcept { return *reinterpret_cast<const std::uint64_t*>(this); }
+
+
+      __host__ __device__
+      constexpr std::uint64_t __get_descriptor() const noexcept {
+#if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED)
+        return cuda::std::is_constant_evaluated() ?
+                    __get_descriptor_cexpr() :
+                    __get_descriptor_non_cexpr();
+#else
+        return __get_descriptor_cexpr();
+#endif
+      }
+    };
+    static_assert(sizeof(__interleave_descriptor_t) == 8, "__interleave_descriptor_t should be 8 bytes");
+    static_assert(sizeof(__interleave_descriptor_t) == sizeof(std::uint64_t), "");
+
+    __host__ __device__
+    static constexpr std::uint64_t __interleave_normal() noexcept {
+      return 0x10F0000000000000;
+    }
+
+    __host__ __device__
+    static constexpr std::uint64_t __interleave_streaming() noexcept {
+      return 0x12F0000000000000;
+    }
+
+    __host__ __device__
+    static constexpr std::uint64_t __interleave_persisting() noexcept {
+      return 0x14F0000000000000;
+    }
+
+    __host__ __device__
+    static constexpr std::uint64_t __interleave_normal_demote() noexcept {
+      return 0x16F0000000000000;
+    }
+
+  } // namespace __sm_80
+
+  __host__ __device__
+  constexpr std::uint64_t __interleave(cudaAccessProperty __hit_prop, float __hit_ratio, cudaAccessProperty __miss_prop = cudaAccessPropertyNormal) {
+      return __sm_80::__interleave_descriptor_t(
+          ((__hit_prop == cudaAccessPropertyNormal) ? __sm_80::__on::__l2_cop_on_t::_L2_EVICT_NORMAL_DEMOTE : static_cast<__sm_80::__on::__l2_cop_on_t>(__hit_prop)),
+          __ap_min((static_cast<std::uint32_t>(__hit_ratio) * __sm_80::__l2_eviction_max_way_t::_CUDA_AMPERE_MAX_L2_WAYS), static_cast<std::uint32_t>(__sm_80::__l2_eviction_max_way_t::_CUDA_AMPERE_MAX_L2_WAYS - 1)),
+          static_cast<__sm_80::__off::__l2_cop_off_t>(__miss_prop)
+          ).__get_descriptor();
+  }
+
+  __host__ __device__
+  constexpr std::uint64_t __block(void* __ptr, std::size_t __hit_bytes, std::size_t __total_bytes, cudaAccessProperty __hit_prop, cudaAccessProperty __miss_prop = cudaAccessPropertyNormal) {
+      return (__total_bytes <= (size_t{0xFFFFFFFF}) & __total_bytes != 0 & __hit_bytes <= __total_bytes) ? __sm_80::__block_descriptor_builder(
+          reinterpret_cast<std::uintptr_t>(__ptr),
+          __hit_bytes,
+          __total_bytes,
+          (__hit_prop == cudaAccessPropertyNormal) ? __sm_80::__on::_L2_EVICT_NORMAL_DEMOTE : static_cast<__sm_80::__on::__l2_cop_on_t>(__hit_prop),
+          static_cast<__sm_80::__off::__l2_cop_off_t>(__miss_prop)
+          ).__get_block().__get_descriptor()
+        : __sm_80::__interleave_normal();
+  }
+} // namespace __detail_ap
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__annotated_ptr b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__annotated_ptr
new file mode 100644
index 000000000000..4c7ad4080615
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__annotated_ptr
@@ -0,0 +1,227 @@
+/*
+ * SPDX-FileCopyrightText: Copyright (c) 2021 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+ *
+ * NVIDIA SOFTWARE LICENSE
+ *
+ * This license is a legal agreement between you and NVIDIA Corporation ("NVIDIA") and governs your use of the NVIDIA/CUDA C++ Library software and materials provided hereunder (“SOFTWARE”).
+ *
+ * This license can be accepted only by an adult of legal age of majority in the country in which the SOFTWARE is used. If you are under the legal age of majority, you must ask your parent or legal guardian to consent to this license. By taking delivery of the SOFTWARE, you affirm that you have reached the legal age of majority, you accept the terms of this license, and you take legal and financial responsibility for the actions of your permitted users.
+ *
+ * You agree to use the SOFTWARE only for purposes that are permitted by (a) this license, and (b) any applicable law, regulation or generally accepted practices or guidelines in the relevant jurisdictions.
+ *
+ * 1. LICENSE. Subject to the terms of this license, NVIDIA grants you a non-exclusive limited license to: (a) install and use the SOFTWARE, and (b) distribute the SOFTWARE subject to the distribution requirements described in this license. NVIDIA reserves all rights, title and interest in and to the SOFTWARE not expressly granted to you under this license.
+ *
+ * 2. DISTRIBUTION REQUIREMENTS. These are the distribution requirements for you to exercise the distribution grant:
+ * a.      The terms under which you distribute the SOFTWARE must be consistent with the terms of this license, including (without limitation) terms relating to the license grant and license restrictions and protection of NVIDIA’s intellectual property rights.
+ * b.      You agree to notify NVIDIA in writing of any known or suspected distribution or use of the SOFTWARE not in compliance with the requirements of this license, and to enforce the terms of your agreements with respect to distributed SOFTWARE.
+ *
+ * 3. LIMITATIONS. Your license to use the SOFTWARE is restricted as follows:
+ * a.      The SOFTWARE is licensed for you to develop applications only for use in systems with NVIDIA GPUs.
+ * b.      You may not reverse engineer, decompile or disassemble, or remove copyright or other proprietary notices from any portion of the SOFTWARE or copies of the SOFTWARE.
+ * c.      You may not modify or create derivative works of any portion of the SOFTWARE.
+ * d.      You may not bypass, disable, or circumvent any technical measure, encryption, security, digital rights management or authentication mechanism in the SOFTWARE.
+ * e.      You may not use the SOFTWARE in any manner that would cause it to become subject to an open source software license. As examples, licenses that require as a condition of use, modification, and/or distribution that the SOFTWARE be (i) disclosed or distributed in source code form; (ii) licensed for the purpose of making derivative works; or (iii) redistributable at no charge.
+ * f.      Unless you have an agreement with NVIDIA for this purpose, you may not use the SOFTWARE with any system or application where the use or failure of the system or application can reasonably be expected to threaten or result in personal injury, death, or catastrophic loss. Examples include use in avionics, navigation, military, medical, life support or other life critical applications. NVIDIA does not design, test or manufacture the SOFTWARE for these critical uses and NVIDIA shall not be liable to you or any third party, in whole or in part, for any claims or damages arising from such uses.
+ * g.      You agree to defend, indemnify and hold harmless NVIDIA and its affiliates, and their respective employees, contractors, agents, officers and directors, from and against any and all claims, damages, obligations, losses, liabilities, costs or debt, fines, restitutions and expenses (including but not limited to attorney’s fees and costs incident to establishing the right of indemnification) arising out of or related to use of the SOFTWARE outside of the scope of this Agreement, or not in compliance with its terms.
+ *
+ * 4. PRE-RELEASE. SOFTWARE versions identified as alpha, beta, preview, early access or otherwise as pre-release may not be fully functional, may contain errors or design flaws, and may have reduced or different security, privacy, availability, and reliability standards relative to commercial versions of NVIDIA software and materials. You may use a pre-release SOFTWARE version at your own risk, understanding that these versions are not intended for use in production or business-critical systems.
+ *
+ * 5. OWNERSHIP. The SOFTWARE and the related intellectual property rights therein are and will remain the sole and exclusive property of NVIDIA or its licensors. The SOFTWARE is copyrighted and protected by the laws of the United States and other countries, and international treaty provisions. NVIDIA may make changes to the SOFTWARE, at any time without notice, but is not obligated to support or update the SOFTWARE.
+ *
+ * 6. COMPONENTS UNDER OTHER LICENSES. The SOFTWARE may include NVIDIA or third-party components with separate legal notices or terms as may be described in proprietary notices accompanying the SOFTWARE. If and to the extent there is a conflict between the terms in this license and the license terms associated with a component, the license terms associated with the components control only to the extent necessary to resolve the conflict.
+ *
+ * 7. FEEDBACK. You may, but don’t have to, provide to NVIDIA any Feedback. “Feedback” means any suggestions, bug fixes, enhancements, modifications, feature requests or other feedback regarding the SOFTWARE. For any Feedback that you voluntarily provide, you hereby grant NVIDIA and its affiliates a perpetual, non-exclusive, worldwide, irrevocable license to use, reproduce, modify, license, sublicense (through multiple tiers of sublicensees), and distribute (through multiple tiers of distributors) the Feedback without the payment of any royalties or fees to you. NVIDIA will use Feedback at its choice.
+ *
+ * 8. NO WARRANTIES. THE SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY EXPRESS OR IMPLIED WARRANTY OF ANY KIND INCLUDING, BUT NOT LIMITED TO, WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, OR FITNESS FOR A PARTICULAR PURPOSE. NVIDIA DOES NOT WARRANT THAT THE SOFTWARE WILL MEET YOUR REQUIREMENTS OR THAT THE OPERATION THEREOF WILL BE UNINTERRUPTED OR ERROR-FREE, OR THAT ALL ERRORS WILL BE CORRECTED.
+ *
+ * 9. LIMITATIONS OF LIABILITY. TO THE MAXIMUM EXTENT PERMITTED BY LAW, NVIDIA AND ITS AFFILIATES SHALL NOT BE LIABLE FOR ANY SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL DAMAGES, OR ANY LOST PROFITS, PROJECT DELAYS, LOSS OF USE, LOSS OF DATA OR LOSS OF GOODWILL, OR THE COSTS OF PROCURING SUBSTITUTE PRODUCTS, ARISING OUT OF OR IN CONNECTION WITH THIS LICENSE OR THE USE OR PERFORMANCE OF THE SOFTWARE, WHETHER SUCH LIABILITY ARISES FROM ANY CLAIM BASED UPON BREACH OF CONTRACT, BREACH OF WARRANTY, TORT (INCLUDING NEGLIGENCE), PRODUCT LIABILITY OR ANY OTHER CAUSE OF ACTION OR THEORY OF LIABILITY, EVEN IF NVIDIA HAS PREVIOUSLY BEEN ADVISED OF, OR COULD REASONABLY HAVE FORESEEN, THE POSSIBILITY OF SUCH DAMAGES. IN NO EVENT WILL NVIDIA’S AND ITS AFFILIATES TOTAL CUMULATIVE LIABILITY UNDER OR ARISING OUT OF THIS LICENSE EXCEED US$10.00. THE NATURE OF THE LIABILITY OR THE NUMBER OF CLAIMS OR SUITS SHALL NOT ENLARGE OR EXTEND THIS LIMIT.
+ *
+ * 10. TERMINATION. Your rights under this license will terminate automatically without notice from NVIDIA if you fail to comply with any term and condition of this license or if you commence or participate in any legal proceeding against NVIDIA with respect to the SOFTWARE. NVIDIA may terminate this license with advance written notice to you if NVIDIA decides to no longer provide the SOFTWARE in a country or, in NVIDIA’s sole discretion, the continued use of it is no longer commercially viable. Upon any termination of this license, you agree to promptly discontinue use of the SOFTWARE and destroy all copies in your possession or control. Your prior distributions in accordance with this license are not affected by the termination of this license. All provisions of this license will survive termination, except for the license granted to you.
+ *
+ * 11. APPLICABLE LAW. This license will be governed in all respects by the laws of the United States and of the State of Delaware as those laws are applied to contracts entered into and performed entirely within Delaware by Delaware residents, without regard to the conflicts of laws principles. The United Nations Convention on Contracts for the International Sale of Goods is specifically disclaimed. You agree to all terms of this Agreement in the English language. The state or federal courts residing in Santa Clara County, California shall have exclusive jurisdiction over any dispute or claim arising out of this license. Notwithstanding this, you agree that NVIDIA shall still be allowed to apply for injunctive remedies or an equivalent type of urgent legal relief in any jurisdiction.
+ *
+ * 12. NO ASSIGNMENT. This license and your rights and obligations thereunder may not be assigned by you by any means or operation of law without NVIDIA’s permission. Any attempted assignment not approved by NVIDIA in writing shall be void and of no effect.
+ *
+ * 13. EXPORT. The SOFTWARE is subject to United States export laws and regulations. You agree that you will not ship, transfer or export the SOFTWARE into any country, or use the SOFTWARE in any manner, prohibited by the United States Bureau of Industry and Security or economic sanctions regulations administered by the U.S. Department of Treasury’s Office of Foreign Assets Control (OFAC), or any applicable export laws, restrictions or regulations. These laws include restrictions on destinations, end users and end use. By accepting this license, you confirm that you are not a resident or citizen of any country currently embargoed by the U.S. and that you are not otherwise prohibited from receiving the SOFTWARE.
+ *
+ * 14. GOVERNMENT USE. The SOFTWARE has been developed entirely at private expense and is “commercial items” consisting of “commercial computer software” and “commercial computer software documentation” provided with RESTRICTED RIGHTS. Use, duplication or disclosure by the U.S. Government or a U.S. Government subcontractor is subject to the restrictions in this license pursuant to DFARS 227.7202-3(a) or as set forth in subparagraphs (b)(1) and (2) of the Commercial Computer Software - Restricted Rights clause at FAR 52.227-19, as applicable. Contractor/manufacturer is NVIDIA, 2788 San Tomas Expressway, Santa Clara, CA 95051.
+ *
+ * 15. ENTIRE AGREEMENT. This license is the final, complete and exclusive agreement between the parties relating to the subject matter of this license and supersedes all prior or contemporaneous understandings and agreements relating to this subject matter, whether oral or written. If any court of competent jurisdiction determines that any provision of this license is illegal, invalid or unenforceable, the remaining provisions will remain in full force and effect. This license may only be modified in a writing signed by an authorized representative of each party.
+ *
+ * (v. August 20, 2021)
+ */
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+namespace __detail_ap {
+
+  template <typename _Property>
+  __device__
+  void* __associate_address_space(void* __ptr, _Property __prop) {
+    if (std::is_same<_Property, access_property::shared>::value == true) {
+      bool __b = __isShared(__ptr);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+      _LIBCUDACXX_DEBUG_ASSERT(__b == true);
+#endif
+      __builtin_assume(__b);
+    } else if (std::is_same<_Property, access_property::global>::value == true ||
+               std::is_same<_Property, access_property::normal>::value == true ||
+               std::is_same<_Property, access_property::persisting>::value == true ||
+               std::is_same<_Property, access_property::streaming>::value == true ||
+               std::is_same<_Property, access_property>::value) {
+      bool __b = __isGlobal(__ptr);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+      _LIBCUDACXX_DEBUG_ASSERT(__b == true);
+#endif
+      __builtin_assume(__b);
+    }
+
+    return __ptr;
+  }
+
+  template <typename __Prop>
+  __device__
+  void* __associate_descriptor(void* __ptr, __Prop __prop) {
+    return __associate_descriptor(__ptr, static_cast<std::uint64_t>(access_property(__prop)));
+  }
+
+  template <>
+  inline __device__
+  void* __associate_descriptor(void* __ptr, std::uint64_t __prop) {
+#if __CUDA_ARCH__ >= 800
+    return __nv_associate_access_property(__ptr, __prop);
+#else
+    return __ptr;
+#endif
+  }
+
+  template<>
+  inline __device__
+  void* __associate_descriptor(void* __ptr, access_property::shared) {
+    return __ptr;
+  }
+
+  template<typename _Type, typename _Property>
+  __host__ __device__
+  _Type* __associate(_Type* __ptr, _Property __prop) {
+#ifdef __CUDA_ARCH__
+    return static_cast<_Type*>(__associate_descriptor(
+	    __associate_address_space(const_cast<void*>(static_cast<const void*>(__ptr)), __prop),
+	    __prop));
+#else
+    return __ptr;
+#endif
+  }
+
+
+  template<typename _Property>
+  class __annotated_ptr_base {
+    using __error = typename _Property::__unknown_access_property_type;
+  };
+
+  template<>
+  class __annotated_ptr_base<access_property::shared> {
+    protected:
+      static constexpr std::uint64_t __prop = 0;
+
+      constexpr __annotated_ptr_base() noexcept = default;
+      constexpr __annotated_ptr_base(__annotated_ptr_base const&) = default;
+      _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 __annotated_ptr_base& operator=(const __annotated_ptr_base&) = default;
+      __host__ __device__ constexpr __annotated_ptr_base(access_property::shared) noexcept {}
+      inline __device__ void* __apply_prop(void* __p) const {
+        return __associate(__p, access_property::shared{});
+      }
+      __host__ __device__ constexpr access_property::shared __get_property() const noexcept {
+	return access_property::shared{};
+      }
+  };
+
+  template<>
+  class __annotated_ptr_base<access_property::global> {
+    protected:
+      static constexpr std::uint64_t __prop = __sm_80::__interleave_normal();
+
+      constexpr __annotated_ptr_base() noexcept = default;
+      constexpr __annotated_ptr_base(__annotated_ptr_base const&) = default;
+      _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 __annotated_ptr_base& operator=(const __annotated_ptr_base&) = default;
+      __host__ __device__ constexpr __annotated_ptr_base(access_property::global) noexcept {}
+      inline __device__ void* __apply_prop(void* __p) const {
+        return __associate(__p, access_property::global{});
+      }
+      __host__ __device__ constexpr access_property::global __get_property() const noexcept {
+	return access_property::global{};
+      }
+  };
+
+  template<>
+  class __annotated_ptr_base<access_property::normal> {
+    protected:
+      static constexpr std::uint64_t __prop = __sm_80::__interleave_normal_demote();
+
+      constexpr __annotated_ptr_base() noexcept = default;
+      constexpr __annotated_ptr_base(__annotated_ptr_base const&) = default;
+      _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 __annotated_ptr_base& operator=(const __annotated_ptr_base&) = default;
+      __host__ __device__ constexpr __annotated_ptr_base(access_property::normal) noexcept {}
+      inline __device__ void* __apply_prop(void* __p) const {
+        return __associate(__p, access_property::normal{});
+      }
+      __host__ __device__ constexpr access_property::normal __get_property() const noexcept {
+	return access_property::normal{};
+      }
+  };
+
+  template<>
+  class __annotated_ptr_base<access_property::persisting> {
+    protected:
+      static constexpr std::uint64_t __prop = __sm_80::__interleave_persisting();
+
+      constexpr __annotated_ptr_base() noexcept = default;
+      constexpr __annotated_ptr_base(__annotated_ptr_base const&) = default;
+      _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 __annotated_ptr_base& operator=(const __annotated_ptr_base&) = default;
+      __host__ __device__ constexpr __annotated_ptr_base(access_property::persisting) noexcept {}
+      inline __device__ void* __apply_prop(void* __p) const {
+        return __associate(__p, access_property::persisting{});
+      }
+      __host__ __device__ constexpr access_property::persisting __get_property() const noexcept {
+	return access_property::persisting{};
+      }
+  };
+
+  template<>
+  class __annotated_ptr_base<access_property::streaming> {
+    protected:
+      static constexpr std::uint64_t __prop = __sm_80::__interleave_streaming();
+
+      constexpr __annotated_ptr_base() noexcept = default;
+      constexpr __annotated_ptr_base(__annotated_ptr_base const&) = default;
+      _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 __annotated_ptr_base& operator=(const __annotated_ptr_base&) = default;
+      __host__ __device__ constexpr __annotated_ptr_base(access_property::streaming) noexcept {}
+      inline __device__ void* __apply_prop(void* __p) const {
+        return __associate(__p, access_property::streaming{});
+      }
+      __host__ __device__ constexpr access_property::streaming __get_property() const noexcept {
+	return access_property::streaming{};
+      }
+  };
+
+  template<>
+  class __annotated_ptr_base<access_property> {
+    protected:
+      std::uint64_t __prop;
+
+      __host__ __device__ constexpr __annotated_ptr_base() noexcept : __prop(access_property()) {}
+      __host__ __device__ constexpr __annotated_ptr_base(std::uint64_t __property) noexcept : __prop(__property) {}
+      __host__ __device__ constexpr __annotated_ptr_base(access_property __property) noexcept
+        : __annotated_ptr_base(static_cast<std::uint64_t>(__property)) {}
+      constexpr __annotated_ptr_base(__annotated_ptr_base const&) = default;
+      _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 __annotated_ptr_base& operator=(const __annotated_ptr_base&) = default;
+      inline __device__ void* __apply_prop(void* __p) const {
+        return __associate(__p, __prop);
+      }
+      __host__ __device__ access_property __get_property() const noexcept {
+	return reinterpret_cast<access_property&>(const_cast<std::uint64_t&>(__prop));
+      }
+  };
+} // namespace __detail_ap
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__config b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__config
new file mode 100644
index 000000000000..039911f2e2ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__config
@@ -0,0 +1,196 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __cuda_std__
+#define __cuda_std__
+
+#ifdef __CUDACC__
+    #if defined(__clang__)
+        #include <cuda_fp16.h>
+        #define __fp16 __half
+    #endif
+    #if defined(__FLT16_MANT_DIG__)
+        #include <cuda_fp16.h>
+        #define _Float16 __half
+    #endif
+    #define _LIBCUDACXX_CUDACC_VER_MAJOR __CUDACC_VER_MAJOR__
+    #define _LIBCUDACXX_CUDACC_VER_MINOR __CUDACC_VER_MINOR__
+    #define _LIBCUDACXX_CUDACC_VER_BUILD __CUDACC_VER_BUILD__
+    #define _LIBCUDACXX_CUDACC_VER                                                  \
+        _LIBCUDACXX_CUDACC_VER_MAJOR * 100000 + _LIBCUDACXX_CUDACC_VER_MINOR * 1000 + \
+        _LIBCUDACXX_CUDACC_VER_BUILD
+
+    #define _LIBCUDACXX_HAS_NO_LONG_DOUBLE
+#else
+    #ifndef __host__
+        #define __host__
+    #endif
+    #ifndef __device__
+        #define __device__
+    #endif
+    #ifndef __forceinline__
+        #define __forceinline__
+    #endif
+#endif
+
+#ifdef _MSC_VER
+    #undef __cpp_lib_transparent_operators
+#endif
+
+// request these outcomes
+#define _LIBCUDACXX_NO_AUTO_LINK
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    #define _LIBCUDACXX_NO_EXCEPTIONS
+#endif
+#ifndef _LIBCUDACXX_NO_RTTI
+    #define _LIBCUDACXX_NO_RTTI
+#endif
+// Disable the nodebug type
+#ifndef _LIBCUDACXX_NODEBUG_TYPE
+    #define _LIBCUDACXX_NODEBUG_TYPE
+#endif
+
+#if defined(__CUDACC_RTC__)
+#  if defined(__CUDACC_RTC_INT128__)
+#    define __SIZEOF_INT128__ 16 // This macro is required in order to use int128 within the library
+#  else
+#    define _LIBCUDACXX_HAS_NO_INT128
+#  endif
+#endif
+
+#if defined(_LIBCUDACXX_COMPILER_MSVC) || (defined(_LIBCUDACXX_CUDACC_VER) && (_LIBCUDACXX_CUDACC_VER < 1105000))
+#  define _LIBCUDACXX_HAS_NO_INT128
+#endif
+
+#ifndef _LIBCUDACXX_ASSERT
+#   define _LIBCUDACXX_ASSERT(x, m) ((void)0)
+#endif
+#define _LIBCUDACXX_FREESTANDING
+#define _LIBCUDACXX_HAS_NO_WCHAR_H
+#define _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+#define _LIBCUDACXX_HAS_NO_PLATFORM_WAIT
+#define _LIBCUDACXX_HAS_NO_MONOTONIC_CLOCK
+#define _LIBCUDACXX_HAS_NO_TREE_BARRIER
+#ifdef __CUDACC_RTC__
+    #define __ELF__
+    #define _LIBCUDACXX_DISABLE_PRAGMA_GCC_SYSTEM_HEADER
+    #define _LIBCUDACXX_HAS_THREAD_API_EXTERNAL
+    #define __alignof(x) alignof(x)
+    #define _LIBCUDACXX_LITTLE_ENDIAN
+    #define _LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS
+    #define _LIBCUDACXX_HAS_NO_PRAGMA_PUSH_POP_MACRO
+#endif
+#define _LIBCUDACXX_HAS_EXTERNAL_ATOMIC_IMP
+#define _LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE(size, ptr) (size <= 8)
+#define _LIBCUDACXX_HAS_NO_CXX20_CHRONO_LITERALS
+
+#define _LIBCUDACXX_SYS_CLOCK_DURATION nanoseconds
+
+#if defined(__PGIC__) && defined(__linux__)
+    #define __ELF__
+#endif
+
+#define _LIBCUDACXX_ENABLE_BIT_BACKPORT
+
+#define _LIBCUDACXX_HAS_CUDA_ATOMIC_EXT
+
+#include "libcxx/include/__config"
+
+#if defined(__CUDA_ARCH__)
+    #define _LIBCUDACXX_HAS_THREAD_API_CUDA
+#elif defined(_LIBCUDACXX_COMPILER_MSVC)
+    #define _LIBCUDACXX_HAS_THREAD_API_WIN32
+#endif
+
+#if _GNUC_VER <= 409
+    #define _LIBCUDACXX_CUDA_HAS_NO_HOST_STD_ATOMIC_INIT
+#endif
+
+// force this outcome
+#undef _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+#define _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER __host__ __device__
+#undef _LIBCUDACXX_ATOMIC_FLAG_TYPE
+#define _LIBCUDACXX_ATOMIC_FLAG_TYPE int
+#undef _LIBCUDACXX_INLINE_VISIBILITY
+#define _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+#undef _LIBCUDACXX_FUNC_VIS
+#define _LIBCUDACXX_FUNC_VIS _LIBCUDACXX_INLINE_VISIBILITY
+#undef _LIBCUDACXX_TYPE_VIS
+#define _LIBCUDACXX_TYPE_VIS
+
+#ifndef _LIBCUDACXX_ABI_UNSTABLE
+#  define _LIBCUDACXX_ABI_UNSTABLE
+#endif
+
+#define _LIBCUDACXX_CUDA_API_VERSION 1008001
+
+#define _LIBCUDACXX_CUDA_API_VERSION_MAJOR \
+  (_LIBCUDACXX_CUDA_API_VERSION / 1000000)
+
+#define _LIBCUDACXX_CUDA_API_VERSION_MINOR \
+  (_LIBCUDACXX_CUDA_API_VERSION / 1000 % 1000)
+
+#define _LIBCUDACXX_CUDA_API_VERSION_PATCH \
+  (_LIBCUDACXX_CUDA_API_VERSION % 1000)
+
+#ifndef _LIBCUDACXX_CUDA_ABI_VERSION_LATEST
+#  define _LIBCUDACXX_CUDA_ABI_VERSION_LATEST 4
+#endif
+
+#ifdef _LIBCUDACXX_CUDA_ABI_VERSION
+#  if _LIBCUDACXX_CUDA_ABI_VERSION != 2 && _LIBCUDACXX_CUDA_ABI_VERSION != 3 && _LIBCUDACXX_CUDA_ABI_VERSION != 4
+#    error Unsupported libcu++ ABI version requested. Please define _LIBCUDACXX_CUDA_ABI_VERSION to either 2 or 3.
+#  endif
+#else
+#  define _LIBCUDACXX_CUDA_ABI_VERSION _LIBCUDACXX_CUDA_ABI_VERSION_LATEST
+#endif
+
+#ifdef _LIBCUDACXX_PIPELINE_ASSUMED_ABI_VERSION
+#  if _LIBCUDACXX_PIPELINE_ASSUMED_ABI_VERSION != _LIBCUDACXX_CUDA_ABI_VERSION
+#    error cuda_pipeline.h has assumed a different libcu++ ABI version than provided by this library. To fix this, please include a libcu++ header before including cuda_pipeline.h, or upgrade to a version of the toolkit this version of libcu++ shipped in.
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_CUDA_ABI_NAMESPACE
+#  define _LIBCUDACXX_CUDA_ABI_NAMESPACE _LIBCUDACXX_CONCAT(__,_LIBCUDACXX_CUDA_ABI_VERSION)
+#endif
+
+#ifndef _LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+#  define _LIBCUDACXX_BEGIN_NAMESPACE_CUDA \
+      namespace cuda { inline namespace _LIBCUDACXX_CUDA_ABI_NAMESPACE {
+#endif
+#ifndef _LIBCUDACXX_END_NAMESPACE_CUDA
+#  define _LIBCUDACXX_END_NAMESPACE_CUDA } }
+#endif
+
+#ifndef _LIBCUDACXX_BEGIN_NAMESPACE_CUDA_DEVICE
+#  define _LIBCUDACXX_BEGIN_NAMESPACE_CUDA_DEVICE \
+     namespace cuda { namespace device { inline namespace _LIBCUDACXX_CUDA_ABI_NAMESPACE {
+#endif
+#ifndef _LIBCUDACXX_END_NAMESPACE_CUDA_DEVICE
+#  define _LIBCUDACXX_END_NAMESPACE_CUDA_DEVICE } } }
+#endif
+
+// redefine namespace std::
+#undef _LIBCUDACXX_BEGIN_NAMESPACE_STD
+#define _LIBCUDACXX_BEGIN_NAMESPACE_STD \
+   namespace cuda { namespace std { inline namespace _LIBCUDACXX_CUDA_ABI_NAMESPACE {
+
+#undef _LIBCUDACXX_END_NAMESPACE_STD
+#define _LIBCUDACXX_END_NAMESPACE_STD } } }
+
+#undef _CUDA_VSTD
+#define _CUDA_VSTD cuda::std::_LIBCUDACXX_CUDA_ABI_NAMESPACE
+
+#undef _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+#define _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION namespace cuda { namespace std {
+#undef _LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+#define _LIBCUDACXX_END_NAMESPACE_STD_NOVERSION } }
+
+#endif //__cuda_std__
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__functional_base b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__functional_base
new file mode 100644
index 000000000000..731645d0066b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__functional_base
@@ -0,0 +1,25 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA___FUNCTIONAL_BASE
+#define _CUDA___FUNCTIONAL_BASE
+
+#ifndef __CUDACC_RTC__
+    #include <errno.h>
+#endif
+
+#include "../chrono"
+#include "../climits"
+
+#include "__config"
+
+#include "libcxx/include/__functional_base"
+
+#endif //_CUDA___FUNCTIONAL_BASE
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_pop b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_pop
new file mode 100644
index 000000000000..a40f5f2bae21
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_pop
@@ -0,0 +1,11 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "libcxx/include/__pragma_pop"
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_push b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_push
new file mode 100644
index 000000000000..e2aae6e48ccb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__pragma_push
@@ -0,0 +1,12 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "libcxx/include/__pragma_push"
+#include "libcxx/include/__undef_macros"
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__threading_support b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__threading_support
new file mode 100644
index 000000000000..0216c5e8f348
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/__threading_support
@@ -0,0 +1,26 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA___THREADING_SUPPORT
+#define _CUDA___THREADING_SUPPORT
+
+#ifndef __CUDACC_RTC__
+    #include <thread>
+    #include <errno.h>
+#endif
+
+
+#include "../chrono"
+#include "../climits"
+
+#include "__config"
+
+#include "libcxx/include/__threading_support"
+
+#endif //_CUDA___THREADING_SUPPORT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt
new file mode 100644
index 000000000000..d59bf2b1c1ed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/CMakeLists.txt
@@ -0,0 +1,277 @@
+set(files
+  __bit_reference
+  __bsd_locale_defaults.h
+  __bsd_locale_fallbacks.h
+  __errc
+  __debug
+  __functional_03
+  __functional_base
+  __functional_base_03
+  __hash_table
+  __libcpp_version
+  __locale
+  __mutex_base
+  __node_handle
+  __nullptr
+  __split_buffer
+  __sso_allocator
+  __std_stream
+  __string
+  __threading_support
+  __tree
+  __tuple
+  __undef_macros
+  algorithm
+  any
+  array
+  atomic
+  barrier
+  bit
+  bitset
+  cassert
+  ccomplex
+  cctype
+  cerrno
+  cfenv
+  cfloat
+  charconv
+  chrono
+  cinttypes
+  ciso646
+  climits
+  clocale
+  cmath
+  codecvt
+  compare
+  complex
+  complex.h
+  condition_variable
+  csetjmp
+  csignal
+  cstdarg
+  cstdbool
+  cstddef
+  cstdint
+  cstdio
+  cstdlib
+  cstring
+  ctgmath
+  ctime
+  ctype.h
+  cwchar
+  cwctype
+  deque
+  errno.h
+  exception
+  execution
+  experimental/__config
+  experimental/__memory
+  experimental/algorithm
+  experimental/coroutine
+  experimental/deque
+  experimental/filesystem
+  experimental/forward_list
+  experimental/functional
+  experimental/iterator
+  experimental/list
+  experimental/map
+  experimental/memory_resource
+  experimental/propagate_const
+  experimental/regex
+  experimental/set
+  experimental/simd
+  experimental/string
+  experimental/type_traits
+  experimental/unordered_map
+  experimental/unordered_set
+  experimental/utility
+  experimental/vector
+  ext/__hash
+  ext/hash_map
+  ext/hash_set
+  fenv.h
+  filesystem
+  float.h
+  forward_list
+  fstream
+  functional
+  future
+  initializer_list
+  inttypes.h
+  iomanip
+  ios
+  iosfwd
+  iostream
+  istream
+  iterator
+  latch
+  limits
+  limits.h
+  list
+  locale
+  locale.h
+  map
+  math.h
+  memory
+  module.modulemap
+  mutex
+  new
+  numeric
+  optional
+  ostream
+  queue
+  random
+  ratio
+  regex
+  scoped_allocator
+  semaphore
+  set
+  setjmp.h
+  shared_mutex
+  span
+  sstream
+  stack
+  stdbool.h
+  stddef.h
+  stdexcept
+  stdint.h
+  stdio.h
+  stdlib.h
+  streambuf
+  string
+  string.h
+  string_view
+  strstream
+  system_error
+  tgmath.h
+  thread
+  tuple
+  type_traits
+  typeindex
+  typeinfo
+  unordered_map
+  unordered_set
+  utility
+  valarray
+  variant
+  vector
+  version
+  wchar.h
+  wctype.h
+  )
+
+if(LIBCXX_INSTALL_SUPPORT_HEADERS)
+  set(files
+    ${files}
+    support/android/locale_bionic.h
+    support/fuchsia/xlocale.h
+    support/ibm/limits.h
+    support/ibm/locale_mgmt_aix.h
+    support/ibm/support.h
+    support/ibm/xlocale.h
+    support/musl/xlocale.h
+    support/newlib/xlocale.h
+    support/solaris/floatingpoint.h
+    support/solaris/wchar.h
+    support/solaris/xlocale.h
+    support/win32/limits_msvc_win32.h
+    support/win32/locale_win32.h
+    support/xlocale/__nop_locale_mgmt.h
+    support/xlocale/__posix_l_fallback.h
+    support/xlocale/__strtonum_fallback.h
+    )
+endif()
+
+if (LIBCXX_NEEDS_SITE_CONFIG)
+  # Generate a custom __config header. The new header is created
+  # by prepending __config_site to the current __config header.
+  add_custom_command(OUTPUT ${LIBCXX_BINARY_DIR}/__generated_config
+    COMMAND ${PYTHON_EXECUTABLE} ${LIBCXX_SOURCE_DIR}/utils/cat_files.py
+      ${LIBCXX_BINARY_DIR}/__config_site
+      ${LIBCXX_SOURCE_DIR}/include/__config
+      -o ${LIBCXX_BINARY_DIR}/__generated_config
+    DEPENDS ${LIBCXX_SOURCE_DIR}/include/__config
+            ${LIBCXX_BINARY_DIR}/__config_site
+  )
+  # Add a target that executes the generation commands.
+  add_custom_target(cxx-generated-config ALL
+    DEPENDS ${LIBCXX_BINARY_DIR}/__generated_config)
+  set(generated_config_deps cxx-generated-config)
+else()
+  set(files
+    ${files}
+    __config
+    )
+endif()
+
+# In some build configuraitons (like boostrapping clang), we need to be able to
+# install the libcxx headers before CMake configuraiton for libcxx runs. Making
+# the name of this target configurable allows LLVM/runtimes/CMakeLists.txt to
+# add this subdirectory to the LLVM build to put libcxx's headers in place
+# before libcxx's build configuration is run.
+if (NOT CXX_HEADER_TARGET)
+  set(CXX_HEADER_TARGET cxx-headers)
+endif()
+if(NOT LIBCXX_USING_INSTALLED_LLVM AND LIBCXX_HEADER_DIR)
+  set(output_dir ${LIBCXX_HEADER_DIR}/include/c++/v1)
+
+  set(out_files)
+  foreach(f ${files})
+    set(src ${CMAKE_CURRENT_SOURCE_DIR}/${f})
+    set(dst ${output_dir}/${f})
+    add_custom_command(OUTPUT ${dst}
+      DEPENDS ${src}
+      COMMAND ${CMAKE_COMMAND} -E copy_if_different ${src} ${dst}
+      COMMENT "Copying CXX header ${f}")
+    list(APPEND out_files ${dst})
+  endforeach()
+
+  if (LIBCXX_NEEDS_SITE_CONFIG)
+    # Copy the generated header as __config into build directory.
+    set(src ${LIBCXX_BINARY_DIR}/__generated_config)
+    set(dst ${output_dir}/__config)
+    add_custom_command(OUTPUT ${dst}
+        DEPENDS ${src} ${generated_config_deps}
+        COMMAND ${CMAKE_COMMAND} -E copy_if_different ${src} ${dst}
+        COMMENT "Copying CXX __config")
+    list(APPEND out_files ${dst})
+  endif()
+
+  add_custom_target(${CXX_HEADER_TARGET} ALL DEPENDS ${out_files} ${LIBCXX_CXX_ABI_HEADER_TARGET})
+else()
+  add_custom_target(${CXX_HEADER_TARGET})
+endif()
+set_target_properties(${CXX_HEADER_TARGET} PROPERTIES FOLDER "Misc")
+
+if (LIBCXX_INSTALL_HEADERS)
+  foreach(file ${files})
+    get_filename_component(dir ${file} DIRECTORY)
+    install(FILES ${file}
+      DESTINATION ${LIBCXX_INSTALL_HEADER_PREFIX}include/c++/v1/${dir}
+      COMPONENT ${CXX_HEADER_TARGET}
+      PERMISSIONS OWNER_READ OWNER_WRITE GROUP_READ WORLD_READ
+    )
+  endforeach()
+
+  if (LIBCXX_NEEDS_SITE_CONFIG)
+    # Install the generated header as __config.
+    install(FILES ${LIBCXX_BINARY_DIR}/__generated_config
+      DESTINATION ${LIBCXX_INSTALL_HEADER_PREFIX}include/c++/v1
+      PERMISSIONS OWNER_READ OWNER_WRITE GROUP_READ WORLD_READ
+      RENAME __config
+      COMPONENT ${CXX_HEADER_TARGET})
+  endif()
+
+  if (NOT CMAKE_CONFIGURATION_TYPES)
+    add_custom_target(install-${CXX_HEADER_TARGET}
+                      DEPENDS ${CXX_HEADER_TARGET} ${generated_config_deps}
+                      COMMAND "${CMAKE_COMMAND}"
+                              -DCMAKE_INSTALL_COMPONENT=${CXX_HEADER_TARGET}
+                              -P "${CMAKE_BINARY_DIR}/cmake_install.cmake")
+    # Stripping is a no-op for headers
+    add_custom_target(install-${CXX_HEADER_TARGET}-stripped DEPENDS install-${CXX_HEADER_TARGET})
+
+    add_custom_target(install-libcxx-headers DEPENDS install-${CXX_HEADER_TARGET})
+    add_custom_target(install-libcxx-headers-stripped DEPENDS install-${CXX_HEADER_TARGET}-stripped)
+  endif()
+endif()
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference
new file mode 100644
index 000000000000..851cbeacfc59
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bit_reference
@@ -0,0 +1,1289 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___BIT_REFERENCE
+#define _LIBCUDACXX___BIT_REFERENCE
+
+#include <__config>
+#include <bit>
+#include <algorithm>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Cp, bool _IsConst, typename _Cp::__storage_type = 0> class __bit_iterator;
+template <class _Cp> class __bit_const_reference;
+
+template <class _Tp>
+struct __has_storage_type
+{
+    static const bool value = false;
+};
+
+template <class _Cp, bool = __has_storage_type<_Cp>::value>
+class __bit_reference
+{
+    typedef typename _Cp::__storage_type    __storage_type;
+    typedef typename _Cp::__storage_pointer __storage_pointer;
+
+    __storage_pointer __seg_;
+    __storage_type    __mask_;
+
+    friend typename _Cp::__self;
+
+    friend class __bit_const_reference<_Cp>;
+    friend class __bit_iterator<_Cp, false>;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_reference(const __bit_reference&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY operator bool() const _NOEXCEPT
+        {return static_cast<bool>(*__seg_ & __mask_);}
+    _LIBCUDACXX_INLINE_VISIBILITY bool operator ~() const _NOEXCEPT
+        {return !static_cast<bool>(*this);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_reference& operator=(bool __x) _NOEXCEPT
+    {
+        if (__x)
+            *__seg_ |= __mask_;
+        else
+            *__seg_ &= ~__mask_;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_reference& operator=(const __bit_reference& __x) _NOEXCEPT
+        {return operator=(static_cast<bool>(__x));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void flip() _NOEXCEPT {*__seg_ ^= __mask_;}
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator<_Cp, false> operator&() const _NOEXCEPT
+        {return __bit_iterator<_Cp, false>(__seg_, static_cast<unsigned>(__libcpp_ctz(__mask_)));}
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_reference(__storage_pointer __s, __storage_type __m) _NOEXCEPT
+        : __seg_(__s), __mask_(__m) {}
+};
+
+template <class _Cp>
+class __bit_reference<_Cp, false>
+{
+};
+
+template <class _Cp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__bit_reference<_Cp> __x, __bit_reference<_Cp> __y) _NOEXCEPT
+{
+    bool __t = __x;
+    __x = __y;
+    __y = __t;
+}
+
+template <class _Cp, class _Dp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__bit_reference<_Cp> __x, __bit_reference<_Dp> __y) _NOEXCEPT
+{
+    bool __t = __x;
+    __x = __y;
+    __y = __t;
+}
+
+template <class _Cp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__bit_reference<_Cp> __x, bool& __y) _NOEXCEPT
+{
+    bool __t = __x;
+    __x = __y;
+    __y = __t;
+}
+
+template <class _Cp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(bool& __x, __bit_reference<_Cp> __y) _NOEXCEPT
+{
+    bool __t = __x;
+    __x = __y;
+    __y = __t;
+}
+
+template <class _Cp>
+class __bit_const_reference
+{
+    typedef typename _Cp::__storage_type          __storage_type;
+    typedef typename _Cp::__const_storage_pointer __storage_pointer;
+
+    __storage_pointer        __seg_;
+    __storage_type __mask_;
+
+    friend typename _Cp::__self;
+    friend class __bit_iterator<_Cp, true>;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_const_reference(const __bit_const_reference&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_const_reference(const __bit_reference<_Cp>& __x) _NOEXCEPT
+        : __seg_(__x.__seg_), __mask_(__x.__mask_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR operator bool() const _NOEXCEPT
+        {return static_cast<bool>(*__seg_ & __mask_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator<_Cp, true> operator&() const _NOEXCEPT
+        {return __bit_iterator<_Cp, true>(__seg_, static_cast<unsigned>(__libcpp_ctz(__mask_)));}
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR
+    __bit_const_reference(__storage_pointer __s, __storage_type __m) _NOEXCEPT
+        : __seg_(__s), __mask_(__m) {}
+
+    __bit_const_reference& operator=(const __bit_const_reference&) = delete;
+};
+
+// find
+
+template <class _Cp, bool _IsConst>
+__bit_iterator<_Cp, _IsConst>
+__find_bool_true(__bit_iterator<_Cp, _IsConst> __first, typename _Cp::size_type __n)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _It;
+    typedef typename _It::__storage_type __storage_type;
+    static const int __bits_per_word = _It::__bits_per_word;
+    // do first partial word
+    if (__first.__ctz_ != 0)
+    {
+        __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+        __storage_type __dn = _CUDA_VSTD::min(__clz_f, __n);
+        __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+        __storage_type __b = *__first.__seg_ & __m;
+        if (__b)
+            return _It(__first.__seg_, static_cast<unsigned>(_CUDA_VSTD::__libcpp_ctz(__b)));
+        if (__n == __dn)
+            return __first + __n;
+        __n -= __dn;
+        ++__first.__seg_;
+    }
+    // do middle whole words
+    for (; __n >= __bits_per_word; ++__first.__seg_, __n -= __bits_per_word)
+        if (*__first.__seg_)
+            return _It(__first.__seg_, static_cast<unsigned>(_CUDA_VSTD::__libcpp_ctz(*__first.__seg_)));
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __storage_type __b = *__first.__seg_ & __m;
+        if (__b)
+            return _It(__first.__seg_, static_cast<unsigned>(_CUDA_VSTD::__libcpp_ctz(__b)));
+    }
+    return _It(__first.__seg_, static_cast<unsigned>(__n));
+}
+
+template <class _Cp, bool _IsConst>
+__bit_iterator<_Cp, _IsConst>
+__find_bool_false(__bit_iterator<_Cp, _IsConst> __first, typename _Cp::size_type __n)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _It;
+    typedef typename _It::__storage_type __storage_type;
+    const int __bits_per_word = _It::__bits_per_word;
+    // do first partial word
+    if (__first.__ctz_ != 0)
+    {
+        __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+        __storage_type __dn = _CUDA_VSTD::min(__clz_f, __n);
+        __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+        __storage_type __b = ~*__first.__seg_ & __m;
+        if (__b)
+            return _It(__first.__seg_, static_cast<unsigned>(_CUDA_VSTD::__libcpp_ctz(__b)));
+        if (__n == __dn)
+            return __first + __n;
+        __n -= __dn;
+        ++__first.__seg_;
+    }
+    // do middle whole words
+    for (; __n >= __bits_per_word; ++__first.__seg_, __n -= __bits_per_word)
+    {
+        __storage_type __b = ~*__first.__seg_;
+        if (__b)
+            return _It(__first.__seg_, static_cast<unsigned>(_CUDA_VSTD::__libcpp_ctz(__b)));
+    }
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __storage_type __b = ~*__first.__seg_ & __m;
+        if (__b)
+            return _It(__first.__seg_, static_cast<unsigned>(_CUDA_VSTD::__libcpp_ctz(__b)));
+    }
+    return _It(__first.__seg_, static_cast<unsigned>(__n));
+}
+
+template <class _Cp, bool _IsConst, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bit_iterator<_Cp, _IsConst>
+find(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, const _Tp& __value_)
+{
+    if (static_cast<bool>(__value_))
+        return __find_bool_true(__first, static_cast<typename _Cp::size_type>(__last - __first));
+    return __find_bool_false(__first, static_cast<typename _Cp::size_type>(__last - __first));
+}
+
+// count
+
+template <class _Cp, bool _IsConst>
+typename __bit_iterator<_Cp, _IsConst>::difference_type
+__count_bool_true(__bit_iterator<_Cp, _IsConst> __first, typename _Cp::size_type __n)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _It;
+    typedef typename _It::__storage_type __storage_type;
+    typedef typename _It::difference_type difference_type;
+    const int __bits_per_word = _It::__bits_per_word;
+    difference_type __r = 0;
+    // do first partial word
+    if (__first.__ctz_ != 0)
+    {
+        __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+        __storage_type __dn = _CUDA_VSTD::min(__clz_f, __n);
+        __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+        __r = _CUDA_VSTD::__libcpp_popcount(*__first.__seg_ & __m);
+        __n -= __dn;
+        ++__first.__seg_;
+    }
+    // do middle whole words
+    for (; __n >= __bits_per_word; ++__first.__seg_, __n -= __bits_per_word)
+        __r += _CUDA_VSTD::__libcpp_popcount(*__first.__seg_);
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __r += _CUDA_VSTD::__libcpp_popcount(*__first.__seg_ & __m);
+    }
+    return __r;
+}
+
+template <class _Cp, bool _IsConst>
+typename __bit_iterator<_Cp, _IsConst>::difference_type
+__count_bool_false(__bit_iterator<_Cp, _IsConst> __first, typename _Cp::size_type __n)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _It;
+    typedef typename _It::__storage_type __storage_type;
+    typedef typename _It::difference_type difference_type;
+    const int __bits_per_word = _It::__bits_per_word;
+    difference_type __r = 0;
+    // do first partial word
+    if (__first.__ctz_ != 0)
+    {
+        __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+        __storage_type __dn = _CUDA_VSTD::min(__clz_f, __n);
+        __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+        __r = _CUDA_VSTD::__libcpp_popcount(~*__first.__seg_ & __m);
+        __n -= __dn;
+        ++__first.__seg_;
+    }
+    // do middle whole words
+    for (; __n >= __bits_per_word; ++__first.__seg_, __n -= __bits_per_word)
+        __r += _CUDA_VSTD::__libcpp_popcount(~*__first.__seg_);
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __r += _CUDA_VSTD::__libcpp_popcount(~*__first.__seg_ & __m);
+    }
+    return __r;
+}
+
+template <class _Cp, bool _IsConst, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __bit_iterator<_Cp, _IsConst>::difference_type
+count(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, const _Tp& __value_)
+{
+    if (static_cast<bool>(__value_))
+        return __count_bool_true(__first, static_cast<typename _Cp::size_type>(__last - __first));
+    return __count_bool_false(__first, static_cast<typename _Cp::size_type>(__last - __first));
+}
+
+// fill_n
+
+template <class _Cp>
+void
+__fill_n_false(__bit_iterator<_Cp, false> __first, typename _Cp::size_type __n)
+{
+    typedef __bit_iterator<_Cp, false> _It;
+    typedef typename _It::__storage_type __storage_type;
+    const int __bits_per_word = _It::__bits_per_word;
+    // do first partial word
+    if (__first.__ctz_ != 0)
+    {
+        __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+        __storage_type __dn = _CUDA_VSTD::min(__clz_f, __n);
+        __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+        *__first.__seg_ &= ~__m;
+        __n -= __dn;
+        ++__first.__seg_;
+    }
+    // do middle whole words
+    __storage_type __nw = __n / __bits_per_word;
+    _CUDA_VSTD::memset(_CUDA_VSTD::__to_raw_pointer(__first.__seg_), 0, __nw * sizeof(__storage_type));
+    __n -= __nw * __bits_per_word;
+    // do last partial word
+    if (__n > 0)
+    {
+        __first.__seg_ += __nw;
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        *__first.__seg_ &= ~__m;
+    }
+}
+
+template <class _Cp>
+void
+__fill_n_true(__bit_iterator<_Cp, false> __first, typename _Cp::size_type __n)
+{
+    typedef __bit_iterator<_Cp, false> _It;
+    typedef typename _It::__storage_type __storage_type;
+    const int __bits_per_word = _It::__bits_per_word;
+    // do first partial word
+    if (__first.__ctz_ != 0)
+    {
+        __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
+        __storage_type __dn = _CUDA_VSTD::min(__clz_f, __n);
+        __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+        *__first.__seg_ |= __m;
+        __n -= __dn;
+        ++__first.__seg_;
+    }
+    // do middle whole words
+    __storage_type __nw = __n / __bits_per_word;
+    _CUDA_VSTD::memset(_CUDA_VSTD::__to_raw_pointer(__first.__seg_), -1, __nw * sizeof(__storage_type));
+    __n -= __nw * __bits_per_word;
+    // do last partial word
+    if (__n > 0)
+    {
+        __first.__seg_ += __nw;
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        *__first.__seg_ |= __m;
+    }
+}
+
+template <class _Cp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+fill_n(__bit_iterator<_Cp, false> __first, typename _Cp::size_type __n, bool __value_)
+{
+    if (__n > 0)
+    {
+        if (__value_)
+            __fill_n_true(__first, __n);
+        else
+            __fill_n_false(__first, __n);
+    }
+}
+
+// fill
+
+template <class _Cp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+fill(__bit_iterator<_Cp, false> __first, __bit_iterator<_Cp, false> __last, bool __value_)
+{
+    _CUDA_VSTD::fill_n(__first, static_cast<typename _Cp::size_type>(__last - __first), __value_);
+}
+
+// copy
+
+template <class _Cp, bool _IsConst>
+__bit_iterator<_Cp, false>
+__copy_aligned(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last,
+                                                     __bit_iterator<_Cp, false> __result)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _In;
+    typedef  typename _In::difference_type difference_type;
+    typedef typename _In::__storage_type __storage_type;
+    const int __bits_per_word = _In::__bits_per_word;
+    difference_type __n = __last - __first;
+    if (__n > 0)
+    {
+        // do first word
+        if (__first.__ctz_ != 0)
+        {
+            unsigned __clz = __bits_per_word - __first.__ctz_;
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__clz), __n);
+            __n -= __dn;
+            __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz - __dn));
+            __storage_type __b = *__first.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b;
+            __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+            __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_)  % __bits_per_word);
+            ++__first.__seg_;
+            // __first.__ctz_ = 0;
+        }
+        // __first.__ctz_ == 0;
+        // do middle words
+        __storage_type __nw = __n / __bits_per_word;
+        _CUDA_VSTD::memmove(_CUDA_VSTD::__to_raw_pointer(__result.__seg_),
+                       _CUDA_VSTD::__to_raw_pointer(__first.__seg_),
+                       __nw * sizeof(__storage_type));
+        __n -= __nw * __bits_per_word;
+        __result.__seg_ += __nw;
+        // do last word
+        if (__n > 0)
+        {
+            __first.__seg_ += __nw;
+            __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+            __storage_type __b = *__first.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b;
+            __result.__ctz_ = static_cast<unsigned>(__n);
+        }
+    }
+    return __result;
+}
+
+template <class _Cp, bool _IsConst>
+__bit_iterator<_Cp, false>
+__copy_unaligned(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last,
+                                                       __bit_iterator<_Cp, false> __result)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _In;
+    typedef  typename _In::difference_type difference_type;
+    typedef typename _In::__storage_type __storage_type;
+    static const int __bits_per_word = _In::__bits_per_word;
+    difference_type __n = __last - __first;
+    if (__n > 0)
+    {
+        // do first word
+        if (__first.__ctz_ != 0)
+        {
+            unsigned __clz_f = __bits_per_word - __first.__ctz_;
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__clz_f), __n);
+            __n -= __dn;
+            __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+            __storage_type __b = *__first.__seg_ & __m;
+            unsigned __clz_r = __bits_per_word - __result.__ctz_;
+            __storage_type __ddn = _CUDA_VSTD::min<__storage_type>(__dn, __clz_r);
+            __m = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __ddn));
+            *__result.__seg_ &= ~__m;
+            if (__result.__ctz_ > __first.__ctz_)
+                *__result.__seg_ |= __b << (__result.__ctz_ - __first.__ctz_);
+            else
+                *__result.__seg_ |= __b >> (__first.__ctz_ - __result.__ctz_);
+            __result.__seg_ += (__ddn + __result.__ctz_) / __bits_per_word;
+            __result.__ctz_ = static_cast<unsigned>((__ddn + __result.__ctz_)  % __bits_per_word);
+            __dn -= __ddn;
+            if (__dn > 0)
+            {
+                __m = ~__storage_type(0) >> (__bits_per_word - __dn);
+                *__result.__seg_ &= ~__m;
+                *__result.__seg_ |= __b >> (__first.__ctz_ + __ddn);
+                __result.__ctz_ = static_cast<unsigned>(__dn);
+            }
+            ++__first.__seg_;
+            // __first.__ctz_ = 0;
+        }
+        // __first.__ctz_ == 0;
+        // do middle words
+        unsigned __clz_r = __bits_per_word - __result.__ctz_;
+        __storage_type __m = ~__storage_type(0) << __result.__ctz_;
+        for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first.__seg_)
+        {
+            __storage_type __b = *__first.__seg_;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b << __result.__ctz_;
+            ++__result.__seg_;
+            *__result.__seg_ &= __m;
+            *__result.__seg_ |= __b >> __clz_r;
+        }
+        // do last word
+        if (__n > 0)
+        {
+            __m = ~__storage_type(0) >> (__bits_per_word - __n);
+            __storage_type __b = *__first.__seg_ & __m;
+            __storage_type __dn = _CUDA_VSTD::min(__n, static_cast<difference_type>(__clz_r));
+            __m = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __dn));
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b << __result.__ctz_;
+            __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+            __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_)  % __bits_per_word);
+            __n -= __dn;
+            if (__n > 0)
+            {
+                __m = ~__storage_type(0) >> (__bits_per_word - __n);
+                *__result.__seg_ &= ~__m;
+                *__result.__seg_ |= __b >> __dn;
+                __result.__ctz_ = static_cast<unsigned>(__n);
+            }
+        }
+    }
+    return __result;
+}
+
+template <class _Cp, bool _IsConst>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bit_iterator<_Cp, false>
+copy(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result)
+{
+    if (__first.__ctz_ == __result.__ctz_)
+        return __copy_aligned(__first, __last, __result);
+    return __copy_unaligned(__first, __last, __result);
+}
+
+// copy_backward
+
+template <class _Cp, bool _IsConst>
+__bit_iterator<_Cp, false>
+__copy_backward_aligned(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last,
+                                                     __bit_iterator<_Cp, false> __result)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _In;
+    typedef  typename _In::difference_type difference_type;
+    typedef typename _In::__storage_type __storage_type;
+    const int __bits_per_word = _In::__bits_per_word;
+    difference_type __n = __last - __first;
+    if (__n > 0)
+    {
+        // do first word
+        if (__last.__ctz_ != 0)
+        {
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__last.__ctz_), __n);
+            __n -= __dn;
+            unsigned __clz = __bits_per_word - __last.__ctz_;
+            __storage_type __m = (~__storage_type(0) << (__last.__ctz_ - __dn)) & (~__storage_type(0) >> __clz);
+            __storage_type __b = *__last.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b;
+            __result.__ctz_ = static_cast<unsigned>(((-__dn & (__bits_per_word - 1)) +
+                                                       __result.__ctz_)  % __bits_per_word);
+            // __last.__ctz_ = 0
+         }
+        // __last.__ctz_ == 0 || __n == 0
+        // __result.__ctz_ == 0 || __n == 0
+        // do middle words
+        __storage_type __nw = __n / __bits_per_word;
+        __result.__seg_ -= __nw;
+        __last.__seg_ -= __nw;
+        _CUDA_VSTD::memmove(_CUDA_VSTD::__to_raw_pointer(__result.__seg_),
+                       _CUDA_VSTD::__to_raw_pointer(__last.__seg_),
+                       __nw * sizeof(__storage_type));
+        __n -= __nw * __bits_per_word;
+        // do last word
+        if (__n > 0)
+        {
+            __storage_type __m = ~__storage_type(0) << (__bits_per_word - __n);
+            __storage_type __b = *--__last.__seg_ & __m;
+            *--__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b;
+            __result.__ctz_ = static_cast<unsigned>(-__n & (__bits_per_word - 1));
+        }
+    }
+    return __result;
+}
+
+template <class _Cp, bool _IsConst>
+__bit_iterator<_Cp, false>
+__copy_backward_unaligned(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last,
+                                                       __bit_iterator<_Cp, false> __result)
+{
+    typedef __bit_iterator<_Cp, _IsConst> _In;
+    typedef  typename _In::difference_type difference_type;
+    typedef typename _In::__storage_type __storage_type;
+    const int __bits_per_word = _In::__bits_per_word;
+    difference_type __n = __last - __first;
+    if (__n > 0)
+    {
+        // do first word
+        if (__last.__ctz_ != 0)
+        {
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__last.__ctz_), __n);
+            __n -= __dn;
+            unsigned __clz_l = __bits_per_word - __last.__ctz_;
+            __storage_type __m = (~__storage_type(0) << (__last.__ctz_ - __dn)) & (~__storage_type(0) >> __clz_l);
+            __storage_type __b = *__last.__seg_ & __m;
+            unsigned __clz_r = __bits_per_word - __result.__ctz_;
+            __storage_type __ddn = _CUDA_VSTD::min(__dn, static_cast<difference_type>(__result.__ctz_));
+            if (__ddn > 0)
+            {
+                __m = (~__storage_type(0) << (__result.__ctz_ - __ddn)) & (~__storage_type(0) >> __clz_r);
+                *__result.__seg_ &= ~__m;
+                if (__result.__ctz_ > __last.__ctz_)
+                    *__result.__seg_ |= __b << (__result.__ctz_ - __last.__ctz_);
+                else
+                    *__result.__seg_ |= __b >> (__last.__ctz_ - __result.__ctz_);
+                __result.__ctz_ = static_cast<unsigned>(((-__ddn & (__bits_per_word - 1)) +
+                                                         __result.__ctz_)  % __bits_per_word);
+                __dn -= __ddn;
+            }
+            if (__dn > 0)
+            {
+                // __result.__ctz_ == 0
+                --__result.__seg_;
+                __result.__ctz_ = static_cast<unsigned>(-__dn & (__bits_per_word - 1));
+                __m = ~__storage_type(0) << __result.__ctz_;
+                *__result.__seg_ &= ~__m;
+                __last.__ctz_ -= __dn + __ddn;
+                *__result.__seg_ |= __b << (__result.__ctz_ - __last.__ctz_);
+            }
+            // __last.__ctz_ = 0
+         }
+        // __last.__ctz_ == 0 || __n == 0
+        // __result.__ctz_ != 0 || __n == 0
+        // do middle words
+        unsigned __clz_r = __bits_per_word - __result.__ctz_;
+        __storage_type __m = ~__storage_type(0) >> __clz_r;
+        for (; __n >= __bits_per_word; __n -= __bits_per_word)
+        {
+            __storage_type __b = *--__last.__seg_;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b >> __clz_r;
+            *--__result.__seg_ &= __m;
+            *__result.__seg_ |= __b << __result.__ctz_;
+        }
+        // do last word
+        if (__n > 0)
+        {
+            __m = ~__storage_type(0) << (__bits_per_word - __n);
+            __storage_type __b = *--__last.__seg_ & __m;
+            __clz_r = __bits_per_word - __result.__ctz_;
+            __storage_type __dn = _CUDA_VSTD::min(__n, static_cast<difference_type>(__result.__ctz_));
+            __m = (~__storage_type(0) << (__result.__ctz_ - __dn)) & (~__storage_type(0) >> __clz_r);
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b >> (__bits_per_word - __result.__ctz_);
+            __result.__ctz_ = static_cast<unsigned>(((-__dn & (__bits_per_word - 1)) +
+                                                     __result.__ctz_)  % __bits_per_word);
+            __n -= __dn;
+            if (__n > 0)
+            {
+                // __result.__ctz_ == 0
+                --__result.__seg_;
+                __result.__ctz_ = static_cast<unsigned>(-__n & (__bits_per_word - 1));
+                __m = ~__storage_type(0) << __result.__ctz_;
+                *__result.__seg_ &= ~__m;
+                *__result.__seg_ |= __b << (__result.__ctz_ - (__bits_per_word - __n - __dn));
+            }
+        }
+    }
+    return __result;
+}
+
+template <class _Cp, bool _IsConst>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bit_iterator<_Cp, false>
+copy_backward(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result)
+{
+    if (__last.__ctz_ == __result.__ctz_)
+        return __copy_backward_aligned(__first, __last, __result);
+    return __copy_backward_unaligned(__first, __last, __result);
+}
+
+// move
+
+template <class _Cp, bool _IsConst>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bit_iterator<_Cp, false>
+move(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result)
+{
+    return _CUDA_VSTD::copy(__first, __last, __result);
+}
+
+// move_backward
+
+template <class _Cp, bool _IsConst>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bit_iterator<_Cp, false>
+move_backward(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, __bit_iterator<_Cp, false> __result)
+{
+    return _CUDA_VSTD::copy_backward(__first, __last, __result);
+}
+
+// swap_ranges
+
+template <class __C1, class __C2>
+__bit_iterator<__C2, false>
+__swap_ranges_aligned(__bit_iterator<__C1, false> __first, __bit_iterator<__C1, false> __last,
+                      __bit_iterator<__C2, false> __result)
+{
+    typedef __bit_iterator<__C1, false> _I1;
+    typedef  typename _I1::difference_type difference_type;
+    typedef typename _I1::__storage_type __storage_type;
+    const int __bits_per_word = _I1::__bits_per_word;
+    difference_type __n = __last - __first;
+    if (__n > 0)
+    {
+        // do first word
+        if (__first.__ctz_ != 0)
+        {
+            unsigned __clz = __bits_per_word - __first.__ctz_;
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__clz), __n);
+            __n -= __dn;
+            __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz - __dn));
+            __storage_type __b1 = *__first.__seg_ & __m;
+            *__first.__seg_ &= ~__m;
+            __storage_type __b2 = *__result.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b1;
+            *__first.__seg_  |= __b2;
+            __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+            __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_)  % __bits_per_word);
+            ++__first.__seg_;
+            // __first.__ctz_ = 0;
+        }
+        // __first.__ctz_ == 0;
+        // do middle words
+        for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first.__seg_, ++__result.__seg_)
+            swap(*__first.__seg_, *__result.__seg_);
+        // do last word
+        if (__n > 0)
+        {
+            __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+            __storage_type __b1 = *__first.__seg_ & __m;
+            *__first.__seg_ &= ~__m;
+            __storage_type __b2 = *__result.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b1;
+            *__first.__seg_  |= __b2;
+            __result.__ctz_ = static_cast<unsigned>(__n);
+        }
+    }
+    return __result;
+}
+
+template <class __C1, class __C2>
+__bit_iterator<__C2, false>
+__swap_ranges_unaligned(__bit_iterator<__C1, false> __first, __bit_iterator<__C1, false> __last,
+                        __bit_iterator<__C2, false> __result)
+{
+    typedef __bit_iterator<__C1, false> _I1;
+    typedef  typename _I1::difference_type difference_type;
+    typedef typename _I1::__storage_type __storage_type;
+    const int __bits_per_word = _I1::__bits_per_word;
+    difference_type __n = __last - __first;
+    if (__n > 0)
+    {
+        // do first word
+        if (__first.__ctz_ != 0)
+        {
+            unsigned __clz_f = __bits_per_word - __first.__ctz_;
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__clz_f), __n);
+            __n -= __dn;
+            __storage_type __m = (~__storage_type(0) << __first.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+            __storage_type __b1 = *__first.__seg_ & __m;
+            *__first.__seg_ &= ~__m;
+            unsigned __clz_r = __bits_per_word - __result.__ctz_;
+            __storage_type __ddn = _CUDA_VSTD::min<__storage_type>(__dn, __clz_r);
+            __m = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __ddn));
+            __storage_type __b2 = *__result.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            if (__result.__ctz_ > __first.__ctz_)
+            {
+                unsigned __s = __result.__ctz_ - __first.__ctz_;
+                *__result.__seg_ |= __b1 << __s;
+                *__first.__seg_  |= __b2 >> __s;
+            }
+            else
+            {
+                unsigned __s = __first.__ctz_ - __result.__ctz_;
+                *__result.__seg_ |= __b1 >> __s;
+                *__first.__seg_  |= __b2 << __s;
+            }
+            __result.__seg_ += (__ddn + __result.__ctz_) / __bits_per_word;
+            __result.__ctz_ = static_cast<unsigned>((__ddn + __result.__ctz_)  % __bits_per_word);
+            __dn -= __ddn;
+            if (__dn > 0)
+            {
+                __m = ~__storage_type(0) >> (__bits_per_word - __dn);
+                __b2 = *__result.__seg_ & __m;
+                *__result.__seg_ &= ~__m;
+                unsigned __s = __first.__ctz_ + __ddn;
+                *__result.__seg_ |= __b1 >> __s;
+                *__first.__seg_  |= __b2 << __s;
+                __result.__ctz_ = static_cast<unsigned>(__dn);
+            }
+            ++__first.__seg_;
+            // __first.__ctz_ = 0;
+        }
+        // __first.__ctz_ == 0;
+        // do middle words
+        __storage_type __m = ~__storage_type(0) << __result.__ctz_;
+        unsigned __clz_r = __bits_per_word - __result.__ctz_;
+        for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first.__seg_)
+        {
+            __storage_type __b1 = *__first.__seg_;
+            __storage_type __b2 = *__result.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b1 << __result.__ctz_;
+            *__first.__seg_  = __b2 >> __result.__ctz_;
+            ++__result.__seg_;
+            __b2 = *__result.__seg_ & ~__m;
+            *__result.__seg_ &= __m;
+            *__result.__seg_ |= __b1 >> __clz_r;
+            *__first.__seg_  |= __b2 << __clz_r;
+        }
+        // do last word
+        if (__n > 0)
+        {
+            __m = ~__storage_type(0) >> (__bits_per_word - __n);
+            __storage_type __b1 = *__first.__seg_ & __m;
+            *__first.__seg_ &= ~__m;
+            __storage_type __dn = _CUDA_VSTD::min<__storage_type>(__n, __clz_r);
+            __m = (~__storage_type(0) << __result.__ctz_) & (~__storage_type(0) >> (__clz_r - __dn));
+            __storage_type __b2 = *__result.__seg_ & __m;
+            *__result.__seg_ &= ~__m;
+            *__result.__seg_ |= __b1 << __result.__ctz_;
+            *__first.__seg_  |= __b2 >> __result.__ctz_;
+            __result.__seg_ += (__dn + __result.__ctz_) / __bits_per_word;
+            __result.__ctz_ = static_cast<unsigned>((__dn + __result.__ctz_)  % __bits_per_word);
+            __n -= __dn;
+            if (__n > 0)
+            {
+                __m = ~__storage_type(0) >> (__bits_per_word - __n);
+                __b2 = *__result.__seg_ & __m;
+                *__result.__seg_ &= ~__m;
+                *__result.__seg_ |= __b1 >> __dn;
+                *__first.__seg_  |= __b2 << __dn;
+                __result.__ctz_ = static_cast<unsigned>(__n);
+            }
+        }
+    }
+    return __result;
+}
+
+template <class __C1, class __C2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bit_iterator<__C2, false>
+swap_ranges(__bit_iterator<__C1, false> __first1, __bit_iterator<__C1, false> __last1,
+            __bit_iterator<__C2, false> __first2)
+{
+    if (__first1.__ctz_ == __first2.__ctz_)
+        return __swap_ranges_aligned(__first1, __last1, __first2);
+    return __swap_ranges_unaligned(__first1, __last1, __first2);
+}
+
+// rotate
+
+template <class _Cp>
+struct __bit_array
+{
+    typedef typename _Cp::difference_type difference_type;
+    typedef typename _Cp::__storage_type  __storage_type;
+    typedef typename _Cp::__storage_pointer __storage_pointer;
+    typedef typename _Cp::iterator        iterator;
+    static const unsigned __bits_per_word = _Cp::__bits_per_word;
+    static const unsigned _Np = 4;
+
+    difference_type __size_;
+    __storage_type __word_[_Np];
+
+    _LIBCUDACXX_INLINE_VISIBILITY static difference_type capacity()
+        {return static_cast<difference_type>(_Np * __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __bit_array(difference_type __s) : __size_(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY iterator begin()
+    {
+        return iterator(pointer_traits<__storage_pointer>::pointer_to(__word_[0]), 0);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY iterator end()
+    {
+        return iterator(pointer_traits<__storage_pointer>::pointer_to(__word_[0]) + __size_ / __bits_per_word,
+                                                  static_cast<unsigned>(__size_ % __bits_per_word));
+    }
+};
+
+template <class _Cp>
+__bit_iterator<_Cp, false>
+rotate(__bit_iterator<_Cp, false> __first, __bit_iterator<_Cp, false> __middle, __bit_iterator<_Cp, false> __last)
+{
+    typedef __bit_iterator<_Cp, false> _I1;
+    typedef  typename _I1::difference_type difference_type;
+    difference_type __d1 = __middle - __first;
+    difference_type __d2 = __last - __middle;
+    _I1 __r = __first + __d2;
+    while (__d1 != 0 && __d2 != 0)
+    {
+        if (__d1 <= __d2)
+        {
+            if (__d1 <= __bit_array<_Cp>::capacity())
+            {
+                __bit_array<_Cp> __b(__d1);
+                _CUDA_VSTD::copy(__first, __middle, __b.begin());
+                _CUDA_VSTD::copy(__b.begin(), __b.end(), _CUDA_VSTD::copy(__middle, __last, __first));
+                break;
+            }
+            else
+            {
+                __bit_iterator<_Cp, false> __mp = _CUDA_VSTD::swap_ranges(__first, __middle, __middle);
+                __first = __middle;
+                __middle = __mp;
+                __d2 -= __d1;
+            }
+        }
+        else
+        {
+            if (__d2 <= __bit_array<_Cp>::capacity())
+            {
+                __bit_array<_Cp> __b(__d2);
+                _CUDA_VSTD::copy(__middle, __last, __b.begin());
+                _CUDA_VSTD::copy_backward(__b.begin(), __b.end(), _CUDA_VSTD::copy_backward(__first, __middle, __last));
+                break;
+            }
+            else
+            {
+                __bit_iterator<_Cp, false> __mp = __first + __d2;
+                _CUDA_VSTD::swap_ranges(__first, __mp, __middle);
+                __first = __mp;
+                __d1 -= __d2;
+            }
+        }
+    }
+    return __r;
+}
+
+// equal
+
+template <class _Cp, bool _IC1, bool _IC2>
+bool
+__equal_unaligned(__bit_iterator<_Cp, _IC1> __first1, __bit_iterator<_Cp, _IC1> __last1,
+                  __bit_iterator<_Cp, _IC2> __first2)
+{
+    typedef __bit_iterator<_Cp, _IC1> _It;
+    typedef  typename _It::difference_type difference_type;
+    typedef typename _It::__storage_type __storage_type;
+    static const int __bits_per_word = _It::__bits_per_word;
+    difference_type __n = __last1 - __first1;
+    if (__n > 0)
+    {
+        // do first word
+        if (__first1.__ctz_ != 0)
+        {
+            unsigned __clz_f = __bits_per_word - __first1.__ctz_;
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__clz_f), __n);
+            __n -= __dn;
+            __storage_type __m = (~__storage_type(0) << __first1.__ctz_) & (~__storage_type(0) >> (__clz_f - __dn));
+            __storage_type __b = *__first1.__seg_ & __m;
+            unsigned __clz_r = __bits_per_word - __first2.__ctz_;
+            __storage_type __ddn = _CUDA_VSTD::min<__storage_type>(__dn, __clz_r);
+            __m = (~__storage_type(0) << __first2.__ctz_) & (~__storage_type(0) >> (__clz_r - __ddn));
+            if (__first2.__ctz_ > __first1.__ctz_)
+            {
+                if ((*__first2.__seg_ & __m) != (__b << (__first2.__ctz_ - __first1.__ctz_)))
+                    return false;
+            }
+            else
+            {
+                if ((*__first2.__seg_ & __m) != (__b >> (__first1.__ctz_ - __first2.__ctz_)))
+                    return false;
+            }
+            __first2.__seg_ += (__ddn + __first2.__ctz_) / __bits_per_word;
+            __first2.__ctz_ = static_cast<unsigned>((__ddn + __first2.__ctz_)  % __bits_per_word);
+            __dn -= __ddn;
+            if (__dn > 0)
+            {
+                __m = ~__storage_type(0) >> (__bits_per_word - __dn);
+                if ((*__first2.__seg_ & __m) != (__b >> (__first1.__ctz_ + __ddn)))
+                    return false;
+                __first2.__ctz_ = static_cast<unsigned>(__dn);
+            }
+            ++__first1.__seg_;
+            // __first1.__ctz_ = 0;
+        }
+        // __first1.__ctz_ == 0;
+        // do middle words
+        unsigned __clz_r = __bits_per_word - __first2.__ctz_;
+        __storage_type __m = ~__storage_type(0) << __first2.__ctz_;
+        for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first1.__seg_)
+        {
+            __storage_type __b = *__first1.__seg_;
+            if ((*__first2.__seg_ & __m) != (__b << __first2.__ctz_))
+                return false;
+            ++__first2.__seg_;
+            if ((*__first2.__seg_ & ~__m) != (__b >> __clz_r))
+                return false;
+        }
+        // do last word
+        if (__n > 0)
+        {
+            __m = ~__storage_type(0) >> (__bits_per_word - __n);
+            __storage_type __b = *__first1.__seg_ & __m;
+            __storage_type __dn = _CUDA_VSTD::min(__n, static_cast<difference_type>(__clz_r));
+            __m = (~__storage_type(0) << __first2.__ctz_) & (~__storage_type(0) >> (__clz_r - __dn));
+            if ((*__first2.__seg_ & __m) != (__b << __first2.__ctz_))
+                return false;
+            __first2.__seg_ += (__dn + __first2.__ctz_) / __bits_per_word;
+            __first2.__ctz_ = static_cast<unsigned>((__dn + __first2.__ctz_)  % __bits_per_word);
+            __n -= __dn;
+            if (__n > 0)
+            {
+                __m = ~__storage_type(0) >> (__bits_per_word - __n);
+                if ((*__first2.__seg_ & __m) != (__b >> __dn))
+                    return false;
+            }
+        }
+    }
+    return true;
+}
+
+template <class _Cp, bool _IC1, bool _IC2>
+bool
+__equal_aligned(__bit_iterator<_Cp, _IC1> __first1, __bit_iterator<_Cp, _IC1> __last1,
+                __bit_iterator<_Cp, _IC2> __first2)
+{
+    typedef __bit_iterator<_Cp, _IC1> _It;
+    typedef  typename _It::difference_type difference_type;
+    typedef typename _It::__storage_type __storage_type;
+    static const int __bits_per_word = _It::__bits_per_word;
+    difference_type __n = __last1 - __first1;
+    if (__n > 0)
+    {
+        // do first word
+        if (__first1.__ctz_ != 0)
+        {
+            unsigned __clz = __bits_per_word - __first1.__ctz_;
+            difference_type __dn = _CUDA_VSTD::min(static_cast<difference_type>(__clz), __n);
+            __n -= __dn;
+            __storage_type __m = (~__storage_type(0) << __first1.__ctz_) & (~__storage_type(0) >> (__clz - __dn));
+            if ((*__first2.__seg_ & __m) != (*__first1.__seg_ & __m))
+                return false;
+            ++__first2.__seg_;
+            ++__first1.__seg_;
+            // __first1.__ctz_ = 0;
+            // __first2.__ctz_ = 0;
+        }
+        // __first1.__ctz_ == 0;
+        // __first2.__ctz_ == 0;
+        // do middle words
+        for (; __n >= __bits_per_word; __n -= __bits_per_word, ++__first1.__seg_, ++__first2.__seg_)
+            if (*__first2.__seg_ != *__first1.__seg_)
+                return false;
+        // do last word
+        if (__n > 0)
+        {
+            __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+            if ((*__first2.__seg_ & __m) != (*__first1.__seg_ & __m))
+                return false;
+        }
+    }
+    return true;
+}
+
+template <class _Cp, bool _IC1, bool _IC2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+equal(__bit_iterator<_Cp, _IC1> __first1, __bit_iterator<_Cp, _IC1> __last1, __bit_iterator<_Cp, _IC2> __first2)
+{
+    if (__first1.__ctz_ == __first2.__ctz_)
+        return __equal_aligned(__first1, __last1, __first2);
+    return __equal_unaligned(__first1, __last1, __first2);
+}
+
+template <class _Cp, bool _IsConst,
+          typename _Cp::__storage_type>
+class __bit_iterator
+{
+public:
+    typedef typename _Cp::difference_type                                                          difference_type;
+    typedef bool                                                                                  value_type;
+    typedef __bit_iterator                                                                        pointer;
+    typedef typename conditional<_IsConst, __bit_const_reference<_Cp>, __bit_reference<_Cp> >::type reference;
+    typedef random_access_iterator_tag                                                            iterator_category;
+
+private:
+    typedef typename _Cp::__storage_type                                           __storage_type;
+    typedef typename conditional<_IsConst, typename _Cp::__const_storage_pointer,
+                                           typename _Cp::__storage_pointer>::type  __storage_pointer;
+    static const unsigned __bits_per_word = _Cp::__bits_per_word;
+
+    __storage_pointer __seg_;
+    unsigned          __ctz_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator() _NOEXCEPT
+#if _LIBCUDACXX_STD_VER > 11
+    : __seg_(nullptr), __ctz_(0)
+#endif
+    {}
+    // avoid re-declaring a copy constructor for the non-const version.
+    using __type_for_copy_to_const =
+      _If<_IsConst, __bit_iterator<_Cp, false>, struct __private_nat>;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_iterator(const __type_for_copy_to_const& __it) _NOEXCEPT
+        : __seg_(__it.__seg_), __ctz_(__it.__ctz_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator*() const _NOEXCEPT
+        {return reference(__seg_, __storage_type(1) << __ctz_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator& operator++()
+    {
+        if (__ctz_ != __bits_per_word-1)
+            ++__ctz_;
+        else
+        {
+            __ctz_ = 0;
+            ++__seg_;
+        }
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator operator++(int)
+    {
+        __bit_iterator __tmp = *this;
+        ++(*this);
+        return __tmp;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator& operator--()
+    {
+        if (__ctz_ != 0)
+            --__ctz_;
+        else
+        {
+            __ctz_ = __bits_per_word - 1;
+            --__seg_;
+        }
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator operator--(int)
+    {
+        __bit_iterator __tmp = *this;
+        --(*this);
+        return __tmp;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator& operator+=(difference_type __n)
+    {
+        if (__n >= 0)
+            __seg_ += (__n + __ctz_) / __bits_per_word;
+        else
+            __seg_ += static_cast<difference_type>(__n - __bits_per_word + __ctz_ + 1)
+                    / static_cast<difference_type>(__bits_per_word);
+        __n &= (__bits_per_word - 1);
+        __ctz_ = static_cast<unsigned>((__n + __ctz_)  % __bits_per_word);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator& operator-=(difference_type __n)
+    {
+        return *this += -__n;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator operator+(difference_type __n) const
+    {
+        __bit_iterator __t(*this);
+        __t += __n;
+        return __t;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __bit_iterator operator-(difference_type __n) const
+    {
+        __bit_iterator __t(*this);
+        __t -= __n;
+        return __t;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend __bit_iterator operator+(difference_type __n, const __bit_iterator& __it) {return __it + __n;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend difference_type operator-(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return (__x.__seg_ - __y.__seg_) * __bits_per_word + __x.__ctz_ - __y.__ctz_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator[](difference_type __n) const {return *(*this + __n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend bool operator==(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return __x.__seg_ == __y.__seg_ && __x.__ctz_ == __y.__ctz_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend bool operator!=(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return !(__x == __y);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend bool operator<(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return __x.__seg_ < __y.__seg_ || (__x.__seg_ == __y.__seg_ && __x.__ctz_ < __y.__ctz_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend bool operator>(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return __y < __x;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend bool operator<=(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return !(__y < __x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend bool operator>=(const __bit_iterator& __x, const __bit_iterator& __y)
+        {return !(__x < __y);}
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bit_iterator(__storage_pointer __s, unsigned __ctz) _NOEXCEPT
+        : __seg_(__s), __ctz_(__ctz) {}
+
+    friend typename _Cp::__self;
+
+    friend class __bit_reference<_Cp>;
+    friend class __bit_const_reference<_Cp>;
+    friend class __bit_iterator<_Cp, true>;
+    template <class _Dp> friend struct __bit_array;
+    template <class _Dp> friend void __fill_n_false(__bit_iterator<_Dp, false> __first, typename _Dp::size_type __n);
+    template <class _Dp> friend void __fill_n_true(__bit_iterator<_Dp, false> __first, typename _Dp::size_type __n);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, false> __copy_aligned(__bit_iterator<_Dp, _IC> __first,
+                                                                                  __bit_iterator<_Dp, _IC> __last,
+                                                                                  __bit_iterator<_Dp, false> __result);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, false> __copy_unaligned(__bit_iterator<_Dp, _IC> __first,
+                                                                                    __bit_iterator<_Dp, _IC> __last,
+                                                                                    __bit_iterator<_Dp, false> __result);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, false> copy(__bit_iterator<_Dp, _IC> __first,
+                                                                        __bit_iterator<_Dp, _IC> __last,
+                                                                        __bit_iterator<_Dp, false> __result);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, false> __copy_backward_aligned(__bit_iterator<_Dp, _IC> __first,
+                                                                                           __bit_iterator<_Dp, _IC> __last,
+                                                                                           __bit_iterator<_Dp, false> __result);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, false> __copy_backward_unaligned(__bit_iterator<_Dp, _IC> __first,
+                                                                                             __bit_iterator<_Dp, _IC> __last,
+                                                                                             __bit_iterator<_Dp, false> __result);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, false> copy_backward(__bit_iterator<_Dp, _IC> __first,
+                                                                                 __bit_iterator<_Dp, _IC> __last,
+                                                                                 __bit_iterator<_Dp, false> __result);
+    template <class __C1, class __C2>friend __bit_iterator<__C2, false> __swap_ranges_aligned(__bit_iterator<__C1, false>,
+                                                                                           __bit_iterator<__C1, false>,
+                                                                                           __bit_iterator<__C2, false>);
+    template <class __C1, class __C2>friend __bit_iterator<__C2, false> __swap_ranges_unaligned(__bit_iterator<__C1, false>,
+                                                                                             __bit_iterator<__C1, false>,
+                                                                                             __bit_iterator<__C2, false>);
+    template <class __C1, class __C2>friend __bit_iterator<__C2, false> swap_ranges(__bit_iterator<__C1, false>,
+                                                                                 __bit_iterator<__C1, false>,
+                                                                                 __bit_iterator<__C2, false>);
+    template <class _Dp> friend __bit_iterator<_Dp, false> rotate(__bit_iterator<_Dp, false>,
+                                                                __bit_iterator<_Dp, false>,
+                                                                __bit_iterator<_Dp, false>);
+    template <class _Dp, bool _IC1, bool _IC2> friend bool __equal_aligned(__bit_iterator<_Dp, _IC1>,
+                                                    __bit_iterator<_Dp, _IC1>,
+                                                    __bit_iterator<_Dp, _IC2>);
+    template <class _Dp, bool _IC1, bool _IC2> friend bool __equal_unaligned(__bit_iterator<_Dp, _IC1>,
+                                                      __bit_iterator<_Dp, _IC1>,
+                                                      __bit_iterator<_Dp, _IC2>);
+    template <class _Dp, bool _IC1, bool _IC2> friend bool equal(__bit_iterator<_Dp, _IC1>,
+                                                                __bit_iterator<_Dp, _IC1>,
+                                                                __bit_iterator<_Dp, _IC2>);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, _IC> __find_bool_true(__bit_iterator<_Dp, _IC>,
+                                                                          typename _Dp::size_type);
+    template <class _Dp, bool _IC> friend __bit_iterator<_Dp, _IC> __find_bool_false(__bit_iterator<_Dp, _IC>,
+                                                                           typename _Dp::size_type);
+    template <class _Dp, bool _IC> friend typename __bit_iterator<_Dp, _IC>::difference_type
+                   __count_bool_true(__bit_iterator<_Dp, _IC>, typename _Dp::size_type);
+    template <class _Dp, bool _IC> friend typename __bit_iterator<_Dp, _IC>::difference_type
+                   __count_bool_false(__bit_iterator<_Dp, _IC>, typename _Dp::size_type);
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX___BIT_REFERENCE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h
new file mode 100644
index 000000000000..609865abd163
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_defaults.h
@@ -0,0 +1,36 @@
+// -*- C++ -*-
+//===---------------------- __bsd_locale_defaults.h -----------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// The BSDs have lots of *_l functions.  We don't want to define those symbols
+// on other platforms though, for fear of conflicts with user code.  So here,
+// we will define the mapping from an internal macro to the real BSD symbol.
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_BSD_LOCALE_DEFAULTS_H
+#define _LIBCUDACXX_BSD_LOCALE_DEFAULTS_H
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#define __libcpp_mb_cur_max_l(loc)                          MB_CUR_MAX_L(loc)
+#define __libcpp_btowc_l(ch, loc)                           btowc_l(ch, loc)
+#define __libcpp_wctob_l(wch, loc)                          wctob_l(wch, loc)
+#define __libcpp_wcsnrtombs_l(dst, src, nwc, len, ps, loc)  wcsnrtombs_l(dst, src, nwc, len, ps, loc)
+#define __libcpp_wcrtomb_l(src, wc, ps, loc)                wcrtomb_l(src, wc, ps, loc)
+#define __libcpp_mbsnrtowcs_l(dst, src, nms, len, ps, loc)  mbsnrtowcs_l(dst, src, nms, len, ps, loc)
+#define __libcpp_mbrtowc_l(pwc, s, n, ps, l)                mbrtowc_l(pwc, s, n, ps, l)
+#define __libcpp_mbtowc_l(pwc, pmb, max, l)                 mbtowc_l(pwc, pmb, max, l)
+#define __libcpp_mbrlen_l(s, n, ps, l)                      mbrlen_l(s, n, ps, l)
+#define __libcpp_localeconv_l(l)                            localeconv_l(l)
+#define __libcpp_mbsrtowcs_l(dest, src, len, ps, l)         mbsrtowcs_l(dest, src, len, ps, l)
+#define __libcpp_snprintf_l(...)                            snprintf_l(__VA_ARGS__)
+#define __libcpp_asprintf_l(...)                            asprintf_l(__VA_ARGS__)
+#define __libcpp_sscanf_l(...)                              sscanf_l(__VA_ARGS__)
+
+#endif // _LIBCUDACXX_BSD_LOCALE_DEFAULTS_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h
new file mode 100644
index 000000000000..58220977d76d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__bsd_locale_fallbacks.h
@@ -0,0 +1,139 @@
+// -*- C++ -*-
+//===---------------------- __bsd_locale_fallbacks.h ----------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// The BSDs have lots of *_l functions.  This file provides reimplementations
+// of those functions for non-BSD platforms.
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_BSD_LOCALE_FALLBACKS_DEFAULTS_H
+#define _LIBCUDACXX_BSD_LOCALE_FALLBACKS_DEFAULTS_H
+
+#include <stdlib.h>
+#include <stdarg.h>
+#include <memory>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+decltype(MB_CUR_MAX) __libcpp_mb_cur_max_l(locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return MB_CUR_MAX;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+wint_t __libcpp_btowc_l(int __c, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return btowc(__c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+int __libcpp_wctob_l(wint_t __c, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return wctob(__c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __libcpp_wcsnrtombs_l(char *__dest, const wchar_t **__src, size_t __nwc,
+                         size_t __len, mbstate_t *__ps, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return wcsnrtombs(__dest, __src, __nwc, __len, __ps);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __libcpp_wcrtomb_l(char *__s, wchar_t __wc, mbstate_t *__ps, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return wcrtomb(__s, __wc, __ps);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __libcpp_mbsnrtowcs_l(wchar_t * __dest, const char **__src, size_t __nms,
+                      size_t __len, mbstate_t *__ps, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return mbsnrtowcs(__dest, __src, __nms, __len, __ps);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __libcpp_mbrtowc_l(wchar_t *__pwc, const char *__s, size_t __n,
+                   mbstate_t *__ps, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return mbrtowc(__pwc, __s, __n, __ps);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+int __libcpp_mbtowc_l(wchar_t *__pwc, const char *__pmb, size_t __max, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return mbtowc(__pwc, __pmb, __max);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __libcpp_mbrlen_l(const char *__s, size_t __n, mbstate_t *__ps, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return mbrlen(__s, __n, __ps);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+lconv *__libcpp_localeconv_l(locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return localeconv();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __libcpp_mbsrtowcs_l(wchar_t *__dest, const char **__src, size_t __len,
+                     mbstate_t *__ps, locale_t __l)
+{
+    __libcpp_locale_guard __current(__l);
+    return mbsrtowcs(__dest, __src, __len, __ps);
+}
+
+inline
+int __libcpp_snprintf_l(char *__s, size_t __n, locale_t __l, const char *__format, ...) {
+    va_list __va;
+    va_start(__va, __format);
+    __libcpp_locale_guard __current(__l);
+    int __res = vsnprintf(__s, __n, __format, __va);
+    va_end(__va);
+    return __res;
+}
+
+inline
+int __libcpp_asprintf_l(char **__s, locale_t __l, const char *__format, ...) {
+    va_list __va;
+    va_start(__va, __format);
+    __libcpp_locale_guard __current(__l);
+    int __res = vasprintf(__s, __format, __va);
+    va_end(__va);
+    return __res;
+}
+
+inline
+int __libcpp_sscanf_l(const char *__s, locale_t __l, const char *__format, ...) {
+    va_list __va;
+    va_start(__va, __format);
+    __libcpp_locale_guard __current(__l);
+    int __res = vsscanf(__s, __format, __va);
+    va_end(__va);
+    return __res;
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // _LIBCUDACXX_BSD_LOCALE_FALLBACKS_DEFAULTS_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config
new file mode 100644
index 000000000000..2714477eca5b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config
@@ -0,0 +1,1943 @@
+// -*- C++ -*-
+//===--------------------------- __config ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CONFIG
+#define _LIBCUDACXX_CONFIG
+
+#if defined(_MSC_VER) && !defined(__clang__)
+  #define _LIBCUDACXX_HAS_PRAGMA_MSVC_WARNING
+  #if !defined(_LIBCUDACXX_DISABLE_PRAGMA_MSVC_WARNING)
+    #define _LIBCUDACXX_USE_PRAGMA_MSVC_WARNING
+  #endif
+#else
+  #define _LIBCUDACXX_HAS_PRAGMA_GCC_SYSTEM_HEADER
+  #if !defined(_LIBCUDACXX_DISABLE_PRAGMA_GCC_SYSTEM_HEADER)
+    #define _LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER
+  #endif
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef __cplusplus
+
+// __config may be included in `extern "C"` contexts, switch back to include <nv/target>
+extern "C++" {
+#include <nv/target>
+}
+
+#ifdef __GNUC__
+#  define _GNUC_VER (__GNUC__ * 100 + __GNUC_MINOR__)
+// The _GNUC_VER_NEW macro better represents the new GCC versioning scheme
+// introduced in GCC 5.0.
+#  define _GNUC_VER_NEW (_GNUC_VER * 10 + __GNUC_PATCHLEVEL__)
+#else
+#  define _GNUC_VER 0
+#  define _GNUC_VER_NEW 0
+#endif
+
+#define _LIBCUDACXX_VERSION 10000
+
+#ifndef _LIBCUDACXX_ABI_VERSION
+#  define _LIBCUDACXX_ABI_VERSION 1
+#endif
+
+#ifndef __STDC_HOSTED__
+#  define _LIBCUDACXX_FREESTANDING
+#endif
+
+#ifndef _LIBCUDACXX_STD_VER
+#  if defined(_MSC_VER)
+#    if   _MSVC_LANG <= 201103L
+#      define _LIBCUDACXX_STD_VER 11
+#    elif _MSVC_LANG <= 201402L
+#      define _LIBCUDACXX_STD_VER 14
+#    elif _MSVC_LANG <= 201703L
+#      define _LIBCUDACXX_STD_VER 17
+#    else
+#      define _LIBCUDACXX_STD_VER 19  // current year, or date of c++2a ratification
+#    endif
+#  else
+#    if   __cplusplus <= 201103L
+#      define _LIBCUDACXX_STD_VER 11
+#    elif __cplusplus <= 201402L
+#      define _LIBCUDACXX_STD_VER 14
+#    elif __cplusplus <= 201703L
+#      define _LIBCUDACXX_STD_VER 17
+#    else
+#      define _LIBCUDACXX_STD_VER 19  // current year, or date of c++2a ratification
+#    endif
+#  endif
+#endif  // _LIBCUDACXX_STD_VER
+
+#if _LIBCUDACXX_STD_VER < 11
+#define _LIBCUDACXX_CXX03_LANG
+#endif
+
+#if defined(__PGIC__) && defined(__linux__)
+    #define __ELF__
+#endif
+
+#if defined(__ELF__)
+#  define _LIBCUDACXX_OBJECT_FORMAT_ELF   1
+#elif defined(__MACH__)
+#  define _LIBCUDACXX_OBJECT_FORMAT_MACHO 1
+#elif defined(_WIN32)
+#  define _LIBCUDACXX_OBJECT_FORMAT_COFF  1
+#elif defined(__wasm__)
+#  define _LIBCUDACXX_OBJECT_FORMAT_WASM  1
+#else
+#  error Unknown object file format
+#endif
+
+#if defined(_LIBCUDACXX_ABI_UNSTABLE) || _LIBCUDACXX_ABI_VERSION >= 2
+// Change short string representation so that string data starts at offset 0,
+// improving its alignment in some cases.
+#  define _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+// Fix deque iterator type in order to support incomplete types.
+#  define _LIBCUDACXX_ABI_INCOMPLETE_TYPES_IN_DEQUE
+// Fix undefined behavior in how std::list stores its linked nodes.
+#  define _LIBCUDACXX_ABI_LIST_REMOVE_NODE_POINTER_UB
+// Fix undefined behavior in  how __tree stores its end and parent nodes.
+#  define _LIBCUDACXX_ABI_TREE_REMOVE_NODE_POINTER_UB
+// Fix undefined behavior in how __hash_table stores its pointer types.
+#  define _LIBCUDACXX_ABI_FIX_UNORDERED_NODE_POINTER_UB
+#  define _LIBCUDACXX_ABI_FORWARD_LIST_REMOVE_NODE_POINTER_UB
+#  define _LIBCUDACXX_ABI_FIX_UNORDERED_CONTAINER_SIZE_TYPE
+// Don't use a nullptr_t simulation type in C++03 instead using C++11 nullptr
+// provided under the alternate keyword __nullptr, which changes the mangling
+// of nullptr_t. This option is ABI incompatible with GCC in C++03 mode.
+#  define _LIBCUDACXX_ABI_ALWAYS_USE_CXX11_NULLPTR
+// Define the `pointer_safety` enum as a C++11 strongly typed enumeration
+// instead of as a class simulating an enum. If this option is enabled
+// `pointer_safety` and `get_pointer_safety()` will no longer be available
+// in C++03.
+#  define _LIBCUDACXX_ABI_POINTER_SAFETY_ENUM_TYPE
+// Define a key function for `bad_function_call` in the library, to centralize
+// its vtable and typeinfo to libc++ rather than having all other libraries
+// using that class define their own copies.
+#  define _LIBCUDACXX_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
+// Enable optimized version of __do_get_(un)signed which avoids redundant copies.
+#  define _LIBCUDACXX_ABI_OPTIMIZED_LOCALE_NUM_GET
+// Use the smallest possible integer type to represent the index of the variant.
+// Previously libc++ used "unsigned int" exclusively.
+#  define _LIBCUDACXX_ABI_VARIANT_INDEX_TYPE_OPTIMIZATION
+// Unstable attempt to provide a more optimized std::function
+#  define _LIBCUDACXX_ABI_OPTIMIZED_FUNCTION
+// All the regex constants must be distinct and nonzero.
+#  define _LIBCUDACXX_ABI_REGEX_CONSTANTS_NONZERO
+#elif _LIBCUDACXX_ABI_VERSION == 1
+#  if !defined(_LIBCUDACXX_OBJECT_FORMAT_COFF)
+// Enable compiling copies of now inline methods into the dylib to support
+// applications compiled against older libraries. This is unnecessary with
+// COFF dllexport semantics, since dllexport forces a non-inline definition
+// of inline functions to be emitted anyway. Our own non-inline copy would
+// conflict with the dllexport-emitted copy, so we disable it.
+#    define _LIBCUDACXX_DEPRECATED_ABI_LEGACY_LIBRARY_DEFINITIONS_FOR_INLINE_FUNCTIONS
+#  endif
+// Feature macros for disabling pre ABI v1 features. All of these options
+// are deprecated.
+#  if defined(__FreeBSD__)
+#    define _LIBCUDACXX_DEPRECATED_ABI_DISABLE_PAIR_TRIVIAL_COPY_CTOR
+#  endif
+#endif
+
+#ifdef _LIBCUDACXX_TRIVIAL_PAIR_COPY_CTOR
+#error "_LIBCUDACXX_TRIVIAL_PAIR_COPY_CTOR" is no longer supported. \
+       use _LIBCUDACXX_DEPRECATED_ABI_DISABLE_PAIR_TRIVIAL_COPY_CTOR instead
+#endif
+
+#define _LIBCUDACXX_CONCAT1(_LIBCUDACXX_X,_LIBCUDACXX_Y) _LIBCUDACXX_X##_LIBCUDACXX_Y
+#define _LIBCUDACXX_CONCAT(_LIBCUDACXX_X,_LIBCUDACXX_Y) _LIBCUDACXX_CONCAT1(_LIBCUDACXX_X,_LIBCUDACXX_Y)
+
+#ifndef _LIBCUDACXX_ABI_NAMESPACE
+# define _LIBCUDACXX_ABI_NAMESPACE _LIBCUDACXX_CONCAT(__,_LIBCUDACXX_ABI_VERSION)
+#endif
+
+#ifndef __has_attribute
+#define __has_attribute(__x) 0
+#endif
+
+#ifndef __has_builtin
+#define __has_builtin(__x) 0
+#endif
+
+#ifndef __has_extension
+#define __has_extension(__x) 0
+#endif
+
+#ifndef __has_feature
+#define __has_feature(__x) 0
+#endif
+
+#ifndef __has_cpp_attribute
+#define __has_cpp_attribute(__x) 0
+#endif
+
+// '__is_identifier' returns '0' if '__x' is a reserved identifier provided by
+// the compiler and '1' otherwise.
+#ifndef __is_identifier
+#define __is_identifier(__x) 1
+#endif
+
+#ifndef __has_declspec_attribute
+#define __has_declspec_attribute(__x) 0
+#endif
+
+#define __has_keyword(__x) !(__is_identifier(__x))
+
+#ifndef __has_include
+#define __has_include(...) 0
+#endif
+
+#if defined(__PGIC__)
+#  define _LIBCUDACXX_COMPILER_PGI
+#elif defined(__clang__)
+#  define _LIBCUDACXX_COMPILER_CLANG
+#  ifndef __apple_build_version__
+#    define _LIBCUDACXX_CLANG_VER (__clang_major__ * 100 + __clang_minor__)
+#  endif
+#elif defined(__GNUC__)
+#  define _LIBCUDACXX_COMPILER_GCC
+#elif defined(_MSC_VER)
+#  define _LIBCUDACXX_COMPILER_MSVC
+#elif defined(__IBMCPP__)
+#  define _LIBCUDACXX_COMPILER_IBM
+#elif defined(__CUDACC_RTC__)
+#  define _LIBCUDACXX_COMPILER_NVRTC
+#endif
+
+#if defined(__NVCC__)
+// This is not mutually exclusive with other compilers, as NVCC uses a host
+// compiler.
+#  define _LIBCUDACXX_COMPILER_NVCC
+#endif
+
+#if !defined(_LIBCUDACXX_COMPILER_NVCC) && !defined(_LIBCUDACXX_COMPILER_NVRTC)
+// If NVCC is not being used <complex> can safely use `long double` without warnings
+#  define _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+// NVCC does not have a way of silencing non '_' prefixed UDLs
+#  define _LIBCUDACXX_HAS_STL_LITERALS
+#endif
+
+#if defined(_LIBCUDACXX_COMPILER_GCC) && __cplusplus < 201103L
+#error "libc++ does not support using GCC with C++03. Please enable C++11"
+#endif
+
+// FIXME: ABI detection should be done via compiler builtin macros. This
+// is just a placeholder until Clang implements such macros. For now assume
+// that Windows compilers pretending to be MSVC++ target the Microsoft ABI,
+// and allow the user to explicitly specify the ABI to handle cases where this
+// heuristic falls short.
+#if defined(_LIBCUDACXX_ABI_FORCE_ITANIUM) && defined(_LIBCUDACXX_ABI_FORCE_MICROSOFT)
+#  error "Only one of _LIBCUDACXX_ABI_FORCE_ITANIUM and _LIBCUDACXX_ABI_FORCE_MICROSOFT can be defined"
+#elif defined(_LIBCUDACXX_ABI_FORCE_ITANIUM)
+#  define _LIBCUDACXX_ABI_ITANIUM
+#elif defined(_LIBCUDACXX_ABI_FORCE_MICROSOFT)
+#  define _LIBCUDACXX_ABI_MICROSOFT
+#else
+#  if defined(_WIN32) && defined(_MSC_VER)
+#    define _LIBCUDACXX_ABI_MICROSOFT
+#  else
+#    define _LIBCUDACXX_ABI_ITANIUM
+#  endif
+#endif
+
+#if defined(_LIBCUDACXX_ABI_MICROSOFT) && !defined(_LIBCUDACXX_NO_VCRUNTIME)
+# define _LIBCUDACXX_ABI_VCRUNTIME
+#endif
+
+// Need to detect which libc we're using if we're on Linux.
+#if defined(__linux__)
+#  include <features.h>
+#  if defined(__GLIBC_PREREQ)
+#    define _LIBCUDACXX_GLIBC_PREREQ(a, b) __GLIBC_PREREQ(a, b)
+#  else
+#    define _LIBCUDACXX_GLIBC_PREREQ(a, b) 0
+#  endif // defined(__GLIBC_PREREQ)
+#endif // defined(__linux__)
+
+#ifdef __LITTLE_ENDIAN__
+#  if __LITTLE_ENDIAN__
+#    define _LIBCUDACXX_LITTLE_ENDIAN
+#  endif  // __LITTLE_ENDIAN__
+#endif  // __LITTLE_ENDIAN__
+
+#ifdef __BIG_ENDIAN__
+#  if __BIG_ENDIAN__
+#    define _LIBCUDACXX_BIG_ENDIAN
+#  endif  // __BIG_ENDIAN__
+#endif  // __BIG_ENDIAN__
+
+#ifdef __BYTE_ORDER__
+#  if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
+#    define _LIBCUDACXX_LITTLE_ENDIAN
+#  elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#    define _LIBCUDACXX_BIG_ENDIAN
+#  endif // __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+#endif // __BYTE_ORDER__
+
+#ifdef __FreeBSD__
+#  include <sys/endian.h>
+#  if _BYTE_ORDER == _LITTLE_ENDIAN
+#    define _LIBCUDACXX_LITTLE_ENDIAN
+#  else  // _BYTE_ORDER == _LITTLE_ENDIAN
+#    define _LIBCUDACXX_BIG_ENDIAN
+#  endif  // _BYTE_ORDER == _LITTLE_ENDIAN
+#  ifndef __LONG_LONG_SUPPORTED
+#    define _LIBCUDACXX_HAS_NO_LONG_LONG
+#  endif  // __LONG_LONG_SUPPORTED
+#endif  // __FreeBSD__
+
+#ifdef __NetBSD__
+#  include <sys/endian.h>
+#  if _BYTE_ORDER == _LITTLE_ENDIAN
+#    define _LIBCUDACXX_LITTLE_ENDIAN
+#  else  // _BYTE_ORDER == _LITTLE_ENDIAN
+#    define _LIBCUDACXX_BIG_ENDIAN
+#  endif  // _BYTE_ORDER == _LITTLE_ENDIAN
+#  define _LIBCUDACXX_HAS_QUICK_EXIT
+#endif  // __NetBSD__
+
+#if defined(_WIN32)
+#  define _LIBCUDACXX_WIN32API
+#  define _LIBCUDACXX_LITTLE_ENDIAN
+#  define _LIBCUDACXX_SHORT_WCHAR   1
+// Both MinGW and native MSVC provide a "MSVC"-like environment
+#  define _LIBCUDACXX_MSVCRT_LIKE
+// If mingw not explicitly detected, assume using MS C runtime only if
+// a MS compatibility version is specified.
+#  if defined(_MSC_VER) && !defined(__MINGW32__)
+#    define _LIBCUDACXX_MSVCRT // Using Microsoft's C Runtime library
+#  endif
+#  if (defined(_M_AMD64) || defined(__x86_64__)) || (defined(_M_ARM) || defined(__arm__))
+#    define _LIBCUDACXX_HAS_BITSCAN64
+#  endif
+#  define _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+#  if defined(_LIBCUDACXX_MSVCRT)
+#    define _LIBCUDACXX_HAS_QUICK_EXIT
+#  endif
+
+// Some CRT APIs are unavailable to store apps
+#  if defined(WINAPI_FAMILY)
+#    include <winapifamily.h>
+#    if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) &&                  \
+        (!defined(WINAPI_PARTITION_SYSTEM) ||                                  \
+         !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_SYSTEM))
+#      define _LIBCUDACXX_WINDOWS_STORE_APP
+#    endif
+#  endif
+#endif // defined(_WIN32)
+
+#ifdef __sun__
+#  include <sys/isa_defs.h>
+#  ifdef _LITTLE_ENDIAN
+#    define _LIBCUDACXX_LITTLE_ENDIAN
+#  else
+#    define _LIBCUDACXX_BIG_ENDIAN
+#  endif
+#endif // __sun__
+
+#if defined(__CloudABI__)
+   // Certain architectures provide arc4random(). Prefer using
+   // arc4random() over /dev/{u,}random to make it possible to obtain
+   // random data even when using sandboxing mechanisms such as chroots,
+   // Capsicum, etc.
+#  define _LIBCUDACXX_USING_ARC4_RANDOM
+#elif defined(__Fuchsia__) || defined(__wasi__)
+#  define _LIBCUDACXX_USING_GETENTROPY
+#elif defined(__native_client__)
+   // NaCl's sandbox (which PNaCl also runs in) doesn't allow filesystem access,
+   // including accesses to the special files under /dev. C++11's
+   // std::random_device is instead exposed through a NaCl syscall.
+#  define _LIBCUDACXX_USING_NACL_RANDOM
+#elif defined(_LIBCUDACXX_WIN32API)
+#  define _LIBCUDACXX_USING_WIN32_RANDOM
+#else
+#  define _LIBCUDACXX_USING_DEV_RANDOM
+#endif
+
+#if !defined(_LIBCUDACXX_LITTLE_ENDIAN) && !defined(_LIBCUDACXX_BIG_ENDIAN)
+#  include <endian.h>
+#  if __BYTE_ORDER == __LITTLE_ENDIAN
+#    define _LIBCUDACXX_LITTLE_ENDIAN
+#  elif __BYTE_ORDER == __BIG_ENDIAN
+#    define _LIBCUDACXX_BIG_ENDIAN
+#  else  // __BYTE_ORDER == __BIG_ENDIAN
+#    error unable to determine endian
+#  endif
+#endif  // !defined(_LIBCUDACXX_LITTLE_ENDIAN) && !defined(_LIBCUDACXX_BIG_ENDIAN)
+
+#if __has_attribute(__no_sanitize__) && !defined(_LIBCUDACXX_COMPILER_GCC)
+#  define _LIBCUDACXX_NO_CFI __attribute__((__no_sanitize__("cfi")))
+#else
+#  define _LIBCUDACXX_NO_CFI
+#endif
+
+#if __ISO_C_VISIBLE >= 2011 || __cplusplus >= 201103L
+#  if defined(__FreeBSD__)
+#    define _LIBCUDACXX_HAS_QUICK_EXIT
+#    define _LIBCUDACXX_HAS_C11_FEATURES
+#  elif defined(__Fuchsia__) || defined(__wasi__)
+#    define _LIBCUDACXX_HAS_QUICK_EXIT
+#    define _LIBCUDACXX_HAS_TIMESPEC_GET
+#    define _LIBCUDACXX_HAS_C11_FEATURES
+#  elif defined(__linux__)
+#    if !defined(_LIBCUDACXX_HAS_MUSL_LIBC)
+#      if _LIBCUDACXX_GLIBC_PREREQ(2, 15) || defined(__BIONIC__)
+#        define _LIBCUDACXX_HAS_QUICK_EXIT
+#      endif
+#      if _LIBCUDACXX_GLIBC_PREREQ(2, 17)
+#        define _LIBCUDACXX_HAS_C11_FEATURES
+#        define _LIBCUDACXX_HAS_TIMESPEC_GET
+#      endif
+#    else // defined(_LIBCUDACXX_HAS_MUSL_LIBC)
+#      define _LIBCUDACXX_HAS_QUICK_EXIT
+#      define _LIBCUDACXX_HAS_TIMESPEC_GET
+#      define _LIBCUDACXX_HAS_C11_FEATURES
+#    endif
+#  endif // __linux__
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+# define _LIBCUDACXX_ALIGNOF(_Tp) alignof(_Tp)
+#elif defined(_LIBCUDACXX_COMPILER_CLANG)
+# define _LIBCUDACXX_ALIGNOF(_Tp) _Alignof(_Tp)
+#else
+// This definition is potentially buggy, but it's only taken with GCC in C++03,
+// which we barely support anyway. See llvm.org/PR39713
+# define _LIBCUDACXX_ALIGNOF(_Tp) __alignof(_Tp)
+#endif
+
+#define _LIBCUDACXX_PREFERRED_ALIGNOF(_Tp) __alignof(_Tp)
+
+#if defined(_LIBCUDACXX_COMPILER_NVRTC)
+#define _LIBCUDACXX_OFFSET_IS_ZERO(type, member) !(&(((type *)0)->member))
+#else
+#define _LIBCUDACXX_OFFSET_IS_ZERO(type, member) !offsetof(type, member)
+#endif
+
+#if defined(_LIBCUDACXX_COMPILER_CLANG)
+
+// _LIBCUDACXX_ALTERNATE_STRING_LAYOUT is an old name for
+// _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT left here for backward compatibility.
+#if (defined(__APPLE__) && !defined(__i386__) && !defined(__x86_64__) &&       \
+     (!defined(__arm__) || __ARM_ARCH_7K__ >= 2)) ||                           \
+    defined(_LIBCUDACXX_ALTERNATE_STRING_LAYOUT)
+#define _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+#endif
+
+#if __has_feature(cxx_alignas)
+#  define _ALIGNAS_TYPE(x) alignas(x)
+#  define _ALIGNAS(x) alignas(x)
+#else
+#  define _ALIGNAS_TYPE(x) __attribute__((__aligned__(_LIBCUDACXX_ALIGNOF(x))))
+#  define _ALIGNAS(x) __attribute__((__aligned__(x)))
+#endif
+
+#if __cplusplus < 201103L
+typedef __char16_t char16_t;
+typedef __char32_t char32_t;
+#endif
+
+#if !(__has_feature(cxx_exceptions)) && !defined(_LIBCUDACXX_NO_EXCEPTIONS)
+#define _LIBCUDACXX_NO_EXCEPTIONS
+#endif
+
+#if !(__has_feature(cxx_rtti)) && !defined(_LIBCUDACXX_NO_RTTI)
+#define _LIBCUDACXX_NO_RTTI
+#endif
+
+#if !(__has_feature(cxx_strong_enums))
+#define _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+#endif
+
+#if __has_feature(cxx_attributes)
+#  define _LIBCUDACXX_NORETURN [[noreturn]]
+#else
+#  define _LIBCUDACXX_NORETURN __attribute__ ((noreturn))
+#endif
+
+#if !(__has_feature(cxx_lambdas))
+#define _LIBCUDACXX_HAS_NO_LAMBDAS
+#endif
+
+#if !(__has_feature(cxx_nullptr))
+#  if (__has_extension(cxx_nullptr) || __has_keyword(__nullptr)) && defined(_LIBCUDACXX_ABI_ALWAYS_USE_CXX11_NULLPTR)
+#    define nullptr __nullptr
+#  else
+#    define _LIBCUDACXX_HAS_NO_NULLPTR
+#  endif
+#endif
+
+#if !(__has_feature(cxx_rvalue_references))
+#define _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+#endif
+
+#if !(__has_feature(cxx_auto_type))
+#define _LIBCUDACXX_HAS_NO_AUTO_TYPE
+#endif
+
+#if !(__has_feature(cxx_variadic_templates))
+#define _LIBCUDACXX_HAS_NO_VARIADICS
+#endif
+
+#if !(__has_feature(cxx_generalized_initializers))
+#define _LIBCUDACXX_HAS_NO_GENERALIZED_INITIALIZERS
+#endif
+
+#if __has_feature(is_base_of)
+#define _LIBCUDACXX_IS_BASE_OF(...) __is_base_of(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_final)
+#define _LIBCUDACXX_IS_FINAL(...) __is_final(__VA_ARGS__)
+#endif
+
+// Objective-C++ features (opt-in)
+#if __has_feature(objc_arc)
+#define _LIBCUDACXX_HAS_OBJC_ARC
+#endif
+
+#if __has_feature(objc_arc_weak)
+#define _LIBCUDACXX_HAS_OBJC_ARC_WEAK
+#endif
+
+#if !(__has_feature(cxx_relaxed_constexpr))
+#define _LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR
+#endif
+
+#if !(__has_feature(cxx_variable_templates))
+#define _LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES
+#endif
+
+#if !(__has_feature(cxx_noexcept))
+#define _LIBCUDACXX_HAS_NO_NOEXCEPT
+#endif
+
+#if __has_feature(underlying_type)
+#define _LIBCUDACXX_UNDERLYING_TYPE(...) __underlying_type(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_constructible)
+#define _LIBCUDACXX_IS_CONSTRUCTIBLE(...) __is_constructible(__VA_ARGS__)
+#endif
+
+#if __has_keyword(__is_destructible)
+#define _LIBCUDACXX_IS_DESTRUCTIBLE(...) __is_destructible(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_trivially_constructible)
+#define _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(...) __is_trivially_constructible(__VA_ARGS__)
+#endif
+
+// https://bugs.llvm.org/show_bug.cgi?id=44517
+#if __has_keyword(__is_trivially_destructible)
+#define _LIBCUDACXX_IS_TRIVIALLY_DESTRUCTIBLE(...) __is_trivially_destructible(__VA_ARGS__)
+#endif
+
+// https://bugs.llvm.org/show_bug.cgi?id=44517
+#if __has_keyword(__is_nothrow_constructible)
+#define _LIBCUDACXX_IS_NOTHROW_CONSTRUCTIBLE(...) __is_nothrow_constructible(__VA_ARGS__)
+#endif
+
+// https://bugs.llvm.org/show_bug.cgi?id=44517
+#if __has_keyword(__is_nothrow_assignable)
+#define _LIBCUDACXX_IS_NOTHROW_ASSIGNABLE(...) __is_nothrow_assignable(__VA_ARGS__)
+#endif
+
+// https://bugs.llvm.org/show_bug.cgi?id=44517
+#if __has_keyword(__is_assignable)
+#define _LIBCUDACXX_IS_ASSIGNABLE(...) __is_assignable(__VA_ARGS__)
+#endif
+
+// https://bugs.llvm.org/show_bug.cgi?id=44517
+#if __has_keyword(__is_same) && !defined(_LIBCUDACXX_COMPILER_NVCC)
+#define _LIBCUDACXX_IS_SAME(...) __is_same(__VA_ARGS__)
+#endif
+
+#if __has_feature(has_trivial_constructor)
+#define _LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR(...) __has_trivial_constructor(__VA_ARGS__)
+#endif
+
+#if __has_feature(has_nothrow_constructor)
+#define _LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR(...) __has_nothrow_constructor(__VA_ARGS__)
+#endif
+
+#if __has_feature(has_nothrow_copy)
+#define _LIBCUDACXX_HAS_NOTHROW_COPY(...) __has_nothrow_copy(__VA_ARGS__)
+#endif
+
+#if __has_feature(has_nothrow_assign)
+#define _LIBCUDACXX_HAS_NOTHROW_ASSIGN(...) __has_nothrow_assign(__VA_ARGS__)
+#endif
+
+#if __has_feature(has_trivial_destructor)
+#define _LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR(...) __has_trivial_destructor(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_trivially_assignable)
+#define _LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE(...) __is_trivially_assignable(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_trivially_copyable)
+#define _LIBCUDACXX_IS_TRIVIALLY_COPYABLE(...) __is_trivially_copyable(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_trivial)
+#define _LIBCUDACXX_IS_TRIVIAL(...) __is_trivial(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_convertible_to)
+#define _LIBCUDACXX_IS_CONVERTIBLE_TO(...) __is_convertible_to(__VA_ARGS__)
+#endif
+
+// https://bugs.llvm.org/show_bug.cgi?id=44517
+#if __has_keyword(__is_function) && !defined(_LIBCUDACXX_COMPILER_NVCC)
+#define _LIBCUDACXX_IS_FUNCTION(...) __is_function(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_union)
+#define _LIBCUDACXX_IS_UNION(...) __is_union(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_class)
+#define _LIBCUDACXX_IS_CLASS(...) __is_class(__VA_ARGS__)
+#endif
+
+#if __has_keyword(__is_aggregate)
+#define _LIBCUDACXX_IS_AGGREGATE(...) __is_aggregate(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_pod)
+#define _LIBCUDACXX_IS_POD(...) __is_pod(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_standard_layout)
+#define _LIBCUDACXX_IS_STANDARD_LAYOUT(...) __is_standard_layout(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_enum)
+#define _LIBCUDACXX_IS_ENUM(...) __is_enum(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_empty)
+#define _LIBCUDACXX_IS_EMPTY(...) __is_empty(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_polymorphic)
+#define _LIBCUDACXX_IS_POLYMORPHIC(...) __is_polymorphic(__VA_ARGS__)
+#endif
+
+#if __has_feature(has_virtual_destructor)
+#define _LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR(...) __has_virtual_destructor(__VA_ARGS__)
+#endif
+
+#if __has_feature(is_literal)
+#define _LIBCUDACXX_IS_LITERAL(...) __is_literal(__VA_ARGS__)
+#endif
+
+#if !defined(_LIBCUDACXX_HAS_NO_ASAN) && !__has_feature(address_sanitizer)
+#define _LIBCUDACXX_HAS_NO_ASAN
+#endif
+
+// Allow for build-time disabling of unsigned integer sanitization
+#if !defined(_LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK) && __has_attribute(no_sanitize)
+#define _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK __attribute__((__no_sanitize__("unsigned-integer-overflow")))
+#endif
+
+#if __has_builtin(__builtin_launder)
+#define _LIBCUDACXX_LAUNDER(...) __builtin_launder(__VA_ARGS__)
+#endif
+
+#if !__is_identifier(__has_unique_object_representations)
+#define _LIBCUDACXX_HAS_UNIQUE_OBJECT_REPRESENTATIONS(...) __has_unique_object_representations(__VA_ARGS__)
+#endif
+
+#if !defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+#define _LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE(...) __atomic_always_lock_free(__VA_ARGS__)
+#endif
+
+// NVCC in C++11 mode freaks out about `__builtin_is_constant_evaluated`.
+#if __has_builtin(__builtin_is_constant_evaluated) && !(defined(_LIBCUDACXX_COMPILER_NVCC) && _LIBCUDACXX_STD_VER < 14)
+#define _LIBCUDACXX_IS_CONSTANT_EVALUATED(...) __builtin_is_constant_evaluated(__VA_ARGS__)
+#endif
+
+#if __has_builtin(__builtin_addressof)
+#define _LIBCUDACXX_ADDRESSOF(...) __builtin_addressof(__VA_ARGS__)
+#endif
+
+#define _LIBCUDACXX_ALWAYS_INLINE __attribute__ ((__always_inline__))
+
+// Literal operators ""d and ""y are supported starting with LLVM Clang 8 and AppleClang 10.0.1
+#if (defined(_LIBCUDACXX_CLANG_VER) && _LIBCUDACXX_CLANG_VER < 800) ||         \
+    (defined(__apple_build_version__) && __apple_build_version__ < 10010000)
+#define _LIBCUDACXX_HAS_NO_CXX20_CHRONO_LITERALS
+#endif
+
+#if __has_feature(cxx_reference_qualified_functions)
+#define _LIBCUDACXX_HAS_REFERENCE_QUALIFIED_FUNCTIONS
+#endif
+
+#define _LIBCUDACXX_DISABLE_EXTENSION_WARNING __extension__
+
+#elif defined(_LIBCUDACXX_COMPILER_GCC)
+
+#define _ALIGNAS(x) __attribute__((__aligned__(x)))
+#define _ALIGNAS_TYPE(x) __attribute__((__aligned__(_LIBCUDACXX_ALIGNOF(x))))
+
+#define _LIBCUDACXX_NORETURN __attribute__((noreturn))
+
+#if _GNUC_VER >= 403
+#define _LIBCUDACXX_IS_UNION(...) __is_union(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CLASS(...) __is_class(__VA_ARGS__)
+#define _LIBCUDACXX_IS_POD(...) __is_pod(__VA_ARGS__)
+#define _LIBCUDACXX_IS_ENUM(...) __is_enum(__VA_ARGS__)
+#define _LIBCUDACXX_IS_EMPTY(...) __is_empty(__VA_ARGS__)
+#define _LIBCUDACXX_IS_POLYMORPHIC(...) __is_polymorphic(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR(...) __has_virtual_destructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_BASE_OF(...) __is_base_of(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR(...) __has_trivial_constructor(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR(...) __has_nothrow_constructor(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_COPY(...) __has_nothrow_copy(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_ASSIGN(...) __has_nothrow_assign(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR(...) __has_trivial_destructor(__VA_ARGS__)
+#endif
+
+#if _GNUC_VER >= 405
+#define _LIBCUDACXX_IS_TRIVIAL(...) __is_trivial(__VA_ARGS__)
+#endif
+
+#if _GNUC_VER >= 406
+#define _LIBCUDACXX_IS_LITERAL(...) __is_literal_type(__VA_ARGS__)
+#endif
+
+#if _GNUC_VER >= 407
+#define _LIBCUDACXX_UNDERLYING_TYPE(...) __underlying_type(__VA_ARGS__)
+#define _LIBCUDACXX_IS_FINAL(...) __is_final(__VA_ARGS__)
+#define _LIBCUDACXX_IS_STANDARD_LAYOUT(...) __is_standard_layout(__VA_ARGS__)
+#endif
+
+#if _GNUC_VER >= 409
+#define _LIBCUDACXX_HAS_REFERENCE_QUALIFIED_FUNCTIONS
+#endif
+
+#if _GNUC_VER >= 501
+#define _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(...) __is_trivially_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE(...) __is_trivially_assignable(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_COPYABLE(...) __is_trivially_copyable(__VA_ARGS__)
+#endif
+
+// FIXME: GCC 8.0 supports this trait, but it has a bug.
+// https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91592
+// https://godbolt.org/z/IljfIw
+//#if _GNUC_VER >= 800
+//#define _LIBCUDACXX_IS_ASSIGNABLE(...) __is_assignable(__VA_ARGS__)
+//#endif
+
+#if _GNUC_VER >= 1000
+// __is_same_as is useless: GCC complains when you try to use it in a SFINAE condition
+// which is the primary place where libc++ tries to use it. Disable.
+// #define _LIBCUDACXX_IS_SAME(...) __is_same_as(__VA_ARGS__)
+#endif
+
+#if !__EXCEPTIONS && !defined(_LIBCUDACXX_NO_EXCEPTIONS)
+#define _LIBCUDACXX_NO_EXCEPTIONS
+#endif
+
+// Determine if GCC supports relaxed constexpr
+#if !defined(__cpp_constexpr) || __cpp_constexpr < 201304L
+#define _LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR
+#endif
+
+// GCC 5 supports variable templates
+#if !defined(__cpp_variable_templates) || __cpp_variable_templates < 201304L
+#define _LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES
+#endif
+
+#ifndef __GXX_EXPERIMENTAL_CXX0X__
+
+#define _LIBCUDACXX_HAS_NO_DECLTYPE
+#define _LIBCUDACXX_HAS_NO_NULLPTR
+#define _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+#define _LIBCUDACXX_HAS_NO_VARIADICS
+#define _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+#define _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+#define _LIBCUDACXX_HAS_NO_NOEXCEPT
+
+#else  // __GXX_EXPERIMENTAL_CXX0X__
+
+#if _GNUC_VER < 403
+#define _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+#endif
+
+#if _GNUC_VER < 404
+#define _LIBCUDACXX_HAS_NO_DECLTYPE
+#define _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+#define _LIBCUDACXX_HAS_NO_VARIADICS
+#define _LIBCUDACXX_HAS_NO_GENERALIZED_INITIALIZERS
+#endif  // _GNUC_VER < 404
+
+#if _GNUC_VER < 406
+#define _LIBCUDACXX_HAS_NO_NOEXCEPT
+#define _LIBCUDACXX_HAS_NO_NULLPTR
+#endif
+
+#endif  // __GXX_EXPERIMENTAL_CXX0X__
+
+#if !defined(_LIBCUDACXX_HAS_NO_ASAN) && !defined(__SANITIZE_ADDRESS__)
+#define _LIBCUDACXX_HAS_NO_ASAN
+#endif
+
+#if _GNUC_VER < 600 && defined(_LIBCUDACXX_COMPILER_NVCC)
+#define _LIBCUDACXX_MISSING_GCC_MATH_INTRINSICS
+#endif
+
+#if _GNUC_VER >= 700
+#define _LIBCUDACXX_LAUNDER(...) __builtin_launder(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_UNIQUE_OBJECT_REPRESENTATIONS(...) __has_unique_object_representations(__VA_ARGS__)
+#define _LIBCUDACXX_IS_AGGREGATE(...) __is_aggregate(__VA_ARGS__)
+#define _LIBCUDACXX_ADDRESSOF(...) __builtin_addressof(__VA_ARGS__)
+#endif
+
+#if !defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+#define _LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE(...) __atomic_always_lock_free(__VA_ARGS__)
+#endif
+
+// NVCC in C++11 mode freaks out about `__builtin_is_constant_evaluated`.
+#if _GNUC_VER >= 900 && !(defined(_LIBCUDACXX_COMPILER_NVCC) && _LIBCUDACXX_STD_VER < 14)
+#define _LIBCUDACXX_IS_CONSTANT_EVALUATED(...) __builtin_is_constant_evaluated(__VA_ARGS__)
+#endif
+
+// NVCC cannot properly handle some deductions occuring within NOEXCEPT
+// C++17 mode causes reference instatiation errors in tuple
+#if (_GNUC_VER >= 702 && _GNUC_VER <= 805)
+#if defined(_LIBCUDACXX_COMPILER_NVCC) && _LIBCUDACXX_STD_VER == 17
+#define _LIBCUDACXX_NO_TUPLE_NOEXCEPT
+#endif
+#endif
+
+#define _LIBCUDACXX_ALWAYS_INLINE __attribute__ ((__always_inline__))
+
+#define _LIBCUDACXX_DISABLE_EXTENSION_WARNING __extension__
+
+#elif defined(_LIBCUDACXX_COMPILER_MSVC)
+
+#define _LIBCUDACXX_TOSTRING2(x) #x
+#define _LIBCUDACXX_TOSTRING(x) _LIBCUDACXX_TOSTRING2(x)
+#define _LIBCUDACXX_WARNING(x) __pragma(message(__FILE__ "(" _LIBCUDACXX_TOSTRING(__LINE__) ") : warning note: " x))
+
+#define _LIBCUDACXX_ADDRESSOF(...) __builtin_addressof(__VA_ARGS__)
+
+// https://github.com/microsoft/STL/blob/master/stl/inc/yvals_core.h#L353
+// warning C4100: 'quack': unreferenced formal parameter
+// warning C4127: conditional expression is constant
+// warning C4180: qualifier applied to function type has no meaning; ignored
+// warning C4197: 'purr': top-level volatile in cast is ignored
+// warning C4324: 'roar': structure was padded due to alignment specifier
+// warning C4455: literal suffix identifiers that do not start with an underscore are reserved
+// warning C4503: 'hum': decorated name length exceeded, name was truncated
+// warning C4522: 'woof' : multiple assignment operators specified
+// warning C4668: 'meow' is not defined as a preprocessor macro, replacing with '0' for '#if/#elif'
+// warning C4800: 'boo': forcing value to bool 'true' or 'false' (performance warning)
+// warning C4996: 'meow': was declared deprecated
+#define _LIBCUDACXX_MSVC_DISABLED_WARNINGS \
+  4100 \
+  4127 \
+  4180 \
+  4197 \
+  4296 \
+  4324 \
+  4455 \
+  4503 \
+  4522 \
+  4668 \
+  4800 \
+  4996 \
+  /**/
+
+#if _MSC_VER < 1900
+#error "MSVC versions prior to Visual Studio 2015 are not supported"
+#endif
+
+// MSVC implemented P0030R1 in 15.7, only available under C++17
+#if _MSC_VER < 1914
+#define _LIBCUDACXX_NO_HOST_CPP17_HYPOT
+#endif
+
+#if _MSC_VER < 1920
+#define _LIBCUDACXX_HAS_NO_NOEXCEPT_SFINAE
+#define _LIBCUDACXX_HAS_NO_PRAGMA_PUSH_POP_MACRO // This is conservative but correct.
+#define _LIBCUDACXX_HAS_NO_LOGICAL_METAFUNCTION_ALIASES
+#endif
+
+// MSVC exposed __iso_volatile intrinsics beginning on 1924 for x86
+#if _MSC_VER < 1924
+    #define _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+#endif
+
+#if _MSC_VER > 1924
+#  define _LIBCUDACXX_IS_CONSTANT_EVALUATED(...) __builtin_is_constant_evaluated(__VA_ARGS__)
+#endif
+
+#define _LIBCUDACXX_UNDERLYING_TYPE(...) __underlying_type(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CONSTRUCTIBLE(...) __is_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_DESTRUCTIBLE(...) __is_destructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(...) __is_trivially_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR(...) __has_trivial_constructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_NOTHROW_CONSTRUCTIBLE(...) __is_nothrow_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR(...) __has_nothrow_constructor(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_COPY(...) __has_nothrow_copy(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_ASSIGN(...) __has_nothrow_assign(__VA_ARGS__)
+#define _LIBCUDACXX_IS_NOTHROW_ASSIGNABLE(...) __is_nothrow_assignable(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR(...) __has_trivial_destructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_NOTHROW_DESTRUCTIBLE(...) __is_nothrow_destructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_DESTRUCTIBLE(...) __is_trivially_destructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE(...) __is_trivially_assignable(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_COPYABLE(...) __is_trivially_copyable(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIAL(...) __is_trivial(__VA_ARGS__)
+#define _LIBCUDACXX_IS_LITERAL(...) __is_literal_type(__VA_ARGS__)
+#define _LIBCUDACXX_IS_UNION(...) __is_union(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CLASS(...) __is_class(__VA_ARGS__)
+#if _MSC_VER > 1914 // MSVC doesn't include __is_aggregate until 19.15
+#define _LIBCUDACXX_IS_AGGREGATE(...) __is_aggregate(__VA_ARGS__)
+#endif // _MSC_VER > 1914
+#define _LIBCUDACXX_IS_POD(...) __is_pod(__VA_ARGS__)
+#define _LIBCUDACXX_IS_STANDARD_LAYOUT(...) __is_standard_layout(__VA_ARGS__)
+#define _LIBCUDACXX_IS_ENUM(...) __is_enum(__VA_ARGS__)
+#define _LIBCUDACXX_IS_EMPTY(...) __is_empty(__VA_ARGS__)
+#define _LIBCUDACXX_IS_POLYMORPHIC(...) __is_polymorphic(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR(...) __has_virtual_destructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_BASE_OF(...) __is_base_of(__VA_ARGS__)
+#define _LIBCUDACXX_IS_FINAL(...) __is_final(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CONVERTIBLE_TO(...) __is_convertible_to(__VA_ARGS__)
+#define _LIBCUDACXX_IS_ASSIGNABLE(...) __is_assignable(__VA_ARGS__)
+#define __alignof__ __alignof
+#define _LIBCUDACXX_NORETURN __declspec(noreturn)
+#define _ALIGNAS(x) __declspec(align(x))
+#define _ALIGNAS_TYPE(x) alignas(x)
+
+#if _LIBCUDACXX_STD_VER < 14
+#define _LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR
+#define _LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES
+#endif
+
+#define _LIBCUDACXX_WEAK
+
+#define _LIBCUDACXX_HAS_NO_ASAN
+
+#define _LIBCUDACXX_ALWAYS_INLINE __forceinline
+
+#define _LIBCUDACXX_HAS_NO_VECTOR_EXTENSION
+
+#define _LIBCUDACXX_HAS_REFERENCE_QUALIFIED_FUNCTIONS
+#define _LIBCUDACXX_DISABLE_EXTENSION_WARNING
+
+#elif defined(_LIBCUDACXX_COMPILER_IBM)
+
+#define _ALIGNAS(x) __attribute__((__aligned__(x)))
+#define _ALIGNAS_TYPE(x) __attribute__((__aligned__(_LIBCUDACXX_ALIGNOF(x))))
+#define _ATTRIBUTE(x) __attribute__((x))
+#define _LIBCUDACXX_NORETURN __attribute__((noreturn))
+
+#define _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+#define _LIBCUDACXX_IS_BASE_OF(...) __is_base_of(__VA_ARGS__)
+#define _LIBCUDACXX_IS_FINAL(...) __is_final(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES
+
+#if defined(_AIX)
+#define __MULTILOCALE_API
+#endif
+
+#define _LIBCUDACXX_HAS_NO_ASAN
+
+#define _LIBCUDACXX_ALWAYS_INLINE __attribute__ ((__always_inline__))
+
+#define _LIBCUDACXX_HAS_NO_VECTOR_EXTENSION
+
+#elif defined(_LIBCUDACXX_COMPILER_NVRTC) || defined(_LIBCUDACXX_COMPILER_PGI)
+
+#if !defined(__cpp_constexpr) || __cpp_constexpr < 201304L
+#define _LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR
+#endif
+
+#if !defined(__cpp_variable_templates) || __cpp_variable_templates < 201304L
+#define _LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES
+#endif
+
+#define _ALIGNAS_TYPE(x) alignas(x)
+#define _ALIGNAS(x) alignas(x)
+
+#define _LIBCUDACXX_UNDERLYING_TYPE(...) __underlying_type(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CONSTRUCTIBLE(...) __is_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(...) __is_trivially_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR(...) __has_trivial_constructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_NOTHROW_CONSTRUCTIBLE(...) __is_nothrow_constructible(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR(...) __has_nothrow_constructor(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_COPY(...) __has_nothrow_copy(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_NOTHROW_ASSIGN(...) __has_nothrow_assign(__VA_ARGS__)
+#define _LIBCUDACXX_IS_NOTHROW_ASSIGNABLE(...) __is_nothrow_assignable(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR(...) __has_trivial_destructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_NOTHROW_DESTRUCTIBLE(...) __is_nothrow_destructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_DESTRUCTIBLE(...) __is_trivially_destructible(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE(...) __is_trivially_assignable(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIALLY_COPYABLE(...) __is_trivially_copyable(__VA_ARGS__)
+#define _LIBCUDACXX_IS_TRIVIAL(...) __is_trivial(__VA_ARGS__)
+#define _LIBCUDACXX_IS_LITERAL(...) __is_literal_type(__VA_ARGS__)
+#define _LIBCUDACXX_IS_UNION(...) __is_union(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CLASS(...) __is_class(__VA_ARGS__)
+#define _LIBCUDACXX_IS_AGGREGATE(...) __is_aggregate(__VA_ARGS__)
+#define _LIBCUDACXX_IS_POD(...) __is_pod(__VA_ARGS__)
+#define _LIBCUDACXX_IS_STANDARD_LAYOUT(...) __is_standard_layout(__VA_ARGS__)
+#define _LIBCUDACXX_IS_ENUM(...) __is_enum(__VA_ARGS__)
+#define _LIBCUDACXX_IS_EMPTY(...) __is_empty(__VA_ARGS__)
+#define _LIBCUDACXX_IS_POLYMORPHIC(...) __is_polymorphic(__VA_ARGS__)
+#define _LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR(...) __has_virtual_destructor(__VA_ARGS__)
+#define _LIBCUDACXX_IS_BASE_OF(...) __is_base_of(__VA_ARGS__)
+#define _LIBCUDACXX_IS_FINAL(...) __is_final(__VA_ARGS__)
+#define _LIBCUDACXX_IS_CONVERTIBLE_TO(...) __is_convertible_to(__VA_ARGS__)
+#define _LIBCUDACXX_DISABLE_EXTENSION_WARNING
+
+#define _LIBCUDACXX_HAS_REFERENCE_QUALIFIED_FUNCTIONS
+
+#define _LIBCUDACXX_HAS_NO_ASAN
+
+#define _LIBCUDACXX_ALWAYS_INLINE __attribute__ ((__always_inline__))
+#define _LIBCUDACXX_NORETURN __attribute__ ((noreturn))
+
+#define _GLIBCXX_INCLUDE_NEXT_C_HEADERS
+
+#endif // _LIBCUDACXX_COMPILER_[CLANG|GCC|MSVC|IBM|NVRTC]
+
+#if defined(_LIBCUDACXX_COMPILER_PGI) && !defined(__cuda_std__)
+// Forcefully disable visibility controls when used as the standard library with NVC++.
+// TODO: reevaluate.
+#define _LIBCUDACXX_HIDE_FROM_ABI
+#ifndef _LIBCUDACXX_DISABLE_EXTERN_TEMPLATE
+#define _LIBCUDACXX_DISABLE_EXTERN_TEMPLATE
+#endif
+#define _LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS
+#endif
+
+#if defined(_LIBCUDACXX_OBJECT_FORMAT_COFF)
+
+#ifdef _DLL
+#  define _LIBCUDACXX_CRT_FUNC __declspec(dllimport)
+#else
+#  define _LIBCUDACXX_CRT_FUNC
+#endif
+
+#if defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#  define _LIBCUDACXX_DLL_VIS
+#  define _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS
+#  define _LIBCUDACXX_CLASS_TEMPLATE_INSTANTIATION_VIS
+#  define _LIBCUDACXX_OVERRIDABLE_FUNC_VIS
+#  define _LIBCUDACXX_EXPORTED_FROM_ABI
+#elif defined(_LIBCUDACXX_BUILDING_LIBRARY)
+#  define _LIBCUDACXX_DLL_VIS __declspec(dllexport)
+#  if defined(__MINGW32__)
+#    define _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS _LIBCUDACXX_DLL_VIS
+#    define _LIBCUDACXX_CLASS_TEMPLATE_INSTANTIATION_VIS
+#  else
+#    define _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS
+#    define _LIBCUDACXX_CLASS_TEMPLATE_INSTANTIATION_VIS _LIBCUDACXX_DLL_VIS
+#  endif
+#  define _LIBCUDACXX_OVERRIDABLE_FUNC_VIS _LIBCUDACXX_DLL_VIS
+#  define _LIBCUDACXX_EXPORTED_FROM_ABI __declspec(dllexport)
+#else
+#  define _LIBCUDACXX_DLL_VIS __declspec(dllimport)
+#  define _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS _LIBCUDACXX_DLL_VIS
+#  define _LIBCUDACXX_CLASS_TEMPLATE_INSTANTIATION_VIS
+#  define _LIBCUDACXX_OVERRIDABLE_FUNC_VIS
+#  define _LIBCUDACXX_EXPORTED_FROM_ABI __declspec(dllimport)
+#endif
+
+#define _LIBCUDACXX_TYPE_VIS            _LIBCUDACXX_DLL_VIS
+#define _LIBCUDACXX_FUNC_VIS            _LIBCUDACXX_DLL_VIS
+#define _LIBCUDACXX_EXCEPTION_ABI       _LIBCUDACXX_DLL_VIS
+#define _LIBCUDACXX_HIDDEN
+#define _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+#define _LIBCUDACXX_TEMPLATE_VIS
+#define _LIBCUDACXX_ENUM_VIS
+
+#endif // defined(_LIBCUDACXX_OBJECT_FORMAT_COFF)
+
+#ifndef _LIBCUDACXX_HIDDEN
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#    define _LIBCUDACXX_HIDDEN __attribute__ ((__visibility__("hidden")))
+#  else
+#    define _LIBCUDACXX_HIDDEN
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+// The inline should be removed once PR32114 is resolved
+#    define _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS inline _LIBCUDACXX_HIDDEN
+#  else
+#    define _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_FUNC_VIS
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#    define _LIBCUDACXX_FUNC_VIS __attribute__ ((__visibility__("default")))
+#  else
+#    define _LIBCUDACXX_FUNC_VIS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_TYPE_VIS
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#    define _LIBCUDACXX_TYPE_VIS __attribute__ ((__visibility__("default")))
+#  else
+#    define _LIBCUDACXX_TYPE_VIS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_TEMPLATE_VIS
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#    if __has_attribute(__type_visibility__)
+#      define _LIBCUDACXX_TEMPLATE_VIS __attribute__ ((__type_visibility__("default")))
+#    else
+#      define _LIBCUDACXX_TEMPLATE_VIS __attribute__ ((__visibility__("default")))
+#    endif
+#  else
+#    define _LIBCUDACXX_TEMPLATE_VIS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_EXPORTED_FROM_ABI
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#    define _LIBCUDACXX_EXPORTED_FROM_ABI __attribute__((__visibility__("default")))
+#  else
+#    define _LIBCUDACXX_EXPORTED_FROM_ABI
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_OVERRIDABLE_FUNC_VIS
+#define _LIBCUDACXX_OVERRIDABLE_FUNC_VIS _LIBCUDACXX_FUNC_VIS
+#endif
+
+#ifndef _LIBCUDACXX_EXCEPTION_ABI
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS)
+#    define _LIBCUDACXX_EXCEPTION_ABI __attribute__ ((__visibility__("default")))
+#  else
+#    define _LIBCUDACXX_EXCEPTION_ABI
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_ENUM_VIS
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS) && __has_attribute(__type_visibility__)
+#    define _LIBCUDACXX_ENUM_VIS __attribute__ ((__type_visibility__("default")))
+#  else
+#    define _LIBCUDACXX_ENUM_VIS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS
+#  if !defined(_LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS) && __has_attribute(__type_visibility__)
+#    define _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __attribute__ ((__visibility__("default")))
+#  else
+#    define _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_CLASS_TEMPLATE_INSTANTIATION_VIS
+#define _LIBCUDACXX_CLASS_TEMPLATE_INSTANTIATION_VIS
+#endif
+
+#if __has_attribute(internal_linkage)
+#  define _LIBCUDACXX_INTERNAL_LINKAGE __attribute__ ((internal_linkage))
+#else
+#  define _LIBCUDACXX_INTERNAL_LINKAGE _LIBCUDACXX_ALWAYS_INLINE
+#endif
+
+#if __has_attribute(exclude_from_explicit_instantiation)
+#  define _LIBCUDACXX_EXCLUDE_FROM_EXPLICIT_INSTANTIATION __attribute__ ((__exclude_from_explicit_instantiation__))
+#else
+   // Try to approximate the effect of exclude_from_explicit_instantiation
+   // (which is that entities are not assumed to be provided by explicit
+   // template instantiations in the dylib) by always inlining those entities.
+#  define _LIBCUDACXX_EXCLUDE_FROM_EXPLICIT_INSTANTIATION _LIBCUDACXX_ALWAYS_INLINE
+#endif
+
+#ifndef _LIBCUDACXX_HIDE_FROM_ABI_PER_TU
+#  ifndef _LIBCUDACXX_HIDE_FROM_ABI_PER_TU_BY_DEFAULT
+#    define _LIBCUDACXX_HIDE_FROM_ABI_PER_TU 0
+#  else
+#    define _LIBCUDACXX_HIDE_FROM_ABI_PER_TU 1
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT
+# ifdef _LIBCUDACXX_OBJECT_FORMAT_COFF // Windows binaries can't merge typeinfos.
+# define _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT 0
+#else
+// TODO: This isn't strictly correct on ELF platforms due to llvm.org/PR37398
+// And we should consider defaulting to OFF.
+# define _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT 1
+#endif
+#endif
+
+#ifndef _LIBCUDACXX_HIDE_FROM_ABI
+#  if _LIBCUDACXX_HIDE_FROM_ABI_PER_TU
+#    define _LIBCUDACXX_HIDE_FROM_ABI _LIBCUDACXX_HIDDEN _LIBCUDACXX_INTERNAL_LINKAGE
+#  else
+#    define _LIBCUDACXX_HIDE_FROM_ABI _LIBCUDACXX_HIDDEN _LIBCUDACXX_EXCLUDE_FROM_EXPLICIT_INSTANTIATION
+#  endif
+#endif
+
+#ifdef _LIBCUDACXX_BUILDING_LIBRARY
+#  if _LIBCUDACXX_ABI_VERSION > 1
+#    define _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1 _LIBCUDACXX_HIDE_FROM_ABI
+#  else
+#    define _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+#  endif
+#else
+#  define _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1 _LIBCUDACXX_HIDE_FROM_ABI
+#endif
+
+// Just so we can migrate to the new macros gradually.
+#define _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_HIDE_FROM_ABI
+
+#define _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+
+// Inline namespaces are available in Clang/GCC/MSVC regardless of C++ dialect.
+#define _LIBCUDACXX_BEGIN_NAMESPACE_STD namespace std { inline namespace _LIBCUDACXX_ABI_NAMESPACE {
+#define _LIBCUDACXX_END_NAMESPACE_STD  } }
+#define _CUDA_VSTD std::_LIBCUDACXX_ABI_NAMESPACE
+_LIBCUDACXX_BEGIN_NAMESPACE_STD _LIBCUDACXX_END_NAMESPACE_STD
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION namespace std {
+#define _LIBCUDACXX_END_NAMESPACE_STD_NOVERSION }
+
+#if _LIBCUDACXX_STD_VER >= 17
+#define _LIBCUDACXX_BEGIN_NAMESPACE_FILESYSTEM \
+  _LIBCUDACXX_BEGIN_NAMESPACE_STD inline namespace __fs { namespace filesystem {
+#else
+#define _LIBCUDACXX_BEGIN_NAMESPACE_FILESYSTEM \
+  _LIBCUDACXX_BEGIN_NAMESPACE_STD namespace __fs { namespace filesystem {
+#endif
+
+#define _LIBCUDACXX_END_NAMESPACE_FILESYSTEM \
+  _LIBCUDACXX_END_NAMESPACE_STD } }
+
+#define _CUDA_VSTD_FS _CUDA_VSTD::__fs::filesystem
+
+#ifndef _LIBCUDACXX_PREFERRED_OVERLOAD
+#  if __has_attribute(__enable_if__)
+#    define _LIBCUDACXX_PREFERRED_OVERLOAD __attribute__ ((__enable_if__(true, "")))
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_NOEXCEPT
+#  define _NOEXCEPT noexcept
+#  define _NOEXCEPT_(x) noexcept(x)
+#else
+#  define _NOEXCEPT throw()
+#  define _NOEXCEPT_(x)
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+typedef unsigned short char16_t;
+typedef unsigned int   char32_t;
+#endif  // _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+
+#ifndef __SIZEOF_INT128__
+#define _LIBCUDACXX_HAS_NO_INT128
+#endif
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+# define static_assert(...) _Static_assert(__VA_ARGS__)
+# define decltype(...) __decltype(__VA_ARGS__)
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+#  define _LIBCUDACXX_CONSTEXPR
+#else
+#  define _LIBCUDACXX_CONSTEXPR constexpr
+#endif
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+#  define _LIBCUDACXX_DEFAULT {}
+#else
+#  define _LIBCUDACXX_DEFAULT = default;
+#endif
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+#  define _LIBCUDACXX_EQUAL_DELETE
+#else
+#  define _LIBCUDACXX_EQUAL_DELETE = delete
+#endif
+
+#ifdef __GNUC__
+#  define _LIBCUDACXX_NOALIAS __attribute__((__malloc__))
+#else
+#  define _LIBCUDACXX_NOALIAS
+#endif
+
+#if __has_feature(cxx_explicit_conversions) || defined(__IBMCPP__) || \
+    (!defined(_LIBCUDACXX_CXX03_LANG) && defined(__GNUC__)) // All supported GCC versions
+#  define _LIBCUDACXX_EXPLICIT explicit
+#else
+#  define _LIBCUDACXX_EXPLICIT
+#endif
+
+#if !__has_builtin(__builtin_operator_new) || !__has_builtin(__builtin_operator_delete)
+#define _LIBCUDACXX_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+#  define _LIBCUDACXX_DECLARE_STRONG_ENUM(x) struct _LIBCUDACXX_TYPE_VIS x { enum __lx
+#  define _LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(x) \
+     __lx __v_; \
+     _LIBCUDACXX_INLINE_VISIBILITY x(__lx __v) : __v_(__v) {} \
+     _LIBCUDACXX_INLINE_VISIBILITY explicit x(int __v) : __v_(static_cast<__lx>(__v)) {} \
+     _LIBCUDACXX_INLINE_VISIBILITY operator int() const {return __v_;} \
+     };
+#else  // _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+#  define _LIBCUDACXX_DECLARE_STRONG_ENUM(x) enum class _LIBCUDACXX_ENUM_VIS x
+#  define _LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(x)
+#endif  // _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+
+#ifdef _LIBCUDACXX_DEBUG
+#  if _LIBCUDACXX_DEBUG == 0
+#    define _LIBCUDACXX_DEBUG_LEVEL 1
+#  elif _LIBCUDACXX_DEBUG == 1
+#    define _LIBCUDACXX_DEBUG_LEVEL 2
+#  else
+#    error Supported values for _LIBCUDACXX_DEBUG are 0 and 1
+#  endif
+#  if !defined(_LIBCUDACXX_BUILDING_LIBRARY)
+#    define _LIBCUDACXX_EXTERN_TEMPLATE(...)
+#  endif
+#endif
+
+#ifdef _LIBCUDACXX_DISABLE_EXTERN_TEMPLATE
+#define _LIBCUDACXX_EXTERN_TEMPLATE(...)
+#define _LIBCUDACXX_EXTERN_TEMPLATE2(...)
+#endif
+
+#ifndef _LIBCUDACXX_EXTERN_TEMPLATE
+#define _LIBCUDACXX_EXTERN_TEMPLATE(...) extern template __VA_ARGS__;
+#endif
+
+#ifndef _LIBCUDACXX_EXTERN_TEMPLATE2
+#define _LIBCUDACXX_EXTERN_TEMPLATE2(...) extern template __VA_ARGS__;
+#endif
+
+#if defined(__APPLE__) || defined(__FreeBSD__) || defined(_LIBCUDACXX_MSVCRT_LIKE) || \
+    defined(__sun__) || defined(__NetBSD__) || defined(__CloudABI__)
+#define _LIBCUDACXX_LOCALE__L_EXTENSIONS 1
+#endif
+
+#if defined(__unix__) || (defined(__APPLE__) && defined(__MACH__))
+// Most unix variants have catopen.  These are the specific ones that don't.
+#  if !defined(__BIONIC__) && !defined(_NEWLIB_VERSION)
+#    define _LIBCUDACXX_HAS_CATOPEN 1
+#  endif
+#endif
+
+#ifdef __FreeBSD__
+#define _DECLARE_C99_LDBL_MATH 1
+#endif
+
+#if defined(_LIBCUDACXX_ABI_MICROSOFT) && !defined(_LIBCUDACXX_NO_VCRUNTIME)
+#  define _LIBCUDACXX_DEFER_NEW_TO_VCRUNTIME
+#endif
+
+// If we are getting operator new from the MSVC CRT, then allocation overloads
+// for align_val_t were added in 19.12, aka VS 2017 version 15.3.
+#if defined(_LIBCUDACXX_MSVCRT) && defined(_MSC_VER) && _MSC_VER < 1912
+#  define _LIBCUDACXX_HAS_NO_LIBRARY_ALIGNED_ALLOCATION
+#elif defined(_LIBCUDACXX_ABI_VCRUNTIME) && !defined(__cpp_aligned_new)
+   // We're deferring to Microsoft's STL to provide aligned new et al. We don't
+   // have it unless the language feature test macro is defined.
+#  define _LIBCUDACXX_HAS_NO_LIBRARY_ALIGNED_ALLOCATION
+#endif
+
+#if defined(__APPLE__)
+#  if !defined(__MAC_OS_X_VERSION_MIN_REQUIRED) && \
+      defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__)
+#    define __MAC_OS_X_VERSION_MIN_REQUIRED __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__
+#  endif
+#endif // defined(__APPLE__)
+
+#if !defined(_LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION) && \
+    (defined(_LIBCUDACXX_HAS_NO_LIBRARY_ALIGNED_ALLOCATION) || \
+    (!defined(__cpp_aligned_new) || __cpp_aligned_new < 201606))
+#  define _LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION
+#endif
+
+#if defined(__APPLE__) || defined(__FreeBSD__)
+#define _LIBCUDACXX_HAS_DEFAULTRUNELOCALE
+#endif
+
+#if defined(__APPLE__) || defined(__FreeBSD__) || defined(__sun__)
+#define _LIBCUDACXX_WCTYPE_IS_MASK
+#endif
+
+#if _LIBCUDACXX_STD_VER <= 17 || !defined(__cpp_char8_t)
+#define _LIBCUDACXX_NO_HAS_CHAR8_T
+#endif
+
+// Deprecation macros.
+//
+// Deprecations warnings are always enabled, except when users explicitly opt-out
+// by defining _LIBCUDACXX_DISABLE_DEPRECATION_WARNINGS.
+#if !defined(_LIBCUDACXX_DISABLE_DEPRECATION_WARNINGS)
+#  if __has_attribute(deprecated)
+#    define _LIBCUDACXX_DEPRECATED __attribute__ ((deprecated))
+#  elif _LIBCUDACXX_STD_VER > 11
+#    define _LIBCUDACXX_DEPRECATED [[deprecated]]
+#  else
+#    define _LIBCUDACXX_DEPRECATED
+#  endif
+#else
+#  define _LIBCUDACXX_DEPRECATED
+#endif
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+#  define _LIBCUDACXX_DEPRECATED_IN_CXX11 _LIBCUDACXX_DEPRECATED
+#else
+#  define _LIBCUDACXX_DEPRECATED_IN_CXX11
+#endif
+
+#if _LIBCUDACXX_STD_VER >= 14
+#  define _LIBCUDACXX_DEPRECATED_IN_CXX14 _LIBCUDACXX_DEPRECATED
+#else
+#  define _LIBCUDACXX_DEPRECATED_IN_CXX14
+#endif
+
+#if _LIBCUDACXX_STD_VER >= 17
+#  define _LIBCUDACXX_DEPRECATED_IN_CXX17 _LIBCUDACXX_DEPRECATED
+#else
+#  define _LIBCUDACXX_DEPRECATED_IN_CXX17
+#endif
+
+#if _LIBCUDACXX_STD_VER <= 11
+#  define _LIBCUDACXX_EXPLICIT_AFTER_CXX11
+#else
+#  define _LIBCUDACXX_EXPLICIT_AFTER_CXX11 explicit
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR)
+#  define _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 constexpr
+#else
+#  define _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+#endif
+
+#if _LIBCUDACXX_STD_VER > 14 && !defined(_LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR)
+#  define _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 constexpr
+#else
+#  define _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17 && !defined(_LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR)
+#  define _LIBCUDACXX_CONSTEXPR_AFTER_CXX17 constexpr
+#else
+#  define _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+#endif
+
+// Macros to enter and leave a state where deprecation warnings are suppressed.
+#if defined(_LIBCUDACXX_COMPILER_CLANG) || defined(_LIBCUDACXX_COMPILER_GCC)
+#   define _LIBCUDACXX_SUPPRESS_DEPRECATED_PUSH \
+        _Pragma("GCC diagnostic push") \
+        _Pragma("GCC diagnostic ignored \"-Wdeprecated\"") \
+        _Pragma("GCC diagnostic ignored \"-Wdeprecated-declarations\"")
+#   define _LIBCUDACXX_SUPPRESS_DEPRECATED_POP \
+        _Pragma("GCC diagnostic pop")
+#else
+#   define _LIBCUDACXX_SUPPRESS_DEPRECATED_PUSH
+#   define _LIBCUDACXX_SUPPRESS_DEPRECATED_POP
+#endif
+
+// The _LIBCUDACXX_NODISCARD_ATTRIBUTE should only be used to define other
+// NODISCARD macros to the correct attribute.
+#if __has_cpp_attribute(nodiscard) || (defined(_LIBCUDACXX_COMPILER_MSVC) && _LIBCUDACXX_STD_VER > 14)
+#  define _LIBCUDACXX_NODISCARD_ATTRIBUTE [[nodiscard]]
+#elif defined(_LIBCUDACXX_COMPILER_CLANG) && !defined(_LIBCUDACXX_CXX03_LANG)
+#  define _LIBCUDACXX_NODISCARD_ATTRIBUTE [[clang::warn_unused_result]]
+#else
+// We can't use GCC's [[gnu::warn_unused_result]] and
+// __attribute__((warn_unused_result)), because GCC does not silence them via
+// (void) cast.
+#  define _LIBCUDACXX_NODISCARD_ATTRIBUTE
+#endif
+
+// _LIBCUDACXX_NODISCARD_EXT may be used to apply [[nodiscard]] to entities not
+// specified as such as an extension.
+#if defined(_LIBCUDACXX_ENABLE_NODISCARD) && !defined(_LIBCUDACXX_DISABLE_NODISCARD_EXT)
+#  define _LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_NODISCARD_ATTRIBUTE
+#else
+#  define _LIBCUDACXX_NODISCARD_EXT
+#endif
+
+#if !defined(_LIBCUDACXX_DISABLE_NODISCARD_AFTER_CXX17) && \
+    (_LIBCUDACXX_STD_VER > 17 || defined(_LIBCUDACXX_ENABLE_NODISCARD))
+#  define _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_NODISCARD_ATTRIBUTE
+#else
+#  define _LIBCUDACXX_NODISCARD_AFTER_CXX17
+#endif
+
+#if _LIBCUDACXX_STD_VER > 14 && defined(__cpp_inline_variables) && (__cpp_inline_variables >= 201606L)
+#  define _LIBCUDACXX_INLINE_VAR inline
+#else
+#  define _LIBCUDACXX_INLINE_VAR
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+#  define _LIBCUDACXX_EXPLICIT_MOVE(x) _CUDA_VSTD::move(x)
+#else
+#  define _LIBCUDACXX_EXPLICIT_MOVE(x) (x)
+#endif
+
+#ifndef _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+#if defined(_LIBCUDACXX_DEBUG) || defined(_LIBCUDACXX_HAS_NO_CXX14_CONSTEXPR)
+#define _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+#else
+#define _LIBCUDACXX_CONSTEXPR_IF_NODEBUG constexpr
+#endif
+#endif
+
+#if __has_attribute(no_destroy)
+#  define _LIBCUDACXX_NO_DESTROY __attribute__((__no_destroy__))
+#else
+#  define _LIBCUDACXX_NO_DESTROY
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_ASAN
+extern "C" _LIBCUDACXX_FUNC_VIS void __sanitizer_annotate_contiguous_container(
+  const void *, const void *, const void *, const void *);
+#endif
+
+// Try to find out if RTTI is disabled.
+// g++ and cl.exe have RTTI on by default and define a macro when it is.
+// g++ only defines the macro in 4.3.2 and onwards.
+#if !defined(_LIBCUDACXX_NO_RTTI)
+#  if defined(__GNUC__) && \
+      ((__GNUC__ >= 5) || \
+       (__GNUC__ == 4 && (__GNUC_MINOR__ >= 3 || __GNUC_PATCHLEVEL__ >= 2))) && \
+      !defined(__GXX_RTTI)
+#    define _LIBCUDACXX_NO_RTTI
+#  elif defined(_LIBCUDACXX_COMPILER_MSVC) && !defined(_CPPRTTI)
+#    define _LIBCUDACXX_NO_RTTI
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_WEAK
+#define _LIBCUDACXX_WEAK __attribute__((__weak__))
+#endif
+
+// Thread API
+#if !defined(_LIBCUDACXX_HAS_NO_THREADS) && \
+    !defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD) && \
+    !defined(_LIBCUDACXX_HAS_THREAD_API_WIN32) && \
+    !defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+#  if defined(__FreeBSD__) || \
+      defined(__Fuchsia__) || \
+      defined(__wasi__) || \
+      defined(__NetBSD__) || \
+      defined(__linux__) || \
+      defined(__GNU__) || \
+      defined(__APPLE__) || \
+      defined(__CloudABI__) || \
+      defined(__sun__) || \
+      (defined(__MINGW32__) && __has_include(<pthread.h>))
+#    define _LIBCUDACXX_HAS_THREAD_API_PTHREAD
+#  elif defined(_LIBCUDACXX_WIN32API)
+#    define _LIBCUDACXX_HAS_THREAD_API_WIN32
+#  else
+#    error "No thread API"
+#  endif // _LIBCUDACXX_HAS_THREAD_API
+#endif // _LIBCUDACXX_HAS_NO_THREADS
+
+#if defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD)
+#if defined(__ANDROID__) && __ANDROID_API__ >= 30
+#define _LIBCUDACXX_HAS_COND_CLOCKWAIT
+#elif defined(_LIBCUDACXX_GLIBC_PREREQ)
+#if _LIBCUDACXX_GLIBC_PREREQ(2, 30)
+#define _LIBCUDACXX_HAS_COND_CLOCKWAIT
+#endif
+#endif
+#endif
+
+#if defined(_LIBCUDACXX_HAS_NO_THREADS) && defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD)
+#error _LIBCUDACXX_HAS_THREAD_API_PTHREAD may only be defined when \
+       _LIBCUDACXX_HAS_NO_THREADS is not defined.
+#endif
+
+#if defined(_LIBCUDACXX_HAS_NO_THREADS) && defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+#error _LIBCUDACXX_HAS_THREAD_API_EXTERNAL may not be defined when \
+       _LIBCUDACXX_HAS_NO_THREADS is defined.
+#endif
+
+#if defined(__STDCPP_THREADS__) && defined(_LIBCUDACXX_HAS_NO_THREADS)
+#error _LIBCUDACXX_HAS_NO_THREADS cannot be set when __STDCPP_THREADS__ is set.
+#endif
+
+#if !defined(_LIBCUDACXX_HAS_NO_THREADS) && !defined(__STDCPP_THREADS__)
+#define __STDCPP_THREADS__ 1
+#endif
+
+// The glibc and Bionic implementation of pthreads implements
+// pthread_mutex_destroy as nop for regular mutexes. Additionally, Win32
+// mutexes have no destroy mechanism.
+//
+// This optimization can't be performed on Apple platforms, where
+// pthread_mutex_destroy can allow the kernel to release resources.
+// See https://llvm.org/D64298 for details.
+//
+// TODO(EricWF): Enable this optimization on Bionic after speaking to their
+//               respective stakeholders.
+#if (defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD) && defined(__GLIBC__)) \
+  || defined(_LIBCUDACXX_HAS_THREAD_API_WIN32)
+# define _LIBCUDACXX_HAS_TRIVIAL_MUTEX_DESTRUCTION
+#endif
+
+// Destroying a condvar is a nop on Windows.
+//
+// This optimization can't be performed on Apple platforms, where
+// pthread_cond_destroy can allow the kernel to release resources.
+// See https://llvm.org/D64298 for details.
+//
+// TODO(EricWF): This is potentially true for some pthread implementations
+// as well.
+#if defined(_LIBCUDACXX_HAS_THREAD_API_WIN32)
+# define _LIBCUDACXX_HAS_TRIVIAL_CONDVAR_DESTRUCTION
+#endif
+
+// Systems that use capability-based security (FreeBSD with Capsicum,
+// Nuxi CloudABI) may only provide local filesystem access (using *at()).
+// Functions like open(), rename(), unlink() and stat() should not be
+// used, as they attempt to access the global filesystem namespace.
+#ifdef __CloudABI__
+#define _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+#endif
+
+// CloudABI is intended for running networked services. Processes do not
+// have standard input and output channels.
+#ifdef __CloudABI__
+#define _LIBCUDACXX_HAS_NO_STDIN
+#define _LIBCUDACXX_HAS_NO_STDOUT
+#endif
+
+// Some systems do not provide gets() in their C library, for security reasons.
+#ifndef _LIBCUDACXX_C_HAS_NO_GETS
+#  if defined(_LIBCUDACXX_MSVCRT) || (defined(__FreeBSD__) && __FreeBSD__ >= 13)
+#    define _LIBCUDACXX_C_HAS_NO_GETS
+#  endif
+#endif
+
+#if defined(__BIONIC__) || defined(__CloudABI__) ||                            \
+    defined(__Fuchsia__) || defined(__wasi__) || defined(_LIBCUDACXX_HAS_MUSL_LIBC)
+#define _LIBCUDACXX_PROVIDES_DEFAULT_RUNE_TABLE
+#endif
+
+// Thread-unsafe functions such as strtok() and localtime()
+// are not available.
+#ifdef __CloudABI__
+#define _LIBCUDACXX_HAS_NO_THREAD_UNSAFE_C_FUNCTIONS
+#endif
+
+// TODO: Support C11 Atomics?
+// #if __has_feature(cxx_atomic) || __has_extension(c_atomic) || __has_keyword(_Atomic)
+// #  define _LIBCUDACXX_HAS_C_ATOMIC_IMP
+#if defined(_LIBCUDACXX_COMPILER_CLANG)
+#  define _LIBCUDACXX_HAS_GCC_ATOMIC_IMP
+#elif defined(_LIBCUDACXX_COMPILER_GCC) || defined(_LIBCUDACXX_COMPILER_PGI)
+#  define _LIBCUDACXX_HAS_GCC_ATOMIC_IMP
+#elif defined(_LIBCUDACXX_COMPILER_PGI)
+#  define _LIBCUDACXX_HAS_GCC_ATOMIC_IMP
+#elif defined(_LIBCUDACXX_COMPILER_MSVC)
+#  define _LIBCUDACXX_HAS_MSVC_ATOMIC_IMPL
+#endif
+
+// CUDA Atomics supersede host atomics in order to insert the host/device dispatch layer
+#if defined(_LIBCUDACXX_COMPILER_NVCC) || defined(_LIBCUDACXX_COMPILER_NVRTC) || defined(_LIBCUDACXX_COMPILER_PGI)
+#  define _LIBCUDACXX_HAS_CUDA_ATOMIC_IMPL
+#endif
+
+#if (!defined(_LIBCUDACXX_HAS_C_ATOMIC_IMP) && \
+     !defined(_LIBCUDACXX_HAS_GCC_ATOMIC_IMP) && \
+     !defined(_LIBCUDACXX_HAS_EXTERNAL_ATOMIC_IMP)) \
+     || defined(_LIBCUDACXX_HAS_NO_THREADS)
+#  define _LIBCUDACXX_HAS_NO_ATOMIC_HEADER
+#else
+#  ifndef _LIBCUDACXX_ATOMIC_FLAG_TYPE
+#    define _LIBCUDACXX_ATOMIC_FLAG_TYPE bool
+#  endif
+#  ifdef _LIBCUDACXX_FREESTANDING
+#    define _LIBCUDACXX_ATOMIC_ONLY_USE_BUILTINS
+#  endif
+#endif
+
+#ifndef _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+#define _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+#endif
+
+#if defined(_LIBCUDACXX_ENABLE_THREAD_SAFETY_ANNOTATIONS)
+#  if defined(__clang__) && __has_attribute(acquire_capability)
+// Work around the attribute handling in clang.  When both __declspec and
+// __attribute__ are present, the processing goes awry preventing the definition
+// of the types.
+#    if !defined(_LIBCUDACXX_OBJECT_FORMAT_COFF)
+#      define _LIBCUDACXX_HAS_THREAD_SAFETY_ANNOTATIONS
+#    endif
+#  endif
+#endif
+
+#if __has_attribute(require_constant_initialization)
+#  define _LIBCUDACXX_SAFE_STATIC __attribute__((__require_constant_initialization__))
+#else
+#  define _LIBCUDACXX_SAFE_STATIC
+#endif
+
+#if !defined(_LIBCUDACXX_HAS_NO_OFF_T_FUNCTIONS)
+#  if defined(_LIBCUDACXX_MSVCRT) || defined(_NEWLIB_VERSION)
+#    define _LIBCUDACXX_HAS_NO_OFF_T_FUNCTIONS
+#  endif
+#endif
+
+#if __has_attribute(diagnose_if) && !defined(_LIBCUDACXX_DISABLE_ADDITIONAL_DIAGNOSTICS)
+#  define _LIBCUDACXX_DIAGNOSE_WARNING(...) \
+     __attribute__((diagnose_if(__VA_ARGS__, "warning")))
+#  define _LIBCUDACXX_DIAGNOSE_ERROR(...) \
+     __attribute__((diagnose_if(__VA_ARGS__, "error")))
+#else
+#  define _LIBCUDACXX_DIAGNOSE_WARNING(...)
+#  define _LIBCUDACXX_DIAGNOSE_ERROR(...)
+#endif
+
+// Use a function like macro to imply that it must be followed by a semicolon
+#if __cplusplus > 201402L && __has_cpp_attribute(fallthrough)
+#  define _LIBCUDACXX_FALLTHROUGH() [[fallthrough]]
+#elif defined(__CUDACC_RTC__)
+#  define _LIBCUDACXX_FALLTHROUGH() ((void)0)
+#elif __has_cpp_attribute(clang::fallthrough)
+#  define _LIBCUDACXX_FALLTHROUGH() [[clang::fallthrough]]
+#elif defined(_LIBCUDACXX_COMPILER_PGI)
+#  define _LIBCUDACXX_FALLTHROUGH()
+#elif __has_attribute(fallthough) || _GNUC_VER >= 700
+#  define _LIBCUDACXX_FALLTHROUGH() __attribute__((__fallthrough__))
+#else
+#  define _LIBCUDACXX_FALLTHROUGH() ((void)0)
+#endif
+
+#if __has_attribute(__nodebug__)
+#define _LIBCUDACXX_NODEBUG __attribute__((__nodebug__))
+#else
+#define _LIBCUDACXX_NODEBUG
+#endif
+
+#ifndef _LIBCUDACXX_NODEBUG_TYPE
+#if __has_attribute(__nodebug__) && \
+    (defined(_LIBCUDACXX_CLANG_VER) && _LIBCUDACXX_CLANG_VER >= 1210)
+#define _LIBCUDACXX_NODEBUG_TYPE __attribute__((nodebug))
+#else
+#define _LIBCUDACXX_NODEBUG_TYPE
+#endif
+#endif // !defined(_LIBCUDACXX_NODEBUG_TYPE)
+
+#if defined(_LIBCUDACXX_ABI_MICROSOFT) && \
+    (defined(_LIBCUDACXX_COMPILER_MSVC) || __has_declspec_attribute(empty_bases))
+#  define _LIBCUDACXX_DECLSPEC_EMPTY_BASES __declspec(empty_bases)
+#else
+#  define _LIBCUDACXX_DECLSPEC_EMPTY_BASES
+#endif
+
+#if defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_FEATURES)
+#define _LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR
+#define _LIBCUDACXX_ENABLE_CXX17_REMOVED_UNEXPECTED_FUNCTIONS
+#define _LIBCUDACXX_ENABLE_CXX17_REMOVED_RANDOM_SHUFFLE
+#define _LIBCUDACXX_ENABLE_CXX17_REMOVED_BINDERS
+#endif // _LIBCUDACXX_ENABLE_CXX17_REMOVED_FEATURES
+
+#if !defined(__cpp_deduction_guides) || __cpp_deduction_guides < 201611
+#define _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+#endif
+
+#if !defined(__cpp_coroutines) || __cpp_coroutines < 201703L
+#define _LIBCUDACXX_HAS_NO_COROUTINES
+#endif
+
+// FIXME: Correct this macro when either (A) a feature test macro for the
+// spaceship operator is provided, or (B) a compiler provides a complete
+// implementation.
+#define _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+
+// Decide whether to use availability macros.
+#if !defined(_LIBCUDACXX_BUILDING_LIBRARY) &&                                      \
+    !defined(_LIBCUDACXX_DISABLE_AVAILABILITY) &&                                  \
+    __has_feature(attribute_availability_with_strict) &&                       \
+    __has_feature(attribute_availability_in_templates) &&                      \
+    __has_extension(pragma_clang_attribute_external_declaration)
+#  ifdef __APPLE__
+#    define _LIBCUDACXX_USE_AVAILABILITY_APPLE
+#  endif
+#endif
+
+// Define availability macros.
+#if defined(_LIBCUDACXX_USE_AVAILABILITY_APPLE)
+#  define _LIBCUDACXX_AVAILABILITY_SHARED_MUTEX                                    \
+     __attribute__((availability(macosx,strict,introduced=10.12)))             \
+     __attribute__((availability(ios,strict,introduced=10.0)))                 \
+     __attribute__((availability(tvos,strict,introduced=10.0)))                \
+     __attribute__((availability(watchos,strict,introduced=3.0)))
+#  define _LIBCUDACXX_AVAILABILITY_BAD_OPTIONAL_ACCESS                             \
+     __attribute__((availability(macosx,strict,introduced=10.14)))             \
+     __attribute__((availability(ios,strict,introduced=12.0)))                 \
+     __attribute__((availability(tvos,strict,introduced=12.0)))                \
+     __attribute__((availability(watchos,strict,introduced=5.0)))
+#  define _LIBCUDACXX_AVAILABILITY_BAD_VARIANT_ACCESS                              \
+     _LIBCUDACXX_AVAILABILITY_BAD_OPTIONAL_ACCESS
+#  define _LIBCUDACXX_AVAILABILITY_BAD_ANY_CAST                                    \
+     _LIBCUDACXX_AVAILABILITY_BAD_OPTIONAL_ACCESS
+#  define _LIBCUDACXX_AVAILABILITY_UNCAUGHT_EXCEPTIONS                             \
+     __attribute__((availability(macosx,strict,introduced=10.12)))             \
+     __attribute__((availability(ios,strict,introduced=10.0)))                 \
+     __attribute__((availability(tvos,strict,introduced=10.0)))                \
+     __attribute__((availability(watchos,strict,introduced=3.0)))
+#  define _LIBCUDACXX_AVAILABILITY_SIZED_NEW_DELETE                                \
+     __attribute__((availability(macosx,strict,introduced=10.12)))             \
+     __attribute__((availability(ios,strict,introduced=10.0)))                 \
+     __attribute__((availability(tvos,strict,introduced=10.0)))                \
+     __attribute__((availability(watchos,strict,introduced=3.0)))
+#  define _LIBCUDACXX_AVAILABILITY_FUTURE_ERROR                                    \
+     __attribute__((availability(ios,strict,introduced=6.0)))
+#  define _LIBCUDACXX_AVAILABILITY_TYPEINFO_VTABLE                                 \
+     __attribute__((availability(macosx,strict,introduced=10.9)))              \
+     __attribute__((availability(ios,strict,introduced=7.0)))
+#  define _LIBCUDACXX_AVAILABILITY_LOCALE_CATEGORY                                 \
+     __attribute__((availability(macosx,strict,introduced=10.9)))              \
+     __attribute__((availability(ios,strict,introduced=7.0)))
+#  define _LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR                               \
+     __attribute__((availability(macosx,strict,introduced=10.9)))              \
+     __attribute__((availability(ios,strict,introduced=7.0)))
+#  define _LIBCUDACXX_AVAILABILITY_FILESYSTEM                                      \
+     __attribute__((availability(macosx,strict,introduced=10.15)))             \
+     __attribute__((availability(ios,strict,introduced=13.0)))                 \
+     __attribute__((availability(tvos,strict,introduced=13.0)))                \
+     __attribute__((availability(watchos,strict,introduced=6.0)))
+#  define _LIBCUDACXX_AVAILABILITY_FILESYSTEM_PUSH                                 \
+     _Pragma("clang attribute push(__attribute__((availability(macosx,strict,introduced=10.15))), apply_to=any(function,record))") \
+     _Pragma("clang attribute push(__attribute__((availability(ios,strict,introduced=13.0))), apply_to=any(function,record))")     \
+     _Pragma("clang attribute push(__attribute__((availability(tvos,strict,introduced=13.0))), apply_to=any(function,record))")    \
+     _Pragma("clang attribute push(__attribute__((availability(watchos,strict,introduced=6.0))), apply_to=any(function,record))")
+#  define _LIBCUDACXX_AVAILABILITY_FILESYSTEM_POP                                  \
+     _Pragma("clang attribute pop")                                            \
+     _Pragma("clang attribute pop")                                            \
+     _Pragma("clang attribute pop")                                            \
+     _Pragma("clang attribute pop")
+#else
+#  define _LIBCUDACXX_AVAILABILITY_SHARED_MUTEX
+#  define _LIBCUDACXX_AVAILABILITY_BAD_VARIANT_ACCESS
+#  define _LIBCUDACXX_AVAILABILITY_BAD_OPTIONAL_ACCESS
+#  define _LIBCUDACXX_AVAILABILITY_BAD_ANY_CAST
+#  define _LIBCUDACXX_AVAILABILITY_UNCAUGHT_EXCEPTIONS
+#  define _LIBCUDACXX_AVAILABILITY_SIZED_NEW_DELETE
+#  define _LIBCUDACXX_AVAILABILITY_FUTURE_ERROR
+#  define _LIBCUDACXX_AVAILABILITY_TYPEINFO_VTABLE
+#  define _LIBCUDACXX_AVAILABILITY_LOCALE_CATEGORY
+#  define _LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+#  define _LIBCUDACXX_AVAILABILITY_FILESYSTEM
+#  define _LIBCUDACXX_AVAILABILITY_FILESYSTEM_PUSH
+#  define _LIBCUDACXX_AVAILABILITY_FILESYSTEM_POP
+#endif
+
+// Define availability that depends on _LIBCUDACXX_NO_EXCEPTIONS.
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+#  define _LIBCUDACXX_AVAILABILITY_FUTURE
+#  define _LIBCUDACXX_AVAILABILITY_THROW_BAD_ANY_CAST
+#  define _LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS
+#  define _LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+#else
+#  define _LIBCUDACXX_AVAILABILITY_FUTURE                    _LIBCUDACXX_AVAILABILITY_FUTURE_ERROR
+#  define _LIBCUDACXX_AVAILABILITY_THROW_BAD_ANY_CAST        _LIBCUDACXX_AVAILABILITY_BAD_ANY_CAST
+#  define _LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS _LIBCUDACXX_AVAILABILITY_BAD_OPTIONAL_ACCESS
+#  define _LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS  _LIBCUDACXX_AVAILABILITY_BAD_VARIANT_ACCESS
+#endif
+
+// The stream API was dropped and re-added in the dylib shipped on macOS
+// and iOS. We can only assume the dylib to provide these definitions for
+// macosx >= 10.9 and ios >= 7.0. Otherwise, the definitions are available
+// from the headers, but not from the dylib. Explicit instantiation
+// declarations for streams exist conditionally to this; if we provide
+// an explicit instantiation declaration and we try to deploy to a dylib
+// that does not provide those symbols, we'll get a load-time error.
+#if !defined(_LIBCUDACXX_BUILDING_LIBRARY) &&                                      \
+    ((defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) &&                \
+      __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ < 1090) ||                 \
+     (defined(__ENVIRONMENT_IPHONE_OS_VERSION_MIN_REQUIRED__) &&               \
+      __ENVIRONMENT_IPHONE_OS_VERSION_MIN_REQUIRED__ < 70000))
+#  define _LIBCUDACXX_DO_NOT_ASSUME_STREAMS_EXPLICIT_INSTANTIATION_IN_DYLIB
+#endif
+
+#if defined(_LIBCUDACXX_COMPILER_IBM)
+#define _LIBCUDACXX_HAS_NO_PRAGMA_PUSH_POP_MACRO
+#endif
+
+#if defined(_LIBCUDACXX_HAS_NO_PRAGMA_PUSH_POP_MACRO)
+#  define _LIBCUDACXX_PUSH_MACROS
+#  define _LIBCUDACXX_POP_MACROS
+#else
+  // Don't warn about macro conflicts when we can restore them at the
+  // end of the header.
+#  ifndef _LIBCUDACXX_DISABLE_MACRO_CONFLICT_WARNINGS
+#    define _LIBCUDACXX_DISABLE_MACRO_CONFLICT_WARNINGS
+#  endif
+#  if defined(_LIBCUDACXX_COMPILER_MSVC)
+#    define _LIBCUDACXX_PUSH_MACROS    \
+       __pragma(push_macro("min")) \
+       __pragma(push_macro("max"))
+#    define _LIBCUDACXX_POP_MACROS     \
+       __pragma(pop_macro("min"))  \
+       __pragma(pop_macro("max"))
+#  else
+#    define _LIBCUDACXX_PUSH_MACROS        \
+       _Pragma("push_macro(\"min\")")  \
+       _Pragma("push_macro(\"max\")")
+#    define _LIBCUDACXX_POP_MACROS         \
+       _Pragma("pop_macro(\"min\")")   \
+       _Pragma("pop_macro(\"max\")")
+#  endif
+#endif // defined(_LIBCUDACXX_HAS_NO_PRAGMA_PUSH_POP_MACRO)
+
+#ifndef _LIBCUDACXX_NO_AUTO_LINK
+#  if defined(_LIBCUDACXX_ABI_MICROSOFT) && !defined(_LIBCUDACXX_BUILDING_LIBRARY)
+#    if defined(_DLL)
+#      pragma comment(lib, "c++.lib")
+#    else
+#      pragma comment(lib, "libc++.lib")
+#    endif
+#  endif // defined(_LIBCUDACXX_ABI_MICROSOFT) && !defined(_LIBCUDACXX_BUILDING_LIBRARY)
+#endif // _LIBCUDACXX_NO_AUTO_LINK
+
+#define _LIBCUDACXX_UNUSED_VAR(x) ((void)(x))
+
+// Configures the fopen close-on-exec mode character, if any. This string will
+// be appended to any mode string used by fstream for fopen/fdopen.
+//
+// Not all platforms support this, but it helps avoid fd-leaks on platforms that
+// do.
+#if defined(__BIONIC__)
+#  define _LIBCUDACXX_FOPEN_CLOEXEC_MODE "e"
+#else
+#  define _LIBCUDACXX_FOPEN_CLOEXEC_MODE
+#endif
+
+#ifndef _LIBCUDACXX_SYS_CLOCK_DURATION
+#  define _LIBCUDACXX_SYS_CLOCK_DURATION microseconds
+#endif
+
+#endif // __cplusplus
+
+#endif // _LIBCUDACXX_CONFIG
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config_site.in b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config_site.in
new file mode 100644
index 000000000000..ab65c9554b0b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__config_site.in
@@ -0,0 +1,36 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CONFIG_SITE
+#define _LIBCUDACXX_CONFIG_SITE
+
+#cmakedefine _LIBCUDACXX_ABI_VERSION @_LIBCUDACXX_ABI_VERSION@
+#cmakedefine _LIBCUDACXX_ABI_UNSTABLE
+#cmakedefine _LIBCUDACXX_ABI_FORCE_ITANIUM
+#cmakedefine _LIBCUDACXX_ABI_FORCE_MICROSOFT
+#cmakedefine _LIBCUDACXX_HIDE_FROM_ABI_PER_TU_BY_DEFAULT
+#cmakedefine _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+#cmakedefine _LIBCUDACXX_HAS_NO_STDIN
+#cmakedefine _LIBCUDACXX_HAS_NO_STDOUT
+#cmakedefine _LIBCUDACXX_HAS_NO_THREADS
+#cmakedefine _LIBCUDACXX_HAS_NO_MONOTONIC_CLOCK
+#cmakedefine _LIBCUDACXX_HAS_NO_THREAD_UNSAFE_C_FUNCTIONS
+#cmakedefine _LIBCUDACXX_HAS_MUSL_LIBC
+#cmakedefine _LIBCUDACXX_HAS_THREAD_API_PTHREAD
+#cmakedefine _LIBCUDACXX_HAS_THREAD_API_EXTERNAL
+#cmakedefine _LIBCUDACXX_HAS_THREAD_API_WIN32
+#cmakedefine _LIBCUDACXX_HAS_THREAD_LIBRARY_EXTERNAL
+#cmakedefine _LIBCUDACXX_DISABLE_VISIBILITY_ANNOTATIONS
+#cmakedefine _LIBCUDACXX_NO_VCRUNTIME
+#cmakedefine01 _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT
+#cmakedefine _LIBCUDACXX_ABI_NAMESPACE @_LIBCUDACXX_ABI_NAMESPACE@
+#cmakedefine _LIBCUDACXX_HAS_PARALLEL_ALGORITHMS
+
+@_LIBCUDACXX_ABI_DEFINES@
+
+#endif // _LIBCUDACXX_CONFIG_SITE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__debug b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__debug
new file mode 100644
index 000000000000..687fce51cdd6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__debug
@@ -0,0 +1,284 @@
+// -*- C++ -*-
+//===--------------------------- __debug ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_DEBUG_H
+#define _LIBCUDACXX_DEBUG_H
+
+#ifndef __cuda_std__
+#include <__config>
+#include <iosfwd>
+#if defined(_LIBCUDACXX_HAS_NO_NULLPTR)
+#  include <cstddef>
+#endif
+#if _LIBCUDACXX_DEBUG_LEVEL >= 1 || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+#  include <cstdlib>
+#  include <cstdio>
+#  include <cstddef>
+#endif
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 1 && !defined(_LIBCUDACXX_ASSERT)
+# define _LIBCUDACXX_ASSERT(x, m) ((x) ? (void)0 : \
+  _CUDA_VSTD::__libcpp_debug_function(_CUDA_VSTD::__libcpp_debug_info(__FILE__, __LINE__, #x, m)))
+#endif
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+#ifndef _LIBCUDACXX_DEBUG_ASSERT
+#define _LIBCUDACXX_DEBUG_ASSERT(x, m) _LIBCUDACXX_ASSERT(x, m)
+#endif
+#define _LIBCUDACXX_DEBUG_MODE(...) __VA_ARGS__
+#endif
+
+#ifndef _LIBCUDACXX_ASSERT
+#   define _LIBCUDACXX_ASSERT(x, m) ((void)0)
+#endif
+#ifndef _LIBCUDACXX_DEBUG_ASSERT
+#   define _LIBCUDACXX_DEBUG_ASSERT(x, m) ((void)0)
+#endif
+#ifndef _LIBCUDACXX_DEBUG_MODE
+#define _LIBCUDACXX_DEBUG_MODE(...) ((void)0)
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+struct _LIBCUDACXX_TEMPLATE_VIS __libcpp_debug_info {
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+  __libcpp_debug_info()
+      : __file_(nullptr), __line_(-1), __pred_(nullptr), __msg_(nullptr) {}
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+  __libcpp_debug_info(const char* __f, int __l, const char* __p, const char* __m)
+    : __file_(__f), __line_(__l), __pred_(__p), __msg_(__m) {}
+
+  _LIBCUDACXX_FUNC_VIS std::string what() const;
+
+  const char* __file_;
+  int __line_;
+  const char* __pred_;
+  const char* __msg_;
+};
+
+/// __libcpp_debug_function_type - The type of the assertion failure handler.
+typedef void(*__libcpp_debug_function_type)(__libcpp_debug_info const&);
+
+/// __libcpp_debug_function - The handler function called when a _LIBCUDACXX_ASSERT
+///    fails.
+extern _LIBCUDACXX_EXPORTED_FROM_ABI __libcpp_debug_function_type __libcpp_debug_function;
+
+/// __libcpp_abort_debug_function - A debug handler that aborts when called.
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS
+void __libcpp_abort_debug_function(__libcpp_debug_info const&);
+
+/// __libcpp_set_debug_function - Set the debug handler to the specified
+///    function.
+_LIBCUDACXX_FUNC_VIS
+bool __libcpp_set_debug_function(__libcpp_debug_function_type __func);
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2 || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+
+struct _LIBCUDACXX_TYPE_VIS __c_node;
+
+struct _LIBCUDACXX_TYPE_VIS __i_node
+{
+    void* __i_;
+    __i_node* __next_;
+    __c_node* __c_;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __i_node(const __i_node&) = delete;
+    __i_node& operator=(const __i_node&) = delete;
+#else
+private:
+    __i_node(const __i_node&);
+    __i_node& operator=(const __i_node&);
+public:
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __i_node(void* __i, __i_node* __next, __c_node* __c)
+        : __i_(__i), __next_(__next), __c_(__c) {}
+    ~__i_node();
+};
+
+struct _LIBCUDACXX_TYPE_VIS __c_node
+{
+    void* __c_;
+    __c_node* __next_;
+    __i_node** beg_;
+    __i_node** end_;
+    __i_node** cap_;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __c_node(const __c_node&) = delete;
+    __c_node& operator=(const __c_node&) = delete;
+#else
+private:
+    __c_node(const __c_node&);
+    __c_node& operator=(const __c_node&);
+public:
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __c_node(void* __c, __c_node* __next)
+        : __c_(__c), __next_(__next), beg_(nullptr), end_(nullptr), cap_(nullptr) {}
+    virtual ~__c_node();
+
+    virtual bool __dereferenceable(const void*) const = 0;
+    virtual bool __decrementable(const void*) const = 0;
+    virtual bool __addable(const void*, ptrdiff_t) const = 0;
+    virtual bool __subscriptable(const void*, ptrdiff_t) const = 0;
+
+    void __add(__i_node* __i);
+    _LIBCUDACXX_HIDDEN void __remove(__i_node* __i);
+};
+
+template <class _Cont>
+struct _C_node
+    : public __c_node
+{
+    _C_node(void* __c, __c_node* __n)
+        : __c_node(__c, __n) {}
+
+    virtual bool __dereferenceable(const void*) const;
+    virtual bool __decrementable(const void*) const;
+    virtual bool __addable(const void*, ptrdiff_t) const;
+    virtual bool __subscriptable(const void*, ptrdiff_t) const;
+};
+
+template <class _Cont>
+inline bool
+_C_node<_Cont>::__dereferenceable(const void* __i) const
+{
+    typedef typename _Cont::const_iterator iterator;
+    const iterator* __j = static_cast<const iterator*>(__i);
+    _Cont* _Cp = static_cast<_Cont*>(__c_);
+    return _Cp->__dereferenceable(__j);
+}
+
+template <class _Cont>
+inline bool
+_C_node<_Cont>::__decrementable(const void* __i) const
+{
+    typedef typename _Cont::const_iterator iterator;
+    const iterator* __j = static_cast<const iterator*>(__i);
+    _Cont* _Cp = static_cast<_Cont*>(__c_);
+    return _Cp->__decrementable(__j);
+}
+
+template <class _Cont>
+inline bool
+_C_node<_Cont>::__addable(const void* __i, ptrdiff_t __n) const
+{
+    typedef typename _Cont::const_iterator iterator;
+    const iterator* __j = static_cast<const iterator*>(__i);
+    _Cont* _Cp = static_cast<_Cont*>(__c_);
+    return _Cp->__addable(__j, __n);
+}
+
+template <class _Cont>
+inline bool
+_C_node<_Cont>::__subscriptable(const void* __i, ptrdiff_t __n) const
+{
+    typedef typename _Cont::const_iterator iterator;
+    const iterator* __j = static_cast<const iterator*>(__i);
+    _Cont* _Cp = static_cast<_Cont*>(__c_);
+    return _Cp->__subscriptable(__j, __n);
+}
+
+class _LIBCUDACXX_TYPE_VIS __libcpp_db
+{
+    __c_node** __cbeg_;
+    __c_node** __cend_;
+    size_t   __csz_;
+    __i_node** __ibeg_;
+    __i_node** __iend_;
+    size_t   __isz_;
+
+    __libcpp_db();
+public:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __libcpp_db(const __libcpp_db&) = delete;
+    __libcpp_db& operator=(const __libcpp_db&) = delete;
+#else
+private:
+    __libcpp_db(const __libcpp_db&);
+    __libcpp_db& operator=(const __libcpp_db&);
+public:
+#endif
+    ~__libcpp_db();
+
+    class __db_c_iterator;
+    class __db_c_const_iterator;
+    class __db_i_iterator;
+    class __db_i_const_iterator;
+
+    __db_c_const_iterator __c_end() const;
+    __db_i_const_iterator __i_end() const;
+
+    typedef __c_node*(_InsertConstruct)(void*, void*, __c_node*);
+
+    template <class _Cont>
+    _LIBCUDACXX_INLINE_VISIBILITY static __c_node* __create_C_node(void *__mem, void *__c, __c_node *__next) {
+        return ::new(__mem) _C_node<_Cont>(__c, __next);
+    }
+
+    template <class _Cont>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __insert_c(_Cont* __c)
+    {
+        __insert_c(static_cast<void*>(__c), &__create_C_node<_Cont>);
+    }
+
+    void __insert_i(void* __i);
+    void __insert_c(void* __c, _InsertConstruct* __fn);
+    void __erase_c(void* __c);
+
+    void __insert_ic(void* __i, const void* __c);
+    void __iterator_copy(void* __i, const void* __i0);
+    void __erase_i(void* __i);
+
+    void* __find_c_from_i(void* __i) const;
+    void __invalidate_all(void* __c);
+    __c_node* __find_c_and_lock(void* __c) const;
+    __c_node* __find_c(void* __c) const;
+    void unlock() const;
+
+    void swap(void* __c1, void* __c2);
+
+
+    bool __dereferenceable(const void* __i) const;
+    bool __decrementable(const void* __i) const;
+    bool __addable(const void* __i, ptrdiff_t __n) const;
+    bool __subscriptable(const void* __i, ptrdiff_t __n) const;
+    bool __less_than_comparable(const void* __i, const void* __j) const;
+private:
+    _LIBCUDACXX_HIDDEN
+    __i_node* __insert_iterator(void* __i);
+    _LIBCUDACXX_HIDDEN
+    __i_node* __find_iterator(const void* __i) const;
+
+    friend _LIBCUDACXX_FUNC_VIS __libcpp_db* __get_db();
+};
+
+_LIBCUDACXX_FUNC_VIS __libcpp_db* __get_db();
+_LIBCUDACXX_FUNC_VIS const __libcpp_db* __get_const_db();
+
+
+#endif // _LIBCUDACXX_DEBUG_LEVEL >= 2 || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+  #include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX_DEBUG_H
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__errc b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__errc
new file mode 100644
index 000000000000..03382fa86500
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__errc
@@ -0,0 +1,217 @@
+// -*- C++ -*-
+//===---------------------------- __errc ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___ERRC
+#define _LIBCUDACXX___ERRC
+
+/*
+    system_error synopsis
+
+namespace std
+{
+
+enum class errc
+{
+    address_family_not_supported,       // EAFNOSUPPORT
+    address_in_use,                     // EADDRINUSE
+    address_not_available,              // EADDRNOTAVAIL
+    already_connected,                  // EISCONN
+    argument_list_too_long,             // E2BIG
+    argument_out_of_domain,             // EDOM
+    bad_address,                        // EFAULT
+    bad_file_descriptor,                // EBADF
+    bad_message,                        // EBADMSG
+    broken_pipe,                        // EPIPE
+    connection_aborted,                 // ECONNABORTED
+    connection_already_in_progress,     // EALREADY
+    connection_refused,                 // ECONNREFUSED
+    connection_reset,                   // ECONNRESET
+    cross_device_link,                  // EXDEV
+    destination_address_required,       // EDESTADDRREQ
+    device_or_resource_busy,            // EBUSY
+    directory_not_empty,                // ENOTEMPTY
+    executable_format_error,            // ENOEXEC
+    file_exists,                        // EEXIST
+    file_too_large,                     // EFBIG
+    filename_too_long,                  // ENAMETOOLONG
+    function_not_supported,             // ENOSYS
+    host_unreachable,                   // EHOSTUNREACH
+    identifier_removed,                 // EIDRM
+    illegal_byte_sequence,              // EILSEQ
+    inappropriate_io_control_operation, // ENOTTY
+    interrupted,                        // EINTR
+    invalid_argument,                   // EINVAL
+    invalid_seek,                       // ESPIPE
+    io_error,                           // EIO
+    is_a_directory,                     // EISDIR
+    message_size,                       // EMSGSIZE
+    network_down,                       // ENETDOWN
+    network_reset,                      // ENETRESET
+    network_unreachable,                // ENETUNREACH
+    no_buffer_space,                    // ENOBUFS
+    no_child_process,                   // ECHILD
+    no_link,                            // ENOLINK
+    no_lock_available,                  // ENOLCK
+    no_message_available,               // ENODATA
+    no_message,                         // ENOMSG
+    no_protocol_option,                 // ENOPROTOOPT
+    no_space_on_device,                 // ENOSPC
+    no_stream_resources,                // ENOSR
+    no_such_device_or_address,          // ENXIO
+    no_such_device,                     // ENODEV
+    no_such_file_or_directory,          // ENOENT
+    no_such_process,                    // ESRCH
+    not_a_directory,                    // ENOTDIR
+    not_a_socket,                       // ENOTSOCK
+    not_a_stream,                       // ENOSTR
+    not_connected,                      // ENOTCONN
+    not_enough_memory,                  // ENOMEM
+    not_supported,                      // ENOTSUP
+    operation_canceled,                 // ECANCELED
+    operation_in_progress,              // EINPROGRESS
+    operation_not_permitted,            // EPERM
+    operation_not_supported,            // EOPNOTSUPP
+    operation_would_block,              // EWOULDBLOCK
+    owner_dead,                         // EOWNERDEAD
+    permission_denied,                  // EACCES
+    protocol_error,                     // EPROTO
+    protocol_not_supported,             // EPROTONOSUPPORT
+    read_only_file_system,              // EROFS
+    resource_deadlock_would_occur,      // EDEADLK
+    resource_unavailable_try_again,     // EAGAIN
+    result_out_of_range,                // ERANGE
+    state_not_recoverable,              // ENOTRECOVERABLE
+    stream_timeout,                     // ETIME
+    text_file_busy,                     // ETXTBSY
+    timed_out,                          // ETIMEDOUT
+    too_many_files_open_in_system,      // ENFILE
+    too_many_files_open,                // EMFILE
+    too_many_links,                     // EMLINK
+    too_many_symbolic_link_levels,      // ELOOP
+    value_too_large,                    // EOVERFLOW
+    wrong_protocol_type                 // EPROTOTYPE
+};
+
+*/
+
+#include <__config>
+#include <cerrno>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// Some error codes are not present on all platforms, so we provide equivalents
+// for them:
+
+//enum class errc
+_LIBCUDACXX_DECLARE_STRONG_ENUM(errc)
+{
+    address_family_not_supported        = EAFNOSUPPORT,
+    address_in_use                      = EADDRINUSE,
+    address_not_available               = EADDRNOTAVAIL,
+    already_connected                   = EISCONN,
+    argument_list_too_long              = E2BIG,
+    argument_out_of_domain              = EDOM,
+    bad_address                         = EFAULT,
+    bad_file_descriptor                 = EBADF,
+    bad_message                         = EBADMSG,
+    broken_pipe                         = EPIPE,
+    connection_aborted                  = ECONNABORTED,
+    connection_already_in_progress      = EALREADY,
+    connection_refused                  = ECONNREFUSED,
+    connection_reset                    = ECONNRESET,
+    cross_device_link                   = EXDEV,
+    destination_address_required        = EDESTADDRREQ,
+    device_or_resource_busy             = EBUSY,
+    directory_not_empty                 = ENOTEMPTY,
+    executable_format_error             = ENOEXEC,
+    file_exists                         = EEXIST,
+    file_too_large                      = EFBIG,
+    filename_too_long                   = ENAMETOOLONG,
+    function_not_supported              = ENOSYS,
+    host_unreachable                    = EHOSTUNREACH,
+    identifier_removed                  = EIDRM,
+    illegal_byte_sequence               = EILSEQ,
+    inappropriate_io_control_operation  = ENOTTY,
+    interrupted                         = EINTR,
+    invalid_argument                    = EINVAL,
+    invalid_seek                        = ESPIPE,
+    io_error                            = EIO,
+    is_a_directory                      = EISDIR,
+    message_size                        = EMSGSIZE,
+    network_down                        = ENETDOWN,
+    network_reset                       = ENETRESET,
+    network_unreachable                 = ENETUNREACH,
+    no_buffer_space                     = ENOBUFS,
+    no_child_process                    = ECHILD,
+    no_link                             = ENOLINK,
+    no_lock_available                   = ENOLCK,
+#ifdef ENODATA
+    no_message_available                = ENODATA,
+#else
+    no_message_available                = ENOMSG,
+#endif
+    no_message                          = ENOMSG,
+    no_protocol_option                  = ENOPROTOOPT,
+    no_space_on_device                  = ENOSPC,
+#ifdef ENOSR
+    no_stream_resources                 = ENOSR,
+#else
+    no_stream_resources                 = ENOMEM,
+#endif
+    no_such_device_or_address           = ENXIO,
+    no_such_device                      = ENODEV,
+    no_such_file_or_directory           = ENOENT,
+    no_such_process                     = ESRCH,
+    not_a_directory                     = ENOTDIR,
+    not_a_socket                        = ENOTSOCK,
+#ifdef ENOSTR
+    not_a_stream                        = ENOSTR,
+#else
+    not_a_stream                        = EINVAL,
+#endif
+    not_connected                       = ENOTCONN,
+    not_enough_memory                   = ENOMEM,
+    not_supported                       = ENOTSUP,
+    operation_canceled                  = ECANCELED,
+    operation_in_progress               = EINPROGRESS,
+    operation_not_permitted             = EPERM,
+    operation_not_supported             = EOPNOTSUPP,
+    operation_would_block               = EWOULDBLOCK,
+    owner_dead                          = EOWNERDEAD,
+    permission_denied                   = EACCES,
+    protocol_error                      = EPROTO,
+    protocol_not_supported              = EPROTONOSUPPORT,
+    read_only_file_system               = EROFS,
+    resource_deadlock_would_occur       = EDEADLK,
+    resource_unavailable_try_again      = EAGAIN,
+    result_out_of_range                 = ERANGE,
+    state_not_recoverable               = ENOTRECOVERABLE,
+#ifdef ETIME
+    stream_timeout                      = ETIME,
+#else
+    stream_timeout                      = ETIMEDOUT,
+#endif
+    text_file_busy                      = ETXTBSY,
+    timed_out                           = ETIMEDOUT,
+    too_many_files_open_in_system       = ENFILE,
+    too_many_files_open                 = EMFILE,
+    too_many_links                      = EMLINK,
+    too_many_symbolic_link_levels       = ELOOP,
+    value_too_large                     = EOVERFLOW,
+    wrong_protocol_type                 = EPROTOTYPE
+};
+_LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(errc)
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX___ERRC
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_03 b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_03
new file mode 100644
index 000000000000..f9610b6ea384
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_03
@@ -0,0 +1,1595 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FUNCTIONAL_03
+#define _LIBCUDACXX_FUNCTIONAL_03
+
+// manual variadic expansion for <functional>
+
+#include <__pragma_push>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+namespace __function {
+
+template<class _Fp> class __base;
+
+template<class _Rp>
+class __base<_Rp()>
+{
+    __base(const __base&);
+    __base& operator=(const __base&);
+public:
+    __base() {}
+    virtual ~__base() {}
+    virtual __base* __clone() const = 0;
+    virtual void __clone(__base*) const = 0;
+    virtual void destroy() = 0;
+    virtual void destroy_deallocate() = 0;
+    virtual _Rp operator()() = 0;
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const = 0;
+    virtual const std::type_info& target_type() const = 0;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp, class _A0>
+class __base<_Rp(_A0)>
+{
+    __base(const __base&);
+    __base& operator=(const __base&);
+public:
+    __base() {}
+    virtual ~__base() {}
+    virtual __base* __clone() const = 0;
+    virtual void __clone(__base*) const = 0;
+    virtual void destroy() = 0;
+    virtual void destroy_deallocate() = 0;
+    virtual _Rp operator()(_A0) = 0;
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const = 0;
+    virtual const std::type_info& target_type() const = 0;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp, class _A0, class _A1>
+class __base<_Rp(_A0, _A1)>
+{
+    __base(const __base&);
+    __base& operator=(const __base&);
+public:
+    __base() {}
+    virtual ~__base() {}
+    virtual __base* __clone() const = 0;
+    virtual void __clone(__base*) const = 0;
+    virtual void destroy() = 0;
+    virtual void destroy_deallocate() = 0;
+    virtual _Rp operator()(_A0, _A1) = 0;
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const = 0;
+    virtual const std::type_info& target_type() const = 0;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp, class _A0, class _A1, class _A2>
+class __base<_Rp(_A0, _A1, _A2)>
+{
+    __base(const __base&);
+    __base& operator=(const __base&);
+public:
+    __base() {}
+    virtual ~__base() {}
+    virtual __base* __clone() const = 0;
+    virtual void __clone(__base*) const = 0;
+    virtual void destroy() = 0;
+    virtual void destroy_deallocate() = 0;
+    virtual _Rp operator()(_A0, _A1, _A2) = 0;
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const = 0;
+    virtual const std::type_info& target_type() const = 0;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _FD, class _Alloc, class _FB> class __func;
+
+template<class _Fp, class _Alloc, class _Rp>
+class __func<_Fp, _Alloc, _Rp()>
+    : public  __base<_Rp()>
+{
+    __compressed_pair<_Fp, _Alloc> __f_;
+public:
+    explicit __func(_Fp __f) : __f_(_CUDA_VSTD::move(__f)) {}
+    explicit __func(_Fp __f, _Alloc __a) : __f_(_CUDA_VSTD::move(__f), _CUDA_VSTD::move(__a)) {}
+    virtual __base<_Rp()>* __clone() const;
+    virtual void __clone(__base<_Rp()>*) const;
+    virtual void destroy();
+    virtual void destroy_deallocate();
+    virtual _Rp operator()();
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const;
+    virtual const std::type_info& target_type() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Fp, class _Alloc, class _Rp>
+__base<_Rp()>*
+__func<_Fp, _Alloc, _Rp()>::__clone() const
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    typedef __allocator_destructor<_Ap> _Dp;
+    unique_ptr<__func, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+    ::new (__hold.get()) __func(__f_.first(), _Alloc(__a));
+    return __hold.release();
+}
+
+template<class _Fp, class _Alloc, class _Rp>
+void
+__func<_Fp, _Alloc, _Rp()>::__clone(__base<_Rp()>* __p) const
+{
+    ::new (__p) __func(__f_.first(), __f_.second());
+}
+
+template<class _Fp, class _Alloc, class _Rp>
+void
+__func<_Fp, _Alloc, _Rp()>::destroy()
+{
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+}
+
+template<class _Fp, class _Alloc, class _Rp>
+void
+__func<_Fp, _Alloc, _Rp()>::destroy_deallocate()
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+    __a.deallocate(this, 1);
+}
+
+template<class _Fp, class _Alloc, class _Rp>
+_Rp
+__func<_Fp, _Alloc, _Rp()>::operator()()
+{
+    typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+    return _Invoker::__call(__f_.first());
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp>
+const void*
+__func<_Fp, _Alloc, _Rp()>::target(const type_info& __ti) const
+{
+    if (__ti == typeid(_Fp))
+        return &__f_.first();
+    return (const void*)0;
+}
+
+template<class _Fp, class _Alloc, class _Rp>
+const std::type_info&
+__func<_Fp, _Alloc, _Rp()>::target_type() const
+{
+    return typeid(_Fp);
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+class __func<_Fp, _Alloc, _Rp(_A0)>
+    : public  __base<_Rp(_A0)>
+{
+    __compressed_pair<_Fp, _Alloc> __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __func(_Fp __f) : __f_(_CUDA_VSTD::move(__f)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __func(_Fp __f, _Alloc __a)
+        : __f_(_CUDA_VSTD::move(__f), _CUDA_VSTD::move(__a)) {}
+    virtual __base<_Rp(_A0)>* __clone() const;
+    virtual void __clone(__base<_Rp(_A0)>*) const;
+    virtual void destroy();
+    virtual void destroy_deallocate();
+    virtual _Rp operator()(_A0);
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const;
+    virtual const std::type_info& target_type() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+__base<_Rp(_A0)>*
+__func<_Fp, _Alloc, _Rp(_A0)>::__clone() const
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    typedef __allocator_destructor<_Ap> _Dp;
+    unique_ptr<__func, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+    ::new (__hold.get()) __func(__f_.first(), _Alloc(__a));
+    return __hold.release();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+void
+__func<_Fp, _Alloc, _Rp(_A0)>::__clone(__base<_Rp(_A0)>* __p) const
+{
+    ::new (__p) __func(__f_.first(), __f_.second());
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+void
+__func<_Fp, _Alloc, _Rp(_A0)>::destroy()
+{
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+void
+__func<_Fp, _Alloc, _Rp(_A0)>::destroy_deallocate()
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+    __a.deallocate(this, 1);
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+_Rp
+__func<_Fp, _Alloc, _Rp(_A0)>::operator()(_A0 __a0)
+{
+    typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+    return _Invoker::__call(__f_.first(), __a0);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+const void*
+__func<_Fp, _Alloc, _Rp(_A0)>::target(const type_info& __ti) const
+{
+    if (__ti == typeid(_Fp))
+        return &__f_.first();
+    return (const void*)0;
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0>
+const std::type_info&
+__func<_Fp, _Alloc, _Rp(_A0)>::target_type() const
+{
+    return typeid(_Fp);
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+class __func<_Fp, _Alloc, _Rp(_A0, _A1)>
+    : public  __base<_Rp(_A0, _A1)>
+{
+    __compressed_pair<_Fp, _Alloc> __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __func(_Fp __f) : __f_(_CUDA_VSTD::move(__f)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __func(_Fp __f, _Alloc __a)
+        : __f_(_CUDA_VSTD::move(__f), _CUDA_VSTD::move(__a)) {}
+    virtual __base<_Rp(_A0, _A1)>* __clone() const;
+    virtual void __clone(__base<_Rp(_A0, _A1)>*) const;
+    virtual void destroy();
+    virtual void destroy_deallocate();
+    virtual _Rp operator()(_A0, _A1);
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const;
+    virtual const std::type_info& target_type() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+__base<_Rp(_A0, _A1)>*
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::__clone() const
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    typedef __allocator_destructor<_Ap> _Dp;
+    unique_ptr<__func, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+    ::new (__hold.get()) __func(__f_.first(), _Alloc(__a));
+    return __hold.release();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+void
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::__clone(__base<_Rp(_A0, _A1)>* __p) const
+{
+    ::new (__p) __func(__f_.first(), __f_.second());
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+void
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::destroy()
+{
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+void
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::destroy_deallocate()
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+    __a.deallocate(this, 1);
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+_Rp
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::operator()(_A0 __a0, _A1 __a1)
+{
+    typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+    return _Invoker::__call(__f_.first(), __a0, __a1);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+const void*
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::target(const type_info& __ti) const
+{
+    if (__ti == typeid(_Fp))
+        return &__f_.first();
+    return (const void*)0;
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1>
+const std::type_info&
+__func<_Fp, _Alloc, _Rp(_A0, _A1)>::target_type() const
+{
+    return typeid(_Fp);
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+class __func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>
+    : public  __base<_Rp(_A0, _A1, _A2)>
+{
+    __compressed_pair<_Fp, _Alloc> __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __func(_Fp __f) : __f_(_CUDA_VSTD::move(__f)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __func(_Fp __f, _Alloc __a)
+        : __f_(_CUDA_VSTD::move(__f), _CUDA_VSTD::move(__a)) {}
+    virtual __base<_Rp(_A0, _A1, _A2)>* __clone() const;
+    virtual void __clone(__base<_Rp(_A0, _A1, _A2)>*) const;
+    virtual void destroy();
+    virtual void destroy_deallocate();
+    virtual _Rp operator()(_A0, _A1, _A2);
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const;
+    virtual const std::type_info& target_type() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+__base<_Rp(_A0, _A1, _A2)>*
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::__clone() const
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    typedef __allocator_destructor<_Ap> _Dp;
+    unique_ptr<__func, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+    ::new (__hold.get()) __func(__f_.first(), _Alloc(__a));
+    return __hold.release();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+void
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::__clone(__base<_Rp(_A0, _A1, _A2)>* __p) const
+{
+    ::new (__p) __func(__f_.first(), __f_.second());
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+void
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::destroy()
+{
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+void
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::destroy_deallocate()
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.second());
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+    __a.deallocate(this, 1);
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+_Rp
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::operator()(_A0 __a0, _A1 __a1, _A2 __a2)
+{
+    typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+    return _Invoker::__call(__f_.first(), __a0, __a1, __a2);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+const void*
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::target(const type_info& __ti) const
+{
+    if (__ti == typeid(_Fp))
+        return &__f_.first();
+    return (const void*)0;
+}
+
+template<class _Fp, class _Alloc, class _Rp, class _A0, class _A1, class _A2>
+const std::type_info&
+__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)>::target_type() const
+{
+    return typeid(_Fp);
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+}  // __function
+
+template<class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS function<_Rp()>
+{
+    typedef __function::__base<_Rp()> __base;
+    aligned_storage<3*sizeof(void*)>::type __buf_;
+    __base* __f_;
+
+public:
+    typedef _Rp result_type;
+
+    // 20.7.16.2.1, construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit function() : __f_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY function(nullptr_t) : __f_(0) {}
+    function(const function&);
+    template<class _Fp>
+      function(_Fp,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&) : __f_(0) {}
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&, nullptr_t) : __f_(0) {}
+    template<class _Alloc>
+      function(allocator_arg_t, const _Alloc&, const function&);
+    template<class _Fp, class _Alloc>
+      function(allocator_arg_t, const _Alloc& __a, _Fp __f,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    function& operator=(const function&);
+    function& operator=(nullptr_t);
+    template<class _Fp>
+      typename enable_if
+      <
+        !is_integral<_Fp>::value,
+        function&
+      >::type
+      operator=(_Fp);
+
+    ~function();
+
+    // 20.7.16.2.2, function modifiers:
+    void swap(function&);
+    template<class _Fp, class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      void assign(_Fp __f, const _Alloc& __a)
+        {function(allocator_arg, __a, __f).swap(*this);}
+
+    // 20.7.16.2.3, function capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY operator bool() const {return __f_;}
+
+private:
+    // deleted overloads close possible hole in the type system
+    template<class _R2>
+      bool operator==(const function<_R2()>&) const;// = delete;
+    template<class _R2>
+      bool operator!=(const function<_R2()>&) const;// = delete;
+public:
+    // 20.7.16.2.4, function invocation:
+    _Rp operator()() const;
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    // 20.7.16.2.5, function target access:
+    const std::type_info& target_type() const;
+    template <typename _Tp> _Tp* target();
+    template <typename _Tp> const _Tp* target() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp>
+function<_Rp()>::function(const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp>
+template<class _Alloc>
+function<_Rp()>::function(allocator_arg_t, const _Alloc&, const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp>
+template <class _Fp>
+function<_Rp()>::function(_Fp __f,
+                                     typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, allocator<_Fp>, _Rp()> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f);
+        }
+        else
+        {
+            typedef allocator<_FF> _Ap;
+            _Ap __a;
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, allocator<_Fp>(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp>
+template <class _Fp, class _Alloc>
+function<_Rp()>::function(allocator_arg_t, const _Alloc& __a0, _Fp __f,
+                                     typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, _Alloc, _Rp()> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f, __a0);
+        }
+        else
+        {
+            typedef typename __rebind_alloc_helper<__alloc_traits, _FF>::type _Ap;
+            _Ap __a(__a0);
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, _Alloc(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp>
+function<_Rp()>&
+function<_Rp()>::operator=(const function& __f)
+{
+    if (__f)
+        function(__f).swap(*this);
+    else
+        *this = nullptr;
+    return *this;
+}
+
+template<class _Rp>
+function<_Rp()>&
+function<_Rp()>::operator=(nullptr_t)
+{
+    __base* __t = __f_;
+    __f_ = 0;
+    if (__t == (__base*)&__buf_)
+        __t->destroy();
+    else if (__t)
+        __t->destroy_deallocate();
+    return *this;
+}
+
+template<class _Rp>
+template <class _Fp>
+typename enable_if
+<
+    !is_integral<_Fp>::value,
+    function<_Rp()>&
+>::type
+function<_Rp()>::operator=(_Fp __f)
+{
+    function(_CUDA_VSTD::move(__f)).swap(*this);
+    return *this;
+}
+
+template<class _Rp>
+function<_Rp()>::~function()
+{
+    if (__f_ == (__base*)&__buf_)
+        __f_->destroy();
+    else if (__f_)
+        __f_->destroy_deallocate();
+}
+
+template<class _Rp>
+void
+function<_Rp()>::swap(function& __f)
+{
+    if (_CUDA_VSTD::addressof(__f) == this)
+      return;
+    if (__f_ == (__base*)&__buf_ && __f.__f_ == (__base*)&__f.__buf_)
+    {
+        typename aligned_storage<sizeof(__buf_)>::type __tempbuf;
+        __base* __t = (__base*)&__tempbuf;
+        __f_->__clone(__t);
+        __f_->destroy();
+        __f_ = 0;
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = 0;
+        __f_ = (__base*)&__buf_;
+        __t->__clone((__base*)&__f.__buf_);
+        __t->destroy();
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f_ == (__base*)&__buf_)
+    {
+        __f_->__clone((__base*)&__f.__buf_);
+        __f_->destroy();
+        __f_ = __f.__f_;
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = __f_;
+        __f_ = (__base*)&__buf_;
+    }
+    else
+        _CUDA_VSTD::swap(__f_, __f.__f_);
+}
+
+template<class _Rp>
+_Rp
+function<_Rp()>::operator()() const
+{
+    if (__f_ == 0)
+        __throw_bad_function_call();
+    return (*__f_)();
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Rp>
+const std::type_info&
+function<_Rp()>::target_type() const
+{
+    if (__f_ == 0)
+        return typeid(void);
+    return __f_->target_type();
+}
+
+template<class _Rp>
+template <typename _Tp>
+_Tp*
+function<_Rp()>::target()
+{
+    if (__f_ == 0)
+        return (_Tp*)0;
+    return (_Tp*) const_cast<void *>(__f_->target(typeid(_Tp)));
+}
+
+template<class _Rp>
+template <typename _Tp>
+const _Tp*
+function<_Rp()>::target() const
+{
+    if (__f_ == 0)
+        return (const _Tp*)0;
+    return (const _Tp*)__f_->target(typeid(_Tp));
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class _A0>
+class _LIBCUDACXX_TEMPLATE_VIS function<_Rp(_A0)>
+    : public unary_function<_A0, _Rp>
+{
+    typedef __function::__base<_Rp(_A0)> __base;
+    aligned_storage<3*sizeof(void*)>::type __buf_;
+    __base* __f_;
+
+public:
+    typedef _Rp result_type;
+
+    // 20.7.16.2.1, construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit function() : __f_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY function(nullptr_t) : __f_(0) {}
+    function(const function&);
+    template<class _Fp>
+      function(_Fp,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&) : __f_(0) {}
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&, nullptr_t) : __f_(0) {}
+    template<class _Alloc>
+      function(allocator_arg_t, const _Alloc&, const function&);
+    template<class _Fp, class _Alloc>
+      function(allocator_arg_t, const _Alloc& __a, _Fp __f,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    function& operator=(const function&);
+    function& operator=(nullptr_t);
+    template<class _Fp>
+      typename enable_if
+      <
+        !is_integral<_Fp>::value,
+        function&
+      >::type
+      operator=(_Fp);
+
+    ~function();
+
+    // 20.7.16.2.2, function modifiers:
+    void swap(function&);
+    template<class _Fp, class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      void assign(_Fp __f, const _Alloc& __a)
+        {function(allocator_arg, __a, __f).swap(*this);}
+
+    // 20.7.16.2.3, function capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY operator bool() const {return __f_;}
+
+private:
+    // deleted overloads close possible hole in the type system
+    template<class _R2, class _B0>
+      bool operator==(const function<_R2(_B0)>&) const;// = delete;
+    template<class _R2, class _B0>
+      bool operator!=(const function<_R2(_B0)>&) const;// = delete;
+public:
+    // 20.7.16.2.4, function invocation:
+    _Rp operator()(_A0) const;
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    // 20.7.16.2.5, function target access:
+    const std::type_info& target_type() const;
+    template <typename _Tp> _Tp* target();
+    template <typename _Tp> const _Tp* target() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp, class _A0>
+function<_Rp(_A0)>::function(const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp, class _A0>
+template<class _Alloc>
+function<_Rp(_A0)>::function(allocator_arg_t, const _Alloc&, const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp, class _A0>
+template <class _Fp>
+function<_Rp(_A0)>::function(_Fp __f,
+                                     typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, allocator<_Fp>, _Rp(_A0)> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f);
+        }
+        else
+        {
+            typedef allocator<_FF> _Ap;
+            _Ap __a;
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, allocator<_Fp>(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp, class _A0>
+template <class _Fp, class _Alloc>
+function<_Rp(_A0)>::function(allocator_arg_t, const _Alloc& __a0, _Fp __f,
+                                     typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, _Alloc, _Rp(_A0)> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f, __a0);
+        }
+        else
+        {
+            typedef typename __rebind_alloc_helper<__alloc_traits, _FF>::type _Ap;
+            _Ap __a(__a0);
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, _Alloc(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp, class _A0>
+function<_Rp(_A0)>&
+function<_Rp(_A0)>::operator=(const function& __f)
+{
+    if (__f)
+        function(__f).swap(*this);
+    else
+        *this = nullptr;
+    return *this;
+}
+
+template<class _Rp, class _A0>
+function<_Rp(_A0)>&
+function<_Rp(_A0)>::operator=(nullptr_t)
+{
+    __base* __t = __f_;
+    __f_ = 0;
+    if (__t == (__base*)&__buf_)
+        __t->destroy();
+    else if (__t)
+        __t->destroy_deallocate();
+    return *this;
+}
+
+template<class _Rp, class _A0>
+template <class _Fp>
+typename enable_if
+<
+    !is_integral<_Fp>::value,
+    function<_Rp(_A0)>&
+>::type
+function<_Rp(_A0)>::operator=(_Fp __f)
+{
+    function(_CUDA_VSTD::move(__f)).swap(*this);
+    return *this;
+}
+
+template<class _Rp, class _A0>
+function<_Rp(_A0)>::~function()
+{
+    if (__f_ == (__base*)&__buf_)
+        __f_->destroy();
+    else if (__f_)
+        __f_->destroy_deallocate();
+}
+
+template<class _Rp, class _A0>
+void
+function<_Rp(_A0)>::swap(function& __f)
+{
+    if (_CUDA_VSTD::addressof(__f) == this)
+      return;
+    if (__f_ == (__base*)&__buf_ && __f.__f_ == (__base*)&__f.__buf_)
+    {
+        typename aligned_storage<sizeof(__buf_)>::type __tempbuf;
+        __base* __t = (__base*)&__tempbuf;
+        __f_->__clone(__t);
+        __f_->destroy();
+        __f_ = 0;
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = 0;
+        __f_ = (__base*)&__buf_;
+        __t->__clone((__base*)&__f.__buf_);
+        __t->destroy();
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f_ == (__base*)&__buf_)
+    {
+        __f_->__clone((__base*)&__f.__buf_);
+        __f_->destroy();
+        __f_ = __f.__f_;
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = __f_;
+        __f_ = (__base*)&__buf_;
+    }
+    else
+        _CUDA_VSTD::swap(__f_, __f.__f_);
+}
+
+template<class _Rp, class _A0>
+_Rp
+function<_Rp(_A0)>::operator()(_A0 __a0) const
+{
+    if (__f_ == 0)
+        __throw_bad_function_call();
+    return (*__f_)(__a0);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class _A0>
+const std::type_info&
+function<_Rp(_A0)>::target_type() const
+{
+    if (__f_ == 0)
+        return typeid(void);
+    return __f_->target_type();
+}
+
+template<class _Rp, class _A0>
+template <typename _Tp>
+_Tp*
+function<_Rp(_A0)>::target()
+{
+    if (__f_ == 0)
+        return (_Tp*)0;
+    return (_Tp*) const_cast<void *>(__f_->target(typeid(_Tp)));
+}
+
+template<class _Rp, class _A0>
+template <typename _Tp>
+const _Tp*
+function<_Rp(_A0)>::target() const
+{
+    if (__f_ == 0)
+        return (const _Tp*)0;
+    return (const _Tp*)__f_->target(typeid(_Tp));
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class _A0, class _A1>
+class _LIBCUDACXX_TEMPLATE_VIS function<_Rp(_A0, _A1)>
+    : public binary_function<_A0, _A1, _Rp>
+{
+    typedef __function::__base<_Rp(_A0, _A1)> __base;
+    aligned_storage<3*sizeof(void*)>::type __buf_;
+    __base* __f_;
+
+public:
+    typedef _Rp result_type;
+
+    // 20.7.16.2.1, construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit function() : __f_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY function(nullptr_t) : __f_(0) {}
+    function(const function&);
+    template<class _Fp>
+      function(_Fp,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&) : __f_(0) {}
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&, nullptr_t) : __f_(0) {}
+    template<class _Alloc>
+      function(allocator_arg_t, const _Alloc&, const function&);
+    template<class _Fp, class _Alloc>
+      function(allocator_arg_t, const _Alloc& __a, _Fp __f,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    function& operator=(const function&);
+    function& operator=(nullptr_t);
+    template<class _Fp>
+      typename enable_if
+      <
+        !is_integral<_Fp>::value,
+        function&
+      >::type
+      operator=(_Fp);
+
+    ~function();
+
+    // 20.7.16.2.2, function modifiers:
+    void swap(function&);
+    template<class _Fp, class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      void assign(_Fp __f, const _Alloc& __a)
+        {function(allocator_arg, __a, __f).swap(*this);}
+
+    // 20.7.16.2.3, function capacity:
+    operator bool() const {return __f_;}
+
+private:
+    // deleted overloads close possible hole in the type system
+    template<class _R2, class _B0, class _B1>
+      bool operator==(const function<_R2(_B0, _B1)>&) const;// = delete;
+    template<class _R2, class _B0, class _B1>
+      bool operator!=(const function<_R2(_B0, _B1)>&) const;// = delete;
+public:
+    // 20.7.16.2.4, function invocation:
+    _Rp operator()(_A0, _A1) const;
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    // 20.7.16.2.5, function target access:
+    const std::type_info& target_type() const;
+    template <typename _Tp> _Tp* target();
+    template <typename _Tp> const _Tp* target() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp, class _A0, class _A1>
+function<_Rp(_A0, _A1)>::function(const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp, class _A0, class _A1>
+template<class _Alloc>
+function<_Rp(_A0, _A1)>::function(allocator_arg_t, const _Alloc&, const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp, class _A0, class _A1>
+template <class _Fp>
+function<_Rp(_A0, _A1)>::function(_Fp __f,
+                                 typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, allocator<_Fp>, _Rp(_A0, _A1)> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f);
+        }
+        else
+        {
+            typedef allocator<_FF> _Ap;
+            _Ap __a;
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, allocator<_Fp>(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp, class _A0, class _A1>
+template <class _Fp, class _Alloc>
+function<_Rp(_A0, _A1)>::function(allocator_arg_t, const _Alloc& __a0, _Fp __f,
+                                 typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, _Alloc, _Rp(_A0, _A1)> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f, __a0);
+        }
+        else
+        {
+            typedef typename __rebind_alloc_helper<__alloc_traits, _FF>::type _Ap;
+            _Ap __a(__a0);
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, _Alloc(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp, class _A0, class _A1>
+function<_Rp(_A0, _A1)>&
+function<_Rp(_A0, _A1)>::operator=(const function& __f)
+{
+    if (__f)
+        function(__f).swap(*this);
+    else
+        *this = nullptr;
+    return *this;
+}
+
+template<class _Rp, class _A0, class _A1>
+function<_Rp(_A0, _A1)>&
+function<_Rp(_A0, _A1)>::operator=(nullptr_t)
+{
+    __base* __t = __f_;
+    __f_ = 0;
+    if (__t == (__base*)&__buf_)
+        __t->destroy();
+    else if (__t)
+        __t->destroy_deallocate();
+    return *this;
+}
+
+template<class _Rp, class _A0, class _A1>
+template <class _Fp>
+typename enable_if
+<
+    !is_integral<_Fp>::value,
+    function<_Rp(_A0, _A1)>&
+>::type
+function<_Rp(_A0, _A1)>::operator=(_Fp __f)
+{
+    function(_CUDA_VSTD::move(__f)).swap(*this);
+    return *this;
+}
+
+template<class _Rp, class _A0, class _A1>
+function<_Rp(_A0, _A1)>::~function()
+{
+    if (__f_ == (__base*)&__buf_)
+        __f_->destroy();
+    else if (__f_)
+        __f_->destroy_deallocate();
+}
+
+template<class _Rp, class _A0, class _A1>
+void
+function<_Rp(_A0, _A1)>::swap(function& __f)
+{
+    if (_CUDA_VSTD::addressof(__f) == this)
+      return;
+    if (__f_ == (__base*)&__buf_ && __f.__f_ == (__base*)&__f.__buf_)
+    {
+        typename aligned_storage<sizeof(__buf_)>::type __tempbuf;
+        __base* __t = (__base*)&__tempbuf;
+        __f_->__clone(__t);
+        __f_->destroy();
+        __f_ = 0;
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = 0;
+        __f_ = (__base*)&__buf_;
+        __t->__clone((__base*)&__f.__buf_);
+        __t->destroy();
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f_ == (__base*)&__buf_)
+    {
+        __f_->__clone((__base*)&__f.__buf_);
+        __f_->destroy();
+        __f_ = __f.__f_;
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = __f_;
+        __f_ = (__base*)&__buf_;
+    }
+    else
+        _CUDA_VSTD::swap(__f_, __f.__f_);
+}
+
+template<class _Rp, class _A0, class _A1>
+_Rp
+function<_Rp(_A0, _A1)>::operator()(_A0 __a0, _A1 __a1) const
+{
+    if (__f_ == 0)
+        __throw_bad_function_call();
+    return (*__f_)(__a0, __a1);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class _A0, class _A1>
+const std::type_info&
+function<_Rp(_A0, _A1)>::target_type() const
+{
+    if (__f_ == 0)
+        return typeid(void);
+    return __f_->target_type();
+}
+
+template<class _Rp, class _A0, class _A1>
+template <typename _Tp>
+_Tp*
+function<_Rp(_A0, _A1)>::target()
+{
+    if (__f_ == 0)
+        return (_Tp*)0;
+    return (_Tp*) const_cast<void *>(__f_->target(typeid(_Tp)));
+}
+
+template<class _Rp, class _A0, class _A1>
+template <typename _Tp>
+const _Tp*
+function<_Rp(_A0, _A1)>::target() const
+{
+    if (__f_ == 0)
+        return (const _Tp*)0;
+    return (const _Tp*)__f_->target(typeid(_Tp));
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class _A0, class _A1, class _A2>
+class _LIBCUDACXX_TEMPLATE_VIS function<_Rp(_A0, _A1, _A2)>
+{
+    typedef __function::__base<_Rp(_A0, _A1, _A2)> __base;
+    aligned_storage<3*sizeof(void*)>::type __buf_;
+    __base* __f_;
+
+public:
+    typedef _Rp result_type;
+
+    // 20.7.16.2.1, construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit function() : __f_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY function(nullptr_t) : __f_(0) {}
+    function(const function&);
+    template<class _Fp>
+      function(_Fp,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&) : __f_(0) {}
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&, nullptr_t) : __f_(0) {}
+    template<class _Alloc>
+      function(allocator_arg_t, const _Alloc&, const function&);
+    template<class _Fp, class _Alloc>
+      function(allocator_arg_t, const _Alloc& __a, _Fp __f,
+               typename enable_if<!is_integral<_Fp>::value>::type* = 0);
+
+    function& operator=(const function&);
+    function& operator=(nullptr_t);
+    template<class _Fp>
+      typename enable_if
+      <
+        !is_integral<_Fp>::value,
+        function&
+      >::type
+      operator=(_Fp);
+
+    ~function();
+
+    // 20.7.16.2.2, function modifiers:
+    void swap(function&);
+    template<class _Fp, class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      void assign(_Fp __f, const _Alloc& __a)
+        {function(allocator_arg, __a, __f).swap(*this);}
+
+    // 20.7.16.2.3, function capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY operator bool() const {return __f_;}
+
+private:
+    // deleted overloads close possible hole in the type system
+    template<class _R2, class _B0, class _B1, class _B2>
+      bool operator==(const function<_R2(_B0, _B1, _B2)>&) const;// = delete;
+    template<class _R2, class _B0, class _B1, class _B2>
+      bool operator!=(const function<_R2(_B0, _B1, _B2)>&) const;// = delete;
+public:
+    // 20.7.16.2.4, function invocation:
+    _Rp operator()(_A0, _A1, _A2) const;
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    // 20.7.16.2.5, function target access:
+    const std::type_info& target_type() const;
+    template <typename _Tp> _Tp* target();
+    template <typename _Tp> const _Tp* target() const;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Rp, class _A0, class _A1, class _A2>
+function<_Rp(_A0, _A1, _A2)>::function(const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+template<class _Alloc>
+function<_Rp(_A0, _A1, _A2)>::function(allocator_arg_t, const _Alloc&,
+                                      const function& __f)
+{
+    if (__f.__f_ == 0)
+        __f_ = 0;
+    else if (__f.__f_ == (const __base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__clone(__f_);
+    }
+    else
+        __f_ = __f.__f_->__clone();
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+template <class _Fp>
+function<_Rp(_A0, _A1, _A2)>::function(_Fp __f,
+                                     typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, allocator<_Fp>, _Rp(_A0, _A1, _A2)> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f);
+        }
+        else
+        {
+            typedef allocator<_FF> _Ap;
+            _Ap __a;
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, allocator<_Fp>(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+template <class _Fp, class _Alloc>
+function<_Rp(_A0, _A1, _A2)>::function(allocator_arg_t, const _Alloc& __a0, _Fp __f,
+                                     typename enable_if<!is_integral<_Fp>::value>::type*)
+    : __f_(0)
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    if (__function::__not_null(__f))
+    {
+        typedef __function::__func<_Fp, _Alloc, _Rp(_A0, _A1, _A2)> _FF;
+        if (sizeof(_FF) <= sizeof(__buf_))
+        {
+            __f_ = (__base*)&__buf_;
+            ::new (__f_) _FF(__f, __a0);
+        }
+        else
+        {
+            typedef typename __rebind_alloc_helper<__alloc_traits, _FF>::type _Ap;
+            _Ap __a(__a0);
+            typedef __allocator_destructor<_Ap> _Dp;
+            unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+            ::new (__hold.get()) _FF(__f, _Alloc(__a));
+            __f_ = __hold.release();
+        }
+    }
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+function<_Rp(_A0, _A1, _A2)>&
+function<_Rp(_A0, _A1, _A2)>::operator=(const function& __f)
+{
+    if (__f)
+        function(__f).swap(*this);
+    else
+        *this = nullptr;
+    return *this;
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+function<_Rp(_A0, _A1, _A2)>&
+function<_Rp(_A0, _A1, _A2)>::operator=(nullptr_t)
+{
+    __base* __t = __f_;
+    __f_ = 0;
+    if (__t == (__base*)&__buf_)
+        __t->destroy();
+    else if (__t)
+        __t->destroy_deallocate();
+    return *this;
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+template <class _Fp>
+typename enable_if
+<
+    !is_integral<_Fp>::value,
+    function<_Rp(_A0, _A1, _A2)>&
+>::type
+function<_Rp(_A0, _A1, _A2)>::operator=(_Fp __f)
+{
+    function(_CUDA_VSTD::move(__f)).swap(*this);
+    return *this;
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+function<_Rp(_A0, _A1, _A2)>::~function()
+{
+    if (__f_ == (__base*)&__buf_)
+        __f_->destroy();
+    else if (__f_)
+        __f_->destroy_deallocate();
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+void
+function<_Rp(_A0, _A1, _A2)>::swap(function& __f)
+{
+    if (_CUDA_VSTD::addressof(__f) == this)
+      return;
+    if (__f_ == (__base*)&__buf_ && __f.__f_ == (__base*)&__f.__buf_)
+    {
+        typename aligned_storage<sizeof(__buf_)>::type __tempbuf;
+        __base* __t = (__base*)&__tempbuf;
+        __f_->__clone(__t);
+        __f_->destroy();
+        __f_ = 0;
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = 0;
+        __f_ = (__base*)&__buf_;
+        __t->__clone((__base*)&__f.__buf_);
+        __t->destroy();
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f_ == (__base*)&__buf_)
+    {
+        __f_->__clone((__base*)&__f.__buf_);
+        __f_->destroy();
+        __f_ = __f.__f_;
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f.__f_->__clone((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = __f_;
+        __f_ = (__base*)&__buf_;
+    }
+    else
+        _CUDA_VSTD::swap(__f_, __f.__f_);
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+_Rp
+function<_Rp(_A0, _A1, _A2)>::operator()(_A0 __a0, _A1 __a1, _A2 __a2) const
+{
+    if (__f_ == 0)
+        __throw_bad_function_call();
+    return (*__f_)(__a0, __a1, __a2);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class _A0, class _A1, class _A2>
+const std::type_info&
+function<_Rp(_A0, _A1, _A2)>::target_type() const
+{
+    if (__f_ == 0)
+        return typeid(void);
+    return __f_->target_type();
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+template <typename _Tp>
+_Tp*
+function<_Rp(_A0, _A1, _A2)>::target()
+{
+    if (__f_ == 0)
+        return (_Tp*)0;
+    return (_Tp*) const_cast<void *>(__f_->target(typeid(_Tp)));
+}
+
+template<class _Rp, class _A0, class _A1, class _A2>
+template <typename _Tp>
+const _Tp*
+function<_Rp(_A0, _A1, _A2)>::target() const
+{
+    if (__f_ == 0)
+        return (const _Tp*)0;
+    return (const _Tp*)__f_->target(typeid(_Tp));
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template <class _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const function<_Fp>& __f, nullptr_t) {return !__f;}
+
+template <class _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(nullptr_t, const function<_Fp>& __f) {return !__f;}
+
+template <class _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const function<_Fp>& __f, nullptr_t) {return (bool)__f;}
+
+template <class _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(nullptr_t, const function<_Fp>& __f) {return (bool)__f;}
+
+template <class _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(function<_Fp>& __x, function<_Fp>& __y)
+{return __x.swap(__y);}
+
+#include <__pragma_pop>
+
+#endif  // _LIBCUDACXX_FUNCTIONAL_03
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base
new file mode 100644
index 000000000000..d6597340624a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base
@@ -0,0 +1,673 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FUNCTIONAL_BASE
+#define _LIBCUDACXX_FUNCTIONAL_BASE
+
+#ifndef __cuda_std__
+#include <__config>
+#include <type_traits>
+#include <typeinfo>
+#include <exception>
+#include <new>
+#include <utility>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Arg1, class _Arg2, class _Result>
+struct _LIBCUDACXX_TEMPLATE_VIS binary_function
+{
+    typedef _Arg1   first_argument_type;
+    typedef _Arg2   second_argument_type;
+    typedef _Result result_type;
+};
+
+template <class _Tp>
+struct __has_result_type
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    template <class _Up> _LIBCUDACXX_INLINE_VISIBILITY static __two __test(...);
+    template <class _Up> _LIBCUDACXX_INLINE_VISIBILITY static char __test(typename _Up::result_type* = 0);
+public:
+    static const bool value = sizeof(__test<_Tp>(0)) == 1;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS less : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x < __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS less<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) < _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) < _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) < _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+// __weak_result_type
+
+template <class _Tp>
+struct __derives_from_unary_function
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static __two __test(...);
+    template <class _Ap, class _Rp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static unary_function<_Ap, _Rp>
+        __test(const volatile unary_function<_Ap, _Rp>*);
+public:
+    static const bool value = !is_same<decltype(__test((_Tp*)0)), __two>::value;
+    typedef decltype(__test((_Tp*)0)) type;
+};
+
+template <class _Tp>
+struct __derives_from_binary_function
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static __two __test(...);
+    template <class _A1, class _A2, class _Rp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static binary_function<_A1, _A2, _Rp>
+        __test(const volatile binary_function<_A1, _A2, _Rp>*);
+public:
+    static const bool value = !is_same<decltype(__test((_Tp*)0)), __two>::value;
+    typedef decltype(__test((_Tp*)0)) type;
+};
+
+template <class _Tp, bool = __derives_from_unary_function<_Tp>::value>
+struct __maybe_derive_from_unary_function  // bool is true
+    : public __derives_from_unary_function<_Tp>::type
+{
+};
+
+template <class _Tp>
+struct __maybe_derive_from_unary_function<_Tp, false>
+{
+};
+
+template <class _Tp, bool = __derives_from_binary_function<_Tp>::value>
+struct __maybe_derive_from_binary_function  // bool is true
+    : public __derives_from_binary_function<_Tp>::type
+{
+};
+
+template <class _Tp>
+struct __maybe_derive_from_binary_function<_Tp, false>
+{
+};
+
+template <class _Tp, bool = __has_result_type<_Tp>::value>
+struct __weak_result_type_imp // bool is true
+    : public __maybe_derive_from_unary_function<_Tp>,
+      public __maybe_derive_from_binary_function<_Tp>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Tp::result_type result_type;
+};
+
+template <class _Tp>
+struct __weak_result_type_imp<_Tp, false>
+    : public __maybe_derive_from_unary_function<_Tp>,
+      public __maybe_derive_from_binary_function<_Tp>
+{
+};
+
+template <class _Tp>
+struct __weak_result_type
+    : public __weak_result_type_imp<_Tp>
+{
+};
+
+// 0 argument case
+
+template <class _Rp>
+struct __weak_result_type<_Rp ()>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE  _Rp result_type;
+};
+
+template <class _Rp>
+struct __weak_result_type<_Rp (&)()>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE  _Rp result_type;
+};
+
+template <class _Rp>
+struct __weak_result_type<_Rp (*)()>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE  _Rp result_type;
+};
+
+// 1 argument case
+
+template <class _Rp, class _A1>
+struct __weak_result_type<_Rp (_A1)>
+    : public unary_function<_A1, _Rp>
+{
+};
+
+template <class _Rp, class _A1>
+struct __weak_result_type<_Rp (&)(_A1)>
+    : public unary_function<_A1, _Rp>
+{
+};
+
+template <class _Rp, class _A1>
+struct __weak_result_type<_Rp (*)(_A1)>
+    : public unary_function<_A1, _Rp>
+{
+};
+
+template <class _Rp, class _Cp>
+struct __weak_result_type<_Rp (_Cp::*)()>
+    : public unary_function<_Cp*, _Rp>
+{
+};
+
+template <class _Rp, class _Cp>
+struct __weak_result_type<_Rp (_Cp::*)() const>
+    : public unary_function<const _Cp*, _Rp>
+{
+};
+
+template <class _Rp, class _Cp>
+struct __weak_result_type<_Rp (_Cp::*)() volatile>
+    : public unary_function<volatile _Cp*, _Rp>
+{
+};
+
+template <class _Rp, class _Cp>
+struct __weak_result_type<_Rp (_Cp::*)() const volatile>
+    : public unary_function<const volatile _Cp*, _Rp>
+{
+};
+
+// 2 argument case
+
+template <class _Rp, class _A1, class _A2>
+struct __weak_result_type<_Rp (_A1, _A2)>
+    : public binary_function<_A1, _A2, _Rp>
+{
+};
+
+template <class _Rp, class _A1, class _A2>
+struct __weak_result_type<_Rp (*)(_A1, _A2)>
+    : public binary_function<_A1, _A2, _Rp>
+{
+};
+
+template <class _Rp, class _A1, class _A2>
+struct __weak_result_type<_Rp (&)(_A1, _A2)>
+    : public binary_function<_A1, _A2, _Rp>
+{
+};
+
+template <class _Rp, class _Cp, class _A1>
+struct __weak_result_type<_Rp (_Cp::*)(_A1)>
+    : public binary_function<_Cp*, _A1, _Rp>
+{
+};
+
+template <class _Rp, class _Cp, class _A1>
+struct __weak_result_type<_Rp (_Cp::*)(_A1) const>
+    : public binary_function<const _Cp*, _A1, _Rp>
+{
+};
+
+template <class _Rp, class _Cp, class _A1>
+struct __weak_result_type<_Rp (_Cp::*)(_A1) volatile>
+    : public binary_function<volatile _Cp*, _A1, _Rp>
+{
+};
+
+template <class _Rp, class _Cp, class _A1>
+struct __weak_result_type<_Rp (_Cp::*)(_A1) const volatile>
+    : public binary_function<const volatile _Cp*, _A1, _Rp>
+{
+};
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+// 3 or more arguments
+
+template <class _Rp, class _A1, class _A2, class _A3, class ..._A4>
+struct __weak_result_type<_Rp (_A1, _A2, _A3, _A4...)>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Rp, class _A1, class _A2, class _A3, class ..._A4>
+struct __weak_result_type<_Rp (&)(_A1, _A2, _A3, _A4...)>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Rp, class _A1, class _A2, class _A3, class ..._A4>
+struct __weak_result_type<_Rp (*)(_A1, _A2, _A3, _A4...)>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Rp, class _Cp, class _A1, class _A2, class ..._A3>
+struct __weak_result_type<_Rp (_Cp::*)(_A1, _A2, _A3...)>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Rp, class _Cp, class _A1, class _A2, class ..._A3>
+struct __weak_result_type<_Rp (_Cp::*)(_A1, _A2, _A3...) const>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Rp, class _Cp, class _A1, class _A2, class ..._A3>
+struct __weak_result_type<_Rp (_Cp::*)(_A1, _A2, _A3...) volatile>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Rp, class _Cp, class _A1, class _A2, class ..._A3>
+struct __weak_result_type<_Rp (_Cp::*)(_A1, _A2, _A3...) const volatile>
+{
+    typedef _Rp result_type;
+};
+
+template <class _Tp, class ..._Args>
+struct __invoke_return
+{
+    typedef decltype(__invoke(_CUDA_VSTD::declval<_Tp>(), _CUDA_VSTD::declval<_Args>()...)) type;
+};
+
+#else // defined(_LIBCUDACXX_CXX03_LANG)
+
+#include <__functional_base_03>
+
+#endif  // !defined(_LIBCUDACXX_CXX03_LANG)
+
+
+template <class _Ret>
+struct __invoke_void_return_wrapper
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Ret __call(_Args&&... __args) {
+        return __invoke(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+#else
+    template <class _Fn>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Ret __call(_Fn __f) {
+        return __invoke(__f);
+    }
+
+    template <class _Fn, class _A0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Ret __call(_Fn __f, _A0& __a0) {
+        return __invoke(__f, __a0);
+    }
+
+    template <class _Fn, class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Ret __call(_Fn __f, _A0& __a0, _A1& __a1) {
+        return __invoke(__f, __a0, __a1);
+    }
+
+    template <class _Fn, class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Ret __call(_Fn __f, _A0& __a0, _A1& __a1, _A2& __a2){
+        return __invoke(__f, __a0, __a1, __a2);
+    }
+#endif
+};
+
+template <>
+struct __invoke_void_return_wrapper<void>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __call(_Args&&... __args) {
+        __invoke(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+#else
+    template <class _Fn>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __call(_Fn __f) {
+        __invoke(__f);
+    }
+
+    template <class _Fn, class _A0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __call(_Fn __f, _A0& __a0) {
+        __invoke(__f, __a0);
+    }
+
+    template <class _Fn, class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __call(_Fn __f, _A0& __a0, _A1& __a1) {
+        __invoke(__f, __a0, __a1);
+    }
+
+    template <class _Fn, class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __call(_Fn __f, _A0& __a0, _A1& __a1, _A2& __a2) {
+        __invoke(__f, __a0, __a1, __a2);
+    }
+#endif
+};
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS reference_wrapper
+    : public __weak_result_type<_Tp>
+{
+public:
+    // types
+    typedef _Tp type;
+private:
+    type* __f_;
+
+public:
+    // construct/copy/destroy
+    _LIBCUDACXX_INLINE_VISIBILITY reference_wrapper(type& __f) _NOEXCEPT
+        : __f_(_CUDA_VSTD::addressof(__f)) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    private: _LIBCUDACXX_INLINE_VISIBILITY reference_wrapper(type&&); public: // = delete; // do not bind to temps
+#endif
+
+    // access
+    _LIBCUDACXX_INLINE_VISIBILITY operator type&    () const _NOEXCEPT {return *__f_;}
+    _LIBCUDACXX_INLINE_VISIBILITY          type& get() const _NOEXCEPT {return *__f_;}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    // invoke
+    template <class... _ArgTypes>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_of<type&, _ArgTypes...>::type
+    operator() (_ArgTypes&&... __args) const {
+        return __invoke(get(), _CUDA_VSTD::forward<_ArgTypes>(__args)...);
+    }
+#else
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return<type>::type
+    operator() () const {
+        return __invoke(get());
+    }
+
+    template <class _A0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return0<type, _A0>::type
+    operator() (_A0& __a0) const {
+        return __invoke(get(), __a0);
+    }
+
+    template <class _A0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return0<type, _A0 const>::type
+    operator() (_A0 const& __a0) const {
+        return __invoke(get(), __a0);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0, _A1>::type
+    operator() (_A0& __a0, _A1& __a1) const {
+        return __invoke(get(), __a0, __a1);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0 const, _A1>::type
+    operator() (_A0 const& __a0, _A1& __a1) const {
+        return __invoke(get(), __a0, __a1);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0, _A1 const>::type
+    operator() (_A0& __a0, _A1 const& __a1) const {
+        return __invoke(get(), __a0, __a1);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0 const, _A1 const>::type
+    operator() (_A0 const& __a0, _A1 const& __a1) const {
+        return __invoke(get(), __a0, __a1);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1, _A2>::type
+    operator() (_A0& __a0, _A1& __a1, _A2& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1, _A2>::type
+    operator() (_A0 const& __a0, _A1& __a1, _A2& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1 const, _A2>::type
+    operator() (_A0& __a0, _A1 const& __a1, _A2& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1, _A2 const>::type
+    operator() (_A0& __a0, _A1& __a1, _A2 const& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1 const, _A2>::type
+    operator() (_A0 const& __a0, _A1 const& __a1, _A2& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1, _A2 const>::type
+    operator() (_A0 const& __a0, _A1& __a1, _A2 const& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1 const, _A2 const>::type
+    operator() (_A0& __a0, _A1 const& __a1, _A2 const& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1 const, _A2 const>::type
+    operator() (_A0 const& __a0, _A1 const& __a1, _A2 const& __a2) const {
+        return __invoke(get(), __a0, __a1, __a2);
+    }
+#endif // _LIBCUDACXX_CXX03_LANG
+};
+
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+reference_wrapper<_Tp>
+ref(_Tp& __t) _NOEXCEPT
+{
+    return reference_wrapper<_Tp>(__t);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+reference_wrapper<_Tp>
+ref(reference_wrapper<_Tp> __t) _NOEXCEPT
+{
+    return ref(__t.get());
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+reference_wrapper<const _Tp>
+cref(const _Tp& __t) _NOEXCEPT
+{
+    return reference_wrapper<const _Tp>(__t);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+reference_wrapper<const _Tp>
+cref(reference_wrapper<_Tp> __t) _NOEXCEPT
+{
+    return cref(__t.get());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp> void ref(const _Tp&&) = delete;
+template <class _Tp> void cref(const _Tp&&) = delete;
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp, class, class = void>
+struct __is_transparent : false_type {};
+
+template <class _Tp, class _Up>
+struct __is_transparent<_Tp, _Up,
+                        typename __void_t<typename _Tp::is_transparent>::type>
+   : true_type {};
+#endif
+
+// allocator_arg_t
+
+struct _LIBCUDACXX_TEMPLATE_VIS allocator_arg_t { explicit allocator_arg_t() = default; };
+
+#if defined(_LIBCUDACXX_CXX03_LANG) || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+extern _LIBCUDACXX_EXPORTED_FROM_ABI const allocator_arg_t allocator_arg;
+#else
+/* _LIBCUDACXX_INLINE_VAR */ constexpr allocator_arg_t allocator_arg = allocator_arg_t();
+#endif
+
+// uses_allocator
+
+template <class _Tp>
+struct __has_allocator_type
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    template <class _Up> _LIBCUDACXX_INLINE_VISIBILITY static __two __test(...);
+    template <class _Up> _LIBCUDACXX_INLINE_VISIBILITY static char __test(typename _Up::allocator_type* = 0);
+public:
+    static const bool value = sizeof(__test<_Tp>(0)) == 1;
+};
+
+template <class _Tp, class _Alloc, bool = __has_allocator_type<_Tp>::value>
+struct __uses_allocator
+    : public integral_constant<bool,
+        is_convertible<_Alloc, typename _Tp::allocator_type>::value>
+{
+};
+
+template <class _Tp, class _Alloc>
+struct __uses_allocator<_Tp, _Alloc, false>
+    : public false_type
+{
+};
+
+template <class _Tp, class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator
+    : public __uses_allocator<_Tp, _Alloc>
+{
+};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp, class _Alloc>
+_LIBCUDACXX_INLINE_VAR constexpr size_t uses_allocator_v = uses_allocator<_Tp, _Alloc>::value;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+// allocator construction
+
+template <class _Tp, class _Alloc, class ..._Args>
+struct __uses_alloc_ctor_imp
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __uncvref<_Alloc>::type _RawAlloc;
+    static const bool __ua = uses_allocator<_Tp, _RawAlloc>::value;
+    static const bool __ic =
+        is_constructible<_Tp, allocator_arg_t, _Alloc, _Args...>::value;
+    static const int value = __ua ? 2 - __ic : 0;
+};
+
+template <class _Tp, class _Alloc, class ..._Args>
+struct __uses_alloc_ctor
+    : integral_constant<int, __uses_alloc_ctor_imp<_Tp, _Alloc, _Args...>::value>
+    {};
+
+template <class _Tp, class _Allocator, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __user_alloc_construct_impl (integral_constant<int, 0>, _Tp *__storage, const _Allocator &, _Args &&... __args )
+{
+    new (__storage) _Tp (_CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+// FIXME: This should have a version which takes a non-const alloc.
+template <class _Tp, class _Allocator, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __user_alloc_construct_impl (integral_constant<int, 1>, _Tp *__storage, const _Allocator &__a, _Args &&... __args )
+{
+    new (__storage) _Tp (allocator_arg, __a, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+// FIXME: This should have a version which takes a non-const alloc.
+template <class _Tp, class _Allocator, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __user_alloc_construct_impl (integral_constant<int, 2>, _Tp *__storage, const _Allocator &__a, _Args &&... __args )
+{
+    new (__storage) _Tp (_CUDA_VSTD::forward<_Args>(__args)..., __a);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX_FUNCTIONAL_BASE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base_03 b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base_03
new file mode 100644
index 000000000000..4187e9f5cf1d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__functional_base_03
@@ -0,0 +1,223 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FUNCTIONAL_BASE_03
+#define _LIBCUDACXX_FUNCTIONAL_BASE_03
+
+// manual variadic expansion for <functional>
+
+// __invoke
+
+template <class _Ret, class _T1, bool _IsFunc, bool _IsBase>
+struct __enable_invoke_imp;
+
+template <class _Ret, class _T1>
+struct __enable_invoke_imp<_Ret, _T1, true, true> {
+    typedef _Ret _Bullet1;
+    typedef _Bullet1 type;
+};
+
+template <class _Ret, class _T1>
+struct __enable_invoke_imp<_Ret, _T1, true, false>  {
+    typedef _Ret _Bullet2;
+    typedef _Bullet2 type;
+};
+
+template <class _Ret, class _T1>
+struct __enable_invoke_imp<_Ret, _T1, false, true>  {
+    typedef typename add_lvalue_reference<
+                typename __apply_cv<_T1, _Ret>::type
+            >::type _Bullet3;
+    typedef _Bullet3 type;
+};
+
+template <class _Ret, class _T1>
+struct __enable_invoke_imp<_Ret, _T1, false, false>  {
+    typedef typename add_lvalue_reference<
+                typename __apply_cv<decltype(*_CUDA_VSTD::declval<_T1>()), _Ret>::type
+            >::type _Bullet4;
+    typedef _Bullet4 type;
+};
+
+template <class _Ret, class _T1>
+struct __enable_invoke_imp<_Ret, _T1*, false, false>  {
+    typedef typename add_lvalue_reference<
+                typename __apply_cv<_T1, _Ret>::type
+            >::type _Bullet4;
+    typedef _Bullet4  type;
+};
+
+template <class _Fn, class _T1,
+          class _Traits = __member_pointer_traits<_Fn>,
+          class _Ret = typename _Traits::_ReturnType,
+          class _Class = typename _Traits::_ClassType>
+struct __enable_invoke : __enable_invoke_imp<
+    _Ret, _T1,
+    is_member_function_pointer<_Fn>::value,
+    is_base_of<_Class, typename remove_reference<_T1>::type>::value>
+{
+};
+
+__nat __invoke(__any, ...);
+
+// first bullet
+
+template <class _Fn, class _T1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet1
+__invoke(_Fn __f, _T1& __t1) {
+    return (__t1.*__f)();
+}
+
+template <class _Fn, class _T1, class _A0>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet1
+__invoke(_Fn __f, _T1& __t1, _A0& __a0) {
+    return (__t1.*__f)(__a0);
+}
+
+template <class _Fn, class _T1, class _A0, class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet1
+__invoke(_Fn __f, _T1& __t1, _A0& __a0, _A1& __a1) {
+    return (__t1.*__f)(__a0, __a1);
+}
+
+template <class _Fn, class _T1, class _A0, class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet1
+__invoke(_Fn __f, _T1& __t1, _A0& __a0, _A1& __a1, _A2& __a2) {
+    return (__t1.*__f)(__a0, __a1, __a2);
+}
+
+template <class _Fn, class _T1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet2
+__invoke(_Fn __f, _T1& __t1) {
+    return ((*__t1).*__f)();
+}
+
+template <class _Fn, class _T1, class _A0>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet2
+__invoke(_Fn __f, _T1& __t1, _A0& __a0) {
+    return ((*__t1).*__f)(__a0);
+}
+
+template <class _Fn, class _T1, class _A0, class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet2
+__invoke(_Fn __f, _T1& __t1, _A0& __a0, _A1& __a1) {
+    return ((*__t1).*__f)(__a0, __a1);
+}
+
+template <class _Fn, class _T1, class _A0, class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet2
+__invoke(_Fn __f, _T1& __t1, _A0& __a0, _A1& __a1, _A2& __a2) {
+    return ((*__t1).*__f)(__a0, __a1, __a2);
+}
+
+template <class _Fn, class _T1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet3
+__invoke(_Fn __f, _T1& __t1) {
+    return __t1.*__f;
+}
+
+template <class _Fn, class _T1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __enable_invoke<_Fn, _T1>::_Bullet4
+__invoke(_Fn __f, _T1& __t1) {
+    return (*__t1).*__f;
+}
+
+// fifth bullet
+
+template <class _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+decltype(_CUDA_VSTD::declval<_Fp&>()())
+__invoke(_Fp& __f)
+{
+    return __f();
+}
+
+template <class _Fp, class _A0>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+decltype(_CUDA_VSTD::declval<_Fp&>()(_CUDA_VSTD::declval<_A0&>()))
+__invoke(_Fp& __f, _A0& __a0)
+{
+    return __f(__a0);
+}
+
+template <class _Fp, class _A0, class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+decltype(_CUDA_VSTD::declval<_Fp&>()(_CUDA_VSTD::declval<_A0&>(), _CUDA_VSTD::declval<_A1&>()))
+__invoke(_Fp& __f, _A0& __a0, _A1& __a1)
+{
+    return __f(__a0, __a1);
+}
+
+template <class _Fp, class _A0, class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+decltype(_CUDA_VSTD::declval<_Fp&>()(_CUDA_VSTD::declval<_A0&>(), _CUDA_VSTD::declval<_A1&>(), _CUDA_VSTD::declval<_A2&>()))
+__invoke(_Fp& __f, _A0& __a0, _A1& __a1, _A2& __a2)
+{
+    return __f(__a0, __a1, __a2);
+}
+
+template <class _Fp, bool = __has_result_type<__weak_result_type<_Fp> >::value>
+struct __invoke_return
+{
+    typedef typename __weak_result_type<_Fp>::result_type type;
+};
+
+template <class _Fp>
+struct __invoke_return<_Fp, false>
+{
+    typedef decltype(__invoke(_CUDA_VSTD::declval<_Fp&>())) type;
+};
+
+template <class _Tp, class _A0>
+struct __invoke_return0
+{
+    typedef decltype(__invoke(_CUDA_VSTD::declval<_Tp&>(), _CUDA_VSTD::declval<_A0&>())) type;
+};
+
+template <class _Rp, class _Tp, class _A0>
+struct __invoke_return0<_Rp _Tp::*, _A0>
+{
+    typedef typename __enable_invoke<_Rp _Tp::*, _A0>::type type;
+};
+
+template <class _Tp, class _A0, class _A1>
+struct __invoke_return1
+{
+    typedef decltype(__invoke(_CUDA_VSTD::declval<_Tp&>(), _CUDA_VSTD::declval<_A0&>(),
+                                                      _CUDA_VSTD::declval<_A1&>())) type;
+};
+
+template <class _Rp, class _Class, class _A0, class _A1>
+struct __invoke_return1<_Rp _Class::*, _A0, _A1> {
+    typedef typename __enable_invoke<_Rp _Class::*, _A0>::type type;
+};
+
+template <class _Tp, class _A0, class _A1, class _A2>
+struct __invoke_return2
+{
+    typedef decltype(__invoke(_CUDA_VSTD::declval<_Tp&>(), _CUDA_VSTD::declval<_A0&>(),
+                                                      _CUDA_VSTD::declval<_A1&>(),
+                                                      _CUDA_VSTD::declval<_A2&>())) type;
+};
+
+template <class _Ret, class _Class, class _A0, class _A1, class _A2>
+struct __invoke_return2<_Ret _Class::*, _A0, _A1, _A2> {
+    typedef typename __enable_invoke<_Ret _Class::*, _A0>::type type;
+};
+#endif  // _LIBCUDACXX_FUNCTIONAL_BASE_03
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table
new file mode 100644
index 000000000000..9399fc210d79
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__hash_table
@@ -0,0 +1,2914 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX__HASH_TABLE
+#define _LIBCUDACXX__HASH_TABLE
+
+#include <__config>
+#include <initializer_list>
+#include <memory>
+#include <iterator>
+#include <algorithm>
+#include <cmath>
+#include <utility>
+#include <type_traits>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Key, class _Tp>
+struct __hash_value_type;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp>
+struct __is_hash_value_type_imp : false_type {};
+
+template <class _Key, class _Value>
+struct __is_hash_value_type_imp<__hash_value_type<_Key, _Value>> : true_type {};
+
+template <class ..._Args>
+struct __is_hash_value_type : false_type {};
+
+template <class _One>
+struct __is_hash_value_type<_One> : __is_hash_value_type_imp<typename __uncvref<_One>::type> {};
+#endif
+
+_LIBCUDACXX_FUNC_VIS
+size_t __next_prime(size_t __n);
+
+template <class _NodePtr>
+struct __hash_node_base
+{
+    typedef typename pointer_traits<_NodePtr>::element_type __node_type;
+    typedef __hash_node_base __first_node;
+    typedef typename __rebind_pointer<_NodePtr, __first_node>::type __node_base_pointer;
+    typedef _NodePtr __node_pointer;
+
+#if defined(_LIBCUDACXX_ABI_FIX_UNORDERED_NODE_POINTER_UB)
+  typedef __node_base_pointer __next_pointer;
+#else
+  typedef typename conditional<
+      is_pointer<__node_pointer>::value,
+      __node_base_pointer,
+      __node_pointer>::type   __next_pointer;
+#endif
+
+    __next_pointer    __next_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __next_pointer __ptr() _NOEXCEPT {
+        return static_cast<__next_pointer>(
+            pointer_traits<__node_base_pointer>::pointer_to(*this));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_pointer __upcast() _NOEXCEPT {
+        return static_cast<__node_pointer>(
+            pointer_traits<__node_base_pointer>::pointer_to(*this));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t __hash() const _NOEXCEPT {
+        return static_cast<__node_type const&>(*this).__hash_;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_node_base() _NOEXCEPT : __next_(nullptr) {}
+};
+
+template <class _Tp, class _VoidPtr>
+struct __hash_node
+    : public __hash_node_base
+             <
+                 typename __rebind_pointer<_VoidPtr, __hash_node<_Tp, _VoidPtr> >::type
+             >
+{
+    typedef _Tp __node_value_type;
+
+    size_t            __hash_;
+    __node_value_type __value_;
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+__is_hash_power2(size_t __bc)
+{
+    return __bc > 2 && !(__bc & (__bc - 1));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t
+__constrain_hash(size_t __h, size_t __bc)
+{
+    return !(__bc & (__bc - 1)) ? __h & (__bc - 1) :
+        (__h < __bc ? __h : __h % __bc);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t
+__next_hash_pow2(size_t __n)
+{
+    return __n < 2 ? __n : (size_t(1) << (std::numeric_limits<size_t>::digits - __libcpp_clz(__n-1)));
+}
+
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc> class __hash_table;
+
+template <class _NodePtr>      class _LIBCUDACXX_TEMPLATE_VIS __hash_iterator;
+template <class _ConstNodePtr> class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator;
+template <class _NodePtr>      class _LIBCUDACXX_TEMPLATE_VIS __hash_local_iterator;
+template <class _ConstNodePtr> class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator;
+template <class _HashIterator> class _LIBCUDACXX_TEMPLATE_VIS __hash_map_iterator;
+template <class _HashIterator> class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator;
+
+template <class _Tp>
+struct __hash_key_value_types {
+  static_assert(!is_reference<_Tp>::value && !is_const<_Tp>::value, "");
+  typedef _Tp key_type;
+  typedef _Tp __node_value_type;
+  typedef _Tp __container_value_type;
+  static const bool __is_map = false;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static key_type const& __get_key(_Tp const& __v) {
+    return __v;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type const& __get_value(__node_value_type const& __v) {
+    return __v;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type* __get_ptr(__node_value_type& __n) {
+    return _CUDA_VSTD::addressof(__n);
+  }
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type&& __move(__node_value_type& __v) {
+    return _CUDA_VSTD::move(__v);
+  }
+#endif
+};
+
+template <class _Key, class _Tp>
+struct __hash_key_value_types<__hash_value_type<_Key, _Tp> > {
+  typedef _Key                                         key_type;
+  typedef _Tp                                          mapped_type;
+  typedef __hash_value_type<_Key, _Tp>                 __node_value_type;
+  typedef pair<const _Key, _Tp>                        __container_value_type;
+  typedef __container_value_type                       __map_value_type;
+  static const bool __is_map = true;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static key_type const& __get_key(__container_value_type const& __v) {
+    return __v.first;
+  }
+
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static typename enable_if<__is_same_uncvref<_Up, __node_value_type>::value,
+      __container_value_type const&>::type
+  __get_value(_Up& __t) {
+    return __t.__get_value();
+  }
+
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static typename enable_if<__is_same_uncvref<_Up, __container_value_type>::value,
+      __container_value_type const&>::type
+  __get_value(_Up& __t) {
+    return __t;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type* __get_ptr(__node_value_type& __n) {
+    return _CUDA_VSTD::addressof(__n.__get_value());
+  }
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static pair<key_type&&, mapped_type&&> __move(__node_value_type& __v) {
+    return __v.__move();
+  }
+#endif
+
+};
+
+template <class _Tp, class _AllocPtr, class _KVTypes = __hash_key_value_types<_Tp>,
+          bool = _KVTypes::__is_map>
+struct __hash_map_pointer_types {};
+
+template <class _Tp, class _AllocPtr, class _KVTypes>
+struct __hash_map_pointer_types<_Tp, _AllocPtr, _KVTypes, true> {
+  typedef typename _KVTypes::__map_value_type   _Mv;
+  typedef typename __rebind_pointer<_AllocPtr, _Mv>::type
+                                                       __map_value_type_pointer;
+  typedef typename __rebind_pointer<_AllocPtr, const _Mv>::type
+                                                 __const_map_value_type_pointer;
+};
+
+template <class _NodePtr, class _NodeT = typename pointer_traits<_NodePtr>::element_type>
+struct __hash_node_types;
+
+template <class _NodePtr, class _Tp, class _VoidPtr>
+struct __hash_node_types<_NodePtr, __hash_node<_Tp, _VoidPtr> >
+    : public __hash_key_value_types<_Tp>, __hash_map_pointer_types<_Tp, _VoidPtr>
+
+{
+  typedef __hash_key_value_types<_Tp>           __base;
+
+public:
+  typedef ptrdiff_t difference_type;
+  typedef size_t size_type;
+
+  typedef typename __rebind_pointer<_NodePtr, void>::type       __void_pointer;
+
+  typedef typename pointer_traits<_NodePtr>::element_type       __node_type;
+  typedef _NodePtr                                              __node_pointer;
+
+  typedef __hash_node_base<__node_pointer>                      __node_base_type;
+  typedef typename __rebind_pointer<_NodePtr, __node_base_type>::type
+                                                             __node_base_pointer;
+
+  typedef typename __node_base_type::__next_pointer          __next_pointer;
+
+  typedef _Tp                                                 __node_value_type;
+  typedef typename __rebind_pointer<_VoidPtr, __node_value_type>::type
+                                                      __node_value_type_pointer;
+  typedef typename __rebind_pointer<_VoidPtr, const __node_value_type>::type
+                                                __const_node_value_type_pointer;
+
+private:
+    static_assert(!is_const<__node_type>::value,
+                "_NodePtr should never be a pointer to const");
+    static_assert((is_same<typename pointer_traits<_VoidPtr>::element_type, void>::value),
+                  "_VoidPtr does not point to unqualified void type");
+    static_assert((is_same<typename __rebind_pointer<_VoidPtr, __node_type>::type,
+                          _NodePtr>::value), "_VoidPtr does not rebind to _NodePtr.");
+};
+
+template <class _HashIterator>
+struct __hash_node_types_from_iterator;
+template <class _NodePtr>
+struct __hash_node_types_from_iterator<__hash_iterator<_NodePtr> > : __hash_node_types<_NodePtr> {};
+template <class _NodePtr>
+struct __hash_node_types_from_iterator<__hash_const_iterator<_NodePtr> > : __hash_node_types<_NodePtr> {};
+template <class _NodePtr>
+struct __hash_node_types_from_iterator<__hash_local_iterator<_NodePtr> > : __hash_node_types<_NodePtr> {};
+template <class _NodePtr>
+struct __hash_node_types_from_iterator<__hash_const_local_iterator<_NodePtr> > : __hash_node_types<_NodePtr> {};
+
+
+template <class _NodeValueTp, class _VoidPtr>
+struct __make_hash_node_types {
+  typedef __hash_node<_NodeValueTp, _VoidPtr> _NodeTp;
+  typedef typename __rebind_pointer<_VoidPtr, _NodeTp>::type _NodePtr;
+  typedef __hash_node_types<_NodePtr> type;
+};
+
+template <class _NodePtr>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_iterator
+{
+    typedef __hash_node_types<_NodePtr> _NodeTypes;
+    typedef _NodePtr                            __node_pointer;
+    typedef typename _NodeTypes::__next_pointer __next_pointer;
+
+    __next_pointer            __node_;
+
+public:
+    typedef forward_iterator_tag                           iterator_category;
+    typedef typename _NodeTypes::__node_value_type         value_type;
+    typedef typename _NodeTypes::difference_type           difference_type;
+    typedef value_type&                                    reference;
+    typedef typename _NodeTypes::__node_value_type_pointer pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_iterator() _NOEXCEPT : __node_(nullptr) {
+        _LIBCUDACXX_DEBUG_MODE(__get_db()->__insert_i(this));
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_iterator(const __hash_iterator& __i)
+        : __node_(__i.__node_)
+    {
+        __get_db()->__iterator_copy(this, &__i);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__hash_iterator()
+    {
+        __get_db()->__erase_i(this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_iterator& operator=(const __hash_iterator& __i)
+    {
+        if (this != &__i)
+        {
+            __get_db()->__iterator_copy(this, &__i);
+            __node_ = __i.__node_;
+        }
+        return *this;
+    }
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                             "Attempted to dereference a non-dereferenceable unordered container iterator");
+        return __node_->__upcast()->__value_;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                           "Attempted to dereference a non-dereferenceable unordered container iterator");
+        return pointer_traits<pointer>::pointer_to(__node_->__upcast()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_iterator& operator++() {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to increment non-incrementable unordered container iterator");
+        __node_ = __node_->__next_;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_iterator operator++(int)
+    {
+        __hash_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __hash_iterator& __x, const __hash_iterator& __y)
+    {
+        return __x.__node_ == __y.__node_;
+    }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __hash_iterator& __x, const __hash_iterator& __y)
+        {return !(__x == __y);}
+
+private:
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_iterator(__next_pointer __node, const void* __c) _NOEXCEPT
+        : __node_(__node)
+        {
+            __get_db()->__insert_ic(this, __c);
+        }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_iterator(__next_pointer __node) _NOEXCEPT
+        : __node_(__node)
+        {}
+#endif
+    template <class, class, class, class> friend class __hash_table;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_map_iterator;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+};
+
+template <class _NodePtr>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator
+{
+    static_assert(!is_const<typename pointer_traits<_NodePtr>::element_type>::value, "");
+    typedef __hash_node_types<_NodePtr> _NodeTypes;
+    typedef _NodePtr                            __node_pointer;
+    typedef typename _NodeTypes::__next_pointer __next_pointer;
+
+    __next_pointer __node_;
+
+public:
+    typedef __hash_iterator<_NodePtr> __non_const_iterator;
+
+    typedef forward_iterator_tag                                 iterator_category;
+    typedef typename _NodeTypes::__node_value_type               value_type;
+    typedef typename _NodeTypes::difference_type                 difference_type;
+    typedef const value_type&                                    reference;
+    typedef typename _NodeTypes::__const_node_value_type_pointer pointer;
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_const_iterator() _NOEXCEPT : __node_(nullptr) {
+        _LIBCUDACXX_DEBUG_MODE(__get_db()->__insert_i(this));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator(const __non_const_iterator& __x) _NOEXCEPT
+        : __node_(__x.__node_)
+    {
+        _LIBCUDACXX_DEBUG_MODE(__get_db()->__iterator_copy(this, &__x));
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator(const __hash_const_iterator& __i)
+        : __node_(__i.__node_)
+    {
+        __get_db()->__iterator_copy(this, &__i);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__hash_const_iterator()
+    {
+        __get_db()->__erase_i(this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator& operator=(const __hash_const_iterator& __i)
+    {
+        if (this != &__i)
+        {
+            __get_db()->__iterator_copy(this, &__i);
+            __node_ = __i.__node_;
+        }
+        return *this;
+    }
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                           "Attempted to dereference a non-dereferenceable unordered container const_iterator");
+        return __node_->__upcast()->__value_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                           "Attempted to dereference a non-dereferenceable unordered container const_iterator");
+        return pointer_traits<pointer>::pointer_to(__node_->__upcast()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator& operator++() {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                             "Attempted to increment non-incrementable unordered container const_iterator");
+        __node_ = __node_->__next_;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator operator++(int)
+    {
+        __hash_const_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __hash_const_iterator& __x, const __hash_const_iterator& __y)
+    {
+        return __x.__node_ == __y.__node_;
+    }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __hash_const_iterator& __x, const __hash_const_iterator& __y)
+        {return !(__x == __y);}
+
+private:
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator(__next_pointer __node, const void* __c) _NOEXCEPT
+        : __node_(__node)
+        {
+            __get_db()->__insert_ic(this, __c);
+        }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_iterator(__next_pointer __node) _NOEXCEPT
+        : __node_(__node)
+        {}
+#endif
+    template <class, class, class, class> friend class __hash_table;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+};
+
+template <class _NodePtr>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_local_iterator
+{
+    typedef __hash_node_types<_NodePtr> _NodeTypes;
+    typedef _NodePtr                            __node_pointer;
+    typedef typename _NodeTypes::__next_pointer __next_pointer;
+
+    __next_pointer         __node_;
+    size_t                 __bucket_;
+    size_t                 __bucket_count_;
+
+public:
+    typedef forward_iterator_tag                                iterator_category;
+    typedef typename _NodeTypes::__node_value_type              value_type;
+    typedef typename _NodeTypes::difference_type                difference_type;
+    typedef value_type&                                         reference;
+    typedef typename _NodeTypes::__node_value_type_pointer      pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_local_iterator() _NOEXCEPT : __node_(nullptr) {
+        _LIBCUDACXX_DEBUG_MODE(__get_db()->__insert_i(this));
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_local_iterator(const __hash_local_iterator& __i)
+        : __node_(__i.__node_),
+          __bucket_(__i.__bucket_),
+          __bucket_count_(__i.__bucket_count_)
+    {
+        __get_db()->__iterator_copy(this, &__i);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__hash_local_iterator()
+    {
+        __get_db()->__erase_i(this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_local_iterator& operator=(const __hash_local_iterator& __i)
+    {
+        if (this != &__i)
+        {
+            __get_db()->__iterator_copy(this, &__i);
+            __node_ = __i.__node_;
+            __bucket_ = __i.__bucket_;
+            __bucket_count_ = __i.__bucket_count_;
+        }
+        return *this;
+    }
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                           "Attempted to dereference a non-dereferenceable unordered container local_iterator");
+        return __node_->__upcast()->__value_;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                             "Attempted to dereference a non-dereferenceable unordered container local_iterator");
+        return pointer_traits<pointer>::pointer_to(__node_->__upcast()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_local_iterator& operator++() {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to increment non-incrementable unordered container local_iterator");
+        __node_ = __node_->__next_;
+        if (__node_ != nullptr && __constrain_hash(__node_->__hash(), __bucket_count_) != __bucket_)
+            __node_ = nullptr;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_local_iterator operator++(int)
+    {
+        __hash_local_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __hash_local_iterator& __x, const __hash_local_iterator& __y)
+    {
+        return __x.__node_ == __y.__node_;
+    }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __hash_local_iterator& __x, const __hash_local_iterator& __y)
+        {return !(__x == __y);}
+
+private:
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_local_iterator(__next_pointer __node, size_t __bucket,
+                          size_t __bucket_count, const void* __c) _NOEXCEPT
+        : __node_(__node),
+          __bucket_(__bucket),
+          __bucket_count_(__bucket_count)
+        {
+            __get_db()->__insert_ic(this, __c);
+            if (__node_ != nullptr)
+                __node_ = __node_->__next_;
+        }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_local_iterator(__next_pointer __node, size_t __bucket,
+                          size_t __bucket_count) _NOEXCEPT
+        : __node_(__node),
+          __bucket_(__bucket),
+          __bucket_count_(__bucket_count)
+        {
+            if (__node_ != nullptr)
+                __node_ = __node_->__next_;
+        }
+#endif
+    template <class, class, class, class> friend class __hash_table;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_map_iterator;
+};
+
+template <class _ConstNodePtr>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator
+{
+    typedef __hash_node_types<_ConstNodePtr> _NodeTypes;
+    typedef _ConstNodePtr                       __node_pointer;
+    typedef typename _NodeTypes::__next_pointer __next_pointer;
+
+    __next_pointer         __node_;
+    size_t                 __bucket_;
+    size_t                 __bucket_count_;
+
+    typedef pointer_traits<__node_pointer>          __pointer_traits;
+    typedef typename __pointer_traits::element_type __node;
+    typedef typename remove_const<__node>::type     __non_const_node;
+    typedef typename __rebind_pointer<__node_pointer, __non_const_node>::type
+        __non_const_node_pointer;
+public:
+    typedef __hash_local_iterator<__non_const_node_pointer>
+                                                    __non_const_iterator;
+
+    typedef forward_iterator_tag                                 iterator_category;
+    typedef typename _NodeTypes::__node_value_type               value_type;
+    typedef typename _NodeTypes::difference_type                 difference_type;
+    typedef const value_type&                                    reference;
+    typedef typename _NodeTypes::__const_node_value_type_pointer pointer;
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_const_local_iterator() _NOEXCEPT : __node_(nullptr) {
+        _LIBCUDACXX_DEBUG_MODE(__get_db()->__insert_i(this));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator(const __non_const_iterator& __x) _NOEXCEPT
+        : __node_(__x.__node_),
+          __bucket_(__x.__bucket_),
+          __bucket_count_(__x.__bucket_count_)
+    {
+        _LIBCUDACXX_DEBUG_MODE(__get_db()->__iterator_copy(this, &__x));
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator(const __hash_const_local_iterator& __i)
+        : __node_(__i.__node_),
+          __bucket_(__i.__bucket_),
+          __bucket_count_(__i.__bucket_count_)
+    {
+        __get_db()->__iterator_copy(this, &__i);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__hash_const_local_iterator()
+    {
+        __get_db()->__erase_i(this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator& operator=(const __hash_const_local_iterator& __i)
+    {
+        if (this != &__i)
+        {
+            __get_db()->__iterator_copy(this, &__i);
+            __node_ = __i.__node_;
+            __bucket_ = __i.__bucket_;
+            __bucket_count_ = __i.__bucket_count_;
+        }
+        return *this;
+    }
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                           "Attempted to dereference a non-dereferenceable unordered container const_local_iterator");
+        return __node_->__upcast()->__value_;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                           "Attempted to dereference a non-dereferenceable unordered container const_local_iterator");
+        return pointer_traits<pointer>::pointer_to(__node_->__upcast()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator& operator++() {
+        _LIBCUDACXX_DEBUG_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to increment non-incrementable unordered container const_local_iterator");
+        __node_ = __node_->__next_;
+        if (__node_ != nullptr && __constrain_hash(__node_->__hash(), __bucket_count_) != __bucket_)
+            __node_ = nullptr;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator operator++(int)
+    {
+        __hash_const_local_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __hash_const_local_iterator& __x, const __hash_const_local_iterator& __y)
+    {
+        return __x.__node_ == __y.__node_;
+    }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __hash_const_local_iterator& __x, const __hash_const_local_iterator& __y)
+        {return !(__x == __y);}
+
+private:
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator(__next_pointer __node, size_t __bucket,
+                                size_t __bucket_count, const void* __c) _NOEXCEPT
+        : __node_(__node),
+          __bucket_(__bucket),
+          __bucket_count_(__bucket_count)
+        {
+            __get_db()->__insert_ic(this, __c);
+            if (__node_ != nullptr)
+                __node_ = __node_->__next_;
+        }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_const_local_iterator(__next_pointer __node, size_t __bucket,
+                                size_t __bucket_count) _NOEXCEPT
+        : __node_(__node),
+          __bucket_(__bucket),
+          __bucket_count_(__bucket_count)
+        {
+            if (__node_ != nullptr)
+                __node_ = __node_->__next_;
+        }
+#endif
+    template <class, class, class, class> friend class __hash_table;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator;
+};
+
+template <class _Alloc>
+class __bucket_list_deallocator
+{
+    typedef _Alloc                                          allocator_type;
+    typedef allocator_traits<allocator_type>                __alloc_traits;
+    typedef typename __alloc_traits::size_type              size_type;
+
+    __compressed_pair<size_type, allocator_type> __data_;
+public:
+    typedef typename __alloc_traits::pointer pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bucket_list_deallocator()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+        : __data_(0) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bucket_list_deallocator(const allocator_type& __a, size_type __size)
+        _NOEXCEPT_(is_nothrow_copy_constructible<allocator_type>::value)
+        : __data_(__size, __a) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bucket_list_deallocator(__bucket_list_deallocator&& __x)
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value)
+        : __data_(_CUDA_VSTD::move(__x.__data_))
+    {
+        __x.size() = 0;
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type& size() _NOEXCEPT {return __data_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type  size() const _NOEXCEPT {return __data_.first();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type& __alloc() _NOEXCEPT {return __data_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const allocator_type& __alloc() const _NOEXCEPT {return __data_.second();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p) _NOEXCEPT
+    {
+        __alloc_traits::deallocate(__alloc(), __p, size());
+    }
+};
+
+template <class _Alloc> class __hash_map_node_destructor;
+
+template <class _Alloc>
+class __hash_node_destructor
+{
+    typedef _Alloc                                          allocator_type;
+    typedef allocator_traits<allocator_type>                __alloc_traits;
+
+public:
+    typedef typename __alloc_traits::pointer                pointer;
+private:
+    typedef __hash_node_types<pointer> _NodeTypes;
+
+    allocator_type& __na_;
+
+public:
+    bool __value_constructed;
+
+    __hash_node_destructor(__hash_node_destructor const&) = default;
+    __hash_node_destructor& operator=(const __hash_node_destructor&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __hash_node_destructor(allocator_type& __na,
+                                    bool __constructed = false) _NOEXCEPT
+        : __na_(__na),
+          __value_constructed(__constructed)
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p) _NOEXCEPT
+    {
+        if (__value_constructed)
+            __alloc_traits::destroy(__na_, _NodeTypes::__get_ptr(__p->__value_));
+        if (__p)
+            __alloc_traits::deallocate(__na_, __p, 1);
+    }
+
+    template <class> friend class __hash_map_node_destructor;
+};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _NodeType, class _Alloc>
+struct __generic_container_node_destructor;
+
+template <class _Tp, class _VoidPtr, class _Alloc>
+struct __generic_container_node_destructor<__hash_node<_Tp, _VoidPtr>, _Alloc>
+    : __hash_node_destructor<_Alloc>
+{
+    using __hash_node_destructor<_Alloc>::__hash_node_destructor;
+};
+#endif
+
+template <class _Key, class _Hash, class _Equal>
+struct __enforce_unordered_container_requirements {
+#ifndef _LIBCUDACXX_CXX03_LANG
+    static_assert(__check_hash_requirements<_Key, _Hash>::value,
+    "the specified hash does not meet the Hash requirements");
+    static_assert(is_copy_constructible<_Equal>::value,
+    "the specified comparator is required to be copy constructible");
+#endif
+    typedef int type;
+};
+
+template <class _Key, class _Hash, class _Equal>
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_DIAGNOSE_WARNING(!__invokable<_Equal const&, _Key const&, _Key const&>::value,
+    "the specified comparator type does not provide a viable const call operator")
+    _LIBCUDACXX_DIAGNOSE_WARNING(!__invokable<_Hash const&, _Key const&>::value,
+    "the specified hash functor does not provide a viable const call operator")
+#endif
+typename __enforce_unordered_container_requirements<_Key, _Hash, _Equal>::type
+__diagnose_unordered_container_requirements(int);
+
+// This dummy overload is used so that the compiler won't emit a spurious
+// "no matching function for call to __diagnose_unordered_xxx" diagnostic
+// when the overload above causes a hard error.
+template <class _Key, class _Hash, class _Equal>
+int __diagnose_unordered_container_requirements(void*);
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+class __hash_table
+{
+public:
+    typedef _Tp    value_type;
+    typedef _Hash  hasher;
+    typedef _Equal key_equal;
+    typedef _Alloc allocator_type;
+
+private:
+    typedef allocator_traits<allocator_type> __alloc_traits;
+    typedef typename
+      __make_hash_node_types<value_type, typename __alloc_traits::void_pointer>::type
+                                                                     _NodeTypes;
+public:
+
+    typedef typename _NodeTypes::__node_value_type           __node_value_type;
+    typedef typename _NodeTypes::__container_value_type      __container_value_type;
+    typedef typename _NodeTypes::key_type                    key_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+#ifndef _LIBCUDACXX_ABI_FIX_UNORDERED_CONTAINER_SIZE_TYPE
+    typedef typename __alloc_traits::size_type       size_type;
+#else
+    typedef typename _NodeTypes::size_type           size_type;
+#endif
+    typedef typename _NodeTypes::difference_type     difference_type;
+public:
+    // Create __node
+
+    typedef typename _NodeTypes::__node_type __node;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __node>::type __node_allocator;
+    typedef allocator_traits<__node_allocator>       __node_traits;
+    typedef typename _NodeTypes::__void_pointer      __void_pointer;
+    typedef typename _NodeTypes::__node_pointer      __node_pointer;
+    typedef typename _NodeTypes::__node_pointer      __node_const_pointer;
+    typedef typename _NodeTypes::__node_base_type    __first_node;
+    typedef typename _NodeTypes::__node_base_pointer __node_base_pointer;
+    typedef typename _NodeTypes::__next_pointer      __next_pointer;
+
+private:
+    // check for sane allocator pointer rebinding semantics. Rebinding the
+    // allocator for a new pointer type should be exactly the same as rebinding
+    // the pointer using 'pointer_traits'.
+    static_assert((is_same<__node_pointer, typename __node_traits::pointer>::value),
+                  "Allocator does not rebind pointers in a sane manner.");
+    typedef typename __rebind_alloc_helper<__node_traits, __first_node>::type
+        __node_base_allocator;
+    typedef allocator_traits<__node_base_allocator> __node_base_traits;
+    static_assert((is_same<__node_base_pointer, typename __node_base_traits::pointer>::value),
+                 "Allocator does not rebind pointers in a sane manner.");
+
+private:
+
+    typedef typename __rebind_alloc_helper<__node_traits, __next_pointer>::type __pointer_allocator;
+    typedef __bucket_list_deallocator<__pointer_allocator> __bucket_list_deleter;
+    typedef unique_ptr<__next_pointer[], __bucket_list_deleter> __bucket_list;
+    typedef allocator_traits<__pointer_allocator>          __pointer_alloc_traits;
+    typedef typename __bucket_list_deleter::pointer       __node_pointer_pointer;
+
+    // --- Member data begin ---
+    __bucket_list                                         __bucket_list_;
+    __compressed_pair<__first_node, __node_allocator>     __p1_;
+    __compressed_pair<size_type, hasher>                  __p2_;
+    __compressed_pair<float, key_equal>                   __p3_;
+    // --- Member data end ---
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type& size() _NOEXCEPT {return __p2_.first();}
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type  size() const _NOEXCEPT {return __p2_.first();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher& hash_function() _NOEXCEPT {return __p2_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const hasher& hash_function() const _NOEXCEPT {return __p2_.second();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float& max_load_factor() _NOEXCEPT {return __p3_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float  max_load_factor() const _NOEXCEPT {return __p3_.first();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal& key_eq() _NOEXCEPT {return __p3_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const key_equal& key_eq() const _NOEXCEPT {return __p3_.second();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_allocator& __node_alloc() _NOEXCEPT {return __p1_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const __node_allocator& __node_alloc() const _NOEXCEPT
+        {return __p1_.second();}
+
+public:
+    typedef __hash_iterator<__node_pointer>                   iterator;
+    typedef __hash_const_iterator<__node_pointer>             const_iterator;
+    typedef __hash_local_iterator<__node_pointer>             local_iterator;
+    typedef __hash_const_local_iterator<__node_pointer>       const_local_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_table()
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<__bucket_list>::value &&
+            is_nothrow_default_constructible<__first_node>::value &&
+            is_nothrow_default_constructible<__node_allocator>::value &&
+            is_nothrow_default_constructible<hasher>::value &&
+            is_nothrow_default_constructible<key_equal>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_table(const hasher& __hf, const key_equal& __eql);
+    __hash_table(const hasher& __hf, const key_equal& __eql,
+                 const allocator_type& __a);
+    explicit __hash_table(const allocator_type& __a);
+    __hash_table(const __hash_table& __u);
+    __hash_table(const __hash_table& __u, const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __hash_table(__hash_table&& __u)
+        _NOEXCEPT_(
+            is_nothrow_move_constructible<__bucket_list>::value &&
+            is_nothrow_move_constructible<__first_node>::value &&
+            is_nothrow_move_constructible<__node_allocator>::value &&
+            is_nothrow_move_constructible<hasher>::value &&
+            is_nothrow_move_constructible<key_equal>::value);
+    __hash_table(__hash_table&& __u, const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    ~__hash_table();
+
+    __hash_table& operator=(const __hash_table& __u);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_table& operator=(__hash_table&& __u)
+        _NOEXCEPT_(
+            __node_traits::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<__node_allocator>::value &&
+            is_nothrow_move_assignable<hasher>::value &&
+            is_nothrow_move_assignable<key_equal>::value);
+#endif
+    template <class _InputIterator>
+        void __assign_unique(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        void __assign_multi(_InputIterator __first, _InputIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT
+    {
+        return std::min<size_type>(
+            __node_traits::max_size(__node_alloc()),
+            numeric_limits<difference_type >::max()
+        );
+    }
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __next_pointer __node_insert_multi_prepare(size_t __cp_hash,
+                                               value_type& __cp_val);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __node_insert_multi_perform(__node_pointer __cp,
+                                     __next_pointer __pn) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __next_pointer __node_insert_unique_prepare(size_t __nd_hash,
+                                                value_type& __nd_val);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __node_insert_unique_perform(__node_pointer __ptr) _NOEXCEPT;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __node_insert_unique(__node_pointer __nd);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator             __node_insert_multi(__node_pointer __nd);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator             __node_insert_multi(const_iterator __p,
+                                             __node_pointer __nd);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Key, class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique_key_args(_Key const& __k, _Args&&... __args);
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique_impl(_Args&&... __args);
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique(_Pp&& __x) {
+      return __emplace_unique_extract_key(_CUDA_VSTD::forward<_Pp>(__x),
+                                          __can_extract_key<_Pp, key_type>());
+    }
+
+    template <class _First, class _Second>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<
+        __can_extract_map_key<_First, key_type, __container_value_type>::value,
+        pair<iterator, bool>
+    >::type __emplace_unique(_First&& __f, _Second&& __s) {
+        return __emplace_unique_key_args(__f, _CUDA_VSTD::forward<_First>(__f),
+                                              _CUDA_VSTD::forward<_Second>(__s));
+    }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique(_Args&&... __args) {
+      return __emplace_unique_impl(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __emplace_unique_extract_key(_Pp&& __x, __extract_key_fail_tag) {
+      return __emplace_unique_impl(_CUDA_VSTD::forward<_Pp>(__x));
+    }
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __emplace_unique_extract_key(_Pp&& __x, __extract_key_self_tag) {
+      return __emplace_unique_key_args(__x, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __emplace_unique_extract_key(_Pp&& __x, __extract_key_first_tag) {
+      return __emplace_unique_key_args(__x.first, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __emplace_multi(_Args&&... __args);
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __emplace_hint_multi(const_iterator __p, _Args&&... __args);
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __insert_unique(__container_value_type&& __x) {
+      return __emplace_unique_key_args(_NodeTypes::__get_key(__x), _CUDA_VSTD::move(__x));
+    }
+
+    template <class _Pp, class = typename enable_if<
+            !__is_same_uncvref<_Pp, __container_value_type>::value
+        >::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __insert_unique(_Pp&& __x) {
+      return __emplace_unique(_CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(_Pp&& __x) {
+      return __emplace_multi(_CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(const_iterator __p, _Pp&& __x) {
+        return __emplace_hint_multi(__p, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+#else  // !defined(_LIBCUDACXX_CXX03_LANG)
+    template <class _Key, class _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique_key_args(_Key const&, _Args& __args);
+
+    iterator __insert_multi(const __container_value_type& __x);
+    iterator __insert_multi(const_iterator __p, const __container_value_type& __x);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __insert_unique(const __container_value_type& __x) {
+        return __emplace_unique_key_args(_NodeTypes::__get_key(__x), __x);
+    }
+
+#if _LIBCUDACXX_STD_VER > 14
+    template <class _NodeHandle, class _InsertReturnType>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _InsertReturnType __node_handle_insert_unique(_NodeHandle&& __nh);
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_handle_insert_unique(const_iterator __hint,
+                                         _NodeHandle&& __nh);
+    template <class _Table>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __node_handle_merge_unique(_Table& __source);
+
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_handle_insert_multi(_NodeHandle&& __nh);
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_handle_insert_multi(const_iterator __hint, _NodeHandle&& __nh);
+    template <class _Table>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __node_handle_merge_multi(_Table& __source);
+
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _NodeHandle __node_handle_extract(key_type const& __key);
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _NodeHandle __node_handle_extract(const_iterator __it);
+#endif
+
+    void clear() _NOEXCEPT;
+    void rehash(size_type __n);
+    _LIBCUDACXX_INLINE_VISIBILITY void reserve(size_type __n)
+        {rehash(static_cast<size_type>(ceil(__n / max_load_factor())));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const _NOEXCEPT
+    {
+        return __bucket_list_.get_deleter().size();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT;
+
+    template <class _Key>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        size_type bucket(const _Key& __k) const
+        {
+            _LIBCUDACXX_ASSERT(bucket_count() > 0,
+                "unordered container::bucket(key) called when bucket_count() == 0");
+            return __constrain_hash(hash_function()(__k), bucket_count());
+        }
+
+    template <class _Key>
+        iterator       find(const _Key& __x);
+    template <class _Key>
+        const_iterator find(const _Key& __x) const;
+
+    typedef __hash_node_destructor<__node_allocator> _Dp;
+    typedef unique_ptr<__node, _Dp> __node_holder;
+
+    iterator erase(const_iterator __p);
+    iterator erase(const_iterator __first, const_iterator __last);
+    template <class _Key>
+        size_type __erase_unique(const _Key& __k);
+    template <class _Key>
+        size_type __erase_multi(const _Key& __k);
+    __node_holder remove(const_iterator __p) _NOEXCEPT;
+
+    template <class _Key>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        size_type __count_unique(const _Key& __k) const;
+    template <class _Key>
+        size_type __count_multi(const _Key& __k) const;
+
+    template <class _Key>
+        pair<iterator, iterator>
+        __equal_range_unique(const _Key& __k);
+    template <class _Key>
+        pair<const_iterator, const_iterator>
+        __equal_range_unique(const _Key& __k) const;
+
+    template <class _Key>
+        pair<iterator, iterator>
+        __equal_range_multi(const _Key& __k);
+    template <class _Key>
+        pair<const_iterator, const_iterator>
+        __equal_range_multi(const _Key& __k) const;
+
+    void swap(__hash_table& __u)
+#if _LIBCUDACXX_STD_VER <= 11
+        _NOEXCEPT_(
+            __is_nothrow_swappable<hasher>::value && __is_nothrow_swappable<key_equal>::value
+            && (!allocator_traits<__pointer_allocator>::propagate_on_container_swap::value
+                  || __is_nothrow_swappable<__pointer_allocator>::value)
+            && (!__node_traits::propagate_on_container_swap::value
+                  || __is_nothrow_swappable<__node_allocator>::value)
+            );
+#else
+     _NOEXCEPT_(__is_nothrow_swappable<hasher>::value && __is_nothrow_swappable<key_equal>::value);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const _NOEXCEPT
+        {return max_size(); }
+    size_type bucket_size(size_type __n) const;
+    _LIBCUDACXX_INLINE_VISIBILITY float load_factor() const _NOEXCEPT
+    {
+        size_type __bc = bucket_count();
+        return __bc != 0 ? (float)size() / __bc : 0.f;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY void max_load_factor(float __mlf) _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__mlf > 0,
+            "unordered container::max_load_factor(lf) called with lf <= 0");
+        max_load_factor() = _CUDA_VSTD::max(__mlf, load_factor());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator
+    begin(size_type __n)
+    {
+        _LIBCUDACXX_ASSERT(__n < bucket_count(),
+            "unordered container::begin(n) called with n >= bucket_count()");
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return local_iterator(__bucket_list_[__n], __n, bucket_count(), this);
+#else
+        return local_iterator(__bucket_list_[__n], __n, bucket_count());
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator
+    end(size_type __n)
+    {
+        _LIBCUDACXX_ASSERT(__n < bucket_count(),
+            "unordered container::end(n) called with n >= bucket_count()");
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return local_iterator(nullptr, __n, bucket_count(), this);
+#else
+        return local_iterator(nullptr, __n, bucket_count());
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator
+    cbegin(size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n < bucket_count(),
+            "unordered container::cbegin(n) called with n >= bucket_count()");
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return const_local_iterator(__bucket_list_[__n], __n, bucket_count(), this);
+#else
+        return const_local_iterator(__bucket_list_[__n], __n, bucket_count());
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator
+    cend(size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n < bucket_count(),
+            "unordered container::cend(n) called with n >= bucket_count()");
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return const_local_iterator(nullptr, __n, bucket_count(), this);
+#else
+        return const_local_iterator(nullptr, __n, bucket_count());
+#endif
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const;
+    bool __decrementable(const const_iterator* __i) const;
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const;
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const;
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+private:
+    void __rehash(size_type __n);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class ..._Args>
+    __node_holder __construct_node(_Args&& ...__args);
+
+    template <class _First, class ..._Rest>
+    __node_holder __construct_node_hash(size_t __hash, _First&& __f, _Rest&&... __rest);
+#else // _LIBCUDACXX_CXX03_LANG
+    __node_holder __construct_node(const __container_value_type& __v);
+    __node_holder __construct_node_hash(size_t __hash, const __container_value_type& __v);
+#endif
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __hash_table& __u)
+        {__copy_assign_alloc(__u, integral_constant<bool,
+             __node_traits::propagate_on_container_copy_assignment::value>());}
+    void __copy_assign_alloc(const __hash_table& __u, true_type);
+    _LIBCUDACXX_INLINE_VISIBILITY
+        void __copy_assign_alloc(const __hash_table&, false_type) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    void __move_assign(__hash_table& __u, false_type);
+    void __move_assign(__hash_table& __u, true_type)
+        _NOEXCEPT_(
+            is_nothrow_move_assignable<__node_allocator>::value &&
+            is_nothrow_move_assignable<hasher>::value &&
+            is_nothrow_move_assignable<key_equal>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__hash_table& __u)
+        _NOEXCEPT_(
+            !__node_traits::propagate_on_container_move_assignment::value ||
+            (is_nothrow_move_assignable<__pointer_allocator>::value &&
+             is_nothrow_move_assignable<__node_allocator>::value))
+        {__move_assign_alloc(__u, integral_constant<bool,
+             __node_traits::propagate_on_container_move_assignment::value>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__hash_table& __u, true_type)
+        _NOEXCEPT_(
+            is_nothrow_move_assignable<__pointer_allocator>::value &&
+            is_nothrow_move_assignable<__node_allocator>::value)
+    {
+        __bucket_list_.get_deleter().__alloc() =
+                _CUDA_VSTD::move(__u.__bucket_list_.get_deleter().__alloc());
+        __node_alloc() = _CUDA_VSTD::move(__u.__node_alloc());
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+        void __move_assign_alloc(__hash_table&, false_type) _NOEXCEPT {}
+#endif // _LIBCUDACXX_CXX03_LANG
+
+    void __deallocate_node(__next_pointer __np) _NOEXCEPT;
+    __next_pointer __detach() _NOEXCEPT;
+
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+};
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table()
+    _NOEXCEPT_(
+        is_nothrow_default_constructible<__bucket_list>::value &&
+        is_nothrow_default_constructible<__first_node>::value &&
+        is_nothrow_default_constructible<__node_allocator>::value &&
+        is_nothrow_default_constructible<hasher>::value &&
+        is_nothrow_default_constructible<key_equal>::value)
+    : __p2_(0),
+      __p3_(1.0f)
+{
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(const hasher& __hf,
+                                                       const key_equal& __eql)
+    : __bucket_list_(nullptr, __bucket_list_deleter()),
+      __p1_(),
+      __p2_(0, __hf),
+      __p3_(1.0f, __eql)
+{
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(const hasher& __hf,
+                                                       const key_equal& __eql,
+                                                       const allocator_type& __a)
+    : __bucket_list_(nullptr, __bucket_list_deleter(__pointer_allocator(__a), 0)),
+      __p1_(__second_tag(), __node_allocator(__a)),
+      __p2_(0, __hf),
+      __p3_(1.0f, __eql)
+{
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(const allocator_type& __a)
+    : __bucket_list_(nullptr, __bucket_list_deleter(__pointer_allocator(__a), 0)),
+      __p1_(__second_tag(), __node_allocator(__a)),
+      __p2_(0),
+      __p3_(1.0f)
+{
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(const __hash_table& __u)
+    : __bucket_list_(nullptr,
+          __bucket_list_deleter(allocator_traits<__pointer_allocator>::
+              select_on_container_copy_construction(
+                  __u.__bucket_list_.get_deleter().__alloc()), 0)),
+      __p1_(__second_tag(), allocator_traits<__node_allocator>::
+          select_on_container_copy_construction(__u.__node_alloc())),
+      __p2_(0, __u.hash_function()),
+      __p3_(__u.__p3_)
+{
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(const __hash_table& __u,
+                                                       const allocator_type& __a)
+    : __bucket_list_(nullptr, __bucket_list_deleter(__pointer_allocator(__a), 0)),
+      __p1_(__second_tag(), __node_allocator(__a)),
+      __p2_(0, __u.hash_function()),
+      __p3_(__u.__p3_)
+{
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(__hash_table&& __u)
+        _NOEXCEPT_(
+            is_nothrow_move_constructible<__bucket_list>::value &&
+            is_nothrow_move_constructible<__first_node>::value &&
+            is_nothrow_move_constructible<__node_allocator>::value &&
+            is_nothrow_move_constructible<hasher>::value &&
+            is_nothrow_move_constructible<key_equal>::value)
+    : __bucket_list_(_CUDA_VSTD::move(__u.__bucket_list_)),
+      __p1_(_CUDA_VSTD::move(__u.__p1_)),
+      __p2_(_CUDA_VSTD::move(__u.__p2_)),
+      __p3_(_CUDA_VSTD::move(__u.__p3_))
+{
+    if (size() > 0)
+    {
+        __bucket_list_[__constrain_hash(__p1_.first().__next_->__hash(), bucket_count())] =
+            __p1_.first().__ptr();
+        __u.__p1_.first().__next_ = nullptr;
+        __u.size() = 0;
+    }
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__hash_table(__hash_table&& __u,
+                                                       const allocator_type& __a)
+    : __bucket_list_(nullptr, __bucket_list_deleter(__pointer_allocator(__a), 0)),
+      __p1_(__second_tag(), __node_allocator(__a)),
+      __p2_(0, _CUDA_VSTD::move(__u.hash_function())),
+      __p3_(_CUDA_VSTD::move(__u.__p3_))
+{
+    if (__a == allocator_type(__u.__node_alloc()))
+    {
+        __bucket_list_.reset(__u.__bucket_list_.release());
+        __bucket_list_.get_deleter().size() = __u.__bucket_list_.get_deleter().size();
+        __u.__bucket_list_.get_deleter().size() = 0;
+        if (__u.size() > 0)
+        {
+            __p1_.first().__next_ = __u.__p1_.first().__next_;
+            __u.__p1_.first().__next_ = nullptr;
+            __bucket_list_[__constrain_hash(__p1_.first().__next_->__hash(), bucket_count())] =
+                __p1_.first().__ptr();
+            size() = __u.size();
+            __u.size() = 0;
+        }
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::~__hash_table()
+{
+#if defined(_LIBCUDACXX_CXX03_LANG)
+    static_assert((is_copy_constructible<key_equal>::value),
+                 "Predicate must be copy-constructible.");
+    static_assert((is_copy_constructible<hasher>::value),
+                 "Hasher must be copy-constructible.");
+#endif
+
+    __deallocate_node(__p1_.first().__next_);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__erase_c(this);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__copy_assign_alloc(
+        const __hash_table& __u, true_type)
+{
+    if (__node_alloc() != __u.__node_alloc())
+    {
+        clear();
+        __bucket_list_.reset();
+        __bucket_list_.get_deleter().size() = 0;
+    }
+    __bucket_list_.get_deleter().__alloc() = __u.__bucket_list_.get_deleter().__alloc();
+    __node_alloc() = __u.__node_alloc();
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>&
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::operator=(const __hash_table& __u)
+{
+    if (this != &__u)
+    {
+        __copy_assign_alloc(__u);
+        hash_function() = __u.hash_function();
+        key_eq() = __u.key_eq();
+        max_load_factor() = __u.max_load_factor();
+        __assign_multi(__u.begin(), __u.end());
+    }
+    return *this;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__deallocate_node(__next_pointer __np)
+    _NOEXCEPT
+{
+    __node_allocator& __na = __node_alloc();
+    while (__np != nullptr)
+    {
+        __next_pointer __next = __np->__next_;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __c_node* __c = __get_db()->__find_c_and_lock(this);
+        for (__i_node** __p = __c->end_; __p != __c->beg_; )
+        {
+            --__p;
+            iterator* __i = static_cast<iterator*>((*__p)->__i_);
+            if (__i->__node_ == __np)
+            {
+                (*__p)->__c_ = nullptr;
+                if (--__c->end_ != __p)
+                    memmove(__p, __p+1, (__c->end_ - __p)*sizeof(__i_node*));
+            }
+        }
+        __get_db()->unlock();
+#endif
+        __node_pointer __real_np = __np->__upcast();
+        __node_traits::destroy(__na, _NodeTypes::__get_ptr(__real_np->__value_));
+        __node_traits::deallocate(__na, __real_np, 1);
+        __np = __next;
+    }
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__next_pointer
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__detach() _NOEXCEPT
+{
+    size_type __bc = bucket_count();
+    for (size_type __i = 0; __i < __bc; ++__i)
+        __bucket_list_[__i] = nullptr;
+    size() = 0;
+    __next_pointer __cache = __p1_.first().__next_;
+    __p1_.first().__next_ = nullptr;
+    return __cache;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__move_assign(
+        __hash_table& __u, true_type)
+    _NOEXCEPT_(
+        is_nothrow_move_assignable<__node_allocator>::value &&
+        is_nothrow_move_assignable<hasher>::value &&
+        is_nothrow_move_assignable<key_equal>::value)
+{
+    clear();
+    __bucket_list_.reset(__u.__bucket_list_.release());
+    __bucket_list_.get_deleter().size() = __u.__bucket_list_.get_deleter().size();
+    __u.__bucket_list_.get_deleter().size() = 0;
+    __move_assign_alloc(__u);
+    size() = __u.size();
+    hash_function() = _CUDA_VSTD::move(__u.hash_function());
+    max_load_factor() = __u.max_load_factor();
+    key_eq() = _CUDA_VSTD::move(__u.key_eq());
+    __p1_.first().__next_ = __u.__p1_.first().__next_;
+    if (size() > 0)
+    {
+        __bucket_list_[__constrain_hash(__p1_.first().__next_->__hash(), bucket_count())] =
+            __p1_.first().__ptr();
+        __u.__p1_.first().__next_ = nullptr;
+        __u.size() = 0;
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__move_assign(
+        __hash_table& __u, false_type)
+{
+    if (__node_alloc() == __u.__node_alloc())
+        __move_assign(__u, true_type());
+    else
+    {
+        hash_function() = _CUDA_VSTD::move(__u.hash_function());
+        key_eq() = _CUDA_VSTD::move(__u.key_eq());
+        max_load_factor() = __u.max_load_factor();
+        if (bucket_count() != 0)
+        {
+            __next_pointer __cache = __detach();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            try
+            {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+                const_iterator __i = __u.begin();
+                while (__cache != nullptr && __u.size() != 0)
+                {
+                    __cache->__upcast()->__value_ =
+                        _CUDA_VSTD::move(__u.remove(__i++)->__value_);
+                    __next_pointer __next = __cache->__next_;
+                    __node_insert_multi(__cache->__upcast());
+                    __cache = __next;
+                }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            }
+            catch (...)
+            {
+                __deallocate_node(__cache);
+                throw;
+            }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            __deallocate_node(__cache);
+        }
+        const_iterator __i = __u.begin();
+        while (__u.size() != 0)
+        {
+            __node_holder __h = __construct_node(_NodeTypes::__move(__u.remove(__i++)->__value_));
+            __node_insert_multi(__h.get());
+            __h.release();
+        }
+    }
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+__hash_table<_Tp, _Hash, _Equal, _Alloc>&
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::operator=(__hash_table&& __u)
+    _NOEXCEPT_(
+        __node_traits::propagate_on_container_move_assignment::value &&
+        is_nothrow_move_assignable<__node_allocator>::value &&
+        is_nothrow_move_assignable<hasher>::value &&
+        is_nothrow_move_assignable<key_equal>::value)
+{
+    __move_assign(__u, integral_constant<bool,
+                  __node_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _InputIterator>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__assign_unique(_InputIterator __first,
+                                                          _InputIterator __last)
+{
+    typedef iterator_traits<_InputIterator> _ITraits;
+    typedef typename _ITraits::value_type _ItValueType;
+    static_assert((is_same<_ItValueType, __container_value_type>::value),
+                  "__assign_unique may only be called with the containers value type");
+
+    if (bucket_count() != 0)
+    {
+        __next_pointer __cache = __detach();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (; __cache != nullptr && __first != __last; ++__first)
+            {
+                __cache->__upcast()->__value_ = *__first;
+                __next_pointer __next = __cache->__next_;
+                __node_insert_unique(__cache->__upcast());
+                __cache = __next;
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __deallocate_node(__cache);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __deallocate_node(__cache);
+    }
+    for (; __first != __last; ++__first)
+        __insert_unique(*__first);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _InputIterator>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__assign_multi(_InputIterator __first,
+                                                         _InputIterator __last)
+{
+    typedef iterator_traits<_InputIterator> _ITraits;
+    typedef typename _ITraits::value_type _ItValueType;
+    static_assert((is_same<_ItValueType, __container_value_type>::value ||
+                  is_same<_ItValueType, __node_value_type>::value),
+                  "__assign_multi may only be called with the containers value type"
+                  " or the nodes value type");
+    if (bucket_count() != 0)
+    {
+        __next_pointer __cache = __detach();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (; __cache != nullptr && __first != __last; ++__first)
+            {
+                __cache->__upcast()->__value_ = *__first;
+                __next_pointer __next = __cache->__next_;
+                __node_insert_multi(__cache->__upcast());
+                __cache = __next;
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __deallocate_node(__cache);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __deallocate_node(__cache);
+    }
+    for (; __first != __last; ++__first)
+        __insert_multi(_NodeTypes::__get_value(*__first));
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::begin() _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__p1_.first().__next_, this);
+#else
+    return iterator(__p1_.first().__next_);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::end() _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(nullptr, this);
+#else
+    return iterator(nullptr);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::begin() const _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return const_iterator(__p1_.first().__next_, this);
+#else
+    return const_iterator(__p1_.first().__next_);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::end() const _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return const_iterator(nullptr, this);
+#else
+    return const_iterator(nullptr);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::clear() _NOEXCEPT
+{
+    if (size() > 0)
+    {
+        __deallocate_node(__p1_.first().__next_);
+        __p1_.first().__next_ = nullptr;
+        size_type __bc = bucket_count();
+        for (size_type __i = 0; __i < __bc; ++__i)
+            __bucket_list_[__i] = nullptr;
+        size() = 0;
+    }
+}
+
+
+// Prepare the container for an insertion of the value __value with the hash
+// __hash. This does a lookup into the container to see if __value is already
+// present, and performs a rehash if necessary. Returns a pointer to the
+// existing element if it exists, otherwise nullptr.
+//
+// Note that this function does forward exceptions if key_eq() throws, and never
+// mutates __value or actually inserts into the map.
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__next_pointer
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_unique_prepare(
+    size_t __hash, value_type& __value)
+{
+    size_type __bc = bucket_count();
+
+    if (__bc != 0)
+    {
+        size_t __chash = __constrain_hash(__hash, __bc);
+        __next_pointer __ndptr = __bucket_list_[__chash];
+        if (__ndptr != nullptr)
+        {
+            for (__ndptr = __ndptr->__next_; __ndptr != nullptr &&
+                                             __constrain_hash(__ndptr->__hash(), __bc) == __chash;
+                                                     __ndptr = __ndptr->__next_)
+            {
+                if (key_eq()(__ndptr->__upcast()->__value_, __value))
+                    return __ndptr;
+            }
+        }
+    }
+    if (size()+1 > __bc * max_load_factor() || __bc == 0)
+    {
+        rehash(_CUDA_VSTD::max<size_type>(2 * __bc + !__is_hash_power2(__bc),
+                                     size_type(ceil(float(size() + 1) / max_load_factor()))));
+    }
+    return nullptr;
+}
+
+// Insert the node __nd into the container by pushing it into the right bucket,
+// and updating size(). Assumes that __nd->__hash is up-to-date, and that
+// rehashing has already occurred and that no element with the same key exists
+// in the map.
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_unique_perform(
+    __node_pointer __nd) _NOEXCEPT
+{
+    size_type __bc = bucket_count();
+    size_t __chash = __constrain_hash(__nd->__hash(), __bc);
+    // insert_after __bucket_list_[__chash], or __first_node if bucket is null
+    __next_pointer __pn = __bucket_list_[__chash];
+    if (__pn == nullptr)
+    {
+        __pn =__p1_.first().__ptr();
+        __nd->__next_ = __pn->__next_;
+        __pn->__next_ = __nd->__ptr();
+        // fix up __bucket_list_
+        __bucket_list_[__chash] = __pn;
+        if (__nd->__next_ != nullptr)
+            __bucket_list_[__constrain_hash(__nd->__next_->__hash(), __bc)] = __nd->__ptr();
+    }
+    else
+    {
+        __nd->__next_ = __pn->__next_;
+        __pn->__next_ = __nd->__ptr();
+    }
+    ++size();
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator, bool>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_unique(__node_pointer __nd)
+{
+    __nd->__hash_ = hash_function()(__nd->__value_);
+    __next_pointer __existing_node =
+        __node_insert_unique_prepare(__nd->__hash(), __nd->__value_);
+
+    // Insert the node, unless it already exists in the container.
+    bool __inserted = false;
+    if (__existing_node == nullptr)
+    {
+        __node_insert_unique_perform(__nd);
+        __existing_node = __nd->__ptr();
+        __inserted = true;
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return pair<iterator, bool>(iterator(__existing_node, this), __inserted);
+#else
+    return pair<iterator, bool>(iterator(__existing_node), __inserted);
+#endif
+}
+
+// Prepare the container for an insertion of the value __cp_val with the hash
+// __cp_hash. This does a lookup into the container to see if __cp_value is
+// already present, and performs a rehash if necessary. Returns a pointer to the
+// last occurance of __cp_val in the map.
+//
+// Note that this function does forward exceptions if key_eq() throws, and never
+// mutates __value or actually inserts into the map.
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__next_pointer
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi_prepare(
+    size_t __cp_hash, value_type& __cp_val)
+{
+    size_type __bc = bucket_count();
+    if (size()+1 > __bc * max_load_factor() || __bc == 0)
+    {
+        rehash(_CUDA_VSTD::max<size_type>(2 * __bc + !__is_hash_power2(__bc),
+                       size_type(ceil(float(size() + 1) / max_load_factor()))));
+        __bc = bucket_count();
+    }
+    size_t __chash = __constrain_hash(__cp_hash, __bc);
+    __next_pointer __pn = __bucket_list_[__chash];
+    if (__pn != nullptr)
+    {
+        for (bool __found = false; __pn->__next_ != nullptr &&
+                                   __constrain_hash(__pn->__next_->__hash(), __bc) == __chash;
+                                                           __pn = __pn->__next_)
+        {
+            //      __found    key_eq()     action
+            //      false       false       loop
+            //      true        true        loop
+            //      false       true        set __found to true
+            //      true        false       break
+            if (__found != (__pn->__next_->__hash() == __cp_hash &&
+                            key_eq()(__pn->__next_->__upcast()->__value_, __cp_val)))
+            {
+                if (!__found)
+                    __found = true;
+                else
+                    break;
+            }
+        }
+    }
+    return __pn;
+}
+
+// Insert the node __cp into the container after __pn (which is the last node in
+// the bucket that compares equal to __cp). Rehashing, and checking for
+// uniqueness has already been performed (in __node_insert_multi_prepare), so
+// all we need to do is update the bucket and size(). Assumes that __cp->__hash
+// is up-to-date.
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi_perform(
+    __node_pointer __cp, __next_pointer __pn) _NOEXCEPT
+{
+    size_type __bc = bucket_count();
+    size_t __chash = __constrain_hash(__cp->__hash_, __bc);
+    if (__pn == nullptr)
+    {
+        __pn =__p1_.first().__ptr();
+        __cp->__next_ = __pn->__next_;
+        __pn->__next_ = __cp->__ptr();
+        // fix up __bucket_list_
+        __bucket_list_[__chash] = __pn;
+        if (__cp->__next_ != nullptr)
+            __bucket_list_[__constrain_hash(__cp->__next_->__hash(), __bc)]
+                = __cp->__ptr();
+    }
+    else
+    {
+        __cp->__next_ = __pn->__next_;
+        __pn->__next_ = __cp->__ptr();
+        if (__cp->__next_ != nullptr)
+        {
+            size_t __nhash = __constrain_hash(__cp->__next_->__hash(), __bc);
+            if (__nhash != __chash)
+                __bucket_list_[__nhash] = __cp->__ptr();
+        }
+    }
+    ++size();
+}
+
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi(__node_pointer __cp)
+{
+    __cp->__hash_ = hash_function()(__cp->__value_);
+    __next_pointer __pn = __node_insert_multi_prepare(__cp->__hash(), __cp->__value_);
+    __node_insert_multi_perform(__cp, __pn);
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__cp->__ptr(), this);
+#else
+    return iterator(__cp->__ptr());
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_insert_multi(
+        const_iterator __p, __node_pointer __cp)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "unordered container::emplace_hint(const_iterator, args...) called with an iterator not"
+        " referring to this unordered container");
+#endif
+    if (__p != end() && key_eq()(*__p, __cp->__value_))
+    {
+        __next_pointer __np = __p.__node_;
+        __cp->__hash_ = __np->__hash();
+        size_type __bc = bucket_count();
+        if (size()+1 > __bc * max_load_factor() || __bc == 0)
+        {
+            rehash(_CUDA_VSTD::max<size_type>(2 * __bc + !__is_hash_power2(__bc),
+                           size_type(ceil(float(size() + 1) / max_load_factor()))));
+            __bc = bucket_count();
+        }
+        size_t __chash = __constrain_hash(__cp->__hash_, __bc);
+        __next_pointer __pp = __bucket_list_[__chash];
+        while (__pp->__next_ != __np)
+            __pp = __pp->__next_;
+        __cp->__next_ = __np;
+        __pp->__next_ = static_cast<__next_pointer>(__cp);
+        ++size();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return iterator(static_cast<__next_pointer>(__cp), this);
+#else
+        return iterator(static_cast<__next_pointer>(__cp));
+#endif
+    }
+    return __node_insert_multi(__cp);
+}
+
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key, class ..._Args>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator, bool>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__emplace_unique_key_args(_Key const& __k, _Args&&... __args)
+#else
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key, class _Args>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator, bool>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__emplace_unique_key_args(_Key const& __k, _Args& __args)
+#endif
+{
+
+    size_t __hash = hash_function()(__k);
+    size_type __bc = bucket_count();
+    bool __inserted = false;
+    __next_pointer __nd;
+    size_t __chash;
+    if (__bc != 0)
+    {
+        __chash = __constrain_hash(__hash, __bc);
+        __nd = __bucket_list_[__chash];
+        if (__nd != nullptr)
+        {
+            for (__nd = __nd->__next_; __nd != nullptr &&
+                (__nd->__hash() == __hash || __constrain_hash(__nd->__hash(), __bc) == __chash);
+                                                           __nd = __nd->__next_)
+            {
+                if (key_eq()(__nd->__upcast()->__value_, __k))
+                    goto __done;
+            }
+        }
+    }
+    {
+#ifndef _LIBCUDACXX_CXX03_LANG
+        __node_holder __h = __construct_node_hash(__hash, _CUDA_VSTD::forward<_Args>(__args)...);
+#else
+        __node_holder __h = __construct_node_hash(__hash, __args);
+#endif
+        if (size()+1 > __bc * max_load_factor() || __bc == 0)
+        {
+            rehash(_CUDA_VSTD::max<size_type>(2 * __bc + !__is_hash_power2(__bc),
+                           size_type(ceil(float(size() + 1) / max_load_factor()))));
+            __bc = bucket_count();
+            __chash = __constrain_hash(__hash, __bc);
+        }
+        // insert_after __bucket_list_[__chash], or __first_node if bucket is null
+        __next_pointer __pn = __bucket_list_[__chash];
+        if (__pn == nullptr)
+        {
+            __pn = __p1_.first().__ptr();
+            __h->__next_ = __pn->__next_;
+            __pn->__next_ = __h.get()->__ptr();
+            // fix up __bucket_list_
+            __bucket_list_[__chash] = __pn;
+            if (__h->__next_ != nullptr)
+                __bucket_list_[__constrain_hash(__h->__next_->__hash(), __bc)]
+                    = __h.get()->__ptr();
+        }
+        else
+        {
+            __h->__next_ = __pn->__next_;
+            __pn->__next_ = static_cast<__next_pointer>(__h.get());
+        }
+        __nd = static_cast<__next_pointer>(__h.release());
+        // increment size
+        ++size();
+        __inserted = true;
+    }
+__done:
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return pair<iterator, bool>(iterator(__nd, this), __inserted);
+#else
+    return pair<iterator, bool>(iterator(__nd), __inserted);
+#endif
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class... _Args>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator, bool>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__emplace_unique_impl(_Args&&... __args)
+{
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    pair<iterator, bool> __r = __node_insert_unique(__h.get());
+    if (__r.second)
+        __h.release();
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class... _Args>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__emplace_multi(_Args&&... __args)
+{
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    iterator __r = __node_insert_multi(__h.get());
+    __h.release();
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class... _Args>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__emplace_hint_multi(
+        const_iterator __p, _Args&&... __args)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "unordered container::emplace_hint(const_iterator, args...) called with an iterator not"
+        " referring to this unordered container");
+#endif
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    iterator __r = __node_insert_multi(__p, __h.get());
+    __h.release();
+    return __r;
+}
+
+#else // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__insert_multi(const __container_value_type& __x)
+{
+    __node_holder __h = __construct_node(__x);
+    iterator __r = __node_insert_multi(__h.get());
+    __h.release();
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__insert_multi(const_iterator __p,
+                                                         const __container_value_type& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "unordered container::insert(const_iterator, lvalue) called with an iterator not"
+        " referring to this unordered container");
+#endif
+    __node_holder __h = __construct_node(__x);
+    iterator __r = __node_insert_multi(__p, __h.get());
+    __h.release();
+    return __r;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _NodeHandle, class _InsertReturnType>
+_LIBCUDACXX_INLINE_VISIBILITY
+_InsertReturnType
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_insert_unique(
+    _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return _InsertReturnType{end(), false, _NodeHandle()};
+    pair<iterator, bool> __result = __node_insert_unique(__nh.__ptr_);
+    if (__result.second)
+        __nh.__release_ptr();
+    return _InsertReturnType{__result.first, __result.second, _CUDA_VSTD::move(__nh)};
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_insert_unique(
+    const_iterator, _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return end();
+    pair<iterator, bool> __result = __node_insert_unique(__nh.__ptr_);
+    if (__result.second)
+        __nh.__release_ptr();
+    return __result.first;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+_NodeHandle
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_extract(
+    key_type const& __key)
+{
+    iterator __i = find(__key);
+    if (__i == end())
+        return _NodeHandle();
+    return __node_handle_extract<_NodeHandle>(__i);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+_NodeHandle
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_extract(
+    const_iterator __p)
+{
+    allocator_type __alloc(__node_alloc());
+    return _NodeHandle(remove(__p).release(), __alloc);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Table>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_merge_unique(
+    _Table& __source)
+{
+    static_assert(is_same<__node, typename _Table::__node>::value, "");
+
+    for (typename _Table::iterator __it = __source.begin();
+         __it != __source.end();)
+    {
+        __node_pointer __src_ptr = __it.__node_->__upcast();
+        size_t __hash = hash_function()(__src_ptr->__value_);
+        __next_pointer __existing_node =
+            __node_insert_unique_prepare(__hash, __src_ptr->__value_);
+        auto __prev_iter = __it++;
+        if (__existing_node == nullptr)
+        {
+            (void)__source.remove(__prev_iter).release();
+            __src_ptr->__hash_ = __hash;
+            __node_insert_unique_perform(__src_ptr);
+        }
+    }
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_insert_multi(
+    _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return end();
+    iterator __result = __node_insert_multi(__nh.__ptr_);
+    __nh.__release_ptr();
+    return __result;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_insert_multi(
+    const_iterator __hint, _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return end();
+    iterator __result = __node_insert_multi(__hint, __nh.__ptr_);
+    __nh.__release_ptr();
+    return __result;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Table>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_handle_merge_multi(
+    _Table& __source)
+{
+    static_assert(is_same<typename _Table::__node, __node>::value, "");
+
+    for (typename _Table::iterator __it = __source.begin();
+         __it != __source.end();)
+    {
+        __node_pointer __src_ptr = __it.__node_->__upcast();
+        size_t __src_hash = hash_function()(__src_ptr->__value_);
+        __next_pointer __pn =
+            __node_insert_multi_prepare(__src_hash, __src_ptr->__value_);
+        (void)__source.remove(__it++).release();
+        __src_ptr->__hash_ = __src_hash;
+        __node_insert_multi_perform(__src_ptr, __pn);
+    }
+}
+#endif  // _LIBCUDACXX_STD_VER > 14
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::rehash(size_type __n)
+_LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+{
+    if (__n == 1)
+        __n = 2;
+    else if (__n & (__n - 1))
+        __n = __next_prime(__n);
+    size_type __bc = bucket_count();
+    if (__n > __bc)
+        __rehash(__n);
+    else if (__n < __bc)
+    {
+        __n = _CUDA_VSTD::max<size_type>
+              (
+                  __n,
+                  __is_hash_power2(__bc) ? __next_hash_pow2(size_t(ceil(float(size()) / max_load_factor()))) :
+                                           __next_prime(size_t(ceil(float(size()) / max_load_factor())))
+              );
+        if (__n < __bc)
+            __rehash(__n);
+    }
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__rehash(size_type __nbc)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__invalidate_all(this);
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __pointer_allocator& __npa = __bucket_list_.get_deleter().__alloc();
+    __bucket_list_.reset(__nbc > 0 ?
+                      __pointer_alloc_traits::allocate(__npa, __nbc) : nullptr);
+    __bucket_list_.get_deleter().size() = __nbc;
+    if (__nbc > 0)
+    {
+        for (size_type __i = 0; __i < __nbc; ++__i)
+            __bucket_list_[__i] = nullptr;
+        __next_pointer __pp = __p1_.first().__ptr();
+        __next_pointer __cp = __pp->__next_;
+        if (__cp != nullptr)
+        {
+            size_type __chash = __constrain_hash(__cp->__hash(), __nbc);
+            __bucket_list_[__chash] = __pp;
+            size_type __phash = __chash;
+            for (__pp = __cp, __cp = __cp->__next_; __cp != nullptr;
+                                                           __cp = __pp->__next_)
+            {
+                __chash = __constrain_hash(__cp->__hash(), __nbc);
+                if (__chash == __phash)
+                    __pp = __cp;
+                else
+                {
+                    if (__bucket_list_[__chash] == nullptr)
+                    {
+                        __bucket_list_[__chash] = __pp;
+                        __pp = __cp;
+                        __phash = __chash;
+                    }
+                    else
+                    {
+                        __next_pointer __np = __cp;
+                        for (; __np->__next_ != nullptr &&
+                               key_eq()(__cp->__upcast()->__value_,
+                                        __np->__next_->__upcast()->__value_);
+                                                           __np = __np->__next_)
+                            ;
+                        __pp->__next_ = __np->__next_;
+                        __np->__next_ = __bucket_list_[__chash]->__next_;
+                        __bucket_list_[__chash]->__next_ = __cp;
+
+                    }
+                }
+            }
+        }
+    }
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::find(const _Key& __k)
+{
+    size_t __hash = hash_function()(__k);
+    size_type __bc = bucket_count();
+    if (__bc != 0)
+    {
+        size_t __chash = __constrain_hash(__hash, __bc);
+        __next_pointer __nd = __bucket_list_[__chash];
+        if (__nd != nullptr)
+        {
+            for (__nd = __nd->__next_; __nd != nullptr &&
+                (__nd->__hash() == __hash
+                  || __constrain_hash(__nd->__hash(), __bc) == __chash);
+                                                           __nd = __nd->__next_)
+            {
+                if ((__nd->__hash() == __hash)
+                    && key_eq()(__nd->__upcast()->__value_, __k))
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                    return iterator(__nd, this);
+#else
+                    return iterator(__nd);
+#endif
+            }
+        }
+    }
+    return end();
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::find(const _Key& __k) const
+{
+    size_t __hash = hash_function()(__k);
+    size_type __bc = bucket_count();
+    if (__bc != 0)
+    {
+        size_t __chash = __constrain_hash(__hash, __bc);
+        __next_pointer __nd = __bucket_list_[__chash];
+        if (__nd != nullptr)
+        {
+            for (__nd = __nd->__next_; __nd != nullptr &&
+                (__hash == __nd->__hash()
+                    || __constrain_hash(__nd->__hash(), __bc) == __chash);
+                                                           __nd = __nd->__next_)
+            {
+                if ((__nd->__hash() == __hash)
+                    && key_eq()(__nd->__upcast()->__value_, __k))
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                    return const_iterator(__nd, this);
+#else
+                    return const_iterator(__nd);
+#endif
+            }
+        }
+
+    }
+    return end();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class ..._Args>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_holder
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__construct_node(_Args&& ...__args)
+{
+    static_assert(!__is_hash_value_type<_Args...>::value,
+                  "Construct cannot be called with a hash value type");
+    __node_allocator& __na = __node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _NodeTypes::__get_ptr(__h->__value_), _CUDA_VSTD::forward<_Args>(__args)...);
+    __h.get_deleter().__value_constructed = true;
+    __h->__hash_ = hash_function()(__h->__value_);
+    __h->__next_ = nullptr;
+    return __h;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _First, class ..._Rest>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_holder
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__construct_node_hash(
+    size_t __hash, _First&& __f, _Rest&& ...__rest)
+{
+    static_assert(!__is_hash_value_type<_First, _Rest...>::value,
+                  "Construct cannot be called with a hash value type");
+    __node_allocator& __na = __node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _NodeTypes::__get_ptr(__h->__value_),
+                             _CUDA_VSTD::forward<_First>(__f),
+                             _CUDA_VSTD::forward<_Rest>(__rest)...);
+    __h.get_deleter().__value_constructed = true;
+    __h->__hash_ = __hash;
+    __h->__next_ = nullptr;
+    return __h;
+}
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_holder
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__construct_node(const __container_value_type& __v)
+{
+    __node_allocator& __na = __node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _NodeTypes::__get_ptr(__h->__value_), __v);
+    __h.get_deleter().__value_constructed = true;
+    __h->__hash_ = hash_function()(__h->__value_);
+    __h->__next_ = nullptr;
+    return _LIBCUDACXX_EXPLICIT_MOVE(__h);  // explicitly moved for C++03
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_holder
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__construct_node_hash(size_t __hash,
+                                                                const __container_value_type& __v)
+{
+    __node_allocator& __na = __node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _NodeTypes::__get_ptr(__h->__value_), __v);
+    __h.get_deleter().__value_constructed = true;
+    __h->__hash_ = __hash;
+    __h->__next_ = nullptr;
+    return _LIBCUDACXX_EXPLICIT_MOVE(__h);  // explicitly moved for C++03
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::erase(const_iterator __p)
+{
+    __next_pointer __np = __p.__node_;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "unordered container erase(iterator) called with an iterator not"
+        " referring to this container");
+    _LIBCUDACXX_ASSERT(__p != end(),
+        "unordered container erase(iterator) called with a non-dereferenceable iterator");
+    iterator __r(__np, this);
+#else
+    iterator __r(__np);
+#endif
+    ++__r;
+    remove(__p);
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::erase(const_iterator __first,
+                                                const_iterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__first) == this,
+        "unodered container::erase(iterator, iterator) called with an iterator not"
+        " referring to this unodered container");
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__last) == this,
+        "unodered container::erase(iterator, iterator) called with an iterator not"
+        " referring to this unodered container");
+#endif
+    for (const_iterator __p = __first; __first != __last; __p = __first)
+    {
+        ++__first;
+        erase(__p);
+    }
+    __next_pointer __np = __last.__node_;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator (__np, this);
+#else
+    return iterator (__np);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::size_type
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__erase_unique(const _Key& __k)
+{
+    iterator __i = find(__k);
+    if (__i == end())
+        return 0;
+    erase(__i);
+    return 1;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::size_type
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__erase_multi(const _Key& __k)
+{
+    size_type __r = 0;
+    iterator __i = find(__k);
+    if (__i != end())
+    {
+        iterator __e = end();
+        do
+        {
+            erase(__i++);
+            ++__r;
+        } while (__i != __e && key_eq()(*__i, __k));
+    }
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::__node_holder
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::remove(const_iterator __p) _NOEXCEPT
+{
+    // current node
+    __next_pointer __cn = __p.__node_;
+    size_type __bc = bucket_count();
+    size_t __chash = __constrain_hash(__cn->__hash(), __bc);
+    // find previous node
+    __next_pointer __pn = __bucket_list_[__chash];
+    for (; __pn->__next_ != __cn; __pn = __pn->__next_)
+        ;
+    // Fix up __bucket_list_
+        // if __pn is not in same bucket (before begin is not in same bucket) &&
+        //    if __cn->__next_ is not in same bucket (nullptr is not in same bucket)
+    if (__pn == __p1_.first().__ptr()
+            || __constrain_hash(__pn->__hash(), __bc) != __chash)
+    {
+        if (__cn->__next_ == nullptr
+            || __constrain_hash(__cn->__next_->__hash(), __bc) != __chash)
+            __bucket_list_[__chash] = nullptr;
+    }
+        // if __cn->__next_ is not in same bucket (nullptr is in same bucket)
+    if (__cn->__next_ != nullptr)
+    {
+        size_t __nhash = __constrain_hash(__cn->__next_->__hash(), __bc);
+        if (__nhash != __chash)
+            __bucket_list_[__nhash] = __pn;
+    }
+    // remove __cn
+    __pn->__next_ = __cn->__next_;
+    __cn->__next_ = nullptr;
+    --size();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __c_node* __c = __get_db()->__find_c_and_lock(this);
+    for (__i_node** __dp = __c->end_; __dp != __c->beg_; )
+    {
+        --__dp;
+        iterator* __i = static_cast<iterator*>((*__dp)->__i_);
+        if (__i->__node_ == __cn)
+        {
+            (*__dp)->__c_ = nullptr;
+            if (--__c->end_ != __dp)
+                memmove(__dp, __dp+1, (__c->end_ - __dp)*sizeof(__i_node*));
+        }
+    }
+    __get_db()->unlock();
+#endif
+    return __node_holder(__cn->__upcast(), _Dp(__node_alloc(), true));
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+inline
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::size_type
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__count_unique(const _Key& __k) const
+{
+    return static_cast<size_type>(find(__k) != end());
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::size_type
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__count_multi(const _Key& __k) const
+{
+    size_type __r = 0;
+    const_iterator __i = find(__k);
+    if (__i != end())
+    {
+        const_iterator __e = end();
+        do
+        {
+            ++__i;
+            ++__r;
+        } while (__i != __e && key_eq()(*__i, __k));
+    }
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator,
+     typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__equal_range_unique(
+        const _Key& __k)
+{
+    iterator __i = find(__k);
+    iterator __j = __i;
+    if (__i != end())
+        ++__j;
+    return pair<iterator, iterator>(__i, __j);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator,
+     typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__equal_range_unique(
+        const _Key& __k) const
+{
+    const_iterator __i = find(__k);
+    const_iterator __j = __i;
+    if (__i != end())
+        ++__j;
+    return pair<const_iterator, const_iterator>(__i, __j);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator,
+     typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::iterator>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__equal_range_multi(
+        const _Key& __k)
+{
+    iterator __i = find(__k);
+    iterator __j = __i;
+    if (__i != end())
+    {
+        iterator __e = end();
+        do
+        {
+            ++__j;
+        } while (__j != __e && key_eq()(*__j, __k));
+    }
+    return pair<iterator, iterator>(__i, __j);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+template <class _Key>
+pair<typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator,
+     typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::const_iterator>
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__equal_range_multi(
+        const _Key& __k) const
+{
+    const_iterator __i = find(__k);
+    const_iterator __j = __i;
+    if (__i != end())
+    {
+        const_iterator __e = end();
+        do
+        {
+            ++__j;
+        } while (__j != __e && key_eq()(*__j, __k));
+    }
+    return pair<const_iterator, const_iterator>(__i, __j);
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+void
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::swap(__hash_table& __u)
+#if _LIBCUDACXX_STD_VER <= 11
+    _NOEXCEPT_(
+        __is_nothrow_swappable<hasher>::value && __is_nothrow_swappable<key_equal>::value
+        && (!allocator_traits<__pointer_allocator>::propagate_on_container_swap::value
+              || __is_nothrow_swappable<__pointer_allocator>::value)
+        && (!__node_traits::propagate_on_container_swap::value
+              || __is_nothrow_swappable<__node_allocator>::value)
+            )
+#else
+  _NOEXCEPT_(__is_nothrow_swappable<hasher>::value && __is_nothrow_swappable<key_equal>::value)
+#endif
+{
+    _LIBCUDACXX_ASSERT(__node_traits::propagate_on_container_swap::value ||
+                   this->__node_alloc() == __u.__node_alloc(),
+                   "list::swap: Either propagate_on_container_swap must be true"
+                   " or the allocators must compare equal");
+    {
+    __node_pointer_pointer __npp = __bucket_list_.release();
+    __bucket_list_.reset(__u.__bucket_list_.release());
+    __u.__bucket_list_.reset(__npp);
+    }
+    _CUDA_VSTD::swap(__bucket_list_.get_deleter().size(), __u.__bucket_list_.get_deleter().size());
+    __swap_allocator(__bucket_list_.get_deleter().__alloc(),
+             __u.__bucket_list_.get_deleter().__alloc());
+    __swap_allocator(__node_alloc(), __u.__node_alloc());
+    _CUDA_VSTD::swap(__p1_.first().__next_, __u.__p1_.first().__next_);
+    __p2_.swap(__u.__p2_);
+    __p3_.swap(__u.__p3_);
+    if (size() > 0)
+        __bucket_list_[__constrain_hash(__p1_.first().__next_->__hash(), bucket_count())] =
+            __p1_.first().__ptr();
+    if (__u.size() > 0)
+        __u.__bucket_list_[__constrain_hash(__u.__p1_.first().__next_->__hash(), __u.bucket_count())] =
+            __u.__p1_.first().__ptr();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+typename __hash_table<_Tp, _Hash, _Equal, _Alloc>::size_type
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::bucket_size(size_type __n) const
+{
+    _LIBCUDACXX_ASSERT(__n < bucket_count(),
+        "unordered container::bucket_size(n) called with n >= bucket_count()");
+    __next_pointer __np = __bucket_list_[__n];
+    size_type __bc = bucket_count();
+    size_type __r = 0;
+    if (__np != nullptr)
+    {
+        for (__np = __np->__next_; __np != nullptr &&
+                                   __constrain_hash(__np->__hash(), __bc) == __n;
+                                                    __np = __np->__next_, ++__r)
+            ;
+    }
+    return __r;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__hash_table<_Tp, _Hash, _Equal, _Alloc>& __x,
+     __hash_table<_Tp, _Hash, _Equal, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+bool
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__dereferenceable(const const_iterator* __i) const
+{
+    return __i->__node_ != nullptr;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+bool
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__decrementable(const const_iterator*) const
+{
+    return false;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+bool
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__addable(const const_iterator*, ptrdiff_t) const
+{
+    return false;
+}
+
+template <class _Tp, class _Hash, class _Equal, class _Alloc>
+bool
+__hash_table<_Tp, _Hash, _Equal, _Alloc>::__subscriptable(const const_iterator*, ptrdiff_t) const
+{
+    return false;
+}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX__HASH_TABLE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__libcpp_version b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__libcpp_version
new file mode 100644
index 000000000000..5caff40c4a0c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__libcpp_version
@@ -0,0 +1 @@
+10000
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__locale b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__locale
new file mode 100644
index 000000000000..2b2df18624ef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__locale
@@ -0,0 +1,1553 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___LOCALE
+#define _LIBCUDACXX___LOCALE
+
+#include <__config>
+#include <string>
+#include <memory>
+#include <utility>
+#include <mutex>
+#include <cstdint>
+#include <cctype>
+#include <locale.h>
+#if defined(_LIBCUDACXX_MSVCRT_LIKE)
+# include <cstring>
+# include <support/win32/locale_win32.h>
+#elif defined(_AIX)
+# include <support/ibm/xlocale.h>
+#elif defined(__ANDROID__)
+# include <support/android/locale_bionic.h>
+#elif defined(__sun__)
+# include <xlocale.h>
+# include <support/solaris/xlocale.h>
+#elif defined(_NEWLIB_VERSION)
+# include <support/newlib/xlocale.h>
+#elif (defined(__APPLE__)      || defined(__FreeBSD__) \
+    || defined(__EMSCRIPTEN__) || defined(__IBMCPP__))
+# include <xlocale.h>
+#elif defined(__Fuchsia__)
+# include <support/fuchsia/xlocale.h>
+#elif defined(__wasi__)
+// WASI libc uses musl's locales support.
+# include <support/musl/xlocale.h>
+#elif defined(_LIBCUDACXX_HAS_MUSL_LIBC)
+# include <support/musl/xlocale.h>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if !defined(_LIBCUDACXX_LOCALE__L_EXTENSIONS)
+struct __libcpp_locale_guard {
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __libcpp_locale_guard(locale_t& __loc) : __old_loc_(uselocale(__loc)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  ~__libcpp_locale_guard() {
+    if (__old_loc_)
+      uselocale(__old_loc_);
+  }
+
+  locale_t __old_loc_;
+private:
+  __libcpp_locale_guard(__libcpp_locale_guard const&);
+  __libcpp_locale_guard& operator=(__libcpp_locale_guard const&);
+};
+#elif defined(_LIBCUDACXX_MSVCRT_LIKE)
+struct __libcpp_locale_guard {
+    __libcpp_locale_guard(locale_t __l) :
+        __status(_configthreadlocale(_ENABLE_PER_THREAD_LOCALE)) {
+      // Setting the locale can be expensive even when the locale given is
+      // already the current locale, so do an explicit check to see if the
+      // current locale is already the one we want.
+      const char* __lc = __setlocale(nullptr);
+      // If every category is the same, the locale string will simply be the
+      // locale name, otherwise it will be a semicolon-separated string listing
+      // each category.  In the second case, we know at least one category won't
+      // be what we want, so we only have to check the first case.
+      if (strcmp(__l.__get_locale(), __lc) != 0) {
+        __locale_all = _strdup(__lc);
+        if (__locale_all == nullptr)
+          __throw_bad_alloc();
+        __setlocale(__l.__get_locale());
+      }
+    }
+    ~__libcpp_locale_guard() {
+      // The CRT documentation doesn't explicitly say, but setlocale() does the
+      // right thing when given a semicolon-separated list of locale settings
+      // for the different categories in the same format as returned by
+      // setlocale(LC_ALL, nullptr).
+      if (__locale_all != nullptr) {
+        __setlocale(__locale_all);
+        free(__locale_all);
+      }
+      _configthreadlocale(__status);
+    }
+    static const char* __setlocale(const char* __locale) {
+      const char* __new_locale = setlocale(LC_ALL, __locale);
+      if (__new_locale == nullptr)
+        __throw_bad_alloc();
+      return __new_locale;
+    }
+    int __status;
+    char* __locale_all = nullptr;
+};
+#endif
+
+
+class _LIBCUDACXX_TYPE_VIS locale;
+
+template <class _Facet>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+has_facet(const locale&) _NOEXCEPT;
+
+template <class _Facet>
+_LIBCUDACXX_INLINE_VISIBILITY
+const _Facet&
+use_facet(const locale&);
+
+class _LIBCUDACXX_TYPE_VIS locale
+{
+public:
+    // types:
+    class _LIBCUDACXX_TYPE_VIS facet;
+    class _LIBCUDACXX_TYPE_VIS id;
+
+    typedef int category;
+    _LIBCUDACXX_AVAILABILITY_LOCALE_CATEGORY
+    static const category // values assigned here are for exposition only
+        none     = 0,
+        collate  = LC_COLLATE_MASK,
+        ctype    = LC_CTYPE_MASK,
+        monetary = LC_MONETARY_MASK,
+        numeric  = LC_NUMERIC_MASK,
+        time     = LC_TIME_MASK,
+        messages = LC_MESSAGES_MASK,
+        all = collate | ctype | monetary | numeric | time | messages;
+
+    // construct/copy/destroy:
+    locale()  _NOEXCEPT;
+    locale(const locale&)  _NOEXCEPT;
+    explicit locale(const char*);
+    explicit locale(const string&);
+    locale(const locale&, const char*, category);
+    locale(const locale&, const string&, category);
+    template <class _Facet>
+        _LIBCUDACXX_INLINE_VISIBILITY locale(const locale&, _Facet*);
+    locale(const locale&, const locale&, category);
+
+    ~locale();
+
+    const locale& operator=(const locale&)  _NOEXCEPT;
+
+    template <class _Facet>
+      _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+      locale combine(const locale&) const;
+
+    // locale operations:
+    string name() const;
+    bool operator==(const locale&) const;
+    bool operator!=(const locale& __y) const {return !(*this == __y);}
+    template <class _CharT, class _Traits, class _Allocator>
+      _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+      bool operator()(const basic_string<_CharT, _Traits, _Allocator>&,
+                      const basic_string<_CharT, _Traits, _Allocator>&) const;
+
+    // global locale objects:
+    static locale global(const locale&);
+    static const locale& classic();
+
+private:
+    class __imp;
+    __imp* __locale_;
+
+    void __install_ctor(const locale&, facet*, long);
+    static locale& __global();
+    bool has_facet(id&) const;
+    const facet* use_facet(id&) const;
+
+    template <class _Facet> friend bool has_facet(const locale&)  _NOEXCEPT;
+    template <class _Facet> friend const _Facet& use_facet(const locale&);
+};
+
+class _LIBCUDACXX_TYPE_VIS locale::facet
+    : public __shared_count
+{
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit facet(size_t __refs = 0)
+        : __shared_count(static_cast<long>(__refs)-1) {}
+
+    virtual ~facet();
+
+//    facet(const facet&) = delete;     // effectively done in __shared_count
+//    void operator=(const facet&) = delete;
+private:
+    virtual void __on_zero_shared() _NOEXCEPT;
+};
+
+class _LIBCUDACXX_TYPE_VIS locale::id
+{
+    once_flag      __flag_;
+    int32_t        __id_;
+
+    static int32_t __next_id;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR id() :__id_(0) {}
+private:
+    void __init();
+    void operator=(const id&); // = delete;
+    id(const id&); // = delete;
+public:  // only needed for tests
+    long __get();
+
+    friend class locale;
+    friend class locale::__imp;
+};
+
+template <class _Facet>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+locale::locale(const locale& __other, _Facet* __f)
+{
+    __install_ctor(__other, __f, __f ? __f->id.__get() : 0);
+}
+
+template <class _Facet>
+locale
+locale::combine(const locale& __other) const
+{
+    if (!_CUDA_VSTD::has_facet<_Facet>(__other))
+        __throw_runtime_error("locale::combine: locale missing facet");
+
+    return locale(*this, &const_cast<_Facet&>(_CUDA_VSTD::use_facet<_Facet>(__other)));
+}
+
+template <class _Facet>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+has_facet(const locale& __l)  _NOEXCEPT
+{
+    return __l.has_facet(_Facet::id);
+}
+
+template <class _Facet>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+const _Facet&
+use_facet(const locale& __l)
+{
+    return static_cast<const _Facet&>(*__l.use_facet(_Facet::id));
+}
+
+// template <class _CharT> class collate;
+
+template <class _CharT>
+class _LIBCUDACXX_TEMPLATE_VIS collate
+    : public locale::facet
+{
+public:
+    typedef _CharT char_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit collate(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(const char_type* __lo1, const char_type* __hi1,
+                const char_type* __lo2, const char_type* __hi2) const
+    {
+        return do_compare(__lo1, __hi1, __lo2, __hi2);
+    }
+
+    // FIXME(EricWF): The _LIBCUDACXX_ALWAYS_INLINE is needed on Windows to work
+    // around a dllimport bug that expects an external instantiation.
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_ALWAYS_INLINE
+    string_type transform(const char_type* __lo, const char_type* __hi) const
+    {
+        return do_transform(__lo, __hi);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    long hash(const char_type* __lo, const char_type* __hi) const
+    {
+        return do_hash(__lo, __hi);
+    }
+
+    static locale::id id;
+
+protected:
+    ~collate();
+    virtual int do_compare(const char_type* __lo1, const char_type* __hi1,
+                           const char_type* __lo2, const char_type* __hi2) const;
+    virtual string_type do_transform(const char_type* __lo, const char_type* __hi) const
+        {return string_type(__lo, __hi);}
+    virtual long do_hash(const char_type* __lo, const char_type* __hi) const;
+};
+
+template <class _CharT> locale::id collate<_CharT>::id;
+
+template <class _CharT>
+collate<_CharT>::~collate()
+{
+}
+
+template <class _CharT>
+int
+collate<_CharT>::do_compare(const char_type* __lo1, const char_type* __hi1,
+                            const char_type* __lo2, const char_type* __hi2) const
+{
+    for (; __lo2 != __hi2; ++__lo1, ++__lo2)
+    {
+        if (__lo1 == __hi1 || *__lo1 < *__lo2)
+            return -1;
+        if (*__lo2 < *__lo1)
+            return 1;
+    }
+    return __lo1 != __hi1;
+}
+
+template <class _CharT>
+long
+collate<_CharT>::do_hash(const char_type* __lo, const char_type* __hi) const
+{
+    size_t __h = 0;
+    const size_t __sr = __CHAR_BIT__ * sizeof(size_t) - 8;
+    const size_t __mask = size_t(0xF) << (__sr + 4);
+    for(const char_type* __p = __lo; __p != __hi; ++__p)
+    {
+        __h = (__h << 4) + static_cast<size_t>(*__p);
+        size_t __g = __h & __mask;
+        __h ^= __g | (__g >> __sr);
+    }
+    return static_cast<long>(__h);
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS collate<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS collate<wchar_t>)
+
+// template <class CharT> class collate_byname;
+
+template <class _CharT> class _LIBCUDACXX_TEMPLATE_VIS collate_byname;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS collate_byname<char>
+    : public collate<char>
+{
+    locale_t __l;
+public:
+    typedef char char_type;
+    typedef basic_string<char_type> string_type;
+
+    explicit collate_byname(const char* __n, size_t __refs = 0);
+    explicit collate_byname(const string& __n, size_t __refs = 0);
+
+protected:
+    ~collate_byname();
+    virtual int do_compare(const char_type* __lo1, const char_type* __hi1,
+                           const char_type* __lo2, const char_type* __hi2) const;
+    virtual string_type do_transform(const char_type* __lo, const char_type* __hi) const;
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS collate_byname<wchar_t>
+    : public collate<wchar_t>
+{
+    locale_t __l;
+public:
+    typedef wchar_t char_type;
+    typedef basic_string<char_type> string_type;
+
+    explicit collate_byname(const char* __n, size_t __refs = 0);
+    explicit collate_byname(const string& __n, size_t __refs = 0);
+
+protected:
+    ~collate_byname();
+
+    virtual int do_compare(const char_type* __lo1, const char_type* __hi1,
+                           const char_type* __lo2, const char_type* __hi2) const;
+    virtual string_type do_transform(const char_type* __lo, const char_type* __hi) const;
+};
+
+template <class _CharT, class _Traits, class _Allocator>
+bool
+locale::operator()(const basic_string<_CharT, _Traits, _Allocator>& __x,
+                   const basic_string<_CharT, _Traits, _Allocator>& __y) const
+{
+    return _CUDA_VSTD::use_facet<_CUDA_VSTD::collate<_CharT> >(*this).compare(
+                                       __x.data(), __x.data() + __x.size(),
+                                       __y.data(), __y.data() + __y.size()) < 0;
+}
+
+// template <class charT> class ctype
+
+class _LIBCUDACXX_TYPE_VIS ctype_base
+{
+public:
+#if defined(__GLIBC__)
+    typedef unsigned short mask;
+    static const mask space  = _ISspace;
+    static const mask print  = _ISprint;
+    static const mask cntrl  = _IScntrl;
+    static const mask upper  = _ISupper;
+    static const mask lower  = _ISlower;
+    static const mask alpha  = _ISalpha;
+    static const mask digit  = _ISdigit;
+    static const mask punct  = _ISpunct;
+    static const mask xdigit = _ISxdigit;
+    static const mask blank  = _ISblank;
+#if defined(__mips__)
+    static const mask __regex_word = static_cast<mask>(_ISbit(15));
+#else
+    static const mask __regex_word = 0x80;
+#endif
+#elif defined(_LIBCUDACXX_MSVCRT_LIKE)
+    typedef unsigned short mask;
+    static const mask space  = _SPACE;
+    static const mask print  = _BLANK|_PUNCT|_ALPHA|_DIGIT;
+    static const mask cntrl  = _CONTROL;
+    static const mask upper  = _UPPER;
+    static const mask lower  = _LOWER;
+    static const mask alpha  = _ALPHA;
+    static const mask digit  = _DIGIT;
+    static const mask punct  = _PUNCT;
+    static const mask xdigit = _HEX;
+    static const mask blank  = _BLANK;
+    static const mask __regex_word = 0x80;
+# define _LIBCUDACXX_CTYPE_MASK_IS_COMPOSITE_PRINT
+#elif defined(__APPLE__) || defined(__FreeBSD__) || defined(__EMSCRIPTEN__) || defined(__NetBSD__)
+# ifdef __APPLE__
+    typedef __uint32_t mask;
+# elif defined(__FreeBSD__)
+    typedef unsigned long mask;
+# elif defined(__EMSCRIPTEN__) || defined(__NetBSD__)
+    typedef unsigned short mask;
+# endif
+    static const mask space  = _CTYPE_S;
+    static const mask print  = _CTYPE_R;
+    static const mask cntrl  = _CTYPE_C;
+    static const mask upper  = _CTYPE_U;
+    static const mask lower  = _CTYPE_L;
+    static const mask alpha  = _CTYPE_A;
+    static const mask digit  = _CTYPE_D;
+    static const mask punct  = _CTYPE_P;
+    static const mask xdigit = _CTYPE_X;
+
+# if defined(__NetBSD__)
+    static const mask blank  = _CTYPE_BL;
+    // NetBSD defines classes up to 0x2000
+    // see sys/ctype_bits.h, _CTYPE_Q
+    static const mask __regex_word = 0x8000;
+# else
+    static const mask blank  = _CTYPE_B;
+    static const mask __regex_word = 0x80;
+# endif
+#elif defined(__sun__) || defined(_AIX)
+    typedef unsigned int mask;
+    static const mask space  = _ISSPACE;
+    static const mask print  = _ISPRINT;
+    static const mask cntrl  = _ISCNTRL;
+    static const mask upper  = _ISUPPER;
+    static const mask lower  = _ISLOWER;
+    static const mask alpha  = _ISALPHA;
+    static const mask digit  = _ISDIGIT;
+    static const mask punct  = _ISPUNCT;
+    static const mask xdigit = _ISXDIGIT;
+    static const mask blank  = _ISBLANK;
+    static const mask __regex_word = 0x80;
+#elif defined(_NEWLIB_VERSION)
+    // Same type as Newlib's _ctype_ array in newlib/libc/include/ctype.h.
+    typedef char mask;
+    static const mask space  = _S;
+    static const mask print  = _P | _U | _L | _N | _B;
+    static const mask cntrl  = _C;
+    static const mask upper  = _U;
+    static const mask lower  = _L;
+    static const mask alpha  = _U | _L;
+    static const mask digit  = _N;
+    static const mask punct  = _P;
+    static const mask xdigit = _X | _N;
+    static const mask blank  = _B;
+    static const mask __regex_word = 0x80;
+# define _LIBCUDACXX_CTYPE_MASK_IS_COMPOSITE_PRINT
+# define _LIBCUDACXX_CTYPE_MASK_IS_COMPOSITE_ALPHA
+# define _LIBCUDACXX_CTYPE_MASK_IS_COMPOSITE_XDIGIT
+#else
+    typedef unsigned long mask;
+    static const mask space  = 1<<0;
+    static const mask print  = 1<<1;
+    static const mask cntrl  = 1<<2;
+    static const mask upper  = 1<<3;
+    static const mask lower  = 1<<4;
+    static const mask alpha  = 1<<5;
+    static const mask digit  = 1<<6;
+    static const mask punct  = 1<<7;
+    static const mask xdigit = 1<<8;
+    static const mask blank  = 1<<9;
+    static const mask __regex_word = 1<<10;
+#endif
+    static const mask alnum  = alpha | digit;
+    static const mask graph  = alnum | punct;
+
+    _LIBCUDACXX_INLINE_VISIBILITY ctype_base() {}
+};
+
+template <class _CharT> class _LIBCUDACXX_TEMPLATE_VIS ctype;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS ctype<wchar_t>
+    : public locale::facet,
+      public ctype_base
+{
+public:
+    typedef wchar_t char_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit ctype(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is(mask __m, char_type __c) const
+    {
+        return do_is(__m, __c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* is(const char_type* __low, const char_type* __high, mask* __vec) const
+    {
+        return do_is(__low, __high, __vec);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* scan_is(mask __m, const char_type* __low, const char_type* __high) const
+    {
+        return do_scan_is(__m, __low, __high);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* scan_not(mask __m, const char_type* __low, const char_type* __high) const
+    {
+        return do_scan_not(__m, __low, __high);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type toupper(char_type __c) const
+    {
+        return do_toupper(__c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* toupper(char_type* __low, const char_type* __high) const
+    {
+        return do_toupper(__low, __high);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type tolower(char_type __c) const
+    {
+        return do_tolower(__c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* tolower(char_type* __low, const char_type* __high) const
+    {
+        return do_tolower(__low, __high);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type widen(char __c) const
+    {
+        return do_widen(__c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char* widen(const char* __low, const char* __high, char_type* __to) const
+    {
+        return do_widen(__low, __high, __to);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char narrow(char_type __c, char __dfault) const
+    {
+        return do_narrow(__c, __dfault);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* narrow(const char_type* __low, const char_type* __high, char __dfault, char* __to) const
+    {
+        return do_narrow(__low, __high, __dfault, __to);
+    }
+
+    static locale::id id;
+
+protected:
+    ~ctype();
+    virtual bool do_is(mask __m, char_type __c) const;
+    virtual const char_type* do_is(const char_type* __low, const char_type* __high, mask* __vec) const;
+    virtual const char_type* do_scan_is(mask __m, const char_type* __low, const char_type* __high) const;
+    virtual const char_type* do_scan_not(mask __m, const char_type* __low, const char_type* __high) const;
+    virtual char_type do_toupper(char_type) const;
+    virtual const char_type* do_toupper(char_type* __low, const char_type* __high) const;
+    virtual char_type do_tolower(char_type) const;
+    virtual const char_type* do_tolower(char_type* __low, const char_type* __high) const;
+    virtual char_type do_widen(char) const;
+    virtual const char* do_widen(const char* __low, const char* __high, char_type* __dest) const;
+    virtual char do_narrow(char_type, char __dfault) const;
+    virtual const char_type* do_narrow(const char_type* __low, const char_type* __high, char __dfault, char* __dest) const;
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS ctype<char>
+    : public locale::facet, public ctype_base
+{
+    const mask* __tab_;
+    bool        __del_;
+public:
+    typedef char char_type;
+
+    explicit ctype(const mask* __tab = 0, bool __del = false, size_t __refs = 0);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is(mask __m, char_type __c) const
+    {
+        return isascii(__c) ? (__tab_[static_cast<int>(__c)] & __m) !=0 : false;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* is(const char_type* __low, const char_type* __high, mask* __vec) const
+    {
+        for (; __low != __high; ++__low, ++__vec)
+            *__vec = isascii(*__low) ? __tab_[static_cast<int>(*__low)] : 0;
+        return __low;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* scan_is (mask __m, const char_type* __low, const char_type* __high) const
+    {
+        for (; __low != __high; ++__low)
+            if (isascii(*__low) && (__tab_[static_cast<int>(*__low)] & __m))
+                break;
+        return __low;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* scan_not(mask __m, const char_type* __low, const char_type* __high) const
+    {
+        for (; __low != __high; ++__low)
+            if (!(isascii(*__low) && (__tab_[static_cast<int>(*__low)] & __m)))
+                break;
+        return __low;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type toupper(char_type __c) const
+    {
+        return do_toupper(__c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* toupper(char_type* __low, const char_type* __high) const
+    {
+        return do_toupper(__low, __high);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type tolower(char_type __c) const
+    {
+        return do_tolower(__c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char_type* tolower(char_type* __low, const char_type* __high) const
+    {
+        return do_tolower(__low, __high);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type widen(char __c) const
+    {
+        return do_widen(__c);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char* widen(const char* __low, const char* __high, char_type* __to) const
+    {
+        return do_widen(__low, __high, __to);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char narrow(char_type __c, char __dfault) const
+    {
+        return do_narrow(__c, __dfault);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char* narrow(const char_type* __low, const char_type* __high, char __dfault, char* __to) const
+    {
+        return do_narrow(__low, __high, __dfault, __to);
+    }
+
+    static locale::id id;
+
+#ifdef _CACHED_RUNES
+    static const size_t table_size = _CACHED_RUNES;
+#else
+    static const size_t table_size = 256;  // FIXME: Don't hardcode this.
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY const mask* table() const  _NOEXCEPT {return __tab_;}
+    static const mask* classic_table()  _NOEXCEPT;
+#if defined(__GLIBC__) || defined(__EMSCRIPTEN__)
+    static const int* __classic_upper_table() _NOEXCEPT;
+    static const int* __classic_lower_table() _NOEXCEPT;
+#endif
+#if defined(__NetBSD__)
+    static const short* __classic_upper_table() _NOEXCEPT;
+    static const short* __classic_lower_table() _NOEXCEPT;
+#endif
+
+protected:
+    ~ctype();
+    virtual char_type do_toupper(char_type __c) const;
+    virtual const char_type* do_toupper(char_type* __low, const char_type* __high) const;
+    virtual char_type do_tolower(char_type __c) const;
+    virtual const char_type* do_tolower(char_type* __low, const char_type* __high) const;
+    virtual char_type do_widen(char __c) const;
+    virtual const char* do_widen(const char* __low, const char* __high, char_type* __to) const;
+    virtual char do_narrow(char_type __c, char __dfault) const;
+    virtual const char* do_narrow(const char_type* __low, const char_type* __high, char __dfault, char* __to) const;
+};
+
+// template <class CharT> class ctype_byname;
+
+template <class _CharT> class _LIBCUDACXX_TEMPLATE_VIS ctype_byname;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS ctype_byname<char>
+    : public ctype<char>
+{
+    locale_t __l;
+
+public:
+    explicit ctype_byname(const char*, size_t = 0);
+    explicit ctype_byname(const string&, size_t = 0);
+
+protected:
+    ~ctype_byname();
+    virtual char_type do_toupper(char_type) const;
+    virtual const char_type* do_toupper(char_type* __low, const char_type* __high) const;
+    virtual char_type do_tolower(char_type) const;
+    virtual const char_type* do_tolower(char_type* __low, const char_type* __high) const;
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS ctype_byname<wchar_t>
+    : public ctype<wchar_t>
+{
+    locale_t __l;
+
+public:
+    explicit ctype_byname(const char*, size_t = 0);
+    explicit ctype_byname(const string&, size_t = 0);
+
+protected:
+    ~ctype_byname();
+    virtual bool do_is(mask __m, char_type __c) const;
+    virtual const char_type* do_is(const char_type* __low, const char_type* __high, mask* __vec) const;
+    virtual const char_type* do_scan_is(mask __m, const char_type* __low, const char_type* __high) const;
+    virtual const char_type* do_scan_not(mask __m, const char_type* __low, const char_type* __high) const;
+    virtual char_type do_toupper(char_type) const;
+    virtual const char_type* do_toupper(char_type* __low, const char_type* __high) const;
+    virtual char_type do_tolower(char_type) const;
+    virtual const char_type* do_tolower(char_type* __low, const char_type* __high) const;
+    virtual char_type do_widen(char) const;
+    virtual const char* do_widen(const char* __low, const char* __high, char_type* __dest) const;
+    virtual char do_narrow(char_type, char __dfault) const;
+    virtual const char_type* do_narrow(const char_type* __low, const char_type* __high, char __dfault, char* __dest) const;
+};
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isspace(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::space, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isprint(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::print, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+iscntrl(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::cntrl, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isupper(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::upper, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+islower(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::lower, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isalpha(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::alpha, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isdigit(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::digit, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+ispunct(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::punct, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isxdigit(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::xdigit, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isalnum(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::alnum, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isgraph(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).is(ctype_base::graph, __c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_CharT
+toupper(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).toupper(__c);
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_CharT
+tolower(_CharT __c, const locale& __loc)
+{
+    return use_facet<ctype<_CharT> >(__loc).tolower(__c);
+}
+
+// codecvt_base
+
+class _LIBCUDACXX_TYPE_VIS codecvt_base
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY codecvt_base() {}
+    enum result {ok, partial, error, noconv};
+};
+
+// template <class internT, class externT, class stateT> class codecvt;
+
+template <class _InternT, class _ExternT, class _StateT> class _LIBCUDACXX_TEMPLATE_VIS codecvt;
+
+// template <> class codecvt<char, char, mbstate_t>
+
+template <>
+class _LIBCUDACXX_TYPE_VIS codecvt<char, char, mbstate_t>
+    : public locale::facet,
+      public codecvt_base
+{
+public:
+    typedef char      intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_out(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_unshift(__st, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const
+    {
+        return do_in(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int encoding() const  _NOEXCEPT
+    {
+        return do_encoding();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool always_noconv() const  _NOEXCEPT
+    {
+        return do_always_noconv();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int length(state_type& __st, const extern_type* __frm, const extern_type* __end, size_t __mx) const
+    {
+        return do_length(__st, __frm, __end, __mx);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int max_length() const  _NOEXCEPT
+    {
+        return do_max_length();
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt(const char*, size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    ~codecvt();
+
+    virtual result do_out(state_type& __st,
+                          const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+                          extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result do_in(state_type& __st,
+                         const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+                         intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result do_unshift(state_type& __st,
+                              extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const  _NOEXCEPT;
+    virtual bool do_always_noconv() const  _NOEXCEPT;
+    virtual int do_length(state_type& __st, const extern_type* __frm, const extern_type* __end, size_t __mx) const;
+    virtual int do_max_length() const  _NOEXCEPT;
+};
+
+// template <> class codecvt<wchar_t, char, mbstate_t>
+
+template <>
+class _LIBCUDACXX_TYPE_VIS codecvt<wchar_t, char, mbstate_t>
+    : public locale::facet,
+      public codecvt_base
+{
+    locale_t __l;
+public:
+    typedef wchar_t   intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    explicit codecvt(size_t __refs = 0);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_out(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_unshift(__st, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const
+    {
+        return do_in(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int encoding() const  _NOEXCEPT
+    {
+        return do_encoding();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool always_noconv() const  _NOEXCEPT
+    {
+        return do_always_noconv();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int length(state_type& __st, const extern_type* __frm, const extern_type* __end, size_t __mx) const
+    {
+        return do_length(__st, __frm, __end, __mx);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int max_length() const  _NOEXCEPT
+    {
+        return do_max_length();
+    }
+
+    static locale::id id;
+
+protected:
+    explicit codecvt(const char*, size_t __refs = 0);
+
+    ~codecvt();
+
+    virtual result do_out(state_type& __st,
+                          const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+                          extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result do_in(state_type& __st,
+                         const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+                         intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result do_unshift(state_type& __st,
+                              extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const  _NOEXCEPT;
+    virtual bool do_always_noconv() const  _NOEXCEPT;
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end, size_t __mx) const;
+    virtual int do_max_length() const  _NOEXCEPT;
+};
+
+// template <> class codecvt<char16_t, char, mbstate_t>
+
+template <>
+class _LIBCUDACXX_TYPE_VIS codecvt<char16_t, char, mbstate_t>
+    : public locale::facet,
+      public codecvt_base
+{
+public:
+    typedef char16_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_out(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_unshift(__st, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const
+    {
+        return do_in(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int encoding() const  _NOEXCEPT
+    {
+        return do_encoding();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool always_noconv() const  _NOEXCEPT
+    {
+        return do_always_noconv();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int length(state_type& __st, const extern_type* __frm, const extern_type* __end, size_t __mx) const
+    {
+        return do_length(__st, __frm, __end, __mx);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int max_length() const  _NOEXCEPT
+    {
+        return do_max_length();
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt(const char*, size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    ~codecvt();
+
+    virtual result do_out(state_type& __st,
+                          const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+                          extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result do_in(state_type& __st,
+                         const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+                         intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result do_unshift(state_type& __st,
+                              extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const  _NOEXCEPT;
+    virtual bool do_always_noconv() const  _NOEXCEPT;
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end, size_t __mx) const;
+    virtual int do_max_length() const  _NOEXCEPT;
+};
+
+// template <> class codecvt<char32_t, char, mbstate_t>
+
+template <>
+class _LIBCUDACXX_TYPE_VIS codecvt<char32_t, char, mbstate_t>
+    : public locale::facet,
+      public codecvt_base
+{
+public:
+    typedef char32_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_out(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const
+    {
+        return do_unshift(__st, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const
+    {
+        return do_in(__st, __frm, __frm_end, __frm_nxt, __to, __to_end, __to_nxt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int encoding() const  _NOEXCEPT
+    {
+        return do_encoding();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool always_noconv() const  _NOEXCEPT
+    {
+        return do_always_noconv();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int length(state_type& __st, const extern_type* __frm, const extern_type* __end, size_t __mx) const
+    {
+        return do_length(__st, __frm, __end, __mx);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int max_length() const  _NOEXCEPT
+    {
+        return do_max_length();
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt(const char*, size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    ~codecvt();
+
+    virtual result do_out(state_type& __st,
+                          const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+                          extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result do_in(state_type& __st,
+                         const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+                         intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result do_unshift(state_type& __st,
+                              extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const  _NOEXCEPT;
+    virtual bool do_always_noconv() const  _NOEXCEPT;
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end, size_t __mx) const;
+    virtual int do_max_length() const  _NOEXCEPT;
+};
+
+// template <class _InternT, class _ExternT, class _StateT> class codecvt_byname
+
+template <class _InternT, class _ExternT, class _StateT>
+class _LIBCUDACXX_TEMPLATE_VIS codecvt_byname
+    : public codecvt<_InternT, _ExternT, _StateT>
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt_byname(const char* __nm, size_t __refs = 0)
+        : codecvt<_InternT, _ExternT, _StateT>(__nm, __refs) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt_byname(const string& __nm, size_t __refs = 0)
+        : codecvt<_InternT, _ExternT, _StateT>(__nm.c_str(), __refs) {}
+protected:
+    ~codecvt_byname();
+};
+
+template <class _InternT, class _ExternT, class _StateT>
+codecvt_byname<_InternT, _ExternT, _StateT>::~codecvt_byname()
+{
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char, char, mbstate_t>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<wchar_t, char, mbstate_t>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char16_t, char, mbstate_t>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char32_t, char, mbstate_t>)
+
+template <size_t _Np>
+struct __narrow_to_utf8
+{
+    template <class _OutputIterator, class _CharT>
+    _OutputIterator
+    operator()(_OutputIterator __s, const _CharT* __wb, const _CharT* __we) const;
+};
+
+template <>
+struct __narrow_to_utf8<8>
+{
+    template <class _OutputIterator, class _CharT>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _OutputIterator
+    operator()(_OutputIterator __s, const _CharT* __wb, const _CharT* __we) const
+    {
+        for (; __wb < __we; ++__wb, ++__s)
+            *__s = *__wb;
+        return __s;
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS __narrow_to_utf8<16>
+    : public codecvt<char16_t, char, mbstate_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __narrow_to_utf8() : codecvt<char16_t, char, mbstate_t>(1) {}
+
+    _LIBCUDACXX_EXPORTED_FROM_ABI ~__narrow_to_utf8();
+
+    template <class _OutputIterator, class _CharT>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _OutputIterator
+    operator()(_OutputIterator __s, const _CharT* __wb, const _CharT* __we) const
+    {
+        result __r = ok;
+        mbstate_t __mb;
+        while (__wb < __we && __r != error)
+        {
+            const int __sz = 32;
+            char __buf[__sz];
+            char* __bn;
+            const char16_t* __wn = (const char16_t*)__wb;
+            __r = do_out(__mb, (const char16_t*)__wb, (const char16_t*)__we, __wn,
+                         __buf, __buf+__sz, __bn);
+            if (__r == codecvt_base::error || __wn == (const char16_t*)__wb)
+                __throw_runtime_error("locale not supported");
+            for (const char* __p = __buf; __p < __bn; ++__p, ++__s)
+                *__s = *__p;
+            __wb = (const _CharT*)__wn;
+        }
+        return __s;
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS __narrow_to_utf8<32>
+    : public codecvt<char32_t, char, mbstate_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __narrow_to_utf8() : codecvt<char32_t, char, mbstate_t>(1) {}
+
+    _LIBCUDACXX_EXPORTED_FROM_ABI ~__narrow_to_utf8();
+
+    template <class _OutputIterator, class _CharT>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _OutputIterator
+    operator()(_OutputIterator __s, const _CharT* __wb, const _CharT* __we) const
+    {
+        result __r = ok;
+        mbstate_t __mb;
+        while (__wb < __we && __r != error)
+        {
+            const int __sz = 32;
+            char __buf[__sz];
+            char* __bn;
+            const char32_t* __wn = (const char32_t*)__wb;
+            __r = do_out(__mb, (const char32_t*)__wb, (const char32_t*)__we, __wn,
+                         __buf, __buf+__sz, __bn);
+            if (__r == codecvt_base::error || __wn == (const char32_t*)__wb)
+                __throw_runtime_error("locale not supported");
+            for (const char* __p = __buf; __p < __bn; ++__p, ++__s)
+                *__s = *__p;
+            __wb = (const _CharT*)__wn;
+        }
+        return __s;
+    }
+};
+
+template <size_t _Np>
+struct __widen_from_utf8
+{
+    template <class _OutputIterator>
+    _OutputIterator
+    operator()(_OutputIterator __s, const char* __nb, const char* __ne) const;
+};
+
+template <>
+struct __widen_from_utf8<8>
+{
+    template <class _OutputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _OutputIterator
+    operator()(_OutputIterator __s, const char* __nb, const char* __ne) const
+    {
+        for (; __nb < __ne; ++__nb, ++__s)
+            *__s = *__nb;
+        return __s;
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS __widen_from_utf8<16>
+    : public codecvt<char16_t, char, mbstate_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __widen_from_utf8() : codecvt<char16_t, char, mbstate_t>(1) {}
+
+    _LIBCUDACXX_EXPORTED_FROM_ABI ~__widen_from_utf8();
+
+    template <class _OutputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _OutputIterator
+    operator()(_OutputIterator __s, const char* __nb, const char* __ne) const
+    {
+        result __r = ok;
+        mbstate_t __mb;
+        while (__nb < __ne && __r != error)
+        {
+            const int __sz = 32;
+            char16_t __buf[__sz];
+            char16_t* __bn;
+            const char* __nn = __nb;
+            __r = do_in(__mb, __nb, __ne - __nb > __sz ? __nb+__sz : __ne, __nn,
+                        __buf, __buf+__sz, __bn);
+            if (__r == codecvt_base::error || __nn == __nb)
+                __throw_runtime_error("locale not supported");
+            for (const char16_t* __p = __buf; __p < __bn; ++__p, ++__s)
+                *__s = (wchar_t)*__p;
+            __nb = __nn;
+        }
+        return __s;
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS __widen_from_utf8<32>
+    : public codecvt<char32_t, char, mbstate_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __widen_from_utf8() : codecvt<char32_t, char, mbstate_t>(1) {}
+
+    _LIBCUDACXX_EXPORTED_FROM_ABI ~__widen_from_utf8();
+
+    template <class _OutputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _OutputIterator
+    operator()(_OutputIterator __s, const char* __nb, const char* __ne) const
+    {
+        result __r = ok;
+        mbstate_t __mb;
+        while (__nb < __ne && __r != error)
+        {
+            const int __sz = 32;
+            char32_t __buf[__sz];
+            char32_t* __bn;
+            const char* __nn = __nb;
+            __r = do_in(__mb, __nb, __ne - __nb > __sz ? __nb+__sz : __ne, __nn,
+                        __buf, __buf+__sz, __bn);
+            if (__r == codecvt_base::error || __nn == __nb)
+                __throw_runtime_error("locale not supported");
+            for (const char32_t* __p = __buf; __p < __bn; ++__p, ++__s)
+                *__s = (wchar_t)*__p;
+            __nb = __nn;
+        }
+        return __s;
+    }
+};
+
+// template <class charT> class numpunct
+
+template <class _CharT> class _LIBCUDACXX_TEMPLATE_VIS numpunct;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS numpunct<char>
+    : public locale::facet
+{
+public:
+    typedef char char_type;
+    typedef basic_string<char_type> string_type;
+
+    explicit numpunct(size_t __refs = 0);
+
+    _LIBCUDACXX_INLINE_VISIBILITY char_type decimal_point() const {return do_decimal_point();}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type thousands_sep() const {return do_thousands_sep();}
+    _LIBCUDACXX_INLINE_VISIBILITY string grouping() const         {return do_grouping();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type truename() const    {return do_truename();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type falsename() const   {return do_falsename();}
+
+    static locale::id id;
+
+protected:
+    ~numpunct();
+    virtual char_type do_decimal_point() const;
+    virtual char_type do_thousands_sep() const;
+    virtual string do_grouping() const;
+    virtual string_type do_truename() const;
+    virtual string_type do_falsename() const;
+
+    char_type __decimal_point_;
+    char_type __thousands_sep_;
+    string __grouping_;
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS numpunct<wchar_t>
+    : public locale::facet
+{
+public:
+    typedef wchar_t char_type;
+    typedef basic_string<char_type> string_type;
+
+    explicit numpunct(size_t __refs = 0);
+
+    _LIBCUDACXX_INLINE_VISIBILITY char_type decimal_point() const {return do_decimal_point();}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type thousands_sep() const {return do_thousands_sep();}
+    _LIBCUDACXX_INLINE_VISIBILITY string grouping() const         {return do_grouping();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type truename() const    {return do_truename();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type falsename() const   {return do_falsename();}
+
+    static locale::id id;
+
+protected:
+    ~numpunct();
+    virtual char_type do_decimal_point() const;
+    virtual char_type do_thousands_sep() const;
+    virtual string do_grouping() const;
+    virtual string_type do_truename() const;
+    virtual string_type do_falsename() const;
+
+    char_type __decimal_point_;
+    char_type __thousands_sep_;
+    string __grouping_;
+};
+
+// template <class charT> class numpunct_byname
+
+template <class _CharT> class _LIBCUDACXX_TEMPLATE_VIS numpunct_byname;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS numpunct_byname<char>
+: public numpunct<char>
+{
+public:
+    typedef char char_type;
+    typedef basic_string<char_type> string_type;
+
+    explicit numpunct_byname(const char* __nm, size_t __refs = 0);
+    explicit numpunct_byname(const string& __nm, size_t __refs = 0);
+
+protected:
+    ~numpunct_byname();
+
+private:
+    void __init(const char*);
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS numpunct_byname<wchar_t>
+: public numpunct<wchar_t>
+{
+public:
+    typedef wchar_t char_type;
+    typedef basic_string<char_type> string_type;
+
+    explicit numpunct_byname(const char* __nm, size_t __refs = 0);
+    explicit numpunct_byname(const string& __nm, size_t __refs = 0);
+
+protected:
+    ~numpunct_byname();
+
+private:
+    void __init(const char*);
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX___LOCALE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base
new file mode 100644
index 000000000000..bae7c7c210c3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__mutex_base
@@ -0,0 +1,541 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___MUTEX_BASE
+#define _LIBCUDACXX___MUTEX_BASE
+
+#include <__config>
+#include <chrono>
+#include <system_error>
+#include <__threading_support>
+
+#include <time.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#ifndef _LIBCUDACXX_HAS_NO_THREADS
+
+#ifndef _LIBCUDACXX_THREAD_SAFETY_ANNOTATION
+#  ifdef _LIBCUDACXX_HAS_THREAD_SAFETY_ANNOTATIONS
+#    define _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(x) __attribute__((x))
+#  else
+#    define _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(x)
+#  endif
+#endif  // _LIBCUDACXX_THREAD_SAFETY_ANNOTATION
+
+
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(capability("mutex")) mutex
+{
+    __libcpp_mutex_t __m_ = _LIBCUDACXX_MUTEX_INITIALIZER;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR mutex() = default;
+
+    mutex(const mutex&) = delete;
+    mutex& operator=(const mutex&) = delete;
+
+#if defined(_LIBCUDACXX_HAS_TRIVIAL_MUTEX_DESTRUCTION)
+    ~mutex() = default;
+#else
+    ~mutex() _NOEXCEPT;
+#endif
+
+    void lock() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(acquire_capability());
+    bool try_lock() _NOEXCEPT _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(try_acquire_capability(true));
+    void unlock() _NOEXCEPT _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(release_capability());
+
+    typedef __libcpp_mutex_t* native_handle_type;
+    _LIBCUDACXX_INLINE_VISIBILITY native_handle_type native_handle() {return &__m_;}
+};
+
+static_assert(is_nothrow_default_constructible<mutex>::value,
+              "the default constructor for std::mutex must be nothrow");
+
+struct _LIBCUDACXX_TYPE_VIS defer_lock_t { explicit defer_lock_t() = default; };
+struct _LIBCUDACXX_TYPE_VIS try_to_lock_t { explicit try_to_lock_t() = default; };
+struct _LIBCUDACXX_TYPE_VIS adopt_lock_t { explicit adopt_lock_t() = default; };
+
+#if defined(_LIBCUDACXX_CXX03_LANG) || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+
+extern _LIBCUDACXX_EXPORTED_FROM_ABI const defer_lock_t  defer_lock;
+extern _LIBCUDACXX_EXPORTED_FROM_ABI const try_to_lock_t try_to_lock;
+extern _LIBCUDACXX_EXPORTED_FROM_ABI const adopt_lock_t  adopt_lock;
+
+#else
+
+/* _LIBCUDACXX_INLINE_VAR */ constexpr defer_lock_t  defer_lock  = defer_lock_t();
+/* _LIBCUDACXX_INLINE_VAR */ constexpr try_to_lock_t try_to_lock = try_to_lock_t();
+/* _LIBCUDACXX_INLINE_VAR */ constexpr adopt_lock_t  adopt_lock  = adopt_lock_t();
+
+#endif
+
+template <class _Mutex>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(scoped_lockable)
+lock_guard
+{
+public:
+    typedef _Mutex mutex_type;
+
+private:
+    mutex_type& __m_;
+public:
+
+    _LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+    explicit lock_guard(mutex_type& __m) _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(acquire_capability(__m))
+        : __m_(__m) {__m_.lock();}
+
+    _LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+    lock_guard(mutex_type& __m, adopt_lock_t) _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(requires_capability(__m))
+        : __m_(__m) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~lock_guard() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(release_capability()) {__m_.unlock();}
+
+private:
+    lock_guard(lock_guard const&) _LIBCUDACXX_EQUAL_DELETE;
+    lock_guard& operator=(lock_guard const&) _LIBCUDACXX_EQUAL_DELETE;
+};
+
+template <class _Mutex>
+class _LIBCUDACXX_TEMPLATE_VIS unique_lock
+{
+public:
+    typedef _Mutex mutex_type;
+
+private:
+    mutex_type* __m_;
+    bool __owns_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unique_lock() _NOEXCEPT : __m_(nullptr), __owns_(false) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit unique_lock(mutex_type& __m)
+        : __m_(_CUDA_VSTD::addressof(__m)), __owns_(true) {__m_->lock();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unique_lock(mutex_type& __m, defer_lock_t) _NOEXCEPT
+        : __m_(_CUDA_VSTD::addressof(__m)), __owns_(false) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unique_lock(mutex_type& __m, try_to_lock_t)
+        : __m_(_CUDA_VSTD::addressof(__m)), __owns_(__m.try_lock()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unique_lock(mutex_type& __m, adopt_lock_t)
+        : __m_(_CUDA_VSTD::addressof(__m)), __owns_(true) {}
+    template <class _Clock, class _Duration>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        unique_lock(mutex_type& __m, const chrono::time_point<_Clock, _Duration>& __t)
+            : __m_(_CUDA_VSTD::addressof(__m)), __owns_(__m.try_lock_until(__t)) {}
+    template <class _Rep, class _Period>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        unique_lock(mutex_type& __m, const chrono::duration<_Rep, _Period>& __d)
+            : __m_(_CUDA_VSTD::addressof(__m)), __owns_(__m.try_lock_for(__d)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~unique_lock()
+    {
+        if (__owns_)
+            __m_->unlock();
+    }
+
+private:
+    unique_lock(unique_lock const&); // = delete;
+    unique_lock& operator=(unique_lock const&); // = delete;
+
+public:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unique_lock(unique_lock&& __u) _NOEXCEPT
+        : __m_(__u.__m_), __owns_(__u.__owns_)
+        {__u.__m_ = nullptr; __u.__owns_ = false;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unique_lock& operator=(unique_lock&& __u) _NOEXCEPT
+        {
+            if (__owns_)
+                __m_->unlock();
+            __m_ = __u.__m_;
+            __owns_ = __u.__owns_;
+            __u.__m_ = nullptr;
+            __u.__owns_ = false;
+            return *this;
+        }
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    void lock();
+    bool try_lock();
+
+    template <class _Rep, class _Period>
+        bool try_lock_for(const chrono::duration<_Rep, _Period>& __d);
+    template <class _Clock, class _Duration>
+        bool try_lock_until(const chrono::time_point<_Clock, _Duration>& __t);
+
+    void unlock();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(unique_lock& __u) _NOEXCEPT
+    {
+        _CUDA_VSTD::swap(__m_, __u.__m_);
+        _CUDA_VSTD::swap(__owns_, __u.__owns_);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mutex_type* release() _NOEXCEPT
+    {
+        mutex_type* __m = __m_;
+        __m_ = nullptr;
+        __owns_ = false;
+        return __m;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool owns_lock() const _NOEXCEPT {return __owns_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_EXPLICIT
+        operator bool () const _NOEXCEPT {return __owns_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mutex_type* mutex() const _NOEXCEPT {return __m_;}
+};
+
+template <class _Mutex>
+void
+unique_lock<_Mutex>::lock()
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "unique_lock::lock: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "unique_lock::lock: already locked");
+    __m_->lock();
+    __owns_ = true;
+}
+
+template <class _Mutex>
+bool
+unique_lock<_Mutex>::try_lock()
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "unique_lock::try_lock: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "unique_lock::try_lock: already locked");
+    __owns_ = __m_->try_lock();
+    return __owns_;
+}
+
+template <class _Mutex>
+template <class _Rep, class _Period>
+bool
+unique_lock<_Mutex>::try_lock_for(const chrono::duration<_Rep, _Period>& __d)
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "unique_lock::try_lock_for: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "unique_lock::try_lock_for: already locked");
+    __owns_ = __m_->try_lock_for(__d);
+    return __owns_;
+}
+
+template <class _Mutex>
+template <class _Clock, class _Duration>
+bool
+unique_lock<_Mutex>::try_lock_until(const chrono::time_point<_Clock, _Duration>& __t)
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "unique_lock::try_lock_until: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "unique_lock::try_lock_until: already locked");
+    __owns_ = __m_->try_lock_until(__t);
+    return __owns_;
+}
+
+template <class _Mutex>
+void
+unique_lock<_Mutex>::unlock()
+{
+    if (!__owns_)
+        __throw_system_error(EPERM, "unique_lock::unlock: not locked");
+    __m_->unlock();
+    __owns_ = false;
+}
+
+template <class _Mutex>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(unique_lock<_Mutex>& __x, unique_lock<_Mutex>& __y) _NOEXCEPT
+    {__x.swap(__y);}
+
+//enum class cv_status
+_LIBCUDACXX_DECLARE_STRONG_ENUM(cv_status)
+{
+    no_timeout,
+    timeout
+};
+_LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(cv_status)
+
+class _LIBCUDACXX_TYPE_VIS condition_variable
+{
+    __libcpp_condvar_t __cv_ = _LIBCUDACXX_CONDVAR_INITIALIZER;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR condition_variable() _NOEXCEPT = default;
+
+#ifdef _LIBCUDACXX_HAS_TRIVIAL_CONDVAR_DESTRUCTION
+    ~condition_variable() = default;
+#else
+    ~condition_variable();
+#endif
+
+    condition_variable(const condition_variable&) = delete;
+    condition_variable& operator=(const condition_variable&) = delete;
+
+    void notify_one() _NOEXCEPT;
+    void notify_all() _NOEXCEPT;
+
+    void wait(unique_lock<mutex>& __lk) _NOEXCEPT;
+    template <class _Predicate>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        void wait(unique_lock<mutex>& __lk, _Predicate __pred);
+
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        cv_status
+        wait_until(unique_lock<mutex>& __lk,
+                   const chrono::time_point<_Clock, _Duration>& __t);
+
+    template <class _Clock, class _Duration, class _Predicate>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        bool
+        wait_until(unique_lock<mutex>& __lk,
+                   const chrono::time_point<_Clock, _Duration>& __t,
+                   _Predicate __pred);
+
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        cv_status
+        wait_for(unique_lock<mutex>& __lk,
+                 const chrono::duration<_Rep, _Period>& __d);
+
+    template <class _Rep, class _Period, class _Predicate>
+        bool
+        _LIBCUDACXX_INLINE_VISIBILITY
+        wait_for(unique_lock<mutex>& __lk,
+                 const chrono::duration<_Rep, _Period>& __d,
+                 _Predicate __pred);
+
+    typedef __libcpp_condvar_t* native_handle_type;
+    _LIBCUDACXX_INLINE_VISIBILITY native_handle_type native_handle() {return &__cv_;}
+
+private:
+    void __do_timed_wait(unique_lock<mutex>& __lk,
+       chrono::time_point<chrono::system_clock, chrono::nanoseconds>) _NOEXCEPT;
+#if defined(_LIBCUDACXX_HAS_COND_CLOCKWAIT)
+    void __do_timed_wait(unique_lock<mutex>& __lk,
+       chrono::time_point<chrono::steady_clock, chrono::nanoseconds>) _NOEXCEPT;
+#endif
+    template <class _Clock>
+    void __do_timed_wait(unique_lock<mutex>& __lk,
+       chrono::time_point<_Clock, chrono::nanoseconds>) _NOEXCEPT;
+};
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+template <class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_floating_point<_Rep>::value,
+    chrono::nanoseconds
+>::type
+__safe_nanosecond_cast(chrono::duration<_Rep, _Period> __d)
+{
+    using namespace chrono;
+    using __ratio = ratio_divide<_Period, nano>;
+    using __ns_rep = nanoseconds::rep;
+    _Rep __result_float = __d.count() * __ratio::num / __ratio::den;
+
+    _Rep __result_max = numeric_limits<__ns_rep>::max();
+    if (__result_float >= __result_max) {
+        return nanoseconds::max();
+    }
+
+    _Rep __result_min = numeric_limits<__ns_rep>::min();
+    if (__result_float <= __result_min) {
+        return nanoseconds::min();
+    }
+
+    return nanoseconds(static_cast<__ns_rep>(__result_float));
+}
+
+template <class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_floating_point<_Rep>::value,
+    chrono::nanoseconds
+>::type
+__safe_nanosecond_cast(chrono::duration<_Rep, _Period> __d)
+{
+    using namespace chrono;
+    if (__d.count() == 0) {
+        return nanoseconds(0);
+    }
+
+    using __ratio = ratio_divide<_Period, nano>;
+    using __ns_rep = nanoseconds::rep;
+    __ns_rep __result_max = std::numeric_limits<__ns_rep>::max();
+    if (__d.count() > 0 && __d.count() > __result_max / __ratio::num) {
+        return nanoseconds::max();
+    }
+
+    __ns_rep __result_min = std::numeric_limits<__ns_rep>::min();
+    if (__d.count() < 0 && __d.count() < __result_min / __ratio::num) {
+        return nanoseconds::min();
+    }
+
+    __ns_rep __result = __d.count() * __ratio::num / __ratio::den;
+    if (__result == 0) {
+        return nanoseconds(1);
+    }
+
+    return nanoseconds(__result);
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_THREADS
+template <class _Predicate>
+void
+condition_variable::wait(unique_lock<mutex>& __lk, _Predicate __pred)
+{
+    while (!__pred())
+        wait(__lk);
+}
+
+template <class _Clock, class _Duration>
+cv_status
+condition_variable::wait_until(unique_lock<mutex>& __lk,
+                               const chrono::time_point<_Clock, _Duration>& __t)
+{
+    using namespace chrono;
+    using __clock_tp_ns = time_point<_Clock, nanoseconds>;
+
+    typename _Clock::time_point __now = _Clock::now();
+    if (__t <= __now)
+        return cv_status::timeout;
+
+    __clock_tp_ns __t_ns = __clock_tp_ns(__safe_nanosecond_cast(__t.time_since_epoch()));
+
+    __do_timed_wait(__lk, __t_ns);
+    return _Clock::now() < __t ? cv_status::no_timeout : cv_status::timeout;
+}
+
+template <class _Clock, class _Duration, class _Predicate>
+bool
+condition_variable::wait_until(unique_lock<mutex>& __lk,
+                   const chrono::time_point<_Clock, _Duration>& __t,
+                   _Predicate __pred)
+{
+    while (!__pred())
+    {
+        if (wait_until(__lk, __t) == cv_status::timeout)
+            return __pred();
+    }
+    return true;
+}
+
+template <class _Rep, class _Period>
+cv_status
+condition_variable::wait_for(unique_lock<mutex>& __lk,
+                             const chrono::duration<_Rep, _Period>& __d)
+{
+    using namespace chrono;
+    if (__d <= __d.zero())
+        return cv_status::timeout;
+    using __ns_rep = nanoseconds::rep;
+    steady_clock::time_point __c_now = steady_clock::now();
+
+#if defined(_LIBCUDACXX_HAS_COND_CLOCKWAIT)
+    using __clock_tp_ns = time_point<steady_clock, nanoseconds>;
+    __ns_rep __now_count_ns = __safe_nanosecond_cast(__c_now.time_since_epoch()).count();
+#else
+    using __clock_tp_ns = time_point<system_clock, nanoseconds>;
+    __ns_rep __now_count_ns = __safe_nanosecond_cast(system_clock::now().time_since_epoch()).count();
+#endif
+
+    __ns_rep __d_ns_count = __safe_nanosecond_cast(__d).count();
+
+    if (__now_count_ns > numeric_limits<__ns_rep>::max() - __d_ns_count) {
+        __do_timed_wait(__lk, __clock_tp_ns::max());
+    } else {
+        __do_timed_wait(__lk, __clock_tp_ns(nanoseconds(__now_count_ns + __d_ns_count)));
+    }
+
+    return steady_clock::now() - __c_now < __d ? cv_status::no_timeout :
+                                                 cv_status::timeout;
+}
+
+template <class _Rep, class _Period, class _Predicate>
+inline
+bool
+condition_variable::wait_for(unique_lock<mutex>& __lk,
+                             const chrono::duration<_Rep, _Period>& __d,
+                             _Predicate __pred)
+{
+    return wait_until(__lk, chrono::steady_clock::now() + __d,
+                      _CUDA_VSTD::move(__pred));
+}
+
+#if defined(_LIBCUDACXX_HAS_COND_CLOCKWAIT)
+inline
+void
+condition_variable::__do_timed_wait(unique_lock<mutex>& __lk,
+     chrono::time_point<chrono::steady_clock, chrono::nanoseconds> __tp) _NOEXCEPT
+{
+    using namespace chrono;
+    if (!__lk.owns_lock())
+        __throw_system_error(EPERM,
+                            "condition_variable::timed wait: mutex not locked");
+    nanoseconds __d = __tp.time_since_epoch();
+    timespec __ts;
+    seconds __s = duration_cast<seconds>(__d);
+    using __ts_sec = decltype(__ts.tv_sec);
+    const __ts_sec __ts_sec_max = numeric_limits<__ts_sec>::max();
+    if (__s.count() < __ts_sec_max)
+    {
+        __ts.tv_sec = static_cast<__ts_sec>(__s.count());
+        __ts.tv_nsec = (__d - __s).count();
+    }
+    else
+    {
+        __ts.tv_sec = __ts_sec_max;
+        __ts.tv_nsec = giga::num - 1;
+    }
+    int __ec = pthread_cond_clockwait(&__cv_, __lk.mutex()->native_handle(), CLOCK_MONOTONIC, &__ts);
+    if (__ec != 0 && __ec != ETIMEDOUT)
+        __throw_system_error(__ec, "condition_variable timed_wait failed");
+}
+#endif // _LIBCUDACXX_HAS_COND_CLOCKWAIT
+
+template <class _Clock>
+inline
+void
+condition_variable::__do_timed_wait(unique_lock<mutex>& __lk,
+     chrono::time_point<_Clock, chrono::nanoseconds> __tp) _NOEXCEPT
+{
+    wait_for(__lk, __tp - _Clock::now());
+}
+
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX___MUTEX_BASE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle
new file mode 100644
index 000000000000..7a6e61af049b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__node_handle
@@ -0,0 +1,208 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___NODE_HANDLE
+#define _LIBCUDACXX___NODE_HANDLE
+
+#include <__config>
+#include <memory>
+#include <optional>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 14
+
+// Specialized in __tree & __hash_table for their _NodeType.
+template <class _NodeType, class _Alloc>
+struct __generic_container_node_destructor;
+
+template <class _NodeType, class _Alloc,
+          template <class, class> class _MapOrSetSpecifics>
+class _LIBCUDACXX_TEMPLATE_VIS __basic_node_handle
+    : public _MapOrSetSpecifics<
+          _NodeType,
+          __basic_node_handle<_NodeType, _Alloc, _MapOrSetSpecifics>>
+{
+    template <class _Tp, class _Compare, class _Allocator>
+        friend class __tree;
+    template <class _Tp, class _Hash, class _Equal, class _Allocator>
+        friend class __hash_table;
+    friend struct _MapOrSetSpecifics<
+        _NodeType, __basic_node_handle<_NodeType, _Alloc, _MapOrSetSpecifics>>;
+
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_pointer<typename __alloc_traits::void_pointer,
+                                      _NodeType>::type
+        __node_pointer_type;
+
+public:
+    typedef _Alloc allocator_type;
+
+private:
+    __node_pointer_type __ptr_ = nullptr;
+    optional<allocator_type> __alloc_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __release_ptr()
+    {
+        __ptr_ = nullptr;
+        __alloc_ = _CUDA_VSTD::nullopt;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __destroy_node_pointer()
+    {
+        if (__ptr_ != nullptr)
+        {
+            typedef typename __allocator_traits_rebind<
+                allocator_type, _NodeType>::type __node_alloc_type;
+            __node_alloc_type __alloc(*__alloc_);
+            __generic_container_node_destructor<_NodeType, __node_alloc_type>(
+                __alloc, true)(__ptr_);
+            __ptr_ = nullptr;
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __basic_node_handle(__node_pointer_type __ptr,
+                        allocator_type const& __alloc)
+            : __ptr_(__ptr), __alloc_(__alloc)
+    {
+    }
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __basic_node_handle() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __basic_node_handle(__basic_node_handle&& __other) noexcept
+            : __ptr_(__other.__ptr_),
+              __alloc_(_CUDA_VSTD::move(__other.__alloc_))
+    {
+        __other.__ptr_ = nullptr;
+        __other.__alloc_ = _CUDA_VSTD::nullopt;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __basic_node_handle& operator=(__basic_node_handle&& __other)
+    {
+        _LIBCUDACXX_ASSERT(
+            __alloc_ == _CUDA_VSTD::nullopt ||
+            __alloc_traits::propagate_on_container_move_assignment::value ||
+            __alloc_ == __other.__alloc_,
+            "node_type with incompatible allocator passed to "
+            "node_type::operator=(node_type&&)");
+
+        __destroy_node_pointer();
+        __ptr_ = __other.__ptr_;
+
+        if (__alloc_traits::propagate_on_container_move_assignment::value ||
+            __alloc_ == _CUDA_VSTD::nullopt)
+            __alloc_ = _CUDA_VSTD::move(__other.__alloc_);
+
+        __other.__ptr_ = nullptr;
+        __other.__alloc_ = _CUDA_VSTD::nullopt;
+
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const { return *__alloc_; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit operator bool() const { return __ptr_ != nullptr; }
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const { return __ptr_ == nullptr; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(__basic_node_handle& __other) noexcept(
+        __alloc_traits::propagate_on_container_swap::value ||
+        __alloc_traits::is_always_equal::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(__ptr_, __other.__ptr_);
+        if (__alloc_traits::propagate_on_container_swap::value ||
+            __alloc_ == _CUDA_VSTD::nullopt || __other.__alloc_ == _CUDA_VSTD::nullopt)
+            swap(__alloc_, __other.__alloc_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend void swap(__basic_node_handle& __a, __basic_node_handle& __b)
+        noexcept(noexcept(__a.swap(__b))) { __a.swap(__b); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__basic_node_handle()
+    {
+        __destroy_node_pointer();
+    }
+};
+
+template <class _NodeType, class _Derived>
+struct __set_node_handle_specifics
+{
+    typedef typename _NodeType::__node_value_type value_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type& value() const
+    {
+        return static_cast<_Derived const*>(this)->__ptr_->__value_;
+    }
+};
+
+template <class _NodeType, class _Derived>
+struct __map_node_handle_specifics
+{
+    typedef typename _NodeType::__node_value_type::key_type key_type;
+    typedef typename _NodeType::__node_value_type::mapped_type mapped_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_type& key() const
+    {
+        return static_cast<_Derived const*>(this)->
+            __ptr_->__value_.__ref().first;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mapped_type& mapped() const
+    {
+        return static_cast<_Derived const*>(this)->
+            __ptr_->__value_.__ref().second;
+    }
+};
+
+template <class _NodeType, class _Alloc>
+using __set_node_handle =
+    __basic_node_handle< _NodeType, _Alloc, __set_node_handle_specifics>;
+
+template <class _NodeType, class _Alloc>
+using __map_node_handle =
+    __basic_node_handle< _NodeType, _Alloc, __map_node_handle_specifics>;
+
+template <class _Iterator, class _NodeType>
+struct _LIBCUDACXX_TEMPLATE_VIS __insert_return_type
+{
+    _Iterator position;
+    bool inserted;
+    _NodeType node;
+};
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_END_NAMESPACE_STD
+_LIBCUDACXX_POP_MACROS
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr
new file mode 100644
index 000000000000..bf3c3c385ac8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__nullptr
@@ -0,0 +1,61 @@
+// -*- C++ -*-
+//===--------------------------- __nullptr --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_NULLPTR
+#define _LIBCUDACXX_NULLPTR
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_NULLPTR
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+struct _LIBCUDACXX_TEMPLATE_VIS nullptr_t
+{
+    void* __lx;
+
+    struct __nat {int __for_bool_;};
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR nullptr_t() : __lx(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR nullptr_t(int __nat::*) : __lx(0) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR operator int __nat::*() const {return 0;}
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+        operator _Tp* () const {return 0;}
+
+    template <class _Tp, class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator _Tp _Up::* () const {return 0;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR bool operator==(nullptr_t, nullptr_t) {return true;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR bool operator!=(nullptr_t, nullptr_t) {return false;}
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR nullptr_t __get_nullptr_t() {return nullptr_t(0);}
+
+#define nullptr _CUDA_VSTD::__get_nullptr_t()
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#else  // _LIBCUDACXX_HAS_NO_NULLPTR
+
+namespace std
+{
+    typedef decltype(nullptr) nullptr_t;
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_NULLPTR
+
+#endif  // _LIBCUDACXX_NULLPTR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_pop b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_pop
new file mode 100644
index 000000000000..27a9a68b4e60
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_pop
@@ -0,0 +1,16 @@
+// -*- C++ -*-
+//===---------------------------- chrono ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_MSVC_WARNING)
+  #pragma warning(pop)
+#endif
+
+#if defined(_LIBCUDACXX_POP_MACROS)
+  _LIBCUDACXX_POP_MACROS
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push
new file mode 100644
index 000000000000..ad34a38d9364
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__pragma_push
@@ -0,0 +1,25 @@
+// -*- C++ -*-
+//===---------------------------- chrono ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+  #pragma GCC system_header
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_MSVC_WARNING)
+  #pragma warning(push)
+  #pragma warning(disable : _LIBCUDACXX_MSVC_DISABLED_WARNINGS)
+#endif
+
+#if defined(_LIBCUDACXX_PUSH_MACROS)
+  _LIBCUDACXX_PUSH_MACROS
+#endif
+
+#ifndef __cuda_std__
+#include <__undef_macros>
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer
new file mode 100644
index 000000000000..d303f9478fbb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__split_buffer
@@ -0,0 +1,644 @@
+// -*- C++ -*-
+#ifndef _LIBCUDACXX_SPLIT_BUFFER
+#define _LIBCUDACXX_SPLIT_BUFFER
+
+#include <__config>
+#include <type_traits>
+#include <algorithm>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <bool>
+class __split_buffer_common
+{
+protected:
+    void __throw_length_error() const;
+    void __throw_out_of_range() const;
+};
+
+template <class _Tp, class _Allocator = allocator<_Tp> >
+struct __split_buffer
+    : private __split_buffer_common<true>
+{
+private:
+    __split_buffer(const __split_buffer&);
+    __split_buffer& operator=(const __split_buffer&);
+public:
+    typedef _Tp                                             value_type;
+    typedef _Allocator                                      allocator_type;
+    typedef typename remove_reference<allocator_type>::type __alloc_rr;
+    typedef allocator_traits<__alloc_rr>                    __alloc_traits;
+    typedef value_type&                                     reference;
+    typedef const value_type&                               const_reference;
+    typedef typename __alloc_traits::size_type              size_type;
+    typedef typename __alloc_traits::difference_type        difference_type;
+    typedef typename __alloc_traits::pointer                pointer;
+    typedef typename __alloc_traits::const_pointer          const_pointer;
+    typedef pointer                                         iterator;
+    typedef const_pointer                                   const_iterator;
+
+    pointer                                         __first_;
+    pointer                                         __begin_;
+    pointer                                         __end_;
+    __compressed_pair<pointer, allocator_type> __end_cap_;
+
+    typedef typename add_lvalue_reference<allocator_type>::type __alloc_ref;
+    typedef typename add_lvalue_reference<allocator_type>::type __alloc_const_ref;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __alloc_rr&           __alloc() _NOEXCEPT         {return __end_cap_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY const __alloc_rr&     __alloc() const _NOEXCEPT   {return __end_cap_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY pointer&              __end_cap() _NOEXCEPT       {return __end_cap_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY const pointer&        __end_cap() const _NOEXCEPT {return __end_cap_.first();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __split_buffer()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __split_buffer(__alloc_rr& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __split_buffer(const __alloc_rr& __a);
+    __split_buffer(size_type __cap, size_type __start, __alloc_rr& __a);
+    ~__split_buffer();
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __split_buffer(__split_buffer&& __c)
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value);
+    __split_buffer(__split_buffer&& __c, const __alloc_rr& __a);
+    __split_buffer& operator=(__split_buffer&& __c)
+        _NOEXCEPT_((__alloc_traits::propagate_on_container_move_assignment::value &&
+                is_nothrow_move_assignable<allocator_type>::value) ||
+               !__alloc_traits::propagate_on_container_move_assignment::value);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY       iterator begin() _NOEXCEPT       {return __begin_;}
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator begin() const _NOEXCEPT {return __begin_;}
+    _LIBCUDACXX_INLINE_VISIBILITY       iterator end() _NOEXCEPT         {return __end_;}
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator end() const _NOEXCEPT   {return __end_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT
+        {__destruct_at_end(__begin_);}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type size() const {return static_cast<size_type>(__end_ - __begin_);}
+    _LIBCUDACXX_INLINE_VISIBILITY bool empty()     const {return __end_ == __begin_;}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type capacity() const {return static_cast<size_type>(__end_cap() - __first_);}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type __front_spare() const {return static_cast<size_type>(__begin_ - __first_);}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type __back_spare() const {return static_cast<size_type>(__end_cap() - __end_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY       reference front()       {return *__begin_;}
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference front() const {return *__begin_;}
+    _LIBCUDACXX_INLINE_VISIBILITY       reference back()        {return *(__end_ - 1);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference back() const  {return *(__end_ - 1);}
+
+    void reserve(size_type __n);
+    void shrink_to_fit() _NOEXCEPT;
+    void push_front(const_reference __x);
+    _LIBCUDACXX_INLINE_VISIBILITY void push_back(const_reference __x);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    void push_front(value_type&& __x);
+    void push_back(value_type&& __x);
+    template <class... _Args>
+        void emplace_back(_Args&&... __args);
+#endif  // !defined(_LIBCUDACXX_CXX03_LANG)
+
+    _LIBCUDACXX_INLINE_VISIBILITY void pop_front() {__destruct_at_begin(__begin_+1);}
+    _LIBCUDACXX_INLINE_VISIBILITY void pop_back() {__destruct_at_end(__end_-1);}
+
+    void __construct_at_end(size_type __n);
+    void __construct_at_end(size_type __n, const_reference __x);
+    template <class _InputIter>
+        typename enable_if
+        <
+            __is_input_iterator<_InputIter>::value &&
+           !__is_forward_iterator<_InputIter>::value,
+            void
+        >::type
+        __construct_at_end(_InputIter __first, _InputIter __last);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value,
+            void
+        >::type
+        __construct_at_end(_ForwardIterator __first, _ForwardIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY void __destruct_at_begin(pointer __new_begin)
+        {__destruct_at_begin(__new_begin, is_trivially_destructible<value_type>());}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void __destruct_at_begin(pointer __new_begin, false_type);
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void __destruct_at_begin(pointer __new_begin, true_type);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __destruct_at_end(pointer __new_last) _NOEXCEPT
+        {__destruct_at_end(__new_last, false_type());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+        void __destruct_at_end(pointer __new_last, false_type) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+        void __destruct_at_end(pointer __new_last, true_type) _NOEXCEPT;
+
+    void swap(__split_buffer& __x)
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value||
+                   __is_nothrow_swappable<__alloc_rr>::value);
+
+    bool __invariants() const;
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__split_buffer& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+        {
+            __alloc() = _CUDA_VSTD::move(__c.__alloc());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__split_buffer&, false_type) _NOEXCEPT
+        {}
+
+    struct _ConstructTransaction {
+      explicit _ConstructTransaction(pointer* __p, size_type __n) _NOEXCEPT
+      : __pos_(*__p), __end_(*__p + __n), __dest_(__p) {
+      }
+      ~_ConstructTransaction() {
+        *__dest_ = __pos_;
+      }
+      pointer __pos_;
+     const pointer __end_;
+    private:
+     pointer *__dest_;
+    };
+};
+
+template <class _Tp, class _Allocator>
+bool
+__split_buffer<_Tp, _Allocator>::__invariants() const
+{
+    if (__first_ == nullptr)
+    {
+        if (__begin_ != nullptr)
+            return false;
+        if (__end_ != nullptr)
+            return false;
+        if (__end_cap() != nullptr)
+            return false;
+    }
+    else
+    {
+        if (__begin_ < __first_)
+            return false;
+        if (__end_ < __begin_)
+            return false;
+        if (__end_cap() < __end_)
+            return false;
+    }
+    return true;
+}
+
+//  Default constructs __n objects starting at __end_
+//  throws if construction throws
+//  Precondition:  __n > 0
+//  Precondition:  size() + __n <= capacity()
+//  Postcondition:  size() == size() + __n
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::__construct_at_end(size_type __n)
+{
+    _ConstructTransaction __tx(&this->__end_, __n);
+    for (; __tx.__pos_ != __tx.__end_; ++__tx.__pos_) {
+        __alloc_traits::construct(this->__alloc(), _CUDA_VSTD::__to_raw_pointer(__tx.__pos_));
+    }
+}
+
+//  Copy constructs __n objects starting at __end_ from __x
+//  throws if construction throws
+//  Precondition:  __n > 0
+//  Precondition:  size() + __n <= capacity()
+//  Postcondition:  size() == old size() + __n
+//  Postcondition:  [i] == __x for all i in [size() - __n, __n)
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::__construct_at_end(size_type __n, const_reference __x)
+{
+    _ConstructTransaction __tx(&this->__end_, __n);
+    for (; __tx.__pos_ != __tx.__end_; ++__tx.__pos_) {
+        __alloc_traits::construct(this->__alloc(),
+            _CUDA_VSTD::__to_raw_pointer(__tx.__pos_), __x);
+    }
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIter>
+typename enable_if
+<
+     __is_input_iterator<_InputIter>::value &&
+    !__is_forward_iterator<_InputIter>::value,
+    void
+>::type
+__split_buffer<_Tp, _Allocator>::__construct_at_end(_InputIter __first, _InputIter __last)
+{
+    __alloc_rr& __a = this->__alloc();
+    for (; __first != __last; ++__first)
+    {
+        if (__end_ == __end_cap())
+        {
+            size_type __old_cap = __end_cap() - __first_;
+            size_type __new_cap = _CUDA_VSTD::max<size_type>(2 * __old_cap, 8);
+            __split_buffer __buf(__new_cap, 0, __a);
+            for (pointer __p = __begin_; __p != __end_; ++__p, ++__buf.__end_)
+                __alloc_traits::construct(__buf.__alloc(),
+                        _CUDA_VSTD::__to_raw_pointer(__buf.__end_), _CUDA_VSTD::move(*__p));
+            swap(__buf);
+        }
+        __alloc_traits::construct(__a, _CUDA_VSTD::__to_raw_pointer(this->__end_), *__first);
+        ++this->__end_;
+    }
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value,
+    void
+>::type
+__split_buffer<_Tp, _Allocator>::__construct_at_end(_ForwardIterator __first, _ForwardIterator __last)
+{
+    _ConstructTransaction __tx(&this->__end_, std::distance(__first, __last));
+    for (; __tx.__pos_ != __tx.__end_; ++__tx.__pos_, ++__first) {
+        __alloc_traits::construct(this->__alloc(),
+            _CUDA_VSTD::__to_raw_pointer(__tx.__pos_), *__first);
+    }
+}
+
+template <class _Tp, class _Allocator>
+inline
+void
+__split_buffer<_Tp, _Allocator>::__destruct_at_begin(pointer __new_begin, false_type)
+{
+    while (__begin_ != __new_begin)
+        __alloc_traits::destroy(__alloc(), __to_raw_pointer(__begin_++));
+}
+
+template <class _Tp, class _Allocator>
+inline
+void
+__split_buffer<_Tp, _Allocator>::__destruct_at_begin(pointer __new_begin, true_type)
+{
+    __begin_ = __new_begin;
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__split_buffer<_Tp, _Allocator>::__destruct_at_end(pointer __new_last, false_type) _NOEXCEPT
+{
+    while (__new_last != __end_)
+        __alloc_traits::destroy(__alloc(), __to_raw_pointer(--__end_));
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__split_buffer<_Tp, _Allocator>::__destruct_at_end(pointer __new_last, true_type) _NOEXCEPT
+{
+    __end_ = __new_last;
+}
+
+template <class _Tp, class _Allocator>
+__split_buffer<_Tp, _Allocator>::__split_buffer(size_type __cap, size_type __start, __alloc_rr& __a)
+    : __end_cap_(nullptr, __a)
+{
+    __first_ = __cap != 0 ? __alloc_traits::allocate(__alloc(), __cap) : nullptr;
+    __begin_ = __end_ = __first_ + __start;
+    __end_cap() = __first_ + __cap;
+}
+
+template <class _Tp, class _Allocator>
+inline
+__split_buffer<_Tp, _Allocator>::__split_buffer()
+    _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+    : __first_(nullptr), __begin_(nullptr), __end_(nullptr), __end_cap_(nullptr)
+{
+}
+
+template <class _Tp, class _Allocator>
+inline
+__split_buffer<_Tp, _Allocator>::__split_buffer(__alloc_rr& __a)
+    : __first_(nullptr), __begin_(nullptr), __end_(nullptr), __end_cap_(nullptr, __a)
+{
+}
+
+template <class _Tp, class _Allocator>
+inline
+__split_buffer<_Tp, _Allocator>::__split_buffer(const __alloc_rr& __a)
+    : __first_(nullptr), __begin_(nullptr), __end_(nullptr), __end_cap_(nullptr, __a)
+{
+}
+
+template <class _Tp, class _Allocator>
+__split_buffer<_Tp, _Allocator>::~__split_buffer()
+{
+    clear();
+    if (__first_)
+        __alloc_traits::deallocate(__alloc(), __first_, capacity());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+__split_buffer<_Tp, _Allocator>::__split_buffer(__split_buffer&& __c)
+    _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value)
+    : __first_(_CUDA_VSTD::move(__c.__first_)),
+      __begin_(_CUDA_VSTD::move(__c.__begin_)),
+      __end_(_CUDA_VSTD::move(__c.__end_)),
+      __end_cap_(_CUDA_VSTD::move(__c.__end_cap_))
+{
+    __c.__first_ = nullptr;
+    __c.__begin_ = nullptr;
+    __c.__end_ = nullptr;
+    __c.__end_cap() = nullptr;
+}
+
+template <class _Tp, class _Allocator>
+__split_buffer<_Tp, _Allocator>::__split_buffer(__split_buffer&& __c, const __alloc_rr& __a)
+    : __end_cap_(__second_tag(), __a)
+{
+    if (__a == __c.__alloc())
+    {
+        __first_ = __c.__first_;
+        __begin_ = __c.__begin_;
+        __end_ = __c.__end_;
+        __end_cap() = __c.__end_cap();
+        __c.__first_ = nullptr;
+        __c.__begin_ = nullptr;
+        __c.__end_ = nullptr;
+        __c.__end_cap() = nullptr;
+    }
+    else
+    {
+        size_type __cap = __c.size();
+        __first_ = __alloc_traits::allocate(__alloc(), __cap);
+        __begin_ = __end_ = __first_;
+        __end_cap() = __first_ + __cap;
+        typedef move_iterator<iterator> _Ip;
+        __construct_at_end(_Ip(__c.begin()), _Ip(__c.end()));
+    }
+}
+
+template <class _Tp, class _Allocator>
+__split_buffer<_Tp, _Allocator>&
+__split_buffer<_Tp, _Allocator>::operator=(__split_buffer&& __c)
+    _NOEXCEPT_((__alloc_traits::propagate_on_container_move_assignment::value &&
+                is_nothrow_move_assignable<allocator_type>::value) ||
+               !__alloc_traits::propagate_on_container_move_assignment::value)
+{
+    clear();
+    shrink_to_fit();
+    __first_ = __c.__first_;
+    __begin_ = __c.__begin_;
+    __end_ = __c.__end_;
+    __end_cap() = __c.__end_cap();
+    __move_assign_alloc(__c,
+        integral_constant<bool,
+                          __alloc_traits::propagate_on_container_move_assignment::value>());
+    __c.__first_ = __c.__begin_ = __c.__end_ = __c.__end_cap() = nullptr;
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::swap(__split_buffer& __x)
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value||
+                   __is_nothrow_swappable<__alloc_rr>::value)
+{
+    _CUDA_VSTD::swap(__first_, __x.__first_);
+    _CUDA_VSTD::swap(__begin_, __x.__begin_);
+    _CUDA_VSTD::swap(__end_, __x.__end_);
+    _CUDA_VSTD::swap(__end_cap(), __x.__end_cap());
+    __swap_allocator(__alloc(), __x.__alloc());
+}
+
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::reserve(size_type __n)
+{
+    if (__n < capacity())
+    {
+        __split_buffer<value_type, __alloc_rr&> __t(__n, 0, __alloc());
+        __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                               move_iterator<pointer>(__end_));
+        _CUDA_VSTD::swap(__first_, __t.__first_);
+        _CUDA_VSTD::swap(__begin_, __t.__begin_);
+        _CUDA_VSTD::swap(__end_, __t.__end_);
+        _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+    }
+}
+
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::shrink_to_fit() _NOEXCEPT
+{
+    if (capacity() > size())
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            __split_buffer<value_type, __alloc_rr&> __t(size(), 0, __alloc());
+            __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                                   move_iterator<pointer>(__end_));
+            __t.__end_ = __t.__begin_ + (__end_ - __begin_);
+            _CUDA_VSTD::swap(__first_, __t.__first_);
+            _CUDA_VSTD::swap(__begin_, __t.__begin_);
+            _CUDA_VSTD::swap(__end_, __t.__end_);
+            _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::push_front(const_reference __x)
+{
+    if (__begin_ == __first_)
+    {
+        if (__end_ < __end_cap())
+        {
+            difference_type __d = __end_cap() - __end_;
+            __d = (__d + 1) / 2;
+            __begin_ = _CUDA_VSTD::move_backward(__begin_, __end_, __end_ + __d);
+            __end_ += __d;
+        }
+        else
+        {
+            size_type __c = max<size_type>(2 * static_cast<size_t>(__end_cap() - __first_), 1);
+            __split_buffer<value_type, __alloc_rr&> __t(__c, (__c + 3) / 4, __alloc());
+            __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                                   move_iterator<pointer>(__end_));
+            _CUDA_VSTD::swap(__first_, __t.__first_);
+            _CUDA_VSTD::swap(__begin_, __t.__begin_);
+            _CUDA_VSTD::swap(__end_, __t.__end_);
+            _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+        }
+    }
+    __alloc_traits::construct(__alloc(), _CUDA_VSTD::__to_raw_pointer(__begin_-1), __x);
+    --__begin_;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::push_front(value_type&& __x)
+{
+    if (__begin_ == __first_)
+    {
+        if (__end_ < __end_cap())
+        {
+            difference_type __d = __end_cap() - __end_;
+            __d = (__d + 1) / 2;
+            __begin_ = _CUDA_VSTD::move_backward(__begin_, __end_, __end_ + __d);
+            __end_ += __d;
+        }
+        else
+        {
+            size_type __c = max<size_type>(2 * static_cast<size_t>(__end_cap() - __first_), 1);
+            __split_buffer<value_type, __alloc_rr&> __t(__c, (__c + 3) / 4, __alloc());
+            __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                                   move_iterator<pointer>(__end_));
+            _CUDA_VSTD::swap(__first_, __t.__first_);
+            _CUDA_VSTD::swap(__begin_, __t.__begin_);
+            _CUDA_VSTD::swap(__end_, __t.__end_);
+            _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+        }
+    }
+    __alloc_traits::construct(__alloc(), _CUDA_VSTD::__to_raw_pointer(__begin_-1),
+            _CUDA_VSTD::move(__x));
+    --__begin_;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__split_buffer<_Tp, _Allocator>::push_back(const_reference __x)
+{
+    if (__end_ == __end_cap())
+    {
+        if (__begin_ > __first_)
+        {
+            difference_type __d = __begin_ - __first_;
+            __d = (__d + 1) / 2;
+            __end_ = _CUDA_VSTD::move(__begin_, __end_, __begin_ - __d);
+            __begin_ -= __d;
+        }
+        else
+        {
+            size_type __c = max<size_type>(2 * static_cast<size_t>(__end_cap() - __first_), 1);
+            __split_buffer<value_type, __alloc_rr&> __t(__c, __c / 4, __alloc());
+            __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                                   move_iterator<pointer>(__end_));
+            _CUDA_VSTD::swap(__first_, __t.__first_);
+            _CUDA_VSTD::swap(__begin_, __t.__begin_);
+            _CUDA_VSTD::swap(__end_, __t.__end_);
+            _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+        }
+    }
+    __alloc_traits::construct(__alloc(), _CUDA_VSTD::__to_raw_pointer(__end_), __x);
+    ++__end_;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+void
+__split_buffer<_Tp, _Allocator>::push_back(value_type&& __x)
+{
+    if (__end_ == __end_cap())
+    {
+        if (__begin_ > __first_)
+        {
+            difference_type __d = __begin_ - __first_;
+            __d = (__d + 1) / 2;
+            __end_ = _CUDA_VSTD::move(__begin_, __end_, __begin_ - __d);
+            __begin_ -= __d;
+        }
+        else
+        {
+            size_type __c = max<size_type>(2 * static_cast<size_t>(__end_cap() - __first_), 1);
+            __split_buffer<value_type, __alloc_rr&> __t(__c, __c / 4, __alloc());
+            __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                                   move_iterator<pointer>(__end_));
+            _CUDA_VSTD::swap(__first_, __t.__first_);
+            _CUDA_VSTD::swap(__begin_, __t.__begin_);
+            _CUDA_VSTD::swap(__end_, __t.__end_);
+            _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+        }
+    }
+    __alloc_traits::construct(__alloc(), _CUDA_VSTD::__to_raw_pointer(__end_),
+            _CUDA_VSTD::move(__x));
+    ++__end_;
+}
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+void
+__split_buffer<_Tp, _Allocator>::emplace_back(_Args&&... __args)
+{
+    if (__end_ == __end_cap())
+    {
+        if (__begin_ > __first_)
+        {
+            difference_type __d = __begin_ - __first_;
+            __d = (__d + 1) / 2;
+            __end_ = _CUDA_VSTD::move(__begin_, __end_, __begin_ - __d);
+            __begin_ -= __d;
+        }
+        else
+        {
+            size_type __c = max<size_type>(2 * static_cast<size_t>(__end_cap() - __first_), 1);
+            __split_buffer<value_type, __alloc_rr&> __t(__c, __c / 4, __alloc());
+            __t.__construct_at_end(move_iterator<pointer>(__begin_),
+                                   move_iterator<pointer>(__end_));
+            _CUDA_VSTD::swap(__first_, __t.__first_);
+            _CUDA_VSTD::swap(__begin_, __t.__begin_);
+            _CUDA_VSTD::swap(__end_, __t.__end_);
+            _CUDA_VSTD::swap(__end_cap(), __t.__end_cap());
+        }
+    }
+    __alloc_traits::construct(__alloc(), _CUDA_VSTD::__to_raw_pointer(__end_),
+                              _CUDA_VSTD::forward<_Args>(__args)...);
+    ++__end_;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__split_buffer<_Tp, _Allocator>& __x, __split_buffer<_Tp, _Allocator>& __y)
+        _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_SPLIT_BUFFER
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator
new file mode 100644
index 000000000000..94f774df4159
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__sso_allocator
@@ -0,0 +1,76 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___SSO_ALLOCATOR
+#define _LIBCUDACXX___SSO_ALLOCATOR
+
+#include <__config>
+#include <type_traits>
+#include <new>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp, size_t _Np> class _LIBCUDACXX_HIDDEN __sso_allocator;
+
+template <size_t _Np>
+class _LIBCUDACXX_HIDDEN __sso_allocator<void, _Np>
+{
+public:
+    typedef const void*       const_pointer;
+    typedef void              value_type;
+};
+
+template <class _Tp, size_t _Np>
+class _LIBCUDACXX_HIDDEN __sso_allocator
+{
+    typename aligned_storage<sizeof(_Tp) * _Np>::type buf_;
+    bool __allocated_;
+public:
+    typedef size_t            size_type;
+    typedef _Tp*              pointer;
+    typedef _Tp               value_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __sso_allocator() throw() : __allocated_(false) {}
+    _LIBCUDACXX_INLINE_VISIBILITY __sso_allocator(const __sso_allocator&) throw() : __allocated_(false) {}
+    template <class _Up> _LIBCUDACXX_INLINE_VISIBILITY __sso_allocator(const __sso_allocator<_Up, _Np>&) throw()
+         : __allocated_(false) {}
+private:
+    __sso_allocator& operator=(const __sso_allocator&);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY pointer allocate(size_type __n, typename __sso_allocator<void, _Np>::const_pointer = 0)
+    {
+        if (!__allocated_ && __n <= _Np)
+        {
+            __allocated_ = true;
+            return (pointer)&buf_;
+        }
+        return static_cast<pointer>(_CUDA_VSTD::__libcpp_allocate(__n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp)));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY void deallocate(pointer __p, size_type __n)
+    {
+        if (__p == (pointer)&buf_)
+            __allocated_ = false;
+        else
+            _CUDA_VSTD::__libcpp_deallocate(__p, __n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY size_type max_size() const throw() {return size_type(~0) / sizeof(_Tp);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(__sso_allocator& __a) const {return &buf_ == &__a.buf_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(__sso_allocator& __a) const {return &buf_ != &__a.buf_;}
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX___SSO_ALLOCATOR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream
new file mode 100644
index 000000000000..996db0387345
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__std_stream
@@ -0,0 +1,361 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___STD_STREAM
+#define _LIBCUDACXX___STD_STREAM
+
+#include <__config>
+#include <ostream>
+#include <istream>
+#include <__locale>
+#include <cstdio>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+static const int __limit = 8;
+
+// __stdinbuf
+
+template <class _CharT>
+class _LIBCUDACXX_HIDDEN __stdinbuf
+    : public basic_streambuf<_CharT, char_traits<_CharT> >
+{
+public:
+    typedef _CharT                           char_type;
+    typedef char_traits<char_type>           traits_type;
+    typedef typename traits_type::int_type   int_type;
+    typedef typename traits_type::pos_type   pos_type;
+    typedef typename traits_type::off_type   off_type;
+    typedef typename traits_type::state_type state_type;
+
+    __stdinbuf(FILE* __fp, state_type* __st);
+
+protected:
+    virtual int_type underflow();
+    virtual int_type uflow();
+    virtual int_type pbackfail(int_type __c = traits_type::eof());
+    virtual void imbue(const locale& __loc);
+
+private:
+
+    FILE* __file_;
+    const codecvt<char_type, char, state_type>* __cv_;
+    state_type* __st_;
+    int __encoding_;
+    int_type __last_consumed_;
+    bool __last_consumed_is_next_;
+    bool __always_noconv_;
+
+    __stdinbuf(const __stdinbuf&);
+    __stdinbuf& operator=(const __stdinbuf&);
+
+    int_type __getchar(bool __consume);
+};
+
+template <class _CharT>
+__stdinbuf<_CharT>::__stdinbuf(FILE* __fp, state_type* __st)
+    : __file_(__fp),
+      __st_(__st),
+      __last_consumed_(traits_type::eof()),
+      __last_consumed_is_next_(false)
+{
+    imbue(this->getloc());
+}
+
+template <class _CharT>
+void
+__stdinbuf<_CharT>::imbue(const locale& __loc)
+{
+    __cv_ = &use_facet<codecvt<char_type, char, state_type> >(__loc);
+    __encoding_ = __cv_->encoding();
+    __always_noconv_ = __cv_->always_noconv();
+    if (__encoding_ > __limit)
+        __throw_runtime_error("unsupported locale for standard input");
+}
+
+template <class _CharT>
+typename __stdinbuf<_CharT>::int_type
+__stdinbuf<_CharT>::underflow()
+{
+    return __getchar(false);
+}
+
+template <class _CharT>
+typename __stdinbuf<_CharT>::int_type
+__stdinbuf<_CharT>::uflow()
+{
+    return __getchar(true);
+}
+
+template <class _CharT>
+typename __stdinbuf<_CharT>::int_type
+__stdinbuf<_CharT>::__getchar(bool __consume)
+{
+    if (__last_consumed_is_next_)
+    {
+        int_type __result = __last_consumed_;
+        if (__consume)
+        {
+            __last_consumed_ = traits_type::eof();
+            __last_consumed_is_next_ = false;
+        }
+        return __result;
+    }
+    char __extbuf[__limit];
+    int __nread = _CUDA_VSTD::max(1, __encoding_);
+    for (int __i = 0; __i < __nread; ++__i)
+    {
+        int __c = getc(__file_);
+        if (__c == EOF)
+            return traits_type::eof();
+        __extbuf[__i] = static_cast<char>(__c);
+    }
+    char_type __1buf;
+    if (__always_noconv_)
+        __1buf = static_cast<char_type>(__extbuf[0]);
+    else
+    {
+        const char* __enxt;
+        char_type* __inxt;
+        codecvt_base::result __r;
+        do
+        {
+            state_type __sv_st = *__st_;
+            __r = __cv_->in(*__st_, __extbuf, __extbuf + __nread, __enxt,
+                                   &__1buf, &__1buf + 1, __inxt);
+            switch (__r)
+            {
+            case _CUDA_VSTD::codecvt_base::ok:
+                break;
+            case codecvt_base::partial:
+                *__st_ = __sv_st;
+                if (__nread == sizeof(__extbuf))
+                    return traits_type::eof();
+                {
+                    int __c = getc(__file_);
+                    if (__c == EOF)
+                        return traits_type::eof();
+                    __extbuf[__nread] = static_cast<char>(__c);
+                }
+                ++__nread;
+                break;
+            case codecvt_base::error:
+                return traits_type::eof();
+            case _CUDA_VSTD::codecvt_base::noconv:
+                __1buf = static_cast<char_type>(__extbuf[0]);
+                break;
+            }
+        } while (__r == _CUDA_VSTD::codecvt_base::partial);
+    }
+    if (!__consume)
+    {
+        for (int __i = __nread; __i > 0;)
+        {
+            if (ungetc(traits_type::to_int_type(__extbuf[--__i]), __file_) == EOF)
+                return traits_type::eof();
+        }
+    }
+    else
+        __last_consumed_ = traits_type::to_int_type(__1buf);
+    return traits_type::to_int_type(__1buf);
+}
+
+template <class _CharT>
+typename __stdinbuf<_CharT>::int_type
+__stdinbuf<_CharT>::pbackfail(int_type __c)
+{
+    if (traits_type::eq_int_type(__c, traits_type::eof()))
+    {
+        if (!__last_consumed_is_next_)
+        {
+            __c = __last_consumed_;
+            __last_consumed_is_next_ = !traits_type::eq_int_type(__last_consumed_,
+                                                                 traits_type::eof());
+        }
+        return __c;
+    }
+    if (__last_consumed_is_next_)
+    {
+        char __extbuf[__limit];
+        char* __enxt;
+        const char_type __ci = traits_type::to_char_type(__last_consumed_);
+        const char_type* __inxt;
+        switch (__cv_->out(*__st_, &__ci, &__ci + 1, __inxt,
+                                  __extbuf, __extbuf + sizeof(__extbuf), __enxt))
+        {
+        case _CUDA_VSTD::codecvt_base::ok:
+            break;
+        case _CUDA_VSTD::codecvt_base::noconv:
+            __extbuf[0] = static_cast<char>(__last_consumed_);
+            __enxt = __extbuf + 1;
+            break;
+        case codecvt_base::partial:
+        case codecvt_base::error:
+            return traits_type::eof();
+        }
+        while (__enxt > __extbuf)
+            if (ungetc(*--__enxt, __file_) == EOF)
+                return traits_type::eof();
+    }
+    __last_consumed_ = __c;
+    __last_consumed_is_next_ = true;
+    return __c;
+}
+
+// __stdoutbuf
+
+template <class _CharT>
+class _LIBCUDACXX_HIDDEN __stdoutbuf
+    : public basic_streambuf<_CharT, char_traits<_CharT> >
+{
+public:
+    typedef _CharT                           char_type;
+    typedef char_traits<char_type>           traits_type;
+    typedef typename traits_type::int_type   int_type;
+    typedef typename traits_type::pos_type   pos_type;
+    typedef typename traits_type::off_type   off_type;
+    typedef typename traits_type::state_type state_type;
+
+    __stdoutbuf(FILE* __fp, state_type* __st);
+
+protected:
+    virtual int_type overflow (int_type __c = traits_type::eof());
+    virtual streamsize xsputn(const char_type* __s, streamsize __n);
+    virtual int sync();
+    virtual void imbue(const locale& __loc);
+
+private:
+    FILE* __file_;
+    const codecvt<char_type, char, state_type>* __cv_;
+    state_type* __st_;
+    bool __always_noconv_;
+
+    __stdoutbuf(const __stdoutbuf&);
+    __stdoutbuf& operator=(const __stdoutbuf&);
+};
+
+template <class _CharT>
+__stdoutbuf<_CharT>::__stdoutbuf(FILE* __fp, state_type* __st)
+    : __file_(__fp),
+      __cv_(&use_facet<codecvt<char_type, char, state_type> >(this->getloc())),
+      __st_(__st),
+      __always_noconv_(__cv_->always_noconv())
+{
+}
+
+template <class _CharT>
+typename __stdoutbuf<_CharT>::int_type
+__stdoutbuf<_CharT>::overflow(int_type __c)
+{
+    char __extbuf[__limit];
+    char_type __1buf;
+    if (!traits_type::eq_int_type(__c, traits_type::eof()))
+    {
+        __1buf = traits_type::to_char_type(__c);
+        if (__always_noconv_)
+        {
+            if (fwrite(&__1buf, sizeof(char_type), 1, __file_) != 1)
+                return traits_type::eof();
+        }
+        else
+        {
+            char* __extbe = __extbuf;
+            codecvt_base::result __r;
+            char_type* pbase = &__1buf;
+            char_type* pptr = pbase + 1;
+            do
+            {
+                const char_type* __e;
+                __r = __cv_->out(*__st_, pbase, pptr, __e,
+                                        __extbuf,
+                                        __extbuf + sizeof(__extbuf),
+                                        __extbe);
+                if (__e == pbase)
+                    return traits_type::eof();
+                if (__r == codecvt_base::noconv)
+                {
+                    if (fwrite(pbase, 1, 1, __file_) != 1)
+                        return traits_type::eof();
+                }
+                else if (__r == codecvt_base::ok || __r == codecvt_base::partial)
+                {
+                    size_t __nmemb = static_cast<size_t>(__extbe - __extbuf);
+                    if (fwrite(__extbuf, 1, __nmemb, __file_) != __nmemb)
+                        return traits_type::eof();
+                    if (__r == codecvt_base::partial)
+                    {
+                        pbase = const_cast<char_type*>(__e);
+                    }
+                }
+                else
+                    return traits_type::eof();
+            } while (__r == codecvt_base::partial);
+        }
+    }
+    return traits_type::not_eof(__c);
+}
+
+template <class _CharT>
+streamsize
+__stdoutbuf<_CharT>::xsputn(const char_type* __s, streamsize __n)
+{
+    if (__always_noconv_)
+        return fwrite(__s, sizeof(char_type), __n, __file_);
+    streamsize __i = 0;
+    for (; __i < __n; ++__i, ++__s)
+        if (overflow(traits_type::to_int_type(*__s)) == traits_type::eof())
+            break;
+    return __i;
+}
+
+template <class _CharT>
+int
+__stdoutbuf<_CharT>::sync()
+{
+    char __extbuf[__limit];
+    codecvt_base::result __r;
+    do
+    {
+        char* __extbe;
+        __r = __cv_->unshift(*__st_, __extbuf,
+                                    __extbuf + sizeof(__extbuf),
+                                    __extbe);
+        size_t __nmemb = static_cast<size_t>(__extbe - __extbuf);
+        if (fwrite(__extbuf, 1, __nmemb, __file_) != __nmemb)
+            return -1;
+    } while (__r == codecvt_base::partial);
+    if (__r == codecvt_base::error)
+        return -1;
+    if (fflush(__file_))
+        return -1;
+    return 0;
+}
+
+template <class _CharT>
+void
+__stdoutbuf<_CharT>::imbue(const locale& __loc)
+{
+    sync();
+    __cv_ = &use_facet<codecvt<char_type, char, state_type> >(__loc);
+    __always_noconv_ = __cv_->always_noconv();
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX___STD_STREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__string b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__string
new file mode 100644
index 000000000000..45dfbf6f7864
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__string
@@ -0,0 +1,985 @@
+// -*- C++ -*-
+//===-------------------------- __string ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___STRING
+#define _LIBCUDACXX___STRING
+
+/*
+    string synopsis
+
+namespace std
+{
+
+template <class charT>
+struct char_traits
+{
+    typedef charT     char_type;
+    typedef ...       int_type;
+    typedef streamoff off_type;
+    typedef streampos pos_type;
+    typedef mbstate_t state_type;
+
+    static constexpr void assign(char_type& c1, const char_type& c2) noexcept;
+    static constexpr bool eq(char_type c1, char_type c2) noexcept;
+    static constexpr bool lt(char_type c1, char_type c2) noexcept;
+
+    static constexpr int    compare(const char_type* s1, const char_type* s2, size_t n);
+    static constexpr size_t length(const char_type* s);
+    static constexpr const char_type* 
+                            find(const char_type* s, size_t n, const char_type& a);
+    static char_type*       move(char_type* s1, const char_type* s2, size_t n);
+    static char_type*       copy(char_type* s1, const char_type* s2, size_t n);
+    static char_type*       assign(char_type* s, size_t n, char_type a);
+
+    static constexpr int_type  not_eof(int_type c) noexcept;
+    static constexpr char_type to_char_type(int_type c) noexcept;
+    static constexpr int_type  to_int_type(char_type c) noexcept;
+    static constexpr bool      eq_int_type(int_type c1, int_type c2) noexcept;
+    static constexpr int_type  eof() noexcept;
+};
+
+template <> struct char_traits<char>;
+template <> struct char_traits<wchar_t>;
+template <> struct char_traits<char8_t>;  // c++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <algorithm>  // for search and min
+#include <cstdio>     // For EOF.
+#include <memory>     // for __murmur2_or_cityhash
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// char_traits
+
+template <class _CharT>
+struct _LIBCUDACXX_TEMPLATE_VIS char_traits
+{
+    typedef _CharT    char_type;
+    typedef int       int_type;
+    typedef streamoff off_type;
+    typedef streampos pos_type;
+    typedef mbstate_t state_type;
+
+    static inline void _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+        assign(char_type& __c1, const char_type& __c2) _NOEXCEPT {__c1 = __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool lt(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 < __c2;}
+
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    int compare(const char_type* __s1, const char_type* __s2, size_t __n);
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    size_t length(const char_type* __s);
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const char_type* find(const char_type* __s, size_t __n, const char_type& __a);
+    static char_type*       move(char_type* __s1, const char_type* __s2, size_t __n);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       copy(char_type* __s1, const char_type* __s2, size_t __n);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       assign(char_type* __s, size_t __n, char_type __a);
+
+    static inline _LIBCUDACXX_CONSTEXPR int_type  not_eof(int_type __c) _NOEXCEPT
+        {return eq_int_type(__c, eof()) ? ~eof() : __c;}
+    static inline _LIBCUDACXX_CONSTEXPR char_type to_char_type(int_type __c) _NOEXCEPT
+        {return char_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR int_type  to_int_type(char_type __c) _NOEXCEPT
+        {return int_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR bool      eq_int_type(int_type __c1, int_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR int_type  eof() _NOEXCEPT
+        {return int_type(EOF);}
+};
+
+template <class _CharT>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX14 int
+char_traits<_CharT>::compare(const char_type* __s1, const char_type* __s2, size_t __n)
+{
+    for (; __n; --__n, ++__s1, ++__s2)
+    {
+        if (lt(*__s1, *__s2))
+            return -1;
+        if (lt(*__s2, *__s1))
+            return 1;
+    }
+    return 0;
+}
+
+template <class _CharT>
+inline
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX14 size_t
+char_traits<_CharT>::length(const char_type* __s)
+{
+    size_t __len = 0;
+    for (; !eq(*__s, char_type(0)); ++__s)
+        ++__len;
+    return __len;
+}
+
+template <class _CharT>
+inline
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX14 const _CharT*
+char_traits<_CharT>::find(const char_type* __s, size_t __n, const char_type& __a)
+{
+    for (; __n; --__n)
+    {
+        if (eq(*__s, __a))
+            return __s;
+        ++__s;
+    }
+    return 0;
+}
+
+template <class _CharT>
+_CharT*
+char_traits<_CharT>::move(char_type* __s1, const char_type* __s2, size_t __n)
+{
+    char_type* __r = __s1;
+    if (__s1 < __s2)
+    {
+        for (; __n; --__n, ++__s1, ++__s2)
+            assign(*__s1, *__s2);
+    }
+    else if (__s2 < __s1)
+    {
+        __s1 += __n;
+        __s2 += __n;
+        for (; __n; --__n)
+            assign(*--__s1, *--__s2);
+    }
+    return __r;
+}
+
+template <class _CharT>
+inline
+_CharT*
+char_traits<_CharT>::copy(char_type* __s1, const char_type* __s2, size_t __n)
+{
+    _LIBCUDACXX_ASSERT(__s2 < __s1 || __s2 >= __s1+__n, "char_traits::copy overlapped range");
+    char_type* __r = __s1;
+    for (; __n; --__n, ++__s1, ++__s2)
+        assign(*__s1, *__s2);
+    return __r;
+}
+
+template <class _CharT>
+inline
+_CharT*
+char_traits<_CharT>::assign(char_type* __s, size_t __n, char_type __a)
+{
+    char_type* __r = __s;
+    for (; __n; --__n, ++__s)
+        assign(*__s, __a);
+    return __r;
+}
+
+// char_traits<char>
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS char_traits<char>
+{
+    typedef char      char_type;
+    typedef int       int_type;
+    typedef streamoff off_type;
+    typedef streampos pos_type;
+    typedef mbstate_t state_type;
+
+    static inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    void assign(char_type& __c1, const char_type& __c2) _NOEXCEPT {__c1 = __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq(char_type __c1, char_type __c2) _NOEXCEPT
+            {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool lt(char_type __c1, char_type __c2) _NOEXCEPT
+        {return (unsigned char)__c1 < (unsigned char)__c2;}
+
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    int compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    static inline size_t _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    length(const char_type* __s)  _NOEXCEPT {return __builtin_strlen(__s);}
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const char_type* find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT;
+    static inline char_type* move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+        {return __n == 0 ? __s1 : (char_type*) memmove(__s1, __s2, __n);}
+    static inline char_type* copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+        {
+            _LIBCUDACXX_ASSERT(__s2 < __s1 || __s2 >= __s1+__n, "char_traits::copy overlapped range");
+            return __n == 0 ? __s1 : (char_type*)memcpy(__s1, __s2, __n);
+        }
+    static inline char_type* assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT
+        {return __n == 0 ? __s : (char_type*)memset(__s, to_int_type(__a), __n);}
+
+    static inline _LIBCUDACXX_CONSTEXPR int_type  not_eof(int_type __c) _NOEXCEPT
+        {return eq_int_type(__c, eof()) ? ~eof() : __c;}
+    static inline _LIBCUDACXX_CONSTEXPR char_type to_char_type(int_type __c) _NOEXCEPT
+        {return char_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR int_type to_int_type(char_type __c) _NOEXCEPT
+        {return int_type((unsigned char)__c);}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq_int_type(int_type __c1, int_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR int_type  eof() _NOEXCEPT
+        {return int_type(EOF);}
+};
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+int
+char_traits<char>::compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    if (__n == 0)
+        return 0;
+#if __has_feature(cxx_constexpr_string_builtins)
+    return __builtin_memcmp(__s1, __s2, __n);
+#elif _LIBCUDACXX_STD_VER <= 14
+    return memcmp(__s1, __s2, __n);
+#else
+    for (; __n; --__n, ++__s1, ++__s2)
+    {
+        if (lt(*__s1, *__s2))
+            return -1;
+        if (lt(*__s2, *__s1))
+            return 1;
+    }
+    return 0;
+#endif
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+const char*
+char_traits<char>::find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT
+{
+    if (__n == 0)
+        return nullptr;
+#if __has_feature(cxx_constexpr_string_builtins)
+    return __builtin_char_memchr(__s, to_int_type(__a), __n);
+#elif _LIBCUDACXX_STD_VER <= 14
+    return (const char_type*) memchr(__s, to_int_type(__a), __n);
+#else
+    for (; __n; --__n)
+    {
+        if (eq(*__s, __a))
+            return __s;
+        ++__s;
+    }
+    return nullptr;
+#endif
+}
+
+
+// char_traits<wchar_t>
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS char_traits<wchar_t>
+{
+    typedef wchar_t   char_type;
+    typedef wint_t    int_type;
+    typedef streamoff off_type;
+    typedef streampos pos_type;
+    typedef mbstate_t state_type;
+
+    static inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    void assign(char_type& __c1, const char_type& __c2) _NOEXCEPT {__c1 = __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool lt(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 < __c2;}
+
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    int compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    size_t length(const char_type* __s) _NOEXCEPT;
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const char_type* find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT;
+    static inline char_type* move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+        {return __n == 0 ? __s1 : (char_type*)wmemmove(__s1, __s2, __n);}
+    static inline char_type* copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+        {
+            _LIBCUDACXX_ASSERT(__s2 < __s1 || __s2 >= __s1+__n, "char_traits::copy overlapped range");
+            return __n == 0 ? __s1 : (char_type*)wmemcpy(__s1, __s2, __n);
+        }
+    static inline char_type* assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT
+        {return __n == 0 ? __s : (char_type*)wmemset(__s, __a, __n);}
+
+    static inline _LIBCUDACXX_CONSTEXPR int_type  not_eof(int_type __c) _NOEXCEPT
+        {return eq_int_type(__c, eof()) ? ~eof() : __c;}
+    static inline _LIBCUDACXX_CONSTEXPR char_type to_char_type(int_type __c) _NOEXCEPT
+        {return char_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR int_type to_int_type(char_type __c) _NOEXCEPT
+        {return int_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq_int_type(int_type __c1, int_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR int_type eof() _NOEXCEPT
+        {return int_type(WEOF);}
+};
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+int
+char_traits<wchar_t>::compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    if (__n == 0)
+        return 0;
+#if __has_feature(cxx_constexpr_string_builtins)
+    return __builtin_wmemcmp(__s1, __s2, __n);
+#elif _LIBCUDACXX_STD_VER <= 14
+    return wmemcmp(__s1, __s2, __n);
+#else
+    for (; __n; --__n, ++__s1, ++__s2)
+    {
+        if (lt(*__s1, *__s2))
+            return -1;
+        if (lt(*__s2, *__s1))
+            return 1;
+    }
+    return 0;
+#endif
+}
+
+
+template <class _Traits>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+inline size_t __char_traits_length_checked(const typename _Traits::char_type* __s) _NOEXCEPT {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 1
+  return __s ? _Traits::length(__s) : (_CUDA_VSTD::__libcpp_debug_function(_CUDA_VSTD::__libcpp_debug_info(__FILE__, __LINE__, "p == nullptr", "null pointer pass to non-null argument of char_traits<...>::length")), 0);
+#else
+  return _Traits::length(__s);
+#endif
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+size_t
+char_traits<wchar_t>::length(const char_type* __s) _NOEXCEPT
+{
+#if __has_feature(cxx_constexpr_string_builtins)
+    return __builtin_wcslen(__s);
+#elif _LIBCUDACXX_STD_VER <= 14
+    return wcslen(__s);
+#else
+    size_t __len = 0;
+    for (; !eq(*__s, char_type(0)); ++__s)
+        ++__len;
+    return __len;
+#endif
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+const wchar_t*
+char_traits<wchar_t>::find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT
+{
+    if (__n == 0)
+        return nullptr;
+#if __has_feature(cxx_constexpr_string_builtins)
+    return __builtin_wmemchr(__s, __a, __n);
+#elif _LIBCUDACXX_STD_VER <= 14
+    return wmemchr(__s, __a, __n);
+#else
+    for (; __n; --__n)
+    {
+        if (eq(*__s, __a))
+            return __s;
+        ++__s;
+    }
+    return nullptr;
+#endif
+}
+
+
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS char_traits<char8_t>
+{
+    typedef char8_t        char_type;
+    typedef unsigned int   int_type;
+    typedef streamoff      off_type;
+    typedef u8streampos    pos_type;
+    typedef mbstate_t      state_type;
+
+    static inline constexpr void assign(char_type& __c1, const char_type& __c2) noexcept
+        {__c1 = __c2;}
+    static inline constexpr bool eq(char_type __c1, char_type __c2) noexcept
+        {return __c1 == __c2;}
+    static inline constexpr bool lt(char_type __c1, char_type __c2) noexcept
+        {return __c1 < __c2;}
+
+    static constexpr
+    int              compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+
+    static constexpr
+    size_t           length(const char_type* __s) _NOEXCEPT;
+ 
+    _LIBCUDACXX_INLINE_VISIBILITY static constexpr
+    const char_type* find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT;
+ 
+    static char_type*       move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+        {return __n == 0 ? __s1 : (char_type*) memmove(__s1, __s2, __n);}
+ 
+    static char_type*       copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+       {
+            _LIBCUDACXX_ASSERT(__s2 < __s1 || __s2 >= __s1+__n, "char_traits::copy overlapped range");
+            return __n == 0 ? __s1 : (char_type*)memcpy(__s1, __s2, __n);
+       }
+ 
+    static char_type*       assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT
+        {return __n == 0 ? __s : (char_type*)memset(__s, to_int_type(__a), __n);}
+
+    static inline constexpr int_type  not_eof(int_type __c) noexcept
+        {return eq_int_type(__c, eof()) ? ~eof() : __c;}
+    static inline constexpr char_type to_char_type(int_type __c) noexcept
+        {return char_type(__c);}
+    static inline constexpr int_type to_int_type(char_type __c) noexcept
+        {return int_type(__c);}
+    static inline constexpr bool eq_int_type(int_type __c1, int_type __c2) noexcept
+        {return __c1 == __c2;}
+    static inline constexpr int_type eof() noexcept
+        {return int_type(EOF);}
+};
+
+// TODO use '__builtin_strlen' if it ever supports char8_t ??
+inline constexpr
+size_t
+char_traits<char8_t>::length(const char_type* __s) _NOEXCEPT
+{
+    size_t __len = 0;
+    for (; !eq(*__s, char_type(0)); ++__s)
+        ++__len;
+    return __len;
+}
+
+inline constexpr
+int
+char_traits<char8_t>::compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+#if __has_feature(cxx_constexpr_string_builtins)
+    return __builtin_memcmp(__s1, __s2, __n);
+#else
+    for (; __n; --__n, ++__s1, ++__s2)
+    {
+        if (lt(*__s1, *__s2))
+            return -1;
+        if (lt(*__s2, *__s1))
+            return 1;
+    }
+    return 0;
+#endif
+}
+
+// TODO use '__builtin_char_memchr' if it ever supports char8_t ??
+inline constexpr
+const char8_t*
+char_traits<char8_t>::find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT
+{
+    for (; __n; --__n)
+    {
+        if (eq(*__s, __a))
+            return __s;
+        ++__s;
+    }
+    return 0;
+}
+
+#endif // #_LIBCUDACXX_NO_HAS_CHAR8_T
+
+#ifndef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS char_traits<char16_t>
+{
+    typedef char16_t       char_type;
+    typedef uint_least16_t int_type;
+    typedef streamoff      off_type;
+    typedef u16streampos   pos_type;
+    typedef mbstate_t      state_type;
+
+    static inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    void assign(char_type& __c1, const char_type& __c2) _NOEXCEPT {__c1 = __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool lt(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 < __c2;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    int              compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    size_t           length(const char_type* __s) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const char_type* find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT;
+
+    static inline _LIBCUDACXX_CONSTEXPR int_type  not_eof(int_type __c) _NOEXCEPT
+        {return eq_int_type(__c, eof()) ? ~eof() : __c;}
+    static inline _LIBCUDACXX_CONSTEXPR char_type to_char_type(int_type __c) _NOEXCEPT
+        {return char_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR int_type to_int_type(char_type __c) _NOEXCEPT
+        {return int_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq_int_type(int_type __c1, int_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR int_type eof() _NOEXCEPT
+        {return int_type(0xFFFF);}
+};
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+int
+char_traits<char16_t>::compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    for (; __n; --__n, ++__s1, ++__s2)
+    {
+        if (lt(*__s1, *__s2))
+            return -1;
+        if (lt(*__s2, *__s1))
+            return 1;
+    }
+    return 0;
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+size_t
+char_traits<char16_t>::length(const char_type* __s) _NOEXCEPT
+{
+    size_t __len = 0;
+    for (; !eq(*__s, char_type(0)); ++__s)
+        ++__len;
+    return __len;
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+const char16_t*
+char_traits<char16_t>::find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT
+{
+    for (; __n; --__n)
+    {
+        if (eq(*__s, __a))
+            return __s;
+        ++__s;
+    }
+    return 0;
+}
+
+inline
+char16_t*
+char_traits<char16_t>::move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    char_type* __r = __s1;
+    if (__s1 < __s2)
+    {
+        for (; __n; --__n, ++__s1, ++__s2)
+            assign(*__s1, *__s2);
+    }
+    else if (__s2 < __s1)
+    {
+        __s1 += __n;
+        __s2 += __n;
+        for (; __n; --__n)
+            assign(*--__s1, *--__s2);
+    }
+    return __r;
+}
+
+inline
+char16_t*
+char_traits<char16_t>::copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s2 < __s1 || __s2 >= __s1+__n, "char_traits::copy overlapped range");
+    char_type* __r = __s1;
+    for (; __n; --__n, ++__s1, ++__s2)
+        assign(*__s1, *__s2);
+    return __r;
+}
+
+inline
+char16_t*
+char_traits<char16_t>::assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT
+{
+    char_type* __r = __s;
+    for (; __n; --__n, ++__s)
+        assign(*__s, __a);
+    return __r;
+}
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS char_traits<char32_t>
+{
+    typedef char32_t       char_type;
+    typedef uint_least32_t int_type;
+    typedef streamoff      off_type;
+    typedef u32streampos   pos_type;
+    typedef mbstate_t      state_type;
+
+    static inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    void assign(char_type& __c1, const char_type& __c2) _NOEXCEPT {__c1 = __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR bool lt(char_type __c1, char_type __c2) _NOEXCEPT
+        {return __c1 < __c2;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    int              compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    size_t           length(const char_type* __s) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const char_type* find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char_type*       assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT;
+
+    static inline _LIBCUDACXX_CONSTEXPR int_type  not_eof(int_type __c) _NOEXCEPT
+        {return eq_int_type(__c, eof()) ? ~eof() : __c;}
+    static inline _LIBCUDACXX_CONSTEXPR char_type to_char_type(int_type __c) _NOEXCEPT
+        {return char_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR int_type to_int_type(char_type __c) _NOEXCEPT
+        {return int_type(__c);}
+    static inline _LIBCUDACXX_CONSTEXPR bool eq_int_type(int_type __c1, int_type __c2) _NOEXCEPT
+        {return __c1 == __c2;}
+    static inline _LIBCUDACXX_CONSTEXPR int_type eof() _NOEXCEPT
+        {return int_type(0xFFFFFFFF);}
+};
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+int
+char_traits<char32_t>::compare(const char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    for (; __n; --__n, ++__s1, ++__s2)
+    {
+        if (lt(*__s1, *__s2))
+            return -1;
+        if (lt(*__s2, *__s1))
+            return 1;
+    }
+    return 0;
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+size_t
+char_traits<char32_t>::length(const char_type* __s) _NOEXCEPT
+{
+    size_t __len = 0;
+    for (; !eq(*__s, char_type(0)); ++__s)
+        ++__len;
+    return __len;
+}
+
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+const char32_t*
+char_traits<char32_t>::find(const char_type* __s, size_t __n, const char_type& __a) _NOEXCEPT
+{
+    for (; __n; --__n)
+    {
+        if (eq(*__s, __a))
+            return __s;
+        ++__s;
+    }
+    return 0;
+}
+
+inline
+char32_t*
+char_traits<char32_t>::move(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    char_type* __r = __s1;
+    if (__s1 < __s2)
+    {
+        for (; __n; --__n, ++__s1, ++__s2)
+            assign(*__s1, *__s2);
+    }
+    else if (__s2 < __s1)
+    {
+        __s1 += __n;
+        __s2 += __n;
+        for (; __n; --__n)
+            assign(*--__s1, *--__s2);
+    }
+    return __r;
+}
+
+inline
+char32_t*
+char_traits<char32_t>::copy(char_type* __s1, const char_type* __s2, size_t __n) _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s2 < __s1 || __s2 >= __s1+__n, "char_traits::copy overlapped range");
+    char_type* __r = __s1;
+    for (; __n; --__n, ++__s1, ++__s2)
+        assign(*__s1, *__s2);
+    return __r;
+}
+
+inline
+char32_t*
+char_traits<char32_t>::assign(char_type* __s, size_t __n, char_type __a) _NOEXCEPT
+{
+    char_type* __r = __s;
+    for (; __n; --__n, ++__s)
+        assign(*__s, __a);
+    return __r;
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+
+// helper fns for basic_string and string_view
+
+// __str_find
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find(const _CharT *__p, _SizeT __sz, 
+             _CharT __c, _SizeT __pos) _NOEXCEPT
+{
+    if (__pos >= __sz)
+        return __npos;
+    const _CharT* __r = _Traits::find(__p + __pos, __sz - __pos, __c);
+    if (__r == 0)
+        return __npos;
+    return static_cast<_SizeT>(__r - __p);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const _CharT *
+__search_substring(const _CharT *__first1, const _CharT *__last1,
+                   const _CharT *__first2, const _CharT *__last2) {
+  // Take advantage of knowing source and pattern lengths.
+  // Stop short when source is smaller than pattern.
+  const ptrdiff_t __len2 = __last2 - __first2;
+  if (__len2 == 0)
+    return __first1;
+
+  ptrdiff_t __len1 = __last1 - __first1;
+  if (__len1 < __len2)
+    return __last1;
+
+  // First element of __first2 is loop invariant.
+  _CharT __f2 = *__first2;
+  while (true) {
+    __len1 = __last1 - __first1;
+    // Check whether __first1 still has at least __len2 bytes.
+    if (__len1 < __len2)
+      return __last1;
+
+    // Find __f2 the first byte matching in __first1.
+    __first1 = _Traits::find(__first1, __len1 - __len2 + 1, __f2);
+    if (__first1 == 0)
+      return __last1;
+
+    // It is faster to compare from the first byte of __first1 even if we
+    // already know that it matches the first byte of __first2: this is because
+    // __first2 is most likely aligned, as it is user's "pattern" string, and
+    // __first1 + 1 is most likely not aligned, as the match is in the middle of
+    // the string.
+    if (_Traits::compare(__first1, __first2, __len2) == 0)
+      return __first1;
+
+    ++__first1;
+  }
+}
+
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find(const _CharT *__p, _SizeT __sz, 
+       const _CharT* __s, _SizeT __pos, _SizeT __n) _NOEXCEPT
+{
+    if (__pos > __sz)
+        return __npos;
+
+    if (__n == 0) // There is nothing to search, just return __pos.
+        return __pos;
+
+    const _CharT *__r = __search_substring<_CharT, _Traits>(
+        __p + __pos, __p + __sz, __s, __s + __n);
+
+    if (__r == __p + __sz)
+        return __npos;
+    return static_cast<_SizeT>(__r - __p);
+}
+
+
+// __str_rfind
+
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_rfind(const _CharT *__p, _SizeT __sz, 
+              _CharT __c, _SizeT __pos) _NOEXCEPT
+{
+    if (__sz < 1)
+        return __npos;
+    if (__pos < __sz)
+        ++__pos;
+    else
+        __pos = __sz;
+    for (const _CharT* __ps = __p + __pos; __ps != __p;)
+    {
+        if (_Traits::eq(*--__ps, __c))
+            return static_cast<_SizeT>(__ps - __p);
+    }
+    return __npos;
+}
+
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_rfind(const _CharT *__p, _SizeT __sz, 
+        const _CharT* __s, _SizeT __pos, _SizeT __n) _NOEXCEPT
+{
+    __pos = _CUDA_VSTD::min(__pos, __sz);
+    if (__n < __sz - __pos)
+        __pos += __n;
+    else
+        __pos = __sz;
+    const _CharT* __r = _CUDA_VSTD::__find_end(
+                  __p, __p + __pos, __s, __s + __n, _Traits::eq, 
+                        random_access_iterator_tag(), random_access_iterator_tag());
+    if (__n > 0 && __r == __p + __pos)
+        return __npos;
+    return static_cast<_SizeT>(__r - __p);
+}
+
+// __str_find_first_of
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find_first_of(const _CharT *__p, _SizeT __sz,
+                const _CharT* __s, _SizeT __pos, _SizeT __n) _NOEXCEPT
+{
+    if (__pos >= __sz || __n == 0)
+        return __npos;
+    const _CharT* __r = _CUDA_VSTD::__find_first_of_ce
+        (__p + __pos, __p + __sz, __s, __s + __n, _Traits::eq );
+    if (__r == __p + __sz)
+        return __npos;
+    return static_cast<_SizeT>(__r - __p);
+}
+
+
+// __str_find_last_of
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find_last_of(const _CharT *__p, _SizeT __sz,
+               const _CharT* __s, _SizeT __pos, _SizeT __n) _NOEXCEPT
+    {
+    if (__n != 0)
+    {
+        if (__pos < __sz)
+            ++__pos;
+        else
+            __pos = __sz;
+        for (const _CharT* __ps = __p + __pos; __ps != __p;)
+        {
+            const _CharT* __r = _Traits::find(__s, __n, *--__ps);
+            if (__r)
+                return static_cast<_SizeT>(__ps - __p);
+        }
+    }
+    return __npos;
+}
+
+
+// __str_find_first_not_of
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find_first_not_of(const _CharT *__p, _SizeT __sz,
+                    const _CharT* __s, _SizeT __pos, _SizeT __n) _NOEXCEPT
+{
+    if (__pos < __sz)
+    {
+        const _CharT* __pe = __p + __sz;
+        for (const _CharT* __ps = __p + __pos; __ps != __pe; ++__ps)
+            if (_Traits::find(__s, __n, *__ps) == 0)
+                return static_cast<_SizeT>(__ps - __p);
+    }
+    return __npos;
+}
+
+
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find_first_not_of(const _CharT *__p, _SizeT __sz,
+                          _CharT __c, _SizeT __pos) _NOEXCEPT
+{
+    if (__pos < __sz)
+    {
+        const _CharT* __pe = __p + __sz;
+        for (const _CharT* __ps = __p + __pos; __ps != __pe; ++__ps)
+            if (!_Traits::eq(*__ps, __c))
+                return static_cast<_SizeT>(__ps - __p);
+    }
+    return __npos;
+}
+
+
+// __str_find_last_not_of
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find_last_not_of(const _CharT *__p, _SizeT __sz,
+                   const _CharT* __s, _SizeT __pos, _SizeT __n) _NOEXCEPT
+{
+    if (__pos < __sz)
+        ++__pos;
+    else
+        __pos = __sz;
+    for (const _CharT* __ps = __p + __pos; __ps != __p;)
+        if (_Traits::find(__s, __n, *--__ps) == 0)
+            return static_cast<_SizeT>(__ps - __p);
+    return __npos;
+}
+
+
+template<class _CharT, class _SizeT, class _Traits, _SizeT __npos>
+inline _SizeT _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+__str_find_last_not_of(const _CharT *__p, _SizeT __sz,
+                         _CharT __c, _SizeT __pos) _NOEXCEPT
+{
+    if (__pos < __sz)
+        ++__pos;
+    else
+        __pos = __sz;
+    for (const _CharT* __ps = __p + __pos; __ps != __p;)
+        if (!_Traits::eq(*--__ps, __c))
+            return static_cast<_SizeT>(__ps - __p);
+    return __npos;
+}
+
+template<class _Ptr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __do_string_hash(_Ptr __p, _Ptr __e)
+{
+    typedef typename iterator_traits<_Ptr>::value_type value_type;
+    return __murmur2_or_cityhash<size_t>()(__p, (__e-__p)*sizeof(value_type));
+}
+
+template <class _CharT, class _Iter, class _Traits=char_traits<_CharT> >
+struct __quoted_output_proxy
+{
+    _Iter  __first;
+    _Iter  __last;
+    _CharT  __delim;
+    _CharT  __escape;
+
+    __quoted_output_proxy(_Iter __f, _Iter __l, _CharT __d, _CharT __e)
+    : __first(__f), __last(__l), __delim(__d), __escape(__e) {}
+    //  This would be a nice place for a string_ref 
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX___STRING
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support
new file mode 100644
index 000000000000..9489f6a6bde0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__threading_support
@@ -0,0 +1,787 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_THREADING_SUPPORT
+#define _LIBCUDACXX_THREADING_SUPPORT
+
+#ifndef __cuda_std__
+#include <__config>
+#include <chrono>
+#include <iosfwd>
+#include <errno.h>
+#include <climits>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+# ifndef __cuda_std__
+#  include <__external_threading>
+# else
+#  define _LIBCUDACXX_THREAD_ABI_VISIBILITY inline _LIBCUDACXX_INLINE_VISIBILITY
+# endif
+#elif !defined(_LIBCUDACXX_HAS_NO_THREADS)
+
+#if defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD)
+# include <pthread.h>
+# include <sched.h>
+# include <semaphore.h>
+# if defined(__APPLE__)
+#  include <dispatch/dispatch.h>
+# endif
+# if defined(__linux__)
+#  include <unistd.h>
+#  include <linux/futex.h>
+#  include <sys/syscall.h>
+# endif
+#endif
+
+#if defined(_LIBCUDACXX_HAS_THREAD_API_WIN32)
+# include <process.h>
+# include <windows.h>
+#endif
+
+#if defined(_LIBCUDACXX_HAS_THREAD_LIBRARY_EXTERNAL) || \
+    defined(_LIBCUDACXX_BUILDING_THREAD_LIBRARY_EXTERNAL)
+#define _LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_FUNC_VIS
+#else
+#define _LIBCUDACXX_THREAD_ABI_VISIBILITY inline _LIBCUDACXX_INLINE_VISIBILITY
+#endif
+
+#if defined(__FreeBSD__) && defined(__clang__) && __has_attribute(no_thread_safety_analysis)
+#define _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS __attribute__((no_thread_safety_analysis))
+#else
+#define _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+#endif
+
+typedef ::timespec __libcpp_timespec_t;
+#endif // !defined(_LIBCUDACXX_HAS_NO_THREADS)
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if !defined(_LIBCUDACXX_HAS_NO_THREADS)
+
+#define _LIBCUDACXX_POLLING_COUNT 16
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline void __libcpp_thread_yield_processor()
+{
+#if defined(__aarch64__)
+# define __LIBCUDACXX_ASM_THREAD_YIELD (asm volatile ("yield" :::);)
+#elif defined(__x86_64__)
+# define __LIBCUDACXX_ASM_THREAD_YIELD (asm volatile ("pause" :::);)
+#elif defined (__powerpc__)
+# define __LIBCUDACXX_ASM_THREAD_YIELD (asm volatile ("or 27,27,27":::);)
+#else
+# define __LIBCUDACXX_ASM_THREAD_YIELD (;)
+#endif
+    NV_IF_TARGET(
+        NV_IS_HOST,
+            __LIBCUDACXX_ASM_THREAD_YIELD
+    )
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void __libcpp_thread_yield();
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void __libcpp_thread_sleep_for(chrono::nanoseconds __ns);
+
+template<class _Fn>
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_poll_with_backoff(_Fn && __f, chrono::nanoseconds __max = chrono::nanoseconds::zero());
+
+#if defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD)
+// Mutex
+typedef pthread_mutex_t __libcpp_mutex_t;
+#define _LIBCUDACXX_MUTEX_INITIALIZER PTHREAD_MUTEX_INITIALIZER
+
+typedef pthread_mutex_t __libcpp_recursive_mutex_t;
+
+// Condition Variable
+typedef pthread_cond_t __libcpp_condvar_t;
+#define _LIBCUDACXX_CONDVAR_INITIALIZER PTHREAD_COND_INITIALIZER
+
+// Semaphore
+#if defined(__APPLE__)
+typedef dispatch_semaphore_t __libcpp_semaphore_t;
+# define _LIBCUDACXX_SEMAPHORE_MAX numeric_limits<long>::max()
+#else
+typedef sem_t __libcpp_semaphore_t;
+# define _LIBCUDACXX_SEMAPHORE_MAX SEM_VALUE_MAX
+#endif
+
+// Execute once
+typedef pthread_once_t __libcpp_exec_once_flag;
+#define _LIBCUDACXX_EXEC_ONCE_INITIALIZER PTHREAD_ONCE_INIT
+
+// Thread id
+typedef pthread_t __libcpp_thread_id;
+
+// Thread
+#define _LIBCUDACXX_NULL_THREAD 0U
+
+typedef pthread_t __libcpp_thread_t;
+
+// Thread Local Storage
+typedef pthread_key_t __libcpp_tls_key;
+
+#define _LIBCUDACXX_TLS_DESTRUCTOR_CC
+#elif !defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+// Mutex
+typedef void* __libcpp_mutex_t;
+#define _LIBCUDACXX_MUTEX_INITIALIZER 0
+
+#if defined(_M_IX86) || defined(__i386__) || defined(_M_ARM) || defined(__arm__)
+typedef void* __libcpp_recursive_mutex_t[6];
+#elif defined(_M_AMD64) || defined(__x86_64__) || defined(_M_ARM64) || defined(__aarch64__)
+typedef void* __libcpp_recursive_mutex_t[5];
+#else
+# error Unsupported architecture
+#endif
+
+// Condition Variable
+typedef void* __libcpp_condvar_t;
+#define _LIBCUDACXX_CONDVAR_INITIALIZER 0
+
+// Semaphore
+typedef void* __libcpp_semaphore_t;
+
+// Execute Once
+typedef void* __libcpp_exec_once_flag;
+#define _LIBCUDACXX_EXEC_ONCE_INITIALIZER 0
+
+// Thread ID
+typedef long __libcpp_thread_id;
+
+// Thread
+#define _LIBCUDACXX_NULL_THREAD 0U
+
+typedef void* __libcpp_thread_t;
+
+// Thread Local Storage
+typedef long __libcpp_tls_key;
+
+#define _LIBCUDACXX_TLS_DESTRUCTOR_CC __stdcall
+#endif // !defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD) && !defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+
+#if !defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+__libcpp_timespec_t __libcpp_to_timespec(const chrono::nanoseconds& __ns);
+
+// Mutex
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_recursive_mutex_init(__libcpp_recursive_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+int __libcpp_recursive_mutex_lock(__libcpp_recursive_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+bool __libcpp_recursive_mutex_trylock(__libcpp_recursive_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+int __libcpp_recursive_mutex_unlock(__libcpp_recursive_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_recursive_mutex_destroy(__libcpp_recursive_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+int __libcpp_mutex_lock(__libcpp_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+bool __libcpp_mutex_trylock(__libcpp_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+int __libcpp_mutex_unlock(__libcpp_mutex_t *__m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_mutex_destroy(__libcpp_mutex_t *__m);
+
+// Condition variable
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_signal(__libcpp_condvar_t* __cv);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_broadcast(__libcpp_condvar_t* __cv);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+int __libcpp_condvar_wait(__libcpp_condvar_t* __cv, __libcpp_mutex_t* __m);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY _LIBCUDACXX_NO_THREAD_SAFETY_ANALYSIS
+int __libcpp_condvar_timedwait(__libcpp_condvar_t *__cv, __libcpp_mutex_t *__m,
+                               __libcpp_timespec_t *__ts);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_destroy(__libcpp_condvar_t* __cv);
+
+// Semaphore
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_init(__libcpp_semaphore_t* __sem, int __init);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_destroy(__libcpp_semaphore_t* __sem);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_post(__libcpp_semaphore_t* __sem);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_wait(__libcpp_semaphore_t* __sem);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_wait_timed(__libcpp_semaphore_t* __sem, chrono::nanoseconds const& __ns);
+
+// Execute once
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_execute_once(__libcpp_exec_once_flag *flag,
+                          void (*init_routine)());
+
+// Thread id
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_id_equal(__libcpp_thread_id t1, __libcpp_thread_id t2);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_id_less(__libcpp_thread_id t1, __libcpp_thread_id t2);
+
+// Thread
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_isnull(const __libcpp_thread_t *__t);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_thread_create(__libcpp_thread_t *__t, void *(*__func)(void *),
+                           void *__arg);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+__libcpp_thread_id __libcpp_thread_get_current_id();
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+__libcpp_thread_id __libcpp_thread_get_id(const __libcpp_thread_t *__t);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_thread_join(__libcpp_thread_t *__t);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_thread_detach(__libcpp_thread_t *__t);
+
+// Thread local storage
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_tls_create(__libcpp_tls_key* __key,
+                        void(_LIBCUDACXX_TLS_DESTRUCTOR_CC* __at_exit)(void*));
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void *__libcpp_tls_get(__libcpp_tls_key __key);
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_tls_set(__libcpp_tls_key __key, void *__p);
+
+#endif // !defined(_LIBCUDACXX_HAS_THREAD_API_EXTERNAL)
+
+#if !defined(_LIBCUDACXX_HAS_THREAD_LIBRARY_EXTERNAL) || defined(_LIBCUDACXX_BUILDING_THREAD_LIBRARY_EXTERNAL)
+
+#if defined(_LIBCUDACXX_HAS_THREAD_API_CUDA)
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void __libcpp_thread_yield() {}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void __libcpp_thread_sleep_for(chrono::nanoseconds __ns)
+{
+    NV_IF_TARGET(
+        NV_IS_DEVICE, (
+            auto const __step = __ns.count();
+            assert(__step < numeric_limits<unsigned>::max());
+            asm volatile("nanosleep.u32 %0;"::"r"((unsigned)__step):);
+        )
+    )
+}
+
+#elif defined(_LIBCUDACXX_HAS_THREAD_API_PTHREAD)
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+__libcpp_timespec_t __libcpp_to_timespec(const chrono::nanoseconds& __ns)
+{
+     using namespace chrono;
+     seconds __s = duration_cast<seconds>(__ns);
+     __libcpp_timespec_t __ts;
+     typedef decltype(__ts.tv_sec) ts_sec;
+     _LIBCUDACXX_CONSTEXPR ts_sec __ts_sec_max = numeric_limits<ts_sec>::max();
+
+     if (__s.count() < __ts_sec_max)
+     {
+         __ts.tv_sec = static_cast<ts_sec>(__s.count());
+         __ts.tv_nsec = static_cast<decltype(__ts.tv_nsec)>((__ns - __s).count());
+     }
+     else
+     {
+         __ts.tv_sec = __ts_sec_max;
+         __ts.tv_nsec = 999999999; // (10^9 - 1)
+     }
+     return __ts;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_recursive_mutex_init(__libcpp_recursive_mutex_t *__m)
+{
+    pthread_mutexattr_t attr;
+    int __ec = pthread_mutexattr_init(&attr);
+    if (__ec)
+        return __ec;
+    __ec = pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE);
+    if (__ec) {
+        pthread_mutexattr_destroy(&attr);
+        return __ec;
+    }
+    __ec = pthread_mutex_init(__m, &attr);
+    if (__ec) {
+        pthread_mutexattr_destroy(&attr);
+        return __ec;
+    }
+    __ec = pthread_mutexattr_destroy(&attr);
+    if (__ec) {
+        pthread_mutex_destroy(__m);
+        return __ec;
+    }
+    return 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_recursive_mutex_lock(__libcpp_recursive_mutex_t *__m)
+{
+    return pthread_mutex_lock(__m);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_recursive_mutex_trylock(__libcpp_recursive_mutex_t *__m)
+{
+    return pthread_mutex_trylock(__m) == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_recursive_mutex_unlock(__libcpp_mutex_t *__m)
+{
+    return pthread_mutex_unlock(__m);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_recursive_mutex_destroy(__libcpp_recursive_mutex_t *__m)
+{
+    return pthread_mutex_destroy(__m);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_mutex_lock(__libcpp_mutex_t *__m)
+{
+    return pthread_mutex_lock(__m);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_mutex_trylock(__libcpp_mutex_t *__m)
+{
+    return pthread_mutex_trylock(__m) == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_mutex_unlock(__libcpp_mutex_t *__m)
+{
+    return pthread_mutex_unlock(__m);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_mutex_destroy(__libcpp_mutex_t *__m)
+{
+  return pthread_mutex_destroy(__m);
+}
+
+// Condition Variable
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_signal(__libcpp_condvar_t *__cv)
+{
+    return pthread_cond_signal(__cv);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_broadcast(__libcpp_condvar_t *__cv)
+{
+    return pthread_cond_broadcast(__cv);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_wait(__libcpp_condvar_t *__cv, __libcpp_mutex_t *__m)
+{
+    return pthread_cond_wait(__cv, __m);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_timedwait(__libcpp_condvar_t *__cv, __libcpp_mutex_t *__m,
+                               __libcpp_timespec_t *__ts)
+{
+    return pthread_cond_timedwait(__cv, __m, __ts);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_condvar_destroy(__libcpp_condvar_t *__cv)
+{
+    return pthread_cond_destroy(__cv);
+}
+
+// Semaphore
+#if defined(__APPLE__)
+
+bool __libcpp_semaphore_init(__libcpp_semaphore_t* __sem, int __init)
+{
+    return (*__sem = dispatch_semaphore_create(__init)) != NULL;
+}
+
+bool __libcpp_semaphore_destroy(__libcpp_semaphore_t* __sem)
+{
+    dispatch_release(*__sem);
+    return true;
+}
+
+bool __libcpp_semaphore_post(__libcpp_semaphore_t* __sem)
+{
+    dispatch_semaphore_signal(*__sem);
+    return true;
+}
+
+bool __libcpp_semaphore_wait(__libcpp_semaphore_t* __sem)
+{
+    return dispatch_semaphore_wait(*__sem, DISPATCH_TIME_FOREVER) == 0;
+}
+
+bool __libcpp_semaphore_wait_timed(__libcpp_semaphore_t* __sem, chrono::nanoseconds const& __ns)
+{
+    return dispatch_semaphore_wait(*__sem, dispatch_time(DISPATCH_TIME_NOW, __ns.count())) == 0;
+}
+
+#else
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_init(__libcpp_semaphore_t* __sem, int __init)
+{
+    return sem_init(__sem, 0, __init) == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_destroy(__libcpp_semaphore_t* __sem)
+{
+    return sem_destroy(__sem) == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_post(__libcpp_semaphore_t* __sem)
+{
+    return sem_post(__sem) == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_wait(__libcpp_semaphore_t* __sem)
+{
+    return sem_wait(__sem) == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_semaphore_wait_timed(__libcpp_semaphore_t* __sem, chrono::nanoseconds const& __ns)
+{
+    __libcpp_timespec_t __ts = __libcpp_to_timespec(__ns);
+    return sem_timedwait(__sem, &__ts) == 0;
+}
+
+#endif //__APPLE__
+
+// Execute once
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_execute_once(__libcpp_exec_once_flag *flag, void (*init_routine)())
+{
+    return pthread_once(flag, init_routine);
+}
+
+// Thread id
+// Returns non-zero if the thread ids are equal, otherwise 0
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_id_equal(__libcpp_thread_id t1, __libcpp_thread_id t2)
+{
+    return pthread_equal(t1, t2) != 0;
+}
+
+// Returns non-zero if t1 < t2, otherwise 0
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_id_less(__libcpp_thread_id t1, __libcpp_thread_id t2)
+{
+    return t1 < t2;
+}
+
+// Thread
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_isnull(const __libcpp_thread_t *__t)
+{
+    return *__t == 0;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_thread_create(__libcpp_thread_t *__t, void *(*__func)(void *),
+                           void *__arg)
+{
+    return pthread_create(__t, 0, __func, __arg);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+__libcpp_thread_id __libcpp_thread_get_current_id()
+{
+    return pthread_self();
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+__libcpp_thread_id __libcpp_thread_get_id(const __libcpp_thread_t *__t)
+{
+    return *__t;
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_thread_join(__libcpp_thread_t *__t)
+{
+    return pthread_join(*__t, 0);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_thread_detach(__libcpp_thread_t *__t)
+{
+    return pthread_detach(*__t);
+}
+
+// Thread local storage
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_tls_create(__libcpp_tls_key *__key, void (*__at_exit)(void *))
+{
+    return pthread_key_create(__key, __at_exit);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void *__libcpp_tls_get(__libcpp_tls_key __key)
+{
+    return pthread_getspecific(__key);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+int __libcpp_tls_set(__libcpp_tls_key __key, void *__p)
+{
+    return pthread_setspecific(__key, __p);
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void __libcpp_thread_yield()
+{
+    sched_yield();
+}
+
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+void __libcpp_thread_sleep_for(chrono::nanoseconds __ns)
+{
+    __libcpp_timespec_t __ts = __libcpp_to_timespec(__ns);
+    while (nanosleep(&__ts, &__ts) == -1 && errno == EINTR);
+}
+
+#if defined(__linux__) && !defined(_LIBCUDACXX_HAS_NO_PLATFORM_WAIT)
+
+#define _LIBCUDACXX_HAS_PLATFORM_WAIT
+
+typedef int __libcpp_platform_wait_t;
+
+template<typename _Tp>
+struct __libcpp_platform_wait_uses_type {
+    enum { __value = is_same<typename remove_cv<_Tp>::type, __libcpp_platform_wait_t>::value };
+};
+
+template <class _Tp, typename enable_if<__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+void __libcpp_platform_wait(_Tp const* ptr, _Tp val, void const* timeout) {
+    syscall(SYS_futex, ptr, FUTEX_WAIT_PRIVATE, val, timeout, 0, 0);
+}
+
+template <class _Tp, typename enable_if<__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+void __libcpp_platform_wake(_Tp const* ptr, bool all) {
+    syscall(SYS_futex, ptr, FUTEX_WAKE_PRIVATE, all ? INT_MAX : 1, 0, 0, 0);
+}
+
+#endif // defined(__linux__) && !defined(_LIBCUDACXX_HAS_NO_PLATFORM_WAIT)
+
+#elif defined(_LIBCUDACXX_HAS_THREAD_API_WIN32)
+
+void __libcpp_thread_yield()
+{
+    SwitchToThread();
+}
+
+void __libcpp_thread_sleep_for(chrono::nanoseconds __ns)
+{
+    using namespace chrono;
+    // round-up to the nearest milisecond
+    milliseconds __ms =
+        duration_cast<milliseconds>(__ns + chrono::nanoseconds(999999));
+    Sleep(static_cast<DWORD>(__ms.count()));
+}
+
+#endif // defined(_LIBCUDACXX_HAS_THREAD_API_WIN32)
+
+#endif // !defined(_LIBCUDACXX_HAS_THREAD_LIBRARY_EXTERNAL) || defined(_LIBCUDACXX_BUILDING_THREAD_LIBRARY_EXTERNAL)
+
+template<class _Fn>
+_LIBCUDACXX_THREAD_ABI_VISIBILITY
+bool __libcpp_thread_poll_with_backoff(_Fn && __f, chrono::nanoseconds __max)
+{
+    chrono::high_resolution_clock::time_point const __start = chrono::high_resolution_clock::now();
+    for(int __count = 0;;) {
+      if(__f())
+        return true;
+      if(__count < _LIBCUDACXX_POLLING_COUNT) {
+        if(__count > (_LIBCUDACXX_POLLING_COUNT >> 1))
+          __libcpp_thread_yield_processor();
+        __count += 1;
+        continue;
+      }
+      chrono::high_resolution_clock::duration const __elapsed = chrono::high_resolution_clock::now() - __start;
+      if(__max != chrono::nanoseconds::zero() &&
+         __max < __elapsed)
+         return false;
+      chrono::nanoseconds const __step = __elapsed / 4;
+      if(__step >= chrono::milliseconds(1))
+        __libcpp_thread_sleep_for(chrono::milliseconds(1));
+      else if(__step >= chrono::microseconds(10))
+        __libcpp_thread_sleep_for(__step);
+      else
+        __libcpp_thread_yield();
+    }
+}
+
+#if _LIBCUDACXX_STD_VER < 11
+# define _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+
+struct alignas(64) __libcpp_contention_t {
+#if defined(_LIBCUDACXX_HAS_PLATFORM_WAIT)
+    ptrdiff_t                __waiters = 0;
+    __libcpp_platform_wait_t __version = 0;
+#else
+    ptrdiff_t                __credit = 0;
+    __libcpp_mutex_t         __mutex = _LIBCUDACXX_MUTEX_INITIALIZER;
+    __libcpp_condvar_t       __condvar = _LIBCUDACXX_CONDVAR_INITIALIZER;
+#endif
+};
+
+_LIBCUDACXX_FUNC_VIS
+__libcpp_contention_t * __libcpp_contention_state(void const volatile * p) _NOEXCEPT;
+
+#endif // _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+
+#if !defined(_LIBCUDACXX_HAS_NO_TREE_BARRIER) && !defined(_LIBCUDACXX_HAS_NO_THREAD_FAVORITE_BARRIER_INDEX)
+
+_LIBCUDACXX_EXPORTED_FROM_ABI
+extern thread_local ptrdiff_t __libcpp_thread_favorite_barrier_index;
+
+#endif
+
+#ifndef __cuda_std__
+
+class _LIBCUDACXX_TYPE_VIS thread;
+class _LIBCUDACXX_TYPE_VIS __thread_id;
+
+namespace this_thread
+{
+
+_LIBCUDACXX_INLINE_VISIBILITY __thread_id get_id() _NOEXCEPT;
+
+}  // this_thread
+
+template<> struct hash<__thread_id>;
+
+class _LIBCUDACXX_TEMPLATE_VIS __thread_id
+{
+    // FIXME: pthread_t is a pointer on Darwin but a long on Linux.
+    // NULL is the no-thread value on Darwin.  Someone needs to check
+    // on other platforms.  We assume 0 works everywhere for now.
+    __libcpp_thread_id __id_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __thread_id() _NOEXCEPT : __id_(0) {}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(__thread_id __x, __thread_id __y) _NOEXCEPT
+        { // don't pass id==0 to underlying routines
+        if (__x.__id_ == 0) return __y.__id_ == 0;
+        if (__y.__id_ == 0) return false;
+        return __libcpp_thread_id_equal(__x.__id_, __y.__id_);
+        }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(__thread_id __x, __thread_id __y) _NOEXCEPT
+        {return !(__x == __y);}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator< (__thread_id __x, __thread_id __y) _NOEXCEPT
+        { // id==0 is always less than any other thread_id
+        if (__x.__id_ == 0) return __y.__id_ != 0;
+        if (__y.__id_ == 0) return false;
+        return  __libcpp_thread_id_less(__x.__id_, __y.__id_);
+        }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator<=(__thread_id __x, __thread_id __y) _NOEXCEPT
+        {return !(__y < __x);}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator> (__thread_id __x, __thread_id __y) _NOEXCEPT
+        {return   __y < __x ;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator>=(__thread_id __x, __thread_id __y) _NOEXCEPT
+        {return !(__x < __y);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __reset() { __id_ = 0; }
+
+#ifndef __cuda_std__
+    template<class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, __thread_id __id);
+#endif
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __thread_id(__libcpp_thread_id __id) : __id_(__id) {}
+
+    friend __thread_id this_thread::get_id() _NOEXCEPT;
+    friend class _LIBCUDACXX_TYPE_VIS thread;
+    friend struct _LIBCUDACXX_TEMPLATE_VIS hash<__thread_id>;
+};
+
+namespace this_thread
+{
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__thread_id
+get_id() _NOEXCEPT
+{
+    return __libcpp_thread_get_current_id();
+}
+
+}  // this_thread
+
+#endif // __cuda_std__
+
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif // _LIBCUDACXX_THREADING_SUPPORT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tree b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tree
new file mode 100644
index 000000000000..14922a74cef0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tree
@@ -0,0 +1,2844 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___TREE
+#define _LIBCUDACXX___TREE
+
+#include <__config>
+#include <iterator>
+#include <memory>
+#include <stdexcept>
+#include <algorithm>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if defined(__GNUC__) && !defined(__clang__) // gcc.gnu.org/PR37804
+template <class, class, class, class> class _LIBCUDACXX_TEMPLATE_VIS map;
+template <class, class, class, class> class _LIBCUDACXX_TEMPLATE_VIS multimap;
+template <class, class, class> class _LIBCUDACXX_TEMPLATE_VIS set;
+template <class, class, class> class _LIBCUDACXX_TEMPLATE_VIS multiset;
+#endif
+
+template <class _Tp, class _Compare, class _Allocator> class __tree;
+template <class _Tp, class _NodePtr, class _DiffType>
+    class _LIBCUDACXX_TEMPLATE_VIS __tree_iterator;
+template <class _Tp, class _ConstNodePtr, class _DiffType>
+    class _LIBCUDACXX_TEMPLATE_VIS __tree_const_iterator;
+
+template <class _Pointer> class __tree_end_node;
+template <class _VoidPtr> class __tree_node_base;
+template <class _Tp, class _VoidPtr> class __tree_node;
+
+template <class _Key, class _Value>
+struct __value_type;
+
+template <class _Allocator> class __map_node_destructor;
+template <class _TreeIterator> class _LIBCUDACXX_TEMPLATE_VIS __map_iterator;
+template <class _TreeIterator> class _LIBCUDACXX_TEMPLATE_VIS __map_const_iterator;
+
+/*
+
+_NodePtr algorithms
+
+The algorithms taking _NodePtr are red black tree algorithms.  Those
+algorithms taking a parameter named __root should assume that __root
+points to a proper red black tree (unless otherwise specified).
+
+Each algorithm herein assumes that __root->__parent_ points to a non-null
+structure which has a member __left_ which points back to __root.  No other
+member is read or written to at __root->__parent_.
+
+__root->__parent_ will be referred to below (in comments only) as end_node.
+end_node->__left_ is an externably accessible lvalue for __root, and can be
+changed by node insertion and removal (without explicit reference to end_node).
+
+All nodes (with the exception of end_node), even the node referred to as
+__root, have a non-null __parent_ field.
+
+*/
+
+// Returns:  true if __x is a left child of its parent, else false
+// Precondition:  __x != nullptr.
+template <class _NodePtr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+__tree_is_left_child(_NodePtr __x) _NOEXCEPT
+{
+    return __x == __x->__parent_->__left_;
+}
+
+// Determines if the subtree rooted at __x is a proper red black subtree.  If
+//    __x is a proper subtree, returns the black height (null counts as 1).  If
+//    __x is an improper subtree, returns 0.
+template <class _NodePtr>
+unsigned
+__tree_sub_invariant(_NodePtr __x)
+{
+    if (__x == nullptr)
+        return 1;
+    // parent consistency checked by caller
+    // check __x->__left_ consistency
+    if (__x->__left_ != nullptr && __x->__left_->__parent_ != __x)
+        return 0;
+    // check __x->__right_ consistency
+    if (__x->__right_ != nullptr && __x->__right_->__parent_ != __x)
+        return 0;
+    // check __x->__left_ != __x->__right_ unless both are nullptr
+    if (__x->__left_ == __x->__right_ && __x->__left_ != nullptr)
+        return 0;
+    // If this is red, neither child can be red
+    if (!__x->__is_black_)
+    {
+        if (__x->__left_ && !__x->__left_->__is_black_)
+            return 0;
+        if (__x->__right_ && !__x->__right_->__is_black_)
+            return 0;
+    }
+    unsigned __h = __tree_sub_invariant(__x->__left_);
+    if (__h == 0)
+        return 0;  // invalid left subtree
+    if (__h != __tree_sub_invariant(__x->__right_))
+        return 0;  // invalid or different height right subtree
+    return __h + __x->__is_black_;  // return black height of this node
+}
+
+// Determines if the red black tree rooted at __root is a proper red black tree.
+//    __root == nullptr is a proper tree.  Returns true is __root is a proper
+//    red black tree, else returns false.
+template <class _NodePtr>
+bool
+__tree_invariant(_NodePtr __root)
+{
+    if (__root == nullptr)
+        return true;
+    // check __x->__parent_ consistency
+    if (__root->__parent_ == nullptr)
+        return false;
+    if (!__tree_is_left_child(__root))
+        return false;
+    // root must be black
+    if (!__root->__is_black_)
+        return false;
+    // do normal node checks
+    return __tree_sub_invariant(__root) != 0;
+}
+
+// Returns:  pointer to the left-most node under __x.
+// Precondition:  __x != nullptr.
+template <class _NodePtr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_NodePtr
+__tree_min(_NodePtr __x) _NOEXCEPT
+{
+    while (__x->__left_ != nullptr)
+        __x = __x->__left_;
+    return __x;
+}
+
+// Returns:  pointer to the right-most node under __x.
+// Precondition:  __x != nullptr.
+template <class _NodePtr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_NodePtr
+__tree_max(_NodePtr __x) _NOEXCEPT
+{
+    while (__x->__right_ != nullptr)
+        __x = __x->__right_;
+    return __x;
+}
+
+// Returns:  pointer to the next in-order node after __x.
+// Precondition:  __x != nullptr.
+template <class _NodePtr>
+_NodePtr
+__tree_next(_NodePtr __x) _NOEXCEPT
+{
+    if (__x->__right_ != nullptr)
+        return __tree_min(__x->__right_);
+    while (!__tree_is_left_child(__x))
+        __x = __x->__parent_unsafe();
+    return __x->__parent_unsafe();
+}
+
+template <class _EndNodePtr, class _NodePtr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_EndNodePtr
+__tree_next_iter(_NodePtr __x) _NOEXCEPT
+{
+    if (__x->__right_ != nullptr)
+        return static_cast<_EndNodePtr>(__tree_min(__x->__right_));
+    while (!__tree_is_left_child(__x))
+        __x = __x->__parent_unsafe();
+    return static_cast<_EndNodePtr>(__x->__parent_);
+}
+
+// Returns:  pointer to the previous in-order node before __x.
+// Precondition:  __x != nullptr.
+// Note: __x may be the end node.
+template <class _NodePtr, class _EndNodePtr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_NodePtr
+__tree_prev_iter(_EndNodePtr __x) _NOEXCEPT
+{
+    if (__x->__left_ != nullptr)
+        return __tree_max(__x->__left_);
+    _NodePtr __xx = static_cast<_NodePtr>(__x);
+    while (__tree_is_left_child(__xx))
+        __xx = __xx->__parent_unsafe();
+    return __xx->__parent_unsafe();
+}
+
+// Returns:  pointer to a node which has no children
+// Precondition:  __x != nullptr.
+template <class _NodePtr>
+_NodePtr
+__tree_leaf(_NodePtr __x) _NOEXCEPT
+{
+    while (true)
+    {
+        if (__x->__left_ != nullptr)
+        {
+            __x = __x->__left_;
+            continue;
+        }
+        if (__x->__right_ != nullptr)
+        {
+            __x = __x->__right_;
+            continue;
+        }
+        break;
+    }
+    return __x;
+}
+
+// Effects:  Makes __x->__right_ the subtree root with __x as its left child
+//           while preserving in-order order.
+// Precondition:  __x->__right_ != nullptr
+template <class _NodePtr>
+void
+__tree_left_rotate(_NodePtr __x) _NOEXCEPT
+{
+    _NodePtr __y = __x->__right_;
+    __x->__right_ = __y->__left_;
+    if (__x->__right_ != nullptr)
+        __x->__right_->__set_parent(__x);
+    __y->__parent_ = __x->__parent_;
+    if (__tree_is_left_child(__x))
+        __x->__parent_->__left_ = __y;
+    else
+        __x->__parent_unsafe()->__right_ = __y;
+    __y->__left_ = __x;
+    __x->__set_parent(__y);
+}
+
+// Effects:  Makes __x->__left_ the subtree root with __x as its right child
+//           while preserving in-order order.
+// Precondition:  __x->__left_ != nullptr
+template <class _NodePtr>
+void
+__tree_right_rotate(_NodePtr __x) _NOEXCEPT
+{
+    _NodePtr __y = __x->__left_;
+    __x->__left_ = __y->__right_;
+    if (__x->__left_ != nullptr)
+        __x->__left_->__set_parent(__x);
+    __y->__parent_ = __x->__parent_;
+    if (__tree_is_left_child(__x))
+        __x->__parent_->__left_ = __y;
+    else
+        __x->__parent_unsafe()->__right_ = __y;
+    __y->__right_ = __x;
+    __x->__set_parent(__y);
+}
+
+// Effects:  Rebalances __root after attaching __x to a leaf.
+// Precondition:  __root != nulptr && __x != nullptr.
+//                __x has no children.
+//                __x == __root or == a direct or indirect child of __root.
+//                If __x were to be unlinked from __root (setting __root to
+//                  nullptr if __root == __x), __tree_invariant(__root) == true.
+// Postcondition: __tree_invariant(end_node->__left_) == true.  end_node->__left_
+//                may be different than the value passed in as __root.
+template <class _NodePtr>
+void
+__tree_balance_after_insert(_NodePtr __root, _NodePtr __x) _NOEXCEPT
+{
+    __x->__is_black_ = __x == __root;
+    while (__x != __root && !__x->__parent_unsafe()->__is_black_)
+    {
+        // __x->__parent_ != __root because __x->__parent_->__is_black == false
+        if (__tree_is_left_child(__x->__parent_unsafe()))
+        {
+            _NodePtr __y = __x->__parent_unsafe()->__parent_unsafe()->__right_;
+            if (__y != nullptr && !__y->__is_black_)
+            {
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = true;
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = __x == __root;
+                __y->__is_black_ = true;
+            }
+            else
+            {
+                if (!__tree_is_left_child(__x))
+                {
+                    __x = __x->__parent_unsafe();
+                    __tree_left_rotate(__x);
+                }
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = true;
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = false;
+                __tree_right_rotate(__x);
+                break;
+            }
+        }
+        else
+        {
+            _NodePtr __y = __x->__parent_unsafe()->__parent_->__left_;
+            if (__y != nullptr && !__y->__is_black_)
+            {
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = true;
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = __x == __root;
+                __y->__is_black_ = true;
+            }
+            else
+            {
+                if (__tree_is_left_child(__x))
+                {
+                    __x = __x->__parent_unsafe();
+                    __tree_right_rotate(__x);
+                }
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = true;
+                __x = __x->__parent_unsafe();
+                __x->__is_black_ = false;
+                __tree_left_rotate(__x);
+                break;
+            }
+        }
+    }
+}
+
+// Precondition:  __root != nullptr && __z != nullptr.
+//                __tree_invariant(__root) == true.
+//                __z == __root or == a direct or indirect child of __root.
+// Effects:  unlinks __z from the tree rooted at __root, rebalancing as needed.
+// Postcondition: __tree_invariant(end_node->__left_) == true && end_node->__left_
+//                nor any of its children refer to __z.  end_node->__left_
+//                may be different than the value passed in as __root.
+template <class _NodePtr>
+void
+__tree_remove(_NodePtr __root, _NodePtr __z) _NOEXCEPT
+{
+    // __z will be removed from the tree.  Client still needs to destruct/deallocate it
+    // __y is either __z, or if __z has two children, __tree_next(__z).
+    // __y will have at most one child.
+    // __y will be the initial hole in the tree (make the hole at a leaf)
+    _NodePtr __y = (__z->__left_ == nullptr || __z->__right_ == nullptr) ?
+                    __z : __tree_next(__z);
+    // __x is __y's possibly null single child
+    _NodePtr __x = __y->__left_ != nullptr ? __y->__left_ : __y->__right_;
+    // __w is __x's possibly null uncle (will become __x's sibling)
+    _NodePtr __w = nullptr;
+    // link __x to __y's parent, and find __w
+    if (__x != nullptr)
+        __x->__parent_ = __y->__parent_;
+    if (__tree_is_left_child(__y))
+    {
+        __y->__parent_->__left_ = __x;
+        if (__y != __root)
+            __w = __y->__parent_unsafe()->__right_;
+        else
+            __root = __x;  // __w == nullptr
+    }
+    else
+    {
+        __y->__parent_unsafe()->__right_ = __x;
+        // __y can't be root if it is a right child
+        __w = __y->__parent_->__left_;
+    }
+    bool __removed_black = __y->__is_black_;
+    // If we didn't remove __z, do so now by splicing in __y for __z,
+    //    but copy __z's color.  This does not impact __x or __w.
+    if (__y != __z)
+    {
+        // __z->__left_ != nulptr but __z->__right_ might == __x == nullptr
+        __y->__parent_ = __z->__parent_;
+        if (__tree_is_left_child(__z))
+            __y->__parent_->__left_ = __y;
+        else
+            __y->__parent_unsafe()->__right_ = __y;
+        __y->__left_ = __z->__left_;
+        __y->__left_->__set_parent(__y);
+        __y->__right_ = __z->__right_;
+        if (__y->__right_ != nullptr)
+            __y->__right_->__set_parent(__y);
+        __y->__is_black_ = __z->__is_black_;
+        if (__root == __z)
+            __root = __y;
+    }
+    // There is no need to rebalance if we removed a red, or if we removed
+    //     the last node.
+    if (__removed_black && __root != nullptr)
+    {
+        // Rebalance:
+        // __x has an implicit black color (transferred from the removed __y)
+        //    associated with it, no matter what its color is.
+        // If __x is __root (in which case it can't be null), it is supposed
+        //    to be black anyway, and if it is doubly black, then the double
+        //    can just be ignored.
+        // If __x is red (in which case it can't be null), then it can absorb
+        //    the implicit black just by setting its color to black.
+        // Since __y was black and only had one child (which __x points to), __x
+        //   is either red with no children, else null, otherwise __y would have
+        //   different black heights under left and right pointers.
+        // if (__x == __root || __x != nullptr && !__x->__is_black_)
+        if (__x != nullptr)
+            __x->__is_black_ = true;
+        else
+        {
+            //  Else __x isn't root, and is "doubly black", even though it may
+            //     be null.  __w can not be null here, else the parent would
+            //     see a black height >= 2 on the __x side and a black height
+            //     of 1 on the __w side (__w must be a non-null black or a red
+            //     with a non-null black child).
+            while (true)
+            {
+                if (!__tree_is_left_child(__w))  // if x is left child
+                {
+                    if (!__w->__is_black_)
+                    {
+                        __w->__is_black_ = true;
+                        __w->__parent_unsafe()->__is_black_ = false;
+                        __tree_left_rotate(__w->__parent_unsafe());
+                        // __x is still valid
+                        // reset __root only if necessary
+                        if (__root == __w->__left_)
+                            __root = __w;
+                        // reset sibling, and it still can't be null
+                        __w = __w->__left_->__right_;
+                    }
+                    // __w->__is_black_ is now true, __w may have null children
+                    if ((__w->__left_  == nullptr || __w->__left_->__is_black_) &&
+                        (__w->__right_ == nullptr || __w->__right_->__is_black_))
+                    {
+                        __w->__is_black_ = false;
+                        __x = __w->__parent_unsafe();
+                        // __x can no longer be null
+                        if (__x == __root || !__x->__is_black_)
+                        {
+                            __x->__is_black_ = true;
+                            break;
+                        }
+                        // reset sibling, and it still can't be null
+                        __w = __tree_is_left_child(__x) ?
+                                    __x->__parent_unsafe()->__right_ :
+                                    __x->__parent_->__left_;
+                        // continue;
+                    }
+                    else  // __w has a red child
+                    {
+                        if (__w->__right_ == nullptr || __w->__right_->__is_black_)
+                        {
+                            // __w left child is non-null and red
+                            __w->__left_->__is_black_ = true;
+                            __w->__is_black_ = false;
+                            __tree_right_rotate(__w);
+                            // __w is known not to be root, so root hasn't changed
+                            // reset sibling, and it still can't be null
+                            __w = __w->__parent_unsafe();
+                        }
+                        // __w has a right red child, left child may be null
+                        __w->__is_black_ = __w->__parent_unsafe()->__is_black_;
+                        __w->__parent_unsafe()->__is_black_ = true;
+                        __w->__right_->__is_black_ = true;
+                        __tree_left_rotate(__w->__parent_unsafe());
+                        break;
+                    }
+                }
+                else
+                {
+                    if (!__w->__is_black_)
+                    {
+                        __w->__is_black_ = true;
+                        __w->__parent_unsafe()->__is_black_ = false;
+                        __tree_right_rotate(__w->__parent_unsafe());
+                        // __x is still valid
+                        // reset __root only if necessary
+                        if (__root == __w->__right_)
+                            __root = __w;
+                        // reset sibling, and it still can't be null
+                        __w = __w->__right_->__left_;
+                    }
+                    // __w->__is_black_ is now true, __w may have null children
+                    if ((__w->__left_  == nullptr || __w->__left_->__is_black_) &&
+                        (__w->__right_ == nullptr || __w->__right_->__is_black_))
+                    {
+                        __w->__is_black_ = false;
+                        __x = __w->__parent_unsafe();
+                        // __x can no longer be null
+                        if (!__x->__is_black_ || __x == __root)
+                        {
+                            __x->__is_black_ = true;
+                            break;
+                        }
+                        // reset sibling, and it still can't be null
+                        __w = __tree_is_left_child(__x) ?
+                                    __x->__parent_unsafe()->__right_ :
+                                    __x->__parent_->__left_;
+                        // continue;
+                    }
+                    else  // __w has a red child
+                    {
+                        if (__w->__left_ == nullptr || __w->__left_->__is_black_)
+                        {
+                            // __w right child is non-null and red
+                            __w->__right_->__is_black_ = true;
+                            __w->__is_black_ = false;
+                            __tree_left_rotate(__w);
+                            // __w is known not to be root, so root hasn't changed
+                            // reset sibling, and it still can't be null
+                            __w = __w->__parent_unsafe();
+                        }
+                        // __w has a left red child, right child may be null
+                        __w->__is_black_ = __w->__parent_unsafe()->__is_black_;
+                        __w->__parent_unsafe()->__is_black_ = true;
+                        __w->__left_->__is_black_ = true;
+                        __tree_right_rotate(__w->__parent_unsafe());
+                        break;
+                    }
+                }
+            }
+        }
+    }
+}
+
+// node traits
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp>
+struct __is_tree_value_type_imp : false_type {};
+
+template <class _Key, class _Value>
+struct __is_tree_value_type_imp<__value_type<_Key, _Value>> : true_type {};
+
+template <class ..._Args>
+struct __is_tree_value_type : false_type {};
+
+template <class _One>
+struct __is_tree_value_type<_One> : __is_tree_value_type_imp<typename __uncvref<_One>::type> {};
+#endif
+
+template <class _Tp>
+struct __tree_key_value_types {
+  typedef _Tp key_type;
+  typedef _Tp __node_value_type;
+  typedef _Tp __container_value_type;
+  static const bool __is_map = false;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static key_type const& __get_key(_Tp const& __v) {
+    return __v;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type const& __get_value(__node_value_type const& __v) {
+    return __v;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type* __get_ptr(__node_value_type& __n) {
+    return _CUDA_VSTD::addressof(__n);
+  }
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type&& __move(__node_value_type& __v) {
+    return _CUDA_VSTD::move(__v);
+  }
+#endif
+};
+
+template <class _Key, class _Tp>
+struct __tree_key_value_types<__value_type<_Key, _Tp> > {
+  typedef _Key                                         key_type;
+  typedef _Tp                                          mapped_type;
+  typedef __value_type<_Key, _Tp>                      __node_value_type;
+  typedef pair<const _Key, _Tp>                        __container_value_type;
+  typedef __container_value_type                       __map_value_type;
+  static const bool __is_map = true;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static key_type const&
+  __get_key(__node_value_type const& __t) {
+    return __t.__get_value().first;
+  }
+
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static typename enable_if<__is_same_uncvref<_Up, __container_value_type>::value,
+      key_type const&>::type
+  __get_key(_Up& __t) {
+    return __t.first;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type const&
+  __get_value(__node_value_type const& __t) {
+    return __t.__get_value();
+  }
+
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static typename enable_if<__is_same_uncvref<_Up, __container_value_type>::value,
+      __container_value_type const&>::type
+  __get_value(_Up& __t) {
+    return __t;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __container_value_type* __get_ptr(__node_value_type& __n) {
+    return _CUDA_VSTD::addressof(__n.__get_value());
+  }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static pair<key_type&&, mapped_type&&> __move(__node_value_type& __v) {
+    return __v.__move();
+  }
+#endif
+};
+
+template <class _VoidPtr>
+struct __tree_node_base_types {
+  typedef _VoidPtr                                               __void_pointer;
+
+  typedef __tree_node_base<__void_pointer>                      __node_base_type;
+  typedef typename __rebind_pointer<_VoidPtr, __node_base_type>::type
+                                                             __node_base_pointer;
+
+  typedef __tree_end_node<__node_base_pointer>                  __end_node_type;
+  typedef typename __rebind_pointer<_VoidPtr, __end_node_type>::type
+                                                             __end_node_pointer;
+#if defined(_LIBCUDACXX_ABI_TREE_REMOVE_NODE_POINTER_UB)
+  typedef __end_node_pointer __parent_pointer;
+#else
+  typedef typename conditional<
+      is_pointer<__end_node_pointer>::value,
+        __end_node_pointer,
+        __node_base_pointer>::type __parent_pointer;
+#endif
+
+private:
+  static_assert((is_same<typename pointer_traits<_VoidPtr>::element_type, void>::value),
+                  "_VoidPtr does not point to unqualified void type");
+};
+
+template <class _Tp, class _AllocPtr, class _KVTypes = __tree_key_value_types<_Tp>,
+         bool = _KVTypes::__is_map>
+struct __tree_map_pointer_types {};
+
+template <class _Tp, class _AllocPtr, class _KVTypes>
+struct __tree_map_pointer_types<_Tp, _AllocPtr, _KVTypes, true> {
+  typedef typename _KVTypes::__map_value_type   _Mv;
+  typedef typename __rebind_pointer<_AllocPtr, _Mv>::type
+                                                       __map_value_type_pointer;
+  typedef typename __rebind_pointer<_AllocPtr, const _Mv>::type
+                                                 __const_map_value_type_pointer;
+};
+
+template <class _NodePtr, class _NodeT = typename pointer_traits<_NodePtr>::element_type>
+struct __tree_node_types;
+
+template <class _NodePtr, class _Tp, class _VoidPtr>
+struct __tree_node_types<_NodePtr, __tree_node<_Tp, _VoidPtr> >
+    : public __tree_node_base_types<_VoidPtr>,
+             __tree_key_value_types<_Tp>,
+             __tree_map_pointer_types<_Tp, _VoidPtr>
+{
+  typedef __tree_node_base_types<_VoidPtr> __base;
+  typedef __tree_key_value_types<_Tp>      __key_base;
+  typedef __tree_map_pointer_types<_Tp, _VoidPtr> __map_pointer_base;
+public:
+
+  typedef typename pointer_traits<_NodePtr>::element_type       __node_type;
+  typedef _NodePtr                                              __node_pointer;
+
+  typedef _Tp                                                 __node_value_type;
+  typedef typename __rebind_pointer<_VoidPtr, __node_value_type>::type
+                                                      __node_value_type_pointer;
+  typedef typename __rebind_pointer<_VoidPtr, const __node_value_type>::type
+                                                __const_node_value_type_pointer;
+#if defined(_LIBCUDACXX_ABI_TREE_REMOVE_NODE_POINTER_UB)
+  typedef typename __base::__end_node_pointer __iter_pointer;
+#else
+  typedef typename conditional<
+      is_pointer<__node_pointer>::value,
+        typename __base::__end_node_pointer,
+        __node_pointer>::type __iter_pointer;
+#endif
+private:
+    static_assert(!is_const<__node_type>::value,
+                "_NodePtr should never be a pointer to const");
+    static_assert((is_same<typename __rebind_pointer<_VoidPtr, __node_type>::type,
+                          _NodePtr>::value), "_VoidPtr does not rebind to _NodePtr.");
+};
+
+template <class _ValueTp, class _VoidPtr>
+struct __make_tree_node_types {
+  typedef typename __rebind_pointer<_VoidPtr, __tree_node<_ValueTp, _VoidPtr> >::type
+                                                                        _NodePtr;
+  typedef __tree_node_types<_NodePtr> type;
+};
+
+// node
+
+template <class _Pointer>
+class __tree_end_node
+{
+public:
+    typedef _Pointer pointer;
+    pointer __left_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_end_node() _NOEXCEPT : __left_() {}
+};
+
+template <class _VoidPtr>
+class __tree_node_base
+    : public __tree_node_base_types<_VoidPtr>::__end_node_type
+{
+    typedef __tree_node_base_types<_VoidPtr> _NodeBaseTypes;
+
+public:
+    typedef typename _NodeBaseTypes::__node_base_pointer pointer;
+    typedef typename _NodeBaseTypes::__parent_pointer __parent_pointer;
+
+    pointer          __right_;
+    __parent_pointer __parent_;
+    bool __is_black_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer __parent_unsafe() const { return static_cast<pointer>(__parent_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_parent(pointer __p) {
+        __parent_ = static_cast<__parent_pointer>(__p);
+    }
+
+private:
+  ~__tree_node_base() _LIBCUDACXX_EQUAL_DELETE;
+  __tree_node_base(__tree_node_base const&) _LIBCUDACXX_EQUAL_DELETE;
+  __tree_node_base& operator=(__tree_node_base const&) _LIBCUDACXX_EQUAL_DELETE;
+};
+
+template <class _Tp, class _VoidPtr>
+class __tree_node
+    : public __tree_node_base<_VoidPtr>
+{
+public:
+    typedef _Tp __node_value_type;
+
+    __node_value_type __value_;
+
+private:
+  ~__tree_node() _LIBCUDACXX_EQUAL_DELETE;
+  __tree_node(__tree_node const&) _LIBCUDACXX_EQUAL_DELETE;
+  __tree_node& operator=(__tree_node const&) _LIBCUDACXX_EQUAL_DELETE;
+};
+
+
+template <class _Allocator>
+class __tree_node_destructor
+{
+    typedef _Allocator                                      allocator_type;
+    typedef allocator_traits<allocator_type>                __alloc_traits;
+
+public:
+    typedef typename __alloc_traits::pointer                pointer;
+private:
+    typedef __tree_node_types<pointer> _NodeTypes;
+    allocator_type& __na_;
+
+public:
+    bool __value_constructed;
+
+    __tree_node_destructor(const __tree_node_destructor &) = default;
+    __tree_node_destructor& operator=(const __tree_node_destructor&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __tree_node_destructor(allocator_type& __na, bool __val = false) _NOEXCEPT
+        : __na_(__na),
+          __value_constructed(__val)
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p) _NOEXCEPT
+    {
+        if (__value_constructed)
+            __alloc_traits::destroy(__na_, _NodeTypes::__get_ptr(__p->__value_));
+        if (__p)
+            __alloc_traits::deallocate(__na_, __p, 1);
+    }
+
+    template <class> friend class __map_node_destructor;
+};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _NodeType, class _Alloc>
+struct __generic_container_node_destructor;
+template <class _Tp, class _VoidPtr, class _Alloc>
+struct __generic_container_node_destructor<__tree_node<_Tp, _VoidPtr>, _Alloc>
+    : __tree_node_destructor<_Alloc>
+{
+    using __tree_node_destructor<_Alloc>::__tree_node_destructor;
+};
+#endif
+
+template <class _Tp, class _NodePtr, class _DiffType>
+class _LIBCUDACXX_TEMPLATE_VIS __tree_iterator
+{
+    typedef __tree_node_types<_NodePtr>                     _NodeTypes;
+    typedef _NodePtr                                        __node_pointer;
+    typedef typename _NodeTypes::__node_base_pointer        __node_base_pointer;
+    typedef typename _NodeTypes::__end_node_pointer         __end_node_pointer;
+    typedef typename _NodeTypes::__iter_pointer             __iter_pointer;
+    typedef pointer_traits<__node_pointer> __pointer_traits;
+
+    __iter_pointer __ptr_;
+
+public:
+    typedef bidirectional_iterator_tag                     iterator_category;
+    typedef _Tp                                            value_type;
+    typedef _DiffType                                      difference_type;
+    typedef value_type&                                    reference;
+    typedef typename _NodeTypes::__node_value_type_pointer pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __tree_iterator() _NOEXCEPT
+#if _LIBCUDACXX_STD_VER > 11
+    : __ptr_(nullptr)
+#endif
+    {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator*() const
+        {return __get_np()->__value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY pointer operator->() const
+        {return pointer_traits<pointer>::pointer_to(__get_np()->__value_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_iterator& operator++() {
+      __ptr_ = static_cast<__iter_pointer>(
+          __tree_next_iter<__end_node_pointer>(static_cast<__node_base_pointer>(__ptr_)));
+      return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_iterator operator++(int)
+        {__tree_iterator __t(*this); ++(*this); return __t;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_iterator& operator--() {
+      __ptr_ = static_cast<__iter_pointer>(__tree_prev_iter<__node_base_pointer>(
+          static_cast<__end_node_pointer>(__ptr_)));
+      return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_iterator operator--(int)
+        {__tree_iterator __t(*this); --(*this); return __t;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const __tree_iterator& __x, const __tree_iterator& __y)
+        {return __x.__ptr_ == __y.__ptr_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const __tree_iterator& __x, const __tree_iterator& __y)
+        {return !(__x == __y);}
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __tree_iterator(__node_pointer __p) _NOEXCEPT : __ptr_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __tree_iterator(__end_node_pointer __p) _NOEXCEPT : __ptr_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_pointer __get_np() const { return static_cast<__node_pointer>(__ptr_); }
+    template <class, class, class> friend class __tree;
+    template <class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS __tree_const_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __map_iterator;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+    template <class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS set;
+    template <class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multiset;
+};
+
+template <class _Tp, class _NodePtr, class _DiffType>
+class _LIBCUDACXX_TEMPLATE_VIS __tree_const_iterator
+{
+    typedef __tree_node_types<_NodePtr>                     _NodeTypes;
+    typedef typename _NodeTypes::__node_pointer             __node_pointer;
+    typedef typename _NodeTypes::__node_base_pointer        __node_base_pointer;
+    typedef typename _NodeTypes::__end_node_pointer         __end_node_pointer;
+    typedef typename _NodeTypes::__iter_pointer             __iter_pointer;
+    typedef pointer_traits<__node_pointer> __pointer_traits;
+
+    __iter_pointer __ptr_;
+
+public:
+    typedef bidirectional_iterator_tag                           iterator_category;
+    typedef _Tp                                                  value_type;
+    typedef _DiffType                                            difference_type;
+    typedef const value_type&                                    reference;
+    typedef typename _NodeTypes::__const_node_value_type_pointer pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __tree_const_iterator() _NOEXCEPT
+#if _LIBCUDACXX_STD_VER > 11
+    : __ptr_(nullptr)
+#endif
+    {}
+
+private:
+    typedef __tree_iterator<value_type, __node_pointer, difference_type>
+                                                           __non_const_iterator;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_const_iterator(__non_const_iterator __p) _NOEXCEPT
+        : __ptr_(__p.__ptr_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator*() const
+        {return __get_np()->__value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY pointer operator->() const
+        {return pointer_traits<pointer>::pointer_to(__get_np()->__value_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_const_iterator& operator++() {
+      __ptr_ = static_cast<__iter_pointer>(
+          __tree_next_iter<__end_node_pointer>(static_cast<__node_base_pointer>(__ptr_)));
+      return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_const_iterator operator++(int)
+        {__tree_const_iterator __t(*this); ++(*this); return __t;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_const_iterator& operator--() {
+      __ptr_ = static_cast<__iter_pointer>(__tree_prev_iter<__node_base_pointer>(
+          static_cast<__end_node_pointer>(__ptr_)));
+      return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tree_const_iterator operator--(int)
+        {__tree_const_iterator __t(*this); --(*this); return __t;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const __tree_const_iterator& __x, const __tree_const_iterator& __y)
+        {return __x.__ptr_ == __y.__ptr_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const __tree_const_iterator& __x, const __tree_const_iterator& __y)
+        {return !(__x == __y);}
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __tree_const_iterator(__node_pointer __p) _NOEXCEPT
+        : __ptr_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __tree_const_iterator(__end_node_pointer __p) _NOEXCEPT
+        : __ptr_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_pointer __get_np() const { return static_cast<__node_pointer>(__ptr_); }
+
+    template <class, class, class> friend class __tree;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+    template <class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS set;
+    template <class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multiset;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __map_const_iterator;
+
+};
+
+template<class _Tp, class _Compare>
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_DIAGNOSE_WARNING(!std::__invokable<_Compare const&, _Tp const&, _Tp const&>::value,
+        "the specified comparator type does not provide a viable const call operator")
+#endif
+int __diagnose_non_const_comparator();
+
+template <class _Tp, class _Compare, class _Allocator>
+class __tree
+{
+public:
+    typedef _Tp                                      value_type;
+    typedef _Compare                                 value_compare;
+    typedef _Allocator                               allocator_type;
+
+private:
+    typedef allocator_traits<allocator_type>         __alloc_traits;
+    typedef typename __make_tree_node_types<value_type,
+        typename __alloc_traits::void_pointer>::type
+                                                    _NodeTypes;
+    typedef typename _NodeTypes::key_type           key_type;
+public:
+    typedef typename _NodeTypes::__node_value_type      __node_value_type;
+    typedef typename _NodeTypes::__container_value_type __container_value_type;
+
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+    typedef typename __alloc_traits::size_type       size_type;
+    typedef typename __alloc_traits::difference_type difference_type;
+
+public:
+    typedef typename _NodeTypes::__void_pointer        __void_pointer;
+
+    typedef typename _NodeTypes::__node_type           __node;
+    typedef typename _NodeTypes::__node_pointer        __node_pointer;
+
+    typedef typename _NodeTypes::__node_base_type      __node_base;
+    typedef typename _NodeTypes::__node_base_pointer   __node_base_pointer;
+
+    typedef typename _NodeTypes::__end_node_type       __end_node_t;
+    typedef typename _NodeTypes::__end_node_pointer    __end_node_ptr;
+
+    typedef typename _NodeTypes::__parent_pointer      __parent_pointer;
+    typedef typename _NodeTypes::__iter_pointer        __iter_pointer;
+
+    typedef typename __rebind_alloc_helper<__alloc_traits, __node>::type __node_allocator;
+    typedef allocator_traits<__node_allocator>         __node_traits;
+
+private:
+    // check for sane allocator pointer rebinding semantics. Rebinding the
+    // allocator for a new pointer type should be exactly the same as rebinding
+    // the pointer using 'pointer_traits'.
+    static_assert((is_same<__node_pointer, typename __node_traits::pointer>::value),
+                  "Allocator does not rebind pointers in a sane manner.");
+    typedef typename __rebind_alloc_helper<__node_traits, __node_base>::type
+        __node_base_allocator;
+    typedef allocator_traits<__node_base_allocator> __node_base_traits;
+    static_assert((is_same<__node_base_pointer, typename __node_base_traits::pointer>::value),
+                 "Allocator does not rebind pointers in a sane manner.");
+
+private:
+    __iter_pointer                                     __begin_node_;
+    __compressed_pair<__end_node_t, __node_allocator>  __pair1_;
+    __compressed_pair<size_type, value_compare>        __pair3_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __iter_pointer __end_node() _NOEXCEPT
+    {
+        return static_cast<__iter_pointer>(
+                pointer_traits<__end_node_ptr>::pointer_to(__pair1_.first())
+        );
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __iter_pointer __end_node() const _NOEXCEPT
+    {
+        return static_cast<__iter_pointer>(
+            pointer_traits<__end_node_ptr>::pointer_to(
+                const_cast<__end_node_t&>(__pair1_.first())
+            )
+        );
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+          __node_allocator& __node_alloc() _NOEXCEPT {return __pair1_.second();}
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const __node_allocator& __node_alloc() const _NOEXCEPT
+        {return __pair1_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          __iter_pointer& __begin_node() _NOEXCEPT {return __begin_node_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const __iter_pointer& __begin_node() const _NOEXCEPT {return __begin_node_;}
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type __alloc() const _NOEXCEPT
+        {return allocator_type(__node_alloc());}
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+          size_type& size() _NOEXCEPT {return __pair3_.first();}
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const size_type& size() const _NOEXCEPT {return __pair3_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          value_compare& value_comp() _NOEXCEPT {return __pair3_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_compare& value_comp() const _NOEXCEPT
+        {return __pair3_.second();}
+public:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_pointer __root() const _NOEXCEPT
+        {return static_cast<__node_pointer>(__end_node()->__left_);}
+
+    __node_base_pointer* __root_ptr() const _NOEXCEPT {
+        return _CUDA_VSTD::addressof(__end_node()->__left_);
+    }
+
+    typedef __tree_iterator<value_type, __node_pointer, difference_type>             iterator;
+    typedef __tree_const_iterator<value_type, __node_pointer, difference_type> const_iterator;
+
+    explicit __tree(const value_compare& __comp)
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<__node_allocator>::value &&
+            is_nothrow_copy_constructible<value_compare>::value);
+    explicit __tree(const allocator_type& __a);
+    __tree(const value_compare& __comp, const allocator_type& __a);
+    __tree(const __tree& __t);
+    __tree& operator=(const __tree& __t);
+    template <class _ForwardIterator>
+        void __assign_unique(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _InputIterator>
+        void __assign_multi(_InputIterator __first, _InputIterator __last);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __tree(__tree&& __t)
+        _NOEXCEPT_(
+            is_nothrow_move_constructible<__node_allocator>::value &&
+            is_nothrow_move_constructible<value_compare>::value);
+    __tree(__tree&& __t, const allocator_type& __a);
+    __tree& operator=(__tree&& __t)
+        _NOEXCEPT_(
+            __node_traits::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<value_compare>::value &&
+            is_nothrow_move_assignable<__node_allocator>::value);
+#endif // _LIBCUDACXX_CXX03_LANG
+
+    ~__tree();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator begin()  _NOEXCEPT {return       iterator(__begin_node());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return const_iterator(__begin_node());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator end() _NOEXCEPT {return       iterator(__end_node());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT {return const_iterator(__end_node());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT
+        {return std::min<size_type>(
+                __node_traits::max_size(__node_alloc()),
+                numeric_limits<difference_type >::max());}
+
+    void clear() _NOEXCEPT;
+
+    void swap(__tree& __t)
+#if _LIBCUDACXX_STD_VER <= 11
+        _NOEXCEPT_(
+            __is_nothrow_swappable<value_compare>::value
+            && (!__node_traits::propagate_on_container_swap::value ||
+                 __is_nothrow_swappable<__node_allocator>::value)
+            );
+#else
+        _NOEXCEPT_(__is_nothrow_swappable<value_compare>::value);
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Key, class ..._Args>
+    pair<iterator, bool>
+    __emplace_unique_key_args(_Key const&, _Args&&... __args);
+    template <class _Key, class ..._Args>
+    iterator
+    __emplace_hint_unique_key_args(const_iterator, _Key const&, _Args&&...);
+
+    template <class... _Args>
+    pair<iterator, bool> __emplace_unique_impl(_Args&&... __args);
+
+    template <class... _Args>
+    iterator __emplace_hint_unique_impl(const_iterator __p, _Args&&... __args);
+
+    template <class... _Args>
+    iterator __emplace_multi(_Args&&... __args);
+
+    template <class... _Args>
+    iterator __emplace_hint_multi(const_iterator __p, _Args&&... __args);
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique(_Pp&& __x) {
+        return __emplace_unique_extract_key(_CUDA_VSTD::forward<_Pp>(__x),
+                                            __can_extract_key<_Pp, key_type>());
+    }
+
+    template <class _First, class _Second>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<
+        __can_extract_map_key<_First, key_type, __container_value_type>::value,
+        pair<iterator, bool>
+    >::type __emplace_unique(_First&& __f, _Second&& __s) {
+        return __emplace_unique_key_args(__f, _CUDA_VSTD::forward<_First>(__f),
+                                              _CUDA_VSTD::forward<_Second>(__s));
+    }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique(_Args&&... __args) {
+        return __emplace_unique_impl(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __emplace_unique_extract_key(_Pp&& __x, __extract_key_fail_tag) {
+      return __emplace_unique_impl(_CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __emplace_unique_extract_key(_Pp&& __x, __extract_key_self_tag) {
+      return __emplace_unique_key_args(__x, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    __emplace_unique_extract_key(_Pp&& __x, __extract_key_first_tag) {
+      return __emplace_unique_key_args(__x.first, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __emplace_hint_unique(const_iterator __p, _Pp&& __x) {
+        return __emplace_hint_unique_extract_key(__p, _CUDA_VSTD::forward<_Pp>(__x),
+                                            __can_extract_key<_Pp, key_type>());
+    }
+
+    template <class _First, class _Second>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<
+        __can_extract_map_key<_First, key_type, __container_value_type>::value,
+        iterator
+    >::type __emplace_hint_unique(const_iterator __p, _First&& __f, _Second&& __s) {
+        return __emplace_hint_unique_key_args(__p, __f,
+                                              _CUDA_VSTD::forward<_First>(__f),
+                                              _CUDA_VSTD::forward<_Second>(__s));
+    }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __emplace_hint_unique(const_iterator __p, _Args&&... __args) {
+        return __emplace_hint_unique_impl(__p, _CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator
+    __emplace_hint_unique_extract_key(const_iterator __p, _Pp&& __x, __extract_key_fail_tag) {
+      return __emplace_hint_unique_impl(__p, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator
+    __emplace_hint_unique_extract_key(const_iterator __p, _Pp&& __x, __extract_key_self_tag) {
+      return __emplace_hint_unique_key_args(__p, __x, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+    template <class _Pp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator
+    __emplace_hint_unique_extract_key(const_iterator __p, _Pp&& __x, __extract_key_first_tag) {
+      return __emplace_hint_unique_key_args(__p, __x.first, _CUDA_VSTD::forward<_Pp>(__x));
+    }
+
+#else
+    template <class _Key, class _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __emplace_unique_key_args(_Key const&, _Args& __args);
+    template <class _Key, class _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __emplace_hint_unique_key_args(const_iterator, _Key const&, _Args&);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __insert_unique(const __container_value_type& __v) {
+        return __emplace_unique_key_args(_NodeTypes::__get_key(__v), __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_unique(const_iterator __p, const __container_value_type& __v) {
+        return __emplace_hint_unique_key_args(__p, _NodeTypes::__get_key(__v), __v);
+    }
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(const __container_value_type& __v);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(const_iterator __p, const __container_value_type& __v);
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __insert_unique(__container_value_type&& __v) {
+        return __emplace_unique_key_args(_NodeTypes::__get_key(__v), _CUDA_VSTD::move(__v));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_unique(const_iterator __p, __container_value_type&& __v) {
+        return __emplace_hint_unique_key_args(__p, _NodeTypes::__get_key(__v), _CUDA_VSTD::move(__v));
+    }
+
+    template <class _Vp, class = typename enable_if<
+            !is_same<typename __unconstref<_Vp>::type,
+                     __container_value_type
+            >::value
+        >::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __insert_unique(_Vp&& __v) {
+        return __emplace_unique(_CUDA_VSTD::forward<_Vp>(__v));
+    }
+
+    template <class _Vp, class = typename enable_if<
+            !is_same<typename __unconstref<_Vp>::type,
+                     __container_value_type
+            >::value
+        >::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_unique(const_iterator __p, _Vp&& __v) {
+        return __emplace_hint_unique(__p, _CUDA_VSTD::forward<_Vp>(__v));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(__container_value_type&& __v) {
+        return __emplace_multi(_CUDA_VSTD::move(__v));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(const_iterator __p, __container_value_type&& __v) {
+        return __emplace_hint_multi(__p, _CUDA_VSTD::move(__v));
+    }
+
+    template <class _Vp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(_Vp&& __v) {
+        return __emplace_multi(_CUDA_VSTD::forward<_Vp>(__v));
+    }
+
+    template <class _Vp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __insert_multi(const_iterator __p, _Vp&& __v) {
+        return __emplace_hint_multi(__p, _CUDA_VSTD::forward<_Vp>(__v));
+    }
+
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> __node_assign_unique(const __container_value_type& __v, __node_pointer __dest);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_insert_multi(__node_pointer __nd);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_insert_multi(const_iterator __p, __node_pointer __nd);
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY iterator
+    __remove_node_pointer(__node_pointer) _NOEXCEPT;
+
+#if _LIBCUDACXX_STD_VER > 14
+    template <class _NodeHandle, class _InsertReturnType>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _InsertReturnType __node_handle_insert_unique(_NodeHandle&&);
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_handle_insert_unique(const_iterator, _NodeHandle&&);
+    template <class _Tree>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __node_handle_merge_unique(_Tree& __source);
+
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_handle_insert_multi(_NodeHandle&&);
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __node_handle_insert_multi(const_iterator, _NodeHandle&&);
+    template <class _Tree>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __node_handle_merge_multi(_Tree& __source);
+
+
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _NodeHandle __node_handle_extract(key_type const&);
+    template <class _NodeHandle>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _NodeHandle __node_handle_extract(const_iterator);
+#endif
+
+    iterator erase(const_iterator __p);
+    iterator erase(const_iterator __f, const_iterator __l);
+    template <class _Key>
+        size_type __erase_unique(const _Key& __k);
+    template <class _Key>
+        size_type __erase_multi(const _Key& __k);
+
+    void __insert_node_at(__parent_pointer     __parent,
+                          __node_base_pointer& __child,
+                          __node_base_pointer __new_node) _NOEXCEPT;
+
+    template <class _Key>
+        iterator find(const _Key& __v);
+    template <class _Key>
+        const_iterator find(const _Key& __v) const;
+
+    template <class _Key>
+        size_type __count_unique(const _Key& __k) const;
+    template <class _Key>
+        size_type __count_multi(const _Key& __k) const;
+
+    template <class _Key>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator lower_bound(const _Key& __v)
+            {return __lower_bound(__v, __root(), __end_node());}
+    template <class _Key>
+        iterator __lower_bound(const _Key& __v,
+                               __node_pointer __root,
+                               __iter_pointer __result);
+    template <class _Key>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        const_iterator lower_bound(const _Key& __v) const
+            {return __lower_bound(__v, __root(), __end_node());}
+    template <class _Key>
+        const_iterator __lower_bound(const _Key& __v,
+                                     __node_pointer __root,
+                                     __iter_pointer __result) const;
+    template <class _Key>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator upper_bound(const _Key& __v)
+            {return __upper_bound(__v, __root(), __end_node());}
+    template <class _Key>
+        iterator __upper_bound(const _Key& __v,
+                               __node_pointer __root,
+                               __iter_pointer __result);
+    template <class _Key>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        const_iterator upper_bound(const _Key& __v) const
+            {return __upper_bound(__v, __root(), __end_node());}
+    template <class _Key>
+        const_iterator __upper_bound(const _Key& __v,
+                                     __node_pointer __root,
+                                     __iter_pointer __result) const;
+    template <class _Key>
+        pair<iterator, iterator>
+        __equal_range_unique(const _Key& __k);
+    template <class _Key>
+        pair<const_iterator, const_iterator>
+        __equal_range_unique(const _Key& __k) const;
+
+    template <class _Key>
+        pair<iterator, iterator>
+        __equal_range_multi(const _Key& __k);
+    template <class _Key>
+        pair<const_iterator, const_iterator>
+        __equal_range_multi(const _Key& __k) const;
+
+    typedef __tree_node_destructor<__node_allocator> _Dp;
+    typedef unique_ptr<__node, _Dp> __node_holder;
+
+    __node_holder remove(const_iterator __p) _NOEXCEPT;
+private:
+    __node_base_pointer&
+        __find_leaf_low(__parent_pointer& __parent, const key_type& __v);
+    __node_base_pointer&
+        __find_leaf_high(__parent_pointer& __parent, const key_type& __v);
+    __node_base_pointer&
+        __find_leaf(const_iterator __hint,
+                    __parent_pointer& __parent, const key_type& __v);
+    // FIXME: Make this function const qualified. Unfortunetly doing so
+    // breaks existing code which uses non-const callable comparators.
+    template <class _Key>
+    __node_base_pointer&
+        __find_equal(__parent_pointer& __parent, const _Key& __v);
+    template <class _Key>
+    _LIBCUDACXX_INLINE_VISIBILITY __node_base_pointer&
+    __find_equal(__parent_pointer& __parent, const _Key& __v) const {
+      return const_cast<__tree*>(this)->__find_equal(__parent, __v);
+    }
+    template <class _Key>
+    __node_base_pointer&
+        __find_equal(const_iterator __hint, __parent_pointer& __parent,
+                     __node_base_pointer& __dummy,
+                     const _Key& __v);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class ..._Args>
+    __node_holder __construct_node(_Args&& ...__args);
+#else
+    __node_holder __construct_node(const __container_value_type& __v);
+#endif
+
+    void destroy(__node_pointer __nd) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __tree& __t)
+        {__copy_assign_alloc(__t, integral_constant<bool,
+             __node_traits::propagate_on_container_copy_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __tree& __t, true_type)
+        {
+        if (__node_alloc() != __t.__node_alloc())
+            clear();
+        __node_alloc() = __t.__node_alloc();
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __tree&, false_type) {}
+
+    void __move_assign(__tree& __t, false_type);
+    void __move_assign(__tree& __t, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<value_compare>::value &&
+                   is_nothrow_move_assignable<__node_allocator>::value);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__tree& __t)
+        _NOEXCEPT_(
+            !__node_traits::propagate_on_container_move_assignment::value ||
+            is_nothrow_move_assignable<__node_allocator>::value)
+        {__move_assign_alloc(__t, integral_constant<bool,
+             __node_traits::propagate_on_container_move_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__tree& __t, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<__node_allocator>::value)
+        {__node_alloc() = _CUDA_VSTD::move(__t.__node_alloc());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__tree&, false_type) _NOEXCEPT {}
+
+    struct _DetachedTreeCache {
+      _LIBCUDACXX_INLINE_VISIBILITY
+      explicit _DetachedTreeCache(__tree *__t) _NOEXCEPT : __t_(__t),
+        __cache_root_(__detach_from_tree(__t)) {
+          __advance();
+        }
+
+      _LIBCUDACXX_INLINE_VISIBILITY
+      __node_pointer __get() const _NOEXCEPT {
+        return __cache_elem_;
+      }
+
+      _LIBCUDACXX_INLINE_VISIBILITY
+      void __advance() _NOEXCEPT {
+        __cache_elem_ = __cache_root_;
+        if (__cache_root_) {
+          __cache_root_ = __detach_next(__cache_root_);
+        }
+      }
+
+      _LIBCUDACXX_INLINE_VISIBILITY
+      ~_DetachedTreeCache() {
+        __t_->destroy(__cache_elem_);
+        if (__cache_root_) {
+          while (__cache_root_->__parent_ != nullptr)
+            __cache_root_ = static_cast<__node_pointer>(__cache_root_->__parent_);
+          __t_->destroy(__cache_root_);
+        }
+      }
+
+       _DetachedTreeCache(_DetachedTreeCache const&) = delete;
+       _DetachedTreeCache& operator=(_DetachedTreeCache const&) = delete;
+
+    private:
+      _LIBCUDACXX_INLINE_VISIBILITY
+      static __node_pointer __detach_from_tree(__tree *__t) _NOEXCEPT;
+      _LIBCUDACXX_INLINE_VISIBILITY
+      static __node_pointer __detach_next(__node_pointer) _NOEXCEPT;
+
+      __tree *__t_;
+      __node_pointer __cache_root_;
+      __node_pointer __cache_elem_;
+    };
+
+
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+};
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::__tree(const value_compare& __comp)
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<__node_allocator>::value &&
+            is_nothrow_copy_constructible<value_compare>::value)
+    : __pair3_(0, __comp)
+{
+    __begin_node() = __end_node();
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::__tree(const allocator_type& __a)
+    : __begin_node_(__iter_pointer()),
+      __pair1_(__second_tag(), __node_allocator(__a)),
+      __pair3_(0)
+{
+    __begin_node() = __end_node();
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::__tree(const value_compare& __comp,
+                                           const allocator_type& __a)
+    : __begin_node_(__iter_pointer()),
+      __pair1_(__second_tag(), __node_allocator(__a)),
+      __pair3_(0, __comp)
+{
+    __begin_node() = __end_node();
+}
+
+// Precondition:  size() != 0
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_pointer
+__tree<_Tp, _Compare, _Allocator>::_DetachedTreeCache::__detach_from_tree(__tree *__t) _NOEXCEPT
+{
+    __node_pointer __cache = static_cast<__node_pointer>(__t->__begin_node());
+    __t->__begin_node() = __t->__end_node();
+    __t->__end_node()->__left_->__parent_ = nullptr;
+    __t->__end_node()->__left_ = nullptr;
+    __t->size() = 0;
+    // __cache->__left_ == nullptr
+    if (__cache->__right_ != nullptr)
+        __cache = static_cast<__node_pointer>(__cache->__right_);
+    // __cache->__left_ == nullptr
+    // __cache->__right_ == nullptr
+    return __cache;
+}
+
+// Precondition:  __cache != nullptr
+//    __cache->left_ == nullptr
+//    __cache->right_ == nullptr
+//    This is no longer a red-black tree
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_pointer
+__tree<_Tp, _Compare, _Allocator>::_DetachedTreeCache::__detach_next(__node_pointer __cache) _NOEXCEPT
+{
+    if (__cache->__parent_ == nullptr)
+        return nullptr;
+    if (__tree_is_left_child(static_cast<__node_base_pointer>(__cache)))
+    {
+        __cache->__parent_->__left_ = nullptr;
+        __cache = static_cast<__node_pointer>(__cache->__parent_);
+        if (__cache->__right_ == nullptr)
+            return __cache;
+        return static_cast<__node_pointer>(__tree_leaf(__cache->__right_));
+    }
+    // __cache is right child
+    __cache->__parent_unsafe()->__right_ = nullptr;
+    __cache = static_cast<__node_pointer>(__cache->__parent_);
+    if (__cache->__left_ == nullptr)
+        return __cache;
+    return static_cast<__node_pointer>(__tree_leaf(__cache->__left_));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>&
+__tree<_Tp, _Compare, _Allocator>::operator=(const __tree& __t)
+{
+    if (this != &__t)
+    {
+        value_comp() = __t.value_comp();
+        __copy_assign_alloc(__t);
+        __assign_multi(__t.begin(), __t.end());
+    }
+    return *this;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _ForwardIterator>
+void
+__tree<_Tp, _Compare, _Allocator>::__assign_unique(_ForwardIterator __first, _ForwardIterator __last)
+{
+    typedef iterator_traits<_ForwardIterator> _ITraits;
+    typedef typename _ITraits::value_type _ItValueType;
+    static_assert((is_same<_ItValueType, __container_value_type>::value),
+                  "__assign_unique may only be called with the containers value type");
+    static_assert(__is_forward_iterator<_ForwardIterator>::value,
+                  "__assign_unique requires a forward iterator");
+    if (size() != 0)
+    {
+        _DetachedTreeCache __cache(this);
+          for (; __cache.__get() != nullptr && __first != __last; ++__first) {
+              if (__node_assign_unique(*__first, __cache.__get()).second)
+                  __cache.__advance();
+            }
+    }
+    for (; __first != __last; ++__first)
+        __insert_unique(*__first);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _InputIterator>
+void
+__tree<_Tp, _Compare, _Allocator>::__assign_multi(_InputIterator __first, _InputIterator __last)
+{
+    typedef iterator_traits<_InputIterator> _ITraits;
+    typedef typename _ITraits::value_type _ItValueType;
+    static_assert((is_same<_ItValueType, __container_value_type>::value ||
+                  is_same<_ItValueType, __node_value_type>::value),
+                  "__assign_multi may only be called with the containers value type"
+                  " or the nodes value type");
+    if (size() != 0)
+    {
+        _DetachedTreeCache __cache(this);
+        for (; __cache.__get() && __first != __last; ++__first) {
+            __cache.__get()->__value_ = *__first;
+            __node_insert_multi(__cache.__get());
+            __cache.__advance();
+        }
+    }
+    for (; __first != __last; ++__first)
+        __insert_multi(_NodeTypes::__get_value(*__first));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::__tree(const __tree& __t)
+    : __begin_node_(__iter_pointer()),
+      __pair1_(__second_tag(), __node_traits::select_on_container_copy_construction(__t.__node_alloc())),
+      __pair3_(0, __t.value_comp())
+{
+    __begin_node() = __end_node();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::__tree(__tree&& __t)
+    _NOEXCEPT_(
+        is_nothrow_move_constructible<__node_allocator>::value &&
+        is_nothrow_move_constructible<value_compare>::value)
+    : __begin_node_(_CUDA_VSTD::move(__t.__begin_node_)),
+      __pair1_(_CUDA_VSTD::move(__t.__pair1_)),
+      __pair3_(_CUDA_VSTD::move(__t.__pair3_))
+{
+    if (size() == 0)
+        __begin_node() = __end_node();
+    else
+    {
+        __end_node()->__left_->__parent_ = static_cast<__parent_pointer>(__end_node());
+        __t.__begin_node() = __t.__end_node();
+        __t.__end_node()->__left_ = nullptr;
+        __t.size() = 0;
+    }
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::__tree(__tree&& __t, const allocator_type& __a)
+    : __pair1_(__second_tag(), __node_allocator(__a)),
+      __pair3_(0, _CUDA_VSTD::move(__t.value_comp()))
+{
+    if (__a == __t.__alloc())
+    {
+        if (__t.size() == 0)
+            __begin_node() = __end_node();
+        else
+        {
+            __begin_node() = __t.__begin_node();
+            __end_node()->__left_ = __t.__end_node()->__left_;
+            __end_node()->__left_->__parent_ = static_cast<__parent_pointer>(__end_node());
+            size() = __t.size();
+            __t.__begin_node() = __t.__end_node();
+            __t.__end_node()->__left_ = nullptr;
+            __t.size() = 0;
+        }
+    }
+    else
+    {
+        __begin_node() = __end_node();
+    }
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+void
+__tree<_Tp, _Compare, _Allocator>::__move_assign(__tree& __t, true_type)
+    _NOEXCEPT_(is_nothrow_move_assignable<value_compare>::value &&
+               is_nothrow_move_assignable<__node_allocator>::value)
+{
+    destroy(static_cast<__node_pointer>(__end_node()->__left_));
+    __begin_node_ = __t.__begin_node_;
+    __pair1_.first() = __t.__pair1_.first();
+    __move_assign_alloc(__t);
+    __pair3_ = _CUDA_VSTD::move(__t.__pair3_);
+    if (size() == 0)
+        __begin_node() = __end_node();
+    else
+    {
+        __end_node()->__left_->__parent_ = static_cast<__parent_pointer>(__end_node());
+        __t.__begin_node() = __t.__end_node();
+        __t.__end_node()->__left_ = nullptr;
+        __t.size() = 0;
+    }
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+void
+__tree<_Tp, _Compare, _Allocator>::__move_assign(__tree& __t, false_type)
+{
+    if (__node_alloc() == __t.__node_alloc())
+        __move_assign(__t, true_type());
+    else
+    {
+        value_comp() = _CUDA_VSTD::move(__t.value_comp());
+        const_iterator __e = end();
+        if (size() != 0)
+        {
+            _DetachedTreeCache __cache(this);
+            while (__cache.__get() != nullptr && __t.size() != 0) {
+              __cache.__get()->__value_ = _CUDA_VSTD::move(__t.remove(__t.begin())->__value_);
+              __node_insert_multi(__cache.__get());
+              __cache.__advance();
+            }
+        }
+        while (__t.size() != 0)
+            __insert_multi(__e, _NodeTypes::__move(__t.remove(__t.begin())->__value_));
+    }
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>&
+__tree<_Tp, _Compare, _Allocator>::operator=(__tree&& __t)
+    _NOEXCEPT_(
+        __node_traits::propagate_on_container_move_assignment::value &&
+        is_nothrow_move_assignable<value_compare>::value &&
+        is_nothrow_move_assignable<__node_allocator>::value)
+
+{
+    __move_assign(__t, integral_constant<bool,
+                  __node_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Compare, class _Allocator>
+__tree<_Tp, _Compare, _Allocator>::~__tree()
+{
+    static_assert((is_copy_constructible<value_compare>::value),
+                 "Comparator must be copy-constructible.");
+  destroy(__root());
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+void
+__tree<_Tp, _Compare, _Allocator>::destroy(__node_pointer __nd) _NOEXCEPT
+{
+    if (__nd != nullptr)
+    {
+        destroy(static_cast<__node_pointer>(__nd->__left_));
+        destroy(static_cast<__node_pointer>(__nd->__right_));
+        __node_allocator& __na = __node_alloc();
+        __node_traits::destroy(__na, _NodeTypes::__get_ptr(__nd->__value_));
+        __node_traits::deallocate(__na, __nd, 1);
+    }
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+void
+__tree<_Tp, _Compare, _Allocator>::swap(__tree& __t)
+#if _LIBCUDACXX_STD_VER <= 11
+        _NOEXCEPT_(
+            __is_nothrow_swappable<value_compare>::value
+            && (!__node_traits::propagate_on_container_swap::value ||
+                 __is_nothrow_swappable<__node_allocator>::value)
+            )
+#else
+        _NOEXCEPT_(__is_nothrow_swappable<value_compare>::value)
+#endif
+{
+    using _CUDA_VSTD::swap;
+    swap(__begin_node_, __t.__begin_node_);
+    swap(__pair1_.first(), __t.__pair1_.first());
+    __swap_allocator(__node_alloc(), __t.__node_alloc());
+    __pair3_.swap(__t.__pair3_);
+    if (size() == 0)
+        __begin_node() = __end_node();
+    else
+        __end_node()->__left_->__parent_ = static_cast<__parent_pointer>(__end_node());
+    if (__t.size() == 0)
+        __t.__begin_node() = __t.__end_node();
+    else
+        __t.__end_node()->__left_->__parent_ = static_cast<__parent_pointer>(__t.__end_node());
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+void
+__tree<_Tp, _Compare, _Allocator>::clear() _NOEXCEPT
+{
+    destroy(__root());
+    size() = 0;
+    __begin_node() = __end_node();
+    __end_node()->__left_ = nullptr;
+}
+
+// Find lower_bound place to insert
+// Set __parent to parent of null leaf
+// Return reference to null leaf
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_base_pointer&
+__tree<_Tp, _Compare, _Allocator>::__find_leaf_low(__parent_pointer& __parent,
+                                                   const key_type& __v)
+{
+    __node_pointer __nd = __root();
+    if (__nd != nullptr)
+    {
+        while (true)
+        {
+            if (value_comp()(__nd->__value_, __v))
+            {
+                if (__nd->__right_ != nullptr)
+                    __nd = static_cast<__node_pointer>(__nd->__right_);
+                else
+                {
+                    __parent = static_cast<__parent_pointer>(__nd);
+                    return __nd->__right_;
+                }
+            }
+            else
+            {
+                if (__nd->__left_ != nullptr)
+                    __nd = static_cast<__node_pointer>(__nd->__left_);
+                else
+                {
+                    __parent = static_cast<__parent_pointer>(__nd);
+                    return __parent->__left_;
+                }
+            }
+        }
+    }
+    __parent = static_cast<__parent_pointer>(__end_node());
+    return __parent->__left_;
+}
+
+// Find upper_bound place to insert
+// Set __parent to parent of null leaf
+// Return reference to null leaf
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_base_pointer&
+__tree<_Tp, _Compare, _Allocator>::__find_leaf_high(__parent_pointer& __parent,
+                                                    const key_type& __v)
+{
+    __node_pointer __nd = __root();
+    if (__nd != nullptr)
+    {
+        while (true)
+        {
+            if (value_comp()(__v, __nd->__value_))
+            {
+                if (__nd->__left_ != nullptr)
+                    __nd = static_cast<__node_pointer>(__nd->__left_);
+                else
+                {
+                    __parent = static_cast<__parent_pointer>(__nd);
+                    return __parent->__left_;
+                }
+            }
+            else
+            {
+                if (__nd->__right_ != nullptr)
+                    __nd = static_cast<__node_pointer>(__nd->__right_);
+                else
+                {
+                    __parent = static_cast<__parent_pointer>(__nd);
+                    return __nd->__right_;
+                }
+            }
+        }
+    }
+    __parent = static_cast<__parent_pointer>(__end_node());
+    return __parent->__left_;
+}
+
+// Find leaf place to insert closest to __hint
+// First check prior to __hint.
+// Next check after __hint.
+// Next do O(log N) search.
+// Set __parent to parent of null leaf
+// Return reference to null leaf
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_base_pointer&
+__tree<_Tp, _Compare, _Allocator>::__find_leaf(const_iterator __hint,
+                                               __parent_pointer& __parent,
+                                               const key_type& __v)
+{
+    if (__hint == end() || !value_comp()(*__hint, __v))  // check before
+    {
+        // __v <= *__hint
+        const_iterator __prior = __hint;
+        if (__prior == begin() || !value_comp()(__v, *--__prior))
+        {
+            // *prev(__hint) <= __v <= *__hint
+            if (__hint.__ptr_->__left_ == nullptr)
+            {
+                __parent = static_cast<__parent_pointer>(__hint.__ptr_);
+                return __parent->__left_;
+            }
+            else
+            {
+                __parent = static_cast<__parent_pointer>(__prior.__ptr_);
+                return static_cast<__node_base_pointer>(__prior.__ptr_)->__right_;
+            }
+        }
+        // __v < *prev(__hint)
+        return __find_leaf_high(__parent, __v);
+    }
+    // else __v > *__hint
+    return __find_leaf_low(__parent, __v);
+}
+
+// Find place to insert if __v doesn't exist
+// Set __parent to parent of null leaf
+// Return reference to null leaf
+// If __v exists, set parent to node of __v and return reference to node of __v
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::__node_base_pointer&
+__tree<_Tp, _Compare, _Allocator>::__find_equal(__parent_pointer& __parent,
+                                                const _Key& __v)
+{
+    __node_pointer __nd = __root();
+    __node_base_pointer* __nd_ptr = __root_ptr();
+    if (__nd != nullptr)
+    {
+        while (true)
+        {
+            if (value_comp()(__v, __nd->__value_))
+            {
+                if (__nd->__left_ != nullptr) {
+                    __nd_ptr = _CUDA_VSTD::addressof(__nd->__left_);
+                    __nd = static_cast<__node_pointer>(__nd->__left_);
+                } else {
+                    __parent = static_cast<__parent_pointer>(__nd);
+                    return __parent->__left_;
+                }
+            }
+            else if (value_comp()(__nd->__value_, __v))
+            {
+                if (__nd->__right_ != nullptr) {
+                    __nd_ptr = _CUDA_VSTD::addressof(__nd->__right_);
+                    __nd = static_cast<__node_pointer>(__nd->__right_);
+                } else {
+                    __parent = static_cast<__parent_pointer>(__nd);
+                    return __nd->__right_;
+                }
+            }
+            else
+            {
+                __parent = static_cast<__parent_pointer>(__nd);
+                return *__nd_ptr;
+            }
+        }
+    }
+    __parent = static_cast<__parent_pointer>(__end_node());
+    return __parent->__left_;
+}
+
+// Find place to insert if __v doesn't exist
+// First check prior to __hint.
+// Next check after __hint.
+// Next do O(log N) search.
+// Set __parent to parent of null leaf
+// Return reference to null leaf
+// If __v exists, set parent to node of __v and return reference to node of __v
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::__node_base_pointer&
+__tree<_Tp, _Compare, _Allocator>::__find_equal(const_iterator __hint,
+                                                __parent_pointer& __parent,
+                                                __node_base_pointer& __dummy,
+                                                const _Key& __v)
+{
+    if (__hint == end() || value_comp()(__v, *__hint))  // check before
+    {
+        // __v < *__hint
+        const_iterator __prior = __hint;
+        if (__prior == begin() || value_comp()(*--__prior, __v))
+        {
+            // *prev(__hint) < __v < *__hint
+            if (__hint.__ptr_->__left_ == nullptr)
+            {
+                __parent = static_cast<__parent_pointer>(__hint.__ptr_);
+                return __parent->__left_;
+            }
+            else
+            {
+                __parent = static_cast<__parent_pointer>(__prior.__ptr_);
+                return static_cast<__node_base_pointer>(__prior.__ptr_)->__right_;
+            }
+        }
+        // __v <= *prev(__hint)
+        return __find_equal(__parent, __v);
+    }
+    else if (value_comp()(*__hint, __v))  // check after
+    {
+        // *__hint < __v
+        const_iterator __next = _CUDA_VSTD::next(__hint);
+        if (__next == end() || value_comp()(__v, *__next))
+        {
+            // *__hint < __v < *_CUDA_VSTD::next(__hint)
+            if (__hint.__get_np()->__right_ == nullptr)
+            {
+                __parent = static_cast<__parent_pointer>(__hint.__ptr_);
+                return static_cast<__node_base_pointer>(__hint.__ptr_)->__right_;
+            }
+            else
+            {
+                __parent = static_cast<__parent_pointer>(__next.__ptr_);
+                return __parent->__left_;
+            }
+        }
+        // *next(__hint) <= __v
+        return __find_equal(__parent, __v);
+    }
+    // else __v == *__hint
+    __parent = static_cast<__parent_pointer>(__hint.__ptr_);
+    __dummy = static_cast<__node_base_pointer>(__hint.__ptr_);
+    return __dummy;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+void __tree<_Tp, _Compare, _Allocator>::__insert_node_at(
+    __parent_pointer __parent, __node_base_pointer& __child,
+    __node_base_pointer __new_node) _NOEXCEPT
+{
+    __new_node->__left_   = nullptr;
+    __new_node->__right_  = nullptr;
+    __new_node->__parent_ = __parent;
+    // __new_node->__is_black_ is initialized in __tree_balance_after_insert
+    __child = __new_node;
+    if (__begin_node()->__left_ != nullptr)
+        __begin_node() = static_cast<__iter_pointer>(__begin_node()->__left_);
+    __tree_balance_after_insert(__end_node()->__left_, __child);
+    ++size();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key, class... _Args>
+pair<typename __tree<_Tp, _Compare, _Allocator>::iterator, bool>
+__tree<_Tp, _Compare, _Allocator>::__emplace_unique_key_args(_Key const& __k, _Args&&... __args)
+#else
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key, class _Args>
+pair<typename __tree<_Tp, _Compare, _Allocator>::iterator, bool>
+__tree<_Tp, _Compare, _Allocator>::__emplace_unique_key_args(_Key const& __k, _Args& __args)
+#endif
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_equal(__parent, __k);
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    bool __inserted = false;
+    if (__child == nullptr)
+    {
+#ifndef _LIBCUDACXX_CXX03_LANG
+        __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+#else
+        __node_holder __h = __construct_node(__args);
+#endif
+        __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+        __r = __h.release();
+        __inserted = true;
+    }
+    return pair<iterator, bool>(iterator(__r), __inserted);
+}
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key, class... _Args>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__emplace_hint_unique_key_args(
+    const_iterator __p, _Key const& __k, _Args&&... __args)
+#else
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key, class _Args>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__emplace_hint_unique_key_args(
+    const_iterator __p, _Key const& __k, _Args& __args)
+#endif
+{
+    __parent_pointer __parent;
+    __node_base_pointer __dummy;
+    __node_base_pointer& __child = __find_equal(__p, __parent, __dummy, __k);
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    if (__child == nullptr)
+    {
+#ifndef _LIBCUDACXX_CXX03_LANG
+        __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+#else
+        __node_holder __h = __construct_node(__args);
+#endif
+        __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+        __r = __h.release();
+    }
+    return iterator(__r);
+}
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class ..._Args>
+typename __tree<_Tp, _Compare, _Allocator>::__node_holder
+__tree<_Tp, _Compare, _Allocator>::__construct_node(_Args&& ...__args)
+{
+    static_assert(!__is_tree_value_type<_Args...>::value,
+                  "Cannot construct from __value_type");
+    __node_allocator& __na = __node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _NodeTypes::__get_ptr(__h->__value_), _CUDA_VSTD::forward<_Args>(__args)...);
+    __h.get_deleter().__value_constructed = true;
+    return __h;
+}
+
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class... _Args>
+pair<typename __tree<_Tp, _Compare, _Allocator>::iterator, bool>
+__tree<_Tp, _Compare, _Allocator>::__emplace_unique_impl(_Args&&... __args)
+{
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_equal(__parent, __h->__value_);
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    bool __inserted = false;
+    if (__child == nullptr)
+    {
+        __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+        __r = __h.release();
+        __inserted = true;
+    }
+    return pair<iterator, bool>(iterator(__r), __inserted);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class... _Args>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__emplace_hint_unique_impl(const_iterator __p, _Args&&... __args)
+{
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    __parent_pointer __parent;
+    __node_base_pointer __dummy;
+    __node_base_pointer& __child = __find_equal(__p, __parent, __dummy, __h->__value_);
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    if (__child == nullptr)
+    {
+        __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+        __r = __h.release();
+    }
+    return iterator(__r);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class... _Args>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__emplace_multi(_Args&&... __args)
+{
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf_high(__parent, _NodeTypes::__get_key(__h->__value_));
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+    return iterator(static_cast<__node_pointer>(__h.release()));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class... _Args>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__emplace_hint_multi(const_iterator __p,
+                                                        _Args&&... __args)
+{
+    __node_holder __h = __construct_node(_CUDA_VSTD::forward<_Args>(__args)...);
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf(__p, __parent, _NodeTypes::__get_key(__h->__value_));
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+    return iterator(static_cast<__node_pointer>(__h.release()));
+}
+
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_holder
+__tree<_Tp, _Compare, _Allocator>::__construct_node(const __container_value_type& __v)
+{
+    __node_allocator& __na = __node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _NodeTypes::__get_ptr(__h->__value_), __v);
+    __h.get_deleter().__value_constructed = true;
+    return _LIBCUDACXX_EXPLICIT_MOVE(__h);  // explicitly moved for C++03
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__insert_multi(const __container_value_type& __v)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf_high(__parent, _NodeTypes::__get_key(__v));
+    __node_holder __h = __construct_node(__v);
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+    return iterator(__h.release());
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__insert_multi(const_iterator __p, const __container_value_type& __v)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf(__p, __parent, _NodeTypes::__get_key(__v));
+    __node_holder __h = __construct_node(__v);
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+    return iterator(__h.release());
+}
+#endif
+
+template <class _Tp, class _Compare, class _Allocator>
+pair<typename __tree<_Tp, _Compare, _Allocator>::iterator, bool>
+__tree<_Tp, _Compare, _Allocator>::__node_assign_unique(const __container_value_type& __v, __node_pointer __nd)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_equal(__parent, _NodeTypes::__get_key(__v));
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    bool __inserted = false;
+    if (__child == nullptr)
+    {
+        __nd->__value_ = __v;
+        __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__nd));
+        __r = __nd;
+        __inserted = true;
+    }
+    return pair<iterator, bool>(iterator(__r), __inserted);
+}
+
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__node_insert_multi(__node_pointer __nd)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf_high(__parent, _NodeTypes::__get_key(__nd->__value_));
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__nd));
+    return iterator(__nd);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__node_insert_multi(const_iterator __p,
+                                                       __node_pointer __nd)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf(__p, __parent, _NodeTypes::__get_key(__nd->__value_));
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__nd));
+    return iterator(__nd);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__remove_node_pointer(__node_pointer __ptr) _NOEXCEPT
+{
+    iterator __r(__ptr);
+    ++__r;
+    if (__begin_node() == __ptr)
+        __begin_node() = __r.__ptr_;
+    --size();
+    __tree_remove(__end_node()->__left_,
+                  static_cast<__node_base_pointer>(__ptr));
+    return __r;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp, class _Compare, class _Allocator>
+template <class _NodeHandle, class _InsertReturnType>
+_LIBCUDACXX_INLINE_VISIBILITY
+_InsertReturnType
+__tree<_Tp, _Compare, _Allocator>::__node_handle_insert_unique(
+    _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return _InsertReturnType{end(), false, _NodeHandle()};
+
+    __node_pointer __ptr = __nh.__ptr_;
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_equal(__parent,
+                                                __ptr->__value_);
+    if (__child != nullptr)
+        return _InsertReturnType{
+            iterator(static_cast<__node_pointer>(__child)),
+            false, _CUDA_VSTD::move(__nh)};
+
+    __insert_node_at(__parent, __child,
+                     static_cast<__node_base_pointer>(__ptr));
+    __nh.__release_ptr();
+    return _InsertReturnType{iterator(__ptr), true, _NodeHandle()};
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__node_handle_insert_unique(
+    const_iterator __hint, _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return end();
+
+    __node_pointer __ptr = __nh.__ptr_;
+    __parent_pointer __parent;
+    __node_base_pointer __dummy;
+    __node_base_pointer& __child = __find_equal(__hint, __parent, __dummy,
+                                                __ptr->__value_);
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    if (__child == nullptr)
+    {
+        __insert_node_at(__parent, __child,
+                         static_cast<__node_base_pointer>(__ptr));
+        __r = __ptr;
+        __nh.__release_ptr();
+    }
+    return iterator(__r);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+_NodeHandle
+__tree<_Tp, _Compare, _Allocator>::__node_handle_extract(key_type const& __key)
+{
+    iterator __it = find(__key);
+    if (__it == end())
+        return _NodeHandle();
+    return __node_handle_extract<_NodeHandle>(__it);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+_NodeHandle
+__tree<_Tp, _Compare, _Allocator>::__node_handle_extract(const_iterator __p)
+{
+    __node_pointer __np = __p.__get_np();
+    __remove_node_pointer(__np);
+    return _NodeHandle(__np, __alloc());
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Tree>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+__tree<_Tp, _Compare, _Allocator>::__node_handle_merge_unique(_Tree& __source)
+{
+    static_assert(is_same<typename _Tree::__node_pointer, __node_pointer>::value, "");
+
+    for (typename _Tree::iterator __i = __source.begin();
+         __i != __source.end();)
+    {
+        __node_pointer __src_ptr = __i.__get_np();
+        __parent_pointer __parent;
+        __node_base_pointer& __child =
+            __find_equal(__parent, _NodeTypes::__get_key(__src_ptr->__value_));
+        ++__i;
+        if (__child != nullptr)
+            continue;
+        __source.__remove_node_pointer(__src_ptr);
+        __insert_node_at(__parent, __child,
+                         static_cast<__node_base_pointer>(__src_ptr));
+    }
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__node_handle_insert_multi(_NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return end();
+    __node_pointer __ptr = __nh.__ptr_;
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf_high(
+        __parent, _NodeTypes::__get_key(__ptr->__value_));
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__ptr));
+    __nh.__release_ptr();
+    return iterator(__ptr);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _NodeHandle>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__node_handle_insert_multi(
+    const_iterator __hint, _NodeHandle&& __nh)
+{
+    if (__nh.empty())
+        return end();
+
+    __node_pointer __ptr = __nh.__ptr_;
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __find_leaf(__hint, __parent,
+                                               _NodeTypes::__get_key(__ptr->__value_));
+    __insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__ptr));
+    __nh.__release_ptr();
+    return iterator(__ptr);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Tree>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+__tree<_Tp, _Compare, _Allocator>::__node_handle_merge_multi(_Tree& __source)
+{
+    static_assert(is_same<typename _Tree::__node_pointer, __node_pointer>::value, "");
+
+    for (typename _Tree::iterator __i = __source.begin();
+         __i != __source.end();)
+    {
+        __node_pointer __src_ptr = __i.__get_np();
+        __parent_pointer __parent;
+        __node_base_pointer& __child = __find_leaf_high(
+            __parent, _NodeTypes::__get_key(__src_ptr->__value_));
+        ++__i;
+        __source.__remove_node_pointer(__src_ptr);
+        __insert_node_at(__parent, __child,
+                         static_cast<__node_base_pointer>(__src_ptr));
+    }
+}
+
+#endif  // _LIBCUDACXX_STD_VER > 14
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::erase(const_iterator __p)
+{
+    __node_pointer __np = __p.__get_np();
+    iterator __r = __remove_node_pointer(__np);
+    __node_allocator& __na = __node_alloc();
+    __node_traits::destroy(__na, _NodeTypes::__get_ptr(
+        const_cast<__node_value_type&>(*__p)));
+    __node_traits::deallocate(__na, __np, 1);
+    return __r;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::erase(const_iterator __f, const_iterator __l)
+{
+    while (__f != __l)
+        __f = erase(__f);
+    return iterator(__l.__ptr_);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::size_type
+__tree<_Tp, _Compare, _Allocator>::__erase_unique(const _Key& __k)
+{
+    iterator __i = find(__k);
+    if (__i == end())
+        return 0;
+    erase(__i);
+    return 1;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::size_type
+__tree<_Tp, _Compare, _Allocator>::__erase_multi(const _Key& __k)
+{
+    pair<iterator, iterator> __p = __equal_range_multi(__k);
+    size_type __r = 0;
+    for (; __p.first != __p.second; ++__r)
+        __p.first = erase(__p.first);
+    return __r;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::find(const _Key& __v)
+{
+    iterator __p = __lower_bound(__v, __root(), __end_node());
+    if (__p != end() && !value_comp()(__v, *__p))
+        return __p;
+    return end();
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::const_iterator
+__tree<_Tp, _Compare, _Allocator>::find(const _Key& __v) const
+{
+    const_iterator __p = __lower_bound(__v, __root(), __end_node());
+    if (__p != end() && !value_comp()(__v, *__p))
+        return __p;
+    return end();
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::size_type
+__tree<_Tp, _Compare, _Allocator>::__count_unique(const _Key& __k) const
+{
+    __node_pointer __rt = __root();
+    while (__rt != nullptr)
+    {
+        if (value_comp()(__k, __rt->__value_))
+        {
+            __rt = static_cast<__node_pointer>(__rt->__left_);
+        }
+        else if (value_comp()(__rt->__value_, __k))
+            __rt = static_cast<__node_pointer>(__rt->__right_);
+        else
+            return 1;
+    }
+    return 0;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::size_type
+__tree<_Tp, _Compare, _Allocator>::__count_multi(const _Key& __k) const
+{
+    __iter_pointer __result = __end_node();
+    __node_pointer __rt = __root();
+    while (__rt != nullptr)
+    {
+        if (value_comp()(__k, __rt->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__rt);
+            __rt = static_cast<__node_pointer>(__rt->__left_);
+        }
+        else if (value_comp()(__rt->__value_, __k))
+            __rt = static_cast<__node_pointer>(__rt->__right_);
+        else
+            return _CUDA_VSTD::distance(
+                __lower_bound(__k, static_cast<__node_pointer>(__rt->__left_), static_cast<__iter_pointer>(__rt)),
+                __upper_bound(__k, static_cast<__node_pointer>(__rt->__right_), __result)
+            );
+    }
+    return 0;
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__lower_bound(const _Key& __v,
+                                                 __node_pointer __root,
+                                                 __iter_pointer __result)
+{
+    while (__root != nullptr)
+    {
+        if (!value_comp()(__root->__value_, __v))
+        {
+            __result = static_cast<__iter_pointer>(__root);
+            __root = static_cast<__node_pointer>(__root->__left_);
+        }
+        else
+            __root = static_cast<__node_pointer>(__root->__right_);
+    }
+    return iterator(__result);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::const_iterator
+__tree<_Tp, _Compare, _Allocator>::__lower_bound(const _Key& __v,
+                                                 __node_pointer __root,
+                                                 __iter_pointer __result) const
+{
+    while (__root != nullptr)
+    {
+        if (!value_comp()(__root->__value_, __v))
+        {
+            __result = static_cast<__iter_pointer>(__root);
+            __root = static_cast<__node_pointer>(__root->__left_);
+        }
+        else
+            __root = static_cast<__node_pointer>(__root->__right_);
+    }
+    return const_iterator(__result);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::iterator
+__tree<_Tp, _Compare, _Allocator>::__upper_bound(const _Key& __v,
+                                                 __node_pointer __root,
+                                                 __iter_pointer __result)
+{
+    while (__root != nullptr)
+    {
+        if (value_comp()(__v, __root->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__root);
+            __root = static_cast<__node_pointer>(__root->__left_);
+        }
+        else
+            __root = static_cast<__node_pointer>(__root->__right_);
+    }
+    return iterator(__result);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+typename __tree<_Tp, _Compare, _Allocator>::const_iterator
+__tree<_Tp, _Compare, _Allocator>::__upper_bound(const _Key& __v,
+                                                 __node_pointer __root,
+                                                 __iter_pointer __result) const
+{
+    while (__root != nullptr)
+    {
+        if (value_comp()(__v, __root->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__root);
+            __root = static_cast<__node_pointer>(__root->__left_);
+        }
+        else
+            __root = static_cast<__node_pointer>(__root->__right_);
+    }
+    return const_iterator(__result);
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+pair<typename __tree<_Tp, _Compare, _Allocator>::iterator,
+     typename __tree<_Tp, _Compare, _Allocator>::iterator>
+__tree<_Tp, _Compare, _Allocator>::__equal_range_unique(const _Key& __k)
+{
+    typedef pair<iterator, iterator> _Pp;
+    __iter_pointer __result = __end_node();
+    __node_pointer __rt = __root();
+    while (__rt != nullptr)
+    {
+        if (value_comp()(__k, __rt->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__rt);
+            __rt = static_cast<__node_pointer>(__rt->__left_);
+        }
+        else if (value_comp()(__rt->__value_, __k))
+            __rt = static_cast<__node_pointer>(__rt->__right_);
+        else
+            return _Pp(iterator(__rt),
+                      iterator(
+                          __rt->__right_ != nullptr ?
+                              static_cast<__iter_pointer>(__tree_min(__rt->__right_))
+                            : __result));
+    }
+    return _Pp(iterator(__result), iterator(__result));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+pair<typename __tree<_Tp, _Compare, _Allocator>::const_iterator,
+     typename __tree<_Tp, _Compare, _Allocator>::const_iterator>
+__tree<_Tp, _Compare, _Allocator>::__equal_range_unique(const _Key& __k) const
+{
+    typedef pair<const_iterator, const_iterator> _Pp;
+    __iter_pointer __result = __end_node();
+    __node_pointer __rt = __root();
+    while (__rt != nullptr)
+    {
+        if (value_comp()(__k, __rt->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__rt);
+            __rt = static_cast<__node_pointer>(__rt->__left_);
+        }
+        else if (value_comp()(__rt->__value_, __k))
+            __rt = static_cast<__node_pointer>(__rt->__right_);
+        else
+            return _Pp(const_iterator(__rt),
+                      const_iterator(
+                          __rt->__right_ != nullptr ?
+                              static_cast<__iter_pointer>(__tree_min(__rt->__right_))
+                            : __result));
+    }
+    return _Pp(const_iterator(__result), const_iterator(__result));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+pair<typename __tree<_Tp, _Compare, _Allocator>::iterator,
+     typename __tree<_Tp, _Compare, _Allocator>::iterator>
+__tree<_Tp, _Compare, _Allocator>::__equal_range_multi(const _Key& __k)
+{
+    typedef pair<iterator, iterator> _Pp;
+    __iter_pointer __result = __end_node();
+    __node_pointer __rt = __root();
+    while (__rt != nullptr)
+    {
+        if (value_comp()(__k, __rt->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__rt);
+            __rt = static_cast<__node_pointer>(__rt->__left_);
+        }
+        else if (value_comp()(__rt->__value_, __k))
+            __rt = static_cast<__node_pointer>(__rt->__right_);
+        else
+            return _Pp(__lower_bound(__k, static_cast<__node_pointer>(__rt->__left_), static_cast<__iter_pointer>(__rt)),
+                      __upper_bound(__k, static_cast<__node_pointer>(__rt->__right_), __result));
+    }
+    return _Pp(iterator(__result), iterator(__result));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+template <class _Key>
+pair<typename __tree<_Tp, _Compare, _Allocator>::const_iterator,
+     typename __tree<_Tp, _Compare, _Allocator>::const_iterator>
+__tree<_Tp, _Compare, _Allocator>::__equal_range_multi(const _Key& __k) const
+{
+    typedef pair<const_iterator, const_iterator> _Pp;
+    __iter_pointer __result = __end_node();
+    __node_pointer __rt = __root();
+    while (__rt != nullptr)
+    {
+        if (value_comp()(__k, __rt->__value_))
+        {
+            __result = static_cast<__iter_pointer>(__rt);
+            __rt = static_cast<__node_pointer>(__rt->__left_);
+        }
+        else if (value_comp()(__rt->__value_, __k))
+            __rt = static_cast<__node_pointer>(__rt->__right_);
+        else
+            return _Pp(__lower_bound(__k, static_cast<__node_pointer>(__rt->__left_), static_cast<__iter_pointer>(__rt)),
+                      __upper_bound(__k, static_cast<__node_pointer>(__rt->__right_), __result));
+    }
+    return _Pp(const_iterator(__result), const_iterator(__result));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+typename __tree<_Tp, _Compare, _Allocator>::__node_holder
+__tree<_Tp, _Compare, _Allocator>::remove(const_iterator __p) _NOEXCEPT
+{
+    __node_pointer __np = __p.__get_np();
+    if (__begin_node() == __p.__ptr_)
+    {
+        if (__np->__right_ != nullptr)
+            __begin_node() = static_cast<__iter_pointer>(__np->__right_);
+        else
+            __begin_node() = static_cast<__iter_pointer>(__np->__parent_);
+    }
+    --size();
+    __tree_remove(__end_node()->__left_,
+                  static_cast<__node_base_pointer>(__np));
+    return __node_holder(__np, _Dp(__node_alloc(), true));
+}
+
+template <class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__tree<_Tp, _Compare, _Allocator>& __x,
+     __tree<_Tp, _Compare, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX___TREE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple
new file mode 100644
index 000000000000..357e3156761e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__tuple
@@ -0,0 +1,567 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX___TUPLE
+#define _LIBCUDACXX___TUPLE
+
+#ifndef __cuda_std__
+#include <__config>
+#include <cstddef>
+#include <type_traits>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS tuple_size;
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+template <class _Tp, class...>
+using __enable_if_tuple_size_imp = _Tp;
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<__enable_if_tuple_size_imp<
+    const _Tp,
+    typename enable_if<!is_volatile<_Tp>::value>::type,
+    integral_constant<size_t, sizeof(tuple_size<_Tp>)>>>
+    : public integral_constant<size_t, tuple_size<_Tp>::value> {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<__enable_if_tuple_size_imp<
+    volatile _Tp,
+    typename enable_if<!is_const<_Tp>::value>::type,
+    integral_constant<size_t, sizeof(tuple_size<_Tp>)>>>
+    : public integral_constant<size_t, tuple_size<_Tp>::value> {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<__enable_if_tuple_size_imp<
+    const volatile _Tp,
+    integral_constant<size_t, sizeof(tuple_size<_Tp>)>>>
+    : public integral_constant<size_t, tuple_size<_Tp>::value> {};
+
+#else
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<const _Tp> : public tuple_size<_Tp> {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<volatile _Tp> : public tuple_size<_Tp> {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<const volatile _Tp> : public tuple_size<_Tp> {};
+#endif
+
+template <size_t _Ip, class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS tuple_element;
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, const _Tp>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename add_const<typename tuple_element<_Ip, _Tp>::type>::type type;
+};
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, volatile _Tp>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename add_volatile<typename tuple_element<_Ip, _Tp>::type>::type type;
+};
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, const volatile _Tp>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename add_cv<typename tuple_element<_Ip, _Tp>::type>::type type;
+};
+
+template <class _Tp> struct __tuple_like : false_type {};
+
+template <class _Tp> struct __tuple_like<const _Tp> : public __tuple_like<_Tp> {};
+template <class _Tp> struct __tuple_like<volatile _Tp> : public __tuple_like<_Tp> {};
+template <class _Tp> struct __tuple_like<const volatile _Tp> : public __tuple_like<_Tp> {};
+
+// tuple specializations
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <size_t...> struct __tuple_indices {};
+
+template <class _IdxType, _IdxType... _Values>
+struct __integer_sequence {
+  template <template <class _OIdxType, _OIdxType...> class _ToIndexSeq, class _ToIndexType>
+  using __convert = _ToIndexSeq<_ToIndexType, _Values...>;
+
+  template <size_t _Sp>
+  using __to_tuple_indices = __tuple_indices<(_Values + _Sp)...>;
+};
+
+#if !__has_builtin(__make_integer_seq) || defined(_LIBCUDACXX_TESTING_FALLBACK_MAKE_INTEGER_SEQUENCE)
+namespace __detail {
+
+template<typename _Tp, size_t ..._Extra> struct __repeat;
+template<typename _Tp, _Tp ..._Np, size_t ..._Extra> struct __repeat<__integer_sequence<_Tp, _Np...>, _Extra...> {
+  typedef _LIBCUDACXX_NODEBUG_TYPE __integer_sequence<_Tp,
+                           _Np...,
+                           sizeof...(_Np) + _Np...,
+                           2 * sizeof...(_Np) + _Np...,
+                           3 * sizeof...(_Np) + _Np...,
+                           4 * sizeof...(_Np) + _Np...,
+                           5 * sizeof...(_Np) + _Np...,
+                           6 * sizeof...(_Np) + _Np...,
+                           7 * sizeof...(_Np) + _Np...,
+                           _Extra...> type;
+};
+
+template<size_t _Np> struct __parity;
+template<size_t _Np> struct __make : __parity<_Np % 8>::template __pmake<_Np> {};
+
+template<> struct __make<0> { typedef __integer_sequence<size_t> type; };
+template<> struct __make<1> { typedef __integer_sequence<size_t, 0> type; };
+template<> struct __make<2> { typedef __integer_sequence<size_t, 0, 1> type; };
+template<> struct __make<3> { typedef __integer_sequence<size_t, 0, 1, 2> type; };
+template<> struct __make<4> { typedef __integer_sequence<size_t, 0, 1, 2, 3> type; };
+template<> struct __make<5> { typedef __integer_sequence<size_t, 0, 1, 2, 3, 4> type; };
+template<> struct __make<6> { typedef __integer_sequence<size_t, 0, 1, 2, 3, 4, 5> type; };
+template<> struct __make<7> { typedef __integer_sequence<size_t, 0, 1, 2, 3, 4, 5, 6> type; };
+
+template<> struct __parity<0> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type> {}; };
+template<> struct __parity<1> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 1> {}; };
+template<> struct __parity<2> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 2, _Np - 1> {}; };
+template<> struct __parity<3> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 3, _Np - 2, _Np - 1> {}; };
+template<> struct __parity<4> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 4, _Np - 3, _Np - 2, _Np - 1> {}; };
+template<> struct __parity<5> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 5, _Np - 4, _Np - 3, _Np - 2, _Np - 1> {}; };
+template<> struct __parity<6> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 6, _Np - 5, _Np - 4, _Np - 3, _Np - 2, _Np - 1> {}; };
+template<> struct __parity<7> { template<size_t _Np> struct __pmake : __repeat<typename __make<_Np / 8>::type, _Np - 7, _Np - 6, _Np - 5, _Np - 4, _Np - 3, _Np - 2, _Np - 1> {}; };
+
+} // namespace detail
+
+#endif  // !__has_builtin(__make_integer_seq) || defined(_LIBCUDACXX_TESTING_FALLBACK_MAKE_INTEGER_SEQUENCE)
+
+#if __has_builtin(__make_integer_seq)
+template <size_t _Ep, size_t _Sp>
+using __make_indices_imp =
+    typename __make_integer_seq<__integer_sequence, size_t, _Ep - _Sp>::template
+    __to_tuple_indices<_Sp>;
+#else
+template <size_t _Ep, size_t _Sp>
+using __make_indices_imp =
+    typename __detail::__make<_Ep - _Sp>::type::template __to_tuple_indices<_Sp>;
+
+#endif
+
+template <size_t _Ep, size_t _Sp = 0>
+struct __make_tuple_indices
+{
+    static_assert(_Sp <= _Ep, "__make_tuple_indices input error");
+    typedef __make_indices_imp<_Ep, _Sp> type;
+};
+
+
+template <class ..._Tp> class _LIBCUDACXX_TEMPLATE_VIS tuple;
+
+template <class... _Tp> struct __tuple_like<tuple<_Tp...> > : true_type {};
+
+template <class ..._Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<tuple<_Tp...> >
+    : public integral_constant<size_t, sizeof...(_Tp)>
+{
+};
+
+template <size_t _Ip, class ..._Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, tuple<_Tp...> >::type&
+get(tuple<_Tp...>&) _NOEXCEPT;
+
+template <size_t _Ip, class ..._Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, tuple<_Tp...> >::type&
+get(const tuple<_Tp...>&) _NOEXCEPT;
+
+template <size_t _Ip, class ..._Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, tuple<_Tp...> >::type&&
+get(tuple<_Tp...>&&) _NOEXCEPT;
+
+template <size_t _Ip, class ..._Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, tuple<_Tp...> >::type&&
+get(const tuple<_Tp...>&&) _NOEXCEPT;
+
+#endif // !defined(_LIBCUDACXX_CXX03_LANG)
+
+// pair specializations
+
+template <class _T1, class _T2> struct __tuple_like<pair<_T1, _T2> > : true_type {};
+
+template <size_t _Ip, class _T1, class _T2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, pair<_T1, _T2> >::type&
+get(pair<_T1, _T2>&) _NOEXCEPT;
+
+template <size_t _Ip, class _T1, class _T2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, pair<_T1, _T2> >::type&
+get(const pair<_T1, _T2>&) _NOEXCEPT;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <size_t _Ip, class _T1, class _T2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, pair<_T1, _T2> >::type&&
+get(pair<_T1, _T2>&&) _NOEXCEPT;
+
+template <size_t _Ip, class _T1, class _T2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, pair<_T1, _T2> >::type&&
+get(const pair<_T1, _T2>&&) _NOEXCEPT;
+#endif
+
+// array specializations
+
+template <class _Tp, size_t _Size> struct _LIBCUDACXX_TEMPLATE_VIS array;
+
+template <class _Tp, size_t _Size> struct __tuple_like<array<_Tp, _Size> > : true_type {};
+
+template <size_t _Ip, class _Tp, size_t _Size>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp&
+get(array<_Tp, _Size>&) _NOEXCEPT;
+
+template <size_t _Ip, class _Tp, size_t _Size>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&
+get(const array<_Tp, _Size>&) _NOEXCEPT;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <size_t _Ip, class _Tp, size_t _Size>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp&&
+get(array<_Tp, _Size>&&) _NOEXCEPT;
+
+template <size_t _Ip, class _Tp, size_t _Size>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&&
+get(const array<_Tp, _Size>&&) _NOEXCEPT;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+// __tuple_types
+
+template <class ..._Tp> struct __tuple_types {};
+
+#if !__has_builtin(__type_pack_element)
+
+namespace __indexer_detail {
+
+template <size_t _Idx, class ..._Types>
+struct _nth_of;
+
+template <class _Head, class ..._Tail>
+struct _nth_of<0, _Head, _Tail...> {
+    using type = _Head;
+};
+
+template <size_t _Idx, class _Head, class ..._Tail>
+struct _nth_of<_Idx, _Head, _Tail...> {
+    using type = typename _nth_of<_Idx-1, _Tail...>::type;
+};
+
+template <size_t _Idx, class ..._Types>
+struct nth_of {
+    static_assert(_Idx < sizeof...(_Types), "");
+    using _impl = _nth_of<_Idx, _Types...>;
+    using type = typename _impl::type;
+};
+
+} // namespace __indexer_detail
+
+template <size_t _Idx, class ..._Types>
+using __type_pack_element _LIBCUDACXX_NODEBUG_TYPE = typename __indexer_detail::nth_of<_Idx, _Types...>::type;
+#endif
+
+template <size_t _Ip, class ..._Types>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, __tuple_types<_Types...>>
+{
+    static_assert(_Ip < sizeof...(_Types), "tuple_element index out of range");
+    typedef _LIBCUDACXX_NODEBUG_TYPE __type_pack_element<_Ip, _Types...> type;
+};
+
+
+template <class ..._Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<__tuple_types<_Tp...> >
+    : public integral_constant<size_t, sizeof...(_Tp)>
+{
+};
+
+template <class... _Tp> struct __tuple_like<__tuple_types<_Tp...> > : true_type {};
+
+template <bool _ApplyLV, bool _ApplyConst, bool _ApplyVolatile>
+struct __apply_cv_mf;
+template <>
+struct __apply_cv_mf<false, false, false> {
+  template <class _Tp> using __apply = _Tp;
+};
+template <>
+struct __apply_cv_mf<false, true, false> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE  = const _Tp;
+};
+template <>
+struct __apply_cv_mf<false, false, true> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE  = volatile _Tp;
+};
+template <>
+struct __apply_cv_mf<false, true, true> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE  = const volatile _Tp;
+};
+template <>
+struct __apply_cv_mf<true, false, false> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE  = _Tp&;
+};
+template <>
+struct __apply_cv_mf<true, true, false> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE  = const _Tp&;
+};
+template <>
+struct __apply_cv_mf<true, false, true> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE  = volatile _Tp&;
+};
+template <>
+struct __apply_cv_mf<true, true, true> {
+  template <class _Tp> using __apply _LIBCUDACXX_NODEBUG_TYPE = const volatile _Tp&;
+};
+template <class _Tp, class _RawTp = typename remove_reference<_Tp>::type>
+using __apply_cv_t _LIBCUDACXX_NODEBUG_TYPE  = __apply_cv_mf<
+    is_lvalue_reference<_Tp>::value,
+    is_const<_RawTp>::value,
+    is_volatile<_RawTp>::value>;
+
+// __make_tuple_types
+
+// __make_tuple_types<_Tuple<_Types...>, _Ep, _Sp>::type is a
+// __tuple_types<_Types...> using only those _Types in the range [_Sp, _Ep).
+// _Sp defaults to 0 and _Ep defaults to tuple_size<_Tuple>.  If _Tuple is a
+// lvalue_reference type, then __tuple_types<_Types&...> is the result.
+
+template <class _TupleTypes, class _TupleIndices>
+struct __make_tuple_types_flat;
+
+template <template <class...> class _Tuple, class ..._Types, size_t ..._Idx>
+struct __make_tuple_types_flat<_Tuple<_Types...>, __tuple_indices<_Idx...>> {
+  // Specialization for pair, tuple, and __tuple_types
+  template <class _Tp, class _ApplyFn = __apply_cv_t<_Tp>>
+  using __apply_quals _LIBCUDACXX_NODEBUG_TYPE = __tuple_types<
+      typename _ApplyFn::template __apply<__type_pack_element<_Idx, _Types...>>...
+    >;
+};
+
+template <class _Vt, size_t _Np, size_t ..._Idx>
+struct __make_tuple_types_flat<array<_Vt, _Np>, __tuple_indices<_Idx...>> {
+  template <size_t>
+  using __value_type = _Vt;
+  template <class _Tp, class _ApplyFn = __apply_cv_t<_Tp>>
+  using __apply_quals = __tuple_types<
+      typename _ApplyFn::template __apply<__value_type<_Idx>>...
+    >;
+};
+
+template <class _Tp, size_t _Ep = tuple_size<typename remove_reference<_Tp>::type>::value,
+          size_t _Sp = 0,
+          bool _SameSize = (_Ep == tuple_size<typename remove_reference<_Tp>::type>::value)>
+struct __make_tuple_types
+{
+    static_assert(_Sp <= _Ep, "__make_tuple_types input error");
+    using _RawTp = typename remove_cv<typename remove_reference<_Tp>::type>::type;
+    using _Maker = __make_tuple_types_flat<_RawTp, typename __make_tuple_indices<_Ep, _Sp>::type>;
+    using type = typename _Maker::template __apply_quals<_Tp>;
+};
+
+template <class ..._Types, size_t _Ep>
+struct __make_tuple_types<tuple<_Types...>, _Ep, 0, true> {
+  typedef _LIBCUDACXX_NODEBUG_TYPE __tuple_types<_Types...> type;
+};
+
+template <class ..._Types, size_t _Ep>
+struct __make_tuple_types<__tuple_types<_Types...>, _Ep, 0, true> {
+  typedef _LIBCUDACXX_NODEBUG_TYPE __tuple_types<_Types...> type;
+};
+
+template <bool ..._Preds>
+struct __all_dummy;
+
+template <bool ..._Pred>
+using __all = _IsSame<__all_dummy<_Pred...>, __all_dummy<((void)_Pred, true)...>>;
+
+struct __tuple_sfinae_base {
+  template <class, class>
+  struct __test_size : false_type {};
+
+  template <class ..._Tp, class ..._Up>
+  struct __test_size<__tuple_types<_Tp...>, __tuple_types<_Up...>>
+    : _BoolConstant<sizeof...(_Tp) == sizeof...(_Up)> {};
+
+  template <template <class, class...> class,
+            class _Tp, class _Up, bool = __test_size<_Tp, _Up>::value>
+  struct __test : false_type {};
+
+  template <template <class, class...> class _Trait,
+            class ..._LArgs, class ..._RArgs>
+  struct __test<_Trait, __tuple_types<_LArgs...>, __tuple_types<_RArgs...>, true>
+      : __all<_Trait<_LArgs, _RArgs>::value...> {};
+
+  template <class _FromArgs, class _ToArgs>
+  using __constructible = __test<is_constructible, _ToArgs, _FromArgs>;
+  template <class _FromArgs, class _ToArgs>
+  using __convertible = __test<is_convertible, _FromArgs, _ToArgs>;
+  template <class _FromArgs, class _ToArgs>
+  using __assignable = __test<is_assignable, _ToArgs, _FromArgs>;
+};
+
+// __tuple_convertible
+
+template <class _Tp, class _Up, bool = __tuple_like<typename remove_reference<_Tp>::type>::value,
+                                bool = __tuple_like<_Up>::value>
+struct __tuple_convertible
+    : public false_type {};
+
+template <class _Tp, class _Up>
+struct __tuple_convertible<_Tp, _Up, true, true>
+    : public __tuple_sfinae_base::__convertible<
+      typename __make_tuple_types<_Tp>::type
+    , typename __make_tuple_types<_Up>::type
+    >
+{};
+
+// __tuple_constructible
+
+template <class _Tp, class _Up, bool = __tuple_like<typename remove_reference<_Tp>::type>::value,
+                                bool = __tuple_like<_Up>::value>
+struct __tuple_constructible
+    : public false_type {};
+
+template <class _Tp, class _Up>
+struct __tuple_constructible<_Tp, _Up, true, true>
+    : public __tuple_sfinae_base::__constructible<
+      typename __make_tuple_types<_Tp>::type
+    , typename __make_tuple_types<_Up>::type
+    >
+{};
+
+// __tuple_assignable
+
+template <class _Tp, class _Up, bool = __tuple_like<typename remove_reference<_Tp>::type>::value,
+                                bool = __tuple_like<_Up>::value>
+struct __tuple_assignable
+    : public false_type {};
+
+template <class _Tp, class _Up>
+struct __tuple_assignable<_Tp, _Up, true, true>
+    : public __tuple_sfinae_base::__assignable<
+      typename __make_tuple_types<_Tp>::type
+    , typename __make_tuple_types<_Up&>::type
+    >
+{};
+
+
+template <size_t _Ip, class ..._Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, tuple<_Tp...> >
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename tuple_element<_Ip, __tuple_types<_Tp...> >::type type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <size_t _Ip, class ..._Tp>
+using tuple_element_t _LIBCUDACXX_NODEBUG_TYPE  = typename tuple_element <_Ip, _Tp...>::type;
+#endif
+
+template <bool _IsTuple, class _SizeTrait, size_t _Expected>
+struct __tuple_like_with_size_imp : false_type {};
+
+template <class _SizeTrait, size_t _Expected>
+struct __tuple_like_with_size_imp<true, _SizeTrait, _Expected>
+    : integral_constant<bool, _SizeTrait::value == _Expected> {};
+
+template <class _Tuple, size_t _ExpectedSize,
+          class _RawTuple = typename __uncvref<_Tuple>::type>
+using __tuple_like_with_size _LIBCUDACXX_NODEBUG_TYPE = __tuple_like_with_size_imp<
+                                   __tuple_like<_RawTuple>::value,
+                                   tuple_size<_RawTuple>, _ExpectedSize
+                              >;
+
+struct _LIBCUDACXX_TYPE_VIS __check_tuple_constructor_fail {
+    template <int&...>
+    using __enable_explicit_default = false_type;
+    template <int&...>
+    using __enable_implicit_default = false_type;
+    template <class ...>
+    using __enable_explicit = false_type;
+    template <class ...>
+    using __enable_implicit = false_type;
+    template <class ...>
+    using __enable_assign = false_type;
+};
+#endif // !defined(_LIBCUDACXX_CXX03_LANG)
+#if _LIBCUDACXX_STD_VER > 14
+
+template <bool _CanCopy, bool _CanMove>
+struct __sfinae_ctor_base {};
+template <>
+struct __sfinae_ctor_base<false, false> {
+  __sfinae_ctor_base() = default;
+  __sfinae_ctor_base(__sfinae_ctor_base const&) = delete;
+  __sfinae_ctor_base(__sfinae_ctor_base &&) = delete;
+  __sfinae_ctor_base& operator=(__sfinae_ctor_base const&) = default;
+  __sfinae_ctor_base& operator=(__sfinae_ctor_base&&) = default;
+};
+template <>
+struct __sfinae_ctor_base<true, false> {
+  __sfinae_ctor_base() = default;
+  __sfinae_ctor_base(__sfinae_ctor_base const&) = default;
+  __sfinae_ctor_base(__sfinae_ctor_base &&) = delete;
+  __sfinae_ctor_base& operator=(__sfinae_ctor_base const&) = default;
+  __sfinae_ctor_base& operator=(__sfinae_ctor_base&&) = default;
+};
+template <>
+struct __sfinae_ctor_base<false, true> {
+  __sfinae_ctor_base() = default;
+  __sfinae_ctor_base(__sfinae_ctor_base const&) = delete;
+  __sfinae_ctor_base(__sfinae_ctor_base &&) = default;
+  __sfinae_ctor_base& operator=(__sfinae_ctor_base const&) = default;
+  __sfinae_ctor_base& operator=(__sfinae_ctor_base&&) = default;
+};
+
+template <bool _CanCopy, bool _CanMove>
+struct __sfinae_assign_base {};
+template <>
+struct __sfinae_assign_base<false, false> {
+  __sfinae_assign_base() = default;
+  __sfinae_assign_base(__sfinae_assign_base const&) = default;
+  __sfinae_assign_base(__sfinae_assign_base &&) = default;
+  __sfinae_assign_base& operator=(__sfinae_assign_base const&) = delete;
+  __sfinae_assign_base& operator=(__sfinae_assign_base&&) = delete;
+};
+template <>
+struct __sfinae_assign_base<true, false> {
+  __sfinae_assign_base() = default;
+  __sfinae_assign_base(__sfinae_assign_base const&) = default;
+  __sfinae_assign_base(__sfinae_assign_base &&) = default;
+  __sfinae_assign_base& operator=(__sfinae_assign_base const&) = default;
+  __sfinae_assign_base& operator=(__sfinae_assign_base&&) = delete;
+};
+template <>
+struct __sfinae_assign_base<false, true> {
+  __sfinae_assign_base() = default;
+  __sfinae_assign_base(__sfinae_assign_base const&) = default;
+  __sfinae_assign_base(__sfinae_assign_base &&) = default;
+  __sfinae_assign_base& operator=(__sfinae_assign_base const&) = delete;
+  __sfinae_assign_base& operator=(__sfinae_assign_base&&) = default;
+};
+#endif // _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX___TUPLE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__undef_macros b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__undef_macros
new file mode 100644
index 000000000000..37594493df25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/__undef_macros
@@ -0,0 +1,33 @@
+// -*- C++ -*-
+//===------------------------ __undef_macros ------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+
+#ifdef min
+#if !defined(_LIBCUDACXX_DISABLE_MACRO_CONFLICT_WARNINGS)
+#if defined(_LIBCUDACXX_WARNING)
+_LIBCUDACXX_WARNING("macro min is incompatible with C++.  Try #define NOMINMAX "
+                "before any Windows header. #undefing min")
+#else
+#warning: macro min is incompatible with C++.  #undefing min
+#endif
+#endif
+#undef min
+#endif
+
+#ifdef max
+#if !defined(_LIBCUDACXX_DISABLE_MACRO_CONFLICT_WARNINGS)
+#if defined(_LIBCUDACXX_WARNING)
+_LIBCUDACXX_WARNING("macro max is incompatible with C++.  Try #define NOMINMAX "
+                "before any Windows header. #undefing max")
+#else
+#warning: macro max is incompatible with C++.  #undefing max
+#endif
+#endif
+#undef max
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm
new file mode 100644
index 000000000000..182fd51d7973
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/algorithm
@@ -0,0 +1,5769 @@
+// -*- C++ -*-
+//===-------------------------- algorithm ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ALGORITHM
+#define _LIBCUDACXX_ALGORITHM
+
+/*
+    algorithm synopsis
+
+#include <initializer_list>
+
+namespace std
+{
+
+template <class InputIterator, class Predicate>
+    constexpr bool     // constexpr in C++20
+    all_of(InputIterator first, InputIterator last, Predicate pred);
+
+template <class InputIterator, class Predicate>
+    constexpr bool     // constexpr in C++20
+    any_of(InputIterator first, InputIterator last, Predicate pred);
+
+template <class InputIterator, class Predicate>
+    constexpr bool     // constexpr in C++20
+    none_of(InputIterator first, InputIterator last, Predicate pred);
+
+template <class InputIterator, class Function>
+    constexpr Function          // constexpr in C++20
+    for_each(InputIterator first, InputIterator last, Function f);
+
+template<class InputIterator, class Size, class Function>
+    constexpr InputIterator     // constexpr in C++20
+    for_each_n(InputIterator first, Size n, Function f); // C++17
+
+template <class InputIterator, class T>
+    constexpr InputIterator     // constexpr in C++20
+    find(InputIterator first, InputIterator last, const T& value);
+
+template <class InputIterator, class Predicate>
+    constexpr InputIterator     // constexpr in C++20
+    find_if(InputIterator first, InputIterator last, Predicate pred);
+
+template<class InputIterator, class Predicate>
+    InputIterator               // constexpr in C++20
+    find_if_not(InputIterator first, InputIterator last, Predicate pred);
+
+template <class ForwardIterator1, class ForwardIterator2>
+    ForwardIterator1            // constexpr in C++20
+    find_end(ForwardIterator1 first1, ForwardIterator1 last1,
+             ForwardIterator2 first2, ForwardIterator2 last2);
+
+template <class ForwardIterator1, class ForwardIterator2, class BinaryPredicate>
+    ForwardIterator1            // constexpr in C++20
+    find_end(ForwardIterator1 first1, ForwardIterator1 last1,
+             ForwardIterator2 first2, ForwardIterator2 last2, BinaryPredicate pred);
+
+template <class ForwardIterator1, class ForwardIterator2>
+    constexpr ForwardIterator1  // constexpr in C++20
+    find_first_of(ForwardIterator1 first1, ForwardIterator1 last1,
+                  ForwardIterator2 first2, ForwardIterator2 last2);
+
+template <class ForwardIterator1, class ForwardIterator2, class BinaryPredicate>
+    constexpr ForwardIterator1  // constexpr in C++20
+    find_first_of(ForwardIterator1 first1, ForwardIterator1 last1,
+                  ForwardIterator2 first2, ForwardIterator2 last2, BinaryPredicate pred);
+
+template <class ForwardIterator>
+    constexpr ForwardIterator   // constexpr in C++20
+    adjacent_find(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class BinaryPredicate>
+    constexpr ForwardIterator   // constexpr in C++20
+    adjacent_find(ForwardIterator first, ForwardIterator last, BinaryPredicate pred);
+
+template <class InputIterator, class T>
+    constexpr typename iterator_traits<InputIterator>::difference_type  // constexpr in C++20
+    count(InputIterator first, InputIterator last, const T& value);
+
+template <class InputIterator, class Predicate>
+    constexpr typename iterator_traits<InputIterator>::difference_type // constexpr in C++20
+    count_if(InputIterator first, InputIterator last, Predicate pred);
+
+template <class InputIterator1, class InputIterator2>
+    constexpr pair<InputIterator1, InputIterator2>   // constexpr in C++20
+    mismatch(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2);
+
+template <class InputIterator1, class InputIterator2>
+    constexpr pair<InputIterator1, InputIterator2>   // constexpr in C++20
+    mismatch(InputIterator1 first1, InputIterator1 last1,
+             InputIterator2 first2, InputIterator2 last2); // **C++14**
+
+template <class InputIterator1, class InputIterator2, class BinaryPredicate>
+    constexpr pair<InputIterator1, InputIterator2>   // constexpr in C++20
+    mismatch(InputIterator1 first1, InputIterator1 last1,
+             InputIterator2 first2, BinaryPredicate pred);
+
+template <class InputIterator1, class InputIterator2, class BinaryPredicate>
+    constexpr pair<InputIterator1, InputIterator2>   // constexpr in C++20
+    mismatch(InputIterator1 first1, InputIterator1 last1,
+             InputIterator2 first2, InputIterator2 last2,
+             BinaryPredicate pred); // **C++14**
+
+template <class InputIterator1, class InputIterator2>
+    constexpr bool      // constexpr in C++20
+    equal(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2);
+
+template <class InputIterator1, class InputIterator2>
+    constexpr bool      // constexpr in C++20
+    equal(InputIterator1 first1, InputIterator1 last1,
+          InputIterator2 first2, InputIterator2 last2); // **C++14**
+
+template <class InputIterator1, class InputIterator2, class BinaryPredicate>
+    constexpr bool      // constexpr in C++20
+    equal(InputIterator1 first1, InputIterator1 last1,
+          InputIterator2 first2, BinaryPredicate pred);
+
+template <class InputIterator1, class InputIterator2, class BinaryPredicate>
+    constexpr bool      // constexpr in C++20
+    equal(InputIterator1 first1, InputIterator1 last1,
+          InputIterator2 first2, InputIterator2 last2,
+          BinaryPredicate pred); // **C++14**
+
+template<class ForwardIterator1, class ForwardIterator2>
+    constexpr bool      // constexpr in C++20
+    is_permutation(ForwardIterator1 first1, ForwardIterator1 last1,
+                   ForwardIterator2 first2);
+
+template<class ForwardIterator1, class ForwardIterator2>
+    constexpr bool      // constexpr in C++20
+    is_permutation(ForwardIterator1 first1, ForwardIterator1 last1,
+                   ForwardIterator2 first2, ForwardIterator2 last2); // **C++14**
+
+template<class ForwardIterator1, class ForwardIterator2, class BinaryPredicate>
+    constexpr bool      // constexpr in C++20
+    is_permutation(ForwardIterator1 first1, ForwardIterator1 last1,
+                   ForwardIterator2 first2, BinaryPredicate pred);
+
+template<class ForwardIterator1, class ForwardIterator2, class BinaryPredicate>
+    constexpr bool      // constexpr in C++20
+    is_permutation(ForwardIterator1 first1, ForwardIterator1 last1,
+                   ForwardIterator2 first2, ForwardIterator2 last2,
+                   BinaryPredicate pred);  // **C++14**
+
+template <class ForwardIterator1, class ForwardIterator2>
+    constexpr ForwardIterator1      // constexpr in C++20
+    search(ForwardIterator1 first1, ForwardIterator1 last1,
+           ForwardIterator2 first2, ForwardIterator2 last2);
+
+template <class ForwardIterator1, class ForwardIterator2, class BinaryPredicate>
+    constexpr ForwardIterator1      // constexpr in C++20
+    search(ForwardIterator1 first1, ForwardIterator1 last1,
+           ForwardIterator2 first2, ForwardIterator2 last2, BinaryPredicate pred);
+
+template <class ForwardIterator, class Size, class T>
+    constexpr ForwardIterator       // constexpr in C++20
+    search_n(ForwardIterator first, ForwardIterator last, Size count, const T& value);
+
+template <class ForwardIterator, class Size, class T, class BinaryPredicate>
+    constexpr ForwardIterator       // constexpr in C++20
+    search_n(ForwardIterator first, ForwardIterator last,
+             Size count, const T& value, BinaryPredicate pred);
+
+template <class InputIterator, class OutputIterator>
+    OutputIterator
+    copy(InputIterator first, InputIterator last, OutputIterator result);
+
+template<class InputIterator, class OutputIterator, class Predicate>
+    OutputIterator
+    copy_if(InputIterator first, InputIterator last,
+            OutputIterator result, Predicate pred);
+
+template<class InputIterator, class Size, class OutputIterator>
+    OutputIterator
+    copy_n(InputIterator first, Size n, OutputIterator result);
+
+template <class BidirectionalIterator1, class BidirectionalIterator2>
+    BidirectionalIterator2
+    copy_backward(BidirectionalIterator1 first, BidirectionalIterator1 last,
+                  BidirectionalIterator2 result);
+
+template <class ForwardIterator1, class ForwardIterator2>
+    ForwardIterator2
+    swap_ranges(ForwardIterator1 first1, ForwardIterator1 last1, ForwardIterator2 first2);
+
+template <class ForwardIterator1, class ForwardIterator2>
+    void
+    iter_swap(ForwardIterator1 a, ForwardIterator2 b);
+
+template <class InputIterator, class OutputIterator, class UnaryOperation>
+    constexpr OutputIterator      // constexpr in C++20
+    transform(InputIterator first, InputIterator last, OutputIterator result, UnaryOperation op);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator, class BinaryOperation>
+    constexpr OutputIterator      // constexpr in C++20
+    transform(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2,
+              OutputIterator result, BinaryOperation binary_op);
+
+template <class ForwardIterator, class T>
+    constexpr void      // constexpr in C++20
+    replace(ForwardIterator first, ForwardIterator last, const T& old_value, const T& new_value);
+
+template <class ForwardIterator, class Predicate, class T>
+    constexpr void      // constexpr in C++20
+    replace_if(ForwardIterator first, ForwardIterator last, Predicate pred, const T& new_value);
+
+template <class InputIterator, class OutputIterator, class T>
+    constexpr OutputIterator      // constexpr in C++20
+    replace_copy(InputIterator first, InputIterator last, OutputIterator result,
+                 const T& old_value, const T& new_value);
+
+template <class InputIterator, class OutputIterator, class Predicate, class T>
+    constexpr OutputIterator      // constexpr in C++20
+    replace_copy_if(InputIterator first, InputIterator last, OutputIterator result, Predicate pred, const T& new_value);
+
+template <class ForwardIterator, class T>
+    constexpr void      // constexpr in C++20
+    fill(ForwardIterator first, ForwardIterator last, const T& value);
+
+template <class OutputIterator, class Size, class T>
+    constexpr OutputIterator      // constexpr in C++20
+    fill_n(OutputIterator first, Size n, const T& value);
+
+template <class ForwardIterator, class Generator>
+    constexpr void      // constexpr in C++20
+    generate(ForwardIterator first, ForwardIterator last, Generator gen);
+
+template <class OutputIterator, class Size, class Generator>
+    constexpr OutputIterator      // constexpr in C++20
+    generate_n(OutputIterator first, Size n, Generator gen);
+
+template <class ForwardIterator, class T>
+    constexpr ForwardIterator     // constexpr in C++20
+    remove(ForwardIterator first, ForwardIterator last, const T& value);
+
+template <class ForwardIterator, class Predicate>
+    constexpr ForwardIterator     // constexpr in C++20
+    remove_if(ForwardIterator first, ForwardIterator last, Predicate pred);
+
+template <class InputIterator, class OutputIterator, class T>
+    constexpr OutputIterator     // constexpr in C++20
+    remove_copy(InputIterator first, InputIterator last, OutputIterator result, const T& value);
+
+template <class InputIterator, class OutputIterator, class Predicate>
+    constexpr OutputIterator     // constexpr in C++20
+    remove_copy_if(InputIterator first, InputIterator last, OutputIterator result, Predicate pred);
+
+template <class ForwardIterator>
+    ForwardIterator
+    unique(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class BinaryPredicate>
+    ForwardIterator
+    unique(ForwardIterator first, ForwardIterator last, BinaryPredicate pred);
+
+template <class InputIterator, class OutputIterator>
+    OutputIterator
+    unique_copy(InputIterator first, InputIterator last, OutputIterator result);
+
+template <class InputIterator, class OutputIterator, class BinaryPredicate>
+    OutputIterator
+    unique_copy(InputIterator first, InputIterator last, OutputIterator result, BinaryPredicate pred);
+
+template <class BidirectionalIterator>
+    void
+    reverse(BidirectionalIterator first, BidirectionalIterator last);
+
+template <class BidirectionalIterator, class OutputIterator>
+    constexpr OutputIterator       // constexpr in C++20
+    reverse_copy(BidirectionalIterator first, BidirectionalIterator last, OutputIterator result);
+
+template <class ForwardIterator>
+    ForwardIterator
+    rotate(ForwardIterator first, ForwardIterator middle, ForwardIterator last);
+
+template <class ForwardIterator, class OutputIterator>
+    OutputIterator
+    rotate_copy(ForwardIterator first, ForwardIterator middle, ForwardIterator last, OutputIterator result);
+
+template <class RandomAccessIterator>
+    void
+    random_shuffle(RandomAccessIterator first, RandomAccessIterator last); // deprecated in C++14, removed in C++17
+
+template <class RandomAccessIterator, class RandomNumberGenerator>
+    void
+    random_shuffle(RandomAccessIterator first, RandomAccessIterator last,
+                   RandomNumberGenerator& rand);  // deprecated in C++14, removed in C++17
+
+template<class PopulationIterator, class SampleIterator,
+         class Distance, class UniformRandomBitGenerator>
+    SampleIterator sample(PopulationIterator first, PopulationIterator last,
+                          SampleIterator out, Distance n,
+                          UniformRandomBitGenerator&& g); // C++17
+
+template<class RandomAccessIterator, class UniformRandomNumberGenerator>
+    void shuffle(RandomAccessIterator first, RandomAccessIterator last,
+                 UniformRandomNumberGenerator&& g);
+
+template <class InputIterator, class Predicate>
+    constexpr bool  // constexpr in C++20
+    is_partitioned(InputIterator first, InputIterator last, Predicate pred);
+
+template <class ForwardIterator, class Predicate>
+    ForwardIterator
+    partition(ForwardIterator first, ForwardIterator last, Predicate pred);
+
+template <class InputIterator, class OutputIterator1,
+          class OutputIterator2, class Predicate>
+    constexpr pair<OutputIterator1, OutputIterator2>   // constexpr in C++20
+    partition_copy(InputIterator first, InputIterator last,
+                   OutputIterator1 out_true, OutputIterator2 out_false,
+                   Predicate pred);
+
+template <class ForwardIterator, class Predicate>
+    ForwardIterator
+    stable_partition(ForwardIterator first, ForwardIterator last, Predicate pred);
+
+template<class ForwardIterator, class Predicate>
+    constexpr ForwardIterator  // constexpr in C++20
+    partition_point(ForwardIterator first, ForwardIterator last, Predicate pred);
+
+template <class ForwardIterator>
+    constexpr bool  // constexpr in C++20
+    is_sorted(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class Compare>
+    bool
+    is_sorted(ForwardIterator first, ForwardIterator last, Compare comp);
+
+template<class ForwardIterator>
+    constexpr ForwardIterator    // constexpr in C++20
+    is_sorted_until(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class Compare>
+    constexpr ForwardIterator    // constexpr in C++20
+    is_sorted_until(ForwardIterator first, ForwardIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    sort(RandomAccessIterator first, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    sort(RandomAccessIterator first, RandomAccessIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    stable_sort(RandomAccessIterator first, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    stable_sort(RandomAccessIterator first, RandomAccessIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    partial_sort(RandomAccessIterator first, RandomAccessIterator middle, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    partial_sort(RandomAccessIterator first, RandomAccessIterator middle, RandomAccessIterator last, Compare comp);
+
+template <class InputIterator, class RandomAccessIterator>
+    RandomAccessIterator
+    partial_sort_copy(InputIterator first, InputIterator last,
+                      RandomAccessIterator result_first, RandomAccessIterator result_last);
+
+template <class InputIterator, class RandomAccessIterator, class Compare>
+    RandomAccessIterator
+    partial_sort_copy(InputIterator first, InputIterator last,
+                      RandomAccessIterator result_first, RandomAccessIterator result_last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    nth_element(RandomAccessIterator first, RandomAccessIterator nth, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    nth_element(RandomAccessIterator first, RandomAccessIterator nth, RandomAccessIterator last, Compare comp);
+
+template <class ForwardIterator, class T>
+    constexpr ForwardIterator                         // constexpr in C++20
+    lower_bound(ForwardIterator first, ForwardIterator last, const T& value);
+
+template <class ForwardIterator, class T, class Compare>
+    constexpr ForwardIterator                         // constexpr in C++20
+    lower_bound(ForwardIterator first, ForwardIterator last, const T& value, Compare comp);
+
+template <class ForwardIterator, class T>
+    constexpr ForwardIterator                         // constexpr in C++20
+    upper_bound(ForwardIterator first, ForwardIterator last, const T& value);
+
+template <class ForwardIterator, class T, class Compare>
+    constexpr ForwardIterator                         // constexpr in C++20
+    upper_bound(ForwardIterator first, ForwardIterator last, const T& value, Compare comp);
+
+template <class ForwardIterator, class T>
+    constexpr pair<ForwardIterator, ForwardIterator>  // constexpr in C++20
+    equal_range(ForwardIterator first, ForwardIterator last, const T& value);
+
+template <class ForwardIterator, class T, class Compare>
+    constexpr pair<ForwardIterator, ForwardIterator>  // constexpr in C++20
+    equal_range(ForwardIterator first, ForwardIterator last, const T& value, Compare comp);
+
+template <class ForwardIterator, class T>
+    constexpr bool                                    // constexpr in C++20
+    binary_search(ForwardIterator first, ForwardIterator last, const T& value);
+
+template <class ForwardIterator, class T, class Compare>
+    constexpr bool                                    // constexpr in C++20
+    binary_search(ForwardIterator first, ForwardIterator last, const T& value, Compare comp);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator>
+    OutputIterator
+    merge(InputIterator1 first1, InputIterator1 last1,
+          InputIterator2 first2, InputIterator2 last2, OutputIterator result);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator, class Compare>
+    OutputIterator
+    merge(InputIterator1 first1, InputIterator1 last1,
+          InputIterator2 first2, InputIterator2 last2, OutputIterator result, Compare comp);
+
+template <class BidirectionalIterator>
+    void
+    inplace_merge(BidirectionalIterator first, BidirectionalIterator middle, BidirectionalIterator last);
+
+template <class BidirectionalIterator, class Compare>
+    void
+    inplace_merge(BidirectionalIterator first, BidirectionalIterator middle, BidirectionalIterator last, Compare comp);
+
+template <class InputIterator1, class InputIterator2>
+    constexpr bool                                    // constexpr in C++20
+    includes(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, InputIterator2 last2);
+
+template <class InputIterator1, class InputIterator2, class Compare>
+    constexpr bool                                    // constexpr in C++20
+    includes(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, InputIterator2 last2, Compare comp);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator>
+    OutputIterator
+    set_union(InputIterator1 first1, InputIterator1 last1,
+              InputIterator2 first2, InputIterator2 last2, OutputIterator result);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator, class Compare>
+    OutputIterator
+    set_union(InputIterator1 first1, InputIterator1 last1,
+              InputIterator2 first2, InputIterator2 last2, OutputIterator result, Compare comp);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator>
+    constexpr OutputIterator                         // constexpr in C++20
+    set_intersection(InputIterator1 first1, InputIterator1 last1,
+                     InputIterator2 first2, InputIterator2 last2, OutputIterator result);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator, class Compare>
+    constexpr OutputIterator                         // constexpr in C++20
+    set_intersection(InputIterator1 first1, InputIterator1 last1,
+                     InputIterator2 first2, InputIterator2 last2, OutputIterator result, Compare comp);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator>
+    OutputIterator
+    set_difference(InputIterator1 first1, InputIterator1 last1,
+                   InputIterator2 first2, InputIterator2 last2, OutputIterator result);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator, class Compare>
+    OutputIterator
+    set_difference(InputIterator1 first1, InputIterator1 last1,
+                   InputIterator2 first2, InputIterator2 last2, OutputIterator result, Compare comp);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator>
+    OutputIterator
+    set_symmetric_difference(InputIterator1 first1, InputIterator1 last1,
+                             InputIterator2 first2, InputIterator2 last2, OutputIterator result);
+
+template <class InputIterator1, class InputIterator2, class OutputIterator, class Compare>
+    OutputIterator
+    set_symmetric_difference(InputIterator1 first1, InputIterator1 last1,
+                             InputIterator2 first2, InputIterator2 last2, OutputIterator result, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    push_heap(RandomAccessIterator first, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    push_heap(RandomAccessIterator first, RandomAccessIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    pop_heap(RandomAccessIterator first, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    pop_heap(RandomAccessIterator first, RandomAccessIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    make_heap(RandomAccessIterator first, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    make_heap(RandomAccessIterator first, RandomAccessIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    void
+    sort_heap(RandomAccessIterator first, RandomAccessIterator last);
+
+template <class RandomAccessIterator, class Compare>
+    void
+    sort_heap(RandomAccessIterator first, RandomAccessIterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    constexpr bool   // constexpr in C++20
+    is_heap(RandomAccessIterator first, RandomAccessiterator last);
+
+template <class RandomAccessIterator, class Compare>
+    constexpr bool   // constexpr in C++20
+    is_heap(RandomAccessIterator first, RandomAccessiterator last, Compare comp);
+
+template <class RandomAccessIterator>
+    constexpr RandomAccessIterator   // constexpr in C++20
+    is_heap_until(RandomAccessIterator first, RandomAccessiterator last);
+
+template <class RandomAccessIterator, class Compare>
+    constexpr RandomAccessIterator   // constexpr in C++20
+    is_heap_until(RandomAccessIterator first, RandomAccessiterator last, Compare comp);
+
+template <class ForwardIterator>
+    ForwardIterator
+    min_element(ForwardIterator first, ForwardIterator last);  // constexpr in C++14
+
+template <class ForwardIterator, class Compare>
+    ForwardIterator
+    min_element(ForwardIterator first, ForwardIterator last, Compare comp);  // constexpr in C++14
+
+template <class T>
+    const T&
+    min(const T& a, const T& b);  // constexpr in C++14
+
+template <class T, class Compare>
+    const T&
+    min(const T& a, const T& b, Compare comp);  // constexpr in C++14
+
+template<class T>
+    T
+    min(::std::initializer_list<T> t);  // constexpr in C++14
+
+template<class T, class Compare>
+    T
+    min(::std::initializer_list<T> t, Compare comp);  // constexpr in C++14
+
+template<class T>
+    constexpr const T& clamp( const T& v, const T& lo, const T& hi );               // C++17
+
+template<class T, class Compare>
+    constexpr const T& clamp( const T& v, const T& lo, const T& hi, Compare comp ); // C++17
+
+template <class ForwardIterator>
+    ForwardIterator
+    max_element(ForwardIterator first, ForwardIterator last);  // constexpr in C++14
+
+template <class ForwardIterator, class Compare>
+    ForwardIterator
+    max_element(ForwardIterator first, ForwardIterator last, Compare comp);  // constexpr in C++14
+
+template <class T>
+    const T&
+    max(const T& a, const T& b); // constexpr in C++14
+
+template <class T, class Compare>
+    const T&
+    max(const T& a, const T& b, Compare comp);  // constexpr in C++14
+
+template<class T>
+    T
+    max(::std::initializer_list<T> t);  // constexpr in C++14
+
+template<class T, class Compare>
+    T
+    max(::std::initializer_list<T> t, Compare comp);  // constexpr in C++14
+
+template<class ForwardIterator>
+    pair<ForwardIterator, ForwardIterator>
+    minmax_element(ForwardIterator first, ForwardIterator last);   // constexpr in C++14
+
+template<class ForwardIterator, class Compare>
+    pair<ForwardIterator, ForwardIterator>
+    minmax_element(ForwardIterator first, ForwardIterator last, Compare comp);   // constexpr in C++14
+
+template<class T>
+    pair<const T&, const T&>
+    minmax(const T& a, const T& b);  // constexpr in C++14
+
+template<class T, class Compare>
+    pair<const T&, const T&>
+    minmax(const T& a, const T& b, Compare comp);  // constexpr in C++14
+
+template<class T>
+    pair<T, T>
+    minmax(::std::initializer_list<T> t);  // constexpr in C++14
+
+template<class T, class Compare>
+    pair<T, T>
+    minmax(::std::initializer_list<T> t, Compare comp);  // constexpr in C++14
+
+template <class InputIterator1, class InputIterator2>
+    constexpr bool     // constexpr in C++20
+    lexicographical_compare(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, InputIterator2 last2);
+
+template <class InputIterator1, class InputIterator2, class Compare>
+    constexpr bool     // constexpr in C++20
+    lexicographical_compare(InputIterator1 first1, InputIterator1 last1,
+                            InputIterator2 first2, InputIterator2 last2, Compare comp);
+
+template <class BidirectionalIterator>
+    bool
+    next_permutation(BidirectionalIterator first, BidirectionalIterator last);
+
+template <class BidirectionalIterator, class Compare>
+    bool
+    next_permutation(BidirectionalIterator first, BidirectionalIterator last, Compare comp);
+
+template <class BidirectionalIterator>
+    bool
+    prev_permutation(BidirectionalIterator first, BidirectionalIterator last);
+
+template <class BidirectionalIterator, class Compare>
+    bool
+    prev_permutation(BidirectionalIterator first, BidirectionalIterator last, Compare comp);
+
+}  // std
+
+*/
+#ifndef __cuda_std__
+#include <__config>
+#include <initializer_list>
+#include <type_traits>
+#include <cstring>
+#include <utility> // needed to provide swap_ranges.
+#include <memory>
+#include <functional>
+#include <iterator>
+#include <cstddef>
+#include <bit>
+#include <version>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+#endif
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// I'd like to replace these with _CUDA_VSTD::equal_to<void>, but can't because:
+//   * That only works with C++14 and later, and
+//   * We haven't included <functional> here.
+template <class _T1, class _T2 = _T1>
+struct __equal_to
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 bool operator()(const _T1& __x, const _T1& __y) const {return __x == __y;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 bool operator()(const _T1& __x, const _T2& __y) const {return __x == __y;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 bool operator()(const _T2& __x, const _T1& __y) const {return __x == __y;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 bool operator()(const _T2& __x, const _T2& __y) const {return __x == __y;}
+};
+
+template <class _T1>
+struct __equal_to<_T1, _T1>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x == __y;}
+};
+
+template <class _T1>
+struct __equal_to<const _T1, _T1>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x == __y;}
+};
+
+template <class _T1>
+struct __equal_to<_T1, const _T1>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x == __y;}
+};
+
+template <class _T1, class _T2 = _T1>
+struct __less
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x < __y;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T2& __y) const {return __x < __y;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T2& __x, const _T1& __y) const {return __x < __y;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T2& __x, const _T2& __y) const {return __x < __y;}
+};
+
+template <class _T1>
+struct __less<_T1, _T1>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x < __y;}
+};
+
+template <class _T1>
+struct __less<const _T1, _T1>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x < __y;}
+};
+
+template <class _T1>
+struct __less<_T1, const _T1>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _T1& __x, const _T1& __y) const {return __x < __y;}
+};
+
+#ifndef __cuda_std__
+
+template <class _Predicate>
+class __invert // invert the sense of a comparison
+{
+private:
+    _Predicate __p_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __invert() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __invert(_Predicate __p) : __p_(__p) {}
+
+    template <class _T1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _T1& __x) {return !__p_(__x);}
+
+    template <class _T1, class _T2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _T1& __x, const _T2& __y) {return __p_(__y, __x);}
+};
+
+// Perform division by two quickly for positive integers (llvm.org/PR39129)
+
+template <typename _Integral>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    is_integral<_Integral>::value,
+    _Integral
+>::type
+__half_positive(_Integral __value)
+{
+    return static_cast<_Integral>(static_cast<typename make_unsigned<_Integral>::type>(__value) / 2);
+}
+
+template <typename _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    !is_integral<_Tp>::value,
+    _Tp
+>::type
+__half_positive(_Tp __value)
+{
+    return __value / 2;
+}
+
+#ifdef _LIBCUDACXX_DEBUG
+
+template <class _Compare>
+struct __debug_less
+{
+    _Compare &__comp_;
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    __debug_less(_Compare& __c) : __comp_(__c) {}
+
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    bool operator()(const _Tp& __x,  const _Up& __y)
+    {
+        bool __r = __comp_(__x, __y);
+        if (__r)
+            __do_compare_assert(0, __y, __x);
+        return __r;
+    }
+
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    bool operator()(_Tp& __x,  _Up& __y)
+    {
+        bool __r = __comp_(__x, __y);
+        if (__r)
+            __do_compare_assert(0, __y, __x);
+        return __r;
+    }
+
+    template <class _LHS, class _RHS>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    decltype((void)_CUDA_VSTD::declval<_Compare&>()(
+        _CUDA_VSTD::declval<_LHS &>(), _CUDA_VSTD::declval<_RHS &>()))
+    __do_compare_assert(int, _LHS & __l, _RHS & __r) {
+        _LIBCUDACXX_ASSERT(!__comp_(__l, __r),
+            "Comparator does not induce a strict weak ordering");
+    }
+
+    template <class _LHS, class _RHS>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void __do_compare_assert(long, _LHS &, _RHS &) {}
+};
+
+#endif // _LIBCUDACXX_DEBUG
+
+#endif // __cuda_std__
+
+template <class _Comp>
+struct __comp_ref_type {
+  // Pass the comparator by lvalue reference. Or in debug mode, using a
+  // debugging wrapper that stores a reference.
+#ifndef _LIBCUDACXX_DEBUG
+  typedef typename add_lvalue_reference<_Comp>::type type;
+#else
+  typedef __debug_less<_Comp> type;
+#endif
+};
+
+#ifndef __cuda_std__
+// all_of
+
+template <class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+all_of(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+        if (!__pred(*__first))
+            return false;
+    return true;
+}
+
+// any_of
+
+template <class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+any_of(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+        if (__pred(*__first))
+            return true;
+    return false;
+}
+
+// none_of
+
+template <class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+none_of(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+        if (__pred(*__first))
+            return false;
+    return true;
+}
+
+// for_each
+
+template <class _InputIterator, class _Function>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_Function
+for_each(_InputIterator __first, _InputIterator __last, _Function __f)
+{
+    for (; __first != __last; ++__first)
+        __f(*__first);
+    return __f;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+// for_each_n
+
+template <class _InputIterator, class _Size, class _Function>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_InputIterator
+for_each_n(_InputIterator __first, _Size __orig_n, _Function __f)
+{
+    typedef decltype(__convert_to_integral(__orig_n)) _IntegralSize;
+    _IntegralSize __n = __orig_n;
+    while (__n > 0)
+    {
+         __f(*__first);
+         ++__first;
+         --__n;
+    }
+    return __first;
+}
+#endif
+
+// find
+
+template <class _InputIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_InputIterator
+find(_InputIterator __first, _InputIterator __last, const _Tp& __value_)
+{
+    for (; __first != __last; ++__first)
+        if (*__first == __value_)
+            break;
+    return __first;
+}
+
+// find_if
+
+template <class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_InputIterator
+find_if(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+        if (__pred(*__first))
+            break;
+    return __first;
+}
+
+// find_if_not
+
+template<class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_InputIterator
+find_if_not(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+        if (!__pred(*__first))
+            break;
+    return __first;
+}
+
+// find_end
+
+template <class _BinaryPredicate, class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator1
+__find_end(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+           _ForwardIterator2 __first2, _ForwardIterator2 __last2, _BinaryPredicate __pred,
+           forward_iterator_tag, forward_iterator_tag)
+{
+    // modeled after search algorithm
+    _ForwardIterator1 __r = __last1;  // __last1 is the "default" answer
+    if (__first2 == __last2)
+        return __r;
+    while (true)
+    {
+        while (true)
+        {
+            if (__first1 == __last1)         // if source exhausted return last correct answer
+                return __r;                  //    (or __last1 if never found)
+            if (__pred(*__first1, *__first2))
+                break;
+            ++__first1;
+        }
+        // *__first1 matches *__first2, now match elements after here
+        _ForwardIterator1 __m1 = __first1;
+        _ForwardIterator2 __m2 = __first2;
+        while (true)
+        {
+            if (++__m2 == __last2)
+            {                         // Pattern exhaused, record answer and search for another one
+                __r = __first1;
+                ++__first1;
+                break;
+            }
+            if (++__m1 == __last1)     // Source exhausted, return last answer
+                return __r;
+            if (!__pred(*__m1, *__m2))  // mismatch, restart with a new __first
+            {
+                ++__first1;
+                break;
+            }  // else there is a match, check next elements
+        }
+    }
+}
+
+template <class _BinaryPredicate, class _BidirectionalIterator1, class _BidirectionalIterator2>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _BidirectionalIterator1
+__find_end(_BidirectionalIterator1 __first1, _BidirectionalIterator1 __last1,
+           _BidirectionalIterator2 __first2, _BidirectionalIterator2 __last2, _BinaryPredicate __pred,
+           bidirectional_iterator_tag, bidirectional_iterator_tag)
+{
+    // modeled after search algorithm (in reverse)
+    if (__first2 == __last2)
+        return __last1;  // Everything matches an empty sequence
+    _BidirectionalIterator1 __l1 = __last1;
+    _BidirectionalIterator2 __l2 = __last2;
+    --__l2;
+    while (true)
+    {
+        // Find last element in sequence 1 that matchs *(__last2-1), with a mininum of loop checks
+        while (true)
+        {
+            if (__first1 == __l1)  // return __last1 if no element matches *__first2
+                return __last1;
+            if (__pred(*--__l1, *__l2))
+                break;
+        }
+        // *__l1 matches *__l2, now match elements before here
+        _BidirectionalIterator1 __m1 = __l1;
+        _BidirectionalIterator2 __m2 = __l2;
+        while (true)
+        {
+            if (__m2 == __first2)  // If pattern exhausted, __m1 is the answer (works for 1 element pattern)
+                return __m1;
+            if (__m1 == __first1)  // Otherwise if source exhaused, pattern not found
+                return __last1;
+            if (!__pred(*--__m1, *--__m2))  // if there is a mismatch, restart with a new __l1
+            {
+                break;
+            }  // else there is a match, check next elements
+        }
+    }
+}
+
+template <class _BinaryPredicate, class _RandomAccessIterator1, class _RandomAccessIterator2>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _RandomAccessIterator1
+__find_end(_RandomAccessIterator1 __first1, _RandomAccessIterator1 __last1,
+           _RandomAccessIterator2 __first2, _RandomAccessIterator2 __last2, _BinaryPredicate __pred,
+           random_access_iterator_tag, random_access_iterator_tag)
+{
+    // Take advantage of knowing source and pattern lengths.  Stop short when source is smaller than pattern
+    typename iterator_traits<_RandomAccessIterator2>::difference_type __len2 = __last2 - __first2;
+    if (__len2 == 0)
+        return __last1;
+    typename iterator_traits<_RandomAccessIterator1>::difference_type __len1 = __last1 - __first1;
+    if (__len1 < __len2)
+        return __last1;
+    const _RandomAccessIterator1 __s = __first1 + (__len2 - 1);  // End of pattern match can't go before here
+    _RandomAccessIterator1 __l1 = __last1;
+    _RandomAccessIterator2 __l2 = __last2;
+    --__l2;
+    while (true)
+    {
+        while (true)
+        {
+            if (__s == __l1)
+                return __last1;
+            if (__pred(*--__l1, *__l2))
+                break;
+        }
+        _RandomAccessIterator1 __m1 = __l1;
+        _RandomAccessIterator2 __m2 = __l2;
+        while (true)
+        {
+            if (__m2 == __first2)
+                return __m1;
+                                 // no need to check range on __m1 because __s guarantees we have enough source
+            if (!__pred(*--__m1, *--__m2))
+            {
+                break;
+            }
+        }
+    }
+}
+
+template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator1
+find_end(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+         _ForwardIterator2 __first2, _ForwardIterator2 __last2, _BinaryPredicate __pred)
+{
+    return _CUDA_VSTD::__find_end<typename add_lvalue_reference<_BinaryPredicate>::type>
+                         (__first1, __last1, __first2, __last2, __pred,
+                          typename iterator_traits<_ForwardIterator1>::iterator_category(),
+                          typename iterator_traits<_ForwardIterator2>::iterator_category());
+}
+
+template <class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator1
+find_end(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+         _ForwardIterator2 __first2, _ForwardIterator2 __last2)
+{
+    typedef typename iterator_traits<_ForwardIterator1>::value_type __v1;
+    typedef typename iterator_traits<_ForwardIterator2>::value_type __v2;
+    return _CUDA_VSTD::find_end(__first1, __last1, __first2, __last2, __equal_to<__v1, __v2>());
+}
+
+// find_first_of
+
+template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _ForwardIterator1
+__find_first_of_ce(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+              _ForwardIterator2 __first2, _ForwardIterator2 __last2, _BinaryPredicate __pred)
+{
+    for (; __first1 != __last1; ++__first1)
+        for (_ForwardIterator2 __j = __first2; __j != __last2; ++__j)
+            if (__pred(*__first1, *__j))
+                return __first1;
+    return __last1;
+}
+
+
+template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator1
+find_first_of(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+              _ForwardIterator2 __first2, _ForwardIterator2 __last2, _BinaryPredicate __pred)
+{
+    return _CUDA_VSTD::__find_first_of_ce(__first1, __last1, __first2, __last2, __pred);
+}
+
+template <class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator1
+find_first_of(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+              _ForwardIterator2 __first2, _ForwardIterator2 __last2)
+{
+    typedef typename iterator_traits<_ForwardIterator1>::value_type __v1;
+    typedef typename iterator_traits<_ForwardIterator2>::value_type __v2;
+    return _CUDA_VSTD::__find_first_of_ce(__first1, __last1, __first2, __last2, __equal_to<__v1, __v2>());
+}
+
+// adjacent_find
+
+template <class _ForwardIterator, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+adjacent_find(_ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+        {
+            if (__pred(*__first, *__i))
+                return __first;
+            __first = __i;
+        }
+    }
+    return __last;
+}
+
+template <class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+adjacent_find(_ForwardIterator __first, _ForwardIterator __last)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type __v;
+    return _CUDA_VSTD::adjacent_find(__first, __last, __equal_to<__v>());
+}
+
+// count
+
+template <class _InputIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+typename iterator_traits<_InputIterator>::difference_type
+count(_InputIterator __first, _InputIterator __last, const _Tp& __value_)
+{
+    typename iterator_traits<_InputIterator>::difference_type __r(0);
+    for (; __first != __last; ++__first)
+        if (*__first == __value_)
+            ++__r;
+    return __r;
+}
+
+// count_if
+
+template <class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+typename iterator_traits<_InputIterator>::difference_type
+count_if(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    typename iterator_traits<_InputIterator>::difference_type __r(0);
+    for (; __first != __last; ++__first)
+        if (__pred(*__first))
+            ++__r;
+    return __r;
+}
+
+// mismatch
+
+template <class _InputIterator1, class _InputIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+pair<_InputIterator1, _InputIterator2>
+mismatch(_InputIterator1 __first1, _InputIterator1 __last1,
+         _InputIterator2 __first2, _BinaryPredicate __pred)
+{
+    for (; __first1 != __last1; ++__first1, (void) ++__first2)
+        if (!__pred(*__first1, *__first2))
+            break;
+    return pair<_InputIterator1, _InputIterator2>(__first1, __first2);
+}
+
+template <class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+pair<_InputIterator1, _InputIterator2>
+mismatch(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2)
+{
+    typedef typename iterator_traits<_InputIterator1>::value_type __v1;
+    typedef typename iterator_traits<_InputIterator2>::value_type __v2;
+    return _CUDA_VSTD::mismatch(__first1, __last1, __first2, __equal_to<__v1, __v2>());
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _InputIterator1, class _InputIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+pair<_InputIterator1, _InputIterator2>
+mismatch(_InputIterator1 __first1, _InputIterator1 __last1,
+         _InputIterator2 __first2, _InputIterator2 __last2,
+         _BinaryPredicate __pred)
+{
+    for (; __first1 != __last1 && __first2 != __last2; ++__first1, (void) ++__first2)
+        if (!__pred(*__first1, *__first2))
+            break;
+    return pair<_InputIterator1, _InputIterator2>(__first1, __first2);
+}
+
+template <class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+pair<_InputIterator1, _InputIterator2>
+mismatch(_InputIterator1 __first1, _InputIterator1 __last1,
+         _InputIterator2 __first2, _InputIterator2 __last2)
+{
+    typedef typename iterator_traits<_InputIterator1>::value_type __v1;
+    typedef typename iterator_traits<_InputIterator2>::value_type __v2;
+    return _CUDA_VSTD::mismatch(__first1, __last1, __first2, __last2, __equal_to<__v1, __v2>());
+}
+#endif
+
+#endif //__cuda_std__
+// equal
+
+template <class _InputIterator1, class _InputIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+equal(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2, _BinaryPredicate __pred)
+{
+    for (; __first1 != __last1; ++__first1, (void) ++__first2)
+        if (!__pred(*__first1, *__first2))
+            return false;
+    return true;
+}
+
+template <class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+equal(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2)
+{
+    typedef typename iterator_traits<_InputIterator1>::value_type __v1;
+    typedef typename iterator_traits<_InputIterator2>::value_type __v2;
+    return _CUDA_VSTD::equal(__first1, __last1, __first2, __equal_to<__v1, __v2>());
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _BinaryPredicate, class _InputIterator1, class _InputIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+__equal(_InputIterator1 __first1, _InputIterator1 __last1,
+        _InputIterator2 __first2, _InputIterator2 __last2, _BinaryPredicate __pred,
+        input_iterator_tag, input_iterator_tag )
+{
+    for (; __first1 != __last1 && __first2 != __last2; ++__first1, (void) ++__first2)
+        if (!__pred(*__first1, *__first2))
+            return false;
+    return __first1 == __last1 && __first2 == __last2;
+}
+
+template <class _BinaryPredicate, class _RandomAccessIterator1, class _RandomAccessIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+__equal(_RandomAccessIterator1 __first1, _RandomAccessIterator1 __last1,
+        _RandomAccessIterator2 __first2, _RandomAccessIterator2 __last2, _BinaryPredicate __pred,
+      random_access_iterator_tag, random_access_iterator_tag )
+{
+    if ( _CUDA_VSTD::distance(__first1, __last1) != _CUDA_VSTD::distance(__first2, __last2))
+        return false;
+    return _CUDA_VSTD::equal<_RandomAccessIterator1, _RandomAccessIterator2,
+                        typename add_lvalue_reference<_BinaryPredicate>::type>
+                       (__first1, __last1, __first2, __pred );
+}
+
+template <class _InputIterator1, class _InputIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+inline bool
+equal(_InputIterator1 __first1, _InputIterator1 __last1,
+      _InputIterator2 __first2, _InputIterator2 __last2, _BinaryPredicate __pred )
+{
+    return _CUDA_VSTD::__equal<typename add_lvalue_reference<_BinaryPredicate>::type>
+       (__first1, __last1, __first2, __last2, __pred,
+        typename iterator_traits<_InputIterator1>::iterator_category(),
+        typename iterator_traits<_InputIterator2>::iterator_category());
+}
+
+template <class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_NODISCARD_EXT
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+inline bool
+equal(_InputIterator1 __first1, _InputIterator1 __last1,
+      _InputIterator2 __first2, _InputIterator2 __last2)
+{
+    typedef typename iterator_traits<_InputIterator1>::value_type __v1;
+    typedef typename iterator_traits<_InputIterator2>::value_type __v2;
+    return _CUDA_VSTD::__equal(__first1, __last1, __first2, __last2, __equal_to<__v1, __v2>(),
+        typename iterator_traits<_InputIterator1>::iterator_category(),
+        typename iterator_traits<_InputIterator2>::iterator_category());
+}
+#endif
+
+#ifndef __cuda_std__
+// is_permutation
+
+template<class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+is_permutation(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+               _ForwardIterator2 __first2, _BinaryPredicate __pred)
+{
+//  shorten sequences as much as possible by lopping of any equal prefix
+    for (; __first1 != __last1; ++__first1, (void) ++__first2)
+        if (!__pred(*__first1, *__first2))
+            break;
+    if (__first1 == __last1)
+        return true;
+
+//  __first1 != __last1 && *__first1 != *__first2
+    typedef typename iterator_traits<_ForwardIterator1>::difference_type _D1;
+    _D1 __l1 = _CUDA_VSTD::distance(__first1, __last1);
+    if (__l1 == _D1(1))
+        return false;
+    _ForwardIterator2 __last2 = _CUDA_VSTD::next(__first2, __l1);
+    // For each element in [f1, l1) see if there are the same number of
+    //    equal elements in [f2, l2)
+    for (_ForwardIterator1 __i = __first1; __i != __last1; ++__i)
+    {
+    //  Have we already counted the number of *__i in [f1, l1)?
+        _ForwardIterator1 __match = __first1;
+        for (; __match != __i; ++__match)
+            if (__pred(*__match, *__i))
+                break;
+        if (__match == __i) {
+            // Count number of *__i in [f2, l2)
+            _D1 __c2 = 0;
+            for (_ForwardIterator2 __j = __first2; __j != __last2; ++__j)
+                if (__pred(*__i, *__j))
+                    ++__c2;
+            if (__c2 == 0)
+                return false;
+            // Count number of *__i in [__i, l1) (we can start with 1)
+            _D1 __c1 = 1;
+            for (_ForwardIterator1 __j = _CUDA_VSTD::next(__i); __j != __last1; ++__j)
+                if (__pred(*__i, *__j))
+                    ++__c1;
+            if (__c1 != __c2)
+                return false;
+        }
+    }
+    return true;
+}
+
+template<class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_NODISCARD_EXT
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+inline bool
+is_permutation(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+               _ForwardIterator2 __first2)
+{
+    typedef typename iterator_traits<_ForwardIterator1>::value_type __v1;
+    typedef typename iterator_traits<_ForwardIterator2>::value_type __v2;
+    return _CUDA_VSTD::is_permutation(__first1, __last1, __first2, __equal_to<__v1, __v2>());
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template<class _BinaryPredicate, class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+__is_permutation(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+                 _ForwardIterator2 __first2, _ForwardIterator2 __last2,
+                 _BinaryPredicate __pred,
+                 forward_iterator_tag, forward_iterator_tag )
+{
+//  shorten sequences as much as possible by lopping of any equal prefix
+    for (; __first1 != __last1 && __first2 != __last2; ++__first1, (void) ++__first2)
+        if (!__pred(*__first1, *__first2))
+            break;
+    if (__first1 == __last1)
+        return __first2 == __last2;
+    else if (__first2 == __last2)
+        return false;
+
+    typedef typename iterator_traits<_ForwardIterator1>::difference_type _D1;
+    _D1 __l1 = _CUDA_VSTD::distance(__first1, __last1);
+
+    typedef typename iterator_traits<_ForwardIterator2>::difference_type _D2;
+    _D2 __l2 = _CUDA_VSTD::distance(__first2, __last2);
+    if (__l1 != __l2)
+        return false;
+
+    // For each element in [f1, l1) see if there are the same number of
+    //    equal elements in [f2, l2)
+    for (_ForwardIterator1 __i = __first1; __i != __last1; ++__i)
+    {
+    //  Have we already counted the number of *__i in [f1, l1)?
+        _ForwardIterator1 __match = __first1;
+        for (; __match != __i; ++__match)
+            if (__pred(*__match, *__i))
+                break;
+        if (__match == __i) {
+            // Count number of *__i in [f2, l2)
+            _D1 __c2 = 0;
+            for (_ForwardIterator2 __j = __first2; __j != __last2; ++__j)
+                if (__pred(*__i, *__j))
+                    ++__c2;
+            if (__c2 == 0)
+                return false;
+            // Count number of *__i in [__i, l1) (we can start with 1)
+            _D1 __c1 = 1;
+            for (_ForwardIterator1 __j = _CUDA_VSTD::next(__i); __j != __last1; ++__j)
+                if (__pred(*__i, *__j))
+                    ++__c1;
+            if (__c1 != __c2)
+                return false;
+        }
+    }
+    return true;
+}
+
+template<class _BinaryPredicate, class _RandomAccessIterator1, class _RandomAccessIterator2>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+__is_permutation(_RandomAccessIterator1 __first1, _RandomAccessIterator2 __last1,
+               _RandomAccessIterator1 __first2, _RandomAccessIterator2 __last2,
+               _BinaryPredicate __pred,
+               random_access_iterator_tag, random_access_iterator_tag )
+{
+    if ( _CUDA_VSTD::distance(__first1, __last1) != _CUDA_VSTD::distance(__first2, __last2))
+        return false;
+    return _CUDA_VSTD::is_permutation<_RandomAccessIterator1, _RandomAccessIterator2,
+                                 typename add_lvalue_reference<_BinaryPredicate>::type>
+                                (__first1, __last1, __first2, __pred );
+}
+
+template<class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+is_permutation(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+               _ForwardIterator2 __first2, _ForwardIterator2 __last2,
+               _BinaryPredicate __pred )
+{
+    return _CUDA_VSTD::__is_permutation<typename add_lvalue_reference<_BinaryPredicate>::type>
+       (__first1, __last1, __first2, __last2, __pred,
+        typename iterator_traits<_ForwardIterator1>::iterator_category(),
+        typename iterator_traits<_ForwardIterator2>::iterator_category());
+}
+
+template<class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+is_permutation(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+               _ForwardIterator2 __first2, _ForwardIterator2 __last2)
+{
+    typedef typename iterator_traits<_ForwardIterator1>::value_type __v1;
+    typedef typename iterator_traits<_ForwardIterator2>::value_type __v2;
+    return _CUDA_VSTD::__is_permutation(__first1, __last1, __first2, __last2,
+        __equal_to<__v1, __v2>(),
+        typename iterator_traits<_ForwardIterator1>::iterator_category(),
+        typename iterator_traits<_ForwardIterator2>::iterator_category());
+}
+#endif
+
+// search
+// __search is in <functional>
+
+template <class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator1
+search(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+       _ForwardIterator2 __first2, _ForwardIterator2 __last2, _BinaryPredicate __pred)
+{
+    return _CUDA_VSTD::__search<typename add_lvalue_reference<_BinaryPredicate>::type>
+                         (__first1, __last1, __first2, __last2, __pred,
+                          typename iterator_traits<_ForwardIterator1>::iterator_category(),
+                          typename iterator_traits<_ForwardIterator2>::iterator_category())
+            .first;
+}
+
+template <class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator1
+search(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+       _ForwardIterator2 __first2, _ForwardIterator2 __last2)
+{
+    typedef typename iterator_traits<_ForwardIterator1>::value_type __v1;
+    typedef typename iterator_traits<_ForwardIterator2>::value_type __v2;
+    return _CUDA_VSTD::search(__first1, __last1, __first2, __last2, __equal_to<__v1, __v2>());
+}
+
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _ForwardIterator, class _Searcher>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator search(_ForwardIterator __f, _ForwardIterator __l, const _Searcher &__s)
+{ return __s(__f, __l).first; }
+#endif
+
+// search_n
+
+template <class _BinaryPredicate, class _ForwardIterator, class _Size, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+__search_n(_ForwardIterator __first, _ForwardIterator __last,
+           _Size __count, const _Tp& __value_, _BinaryPredicate __pred, forward_iterator_tag)
+{
+    if (__count <= 0)
+        return __first;
+    while (true)
+    {
+        // Find first element in sequence that matchs __value_, with a mininum of loop checks
+        while (true)
+        {
+            if (__first == __last)  // return __last if no element matches __value_
+                return __last;
+            if (__pred(*__first, __value_))
+                break;
+            ++__first;
+        }
+        // *__first matches __value_, now match elements after here
+        _ForwardIterator __m = __first;
+        _Size __c(0);
+        while (true)
+        {
+            if (++__c == __count)  // If pattern exhausted, __first is the answer (works for 1 element pattern)
+                return __first;
+            if (++__m == __last)  // Otherwise if source exhaused, pattern not found
+                return __last;
+            if (!__pred(*__m, __value_))  // if there is a mismatch, restart with a new __first
+            {
+                __first = __m;
+                ++__first;
+                break;
+            }  // else there is a match, check next elements
+        }
+    }
+}
+
+template <class _BinaryPredicate, class _RandomAccessIterator, class _Size, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _RandomAccessIterator
+__search_n(_RandomAccessIterator __first, _RandomAccessIterator __last,
+           _Size __count, const _Tp& __value_, _BinaryPredicate __pred, random_access_iterator_tag)
+{
+    if (__count <= 0)
+        return __first;
+    _Size __len = static_cast<_Size>(__last - __first);
+    if (__len < __count)
+        return __last;
+    const _RandomAccessIterator __s = __last - (__count - 1);  // Start of pattern match can't go beyond here
+    while (true)
+    {
+        // Find first element in sequence that matchs __value_, with a mininum of loop checks
+        while (true)
+        {
+            if (__first >= __s)  // return __last if no element matches __value_
+                return __last;
+            if (__pred(*__first, __value_))
+                break;
+            ++__first;
+        }
+        // *__first matches __value_, now match elements after here
+        _RandomAccessIterator __m = __first;
+        _Size __c(0);
+        while (true)
+        {
+            if (++__c == __count)  // If pattern exhausted, __first is the answer (works for 1 element pattern)
+                return __first;
+             ++__m;          // no need to check range on __m because __s guarantees we have enough source
+            if (!__pred(*__m, __value_))  // if there is a mismatch, restart with a new __first
+            {
+                __first = __m;
+                ++__first;
+                break;
+            }  // else there is a match, check next elements
+        }
+    }
+}
+
+template <class _ForwardIterator, class _Size, class _Tp, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+search_n(_ForwardIterator __first, _ForwardIterator __last,
+         _Size __count, const _Tp& __value_, _BinaryPredicate __pred)
+{
+    return _CUDA_VSTD::__search_n<typename add_lvalue_reference<_BinaryPredicate>::type>
+           (__first, __last, __convert_to_integral(__count), __value_, __pred,
+           typename iterator_traits<_ForwardIterator>::iterator_category());
+}
+
+template <class _ForwardIterator, class _Size, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+search_n(_ForwardIterator __first, _ForwardIterator __last, _Size __count, const _Tp& __value_)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type __v;
+    return _CUDA_VSTD::search_n(__first, __last, __convert_to_integral(__count),
+                           __value_, __equal_to<__v, _Tp>());
+}
+
+// copy
+template <class _Iter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Iter
+__unwrap_iter(_Iter __i)
+{
+    return __i;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_trivially_copy_assignable<_Tp>::value,
+    _Tp*
+>::type
+__unwrap_iter(move_iterator<_Tp*> __i)
+{
+    return __i.base();
+}
+
+#if _LIBCUDACXX_DEBUG_LEVEL < 2
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+typename enable_if
+<
+    is_trivially_copy_assignable<_Tp>::value,
+    _Tp*
+>::type
+__unwrap_iter(__wrap_iter<_Tp*> __i)
+{
+    return __i.base();
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+typename enable_if
+<
+    is_trivially_copy_assignable<_Tp>::value,
+    const _Tp*
+>::type
+__unwrap_iter(__wrap_iter<const _Tp*> __i)
+{
+    return __i.base();
+}
+
+#else
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+typename enable_if
+<
+    is_trivially_copy_assignable<_Tp>::value,
+    __wrap_iter<_Tp*>
+>::type
+__unwrap_iter(__wrap_iter<_Tp*> __i)
+{
+    return __i;
+}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL < 2
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+__copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    for (; __first != __last; ++__first, (void) ++__result)
+        *__result = *__first;
+    return __result;
+}
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_same<typename remove_const<_Tp>::type, _Up>::value &&
+    is_trivially_copy_assignable<_Up>::value,
+    _Up*
+>::type
+__copy(_Tp* __first, _Tp* __last, _Up* __result)
+{
+    const size_t __n = static_cast<size_t>(__last - __first);
+    if (__n > 0)
+        _CUDA_VSTD::memmove(__result, __first, __n * sizeof(_Up));
+    return __result + __n;
+}
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    return _CUDA_VSTD::__copy(__unwrap_iter(__first), __unwrap_iter(__last), __unwrap_iter(__result));
+}
+
+// copy_backward
+
+template <class _BidirectionalIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+__copy_backward(_BidirectionalIterator __first, _BidirectionalIterator __last, _OutputIterator __result)
+{
+    while (__first != __last)
+        *--__result = *--__last;
+    return __result;
+}
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_same<typename remove_const<_Tp>::type, _Up>::value &&
+    is_trivially_copy_assignable<_Up>::value,
+    _Up*
+>::type
+__copy_backward(_Tp* __first, _Tp* __last, _Up* __result)
+{
+    const size_t __n = static_cast<size_t>(__last - __first);
+    if (__n > 0)
+    {
+        __result -= __n;
+        _CUDA_VSTD::memmove(__result, __first, __n * sizeof(_Up));
+    }
+    return __result;
+}
+
+template <class _BidirectionalIterator1, class _BidirectionalIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_BidirectionalIterator2
+copy_backward(_BidirectionalIterator1 __first, _BidirectionalIterator1 __last,
+              _BidirectionalIterator2 __result)
+{
+    return _CUDA_VSTD::__copy_backward(__unwrap_iter(__first),
+                                  __unwrap_iter(__last),
+                                  __unwrap_iter(__result));
+}
+
+// copy_if
+
+template<class _InputIterator, class _OutputIterator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+copy_if(_InputIterator __first, _InputIterator __last,
+        _OutputIterator __result, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+    {
+        if (__pred(*__first))
+        {
+            *__result = *__first;
+            ++__result;
+        }
+    }
+    return __result;
+}
+
+// copy_n
+
+template<class _InputIterator, class _Size, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_input_iterator<_InputIterator>::value &&
+   !__is_random_access_iterator<_InputIterator>::value,
+    _OutputIterator
+>::type
+copy_n(_InputIterator __first, _Size __orig_n, _OutputIterator __result)
+{
+    typedef decltype(__convert_to_integral(__orig_n)) _IntegralSize;
+    _IntegralSize __n = __orig_n;
+    if (__n > 0)
+    {
+        *__result = *__first;
+        ++__result;
+        for (--__n; __n > 0; --__n)
+        {
+            ++__first;
+            *__result = *__first;
+            ++__result;
+        }
+    }
+    return __result;
+}
+
+template<class _InputIterator, class _Size, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_random_access_iterator<_InputIterator>::value,
+    _OutputIterator
+>::type
+copy_n(_InputIterator __first, _Size __orig_n, _OutputIterator __result)
+{
+    typedef decltype(__convert_to_integral(__orig_n)) _IntegralSize;
+    _IntegralSize __n = __orig_n;
+    return _CUDA_VSTD::copy(__first, __first + __n, __result);
+}
+
+// move
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+__move(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    for (; __first != __last; ++__first, (void) ++__result)
+        *__result = _CUDA_VSTD::move(*__first);
+    return __result;
+}
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_same<typename remove_const<_Tp>::type, _Up>::value &&
+    is_trivially_copy_assignable<_Up>::value,
+    _Up*
+>::type
+__move(_Tp* __first, _Tp* __last, _Up* __result)
+{
+    const size_t __n = static_cast<size_t>(__last - __first);
+    if (__n > 0)
+        _CUDA_VSTD::memmove(__result, __first, __n * sizeof(_Up));
+    return __result + __n;
+}
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+move(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    return _CUDA_VSTD::__move(__unwrap_iter(__first), __unwrap_iter(__last), __unwrap_iter(__result));
+}
+
+// move_backward
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+__move_backward(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    while (__first != __last)
+        *--__result = _CUDA_VSTD::move(*--__last);
+    return __result;
+}
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_same<typename remove_const<_Tp>::type, _Up>::value &&
+    is_trivially_copy_assignable<_Up>::value,
+    _Up*
+>::type
+__move_backward(_Tp* __first, _Tp* __last, _Up* __result)
+{
+    const size_t __n = static_cast<size_t>(__last - __first);
+    if (__n > 0)
+    {
+        __result -= __n;
+        _CUDA_VSTD::memmove(__result, __first, __n * sizeof(_Up));
+    }
+    return __result;
+}
+
+template <class _BidirectionalIterator1, class _BidirectionalIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_BidirectionalIterator2
+move_backward(_BidirectionalIterator1 __first, _BidirectionalIterator1 __last,
+              _BidirectionalIterator2 __result)
+{
+    return _CUDA_VSTD::__move_backward(__unwrap_iter(__first), __unwrap_iter(__last), __unwrap_iter(__result));
+}
+
+// iter_swap
+
+// moved to <type_traits> for better swap / noexcept support
+
+// transform
+
+template <class _InputIterator, class _OutputIterator, class _UnaryOperation>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+transform(_InputIterator __first, _InputIterator __last, _OutputIterator __result, _UnaryOperation __op)
+{
+    for (; __first != __last; ++__first, (void) ++__result)
+        *__result = __op(*__first);
+    return __result;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _BinaryOperation>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+transform(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2,
+          _OutputIterator __result, _BinaryOperation __binary_op)
+{
+    for (; __first1 != __last1; ++__first1, (void) ++__first2, ++__result)
+        *__result = __binary_op(*__first1, *__first2);
+    return __result;
+}
+
+// replace
+
+template <class _ForwardIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+replace(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __old_value, const _Tp& __new_value)
+{
+    for (; __first != __last; ++__first)
+        if (*__first == __old_value)
+            *__first = __new_value;
+}
+
+// replace_if
+
+template <class _ForwardIterator, class _Predicate, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+replace_if(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred, const _Tp& __new_value)
+{
+    for (; __first != __last; ++__first)
+        if (__pred(*__first))
+            *__first = __new_value;
+}
+
+// replace_copy
+
+template <class _InputIterator, class _OutputIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+replace_copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result,
+             const _Tp& __old_value, const _Tp& __new_value)
+{
+    for (; __first != __last; ++__first, (void) ++__result)
+        if (*__first == __old_value)
+            *__result = __new_value;
+        else
+            *__result = *__first;
+    return __result;
+}
+
+// replace_copy_if
+
+template <class _InputIterator, class _OutputIterator, class _Predicate, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+inline _OutputIterator
+replace_copy_if(_InputIterator __first, _InputIterator __last, _OutputIterator __result,
+                _Predicate __pred, const _Tp& __new_value)
+{
+    for (; __first != __last; ++__first, (void) ++__result)
+        if (__pred(*__first))
+            *__result = __new_value;
+        else
+            *__result = *__first;
+    return __result;
+}
+
+// fill_n
+#endif // __cuda_std__
+
+template <class _OutputIterator, class _Size, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+inline _OutputIterator
+__fill_n(_OutputIterator __first, _Size __n, const _Tp& __value_)
+{
+    for (; __n > 0; ++__first, (void) --__n)
+        *__first = __value_;
+    return __first;
+}
+
+template <class _OutputIterator, class _Size, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+inline _OutputIterator
+fill_n(_OutputIterator __first, _Size __n, const _Tp& __value_)
+{
+   return _CUDA_VSTD::__fill_n(__first, __convert_to_integral(__n), __value_);
+}
+
+// fill
+
+template <class _ForwardIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+__fill(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, forward_iterator_tag)
+{
+    for (; __first != __last; ++__first)
+        *__first = __value_;
+}
+
+template <class _RandomAccessIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+__fill(_RandomAccessIterator __first, _RandomAccessIterator __last, const _Tp& __value_, random_access_iterator_tag)
+{
+    _CUDA_VSTD::fill_n(__first, __last - __first, __value_);
+}
+
+template <class _ForwardIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+fill(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_)
+{
+    _CUDA_VSTD::__fill(__first, __last, __value_, typename iterator_traits<_ForwardIterator>::iterator_category());
+}
+
+#ifndef __cuda_std__
+// generate
+
+template <class _ForwardIterator, class _Generator>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+generate(_ForwardIterator __first, _ForwardIterator __last, _Generator __gen)
+{
+    for (; __first != __last; ++__first)
+        *__first = __gen();
+}
+
+// generate_n
+
+template <class _OutputIterator, class _Size, class _Generator>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+generate_n(_OutputIterator __first, _Size __orig_n, _Generator __gen)
+{
+    typedef decltype(__convert_to_integral(__orig_n)) _IntegralSize;
+    _IntegralSize __n = __orig_n;
+    for (; __n > 0; ++__first, (void) --__n)
+        *__first = __gen();
+    return __first;
+}
+
+// remove
+
+template <class _ForwardIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+remove(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_)
+{
+    __first = _CUDA_VSTD::find(__first, __last, __value_);
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+        {
+            if (!(*__i == __value_))
+            {
+                *__first = _CUDA_VSTD::move(*__i);
+                ++__first;
+            }
+        }
+    }
+    return __first;
+}
+
+// remove_if
+
+template <class _ForwardIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+remove_if(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred)
+{
+    __first = _CUDA_VSTD::find_if<_ForwardIterator, typename add_lvalue_reference<_Predicate>::type>
+                           (__first, __last, __pred);
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+        {
+            if (!__pred(*__i))
+            {
+                *__first = _CUDA_VSTD::move(*__i);
+                ++__first;
+            }
+        }
+    }
+    return __first;
+}
+
+// remove_copy
+
+template <class _InputIterator, class _OutputIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+remove_copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result, const _Tp& __value_)
+{
+    for (; __first != __last; ++__first)
+    {
+        if (!(*__first == __value_))
+        {
+            *__result = *__first;
+            ++__result;
+        }
+    }
+    return __result;
+}
+
+// remove_copy_if
+
+template <class _InputIterator, class _OutputIterator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+remove_copy_if(_InputIterator __first, _InputIterator __last, _OutputIterator __result, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+    {
+        if (!__pred(*__first))
+        {
+            *__result = *__first;
+            ++__result;
+        }
+    }
+    return __result;
+}
+
+// unique
+
+template <class _ForwardIterator, class _BinaryPredicate>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+unique(_ForwardIterator __first, _ForwardIterator __last, _BinaryPredicate __pred)
+{
+    __first = _CUDA_VSTD::adjacent_find<_ForwardIterator, typename add_lvalue_reference<_BinaryPredicate>::type>
+                                 (__first, __last, __pred);
+    if (__first != __last)
+    {
+        // ...  a  a  ?  ...
+        //      f     i
+        _ForwardIterator __i = __first;
+        for (++__i; ++__i != __last;)
+            if (!__pred(*__first, *__i))
+                *++__first = _CUDA_VSTD::move(*__i);
+        ++__first;
+    }
+    return __first;
+}
+
+template <class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+unique(_ForwardIterator __first, _ForwardIterator __last)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type __v;
+    return _CUDA_VSTD::unique(__first, __last, __equal_to<__v>());
+}
+
+// unique_copy
+
+template <class _BinaryPredicate, class _InputIterator, class _OutputIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _OutputIterator
+__unique_copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result, _BinaryPredicate __pred,
+              input_iterator_tag, output_iterator_tag)
+{
+    if (__first != __last)
+    {
+        typename iterator_traits<_InputIterator>::value_type __t(*__first);
+        *__result = __t;
+        ++__result;
+        while (++__first != __last)
+        {
+            if (!__pred(__t, *__first))
+            {
+                __t = *__first;
+                *__result = __t;
+                ++__result;
+            }
+        }
+    }
+    return __result;
+}
+
+template <class _BinaryPredicate, class _ForwardIterator, class _OutputIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _OutputIterator
+__unique_copy(_ForwardIterator __first, _ForwardIterator __last, _OutputIterator __result, _BinaryPredicate __pred,
+              forward_iterator_tag, output_iterator_tag)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        *__result = *__i;
+        ++__result;
+        while (++__first != __last)
+        {
+            if (!__pred(*__i, *__first))
+            {
+                *__result = *__first;
+                ++__result;
+                __i = __first;
+            }
+        }
+    }
+    return __result;
+}
+
+template <class _BinaryPredicate, class _InputIterator, class _ForwardIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+__unique_copy(_InputIterator __first, _InputIterator __last, _ForwardIterator __result, _BinaryPredicate __pred,
+              input_iterator_tag, forward_iterator_tag)
+{
+    if (__first != __last)
+    {
+        *__result = *__first;
+        while (++__first != __last)
+            if (!__pred(*__result, *__first))
+                *++__result = *__first;
+        ++__result;
+    }
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _BinaryPredicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+unique_copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result, _BinaryPredicate __pred)
+{
+    return _CUDA_VSTD::__unique_copy<typename add_lvalue_reference<_BinaryPredicate>::type>
+                              (__first, __last, __result, __pred,
+                               typename iterator_traits<_InputIterator>::iterator_category(),
+                               typename iterator_traits<_OutputIterator>::iterator_category());
+}
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+unique_copy(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    typedef typename iterator_traits<_InputIterator>::value_type __v;
+    return _CUDA_VSTD::unique_copy(__first, __last, __result, __equal_to<__v>());
+}
+
+// reverse
+
+template <class _BidirectionalIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__reverse(_BidirectionalIterator __first, _BidirectionalIterator __last, bidirectional_iterator_tag)
+{
+    while (__first != __last)
+    {
+        if (__first == --__last)
+            break;
+        _CUDA_VSTD::iter_swap(__first, __last);
+        ++__first;
+    }
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__reverse(_RandomAccessIterator __first, _RandomAccessIterator __last, random_access_iterator_tag)
+{
+    if (__first != __last)
+        for (; __first < --__last; ++__first)
+            _CUDA_VSTD::iter_swap(__first, __last);
+}
+
+template <class _BidirectionalIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+reverse(_BidirectionalIterator __first, _BidirectionalIterator __last)
+{
+    _CUDA_VSTD::__reverse(__first, __last, typename iterator_traits<_BidirectionalIterator>::iterator_category());
+}
+
+// reverse_copy
+
+template <class _BidirectionalIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+reverse_copy(_BidirectionalIterator __first, _BidirectionalIterator __last, _OutputIterator __result)
+{
+    for (; __first != __last; ++__result)
+        *__result = *--__last;
+    return __result;
+}
+
+// rotate
+
+template <class _ForwardIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+__rotate_left(_ForwardIterator __first, _ForwardIterator __last)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+    value_type __tmp = _CUDA_VSTD::move(*__first);
+    _ForwardIterator __lm1 = _CUDA_VSTD::move(_CUDA_VSTD::next(__first), __last, __first);
+    *__lm1 = _CUDA_VSTD::move(__tmp);
+    return __lm1;
+}
+
+template <class _BidirectionalIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_BidirectionalIterator
+__rotate_right(_BidirectionalIterator __first, _BidirectionalIterator __last)
+{
+    typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
+    _BidirectionalIterator __lm1 = _CUDA_VSTD::prev(__last);
+    value_type __tmp = _CUDA_VSTD::move(*__lm1);
+    _BidirectionalIterator __fp1 = _CUDA_VSTD::move_backward(__first, __lm1, __last);
+    *__first = _CUDA_VSTD::move(__tmp);
+    return __fp1;
+}
+
+template <class _ForwardIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+__rotate_forward(_ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last)
+{
+    _ForwardIterator __i = __middle;
+    while (true)
+    {
+        swap(*__first, *__i);
+        ++__first;
+        if (++__i == __last)
+            break;
+        if (__first == __middle)
+            __middle = __i;
+    }
+    _ForwardIterator __r = __first;
+    if (__first != __middle)
+    {
+        __i = __middle;
+        while (true)
+        {
+            swap(*__first, *__i);
+            ++__first;
+            if (++__i == __last)
+            {
+                if (__first == __middle)
+                    break;
+                __i = __middle;
+            }
+            else if (__first == __middle)
+                __middle = __i;
+        }
+    }
+    return __r;
+}
+
+template<typename _Integral>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Integral
+__algo_gcd(_Integral __x, _Integral __y)
+{
+    do
+    {
+        _Integral __t = __x % __y;
+        __x = __y;
+        __y = __t;
+    } while (__y);
+    return __x;
+}
+
+template<typename _RandomAccessIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_RandomAccessIterator
+__rotate_gcd(_RandomAccessIterator __first, _RandomAccessIterator __middle, _RandomAccessIterator __last)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+
+    const difference_type __m1 = __middle - __first;
+    const difference_type __m2 = __last - __middle;
+    if (__m1 == __m2)
+    {
+        _CUDA_VSTD::swap_ranges(__first, __middle, __middle);
+        return __middle;
+    }
+    const difference_type __g = _CUDA_VSTD::__algo_gcd(__m1, __m2);
+    for (_RandomAccessIterator __p = __first + __g; __p != __first;)
+    {
+        value_type __t(_CUDA_VSTD::move(*--__p));
+        _RandomAccessIterator __p1 = __p;
+        _RandomAccessIterator __p2 = __p1 + __m1;
+        do
+        {
+            *__p1 = _CUDA_VSTD::move(*__p2);
+            __p1 = __p2;
+            const difference_type __d = __last - __p2;
+            if (__m1 < __d)
+                __p2 += __m1;
+            else
+                __p2 = __first + (__m1 - __d);
+        } while (__p2 != __p);
+        *__p1 = _CUDA_VSTD::move(__t);
+    }
+    return __first + __m2;
+}
+
+template <class _ForwardIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+__rotate(_ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last,
+         _CUDA_VSTD::forward_iterator_tag)
+{
+    typedef typename _CUDA_VSTD::iterator_traits<_ForwardIterator>::value_type value_type;
+    if (_CUDA_VSTD::is_trivially_move_assignable<value_type>::value)
+    {
+        if (_CUDA_VSTD::next(__first) == __middle)
+            return _CUDA_VSTD::__rotate_left(__first, __last);
+    }
+    return _CUDA_VSTD::__rotate_forward(__first, __middle, __last);
+}
+
+template <class _BidirectionalIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_BidirectionalIterator
+__rotate(_BidirectionalIterator __first, _BidirectionalIterator __middle, _BidirectionalIterator __last,
+         _CUDA_VSTD::bidirectional_iterator_tag)
+{
+    typedef typename _CUDA_VSTD::iterator_traits<_BidirectionalIterator>::value_type value_type;
+    if (_CUDA_VSTD::is_trivially_move_assignable<value_type>::value)
+    {
+        if (_CUDA_VSTD::next(__first) == __middle)
+            return _CUDA_VSTD::__rotate_left(__first, __last);
+        if (_CUDA_VSTD::next(__middle) == __last)
+            return _CUDA_VSTD::__rotate_right(__first, __last);
+    }
+    return _CUDA_VSTD::__rotate_forward(__first, __middle, __last);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_RandomAccessIterator
+__rotate(_RandomAccessIterator __first, _RandomAccessIterator __middle, _RandomAccessIterator __last,
+         _CUDA_VSTD::random_access_iterator_tag)
+{
+    typedef typename _CUDA_VSTD::iterator_traits<_RandomAccessIterator>::value_type value_type;
+    if (_CUDA_VSTD::is_trivially_move_assignable<value_type>::value)
+    {
+        if (_CUDA_VSTD::next(__first) == __middle)
+            return _CUDA_VSTD::__rotate_left(__first, __last);
+        if (_CUDA_VSTD::next(__middle) == __last)
+            return _CUDA_VSTD::__rotate_right(__first, __last);
+        return _CUDA_VSTD::__rotate_gcd(__first, __middle, __last);
+    }
+    return _CUDA_VSTD::__rotate_forward(__first, __middle, __last);
+}
+
+template <class _ForwardIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+rotate(_ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last)
+{
+    if (__first == __middle)
+        return __last;
+    if (__middle == __last)
+        return __first;
+    return _CUDA_VSTD::__rotate(__first, __middle, __last,
+                           typename _CUDA_VSTD::iterator_traits<_ForwardIterator>::iterator_category());
+}
+
+// rotate_copy
+
+template <class _ForwardIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+rotate_copy(_ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last, _OutputIterator __result)
+{
+    return _CUDA_VSTD::copy(__first, __middle, _CUDA_VSTD::copy(__middle, __last, __result));
+}
+
+#endif // __cuda_std__
+// min_element
+
+template <class _ForwardIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_ForwardIterator
+min_element(_ForwardIterator __first, _ForwardIterator __last, _Compare __comp)
+{
+    static_assert(__is_forward_iterator<_ForwardIterator>::value,
+        "std::min_element requires a ForwardIterator");
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+            if (__comp(*__i, *__first))
+                __first = __i;
+    }
+    return __first;
+}
+
+template <class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_ForwardIterator
+min_element(_ForwardIterator __first, _ForwardIterator __last)
+{
+    return _CUDA_VSTD::min_element(__first, __last,
+              __less<typename iterator_traits<_ForwardIterator>::value_type>());
+}
+
+// min
+
+template <class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&
+min(const _Tp& __a, const _Tp& __b, _Compare __comp)
+{
+    return __comp(__b, __a) ? __b : __a;
+}
+
+template <class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&
+min(const _Tp& __a, const _Tp& __b)
+{
+    return _CUDA_VSTD::min(__a, __b, __less<_Tp>());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp
+min(::std::initializer_list<_Tp> __t, _Compare __comp)
+{
+    return *_CUDA_VSTD::min_element(__t.begin(), __t.end(), __comp);
+}
+
+template<class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp
+min(::std::initializer_list<_Tp> __t)
+{
+    return *_CUDA_VSTD::min_element(__t.begin(), __t.end(), __less<_Tp>());
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+// max_element
+
+template <class _ForwardIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_ForwardIterator
+max_element(_ForwardIterator __first, _ForwardIterator __last, _Compare __comp)
+{
+    static_assert(__is_forward_iterator<_ForwardIterator>::value,
+        "std::max_element requires a ForwardIterator");
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+            if (__comp(*__first, *__i))
+                __first = __i;
+    }
+    return __first;
+}
+
+
+template <class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_ForwardIterator
+max_element(_ForwardIterator __first, _ForwardIterator __last)
+{
+    return _CUDA_VSTD::max_element(__first, __last,
+              __less<typename iterator_traits<_ForwardIterator>::value_type>());
+}
+
+// max
+
+template <class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&
+max(const _Tp& __a, const _Tp& __b, _Compare __comp)
+{
+    return __comp(__a, __b) ? __b : __a;
+}
+
+template <class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&
+max(const _Tp& __a, const _Tp& __b)
+{
+    return _CUDA_VSTD::max(__a, __b, __less<_Tp>());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp
+max(::std::initializer_list<_Tp> __t, _Compare __comp)
+{
+    return *_CUDA_VSTD::max_element(__t.begin(), __t.end(), __comp);
+}
+
+template<class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp
+max(::std::initializer_list<_Tp> __t)
+{
+    return *_CUDA_VSTD::max_element(__t.begin(), __t.end(), __less<_Tp>());
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 14
+// clamp
+template<class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Tp&
+clamp(const _Tp& __v, const _Tp& __lo, const _Tp& __hi, _Compare __comp)
+{
+    _LIBCUDACXX_ASSERT(!__comp(__hi, __lo), "Bad bounds passed to std::clamp");
+    return __comp(__v, __lo) ? __lo : __comp(__hi, __v) ? __hi : __v;
+
+}
+
+template<class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Tp&
+clamp(const _Tp& __v, const _Tp& __lo, const _Tp& __hi)
+{
+    return _CUDA_VSTD::clamp(__v, __lo, __hi, __less<_Tp>());
+}
+#endif
+
+// minmax_element
+
+template <class _ForwardIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+std::pair<_ForwardIterator, _ForwardIterator>
+minmax_element(_ForwardIterator __first, _ForwardIterator __last, _Compare __comp)
+{
+  static_assert(__is_forward_iterator<_ForwardIterator>::value,
+        "std::minmax_element requires a ForwardIterator");
+  std::pair<_ForwardIterator, _ForwardIterator> __result(__first, __first);
+  if (__first != __last)
+  {
+      if (++__first != __last)
+      {
+          if (__comp(*__first, *__result.first))
+              __result.first = __first;
+          else
+              __result.second = __first;
+          while (++__first != __last)
+          {
+              _ForwardIterator __i = __first;
+              if (++__first == __last)
+              {
+                  if (__comp(*__i, *__result.first))
+                      __result.first = __i;
+                  else if (!__comp(*__i, *__result.second))
+                      __result.second = __i;
+                  break;
+              }
+              else
+              {
+                  if (__comp(*__first, *__i))
+                  {
+                      if (__comp(*__first, *__result.first))
+                          __result.first = __first;
+                      if (!__comp(*__i, *__result.second))
+                          __result.second = __i;
+                  }
+                  else
+                  {
+                      if (__comp(*__i, *__result.first))
+                          __result.first = __i;
+                      if (!__comp(*__first, *__result.second))
+                          __result.second = __first;
+                  }
+              }
+          }
+      }
+  }
+  return __result;
+}
+
+template <class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+std::pair<_ForwardIterator, _ForwardIterator>
+minmax_element(_ForwardIterator __first, _ForwardIterator __last)
+{
+    return _CUDA_VSTD::minmax_element(__first, __last,
+              __less<typename iterator_traits<_ForwardIterator>::value_type>());
+}
+
+// minmax
+
+template<class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<const _Tp&, const _Tp&>
+minmax(const _Tp& __a, const _Tp& __b, _Compare __comp)
+{
+    return __comp(__b, __a) ? pair<const _Tp&, const _Tp&>(__b, __a) :
+                              pair<const _Tp&, const _Tp&>(__a, __b);
+}
+
+template<class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<const _Tp&, const _Tp&>
+minmax(const _Tp& __a, const _Tp& __b)
+{
+    return _CUDA_VSTD::minmax(__a, __b, __less<_Tp>());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<_Tp, _Tp>
+minmax(::std::initializer_list<_Tp> __t, _Compare __comp)
+{
+    typedef typename ::std::initializer_list<_Tp>::const_iterator _Iter;
+    _Iter __first = __t.begin();
+    _Iter __last  = __t.end();
+    std::pair<_Tp, _Tp> __result(*__first, *__first);
+
+    ++__first;
+    if (__t.size() % 2 == 0)
+    {
+        if (__comp(*__first,  __result.first))
+            __result.first  = *__first;
+        else
+            __result.second = *__first;
+        ++__first;
+    }
+
+    while (__first != __last)
+    {
+        _Tp __prev = *__first++;
+        if (__comp(*__first, __prev)) {
+            if ( __comp(*__first, __result.first)) __result.first  = *__first;
+            if (!__comp(__prev, __result.second))  __result.second = __prev;
+            }
+        else {
+            if ( __comp(__prev, __result.first))    __result.first  = __prev;
+            if (!__comp(*__first, __result.second)) __result.second = *__first;
+            }
+
+        __first++;
+    }
+    return __result;
+}
+
+template<class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<_Tp, _Tp>
+minmax(::std::initializer_list<_Tp> __t)
+{
+    return _CUDA_VSTD::minmax(__t, __less<_Tp>());
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#ifndef __cuda_std__
+
+// random_shuffle
+
+// __independent_bits_engine
+
+template <unsigned long long _Xp, size_t _Rp>
+struct __log2_imp
+{
+    static const size_t value = _Xp & ((unsigned long long)(1) << _Rp) ? _Rp
+                                           : __log2_imp<_Xp, _Rp - 1>::value;
+};
+
+template <unsigned long long _Xp>
+struct __log2_imp<_Xp, 0>
+{
+    static const size_t value = 0;
+};
+
+template <size_t _Rp>
+struct __log2_imp<0, _Rp>
+{
+    static const size_t value = _Rp + 1;
+};
+
+template <class _UIntType, _UIntType _Xp>
+struct __log2
+{
+    static const size_t value = __log2_imp<_Xp,
+                                         sizeof(_UIntType) * __CHAR_BIT__ - 1>::value;
+};
+
+template<class _Engine, class _UIntType>
+class __independent_bits_engine
+{
+public:
+    // types
+    typedef _UIntType result_type;
+
+private:
+    typedef typename _Engine::result_type _Engine_result_type;
+    typedef typename conditional
+        <
+            sizeof(_Engine_result_type) <= sizeof(result_type),
+                result_type,
+                _Engine_result_type
+        >::type _Working_result_type;
+
+    _Engine& __e_;
+    size_t __w_;
+    size_t __w0_;
+    size_t __n_;
+    size_t __n0_;
+    _Working_result_type __y0_;
+    _Working_result_type __y1_;
+    _Engine_result_type __mask0_;
+    _Engine_result_type __mask1_;
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    static const _Working_result_type _Rp = _Engine::_Max - _Engine::_Min
+                                          + _Working_result_type(1);
+#else
+    static _LIBCUDACXX_CONSTEXPR const _Working_result_type _Rp = _Engine::max() - _Engine::min()
+                                                      + _Working_result_type(1);
+#endif
+    static _LIBCUDACXX_CONSTEXPR const size_t __m = __log2<_Working_result_type, _Rp>::value;
+    static _LIBCUDACXX_CONSTEXPR const size_t _WDt = numeric_limits<_Working_result_type>::digits;
+    static _LIBCUDACXX_CONSTEXPR const size_t _EDt = numeric_limits<_Engine_result_type>::digits;
+
+public:
+    // constructors and seeding functions
+    __independent_bits_engine(_Engine& __e, size_t __w);
+
+    // generating functions
+    result_type operator()() {return __eval(integral_constant<bool, _Rp != 0>());}
+
+private:
+    result_type __eval(false_type);
+    result_type __eval(true_type);
+};
+
+template<class _Engine, class _UIntType>
+__independent_bits_engine<_Engine, _UIntType>
+    ::__independent_bits_engine(_Engine& __e, size_t __w)
+        : __e_(__e),
+          __w_(__w)
+{
+    __n_ = __w_ / __m + (__w_ % __m != 0);
+    __w0_ = __w_ / __n_;
+    if (_Rp == 0)
+        __y0_ = _Rp;
+    else if (__w0_ < _WDt)
+        __y0_ = (_Rp >> __w0_) << __w0_;
+    else
+        __y0_ = 0;
+    if (_Rp - __y0_ > __y0_ / __n_)
+    {
+        ++__n_;
+        __w0_ = __w_ / __n_;
+        if (__w0_ < _WDt)
+            __y0_ = (_Rp >> __w0_) << __w0_;
+        else
+            __y0_ = 0;
+    }
+    __n0_ = __n_ - __w_ % __n_;
+    if (__w0_ < _WDt - 1)
+        __y1_ = (_Rp >> (__w0_ + 1)) << (__w0_ + 1);
+    else
+        __y1_ = 0;
+    __mask0_ = __w0_ > 0 ? _Engine_result_type(~0) >> (_EDt - __w0_) :
+                          _Engine_result_type(0);
+    __mask1_ = __w0_ < _EDt - 1 ?
+                               _Engine_result_type(~0) >> (_EDt - (__w0_ + 1)) :
+                               _Engine_result_type(~0);
+}
+
+template<class _Engine, class _UIntType>
+inline
+_UIntType
+__independent_bits_engine<_Engine, _UIntType>::__eval(false_type)
+{
+    return static_cast<result_type>(__e_() & __mask0_);
+}
+
+template<class _Engine, class _UIntType>
+_UIntType
+__independent_bits_engine<_Engine, _UIntType>::__eval(true_type)
+{
+    const size_t _WRt = numeric_limits<result_type>::digits;
+    result_type _Sp = 0;
+    for (size_t __k = 0; __k < __n0_; ++__k)
+    {
+        _Engine_result_type __u;
+        do
+        {
+            __u = __e_() - _Engine::min();
+        } while (__u >= __y0_);
+        if (__w0_ < _WRt)
+            _Sp <<= __w0_;
+        else
+            _Sp = 0;
+        _Sp += __u & __mask0_;
+    }
+    for (size_t __k = __n0_; __k < __n_; ++__k)
+    {
+        _Engine_result_type __u;
+        do
+        {
+            __u = __e_() - _Engine::min();
+        } while (__u >= __y1_);
+        if (__w0_ < _WRt - 1)
+            _Sp <<= __w0_ + 1;
+        else
+            _Sp = 0;
+        _Sp += __u & __mask1_;
+    }
+    return _Sp;
+}
+
+// uniform_int_distribution
+
+template<class _IntType = int>
+class uniform_int_distribution
+{
+public:
+    // types
+    typedef _IntType result_type;
+
+    class param_type
+    {
+        result_type __a_;
+        result_type __b_;
+    public:
+        typedef uniform_int_distribution distribution_type;
+
+        explicit param_type(result_type __a = 0,
+                            result_type __b = numeric_limits<result_type>::max())
+            : __a_(__a), __b_(__b) {}
+
+        result_type a() const {return __a_;}
+        result_type b() const {return __b_;}
+
+        friend bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__a_ == __y.__a_ && __x.__b_ == __y.__b_;}
+        friend bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    explicit uniform_int_distribution(result_type __a = 0,
+                                      result_type __b = numeric_limits<result_type>::max())
+        : __p_(param_type(__a, __b)) {}
+    explicit uniform_int_distribution(const param_type& __p) : __p_(__p) {}
+    void reset() {}
+
+    // generating functions
+    template<class _URNG> result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    result_type a() const {return __p_.a();}
+    result_type b() const {return __p_.b();}
+
+    param_type param() const {return __p_;}
+    void param(const param_type& __p) {__p_ = __p;}
+
+    result_type min() const {return a();}
+    result_type max() const {return b();}
+
+    friend bool operator==(const uniform_int_distribution& __x,
+                           const uniform_int_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend bool operator!=(const uniform_int_distribution& __x,
+                           const uniform_int_distribution& __y)
+            {return !(__x == __y);}
+};
+
+template<class _IntType>
+template<class _URNG>
+typename uniform_int_distribution<_IntType>::result_type
+uniform_int_distribution<_IntType>::operator()(_URNG& __g, const param_type& __p)
+_LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+{
+    typedef typename conditional<sizeof(result_type) <= sizeof(uint32_t),
+                                            uint32_t, uint64_t>::type _UIntType;
+    const _UIntType _Rp = _UIntType(__p.b()) - _UIntType(__p.a()) + _UIntType(1);
+    if (_Rp == 1)
+        return __p.a();
+    const size_t _Dt = numeric_limits<_UIntType>::digits;
+    typedef __independent_bits_engine<_URNG, _UIntType> _Eng;
+    if (_Rp == 0)
+        return static_cast<result_type>(_Eng(__g, _Dt)());
+    size_t __w = _Dt - __libcpp_clz(_Rp) - 1;
+    if ((_Rp & (std::numeric_limits<_UIntType>::max() >> (_Dt - __w))) != 0)
+        ++__w;
+    _Eng __e(__g, __w);
+    _UIntType __u;
+    do
+    {
+        __u = __e();
+    } while (__u >= _Rp);
+    return static_cast<result_type>(__u + __p.a());
+}
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_RANDOM_SHUFFLE) \
+  || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+class _LIBCUDACXX_TYPE_VIS __rs_default;
+
+_LIBCUDACXX_FUNC_VIS __rs_default __rs_get();
+
+class _LIBCUDACXX_TYPE_VIS __rs_default
+{
+    static unsigned __c_;
+
+    __rs_default();
+public:
+    typedef uint_fast32_t result_type;
+
+    static const result_type _Min = 0;
+    static const result_type _Max = 0xFFFFFFFF;
+
+    __rs_default(const __rs_default&);
+    ~__rs_default();
+
+    result_type operator()();
+
+    static _LIBCUDACXX_CONSTEXPR result_type min() {return _Min;}
+    static _LIBCUDACXX_CONSTEXPR result_type max() {return _Max;}
+
+    friend _LIBCUDACXX_FUNC_VIS __rs_default __rs_get();
+};
+
+_LIBCUDACXX_FUNC_VIS __rs_default __rs_get();
+
+template <class _RandomAccessIterator>
+_LIBCUDACXX_DEPRECATED_IN_CXX14 void
+random_shuffle(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    typedef uniform_int_distribution<ptrdiff_t> _Dp;
+    typedef typename _Dp::param_type _Pp;
+    difference_type __d = __last - __first;
+    if (__d > 1)
+    {
+        _Dp __uid;
+        __rs_default __g = __rs_get();
+        for (--__last, (void) --__d; __first < __last; ++__first, (void) --__d)
+        {
+            difference_type __i = __uid(__g, _Pp(0, __d));
+            if (__i != difference_type(0))
+                swap(*__first, *(__first + __i));
+        }
+    }
+}
+
+template <class _RandomAccessIterator, class _RandomNumberGenerator>
+_LIBCUDACXX_DEPRECATED_IN_CXX14 void
+random_shuffle(_RandomAccessIterator __first, _RandomAccessIterator __last,
+#ifndef _LIBCUDACXX_CXX03_LANG
+               _RandomNumberGenerator&& __rand)
+#else
+               _RandomNumberGenerator& __rand)
+#endif
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    difference_type __d = __last - __first;
+    if (__d > 1)
+    {
+        for (--__last; __first < __last; ++__first, (void) --__d)
+        {
+            difference_type __i = __rand(__d);
+            if (__i != difference_type(0))
+              swap(*__first, *(__first + __i));
+        }
+    }
+}
+#endif
+
+template <class _PopulationIterator, class _SampleIterator, class _Distance,
+          class _UniformRandomNumberGenerator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_SampleIterator __sample(_PopulationIterator __first,
+                         _PopulationIterator __last, _SampleIterator __output_iter,
+                         _Distance __n,
+                         _UniformRandomNumberGenerator & __g,
+                         input_iterator_tag) {
+
+  _Distance __k = 0;
+  for (; __first != __last && __k < __n; ++__first, (void) ++__k)
+    __output_iter[__k] = *__first;
+  _Distance __sz = __k;
+  for (; __first != __last; ++__first, (void) ++__k) {
+    _Distance __r = _CUDA_VSTD::uniform_int_distribution<_Distance>(0, __k)(__g);
+    if (__r < __sz)
+      __output_iter[__r] = *__first;
+  }
+  return __output_iter + _CUDA_VSTD::min(__n, __k);
+}
+
+template <class _PopulationIterator, class _SampleIterator, class _Distance,
+          class _UniformRandomNumberGenerator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_SampleIterator __sample(_PopulationIterator __first,
+                         _PopulationIterator __last, _SampleIterator __output_iter,
+                         _Distance __n,
+                         _UniformRandomNumberGenerator& __g,
+                         forward_iterator_tag) {
+  _Distance __unsampled_sz = _CUDA_VSTD::distance(__first, __last);
+  for (__n = _CUDA_VSTD::min(__n, __unsampled_sz); __n != 0; ++__first) {
+    _Distance __r =
+        _CUDA_VSTD::uniform_int_distribution<_Distance>(0, --__unsampled_sz)(__g);
+    if (__r < __n) {
+      *__output_iter++ = *__first;
+      --__n;
+    }
+  }
+  return __output_iter;
+}
+
+template <class _PopulationIterator, class _SampleIterator, class _Distance,
+          class _UniformRandomNumberGenerator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_SampleIterator __sample(_PopulationIterator __first,
+                         _PopulationIterator __last, _SampleIterator __output_iter,
+                         _Distance __n, _UniformRandomNumberGenerator& __g) {
+  typedef typename iterator_traits<_PopulationIterator>::iterator_category
+        _PopCategory;
+  typedef typename iterator_traits<_PopulationIterator>::difference_type
+        _Difference;
+  static_assert(__is_forward_iterator<_PopulationIterator>::value ||
+                __is_random_access_iterator<_SampleIterator>::value,
+                "SampleIterator must meet the requirements of RandomAccessIterator");
+  typedef typename common_type<_Distance, _Difference>::type _CommonType;
+  _LIBCUDACXX_ASSERT(__n >= 0, "N must be a positive number.");
+  return _CUDA_VSTD::__sample(
+      __first, __last, __output_iter, _CommonType(__n),
+      __g, _PopCategory());
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _PopulationIterator, class _SampleIterator, class _Distance,
+          class _UniformRandomNumberGenerator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_SampleIterator sample(_PopulationIterator __first,
+                       _PopulationIterator __last, _SampleIterator __output_iter,
+                       _Distance __n, _UniformRandomNumberGenerator&& __g) {
+    return _CUDA_VSTD::__sample(__first, __last, __output_iter, __n, __g);
+}
+#endif // _LIBCUDACXX_STD_VER > 14
+
+template<class _RandomAccessIterator, class _UniformRandomNumberGenerator>
+_LIBCUDACXX_INLINE_VISIBILITY
+    void shuffle(_RandomAccessIterator __first, _RandomAccessIterator __last,
+                 _UniformRandomNumberGenerator&& __g)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    typedef uniform_int_distribution<ptrdiff_t> _Dp;
+    typedef typename _Dp::param_type _Pp;
+    difference_type __d = __last - __first;
+    if (__d > 1)
+    {
+        _Dp __uid;
+        for (--__last, (void) --__d; __first < __last; ++__first, (void) --__d)
+        {
+            difference_type __i = __uid(__g, _Pp(0, __d));
+            if (__i != difference_type(0))
+                swap(*__first, *(__first + __i));
+        }
+    }
+}
+
+template <class _InputIterator, class _Predicate>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+is_partitioned(_InputIterator __first, _InputIterator __last, _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+        if (!__pred(*__first))
+            break;
+    if ( __first == __last )
+        return true;
+    ++__first;
+    for (; __first != __last; ++__first)
+        if (__pred(*__first))
+            return false;
+    return true;
+}
+
+// partition
+
+template <class _Predicate, class _ForwardIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+__partition(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred, forward_iterator_tag)
+{
+    while (true)
+    {
+        if (__first == __last)
+            return __first;
+        if (!__pred(*__first))
+            break;
+        ++__first;
+    }
+    for (_ForwardIterator __p = __first; ++__p != __last;)
+    {
+        if (__pred(*__p))
+        {
+            swap(*__first, *__p);
+            ++__first;
+        }
+    }
+    return __first;
+}
+
+template <class _Predicate, class _BidirectionalIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_BidirectionalIterator
+__partition(_BidirectionalIterator __first, _BidirectionalIterator __last, _Predicate __pred,
+            bidirectional_iterator_tag)
+{
+    while (true)
+    {
+        while (true)
+        {
+            if (__first == __last)
+                return __first;
+            if (!__pred(*__first))
+                break;
+            ++__first;
+        }
+        do
+        {
+            if (__first == --__last)
+                return __first;
+        } while (!__pred(*__last));
+        swap(*__first, *__last);
+        ++__first;
+    }
+}
+
+template <class _ForwardIterator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+partition(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred)
+{
+    return _CUDA_VSTD::__partition<typename add_lvalue_reference<_Predicate>::type>
+                            (__first, __last, __pred, typename iterator_traits<_ForwardIterator>::iterator_category());
+}
+
+// partition_copy
+
+template <class _InputIterator, class _OutputIterator1,
+          class _OutputIterator2, class _Predicate>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 pair<_OutputIterator1, _OutputIterator2>
+partition_copy(_InputIterator __first, _InputIterator __last,
+               _OutputIterator1 __out_true, _OutputIterator2 __out_false,
+               _Predicate __pred)
+{
+    for (; __first != __last; ++__first)
+    {
+        if (__pred(*__first))
+        {
+            *__out_true = *__first;
+            ++__out_true;
+        }
+        else
+        {
+            *__out_false = *__first;
+            ++__out_false;
+        }
+    }
+    return pair<_OutputIterator1, _OutputIterator2>(__out_true, __out_false);
+}
+
+// partition_point
+
+template<class _ForwardIterator, class _Predicate>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+partition_point(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred)
+{
+    typedef typename iterator_traits<_ForwardIterator>::difference_type difference_type;
+    difference_type __len = _CUDA_VSTD::distance(__first, __last);
+    while (__len != 0)
+    {
+        difference_type __l2 = _CUDA_VSTD::__half_positive(__len);
+        _ForwardIterator __m = __first;
+        _CUDA_VSTD::advance(__m, __l2);
+        if (__pred(*__m))
+        {
+            __first = ++__m;
+            __len -= __l2 + 1;
+        }
+        else
+            __len = __l2;
+    }
+    return __first;
+}
+
+// stable_partition
+
+template <class _Predicate, class _ForwardIterator, class _Distance, class _Pair>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_ForwardIterator
+__stable_partition(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred,
+                   _Distance __len, _Pair __p, forward_iterator_tag __fit)
+{
+    // *__first is known to be false
+    // __len >= 1
+    if (__len == 1)
+        return __first;
+    if (__len == 2)
+    {
+        _ForwardIterator __m = __first;
+        if (__pred(*++__m))
+        {
+            swap(*__first, *__m);
+            return __m;
+        }
+        return __first;
+    }
+    if (__len <= __p.second)
+    {   // The buffer is big enough to use
+        typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+        __destruct_n __d(0);
+        unique_ptr<value_type, __destruct_n&> __h(__p.first, __d);
+        // Move the falses into the temporary buffer, and the trues to the front of the line
+        // Update __first to always point to the end of the trues
+        value_type* __t = __p.first;
+        ::new(__t) value_type(_CUDA_VSTD::move(*__first));
+        __d.__incr((value_type*)0);
+        ++__t;
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+        {
+            if (__pred(*__i))
+            {
+                *__first = _CUDA_VSTD::move(*__i);
+                ++__first;
+            }
+            else
+            {
+                ::new(__t) value_type(_CUDA_VSTD::move(*__i));
+                __d.__incr((value_type*)0);
+                ++__t;
+            }
+        }
+        // All trues now at start of range, all falses in buffer
+        // Move falses back into range, but don't mess up __first which points to first false
+        __i = __first;
+        for (value_type* __t2 = __p.first; __t2 < __t; ++__t2, (void) ++__i)
+            *__i = _CUDA_VSTD::move(*__t2);
+        // __h destructs moved-from values out of the temp buffer, but doesn't deallocate buffer
+        return __first;
+    }
+    // Else not enough buffer, do in place
+    // __len >= 3
+    _ForwardIterator __m = __first;
+    _Distance __len2 = __len / 2;  // __len2 >= 2
+    _CUDA_VSTD::advance(__m, __len2);
+    // recurse on [__first, __m), *__first know to be false
+    // F?????????????????
+    // f       m         l
+    typedef typename add_lvalue_reference<_Predicate>::type _PredRef;
+    _ForwardIterator __first_false = __stable_partition<_PredRef>(__first, __m, __pred, __len2, __p, __fit);
+    // TTTFFFFF??????????
+    // f  ff   m         l
+    // recurse on [__m, __last], except increase __m until *(__m) is false, *__last know to be true
+    _ForwardIterator __m1 = __m;
+    _ForwardIterator __second_false = __last;
+    _Distance __len_half = __len - __len2;
+    while (__pred(*__m1))
+    {
+        if (++__m1 == __last)
+            goto __second_half_done;
+        --__len_half;
+    }
+    // TTTFFFFFTTTF??????
+    // f  ff   m  m1     l
+    __second_false = __stable_partition<_PredRef>(__m1, __last, __pred, __len_half, __p, __fit);
+__second_half_done:
+    // TTTFFFFFTTTTTFFFFF
+    // f  ff   m    sf   l
+    return _CUDA_VSTD::rotate(__first_false, __m, __second_false);
+    // TTTTTTTTFFFFFFFFFF
+    //         |
+}
+
+struct __return_temporary_buffer
+{
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY void operator()(_Tp* __p) const {_CUDA_VSTD::return_temporary_buffer(__p);}
+};
+
+template <class _Predicate, class _ForwardIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_ForwardIterator
+__stable_partition(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred,
+                   forward_iterator_tag)
+{
+    const unsigned __alloc_limit = 3;  // might want to make this a function of trivial assignment
+    // Either prove all true and return __first or point to first false
+    while (true)
+    {
+        if (__first == __last)
+            return __first;
+        if (!__pred(*__first))
+            break;
+        ++__first;
+    }
+    // We now have a reduced range [__first, __last)
+    // *__first is known to be false
+    typedef typename iterator_traits<_ForwardIterator>::difference_type difference_type;
+    typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+    difference_type __len = _CUDA_VSTD::distance(__first, __last);
+    pair<value_type*, ptrdiff_t> __p(0, 0);
+    unique_ptr<value_type, __return_temporary_buffer> __h;
+    if (__len >= __alloc_limit)
+    {
+        __p = _CUDA_VSTD::get_temporary_buffer<value_type>(__len);
+        __h.reset(__p.first);
+    }
+    return __stable_partition<typename add_lvalue_reference<_Predicate>::type>
+                             (__first, __last, __pred, __len, __p, forward_iterator_tag());
+}
+
+template <class _Predicate, class _BidirectionalIterator, class _Distance, class _Pair>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_BidirectionalIterator
+__stable_partition(_BidirectionalIterator __first, _BidirectionalIterator __last, _Predicate __pred,
+                   _Distance __len, _Pair __p, bidirectional_iterator_tag __bit)
+{
+    // *__first is known to be false
+    // *__last is known to be true
+    // __len >= 2
+    if (__len == 2)
+    {
+        swap(*__first, *__last);
+        return __last;
+    }
+    if (__len == 3)
+    {
+        _BidirectionalIterator __m = __first;
+        if (__pred(*++__m))
+        {
+            swap(*__first, *__m);
+            swap(*__m, *__last);
+            return __last;
+        }
+        swap(*__m, *__last);
+        swap(*__first, *__m);
+        return __m;
+    }
+    if (__len <= __p.second)
+    {   // The buffer is big enough to use
+        typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
+        __destruct_n __d(0);
+        unique_ptr<value_type, __destruct_n&> __h(__p.first, __d);
+        // Move the falses into the temporary buffer, and the trues to the front of the line
+        // Update __first to always point to the end of the trues
+        value_type* __t = __p.first;
+        ::new(__t) value_type(_CUDA_VSTD::move(*__first));
+        __d.__incr((value_type*)0);
+        ++__t;
+        _BidirectionalIterator __i = __first;
+        while (++__i != __last)
+        {
+            if (__pred(*__i))
+            {
+                *__first = _CUDA_VSTD::move(*__i);
+                ++__first;
+            }
+            else
+            {
+                ::new(__t) value_type(_CUDA_VSTD::move(*__i));
+                __d.__incr((value_type*)0);
+                ++__t;
+            }
+        }
+        // move *__last, known to be true
+        *__first = _CUDA_VSTD::move(*__i);
+        __i = ++__first;
+        // All trues now at start of range, all falses in buffer
+        // Move falses back into range, but don't mess up __first which points to first false
+        for (value_type* __t2 = __p.first; __t2 < __t; ++__t2, (void) ++__i)
+            *__i = _CUDA_VSTD::move(*__t2);
+        // __h destructs moved-from values out of the temp buffer, but doesn't deallocate buffer
+        return __first;
+    }
+    // Else not enough buffer, do in place
+    // __len >= 4
+    _BidirectionalIterator __m = __first;
+    _Distance __len2 = __len / 2;  // __len2 >= 2
+    _CUDA_VSTD::advance(__m, __len2);
+    // recurse on [__first, __m-1], except reduce __m-1 until *(__m-1) is true, *__first know to be false
+    // F????????????????T
+    // f       m        l
+    _BidirectionalIterator __m1 = __m;
+    _BidirectionalIterator __first_false = __first;
+    _Distance __len_half = __len2;
+    while (!__pred(*--__m1))
+    {
+        if (__m1 == __first)
+            goto __first_half_done;
+        --__len_half;
+    }
+    // F???TFFF?????????T
+    // f   m1  m        l
+    typedef typename add_lvalue_reference<_Predicate>::type _PredRef;
+    __first_false = __stable_partition<_PredRef>(__first, __m1, __pred, __len_half, __p, __bit);
+__first_half_done:
+    // TTTFFFFF?????????T
+    // f  ff   m        l
+    // recurse on [__m, __last], except increase __m until *(__m) is false, *__last know to be true
+    __m1 = __m;
+    _BidirectionalIterator __second_false = __last;
+    ++__second_false;
+    __len_half = __len - __len2;
+    while (__pred(*__m1))
+    {
+        if (++__m1 == __last)
+            goto __second_half_done;
+        --__len_half;
+    }
+    // TTTFFFFFTTTF?????T
+    // f  ff   m  m1    l
+    __second_false = __stable_partition<_PredRef>(__m1, __last, __pred, __len_half, __p, __bit);
+__second_half_done:
+    // TTTFFFFFTTTTTFFFFF
+    // f  ff   m    sf  l
+    return _CUDA_VSTD::rotate(__first_false, __m, __second_false);
+    // TTTTTTTTFFFFFFFFFF
+    //         |
+}
+
+template <class _Predicate, class _BidirectionalIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_BidirectionalIterator
+__stable_partition(_BidirectionalIterator __first, _BidirectionalIterator __last, _Predicate __pred,
+                   bidirectional_iterator_tag)
+{
+    typedef typename iterator_traits<_BidirectionalIterator>::difference_type difference_type;
+    typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
+    const difference_type __alloc_limit = 4;  // might want to make this a function of trivial assignment
+    // Either prove all true and return __first or point to first false
+    while (true)
+    {
+        if (__first == __last)
+            return __first;
+        if (!__pred(*__first))
+            break;
+        ++__first;
+    }
+    // __first points to first false, everything prior to __first is already set.
+    // Either prove [__first, __last) is all false and return __first, or point __last to last true
+    do
+    {
+        if (__first == --__last)
+            return __first;
+    } while (!__pred(*__last));
+    // We now have a reduced range [__first, __last]
+    // *__first is known to be false
+    // *__last is known to be true
+    // __len >= 2
+    difference_type __len = _CUDA_VSTD::distance(__first, __last) + 1;
+    pair<value_type*, ptrdiff_t> __p(0, 0);
+    unique_ptr<value_type, __return_temporary_buffer> __h;
+    if (__len >= __alloc_limit)
+    {
+        __p = _CUDA_VSTD::get_temporary_buffer<value_type>(__len);
+        __h.reset(__p.first);
+    }
+    return __stable_partition<typename add_lvalue_reference<_Predicate>::type>
+                             (__first, __last, __pred, __len, __p, bidirectional_iterator_tag());
+}
+
+template <class _ForwardIterator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator
+stable_partition(_ForwardIterator __first, _ForwardIterator __last, _Predicate __pred)
+{
+    return __stable_partition<typename add_lvalue_reference<_Predicate>::type>
+                             (__first, __last, __pred, typename iterator_traits<_ForwardIterator>::iterator_category());
+}
+
+// is_sorted_until
+
+template <class _ForwardIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+is_sorted_until(_ForwardIterator __first, _ForwardIterator __last, _Compare __comp)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __i = __first;
+        while (++__i != __last)
+        {
+            if (__comp(*__i, *__first))
+                return __i;
+            __first = __i;
+        }
+    }
+    return __last;
+}
+
+template<class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+is_sorted_until(_ForwardIterator __first, _ForwardIterator __last)
+{
+    return _CUDA_VSTD::is_sorted_until(__first, __last, __less<typename iterator_traits<_ForwardIterator>::value_type>());
+}
+
+// is_sorted
+
+template <class _ForwardIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+is_sorted(_ForwardIterator __first, _ForwardIterator __last, _Compare __comp)
+{
+    return _CUDA_VSTD::is_sorted_until(__first, __last, __comp) == __last;
+}
+
+template<class _ForwardIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+is_sorted(_ForwardIterator __first, _ForwardIterator __last)
+{
+    return _CUDA_VSTD::is_sorted(__first, __last, __less<typename iterator_traits<_ForwardIterator>::value_type>());
+}
+
+// sort
+
+// stable, 2-3 compares, 0-2 swaps
+
+template <class _Compare, class _ForwardIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+unsigned
+__sort3(_ForwardIterator __x, _ForwardIterator __y, _ForwardIterator __z, _Compare __c)
+{
+    unsigned __r = 0;
+    if (!__c(*__y, *__x))          // if x <= y
+    {
+        if (!__c(*__z, *__y))      // if y <= z
+            return __r;            // x <= y && y <= z
+                                   // x <= y && y > z
+        swap(*__y, *__z);          // x <= z && y < z
+        __r = 1;
+        if (__c(*__y, *__x))       // if x > y
+        {
+            swap(*__x, *__y);      // x < y && y <= z
+            __r = 2;
+        }
+        return __r;                // x <= y && y < z
+    }
+    if (__c(*__z, *__y))           // x > y, if y > z
+    {
+        swap(*__x, *__z);          // x < y && y < z
+        __r = 1;
+        return __r;
+    }
+    swap(*__x, *__y);              // x > y && y <= z
+    __r = 1;                       // x < y && x <= z
+    if (__c(*__z, *__y))           // if y > z
+    {
+        swap(*__y, *__z);          // x <= y && y < z
+        __r = 2;
+    }
+    return __r;
+}                                  // x <= y && y <= z
+
+// stable, 3-6 compares, 0-5 swaps
+
+template <class _Compare, class _ForwardIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+unsigned
+__sort4(_ForwardIterator __x1, _ForwardIterator __x2, _ForwardIterator __x3,
+            _ForwardIterator __x4, _Compare __c)
+{
+    unsigned __r = __sort3<_Compare>(__x1, __x2, __x3, __c);
+    if (__c(*__x4, *__x3))
+    {
+        swap(*__x3, *__x4);
+        ++__r;
+        if (__c(*__x3, *__x2))
+        {
+            swap(*__x2, *__x3);
+            ++__r;
+            if (__c(*__x2, *__x1))
+            {
+                swap(*__x1, *__x2);
+                ++__r;
+            }
+        }
+    }
+    return __r;
+}
+
+// stable, 4-10 compares, 0-9 swaps
+
+template <class _Compare, class _ForwardIterator>
+_LIBCUDACXX_HIDDEN
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+unsigned
+__sort5(_ForwardIterator __x1, _ForwardIterator __x2, _ForwardIterator __x3,
+            _ForwardIterator __x4, _ForwardIterator __x5, _Compare __c)
+{
+    unsigned __r = __sort4<_Compare>(__x1, __x2, __x3, __x4, __c);
+    if (__c(*__x5, *__x4))
+    {
+        swap(*__x4, *__x5);
+        ++__r;
+        if (__c(*__x4, *__x3))
+        {
+            swap(*__x3, *__x4);
+            ++__r;
+            if (__c(*__x3, *__x2))
+            {
+                swap(*__x2, *__x3);
+                ++__r;
+                if (__c(*__x2, *__x1))
+                {
+                    swap(*__x1, *__x2);
+                    ++__r;
+                }
+            }
+        }
+    }
+    return __r;
+}
+
+// Assumes size > 0
+template <class _Compare, class _BirdirectionalIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__selection_sort(_BirdirectionalIterator __first, _BirdirectionalIterator __last, _Compare __comp)
+{
+    _BirdirectionalIterator __lm1 = __last;
+    for (--__lm1; __first != __lm1; ++__first)
+    {
+        _BirdirectionalIterator __i = _CUDA_VSTD::min_element<_BirdirectionalIterator,
+                                                        typename add_lvalue_reference<_Compare>::type>
+                                                       (__first, __last, __comp);
+        if (__i != __first)
+            swap(*__first, *__i);
+    }
+}
+
+template <class _Compare, class _BirdirectionalIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__insertion_sort(_BirdirectionalIterator __first, _BirdirectionalIterator __last, _Compare __comp)
+{
+    typedef typename iterator_traits<_BirdirectionalIterator>::value_type value_type;
+    if (__first != __last)
+    {
+        _BirdirectionalIterator __i = __first;
+        for (++__i; __i != __last; ++__i)
+        {
+            _BirdirectionalIterator __j = __i;
+            value_type __t(_CUDA_VSTD::move(*__j));
+            for (_BirdirectionalIterator __k = __i; __k != __first && __comp(__t,  *--__k); --__j)
+                *__j = _CUDA_VSTD::move(*__k);
+            *__j = _CUDA_VSTD::move(__t);
+        }
+    }
+}
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__insertion_sort_3(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    _RandomAccessIterator __j = __first+2;
+    __sort3<_Compare>(__first, __first+1, __j, __comp);
+    for (_RandomAccessIterator __i = __j+1; __i != __last; ++__i)
+    {
+        if (__comp(*__i, *__j))
+        {
+            value_type __t(_CUDA_VSTD::move(*__i));
+            _RandomAccessIterator __k = __j;
+            __j = __i;
+            do
+            {
+                *__j = _CUDA_VSTD::move(*__k);
+                __j = __k;
+            } while (__j != __first && __comp(__t, *--__k));
+            *__j = _CUDA_VSTD::move(__t);
+        }
+        __j = __i;
+    }
+}
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+bool
+__insertion_sort_incomplete(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    switch (__last - __first)
+    {
+    case 0:
+    case 1:
+        return true;
+    case 2:
+        if (__comp(*--__last, *__first))
+            swap(*__first, *__last);
+        return true;
+    case 3:
+        _CUDA_VSTD::__sort3<_Compare>(__first, __first+1, --__last, __comp);
+        return true;
+    case 4:
+        _CUDA_VSTD::__sort4<_Compare>(__first, __first+1, __first+2, --__last, __comp);
+        return true;
+    case 5:
+        _CUDA_VSTD::__sort5<_Compare>(__first, __first+1, __first+2, __first+3, --__last, __comp);
+        return true;
+    }
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    _RandomAccessIterator __j = __first+2;
+    __sort3<_Compare>(__first, __first+1, __j, __comp);
+    const unsigned __limit = 8;
+    unsigned __count = 0;
+    for (_RandomAccessIterator __i = __j+1; __i != __last; ++__i)
+    {
+        if (__comp(*__i, *__j))
+        {
+            value_type __t(_CUDA_VSTD::move(*__i));
+            _RandomAccessIterator __k = __j;
+            __j = __i;
+            do
+            {
+                *__j = _CUDA_VSTD::move(*__k);
+                __j = __k;
+            } while (__j != __first && __comp(__t, *--__k));
+            *__j = _CUDA_VSTD::move(__t);
+            if (++__count == __limit)
+                return ++__i == __last;
+        }
+        __j = __i;
+    }
+    return true;
+}
+
+template <class _Compare, class _BirdirectionalIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__insertion_sort_move(_BirdirectionalIterator __first1, _BirdirectionalIterator __last1,
+                      typename iterator_traits<_BirdirectionalIterator>::value_type* __first2, _Compare __comp)
+{
+    typedef typename iterator_traits<_BirdirectionalIterator>::value_type value_type;
+    if (__first1 != __last1)
+    {
+        __destruct_n __d(0);
+        unique_ptr<value_type, __destruct_n&> __h(__first2, __d);
+        value_type* __last2 = __first2;
+        ::new(__last2) value_type(_CUDA_VSTD::move(*__first1));
+        __d.__incr((value_type*)0);
+        for (++__last2; ++__first1 != __last1; ++__last2)
+        {
+            value_type* __j2 = __last2;
+            value_type* __i2 = __j2;
+            if (__comp(*__first1, *--__i2))
+            {
+                ::new(__j2) value_type(_CUDA_VSTD::move(*__i2));
+                __d.__incr((value_type*)0);
+                for (--__j2; __i2 != __first2 && __comp(*__first1,  *--__i2); --__j2)
+                    *__j2 = _CUDA_VSTD::move(*__i2);
+                *__j2 = _CUDA_VSTD::move(*__first1);
+            }
+            else
+            {
+                ::new(__j2) value_type(_CUDA_VSTD::move(*__first1));
+                __d.__incr((value_type*)0);
+            }
+        }
+        __h.release();
+    }
+}
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    // _Compare is known to be a reference type
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    const difference_type __limit = is_trivially_copy_constructible<value_type>::value &&
+                                    is_trivially_copy_assignable<value_type>::value ? 30 : 6;
+    while (true)
+    {
+    __restart:
+        difference_type __len = __last - __first;
+        switch (__len)
+        {
+        case 0:
+        case 1:
+            return;
+        case 2:
+            if (__comp(*--__last, *__first))
+                swap(*__first, *__last);
+            return;
+        case 3:
+            _CUDA_VSTD::__sort3<_Compare>(__first, __first+1, --__last, __comp);
+            return;
+        case 4:
+            _CUDA_VSTD::__sort4<_Compare>(__first, __first+1, __first+2, --__last, __comp);
+            return;
+        case 5:
+            _CUDA_VSTD::__sort5<_Compare>(__first, __first+1, __first+2, __first+3, --__last, __comp);
+            return;
+        }
+        if (__len <= __limit)
+        {
+            _CUDA_VSTD::__insertion_sort_3<_Compare>(__first, __last, __comp);
+            return;
+        }
+        // __len > 5
+        _RandomAccessIterator __m = __first;
+        _RandomAccessIterator __lm1 = __last;
+        --__lm1;
+        unsigned __n_swaps;
+        {
+        difference_type __delta;
+        if (__len >= 1000)
+        {
+            __delta = __len/2;
+            __m += __delta;
+            __delta /= 2;
+            __n_swaps = _CUDA_VSTD::__sort5<_Compare>(__first, __first + __delta, __m, __m+__delta, __lm1, __comp);
+        }
+        else
+        {
+            __delta = __len/2;
+            __m += __delta;
+            __n_swaps = _CUDA_VSTD::__sort3<_Compare>(__first, __m, __lm1, __comp);
+        }
+        }
+        // *__m is median
+        // partition [__first, __m) < *__m and *__m <= [__m, __last)
+        // (this inhibits tossing elements equivalent to __m around unnecessarily)
+        _RandomAccessIterator __i = __first;
+        _RandomAccessIterator __j = __lm1;
+        // j points beyond range to be tested, *__m is known to be <= *__lm1
+        // The search going up is known to be guarded but the search coming down isn't.
+        // Prime the downward search with a guard.
+        if (!__comp(*__i, *__m))  // if *__first == *__m
+        {
+            // *__first == *__m, *__first doesn't go in first part
+            // manually guard downward moving __j against __i
+            while (true)
+            {
+                if (__i == --__j)
+                {
+                    // *__first == *__m, *__m <= all other elements
+                    // Parition instead into [__first, __i) == *__first and *__first < [__i, __last)
+                    ++__i;  // __first + 1
+                    __j = __last;
+                    if (!__comp(*__first, *--__j))  // we need a guard if *__first == *(__last-1)
+                    {
+                        while (true)
+                        {
+                            if (__i == __j)
+                                return;  // [__first, __last) all equivalent elements
+                            if (__comp(*__first, *__i))
+                            {
+                                swap(*__i, *__j);
+                                ++__n_swaps;
+                                ++__i;
+                                break;
+                            }
+                            ++__i;
+                        }
+                    }
+                    // [__first, __i) == *__first and *__first < [__j, __last) and __j == __last - 1
+                    if (__i == __j)
+                        return;
+                    while (true)
+                    {
+                        while (!__comp(*__first, *__i))
+                            ++__i;
+                        while (__comp(*__first, *--__j))
+                            ;
+                        if (__i >= __j)
+                            break;
+                        swap(*__i, *__j);
+                        ++__n_swaps;
+                        ++__i;
+                    }
+                    // [__first, __i) == *__first and *__first < [__i, __last)
+                    // The first part is sorted, sort the secod part
+                    // _CUDA_VSTD::__sort<_Compare>(__i, __last, __comp);
+                    __first = __i;
+                    goto __restart;
+                }
+                if (__comp(*__j, *__m))
+                {
+                    swap(*__i, *__j);
+                    ++__n_swaps;
+                    break;  // found guard for downward moving __j, now use unguarded partition
+                }
+            }
+        }
+        // It is known that *__i < *__m
+        ++__i;
+        // j points beyond range to be tested, *__m is known to be <= *__lm1
+        // if not yet partitioned...
+        if (__i < __j)
+        {
+            // known that *(__i - 1) < *__m
+            // known that __i <= __m
+            while (true)
+            {
+                // __m still guards upward moving __i
+                while (__comp(*__i, *__m))
+                    ++__i;
+                // It is now known that a guard exists for downward moving __j
+                while (!__comp(*--__j, *__m))
+                    ;
+                if (__i > __j)
+                    break;
+                swap(*__i, *__j);
+                ++__n_swaps;
+                // It is known that __m != __j
+                // If __m just moved, follow it
+                if (__m == __i)
+                    __m = __j;
+                ++__i;
+            }
+        }
+        // [__first, __i) < *__m and *__m <= [__i, __last)
+        if (__i != __m && __comp(*__m, *__i))
+        {
+            swap(*__i, *__m);
+            ++__n_swaps;
+        }
+        // [__first, __i) < *__i and *__i <= [__i+1, __last)
+        // If we were given a perfect partition, see if insertion sort is quick...
+        if (__n_swaps == 0)
+        {
+            bool __fs = _CUDA_VSTD::__insertion_sort_incomplete<_Compare>(__first, __i, __comp);
+            if (_CUDA_VSTD::__insertion_sort_incomplete<_Compare>(__i+1, __last, __comp))
+            {
+                if (__fs)
+                    return;
+                __last = __i;
+                continue;
+            }
+            else
+            {
+                if (__fs)
+                {
+                    __first = ++__i;
+                    continue;
+                }
+            }
+        }
+        // sort smaller range with recursive call and larger with tail recursion elimination
+        if (__i - __first < __last - __i)
+        {
+            _CUDA_VSTD::__sort<_Compare>(__first, __i, __comp);
+            // _CUDA_VSTD::__sort<_Compare>(__i+1, __last, __comp);
+            __first = ++__i;
+        }
+        else
+        {
+            _CUDA_VSTD::__sort<_Compare>(__i+1, __last, __comp);
+            // _CUDA_VSTD::__sort<_Compare>(__first, __i, __comp);
+            __last = __i;
+        }
+    }
+}
+
+// This forwarder keeps the top call and the recursive calls using the same instantiation, forcing a reference _Compare
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    _CUDA_VSTD::__sort<_Comp_ref>(__first, __last, _Comp_ref(__comp));
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::sort(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort(_Tp** __first, _Tp** __last)
+{
+    _CUDA_VSTD::sort((size_t*)__first, (size_t*)__last, __less<size_t>());
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort(__wrap_iter<_Tp*> __first, __wrap_iter<_Tp*> __last)
+{
+    _CUDA_VSTD::sort(__first.base(), __last.base());
+}
+
+template <class _Tp, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort(__wrap_iter<_Tp*> __first, __wrap_iter<_Tp*> __last, _Compare __comp)
+{
+    typedef typename add_lvalue_reference<_Compare>::type _Comp_ref;
+    _CUDA_VSTD::sort<_Tp*, _Comp_ref>(__first.base(), __last.base(), __comp);
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<char>&, char*>(char*, char*, __less<char>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<wchar_t>&, wchar_t*>(wchar_t*, wchar_t*, __less<wchar_t>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<signed char>&, signed char*>(signed char*, signed char*, __less<signed char>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<unsigned char>&, unsigned char*>(unsigned char*, unsigned char*, __less<unsigned char>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<short>&, short*>(short*, short*, __less<short>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<unsigned short>&, unsigned short*>(unsigned short*, unsigned short*, __less<unsigned short>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<int>&, int*>(int*, int*, __less<int>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<unsigned>&, unsigned*>(unsigned*, unsigned*, __less<unsigned>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<long>&, long*>(long*, long*, __less<long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<unsigned long>&, unsigned long*>(unsigned long*, unsigned long*, __less<unsigned long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<long long>&, long long*>(long long*, long long*, __less<long long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<unsigned long long>&, unsigned long long*>(unsigned long long*, unsigned long long*, __less<unsigned long long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<float>&, float*>(float*, float*, __less<float>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<double>&, double*>(double*, double*, __less<double>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void __sort<__less<long double>&, long double*>(long double*, long double*, __less<long double>&))
+
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<char>&, char*>(char*, char*, __less<char>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<wchar_t>&, wchar_t*>(wchar_t*, wchar_t*, __less<wchar_t>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<signed char>&, signed char*>(signed char*, signed char*, __less<signed char>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned char>&, unsigned char*>(unsigned char*, unsigned char*, __less<unsigned char>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<short>&, short*>(short*, short*, __less<short>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned short>&, unsigned short*>(unsigned short*, unsigned short*, __less<unsigned short>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<int>&, int*>(int*, int*, __less<int>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned>&, unsigned*>(unsigned*, unsigned*, __less<unsigned>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<long>&, long*>(long*, long*, __less<long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned long>&, unsigned long*>(unsigned long*, unsigned long*, __less<unsigned long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<long long>&, long long*>(long long*, long long*, __less<long long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<unsigned long long>&, unsigned long long*>(unsigned long long*, unsigned long long*, __less<unsigned long long>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<float>&, float*>(float*, float*, __less<float>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<double>&, double*>(double*, double*, __less<double>&))
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS bool __insertion_sort_incomplete<__less<long double>&, long double*>(long double*, long double*, __less<long double>&))
+
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS unsigned __sort5<__less<long double>&, long double*>(long double*, long double*, long double*, long double*, long double*, __less<long double>&))
+
+// lower_bound
+
+template <class _Compare, class _ForwardIterator, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+__lower_bound(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename iterator_traits<_ForwardIterator>::difference_type difference_type;
+    difference_type __len = _CUDA_VSTD::distance(__first, __last);
+    while (__len != 0)
+    {
+        difference_type __l2 = _CUDA_VSTD::__half_positive(__len);
+        _ForwardIterator __m = __first;
+        _CUDA_VSTD::advance(__m, __l2);
+        if (__comp(*__m, __value_))
+        {
+            __first = ++__m;
+            __len -= __l2 + 1;
+        }
+        else
+            __len = __l2;
+    }
+    return __first;
+}
+
+template <class _ForwardIterator, class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+lower_bound(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename add_lvalue_reference<_Compare>::type _Comp_ref;
+    return __lower_bound<_Comp_ref>(__first, __last, __value_, __comp);
+}
+
+template <class _ForwardIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+lower_bound(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_)
+{
+    return _CUDA_VSTD::lower_bound(__first, __last, __value_,
+                             __less<typename iterator_traits<_ForwardIterator>::value_type, _Tp>());
+}
+
+// upper_bound
+
+template <class _Compare, class _ForwardIterator, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _ForwardIterator
+__upper_bound(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename iterator_traits<_ForwardIterator>::difference_type difference_type;
+    difference_type __len = _CUDA_VSTD::distance(__first, __last);
+    while (__len != 0)
+    {
+        difference_type __l2 = _CUDA_VSTD::__half_positive(__len);
+        _ForwardIterator __m = __first;
+        _CUDA_VSTD::advance(__m, __l2);
+        if (__comp(__value_, *__m))
+            __len = __l2;
+        else
+        {
+            __first = ++__m;
+            __len -= __l2 + 1;
+        }
+    }
+    return __first;
+}
+
+template <class _ForwardIterator, class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+upper_bound(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename add_lvalue_reference<_Compare>::type _Comp_ref;
+    return __upper_bound<_Comp_ref>(__first, __last, __value_, __comp);
+}
+
+template <class _ForwardIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator
+upper_bound(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_)
+{
+    return _CUDA_VSTD::upper_bound(__first, __last, __value_,
+                             __less<_Tp, typename iterator_traits<_ForwardIterator>::value_type>());
+}
+
+// equal_range
+
+template <class _Compare, class _ForwardIterator, class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 pair<_ForwardIterator, _ForwardIterator>
+__equal_range(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename iterator_traits<_ForwardIterator>::difference_type difference_type;
+    difference_type __len = _CUDA_VSTD::distance(__first, __last);
+    while (__len != 0)
+    {
+        difference_type __l2 = _CUDA_VSTD::__half_positive(__len);
+        _ForwardIterator __m = __first;
+        _CUDA_VSTD::advance(__m, __l2);
+        if (__comp(*__m, __value_))
+        {
+            __first = ++__m;
+            __len -= __l2 + 1;
+        }
+        else if (__comp(__value_, *__m))
+        {
+            __last = __m;
+            __len = __l2;
+        }
+        else
+        {
+            _ForwardIterator __mp1 = __m;
+            return pair<_ForwardIterator, _ForwardIterator>
+                   (
+                      __lower_bound<_Compare>(__first, __m, __value_, __comp),
+                      __upper_bound<_Compare>(++__mp1, __last, __value_, __comp)
+                   );
+        }
+    }
+    return pair<_ForwardIterator, _ForwardIterator>(__first, __first);
+}
+
+template <class _ForwardIterator, class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+pair<_ForwardIterator, _ForwardIterator>
+equal_range(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __equal_range<_Comp_ref>(__first, __last, __value_, __comp);
+}
+
+template <class _ForwardIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+pair<_ForwardIterator, _ForwardIterator>
+equal_range(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_)
+{
+    return _CUDA_VSTD::equal_range(__first, __last, __value_,
+                             __less<typename iterator_traits<_ForwardIterator>::value_type, _Tp>());
+}
+
+// binary_search
+
+template <class _Compare, class _ForwardIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+__binary_search(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    __first = __lower_bound<_Compare>(__first, __last, __value_, __comp);
+    return __first != __last && !__comp(__value_, *__first);
+}
+
+template <class _ForwardIterator, class _Tp, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+binary_search(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __binary_search<_Comp_ref>(__first, __last, __value_, __comp);
+}
+
+template <class _ForwardIterator, class _Tp>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+binary_search(_ForwardIterator __first, _ForwardIterator __last, const _Tp& __value_)
+{
+    return _CUDA_VSTD::binary_search(__first, __last, __value_,
+                             __less<typename iterator_traits<_ForwardIterator>::value_type, _Tp>());
+}
+
+// merge
+
+template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+__merge(_InputIterator1 __first1, _InputIterator1 __last1,
+        _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    for (; __first1 != __last1; ++__result)
+    {
+        if (__first2 == __last2)
+            return _CUDA_VSTD::copy(__first1, __last1, __result);
+        if (__comp(*__first2, *__first1))
+        {
+            *__result = *__first2;
+            ++__first2;
+        }
+        else
+        {
+            *__result = *__first1;
+            ++__first1;
+        }
+    }
+    return _CUDA_VSTD::copy(__first2, __last2, __result);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+merge(_InputIterator1 __first1, _InputIterator1 __last1,
+      _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return _CUDA_VSTD::__merge<_Comp_ref>(__first1, __last1, __first2, __last2, __result, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+merge(_InputIterator1 __first1, _InputIterator1 __last1,
+      _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result)
+{
+    typedef typename iterator_traits<_InputIterator1>::value_type __v1;
+    typedef typename iterator_traits<_InputIterator2>::value_type __v2;
+    return _CUDA_VSTD::merge(__first1, __last1, __first2, __last2, __result, __less<__v1, __v2>());
+}
+
+// inplace_merge
+
+template <class _Compare, class _InputIterator1, class _InputIterator2,
+          class _OutputIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void __half_inplace_merge(_InputIterator1 __first1, _InputIterator1 __last1,
+                          _InputIterator2 __first2, _InputIterator2 __last2,
+                          _OutputIterator __result, _Compare __comp)
+{
+    for (; __first1 != __last1; ++__result)
+    {
+        if (__first2 == __last2)
+        {
+            _CUDA_VSTD::move(__first1, __last1, __result);
+            return;
+        }
+
+        if (__comp(*__first2, *__first1))
+        {
+            *__result = _CUDA_VSTD::move(*__first2);
+            ++__first2;
+        }
+        else
+        {
+            *__result = _CUDA_VSTD::move(*__first1);
+            ++__first1;
+        }
+    }
+    // __first2 through __last2 are already in the right spot.
+}
+
+template <class _Compare, class _BidirectionalIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__buffered_inplace_merge(_BidirectionalIterator __first, _BidirectionalIterator __middle, _BidirectionalIterator __last,
+                _Compare __comp, typename iterator_traits<_BidirectionalIterator>::difference_type __len1,
+                                 typename iterator_traits<_BidirectionalIterator>::difference_type __len2,
+                typename iterator_traits<_BidirectionalIterator>::value_type* __buff)
+{
+    typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
+    __destruct_n __d(0);
+    unique_ptr<value_type, __destruct_n&> __h2(__buff, __d);
+    if (__len1 <= __len2)
+    {
+        value_type* __p = __buff;
+        for (_BidirectionalIterator __i = __first; __i != __middle; __d.__incr((value_type*)0), (void) ++__i, (void) ++__p)
+            ::new(__p) value_type(_CUDA_VSTD::move(*__i));
+        __half_inplace_merge(__buff, __p, __middle, __last, __first, __comp);
+    }
+    else
+    {
+        value_type* __p = __buff;
+        for (_BidirectionalIterator __i = __middle; __i != __last; __d.__incr((value_type*)0), (void) ++__i, (void) ++__p)
+            ::new(__p) value_type(_CUDA_VSTD::move(*__i));
+        typedef reverse_iterator<_BidirectionalIterator> _RBi;
+        typedef reverse_iterator<value_type*> _Rv;
+        __half_inplace_merge(_Rv(__p), _Rv(__buff),
+                             _RBi(__middle), _RBi(__first),
+                             _RBi(__last), __invert<_Compare>(__comp));
+    }
+}
+
+template <class _Compare, class _BidirectionalIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__inplace_merge(_BidirectionalIterator __first, _BidirectionalIterator __middle, _BidirectionalIterator __last,
+                _Compare __comp, typename iterator_traits<_BidirectionalIterator>::difference_type __len1,
+                                 typename iterator_traits<_BidirectionalIterator>::difference_type __len2,
+                typename iterator_traits<_BidirectionalIterator>::value_type* __buff, ptrdiff_t __buff_size)
+{
+    typedef typename iterator_traits<_BidirectionalIterator>::difference_type difference_type;
+    while (true)
+    {
+        // if __middle == __last, we're done
+        if (__len2 == 0)
+            return;
+        if (__len1 <= __buff_size || __len2 <= __buff_size)
+            return __buffered_inplace_merge<_Compare>
+                   (__first, __middle, __last, __comp, __len1, __len2, __buff);
+        // shrink [__first, __middle) as much as possible (with no moves), returning if it shrinks to 0
+        for (; true; ++__first, (void) --__len1)
+        {
+            if (__len1 == 0)
+                return;
+            if (__comp(*__middle, *__first))
+                break;
+        }
+        // __first < __middle < __last
+        // *__first > *__middle
+        // partition [__first, __m1) [__m1, __middle) [__middle, __m2) [__m2, __last) such that
+        //     all elements in:
+        //         [__first, __m1)  <= [__middle, __m2)
+        //         [__middle, __m2) <  [__m1, __middle)
+        //         [__m1, __middle) <= [__m2, __last)
+        //     and __m1 or __m2 is in the middle of its range
+        _BidirectionalIterator __m1;  // "median" of [__first, __middle)
+        _BidirectionalIterator __m2;  // "median" of [__middle, __last)
+        difference_type __len11;      // distance(__first, __m1)
+        difference_type __len21;      // distance(__middle, __m2)
+        // binary search smaller range
+        if (__len1 < __len2)
+        {   // __len >= 1, __len2 >= 2
+            __len21 = __len2 / 2;
+            __m2 = __middle;
+            _CUDA_VSTD::advance(__m2, __len21);
+            __m1 = __upper_bound<_Compare>(__first, __middle, *__m2, __comp);
+            __len11 = _CUDA_VSTD::distance(__first, __m1);
+        }
+        else
+        {
+            if (__len1 == 1)
+            {   // __len1 >= __len2 && __len2 > 0, therefore __len2 == 1
+                // It is known *__first > *__middle
+                swap(*__first, *__middle);
+                return;
+            }
+            // __len1 >= 2, __len2 >= 1
+            __len11 = __len1 / 2;
+            __m1 = __first;
+            _CUDA_VSTD::advance(__m1, __len11);
+            __m2 = __lower_bound<_Compare>(__middle, __last, *__m1, __comp);
+            __len21 = _CUDA_VSTD::distance(__middle, __m2);
+        }
+        difference_type __len12 = __len1 - __len11;  // distance(__m1, __middle)
+        difference_type __len22 = __len2 - __len21;  // distance(__m2, __last)
+        // [__first, __m1) [__m1, __middle) [__middle, __m2) [__m2, __last)
+        // swap middle two partitions
+        __middle = _CUDA_VSTD::rotate(__m1, __middle, __m2);
+        // __len12 and __len21 now have swapped meanings
+        // merge smaller range with recurisve call and larger with tail recursion elimination
+        if (__len11 + __len21 < __len12 + __len22)
+        {
+            __inplace_merge<_Compare>(__first, __m1, __middle, __comp, __len11, __len21, __buff, __buff_size);
+//          __inplace_merge<_Compare>(__middle, __m2, __last, __comp, __len12, __len22, __buff, __buff_size);
+            __first = __middle;
+            __middle = __m2;
+            __len1 = __len12;
+            __len2 = __len22;
+        }
+        else
+        {
+            __inplace_merge<_Compare>(__middle, __m2, __last, __comp, __len12, __len22, __buff, __buff_size);
+//          __inplace_merge<_Compare>(__first, __m1, __middle, __comp, __len11, __len21, __buff, __buff_size);
+            __last = __middle;
+            __middle = __m1;
+            __len1 = __len11;
+            __len2 = __len21;
+        }
+    }
+}
+
+template <class _BidirectionalIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+inplace_merge(_BidirectionalIterator __first, _BidirectionalIterator __middle, _BidirectionalIterator __last,
+              _Compare __comp)
+{
+    typedef typename iterator_traits<_BidirectionalIterator>::value_type value_type;
+    typedef typename iterator_traits<_BidirectionalIterator>::difference_type difference_type;
+    difference_type __len1 = _CUDA_VSTD::distance(__first, __middle);
+    difference_type __len2 = _CUDA_VSTD::distance(__middle, __last);
+    difference_type __buf_size = _CUDA_VSTD::min(__len1, __len2);
+    pair<value_type*, ptrdiff_t> __buf = _CUDA_VSTD::get_temporary_buffer<value_type>(__buf_size);
+    unique_ptr<value_type, __return_temporary_buffer> __h(__buf.first);
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return _CUDA_VSTD::__inplace_merge<_Comp_ref>(__first, __middle, __last, __comp, __len1, __len2,
+                                            __buf.first, __buf.second);
+}
+
+template <class _BidirectionalIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+inplace_merge(_BidirectionalIterator __first, _BidirectionalIterator __middle, _BidirectionalIterator __last)
+{
+    _CUDA_VSTD::inplace_merge(__first, __middle, __last,
+                        __less<typename iterator_traits<_BidirectionalIterator>::value_type>());
+}
+
+// stable_sort
+
+template <class _Compare, class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__merge_move_construct(_InputIterator1 __first1, _InputIterator1 __last1,
+        _InputIterator2 __first2, _InputIterator2 __last2,
+        typename iterator_traits<_InputIterator1>::value_type* __result, _Compare __comp)
+{
+    typedef typename iterator_traits<_InputIterator1>::value_type value_type;
+    __destruct_n __d(0);
+    unique_ptr<value_type, __destruct_n&> __h(__result, __d);
+    for (; true; ++__result)
+    {
+        if (__first1 == __last1)
+        {
+            for (; __first2 != __last2; ++__first2, ++__result, (void) __d.__incr((value_type*)0))
+                ::new (__result) value_type(_CUDA_VSTD::move(*__first2));
+            __h.release();
+            return;
+        }
+        if (__first2 == __last2)
+        {
+            for (; __first1 != __last1; ++__first1, ++__result, (void) __d.__incr((value_type*)0))
+                ::new (__result) value_type(_CUDA_VSTD::move(*__first1));
+            __h.release();
+            return;
+        }
+        if (__comp(*__first2, *__first1))
+        {
+            ::new (__result) value_type(_CUDA_VSTD::move(*__first2));
+            __d.__incr((value_type*)0);
+            ++__first2;
+        }
+        else
+        {
+            ::new (__result) value_type(_CUDA_VSTD::move(*__first1));
+            __d.__incr((value_type*)0);
+            ++__first1;
+        }
+    }
+}
+
+template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__merge_move_assign(_InputIterator1 __first1, _InputIterator1 __last1,
+        _InputIterator2 __first2, _InputIterator2 __last2,
+        _OutputIterator __result, _Compare __comp)
+{
+    for (; __first1 != __last1; ++__result)
+    {
+        if (__first2 == __last2)
+        {
+            for (; __first1 != __last1; ++__first1, (void) ++__result)
+                *__result = _CUDA_VSTD::move(*__first1);
+            return;
+        }
+        if (__comp(*__first2, *__first1))
+        {
+            *__result = _CUDA_VSTD::move(*__first2);
+            ++__first2;
+        }
+        else
+        {
+            *__result = _CUDA_VSTD::move(*__first1);
+            ++__first1;
+        }
+    }
+    for (; __first2 != __last2; ++__first2, (void) ++__result)
+        *__result = _CUDA_VSTD::move(*__first2);
+}
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp,
+              typename iterator_traits<_RandomAccessIterator>::difference_type __len,
+              typename iterator_traits<_RandomAccessIterator>::value_type* __buff, ptrdiff_t __buff_size);
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__stable_sort_move(_RandomAccessIterator __first1, _RandomAccessIterator __last1, _Compare __comp,
+                   typename iterator_traits<_RandomAccessIterator>::difference_type __len,
+                   typename iterator_traits<_RandomAccessIterator>::value_type* __first2)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    switch (__len)
+    {
+    case 0:
+        return;
+    case 1:
+        ::new(__first2) value_type(_CUDA_VSTD::move(*__first1));
+        return;
+    case 2:
+        __destruct_n __d(0);
+        unique_ptr<value_type, __destruct_n&> __h2(__first2, __d);
+        if (__comp(*--__last1, *__first1))
+        {
+            ::new(__first2) value_type(_CUDA_VSTD::move(*__last1));
+            __d.__incr((value_type*)0);
+            ++__first2;
+            ::new(__first2) value_type(_CUDA_VSTD::move(*__first1));
+        }
+        else
+        {
+            ::new(__first2) value_type(_CUDA_VSTD::move(*__first1));
+            __d.__incr((value_type*)0);
+            ++__first2;
+            ::new(__first2) value_type(_CUDA_VSTD::move(*__last1));
+        }
+        __h2.release();
+        return;
+    }
+    if (__len <= 8)
+    {
+        __insertion_sort_move<_Compare>(__first1, __last1, __first2, __comp);
+        return;
+    }
+    typename iterator_traits<_RandomAccessIterator>::difference_type __l2 = __len / 2;
+    _RandomAccessIterator __m = __first1 + __l2;
+    __stable_sort<_Compare>(__first1, __m, __comp, __l2, __first2, __l2);
+    __stable_sort<_Compare>(__m, __last1, __comp, __len - __l2, __first2 + __l2, __len - __l2);
+    __merge_move_construct<_Compare>(__first1, __m, __m, __last1, __first2, __comp);
+}
+
+template <class _Tp>
+struct __stable_sort_switch
+{
+    static const unsigned value = 128*is_trivially_copy_assignable<_Tp>::value;
+};
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp,
+              typename iterator_traits<_RandomAccessIterator>::difference_type __len,
+              typename iterator_traits<_RandomAccessIterator>::value_type* __buff, ptrdiff_t __buff_size)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    switch (__len)
+    {
+    case 0:
+    case 1:
+        return;
+    case 2:
+        if (__comp(*--__last, *__first))
+            swap(*__first, *__last);
+        return;
+    }
+    if (__len <= static_cast<difference_type>(__stable_sort_switch<value_type>::value))
+    {
+        __insertion_sort<_Compare>(__first, __last, __comp);
+        return;
+    }
+    typename iterator_traits<_RandomAccessIterator>::difference_type __l2 = __len / 2;
+    _RandomAccessIterator __m = __first + __l2;
+    if (__len <= __buff_size)
+    {
+        __destruct_n __d(0);
+        unique_ptr<value_type, __destruct_n&> __h2(__buff, __d);
+        __stable_sort_move<_Compare>(__first, __m, __comp, __l2, __buff);
+        __d.__set(__l2, (value_type*)0);
+        __stable_sort_move<_Compare>(__m, __last, __comp, __len - __l2, __buff + __l2);
+        __d.__set(__len, (value_type*)0);
+        __merge_move_assign<_Compare>(__buff, __buff + __l2, __buff + __l2, __buff + __len, __first, __comp);
+//         __merge<_Compare>(move_iterator<value_type*>(__buff),
+//                           move_iterator<value_type*>(__buff + __l2),
+//                           move_iterator<_RandomAccessIterator>(__buff + __l2),
+//                           move_iterator<_RandomAccessIterator>(__buff + __len),
+//                           __first, __comp);
+        return;
+    }
+    __stable_sort<_Compare>(__first, __m, __comp, __l2, __buff, __buff_size);
+    __stable_sort<_Compare>(__m, __last, __comp, __len - __l2, __buff, __buff_size);
+    __inplace_merge<_Compare>(__first, __m, __last, __comp, __l2, __len - __l2, __buff, __buff_size);
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    difference_type __len = __last - __first;
+    pair<value_type*, ptrdiff_t> __buf(0, 0);
+    unique_ptr<value_type, __return_temporary_buffer> __h;
+    if (__len > static_cast<difference_type>(__stable_sort_switch<value_type>::value))
+    {
+        __buf = _CUDA_VSTD::get_temporary_buffer<value_type>(__len);
+        __h.reset(__buf.first);
+    }
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __stable_sort<_Comp_ref>(__first, __last, __comp, __len, __buf.first, __buf.second);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::stable_sort(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// is_heap_until
+
+template <class _RandomAccessIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _RandomAccessIterator
+is_heap_until(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename _CUDA_VSTD::iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    difference_type __len = __last - __first;
+    difference_type __p = 0;
+    difference_type __c = 1;
+    _RandomAccessIterator __pp = __first;
+    while (__c < __len)
+    {
+        _RandomAccessIterator __cp = __first + __c;
+        if (__comp(*__pp, *__cp))
+            return __cp;
+        ++__c;
+        ++__cp;
+        if (__c == __len)
+            return __last;
+        if (__comp(*__pp, *__cp))
+            return __cp;
+        ++__p;
+        ++__pp;
+        __c = 2 * __p + 1;
+    }
+    return __last;
+}
+
+template<class _RandomAccessIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_RandomAccessIterator
+is_heap_until(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    return _CUDA_VSTD::is_heap_until(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// is_heap
+
+template <class _RandomAccessIterator, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+is_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    return _CUDA_VSTD::is_heap_until(__first, __last, __comp) == __last;
+}
+
+template<class _RandomAccessIterator>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+is_heap(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    return _CUDA_VSTD::is_heap(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// push_heap
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__sift_up(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp,
+          typename iterator_traits<_RandomAccessIterator>::difference_type __len)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    if (__len > 1)
+    {
+        __len = (__len - 2) / 2;
+        _RandomAccessIterator __ptr = __first + __len;
+        if (__comp(*__ptr, *--__last))
+        {
+            value_type __t(_CUDA_VSTD::move(*__last));
+            do
+            {
+                *__last = _CUDA_VSTD::move(*__ptr);
+                __last = __ptr;
+                if (__len == 0)
+                    break;
+                __len = (__len - 1) / 2;
+                __ptr = __first + __len;
+            } while (__comp(*__ptr, __t));
+            *__last = _CUDA_VSTD::move(__t);
+        }
+    }
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+push_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __sift_up<_Comp_ref>(__first, __last, __comp, __last - __first);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+push_heap(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::push_heap(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// pop_heap
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__sift_down(_RandomAccessIterator __first, _RandomAccessIterator /*__last*/,
+            _Compare __comp,
+            typename iterator_traits<_RandomAccessIterator>::difference_type __len,
+            _RandomAccessIterator __start)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    typedef typename iterator_traits<_RandomAccessIterator>::value_type value_type;
+    // left-child of __start is at 2 * __start + 1
+    // right-child of __start is at 2 * __start + 2
+    difference_type __child = __start - __first;
+
+    if (__len < 2 || (__len - 2) / 2 < __child)
+        return;
+
+    __child = 2 * __child + 1;
+    _RandomAccessIterator __child_i = __first + __child;
+
+    if ((__child + 1) < __len && __comp(*__child_i, *(__child_i + 1))) {
+        // right-child exists and is greater than left-child
+        ++__child_i;
+        ++__child;
+    }
+
+    // check if we are in heap-order
+    if (__comp(*__child_i, *__start))
+        // we are, __start is larger than it's largest child
+        return;
+
+    value_type __top(_CUDA_VSTD::move(*__start));
+    do
+    {
+        // we are not in heap-order, swap the parent with it's largest child
+        *__start = _CUDA_VSTD::move(*__child_i);
+        __start = __child_i;
+
+        if ((__len - 2) / 2 < __child)
+            break;
+
+        // recompute the child based off of the updated parent
+        __child = 2 * __child + 1;
+        __child_i = __first + __child;
+
+        if ((__child + 1) < __len && __comp(*__child_i, *(__child_i + 1))) {
+            // right-child exists and is greater than left-child
+            ++__child_i;
+            ++__child;
+        }
+
+        // check if we are in heap-order
+    } while (!__comp(*__child_i, __top));
+    *__start = _CUDA_VSTD::move(__top);
+}
+
+template <class _Compare, class _RandomAccessIterator>
+inline _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__pop_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp,
+           typename iterator_traits<_RandomAccessIterator>::difference_type __len)
+{
+    if (__len > 1)
+    {
+        swap(*__first, *--__last);
+        __sift_down<_Compare>(__first, __last, __comp, __len - 1, __first);
+    }
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+pop_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __pop_heap<_Comp_ref>(__first, __last, __comp, __last - __first);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+pop_heap(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::pop_heap(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// make_heap
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+__make_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    difference_type __n = __last - __first;
+    if (__n > 1)
+    {
+        // start from the first parent, there is no need to consider children
+        for (difference_type __start = (__n - 2) / 2; __start >= 0; --__start)
+        {
+            __sift_down<_Compare>(__first, __last, __comp, __n, __first + __start);
+        }
+    }
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+make_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __make_heap<_Comp_ref>(__first, __last, __comp);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+make_heap(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::make_heap(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// sort_heap
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__sort_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    for (difference_type __n = __last - __first; __n > 1; --__last, (void) --__n)
+        __pop_heap<_Compare>(__first, __last, __comp, __n);
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort_heap(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __sort_heap<_Comp_ref>(__first, __last, __comp);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sort_heap(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::sort_heap(__first, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// partial_sort
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__partial_sort(_RandomAccessIterator __first, _RandomAccessIterator __middle, _RandomAccessIterator __last,
+             _Compare __comp)
+{
+    __make_heap<_Compare>(__first, __middle, __comp);
+    typename iterator_traits<_RandomAccessIterator>::difference_type __len = __middle - __first;
+    for (_RandomAccessIterator __i = __middle; __i != __last; ++__i)
+    {
+        if (__comp(*__i, *__first))
+        {
+            swap(*__i, *__first);
+            __sift_down<_Compare>(__first, __middle, __comp, __len, __first);
+        }
+    }
+    __sort_heap<_Compare>(__first, __middle, __comp);
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+partial_sort(_RandomAccessIterator __first, _RandomAccessIterator __middle, _RandomAccessIterator __last,
+             _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __partial_sort<_Comp_ref>(__first, __middle, __last, __comp);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+partial_sort(_RandomAccessIterator __first, _RandomAccessIterator __middle, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::partial_sort(__first, __middle, __last,
+                       __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// partial_sort_copy
+
+template <class _Compare, class _InputIterator, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_RandomAccessIterator
+__partial_sort_copy(_InputIterator __first, _InputIterator __last,
+                    _RandomAccessIterator __result_first, _RandomAccessIterator __result_last, _Compare __comp)
+{
+    _RandomAccessIterator __r = __result_first;
+    if (__r != __result_last)
+    {
+        for (; __first != __last && __r != __result_last; ++__first, (void) ++__r)
+            *__r = *__first;
+        __make_heap<_Compare>(__result_first, __r, __comp);
+        typename iterator_traits<_RandomAccessIterator>::difference_type __len = __r - __result_first;
+        for (; __first != __last; ++__first)
+            if (__comp(*__first, *__result_first))
+            {
+                *__result_first = *__first;
+                __sift_down<_Compare>(__result_first, __r, __comp, __len, __result_first);
+            }
+        __sort_heap<_Compare>(__result_first, __r, __comp);
+    }
+    return __r;
+}
+
+template <class _InputIterator, class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_RandomAccessIterator
+partial_sort_copy(_InputIterator __first, _InputIterator __last,
+                  _RandomAccessIterator __result_first, _RandomAccessIterator __result_last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __partial_sort_copy<_Comp_ref>(__first, __last, __result_first, __result_last, __comp);
+}
+
+template <class _InputIterator, class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_RandomAccessIterator
+partial_sort_copy(_InputIterator __first, _InputIterator __last,
+                  _RandomAccessIterator __result_first, _RandomAccessIterator __result_last)
+{
+    return _CUDA_VSTD::partial_sort_copy(__first, __last, __result_first, __result_last,
+                                   __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// nth_element
+
+template <class _Compare, class _RandomAccessIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+void
+__nth_element(_RandomAccessIterator __first, _RandomAccessIterator __nth, _RandomAccessIterator __last, _Compare __comp)
+{
+    // _Compare is known to be a reference type
+    typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
+    const difference_type __limit = 7;
+    while (true)
+    {
+    __restart:
+        if (__nth == __last)
+            return;
+        difference_type __len = __last - __first;
+        switch (__len)
+        {
+        case 0:
+        case 1:
+            return;
+        case 2:
+            if (__comp(*--__last, *__first))
+                swap(*__first, *__last);
+            return;
+        case 3:
+            {
+            _RandomAccessIterator __m = __first;
+            _CUDA_VSTD::__sort3<_Compare>(__first, ++__m, --__last, __comp);
+            return;
+            }
+        }
+        if (__len <= __limit)
+        {
+            __selection_sort<_Compare>(__first, __last, __comp);
+            return;
+        }
+        // __len > __limit >= 3
+        _RandomAccessIterator __m = __first + __len/2;
+        _RandomAccessIterator __lm1 = __last;
+        unsigned __n_swaps = _CUDA_VSTD::__sort3<_Compare>(__first, __m, --__lm1, __comp);
+        // *__m is median
+        // partition [__first, __m) < *__m and *__m <= [__m, __last)
+        // (this inhibits tossing elements equivalent to __m around unnecessarily)
+        _RandomAccessIterator __i = __first;
+        _RandomAccessIterator __j = __lm1;
+        // j points beyond range to be tested, *__lm1 is known to be <= *__m
+        // The search going up is known to be guarded but the search coming down isn't.
+        // Prime the downward search with a guard.
+        if (!__comp(*__i, *__m))  // if *__first == *__m
+        {
+            // *__first == *__m, *__first doesn't go in first part
+            // manually guard downward moving __j against __i
+            while (true)
+            {
+                if (__i == --__j)
+                {
+                    // *__first == *__m, *__m <= all other elements
+                    // Parition instead into [__first, __i) == *__first and *__first < [__i, __last)
+                    ++__i;  // __first + 1
+                    __j = __last;
+                    if (!__comp(*__first, *--__j))  // we need a guard if *__first == *(__last-1)
+                    {
+                        while (true)
+                        {
+                            if (__i == __j)
+                                return;  // [__first, __last) all equivalent elements
+                            if (__comp(*__first, *__i))
+                            {
+                                swap(*__i, *__j);
+                                ++__n_swaps;
+                                ++__i;
+                                break;
+                            }
+                            ++__i;
+                        }
+                    }
+                    // [__first, __i) == *__first and *__first < [__j, __last) and __j == __last - 1
+                    if (__i == __j)
+                        return;
+                    while (true)
+                    {
+                        while (!__comp(*__first, *__i))
+                            ++__i;
+                        while (__comp(*__first, *--__j))
+                            ;
+                        if (__i >= __j)
+                            break;
+                        swap(*__i, *__j);
+                        ++__n_swaps;
+                        ++__i;
+                    }
+                    // [__first, __i) == *__first and *__first < [__i, __last)
+                    // The first part is sorted,
+                    if (__nth < __i)
+                        return;
+                    // __nth_element the secod part
+                    // __nth_element<_Compare>(__i, __nth, __last, __comp);
+                    __first = __i;
+                    goto __restart;
+                }
+                if (__comp(*__j, *__m))
+                {
+                    swap(*__i, *__j);
+                    ++__n_swaps;
+                    break;  // found guard for downward moving __j, now use unguarded partition
+                }
+            }
+        }
+        ++__i;
+        // j points beyond range to be tested, *__lm1 is known to be <= *__m
+        // if not yet partitioned...
+        if (__i < __j)
+        {
+            // known that *(__i - 1) < *__m
+            while (true)
+            {
+                // __m still guards upward moving __i
+                while (__comp(*__i, *__m))
+                    ++__i;
+                // It is now known that a guard exists for downward moving __j
+                while (!__comp(*--__j, *__m))
+                    ;
+                if (__i >= __j)
+                    break;
+                swap(*__i, *__j);
+                ++__n_swaps;
+                // It is known that __m != __j
+                // If __m just moved, follow it
+                if (__m == __i)
+                    __m = __j;
+                ++__i;
+            }
+        }
+        // [__first, __i) < *__m and *__m <= [__i, __last)
+        if (__i != __m && __comp(*__m, *__i))
+        {
+            swap(*__i, *__m);
+            ++__n_swaps;
+        }
+        // [__first, __i) < *__i and *__i <= [__i+1, __last)
+        if (__nth == __i)
+            return;
+        if (__n_swaps == 0)
+        {
+            // We were given a perfectly partitioned sequence.  Coincidence?
+            if (__nth < __i)
+            {
+                // Check for [__first, __i) already sorted
+                __j = __m = __first;
+                while (++__j != __i)
+                {
+                    if (__comp(*__j, *__m))
+                        // not yet sorted, so sort
+                        goto not_sorted;
+                    __m = __j;
+                }
+                // [__first, __i) sorted
+                return;
+            }
+            else
+            {
+                // Check for [__i, __last) already sorted
+                __j = __m = __i;
+                while (++__j != __last)
+                {
+                    if (__comp(*__j, *__m))
+                        // not yet sorted, so sort
+                        goto not_sorted;
+                    __m = __j;
+                }
+                // [__i, __last) sorted
+                return;
+            }
+        }
+not_sorted:
+        // __nth_element on range containing __nth
+        if (__nth < __i)
+        {
+            // __nth_element<_Compare>(__first, __nth, __i, __comp);
+            __last = __i;
+        }
+        else
+        {
+            // __nth_element<_Compare>(__i+1, __nth, __last, __comp);
+            __first = ++__i;
+        }
+    }
+}
+
+template <class _RandomAccessIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+nth_element(_RandomAccessIterator __first, _RandomAccessIterator __nth, _RandomAccessIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    __nth_element<_Comp_ref>(__first, __nth, __last, __comp);
+}
+
+template <class _RandomAccessIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+nth_element(_RandomAccessIterator __first, _RandomAccessIterator __nth, _RandomAccessIterator __last)
+{
+    _CUDA_VSTD::nth_element(__first, __nth, __last, __less<typename iterator_traits<_RandomAccessIterator>::value_type>());
+}
+
+// includes
+
+template <class _Compare, class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+__includes(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2, _InputIterator2 __last2,
+           _Compare __comp)
+{
+    for (; __first2 != __last2; ++__first1)
+    {
+        if (__first1 == __last1 || __comp(*__first2, *__first1))
+            return false;
+        if (!__comp(*__first1, *__first2))
+            ++__first2;
+    }
+    return true;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+includes(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2, _InputIterator2 __last2,
+         _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __includes<_Comp_ref>(__first1, __last1, __first2, __last2, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+includes(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2, _InputIterator2 __last2)
+{
+    return _CUDA_VSTD::includes(__first1, __last1, __first2, __last2,
+                          __less<typename iterator_traits<_InputIterator1>::value_type,
+                                 typename iterator_traits<_InputIterator2>::value_type>());
+}
+
+// set_union
+
+template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_OutputIterator
+__set_union(_InputIterator1 __first1, _InputIterator1 __last1,
+            _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    for (; __first1 != __last1; ++__result)
+    {
+        if (__first2 == __last2)
+            return _CUDA_VSTD::copy(__first1, __last1, __result);
+        if (__comp(*__first2, *__first1))
+        {
+            *__result = *__first2;
+            ++__first2;
+        }
+        else
+        {
+            if (!__comp(*__first1, *__first2))
+                ++__first2;
+            *__result = *__first1;
+            ++__first1;
+        }
+    }
+    return _CUDA_VSTD::copy(__first2, __last2, __result);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+set_union(_InputIterator1 __first1, _InputIterator1 __last1,
+          _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __set_union<_Comp_ref>(__first1, __last1, __first2, __last2, __result, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+set_union(_InputIterator1 __first1, _InputIterator1 __last1,
+          _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result)
+{
+    return _CUDA_VSTD::set_union(__first1, __last1, __first2, __last2, __result,
+                          __less<typename iterator_traits<_InputIterator1>::value_type,
+                                 typename iterator_traits<_InputIterator2>::value_type>());
+}
+
+// set_intersection
+
+template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 _OutputIterator
+__set_intersection(_InputIterator1 __first1, _InputIterator1 __last1,
+                   _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    while (__first1 != __last1 && __first2 != __last2)
+    {
+        if (__comp(*__first1, *__first2))
+            ++__first1;
+        else
+        {
+            if (!__comp(*__first2, *__first1))
+            {
+                *__result = *__first1;
+                ++__result;
+                ++__first1;
+            }
+            ++__first2;
+        }
+    }
+    return __result;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+set_intersection(_InputIterator1 __first1, _InputIterator1 __last1,
+                 _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __set_intersection<_Comp_ref>(__first1, __last1, __first2, __last2, __result, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_OutputIterator
+set_intersection(_InputIterator1 __first1, _InputIterator1 __last1,
+                 _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result)
+{
+    return _CUDA_VSTD::set_intersection(__first1, __last1, __first2, __last2, __result,
+                                  __less<typename iterator_traits<_InputIterator1>::value_type,
+                                         typename iterator_traits<_InputIterator2>::value_type>());
+}
+
+// set_difference
+
+template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_OutputIterator
+__set_difference(_InputIterator1 __first1, _InputIterator1 __last1,
+                 _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    while (__first1 != __last1)
+    {
+        if (__first2 == __last2)
+            return _CUDA_VSTD::copy(__first1, __last1, __result);
+        if (__comp(*__first1, *__first2))
+        {
+            *__result = *__first1;
+            ++__result;
+            ++__first1;
+        }
+        else
+        {
+            if (!__comp(*__first2, *__first1))
+                ++__first1;
+            ++__first2;
+        }
+    }
+    return __result;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+set_difference(_InputIterator1 __first1, _InputIterator1 __last1,
+               _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __set_difference<_Comp_ref>(__first1, __last1, __first2, __last2, __result, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+set_difference(_InputIterator1 __first1, _InputIterator1 __last1,
+               _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result)
+{
+    return _CUDA_VSTD::set_difference(__first1, __last1, __first2, __last2, __result,
+                                __less<typename iterator_traits<_InputIterator1>::value_type,
+                                       typename iterator_traits<_InputIterator2>::value_type>());
+}
+
+// set_symmetric_difference
+
+template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+_OutputIterator
+__set_symmetric_difference(_InputIterator1 __first1, _InputIterator1 __last1,
+                           _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    while (__first1 != __last1)
+    {
+        if (__first2 == __last2)
+            return _CUDA_VSTD::copy(__first1, __last1, __result);
+        if (__comp(*__first1, *__first2))
+        {
+            *__result = *__first1;
+            ++__result;
+            ++__first1;
+        }
+        else
+        {
+            if (__comp(*__first2, *__first1))
+            {
+                *__result = *__first2;
+                ++__result;
+            }
+            else
+                ++__first1;
+            ++__first2;
+        }
+    }
+    return _CUDA_VSTD::copy(__first2, __last2, __result);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+set_symmetric_difference(_InputIterator1 __first1, _InputIterator1 __last1,
+                         _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __set_symmetric_difference<_Comp_ref>(__first1, __last1, __first2, __last2, __result, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+set_symmetric_difference(_InputIterator1 __first1, _InputIterator1 __last1,
+                         _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result)
+{
+    return _CUDA_VSTD::set_symmetric_difference(__first1, __last1, __first2, __last2, __result,
+                                          __less<typename iterator_traits<_InputIterator1>::value_type,
+                                                 typename iterator_traits<_InputIterator2>::value_type>());
+}
+
+// lexicographical_compare
+#endif // __cuda_std__
+
+template <class _Compare, class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+__lexicographical_compare(_InputIterator1 __first1, _InputIterator1 __last1,
+                          _InputIterator2 __first2, _InputIterator2 __last2, _Compare __comp)
+{
+    for (; __first2 != __last2; ++__first1, (void) ++__first2)
+    {
+        if (__first1 == __last1 || __comp(*__first1, *__first2))
+            return true;
+        if (__comp(*__first2, *__first1))
+            return false;
+    }
+    return false;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _Compare>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+lexicographical_compare(_InputIterator1 __first1, _InputIterator1 __last1,
+                        _InputIterator2 __first2, _InputIterator2 __last2, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __lexicographical_compare<_Comp_ref>(__first1, __last1, __first2, __last2, __comp);
+}
+
+template <class _InputIterator1, class _InputIterator2>
+_LIBCUDACXX_NODISCARD_EXT inline
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+bool
+lexicographical_compare(_InputIterator1 __first1, _InputIterator1 __last1,
+                        _InputIterator2 __first2, _InputIterator2 __last2)
+{
+    return _CUDA_VSTD::lexicographical_compare(__first1, __last1, __first2, __last2,
+                                         __less<typename iterator_traits<_InputIterator1>::value_type,
+                                                typename iterator_traits<_InputIterator2>::value_type>());
+}
+
+#ifndef __cuda_std__
+// next_permutation
+
+template <class _Compare, class _BidirectionalIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__next_permutation(_BidirectionalIterator __first, _BidirectionalIterator __last, _Compare __comp)
+{
+    _BidirectionalIterator __i = __last;
+    if (__first == __last || __first == --__i)
+        return false;
+    while (true)
+    {
+        _BidirectionalIterator __ip1 = __i;
+        if (__comp(*--__i, *__ip1))
+        {
+            _BidirectionalIterator __j = __last;
+            while (!__comp(*__i, *--__j))
+                ;
+            swap(*__i, *__j);
+            _CUDA_VSTD::reverse(__ip1, __last);
+            return true;
+        }
+        if (__i == __first)
+        {
+            _CUDA_VSTD::reverse(__first, __last);
+            return false;
+        }
+    }
+}
+
+template <class _BidirectionalIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+next_permutation(_BidirectionalIterator __first, _BidirectionalIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __next_permutation<_Comp_ref>(__first, __last, __comp);
+}
+
+template <class _BidirectionalIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+next_permutation(_BidirectionalIterator __first, _BidirectionalIterator __last)
+{
+    return _CUDA_VSTD::next_permutation(__first, __last,
+                                  __less<typename iterator_traits<_BidirectionalIterator>::value_type>());
+}
+
+// prev_permutation
+
+template <class _Compare, class _BidirectionalIterator>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__prev_permutation(_BidirectionalIterator __first, _BidirectionalIterator __last, _Compare __comp)
+{
+    _BidirectionalIterator __i = __last;
+    if (__first == __last || __first == --__i)
+        return false;
+    while (true)
+    {
+        _BidirectionalIterator __ip1 = __i;
+        if (__comp(*__ip1, *--__i))
+        {
+            _BidirectionalIterator __j = __last;
+            while (!__comp(*--__j, *__i))
+                ;
+            swap(*__i, *__j);
+            _CUDA_VSTD::reverse(__ip1, __last);
+            return true;
+        }
+        if (__i == __first)
+        {
+            _CUDA_VSTD::reverse(__first, __last);
+            return false;
+        }
+    }
+}
+
+template <class _BidirectionalIterator, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+prev_permutation(_BidirectionalIterator __first, _BidirectionalIterator __last, _Compare __comp)
+{
+    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+    return __prev_permutation<_Comp_ref>(__first, __last, __comp);
+}
+
+template <class _BidirectionalIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+prev_permutation(_BidirectionalIterator __first, _BidirectionalIterator __last)
+{
+    return _CUDA_VSTD::prev_permutation(__first, __last,
+                                  __less<typename iterator_traits<_BidirectionalIterator>::value_type>());
+}
+
+#endif
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+_LIBCUDACXX_POP_MACROS
+#endif
+
+#if defined(_LIBCUDACXX_HAS_PARALLEL_ALGORITHMS) && _LIBCUDACXX_STD_VER >= 17
+#   include <__pstl_algorithm>
+#endif
+
+#endif  // _LIBCUDACXX_ALGORITHM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/any b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/any
new file mode 100644
index 000000000000..b0e10bd490fb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/any
@@ -0,0 +1,671 @@
+// -*- C++ -*-
+//===------------------------------ any -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ANY
+#define _LIBCUDACXX_ANY
+
+/*
+   any synopsis
+
+namespace std {
+
+  class bad_any_cast : public bad_cast
+  {
+  public:
+    virtual const char* what() const noexcept;
+  };
+
+  class any
+  {
+  public:
+
+    // 6.3.1 any construct/destruct
+    any() noexcept;
+
+    any(const any& other);
+    any(any&& other) noexcept;
+
+    template <class ValueType>
+      any(ValueType&& value);
+
+    ~any();
+
+    // 6.3.2 any assignments
+    any& operator=(const any& rhs);
+    any& operator=(any&& rhs) noexcept;
+
+    template <class ValueType>
+      any& operator=(ValueType&& rhs);
+
+    // 6.3.3 any modifiers
+    template <class ValueType, class... Args>
+      decay_t<ValueType>& emplace(Args&&... args);
+    template <class ValueType, class U, class... Args>
+      decay_t<ValueType>& emplace(initializer_list<U>, Args&&...);
+    void reset() noexcept;
+    void swap(any& rhs) noexcept;
+
+    // 6.3.4 any observers
+    bool has_value() const noexcept;
+    const type_info& type() const noexcept;
+  };
+
+   // 6.4 Non-member functions
+  void swap(any& x, any& y) noexcept;
+
+  template <class T, class ...Args>
+    any make_any(Args&& ...args);
+  template <class T, class U, class ...Args>
+    any make_any(initializer_list<U>, Args&& ...args);
+
+  template<class ValueType>
+    ValueType any_cast(const any& operand);
+  template<class ValueType>
+    ValueType any_cast(any& operand);
+  template<class ValueType>
+    ValueType any_cast(any&& operand);
+
+  template<class ValueType>
+    const ValueType* any_cast(const any* operand) noexcept;
+  template<class ValueType>
+    ValueType* any_cast(any* operand) noexcept;
+
+} // namespace std
+
+*/
+
+#include <experimental/__config>
+#include <memory>
+#include <new>
+#include <typeinfo>
+#include <type_traits>
+#include <cstdlib>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+namespace std {
+class _LIBCUDACXX_EXCEPTION_ABI _LIBCUDACXX_AVAILABILITY_BAD_ANY_CAST bad_any_cast : public bad_cast
+{
+public:
+    virtual const char* what() const _NOEXCEPT;
+};
+} // namespace std
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_ANY_CAST
+void __throw_bad_any_cast()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw bad_any_cast();
+#else
+    _CUDA_VSTD::abort();
+#endif
+}
+
+// Forward declarations
+class _LIBCUDACXX_TEMPLATE_VIS any;
+
+template <class _ValueType>
+_LIBCUDACXX_INLINE_VISIBILITY
+add_pointer_t<add_const_t<_ValueType>>
+any_cast(any const *) _NOEXCEPT;
+
+template <class _ValueType>
+_LIBCUDACXX_INLINE_VISIBILITY
+add_pointer_t<_ValueType> any_cast(any *) _NOEXCEPT;
+
+namespace __any_imp
+{
+  using _Buffer = aligned_storage_t<3*sizeof(void*), alignment_of<void*>::value>;
+
+  template <class _Tp>
+  using _IsSmallObject = integral_constant<bool
+        , sizeof(_Tp) <= sizeof(_Buffer)
+          && alignment_of<_Buffer>::value
+             % alignment_of<_Tp>::value == 0
+          && is_nothrow_move_constructible<_Tp>::value
+        >;
+
+  enum class _Action {
+    _Destroy,
+    _Copy,
+    _Move,
+    _Get,
+    _TypeInfo
+  };
+
+  template <class _Tp> struct _SmallHandler;
+  template <class _Tp> struct _LargeHandler;
+
+  template <class _Tp>
+  struct  _LIBCUDACXX_TEMPLATE_VIS __unique_typeinfo { static constexpr int __id = 0; };
+  template <class _Tp> constexpr int __unique_typeinfo<_Tp>::__id;
+
+  template <class _Tp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr const void* __get_fallback_typeid() {
+      return &__unique_typeinfo<decay_t<_Tp>>::__id;
+  }
+
+  template <class _Tp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  bool __compare_typeid(type_info const* __id, const void* __fallback_id)
+  {
+#if !defined(_LIBCUDACXX_NO_RTTI)
+      if (__id && *__id == typeid(_Tp))
+          return true;
+#endif
+      if (!__id && __fallback_id == __any_imp::__get_fallback_typeid<_Tp>())
+          return true;
+      return false;
+  }
+
+  template <class _Tp>
+  using _Handler = conditional_t<
+    _IsSmallObject<_Tp>::value, _SmallHandler<_Tp>, _LargeHandler<_Tp>>;
+
+} // namespace __any_imp
+
+class _LIBCUDACXX_TEMPLATE_VIS any
+{
+public:
+  // construct/destruct
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr any() _NOEXCEPT : __h(nullptr) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  any(any const & __other) : __h(nullptr)
+  {
+    if (__other.__h) __other.__call(_Action::_Copy, this);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  any(any && __other) _NOEXCEPT : __h(nullptr)
+  {
+    if (__other.__h) __other.__call(_Action::_Move, this);
+  }
+
+  template <
+      class _ValueType
+    , class _Tp = decay_t<_ValueType>
+    , class = enable_if_t<
+        !is_same<_Tp, any>::value &&
+        !__is_inplace_type<_ValueType>::value &&
+        is_copy_constructible<_Tp>::value>
+    >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  any(_ValueType && __value);
+
+  template <class _ValueType, class ..._Args,
+    class _Tp = decay_t<_ValueType>,
+    class = enable_if_t<
+        is_constructible<_Tp, _Args...>::value &&
+        is_copy_constructible<_Tp>::value
+    >
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit any(in_place_type_t<_ValueType>, _Args&&... __args);
+
+  template <class _ValueType, class _Up, class ..._Args,
+    class _Tp = decay_t<_ValueType>,
+    class = enable_if_t<
+        is_constructible<_Tp, initializer_list<_Up>&, _Args...>::value &&
+        is_copy_constructible<_Tp>::value>
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit any(in_place_type_t<_ValueType>, initializer_list<_Up>, _Args&&... __args);
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  ~any() { this->reset(); }
+
+  // assignments
+  _LIBCUDACXX_INLINE_VISIBILITY
+  any & operator=(any const & __rhs) {
+    any(__rhs).swap(*this);
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  any & operator=(any && __rhs) _NOEXCEPT {
+    any(_CUDA_VSTD::move(__rhs)).swap(*this);
+    return *this;
+  }
+
+  template <
+      class _ValueType
+    , class _Tp = decay_t<_ValueType>
+    , class = enable_if_t<
+          !is_same<_Tp, any>::value
+          && is_copy_constructible<_Tp>::value>
+    >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  any & operator=(_ValueType && __rhs);
+
+  template <class _ValueType, class ..._Args,
+    class _Tp = decay_t<_ValueType>,
+    class = enable_if_t<
+        is_constructible<_Tp, _Args...>::value &&
+        is_copy_constructible<_Tp>::value>
+    >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _Tp& emplace(_Args&&... args);
+
+  template <class _ValueType, class _Up, class ..._Args,
+    class _Tp = decay_t<_ValueType>,
+    class = enable_if_t<
+        is_constructible<_Tp, initializer_list<_Up>&, _Args...>::value &&
+        is_copy_constructible<_Tp>::value>
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _Tp& emplace(initializer_list<_Up>, _Args&&...);
+
+  // 6.3.3 any modifiers
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void reset() _NOEXCEPT { if (__h) this->__call(_Action::_Destroy); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void swap(any & __rhs) _NOEXCEPT;
+
+  // 6.3.4 any observers
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool has_value() const _NOEXCEPT { return __h != nullptr; }
+
+#if !defined(_LIBCUDACXX_NO_RTTI)
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const type_info & type() const _NOEXCEPT {
+    if (__h) {
+        return *static_cast<type_info const *>(this->__call(_Action::_TypeInfo));
+    } else {
+        return typeid(void);
+    }
+  }
+#endif
+
+private:
+    typedef __any_imp::_Action _Action;
+    using _HandleFuncPtr =  void* (*)(_Action, any const *, any *, const type_info *,
+      const void* __fallback_info);
+
+    union _Storage {
+        constexpr _Storage() : __ptr(nullptr) {}
+        void *  __ptr;
+        __any_imp::_Buffer __buf;
+    };
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void * __call(_Action __a, any * __other = nullptr,
+                  type_info const * __info = nullptr,
+                   const void* __fallback_info = nullptr) const
+    {
+        return __h(__a, this, __other, __info, __fallback_info);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void * __call(_Action __a, any * __other = nullptr,
+                  type_info const * __info = nullptr,
+                  const void* __fallback_info = nullptr)
+    {
+        return __h(__a, this, __other, __info, __fallback_info);
+    }
+
+    template <class>
+    friend struct __any_imp::_SmallHandler;
+    template <class>
+    friend struct __any_imp::_LargeHandler;
+
+    template <class _ValueType>
+    friend add_pointer_t<add_const_t<_ValueType>>
+    any_cast(any const *) _NOEXCEPT;
+
+    template <class _ValueType>
+    friend add_pointer_t<_ValueType>
+    any_cast(any *) _NOEXCEPT;
+
+    _HandleFuncPtr __h = nullptr;
+    _Storage __s;
+};
+
+namespace __any_imp
+{
+  template <class _Tp>
+  struct _LIBCUDACXX_TEMPLATE_VIS _SmallHandler
+  {
+     _LIBCUDACXX_INLINE_VISIBILITY
+     static void* __handle(_Action __act, any const * __this, any * __other,
+                           type_info const * __info, const void* __fallback_info)
+     {
+        switch (__act)
+        {
+        case _Action::_Destroy:
+          __destroy(const_cast<any &>(*__this));
+          return nullptr;
+        case _Action::_Copy:
+            __copy(*__this, *__other);
+            return nullptr;
+        case _Action::_Move:
+          __move(const_cast<any &>(*__this), *__other);
+          return nullptr;
+        case _Action::_Get:
+            return __get(const_cast<any &>(*__this), __info, __fallback_info);
+        case _Action::_TypeInfo:
+          return __type_info();
+        }
+    }
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Tp& __create(any & __dest, _Args&&... __args) {
+        _Tp* __ret = ::new (static_cast<void*>(&__dest.__s.__buf)) _Tp(_CUDA_VSTD::forward<_Args>(__args)...);
+        __dest.__h = &_SmallHandler::__handle;
+        return *__ret;
+    }
+
+  private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __destroy(any & __this) {
+        _Tp & __value = *static_cast<_Tp *>(static_cast<void*>(&__this.__s.__buf));
+        __value.~_Tp();
+        __this.__h = nullptr;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __copy(any const & __this, any & __dest) {
+        _SmallHandler::__create(__dest, *static_cast<_Tp const *>(
+            static_cast<void const *>(&__this.__s.__buf)));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __move(any & __this, any & __dest) {
+        _SmallHandler::__create(__dest, _CUDA_VSTD::move(
+            *static_cast<_Tp*>(static_cast<void*>(&__this.__s.__buf))));
+        __destroy(__this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void* __get(any & __this,
+                       type_info const * __info,
+                       const void* __fallback_id)
+    {
+        if (__any_imp::__compare_typeid<_Tp>(__info, __fallback_id))
+            return static_cast<void*>(&__this.__s.__buf);
+        return nullptr;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void* __type_info()
+    {
+#if !defined(_LIBCUDACXX_NO_RTTI)
+        return const_cast<void*>(static_cast<void const *>(&typeid(_Tp)));
+#else
+        return nullptr;
+#endif
+    }
+  };
+
+  template <class _Tp>
+  struct _LIBCUDACXX_TEMPLATE_VIS _LargeHandler
+  {
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void* __handle(_Action __act, any const * __this,
+                          any * __other, type_info const * __info,
+                          void const* __fallback_info)
+    {
+        switch (__act)
+        {
+        case _Action::_Destroy:
+          __destroy(const_cast<any &>(*__this));
+          return nullptr;
+        case _Action::_Copy:
+          __copy(*__this, *__other);
+          return nullptr;
+        case _Action::_Move:
+          __move(const_cast<any &>(*__this), *__other);
+          return nullptr;
+        case _Action::_Get:
+            return __get(const_cast<any &>(*__this), __info, __fallback_info);
+        case _Action::_TypeInfo:
+          return __type_info();
+        }
+    }
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Tp& __create(any & __dest, _Args&&... __args) {
+        typedef allocator<_Tp> _Alloc;
+        typedef __allocator_destructor<_Alloc> _Dp;
+        _Alloc __a;
+        unique_ptr<_Tp, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+        _Tp* __ret = ::new ((void*)__hold.get()) _Tp(_CUDA_VSTD::forward<_Args>(__args)...);
+        __dest.__s.__ptr = __hold.release();
+        __dest.__h = &_LargeHandler::__handle;
+        return *__ret;
+    }
+
+  private:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __destroy(any & __this){
+        delete static_cast<_Tp*>(__this.__s.__ptr);
+        __this.__h = nullptr;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __copy(any const & __this, any & __dest) {
+        _LargeHandler::__create(__dest, *static_cast<_Tp const *>(__this.__s.__ptr));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __move(any & __this, any & __dest) {
+      __dest.__s.__ptr = __this.__s.__ptr;
+      __dest.__h = &_LargeHandler::__handle;
+      __this.__h = nullptr;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void* __get(any & __this, type_info const * __info,
+                       void const* __fallback_info)
+    {
+        if (__any_imp::__compare_typeid<_Tp>(__info, __fallback_info))
+            return static_cast<void*>(__this.__s.__ptr);
+        return nullptr;
+
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void* __type_info()
+    {
+#if !defined(_LIBCUDACXX_NO_RTTI)
+        return const_cast<void*>(static_cast<void const *>(&typeid(_Tp)));
+#else
+        return nullptr;
+#endif
+    }
+  };
+
+} // namespace __any_imp
+
+
+template <class _ValueType, class _Tp, class>
+any::any(_ValueType && __v) : __h(nullptr)
+{
+  __any_imp::_Handler<_Tp>::__create(*this, _CUDA_VSTD::forward<_ValueType>(__v));
+}
+
+template <class _ValueType, class ..._Args, class _Tp, class>
+any::any(in_place_type_t<_ValueType>, _Args&&... __args) {
+  __any_imp::_Handler<_Tp>::__create(*this, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _ValueType, class _Up, class ..._Args, class _Tp, class>
+any::any(in_place_type_t<_ValueType>, initializer_list<_Up> __il, _Args&&... __args) {
+  __any_imp::_Handler<_Tp>::__create(*this, __il, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _ValueType, class, class>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+any & any::operator=(_ValueType && __v)
+{
+  any(_CUDA_VSTD::forward<_ValueType>(__v)).swap(*this);
+  return *this;
+}
+
+template <class _ValueType, class ..._Args, class _Tp, class>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp& any::emplace(_Args&&... __args) {
+  reset();
+  return __any_imp::_Handler<_Tp>::__create(*this, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _ValueType, class _Up, class ..._Args, class _Tp, class>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp& any::emplace(initializer_list<_Up> __il, _Args&&... __args) {
+  reset();
+  return __any_imp::_Handler<_Tp>::__create(*this, __il, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void any::swap(any & __rhs) _NOEXCEPT
+{
+    if (this == &__rhs)
+      return;
+    if (__h && __rhs.__h) {
+        any __tmp;
+        __rhs.__call(_Action::_Move, &__tmp);
+        this->__call(_Action::_Move, &__rhs);
+        __tmp.__call(_Action::_Move, this);
+    }
+    else if (__h) {
+        this->__call(_Action::_Move, &__rhs);
+    }
+    else if (__rhs.__h) {
+        __rhs.__call(_Action::_Move, this);
+    }
+}
+
+// 6.4 Non-member functions
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void swap(any & __lhs, any & __rhs) _NOEXCEPT
+{
+    __lhs.swap(__rhs);
+}
+
+template <class _Tp, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+any make_any(_Args&&... __args) {
+    return any(in_place_type<_Tp>, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _Tp, class _Up, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+any make_any(initializer_list<_Up> __il, _Args&&... __args) {
+    return any(in_place_type<_Tp>, __il, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _ValueType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_ANY_CAST
+_ValueType any_cast(any const & __v)
+{
+    using _RawValueType = __uncvref_t<_ValueType>;
+    static_assert(is_constructible<_ValueType, _RawValueType const &>::value,
+                  "ValueType is required to be a const lvalue reference "
+                  "or a CopyConstructible type");
+    auto __tmp = _CUDA_VSTD::any_cast<add_const_t<_RawValueType>>(&__v);
+    if (__tmp == nullptr)
+        __throw_bad_any_cast();
+    return static_cast<_ValueType>(*__tmp);
+}
+
+template <class _ValueType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_ANY_CAST
+_ValueType any_cast(any & __v)
+{
+    using _RawValueType = __uncvref_t<_ValueType>;
+    static_assert(is_constructible<_ValueType, _RawValueType &>::value,
+                  "ValueType is required to be an lvalue reference "
+                  "or a CopyConstructible type");
+    auto __tmp = _CUDA_VSTD::any_cast<_RawValueType>(&__v);
+    if (__tmp == nullptr)
+        __throw_bad_any_cast();
+    return static_cast<_ValueType>(*__tmp);
+}
+
+template <class _ValueType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_ANY_CAST
+_ValueType any_cast(any && __v)
+{
+    using _RawValueType = __uncvref_t<_ValueType>;
+    static_assert(is_constructible<_ValueType, _RawValueType>::value,
+                  "ValueType is required to be an rvalue reference "
+                  "or a CopyConstructible type");
+    auto __tmp = _CUDA_VSTD::any_cast<_RawValueType>(&__v);
+    if (__tmp == nullptr)
+        __throw_bad_any_cast();
+    return static_cast<_ValueType>(_CUDA_VSTD::move(*__tmp));
+}
+
+template <class _ValueType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+add_pointer_t<add_const_t<_ValueType>>
+any_cast(any const * __any) _NOEXCEPT
+{
+    static_assert(!is_reference<_ValueType>::value,
+                  "_ValueType may not be a reference.");
+    return _CUDA_VSTD::any_cast<_ValueType>(const_cast<any *>(__any));
+}
+
+template <class _RetType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_RetType __pointer_or_func_cast(void* __p, /*IsFunction*/false_type) noexcept {
+  return static_cast<_RetType>(__p);
+}
+
+template <class _RetType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_RetType __pointer_or_func_cast(void*, /*IsFunction*/true_type) noexcept {
+  return nullptr;
+}
+
+template <class _ValueType>
+add_pointer_t<_ValueType>
+any_cast(any * __any) _NOEXCEPT
+{
+    using __any_imp::_Action;
+    static_assert(!is_reference<_ValueType>::value,
+                  "_ValueType may not be a reference.");
+    typedef typename add_pointer<_ValueType>::type _ReturnType;
+    if (__any && __any->__h) {
+      void *__p = __any->__call(_Action::_Get, nullptr,
+#if !defined(_LIBCUDACXX_NO_RTTI)
+                          &typeid(_ValueType),
+#else
+                          nullptr,
+#endif
+                          __any_imp::__get_fallback_typeid<_ValueType>());
+        return _CUDA_VSTD::__pointer_or_func_cast<_ReturnType>(
+            __p, is_function<_ValueType>{});
+    }
+    return nullptr;
+}
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // _LIBCUDACXX_ANY
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/array b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/array
new file mode 100644
index 000000000000..7ee5e0926a73
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/array
@@ -0,0 +1,486 @@
+// -*- C++ -*-
+//===---------------------------- array -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ARRAY
+#define _LIBCUDACXX_ARRAY
+
+/*
+    array synopsis
+
+namespace std
+{
+template <class T, size_t N >
+struct array
+{
+    // types:
+    typedef T & reference;
+    typedef const T & const_reference;
+    typedef implementation defined iterator;
+    typedef implementation defined const_iterator;
+    typedef size_t size_type;
+    typedef ptrdiff_t difference_type;
+    typedef T value_type;
+    typedef T* pointer;
+    typedef const T* const_pointer;
+    typedef std::reverse_iterator<iterator> reverse_iterator;
+    typedef std::reverse_iterator<const_iterator> const_reverse_iterator;
+
+    // No explicit construct/copy/destroy for aggregate type
+    void fill(const T& u);
+    void swap(array& a) noexcept(is_nothrow_swappable_v<T>);
+
+    // iterators:
+    iterator begin() noexcept;
+    const_iterator begin() const noexcept;
+    iterator end() noexcept;
+    const_iterator end() const noexcept;
+
+    reverse_iterator rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+    reverse_iterator rend() noexcept;
+    const_reverse_iterator rend() const noexcept;
+
+    const_iterator cbegin() const noexcept;
+    const_iterator cend() const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend() const noexcept;
+
+    // capacity:
+    constexpr size_type size() const noexcept;
+    constexpr size_type max_size() const noexcept;
+    constexpr bool empty() const noexcept;
+
+    // element access:
+    reference operator[](size_type n);
+    const_reference operator[](size_type n) const; // constexpr in C++14
+    const_reference at(size_type n) const; // constexpr in C++14
+    reference at(size_type n);
+
+    reference front();
+    const_reference front() const; // constexpr in C++14
+    reference back();
+    const_reference back() const; // constexpr in C++14
+
+    T* data() noexcept;
+    const T* data() const noexcept;
+};
+
+  template <class T, class... U>
+    array(T, U...) -> array<T, 1 + sizeof...(U)>;
+
+template <class T, size_t N>
+  bool operator==(const array<T,N>& x, const array<T,N>& y);
+template <class T, size_t N>
+  bool operator!=(const array<T,N>& x, const array<T,N>& y);
+template <class T, size_t N>
+  bool operator<(const array<T,N>& x, const array<T,N>& y);
+template <class T, size_t N>
+  bool operator>(const array<T,N>& x, const array<T,N>& y);
+template <class T, size_t N>
+  bool operator<=(const array<T,N>& x, const array<T,N>& y);
+template <class T, size_t N>
+  bool operator>=(const array<T,N>& x, const array<T,N>& y);
+
+template <class T, size_t N >
+  void swap(array<T,N>& x, array<T,N>& y) noexcept(noexcept(x.swap(y))); // C++17
+
+template <class T> struct tuple_size;
+template <size_t I, class T> struct tuple_element;
+template <class T, size_t N> struct tuple_size<array<T, N>>;
+template <size_t I, class T, size_t N> struct tuple_element<I, array<T, N>>;
+template <size_t I, class T, size_t N> T& get(array<T, N>&) noexcept; // constexpr in C++14
+template <size_t I, class T, size_t N> const T& get(const array<T, N>&) noexcept; // constexpr in C++14
+template <size_t I, class T, size_t N> T&& get(array<T, N>&&) noexcept; // constexpr in C++14
+template <size_t I, class T, size_t N> const T&& get(const array<T, N>&&) noexcept; // constexpr in C++14
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__tuple>
+#include <type_traits>
+#include <utility>
+#include <iterator>
+#include <algorithm>
+#include <stdexcept>
+#include <cstdlib> // for _LIBCUDACXX_UNREACHABLE
+#include <version>
+#include <__debug>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+
+template <class _Tp, size_t _Size>
+struct _LIBCUDACXX_TEMPLATE_VIS array
+{
+    // types:
+    typedef array __self;
+    typedef _Tp                                          value_type;
+    typedef value_type&                                  reference;
+    typedef const value_type&                            const_reference;
+    typedef value_type*                                  iterator;
+    typedef const value_type*                            const_iterator;
+    typedef value_type*                                  pointer;
+    typedef const value_type*                            const_pointer;
+    typedef size_t                                       size_type;
+    typedef ptrdiff_t                                    difference_type;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator> const_reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>       reverse_iterator;
+
+    _Tp __elems_[_Size];
+
+    // No explicit construct/copy/destroy for aggregate type
+    _LIBCUDACXX_INLINE_VISIBILITY void fill(const value_type& __u) {
+      _CUDA_VSTD::fill_n(__elems_, _Size, __u);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(array& __a) _NOEXCEPT_(__is_nothrow_swappable<_Tp>::value) {
+      _CUDA_VSTD::swap_ranges(__elems_, __elems_ + _Size, __a.__elems_);
+    }
+
+    // iterators:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    iterator begin() _NOEXCEPT {return iterator(data());}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_iterator begin() const _NOEXCEPT {return const_iterator(data());}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    iterator end() _NOEXCEPT {return iterator(data() + _Size);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_iterator end() const _NOEXCEPT {return const_iterator(data() + _Size);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator rbegin() _NOEXCEPT {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_reverse_iterator rbegin() const _NOEXCEPT {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator rend() _NOEXCEPT {return reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_reverse_iterator rend() const _NOEXCEPT {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_iterator cbegin() const _NOEXCEPT {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_iterator cend() const _NOEXCEPT {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_reverse_iterator crbegin() const _NOEXCEPT {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const_reverse_iterator crend() const _NOEXCEPT {return rend();}
+
+    // capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR size_type size() const _NOEXCEPT {return _Size;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR size_type max_size() const _NOEXCEPT {return _Size;}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR bool empty() const _NOEXCEPT {return false; }
+
+    // element access:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reference operator[](size_type __n)             _NOEXCEPT {return __elems_[__n];}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const_reference operator[](size_type __n) const _NOEXCEPT {return __elems_[__n];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14       reference at(size_type __n);
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const_reference at(size_type __n) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 reference front()             _NOEXCEPT {return __elems_[0];}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const_reference front() const _NOEXCEPT {return __elems_[0];}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 reference back()              _NOEXCEPT {return __elems_[_Size - 1];}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const_reference back() const  _NOEXCEPT {return __elems_[_Size - 1];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    value_type* data() _NOEXCEPT {return __elems_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    const value_type* data() const _NOEXCEPT {return __elems_;}
+};
+
+
+template <class _Tp, size_t _Size>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+typename array<_Tp, _Size>::reference
+array<_Tp, _Size>::at(size_type __n)
+{
+    if (__n >= _Size)
+        __throw_out_of_range("array::at");
+
+    return __elems_[__n];
+}
+
+template <class _Tp, size_t _Size>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename array<_Tp, _Size>::const_reference
+array<_Tp, _Size>::at(size_type __n) const
+{
+    if (__n >= _Size)
+        __throw_out_of_range("array::at");
+    return __elems_[__n];
+}
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS array<_Tp, 0>
+{
+    // types:
+    typedef array __self;
+    typedef _Tp                                   value_type;
+    typedef value_type&                           reference;
+    typedef const value_type&                     const_reference;
+    typedef value_type*                           iterator;
+    typedef const value_type*                     const_iterator;
+    typedef value_type*                           pointer;
+    typedef const value_type*                     const_pointer;
+    typedef size_t                                size_type;
+    typedef ptrdiff_t                             difference_type;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>      const_reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>            reverse_iterator;
+
+    typedef typename conditional<is_const<_Tp>::value, const char,
+                                char>::type _CharType;
+
+    struct  _ArrayInStructT { _Tp __data_[1]; };
+    _ALIGNAS_TYPE(_ArrayInStructT) _CharType __elems_[sizeof(_ArrayInStructT)];
+
+    // No explicit construct/copy/destroy for aggregate type
+    _LIBCUDACXX_INLINE_VISIBILITY void fill(const value_type&) {
+      static_assert(!is_const<_Tp>::value,
+                    "cannot fill zero-sized array of type 'const T'");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(array&) _NOEXCEPT {
+      static_assert(!is_const<_Tp>::value,
+                    "cannot swap zero-sized array of type 'const T'");
+    }
+
+    // iterators:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator begin() _NOEXCEPT {return iterator(data());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return const_iterator(data());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator end() _NOEXCEPT {return iterator(data());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT {return const_iterator(data());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator rbegin() _NOEXCEPT {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator rend() _NOEXCEPT {return reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend() const _NOEXCEPT {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend() const _NOEXCEPT {return rend();}
+
+    // capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR size_type size() const _NOEXCEPT {return 0; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR size_type max_size() const _NOEXCEPT {return 0;}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR bool empty() const _NOEXCEPT {return true;}
+
+    // element access:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator[](size_type) _NOEXCEPT {
+      _LIBCUDACXX_ASSERT(false, "cannot call array<T, 0>::operator[] on a zero-sized array");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const_reference operator[](size_type) const _NOEXCEPT {
+      _LIBCUDACXX_ASSERT(false, "cannot call array<T, 0>::operator[] on a zero-sized array");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference at(size_type) {
+      __throw_out_of_range("array<T, 0>::at");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference at(size_type) const {
+      __throw_out_of_range("array<T, 0>::at");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference front() _NOEXCEPT {
+      _LIBCUDACXX_ASSERT(false, "cannot call array<T, 0>::front() on a zero-sized array");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference front() const _NOEXCEPT {
+      _LIBCUDACXX_ASSERT(false, "cannot call array<T, 0>::front() on a zero-sized array");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference back() _NOEXCEPT {
+      _LIBCUDACXX_ASSERT(false, "cannot call array<T, 0>::back() on a zero-sized array");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference back() const _NOEXCEPT {
+      _LIBCUDACXX_ASSERT(false, "cannot call array<T, 0>::back() on a zero-sized array");
+      _LIBCUDACXX_UNREACHABLE();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type* data() _NOEXCEPT {return reinterpret_cast<value_type*>(__elems_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type* data() const _NOEXCEPT {return reinterpret_cast<const value_type*>(__elems_);}
+};
+
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _Tp, class... _Args,
+         class = typename enable_if<(is_same_v<_Tp, _Args> && ...), void>::type
+         >
+array(_Tp, _Args...)
+  -> array<_Tp, 1 + sizeof...(_Args)>;
+#endif
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+operator==(const array<_Tp, _Size>& __x, const array<_Tp, _Size>& __y)
+{
+    return _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+operator!=(const array<_Tp, _Size>& __x, const array<_Tp, _Size>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+operator<(const array<_Tp, _Size>& __x, const array<_Tp, _Size>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(),
+                                          __y.begin(), __y.end());
+}
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+operator>(const array<_Tp, _Size>& __x, const array<_Tp, _Size>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+operator<=(const array<_Tp, _Size>& __x, const array<_Tp, _Size>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17 bool
+operator>=(const array<_Tp, _Size>& __x, const array<_Tp, _Size>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    _Size == 0 ||
+    __is_swappable<_Tp>::value,
+    void
+>::type
+swap(array<_Tp, _Size>& __x, array<_Tp, _Size>& __y)
+                                  _NOEXCEPT_(noexcept(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Tp, size_t _Size>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<array<_Tp, _Size> >
+    : public integral_constant<size_t, _Size> {};
+
+template <size_t _Ip, class _Tp, size_t _Size>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, array<_Tp, _Size> >
+{
+    static_assert(_Ip < _Size, "Index out of bounds in std::tuple_element<> (std::array)");
+    typedef _Tp type;
+};
+
+template <size_t _Ip, class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp&
+get(array<_Tp, _Size>& __a) _NOEXCEPT
+{
+    static_assert(_Ip < _Size, "Index out of bounds in std::get<> (std::array)");
+    return __a.__elems_[_Ip];
+}
+
+template <size_t _Ip, class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&
+get(const array<_Tp, _Size>& __a) _NOEXCEPT
+{
+    static_assert(_Ip < _Size, "Index out of bounds in std::get<> (const std::array)");
+    return __a.__elems_[_Ip];
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <size_t _Ip, class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp&&
+get(array<_Tp, _Size>&& __a) _NOEXCEPT
+{
+    static_assert(_Ip < _Size, "Index out of bounds in std::get<> (std::array &&)");
+    return _CUDA_VSTD::move(__a.__elems_[_Ip]);
+}
+
+template <size_t _Ip, class _Tp, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Tp&&
+get(const array<_Tp, _Size>&& __a) _NOEXCEPT
+{
+    static_assert(_Ip < _Size, "Index out of bounds in std::get<> (const std::array &&)");
+    return _CUDA_VSTD::move(__a.__elems_[_Ip]);
+}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_ARRAY
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/atomic b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/atomic
new file mode 100644
index 000000000000..b97738ffa635
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/atomic
@@ -0,0 +1,2763 @@
+// -*- C++ -*-
+//===--------------------------- atomic -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ATOMIC
+#define _LIBCUDACXX_ATOMIC
+
+/*
+    atomic synopsis
+
+namespace std
+{
+
+// feature test macro
+
+#define __cpp_lib_atomic_is_always_lock_free // as specified by SG10
+
+ // order and consistency
+
+ enum memory_order: unspecified // enum class in C++20
+ {
+    relaxed,
+    consume, // load-consume
+    acquire, // load-acquire
+    release, // store-release
+    acq_rel, // store-release load-acquire
+    seq_cst // store-release load-acquire
+ };
+
+ inline constexpr auto memory_order_relaxed = memory_order::relaxed;
+ inline constexpr auto memory_order_consume = memory_order::consume;
+ inline constexpr auto memory_order_acquire = memory_order::acquire;
+ inline constexpr auto memory_order_release = memory_order::release;
+ inline constexpr auto memory_order_acq_rel = memory_order::acq_rel;
+ inline constexpr auto memory_order_seq_cst = memory_order::seq_cst;
+
+template <class T> T kill_dependency(T y) noexcept;
+
+// lock-free property
+
+#define ATOMIC_BOOL_LOCK_FREE unspecified
+#define ATOMIC_CHAR_LOCK_FREE unspecified
+#define ATOMIC_CHAR16_T_LOCK_FREE unspecified
+#define ATOMIC_CHAR32_T_LOCK_FREE unspecified
+#define ATOMIC_WCHAR_T_LOCK_FREE unspecified
+#define ATOMIC_SHORT_LOCK_FREE unspecified
+#define ATOMIC_INT_LOCK_FREE unspecified
+#define ATOMIC_LONG_LOCK_FREE unspecified
+#define ATOMIC_LLONG_LOCK_FREE unspecified
+#define ATOMIC_POINTER_LOCK_FREE unspecified
+
+// flag type and operations
+
+typedef struct atomic_flag
+{
+    bool test_and_set(memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool test_and_set(memory_order m = memory_order_seq_cst) noexcept;
+    void clear(memory_order m = memory_order_seq_cst) volatile noexcept;
+    void clear(memory_order m = memory_order_seq_cst) noexcept;
+    atomic_flag()  noexcept = default;
+    atomic_flag(const atomic_flag&) = delete;
+    atomic_flag& operator=(const atomic_flag&) = delete;
+    atomic_flag& operator=(const atomic_flag&) volatile = delete;
+} atomic_flag;
+
+bool
+    atomic_flag_test_and_set(volatile atomic_flag* obj) noexcept;
+
+bool
+    atomic_flag_test_and_set(atomic_flag* obj) noexcept;
+
+bool
+    atomic_flag_test_and_set_explicit(volatile atomic_flag* obj,
+                                      memory_order m) noexcept;
+
+bool
+    atomic_flag_test_and_set_explicit(atomic_flag* obj, memory_order m) noexcept;
+
+void
+    atomic_flag_clear(volatile atomic_flag* obj) noexcept;
+
+void
+    atomic_flag_clear(atomic_flag* obj) noexcept;
+
+void
+    atomic_flag_clear_explicit(volatile atomic_flag* obj, memory_order m) noexcept;
+
+void
+    atomic_flag_clear_explicit(atomic_flag* obj, memory_order m) noexcept;
+
+#define ATOMIC_FLAG_INIT see below
+#define ATOMIC_VAR_INIT(value) see below
+
+template <class T>
+struct atomic
+{
+    static constexpr bool is_always_lock_free;
+    bool is_lock_free() const volatile noexcept;
+    bool is_lock_free() const noexcept;
+    void store(T desr, memory_order m = memory_order_seq_cst) volatile noexcept;
+    void store(T desr, memory_order m = memory_order_seq_cst) noexcept;
+    T load(memory_order m = memory_order_seq_cst) const volatile noexcept;
+    T load(memory_order m = memory_order_seq_cst) const noexcept;
+    operator T() const volatile noexcept;
+    operator T() const noexcept;
+    T exchange(T desr, memory_order m = memory_order_seq_cst) volatile noexcept;
+    T exchange(T desr, memory_order m = memory_order_seq_cst) noexcept;
+    bool compare_exchange_weak(T& expc, T desr,
+                               memory_order s, memory_order f) volatile noexcept;
+    bool compare_exchange_weak(T& expc, T desr, memory_order s, memory_order f) noexcept;
+    bool compare_exchange_strong(T& expc, T desr,
+                                 memory_order s, memory_order f) volatile noexcept;
+    bool compare_exchange_strong(T& expc, T desr,
+                                 memory_order s, memory_order f) noexcept;
+    bool compare_exchange_weak(T& expc, T desr,
+                               memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool compare_exchange_weak(T& expc, T desr,
+                               memory_order m = memory_order_seq_cst) noexcept;
+    bool compare_exchange_strong(T& expc, T desr,
+                                memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool compare_exchange_strong(T& expc, T desr,
+                                 memory_order m = memory_order_seq_cst) noexcept;
+
+    atomic() noexcept = default;
+    constexpr atomic(T desr) noexcept;
+    atomic(const atomic&) = delete;
+    atomic& operator=(const atomic&) = delete;
+    atomic& operator=(const atomic&) volatile = delete;
+    T operator=(T) volatile noexcept;
+    T operator=(T) noexcept;
+};
+
+template <>
+struct atomic<integral>
+{
+    static constexpr bool is_always_lock_free;
+    bool is_lock_free() const volatile noexcept;
+    bool is_lock_free() const noexcept;
+    void store(integral desr, memory_order m = memory_order_seq_cst) volatile noexcept;
+    void store(integral desr, memory_order m = memory_order_seq_cst) noexcept;
+    integral load(memory_order m = memory_order_seq_cst) const volatile noexcept;
+    integral load(memory_order m = memory_order_seq_cst) const noexcept;
+    operator integral() const volatile noexcept;
+    operator integral() const noexcept;
+    integral exchange(integral desr,
+                      memory_order m = memory_order_seq_cst) volatile noexcept;
+    integral exchange(integral desr, memory_order m = memory_order_seq_cst) noexcept;
+    bool compare_exchange_weak(integral& expc, integral desr,
+                               memory_order s, memory_order f) volatile noexcept;
+    bool compare_exchange_weak(integral& expc, integral desr,
+                               memory_order s, memory_order f) noexcept;
+    bool compare_exchange_strong(integral& expc, integral desr,
+                                 memory_order s, memory_order f) volatile noexcept;
+    bool compare_exchange_strong(integral& expc, integral desr,
+                                 memory_order s, memory_order f) noexcept;
+    bool compare_exchange_weak(integral& expc, integral desr,
+                               memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool compare_exchange_weak(integral& expc, integral desr,
+                               memory_order m = memory_order_seq_cst) noexcept;
+    bool compare_exchange_strong(integral& expc, integral desr,
+                                memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool compare_exchange_strong(integral& expc, integral desr,
+                                 memory_order m = memory_order_seq_cst) noexcept;
+
+    integral
+        fetch_add(integral op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    integral fetch_add(integral op, memory_order m = memory_order_seq_cst) noexcept;
+    integral
+        fetch_sub(integral op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    integral fetch_sub(integral op, memory_order m = memory_order_seq_cst) noexcept;
+    integral
+        fetch_and(integral op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    integral fetch_and(integral op, memory_order m = memory_order_seq_cst) noexcept;
+    integral
+        fetch_or(integral op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    integral fetch_or(integral op, memory_order m = memory_order_seq_cst) noexcept;
+    integral
+        fetch_xor(integral op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    integral fetch_xor(integral op, memory_order m = memory_order_seq_cst) noexcept;
+
+    atomic() noexcept = default;
+    constexpr atomic(integral desr) noexcept;
+    atomic(const atomic&) = delete;
+    atomic& operator=(const atomic&) = delete;
+    atomic& operator=(const atomic&) volatile = delete;
+    integral operator=(integral desr) volatile noexcept;
+    integral operator=(integral desr) noexcept;
+
+    integral operator++(int) volatile noexcept;
+    integral operator++(int) noexcept;
+    integral operator--(int) volatile noexcept;
+    integral operator--(int) noexcept;
+    integral operator++() volatile noexcept;
+    integral operator++() noexcept;
+    integral operator--() volatile noexcept;
+    integral operator--() noexcept;
+    integral operator+=(integral op) volatile noexcept;
+    integral operator+=(integral op) noexcept;
+    integral operator-=(integral op) volatile noexcept;
+    integral operator-=(integral op) noexcept;
+    integral operator&=(integral op) volatile noexcept;
+    integral operator&=(integral op) noexcept;
+    integral operator|=(integral op) volatile noexcept;
+    integral operator|=(integral op) noexcept;
+    integral operator^=(integral op) volatile noexcept;
+    integral operator^=(integral op) noexcept;
+};
+
+template <class T>
+struct atomic<T*>
+{
+    static constexpr bool is_always_lock_free;
+    bool is_lock_free() const volatile noexcept;
+    bool is_lock_free() const noexcept;
+    void store(T* desr, memory_order m = memory_order_seq_cst) volatile noexcept;
+    void store(T* desr, memory_order m = memory_order_seq_cst) noexcept;
+    T* load(memory_order m = memory_order_seq_cst) const volatile noexcept;
+    T* load(memory_order m = memory_order_seq_cst) const noexcept;
+    operator T*() const volatile noexcept;
+    operator T*() const noexcept;
+    T* exchange(T* desr, memory_order m = memory_order_seq_cst) volatile noexcept;
+    T* exchange(T* desr, memory_order m = memory_order_seq_cst) noexcept;
+    bool compare_exchange_weak(T*& expc, T* desr,
+                               memory_order s, memory_order f) volatile noexcept;
+    bool compare_exchange_weak(T*& expc, T* desr,
+                               memory_order s, memory_order f) noexcept;
+    bool compare_exchange_strong(T*& expc, T* desr,
+                                 memory_order s, memory_order f) volatile noexcept;
+    bool compare_exchange_strong(T*& expc, T* desr,
+                                 memory_order s, memory_order f) noexcept;
+    bool compare_exchange_weak(T*& expc, T* desr,
+                               memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool compare_exchange_weak(T*& expc, T* desr,
+                               memory_order m = memory_order_seq_cst) noexcept;
+    bool compare_exchange_strong(T*& expc, T* desr,
+                                memory_order m = memory_order_seq_cst) volatile noexcept;
+    bool compare_exchange_strong(T*& expc, T* desr,
+                                 memory_order m = memory_order_seq_cst) noexcept;
+    T* fetch_add(ptrdiff_t op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    T* fetch_add(ptrdiff_t op, memory_order m = memory_order_seq_cst) noexcept;
+    T* fetch_sub(ptrdiff_t op, memory_order m = memory_order_seq_cst) volatile noexcept;
+    T* fetch_sub(ptrdiff_t op, memory_order m = memory_order_seq_cst) noexcept;
+
+    atomic() noexcept = default;
+    constexpr atomic(T* desr) noexcept;
+    atomic(const atomic&) = delete;
+    atomic& operator=(const atomic&) = delete;
+    atomic& operator=(const atomic&) volatile = delete;
+
+    T* operator=(T*) volatile noexcept;
+    T* operator=(T*) noexcept;
+    T* operator++(int) volatile noexcept;
+    T* operator++(int) noexcept;
+    T* operator--(int) volatile noexcept;
+    T* operator--(int) noexcept;
+    T* operator++() volatile noexcept;
+    T* operator++() noexcept;
+    T* operator--() volatile noexcept;
+    T* operator--() noexcept;
+    T* operator+=(ptrdiff_t op) volatile noexcept;
+    T* operator+=(ptrdiff_t op) noexcept;
+    T* operator-=(ptrdiff_t op) volatile noexcept;
+    T* operator-=(ptrdiff_t op) noexcept;
+};
+
+
+template <class T>
+    bool
+    atomic_is_lock_free(const volatile atomic<T>* obj) noexcept;
+
+template <class T>
+    bool
+    atomic_is_lock_free(const atomic<T>* obj) noexcept;
+
+template <class T>
+    void
+    atomic_init(volatile atomic<T>* obj, T desr) noexcept;
+
+template <class T>
+    void
+    atomic_init(atomic<T>* obj, T desr) noexcept;
+
+template <class T>
+    void
+    atomic_store(volatile atomic<T>* obj, T desr) noexcept;
+
+template <class T>
+    void
+    atomic_store(atomic<T>* obj, T desr) noexcept;
+
+template <class T>
+    void
+    atomic_store_explicit(volatile atomic<T>* obj, T desr, memory_order m) noexcept;
+
+template <class T>
+    void
+    atomic_store_explicit(atomic<T>* obj, T desr, memory_order m) noexcept;
+
+template <class T>
+    T
+    atomic_load(const volatile atomic<T>* obj) noexcept;
+
+template <class T>
+    T
+    atomic_load(const atomic<T>* obj) noexcept;
+
+template <class T>
+    T
+    atomic_load_explicit(const volatile atomic<T>* obj, memory_order m) noexcept;
+
+template <class T>
+    T
+    atomic_load_explicit(const atomic<T>* obj, memory_order m) noexcept;
+
+template <class T>
+    T
+    atomic_exchange(volatile atomic<T>* obj, T desr) noexcept;
+
+template <class T>
+    T
+    atomic_exchange(atomic<T>* obj, T desr) noexcept;
+
+template <class T>
+    T
+    atomic_exchange_explicit(volatile atomic<T>* obj, T desr, memory_order m) noexcept;
+
+template <class T>
+    T
+    atomic_exchange_explicit(atomic<T>* obj, T desr, memory_order m) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_weak(volatile atomic<T>* obj, T* expc, T desr) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_weak(atomic<T>* obj, T* expc, T desr) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_strong(volatile atomic<T>* obj, T* expc, T desr) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_strong(atomic<T>* obj, T* expc, T desr) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_weak_explicit(volatile atomic<T>* obj, T* expc,
+                                          T desr,
+                                          memory_order s, memory_order f) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_weak_explicit(atomic<T>* obj, T* expc, T desr,
+                                          memory_order s, memory_order f) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_strong_explicit(volatile atomic<T>* obj,
+                                            T* expc, T desr,
+                                            memory_order s, memory_order f) noexcept;
+
+template <class T>
+    bool
+    atomic_compare_exchange_strong_explicit(atomic<T>* obj, T* expc,
+                                            T desr,
+                                            memory_order s, memory_order f) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_add(volatile atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_add(atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_add_explicit(volatile atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_add_explicit(atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_sub(volatile atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_sub(atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_sub_explicit(volatile atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_sub_explicit(atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_and(volatile atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_and(atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_and_explicit(volatile atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_and_explicit(atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_or(volatile atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_or(atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_or_explicit(volatile atomic<Integral>* obj, Integral op,
+                             memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_or_explicit(atomic<Integral>* obj, Integral op,
+                             memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_xor(volatile atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_xor(atomic<Integral>* obj, Integral op) noexcept;
+
+template <class Integral>
+    Integral
+    atomic_fetch_xor_explicit(volatile atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+template <class Integral>
+    Integral
+    atomic_fetch_xor_explicit(atomic<Integral>* obj, Integral op,
+                              memory_order m) noexcept;
+
+template <class T>
+    T*
+    atomic_fetch_add(volatile atomic<T*>* obj, ptrdiff_t op) noexcept;
+
+template <class T>
+    T*
+    atomic_fetch_add(atomic<T*>* obj, ptrdiff_t op) noexcept;
+
+template <class T>
+    T*
+    atomic_fetch_add_explicit(volatile atomic<T*>* obj, ptrdiff_t op,
+                              memory_order m) noexcept;
+template <class T>
+    T*
+    atomic_fetch_add_explicit(atomic<T*>* obj, ptrdiff_t op, memory_order m) noexcept;
+
+template <class T>
+    T*
+    atomic_fetch_sub(volatile atomic<T*>* obj, ptrdiff_t op) noexcept;
+
+template <class T>
+    T*
+    atomic_fetch_sub(atomic<T*>* obj, ptrdiff_t op) noexcept;
+
+template <class T>
+    T*
+    atomic_fetch_sub_explicit(volatile atomic<T*>* obj, ptrdiff_t op,
+                              memory_order m) noexcept;
+template <class T>
+    T*
+    atomic_fetch_sub_explicit(atomic<T*>* obj, ptrdiff_t op, memory_order m) noexcept;
+
+// Atomics for standard typedef types
+
+typedef atomic<bool>               atomic_bool;
+typedef atomic<char>               atomic_char;
+typedef atomic<signed char>        atomic_schar;
+typedef atomic<unsigned char>      atomic_uchar;
+typedef atomic<short>              atomic_short;
+typedef atomic<unsigned short>     atomic_ushort;
+typedef atomic<int>                atomic_int;
+typedef atomic<unsigned int>       atomic_uint;
+typedef atomic<long>               atomic_long;
+typedef atomic<unsigned long>      atomic_ulong;
+typedef atomic<long long>          atomic_llong;
+typedef atomic<unsigned long long> atomic_ullong;
+typedef atomic<char16_t>           atomic_char16_t;
+typedef atomic<char32_t>           atomic_char32_t;
+typedef atomic<wchar_t>            atomic_wchar_t;
+
+typedef atomic<int_least8_t>   atomic_int_least8_t;
+typedef atomic<uint_least8_t>  atomic_uint_least8_t;
+typedef atomic<int_least16_t>  atomic_int_least16_t;
+typedef atomic<uint_least16_t> atomic_uint_least16_t;
+typedef atomic<int_least32_t>  atomic_int_least32_t;
+typedef atomic<uint_least32_t> atomic_uint_least32_t;
+typedef atomic<int_least64_t>  atomic_int_least64_t;
+typedef atomic<uint_least64_t> atomic_uint_least64_t;
+
+typedef atomic<int_fast8_t>   atomic_int_fast8_t;
+typedef atomic<uint_fast8_t>  atomic_uint_fast8_t;
+typedef atomic<int_fast16_t>  atomic_int_fast16_t;
+typedef atomic<uint_fast16_t> atomic_uint_fast16_t;
+typedef atomic<int_fast32_t>  atomic_int_fast32_t;
+typedef atomic<uint_fast32_t> atomic_uint_fast32_t;
+typedef atomic<int_fast64_t>  atomic_int_fast64_t;
+typedef atomic<uint_fast64_t> atomic_uint_fast64_t;
+
+typedef atomic<int8_t>   atomic_int8_t;
+typedef atomic<uint8_t>  atomic_uint8_t;
+typedef atomic<int16_t>  atomic_int16_t;
+typedef atomic<uint16_t> atomic_uint16_t;
+typedef atomic<int32_t>  atomic_int32_t;
+typedef atomic<uint32_t> atomic_uint32_t;
+typedef atomic<int64_t>  atomic_int64_t;
+typedef atomic<uint64_t> atomic_uint64_t;
+
+typedef atomic<intptr_t>  atomic_intptr_t;
+typedef atomic<uintptr_t> atomic_uintptr_t;
+typedef atomic<size_t>    atomic_size_t;
+typedef atomic<ptrdiff_t> atomic_ptrdiff_t;
+typedef atomic<intmax_t>  atomic_intmax_t;
+typedef atomic<uintmax_t> atomic_uintmax_t;
+
+// fences
+
+void atomic_thread_fence(memory_order m) noexcept;
+void atomic_signal_fence(memory_order m) noexcept;
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__threading_support>
+#include <cstddef>
+#include <cstdint>
+#include <cstring>
+#include <type_traits>
+#include <version>
+#include <__pragma_push>
+
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+# error <atomic> is not supported on this single threaded system
+#endif
+#ifdef _LIBCUDACXX_HAS_NO_ATOMIC_HEADER
+# error <atomic> is not implemented
+#endif
+#ifdef kill_dependency
+# error C++ standard library is incompatible with <stdatomic.h>
+#endif
+
+#define _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m) \
+  _LIBCUDACXX_DIAGNOSE_WARNING(__m == memory_order_consume || \
+                           __m == memory_order_acquire || \
+                           __m == memory_order_acq_rel,   \
+                        "memory order argument to atomic operation is invalid")
+
+#define _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m) \
+  _LIBCUDACXX_DIAGNOSE_WARNING(__m == memory_order_release || \
+                           __m == memory_order_acq_rel,   \
+                        "memory order argument to atomic operation is invalid")
+
+#define _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__m, __f) \
+  _LIBCUDACXX_DIAGNOSE_WARNING(__f == memory_order_release || \
+                           __f == memory_order_acq_rel,   \
+                        "memory order argument to atomic operation is invalid")
+
+#if defined(_LIBCUDACXX_HAS_MSVC_ATOMIC_IMPL)
+#  include <intrin.h>
+#endif
+
+#if !defined(_LIBCUDACXX_COMPILER_NVRTC)
+#  include <string.h>
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// Figure out what the underlying type for `memory_order` would be if it were
+// declared as an unscoped enum (accounting for -fshort-enums). Use this result
+// to pin the underlying type in C++20.
+enum __legacy_memory_order {
+    __mo_relaxed,
+    __mo_consume,
+    __mo_acquire,
+    __mo_release,
+    __mo_acq_rel,
+    __mo_seq_cst
+};
+
+typedef underlying_type<__legacy_memory_order>::type __memory_order_underlying_t;
+
+#if _LIBCUDACXX_STD_VER > 17
+
+enum class memory_order : __memory_order_underlying_t {
+  relaxed = __mo_relaxed,
+  consume = __mo_consume,
+  acquire = __mo_acquire,
+  release = __mo_release,
+  acq_rel = __mo_acq_rel,
+  seq_cst = __mo_seq_cst
+};
+
+inline constexpr auto memory_order_relaxed = memory_order::relaxed;
+inline constexpr auto memory_order_consume = memory_order::consume;
+inline constexpr auto memory_order_acquire = memory_order::acquire;
+inline constexpr auto memory_order_release = memory_order::release;
+inline constexpr auto memory_order_acq_rel = memory_order::acq_rel;
+inline constexpr auto memory_order_seq_cst = memory_order::seq_cst;
+
+#else
+
+typedef enum memory_order {
+  memory_order_relaxed = __mo_relaxed,
+  memory_order_consume = __mo_consume,
+  memory_order_acquire = __mo_acquire,
+  memory_order_release = __mo_release,
+  memory_order_acq_rel = __mo_acq_rel,
+  memory_order_seq_cst = __mo_seq_cst,
+} memory_order;
+
+#endif // _LIBCUDACXX_STD_VER > 17
+
+template <typename _Tp> _LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_nonatomic_compare_equal(_Tp const& __lhs, _Tp const& __rhs) {
+#if defined(_LIBCUDACXX_COMPILER_NVCC) || defined(_LIBCUDACXX_COMPILER_NVRTC)
+    return __lhs == __rhs;
+#else
+    return memcmp(&__lhs, &__rhs, sizeof(_Tp)) == 0;
+#endif
+}
+
+static_assert((is_same<underlying_type<memory_order>::type, __memory_order_underlying_t>::value),
+  "unexpected underlying type for std::memory_order");
+
+#if defined(_LIBCUDACXX_HAS_GCC_ATOMIC_IMP) || \
+    defined(_LIBCUDACXX_ATOMIC_ONLY_USE_BUILTINS)
+
+// [atomics.types.generic]p1 guarantees _Tp is trivially copyable. Because
+// the default operator= in an object is not volatile, a byte-by-byte copy
+// is required.
+template <typename _Tp, typename _Tv> _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<is_assignable<_Tp&, _Tv>::value>::type
+__cxx_atomic_assign_volatile(_Tp& __a_value, _Tv const& __val) {
+  __a_value = __val;
+}
+template <typename _Tp, typename _Tv> _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<is_assignable<_Tp&, _Tv>::value>::type
+__cxx_atomic_assign_volatile(_Tp volatile& __a_value, _Tv volatile const& __val) {
+  volatile char* __to = reinterpret_cast<volatile char*>(&__a_value);
+  volatile char* __end = __to + sizeof(_Tp);
+  volatile const char* __from = reinterpret_cast<volatile const char*>(&__val);
+  while (__to != __end)
+    *__to++ = *__from++;
+}
+
+#endif
+
+// Headers are wrapped like so: (cuda::std::|std::)detail
+namespace __detail {
+#if defined(_LIBCUDACXX_HAS_CUDA_ATOMIC_EXT)
+#  include "support/atomic/atomic_scopes.h"
+#endif
+
+#if defined(_LIBCUDACXX_HAS_CUDA_ATOMIC_IMPL)
+#  include "support/atomic/atomic_cuda.h"
+#elif defined(_LIBCUDACXX_HAS_MSVC_ATOMIC_IMPL)
+#  include "support/atomic/atomic_msvc.h"
+#elif defined(_LIBCUDACXX_HAS_GCC_ATOMIC_IMP)
+#  include "support/atomic/atomic_gcc.h"
+#elif defined(_LIBCUDACXX_HAS_C_ATOMIC_IMP)
+// TODO: Maybe support C11 atomics?
+// #include "support/atomic/atomic_c11.h"
+#endif // _LIBCUDACXX_HAS_GCC_ATOMIC_IMP, _LIBCUDACXX_HAS_C_ATOMIC_IMP
+}
+
+using __detail::__cxx_atomic_base_impl;
+using __detail::__cxx_atomic_ref_base_impl;
+using __detail::__cxx_atomic_thread_fence;
+using __detail::__cxx_atomic_signal_fence;
+using __detail::__cxx_atomic_load;
+using __detail::__cxx_atomic_store;
+using __detail::__cxx_atomic_exchange;
+using __detail::__cxx_atomic_compare_exchange_weak;
+using __detail::__cxx_atomic_compare_exchange_strong;
+using __detail::__cxx_atomic_fetch_add;
+using __detail::__cxx_atomic_fetch_sub;
+using __detail::__cxx_atomic_fetch_or;
+using __detail::__cxx_atomic_fetch_and;
+using __detail::__cxx_atomic_fetch_xor;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp kill_dependency(_Tp __y) _NOEXCEPT
+{
+    return __y;
+}
+
+#if defined(__CLANG_ATOMIC_BOOL_LOCK_FREE)
+# define ATOMIC_BOOL_LOCK_FREE      __CLANG_ATOMIC_BOOL_LOCK_FREE
+# define ATOMIC_CHAR_LOCK_FREE      __CLANG_ATOMIC_CHAR_LOCK_FREE
+# define ATOMIC_CHAR16_T_LOCK_FREE  __CLANG_ATOMIC_CHAR16_T_LOCK_FREE
+# define ATOMIC_CHAR32_T_LOCK_FREE  __CLANG_ATOMIC_CHAR32_T_LOCK_FREE
+# define ATOMIC_WCHAR_T_LOCK_FREE   __CLANG_ATOMIC_WCHAR_T_LOCK_FREE
+# define ATOMIC_SHORT_LOCK_FREE     __CLANG_ATOMIC_SHORT_LOCK_FREE
+# define ATOMIC_INT_LOCK_FREE       __CLANG_ATOMIC_INT_LOCK_FREE
+# define ATOMIC_LONG_LOCK_FREE      __CLANG_ATOMIC_LONG_LOCK_FREE
+# define ATOMIC_LLONG_LOCK_FREE     __CLANG_ATOMIC_LLONG_LOCK_FREE
+# define ATOMIC_POINTER_LOCK_FREE   __CLANG_ATOMIC_POINTER_LOCK_FREE
+#elif defined(__GCC_ATOMIC_BOOL_LOCK_FREE)
+# define ATOMIC_BOOL_LOCK_FREE      __GCC_ATOMIC_BOOL_LOCK_FREE
+# define ATOMIC_CHAR_LOCK_FREE      __GCC_ATOMIC_CHAR_LOCK_FREE
+# define ATOMIC_CHAR16_T_LOCK_FREE  __GCC_ATOMIC_CHAR16_T_LOCK_FREE
+# define ATOMIC_CHAR32_T_LOCK_FREE  __GCC_ATOMIC_CHAR32_T_LOCK_FREE
+# define ATOMIC_WCHAR_T_LOCK_FREE   __GCC_ATOMIC_WCHAR_T_LOCK_FREE
+# define ATOMIC_SHORT_LOCK_FREE     __GCC_ATOMIC_SHORT_LOCK_FREE
+# define ATOMIC_INT_LOCK_FREE       __GCC_ATOMIC_INT_LOCK_FREE
+# define ATOMIC_LONG_LOCK_FREE      __GCC_ATOMIC_LONG_LOCK_FREE
+# define ATOMIC_LLONG_LOCK_FREE     __GCC_ATOMIC_LLONG_LOCK_FREE
+# define ATOMIC_POINTER_LOCK_FREE   __GCC_ATOMIC_POINTER_LOCK_FREE
+#endif
+
+#ifdef _LIBCUDACXX_ATOMIC_ONLY_USE_BUILTINS
+
+template<typename _Tp, int _Sco>
+struct __cxx_atomic_lock_impl {
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __cxx_atomic_lock_impl() _NOEXCEPT
+    : __a_value(), __a_lock(0) {}
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit
+  __cxx_atomic_lock_impl(_Tp value) _NOEXCEPT
+    : __a_value(value), __a_lock(0) {}
+
+  __cxx_atomic_lock_impl(const __cxx_atomic_lock_impl&) _NOEXCEPT = default;
+
+  _Tp __a_value;
+  mutable __cxx_atomic_base_impl<_LIBCUDACXX_ATOMIC_FLAG_TYPE, _Sco> __a_lock;
+
+  _LIBCUDACXX_INLINE_VISIBILITY void __lock() const volatile {
+    while(1 == __cxx_atomic_exchange(&__a_lock, _LIBCUDACXX_ATOMIC_FLAG_TYPE(true), memory_order_acquire))
+        /*spin*/;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY void __lock() const {
+    while(1 == __cxx_atomic_exchange(&__a_lock, _LIBCUDACXX_ATOMIC_FLAG_TYPE(true), memory_order_acquire))
+        /*spin*/;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY void __unlock() const volatile {
+    __cxx_atomic_store(&__a_lock, _LIBCUDACXX_ATOMIC_FLAG_TYPE(false), memory_order_release);
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY void __unlock() const {
+    __cxx_atomic_store(&__a_lock, _LIBCUDACXX_ATOMIC_FLAG_TYPE(false), memory_order_release);
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY _Tp __read() const volatile {
+    __lock();
+    _Tp __old;
+    __cxx_atomic_assign_volatile(__old, __a_value);
+    __unlock();
+    return __old;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY _Tp __read() const {
+    __lock();
+    _Tp __old = __a_value;
+    __unlock();
+    return __old;
+  }
+};
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_init(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,  _Tp __val) {
+  __cxx_atomic_assign_volatile(__a->__a_value, __val);
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_init(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,  _Tp __val) {
+  __a->__a_value = __val;
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_store(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,  _Tp __val, memory_order) {
+  __a->__lock();
+  __cxx_atomic_assign_volatile(__a->__a_value, __val);
+  __a->__unlock();
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_store(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,  _Tp __val, memory_order) {
+  __a->__lock();
+  __a->__a_value = __val;
+  __a->__unlock();
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_load(const volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a, memory_order) {
+  return __a->__read();
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_load(const __cxx_atomic_lock_impl<_Tp, _Sco>* __a, memory_order) {
+  return __a->__read();
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_exchange(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a, _Tp __value, memory_order) {
+  __a->__lock();
+  _Tp __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, __value);
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_exchange(__cxx_atomic_lock_impl<_Tp, _Sco>* __a, _Tp __value, memory_order) {
+  __a->__lock();
+  _Tp __old = __a->__a_value;
+  __a->__a_value = __value;
+  __a->__unlock();
+  return __old;
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_strong(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                                          _Tp* __expected, _Tp __value, memory_order, memory_order) {
+  __a->__lock();
+  _Tp __temp;
+  __cxx_atomic_assign_volatile(__temp, __a->__a_value);
+  bool __ret = __temp == *__expected;
+  if(__ret)
+    __cxx_atomic_assign_volatile(__a->__a_value, __value);
+  else
+    __cxx_atomic_assign_volatile(*__expected, __a->__a_value);
+  __a->__unlock();
+  return __ret;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_strong(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                                          _Tp* __expected, _Tp __value, memory_order, memory_order) {
+  __a->__lock();
+  bool __ret = __a->__a_value == *__expected;
+  if(__ret)
+    __a->__a_value = __value;
+  else
+    *__expected = __a->__a_value;
+  __a->__unlock();
+  return __ret;
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_weak(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                                        _Tp* __expected, _Tp __value, memory_order, memory_order) {
+  __a->__lock();
+  _Tp __temp;
+  __cxx_atomic_assign_volatile(__temp, __a->__a_value);
+  bool __ret = __temp == *__expected;
+  if(__ret)
+    __cxx_atomic_assign_volatile(__a->__a_value, __value);
+  else
+    __cxx_atomic_assign_volatile(*__expected, __a->__a_value);
+  __a->__unlock();
+  return __ret;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_weak(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                                        _Tp* __expected, _Tp __value, memory_order, memory_order) {
+  __a->__lock();
+  bool __ret = __a->__a_value == *__expected;
+  if(__ret)
+    __a->__a_value = __value;
+  else
+    *__expected = __a->__a_value;
+  __a->__unlock();
+  return __ret;
+}
+
+template <typename _Tp, typename _Td, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_add(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Td __delta, memory_order) {
+  __a->__lock();
+  _Tp __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, _Tp(__old + __delta));
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, typename _Td, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_add(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Td __delta, memory_order) {
+  __a->__lock();
+  _Tp __old = __a->__a_value;
+  __a->__a_value += __delta;
+  __a->__unlock();
+  return __old;
+}
+
+template <typename _Tp, typename _Td, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp* __cxx_atomic_fetch_add(volatile __cxx_atomic_lock_impl<_Tp*, _Sco>* __a,
+                           ptrdiff_t __delta, memory_order) {
+  __a->__lock();
+  _Tp* __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, __old + __delta);
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, typename _Td, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp* __cxx_atomic_fetch_add(__cxx_atomic_lock_impl<_Tp*, _Sco>* __a,
+                            ptrdiff_t __delta, memory_order) {
+  __a->__lock();
+  _Tp* __old = __a->__a_value;
+  __a->__a_value += __delta;
+  __a->__unlock();
+  return __old;
+}
+
+template <typename _Tp, typename _Td, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_sub(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Td __delta, memory_order) {
+  __a->__lock();
+  _Tp __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, _Tp(__old - __delta));
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, typename _Td, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_sub(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Td __delta, memory_order) {
+  __a->__lock();
+  _Tp __old = __a->__a_value;
+  __a->__a_value -= __delta;
+  __a->__unlock();
+  return __old;
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_and(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Tp __pattern, memory_order) {
+  __a->__lock();
+  _Tp __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, _Tp(__old & __pattern));
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_and(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Tp __pattern, memory_order) {
+  __a->__lock();
+  _Tp __old = __a->__a_value;
+  __a->__a_value &= __pattern;
+  __a->__unlock();
+  return __old;
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_or(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                          _Tp __pattern, memory_order) {
+  __a->__lock();
+  _Tp __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, _Tp(__old | __pattern));
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_or(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                          _Tp __pattern, memory_order) {
+  __a->__lock();
+  _Tp __old = __a->__a_value;
+  __a->__a_value |= __pattern;
+  __a->__unlock();
+  return __old;
+}
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_xor(volatile __cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Tp __pattern, memory_order) {
+  __a->__lock();
+  _Tp __old;
+  __cxx_atomic_assign_volatile(__old, __a->__a_value);
+  __cxx_atomic_assign_volatile(__a->__a_value, _Tp(__old ^ __pattern));
+  __a->__unlock();
+  return __old;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_xor(__cxx_atomic_lock_impl<_Tp, _Sco>* __a,
+                           _Tp __pattern, memory_order) {
+  __a->__lock();
+  _Tp __old = __a->__a_value;
+  __a->__a_value ^= __pattern;
+  __a->__unlock();
+  return __old;
+}
+
+#if defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+
+template<typename _Tp> struct __cxx_is_always_lock_free {
+    enum { __value = _LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE(sizeof(_Tp), 0) }; };
+
+#else
+
+template<typename _Tp> struct __cxx_is_always_lock_free {
+    enum { __value = sizeof(_Tp) <= 8 }; };
+
+#endif // defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+
+template <typename _Tp, int _Sco,
+          typename _Base = typename conditional<__cxx_is_always_lock_free<_Tp>::__value,
+                                                __cxx_atomic_base_impl<_Tp, _Sco>,
+                                                __cxx_atomic_lock_impl<_Tp, _Sco> >::type>
+#else
+template <typename _Tp, int _Sco,
+          typename _Base = __cxx_atomic_base_impl<_Tp, _Sco> >
+#endif //_LIBCUDACXX_ATOMIC_ONLY_USE_BUILTINS
+struct __cxx_atomic_impl : public _Base {
+  __cxx_atomic_impl() _NOEXCEPT _LIBCUDACXX_DEFAULT
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit __cxx_atomic_impl(_Tp value) _NOEXCEPT
+    : _Base(value) {}
+};
+
+
+template<int _Sco, typename _Tp = int>
+_LIBCUDACXX_INLINE_VISIBILITY
+__cxx_atomic_impl<_Tp, _Sco>* __cxx_atomic_rebind(_Tp* __inst) {
+    static_assert(sizeof(__cxx_atomic_impl<_Tp, _Sco>) == sizeof(_Tp),"");
+    static_assert(alignof(__cxx_atomic_impl<_Tp, _Sco>) == alignof(_Tp),"");
+    return (__cxx_atomic_impl<_Tp, _Sco>*)__inst;
+}
+
+template <typename _Tp, int _Sco>
+using __cxx_atomic_ref_impl = __cxx_atomic_ref_base_impl<_Tp, _Sco>;
+
+#ifdef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+
+template <class _Tp, int _Sco>
+struct __cxx_atomic_poll_tester {
+    __cxx_atomic_impl<_Tp, _Sco> const volatile* __a;
+    _Tp __val;
+    memory_order __order;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __cxx_atomic_poll_tester(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a_, _Tp __val_, memory_order __order_)
+      : __a(__a_)
+      , __val(__val_)
+      , __order(__order_)
+    {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY bool operator()() const {
+      return !(__cxx_atomic_load(__a, __order) == __val);
+    }
+};
+
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_try_wait_slow_fallback(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a, _Tp __val, memory_order __order) {
+    __libcpp_thread_poll_with_backoff(__cxx_atomic_poll_tester<_Tp, _Sco>(__a, __val, __order));
+}
+
+#endif
+
+#ifdef _LIBCUDACXX_HAS_PLATFORM_WAIT
+
+template <class _Tp, int _Sco, typename enable_if<!__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_all(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a) {
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+    auto * const __c = __libcpp_contention_state(__a);
+    __cxx_atomic_fetch_add(__cxx_atomic_rebind<_Sco>(&__c->__version), (__libcpp_platform_wait_t)1, memory_order_relaxed);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+    if (0 != __cxx_atomic_exchange(__cxx_atomic_rebind<_Sco>(&__c->__waiters), (ptrdiff_t)0, memory_order_relaxed))
+        __libcpp_platform_wake(&__c->__version, true);
+#endif
+}
+template <class _Tp, int _Sco, typename enable_if<!__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_one(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a) {
+    __cxx_atomic_notify_all(__a);
+}
+template <class _Tp, int _Sco, typename enable_if<!__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_try_wait_slow(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a, _Tp const __val, memory_order __order) {
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+    auto * const __c = __libcpp_contention_state(__a);
+    __cxx_atomic_store(__cxx_atomic_rebind<_Sco>(&__c->__waiters), (ptrdiff_t)1, memory_order_relaxed);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+    auto const __version = __cxx_atomic_load(__cxx_atomic_rebind<_Sco>(&__c->__version), memory_order_relaxed);
+    if (!__cxx_nonatomic_compare_equal(__cxx_atomic_load(__a, __order), __val))
+        return;
+    if(sizeof(__libcpp_platform_wait_t) < 8) {
+        constexpr timespec __timeout = { 2, 0 }; // Hedge on rare 'int version' aliasing.
+        __libcpp_platform_wait(&__c->__version, __version, &__timeout);
+    }
+    else
+        __libcpp_platform_wait(&__c->__version, __version, nullptr);
+#else
+    __cxx_atomic_try_wait_slow_fallback(__a, __val, __order);
+#endif // _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+}
+
+template <class _Tp, int _Sco, typename enable_if<__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_try_wait_slow(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a, _Tp __val, memory_order) {
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+    auto * const __c = __libcpp_contention_state(__a);
+    __cxx_atomic_fetch_add(__cxx_atomic_rebind<_Sco>(&__c->__waiters), (ptrdiff_t)1, memory_order_relaxed);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+#endif
+    __libcpp_platform_wait((_Tp*)__a, __val, nullptr);
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+    __cxx_atomic_fetch_sub(__cxx_atomic_rebind<_Sco>(&__c->__waiters), (ptrdiff_t)1, memory_order_relaxed);
+#endif
+}
+template <class _Tp, int _Sco, typename enable_if<__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_all(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a) {
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+    auto * const __c = __libcpp_contention_state(__a);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+    if (0 != __cxx_atomic_load(__cxx_atomic_rebind<_Sco>(&__c->__waiters), memory_order_relaxed))
+#endif
+        __libcpp_platform_wake((_Tp*)__a, true);
+}
+template <class _Tp, int _Sco, typename enable_if<__libcpp_platform_wait_uses_type<_Tp>::__value, int>::type = 1>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_one(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a) {
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE
+    auto * const __c = __libcpp_contention_state(__a);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+    if (0 != __cxx_atomic_load(__cxx_atomic_rebind<_Sco>(&__c->__waiters), memory_order_relaxed))
+#endif
+        __libcpp_platform_wake((_Tp*)__a, false);
+}
+
+#elif !defined(_LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE)
+
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_all(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a) {
+    auto * const __c = __libcpp_contention_state(__a);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+    if(0 == __cxx_atomic_load(__cxx_atomic_rebind<_Sco>(&__c->__credit), memory_order_relaxed))
+        return;
+    if(0 != __cxx_atomic_exchange(__cxx_atomic_rebind<_Sco>(&__c->__credit), (ptrdiff_t)0, memory_order_relaxed)) {
+        __libcpp_mutex_lock(&__c->__mutex);
+        __libcpp_mutex_unlock(&__c->__mutex);
+        __libcpp_condvar_broadcast(&__c->__condvar);
+    }
+}
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_one(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a) {
+    __cxx_atomic_notify_all(__a);
+}
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_try_wait_slow(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a, _Tp const __val, memory_order __order) {
+    auto * const __c = __libcpp_contention_state(__a);
+    __libcpp_mutex_lock(&__c->__mutex);
+    __cxx_atomic_store(__cxx_atomic_rebind<_Sco>(&__c->__credit), (ptrdiff_t)1, memory_order_relaxed);
+    __cxx_atomic_thread_fence(memory_order_seq_cst);
+    if (__cxx_nonatomic_compare_equal(__cxx_atomic_load(__a, __order), __val))
+        __libcpp_condvar_wait(&__c->__condvar, &__c->__mutex);
+    __libcpp_mutex_unlock(&__c->__mutex);
+}
+
+#else
+
+template<typename T>
+struct __atomic_wait_and_notify_supported
+#if defined(__CUDA_MINIMUM_ARCH__) && __CUDA_MINIMUM_ARCH__ < 700
+    : false_type
+#else
+    : true_type
+#endif
+{};
+
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_try_wait_slow(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a, _Tp __val, memory_order __order) {
+    static_assert(__atomic_wait_and_notify_supported<_Tp>::value, "atomic wait operations are unsupported on Pascal");
+    __cxx_atomic_try_wait_slow_fallback(__a, __val, __order);
+}
+
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_one(__cxx_atomic_impl<_Tp, _Sco> const volatile*) {
+    static_assert(__atomic_wait_and_notify_supported<_Tp>::value, "atomic notify-one operations are unsupported on Pascal");
+}
+
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_notify_all(__cxx_atomic_impl<_Tp, _Sco> const volatile*) {
+    static_assert(__atomic_wait_and_notify_supported<_Tp>::value, "atomic notify-all operations are unsupported on Pascal");
+}
+
+#endif // _LIBCUDACXX_HAS_PLATFORM_WAIT || !defined(_LIBCUDACXX_HAS_NO_THREAD_CONTENTION_TABLE)
+
+template <class _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY void __cxx_atomic_wait(__cxx_atomic_impl<_Tp, _Sco> const volatile* __a, _Tp const __val, memory_order __order) {
+    for(int __i = 0; __i < _LIBCUDACXX_POLLING_COUNT; ++__i) {
+        if(!__cxx_nonatomic_compare_equal(__cxx_atomic_load(__a, __order), __val))
+            return;
+        if(__i < 12)
+            __libcpp_thread_yield_processor();
+        else
+            __libcpp_thread_yield();
+    }
+    while(__cxx_nonatomic_compare_equal(__cxx_atomic_load(__a, __order), __val))
+        __cxx_atomic_try_wait_slow(__a, __val, __order);
+}
+
+// general atomic<T>/atomic_ref<T>
+template <class _Tp, int _Sco = 0, bool = is_integral<_Tp>::value && !is_same<_Tp, bool>::value>
+struct __atomic_base {
+    mutable __cxx_atomic_impl<_Tp, _Sco> __a_;
+
+    __atomic_base() = default;
+    __atomic_base(const __atomic_base&) = delete;
+    __atomic_base(__atomic_base&&) = delete;
+
+    __atomic_base& operator=(const __atomic_base&) = delete;
+    __atomic_base& operator=(__atomic_base&&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __atomic_base(const _Tp& __a) _NOEXCEPT : __a_(__a) {}
+
+#if defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+    static _LIBCUDACXX_CONSTEXPR bool is_always_lock_free = _LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE(sizeof(decltype(__a_)), 0);
+#endif // defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_lock_free() const volatile _NOEXCEPT
+        {return _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(sizeof(_Tp));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_lock_free() const _NOEXCEPT
+        {return static_cast<__atomic_base const volatile*>(this)->is_lock_free();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void store(_Tp __d, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m)
+        {__cxx_atomic_store(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void store(_Tp __d, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+      _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m)
+        {__cxx_atomic_store(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp load(memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+        {return __cxx_atomic_load(&__a_, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp load(memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+      _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+        {return __cxx_atomic_load(&__a_, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator _Tp() const volatile _NOEXCEPT {return load();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator _Tp() const _NOEXCEPT          {return load();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp exchange(_Tp __d, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_exchange(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp exchange(_Tp __d, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_exchange(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                               memory_order __s, memory_order __f) volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                               memory_order __s, memory_order __f) _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                                 memory_order __s, memory_order __f) volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                                 memory_order __s, memory_order __f) _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                              memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT {
+        if (memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if (memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, __m);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                               memory_order __m = memory_order_seq_cst) _NOEXCEPT {
+        if(memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if(memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, __m);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                              memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT {
+        if (memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if (memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, __m);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                                 memory_order __m = memory_order_seq_cst) _NOEXCEPT {
+        if (memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if (memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, __m);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY void wait(_Tp __v, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {__cxx_atomic_wait(&__a_, __v, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY void wait(_Tp __v, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {__cxx_atomic_wait(&__a_, __v, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_one() volatile _NOEXCEPT
+        {__cxx_atomic_notify_one(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_one() _NOEXCEPT
+        {__cxx_atomic_notify_one(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_all() volatile _NOEXCEPT
+        {__cxx_atomic_notify_all(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_all() _NOEXCEPT
+        {__cxx_atomic_notify_all(&__a_);}
+};
+
+template <class _Tp, int _Sco = 0, bool = is_integral<_Tp>() && !is_same<_Tp, bool>()>
+struct __atomic_base_ref {
+    mutable __cxx_atomic_ref_impl<_Tp, _Sco> __a_;
+
+    __atomic_base_ref() = default;
+    __atomic_base_ref(const __atomic_base_ref&) = default;
+    __atomic_base_ref(__atomic_base_ref&&) = default;
+
+    __atomic_base_ref& operator=(const __atomic_base_ref&) = delete;
+    __atomic_base_ref& operator=(__atomic_base_ref&&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __atomic_base_ref(_Tp& __a) _NOEXCEPT : __a_(__a) {}
+
+#if defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+    static _LIBCUDACXX_CONSTEXPR bool is_always_lock_free = _LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE(sizeof(decltype(__a_)), 0);
+#endif // defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_lock_free() const volatile _NOEXCEPT
+        {return _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(sizeof(_Tp));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_lock_free() const _NOEXCEPT
+        {return static_cast<__atomic_base_ref const volatile*>(this)->is_lock_free();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void store(_Tp __d, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m)
+        {__cxx_atomic_store(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void store(_Tp __d, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+      _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m)
+        {__cxx_atomic_store(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp load(memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+        {return __cxx_atomic_load(&__a_, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp load(memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+      _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+        {return __cxx_atomic_load(&__a_, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator _Tp() const volatile _NOEXCEPT {return load();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator _Tp() const _NOEXCEPT          {return load();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp exchange(_Tp __d, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return __cxx_atomic_exchange(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp exchange(_Tp __d, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return __cxx_atomic_exchange(&__a_, __d, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                               memory_order __s, memory_order __f) const volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                               memory_order __s, memory_order __f) const _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                                 memory_order __s, memory_order __f) const volatile _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                                 memory_order __s, memory_order __f) const _NOEXCEPT
+      _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+        {return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __s, __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                              memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT {
+        if (memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if (memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, __m);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_weak(_Tp& __e, _Tp __d,
+                               memory_order __m = memory_order_seq_cst) const _NOEXCEPT {
+        if(memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if(memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_weak(&__a_, &__e, __d, __m, __m);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                              memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT {
+        if (memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if (memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, __m);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool compare_exchange_strong(_Tp& __e, _Tp __d,
+                                 memory_order __m = memory_order_seq_cst) const _NOEXCEPT {
+        if (memory_order_acq_rel == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_acquire);
+        else if (memory_order_release == __m)
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, memory_order_relaxed);
+        else
+            return __cxx_atomic_compare_exchange_strong(&__a_, &__e, __d, __m, __m);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY void wait(_Tp __v, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {__cxx_atomic_wait(&__a_, __v, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY void wait(_Tp __v, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {__cxx_atomic_wait(&__a_, __v, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_one() const volatile _NOEXCEPT
+        {__cxx_atomic_notify_one(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_one() const _NOEXCEPT
+        {__cxx_atomic_notify_one(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_all() const volatile _NOEXCEPT
+        {__cxx_atomic_notify_all(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY void notify_all() const _NOEXCEPT
+        {__cxx_atomic_notify_all(&__a_);}
+};
+
+// atomic/atomic_ref<integral>
+template <class _Tp, int _Sco>
+struct __atomic_base<_Tp, _Sco, true> : public __atomic_base<_Tp, _Sco, false> {
+    __atomic_base() = default;
+    __atomic_base(const __atomic_base&) = delete;
+    __atomic_base(__atomic_base&&) = delete;
+
+    __atomic_base& operator=(const __atomic_base&) = delete;
+    __atomic_base& operator=(__atomic_base&&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __atomic_base(const _Tp& __a) _NOEXCEPT : __atomic_base<_Tp, _Sco, false>(__a) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_add(_Tp __op, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_add(_Tp __op, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_sub(_Tp __op, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_sub(_Tp __op, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_and(_Tp __op, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_and(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_and(_Tp __op, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_fetch_and(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_or(_Tp __op, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_or(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_or(_Tp __op, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_fetch_or(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_xor(_Tp __op, memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_xor(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_xor(_Tp __op, memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_fetch_xor(&this->__a_, __op, __m);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++(int) volatile _NOEXCEPT      {return fetch_add(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++(int) _NOEXCEPT               {return fetch_add(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--(int) volatile _NOEXCEPT      {return fetch_sub(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--(int) _NOEXCEPT               {return fetch_sub(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++() volatile _NOEXCEPT         {return fetch_add(_Tp(1)) + _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++() _NOEXCEPT                  {return fetch_add(_Tp(1)) + _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--() volatile _NOEXCEPT         {return fetch_sub(_Tp(1)) - _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--() _NOEXCEPT                  {return fetch_sub(_Tp(1)) - _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator+=(_Tp __op) volatile _NOEXCEPT {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator+=(_Tp __op) _NOEXCEPT          {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator-=(_Tp __op) volatile _NOEXCEPT {return fetch_sub(__op) - __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator-=(_Tp __op) _NOEXCEPT          {return fetch_sub(__op) - __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator&=(_Tp __op) volatile _NOEXCEPT {return fetch_and(__op) & __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator&=(_Tp __op) _NOEXCEPT          {return fetch_and(__op) & __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator|=(_Tp __op) volatile _NOEXCEPT {return fetch_or(__op) | __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator|=(_Tp __op) _NOEXCEPT          {return fetch_or(__op) | __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator^=(_Tp __op) volatile _NOEXCEPT {return fetch_xor(__op) ^ __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator^=(_Tp __op) _NOEXCEPT          {return fetch_xor(__op) ^ __op;}
+};
+
+template <class _Tp, int _Sco>
+struct __atomic_base_ref<_Tp, _Sco, true> : public __atomic_base_ref<_Tp, _Sco, false> {
+    __atomic_base_ref() = default;
+    __atomic_base_ref(const __atomic_base_ref&) = default;
+    __atomic_base_ref(__atomic_base_ref&&) = default;
+
+    __atomic_base_ref& operator=(const __atomic_base_ref&) = delete;
+    __atomic_base_ref& operator=(__atomic_base_ref&&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __atomic_base_ref(_Tp& __a) _NOEXCEPT : __atomic_base_ref<_Tp, _Sco, false>(__a) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_add(_Tp __op, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_add(_Tp __op, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_sub(_Tp __op, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_sub(_Tp __op, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_and(_Tp __op, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_and(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_and(_Tp __op, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return __cxx_atomic_fetch_and(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_or(_Tp __op, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_or(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_or(_Tp __op, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return __cxx_atomic_fetch_or(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_xor(_Tp __op, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_xor(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp fetch_xor(_Tp __op, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return __cxx_atomic_fetch_xor(&this->__a_, __op, __m);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++(int) const volatile _NOEXCEPT      {return fetch_add(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++(int) const _NOEXCEPT               {return fetch_add(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--(int) const volatile _NOEXCEPT      {return fetch_sub(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--(int) const _NOEXCEPT               {return fetch_sub(_Tp(1));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++() const volatile _NOEXCEPT         {return fetch_add(_Tp(1)) + _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator++() const _NOEXCEPT                  {return fetch_add(_Tp(1)) + _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--() const volatile _NOEXCEPT         {return fetch_sub(_Tp(1)) - _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator--() const _NOEXCEPT                  {return fetch_sub(_Tp(1)) - _Tp(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator+=(_Tp __op) const volatile _NOEXCEPT {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator+=(_Tp __op) const _NOEXCEPT          {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator-=(_Tp __op) const volatile _NOEXCEPT {return fetch_sub(__op) - __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator-=(_Tp __op) const _NOEXCEPT          {return fetch_sub(__op) - __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator&=(_Tp __op) const volatile _NOEXCEPT {return fetch_and(__op) & __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator&=(_Tp __op) const _NOEXCEPT          {return fetch_and(__op) & __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator|=(_Tp __op) const volatile _NOEXCEPT {return fetch_or(__op) | __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator|=(_Tp __op) const _NOEXCEPT          {return fetch_or(__op) | __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator^=(_Tp __op) const volatile _NOEXCEPT {return fetch_xor(__op) ^ __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator^=(_Tp __op) const _NOEXCEPT          {return fetch_xor(__op) ^ __op;}
+};
+
+#if defined(_LIBCUDACXX_ATOMIC_ALWAYS_LOCK_FREE)
+template <class _Tp, int _Sco, bool _Integral>
+_LIBCUDACXX_CONSTEXPR bool __atomic_base<_Tp, _Sco, _Integral>::is_always_lock_free;
+
+template <class _Tp, int _Sco, bool _Integral>
+_LIBCUDACXX_CONSTEXPR bool __atomic_base_ref<_Tp, _Sco, _Integral>::is_always_lock_free;
+#endif
+
+// atomic<T>
+
+template <class _Tp>
+struct atomic
+    : public __atomic_base<_Tp>
+{
+    typedef __atomic_base<_Tp> __base;
+    using value_type = _Tp;
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+#endif
+    atomic() _NOEXCEPT _LIBCUDACXX_DEFAULT
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR atomic(_Tp __d) _NOEXCEPT : __base(__d) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator=(_Tp __d) volatile _NOEXCEPT
+        {__base::store(__d); return __d;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator=(_Tp __d) _NOEXCEPT
+        {__base::store(__d); return __d;}
+};
+
+// atomic<T*>
+
+template <class _Tp>
+struct atomic<_Tp*>
+    : public __atomic_base<_Tp*>
+{
+    typedef __atomic_base<_Tp*> __base;
+    using value_type = _Tp*;
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+#endif
+    atomic() _NOEXCEPT _LIBCUDACXX_DEFAULT
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR atomic(_Tp* __d) _NOEXCEPT : __base(__d) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator=(_Tp* __d) volatile _NOEXCEPT
+        {__base::store(__d); return __d;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator=(_Tp* __d) _NOEXCEPT
+        {__base::store(__d); return __d;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++(int) volatile _NOEXCEPT            {return fetch_add(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++(int) _NOEXCEPT                     {return fetch_add(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--(int) volatile _NOEXCEPT            {return fetch_sub(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--(int) _NOEXCEPT                     {return fetch_sub(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++() volatile _NOEXCEPT               {return fetch_add(1) + 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++() _NOEXCEPT                        {return fetch_add(1) + 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--() volatile _NOEXCEPT               {return fetch_sub(1) - 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--() _NOEXCEPT                        {return fetch_sub(1) - 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator+=(ptrdiff_t __op) volatile _NOEXCEPT {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator+=(ptrdiff_t __op) _NOEXCEPT          {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator-=(ptrdiff_t __op) volatile _NOEXCEPT {return fetch_sub(__op) - __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator-=(ptrdiff_t __op) _NOEXCEPT          {return fetch_sub(__op) - __op;}
+};
+
+// atomic_ref<T>
+
+template <class _Tp>
+ struct atomic_ref
+    : public __atomic_base_ref<_Tp>
+{
+    typedef __atomic_base_ref<_Tp> __base;
+    using value_type = _Tp;
+
+    static constexpr size_t required_alignment = sizeof(_Tp);
+
+    static constexpr bool is_always_lock_free = sizeof(_Tp) <= 8;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit atomic_ref(_Tp& __ref) : __base(__ref) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator=(_Tp __v) const noexcept {__base::store(__v); return __v;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator=(_Tp __v) const volatile noexcept {__base::store(__v); return __v;}
+};
+
+// atomic_ref<T*>
+
+template <class _Tp>
+ struct atomic_ref<_Tp*>
+    : public __atomic_base_ref<_Tp*>
+{
+    typedef __atomic_base_ref<_Tp*> __base;
+    using value_type = _Tp*;
+
+    static constexpr size_t required_alignment = sizeof(_Tp*);
+
+    static constexpr bool is_always_lock_free = sizeof(_Tp*) <= 8;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit atomic_ref(_Tp*& __ref) : __base(__ref) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator=(_Tp* __v) const noexcept {__base::store(__v); return __v;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator=(_Tp* __v) const volatile noexcept {__base::store(__v); return __v;}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_add(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        const _NOEXCEPT
+        {return __cxx_atomic_fetch_add(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        const volatile _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* fetch_sub(ptrdiff_t __op, memory_order __m = memory_order_seq_cst)
+                                                                        const _NOEXCEPT
+        {return __cxx_atomic_fetch_sub(&this->__a_, __op, __m);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++(int) const volatile _NOEXCEPT            {return fetch_add(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++(int) const _NOEXCEPT                     {return fetch_add(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--(int) const volatile _NOEXCEPT            {return fetch_sub(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--(int) const _NOEXCEPT                     {return fetch_sub(1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++() const volatile _NOEXCEPT               {return fetch_add(1) + 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator++() const _NOEXCEPT                        {return fetch_add(1) + 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--() const volatile _NOEXCEPT               {return fetch_sub(1) - 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator--() const _NOEXCEPT                        {return fetch_sub(1) - 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator+=(ptrdiff_t __op) const volatile _NOEXCEPT {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator+=(ptrdiff_t __op) const _NOEXCEPT          {return fetch_add(__op) + __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator-=(ptrdiff_t __op) const volatile _NOEXCEPT {return fetch_sub(__op) - __op;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* operator-=(ptrdiff_t __op) const _NOEXCEPT          {return fetch_sub(__op) - __op;}
+};
+
+// atomic_is_lock_free
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_is_lock_free(const volatile atomic<_Tp>* __o) _NOEXCEPT
+{
+    return __o->is_lock_free();
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_is_lock_free(const atomic<_Tp>* __o) _NOEXCEPT
+{
+    return __o->is_lock_free();
+}
+
+// atomic_init
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_init(volatile atomic<_Tp>* __o, _Tp __d) _NOEXCEPT
+{
+    __cxx_atomic_init(&__o->__a_, __d);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_init(atomic<_Tp>* __o, _Tp __d) _NOEXCEPT
+{
+    __cxx_atomic_init(&__o->__a_, __d);
+}
+
+// atomic_store
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_store(volatile atomic<_Tp>* __o, _Tp __d) _NOEXCEPT
+{
+    __o->store(__d);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_store(atomic<_Tp>* __o, _Tp __d) _NOEXCEPT
+{
+    __o->store(__d);
+}
+
+// atomic_store_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_store_explicit(volatile atomic<_Tp>* __o, _Tp __d, memory_order __m) _NOEXCEPT
+  _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m)
+{
+    __o->store(__d, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_store_explicit(atomic<_Tp>* __o, _Tp __d, memory_order __m) _NOEXCEPT
+  _LIBCUDACXX_CHECK_STORE_MEMORY_ORDER(__m)
+{
+    __o->store(__d, __m);
+}
+
+// atomic_load
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_load(const volatile atomic<_Tp>* __o) _NOEXCEPT
+{
+    return __o->load();
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_load(const atomic<_Tp>* __o) _NOEXCEPT
+{
+    return __o->load();
+}
+
+// atomic_load_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_load_explicit(const volatile atomic<_Tp>* __o, memory_order __m) _NOEXCEPT
+  _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+{
+    return __o->load(__m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_load_explicit(const atomic<_Tp>* __o, memory_order __m) _NOEXCEPT
+  _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+{
+    return __o->load(__m);
+}
+
+// atomic_exchange
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_exchange(volatile atomic<_Tp>* __o, _Tp __d) _NOEXCEPT
+{
+    return __o->exchange(__d);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_exchange(atomic<_Tp>* __o, _Tp __d) _NOEXCEPT
+{
+    return __o->exchange(__d);
+}
+
+// atomic_exchange_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_exchange_explicit(volatile atomic<_Tp>* __o, _Tp __d, memory_order __m) _NOEXCEPT
+{
+    return __o->exchange(__d, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+atomic_exchange_explicit(atomic<_Tp>* __o, _Tp __d, memory_order __m) _NOEXCEPT
+{
+    return __o->exchange(__d, __m);
+}
+
+// atomic_compare_exchange_weak
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_weak(volatile atomic<_Tp>* __o, _Tp* __e, _Tp __d) _NOEXCEPT
+{
+    return __o->compare_exchange_weak(*__e, __d);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_weak(atomic<_Tp>* __o, _Tp* __e, _Tp __d) _NOEXCEPT
+{
+    return __o->compare_exchange_weak(*__e, __d);
+}
+
+// atomic_compare_exchange_strong
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_strong(volatile atomic<_Tp>* __o, _Tp* __e, _Tp __d) _NOEXCEPT
+{
+    return __o->compare_exchange_strong(*__e, __d);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_strong(atomic<_Tp>* __o, _Tp* __e, _Tp __d) _NOEXCEPT
+{
+    return __o->compare_exchange_strong(*__e, __d);
+}
+
+// atomic_compare_exchange_weak_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_weak_explicit(volatile atomic<_Tp>* __o, _Tp* __e,
+                                      _Tp __d,
+                                      memory_order __s, memory_order __f) _NOEXCEPT
+  _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+{
+    return __o->compare_exchange_weak(*__e, __d, __s, __f);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_weak_explicit(atomic<_Tp>* __o, _Tp* __e, _Tp __d,
+                                      memory_order __s, memory_order __f) _NOEXCEPT
+  _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+{
+    return __o->compare_exchange_weak(*__e, __d, __s, __f);
+}
+
+// atomic_compare_exchange_strong_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_strong_explicit(volatile atomic<_Tp>* __o,
+                                        _Tp* __e, _Tp __d,
+                                        memory_order __s, memory_order __f) _NOEXCEPT
+  _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+{
+    return __o->compare_exchange_strong(*__e, __d, __s, __f);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_compare_exchange_strong_explicit(atomic<_Tp>* __o, _Tp* __e,
+                                        _Tp __d,
+                                        memory_order __s, memory_order __f) _NOEXCEPT
+  _LIBCUDACXX_CHECK_EXCHANGE_MEMORY_ORDER(__s, __f)
+{
+    return __o->compare_exchange_strong(*__e, __d, __s, __f);
+}
+
+// atomic_wait
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_wait(const volatile atomic<_Tp>* __o,
+                    typename atomic<_Tp>::value_type __v) _NOEXCEPT
+{
+    return __o->wait(__v);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_wait(const atomic<_Tp>* __o,
+                    typename atomic<_Tp>::value_type __v) _NOEXCEPT
+{
+    return __o->wait(__v);
+}
+
+// atomic_wait_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_wait_explicit(const volatile atomic<_Tp>* __o,
+                            typename atomic<_Tp>::value_type __v,
+                            memory_order __m) _NOEXCEPT
+  _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+{
+    return __o->wait(__v, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_wait_explicit(const atomic<_Tp>* __o,
+                            typename atomic<_Tp>::value_type __v,
+                            memory_order __m) _NOEXCEPT
+  _LIBCUDACXX_CHECK_LOAD_MEMORY_ORDER(__m)
+{
+    return __o->wait(__v, __m);
+}
+
+// atomic_notify_one
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_notify_one(volatile atomic<_Tp>* __o) _NOEXCEPT
+{
+    __o->notify_one();
+}
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_notify_one(atomic<_Tp>* __o) _NOEXCEPT
+{
+    __o->notify_one();
+}
+
+// atomic_notify_one
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_notify_all(volatile atomic<_Tp>* __o) _NOEXCEPT
+{
+    __o->notify_all();
+}
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void atomic_notify_all(atomic<_Tp>* __o) _NOEXCEPT
+{
+    __o->notify_all();
+}
+
+// atomic_fetch_add
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_add(volatile atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_add(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_add(atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_add(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_add(volatile atomic<_Tp*>* __o, ptrdiff_t __op) _NOEXCEPT
+{
+    return __o->fetch_add(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_add(atomic<_Tp*>* __o, ptrdiff_t __op) _NOEXCEPT
+{
+    return __o->fetch_add(__op);
+}
+
+// atomic_fetch_add_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_add_explicit(volatile atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_add(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_add_explicit(atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_add(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_add_explicit(volatile atomic<_Tp*>* __o, ptrdiff_t __op,
+                          memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_add(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_add_explicit(atomic<_Tp*>* __o, ptrdiff_t __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_add(__op, __m);
+}
+
+// atomic_fetch_sub
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_sub(volatile atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_sub(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_sub(atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_sub(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_sub(volatile atomic<_Tp*>* __o, ptrdiff_t __op) _NOEXCEPT
+{
+    return __o->fetch_sub(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_sub(atomic<_Tp*>* __o, ptrdiff_t __op) _NOEXCEPT
+{
+    return __o->fetch_sub(__op);
+}
+
+// atomic_fetch_sub_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_sub_explicit(volatile atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_sub(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_sub_explicit(atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_sub(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_sub_explicit(volatile atomic<_Tp*>* __o, ptrdiff_t __op,
+                          memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_sub(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+atomic_fetch_sub_explicit(atomic<_Tp*>* __o, ptrdiff_t __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_sub(__op, __m);
+}
+
+// atomic_fetch_and
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_and(volatile atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_and(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_and(atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_and(__op);
+}
+
+// atomic_fetch_and_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_and_explicit(volatile atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_and(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_and_explicit(atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_and(__op, __m);
+}
+
+// atomic_fetch_or
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_or(volatile atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_or(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_or(atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_or(__op);
+}
+
+// atomic_fetch_or_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_or_explicit(volatile atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_or(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_or_explicit(atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_or(__op, __m);
+}
+
+// atomic_fetch_xor
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_xor(volatile atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_xor(__op);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_xor(atomic<_Tp>* __o, _Tp __op) _NOEXCEPT
+{
+    return __o->fetch_xor(__op);
+}
+
+// atomic_fetch_xor_explicit
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_xor_explicit(volatile atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_xor(__op, __m);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value && !is_same<_Tp, bool>::value,
+    _Tp
+>::type
+atomic_fetch_xor_explicit(atomic<_Tp>* __o, _Tp __op, memory_order __m) _NOEXCEPT
+{
+    return __o->fetch_xor(__op, __m);
+}
+
+// flag type and operations
+
+typedef struct atomic_flag
+{
+    __cxx_atomic_impl<_LIBCUDACXX_ATOMIC_FLAG_TYPE, 0> __a_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool test(memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {return _LIBCUDACXX_ATOMIC_FLAG_TYPE(true)==__cxx_atomic_load(&__a_, __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool test(memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {return _LIBCUDACXX_ATOMIC_FLAG_TYPE(true)==__cxx_atomic_load(&__a_, __m);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool test_and_set(memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {return __cxx_atomic_exchange(&__a_, _LIBCUDACXX_ATOMIC_FLAG_TYPE(true), __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool test_and_set(memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {return __cxx_atomic_exchange(&__a_, _LIBCUDACXX_ATOMIC_FLAG_TYPE(true), __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear(memory_order __m = memory_order_seq_cst) volatile _NOEXCEPT
+        {__cxx_atomic_store(&__a_, _LIBCUDACXX_ATOMIC_FLAG_TYPE(false), __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear(memory_order __m = memory_order_seq_cst) _NOEXCEPT
+        {__cxx_atomic_store(&__a_, _LIBCUDACXX_ATOMIC_FLAG_TYPE(false), __m);}
+
+#if !defined(__CUDA_MINIMUM_ARCH__) || __CUDA_MINIMUM_ARCH__ >= 700
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait(bool __v, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT
+        {__cxx_atomic_wait(&__a_, _LIBCUDACXX_ATOMIC_FLAG_TYPE(__v), __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait(bool __v, memory_order __m = memory_order_seq_cst) const _NOEXCEPT
+        {__cxx_atomic_wait(&__a_, _LIBCUDACXX_ATOMIC_FLAG_TYPE(__v), __m);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void notify_one() volatile _NOEXCEPT
+        {__cxx_atomic_notify_one(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void notify_one() _NOEXCEPT
+        {__cxx_atomic_notify_one(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void notify_all() volatile _NOEXCEPT
+        {__cxx_atomic_notify_all(&__a_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void notify_all() _NOEXCEPT
+        {__cxx_atomic_notify_all(&__a_);}
+#endif
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+#endif
+    atomic_flag() _NOEXCEPT _LIBCUDACXX_DEFAULT
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    atomic_flag(bool __b) _NOEXCEPT : __a_(__b) {} // EXTENSION
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    atomic_flag(const atomic_flag&) = delete;
+    atomic_flag& operator=(const atomic_flag&) = delete;
+    atomic_flag& operator=(const atomic_flag&) volatile = delete;
+#else
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    atomic_flag(const atomic_flag&);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    atomic_flag& operator=(const atomic_flag&);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    atomic_flag& operator=(const atomic_flag&) volatile;
+#endif
+} atomic_flag;
+
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test(const volatile atomic_flag* __o) _NOEXCEPT
+{
+    return __o->test();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test(const atomic_flag* __o) _NOEXCEPT
+{
+    return __o->test();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test_explicit(const volatile atomic_flag* __o, memory_order __m) _NOEXCEPT
+{
+    return __o->test(__m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test_explicit(const atomic_flag* __o, memory_order __m) _NOEXCEPT
+{
+    return __o->test(__m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test_and_set(volatile atomic_flag* __o) _NOEXCEPT
+{
+    return __o->test_and_set();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test_and_set(atomic_flag* __o) _NOEXCEPT
+{
+    return __o->test_and_set();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test_and_set_explicit(volatile atomic_flag* __o, memory_order __m) _NOEXCEPT
+{
+    return __o->test_and_set(__m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_flag_test_and_set_explicit(atomic_flag* __o, memory_order __m) _NOEXCEPT
+{
+    return __o->test_and_set(__m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_clear(volatile atomic_flag* __o) _NOEXCEPT
+{
+    __o->clear();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_clear(atomic_flag* __o) _NOEXCEPT
+{
+    __o->clear();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_clear_explicit(volatile atomic_flag* __o, memory_order __m) _NOEXCEPT
+{
+    __o->clear(__m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_clear_explicit(atomic_flag* __o, memory_order __m) _NOEXCEPT
+{
+    __o->clear(__m);
+}
+
+#if !defined(__CUDA_MINIMUM_ARCH__) || __CUDA_MINIMUM_ARCH__ >= 700
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_wait(const volatile atomic_flag* __o, bool __v) _NOEXCEPT
+{
+    __o->wait(__v);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_wait(const atomic_flag* __o, bool __v) _NOEXCEPT
+{
+    __o->wait(__v);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_wait_explicit(const volatile atomic_flag* __o,
+                          bool __v, memory_order __m) _NOEXCEPT
+{
+    __o->wait(__v, __m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_wait_explicit(const atomic_flag* __o,
+                          bool __v, memory_order __m) _NOEXCEPT
+{
+    __o->wait(__v, __m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_notify_one(volatile atomic_flag* __o) _NOEXCEPT
+{
+    __o->notify_one();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_notify_one(atomic_flag* __o) _NOEXCEPT
+{
+    __o->notify_one();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_notify_all(volatile atomic_flag* __o) _NOEXCEPT
+{
+    __o->notify_all();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_flag_notify_all(atomic_flag* __o) _NOEXCEPT
+{
+    __o->notify_all();
+}
+
+#endif
+
+// fences
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_thread_fence(memory_order __m) _NOEXCEPT
+{
+    __cxx_atomic_thread_fence(__m);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+atomic_signal_fence(memory_order __m) _NOEXCEPT
+{
+    __cxx_atomic_signal_fence(__m);
+}
+
+// Atomics for standard typedef types
+
+typedef atomic<bool>               atomic_bool;
+typedef atomic<char>               atomic_char;
+typedef atomic<signed char>        atomic_schar;
+typedef atomic<unsigned char>      atomic_uchar;
+typedef atomic<short>              atomic_short;
+typedef atomic<unsigned short>     atomic_ushort;
+typedef atomic<int>                atomic_int;
+typedef atomic<unsigned int>       atomic_uint;
+typedef atomic<long>               atomic_long;
+typedef atomic<unsigned long>      atomic_ulong;
+typedef atomic<long long>          atomic_llong;
+typedef atomic<unsigned long long> atomic_ullong;
+typedef atomic<char16_t>           atomic_char16_t;
+typedef atomic<char32_t>           atomic_char32_t;
+typedef atomic<wchar_t>            atomic_wchar_t;
+
+typedef atomic<int_least8_t>   atomic_int_least8_t;
+typedef atomic<uint_least8_t>  atomic_uint_least8_t;
+typedef atomic<int_least16_t>  atomic_int_least16_t;
+typedef atomic<uint_least16_t> atomic_uint_least16_t;
+typedef atomic<int_least32_t>  atomic_int_least32_t;
+typedef atomic<uint_least32_t> atomic_uint_least32_t;
+typedef atomic<int_least64_t>  atomic_int_least64_t;
+typedef atomic<uint_least64_t> atomic_uint_least64_t;
+
+typedef atomic<int_fast8_t>   atomic_int_fast8_t;
+typedef atomic<uint_fast8_t>  atomic_uint_fast8_t;
+typedef atomic<int_fast16_t>  atomic_int_fast16_t;
+typedef atomic<uint_fast16_t> atomic_uint_fast16_t;
+typedef atomic<int_fast32_t>  atomic_int_fast32_t;
+typedef atomic<uint_fast32_t> atomic_uint_fast32_t;
+typedef atomic<int_fast64_t>  atomic_int_fast64_t;
+typedef atomic<uint_fast64_t> atomic_uint_fast64_t;
+
+typedef atomic< int8_t>  atomic_int8_t;
+typedef atomic<uint8_t>  atomic_uint8_t;
+typedef atomic< int16_t> atomic_int16_t;
+typedef atomic<uint16_t> atomic_uint16_t;
+typedef atomic< int32_t> atomic_int32_t;
+typedef atomic<uint32_t> atomic_uint32_t;
+typedef atomic< int64_t> atomic_int64_t;
+typedef atomic<uint64_t> atomic_uint64_t;
+
+typedef atomic<intptr_t>  atomic_intptr_t;
+typedef atomic<uintptr_t> atomic_uintptr_t;
+typedef atomic<size_t>    atomic_size_t;
+typedef atomic<ptrdiff_t> atomic_ptrdiff_t;
+typedef atomic<intmax_t>  atomic_intmax_t;
+typedef atomic<uintmax_t> atomic_uintmax_t;
+
+static_assert(ATOMIC_INT_LOCK_FREE, "This library assumes atomic<int> is lock-free.");
+
+typedef atomic<int>       atomic_signed_lock_free;
+typedef atomic<unsigned>  atomic_unsigned_lock_free;
+
+#define ATOMIC_FLAG_INIT {false}
+#define ATOMIC_VAR_INIT(__v) {__v}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX_ATOMIC
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/barrier b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/barrier
new file mode 100644
index 000000000000..762df1934d09
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/barrier
@@ -0,0 +1,384 @@
+// -*- C++ -*-
+//===--------------------------- barrier ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_BARRIER
+#define _LIBCUDACXX_BARRIER
+
+/*
+    barrier synopsis
+
+namespace std
+{
+
+  template<class CompletionFunction = see below>
+  class barrier
+  {
+  public:
+    using arrival_token = see below;
+
+    constexpr explicit barrier(ptrdiff_t phase_count,
+                               CompletionFunction f = CompletionFunction());
+    ~barrier();
+
+    barrier(const barrier&) = delete;
+    barrier& operator=(const barrier&) = delete;
+
+    [[nodiscard]] arrival_token arrive(ptrdiff_t update = 1);
+    void wait(arrival_token&& arrival) const;
+
+    void arrive_and_wait();
+    void arrive_and_drop();
+
+  private:
+    CompletionFunction __completion; // exposition only
+  };
+
+}
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__threading_support>
+#include <atomic>
+#include <cassert>
+#ifndef _LIBCUDACXX_HAS_NO_TREE_BARRIER
+#  include <thread>
+#  include <vector>
+#endif
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+# error <barrier> is not supported on this single threaded system
+#endif
+
+#if _LIBCUDACXX_STD_VER < 11
+# error <barrier> is requires C++11 or later
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+struct __empty_completion
+{
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()() noexcept { }
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_TREE_BARRIER
+
+template<class _CompletionF = __empty_completion, int _Sco = 0>
+class alignas(64) __barrier_base {
+
+    ptrdiff_t                       __expected;
+    __atomic_base<ptrdiff_t, _Sco>  __expected_adjustment;
+    _CompletionF                    __completion;
+
+    using __phase_t = uint8_t;
+    __atomic_base<__phase_t, _Sco>  __phase;
+
+    struct alignas(64) __state_t
+    {
+        struct {
+            __atomic_base<__phase_t, _Sco> __phase = ATOMIC_VAR_INIT(0);
+        } __tickets[64];
+    };
+    ::std::vector<__state_t>   __state;
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    bool __arrive(__phase_t const __old_phase)
+    {
+        __phase_t const __half_step = __old_phase + 1, __full_step = __old_phase + 2;
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_FAVORITE_BARRIER_INDEX
+        ptrdiff_t __current = __libcpp_thread_favorite_barrier_index,
+#else
+        ptrdiff_t __current = 0,
+#endif
+                  __current_expected = __expected,
+                  __last_node = (__current_expected >> 1);
+        for(size_t __round = 0;; ++__round) {
+            assert(__round <= 63);
+            if(__current_expected == 1)
+                return true;
+            for(;;++__current) {
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_FAVORITE_BARRIER_INDEX
+                if(0 == __round) {
+                    if(__current >= __current_expected)
+                        __current = 0;
+                    __libcpp_thread_favorite_barrier_index = __current;
+                }
+#endif
+                assert(__current <= __last_node);
+                __phase_t expect = __old_phase;
+                if(__current == __last_node && (__current_expected & 1))
+                {
+                    if(__state[__current].__tickets[__round].__phase.compare_exchange_strong(expect, __full_step, memory_order_acq_rel))
+                        break;    // I'm 1 in 1, go to next __round
+                    assert(expect == __full_step);
+                }
+                else if(__state[__current].__tickets[__round].__phase.compare_exchange_strong(expect, __half_step, memory_order_acq_rel))
+                {
+                    return false; // I'm 1 in 2, done with arrival
+                }
+                else if(expect == __half_step)
+                {
+                    if(__state[__current].__tickets[__round].__phase.compare_exchange_strong(expect, __full_step, memory_order_acq_rel))
+                        break;    // I'm 2 in 2, go to next __round
+                    assert(expect == __full_step);
+                }
+                assert(__round == 0 && expect == __full_step);
+            }
+            __current_expected = (__current_expected >> 1) + (__current_expected & 1);
+            __current &= ~( 1 << __round );
+            __last_node &= ~( 1 << __round );
+        }
+    }
+
+public:
+    using arrival_token = __phase_t;
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    __barrier_base(ptrdiff_t __expected, _CompletionF __completion = _CompletionF())
+            : __expected(__expected), __expected_adjustment(0), __completion(__completion),
+              __phase(0), __state((__expected+1) >> 1)
+    {
+        assert(__expected >= 0);
+    }
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    ~__barrier_base() = default;
+
+    __barrier_base(__barrier_base const&) = delete;
+    __barrier_base& operator=(__barrier_base const&) = delete;
+
+     _LIBCUDACXX_NODISCARD_ATTRIBUTE inline _LIBCUDACXX_INLINE_VISIBILITY
+    arrival_token arrive(ptrdiff_t update = 1)
+    {
+        assert(update > 0);
+        auto __old_phase = __phase.load(memory_order_relaxed);
+        for(; update; --update)
+            if(__arrive(__old_phase)) {
+                __completion();
+                __expected += __expected_adjustment.load(memory_order_relaxed);
+                __expected_adjustment.store(0, memory_order_relaxed);
+                __phase.store(__old_phase + 2, memory_order_release);
+            }
+        return __old_phase;
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void wait(arrival_token&& __old_phase) const
+    {
+        __libcpp_thread_poll_with_backoff([=]() -> bool {
+            return __phase.load(memory_order_acquire) != __old_phase;
+        });
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_wait()
+    {
+        wait(arrive());
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_drop()
+    {
+        __expected_adjustment.fetch_sub(1, memory_order_relaxed);
+        (void)arrive();
+    }
+};
+
+#else
+
+# if _LIBCUDACXX_CUDA_ABI_VERSION < 3
+#  define _LIBCUDACXX_BARRIER_ALIGNMENTS alignas(64)
+# else
+#  define _LIBCUDACXX_BARRIER_ALIGNMENTS
+# endif
+
+template<class _CompletionF, int _Sco = 0>
+class __barrier_base {
+
+    _LIBCUDACXX_BARRIER_ALIGNMENTS __atomic_base<ptrdiff_t, _Sco> __expected, __arrived;
+    _LIBCUDACXX_BARRIER_ALIGNMENTS _CompletionF                   __completion;
+    _LIBCUDACXX_BARRIER_ALIGNMENTS __atomic_base<bool, _Sco>      __phase;
+public:
+    using arrival_token = bool;
+
+    __barrier_base() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __barrier_base(ptrdiff_t __expected, _CompletionF __completion = _CompletionF())
+        : __phase(false), __expected(__expected), __arrived(__expected), __completion(__completion)
+    {
+    }
+
+    ~__barrier_base() = default;
+
+    __barrier_base(__barrier_base const&) = delete;
+    __barrier_base& operator=(__barrier_base const&) = delete;
+
+    _LIBCUDACXX_NODISCARD_ATTRIBUTE _LIBCUDACXX_INLINE_VISIBILITY
+    arrival_token arrive(ptrdiff_t __update = 1)
+    {
+        auto const __old_phase = __phase.load(memory_order_relaxed);
+        auto const __result = __arrived.fetch_sub(__update, memory_order_acq_rel) - __update;
+        assert(__result >= 0);
+        auto const __new_expected = __expected.load(memory_order_relaxed);
+        if(0 == __result) {
+            __completion();
+            __arrived.store(__new_expected, memory_order_relaxed);
+            __phase.store(!__old_phase, memory_order_release);
+            __cxx_atomic_notify_all(&__phase.__a_);
+        }
+        return __old_phase;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __try_wait(arrival_token __old_phase) const
+    {
+        return __phase != __old_phase;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait(arrival_token&& __old_phase) const
+    {
+        __phase.wait(__old_phase, memory_order_acquire);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_wait()
+    {
+        wait(arrive());
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_drop()
+    {
+        __expected.fetch_sub(1, memory_order_relaxed);
+        (void)arrive();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr ptrdiff_t max() noexcept
+    {
+        return numeric_limits<ptrdiff_t>::max();
+    }
+};
+
+template<int _Sco>
+class __barrier_base<__empty_completion, _Sco> {
+
+    static constexpr uint64_t __expected_unit = 1ull;
+    static constexpr uint64_t __arrived_unit = 1ull << 32;
+    static constexpr uint64_t __expected_mask = __arrived_unit - 1;
+    static constexpr uint64_t __phase_bit = 1ull << 63;
+    static constexpr uint64_t __arrived_mask = (__phase_bit - 1) & ~__expected_mask;
+
+    _LIBCUDACXX_BARRIER_ALIGNMENTS __atomic_base<uint64_t, _Sco> __phase_arrived_expected;
+
+public:
+    using arrival_token = uint64_t;
+
+private:
+    struct __poll_tester {
+        __barrier_base const* __this;
+        arrival_token __phase;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __poll_tester(__barrier_base const* __this_, arrival_token&& __phase_)
+          : __this(__this_)
+          , __phase(_CUDA_VSTD::move(__phase_))
+        {}
+
+        inline _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator()() const
+        {
+            return __this->__try_wait(__phase);
+        }
+    };
+
+    static inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    uint64_t __init(ptrdiff_t __count) _NOEXCEPT
+    {
+        return (((1u << 31) - __count) << 32)
+              | ((1u << 31) - __count);
+    }
+
+public:
+    __barrier_base() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __barrier_base(ptrdiff_t __count, __empty_completion = __empty_completion())
+        : __phase_arrived_expected(__init(__count)) {
+    }
+
+    ~__barrier_base() = default;
+
+    __barrier_base(__barrier_base const&) = delete;
+    __barrier_base& operator=(__barrier_base const&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __try_wait(arrival_token __phase) const
+    {
+        uint64_t const __current = __phase_arrived_expected.load(memory_order_acquire);
+        return ((__current & __phase_bit) != __phase);
+    }
+
+    _LIBCUDACXX_NODISCARD_ATTRIBUTE inline _LIBCUDACXX_INLINE_VISIBILITY
+    arrival_token arrive(ptrdiff_t __update = 1)
+    {
+        auto const __inc = __arrived_unit * __update;
+        auto const __old = __phase_arrived_expected.fetch_add(__inc, memory_order_acq_rel);
+        if((__old ^ (__old + __inc)) & __phase_bit) {
+            __phase_arrived_expected.fetch_add((__old & __expected_mask) << 32, memory_order_relaxed);
+            __phase_arrived_expected.notify_all();
+        }
+        return __old & __phase_bit;
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void wait(arrival_token&& __phase) const
+    {
+        __libcpp_thread_poll_with_backoff(__poll_tester(this, _CUDA_VSTD::move(__phase)));
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_wait()
+    {
+        wait(arrive());
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_drop()
+    {
+        __phase_arrived_expected.fetch_add(__expected_unit, memory_order_relaxed);
+        (void)arrive();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr ptrdiff_t max() noexcept
+    {
+        return numeric_limits<int32_t>::max();
+    }
+};
+
+#endif //_LIBCUDACXX_HAS_NO_TREE_BARRIER
+
+template<class _CompletionF = __empty_completion>
+class barrier : public __barrier_base<_CompletionF> {
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    barrier(ptrdiff_t __count, _CompletionF __completion = _CompletionF())
+        : __barrier_base<_CompletionF>(__count, __completion) {
+    }
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif //_LIBCUDACXX_BARRIER
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bit b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bit
new file mode 100644
index 000000000000..262282e3e7ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bit
@@ -0,0 +1,895 @@
+// -*- C++ -*-
+//===------------------------------ bit ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_BIT
+#define _LIBCUDACXX_BIT
+
+/*
+    bit synopsis
+
+namespace std {
+
+  template <class T>
+    constexpr bool has_single_bit(T x) noexcept; // C++20
+  template <class T>
+    constexpr T bit_ceil(T x);                   // C++20
+  template <class T>
+    constexpr T bit_floor(T x) noexcept;         // C++20
+  template <class T>
+    constexpr T bit_width(T x) noexcept;         // C++20
+
+  // 23.20.2, rotating
+  template<class T>
+    constexpr T rotl(T x, unsigned int s) noexcept; // C++20
+  template<class T>
+    constexpr T rotr(T x, unsigned int s) noexcept; // C++20
+
+  // 23.20.3, counting
+  template<class T>
+    constexpr int countl_zero(T x) noexcept;  // C++20
+  template<class T>
+    constexpr int countl_one(T x) noexcept;   // C++20
+  template<class T>
+    constexpr int countr_zero(T x) noexcept;  // C++20
+  template<class T>
+    constexpr int countr_one(T x) noexcept;   // C++20
+  template<class T>
+    constexpr int popcount(T x) noexcept;     // C++20
+
+  // 20.15.9, endian
+  enum class endian {
+    little = see below,        // C++20
+    big = see below,           // C++20
+    native = see below         // C++20
+};
+
+} // namespace std
+
+*/
+
+#ifndef __cuda_std__
+
+#include <__config>
+#include <cstdint>
+#include <limits>
+#include <type_traits>
+#include <version>
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+#endif // __cuda_std__
+
+#if defined(__IBMCPP__)
+#include "support/ibm/support.h"
+#endif
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
+#include <intrin.h>
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#define _LIBCUDACXX_BIT_CONSTEXPR constexpr
+
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_ctz2(uint64_t __x, int __c)           _NOEXCEPT {
+  return (__x & 0x1) ? __c : __c + 1;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_ctz4(uint64_t __x, int __c)           _NOEXCEPT {
+  return __binary_ctz2(
+            __x >> 2*!(__x & 0x3),
+            __c + 2*!(__x & 0x3));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_ctz8(uint64_t __x, int __c)           _NOEXCEPT {
+  return __binary_ctz4(
+            __x >> 4*!(__x & 0x0F),
+            __c + 4*!(__x & 0x0F));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_ctz16(uint64_t __x, int __c)           _NOEXCEPT {
+  return __binary_ctz8(
+            __x >> 8*!(__x & 0x00FF),
+            __c + 8*!(__x & 0x00FF));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_ctz32(uint64_t __x, int __c)           _NOEXCEPT {
+  return __binary_ctz16(
+            __x >> 16*!(__x & 0x0000FFFF),
+            __c + 16*!(__x & 0x0000FFFF));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_ctz64(uint64_t __x)           _NOEXCEPT {
+  return __binary_ctz32(
+            __x >> 32*!(__x & 0x00000000FFFFFFFF),
+            32*!(__x & 0x00000000FFFFFFFF));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_clz2(uint64_t __x, int __c) {
+  return !!(~__x & 0x2) ^ __c;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_clz4(uint64_t __x, int __c) {
+  return __binary_clz2(
+      __x >> 2*!!(__x & 0xC),
+      __c + 2*!(__x & 0xC));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_clz8(uint64_t __x, int __c) {
+  return __binary_clz4(
+      __x >> 4*!!(__x & 0xF0),
+      __c + 4*!(__x & 0xF0));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_clz16(uint64_t __x, int __c) {
+  return __binary_clz8(
+      __x >> 8*!!(__x & 0xFF00),
+      __c + 8*!(__x & 0xFF00));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_clz32(uint64_t __x, int __c) {
+  return __binary_clz16(
+      __x >> 16*!!(__x & 0xFFFF0000),
+      __c + 16*!(__x & 0xFFFF0000));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __binary_clz64(uint64_t __x) {
+  return __binary_clz32(
+      __x >> 32*!!(__x & 0xFFFFFFFF00000000),
+      32*!(__x & 0xFFFFFFFF00000000));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __fallback_popc8(uint64_t __x) {
+  return static_cast<int>((__x * 0x0101010101010101) >> 56);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __fallback_popc16(uint64_t __x) {
+  return __fallback_popc8(
+      __x + (__x >> 4) & 0x0f0f0f0f0f0f0f0f);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __fallback_popc32(uint64_t __x) {
+  return __fallback_popc16(
+      (__x & 0x3333333333333333) + ((__x >> 2) & 0x3333333333333333));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __fallback_popc64(uint64_t __x) {
+  return __fallback_popc32(
+      __x - ((__x >> 1) & 0x5555555555555555));
+}
+
+#ifndef _LIBCUDACXX_COMPILER_MSVC
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_ctz(unsigned __x)           _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_ctz32(static_cast<uint64_t>(__x), 0);
+  }
+  else {
+    return (!__x) ? 32 : __ffs(__x) - 1;
+  }
+#  else
+  return __binary_ctz32(static_cast<uint64_t>(__x), 0);
+#  endif
+#else
+  return __builtin_ctz(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_ctz(unsigned long __x)      _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_ctz64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return (!__x) ? 32 : __ffsll(__x) - 1;
+  }
+#  else
+  return __binary_ctz64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_ctzl(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_ctz(unsigned long long __x) _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_ctz64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return (!__x) ? 32 : __ffsll(__x) - 1;
+  }
+#  else
+  return __binary_ctz64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_ctzll(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_clz(unsigned __x)           _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_clz32(static_cast<uint64_t>(__x), 0);
+  }
+  else {
+    return __clz(__x);
+  }
+#  else
+  return __binary_clz32(static_cast<uint64_t>(__x), 0);
+#  endif
+#else
+  return __builtin_clz(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_clz(unsigned long __x)      _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_clz64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return __clzll(__x);
+  }
+#  else
+  return __binary_clz64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_clzl(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_clz(unsigned long long __x) _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_clz64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return __clzll(__x);
+  }
+#  else
+  return __binary_clz64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_clzll(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_popcount(unsigned __x)           _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return __popc(__x);
+  }
+#  else
+  return __fallback_popc64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_popcount(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_popcount(unsigned long __x)      _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return __popcll(__x);
+  }
+#  else
+  return __fallback_popc64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_popcountl(__x);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __libcpp_popcount(unsigned long long __x) _NOEXCEPT {
+#if defined(__CUDA_ARCH__)
+#  if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) && (_LIBCUDACXX_STD_VER >= 14)
+  if (__libcpp_is_constant_evaluated()) {
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return __popcll(__x);
+  }
+#  else
+  return __fallback_popc64(static_cast<uint64_t>(__x));
+#  endif
+#else
+  return __builtin_popcountl(__x);
+#endif
+}
+
+#else  // _LIBCUDACXX_COMPILER_MSVC
+
+// Precondition:  __x != 0
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_ctz(unsigned __x) {
+  static_assert(sizeof(unsigned) == sizeof(unsigned long), "");
+  static_assert(sizeof(unsigned long) == 4, "");
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __binary_ctz32(static_cast<uint64_t>(__x), 0);
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_ctz32(static_cast<uint64_t>(__x), 0);
+  }
+  else {
+    unsigned long __where = 0;
+    if (_BitScanForward(&__where, __x))
+      return static_cast<int>(__where);
+    return 32;
+  }
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_ctz(unsigned long __x) {
+    static_assert(sizeof(unsigned long) == sizeof(unsigned), "");
+    return __libcpp_ctz(static_cast<unsigned>(__x));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_ctz(unsigned long long __x) {
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __binary_ctz64(__x);
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_ctz64(__x);
+  }
+  else {
+    unsigned long __where = 0;
+#  if defined(_LIBCUDACXX_HAS_BITSCAN64) && (defined(_M_AMD64) || defined(__x86_64__))
+    if (_BitScanForward64(&__where, __x))
+      return static_cast<int>(__where);
+#  else
+    // Win32 doesn't have _BitScanForward64 so emulate it with two 32 bit calls.
+    if (_BitScanForward(&__where, static_cast<unsigned long>(__x)))
+      return static_cast<int>(__where);
+    if (_BitScanForward(&__where, static_cast<unsigned long>(__x >> 32)))
+      return static_cast<int>(__where + 32);
+#  endif
+    return 64;
+  }
+#endif
+}
+
+// Precondition:  __x != 0
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_clz(unsigned __x) {
+  static_assert(sizeof(unsigned) == sizeof(unsigned long), "");
+  static_assert(sizeof(unsigned long) == 4, "");
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __binary_clz32(static_cast<uint64_t>(__x), 0);
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_clz32(static_cast<uint64_t>(__x), 0);
+  }
+  else {
+    unsigned long __where = 0;
+    if (_BitScanReverse(&__where, __x))
+      return static_cast<int>(31 - __where);
+    return 32; // Undefined Behavior.
+  }
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_clz(unsigned long __x) {
+    static_assert(sizeof(unsigned) == sizeof(unsigned long), "");
+    return __libcpp_clz(static_cast<unsigned>(__x));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_clz(unsigned long long __x) {
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __binary_clz64(static_cast<uint64_t>(__x));
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __binary_clz64(static_cast<uint64_t>(__x));
+  }
+  else {
+    unsigned long __where = 0;
+#  if defined(_LIBCUDACXX_HAS_BITSCAN64)
+    if (_BitScanReverse64(&__where, __x))
+      return static_cast<int>(63 - __where);
+#  else
+    // Win32 doesn't have _BitScanReverse64 so emulate it with two 32 bit calls.
+    if (_BitScanReverse(&__where, static_cast<unsigned long>(__x >> 32)))
+      return static_cast<int>(63 - (__where + 32));
+    if (_BitScanReverse(&__where, static_cast<unsigned long>(__x)))
+      return static_cast<int>(63 - __where);
+#  endif
+    return 64; // Undefined Behavior.
+  }
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_popcount(unsigned __x) {
+  static_assert(sizeof(unsigned) == 4, "");
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return static_cast<int>(__popcnt(__x));
+  }
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_popcount(unsigned long __x) {
+  static_assert(sizeof(unsigned long) == 4, "");
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return static_cast<int>(__popcnt(__x));
+  }
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr int __libcpp_popcount(unsigned long long __x) {
+  static_assert(sizeof(unsigned long long) == 8, "");
+#if !defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED) || defined(__CUDA_ARCH__)
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+#else
+  if (__libcpp_is_constant_evaluated()) {
+    return __fallback_popc64(static_cast<uint64_t>(__x));
+  }
+  else {
+    return static_cast<int>(__popcnt64(__x));
+  }
+#endif
+}
+
+#endif // _LIBCUDACXX_COMPILER_MSVC
+
+template <class _Tp>
+using __bitop_unsigned_integer _LIBCUDACXX_NODEBUG_TYPE = integral_constant<bool,
+         is_integral<_Tp>::value &&
+         is_unsigned<_Tp>::value &&
+        _IsNotSame<typename remove_cv<_Tp>::type, bool>::value &&
+        _IsNotSame<typename remove_cv<_Tp>::type, signed char>::value &&
+        _IsNotSame<typename remove_cv<_Tp>::type, wchar_t>::value &&
+        _IsNotSame<typename remove_cv<_Tp>::type, char16_t>::value &&
+        _IsNotSame<typename remove_cv<_Tp>::type, char32_t>::value
+    >;
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_Tp __rotl(_Tp __t, unsigned int __cnt) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__rotl requires unsigned");
+    using __nlt = numeric_limits<_Tp>;
+
+    return ((__cnt % __nlt::digits) == 0) ?
+      __t :
+      (__t << (__cnt % __nlt::digits)) | (__t >> (__nlt::digits - (__cnt % __nlt::digits)));
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_Tp __rotr(_Tp __t, unsigned int __cnt) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__rotr requires unsigned");
+    using __nlt = numeric_limits<_Tp>;
+
+    return ((__cnt % __nlt::digits) == 0) ?
+        __t :
+        (__t >> (__cnt % __nlt::digits)) | (__t << (__nlt::digits - (__cnt % __nlt::digits)));
+}
+
+// Forward decl for recursive use in split word operations
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __countr_zero(_Tp __t) _NOEXCEPT;
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) <= sizeof(unsigned int), int> __countr_zero_dispatch(_Tp __t) _NOEXCEPT
+{
+  return __libcpp_ctz(static_cast<unsigned int>(__t));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) == sizeof(unsigned long long), int> __countr_zero_dispatch(_Tp __t) _NOEXCEPT
+{
+  return __libcpp_ctz(static_cast<unsigned long long>(__t));
+}
+
+template <typename _Tp, int _St = sizeof(_Tp)/sizeof(unsigned long long)>
+struct __countr_zero_rsh_impl {
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __short_circuit(_Tp __t, int __cur, int __count) {
+      // Stops processing early if non-zero
+      return (__cur == numeric_limits<unsigned long long>::digits) ?
+            __countr_zero_rsh_impl<_Tp, _St-1>::__count(__t, __cur + __count) :
+            __cur + __count;
+    }
+
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __count(_Tp __t, int __count) {
+        return __short_circuit(
+            __t >> numeric_limits<unsigned long long>::digits,
+            __countr_zero(static_cast<unsigned long long>(__t)),
+            __count);
+    }
+};
+
+template <typename _Tp>
+struct __countr_zero_rsh_impl<_Tp, 1> {
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __count(_Tp __t, int __count) {
+        return __count + __countr_zero(static_cast<unsigned long long>(__t));
+    }
+};
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<(sizeof(_Tp) > sizeof(unsigned long long)), int> __countr_zero_dispatch(_Tp __t) _NOEXCEPT
+{ return __countr_zero_rsh_impl<_Tp>::__count(__t, 0); }
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __countr_zero(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__countr_zero requires unsigned");
+
+    return __t ? __countr_zero_dispatch(__t) : numeric_limits<_Tp>::digits;
+}
+
+// Forward decl for recursive use in split word operations
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __countl_zero(_Tp __t) _NOEXCEPT;
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) <= sizeof(unsigned int), int> __countl_zero_dispatch(_Tp __t) _NOEXCEPT
+{
+  return __libcpp_clz(static_cast<unsigned int>(__t))
+              - (numeric_limits<unsigned int>::digits - numeric_limits<_Tp>::digits);
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) == sizeof(unsigned long long), int> __countl_zero_dispatch(_Tp __t) _NOEXCEPT
+{
+  return __libcpp_clz(static_cast<unsigned long long>(__t))
+              - (numeric_limits<unsigned long long>::digits - numeric_limits<_Tp>::digits);
+}
+
+template <typename _Tp, int _St = sizeof(_Tp)/sizeof(unsigned long long)>
+struct __countl_zero_rotl_impl {
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __short_circuit(_Tp __t, int __cur) {
+      // This stops processing early if the current word is not empty
+      return (__cur == numeric_limits<unsigned long long>::digits) ?
+            __cur + __countl_zero_rotl_impl<_Tp, _St-1>::__count(__t) :
+            __cur;
+    }
+
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __countl_iter(_Tp __t) {
+      // After rotating pass result of clz to another step for processing
+      return __short_circuit(
+            __t,
+            __countl_zero(static_cast<unsigned long long>(__t)));
+    }
+
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __count(_Tp __t) {
+        return __countl_iter(
+            __rotl(__t, numeric_limits<unsigned long long>::digits));
+    }
+};
+
+template <typename _Tp>
+struct __countl_zero_rotl_impl<_Tp, 1> {
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __count(_Tp __t) {
+        return __countl_zero(static_cast<unsigned long long>(__rotl(__t, numeric_limits<unsigned long long>::digits)));
+    }
+};
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<(sizeof(_Tp) > sizeof(unsigned long long)), int> __countl_zero_dispatch(_Tp __t) _NOEXCEPT
+{ return __countl_zero_rotl_impl<_Tp>::__count(__t); }
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __countl_zero(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__countl_zero requires unsigned");
+    return __t ? __countl_zero_dispatch(__t) : numeric_limits<_Tp>::digits;
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __countl_one(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__countl_one requires unsigned");
+    return __t != numeric_limits<_Tp>::max()
+        ? __countl_zero(static_cast<_Tp>(~__t))
+        : numeric_limits<_Tp>::digits;
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __countr_one(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__countr_one requires unsigned");
+    return __t != numeric_limits<_Tp>::max()
+        ? __countr_zero(static_cast<_Tp>(~__t))
+        : numeric_limits<_Tp>::digits;
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) <= sizeof(unsigned int), int> __popcount_dispatch(_Tp __t) _NOEXCEPT
+{ return __libcpp_popcount(static_cast<unsigned int>(__t)); }
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) == sizeof(unsigned long long), int> __popcount_dispatch(_Tp __t) _NOEXCEPT
+{ return __libcpp_popcount(static_cast<unsigned long long>(__t)); }
+
+template <typename _Tp, int _St = sizeof(_Tp)/sizeof(unsigned long long)>
+struct __popcount_rsh_impl {
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __count(_Tp __t) {
+        return __popcount_rsh_impl<_Tp, _St-1>::__count(
+                    __t >> numeric_limits<unsigned long long>::digits) +
+               __libcpp_popcount(static_cast<unsigned long long>(__t));
+    }
+};
+
+template <typename _Tp>
+struct __popcount_rsh_impl<_Tp, 1> {
+    static _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+    int __count(_Tp __t) {
+        return __libcpp_popcount(static_cast<unsigned long long>(__t));
+    }
+};
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<(sizeof(_Tp) > sizeof(unsigned long long)), int> __popcount_dispatch(_Tp __t) _NOEXCEPT
+{ return __popcount_rsh_impl<_Tp>::__count(__t); }
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+int __popcount(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__libcpp_popcount requires unsigned");
+
+    return __popcount_dispatch(__t);
+}
+
+
+// integral log base 2
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+unsigned __bit_log2(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__bit_log2 requires unsigned");
+    return std::numeric_limits<_Tp>::digits - 1 - __countl_zero(__t);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+bool __has_single_bit(_Tp __t) _NOEXCEPT
+{
+    static_assert(__bitop_unsigned_integer<_Tp>::value, "__has_single_bit requires unsigned");
+	return __t != 0 && (((__t & (__t - 1)) == 0));
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) >= sizeof(unsigned), _Tp>
+__ceil2(_Tp __t) noexcept
+{
+    // const unsigned __n = numeric_limits<_Tp>::digits - countl_zero((_Tp)(__t - 1u));
+    // _LIBCUDACXX_DEBUG_ASSERT(__libcpp_is_constant_evaluated() || __n != numeric_limits<_Tp>::digits, "Bad input to ceil2");
+    return _Tp{1} << (numeric_limits<_Tp>::digits - __countl_zero((_Tp)(__t - 1u)));
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<sizeof(_Tp) < sizeof(unsigned), _Tp>
+__ceil2(_Tp __t) noexcept
+{
+    // const unsigned __n = numeric_limits<_Tp>::digits - countl_zero((_Tp)(__t - 1u));
+    // _LIBCUDACXX_DEBUG_ASSERT(__libcpp_is_constant_evaluated() || __n != numeric_limits<_Tp>::digits, "Bad input to ceil2");
+
+    // const unsigned __extra = numeric_limits<unsigned>::digits  - numeric_limits<_Tp>::digits;
+    // const unsigned __retVal = 1u << (__n + __extra);
+    return (_Tp)
+        ((1u << ((numeric_limits<_Tp>::digits - __countl_zero((_Tp)(__t - 1u))) + (numeric_limits<unsigned>::digits  - numeric_limits<_Tp>::digits))) >>
+        (numeric_limits<unsigned>::digits  - numeric_limits<_Tp>::digits));
+}
+
+
+#if (_LIBCUDACXX_STD_VER > 17) || defined(_LIBCUDACXX_ENABLE_BIT_BACKPORT)
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, _Tp>
+rotl(_Tp __t, unsigned int __cnt) noexcept
+{
+    return __rotl(__t, __cnt);
+}
+
+
+// rotr
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, _Tp>
+rotr(_Tp __t, unsigned int __cnt) noexcept
+{
+    return __rotr(__t, __cnt);
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, int>
+countl_zero(_Tp __t) noexcept
+{
+    return __countl_zero(__t);
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, int>
+countl_one(_Tp __t) noexcept
+{
+    return __countl_one(__t);
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, int>
+countr_zero(_Tp __t) noexcept
+{
+	return __countr_zero(__t);
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, int>
+countr_one(_Tp __t) noexcept
+{
+    return __countr_one(__t);
+}
+
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, int>
+popcount(_Tp __t) noexcept
+{
+    return __popcount(__t);
+}
+
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, bool>
+has_single_bit(_Tp __t) noexcept
+{
+    return __has_single_bit(__t);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, _Tp>
+bit_floor(_Tp __t) noexcept
+{
+    return __t == 0 ? 0 : static_cast<_Tp>(_Tp{1} << __bit_log2(__t));
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, _Tp>
+bit_ceil(_Tp __t) noexcept
+{
+    return (__t < 2) ? 1 : static_cast<_Tp>(__ceil2(__t));
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_BIT_CONSTEXPR
+_EnableIf<__bitop_unsigned_integer<_Tp>::value, _Tp>
+bit_width(_Tp __t) noexcept
+{
+    return __t == 0 ? 0 : static_cast<_Tp>(__bit_log2(__t) + 1);
+}
+
+
+enum class endian
+{
+    little = 0xDEAD,
+    big    = 0xFACE,
+#if defined(_LIBCUDACXX_LITTLE_ENDIAN)
+    native = little
+#elif defined(_LIBCUDACXX_BIG_ENDIAN)
+    native = big
+#else
+    native = 0xCAFE
+#endif
+};
+
+#endif // _LIBCUDACXX_STD_VER > 17
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+_LIBCUDACXX_POP_MACROS
+#endif
+
+#endif // _LIBCUDACXX_BIT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bitset b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bitset
new file mode 100644
index 000000000000..bf529468219a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/bitset
@@ -0,0 +1,1109 @@
+// -*- C++ -*-
+//===---------------------------- bitset ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_BITSET
+#define _LIBCUDACXX_BITSET
+
+/*
+    bitset synopsis
+
+namespace std
+{
+
+namespace std {
+
+template <size_t N>
+class bitset
+{
+public:
+    // bit reference:
+    class reference
+    {
+        friend class bitset;
+        reference() noexcept;
+    public:
+        ~reference() noexcept;
+        reference& operator=(bool x) noexcept;           // for b[i] = x;
+        reference& operator=(const reference&) noexcept; // for b[i] = b[j];
+        bool operator~() const noexcept;                 // flips the bit
+        operator bool() const noexcept;                  // for x = b[i];
+        reference& flip() noexcept;                      // for b[i].flip();
+    };
+
+    // 23.3.5.1 constructors:
+    constexpr bitset() noexcept;
+    constexpr bitset(unsigned long long val) noexcept;
+    template <class charT>
+        explicit bitset(const charT* str,
+                        typename basic_string<charT>::size_type n = basic_string<charT>::npos,
+                        charT zero = charT('0'), charT one = charT('1'));
+    template<class charT, class traits, class Allocator>
+        explicit bitset(const basic_string<charT,traits,Allocator>& str,
+                        typename basic_string<charT,traits,Allocator>::size_type pos = 0,
+                        typename basic_string<charT,traits,Allocator>::size_type n =
+                                 basic_string<charT,traits,Allocator>::npos,
+                        charT zero = charT('0'), charT one = charT('1'));
+
+    // 23.3.5.2 bitset operations:
+    bitset& operator&=(const bitset& rhs) noexcept;
+    bitset& operator|=(const bitset& rhs) noexcept;
+    bitset& operator^=(const bitset& rhs) noexcept;
+    bitset& operator<<=(size_t pos) noexcept;
+    bitset& operator>>=(size_t pos) noexcept;
+    bitset& set() noexcept;
+    bitset& set(size_t pos, bool val = true);
+    bitset& reset() noexcept;
+    bitset& reset(size_t pos);
+    bitset operator~() const noexcept;
+    bitset& flip() noexcept;
+    bitset& flip(size_t pos);
+
+    // element access:
+    constexpr bool operator[](size_t pos) const; // for b[i];
+    reference operator[](size_t pos);            // for b[i];
+    unsigned long to_ulong() const;
+    unsigned long long to_ullong() const;
+    template <class charT, class traits, class Allocator>
+        basic_string<charT, traits, Allocator> to_string(charT zero = charT('0'), charT one = charT('1')) const;
+    template <class charT, class traits>
+        basic_string<charT, traits, allocator<charT> > to_string(charT zero = charT('0'), charT one = charT('1')) const;
+    template <class charT>
+        basic_string<charT, char_traits<charT>, allocator<charT> > to_string(charT zero = charT('0'), charT one = charT('1')) const;
+    basic_string<char, char_traits<char>, allocator<char> > to_string(char zero = '0', char one = '1') const;
+    size_t count() const noexcept;
+    constexpr size_t size() const noexcept;
+    bool operator==(const bitset& rhs) const noexcept;
+    bool operator!=(const bitset& rhs) const noexcept;
+    bool test(size_t pos) const;
+    bool all() const noexcept;
+    bool any() const noexcept;
+    bool none() const noexcept;
+    bitset operator<<(size_t pos) const noexcept;
+    bitset operator>>(size_t pos) const noexcept;
+};
+
+// 23.3.5.3 bitset operators:
+template <size_t N>
+bitset<N> operator&(const bitset<N>&, const bitset<N>&) noexcept;
+
+template <size_t N>
+bitset<N> operator|(const bitset<N>&, const bitset<N>&) noexcept;
+
+template <size_t N>
+bitset<N> operator^(const bitset<N>&, const bitset<N>&) noexcept;
+
+template <class charT, class traits, size_t N>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is, bitset<N>& x);
+
+template <class charT, class traits, size_t N>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os, const bitset<N>& x);
+
+template <size_t N> struct hash<std::bitset<N>>;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__bit_reference>
+#include <cstddef>
+#include <climits>
+#include <string>
+#include <stdexcept>
+#include <iosfwd>
+#include <__functional_base>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <size_t _N_words, size_t _Size>
+class __bitset;
+
+template <size_t _N_words, size_t _Size>
+struct __has_storage_type<__bitset<_N_words, _Size> >
+{
+    static const bool value = true;
+};
+
+template <size_t _N_words, size_t _Size>
+class __bitset
+{
+public:
+    typedef ptrdiff_t              difference_type;
+    typedef size_t                 size_type;
+    typedef size_type              __storage_type;
+protected:
+    typedef __bitset __self;
+    typedef       __storage_type*  __storage_pointer;
+    typedef const __storage_type*  __const_storage_pointer;
+    static const unsigned __bits_per_word = static_cast<unsigned>(sizeof(__storage_type) * CHAR_BIT);
+
+    friend class __bit_reference<__bitset>;
+    friend class __bit_const_reference<__bitset>;
+    friend class __bit_iterator<__bitset, false>;
+    friend class __bit_iterator<__bitset, true>;
+    friend struct __bit_array<__bitset>;
+
+    __storage_type __first_[_N_words];
+
+    typedef __bit_reference<__bitset>                  reference;
+    typedef __bit_const_reference<__bitset>            const_reference;
+    typedef __bit_iterator<__bitset, false>            iterator;
+    typedef __bit_iterator<__bitset, true>             const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR __bitset() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit _LIBCUDACXX_CONSTEXPR __bitset(unsigned long long __v) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference __make_ref(size_t __pos) _NOEXCEPT
+        {return reference(__first_ + __pos / __bits_per_word, __storage_type(1) << __pos % __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR const_reference __make_ref(size_t __pos) const _NOEXCEPT
+        {return const_reference(__first_ + __pos / __bits_per_word, __storage_type(1) << __pos % __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY iterator __make_iter(size_t __pos) _NOEXCEPT
+        {return iterator(__first_ + __pos / __bits_per_word, __pos % __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator __make_iter(size_t __pos) const _NOEXCEPT
+        {return const_iterator(__first_ + __pos / __bits_per_word, __pos % __bits_per_word);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator&=(const __bitset& __v) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator|=(const __bitset& __v) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator^=(const __bitset& __v) _NOEXCEPT;
+
+    void flip() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY unsigned long to_ulong() const
+        {return to_ulong(integral_constant<bool, _Size < sizeof(unsigned long) * CHAR_BIT>());}
+    _LIBCUDACXX_INLINE_VISIBILITY unsigned long long to_ullong() const
+        {return to_ullong(integral_constant<bool, _Size < sizeof(unsigned long long) * CHAR_BIT>());}
+
+    bool all() const _NOEXCEPT;
+    bool any() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t __hash_code() const _NOEXCEPT;
+private:
+#ifdef _LIBCUDACXX_CXX03_LANG
+    void __init(unsigned long long __v, false_type) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __init(unsigned long long __v, true_type) _NOEXCEPT;
+#endif  // _LIBCUDACXX_CXX03_LANG
+    unsigned long to_ulong(false_type) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long to_ulong(true_type) const;
+    unsigned long long to_ullong(false_type) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long long to_ullong(true_type) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long long to_ullong(true_type, false_type) const;
+    unsigned long long to_ullong(true_type, true_type) const;
+};
+
+template <size_t _N_words, size_t _Size>
+inline
+_LIBCUDACXX_CONSTEXPR
+__bitset<_N_words, _Size>::__bitset() _NOEXCEPT
+#ifndef _LIBCUDACXX_CXX03_LANG
+    : __first_{0}
+#endif
+{
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _CUDA_VSTD::fill_n(__first_, _N_words, __storage_type(0));
+#endif
+}
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+
+template <size_t _N_words, size_t _Size>
+void
+__bitset<_N_words, _Size>::__init(unsigned long long __v, false_type) _NOEXCEPT
+{
+    __storage_type __t[sizeof(unsigned long long) / sizeof(__storage_type)];
+    size_t __sz = _Size;
+    for (size_t __i = 0; __i < sizeof(__t)/sizeof(__t[0]); ++__i, __v >>= __bits_per_word, __sz -= __bits_per_word )
+        if ( __sz < __bits_per_word)
+            __t[__i] = static_cast<__storage_type>(__v) & ( 1ULL << __sz ) - 1;
+        else
+            __t[__i] = static_cast<__storage_type>(__v);
+
+    _CUDA_VSTD::copy(__t, __t + sizeof(__t)/sizeof(__t[0]), __first_);
+    _CUDA_VSTD::fill(__first_ + sizeof(__t)/sizeof(__t[0]), __first_ + sizeof(__first_)/sizeof(__first_[0]),
+               __storage_type(0));
+}
+
+template <size_t _N_words, size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__bitset<_N_words, _Size>::__init(unsigned long long __v, true_type) _NOEXCEPT
+{
+    __first_[0] = __v;
+    if (_Size < __bits_per_word)
+        __first_[0] &= ( 1ULL << _Size ) - 1;
+
+    _CUDA_VSTD::fill(__first_ + 1, __first_ + sizeof(__first_)/sizeof(__first_[0]), __storage_type(0));
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <size_t _N_words, size_t _Size>
+inline
+_LIBCUDACXX_CONSTEXPR
+__bitset<_N_words, _Size>::__bitset(unsigned long long __v) _NOEXCEPT
+#ifndef _LIBCUDACXX_CXX03_LANG
+#if __SIZEOF_SIZE_T__ == 8
+    : __first_{__v}
+#elif __SIZEOF_SIZE_T__ == 4
+    : __first_{static_cast<__storage_type>(__v),
+                _Size >= 2 * __bits_per_word ? static_cast<__storage_type>(__v >> __bits_per_word)
+                : static_cast<__storage_type>((__v >> __bits_per_word) & (__storage_type(1) << (_Size - __bits_per_word)) - 1)}
+#else
+#error This constructor has not been ported to this platform
+#endif
+#endif
+{
+#ifdef _LIBCUDACXX_CXX03_LANG
+    __init(__v, integral_constant<bool, sizeof(unsigned long long) == sizeof(__storage_type)>());
+#endif
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+void
+__bitset<_N_words, _Size>::operator&=(const __bitset& __v) _NOEXCEPT
+{
+    for (size_type __i = 0; __i < _N_words; ++__i)
+        __first_[__i] &= __v.__first_[__i];
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+void
+__bitset<_N_words, _Size>::operator|=(const __bitset& __v) _NOEXCEPT
+{
+    for (size_type __i = 0; __i < _N_words; ++__i)
+        __first_[__i] |= __v.__first_[__i];
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+void
+__bitset<_N_words, _Size>::operator^=(const __bitset& __v) _NOEXCEPT
+{
+    for (size_type __i = 0; __i < _N_words; ++__i)
+        __first_[__i] ^= __v.__first_[__i];
+}
+
+template <size_t _N_words, size_t _Size>
+void
+__bitset<_N_words, _Size>::flip() _NOEXCEPT
+{
+    // do middle whole words
+    size_type __n = _Size;
+    __storage_pointer __p = __first_;
+    for (; __n >= __bits_per_word; ++__p, __n -= __bits_per_word)
+        *__p = ~*__p;
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __storage_type __b = *__p & __m;
+        *__p &= ~__m;
+        *__p |= ~__b & __m;
+    }
+}
+
+template <size_t _N_words, size_t _Size>
+unsigned long
+__bitset<_N_words, _Size>::to_ulong(false_type) const
+{
+    const_iterator __e = __make_iter(_Size);
+    const_iterator __i = _CUDA_VSTD::find(__make_iter(sizeof(unsigned long) * CHAR_BIT), __e, true);
+    if (__i != __e)
+        __throw_overflow_error("bitset to_ulong overflow error");
+
+    return __first_[0];
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+unsigned long
+__bitset<_N_words, _Size>::to_ulong(true_type) const
+{
+    return __first_[0];
+}
+
+template <size_t _N_words, size_t _Size>
+unsigned long long
+__bitset<_N_words, _Size>::to_ullong(false_type) const
+{
+    const_iterator __e = __make_iter(_Size);
+    const_iterator __i = _CUDA_VSTD::find(__make_iter(sizeof(unsigned long long) * CHAR_BIT), __e, true);
+    if (__i != __e)
+        __throw_overflow_error("bitset to_ullong overflow error");
+
+    return to_ullong(true_type());
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+unsigned long long
+__bitset<_N_words, _Size>::to_ullong(true_type) const
+{
+    return to_ullong(true_type(), integral_constant<bool, sizeof(__storage_type) < sizeof(unsigned long long)>());
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+unsigned long long
+__bitset<_N_words, _Size>::to_ullong(true_type, false_type) const
+{
+    return __first_[0];
+}
+
+template <size_t _N_words, size_t _Size>
+unsigned long long
+__bitset<_N_words, _Size>::to_ullong(true_type, true_type) const
+{
+    unsigned long long __r = __first_[0];
+    for (std::size_t __i = 1; __i < sizeof(unsigned long long) / sizeof(__storage_type); ++__i)
+        __r |= static_cast<unsigned long long>(__first_[__i]) << (sizeof(__storage_type) * CHAR_BIT);
+    return __r;
+}
+
+template <size_t _N_words, size_t _Size>
+bool
+__bitset<_N_words, _Size>::all() const _NOEXCEPT
+{
+    // do middle whole words
+    size_type __n = _Size;
+    __const_storage_pointer __p = __first_;
+    for (; __n >= __bits_per_word; ++__p, __n -= __bits_per_word)
+        if (~*__p)
+            return false;
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        if (~*__p & __m)
+            return false;
+    }
+    return true;
+}
+
+template <size_t _N_words, size_t _Size>
+bool
+__bitset<_N_words, _Size>::any() const _NOEXCEPT
+{
+    // do middle whole words
+    size_type __n = _Size;
+    __const_storage_pointer __p = __first_;
+    for (; __n >= __bits_per_word; ++__p, __n -= __bits_per_word)
+        if (*__p)
+            return true;
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        if (*__p & __m)
+            return true;
+    }
+    return false;
+}
+
+template <size_t _N_words, size_t _Size>
+inline
+size_t
+__bitset<_N_words, _Size>::__hash_code() const _NOEXCEPT
+{
+    size_t __h = 0;
+    for (size_type __i = 0; __i < _N_words; ++__i)
+        __h ^= __first_[__i];
+    return __h;
+}
+
+template <size_t _Size>
+class __bitset<1, _Size>
+{
+public:
+    typedef ptrdiff_t              difference_type;
+    typedef size_t                 size_type;
+    typedef size_type              __storage_type;
+protected:
+    typedef __bitset __self;
+    typedef       __storage_type*  __storage_pointer;
+    typedef const __storage_type*  __const_storage_pointer;
+    static const unsigned __bits_per_word = static_cast<unsigned>(sizeof(__storage_type) * CHAR_BIT);
+
+    friend class __bit_reference<__bitset>;
+    friend class __bit_const_reference<__bitset>;
+    friend class __bit_iterator<__bitset, false>;
+    friend class __bit_iterator<__bitset, true>;
+    friend struct __bit_array<__bitset>;
+
+    __storage_type __first_;
+
+    typedef __bit_reference<__bitset>                  reference;
+    typedef __bit_const_reference<__bitset>            const_reference;
+    typedef __bit_iterator<__bitset, false>            iterator;
+    typedef __bit_iterator<__bitset, true>             const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR __bitset() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit _LIBCUDACXX_CONSTEXPR __bitset(unsigned long long __v) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference __make_ref(size_t __pos) _NOEXCEPT
+        {return reference(&__first_, __storage_type(1) << __pos);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR const_reference __make_ref(size_t __pos) const _NOEXCEPT
+        {return const_reference(&__first_, __storage_type(1) << __pos);}
+    _LIBCUDACXX_INLINE_VISIBILITY iterator __make_iter(size_t __pos) _NOEXCEPT
+        {return iterator(&__first_ + __pos / __bits_per_word, __pos % __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator __make_iter(size_t __pos) const _NOEXCEPT
+        {return const_iterator(&__first_ + __pos / __bits_per_word, __pos % __bits_per_word);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator&=(const __bitset& __v) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator|=(const __bitset& __v) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator^=(const __bitset& __v) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void flip() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long to_ulong() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long long to_ullong() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool all() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool any() const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t __hash_code() const _NOEXCEPT;
+};
+
+template <size_t _Size>
+inline
+_LIBCUDACXX_CONSTEXPR
+__bitset<1, _Size>::__bitset() _NOEXCEPT
+    : __first_(0)
+{
+}
+
+template <size_t _Size>
+inline
+_LIBCUDACXX_CONSTEXPR
+__bitset<1, _Size>::__bitset(unsigned long long __v) _NOEXCEPT
+    : __first_(
+        _Size == __bits_per_word ? static_cast<__storage_type>(__v)
+                                 : static_cast<__storage_type>(__v) & ((__storage_type(1) << _Size) - 1)
+    )
+{
+}
+
+template <size_t _Size>
+inline
+void
+__bitset<1, _Size>::operator&=(const __bitset& __v) _NOEXCEPT
+{
+    __first_ &= __v.__first_;
+}
+
+template <size_t _Size>
+inline
+void
+__bitset<1, _Size>::operator|=(const __bitset& __v) _NOEXCEPT
+{
+    __first_ |= __v.__first_;
+}
+
+template <size_t _Size>
+inline
+void
+__bitset<1, _Size>::operator^=(const __bitset& __v) _NOEXCEPT
+{
+    __first_ ^= __v.__first_;
+}
+
+template <size_t _Size>
+inline
+void
+__bitset<1, _Size>::flip() _NOEXCEPT
+{
+    __storage_type __m = ~__storage_type(0) >> (__bits_per_word - _Size);
+    __first_ = ~__first_;
+    __first_ &= __m;
+}
+
+template <size_t _Size>
+inline
+unsigned long
+__bitset<1, _Size>::to_ulong() const
+{
+    return __first_;
+}
+
+template <size_t _Size>
+inline
+unsigned long long
+__bitset<1, _Size>::to_ullong() const
+{
+    return __first_;
+}
+
+template <size_t _Size>
+inline
+bool
+__bitset<1, _Size>::all() const _NOEXCEPT
+{
+    __storage_type __m = ~__storage_type(0) >> (__bits_per_word - _Size);
+    return !(~__first_ & __m);
+}
+
+template <size_t _Size>
+inline
+bool
+__bitset<1, _Size>::any() const _NOEXCEPT
+{
+    __storage_type __m = ~__storage_type(0) >> (__bits_per_word - _Size);
+    return __first_ & __m;
+}
+
+template <size_t _Size>
+inline
+size_t
+__bitset<1, _Size>::__hash_code() const _NOEXCEPT
+{
+    return __first_;
+}
+
+template <>
+class __bitset<0, 0>
+{
+public:
+    typedef ptrdiff_t              difference_type;
+    typedef size_t                 size_type;
+    typedef size_type              __storage_type;
+protected:
+    typedef __bitset __self;
+    typedef       __storage_type*  __storage_pointer;
+    typedef const __storage_type*  __const_storage_pointer;
+    static const unsigned __bits_per_word = static_cast<unsigned>(sizeof(__storage_type) * CHAR_BIT);
+
+    friend class __bit_reference<__bitset>;
+    friend class __bit_const_reference<__bitset>;
+    friend class __bit_iterator<__bitset, false>;
+    friend class __bit_iterator<__bitset, true>;
+    friend struct __bit_array<__bitset>;
+
+    typedef __bit_reference<__bitset>                  reference;
+    typedef __bit_const_reference<__bitset>            const_reference;
+    typedef __bit_iterator<__bitset, false>            iterator;
+    typedef __bit_iterator<__bitset, true>             const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR __bitset() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit _LIBCUDACXX_CONSTEXPR __bitset(unsigned long long) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference __make_ref(size_t) _NOEXCEPT
+        {return reference(0, 1);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR const_reference __make_ref(size_t) const _NOEXCEPT
+        {return const_reference(0, 1);}
+    _LIBCUDACXX_INLINE_VISIBILITY iterator __make_iter(size_t) _NOEXCEPT
+        {return iterator(0, 0);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator __make_iter(size_t) const _NOEXCEPT
+        {return const_iterator(0, 0);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void operator&=(const __bitset&) _NOEXCEPT {}
+    _LIBCUDACXX_INLINE_VISIBILITY void operator|=(const __bitset&) _NOEXCEPT {}
+    _LIBCUDACXX_INLINE_VISIBILITY void operator^=(const __bitset&) _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void flip() _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY unsigned long to_ulong() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY unsigned long long to_ullong() const {return 0;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY bool all() const _NOEXCEPT {return true;}
+    _LIBCUDACXX_INLINE_VISIBILITY bool any() const _NOEXCEPT {return false;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY size_t __hash_code() const _NOEXCEPT {return 0;}
+};
+
+inline
+_LIBCUDACXX_CONSTEXPR
+__bitset<0, 0>::__bitset() _NOEXCEPT
+{
+}
+
+inline
+_LIBCUDACXX_CONSTEXPR
+__bitset<0, 0>::__bitset(unsigned long long) _NOEXCEPT
+{
+}
+
+template <size_t _Size> class _LIBCUDACXX_TEMPLATE_VIS bitset;
+template <size_t _Size> struct hash<bitset<_Size> >;
+
+template <size_t _Size>
+class _LIBCUDACXX_TEMPLATE_VIS bitset
+    : private __bitset<_Size == 0 ? 0 : (_Size - 1) / (sizeof(size_t) * CHAR_BIT) + 1, _Size>
+{
+public:
+    static const unsigned __n_words = _Size == 0 ? 0 : (_Size - 1) / (sizeof(size_t) * CHAR_BIT) + 1;
+    typedef __bitset<__n_words, _Size> base;
+
+public:
+    typedef typename base::reference       reference;
+    typedef typename base::const_reference const_reference;
+
+    // 23.3.5.1 constructors:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR bitset() _NOEXCEPT {}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+        bitset(unsigned long long __v) _NOEXCEPT : base(__v) {}
+    template<class _CharT, class = _EnableIf<_IsCharLikeType<_CharT>::value> >
+        explicit bitset(const _CharT* __str,
+                        typename basic_string<_CharT>::size_type __n = basic_string<_CharT>::npos,
+                        _CharT __zero = _CharT('0'), _CharT __one = _CharT('1'));
+    template<class _CharT, class _Traits, class _Allocator>
+        explicit bitset(const basic_string<_CharT,_Traits,_Allocator>& __str,
+                        typename basic_string<_CharT,_Traits,_Allocator>::size_type __pos = 0,
+                        typename basic_string<_CharT,_Traits,_Allocator>::size_type __n =
+                                (basic_string<_CharT,_Traits,_Allocator>::npos),
+                        _CharT __zero = _CharT('0'), _CharT __one = _CharT('1'));
+
+    // 23.3.5.2 bitset operations:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset& operator&=(const bitset& __rhs) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset& operator|=(const bitset& __rhs) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset& operator^=(const bitset& __rhs) _NOEXCEPT;
+    bitset& operator<<=(size_t __pos) _NOEXCEPT;
+    bitset& operator>>=(size_t __pos) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset& set() _NOEXCEPT;
+    bitset& set(size_t __pos, bool __val = true);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset& reset() _NOEXCEPT;
+    bitset& reset(size_t __pos);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset  operator~() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset& flip() _NOEXCEPT;
+    bitset& flip(size_t __pos);
+
+    // element access:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+                              const_reference operator[](size_t __p) const {return base::__make_ref(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY       reference operator[](size_t __p)       {return base::__make_ref(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long to_ulong() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned long long to_ullong() const;
+    template <class _CharT, class _Traits, class _Allocator>
+        basic_string<_CharT, _Traits, _Allocator> to_string(_CharT __zero = _CharT('0'),
+                                                            _CharT __one = _CharT('1')) const;
+    template <class _CharT, class _Traits>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string<_CharT, _Traits, allocator<_CharT> > to_string(_CharT __zero = _CharT('0'),
+                                                                    _CharT __one = _CharT('1')) const;
+    template <class _CharT>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string<_CharT, char_traits<_CharT>, allocator<_CharT> > to_string(_CharT __zero = _CharT('0'),
+                                                                                _CharT __one = _CharT('1')) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string<char, char_traits<char>, allocator<char> > to_string(char __zero = '0',
+                                                                      char __one = '1') const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t count() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR size_t size() const _NOEXCEPT {return _Size;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const bitset& __rhs) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const bitset& __rhs) const _NOEXCEPT;
+    bool test(size_t __pos) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool all() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool any() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY bool none() const _NOEXCEPT {return !any();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset operator<<(size_t __pos) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bitset operator>>(size_t __pos) const _NOEXCEPT;
+
+private:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t __hash_code() const _NOEXCEPT {return base::__hash_code();}
+
+    friend struct hash<bitset>;
+};
+
+template <size_t _Size>
+template<class _CharT, class>
+bitset<_Size>::bitset(const _CharT* __str,
+                      typename basic_string<_CharT>::size_type __n,
+                      _CharT __zero, _CharT __one)
+{
+    size_t __rlen = _CUDA_VSTD::min(__n, char_traits<_CharT>::length(__str));
+    for (size_t __i = 0; __i < __rlen; ++__i)
+        if (__str[__i] != __zero && __str[__i] != __one)
+            __throw_invalid_argument("bitset string ctor has invalid argument");
+
+    size_t _Mp = _CUDA_VSTD::min(__rlen, _Size);
+    size_t __i = 0;
+    for (; __i < _Mp; ++__i)
+    {
+        _CharT __c = __str[_Mp - 1 - __i];
+        if (__c == __zero)
+            (*this)[__i] = false;
+        else
+            (*this)[__i] = true;
+    }
+    _CUDA_VSTD::fill(base::__make_iter(__i), base::__make_iter(_Size), false);
+}
+
+template <size_t _Size>
+template<class _CharT, class _Traits, class _Allocator>
+bitset<_Size>::bitset(const basic_string<_CharT,_Traits,_Allocator>& __str,
+       typename basic_string<_CharT,_Traits,_Allocator>::size_type __pos,
+       typename basic_string<_CharT,_Traits,_Allocator>::size_type __n,
+       _CharT __zero, _CharT __one)
+{
+    if (__pos > __str.size())
+        __throw_out_of_range("bitset string pos out of range");
+
+    size_t __rlen = _CUDA_VSTD::min(__n, __str.size() - __pos);
+    for (size_t __i = __pos; __i < __pos + __rlen; ++__i)
+        if (!_Traits::eq(__str[__i], __zero) && !_Traits::eq(__str[__i], __one))
+            __throw_invalid_argument("bitset string ctor has invalid argument");
+
+    size_t _Mp = _CUDA_VSTD::min(__rlen, _Size);
+    size_t __i = 0;
+    for (; __i < _Mp; ++__i)
+    {
+        _CharT __c = __str[__pos + _Mp - 1 - __i];
+        if (_Traits::eq(__c, __zero))
+            (*this)[__i] = false;
+        else
+            (*this)[__i] = true;
+    }
+    _CUDA_VSTD::fill(base::__make_iter(__i), base::__make_iter(_Size), false);
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>&
+bitset<_Size>::operator&=(const bitset& __rhs) _NOEXCEPT
+{
+    base::operator&=(__rhs);
+    return *this;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>&
+bitset<_Size>::operator|=(const bitset& __rhs) _NOEXCEPT
+{
+    base::operator|=(__rhs);
+    return *this;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>&
+bitset<_Size>::operator^=(const bitset& __rhs) _NOEXCEPT
+{
+    base::operator^=(__rhs);
+    return *this;
+}
+
+template <size_t _Size>
+bitset<_Size>&
+bitset<_Size>::operator<<=(size_t __pos) _NOEXCEPT
+{
+    __pos = _CUDA_VSTD::min(__pos, _Size);
+    _CUDA_VSTD::copy_backward(base::__make_iter(0), base::__make_iter(_Size - __pos), base::__make_iter(_Size));
+    _CUDA_VSTD::fill_n(base::__make_iter(0), __pos, false);
+    return *this;
+}
+
+template <size_t _Size>
+bitset<_Size>&
+bitset<_Size>::operator>>=(size_t __pos) _NOEXCEPT
+{
+    __pos = _CUDA_VSTD::min(__pos, _Size);
+    _CUDA_VSTD::copy(base::__make_iter(__pos), base::__make_iter(_Size), base::__make_iter(0));
+    _CUDA_VSTD::fill_n(base::__make_iter(_Size - __pos), __pos, false);
+    return *this;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>&
+bitset<_Size>::set() _NOEXCEPT
+{
+    _CUDA_VSTD::fill_n(base::__make_iter(0), _Size, true);
+    return *this;
+}
+
+template <size_t _Size>
+bitset<_Size>&
+bitset<_Size>::set(size_t __pos, bool __val)
+{
+    if (__pos >= _Size)
+        __throw_out_of_range("bitset set argument out of range");
+
+    (*this)[__pos] = __val;
+    return *this;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>&
+bitset<_Size>::reset() _NOEXCEPT
+{
+    _CUDA_VSTD::fill_n(base::__make_iter(0), _Size, false);
+    return *this;
+}
+
+template <size_t _Size>
+bitset<_Size>&
+bitset<_Size>::reset(size_t __pos)
+{
+    if (__pos >= _Size)
+        __throw_out_of_range("bitset reset argument out of range");
+
+    (*this)[__pos] = false;
+    return *this;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>
+bitset<_Size>::operator~() const _NOEXCEPT
+{
+    bitset __x(*this);
+    __x.flip();
+    return __x;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>&
+bitset<_Size>::flip() _NOEXCEPT
+{
+    base::flip();
+    return *this;
+}
+
+template <size_t _Size>
+bitset<_Size>&
+bitset<_Size>::flip(size_t __pos)
+{
+    if (__pos >= _Size)
+        __throw_out_of_range("bitset flip argument out of range");
+
+    reference r = base::__make_ref(__pos);
+    r = ~r;
+    return *this;
+}
+
+template <size_t _Size>
+inline
+unsigned long
+bitset<_Size>::to_ulong() const
+{
+    return base::to_ulong();
+}
+
+template <size_t _Size>
+inline
+unsigned long long
+bitset<_Size>::to_ullong() const
+{
+    return base::to_ullong();
+}
+
+template <size_t _Size>
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+bitset<_Size>::to_string(_CharT __zero, _CharT __one) const
+{
+    basic_string<_CharT, _Traits, _Allocator> __r(_Size, __zero);
+    for (size_t __i = 0; __i < _Size; ++__i)
+    {
+        if ((*this)[__i])
+            __r[_Size - 1 - __i] = __one;
+    }
+    return __r;
+}
+
+template <size_t _Size>
+template <class _CharT, class _Traits>
+inline
+basic_string<_CharT, _Traits, allocator<_CharT> >
+bitset<_Size>::to_string(_CharT __zero, _CharT __one) const
+{
+    return to_string<_CharT, _Traits, allocator<_CharT> >(__zero, __one);
+}
+
+template <size_t _Size>
+template <class _CharT>
+inline
+basic_string<_CharT, char_traits<_CharT>, allocator<_CharT> >
+bitset<_Size>::to_string(_CharT __zero, _CharT __one) const
+{
+    return to_string<_CharT, char_traits<_CharT>, allocator<_CharT> >(__zero, __one);
+}
+
+template <size_t _Size>
+inline
+basic_string<char, char_traits<char>, allocator<char> >
+bitset<_Size>::to_string(char __zero, char __one) const
+{
+    return to_string<char, char_traits<char>, allocator<char> >(__zero, __one);
+}
+
+template <size_t _Size>
+inline
+size_t
+bitset<_Size>::count() const _NOEXCEPT
+{
+    return static_cast<size_t>(__count_bool_true(base::__make_iter(0), _Size));
+}
+
+template <size_t _Size>
+inline
+bool
+bitset<_Size>::operator==(const bitset& __rhs) const _NOEXCEPT
+{
+    return _CUDA_VSTD::equal(base::__make_iter(0), base::__make_iter(_Size), __rhs.__make_iter(0));
+}
+
+template <size_t _Size>
+inline
+bool
+bitset<_Size>::operator!=(const bitset& __rhs) const _NOEXCEPT
+{
+    return !(*this == __rhs);
+}
+
+template <size_t _Size>
+bool
+bitset<_Size>::test(size_t __pos) const
+{
+    if (__pos >= _Size)
+        __throw_out_of_range("bitset test argument out of range");
+
+    return (*this)[__pos];
+}
+
+template <size_t _Size>
+inline
+bool
+bitset<_Size>::all() const _NOEXCEPT
+{
+    return base::all();
+}
+
+template <size_t _Size>
+inline
+bool
+bitset<_Size>::any() const _NOEXCEPT
+{
+    return base::any();
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>
+bitset<_Size>::operator<<(size_t __pos) const _NOEXCEPT
+{
+    bitset __r = *this;
+    __r <<= __pos;
+    return __r;
+}
+
+template <size_t _Size>
+inline
+bitset<_Size>
+bitset<_Size>::operator>>(size_t __pos) const _NOEXCEPT
+{
+    bitset __r = *this;
+    __r >>= __pos;
+    return __r;
+}
+
+template <size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bitset<_Size>
+operator&(const bitset<_Size>& __x, const bitset<_Size>& __y) _NOEXCEPT
+{
+    bitset<_Size> __r = __x;
+    __r &= __y;
+    return __r;
+}
+
+template <size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bitset<_Size>
+operator|(const bitset<_Size>& __x, const bitset<_Size>& __y) _NOEXCEPT
+{
+    bitset<_Size> __r = __x;
+    __r |= __y;
+    return __r;
+}
+
+template <size_t _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bitset<_Size>
+operator^(const bitset<_Size>& __x, const bitset<_Size>& __y) _NOEXCEPT
+{
+    bitset<_Size> __r = __x;
+    __r ^= __y;
+    return __r;
+}
+
+template <size_t _Size>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<bitset<_Size> >
+    : public unary_function<bitset<_Size>, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const bitset<_Size>& __bs) const _NOEXCEPT
+        {return __bs.__hash_code();}
+};
+
+template <class _CharT, class _Traits, size_t _Size>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, bitset<_Size>& __x);
+
+template <class _CharT, class _Traits, size_t _Size>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const bitset<_Size>& __x);
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_BITSET
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cassert b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cassert
new file mode 100644
index 000000000000..d5b1302b1ec1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cassert
@@ -0,0 +1,25 @@
+// -*- C++ -*-
+//===-------------------------- cassert -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+/*
+    cassert synopsis
+
+Macros:
+
+    assert
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__pragma_push>
+#include <assert.h>
+#include <__pragma_pop>
+#endif //__cuda_std__
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex
new file mode 100644
index 000000000000..728af56ffafc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ccomplex
@@ -0,0 +1,28 @@
+// -*- C++ -*-
+//===--------------------------- ccomplex ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CCOMPLEX
+#define _LIBCUDACXX_CCOMPLEX
+
+/*
+    ccomplex synopsis
+
+#include <complex>
+
+*/
+
+#include <complex>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+// hh 080623 Created
+
+#endif  // _LIBCUDACXX_CCOMPLEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cctype b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cctype
new file mode 100644
index 000000000000..16173ec7674c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cctype
@@ -0,0 +1,120 @@
+// -*- C++ -*-
+//===---------------------------- cctype ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CCTYPE
+#define _LIBCUDACXX_CCTYPE
+
+/*
+    cctype synopsis
+
+namespace std
+{
+
+int isalnum(int c);
+int isalpha(int c);
+int isblank(int c);  // C99
+int iscntrl(int c);
+int isdigit(int c);
+int isgraph(int c);
+int islower(int c);
+int isprint(int c);
+int ispunct(int c);
+int isspace(int c);
+int isupper(int c);
+int isxdigit(int c);
+int tolower(int c);
+int toupper(int c);
+
+}  // std
+*/
+
+#include <__config>
+#include <ctype.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#ifdef isalnum
+#undef isalnum
+#endif
+
+#ifdef isalpha
+#undef isalpha
+#endif
+
+#ifdef isblank
+#undef isblank
+#endif
+
+#ifdef iscntrl
+#undef iscntrl
+#endif
+
+#ifdef isdigit
+#undef isdigit
+#endif
+
+#ifdef isgraph
+#undef isgraph
+#endif
+
+#ifdef islower
+#undef islower
+#endif
+
+#ifdef isprint
+#undef isprint
+#endif
+
+#ifdef ispunct
+#undef ispunct
+#endif
+
+#ifdef isspace
+#undef isspace
+#endif
+
+#ifdef isupper
+#undef isupper
+#endif
+
+#ifdef isxdigit
+#undef isxdigit
+#endif
+
+#ifdef tolower
+#undef tolower
+#endif
+
+#ifdef toupper
+#undef toupper
+#endif
+
+
+using ::isalnum;
+using ::isalpha;
+using ::isblank;
+using ::iscntrl;
+using ::isdigit;
+using ::isgraph;
+using ::islower;
+using ::isprint;
+using ::ispunct;
+using ::isspace;
+using ::isupper;
+using ::isxdigit;
+using ::tolower;
+using ::toupper;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CCTYPE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno
new file mode 100644
index 000000000000..717f0f481d3f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cerrno
@@ -0,0 +1,32 @@
+// -*- C++ -*-
+//===-------------------------- cerrno ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CERRNO
+#define _LIBCUDACXX_CERRNO
+
+/*
+    cerrno synopsis
+
+Macros:
+
+    EDOM
+    EILSEQ  // C99
+    ERANGE
+    errno
+
+*/
+
+#include <__config>
+#include <errno.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#endif  // _LIBCUDACXX_CERRNO
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv
new file mode 100644
index 000000000000..65e776ff64c6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfenv
@@ -0,0 +1,81 @@
+// -*- C++ -*-
+//===---------------------------- cfenv -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CFENV
+#define _LIBCUDACXX_CFENV
+
+/*
+    cfenv synopsis
+
+This entire header is C99 / C++0X
+
+Macros:
+
+    FE_DIVBYZERO
+    FE_INEXACT
+    FE_INVALID
+    FE_OVERFLOW
+    FE_UNDERFLOW
+    FE_ALL_EXCEPT
+    FE_DOWNWARD
+    FE_TONEAREST
+    FE_TOWARDZERO
+    FE_UPWARD
+    FE_DFL_ENV
+
+namespace std
+{
+
+Types:
+
+    fenv_t
+    fexcept_t
+
+int feclearexcept(int excepts);
+int fegetexceptflag(fexcept_t* flagp, int excepts);
+int feraiseexcept(int excepts);
+int fesetexceptflag(const fexcept_t* flagp, int excepts);
+int fetestexcept(int excepts);
+int fegetround();
+int fesetround(int round);
+int fegetenv(fenv_t* envp);
+int feholdexcept(fenv_t* envp);
+int fesetenv(const fenv_t* envp);
+int feupdateenv(const fenv_t* envp);
+
+}  // std
+*/
+
+#include <__config>
+#include <fenv.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::fenv_t;
+using ::fexcept_t;
+
+using ::feclearexcept;
+using ::fegetexceptflag;
+using ::feraiseexcept;
+using ::fesetexceptflag;
+using ::fetestexcept;
+using ::fegetround;
+using ::fesetround;
+using ::fegetenv;
+using ::feholdexcept;
+using ::fesetenv;
+using ::feupdateenv;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CFENV
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfloat b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfloat
new file mode 100644
index 000000000000..36be66f00c79
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cfloat
@@ -0,0 +1,79 @@
+// -*- C++ -*-
+//===--------------------------- cfloat -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CFLOAT
+#define _LIBCUDACXX_CFLOAT
+
+/*
+    cfloat synopsis
+
+Macros:
+
+    FLT_ROUNDS
+    FLT_EVAL_METHOD     // C99
+    FLT_RADIX
+
+    FLT_HAS_SUBNORM     // C11
+    DBL_HAS_SUBNORM     // C11
+    LDBL_HAS_SUBNORM    // C11
+
+    FLT_MANT_DIG
+    DBL_MANT_DIG
+    LDBL_MANT_DIG
+
+    DECIMAL_DIG         // C99
+    FLT_DECIMAL_DIG     // C11
+    DBL_DECIMAL_DIG     // C11
+    LDBL_DECIMAL_DIG    // C11
+
+    FLT_DIG
+    DBL_DIG
+    LDBL_DIG
+
+    FLT_MIN_EXP
+    DBL_MIN_EXP
+    LDBL_MIN_EXP
+
+    FLT_MIN_10_EXP
+    DBL_MIN_10_EXP
+    LDBL_MIN_10_EXP
+
+    FLT_MAX_EXP
+    DBL_MAX_EXP
+    LDBL_MAX_EXP
+
+    FLT_MAX_10_EXP
+    DBL_MAX_10_EXP
+    LDBL_MAX_10_EXP
+
+    FLT_MAX
+    DBL_MAX
+    LDBL_MAX
+
+    FLT_EPSILON
+    DBL_EPSILON
+    LDBL_EPSILON
+
+    FLT_MIN
+    DBL_MIN
+    LDBL_MIN
+
+    FLT_TRUE_MIN        // C11
+    DBL_TRUE_MIN        // C11
+    LDBL_TRUE_MIN       // C11
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__pragma_push>
+#include <float.h>
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CFLOAT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/charconv b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/charconv
new file mode 100644
index 000000000000..9e5d1dda58de
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/charconv
@@ -0,0 +1,616 @@
+// -*- C++ -*-
+//===------------------------------ charconv ------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CHARCONV
+#define _LIBCUDACXX_CHARCONV
+
+/*
+    charconv synopsis
+
+namespace std {
+
+  // floating-point format for primitive numerical conversion
+  enum class chars_format {
+    scientific = unspecified,
+    fixed = unspecified,
+    hex = unspecified,
+    general = fixed | scientific
+  };
+
+  // 23.20.2, primitive numerical output conversion
+  struct to_chars_result {
+    char* ptr;
+    errc ec;
+  };
+
+  to_chars_result to_chars(char* first, char* last, see below value,
+                           int base = 10);
+
+  to_chars_result to_chars(char* first, char* last, float value);
+  to_chars_result to_chars(char* first, char* last, double value);
+  to_chars_result to_chars(char* first, char* last, long double value);
+
+  to_chars_result to_chars(char* first, char* last, float value,
+                           chars_format fmt);
+  to_chars_result to_chars(char* first, char* last, double value,
+                           chars_format fmt);
+  to_chars_result to_chars(char* first, char* last, long double value,
+                           chars_format fmt);
+
+  to_chars_result to_chars(char* first, char* last, float value,
+                           chars_format fmt, int precision);
+  to_chars_result to_chars(char* first, char* last, double value,
+                           chars_format fmt, int precision);
+  to_chars_result to_chars(char* first, char* last, long double value,
+                           chars_format fmt, int precision);
+
+  // 23.20.3, primitive numerical input conversion
+  struct from_chars_result {
+    const char* ptr;
+    errc ec;
+  };
+
+  from_chars_result from_chars(const char* first, const char* last,
+                               see below& value, int base = 10);
+
+  from_chars_result from_chars(const char* first, const char* last,
+                               float& value,
+                               chars_format fmt = chars_format::general);
+  from_chars_result from_chars(const char* first, const char* last,
+                               double& value,
+                               chars_format fmt = chars_format::general);
+  from_chars_result from_chars(const char* first, const char* last,
+                               long double& value,
+                               chars_format fmt = chars_format::general);
+
+} // namespace std
+
+*/
+
+#include <__errc>
+#include <type_traits>
+#include <limits>
+#include <stdint.h>
+#include <string.h>
+#include <math.h>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+namespace __itoa {
+_LIBCUDACXX_FUNC_VIS char* __u64toa(uint64_t __value, char* __buffer);
+_LIBCUDACXX_FUNC_VIS char* __u32toa(uint32_t __value, char* __buffer);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+enum class _LIBCUDACXX_ENUM_VIS chars_format
+{
+    scientific = 0x1,
+    fixed = 0x2,
+    hex = 0x4,
+    general = fixed | scientific
+};
+
+struct _LIBCUDACXX_TYPE_VIS to_chars_result
+{
+    char* ptr;
+    errc ec;
+};
+
+struct _LIBCUDACXX_TYPE_VIS from_chars_result
+{
+    const char* ptr;
+    errc ec;
+};
+
+void to_chars(char*, char*, bool, int = 10) = delete;
+void from_chars(const char*, const char*, bool, int = 10) = delete;
+
+namespace __itoa
+{
+
+static _LIBCUDACXX_CONSTEXPR uint64_t __pow10_64[] = {
+    UINT64_C(0),
+    UINT64_C(10),
+    UINT64_C(100),
+    UINT64_C(1000),
+    UINT64_C(10000),
+    UINT64_C(100000),
+    UINT64_C(1000000),
+    UINT64_C(10000000),
+    UINT64_C(100000000),
+    UINT64_C(1000000000),
+    UINT64_C(10000000000),
+    UINT64_C(100000000000),
+    UINT64_C(1000000000000),
+    UINT64_C(10000000000000),
+    UINT64_C(100000000000000),
+    UINT64_C(1000000000000000),
+    UINT64_C(10000000000000000),
+    UINT64_C(100000000000000000),
+    UINT64_C(1000000000000000000),
+    UINT64_C(10000000000000000000),
+};
+
+static _LIBCUDACXX_CONSTEXPR uint32_t __pow10_32[] = {
+    UINT32_C(0),          UINT32_C(10),       UINT32_C(100),
+    UINT32_C(1000),       UINT32_C(10000),    UINT32_C(100000),
+    UINT32_C(1000000),    UINT32_C(10000000), UINT32_C(100000000),
+    UINT32_C(1000000000),
+};
+
+template <typename _Tp, typename = void>
+struct _LIBCUDACXX_HIDDEN __traits_base
+{
+    using type = uint64_t;
+
+#if !defined(_LIBCUDACXX_COMPILER_MSVC)
+    static _LIBCUDACXX_INLINE_VISIBILITY int __width(_Tp __v)
+    {
+        auto __t = (64 - __builtin_clzll(__v | 1)) * 1233 >> 12;
+        return __t - (__v < __pow10_64[__t]) + 1;
+    }
+#endif
+
+    static _LIBCUDACXX_INLINE_VISIBILITY char* __convert(_Tp __v, char* __p)
+    {
+        return __u64toa(__v, __p);
+    }
+
+    static _LIBCUDACXX_INLINE_VISIBILITY decltype(__pow10_64)& __pow() { return __pow10_64; }
+};
+
+template <typename _Tp>
+struct _LIBCUDACXX_HIDDEN
+    __traits_base<_Tp, decltype(void(uint32_t{declval<_Tp>()}))>
+{
+    using type = uint32_t;
+
+#if !defined(_LIBCUDACXX_COMPILER_MSVC)
+    static _LIBCUDACXX_INLINE_VISIBILITY int __width(_Tp __v)
+    {
+        auto __t = (32 - __builtin_clz(__v | 1)) * 1233 >> 12;
+        return __t - (__v < __pow10_32[__t]) + 1;
+    }
+#endif
+
+    static _LIBCUDACXX_INLINE_VISIBILITY char* __convert(_Tp __v, char* __p)
+    {
+        return __u32toa(__v, __p);
+    }
+
+    static _LIBCUDACXX_INLINE_VISIBILITY decltype(__pow10_32)& __pow() { return __pow10_32; }
+};
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+__mul_overflowed(unsigned char __a, _Tp __b, unsigned char& __r)
+{
+    auto __c = __a * __b;
+    __r = __c;
+    return __c > (numeric_limits<unsigned char>::max)();
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+__mul_overflowed(unsigned short __a, _Tp __b, unsigned short& __r)
+{
+    auto __c = __a * __b;
+    __r = __c;
+    return __c > (numeric_limits<unsigned short>::max)();
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+__mul_overflowed(_Tp __a, _Tp __b, _Tp& __r)
+{
+    static_assert(is_unsigned<_Tp>::value, "");
+#if !defined(_LIBCUDACXX_COMPILER_MSVC) && !defined(_LIBCUDACXX_COMPILER_PGI)
+    return __builtin_mul_overflow(__a, __b, &__r);
+#else
+    bool __did = __b && ((numeric_limits<_Tp>::max)() / __b) < __a;
+    __r = __a * __b;
+    return __did;
+#endif
+}
+
+template <typename _Tp, typename _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+__mul_overflowed(_Tp __a, _Up __b, _Tp& __r)
+{
+    return __mul_overflowed(__a, static_cast<_Tp>(__b), __r);
+}
+
+template <typename _Tp>
+struct _LIBCUDACXX_HIDDEN __traits : __traits_base<_Tp>
+{
+    static _LIBCUDACXX_CONSTEXPR int digits = numeric_limits<_Tp>::digits10 + 1;
+    using __traits_base<_Tp>::__pow;
+    using typename __traits_base<_Tp>::type;
+
+    // precondition: at least one non-zero character available
+    static _LIBCUDACXX_INLINE_VISIBILITY char const*
+    __read(char const* __p, char const* __ep, type& __a, type& __b)
+    {
+        type __cprod[digits];
+        int __j = digits - 1;
+        int __i = digits;
+        do
+        {
+            if (!('0' <= *__p && *__p <= '9'))
+                break;
+            __cprod[--__i] = *__p++ - '0';
+        } while (__p != __ep && __i != 0);
+
+        __a = __inner_product(__cprod + __i + 1, __cprod + __j, __pow() + 1,
+                              __cprod[__i]);
+        if (__mul_overflowed(__cprod[__j], __pow()[__j - __i], __b))
+            --__p;
+        return __p;
+    }
+
+    template <typename _It1, typename _It2, class _Up>
+    static _LIBCUDACXX_INLINE_VISIBILITY _Up
+    __inner_product(_It1 __first1, _It1 __last1, _It2 __first2, _Up __init)
+    {
+        for (; __first1 < __last1; ++__first1, ++__first2)
+            __init = __init + *__first1 * *__first2;
+        return __init;
+    }
+};
+
+}  // namespace __itoa
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _Tp
+__complement(_Tp __x)
+{
+    static_assert(is_unsigned<_Tp>::value, "cast to unsigned first");
+    return _Tp(~__x + 1);
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY typename make_unsigned<_Tp>::type
+__to_unsigned(_Tp __x)
+{
+    return static_cast<typename make_unsigned<_Tp>::type>(__x);
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY to_chars_result
+__to_chars_itoa(char* __first, char* __last, _Tp __value, true_type)
+{
+    auto __x = __to_unsigned(__value);
+    if (__value < 0 && __first != __last)
+    {
+        *__first++ = '-';
+        __x = __complement(__x);
+    }
+
+    return __to_chars_itoa(__first, __last, __x, false_type());
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY to_chars_result
+__to_chars_itoa(char* __first, char* __last, _Tp __value, false_type)
+{
+    using __tx = __itoa::__traits<_Tp>;
+    auto __diff = __last - __first;
+
+#if !defined(_LIBCUDACXX_COMPILER_MSVC)
+    if (__tx::digits <= __diff || __tx::__width(__value) <= __diff)
+        return {__tx::__convert(__value, __first), errc(0)};
+    else
+        return {__last, errc::value_too_large};
+#else
+    if (__tx::digits <= __diff)
+        return {__tx::__convert(__value, __first), {}};
+    else
+    {
+        char __buf[__tx::digits];
+        auto __p = __tx::__convert(__value, __buf);
+        auto __len = __p - __buf;
+        if (__len <= __diff)
+        {
+            memcpy(__first, __buf, __len);
+            return {__first + __len, {}};
+        }
+        else
+            return {__last, errc::value_too_large};
+    }
+#endif
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY to_chars_result
+__to_chars_integral(char* __first, char* __last, _Tp __value, int __base,
+                    true_type)
+{
+    auto __x = __to_unsigned(__value);
+    if (__value < 0 && __first != __last)
+    {
+        *__first++ = '-';
+        __x = __complement(__x);
+    }
+
+    return __to_chars_integral(__first, __last, __x, __base, false_type());
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY to_chars_result
+__to_chars_integral(char* __first, char* __last, _Tp __value, int __base,
+                    false_type)
+{
+    if (__base == 10)
+        return __to_chars_itoa(__first, __last, __value, false_type());
+
+    auto __p = __last;
+    while (__p != __first)
+    {
+        auto __c = __value % __base;
+        __value /= __base;
+        *--__p = "0123456789abcdefghijklmnopqrstuvwxyz"[__c];
+        if (__value == 0)
+            break;
+    }
+
+    auto __len = __last - __p;
+    if (__value != 0 || !__len)
+        return {__last, errc::value_too_large};
+    else
+    {
+        memmove(__first, __p, __len);
+        return {__first + __len, {}};
+    }
+}
+
+template <typename _Tp, typename enable_if<is_integral<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY to_chars_result
+to_chars(char* __first, char* __last, _Tp __value)
+{
+    return __to_chars_itoa(__first, __last, __value, is_signed<_Tp>());
+}
+
+template <typename _Tp, typename enable_if<is_integral<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY to_chars_result
+to_chars(char* __first, char* __last, _Tp __value, int __base)
+{
+    _LIBCUDACXX_ASSERT(2 <= __base && __base <= 36, "base not in [2, 36]");
+    return __to_chars_integral(__first, __last, __value, __base,
+                               is_signed<_Tp>());
+}
+
+template <typename _It, typename _Tp, typename _Fn, typename... _Ts>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+__sign_combinator(_It __first, _It __last, _Tp& __value, _Fn __f, _Ts... __args)
+{
+    using __tl = numeric_limits<_Tp>;
+    decltype(__to_unsigned(__value)) __x;
+
+    bool __neg = (__first != __last && *__first == '-');
+    auto __r = __f(__neg ? __first + 1 : __first, __last, __x, __args...);
+    switch (__r.ec)
+    {
+    case errc::invalid_argument:
+        return {__first, __r.ec};
+    case errc::result_out_of_range:
+        return __r;
+    default:
+        break;
+    }
+
+    if (__neg)
+    {
+        if (__x <= __complement(__to_unsigned(__tl::min())))
+        {
+            __x = __complement(__x);
+            memcpy(&__value, &__x, sizeof(__x));
+            return __r;
+        }
+    }
+    else
+    {
+        if (__x <= (__tl::max)())
+        {
+            __value = __x;
+            return __r;
+        }
+    }
+
+    return {__r.ptr, errc::result_out_of_range};
+}
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+__in_pattern(_Tp __c)
+{
+    return '0' <= __c && __c <= '9';
+}
+
+struct _LIBCUDACXX_HIDDEN __in_pattern_result
+{
+    bool __ok;
+    int __val;
+
+    explicit _LIBCUDACXX_INLINE_VISIBILITY operator bool() const { return __ok; }
+};
+
+template <typename _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY __in_pattern_result
+__in_pattern(_Tp __c, int __base)
+{
+    if (__base <= 10)
+        return {'0' <= __c && __c < '0' + __base, __c - '0'};
+    else if (__in_pattern(__c))
+        return {true, __c - '0'};
+    else if ('a' <= __c && __c < 'a' + __base - 10)
+        return {true, __c - 'a' + 10};
+    else
+        return {'A' <= __c && __c < 'A' + __base - 10, __c - 'A' + 10};
+}
+
+template <typename _It, typename _Tp, typename _Fn, typename... _Ts>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+__subject_seq_combinator(_It __first, _It __last, _Tp& __value, _Fn __f,
+                         _Ts... __args)
+{
+    auto __find_non_zero = [](_It __first, _It __last) {
+        for (; __first != __last; ++__first)
+            if (*__first != '0')
+                break;
+        return __first;
+    };
+
+    auto __p = __find_non_zero(__first, __last);
+    if (__p == __last || !__in_pattern(*__p, __args...))
+    {
+        if (__p == __first)
+            return {__first, errc::invalid_argument};
+        else
+        {
+            __value = 0;
+            return {__p, {}};
+        }
+    }
+
+    auto __r = __f(__p, __last, __value, __args...);
+    if (__r.ec == errc::result_out_of_range)
+    {
+        for (; __r.ptr != __last; ++__r.ptr)
+        {
+            if (!__in_pattern(*__r.ptr, __args...))
+                break;
+        }
+    }
+
+    return __r;
+}
+
+template <typename _Tp, typename enable_if<is_unsigned<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+__from_chars_atoi(const char* __first, const char* __last, _Tp& __value)
+{
+    using __tx = __itoa::__traits<_Tp>;
+    using __output_type = typename __tx::type;
+
+    return __subject_seq_combinator(
+        __first, __last, __value,
+        [](const char* __first, const char* __last,
+           _Tp& __value) -> from_chars_result {
+            __output_type __a, __b;
+            auto __p = __tx::__read(__first, __last, __a, __b);
+            if (__p == __last || !__in_pattern(*__p))
+            {
+                __output_type __m = (numeric_limits<_Tp>::max)();
+                if (__m >= __a && __m - __a >= __b)
+                {
+                    __value = __a + __b;
+                    return {__p, {}};
+                }
+            }
+            return {__p, errc::result_out_of_range};
+        });
+}
+
+template <typename _Tp, typename enable_if<is_signed<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+__from_chars_atoi(const char* __first, const char* __last, _Tp& __value)
+{
+    using __t = decltype(__to_unsigned(__value));
+    return __sign_combinator(__first, __last, __value, __from_chars_atoi<__t>);
+}
+
+template <typename _Tp, typename enable_if<is_unsigned<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+__from_chars_integral(const char* __first, const char* __last, _Tp& __value,
+                      int __base)
+{
+    if (__base == 10)
+        return __from_chars_atoi(__first, __last, __value);
+
+    return __subject_seq_combinator(
+        __first, __last, __value,
+        [](const char* __p, const char* __last, _Tp& __value,
+           int __base) -> from_chars_result {
+            using __tl = numeric_limits<_Tp>;
+            auto __digits = __tl::digits / log2f(float(__base));
+            _Tp __a = __in_pattern(*__p++, __base).__val, __b = 0;
+
+            for (int __i = 1; __p != __last; ++__i, ++__p)
+            {
+                if (auto __c = __in_pattern(*__p, __base))
+                {
+                    if (__i < __digits - 1)
+                        __a = __a * __base + __c.__val;
+                    else
+                    {
+                        if (!__itoa::__mul_overflowed(__a, __base, __a))
+                            ++__p;
+                        __b = __c.__val;
+                        break;
+                    }
+                }
+                else
+                    break;
+            }
+
+            if (__p == __last || !__in_pattern(*__p, __base))
+            {
+                if ((__tl::max)() - __a >= __b)
+                {
+                    __value = __a + __b;
+                    return {__p, {}};
+                }
+            }
+            return {__p, errc::result_out_of_range};
+        },
+        __base);
+}
+
+template <typename _Tp, typename enable_if<is_signed<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+__from_chars_integral(const char* __first, const char* __last, _Tp& __value,
+                      int __base)
+{
+    using __t = decltype(__to_unsigned(__value));
+    return __sign_combinator(__first, __last, __value,
+                             __from_chars_integral<__t>, __base);
+}
+
+template <typename _Tp, typename enable_if<is_integral<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+from_chars(const char* __first, const char* __last, _Tp& __value)
+{
+    return __from_chars_atoi(__first, __last, __value);
+}
+
+template <typename _Tp, typename enable_if<is_integral<_Tp>::value, int>::type = 0>
+inline _LIBCUDACXX_INLINE_VISIBILITY from_chars_result
+from_chars(const char* __first, const char* __last, _Tp& __value, int __base)
+{
+    _LIBCUDACXX_ASSERT(2 <= __base && __base <= 36, "base not in [2, 36]");
+    return __from_chars_integral(__first, __last, __value, __base);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_CHARCONV
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/chrono b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/chrono
new file mode 100644
index 000000000000..d64d57eb0ea5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/chrono
@@ -0,0 +1,3348 @@
+// -*- C++ -*-
+//===---------------------------- chrono ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CHRONO
+#define _LIBCUDACXX_CHRONO
+
+/*
+    chrono synopsis
+
+namespace std
+{
+namespace chrono
+{
+
+template <class ToDuration, class Rep, class Period>
+constexpr
+ToDuration
+duration_cast(const duration<Rep, Period>& fd);
+
+template <class Rep> struct treat_as_floating_point : is_floating_point<Rep> {};
+
+template <class Rep> inline constexpr bool treat_as_floating_point_v
+    = treat_as_floating_point<Rep>::value;                       // C++17
+
+template <class Rep>
+struct duration_values
+{
+public:
+    static constexpr Rep zero(); // noexcept in C++20
+    static constexpr Rep max();  // noexcept in C++20
+    static constexpr Rep min();  // noexcept in C++20
+};
+
+// duration
+
+template <class Rep, class Period = ratio<1>>
+class duration
+{
+    static_assert(!__is_duration<Rep>::value, "A duration representation can not be a duration");
+    static_assert(__is_ratio<Period>::value, "Second template parameter of duration must be a std::ratio");
+    static_assert(Period::num > 0, "duration period must be positive");
+public:
+    typedef Rep rep;
+    typedef typename _Period::type period;
+
+    constexpr duration() = default;
+    template <class Rep2>
+        constexpr explicit duration(const Rep2& r,
+            typename enable_if
+            <
+               is_convertible<Rep2, rep>::value &&
+               (treat_as_floating_point<rep>::value ||
+               !treat_as_floating_point<rep>::value && !treat_as_floating_point<Rep2>::value)
+            >::type* = 0);
+
+    // conversions
+    template <class Rep2, class Period2>
+        constexpr duration(const duration<Rep2, Period2>& d,
+            typename enable_if
+            <
+                treat_as_floating_point<rep>::value ||
+                ratio_divide<Period2, period>::type::den == 1
+            >::type* = 0);
+
+    // observer
+
+    constexpr rep count() const;
+
+    // arithmetic
+
+    constexpr common_type<duration>::type  operator+() const;
+    constexpr common_type<duration>::type  operator-() const;
+    constexpr duration& operator++();    // constexpr in C++17
+    constexpr duration  operator++(int); // constexpr in C++17
+    constexpr duration& operator--();    // constexpr in C++17
+    constexpr duration  operator--(int); // constexpr in C++17
+
+    constexpr duration& operator+=(const duration& d);  // constexpr in C++17
+    constexpr duration& operator-=(const duration& d);  // constexpr in C++17
+
+    duration& operator*=(const rep& rhs);       // constexpr in C++17
+    duration& operator/=(const rep& rhs);       // constexpr in C++17
+    duration& operator%=(const rep& rhs);       // constexpr in C++17
+    duration& operator%=(const duration& rhs);  // constexpr in C++17
+
+    // special values
+
+    static constexpr duration zero(); // noexcept in C++20
+    static constexpr duration min();  // noexcept in C++20
+    static constexpr duration max();  // noexcept in C++20
+};
+
+typedef duration<long long,         nano> nanoseconds;
+typedef duration<long long,        micro> microseconds;
+typedef duration<long long,        milli> milliseconds;
+typedef duration<long long              > seconds;
+typedef duration<     long, ratio<  60> > minutes;
+typedef duration<     long, ratio<3600> > hours;
+
+template <class Clock, class Duration = typename Clock::duration>
+class time_point
+{
+public:
+    typedef Clock                     clock;
+    typedef Duration                  duration;
+    typedef typename duration::rep    rep;
+    typedef typename duration::period period;
+private:
+    duration d_;  // exposition only
+
+public:
+    time_point();  // has value "epoch" // constexpr in C++14
+    explicit time_point(const duration& d);  // same as time_point() + d // constexpr in C++14
+
+    // conversions
+    template <class Duration2>
+       time_point(const time_point<clock, Duration2>& t); // constexpr in C++14
+
+    // observer
+
+    duration time_since_epoch() const; // constexpr in C++14
+
+    // arithmetic
+
+    time_point& operator+=(const duration& d); // constexpr in C++17
+    time_point& operator-=(const duration& d); // constexpr in C++17
+
+    // special values
+
+    static constexpr time_point min();  // noexcept in C++20
+    static constexpr time_point max();  // noexcept in C++20
+};
+
+} // chrono
+
+// common_type traits
+template <class Rep1, class Period1, class Rep2, class Period2>
+  struct common_type<chrono::duration<Rep1, Period1>, chrono::duration<Rep2, Period2>>;
+
+template <class Clock, class Duration1, class Duration2>
+  struct common_type<chrono::time_point<Clock, Duration1>, chrono::time_point<Clock, Duration2>>;
+
+namespace chrono {
+
+
+template<class T> struct is_clock;  // C++20
+template<class T> inline constexpr bool is_clock_v = is_clock<T>::value;   // C++20
+
+
+// duration arithmetic
+template <class Rep1, class Period1, class Rep2, class Period2>
+  constexpr
+  typename common_type<duration<Rep1, Period1>, duration<Rep2, Period2>>::type
+  operator+(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Rep2, class Period2>
+  constexpr
+  typename common_type<duration<Rep1, Period1>, duration<Rep2, Period2>>::type
+  operator-(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period, class Rep2>
+  constexpr
+  duration<typename common_type<Rep1, Rep2>::type, Period>
+  operator*(const duration<Rep1, Period>& d, const Rep2& s);
+template <class Rep1, class Period, class Rep2>
+  constexpr
+  duration<typename common_type<Rep1, Rep2>::type, Period>
+  operator*(const Rep1& s, const duration<Rep2, Period>& d);
+template <class Rep1, class Period, class Rep2>
+  constexpr
+  duration<typename common_type<Rep1, Rep2>::type, Period>
+  operator/(const duration<Rep1, Period>& d, const Rep2& s);
+template <class Rep1, class Period1, class Rep2, class Period2>
+  constexpr
+  typename common_type<Rep1, Rep2>::type
+  operator/(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+
+// duration comparisons
+template <class Rep1, class Period1, class Rep2, class Period2>
+   constexpr
+   bool operator==(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Rep2, class Period2>
+   constexpr
+   bool operator!=(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Rep2, class Period2>
+   constexpr
+   bool operator< (const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Rep2, class Period2>
+   constexpr
+   bool operator<=(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Rep2, class Period2>
+   constexpr
+   bool operator> (const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Rep2, class Period2>
+   constexpr
+   bool operator>=(const duration<Rep1, Period1>& lhs, const duration<Rep2, Period2>& rhs);
+
+// duration_cast
+template <class ToDuration, class Rep, class Period>
+  ToDuration duration_cast(const duration<Rep, Period>& d);
+
+template <class ToDuration, class Rep, class Period>
+    constexpr ToDuration floor(const duration<Rep, Period>& d);    // C++17
+template <class ToDuration, class Rep, class Period>
+    constexpr ToDuration ceil(const duration<Rep, Period>& d);     // C++17
+template <class ToDuration, class Rep, class Period>
+    constexpr ToDuration round(const duration<Rep, Period>& d);    // C++17
+
+// duration I/O is elsewhere
+
+// time_point arithmetic (all constexpr in C++14)
+template <class Clock, class Duration1, class Rep2, class Period2>
+  time_point<Clock, typename common_type<Duration1, duration<Rep2, Period2>>::type>
+  operator+(const time_point<Clock, Duration1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Rep1, class Period1, class Clock, class Duration2>
+  time_point<Clock, typename common_type<duration<Rep1, Period1>, Duration2>::type>
+  operator+(const duration<Rep1, Period1>& lhs, const time_point<Clock, Duration2>& rhs);
+template <class Clock, class Duration1, class Rep2, class Period2>
+  time_point<Clock, typename common_type<Duration1, duration<Rep2, Period2>>::type>
+  operator-(const time_point<Clock, Duration1>& lhs, const duration<Rep2, Period2>& rhs);
+template <class Clock, class Duration1, class Duration2>
+  typename common_type<Duration1, Duration2>::type
+  operator-(const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+
+// time_point comparisons (all constexpr in C++14)
+template <class Clock, class Duration1, class Duration2>
+   bool operator==(const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+template <class Clock, class Duration1, class Duration2>
+   bool operator!=(const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+template <class Clock, class Duration1, class Duration2>
+   bool operator< (const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+template <class Clock, class Duration1, class Duration2>
+   bool operator<=(const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+template <class Clock, class Duration1, class Duration2>
+   bool operator> (const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+template <class Clock, class Duration1, class Duration2>
+   bool operator>=(const time_point<Clock, Duration1>& lhs, const time_point<Clock, Duration2>& rhs);
+
+// time_point_cast (constexpr in C++14)
+
+template <class ToDuration, class Clock, class Duration>
+  time_point<Clock, ToDuration> time_point_cast(const time_point<Clock, Duration>& t);
+
+template <class ToDuration, class Clock, class Duration>
+    constexpr time_point<Clock, ToDuration>
+    floor(const time_point<Clock, Duration>& tp);                  // C++17
+
+template <class ToDuration, class Clock, class Duration>
+    constexpr time_point<Clock, ToDuration>
+    ceil(const time_point<Clock, Duration>& tp);                   // C++17
+
+template <class ToDuration, class Clock, class Duration>
+    constexpr time_point<Clock, ToDuration>
+    round(const time_point<Clock, Duration>& tp);                  // C++17
+
+template <class Rep, class Period>
+    constexpr duration<Rep, Period> abs(duration<Rep, Period> d);  // C++17
+
+// Clocks
+
+class system_clock
+{
+public:
+    typedef microseconds                     duration;
+    typedef duration::rep                    rep;
+    typedef duration::period                 period;
+    typedef chrono::time_point<system_clock> time_point;
+    static const bool is_steady =            false; // constexpr in C++14
+
+    static time_point now() noexcept;
+    static time_t     to_time_t  (const time_point& __t) noexcept;
+    static time_point from_time_t(time_t __t) noexcept;
+};
+
+template <class Duration>
+  using sys_time  = time_point<system_clock, Duration>; // C++20
+using sys_seconds = sys_time<seconds>;                  // C++20
+using sys_days    = sys_time<days>;                     // C++20
+
+class utc_clock;                                        // C++20
+
+template <class Duration>
+  using utc_time  = time_point<utc_clock, Duration>;    // C++20
+using utc_seconds = utc_time<seconds>;                  // C++20
+
+class tai_clock;                                        // C++20
+
+template <class Duration>
+  using tai_time  = time_point<tai_clock, Duration>;    // C++20
+using tai_seconds = tai_time<seconds>;                  // C++20
+
+class file_clock;                                       // C++20
+
+template<class Duration>
+  using file_time = time_point<file_clock, Duration>;   // C++20
+
+class steady_clock
+{
+public:
+    typedef nanoseconds                                   duration;
+    typedef duration::rep                                 rep;
+    typedef duration::period                              period;
+    typedef chrono::time_point<steady_clock, duration>    time_point;
+    static const bool is_steady =                         true; // constexpr in C++14
+
+    static time_point now() noexcept;
+};
+
+typedef steady_clock high_resolution_clock;
+
+// 25.7.8, local time           // C++20
+struct local_t {};
+template<class Duration>
+  using local_time  = time_point<local_t, Duration>;
+using local_seconds = local_time<seconds>;
+using local_days    = local_time<days>;
+
+// 25.7.9, time_point conversions template<class DestClock, class SourceClock>    // C++20
+struct clock_time_conversion;
+
+template<class DestClock, class SourceClock, class Duration>
+  auto clock_cast(const time_point<SourceClock, Duration>& t);
+
+// 25.8.2, class last_spec    // C++20
+struct last_spec;
+
+// 25.8.3, class day          // C++20
+
+class day;
+constexpr bool operator==(const day& x, const day& y) noexcept;
+constexpr bool operator!=(const day& x, const day& y) noexcept;
+constexpr bool operator< (const day& x, const day& y) noexcept;
+constexpr bool operator> (const day& x, const day& y) noexcept;
+constexpr bool operator<=(const day& x, const day& y) noexcept;
+constexpr bool operator>=(const day& x, const day& y) noexcept;
+constexpr day  operator+(const day&  x, const days& y) noexcept;
+constexpr day  operator+(const days& x, const day&  y) noexcept;
+constexpr day  operator-(const day&  x, const days& y) noexcept;
+constexpr days operator-(const day&  x, const day&  y) noexcept;
+
+// 25.8.4, class month    // C++20
+class month;
+constexpr bool operator==(const month& x, const month& y) noexcept;
+constexpr bool operator!=(const month& x, const month& y) noexcept;
+constexpr bool operator< (const month& x, const month& y) noexcept;
+constexpr bool operator> (const month& x, const month& y) noexcept;
+constexpr bool operator<=(const month& x, const month& y) noexcept;
+constexpr bool operator>=(const month& x, const month& y) noexcept;
+constexpr month  operator+(const month&  x, const months& y) noexcept;
+constexpr month  operator+(const months& x,  const month& y) noexcept;
+constexpr month  operator-(const month&  x, const months& y) noexcept;
+constexpr months operator-(const month&  x,  const month& y) noexcept;
+
+// 25.8.5, class year    // C++20
+class year;
+constexpr bool operator==(const year& x, const year& y) noexcept;
+constexpr bool operator!=(const year& x, const year& y) noexcept;
+constexpr bool operator< (const year& x, const year& y) noexcept;
+constexpr bool operator> (const year& x, const year& y) noexcept;
+constexpr bool operator<=(const year& x, const year& y) noexcept;
+constexpr bool operator>=(const year& x, const year& y) noexcept;
+constexpr year  operator+(const year&  x, const years& y) noexcept;
+constexpr year  operator+(const years& x, const year&  y) noexcept;
+constexpr year  operator-(const year&  x, const years& y) noexcept;
+constexpr years operator-(const year&  x, const year&  y) noexcept;
+
+// 25.8.6, class weekday    // C++20
+class weekday;
+
+constexpr bool operator==(const weekday& x, const weekday& y) noexcept;
+constexpr bool operator!=(const weekday& x, const weekday& y) noexcept;
+constexpr weekday operator+(const weekday& x, const days&    y) noexcept;
+constexpr weekday operator+(const days&    x, const weekday& y) noexcept;
+constexpr weekday operator-(const weekday& x, const days&    y) noexcept;
+constexpr days    operator-(const weekday& x, const weekday& y) noexcept;
+
+// 25.8.7, class weekday_indexed    // C++20
+
+class weekday_indexed;
+constexpr bool operator==(const weekday_indexed& x, const weekday_indexed& y) noexcept;
+constexpr bool operator!=(const weekday_indexed& x, const weekday_indexed& y) noexcept;
+
+// 25.8.8, class weekday_last    // C++20
+class weekday_last;
+
+constexpr bool operator==(const weekday_last& x, const weekday_last& y) noexcept;
+constexpr bool operator!=(const weekday_last& x, const weekday_last& y) noexcept;
+
+// 25.8.9, class month_day    // C++20
+class month_day;
+
+constexpr bool operator==(const month_day& x, const month_day& y) noexcept;
+constexpr bool operator!=(const month_day& x, const month_day& y) noexcept;
+constexpr bool operator< (const month_day& x, const month_day& y) noexcept;
+constexpr bool operator> (const month_day& x, const month_day& y) noexcept;
+constexpr bool operator<=(const month_day& x, const month_day& y) noexcept;
+constexpr bool operator>=(const month_day& x, const month_day& y) noexcept;
+
+
+// 25.8.10, class month_day_last    // C++20
+class month_day_last;
+
+constexpr bool operator==(const month_day_last& x, const month_day_last& y) noexcept;
+constexpr bool operator!=(const month_day_last& x, const month_day_last& y) noexcept;
+constexpr bool operator< (const month_day_last& x, const month_day_last& y) noexcept;
+constexpr bool operator> (const month_day_last& x, const month_day_last& y) noexcept;
+constexpr bool operator<=(const month_day_last& x, const month_day_last& y) noexcept;
+constexpr bool operator>=(const month_day_last& x, const month_day_last& y) noexcept;
+
+// 25.8.11, class month_weekday    // C++20
+class month_weekday;
+
+constexpr bool operator==(const month_weekday& x, const month_weekday& y) noexcept;
+constexpr bool operator!=(const month_weekday& x, const month_weekday& y) noexcept;
+
+// 25.8.12, class month_weekday_last    // C++20
+class month_weekday_last;
+
+constexpr bool operator==(const month_weekday_last& x, const month_weekday_last& y) noexcept;
+constexpr bool operator!=(const month_weekday_last& x, const month_weekday_last& y) noexcept;
+
+
+// 25.8.13, class year_month    // C++20
+class year_month;
+
+constexpr bool operator==(const year_month& x, const year_month& y) noexcept;
+constexpr bool operator!=(const year_month& x, const year_month& y) noexcept;
+constexpr bool operator< (const year_month& x, const year_month& y) noexcept;
+constexpr bool operator> (const year_month& x, const year_month& y) noexcept;
+constexpr bool operator<=(const year_month& x, const year_month& y) noexcept;
+constexpr bool operator>=(const year_month& x, const year_month& y) noexcept;
+
+constexpr year_month operator+(const year_month& ym, const months& dm) noexcept;
+constexpr year_month operator+(const months& dm, const year_month& ym) noexcept;
+constexpr year_month operator-(const year_month& ym, const months& dm) noexcept;
+constexpr months operator-(const year_month& x, const year_month& y) noexcept;
+constexpr year_month operator+(const year_month& ym, const years& dy) noexcept;
+constexpr year_month operator+(const years& dy, const year_month& ym) noexcept;
+constexpr year_month operator-(const year_month& ym, const years& dy) noexcept;
+
+// 25.8.14, class year_month_day class    // C++20
+year_month_day;
+
+constexpr bool operator==(const year_month_day& x, const year_month_day& y) noexcept;
+constexpr bool operator!=(const year_month_day& x, const year_month_day& y) noexcept;
+constexpr bool operator< (const year_month_day& x, const year_month_day& y) noexcept;
+constexpr bool operator> (const year_month_day& x, const year_month_day& y) noexcept;
+constexpr bool operator<=(const year_month_day& x, const year_month_day& y) noexcept;
+constexpr bool operator>=(const year_month_day& x, const year_month_day& y) noexcept;
+
+constexpr year_month_day operator+(const year_month_day& ymd, const months& dm) noexcept;
+constexpr year_month_day operator+(const months& dm, const year_month_day& ymd) noexcept;
+constexpr year_month_day operator+(const year_month_day& ymd, const years& dy) noexcept;
+constexpr year_month_day operator+(const years& dy, const year_month_day& ymd) noexcept;
+constexpr year_month_day operator-(const year_month_day& ymd, const months& dm) noexcept;
+constexpr year_month_day operator-(const year_month_day& ymd, const years& dy) noexcept;
+
+
+// 25.8.15, class year_month_day_last    // C++20
+class year_month_day_last;
+
+constexpr bool operator==(const year_month_day_last& x,
+                          const year_month_day_last& y) noexcept;
+constexpr bool operator!=(const year_month_day_last& x,
+                          const year_month_day_last& y) noexcept;
+constexpr bool operator< (const year_month_day_last& x,
+                          const year_month_day_last& y) noexcept;
+constexpr bool operator> (const year_month_day_last& x,
+                          const year_month_day_last& y) noexcept;
+constexpr bool operator<=(const year_month_day_last& x,
+                          const year_month_day_last& y) noexcept;
+constexpr bool operator>=(const year_month_day_last& x,
+                          const year_month_day_last& y) noexcept;
+
+constexpr year_month_day_last
+  operator+(const year_month_day_last& ymdl, const months& dm) noexcept;
+constexpr year_month_day_last
+  operator+(const months& dm, const year_month_day_last& ymdl) noexcept;
+constexpr year_month_day_last
+  operator+(const year_month_day_last& ymdl, const years& dy) noexcept;
+constexpr year_month_day_last
+  operator+(const years& dy, const year_month_day_last& ymdl) noexcept;
+constexpr year_month_day_last
+  operator-(const year_month_day_last& ymdl, const months& dm) noexcept;
+constexpr year_month_day_last
+  operator-(const year_month_day_last& ymdl, const years& dy) noexcept;
+
+// 25.8.16, class year_month_weekday    // C++20
+class year_month_weekday;
+
+constexpr bool operator==(const year_month_weekday& x,
+                          const year_month_weekday& y) noexcept;
+constexpr bool operator!=(const year_month_weekday& x,
+                          const year_month_weekday& y) noexcept;
+
+constexpr year_month_weekday
+  operator+(const year_month_weekday& ymwd, const months& dm) noexcept;
+constexpr year_month_weekday
+  operator+(const months& dm, const year_month_weekday& ymwd) noexcept;
+constexpr year_month_weekday
+  operator+(const year_month_weekday& ymwd, const years& dy) noexcept;
+constexpr year_month_weekday
+  operator+(const years& dy, const year_month_weekday& ymwd) noexcept;
+constexpr year_month_weekday
+  operator-(const year_month_weekday& ymwd, const months& dm) noexcept;
+constexpr year_month_weekday
+  operator-(const year_month_weekday& ymwd, const years& dy) noexcept;
+
+// 25.8.17, class year_month_weekday_last    // C++20
+class year_month_weekday_last;
+
+constexpr bool operator==(const year_month_weekday_last& x,
+                          const year_month_weekday_last& y) noexcept;
+constexpr bool operator!=(const year_month_weekday_last& x,
+                          const year_month_weekday_last& y) noexcept;
+constexpr year_month_weekday_last
+  operator+(const year_month_weekday_last& ymwdl, const months& dm) noexcept;
+constexpr year_month_weekday_last
+  operator+(const months& dm, const year_month_weekday_last& ymwdl) noexcept;
+constexpr year_month_weekday_last
+  operator+(const year_month_weekday_last& ymwdl, const years& dy) noexcept;
+constexpr year_month_weekday_last
+  operator+(const years& dy, const year_month_weekday_last& ymwdl) noexcept;
+constexpr year_month_weekday_last
+  operator-(const year_month_weekday_last& ymwdl, const months& dm) noexcept;
+constexpr year_month_weekday_last
+  operator-(const year_month_weekday_last& ymwdl, const years& dy) noexcept;
+
+// 25.8.18, civil calendar conventional syntax operators    // C++20
+constexpr year_month
+  operator/(const year& y, const month& m) noexcept;
+constexpr year_month
+  operator/(const year& y, int m) noexcept;
+constexpr month_day
+  operator/(const month& m, const day& d) noexcept;
+constexpr month_day
+  operator/(const month& m, int d) noexcept;
+constexpr month_day
+  operator/(int m, const day& d) noexcept;
+constexpr month_day
+  operator/(const day& d, const month& m) noexcept;
+constexpr month_day
+  operator/(const day& d, int m) noexcept;
+constexpr month_day_last
+  operator/(const month& m, last_spec) noexcept;
+constexpr month_day_last
+  operator/(int m, last_spec) noexcept;
+constexpr month_day_last
+  operator/(last_spec, const month& m) noexcept;
+constexpr month_day_last
+  operator/(last_spec, int m) noexcept;
+constexpr month_weekday
+  operator/(const month& m, const weekday_indexed& wdi) noexcept;
+constexpr month_weekday
+  operator/(int m, const weekday_indexed& wdi) noexcept;
+constexpr month_weekday
+  operator/(const weekday_indexed& wdi, const month& m) noexcept;
+constexpr month_weekday
+  operator/(const weekday_indexed& wdi, int m) noexcept;
+constexpr month_weekday_last
+  operator/(const month& m, const weekday_last& wdl) noexcept;
+constexpr month_weekday_last
+  operator/(int m, const weekday_last& wdl) noexcept;
+constexpr month_weekday_last
+  operator/(const weekday_last& wdl, const month& m) noexcept;
+constexpr month_weekday_last
+  operator/(const weekday_last& wdl, int m) noexcept;
+constexpr year_month_day
+  operator/(const year_month& ym, const day& d) noexcept;
+constexpr year_month_day
+  operator/(const year_month& ym, int d) noexcept;
+constexpr year_month_day
+  operator/(const year& y, const month_day& md) noexcept;
+constexpr year_month_day
+  operator/(int y, const month_day& md) noexcept;
+constexpr year_month_day
+  operator/(const month_day& md, const year& y) noexcept;
+constexpr year_month_day
+  operator/(const month_day& md, int y) noexcept;
+constexpr year_month_day_last
+  operator/(const year_month& ym, last_spec) noexcept;
+constexpr year_month_day_last
+  operator/(const year& y, const month_day_last& mdl) noexcept;
+constexpr year_month_day_last
+  operator/(int y, const month_day_last& mdl) noexcept;
+constexpr year_month_day_last
+  operator/(const month_day_last& mdl, const year& y) noexcept;
+constexpr year_month_day_last
+  operator/(const month_day_last& mdl, int y) noexcept;
+constexpr year_month_weekday
+  operator/(const year_month& ym, const weekday_indexed& wdi) noexcept;
+constexpr year_month_weekday
+  operator/(const year& y, const month_weekday& mwd) noexcept;
+constexpr year_month_weekday
+  operator/(int y, const month_weekday& mwd) noexcept;
+constexpr year_month_weekday
+  operator/(const month_weekday& mwd, const year& y) noexcept;
+constexpr year_month_weekday
+  operator/(const month_weekday& mwd, int y) noexcept;
+constexpr year_month_weekday_last
+  operator/(const year_month& ym, const weekday_last& wdl) noexcept;
+constexpr year_month_weekday_last
+  operator/(const year& y, const month_weekday_last& mwdl) noexcept;
+constexpr year_month_weekday_last
+  operator/(int y, const month_weekday_last& mwdl) noexcept;
+constexpr year_month_weekday_last
+  operator/(const month_weekday_last& mwdl, const year& y) noexcept;
+constexpr year_month_weekday_last
+  operator/(const month_weekday_last& mwdl, int y) noexcept;
+
+// 26.9, class template hh_mm_ss
+template <class Duration>
+class hh_mm_ss
+{
+    bool            is_neg; // exposition only
+    chrono::hours   h;      // exposition only
+    chrono::minutes m;      // exposition only
+    chrono::seconds s;      // exposition only
+    precision       ss;     // exposition only
+
+public:
+    static unsigned constexpr fractional_width = see below;
+    using precision                            = see below;
+
+    constexpr hh_mm_ss() noexcept : hh_mm_ss{Duration::zero()} {}
+    constexpr explicit hh_mm_ss(Duration d) noexcept;
+
+    constexpr bool is_negative() const noexcept;
+    constexpr chrono::hours hours() const noexcept;
+    constexpr chrono::minutes minutes() const noexcept;
+    constexpr chrono::seconds seconds() const noexcept;
+    constexpr precision subseconds() const noexcept;
+
+    constexpr explicit operator  precision()   const noexcept;
+    constexpr          precision to_duration() const noexcept;
+};
+
+template <class charT, class traits, class Duration>
+  basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os, hh_mm_ss<Duration> const& hms);
+
+// 26.10, 12/24 hour functions
+constexpr bool is_am(hours const& h) noexcept;
+constexpr bool is_pm(hours const& h) noexcept;
+constexpr hours make12(const hours& h) noexcept;
+constexpr hours make24(const hours& h, bool is_pm) noexcept;
+
+
+// 25.10.2, time zone database     // C++20
+struct tzdb;
+class tzdb_list;
+
+// 25.10.2.3, time zone database access    // C++20
+const tzdb& get_tzdb();
+tzdb_list& get_tzdb_list();
+const time_zone* locate_zone(string_view tz_name);
+const time_zone* current_zone();
+
+// 25.10.2.4, remote time zone database support    // C++20
+const tzdb& reload_tzdb();
+string remote_version();
+
+// 25.10.3, exception classes    // C++20
+class nonexistent_local_time;
+class ambiguous_local_time;
+
+// 25.10.4, information classes    // C++20
+struct sys_info;
+struct local_info;
+
+// 25.10.5, class time_zone    // C++20
+enum class choose {earliest, latest};
+class time_zone;
+bool operator==(const time_zone& x, const time_zone& y) noexcept;
+bool operator!=(const time_zone& x, const time_zone& y) noexcept;
+bool operator<(const time_zone& x, const time_zone& y) noexcept;
+bool operator>(const time_zone& x, const time_zone& y) noexcept;
+bool operator<=(const time_zone& x, const time_zone& y) noexcept;
+bool operator>=(const time_zone& x, const time_zone& y) noexcept;
+
+// 25.10.6, class template zoned_traits    // C++20
+template<class T> struct zoned_traits;
+
+// 25.10.7, class template zoned_time    // C++20
+template<class Duration, class TimeZonePtr = const time_zone*> class zoned_time;
+using zoned_seconds = zoned_time<seconds>;
+
+template<class Duration1, class Duration2, class TimeZonePtr>
+  bool operator==(const zoned_time<Duration1, TimeZonePtr>& x,
+                  const zoned_time<Duration2, TimeZonePtr>& y);
+template<class Duration1, class Duration2, class TimeZonePtr>
+  bool operator!=(const zoned_time<Duration1, TimeZonePtr>& x,
+                  const zoned_time<Duration2, TimeZonePtr>& y);
+
+// 25.10.8, leap second support    // C++20
+class leap;
+
+bool operator==(const leap& x, const leap& y);
+bool operator!=(const leap& x, const leap& y);
+bool operator< (const leap& x, const leap& y);
+bool operator> (const leap& x, const leap& y);
+bool operator<=(const leap& x, const leap& y);
+bool operator>=(const leap& x, const leap& y);
+template<class Duration>
+  bool operator==(const leap& x, const sys_time<Duration>& y);
+template<class Duration>
+  bool operator==(const sys_time<Duration>& x, const leap& y);
+template<class Duration>
+  bool operator!=(const leap& x, const sys_time<Duration>& y);
+template<class Duration>
+  bool operator!=(const sys_time<Duration>& x, const leap& y);
+template<class Duration>
+  bool operator< (const leap& x, const sys_time<Duration>& y);
+template<class Duration>
+  bool operator< (const sys_time<Duration>& x, const leap& y);
+template<class Duration>
+  bool operator> (const leap& x, const sys_time<Duration>& y);
+template<class Duration>
+  bool operator> (const sys_time<Duration>& x, const leap& y);
+template<class Duration>
+  bool operator<=(const leap& x, const sys_time<Duration>& y);
+template<class Duration>
+  bool operator<=(const sys_time<Duration>& x, const leap& y);
+template<class Duration>
+  bool operator>=(const leap& x, const sys_time<Duration>& y);
+template<class Duration>
+  bool operator>=(const sys_time<Duration>& x, const leap& y);
+
+// 25.10.9, class link    // C++20
+class link;
+bool operator==(const link& x, const link& y);
+bool operator!=(const link& x, const link& y);
+bool operator< (const link& x, const link& y);
+bool operator> (const link& x, const link& y);
+bool operator<=(const link& x, const link& y);
+bool operator>=(const link& x, const link& y);
+
+// 25.11, formatting    // C++20
+template<class charT, class Streamable>
+  basic_string<charT>
+    format(const charT* fmt, const Streamable& s);
+
+template<class charT, class Streamable>
+  basic_string<charT>
+    format(const locale& loc, const charT* fmt, const Streamable& s);
+
+template<class charT, class traits, class Alloc, class Streamable>
+  basic_string<charT, traits, Alloc>
+    format(const basic_string<charT, traits, Alloc>& fmt, const Streamable& s);
+
+template<class charT, class traits, class Alloc, class Streamable>
+  basic_string<charT, traits, Alloc>
+    format(const locale& loc, const basic_string<charT, traits, Alloc>& fmt,
+           const Streamable& s);
+
+// 25.12, parsing    // C++20
+template<class charT, class traits, class Alloc, class Parsable>
+unspecified
+    parse(const basic_string<charT, traits, Alloc>& format, Parsable& tp);
+
+template<class charT, class traits, class Alloc, class Parsable>
+unspecified
+    parse(const basic_string<charT, traits, Alloc>& format, Parsable& tp,
+          basic_string<charT, traits, Alloc>& abbrev);
+
+template<class charT, class traits, class Alloc, class Parsable>
+unspecified
+    parse(const basic_string<charT, traits, Alloc>& format, Parsable& tp,
+          minutes& offset);
+
+template<class charT, class traits, class Alloc, class Parsable>
+unspecified
+    parse(const basic_string<charT, traits, Alloc>& format, Parsable& tp,
+          basic_string<charT, traits, Alloc>& abbrev, minutes& offset);
+
+// calendrical constants
+inline constexpr last_spec                              last{};       // C++20
+inline constexpr chrono::weekday                        Sunday{0};    // C++20
+inline constexpr chrono::weekday                        Monday{1};    // C++20
+inline constexpr chrono::weekday                        Tuesday{2};   // C++20
+inline constexpr chrono::weekday                        Wednesday{3}; // C++20
+inline constexpr chrono::weekday                        Thursday{4};  // C++20
+inline constexpr chrono::weekday                        Friday{5};    // C++20
+inline constexpr chrono::weekday                        Saturday{6};  // C++20
+
+inline constexpr chrono::month                          January{1};   // C++20
+inline constexpr chrono::month                          February{2};  // C++20
+inline constexpr chrono::month                          March{3};     // C++20
+inline constexpr chrono::month                          April{4};     // C++20
+inline constexpr chrono::month                          May{5};       // C++20
+inline constexpr chrono::month                          June{6};      // C++20
+inline constexpr chrono::month                          July{7};      // C++20
+inline constexpr chrono::month                          August{8};    // C++20
+inline constexpr chrono::month                          September{9}; // C++20
+inline constexpr chrono::month                          October{10};  // C++20
+inline constexpr chrono::month                          November{11}; // C++20
+inline constexpr chrono::month                          December{12}; // C++20
+}  // chrono
+
+inline namespace literals {
+  inline namespace chrono_literals {
+constexpr chrono::hours                                 operator ""h(unsigned long long); // C++14
+constexpr chrono::duration<unspecified , ratio<3600,1>> operator ""h(long double); // C++14
+constexpr chrono::minutes                               operator ""min(unsigned long long); // C++14
+constexpr chrono::duration<unspecified , ratio<60,1>>   operator ""min(long double); // C++14
+constexpr chrono::seconds                               operator ""s(unsigned long long); // C++14
+constexpr chrono::duration<unspecified >                operator ""s(long double); // C++14
+constexpr chrono::milliseconds                          operator ""ms(unsigned long long); // C++14
+constexpr chrono::duration<unspecified , milli>         operator ""ms(long double); // C++14
+constexpr chrono::microseconds                          operator ""us(unsigned long long); // C++14
+constexpr chrono::duration<unspecified , micro>         operator ""us(long double); // C++14
+constexpr chrono::nanoseconds                           operator ""ns(unsigned long long); // C++14
+constexpr chrono::duration<unspecified , nano>          operator ""ns(long double); // C++14
+constexpr chrono::day                                   operator ""d(unsigned long long d) noexcept; // C++20
+constexpr chrono::year                                  operator ""y(unsigned long long y) noexcept; // C++20
+}  // chrono_literals
+}  // literals
+
+}  // std
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <ctime>
+#include <type_traits>
+#include <ratio>
+#include <limits>
+#include <version>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+_LIBCUDACXX_BEGIN_NAMESPACE_FILESYSTEM
+struct _FilesystemClock;
+_LIBCUDACXX_END_NAMESPACE_FILESYSTEM
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+# if _LIBCUDACXX_CUDA_ABI_VERSION > 3
+#  define _LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T double
+# else
+#  define _LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T long double
+# endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+namespace chrono
+{
+
+template <class _Rep, class _Period = ratio<1> > class _LIBCUDACXX_TEMPLATE_VIS duration;
+
+template <class _Tp>
+struct __is_duration : false_type {};
+
+template <class _Rep, class _Period>
+struct __is_duration<duration<_Rep, _Period> > : true_type  {};
+
+template <class _Rep, class _Period>
+struct __is_duration<const duration<_Rep, _Period> > : true_type  {};
+
+template <class _Rep, class _Period>
+struct __is_duration<volatile duration<_Rep, _Period> > : true_type  {};
+
+template <class _Rep, class _Period>
+struct __is_duration<const volatile duration<_Rep, _Period> > : true_type  {};
+
+} // chrono
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+struct _LIBCUDACXX_TEMPLATE_VIS common_type<chrono::duration<_Rep1, _Period1>,
+                                         chrono::duration<_Rep2, _Period2> >
+{
+    typedef chrono::duration<typename common_type<_Rep1, _Rep2>::type,
+                             typename __ratio_gcd<_Period1, _Period2>::type> type;
+};
+
+namespace chrono {
+
+// duration_cast
+
+template <class _FromDuration, class _ToDuration,
+          class _Period = typename ratio_divide<typename _FromDuration::period, typename _ToDuration::period>::type,
+          bool = _Period::num == 1,
+          bool = _Period::den == 1>
+struct __duration_cast;
+
+template <class _FromDuration, class _ToDuration, class _Period>
+struct __duration_cast<_FromDuration, _ToDuration, _Period, true, true>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    _ToDuration operator()(const _FromDuration& __fd) const
+    {
+        return _ToDuration(static_cast<typename _ToDuration::rep>(__fd.count()));
+    }
+};
+
+template <class _FromDuration, class _ToDuration, class _Period>
+struct __duration_cast<_FromDuration, _ToDuration, _Period, true, false>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    _ToDuration operator()(const _FromDuration& __fd) const
+    {
+        typedef typename common_type<typename _ToDuration::rep, typename _FromDuration::rep, intmax_t>::type _Ct;
+        return _ToDuration(static_cast<typename _ToDuration::rep>(
+                           static_cast<_Ct>(__fd.count()) / static_cast<_Ct>(_Period::den)));
+    }
+};
+
+template <class _FromDuration, class _ToDuration, class _Period>
+struct __duration_cast<_FromDuration, _ToDuration, _Period, false, true>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    _ToDuration operator()(const _FromDuration& __fd) const
+    {
+        typedef typename common_type<typename _ToDuration::rep, typename _FromDuration::rep, intmax_t>::type _Ct;
+        return _ToDuration(static_cast<typename _ToDuration::rep>(
+                           static_cast<_Ct>(__fd.count()) * static_cast<_Ct>(_Period::num)));
+    }
+};
+
+template <class _FromDuration, class _ToDuration, class _Period>
+struct __duration_cast<_FromDuration, _ToDuration, _Period, false, false>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    _ToDuration operator()(const _FromDuration& __fd) const
+    {
+        typedef typename common_type<typename _ToDuration::rep, typename _FromDuration::rep, intmax_t>::type _Ct;
+        return _ToDuration(static_cast<typename _ToDuration::rep>(
+                           static_cast<_Ct>(__fd.count()) * static_cast<_Ct>(_Period::num)
+                                                          / static_cast<_Ct>(_Period::den)));
+    }
+};
+
+template <class _ToDuration, class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    _ToDuration
+>::type
+duration_cast(const duration<_Rep, _Period>& __fd)
+{
+    return __duration_cast<duration<_Rep, _Period>, _ToDuration>()(__fd);
+}
+
+template <class _Rep>
+struct _LIBCUDACXX_TEMPLATE_VIS treat_as_floating_point : is_floating_point<_Rep> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Rep>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool treat_as_floating_point_v
+    = treat_as_floating_point<_Rep>::value;
+#endif // _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+
+template <class _Rep>
+struct _LIBCUDACXX_TEMPLATE_VIS duration_values
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR _Rep zero() _NOEXCEPT {return _Rep(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR _Rep max()  _NOEXCEPT {return numeric_limits<_Rep>::max();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR _Rep min()  _NOEXCEPT {return numeric_limits<_Rep>::lowest();}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _ToDuration, class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    _ToDuration
+>::type
+floor(const duration<_Rep, _Period>& __d)
+{
+    _ToDuration __t = duration_cast<_ToDuration>(__d);
+    if (__t > __d)
+        __t = __t - _ToDuration{1};
+    return __t;
+}
+
+template <class _ToDuration, class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    _ToDuration
+>::type
+ceil(const duration<_Rep, _Period>& __d)
+{
+    _ToDuration __t = duration_cast<_ToDuration>(__d);
+    if (__t < __d)
+        __t = __t + _ToDuration{1};
+    return __t;
+}
+
+template <class _ToDuration, class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    _ToDuration
+>::type
+round(const duration<_Rep, _Period>& __d)
+{
+    _ToDuration __lower = floor<_ToDuration>(__d);
+    _ToDuration __upper = __lower + _ToDuration{1};
+    auto __lowerDiff = __d - __lower;
+    auto __upperDiff = __upper - __d;
+    if (__lowerDiff < __upperDiff)
+        return __lower;
+    if (__lowerDiff > __upperDiff)
+        return __upper;
+    return __lower.count() & 1 ? __upper : __lower;
+}
+#endif // _LIBCUDACXX_STD_VER > 11
+
+// duration
+
+template <class _Rep, class _Period>
+class _LIBCUDACXX_TEMPLATE_VIS duration
+{
+    static_assert(!__is_duration<_Rep>::value, "A duration representation can not be a duration");
+    static_assert(__is_ratio<_Period>::value, "Second template parameter of duration must be a std::ratio");
+    static_assert(_Period::num > 0, "duration period must be positive");
+
+    template <class _R1, class _R2>
+    struct __no_overflow
+    {
+    private:
+        static const intmax_t __gcd_n1_n2 = __static_gcd<_R1::num, _R2::num>::value;
+        static const intmax_t __gcd_d1_d2 = __static_gcd<_R1::den, _R2::den>::value;
+        static const intmax_t __n1 = _R1::num / __gcd_n1_n2;
+        static const intmax_t __d1 = _R1::den / __gcd_d1_d2;
+        static const intmax_t __n2 = _R2::num / __gcd_n1_n2;
+        static const intmax_t __d2 = _R2::den / __gcd_d1_d2;
+        static const intmax_t max = -((intmax_t(1) << (sizeof(intmax_t) * CHAR_BIT - 1)) + 1);
+
+        template <intmax_t _Xp, intmax_t _Yp, bool __overflow>
+        struct __mul    // __overflow == false
+        {
+            static const intmax_t value = _Xp * _Yp;
+        };
+
+        template <intmax_t _Xp, intmax_t _Yp>
+        struct __mul<_Xp, _Yp, true>
+        {
+            static const intmax_t value = 1;
+        };
+
+    public:
+        static const bool value = (__n1 <= max / __d2) && (__n2 <= max / __d1);
+        typedef ratio<__mul<__n1, __d2, !value>::value,
+                      __mul<__n2, __d1, !value>::value> type;
+    };
+
+public:
+    typedef _Rep rep;
+    typedef typename _Period::type period;
+private:
+    rep __rep_;
+public:
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+    _LIBCUDACXX_CONSTEXPR
+        duration() = default;
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+        duration() {}
+#endif
+
+    template <class _Rep2>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+        explicit duration(const _Rep2& __r,
+            typename enable_if
+            <
+               is_convertible<_Rep2, rep>::value &&
+               (treat_as_floating_point<rep>::value ||
+               !treat_as_floating_point<_Rep2>::value)
+            >::type* = 0)
+                : __rep_(static_cast<rep>(__r)) {}
+
+    // conversions
+    template <class _Rep2, class _Period2>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+        duration(const duration<_Rep2, _Period2>& __d,
+            typename enable_if
+            <
+                __no_overflow<_Period2, period>::value && (
+                treat_as_floating_point<rep>::value ||
+                (__no_overflow<_Period2, period>::type::den == 1 &&
+                 !treat_as_floating_point<_Rep2>::value))
+            >::type* = 0)
+                : __rep_(_CUDA_VSTD::chrono::duration_cast<duration>(__d).count()) {}
+
+    // observer
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR rep count() const {return __rep_;}
+
+    // arithmetic
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR typename common_type<duration>::type operator+() const {return typename common_type<duration>::type(*this);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR typename common_type<duration>::type operator-() const {return typename common_type<duration>::type(-__rep_);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator++()      {++__rep_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration  operator++(int)   {return duration(__rep_++);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator--()      {--__rep_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration  operator--(int)   {return duration(__rep_--);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator+=(const duration& __d) {__rep_ += __d.count(); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator-=(const duration& __d) {__rep_ -= __d.count(); return *this;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator*=(const rep& rhs) {__rep_ *= rhs; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator/=(const rep& rhs) {__rep_ /= rhs; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator%=(const rep& rhs) {__rep_ %= rhs; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 duration& operator%=(const duration& rhs) {__rep_ %= rhs.count(); return *this;}
+
+    // special values
+
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR duration zero() _NOEXCEPT {return duration(duration_values<rep>::zero());}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR duration min()  _NOEXCEPT {return duration(duration_values<rep>::min());}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR duration max()  _NOEXCEPT {return duration(duration_values<rep>::max());}
+};
+
+typedef duration<long long,         nano> nanoseconds;
+typedef duration<long long,        micro> microseconds;
+typedef duration<long long,        milli> milliseconds;
+typedef duration<long long              > seconds;
+typedef duration<     long, ratio<  60> > minutes;
+typedef duration<     long, ratio<3600> > hours;
+#if _LIBCUDACXX_STD_VER > 11
+typedef duration<     int, ratio_multiply<ratio<24>, hours::period>>         days;
+typedef duration<     int, ratio_multiply<ratio<7>,   days::period>>         weeks;
+typedef duration<     int, ratio_multiply<ratio<146097, 400>, days::period>> years;
+typedef duration<     int, ratio_divide<years::period, ratio<12>>>           months;
+#endif // _LIBCUDACXX_STD_VER > 11
+// Duration ==
+
+template <class _LhsDuration, class _RhsDuration>
+struct __duration_eq
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    bool operator()(const _LhsDuration& __lhs, const _RhsDuration& __rhs) const
+        {
+            typedef typename common_type<_LhsDuration, _RhsDuration>::type _Ct;
+            return _Ct(__lhs).count() == _Ct(__rhs).count();
+        }
+};
+
+template <class _LhsDuration>
+struct __duration_eq<_LhsDuration, _LhsDuration>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    bool operator()(const _LhsDuration& __lhs, const _LhsDuration& __rhs) const
+        {return __lhs.count() == __rhs.count();}
+};
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+bool
+operator==(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    return __duration_eq<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >()(__lhs, __rhs);
+}
+
+// Duration !=
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+bool
+operator!=(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    return !(__lhs == __rhs);
+}
+
+// Duration <
+
+template <class _LhsDuration, class _RhsDuration>
+struct __duration_lt
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    bool operator()(const _LhsDuration& __lhs, const _RhsDuration& __rhs) const
+        {
+            typedef typename common_type<_LhsDuration, _RhsDuration>::type _Ct;
+            return _Ct(__lhs).count() < _Ct(__rhs).count();
+        }
+};
+
+template <class _LhsDuration>
+struct __duration_lt<_LhsDuration, _LhsDuration>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    bool operator()(const _LhsDuration& __lhs, const _LhsDuration& __rhs) const
+        {return __lhs.count() < __rhs.count();}
+};
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+bool
+operator< (const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    return __duration_lt<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >()(__lhs, __rhs);
+}
+
+// Duration >
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+bool
+operator> (const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    return __rhs < __lhs;
+}
+
+// Duration <=
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+bool
+operator<=(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    return !(__rhs < __lhs);
+}
+
+// Duration >=
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+bool
+operator>=(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    return !(__lhs < __rhs);
+}
+
+// Duration +
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type
+operator+(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    typedef typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type _Cd;
+    return _Cd(_Cd(__lhs).count() + _Cd(__rhs).count());
+}
+
+// Duration -
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type
+operator-(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    typedef typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type _Cd;
+    return _Cd(_Cd(__lhs).count() - _Cd(__rhs).count());
+}
+
+// Duration *
+
+template <class _Rep1, class _Period, class _Rep2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    is_convertible<_Rep2, typename common_type<_Rep1, _Rep2>::type>::value,
+    duration<typename common_type<_Rep1, _Rep2>::type, _Period>
+>::type
+operator*(const duration<_Rep1, _Period>& __d, const _Rep2& __s)
+{
+    typedef typename common_type<_Rep1, _Rep2>::type _Cr;
+    typedef duration<_Cr, _Period> _Cd;
+    return _Cd(_Cd(__d).count() * static_cast<_Cr>(__s));
+}
+
+template <class _Rep1, class _Period, class _Rep2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    is_convertible<_Rep1, typename common_type<_Rep1, _Rep2>::type>::value,
+    duration<typename common_type<_Rep1, _Rep2>::type, _Period>
+>::type
+operator*(const _Rep1& __s, const duration<_Rep2, _Period>& __d)
+{
+    return __d * __s;
+}
+
+// Duration /
+
+template <class _Rep1, class _Period, class _Rep2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    !__is_duration<_Rep2>::value &&
+      is_convertible<_Rep2, typename common_type<_Rep1, _Rep2>::type>::value,
+    duration<typename common_type<_Rep1, _Rep2>::type, _Period>
+>::type
+operator/(const duration<_Rep1, _Period>& __d, const _Rep2& __s)
+{
+    typedef typename common_type<_Rep1, _Rep2>::type _Cr;
+    typedef duration<_Cr, _Period> _Cd;
+    return _Cd(_Cd(__d).count() / static_cast<_Cr>(__s));
+}
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename common_type<_Rep1, _Rep2>::type
+operator/(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    typedef typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type _Ct;
+    return _Ct(__lhs).count() / _Ct(__rhs).count();
+}
+
+// Duration %
+
+template <class _Rep1, class _Period, class _Rep2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    !__is_duration<_Rep2>::value &&
+      is_convertible<_Rep2, typename common_type<_Rep1, _Rep2>::type>::value,
+    duration<typename common_type<_Rep1, _Rep2>::type, _Period>
+>::type
+operator%(const duration<_Rep1, _Period>& __d, const _Rep2& __s)
+{
+    typedef typename common_type<_Rep1, _Rep2>::type _Cr;
+    typedef duration<_Cr, _Period> _Cd;
+    return _Cd(_Cd(__d).count() % static_cast<_Cr>(__s));
+}
+
+template <class _Rep1, class _Period1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type
+operator%(const duration<_Rep1, _Period1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    typedef typename common_type<_Rep1, _Rep2>::type _Cr;
+    typedef typename common_type<duration<_Rep1, _Period1>, duration<_Rep2, _Period2> >::type _Cd;
+    return _Cd(static_cast<_Cr>(_Cd(__lhs).count()) % static_cast<_Cr>(_Cd(__rhs).count()));
+}
+
+//////////////////////////////////////////////////////////
+///////////////////// time_point /////////////////////////
+//////////////////////////////////////////////////////////
+
+template <class _Clock, class _Duration = typename _Clock::duration>
+class _LIBCUDACXX_TEMPLATE_VIS time_point
+{
+    static_assert(__is_duration<_Duration>::value,
+                  "Second template parameter of time_point must be a std::chrono::duration");
+public:
+    typedef _Clock                    clock;
+    typedef _Duration                 duration;
+    typedef typename duration::rep    rep;
+    typedef typename duration::period period;
+private:
+    duration __d_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 time_point() : __d_(duration::zero()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 explicit time_point(const duration& __d) : __d_(__d) {}
+
+    // conversions
+    template <class _Duration2>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    time_point(const time_point<clock, _Duration2>& t,
+        typename enable_if
+        <
+            is_convertible<_Duration2, duration>::value
+        >::type* = 0)
+            : __d_(t.time_since_epoch()) {}
+
+    // observer
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 duration time_since_epoch() const {return __d_;}
+
+    // arithmetic
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 time_point& operator+=(const duration& __d) {__d_ += __d; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 time_point& operator-=(const duration& __d) {__d_ -= __d; return *this;}
+
+    // special values
+
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR time_point min() _NOEXCEPT {return time_point(duration::min());}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR time_point max() _NOEXCEPT {return time_point(duration::max());}
+};
+
+} // chrono
+
+template <class _Clock, class _Duration1, class _Duration2>
+struct _LIBCUDACXX_TEMPLATE_VIS common_type<chrono::time_point<_Clock, _Duration1>,
+                                         chrono::time_point<_Clock, _Duration2> >
+{
+    typedef chrono::time_point<_Clock, typename common_type<_Duration1, _Duration2>::type> type;
+};
+
+namespace chrono {
+
+template <class _ToDuration, class _Clock, class _Duration>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+time_point<_Clock, _ToDuration>
+time_point_cast(const time_point<_Clock, _Duration>& __t)
+{
+    return time_point<_Clock, _ToDuration>(_CUDA_VSTD::chrono::duration_cast<_ToDuration>(__t.time_since_epoch()));
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _ToDuration, class _Clock, class _Duration>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    time_point<_Clock, _ToDuration>
+>::type
+floor(const time_point<_Clock, _Duration>& __t)
+{
+    return time_point<_Clock, _ToDuration>{floor<_ToDuration>(__t.time_since_epoch())};
+}
+
+template <class _ToDuration, class _Clock, class _Duration>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    time_point<_Clock, _ToDuration>
+>::type
+ceil(const time_point<_Clock, _Duration>& __t)
+{
+    return time_point<_Clock, _ToDuration>{ceil<_ToDuration>(__t.time_since_epoch())};
+}
+
+template <class _ToDuration, class _Clock, class _Duration>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    __is_duration<_ToDuration>::value,
+    time_point<_Clock, _ToDuration>
+>::type
+round(const time_point<_Clock, _Duration>& __t)
+{
+    return time_point<_Clock, _ToDuration>{round<_ToDuration>(__t.time_since_epoch())};
+}
+
+template <class _Rep, class _Period>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if
+<
+    numeric_limits<_Rep>::is_signed,
+    duration<_Rep, _Period>
+>::type
+abs(duration<_Rep, _Period> __d)
+{
+    return __d >= __d.zero() ? +__d : -__d;
+}
+#endif // _LIBCUDACXX_STD_VER > 11
+
+// time_point ==
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator==(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return __lhs.time_since_epoch() == __rhs.time_since_epoch();
+}
+
+// time_point !=
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator!=(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return !(__lhs == __rhs);
+}
+
+// time_point <
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator<(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return __lhs.time_since_epoch() < __rhs.time_since_epoch();
+}
+
+// time_point >
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator>(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return __rhs < __lhs;
+}
+
+// time_point <=
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator<=(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return !(__rhs < __lhs);
+}
+
+// time_point >=
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator>=(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return !(__lhs < __rhs);
+}
+
+// time_point operator+(time_point x, duration y);
+
+template <class _Clock, class _Duration1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+time_point<_Clock, typename common_type<_Duration1, duration<_Rep2, _Period2> >::type>
+operator+(const time_point<_Clock, _Duration1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    typedef time_point<_Clock, typename common_type<_Duration1, duration<_Rep2, _Period2> >::type> _Tr;
+    return _Tr (__lhs.time_since_epoch() + __rhs);
+}
+
+// time_point operator+(duration x, time_point y);
+
+template <class _Rep1, class _Period1, class _Clock, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+time_point<_Clock, typename common_type<duration<_Rep1, _Period1>, _Duration2>::type>
+operator+(const duration<_Rep1, _Period1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return __rhs + __lhs;
+}
+
+// time_point operator-(time_point x, duration y);
+
+template <class _Clock, class _Duration1, class _Rep2, class _Period2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+time_point<_Clock, typename common_type<_Duration1, duration<_Rep2, _Period2> >::type>
+operator-(const time_point<_Clock, _Duration1>& __lhs, const duration<_Rep2, _Period2>& __rhs)
+{
+    typedef time_point<_Clock, typename common_type<_Duration1, duration<_Rep2, _Period2> >::type> _Ret;
+    return _Ret(__lhs.time_since_epoch() -__rhs);
+}
+
+// duration operator-(time_point x, time_point y);
+
+template <class _Clock, class _Duration1, class _Duration2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename common_type<_Duration1, _Duration2>::type
+operator-(const time_point<_Clock, _Duration1>& __lhs, const time_point<_Clock, _Duration2>& __rhs)
+{
+    return __lhs.time_since_epoch() - __rhs.time_since_epoch();
+}
+
+//////////////////////////////////////////////////////////
+/////////////////////// clocks ///////////////////////////
+//////////////////////////////////////////////////////////
+
+#ifdef __cuda_std__
+#define _LIBCUDACXX_CLOCK_API __host__ __device__
+#else
+#define _LIBCUDACXX_CLOCK_API
+#endif
+
+class _LIBCUDACXX_TYPE_VIS system_clock
+{
+public:
+    typedef _LIBCUDACXX_SYS_CLOCK_DURATION   duration;
+    typedef duration::rep                    rep;
+    typedef duration::period                 period;
+    typedef chrono::time_point<system_clock> time_point;
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const bool is_steady = false;
+
+    _LIBCUDACXX_CLOCK_API
+    static time_point now() _NOEXCEPT;
+    _LIBCUDACXX_CLOCK_API
+    static time_t     to_time_t  (const time_point& __t) _NOEXCEPT;
+    _LIBCUDACXX_CLOCK_API
+    static time_point from_time_t(time_t __t) _NOEXCEPT;
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_MONOTONIC_CLOCK
+class _LIBCUDACXX_TYPE_VIS steady_clock
+{
+public:
+    typedef nanoseconds                                   duration;
+    typedef duration::rep                                 rep;
+    typedef duration::period                              period;
+    typedef chrono::time_point<steady_clock, duration>    time_point;
+    static _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const bool is_steady = true;
+
+    static time_point now() _NOEXCEPT;
+};
+
+typedef steady_clock high_resolution_clock;
+#else
+typedef system_clock high_resolution_clock;
+#endif
+
+#undef _LIBCUDACXX_CLOCK_API
+
+#if _LIBCUDACXX_STD_VER > 11
+
+// [time.clock.file], type file_clock
+using file_clock = _CUDA_VSTD_FS::_FilesystemClock;
+
+template<class _Duration>
+using file_time = time_point<file_clock, _Duration>;
+
+
+template <class _Duration>
+using sys_time    = time_point<system_clock, _Duration>;
+using sys_seconds = sys_time<seconds>;
+using sys_days    = sys_time<days>;
+
+struct local_t {};
+template<class Duration>
+using local_time  = time_point<local_t, Duration>;
+using local_seconds = local_time<seconds>;
+using local_days    = local_time<days>;
+
+struct last_spec { explicit last_spec() = default; };
+
+class day {
+private:
+    unsigned char __d;
+public:
+    day() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit inline constexpr day(unsigned __val) noexcept : __d(static_cast<unsigned char>(__val)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr day& operator++()    noexcept { ++__d; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr day  operator++(int) noexcept { day __tmp = *this; ++(*this); return __tmp; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr day& operator--()    noexcept { --__d; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr day  operator--(int) noexcept { day __tmp = *this; --(*this); return __tmp; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr day& operator+=(const days& __dd) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr day& operator-=(const days& __dd) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit inline constexpr operator unsigned() const noexcept { return __d; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept { return __d >= 1 && __d <= 31; }
+  };
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const day& __lhs, const day& __rhs) noexcept
+{ return static_cast<unsigned>(__lhs) == static_cast<unsigned>(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const day& __lhs, const day& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const day& __lhs, const day& __rhs) noexcept
+{ return static_cast<unsigned>(__lhs) <  static_cast<unsigned>(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const day& __lhs, const day& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const day& __lhs, const day& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const day& __lhs, const day& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+day operator+ (const day& __lhs, const days& __rhs) noexcept
+{ return day(static_cast<unsigned>(__lhs) + static_cast<unsigned>(__rhs.count())); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+day operator+ (const days& __lhs, const day& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+day operator- (const day& __lhs, const days& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+days operator-(const day& __lhs, const day& __rhs) noexcept
+{ return days(static_cast<int>(static_cast<unsigned>(__lhs)) -
+              static_cast<int>(static_cast<unsigned>(__rhs))); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr day& day::operator+=(const days& __dd) noexcept
+{ *this = *this + __dd; return *this; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr day& day::operator-=(const days& __dd) noexcept
+{ *this = *this - __dd; return *this; }
+
+
+class month {
+private:
+    unsigned char __m;
+public:
+    month() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit inline constexpr month(unsigned __val) noexcept : __m(static_cast<unsigned char>(__val)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr month& operator++()    noexcept { ++__m; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr month  operator++(int) noexcept { month __tmp = *this; ++(*this); return __tmp; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr month& operator--()    noexcept { --__m; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr month  operator--(int) noexcept { month __tmp = *this; --(*this); return __tmp; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr month& operator+=(const months& __m1) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr month& operator-=(const months& __m1) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit inline constexpr operator unsigned() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept { return __m >= 1 && __m <= 12; }
+};
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const month& __lhs, const month& __rhs) noexcept
+{ return static_cast<unsigned>(__lhs) == static_cast<unsigned>(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const month& __lhs, const month& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const month& __lhs, const month& __rhs) noexcept
+{ return static_cast<unsigned>(__lhs)  < static_cast<unsigned>(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const month& __lhs, const month& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const month& __lhs, const month& __rhs) noexcept
+{ return !(__rhs < __lhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const month& __lhs, const month& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month operator+ (const month& __lhs, const months& __rhs) noexcept
+{
+    auto const __mu = static_cast<long long>(static_cast<unsigned>(__lhs)) + (__rhs.count() - 1);
+    auto const __yr = (__mu >= 0 ? __mu : __mu - 11) / 12;
+    return month{static_cast<unsigned>(__mu - __yr * 12 + 1)};
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month operator+ (const months& __lhs, const month& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month operator- (const month& __lhs, const months& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+months operator-(const month& __lhs, const month& __rhs) noexcept
+{
+    auto const __dm = static_cast<unsigned>(__lhs) - static_cast<unsigned>(__rhs);
+    return months(__dm <= 11 ? __dm : __dm + 12);
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr month& month::operator+=(const months& __dm) noexcept
+{ *this = *this + __dm; return *this; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr month& month::operator-=(const months& __dm) noexcept
+{ *this = *this - __dm; return *this; }
+
+
+class year {
+private:
+    short __y;
+public:
+    year() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit inline constexpr year(int __val) noexcept : __y(static_cast<short>(__val)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year& operator++()    noexcept { ++__y; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year  operator++(int) noexcept { year __tmp = *this; ++(*this); return __tmp; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year& operator--()    noexcept { --__y; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year  operator--(int) noexcept { year __tmp = *this; --(*this); return __tmp; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr year& operator+=(const years& __dy) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr year& operator-=(const years& __dy) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year operator+() const noexcept { return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year operator-() const noexcept { return year{-__y}; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool is_leap() const noexcept { return __y % 4 == 0 && (__y % 100 != 0 || __y % 400 == 0); }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit inline constexpr operator int() const noexcept { return __y; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+           constexpr bool ok() const noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static inline constexpr year min() noexcept { return year{-32767}; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static inline constexpr year max() noexcept { return year{ 32767}; }
+};
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const year& __lhs, const year& __rhs) noexcept
+{ return static_cast<int>(__lhs) == static_cast<int>(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const year& __lhs, const year& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const year& __lhs, const year& __rhs) noexcept
+{ return static_cast<int>(__lhs)  < static_cast<int>(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const year& __lhs, const year& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const year& __lhs, const year& __rhs) noexcept
+{ return !(__rhs < __lhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const year& __lhs, const year& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year operator+ (const year& __lhs, const years& __rhs) noexcept
+{ return year(static_cast<int>(__lhs) + __rhs.count()); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year operator+ (const years& __lhs, const year& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year operator- (const year& __lhs, const years& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+years operator-(const year& __lhs, const year& __rhs) noexcept
+{ return years{static_cast<int>(__lhs) - static_cast<int>(__rhs)}; }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year& year::operator+=(const years& __dy) noexcept
+{ *this = *this + __dy; return *this; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year& year::operator-=(const years& __dy) noexcept
+{ *this = *this - __dy; return *this; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool year::ok() const noexcept
+{ return static_cast<int>(min()) <= __y && __y <= static_cast<int>(max()); }
+
+class weekday_indexed;
+class weekday_last;
+
+class weekday {
+private:
+    unsigned char __wd;
+public:
+  weekday() = default;
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline explicit constexpr weekday(unsigned __val) noexcept : __wd(static_cast<unsigned char>(__val == 7 ? 0 : __val)) {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr          weekday(const sys_days& __sysd) noexcept
+          : __wd(__weekday_from_days(__sysd.time_since_epoch().count())) {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline explicit constexpr weekday(const local_days& __locd) noexcept
+          : __wd(__weekday_from_days(__locd.time_since_epoch().count())) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr weekday& operator++()    noexcept { __wd = (__wd == 6 ? 0 : __wd + 1); return *this; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr weekday  operator++(int) noexcept { weekday __tmp = *this; ++(*this); return __tmp; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr weekday& operator--()    noexcept { __wd = (__wd == 0 ? 6 : __wd - 1); return *this; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr weekday  operator--(int) noexcept { weekday __tmp = *this; --(*this); return __tmp; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+         constexpr weekday& operator+=(const days& __dd) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY
+         constexpr weekday& operator-=(const days& __dd) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr unsigned c_encoding()   const noexcept { return __wd; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr unsigned iso_encoding() const noexcept { return __wd == 0u ? 7 : __wd; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  inline constexpr bool ok() const noexcept { return __wd <= 6; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+         constexpr weekday_indexed operator[](unsigned __index) const noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY
+         constexpr weekday_last    operator[](last_spec) const noexcept;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr unsigned char __weekday_from_days(int __days) noexcept;
+};
+
+
+// https://howardhinnant.github.io/date_algorithms.html#weekday_from_days
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+unsigned char weekday::__weekday_from_days(int __days) noexcept
+{
+    return static_cast<unsigned char>(
+              static_cast<unsigned>(__days >= -4 ? (__days+4) % 7 : (__days+5) % 7 + 6)
+           );
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const weekday& __lhs, const weekday& __rhs) noexcept
+{ return __lhs.c_encoding() == __rhs.c_encoding(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const weekday& __lhs, const weekday& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const weekday& __lhs, const weekday& __rhs) noexcept
+{ return __lhs.c_encoding() < __rhs.c_encoding(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const weekday& __lhs, const weekday& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const weekday& __lhs, const weekday& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const weekday& __lhs, const weekday& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr weekday operator+(const weekday& __lhs, const days& __rhs) noexcept
+{
+    auto const __mu = static_cast<long long>(__lhs.c_encoding()) + __rhs.count();
+    auto const __yr = (__mu >= 0 ? __mu : __mu - 6) / 7;
+    return weekday{static_cast<unsigned>(__mu - __yr * 7)};
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr weekday operator+(const days& __lhs, const weekday& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr weekday operator-(const weekday& __lhs, const days& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr days operator-(const weekday& __lhs, const weekday& __rhs) noexcept
+{
+    // casts are required to work around nvcc bug 3145483
+    const int __wdu = static_cast<int>(__lhs.c_encoding()) - static_cast<int>(__rhs.c_encoding());
+    const int __wk = (__wdu >= 0 ? __wdu : __wdu-6) / 7;
+    return days{__wdu - __wk * 7};
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr weekday& weekday::operator+=(const days& __dd) noexcept
+{ *this = *this + __dd; return *this; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr weekday& weekday::operator-=(const days& __dd) noexcept
+{ *this = *this - __dd; return *this; }
+
+
+class weekday_indexed {
+private:
+    _CUDA_VSTD::chrono::weekday __wd;
+    unsigned char          __idx;
+public:
+    weekday_indexed() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr weekday_indexed(const _CUDA_VSTD::chrono::weekday& __wdval, unsigned __idxval) noexcept
+        : __wd{__wdval}, __idx(static_cast<unsigned char>(__idxval)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr _CUDA_VSTD::chrono::weekday weekday() const noexcept { return __wd; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr unsigned                 index() const noexcept { return __idx; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept { return __wd.ok() && __idx >= 1 && __idx <= 5; }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const weekday_indexed& __lhs, const weekday_indexed& __rhs) noexcept
+{ return __lhs.weekday() == __rhs.weekday() && __lhs.index() == __rhs.index(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const weekday_indexed& __lhs, const weekday_indexed& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+
+class weekday_last {
+private:
+    _CUDA_VSTD::chrono::weekday __wd;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit constexpr weekday_last(const _CUDA_VSTD::chrono::weekday& __val) noexcept
+        : __wd{__val} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr _CUDA_VSTD::chrono::weekday weekday() const noexcept { return __wd; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr bool ok() const noexcept { return __wd.ok(); }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const weekday_last& __lhs, const weekday_last& __rhs) noexcept
+{ return __lhs.weekday() == __rhs.weekday(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const weekday_last& __lhs, const weekday_last& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+weekday_indexed weekday::operator[](unsigned __index) const noexcept { return weekday_indexed{*this, __index}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+weekday_last    weekday::operator[](last_spec) const noexcept { return weekday_last{*this}; }
+
+
+_LIBCUDACXX_INLINE_VAR constexpr last_spec last{};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Sunday{0};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Monday{1};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Tuesday{2};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Wednesday{3};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Thursday{4};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Friday{5};
+_LIBCUDACXX_INLINE_VAR constexpr weekday   Saturday{6};
+
+_LIBCUDACXX_INLINE_VAR constexpr month January{1};
+_LIBCUDACXX_INLINE_VAR constexpr month February{2};
+_LIBCUDACXX_INLINE_VAR constexpr month March{3};
+_LIBCUDACXX_INLINE_VAR constexpr month April{4};
+_LIBCUDACXX_INLINE_VAR constexpr month May{5};
+_LIBCUDACXX_INLINE_VAR constexpr month June{6};
+_LIBCUDACXX_INLINE_VAR constexpr month July{7};
+_LIBCUDACXX_INLINE_VAR constexpr month August{8};
+_LIBCUDACXX_INLINE_VAR constexpr month September{9};
+_LIBCUDACXX_INLINE_VAR constexpr month October{10};
+_LIBCUDACXX_INLINE_VAR constexpr month November{11};
+_LIBCUDACXX_INLINE_VAR constexpr month December{12};
+
+
+class month_day {
+private:
+   chrono::month __m;
+   chrono::day   __d;
+public:
+    month_day() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr month_day(const chrono::month& __mval, const chrono::day& __dval) noexcept
+        : __m{__mval}, __d{__dval} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::day   day()   const noexcept { return __d; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr bool ok() const noexcept;
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool month_day::ok() const noexcept
+{
+    if (!__m.ok()) return false;
+    const unsigned __dval = static_cast<unsigned>(__d);
+    if (__dval < 1 || __dval > 31) return false;
+    if (__dval <= 29) return true;
+//  Now we've got either 30 or 31
+    const unsigned __mval = static_cast<unsigned>(__m);
+    if (__mval == 2) return false;
+    if (__mval == 4 || __mval == 6 || __mval == 9 || __mval == 11)
+        return __dval == 30;
+    return true;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const month_day& __lhs, const month_day& __rhs) noexcept
+{ return __lhs.month() == __rhs.month() && __lhs.day() == __rhs.day(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const month_day& __lhs, const month_day& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_day operator/(const month& __lhs, const day& __rhs) noexcept
+{ return month_day{__lhs, __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr
+month_day operator/(const day& __lhs, const month& __rhs) noexcept
+{ return __rhs / __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_day operator/(const month& __lhs, int __rhs) noexcept
+{ return __lhs / day(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr
+month_day operator/(int __lhs, const day& __rhs) noexcept
+{ return month(__lhs) / __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr
+month_day operator/(const day& __lhs, int __rhs) noexcept
+{ return month(__rhs) / __lhs; }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const month_day& __lhs, const month_day& __rhs) noexcept
+{ return __lhs.month() != __rhs.month() ? __lhs.month() < __rhs.month() : __lhs.day() < __rhs.day(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const month_day& __lhs, const month_day& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const month_day& __lhs, const month_day& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const month_day& __lhs, const month_day& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+
+
+class month_day_last {
+private:
+    chrono::month __m;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit constexpr month_day_last(const chrono::month& __val) noexcept
+        : __m{__val} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept { return __m.ok(); }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const month_day_last& __lhs, const month_day_last& __rhs) noexcept
+{ return __lhs.month() == __rhs.month(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const month_day_last& __lhs, const month_day_last& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const month_day_last& __lhs, const month_day_last& __rhs) noexcept
+{ return __lhs.month() < __rhs.month(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const month_day_last& __lhs, const month_day_last& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const month_day_last& __lhs, const month_day_last& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const month_day_last& __lhs, const month_day_last& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_day_last operator/(const month& __lhs, last_spec) noexcept
+{ return month_day_last{__lhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_day_last operator/(last_spec, const month& __rhs) noexcept
+{ return month_day_last{__rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_day_last operator/(int __lhs, last_spec) noexcept
+{ return month_day_last{month(__lhs)}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_day_last operator/(last_spec, int __rhs) noexcept
+{ return month_day_last{month(__rhs)}; }
+
+
+class month_weekday {
+private:
+    chrono::month __m;
+    chrono::weekday_indexed __wdi;
+public:
+    month_weekday() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr month_weekday(const chrono::month& __mval, const chrono::weekday_indexed& __wdival) noexcept
+        : __m{__mval}, __wdi{__wdival} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month                     month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::weekday_indexed weekday_indexed() const noexcept { return __wdi; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool                                 ok() const noexcept { return __m.ok() && __wdi.ok(); }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const month_weekday& __lhs, const month_weekday& __rhs) noexcept
+{ return __lhs.month() == __rhs.month() && __lhs.weekday_indexed() == __rhs.weekday_indexed(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const month_weekday& __lhs, const month_weekday& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday operator/(const month& __lhs, const weekday_indexed& __rhs) noexcept
+{ return month_weekday{__lhs, __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday operator/(int __lhs, const weekday_indexed& __rhs) noexcept
+{ return month_weekday{month(__lhs), __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday operator/(const weekday_indexed& __lhs, const month& __rhs) noexcept
+{ return month_weekday{__rhs, __lhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday operator/(const weekday_indexed& __lhs, int __rhs) noexcept
+{ return month_weekday{month(__rhs), __lhs}; }
+
+
+class month_weekday_last {
+    chrono::month        __m;
+    chrono::weekday_last __wdl;
+  public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr month_weekday_last(const chrono::month& __mval, const chrono::weekday_last& __wdlval) noexcept
+        : __m{__mval}, __wdl{__wdlval} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month               month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::weekday_last weekday_last() const noexcept { return __wdl; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool                           ok() const noexcept { return __m.ok() && __wdl.ok(); }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const month_weekday_last& __lhs, const month_weekday_last& __rhs) noexcept
+{ return __lhs.month() == __rhs.month() && __lhs.weekday_last() == __rhs.weekday_last(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const month_weekday_last& __lhs, const month_weekday_last& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday_last operator/(const month& __lhs, const weekday_last& __rhs) noexcept
+{ return month_weekday_last{__lhs, __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday_last operator/(int __lhs, const weekday_last& __rhs) noexcept
+{ return month_weekday_last{month(__lhs), __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday_last operator/(const weekday_last& __lhs, const month& __rhs) noexcept
+{ return month_weekday_last{__rhs, __lhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+month_weekday_last operator/(const weekday_last& __lhs, int __rhs) noexcept
+{ return month_weekday_last{month(__rhs), __lhs}; }
+
+
+class year_month {
+    chrono::year  __y;
+    chrono::month __m;
+public:
+    year_month() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month(const chrono::year& __yval, const chrono::month& __mval) noexcept
+        : __y{__yval}, __m{__mval} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::year  year()  const noexcept { return __y; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year_month& operator+=(const months& __dm) noexcept { this->__m += __dm; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year_month& operator-=(const months& __dm) noexcept { this->__m -= __dm; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year_month& operator+=(const years& __dy)  noexcept { this->__y += __dy; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr year_month& operator-=(const years& __dy)  noexcept { this->__y -= __dy; return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept { return __y.ok() && __m.ok(); }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month operator/(const year& __y, const month& __m) noexcept { return year_month{__y, __m}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month operator/(const year& __y, int __m) noexcept { return year_month{__y, month(__m)}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const year_month& __lhs, const year_month& __rhs) noexcept
+{ return __lhs.year() == __rhs.year() && __lhs.month() == __rhs.month(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const year_month& __lhs, const year_month& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const year_month& __lhs, const year_month& __rhs) noexcept
+{ return __lhs.year() != __rhs.year() ? __lhs.year() < __rhs.year() : __lhs.month() < __rhs.month(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const year_month& __lhs, const year_month& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const year_month& __lhs, const year_month& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const year_month& __lhs, const year_month& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr year_month operator+(const year_month& __lhs, const months& __rhs) noexcept
+{
+    int __dmi = static_cast<int>(static_cast<unsigned>(__lhs.month())) - 1 + __rhs.count();
+    const int __dy = (__dmi >= 0 ? __dmi : __dmi-11) / 12;
+    __dmi = __dmi - __dy * 12 + 1;
+    return (__lhs.year() + years(__dy)) / month(static_cast<unsigned>(__dmi));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr year_month operator+(const months& __lhs, const year_month& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr year_month operator+(const year_month& __lhs, const years& __rhs) noexcept
+{ return (__lhs.year() + __rhs) / __lhs.month(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr year_month operator+(const years& __lhs, const year_month& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr months     operator-(const year_month& __lhs, const year_month& __rhs) noexcept
+{ return (__lhs.year() - __rhs.year()) + months(static_cast<unsigned>(__lhs.month()) - static_cast<unsigned>(__rhs.month())); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr year_month operator-(const year_month& __lhs, const months& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr year_month operator-(const year_month& __lhs, const years& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+class year_month_day_last;
+
+class year_month_day {
+private:
+    chrono::year  __y;
+    chrono::month __m;
+    chrono::day   __d;
+public:
+     year_month_day() = default;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr year_month_day(
+            const chrono::year& __yval, const chrono::month& __mval, const chrono::day& __dval) noexcept
+            : __y{__yval}, __m{__mval}, __d{__dval} {}
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr year_month_day(const year_month_day_last& __ymdl) noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr year_month_day(const sys_days& __sysd) noexcept
+            : year_month_day(__from_days(__sysd.time_since_epoch())) {}
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline explicit constexpr year_month_day(const local_days& __locd) noexcept
+            : year_month_day(__from_days(__locd.time_since_epoch())) {}
+
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr year_month_day& operator+=(const months& __dm) noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr year_month_day& operator-=(const months& __dm) noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr year_month_day& operator+=(const years& __dy)  noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr year_month_day& operator-=(const years& __dy)  noexcept;
+
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr chrono::year   year() const noexcept { return __y; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr chrono::month month() const noexcept { return __m; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr chrono::day     day() const noexcept { return __d; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr operator   sys_days() const noexcept          { return   sys_days{__to_days()}; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline explicit constexpr operator local_days() const noexcept { return local_days{__to_days()}; }
+
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr bool             ok() const noexcept;
+
+     _LIBCUDACXX_INLINE_VISIBILITY
+     static constexpr year_month_day __from_days(days __d) noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     constexpr days __to_days() const noexcept;
+};
+
+
+// https://howardhinnant.github.io/date_algorithms.html#civil_from_days
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day
+year_month_day::__from_days(days __d) noexcept
+{
+    static_assert(std::numeric_limits<unsigned>::digits >= 18, "");
+    static_assert(std::numeric_limits<int>::digits >= 20     , "");
+    const int      __z = __d.count() + 719468;
+    const int      __era = (__z >= 0 ? __z : __z - 146096) / 146097;
+    const unsigned __doe = static_cast<unsigned>(__z - __era * 146097);              // [0, 146096]
+    const unsigned __yoe = (__doe - __doe/1460 + __doe/36524 - __doe/146096) / 365;  // [0, 399]
+    const int      __yr = static_cast<int>(__yoe) + __era * 400;
+    const unsigned __doy = __doe - (365 * __yoe + __yoe/4 - __yoe/100);              // [0, 365]
+    const unsigned __mp = (5 * __doy + 2)/153;                                       // [0, 11]
+    const unsigned __dy = __doy - (153 * __mp + 2)/5 + 1;                            // [1, 31]
+    const unsigned __mth = __mp + static_cast<unsigned>(__mp < 10 ? 3 : -9);         // [1, 12]
+    return year_month_day{chrono::year{__yr + (__mth <= 2)}, chrono::month{__mth}, chrono::day{__dy}};
+}
+
+// https://howardhinnant.github.io/date_algorithms.html#days_from_civil
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr days year_month_day::__to_days() const noexcept
+{
+    static_assert(std::numeric_limits<unsigned>::digits >= 18, "");
+    static_assert(std::numeric_limits<int>::digits >= 20     , "");
+
+    // nvcc doesn't allow ODR using constexpr globals. Therefore,
+    // make a temporary initialized from the global
+    auto constexpr __Feb = February;
+    const int      __yr  = static_cast<int>(__y) - (__m <= __Feb);
+    const unsigned __mth = static_cast<unsigned>(__m);
+    const unsigned __dy  = static_cast<unsigned>(__d);
+
+    const int      __era = (__yr >= 0 ? __yr : __yr - 399) / 400;
+    const unsigned __yoe = static_cast<unsigned>(__yr - __era * 400);                  // [0, 399]
+    const unsigned __doy = static_cast<unsigned>(
+        (153 * (__mth + static_cast<unsigned>(__mth > 2 ? -3 : 9)) + 2) / 5 + __dy-1); // [0, 365]
+    const unsigned __doe = __yoe * 365 + __yoe/4 - __yoe/100 + __doy;                  // [0, 146096]
+    return days{__era * 146097 + static_cast<int>(__doe) - 719468};
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const year_month_day& __lhs, const year_month_day& __rhs) noexcept
+{ return __lhs.year() == __rhs.year() && __lhs.month() == __rhs.month() && __lhs.day() == __rhs.day(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const year_month_day& __lhs, const year_month_day& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const year_month_day& __lhs, const year_month_day& __rhs) noexcept
+{
+    if (__lhs.year() < __rhs.year()) return true;
+    if (__lhs.year() > __rhs.year()) return false;
+    if (__lhs.month() < __rhs.month()) return true;
+    if (__lhs.month() > __rhs.month()) return false;
+    return __lhs.day() < __rhs.day();
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const year_month_day& __lhs, const year_month_day& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const year_month_day& __lhs, const year_month_day& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const year_month_day& __lhs, const year_month_day& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator/(const year_month& __lhs, const day& __rhs) noexcept
+{ return year_month_day{__lhs.year(), __lhs.month(), __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator/(const year_month& __lhs, int __rhs) noexcept
+{ return __lhs / day(__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator/(const year& __lhs, const month_day& __rhs) noexcept
+{ return __lhs / __rhs.month() / __rhs.day(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator/(int __lhs, const month_day& __rhs) noexcept
+{ return year(__lhs) / __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator/(const month_day& __lhs, const year& __rhs) noexcept
+{ return __rhs / __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator/(const month_day& __lhs, int __rhs) noexcept
+{ return year(__rhs) / __lhs; }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator+(const year_month_day& __lhs, const months& __rhs) noexcept
+{ return (__lhs.year()/__lhs.month() + __rhs)/__lhs.day(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator+(const months& __lhs, const year_month_day& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator-(const year_month_day& __lhs, const months& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator+(const year_month_day& __lhs, const years& __rhs) noexcept
+{ return (__lhs.year() + __rhs) / __lhs.month() / __lhs.day(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator+(const years& __lhs, const year_month_day& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day operator-(const year_month_day& __lhs, const years& __rhs) noexcept
+{ return __lhs + -__rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day& year_month_day::operator+=(const months& __dm) noexcept { *this = *this + __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day& year_month_day::operator-=(const months& __dm) noexcept { *this = *this - __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day& year_month_day::operator+=(const years& __dy)  noexcept { *this = *this + __dy; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day& year_month_day::operator-=(const years& __dy)  noexcept { *this = *this - __dy; return *this; }
+
+class year_month_day_last {
+private:
+    chrono::year           __y;
+    chrono::month_day_last __mdl;
+public:
+     _LIBCUDACXX_INLINE_VISIBILITY
+     constexpr year_month_day_last(const year& __yval, const month_day_last& __mdlval) noexcept
+        : __y{__yval}, __mdl{__mdlval} {}
+
+     _LIBCUDACXX_INLINE_VISIBILITY
+     constexpr year_month_day_last& operator+=(const months& __m) noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     constexpr year_month_day_last& operator-=(const months& __m) noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     constexpr year_month_day_last& operator+=(const years& __y)  noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     constexpr year_month_day_last& operator-=(const years& __y)  noexcept;
+
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr chrono::year                     year() const noexcept { return __y; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr chrono::month                   month() const noexcept { return __mdl.month(); }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr chrono::month_day_last month_day_last() const noexcept { return __mdl; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+            constexpr chrono::day                       day() const noexcept;
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr operator                     sys_days() const noexcept { return   sys_days{year()/month()/day()}; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline explicit constexpr operator          local_days() const noexcept { return local_days{year()/month()/day()}; }
+     _LIBCUDACXX_INLINE_VISIBILITY
+     inline constexpr bool                               ok() const noexcept { return __y.ok() && __mdl.ok(); }
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+chrono::day year_month_day_last::day() const noexcept
+{
+    constexpr chrono::day __d[] =
+    {
+        chrono::day(31), chrono::day(28), chrono::day(31),
+        chrono::day(30), chrono::day(31), chrono::day(30),
+        chrono::day(31), chrono::day(31), chrono::day(30),
+        chrono::day(31), chrono::day(30), chrono::day(31)
+    };
+
+    // nvcc doesn't allow ODR using constexpr globals. Therefore,
+    // make a temporary initialized from the global
+    auto constexpr __Feb = February;
+    return month() != __Feb || !__y.is_leap() ?
+        __d[static_cast<unsigned>(month()) - 1] : chrono::day{29};
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const year_month_day_last& __lhs, const year_month_day_last& __rhs) noexcept
+{ return __lhs.year() == __rhs.year() && __lhs.month_day_last() == __rhs.month_day_last(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const year_month_day_last& __lhs, const year_month_day_last& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator< (const year_month_day_last& __lhs, const year_month_day_last& __rhs) noexcept
+{
+    if (__lhs.year() < __rhs.year()) return true;
+    if (__lhs.year() > __rhs.year()) return false;
+    return __lhs.month_day_last() < __rhs.month_day_last();
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator> (const year_month_day_last& __lhs, const year_month_day_last& __rhs) noexcept
+{ return __rhs < __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator<=(const year_month_day_last& __lhs, const year_month_day_last& __rhs) noexcept
+{ return !(__rhs < __lhs);}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator>=(const year_month_day_last& __lhs, const year_month_day_last& __rhs) noexcept
+{ return !(__lhs < __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last operator/(const year_month& __lhs, last_spec) noexcept
+{ return year_month_day_last{__lhs.year(), month_day_last{__lhs.month()}}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last operator/(const year& __lhs, const month_day_last& __rhs) noexcept
+{ return year_month_day_last{__lhs, __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last operator/(int __lhs, const month_day_last& __rhs) noexcept
+{ return year_month_day_last{year{__lhs}, __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last operator/(const month_day_last& __lhs, const year& __rhs) noexcept
+{ return __rhs / __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last operator/(const month_day_last& __lhs, int __rhs) noexcept
+{ return year{__rhs} / __lhs; }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day_last operator+(const year_month_day_last& __lhs, const months& __rhs) noexcept
+{ return (__lhs.year() / __lhs.month() + __rhs) / last; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day_last operator+(const months& __lhs, const year_month_day_last& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day_last operator-(const year_month_day_last& __lhs, const months& __rhs) noexcept
+{ return __lhs + (-__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day_last operator+(const year_month_day_last& __lhs, const years& __rhs) noexcept
+{ return year_month_day_last{__lhs.year() + __rhs, __lhs.month_day_last()}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day_last operator+(const years& __lhs, const year_month_day_last& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_day_last operator-(const year_month_day_last& __lhs, const years& __rhs) noexcept
+{ return __lhs + (-__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last& year_month_day_last::operator+=(const months& __dm) noexcept { *this = *this + __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last& year_month_day_last::operator-=(const months& __dm) noexcept { *this = *this - __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last& year_month_day_last::operator+=(const years& __dy)  noexcept { *this = *this + __dy; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day_last& year_month_day_last::operator-=(const years& __dy)  noexcept { *this = *this - __dy; return *this; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_day::year_month_day(const year_month_day_last& __ymdl) noexcept
+    : __y{__ymdl.year()}, __m{__ymdl.month()}, __d{__ymdl.day()} {}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool year_month_day::ok() const noexcept
+{
+    if (!__y.ok() || !__m.ok()) return false;
+    return chrono::day{1} <= __d && __d <= (__y / __m / last).day();
+}
+
+class year_month_weekday {
+    chrono::year            __y;
+    chrono::month           __m;
+    chrono::weekday_indexed __wdi;
+public:
+    year_month_weekday() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday(const chrono::year& __yval, const chrono::month& __mval,
+                               const chrono::weekday_indexed& __wdival) noexcept
+        : __y{__yval}, __m{__mval}, __wdi{__wdival} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday(const sys_days& __sysd) noexcept
+            : year_month_weekday(__from_days(__sysd.time_since_epoch())) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline explicit constexpr year_month_weekday(const local_days& __locd) noexcept
+            : year_month_weekday(__from_days(__locd.time_since_epoch())) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday& operator+=(const months& m) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday& operator-=(const months& m) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday& operator+=(const years& y)  noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday& operator-=(const years& y)  noexcept;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::year                       year() const noexcept { return __y; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month                     month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::weekday                 weekday() const noexcept { return __wdi.weekday(); }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr unsigned                          index() const noexcept { return __wdi.index(); }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::weekday_indexed weekday_indexed() const noexcept { return __wdi; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr                       operator sys_days() const noexcept { return   sys_days{__to_days()}; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline explicit constexpr operator            local_days() const noexcept { return local_days{__to_days()}; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept
+    {
+        if (!__y.ok() || !__m.ok() || !__wdi.ok()) return false;
+    //  TODO: make sure it's a valid date
+        return true;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr year_month_weekday __from_days(days __d) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr days __to_days() const noexcept;
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday year_month_weekday::__from_days(days __d) noexcept
+{
+    const sys_days      __sysd{__d};
+    const chrono::weekday __wd = chrono::weekday(__sysd);
+    const year_month_day __ymd = year_month_day(__sysd);
+    return year_month_weekday{__ymd.year(), __ymd.month(),
+                              __wd[(static_cast<unsigned>(__ymd.day())-1)/7+1]};
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+days year_month_weekday::__to_days() const noexcept
+{
+    const sys_days __sysd = sys_days(__y/__m/1);
+    return (__sysd + (__wdi.weekday() - chrono::weekday(__sysd) + days{(__wdi.index()-1)*7}))
+                .time_since_epoch();
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const year_month_weekday& __lhs, const year_month_weekday& __rhs) noexcept
+{ return __lhs.year() == __rhs.year() && __lhs.month() == __rhs.month() && __lhs.weekday_indexed() == __rhs.weekday_indexed(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const year_month_weekday& __lhs, const year_month_weekday& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator/(const year_month& __lhs, const weekday_indexed& __rhs) noexcept
+{ return year_month_weekday{__lhs.year(), __lhs.month(), __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator/(const year& __lhs, const month_weekday& __rhs) noexcept
+{ return year_month_weekday{__lhs, __rhs.month(), __rhs.weekday_indexed()}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator/(int __lhs, const month_weekday& __rhs) noexcept
+{ return year(__lhs) / __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator/(const month_weekday& __lhs, const year& __rhs) noexcept
+{ return __rhs / __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator/(const month_weekday& __lhs, int __rhs) noexcept
+{ return year(__rhs) / __lhs; }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator+(const year_month_weekday& __lhs, const months& __rhs) noexcept
+{ return (__lhs.year() / __lhs.month() + __rhs) / __lhs.weekday_indexed(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator+(const months& __lhs, const year_month_weekday& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator-(const year_month_weekday& __lhs, const months& __rhs) noexcept
+{ return __lhs + (-__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator+(const year_month_weekday& __lhs, const years& __rhs) noexcept
+{ return year_month_weekday{__lhs.year() + __rhs, __lhs.month(), __lhs.weekday_indexed()}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator+(const years& __lhs, const year_month_weekday& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday operator-(const year_month_weekday& __lhs, const years& __rhs) noexcept
+{ return __lhs + (-__rhs); }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday& year_month_weekday::operator+=(const months& __dm) noexcept { *this = *this + __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday& year_month_weekday::operator-=(const months& __dm) noexcept { *this = *this - __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday& year_month_weekday::operator+=(const years& __dy)  noexcept { *this = *this + __dy; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday& year_month_weekday::operator-=(const years& __dy)  noexcept { *this = *this - __dy; return *this; }
+
+class year_month_weekday_last {
+private:
+    chrono::year         __y;
+    chrono::month        __m;
+    chrono::weekday_last __wdl;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday_last(const chrono::year& __yval, const chrono::month& __mval,
+                                      const chrono::weekday_last& __wdlval) noexcept
+                : __y{__yval}, __m{__mval}, __wdl{__wdlval} {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday_last& operator+=(const months& __dm) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday_last& operator-=(const months& __dm) noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday_last& operator+=(const years& __dy)  noexcept;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr year_month_weekday_last& operator-=(const years& __dy)  noexcept;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::year                 year() const noexcept { return __y; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::month               month() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::weekday           weekday() const noexcept { return __wdl.weekday(); }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr chrono::weekday_last weekday_last() const noexcept { return __wdl; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr operator                 sys_days() const noexcept { return   sys_days{__to_days()}; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline explicit constexpr operator      local_days() const noexcept { return local_days{__to_days()}; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline constexpr bool ok() const noexcept { return __y.ok() && __m.ok() && __wdl.ok(); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr days __to_days() const noexcept;
+
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+days year_month_weekday_last::__to_days() const noexcept
+{
+    const sys_days __last = sys_days{__y/__m/last};
+    return (__last - (chrono::weekday{__last} - __wdl.weekday())).time_since_epoch();
+
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator==(const year_month_weekday_last& __lhs, const year_month_weekday_last& __rhs) noexcept
+{ return __lhs.year() == __rhs.year() && __lhs.month() == __rhs.month() && __lhs.weekday_last() == __rhs.weekday_last(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+bool operator!=(const year_month_weekday_last& __lhs, const year_month_weekday_last& __rhs) noexcept
+{ return !(__lhs == __rhs); }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator/(const year_month& __lhs, const weekday_last& __rhs) noexcept
+{ return year_month_weekday_last{__lhs.year(), __lhs.month(), __rhs}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator/(const year& __lhs, const month_weekday_last& __rhs) noexcept
+{ return year_month_weekday_last{__lhs, __rhs.month(), __rhs.weekday_last()}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator/(int __lhs, const month_weekday_last& __rhs) noexcept
+{ return year(__lhs) / __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator/(const month_weekday_last& __lhs, const year& __rhs) noexcept
+{ return __rhs / __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator/(const month_weekday_last& __lhs, int __rhs) noexcept
+{ return year(__rhs) / __lhs; }
+
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator+(const year_month_weekday_last& __lhs, const months& __rhs) noexcept
+{ return (__lhs.year() / __lhs.month() + __rhs) / __lhs.weekday_last(); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator+(const months& __lhs, const year_month_weekday_last& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator-(const year_month_weekday_last& __lhs, const months& __rhs) noexcept
+{ return __lhs + (-__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator+(const year_month_weekday_last& __lhs, const years& __rhs) noexcept
+{ return year_month_weekday_last{__lhs.year() + __rhs, __lhs.month(), __lhs.weekday_last()}; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator+(const years& __lhs, const year_month_weekday_last& __rhs) noexcept
+{ return __rhs + __lhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr
+year_month_weekday_last operator-(const year_month_weekday_last& __lhs, const years& __rhs) noexcept
+{ return __lhs + (-__rhs); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday_last& year_month_weekday_last::operator+=(const months& __dm) noexcept { *this = *this + __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday_last& year_month_weekday_last::operator-=(const months& __dm) noexcept { *this = *this - __dm; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday_last& year_month_weekday_last::operator+=(const years& __dy)  noexcept { *this = *this + __dy; return *this; }
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr year_month_weekday_last& year_month_weekday_last::operator-=(const years& __dy)  noexcept { *this = *this - __dy; return *this; }
+
+template <class _Duration>
+class hh_mm_ss
+{
+private:
+    static_assert(__is_duration<_Duration>::value, "template parameter of hh_mm_ss must be a std::chrono::duration");
+    using __CommonType = typename common_type<_Duration, chrono::seconds>::type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr uint64_t __pow10(unsigned __exp)
+    {
+        uint64_t __ret = 1;
+        for (unsigned __i = 0; __i < __exp; ++__i)
+            __ret *= 10U;
+        return __ret;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr unsigned __width(uint64_t __n, uint64_t __d = 10, unsigned __w = 0)
+    {
+        if (__n >= 2 && __d != 0 && __w < 19)
+            return 1 + __width(__n, __d % __n * 10, __w+1);
+        return 0;
+    }
+
+public:
+    static unsigned constexpr fractional_width = __width(__CommonType::period::den) < 19 ?
+                                                 __width(__CommonType::period::den) : 6u;
+    using precision = duration<typename __CommonType::rep, ratio<1, __pow10(fractional_width)>>;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr hh_mm_ss() noexcept : hh_mm_ss{_Duration::zero()} {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit hh_mm_ss(_Duration __d) noexcept :
+        __is_neg(__d < _Duration(0)),
+        __h(duration_cast<chrono::hours>  (abs(__d))),
+        __m(duration_cast<chrono::minutes>(abs(__d) - hours())),
+        __s(duration_cast<chrono::seconds>(abs(__d) - hours() - minutes())),
+        __f(duration_cast<precision>      (abs(__d) - hours() - minutes() - seconds()))
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr bool is_negative()        const noexcept { return __is_neg; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::hours hours()     const noexcept { return __h; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::minutes minutes() const noexcept { return __m; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::seconds seconds() const noexcept { return __s; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr precision subseconds()    const noexcept { return __f; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr precision to_duration() const noexcept
+    {
+        auto __dur = __h + __m + __s + __f;
+        return __is_neg ? -__dur : __dur;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit operator precision() const noexcept { return to_duration(); }
+
+private:
+    bool            __is_neg;
+    chrono::hours   __h;
+    chrono::minutes __m;
+    chrono::seconds __s;
+    precision       __f;
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_am(const hours& __h) noexcept { return __h >= hours( 0) && __h < hours(12); }
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_pm(const hours& __h) noexcept { return __h >= hours(12) && __h < hours(24); }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr hours make12(const hours& __h) noexcept
+{
+    if      (__h == hours( 0)) return hours(12);
+    else if (__h <= hours(12)) return __h;
+    else                       return __h - hours(12);
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr hours make24(const hours& __h, bool __is_pm) noexcept
+{
+    if (__is_pm)
+        return __h == hours(12) ? __h : __h + hours(12);
+    else
+        return __h == hours(12) ? hours(0) : __h;
+}
+#endif // _LIBCUDACXX_STD_VER > 11
+} // chrono
+#if _LIBCUDACXX_STD_VER > 11
+
+// GCC 5 and 6 warn (and then error) on us using the standard reserved UDL names,
+// but have no way of disabling that. Use the system_header pragma on those GCC versions
+// for the remainder of this file, even if it has been requested to disable the pragma
+// earlier.
+#if defined(_LIBCUDACXX_COMPILER_GCC) && (__GNUC__ == 5 || __GNUC__ == 6)
+#pragma GCC system_header
+#endif
+
+// Suffixes for duration literals [time.duration.literals]
+inline namespace literals
+{
+  inline namespace chrono_literals
+  {
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::hours operator""h(unsigned long long __h)
+    {
+        return chrono::hours(static_cast<chrono::hours::rep>(__h));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, ratio<3600,1>> operator""h(long double __h)
+    {
+        return chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, ratio<3600,1>>(__h);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::minutes operator""min(unsigned long long __m)
+    {
+        return chrono::minutes(static_cast<chrono::minutes::rep>(__m));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, ratio<60,1>> operator""min(long double __m)
+    {
+        return chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, ratio<60,1>> (__m);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::seconds operator""s(unsigned long long __s)
+    {
+        return chrono::seconds(static_cast<chrono::seconds::rep>(__s));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T> operator""s(long double __s)
+    {
+        return chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T> (__s);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::milliseconds operator""ms(unsigned long long __ms)
+    {
+        return chrono::milliseconds(static_cast<chrono::milliseconds::rep>(__ms));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, milli> operator""ms(long double __ms)
+    {
+        return chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, milli>(__ms);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::microseconds operator""us(unsigned long long __us)
+    {
+        return chrono::microseconds(static_cast<chrono::microseconds::rep>(__us));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, micro> operator""us(long double __us)
+    {
+        return chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, micro> (__us);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::nanoseconds operator""ns(unsigned long long __ns)
+    {
+        return chrono::nanoseconds(static_cast<chrono::nanoseconds::rep>(__ns));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, nano> operator""ns(long double __ns)
+    {
+        return chrono::duration<_LIBCUDACXX_CHRONO_LITERAL_INTERNAL_T, nano> (__ns);
+    }
+
+#if _LIBCUDACXX_STD_VER > 17 && !defined(_LIBCUDACXX_HAS_NO_CXX20_CHRONO_LITERALS)
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::day operator""d(unsigned long long __d) noexcept
+    {
+        return chrono::day(static_cast<unsigned>(__d));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr chrono::year operator""y(unsigned long long __y) noexcept
+    {
+        return chrono::year(static_cast<int>(__y));
+    }
+#endif //_LIBCUDACXX_STD_VER > 17 && !defined(_LIBCUDACXX_HAS_NO_CXX20_CHRONO_LITERALS)
+}}
+
+namespace chrono { // hoist the literals into namespace std::chrono
+   using namespace literals::chrono_literals;
+}
+
+#endif // _LIBCUDACXX_STD_VER > 11
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+_LIBCUDACXX_BEGIN_NAMESPACE_FILESYSTEM
+struct _FilesystemClock {
+#if !defined(_LIBCUDACXX_HAS_NO_INT128)
+  typedef __int128_t rep;
+  typedef nano period;
+#else
+  typedef long long rep;
+  typedef nano period;
+#endif
+
+  typedef chrono::duration<rep, period> duration;
+  typedef chrono::time_point<_FilesystemClock> time_point;
+
+  _LIBCUDACXX_EXPORTED_FROM_ABI
+  static _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const bool is_steady = false;
+
+  _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_FUNC_VIS static time_point now() noexcept;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static time_t to_time_t(const time_point& __t) noexcept {
+      typedef chrono::duration<rep> __secs;
+      return time_t(
+          chrono::duration_cast<__secs>(__t.time_since_epoch()).count());
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static time_point from_time_t(time_t __t) noexcept {
+      typedef chrono::duration<rep> __secs;
+      return time_point(__secs(__t));
+  }
+};
+_LIBCUDACXX_END_NAMESPACE_FILESYSTEM
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+#endif
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX_CHRONO
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes
new file mode 100644
index 000000000000..50ea06403a8e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cinttypes
@@ -0,0 +1,257 @@
+// -*- C++ -*-
+//===--------------------------- cinttypes --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CINTTYPES
+#define _LIBCUDACXX_CINTTYPES
+
+/*
+    cinttypes synopsis
+
+This entire header is C99 / C++0X
+
+#include <cstdint>  // <cinttypes> includes <cstdint>
+
+Macros:
+
+    PRId8
+    PRId16
+    PRId32
+    PRId64
+
+    PRIdLEAST8
+    PRIdLEAST16
+    PRIdLEAST32
+    PRIdLEAST64
+
+    PRIdFAST8
+    PRIdFAST16
+    PRIdFAST32
+    PRIdFAST64
+
+    PRIdMAX
+    PRIdPTR
+
+    PRIi8
+    PRIi16
+    PRIi32
+    PRIi64
+
+    PRIiLEAST8
+    PRIiLEAST16
+    PRIiLEAST32
+    PRIiLEAST64
+
+    PRIiFAST8
+    PRIiFAST16
+    PRIiFAST32
+    PRIiFAST64
+
+    PRIiMAX
+    PRIiPTR
+
+    PRIo8
+    PRIo16
+    PRIo32
+    PRIo64
+
+    PRIoLEAST8
+    PRIoLEAST16
+    PRIoLEAST32
+    PRIoLEAST64
+
+    PRIoFAST8
+    PRIoFAST16
+    PRIoFAST32
+    PRIoFAST64
+
+    PRIoMAX
+    PRIoPTR
+
+    PRIu8
+    PRIu16
+    PRIu32
+    PRIu64
+
+    PRIuLEAST8
+    PRIuLEAST16
+    PRIuLEAST32
+    PRIuLEAST64
+
+    PRIuFAST8
+    PRIuFAST16
+    PRIuFAST32
+    PRIuFAST64
+
+    PRIuMAX
+    PRIuPTR
+
+    PRIx8
+    PRIx16
+    PRIx32
+    PRIx64
+
+    PRIxLEAST8
+    PRIxLEAST16
+    PRIxLEAST32
+    PRIxLEAST64
+
+    PRIxFAST8
+    PRIxFAST16
+    PRIxFAST32
+    PRIxFAST64
+
+    PRIxMAX
+    PRIxPTR
+
+    PRIX8
+    PRIX16
+    PRIX32
+    PRIX64
+
+    PRIXLEAST8
+    PRIXLEAST16
+    PRIXLEAST32
+    PRIXLEAST64
+
+    PRIXFAST8
+    PRIXFAST16
+    PRIXFAST32
+    PRIXFAST64
+
+    PRIXMAX
+    PRIXPTR
+
+    SCNd8
+    SCNd16
+    SCNd32
+    SCNd64
+
+    SCNdLEAST8
+    SCNdLEAST16
+    SCNdLEAST32
+    SCNdLEAST64
+
+    SCNdFAST8
+    SCNdFAST16
+    SCNdFAST32
+    SCNdFAST64
+
+    SCNdMAX
+    SCNdPTR
+
+    SCNi8
+    SCNi16
+    SCNi32
+    SCNi64
+
+    SCNiLEAST8
+    SCNiLEAST16
+    SCNiLEAST32
+    SCNiLEAST64
+
+    SCNiFAST8
+    SCNiFAST16
+    SCNiFAST32
+    SCNiFAST64
+
+    SCNiMAX
+    SCNiPTR
+
+    SCNo8
+    SCNo16
+    SCNo32
+    SCNo64
+
+    SCNoLEAST8
+    SCNoLEAST16
+    SCNoLEAST32
+    SCNoLEAST64
+
+    SCNoFAST8
+    SCNoFAST16
+    SCNoFAST32
+    SCNoFAST64
+
+    SCNoMAX
+    SCNoPTR
+
+    SCNu8
+    SCNu16
+    SCNu32
+    SCNu64
+
+    SCNuLEAST8
+    SCNuLEAST16
+    SCNuLEAST32
+    SCNuLEAST64
+
+    SCNuFAST8
+    SCNuFAST16
+    SCNuFAST32
+    SCNuFAST64
+
+    SCNuMAX
+    SCNuPTR
+
+    SCNx8
+    SCNx16
+    SCNx32
+    SCNx64
+
+    SCNxLEAST8
+    SCNxLEAST16
+    SCNxLEAST32
+    SCNxLEAST64
+
+    SCNxFAST8
+    SCNxFAST16
+    SCNxFAST32
+    SCNxFAST64
+
+    SCNxMAX
+    SCNxPTR
+
+namespace std
+{
+
+Types:
+
+    imaxdiv_t
+
+intmax_t  imaxabs(intmax_t j);
+imaxdiv_t imaxdiv(intmax_t numer, intmax_t denom);
+intmax_t  strtoimax(const char* restrict nptr, char** restrict endptr, int base);
+uintmax_t strtoumax(const char* restrict nptr, char** restrict endptr, int base);
+intmax_t  wcstoimax(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+uintmax_t wcstoumax(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+
+}  // std
+*/
+
+#include <__config>
+#include <cstdint>
+#include <inttypes.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using::imaxdiv_t;
+using::imaxabs;
+using::imaxdiv;
+using::strtoimax;
+using::strtoumax;
+using::wcstoimax;
+using::wcstoumax;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CINTTYPES
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ciso646 b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ciso646
new file mode 100644
index 000000000000..40beccb215b4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ciso646
@@ -0,0 +1,24 @@
+// -*- C++ -*-
+//===--------------------------- ciso646 ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CISO646
+#define _LIBCUDACXX_CISO646
+
+/*
+    ciso646 synopsis
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__pragma_push>
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CISO646
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/climits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/climits
new file mode 100644
index 000000000000..0ff07d9e024d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/climits
@@ -0,0 +1,62 @@
+// -*- C++ -*-
+//===--------------------------- climits ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CLIMITS
+#define _LIBCUDACXX_CLIMITS
+
+/*
+    climits synopsis
+
+Macros:
+
+    CHAR_BIT
+    SCHAR_MIN
+    SCHAR_MAX
+    UCHAR_MAX
+    CHAR_MIN
+    CHAR_MAX
+    MB_LEN_MAX
+    SHRT_MIN
+    SHRT_MAX
+    USHRT_MAX
+    INT_MIN
+    INT_MAX
+    UINT_MAX
+    LONG_MIN
+    LONG_MAX
+    ULONG_MAX
+    LLONG_MIN   // C99
+    LLONG_MAX   // C99
+    ULLONG_MAX  // C99
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <limits.h>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
+#include "support/win32/limits_msvc_win32.h"
+#endif // _LIBCUDACXX_MSVCRT
+
+#if defined(__IBMCPP__)
+#include "support/ibm/limits.h"
+#endif // __IBMCPP__
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX_CLIMITS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/clocale b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/clocale
new file mode 100644
index 000000000000..64d6fe377744
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/clocale
@@ -0,0 +1,54 @@
+// -*- C++ -*-
+//===--------------------------- clocale ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CLOCALE
+#define _LIBCUDACXX_CLOCALE
+
+/*
+    clocale synopsis
+
+Macros:
+
+    LC_ALL
+    LC_COLLATE
+    LC_CTYPE
+    LC_MONETARY
+    LC_NUMERIC
+    LC_TIME
+    NULL
+
+namespace std
+{
+
+struct lconv;
+char* setlocale(int category, const char* locale);
+lconv* localeconv();
+
+}  // std
+
+*/
+
+#include <__config>
+#include <locale.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::lconv;
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_UNSAFE_C_FUNCTIONS
+using ::setlocale;
+#endif
+using ::localeconv;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CLOCALE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cmath b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cmath
new file mode 100644
index 000000000000..515164c7adf0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cmath
@@ -0,0 +1,725 @@
+// -*- C++ -*-
+//===---------------------------- cmath -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CMATH
+#define _LIBCUDACXX_CMATH
+
+/*
+    cmath synopsis
+
+Macros:
+
+    HUGE_VAL
+    HUGE_VALF               // C99
+    HUGE_VALL               // C99
+    INFINITY                // C99
+    NAN                     // C99
+    FP_INFINITE             // C99
+    FP_NAN                  // C99
+    FP_NORMAL               // C99
+    FP_SUBNORMAL            // C99
+    FP_ZERO                 // C99
+    FP_FAST_FMA             // C99
+    FP_FAST_FMAF            // C99
+    FP_FAST_FMAL            // C99
+    FP_ILOGB0               // C99
+    FP_ILOGBNAN             // C99
+    MATH_ERRNO              // C99
+    MATH_ERREXCEPT          // C99
+    math_errhandling        // C99
+
+namespace std
+{
+
+Types:
+
+    float_t                 // C99
+    double_t                // C99
+
+// C90
+
+floating_point abs(floating_point x);
+
+floating_point acos (arithmetic x);
+float          acosf(float x);
+long double    acosl(long double x);
+
+floating_point asin (arithmetic x);
+float          asinf(float x);
+long double    asinl(long double x);
+
+floating_point atan (arithmetic x);
+float          atanf(float x);
+long double    atanl(long double x);
+
+floating_point atan2 (arithmetic y, arithmetic x);
+float          atan2f(float y, float x);
+long double    atan2l(long double y, long double x);
+
+floating_point ceil (arithmetic x);
+float          ceilf(float x);
+long double    ceill(long double x);
+
+floating_point cos (arithmetic x);
+float          cosf(float x);
+long double    cosl(long double x);
+
+floating_point cosh (arithmetic x);
+float          coshf(float x);
+long double    coshl(long double x);
+
+floating_point exp (arithmetic x);
+float          expf(float x);
+long double    expl(long double x);
+
+floating_point fabs (arithmetic x);
+float          fabsf(float x);
+long double    fabsl(long double x);
+
+floating_point floor (arithmetic x);
+float          floorf(float x);
+long double    floorl(long double x);
+
+floating_point fmod (arithmetic x, arithmetic y);
+float          fmodf(float x, float y);
+long double    fmodl(long double x, long double y);
+
+floating_point frexp (arithmetic value, int* exp);
+float          frexpf(float value, int* exp);
+long double    frexpl(long double value, int* exp);
+
+floating_point ldexp (arithmetic value, int exp);
+float          ldexpf(float value, int exp);
+long double    ldexpl(long double value, int exp);
+
+floating_point log (arithmetic x);
+float          logf(float x);
+long double    logl(long double x);
+
+floating_point log10 (arithmetic x);
+float          log10f(float x);
+long double    log10l(long double x);
+
+floating_point modf (floating_point value, floating_point* iptr);
+float          modff(float value, float* iptr);
+long double    modfl(long double value, long double* iptr);
+
+floating_point pow (arithmetic x, arithmetic y);
+float          powf(float x, float y);
+long double    powl(long double x, long double y);
+
+floating_point sin (arithmetic x);
+float          sinf(float x);
+long double    sinl(long double x);
+
+floating_point sinh (arithmetic x);
+float          sinhf(float x);
+long double    sinhl(long double x);
+
+floating_point sqrt (arithmetic x);
+float          sqrtf(float x);
+long double    sqrtl(long double x);
+
+floating_point tan (arithmetic x);
+float          tanf(float x);
+long double    tanl(long double x);
+
+floating_point tanh (arithmetic x);
+float          tanhf(float x);
+long double    tanhl(long double x);
+
+//  C99
+
+bool signbit(arithmetic x);
+
+int fpclassify(arithmetic x);
+
+bool isfinite(arithmetic x);
+bool isinf(arithmetic x);
+bool isnan(arithmetic x);
+bool isnormal(arithmetic x);
+
+bool isgreater(arithmetic x, arithmetic y);
+bool isgreaterequal(arithmetic x, arithmetic y);
+bool isless(arithmetic x, arithmetic y);
+bool islessequal(arithmetic x, arithmetic y);
+bool islessgreater(arithmetic x, arithmetic y);
+bool isunordered(arithmetic x, arithmetic y);
+
+floating_point acosh (arithmetic x);
+float          acoshf(float x);
+long double    acoshl(long double x);
+
+floating_point asinh (arithmetic x);
+float          asinhf(float x);
+long double    asinhl(long double x);
+
+floating_point atanh (arithmetic x);
+float          atanhf(float x);
+long double    atanhl(long double x);
+
+floating_point cbrt (arithmetic x);
+float          cbrtf(float x);
+long double    cbrtl(long double x);
+
+floating_point copysign (arithmetic x, arithmetic y);
+float          copysignf(float x, float y);
+long double    copysignl(long double x, long double y);
+
+floating_point erf (arithmetic x);
+float          erff(float x);
+long double    erfl(long double x);
+
+floating_point erfc (arithmetic x);
+float          erfcf(float x);
+long double    erfcl(long double x);
+
+floating_point exp2 (arithmetic x);
+float          exp2f(float x);
+long double    exp2l(long double x);
+
+floating_point expm1 (arithmetic x);
+float          expm1f(float x);
+long double    expm1l(long double x);
+
+floating_point fdim (arithmetic x, arithmetic y);
+float          fdimf(float x, float y);
+long double    fdiml(long double x, long double y);
+
+floating_point fma (arithmetic x, arithmetic y, arithmetic z);
+float          fmaf(float x, float y, float z);
+long double    fmal(long double x, long double y, long double z);
+
+floating_point fmax (arithmetic x, arithmetic y);
+float          fmaxf(float x, float y);
+long double    fmaxl(long double x, long double y);
+
+floating_point fmin (arithmetic x, arithmetic y);
+float          fminf(float x, float y);
+long double    fminl(long double x, long double y);
+
+floating_point hypot (arithmetic x, arithmetic y);
+float          hypotf(float x, float y);
+long double    hypotl(long double x, long double y);
+
+double       hypot(double x, double y, double z);                // C++17
+float        hypot(float x, float y, float z);                   // C++17
+long double  hypot(long double x, long double y, long double z); // C++17
+
+int ilogb (arithmetic x);
+int ilogbf(float x);
+int ilogbl(long double x);
+
+floating_point lgamma (arithmetic x);
+float          lgammaf(float x);
+long double    lgammal(long double x);
+
+long long llrint (arithmetic x);
+long long llrintf(float x);
+long long llrintl(long double x);
+
+long long llround (arithmetic x);
+long long llroundf(float x);
+long long llroundl(long double x);
+
+floating_point log1p (arithmetic x);
+float          log1pf(float x);
+long double    log1pl(long double x);
+
+floating_point log2 (arithmetic x);
+float          log2f(float x);
+long double    log2l(long double x);
+
+floating_point logb (arithmetic x);
+float          logbf(float x);
+long double    logbl(long double x);
+
+long lrint (arithmetic x);
+long lrintf(float x);
+long lrintl(long double x);
+
+long lround (arithmetic x);
+long lroundf(float x);
+long lroundl(long double x);
+
+double      nan (const char* str);
+float       nanf(const char* str);
+long double nanl(const char* str);
+
+floating_point nearbyint (arithmetic x);
+float          nearbyintf(float x);
+long double    nearbyintl(long double x);
+
+floating_point nextafter (arithmetic x, arithmetic y);
+float          nextafterf(float x, float y);
+long double    nextafterl(long double x, long double y);
+
+floating_point nexttoward (arithmetic x, long double y);
+float          nexttowardf(float x, long double y);
+long double    nexttowardl(long double x, long double y);
+
+floating_point remainder (arithmetic x, arithmetic y);
+float          remainderf(float x, float y);
+long double    remainderl(long double x, long double y);
+
+floating_point remquo (arithmetic x, arithmetic y, int* pquo);
+float          remquof(float x, float y, int* pquo);
+long double    remquol(long double x, long double y, int* pquo);
+
+floating_point rint (arithmetic x);
+float          rintf(float x);
+long double    rintl(long double x);
+
+floating_point round (arithmetic x);
+float          roundf(float x);
+long double    roundl(long double x);
+
+floating_point scalbln (arithmetic x, long ex);
+float          scalblnf(float x, long ex);
+long double    scalblnl(long double x, long ex);
+
+floating_point scalbn (arithmetic x, int ex);
+float          scalbnf(float x, int ex);
+long double    scalbnl(long double x, int ex);
+
+floating_point tgamma (arithmetic x);
+float          tgammaf(float x);
+long double    tgammal(long double x);
+
+floating_point trunc (arithmetic x);
+float          truncf(float x);
+long double    truncl(long double x);
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <math.h>
+#include <version>
+#include <type_traits>
+#include <__pragma_push>
+#endif //stdexcept
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_COMPILER_NVRTC
+#include "climits"
+#define INFINITY __builtin_huge_val()
+#define NAN      __builtin_nan()
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::signbit;
+using ::isfinite;
+using ::isinf;
+using ::isnan;
+
+using ::acos;
+using ::acosf;
+using ::asin;
+using ::asinf;
+using ::atan;
+using ::atanf;
+using ::atan2;
+using ::atan2f;
+using ::ceil;
+using ::ceilf;
+using ::cos;
+using ::cosf;
+using ::cosh;
+using ::coshf;
+
+using ::exp;
+using ::expf;
+
+using ::pow;
+using ::powf;
+
+using ::sin;
+using ::sinf;
+using ::sinh;
+using ::sinhf;
+
+using ::sqrt;
+using ::sqrtf;
+using ::tan;
+using ::tanf;
+
+using ::tanh;
+using ::tanhf;
+
+using ::acosh;
+using ::acoshf;
+using ::asinh;
+using ::asinhf;
+using ::atanh;
+using ::atanhf;
+
+using ::log;
+using ::logf;
+
+using ::hypot;
+using ::hypotf;
+
+#ifndef _AIX
+using ::abs;
+#endif
+
+#ifndef _LIBCUDACXX_COMPILER_NVRTC
+// GCC <= 50X is missing some math intrinsics in the global namespace
+#if !defined(_LIBCUDACXX_MISSING_GCC_MATH_INTRINSICS)
+using ::fpclassify;
+using ::isnormal;
+using ::isgreater;
+using ::isgreaterequal;
+using ::isless;
+using ::islessequal;
+using ::islessgreater;
+using ::isunordered;
+using ::isunordered;
+#endif
+
+using ::float_t;
+using ::double_t;
+
+using ::fabs;
+using ::fabsf;
+using ::floor;
+using ::floorf;
+
+using ::fmod;
+using ::fmodf;
+
+using ::frexp;
+using ::frexpf;
+using ::ldexp;
+using ::ldexpf;
+
+using ::log;
+using ::logf;
+
+using ::log10;
+using ::log10f;
+using ::modf;
+using ::modff;
+
+using ::pow;
+using ::powf;
+
+using ::sin;
+using ::sinf;
+using ::sinh;
+using ::sinhf;
+
+using ::sqrt;
+using ::sqrtf;
+using ::tan;
+using ::tanf;
+
+using ::tanh;
+using ::tanhf;
+
+using ::acosh;
+using ::acoshf;
+using ::asinh;
+using ::asinhf;
+using ::atanh;
+using ::atanhf;
+using ::cbrt;
+using ::cbrtf;
+
+using ::copysign;
+using ::copysignf;
+
+using ::erf;
+using ::erff;
+using ::erfc;
+using ::erfcf;
+using ::exp2;
+using ::exp2f;
+using ::expm1;
+using ::expm1f;
+using ::fdim;
+using ::fdimf;
+using ::fmaf;
+using ::fma;
+using ::fmax;
+using ::fmaxf;
+using ::fmin;
+using ::fminf;
+using ::ilogb;
+using ::ilogbf;
+using ::lgamma;
+using ::lgammaf;
+using ::llrint;
+using ::llrintf;
+using ::llround;
+using ::llroundf;
+using ::log1p;
+using ::log1pf;
+using ::log2;
+using ::log2f;
+using ::logb;
+using ::logbf;
+using ::lrint;
+using ::lrintf;
+using ::lround;
+using ::lroundf;
+
+using ::nan;
+using ::nanf;
+
+using ::nearbyint;
+using ::nearbyintf;
+using ::nextafter;
+using ::nextafterf;
+using ::nexttoward;
+using ::nexttowardf;
+using ::remainder;
+using ::remainderf;
+using ::remquo;
+using ::remquof;
+using ::rint;
+using ::rintf;
+using ::round;
+using ::roundf;
+using ::scalbln;
+using ::scalblnf;
+using ::scalbn;
+using ::scalbnf;
+using ::tgamma;
+using ::tgammaf;
+using ::trunc;
+using ::truncf;
+
+using ::acosl;
+using ::asinl;
+using ::atanl;
+using ::atan2l;
+using ::ceill;
+using ::cosl;
+using ::coshl;
+using ::expl;
+using ::fabsl;
+using ::floorl;
+using ::fmodl;
+using ::frexpl;
+using ::ldexpl;
+using ::logl;
+using ::log10l;
+using ::modfl;
+using ::powl;
+using ::sinl;
+using ::sinhl;
+using ::sqrtl;
+using ::tanl;
+
+using ::tanhl;
+using ::acoshl;
+using ::asinhl;
+using ::atanhl;
+using ::cbrtl;
+
+using ::copysignl;
+
+using ::erfl;
+using ::erfcl;
+using ::exp2l;
+using ::expm1l;
+using ::fdiml;
+using ::fmal;
+using ::fmaxl;
+using ::fminl;
+using ::hypotl;
+using ::ilogbl;
+using ::lgammal;
+using ::llrintl;
+using ::llroundl;
+using ::log1pl;
+using ::log2l;
+using ::logbl;
+using ::lrintl;
+using ::lroundl;
+using ::nanl;
+using ::nearbyintl;
+using ::nextafterl;
+using ::nexttowardl;
+using ::remainderl;
+using ::remquol;
+using ::rintl;
+using ::roundl;
+using ::scalblnl;
+using ::scalbnl;
+using ::tgammal;
+using ::truncl;
+#endif
+
+#if _LIBCUDACXX_STD_VER > 14 && !defined(__cuda_std__)
+inline _LIBCUDACXX_INLINE_VISIBILITY float       hypot(       float x,       float y,       float z ) { return sqrt(x*x + y*y + z*z); }
+inline _LIBCUDACXX_INLINE_VISIBILITY double      hypot(      double x,      double y,      double z ) { return sqrt(x*x + y*y + z*z); }
+#ifdef _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+inline _LIBCUDACXX_INLINE_VISIBILITY long double hypot( long double x, long double y, long double z ) { return sqrt(x*x + y*y + z*z); }
+#endif
+
+template <class _A1, class _A2, class _A3>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename _EnableIf
+<
+    is_arithmetic<_A1>::value &&
+    is_arithmetic<_A2>::value &&
+    is_arithmetic<_A3>::value,
+    __promote<_A1, _A2, _A3>
+>::type
+hypot(_A1 __lcpp_x, _A2 __lcpp_y, _A3 __lcpp_z) _NOEXCEPT
+{
+    typedef typename __promote<_A1, _A2, _A3>::type __result_type;
+    static_assert((!(is_same<_A1, __result_type>::value &&
+                     is_same<_A2, __result_type>::value &&
+                     is_same<_A3, __result_type>::value)), "");
+    return hypot((__result_type)__lcpp_x, (__result_type)__lcpp_y, (__result_type)__lcpp_z);
+}
+#endif
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR typename enable_if<is_floating_point<_A1>::value, bool>::type
+__libcpp_isnan_or_builtin(_A1 __lcpp_x) _NOEXCEPT
+{
+#if defined(__CUDA_ARCH__)
+    return __isnan(__lcpp_x);
+#elif __has_builtin(__builtin_isnan)
+    return __builtin_isnan(__lcpp_x);
+#else
+    return isnan(__lcpp_x);
+#endif
+}
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR typename enable_if<!is_floating_point<_A1>::value, bool>::type
+__libcpp_isnan_or_builtin(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isnan(__lcpp_x);
+}
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR typename enable_if<is_floating_point<_A1>::value, bool>::type
+__libcpp_isinf_or_builtin(_A1 __lcpp_x) _NOEXCEPT
+{
+#if defined(__CUDA_ARCH__)
+    return __isinf(__lcpp_x);
+#elif __has_builtin(__builtin_isnan)
+    return __builtin_isinf(__lcpp_x);
+#else
+    return isinf(__lcpp_x);
+#endif
+}
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR typename enable_if<!is_floating_point<_A1>::value, bool>::type
+__libcpp_isinf_or_builtin(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isinf(__lcpp_x);
+}
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR typename enable_if<is_floating_point<_A1>::value, bool>::type
+__libcpp_isfinite_or_builtin(_A1 __lcpp_x) _NOEXCEPT
+{
+#if !defined(__CUDA_ARCH__) && __has_builtin(__builtin_isnan)
+    return __builtin_isfinite(__lcpp_x);
+#else
+    return isfinite(__lcpp_x);
+#endif
+}
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR typename enable_if<!is_floating_point<_A1>::value, bool>::type
+__libcpp_isfinite_or_builtin(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isfinite(__lcpp_x);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <typename _Fp>
+constexpr
+_Fp __lerp(_Fp __a, _Fp __b, _Fp __t) noexcept {
+    if ((__a <= 0 && __b >= 0) || (__a >= 0 && __b <= 0))
+        return __t * __b + (1 - __t) * __a;
+
+    if (__t == 1) return __b;
+    const _Fp __x = __a + __t * (__b - __a);
+    if (__t > 1 == __b > __a)
+    	return __b < __x ? __x : __b;
+    else
+    	return __x < __b ? __x : __b;
+}
+
+constexpr float
+lerp(float __a, float __b, float __t)                   _NOEXCEPT { return __lerp(__a, __b, __t); }
+
+constexpr double
+lerp(double __a, double __b, double __t)                _NOEXCEPT { return __lerp(__a, __b, __t); }
+
+constexpr long double
+lerp(long double __a, long double __b, long double __t) _NOEXCEPT { return __lerp(__a, __b, __t); }
+
+#endif // _LIBCUDACXX_STD_VER > 17
+
+template <class _IntT, class _FloatT,
+    bool _FloatBigger = (numeric_limits<_FloatT>::digits > numeric_limits<_IntT>::digits),
+    int _Bits = (numeric_limits<_IntT>::digits - numeric_limits<_FloatT>::digits)>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR _IntT __max_representable_int_for_float() _NOEXCEPT {
+  static_assert(is_floating_point<_FloatT>::value, "must be a floating point type");
+  static_assert(is_integral<_IntT>::value, "must be an integral type");
+  static_assert(numeric_limits<_FloatT>::radix == 2, "FloatT has incorrect radix");
+#ifdef _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+  static_assert((_IsSame<_FloatT, float>::value || _IsSame<_FloatT, double>::value
+                 || _IsSame<_FloatT,long double>::value), "unsupported floating point type");
+#else
+  static_assert((_IsSame<_FloatT, float>::value || _IsSame<_FloatT, double>::value),
+                 "unsupported floating point type");
+#endif
+  return _FloatBigger ? numeric_limits<_IntT>::max() :  (numeric_limits<_IntT>::max() >> _Bits << _Bits);
+}
+
+// Convert a floating point number to the specified integral type after
+// clamping to the integral types representable range.
+//
+// The behavior is undefined if `__r` is NaN.
+template <class _IntT, class _RealT>
+_LIBCUDACXX_INLINE_VISIBILITY
+_IntT __clamp_to_integral(_RealT __r) _NOEXCEPT {
+  using _Lim = std::numeric_limits<_IntT>;
+  const _IntT _MaxVal = std::__max_representable_int_for_float<_IntT, _RealT>();
+  if (__r >= ::nextafter(static_cast<_RealT>(_MaxVal), INFINITY)) {
+    return _Lim::max();
+  } else if (__r <= _Lim::lowest()) {
+    return _Lim::min();
+  }
+  return static_cast<_IntT>(__r);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif
+
+#endif  // _LIBCUDACXX_CMATH
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt
new file mode 100644
index 000000000000..cbb09db11ab2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/codecvt
@@ -0,0 +1,549 @@
+// -*- C++ -*-
+//===-------------------------- codecvt -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CODECVT
+#define _LIBCUDACXX_CODECVT
+
+/*
+    codecvt synopsis
+
+namespace std
+{
+
+enum codecvt_mode
+{
+    consume_header = 4,
+    generate_header = 2,
+    little_endian = 1
+};
+
+template <class Elem, unsigned long Maxcode = 0x10ffff,
+          codecvt_mode Mode = (codecvt_mode)0>
+class codecvt_utf8
+    : public codecvt<Elem, char, mbstate_t>
+{
+    explicit codecvt_utf8(size_t refs = 0);
+    ~codecvt_utf8();
+};
+
+template <class Elem, unsigned long Maxcode = 0x10ffff,
+          codecvt_mode Mode = (codecvt_mode)0>
+class codecvt_utf16
+    : public codecvt<Elem, char, mbstate_t>
+{
+    explicit codecvt_utf16(size_t refs = 0);
+    ~codecvt_utf16();
+};
+
+template <class Elem, unsigned long Maxcode = 0x10ffff,
+          codecvt_mode Mode = (codecvt_mode)0>
+class codecvt_utf8_utf16
+    : public codecvt<Elem, char, mbstate_t>
+{
+    explicit codecvt_utf8_utf16(size_t refs = 0);
+    ~codecvt_utf8_utf16();
+};
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__locale>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+enum codecvt_mode
+{
+    consume_header = 4,
+    generate_header = 2,
+    little_endian = 1
+};
+
+// codecvt_utf8
+
+template <class _Elem> class __codecvt_utf8;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf8<wchar_t>
+    : public codecvt<wchar_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef wchar_t   intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf8(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<wchar_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf8<char16_t>
+    : public codecvt<char16_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char16_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf8(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char16_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf8<char32_t>
+    : public codecvt<char32_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char32_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf8(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char32_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <class _Elem, unsigned long _Maxcode = 0x10ffff,
+          codecvt_mode _Mode = (codecvt_mode)0>
+class _LIBCUDACXX_TEMPLATE_VIS codecvt_utf8
+    : public __codecvt_utf8<_Elem>
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt_utf8(size_t __refs = 0)
+        : __codecvt_utf8<_Elem>(__refs, _Maxcode, _Mode) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~codecvt_utf8() {}
+};
+
+// codecvt_utf16
+
+template <class _Elem, bool _LittleEndian> class __codecvt_utf16;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf16<wchar_t, false>
+    : public codecvt<wchar_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef wchar_t   intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<wchar_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf16<wchar_t, true>
+    : public codecvt<wchar_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef wchar_t   intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<wchar_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf16<char16_t, false>
+    : public codecvt<char16_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char16_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char16_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf16<char16_t, true>
+    : public codecvt<char16_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char16_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char16_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf16<char32_t, false>
+    : public codecvt<char32_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char32_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char32_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf16<char32_t, true>
+    : public codecvt<char32_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char32_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char32_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <class _Elem, unsigned long _Maxcode = 0x10ffff,
+          codecvt_mode _Mode = (codecvt_mode)0>
+class _LIBCUDACXX_TEMPLATE_VIS codecvt_utf16
+    : public __codecvt_utf16<_Elem, _Mode & little_endian>
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt_utf16(size_t __refs = 0)
+        : __codecvt_utf16<_Elem, _Mode & little_endian>(__refs, _Maxcode, _Mode) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~codecvt_utf16() {}
+};
+
+// codecvt_utf8_utf16
+
+template <class _Elem> class __codecvt_utf8_utf16;
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf8_utf16<wchar_t>
+    : public codecvt<wchar_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef wchar_t   intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf8_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<wchar_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf8_utf16<char32_t>
+    : public codecvt<char32_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char32_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf8_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char32_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <>
+class _LIBCUDACXX_TYPE_VIS __codecvt_utf8_utf16<char16_t>
+    : public codecvt<char16_t, char, mbstate_t>
+{
+    unsigned long _Maxcode_;
+    codecvt_mode _Mode_;
+public:
+    typedef char16_t  intern_type;
+    typedef char      extern_type;
+    typedef mbstate_t state_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __codecvt_utf8_utf16(size_t __refs, unsigned long _Maxcode,
+                            codecvt_mode _Mode)
+        : codecvt<char16_t, char, mbstate_t>(__refs), _Maxcode_(_Maxcode),
+          _Mode_(_Mode) {}
+protected:
+    virtual result
+        do_out(state_type& __st,
+               const intern_type* __frm, const intern_type* __frm_end, const intern_type*& __frm_nxt,
+               extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual result
+        do_in(state_type& __st,
+              const extern_type* __frm, const extern_type* __frm_end, const extern_type*& __frm_nxt,
+              intern_type* __to, intern_type* __to_end, intern_type*& __to_nxt) const;
+    virtual result
+        do_unshift(state_type& __st,
+                   extern_type* __to, extern_type* __to_end, extern_type*& __to_nxt) const;
+    virtual int do_encoding() const throw();
+    virtual bool do_always_noconv() const throw();
+    virtual int do_length(state_type&, const extern_type* __frm, const extern_type* __end,
+                          size_t __mx) const;
+    virtual int do_max_length() const throw();
+};
+
+template <class _Elem, unsigned long _Maxcode = 0x10ffff,
+          codecvt_mode _Mode = (codecvt_mode)0>
+class _LIBCUDACXX_TEMPLATE_VIS codecvt_utf8_utf16
+    : public __codecvt_utf8_utf16<_Elem>
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit codecvt_utf8_utf16(size_t __refs = 0)
+        : __codecvt_utf8_utf16<_Elem>(__refs, _Maxcode, _Mode) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~codecvt_utf8_utf16() {}
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CODECVT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/compare b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/compare
new file mode 100644
index 000000000000..d3482f0e23c9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/compare
@@ -0,0 +1,678 @@
+// -*- C++ -*-
+//===-------------------------- compare -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_COMPARE
+#define _LIBCUDACXX_COMPARE
+
+/*
+    compare synopsis
+
+namespace std {
+  // [cmp.categories], comparison category types
+  class weak_equality;
+  class strong_equality;
+  class partial_ordering;
+  class weak_ordering;
+  class strong_ordering;
+
+  // named comparison functions
+  constexpr bool is_eq  (weak_equality cmp) noexcept    { return cmp == 0; }
+  constexpr bool is_neq (weak_equality cmp) noexcept    { return cmp != 0; }
+  constexpr bool is_lt  (partial_ordering cmp) noexcept { return cmp < 0; }
+  constexpr bool is_lteq(partial_ordering cmp) noexcept { return cmp <= 0; }
+  constexpr bool is_gt  (partial_ordering cmp) noexcept { return cmp > 0; }
+  constexpr bool is_gteq(partial_ordering cmp) noexcept { return cmp >= 0; }
+
+  // [cmp.common], common comparison category type
+  template<class... Ts>
+  struct common_comparison_category {
+    using type = see below;
+  };
+  template<class... Ts>
+    using common_comparison_category_t = typename common_comparison_category<Ts...>::type;
+
+  // [cmp.alg], comparison algorithms
+  template<class T> constexpr strong_ordering strong_order(const T& a, const T& b);
+  template<class T> constexpr weak_ordering weak_order(const T& a, const T& b);
+  template<class T> constexpr partial_ordering partial_order(const T& a, const T& b);
+  template<class T> constexpr strong_equality strong_equal(const T& a, const T& b);
+  template<class T> constexpr weak_equality weak_equal(const T& a, const T& b);
+}
+*/
+
+#include <__config>
+#include <type_traits>
+#include <array>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 17
+
+// exposition only
+enum class _LIBCUDACXX_ENUM_VIS _EqResult : unsigned char {
+  __zero = 0,
+  __equal = __zero,
+  __equiv = __equal,
+  __nonequal = 1,
+  __nonequiv = __nonequal
+};
+
+enum class _LIBCUDACXX_ENUM_VIS _OrdResult : signed char {
+  __less = -1,
+  __greater = 1
+};
+
+enum class _LIBCUDACXX_ENUM_VIS _NCmpResult : signed char {
+  __unordered = -127
+};
+
+struct _CmpUnspecifiedType;
+using _CmpUnspecifiedParam = void (_CmpUnspecifiedType::*)();
+
+class  weak_equality {
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr explicit weak_equality(_EqResult __val) noexcept : __value_(__val) {}
+
+public:
+  static const weak_equality equivalent;
+  static const weak_equality nonequivalent;
+
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(weak_equality __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(_CmpUnspecifiedParam, weak_equality __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(weak_equality __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(_CmpUnspecifiedParam, weak_equality __v) noexcept;
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr weak_equality operator<=>(weak_equality __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr weak_equality operator<=>(_CmpUnspecifiedParam, weak_equality __v) noexcept;
+#endif
+
+private:
+  _EqResult __value_;
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr weak_equality weak_equality::equivalent(_EqResult::__equiv);
+_LIBCUDACXX_INLINE_VAR constexpr weak_equality weak_equality::nonequivalent(_EqResult::__nonequiv);
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool operator==(weak_equality __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ == _EqResult::__zero;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool operator==(_CmpUnspecifiedParam, weak_equality __v) noexcept {
+  return __v.__value_ == _EqResult::__zero;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool operator!=(weak_equality __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ != _EqResult::__zero;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool operator!=(_CmpUnspecifiedParam, weak_equality __v) noexcept {
+  return __v.__value_ != _EqResult::__zero;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr weak_equality operator<=>(weak_equality __v, _CmpUnspecifiedParam) noexcept {
+  return __v;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr weak_equality operator<=>(_CmpUnspecifiedParam, weak_equality __v) noexcept {
+  return __v;
+}
+#endif
+
+class strong_equality {
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr strong_equality(_EqResult __val) noexcept : __value_(__val) {}
+
+public:
+  static const strong_equality equal;
+  static const strong_equality nonequal;
+  static const strong_equality equivalent;
+  static const strong_equality nonequivalent;
+
+  // conversion
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr operator weak_equality() const noexcept {
+    return __value_ == _EqResult::__zero ? weak_equality::equivalent
+          : weak_equality::nonequivalent;
+  }
+
+  // comparisons
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(strong_equality __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(strong_equality __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(_CmpUnspecifiedParam, strong_equality __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(_CmpUnspecifiedParam, strong_equality __v) noexcept;
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr strong_equality operator<=>(strong_equality __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr strong_equality operator<=>(_CmpUnspecifiedParam, strong_equality __v) noexcept;
+#endif
+private:
+  _EqResult __value_;
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr strong_equality strong_equality::equal(_EqResult::__equal);
+_LIBCUDACXX_INLINE_VAR constexpr strong_equality strong_equality::nonequal(_EqResult::__nonequal);
+_LIBCUDACXX_INLINE_VAR constexpr strong_equality strong_equality::equivalent(_EqResult::__equiv);
+_LIBCUDACXX_INLINE_VAR constexpr strong_equality strong_equality::nonequivalent(_EqResult::__nonequiv);
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(strong_equality __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ == _EqResult::__zero;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(_CmpUnspecifiedParam, strong_equality __v) noexcept {
+  return __v.__value_ == _EqResult::__zero;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(strong_equality __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ != _EqResult::__zero;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(_CmpUnspecifiedParam, strong_equality __v) noexcept {
+  return __v.__value_ != _EqResult::__zero;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr strong_equality operator<=>(strong_equality __v, _CmpUnspecifiedParam) noexcept {
+  return __v;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr strong_equality operator<=>(_CmpUnspecifiedParam, strong_equality __v) noexcept {
+  return __v;
+}
+#endif // _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+
+class partial_ordering {
+  using _ValueT = signed char;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr partial_ordering(_EqResult __v) noexcept
+      : __value_(_ValueT(__v)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr partial_ordering(_OrdResult __v) noexcept
+      : __value_(_ValueT(__v)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr partial_ordering(_NCmpResult __v) noexcept
+      : __value_(_ValueT(__v)) {}
+
+  constexpr bool __is_ordered() const noexcept {
+    return __value_ != _ValueT(_NCmpResult::__unordered);
+  }
+public:
+  // valid values
+  static const partial_ordering less;
+  static const partial_ordering equivalent;
+  static const partial_ordering greater;
+  static const partial_ordering unordered;
+
+  // conversion
+  constexpr operator weak_equality() const noexcept {
+    return __value_ == 0 ? weak_equality::equivalent : weak_equality::nonequivalent;
+  }
+
+  // comparisons
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator< (partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator<=(partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator> (partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator>=(partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator< (_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator<=(_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator> (_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator>=(_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr partial_ordering operator<=>(partial_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr partial_ordering operator<=>(_CmpUnspecifiedParam, partial_ordering __v) noexcept;
+#endif
+
+private:
+  _ValueT __value_;
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr partial_ordering partial_ordering::less(_OrdResult::__less);
+_LIBCUDACXX_INLINE_VAR constexpr partial_ordering partial_ordering::equivalent(_EqResult::__equiv);
+_LIBCUDACXX_INLINE_VAR constexpr partial_ordering partial_ordering::greater(_OrdResult::__greater);
+_LIBCUDACXX_INLINE_VAR constexpr partial_ordering partial_ordering::unordered(_NCmpResult ::__unordered);
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__is_ordered() && __v.__value_ == 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator< (partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__is_ordered() && __v.__value_ < 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__is_ordered() && __v.__value_ <= 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator> (partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__is_ordered() && __v.__value_ > 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__is_ordered() && __v.__value_ >= 0;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return __v.__is_ordered() && 0 == __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator< (_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return __v.__is_ordered() && 0 < __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return __v.__is_ordered() && 0 <= __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator> (_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return __v.__is_ordered() && 0 > __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return __v.__is_ordered() && 0 >= __v.__value_;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return !__v.__is_ordered() || __v.__value_ != 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return !__v.__is_ordered() || __v.__value_ != 0;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr partial_ordering operator<=>(partial_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr partial_ordering operator<=>(_CmpUnspecifiedParam, partial_ordering __v) noexcept {
+  return __v < 0 ? partial_ordering::greater : (__v > 0 ? partial_ordering::less : __v);
+}
+#endif // _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+
+class weak_ordering {
+  using _ValueT = signed char;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr weak_ordering(_EqResult __v) noexcept : __value_(_ValueT(__v)) {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr weak_ordering(_OrdResult __v) noexcept : __value_(_ValueT(__v)) {}
+
+public:
+  static const weak_ordering less;
+  static const weak_ordering equivalent;
+  static const weak_ordering greater;
+
+  // conversions
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr operator weak_equality() const noexcept {
+    return __value_ == 0 ? weak_equality::equivalent
+                         : weak_equality::nonequivalent;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr operator partial_ordering() const noexcept {
+    return __value_ == 0 ? partial_ordering::equivalent
+        : (__value_ < 0 ? partial_ordering::less : partial_ordering::greater);
+  }
+
+  // comparisons
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator< (weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator<=(weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator> (weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator>=(weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator< (_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator<=(_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator> (_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator>=(_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr weak_ordering operator<=>(weak_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr weak_ordering operator<=>(_CmpUnspecifiedParam, weak_ordering __v) noexcept;
+#endif
+
+private:
+  _ValueT __value_;
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr weak_ordering weak_ordering::less(_OrdResult::__less);
+_LIBCUDACXX_INLINE_VAR constexpr weak_ordering weak_ordering::equivalent(_EqResult::__equiv);
+_LIBCUDACXX_INLINE_VAR constexpr weak_ordering weak_ordering::greater(_OrdResult::__greater);
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ == 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ != 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator< (weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ < 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ <= 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator> (weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ > 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ >= 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return 0 == __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return 0 != __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator< (_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return 0 < __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return 0 <= __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator> (_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return 0 > __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return 0 >= __v.__value_;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr weak_ordering operator<=>(weak_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr weak_ordering operator<=>(_CmpUnspecifiedParam, weak_ordering __v) noexcept {
+  return __v < 0 ? weak_ordering::greater : (__v > 0 ? weak_ordering::less : __v);
+}
+#endif // _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+
+class strong_ordering {
+  using _ValueT = signed char;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr strong_ordering(_EqResult __v) noexcept : __value_(_ValueT(__v)) {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr strong_ordering(_OrdResult __v) noexcept : __value_(_ValueT(__v)) {}
+
+public:
+  static const strong_ordering less;
+  static const strong_ordering equal;
+  static const strong_ordering equivalent;
+  static const strong_ordering greater;
+
+  // conversions
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr operator weak_equality() const noexcept {
+    return __value_ == 0 ? weak_equality::equivalent
+                         : weak_equality::nonequivalent;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr operator strong_equality() const noexcept {
+    return __value_ == 0 ? strong_equality::equal
+                         : strong_equality::nonequal;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr operator partial_ordering() const noexcept {
+    return __value_ == 0 ? partial_ordering::equivalent
+        : (__value_ < 0 ? partial_ordering::less : partial_ordering::greater);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr operator weak_ordering() const noexcept {
+    return __value_ == 0 ? weak_ordering::equivalent
+        : (__value_ < 0 ? weak_ordering::less : weak_ordering::greater);
+  }
+
+  // comparisons
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator< (strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator<=(strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator> (strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator>=(strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator==(_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator!=(_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator< (_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator<=(_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator> (_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr bool operator>=(_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr strong_ordering operator<=>(strong_ordering __v, _CmpUnspecifiedParam) noexcept;
+  _LIBCUDACXX_INLINE_VISIBILITY friend constexpr strong_ordering operator<=>(_CmpUnspecifiedParam, strong_ordering __v) noexcept;
+#endif
+
+private:
+  _ValueT __value_;
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr strong_ordering strong_ordering::less(_OrdResult::__less);
+_LIBCUDACXX_INLINE_VAR constexpr strong_ordering strong_ordering::equal(_EqResult::__equal);
+_LIBCUDACXX_INLINE_VAR constexpr strong_ordering strong_ordering::equivalent(_EqResult::__equiv);
+_LIBCUDACXX_INLINE_VAR constexpr strong_ordering strong_ordering::greater(_OrdResult::__greater);
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ == 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ != 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator< (strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ < 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ <= 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator> (strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ > 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v.__value_ >= 0;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return 0 == __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return 0 != __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator< (_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return 0 < __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return 0 <= __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator> (_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return 0 > __v.__value_;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return 0 >= __v.__value_;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr strong_ordering operator<=>(strong_ordering __v, _CmpUnspecifiedParam) noexcept {
+  return __v;
+}
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr strong_ordering operator<=>(_CmpUnspecifiedParam, strong_ordering __v) noexcept {
+  return __v < 0 ? strong_ordering::greater : (__v > 0 ? strong_ordering::less : __v);
+}
+#endif // _LIBCUDACXX_HAS_NO_SPACESHIP_OPERATOR
+
+// named comparison functions
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_eq(weak_equality __cmp) noexcept    { return __cmp == 0; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_neq(weak_equality __cmp) noexcept    { return __cmp != 0; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_lt(partial_ordering __cmp) noexcept { return __cmp < 0; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_lteq(partial_ordering __cmp) noexcept { return __cmp <= 0; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_gt(partial_ordering __cmp) noexcept { return __cmp > 0; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool is_gteq(partial_ordering __cmp) noexcept { return __cmp >= 0; }
+
+namespace __comp_detail {
+
+enum _ClassifyCompCategory : unsigned{
+  _None,
+  _WeakEq,
+  _StrongEq,
+  _PartialOrd,
+  _WeakOrd,
+  _StrongOrd,
+  _CCC_Size
+};
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr _ClassifyCompCategory __type_to_enum() noexcept {
+  if (is_same_v<_Tp, weak_equality>)
+    return _WeakEq;
+  if (is_same_v<_Tp, strong_equality>)
+    return _StrongEq;
+  if (is_same_v<_Tp, partial_ordering>)
+    return _PartialOrd;
+  if (is_same_v<_Tp, weak_ordering>)
+    return _WeakOrd;
+  if (is_same_v<_Tp, strong_ordering>)
+    return _StrongOrd;
+  return _None;
+}
+
+template <size_t _Size>
+constexpr _ClassifyCompCategory
+__compute_comp_type(std::array<_ClassifyCompCategory, _Size> __types) {
+  std::array<int, _CCC_Size> __seen = {};
+  for (auto __type : __types)
+    ++__seen[__type];
+  if (__seen[_None])
+    return _None;
+  if (__seen[_WeakEq])
+    return _WeakEq;
+  if (__seen[_StrongEq] && (__seen[_PartialOrd] || __seen[_WeakOrd]))
+    return _WeakEq;
+  if (__seen[_StrongEq])
+    return _StrongEq;
+  if (__seen[_PartialOrd])
+    return _PartialOrd;
+  if (__seen[_WeakOrd])
+    return _WeakOrd;
+  return _StrongOrd;
+}
+
+template <class ..._Ts>
+constexpr auto __get_comp_type() {
+  using _CCC = _ClassifyCompCategory;
+  constexpr array<_CCC, sizeof...(_Ts)> __type_kinds{{__comp_detail::__type_to_enum<_Ts>()...}};
+  constexpr _CCC _Cat = sizeof...(_Ts) == 0 ? _StrongOrd
+      : __compute_comp_type(__type_kinds);
+  if constexpr (_Cat == _None)
+    return void();
+  else if constexpr (_Cat == _WeakEq)
+    return weak_equality::equivalent;
+  else if constexpr (_Cat == _StrongEq)
+    return strong_equality::equivalent;
+  else if constexpr (_Cat == _PartialOrd)
+    return partial_ordering::equivalent;
+  else if constexpr (_Cat == _WeakOrd)
+    return weak_ordering::equivalent;
+  else if constexpr (_Cat == _StrongOrd)
+    return strong_ordering::equivalent;
+  else
+    static_assert(_Cat != _Cat, "unhandled case");
+}
+} // namespace __comp_detail
+
+// [cmp.common], common comparison category type
+template<class... _Ts>
+struct _LIBCUDACXX_TEMPLATE_VIS common_comparison_category {
+  using type = decltype(__comp_detail::__get_comp_type<_Ts...>());
+};
+
+template<class... _Ts>
+using common_comparison_category_t = typename common_comparison_category<_Ts...>::type;
+
+// [cmp.alg], comparison algorithms
+// TODO: unimplemented
+template<class _Tp> constexpr strong_ordering strong_order(const _Tp& __lhs, const _Tp& __rhs);
+template<class _Tp> constexpr weak_ordering weak_order(const _Tp& __lhs, const _Tp& __rhs);
+template<class _Tp> constexpr partial_ordering partial_order(const _Tp& __lhs, const _Tp& __rhs);
+template<class _Tp> constexpr strong_equality strong_equal(const _Tp& __lhs, const _Tp& __rhs);
+template<class _Tp> constexpr weak_equality weak_equal(const _Tp& __lhs, const _Tp& __rhs);
+
+#endif // _LIBCUDACXX_STD_VER > 17
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // _LIBCUDACXX_COMPARE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex
new file mode 100644
index 000000000000..eccde954ef25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex
@@ -0,0 +1,1590 @@
+// -*- C++ -*-
+//===--------------------------- complex ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_COMPLEX
+#define _LIBCUDACXX_COMPLEX
+
+/*
+    complex synopsis
+
+namespace std
+{
+
+template<class T>
+class complex
+{
+public:
+    typedef T value_type;
+
+    complex(const T& re = T(), const T& im = T()); // constexpr in C++14
+    complex(const complex&);  // constexpr in C++14
+    template<class X> complex(const complex<X>&);  // constexpr in C++14
+
+    T real() const; // constexpr in C++14
+    T imag() const; // constexpr in C++14
+
+    void real(T);
+    void imag(T);
+
+    complex<T>& operator= (const T&);
+    complex<T>& operator+=(const T&);
+    complex<T>& operator-=(const T&);
+    complex<T>& operator*=(const T&);
+    complex<T>& operator/=(const T&);
+
+    complex& operator=(const complex&);
+    template<class X> complex<T>& operator= (const complex<X>&);
+    template<class X> complex<T>& operator+=(const complex<X>&);
+    template<class X> complex<T>& operator-=(const complex<X>&);
+    template<class X> complex<T>& operator*=(const complex<X>&);
+    template<class X> complex<T>& operator/=(const complex<X>&);
+};
+
+template<>
+class complex<float>
+{
+public:
+    typedef float value_type;
+
+    constexpr complex(float re = 0.0f, float im = 0.0f);
+    explicit constexpr complex(const complex<double>&);
+    explicit constexpr complex(const complex<long double>&);
+
+    constexpr float real() const;
+    void real(float);
+    constexpr float imag() const;
+    void imag(float);
+
+    complex<float>& operator= (float);
+    complex<float>& operator+=(float);
+    complex<float>& operator-=(float);
+    complex<float>& operator*=(float);
+    complex<float>& operator/=(float);
+
+    complex<float>& operator=(const complex<float>&);
+    template<class X> complex<float>& operator= (const complex<X>&);
+    template<class X> complex<float>& operator+=(const complex<X>&);
+    template<class X> complex<float>& operator-=(const complex<X>&);
+    template<class X> complex<float>& operator*=(const complex<X>&);
+    template<class X> complex<float>& operator/=(const complex<X>&);
+};
+
+template<>
+class complex<double>
+{
+public:
+    typedef double value_type;
+
+    constexpr complex(double re = 0.0, double im = 0.0);
+    constexpr complex(const complex<float>&);
+    explicit constexpr complex(const complex<long double>&);
+
+    constexpr double real() const;
+    void real(double);
+    constexpr double imag() const;
+    void imag(double);
+
+    complex<double>& operator= (double);
+    complex<double>& operator+=(double);
+    complex<double>& operator-=(double);
+    complex<double>& operator*=(double);
+    complex<double>& operator/=(double);
+    complex<double>& operator=(const complex<double>&);
+
+    template<class X> complex<double>& operator= (const complex<X>&);
+    template<class X> complex<double>& operator+=(const complex<X>&);
+    template<class X> complex<double>& operator-=(const complex<X>&);
+    template<class X> complex<double>& operator*=(const complex<X>&);
+    template<class X> complex<double>& operator/=(const complex<X>&);
+};
+
+template<>
+class complex<long double>
+{
+public:
+    typedef long double value_type;
+
+    constexpr complex(long double re = 0.0L, long double im = 0.0L);
+    constexpr complex(const complex<float>&);
+    constexpr complex(const complex<double>&);
+
+    constexpr long double real() const;
+    void real(long double);
+    constexpr long double imag() const;
+    void imag(long double);
+
+    complex<long double>& operator=(const complex<long double>&);
+    complex<long double>& operator= (long double);
+    complex<long double>& operator+=(long double);
+    complex<long double>& operator-=(long double);
+    complex<long double>& operator*=(long double);
+    complex<long double>& operator/=(long double);
+
+    template<class X> complex<long double>& operator= (const complex<X>&);
+    template<class X> complex<long double>& operator+=(const complex<X>&);
+    template<class X> complex<long double>& operator-=(const complex<X>&);
+    template<class X> complex<long double>& operator*=(const complex<X>&);
+    template<class X> complex<long double>& operator/=(const complex<X>&);
+};
+
+// 26.3.6 operators:
+template<class T> complex<T> operator+(const complex<T>&, const complex<T>&);
+template<class T> complex<T> operator+(const complex<T>&, const T&);
+template<class T> complex<T> operator+(const T&, const complex<T>&);
+template<class T> complex<T> operator-(const complex<T>&, const complex<T>&);
+template<class T> complex<T> operator-(const complex<T>&, const T&);
+template<class T> complex<T> operator-(const T&, const complex<T>&);
+template<class T> complex<T> operator*(const complex<T>&, const complex<T>&);
+template<class T> complex<T> operator*(const complex<T>&, const T&);
+template<class T> complex<T> operator*(const T&, const complex<T>&);
+template<class T> complex<T> operator/(const complex<T>&, const complex<T>&);
+template<class T> complex<T> operator/(const complex<T>&, const T&);
+template<class T> complex<T> operator/(const T&, const complex<T>&);
+template<class T> complex<T> operator+(const complex<T>&);
+template<class T> complex<T> operator-(const complex<T>&);
+template<class T> bool operator==(const complex<T>&, const complex<T>&); // constexpr in C++14
+template<class T> bool operator==(const complex<T>&, const T&); // constexpr in C++14
+template<class T> bool operator==(const T&, const complex<T>&); // constexpr in C++14
+template<class T> bool operator!=(const complex<T>&, const complex<T>&); // constexpr in C++14
+template<class T> bool operator!=(const complex<T>&, const T&); // constexpr in C++14
+template<class T> bool operator!=(const T&, const complex<T>&); // constexpr in C++14
+
+template<class T, class charT, class traits>
+  basic_istream<charT, traits>&
+  operator>>(basic_istream<charT, traits>&, complex<T>&);
+template<class T, class charT, class traits>
+  basic_ostream<charT, traits>&
+  operator<<(basic_ostream<charT, traits>&, const complex<T>&);
+
+// 26.3.7 values:
+
+template<class T>              T real(const complex<T>&); // constexpr in C++14
+                     long double real(long double);       // constexpr in C++14
+                          double real(double);            // constexpr in C++14
+template<Integral T>      double real(T);                 // constexpr in C++14
+                          float  real(float);             // constexpr in C++14
+
+template<class T>              T imag(const complex<T>&); // constexpr in C++14
+                     long double imag(long double);       // constexpr in C++14
+                          double imag(double);            // constexpr in C++14
+template<Integral T>      double imag(T);                 // constexpr in C++14
+                          float  imag(float);             // constexpr in C++14
+
+template<class T> T abs(const complex<T>&);
+
+template<class T>              T arg(const complex<T>&);
+                     long double arg(long double);
+                          double arg(double);
+template<Integral T>      double arg(T);
+                          float  arg(float);
+
+template<class T>              T norm(const complex<T>&);
+                     long double norm(long double);
+                          double norm(double);
+template<Integral T>      double norm(T);
+                          float  norm(float);
+
+template<class T>      complex<T>           conj(const complex<T>&);
+                       complex<long double> conj(long double);
+                       complex<double>      conj(double);
+template<Integral T>   complex<double>      conj(T);
+                       complex<float>       conj(float);
+
+template<class T>    complex<T>           proj(const complex<T>&);
+                     complex<long double> proj(long double);
+                     complex<double>      proj(double);
+template<Integral T> complex<double>      proj(T);
+                     complex<float>       proj(float);
+
+template<class T> complex<T> polar(const T&, const T& = T());
+
+// 26.3.8 transcendentals:
+template<class T> complex<T> acos(const complex<T>&);
+template<class T> complex<T> asin(const complex<T>&);
+template<class T> complex<T> atan(const complex<T>&);
+template<class T> complex<T> acosh(const complex<T>&);
+template<class T> complex<T> asinh(const complex<T>&);
+template<class T> complex<T> atanh(const complex<T>&);
+template<class T> complex<T> cos (const complex<T>&);
+template<class T> complex<T> cosh (const complex<T>&);
+template<class T> complex<T> exp (const complex<T>&);
+template<class T> complex<T> log (const complex<T>&);
+template<class T> complex<T> log10(const complex<T>&);
+
+template<class T> complex<T> pow(const complex<T>&, const T&);
+template<class T> complex<T> pow(const complex<T>&, const complex<T>&);
+template<class T> complex<T> pow(const T&, const complex<T>&);
+
+template<class T> complex<T> sin (const complex<T>&);
+template<class T> complex<T> sinh (const complex<T>&);
+template<class T> complex<T> sqrt (const complex<T>&);
+template<class T> complex<T> tan (const complex<T>&);
+template<class T> complex<T> tanh (const complex<T>&);
+
+template<class T, class charT, class traits>
+  basic_istream<charT, traits>&
+  operator>>(basic_istream<charT, traits>& is, complex<T>& x);
+
+template<class T, class charT, class traits>
+  basic_ostream<charT, traits>&
+  operator<<(basic_ostream<charT, traits>& o, const complex<T>& x);
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <type_traits>
+#include <stdexcept>
+#include <cmath>
+#include <sstream>
+#include <version>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+# if _LIBCUDACXX_CUDA_ABI_VERSION > 3
+#  define _LIBCUDACXX_COMPLEX_ALIGNAS(V) _ALIGNAS(V)
+# else
+#  define _LIBCUDACXX_COMPLEX_ALIGNAS(V)
+# endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template<class _Tp> class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_COMPLEX_ALIGNAS(2*sizeof(_Tp)) complex;
+
+template<class _Tp> _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp> operator*(const complex<_Tp>& __z, const complex<_Tp>& __w);
+
+template<class _Tp> _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp> operator/(const complex<_Tp>& __x, const complex<_Tp>& __y);
+
+template<class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_COMPLEX_ALIGNAS(2*sizeof(_Tp)) complex
+{
+public:
+    typedef _Tp value_type;
+private:
+    value_type __re_;
+    value_type __im_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    complex(const value_type& __re = value_type(), const value_type& __im = value_type())
+        : __re_(__re), __im_(__im) {}
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    complex(const complex<_Xp>& __c)
+        : __re_(__c.real()), __im_(__c.imag()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 value_type real() const {return __re_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 value_type imag() const {return __im_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void real(value_type __re) {__re_ = __re;}
+    _LIBCUDACXX_INLINE_VISIBILITY void imag(value_type __im) {__im_ = __im;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (const value_type& __re)
+        {__re_ = __re; __im_ = value_type(); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(const value_type& __re) {__re_ += __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(const value_type& __re) {__re_ -= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(const value_type& __re) {__re_ *= __re; __im_ *= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(const value_type& __re) {__re_ /= __re; __im_ /= __re; return *this;}
+
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (const complex<_Xp>& __c)
+        {
+            __re_ = __c.real();
+            __im_ = __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(const complex<_Xp>& __c)
+        {
+            __re_ += __c.real();
+            __im_ += __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(const complex<_Xp>& __c)
+        {
+            __re_ -= __c.real();
+            __im_ -= __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(const complex<_Xp>& __c)
+        {
+            *this = *this * complex(__c.real(), __c.imag());
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(const complex<_Xp>& __c)
+        {
+            *this = *this / complex(__c.real(), __c.imag());
+            return *this;
+        }
+};
+
+template<> class complex<double>;
+#ifdef _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+template<> class complex<long double>;
+#endif // _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+
+template<>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_COMPLEX_ALIGNAS(2*sizeof(float)) complex<float>
+{
+    float __re_;
+    float __im_;
+public:
+    typedef float value_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR complex(float __re = 0.0f, float __im = 0.0f)
+        : __re_(__re), __im_(__im) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit _LIBCUDACXX_CONSTEXPR complex(const complex<double>& __c);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit _LIBCUDACXX_CONSTEXPR complex(const complex<long double>& __c);
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR float real() const {return __re_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR float imag() const {return __im_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void real(value_type __re) {__re_ = __re;}
+    _LIBCUDACXX_INLINE_VISIBILITY void imag(value_type __im) {__im_ = __im;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (float __re)
+        {__re_ = __re; __im_ = value_type(); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(float __re) {__re_ += __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(float __re) {__re_ -= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(float __re) {__re_ *= __re; __im_ *= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(float __re) {__re_ /= __re; __im_ /= __re; return *this;}
+
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (const complex<_Xp>& __c)
+        {
+            __re_ = __c.real();
+            __im_ = __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(const complex<_Xp>& __c)
+        {
+            __re_ += __c.real();
+            __im_ += __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(const complex<_Xp>& __c)
+        {
+            __re_ -= __c.real();
+            __im_ -= __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(const complex<_Xp>& __c)
+        {
+            *this = *this * complex(__c.real(), __c.imag());
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(const complex<_Xp>& __c)
+        {
+            *this = *this / complex(__c.real(), __c.imag());
+            return *this;
+        }
+};
+
+template<>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_COMPLEX_ALIGNAS(2*sizeof(double)) complex<double>
+{
+    double __re_;
+    double __im_;
+public:
+    typedef double value_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR complex(double __re = 0.0, double __im = 0.0)
+        : __re_(__re), __im_(__im) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR complex(const complex<float>& __c);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit _LIBCUDACXX_CONSTEXPR complex(const complex<long double>& __c);
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR double real() const {return __re_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR double imag() const {return __im_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void real(value_type __re) {__re_ = __re;}
+    _LIBCUDACXX_INLINE_VISIBILITY void imag(value_type __im) {__im_ = __im;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (double __re)
+        {__re_ = __re; __im_ = value_type(); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(double __re) {__re_ += __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(double __re) {__re_ -= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(double __re) {__re_ *= __re; __im_ *= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(double __re) {__re_ /= __re; __im_ /= __re; return *this;}
+
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (const complex<_Xp>& __c)
+        {
+            __re_ = __c.real();
+            __im_ = __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(const complex<_Xp>& __c)
+        {
+            __re_ += __c.real();
+            __im_ += __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(const complex<_Xp>& __c)
+        {
+            __re_ -= __c.real();
+            __im_ -= __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(const complex<_Xp>& __c)
+        {
+            *this = *this * complex(__c.real(), __c.imag());
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(const complex<_Xp>& __c)
+        {
+            *this = *this / complex(__c.real(), __c.imag());
+            return *this;
+        }
+};
+
+template<>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_COMPLEX_ALIGNAS(2*sizeof(long double)) complex<long double>
+{
+#ifndef _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+public:
+    template <typename _Dummy = void>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR complex(long double __re = 0.0, long double __im = 0.0)
+        {static_assert(is_same<_Dummy, void>::value, "complex<long double> is not supported");}
+
+    template <typename _Tp, typename _Dummy = void>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR complex(const complex<_Tp> &__c)
+        {static_assert(is_same<_Dummy, void>::value, "complex<long double> is not supported");}
+
+#else
+    long double __re_;
+    long double __im_;
+public:
+    typedef long double value_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR complex(long double __re = 0.0L, long double __im = 0.0L)
+        : __re_(__re), __im_(__im) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR complex(const complex<float>& __c);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR complex(const complex<double>& __c);
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR long double real() const {return __re_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR long double imag() const {return __im_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void real(value_type __re) {__re_ = __re;}
+    _LIBCUDACXX_INLINE_VISIBILITY void imag(value_type __im) {__im_ = __im;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (long double __re)
+        {__re_ = __re; __im_ = value_type(); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(long double __re) {__re_ += __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(long double __re) {__re_ -= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(long double __re) {__re_ *= __re; __im_ *= __re; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(long double __re) {__re_ /= __re; __im_ /= __re; return *this;}
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator= (const complex<_Xp>& __c)
+        {
+            __re_ = __c.real();
+            __im_ = __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator+=(const complex<_Xp>& __c)
+        {
+            __re_ += __c.real();
+            __im_ += __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator-=(const complex<_Xp>& __c)
+        {
+            __re_ -= __c.real();
+            __im_ -= __c.imag();
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator*=(const complex<_Xp>& __c)
+        {
+            *this = *this * complex(__c.real(), __c.imag());
+            return *this;
+        }
+    template<class _Xp> _LIBCUDACXX_INLINE_VISIBILITY complex& operator/=(const complex<_Xp>& __c)
+        {
+            *this = *this / complex(__c.real(), __c.imag());
+            return *this;
+        }
+#endif // _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+};
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_MSVC_WARNING)
+  // MSVC complains about narrowing conversions on these copy constructors regardless if they are used
+  #pragma warning(push)
+  #pragma warning(disable : 4244)
+#endif
+
+inline
+_LIBCUDACXX_CONSTEXPR
+complex<float>::complex(const complex<double>& __c)
+    : __re_(__c.real()), __im_(__c.imag()) {}
+
+inline
+_LIBCUDACXX_CONSTEXPR
+complex<double>::complex(const complex<float>& __c)
+    : __re_(__c.real()), __im_(__c.imag()) {}
+
+#ifdef _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+inline
+_LIBCUDACXX_CONSTEXPR
+complex<float>::complex(const complex<long double>& __c)
+    : __re_(__c.real()), __im_(__c.imag()) {}
+
+inline
+_LIBCUDACXX_CONSTEXPR
+complex<double>::complex(const complex<long double>& __c)
+    : __re_(__c.real()), __im_(__c.imag()) {}
+
+inline
+_LIBCUDACXX_CONSTEXPR
+complex<long double>::complex(const complex<float>& __c)
+    : __re_(__c.real()), __im_(__c.imag()) {}
+
+inline
+_LIBCUDACXX_CONSTEXPR
+complex<long double>::complex(const complex<double>& __c)
+    : __re_(__c.real()), __im_(__c.imag()) {}
+#endif // _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_MSVC_WARNING)
+  #pragma warning(pop)
+#endif
+
+// 26.3.6 operators:
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator+(const complex<_Tp>& __x, const complex<_Tp>& __y)
+{
+    complex<_Tp> __t(__x);
+    __t += __y;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator+(const complex<_Tp>& __x, const _Tp& __y)
+{
+    complex<_Tp> __t(__x);
+    __t += __y;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator+(const _Tp& __x, const complex<_Tp>& __y)
+{
+    complex<_Tp> __t(__y);
+    __t += __x;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator-(const complex<_Tp>& __x, const complex<_Tp>& __y)
+{
+    complex<_Tp> __t(__x);
+    __t -= __y;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator-(const complex<_Tp>& __x, const _Tp& __y)
+{
+    complex<_Tp> __t(__x);
+    __t -= __y;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator-(const _Tp& __x, const complex<_Tp>& __y)
+{
+    complex<_Tp> __t(-__y);
+    __t += __x;
+    return __t;
+}
+
+template<class _Tp>
+complex<_Tp>
+operator*(const complex<_Tp>& __z, const complex<_Tp>& __w)
+{
+    _Tp __a = __z.real();
+    _Tp __b = __z.imag();
+    _Tp __c = __w.real();
+    _Tp __d = __w.imag();
+    _Tp __ac = __a * __c;
+    _Tp __bd = __b * __d;
+    _Tp __ad = __a * __d;
+    _Tp __bc = __b * __c;
+    _Tp __x = __ac - __bd;
+    _Tp __y = __ad + __bc;
+    if (__libcpp_isnan_or_builtin(__x) && __libcpp_isnan_or_builtin(__y))
+    {
+        bool __recalc = false;
+        if (__libcpp_isinf_or_builtin(__a) || __libcpp_isinf_or_builtin(__b))
+        {
+            __a = copysign(__libcpp_isinf_or_builtin(__a) ? _Tp(1) : _Tp(0), __a);
+            __b = copysign(__libcpp_isinf_or_builtin(__b) ? _Tp(1) : _Tp(0), __b);
+            if (__libcpp_isnan_or_builtin(__c))
+                __c = copysign(_Tp(0), __c);
+            if (__libcpp_isnan_or_builtin(__d))
+                __d = copysign(_Tp(0), __d);
+            __recalc = true;
+        }
+        if (__libcpp_isinf_or_builtin(__c) || __libcpp_isinf_or_builtin(__d))
+        {
+            __c = copysign(__libcpp_isinf_or_builtin(__c) ? _Tp(1) : _Tp(0), __c);
+            __d = copysign(__libcpp_isinf_or_builtin(__d) ? _Tp(1) : _Tp(0), __d);
+            if (__libcpp_isnan_or_builtin(__a))
+                __a = copysign(_Tp(0), __a);
+            if (__libcpp_isnan_or_builtin(__b))
+                __b = copysign(_Tp(0), __b);
+            __recalc = true;
+        }
+        if (!__recalc && (__libcpp_isinf_or_builtin(__ac) || __libcpp_isinf_or_builtin(__bd) ||
+                          __libcpp_isinf_or_builtin(__ad) || __libcpp_isinf_or_builtin(__bc)))
+        {
+            if (__libcpp_isnan_or_builtin(__a))
+                __a = copysign(_Tp(0), __a);
+            if (__libcpp_isnan_or_builtin(__b))
+                __b = copysign(_Tp(0), __b);
+            if (__libcpp_isnan_or_builtin(__c))
+                __c = copysign(_Tp(0), __c);
+            if (__libcpp_isnan_or_builtin(__d))
+                __d = copysign(_Tp(0), __d);
+            __recalc = true;
+        }
+        if (__recalc)
+        {
+            __x = _Tp(INFINITY) * (__a * __c - __b * __d);
+            __y = _Tp(INFINITY) * (__a * __d + __b * __c);
+        }
+    }
+    return complex<_Tp>(__x, __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator*(const complex<_Tp>& __x, const _Tp& __y)
+{
+    complex<_Tp> __t(__x);
+    __t *= __y;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator*(const _Tp& __x, const complex<_Tp>& __y)
+{
+    complex<_Tp> __t(__y);
+    __t *= __x;
+    return __t;
+}
+
+namespace detail {
+    template <class _Tp>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp __scalbn(_Tp __x, int __i) {
+        return static_cast<_Tp>(scalbn(static_cast<double>(__x), __i));
+    }
+
+    template <>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    float __scalbn<float>(float __x, int __i) {
+        return scalbnf(__x, __i);
+    }
+
+    template <>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    double __scalbn<double>(double __x, int __i) {
+        return scalbn(__x, __i);
+    }
+
+#ifndef _LIBCUDACXX_COMPILER_NVRTC
+    template <>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    long double __scalbn<long double>(long double __x, int __i) {
+        return scalbnl(__x, __i);
+    }
+#endif
+}
+
+template<class _Tp>
+complex<_Tp>
+operator/(const complex<_Tp>& __z, const complex<_Tp>& __w)
+{
+    int __ilogbw = 0;
+    _Tp __a = __z.real();
+    _Tp __b = __z.imag();
+    _Tp __c = __w.real();
+    _Tp __d = __w.imag();
+    _Tp __logbw = logb(fmax(fabs(__c), fabs(__d)));
+    if (__libcpp_isfinite_or_builtin(__logbw))
+    {
+        __ilogbw = static_cast<int>(__logbw);
+        __c = detail::__scalbn(__c, -__ilogbw);
+        __d = detail::__scalbn(__d, -__ilogbw);
+    }
+    _Tp __denom = __c * __c + __d * __d;
+    _Tp __x = detail::__scalbn((__a * __c + __b * __d) / __denom, -__ilogbw);
+    _Tp __y = detail::__scalbn((__b * __c - __a * __d) / __denom, -__ilogbw);
+    if (__libcpp_isnan_or_builtin(__x) && __libcpp_isnan_or_builtin(__y))
+    {
+        if ((__denom == _Tp(0)) && (!__libcpp_isnan_or_builtin(__a) || !__libcpp_isnan_or_builtin(__b)))
+        {
+            __x = copysign(_Tp(INFINITY), __c) * __a;
+            __y = copysign(_Tp(INFINITY), __c) * __b;
+        }
+        else if ((__libcpp_isinf_or_builtin(__a) || __libcpp_isinf_or_builtin(__b)) && __libcpp_isfinite_or_builtin(__c) && __libcpp_isfinite_or_builtin(__d))
+        {
+            __a = copysign(__libcpp_isinf_or_builtin(__a) ? _Tp(1) : _Tp(0), __a);
+            __b = copysign(__libcpp_isinf_or_builtin(__b) ? _Tp(1) : _Tp(0), __b);
+            __x = _Tp(INFINITY) * (__a * __c + __b * __d);
+            __y = _Tp(INFINITY) * (__b * __c - __a * __d);
+        }
+        else if (__libcpp_isinf_or_builtin(__logbw) && __logbw > _Tp(0) && __libcpp_isfinite_or_builtin(__a) && __libcpp_isfinite_or_builtin(__b))
+        {
+            __c = copysign(__libcpp_isinf_or_builtin(__c) ? _Tp(1) : _Tp(0), __c);
+            __d = copysign(__libcpp_isinf_or_builtin(__d) ? _Tp(1) : _Tp(0), __d);
+            __x = _Tp(0) * (__a * __c + __b * __d);
+            __y = _Tp(0) * (__b * __c - __a * __d);
+        }
+    }
+    return complex<_Tp>(__x, __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator/(const complex<_Tp>& __x, const _Tp& __y)
+{
+    return complex<_Tp>(__x.real() / __y, __x.imag() / __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator/(const _Tp& __x, const complex<_Tp>& __y)
+{
+    complex<_Tp> __t(__x);
+    __t /= __y;
+    return __t;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator+(const complex<_Tp>& __x)
+{
+    return __x;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+operator-(const complex<_Tp>& __x)
+{
+    return complex<_Tp>(-__x.real(), -__x.imag());
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator==(const complex<_Tp>& __x, const complex<_Tp>& __y)
+{
+    return __x.real() == __y.real() && __x.imag() == __y.imag();
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator==(const complex<_Tp>& __x, const _Tp& __y)
+{
+    return __x.real() == __y && __x.imag() == 0;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator==(const _Tp& __x, const complex<_Tp>& __y)
+{
+    return __x == __y.real() && 0 == __y.imag();
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator!=(const complex<_Tp>& __x, const complex<_Tp>& __y)
+{
+    return !(__x == __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator!=(const complex<_Tp>& __x, const _Tp& __y)
+{
+    return !(__x == __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator!=(const _Tp& __x, const complex<_Tp>& __y)
+{
+    return !(__x == __y);
+}
+
+// 26.3.7 values:
+
+template <class _Tp, bool = is_integral<_Tp>::value,
+                     bool = is_floating_point<_Tp>::value
+                     >
+struct __libcpp_complex_overload_traits {};
+
+// Integral Types
+template <class _Tp>
+struct __libcpp_complex_overload_traits<_Tp, true, false>
+{
+    typedef double _ValueType;
+    typedef complex<double> _ComplexType;
+};
+
+// Floating point types
+template <class _Tp>
+struct __libcpp_complex_overload_traits<_Tp, false, true>
+{
+    typedef _Tp _ValueType;
+    typedef complex<_Tp> _ComplexType;
+};
+
+// real
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp
+real(const complex<_Tp>& __c)
+{
+    return __c.real();
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename __libcpp_complex_overload_traits<_Tp>::_ValueType
+real(_Tp __re)
+{
+    return __re;
+}
+
+// imag
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp
+imag(const complex<_Tp>& __c)
+{
+    return __c.imag();
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename __libcpp_complex_overload_traits<_Tp>::_ValueType
+imag(_Tp)
+{
+    return 0;
+}
+
+// abs
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+abs(const complex<_Tp>& __c)
+{
+    return hypot(__c.real(), __c.imag());
+}
+
+// arg
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+arg(const complex<_Tp>& __c)
+{
+    return atan2(__c.imag(), __c.real());
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<
+    is_same<_Tp, long double>::value,
+    long double
+>::type
+arg(_Tp __re)
+{
+    return atan2l(0.L, __re);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value || is_same<_Tp, double>::value,
+    double
+>::type
+arg(_Tp __re)
+{
+    // integrals need to be promoted to double
+    return atan2(0., static_cast<double>(__re));
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<
+    is_same<_Tp, float>::value,
+    float
+>::type
+arg(_Tp __re)
+{
+    return atan2f(0.F, __re);
+}
+
+// norm
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+norm(const complex<_Tp>& __c)
+{
+    if (__libcpp_isinf_or_builtin(__c.real()))
+        return abs(__c.real());
+    if (__libcpp_isinf_or_builtin(__c.imag()))
+        return abs(__c.imag());
+    return __c.real() * __c.real() + __c.imag() * __c.imag();
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __libcpp_complex_overload_traits<_Tp>::_ValueType
+norm(_Tp __re)
+{
+    typedef typename __libcpp_complex_overload_traits<_Tp>::_ValueType _ValueType;
+    return static_cast<_ValueType>(__re) * __re;
+}
+
+// conj
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+conj(const complex<_Tp>& __c)
+{
+    return complex<_Tp>(__c.real(), -__c.imag());
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __libcpp_complex_overload_traits<_Tp>::_ComplexType
+conj(_Tp __re)
+{
+    typedef typename __libcpp_complex_overload_traits<_Tp>::_ComplexType _ComplexType;
+    return _ComplexType(__re);
+}
+
+
+
+// proj
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+proj(const complex<_Tp>& __c)
+{
+    std::complex<_Tp> __r = __c;
+    if (__libcpp_isinf_or_builtin(__c.real()) || __libcpp_isinf_or_builtin(__c.imag()))
+        __r = complex<_Tp>(INFINITY, copysign(_Tp(0), __c.imag()));
+    return __r;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_floating_point<_Tp>::value,
+    typename __libcpp_complex_overload_traits<_Tp>::_ComplexType
+>::type
+proj(_Tp __re)
+{
+    if (__libcpp_isinf_or_builtin(__re))
+        __re = abs(__re);
+    return complex<_Tp>(__re);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_integral<_Tp>::value,
+    typename __libcpp_complex_overload_traits<_Tp>::_ComplexType
+>::type
+proj(_Tp __re)
+{
+    typedef typename __libcpp_complex_overload_traits<_Tp>::_ComplexType _ComplexType;
+    return _ComplexType(__re);
+}
+
+// polar
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+polar(const _Tp& __rho, const _Tp& __theta = _Tp())
+{
+    if (__libcpp_isnan_or_builtin(__rho) || signbit(__rho))
+        return complex<_Tp>(_Tp(NAN), _Tp(NAN));
+    if (__libcpp_isnan_or_builtin(__theta))
+    {
+        if (__libcpp_isinf_or_builtin(__rho))
+            return complex<_Tp>(__rho, __theta);
+        return complex<_Tp>(__theta, __theta);
+    }
+    if (__libcpp_isinf_or_builtin(__theta))
+    {
+        if (__libcpp_isinf_or_builtin(__rho))
+            return complex<_Tp>(__rho, _Tp(NAN));
+        return complex<_Tp>(_Tp(NAN), _Tp(NAN));
+    }
+    _Tp __x = __rho * cos(__theta);
+    if (__libcpp_isnan_or_builtin(__x))
+        __x = 0;
+    _Tp __y = __rho * sin(__theta);
+    if (__libcpp_isnan_or_builtin(__y))
+        __y = 0;
+    return complex<_Tp>(__x, __y);
+}
+
+// log
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+log(const complex<_Tp>& __x)
+{
+    return complex<_Tp>(log(abs(__x)), arg(__x));
+}
+
+// log10
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+log10(const complex<_Tp>& __x)
+{
+    return log(__x) / log(_Tp(10));
+}
+
+// sqrt
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+sqrt(const complex<_Tp>& __x)
+{
+    if (__libcpp_isinf_or_builtin(__x.imag()))
+        return complex<_Tp>(_Tp(INFINITY), __x.imag());
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        if (__x.real() > _Tp(0))
+            return complex<_Tp>(__x.real(), __libcpp_isnan_or_builtin(__x.imag()) ? __x.imag() : copysign(_Tp(0), __x.imag()));
+        return complex<_Tp>(__libcpp_isnan_or_builtin(__x.imag()) ? __x.imag() : _Tp(0), copysign(__x.real(), __x.imag()));
+    }
+    return polar(sqrt(abs(__x)), arg(__x) / _Tp(2));
+}
+
+// exp
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+exp(const complex<_Tp>& __x)
+{
+    _Tp __i = __x.imag();
+    if (__i == 0) {
+        return complex<_Tp>(exp(__x.real()), copysign(_Tp(0), __x.imag()));
+    }
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        if (__x.real() < _Tp(0))
+        {
+            if (!__libcpp_isfinite_or_builtin(__i))
+                __i = _Tp(1);
+        }
+        else if (__i == 0 || !__libcpp_isfinite_or_builtin(__i))
+        {
+            if (__libcpp_isinf_or_builtin(__i))
+                __i = _Tp(NAN);
+            return complex<_Tp>(__x.real(), __i);
+        }
+    }
+    _Tp __e = exp(__x.real());
+    return complex<_Tp>(__e * cos(__i), __e * sin(__i));
+}
+
+// pow
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+pow(const complex<_Tp>& __x, const complex<_Tp>& __y)
+{
+    return exp(__y * log(__x));
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<typename __promote<_Tp, _Up>::type>
+pow(const complex<_Tp>& __x, const complex<_Up>& __y)
+{
+    typedef complex<typename __promote<_Tp, _Up>::type> result_type;
+    return _CUDA_VSTD::pow(result_type(__x), result_type(__y));
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_arithmetic<_Up>::value,
+    complex<typename __promote<_Tp, _Up>::type>
+>::type
+pow(const complex<_Tp>& __x, const _Up& __y)
+{
+    typedef complex<typename __promote<_Tp, _Up>::type> result_type;
+    return _CUDA_VSTD::pow(result_type(__x), result_type(__y));
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_arithmetic<_Tp>::value,
+    complex<typename __promote<_Tp, _Up>::type>
+>::type
+pow(const _Tp& __x, const complex<_Up>& __y)
+{
+    typedef complex<typename __promote<_Tp, _Up>::type> result_type;
+    return _CUDA_VSTD::pow(result_type(__x), result_type(__y));
+}
+
+// __sqr, computes pow(x, 2)
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+__sqr(const complex<_Tp>& __x)
+{
+    return complex<_Tp>((__x.real() - __x.imag()) * (__x.real() + __x.imag()),
+                        _Tp(2) * __x.real() * __x.imag());
+}
+
+// asinh
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+asinh(const complex<_Tp>& __x)
+{
+    const _Tp __pi(static_cast<_Tp>(atan2(+0., -0.)));
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        if (__libcpp_isnan_or_builtin(__x.imag()))
+            return __x;
+        if (__libcpp_isinf_or_builtin(__x.imag()))
+            return complex<_Tp>(__x.real(), copysign(__pi * _Tp(0.25), __x.imag()));
+        return complex<_Tp>(__x.real(), copysign(_Tp(0), __x.imag()));
+    }
+    if (__libcpp_isnan_or_builtin(__x.real()))
+    {
+        if (__libcpp_isinf_or_builtin(__x.imag()))
+            return complex<_Tp>(__x.imag(), __x.real());
+        if (__x.imag() == 0)
+            return __x;
+        return complex<_Tp>(__x.real(), __x.real());
+    }
+    if (__libcpp_isinf_or_builtin(__x.imag()))
+        return complex<_Tp>(copysign(__x.imag(), __x.real()), copysign(__pi/_Tp(2), __x.imag()));
+    complex<_Tp> __z = log(__x + sqrt(__sqr(__x) + _Tp(1)));
+    return complex<_Tp>(copysign(__z.real(), __x.real()), copysign(__z.imag(), __x.imag()));
+}
+
+// acosh
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+acosh(const complex<_Tp>& __x)
+{
+    const _Tp __pi(static_cast<_Tp>(atan2(+0., -0.)));
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        if (__libcpp_isnan_or_builtin(__x.imag()))
+            return complex<_Tp>(abs(__x.real()), __x.imag());
+        if (__libcpp_isinf_or_builtin(__x.imag()))
+        {
+            if (__x.real() > 0)
+                return complex<_Tp>(__x.real(), copysign(__pi * _Tp(0.25), __x.imag()));
+            else
+                return complex<_Tp>(-__x.real(), copysign(__pi * _Tp(0.75), __x.imag()));
+        }
+        if (__x.real() < 0)
+            return complex<_Tp>(-__x.real(), copysign(__pi, __x.imag()));
+        return complex<_Tp>(__x.real(), copysign(_Tp(0), __x.imag()));
+    }
+    if (__libcpp_isnan_or_builtin(__x.real()))
+    {
+        if (__libcpp_isinf_or_builtin(__x.imag()))
+            return complex<_Tp>(abs(__x.imag()), __x.real());
+        return complex<_Tp>(__x.real(), __x.real());
+    }
+    if (__libcpp_isinf_or_builtin(__x.imag()))
+        return complex<_Tp>(abs(__x.imag()), copysign(__pi/_Tp(2), __x.imag()));
+    complex<_Tp> __z = log(__x + sqrt(__sqr(__x) - _Tp(1)));
+    return complex<_Tp>(copysign(__z.real(), _Tp(0)), copysign(__z.imag(), __x.imag()));
+}
+
+// atanh
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+atanh(const complex<_Tp>& __x)
+{
+    const _Tp __pi(static_cast<_Tp>(atan2(+0., -0.)));
+    if (__libcpp_isinf_or_builtin(__x.imag()))
+    {
+        return complex<_Tp>(copysign(_Tp(0), __x.real()), copysign(__pi/_Tp(2), __x.imag()));
+    }
+    if (__libcpp_isnan_or_builtin(__x.imag()))
+    {
+        if (__libcpp_isinf_or_builtin(__x.real()) || __x.real() == 0)
+            return complex<_Tp>(copysign(_Tp(0), __x.real()), __x.imag());
+        return complex<_Tp>(__x.imag(), __x.imag());
+    }
+    if (__libcpp_isnan_or_builtin(__x.real()))
+    {
+        return complex<_Tp>(__x.real(), __x.real());
+    }
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        return complex<_Tp>(copysign(_Tp(0), __x.real()), copysign(__pi/_Tp(2), __x.imag()));
+    }
+    if (abs(__x.real()) == _Tp(1) && __x.imag() == _Tp(0))
+    {
+        return complex<_Tp>(copysign(_Tp(INFINITY), __x.real()), copysign(_Tp(0), __x.imag()));
+    }
+    complex<_Tp> __z = log((_Tp(1) + __x) / (_Tp(1) - __x)) / _Tp(2);
+    return complex<_Tp>(copysign(__z.real(), __x.real()), copysign(__z.imag(), __x.imag()));
+}
+
+// sinh
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+sinh(const complex<_Tp>& __x)
+{
+    if (__libcpp_isinf_or_builtin(__x.real()) && !__libcpp_isfinite_or_builtin(__x.imag()))
+        return complex<_Tp>(__x.real(), _Tp(NAN));
+    if (__x.real() == 0 && !__libcpp_isfinite_or_builtin(__x.imag()))
+        return complex<_Tp>(__x.real(), _Tp(NAN));
+    if (__x.imag() == 0 && !__libcpp_isfinite_or_builtin(__x.real()))
+        return __x;
+    return complex<_Tp>(sinh(__x.real()) * cos(__x.imag()), cosh(__x.real()) * sin(__x.imag()));
+}
+
+// cosh
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+cosh(const complex<_Tp>& __x)
+{
+    if (__libcpp_isinf_or_builtin(__x.real()) && !__libcpp_isfinite_or_builtin(__x.imag()))
+        return complex<_Tp>(abs(__x.real()), _Tp(NAN));
+    if (__x.real() == 0 && !__libcpp_isfinite_or_builtin(__x.imag()))
+        return complex<_Tp>(_Tp(NAN), __x.real());
+    if (__x.real() == 0 && __x.imag() == 0)
+        return complex<_Tp>(_Tp(1), __x.imag());
+    if (__x.imag() == 0 && !__libcpp_isfinite_or_builtin(__x.real()))
+        return complex<_Tp>(abs(__x.real()), __x.imag());
+    return complex<_Tp>(cosh(__x.real()) * cos(__x.imag()), sinh(__x.real()) * sin(__x.imag()));
+}
+
+// tanh
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+tanh(const complex<_Tp>& __x)
+{
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        if (!__libcpp_isfinite_or_builtin(__x.imag()))
+            return complex<_Tp>(copysign(_Tp(1), __x.real()), _Tp(0));
+        return complex<_Tp>(copysign(_Tp(1), __x.real()), copysign(_Tp(0), sin(_Tp(2) * __x.imag())));
+    }
+    if (__libcpp_isnan_or_builtin(__x.real()) && __x.imag() == 0)
+        return __x;
+    _Tp __2r(_Tp(2) * __x.real());
+    _Tp __2i(_Tp(2) * __x.imag());
+    _Tp __d(cosh(__2r) + cos(__2i));
+    _Tp __2rsh(sinh(__2r));
+    if (__libcpp_isinf_or_builtin(__2rsh) && __libcpp_isinf_or_builtin(__d))
+        return complex<_Tp>(__2rsh > _Tp(0) ? _Tp(1) : _Tp(-1),
+                            __2i > _Tp(0) ? _Tp(0) : _Tp(-0.));
+    return  complex<_Tp>(__2rsh/__d, sin(__2i)/__d);
+}
+
+// asin
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+asin(const complex<_Tp>& __x)
+{
+    complex<_Tp> __z = asinh(complex<_Tp>(-__x.imag(), __x.real()));
+    return complex<_Tp>(__z.imag(), -__z.real());
+}
+
+// acos
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+acos(const complex<_Tp>& __x)
+{
+    const _Tp __pi(static_cast<_Tp>(atan2(+0., -0.)));
+    if (__libcpp_isinf_or_builtin(__x.real()))
+    {
+        if (__libcpp_isnan_or_builtin(__x.imag()))
+            return complex<_Tp>(__x.imag(), __x.real());
+        if (__libcpp_isinf_or_builtin(__x.imag()))
+        {
+            if (__x.real() < _Tp(0))
+                return complex<_Tp>(_Tp(0.75) * __pi, -__x.imag());
+            return complex<_Tp>(_Tp(0.25) * __pi, -__x.imag());
+        }
+        if (__x.real() < _Tp(0))
+            return complex<_Tp>(__pi, signbit(__x.imag()) ? -__x.real() : __x.real());
+        return complex<_Tp>(_Tp(0), signbit(__x.imag()) ? __x.real() : -__x.real());
+    }
+    if (__libcpp_isnan_or_builtin(__x.real()))
+    {
+        if (__libcpp_isinf_or_builtin(__x.imag()))
+            return complex<_Tp>(__x.real(), -__x.imag());
+        return complex<_Tp>(__x.real(), __x.real());
+    }
+    if (__libcpp_isinf_or_builtin(__x.imag()))
+        return complex<_Tp>(__pi/_Tp(2), -__x.imag());
+    if (__x.real() == 0 && (__x.imag() == 0 || isnan(__x.imag())))
+        return complex<_Tp>(__pi/_Tp(2), -__x.imag());
+    complex<_Tp> __z = log(__x + sqrt(__sqr(__x) - _Tp(1)));
+    if (signbit(__x.imag()))
+        return complex<_Tp>(abs(__z.imag()), abs(__z.real()));
+    return complex<_Tp>(abs(__z.imag()), -abs(__z.real()));
+}
+
+// atan
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+atan(const complex<_Tp>& __x)
+{
+    complex<_Tp> __z = atanh(complex<_Tp>(-__x.imag(), __x.real()));
+    return complex<_Tp>(__z.imag(), -__z.real());
+}
+
+// sin
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+sin(const complex<_Tp>& __x)
+{
+    complex<_Tp> __z = sinh(complex<_Tp>(-__x.imag(), __x.real()));
+    return complex<_Tp>(__z.imag(), -__z.real());
+}
+
+// cos
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+complex<_Tp>
+cos(const complex<_Tp>& __x)
+{
+    return cosh(complex<_Tp>(-__x.imag(), __x.real()));
+}
+
+// tan
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY complex<_Tp>
+tan(const complex<_Tp>& __x)
+{
+    complex<_Tp> __z = tanh(complex<_Tp>(-__x.imag(), __x.real()));
+    return complex<_Tp>(__z.imag(), -__z.real());
+}
+
+#ifndef __cuda_std__
+
+template<class _Tp, class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, complex<_Tp>& __x)
+{
+    if (__is.good())
+    {
+        ws(__is);
+        if (__is.peek() == _CharT('('))
+        {
+            __is.get();
+            _Tp __r;
+            __is >> __r;
+            if (!__is.fail())
+            {
+                ws(__is);
+                _CharT __c = __is.peek();
+                if (__c == _CharT(','))
+                {
+                    __is.get();
+                    _Tp __i;
+                    __is >> __i;
+                    if (!__is.fail())
+                    {
+                        ws(__is);
+                        __c = __is.peek();
+                        if (__c == _CharT(')'))
+                        {
+                            __is.get();
+                            __x = complex<_Tp>(__r, __i);
+                        }
+                        else
+                            __is.setstate(ios_base::failbit);
+                    }
+                    else
+                        __is.setstate(ios_base::failbit);
+                }
+                else if (__c == _CharT(')'))
+                {
+                    __is.get();
+                    __x = complex<_Tp>(__r, _Tp(0));
+                }
+                else
+                    __is.setstate(ios_base::failbit);
+            }
+            else
+                __is.setstate(ios_base::failbit);
+        }
+        else
+        {
+            _Tp __r;
+            __is >> __r;
+            if (!__is.fail())
+                __x = complex<_Tp>(__r, _Tp(0));
+            else
+                __is.setstate(ios_base::failbit);
+        }
+    }
+    else
+        __is.setstate(ios_base::failbit);
+    return __is;
+}
+
+template<class _Tp, class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const complex<_Tp>& __x)
+{
+    basic_ostringstream<_CharT, _Traits> __s;
+    __s.flags(__os.flags());
+    __s.imbue(__os.getloc());
+    __s.precision(__os.precision());
+    __s << '(' << __x.real() << ',' << __x.imag() << ')';
+    return __os << __s.str();
+}
+
+#endif // __cuda_std__
+
+#if _LIBCUDACXX_STD_VER > 11 && defined(_LIBCUDACXX_HAS_STL_LITERALS)
+// Literal suffix for complex number literals [complex.literals]
+inline namespace literals
+{
+  inline namespace complex_literals
+  {
+#ifdef _LIBCUDACXX_HAS_COMPLEX_LONG_DOUBLE
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<long double> operator""il(long double __im)
+    {
+        return { 0.0l, __im };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<long double> operator""il(unsigned long long __im)
+    {
+        return { 0.0l, static_cast<long double>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<double> operator""i(long double __im)
+    {
+        return { 0.0, static_cast<double>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<double> operator""i(unsigned long long __im)
+    {
+        return { 0.0, static_cast<double>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<float> operator""if(long double __im)
+    {
+        return { 0.0f, static_cast<float>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<float> operator""if(unsigned long long __im)
+    {
+        return { 0.0f, static_cast<float>(__im) };
+    }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<double> operator""i(double __im)
+    {
+        return { 0.0, static_cast<double>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<double> operator""i(unsigned long long __im)
+    {
+        return { 0.0, static_cast<double>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<float> operator""if(double __im)
+    {
+        return { 0.0f, static_cast<float>(__im) };
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr complex<float> operator""if(unsigned long long __im)
+    {
+        return { 0.0f, static_cast<float>(__im) };
+    }
+#endif
+  }
+}
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_COMPLEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h
new file mode 100644
index 000000000000..505d11cef35b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/complex.h
@@ -0,0 +1,36 @@
+// -*- C++ -*-
+//===--------------------------- complex.h --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_COMPLEX_H
+#define _LIBCUDACXX_COMPLEX_H
+
+/*
+    complex.h synopsis
+
+#include <ccomplex>
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef __cplusplus
+
+#include <ccomplex>
+
+#else  // __cplusplus
+
+#include_next <complex.h>
+
+#endif  // __cplusplus
+
+#endif  // _LIBCUDACXX_COMPLEX_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable
new file mode 100644
index 000000000000..8f4bf7e1af29
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/condition_variable
@@ -0,0 +1,268 @@
+// -*- C++ -*-
+//===---------------------- condition_variable ----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CONDITION_VARIABLE
+#define _LIBCUDACXX_CONDITION_VARIABLE
+
+/*
+    condition_variable synopsis
+
+namespace std
+{
+
+enum class cv_status { no_timeout, timeout };
+
+class condition_variable
+{
+public:
+    condition_variable();
+    ~condition_variable();
+
+    condition_variable(const condition_variable&) = delete;
+    condition_variable& operator=(const condition_variable&) = delete;
+
+    void notify_one() noexcept;
+    void notify_all() noexcept;
+
+    void wait(unique_lock<mutex>& lock);
+    template <class Predicate>
+        void wait(unique_lock<mutex>& lock, Predicate pred);
+
+    template <class Clock, class Duration>
+        cv_status
+        wait_until(unique_lock<mutex>& lock,
+                   const chrono::time_point<Clock, Duration>& abs_time);
+
+    template <class Clock, class Duration, class Predicate>
+        bool
+        wait_until(unique_lock<mutex>& lock,
+                   const chrono::time_point<Clock, Duration>& abs_time,
+                   Predicate pred);
+
+    template <class Rep, class Period>
+        cv_status
+        wait_for(unique_lock<mutex>& lock,
+                 const chrono::duration<Rep, Period>& rel_time);
+
+    template <class Rep, class Period, class Predicate>
+        bool
+        wait_for(unique_lock<mutex>& lock,
+                 const chrono::duration<Rep, Period>& rel_time,
+                 Predicate pred);
+
+    typedef pthread_cond_t* native_handle_type;
+    native_handle_type native_handle();
+};
+
+void notify_all_at_thread_exit(condition_variable& cond, unique_lock<mutex> lk);
+
+class condition_variable_any
+{
+public:
+    condition_variable_any();
+    ~condition_variable_any();
+
+    condition_variable_any(const condition_variable_any&) = delete;
+    condition_variable_any& operator=(const condition_variable_any&) = delete;
+
+    void notify_one() noexcept;
+    void notify_all() noexcept;
+
+    template <class Lock>
+        void wait(Lock& lock);
+    template <class Lock, class Predicate>
+        void wait(Lock& lock, Predicate pred);
+
+    template <class Lock, class Clock, class Duration>
+        cv_status
+        wait_until(Lock& lock,
+                   const chrono::time_point<Clock, Duration>& abs_time);
+
+    template <class Lock, class Clock, class Duration, class Predicate>
+        bool
+        wait_until(Lock& lock,
+                   const chrono::time_point<Clock, Duration>& abs_time,
+                   Predicate pred);
+
+    template <class Lock, class Rep, class Period>
+        cv_status
+        wait_for(Lock& lock,
+                 const chrono::duration<Rep, Period>& rel_time);
+
+    template <class Lock, class Rep, class Period, class Predicate>
+        bool
+        wait_for(Lock& lock,
+                 const chrono::duration<Rep, Period>& rel_time,
+                 Predicate pred);
+};
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__mutex_base>
+#include <memory>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+class _LIBCUDACXX_TYPE_VIS condition_variable_any
+{
+    condition_variable __cv_;
+    shared_ptr<mutex>  __mut_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    condition_variable_any();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void notify_one() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void notify_all() _NOEXCEPT;
+
+    template <class _Lock>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        void wait(_Lock& __lock);
+    template <class _Lock, class _Predicate>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void wait(_Lock& __lock, _Predicate __pred);
+
+    template <class _Lock, class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        cv_status
+        wait_until(_Lock& __lock,
+                   const chrono::time_point<_Clock, _Duration>& __t);
+
+    template <class _Lock, class _Clock, class _Duration, class _Predicate>
+        bool
+        _LIBCUDACXX_INLINE_VISIBILITY
+        wait_until(_Lock& __lock,
+                   const chrono::time_point<_Clock, _Duration>& __t,
+                   _Predicate __pred);
+
+    template <class _Lock, class _Rep, class _Period>
+        cv_status
+        _LIBCUDACXX_INLINE_VISIBILITY
+        wait_for(_Lock& __lock,
+                 const chrono::duration<_Rep, _Period>& __d);
+
+    template <class _Lock, class _Rep, class _Period, class _Predicate>
+        bool
+        _LIBCUDACXX_INLINE_VISIBILITY
+        wait_for(_Lock& __lock,
+                 const chrono::duration<_Rep, _Period>& __d,
+                 _Predicate __pred);
+};
+
+inline
+condition_variable_any::condition_variable_any()
+    : __mut_(make_shared<mutex>()) {}
+
+inline
+void
+condition_variable_any::notify_one() _NOEXCEPT
+{
+    {lock_guard<mutex> __lx(*__mut_);}
+    __cv_.notify_one();
+}
+
+inline
+void
+condition_variable_any::notify_all() _NOEXCEPT
+{
+    {lock_guard<mutex> __lx(*__mut_);}
+    __cv_.notify_all();
+}
+
+struct __lock_external
+{
+    template <class _Lock>
+    void operator()(_Lock* __m) {__m->lock();}
+};
+
+template <class _Lock>
+void
+condition_variable_any::wait(_Lock& __lock)
+{
+    shared_ptr<mutex> __mut = __mut_;
+    unique_lock<mutex> __lk(*__mut);
+    __lock.unlock();
+    unique_ptr<_Lock, __lock_external> __lxx(&__lock);
+    lock_guard<unique_lock<mutex> > __lx(__lk, adopt_lock);
+    __cv_.wait(__lk);
+}  // __mut_.unlock(), __lock.lock()
+
+template <class _Lock, class _Predicate>
+inline
+void
+condition_variable_any::wait(_Lock& __lock, _Predicate __pred)
+{
+    while (!__pred())
+        wait(__lock);
+}
+
+template <class _Lock, class _Clock, class _Duration>
+cv_status
+condition_variable_any::wait_until(_Lock& __lock,
+                                   const chrono::time_point<_Clock, _Duration>& __t)
+{
+    shared_ptr<mutex> __mut = __mut_;
+    unique_lock<mutex> __lk(*__mut);
+    __lock.unlock();
+    unique_ptr<_Lock, __lock_external> __lxx(&__lock);
+    lock_guard<unique_lock<mutex> > __lx(__lk, adopt_lock);
+    return __cv_.wait_until(__lk, __t);
+}  // __mut_.unlock(), __lock.lock()
+
+template <class _Lock, class _Clock, class _Duration, class _Predicate>
+inline
+bool
+condition_variable_any::wait_until(_Lock& __lock,
+                                   const chrono::time_point<_Clock, _Duration>& __t,
+                                   _Predicate __pred)
+{
+    while (!__pred())
+        if (wait_until(__lock, __t) == cv_status::timeout)
+            return __pred();
+    return true;
+}
+
+template <class _Lock, class _Rep, class _Period>
+inline
+cv_status
+condition_variable_any::wait_for(_Lock& __lock,
+                                 const chrono::duration<_Rep, _Period>& __d)
+{
+    return wait_until(__lock, chrono::steady_clock::now() + __d);
+}
+
+template <class _Lock, class _Rep, class _Period, class _Predicate>
+inline
+bool
+condition_variable_any::wait_for(_Lock& __lock,
+                                 const chrono::duration<_Rep, _Period>& __d,
+                                 _Predicate __pred)
+{
+    return wait_until(__lock, chrono::steady_clock::now() + __d,
+                      _CUDA_VSTD::move(__pred));
+}
+
+_LIBCUDACXX_FUNC_VIS
+void notify_all_at_thread_exit(condition_variable& cond, unique_lock<mutex> lk);
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+#endif  // _LIBCUDACXX_CONDITION_VARIABLE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp
new file mode 100644
index 000000000000..d6b46eff984e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csetjmp
@@ -0,0 +1,47 @@
+// -*- C++ -*-
+//===--------------------------- csetjmp ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSETJMP
+#define _LIBCUDACXX_CSETJMP
+
+/*
+    csetjmp synopsis
+
+Macros:
+
+    setjmp
+
+namespace std
+{
+
+Types:
+
+    jmp_buf
+
+void longjmp(jmp_buf env, int val);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <setjmp.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::jmp_buf;
+using ::longjmp;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CSETJMP
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csignal b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csignal
new file mode 100644
index 000000000000..39c13590e024
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/csignal
@@ -0,0 +1,57 @@
+// -*- C++ -*-
+//===--------------------------- csignal ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSIGNAL
+#define _LIBCUDACXX_CSIGNAL
+
+/*
+    csignal synopsis
+
+Macros:
+
+    SIG_DFL
+    SIG_ERR
+    SIG_IGN
+    SIGABRT
+    SIGFPE
+    SIGILL
+    SIGINT
+    SIGSEGV
+    SIGTERM
+
+namespace std
+{
+
+Types:
+
+    sig_atomic_t
+
+void (*signal(int sig, void (*func)(int)))(int);
+int raise(int sig);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <signal.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::sig_atomic_t;
+using ::signal;
+using ::raise;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CSIGNAL
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg
new file mode 100644
index 000000000000..a7b2fb9a1d04
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdarg
@@ -0,0 +1,54 @@
+// -*- C++ -*-
+//===--------------------------- cstdarg ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTDARG
+#define _LIBCUDACXX_CSTDARG
+
+/*
+    cstdarg synopsis
+
+Macros:
+
+    type va_arg(va_list ap, type);
+    void va_copy(va_list dest, va_list src);  // C99
+    void va_end(va_list ap);
+    void va_start(va_list ap, parmN);
+
+namespace std
+{
+
+Types:
+
+    va_list
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <stdarg.h>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::va_list;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CSTDARG
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool
new file mode 100644
index 000000000000..026f9574fcd9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdbool
@@ -0,0 +1,38 @@
+// -*- C++ -*-
+//===--------------------------- cstdbool ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTDBOOL
+#define _LIBCUDACXX_CSTDBOOL
+
+/*
+    cstdbool synopsis
+
+Macros:
+
+    __bool_true_false_are_defined
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#undef __bool_true_false_are_defined
+#define __bool_true_false_are_defined 1
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CSTDBOOL
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef
new file mode 100644
index 000000000000..9f1ce122fb60
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstddef
@@ -0,0 +1,143 @@
+// -*- C++ -*-
+//===--------------------------- cstddef ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTDDEF
+#define _LIBCUDACXX_CSTDDEF
+
+/*
+    cstddef synopsis
+
+Macros:
+
+    offsetof(type,member-designator)
+    NULL
+
+namespace std
+{
+
+Types:
+
+    ptrdiff_t
+    size_t
+    max_align_t
+    nullptr_t
+    byte // C++17
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <version>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifndef __cuda_std__
+// Don't include our own <stddef.h>; we don't want to declare ::nullptr_t.
+#ifdef _MSC_VER
+    #include <../ucrt/stddef.h>
+#else
+    #include_next <stddef.h>
+#endif
+#include <__nullptr>
+#endif //__cuda_std__
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::ptrdiff_t;
+using ::size_t;
+
+#if defined(__CLANG_MAX_ALIGN_T_DEFINED) || defined(_GCC_MAX_ALIGN_T) || \
+    defined(__DEFINED_max_align_t) || defined(__NetBSD__)
+// Re-use the compiler's <stddef.h> max_align_t where possible.
+using ::max_align_t;
+#else
+typedef long double max_align_t;
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 11
+#ifdef _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+_LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+#else
+namespace std  // purposefully not versioned
+{
+#endif //_LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+enum class byte : unsigned char {};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte  operator| (byte  __lhs, byte __rhs) noexcept
+{
+    return static_cast<byte>(
+      static_cast<unsigned char>(
+         static_cast<unsigned int>(__lhs) | static_cast<unsigned int>(__rhs)
+    ));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte& operator|=(byte& __lhs, byte __rhs) noexcept
+{ return __lhs = __lhs | __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte  operator& (byte  __lhs, byte __rhs) noexcept
+{
+    return static_cast<byte>(
+      static_cast<unsigned char>(
+         static_cast<unsigned int>(__lhs) & static_cast<unsigned int>(__rhs)
+    ));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte& operator&=(byte& __lhs, byte __rhs) noexcept
+{ return __lhs = __lhs & __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte  operator^ (byte  __lhs, byte __rhs) noexcept
+{
+    return static_cast<byte>(
+      static_cast<unsigned char>(
+         static_cast<unsigned int>(__lhs) ^ static_cast<unsigned int>(__rhs)
+    ));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte& operator^=(byte& __lhs, byte __rhs) noexcept
+{ return __lhs = __lhs ^ __rhs; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr byte  operator~ (byte __b) noexcept
+{
+    return static_cast<byte>(
+      static_cast<unsigned char>(
+        ~static_cast<unsigned int>(__b)
+    ));
+}
+
+#ifdef _LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+_LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+#else
+}
+#endif //_LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+
+#ifndef __cuda_std__
+#include <type_traits>  // rest of byte
+#endif
+#endif
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CSTDDEF
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint
new file mode 100644
index 000000000000..992a227b86ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdint
@@ -0,0 +1,197 @@
+// -*- C++ -*-
+//===--------------------------- cstdint ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTDINT
+#define _LIBCUDACXX_CSTDINT
+
+/*
+    cstdint synopsis
+
+Macros:
+
+    INT8_MIN
+    INT16_MIN
+    INT32_MIN
+    INT64_MIN
+
+    INT8_MAX
+    INT16_MAX
+    INT32_MAX
+    INT64_MAX
+
+    UINT8_MAX
+    UINT16_MAX
+    UINT32_MAX
+    UINT64_MAX
+
+    INT_LEAST8_MIN
+    INT_LEAST16_MIN
+    INT_LEAST32_MIN
+    INT_LEAST64_MIN
+
+    INT_LEAST8_MAX
+    INT_LEAST16_MAX
+    INT_LEAST32_MAX
+    INT_LEAST64_MAX
+
+    UINT_LEAST8_MAX
+    UINT_LEAST16_MAX
+    UINT_LEAST32_MAX
+    UINT_LEAST64_MAX
+
+    INT_FAST8_MIN
+    INT_FAST16_MIN
+    INT_FAST32_MIN
+    INT_FAST64_MIN
+
+    INT_FAST8_MAX
+    INT_FAST16_MAX
+    INT_FAST32_MAX
+    INT_FAST64_MAX
+
+    UINT_FAST8_MAX
+    UINT_FAST16_MAX
+    UINT_FAST32_MAX
+    UINT_FAST64_MAX
+
+    INTPTR_MIN
+    INTPTR_MAX
+    UINTPTR_MAX
+
+    INTMAX_MIN
+    INTMAX_MAX
+
+    UINTMAX_MAX
+
+    PTRDIFF_MIN
+    PTRDIFF_MAX
+
+    SIG_ATOMIC_MIN
+    SIG_ATOMIC_MAX
+
+    SIZE_MAX
+
+    WCHAR_MIN
+    WCHAR_MAX
+
+    WINT_MIN
+    WINT_MAX
+
+    INT8_C(value)
+    INT16_C(value)
+    INT32_C(value)
+    INT64_C(value)
+
+    UINT8_C(value)
+    UINT16_C(value)
+    UINT32_C(value)
+    UINT64_C(value)
+
+    INTMAX_C(value)
+    UINTMAX_C(value)
+
+namespace std
+{
+
+Types:
+
+    int8_t
+    int16_t
+    int32_t
+    int64_t
+
+    uint8_t
+    uint16_t
+    uint32_t
+    uint64_t
+
+    int_least8_t
+    int_least16_t
+    int_least32_t
+    int_least64_t
+
+    uint_least8_t
+    uint_least16_t
+    uint_least32_t
+    uint_least64_t
+
+    int_fast8_t
+    int_fast16_t
+    int_fast32_t
+    int_fast64_t
+
+    uint_fast8_t
+    uint_fast16_t
+    uint_fast32_t
+    uint_fast64_t
+
+    intptr_t
+    uintptr_t
+
+    intmax_t
+    uintmax_t
+
+}  // std
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <stdint.h>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using::int8_t;
+using::int16_t;
+using::int32_t;
+using::int64_t;
+
+using::uint8_t;
+using::uint16_t;
+using::uint32_t;
+using::uint64_t;
+
+using::int_least8_t;
+using::int_least16_t;
+using::int_least32_t;
+using::int_least64_t;
+
+using::uint_least8_t;
+using::uint_least16_t;
+using::uint_least32_t;
+using::uint_least64_t;
+
+using::int_fast8_t;
+using::int_fast16_t;
+using::int_fast32_t;
+using::int_fast64_t;
+
+using::uint_fast8_t;
+using::uint_fast16_t;
+using::uint_fast32_t;
+using::uint_fast64_t;
+
+using::intptr_t;
+using::uintptr_t;
+
+using::intmax_t;
+using::uintmax_t;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CSTDINT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio
new file mode 100644
index 000000000000..e17d75386e00
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdio
@@ -0,0 +1,171 @@
+// -*- C++ -*-
+//===---------------------------- cstdio ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTDIO
+#define _LIBCUDACXX_CSTDIO
+
+/*
+    cstdio synopsis
+
+Macros:
+
+    BUFSIZ
+    EOF
+    FILENAME_MAX
+    FOPEN_MAX
+    L_tmpnam
+    NULL
+    SEEK_CUR
+    SEEK_END
+    SEEK_SET
+    TMP_MAX
+    _IOFBF
+    _IOLBF
+    _IONBF
+    stderr
+    stdin
+    stdout
+
+namespace std
+{
+
+Types:
+
+FILE
+fpos_t
+size_t
+
+int remove(const char* filename);
+int rename(const char* old, const char* new);
+FILE* tmpfile(void);
+char* tmpnam(char* s);
+int fclose(FILE* stream);
+int fflush(FILE* stream);
+FILE* fopen(const char* restrict filename, const char* restrict mode);
+FILE* freopen(const char* restrict filename, const char * restrict mode,
+              FILE * restrict stream);
+void setbuf(FILE* restrict stream, char* restrict buf);
+int setvbuf(FILE* restrict stream, char* restrict buf, int mode, size_t size);
+int fprintf(FILE* restrict stream, const char* restrict format, ...);
+int fscanf(FILE* restrict stream, const char * restrict format, ...);
+int printf(const char* restrict format, ...);
+int scanf(const char* restrict format, ...);
+int snprintf(char* restrict s, size_t n, const char* restrict format, ...);    // C99
+int sprintf(char* restrict s, const char* restrict format, ...);
+int sscanf(const char* restrict s, const char* restrict format, ...);
+int vfprintf(FILE* restrict stream, const char* restrict format, va_list arg);
+int vfscanf(FILE* restrict stream, const char* restrict format, va_list arg);  // C99
+int vprintf(const char* restrict format, va_list arg);
+int vscanf(const char* restrict format, va_list arg);                          // C99
+int vsnprintf(char* restrict s, size_t n, const char* restrict format,         // C99
+              va_list arg);
+int vsprintf(char* restrict s, const char* restrict format, va_list arg);
+int vsscanf(const char* restrict s, const char* restrict format, va_list arg); // C99
+int fgetc(FILE* stream);
+char* fgets(char* restrict s, int n, FILE* restrict stream);
+int fputc(int c, FILE* stream);
+int fputs(const char* restrict s, FILE* restrict stream);
+int getc(FILE* stream);
+int getchar(void);
+char* gets(char* s);  // removed in C++14
+int putc(int c, FILE* stream);
+int putchar(int c);
+int puts(const char* s);
+int ungetc(int c, FILE* stream);
+size_t fread(void* restrict ptr, size_t size, size_t nmemb,
+             FILE* restrict stream);
+size_t fwrite(const void* restrict ptr, size_t size, size_t nmemb,
+              FILE* restrict stream);
+int fgetpos(FILE* restrict stream, fpos_t* restrict pos);
+int fseek(FILE* stream, long offset, int whence);
+int fsetpos(FILE*stream, const fpos_t* pos);
+long ftell(FILE* stream);
+void rewind(FILE* stream);
+void clearerr(FILE* stream);
+int feof(FILE* stream);
+int ferror(FILE* stream);
+void perror(const char* s);
+
+}  // std
+*/
+
+#include <__config>
+#include <stdio.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::FILE;
+using ::fpos_t;
+using ::size_t;
+
+using ::fclose;
+using ::fflush;
+using ::setbuf;
+using ::setvbuf;
+using ::fprintf;
+using ::fscanf;
+using ::snprintf;
+using ::sprintf;
+using ::sscanf;
+using ::vfprintf;
+using ::vfscanf;
+using ::vsscanf;
+using ::vsnprintf;
+using ::vsprintf;
+using ::fgetc;
+using ::fgets;
+using ::fputc;
+using ::fputs;
+using ::getc;
+using ::putc;
+using ::ungetc;
+using ::fread;
+using ::fwrite;
+using ::fgetpos;
+using ::fseek;
+using ::fsetpos;
+using ::ftell;
+using ::rewind;
+using ::clearerr;
+using ::feof;
+using ::ferror;
+using ::perror;
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+using ::fopen;
+using ::freopen;
+using ::remove;
+using ::rename;
+using ::tmpfile;
+using ::tmpnam;
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_STDIN
+using ::getchar;
+#if _LIBCUDACXX_STD_VER <= 11 && !defined(_LIBCUDACXX_C_HAS_NO_GETS)
+using ::gets;
+#endif
+using ::scanf;
+using ::vscanf;
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_STDOUT
+using ::printf;
+using ::putchar;
+using ::puts;
+using ::vprintf;
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CSTDIO
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib
new file mode 100644
index 000000000000..a6be52c9a6b5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstdlib
@@ -0,0 +1,182 @@
+// -*- C++ -*-
+//===--------------------------- cstdlib ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTDLIB
+#define _LIBCUDACXX_CSTDLIB
+
+/*
+    cstdlib synopsis
+
+Macros:
+
+    EXIT_FAILURE
+    EXIT_SUCCESS
+    MB_CUR_MAX
+    NULL
+    RAND_MAX
+
+namespace std
+{
+
+Types:
+
+    size_t
+    div_t
+    ldiv_t
+    lldiv_t                                                               // C99
+
+double    atof (const char* nptr);
+int       atoi (const char* nptr);
+long      atol (const char* nptr);
+long long atoll(const char* nptr);                                        // C99
+double             strtod  (const char* restrict nptr, char** restrict endptr);
+float              strtof  (const char* restrict nptr, char** restrict endptr); // C99
+long double        strtold (const char* restrict nptr, char** restrict endptr); // C99
+long               strtol  (const char* restrict nptr, char** restrict endptr, int base);
+long long          strtoll (const char* restrict nptr, char** restrict endptr, int base); // C99
+unsigned long      strtoul (const char* restrict nptr, char** restrict endptr, int base);
+unsigned long long strtoull(const char* restrict nptr, char** restrict endptr, int base); // C99
+int rand(void);
+void srand(unsigned int seed);
+void* calloc(size_t nmemb, size_t size);
+void free(void* ptr);
+void* malloc(size_t size);
+void* realloc(void* ptr, size_t size);
+void abort(void);
+int atexit(void (*func)(void));
+void exit(int status);
+void _Exit(int status);
+char* getenv(const char* name);
+int system(const char* string);
+void* bsearch(const void* key, const void* base, size_t nmemb, size_t size,
+              int (*compar)(const void *, const void *));
+void qsort(void* base, size_t nmemb, size_t size,
+           int (*compar)(const void *, const void *));
+int         abs(      int j);
+long        abs(     long j);
+long long   abs(long long j);                                             // C++0X
+long       labs(     long j);
+long long llabs(long long j);                                             // C99
+div_t     div(      int numer,       int denom);
+ldiv_t    div(     long numer,      long denom);
+lldiv_t   div(long long numer, long long denom);                          // C++0X
+ldiv_t   ldiv(     long numer,      long denom);
+lldiv_t lldiv(long long numer, long long denom);                          // C99
+int mblen(const char* s, size_t n);
+int mbtowc(wchar_t* restrict pwc, const char* restrict s, size_t n);
+int wctomb(char* s, wchar_t wchar);
+size_t mbstowcs(wchar_t* restrict pwcs, const char* restrict s, size_t n);
+size_t wcstombs(char* restrict s, const wchar_t* restrict pwcs, size_t n);
+int at_quick_exit(void (*func)(void))                                     // C++11
+void quick_exit(int status);                                              // C++11
+void *aligned_alloc(size_t alignment, size_t size);                       // C11
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <stdlib.h>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef __GNUC__
+#define _LIBCUDACXX_UNREACHABLE() __builtin_unreachable()
+#else
+#if defined(__CUDA_ARCH__)
+#  define _LIBCUDACXX_UNREACHABLE() __trap()
+#else
+#  define _LIBCUDACXX_UNREACHABLE() _CUDA_VSTD::abort()
+#endif
+#endif
+
+#ifdef _LIBCUDACXX_COMPILER_PGI
+#define _LIBCUDACXX_UNREACHABLE_AFTER_SWITCH()
+#else
+#define _LIBCUDACXX_UNREACHABLE_AFTER_SWITCH() _LIBCUDACXX_UNREACHABLE()
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if !defined(_LIBCUDACXX_COMPILER_NVRTC)
+using ::size_t;
+using ::div_t;
+using ::ldiv_t;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::lldiv_t;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::atof;
+using ::atoi;
+using ::atol;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::atoll;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::strtod;
+using ::strtof;
+using ::strtold;
+using ::strtol;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::strtoll;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::strtoul;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::strtoull;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::rand;
+using ::srand;
+using ::calloc;
+using ::free;
+using ::malloc;
+using ::realloc;
+using ::abort;
+using ::atexit;
+using ::exit;
+using ::_Exit;
+#ifndef _LIBCUDACXX_WINDOWS_STORE_APP
+using ::getenv;
+using ::system;
+#endif
+using ::bsearch;
+using ::qsort;
+using ::abs;
+using ::labs;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::llabs;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::div;
+using ::ldiv;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::lldiv;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::mblen;
+using ::mbtowc;
+using ::wctomb;
+using ::mbstowcs;
+using ::wcstombs;
+#if !defined(_LIBCUDACXX_CXX03_LANG) && defined(_LIBCUDACXX_HAS_QUICK_EXIT)
+using ::at_quick_exit;
+using ::quick_exit;
+#endif
+#if _LIBCUDACXX_STD_VER > 14 && defined(_LIBCUDACXX_HAS_C11_FEATURES)
+using ::aligned_alloc;
+#endif
+#endif // !defined(_LIBCUDACXX_COMPILER_NVRTC)
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CSTDLIB
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstring b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstring
new file mode 100644
index 000000000000..06e1c95f8b3a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cstring
@@ -0,0 +1,103 @@
+// -*- C++ -*-
+//===--------------------------- cstring ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CSTRING
+#define _LIBCUDACXX_CSTRING
+
+/*
+    cstring synopsis
+
+Macros:
+
+    NULL
+
+namespace std
+{
+
+Types:
+
+    size_t
+
+void* memcpy(void* restrict s1, const void* restrict s2, size_t n);
+void* memmove(void* s1, const void* s2, size_t n);
+char* strcpy (char* restrict s1, const char* restrict s2);
+char* strncpy(char* restrict s1, const char* restrict s2, size_t n);
+char* strcat (char* restrict s1, const char* restrict s2);
+char* strncat(char* restrict s1, const char* restrict s2, size_t n);
+int memcmp(const void* s1, const void* s2, size_t n);
+int strcmp (const char* s1, const char* s2);
+int strncmp(const char* s1, const char* s2, size_t n);
+int strcoll(const char* s1, const char* s2);
+size_t strxfrm(char* restrict s1, const char* restrict s2, size_t n);
+const void* memchr(const void* s, int c, size_t n);
+      void* memchr(      void* s, int c, size_t n);
+const char* strchr(const char* s, int c);
+      char* strchr(      char* s, int c);
+size_t strcspn(const char* s1, const char* s2);
+const char* strpbrk(const char* s1, const char* s2);
+      char* strpbrk(      char* s1, const char* s2);
+const char* strrchr(const char* s, int c);
+      char* strrchr(      char* s, int c);
+size_t strspn(const char* s1, const char* s2);
+const char* strstr(const char* s1, const char* s2);
+      char* strstr(      char* s1, const char* s2);
+char* strtok(char* restrict s1, const char* restrict s2);
+void* memset(void* s, int c, size_t n);
+char* strerror(int errnum);
+size_t strlen(const char* s);
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <string.h>
+#include <__pragma_push>
+#endif //__simet__ <- a tribute to ogiroux
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::size_t;
+using ::memcpy;
+using ::memmove;
+using ::strcpy;
+using ::strncpy;
+using ::strcat;
+using ::strncat;
+using ::memcmp;
+using ::strcmp;
+using ::strncmp;
+using ::strcoll;
+using ::strxfrm;
+using ::memchr;
+using ::strchr;
+using ::strcspn;
+using ::strpbrk;
+using ::strrchr;
+using ::strspn;
+using ::strstr;
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_UNSAFE_C_FUNCTIONS
+using ::strtok;
+#endif
+using ::memset;
+using ::strerror;
+using ::strlen;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CSTRING
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath
new file mode 100644
index 000000000000..942bc90b4069
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctgmath
@@ -0,0 +1,28 @@
+// -*- C++ -*-
+//===-------------------------- ctgmath -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CTGMATH
+#define _LIBCUDACXX_CTGMATH
+
+/*
+    ctgmath synopsis
+
+#include <ccomplex>
+#include <cmath>
+
+*/
+
+#include <ccomplex>
+#include <cmath>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#endif  // _LIBCUDACXX_CTGMATH
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctime b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctime
new file mode 100644
index 000000000000..8cce5463c294
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctime
@@ -0,0 +1,94 @@
+// -*- C++ -*-
+//===---------------------------- ctime -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CTIME
+#define _LIBCUDACXX_CTIME
+
+/*
+    ctime synopsis
+
+Macros:
+
+    NULL
+    CLOCKS_PER_SEC
+    TIME_UTC // C++17
+
+namespace std
+{
+
+Types:
+
+    clock_t
+    size_t
+    time_t
+    tm
+    timespec // C++17
+
+clock_t clock();
+double difftime(time_t time1, time_t time0);
+time_t mktime(tm* timeptr);
+time_t time(time_t* timer);
+char* asctime(const tm* timeptr);
+char* ctime(const time_t* timer);
+tm*    gmtime(const time_t* timer);
+tm* localtime(const time_t* timer);
+size_t strftime(char* restrict s, size_t maxsize, const char* restrict format,
+                const tm* restrict timeptr);
+int timespec_get( struct timespec *ts, int base); // C++17
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <time.h>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::clock_t;
+using ::size_t;
+using ::time_t;
+
+#ifndef __CUDACC_RTC__
+
+using ::tm;
+#if _LIBCUDACXX_STD_VER > 14 && defined(_LIBCUDACXX_HAS_C11_FEATURES)
+using ::timespec;
+#endif
+using ::clock;
+using ::difftime;
+using ::mktime;
+using ::time;
+#ifndef _LIBCUDACXX_HAS_NO_THREAD_UNSAFE_C_FUNCTIONS
+using ::asctime;
+using ::ctime;
+using ::gmtime;
+using ::localtime;
+#endif
+using ::strftime;
+#if _LIBCUDACXX_STD_VER > 14 && defined(_LIBCUDACXX_HAS_TIMESPEC_GET)
+using ::timespec_get;
+
+#endif
+
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_CTIME
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h
new file mode 100644
index 000000000000..4957f6eb1340
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ctype.h
@@ -0,0 +1,59 @@
+// -*- C++ -*-
+//===---------------------------- ctype.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CTYPE_H
+#define _LIBCUDACXX_CTYPE_H
+
+/*
+    ctype.h synopsis
+
+int isalnum(int c);
+int isalpha(int c);
+int isblank(int c);  // C99
+int iscntrl(int c);
+int isdigit(int c);
+int isgraph(int c);
+int islower(int c);
+int isprint(int c);
+int ispunct(int c);
+int isspace(int c);
+int isupper(int c);
+int isxdigit(int c);
+int tolower(int c);
+int toupper(int c);
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <ctype.h>
+
+#ifdef __cplusplus
+
+#undef isalnum
+#undef isalpha
+#undef isblank
+#undef iscntrl
+#undef isdigit
+#undef isgraph
+#undef islower
+#undef isprint
+#undef ispunct
+#undef isspace
+#undef isupper
+#undef isxdigit
+#undef tolower
+#undef toupper
+
+#endif
+
+#endif  // _LIBCUDACXX_CTYPE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar
new file mode 100644
index 000000000000..750b774c6a1e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwchar
@@ -0,0 +1,192 @@
+// -*- C++ -*-
+//===--------------------------- cwchar -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CWCHAR
+#define _LIBCUDACXX_CWCHAR
+
+/*
+    cwchar synopsis
+
+Macros:
+
+    NULL
+    WCHAR_MAX
+    WCHAR_MIN
+    WEOF
+
+namespace std
+{
+
+Types:
+
+    mbstate_t
+    size_t
+    tm
+    wint_t
+
+int fwprintf(FILE* restrict stream, const wchar_t* restrict format, ...);
+int fwscanf(FILE* restrict stream, const wchar_t* restrict format, ...);
+int swprintf(wchar_t* restrict s, size_t n, const wchar_t* restrict format, ...);
+int swscanf(const wchar_t* restrict s, const wchar_t* restrict format, ...);
+int vfwprintf(FILE* restrict stream, const wchar_t* restrict format, va_list arg);
+int vfwscanf(FILE* restrict stream, const wchar_t* restrict format, va_list arg);  // C99
+int vswprintf(wchar_t* restrict s, size_t n, const wchar_t* restrict format, va_list arg);
+int vswscanf(const wchar_t* restrict s, const wchar_t* restrict format, va_list arg);  // C99
+int vwprintf(const wchar_t* restrict format, va_list arg);
+int vwscanf(const wchar_t* restrict format, va_list arg);  // C99
+int wprintf(const wchar_t* restrict format, ...);
+int wscanf(const wchar_t* restrict format, ...);
+wint_t fgetwc(FILE* stream);
+wchar_t* fgetws(wchar_t* restrict s, int n, FILE* restrict stream);
+wint_t fputwc(wchar_t c, FILE* stream);
+int fputws(const wchar_t* restrict s, FILE* restrict stream);
+int fwide(FILE* stream, int mode);
+wint_t getwc(FILE* stream);
+wint_t getwchar();
+wint_t putwc(wchar_t c, FILE* stream);
+wint_t putwchar(wchar_t c);
+wint_t ungetwc(wint_t c, FILE* stream);
+double wcstod(const wchar_t* restrict nptr, wchar_t** restrict endptr);
+float wcstof(const wchar_t* restrict nptr, wchar_t** restrict endptr);         // C99
+long double wcstold(const wchar_t* restrict nptr, wchar_t** restrict endptr);  // C99
+long wcstol(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+long long wcstoll(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);  // C99
+unsigned long wcstoul(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+unsigned long long wcstoull(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);  // C99
+wchar_t* wcscpy(wchar_t* restrict s1, const wchar_t* restrict s2);
+wchar_t* wcsncpy(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+wchar_t* wcscat(wchar_t* restrict s1, const wchar_t* restrict s2);
+wchar_t* wcsncat(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+int wcscmp(const wchar_t* s1, const wchar_t* s2);
+int wcscoll(const wchar_t* s1, const wchar_t* s2);
+int wcsncmp(const wchar_t* s1, const wchar_t* s2, size_t n);
+size_t wcsxfrm(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+const wchar_t* wcschr(const wchar_t* s, wchar_t c);
+      wchar_t* wcschr(      wchar_t* s, wchar_t c);
+size_t wcscspn(const wchar_t* s1, const wchar_t* s2);
+size_t wcslen(const wchar_t* s);
+const wchar_t* wcspbrk(const wchar_t* s1, const wchar_t* s2);
+      wchar_t* wcspbrk(      wchar_t* s1, const wchar_t* s2);
+const wchar_t* wcsrchr(const wchar_t* s, wchar_t c);
+      wchar_t* wcsrchr(      wchar_t* s, wchar_t c);
+size_t wcsspn(const wchar_t* s1, const wchar_t* s2);
+const wchar_t* wcsstr(const wchar_t* s1, const wchar_t* s2);
+      wchar_t* wcsstr(      wchar_t* s1, const wchar_t* s2);
+wchar_t* wcstok(wchar_t* restrict s1, const wchar_t* restrict s2, wchar_t** restrict ptr);
+const wchar_t* wmemchr(const wchar_t* s, wchar_t c, size_t n);
+      wchar_t* wmemchr(      wchar_t* s, wchar_t c, size_t n);
+int wmemcmp(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+wchar_t* wmemcpy(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+wchar_t* wmemmove(wchar_t* s1, const wchar_t* s2, size_t n);
+wchar_t* wmemset(wchar_t* s, wchar_t c, size_t n);
+size_t wcsftime(wchar_t* restrict s, size_t maxsize, const wchar_t* restrict format,
+                const tm* restrict timeptr);
+wint_t btowc(int c);
+int wctob(wint_t c);
+int mbsinit(const mbstate_t* ps);
+size_t mbrlen(const char* restrict s, size_t n, mbstate_t* restrict ps);
+size_t mbrtowc(wchar_t* restrict pwc, const char* restrict s, size_t n, mbstate_t* restrict ps);
+size_t wcrtomb(char* restrict s, wchar_t wc, mbstate_t* restrict ps);
+size_t mbsrtowcs(wchar_t* restrict dst, const char** restrict src, size_t len,
+                 mbstate_t* restrict ps);
+size_t wcsrtombs(char* restrict dst, const wchar_t** restrict src, size_t len,
+                 mbstate_t* restrict ps);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <cwctype>
+#include <wchar.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::mbstate_t;
+using ::size_t;
+using ::tm;
+using ::wint_t;
+using ::FILE;
+using ::fwprintf;
+using ::fwscanf;
+using ::swprintf;
+using ::vfwprintf;
+using ::vswprintf;
+using ::swscanf;
+using ::vfwscanf;
+using ::vswscanf;
+using ::fgetwc;
+using ::fgetws;
+using ::fputwc;
+using ::fputws;
+using ::fwide;
+using ::getwc;
+using ::putwc;
+using ::ungetwc;
+using ::wcstod;
+using ::wcstof;
+using ::wcstold;
+using ::wcstol;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::wcstoll;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::wcstoul;
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::wcstoull;
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+using ::wcscpy;
+using ::wcsncpy;
+using ::wcscat;
+using ::wcsncat;
+using ::wcscmp;
+using ::wcscoll;
+using ::wcsncmp;
+using ::wcsxfrm;
+using ::wcschr;
+using ::wcspbrk;
+using ::wcsrchr;
+using ::wcsstr;
+using ::wmemchr;
+using ::wcscspn;
+using ::wcslen;
+using ::wcsspn;
+using ::wcstok;
+using ::wmemcmp;
+using ::wmemcpy;
+using ::wmemmove;
+using ::wmemset;
+using ::wcsftime;
+using ::btowc;
+using ::wctob;
+using ::mbsinit;
+using ::mbrlen;
+using ::mbrtowc;
+using ::wcrtomb;
+using ::mbsrtowcs;
+using ::wcsrtombs;
+
+#ifndef _LIBCUDACXX_HAS_NO_STDIN
+using ::getwchar;
+using ::vwscanf;
+using ::wscanf;
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_STDOUT
+using ::putwchar;
+using ::vwprintf;
+using ::wprintf;
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CWCHAR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype
new file mode 100644
index 000000000000..ce8a7933c296
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/cwctype
@@ -0,0 +1,86 @@
+// -*- C++ -*-
+//===--------------------------- cwctype ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CWCTYPE
+#define _LIBCUDACXX_CWCTYPE
+
+/*
+    cwctype synopsis
+
+Macros:
+
+    WEOF
+
+namespace std
+{
+
+Types:
+
+    wint_t
+    wctrans_t
+    wctype_t
+
+int iswalnum(wint_t wc);
+int iswalpha(wint_t wc);
+int iswblank(wint_t wc);  // C99
+int iswcntrl(wint_t wc);
+int iswdigit(wint_t wc);
+int iswgraph(wint_t wc);
+int iswlower(wint_t wc);
+int iswprint(wint_t wc);
+int iswpunct(wint_t wc);
+int iswspace(wint_t wc);
+int iswupper(wint_t wc);
+int iswxdigit(wint_t wc);
+int iswctype(wint_t wc, wctype_t desc);
+wctype_t wctype(const char* property);
+wint_t towlower(wint_t wc);
+wint_t towupper(wint_t wc);
+wint_t towctrans(wint_t wc, wctrans_t desc);
+wctrans_t wctrans(const char* property);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <cctype>
+#include <wctype.h>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+using ::wint_t;
+using ::wctrans_t;
+using ::wctype_t;
+using ::iswalnum;
+using ::iswalpha;
+using ::iswblank;
+using ::iswcntrl;
+using ::iswdigit;
+using ::iswgraph;
+using ::iswlower;
+using ::iswprint;
+using ::iswpunct;
+using ::iswspace;
+using ::iswupper;
+using ::iswxdigit;
+using ::iswctype;
+using ::wctype;
+using ::towlower;
+using ::towupper;
+using ::towctrans;
+using ::wctrans;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_CWCTYPE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/deque b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/deque
new file mode 100644
index 000000000000..457823ca9f52
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/deque
@@ -0,0 +1,3039 @@
+// -*- C++ -*-
+//===---------------------------- deque -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_DEQUE
+#define _LIBCUDACXX_DEQUE
+
+/*
+    deque synopsis
+
+namespace std
+{
+
+template <class T, class Allocator = allocator<T> >
+class deque
+{
+public:
+    // types:
+    typedef T value_type;
+    typedef Allocator allocator_type;
+
+    typedef typename allocator_type::reference       reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+
+    typedef typename allocator_type::pointer         pointer;
+    typedef typename allocator_type::const_pointer   const_pointer;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+
+    // construct/copy/destroy:
+    deque() noexcept(is_nothrow_default_constructible<allocator_type>::value);
+    explicit deque(const allocator_type& a);
+    explicit deque(size_type n);
+    explicit deque(size_type n, const allocator_type& a); // C++14
+    deque(size_type n, const value_type& v);
+    deque(size_type n, const value_type& v, const allocator_type& a);
+    template <class InputIterator>
+        deque(InputIterator f, InputIterator l);
+    template <class InputIterator>
+        deque(InputIterator f, InputIterator l, const allocator_type& a);
+    deque(const deque& c);
+    deque(deque&& c)
+        noexcept(is_nothrow_move_constructible<allocator_type>::value);
+    deque(initializer_list<value_type> il, const Allocator& a = allocator_type());
+    deque(const deque& c, const allocator_type& a);
+    deque(deque&& c, const allocator_type& a);
+    ~deque();
+
+    deque& operator=(const deque& c);
+    deque& operator=(deque&& c)
+        noexcept(
+             allocator_type::propagate_on_container_move_assignment::value &&
+             is_nothrow_move_assignable<allocator_type>::value);
+    deque& operator=(initializer_list<value_type> il);
+
+    template <class InputIterator>
+        void assign(InputIterator f, InputIterator l);
+    void assign(size_type n, const value_type& v);
+    void assign(initializer_list<value_type> il);
+
+    allocator_type get_allocator() const noexcept;
+
+    // iterators:
+
+    iterator       begin() noexcept;
+    const_iterator begin() const noexcept;
+    iterator       end() noexcept;
+    const_iterator end() const noexcept;
+
+    reverse_iterator       rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+    reverse_iterator       rend() noexcept;
+    const_reverse_iterator rend() const noexcept;
+
+    const_iterator         cbegin() const noexcept;
+    const_iterator         cend() const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend() const noexcept;
+
+    // capacity:
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+    void resize(size_type n);
+    void resize(size_type n, const value_type& v);
+    void shrink_to_fit();
+    bool empty() const noexcept;
+
+    // element access:
+    reference operator[](size_type i);
+    const_reference operator[](size_type i) const;
+    reference at(size_type i);
+    const_reference at(size_type i) const;
+    reference front();
+    const_reference front() const;
+    reference back();
+    const_reference back() const;
+
+    // modifiers:
+    void push_front(const value_type& v);
+    void push_front(value_type&& v);
+    void push_back(const value_type& v);
+    void push_back(value_type&& v);
+    template <class... Args> reference emplace_front(Args&&... args);  // reference in C++17
+    template <class... Args> reference emplace_back(Args&&... args);   // reference in C++17
+    template <class... Args> iterator emplace(const_iterator p, Args&&... args);
+    iterator insert(const_iterator p, const value_type& v);
+    iterator insert(const_iterator p, value_type&& v);
+    iterator insert(const_iterator p, size_type n, const value_type& v);
+    template <class InputIterator>
+        iterator insert(const_iterator p, InputIterator f, InputIterator l);
+    iterator insert(const_iterator p, initializer_list<value_type> il);
+    void pop_front();
+    void pop_back();
+    iterator erase(const_iterator p);
+    iterator erase(const_iterator f, const_iterator l);
+    void swap(deque& c)
+        noexcept(allocator_traits<allocator_type>::is_always_equal::value);  // C++17
+    void clear() noexcept;
+};
+
+template <class InputIterator, class Allocator = allocator<typename iterator_traits<InputIterator>::value_type>>
+   deque(InputIterator, InputIterator, Allocator = Allocator())
+   -> deque<typename iterator_traits<InputIterator>::value_type, Allocator>;
+
+template <class T, class Allocator>
+    bool operator==(const deque<T,Allocator>& x, const deque<T,Allocator>& y);
+template <class T, class Allocator>
+    bool operator< (const deque<T,Allocator>& x, const deque<T,Allocator>& y);
+template <class T, class Allocator>
+    bool operator!=(const deque<T,Allocator>& x, const deque<T,Allocator>& y);
+template <class T, class Allocator>
+    bool operator> (const deque<T,Allocator>& x, const deque<T,Allocator>& y);
+template <class T, class Allocator>
+    bool operator>=(const deque<T,Allocator>& x, const deque<T,Allocator>& y);
+template <class T, class Allocator>
+    bool operator<=(const deque<T,Allocator>& x, const deque<T,Allocator>& y);
+
+// specialized algorithms:
+template <class T, class Allocator>
+    void swap(deque<T,Allocator>& x, deque<T,Allocator>& y)
+         noexcept(noexcept(x.swap(y)));
+
+template <class T, class Allocator, class U>
+    void erase(deque<T, Allocator>& c, const U& value);       // C++20
+template <class T, class Allocator, class Predicate>
+    void erase_if(deque<T, Allocator>& c, Predicate pred);    // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__split_buffer>
+#include <type_traits>
+#include <initializer_list>
+#include <iterator>
+#include <algorithm>
+#include <stdexcept>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp, class _Allocator> class __deque_base;
+template <class _Tp, class _Allocator = allocator<_Tp> > class _LIBCUDACXX_TEMPLATE_VIS deque;
+
+template <class _ValueType, class _Pointer, class _Reference, class _MapPointer,
+          class _DiffType, _DiffType _BlockSize>
+class _LIBCUDACXX_TEMPLATE_VIS __deque_iterator;
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy(_RAIter __f,
+     _RAIter __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+     typename enable_if<__is_random_access_iterator<_RAIter>::value>::type* = 0);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+copy(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     _OutputIterator __r);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy_backward(_RAIter __f,
+              _RAIter __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+              typename enable_if<__is_random_access_iterator<_RAIter>::value>::type* = 0);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+copy_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              _OutputIterator __r);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move(_RAIter __f,
+     _RAIter __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+     typename enable_if<__is_random_access_iterator<_RAIter>::value>::type* = 0);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+move(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     _OutputIterator __r);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move_backward(_RAIter __f,
+              _RAIter __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+              typename enable_if<__is_random_access_iterator<_RAIter>::value>::type* = 0);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+move_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              _OutputIterator __r);
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+template <class _ValueType, class _DiffType>
+struct __deque_block_size {
+  static const _DiffType value = sizeof(_ValueType) < 256 ? 4096 / sizeof(_ValueType) : 16;
+};
+
+template <class _ValueType, class _Pointer, class _Reference, class _MapPointer,
+          class _DiffType, _DiffType _BS =
+#ifdef _LIBCUDACXX_ABI_INCOMPLETE_TYPES_IN_DEQUE
+// Keep template parameter to avoid changing all template declarations thoughout
+// this file.
+                               0
+#else
+                               __deque_block_size<_ValueType, _DiffType>::value
+#endif
+          >
+class _LIBCUDACXX_TEMPLATE_VIS __deque_iterator
+{
+    typedef _MapPointer __map_iterator;
+public:
+    typedef _Pointer  pointer;
+    typedef _DiffType difference_type;
+private:
+    __map_iterator __m_iter_;
+    pointer        __ptr_;
+
+    static const difference_type __block_size;
+public:
+    typedef _ValueType                  value_type;
+    typedef random_access_iterator_tag  iterator_category;
+    typedef _Reference                  reference;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator() _NOEXCEPT
+#if _LIBCUDACXX_STD_VER > 11
+     : __m_iter_(nullptr), __ptr_(nullptr)
+#endif
+     {}
+
+    template <class _Pp, class _Rp, class _MP>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __deque_iterator(const __deque_iterator<value_type, _Pp, _Rp, _MP, difference_type, _BS>& __it,
+                typename enable_if<is_convertible<_Pp, pointer>::value>::type* = 0) _NOEXCEPT
+        : __m_iter_(__it.__m_iter_), __ptr_(__it.__ptr_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator*() const {return *__ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY pointer operator->() const {return __ptr_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator& operator++()
+    {
+        if (++__ptr_ - *__m_iter_ == __block_size)
+        {
+            ++__m_iter_;
+            __ptr_ = *__m_iter_;
+        }
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator operator++(int)
+    {
+        __deque_iterator __tmp = *this;
+        ++(*this);
+        return __tmp;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator& operator--()
+    {
+        if (__ptr_ == *__m_iter_)
+        {
+            --__m_iter_;
+            __ptr_ = *__m_iter_ + __block_size;
+        }
+        --__ptr_;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator operator--(int)
+    {
+        __deque_iterator __tmp = *this;
+        --(*this);
+        return __tmp;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator& operator+=(difference_type __n)
+    {
+        if (__n != 0)
+        {
+            __n += __ptr_ - *__m_iter_;
+            if (__n > 0)
+            {
+                __m_iter_ += __n / __block_size;
+                __ptr_ = *__m_iter_ + __n % __block_size;
+            }
+            else // (__n < 0)
+            {
+                difference_type __z = __block_size - 1 - __n;
+                __m_iter_ -= __z / __block_size;
+                __ptr_ = *__m_iter_ + (__block_size - 1 - __z % __block_size);
+            }
+        }
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator& operator-=(difference_type __n)
+    {
+        return *this += -__n;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator operator+(difference_type __n) const
+    {
+        __deque_iterator __t(*this);
+        __t += __n;
+        return __t;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator operator-(difference_type __n) const
+    {
+        __deque_iterator __t(*this);
+        __t -= __n;
+        return __t;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend __deque_iterator operator+(difference_type __n, const __deque_iterator& __it)
+        {return __it + __n;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    friend difference_type operator-(const __deque_iterator& __x, const __deque_iterator& __y)
+    {
+        if (__x != __y)
+            return (__x.__m_iter_ - __y.__m_iter_) * __block_size
+                 + (__x.__ptr_ - *__x.__m_iter_)
+                 - (__y.__ptr_ - *__y.__m_iter_);
+        return 0;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator[](difference_type __n) const
+        {return *(*this + __n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend
+        bool operator==(const __deque_iterator& __x, const __deque_iterator& __y)
+        {return __x.__ptr_ == __y.__ptr_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend
+        bool operator!=(const __deque_iterator& __x, const __deque_iterator& __y)
+        {return !(__x == __y);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend
+        bool operator<(const __deque_iterator& __x, const __deque_iterator& __y)
+        {return __x.__m_iter_ < __y.__m_iter_ ||
+               (__x.__m_iter_ == __y.__m_iter_ && __x.__ptr_ < __y.__ptr_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend
+        bool operator>(const __deque_iterator& __x, const __deque_iterator& __y)
+        {return __y < __x;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend
+        bool operator<=(const __deque_iterator& __x, const __deque_iterator& __y)
+        {return !(__y < __x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY friend
+        bool operator>=(const __deque_iterator& __x, const __deque_iterator& __y)
+        {return !(__x < __y);}
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY __deque_iterator(__map_iterator __m, pointer __p) _NOEXCEPT
+        : __m_iter_(__m), __ptr_(__p) {}
+
+    template <class _Tp, class _Ap> friend class __deque_base;
+    template <class _Tp, class _Ap> friend class _LIBCUDACXX_TEMPLATE_VIS deque;
+    template <class _Vp, class _Pp, class _Rp, class _MP, class _Dp, _Dp>
+        friend class _LIBCUDACXX_TEMPLATE_VIS __deque_iterator;
+
+    template <class _RAIter,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    copy(_RAIter __f,
+         _RAIter __l,
+         __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+         typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _OutputIterator>
+    friend
+    _OutputIterator
+    copy(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+         __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+         _OutputIterator __r);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    copy(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+         __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+         __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+    template <class _RAIter,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    copy_backward(_RAIter __f,
+                  _RAIter __l,
+                  __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+                  typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _OutputIterator>
+    friend
+    _OutputIterator
+    copy_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+                  __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+                  _OutputIterator __r);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    copy_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+                  __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+                  __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+    template <class _RAIter,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    move(_RAIter __f,
+         _RAIter __l,
+         __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+         typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _OutputIterator>
+    friend
+    _OutputIterator
+    move(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+         __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+         _OutputIterator __r);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    move(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+         __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+         __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+
+    template <class _RAIter,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    move_backward(_RAIter __f,
+                  _RAIter __l,
+                  __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+                  typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _OutputIterator>
+    friend
+    _OutputIterator
+    move_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+                  __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+                  _OutputIterator __r);
+
+    template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+              class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+    friend
+    __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+    move_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+                  __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+                  __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r);
+};
+
+template <class _ValueType, class _Pointer, class _Reference, class _MapPointer,
+          class _DiffType, _DiffType _BlockSize>
+const _DiffType __deque_iterator<_ValueType, _Pointer, _Reference, _MapPointer,
+                                 _DiffType, _BlockSize>::__block_size =
+    __deque_block_size<_ValueType, _DiffType>::value;
+
+// copy
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy(_RAIter __f,
+     _RAIter __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+     typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*)
+{
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::difference_type difference_type;
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::pointer pointer;
+    const difference_type __block_size = __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::__block_size;
+    while (__f != __l)
+    {
+        pointer __rb = __r.__ptr_;
+        pointer __re = *__r.__m_iter_ + __block_size;
+        difference_type __bs = __re - __rb;
+        difference_type __n = __l - __f;
+        _RAIter __m = __l;
+        if (__n > __bs)
+        {
+            __n = __bs;
+            __m = __f + __n;
+        }
+        _CUDA_VSTD::copy(__f, __m, __rb);
+        __f = __m;
+        __r += __n;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+copy(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     _OutputIterator __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    const difference_type __block_size = __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::__block_size;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        pointer __fb = __f.__ptr_;
+        pointer __fe = *__f.__m_iter_ + __block_size;
+        difference_type __bs = __fe - __fb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __fe = __fb + __bs;
+        }
+        __r = _CUDA_VSTD::copy(__fb, __fe, __r);
+        __n -= __bs;
+        __f += __bs;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    const difference_type __block_size = __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::__block_size;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        pointer __fb = __f.__ptr_;
+        pointer __fe = *__f.__m_iter_ + __block_size;
+        difference_type __bs = __fe - __fb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __fe = __fb + __bs;
+        }
+        __r = _CUDA_VSTD::copy(__fb, __fe, __r);
+        __n -= __bs;
+        __f += __bs;
+    }
+    return __r;
+}
+
+// copy_backward
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy_backward(_RAIter __f,
+              _RAIter __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+              typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*)
+{
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::difference_type difference_type;
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::pointer pointer;
+    while (__f != __l)
+    {
+        __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __rp = _CUDA_VSTD::prev(__r);
+        pointer __rb = *__rp.__m_iter_;
+        pointer __re = __rp.__ptr_ + 1;
+        difference_type __bs = __re - __rb;
+        difference_type __n = __l - __f;
+        _RAIter __m = __f;
+        if (__n > __bs)
+        {
+            __n = __bs;
+            __m = __l - __n;
+        }
+        _CUDA_VSTD::copy_backward(__m, __l, __re);
+        __l = __m;
+        __r -= __n;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+copy_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              _OutputIterator __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        --__l;
+        pointer __lb = *__l.__m_iter_;
+        pointer __le = __l.__ptr_ + 1;
+        difference_type __bs = __le - __lb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __lb = __le - __bs;
+        }
+        __r = _CUDA_VSTD::copy_backward(__lb, __le, __r);
+        __n -= __bs;
+        __l -= __bs - 1;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+copy_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        --__l;
+        pointer __lb = *__l.__m_iter_;
+        pointer __le = __l.__ptr_ + 1;
+        difference_type __bs = __le - __lb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __lb = __le - __bs;
+        }
+        __r = _CUDA_VSTD::copy_backward(__lb, __le, __r);
+        __n -= __bs;
+        __l -= __bs - 1;
+    }
+    return __r;
+}
+
+// move
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move(_RAIter __f,
+     _RAIter __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+     typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*)
+{
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::difference_type difference_type;
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::pointer pointer;
+    const difference_type __block_size = __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::__block_size;
+    while (__f != __l)
+    {
+        pointer __rb = __r.__ptr_;
+        pointer __re = *__r.__m_iter_ + __block_size;
+        difference_type __bs = __re - __rb;
+        difference_type __n = __l - __f;
+        _RAIter __m = __l;
+        if (__n > __bs)
+        {
+            __n = __bs;
+            __m = __f + __n;
+        }
+        _CUDA_VSTD::move(__f, __m, __rb);
+        __f = __m;
+        __r += __n;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+move(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     _OutputIterator __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    const difference_type __block_size = __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::__block_size;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        pointer __fb = __f.__ptr_;
+        pointer __fe = *__f.__m_iter_ + __block_size;
+        difference_type __bs = __fe - __fb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __fe = __fb + __bs;
+        }
+        __r = _CUDA_VSTD::move(__fb, __fe, __r);
+        __n -= __bs;
+        __f += __bs;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+     __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+     __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    const difference_type __block_size = __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::__block_size;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        pointer __fb = __f.__ptr_;
+        pointer __fe = *__f.__m_iter_ + __block_size;
+        difference_type __bs = __fe - __fb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __fe = __fb + __bs;
+        }
+        __r = _CUDA_VSTD::move(__fb, __fe, __r);
+        __n -= __bs;
+        __f += __bs;
+    }
+    return __r;
+}
+
+// move_backward
+
+template <class _RAIter,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move_backward(_RAIter __f,
+              _RAIter __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r,
+              typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*)
+{
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::difference_type difference_type;
+    typedef typename __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>::pointer pointer;
+    while (__f != __l)
+    {
+        __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __rp = _CUDA_VSTD::prev(__r);
+        pointer __rb = *__rp.__m_iter_;
+        pointer __re = __rp.__ptr_ + 1;
+        difference_type __bs = __re - __rb;
+        difference_type __n = __l - __f;
+        _RAIter __m = __f;
+        if (__n > __bs)
+        {
+            __n = __bs;
+            __m = __l - __n;
+        }
+        _CUDA_VSTD::move_backward(__m, __l, __re);
+        __l = __m;
+        __r -= __n;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _OutputIterator>
+_OutputIterator
+move_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              _OutputIterator __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        --__l;
+        pointer __lb = *__l.__m_iter_;
+        pointer __le = __l.__ptr_ + 1;
+        difference_type __bs = __le - __lb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __lb = __le - __bs;
+        }
+        __r = _CUDA_VSTD::move_backward(__lb, __le, __r);
+        __n -= __bs;
+        __l -= __bs - 1;
+    }
+    return __r;
+}
+
+template <class _V1, class _P1, class _R1, class _M1, class _D1, _D1 _B1,
+          class _V2, class _P2, class _R2, class _M2, class _D2, _D2 _B2>
+__deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2>
+move_backward(__deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __f,
+              __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1> __l,
+              __deque_iterator<_V2, _P2, _R2, _M2, _D2, _B2> __r)
+{
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::difference_type difference_type;
+    typedef typename __deque_iterator<_V1, _P1, _R1, _M1, _D1, _B1>::pointer pointer;
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        --__l;
+        pointer __lb = *__l.__m_iter_;
+        pointer __le = __l.__ptr_ + 1;
+        difference_type __bs = __le - __lb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __lb = __le - __bs;
+        }
+        __r = _CUDA_VSTD::move_backward(__lb, __le, __r);
+        __n -= __bs;
+        __l -= __bs - 1;
+    }
+    return __r;
+}
+
+template <bool>
+class __deque_base_common
+{
+protected:
+    _LIBCUDACXX_NORETURN void __throw_length_error() const;
+    _LIBCUDACXX_NORETURN void __throw_out_of_range() const;
+};
+
+template <bool __b>
+void
+__deque_base_common<__b>::__throw_length_error() const
+{
+    _CUDA_VSTD::__throw_length_error("deque");
+}
+
+template <bool __b>
+void
+__deque_base_common<__b>::__throw_out_of_range() const
+{
+    _CUDA_VSTD::__throw_out_of_range("deque");
+}
+
+template <class _Tp, class _Allocator>
+class __deque_base
+    : protected __deque_base_common<true>
+{
+    __deque_base(const __deque_base& __c);
+    __deque_base& operator=(const __deque_base& __c);
+public:
+    typedef _Allocator                               allocator_type;
+    typedef allocator_traits<allocator_type>         __alloc_traits;
+    typedef typename __alloc_traits::size_type       size_type;
+
+    typedef _Tp                                      value_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+    typedef typename __alloc_traits::difference_type difference_type;
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+
+    static const difference_type __block_size;
+
+    typedef typename __rebind_alloc_helper<__alloc_traits, pointer>::type __pointer_allocator;
+    typedef allocator_traits<__pointer_allocator>        __map_traits;
+    typedef typename __map_traits::pointer               __map_pointer;
+    typedef typename __rebind_alloc_helper<__alloc_traits, const_pointer>::type __const_pointer_allocator;
+    typedef typename allocator_traits<__const_pointer_allocator>::const_pointer __map_const_pointer;
+    typedef __split_buffer<pointer, __pointer_allocator> __map;
+
+    typedef __deque_iterator<value_type, pointer, reference, __map_pointer,
+                             difference_type>    iterator;
+    typedef __deque_iterator<value_type, const_pointer, const_reference, __map_const_pointer,
+                             difference_type>    const_iterator;
+
+    struct __deque_block_range {
+      explicit __deque_block_range(pointer __b, pointer __e) _NOEXCEPT : __begin_(__b), __end_(__e) {}
+      const pointer __begin_;
+      const pointer __end_;
+    };
+
+    struct __deque_range {
+      iterator __pos_;
+      const iterator __end_;
+
+      __deque_range(iterator __pos, iterator __e) _NOEXCEPT
+        : __pos_(__pos), __end_(__e) {}
+
+      explicit operator bool() const _NOEXCEPT {
+        return __pos_ != __end_;
+      }
+
+      __deque_range begin() const {
+        return *this;
+      }
+
+      __deque_range end() const {
+        return __deque_range(__end_, __end_);
+      }
+      __deque_block_range operator*() const _NOEXCEPT {
+         if (__pos_.__m_iter_ == __end_.__m_iter_) {
+          return __deque_block_range(__pos_.__ptr_, __end_.__ptr_);
+        }
+        return __deque_block_range(__pos_.__ptr_, *__pos_.__m_iter_ + __block_size);
+      }
+
+      __deque_range& operator++() _NOEXCEPT {
+        if (__pos_.__m_iter_ == __end_.__m_iter_) {
+          __pos_ = __end_;
+        } else {
+          ++__pos_.__m_iter_;
+          __pos_.__ptr_ = *__pos_.__m_iter_;
+        }
+        return *this;
+      }
+
+
+      friend bool operator==(__deque_range const& __lhs, __deque_range const& __rhs) {
+        return __lhs.__pos_ == __rhs.__pos_;
+      }
+      friend bool operator!=(__deque_range const& __lhs, __deque_range const& __rhs) {
+        return !(__lhs == __rhs);
+      }
+    };
+
+
+
+    struct _ConstructTransaction {
+      _ConstructTransaction(__deque_base* __db, __deque_block_range& __r)
+        : __pos_(__r.__begin_), __end_(__r.__end_), __begin_(__r.__begin_), __base_(__db) {}
+
+
+      ~_ConstructTransaction() {
+        __base_->size() += (__pos_ - __begin_);
+      }
+
+      pointer __pos_;
+      const pointer __end_;
+    private:
+      const pointer __begin_;
+      __deque_base * const __base_;
+    };
+
+protected:
+    __map __map_;
+    size_type __start_;
+    __compressed_pair<size_type, allocator_type> __size_;
+
+    iterator       begin() _NOEXCEPT;
+    const_iterator begin() const _NOEXCEPT;
+    iterator       end() _NOEXCEPT;
+    const_iterator end() const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY size_type&            size()          {return __size_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const size_type& size() const _NOEXCEPT {return __size_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY allocator_type&       __alloc()       {return __size_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const allocator_type& __alloc() const _NOEXCEPT {return __size_.second();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __deque_base()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __deque_base(const allocator_type& __a);
+public:
+    ~__deque_base();
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __deque_base(__deque_base&& __c)
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value);
+    __deque_base(__deque_base&& __c, const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    void swap(__deque_base& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value);
+#endif
+protected:
+    void clear() _NOEXCEPT;
+
+    bool __invariants() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign(__deque_base& __c)
+        _NOEXCEPT_(__alloc_traits::propagate_on_container_move_assignment::value &&
+                   is_nothrow_move_assignable<allocator_type>::value)
+    {
+        __map_ = _CUDA_VSTD::move(__c.__map_);
+        __start_ = __c.__start_;
+        size() = __c.size();
+        __move_assign_alloc(__c);
+        __c.__start_ = __c.size() = 0;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__deque_base& __c)
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_move_assignment::value ||
+                   is_nothrow_move_assignable<allocator_type>::value)
+        {__move_assign_alloc(__c, integral_constant<bool,
+                      __alloc_traits::propagate_on_container_move_assignment::value>());}
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__deque_base& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+        {
+            __alloc() = _CUDA_VSTD::move(__c.__alloc());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__deque_base&, false_type) _NOEXCEPT
+        {}
+};
+
+template <class _Tp, class _Allocator>
+const typename __deque_base<_Tp, _Allocator>::difference_type
+    __deque_base<_Tp, _Allocator>::__block_size =
+        __deque_block_size<value_type, difference_type>::value;
+
+template <class _Tp, class _Allocator>
+bool
+__deque_base<_Tp, _Allocator>::__invariants() const
+{
+    if (!__map_.__invariants())
+        return false;
+    if (__map_.size() >= size_type(-1) / __block_size)
+        return false;
+    for (typename __map::const_iterator __i = __map_.begin(), __e = __map_.end();
+         __i != __e; ++__i)
+        if (*__i == nullptr)
+            return false;
+    if (__map_.size() != 0)
+    {
+        if (size() >= __map_.size() * __block_size)
+            return false;
+        if (__start_ >= __map_.size() * __block_size - size())
+            return false;
+    }
+    else
+    {
+        if (size() != 0)
+            return false;
+        if (__start_ != 0)
+            return false;
+    }
+    return true;
+}
+
+template <class _Tp, class _Allocator>
+typename __deque_base<_Tp, _Allocator>::iterator
+__deque_base<_Tp, _Allocator>::begin() _NOEXCEPT
+{
+    __map_pointer __mp = __map_.begin() + __start_ / __block_size;
+    return iterator(__mp, __map_.empty() ? 0 : *__mp + __start_ % __block_size);
+}
+
+template <class _Tp, class _Allocator>
+typename __deque_base<_Tp, _Allocator>::const_iterator
+__deque_base<_Tp, _Allocator>::begin() const _NOEXCEPT
+{
+    __map_const_pointer __mp = static_cast<__map_const_pointer>(__map_.begin() + __start_ / __block_size);
+    return const_iterator(__mp, __map_.empty() ? 0 : *__mp + __start_ % __block_size);
+}
+
+template <class _Tp, class _Allocator>
+typename __deque_base<_Tp, _Allocator>::iterator
+__deque_base<_Tp, _Allocator>::end() _NOEXCEPT
+{
+    size_type __p = size() + __start_;
+    __map_pointer __mp = __map_.begin() + __p / __block_size;
+    return iterator(__mp, __map_.empty() ? 0 : *__mp + __p % __block_size);
+}
+
+template <class _Tp, class _Allocator>
+typename __deque_base<_Tp, _Allocator>::const_iterator
+__deque_base<_Tp, _Allocator>::end() const _NOEXCEPT
+{
+    size_type __p = size() + __start_;
+    __map_const_pointer __mp = static_cast<__map_const_pointer>(__map_.begin() + __p / __block_size);
+    return const_iterator(__mp, __map_.empty() ? 0 : *__mp + __p % __block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+__deque_base<_Tp, _Allocator>::__deque_base()
+    _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+    : __start_(0), __size_(0) {}
+
+template <class _Tp, class _Allocator>
+inline
+__deque_base<_Tp, _Allocator>::__deque_base(const allocator_type& __a)
+    : __map_(__pointer_allocator(__a)), __start_(0), __size_(0, __a) {}
+
+template <class _Tp, class _Allocator>
+__deque_base<_Tp, _Allocator>::~__deque_base()
+{
+    clear();
+    typename __map::iterator __i = __map_.begin();
+    typename __map::iterator __e = __map_.end();
+    for (; __i != __e; ++__i)
+        __alloc_traits::deallocate(__alloc(), *__i, __block_size);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+__deque_base<_Tp, _Allocator>::__deque_base(__deque_base&& __c)
+    _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value)
+    : __map_(_CUDA_VSTD::move(__c.__map_)),
+      __start_(_CUDA_VSTD::move(__c.__start_)),
+      __size_(_CUDA_VSTD::move(__c.__size_))
+{
+    __c.__start_ = 0;
+    __c.size() = 0;
+}
+
+template <class _Tp, class _Allocator>
+__deque_base<_Tp, _Allocator>::__deque_base(__deque_base&& __c, const allocator_type& __a)
+    : __map_(_CUDA_VSTD::move(__c.__map_), __pointer_allocator(__a)),
+      __start_(_CUDA_VSTD::move(__c.__start_)),
+      __size_(_CUDA_VSTD::move(__c.size()), __a)
+{
+    if (__a == __c.__alloc())
+    {
+        __c.__start_ = 0;
+        __c.size() = 0;
+    }
+    else
+    {
+        __map_.clear();
+        __start_ = 0;
+        size() = 0;
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+void
+__deque_base<_Tp, _Allocator>::swap(__deque_base& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value)
+#endif
+{
+    __map_.swap(__c.__map_);
+    _CUDA_VSTD::swap(__start_, __c.__start_);
+    _CUDA_VSTD::swap(size(), __c.size());
+    __swap_allocator(__alloc(), __c.__alloc());
+}
+
+template <class _Tp, class _Allocator>
+void
+__deque_base<_Tp, _Allocator>::clear() _NOEXCEPT
+{
+    allocator_type& __a = __alloc();
+    for (iterator __i = begin(), __e = end(); __i != __e; ++__i)
+        __alloc_traits::destroy(__a, _CUDA_VSTD::addressof(*__i));
+    size() = 0;
+    while (__map_.size() > 2)
+    {
+        __alloc_traits::deallocate(__a, __map_.front(), __block_size);
+        __map_.pop_front();
+    }
+    switch (__map_.size())
+    {
+    case 1:
+        __start_ = __block_size / 2;
+        break;
+    case 2:
+        __start_ = __block_size;
+        break;
+    }
+}
+
+template <class _Tp, class _Allocator /*= allocator<_Tp>*/>
+class _LIBCUDACXX_TEMPLATE_VIS deque
+    : private __deque_base<_Tp, _Allocator>
+{
+public:
+    // types:
+
+    typedef _Tp value_type;
+    typedef _Allocator allocator_type;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+    typedef __deque_base<value_type, allocator_type> __base;
+
+    typedef typename __base::__alloc_traits        __alloc_traits;
+    typedef typename __base::reference             reference;
+    typedef typename __base::const_reference       const_reference;
+    typedef typename __base::iterator              iterator;
+    typedef typename __base::const_iterator        const_iterator;
+    typedef typename __base::size_type             size_type;
+    typedef typename __base::difference_type       difference_type;
+
+    typedef typename __base::pointer               pointer;
+    typedef typename __base::const_pointer         const_pointer;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>       reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator> const_reverse_iterator;
+
+    using typename __base::__deque_range;
+    using typename __base::__deque_block_range;
+    using typename __base::_ConstructTransaction;
+
+    // construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    deque()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+        {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit deque(const allocator_type& __a) : __base(__a) {}
+    explicit deque(size_type __n);
+#if _LIBCUDACXX_STD_VER > 11
+    explicit deque(size_type __n, const _Allocator& __a);
+#endif
+    deque(size_type __n, const value_type& __v);
+    deque(size_type __n, const value_type& __v, const allocator_type& __a);
+    template <class _InputIter>
+        deque(_InputIter __f, _InputIter __l,
+              typename enable_if<__is_input_iterator<_InputIter>::value>::type* = 0);
+    template <class _InputIter>
+        deque(_InputIter __f, _InputIter __l, const allocator_type& __a,
+              typename enable_if<__is_input_iterator<_InputIter>::value>::type* = 0);
+    deque(const deque& __c);
+    deque(const deque& __c, const allocator_type& __a);
+
+    deque& operator=(const deque& __c);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    deque(initializer_list<value_type> __il);
+    deque(initializer_list<value_type> __il, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    deque& operator=(initializer_list<value_type> __il) {assign(__il); return *this;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    deque(deque&& __c) _NOEXCEPT_(is_nothrow_move_constructible<__base>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    deque(deque&& __c, const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    deque& operator=(deque&& __c)
+        _NOEXCEPT_(__alloc_traits::propagate_on_container_move_assignment::value &&
+                   is_nothrow_move_assignable<allocator_type>::value);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(initializer_list<value_type> __il) {assign(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template <class _InputIter>
+        void assign(_InputIter __f, _InputIter __l,
+                    typename enable_if<__is_input_iterator<_InputIter>::value &&
+                                      !__is_random_access_iterator<_InputIter>::value>::type* = 0);
+    template <class _RAIter>
+        void assign(_RAIter __f, _RAIter __l,
+                    typename enable_if<__is_random_access_iterator<_RAIter>::value>::type* = 0);
+    void assign(size_type __n, const value_type& __v);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT;
+
+    // iterators:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT       {return __base::begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return __base::begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT         {return __base::end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()   const _NOEXCEPT {return __base::end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator       rbegin() _NOEXCEPT
+        {return       reverse_iterator(__base::end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(__base::end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator       rend() _NOEXCEPT
+        {return       reverse_iterator(__base::begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend()   const _NOEXCEPT
+        {return const_reverse_iterator(__base::begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator         cbegin()  const _NOEXCEPT
+        {return __base::begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator         cend()    const _NOEXCEPT
+        {return __base::end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT
+        {return const_reverse_iterator(__base::end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend()   const _NOEXCEPT
+        {return const_reverse_iterator(__base::begin());}
+
+    // capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT {return __base::size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT
+        {return std::min<size_type>(
+            __alloc_traits::max_size(__base::__alloc()),
+            numeric_limits<difference_type>::max());}
+    void resize(size_type __n);
+    void resize(size_type __n, const value_type& __v);
+    void shrink_to_fit() _NOEXCEPT;
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return __base::size() == 0;}
+
+    // element access:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator[](size_type __i) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference operator[](size_type __i) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference at(size_type __i);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference at(size_type __i) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference front() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference front() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference back() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference back() const _NOEXCEPT;
+
+    // 23.2.2.3 modifiers:
+    void push_front(const value_type& __v);
+    void push_back(const value_type& __v);
+#ifndef _LIBCUDACXX_CXX03_LANG
+#if _LIBCUDACXX_STD_VER > 14
+    template <class... _Args> reference emplace_front(_Args&&... __args);
+    template <class... _Args> reference emplace_back (_Args&&... __args);
+#else
+    template <class... _Args> void      emplace_front(_Args&&... __args);
+    template <class... _Args> void      emplace_back (_Args&&... __args);
+#endif
+    template <class... _Args> iterator emplace(const_iterator __p, _Args&&... __args);
+
+    void push_front(value_type&& __v);
+    void push_back(value_type&& __v);
+    iterator insert(const_iterator __p, value_type&& __v);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, initializer_list<value_type> __il)
+        {return insert(__p, __il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    iterator insert(const_iterator __p, const value_type& __v);
+    iterator insert(const_iterator __p, size_type __n, const value_type& __v);
+    template <class _InputIter>
+        iterator insert(const_iterator __p, _InputIter __f, _InputIter __l,
+                         typename enable_if<__is_input_iterator<_InputIter>::value
+                                         &&!__is_forward_iterator<_InputIter>::value>::type* = 0);
+    template <class _ForwardIterator>
+        iterator insert(const_iterator __p, _ForwardIterator __f, _ForwardIterator __l,
+                               typename enable_if<__is_forward_iterator<_ForwardIterator>::value
+                                         &&!__is_bidirectional_iterator<_ForwardIterator>::value>::type* = 0);
+    template <class _BiIter>
+        iterator insert(const_iterator __p, _BiIter __f, _BiIter __l,
+                         typename enable_if<__is_bidirectional_iterator<_BiIter>::value>::type* = 0);
+
+    void pop_front();
+    void pop_back();
+    iterator erase(const_iterator __p);
+    iterator erase(const_iterator __f, const_iterator __l);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(deque& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                   __is_nothrow_swappable<allocator_type>::value);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __invariants() const {return __base::__invariants();}
+
+    typedef typename __base::__map_const_pointer __map_const_pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type __recommend_blocks(size_type __n)
+    {
+        return __n / __base::__block_size + (__n % __base::__block_size != 0);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __capacity() const
+    {
+        return __base::__map_.size() == 0 ? 0 : __base::__map_.size() * __base::__block_size - 1;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __block_count() const
+    {
+        return __base::__map_.size();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __front_spare() const
+    {
+        return __base::__start_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __front_spare_blocks() const {
+      return __front_spare() / __base::__block_size;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __back_spare() const
+    {
+        return __capacity() - (__base::__start_ + __base::size());
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __back_spare_blocks() const {
+      return __back_spare() / __base::__block_size;
+    }
+
+ private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __maybe_remove_front_spare(bool __keep_one = true) {
+      if (__front_spare_blocks() >= 2 || (!__keep_one && __front_spare_blocks())) {
+        __alloc_traits::deallocate(__base::__alloc(), __base::__map_.front(),
+                                   __base::__block_size);
+        __base::__map_.pop_front();
+        __base::__start_ -= __base::__block_size;
+        return true;
+      }
+      return false;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __maybe_remove_back_spare(bool __keep_one = true) {
+      if (__back_spare_blocks() >= 2 || (!__keep_one && __back_spare_blocks())) {
+        __alloc_traits::deallocate(__base::__alloc(), __base::__map_.back(),
+                                   __base::__block_size);
+        __base::__map_.pop_back();
+        return true;
+      }
+      return false;
+    }
+
+    template <class _InpIter>
+        void __append(_InpIter __f, _InpIter __l,
+                 typename enable_if<__is_input_iterator<_InpIter>::value &&
+                                   !__is_forward_iterator<_InpIter>::value>::type* = 0);
+    template <class _ForIter>
+        void __append(_ForIter __f, _ForIter __l,
+                      typename enable_if<__is_forward_iterator<_ForIter>::value>::type* = 0);
+    void __append(size_type __n);
+    void __append(size_type __n, const value_type& __v);
+    void __erase_to_end(const_iterator __f);
+    void __add_front_capacity();
+    void __add_front_capacity(size_type __n);
+    void __add_back_capacity();
+    void __add_back_capacity(size_type __n);
+    iterator __move_and_check(iterator __f, iterator __l, iterator __r,
+                              const_pointer& __vt);
+    iterator __move_backward_and_check(iterator __f, iterator __l, iterator __r,
+                                       const_pointer& __vt);
+    void __move_construct_and_check(iterator __f, iterator __l,
+                                    iterator __r, const_pointer& __vt);
+    void __move_construct_backward_and_check(iterator __f, iterator __l,
+                                             iterator __r, const_pointer& __vt);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const deque& __c)
+        {__copy_assign_alloc(__c, integral_constant<bool,
+                      __alloc_traits::propagate_on_container_copy_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const deque& __c, true_type)
+        {
+            if (__base::__alloc() != __c.__alloc())
+            {
+                clear();
+                shrink_to_fit();
+            }
+            __base::__alloc() = __c.__alloc();
+            __base::__map_.__alloc() = __c.__map_.__alloc();
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const deque&, false_type)
+        {}
+
+    void __move_assign(deque& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value);
+    void __move_assign(deque& __c, false_type);
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Alloc = typename std::allocator<typename iterator_traits<_InputIterator>::value_type>,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+deque(_InputIterator, _InputIterator)
+  -> deque<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+
+template<class _InputIterator,
+         class _Alloc,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+deque(_InputIterator, _InputIterator, _Alloc)
+  -> deque<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+#endif
+
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(size_type __n)
+{
+    if (__n > 0)
+        __append(__n);
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(size_type __n, const _Allocator& __a)
+    : __base(__a)
+{
+    if (__n > 0)
+        __append(__n);
+}
+#endif
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(size_type __n, const value_type& __v)
+{
+    if (__n > 0)
+        __append(__n, __v);
+}
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(size_type __n, const value_type& __v, const allocator_type& __a)
+    : __base(__a)
+{
+    if (__n > 0)
+        __append(__n, __v);
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIter>
+deque<_Tp, _Allocator>::deque(_InputIter __f, _InputIter __l,
+              typename enable_if<__is_input_iterator<_InputIter>::value>::type*)
+{
+    __append(__f, __l);
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIter>
+deque<_Tp, _Allocator>::deque(_InputIter __f, _InputIter __l, const allocator_type& __a,
+              typename enable_if<__is_input_iterator<_InputIter>::value>::type*)
+    : __base(__a)
+{
+    __append(__f, __l);
+}
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(const deque& __c)
+    : __base(__alloc_traits::select_on_container_copy_construction(__c.__alloc()))
+{
+    __append(__c.begin(), __c.end());
+}
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(const deque& __c, const allocator_type& __a)
+    : __base(__a)
+{
+    __append(__c.begin(), __c.end());
+}
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>&
+deque<_Tp, _Allocator>::operator=(const deque& __c)
+{
+    if (this != &__c)
+    {
+        __copy_assign_alloc(__c);
+        assign(__c.begin(), __c.end());
+    }
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(initializer_list<value_type> __il)
+{
+    __append(__il.begin(), __il.end());
+}
+
+template <class _Tp, class _Allocator>
+deque<_Tp, _Allocator>::deque(initializer_list<value_type> __il, const allocator_type& __a)
+    : __base(__a)
+{
+    __append(__il.begin(), __il.end());
+}
+
+template <class _Tp, class _Allocator>
+inline
+deque<_Tp, _Allocator>::deque(deque&& __c)
+    _NOEXCEPT_(is_nothrow_move_constructible<__base>::value)
+    : __base(_CUDA_VSTD::move(__c))
+{
+}
+
+template <class _Tp, class _Allocator>
+inline
+deque<_Tp, _Allocator>::deque(deque&& __c, const allocator_type& __a)
+    : __base(_CUDA_VSTD::move(__c), __a)
+{
+    if (__a != __c.__alloc())
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__c.begin()), _Ip(__c.end()));
+    }
+}
+
+template <class _Tp, class _Allocator>
+inline
+deque<_Tp, _Allocator>&
+deque<_Tp, _Allocator>::operator=(deque&& __c)
+        _NOEXCEPT_(__alloc_traits::propagate_on_container_move_assignment::value &&
+                   is_nothrow_move_assignable<allocator_type>::value)
+{
+    __move_assign(__c, integral_constant<bool,
+          __alloc_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__move_assign(deque& __c, false_type)
+{
+    if (__base::__alloc() != __c.__alloc())
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__c.begin()), _Ip(__c.end()));
+    }
+    else
+        __move_assign(__c, true_type());
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__move_assign(deque& __c, true_type)
+    _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+{
+    clear();
+    shrink_to_fit();
+    __base::__move_assign(__c);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+template <class _InputIter>
+void
+deque<_Tp, _Allocator>::assign(_InputIter __f, _InputIter __l,
+                               typename enable_if<__is_input_iterator<_InputIter>::value &&
+                                                 !__is_random_access_iterator<_InputIter>::value>::type*)
+{
+    iterator __i = __base::begin();
+    iterator __e = __base::end();
+    for (; __f != __l && __i != __e; ++__f, (void) ++__i)
+        *__i = *__f;
+    if (__f != __l)
+        __append(__f, __l);
+    else
+        __erase_to_end(__i);
+}
+
+template <class _Tp, class _Allocator>
+template <class _RAIter>
+void
+deque<_Tp, _Allocator>::assign(_RAIter __f, _RAIter __l,
+                               typename enable_if<__is_random_access_iterator<_RAIter>::value>::type*)
+{
+    if (static_cast<size_type>(__l - __f) > __base::size())
+    {
+        _RAIter __m = __f + __base::size();
+        _CUDA_VSTD::copy(__f, __m, __base::begin());
+        __append(__m, __l);
+    }
+    else
+        __erase_to_end(_CUDA_VSTD::copy(__f, __l, __base::begin()));
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::assign(size_type __n, const value_type& __v)
+{
+    if (__n > __base::size())
+    {
+        _CUDA_VSTD::fill_n(__base::begin(), __base::size(), __v);
+        __n -= __base::size();
+        __append(__n, __v);
+    }
+    else
+        __erase_to_end(_CUDA_VSTD::fill_n(__base::begin(), __n, __v));
+}
+
+template <class _Tp, class _Allocator>
+inline
+_Allocator
+deque<_Tp, _Allocator>::get_allocator() const _NOEXCEPT
+{
+    return __base::__alloc();
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::resize(size_type __n)
+{
+    if (__n > __base::size())
+        __append(__n - __base::size());
+    else if (__n < __base::size())
+        __erase_to_end(__base::begin() + __n);
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::resize(size_type __n, const value_type& __v)
+{
+    if (__n > __base::size())
+        __append(__n - __base::size(), __v);
+    else if (__n < __base::size())
+        __erase_to_end(__base::begin() + __n);
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::shrink_to_fit() _NOEXCEPT
+{
+    allocator_type& __a = __base::__alloc();
+    if (empty())
+    {
+        while (__base::__map_.size() > 0)
+        {
+            __alloc_traits::deallocate(__a, __base::__map_.back(), __base::__block_size);
+            __base::__map_.pop_back();
+        }
+        __base::__start_ = 0;
+    }
+    else
+    {
+      __maybe_remove_front_spare(/*__keep_one=*/false);
+      __maybe_remove_back_spare(/*__keep_one=*/false);
+    }
+    __base::__map_.shrink_to_fit();
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::reference
+deque<_Tp, _Allocator>::operator[](size_type __i) _NOEXCEPT
+{
+    size_type __p = __base::__start_ + __i;
+    return *(*(__base::__map_.begin() + __p / __base::__block_size) + __p % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::const_reference
+deque<_Tp, _Allocator>::operator[](size_type __i) const _NOEXCEPT
+{
+    size_type __p = __base::__start_ + __i;
+    return *(*(__base::__map_.begin() + __p / __base::__block_size) + __p % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::reference
+deque<_Tp, _Allocator>::at(size_type __i)
+{
+    if (__i >= __base::size())
+        __base::__throw_out_of_range();
+    size_type __p = __base::__start_ + __i;
+    return *(*(__base::__map_.begin() + __p / __base::__block_size) + __p % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::const_reference
+deque<_Tp, _Allocator>::at(size_type __i) const
+{
+    if (__i >= __base::size())
+        __base::__throw_out_of_range();
+    size_type __p = __base::__start_ + __i;
+    return *(*(__base::__map_.begin() + __p / __base::__block_size) + __p % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::reference
+deque<_Tp, _Allocator>::front() _NOEXCEPT
+{
+    return *(*(__base::__map_.begin() + __base::__start_ / __base::__block_size)
+                                      + __base::__start_ % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::const_reference
+deque<_Tp, _Allocator>::front() const _NOEXCEPT
+{
+    return *(*(__base::__map_.begin() + __base::__start_ / __base::__block_size)
+                                      + __base::__start_ % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::reference
+deque<_Tp, _Allocator>::back() _NOEXCEPT
+{
+    size_type __p = __base::size() + __base::__start_ - 1;
+    return *(*(__base::__map_.begin() + __p / __base::__block_size) + __p % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+inline
+typename deque<_Tp, _Allocator>::const_reference
+deque<_Tp, _Allocator>::back() const _NOEXCEPT
+{
+    size_type __p = __base::size() + __base::__start_ - 1;
+    return *(*(__base::__map_.begin() + __p / __base::__block_size) + __p % __base::__block_size);
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::push_back(const value_type& __v)
+{
+    allocator_type& __a = __base::__alloc();
+    if (__back_spare() == 0)
+        __add_back_capacity();
+    // __back_spare() >= 1
+    __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__base::end()), __v);
+    ++__base::size();
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::push_front(const value_type& __v)
+{
+    allocator_type& __a = __base::__alloc();
+    if (__front_spare() == 0)
+        __add_front_capacity();
+    // __front_spare() >= 1
+    __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__base::begin()), __v);
+    --__base::__start_;
+    ++__base::size();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::push_back(value_type&& __v)
+{
+    allocator_type& __a = __base::__alloc();
+    if (__back_spare() == 0)
+        __add_back_capacity();
+    // __back_spare() >= 1
+    __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__base::end()), _CUDA_VSTD::move(__v));
+    ++__base::size();
+}
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+typename deque<_Tp, _Allocator>::reference
+#else
+void
+#endif
+deque<_Tp, _Allocator>::emplace_back(_Args&&... __args)
+{
+    allocator_type& __a = __base::__alloc();
+    if (__back_spare() == 0)
+        __add_back_capacity();
+    // __back_spare() >= 1
+    __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__base::end()),
+                              _CUDA_VSTD::forward<_Args>(__args)...);
+    ++__base::size();
+#if _LIBCUDACXX_STD_VER > 14
+    return *--__base::end();
+#endif
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::push_front(value_type&& __v)
+{
+    allocator_type& __a = __base::__alloc();
+    if (__front_spare() == 0)
+        __add_front_capacity();
+    // __front_spare() >= 1
+    __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__base::begin()), _CUDA_VSTD::move(__v));
+    --__base::__start_;
+    ++__base::size();
+}
+
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+typename deque<_Tp, _Allocator>::reference
+#else
+void
+#endif
+deque<_Tp, _Allocator>::emplace_front(_Args&&... __args)
+{
+    allocator_type& __a = __base::__alloc();
+    if (__front_spare() == 0)
+        __add_front_capacity();
+    // __front_spare() >= 1
+    __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__base::begin()), _CUDA_VSTD::forward<_Args>(__args)...);
+    --__base::__start_;
+    ++__base::size();
+#if _LIBCUDACXX_STD_VER > 14
+    return *__base::begin();
+#endif
+}
+
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::insert(const_iterator __p, value_type&& __v)
+{
+    size_type __pos = __p - __base::begin();
+    size_type __to_end = __base::size() - __pos;
+    allocator_type& __a = __base::__alloc();
+    if (__pos < __to_end)
+    {   // insert by shifting things backward
+        if (__front_spare() == 0)
+            __add_front_capacity();
+        // __front_spare() >= 1
+        if (__pos == 0)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__base::begin()), _CUDA_VSTD::move(__v));
+            --__base::__start_;
+            ++__base::size();
+        }
+        else
+        {
+            iterator __b = __base::begin();
+            iterator __bm1 = _CUDA_VSTD::prev(__b);
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__bm1), _CUDA_VSTD::move(*__b));
+            --__base::__start_;
+            ++__base::size();
+            if (__pos > 1)
+                __b = _CUDA_VSTD::move(_CUDA_VSTD::next(__b), __b + __pos, __b);
+            *__b = _CUDA_VSTD::move(__v);
+        }
+    }
+    else
+    {   // insert by shifting things forward
+        if (__back_spare() == 0)
+            __add_back_capacity();
+        // __back_capacity >= 1
+        size_type __de = __base::size() - __pos;
+        if (__de == 0)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__base::end()), _CUDA_VSTD::move(__v));
+            ++__base::size();
+        }
+        else
+        {
+            iterator __e = __base::end();
+            iterator __em1 = _CUDA_VSTD::prev(__e);
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__e), _CUDA_VSTD::move(*__em1));
+            ++__base::size();
+            if (__de > 1)
+                __e = _CUDA_VSTD::move_backward(__e - __de, __em1, __e);
+            *--__e = _CUDA_VSTD::move(__v);
+        }
+    }
+    return __base::begin() + __pos;
+}
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::emplace(const_iterator __p, _Args&&... __args)
+{
+    size_type __pos = __p - __base::begin();
+    size_type __to_end = __base::size() - __pos;
+    allocator_type& __a = __base::__alloc();
+    if (__pos < __to_end)
+    {   // insert by shifting things backward
+        if (__front_spare() == 0)
+            __add_front_capacity();
+        // __front_spare() >= 1
+        if (__pos == 0)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__base::begin()), _CUDA_VSTD::forward<_Args>(__args)...);
+            --__base::__start_;
+            ++__base::size();
+        }
+        else
+        {
+            __temp_value<value_type, _Allocator> __tmp(this->__alloc(), _CUDA_VSTD::forward<_Args>(__args)...);
+            iterator __b = __base::begin();
+            iterator __bm1 = _CUDA_VSTD::prev(__b);
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__bm1), _CUDA_VSTD::move(*__b));
+            --__base::__start_;
+            ++__base::size();
+            if (__pos > 1)
+                __b = _CUDA_VSTD::move(_CUDA_VSTD::next(__b), __b + __pos, __b);
+            *__b = _CUDA_VSTD::move(__tmp.get());
+        }
+    }
+    else
+    {   // insert by shifting things forward
+        if (__back_spare() == 0)
+            __add_back_capacity();
+        // __back_capacity >= 1
+        size_type __de = __base::size() - __pos;
+        if (__de == 0)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__base::end()), _CUDA_VSTD::forward<_Args>(__args)...);
+            ++__base::size();
+        }
+        else
+        {
+            __temp_value<value_type, _Allocator> __tmp(this->__alloc(), _CUDA_VSTD::forward<_Args>(__args)...);
+            iterator __e = __base::end();
+            iterator __em1 = _CUDA_VSTD::prev(__e);
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__e), _CUDA_VSTD::move(*__em1));
+            ++__base::size();
+            if (__de > 1)
+                __e = _CUDA_VSTD::move_backward(__e - __de, __em1, __e);
+            *--__e = _CUDA_VSTD::move(__tmp.get());
+        }
+    }
+    return __base::begin() + __pos;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::insert(const_iterator __p, const value_type& __v)
+{
+    size_type __pos = __p - __base::begin();
+    size_type __to_end = __base::size() - __pos;
+    allocator_type& __a = __base::__alloc();
+    if (__pos < __to_end)
+    {   // insert by shifting things backward
+        if (__front_spare() == 0)
+            __add_front_capacity();
+        // __front_spare() >= 1
+        if (__pos == 0)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__base::begin()), __v);
+            --__base::__start_;
+            ++__base::size();
+        }
+        else
+        {
+            const_pointer __vt = pointer_traits<const_pointer>::pointer_to(__v);
+            iterator __b = __base::begin();
+            iterator __bm1 = _CUDA_VSTD::prev(__b);
+            if (__vt == pointer_traits<const_pointer>::pointer_to(*__b))
+                __vt = pointer_traits<const_pointer>::pointer_to(*__bm1);
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__bm1), _CUDA_VSTD::move(*__b));
+            --__base::__start_;
+            ++__base::size();
+            if (__pos > 1)
+                __b = __move_and_check(_CUDA_VSTD::next(__b), __b + __pos, __b, __vt);
+            *__b = *__vt;
+        }
+    }
+    else
+    {   // insert by shifting things forward
+        if (__back_spare() == 0)
+            __add_back_capacity();
+        // __back_capacity >= 1
+        size_type __de = __base::size() - __pos;
+        if (__de == 0)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__base::end()), __v);
+            ++__base::size();
+        }
+        else
+        {
+            const_pointer __vt = pointer_traits<const_pointer>::pointer_to(__v);
+            iterator __e = __base::end();
+            iterator __em1 = _CUDA_VSTD::prev(__e);
+            if (__vt == pointer_traits<const_pointer>::pointer_to(*__em1))
+                __vt = pointer_traits<const_pointer>::pointer_to(*__e);
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__e), _CUDA_VSTD::move(*__em1));
+            ++__base::size();
+            if (__de > 1)
+                __e = __move_backward_and_check(__e - __de, __em1, __e, __vt);
+            *--__e = *__vt;
+        }
+    }
+    return __base::begin() + __pos;
+}
+
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::insert(const_iterator __p, size_type __n, const value_type& __v)
+{
+    size_type __pos = __p - __base::begin();
+    size_type __to_end = __base::size() - __pos;
+    allocator_type& __a = __base::__alloc();
+    if (__pos < __to_end)
+    {   // insert by shifting things backward
+        if (__n > __front_spare())
+            __add_front_capacity(__n - __front_spare());
+        // __n <= __front_spare()
+        iterator __old_begin = __base::begin();
+        iterator __i = __old_begin;
+        if (__n > __pos)
+        {
+            for (size_type __m = __n - __pos; __m; --__m, --__base::__start_, ++__base::size())
+                __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__i), __v);
+            __n = __pos;
+        }
+        if (__n > 0)
+        {
+            const_pointer __vt = pointer_traits<const_pointer>::pointer_to(__v);
+            iterator __obn = __old_begin + __n;
+            __move_construct_backward_and_check(__old_begin, __obn, __i, __vt);
+            if (__n < __pos)
+                __old_begin = __move_and_check(__obn, __old_begin + __pos, __old_begin, __vt);
+            _CUDA_VSTD::fill_n(__old_begin, __n, *__vt);
+        }
+    }
+    else
+    {   // insert by shifting things forward
+        size_type __back_capacity = __back_spare();
+        if (__n > __back_capacity)
+            __add_back_capacity(__n - __back_capacity);
+        // __n <= __back_capacity
+        iterator __old_end = __base::end();
+        iterator __i = __old_end;
+        size_type __de = __base::size() - __pos;
+        if (__n > __de)
+        {
+            for (size_type __m = __n - __de; __m; --__m, ++__i, ++__base::size())
+                __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__i), __v);
+            __n = __de;
+        }
+        if (__n > 0)
+        {
+            const_pointer __vt = pointer_traits<const_pointer>::pointer_to(__v);
+            iterator __oen = __old_end - __n;
+            __move_construct_and_check(__oen, __old_end, __i, __vt);
+            if (__n < __de)
+                __old_end = __move_backward_and_check(__old_end - __de, __oen, __old_end, __vt);
+            _CUDA_VSTD::fill_n(__old_end - __n, __n, *__vt);
+        }
+    }
+    return __base::begin() + __pos;
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIter>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::insert(const_iterator __p, _InputIter __f, _InputIter __l,
+                               typename enable_if<__is_input_iterator<_InputIter>::value
+                                               &&!__is_forward_iterator<_InputIter>::value>::type*)
+{
+    __split_buffer<value_type, allocator_type&> __buf(__base::__alloc());
+    __buf.__construct_at_end(__f, __l);
+    typedef typename __split_buffer<value_type, allocator_type&>::iterator __bi;
+    return insert(__p, move_iterator<__bi>(__buf.begin()), move_iterator<__bi>(__buf.end()));
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::insert(const_iterator __p, _ForwardIterator __f, _ForwardIterator __l,
+                               typename enable_if<__is_forward_iterator<_ForwardIterator>::value
+                                               &&!__is_bidirectional_iterator<_ForwardIterator>::value>::type*)
+{
+    size_type __n = _CUDA_VSTD::distance(__f, __l);
+    __split_buffer<value_type, allocator_type&> __buf(__n, 0, __base::__alloc());
+    __buf.__construct_at_end(__f, __l);
+    typedef typename __split_buffer<value_type, allocator_type&>::iterator __fwd;
+    return insert(__p, move_iterator<__fwd>(__buf.begin()), move_iterator<__fwd>(__buf.end()));
+}
+
+template <class _Tp, class _Allocator>
+template <class _BiIter>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::insert(const_iterator __p, _BiIter __f, _BiIter __l,
+                               typename enable_if<__is_bidirectional_iterator<_BiIter>::value>::type*)
+{
+    size_type __n = _CUDA_VSTD::distance(__f, __l);
+    size_type __pos = __p - __base::begin();
+    size_type __to_end = __base::size() - __pos;
+    allocator_type& __a = __base::__alloc();
+    if (__pos < __to_end)
+    {   // insert by shifting things backward
+        if (__n > __front_spare())
+            __add_front_capacity(__n - __front_spare());
+        // __n <= __front_spare()
+        iterator __old_begin = __base::begin();
+        iterator __i = __old_begin;
+        _BiIter __m = __f;
+        if (__n > __pos)
+        {
+            __m = __pos < __n / 2 ? _CUDA_VSTD::prev(__l, __pos) : _CUDA_VSTD::next(__f, __n - __pos);
+            for (_BiIter __j = __m; __j != __f; --__base::__start_, ++__base::size())
+                __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__i), *--__j);
+            __n = __pos;
+        }
+        if (__n > 0)
+        {
+            iterator __obn = __old_begin + __n;
+            for (iterator __j = __obn; __j != __old_begin;)
+            {
+                __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__i), _CUDA_VSTD::move(*--__j));
+                --__base::__start_;
+                ++__base::size();
+            }
+            if (__n < __pos)
+                __old_begin = _CUDA_VSTD::move(__obn, __old_begin + __pos, __old_begin);
+            _CUDA_VSTD::copy(__m, __l, __old_begin);
+        }
+    }
+    else
+    {   // insert by shifting things forward
+        size_type __back_capacity = __back_spare();
+        if (__n > __back_capacity)
+            __add_back_capacity(__n - __back_capacity);
+        // __n <= __back_capacity
+        iterator __old_end = __base::end();
+        iterator __i = __old_end;
+        _BiIter __m = __l;
+        size_type __de = __base::size() - __pos;
+        if (__n > __de)
+        {
+            __m = __de < __n / 2 ? _CUDA_VSTD::next(__f, __de) : _CUDA_VSTD::prev(__l, __n - __de);
+            for (_BiIter __j = __m; __j != __l; ++__i, (void) ++__j, ++__base::size())
+                __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__i), *__j);
+            __n = __de;
+        }
+        if (__n > 0)
+        {
+            iterator __oen = __old_end - __n;
+            for (iterator __j = __oen; __j != __old_end; ++__i, ++__j, ++__base::size())
+                __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__i), _CUDA_VSTD::move(*__j));
+            if (__n < __de)
+                __old_end = _CUDA_VSTD::move_backward(__old_end - __de, __oen, __old_end);
+            _CUDA_VSTD::copy_backward(__f, __m, __old_end);
+        }
+    }
+    return __base::begin() + __pos;
+}
+
+template <class _Tp, class _Allocator>
+template <class _InpIter>
+void
+deque<_Tp, _Allocator>::__append(_InpIter __f, _InpIter __l,
+                                 typename enable_if<__is_input_iterator<_InpIter>::value &&
+                                                   !__is_forward_iterator<_InpIter>::value>::type*)
+{
+    for (; __f != __l; ++__f)
+#ifdef _LIBCUDACXX_CXX03_LANG
+        push_back(*__f);
+#else
+        emplace_back(*__f);
+#endif
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForIter>
+void
+deque<_Tp, _Allocator>::__append(_ForIter __f, _ForIter __l,
+                                 typename enable_if<__is_forward_iterator<_ForIter>::value>::type*)
+{
+    size_type __n = _CUDA_VSTD::distance(__f, __l);
+    allocator_type& __a = __base::__alloc();
+    size_type __back_capacity = __back_spare();
+    if (__n > __back_capacity)
+        __add_back_capacity(__n - __back_capacity);
+    // __n <= __back_capacity
+    for (__deque_block_range __br : __deque_range(__base::end(), __base::end() + __n)) {
+      _ConstructTransaction __tx(this, __br);
+      for (; __tx.__pos_ != __tx.__end_; ++__tx.__pos_, (void)++__f) {
+        __alloc_traits::construct(__a, std::__to_raw_pointer(__tx.__pos_), *__f);
+      }
+    }
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__append(size_type __n)
+{
+    allocator_type& __a = __base::__alloc();
+    size_type __back_capacity = __back_spare();
+    if (__n > __back_capacity)
+        __add_back_capacity(__n - __back_capacity);
+    // __n <= __back_capacity
+    for (__deque_block_range __br : __deque_range(__base::end(), __base::end() + __n)) {
+      _ConstructTransaction __tx(this, __br);
+      for (; __tx.__pos_ != __tx.__end_; ++__tx.__pos_) {
+        __alloc_traits::construct(__a, std::__to_raw_pointer(__tx.__pos_));
+      }
+    }
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__append(size_type __n, const value_type& __v)
+{
+    allocator_type& __a = __base::__alloc();
+    size_type __back_capacity = __back_spare();
+    if (__n > __back_capacity)
+        __add_back_capacity(__n - __back_capacity);
+    // __n <= __back_capacity
+    for (__deque_block_range __br : __deque_range(__base::end(), __base::end() + __n)) {
+      _ConstructTransaction __tx(this, __br);
+      for (; __tx.__pos_ != __tx.__end_; ++__tx.__pos_) {
+        __alloc_traits::construct(__a, std::__to_raw_pointer(__tx.__pos_), __v);
+      }
+    }
+
+}
+
+// Create front capacity for one block of elements.
+// Strong guarantee.  Either do it or don't touch anything.
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__add_front_capacity()
+{
+    allocator_type& __a = __base::__alloc();
+    if (__back_spare() >= __base::__block_size)
+    {
+        __base::__start_ += __base::__block_size;
+        pointer __pt = __base::__map_.back();
+        __base::__map_.pop_back();
+        __base::__map_.push_front(__pt);
+    }
+    // Else if __base::__map_.size() < __base::__map_.capacity() then we need to allocate 1 buffer
+    else if (__base::__map_.size() < __base::__map_.capacity())
+    {   // we can put the new buffer into the map, but don't shift things around
+        // until all buffers are allocated.  If we throw, we don't need to fix
+        // anything up (any added buffers are undetectible)
+        if (__base::__map_.__front_spare() > 0)
+            __base::__map_.push_front(__alloc_traits::allocate(__a, __base::__block_size));
+        else
+        {
+            __base::__map_.push_back(__alloc_traits::allocate(__a, __base::__block_size));
+            // Done allocating, reorder capacity
+            pointer __pt = __base::__map_.back();
+            __base::__map_.pop_back();
+            __base::__map_.push_front(__pt);
+        }
+        __base::__start_ = __base::__map_.size() == 1 ?
+                               __base::__block_size / 2 :
+                               __base::__start_ + __base::__block_size;
+    }
+    // Else need to allocate 1 buffer, *and* we need to reallocate __map_.
+    else
+    {
+        __split_buffer<pointer, typename __base::__pointer_allocator&>
+            __buf(max<size_type>(2 * __base::__map_.capacity(), 1),
+                  0, __base::__map_.__alloc());
+
+        typedef __allocator_destructor<_Allocator> _Dp;
+        unique_ptr<pointer, _Dp> __hold(
+            __alloc_traits::allocate(__a, __base::__block_size),
+                _Dp(__a, __base::__block_size));
+        __buf.push_back(__hold.get());
+        __hold.release();
+
+        for (typename __base::__map_pointer __i = __base::__map_.begin();
+                __i != __base::__map_.end(); ++__i)
+            __buf.push_back(*__i);
+        _CUDA_VSTD::swap(__base::__map_.__first_, __buf.__first_);
+        _CUDA_VSTD::swap(__base::__map_.__begin_, __buf.__begin_);
+        _CUDA_VSTD::swap(__base::__map_.__end_, __buf.__end_);
+        _CUDA_VSTD::swap(__base::__map_.__end_cap(), __buf.__end_cap());
+        __base::__start_ = __base::__map_.size() == 1 ?
+                               __base::__block_size / 2 :
+                               __base::__start_ + __base::__block_size;
+    }
+}
+
+// Create front capacity for __n elements.
+// Strong guarantee.  Either do it or don't touch anything.
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__add_front_capacity(size_type __n)
+{
+    allocator_type& __a = __base::__alloc();
+    size_type __nb = __recommend_blocks(__n + __base::__map_.empty());
+    // Number of unused blocks at back:
+    size_type __back_capacity = __back_spare() / __base::__block_size;
+    __back_capacity = _CUDA_VSTD::min(__back_capacity, __nb);  // don't take more than you need
+    __nb -= __back_capacity;  // number of blocks need to allocate
+    // If __nb == 0, then we have sufficient capacity.
+    if (__nb == 0)
+    {
+        __base::__start_ += __base::__block_size * __back_capacity;
+        for (; __back_capacity > 0; --__back_capacity)
+        {
+            pointer __pt = __base::__map_.back();
+            __base::__map_.pop_back();
+            __base::__map_.push_front(__pt);
+        }
+    }
+    // Else if __nb <= __map_.capacity() - __map_.size() then we need to allocate __nb buffers
+    else if (__nb <= __base::__map_.capacity() - __base::__map_.size())
+    {   // we can put the new buffers into the map, but don't shift things around
+        // until all buffers are allocated.  If we throw, we don't need to fix
+        // anything up (any added buffers are undetectible)
+        for (; __nb > 0; --__nb, __base::__start_ += __base::__block_size - (__base::__map_.size() == 1))
+        {
+            if (__base::__map_.__front_spare() == 0)
+                break;
+            __base::__map_.push_front(__alloc_traits::allocate(__a, __base::__block_size));
+        }
+        for (; __nb > 0; --__nb, ++__back_capacity)
+            __base::__map_.push_back(__alloc_traits::allocate(__a, __base::__block_size));
+        // Done allocating, reorder capacity
+        __base::__start_ += __back_capacity * __base::__block_size;
+        for (; __back_capacity > 0; --__back_capacity)
+        {
+            pointer __pt = __base::__map_.back();
+            __base::__map_.pop_back();
+            __base::__map_.push_front(__pt);
+        }
+    }
+    // Else need to allocate __nb buffers, *and* we need to reallocate __map_.
+    else
+    {
+        size_type __ds = (__nb + __back_capacity) * __base::__block_size - __base::__map_.empty();
+        __split_buffer<pointer, typename __base::__pointer_allocator&>
+            __buf(max<size_type>(2* __base::__map_.capacity(),
+                                 __nb + __base::__map_.size()),
+                  0, __base::__map_.__alloc());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (; __nb > 0; --__nb)
+                __buf.push_back(__alloc_traits::allocate(__a, __base::__block_size));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            for (typename __base::__map_pointer __i = __buf.begin();
+                    __i != __buf.end(); ++__i)
+                __alloc_traits::deallocate(__a, *__i, __base::__block_size);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        for (; __back_capacity > 0; --__back_capacity)
+        {
+            __buf.push_back(__base::__map_.back());
+            __base::__map_.pop_back();
+        }
+        for (typename __base::__map_pointer __i = __base::__map_.begin();
+                __i != __base::__map_.end(); ++__i)
+            __buf.push_back(*__i);
+        _CUDA_VSTD::swap(__base::__map_.__first_, __buf.__first_);
+        _CUDA_VSTD::swap(__base::__map_.__begin_, __buf.__begin_);
+        _CUDA_VSTD::swap(__base::__map_.__end_, __buf.__end_);
+        _CUDA_VSTD::swap(__base::__map_.__end_cap(), __buf.__end_cap());
+        __base::__start_ += __ds;
+    }
+}
+
+// Create back capacity for one block of elements.
+// Strong guarantee.  Either do it or don't touch anything.
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__add_back_capacity()
+{
+    allocator_type& __a = __base::__alloc();
+    if (__front_spare() >= __base::__block_size)
+    {
+        __base::__start_ -= __base::__block_size;
+        pointer __pt = __base::__map_.front();
+        __base::__map_.pop_front();
+        __base::__map_.push_back(__pt);
+    }
+    // Else if __nb <= __map_.capacity() - __map_.size() then we need to allocate __nb buffers
+    else if (__base::__map_.size() < __base::__map_.capacity())
+    {   // we can put the new buffer into the map, but don't shift things around
+        // until it is allocated.  If we throw, we don't need to fix
+        // anything up (any added buffers are undetectible)
+        if (__base::__map_.__back_spare() != 0)
+            __base::__map_.push_back(__alloc_traits::allocate(__a, __base::__block_size));
+        else
+        {
+            __base::__map_.push_front(__alloc_traits::allocate(__a, __base::__block_size));
+            // Done allocating, reorder capacity
+            pointer __pt = __base::__map_.front();
+            __base::__map_.pop_front();
+            __base::__map_.push_back(__pt);
+        }
+    }
+    // Else need to allocate 1 buffer, *and* we need to reallocate __map_.
+    else
+    {
+        __split_buffer<pointer, typename __base::__pointer_allocator&>
+            __buf(max<size_type>(2* __base::__map_.capacity(), 1),
+                  __base::__map_.size(),
+                  __base::__map_.__alloc());
+
+        typedef __allocator_destructor<_Allocator> _Dp;
+        unique_ptr<pointer, _Dp> __hold(
+            __alloc_traits::allocate(__a, __base::__block_size),
+                _Dp(__a, __base::__block_size));
+        __buf.push_back(__hold.get());
+        __hold.release();
+
+        for (typename __base::__map_pointer __i = __base::__map_.end();
+                __i != __base::__map_.begin();)
+            __buf.push_front(*--__i);
+        _CUDA_VSTD::swap(__base::__map_.__first_, __buf.__first_);
+        _CUDA_VSTD::swap(__base::__map_.__begin_, __buf.__begin_);
+        _CUDA_VSTD::swap(__base::__map_.__end_, __buf.__end_);
+        _CUDA_VSTD::swap(__base::__map_.__end_cap(), __buf.__end_cap());
+    }
+}
+
+// Create back capacity for __n elements.
+// Strong guarantee.  Either do it or don't touch anything.
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__add_back_capacity(size_type __n)
+{
+    allocator_type& __a = __base::__alloc();
+    size_type __nb = __recommend_blocks(__n + __base::__map_.empty());
+    // Number of unused blocks at front:
+    size_type __front_capacity = __front_spare() / __base::__block_size;
+    __front_capacity = _CUDA_VSTD::min(__front_capacity, __nb);  // don't take more than you need
+    __nb -= __front_capacity;  // number of blocks need to allocate
+    // If __nb == 0, then we have sufficient capacity.
+    if (__nb == 0)
+    {
+        __base::__start_ -= __base::__block_size * __front_capacity;
+        for (; __front_capacity > 0; --__front_capacity)
+        {
+            pointer __pt = __base::__map_.front();
+            __base::__map_.pop_front();
+            __base::__map_.push_back(__pt);
+        }
+    }
+    // Else if __nb <= __map_.capacity() - __map_.size() then we need to allocate __nb buffers
+    else if (__nb <= __base::__map_.capacity() - __base::__map_.size())
+    {   // we can put the new buffers into the map, but don't shift things around
+        // until all buffers are allocated.  If we throw, we don't need to fix
+        // anything up (any added buffers are undetectible)
+        for (; __nb > 0; --__nb)
+        {
+            if (__base::__map_.__back_spare() == 0)
+                break;
+            __base::__map_.push_back(__alloc_traits::allocate(__a, __base::__block_size));
+        }
+        for (; __nb > 0; --__nb, ++__front_capacity, __base::__start_ +=
+                                 __base::__block_size - (__base::__map_.size() == 1))
+            __base::__map_.push_front(__alloc_traits::allocate(__a, __base::__block_size));
+        // Done allocating, reorder capacity
+        __base::__start_ -= __base::__block_size * __front_capacity;
+        for (; __front_capacity > 0; --__front_capacity)
+        {
+            pointer __pt = __base::__map_.front();
+            __base::__map_.pop_front();
+            __base::__map_.push_back(__pt);
+        }
+    }
+    // Else need to allocate __nb buffers, *and* we need to reallocate __map_.
+    else
+    {
+        size_type __ds = __front_capacity * __base::__block_size;
+        __split_buffer<pointer, typename __base::__pointer_allocator&>
+            __buf(max<size_type>(2* __base::__map_.capacity(),
+                                 __nb + __base::__map_.size()),
+                  __base::__map_.size() - __front_capacity,
+                  __base::__map_.__alloc());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (; __nb > 0; --__nb)
+                __buf.push_back(__alloc_traits::allocate(__a, __base::__block_size));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            for (typename __base::__map_pointer __i = __buf.begin();
+                    __i != __buf.end(); ++__i)
+                __alloc_traits::deallocate(__a, *__i, __base::__block_size);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        for (; __front_capacity > 0; --__front_capacity)
+        {
+            __buf.push_back(__base::__map_.front());
+            __base::__map_.pop_front();
+        }
+        for (typename __base::__map_pointer __i = __base::__map_.end();
+                __i != __base::__map_.begin();)
+            __buf.push_front(*--__i);
+        _CUDA_VSTD::swap(__base::__map_.__first_, __buf.__first_);
+        _CUDA_VSTD::swap(__base::__map_.__begin_, __buf.__begin_);
+        _CUDA_VSTD::swap(__base::__map_.__end_, __buf.__end_);
+        _CUDA_VSTD::swap(__base::__map_.__end_cap(), __buf.__end_cap());
+        __base::__start_ -= __ds;
+    }
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::pop_front()
+{
+    allocator_type& __a = __base::__alloc();
+    __alloc_traits::destroy(__a, __to_raw_pointer(*(__base::__map_.begin() +
+                                                    __base::__start_ / __base::__block_size) +
+                                                    __base::__start_ % __base::__block_size));
+    --__base::size();
+    ++__base::__start_;
+    __maybe_remove_front_spare();
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::pop_back()
+{
+    _LIBCUDACXX_ASSERT(!empty(), "deque::pop_back called for empty deque");
+    allocator_type& __a = __base::__alloc();
+    size_type __p = __base::size() + __base::__start_ - 1;
+    __alloc_traits::destroy(__a, __to_raw_pointer(*(__base::__map_.begin() +
+                                                    __p / __base::__block_size) +
+                                                    __p % __base::__block_size));
+    --__base::size();
+    __maybe_remove_back_spare();
+}
+
+// move assign [__f, __l) to [__r, __r + (__l-__f)).
+// If __vt points into [__f, __l), then subtract (__f - __r) from __vt.
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::__move_and_check(iterator __f, iterator __l, iterator __r,
+                                         const_pointer& __vt)
+{
+    // as if
+    //   for (; __f != __l; ++__f, ++__r)
+    //       *__r = _CUDA_VSTD::move(*__f);
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        pointer __fb = __f.__ptr_;
+        pointer __fe = *__f.__m_iter_ + __base::__block_size;
+        difference_type __bs = __fe - __fb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __fe = __fb + __bs;
+        }
+        if (__fb <= __vt && __vt < __fe)
+            __vt = (const_iterator(static_cast<__map_const_pointer>(__f.__m_iter_), __vt) -= __f - __r).__ptr_;
+        __r = _CUDA_VSTD::move(__fb, __fe, __r);
+        __n -= __bs;
+        __f += __bs;
+    }
+    return __r;
+}
+
+// move assign [__f, __l) to [__r - (__l-__f), __r) backwards.
+// If __vt points into [__f, __l), then add (__r - __l) to __vt.
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::__move_backward_and_check(iterator __f, iterator __l, iterator __r,
+                                                  const_pointer& __vt)
+{
+    // as if
+    //   while (__f != __l)
+    //       *--__r = _CUDA_VSTD::move(*--__l);
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        --__l;
+        pointer __lb = *__l.__m_iter_;
+        pointer __le = __l.__ptr_ + 1;
+        difference_type __bs = __le - __lb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __lb = __le - __bs;
+        }
+        if (__lb <= __vt && __vt < __le)
+            __vt = (const_iterator(static_cast<__map_const_pointer>(__l.__m_iter_), __vt) += __r - __l - 1).__ptr_;
+        __r = _CUDA_VSTD::move_backward(__lb, __le, __r);
+        __n -= __bs;
+        __l -= __bs - 1;
+    }
+    return __r;
+}
+
+// move construct [__f, __l) to [__r, __r + (__l-__f)).
+// If __vt points into [__f, __l), then add (__r - __f) to __vt.
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__move_construct_and_check(iterator __f, iterator __l,
+                                                   iterator __r, const_pointer& __vt)
+{
+    allocator_type& __a = __base::__alloc();
+    // as if
+    //   for (; __f != __l; ++__r, ++__f, ++__base::size())
+    //       __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__r), _CUDA_VSTD::move(*__f));
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        pointer __fb = __f.__ptr_;
+        pointer __fe = *__f.__m_iter_ + __base::__block_size;
+        difference_type __bs = __fe - __fb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __fe = __fb + __bs;
+        }
+        if (__fb <= __vt && __vt < __fe)
+            __vt = (const_iterator(static_cast<__map_const_pointer>(__f.__m_iter_), __vt) += __r - __f).__ptr_;
+        for (; __fb != __fe; ++__fb, ++__r, ++__base::size())
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*__r), _CUDA_VSTD::move(*__fb));
+        __n -= __bs;
+        __f += __bs;
+    }
+}
+
+// move construct [__f, __l) to [__r - (__l-__f), __r) backwards.
+// If __vt points into [__f, __l), then subtract (__l - __r) from __vt.
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__move_construct_backward_and_check(iterator __f, iterator __l,
+                                                            iterator __r, const_pointer& __vt)
+{
+    allocator_type& __a = __base::__alloc();
+    // as if
+    //   for (iterator __j = __l; __j != __f;)
+    //   {
+    //       __alloc_traitsconstruct(__a, _CUDA_VSTD::addressof(*--__r), _CUDA_VSTD::move(*--__j));
+    //       --__base::__start_;
+    //       ++__base::size();
+    //   }
+    difference_type __n = __l - __f;
+    while (__n > 0)
+    {
+        --__l;
+        pointer __lb = *__l.__m_iter_;
+        pointer __le = __l.__ptr_ + 1;
+        difference_type __bs = __le - __lb;
+        if (__bs > __n)
+        {
+            __bs = __n;
+            __lb = __le - __bs;
+        }
+        if (__lb <= __vt && __vt < __le)
+            __vt = (const_iterator(static_cast<__map_const_pointer>(__l.__m_iter_), __vt) -= __l - __r + 1).__ptr_;
+        while (__le != __lb)
+        {
+            __alloc_traits::construct(__a, _CUDA_VSTD::addressof(*--__r), _CUDA_VSTD::move(*--__le));
+            --__base::__start_;
+            ++__base::size();
+        }
+        __n -= __bs;
+        __l -= __bs - 1;
+    }
+}
+
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::erase(const_iterator __f)
+{
+    iterator __b = __base::begin();
+    difference_type __pos = __f - __b;
+    iterator __p = __b + __pos;
+    allocator_type& __a = __base::__alloc();
+    if (static_cast<size_t>(__pos) <= (__base::size() - 1) / 2)
+    {   // erase from front
+        _CUDA_VSTD::move_backward(__b, __p, _CUDA_VSTD::next(__p));
+        __alloc_traits::destroy(__a, _CUDA_VSTD::addressof(*__b));
+        --__base::size();
+        ++__base::__start_;
+        __maybe_remove_front_spare();
+    }
+    else
+    {   // erase from back
+        iterator __i = _CUDA_VSTD::move(_CUDA_VSTD::next(__p), __base::end(), __p);
+        __alloc_traits::destroy(__a, _CUDA_VSTD::addressof(*__i));
+        --__base::size();
+        __maybe_remove_back_spare();
+    }
+    return __base::begin() + __pos;
+}
+
+template <class _Tp, class _Allocator>
+typename deque<_Tp, _Allocator>::iterator
+deque<_Tp, _Allocator>::erase(const_iterator __f, const_iterator __l)
+{
+    difference_type __n = __l - __f;
+    iterator __b = __base::begin();
+    difference_type __pos = __f - __b;
+    iterator __p = __b + __pos;
+    if (__n > 0)
+    {
+        allocator_type& __a = __base::__alloc();
+        if (static_cast<size_t>(__pos) <= (__base::size() - __n) / 2)
+        {   // erase from front
+            iterator __i = _CUDA_VSTD::move_backward(__b, __p, __p + __n);
+            for (; __b != __i; ++__b)
+                __alloc_traits::destroy(__a, _CUDA_VSTD::addressof(*__b));
+            __base::size() -= __n;
+            __base::__start_ += __n;
+            while (__maybe_remove_front_spare()) {
+            }
+        }
+        else
+        {   // erase from back
+            iterator __i = _CUDA_VSTD::move(__p + __n, __base::end(), __p);
+            for (iterator __e = __base::end(); __i != __e; ++__i)
+                __alloc_traits::destroy(__a, _CUDA_VSTD::addressof(*__i));
+            __base::size() -= __n;
+            while (__maybe_remove_back_spare()) {
+            }
+        }
+    }
+    return __base::begin() + __pos;
+}
+
+template <class _Tp, class _Allocator>
+void
+deque<_Tp, _Allocator>::__erase_to_end(const_iterator __f)
+{
+    iterator __e = __base::end();
+    difference_type __n = __e - __f;
+    if (__n > 0)
+    {
+        allocator_type& __a = __base::__alloc();
+        iterator __b = __base::begin();
+        difference_type __pos = __f - __b;
+        for (iterator __p = __b + __pos; __p != __e; ++__p)
+            __alloc_traits::destroy(__a, _CUDA_VSTD::addressof(*__p));
+        __base::size() -= __n;
+        while (__maybe_remove_back_spare()) {
+        }
+    }
+}
+
+template <class _Tp, class _Allocator>
+inline
+void
+deque<_Tp, _Allocator>::swap(deque& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value)
+#endif
+{
+    __base::swap(__c);
+}
+
+template <class _Tp, class _Allocator>
+inline
+void
+deque<_Tp, _Allocator>::clear() _NOEXCEPT
+{
+    __base::clear();
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const deque<_Tp, _Allocator>& __x, const deque<_Tp, _Allocator>& __y)
+{
+    const typename deque<_Tp, _Allocator>::size_type __sz = __x.size();
+    return __sz == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const deque<_Tp, _Allocator>& __x, const deque<_Tp, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const deque<_Tp, _Allocator>& __x, const deque<_Tp, _Allocator>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const deque<_Tp, _Allocator>& __x, const deque<_Tp, _Allocator>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const deque<_Tp, _Allocator>& __x, const deque<_Tp, _Allocator>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const deque<_Tp, _Allocator>& __x, const deque<_Tp, _Allocator>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(deque<_Tp, _Allocator>& __x, deque<_Tp, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp, class _Allocator, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase(deque<_Tp, _Allocator>& __c, const _Up& __v)
+{ __c.erase(_CUDA_VSTD::remove(__c.begin(), __c.end(), __v), __c.end()); }
+
+template <class _Tp, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(deque<_Tp, _Allocator>& __c, _Predicate __pred)
+{ __c.erase(_CUDA_VSTD::remove_if(__c.begin(), __c.end(), __pred), __c.end()); }
+#endif
+
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_DEQUE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h
new file mode 100644
index 000000000000..b35af2570a02
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/errno.h
@@ -0,0 +1,397 @@
+// -*- C++ -*-
+//===-------------------------- errno.h -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ERRNO_H
+#define _LIBCUDACXX_ERRNO_H
+
+/*
+    errno.h synopsis
+
+Macros:
+
+    EDOM
+    EILSEQ  // C99
+    ERANGE
+    errno
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <errno.h>
+
+#ifdef __cplusplus
+
+#if !defined(EOWNERDEAD) || !defined(ENOTRECOVERABLE)
+
+#ifdef ELAST
+
+static const int __elast1 = ELAST+1;
+static const int __elast2 = ELAST+2;
+
+#else
+
+static const int __elast1 = 104;
+static const int __elast2 = 105;
+
+#endif
+
+#ifdef ENOTRECOVERABLE
+
+#define EOWNERDEAD __elast1
+
+#ifdef ELAST
+#undef ELAST
+#define ELAST EOWNERDEAD
+#endif
+
+#elif defined(EOWNERDEAD)
+
+#define ENOTRECOVERABLE __elast1
+#ifdef ELAST
+#undef ELAST
+#define ELAST ENOTRECOVERABLE
+#endif
+
+#else  // defined(EOWNERDEAD)
+
+#define EOWNERDEAD __elast1
+#define ENOTRECOVERABLE __elast2
+#ifdef ELAST
+#undef ELAST
+#define ELAST ENOTRECOVERABLE
+#endif
+
+#endif  // defined(EOWNERDEAD)
+
+#endif  // !defined(EOWNERDEAD) || !defined(ENOTRECOVERABLE)
+
+//  supply errno values likely to be missing, particularly on Windows
+
+#ifndef EAFNOSUPPORT
+#define EAFNOSUPPORT 9901
+#endif
+
+#ifndef EADDRINUSE
+#define EADDRINUSE 9902
+#endif
+
+#ifndef EADDRNOTAVAIL
+#define EADDRNOTAVAIL 9903
+#endif
+
+#ifndef EISCONN
+#define EISCONN 9904
+#endif
+
+#ifndef EBADMSG
+#define EBADMSG 9905
+#endif
+
+#ifndef ECONNABORTED
+#define ECONNABORTED 9906
+#endif
+
+#ifndef EALREADY
+#define EALREADY 9907
+#endif
+
+#ifndef ECONNREFUSED
+#define ECONNREFUSED 9908
+#endif
+
+#ifndef ECONNRESET
+#define ECONNRESET 9909
+#endif
+
+#ifndef EDESTADDRREQ
+#define EDESTADDRREQ 9910
+#endif
+
+#ifndef EHOSTUNREACH
+#define EHOSTUNREACH 9911
+#endif
+
+#ifndef EIDRM
+#define EIDRM 9912
+#endif
+
+#ifndef EMSGSIZE
+#define EMSGSIZE 9913
+#endif
+
+#ifndef ENETDOWN
+#define ENETDOWN 9914
+#endif
+
+#ifndef ENETRESET
+#define ENETRESET 9915
+#endif
+
+#ifndef ENETUNREACH
+#define ENETUNREACH 9916
+#endif
+
+#ifndef ENOBUFS
+#define ENOBUFS 9917
+#endif
+
+#ifndef ENOLINK
+#define ENOLINK 9918
+#endif
+
+#ifndef ENODATA
+#define ENODATA 9919
+#endif
+
+#ifndef ENOMSG
+#define ENOMSG 9920
+#endif
+
+#ifndef ENOPROTOOPT
+#define ENOPROTOOPT 9921
+#endif
+
+#ifndef ENOSR
+#define ENOSR 9922
+#endif
+
+#ifndef ENOTSOCK
+#define ENOTSOCK 9923
+#endif
+
+#ifndef ENOSTR
+#define ENOSTR 9924
+#endif
+
+#ifndef ENOTCONN
+#define ENOTCONN 9925
+#endif
+
+#ifndef ENOTSUP
+#define ENOTSUP 9926
+#endif
+
+#ifndef ECANCELED
+#define ECANCELED 9927
+#endif
+
+#ifndef EINPROGRESS
+#define EINPROGRESS 9928
+#endif
+
+#ifndef EOPNOTSUPP
+#define EOPNOTSUPP 9929
+#endif
+
+#ifndef EWOULDBLOCK
+#define EWOULDBLOCK 9930
+#endif
+
+#ifndef EOWNERDEAD
+#define EOWNERDEAD  9931
+#endif
+
+#ifndef EPROTO
+#define EPROTO 9932
+#endif
+
+#ifndef EPROTONOSUPPORT
+#define EPROTONOSUPPORT 9933
+#endif
+
+#ifndef ENOTRECOVERABLE
+#define ENOTRECOVERABLE 9934
+#endif
+
+#ifndef ETIME
+#define ETIME 9935
+#endif
+
+#ifndef ETXTBSY
+#define ETXTBSY 9936
+#endif
+
+#ifndef ETIMEDOUT
+#define ETIMEDOUT 9938
+#endif
+
+#ifndef ELOOP
+#define ELOOP 9939
+#endif
+
+#ifndef EOVERFLOW
+#define EOVERFLOW 9940
+#endif
+
+#ifndef EPROTOTYPE
+#define EPROTOTYPE 9941
+#endif
+
+#ifndef ENOSYS
+#define ENOSYS 9942
+#endif
+
+#ifndef EINVAL
+#define EINVAL 9943
+#endif
+
+#ifndef ERANGE
+#define ERANGE 9944
+#endif
+
+#ifndef EILSEQ
+#define EILSEQ 9945
+#endif
+
+//  Windows Mobile doesn't appear to define these:
+
+#ifndef E2BIG
+#define E2BIG 9946
+#endif
+
+#ifndef EDOM
+#define EDOM 9947
+#endif
+
+#ifndef EFAULT
+#define EFAULT 9948
+#endif
+
+#ifndef EBADF
+#define EBADF 9949
+#endif
+
+#ifndef EPIPE
+#define EPIPE 9950
+#endif
+
+#ifndef EXDEV
+#define EXDEV 9951
+#endif
+
+#ifndef EBUSY
+#define EBUSY 9952
+#endif
+
+#ifndef ENOTEMPTY
+#define ENOTEMPTY 9953
+#endif
+
+#ifndef ENOEXEC
+#define ENOEXEC 9954
+#endif
+
+#ifndef EEXIST
+#define EEXIST 9955
+#endif
+
+#ifndef EFBIG
+#define EFBIG 9956
+#endif
+
+#ifndef ENAMETOOLONG
+#define ENAMETOOLONG 9957
+#endif
+
+#ifndef ENOTTY
+#define ENOTTY 9958
+#endif
+
+#ifndef EINTR
+#define EINTR 9959
+#endif
+
+#ifndef ESPIPE
+#define ESPIPE 9960
+#endif
+
+#ifndef EIO
+#define EIO 9961
+#endif
+
+#ifndef EISDIR
+#define EISDIR 9962
+#endif
+
+#ifndef ECHILD
+#define ECHILD 9963
+#endif
+
+#ifndef ENOLCK
+#define ENOLCK 9964
+#endif
+
+#ifndef ENOSPC
+#define ENOSPC 9965
+#endif
+
+#ifndef ENXIO
+#define ENXIO 9966
+#endif
+
+#ifndef ENODEV
+#define ENODEV 9967
+#endif
+
+#ifndef ENOENT
+#define ENOENT 9968
+#endif
+
+#ifndef ESRCH
+#define ESRCH 9969
+#endif
+
+#ifndef ENOTDIR
+#define ENOTDIR 9970
+#endif
+
+#ifndef ENOMEM
+#define ENOMEM 9971
+#endif
+
+#ifndef EPERM
+#define EPERM 9972
+#endif
+
+#ifndef EACCES
+#define EACCES 9973
+#endif
+
+#ifndef EROFS
+#define EROFS 9974
+#endif
+
+#ifndef EDEADLK
+#define EDEADLK 9975
+#endif
+
+#ifndef EAGAIN
+#define EAGAIN 9976
+#endif
+
+#ifndef ENFILE
+#define ENFILE 9977
+#endif
+
+#ifndef EMFILE
+#define EMFILE 9978
+#endif
+
+#ifndef EMLINK
+#define EMLINK 9979
+#endif
+
+#endif // __cplusplus
+
+#endif  // _LIBCUDACXX_ERRNO_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/exception b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/exception
new file mode 100644
index 000000000000..bd317e22e5ac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/exception
@@ -0,0 +1,329 @@
+// -*- C++ -*-
+//===-------------------------- exception ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXCEPTION
+#define _LIBCUDACXX_EXCEPTION
+
+/*
+    exception synopsis
+
+namespace std
+{
+
+class exception
+{
+public:
+    exception() noexcept;
+    exception(const exception&) noexcept;
+    exception& operator=(const exception&) noexcept;
+    virtual ~exception() noexcept;
+    virtual const char* what() const noexcept;
+};
+
+class bad_exception
+    : public exception
+{
+public:
+    bad_exception() noexcept;
+    bad_exception(const bad_exception&) noexcept;
+    bad_exception& operator=(const bad_exception&) noexcept;
+    virtual ~bad_exception() noexcept;
+    virtual const char* what() const noexcept;
+};
+
+typedef void (*unexpected_handler)();
+unexpected_handler set_unexpected(unexpected_handler  f ) noexcept;
+unexpected_handler get_unexpected() noexcept;
+[[noreturn]] void unexpected();
+
+typedef void (*terminate_handler)();
+terminate_handler set_terminate(terminate_handler  f ) noexcept;
+terminate_handler get_terminate() noexcept;
+[[noreturn]] void terminate() noexcept;
+
+bool uncaught_exception()  noexcept;
+int  uncaught_exceptions() noexcept;  // C++17
+
+typedef unspecified exception_ptr;
+
+exception_ptr current_exception() noexcept;
+void rethrow_exception [[noreturn]] (exception_ptr p);
+template<class E> exception_ptr make_exception_ptr(E e) noexcept;
+
+class nested_exception
+{
+public:
+    nested_exception() noexcept;
+    nested_exception(const nested_exception&) noexcept = default;
+    nested_exception& operator=(const nested_exception&) noexcept = default;
+    virtual ~nested_exception() = default;
+
+    // access functions
+    [[noreturn]] void rethrow_nested() const;
+    exception_ptr nested_ptr() const noexcept;
+};
+
+template <class T> [[noreturn]] void throw_with_nested(T&& t);
+template <class E> void rethrow_if_nested(const E& e);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <cstddef>
+#include <cstdlib>
+#include <type_traits>
+#include <version>
+
+#if defined(_LIBCUDACXX_ABI_VCRUNTIME)
+#include <vcruntime_exception.h>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+namespace std  // purposefully not using versioning namespace
+{
+
+#if !defined(_LIBCUDACXX_ABI_VCRUNTIME)
+class _LIBCUDACXX_EXCEPTION_ABI exception
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY exception() _NOEXCEPT {}
+    virtual ~exception() _NOEXCEPT;
+    virtual const char* what() const _NOEXCEPT;
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_exception
+    : public exception
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY bad_exception() _NOEXCEPT {}
+    virtual ~bad_exception() _NOEXCEPT;
+    virtual const char* what() const _NOEXCEPT;
+};
+#endif // !_LIBCUDACXX_ABI_VCRUNTIME
+
+#if _LIBCUDACXX_STD_VER <= 14 \
+    || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_UNEXPECTED_FUNCTIONS) \
+    || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+typedef void (*unexpected_handler)();
+_LIBCUDACXX_FUNC_VIS unexpected_handler set_unexpected(unexpected_handler) _NOEXCEPT;
+_LIBCUDACXX_FUNC_VIS unexpected_handler get_unexpected() _NOEXCEPT;
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS void unexpected();
+#endif
+
+typedef void (*terminate_handler)();
+_LIBCUDACXX_FUNC_VIS terminate_handler set_terminate(terminate_handler) _NOEXCEPT;
+_LIBCUDACXX_FUNC_VIS terminate_handler get_terminate() _NOEXCEPT;
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS void terminate() _NOEXCEPT;
+
+_LIBCUDACXX_FUNC_VIS bool uncaught_exception() _NOEXCEPT;
+_LIBCUDACXX_FUNC_VIS _LIBCUDACXX_AVAILABILITY_UNCAUGHT_EXCEPTIONS int uncaught_exceptions() _NOEXCEPT;
+
+class _LIBCUDACXX_TYPE_VIS exception_ptr;
+
+_LIBCUDACXX_FUNC_VIS exception_ptr current_exception() _NOEXCEPT;
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS void rethrow_exception(exception_ptr);
+
+#ifndef _LIBCUDACXX_ABI_MICROSOFT
+
+class _LIBCUDACXX_TYPE_VIS exception_ptr
+{
+    void* __ptr_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY exception_ptr() _NOEXCEPT : __ptr_() {}
+    _LIBCUDACXX_INLINE_VISIBILITY exception_ptr(nullptr_t) _NOEXCEPT : __ptr_() {}
+
+    exception_ptr(const exception_ptr&) _NOEXCEPT;
+    exception_ptr& operator=(const exception_ptr&) _NOEXCEPT;
+    ~exception_ptr() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT
+    {return __ptr_ != nullptr;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const exception_ptr& __x, const exception_ptr& __y) _NOEXCEPT
+        {return __x.__ptr_ == __y.__ptr_;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const exception_ptr& __x, const exception_ptr& __y) _NOEXCEPT
+        {return !(__x == __y);}
+
+    friend _LIBCUDACXX_FUNC_VIS exception_ptr current_exception() _NOEXCEPT;
+    friend _LIBCUDACXX_FUNC_VIS void rethrow_exception(exception_ptr);
+};
+
+template<class _Ep>
+_LIBCUDACXX_INLINE_VISIBILITY exception_ptr
+make_exception_ptr(_Ep __e) _NOEXCEPT
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+        throw __e;
+    }
+    catch (...)
+    {
+        return current_exception();
+    }
+#else
+    ((void)__e);
+    _CUDA_VSTD::abort();
+#endif
+}
+
+#else // _LIBCUDACXX_ABI_MICROSOFT
+
+class _LIBCUDACXX_TYPE_VIS exception_ptr
+{
+#if defined(__clang__)
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunused-private-field"
+#endif
+    void* __ptr1_;
+    void* __ptr2_;
+#if defined(__clang__)
+#pragma clang diagnostic pop
+#endif
+public:
+    exception_ptr() _NOEXCEPT;
+    exception_ptr(nullptr_t) _NOEXCEPT;
+    exception_ptr(const exception_ptr& __other) _NOEXCEPT;
+    exception_ptr& operator=(const exception_ptr& __other) _NOEXCEPT;
+    exception_ptr& operator=(nullptr_t) _NOEXCEPT;
+    ~exception_ptr() _NOEXCEPT;
+    _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT;
+};
+
+_LIBCUDACXX_FUNC_VIS
+bool operator==(const exception_ptr& __x, const exception_ptr& __y) _NOEXCEPT;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(const exception_ptr& __x, const exception_ptr& __y) _NOEXCEPT
+    {return !(__x == __y);}
+
+_LIBCUDACXX_FUNC_VIS void swap(exception_ptr&, exception_ptr&) _NOEXCEPT;
+
+_LIBCUDACXX_FUNC_VIS exception_ptr __copy_exception_ptr(void *__except, const void* __ptr);
+_LIBCUDACXX_FUNC_VIS exception_ptr current_exception() _NOEXCEPT;
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS void rethrow_exception(exception_ptr p);
+
+// This is a built-in template function which automagically extracts the required
+// information.
+template <class _E> void *__GetExceptionInfo(_E);
+
+template<class _Ep>
+_LIBCUDACXX_INLINE_VISIBILITY exception_ptr
+make_exception_ptr(_Ep __e) _NOEXCEPT
+{
+  return __copy_exception_ptr(_CUDA_VSTD::addressof(__e), __GetExceptionInfo(__e));
+}
+
+#endif // _LIBCUDACXX_ABI_MICROSOFT
+// nested_exception
+
+class _LIBCUDACXX_EXCEPTION_ABI nested_exception
+{
+    exception_ptr __ptr_;
+public:
+    nested_exception() _NOEXCEPT;
+//     nested_exception(const nested_exception&) noexcept = default;
+//     nested_exception& operator=(const nested_exception&) noexcept = default;
+    virtual ~nested_exception() _NOEXCEPT;
+
+    // access functions
+    _LIBCUDACXX_NORETURN void rethrow_nested() const;
+    _LIBCUDACXX_INLINE_VISIBILITY exception_ptr nested_ptr() const _NOEXCEPT {return __ptr_;}
+};
+
+template <class _Tp>
+struct __nested
+    : public _Tp,
+      public nested_exception
+{
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __nested(const _Tp& __t) : _Tp(__t) {}
+};
+
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+template <class _Tp, class _Up, bool>
+struct __throw_with_nested;
+
+template <class _Tp, class _Up>
+struct __throw_with_nested<_Tp, _Up, true> {
+    _LIBCUDACXX_NORETURN static inline _LIBCUDACXX_INLINE_VISIBILITY void
+    __do_throw(_Tp&& __t)
+    {
+        throw __nested<_Up>(_CUDA_VSTD::forward<_Tp>(__t));
+    }
+};
+
+template <class _Tp, class _Up>
+struct __throw_with_nested<_Tp, _Up, false> {
+    _LIBCUDACXX_NORETURN static inline _LIBCUDACXX_INLINE_VISIBILITY void
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __do_throw(_Tp&& __t)
+#else
+    __do_throw (_Tp& __t)
+#endif  // _LIBCUDACXX_CXX03_LANG
+    {
+        throw _CUDA_VSTD::forward<_Tp>(__t);
+    }
+};
+#endif
+
+template <class _Tp>
+_LIBCUDACXX_NORETURN
+void
+throw_with_nested(_Tp&& __t)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    typedef typename decay<_Tp>::type _Up;
+    static_assert( is_copy_constructible<_Up>::value, "type thrown must be CopyConstructible");
+    __throw_with_nested<_Tp, _Up,
+        is_class<_Up>::value &&
+        !is_base_of<nested_exception, _Up>::value &&
+        !__libcpp_is_final<_Up>::value>::
+            __do_throw(_CUDA_VSTD::forward<_Tp>(__t));
+#else
+    ((void)__t);
+    // FIXME: Make this abort
+#endif
+}
+
+template <class _From, class _To>
+struct __can_dynamic_cast : public _LIBCUDACXX_BOOL_CONSTANT(
+              is_polymorphic<_From>::value &&
+                 (!is_base_of<_To, _From>::value ||
+                   is_convertible<const _From*, const _To*>::value)) {};
+
+template <class _Ep>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+rethrow_if_nested(const _Ep& __e,
+                  typename enable_if< __can_dynamic_cast<_Ep, nested_exception>::value>::type* = 0)
+{
+    const nested_exception* __nep = dynamic_cast<const nested_exception*>(_CUDA_VSTD::addressof(__e));
+    if (__nep)
+        __nep->rethrow_nested();
+}
+
+template <class _Ep>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+rethrow_if_nested(const _Ep&,
+                  typename enable_if<!__can_dynamic_cast<_Ep, nested_exception>::value>::type* = 0)
+{
+}
+
+}  // std
+
+#endif  // _LIBCUDACXX_EXCEPTION
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/execution b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/execution
new file mode 100644
index 000000000000..298bb5befe98
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/execution
@@ -0,0 +1,19 @@
+// -*- C++ -*-
+//===------------------------- execution ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXECUTION
+#define _LIBCUDACXX_EXECUTION
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_HAS_PARALLEL_ALGORITHMS) && _LIBCUDACXX_STD_VER >= 17
+#   include <__pstl_execution>
+#endif
+
+#endif // _LIBCUDACXX_EXECUTION
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config
new file mode 100644
index 000000000000..4571a0892400
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__config
@@ -0,0 +1,79 @@
+// -*- C++ -*-
+//===--------------------------- __config ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_CONFIG
+#define _LIBCUDACXX_EXPERIMENTAL_CONFIG
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL namespace std { namespace experimental {
+#define _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL  } }
+#define _CUDA_VSTD_EXPERIMENTAL std::experimental
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_LFTS _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL inline namespace fundamentals_v1 {
+#define _LIBCUDACXX_END_NAMESPACE_LFTS  } } }
+#define _CUDA_VSTD_LFTS _CUDA_VSTD_EXPERIMENTAL::fundamentals_v1
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_V2 _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL inline namespace fundamentals_v2 {
+#define _LIBCUDACXX_END_NAMESPACE_LFTS_V2  } } }
+#define _CUDA_VSTD_LFTS_V2 _CUDA_VSTD_EXPERIMENTAL::fundamentals_v2
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR _LIBCUDACXX_BEGIN_NAMESPACE_LFTS namespace pmr {
+#define _LIBCUDACXX_END_NAMESPACE_LFTS_PMR _LIBCUDACXX_END_NAMESPACE_LFTS }
+#define _CUDA_VSTD_LFTS_PMR _CUDA_VSTD_LFTS::pmr
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_CHRONO_LFTS _LIBCUDACXX_BEGIN_NAMESPACE_STD        \
+  namespace chrono { namespace experimental { inline namespace fundamentals_v1 {
+#define _LIBCUDACXX_END_NAMESPACE_CHRONO_LFTS _LIBCUDACXX_END_NAMESPACE_STD } } }
+
+#if defined(_LIBCUDACXX_NO_EXPERIMENTAL_DEPRECATION_WARNING_FILESYSTEM)
+#   define _LIBCUDACXX_DEPRECATED_EXPERIMENTAL_FILESYSTEM /* nothing */
+#else
+#   define _LIBCUDACXX_DEPRECATED_EXPERIMENTAL_FILESYSTEM __attribute__((deprecated("std::experimental::filesystem has now been deprecated in favor of C++17's std::filesystem. Please stop using it and start using std::filesystem. This experimental version will be removed in LLVM 11. You can remove this warning by defining the _LIBCUDACXX_NO_EXPERIMENTAL_DEPRECATION_WARNING_FILESYSTEM macro.")))
+#endif
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_FILESYSTEM \
+    _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL namespace filesystem _LIBCUDACXX_DEPRECATED_EXPERIMENTAL_FILESYSTEM { \
+    inline namespace v1 {
+
+#define _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_FILESYSTEM \
+    } } _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_COROUTINES \
+  _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL inline namespace coroutines_v1 {
+
+#define _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_COROUTINES \
+  } _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL
+
+#define _CUDA_VSTD_CORO _CUDA_VSTD_EXPERIMENTAL::coroutines_v1
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD \
+    _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL inline namespace parallelism_v2 {
+
+#define _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_SIMD \
+    } _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL
+
+#define _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD_ABI \
+    _LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD namespace simd_abi {
+
+#define _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_SIMD_ABI \
+    } _LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_SIMD
+
+// TODO: support more targets
+#if defined(__AVX__)
+#define _LIBCUDACXX_NATIVE_SIMD_WIDTH_IN_BYTES 32
+#else
+#define _LIBCUDACXX_NATIVE_SIMD_WIDTH_IN_BYTES 16
+#endif
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__memory b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__memory
new file mode 100644
index 000000000000..db27511e7450
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/__memory
@@ -0,0 +1,89 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL___MEMORY
+#define _LIBCUDACXX_EXPERIMENTAL___MEMORY
+
+#include <experimental/__config>
+#include <experimental/utility> // for erased_type
+#include <__functional_base>
+#include <type_traits>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS
+
+template <
+    class _Tp, class _Alloc
+  , bool = uses_allocator<_Tp, _Alloc>::value
+  , bool = __has_allocator_type<_Tp>::value
+  >
+struct __lfts_uses_allocator : public false_type {};
+
+template <class _Tp, class _Alloc>
+struct __lfts_uses_allocator<_Tp, _Alloc, false, false> : public false_type {};
+
+template <class _Tp, class _Alloc, bool HasAlloc>
+struct __lfts_uses_allocator<_Tp, _Alloc, true, HasAlloc> : public true_type {};
+
+template <class _Tp, class _Alloc>
+struct __lfts_uses_allocator<_Tp, _Alloc, false, true>
+  : public integral_constant<bool
+    , is_convertible<_Alloc, typename _Tp::allocator_type>::value
+      || is_same<erased_type, typename _Tp::allocator_type>::value
+    >
+{};
+
+template <bool _UsesAlloc, class _Tp, class _Alloc, class ..._Args>
+struct __lfts_uses_alloc_ctor_imp
+{
+    static const int value = 0;
+};
+
+template <class _Tp, class _Alloc, class ..._Args>
+struct __lfts_uses_alloc_ctor_imp<true, _Tp, _Alloc, _Args...>
+{
+    static const bool __ic_first
+        = is_constructible<_Tp, allocator_arg_t, _Alloc, _Args...>::value;
+
+    static const bool __ic_second =
+        conditional<
+            __ic_first,
+            false_type,
+            is_constructible<_Tp, _Args..., _Alloc>
+        >::type::value;
+
+    static_assert(__ic_first || __ic_second,
+                  "Request for uses allocator construction is ill-formed");
+
+    static const int value = __ic_first ? 1 : 2;
+};
+
+template <class _Tp, class _Alloc, class ..._Args>
+struct __lfts_uses_alloc_ctor
+  : integral_constant<int,
+        __lfts_uses_alloc_ctor_imp<
+            __lfts_uses_allocator<_Tp, _Alloc>::value
+          , _Tp, _Alloc, _Args...
+        >::value
+    >
+{};
+
+template <class _Tp, class _Alloc, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __lfts_user_alloc_construct(
+    _Tp * __store, const _Alloc & __a, _Args &&... __args)
+{
+    _CUDA_VSTD::__user_alloc_construct_impl(
+        typename __lfts_uses_alloc_ctor<_Tp, _Alloc, _Args...>::type()
+       , __store, __a, _CUDA_VSTD::forward<_Args>(__args)...
+       );
+}
+
+_LIBCUDACXX_END_NAMESPACE_LFTS
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL___MEMORY */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm
new file mode 100644
index 000000000000..7fe305cf0b47
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/algorithm
@@ -0,0 +1,59 @@
+// -*- C++ -*-
+//===-------------------------- algorithm ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_ALGORITHM
+#define _LIBCUDACXX_EXPERIMENTAL_ALGORITHM
+
+/*
+   experimental/algorithm synopsis
+
+#include <algorithm>
+
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+
+template <class ForwardIterator, class Searcher>
+ForwardIterator search(ForwardIterator first, ForwardIterator last,
+                       const Searcher &searcher);
+
+// sample removed because it's now part of C++17
+
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+*/
+
+#include <experimental/__config>
+#include <algorithm>
+#include <type_traits>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS
+
+template <class _ForwardIterator, class _Searcher>
+_LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator search(_ForwardIterator __f, _ForwardIterator __l, const _Searcher &__s)
+{ return __s(__f, __l).first; }
+
+_LIBCUDACXX_END_NAMESPACE_LFTS
+
+_LIBCUDACXX_POP_MACROS
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_ALGORITHM */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine
new file mode 100644
index 000000000000..f1df755cff18
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/coroutine
@@ -0,0 +1,334 @@
+// -*- C++ -*-
+//===----------------------------- coroutine -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_COROUTINE
+#define _LIBCUDACXX_EXPERIMENTAL_COROUTINE
+
+/**
+    experimental/coroutine synopsis
+
+// C++next
+
+namespace std {
+namespace experimental {
+inline namespace coroutines_v1 {
+
+  // 18.11.1 coroutine traits
+template <typename R, typename... ArgTypes>
+class coroutine_traits;
+// 18.11.2 coroutine handle
+template <typename Promise = void>
+class coroutine_handle;
+// 18.11.2.7 comparison operators:
+bool operator==(coroutine_handle<> x, coroutine_handle<> y) _NOEXCEPT;
+bool operator!=(coroutine_handle<> x, coroutine_handle<> y) _NOEXCEPT;
+bool operator<(coroutine_handle<> x, coroutine_handle<> y) _NOEXCEPT;
+bool operator<=(coroutine_handle<> x, coroutine_handle<> y) _NOEXCEPT;
+bool operator>=(coroutine_handle<> x, coroutine_handle<> y) _NOEXCEPT;
+bool operator>(coroutine_handle<> x, coroutine_handle<> y) _NOEXCEPT;
+// 18.11.3 trivial awaitables
+struct suspend_never;
+struct suspend_always;
+// 18.11.2.8 hash support:
+template <class T> struct hash;
+template <class P> struct hash<coroutine_handle<P>>;
+
+} // namespace coroutines_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <new>
+#include <type_traits>
+#include <functional>
+#include <memory> // for hash<T*>
+#include <cstddef>
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_COROUTINES
+# if defined(_LIBCUDACXX_WARNING)
+    _LIBCUDACXX_WARNING("<experimental/coroutine> cannot be used with this compiler")
+# else
+#   warning <experimental/coroutine> cannot be used with this compiler
+# endif
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_COROUTINES
+
+_LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_COROUTINES
+
+template <class _Tp, class = void>
+struct __coroutine_traits_sfinae {};
+
+template <class _Tp>
+struct __coroutine_traits_sfinae<
+    _Tp, typename __void_t<typename _Tp::promise_type>::type>
+{
+  using promise_type = typename _Tp::promise_type;
+};
+
+template <typename _Ret, typename... _Args>
+struct coroutine_traits
+    : public __coroutine_traits_sfinae<_Ret>
+{
+};
+
+template <typename _Promise = void>
+class _LIBCUDACXX_TEMPLATE_VIS coroutine_handle;
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS coroutine_handle<void> {
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR coroutine_handle() _NOEXCEPT : __handle_(nullptr) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR coroutine_handle(nullptr_t) _NOEXCEPT : __handle_(nullptr) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    coroutine_handle& operator=(nullptr_t) _NOEXCEPT {
+        __handle_ = nullptr;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR void* address() const _NOEXCEPT { return __handle_; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR explicit operator bool() const _NOEXCEPT { return __handle_; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()() { resume(); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void resume() {
+      _LIBCUDACXX_ASSERT(__is_suspended(),
+                     "resume() can only be called on suspended coroutines");
+      _LIBCUDACXX_ASSERT(!done(),
+                "resume() has undefined behavior when the coroutine is done");
+      __builtin_coro_resume(__handle_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void destroy() {
+      _LIBCUDACXX_ASSERT(__is_suspended(),
+                     "destroy() can only be called on suspended coroutines");
+      __builtin_coro_destroy(__handle_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool done() const {
+      _LIBCUDACXX_ASSERT(__is_suspended(),
+                     "done() can only be called on suspended coroutines");
+      return __builtin_coro_done(__handle_);
+    }
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static coroutine_handle from_address(void* __addr) _NOEXCEPT {
+        coroutine_handle __tmp;
+        __tmp.__handle_ = __addr;
+        return __tmp;
+    }
+
+    // FIXME: Should from_address(nullptr) be allowed?
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static coroutine_handle from_address(nullptr_t) _NOEXCEPT {
+      return coroutine_handle(nullptr);
+    }
+
+    template <class _Tp, bool _CallIsValid = false>
+    static coroutine_handle from_address(_Tp*) {
+      static_assert(_CallIsValid,
+       "coroutine_handle<void>::from_address cannot be called with "
+        "non-void pointers");
+    }
+
+private:
+  bool __is_suspended() const _NOEXCEPT  {
+    // FIXME actually implement a check for if the coro is suspended.
+    return __handle_;
+  }
+
+  template <class _PromiseT> friend class coroutine_handle;
+  void* __handle_;
+};
+
+// 18.11.2.7 comparison operators:
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(coroutine_handle<> __x, coroutine_handle<> __y) _NOEXCEPT {
+    return __x.address() == __y.address();
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(coroutine_handle<> __x, coroutine_handle<> __y) _NOEXCEPT {
+    return !(__x == __y);
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<(coroutine_handle<> __x, coroutine_handle<> __y) _NOEXCEPT {
+    return less<void*>()(__x.address(), __y.address());
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>(coroutine_handle<> __x, coroutine_handle<> __y) _NOEXCEPT {
+    return __y < __x;
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<=(coroutine_handle<> __x, coroutine_handle<> __y) _NOEXCEPT {
+    return !(__x > __y);
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>=(coroutine_handle<> __x, coroutine_handle<> __y) _NOEXCEPT {
+    return !(__x < __y);
+}
+
+template <typename _Promise>
+class _LIBCUDACXX_TEMPLATE_VIS coroutine_handle : public coroutine_handle<> {
+    using _Base = coroutine_handle<>;
+public:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    // 18.11.2.1 construct/reset
+    using coroutine_handle<>::coroutine_handle;
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY coroutine_handle() _NOEXCEPT : _Base() {}
+    _LIBCUDACXX_INLINE_VISIBILITY coroutine_handle(nullptr_t) _NOEXCEPT : _Base(nullptr) {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    coroutine_handle& operator=(nullptr_t) _NOEXCEPT {
+        _Base::operator=(nullptr);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Promise& promise() const {
+        return *static_cast<_Promise*>(
+            __builtin_coro_promise(this->__handle_, _LIBCUDACXX_ALIGNOF(_Promise), false));
+    }
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static coroutine_handle from_address(void* __addr) _NOEXCEPT {
+        coroutine_handle __tmp;
+        __tmp.__handle_ = __addr;
+        return __tmp;
+    }
+
+    // NOTE: this overload isn't required by the standard but is needed so
+    // the deleted _Promise* overload doesn't make from_address(nullptr)
+    // ambiguous.
+    // FIXME: should from_address work with nullptr?
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static coroutine_handle from_address(nullptr_t) _NOEXCEPT {
+      return coroutine_handle(nullptr);
+    }
+
+    template <class _Tp, bool _CallIsValid = false>
+    static coroutine_handle from_address(_Tp*) {
+      static_assert(_CallIsValid,
+       "coroutine_handle<promise_type>::from_address cannot be called with "
+        "non-void pointers");
+    }
+
+    template <bool _CallIsValid = false>
+    static coroutine_handle from_address(_Promise*) {
+      static_assert(_CallIsValid,
+       "coroutine_handle<promise_type>::from_address cannot be used with "
+        "pointers to the coroutine's promise type; use 'from_promise' instead");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static coroutine_handle from_promise(_Promise& __promise) _NOEXCEPT {
+        typedef typename remove_cv<_Promise>::type _RawPromise;
+        coroutine_handle __tmp;
+        __tmp.__handle_ = __builtin_coro_promise(
+            _CUDA_VSTD::addressof(const_cast<_RawPromise&>(__promise)),
+             _LIBCUDACXX_ALIGNOF(_Promise), true);
+        return __tmp;
+    }
+};
+
+#if __has_builtin(__builtin_coro_noop)
+struct noop_coroutine_promise {};
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS coroutine_handle<noop_coroutine_promise>
+    : public coroutine_handle<> {
+  using _Base = coroutine_handle<>;
+  using _Promise = noop_coroutine_promise;
+public:
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _Promise& promise() const {
+    return *static_cast<_Promise*>(
+      __builtin_coro_promise(this->__handle_, _LIBCUDACXX_ALIGNOF(_Promise), false));
+  }
+
+  _LIBCUDACXX_CONSTEXPR explicit operator bool() const _NOEXCEPT { return true; }
+  _LIBCUDACXX_CONSTEXPR bool done() const _NOEXCEPT { return false; }
+
+  _LIBCUDACXX_CONSTEXPR_AFTER_CXX17 void operator()() const _NOEXCEPT {}
+  _LIBCUDACXX_CONSTEXPR_AFTER_CXX17 void resume() const _NOEXCEPT {}
+  _LIBCUDACXX_CONSTEXPR_AFTER_CXX17 void destroy() const _NOEXCEPT {}
+
+private:
+  _LIBCUDACXX_INLINE_VISIBILITY
+  friend coroutine_handle<noop_coroutine_promise> noop_coroutine() _NOEXCEPT;
+
+  _LIBCUDACXX_INLINE_VISIBILITY coroutine_handle() _NOEXCEPT {
+    this->__handle_ = __builtin_coro_noop();
+  }
+};
+
+using noop_coroutine_handle = coroutine_handle<noop_coroutine_promise>;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+noop_coroutine_handle noop_coroutine() _NOEXCEPT {
+  return noop_coroutine_handle();
+}
+#endif // __has_builtin(__builtin_coro_noop)
+
+struct suspend_never {
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool await_ready() const _NOEXCEPT { return true; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void await_suspend(coroutine_handle<>) const _NOEXCEPT {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void await_resume() const _NOEXCEPT {}
+};
+
+struct suspend_always {
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool await_ready() const _NOEXCEPT { return false; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void await_suspend(coroutine_handle<>) const _NOEXCEPT {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void await_resume() const _NOEXCEPT {}
+};
+
+_LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_COROUTINES
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp>
+struct hash<_CUDA_VSTD_CORO::coroutine_handle<_Tp> > {
+    using __arg_type = _CUDA_VSTD_CORO::coroutine_handle<_Tp>;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(__arg_type const& __v) const _NOEXCEPT
+    {return hash<void*>()(__v.address());}
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // !defined(_LIBCUDACXX_HAS_NO_COROUTINES)
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_COROUTINE */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque
new file mode 100644
index 000000000000..1f2291796db9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/deque
@@ -0,0 +1,46 @@
+// -*- C++ -*-
+//===--------------------------- deque ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_DEQUE
+#define _LIBCUDACXX_EXPERIMENTAL_DEQUE
+/*
+    experimental/deque synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class T>
+  using deque = std::deque<T,polymorphic_allocator<T>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <deque>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _ValueT>
+using deque = _CUDA_VSTD::deque<_ValueT, polymorphic_allocator<_ValueT>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_DEQUE */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem
new file mode 100644
index 000000000000..2b0d1dba70fa
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/filesystem
@@ -0,0 +1,256 @@
+// -*- C++ -*-
+//===--------------------------- filesystem -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef _LIBCUDACXX_EXPERIMENTAL_FILESYSTEM
+#define _LIBCUDACXX_EXPERIMENTAL_FILESYSTEM
+/*
+    filesystem synopsis
+
+    namespace std { namespace experimental { namespace filesystem { inline namespace v1 {
+
+    class path;
+
+    void swap(path& lhs, path& rhs) noexcept;
+    size_t hash_value(const path& p) noexcept;
+
+    bool operator==(const path& lhs, const path& rhs) noexcept;
+    bool operator!=(const path& lhs, const path& rhs) noexcept;
+    bool operator< (const path& lhs, const path& rhs) noexcept;
+    bool operator<=(const path& lhs, const path& rhs) noexcept;
+    bool operator> (const path& lhs, const path& rhs) noexcept;
+    bool operator>=(const path& lhs, const path& rhs) noexcept;
+
+    path operator/ (const path& lhs, const path& rhs);
+
+    // fs.path.io operators are friends of path.
+    template <class charT, class traits>
+    friend basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os, const path& p);
+
+    template <class charT, class traits>
+    friend basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is, path& p);
+
+    template <class Source>
+      path u8path(const Source& source);
+    template <class InputIterator>
+      path u8path(InputIterator first, InputIterator last);
+
+    class filesystem_error;
+    class directory_entry;
+
+    class directory_iterator;
+
+    // enable directory_iterator range-based for statements
+    directory_iterator begin(directory_iterator iter) noexcept;
+    directory_iterator end(const directory_iterator&) noexcept;
+
+    class recursive_directory_iterator;
+
+    // enable recursive_directory_iterator range-based for statements
+    recursive_directory_iterator begin(recursive_directory_iterator iter) noexcept;
+    recursive_directory_iterator end(const recursive_directory_iterator&) noexcept;
+
+    class file_status;
+
+    struct space_info
+    {
+      uintmax_t capacity;
+      uintmax_t free;
+      uintmax_t available;
+    };
+
+    enum class file_type;
+    enum class perms;
+    enum class perm_options;
+    enum class copy_options;
+    enum class directory_options;
+
+    typedef chrono::time_point<trivial-clock>  file_time_type;
+
+    // operational functions
+
+    path absolute(const path& p);
+    path absolute(const path& p, error_code &ec);
+
+    path canonical(const path& p);
+    path canonical(const path& p, error_code& ec);
+
+    void copy(const path& from, const path& to);
+    void copy(const path& from, const path& to, error_code& ec);
+    void copy(const path& from, const path& to, copy_options options);
+    void copy(const path& from, const path& to, copy_options options,
+                   error_code& ec);
+
+    bool copy_file(const path& from, const path& to);
+    bool copy_file(const path& from, const path& to, error_code& ec);
+    bool copy_file(const path& from, const path& to, copy_options option);
+    bool copy_file(const path& from, const path& to, copy_options option,
+                           error_code& ec);
+
+    void copy_symlink(const path& existing_symlink, const path& new_symlink);
+    void copy_symlink(const path& existing_symlink, const path& new_symlink,
+                              error_code& ec) noexcept;
+
+    bool create_directories(const path& p);
+    bool create_directories(const path& p, error_code& ec);
+
+    bool create_directory(const path& p);
+    bool create_directory(const path& p, error_code& ec) noexcept;
+
+    bool create_directory(const path& p, const path& attributes);
+    bool create_directory(const path& p, const path& attributes,
+                                  error_code& ec) noexcept;
+
+    void create_directory_symlink(const path& to, const path& new_symlink);
+    void create_directory_symlink(const path& to, const path& new_symlink,
+                                          error_code& ec) noexcept;
+
+    void create_hard_link(const path& to, const path& new_hard_link);
+    void create_hard_link(const path& to, const path& new_hard_link,
+                                  error_code& ec) noexcept;
+
+    void create_symlink(const path& to, const path& new_symlink);
+    void create_symlink(const path& to, const path& new_symlink,
+                                error_code& ec) noexcept;
+
+    path current_path();
+    path current_path(error_code& ec);
+    void current_path(const path& p);
+    void current_path(const path& p, error_code& ec) noexcept;
+
+    bool exists(file_status s) noexcept;
+    bool exists(const path& p);
+    bool exists(const path& p, error_code& ec) noexcept;
+
+    bool equivalent(const path& p1, const path& p2);
+    bool equivalent(const path& p1, const path& p2, error_code& ec) noexcept;
+
+    uintmax_t    file_size(const path& p);
+    uintmax_t    file_size(const path& p, error_code& ec) noexcept;
+
+    uintmax_t    hard_link_count(const path& p);
+    uintmax_t    hard_link_count(const path& p, error_code& ec) noexcept;
+
+    bool is_block_file(file_status s) noexcept;
+    bool is_block_file(const path& p);
+    bool is_block_file(const path& p, error_code& ec) noexcept;
+
+    bool is_character_file(file_status s) noexcept;
+    bool is_character_file(const path& p);
+    bool is_character_file(const path& p, error_code& ec) noexcept;
+
+    bool is_directory(file_status s) noexcept;
+    bool is_directory(const path& p);
+    bool is_directory(const path& p, error_code& ec) noexcept;
+
+    bool is_empty(const path& p);
+    bool is_empty(const path& p, error_code& ec) noexcept;
+
+    bool is_fifo(file_status s) noexcept;
+    bool is_fifo(const path& p);
+    bool is_fifo(const path& p, error_code& ec) noexcept;
+
+    bool is_other(file_status s) noexcept;
+    bool is_other(const path& p);
+    bool is_other(const path& p, error_code& ec) noexcept;
+
+    bool is_regular_file(file_status s) noexcept;
+    bool is_regular_file(const path& p);
+    bool is_regular_file(const path& p, error_code& ec) noexcept;
+
+    bool is_socket(file_status s) noexcept;
+    bool is_socket(const path& p);
+    bool is_socket(const path& p, error_code& ec) noexcept;
+
+    bool is_symlink(file_status s) noexcept;
+    bool is_symlink(const path& p);
+    bool is_symlink(const path& p, error_code& ec) noexcept;
+
+    file_time_type  last_write_time(const path& p);
+    file_time_type  last_write_time(const path& p, error_code& ec) noexcept;
+    void last_write_time(const path& p, file_time_type new_time);
+    void last_write_time(const path& p, file_time_type new_time,
+                                 error_code& ec) noexcept;
+
+    void permissions(const path& p, perms prms,
+                     perm_options opts=perm_options::replace);
+    void permissions(const path& p, perms prms, error_code& ec) noexcept;
+    void permissions(const path& p, perms prms, perm_options opts,
+                     error_code& ec);
+
+    path proximate(const path& p, error_code& ec);
+    path proximate(const path& p, const path& base = current_path());
+    path proximate(const path& p, const path& base, error_code &ec);
+
+    path read_symlink(const path& p);
+    path read_symlink(const path& p, error_code& ec);
+
+    path relative(const path& p, error_code& ec);
+    path relative(const path& p, const path& base=current_path());
+    path relative(const path& p, const path& base, error_code& ec);
+
+    bool remove(const path& p);
+    bool remove(const path& p, error_code& ec) noexcept;
+
+    uintmax_t    remove_all(const path& p);
+    uintmax_t    remove_all(const path& p, error_code& ec);
+
+    void rename(const path& from, const path& to);
+    void rename(const path& from, const path& to, error_code& ec) noexcept;
+
+    void resize_file(const path& p, uintmax_t size);
+    void resize_file(const path& p, uintmax_t size, error_code& ec) noexcept;
+
+    space_info   space(const path& p);
+    space_info   space(const path& p, error_code& ec) noexcept;
+
+    file_status  status(const path& p);
+    file_status  status(const path& p, error_code& ec) noexcept;
+
+    bool status_known(file_status s) noexcept;
+
+    file_status  symlink_status(const path& p);
+    file_status  symlink_status(const path& p, error_code& ec) noexcept;
+
+    path temp_directory_path();
+    path temp_directory_path(error_code& ec);
+
+    path weakly_canonical(path const& p);
+    path weakly_canonical(path const& p, error_code& ec);
+
+
+} } } }  // namespaces std::experimental::filesystem::v1
+
+*/
+
+#include <experimental/__config>
+#include <filesystem>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+#define __cpp_lib_experimental_filesystem 201406
+
+_LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_FILESYSTEM
+
+using namespace _CUDA_VSTD_FS;
+
+_LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_FILESYSTEM
+
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_POP_MACROS
+
+#endif // _LIBCUDACXX_EXPERIMENTAL_FILESYSTEM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list
new file mode 100644
index 000000000000..585ca5a8d6fc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/forward_list
@@ -0,0 +1,46 @@
+// -*- C++ -*-
+//===--------------------------- forward_list -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_FORWARD_LIST
+#define _LIBCUDACXX_EXPERIMENTAL_FORWARD_LIST
+/*
+    experimental/forward_list synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class T>
+  using forward_list = std::forward_list<T,polymorphic_allocator<T>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <forward_list>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _ValueT>
+using forward_list = _CUDA_VSTD::forward_list<_ValueT, polymorphic_allocator<_ValueT>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_FORWARD_LIST */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional
new file mode 100644
index 000000000000..4004c7df44c9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/functional
@@ -0,0 +1,458 @@
+// -*- C++ -*-
+//===-------------------------- functional --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_FUNCTIONAL
+#define _LIBCUDACXX_EXPERIMENTAL_FUNCTIONAL
+
+/*
+   experimental/functional synopsis
+
+#include <algorithm>
+
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+
+    // See C++14 20.9.9, Function object binders
+    template <class T> constexpr bool is_bind_expression_v
+      = is_bind_expression<T>::value;
+    template <class T> constexpr int is_placeholder_v
+      = is_placeholder<T>::value;
+
+    // 4.2, Class template function
+    template<class> class function; // undefined
+    template<class R, class... ArgTypes> class function<R(ArgTypes...)>;
+
+    template<class R, class... ArgTypes>
+    void swap(function<R(ArgTypes...)>&, function<R(ArgTypes...)>&);
+
+    template<class R, class... ArgTypes>
+    bool operator==(const function<R(ArgTypes...)>&, nullptr_t) noexcept;
+    template<class R, class... ArgTypes>
+    bool operator==(nullptr_t, const function<R(ArgTypes...)>&) noexcept;
+    template<class R, class... ArgTypes>
+    bool operator!=(const function<R(ArgTypes...)>&, nullptr_t) noexcept;
+    template<class R, class... ArgTypes>
+    bool operator!=(nullptr_t, const function<R(ArgTypes...)>&) noexcept;
+
+    // 4.3, Searchers
+    template<class ForwardIterator, class BinaryPredicate = equal_to<>>
+      class default_searcher;
+
+    template<class RandomAccessIterator,
+             class Hash = hash<typename iterator_traits<RandomAccessIterator>::value_type>,
+             class BinaryPredicate = equal_to<>>
+      class boyer_moore_searcher;
+
+    template<class RandomAccessIterator,
+             class Hash = hash<typename iterator_traits<RandomAccessIterator>::value_type>,
+             class BinaryPredicate = equal_to<>>
+      class boyer_moore_horspool_searcher;
+
+    template<class ForwardIterator, class BinaryPredicate = equal_to<>>
+    default_searcher<ForwardIterator, BinaryPredicate>
+    make_default_searcher(ForwardIterator pat_first, ForwardIterator pat_last,
+                          BinaryPredicate pred = BinaryPredicate());
+
+    template<class RandomAccessIterator,
+             class Hash = hash<typename iterator_traits<RandomAccessIterator>::value_type>,
+             class BinaryPredicate = equal_to<>>
+    boyer_moore_searcher<RandomAccessIterator, Hash, BinaryPredicate>
+    make_boyer_moore_searcher(
+        RandomAccessIterator pat_first, RandomAccessIterator pat_last,
+        Hash hf = Hash(), BinaryPredicate pred = BinaryPredicate());
+
+    template<class RandomAccessIterator,
+             class Hash = hash<typename iterator_traits<RandomAccessIterator>::value_type>,
+             class BinaryPredicate = equal_to<>>
+    boyer_moore_horspool_searcher<RandomAccessIterator, Hash, BinaryPredicate>
+    make_boyer_moore_horspool_searcher(
+        RandomAccessIterator pat_first, RandomAccessIterator pat_last,
+        Hash hf = Hash(), BinaryPredicate pred = BinaryPredicate());
+
+  } // namespace fundamentals_v1
+  } // namespace experimental
+
+  template<class R, class... ArgTypes, class Alloc>
+  struct uses_allocator<experimental::function<R(ArgTypes...)>, Alloc>;
+
+} // namespace std
+
+*/
+
+#include <experimental/__config>
+#include <functional>
+#include <algorithm>
+#include <type_traits>
+#include <vector>
+#include <array>
+#include <unordered_map>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS
+
+#if _LIBCUDACXX_STD_VER > 11
+// default searcher
+template<class _ForwardIterator, class _BinaryPredicate = equal_to<>>
+class _LIBCUDACXX_TYPE_VIS default_searcher {
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    default_searcher(_ForwardIterator __f, _ForwardIterator __l, 
+                       _BinaryPredicate __p = _BinaryPredicate())
+        : __first_(__f), __last_(__l), __pred_(__p) {}
+
+    template <typename _ForwardIterator2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<_ForwardIterator2, _ForwardIterator2>
+    operator () (_ForwardIterator2 __f, _ForwardIterator2 __l) const
+    {
+        return _CUDA_VSTD::__search(__f, __l, __first_, __last_, __pred_,
+            typename _CUDA_VSTD::iterator_traits<_ForwardIterator>::iterator_category(),
+            typename _CUDA_VSTD::iterator_traits<_ForwardIterator2>::iterator_category());
+    }
+
+private:
+    _ForwardIterator __first_;
+    _ForwardIterator __last_;
+    _BinaryPredicate __pred_;
+    };
+
+template<class _ForwardIterator, class _BinaryPredicate = equal_to<>>
+_LIBCUDACXX_INLINE_VISIBILITY
+default_searcher<_ForwardIterator, _BinaryPredicate>
+make_default_searcher( _ForwardIterator __f, _ForwardIterator __l, _BinaryPredicate __p = _BinaryPredicate ())
+{
+    return default_searcher<_ForwardIterator, _BinaryPredicate>(__f, __l, __p);
+}
+
+template<class _Key, class _Value, class _Hash, class _BinaryPredicate, bool /*useArray*/> class _BMSkipTable;
+
+//  General case for BM data searching; use a map
+template<class _Key, typename _Value, class _Hash, class _BinaryPredicate>
+class _BMSkipTable<_Key, _Value, _Hash, _BinaryPredicate, false> {
+public: // TODO private:
+    typedef _Value value_type;
+    typedef _Key   key_type;
+
+    const _Value __default_value_;
+    std::unordered_map<_Key, _Value, _Hash, _BinaryPredicate> __table;
+    
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _BMSkipTable(std::size_t __sz, _Value __default, _Hash __hf, _BinaryPredicate __pred)
+        : __default_value_(__default), __table(__sz, __hf, __pred) {}
+    
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(const key_type &__key, value_type __val)
+    {
+        __table [__key] = __val;    // Would skip_.insert (val) be better here?
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type operator [](const key_type & __key) const
+    {
+        auto __it = __table.find (__key);
+        return __it == __table.end() ? __default_value_ : __it->second;
+    }
+};
+    
+
+//  Special case small numeric values; use an array
+template<class _Key, typename _Value, class _Hash, class _BinaryPredicate>
+class _BMSkipTable<_Key, _Value, _Hash, _BinaryPredicate, true> {
+private:
+    typedef _Value value_type;
+    typedef _Key   key_type;
+
+    typedef typename std::make_unsigned<key_type>::type unsigned_key_type;
+    typedef std::array<value_type, _CUDA_VSTD::numeric_limits<unsigned_key_type>::max()> skip_map;
+    skip_map __table;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _BMSkipTable(std::size_t /*__sz*/, _Value __default, _Hash /*__hf*/, _BinaryPredicate /*__pred*/)
+    {
+        std::fill_n(__table.begin(), __table.size(), __default);
+    }
+    
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(key_type __key, value_type __val)
+    {
+        __table[static_cast<unsigned_key_type>(__key)] = __val;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type operator [](key_type __key) const
+    {
+        return __table[static_cast<unsigned_key_type>(__key)];
+    }
+};
+
+
+template <class _RandomAccessIterator1, 
+          class _Hash = hash<typename iterator_traits<_RandomAccessIterator1>::value_type>, 
+          class _BinaryPredicate = equal_to<>>
+class _LIBCUDACXX_TYPE_VIS boyer_moore_searcher {
+private:
+    typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type difference_type;
+    typedef typename std::iterator_traits<_RandomAccessIterator1>::value_type      value_type;
+    typedef _BMSkipTable<value_type, difference_type, _Hash, _BinaryPredicate,
+                    _CUDA_VSTD::is_integral<value_type>::value && // what about enums?
+                    sizeof(value_type) == 1 &&
+                    is_same<_Hash, hash<value_type>>::value &&
+                    is_same<_BinaryPredicate, equal_to<>>::value
+            > skip_table_type;
+    
+public:
+    boyer_moore_searcher(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l, 
+                _Hash __hf = _Hash(), _BinaryPredicate __pred = _BinaryPredicate())
+            : __first_(__f), __last_(__l), __pred_(__pred),
+              __pattern_length_(_CUDA_VSTD::distance(__first_, __last_)),
+              __skip_{make_shared<skip_table_type>(__pattern_length_, -1, __hf, __pred_)},
+              __suffix_{make_shared<vector<difference_type>>(__pattern_length_ + 1)}
+        {
+    //  build the skip table
+        for ( difference_type __i = 0; __f != __l; ++__f, (void) ++__i )
+            __skip_->insert(*__f, __i);
+
+        this->__build_suffix_table ( __first_, __last_, __pred_ );
+        }
+        
+    template <typename _RandomAccessIterator2>
+    pair<_RandomAccessIterator2, _RandomAccessIterator2>
+    operator ()(_RandomAccessIterator2 __f, _RandomAccessIterator2 __l) const
+    {
+        static_assert ( std::is_same<
+                typename std::__uncvref<typename std::iterator_traits<_RandomAccessIterator1>::value_type>::type, 
+                typename std::__uncvref<typename std::iterator_traits<_RandomAccessIterator2>::value_type>::type
+                    >::value,
+                "Corpus and Pattern iterators must point to the same type" );
+
+        if (__f      == __l )    return make_pair(__l, __l); // empty corpus
+        if (__first_ == __last_) return make_pair(__f, __f); // empty pattern
+
+    //  If the pattern is larger than the corpus, we can't find it!
+        if ( __pattern_length_ > _CUDA_VSTD::distance (__f, __l)) 
+            return make_pair(__l, __l);
+
+    //  Do the search 
+        return this->__search(__f, __l);
+    }
+        
+public: // TODO private:
+    _RandomAccessIterator1               __first_;
+    _RandomAccessIterator1               __last_;
+    _BinaryPredicate                     __pred_;
+    difference_type                      __pattern_length_;
+    shared_ptr<skip_table_type>          __skip_;
+    shared_ptr<vector<difference_type>>  __suffix_;
+
+    template <typename _RandomAccessIterator2>
+    pair<_RandomAccessIterator2, _RandomAccessIterator2>
+    __search(_RandomAccessIterator2 __f, _RandomAccessIterator2 __l) const
+    {
+        _RandomAccessIterator2 __cur = __f;
+        const _RandomAccessIterator2 __last = __l - __pattern_length_;
+        const skip_table_type &         __skip   = *__skip_.get();
+        const vector<difference_type> & __suffix = *__suffix_.get();
+        
+        while (__cur <= __last)
+        {
+
+        //  Do we match right where we are?
+            difference_type __j = __pattern_length_;
+            while (__pred_(__first_ [__j-1], __cur [__j-1])) {
+                __j--;
+            //  We matched - we're done!
+                if ( __j == 0 )
+                    return make_pair(__cur, __cur + __pattern_length_);
+                }
+            
+        //  Since we didn't match, figure out how far to skip forward
+            difference_type __k = __skip[__cur [ __j - 1 ]];
+            difference_type __m = __j - __k - 1;
+            if (__k < __j && __m > __suffix[ __j ])
+                __cur += __m;
+            else
+                __cur += __suffix[ __j ];
+        }
+    
+        return make_pair(__l, __l);     // We didn't find anything
+    }
+
+
+    template<typename _Iterator, typename _Container>
+    void __compute_bm_prefix ( _Iterator __f, _Iterator __l, _BinaryPredicate __pred, _Container &__prefix )
+    {
+        const std::size_t __count = _CUDA_VSTD::distance(__f, __l);
+                        
+        __prefix[0] = 0;
+        std::size_t __k = 0;
+        for ( std::size_t __i = 1; __i < __count; ++__i )
+        {
+            while ( __k > 0 && !__pred ( __f[__k], __f[__i] ))
+                __k = __prefix [ __k - 1 ];
+                
+            if ( __pred ( __f[__k], __f[__i] ))
+                __k++;
+            __prefix [ __i ] = __k;
+        }
+    }
+
+    void __build_suffix_table(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l, 
+                                                    _BinaryPredicate __pred)
+    {
+        const std::size_t __count = _CUDA_VSTD::distance(__f, __l);
+        vector<difference_type> & __suffix = *__suffix_.get();
+        if (__count > 0)
+        {
+            _CUDA_VSTD::vector<value_type> __scratch(__count);
+            
+            __compute_bm_prefix(__f, __l, __pred, __scratch);
+            for ( std::size_t __i = 0; __i <= __count; __i++ )
+                __suffix[__i] = __count - __scratch[__count-1];
+    
+            typedef _CUDA_VSTD::reverse_iterator<_RandomAccessIterator1> _RevIter;
+            __compute_bm_prefix(_RevIter(__l), _RevIter(__f), __pred, __scratch);
+     
+            for ( std::size_t __i = 0; __i < __count; __i++ )
+            {
+                const std::size_t     __j = __count - __scratch[__i];
+                const difference_type __k = __i     - __scratch[__i] + 1;
+     
+                if (__suffix[__j] > __k)
+                    __suffix[__j] = __k;
+            }
+        }
+    }
+
+};
+
+template<class _RandomAccessIterator, 
+         class _Hash = hash<typename iterator_traits<_RandomAccessIterator>::value_type>, 
+         class _BinaryPredicate = equal_to<>>
+_LIBCUDACXX_INLINE_VISIBILITY
+boyer_moore_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>
+make_boyer_moore_searcher( _RandomAccessIterator __f, _RandomAccessIterator __l, 
+                    _Hash __hf = _Hash(), _BinaryPredicate __p = _BinaryPredicate ())
+{
+    return boyer_moore_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>(__f, __l, __hf, __p);
+}
+
+// boyer-moore-horspool
+template <class _RandomAccessIterator1, 
+          class _Hash = hash<typename iterator_traits<_RandomAccessIterator1>::value_type>, 
+          class _BinaryPredicate = equal_to<>>
+class _LIBCUDACXX_TYPE_VIS boyer_moore_horspool_searcher {
+private:
+    typedef typename std::iterator_traits<_RandomAccessIterator1>::difference_type difference_type;
+    typedef typename std::iterator_traits<_RandomAccessIterator1>::value_type      value_type;
+    typedef _BMSkipTable<value_type, difference_type, _Hash, _BinaryPredicate,
+                    _CUDA_VSTD::is_integral<value_type>::value && // what about enums?
+                    sizeof(value_type) == 1 &&
+                    is_same<_Hash, hash<value_type>>::value &&
+                    is_same<_BinaryPredicate, equal_to<>>::value
+            > skip_table_type;
+
+public:
+    boyer_moore_horspool_searcher(_RandomAccessIterator1 __f, _RandomAccessIterator1 __l, 
+                _Hash __hf = _Hash(), _BinaryPredicate __pred = _BinaryPredicate())
+            : __first_(__f), __last_(__l), __pred_(__pred),
+              __pattern_length_(_CUDA_VSTD::distance(__first_, __last_)),
+              __skip_{_CUDA_VSTD::make_shared<skip_table_type>(__pattern_length_, __pattern_length_, __hf, __pred_)}
+        {
+    //  build the skip table
+            if ( __f != __l )
+            {
+                __l = __l - 1;
+                for ( difference_type __i = 0; __f != __l; ++__f, (void) ++__i )
+                    __skip_->insert(*__f, __pattern_length_ - 1 - __i);
+            }
+        }
+            
+    template <typename _RandomAccessIterator2>
+    pair<_RandomAccessIterator2, _RandomAccessIterator2>
+    operator ()(_RandomAccessIterator2 __f, _RandomAccessIterator2 __l) const
+    {
+        static_assert ( std::is_same<
+                typename std::__uncvref<typename std::iterator_traits<_RandomAccessIterator1>::value_type>::type, 
+                typename std::__uncvref<typename std::iterator_traits<_RandomAccessIterator2>::value_type>::type
+                    >::value,
+                "Corpus and Pattern iterators must point to the same type" );
+
+        if (__f      == __l )    return make_pair(__l, __l); // empty corpus
+        if (__first_ == __last_) return make_pair(__f, __f); // empty pattern
+
+    //  If the pattern is larger than the corpus, we can't find it!
+        if ( __pattern_length_ > _CUDA_VSTD::distance (__f, __l)) 
+            return make_pair(__l, __l);
+
+    //  Do the search 
+        return this->__search(__f, __l);
+    }
+        
+private:
+    _RandomAccessIterator1      __first_;
+    _RandomAccessIterator1      __last_;
+    _BinaryPredicate            __pred_;
+    difference_type             __pattern_length_;
+    shared_ptr<skip_table_type> __skip_;
+
+    template <typename _RandomAccessIterator2>
+    pair<_RandomAccessIterator2, _RandomAccessIterator2>
+    __search ( _RandomAccessIterator2 __f, _RandomAccessIterator2 __l ) const {
+        _RandomAccessIterator2 __cur = __f;
+        const _RandomAccessIterator2 __last = __l - __pattern_length_;
+        const skip_table_type & __skip = *__skip_.get();
+
+        while (__cur <= __last)
+        {
+        //  Do we match right where we are?
+            difference_type __j = __pattern_length_;
+            while (__pred_(__first_[__j-1], __cur[__j-1]))
+            {
+                __j--;
+            //  We matched - we're done!
+                if ( __j == 0 )
+                    return make_pair(__cur, __cur + __pattern_length_);
+            }
+            __cur += __skip[__cur[__pattern_length_-1]];
+        }
+        
+        return make_pair(__l, __l);
+    }
+};
+
+template<class _RandomAccessIterator, 
+         class _Hash = hash<typename iterator_traits<_RandomAccessIterator>::value_type>, 
+         class _BinaryPredicate = equal_to<>>
+_LIBCUDACXX_INLINE_VISIBILITY
+boyer_moore_horspool_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>
+make_boyer_moore_horspool_searcher( _RandomAccessIterator __f, _RandomAccessIterator __l, 
+                    _Hash __hf = _Hash(), _BinaryPredicate __p = _BinaryPredicate ())
+{
+    return boyer_moore_horspool_searcher<_RandomAccessIterator, _Hash, _BinaryPredicate>(__f, __l, __hf, __p);
+}
+
+#endif // _LIBCUDACXX_STD_VER > 11
+
+_LIBCUDACXX_END_NAMESPACE_LFTS
+
+_LIBCUDACXX_POP_MACROS
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_FUNCTIONAL */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/iterator b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/iterator
new file mode 100644
index 000000000000..acc8350c98e4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/iterator
@@ -0,0 +1,113 @@
+// -*- C++ -*-
+//===----------------------------- iterator -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_ITERATOR
+#define _LIBCUDACXX_EXPERIMENTAL_ITERATOR
+
+/*
+namespace std {
+  namespace experimental {
+    inline namespace fundamentals_v2 {
+
+    template <class DelimT, class charT = char, class traits = char_traits<charT>>
+        class ostream_joiner {
+        public:
+         typedef charT                        char_type;
+         typedef traits                       traits_type;
+         typedef basic_ostream<charT, traits> ostream_type;
+         typedef output_iterator_tag          iterator_category;
+         typedef void                         value_type;
+         typedef void                         difference_type;
+         typedef void                         pointer;
+         typedef void                         reference;
+      
+         ostream_joiner(ostream_type& s, const DelimT& delimiter);
+         ostream_joiner(ostream_type& s, DelimT&& delimiter);
+
+         template<typename T>  
+         ostream_joiner& operator=(const T& value);
+
+         ostream_joiner& operator*() noexcept;
+         ostream_joiner& operator++() noexcept;
+         ostream_joiner& operator++(int) noexcept;
+   private:
+      ostream_type* out_stream;   // exposition only 
+      DelimT delim;               // exposition only 
+      bool first_element;         // exposition only
+   };
+
+  template <class charT, class traits, class DelimT>
+    ostream_joiner<decay_t<DelimT>, charT, traits>
+    make_ostream_joiner(basic_ostream<charT, traits>& os, DelimT&& delimiter);
+
+    } // inline namespace fundamentals_v2
+  } // namespace experimental
+} // namespace std
+
+*/
+
+#include <experimental/__config>
+
+#if _LIBCUDACXX_STD_VER > 11
+
+#include <iterator>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS
+
+template <class _Delim, class _CharT = char, class _Traits = char_traits<_CharT>>
+class ostream_joiner {
+public:
+
+    typedef _CharT                               char_type;
+    typedef _Traits                              traits_type;
+    typedef basic_ostream<char_type,traits_type> ostream_type;
+    typedef output_iterator_tag                  iterator_category;
+    typedef void                                 value_type;
+    typedef void                                 difference_type;
+    typedef void                                 pointer;
+    typedef void                                 reference;
+
+    ostream_joiner(ostream_type& __os, _Delim&& __d)
+        : __output_iter(_CUDA_VSTD::addressof(__os)), __delim(_CUDA_VSTD::move(__d)), __first(true) {}
+        
+    ostream_joiner(ostream_type& __os, const _Delim& __d)
+        : __output_iter(_CUDA_VSTD::addressof(__os)), __delim(__d), __first(true) {}
+    
+
+    template<typename _Tp>
+    ostream_joiner& operator=(const _Tp& __v)
+    {
+        if (!__first)
+            *__output_iter << __delim;
+        __first = false;
+        *__output_iter << __v;
+        return *this;
+    }
+
+    ostream_joiner& operator*()     _NOEXCEPT { return *this; }
+    ostream_joiner& operator++()    _NOEXCEPT { return *this; }
+    ostream_joiner& operator++(int) _NOEXCEPT { return *this; }
+
+private:
+    ostream_type*   __output_iter;
+    _Delim          __delim;
+    bool            __first;
+};
+
+
+template <class _CharT, class _Traits, class _Delim>
+ostream_joiner<typename decay<_Delim>::type, _CharT, _Traits>
+make_ostream_joiner(basic_ostream<_CharT, _Traits>& __os, _Delim && __d)
+{ return ostream_joiner<typename decay<_Delim>::type, _CharT, _Traits>(__os, _CUDA_VSTD::forward<_Delim>(__d)); }
+
+_LIBCUDACXX_END_NAMESPACE_LFTS
+
+#endif /* _LIBCUDACXX_STD_VER > 11 */
+
+#endif // _LIBCUDACXX_EXPERIMENTAL_ITERATOR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list
new file mode 100644
index 000000000000..0600d27e65a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/list
@@ -0,0 +1,46 @@
+// -*- C++ -*-
+//===--------------------------- list ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_LIST
+#define _LIBCUDACXX_EXPERIMENTAL_LIST
+/*
+    experimental/list synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class T>
+  using list = std::list<T,polymorphic_allocator<T>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <list>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _ValueT>
+using list = _CUDA_VSTD::list<_ValueT, polymorphic_allocator<_ValueT>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_LIST */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map
new file mode 100644
index 000000000000..222836e18011
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/map
@@ -0,0 +1,56 @@
+// -*- C++ -*-
+//===----------------------------- map ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_MAP
+#define _LIBCUDACXX_EXPERIMENTAL_MAP
+/*
+    experimental/map synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class Key, class T, class Compare = less<Key>>
+  using map = std::map<Key, T, Compare,
+                       polymorphic_allocator<pair<const Key,T>>>;
+
+  template <class Key, class T, class Compare = less<Key>>
+  using multimap = std::multimap<Key, T, Compare,
+                                 polymorphic_allocator<pair<const Key,T>>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <map>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _Key, class _Value, class  _Compare = less<_Key>>
+using map = _CUDA_VSTD::map<_Key, _Value, _Compare,
+                        polymorphic_allocator<pair<const _Key, _Value>>>;
+
+template <class _Key, class _Value, class  _Compare = less<_Key>>
+using multimap = _CUDA_VSTD::multimap<_Key, _Value, _Compare,
+                        polymorphic_allocator<pair<const _Key, _Value>>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_MAP */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource
new file mode 100644
index 000000000000..843595dbf017
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/memory_resource
@@ -0,0 +1,426 @@
+// -*- C++ -*-
+//===------------------------ memory_resource -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_MEMORY_RESOURCE
+#define _LIBCUDACXX_EXPERIMENTAL_MEMORY_RESOURCE
+
+/**
+    experimental/memory_resource synopsis
+
+// C++1y
+
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  class memory_resource;
+
+  bool operator==(const memory_resource& a,
+                  const memory_resource& b) noexcept;
+  bool operator!=(const memory_resource& a,
+                  const memory_resource& b) noexcept;
+
+  template <class Tp> class polymorphic_allocator;
+
+  template <class T1, class T2>
+  bool operator==(const polymorphic_allocator<T1>& a,
+                  const polymorphic_allocator<T2>& b) noexcept;
+  template <class T1, class T2>
+  bool operator!=(const polymorphic_allocator<T1>& a,
+                  const polymorphic_allocator<T2>& b) noexcept;
+
+  // The name resource_adaptor_imp is for exposition only.
+  template <class Allocator> class resource_adaptor_imp;
+
+  template <class Allocator>
+    using resource_adaptor = resource_adaptor_imp<
+      allocator_traits<Allocator>::rebind_alloc<char>>;
+
+  // Global memory resources
+  memory_resource* new_delete_resource() noexcept;
+  memory_resource* null_memory_resource() noexcept;
+
+  // The default memory resource
+  memory_resource* set_default_resource(memory_resource* r) noexcept;
+  memory_resource* get_default_resource() noexcept;
+
+  // Standard memory resources
+  struct pool_options;
+  class synchronized_pool_resource;
+  class unsynchronized_pool_resource;
+  class monotonic_buffer_resource;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <experimental/__memory>
+#include <limits>
+#include <memory>
+#include <new>
+#include <stdexcept>
+#include <__tuple>
+#include <type_traits>
+#include <utility>
+#include <cstddef>
+#include <cstdlib>
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+// Round __s up to next multiple of __a.
+inline _LIBCUDACXX_INLINE_VISIBILITY
+size_t __aligned_allocation_size(size_t __s, size_t __a) _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s + __a > __s, "aligned allocation size overflows");
+    return (__s + __a - 1) & ~(__a - 1);
+}
+
+// 8.5, memory.resource
+class _LIBCUDACXX_TYPE_VIS memory_resource
+{
+    static const size_t __max_align = _LIBCUDACXX_ALIGNOF(max_align_t);
+
+// 8.5.2, memory.resource.public
+public:
+    virtual ~memory_resource() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void* allocate(size_t __bytes, size_t __align = __max_align)
+        { return do_allocate(__bytes, __align); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void deallocate(void * __p, size_t __bytes, size_t __align = __max_align)
+        { do_deallocate(__p, __bytes, __align); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_equal(memory_resource const & __other) const _NOEXCEPT
+        { return do_is_equal(__other); }
+
+// 8.5.3, memory.resource.priv
+protected:
+    virtual void* do_allocate(size_t, size_t) = 0;
+    virtual void do_deallocate(void*, size_t, size_t) = 0;
+    virtual bool do_is_equal(memory_resource const &) const _NOEXCEPT = 0;
+};
+
+// 8.5.4, memory.resource.eq
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(memory_resource const & __lhs,
+                memory_resource const & __rhs) _NOEXCEPT
+{
+    return &__lhs == &__rhs || __lhs.is_equal(__rhs);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(memory_resource const & __lhs,
+                memory_resource const & __rhs) _NOEXCEPT
+{
+    return !(__lhs == __rhs);
+}
+
+_LIBCUDACXX_FUNC_VIS
+memory_resource * new_delete_resource() _NOEXCEPT;
+
+_LIBCUDACXX_FUNC_VIS
+memory_resource * null_memory_resource() _NOEXCEPT;
+
+_LIBCUDACXX_FUNC_VIS
+memory_resource * get_default_resource() _NOEXCEPT;
+
+_LIBCUDACXX_FUNC_VIS
+memory_resource * set_default_resource(memory_resource * __new_res) _NOEXCEPT;
+
+// 8.6, memory.polymorphic.allocator.class
+
+// 8.6.1, memory.polymorphic.allocator.overview
+template <class _ValueType>
+class _LIBCUDACXX_TEMPLATE_VIS polymorphic_allocator
+{
+public:
+    typedef _ValueType value_type;
+
+    // 8.6.2, memory.polymorphic.allocator.ctor
+    _LIBCUDACXX_INLINE_VISIBILITY
+    polymorphic_allocator() _NOEXCEPT
+      : __res_(_CUDA_VSTD_LFTS_PMR::get_default_resource())
+    {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    polymorphic_allocator(memory_resource * __r) _NOEXCEPT
+      : __res_(__r)
+    {}
+
+    polymorphic_allocator(polymorphic_allocator const &) = default;
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    polymorphic_allocator(polymorphic_allocator<_Tp> const & __other) _NOEXCEPT
+      : __res_(__other.resource())
+    {}
+
+    polymorphic_allocator &
+    operator=(polymorphic_allocator const &) = delete;
+
+    // 8.6.3, memory.polymorphic.allocator.mem
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _ValueType* allocate(size_t __n) {
+        if (__n > __max_size()) {
+            __throw_length_error(
+                "std::experimental::pmr::polymorphic_allocator<T>::allocate(size_t n)"
+                " 'n' exceeds maximum supported size");
+        }
+        return static_cast<_ValueType*>(
+            __res_->allocate(__n * sizeof(_ValueType), _LIBCUDACXX_ALIGNOF(_ValueType))
+        );
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void deallocate(_ValueType * __p, size_t __n) _NOEXCEPT {
+        _LIBCUDACXX_ASSERT(__n <= __max_size(),
+                       "deallocate called for size which exceeds max_size()");
+        __res_->deallocate(__p, __n * sizeof(_ValueType), _LIBCUDACXX_ALIGNOF(_ValueType));
+    }
+
+    template <class _Tp, class ..._Ts>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void construct(_Tp* __p, _Ts &&... __args)
+    {
+        _CUDA_VSTD_LFTS::__lfts_user_alloc_construct(
+            __p, *this, _CUDA_VSTD::forward<_Ts>(__args)...
+          );
+    }
+
+    template <class _T1, class _T2, class ..._Args1, class ..._Args2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void construct(pair<_T1, _T2>* __p, piecewise_construct_t,
+                   tuple<_Args1...> __x, tuple<_Args2...> __y)
+    {
+        ::new ((void*)__p) pair<_T1, _T2>(piecewise_construct
+          , __transform_tuple(
+              typename __lfts_uses_alloc_ctor<
+                  _T1, polymorphic_allocator&, _Args1...
+              >::type()
+            , _CUDA_VSTD::move(__x)
+            , typename __make_tuple_indices<sizeof...(_Args1)>::type{}
+          )
+          , __transform_tuple(
+              typename __lfts_uses_alloc_ctor<
+                  _T2, polymorphic_allocator&, _Args2...
+              >::type()
+            , _CUDA_VSTD::move(__y)
+            , typename __make_tuple_indices<sizeof...(_Args2)>::type{}
+          )
+        );
+    }
+
+    template <class _T1, class _T2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void construct(pair<_T1, _T2>* __p) {
+        construct(__p, piecewise_construct, tuple<>(), tuple<>());
+    }
+
+    template <class _T1, class _T2, class _Up, class _Vp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void construct(pair<_T1, _T2> * __p, _Up && __u, _Vp && __v) {
+        construct(__p, piecewise_construct
+          , _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Up>(__u))
+          , _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Vp>(__v)));
+    }
+
+    template <class _T1, class _T2, class _U1, class _U2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void construct(pair<_T1, _T2> * __p, pair<_U1, _U2> const & __pr) {
+        construct(__p, piecewise_construct
+            , _CUDA_VSTD::forward_as_tuple(__pr.first)
+            , _CUDA_VSTD::forward_as_tuple(__pr.second));
+    }
+
+    template <class _T1, class _T2, class _U1, class _U2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void construct(pair<_T1, _T2> * __p, pair<_U1, _U2> && __pr){
+        construct(__p, piecewise_construct
+            , _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_U1>(__pr.first))
+            , _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_U2>(__pr.second)));
+    }
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void destroy(_Tp * __p) _NOEXCEPT
+        { __p->~_Tp(); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    polymorphic_allocator
+    select_on_container_copy_construction() const _NOEXCEPT
+        { return polymorphic_allocator(); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    memory_resource * resource() const _NOEXCEPT
+        { return __res_; }
+
+private:
+    template <class ..._Args, size_t ..._Idx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple<_Args&&...>
+    __transform_tuple(integral_constant<int, 0>, tuple<_Args...>&& __t,
+                      __tuple_indices<_Idx...>) const
+    {
+        return _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::get<_Idx>(_CUDA_VSTD::move(__t))...);
+    }
+
+    template <class ..._Args, size_t ..._Idx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple<allocator_arg_t const&, polymorphic_allocator&, _Args&&...>
+    __transform_tuple(integral_constant<int, 1>, tuple<_Args...> && __t,
+                      __tuple_indices<_Idx...>)
+    {
+        using _Tup = tuple<allocator_arg_t const&, polymorphic_allocator&, _Args&&...>;
+        return _Tup(allocator_arg, *this,
+                    _CUDA_VSTD::get<_Idx>(_CUDA_VSTD::move(__t))...);
+    }
+
+    template <class ..._Args, size_t ..._Idx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple<_Args&&..., polymorphic_allocator&>
+    __transform_tuple(integral_constant<int, 2>, tuple<_Args...> && __t,
+                      __tuple_indices<_Idx...>)
+    {
+        using _Tup = tuple<_Args&&..., polymorphic_allocator&>;
+        return _Tup(_CUDA_VSTD::get<_Idx>(_CUDA_VSTD::move(__t))..., *this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t __max_size() const _NOEXCEPT
+        { return numeric_limits<size_t>::max() / sizeof(value_type); }
+
+    memory_resource * __res_;
+};
+
+// 8.6.4, memory.polymorphic.allocator.eq
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(polymorphic_allocator<_Tp> const & __lhs,
+                polymorphic_allocator<_Up> const & __rhs) _NOEXCEPT
+{
+    return *__lhs.resource() == *__rhs.resource();
+}
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(polymorphic_allocator<_Tp> const & __lhs,
+                polymorphic_allocator<_Up> const & __rhs) _NOEXCEPT
+{
+    return !(__lhs == __rhs);
+}
+
+// 8.7, memory.resource.adaptor
+
+// 8.7.1, memory.resource.adaptor.overview
+template <class _CharAlloc>
+class _LIBCUDACXX_TEMPLATE_VIS __resource_adaptor_imp
+  : public memory_resource
+{
+    using _CTraits = allocator_traits<_CharAlloc>;
+    static_assert(is_same<typename _CTraits::value_type, char>::value
+               && is_same<typename _CTraits::pointer, char*>::value
+               && is_same<typename _CTraits::void_pointer, void*>::value, "");
+
+    static const size_t _MaxAlign = _LIBCUDACXX_ALIGNOF(max_align_t);
+
+    using _Alloc = typename _CTraits::template rebind_alloc<
+            typename aligned_storage<_MaxAlign, _MaxAlign>::type
+        >;
+
+    using _ValueType = typename _Alloc::value_type;
+
+    _Alloc __alloc_;
+
+public:
+    typedef _CharAlloc allocator_type;
+
+    __resource_adaptor_imp() = default;
+    __resource_adaptor_imp(__resource_adaptor_imp const &) = default;
+    __resource_adaptor_imp(__resource_adaptor_imp &&) = default;
+
+    // 8.7.2, memory.resource.adaptor.ctor
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __resource_adaptor_imp(allocator_type const & __a)
+      : __alloc_(__a)
+    {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __resource_adaptor_imp(allocator_type && __a)
+      : __alloc_(_CUDA_VSTD::move(__a))
+    {}
+
+    __resource_adaptor_imp &
+    operator=(__resource_adaptor_imp const &) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const
+    { return __alloc_; }
+
+// 8.7.3, memory.resource.adaptor.mem
+protected:
+    virtual void * do_allocate(size_t __bytes, size_t)
+    {
+        if (__bytes > __max_size()) {
+            __throw_length_error(
+                "std::experimental::pmr::resource_adaptor<T>::do_allocate(size_t bytes, size_t align)"
+                " 'bytes' exceeds maximum supported size");
+        }
+        size_t __s = __aligned_allocation_size(__bytes, _MaxAlign) / _MaxAlign;
+        return __alloc_.allocate(__s);
+    }
+
+    virtual void do_deallocate(void * __p, size_t __bytes, size_t)
+    {
+        _LIBCUDACXX_ASSERT(__bytes <= __max_size(),
+            "do_deallocate called for size which exceeds the maximum allocation size");
+        size_t __s = __aligned_allocation_size(__bytes, _MaxAlign) / _MaxAlign;
+        __alloc_.deallocate((_ValueType*)__p, __s);
+    }
+
+    virtual bool do_is_equal(memory_resource const & __other) const _NOEXCEPT {
+        __resource_adaptor_imp const * __p
+          = dynamic_cast<__resource_adaptor_imp const *>(&__other);
+        return __p  ? __alloc_ == __p->__alloc_ : false;
+    }
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t __max_size() const _NOEXCEPT {
+        return numeric_limits<size_t>::max() - _MaxAlign;
+    }
+};
+
+template <class _Alloc>
+using resource_adaptor = __resource_adaptor_imp<
+    typename allocator_traits<_Alloc>::template rebind_alloc<char>
+  >;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+_LIBCUDACXX_POP_MACROS
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_MEMORY_RESOURCE */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const
new file mode 100644
index 000000000000..feea80013494
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/propagate_const
@@ -0,0 +1,578 @@
+// -*- C++ -*-
+//===------------------------ propagate_const -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_PROPAGATE_CONST
+#define _LIBCUDACXX_EXPERIMENTAL_PROPAGATE_CONST
+/*
+    propagate_const synopsis
+
+    namespace std { namespace experimental { inline namespace fundamentals_v2 {
+
+    // [propagate_const]
+    template <class T> class propagate_const;
+
+    // [propagate_const.underlying], underlying pointer access
+    constexpr const _Tp& _CUDA_VSTD_LFTS_V2::get_underlying(const propagate_const<T>& pt) noexcept;
+    constexpr T& _CUDA_VSTD_LFTS_V2::get_underlying(propagate_const<T>& pt) noexcept;
+
+    // [propagate_const.relational], relational operators
+    template <class T> constexpr bool operator==(const propagate_const<T>& pt, nullptr_t);
+    template <class T> constexpr bool operator==(nullptr_t, const propagate_const<T>& pu);
+    template <class T> constexpr bool operator!=(const propagate_const<T>& pt, nullptr_t);
+    template <class T> constexpr bool operator!=(nullptr_t, const propagate_const<T>& pu);
+    template <class T, class U> constexpr bool operator==(const propagate_const<T>& pt, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator!=(const propagate_const<T>& pt, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator<(const propagate_const<T>& pt, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator>(const propagate_const<T>& pt, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator<=(const propagate_const<T>& pt, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator>=(const propagate_const<T>& pt, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator==(const propagate_const<T>& pt, const _Up& u);
+    template <class T, class U> constexpr bool operator!=(const propagate_const<T>& pt, const _Up& u);
+    template <class T, class U> constexpr bool operator<(const propagate_const<T>& pt, const _Up& u);
+    template <class T, class U> constexpr bool operator>(const propagate_const<T>& pt, const _Up& u);
+    template <class T, class U> constexpr bool operator<=(const propagate_const<T>& pt, const _Up& u);
+    template <class T, class U> constexpr bool operator>=(const propagate_const<T>& pt, const _Up& u);
+    template <class T, class U> constexpr bool operator==(const _Tp& t, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator!=(const _Tp& t, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator<(const _Tp& t, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator>(const _Tp& t, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator<=(const _Tp& t, const propagate_const<_Up>& pu);
+    template <class T, class U> constexpr bool operator>=(const _Tp& t, const propagate_const<_Up>& pu);
+
+    // [propagate_const.algorithms], specialized algorithms
+    template <class T> constexpr void swap(propagate_const<T>& pt, propagate_const<T>& pu) noexcept(see below);
+
+    template <class T>
+    class propagate_const
+    {
+
+    public:
+      typedef remove_reference_t<decltype(*declval<T&>())> element_type;
+
+      // [propagate_const.ctor], constructors
+      constexpr propagate_const() = default;
+      propagate_const(const propagate_const& p) = delete;
+      constexpr propagate_const(propagate_const&& p) = default;
+      template <class U> EXPLICIT constexpr propagate_const(propagate_const<_Up>&& pu); // see below
+      template <class U> EXPLICIT constexpr propagate_const(U&& u); // see below
+
+      // [propagate_const.assignment], assignment
+      propagate_const& operator=(const propagate_const& p) = delete;
+      constexpr propagate_const& operator=(propagate_const&& p) = default;
+      template <class U> constexpr propagate_const& operator=(propagate_const<_Up>&& pu);
+      template <class U> constexpr propagate_const& operator=(U&& u); // see below
+
+      // [propagate_const.const_observers], const observers
+      explicit constexpr operator bool() const;
+      constexpr const element_type* operator->() const;
+      constexpr operator const element_type*() const; // Not always defined
+      constexpr const element_type& operator*() const;
+      constexpr const element_type* get() const;
+
+      // [propagate_const.non_const_observers], non-const observers
+      constexpr element_type* operator->();
+      constexpr operator element_type*(); // Not always defined
+      constexpr element_type& operator*();
+      constexpr element_type* get();
+
+      // [propagate_const.modifiers], modifiers
+      constexpr void swap(propagate_const& pt) noexcept(see below)
+
+    private:
+      T t_; // exposition only
+    };
+
+  } // namespace fundamentals_v2
+  } // namespace experimental
+
+  // [propagate_const.hash], hash support
+  template <class T> struct hash<experimental::fundamentals_v2::propagate_const<T>>;
+
+  // [propagate_const.comparison_function_objects], comparison function objects
+  template <class T> struct equal_to<experimental::fundamentals_v2::propagate_const<T>>;
+  template <class T> struct not_equal_to<experimental::fundamentals_v2::propagate_const<T>>;
+  template <class T> struct less<experimental::fundamentals_v2::propagate_const<T>>;
+  template <class T> struct greater<experimental::fundamentals_v2::propagate_const<T>>;
+  template <class T> struct less_equal<experimental::fundamentals_v2::propagate_const<T>>;
+  template <class T> struct greater_equal<experimental::fundamentals_v2::propagate_const<T>>;
+
+} // namespace std
+
+*/
+
+#include <experimental/__config>
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11
+
+#include <type_traits>
+#include <utility>
+#include <functional>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_V2
+
+
+template <class _Tp>
+class propagate_const;
+
+template <class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Up& get_underlying(const propagate_const<_Up>& __pu) _NOEXCEPT;
+
+template <class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Up& get_underlying(propagate_const<_Up>& __pu) _NOEXCEPT;
+
+template <class _Tp>
+class propagate_const
+{
+public:
+  typedef remove_reference_t<decltype(*_CUDA_VSTD::declval<_Tp&>())> element_type;
+
+  static_assert(!is_array<_Tp>::value,
+      "Instantiation of propagate_const with an array type is ill-formed.");
+  static_assert(!is_reference<_Tp>::value,
+      "Instantiation of propagate_const with a reference type is ill-formed.");
+  static_assert(!(is_pointer<_Tp>::value && is_function<typename remove_pointer<_Tp>::type>::value),
+      "Instantiation of propagate_const with a function-pointer type is ill-formed.");
+  static_assert(!(is_pointer<_Tp>::value && is_same<typename remove_cv<typename remove_pointer<_Tp>::type>::type, void>::value),
+      "Instantiation of propagate_const with a pointer to (possibly cv-qualified) void is ill-formed.");
+
+private:
+  template <class _Up>
+  static _LIBCUDACXX_CONSTEXPR element_type* __get_pointer(_Up* __u)
+  {
+    return __u;
+  }
+
+  template <class _Up>
+  static _LIBCUDACXX_CONSTEXPR element_type* __get_pointer(_Up& __u)
+  {
+    return __get_pointer(__u.get());
+  }
+
+  template <class _Up>
+  static _LIBCUDACXX_CONSTEXPR const element_type* __get_pointer(const _Up* __u)
+  {
+    return __u;
+  }
+
+  template <class _Up>
+  static _LIBCUDACXX_CONSTEXPR const element_type* __get_pointer(const _Up& __u)
+  {
+    return __get_pointer(__u.get());
+  }
+
+  template <class _Up>
+  struct __is_propagate_const : false_type
+  {
+  };
+
+  template <class _Up>
+  struct __is_propagate_const<propagate_const<_Up>> : true_type
+  {
+  };
+
+  _Tp __t_;
+
+public:
+
+  template <class _Up> friend _LIBCUDACXX_CONSTEXPR const _Up& ::_CUDA_VSTD_LFTS_V2::get_underlying(const propagate_const<_Up>& __pu) _NOEXCEPT;
+  template <class _Up> friend _LIBCUDACXX_CONSTEXPR _Up& ::_CUDA_VSTD_LFTS_V2::get_underlying(propagate_const<_Up>& __pu) _NOEXCEPT;
+
+  _LIBCUDACXX_CONSTEXPR propagate_const() = default;
+
+  propagate_const(const propagate_const&) = delete;
+
+  _LIBCUDACXX_CONSTEXPR propagate_const(propagate_const&&) = default;
+
+  template <class _Up, enable_if_t<!is_convertible<_Up, _Tp>::value &&
+                                 is_constructible<_Tp, _Up&&>::value,bool> = true>
+  explicit _LIBCUDACXX_CONSTEXPR propagate_const(propagate_const<_Up>&& __pu)
+      : __t_(std::move(_CUDA_VSTD_LFTS_V2::get_underlying(__pu)))
+  {
+  }
+
+  template <class _Up, enable_if_t<is_convertible<_Up&&, _Tp>::value &&
+                                 is_constructible<_Tp, _Up&&>::value,bool> = false>
+  _LIBCUDACXX_CONSTEXPR propagate_const(propagate_const<_Up>&& __pu)
+      : __t_(std::move(_CUDA_VSTD_LFTS_V2::get_underlying(__pu)))
+  {
+  }
+
+  template <class _Up, enable_if_t<!is_convertible<_Up&&, _Tp>::value &&
+                                 is_constructible<_Tp, _Up&&>::value &&
+                                 !__is_propagate_const<decay_t<_Up>>::value,bool> = true>
+  explicit _LIBCUDACXX_CONSTEXPR propagate_const(_Up&& __u)
+      : __t_(std::forward<_Up>(__u))
+  {
+  }
+
+  template <class _Up, enable_if_t<is_convertible<_Up&&, _Tp>::value &&
+                                 is_constructible<_Tp, _Up&&>::value &&
+                                 !__is_propagate_const<decay_t<_Up>>::value,bool> = false>
+  _LIBCUDACXX_CONSTEXPR propagate_const(_Up&& __u)
+      : __t_(std::forward<_Up>(__u))
+  {
+  }
+
+  propagate_const& operator=(const propagate_const&) = delete;
+
+  _LIBCUDACXX_CONSTEXPR propagate_const& operator=(propagate_const&&) = default;
+
+  template <class _Up>
+  _LIBCUDACXX_CONSTEXPR propagate_const& operator=(propagate_const<_Up>&& __pu)
+  {
+    __t_ = std::move(_CUDA_VSTD_LFTS_V2::get_underlying(__pu));
+    return *this;
+  }
+
+  template <class _Up, class _Vp = enable_if_t<!__is_propagate_const<decay_t<_Up>>::value>>
+  _LIBCUDACXX_CONSTEXPR propagate_const& operator=(_Up&& __u)
+  {
+    __t_ = std::forward<_Up>(__u);
+    return *this;
+  }
+
+  _LIBCUDACXX_CONSTEXPR const element_type* get() const
+  {
+    return __get_pointer(__t_);
+  }
+
+  _LIBCUDACXX_CONSTEXPR element_type* get()
+  {
+    return __get_pointer(__t_);
+  }
+
+  explicit _LIBCUDACXX_CONSTEXPR operator bool() const
+  {
+    return get() != nullptr;
+  }
+
+  _LIBCUDACXX_CONSTEXPR const element_type* operator->() const
+  {
+    return get();
+  }
+
+  template <class _Tp_ = _Tp, class _Up = enable_if_t<is_convertible<
+                                  const _Tp_, const element_type *>::value>>
+  _LIBCUDACXX_CONSTEXPR operator const element_type *() const {
+    return get();
+  }
+
+  _LIBCUDACXX_CONSTEXPR const element_type& operator*() const
+  {
+    return *get();
+  }
+
+  _LIBCUDACXX_CONSTEXPR element_type* operator->()
+  {
+    return get();
+  }
+
+  template <class _Tp_ = _Tp, class _Up = enable_if_t<
+                                  is_convertible<_Tp_, element_type *>::value>>
+  _LIBCUDACXX_CONSTEXPR operator element_type *() {
+    return get();
+  }
+
+  _LIBCUDACXX_CONSTEXPR element_type& operator*()
+  {
+    return *get();
+  }
+
+  _LIBCUDACXX_CONSTEXPR void swap(propagate_const& __pt) _NOEXCEPT_(__is_nothrow_swappable<_Tp>::value)
+  {
+    using _CUDA_VSTD::swap;
+    swap(__t_, __pt.__t_);
+  }
+};
+
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator==(const propagate_const<_Tp>& __pt, nullptr_t)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) == nullptr;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator==(nullptr_t, const propagate_const<_Tp>& __pt)
+{
+  return nullptr == _CUDA_VSTD_LFTS_V2::get_underlying(__pt);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator!=(const propagate_const<_Tp>& __pt, nullptr_t)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) != nullptr;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator!=(nullptr_t, const propagate_const<_Tp>& __pt)
+{
+  return nullptr != _CUDA_VSTD_LFTS_V2::get_underlying(__pt);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator==(const propagate_const<_Tp>& __pt,
+                          const propagate_const<_Up>& __pu)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) == _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator!=(const propagate_const<_Tp>& __pt,
+                          const propagate_const<_Up>& __pu)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) != _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator<(const propagate_const<_Tp>& __pt,
+                         const propagate_const<_Up>& __pu)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) < _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator>(const propagate_const<_Tp>& __pt,
+                         const propagate_const<_Up>& __pu)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) > _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator<=(const propagate_const<_Tp>& __pt,
+                          const propagate_const<_Up>& __pu)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) <= _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator>=(const propagate_const<_Tp>& __pt,
+                          const propagate_const<_Up>& __pu)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) >= _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator==(const propagate_const<_Tp>& __pt, const _Up& __u)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) == __u;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator!=(const propagate_const<_Tp>& __pt, const _Up& __u)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) != __u;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator<(const propagate_const<_Tp>& __pt, const _Up& __u)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) < __u;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator>(const propagate_const<_Tp>& __pt, const _Up& __u)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) > __u;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator<=(const propagate_const<_Tp>& __pt, const _Up& __u)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) <= __u;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator>=(const propagate_const<_Tp>& __pt, const _Up& __u)
+{
+  return _CUDA_VSTD_LFTS_V2::get_underlying(__pt) >= __u;
+}
+
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator==(const _Tp& __t, const propagate_const<_Up>& __pu)
+{
+  return __t == _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator!=(const _Tp& __t, const propagate_const<_Up>& __pu)
+{
+  return __t != _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator<(const _Tp& __t, const propagate_const<_Up>& __pu)
+{
+  return __t < _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator>(const _Tp& __t, const propagate_const<_Up>& __pu)
+{
+  return __t > _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator<=(const _Tp& __t, const propagate_const<_Up>& __pu)
+{
+  return __t <= _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR bool operator>=(const _Tp& __t, const propagate_const<_Up>& __pu)
+{
+  return __t >= _CUDA_VSTD_LFTS_V2::get_underlying(__pu);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR void swap(propagate_const<_Tp>& __pc1, propagate_const<_Tp>& __pc2) _NOEXCEPT_(__is_nothrow_swappable<_Tp>::value)
+{
+  __pc1.swap(__pc2);
+}
+
+template <class _Tp>
+_LIBCUDACXX_CONSTEXPR const _Tp& get_underlying(const propagate_const<_Tp>& __pt) _NOEXCEPT
+{
+  return __pt.__t_;
+}
+
+template <class _Tp>
+_LIBCUDACXX_CONSTEXPR _Tp& get_underlying(propagate_const<_Tp>& __pt) _NOEXCEPT
+{
+  return __pt.__t_;
+}
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_V2
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp>
+struct hash<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef size_t result_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> argument_type;
+
+  size_t operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1) const
+  {
+    return std::hash<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1));
+  }
+};
+
+template <class _Tp>
+struct equal_to<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> first_argument_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> second_argument_type;
+
+  bool operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1,
+      const experimental::fundamentals_v2::propagate_const<_Tp>& __pc2) const
+  {
+    return std::equal_to<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1), _CUDA_VSTD_LFTS_V2::get_underlying(__pc2));
+  }
+};
+
+template <class _Tp>
+struct not_equal_to<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> first_argument_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> second_argument_type;
+
+  bool operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1,
+      const experimental::fundamentals_v2::propagate_const<_Tp>& __pc2) const
+  {
+    return std::not_equal_to<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1), _CUDA_VSTD_LFTS_V2::get_underlying(__pc2));
+  }
+};
+
+template <class _Tp>
+struct less<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> first_argument_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> second_argument_type;
+
+  bool operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1,
+      const experimental::fundamentals_v2::propagate_const<_Tp>& __pc2) const
+  {
+    return std::less<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1), _CUDA_VSTD_LFTS_V2::get_underlying(__pc2));
+  }
+};
+
+template <class _Tp>
+struct greater<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> first_argument_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> second_argument_type;
+
+  bool operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1,
+      const experimental::fundamentals_v2::propagate_const<_Tp>& __pc2) const
+  {
+    return std::greater<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1), _CUDA_VSTD_LFTS_V2::get_underlying(__pc2));
+  }
+};
+
+template <class _Tp>
+struct less_equal<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> first_argument_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> second_argument_type;
+
+  bool operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1,
+      const experimental::fundamentals_v2::propagate_const<_Tp>& __pc2) const
+  {
+    return std::less_equal<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1), _CUDA_VSTD_LFTS_V2::get_underlying(__pc2));
+  }
+};
+
+template <class _Tp>
+struct greater_equal<experimental::fundamentals_v2::propagate_const<_Tp>>
+{
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> first_argument_type;
+  typedef experimental::fundamentals_v2::propagate_const<_Tp> second_argument_type;
+
+  bool operator()(const experimental::fundamentals_v2::propagate_const<_Tp>& __pc1,
+      const experimental::fundamentals_v2::propagate_const<_Tp>& __pc2) const
+  {
+    return std::greater_equal<_Tp>()(_CUDA_VSTD_LFTS_V2::get_underlying(__pc1), _CUDA_VSTD_LFTS_V2::get_underlying(__pc2));
+  }
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // _LIBCUDACXX_STD_VER > 11
+#endif // _LIBCUDACXX_EXPERIMENTAL_PROPAGATE_CONST
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex
new file mode 100644
index 000000000000..3a7cd32ccecb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/regex
@@ -0,0 +1,61 @@
+// -*- C++ -*-
+//===----------------------------- regex ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_REGEX
+#define _LIBCUDACXX_EXPERIMENTAL_REGEX
+/*
+    experimental/regex synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class BidirectionalIterator>
+  using match_results =
+    std::match_results<BidirectionalIterator,
+                       polymorphic_allocator<sub_match<BidirectionalIterator>>>;
+
+  typedef match_results<const char*> cmatch;
+  typedef match_results<const wchar_t*> wcmatch;
+  typedef match_results<string::const_iterator> smatch;
+  typedef match_results<wstring::const_iterator> wsmatch;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <regex>
+#include <experimental/string>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _BiDirIter>
+using match_results =
+    _CUDA_VSTD::match_results<_BiDirIter,
+        polymorphic_allocator<_CUDA_VSTD::sub_match<_BiDirIter>>>;
+
+typedef match_results<const char*> cmatch;
+typedef match_results<const wchar_t*> wcmatch;
+typedef match_results<_CUDA_VSTD_LFTS_PMR::string::const_iterator> smatch;
+typedef match_results<_CUDA_VSTD_LFTS_PMR::wstring::const_iterator> wsmatch;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_REGEX */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set
new file mode 100644
index 000000000000..995d50a3d489
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/set
@@ -0,0 +1,56 @@
+// -*- C++ -*-
+//===--------------------------- list ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_SET
+#define _LIBCUDACXX_EXPERIMENTAL_SET
+/*
+    experimental/set synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class Key, class T, class Compare = less<Key>>
+  using set = std::set<Key, T, Compare,
+                       polymorphic_allocator<pair<const Key,T>>>;
+
+  template <class Key, class T, class Compare = less<Key>>
+  using multiset = std::multiset<Key, T, Compare,
+                                 polymorphic_allocator<pair<const Key,T>>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <set>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _Value, class  _Compare = less<_Value>>
+using set = _CUDA_VSTD::set<_Value, _Compare,
+                        polymorphic_allocator<_Value>>;
+
+template <class _Value, class  _Compare = less<_Value>>
+using multiset = _CUDA_VSTD::multiset<_Value, _Compare,
+                        polymorphic_allocator<_Value>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_SET */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd
new file mode 100644
index 000000000000..01fdbb8d3e41
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/simd
@@ -0,0 +1,1569 @@
+// -*- C++ -*-
+//===------------------------------- simd ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef _LIBCUDACXX_EXPERIMENTAL_SIMD
+#define _LIBCUDACXX_EXPERIMENTAL_SIMD
+
+/*
+    experimental/simd synopsis
+
+namespace std::experimental {
+
+inline namespace parallelism_v2 {
+
+namespace simd_abi {
+
+struct scalar {};
+template <int N> struct fixed_size {};
+template <typename T> inline constexpr int max_fixed_size = implementation-defined;
+template <typename T> using compatible = implementation-defined;
+template <typename T> using native = implementation-defined;
+
+} // simd_abi
+
+struct element_aligned_tag {};
+struct vector_aligned_tag {};
+template <size_t> struct overaligned_tag {};
+inline constexpr element_aligned_tag element_aligned{};
+inline constexpr vector_aligned_tag vector_aligned{};
+template <size_t N> inline constexpr overaligned_tag<N> overaligned{};
+
+// traits [simd.traits]
+template <class T> struct is_abi_tag;
+template <class T> inline constexpr bool is_abi_tag_v = is_abi_tag<T>::value;
+
+template <class T> struct is_simd;
+template <class T> inline constexpr bool is_simd_v = is_simd<T>::value;
+
+template <class T> struct is_simd_mask;
+template <class T> inline constexpr bool is_simd_mask_v = is_simd_mask<T>::value;
+
+template <class T> struct is_simd_flag_type;
+template <class T> inline constexpr bool is_simd_flag_type_v = is_simd_flag_type<T>::value;
+
+template <class T, size_t N> struct abi_for_size { using type = see below; };
+template <class T, size_t N> using abi_for_size_t = typename abi_for_size<T, N>::type;
+
+template <class T, class Abi = simd_abi::compatible<T>> struct simd_size;
+template <class T, class Abi = simd_abi::compatible<T>>
+inline constexpr size_t simd_size_v = simd_size<T, Abi>::value;
+
+template <class T, class U = typename T::value_type> struct memory_alignment;
+template <class T, class U = typename T::value_type>
+inline constexpr size_t memory_alignment_v = memory_alignment<T, U>::value;
+
+// class template simd [simd.class]
+template <class T, class Abi = simd_abi::compatible<T>> class simd;
+template <class T> using native_simd = simd<T, simd_abi::native<T>>;
+template <class T, int N> using fixed_size_simd = simd<T, simd_abi::fixed_size<N>>;
+
+// class template simd_mask [simd.mask.class]
+template <class T, class Abi = simd_abi::compatible<T>> class simd_mask;
+template <class T> using native_simd_mask = simd_mask<T, simd_abi::native<T>>;
+template <class T, int N> using fixed_size_simd_mask = simd_mask<T, simd_abi::fixed_size<N>>;
+
+// casts [simd.casts]
+template <class T, class U, class Abi> see below simd_cast(const simd<U, Abi>&);
+template <class T, class U, class Abi> see below static_simd_cast(const simd<U, Abi>&);
+
+template <class T, class Abi>
+fixed_size_simd<T, simd_size_v<T, Abi>> to_fixed_size(const simd<T, Abi>&) noexcept;
+template <class T, class Abi>
+fixed_size_simd_mask<T, simd_size_v<T, Abi>> to_fixed_size(const simd_mask<T, Abi>&) noexcept;
+template <class T, size_t N> native_simd<T> to_native(const fixed_size_simd<T, N>&) noexcept;
+template <class T, size_t N>
+native_simd_mask<T> to_native(const fixed_size_simd_mask<T, N>> &) noexcept;
+template <class T, size_t N> simd<T> to_compatible(const fixed_size_simd<T, N>&) noexcept;
+template <class T, size_t N> simd_mask<T> to_compatible(const fixed_size_simd_mask<T, N>&) noexcept;
+
+template <size_t... Sizes, class T, class Abi>
+tuple<simd<T, abi_for_size_t<Sizes>>...> split(const simd<T, Abi>&);
+template <size_t... Sizes, class T, class Abi>
+tuple<simd_mask<T, abi_for_size_t<Sizes>>...> split(const simd_mask<T, Abi>&);
+template <class V, class Abi>
+array<V, simd_size_v<typename V::value_type, Abi> / V::size()> split(
+const simd<typename V::value_type, Abi>&);
+template <class V, class Abi>
+array<V, simd_size_v<typename V::value_type, Abi> / V::size()> split(
+const simd_mask<typename V::value_type, Abi>&);
+
+template <class T, class... Abis>
+simd<T, abi_for_size_t<T, (simd_size_v<T, Abis> + ...)>> concat(const simd<T, Abis>&...);
+template <class T, class... Abis>
+simd_mask<T, abi_for_size_t<T, (simd_size_v<T, Abis> + ...)>> concat(const simd_mask<T, Abis>&...);
+
+// reductions [simd.mask.reductions]
+template <class T, class Abi> bool all_of(const simd_mask<T, Abi>&) noexcept;
+template <class T, class Abi> bool any_of(const simd_mask<T, Abi>&) noexcept;
+template <class T, class Abi> bool none_of(const simd_mask<T, Abi>&) noexcept;
+template <class T, class Abi> bool some_of(const simd_mask<T, Abi>&) noexcept;
+template <class T, class Abi> int popcount(const simd_mask<T, Abi>&) noexcept;
+template <class T, class Abi> int find_first_set(const simd_mask<T, Abi>&);
+template <class T, class Abi> int find_last_set(const simd_mask<T, Abi>&);
+
+bool all_of(see below) noexcept;
+bool any_of(see below) noexcept;
+bool none_of(see below) noexcept;
+bool some_of(see below) noexcept;
+int popcount(see below) noexcept;
+int find_first_set(see below) noexcept;
+int find_last_set(see below) noexcept;
+
+// masked assignment [simd.whereexpr]
+template <class M, class T> class const_where_expression;
+template <class M, class T> class where_expression;
+
+// masked assignment [simd.mask.where]
+template <class T> struct nodeduce { using type = T; }; // exposition only
+
+template <class T> using nodeduce_t = typename nodeduce<T>::type; // exposition only
+
+template <class T, class Abi>
+where_expression<simd_mask<T, Abi>, simd<T, Abi>>
+where(const typename simd<T, Abi>::mask_type&, simd<T, Abi>&) noexcept;
+
+template <class T, class Abi>
+const_where_expression<simd_mask<T, Abi>, const simd<T, Abi>>
+where(const typename simd<T, Abi>::mask_type&, const simd<T, Abi>&) noexcept;
+
+template <class T, class Abi>
+where_expression<simd_mask<T, Abi>, simd_mask<T, Abi>>
+where(const nodeduce_t<simd_mask<T, Abi>>&, simd_mask<T, Abi>&) noexcept;
+
+template <class T, class Abi>
+const_where_expression<simd_mask<T, Abi>, const simd_mask<T, Abi>>
+where(const nodeduce_t<simd_mask<T, Abi>>&, const simd_mask<T, Abi>&) noexcept;
+
+template <class T> where_expression<bool, T> where(see below k, T& d) noexcept;
+
+template <class T>
+const_where_expression<bool, const T> where(see below k, const T& d) noexcept;
+
+// reductions [simd.reductions]
+template <class T, class Abi, class BinaryOperation = std::plus<>>
+T reduce(const simd<T, Abi>&, BinaryOperation = BinaryOperation());
+
+template <class M, class V, class BinaryOperation>
+typename V::value_type reduce(const const_where_expression<M, V>& x,
+typename V::value_type neutral_element, BinaryOperation binary_op);
+
+template <class M, class V>
+typename V::value_type reduce(const const_where_expression<M, V>& x, plus<> binary_op = plus<>());
+
+template <class M, class V>
+typename V::value_type reduce(const const_where_expression<M, V>& x, multiplies<> binary_op);
+
+template <class M, class V>
+typename V::value_type reduce(const const_where_expression<M, V>& x, bit_and<> binary_op);
+
+template <class M, class V>
+typename V::value_type reduce(const const_where_expression<M, V>& x, bit_or<> binary_op);
+
+template <class M, class V>
+typename V::value_type reduce(const const_where_expression<M, V>& x, bit_xor<> binary_op);
+
+template <class T, class Abi> T hmin(const simd<T, Abi>&);
+template <class M, class V> T hmin(const const_where_expression<M, V>&);
+template <class T, class Abi> T hmax(const simd<T, Abi>&);
+template <class M, class V> T hmax(const const_where_expression<M, V>&);
+
+// algorithms [simd.alg]
+template <class T, class Abi> simd<T, Abi> min(const simd<T, Abi>&, const simd<T, Abi>&) noexcept;
+
+template <class T, class Abi> simd<T, Abi> max(const simd<T, Abi>&, const simd<T, Abi>&) noexcept;
+
+template <class T, class Abi>
+std::pair<simd<T, Abi>, simd<T, Abi>> minmax(const simd<T, Abi>&, const simd<T, Abi>&) noexcept;
+
+template <class T, class Abi>
+simd<T, Abi> clamp(const simd<T, Abi>& v, const simd<T, Abi>& lo, const simd<T, Abi>& hi);
+
+// [simd.whereexpr]
+template <class M, class T>
+class const_where_expression {
+  const M& mask; // exposition only
+  T& data; // exposition only
+public:
+  const_where_expression(const const_where_expression&) = delete;
+  const_where_expression& operator=(const const_where_expression&) = delete;
+  remove_const_t<T> operator-() const &&;
+  template <class U, class Flags> void copy_to(U* mem, Flags f) const &&;
+};
+
+template <class M, class T>
+class where_expression : public const_where_expression<M, T> {
+public:
+  where_expression(const where_expression&) = delete;
+  where_expression& operator=(const where_expression&) = delete;
+  template <class U> void operator=(U&& x);
+  template <class U> void operator+=(U&& x);
+  template <class U> void operator-=(U&& x);
+  template <class U> void operator*=(U&& x);
+  template <class U> void operator/=(U&& x);
+  template <class U> void operator%=(U&& x);
+  template <class U> void operator&=(U&& x);
+  template <class U> void operator|=(U&& x);
+  template <class U> void operator^=(U&& x);
+  template <class U> void operator<<=(U&& x);
+  template <class U> void operator>>=(U&& x);
+  void operator++();
+  void operator++(int);
+  void operator--();
+  void operator--(int);
+  template <class U, class Flags> void copy_from(const U* mem, Flags);
+};
+
+// [simd.class]
+template <class T, class Abi> class simd {
+public:
+  using value_type = T;
+  using reference = see below;
+  using mask_type = simd_mask<T, Abi>;
+
+  using abi_type = Abi;
+  static constexpr size_t size() noexcept;
+  simd() = default;
+
+  // implicit type conversion constructor
+  template <class U> simd(const simd<U, simd_abi::fixed_size<size()>>&);
+
+  // implicit broadcast constructor (see below for constraints)
+  template <class U> simd(U&& value);
+
+  // generator constructor (see below for constraints)
+  template <class G> explicit simd(G&& gen);
+
+  // load constructor
+  template <class U, class Flags> simd(const U* mem, Flags f);
+
+  // loads [simd.load]
+  template <class U, class Flags> void copy_from(const U* mem, Flags f);
+
+  // stores [simd.store]
+  template <class U, class Flags> void copy_to(U* mem, Flags f) const;
+
+  // scalar access [simd.subscr]
+  reference operator[](size_t);
+  value_type operator[](size_t) const;
+
+  // unary operators [simd.unary]
+  simd& operator++();
+  simd operator++(int);
+  simd& operator--();
+  simd operator--(int);
+  mask_type operator!() const;
+  simd operator~() const; // see below
+  simd operator+() const;
+  simd operator-() const;
+
+  // binary operators [simd.binary]
+  friend simd operator+ (const simd&, const simd&);
+  friend simd operator- (const simd&, const simd&);
+  friend simd operator* (const simd&, const simd&);
+  friend simd operator/ (const simd&, const simd&);
+  friend simd operator% (const simd&, const simd&);
+  friend simd operator& (const simd&, const simd&);
+  friend simd operator| (const simd&, const simd&);
+  friend simd operator^ (const simd&, const simd&);
+  friend simd operator<<(const simd&, const simd&);
+  friend simd operator>>(const simd&, const simd&);
+  friend simd operator<<(const simd&, int);
+  friend simd operator>>(const simd&, int);
+
+  // compound assignment [simd.cassign]
+  friend simd& operator+= (simd&, const simd&);
+  friend simd& operator-= (simd&, const simd&);
+  friend simd& operator*= (simd&, const simd&);
+  friend simd& operator/= (simd&, const simd&);
+  friend simd& operator%= (simd&, const simd&);
+
+  friend simd& operator&= (simd&, const simd&);
+  friend simd& operator|= (simd&, const simd&);
+  friend simd& operator^= (simd&, const simd&);
+  friend simd& operator<<=(simd&, const simd&);
+  friend simd& operator>>=(simd&, const simd&);
+  friend simd& operator<<=(simd&, int);
+  friend simd& operator>>=(simd&, int);
+
+  // compares [simd.comparison]
+  friend mask_type operator==(const simd&, const simd&);
+  friend mask_type operator!=(const simd&, const simd&);
+  friend mask_type operator>=(const simd&, const simd&);
+  friend mask_type operator<=(const simd&, const simd&);
+  friend mask_type operator> (const simd&, const simd&);
+  friend mask_type operator< (const simd&, const simd&);
+};
+
+// [simd.math]
+template <class Abi> using scharv = simd<signed char, Abi>; // exposition only
+template <class Abi> using shortv = simd<short, Abi>; // exposition only
+template <class Abi> using intv = simd<int, Abi>; // exposition only
+template <class Abi> using longv = simd<long int, Abi>; // exposition only
+template <class Abi> using llongv = simd<long long int, Abi>; // exposition only
+template <class Abi> using floatv = simd<float, Abi>; // exposition only
+template <class Abi> using doublev = simd<double, Abi>; // exposition only
+template <class Abi> using ldoublev = simd<long double, Abi>; // exposition only
+template <class T, class V> using samesize = fixed_size_simd<T, V::size()>; // exposition only
+
+template <class Abi> floatv<Abi> acos(floatv<Abi> x);
+template <class Abi> doublev<Abi> acos(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> acos(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> asin(floatv<Abi> x);
+template <class Abi> doublev<Abi> asin(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> asin(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> atan(floatv<Abi> x);
+template <class Abi> doublev<Abi> atan(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> atan(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> atan2(floatv<Abi> y, floatv<Abi> x);
+template <class Abi> doublev<Abi> atan2(doublev<Abi> y, doublev<Abi> x);
+template <class Abi> ldoublev<Abi> atan2(ldoublev<Abi> y, ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> cos(floatv<Abi> x);
+template <class Abi> doublev<Abi> cos(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> cos(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> sin(floatv<Abi> x);
+template <class Abi> doublev<Abi> sin(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> sin(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> tan(floatv<Abi> x);
+template <class Abi> doublev<Abi> tan(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> tan(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> acosh(floatv<Abi> x);
+template <class Abi> doublev<Abi> acosh(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> acosh(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> asinh(floatv<Abi> x);
+template <class Abi> doublev<Abi> asinh(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> asinh(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> atanh(floatv<Abi> x);
+template <class Abi> doublev<Abi> atanh(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> atanh(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> cosh(floatv<Abi> x);
+template <class Abi> doublev<Abi> cosh(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> cosh(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> sinh(floatv<Abi> x);
+template <class Abi> doublev<Abi> sinh(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> sinh(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> tanh(floatv<Abi> x);
+template <class Abi> doublev<Abi> tanh(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> tanh(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> exp(floatv<Abi> x);
+template <class Abi> doublev<Abi> exp(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> exp(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> exp2(floatv<Abi> x);
+template <class Abi> doublev<Abi> exp2(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> exp2(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> expm1(floatv<Abi> x);
+template <class Abi> doublev<Abi> expm1(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> expm1(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> frexp(floatv<Abi> value, samesize<int, floatv<Abi>>* exp);
+template <class Abi> doublev<Abi> frexp(doublev<Abi> value, samesize<int, doublev<Abi>>* exp);
+template <class Abi> ldoublev<Abi> frexp(ldoublev<Abi> value, samesize<int, ldoublev<Abi>>* exp);
+
+template <class Abi> samesize<int, floatv<Abi>> ilogb(floatv<Abi> x);
+template <class Abi> samesize<int, doublev<Abi>> ilogb(doublev<Abi> x);
+template <class Abi> samesize<int, ldoublev<Abi>> ilogb(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> ldexp(floatv<Abi> x, samesize<int, floatv<Abi>> exp);
+template <class Abi> doublev<Abi> ldexp(doublev<Abi> x, samesize<int, doublev<Abi>> exp);
+template <class Abi> ldoublev<Abi> ldexp(ldoublev<Abi> x, samesize<int, ldoublev<Abi>> exp);
+
+template <class Abi> floatv<Abi> log(floatv<Abi> x);
+template <class Abi> doublev<Abi> log(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> log(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> log10(floatv<Abi> x);
+template <class Abi> doublev<Abi> log10(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> log10(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> log1p(floatv<Abi> x);
+template <class Abi> doublev<Abi> log1p(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> log1p(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> log2(floatv<Abi> x);
+template <class Abi> doublev<Abi> log2(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> log2(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> logb(floatv<Abi> x);
+template <class Abi> doublev<Abi> logb(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> logb(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> modf(floatv<Abi> value, floatv<Abi>* iptr);
+template <class Abi> doublev<Abi> modf(doublev<Abi> value, doublev<Abi>* iptr);
+template <class Abi> ldoublev<Abi> modf(ldoublev<Abi> value, ldoublev<Abi>* iptr);
+
+template <class Abi> floatv<Abi> scalbn(floatv<Abi> x, samesize<int, floatv<Abi>> n);
+template <class Abi> doublev<Abi> scalbn(doublev<Abi> x, samesize<int, doublev<Abi>> n);
+template <class Abi> ldoublev<Abi> scalbn(ldoublev<Abi> x, samesize<int, ldoublev<Abi>> n);
+template <class Abi> floatv<Abi> scalbln(floatv<Abi> x, samesize<long int, floatv<Abi>> n);
+template <class Abi> doublev<Abi> scalbln(doublev<Abi> x, samesize<long int, doublev<Abi>> n);
+template <class Abi> ldoublev<Abi> scalbln(ldoublev<Abi> x, samesize<long int, ldoublev<Abi>> n);
+
+template <class Abi> floatv<Abi> cbrt(floatv<Abi> x);
+template <class Abi> doublev<Abi> cbrt(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> cbrt(ldoublev<Abi> x);
+
+template <class Abi> scharv<Abi> abs(scharv<Abi> j);
+template <class Abi> shortv<Abi> abs(shortv<Abi> j);
+template <class Abi> intv<Abi> abs(intv<Abi> j);
+template <class Abi> longv<Abi> abs(longv<Abi> j);
+template <class Abi> llongv<Abi> abs(llongv<Abi> j);
+template <class Abi> floatv<Abi> abs(floatv<Abi> j);
+template <class Abi> doublev<Abi> abs(doublev<Abi> j);
+template <class Abi> ldoublev<Abi> abs(ldoublev<Abi> j);
+
+template <class Abi> floatv<Abi> hypot(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> hypot(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> hypot(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> floatv<Abi> hypot(floatv<Abi> x, floatv<Abi> y, floatv<Abi> z);
+template <class Abi> doublev<Abi> hypot(doublev<Abi> x, doublev<Abi> y, doublev<Abi> z);
+template <class Abi> ldoublev<Abi> hypot(ldoublev<Abi> x, ldoublev<Abi> y, ldoublev<Abi> z);
+
+template <class Abi> floatv<Abi> pow(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> pow(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> pow(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> sqrt(floatv<Abi> x);
+template <class Abi> doublev<Abi> sqrt(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> sqrt(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> erf(floatv<Abi> x);
+template <class Abi> doublev<Abi> erf(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> erf(ldoublev<Abi> x);
+template <class Abi> floatv<Abi> erfc(floatv<Abi> x);
+template <class Abi> doublev<Abi> erfc(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> erfc(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> lgamma(floatv<Abi> x);
+template <class Abi> doublev<Abi> lgamma(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> lgamma(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> tgamma(floatv<Abi> x);
+template <class Abi> doublev<Abi> tgamma(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> tgamma(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> ceil(floatv<Abi> x);
+template <class Abi> doublev<Abi> ceil(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> ceil(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> floor(floatv<Abi> x);
+template <class Abi> doublev<Abi> floor(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> floor(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> nearbyint(floatv<Abi> x);
+template <class Abi> doublev<Abi> nearbyint(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> nearbyint(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> rint(floatv<Abi> x);
+template <class Abi> doublev<Abi> rint(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> rint(ldoublev<Abi> x);
+
+template <class Abi> samesize<long int, floatv<Abi>> lrint(floatv<Abi> x);
+template <class Abi> samesize<long int, doublev<Abi>> lrint(doublev<Abi> x);
+template <class Abi> samesize<long int, ldoublev<Abi>> lrint(ldoublev<Abi> x);
+template <class Abi> samesize<long long int, floatv<Abi>> llrint(floatv<Abi> x);
+template <class Abi> samesize<long long int, doublev<Abi>> llrint(doublev<Abi> x);
+template <class Abi> samesize<long long int, ldoublev<Abi>> llrint(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> round(floatv<Abi> x);
+template <class Abi> doublev<Abi> round(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> round(ldoublev<Abi> x);
+template <class Abi> samesize<long int, floatv<Abi>> lround(floatv<Abi> x);
+template <class Abi> samesize<long int, doublev<Abi>> lround(doublev<Abi> x);
+template <class Abi> samesize<long int, ldoublev<Abi>> lround(ldoublev<Abi> x);
+template <class Abi> samesize<long long int, floatv<Abi>> llround(floatv<Abi> x);
+template <class Abi> samesize<long long int, doublev<Abi>> llround(doublev<Abi> x);
+template <class Abi> samesize<long long int, ldoublev<Abi>> llround(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> trunc(floatv<Abi> x);
+template <class Abi> doublev<Abi> trunc(doublev<Abi> x);
+template <class Abi> ldoublev<Abi> trunc(ldoublev<Abi> x);
+
+template <class Abi> floatv<Abi> fmod(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> fmod(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> fmod(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> remainder(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> remainder(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> remainder(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> remquo(floatv<Abi> x, floatv<Abi> y, samesize<int, floatv<Abi>>* quo);
+template <class Abi> doublev<Abi> remquo(doublev<Abi> x, doublev<Abi> y, samesize<int, doublev<Abi>>* quo);
+template <class Abi> ldoublev<Abi> remquo(ldoublev<Abi> x, ldoublev<Abi> y, samesize<int, ldoublev<Abi>>* quo);
+
+template <class Abi> floatv<Abi> copysign(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> copysign(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> copysign(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> doublev<Abi> nan(const char* tagp);
+template <class Abi> floatv<Abi> nanf(const char* tagp);
+template <class Abi> ldoublev<Abi> nanl(const char* tagp);
+
+template <class Abi> floatv<Abi> nextafter(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> nextafter(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> nextafter(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> nexttoward(floatv<Abi> x, ldoublev<Abi> y);
+template <class Abi> doublev<Abi> nexttoward(doublev<Abi> x, ldoublev<Abi> y);
+template <class Abi> ldoublev<Abi> nexttoward(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> fdim(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> fdim(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> fdim(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> fmax(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> fmax(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> fmax(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> fmin(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> doublev<Abi> fmin(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> ldoublev<Abi> fmin(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> floatv<Abi> fma(floatv<Abi> x, floatv<Abi> y, floatv<Abi> z);
+template <class Abi> doublev<Abi> fma(doublev<Abi> x, doublev<Abi> y, doublev<Abi> z);
+template <class Abi> ldoublev<Abi> fma(ldoublev<Abi> x, ldoublev<Abi> y, ldoublev<Abi> z);
+
+template <class Abi> samesize<int, floatv<Abi>> fpclassify(floatv<Abi> x);
+template <class Abi> samesize<int, doublev<Abi>> fpclassify(doublev<Abi> x);
+template <class Abi> samesize<int, ldoublev<Abi>> fpclassify(ldoublev<Abi> x);
+
+template <class Abi> simd_mask<float, Abi> isfinite(floatv<Abi> x);
+template <class Abi> simd_mask<double, Abi> isfinite(doublev<Abi> x);
+template <class Abi> simd_mask<long double, Abi> isfinite(ldoublev<Abi> x);
+
+template <class Abi> simd_mask<float, Abi> isinf(floatv<Abi> x);
+template <class Abi> simd_mask<double, Abi> isinf(doublev<Abi> x);
+template <class Abi> simd_mask<long double, Abi> isinf(ldoublev<Abi> x);
+
+template <class Abi> simd_mask<float, Abi> isnan(floatv<Abi> x);
+template <class Abi> simd_mask<double, Abi> isnan(doublev<Abi> x);
+template <class Abi> simd_mask<long double, Abi> isnan(ldoublev<Abi> x);
+
+template <class Abi> simd_mask<float, Abi> isnormal(floatv<Abi> x);
+template <class Abi> simd_mask<double, Abi> isnormal(doublev<Abi> x);
+template <class Abi> simd_mask<long double, Abi> isnormal(ldoublev<Abi> x);
+
+template <class Abi> simd_mask<float, Abi> signbit(floatv<Abi> x);
+template <class Abi> simd_mask<double, Abi> signbit(doublev<Abi> x);
+template <class Abi> simd_mask<long double, Abi> signbit(ldoublev<Abi> x);
+
+template <class Abi> simd_mask<float, Abi> isgreater(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> simd_mask<double, Abi> isgreater(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> simd_mask<long double, Abi> isgreater(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> simd_mask<float, Abi> isgreaterequal(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> simd_mask<double, Abi> isgreaterequal(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> simd_mask<long double, Abi> isgreaterequal(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> simd_mask<float, Abi> isless(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> simd_mask<double, Abi> isless(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> simd_mask<long double, Abi> isless(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> simd_mask<float, Abi> islessequal(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> simd_mask<double, Abi> islessequal(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> simd_mask<long double, Abi> islessequal(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> simd_mask<float, Abi> islessgreater(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> simd_mask<double, Abi> islessgreater(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> simd_mask<long double, Abi> islessgreater(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class Abi> simd_mask<float, Abi> isunordered(floatv<Abi> x, floatv<Abi> y);
+template <class Abi> simd_mask<double, Abi> isunordered(doublev<Abi> x, doublev<Abi> y);
+template <class Abi> simd_mask<long double, Abi> isunordered(ldoublev<Abi> x, ldoublev<Abi> y);
+
+template <class V> struct simd_div_t { V quot, rem; };
+template <class Abi> simd_div_t<scharv<Abi>> div(scharv<Abi> numer, scharv<Abi> denom);
+template <class Abi> simd_div_t<shortv<Abi>> div(shortv<Abi> numer, shortv<Abi> denom);
+template <class Abi> simd_div_t<intv<Abi>> div(intv<Abi> numer, intv<Abi> denom);
+template <class Abi> simd_div_t<longv<Abi>> div(longv<Abi> numer, longv<Abi> denom);
+template <class Abi> simd_div_t<llongv<Abi>> div(llongv<Abi> numer, llongv<Abi> denom);
+
+// [simd.mask.class]
+template <class T, class Abi>
+class simd_mask {
+public:
+  using value_type = bool;
+  using reference = see below;
+  using simd_type = simd<T, Abi>;
+  using abi_type = Abi;
+  static constexpr size_t size() noexcept;
+  simd_mask() = default;
+
+  // broadcast constructor
+  explicit simd_mask(value_type) noexcept;
+
+  // implicit type conversion constructor
+  template <class U> simd_mask(const simd_mask<U, simd_abi::fixed_size<size()>>&) noexcept;
+
+  // load constructor
+  template <class Flags> simd_mask(const value_type* mem, Flags);
+
+  // loads [simd.mask.copy]
+  template <class Flags> void copy_from(const value_type* mem, Flags);
+  template <class Flags> void copy_to(value_type* mem, Flags) const;
+
+  // scalar access [simd.mask.subscr]
+  reference operator[](size_t);
+  value_type operator[](size_t) const;
+
+  // unary operators [simd.mask.unary]
+  simd_mask operator!() const noexcept;
+
+  // simd_mask binary operators [simd.mask.binary]
+  friend simd_mask operator&&(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator||(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator& (const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator| (const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator^ (const simd_mask&, const simd_mask&) noexcept;
+
+  // simd_mask compound assignment [simd.mask.cassign]
+  friend simd_mask& operator&=(simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask& operator|=(simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask& operator^=(simd_mask&, const simd_mask&) noexcept;
+
+  // simd_mask compares [simd.mask.comparison]
+  friend simd_mask operator==(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator!=(const simd_mask&, const simd_mask&) noexcept;
+};
+
+} // parallelism_v2
+} // std::experimental
+
+*/
+
+#include <experimental/__config>
+#include <algorithm>
+#include <array>
+#include <cstddef>
+#include <functional>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD
+
+#if _LIBCUDACXX_STD_VER >= 17
+
+enum class _StorageKind {
+  _Scalar,
+  _Array,
+  _VecExt,
+};
+
+template <_StorageKind __kind, int _Np>
+struct __simd_abi {};
+
+template <class _Tp, class _Abi>
+class __simd_storage {};
+
+template <class _Tp, int __num_element>
+class __simd_storage<_Tp, __simd_abi<_StorageKind::_Array, __num_element>> {
+  std::array<_Tp, __num_element> __storage_;
+
+  template <class, class>
+  friend struct simd;
+
+  template <class, class>
+  friend struct simd_mask;
+
+public:
+  _Tp __get(size_t __index) const noexcept { return __storage_[__index]; };
+  void __set(size_t __index, _Tp __val) noexcept {
+    __storage_[__index] = __val;
+  }
+};
+
+template <class _Tp>
+class __simd_storage<_Tp, __simd_abi<_StorageKind::_Scalar, 1>> {
+  _Tp __storage_;
+
+  template <class, class>
+  friend struct simd;
+
+  template <class, class>
+  friend struct simd_mask;
+
+public:
+  _Tp __get(size_t __index) const noexcept { return (&__storage_)[__index]; };
+  void __set(size_t __index, _Tp __val) noexcept {
+    (&__storage_)[__index] = __val;
+  }
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_VECTOR_EXTENSION
+
+constexpr size_t __floor_pow_of_2(size_t __val) {
+  return ((__val - 1) & __val) == 0 ? __val
+                                    : __floor_pow_of_2((__val - 1) & __val);
+}
+
+constexpr size_t __ceil_pow_of_2(size_t __val) {
+  return __val == 1 ? 1 : __floor_pow_of_2(__val - 1) << 1;
+}
+
+template <class _Tp, size_t __bytes>
+struct __vec_ext_traits {
+#if !defined(_LIBCUDACXX_COMPILER_CLANG)
+  typedef _Tp type __attribute__((vector_size(__ceil_pow_of_2(__bytes))));
+#endif
+};
+
+#if defined(_LIBCUDACXX_COMPILER_CLANG)
+#define _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, _NUM_ELEMENT)                        \
+  template <>                                                                  \
+  struct __vec_ext_traits<_TYPE, sizeof(_TYPE) * _NUM_ELEMENT> {               \
+    using type =                                                               \
+        _TYPE __attribute__((vector_size(sizeof(_TYPE) * _NUM_ELEMENT)));      \
+  }
+
+#define _LIBCUDACXX_SPECIALIZE_VEC_EXT_32(_TYPE)                                   \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 1);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 2);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 3);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 4);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 5);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 6);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 7);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 8);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 9);                                        \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 10);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 11);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 12);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 13);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 14);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 15);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 16);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 17);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 18);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 19);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 20);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 21);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 22);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 23);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 24);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 25);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 26);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 27);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 28);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 29);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 30);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 31);                                       \
+  _LIBCUDACXX_SPECIALIZE_VEC_EXT(_TYPE, 32);
+
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(char);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(char16_t);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(char32_t);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(wchar_t);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(signed char);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(signed short);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(signed int);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(signed long);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(signed long long);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(unsigned char);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(unsigned short);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(unsigned int);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(unsigned long);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(unsigned long long);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(float);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(double);
+_LIBCUDACXX_SPECIALIZE_VEC_EXT_32(long double);
+
+#undef _LIBCUDACXX_SPECIALIZE_VEC_EXT_32
+#undef _LIBCUDACXX_SPECIALIZE_VEC_EXT
+#endif
+
+template <class _Tp, int __num_element>
+class __simd_storage<_Tp, __simd_abi<_StorageKind::_VecExt, __num_element>> {
+  using _StorageType =
+      typename __vec_ext_traits<_Tp, sizeof(_Tp) * __num_element>::type;
+
+  _StorageType __storage_;
+
+  template <class, class>
+  friend struct simd;
+
+  template <class, class>
+  friend struct simd_mask;
+
+public:
+  _Tp __get(size_t __index) const noexcept { return __storage_[__index]; };
+  void __set(size_t __index, _Tp __val) noexcept {
+    __storage_[__index] = __val;
+  }
+};
+
+#endif // _LIBCUDACXX_HAS_NO_VECTOR_EXTENSION
+
+template <class _Vp, class _Tp, class _Abi>
+class __simd_reference {
+  static_assert(std::is_same<_Vp, _Tp>::value, "");
+
+  template <class, class>
+  friend struct simd;
+
+  template <class, class>
+  friend struct simd_mask;
+
+  __simd_storage<_Tp, _Abi>* __ptr_;
+  size_t __index_;
+
+  __simd_reference(__simd_storage<_Tp, _Abi>* __ptr, size_t __index)
+      : __ptr_(__ptr), __index_(__index) {}
+
+  __simd_reference(const __simd_reference&) = default;
+
+public:
+  __simd_reference() = delete;
+  __simd_reference& operator=(const __simd_reference&) = delete;
+
+  operator _Vp() const { return __ptr_->__get(__index_); }
+
+  __simd_reference operator=(_Vp __value) && {
+    __ptr_->__set(__index_, __value);
+    return *this;
+  }
+
+  __simd_reference operator++() && {
+    return std::move(*this) = __ptr_->__get(__index_) + 1;
+  }
+
+  _Vp operator++(int) && {
+    auto __val = __ptr_->__get(__index_);
+    __ptr_->__set(__index_, __val + 1);
+    return __val;
+  }
+
+  __simd_reference operator--() && {
+    return std::move(*this) = __ptr_->__get(__index_) - 1;
+  }
+
+  _Vp operator--(int) && {
+    auto __val = __ptr_->__get(__index_);
+    __ptr_->__set(__index_, __val - 1);
+    return __val;
+  }
+
+  __simd_reference operator+=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) + __value;
+  }
+
+  __simd_reference operator-=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) - __value;
+  }
+
+  __simd_reference operator*=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) * __value;
+  }
+
+  __simd_reference operator/=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) / __value;
+  }
+
+  __simd_reference operator%=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) % __value;
+  }
+
+  __simd_reference operator>>=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) >> __value;
+  }
+
+  __simd_reference operator<<=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) << __value;
+  }
+
+  __simd_reference operator&=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) & __value;
+  }
+
+  __simd_reference operator|=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) | __value;
+  }
+
+  __simd_reference operator^=(_Vp __value) && {
+    return std::move(*this) = __ptr_->__get(__index_) ^ __value;
+  }
+};
+
+template <class _To, class _From>
+constexpr decltype(_To{std::declval<_From>()}, true)
+__is_non_narrowing_convertible_impl(_From) {
+  return true;
+}
+
+template <class _To>
+constexpr bool __is_non_narrowing_convertible_impl(...) {
+  return false;
+}
+
+template <class _From, class _To>
+constexpr typename std::enable_if<std::is_arithmetic<_To>::value &&
+                                      std::is_arithmetic<_From>::value,
+                                  bool>::type
+__is_non_narrowing_arithmetic_convertible() {
+  return __is_non_narrowing_convertible_impl<_To>(_From{});
+}
+
+template <class _From, class _To>
+constexpr typename std::enable_if<!(std::is_arithmetic<_To>::value &&
+                                    std::is_arithmetic<_From>::value),
+                                  bool>::type
+__is_non_narrowing_arithmetic_convertible() {
+  return false;
+}
+
+template <class _Tp>
+constexpr _Tp __variadic_sum() {
+  return _Tp{};
+}
+
+template <class _Tp, class _Up, class... _Args>
+constexpr _Tp __variadic_sum(_Up __first, _Args... __rest) {
+  return static_cast<_Tp>(__first) + __variadic_sum<_Tp>(__rest...);
+}
+
+template <class _Tp>
+struct __nodeduce {
+  using type = _Tp;
+};
+
+template <class _Tp>
+constexpr bool __vectorizable() {
+  return std::is_arithmetic<_Tp>::value && !std::is_const<_Tp>::value &&
+         !std::is_volatile<_Tp>::value && !std::is_same<_Tp, bool>::value;
+}
+
+_LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_SIMD
+_LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD_ABI
+
+using scalar = __simd_abi<_StorageKind::_Scalar, 1>;
+
+template <int _Np>
+using fixed_size = __simd_abi<_StorageKind::_Array, _Np>;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t max_fixed_size = 32;
+
+template <class _Tp>
+using compatible = fixed_size<16 / sizeof(_Tp)>;
+
+#ifndef _LIBCUDACXX_HAS_NO_VECTOR_EXTENSION
+template <class _Tp>
+using native = __simd_abi<_StorageKind::_VecExt,
+                          _LIBCUDACXX_NATIVE_SIMD_WIDTH_IN_BYTES / sizeof(_Tp)>;
+#else
+template <class _Tp>
+using native =
+    fixed_size<_Tp, _LIBCUDACXX_NATIVE_SIMD_WIDTH_IN_BYTES / sizeof(_Tp)>;
+#endif // _LIBCUDACXX_HAS_NO_VECTOR_EXTENSION
+
+_LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_SIMD_ABI
+_LIBCUDACXX_BEGIN_NAMESPACE_EXPERIMENTAL_SIMD
+
+template <class _Tp, class _Abi = simd_abi::compatible<_Tp>>
+class simd;
+template <class _Tp, class _Abi = simd_abi::compatible<_Tp>>
+class simd_mask;
+
+struct element_aligned_tag {};
+struct vector_aligned_tag {};
+template <size_t>
+struct overaligned_tag {};
+_LIBCUDACXX_INLINE_VAR constexpr element_aligned_tag element_aligned{};
+_LIBCUDACXX_INLINE_VAR constexpr vector_aligned_tag vector_aligned{};
+template <size_t _Np>
+_LIBCUDACXX_INLINE_VAR constexpr overaligned_tag<_Np> overaligned{};
+
+// traits [simd.traits]
+template <class _Tp>
+struct is_abi_tag : std::integral_constant<bool, false> {};
+
+template <_StorageKind __kind, int _Np>
+struct is_abi_tag<__simd_abi<__kind, _Np>>
+    : std::integral_constant<bool, true> {};
+
+template <class _Tp>
+struct is_simd : std::integral_constant<bool, false> {};
+
+template <class _Tp, class _Abi>
+struct is_simd<simd<_Tp, _Abi>> : std::integral_constant<bool, true> {};
+
+template <class _Tp>
+struct is_simd_mask : std::integral_constant<bool, false> {};
+
+template <class _Tp, class _Abi>
+struct is_simd_mask<simd_mask<_Tp, _Abi>> : std::integral_constant<bool, true> {
+};
+
+template <class _Tp>
+struct is_simd_flag_type : std::integral_constant<bool, false> {};
+
+template <>
+struct is_simd_flag_type<element_aligned_tag>
+    : std::integral_constant<bool, true> {};
+
+template <>
+struct is_simd_flag_type<vector_aligned_tag>
+    : std::integral_constant<bool, true> {};
+
+template <size_t _Align>
+struct is_simd_flag_type<overaligned_tag<_Align>>
+    : std::integral_constant<bool, true> {};
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_abi_tag_v = is_abi_tag<_Tp>::value;
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_simd_v = is_simd<_Tp>::value;
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_simd_mask_v = is_simd_mask<_Tp>::value;
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_simd_flag_type_v =
+    is_simd_flag_type<_Tp>::value;
+template <class _Tp, size_t _Np>
+struct abi_for_size {
+  using type = simd_abi::fixed_size<_Np>;
+};
+template <class _Tp, size_t _Np>
+using abi_for_size_t = typename abi_for_size<_Tp, _Np>::type;
+
+template <class _Tp, class _Abi = simd_abi::compatible<_Tp>>
+struct simd_size;
+
+template <class _Tp, _StorageKind __kind, int _Np>
+struct simd_size<_Tp, __simd_abi<__kind, _Np>>
+    : std::integral_constant<size_t, _Np> {
+  static_assert(
+      std::is_arithmetic<_Tp>::value &&
+          !std::is_same<typename std::remove_const<_Tp>::type, bool>::value,
+      "Element type should be vectorizable");
+};
+
+// TODO: implement it.
+template <class _Tp, class _Up = typename _Tp::value_type>
+struct memory_alignment;
+
+template <class _Tp, class _Abi = simd_abi::compatible<_Tp>>
+_LIBCUDACXX_INLINE_VAR constexpr size_t simd_size_v = simd_size<_Tp, _Abi>::value;
+
+template <class _Tp, class _Up = typename _Tp::value_type>
+_LIBCUDACXX_INLINE_VAR constexpr size_t memory_alignment_v =
+    memory_alignment<_Tp, _Up>::value;
+
+// class template simd [simd.class]
+template <class _Tp>
+using native_simd = simd<_Tp, simd_abi::native<_Tp>>;
+template <class _Tp, int _Np>
+using fixed_size_simd = simd<_Tp, simd_abi::fixed_size<_Np>>;
+
+// class template simd_mask [simd.mask.class]
+template <class _Tp>
+using native_simd_mask = simd_mask<_Tp, simd_abi::native<_Tp>>;
+
+template <class _Tp, int _Np>
+using fixed_size_simd_mask = simd_mask<_Tp, simd_abi::fixed_size<_Np>>;
+
+// casts [simd.casts]
+template <class _Tp>
+struct __static_simd_cast_traits {
+  template <class _Up, class _Abi>
+  static simd<_Tp, _Abi> __apply(const simd<_Up, _Abi>& __v);
+};
+
+template <class _Tp, class _NewAbi>
+struct __static_simd_cast_traits<simd<_Tp, _NewAbi>> {
+  template <class _Up, class _Abi>
+  static typename std::enable_if<simd<_Up, _Abi>::size() ==
+                                     simd<_Tp, _NewAbi>::size(),
+                                 simd<_Tp, _NewAbi>>::type
+  __apply(const simd<_Up, _Abi>& __v);
+};
+
+template <class _Tp>
+struct __simd_cast_traits {
+  template <class _Up, class _Abi>
+  static typename std::enable_if<
+      __is_non_narrowing_arithmetic_convertible<_Up, _Tp>(),
+      simd<_Tp, _Abi>>::type
+  __apply(const simd<_Up, _Abi>& __v);
+};
+
+template <class _Tp, class _NewAbi>
+struct __simd_cast_traits<simd<_Tp, _NewAbi>> {
+  template <class _Up, class _Abi>
+  static typename std::enable_if<
+      __is_non_narrowing_arithmetic_convertible<_Up, _Tp>() &&
+          simd<_Up, _Abi>::size() == simd<_Tp, _NewAbi>::size(),
+      simd<_Tp, _NewAbi>>::type
+  __apply(const simd<_Up, _Abi>& __v);
+};
+
+template <class _Tp, class _Up, class _Abi>
+auto simd_cast(const simd<_Up, _Abi>& __v)
+    -> decltype(__simd_cast_traits<_Tp>::__apply(__v)) {
+  return __simd_cast_traits<_Tp>::__apply(__v);
+}
+
+template <class _Tp, class _Up, class _Abi>
+auto static_simd_cast(const simd<_Up, _Abi>& __v)
+    -> decltype(__static_simd_cast_traits<_Tp>::__apply(__v)) {
+  return __static_simd_cast_traits<_Tp>::__apply(__v);
+}
+
+template <class _Tp, class _Abi>
+fixed_size_simd<_Tp, simd_size<_Tp, _Abi>::value>
+to_fixed_size(const simd<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+fixed_size_simd_mask<_Tp, simd_size<_Tp, _Abi>::value>
+to_fixed_size(const simd_mask<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, size_t _Np>
+native_simd<_Tp> to_native(const fixed_size_simd<_Tp, _Np>&) noexcept;
+
+template <class _Tp, size_t _Np>
+native_simd_mask<_Tp> to_native(const fixed_size_simd_mask<_Tp, _Np>&) noexcept;
+
+template <class _Tp, size_t _Np>
+simd<_Tp> to_compatible(const fixed_size_simd<_Tp, _Np>&) noexcept;
+
+template <class _Tp, size_t _Np>
+simd_mask<_Tp> to_compatible(const fixed_size_simd_mask<_Tp, _Np>&) noexcept;
+
+template <size_t... __sizes, class _Tp, class _Abi>
+tuple<simd<_Tp, abi_for_size_t<_Tp, __sizes>>...> split(const simd<_Tp, _Abi>&);
+
+template <size_t... __sizes, class _Tp, class _Abi>
+tuple<simd_mask<_Tp, abi_for_size_t<_Tp, __sizes>>...>
+split(const simd_mask<_Tp, _Abi>&);
+
+template <class _SimdType, class _Abi>
+array<_SimdType, simd_size<typename _SimdType::value_type, _Abi>::value /
+                     _SimdType::size()>
+split(const simd<typename _SimdType::value_type, _Abi>&);
+
+template <class _SimdType, class _Abi>
+array<_SimdType, simd_size<typename _SimdType::value_type, _Abi>::value /
+                     _SimdType::size()>
+split(const simd_mask<typename _SimdType::value_type, _Abi>&);
+
+template <class _Tp, class... _Abis>
+simd<_Tp, abi_for_size_t<_Tp, __variadic_sum(simd_size<_Tp, _Abis>::value...)>>
+concat(const simd<_Tp, _Abis>&...);
+
+template <class _Tp, class... _Abis>
+simd_mask<_Tp,
+          abi_for_size_t<_Tp, __variadic_sum(simd_size<_Tp, _Abis>::value...)>>
+concat(const simd_mask<_Tp, _Abis>&...);
+
+// reductions [simd.mask.reductions]
+template <class _Tp, class _Abi>
+bool all_of(const simd_mask<_Tp, _Abi>&) noexcept;
+template <class _Tp, class _Abi>
+bool any_of(const simd_mask<_Tp, _Abi>&) noexcept;
+template <class _Tp, class _Abi>
+bool none_of(const simd_mask<_Tp, _Abi>&) noexcept;
+template <class _Tp, class _Abi>
+bool some_of(const simd_mask<_Tp, _Abi>&) noexcept;
+template <class _Tp, class _Abi>
+int popcount(const simd_mask<_Tp, _Abi>&) noexcept;
+template <class _Tp, class _Abi>
+int find_first_set(const simd_mask<_Tp, _Abi>&);
+template <class _Tp, class _Abi>
+int find_last_set(const simd_mask<_Tp, _Abi>&);
+bool all_of(bool) noexcept;
+bool any_of(bool) noexcept;
+bool none_of(bool) noexcept;
+bool some_of(bool) noexcept;
+int popcount(bool) noexcept;
+int find_first_set(bool) noexcept;
+int find_last_set(bool) noexcept;
+
+// masked assignment [simd.whereexpr]
+template <class _MaskType, class _Tp>
+class const_where_expression;
+template <class _MaskType, class _Tp>
+class where_expression;
+
+// masked assignment [simd.mask.where]
+template <class _Tp, class _Abi>
+where_expression<simd_mask<_Tp, _Abi>, simd<_Tp, _Abi>>
+where(const typename simd<_Tp, _Abi>::mask_type&, simd<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+const_where_expression<simd_mask<_Tp, _Abi>, const simd<_Tp, _Abi>>
+where(const typename simd<_Tp, _Abi>::mask_type&,
+      const simd<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+where_expression<simd_mask<_Tp, _Abi>, simd_mask<_Tp, _Abi>>
+where(const typename __nodeduce<simd_mask<_Tp, _Abi>>::type&,
+      simd_mask<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+const_where_expression<simd_mask<_Tp, _Abi>, const simd_mask<_Tp, _Abi>>
+where(const typename __nodeduce<simd_mask<_Tp, _Abi>>::type&,
+      const simd_mask<_Tp, _Abi>&) noexcept;
+
+template <class _Tp>
+where_expression<bool, _Tp> where(bool, _Tp&) noexcept;
+
+template <class _Tp>
+const_where_expression<bool, const _Tp> where(bool, const _Tp&) noexcept;
+
+// reductions [simd.reductions]
+template <class _Tp, class _Abi, class _BinaryOp = std::plus<_Tp>>
+_Tp reduce(const simd<_Tp, _Abi>&, _BinaryOp = _BinaryOp());
+
+template <class _MaskType, class _SimdType, class _BinaryOp>
+typename _SimdType::value_type
+reduce(const const_where_expression<_MaskType, _SimdType>&,
+       typename _SimdType::value_type neutral_element, _BinaryOp binary_op);
+
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+reduce(const const_where_expression<_MaskType, _SimdType>&,
+       plus<typename _SimdType::value_type> binary_op = {});
+
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+reduce(const const_where_expression<_MaskType, _SimdType>&,
+       multiplies<typename _SimdType::value_type> binary_op);
+
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+reduce(const const_where_expression<_MaskType, _SimdType>&,
+       bit_and<typename _SimdType::value_type> binary_op);
+
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+reduce(const const_where_expression<_MaskType, _SimdType>&,
+       bit_or<typename _SimdType::value_type> binary_op);
+
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+reduce(const const_where_expression<_MaskType, _SimdType>&,
+       bit_xor<typename _SimdType::value_type> binary_op);
+
+template <class _Tp, class _Abi>
+_Tp hmin(const simd<_Tp, _Abi>&);
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+hmin(const const_where_expression<_MaskType, _SimdType>&);
+template <class _Tp, class _Abi>
+_Tp hmax(const simd<_Tp, _Abi>&);
+template <class _MaskType, class _SimdType>
+typename _SimdType::value_type
+hmax(const const_where_expression<_MaskType, _SimdType>&);
+
+// algorithms [simd.alg]
+template <class _Tp, class _Abi>
+simd<_Tp, _Abi> min(const simd<_Tp, _Abi>&, const simd<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+simd<_Tp, _Abi> max(const simd<_Tp, _Abi>&, const simd<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+std::pair<simd<_Tp, _Abi>, simd<_Tp, _Abi>>
+minmax(const simd<_Tp, _Abi>&, const simd<_Tp, _Abi>&) noexcept;
+
+template <class _Tp, class _Abi>
+simd<_Tp, _Abi> clamp(const simd<_Tp, _Abi>&, const simd<_Tp, _Abi>&,
+                      const simd<_Tp, _Abi>&);
+
+// [simd.whereexpr]
+// TODO implement where expressions.
+template <class _MaskType, class _Tp>
+class const_where_expression {
+public:
+  const_where_expression(const const_where_expression&) = delete;
+  const_where_expression& operator=(const const_where_expression&) = delete;
+  typename remove_const<_Tp>::type operator-() const&&;
+  template <class _Up, class _Flags>
+  void copy_to(_Up*, _Flags) const&&;
+};
+
+template <class _MaskType, class _Tp>
+class where_expression : public const_where_expression<_MaskType, _Tp> {
+public:
+  where_expression(const where_expression&) = delete;
+  where_expression& operator=(const where_expression&) = delete;
+  template <class _Up>
+  void operator=(_Up&&);
+  template <class _Up>
+  void operator+=(_Up&&);
+  template <class _Up>
+  void operator-=(_Up&&);
+  template <class _Up>
+  void operator*=(_Up&&);
+  template <class _Up>
+  void operator/=(_Up&&);
+  template <class _Up>
+  void operator%=(_Up&&);
+  template <class _Up>
+  void operator&=(_Up&&);
+  template <class _Up>
+  void operator|=(_Up&&);
+  template <class _Up>
+  void operator^=(_Up&&);
+  template <class _Up>
+  void operator<<=(_Up&&);
+  template <class _Up>
+  void operator>>=(_Up&&);
+  void operator++();
+  void operator++(int);
+  void operator--();
+  void operator--(int);
+  template <class _Up, class _Flags>
+  void copy_from(const _Up*, _Flags);
+};
+
+// [simd.class]
+// TODO: implement simd
+template <class _Tp, class _Abi>
+class simd {
+public:
+  using value_type = _Tp;
+  using reference = __simd_reference<_Tp, _Tp, _Abi>;
+  using mask_type = simd_mask<_Tp, _Abi>;
+  using abi_type = _Abi;
+
+  simd() = default;
+  simd(const simd&) = default;
+  simd& operator=(const simd&) = default;
+
+  static constexpr size_t size() noexcept {
+    return simd_size<_Tp, _Abi>::value;
+  }
+
+private:
+  __simd_storage<_Tp, _Abi> __s_;
+
+  template <class _Up>
+  static constexpr bool __can_broadcast() {
+    return (std::is_arithmetic<_Up>::value &&
+            __is_non_narrowing_arithmetic_convertible<_Up, _Tp>()) ||
+           (!std::is_arithmetic<_Up>::value &&
+            std::is_convertible<_Up, _Tp>::value) ||
+           std::is_same<typename std::remove_const<_Up>::type, int>::value ||
+           (std::is_same<typename std::remove_const<_Up>::type,
+                         unsigned int>::value &&
+            std::is_unsigned<_Tp>::value);
+  }
+
+  template <class _Generator, size_t... __indicies>
+  static constexpr decltype(
+      std::forward_as_tuple(std::declval<_Generator>()(
+          std::integral_constant<size_t, __indicies>())...),
+      bool())
+  __can_generate(std::index_sequence<__indicies...>) {
+    return !__variadic_sum<bool>(
+        !__can_broadcast<decltype(std::declval<_Generator>()(
+            std::integral_constant<size_t, __indicies>()))>()...);
+  }
+
+  template <class _Generator>
+  static bool __can_generate(...) {
+    return false;
+  }
+
+  template <class _Generator, size_t... __indicies>
+  void __generator_init(_Generator&& __g, std::index_sequence<__indicies...>) {
+    int __not_used[]{((*this)[__indicies] =
+                          __g(std::integral_constant<size_t, __indicies>()),
+                      0)...};
+    (void)__not_used;
+  }
+
+public:
+  // implicit type conversion constructor
+  template <class _Up,
+            class = typename std::enable_if<
+                std::is_same<_Abi, simd_abi::fixed_size<size()>>::value &&
+                __is_non_narrowing_arithmetic_convertible<_Up, _Tp>()>::type>
+  simd(const simd<_Up, simd_abi::fixed_size<size()>>& __v) {
+    for (size_t __i = 0; __i < size(); __i++) {
+      (*this)[__i] = static_cast<_Tp>(__v[__i]);
+    }
+  }
+
+  // implicit broadcast constructor
+  template <class _Up,
+            class = typename std::enable_if<__can_broadcast<_Up>()>::type>
+  simd(_Up&& __rv) {
+    auto __v = static_cast<_Tp>(__rv);
+    for (size_t __i = 0; __i < size(); __i++) {
+      (*this)[__i] = __v;
+    }
+  }
+
+  // generator constructor
+  template <class _Generator,
+            int = typename std::enable_if<
+                __can_generate<_Generator>(std::make_index_sequence<size()>()),
+                int>::type()>
+  explicit simd(_Generator&& __g) {
+    __generator_init(std::forward<_Generator>(__g),
+                     std::make_index_sequence<size()>());
+  }
+
+  // load constructor
+  template <
+      class _Up, class _Flags,
+      class = typename std::enable_if<__vectorizable<_Up>()>::type,
+      class = typename std::enable_if<is_simd_flag_type<_Flags>::value>::type>
+  simd(const _Up* __buffer, _Flags) {
+    // TODO: optimize for overaligned flags
+    for (size_t __i = 0; __i < size(); __i++) {
+      (*this)[__i] = static_cast<_Tp>(__buffer[__i]);
+    }
+  }
+
+  // loads [simd.load]
+  template <class _Up, class _Flags>
+  typename std::enable_if<__vectorizable<_Up>() &&
+                          is_simd_flag_type<_Flags>::value>::type
+  copy_from(const _Up* __buffer, _Flags) {
+    *this = simd(__buffer, _Flags());
+  }
+
+  // stores [simd.store]
+  template <class _Up, class _Flags>
+  typename std::enable_if<__vectorizable<_Up>() &&
+                          is_simd_flag_type<_Flags>::value>::type
+  copy_to(_Up* __buffer, _Flags) const {
+    // TODO: optimize for overaligned flags
+    for (size_t __i = 0; __i < size(); __i++) {
+      __buffer[__i] = static_cast<_Up>((*this)[__i]);
+    }
+  }
+
+  // scalar access [simd.subscr]
+  reference operator[](size_t __i) { return reference(&__s_, __i); }
+
+  value_type operator[](size_t __i) const { return __s_.__get(__i); }
+
+  // unary operators [simd.unary]
+  simd& operator++();
+  simd operator++(int);
+  simd& operator--();
+  simd operator--(int);
+  mask_type operator!() const;
+  simd operator~() const;
+  simd operator+() const;
+  simd operator-() const;
+
+  // binary operators [simd.binary]
+  friend simd operator+(const simd&, const simd&);
+  friend simd operator-(const simd&, const simd&);
+  friend simd operator*(const simd&, const simd&);
+  friend simd operator/(const simd&, const simd&);
+  friend simd operator%(const simd&, const simd&);
+  friend simd operator&(const simd&, const simd&);
+  friend simd operator|(const simd&, const simd&);
+  friend simd operator^(const simd&, const simd&);
+  friend simd operator<<(const simd&, const simd&);
+  friend simd operator>>(const simd&, const simd&);
+  friend simd operator<<(const simd&, int);
+  friend simd operator>>(const simd&, int);
+
+  // compound assignment [simd.cassign]
+  friend simd& operator+=(simd&, const simd&);
+  friend simd& operator-=(simd&, const simd&);
+  friend simd& operator*=(simd&, const simd&);
+  friend simd& operator/=(simd&, const simd&);
+  friend simd& operator%=(simd&, const simd&);
+
+  friend simd& operator&=(simd&, const simd&);
+  friend simd& operator|=(simd&, const simd&);
+  friend simd& operator^=(simd&, const simd&);
+  friend simd& operator<<=(simd&, const simd&);
+  friend simd& operator>>=(simd&, const simd&);
+  friend simd& operator<<=(simd&, int);
+  friend simd& operator>>=(simd&, int);
+
+  // compares [simd.comparison]
+  friend mask_type operator==(const simd&, const simd&);
+  friend mask_type operator!=(const simd&, const simd&);
+  friend mask_type operator>=(const simd&, const simd&);
+  friend mask_type operator<=(const simd&, const simd&);
+  friend mask_type operator>(const simd&, const simd&);
+  friend mask_type operator<(const simd&, const simd&);
+};
+
+// [simd.mask.class]
+template <class _Tp, class _Abi>
+// TODO: implement simd_mask
+class simd_mask {
+public:
+  using value_type = bool;
+  // TODO: this is strawman implementation. Turn it into a proxy type.
+  using reference = bool&;
+  using simd_type = simd<_Tp, _Abi>;
+  using abi_type = _Abi;
+  static constexpr size_t size() noexcept;
+  simd_mask() = default;
+
+  // broadcast constructor
+  explicit simd_mask(value_type) noexcept;
+
+  // implicit type conversion constructor
+  template <class _Up>
+  simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>&) noexcept;
+
+  // load constructor
+  template <class _Flags>
+  simd_mask(const value_type*, _Flags);
+
+  // loads [simd.mask.copy]
+  template <class _Flags>
+  void copy_from(const value_type*, _Flags);
+  template <class _Flags>
+  void copy_to(value_type*, _Flags) const;
+
+  // scalar access [simd.mask.subscr]
+  reference operator[](size_t);
+  value_type operator[](size_t) const;
+
+  // unary operators [simd.mask.unary]
+  simd_mask operator!() const noexcept;
+
+  // simd_mask binary operators [simd.mask.binary]
+  friend simd_mask operator&&(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator||(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator&(const simd_mask&, const simd_mask&)noexcept;
+  friend simd_mask operator|(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator^(const simd_mask&, const simd_mask&) noexcept;
+
+  // simd_mask compound assignment [simd.mask.cassign]
+  friend simd_mask& operator&=(simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask& operator|=(simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask& operator^=(simd_mask&, const simd_mask&) noexcept;
+
+  // simd_mask compares [simd.mask.comparison]
+  friend simd_mask operator==(const simd_mask&, const simd_mask&) noexcept;
+  friend simd_mask operator!=(const simd_mask&, const simd_mask&) noexcept;
+};
+
+#endif // _LIBCUDACXX_STD_VER >= 17
+
+_LIBCUDACXX_END_NAMESPACE_EXPERIMENTAL_SIMD
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_SIMD */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string
new file mode 100644
index 000000000000..0e502e823b3f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/string
@@ -0,0 +1,61 @@
+// -*- C++ -*-
+//===--------------------------- string ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_STRING
+#define _LIBCUDACXX_EXPERIMENTAL_STRING
+/*
+    experimental/string synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  // basic_string using polymorphic allocator in namespace pmr
+  template <class charT, class traits = char_traits<charT>>
+   using basic_string =
+     std::basic_string<charT, traits, polymorphic_allocator<charT>>;
+
+  // basic_string typedef names using polymorphic allocator in namespace
+  // std::experimental::pmr
+  typedef basic_string<char> string;
+  typedef basic_string<char16_t> u16string;
+  typedef basic_string<char32_t> u32string;
+  typedef basic_string<wchar_t> wstring;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <string>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _CharT, class _Traits = char_traits<_CharT>>
+using basic_string =
+    _CUDA_VSTD::basic_string<_CharT, _Traits, polymorphic_allocator<_CharT>>;
+
+typedef basic_string<char> string;
+typedef basic_string<char16_t> u16string;
+typedef basic_string<char32_t> u32string;
+typedef basic_string<wchar_t> wstring;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_STRING */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits
new file mode 100644
index 000000000000..81555c2b2968
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/type_traits
@@ -0,0 +1,154 @@
+// -*- C++ -*-
+//===-------------------------- type_traits -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_TYPE_TRAITS
+#define _LIBCUDACXX_EXPERIMENTAL_TYPE_TRAITS
+
+/**
+    experimental/type_traits synopsis
+
+// C++1y
+#include <type_traits>
+
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+
+  // 3.3.2, Other type transformations
+  template <class> class invocation_type; // not defined
+  template <class F, class... ArgTypes> class invocation_type<F(ArgTypes...)>;
+  template <class> class raw_invocation_type; // not defined
+  template <class F, class... ArgTypes> class raw_invocation_type<F(ArgTypes...)>;
+
+  template <class T>
+    using invocation_type_t = typename invocation_type<T>::type;
+  template <class T>
+    using raw_invocation_type_t = typename raw_invocation_type<T>::type;
+
+  // 3.3.4, Detection idiom
+  template <class...> using void_t = void;
+
+  struct nonesuch {
+    nonesuch() = delete;
+    ~nonesuch() = delete;
+    nonesuch(nonesuch const&) = delete;
+    void operator=(nonesuch const&) = delete;
+  };
+
+  template <template<class...> class Op, class... Args>
+    using is_detected = see below;
+  template <template<class...> class Op, class... Args>
+    constexpr bool is_detected_v = is_detected<Op, Args...>::value;
+  template <template<class...> class Op, class... Args>
+    using detected_t = see below;
+  template <class Default, template<class...> class Op, class... Args>
+    using detected_or = see below;
+  template <class Default, template<class...> class Op, class... Args>
+    using detected_or_t = typename detected_or<Default, Op, Args...>::type;
+  template <class Expected, template<class...> class Op, class... Args>
+    using is_detected_exact = is_same<Expected, detected_t<Op, Args...>>;
+  template <class Expected, template<class...> class Op, class... Args>
+    constexpr bool is_detected_exact_v
+      = is_detected_exact<Expected, Op, Args...>::value;
+  template <class To, template<class...> class Op, class... Args>
+     using is_detected_convertible = is_convertible<detected_t<Op, Args...>, To>;
+  template <class To, template<class...> class Op, class... Args>
+     constexpr bool is_detected_convertible_v
+       = is_detected_convertible<To, Op, Args...>::value;  
+
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+
+#if _LIBCUDACXX_STD_VER > 11
+
+#include <initializer_list>
+#include <type_traits>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS
+
+// 3.3.2, Other type transformations
+/*
+template <class>
+class _LIBCUDACXX_TEMPLATE_VIS raw_invocation_type;
+
+template <class _Fn, class ..._Args>
+class _LIBCUDACXX_TEMPLATE_VIS raw_invocation_type<_Fn(_Args...)>;
+
+template <class>
+class _LIBCUDACXX_TEMPLATE_VIS invokation_type;
+
+template <class _Fn, class ..._Args>
+class _LIBCUDACXX_TEMPLATE_VIS invokation_type<_Fn(_Args...)>;
+
+template <class _Tp>
+using invokation_type_t = typename invokation_type<_Tp>::type;
+
+template <class _Tp>
+using raw_invocation_type_t = typename raw_invocation_type<_Tp>::type;
+*/
+
+// 3.3.4, Detection idiom
+template <class...> using void_t = void;
+
+struct nonesuch : private _CUDA_VSTD::__nat { // make nonesuch "not an aggregate"
+  ~nonesuch() = delete;
+  nonesuch      (nonesuch const&) = delete;
+  void operator=(nonesuch const&) = delete;
+  };
+
+template <class _Default, class _AlwaysVoid, template <class...> class _Op, class... _Args>
+struct _DETECTOR {
+   using value_t = false_type;
+   using type = _Default;
+   };
+
+template <class _Default, template <class...> class _Op, class... _Args>
+struct _DETECTOR<_Default, void_t<_Op<_Args...>>, _Op, _Args...> {
+   using value_t = true_type;
+   using type = _Op<_Args...>;
+   };
+     
+
+template <template<class...> class _Op, class... _Args>
+  using is_detected = typename _DETECTOR<nonesuch, void, _Op, _Args...>::value_t;
+template <template<class...> class _Op, class... _Args>
+  using detected_t = typename _DETECTOR<nonesuch, void, _Op, _Args...>::type;
+template <template<class...> class _Op, class... _Args>
+  _LIBCUDACXX_CONSTEXPR bool is_detected_v = is_detected<_Op, _Args...>::value;
+
+template <class Default, template<class...> class _Op, class... _Args>
+  using detected_or = _DETECTOR<Default, void, _Op, _Args...>;
+template <class Default, template<class...> class _Op, class... _Args>
+  using detected_or_t = typename detected_or<Default, _Op, _Args...>::type;
+
+template <class Expected, template<class...> class _Op, class... _Args>
+  using is_detected_exact = is_same<Expected, detected_t<_Op, _Args...>>;
+template <class Expected, template<class...> class _Op, class... _Args>
+  _LIBCUDACXX_CONSTEXPR bool is_detected_exact_v = is_detected_exact<Expected, _Op, _Args...>::value;
+
+template <class To, template<class...> class _Op, class... _Args>
+  using is_detected_convertible = is_convertible<detected_t<_Op, _Args...>, To>;
+template <class To, template<class...> class _Op, class... _Args>
+  _LIBCUDACXX_CONSTEXPR bool is_detected_convertible_v = is_detected_convertible<To, _Op, _Args...>::value;  
+
+
+_LIBCUDACXX_END_NAMESPACE_LFTS
+
+#endif /* _LIBCUDACXX_STD_VER > 11 */
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_TYPE_TRAITS */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map
new file mode 100644
index 000000000000..221156a3e84b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_map
@@ -0,0 +1,64 @@
+// -*- C++ -*-
+//===------------------------- unordered_map ------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_UNORDERED_MAP
+#define _LIBCUDACXX_EXPERIMENTAL_UNORDERED_MAP
+/*
+    experimental/unordered_map synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class Key, class T,
+            class Hash = hash<Key>,
+            class Pred = equal_to<Key>>
+  using unordered_map =
+    std::unordered_map<Key, T, Hash, Pred,
+                       polymorphic_allocator<pair<const Key,T>>>;
+
+  template <class Key, class T,
+            class Hash = hash<Key>,
+            class Pred = equal_to<Key>>
+  using unordered_multimap =
+    std::unordered_multimap<Key, T, Hash, Pred,
+                            polymorphic_allocator<pair<const Key,T>>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <unordered_map>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _Key, class _Value,
+          class _Hash = hash<_Key>, class _Pred = equal_to<_Key>>
+using unordered_map = _CUDA_VSTD::unordered_map<_Key, _Value, _Hash, _Pred,
+                        polymorphic_allocator<pair<const _Key, _Value>>>;
+
+template <class _Key, class _Value,
+          class _Hash = hash<_Key>, class _Pred = equal_to<_Key>>
+using unordered_multimap = _CUDA_VSTD::unordered_multimap<_Key, _Value, _Hash, _Pred,
+                        polymorphic_allocator<pair<const _Key, _Value>>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_UNORDERED_MAP */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set
new file mode 100644
index 000000000000..2c3243e06c71
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/unordered_set
@@ -0,0 +1,58 @@
+// -*- C++ -*-
+//===------------------------- unordered_set ------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_UNORDERED_SET
+#define _LIBCUDACXX_EXPERIMENTAL_UNORDERED_SET
+/*
+    experimental/unordered_set synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class T, class Hash = hash<T>, class Pred = equal_to<T>>
+  using unordered_set = std::unordered_set<T, Hash, Pred,
+                       polymorphic_allocator<T>>;
+
+  template <class T, class Hash = hash<T>, class Pred = equal_to<T>>
+  using unordered_multiset = std::unordered_multiset<T, Hash, Pred,
+                       polymorphic_allocator<T>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <unordered_set>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _Value,
+          class _Hash = hash<_Value>, class _Pred = equal_to<_Value>>
+using unordered_set = _CUDA_VSTD::unordered_set<_Value, _Hash, _Pred,
+                        polymorphic_allocator<_Value>>;
+
+template <class _Value,
+          class _Hash = hash<_Value>, class _Pred = equal_to<_Value>>
+using unordered_multiset = _CUDA_VSTD::unordered_multiset<_Value, _Hash, _Pred,
+                        polymorphic_allocator<_Value>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_UNORDERED_SET */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility
new file mode 100644
index 000000000000..8f7f7a149d9f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/utility
@@ -0,0 +1,46 @@
+// -*- C++ -*-
+//===-------------------------- utility ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_UTILITY
+#define _LIBCUDACXX_EXPERIMENTAL_UTILITY
+
+/*
+    experimental/utility synopsis
+
+// C++1y
+
+#include <utility>
+
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+
+  3.1.2, erased-type placeholder
+  struct erased_type { };
+
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <utility>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS
+
+  struct _LIBCUDACXX_TEMPLATE_VIS erased_type { };
+
+_LIBCUDACXX_END_NAMESPACE_LFTS
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_UTILITY */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector
new file mode 100644
index 000000000000..d3c593467c55
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/experimental/vector
@@ -0,0 +1,46 @@
+// -*- C++ -*-
+//===--------------------------- vector ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXPERIMENTAL_VECTOR
+#define _LIBCUDACXX_EXPERIMENTAL_VECTOR
+/*
+    experimental/vector synopsis
+
+// C++1z
+namespace std {
+namespace experimental {
+inline namespace fundamentals_v1 {
+namespace pmr {
+
+  template <class T>
+  using vector = std::vector<T, polymorphic_allocator<T>>;
+
+} // namespace pmr
+} // namespace fundamentals_v1
+} // namespace experimental
+} // namespace std
+
+ */
+
+#include <experimental/__config>
+#include <vector>
+#include <experimental/memory_resource>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_LFTS_PMR
+
+template <class _ValueT>
+using vector = _CUDA_VSTD::vector<_ValueT, polymorphic_allocator<_ValueT>>;
+
+_LIBCUDACXX_END_NAMESPACE_LFTS_PMR
+
+#endif /* _LIBCUDACXX_EXPERIMENTAL_VECTOR */
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/__hash b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/__hash
new file mode 100644
index 000000000000..366df8336af0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/__hash
@@ -0,0 +1,133 @@
+// -*- C++ -*-
+//===------------------------- hash_set ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_EXT_HASH
+#define _LIBCUDACXX_EXT_HASH
+
+#pragma GCC system_header
+
+#include <string>
+#include <cstring>
+
+namespace __gnu_cxx {
+
+template <typename _Tp> struct _LIBCUDACXX_TEMPLATE_VIS hash { };
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<const char*>
+ : public std::unary_function<const char*, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const char *__c) const _NOEXCEPT
+    {
+        return std::__do_string_hash(__c, __c + strlen(__c));
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<char *>
+ : public std::unary_function<char*, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(char *__c) const _NOEXCEPT
+    {
+        return std::__do_string_hash<const char *>(__c, __c + strlen(__c));
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<char>
+ : public std::unary_function<char, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(char __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<signed char>
+ : public std::unary_function<signed char, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(signed char __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned char>
+ : public std::unary_function<unsigned char, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned char __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<short>
+ : public std::unary_function<short, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(short __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned short>
+ : public std::unary_function<unsigned short, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned short __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<int>
+    : public std::unary_function<int, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(int __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned int>
+    : public std::unary_function<unsigned int, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned int __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<long>
+    : public std::unary_function<long, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(long __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned long>
+    : public std::unary_function<unsigned long, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned long __c) const _NOEXCEPT
+    {
+        return __c;
+    }
+};
+}
+
+#endif  // _LIBCUDACXX_EXT_HASH
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map
new file mode 100644
index 000000000000..799df8a50fcf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_map
@@ -0,0 +1,984 @@
+// -*- C++ -*-
+//===-------------------------- hash_map ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_HASH_MAP
+#define _LIBCUDACXX_HASH_MAP
+
+/*
+
+    hash_map synopsis
+
+namespace __gnu_cxx
+{
+
+template <class Key, class T, class Hash = hash<Key>, class Pred = equal_to<Key>,
+          class Alloc = allocator<pair<const Key, T>>>
+class hash_map
+{
+public:
+    // types
+    typedef Key                                                        key_type;
+    typedef T                                                          mapped_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef pair<const key_type, mapped_type>                          value_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+
+    hash_map();
+    explicit hash_map(size_type n, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+    hash_map(InputIterator f, InputIterator l);
+    template <class InputIterator>
+    hash_map(InputIterator f, InputIterator l,
+                size_type n, const hasher& hf = hasher(),
+                const key_equal& eql = key_equal(),
+                const allocator_type& a = allocator_type());
+    hash_map(const hash_map&);
+    ~hash_map();
+    hash_map& operator=(const hash_map&);
+
+    allocator_type get_allocator() const;
+
+    bool      empty() const;
+    size_type size() const;
+    size_type max_size() const;
+
+    iterator       begin();
+    iterator       end();
+    const_iterator begin()  const;
+    const_iterator end()    const;
+
+    pair<iterator, bool> insert(const value_type& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+
+    void erase(const_iterator position);
+    size_type erase(const key_type& k);
+    void erase(const_iterator first, const_iterator last);
+    void clear();
+
+    void swap(hash_map&);
+
+    hasher hash_funct() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    mapped_type& operator[](const key_type& k);
+
+    size_type bucket_count() const;
+    size_type max_bucket_count() const;
+
+    size_type elems_in_bucket(size_type n) const;
+
+    void resize(size_type n);
+};
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    void swap(hash_map<Key, T, Hash, Pred, Alloc>& x,
+              hash_map<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const hash_map<Key, T, Hash, Pred, Alloc>& x,
+               const hash_map<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const hash_map<Key, T, Hash, Pred, Alloc>& x,
+               const hash_map<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash = hash<Key>, class Pred = equal_to<Key>,
+          class Alloc = allocator<pair<const Key, T>>>
+class hash_multimap
+{
+public:
+    // types
+    typedef Key                                                        key_type;
+    typedef T                                                          mapped_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef pair<const key_type, mapped_type>                          value_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+
+    explicit hash_multimap(size_type n = 193, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        hash_multimap(InputIterator f, InputIterator l,
+                      size_type n = 193, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    explicit hash_multimap(const allocator_type&);
+    hash_multimap(const hash_multimap&);
+    ~hash_multimap();
+    hash_multimap& operator=(const hash_multimap&);
+
+    allocator_type get_allocator() const;
+
+    bool      empty() const;
+    size_type size() const;
+    size_type max_size() const;
+
+    iterator       begin();
+    iterator       end();
+    const_iterator begin()  const;
+    const_iterator end()    const;
+
+    iterator insert(const value_type& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+
+    void erase(const_iterator position);
+    size_type erase(const key_type& k);
+    void erase(const_iterator first, const_iterator last);
+    void clear();
+
+    void swap(hash_multimap&);
+
+    hasher hash_funct() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    size_type bucket_count() const;
+    size_type max_bucket_count() const;
+
+    size_type elems_in_bucket(size_type n) const;
+
+    void resize(size_type n);
+};
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    void swap(hash_multimap<Key, T, Hash, Pred, Alloc>& x,
+              hash_multimap<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const hash_multimap<Key, T, Hash, Pred, Alloc>& x,
+               const hash_multimap<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const hash_multimap<Key, T, Hash, Pred, Alloc>& x,
+               const hash_multimap<Key, T, Hash, Pred, Alloc>& y);
+
+}  // __gnu_cxx
+
+*/
+
+#include <__config>
+#include <__hash_table>
+#include <functional>
+#include <stdexcept>
+#include <type_traits>
+#include <ext/__hash>
+
+#if __DEPRECATED
+#if defined(_LIBCUDACXX_WARNING)
+    _LIBCUDACXX_WARNING("Use of the header <ext/hash_map> is deprecated.  Migrate to <unordered_map>")
+#else
+#   warning Use of the header <ext/hash_map> is deprecated.  Migrate to <unordered_map>
+#endif
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+namespace __gnu_cxx {
+
+template <class _Tp, class _Hash,
+          bool = std::is_empty<_Hash>::value && !std::__libcpp_is_final<_Hash>::value
+        >
+class __hash_map_hasher
+    : private _Hash
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_hasher() : _Hash() {}
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_hasher(const _Hash& __h) : _Hash(__h) {}
+    _LIBCUDACXX_INLINE_VISIBILITY const _Hash& hash_function() const {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const _Tp& __x) const
+        {return static_cast<const _Hash&>(*this)(__x.first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const typename _Tp::first_type& __x) const
+        {return static_cast<const _Hash&>(*this)(__x);}
+};
+
+template <class _Tp, class _Hash>
+class __hash_map_hasher<_Tp, _Hash, false>
+{
+    _Hash __hash_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_hasher() : __hash_() {}
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_hasher(const _Hash& __h) : __hash_(__h) {}
+    _LIBCUDACXX_INLINE_VISIBILITY const _Hash& hash_function() const {return __hash_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const _Tp& __x) const
+        {return __hash_(__x.first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const typename _Tp::first_type& __x) const
+        {return __hash_(__x);}
+};
+
+template <class _Tp, class _Pred,
+          bool = std::is_empty<_Pred>::value && !std::__libcpp_is_final<_Pred>::value
+         >
+class __hash_map_equal
+    : private _Pred
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_equal() : _Pred() {}
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_equal(const _Pred& __p) : _Pred(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY const _Pred& key_eq() const {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return static_cast<const _Pred&>(*this)(__x.first, __y.first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const typename _Tp::first_type& __x, const _Tp& __y) const
+        {return static_cast<const _Pred&>(*this)(__x, __y.first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const typename _Tp::first_type& __y) const
+        {return static_cast<const _Pred&>(*this)(__x.first, __y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const typename _Tp::first_type& __x,
+                    const typename _Tp::first_type& __y) const
+        {return static_cast<const _Pred&>(*this)(__x, __y);}
+};
+
+template <class _Tp, class _Pred>
+class __hash_map_equal<_Tp, _Pred, false>
+{
+    _Pred __pred_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_equal() : __pred_() {}
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_equal(const _Pred& __p) : __pred_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY const _Pred& key_eq() const {return __pred_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __pred_(__x.first, __y.first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const typename _Tp::first_type& __x, const _Tp& __y) const
+        {return __pred_(__x, __y.first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const typename _Tp::first_type& __y) const
+        {return __pred_(__x.first, __y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const typename _Tp::first_type& __x,
+                    const typename _Tp::first_type& __y) const
+        {return __pred_(__x, __y);}
+};
+
+template <class _Alloc>
+class __hash_map_node_destructor
+{
+    typedef _Alloc                              allocator_type;
+    typedef std::allocator_traits<allocator_type>    __alloc_traits;
+    typedef typename __alloc_traits::value_type::__node_value_type value_type;
+public:
+    typedef typename __alloc_traits::pointer    pointer;
+private:
+    typedef typename value_type::first_type     first_type;
+    typedef typename value_type::second_type    second_type;
+
+    allocator_type& __na_;
+
+public:
+    bool __first_constructed;
+    bool __second_constructed;
+
+    __hash_map_node_destructor(__hash_map_node_destructor const&) = default;
+    __hash_map_node_destructor& operator=(const __hash_map_node_destructor&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __hash_map_node_destructor(allocator_type& __na)
+        : __na_(__na),
+          __first_constructed(false),
+          __second_constructed(false)
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_node_destructor(std::__hash_node_destructor<allocator_type>&& __x)
+        : __na_(__x.__na_),
+          __first_constructed(__x.__value_constructed),
+          __second_constructed(__x.__value_constructed)
+        {
+            __x.__value_constructed = false;
+        }
+#else  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_node_destructor(const std::__hash_node_destructor<allocator_type>& __x)
+        : __na_(__x.__na_),
+          __first_constructed(__x.__value_constructed),
+          __second_constructed(__x.__value_constructed)
+        {
+            const_cast<bool&>(__x.__value_constructed) = false;
+        }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p)
+    {
+        if (__second_constructed)
+            __alloc_traits::destroy(__na_, _CUDA_VSTD::addressof(__p->__value_.second));
+        if (__first_constructed)
+            __alloc_traits::destroy(__na_, _CUDA_VSTD::addressof(__p->__value_.first));
+        if (__p)
+            __alloc_traits::deallocate(__na_, __p, 1);
+    }
+};
+
+template <class _HashIterator>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_map_iterator
+{
+    _HashIterator __i_;
+
+    typedef const typename _HashIterator::value_type::first_type key_type;
+    typedef typename _HashIterator::value_type::second_type      mapped_type;
+public:
+    typedef std::forward_iterator_tag                            iterator_category;
+    typedef std::pair<key_type, mapped_type>                     value_type;
+    typedef typename _HashIterator::difference_type              difference_type;
+    typedef value_type&                                          reference;
+    typedef typename std::__rebind_pointer<typename _HashIterator::pointer, value_type>::type
+        pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_iterator() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_iterator(_HashIterator __i) : __i_(__i) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference operator*() const {return *operator->();}
+    _LIBCUDACXX_INLINE_VISIBILITY pointer operator->() const {return (pointer)__i_.operator->();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_iterator& operator++() {++__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_iterator operator++(int)
+    {
+        __hash_map_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __hash_map_iterator& __x, const __hash_map_iterator& __y)
+        {return __x.__i_ == __y.__i_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __hash_map_iterator& __x, const __hash_map_iterator& __y)
+        {return __x.__i_ != __y.__i_;}
+
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS hash_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS hash_multimap;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator;
+};
+
+template <class _HashIterator>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator
+{
+    _HashIterator __i_;
+
+    typedef const typename _HashIterator::value_type::first_type key_type;
+    typedef typename _HashIterator::value_type::second_type      mapped_type;
+public:
+    typedef std::forward_iterator_tag                            iterator_category;
+    typedef std::pair<key_type, mapped_type>                     value_type;
+    typedef typename _HashIterator::difference_type              difference_type;
+    typedef const value_type&                                    reference;
+    typedef typename std::__rebind_pointer<typename _HashIterator::pointer, const value_type>::type
+        pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __hash_map_const_iterator() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator(_HashIterator __i) : __i_(__i) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator(
+            __hash_map_iterator<typename _HashIterator::__non_const_iterator> __i)
+                : __i_(__i.__i_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return *operator->();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return (pointer)__i_.operator->();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator& operator++() {++__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator operator++(int)
+    {
+        __hash_map_const_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __hash_map_const_iterator& __x, const __hash_map_const_iterator& __y)
+        {return __x.__i_ == __y.__i_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __hash_map_const_iterator& __x, const __hash_map_const_iterator& __y)
+        {return __x.__i_ != __y.__i_;}
+
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS hash_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS hash_multimap;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator;
+};
+
+template <class _Key, class _Tp, class _Hash = hash<_Key>, class _Pred = std::equal_to<_Key>,
+          class _Alloc = std::allocator<std::pair<const _Key, _Tp> > >
+class _LIBCUDACXX_TEMPLATE_VIS hash_map
+{
+public:
+    // types
+    typedef _Key                                           key_type;
+    typedef _Tp                                            mapped_type;
+    typedef _Tp                                            data_type;
+    typedef _Hash                                          hasher;
+    typedef _Pred                                          key_equal;
+    typedef _Alloc                                         allocator_type;
+    typedef std::pair<const key_type, mapped_type>         value_type;
+    typedef value_type&                                    reference;
+    typedef const value_type&                              const_reference;
+
+private:
+    typedef std::pair<key_type, mapped_type>                    __value_type;
+    typedef __hash_map_hasher<__value_type, hasher>   __hasher;
+    typedef __hash_map_equal<__value_type, key_equal> __key_equal;
+    typedef typename std::__rebind_alloc_helper<
+        std::allocator_traits<allocator_type>, __value_type>::type __allocator_type;
+
+    typedef std::__hash_table<__value_type, __hasher,
+                         __key_equal,  __allocator_type>   __table;
+
+    __table __table_;
+
+    typedef typename __table::__node_pointer               __node_pointer;
+    typedef typename __table::__node_const_pointer         __node_const_pointer;
+    typedef typename __table::__node_traits                __node_traits;
+    typedef typename __table::__node_allocator             __node_allocator;
+    typedef typename __table::__node                       __node;
+    typedef __hash_map_node_destructor<__node_allocator>   _Dp;
+    typedef std::unique_ptr<__node, _Dp>                   __node_holder;
+    typedef std::allocator_traits<allocator_type>          __alloc_traits;
+public:
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+    typedef typename __alloc_traits::size_type       size_type;
+    typedef typename __alloc_traits::difference_type difference_type;
+
+    typedef __hash_map_iterator<typename __table::iterator>       iterator;
+    typedef __hash_map_const_iterator<typename __table::const_iterator> const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY hash_map() { }
+    explicit hash_map(size_type __n, const hasher& __hf = hasher(),
+                           const key_equal& __eql = key_equal());
+    hash_map(size_type __n, const hasher& __hf,
+                  const key_equal& __eql,
+                  const allocator_type& __a);
+    template <class _InputIterator>
+        hash_map(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        hash_map(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        hash_map(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf,
+                      const key_equal& __eql,
+                      const allocator_type& __a);
+    hash_map(const hash_map& __u);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin()        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end()          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const {return __table_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<iterator, bool> insert(const value_type& __x)
+        {return __table_.__insert_unique(__x);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator, const value_type& __x) {return insert(__x).first;}
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __p) {__table_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __first, const_iterator __last)
+        {__table_.erase(__first.__i_, __last.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() {__table_.clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(hash_map& __u) {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_funct() const
+        {return __table_.hash_function().hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const
+        {return __table_.key_eq().key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_unique(__k);}
+
+    mapped_type& operator[](const key_type& __k);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type elems_in_bucket(size_type __n) const
+        {return __table_.bucket_size(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void resize(size_type __n) {__table_.rehash(__n);}
+
+private:
+    __node_holder __construct_node(const key_type& __k);
+};
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_map(
+        size_type __n, const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_map(
+        size_type __n, const hasher& __hf, const key_equal& __eql,
+        const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_map(
+        _InputIterator __first, _InputIterator __last)
+{
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_map(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_map(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_map(
+        const hash_map& __u)
+    : __table_(__u.__table_)
+{
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+typename hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::__node_holder
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::__construct_node(const key_type& __k)
+{
+    __node_allocator& __na = __table_.__node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _CUDA_VSTD::addressof(__h->__value_.first), __k);
+    __h.get_deleter().__first_constructed = true;
+    __node_traits::construct(__na, _CUDA_VSTD::addressof(__h->__value_.second));
+    __h.get_deleter().__second_constructed = true;
+    return _LIBCUDACXX_EXPLICIT_MOVE(__h);  // explicitly moved for C++03
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                       _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_unique(*__first);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+_Tp&
+hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::operator[](const key_type& __k)
+{
+    iterator __i = find(__k);
+    if (__i != end())
+        return __i->second;
+    __node_holder __h = __construct_node(__k);
+    std::pair<iterator, bool> __r = __table_.__node_insert_unique(__h.get());
+    __h.release();
+    return __r.first->second;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+     hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(), __ey = __y.end();
+            __i != __ex; ++__i)
+    {
+        const_iterator __j = __y.find(__i->first);
+        if (__j == __ey || !(*__i == *__j))
+            return false;
+    }
+    return true;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const hash_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Key, class _Tp, class _Hash = hash<_Key>, class _Pred = std::equal_to<_Key>,
+          class _Alloc = std::allocator<std::pair<const _Key, _Tp> > >
+class _LIBCUDACXX_TEMPLATE_VIS hash_multimap
+{
+public:
+    // types
+    typedef _Key                                           key_type;
+    typedef _Tp                                            mapped_type;
+    typedef _Tp                                            data_type;
+    typedef _Hash                                          hasher;
+    typedef _Pred                                          key_equal;
+    typedef _Alloc                                         allocator_type;
+    typedef std::pair<const key_type, mapped_type>         value_type;
+    typedef value_type&                                    reference;
+    typedef const value_type&                              const_reference;
+
+private:
+    typedef std::pair<key_type, mapped_type>               __value_type;
+    typedef __hash_map_hasher<__value_type, hasher>   __hasher;
+    typedef __hash_map_equal<__value_type, key_equal> __key_equal;
+    typedef typename std::__rebind_alloc_helper<std::allocator_traits<allocator_type>, __value_type>::type __allocator_type;
+
+    typedef std::__hash_table<__value_type, __hasher,
+                         __key_equal,  __allocator_type>   __table;
+
+    __table __table_;
+
+    typedef typename __table::__node_traits                __node_traits;
+    typedef typename __table::__node_allocator             __node_allocator;
+    typedef typename __table::__node                       __node;
+    typedef __hash_map_node_destructor<__node_allocator>   _Dp;
+    typedef std::unique_ptr<__node, _Dp>                         __node_holder;
+    typedef std::allocator_traits<allocator_type>               __alloc_traits;
+public:
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+    typedef typename __alloc_traits::size_type       size_type;
+    typedef typename __alloc_traits::difference_type difference_type;
+
+    typedef __hash_map_iterator<typename __table::iterator>       iterator;
+    typedef __hash_map_const_iterator<typename __table::const_iterator> const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hash_multimap() { }
+    explicit hash_multimap(size_type __n, const hasher& __hf = hasher(),
+                                const key_equal& __eql = key_equal());
+    hash_multimap(size_type __n, const hasher& __hf,
+                                const key_equal& __eql,
+                                const allocator_type& __a);
+    template <class _InputIterator>
+        hash_multimap(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        hash_multimap(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        hash_multimap(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf,
+                      const key_equal& __eql,
+                      const allocator_type& __a);
+    hash_multimap(const hash_multimap& __u);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin()        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end()          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const {return __table_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const value_type& __x) {return __table_.__insert_multi(__x);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator, const value_type& __x) {return insert(__x);}
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __p) {__table_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __first, const_iterator __last)
+        {__table_.erase(__first.__i_, __last.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() {__table_.clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(hash_multimap& __u) {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_funct() const
+        {return __table_.hash_function().hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const
+        {return __table_.key_eq().key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_multi(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type elems_in_bucket(size_type __n) const
+        {return __table_.bucket_size(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void resize(size_type __n) {__table_.rehash(__n);}
+};
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_multimap(
+        size_type __n, const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_multimap(
+        size_type __n, const hasher& __hf, const key_equal& __eql,
+        const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_multimap(
+        _InputIterator __first, _InputIterator __last)
+{
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_multimap(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_multimap(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::hash_multimap(
+        const hash_multimap& __u)
+    : __table_(__u.__table_)
+{
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                            _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_multi(*__first);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+     hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    typedef std::pair<const_iterator, const_iterator> _EqRng;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(); __i != __ex;)
+    {
+        _EqRng __xeq = __x.equal_range(__i->first);
+        _EqRng __yeq = __y.equal_range(__i->first);
+        if (_CUDA_VSTD::distance(__xeq.first, __xeq.second) !=
+            _CUDA_VSTD::distance(__yeq.first, __yeq.second) ||
+                  !_CUDA_VSTD::is_permutation(__xeq.first, __xeq.second, __yeq.first))
+            return false;
+        __i = __xeq.second;
+    }
+    return true;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const hash_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+} // __gnu_cxx
+
+#endif  // _LIBCUDACXX_HASH_MAP
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_set b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_set
new file mode 100644
index 000000000000..2b0594a80f7f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ext/hash_set
@@ -0,0 +1,659 @@
+// -*- C++ -*-
+//===------------------------- hash_set ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_HASH_SET
+#define _LIBCUDACXX_HASH_SET
+
+/*
+
+    hash_set synopsis
+
+namespace __gnu_cxx
+{
+
+template <class Value, class Hash = hash<Value>, class Pred = equal_to<Value>,
+          class Alloc = allocator<Value>>
+class hash_set
+{
+public:
+    // types
+    typedef Value                                                      key_type;
+    typedef key_type                                                   value_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+
+    explicit hash_set(size_type n = 193, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        hash_set(InputIterator f, InputIterator l,
+                      size_type n = 193, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    hash_set(const hash_set&);
+    ~hash_set();
+    hash_set& operator=(const hash_set&);
+
+    allocator_type get_allocator() const;
+
+    bool      empty() const;
+    size_type size() const;
+    size_type max_size() const;
+
+    iterator       begin();
+    iterator       end();
+    const_iterator begin()  const;
+    const_iterator end()    const;
+
+    pair<iterator, bool> insert(const value_type& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+
+    void erase(const_iterator position);
+    size_type erase(const key_type& k);
+    void erase(const_iterator first, const_iterator last);
+    void clear();
+
+    void swap(hash_set&);
+
+    hasher hash_funct() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    size_type bucket_count() const;
+    size_type max_bucket_count() const;
+
+    size_type elems_in_bucket(size_type n) const;
+
+    void resize(size_type n);
+};
+
+template <class Value, class Hash, class Pred, class Alloc>
+    void swap(hash_set<Value, Hash, Pred, Alloc>& x,
+              hash_set<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const hash_set<Value, Hash, Pred, Alloc>& x,
+               const hash_set<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const hash_set<Value, Hash, Pred, Alloc>& x,
+               const hash_set<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash = hash<Value>, class Pred = equal_to<Value>,
+          class Alloc = allocator<Value>>
+class hash_multiset
+{
+public:
+    // types
+    typedef Value                                                      key_type;
+    typedef key_type                                                   value_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+
+    explicit hash_multiset(size_type n = 193, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        hash_multiset(InputIterator f, InputIterator l,
+                      size_type n = 193, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    hash_multiset(const hash_multiset&);
+    ~hash_multiset();
+    hash_multiset& operator=(const hash_multiset&);
+
+    allocator_type get_allocator() const;
+
+    bool      empty() const;
+    size_type size() const;
+    size_type max_size() const;
+
+    iterator       begin();
+    iterator       end();
+    const_iterator begin()  const;
+    const_iterator end()    const;
+
+    iterator insert(const value_type& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+
+    void erase(const_iterator position);
+    size_type erase(const key_type& k);
+    void erase(const_iterator first, const_iterator last);
+    void clear();
+
+    void swap(hash_multiset&);
+
+    hasher hash_funct() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    size_type bucket_count() const;
+    size_type max_bucket_count() const;
+
+    size_type elems_in_bucket(size_type n) const;
+
+    void resize(size_type n);
+};
+
+template <class Value, class Hash, class Pred, class Alloc>
+    void swap(hash_multiset<Value, Hash, Pred, Alloc>& x,
+              hash_multiset<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const hash_multiset<Value, Hash, Pred, Alloc>& x,
+               const hash_multiset<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const hash_multiset<Value, Hash, Pred, Alloc>& x,
+               const hash_multiset<Value, Hash, Pred, Alloc>& y);
+}  // __gnu_cxx
+
+*/
+
+#include <__config>
+#include <__hash_table>
+#include <functional>
+#include <ext/__hash>
+
+#if __DEPRECATED
+#if defined(_LIBCUDACXX_WARNING)
+    _LIBCUDACXX_WARNING("Use of the header <ext/hash_set> is deprecated.  Migrate to <unordered_set>")
+#else
+#   warning Use of the header <ext/hash_set> is deprecated.  Migrate to <unordered_set>
+#endif
+#endif
+
+namespace __gnu_cxx {
+
+
+template <class _Value, class _Hash = hash<_Value>, class _Pred = std::equal_to<_Value>,
+          class _Alloc = std::allocator<_Value> >
+class _LIBCUDACXX_TEMPLATE_VIS hash_set
+{
+public:
+    // types
+    typedef _Value                                                     key_type;
+    typedef key_type                                                   value_type;
+    typedef _Hash                                                      hasher;
+    typedef _Pred                                                      key_equal;
+    typedef _Alloc                                                     allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+
+private:
+    typedef std::__hash_table<value_type, hasher, key_equal, allocator_type> __table;
+
+    __table __table_;
+
+public:
+    typedef typename __table::pointer         pointer;
+    typedef typename __table::const_pointer   const_pointer;
+    typedef typename __table::size_type       size_type;
+    typedef typename __table::difference_type difference_type;
+
+    typedef typename __table::const_iterator       iterator;
+    typedef typename __table::const_iterator       const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hash_set() { }
+    explicit hash_set(size_type __n, const hasher& __hf = hasher(),
+                           const key_equal& __eql = key_equal());
+    hash_set(size_type __n, const hasher& __hf, const key_equal& __eql,
+                  const allocator_type& __a);
+    template <class _InputIterator>
+        hash_set(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        hash_set(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        hash_set(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf, const key_equal& __eql,
+                      const allocator_type& __a);
+    hash_set(const hash_set& __u);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin()        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end()          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const {return __table_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<iterator, bool> insert(const value_type& __x)
+        {return __table_.__insert_unique(__x);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator, const value_type& __x) {return insert(__x).first;}
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __p) {__table_.erase(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __first, const_iterator __last)
+        {__table_.erase(__first, __last);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() {__table_.clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(hash_set& __u) {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_funct() const {return __table_.hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const {return __table_.key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_unique(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type elems_in_bucket(size_type __n) const {return __table_.bucket_size(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void resize(size_type __n) {__table_.rehash(__n);}
+};
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+hash_set<_Value, _Hash, _Pred, _Alloc>::hash_set(size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+hash_set<_Value, _Hash, _Pred, _Alloc>::hash_set(size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_set<_Value, _Hash, _Pred, _Alloc>::hash_set(
+        _InputIterator __first, _InputIterator __last)
+{
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_set<_Value, _Hash, _Pred, _Alloc>::hash_set(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_set<_Value, _Hash, _Pred, _Alloc>::hash_set(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+hash_set<_Value, _Hash, _Pred, _Alloc>::hash_set(
+        const hash_set& __u)
+    : __table_(__u.__table_)
+{
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+hash_set<_Value, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                    _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_unique(*__first);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(hash_set<_Value, _Hash, _Pred, _Alloc>& __x,
+     hash_set<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const hash_set<_Value, _Hash, _Pred, _Alloc>& __x,
+           const hash_set<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename hash_set<_Value, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(), __ey = __y.end();
+            __i != __ex; ++__i)
+    {
+        const_iterator __j = __y.find(*__i);
+        if (__j == __ey || !(*__i == *__j))
+            return false;
+    }
+    return true;
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const hash_set<_Value, _Hash, _Pred, _Alloc>& __x,
+           const hash_set<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Value, class _Hash = hash<_Value>, class _Pred = std::equal_to<_Value>,
+          class _Alloc = std::allocator<_Value> >
+class _LIBCUDACXX_TEMPLATE_VIS hash_multiset
+{
+public:
+    // types
+    typedef _Value                                                     key_type;
+    typedef key_type                                                   value_type;
+    typedef _Hash                                                      hasher;
+    typedef _Pred                                                      key_equal;
+    typedef _Alloc                                                     allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+
+private:
+    typedef std::__hash_table<value_type, hasher, key_equal, allocator_type> __table;
+
+    __table __table_;
+
+public:
+    typedef typename __table::pointer         pointer;
+    typedef typename __table::const_pointer   const_pointer;
+    typedef typename __table::size_type       size_type;
+    typedef typename __table::difference_type difference_type;
+
+    typedef typename __table::const_iterator       iterator;
+    typedef typename __table::const_iterator       const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hash_multiset() { }
+    explicit hash_multiset(size_type __n, const hasher& __hf = hasher(),
+                                const key_equal& __eql = key_equal());
+    hash_multiset(size_type __n, const hasher& __hf,
+                       const key_equal& __eql, const allocator_type& __a);
+    template <class _InputIterator>
+        hash_multiset(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        hash_multiset(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        hash_multiset(_InputIterator __first, _InputIterator __last,
+                      size_type __n , const hasher& __hf,
+                      const key_equal& __eql, const allocator_type& __a);
+    hash_multiset(const hash_multiset& __u);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin()        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end()          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const {return __table_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const value_type& __x) {return __table_.__insert_multi(__x);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator, const value_type& __x) {return insert(__x);}
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __p) {__table_.erase(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void erase(const_iterator __first, const_iterator __last)
+        {__table_.erase(__first, __last);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() {__table_.clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(hash_multiset& __u) {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_funct() const {return __table_.hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const {return __table_.key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    std::pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_multi(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type elems_in_bucket(size_type __n) const {return __table_.bucket_size(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void resize(size_type __n) {__table_.rehash(__n);}
+};
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::hash_multiset(
+        size_type __n, const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::hash_multiset(
+        size_type __n, const hasher& __hf, const key_equal& __eql,
+        const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::hash_multiset(
+        _InputIterator __first, _InputIterator __last)
+{
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::hash_multiset(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::hash_multiset(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::hash_multiset(
+        const hash_multiset& __u)
+    : __table_(__u.__table_)
+{
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+hash_multiset<_Value, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                    _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_multi(*__first);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(hash_multiset<_Value, _Hash, _Pred, _Alloc>& __x,
+     hash_multiset<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const hash_multiset<_Value, _Hash, _Pred, _Alloc>& __x,
+           const hash_multiset<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename hash_multiset<_Value, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    typedef std::pair<const_iterator, const_iterator> _EqRng;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(); __i != __ex;)
+    {
+        _EqRng __xeq = __x.equal_range(*__i);
+        _EqRng __yeq = __y.equal_range(*__i);
+        if (_CUDA_VSTD::distance(__xeq.first, __xeq.second) !=
+            _CUDA_VSTD::distance(__yeq.first, __yeq.second) ||
+                  !_CUDA_VSTD::is_permutation(__xeq.first, __xeq.second, __yeq.first))
+            return false;
+        __i = __xeq.second;
+    }
+    return true;
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const hash_multiset<_Value, _Hash, _Pred, _Alloc>& __x,
+           const hash_multiset<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+} // __gnu_cxx
+
+#endif  // _LIBCUDACXX_HASH_SET
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h
new file mode 100644
index 000000000000..30b1eff084da
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fenv.h
@@ -0,0 +1,116 @@
+// -*- C++ -*-
+//===---------------------------- math.h ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FENV_H
+#define _LIBCUDACXX_FENV_H
+
+
+/*
+    fenv.h synopsis
+
+This entire header is C99 / C++0X
+
+Macros:
+
+    FE_DIVBYZERO
+    FE_INEXACT
+    FE_INVALID
+    FE_OVERFLOW
+    FE_UNDERFLOW
+    FE_ALL_EXCEPT
+    FE_DOWNWARD
+    FE_TONEAREST
+    FE_TOWARDZERO
+    FE_UPWARD
+    FE_DFL_ENV
+
+Types:
+
+    fenv_t
+    fexcept_t
+
+int feclearexcept(int excepts);
+int fegetexceptflag(fexcept_t* flagp, int excepts);
+int feraiseexcept(int excepts);
+int fesetexceptflag(const fexcept_t* flagp, int excepts);
+int fetestexcept(int excepts);
+int fegetround();
+int fesetround(int round);
+int fegetenv(fenv_t* envp);
+int feholdexcept(fenv_t* envp);
+int fesetenv(const fenv_t* envp);
+int feupdateenv(const fenv_t* envp);
+
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <fenv.h>
+
+#ifdef __cplusplus
+
+extern "C++" {
+
+#ifdef feclearexcept
+#undef feclearexcept
+#endif
+
+#ifdef fegetexceptflag
+#undef fegetexceptflag
+#endif
+
+
+#ifdef feraiseexcept
+#undef feraiseexcept
+#endif
+
+#ifdef fesetexceptflag
+#undef fesetexceptflag
+#endif
+
+
+#ifdef fetestexcept
+#undef fetestexcept
+#endif
+
+#ifdef fegetround
+#undef fegetround
+#endif
+
+#ifdef fesetround
+#undef fesetround
+#endif
+
+#ifdef fegetenv
+#undef fegetenv
+#endif
+
+#ifdef feholdexcept
+#undef feholdexcept
+#endif
+
+
+#ifdef fesetenv
+#undef fesetenv
+#endif
+
+#ifdef feupdateenv
+#undef feupdateenv
+#endif
+
+} // extern "C++"
+
+#endif // defined(__cplusplus)
+
+#endif // _LIBCUDACXX_FENV_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem
new file mode 100644
index 000000000000..1a1f25d95e4d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/filesystem
@@ -0,0 +1,2644 @@
+// -*- C++ -*-
+//===--------------------------- filesystem -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef _LIBCUDACXX_FILESYSTEM
+#define _LIBCUDACXX_FILESYSTEM
+/*
+    filesystem synopsis
+
+    namespace std { namespace filesystem {
+
+    class path;
+
+    void swap(path& lhs, path& rhs) noexcept;
+    size_t hash_value(const path& p) noexcept;
+
+    bool operator==(const path& lhs, const path& rhs) noexcept;
+    bool operator!=(const path& lhs, const path& rhs) noexcept;
+    bool operator< (const path& lhs, const path& rhs) noexcept;
+    bool operator<=(const path& lhs, const path& rhs) noexcept;
+    bool operator> (const path& lhs, const path& rhs) noexcept;
+    bool operator>=(const path& lhs, const path& rhs) noexcept;
+
+    path operator/ (const path& lhs, const path& rhs);
+
+    // fs.path.io operators are friends of path.
+    template <class charT, class traits>
+    friend basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os, const path& p);
+
+    template <class charT, class traits>
+    friend basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is, path& p);
+
+    template <class Source>
+      path u8path(const Source& source);
+    template <class InputIterator>
+      path u8path(InputIterator first, InputIterator last);
+
+    class filesystem_error;
+    class directory_entry;
+
+    class directory_iterator;
+
+    // enable directory_iterator range-based for statements
+    directory_iterator begin(directory_iterator iter) noexcept;
+    directory_iterator end(const directory_iterator&) noexcept;
+
+    class recursive_directory_iterator;
+
+    // enable recursive_directory_iterator range-based for statements
+    recursive_directory_iterator begin(recursive_directory_iterator iter) noexcept;
+    recursive_directory_iterator end(const recursive_directory_iterator&) noexcept;
+
+    class file_status;
+
+    struct space_info
+    {
+      uintmax_t capacity;
+      uintmax_t free;
+      uintmax_t available;
+    };
+
+    enum class file_type;
+    enum class perms;
+    enum class perm_options;
+    enum class copy_options;
+    enum class directory_options;
+
+    typedef chrono::time_point<trivial-clock>  file_time_type;
+
+    // operational functions
+
+    path absolute(const path& p);
+    path absolute(const path& p, error_code &ec);
+
+    path canonical(const path& p);
+    path canonical(const path& p, error_code& ec);
+
+    void copy(const path& from, const path& to);
+    void copy(const path& from, const path& to, error_code& ec);
+    void copy(const path& from, const path& to, copy_options options);
+    void copy(const path& from, const path& to, copy_options options,
+                   error_code& ec);
+
+    bool copy_file(const path& from, const path& to);
+    bool copy_file(const path& from, const path& to, error_code& ec);
+    bool copy_file(const path& from, const path& to, copy_options option);
+    bool copy_file(const path& from, const path& to, copy_options option,
+                           error_code& ec);
+
+    void copy_symlink(const path& existing_symlink, const path& new_symlink);
+    void copy_symlink(const path& existing_symlink, const path& new_symlink,
+                              error_code& ec) noexcept;
+
+    bool create_directories(const path& p);
+    bool create_directories(const path& p, error_code& ec);
+
+    bool create_directory(const path& p);
+    bool create_directory(const path& p, error_code& ec) noexcept;
+
+    bool create_directory(const path& p, const path& attributes);
+    bool create_directory(const path& p, const path& attributes,
+                                  error_code& ec) noexcept;
+
+    void create_directory_symlink(const path& to, const path& new_symlink);
+    void create_directory_symlink(const path& to, const path& new_symlink,
+                                          error_code& ec) noexcept;
+
+    void create_hard_link(const path& to, const path& new_hard_link);
+    void create_hard_link(const path& to, const path& new_hard_link,
+                                  error_code& ec) noexcept;
+
+    void create_symlink(const path& to, const path& new_symlink);
+    void create_symlink(const path& to, const path& new_symlink,
+                                error_code& ec) noexcept;
+
+    path current_path();
+    path current_path(error_code& ec);
+    void current_path(const path& p);
+    void current_path(const path& p, error_code& ec) noexcept;
+
+    bool exists(file_status s) noexcept;
+    bool exists(const path& p);
+    bool exists(const path& p, error_code& ec) noexcept;
+
+    bool equivalent(const path& p1, const path& p2);
+    bool equivalent(const path& p1, const path& p2, error_code& ec) noexcept;
+
+    uintmax_t    file_size(const path& p);
+    uintmax_t    file_size(const path& p, error_code& ec) noexcept;
+
+    uintmax_t    hard_link_count(const path& p);
+    uintmax_t    hard_link_count(const path& p, error_code& ec) noexcept;
+
+    bool is_block_file(file_status s) noexcept;
+    bool is_block_file(const path& p);
+    bool is_block_file(const path& p, error_code& ec) noexcept;
+
+    bool is_character_file(file_status s) noexcept;
+    bool is_character_file(const path& p);
+    bool is_character_file(const path& p, error_code& ec) noexcept;
+
+    bool is_directory(file_status s) noexcept;
+    bool is_directory(const path& p);
+    bool is_directory(const path& p, error_code& ec) noexcept;
+
+    bool is_empty(const path& p);
+    bool is_empty(const path& p, error_code& ec) noexcept;
+
+    bool is_fifo(file_status s) noexcept;
+    bool is_fifo(const path& p);
+    bool is_fifo(const path& p, error_code& ec) noexcept;
+
+    bool is_other(file_status s) noexcept;
+    bool is_other(const path& p);
+    bool is_other(const path& p, error_code& ec) noexcept;
+
+    bool is_regular_file(file_status s) noexcept;
+    bool is_regular_file(const path& p);
+    bool is_regular_file(const path& p, error_code& ec) noexcept;
+
+    bool is_socket(file_status s) noexcept;
+    bool is_socket(const path& p);
+    bool is_socket(const path& p, error_code& ec) noexcept;
+
+    bool is_symlink(file_status s) noexcept;
+    bool is_symlink(const path& p);
+    bool is_symlink(const path& p, error_code& ec) noexcept;
+
+    file_time_type  last_write_time(const path& p);
+    file_time_type  last_write_time(const path& p, error_code& ec) noexcept;
+    void last_write_time(const path& p, file_time_type new_time);
+    void last_write_time(const path& p, file_time_type new_time,
+                                 error_code& ec) noexcept;
+
+    void permissions(const path& p, perms prms,
+                     perm_options opts=perm_options::replace);
+    void permissions(const path& p, perms prms, error_code& ec) noexcept;
+    void permissions(const path& p, perms prms, perm_options opts,
+                     error_code& ec);
+
+    path proximate(const path& p, error_code& ec);
+    path proximate(const path& p, const path& base = current_path());
+    path proximate(const path& p, const path& base, error_code &ec);
+
+    path read_symlink(const path& p);
+    path read_symlink(const path& p, error_code& ec);
+
+    path relative(const path& p, error_code& ec);
+    path relative(const path& p, const path& base=current_path());
+    path relative(const path& p, const path& base, error_code& ec);
+
+    bool remove(const path& p);
+    bool remove(const path& p, error_code& ec) noexcept;
+
+    uintmax_t    remove_all(const path& p);
+    uintmax_t    remove_all(const path& p, error_code& ec);
+
+    void rename(const path& from, const path& to);
+    void rename(const path& from, const path& to, error_code& ec) noexcept;
+
+    void resize_file(const path& p, uintmax_t size);
+    void resize_file(const path& p, uintmax_t size, error_code& ec) noexcept;
+
+    space_info   space(const path& p);
+    space_info   space(const path& p, error_code& ec) noexcept;
+
+    file_status  status(const path& p);
+    file_status  status(const path& p, error_code& ec) noexcept;
+
+    bool status_known(file_status s) noexcept;
+
+    file_status  symlink_status(const path& p);
+    file_status  symlink_status(const path& p, error_code& ec) noexcept;
+
+    path temp_directory_path();
+    path temp_directory_path(error_code& ec);
+
+    path weakly_canonical(path const& p);
+    path weakly_canonical(path const& p, error_code& ec);
+
+
+} }  // namespaces std::filesystem
+
+*/
+
+#include <__config>
+#include <cstddef>
+#include <cstdlib>
+#include <chrono>
+#include <iterator>
+#include <iosfwd>
+#include <locale>
+#include <memory>
+#include <stack>
+#include <string>
+#include <system_error>
+#include <utility>
+#include <iomanip> // for quoted
+#include <string_view>
+#include <version>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_BEGIN_NAMESPACE_FILESYSTEM
+
+_LIBCUDACXX_AVAILABILITY_FILESYSTEM_PUSH
+
+typedef chrono::time_point<_FilesystemClock> file_time_type;
+
+struct _LIBCUDACXX_TYPE_VIS space_info {
+  uintmax_t capacity;
+  uintmax_t free;
+  uintmax_t available;
+};
+
+enum class _LIBCUDACXX_ENUM_VIS file_type : signed char {
+  none = 0,
+  not_found = -1,
+  regular = 1,
+  directory = 2,
+  symlink = 3,
+  block = 4,
+  character = 5,
+  fifo = 6,
+  socket = 7,
+  unknown = 8
+};
+
+enum class _LIBCUDACXX_ENUM_VIS perms : unsigned {
+  none = 0,
+
+  owner_read = 0400,
+  owner_write = 0200,
+  owner_exec = 0100,
+  owner_all = 0700,
+
+  group_read = 040,
+  group_write = 020,
+  group_exec = 010,
+  group_all = 070,
+
+  others_read = 04,
+  others_write = 02,
+  others_exec = 01,
+  others_all = 07,
+
+  all = 0777,
+
+  set_uid = 04000,
+  set_gid = 02000,
+  sticky_bit = 01000,
+  mask = 07777,
+  unknown = 0xFFFF,
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perms operator&(perms _LHS, perms _RHS) {
+  return static_cast<perms>(static_cast<unsigned>(_LHS) &
+                            static_cast<unsigned>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perms operator|(perms _LHS, perms _RHS) {
+  return static_cast<perms>(static_cast<unsigned>(_LHS) |
+                            static_cast<unsigned>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perms operator^(perms _LHS, perms _RHS) {
+  return static_cast<perms>(static_cast<unsigned>(_LHS) ^
+                            static_cast<unsigned>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perms operator~(perms _LHS) {
+  return static_cast<perms>(~static_cast<unsigned>(_LHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline perms& operator&=(perms& _LHS, perms _RHS) { return _LHS = _LHS & _RHS; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline perms& operator|=(perms& _LHS, perms _RHS) { return _LHS = _LHS | _RHS; }
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline perms& operator^=(perms& _LHS, perms _RHS) { return _LHS = _LHS ^ _RHS; }
+
+enum class _LIBCUDACXX_ENUM_VIS perm_options : unsigned char {
+  replace = 1,
+  add = 2,
+  remove = 4,
+  nofollow = 8
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perm_options operator&(perm_options _LHS, perm_options _RHS) {
+  return static_cast<perm_options>(static_cast<unsigned>(_LHS) &
+                                   static_cast<unsigned>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perm_options operator|(perm_options _LHS, perm_options _RHS) {
+  return static_cast<perm_options>(static_cast<unsigned>(_LHS) |
+                                   static_cast<unsigned>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perm_options operator^(perm_options _LHS, perm_options _RHS) {
+  return static_cast<perm_options>(static_cast<unsigned>(_LHS) ^
+                                   static_cast<unsigned>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr perm_options operator~(perm_options _LHS) {
+  return static_cast<perm_options>(~static_cast<unsigned>(_LHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline perm_options& operator&=(perm_options& _LHS, perm_options _RHS) {
+  return _LHS = _LHS & _RHS;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline perm_options& operator|=(perm_options& _LHS, perm_options _RHS) {
+  return _LHS = _LHS | _RHS;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline perm_options& operator^=(perm_options& _LHS, perm_options _RHS) {
+  return _LHS = _LHS ^ _RHS;
+}
+
+enum class _LIBCUDACXX_ENUM_VIS copy_options : unsigned short {
+  none = 0,
+  skip_existing = 1,
+  overwrite_existing = 2,
+  update_existing = 4,
+  recursive = 8,
+  copy_symlinks = 16,
+  skip_symlinks = 32,
+  directories_only = 64,
+  create_symlinks = 128,
+  create_hard_links = 256,
+  __in_recursive_copy = 512,
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr copy_options operator&(copy_options _LHS, copy_options _RHS) {
+  return static_cast<copy_options>(static_cast<unsigned short>(_LHS) &
+                                   static_cast<unsigned short>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr copy_options operator|(copy_options _LHS, copy_options _RHS) {
+  return static_cast<copy_options>(static_cast<unsigned short>(_LHS) |
+                                   static_cast<unsigned short>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr copy_options operator^(copy_options _LHS, copy_options _RHS) {
+  return static_cast<copy_options>(static_cast<unsigned short>(_LHS) ^
+                                   static_cast<unsigned short>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr copy_options operator~(copy_options _LHS) {
+  return static_cast<copy_options>(~static_cast<unsigned short>(_LHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline copy_options& operator&=(copy_options& _LHS, copy_options _RHS) {
+  return _LHS = _LHS & _RHS;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline copy_options& operator|=(copy_options& _LHS, copy_options _RHS) {
+  return _LHS = _LHS | _RHS;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline copy_options& operator^=(copy_options& _LHS, copy_options _RHS) {
+  return _LHS = _LHS ^ _RHS;
+}
+
+enum class _LIBCUDACXX_ENUM_VIS directory_options : unsigned char {
+  none = 0,
+  follow_directory_symlink = 1,
+  skip_permission_denied = 2
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr directory_options operator&(directory_options _LHS,
+                                             directory_options _RHS) {
+  return static_cast<directory_options>(static_cast<unsigned char>(_LHS) &
+                                        static_cast<unsigned char>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr directory_options operator|(directory_options _LHS,
+                                             directory_options _RHS) {
+  return static_cast<directory_options>(static_cast<unsigned char>(_LHS) |
+                                        static_cast<unsigned char>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr directory_options operator^(directory_options _LHS,
+                                             directory_options _RHS) {
+  return static_cast<directory_options>(static_cast<unsigned char>(_LHS) ^
+                                        static_cast<unsigned char>(_RHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr directory_options operator~(directory_options _LHS) {
+  return static_cast<directory_options>(~static_cast<unsigned char>(_LHS));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline directory_options& operator&=(directory_options& _LHS,
+                                     directory_options _RHS) {
+  return _LHS = _LHS & _RHS;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline directory_options& operator|=(directory_options& _LHS,
+                                     directory_options _RHS) {
+  return _LHS = _LHS | _RHS;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline directory_options& operator^=(directory_options& _LHS,
+                                     directory_options _RHS) {
+  return _LHS = _LHS ^ _RHS;
+}
+
+class _LIBCUDACXX_TYPE_VIS file_status {
+public:
+  // constructors
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status() noexcept : file_status(file_type::none) {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit file_status(file_type __ft, perms __prms = perms::unknown) noexcept
+      : __ft_(__ft),
+        __prms_(__prms) {}
+
+  file_status(const file_status&) noexcept = default;
+  file_status(file_status&&) noexcept = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  ~file_status() {}
+
+  file_status& operator=(const file_status&) noexcept = default;
+  file_status& operator=(file_status&&) noexcept = default;
+
+  // observers
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_type type() const noexcept { return __ft_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  perms permissions() const noexcept { return __prms_; }
+
+  // modifiers
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void type(file_type __ft) noexcept { __ft_ = __ft; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void permissions(perms __p) noexcept { __prms_ = __p; }
+
+private:
+  file_type __ft_;
+  perms __prms_;
+};
+
+class _LIBCUDACXX_TYPE_VIS directory_entry;
+
+template <class _Tp>
+struct __can_convert_char {
+  static const bool value = false;
+};
+template <class _Tp>
+struct __can_convert_char<const _Tp> : public __can_convert_char<_Tp> {};
+template <>
+struct __can_convert_char<char> {
+  static const bool value = true;
+  using __char_type = char;
+};
+template <>
+struct __can_convert_char<wchar_t> {
+  static const bool value = true;
+  using __char_type = wchar_t;
+};
+template <>
+struct __can_convert_char<char16_t> {
+  static const bool value = true;
+  using __char_type = char16_t;
+};
+template <>
+struct __can_convert_char<char32_t> {
+  static const bool value = true;
+  using __char_type = char32_t;
+};
+
+template <class _ECharT>
+typename enable_if<__can_convert_char<_ECharT>::value, bool>::type
+__is_separator(_ECharT __e) {
+  return __e == _ECharT('/');
+}
+
+struct _NullSentinal {};
+
+template <class _Tp>
+using _Void = void;
+
+template <class _Tp, class = void>
+struct __is_pathable_string : public false_type {};
+
+template <class _ECharT, class _Traits, class _Alloc>
+struct __is_pathable_string<
+    basic_string<_ECharT, _Traits, _Alloc>,
+    _Void<typename __can_convert_char<_ECharT>::__char_type> >
+    : public __can_convert_char<_ECharT> {
+  using _Str = basic_string<_ECharT, _Traits, _Alloc>;
+  using _Base = __can_convert_char<_ECharT>;
+  static _ECharT const* __range_begin(_Str const& __s) { return __s.data(); }
+  static _ECharT const* __range_end(_Str const& __s) {
+    return __s.data() + __s.length();
+  }
+  static _ECharT __first_or_null(_Str const& __s) {
+    return __s.empty() ? _ECharT{} : __s[0];
+  }
+};
+
+template <class _ECharT, class _Traits>
+struct __is_pathable_string<
+    basic_string_view<_ECharT, _Traits>,
+    _Void<typename __can_convert_char<_ECharT>::__char_type> >
+    : public __can_convert_char<_ECharT> {
+  using _Str = basic_string_view<_ECharT, _Traits>;
+  using _Base = __can_convert_char<_ECharT>;
+  static _ECharT const* __range_begin(_Str const& __s) { return __s.data(); }
+  static _ECharT const* __range_end(_Str const& __s) {
+    return __s.data() + __s.length();
+  }
+  static _ECharT __first_or_null(_Str const& __s) {
+    return __s.empty() ? _ECharT{} : __s[0];
+  }
+};
+
+template <class _Source, class _DS = typename decay<_Source>::type,
+          class _UnqualPtrType =
+              typename remove_const<typename remove_pointer<_DS>::type>::type,
+          bool _IsCharPtr = is_pointer<_DS>::value&&
+              __can_convert_char<_UnqualPtrType>::value>
+struct __is_pathable_char_array : false_type {};
+
+template <class _Source, class _ECharT, class _UPtr>
+struct __is_pathable_char_array<_Source, _ECharT*, _UPtr, true>
+    : __can_convert_char<typename remove_const<_ECharT>::type> {
+  using _Base = __can_convert_char<typename remove_const<_ECharT>::type>;
+
+  static _ECharT const* __range_begin(const _ECharT* __b) { return __b; }
+  static _ECharT const* __range_end(const _ECharT* __b) {
+    using _Iter = const _ECharT*;
+    const _ECharT __sentinal = _ECharT{};
+    _Iter __e = __b;
+    for (; *__e != __sentinal; ++__e)
+      ;
+    return __e;
+  }
+
+  static _ECharT __first_or_null(const _ECharT* __b) { return *__b; }
+};
+
+template <class _Iter, bool _IsIt = __is_input_iterator<_Iter>::value,
+          class = void>
+struct __is_pathable_iter : false_type {};
+
+template <class _Iter>
+struct __is_pathable_iter<
+    _Iter, true,
+    _Void<typename __can_convert_char<
+        typename iterator_traits<_Iter>::value_type>::__char_type> >
+    : __can_convert_char<typename iterator_traits<_Iter>::value_type> {
+  using _ECharT = typename iterator_traits<_Iter>::value_type;
+  using _Base = __can_convert_char<_ECharT>;
+
+  static _Iter __range_begin(_Iter __b) { return __b; }
+  static _NullSentinal __range_end(_Iter) { return _NullSentinal{}; }
+
+  static _ECharT __first_or_null(_Iter __b) { return *__b; }
+};
+
+template <class _Tp, bool _IsStringT = __is_pathable_string<_Tp>::value,
+          bool _IsCharIterT = __is_pathable_char_array<_Tp>::value,
+          bool _IsIterT = !_IsCharIterT && __is_pathable_iter<_Tp>::value>
+struct __is_pathable : false_type {
+  static_assert(!_IsStringT && !_IsCharIterT && !_IsIterT, "Must all be false");
+};
+
+template <class _Tp>
+struct __is_pathable<_Tp, true, false, false> : __is_pathable_string<_Tp> {};
+
+template <class _Tp>
+struct __is_pathable<_Tp, false, true, false> : __is_pathable_char_array<_Tp> {
+};
+
+template <class _Tp>
+struct __is_pathable<_Tp, false, false, true> : __is_pathable_iter<_Tp> {};
+
+template <class _ECharT>
+struct _PathCVT {
+  static_assert(__can_convert_char<_ECharT>::value,
+                "Char type not convertible");
+
+  typedef __narrow_to_utf8<sizeof(_ECharT) * __CHAR_BIT__> _Narrower;
+
+  static void __append_range(string& __dest, _ECharT const* __b,
+                             _ECharT const* __e) {
+    _Narrower()(back_inserter(__dest), __b, __e);
+  }
+
+  template <class _Iter>
+  static void __append_range(string& __dest, _Iter __b, _Iter __e) {
+    static_assert(!is_same<_Iter, _ECharT*>::value, "Call const overload");
+    if (__b == __e)
+      return;
+    basic_string<_ECharT> __tmp(__b, __e);
+    _Narrower()(back_inserter(__dest), __tmp.data(),
+                __tmp.data() + __tmp.length());
+  }
+
+  template <class _Iter>
+  static void __append_range(string& __dest, _Iter __b, _NullSentinal) {
+    static_assert(!is_same<_Iter, _ECharT*>::value, "Call const overload");
+    const _ECharT __sentinal = _ECharT{};
+    if (*__b == __sentinal)
+      return;
+    basic_string<_ECharT> __tmp;
+    for (; *__b != __sentinal; ++__b)
+      __tmp.push_back(*__b);
+    _Narrower()(back_inserter(__dest), __tmp.data(),
+                __tmp.data() + __tmp.length());
+  }
+
+  template <class _Source>
+  static void __append_source(string& __dest, _Source const& __s) {
+    using _Traits = __is_pathable<_Source>;
+    __append_range(__dest, _Traits::__range_begin(__s),
+                   _Traits::__range_end(__s));
+  }
+};
+
+template <>
+struct _PathCVT<char> {
+
+  template <class _Iter>
+  static typename enable_if<__is_exactly_input_iterator<_Iter>::value>::type
+  __append_range(string& __dest, _Iter __b, _Iter __e) {
+    for (; __b != __e; ++__b)
+      __dest.push_back(*__b);
+  }
+
+  template <class _Iter>
+  static typename enable_if<__is_forward_iterator<_Iter>::value>::type
+  __append_range(string& __dest, _Iter __b, _Iter __e) {
+    __dest.__append_forward_unsafe(__b, __e);
+  }
+
+  template <class _Iter>
+  static void __append_range(string& __dest, _Iter __b, _NullSentinal) {
+    const char __sentinal = char{};
+    for (; *__b != __sentinal; ++__b)
+      __dest.push_back(*__b);
+  }
+
+  template <class _Source>
+  static void __append_source(string& __dest, _Source const& __s) {
+    using _Traits = __is_pathable<_Source>;
+    __append_range(__dest, _Traits::__range_begin(__s),
+                   _Traits::__range_end(__s));
+  }
+};
+
+class _LIBCUDACXX_TYPE_VIS path {
+  template <class _SourceOrIter, class _Tp = path&>
+  using _EnableIfPathable =
+      typename enable_if<__is_pathable<_SourceOrIter>::value, _Tp>::type;
+
+  template <class _Tp>
+  using _SourceChar = typename __is_pathable<_Tp>::__char_type;
+
+  template <class _Tp>
+  using _SourceCVT = _PathCVT<_SourceChar<_Tp> >;
+
+public:
+  typedef char value_type;
+  typedef basic_string<value_type> string_type;
+  typedef _CUDA_VSTD::string_view __string_view;
+  static constexpr value_type preferred_separator = '/';
+
+  enum class _LIBCUDACXX_ENUM_VIS format : unsigned char {
+    auto_format,
+    native_format,
+    generic_format
+  };
+
+  // constructors and destructor
+  _LIBCUDACXX_INLINE_VISIBILITY path() noexcept {}
+  _LIBCUDACXX_INLINE_VISIBILITY path(const path& __p) : __pn_(__p.__pn_) {}
+  _LIBCUDACXX_INLINE_VISIBILITY path(path&& __p) noexcept
+      : __pn_(_CUDA_VSTD::move(__p.__pn_)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path(string_type&& __s, format = format::auto_format) noexcept
+      : __pn_(_CUDA_VSTD::move(__s)) {}
+
+  template <class _Source, class = _EnableIfPathable<_Source, void> >
+  path(const _Source& __src, format = format::auto_format) {
+    _SourceCVT<_Source>::__append_source(__pn_, __src);
+  }
+
+  template <class _InputIt>
+  path(_InputIt __first, _InputIt __last, format = format::auto_format) {
+    typedef typename iterator_traits<_InputIt>::value_type _ItVal;
+    _PathCVT<_ItVal>::__append_range(__pn_, __first, __last);
+  }
+
+  // TODO Implement locale conversions.
+  template <class _Source, class = _EnableIfPathable<_Source, void> >
+  path(const _Source& __src, const locale& __loc, format = format::auto_format);
+  template <class _InputIt>
+  path(_InputIt __first, _InputIt _last, const locale& __loc,
+       format = format::auto_format);
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  ~path() = default;
+
+  // assignments
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator=(const path& __p) {
+    __pn_ = __p.__pn_;
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator=(path&& __p) noexcept {
+    __pn_ = _CUDA_VSTD::move(__p.__pn_);
+    return *this;
+  }
+
+  template <class = void>
+  _LIBCUDACXX_INLINE_VISIBILITY path& operator=(string_type&& __s) noexcept {
+    __pn_ = _CUDA_VSTD::move(__s);
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& assign(string_type&& __s) noexcept {
+    __pn_ = _CUDA_VSTD::move(__s);
+    return *this;
+  }
+
+  template <class _Source>
+  _LIBCUDACXX_INLINE_VISIBILITY _EnableIfPathable<_Source>
+  operator=(const _Source& __src) {
+    return this->assign(__src);
+  }
+
+  template <class _Source>
+  _EnableIfPathable<_Source> assign(const _Source& __src) {
+    __pn_.clear();
+    _SourceCVT<_Source>::__append_source(__pn_, __src);
+    return *this;
+  }
+
+  template <class _InputIt>
+  path& assign(_InputIt __first, _InputIt __last) {
+    typedef typename iterator_traits<_InputIt>::value_type _ItVal;
+    __pn_.clear();
+    _PathCVT<_ItVal>::__append_range(__pn_, __first, __last);
+    return *this;
+  }
+
+private:
+  template <class _ECharT>
+  static bool __source_is_absolute(_ECharT __first_or_null) {
+    return __is_separator(__first_or_null);
+  }
+
+public:
+  // appends
+  path& operator/=(const path& __p) {
+    if (__p.is_absolute()) {
+      __pn_ = __p.__pn_;
+      return *this;
+    }
+    if (has_filename())
+      __pn_ += preferred_separator;
+    __pn_ += __p.native();
+    return *this;
+  }
+
+  // FIXME: Use _LIBCUDACXX_DIAGNOSE_WARNING to produce a diagnostic when __src
+  // is known at compile time to be "/' since the user almost certainly intended
+  // to append a separator instead of overwriting the path with "/"
+  template <class _Source>
+  _LIBCUDACXX_INLINE_VISIBILITY _EnableIfPathable<_Source>
+  operator/=(const _Source& __src) {
+    return this->append(__src);
+  }
+
+  template <class _Source>
+  _EnableIfPathable<_Source> append(const _Source& __src) {
+    using _Traits = __is_pathable<_Source>;
+    using _CVT = _PathCVT<_SourceChar<_Source> >;
+    if (__source_is_absolute(_Traits::__first_or_null(__src)))
+      __pn_.clear();
+    else if (has_filename())
+      __pn_ += preferred_separator;
+    _CVT::__append_source(__pn_, __src);
+    return *this;
+  }
+
+  template <class _InputIt>
+  path& append(_InputIt __first, _InputIt __last) {
+    typedef typename iterator_traits<_InputIt>::value_type _ItVal;
+    static_assert(__can_convert_char<_ItVal>::value, "Must convertible");
+    using _CVT = _PathCVT<_ItVal>;
+    if (__first != __last && __source_is_absolute(*__first))
+      __pn_.clear();
+    else if (has_filename())
+      __pn_ += preferred_separator;
+    _CVT::__append_range(__pn_, __first, __last);
+    return *this;
+  }
+
+  // concatenation
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator+=(const path& __x) {
+    __pn_ += __x.__pn_;
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator+=(const string_type& __x) {
+    __pn_ += __x;
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator+=(__string_view __x) {
+    __pn_ += __x;
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator+=(const value_type* __x) {
+    __pn_ += __x;
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& operator+=(value_type __x) {
+    __pn_ += __x;
+    return *this;
+  }
+
+  template <class _ECharT>
+  typename enable_if<__can_convert_char<_ECharT>::value, path&>::type
+  operator+=(_ECharT __x) {
+    basic_string<_ECharT> __tmp;
+    __tmp += __x;
+    _PathCVT<_ECharT>::__append_source(__pn_, __tmp);
+    return *this;
+  }
+
+  template <class _Source>
+  _EnableIfPathable<_Source> operator+=(const _Source& __x) {
+    return this->concat(__x);
+  }
+
+  template <class _Source>
+  _EnableIfPathable<_Source> concat(const _Source& __x) {
+    _SourceCVT<_Source>::__append_source(__pn_, __x);
+    return *this;
+  }
+
+  template <class _InputIt>
+  path& concat(_InputIt __first, _InputIt __last) {
+    typedef typename iterator_traits<_InputIt>::value_type _ItVal;
+    _PathCVT<_ItVal>::__append_range(__pn_, __first, __last);
+    return *this;
+  }
+
+  // modifiers
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void clear() noexcept { __pn_.clear(); }
+
+  path& make_preferred() { return *this; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  path& remove_filename() {
+    auto __fname = __filename();
+    if (!__fname.empty())
+      __pn_.erase(__fname.data() - __pn_.data());
+    return *this;
+  }
+
+  path& replace_filename(const path& __replacement) {
+    remove_filename();
+    return (*this /= __replacement);
+  }
+
+  path& replace_extension(const path& __replacement = path());
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void swap(path& __rhs) noexcept { __pn_.swap(__rhs.__pn_); }
+
+  // private helper to allow reserving memory in the path
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void __reserve(size_t __s) { __pn_.reserve(__s); }
+
+  // native format observers
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const string_type& native() const noexcept { return __pn_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const value_type* c_str() const noexcept { return __pn_.c_str(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY operator string_type() const { return __pn_; }
+
+  template <class _ECharT, class _Traits = char_traits<_ECharT>,
+            class _Allocator = allocator<_ECharT> >
+  basic_string<_ECharT, _Traits, _Allocator>
+  string(const _Allocator& __a = _Allocator()) const {
+    using _CVT = __widen_from_utf8<sizeof(_ECharT) * __CHAR_BIT__>;
+    using _Str = basic_string<_ECharT, _Traits, _Allocator>;
+    _Str __s(__a);
+    __s.reserve(__pn_.size());
+    _CVT()(back_inserter(__s), __pn_.data(), __pn_.data() + __pn_.size());
+    return __s;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY std::string string() const { return __pn_; }
+  _LIBCUDACXX_INLINE_VISIBILITY std::wstring wstring() const {
+    return string<wchar_t>();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY std::string u8string() const { return __pn_; }
+  _LIBCUDACXX_INLINE_VISIBILITY std::u16string u16string() const {
+    return string<char16_t>();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY std::u32string u32string() const {
+    return string<char32_t>();
+  }
+
+  // generic format observers
+  template <class _ECharT, class _Traits = char_traits<_ECharT>,
+            class _Allocator = allocator<_ECharT> >
+  basic_string<_ECharT, _Traits, _Allocator>
+  generic_string(const _Allocator& __a = _Allocator()) const {
+    return string<_ECharT, _Traits, _Allocator>(__a);
+  }
+
+  std::string generic_string() const { return __pn_; }
+  std::wstring generic_wstring() const { return string<wchar_t>(); }
+  std::string generic_u8string() const { return __pn_; }
+  std::u16string generic_u16string() const { return string<char16_t>(); }
+  std::u32string generic_u32string() const { return string<char32_t>(); }
+
+private:
+  int __compare(__string_view) const;
+  __string_view __root_name() const;
+  __string_view __root_directory() const;
+  __string_view __root_path_raw() const;
+  __string_view __relative_path() const;
+  __string_view __parent_path() const;
+  __string_view __filename() const;
+  __string_view __stem() const;
+  __string_view __extension() const;
+
+public:
+  // compare
+  _LIBCUDACXX_INLINE_VISIBILITY int compare(const path& __p) const noexcept {
+    return __compare(__p.__pn_);
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY int compare(const string_type& __s) const {
+    return __compare(__s);
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY int compare(__string_view __s) const {
+    return __compare(__s);
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY int compare(const value_type* __s) const {
+    return __compare(__s);
+  }
+
+  // decomposition
+  _LIBCUDACXX_INLINE_VISIBILITY path root_name() const {
+    return string_type(__root_name());
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY path root_directory() const {
+    return string_type(__root_directory());
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY path root_path() const {
+    return root_name().append(string_type(__root_directory()));
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY path relative_path() const {
+    return string_type(__relative_path());
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY path parent_path() const {
+    return string_type(__parent_path());
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY path filename() const {
+    return string_type(__filename());
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY path stem() const { return string_type(__stem()); }
+  _LIBCUDACXX_INLINE_VISIBILITY path extension() const {
+    return string_type(__extension());
+  }
+
+  // query
+  _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY bool
+  empty() const noexcept {
+    return __pn_.empty();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_root_name() const {
+    return !__root_name().empty();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_root_directory() const {
+    return !__root_directory().empty();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_root_path() const {
+    return !__root_path_raw().empty();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_relative_path() const {
+    return !__relative_path().empty();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_parent_path() const {
+    return !__parent_path().empty();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_filename() const {
+    return !__filename().empty();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_stem() const { return !__stem().empty(); }
+  _LIBCUDACXX_INLINE_VISIBILITY bool has_extension() const {
+    return !__extension().empty();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY bool is_absolute() const {
+    return has_root_directory();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY bool is_relative() const { return !is_absolute(); }
+
+  // relative paths
+  path lexically_normal() const;
+  path lexically_relative(const path& __base) const;
+
+  _LIBCUDACXX_INLINE_VISIBILITY path lexically_proximate(const path& __base) const {
+    path __result = this->lexically_relative(__base);
+    if (__result.native().empty())
+      return *this;
+    return __result;
+  }
+
+  // iterators
+  class _LIBCUDACXX_TYPE_VIS iterator;
+  typedef iterator const_iterator;
+
+  iterator begin() const;
+  iterator end() const;
+
+  template <class _CharT, class _Traits>
+  _LIBCUDACXX_INLINE_VISIBILITY friend
+      typename enable_if<is_same<_CharT, char>::value &&
+                             is_same<_Traits, char_traits<char> >::value,
+                         basic_ostream<_CharT, _Traits>&>::type
+      operator<<(basic_ostream<_CharT, _Traits>& __os, const path& __p) {
+    __os << std::__quoted(__p.native());
+    return __os;
+  }
+
+  template <class _CharT, class _Traits>
+  _LIBCUDACXX_INLINE_VISIBILITY friend
+      typename enable_if<!is_same<_CharT, char>::value ||
+                             !is_same<_Traits, char_traits<char> >::value,
+                         basic_ostream<_CharT, _Traits>&>::type
+      operator<<(basic_ostream<_CharT, _Traits>& __os, const path& __p) {
+    __os << std::__quoted(__p.string<_CharT, _Traits>());
+    return __os;
+  }
+
+  template <class _CharT, class _Traits>
+  _LIBCUDACXX_INLINE_VISIBILITY friend basic_istream<_CharT, _Traits>&
+  operator>>(basic_istream<_CharT, _Traits>& __is, path& __p) {
+    basic_string<_CharT, _Traits> __tmp;
+    __is >> __quoted(__tmp);
+    __p = __tmp;
+    return __is;
+  }
+
+  friend _LIBCUDACXX_INLINE_VISIBILITY bool operator==(const path& __lhs, const path& __rhs) noexcept {
+    return __lhs.compare(__rhs) == 0;
+  }
+  friend _LIBCUDACXX_INLINE_VISIBILITY bool operator!=(const path& __lhs, const path& __rhs) noexcept {
+    return __lhs.compare(__rhs) != 0;
+  }
+  friend _LIBCUDACXX_INLINE_VISIBILITY bool operator<(const path& __lhs, const path& __rhs) noexcept {
+    return __lhs.compare(__rhs) < 0;
+  }
+  friend _LIBCUDACXX_INLINE_VISIBILITY bool operator<=(const path& __lhs, const path& __rhs) noexcept {
+    return __lhs.compare(__rhs) <= 0;
+  }
+  friend _LIBCUDACXX_INLINE_VISIBILITY bool operator>(const path& __lhs, const path& __rhs) noexcept {
+    return __lhs.compare(__rhs) > 0;
+  }
+  friend _LIBCUDACXX_INLINE_VISIBILITY bool operator>=(const path& __lhs, const path& __rhs) noexcept {
+    return __lhs.compare(__rhs) >= 0;
+  }
+
+  friend _LIBCUDACXX_INLINE_VISIBILITY path operator/(const path& __lhs,
+                                                  const path& __rhs) {
+    path __result(__lhs);
+    __result /= __rhs;
+    return __result;
+  }
+private:
+  inline _LIBCUDACXX_INLINE_VISIBILITY path&
+  __assign_view(__string_view const& __s) noexcept {
+    __pn_ = string_type(__s);
+    return *this;
+  }
+  string_type __pn_;
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void swap(path& __lhs, path& __rhs) noexcept {
+  __lhs.swap(__rhs);
+}
+
+_LIBCUDACXX_FUNC_VIS
+size_t hash_value(const path& __p) noexcept;
+
+template <class _Source>
+_LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_pathable<_Source>::value, path>::type
+    u8path(const _Source& __s) {
+  static_assert(
+      is_same<typename __is_pathable<_Source>::__char_type, char>::value,
+      "u8path(Source const&) requires Source have a character type of type "
+      "'char'");
+  return path(__s);
+}
+
+template <class _InputIt>
+_LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_pathable<_InputIt>::value, path>::type
+    u8path(_InputIt __f, _InputIt __l) {
+  static_assert(
+      is_same<typename __is_pathable<_InputIt>::__char_type, char>::value,
+      "u8path(Iter, Iter) requires Iter have a value_type of type 'char'");
+  return path(__f, __l);
+}
+
+class _LIBCUDACXX_TYPE_VIS path::iterator {
+public:
+  enum _ParserState : unsigned char {
+    _Singular,
+    _BeforeBegin,
+    _InRootName,
+    _InRootDir,
+    _InFilenames,
+    _InTrailingSep,
+    _AtEnd
+  };
+
+public:
+  typedef bidirectional_iterator_tag iterator_category;
+
+  typedef path value_type;
+  typedef std::ptrdiff_t difference_type;
+  typedef const path* pointer;
+  typedef const path& reference;
+
+  typedef void
+      __stashing_iterator_tag; // See reverse_iterator and __is_stashing_iterator
+
+public:
+  _LIBCUDACXX_INLINE_VISIBILITY
+  iterator()
+      : __stashed_elem_(), __path_ptr_(nullptr), __entry_(),
+        __state_(_Singular) {}
+
+  iterator(const iterator&) = default;
+  ~iterator() = default;
+
+  iterator& operator=(const iterator&) = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  reference operator*() const { return __stashed_elem_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  pointer operator->() const { return &__stashed_elem_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  iterator& operator++() {
+    _LIBCUDACXX_ASSERT(__state_ != _Singular,
+                   "attempting to increment a singular iterator");
+    _LIBCUDACXX_ASSERT(__state_ != _AtEnd,
+                   "attempting to increment the end iterator");
+    return __increment();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  iterator operator++(int) {
+    iterator __it(*this);
+    this->operator++();
+    return __it;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  iterator& operator--() {
+    _LIBCUDACXX_ASSERT(__state_ != _Singular,
+                   "attempting to decrement a singular iterator");
+    _LIBCUDACXX_ASSERT(__entry_.data() != __path_ptr_->native().data(),
+                   "attempting to decrement the begin iterator");
+    return __decrement();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  iterator operator--(int) {
+    iterator __it(*this);
+    this->operator--();
+    return __it;
+  }
+
+private:
+  friend class path;
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY friend bool operator==(const iterator&,
+                                                          const iterator&);
+
+  iterator& __increment();
+  iterator& __decrement();
+
+  path __stashed_elem_;
+  const path* __path_ptr_;
+  path::__string_view __entry_;
+  _ParserState __state_;
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool operator==(const path::iterator& __lhs,
+                                                 const path::iterator& __rhs) {
+  return __lhs.__path_ptr_ == __rhs.__path_ptr_ &&
+         __lhs.__entry_.data() == __rhs.__entry_.data();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool operator!=(const path::iterator& __lhs,
+                                                 const path::iterator& __rhs) {
+  return !(__lhs == __rhs);
+}
+
+// TODO(ldionne): We need to pop the pragma and push it again after
+//                filesystem_error to work around PR41078.
+_LIBCUDACXX_AVAILABILITY_FILESYSTEM_POP
+
+class _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_EXCEPTION_ABI filesystem_error : public system_error {
+public:
+  _LIBCUDACXX_INLINE_VISIBILITY
+  filesystem_error(const string& __what, error_code __ec)
+      : system_error(__ec, __what),
+        __storage_(make_shared<_Storage>(path(), path())) {
+    __create_what(0);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  filesystem_error(const string& __what, const path& __p1, error_code __ec)
+      : system_error(__ec, __what),
+        __storage_(make_shared<_Storage>(__p1, path())) {
+    __create_what(1);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  filesystem_error(const string& __what, const path& __p1, const path& __p2,
+                   error_code __ec)
+      : system_error(__ec, __what),
+        __storage_(make_shared<_Storage>(__p1, __p2)) {
+    __create_what(2);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const path& path1() const noexcept { return __storage_->__p1_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const path& path2() const noexcept { return __storage_->__p2_; }
+
+  ~filesystem_error() override; // key function
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const char* what() const noexcept override {
+    return __storage_->__what_.c_str();
+  }
+
+  void __create_what(int __num_paths);
+
+private:
+  struct _LIBCUDACXX_HIDDEN _Storage {
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Storage(const path& __p1, const path& __p2) : __p1_(__p1), __p2_(__p2) {}
+
+    path __p1_;
+    path __p2_;
+    string __what_;
+  };
+  shared_ptr<_Storage> __storage_;
+};
+
+_LIBCUDACXX_AVAILABILITY_FILESYSTEM_PUSH
+
+template <class... _Args>
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+void __throw_filesystem_error(_Args&&... __args) {
+  throw filesystem_error(std::forward<_Args>(__args)...);
+}
+#else
+void __throw_filesystem_error(_Args&&...) {
+  _CUDA_VSTD::abort();
+}
+#endif
+
+// operational functions
+
+_LIBCUDACXX_FUNC_VIS
+path __absolute(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+path __canonical(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __copy(const path& __from, const path& __to, copy_options __opt,
+            error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __copy_file(const path& __from, const path& __to, copy_options __opt,
+                 error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __copy_symlink(const path& __existing_symlink, const path& __new_symlink,
+                    error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __create_directories(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __create_directory(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __create_directory(const path& p, const path& attributes,
+                        error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __create_directory_symlink(const path& __to, const path& __new_symlink,
+                                error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __create_hard_link(const path& __to, const path& __new_hard_link,
+                        error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __create_symlink(const path& __to, const path& __new_symlink,
+                      error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+path __current_path(error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __current_path(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __equivalent(const path&, const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+uintmax_t __file_size(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+uintmax_t __hard_link_count(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __fs_is_empty(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+file_time_type __last_write_time(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __last_write_time(const path& p, file_time_type new_time,
+                       error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __permissions(const path&, perms, perm_options, error_code* = nullptr);
+_LIBCUDACXX_FUNC_VIS
+path __read_symlink(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+bool __remove(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+uintmax_t __remove_all(const path& p, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __rename(const path& from, const path& to, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+void __resize_file(const path& p, uintmax_t size, error_code* ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+space_info __space(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+file_status __status(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+file_status __symlink_status(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+path __system_complete(const path&, error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+path __temp_directory_path(error_code* __ec = nullptr);
+_LIBCUDACXX_FUNC_VIS
+path __weakly_canonical(path const& __p, error_code* __ec = nullptr);
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path current_path() {
+  return __current_path();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path current_path(error_code& __ec) {
+  return __current_path(&__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void current_path(const path& __p) {
+  __current_path(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void current_path(const path& __p,
+                                                   error_code& __ec) noexcept {
+  __current_path(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path absolute(const path& __p) {
+  return __absolute(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path absolute(const path& __p,
+                                               error_code& __ec) {
+  return __absolute(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path canonical(const path& __p) {
+  return __canonical(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path canonical(const path& __p,
+                                                error_code& __ec) {
+  return __canonical(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void copy(const path& __from,
+                                           const path& __to) {
+  __copy(__from, __to, copy_options::none);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void copy(const path& __from, const path& __to,
+                                           error_code& __ec) {
+  __copy(__from, __to, copy_options::none, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void copy(const path& __from, const path& __to,
+                                           copy_options __opt) {
+  __copy(__from, __to, __opt);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void copy(const path& __from, const path& __to,
+                                           copy_options __opt,
+                                           error_code& __ec) {
+  __copy(__from, __to, __opt, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool copy_file(const path& __from,
+                                                const path& __to) {
+  return __copy_file(__from, __to, copy_options::none);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+copy_file(const path& __from, const path& __to, error_code& __ec) {
+  return __copy_file(__from, __to, copy_options::none, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+copy_file(const path& __from, const path& __to, copy_options __opt) {
+  return __copy_file(__from, __to, __opt);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool copy_file(const path& __from,
+                                                const path& __to,
+                                                copy_options __opt,
+                                                error_code& __ec) {
+  return __copy_file(__from, __to, __opt, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void copy_symlink(const path& __existing,
+                                                   const path& __new) {
+  __copy_symlink(__existing, __new);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+copy_symlink(const path& __ext, const path& __new, error_code& __ec) noexcept {
+  __copy_symlink(__ext, __new, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool create_directories(const path& __p) {
+  return __create_directories(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool create_directories(const path& __p,
+                                                         error_code& __ec) {
+  return __create_directories(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool create_directory(const path& __p) {
+  return __create_directory(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+create_directory(const path& __p, error_code& __ec) noexcept {
+  return __create_directory(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool create_directory(const path& __p,
+                                                       const path& __attrs) {
+  return __create_directory(__p, __attrs);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+create_directory(const path& __p, const path& __attrs,
+                 error_code& __ec) noexcept {
+  return __create_directory(__p, __attrs, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+create_directory_symlink(const path& __to, const path& __new) {
+  __create_directory_symlink(__to, __new);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+create_directory_symlink(const path& __to, const path& __new,
+                         error_code& __ec) noexcept {
+  __create_directory_symlink(__to, __new, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void create_hard_link(const path& __to,
+                                                       const path& __new) {
+  __create_hard_link(__to, __new);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+create_hard_link(const path& __to, const path& __new,
+                 error_code& __ec) noexcept {
+  __create_hard_link(__to, __new, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void create_symlink(const path& __to,
+                                                     const path& __new) {
+  __create_symlink(__to, __new);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+create_symlink(const path& __to, const path& __new, error_code& __ec) noexcept {
+  return __create_symlink(__to, __new, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool status_known(file_status __s) noexcept {
+  return __s.type() != file_type::none;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool exists(file_status __s) noexcept {
+  return status_known(__s) && __s.type() != file_type::not_found;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool exists(const path& __p) {
+  return exists(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool exists(const path& __p,
+                                             error_code& __ec) noexcept {
+  auto __s = __status(__p, &__ec);
+  if (status_known(__s))
+    __ec.clear();
+  return exists(__s);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool equivalent(const path& __p1,
+                                                 const path& __p2) {
+  return __equivalent(__p1, __p2);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+equivalent(const path& __p1, const path& __p2, error_code& __ec) noexcept {
+  return __equivalent(__p1, __p2, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY uintmax_t file_size(const path& __p) {
+  return __file_size(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY uintmax_t
+file_size(const path& __p, error_code& __ec) noexcept {
+  return __file_size(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY uintmax_t hard_link_count(const path& __p) {
+  return __hard_link_count(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY uintmax_t
+hard_link_count(const path& __p, error_code& __ec) noexcept {
+  return __hard_link_count(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_block_file(file_status __s) noexcept {
+  return __s.type() == file_type::block;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_block_file(const path& __p) {
+  return is_block_file(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_block_file(const path& __p,
+                                                    error_code& __ec) noexcept {
+  return is_block_file(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+is_character_file(file_status __s) noexcept {
+  return __s.type() == file_type::character;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_character_file(const path& __p) {
+  return is_character_file(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+is_character_file(const path& __p, error_code& __ec) noexcept {
+  return is_character_file(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_directory(file_status __s) noexcept {
+  return __s.type() == file_type::directory;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_directory(const path& __p) {
+  return is_directory(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_directory(const path& __p,
+                                                   error_code& __ec) noexcept {
+  return is_directory(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_empty(const path& __p) {
+  return __fs_is_empty(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_empty(const path& __p,
+                                               error_code& __ec) {
+  return __fs_is_empty(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_fifo(file_status __s) noexcept {
+  return __s.type() == file_type::fifo;
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_fifo(const path& __p) {
+  return is_fifo(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_fifo(const path& __p,
+                                              error_code& __ec) noexcept {
+  return is_fifo(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+is_regular_file(file_status __s) noexcept {
+  return __s.type() == file_type::regular;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_regular_file(const path& __p) {
+  return is_regular_file(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+is_regular_file(const path& __p, error_code& __ec) noexcept {
+  return is_regular_file(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_socket(file_status __s) noexcept {
+  return __s.type() == file_type::socket;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_socket(const path& __p) {
+  return is_socket(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_socket(const path& __p,
+                                                error_code& __ec) noexcept {
+  return is_socket(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_symlink(file_status __s) noexcept {
+  return __s.type() == file_type::symlink;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_symlink(const path& __p) {
+  return is_symlink(__symlink_status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_symlink(const path& __p,
+                                                 error_code& __ec) noexcept {
+  return is_symlink(__symlink_status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_other(file_status __s) noexcept {
+  return exists(__s) && !is_regular_file(__s) && !is_directory(__s) &&
+         !is_symlink(__s);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_other(const path& __p) {
+  return is_other(__status(__p));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool is_other(const path& __p,
+                                               error_code& __ec) noexcept {
+  return is_other(__status(__p, &__ec));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY file_time_type
+last_write_time(const path& __p) {
+  return __last_write_time(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY file_time_type
+last_write_time(const path& __p, error_code& __ec) noexcept {
+  return __last_write_time(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void last_write_time(const path& __p,
+                                                      file_time_type __t) {
+  __last_write_time(__p, __t);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+last_write_time(const path& __p, file_time_type __t,
+                error_code& __ec) noexcept {
+  __last_write_time(__p, __t, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+permissions(const path& __p, perms __prms,
+            perm_options __opts = perm_options::replace) {
+  __permissions(__p, __prms, __opts);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void permissions(const path& __p, perms __prms,
+                                                  error_code& __ec) noexcept {
+  __permissions(__p, __prms, perm_options::replace, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void permissions(const path& __p, perms __prms,
+                                                  perm_options __opts,
+                                                  error_code& __ec) {
+  __permissions(__p, __prms, __opts, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path proximate(const path& __p,
+                                                const path& __base,
+                                                error_code& __ec) {
+  path __tmp = __weakly_canonical(__p, &__ec);
+  if (__ec)
+    return {};
+  path __tmp_base = __weakly_canonical(__base, &__ec);
+  if (__ec)
+    return {};
+  return __tmp.lexically_proximate(__tmp_base);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path proximate(const path& __p,
+                                                error_code& __ec) {
+  return proximate(__p, current_path(), __ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path
+proximate(const path& __p, const path& __base = current_path()) {
+  return __weakly_canonical(__p).lexically_proximate(
+      __weakly_canonical(__base));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path read_symlink(const path& __p) {
+  return __read_symlink(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path read_symlink(const path& __p,
+                                                   error_code& __ec) {
+  return __read_symlink(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path relative(const path& __p,
+                                               const path& __base,
+                                               error_code& __ec) {
+  path __tmp = __weakly_canonical(__p, &__ec);
+  if (__ec)
+    return path();
+  path __tmpbase = __weakly_canonical(__base, &__ec);
+  if (__ec)
+    return path();
+  return __tmp.lexically_relative(__tmpbase);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path relative(const path& __p,
+                                               error_code& __ec) {
+  return relative(__p, current_path(), __ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path
+relative(const path& __p, const path& __base = current_path()) {
+  return __weakly_canonical(__p).lexically_relative(__weakly_canonical(__base));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool remove(const path& __p) {
+  return __remove(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool remove(const path& __p,
+                                             error_code& __ec) noexcept {
+  return __remove(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY uintmax_t remove_all(const path& __p) {
+  return __remove_all(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY uintmax_t remove_all(const path& __p,
+                                                      error_code& __ec) {
+  return __remove_all(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void rename(const path& __from,
+                                             const path& __to) {
+  return __rename(__from, __to);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+rename(const path& __from, const path& __to, error_code& __ec) noexcept {
+  return __rename(__from, __to, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void resize_file(const path& __p,
+                                                  uintmax_t __ns) {
+  return __resize_file(__p, __ns);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void
+resize_file(const path& __p, uintmax_t __ns, error_code& __ec) noexcept {
+  return __resize_file(__p, __ns, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY space_info space(const path& __p) {
+  return __space(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY space_info space(const path& __p,
+                                                  error_code& __ec) noexcept {
+  return __space(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY file_status status(const path& __p) {
+  return __status(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY file_status status(const path& __p,
+                                                    error_code& __ec) noexcept {
+  return __status(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY file_status symlink_status(const path& __p) {
+  return __symlink_status(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY file_status
+symlink_status(const path& __p, error_code& __ec) noexcept {
+  return __symlink_status(__p, &__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path temp_directory_path() {
+  return __temp_directory_path();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path temp_directory_path(error_code& __ec) {
+  return __temp_directory_path(&__ec);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path weakly_canonical(path const& __p) {
+  return __weakly_canonical(__p);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY path weakly_canonical(path const& __p,
+                                                       error_code& __ec) {
+  return __weakly_canonical(__p, &__ec);
+}
+
+class directory_iterator;
+class recursive_directory_iterator;
+class _LIBCUDACXX_HIDDEN __dir_stream;
+
+class directory_entry {
+  typedef _CUDA_VSTD_FS::path _Path;
+
+public:
+  // constructors and destructors
+  directory_entry() noexcept = default;
+  directory_entry(directory_entry const&) = default;
+  directory_entry(directory_entry&&) noexcept = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit directory_entry(_Path const& __p) : __p_(__p) {
+    error_code __ec;
+    __refresh(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  directory_entry(_Path const& __p, error_code& __ec) : __p_(__p) {
+    __refresh(&__ec);
+  }
+
+  ~directory_entry() {}
+
+  directory_entry& operator=(directory_entry const&) = default;
+  directory_entry& operator=(directory_entry&&) noexcept = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void assign(_Path const& __p) {
+    __p_ = __p;
+    error_code __ec;
+    __refresh(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void assign(_Path const& __p, error_code& __ec) {
+    __p_ = __p;
+    __refresh(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void replace_filename(_Path const& __p) {
+    __p_.replace_filename(__p);
+    error_code __ec;
+    __refresh(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void replace_filename(_Path const& __p, error_code& __ec) {
+    __p_ = __p_.parent_path() / __p;
+    __refresh(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void refresh() { __refresh(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void refresh(error_code& __ec) noexcept { __refresh(&__ec); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _Path const& path() const noexcept { return __p_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  operator const _Path&() const noexcept { return __p_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool exists() const { return _CUDA_VSTD_FS::exists(file_status{__get_ft()}); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool exists(error_code& __ec) const noexcept {
+    return _CUDA_VSTD_FS::exists(file_status{__get_ft(&__ec)});
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_block_file() const { return __get_ft() == file_type::block; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_block_file(error_code& __ec) const noexcept {
+    return __get_ft(&__ec) == file_type::block;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_character_file() const { return __get_ft() == file_type::character; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_character_file(error_code& __ec) const noexcept {
+    return __get_ft(&__ec) == file_type::character;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_directory() const { return __get_ft() == file_type::directory; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_directory(error_code& __ec) const noexcept {
+    return __get_ft(&__ec) == file_type::directory;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_fifo() const { return __get_ft() == file_type::fifo; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_fifo(error_code& __ec) const noexcept {
+    return __get_ft(&__ec) == file_type::fifo;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_other() const { return _CUDA_VSTD_FS::is_other(file_status{__get_ft()}); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_other(error_code& __ec) const noexcept {
+    return _CUDA_VSTD_FS::is_other(file_status{__get_ft(&__ec)});
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_regular_file() const { return __get_ft() == file_type::regular; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_regular_file(error_code& __ec) const noexcept {
+    return __get_ft(&__ec) == file_type::regular;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_socket() const { return __get_ft() == file_type::socket; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_socket(error_code& __ec) const noexcept {
+    return __get_ft(&__ec) == file_type::socket;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_symlink() const { return __get_sym_ft() == file_type::symlink; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool is_symlink(error_code& __ec) const noexcept {
+    return __get_sym_ft(&__ec) == file_type::symlink;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  uintmax_t file_size() const { return __get_size(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  uintmax_t file_size(error_code& __ec) const noexcept {
+    return __get_size(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  uintmax_t hard_link_count() const { return __get_nlink(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  uintmax_t hard_link_count(error_code& __ec) const noexcept {
+    return __get_nlink(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_time_type last_write_time() const { return __get_write_time(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_time_type last_write_time(error_code& __ec) const noexcept {
+    return __get_write_time(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status status() const { return __get_status(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status status(error_code& __ec) const noexcept {
+    return __get_status(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status symlink_status() const { return __get_symlink_status(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status symlink_status(error_code& __ec) const noexcept {
+    return __get_symlink_status(&__ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool operator<(directory_entry const& __rhs) const noexcept {
+    return __p_ < __rhs.__p_;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool operator==(directory_entry const& __rhs) const noexcept {
+    return __p_ == __rhs.__p_;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool operator!=(directory_entry const& __rhs) const noexcept {
+    return __p_ != __rhs.__p_;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool operator<=(directory_entry const& __rhs) const noexcept {
+    return __p_ <= __rhs.__p_;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool operator>(directory_entry const& __rhs) const noexcept {
+    return __p_ > __rhs.__p_;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool operator>=(directory_entry const& __rhs) const noexcept {
+    return __p_ >= __rhs.__p_;
+  }
+
+private:
+  friend class directory_iterator;
+  friend class recursive_directory_iterator;
+  friend class __dir_stream;
+
+  enum _CacheType : unsigned char {
+    _Empty,
+    _IterSymlink,
+    _IterNonSymlink,
+    _RefreshSymlink,
+    _RefreshSymlinkUnresolved,
+    _RefreshNonSymlink
+  };
+
+  struct __cached_data {
+    uintmax_t __size_;
+    uintmax_t __nlink_;
+    file_time_type __write_time_;
+    perms __sym_perms_;
+    perms __non_sym_perms_;
+    file_type __type_;
+    _CacheType __cache_type_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __cached_data() noexcept { __reset(); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __reset() {
+      __cache_type_ = _Empty;
+      __type_ = file_type::none;
+      __sym_perms_ = __non_sym_perms_ = perms::unknown;
+      __size_ = __nlink_ = uintmax_t(-1);
+      __write_time_ = file_time_type::min();
+    }
+  };
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __cached_data __create_iter_result(file_type __ft) {
+    __cached_data __data;
+    __data.__type_ = __ft;
+    __data.__cache_type_ = [&]() {
+      switch (__ft) {
+      case file_type::none:
+        return _Empty;
+      case file_type::symlink:
+        return _IterSymlink;
+      default:
+        return _IterNonSymlink;
+      }
+    }();
+    return __data;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void __assign_iter_entry(_Path&& __p, __cached_data __dt) {
+    __p_ = std::move(__p);
+    __data_ = __dt;
+  }
+
+  _LIBCUDACXX_FUNC_VIS
+  error_code __do_refresh() noexcept;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static bool __is_dne_error(error_code const& __ec) {
+    if (!__ec)
+      return true;
+    switch (static_cast<errc>(__ec.value())) {
+    case errc::no_such_file_or_directory:
+    case errc::not_a_directory:
+      return true;
+    default:
+      return false;
+    }
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void __handle_error(const char* __msg, error_code* __dest_ec,
+                      error_code const& __ec, bool __allow_dne = false) const {
+    if (__dest_ec) {
+      *__dest_ec = __ec;
+      return;
+    }
+    if (__ec && (!__allow_dne || !__is_dne_error(__ec)))
+      __throw_filesystem_error(__msg, __p_, __ec);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void __refresh(error_code* __ec = nullptr) {
+    __handle_error("in directory_entry::refresh", __ec, __do_refresh(),
+                   /*allow_dne*/ true);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_type __get_sym_ft(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+      return __symlink_status(__p_, __ec).type();
+    case _IterSymlink:
+    case _RefreshSymlink:
+    case _RefreshSymlinkUnresolved:
+      if (__ec)
+        __ec->clear();
+      return file_type::symlink;
+    case _IterNonSymlink:
+    case _RefreshNonSymlink:
+      file_status __st(__data_.__type_);
+      if (__ec && !_CUDA_VSTD_FS::exists(__st))
+        *__ec = make_error_code(errc::no_such_file_or_directory);
+      else if (__ec)
+        __ec->clear();
+      return __data_.__type_;
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_type __get_ft(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+    case _IterSymlink:
+    case _RefreshSymlinkUnresolved:
+      return __status(__p_, __ec).type();
+    case _IterNonSymlink:
+    case _RefreshNonSymlink:
+    case _RefreshSymlink: {
+      file_status __st(__data_.__type_);
+      if (__ec && !_CUDA_VSTD_FS::exists(__st))
+        *__ec = make_error_code(errc::no_such_file_or_directory);
+      else if (__ec)
+        __ec->clear();
+      return __data_.__type_;
+    }
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status __get_status(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+    case _IterNonSymlink:
+    case _IterSymlink:
+    case _RefreshSymlinkUnresolved:
+      return __status(__p_, __ec);
+    case _RefreshNonSymlink:
+    case _RefreshSymlink:
+      return file_status(__get_ft(__ec), __data_.__non_sym_perms_);
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_status __get_symlink_status(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+    case _IterNonSymlink:
+    case _IterSymlink:
+      return __symlink_status(__p_, __ec);
+    case _RefreshNonSymlink:
+      return file_status(__get_sym_ft(__ec), __data_.__non_sym_perms_);
+    case _RefreshSymlink:
+    case _RefreshSymlinkUnresolved:
+      return file_status(__get_sym_ft(__ec), __data_.__sym_perms_);
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  uintmax_t __get_size(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+    case _IterNonSymlink:
+    case _IterSymlink:
+    case _RefreshSymlinkUnresolved:
+      return _CUDA_VSTD_FS::__file_size(__p_, __ec);
+    case _RefreshSymlink:
+    case _RefreshNonSymlink: {
+      error_code __m_ec;
+      file_status __st(__get_ft(&__m_ec));
+      __handle_error("in directory_entry::file_size", __ec, __m_ec);
+      if (_CUDA_VSTD_FS::exists(__st) && !_CUDA_VSTD_FS::is_regular_file(__st)) {
+        errc __err_kind = _CUDA_VSTD_FS::is_directory(__st) ? errc::is_a_directory
+                                                       : errc::not_supported;
+        __handle_error("in directory_entry::file_size", __ec,
+                       make_error_code(__err_kind));
+      }
+      return __data_.__size_;
+    }
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  uintmax_t __get_nlink(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+    case _IterNonSymlink:
+    case _IterSymlink:
+    case _RefreshSymlinkUnresolved:
+      return _CUDA_VSTD_FS::__hard_link_count(__p_, __ec);
+    case _RefreshSymlink:
+    case _RefreshNonSymlink: {
+      error_code __m_ec;
+      (void)__get_ft(&__m_ec);
+      __handle_error("in directory_entry::hard_link_count", __ec, __m_ec);
+      return __data_.__nlink_;
+    }
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  file_time_type __get_write_time(error_code* __ec = nullptr) const {
+    switch (__data_.__cache_type_) {
+    case _Empty:
+    case _IterNonSymlink:
+    case _IterSymlink:
+    case _RefreshSymlinkUnresolved:
+      return _CUDA_VSTD_FS::__last_write_time(__p_, __ec);
+    case _RefreshSymlink:
+    case _RefreshNonSymlink: {
+      error_code __m_ec;
+      file_status __st(__get_ft(&__m_ec));
+      __handle_error("in directory_entry::last_write_time", __ec, __m_ec);
+      if (_CUDA_VSTD_FS::exists(__st) &&
+          __data_.__write_time_ == file_time_type::min())
+        __handle_error("in directory_entry::last_write_time", __ec,
+                       make_error_code(errc::value_too_large));
+      return __data_.__write_time_;
+    }
+    }
+    _LIBCUDACXX_UNREACHABLE();
+  }
+
+private:
+  _Path __p_;
+  __cached_data __data_;
+};
+
+class __dir_element_proxy {
+public:
+  inline _LIBCUDACXX_INLINE_VISIBILITY directory_entry operator*() {
+    return _CUDA_VSTD::move(__elem_);
+  }
+
+private:
+  friend class directory_iterator;
+  friend class recursive_directory_iterator;
+  explicit __dir_element_proxy(directory_entry const& __e) : __elem_(__e) {}
+  __dir_element_proxy(__dir_element_proxy&& __o)
+      : __elem_(_CUDA_VSTD::move(__o.__elem_)) {}
+  directory_entry __elem_;
+};
+
+class directory_iterator {
+public:
+  typedef directory_entry value_type;
+  typedef ptrdiff_t difference_type;
+  typedef value_type const* pointer;
+  typedef value_type const& reference;
+  typedef input_iterator_tag iterator_category;
+
+public:
+  //ctor & dtor
+  directory_iterator() noexcept {}
+
+  explicit directory_iterator(const path& __p)
+      : directory_iterator(__p, nullptr) {}
+
+  directory_iterator(const path& __p, directory_options __opts)
+      : directory_iterator(__p, nullptr, __opts) {}
+
+  directory_iterator(const path& __p, error_code& __ec)
+      : directory_iterator(__p, &__ec) {}
+
+  directory_iterator(const path& __p, directory_options __opts,
+                     error_code& __ec)
+      : directory_iterator(__p, &__ec, __opts) {}
+
+  directory_iterator(const directory_iterator&) = default;
+  directory_iterator(directory_iterator&&) = default;
+  directory_iterator& operator=(const directory_iterator&) = default;
+
+  directory_iterator& operator=(directory_iterator&& __o) noexcept {
+    // non-default implementation provided to support self-move assign.
+    if (this != &__o) {
+      __imp_ = _CUDA_VSTD::move(__o.__imp_);
+    }
+    return *this;
+  }
+
+  ~directory_iterator() = default;
+
+  const directory_entry& operator*() const {
+    _LIBCUDACXX_ASSERT(__imp_, "The end iterator cannot be dereferenced");
+    return __dereference();
+  }
+
+  const directory_entry* operator->() const { return &**this; }
+
+  directory_iterator& operator++() { return __increment(); }
+
+  __dir_element_proxy operator++(int) {
+    __dir_element_proxy __p(**this);
+    __increment();
+    return __p;
+  }
+
+  directory_iterator& increment(error_code& __ec) { return __increment(&__ec); }
+
+private:
+  inline _LIBCUDACXX_INLINE_VISIBILITY friend bool
+  operator==(const directory_iterator& __lhs,
+             const directory_iterator& __rhs) noexcept;
+
+  // construct the dir_stream
+  _LIBCUDACXX_FUNC_VIS
+  directory_iterator(const path&, error_code*,
+                     directory_options = directory_options::none);
+
+  _LIBCUDACXX_FUNC_VIS
+  directory_iterator& __increment(error_code* __ec = nullptr);
+
+  _LIBCUDACXX_FUNC_VIS
+  const directory_entry& __dereference() const;
+
+private:
+  shared_ptr<__dir_stream> __imp_;
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+operator==(const directory_iterator& __lhs,
+           const directory_iterator& __rhs) noexcept {
+  return __lhs.__imp_ == __rhs.__imp_;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+operator!=(const directory_iterator& __lhs,
+           const directory_iterator& __rhs) noexcept {
+  return !(__lhs == __rhs);
+}
+
+// enable directory_iterator range-based for statements
+inline _LIBCUDACXX_INLINE_VISIBILITY directory_iterator
+begin(directory_iterator __iter) noexcept {
+  return __iter;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY directory_iterator
+end(const directory_iterator&) noexcept {
+  return directory_iterator();
+}
+
+class recursive_directory_iterator {
+public:
+  using value_type = directory_entry;
+  using difference_type = std::ptrdiff_t;
+  using pointer = directory_entry const*;
+  using reference = directory_entry const&;
+  using iterator_category = std::input_iterator_tag;
+
+public:
+  // constructors and destructor
+  _LIBCUDACXX_INLINE_VISIBILITY
+  recursive_directory_iterator() noexcept : __rec_(false) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit recursive_directory_iterator(
+      const path& __p, directory_options __xoptions = directory_options::none)
+      : recursive_directory_iterator(__p, __xoptions, nullptr) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  recursive_directory_iterator(const path& __p, directory_options __xoptions,
+                               error_code& __ec)
+      : recursive_directory_iterator(__p, __xoptions, &__ec) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  recursive_directory_iterator(const path& __p, error_code& __ec)
+      : recursive_directory_iterator(__p, directory_options::none, &__ec) {}
+
+  recursive_directory_iterator(const recursive_directory_iterator&) = default;
+  recursive_directory_iterator(recursive_directory_iterator&&) = default;
+
+  recursive_directory_iterator&
+  operator=(const recursive_directory_iterator&) = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  recursive_directory_iterator&
+  operator=(recursive_directory_iterator&& __o) noexcept {
+    // non-default implementation provided to support self-move assign.
+    if (this != &__o) {
+      __imp_ = _CUDA_VSTD::move(__o.__imp_);
+      __rec_ = __o.__rec_;
+    }
+    return *this;
+  }
+
+  ~recursive_directory_iterator() = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const directory_entry& operator*() const { return __dereference(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const directory_entry* operator->() const { return &__dereference(); }
+
+  recursive_directory_iterator& operator++() { return __increment(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __dir_element_proxy operator++(int) {
+    __dir_element_proxy __p(**this);
+    __increment();
+    return __p;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  recursive_directory_iterator& increment(error_code& __ec) {
+    return __increment(&__ec);
+  }
+
+  _LIBCUDACXX_FUNC_VIS directory_options options() const;
+  _LIBCUDACXX_FUNC_VIS int depth() const;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void pop() { __pop(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void pop(error_code& __ec) { __pop(&__ec); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  bool recursion_pending() const { return __rec_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void disable_recursion_pending() { __rec_ = false; }
+
+private:
+  _LIBCUDACXX_FUNC_VIS
+  recursive_directory_iterator(const path& __p, directory_options __opt,
+                               error_code* __ec);
+
+  _LIBCUDACXX_FUNC_VIS
+  const directory_entry& __dereference() const;
+
+  _LIBCUDACXX_FUNC_VIS
+  bool __try_recursion(error_code* __ec);
+
+  _LIBCUDACXX_FUNC_VIS
+  void __advance(error_code* __ec = nullptr);
+
+  _LIBCUDACXX_FUNC_VIS
+  recursive_directory_iterator& __increment(error_code* __ec = nullptr);
+
+  _LIBCUDACXX_FUNC_VIS
+  void __pop(error_code* __ec = nullptr);
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY friend bool
+  operator==(const recursive_directory_iterator&,
+             const recursive_directory_iterator&) noexcept;
+
+  struct _LIBCUDACXX_HIDDEN __shared_imp;
+  shared_ptr<__shared_imp> __imp_;
+  bool __rec_;
+}; // class recursive_directory_iterator
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool
+operator==(const recursive_directory_iterator& __lhs,
+           const recursive_directory_iterator& __rhs) noexcept {
+  return __lhs.__imp_ == __rhs.__imp_;
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline bool operator!=(const recursive_directory_iterator& __lhs,
+                       const recursive_directory_iterator& __rhs) noexcept {
+  return !(__lhs == __rhs);
+}
+// enable recursive_directory_iterator range-based for statements
+inline _LIBCUDACXX_INLINE_VISIBILITY recursive_directory_iterator
+begin(recursive_directory_iterator __iter) noexcept {
+  return __iter;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY recursive_directory_iterator
+end(const recursive_directory_iterator&) noexcept {
+  return recursive_directory_iterator();
+}
+
+_LIBCUDACXX_AVAILABILITY_FILESYSTEM_POP
+
+_LIBCUDACXX_END_NAMESPACE_FILESYSTEM
+
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_POP_MACROS
+
+#endif // _LIBCUDACXX_FILESYSTEM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/float.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/float.h
new file mode 100644
index 000000000000..afd612c76e10
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/float.h
@@ -0,0 +1,97 @@
+// -*- C++ -*-
+//===--------------------------- float.h ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FLOAT_H
+#define _LIBCUDACXX_FLOAT_H
+
+/*
+    float.h synopsis
+
+Macros:
+
+    FLT_ROUNDS
+    FLT_EVAL_METHOD     // C99
+    FLT_RADIX
+
+    FLT_MANT_DIG
+    DBL_MANT_DIG
+    LDBL_MANT_DIG
+
+    FLT_HAS_SUBNORM     // C11
+    DBL_HAS_SUBNORM     // C11
+    LDBL_HAS_SUBNORM    // C11
+
+    DECIMAL_DIG         // C99
+    FLT_DECIMAL_DIG     // C11
+    DBL_DECIMAL_DIG     // C11
+    LDBL_DECIMAL_DIG    // C11
+
+    FLT_DIG
+    DBL_DIG
+    LDBL_DIG
+
+    FLT_MIN_EXP
+    DBL_MIN_EXP
+    LDBL_MIN_EXP
+
+    FLT_MIN_10_EXP
+    DBL_MIN_10_EXP
+    LDBL_MIN_10_EXP
+
+    FLT_MAX_EXP
+    DBL_MAX_EXP
+    LDBL_MAX_EXP
+
+    FLT_MAX_10_EXP
+    DBL_MAX_10_EXP
+    LDBL_MAX_10_EXP
+
+    FLT_MAX
+    DBL_MAX
+    LDBL_MAX
+
+    FLT_EPSILON
+    DBL_EPSILON
+    LDBL_EPSILON
+
+    FLT_MIN
+    DBL_MIN
+    LDBL_MIN
+
+    FLT_TRUE_MIN        // C11
+    DBL_TRUE_MIN        // C11
+    LDBL_TRUE_MIN       // C11
+
+*/
+
+#include <__config>
+
+#include <__pragma_push>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <float.h>
+
+#ifdef __cplusplus
+
+#ifndef FLT_EVAL_METHOD
+#define FLT_EVAL_METHOD __FLT_EVAL_METHOD__
+#endif
+
+#ifndef DECIMAL_DIG
+#define DECIMAL_DIG __DECIMAL_DIG__
+#endif
+
+#endif // __cplusplus
+
+#include <__pragma_pop>
+
+#endif  // _LIBCUDACXX_FLOAT_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list
new file mode 100644
index 000000000000..cfb6a24c65ee
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/forward_list
@@ -0,0 +1,1781 @@
+// -*- C++ -*-
+//===----------------------- forward_list ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FORWARD_LIST
+#define _LIBCUDACXX_FORWARD_LIST
+
+/*
+    forward_list synopsis
+
+namespace std
+{
+
+template <class T, class Allocator = allocator<T>>
+class forward_list
+{
+public:
+    typedef T         value_type;
+    typedef Allocator allocator_type;
+
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef <details> iterator;
+    typedef <details> const_iterator;
+
+    forward_list()
+        noexcept(is_nothrow_default_constructible<allocator_type>::value);
+    explicit forward_list(const allocator_type& a);
+    explicit forward_list(size_type n);
+    explicit forward_list(size_type n, const allocator_type& a); // C++14
+    forward_list(size_type n, const value_type& v);
+    forward_list(size_type n, const value_type& v, const allocator_type& a);
+    template <class InputIterator>
+        forward_list(InputIterator first, InputIterator last);
+    template <class InputIterator>
+        forward_list(InputIterator first, InputIterator last, const allocator_type& a);
+    forward_list(const forward_list& x);
+    forward_list(const forward_list& x, const allocator_type& a);
+    forward_list(forward_list&& x)
+        noexcept(is_nothrow_move_constructible<allocator_type>::value);
+    forward_list(forward_list&& x, const allocator_type& a);
+    forward_list(initializer_list<value_type> il);
+    forward_list(initializer_list<value_type> il, const allocator_type& a);
+
+    ~forward_list();
+
+    forward_list& operator=(const forward_list& x);
+    forward_list& operator=(forward_list&& x)
+        noexcept(
+             allocator_type::propagate_on_container_move_assignment::value &&
+             is_nothrow_move_assignable<allocator_type>::value);
+    forward_list& operator=(initializer_list<value_type> il);
+
+    template <class InputIterator>
+        void assign(InputIterator first, InputIterator last);
+    void assign(size_type n, const value_type& v);
+    void assign(initializer_list<value_type> il);
+
+    allocator_type get_allocator() const noexcept;
+
+    iterator       begin() noexcept;
+    const_iterator begin() const noexcept;
+    iterator       end() noexcept;
+    const_iterator end() const noexcept;
+
+    const_iterator cbegin() const noexcept;
+    const_iterator cend() const noexcept;
+
+    iterator       before_begin() noexcept;
+    const_iterator before_begin() const noexcept;
+    const_iterator cbefore_begin() const noexcept;
+
+    bool empty() const noexcept;
+    size_type max_size() const noexcept;
+
+    reference       front();
+    const_reference front() const;
+
+    template <class... Args> reference emplace_front(Args&&... args);  // reference in C++17
+    void push_front(const value_type& v);
+    void push_front(value_type&& v);
+
+    void pop_front();
+
+    template <class... Args>
+        iterator emplace_after(const_iterator p, Args&&... args);
+    iterator insert_after(const_iterator p, const value_type& v);
+    iterator insert_after(const_iterator p, value_type&& v);
+    iterator insert_after(const_iterator p, size_type n, const value_type& v);
+    template <class InputIterator>
+        iterator insert_after(const_iterator p,
+                              InputIterator first, InputIterator last);
+    iterator insert_after(const_iterator p, initializer_list<value_type> il);
+
+    iterator erase_after(const_iterator p);
+    iterator erase_after(const_iterator first, const_iterator last);
+
+    void swap(forward_list& x)
+        noexcept(allocator_traits<allocator_type>::is_always_equal::value);  // C++17
+
+    void resize(size_type n);
+    void resize(size_type n, const value_type& v);
+    void clear() noexcept;
+
+    void splice_after(const_iterator p, forward_list& x);
+    void splice_after(const_iterator p, forward_list&& x);
+    void splice_after(const_iterator p, forward_list& x, const_iterator i);
+    void splice_after(const_iterator p, forward_list&& x, const_iterator i);
+    void splice_after(const_iterator p, forward_list& x,
+                      const_iterator first, const_iterator last);
+    void splice_after(const_iterator p, forward_list&& x,
+                      const_iterator first, const_iterator last);
+    size_type remove(const value_type& v);           // void before C++20
+    template <class Predicate>
+      size_type remove_if(Predicate pred);           // void before C++20
+    size_type unique();                              // void before C++20
+    template <class BinaryPredicate>
+      size_type unique(BinaryPredicate binary_pred); // void before C++20
+    void merge(forward_list& x);
+    void merge(forward_list&& x);
+    template <class Compare> void merge(forward_list& x, Compare comp);
+    template <class Compare> void merge(forward_list&& x, Compare comp);
+    void sort();
+    template <class Compare> void sort(Compare comp);
+    void reverse() noexcept;
+};
+
+
+template <class InputIterator, class Allocator = allocator<typename iterator_traits<InputIterator>::value_type>>
+    forward_list(InputIterator, InputIterator, Allocator = Allocator())
+    -> forward_list<typename iterator_traits<InputIterator>::value_type, Allocator>;  // C++17
+
+template <class T, class Allocator>
+    bool operator==(const forward_list<T, Allocator>& x,
+                    const forward_list<T, Allocator>& y);
+
+template <class T, class Allocator>
+    bool operator< (const forward_list<T, Allocator>& x,
+                    const forward_list<T, Allocator>& y);
+
+template <class T, class Allocator>
+    bool operator!=(const forward_list<T, Allocator>& x,
+                    const forward_list<T, Allocator>& y);
+
+template <class T, class Allocator>
+    bool operator> (const forward_list<T, Allocator>& x,
+                    const forward_list<T, Allocator>& y);
+
+template <class T, class Allocator>
+    bool operator>=(const forward_list<T, Allocator>& x,
+                    const forward_list<T, Allocator>& y);
+
+template <class T, class Allocator>
+    bool operator<=(const forward_list<T, Allocator>& x,
+                    const forward_list<T, Allocator>& y);
+
+template <class T, class Allocator>
+    void swap(forward_list<T, Allocator>& x, forward_list<T, Allocator>& y)
+         noexcept(noexcept(x.swap(y)));
+
+template <class T, class Allocator, class U>
+    void erase(forward_list<T, Allocator>& c, const U& value);       // C++20
+template <class T, class Allocator, class Predicate>
+    void erase_if(forward_list<T, Allocator>& c, Predicate pred);    // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <initializer_list>
+#include <memory>
+#include <limits>
+#include <iterator>
+#include <algorithm>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp, class _VoidPtr> struct __forward_list_node;
+template <class _NodePtr> struct __forward_begin_node;
+
+
+template <class>
+struct __forward_list_node_value_type;
+
+template <class _Tp, class _VoidPtr>
+struct __forward_list_node_value_type<__forward_list_node<_Tp, _VoidPtr> > {
+  typedef _Tp type;
+};
+
+template <class _NodePtr>
+struct __forward_node_traits {
+
+  typedef typename remove_cv<
+        typename pointer_traits<_NodePtr>::element_type>::type  __node;
+  typedef typename __forward_list_node_value_type<__node>::type __node_value_type;
+  typedef _NodePtr                                              __node_pointer;
+  typedef __forward_begin_node<_NodePtr>                        __begin_node;
+  typedef typename __rebind_pointer<_NodePtr, __begin_node>::type
+                                                                __begin_node_pointer;
+  typedef typename __rebind_pointer<_NodePtr, void>::type       __void_pointer;
+
+#if defined(_LIBCUDACXX_ABI_FORWARD_LIST_REMOVE_NODE_POINTER_UB)
+  typedef __begin_node_pointer __iter_node_pointer;
+#else
+  typedef typename conditional<
+          is_pointer<__void_pointer>::value,
+          __begin_node_pointer,
+          __node_pointer
+    >::type __iter_node_pointer;
+#endif
+
+  typedef typename conditional<
+          is_same<__iter_node_pointer, __node_pointer>::value,
+          __begin_node_pointer,
+          __node_pointer
+    >::type __non_iter_node_pointer;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __iter_node_pointer __as_iter_node(__iter_node_pointer __p) {
+      return __p;
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __iter_node_pointer __as_iter_node(__non_iter_node_pointer __p) {
+      return static_cast<__iter_node_pointer>(static_cast<__void_pointer>(__p));
+  }
+};
+
+template <class _NodePtr>
+struct __forward_begin_node
+{
+    typedef _NodePtr pointer;
+    typedef typename __rebind_pointer<_NodePtr, __forward_begin_node>::type __begin_node_pointer;
+
+    pointer __next_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __forward_begin_node() : __next_(nullptr) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __begin_node_pointer __next_as_begin() const {
+        return static_cast<__begin_node_pointer>(__next_);
+    }
+};
+
+template <class _Tp, class _VoidPtr>
+struct _LIBCUDACXX_HIDDEN __begin_node_of
+{
+    typedef __forward_begin_node<
+        typename __rebind_pointer<_VoidPtr, __forward_list_node<_Tp, _VoidPtr> >::type
+    > type;
+};
+
+template <class _Tp, class _VoidPtr>
+struct __forward_list_node
+    : public __begin_node_of<_Tp, _VoidPtr>::type
+{
+    typedef _Tp value_type;
+
+    value_type __value_;
+};
+
+
+template <class _Tp, class _Alloc = allocator<_Tp> > class _LIBCUDACXX_TEMPLATE_VIS forward_list;
+template<class _NodeConstPtr> class _LIBCUDACXX_TEMPLATE_VIS __forward_list_const_iterator;
+
+template <class _NodePtr>
+class _LIBCUDACXX_TEMPLATE_VIS __forward_list_iterator
+{
+    typedef __forward_node_traits<_NodePtr>         __traits;
+    typedef typename __traits::__node_pointer       __node_pointer;
+    typedef typename __traits::__begin_node_pointer __begin_node_pointer;
+    typedef typename __traits::__iter_node_pointer  __iter_node_pointer;
+    typedef typename __traits::__void_pointer       __void_pointer;
+
+    __iter_node_pointer __ptr_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __begin_node_pointer __get_begin() const {
+        return static_cast<__begin_node_pointer>(
+                static_cast<__void_pointer>(__ptr_));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_pointer __get_unsafe_node_pointer() const {
+        return static_cast<__node_pointer>(
+                static_cast<__void_pointer>(__ptr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_iterator(nullptr_t) _NOEXCEPT : __ptr_(nullptr) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_iterator(__begin_node_pointer __p) _NOEXCEPT
+        : __ptr_(__traits::__as_iter_node(__p)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_iterator(__node_pointer __p) _NOEXCEPT
+        : __ptr_(__traits::__as_iter_node(__p)) {}
+
+    template<class, class> friend class _LIBCUDACXX_TEMPLATE_VIS forward_list;
+    template<class> friend class _LIBCUDACXX_TEMPLATE_VIS __forward_list_const_iterator;
+
+public:
+    typedef forward_iterator_tag                              iterator_category;
+    typedef typename __traits::__node_value_type              value_type;
+    typedef value_type&                                       reference;
+    typedef typename pointer_traits<__node_pointer>::difference_type
+                                                              difference_type;
+    typedef typename __rebind_pointer<__node_pointer, value_type>::type pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_iterator() _NOEXCEPT : __ptr_(nullptr) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return __get_unsafe_node_pointer()->__value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {
+        return pointer_traits<pointer>::pointer_to(__get_unsafe_node_pointer()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_iterator& operator++()
+    {
+        __ptr_ = __traits::__as_iter_node(__ptr_->__next_);
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_iterator operator++(int)
+    {
+        __forward_list_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __forward_list_iterator& __x,
+                    const __forward_list_iterator& __y)
+        {return __x.__ptr_ == __y.__ptr_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __forward_list_iterator& __x,
+                    const __forward_list_iterator& __y)
+        {return !(__x == __y);}
+};
+
+template <class _NodeConstPtr>
+class _LIBCUDACXX_TEMPLATE_VIS __forward_list_const_iterator
+{
+    static_assert((!is_const<typename pointer_traits<_NodeConstPtr>::element_type>::value), "");
+    typedef _NodeConstPtr _NodePtr;
+
+    typedef __forward_node_traits<_NodePtr>         __traits;
+    typedef typename __traits::__node               __node;
+    typedef typename __traits::__node_pointer       __node_pointer;
+    typedef typename __traits::__begin_node_pointer __begin_node_pointer;
+    typedef typename __traits::__iter_node_pointer  __iter_node_pointer;
+    typedef typename __traits::__void_pointer       __void_pointer;
+
+    __iter_node_pointer __ptr_;
+
+    __begin_node_pointer __get_begin() const {
+        return static_cast<__begin_node_pointer>(
+                static_cast<__void_pointer>(__ptr_));
+    }
+    __node_pointer __get_unsafe_node_pointer() const {
+        return static_cast<__node_pointer>(
+                static_cast<__void_pointer>(__ptr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_const_iterator(nullptr_t) _NOEXCEPT
+        : __ptr_(nullptr) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_const_iterator(__begin_node_pointer __p) _NOEXCEPT
+        : __ptr_(__traits::__as_iter_node(__p)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_const_iterator(__node_pointer __p) _NOEXCEPT
+        : __ptr_(__traits::__as_iter_node(__p)) {}
+
+
+    template<class, class> friend class forward_list;
+
+public:
+    typedef forward_iterator_tag                              iterator_category;
+    typedef typename __traits::__node_value_type              value_type;
+    typedef const value_type&                                 reference;
+    typedef typename pointer_traits<__node_pointer>::difference_type
+                                                              difference_type;
+    typedef typename __rebind_pointer<__node_pointer, const value_type>::type
+                                                              pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_const_iterator() _NOEXCEPT : __ptr_(nullptr) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_const_iterator(__forward_list_iterator<__node_pointer> __p) _NOEXCEPT
+        : __ptr_(__p.__ptr_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return __get_unsafe_node_pointer()->__value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return pointer_traits<pointer>::pointer_to(
+                __get_unsafe_node_pointer()->__value_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_const_iterator& operator++()
+    {
+        __ptr_ = __traits::__as_iter_node(__ptr_->__next_);
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_const_iterator operator++(int)
+    {
+        __forward_list_const_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __forward_list_const_iterator& __x,
+                    const __forward_list_const_iterator& __y)
+        {return __x.__ptr_ == __y.__ptr_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __forward_list_const_iterator& __x,
+                           const __forward_list_const_iterator& __y)
+        {return !(__x == __y);}
+};
+
+template <class _Tp, class _Alloc>
+class __forward_list_base
+{
+protected:
+    typedef _Tp    value_type;
+    typedef _Alloc allocator_type;
+
+    typedef typename allocator_traits<allocator_type>::void_pointer  void_pointer;
+    typedef __forward_list_node<value_type, void_pointer>            __node;
+    typedef typename __begin_node_of<value_type, void_pointer>::type __begin_node;
+    typedef typename __rebind_alloc_helper<allocator_traits<allocator_type>, __node>::type __node_allocator;
+    typedef allocator_traits<__node_allocator>        __node_traits;
+    typedef typename __node_traits::pointer           __node_pointer;
+
+    typedef typename __rebind_alloc_helper<
+        allocator_traits<allocator_type>, __begin_node
+    >::type                                           __begin_node_allocator;
+    typedef typename allocator_traits<__begin_node_allocator>::pointer
+                                                      __begin_node_pointer;
+
+    static_assert((!is_same<allocator_type, __node_allocator>::value),
+                  "internal allocator type must differ from user-specified "
+                  "type; otherwise overload resolution breaks");
+
+    __compressed_pair<__begin_node, __node_allocator> __before_begin_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __begin_node_pointer        __before_begin() _NOEXCEPT
+        {return pointer_traits<__begin_node_pointer>::pointer_to(__before_begin_.first());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __begin_node_pointer __before_begin() const _NOEXCEPT
+        {return pointer_traits<__begin_node_pointer>::pointer_to(const_cast<__begin_node&>(__before_begin_.first()));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          __node_allocator& __alloc() _NOEXCEPT
+            {return __before_begin_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const __node_allocator& __alloc() const _NOEXCEPT
+        {return __before_begin_.second();}
+
+    typedef __forward_list_iterator<__node_pointer>             iterator;
+    typedef __forward_list_const_iterator<__node_pointer>       const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_base()
+        _NOEXCEPT_(is_nothrow_default_constructible<__node_allocator>::value)
+        : __before_begin_(__begin_node()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_base(const allocator_type& __a)
+        : __before_begin_(__begin_node(), __node_allocator(__a)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __forward_list_base(const __node_allocator& __a)
+        : __before_begin_(__begin_node(), __a) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_base(__forward_list_base&& __x)
+        _NOEXCEPT_(is_nothrow_move_constructible<__node_allocator>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __forward_list_base(__forward_list_base&& __x, const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+private:
+    __forward_list_base(const __forward_list_base&);
+    __forward_list_base& operator=(const __forward_list_base&);
+
+public:
+    ~__forward_list_base();
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __forward_list_base& __x)
+        {__copy_assign_alloc(__x, integral_constant<bool,
+              __node_traits::propagate_on_container_copy_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__forward_list_base& __x)
+        _NOEXCEPT_(!__node_traits::propagate_on_container_move_assignment::value ||
+                   is_nothrow_move_assignable<__node_allocator>::value)
+        {__move_assign_alloc(__x, integral_constant<bool,
+              __node_traits::propagate_on_container_move_assignment::value>());}
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(__forward_list_base& __x)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__node_traits::propagate_on_container_move_assignment::value ||
+                    __is_nothrow_swappable<__node_allocator>::value);
+#endif
+protected:
+    void clear() _NOEXCEPT;
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __forward_list_base&, false_type) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __forward_list_base& __x, true_type)
+    {
+        if (__alloc() != __x.__alloc())
+            clear();
+        __alloc() = __x.__alloc();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__forward_list_base&, false_type) _NOEXCEPT
+        {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__forward_list_base& __x, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<__node_allocator>::value)
+        {__alloc() = _CUDA_VSTD::move(__x.__alloc());}
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+inline
+__forward_list_base<_Tp, _Alloc>::__forward_list_base(__forward_list_base&& __x)
+        _NOEXCEPT_(is_nothrow_move_constructible<__node_allocator>::value)
+    : __before_begin_(_CUDA_VSTD::move(__x.__before_begin_))
+{
+    __x.__before_begin()->__next_ = nullptr;
+}
+
+template <class _Tp, class _Alloc>
+inline
+__forward_list_base<_Tp, _Alloc>::__forward_list_base(__forward_list_base&& __x,
+                                                      const allocator_type& __a)
+    : __before_begin_(__begin_node(), __node_allocator(__a))
+{
+    if (__alloc() == __x.__alloc())
+    {
+        __before_begin()->__next_ = __x.__before_begin()->__next_;
+        __x.__before_begin()->__next_ = nullptr;
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+__forward_list_base<_Tp, _Alloc>::~__forward_list_base()
+{
+    clear();
+}
+
+template <class _Tp, class _Alloc>
+inline
+void
+__forward_list_base<_Tp, _Alloc>::swap(__forward_list_base& __x)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__node_traits::propagate_on_container_move_assignment::value ||
+                    __is_nothrow_swappable<__node_allocator>::value)
+#endif
+{
+    __swap_allocator(__alloc(), __x.__alloc(),
+            integral_constant<bool, __node_traits::propagate_on_container_swap::value>());
+    using _CUDA_VSTD::swap;
+    swap(__before_begin()->__next_, __x.__before_begin()->__next_);
+}
+
+template <class _Tp, class _Alloc>
+void
+__forward_list_base<_Tp, _Alloc>::clear() _NOEXCEPT
+{
+    __node_allocator& __a = __alloc();
+    for (__node_pointer __p = __before_begin()->__next_; __p != nullptr;)
+    {
+        __node_pointer __next = __p->__next_;
+        __node_traits::destroy(__a, _CUDA_VSTD::addressof(__p->__value_));
+        __node_traits::deallocate(__a, __p, 1);
+        __p = __next;
+    }
+    __before_begin()->__next_ = nullptr;
+}
+
+template <class _Tp, class _Alloc /*= allocator<_Tp>*/>
+class _LIBCUDACXX_TEMPLATE_VIS forward_list
+    : private __forward_list_base<_Tp, _Alloc>
+{
+    typedef __forward_list_base<_Tp, _Alloc> base;
+    typedef typename base::__node_allocator  __node_allocator;
+    typedef typename base::__node               __node;
+    typedef typename base::__node_traits        __node_traits;
+    typedef typename base::__node_pointer       __node_pointer;
+    typedef typename base::__begin_node_pointer __begin_node_pointer;
+
+public:
+    typedef _Tp    value_type;
+    typedef _Alloc allocator_type;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef typename base::iterator       iterator;
+    typedef typename base::const_iterator const_iterator;
+#if _LIBCUDACXX_STD_VER > 17
+    typedef size_type                                __remove_return_type;
+#else
+    typedef void                                     __remove_return_type;
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    forward_list()
+        _NOEXCEPT_(is_nothrow_default_constructible<__node_allocator>::value)
+        {} // = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit forward_list(const allocator_type& __a);
+    explicit forward_list(size_type __n);
+#if _LIBCUDACXX_STD_VER > 11
+    explicit forward_list(size_type __n, const allocator_type& __a);
+#endif
+    forward_list(size_type __n, const value_type& __v);
+    forward_list(size_type __n, const value_type& __v, const allocator_type& __a);
+    template <class _InputIterator>
+        forward_list(_InputIterator __f, _InputIterator __l,
+                     typename enable_if<
+                       __is_input_iterator<_InputIterator>::value
+                     >::type* = nullptr);
+    template <class _InputIterator>
+        forward_list(_InputIterator __f, _InputIterator __l,
+                     const allocator_type& __a,
+                     typename enable_if<
+                       __is_input_iterator<_InputIterator>::value
+                     >::type* = nullptr);
+    forward_list(const forward_list& __x);
+    forward_list(const forward_list& __x, const allocator_type& __a);
+
+    forward_list& operator=(const forward_list& __x);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    forward_list(forward_list&& __x)
+        _NOEXCEPT_(is_nothrow_move_constructible<base>::value)
+        : base(_CUDA_VSTD::move(__x)) {}
+    forward_list(forward_list&& __x, const allocator_type& __a);
+
+    forward_list(initializer_list<value_type> __il);
+    forward_list(initializer_list<value_type> __il, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    forward_list& operator=(forward_list&& __x)
+        _NOEXCEPT_(
+             __node_traits::propagate_on_container_move_assignment::value &&
+             is_nothrow_move_assignable<allocator_type>::value);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    forward_list& operator=(initializer_list<value_type> __il);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(initializer_list<value_type> __il);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    // ~forward_list() = default;
+
+    template <class _InputIterator>
+        typename enable_if
+        <
+            __is_input_iterator<_InputIterator>::value,
+            void
+        >::type
+        assign(_InputIterator __f, _InputIterator __l);
+    void assign(size_type __n, const value_type& __v);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT
+        {return allocator_type(base::__alloc());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT
+        {return       iterator(base::__before_begin()->__next_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT
+        {return const_iterator(base::__before_begin()->__next_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT
+        {return       iterator(nullptr);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT
+        {return const_iterator(nullptr);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT
+        {return const_iterator(base::__before_begin()->__next_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT
+        {return const_iterator(nullptr);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       before_begin() _NOEXCEPT
+        {return       iterator(base::__before_begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator before_begin() const _NOEXCEPT
+        {return const_iterator(base::__before_begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbefore_begin() const _NOEXCEPT
+        {return const_iterator(base::__before_begin());}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT
+        {return base::__before_begin()->__next_ == nullptr;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {
+        return std::min<size_type>(
+            __node_traits::max_size(base::__alloc()),
+            numeric_limits<difference_type>::max());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference       front()       {return base::__before_begin()->__next_->__value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference front() const {return base::__before_begin()->__next_->__value_;}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+#if _LIBCUDACXX_STD_VER > 14
+    template <class... _Args> reference emplace_front(_Args&&... __args);
+#else
+    template <class... _Args> void      emplace_front(_Args&&... __args);
+#endif
+    void push_front(value_type&& __v);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    void push_front(const value_type& __v);
+
+    void pop_front();
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class... _Args>
+        iterator emplace_after(const_iterator __p, _Args&&... __args);
+
+    iterator insert_after(const_iterator __p, value_type&& __v);
+    iterator insert_after(const_iterator __p, initializer_list<value_type> __il)
+        {return insert_after(__p, __il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    iterator insert_after(const_iterator __p, const value_type& __v);
+    iterator insert_after(const_iterator __p, size_type __n, const value_type& __v);
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_input_iterator<_InputIterator>::value,
+            iterator
+        >::type
+        insert_after(const_iterator __p, _InputIterator __f, _InputIterator __l);
+
+    iterator erase_after(const_iterator __p);
+    iterator erase_after(const_iterator __f, const_iterator __l);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(forward_list& __x)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__node_traits::propagate_on_container_swap::value ||
+                   __is_nothrow_swappable<__node_allocator>::value)
+#endif
+        {base::swap(__x);}
+
+    void resize(size_type __n);
+    void resize(size_type __n, const value_type& __v);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {base::clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void splice_after(const_iterator __p, forward_list&& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void splice_after(const_iterator __p, forward_list&& __x, const_iterator __i);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void splice_after(const_iterator __p, forward_list&& __x,
+                      const_iterator __f, const_iterator __l);
+    void splice_after(const_iterator __p, forward_list& __x);
+    void splice_after(const_iterator __p, forward_list& __x, const_iterator __i);
+    void splice_after(const_iterator __p, forward_list& __x,
+                      const_iterator __f, const_iterator __l);
+    __remove_return_type remove(const value_type& __v);
+    template <class _Predicate> __remove_return_type remove_if(_Predicate __pred);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __remove_return_type unique() {return unique(__equal_to<value_type>());}
+    template <class _BinaryPredicate> __remove_return_type unique(_BinaryPredicate __binary_pred);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(forward_list&& __x) {merge(__x, __less<value_type>());}
+    template <class _Compare>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void merge(forward_list&& __x, _Compare __comp)
+        {merge(__x, _CUDA_VSTD::move(__comp));}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(forward_list& __x) {merge(__x, __less<value_type>());}
+    template <class _Compare> void merge(forward_list& __x, _Compare __comp);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void sort() {sort(__less<value_type>());}
+    template <class _Compare> _LIBCUDACXX_INLINE_VISIBILITY void sort(_Compare __comp);
+    void reverse() _NOEXCEPT;
+
+private:
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    void __move_assign(forward_list& __x, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value);
+    void __move_assign(forward_list& __x, false_type);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template <class _Compare>
+        static
+        __node_pointer
+        __merge(__node_pointer __f1, __node_pointer __f2, _Compare& __comp);
+
+    template <class _Compare>
+        static
+        __node_pointer
+        __sort(__node_pointer __f, difference_type __sz, _Compare& __comp);
+};
+
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Alloc = typename std::allocator<typename iterator_traits<_InputIterator>::value_type>,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+forward_list(_InputIterator, _InputIterator)
+  -> forward_list<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+
+template<class _InputIterator,
+         class _Alloc,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+forward_list(_InputIterator, _InputIterator, _Alloc)
+  -> forward_list<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+#endif
+
+template <class _Tp, class _Alloc>
+inline
+forward_list<_Tp, _Alloc>::forward_list(const allocator_type& __a)
+    : base(__a)
+{
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(size_type __n)
+{
+    if (__n > 0)
+    {
+        __node_allocator& __a = base::__alloc();
+        typedef __allocator_destructor<__node_allocator> _Dp;
+        unique_ptr<__node, _Dp> __h(nullptr, _Dp(__a, 1));
+        for (__begin_node_pointer __p = base::__before_begin(); __n > 0; --__n,
+                                                             __p = __p->__next_as_begin())
+        {
+            __h.reset(__node_traits::allocate(__a, 1));
+            __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_));
+            __h->__next_ = nullptr;
+            __p->__next_ = __h.release();
+        }
+    }
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(size_type __n,
+                                        const allocator_type& __base_alloc)
+    : base ( __base_alloc )
+{
+    if (__n > 0)
+    {
+        __node_allocator& __a = base::__alloc();
+        typedef __allocator_destructor<__node_allocator> _Dp;
+        unique_ptr<__node, _Dp> __h(nullptr, _Dp(__a, 1));
+        for (__begin_node_pointer __p = base::__before_begin(); __n > 0; --__n,
+                                                             __p = __p->__next_as_begin())
+        {
+            __h.reset(__node_traits::allocate(__a, 1));
+            __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_));
+            __h->__next_ = nullptr;
+            __p->__next_ = __h.release();
+        }
+    }
+}
+#endif
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(size_type __n, const value_type& __v)
+{
+    insert_after(cbefore_begin(), __n, __v);
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(size_type __n, const value_type& __v,
+                                        const allocator_type& __a)
+    : base(__a)
+{
+    insert_after(cbefore_begin(), __n, __v);
+}
+
+template <class _Tp, class _Alloc>
+template <class _InputIterator>
+forward_list<_Tp, _Alloc>::forward_list(_InputIterator __f, _InputIterator __l,
+                                        typename enable_if<
+                                          __is_input_iterator<_InputIterator>::value
+                                        >::type*)
+{
+    insert_after(cbefore_begin(), __f, __l);
+}
+
+template <class _Tp, class _Alloc>
+template <class _InputIterator>
+forward_list<_Tp, _Alloc>::forward_list(_InputIterator __f, _InputIterator __l,
+                                        const allocator_type& __a,
+                                        typename enable_if<
+                                          __is_input_iterator<_InputIterator>::value
+                                        >::type*)
+    : base(__a)
+{
+    insert_after(cbefore_begin(), __f, __l);
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(const forward_list& __x)
+    : base(
+          __node_traits::select_on_container_copy_construction(__x.__alloc())) {
+  insert_after(cbefore_begin(), __x.begin(), __x.end());
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(const forward_list& __x,
+                                        const allocator_type& __a)
+    : base(__a)
+{
+    insert_after(cbefore_begin(), __x.begin(), __x.end());
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>&
+forward_list<_Tp, _Alloc>::operator=(const forward_list& __x)
+{
+    if (this != &__x)
+    {
+        base::__copy_assign_alloc(__x);
+        assign(__x.begin(), __x.end());
+    }
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(forward_list&& __x,
+                                        const allocator_type& __a)
+    : base(_CUDA_VSTD::move(__x), __a)
+{
+    if (base::__alloc() != __x.__alloc())
+    {
+        typedef move_iterator<iterator> _Ip;
+        insert_after(cbefore_begin(), _Ip(__x.begin()), _Ip(__x.end()));
+    }
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(initializer_list<value_type> __il)
+{
+    insert_after(cbefore_begin(), __il.begin(), __il.end());
+}
+
+template <class _Tp, class _Alloc>
+forward_list<_Tp, _Alloc>::forward_list(initializer_list<value_type> __il,
+                                        const allocator_type& __a)
+    : base(__a)
+{
+    insert_after(cbefore_begin(), __il.begin(), __il.end());
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::__move_assign(forward_list& __x, true_type)
+    _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+{
+    clear();
+    base::__move_assign_alloc(__x);
+    base::__before_begin()->__next_ = __x.__before_begin()->__next_;
+    __x.__before_begin()->__next_ = nullptr;
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::__move_assign(forward_list& __x, false_type)
+{
+    if (base::__alloc() == __x.__alloc())
+        __move_assign(__x, true_type());
+    else
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__x.begin()), _Ip(__x.end()));
+    }
+}
+
+template <class _Tp, class _Alloc>
+inline
+forward_list<_Tp, _Alloc>&
+forward_list<_Tp, _Alloc>::operator=(forward_list&& __x)
+    _NOEXCEPT_(
+             __node_traits::propagate_on_container_move_assignment::value &&
+             is_nothrow_move_assignable<allocator_type>::value)
+{
+    __move_assign(__x, integral_constant<bool,
+          __node_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+template <class _Tp, class _Alloc>
+inline
+forward_list<_Tp, _Alloc>&
+forward_list<_Tp, _Alloc>::operator=(initializer_list<value_type> __il)
+{
+    assign(__il.begin(), __il.end());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+template <class _InputIterator>
+typename enable_if
+<
+    __is_input_iterator<_InputIterator>::value,
+    void
+>::type
+forward_list<_Tp, _Alloc>::assign(_InputIterator __f, _InputIterator __l)
+{
+    iterator __i = before_begin();
+    iterator __j = _CUDA_VSTD::next(__i);
+    iterator __e = end();
+    for (; __j != __e && __f != __l; ++__i, (void) ++__j, ++__f)
+        *__j = *__f;
+    if (__j == __e)
+        insert_after(__i, __f, __l);
+    else
+        erase_after(__i, __e);
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::assign(size_type __n, const value_type& __v)
+{
+    iterator __i = before_begin();
+    iterator __j = _CUDA_VSTD::next(__i);
+    iterator __e = end();
+    for (; __j != __e && __n > 0; --__n, ++__i, ++__j)
+        *__j = __v;
+    if (__j == __e)
+        insert_after(__i, __n, __v);
+    else
+        erase_after(__i, __e);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+inline
+void
+forward_list<_Tp, _Alloc>::assign(initializer_list<value_type> __il)
+{
+    assign(__il.begin(), __il.end());
+}
+
+template <class _Tp, class _Alloc>
+template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+typename forward_list<_Tp, _Alloc>::reference
+#else
+void
+#endif
+forward_list<_Tp, _Alloc>::emplace_front(_Args&&... __args)
+{
+    __node_allocator& __a = base::__alloc();
+    typedef __allocator_destructor<__node_allocator> _Dp;
+    unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+    __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_),
+                                  _CUDA_VSTD::forward<_Args>(__args)...);
+    __h->__next_ = base::__before_begin()->__next_;
+    base::__before_begin()->__next_ = __h.release();
+#if _LIBCUDACXX_STD_VER > 14
+    return base::__before_begin()->__next_->__value_;
+#endif
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::push_front(value_type&& __v)
+{
+    __node_allocator& __a = base::__alloc();
+    typedef __allocator_destructor<__node_allocator> _Dp;
+    unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+    __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), _CUDA_VSTD::move(__v));
+    __h->__next_ = base::__before_begin()->__next_;
+    base::__before_begin()->__next_ = __h.release();
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::push_front(const value_type& __v)
+{
+    __node_allocator& __a = base::__alloc();
+    typedef __allocator_destructor<__node_allocator> _Dp;
+    unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+    __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), __v);
+    __h->__next_ = base::__before_begin()->__next_;
+    base::__before_begin()->__next_ = __h.release();
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::pop_front()
+{
+    __node_allocator& __a = base::__alloc();
+    __node_pointer __p = base::__before_begin()->__next_;
+    base::__before_begin()->__next_ = __p->__next_;
+    __node_traits::destroy(__a, _CUDA_VSTD::addressof(__p->__value_));
+    __node_traits::deallocate(__a, __p, 1);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+template <class... _Args>
+typename forward_list<_Tp, _Alloc>::iterator
+forward_list<_Tp, _Alloc>::emplace_after(const_iterator __p, _Args&&... __args)
+{
+    __begin_node_pointer const __r = __p.__get_begin();
+    __node_allocator& __a = base::__alloc();
+    typedef __allocator_destructor<__node_allocator> _Dp;
+    unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+    __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_),
+                                  _CUDA_VSTD::forward<_Args>(__args)...);
+    __h->__next_ = __r->__next_;
+    __r->__next_ = __h.release();
+    return iterator(__r->__next_);
+}
+
+template <class _Tp, class _Alloc>
+typename forward_list<_Tp, _Alloc>::iterator
+forward_list<_Tp, _Alloc>::insert_after(const_iterator __p, value_type&& __v)
+{
+    __begin_node_pointer const __r = __p.__get_begin();
+    __node_allocator& __a = base::__alloc();
+    typedef __allocator_destructor<__node_allocator> _Dp;
+    unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+    __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), _CUDA_VSTD::move(__v));
+    __h->__next_ = __r->__next_;
+    __r->__next_ = __h.release();
+    return iterator(__r->__next_);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+typename forward_list<_Tp, _Alloc>::iterator
+forward_list<_Tp, _Alloc>::insert_after(const_iterator __p, const value_type& __v)
+{
+    __begin_node_pointer const __r = __p.__get_begin();
+    __node_allocator& __a = base::__alloc();
+    typedef __allocator_destructor<__node_allocator> _Dp;
+    unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+    __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), __v);
+    __h->__next_ = __r->__next_;
+    __r->__next_ = __h.release();
+    return iterator(__r->__next_);
+}
+
+template <class _Tp, class _Alloc>
+typename forward_list<_Tp, _Alloc>::iterator
+forward_list<_Tp, _Alloc>::insert_after(const_iterator __p, size_type __n,
+                                        const value_type& __v)
+{
+    __begin_node_pointer __r = __p.__get_begin();
+    if (__n > 0)
+    {
+        __node_allocator& __a = base::__alloc();
+        typedef __allocator_destructor<__node_allocator> _Dp;
+        unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+        __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), __v);
+        __node_pointer __first = __h.release();
+        __node_pointer __last = __first;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (--__n; __n != 0; --__n, __last = __last->__next_)
+            {
+                __h.reset(__node_traits::allocate(__a, 1));
+                __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), __v);
+                __last->__next_ = __h.release();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            while (__first != nullptr)
+            {
+                __node_pointer __next = __first->__next_;
+                __node_traits::destroy(__a, _CUDA_VSTD::addressof(__first->__value_));
+                __node_traits::deallocate(__a, __first, 1);
+                __first = __next;
+            }
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __last->__next_ = __r->__next_;
+        __r->__next_ = __first;
+        __r = static_cast<__begin_node_pointer>(__last);
+    }
+    return iterator(__r);
+}
+
+template <class _Tp, class _Alloc>
+template <class _InputIterator>
+typename enable_if
+<
+    __is_input_iterator<_InputIterator>::value,
+    typename forward_list<_Tp, _Alloc>::iterator
+>::type
+forward_list<_Tp, _Alloc>::insert_after(const_iterator __p,
+                                        _InputIterator __f, _InputIterator __l)
+{
+    __begin_node_pointer __r = __p.__get_begin();
+    if (__f != __l)
+    {
+        __node_allocator& __a = base::__alloc();
+        typedef __allocator_destructor<__node_allocator> _Dp;
+        unique_ptr<__node, _Dp> __h(__node_traits::allocate(__a, 1), _Dp(__a, 1));
+        __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), *__f);
+        __node_pointer __first = __h.release();
+        __node_pointer __last = __first;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (++__f; __f != __l; ++__f, ((void)(__last = __last->__next_)))
+            {
+                __h.reset(__node_traits::allocate(__a, 1));
+                __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), *__f);
+                __last->__next_ = __h.release();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            while (__first != nullptr)
+            {
+                __node_pointer __next = __first->__next_;
+                __node_traits::destroy(__a, _CUDA_VSTD::addressof(__first->__value_));
+                __node_traits::deallocate(__a, __first, 1);
+                __first = __next;
+            }
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __last->__next_ = __r->__next_;
+        __r->__next_ = __first;
+        __r = static_cast<__begin_node_pointer>(__last);
+    }
+    return iterator(__r);
+}
+
+template <class _Tp, class _Alloc>
+typename forward_list<_Tp, _Alloc>::iterator
+forward_list<_Tp, _Alloc>::erase_after(const_iterator __f)
+{
+    __begin_node_pointer __p = __f.__get_begin();
+    __node_pointer __n = __p->__next_;
+    __p->__next_ = __n->__next_;
+    __node_allocator& __a = base::__alloc();
+    __node_traits::destroy(__a, _CUDA_VSTD::addressof(__n->__value_));
+    __node_traits::deallocate(__a, __n, 1);
+    return iterator(__p->__next_);
+}
+
+template <class _Tp, class _Alloc>
+typename forward_list<_Tp, _Alloc>::iterator
+forward_list<_Tp, _Alloc>::erase_after(const_iterator __f, const_iterator __l)
+{
+    __node_pointer __e = __l.__get_unsafe_node_pointer();
+    if (__f != __l)
+    {
+        __begin_node_pointer __bp = __f.__get_begin();
+
+        __node_pointer __n = __bp->__next_;
+        if (__n != __e)
+        {
+            __bp->__next_ = __e;
+            __node_allocator& __a = base::__alloc();
+            do
+            {
+                __node_pointer __tmp = __n->__next_;
+                __node_traits::destroy(__a, _CUDA_VSTD::addressof(__n->__value_));
+                __node_traits::deallocate(__a, __n, 1);
+                __n = __tmp;
+            } while (__n != __e);
+        }
+    }
+    return iterator(__e);
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::resize(size_type __n)
+{
+    size_type __sz = 0;
+    iterator __p = before_begin();
+    iterator __i = begin();
+    iterator __e = end();
+    for (; __i != __e && __sz < __n; ++__p, ++__i, ++__sz)
+        ;
+    if (__i != __e)
+        erase_after(__p, __e);
+    else
+    {
+        __n -= __sz;
+        if (__n > 0)
+        {
+            __node_allocator& __a = base::__alloc();
+            typedef __allocator_destructor<__node_allocator> _Dp;
+            unique_ptr<__node, _Dp> __h(nullptr, _Dp(__a, 1));
+            for (__begin_node_pointer __ptr = __p.__get_begin(); __n > 0; --__n,
+                                                         __ptr = __ptr->__next_as_begin())
+            {
+                __h.reset(__node_traits::allocate(__a, 1));
+                __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_));
+                __h->__next_ = nullptr;
+                __ptr->__next_ = __h.release();
+            }
+        }
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::resize(size_type __n, const value_type& __v)
+{
+    size_type __sz = 0;
+    iterator __p = before_begin();
+    iterator __i = begin();
+    iterator __e = end();
+    for (; __i != __e && __sz < __n; ++__p, ++__i, ++__sz)
+        ;
+    if (__i != __e)
+        erase_after(__p, __e);
+    else
+    {
+        __n -= __sz;
+        if (__n > 0)
+        {
+            __node_allocator& __a = base::__alloc();
+            typedef __allocator_destructor<__node_allocator> _Dp;
+            unique_ptr<__node, _Dp> __h(nullptr, _Dp(__a, 1));
+            for (__begin_node_pointer __ptr = __p.__get_begin(); __n > 0; --__n,
+                                                         __ptr = __ptr->__next_as_begin())
+            {
+                __h.reset(__node_traits::allocate(__a, 1));
+                __node_traits::construct(__a, _CUDA_VSTD::addressof(__h->__value_), __v);
+                __h->__next_ = nullptr;
+                __ptr->__next_ = __h.release();
+            }
+        }
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::splice_after(const_iterator __p,
+                                        forward_list& __x)
+{
+    if (!__x.empty())
+    {
+        if (__p.__get_begin()->__next_ != nullptr)
+        {
+            const_iterator __lm1 = __x.before_begin();
+            while (__lm1.__get_begin()->__next_ != nullptr)
+                ++__lm1;
+            __lm1.__get_begin()->__next_ = __p.__get_begin()->__next_;
+        }
+        __p.__get_begin()->__next_ = __x.__before_begin()->__next_;
+        __x.__before_begin()->__next_ = nullptr;
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::splice_after(const_iterator __p,
+                                        forward_list& /*__other*/,
+                                        const_iterator __i)
+{
+    const_iterator __lm1 = _CUDA_VSTD::next(__i);
+    if (__p != __i && __p != __lm1)
+    {
+        __i.__get_begin()->__next_ = __lm1.__get_begin()->__next_;
+        __lm1.__get_begin()->__next_ = __p.__get_begin()->__next_;
+        __p.__get_begin()->__next_ = __lm1.__get_unsafe_node_pointer();
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::splice_after(const_iterator __p,
+                                        forward_list& /*__other*/,
+                                        const_iterator __f, const_iterator __l)
+{
+    if (__f != __l && __p != __f)
+    {
+        const_iterator __lm1 = __f;
+        while (__lm1.__get_begin()->__next_ != __l.__get_begin())
+            ++__lm1;
+        if (__f != __lm1)
+        {
+            __lm1.__get_begin()->__next_ = __p.__get_begin()->__next_;
+            __p.__get_begin()->__next_ = __f.__get_begin()->__next_;
+            __f.__get_begin()->__next_ = __l.__get_unsafe_node_pointer();
+        }
+    }
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+forward_list<_Tp, _Alloc>::splice_after(const_iterator __p,
+                                        forward_list&& __x)
+{
+    splice_after(__p, __x);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+forward_list<_Tp, _Alloc>::splice_after(const_iterator __p,
+                                        forward_list&& __x,
+                                        const_iterator __i)
+{
+    splice_after(__p, __x, __i);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+forward_list<_Tp, _Alloc>::splice_after(const_iterator __p,
+                                        forward_list&& __x,
+                                        const_iterator __f, const_iterator __l)
+{
+    splice_after(__p, __x, __f, __l);
+}
+
+template <class _Tp, class _Alloc>
+typename forward_list<_Tp, _Alloc>::__remove_return_type
+forward_list<_Tp, _Alloc>::remove(const value_type& __v)
+{
+    forward_list<_Tp, _Alloc> __deleted_nodes(get_allocator()); // collect the nodes we're removing
+    typename forward_list<_Tp, _Alloc>::size_type __count_removed = 0;
+    const iterator __e = end();
+    for (iterator __i = before_begin(); __i.__get_begin()->__next_ != nullptr;)
+    {
+        if (__i.__get_begin()->__next_->__value_ == __v)
+        {
+            ++__count_removed;
+            iterator __j = _CUDA_VSTD::next(__i, 2);
+            for (; __j != __e && *__j == __v; ++__j)
+                ++__count_removed;
+            __deleted_nodes.splice_after(__deleted_nodes.before_begin(), *this, __i, __j);
+            if (__j == __e)
+                break;
+            __i = __j;
+        }
+        else
+            ++__i;
+    }
+    
+    return (__remove_return_type) __count_removed;
+}
+
+template <class _Tp, class _Alloc>
+template <class _Predicate>
+typename forward_list<_Tp, _Alloc>::__remove_return_type
+forward_list<_Tp, _Alloc>::remove_if(_Predicate __pred)
+{
+    forward_list<_Tp, _Alloc> __deleted_nodes(get_allocator()); // collect the nodes we're removing
+    typename forward_list<_Tp, _Alloc>::size_type __count_removed = 0;
+    const iterator __e = end();
+    for (iterator __i = before_begin(); __i.__get_begin()->__next_ != nullptr;)
+    {
+        if (__pred(__i.__get_begin()->__next_->__value_))
+        {
+            ++__count_removed;
+            iterator __j = _CUDA_VSTD::next(__i, 2);
+            for (; __j != __e && __pred(*__j); ++__j)
+                ++__count_removed;
+            __deleted_nodes.splice_after(__deleted_nodes.before_begin(), *this, __i, __j);
+            if (__j == __e)
+                break;
+            __i = __j;
+        }
+        else
+            ++__i;
+    }
+    
+    return (__remove_return_type) __count_removed;
+}
+
+template <class _Tp, class _Alloc>
+template <class _BinaryPredicate>
+typename forward_list<_Tp, _Alloc>::__remove_return_type
+forward_list<_Tp, _Alloc>::unique(_BinaryPredicate __binary_pred)
+{
+    forward_list<_Tp, _Alloc> __deleted_nodes(get_allocator()); // collect the nodes we're removing
+    typename forward_list<_Tp, _Alloc>::size_type __count_removed = 0;
+    for (iterator __i = begin(), __e = end(); __i != __e;)
+    {
+        iterator __j = _CUDA_VSTD::next(__i);
+        for (; __j != __e && __binary_pred(*__i, *__j); ++__j)
+            ++__count_removed;
+        if (__i.__get_begin()->__next_ != __j.__get_unsafe_node_pointer())
+            __deleted_nodes.splice_after(__deleted_nodes.before_begin(), *this, __i, __j);
+        __i = __j;
+    }
+    
+    return (__remove_return_type) __count_removed;
+}
+
+template <class _Tp, class _Alloc>
+template <class _Compare>
+void
+forward_list<_Tp, _Alloc>::merge(forward_list& __x, _Compare __comp)
+{
+    if (this != &__x)
+    {
+        base::__before_begin()->__next_ = __merge(base::__before_begin()->__next_,
+                                                    __x.__before_begin()->__next_,
+                                                    __comp);
+        __x.__before_begin()->__next_ = nullptr;
+    }
+}
+
+template <class _Tp, class _Alloc>
+template <class _Compare>
+typename forward_list<_Tp, _Alloc>::__node_pointer
+forward_list<_Tp, _Alloc>::__merge(__node_pointer __f1, __node_pointer __f2,
+                                   _Compare& __comp)
+{
+    if (__f1 == nullptr)
+        return __f2;
+    if (__f2 == nullptr)
+        return __f1;
+    __node_pointer __r;
+    if (__comp(__f2->__value_, __f1->__value_))
+    {
+        __node_pointer __t = __f2;
+        while (__t->__next_ != nullptr &&
+                             __comp(__t->__next_->__value_, __f1->__value_))
+            __t = __t->__next_;
+        __r = __f2;
+        __f2 = __t->__next_;
+        __t->__next_ = __f1;
+    }
+    else
+        __r = __f1;
+    __node_pointer __p = __f1;
+    __f1 = __f1->__next_;
+    while (__f1 != nullptr && __f2 != nullptr)
+    {
+        if (__comp(__f2->__value_, __f1->__value_))
+        {
+            __node_pointer __t = __f2;
+            while (__t->__next_ != nullptr &&
+                                 __comp(__t->__next_->__value_, __f1->__value_))
+                __t = __t->__next_;
+            __p->__next_ = __f2;
+            __f2 = __t->__next_;
+            __t->__next_ = __f1;
+        }
+        __p = __f1;
+        __f1 = __f1->__next_;
+    }
+    if (__f2 != nullptr)
+        __p->__next_ = __f2;
+    return __r;
+}
+
+template <class _Tp, class _Alloc>
+template <class _Compare>
+inline
+void
+forward_list<_Tp, _Alloc>::sort(_Compare __comp)
+{
+    base::__before_begin()->__next_ = __sort(base::__before_begin()->__next_,
+                                       _CUDA_VSTD::distance(begin(), end()), __comp);
+}
+
+template <class _Tp, class _Alloc>
+template <class _Compare>
+typename forward_list<_Tp, _Alloc>::__node_pointer
+forward_list<_Tp, _Alloc>::__sort(__node_pointer __f1, difference_type __sz,
+                                  _Compare& __comp)
+{
+    switch (__sz)
+    {
+    case 0:
+    case 1:
+        return __f1;
+    case 2:
+        if (__comp(__f1->__next_->__value_, __f1->__value_))
+        {
+            __node_pointer __t = __f1->__next_;
+            __t->__next_ = __f1;
+            __f1->__next_ = nullptr;
+            __f1 = __t;
+        }
+        return __f1;
+    }
+    difference_type __sz1 = __sz / 2;
+    difference_type __sz2 = __sz - __sz1;
+    __node_pointer __t = _CUDA_VSTD::next(iterator(__f1), __sz1 - 1).__get_unsafe_node_pointer();
+    __node_pointer __f2 = __t->__next_;
+    __t->__next_ = nullptr;
+    return __merge(__sort(__f1, __sz1, __comp),
+                   __sort(__f2, __sz2, __comp), __comp);
+}
+
+template <class _Tp, class _Alloc>
+void
+forward_list<_Tp, _Alloc>::reverse() _NOEXCEPT
+{
+    __node_pointer __p = base::__before_begin()->__next_;
+    if (__p != nullptr)
+    {
+        __node_pointer __f = __p->__next_;
+        __p->__next_ = nullptr;
+        while (__f != nullptr)
+        {
+            __node_pointer __t = __f->__next_;
+            __f->__next_ = __p;
+            __p = __f;
+            __f = __t;
+        }
+        base::__before_begin()->__next_ = __p;
+    }
+}
+
+template <class _Tp, class _Alloc>
+bool operator==(const forward_list<_Tp, _Alloc>& __x,
+                const forward_list<_Tp, _Alloc>& __y)
+{
+    typedef forward_list<_Tp, _Alloc> _Cp;
+    typedef typename _Cp::const_iterator _Ip;
+    _Ip __ix = __x.begin();
+    _Ip __ex = __x.end();
+    _Ip __iy = __y.begin();
+    _Ip __ey = __y.end();
+    for (; __ix != __ex && __iy != __ey; ++__ix, ++__iy)
+        if (!(*__ix == *__iy))
+            return false;
+    return (__ix == __ex) == (__iy == __ey);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(const forward_list<_Tp, _Alloc>& __x,
+                const forward_list<_Tp, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator< (const forward_list<_Tp, _Alloc>& __x,
+                const forward_list<_Tp, _Alloc>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(),
+                                         __y.begin(), __y.end());
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator> (const forward_list<_Tp, _Alloc>& __x,
+                const forward_list<_Tp, _Alloc>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>=(const forward_list<_Tp, _Alloc>& __x,
+                const forward_list<_Tp, _Alloc>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<=(const forward_list<_Tp, _Alloc>& __x,
+                const forward_list<_Tp, _Alloc>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(forward_list<_Tp, _Alloc>& __x, forward_list<_Tp, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(forward_list<_Tp, _Allocator>& __c, _Predicate __pred)
+{ __c.remove_if(__pred); }
+
+template <class _Tp, class _Allocator, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase(forward_list<_Tp, _Allocator>& __c, const _Up& __v)
+{ _CUDA_VSTD::erase_if(__c, [&](auto& __elem) { return __elem == __v; }); }
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_FORWARD_LIST
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fstream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fstream
new file mode 100644
index 000000000000..3d2259bff08e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/fstream
@@ -0,0 +1,1763 @@
+// -*- C++ -*-
+//===------------------------- fstream ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FSTREAM
+#define _LIBCUDACXX_FSTREAM
+
+/*
+    fstream synopsis
+
+template <class charT, class traits = char_traits<charT> >
+class basic_filebuf
+    : public basic_streambuf<charT, traits>
+{
+public:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // 27.9.1.2 Constructors/destructor:
+    basic_filebuf();
+    basic_filebuf(basic_filebuf&& rhs);
+    virtual ~basic_filebuf();
+
+    // 27.9.1.3 Assign/swap:
+    basic_filebuf& operator=(basic_filebuf&& rhs);
+    void swap(basic_filebuf& rhs);
+
+    // 27.9.1.4 Members:
+    bool is_open() const;
+    basic_filebuf* open(const char* s, ios_base::openmode mode);
+    basic_filebuf* open(const string& s, ios_base::openmode mode);
+    basic_filebuf* open(const filesystem::path& p, ios_base::openmode mode); // C++17
+    basic_filebuf* close();
+
+protected:
+    // 27.9.1.5 Overridden virtual functions:
+    virtual streamsize showmanyc();
+    virtual int_type underflow();
+    virtual int_type uflow();
+    virtual int_type pbackfail(int_type c = traits_type::eof());
+    virtual int_type overflow (int_type c = traits_type::eof());
+    virtual basic_streambuf<char_type, traits_type>* setbuf(char_type* s, streamsize n);
+    virtual pos_type seekoff(off_type off, ios_base::seekdir way,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type sp,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual int sync();
+    virtual void imbue(const locale& loc);
+};
+
+template <class charT, class traits>
+  void
+  swap(basic_filebuf<charT, traits>& x, basic_filebuf<charT, traits>& y);
+
+typedef basic_filebuf<char>    filebuf;
+typedef basic_filebuf<wchar_t> wfilebuf;
+
+template <class charT, class traits = char_traits<charT> >
+class basic_ifstream
+    : public basic_istream<charT,traits>
+{
+public:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    basic_ifstream();
+    explicit basic_ifstream(const char* s, ios_base::openmode mode = ios_base::in);
+    explicit basic_ifstream(const string& s, ios_base::openmode mode = ios_base::in);
+    explicit basic_ifstream(const filesystem::path& p,
+                            ios_base::openmode mode = ios_base::in); // C++17
+    basic_ifstream(basic_ifstream&& rhs);
+
+    basic_ifstream& operator=(basic_ifstream&& rhs);
+    void swap(basic_ifstream& rhs);
+
+    basic_filebuf<char_type, traits_type>* rdbuf() const;
+    bool is_open() const;
+    void open(const char* s, ios_base::openmode mode = ios_base::in);
+    void open(const string& s, ios_base::openmode mode = ios_base::in);
+    void open(const filesystem::path& s, ios_base::openmode mode = ios_base::in); // C++17
+
+    void close();
+};
+
+template <class charT, class traits>
+  void
+  swap(basic_ifstream<charT, traits>& x, basic_ifstream<charT, traits>& y);
+
+typedef basic_ifstream<char>    ifstream;
+typedef basic_ifstream<wchar_t> wifstream;
+
+template <class charT, class traits = char_traits<charT> >
+class basic_ofstream
+    : public basic_ostream<charT,traits>
+{
+public:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    basic_ofstream();
+    explicit basic_ofstream(const char* s, ios_base::openmode mode = ios_base::out);
+    explicit basic_ofstream(const string& s, ios_base::openmode mode = ios_base::out);
+    explicit basic_ofstream(const filesystem::path& p,
+                            ios_base::openmode mode = ios_base::out); // C++17
+    basic_ofstream(basic_ofstream&& rhs);
+
+    basic_ofstream& operator=(basic_ofstream&& rhs);
+    void swap(basic_ofstream& rhs);
+
+    basic_filebuf<char_type, traits_type>* rdbuf() const;
+    bool is_open() const;
+    void open(const char* s, ios_base::openmode mode = ios_base::out);
+    void open(const string& s, ios_base::openmode mode = ios_base::out);
+    void open(const filesystem::path& p,
+              ios_base::openmode mode = ios_base::out); // C++17
+
+    void close();
+};
+
+template <class charT, class traits>
+  void
+  swap(basic_ofstream<charT, traits>& x, basic_ofstream<charT, traits>& y);
+
+typedef basic_ofstream<char>    ofstream;
+typedef basic_ofstream<wchar_t> wofstream;
+
+template <class charT, class traits=char_traits<charT> >
+class basic_fstream
+    : public basic_iostream<charT,traits>
+{
+public:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    basic_fstream();
+    explicit basic_fstream(const char* s, ios_base::openmode mode = ios_base::in|ios_base::out);
+    explicit basic_fstream(const string& s, ios_base::openmode mode = ios_base::in|ios_base::out);
+    explicit basic_fstream(const filesystem::path& p,
+                           ios_base::openmode mode = ios_base::in|ios_base::out); C++17
+    basic_fstream(basic_fstream&& rhs);
+
+    basic_fstream& operator=(basic_fstream&& rhs);
+    void swap(basic_fstream& rhs);
+
+    basic_filebuf<char_type, traits_type>* rdbuf() const;
+    bool is_open() const;
+    void open(const char* s, ios_base::openmode mode = ios_base::in|ios_base::out);
+    void open(const string& s, ios_base::openmode mode = ios_base::in|ios_base::out);
+    void open(const filesystem::path& s,
+              ios_base::openmode mode = ios_base::in|ios_base::out); // C++17
+
+    void close();
+};
+
+template <class charT, class traits>
+  void swap(basic_fstream<charT, traits>& x, basic_fstream<charT, traits>& y);
+
+typedef basic_fstream<char>    fstream;
+typedef basic_fstream<wchar_t> wfstream;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <ostream>
+#include <istream>
+#include <__locale>
+#include <cstdio>
+#include <cstdlib>
+#include <filesystem>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_filebuf
+    : public basic_streambuf<_CharT, _Traits>
+{
+public:
+    typedef _CharT                           char_type;
+    typedef _Traits                          traits_type;
+    typedef typename traits_type::int_type   int_type;
+    typedef typename traits_type::pos_type   pos_type;
+    typedef typename traits_type::off_type   off_type;
+    typedef typename traits_type::state_type state_type;
+
+    // 27.9.1.2 Constructors/destructor:
+    basic_filebuf();
+#ifndef _LIBCUDACXX_CXX03_LANG
+    basic_filebuf(basic_filebuf&& __rhs);
+#endif
+    virtual ~basic_filebuf();
+
+    // 27.9.1.3 Assign/swap:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf& operator=(basic_filebuf&& __rhs);
+#endif
+    void swap(basic_filebuf& __rhs);
+
+    // 27.9.1.4 Members:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_open() const;
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+    basic_filebuf* open(const char* __s, ios_base::openmode __mode);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    basic_filebuf* open(const wchar_t* __s, ios_base::openmode __mode);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf* open(const string& __s, ios_base::openmode __mode);
+
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf* open(const _CUDA_VSTD_FS::path& __p, ios_base::openmode __mode) {
+      return open(__p.c_str(), __mode);
+    }
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf* __open(int __fd, ios_base::openmode __mode);
+#endif
+    basic_filebuf* close();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline static const char*
+    __make_mdstring(ios_base::openmode __mode) _NOEXCEPT;
+
+  protected:
+    // 27.9.1.5 Overridden virtual functions:
+    virtual int_type underflow();
+    virtual int_type pbackfail(int_type __c = traits_type::eof());
+    virtual int_type overflow (int_type __c = traits_type::eof());
+    virtual basic_streambuf<char_type, traits_type>* setbuf(char_type* __s, streamsize __n);
+    virtual pos_type seekoff(off_type __off, ios_base::seekdir __way,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type __sp,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+    virtual int sync();
+    virtual void imbue(const locale& __loc);
+
+private:
+  char* __extbuf_;
+  const char* __extbufnext_;
+  const char* __extbufend_;
+  char __extbuf_min_[8];
+  size_t __ebs_;
+  char_type* __intbuf_;
+  size_t __ibs_;
+  FILE* __file_;
+  const codecvt<char_type, char, state_type>* __cv_;
+  state_type __st_;
+  state_type __st_last_;
+  ios_base::openmode __om_;
+  ios_base::openmode __cm_;
+  bool __owns_eb_;
+  bool __owns_ib_;
+  bool __always_noconv_;
+
+  bool __read_mode();
+  void __write_mode();
+};
+
+template <class _CharT, class _Traits>
+basic_filebuf<_CharT, _Traits>::basic_filebuf()
+    : __extbuf_(0),
+      __extbufnext_(0),
+      __extbufend_(0),
+      __ebs_(0),
+      __intbuf_(0),
+      __ibs_(0),
+      __file_(0),
+      __cv_(nullptr),
+      __st_(),
+      __st_last_(),
+      __om_(0),
+      __cm_(0),
+      __owns_eb_(false),
+      __owns_ib_(false),
+      __always_noconv_(false)
+{
+    if (has_facet<codecvt<char_type, char, state_type> >(this->getloc()))
+    {
+        __cv_ = &use_facet<codecvt<char_type, char, state_type> >(this->getloc());
+        __always_noconv_ = __cv_->always_noconv();
+    }
+    setbuf(0, 4096);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_filebuf<_CharT, _Traits>::basic_filebuf(basic_filebuf&& __rhs)
+    : basic_streambuf<_CharT, _Traits>(__rhs)
+{
+    if (__rhs.__extbuf_ == __rhs.__extbuf_min_)
+    {
+        __extbuf_ = __extbuf_min_;
+        __extbufnext_ = __extbuf_ + (__rhs.__extbufnext_ - __rhs.__extbuf_);
+        __extbufend_ = __extbuf_ + (__rhs.__extbufend_ - __rhs.__extbuf_);
+    }
+    else
+    {
+        __extbuf_ = __rhs.__extbuf_;
+        __extbufnext_ = __rhs.__extbufnext_;
+        __extbufend_ = __rhs.__extbufend_;
+    }
+    __ebs_ = __rhs.__ebs_;
+    __intbuf_ = __rhs.__intbuf_;
+    __ibs_ = __rhs.__ibs_;
+    __file_ = __rhs.__file_;
+    __cv_ = __rhs.__cv_;
+    __st_ = __rhs.__st_;
+    __st_last_ = __rhs.__st_last_;
+    __om_ = __rhs.__om_;
+    __cm_ = __rhs.__cm_;
+    __owns_eb_ = __rhs.__owns_eb_;
+    __owns_ib_ = __rhs.__owns_ib_;
+    __always_noconv_ = __rhs.__always_noconv_;
+    if (__rhs.pbase())
+    {
+        if (__rhs.pbase() == __rhs.__intbuf_)
+            this->setp(__intbuf_, __intbuf_ + (__rhs. epptr() - __rhs.pbase()));
+        else
+            this->setp((char_type*)__extbuf_,
+                       (char_type*)__extbuf_ + (__rhs. epptr() - __rhs.pbase()));
+        this->__pbump(__rhs. pptr() - __rhs.pbase());
+    }
+    else if (__rhs.eback())
+    {
+        if (__rhs.eback() == __rhs.__intbuf_)
+            this->setg(__intbuf_, __intbuf_ + (__rhs.gptr() - __rhs.eback()),
+                                  __intbuf_ + (__rhs.egptr() - __rhs.eback()));
+        else
+            this->setg((char_type*)__extbuf_,
+                       (char_type*)__extbuf_ + (__rhs.gptr() - __rhs.eback()),
+                       (char_type*)__extbuf_ + (__rhs.egptr() - __rhs.eback()));
+    }
+    __rhs.__extbuf_ = 0;
+    __rhs.__extbufnext_ = 0;
+    __rhs.__extbufend_ = 0;
+    __rhs.__ebs_ = 0;
+    __rhs.__intbuf_ = 0;
+    __rhs.__ibs_ = 0;
+    __rhs.__file_ = 0;
+    __rhs.__st_ = state_type();
+    __rhs.__st_last_ = state_type();
+    __rhs.__om_ = 0;
+    __rhs.__cm_ = 0;
+    __rhs.__owns_eb_ = false;
+    __rhs.__owns_ib_ = false;
+    __rhs.setg(0, 0, 0);
+    __rhs.setp(0, 0);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_filebuf<_CharT, _Traits>&
+basic_filebuf<_CharT, _Traits>::operator=(basic_filebuf&& __rhs)
+{
+    close();
+    swap(__rhs);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_filebuf<_CharT, _Traits>::~basic_filebuf()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        close();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    if (__owns_eb_)
+        delete [] __extbuf_;
+    if (__owns_ib_)
+        delete [] __intbuf_;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_filebuf<_CharT, _Traits>::swap(basic_filebuf& __rhs)
+{
+    basic_streambuf<char_type, traits_type>::swap(__rhs);
+    if (__extbuf_ != __extbuf_min_ && __rhs.__extbuf_ != __rhs.__extbuf_min_)
+    {
+        _CUDA_VSTD::swap(__extbuf_, __rhs.__extbuf_);
+        _CUDA_VSTD::swap(__extbufnext_, __rhs.__extbufnext_);
+        _CUDA_VSTD::swap(__extbufend_, __rhs.__extbufend_);
+    }
+    else
+    {
+        ptrdiff_t __ln = __extbufnext_ - __extbuf_;
+        ptrdiff_t __le = __extbufend_ - __extbuf_;
+        ptrdiff_t __rn = __rhs.__extbufnext_ - __rhs.__extbuf_;
+        ptrdiff_t __re = __rhs.__extbufend_ - __rhs.__extbuf_;
+        if (__extbuf_ == __extbuf_min_ && __rhs.__extbuf_ != __rhs.__extbuf_min_)
+        {
+            __extbuf_ = __rhs.__extbuf_;
+            __rhs.__extbuf_ = __rhs.__extbuf_min_;
+        }
+        else if (__extbuf_ != __extbuf_min_ && __rhs.__extbuf_ == __rhs.__extbuf_min_)
+        {
+            __rhs.__extbuf_ = __extbuf_;
+            __extbuf_ = __extbuf_min_;
+        }
+        __extbufnext_ = __extbuf_ + __rn;
+        __extbufend_ = __extbuf_ + __re;
+        __rhs.__extbufnext_ = __rhs.__extbuf_ + __ln;
+        __rhs.__extbufend_ = __rhs.__extbuf_ + __le;
+    }
+    _CUDA_VSTD::swap(__ebs_, __rhs.__ebs_);
+    _CUDA_VSTD::swap(__intbuf_, __rhs.__intbuf_);
+    _CUDA_VSTD::swap(__ibs_, __rhs.__ibs_);
+    _CUDA_VSTD::swap(__file_, __rhs.__file_);
+    _CUDA_VSTD::swap(__cv_, __rhs.__cv_);
+    _CUDA_VSTD::swap(__st_, __rhs.__st_);
+    _CUDA_VSTD::swap(__st_last_, __rhs.__st_last_);
+    _CUDA_VSTD::swap(__om_, __rhs.__om_);
+    _CUDA_VSTD::swap(__cm_, __rhs.__cm_);
+    _CUDA_VSTD::swap(__owns_eb_, __rhs.__owns_eb_);
+    _CUDA_VSTD::swap(__owns_ib_, __rhs.__owns_ib_);
+    _CUDA_VSTD::swap(__always_noconv_, __rhs.__always_noconv_);
+    if (this->eback() == (char_type*)__rhs.__extbuf_min_)
+    {
+        ptrdiff_t __n = this->gptr() - this->eback();
+        ptrdiff_t __e = this->egptr() - this->eback();
+        this->setg((char_type*)__extbuf_min_,
+                   (char_type*)__extbuf_min_ + __n,
+                   (char_type*)__extbuf_min_ + __e);
+    }
+    else if (this->pbase() == (char_type*)__rhs.__extbuf_min_)
+    {
+        ptrdiff_t __n = this->pptr() - this->pbase();
+        ptrdiff_t __e = this->epptr() - this->pbase();
+        this->setp((char_type*)__extbuf_min_,
+                   (char_type*)__extbuf_min_ + __e);
+        this->__pbump(__n);
+    }
+    if (__rhs.eback() == (char_type*)__extbuf_min_)
+    {
+        ptrdiff_t __n = __rhs.gptr() - __rhs.eback();
+        ptrdiff_t __e = __rhs.egptr() - __rhs.eback();
+        __rhs.setg((char_type*)__rhs.__extbuf_min_,
+                   (char_type*)__rhs.__extbuf_min_ + __n,
+                   (char_type*)__rhs.__extbuf_min_ + __e);
+    }
+    else if (__rhs.pbase() == (char_type*)__extbuf_min_)
+    {
+        ptrdiff_t __n = __rhs.pptr() - __rhs.pbase();
+        ptrdiff_t __e = __rhs.epptr() - __rhs.pbase();
+        __rhs.setp((char_type*)__rhs.__extbuf_min_,
+                   (char_type*)__rhs.__extbuf_min_ + __e);
+        __rhs.__pbump(__n);
+    }
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_filebuf<_CharT, _Traits>& __x, basic_filebuf<_CharT, _Traits>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits>
+inline
+bool
+basic_filebuf<_CharT, _Traits>::is_open() const
+{
+    return __file_ != 0;
+}
+
+template <class _CharT, class _Traits>
+const char* basic_filebuf<_CharT, _Traits>::__make_mdstring(
+    ios_base::openmode __mode) _NOEXCEPT {
+  switch (__mode & ~ios_base::ate) {
+  case ios_base::out:
+  case ios_base::out | ios_base::trunc:
+    return "w" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::out | ios_base::app:
+  case ios_base::app:
+    return "a" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in:
+    return "r" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::out:
+    return "r+" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::out | ios_base::trunc:
+    return "w+" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::out | ios_base::app:
+  case ios_base::in | ios_base::app:
+    return "a+" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::out | ios_base::binary:
+  case ios_base::out | ios_base::trunc | ios_base::binary:
+    return "wb" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::out | ios_base::app | ios_base::binary:
+  case ios_base::app | ios_base::binary:
+    return "ab" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::binary:
+    return "rb" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::out | ios_base::binary:
+    return "r+b" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::out | ios_base::trunc | ios_base::binary:
+    return "w+b" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  case ios_base::in | ios_base::out | ios_base::app | ios_base::binary:
+  case ios_base::in | ios_base::app | ios_base::binary:
+    return "a+b" _LIBCUDACXX_FOPEN_CLOEXEC_MODE;
+  default:
+    return nullptr;
+  }
+  _LIBCUDACXX_UNREACHABLE_AFTER_SWITCH();
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+basic_filebuf<_CharT, _Traits>*
+basic_filebuf<_CharT, _Traits>::open(const char* __s, ios_base::openmode __mode)
+{
+    basic_filebuf<_CharT, _Traits>* __rt = 0;
+    if (__file_ == 0)
+    {
+      if (const char* __mdstr = __make_mdstring(__mode)) {
+        __rt = this;
+        __file_ = fopen(__s, __mdstr);
+        if (__file_) {
+          __om_ = __mode;
+          if (__mode & ios_base::ate) {
+            if (fseek(__file_, 0, SEEK_END)) {
+              fclose(__file_);
+              __file_ = 0;
+              __rt = 0;
+            }
+          }
+        } else
+          __rt = 0;
+      }
+    }
+    return __rt;
+}
+
+template <class _CharT, class _Traits>
+_LIBCUDACXX_INLINE_VISIBILITY basic_filebuf<_CharT, _Traits>*
+basic_filebuf<_CharT, _Traits>::__open(int __fd, ios_base::openmode __mode) {
+  basic_filebuf<_CharT, _Traits>* __rt = 0;
+  if (__file_ == 0) {
+    if (const char* __mdstr = __make_mdstring(__mode)) {
+      __rt = this;
+      __file_ = fdopen(__fd, __mdstr);
+      if (__file_) {
+        __om_ = __mode;
+        if (__mode & ios_base::ate) {
+          if (fseek(__file_, 0, SEEK_END)) {
+            fclose(__file_);
+            __file_ = 0;
+            __rt = 0;
+          }
+        }
+      } else
+        __rt = 0;
+    }
+  }
+  return __rt;
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+// This is basically the same as the char* overload except that it uses _wfopen
+// and long mode strings.
+template <class _CharT, class _Traits>
+basic_filebuf<_CharT, _Traits>*
+basic_filebuf<_CharT, _Traits>::open(const wchar_t* __s, ios_base::openmode __mode)
+{
+    basic_filebuf<_CharT, _Traits>* __rt = 0;
+    if (__file_ == 0)
+    {
+        __rt = this;
+        const wchar_t* __mdstr;
+        switch (__mode & ~ios_base::ate)
+        {
+        case ios_base::out:
+        case ios_base::out | ios_base::trunc:
+            __mdstr = L"w";
+            break;
+        case ios_base::out | ios_base::app:
+        case ios_base::app:
+            __mdstr = L"a";
+            break;
+        case ios_base::in:
+            __mdstr = L"r";
+            break;
+        case ios_base::in | ios_base::out:
+            __mdstr = L"r+";
+            break;
+        case ios_base::in | ios_base::out | ios_base::trunc:
+            __mdstr = L"w+";
+            break;
+        case ios_base::in | ios_base::out | ios_base::app:
+        case ios_base::in | ios_base::app:
+            __mdstr = L"a+";
+            break;
+        case ios_base::out | ios_base::binary:
+        case ios_base::out | ios_base::trunc | ios_base::binary:
+            __mdstr = L"wb";
+            break;
+        case ios_base::out | ios_base::app | ios_base::binary:
+        case ios_base::app | ios_base::binary:
+            __mdstr = L"ab";
+            break;
+        case ios_base::in | ios_base::binary:
+            __mdstr = L"rb";
+            break;
+        case ios_base::in | ios_base::out | ios_base::binary:
+            __mdstr = L"r+b";
+            break;
+        case ios_base::in | ios_base::out | ios_base::trunc | ios_base::binary:
+            __mdstr = L"w+b";
+            break;
+        case ios_base::in | ios_base::out | ios_base::app | ios_base::binary:
+        case ios_base::in | ios_base::app | ios_base::binary:
+            __mdstr = L"a+b";
+            break;
+        default:
+            __rt = 0;
+            break;
+        }
+        if (__rt)
+        {
+            __file_ = _wfopen(__s, __mdstr);
+            if (__file_)
+            {
+                __om_ = __mode;
+                if (__mode & ios_base::ate)
+                {
+                    if (fseek(__file_, 0, SEEK_END))
+                    {
+                        fclose(__file_);
+                        __file_ = 0;
+                        __rt = 0;
+                    }
+                }
+            }
+            else
+                __rt = 0;
+        }
+    }
+    return __rt;
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+basic_filebuf<_CharT, _Traits>*
+basic_filebuf<_CharT, _Traits>::open(const string& __s, ios_base::openmode __mode)
+{
+    return open(__s.c_str(), __mode);
+}
+#endif
+
+template <class _CharT, class _Traits>
+basic_filebuf<_CharT, _Traits>*
+basic_filebuf<_CharT, _Traits>::close()
+{
+    basic_filebuf<_CharT, _Traits>* __rt = 0;
+    if (__file_)
+    {
+        __rt = this;
+        unique_ptr<FILE, int(*)(FILE*)> __h(__file_, fclose);
+        if (sync())
+            __rt = 0;
+        if (fclose(__h.release()))
+            __rt = 0;
+        __file_ = 0;
+        setbuf(0, 0);
+    }
+    return __rt;
+}
+
+template <class _CharT, class _Traits>
+typename basic_filebuf<_CharT, _Traits>::int_type
+basic_filebuf<_CharT, _Traits>::underflow()
+{
+    if (__file_ == 0)
+        return traits_type::eof();
+    bool __initial = __read_mode();
+    char_type __1buf;
+    if (this->gptr() == 0)
+        this->setg(&__1buf, &__1buf+1, &__1buf+1);
+    const size_t __unget_sz = __initial ? 0 : min<size_t>((this->egptr() - this->eback()) / 2, 4);
+    int_type __c = traits_type::eof();
+    if (this->gptr() == this->egptr())
+    {
+        memmove(this->eback(), this->egptr() - __unget_sz, __unget_sz * sizeof(char_type));
+        if (__always_noconv_)
+        {
+            size_t __nmemb = static_cast<size_t>(this->egptr() - this->eback() - __unget_sz);
+            __nmemb = fread(this->eback() + __unget_sz, 1, __nmemb, __file_);
+            if (__nmemb != 0)
+            {
+                this->setg(this->eback(),
+                           this->eback() + __unget_sz,
+                           this->eback() + __unget_sz + __nmemb);
+                __c = traits_type::to_int_type(*this->gptr());
+            }
+        }
+        else
+        {
+            _LIBCUDACXX_ASSERT ( !(__extbufnext_ == NULL && (__extbufend_ != __extbufnext_)), "underflow moving from NULL" );
+            if (__extbufend_ != __extbufnext_)
+                memmove(__extbuf_, __extbufnext_, __extbufend_ - __extbufnext_);
+            __extbufnext_ = __extbuf_ + (__extbufend_ - __extbufnext_);
+            __extbufend_ = __extbuf_ + (__extbuf_ == __extbuf_min_ ? sizeof(__extbuf_min_) : __ebs_);
+            size_t __nmemb = _CUDA_VSTD::min(static_cast<size_t>(__ibs_ - __unget_sz),
+                                 static_cast<size_t>(__extbufend_ - __extbufnext_));
+            codecvt_base::result __r;
+            __st_last_ = __st_;
+            size_t __nr = fread((void*) const_cast<char *>(__extbufnext_), 1, __nmemb, __file_);
+            if (__nr != 0)
+            {
+                if (!__cv_)
+                    __throw_bad_cast();
+
+                __extbufend_ = __extbufnext_ + __nr;
+                char_type*  __inext;
+                __r = __cv_->in(__st_, __extbuf_, __extbufend_, __extbufnext_,
+                                       this->eback() + __unget_sz,
+                                       this->eback() + __ibs_, __inext);
+                if (__r == codecvt_base::noconv)
+                {
+                    this->setg((char_type*)__extbuf_, (char_type*)__extbuf_,
+                                          (char_type*)const_cast<char *>(__extbufend_));
+                    __c = traits_type::to_int_type(*this->gptr());
+                }
+                else if (__inext != this->eback() + __unget_sz)
+                {
+                    this->setg(this->eback(), this->eback() + __unget_sz, __inext);
+                    __c = traits_type::to_int_type(*this->gptr());
+                }
+            }
+        }
+    }
+    else
+        __c = traits_type::to_int_type(*this->gptr());
+    if (this->eback() == &__1buf)
+        this->setg(0, 0, 0);
+    return __c;
+}
+
+template <class _CharT, class _Traits>
+typename basic_filebuf<_CharT, _Traits>::int_type
+basic_filebuf<_CharT, _Traits>::pbackfail(int_type __c)
+{
+    if (__file_ && this->eback() < this->gptr())
+    {
+        if (traits_type::eq_int_type(__c, traits_type::eof()))
+        {
+            this->gbump(-1);
+            return traits_type::not_eof(__c);
+        }
+        if ((__om_ & ios_base::out) ||
+            traits_type::eq(traits_type::to_char_type(__c), this->gptr()[-1]))
+        {
+            this->gbump(-1);
+            *this->gptr() = traits_type::to_char_type(__c);
+            return __c;
+        }
+    }
+    return traits_type::eof();
+}
+
+template <class _CharT, class _Traits>
+typename basic_filebuf<_CharT, _Traits>::int_type
+basic_filebuf<_CharT, _Traits>::overflow(int_type __c)
+{
+    if (__file_ == 0)
+        return traits_type::eof();
+    __write_mode();
+    char_type __1buf;
+    char_type* __pb_save = this->pbase();
+    char_type* __epb_save = this->epptr();
+    if (!traits_type::eq_int_type(__c, traits_type::eof()))
+    {
+        if (this->pptr() == 0)
+            this->setp(&__1buf, &__1buf+1);
+        *this->pptr() = traits_type::to_char_type(__c);
+        this->pbump(1);
+    }
+    if (this->pptr() != this->pbase())
+    {
+        if (__always_noconv_)
+        {
+            size_t __nmemb = static_cast<size_t>(this->pptr() - this->pbase());
+            if (fwrite(this->pbase(), sizeof(char_type), __nmemb, __file_) != __nmemb)
+                return traits_type::eof();
+        }
+        else
+        {
+            char* __extbe = __extbuf_;
+            codecvt_base::result __r;
+            do
+            {
+                if (!__cv_)
+                    __throw_bad_cast();
+
+                const char_type* __e;
+                __r = __cv_->out(__st_, this->pbase(), this->pptr(), __e,
+                                        __extbuf_, __extbuf_ + __ebs_, __extbe);
+                if (__e == this->pbase())
+                    return traits_type::eof();
+                if (__r == codecvt_base::noconv)
+                {
+                    size_t __nmemb = static_cast<size_t>(this->pptr() - this->pbase());
+                    if (fwrite(this->pbase(), 1, __nmemb, __file_) != __nmemb)
+                        return traits_type::eof();
+                }
+                else if (__r == codecvt_base::ok || __r == codecvt_base::partial)
+                {
+                    size_t __nmemb = static_cast<size_t>(__extbe - __extbuf_);
+                    if (fwrite(__extbuf_, 1, __nmemb, __file_) != __nmemb)
+                        return traits_type::eof();
+                    if (__r == codecvt_base::partial)
+                    {
+                        this->setp(const_cast<char_type*>(__e), this->pptr());
+                        this->__pbump(this->epptr() - this->pbase());
+                    }
+                }
+                else
+                    return traits_type::eof();
+            } while (__r == codecvt_base::partial);
+        }
+        this->setp(__pb_save, __epb_save);
+    }
+    return traits_type::not_eof(__c);
+}
+
+template <class _CharT, class _Traits>
+basic_streambuf<_CharT, _Traits>*
+basic_filebuf<_CharT, _Traits>::setbuf(char_type* __s, streamsize __n)
+{
+    this->setg(0, 0, 0);
+    this->setp(0, 0);
+    if (__owns_eb_)
+        delete [] __extbuf_;
+    if (__owns_ib_)
+        delete [] __intbuf_;
+    __ebs_ = __n;
+    if (__ebs_ > sizeof(__extbuf_min_))
+    {
+        if (__always_noconv_ && __s)
+        {
+            __extbuf_ = (char*)__s;
+            __owns_eb_ = false;
+        }
+        else
+        {
+            __extbuf_ = new char[__ebs_];
+            __owns_eb_ = true;
+        }
+    }
+    else
+    {
+        __extbuf_ = __extbuf_min_;
+        __ebs_ = sizeof(__extbuf_min_);
+        __owns_eb_ = false;
+    }
+    if (!__always_noconv_)
+    {
+        __ibs_ = max<streamsize>(__n, sizeof(__extbuf_min_));
+        if (__s && __ibs_ >= sizeof(__extbuf_min_))
+        {
+            __intbuf_ = __s;
+            __owns_ib_ = false;
+        }
+        else
+        {
+            __intbuf_ = new char_type[__ibs_];
+            __owns_ib_ = true;
+        }
+    }
+    else
+    {
+        __ibs_ = 0;
+        __intbuf_ = 0;
+        __owns_ib_ = false;
+    }
+    return this;
+}
+
+template <class _CharT, class _Traits>
+typename basic_filebuf<_CharT, _Traits>::pos_type
+basic_filebuf<_CharT, _Traits>::seekoff(off_type __off, ios_base::seekdir __way,
+                                        ios_base::openmode)
+{
+    if (!__cv_)
+        __throw_bad_cast();
+
+    int __width = __cv_->encoding();
+    if (__file_ == 0 || (__width <= 0 && __off != 0) || sync())
+        return pos_type(off_type(-1));
+    // __width > 0 || __off == 0
+    int __whence;
+    switch (__way)
+    {
+    case ios_base::beg:
+        __whence = SEEK_SET;
+        break;
+    case ios_base::cur:
+        __whence = SEEK_CUR;
+        break;
+    case ios_base::end:
+        __whence = SEEK_END;
+        break;
+    default:
+        return pos_type(off_type(-1));
+    }
+#if defined(_LIBCUDACXX_HAS_NO_OFF_T_FUNCTIONS)
+    if (fseek(__file_, __width > 0 ? __width * __off : 0, __whence))
+        return pos_type(off_type(-1));
+    pos_type __r = ftell(__file_);
+#else
+    if (fseeko(__file_, __width > 0 ? __width * __off : 0, __whence))
+        return pos_type(off_type(-1));
+    pos_type __r = ftello(__file_);
+#endif
+    __r.state(__st_);
+    return __r;
+}
+
+template <class _CharT, class _Traits>
+typename basic_filebuf<_CharT, _Traits>::pos_type
+basic_filebuf<_CharT, _Traits>::seekpos(pos_type __sp, ios_base::openmode)
+{
+    if (__file_ == 0 || sync())
+        return pos_type(off_type(-1));
+#if defined(_LIBCUDACXX_HAS_NO_OFF_T_FUNCTIONS)
+    if (fseek(__file_, __sp, SEEK_SET))
+        return pos_type(off_type(-1));
+#else
+    if (fseeko(__file_, __sp, SEEK_SET))
+        return pos_type(off_type(-1));
+#endif
+    __st_ = __sp.state();
+    return __sp;
+}
+
+template <class _CharT, class _Traits>
+int
+basic_filebuf<_CharT, _Traits>::sync()
+{
+    if (__file_ == 0)
+        return 0;
+    if (!__cv_)
+        __throw_bad_cast();
+
+    if (__cm_ & ios_base::out)
+    {
+        if (this->pptr() != this->pbase())
+            if (overflow() == traits_type::eof())
+                return -1;
+        codecvt_base::result __r;
+        do
+        {
+            char* __extbe;
+            __r = __cv_->unshift(__st_, __extbuf_, __extbuf_ + __ebs_, __extbe);
+            size_t __nmemb = static_cast<size_t>(__extbe - __extbuf_);
+            if (fwrite(__extbuf_, 1, __nmemb, __file_) != __nmemb)
+                return -1;
+        } while (__r == codecvt_base::partial);
+        if (__r == codecvt_base::error)
+            return -1;
+        if (fflush(__file_))
+            return -1;
+    }
+    else if (__cm_ & ios_base::in)
+    {
+        off_type __c;
+        state_type __state = __st_last_;
+        bool __update_st = false;
+        if (__always_noconv_)
+            __c = this->egptr() - this->gptr();
+        else
+        {
+            int __width = __cv_->encoding();
+            __c = __extbufend_ - __extbufnext_;
+            if (__width > 0)
+                __c += __width * (this->egptr() - this->gptr());
+            else
+            {
+                if (this->gptr() != this->egptr())
+                {
+                    const int __off =  __cv_->length(__state, __extbuf_,
+                                                     __extbufnext_,
+                                                     this->gptr() - this->eback());
+                    __c += __extbufnext_ - __extbuf_ - __off;
+                    __update_st = true;
+                }
+            }
+        }
+#if defined(_LIBCUDACXX_HAS_NO_OFF_T_FUNCTIONS)
+        if (fseek(__file_, -__c, SEEK_CUR))
+            return -1;
+#else
+        if (fseeko(__file_, -__c, SEEK_CUR))
+            return -1;
+#endif
+        if (__update_st)
+            __st_ = __state;
+        __extbufnext_ = __extbufend_ = __extbuf_;
+        this->setg(0, 0, 0);
+        __cm_ = 0;
+    }
+    return 0;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_filebuf<_CharT, _Traits>::imbue(const locale& __loc)
+{
+    sync();
+    __cv_ = &use_facet<codecvt<char_type, char, state_type> >(__loc);
+    bool __old_anc = __always_noconv_;
+    __always_noconv_ = __cv_->always_noconv();
+    if (__old_anc != __always_noconv_)
+    {
+        this->setg(0, 0, 0);
+        this->setp(0, 0);
+        // invariant, char_type is char, else we couldn't get here
+        if (__always_noconv_)  // need to dump __intbuf_
+        {
+            if (__owns_eb_)
+                delete [] __extbuf_;
+            __owns_eb_ = __owns_ib_;
+            __ebs_ = __ibs_;
+            __extbuf_ = (char*)__intbuf_;
+            __ibs_ = 0;
+            __intbuf_ = 0;
+            __owns_ib_ = false;
+        }
+        else  // need to obtain an __intbuf_.
+        {     // If __extbuf_ is user-supplied, use it, else new __intbuf_
+            if (!__owns_eb_ && __extbuf_ != __extbuf_min_)
+            {
+                __ibs_ = __ebs_;
+                __intbuf_ = (char_type*)__extbuf_;
+                __owns_ib_ = false;
+                __extbuf_ = new char[__ebs_];
+                __owns_eb_ = true;
+            }
+            else
+            {
+                __ibs_ = __ebs_;
+                __intbuf_ = new char_type[__ibs_];
+                __owns_ib_ = true;
+            }
+        }
+    }
+}
+
+template <class _CharT, class _Traits>
+bool
+basic_filebuf<_CharT, _Traits>::__read_mode()
+{
+    if (!(__cm_ & ios_base::in))
+    {
+        this->setp(0, 0);
+        if (__always_noconv_)
+            this->setg((char_type*)__extbuf_,
+                       (char_type*)__extbuf_ + __ebs_,
+                       (char_type*)__extbuf_ + __ebs_);
+        else
+            this->setg(__intbuf_, __intbuf_ + __ibs_, __intbuf_ + __ibs_);
+        __cm_ = ios_base::in;
+        return true;
+    }
+    return false;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_filebuf<_CharT, _Traits>::__write_mode()
+{
+    if (!(__cm_ & ios_base::out))
+    {
+        this->setg(0, 0, 0);
+        if (__ebs_ > sizeof(__extbuf_min_))
+        {
+            if (__always_noconv_)
+                this->setp((char_type*)__extbuf_,
+                           (char_type*)__extbuf_ + (__ebs_ - 1));
+            else
+                this->setp(__intbuf_, __intbuf_ + (__ibs_ - 1));
+        }
+        else
+            this->setp(0, 0);
+        __cm_ = ios_base::out;
+    }
+}
+
+// basic_ifstream
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_ifstream
+    : public basic_istream<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ifstream();
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ifstream(const char* __s, ios_base::openmode __mode = ios_base::in);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ifstream(const wchar_t* __s, ios_base::openmode __mode = ios_base::in);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ifstream(const string& __s, ios_base::openmode __mode = ios_base::in);
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ifstream(const filesystem::path& __p, ios_base::openmode __mode = ios_base::in)
+      : basic_ifstream(__p.c_str(), __mode) {}
+#endif // _LIBCUDACXX_STD_VER >= 17
+#endif
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ifstream(basic_ifstream&& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ifstream& operator=(basic_ifstream&& __rhs);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_ifstream& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf<char_type, traits_type>* rdbuf() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_open() const;
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+    void open(const char* __s, ios_base::openmode __mode = ios_base::in);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    void open(const wchar_t* __s, ios_base::openmode __mode = ios_base::in);
+#endif
+    void open(const string& __s, ios_base::openmode __mode = ios_base::in);
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    void open(const filesystem::path& __p,
+              ios_base::openmode __mode = ios_base::in) {
+      return open(__p.c_str(), __mode);
+    }
+#endif // _LIBCUDACXX_STD_VER >= 17
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __open(int __fd, ios_base::openmode __mode);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void close();
+
+private:
+    basic_filebuf<char_type, traits_type> __sb_;
+};
+
+template <class _CharT, class _Traits>
+inline
+basic_ifstream<_CharT, _Traits>::basic_ifstream()
+    : basic_istream<char_type, traits_type>(&__sb_)
+{
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+inline
+basic_ifstream<_CharT, _Traits>::basic_ifstream(const char* __s, ios_base::openmode __mode)
+    : basic_istream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode | ios_base::in) == 0)
+        this->setstate(ios_base::failbit);
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+template <class _CharT, class _Traits>
+inline
+basic_ifstream<_CharT, _Traits>::basic_ifstream(const wchar_t* __s, ios_base::openmode __mode)
+    : basic_istream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode | ios_base::in) == 0)
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+basic_ifstream<_CharT, _Traits>::basic_ifstream(const string& __s, ios_base::openmode __mode)
+    : basic_istream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode | ios_base::in) == 0)
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+inline
+basic_ifstream<_CharT, _Traits>::basic_ifstream(basic_ifstream&& __rhs)
+    : basic_istream<char_type, traits_type>(_CUDA_VSTD::move(__rhs)),
+      __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+{
+    this->set_rdbuf(&__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_ifstream<_CharT, _Traits>&
+basic_ifstream<_CharT, _Traits>::operator=(basic_ifstream&& __rhs)
+{
+    basic_istream<char_type, traits_type>::operator=(_CUDA_VSTD::move(__rhs));
+    __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+inline
+void
+basic_ifstream<_CharT, _Traits>::swap(basic_ifstream& __rhs)
+{
+    basic_istream<char_type, traits_type>::swap(__rhs);
+    __sb_.swap(__rhs.__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_ifstream<_CharT, _Traits>& __x, basic_ifstream<_CharT, _Traits>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_filebuf<_CharT, _Traits>*
+basic_ifstream<_CharT, _Traits>::rdbuf() const
+{
+    return const_cast<basic_filebuf<char_type, traits_type>*>(&__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline
+bool
+basic_ifstream<_CharT, _Traits>::is_open() const
+{
+    return __sb_.is_open();
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+void
+basic_ifstream<_CharT, _Traits>::open(const char* __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode | ios_base::in))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+template <class _CharT, class _Traits>
+void
+basic_ifstream<_CharT, _Traits>::open(const wchar_t* __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode | ios_base::in))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+void
+basic_ifstream<_CharT, _Traits>::open(const string& __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode | ios_base::in))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+
+template <class _CharT, class _Traits>
+void basic_ifstream<_CharT, _Traits>::__open(int __fd,
+                                             ios_base::openmode __mode) {
+  if (__sb_.__open(__fd, __mode | ios_base::in))
+    this->clear();
+  else
+    this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+void
+basic_ifstream<_CharT, _Traits>::close()
+{
+    if (__sb_.close() == 0)
+        this->setstate(ios_base::failbit);
+}
+
+// basic_ofstream
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_ofstream
+    : public basic_ostream<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ofstream();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ofstream(const char* __s, ios_base::openmode __mode = ios_base::out);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ofstream(const wchar_t* __s, ios_base::openmode __mode = ios_base::out);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ofstream(const string& __s, ios_base::openmode __mode = ios_base::out);
+
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ofstream(const filesystem::path& __p, ios_base::openmode __mode = ios_base::out)
+      : basic_ofstream(__p.c_str(), __mode) {}
+#endif // _LIBCUDACXX_STD_VER >= 17
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ofstream(basic_ofstream&& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ofstream& operator=(basic_ofstream&& __rhs);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_ofstream& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf<char_type, traits_type>* rdbuf() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_open() const;
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+    void open(const char* __s, ios_base::openmode __mode = ios_base::out);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    void open(const wchar_t* __s, ios_base::openmode __mode = ios_base::out);
+#endif
+    void open(const string& __s, ios_base::openmode __mode = ios_base::out);
+
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    void open(const filesystem::path& __p, ios_base::openmode __mode = ios_base::out)
+    { return open(__p.c_str(), __mode); }
+#endif // _LIBCUDACXX_STD_VER >= 17
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __open(int __fd, ios_base::openmode __mode);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void close();
+
+private:
+    basic_filebuf<char_type, traits_type> __sb_;
+};
+
+template <class _CharT, class _Traits>
+inline
+basic_ofstream<_CharT, _Traits>::basic_ofstream()
+    : basic_ostream<char_type, traits_type>(&__sb_)
+{
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+inline
+basic_ofstream<_CharT, _Traits>::basic_ofstream(const char* __s, ios_base::openmode __mode)
+    : basic_ostream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode | ios_base::out) == 0)
+        this->setstate(ios_base::failbit);
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+template <class _CharT, class _Traits>
+inline
+basic_ofstream<_CharT, _Traits>::basic_ofstream(const wchar_t* __s, ios_base::openmode __mode)
+    : basic_ostream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode | ios_base::out) == 0)
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+basic_ofstream<_CharT, _Traits>::basic_ofstream(const string& __s, ios_base::openmode __mode)
+    : basic_ostream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode | ios_base::out) == 0)
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+inline
+basic_ofstream<_CharT, _Traits>::basic_ofstream(basic_ofstream&& __rhs)
+    : basic_ostream<char_type, traits_type>(_CUDA_VSTD::move(__rhs)),
+      __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+{
+    this->set_rdbuf(&__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_ofstream<_CharT, _Traits>&
+basic_ofstream<_CharT, _Traits>::operator=(basic_ofstream&& __rhs)
+{
+    basic_ostream<char_type, traits_type>::operator=(_CUDA_VSTD::move(__rhs));
+    __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+inline
+void
+basic_ofstream<_CharT, _Traits>::swap(basic_ofstream& __rhs)
+{
+    basic_ostream<char_type, traits_type>::swap(__rhs);
+    __sb_.swap(__rhs.__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_ofstream<_CharT, _Traits>& __x, basic_ofstream<_CharT, _Traits>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_filebuf<_CharT, _Traits>*
+basic_ofstream<_CharT, _Traits>::rdbuf() const
+{
+    return const_cast<basic_filebuf<char_type, traits_type>*>(&__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline
+bool
+basic_ofstream<_CharT, _Traits>::is_open() const
+{
+    return __sb_.is_open();
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+void
+basic_ofstream<_CharT, _Traits>::open(const char* __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode | ios_base::out))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+template <class _CharT, class _Traits>
+void
+basic_ofstream<_CharT, _Traits>::open(const wchar_t* __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode | ios_base::out))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+void
+basic_ofstream<_CharT, _Traits>::open(const string& __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode | ios_base::out))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+
+template <class _CharT, class _Traits>
+void basic_ofstream<_CharT, _Traits>::__open(int __fd,
+                                             ios_base::openmode __mode) {
+  if (__sb_.__open(__fd, __mode | ios_base::out))
+    this->clear();
+  else
+    this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+void
+basic_ofstream<_CharT, _Traits>::close()
+{
+    if (__sb_.close() == 0)
+        this->setstate(ios_base::failbit);
+}
+
+// basic_fstream
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_fstream
+    : public basic_iostream<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_fstream();
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_fstream(const char* __s, ios_base::openmode __mode = ios_base::in | ios_base::out);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_fstream(const wchar_t* __s, ios_base::openmode __mode = ios_base::in | ios_base::out);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_fstream(const string& __s, ios_base::openmode __mode = ios_base::in | ios_base::out);
+
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_fstream(const filesystem::path& __p, ios_base::openmode __mode = ios_base::in | ios_base::out)
+      : basic_fstream(__p.c_str(), __mode) {}
+#endif // _LIBCUDACXX_STD_VER >= 17
+
+#endif
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_fstream(basic_fstream&& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_fstream& operator=(basic_fstream&& __rhs);
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_fstream& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_filebuf<char_type, traits_type>* rdbuf() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool is_open() const;
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+    void open(const char* __s, ios_base::openmode __mode = ios_base::in | ios_base::out);
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+    void open(const wchar_t* __s, ios_base::openmode __mode = ios_base::in | ios_base::out);
+#endif
+    void open(const string& __s, ios_base::openmode __mode = ios_base::in | ios_base::out);
+
+#if _LIBCUDACXX_STD_VER >= 17
+    _LIBCUDACXX_AVAILABILITY_FILESYSTEM _LIBCUDACXX_INLINE_VISIBILITY
+    void open(const filesystem::path& __p, ios_base::openmode __mode = ios_base::in|ios_base::out)
+    { return open(__p.c_str(), __mode); }
+#endif // _LIBCUDACXX_STD_VER >= 17
+
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void close();
+
+private:
+    basic_filebuf<char_type, traits_type> __sb_;
+};
+
+template <class _CharT, class _Traits>
+inline
+basic_fstream<_CharT, _Traits>::basic_fstream()
+    : basic_iostream<char_type, traits_type>(&__sb_)
+{
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+inline
+basic_fstream<_CharT, _Traits>::basic_fstream(const char* __s, ios_base::openmode __mode)
+    : basic_iostream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode) == 0)
+        this->setstate(ios_base::failbit);
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+template <class _CharT, class _Traits>
+inline
+basic_fstream<_CharT, _Traits>::basic_fstream(const wchar_t* __s, ios_base::openmode __mode)
+    : basic_iostream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode) == 0)
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+basic_fstream<_CharT, _Traits>::basic_fstream(const string& __s, ios_base::openmode __mode)
+    : basic_iostream<char_type, traits_type>(&__sb_)
+{
+    if (__sb_.open(__s, __mode) == 0)
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+inline
+basic_fstream<_CharT, _Traits>::basic_fstream(basic_fstream&& __rhs)
+    : basic_iostream<char_type, traits_type>(_CUDA_VSTD::move(__rhs)),
+      __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+{
+    this->set_rdbuf(&__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_fstream<_CharT, _Traits>&
+basic_fstream<_CharT, _Traits>::operator=(basic_fstream&& __rhs)
+{
+    basic_iostream<char_type, traits_type>::operator=(_CUDA_VSTD::move(__rhs));
+    __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+inline
+void
+basic_fstream<_CharT, _Traits>::swap(basic_fstream& __rhs)
+{
+    basic_iostream<char_type, traits_type>::swap(__rhs);
+    __sb_.swap(__rhs.__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_fstream<_CharT, _Traits>& __x, basic_fstream<_CharT, _Traits>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits>
+inline
+basic_filebuf<_CharT, _Traits>*
+basic_fstream<_CharT, _Traits>::rdbuf() const
+{
+    return const_cast<basic_filebuf<char_type, traits_type>*>(&__sb_);
+}
+
+template <class _CharT, class _Traits>
+inline
+bool
+basic_fstream<_CharT, _Traits>::is_open() const
+{
+    return __sb_.is_open();
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE
+template <class _CharT, class _Traits>
+void
+basic_fstream<_CharT, _Traits>::open(const char* __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+
+#ifdef _LIBCUDACXX_HAS_OPEN_WITH_WCHAR
+template <class _CharT, class _Traits>
+void
+basic_fstream<_CharT, _Traits>::open(const wchar_t* __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+void
+basic_fstream<_CharT, _Traits>::open(const string& __s, ios_base::openmode __mode)
+{
+    if (__sb_.open(__s, __mode))
+        this->clear();
+    else
+        this->setstate(ios_base::failbit);
+}
+#endif
+
+template <class _CharT, class _Traits>
+inline
+void
+basic_fstream<_CharT, _Traits>::close()
+{
+    if (__sb_.close() == 0)
+        this->setstate(ios_base::failbit);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_FSTREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/functional b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/functional
new file mode 100644
index 000000000000..5d4bac63ee1e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/functional
@@ -0,0 +1,3126 @@
+// -*- C++ -*-
+//===------------------------ functional ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FUNCTIONAL
+#define _LIBCUDACXX_FUNCTIONAL
+
+/*
+    functional synopsis
+
+namespace std
+{
+
+template <class Arg, class Result>
+struct unary_function
+{
+    typedef Arg    argument_type;
+    typedef Result result_type;
+};
+
+template <class Arg1, class Arg2, class Result>
+struct binary_function
+{
+    typedef Arg1   first_argument_type;
+    typedef Arg2   second_argument_type;
+    typedef Result result_type;
+};
+
+template <class T>
+class reference_wrapper
+    : public unary_function<T1, R> // if wrapping a unary functor
+    : public binary_function<T1, T2, R> // if wraping a binary functor
+{
+public:
+    // types
+    typedef T type;
+    typedef see below result_type; // Not always defined
+
+    // construct/copy/destroy
+    reference_wrapper(T&) noexcept;
+    reference_wrapper(T&&) = delete; // do not bind to temps
+    reference_wrapper(const reference_wrapper<T>& x) noexcept;
+
+    // assignment
+    reference_wrapper& operator=(const reference_wrapper<T>& x) noexcept;
+
+    // access
+    operator T& () const noexcept;
+    T& get() const noexcept;
+
+    // invoke
+    template <class... ArgTypes>
+      typename result_of<T&(ArgTypes&&...)>::type
+          operator() (ArgTypes&&...) const;
+};
+
+template <class T> reference_wrapper<T> ref(T& t) noexcept;
+template <class T> void ref(const T&& t) = delete;
+template <class T> reference_wrapper<T> ref(reference_wrapper<T>t) noexcept;
+
+template <class T> reference_wrapper<const T> cref(const T& t) noexcept;
+template <class T> void cref(const T&& t) = delete;
+template <class T> reference_wrapper<const T> cref(reference_wrapper<T> t) noexcept;
+
+template <class T> struct unwrap_reference;                                       // since C++20
+template <class T> struct unwrap_ref_decay : unwrap_reference<decay_t<T>> { };    // since C++20
+template <class T> using unwrap_reference_t = typename unwrap_reference<T>::type; // since C++20
+template <class T> using unwrap_ref_decay_t = typename unwrap_ref_decay<T>::type; // since C++20
+
+template <class T> // <class T=void> in C++14
+struct plus : binary_function<T, T, T>
+{
+    T operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct minus : binary_function<T, T, T>
+{
+    T operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct multiplies : binary_function<T, T, T>
+{
+    T operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct divides : binary_function<T, T, T>
+{
+    T operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct modulus : binary_function<T, T, T>
+{
+    T operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct negate : unary_function<T, T>
+{
+    T operator()(const T& x) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct equal_to : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct not_equal_to : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct greater : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct less : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct greater_equal : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct less_equal : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct logical_and : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct logical_or : binary_function<T, T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct logical_not : unary_function<T, bool>
+{
+    bool operator()(const T& x) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct bit_and : unary_function<T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct bit_or : unary_function<T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T> // <class T=void> in C++14
+struct bit_xor : unary_function<T, bool>
+{
+    bool operator()(const T& x, const T& y) const;
+};
+
+template <class T=void> // C++14
+struct bit_xor : unary_function<T, bool>
+{
+    bool operator()(const T& x) const;
+};
+
+template <class Predicate>
+class unary_negate // deprecated in C++17
+    : public unary_function<typename Predicate::argument_type, bool>
+{
+public:
+    explicit unary_negate(const Predicate& pred);
+    bool operator()(const typename Predicate::argument_type& x) const;
+};
+
+template <class Predicate> // deprecated in C++17
+unary_negate<Predicate> not1(const Predicate& pred);
+
+template <class Predicate>
+class binary_negate // deprecated in C++17
+    : public binary_function<typename Predicate::first_argument_type,
+                             typename Predicate::second_argument_type,
+                             bool>
+{
+public:
+    explicit binary_negate(const Predicate& pred);
+    bool operator()(const typename Predicate::first_argument_type& x,
+                    const typename Predicate::second_argument_type& y) const;
+};
+
+template <class Predicate> // deprecated in C++17
+binary_negate<Predicate> not2(const Predicate& pred);
+
+template <class F> unspecified not_fn(F&& f); // C++17
+
+template<class T> struct is_bind_expression;
+template<class T> struct is_placeholder;
+
+    // See C++14 20.9.9, Function object binders
+template <class T> inline constexpr bool is_bind_expression_v
+  = is_bind_expression<T>::value; // C++17
+template <class T> inline constexpr int is_placeholder_v
+  = is_placeholder<T>::value; // C++17
+
+
+template<class Fn, class... BoundArgs>
+  unspecified bind(Fn&&, BoundArgs&&...);
+template<class R, class Fn, class... BoundArgs>
+  unspecified bind(Fn&&, BoundArgs&&...);
+
+template<class F, class... Args>
+ invoke_result_t<F, Args...> invoke(F&& f, Args&&... args) // C++17
+    noexcept(is_nothrow_invocable_v<F, Args...>);
+
+namespace placeholders {
+  // M is the implementation-defined number of placeholders
+  extern unspecified _1;
+  extern unspecified _2;
+  .
+  .
+  .
+  extern unspecified _Mp;
+}
+
+template <class Operation>
+class binder1st     // deprecated in C++11, removed in C++17
+    : public unary_function<typename Operation::second_argument_type,
+                            typename Operation::result_type>
+{
+protected:
+    Operation                               op;
+    typename Operation::first_argument_type value;
+public:
+    binder1st(const Operation& x, const typename Operation::first_argument_type y);
+    typename Operation::result_type operator()(      typename Operation::second_argument_type& x) const;
+    typename Operation::result_type operator()(const typename Operation::second_argument_type& x) const;
+};
+
+template <class Operation, class T>
+binder1st<Operation> bind1st(const Operation& op, const T& x);  // deprecated in C++11, removed in C++17
+
+template <class Operation>
+class binder2nd     // deprecated in C++11, removed in C++17
+    : public unary_function<typename Operation::first_argument_type,
+                            typename Operation::result_type>
+{
+protected:
+    Operation                                op;
+    typename Operation::second_argument_type value;
+public:
+    binder2nd(const Operation& x, const typename Operation::second_argument_type y);
+    typename Operation::result_type operator()(      typename Operation::first_argument_type& x) const;
+    typename Operation::result_type operator()(const typename Operation::first_argument_type& x) const;
+};
+
+template <class Operation, class T>
+binder2nd<Operation> bind2nd(const Operation& op, const T& x);  // deprecated in C++11, removed in C++17
+
+template <class Arg, class Result>      // deprecated in C++11, removed in C++17
+class pointer_to_unary_function : public unary_function<Arg, Result>
+{
+public:
+    explicit pointer_to_unary_function(Result (*f)(Arg));
+    Result operator()(Arg x) const;
+};
+
+template <class Arg, class Result>
+pointer_to_unary_function<Arg,Result> ptr_fun(Result (*f)(Arg));      // deprecated in C++11, removed in C++17
+
+template <class Arg1, class Arg2, class Result>      // deprecated in C++11, removed in C++17
+class pointer_to_binary_function : public binary_function<Arg1, Arg2, Result>
+{
+public:
+    explicit pointer_to_binary_function(Result (*f)(Arg1, Arg2));
+    Result operator()(Arg1 x, Arg2 y) const;
+};
+
+template <class Arg1, class Arg2, class Result>
+pointer_to_binary_function<Arg1,Arg2,Result> ptr_fun(Result (*f)(Arg1,Arg2));      // deprecated in C++11, removed in C++17
+
+template<class S, class T>      // deprecated in C++11, removed in C++17
+class mem_fun_t : public unary_function<T*, S>
+{
+public:
+    explicit mem_fun_t(S (T::*p)());
+    S operator()(T* p) const;
+};
+
+template<class S, class T, class A>
+class mem_fun1_t : public binary_function<T*, A, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit mem_fun1_t(S (T::*p)(A));
+    S operator()(T* p, A x) const;
+};
+
+template<class S, class T>          mem_fun_t<S,T>    mem_fun(S (T::*f)());      // deprecated in C++11, removed in C++17
+template<class S, class T, class A> mem_fun1_t<S,T,A> mem_fun(S (T::*f)(A));     // deprecated in C++11, removed in C++17
+
+template<class S, class T>
+class mem_fun_ref_t : public unary_function<T, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit mem_fun_ref_t(S (T::*p)());
+    S operator()(T& p) const;
+};
+
+template<class S, class T, class A>
+class mem_fun1_ref_t : public binary_function<T, A, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit mem_fun1_ref_t(S (T::*p)(A));
+    S operator()(T& p, A x) const;
+};
+
+template<class S, class T>          mem_fun_ref_t<S,T>    mem_fun_ref(S (T::*f)());      // deprecated in C++11, removed in C++17
+template<class S, class T, class A> mem_fun1_ref_t<S,T,A> mem_fun_ref(S (T::*f)(A));     // deprecated in C++11, removed in C++17
+
+template <class S, class T>
+class const_mem_fun_t : public unary_function<const T*, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit const_mem_fun_t(S (T::*p)() const);
+    S operator()(const T* p) const;
+};
+
+template <class S, class T, class A>
+class const_mem_fun1_t : public binary_function<const T*, A, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit const_mem_fun1_t(S (T::*p)(A) const);
+    S operator()(const T* p, A x) const;
+};
+
+template <class S, class T>          const_mem_fun_t<S,T>    mem_fun(S (T::*f)() const);      // deprecated in C++11, removed in C++17
+template <class S, class T, class A> const_mem_fun1_t<S,T,A> mem_fun(S (T::*f)(A) const);     // deprecated in C++11, removed in C++17
+
+template <class S, class T>
+class const_mem_fun_ref_t : public unary_function<T, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit const_mem_fun_ref_t(S (T::*p)() const);
+    S operator()(const T& p) const;
+};
+
+template <class S, class T, class A>
+class const_mem_fun1_ref_t : public binary_function<T, A, S>      // deprecated in C++11, removed in C++17
+{
+public:
+    explicit const_mem_fun1_ref_t(S (T::*p)(A) const);
+    S operator()(const T& p, A x) const;
+};
+
+template <class S, class T>          const_mem_fun_ref_t<S,T>    mem_fun_ref(S (T::*f)() const);   // deprecated in C++11, removed in C++17
+template <class S, class T, class A> const_mem_fun1_ref_t<S,T,A> mem_fun_ref(S (T::*f)(A) const);  // deprecated in C++11, removed in C++17
+
+template<class R, class T> unspecified mem_fn(R T::*);
+
+class bad_function_call
+    : public exception
+{
+};
+
+template<class> class function; // undefined
+
+template<class R, class... ArgTypes>
+class function<R(ArgTypes...)>
+  : public unary_function<T1, R>      // iff sizeof...(ArgTypes) == 1 and
+                                      // ArgTypes contains T1
+  : public binary_function<T1, T2, R> // iff sizeof...(ArgTypes) == 2 and
+                                      // ArgTypes contains T1 and T2
+{
+public:
+    typedef R result_type;
+
+    // construct/copy/destroy:
+    function() noexcept;
+    function(nullptr_t) noexcept;
+    function(const function&);
+    function(function&&) noexcept;
+    template<class F>
+      function(F);
+    template<Allocator Alloc>
+      function(allocator_arg_t, const Alloc&) noexcept;            // removed in C++17
+    template<Allocator Alloc>
+      function(allocator_arg_t, const Alloc&, nullptr_t) noexcept; // removed in C++17
+    template<Allocator Alloc>
+      function(allocator_arg_t, const Alloc&, const function&);    // removed in C++17
+    template<Allocator Alloc>
+      function(allocator_arg_t, const Alloc&, function&&);         // removed in C++17
+    template<class F, Allocator Alloc>
+      function(allocator_arg_t, const Alloc&, F);                  // removed in C++17
+
+    function& operator=(const function&);
+    function& operator=(function&&) noexcept;
+    function& operator=(nullptr_t) noexcept;
+    template<class F>
+      function& operator=(F&&);
+    template<class F>
+      function& operator=(reference_wrapper<F>) noexcept;
+
+    ~function();
+
+    // function modifiers:
+    void swap(function&) noexcept;
+    template<class F, class Alloc>
+      void assign(F&&, const Alloc&);                 // Removed in C++17
+
+    // function capacity:
+    explicit operator bool() const noexcept;
+
+    // function invocation:
+    R operator()(ArgTypes...) const;
+
+    // function target access:
+    const std::type_info& target_type() const noexcept;
+    template <typename T>       T* target() noexcept;
+    template <typename T> const T* target() const noexcept;
+};
+
+// Deduction guides
+template<class R, class ...Args>
+function(R(*)(Args...)) -> function<R(Args...)>; // since C++17
+
+template<class F>
+function(F) -> function<see-below>; // since C++17
+
+// Null pointer comparisons:
+template <class R, class ... ArgTypes>
+  bool operator==(const function<R(ArgTypes...)>&, nullptr_t) noexcept;
+
+template <class R, class ... ArgTypes>
+  bool operator==(nullptr_t, const function<R(ArgTypes...)>&) noexcept;
+
+template <class R, class ... ArgTypes>
+  bool operator!=(const function<R(ArgTypes...)>&, nullptr_t) noexcept;
+
+template <class  R, class ... ArgTypes>
+  bool operator!=(nullptr_t, const function<R(ArgTypes...)>&) noexcept;
+
+// specialized algorithms:
+template <class  R, class ... ArgTypes>
+  void swap(function<R(ArgTypes...)>&, function<R(ArgTypes...)>&) noexcept;
+
+template <class T> struct hash;
+
+template <> struct hash<bool>;
+template <> struct hash<char>;
+template <> struct hash<signed char>;
+template <> struct hash<unsigned char>;
+template <> struct hash<char16_t>;
+template <> struct hash<char32_t>;
+template <> struct hash<wchar_t>;
+template <> struct hash<short>;
+template <> struct hash<unsigned short>;
+template <> struct hash<int>;
+template <> struct hash<unsigned int>;
+template <> struct hash<long>;
+template <> struct hash<long long>;
+template <> struct hash<unsigned long>;
+template <> struct hash<unsigned long long>;
+
+template <> struct hash<float>;
+template <> struct hash<double>;
+template <> struct hash<long double>;
+
+template<class T> struct hash<T*>;
+template <> struct hash<nullptr_t>;  // C++17
+
+}  // std
+
+POLICY:  For non-variadic implementations, the number of arguments is limited
+         to 3.  It is hoped that the need for non-variadic implementations
+         will be minimal.
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <type_traits>
+#include <typeinfo>
+#include <exception>
+#include <memory>
+#include <tuple>
+#include <utility>
+#include <version>
+#include <__functional_base>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS plus : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x + __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS plus<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) + _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) + _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) + _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS minus : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x - __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS minus<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) - _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) - _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) - _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS multiplies : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x * __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS multiplies<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) * _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) * _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) * _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS divides : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x / __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS divides<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) / _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) / _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) / _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS modulus : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x % __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS modulus<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) % _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) % _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) % _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS negate : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return -__x;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS negate<void>
+{
+    template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Tp&& __x) const
+    _NOEXCEPT_(noexcept(- _CUDA_VSTD::forward<_Tp>(__x)))
+    -> decltype        (- _CUDA_VSTD::forward<_Tp>(__x))
+        { return        - _CUDA_VSTD::forward<_Tp>(__x); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS equal_to : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x == __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS equal_to<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) == _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) == _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) == _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS not_equal_to : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x != __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS not_equal_to<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) != _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) != _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) != _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS greater : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x > __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS greater<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) > _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) > _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) > _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+// less in <__functional_base>
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS greater_equal : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x >= __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS greater_equal<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) >= _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) >= _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) >= _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS less_equal : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x <= __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS less_equal<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) <= _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) <= _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) <= _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS logical_and : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x && __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS logical_and<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) && _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) && _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) && _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS logical_or : binary_function<_Tp, _Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x || __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS logical_or<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) || _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) || _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) || _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS logical_not : unary_function<_Tp, bool>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Tp& __x) const
+        {return !__x;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS logical_not<void>
+{
+    template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Tp&& __x) const
+    _NOEXCEPT_(noexcept(!_CUDA_VSTD::forward<_Tp>(__x)))
+    -> decltype        (!_CUDA_VSTD::forward<_Tp>(__x))
+        { return        !_CUDA_VSTD::forward<_Tp>(__x); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS bit_and : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x & __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS bit_and<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) & _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) & _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) & _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS bit_or : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x | __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS bit_or<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) | _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) | _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) | _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+#else
+template <class _Tp>
+#endif
+struct _LIBCUDACXX_TEMPLATE_VIS bit_xor : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x ^ __y;}
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS bit_xor<void>
+{
+    template <class _T1, class _T2>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_T1&& __t, _T2&& __u) const
+    _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) ^ _CUDA_VSTD::forward<_T2>(__u)))
+    -> decltype        (_CUDA_VSTD::forward<_T1>(__t) ^ _CUDA_VSTD::forward<_T2>(__u))
+        { return        _CUDA_VSTD::forward<_T1>(__t) ^ _CUDA_VSTD::forward<_T2>(__u); }
+    typedef void is_transparent;
+};
+#endif
+
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp = void>
+struct _LIBCUDACXX_TEMPLATE_VIS bit_not : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return ~__x;}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS bit_not<void>
+{
+    template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Tp&& __x) const
+    _NOEXCEPT_(noexcept(~_CUDA_VSTD::forward<_Tp>(__x)))
+    -> decltype        (~_CUDA_VSTD::forward<_Tp>(__x))
+        { return        ~_CUDA_VSTD::forward<_Tp>(__x); }
+    typedef void is_transparent;
+};
+#endif
+
+template <class _Predicate>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX17 unary_negate
+    : public unary_function<typename _Predicate::argument_type, bool>
+{
+    _Predicate __pred_;
+public:
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    explicit unary_negate(const _Predicate& __pred)
+        : __pred_(__pred) {}
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const typename _Predicate::argument_type& __x) const
+        {return !__pred_(__x);}
+};
+
+template <class _Predicate>
+_LIBCUDACXX_DEPRECATED_IN_CXX17 inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+unary_negate<_Predicate>
+not1(const _Predicate& __pred) {return unary_negate<_Predicate>(__pred);}
+
+template <class _Predicate>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX17 binary_negate
+    : public binary_function<typename _Predicate::first_argument_type,
+                             typename _Predicate::second_argument_type,
+                             bool>
+{
+    _Predicate __pred_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    binary_negate(const _Predicate& __pred) : __pred_(__pred) {}
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const typename _Predicate::first_argument_type& __x,
+                    const typename _Predicate::second_argument_type& __y) const
+        {return !__pred_(__x, __y);}
+};
+
+template <class _Predicate>
+_LIBCUDACXX_DEPRECATED_IN_CXX17 inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+binary_negate<_Predicate>
+not2(const _Predicate& __pred) {return binary_negate<_Predicate>(__pred);}
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_BINDERS)
+template <class __Operation>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 binder1st
+    : public unary_function<typename __Operation::second_argument_type,
+                            typename __Operation::result_type>
+{
+protected:
+    __Operation                               op;
+    typename __Operation::first_argument_type value;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY binder1st(const __Operation& __x,
+                               const typename __Operation::first_argument_type __y)
+        : op(__x), value(__y) {}
+    _LIBCUDACXX_INLINE_VISIBILITY typename __Operation::result_type operator()
+        (typename __Operation::second_argument_type& __x) const
+            {return op(value, __x);}
+    _LIBCUDACXX_INLINE_VISIBILITY typename __Operation::result_type operator()
+        (const typename __Operation::second_argument_type& __x) const
+            {return op(value, __x);}
+};
+
+template <class __Operation, class _Tp>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+binder1st<__Operation>
+bind1st(const __Operation& __op, const _Tp& __x)
+    {return binder1st<__Operation>(__op, __x);}
+
+template <class __Operation>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 binder2nd
+    : public unary_function<typename __Operation::first_argument_type,
+                            typename __Operation::result_type>
+{
+protected:
+    __Operation                                op;
+    typename __Operation::second_argument_type value;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    binder2nd(const __Operation& __x, const typename __Operation::second_argument_type __y)
+        : op(__x), value(__y) {}
+    _LIBCUDACXX_INLINE_VISIBILITY typename __Operation::result_type operator()
+        (      typename __Operation::first_argument_type& __x) const
+            {return op(__x, value);}
+    _LIBCUDACXX_INLINE_VISIBILITY typename __Operation::result_type operator()
+        (const typename __Operation::first_argument_type& __x) const
+            {return op(__x, value);}
+};
+
+template <class __Operation, class _Tp>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+binder2nd<__Operation>
+bind2nd(const __Operation& __op, const _Tp& __x)
+    {return binder2nd<__Operation>(__op, __x);}
+
+template <class _Arg, class _Result>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 pointer_to_unary_function
+    : public unary_function<_Arg, _Result>
+{
+    _Result (*__f_)(_Arg);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit pointer_to_unary_function(_Result (*__f)(_Arg))
+        : __f_(__f) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Result operator()(_Arg __x) const
+        {return __f_(__x);}
+};
+
+template <class _Arg, class _Result>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+pointer_to_unary_function<_Arg,_Result>
+ptr_fun(_Result (*__f)(_Arg))
+    {return pointer_to_unary_function<_Arg,_Result>(__f);}
+
+template <class _Arg1, class _Arg2, class _Result>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 pointer_to_binary_function
+    : public binary_function<_Arg1, _Arg2, _Result>
+{
+    _Result (*__f_)(_Arg1, _Arg2);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit pointer_to_binary_function(_Result (*__f)(_Arg1, _Arg2))
+        : __f_(__f) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Result operator()(_Arg1 __x, _Arg2 __y) const
+        {return __f_(__x, __y);}
+};
+
+template <class _Arg1, class _Arg2, class _Result>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+pointer_to_binary_function<_Arg1,_Arg2,_Result>
+ptr_fun(_Result (*__f)(_Arg1,_Arg2))
+    {return pointer_to_binary_function<_Arg1,_Arg2,_Result>(__f);}
+
+template<class _Sp, class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 mem_fun_t
+    : public unary_function<_Tp*, _Sp>
+{
+    _Sp (_Tp::*__p_)();
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit mem_fun_t(_Sp (_Tp::*__p)())
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(_Tp* __p) const
+        {return (__p->*__p_)();}
+};
+
+template<class _Sp, class _Tp, class _Ap>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 mem_fun1_t
+    : public binary_function<_Tp*, _Ap, _Sp>
+{
+    _Sp (_Tp::*__p_)(_Ap);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit mem_fun1_t(_Sp (_Tp::*__p)(_Ap))
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(_Tp* __p, _Ap __x) const
+        {return (__p->*__p_)(__x);}
+};
+
+template<class _Sp, class _Tp>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+mem_fun_t<_Sp,_Tp>
+mem_fun(_Sp (_Tp::*__f)())
+    {return mem_fun_t<_Sp,_Tp>(__f);}
+
+template<class _Sp, class _Tp, class _Ap>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+mem_fun1_t<_Sp,_Tp,_Ap>
+mem_fun(_Sp (_Tp::*__f)(_Ap))
+    {return mem_fun1_t<_Sp,_Tp,_Ap>(__f);}
+
+template<class _Sp, class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 mem_fun_ref_t
+    : public unary_function<_Tp, _Sp>
+{
+    _Sp (_Tp::*__p_)();
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit mem_fun_ref_t(_Sp (_Tp::*__p)())
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(_Tp& __p) const
+        {return (__p.*__p_)();}
+};
+
+template<class _Sp, class _Tp, class _Ap>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 mem_fun1_ref_t
+    : public binary_function<_Tp, _Ap, _Sp>
+{
+    _Sp (_Tp::*__p_)(_Ap);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit mem_fun1_ref_t(_Sp (_Tp::*__p)(_Ap))
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(_Tp& __p, _Ap __x) const
+        {return (__p.*__p_)(__x);}
+};
+
+template<class _Sp, class _Tp>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+mem_fun_ref_t<_Sp,_Tp>
+mem_fun_ref(_Sp (_Tp::*__f)())
+    {return mem_fun_ref_t<_Sp,_Tp>(__f);}
+
+template<class _Sp, class _Tp, class _Ap>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+mem_fun1_ref_t<_Sp,_Tp,_Ap>
+mem_fun_ref(_Sp (_Tp::*__f)(_Ap))
+    {return mem_fun1_ref_t<_Sp,_Tp,_Ap>(__f);}
+
+template <class _Sp, class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 const_mem_fun_t
+    : public unary_function<const _Tp*, _Sp>
+{
+    _Sp (_Tp::*__p_)() const;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit const_mem_fun_t(_Sp (_Tp::*__p)() const)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(const _Tp* __p) const
+        {return (__p->*__p_)();}
+};
+
+template <class _Sp, class _Tp, class _Ap>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 const_mem_fun1_t
+    : public binary_function<const _Tp*, _Ap, _Sp>
+{
+    _Sp (_Tp::*__p_)(_Ap) const;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit const_mem_fun1_t(_Sp (_Tp::*__p)(_Ap) const)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(const _Tp* __p, _Ap __x) const
+        {return (__p->*__p_)(__x);}
+};
+
+template <class _Sp, class _Tp>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+const_mem_fun_t<_Sp,_Tp>
+mem_fun(_Sp (_Tp::*__f)() const)
+    {return const_mem_fun_t<_Sp,_Tp>(__f);}
+
+template <class _Sp, class _Tp, class _Ap>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+const_mem_fun1_t<_Sp,_Tp,_Ap>
+mem_fun(_Sp (_Tp::*__f)(_Ap) const)
+    {return const_mem_fun1_t<_Sp,_Tp,_Ap>(__f);}
+
+template <class _Sp, class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 const_mem_fun_ref_t
+    : public unary_function<_Tp, _Sp>
+{
+    _Sp (_Tp::*__p_)() const;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit const_mem_fun_ref_t(_Sp (_Tp::*__p)() const)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(const _Tp& __p) const
+        {return (__p.*__p_)();}
+};
+
+template <class _Sp, class _Tp, class _Ap>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 const_mem_fun1_ref_t
+    : public binary_function<_Tp, _Ap, _Sp>
+{
+    _Sp (_Tp::*__p_)(_Ap) const;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit const_mem_fun1_ref_t(_Sp (_Tp::*__p)(_Ap) const)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _Sp operator()(const _Tp& __p, _Ap __x) const
+        {return (__p.*__p_)(__x);}
+};
+
+template <class _Sp, class _Tp>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+const_mem_fun_ref_t<_Sp,_Tp>
+mem_fun_ref(_Sp (_Tp::*__f)() const)
+    {return const_mem_fun_ref_t<_Sp,_Tp>(__f);}
+
+template <class _Sp, class _Tp, class _Ap>
+_LIBCUDACXX_DEPRECATED_IN_CXX11 inline _LIBCUDACXX_INLINE_VISIBILITY
+const_mem_fun1_ref_t<_Sp,_Tp,_Ap>
+mem_fun_ref(_Sp (_Tp::*__f)(_Ap) const)
+    {return const_mem_fun1_ref_t<_Sp,_Tp,_Ap>(__f);}
+#endif
+
+////////////////////////////////////////////////////////////////////////////////
+//                                MEMFUN
+//==============================================================================
+
+template <class _Tp>
+class __mem_fn
+    : public __weak_result_type<_Tp>
+{
+public:
+    // types
+    typedef _Tp type;
+private:
+    type __f_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __mem_fn(type __f) _NOEXCEPT : __f_(__f) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    // invoke
+    template <class... _ArgTypes>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return<type, _ArgTypes...>::type
+    operator() (_ArgTypes&&... __args) const {
+        return __invoke(__f_, _CUDA_VSTD::forward<_ArgTypes>(__args)...);
+    }
+#else
+
+    template <class _A0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return0<type, _A0>::type
+    operator() (_A0& __a0) const {
+        return __invoke(__f_, __a0);
+    }
+
+    template <class _A0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return0<type, _A0 const>::type
+    operator() (_A0 const& __a0) const {
+        return __invoke(__f_, __a0);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0, _A1>::type
+    operator() (_A0& __a0, _A1& __a1) const {
+        return __invoke(__f_, __a0, __a1);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0 const, _A1>::type
+    operator() (_A0 const& __a0, _A1& __a1) const {
+        return __invoke(__f_, __a0, __a1);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0, _A1 const>::type
+    operator() (_A0& __a0, _A1 const& __a1) const {
+        return __invoke(__f_, __a0, __a1);
+    }
+
+    template <class _A0, class _A1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return1<type, _A0 const, _A1 const>::type
+    operator() (_A0 const& __a0, _A1 const& __a1) const {
+        return __invoke(__f_, __a0, __a1);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1, _A2>::type
+    operator() (_A0& __a0, _A1& __a1, _A2& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1, _A2>::type
+    operator() (_A0 const& __a0, _A1& __a1, _A2& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1 const, _A2>::type
+    operator() (_A0& __a0, _A1 const& __a1, _A2& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1, _A2 const>::type
+    operator() (_A0& __a0, _A1& __a1, _A2 const& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1 const, _A2>::type
+    operator() (_A0 const& __a0, _A1 const& __a1, _A2& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1, _A2 const>::type
+    operator() (_A0 const& __a0, _A1& __a1, _A2 const& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0, _A1 const, _A2 const>::type
+    operator() (_A0& __a0, _A1 const& __a1, _A2 const& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+
+    template <class _A0, class _A1, class _A2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename __invoke_return2<type, _A0 const, _A1 const, _A2 const>::type
+    operator() (_A0 const& __a0, _A1 const& __a1, _A2 const& __a2) const {
+        return __invoke(__f_, __a0, __a1, __a2);
+    }
+#endif
+};
+
+template<class _Rp, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__mem_fn<_Rp _Tp::*>
+mem_fn(_Rp _Tp::* __pm) _NOEXCEPT
+{
+    return __mem_fn<_Rp _Tp::*>(__pm);
+}
+
+// `cuda::std::function` is not supported at this time.
+
+#ifndef __cuda_std__
+
+////////////////////////////////////////////////////////////////////////////////
+//                                FUNCTION
+//==============================================================================
+
+// bad_function_call
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_function_call
+    : public exception
+{
+#ifdef _LIBCUDACXX_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
+public:
+    virtual ~bad_function_call() _NOEXCEPT;
+
+    virtual const char* what() const _NOEXCEPT;
+#endif
+};
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_bad_function_call()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw bad_function_call();
+#else
+    _CUDA_VSTD::abort();
+#endif
+}
+
+template<class _Fp> class _LIBCUDACXX_TEMPLATE_VIS function; // undefined
+
+namespace __function
+{
+
+template<class _Rp>
+struct __maybe_derive_from_unary_function
+{
+};
+
+template<class _Rp, class _A1>
+struct __maybe_derive_from_unary_function<_Rp(_A1)>
+    : public unary_function<_A1, _Rp>
+{
+};
+
+template<class _Rp>
+struct __maybe_derive_from_binary_function
+{
+};
+
+template<class _Rp, class _A1, class _A2>
+struct __maybe_derive_from_binary_function<_Rp(_A1, _A2)>
+    : public binary_function<_A1, _A2, _Rp>
+{
+};
+
+template <class _Fp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __not_null(_Fp const&) { return true; }
+
+template <class _Fp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __not_null(_Fp* __ptr) { return __ptr; }
+
+template <class _Ret, class _Class>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __not_null(_Ret _Class::*__ptr) { return __ptr; }
+
+template <class _Fp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __not_null(function<_Fp> const& __f) { return !!__f; }
+
+} // namespace __function
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+namespace __function {
+
+// __alloc_func holds a functor and an allocator.
+
+template <class _Fp, class _Ap, class _FB> class __alloc_func;
+template <class _Fp, class _FB>
+class __default_alloc_func;
+
+template <class _Fp, class _Ap, class _Rp, class... _ArgTypes>
+class __alloc_func<_Fp, _Ap, _Rp(_ArgTypes...)>
+{
+    __compressed_pair<_Fp, _Ap> __f_;
+
+  public:
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Fp _Target;
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Ap _Alloc;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Target& __target() const { return __f_.first(); }
+
+    // WIN32 APIs may define __allocator, so use __get_allocator instead.
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Alloc& __get_allocator() const { return __f_.second(); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __alloc_func(_Target&& __f)
+        : __f_(piecewise_construct, _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__f)),
+               _CUDA_VSTD::forward_as_tuple())
+    {
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __alloc_func(const _Target& __f, const _Alloc& __a)
+        : __f_(piecewise_construct, _CUDA_VSTD::forward_as_tuple(__f),
+               _CUDA_VSTD::forward_as_tuple(__a))
+    {
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __alloc_func(const _Target& __f, _Alloc&& __a)
+        : __f_(piecewise_construct, _CUDA_VSTD::forward_as_tuple(__f),
+               _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__a)))
+    {
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __alloc_func(_Target&& __f, _Alloc&& __a)
+        : __f_(piecewise_construct, _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__f)),
+               _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__a)))
+    {
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Rp operator()(_ArgTypes&&... __arg)
+    {
+        typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+        return _Invoker::__call(__f_.first(),
+                                _CUDA_VSTD::forward<_ArgTypes>(__arg)...);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __alloc_func* __clone() const
+    {
+        typedef allocator_traits<_Alloc> __alloc_traits;
+        typedef
+            typename __rebind_alloc_helper<__alloc_traits, __alloc_func>::type
+                _AA;
+        _AA __a(__f_.second());
+        typedef __allocator_destructor<_AA> _Dp;
+        unique_ptr<__alloc_func, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+        ::new ((void*)__hold.get()) __alloc_func(__f_.first(), _Alloc(__a));
+        return __hold.release();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void destroy() _NOEXCEPT { __f_.~__compressed_pair<_Target, _Alloc>(); }
+
+    static void __destroy_and_delete(__alloc_func* __f) {
+      typedef allocator_traits<_Alloc> __alloc_traits;
+      typedef typename __rebind_alloc_helper<__alloc_traits, __alloc_func>::type
+          _FunAlloc;
+      _FunAlloc __a(__f->__get_allocator());
+      __f->destroy();
+      __a.deallocate(__f, 1);
+    }
+};
+
+template <class _Fp, class _Rp, class... _ArgTypes>
+class __default_alloc_func<_Fp, _Rp(_ArgTypes...)> {
+  _Fp __f_;
+
+public:
+  typedef _LIBCUDACXX_NODEBUG_TYPE _Fp _Target;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const _Target& __target() const { return __f_; }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit __default_alloc_func(_Target&& __f) : __f_(std::move(__f)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit __default_alloc_func(const _Target& __f) : __f_(__f) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _Rp operator()(_ArgTypes&&... __arg) {
+    typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+    return _Invoker::__call(__f_, _CUDA_VSTD::forward<_ArgTypes>(__arg)...);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __default_alloc_func* __clone() const {
+      __builtin_new_allocator::__holder_t __hold =
+        __builtin_new_allocator::__allocate_type<__default_alloc_func>(1);
+    __default_alloc_func* __res =
+        ::new (__hold.get()) __default_alloc_func(__f_);
+    (void)__hold.release();
+    return __res;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void destroy() _NOEXCEPT { __f_.~_Target(); }
+
+  static void __destroy_and_delete(__default_alloc_func* __f) {
+    __f->destroy();
+      __builtin_new_allocator::__deallocate_type<__default_alloc_func>(__f, 1);
+  }
+};
+
+// __base provides an abstract interface for copyable functors.
+
+template<class _Fp> class __base;
+
+template<class _Rp, class ..._ArgTypes>
+class __base<_Rp(_ArgTypes...)>
+{
+    __base(const __base&);
+    __base& operator=(const __base&);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __base() {}
+    _LIBCUDACXX_INLINE_VISIBILITY virtual ~__base() {}
+    virtual __base* __clone() const = 0;
+    virtual void __clone(__base*) const = 0;
+    virtual void destroy() _NOEXCEPT = 0;
+    virtual void destroy_deallocate() _NOEXCEPT = 0;
+    virtual _Rp operator()(_ArgTypes&& ...) = 0;
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const _NOEXCEPT = 0;
+    virtual const std::type_info& target_type() const _NOEXCEPT = 0;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+// __func implements __base for a given functor type.
+
+template<class _FD, class _Alloc, class _FB> class __func;
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+class __func<_Fp, _Alloc, _Rp(_ArgTypes...)>
+    : public  __base<_Rp(_ArgTypes...)>
+{
+    __alloc_func<_Fp, _Alloc, _Rp(_ArgTypes...)> __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __func(_Fp&& __f)
+        : __f_(_CUDA_VSTD::move(__f)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __func(const _Fp& __f, const _Alloc& __a)
+        : __f_(__f, __a) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __func(const _Fp& __f, _Alloc&& __a)
+        : __f_(__f, _CUDA_VSTD::move(__a)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __func(_Fp&& __f, _Alloc&& __a)
+        : __f_(_CUDA_VSTD::move(__f), _CUDA_VSTD::move(__a)) {}
+
+    virtual __base<_Rp(_ArgTypes...)>* __clone() const;
+    virtual void __clone(__base<_Rp(_ArgTypes...)>*) const;
+    virtual void destroy() _NOEXCEPT;
+    virtual void destroy_deallocate() _NOEXCEPT;
+    virtual _Rp operator()(_ArgTypes&&... __arg);
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* target(const type_info&) const _NOEXCEPT;
+    virtual const std::type_info& target_type() const _NOEXCEPT;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+__base<_Rp(_ArgTypes...)>*
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::__clone() const
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.__get_allocator());
+    typedef __allocator_destructor<_Ap> _Dp;
+    unique_ptr<__func, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+    ::new ((void*)__hold.get()) __func(__f_.__target(), _Alloc(__a));
+    return __hold.release();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+void
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::__clone(__base<_Rp(_ArgTypes...)>* __p) const
+{
+    ::new (__p) __func(__f_.__target(), __f_.__get_allocator());
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+void
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::destroy() _NOEXCEPT
+{
+    __f_.destroy();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+void
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::destroy_deallocate() _NOEXCEPT
+{
+    typedef allocator_traits<_Alloc> __alloc_traits;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __func>::type _Ap;
+    _Ap __a(__f_.__get_allocator());
+    __f_.destroy();
+    __a.deallocate(this, 1);
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+_Rp
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::operator()(_ArgTypes&& ... __arg)
+{
+    return __f_(_CUDA_VSTD::forward<_ArgTypes>(__arg)...);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+const void*
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::target(const type_info& __ti) const _NOEXCEPT
+{
+    if (__ti == typeid(_Fp))
+        return &__f_.__target();
+    return (const void*)0;
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+const std::type_info&
+__func<_Fp, _Alloc, _Rp(_ArgTypes...)>::target_type() const _NOEXCEPT
+{
+    return typeid(_Fp);
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+// __value_func creates a value-type from a __func.
+
+template <class _Fp> class __value_func;
+
+template <class _Rp, class... _ArgTypes> class __value_func<_Rp(_ArgTypes...)>
+{
+    typename aligned_storage<3 * sizeof(void*)>::type __buf_;
+
+    typedef __base<_Rp(_ArgTypes...)> __func;
+    __func* __f_;
+
+    _LIBCUDACXX_NO_CFI static __func* __as_base(void* p)
+    {
+        return reinterpret_cast<__func*>(p);
+    }
+
+  public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_func() _NOEXCEPT : __f_(0) {}
+
+    template <class _Fp, class _Alloc>
+    _LIBCUDACXX_INLINE_VISIBILITY __value_func(_Fp&& __f, const _Alloc& __a)
+        : __f_(0)
+    {
+        typedef allocator_traits<_Alloc> __alloc_traits;
+        typedef __function::__func<_Fp, _Alloc, _Rp(_ArgTypes...)> _Fun;
+        typedef typename __rebind_alloc_helper<__alloc_traits, _Fun>::type
+            _FunAlloc;
+
+        if (__function::__not_null(__f))
+        {
+            _FunAlloc __af(__a);
+            if (sizeof(_Fun) <= sizeof(__buf_) &&
+                is_nothrow_copy_constructible<_Fp>::value &&
+                is_nothrow_copy_constructible<_FunAlloc>::value)
+            {
+                __f_ =
+                    ::new ((void*)&__buf_) _Fun(_CUDA_VSTD::move(__f), _Alloc(__af));
+            }
+            else
+            {
+                typedef __allocator_destructor<_FunAlloc> _Dp;
+                unique_ptr<__func, _Dp> __hold(__af.allocate(1), _Dp(__af, 1));
+                ::new ((void*)__hold.get()) _Fun(_CUDA_VSTD::move(__f), _Alloc(__a));
+                __f_ = __hold.release();
+            }
+        }
+    }
+
+    template <class _Fp,
+        class = typename enable_if<!is_same<typename decay<_Fp>::type, __value_func>::value>::type>
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __value_func(_Fp&& __f)
+        : __value_func(std::forward<_Fp>(__f), allocator<_Fp>()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_func(const __value_func& __f)
+    {
+        if (__f.__f_ == 0)
+            __f_ = 0;
+        else if ((void*)__f.__f_ == &__f.__buf_)
+        {
+            __f_ = __as_base(&__buf_);
+            __f.__f_->__clone(__f_);
+        }
+        else
+            __f_ = __f.__f_->__clone();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_func(__value_func&& __f) _NOEXCEPT
+    {
+        if (__f.__f_ == 0)
+            __f_ = 0;
+        else if ((void*)__f.__f_ == &__f.__buf_)
+        {
+            __f_ = __as_base(&__buf_);
+            __f.__f_->__clone(__f_);
+        }
+        else
+        {
+            __f_ = __f.__f_;
+            __f.__f_ = 0;
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__value_func()
+    {
+        if ((void*)__f_ == &__buf_)
+            __f_->destroy();
+        else if (__f_)
+            __f_->destroy_deallocate();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_func& operator=(__value_func&& __f)
+    {
+        *this = nullptr;
+        if (__f.__f_ == 0)
+            __f_ = 0;
+        else if ((void*)__f.__f_ == &__f.__buf_)
+        {
+            __f_ = __as_base(&__buf_);
+            __f.__f_->__clone(__f_);
+        }
+        else
+        {
+            __f_ = __f.__f_;
+            __f.__f_ = 0;
+        }
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_func& operator=(nullptr_t)
+    {
+        __func* __f = __f_;
+        __f_ = 0;
+        if ((void*)__f == &__buf_)
+            __f->destroy();
+        else if (__f)
+            __f->destroy_deallocate();
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Rp operator()(_ArgTypes&&... __args) const
+    {
+        if (__f_ == 0)
+            __throw_bad_function_call();
+        return (*__f_)(_CUDA_VSTD::forward<_ArgTypes>(__args)...);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(__value_func& __f) _NOEXCEPT
+    {
+        if (&__f == this)
+            return;
+        if ((void*)__f_ == &__buf_ && (void*)__f.__f_ == &__f.__buf_)
+        {
+            typename aligned_storage<sizeof(__buf_)>::type __tempbuf;
+            __func* __t = __as_base(&__tempbuf);
+            __f_->__clone(__t);
+            __f_->destroy();
+            __f_ = 0;
+            __f.__f_->__clone(__as_base(&__buf_));
+            __f.__f_->destroy();
+            __f.__f_ = 0;
+            __f_ = __as_base(&__buf_);
+            __t->__clone(__as_base(&__f.__buf_));
+            __t->destroy();
+            __f.__f_ = __as_base(&__f.__buf_);
+        }
+        else if ((void*)__f_ == &__buf_)
+        {
+            __f_->__clone(__as_base(&__f.__buf_));
+            __f_->destroy();
+            __f_ = __f.__f_;
+            __f.__f_ = __as_base(&__f.__buf_);
+        }
+        else if ((void*)__f.__f_ == &__f.__buf_)
+        {
+            __f.__f_->__clone(__as_base(&__buf_));
+            __f.__f_->destroy();
+            __f.__f_ = __f_;
+            __f_ = __as_base(&__buf_);
+        }
+        else
+            _CUDA_VSTD::swap(__f_, __f.__f_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT { return __f_ != 0; }
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const std::type_info& target_type() const _NOEXCEPT
+    {
+        if (__f_ == 0)
+            return typeid(void);
+        return __f_->target_type();
+    }
+
+    template <typename _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY const _Tp* target() const _NOEXCEPT
+    {
+        if (__f_ == 0)
+            return 0;
+        return (const _Tp*)__f_->target(typeid(_Tp));
+    }
+#endif // _LIBCUDACXX_NO_RTTI
+};
+
+// Storage for a functor object, to be used with __policy to manage copy and
+// destruction.
+union __policy_storage
+{
+    mutable char __small[sizeof(void*) * 2];
+    void* __large;
+};
+
+// True if _Fun can safely be held in __policy_storage.__small.
+template <typename _Fun>
+struct __use_small_storage
+    : public _CUDA_VSTD::integral_constant<
+          bool, sizeof(_Fun) <= sizeof(__policy_storage) &&
+                    _LIBCUDACXX_ALIGNOF(_Fun) <= _LIBCUDACXX_ALIGNOF(__policy_storage) &&
+                    _CUDA_VSTD::is_trivially_copy_constructible<_Fun>::value &&
+                    _CUDA_VSTD::is_trivially_destructible<_Fun>::value> {};
+
+// Policy contains information about how to copy, destroy, and move the
+// underlying functor. You can think of it as a vtable of sorts.
+struct __policy
+{
+    // Used to copy or destroy __large values. null for trivial objects.
+    void* (*const __clone)(const void*);
+    void (*const __destroy)(void*);
+
+    // True if this is the null policy (no value).
+    const bool __is_null;
+
+    // The target type. May be null if RTTI is disabled.
+    const std::type_info* const __type_info;
+
+    // Returns a pointer to a static policy object suitable for the functor
+    // type.
+    template <typename _Fun>
+    _LIBCUDACXX_INLINE_VISIBILITY static const __policy* __create()
+    {
+        return __choose_policy<_Fun>(__use_small_storage<_Fun>());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static const __policy* __create_empty()
+    {
+        static const _LIBCUDACXX_CONSTEXPR __policy __policy_ = {nullptr, nullptr,
+                                                             true,
+#ifndef _LIBCUDACXX_NO_RTTI
+                                                             &typeid(void)
+#else
+                                                             nullptr
+#endif
+        };
+        return &__policy_;
+    }
+
+  private:
+    template <typename _Fun> static void* __large_clone(const void* __s)
+    {
+        const _Fun* __f = static_cast<const _Fun*>(__s);
+        return __f->__clone();
+    }
+
+    template <typename _Fun>
+    static void __large_destroy(void* __s) {
+      _Fun::__destroy_and_delete(static_cast<_Fun*>(__s));
+    }
+
+    template <typename _Fun>
+    _LIBCUDACXX_INLINE_VISIBILITY static const __policy*
+    __choose_policy(/* is_small = */ false_type) {
+      static const _LIBCUDACXX_CONSTEXPR __policy __policy_ = {
+          &__large_clone<_Fun>, &__large_destroy<_Fun>, false,
+#ifndef _LIBCUDACXX_NO_RTTI
+          &typeid(typename _Fun::_Target)
+#else
+          nullptr
+#endif
+      };
+        return &__policy_;
+    }
+
+    template <typename _Fun>
+    _LIBCUDACXX_INLINE_VISIBILITY static const __policy*
+        __choose_policy(/* is_small = */ true_type)
+    {
+        static const _LIBCUDACXX_CONSTEXPR __policy __policy_ = {
+            nullptr, nullptr, false,
+#ifndef _LIBCUDACXX_NO_RTTI
+            &typeid(typename _Fun::_Target)
+#else
+            nullptr
+#endif
+        };
+        return &__policy_;
+    }
+};
+
+// Used to choose between perfect forwarding or pass-by-value. Pass-by-value is
+// faster for types that can be passed in registers.
+template <typename _Tp>
+using __fast_forward =
+    typename _CUDA_VSTD::conditional<_CUDA_VSTD::is_scalar<_Tp>::value, _Tp, _Tp&&>::type;
+
+// __policy_invoker calls an instance of __alloc_func held in __policy_storage.
+
+template <class _Fp> struct __policy_invoker;
+
+template <class _Rp, class... _ArgTypes>
+struct __policy_invoker<_Rp(_ArgTypes...)>
+{
+    typedef _Rp (*__Call)(const __policy_storage*,
+                          __fast_forward<_ArgTypes>...);
+
+    __Call __call_;
+
+    // Creates an invoker that throws bad_function_call.
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __policy_invoker() : __call_(&__call_empty) {}
+
+    // Creates an invoker that calls the given instance of __func.
+    template <typename _Fun>
+    _LIBCUDACXX_INLINE_VISIBILITY static __policy_invoker __create()
+    {
+        return __policy_invoker(&__call_impl<_Fun>);
+    }
+
+  private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __policy_invoker(__Call __c) : __call_(__c) {}
+
+    static _Rp __call_empty(const __policy_storage*,
+                            __fast_forward<_ArgTypes>...)
+    {
+        __throw_bad_function_call();
+    }
+
+    template <typename _Fun>
+    static _Rp __call_impl(const __policy_storage* __buf,
+                           __fast_forward<_ArgTypes>... __args)
+    {
+        _Fun* __f = reinterpret_cast<_Fun*>(__use_small_storage<_Fun>::value
+                                                ? &__buf->__small
+                                                : __buf->__large);
+        return (*__f)(_CUDA_VSTD::forward<_ArgTypes>(__args)...);
+    }
+};
+
+// __policy_func uses a __policy and __policy_invoker to create a type-erased,
+// copyable functor.
+
+template <class _Fp> class __policy_func;
+
+template <class _Rp, class... _ArgTypes> class __policy_func<_Rp(_ArgTypes...)>
+{
+    // Inline storage for small objects.
+    __policy_storage __buf_;
+
+    // Calls the value stored in __buf_. This could technically be part of
+    // policy, but storing it here eliminates a level of indirection inside
+    // operator().
+    typedef __function::__policy_invoker<_Rp(_ArgTypes...)> __invoker;
+    __invoker __invoker_;
+
+    // The policy that describes how to move / copy / destroy __buf_. Never
+    // null, even if the function is empty.
+    const __policy* __policy_;
+
+  public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __policy_func() : __policy_(__policy::__create_empty()) {}
+
+    template <class _Fp, class _Alloc>
+    _LIBCUDACXX_INLINE_VISIBILITY __policy_func(_Fp&& __f, const _Alloc& __a)
+        : __policy_(__policy::__create_empty())
+    {
+        typedef __alloc_func<_Fp, _Alloc, _Rp(_ArgTypes...)> _Fun;
+        typedef allocator_traits<_Alloc> __alloc_traits;
+        typedef typename __rebind_alloc_helper<__alloc_traits, _Fun>::type
+            _FunAlloc;
+
+        if (__function::__not_null(__f))
+        {
+            __invoker_ = __invoker::template __create<_Fun>();
+            __policy_ = __policy::__create<_Fun>();
+
+            _FunAlloc __af(__a);
+            if (__use_small_storage<_Fun>())
+            {
+                ::new ((void*)&__buf_.__small)
+                    _Fun(_CUDA_VSTD::move(__f), _Alloc(__af));
+            }
+            else
+            {
+                typedef __allocator_destructor<_FunAlloc> _Dp;
+                unique_ptr<_Fun, _Dp> __hold(__af.allocate(1), _Dp(__af, 1));
+                ::new ((void*)__hold.get())
+                    _Fun(_CUDA_VSTD::move(__f), _Alloc(__af));
+                __buf_.__large = __hold.release();
+            }
+        }
+    }
+
+    template <class _Fp, class = typename enable_if<!is_same<typename decay<_Fp>::type, __policy_func>::value>::type>
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __policy_func(_Fp&& __f)
+        : __policy_(__policy::__create_empty()) {
+      typedef __default_alloc_func<_Fp, _Rp(_ArgTypes...)> _Fun;
+
+      if (__function::__not_null(__f)) {
+        __invoker_ = __invoker::template __create<_Fun>();
+        __policy_ = __policy::__create<_Fun>();
+        if (__use_small_storage<_Fun>()) {
+          ::new ((void*)&__buf_.__small) _Fun(_CUDA_VSTD::move(__f));
+        } else {
+          __builtin_new_allocator::__holder_t __hold =
+              __builtin_new_allocator::__allocate_type<_Fun>(1);
+          __buf_.__large = ::new (__hold.get()) _Fun(_CUDA_VSTD::move(__f));
+          (void)__hold.release();
+        }
+      }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __policy_func(const __policy_func& __f)
+        : __buf_(__f.__buf_), __invoker_(__f.__invoker_),
+          __policy_(__f.__policy_)
+    {
+        if (__policy_->__clone)
+            __buf_.__large = __policy_->__clone(__f.__buf_.__large);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __policy_func(__policy_func&& __f)
+        : __buf_(__f.__buf_), __invoker_(__f.__invoker_),
+          __policy_(__f.__policy_)
+    {
+        if (__policy_->__destroy)
+        {
+            __f.__policy_ = __policy::__create_empty();
+            __f.__invoker_ = __invoker();
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__policy_func()
+    {
+        if (__policy_->__destroy)
+            __policy_->__destroy(__buf_.__large);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __policy_func& operator=(__policy_func&& __f)
+    {
+        *this = nullptr;
+        __buf_ = __f.__buf_;
+        __invoker_ = __f.__invoker_;
+        __policy_ = __f.__policy_;
+        __f.__policy_ = __policy::__create_empty();
+        __f.__invoker_ = __invoker();
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __policy_func& operator=(nullptr_t)
+    {
+        const __policy* __p = __policy_;
+        __policy_ = __policy::__create_empty();
+        __invoker_ = __invoker();
+        if (__p->__destroy)
+            __p->__destroy(__buf_.__large);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Rp operator()(_ArgTypes&&... __args) const
+    {
+        return __invoker_.__call_(_CUDA_VSTD::addressof(__buf_),
+                                  _CUDA_VSTD::forward<_ArgTypes>(__args)...);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(__policy_func& __f)
+    {
+        _CUDA_VSTD::swap(__invoker_, __f.__invoker_);
+        _CUDA_VSTD::swap(__policy_, __f.__policy_);
+        _CUDA_VSTD::swap(__buf_, __f.__buf_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit operator bool() const _NOEXCEPT
+    {
+        return !__policy_->__is_null;
+    }
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const std::type_info& target_type() const _NOEXCEPT
+    {
+        return *__policy_->__type_info;
+    }
+
+    template <typename _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY const _Tp* target() const _NOEXCEPT
+    {
+        if (__policy_->__is_null || typeid(_Tp) != *__policy_->__type_info)
+            return nullptr;
+        if (__policy_->__clone) // Out of line storage.
+            return reinterpret_cast<const _Tp*>(__buf_.__large);
+        else
+            return reinterpret_cast<const _Tp*>(&__buf_.__small);
+    }
+#endif // _LIBCUDACXX_NO_RTTI
+};
+
+}  // __function
+
+template<class _Rp, class ..._ArgTypes>
+class _LIBCUDACXX_TEMPLATE_VIS function<_Rp(_ArgTypes...)>
+    : public __function::__maybe_derive_from_unary_function<_Rp(_ArgTypes...)>,
+      public __function::__maybe_derive_from_binary_function<_Rp(_ArgTypes...)>
+{
+#ifndef _LIBCUDACXX_ABI_OPTIMIZED_FUNCTION
+    typedef __function::__value_func<_Rp(_ArgTypes...)> __func;
+#else
+    typedef __function::__policy_func<_Rp(_ArgTypes...)> __func;
+#endif
+
+    __func __f_;
+
+    template <class _Fp, bool = _And<
+        _IsNotSame<__uncvref_t<_Fp>, function>,
+        __invokable<_Fp&, _ArgTypes...>
+    >::value>
+    struct __callable;
+    template <class _Fp>
+        struct __callable<_Fp, true>
+        {
+            static const bool value = is_same<void, _Rp>::value ||
+                is_convertible<typename __invoke_of<_Fp&, _ArgTypes...>::type,
+                               _Rp>::value;
+        };
+    template <class _Fp>
+        struct __callable<_Fp, false>
+        {
+            static const bool value = false;
+        };
+
+  template <class _Fp>
+  using _EnableIfCallable = typename enable_if<__callable<_Fp>::value>::type;
+public:
+    typedef _Rp result_type;
+
+    // construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    function() _NOEXCEPT { }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    function(nullptr_t) _NOEXCEPT {}
+    function(const function&);
+    function(function&&) _NOEXCEPT;
+    template<class _Fp, class = _EnableIfCallable<_Fp>>
+    function(_Fp);
+
+#if _LIBCUDACXX_STD_VER <= 14
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&) _NOEXCEPT {}
+    template<class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      function(allocator_arg_t, const _Alloc&, nullptr_t) _NOEXCEPT {}
+    template<class _Alloc>
+      function(allocator_arg_t, const _Alloc&, const function&);
+    template<class _Alloc>
+      function(allocator_arg_t, const _Alloc&, function&&);
+    template<class _Fp, class _Alloc, class = _EnableIfCallable<_Fp>>
+      function(allocator_arg_t, const _Alloc& __a, _Fp __f);
+#endif
+
+    function& operator=(const function&);
+    function& operator=(function&&) _NOEXCEPT;
+    function& operator=(nullptr_t) _NOEXCEPT;
+    template<class _Fp, class = _EnableIfCallable<_Fp>>
+    function& operator=(_Fp&&);
+
+    ~function();
+
+    // function modifiers:
+    void swap(function&) _NOEXCEPT;
+
+#if _LIBCUDACXX_STD_VER <= 14
+    template<class _Fp, class _Alloc>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      void assign(_Fp&& __f, const _Alloc& __a)
+        {function(allocator_arg, __a, _CUDA_VSTD::forward<_Fp>(__f)).swap(*this);}
+#endif
+
+    // function capacity:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT {
+      return static_cast<bool>(__f_);
+    }
+
+    // deleted overloads close possible hole in the type system
+    template<class _R2, class... _ArgTypes2>
+      bool operator==(const function<_R2(_ArgTypes2...)>&) const = delete;
+    template<class _R2, class... _ArgTypes2>
+      bool operator!=(const function<_R2(_ArgTypes2...)>&) const = delete;
+public:
+    // function invocation:
+    _Rp operator()(_ArgTypes...) const;
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    // function target access:
+    const std::type_info& target_type() const _NOEXCEPT;
+    template <typename _Tp> _Tp* target() _NOEXCEPT;
+    template <typename _Tp> const _Tp* target() const _NOEXCEPT;
+#endif  // _LIBCUDACXX_NO_RTTI
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _Rp, class ..._Ap>
+function(_Rp(*)(_Ap...)) -> function<_Rp(_Ap...)>;
+
+template<class _Fp>
+struct __strip_signature;
+
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...)> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) volatile> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const volatile> { using type = _Rp(_Ap...); };
+
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) &> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const &> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) volatile &> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const volatile &> { using type = _Rp(_Ap...); };
+
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) noexcept> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const noexcept> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) volatile noexcept> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const volatile noexcept> { using type = _Rp(_Ap...); };
+
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) & noexcept> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const & noexcept> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) volatile & noexcept> { using type = _Rp(_Ap...); };
+template<class _Rp, class _Gp, class ..._Ap>
+struct __strip_signature<_Rp (_Gp::*) (_Ap...) const volatile & noexcept> { using type = _Rp(_Ap...); };
+
+template<class _Fp, class _Stripped = typename __strip_signature<decltype(&_Fp::operator())>::type>
+function(_Fp) -> function<_Stripped>;
+#endif // !_LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+
+template<class _Rp, class ..._ArgTypes>
+function<_Rp(_ArgTypes...)>::function(const function& __f) : __f_(__f.__f_) {}
+
+#if _LIBCUDACXX_STD_VER <= 14
+template<class _Rp, class ..._ArgTypes>
+template <class _Alloc>
+function<_Rp(_ArgTypes...)>::function(allocator_arg_t, const _Alloc&,
+                                     const function& __f) : __f_(__f.__f_) {}
+#endif
+
+template <class _Rp, class... _ArgTypes>
+function<_Rp(_ArgTypes...)>::function(function&& __f) _NOEXCEPT
+    : __f_(_CUDA_VSTD::move(__f.__f_)) {}
+
+#if _LIBCUDACXX_STD_VER <= 14
+template<class _Rp, class ..._ArgTypes>
+template <class _Alloc>
+function<_Rp(_ArgTypes...)>::function(allocator_arg_t, const _Alloc&,
+                                      function&& __f)
+    : __f_(_CUDA_VSTD::move(__f.__f_)) {}
+#endif
+
+template <class _Rp, class... _ArgTypes>
+template <class _Fp, class>
+function<_Rp(_ArgTypes...)>::function(_Fp __f) : __f_(_CUDA_VSTD::move(__f)) {}
+
+#if _LIBCUDACXX_STD_VER <= 14
+template <class _Rp, class... _ArgTypes>
+template <class _Fp, class _Alloc, class>
+function<_Rp(_ArgTypes...)>::function(allocator_arg_t, const _Alloc& __a,
+                                      _Fp __f)
+    : __f_(_CUDA_VSTD::move(__f), __a) {}
+#endif
+
+template<class _Rp, class ..._ArgTypes>
+function<_Rp(_ArgTypes...)>&
+function<_Rp(_ArgTypes...)>::operator=(const function& __f)
+{
+    function(__f).swap(*this);
+    return *this;
+}
+
+template<class _Rp, class ..._ArgTypes>
+function<_Rp(_ArgTypes...)>&
+function<_Rp(_ArgTypes...)>::operator=(function&& __f) _NOEXCEPT
+{
+    __f_ = std::move(__f.__f_);
+    return *this;
+}
+
+template<class _Rp, class ..._ArgTypes>
+function<_Rp(_ArgTypes...)>&
+function<_Rp(_ArgTypes...)>::operator=(nullptr_t) _NOEXCEPT
+{
+    __f_ = nullptr;
+    return *this;
+}
+
+template<class _Rp, class ..._ArgTypes>
+template <class _Fp, class>
+function<_Rp(_ArgTypes...)>&
+function<_Rp(_ArgTypes...)>::operator=(_Fp&& __f)
+{
+    function(_CUDA_VSTD::forward<_Fp>(__f)).swap(*this);
+    return *this;
+}
+
+template<class _Rp, class ..._ArgTypes>
+function<_Rp(_ArgTypes...)>::~function() {}
+
+template<class _Rp, class ..._ArgTypes>
+void
+function<_Rp(_ArgTypes...)>::swap(function& __f) _NOEXCEPT
+{
+    __f_.swap(__f.__f_);
+}
+
+template<class _Rp, class ..._ArgTypes>
+_Rp
+function<_Rp(_ArgTypes...)>::operator()(_ArgTypes... __arg) const
+{
+    return __f_(_CUDA_VSTD::forward<_ArgTypes>(__arg)...);
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Rp, class ..._ArgTypes>
+const std::type_info&
+function<_Rp(_ArgTypes...)>::target_type() const _NOEXCEPT
+{
+    return __f_.target_type();
+}
+
+template<class _Rp, class ..._ArgTypes>
+template <typename _Tp>
+_Tp*
+function<_Rp(_ArgTypes...)>::target() _NOEXCEPT
+{
+    return (_Tp*)(__f_.template target<_Tp>());
+}
+
+template<class _Rp, class ..._ArgTypes>
+template <typename _Tp>
+const _Tp*
+function<_Rp(_ArgTypes...)>::target() const _NOEXCEPT
+{
+    return __f_.template target<_Tp>();
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template <class _Rp, class... _ArgTypes>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const function<_Rp(_ArgTypes...)>& __f, nullptr_t) _NOEXCEPT {return !__f;}
+
+template <class _Rp, class... _ArgTypes>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(nullptr_t, const function<_Rp(_ArgTypes...)>& __f) _NOEXCEPT {return !__f;}
+
+template <class _Rp, class... _ArgTypes>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const function<_Rp(_ArgTypes...)>& __f, nullptr_t) _NOEXCEPT {return (bool)__f;}
+
+template <class _Rp, class... _ArgTypes>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(nullptr_t, const function<_Rp(_ArgTypes...)>& __f) _NOEXCEPT {return (bool)__f;}
+
+template <class _Rp, class... _ArgTypes>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(function<_Rp(_ArgTypes...)>& __x, function<_Rp(_ArgTypes...)>& __y) _NOEXCEPT
+{return __x.swap(__y);}
+
+#else // _LIBCUDACXX_CXX03_LANG
+
+#include <__functional_03>
+
+#endif
+
+#endif // __cuda_std__
+
+////////////////////////////////////////////////////////////////////////////////
+//                                  BIND
+//==============================================================================
+
+// `cuda::std::bind` is not currently supported.
+
+#ifndef __cuda_std__
+
+template<class _Tp> struct __is_bind_expression : public false_type {};
+template<class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_bind_expression
+    : public __is_bind_expression<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t is_bind_expression_v = is_bind_expression<_Tp>::value;
+#endif
+
+template<class _Tp> struct __is_placeholder : public integral_constant<int, 0> {};
+template<class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_placeholder
+    : public __is_placeholder<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t is_placeholder_v = is_placeholder<_Tp>::value;
+#endif
+
+namespace placeholders
+{
+
+template <int _Np> struct __ph {};
+
+#if defined(_LIBCUDACXX_CXX03_LANG) || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+_LIBCUDACXX_FUNC_VIS extern const __ph<1>   _1;
+_LIBCUDACXX_FUNC_VIS extern const __ph<2>   _2;
+_LIBCUDACXX_FUNC_VIS extern const __ph<3>   _3;
+_LIBCUDACXX_FUNC_VIS extern const __ph<4>   _4;
+_LIBCUDACXX_FUNC_VIS extern const __ph<5>   _5;
+_LIBCUDACXX_FUNC_VIS extern const __ph<6>   _6;
+_LIBCUDACXX_FUNC_VIS extern const __ph<7>   _7;
+_LIBCUDACXX_FUNC_VIS extern const __ph<8>   _8;
+_LIBCUDACXX_FUNC_VIS extern const __ph<9>   _9;
+_LIBCUDACXX_FUNC_VIS extern const __ph<10> _10;
+#else
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<1>   _1{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<2>   _2{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<3>   _3{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<4>   _4{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<5>   _5{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<6>   _6{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<7>   _7{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<8>   _8{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<9>   _9{};
+/* _LIBCUDACXX_INLINE_VAR */ constexpr __ph<10> _10{};
+#endif // defined(_LIBCUDACXX_CXX03_LANG) || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+
+}  // placeholders
+
+template<int _Np>
+struct __is_placeholder<placeholders::__ph<_Np> >
+    : public integral_constant<int, _Np> {};
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Uj>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp&
+__mu(reference_wrapper<_Tp> __t, _Uj&)
+{
+    return __t.get();
+}
+
+template <class _Ti, class ..._Uj, size_t ..._Indx>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __invoke_of<_Ti&, _Uj...>::type
+__mu_expand(_Ti& __ti, tuple<_Uj...>& __uj, __tuple_indices<_Indx...>)
+{
+    return __ti(_CUDA_VSTD::forward<_Uj>(_CUDA_VSTD::get<_Indx>(__uj))...);
+}
+
+template <class _Ti, class ..._Uj>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename _EnableIf
+<
+    is_bind_expression<_Ti>::value,
+    __invoke_of<_Ti&, _Uj...>
+>::type
+__mu(_Ti& __ti, tuple<_Uj...>& __uj)
+{
+    typedef typename __make_tuple_indices<sizeof...(_Uj)>::type __indices;
+    return  __mu_expand(__ti, __uj, __indices());
+}
+
+template <bool IsPh, class _Ti, class _Uj>
+struct __mu_return2 {};
+
+template <class _Ti, class _Uj>
+struct __mu_return2<true, _Ti, _Uj>
+{
+    typedef typename tuple_element<is_placeholder<_Ti>::value - 1, _Uj>::type type;
+};
+
+template <class _Ti, class _Uj>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    0 < is_placeholder<_Ti>::value,
+    typename __mu_return2<0 < is_placeholder<_Ti>::value, _Ti, _Uj>::type
+>::type
+__mu(_Ti&, _Uj& __uj)
+{
+    const size_t _Indx = is_placeholder<_Ti>::value - 1;
+    return _CUDA_VSTD::forward<typename tuple_element<_Indx, _Uj>::type>(_CUDA_VSTD::get<_Indx>(__uj));
+}
+
+template <class _Ti, class _Uj>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_bind_expression<_Ti>::value &&
+    is_placeholder<_Ti>::value == 0 &&
+    !__is_reference_wrapper<_Ti>::value,
+    _Ti&
+>::type
+__mu(_Ti& __ti, _Uj&)
+{
+    return __ti;
+}
+
+template <class _Ti, bool IsReferenceWrapper, bool IsBindEx, bool IsPh,
+          class _TupleUj>
+struct __mu_return_impl;
+
+template <bool _Invokable, class _Ti, class ..._Uj>
+struct __mu_return_invokable  // false
+{
+    typedef __nat type;
+};
+
+template <class _Ti, class ..._Uj>
+struct __mu_return_invokable<true, _Ti, _Uj...>
+{
+    typedef typename __invoke_of<_Ti&, _Uj...>::type type;
+};
+
+template <class _Ti, class ..._Uj>
+struct __mu_return_impl<_Ti, false, true, false, tuple<_Uj...> >
+    : public __mu_return_invokable<__invokable<_Ti&, _Uj...>::value, _Ti, _Uj...>
+{
+};
+
+template <class _Ti, class _TupleUj>
+struct __mu_return_impl<_Ti, false, false, true, _TupleUj>
+{
+    typedef typename tuple_element<is_placeholder<_Ti>::value - 1,
+                                   _TupleUj>::type&& type;
+};
+
+template <class _Ti, class _TupleUj>
+struct __mu_return_impl<_Ti, true, false, false, _TupleUj>
+{
+    typedef typename _Ti::type& type;
+};
+
+template <class _Ti, class _TupleUj>
+struct __mu_return_impl<_Ti, false, false, false, _TupleUj>
+{
+    typedef _Ti& type;
+};
+
+template <class _Ti, class _TupleUj>
+struct __mu_return
+    : public __mu_return_impl<_Ti,
+                              __is_reference_wrapper<_Ti>::value,
+                              is_bind_expression<_Ti>::value,
+                              0 < is_placeholder<_Ti>::value &&
+                              is_placeholder<_Ti>::value <= tuple_size<_TupleUj>::value,
+                              _TupleUj>
+{
+};
+
+template <class _Fp, class _BoundArgs, class _TupleUj>
+struct __is_valid_bind_return
+{
+    static const bool value = false;
+};
+
+template <class _Fp, class ..._BoundArgs, class _TupleUj>
+struct __is_valid_bind_return<_Fp, tuple<_BoundArgs...>, _TupleUj>
+{
+    static const bool value = __invokable<_Fp,
+                    typename __mu_return<_BoundArgs, _TupleUj>::type...>::value;
+};
+
+template <class _Fp, class ..._BoundArgs, class _TupleUj>
+struct __is_valid_bind_return<_Fp, const tuple<_BoundArgs...>, _TupleUj>
+{
+    static const bool value = __invokable<_Fp,
+                    typename __mu_return<const _BoundArgs, _TupleUj>::type...>::value;
+};
+
+template <class _Fp, class _BoundArgs, class _TupleUj,
+          bool = __is_valid_bind_return<_Fp, _BoundArgs, _TupleUj>::value>
+struct __bind_return;
+
+template <class _Fp, class ..._BoundArgs, class _TupleUj>
+struct __bind_return<_Fp, tuple<_BoundArgs...>, _TupleUj, true>
+{
+    typedef typename __invoke_of
+    <
+        _Fp&,
+        typename __mu_return
+        <
+            _BoundArgs,
+            _TupleUj
+        >::type...
+    >::type type;
+};
+
+template <class _Fp, class ..._BoundArgs, class _TupleUj>
+struct __bind_return<_Fp, const tuple<_BoundArgs...>, _TupleUj, true>
+{
+    typedef typename __invoke_of
+    <
+        _Fp&,
+        typename __mu_return
+        <
+            const _BoundArgs,
+            _TupleUj
+        >::type...
+    >::type type;
+};
+
+template <class _Fp, class _BoundArgs, size_t ..._Indx, class _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __bind_return<_Fp, _BoundArgs, _Args>::type
+__apply_functor(_Fp& __f, _BoundArgs& __bound_args, __tuple_indices<_Indx...>,
+                _Args&& __args)
+{
+    return _CUDA_VSTD::__invoke(__f, _CUDA_VSTD::__mu(_CUDA_VSTD::get<_Indx>(__bound_args), __args)...);
+}
+
+template<class _Fp, class ..._BoundArgs>
+class __bind
+    : public __weak_result_type<typename decay<_Fp>::type>
+{
+protected:
+    typedef typename decay<_Fp>::type _Fd;
+    typedef tuple<typename decay<_BoundArgs>::type...> _Td;
+private:
+    _Fd __f_;
+    _Td __bound_args_;
+
+    typedef typename __make_tuple_indices<sizeof...(_BoundArgs)>::type __indices;
+public:
+    template <class _Gp, class ..._BA,
+              class = typename enable_if
+                               <
+                                  is_constructible<_Fd, _Gp>::value &&
+                                  !is_same<typename remove_reference<_Gp>::type,
+                                           __bind>::value
+                               >::type>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      explicit __bind(_Gp&& __f, _BA&& ...__bound_args)
+        : __f_(_CUDA_VSTD::forward<_Gp>(__f)),
+          __bound_args_(_CUDA_VSTD::forward<_BA>(__bound_args)...) {}
+
+    template <class ..._Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename __bind_return<_Fd, _Td, tuple<_Args&&...> >::type
+        operator()(_Args&& ...__args)
+        {
+            return _CUDA_VSTD::__apply_functor(__f_, __bound_args_, __indices(),
+                                  tuple<_Args&&...>(_CUDA_VSTD::forward<_Args>(__args)...));
+        }
+
+    template <class ..._Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename __bind_return<const _Fd, const _Td, tuple<_Args&&...> >::type
+        operator()(_Args&& ...__args) const
+        {
+            return _CUDA_VSTD::__apply_functor(__f_, __bound_args_, __indices(),
+                                   tuple<_Args&&...>(_CUDA_VSTD::forward<_Args>(__args)...));
+        }
+};
+
+template<class _Fp, class ..._BoundArgs>
+struct __is_bind_expression<__bind<_Fp, _BoundArgs...> > : public true_type {};
+
+template<class _Rp, class _Fp, class ..._BoundArgs>
+class __bind_r
+    : public __bind<_Fp, _BoundArgs...>
+{
+    typedef __bind<_Fp, _BoundArgs...> base;
+    typedef typename base::_Fd _Fd;
+    typedef typename base::_Td _Td;
+public:
+    typedef _Rp result_type;
+
+
+    template <class _Gp, class ..._BA,
+              class = typename enable_if
+                               <
+                                  is_constructible<_Fd, _Gp>::value &&
+                                  !is_same<typename remove_reference<_Gp>::type,
+                                           __bind_r>::value
+                               >::type>
+      _LIBCUDACXX_INLINE_VISIBILITY
+      explicit __bind_r(_Gp&& __f, _BA&& ...__bound_args)
+        : base(_CUDA_VSTD::forward<_Gp>(__f),
+               _CUDA_VSTD::forward<_BA>(__bound_args)...) {}
+
+    template <class ..._Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            is_convertible<typename __bind_return<_Fd, _Td, tuple<_Args&&...> >::type,
+                           result_type>::value || is_void<_Rp>::value,
+            result_type
+        >::type
+        operator()(_Args&& ...__args)
+        {
+            typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+            return _Invoker::__call(static_cast<base&>(*this), _CUDA_VSTD::forward<_Args>(__args)...);
+        }
+
+    template <class ..._Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            is_convertible<typename __bind_return<const _Fd, const _Td, tuple<_Args&&...> >::type,
+                           result_type>::value || is_void<_Rp>::value,
+            result_type
+        >::type
+        operator()(_Args&& ...__args) const
+        {
+            typedef __invoke_void_return_wrapper<_Rp> _Invoker;
+            return _Invoker::__call(static_cast<base const&>(*this), _CUDA_VSTD::forward<_Args>(__args)...);
+        }
+};
+
+template<class _Rp, class _Fp, class ..._BoundArgs>
+struct __is_bind_expression<__bind_r<_Rp, _Fp, _BoundArgs...> > : public true_type {};
+
+template<class _Fp, class ..._BoundArgs>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bind<_Fp, _BoundArgs...>
+bind(_Fp&& __f, _BoundArgs&&... __bound_args)
+{
+    typedef __bind<_Fp, _BoundArgs...> type;
+    return type(_CUDA_VSTD::forward<_Fp>(__f), _CUDA_VSTD::forward<_BoundArgs>(__bound_args)...);
+}
+
+template<class _Rp, class _Fp, class ..._BoundArgs>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__bind_r<_Rp, _Fp, _BoundArgs...>
+bind(_Fp&& __f, _BoundArgs&&... __bound_args)
+{
+    typedef __bind_r<_Rp, _Fp, _BoundArgs...> type;
+    return type(_CUDA_VSTD::forward<_Fp>(__f), _CUDA_VSTD::forward<_BoundArgs>(__bound_args)...);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#endif // __cuda_std__
+
+#if _LIBCUDACXX_STD_VER > 14
+
+template <class _Fn, class ..._Args>
+_LIBCUDACXX_INLINE_VISIBILITY
+invoke_result_t<_Fn, _Args...>
+invoke(_Fn&& __f, _Args&&... __args)
+    noexcept(is_nothrow_invocable_v<_Fn, _Args...>)
+{
+    return _CUDA_VSTD::__invoke(_CUDA_VSTD::forward<_Fn>(__f), _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _DecayFunc>
+class _LIBCUDACXX_TEMPLATE_VIS __not_fn_imp {
+  _DecayFunc __fd;
+
+public:
+    __not_fn_imp() = delete;
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Args&& ...__args) &
+            noexcept(noexcept(!_CUDA_VSTD::invoke(__fd, _CUDA_VSTD::forward<_Args>(__args)...)))
+        -> decltype(          !_CUDA_VSTD::invoke(__fd, _CUDA_VSTD::forward<_Args>(__args)...))
+        { return              !_CUDA_VSTD::invoke(__fd, _CUDA_VSTD::forward<_Args>(__args)...); }
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Args&& ...__args) &&
+            noexcept(noexcept(!_CUDA_VSTD::invoke(_CUDA_VSTD::move(__fd), _CUDA_VSTD::forward<_Args>(__args)...)))
+        -> decltype(          !_CUDA_VSTD::invoke(_CUDA_VSTD::move(__fd), _CUDA_VSTD::forward<_Args>(__args)...))
+        { return              !_CUDA_VSTD::invoke(_CUDA_VSTD::move(__fd), _CUDA_VSTD::forward<_Args>(__args)...); }
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Args&& ...__args) const&
+            noexcept(noexcept(!_CUDA_VSTD::invoke(__fd, _CUDA_VSTD::forward<_Args>(__args)...)))
+        -> decltype(          !_CUDA_VSTD::invoke(__fd, _CUDA_VSTD::forward<_Args>(__args)...))
+        { return              !_CUDA_VSTD::invoke(__fd, _CUDA_VSTD::forward<_Args>(__args)...); }
+
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    auto operator()(_Args&& ...__args) const&&
+            noexcept(noexcept(!_CUDA_VSTD::invoke(_CUDA_VSTD::move(__fd), _CUDA_VSTD::forward<_Args>(__args)...)))
+        -> decltype(          !_CUDA_VSTD::invoke(_CUDA_VSTD::move(__fd), _CUDA_VSTD::forward<_Args>(__args)...))
+        { return              !_CUDA_VSTD::invoke(_CUDA_VSTD::move(__fd), _CUDA_VSTD::forward<_Args>(__args)...); }
+
+private:
+    template <class _RawFunc,
+              class = enable_if_t<!is_same<decay_t<_RawFunc>, __not_fn_imp>::value>>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __not_fn_imp(_RawFunc&& __rf)
+        : __fd(_CUDA_VSTD::forward<_RawFunc>(__rf)) {}
+
+    template <class _RawFunc>
+    friend inline _LIBCUDACXX_INLINE_VISIBILITY
+    __not_fn_imp<decay_t<_RawFunc>> not_fn(_RawFunc&&);
+};
+
+template <class _RawFunc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__not_fn_imp<decay_t<_RawFunc>> not_fn(_RawFunc&& __fn) {
+    return __not_fn_imp<decay_t<_RawFunc>>(_CUDA_VSTD::forward<_RawFunc>(__fn));
+}
+
+#endif
+
+// struct hash<T*> in <memory>
+
+// Searchers are not currently supported.
+
+
+template <class _BinaryPredicate, class _ForwardIterator1, class _ForwardIterator2>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<_ForwardIterator1, _ForwardIterator1>
+__search(_ForwardIterator1 __first1, _ForwardIterator1 __last1,
+         _ForwardIterator2 __first2, _ForwardIterator2 __last2, _BinaryPredicate __pred,
+         forward_iterator_tag, forward_iterator_tag)
+{
+    if (__first2 == __last2)
+        return make_pair(__first1, __first1);  // Everything matches an empty sequence
+    while (true)
+    {
+        // Find first element in sequence 1 that matchs *__first2, with a mininum of loop checks
+        while (true)
+        {
+            if (__first1 == __last1)  // return __last1 if no element matches *__first2
+                return make_pair(__last1, __last1);
+            if (__pred(*__first1, *__first2))
+                break;
+            ++__first1;
+        }
+        // *__first1 matches *__first2, now match elements after here
+        _ForwardIterator1 __m1 = __first1;
+        _ForwardIterator2 __m2 = __first2;
+        while (true)
+        {
+            if (++__m2 == __last2)  // If pattern exhausted, __first1 is the answer (works for 1 element pattern)
+                return make_pair(__first1, __m1);
+            if (++__m1 == __last1)  // Otherwise if source exhaused, pattern not found
+                return make_pair(__last1, __last1);
+            if (!__pred(*__m1, *__m2))  // if there is a mismatch, restart with a new __first1
+            {
+                ++__first1;
+                break;
+            }  // else there is a match, check next elements
+        }
+    }
+}
+
+template <class _BinaryPredicate, class _RandomAccessIterator1, class _RandomAccessIterator2>
+_LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<_RandomAccessIterator1, _RandomAccessIterator1>
+__search(_RandomAccessIterator1 __first1, _RandomAccessIterator1 __last1,
+         _RandomAccessIterator2 __first2, _RandomAccessIterator2 __last2, _BinaryPredicate __pred,
+           random_access_iterator_tag, random_access_iterator_tag)
+{
+    typedef typename iterator_traits<_RandomAccessIterator1>::difference_type _D1;
+    typedef typename iterator_traits<_RandomAccessIterator2>::difference_type _D2;
+    // Take advantage of knowing source and pattern lengths.  Stop short when source is smaller than pattern
+    const _D2 __len2 = __last2 - __first2;
+    if (__len2 == 0)
+        return make_pair(__first1, __first1);
+    const _D1 __len1 = __last1 - __first1;
+    if (__len1 < __len2)
+        return make_pair(__last1, __last1);
+    const _RandomAccessIterator1 __s = __last1 - (__len2 - 1);  // Start of pattern match can't go beyond here
+
+    while (true)
+    {
+        while (true)
+        {
+            if (__first1 == __s)
+                return make_pair(__last1, __last1);
+            if (__pred(*__first1, *__first2))
+                break;
+            ++__first1;
+        }
+
+        _RandomAccessIterator1 __m1 = __first1;
+        _RandomAccessIterator2 __m2 = __first2;
+         while (true)
+         {
+             if (++__m2 == __last2)
+                 return make_pair(__first1, __first1 + __len2);
+             ++__m1;          // no need to check range on __m1 because __s guarantees we have enough source
+             if (!__pred(*__m1, *__m2))
+             {
+                 ++__first1;
+                 break;
+             }
+         }
+    }
+}
+
+#ifndef __cuda_std__
+#if _LIBCUDACXX_STD_VER > 14
+
+// default searcher
+template<class _ForwardIterator, class _BinaryPredicate = equal_to<>>
+class _LIBCUDACXX_TYPE_VIS default_searcher {
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    default_searcher(_ForwardIterator __f, _ForwardIterator __l,
+                       _BinaryPredicate __p = _BinaryPredicate())
+        : __first_(__f), __last_(__l), __pred_(__p) {}
+
+    template <typename _ForwardIterator2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<_ForwardIterator2, _ForwardIterator2>
+    operator () (_ForwardIterator2 __f, _ForwardIterator2 __l) const
+    {
+        return _CUDA_VSTD::__search(__f, __l, __first_, __last_, __pred_,
+            typename _CUDA_VSTD::iterator_traits<_ForwardIterator>::iterator_category(),
+            typename _CUDA_VSTD::iterator_traits<_ForwardIterator2>::iterator_category());
+    }
+
+private:
+    _ForwardIterator __first_;
+    _ForwardIterator __last_;
+    _BinaryPredicate __pred_;
+    };
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+#endif // __cuda_std__
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp>
+using unwrap_reference_t = typename unwrap_reference<_Tp>::type;
+
+template <class _Tp>
+using unwrap_ref_decay_t = typename unwrap_ref_decay<_Tp>::type;
+#endif // > C++17
+
+#ifndef __cuda_std__
+
+template <class _Container, class _Predicate>
+inline void __libcpp_erase_if_container( _Container& __c, _Predicate __pred)
+{
+	for (typename _Container::iterator __iter = __c.begin(), __last = __c.end(); __iter != __last;)
+	{
+		if (__pred(*__iter))
+			__iter = __c.erase(__iter);
+		else
+			++__iter;
+	}
+}
+
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_FUNCTIONAL
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/future b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/future
new file mode 100644
index 000000000000..df4dcca4eb69
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/future
@@ -0,0 +1,2608 @@
+// -*- C++ -*-
+//===--------------------------- future -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_FUTURE
+#define _LIBCUDACXX_FUTURE
+
+/*
+    future synopsis
+
+namespace std
+{
+
+enum class future_errc
+{
+    future_already_retrieved = 1,
+    promise_already_satisfied,
+    no_state,
+    broken_promise
+};
+
+enum class launch
+{
+    async = 1,
+    deferred = 2,
+    any = async | deferred
+};
+
+enum class future_status
+{
+    ready,
+    timeout,
+    deferred
+};
+
+template <> struct is_error_code_enum<future_errc> : public true_type { };
+error_code make_error_code(future_errc e) noexcept;
+error_condition make_error_condition(future_errc e) noexcept;
+
+const error_category& future_category() noexcept;
+
+class future_error
+    : public logic_error
+{
+public:
+    future_error(error_code ec);  // exposition only
+    explicit future_error(future_errc); // C++17
+    const error_code& code() const noexcept;
+    const char*       what() const noexcept;
+};
+
+template <class R>
+class promise
+{
+public:
+    promise();
+    template <class Allocator>
+        promise(allocator_arg_t, const Allocator& a);
+    promise(promise&& rhs) noexcept;
+    promise(const promise& rhs) = delete;
+    ~promise();
+
+    // assignment
+    promise& operator=(promise&& rhs) noexcept;
+    promise& operator=(const promise& rhs) = delete;
+    void swap(promise& other) noexcept;
+
+    // retrieving the result
+    future<R> get_future();
+
+    // setting the result
+    void set_value(const R& r);
+    void set_value(R&& r);
+    void set_exception(exception_ptr p);
+
+    // setting the result with deferred notification
+    void set_value_at_thread_exit(const R& r);
+    void set_value_at_thread_exit(R&& r);
+    void set_exception_at_thread_exit(exception_ptr p);
+};
+
+template <class R>
+class promise<R&>
+{
+public:
+    promise();
+    template <class Allocator>
+        promise(allocator_arg_t, const Allocator& a);
+    promise(promise&& rhs) noexcept;
+    promise(const promise& rhs) = delete;
+    ~promise();
+
+    // assignment
+    promise& operator=(promise&& rhs) noexcept;
+    promise& operator=(const promise& rhs) = delete;
+    void swap(promise& other) noexcept;
+
+    // retrieving the result
+    future<R&> get_future();
+
+    // setting the result
+    void set_value(R& r);
+    void set_exception(exception_ptr p);
+
+    // setting the result with deferred notification
+    void set_value_at_thread_exit(R&);
+    void set_exception_at_thread_exit(exception_ptr p);
+};
+
+template <>
+class promise<void>
+{
+public:
+    promise();
+    template <class Allocator>
+        promise(allocator_arg_t, const Allocator& a);
+    promise(promise&& rhs) noexcept;
+    promise(const promise& rhs) = delete;
+    ~promise();
+
+    // assignment
+    promise& operator=(promise&& rhs) noexcept;
+    promise& operator=(const promise& rhs) = delete;
+    void swap(promise& other) noexcept;
+
+    // retrieving the result
+    future<void> get_future();
+
+    // setting the result
+    void set_value();
+    void set_exception(exception_ptr p);
+
+    // setting the result with deferred notification
+    void set_value_at_thread_exit();
+    void set_exception_at_thread_exit(exception_ptr p);
+};
+
+template <class R> void swap(promise<R>& x, promise<R>& y) noexcept;
+
+template <class R, class Alloc>
+    struct uses_allocator<promise<R>, Alloc> : public true_type {};
+
+template <class R>
+class future
+{
+public:
+    future() noexcept;
+    future(future&&) noexcept;
+    future(const future& rhs) = delete;
+    ~future();
+    future& operator=(const future& rhs) = delete;
+    future& operator=(future&&) noexcept;
+    shared_future<R> share() noexcept;
+
+    // retrieving the value
+    R get();
+
+    // functions to check state
+    bool valid() const noexcept;
+
+    void wait() const;
+    template <class Rep, class Period>
+        future_status
+        wait_for(const chrono::duration<Rep, Period>& rel_time) const;
+    template <class Clock, class Duration>
+        future_status
+        wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
+};
+
+template <class R>
+class future<R&>
+{
+public:
+    future() noexcept;
+    future(future&&) noexcept;
+    future(const future& rhs) = delete;
+    ~future();
+    future& operator=(const future& rhs) = delete;
+    future& operator=(future&&) noexcept;
+    shared_future<R&> share() noexcept;
+
+    // retrieving the value
+    R& get();
+
+    // functions to check state
+    bool valid() const noexcept;
+
+    void wait() const;
+    template <class Rep, class Period>
+        future_status
+        wait_for(const chrono::duration<Rep, Period>& rel_time) const;
+    template <class Clock, class Duration>
+        future_status
+        wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
+};
+
+template <>
+class future<void>
+{
+public:
+    future() noexcept;
+    future(future&&) noexcept;
+    future(const future& rhs) = delete;
+    ~future();
+    future& operator=(const future& rhs) = delete;
+    future& operator=(future&&) noexcept;
+    shared_future<void> share() noexcept;
+
+    // retrieving the value
+    void get();
+
+    // functions to check state
+    bool valid() const noexcept;
+
+    void wait() const;
+    template <class Rep, class Period>
+        future_status
+        wait_for(const chrono::duration<Rep, Period>& rel_time) const;
+    template <class Clock, class Duration>
+        future_status
+        wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
+};
+
+template <class R>
+class shared_future
+{
+public:
+    shared_future() noexcept;
+    shared_future(const shared_future& rhs);
+    shared_future(future<R>&&) noexcept;
+    shared_future(shared_future&& rhs) noexcept;
+    ~shared_future();
+    shared_future& operator=(const shared_future& rhs);
+    shared_future& operator=(shared_future&& rhs) noexcept;
+
+    // retrieving the value
+    const R& get() const;
+
+    // functions to check state
+    bool valid() const noexcept;
+
+    void wait() const;
+    template <class Rep, class Period>
+        future_status
+        wait_for(const chrono::duration<Rep, Period>& rel_time) const;
+    template <class Clock, class Duration>
+        future_status
+        wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
+};
+
+template <class R>
+class shared_future<R&>
+{
+public:
+    shared_future() noexcept;
+    shared_future(const shared_future& rhs);
+    shared_future(future<R&>&&) noexcept;
+    shared_future(shared_future&& rhs) noexcept;
+    ~shared_future();
+    shared_future& operator=(const shared_future& rhs);
+    shared_future& operator=(shared_future&& rhs) noexcept;
+
+    // retrieving the value
+    R& get() const;
+
+    // functions to check state
+    bool valid() const noexcept;
+
+    void wait() const;
+    template <class Rep, class Period>
+        future_status
+        wait_for(const chrono::duration<Rep, Period>& rel_time) const;
+    template <class Clock, class Duration>
+        future_status
+        wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
+};
+
+template <>
+class shared_future<void>
+{
+public:
+    shared_future() noexcept;
+    shared_future(const shared_future& rhs);
+    shared_future(future<void>&&) noexcept;
+    shared_future(shared_future&& rhs) noexcept;
+    ~shared_future();
+    shared_future& operator=(const shared_future& rhs);
+    shared_future& operator=(shared_future&& rhs) noexcept;
+
+    // retrieving the value
+    void get() const;
+
+    // functions to check state
+    bool valid() const noexcept;
+
+    void wait() const;
+    template <class Rep, class Period>
+        future_status
+        wait_for(const chrono::duration<Rep, Period>& rel_time) const;
+    template <class Clock, class Duration>
+        future_status
+        wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
+};
+
+template <class F, class... Args>
+  future<typename result_of<typename decay<F>::type(typename decay<Args>::type...)>::type>
+  async(F&& f, Args&&... args);
+
+template <class F, class... Args>
+  future<typename result_of<typename decay<F>::type(typename decay<Args>::type...)>::type>
+  async(launch policy, F&& f, Args&&... args);
+
+template <class> class packaged_task; // undefined
+
+template <class R, class... ArgTypes>
+class packaged_task<R(ArgTypes...)>
+{
+public:
+    typedef R result_type; // extension
+
+    // construction and destruction
+    packaged_task() noexcept;
+    template <class F>
+        explicit packaged_task(F&& f);
+    template <class F, class Allocator>
+        packaged_task(allocator_arg_t, const Allocator& a, F&& f);
+    ~packaged_task();
+
+    // no copy
+    packaged_task(const packaged_task&) = delete;
+    packaged_task& operator=(const packaged_task&) = delete;
+
+    // move support
+    packaged_task(packaged_task&& other) noexcept;
+    packaged_task& operator=(packaged_task&& other) noexcept;
+    void swap(packaged_task& other) noexcept;
+
+    bool valid() const noexcept;
+
+    // result retrieval
+    future<R> get_future();
+
+    // execution
+    void operator()(ArgTypes... );
+    void make_ready_at_thread_exit(ArgTypes...);
+
+    void reset();
+};
+
+template <class R>
+  void swap(packaged_task<R(ArgTypes...)&, packaged_task<R(ArgTypes...)>&) noexcept;
+
+template <class R, class Alloc> struct uses_allocator<packaged_task<R>, Alloc>;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <system_error>
+#include <memory>
+#include <chrono>
+#include <exception>
+#include <mutex>
+#include <thread>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+#error <future> is not supported on this single threaded system
+#else // !_LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+//enum class future_errc
+_LIBCUDACXX_DECLARE_STRONG_ENUM(future_errc)
+{
+    future_already_retrieved = 1,
+    promise_already_satisfied,
+    no_state,
+    broken_promise
+};
+_LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(future_errc)
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_code_enum<future_errc> : public true_type {};
+
+#ifdef _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_code_enum<future_errc::__lx> : public true_type { };
+#endif
+
+//enum class launch
+_LIBCUDACXX_DECLARE_STRONG_ENUM(launch)
+{
+    async = 1,
+    deferred = 2,
+    any = async | deferred
+};
+_LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(launch)
+
+#ifndef _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+
+typedef underlying_type<launch>::type __launch_underlying_type;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+launch
+operator&(launch __x, launch __y)
+{
+    return static_cast<launch>(static_cast<__launch_underlying_type>(__x) &
+                               static_cast<__launch_underlying_type>(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+launch
+operator|(launch __x, launch __y)
+{
+    return static_cast<launch>(static_cast<__launch_underlying_type>(__x) |
+                               static_cast<__launch_underlying_type>(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+launch
+operator^(launch __x, launch __y)
+{
+    return static_cast<launch>(static_cast<__launch_underlying_type>(__x) ^
+                               static_cast<__launch_underlying_type>(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+launch
+operator~(launch __x)
+{
+    return static_cast<launch>(~static_cast<__launch_underlying_type>(__x) & 3);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+launch&
+operator&=(launch& __x, launch __y)
+{
+    __x = __x & __y; return __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+launch&
+operator|=(launch& __x, launch __y)
+{
+    __x = __x | __y; return __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+launch&
+operator^=(launch& __x, launch __y)
+{
+    __x = __x ^ __y; return __x;
+}
+
+#endif  // !_LIBCUDACXX_HAS_NO_STRONG_ENUMS
+
+//enum class future_status
+_LIBCUDACXX_DECLARE_STRONG_ENUM(future_status)
+{
+    ready,
+    timeout,
+    deferred
+};
+_LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(future_status)
+
+_LIBCUDACXX_FUNC_VIS
+const error_category& future_category() _NOEXCEPT;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+error_code
+make_error_code(future_errc __e) _NOEXCEPT
+{
+    return error_code(static_cast<int>(__e), future_category());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+error_condition
+make_error_condition(future_errc __e) _NOEXCEPT
+{
+    return error_condition(static_cast<int>(__e), future_category());
+}
+
+class _LIBCUDACXX_EXCEPTION_ABI _LIBCUDACXX_AVAILABILITY_FUTURE_ERROR future_error
+    : public logic_error
+{
+    error_code __ec_;
+public:
+    future_error(error_code __ec);
+#if _LIBCUDACXX_STD_VERS > 14
+    explicit future_error(future_errc _Ev) : logic_error(), __ec_(make_error_code(_Ev)) {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const error_code& code() const _NOEXCEPT {return __ec_;}
+
+    virtual ~future_error() _NOEXCEPT;
+};
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+_LIBCUDACXX_AVAILABILITY_FUTURE_ERROR
+#endif
+void __throw_future_error(future_errc _Ev)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw future_error(make_error_code(_Ev));
+#else
+    ((void)_Ev);
+    _CUDA_VSTD::abort();
+#endif
+}
+
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE __assoc_sub_state
+    : public __shared_count
+{
+protected:
+    exception_ptr __exception_;
+    mutable mutex __mut_;
+    mutable condition_variable __cv_;
+    unsigned __state_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+    void __sub_wait(unique_lock<mutex>& __lk);
+public:
+    enum
+    {
+        __constructed = 1,
+        __future_attached = 2,
+        ready = 4,
+        deferred = 8
+    };
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __assoc_sub_state() : __state_(0) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __has_value() const
+        {return (__state_ & __constructed) || (__exception_ != nullptr);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __attach_future() {
+        lock_guard<mutex> __lk(__mut_);
+        bool __has_future_attached = (__state_ & __future_attached) != 0;
+        if (__has_future_attached)
+            __throw_future_error(future_errc::future_already_retrieved);
+        this->__add_shared();
+        __state_ |= __future_attached;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_deferred() {__state_ |= deferred;}
+
+    void __make_ready();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __is_ready() const {return (__state_ & ready) != 0;}
+
+    void set_value();
+    void set_value_at_thread_exit();
+
+    void set_exception(exception_ptr __p);
+    void set_exception_at_thread_exit(exception_ptr __p);
+
+    void copy();
+
+    void wait();
+    template <class _Rep, class _Period>
+        future_status
+        _LIBCUDACXX_INLINE_VISIBILITY
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const;
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const;
+
+    virtual void __execute();
+};
+
+template <class _Clock, class _Duration>
+future_status
+__assoc_sub_state::wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+{
+    unique_lock<mutex> __lk(__mut_);
+    if (__state_ & deferred)
+        return future_status::deferred;
+    while (!(__state_ & ready) && _Clock::now() < __abs_time)
+        __cv_.wait_until(__lk, __abs_time);
+    if (__state_ & ready)
+        return future_status::ready;
+    return future_status::timeout;
+}
+
+template <class _Rep, class _Period>
+inline
+future_status
+__assoc_sub_state::wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+{
+    return wait_until(chrono::steady_clock::now() + __rel_time);
+}
+
+template <class _Rp>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __assoc_state
+    : public __assoc_sub_state
+{
+    typedef __assoc_sub_state base;
+    typedef typename aligned_storage<sizeof(_Rp), alignment_of<_Rp>::value>::type _Up;
+protected:
+    _Up __value_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+
+    template <class _Arg>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+        void set_value(_Arg&& __arg);
+#else
+        void set_value(_Arg& __arg);
+#endif
+
+    template <class _Arg>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+        void set_value_at_thread_exit(_Arg&& __arg);
+#else
+        void set_value_at_thread_exit(_Arg& __arg);
+#endif
+
+    _Rp move();
+    typename add_lvalue_reference<_Rp>::type copy();
+};
+
+template <class _Rp>
+void
+__assoc_state<_Rp>::__on_zero_shared() _NOEXCEPT
+{
+    if (this->__state_ & base::__constructed)
+        reinterpret_cast<_Rp*>(&__value_)->~_Rp();
+    delete this;
+}
+
+template <class _Rp>
+template <class _Arg>
+_LIBCUDACXX_AVAILABILITY_FUTURE
+void
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+__assoc_state<_Rp>::set_value(_Arg&& __arg)
+#else
+__assoc_state<_Rp>::set_value(_Arg& __arg)
+#endif
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    if (this->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+    ::new(&__value_) _Rp(_CUDA_VSTD::forward<_Arg>(__arg));
+    this->__state_ |= base::__constructed | base::ready;
+    __cv_.notify_all();
+}
+
+template <class _Rp>
+template <class _Arg>
+void
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+__assoc_state<_Rp>::set_value_at_thread_exit(_Arg&& __arg)
+#else
+__assoc_state<_Rp>::set_value_at_thread_exit(_Arg& __arg)
+#endif
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    if (this->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+    ::new(&__value_) _Rp(_CUDA_VSTD::forward<_Arg>(__arg));
+    this->__state_ |= base::__constructed;
+    __thread_local_data()->__make_ready_at_thread_exit(this);
+}
+
+template <class _Rp>
+_Rp
+__assoc_state<_Rp>::move()
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    this->__sub_wait(__lk);
+    if (this->__exception_ != nullptr)
+        rethrow_exception(this->__exception_);
+    return _CUDA_VSTD::move(*reinterpret_cast<_Rp*>(&__value_));
+}
+
+template <class _Rp>
+typename add_lvalue_reference<_Rp>::type
+__assoc_state<_Rp>::copy()
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    this->__sub_wait(__lk);
+    if (this->__exception_ != nullptr)
+        rethrow_exception(this->__exception_);
+    return *reinterpret_cast<_Rp*>(&__value_);
+}
+
+template <class _Rp>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __assoc_state<_Rp&>
+    : public __assoc_sub_state
+{
+    typedef __assoc_sub_state base;
+    typedef _Rp* _Up;
+protected:
+    _Up __value_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+
+    void set_value(_Rp& __arg);
+    void set_value_at_thread_exit(_Rp& __arg);
+
+    _Rp& copy();
+};
+
+template <class _Rp>
+void
+__assoc_state<_Rp&>::__on_zero_shared() _NOEXCEPT
+{
+    delete this;
+}
+
+template <class _Rp>
+void
+__assoc_state<_Rp&>::set_value(_Rp& __arg)
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    if (this->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+    __value_ = _CUDA_VSTD::addressof(__arg);
+    this->__state_ |= base::__constructed | base::ready;
+    __cv_.notify_all();
+}
+
+template <class _Rp>
+void
+__assoc_state<_Rp&>::set_value_at_thread_exit(_Rp& __arg)
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    if (this->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+    __value_ = _CUDA_VSTD::addressof(__arg);
+    this->__state_ |= base::__constructed;
+    __thread_local_data()->__make_ready_at_thread_exit(this);
+}
+
+template <class _Rp>
+_Rp&
+__assoc_state<_Rp&>::copy()
+{
+    unique_lock<mutex> __lk(this->__mut_);
+    this->__sub_wait(__lk);
+    if (this->__exception_ != nullptr)
+        rethrow_exception(this->__exception_);
+    return *__value_;
+}
+
+template <class _Rp, class _Alloc>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __assoc_state_alloc
+    : public __assoc_state<_Rp>
+{
+    typedef __assoc_state<_Rp> base;
+    _Alloc __alloc_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __assoc_state_alloc(const _Alloc& __a)
+        : __alloc_(__a) {}
+};
+
+template <class _Rp, class _Alloc>
+void
+__assoc_state_alloc<_Rp, _Alloc>::__on_zero_shared() _NOEXCEPT
+{
+    if (this->__state_ & base::__constructed)
+        reinterpret_cast<_Rp*>(_CUDA_VSTD::addressof(this->__value_))->~_Rp();
+    typedef typename __allocator_traits_rebind<_Alloc, __assoc_state_alloc>::type _Al;
+    typedef allocator_traits<_Al> _ATraits;
+    typedef pointer_traits<typename _ATraits::pointer> _PTraits;
+    _Al __a(__alloc_);
+    this->~__assoc_state_alloc();
+    __a.deallocate(_PTraits::pointer_to(*this), 1);
+}
+
+template <class _Rp, class _Alloc>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __assoc_state_alloc<_Rp&, _Alloc>
+    : public __assoc_state<_Rp&>
+{
+    typedef __assoc_state<_Rp&> base;
+    _Alloc __alloc_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __assoc_state_alloc(const _Alloc& __a)
+        : __alloc_(__a) {}
+};
+
+template <class _Rp, class _Alloc>
+void
+__assoc_state_alloc<_Rp&, _Alloc>::__on_zero_shared() _NOEXCEPT
+{
+    typedef typename __allocator_traits_rebind<_Alloc, __assoc_state_alloc>::type _Al;
+    typedef allocator_traits<_Al> _ATraits;
+    typedef pointer_traits<typename _ATraits::pointer> _PTraits;
+    _Al __a(__alloc_);
+    this->~__assoc_state_alloc();
+    __a.deallocate(_PTraits::pointer_to(*this), 1);
+}
+
+template <class _Alloc>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __assoc_sub_state_alloc
+    : public __assoc_sub_state
+{
+    typedef __assoc_sub_state base;
+    _Alloc __alloc_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __assoc_sub_state_alloc(const _Alloc& __a)
+        : __alloc_(__a) {}
+};
+
+template <class _Alloc>
+void
+__assoc_sub_state_alloc<_Alloc>::__on_zero_shared() _NOEXCEPT
+{
+    typedef typename __allocator_traits_rebind<_Alloc, __assoc_sub_state_alloc>::type _Al;
+    typedef allocator_traits<_Al> _ATraits;
+    typedef pointer_traits<typename _ATraits::pointer> _PTraits;
+    _Al __a(__alloc_);
+    this->~__assoc_sub_state_alloc();
+    __a.deallocate(_PTraits::pointer_to(*this), 1);
+}
+
+template <class _Rp, class _Fp>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __deferred_assoc_state
+    : public __assoc_state<_Rp>
+{
+    typedef __assoc_state<_Rp> base;
+
+    _Fp __func_;
+
+public:
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __deferred_assoc_state(_Fp&& __f);
+#endif
+
+    virtual void __execute();
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp, class _Fp>
+inline
+__deferred_assoc_state<_Rp, _Fp>::__deferred_assoc_state(_Fp&& __f)
+    : __func_(_CUDA_VSTD::forward<_Fp>(__f))
+{
+    this->__set_deferred();
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp, class _Fp>
+void
+__deferred_assoc_state<_Rp, _Fp>::__execute()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->set_value(__func_());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->set_exception(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _Fp>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __deferred_assoc_state<void, _Fp>
+    : public __assoc_sub_state
+{
+    typedef __assoc_sub_state base;
+
+    _Fp __func_;
+
+public:
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __deferred_assoc_state(_Fp&& __f);
+#endif
+
+    virtual void __execute();
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Fp>
+inline
+__deferred_assoc_state<void, _Fp>::__deferred_assoc_state(_Fp&& __f)
+    : __func_(_CUDA_VSTD::forward<_Fp>(__f))
+{
+    this->__set_deferred();
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Fp>
+void
+__deferred_assoc_state<void, _Fp>::__execute()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __func_();
+        this->set_value();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->set_exception(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _Rp, class _Fp>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __async_assoc_state
+    : public __assoc_state<_Rp>
+{
+    typedef __assoc_state<_Rp> base;
+
+    _Fp __func_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __async_assoc_state(_Fp&& __f);
+#endif
+
+    virtual void __execute();
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp, class _Fp>
+inline
+__async_assoc_state<_Rp, _Fp>::__async_assoc_state(_Fp&& __f)
+    : __func_(_CUDA_VSTD::forward<_Fp>(__f))
+{
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp, class _Fp>
+void
+__async_assoc_state<_Rp, _Fp>::__execute()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->set_value(__func_());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->set_exception(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _Rp, class _Fp>
+void
+__async_assoc_state<_Rp, _Fp>::__on_zero_shared() _NOEXCEPT
+{
+    this->wait();
+    base::__on_zero_shared();
+}
+
+template <class _Fp>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __async_assoc_state<void, _Fp>
+    : public __assoc_sub_state
+{
+    typedef __assoc_sub_state base;
+
+    _Fp __func_;
+
+    virtual void __on_zero_shared() _NOEXCEPT;
+public:
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __async_assoc_state(_Fp&& __f);
+#endif
+
+    virtual void __execute();
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Fp>
+inline
+__async_assoc_state<void, _Fp>::__async_assoc_state(_Fp&& __f)
+    : __func_(_CUDA_VSTD::forward<_Fp>(__f))
+{
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Fp>
+void
+__async_assoc_state<void, _Fp>::__execute()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __func_();
+        this->set_value();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->set_exception(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _Fp>
+void
+__async_assoc_state<void, _Fp>::__on_zero_shared() _NOEXCEPT
+{
+    this->wait();
+    base::__on_zero_shared();
+}
+
+template <class _Rp> class _LIBCUDACXX_TEMPLATE_VIS promise;
+template <class _Rp> class _LIBCUDACXX_TEMPLATE_VIS shared_future;
+
+// future
+
+template <class _Rp> class _LIBCUDACXX_TEMPLATE_VIS future;
+
+template <class _Rp, class _Fp>
+future<_Rp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+__make_deferred_assoc_state(_Fp&& __f);
+#else
+__make_deferred_assoc_state(_Fp __f);
+#endif
+
+template <class _Rp, class _Fp>
+future<_Rp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+__make_async_assoc_state(_Fp&& __f);
+#else
+__make_async_assoc_state(_Fp __f);
+#endif
+
+template <class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE future
+{
+    __assoc_state<_Rp>* __state_;
+
+    explicit future(__assoc_state<_Rp>* __state);
+
+    template <class> friend class promise;
+    template <class> friend class shared_future;
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_deferred_assoc_state(_Fp&& __f);
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_async_assoc_state(_Fp&& __f);
+#else
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_deferred_assoc_state(_Fp __f);
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_async_assoc_state(_Fp __f);
+#endif
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future() _NOEXCEPT : __state_(nullptr) {}
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future(future&& __rhs) _NOEXCEPT
+        : __state_(__rhs.__state_) {__rhs.__state_ = nullptr;}
+    future(const future&) = delete;
+    future& operator=(const future&) = delete;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future& operator=(future&& __rhs) _NOEXCEPT
+        {
+            future(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    future(const future&);
+    future& operator=(const future&);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~future();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future<_Rp> share() _NOEXCEPT;
+
+    // retrieving the value
+    _Rp get();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(future& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // functions to check state
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __state_ != nullptr;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const {__state_->wait();}
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+            {return __state_->wait_for(__rel_time);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+            {return __state_->wait_until(__abs_time);}
+};
+
+template <class _Rp>
+future<_Rp>::future(__assoc_state<_Rp>* __state)
+    : __state_(__state)
+{
+    __state_->__attach_future();
+}
+
+struct __release_shared_count
+{
+    void operator()(__shared_count* p) {p->__release_shared();}
+};
+
+template <class _Rp>
+future<_Rp>::~future()
+{
+    if (__state_)
+        __state_->__release_shared();
+}
+
+template <class _Rp>
+_Rp
+future<_Rp>::get()
+{
+    unique_ptr<__shared_count, __release_shared_count> __(__state_);
+    __assoc_state<_Rp>* __s = __state_;
+    __state_ = nullptr;
+    return __s->move();
+}
+
+template <class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE future<_Rp&>
+{
+    __assoc_state<_Rp&>* __state_;
+
+    explicit future(__assoc_state<_Rp&>* __state);
+
+    template <class> friend class promise;
+    template <class> friend class shared_future;
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_deferred_assoc_state(_Fp&& __f);
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_async_assoc_state(_Fp&& __f);
+#else
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_deferred_assoc_state(_Fp __f);
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_async_assoc_state(_Fp __f);
+#endif
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future() _NOEXCEPT : __state_(nullptr) {}
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future(future&& __rhs) _NOEXCEPT
+        : __state_(__rhs.__state_) {__rhs.__state_ = nullptr;}
+    future(const future&) = delete;
+    future& operator=(const future&) = delete;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future& operator=(future&& __rhs) _NOEXCEPT
+        {
+            future(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    future(const future&);
+    future& operator=(const future&);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~future();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future<_Rp&> share() _NOEXCEPT;
+
+    // retrieving the value
+    _Rp& get();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(future& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // functions to check state
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __state_ != nullptr;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const {__state_->wait();}
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+            {return __state_->wait_for(__rel_time);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+            {return __state_->wait_until(__abs_time);}
+};
+
+template <class _Rp>
+future<_Rp&>::future(__assoc_state<_Rp&>* __state)
+    : __state_(__state)
+{
+    __state_->__attach_future();
+}
+
+template <class _Rp>
+future<_Rp&>::~future()
+{
+    if (__state_)
+        __state_->__release_shared();
+}
+
+template <class _Rp>
+_Rp&
+future<_Rp&>::get()
+{
+    unique_ptr<__shared_count, __release_shared_count> __(__state_);
+    __assoc_state<_Rp&>* __s = __state_;
+    __state_ = nullptr;
+    return __s->copy();
+}
+
+template <>
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE future<void>
+{
+    __assoc_sub_state* __state_;
+
+    explicit future(__assoc_sub_state* __state);
+
+    template <class> friend class promise;
+    template <class> friend class shared_future;
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_deferred_assoc_state(_Fp&& __f);
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_async_assoc_state(_Fp&& __f);
+#else
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_deferred_assoc_state(_Fp __f);
+    template <class _R1, class _Fp>
+        friend future<_R1> __make_async_assoc_state(_Fp __f);
+#endif
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future() _NOEXCEPT : __state_(nullptr) {}
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future(future&& __rhs) _NOEXCEPT
+        : __state_(__rhs.__state_) {__rhs.__state_ = nullptr;}
+    future(const future&) = delete;
+    future& operator=(const future&) = delete;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future& operator=(future&& __rhs) _NOEXCEPT
+        {
+            future(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    future(const future&);
+    future& operator=(const future&);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~future();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future<void> share() _NOEXCEPT;
+
+    // retrieving the value
+    void get();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(future& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // functions to check state
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __state_ != nullptr;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const {__state_->wait();}
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+            {return __state_->wait_for(__rel_time);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+            {return __state_->wait_until(__abs_time);}
+};
+
+template <class _Rp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(future<_Rp>& __x, future<_Rp>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+// promise<R>
+
+template <class _Callable> class packaged_task;
+
+template <class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE promise
+{
+    __assoc_state<_Rp>* __state_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit promise(nullptr_t) _NOEXCEPT : __state_(nullptr) {}
+
+    template <class> friend class packaged_task;
+public:
+    promise();
+    template <class _Alloc>
+        promise(allocator_arg_t, const _Alloc& __a);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    promise(promise&& __rhs) _NOEXCEPT
+        : __state_(__rhs.__state_) {__rhs.__state_ = nullptr;}
+    promise(const promise& __rhs) = delete;
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    promise(const promise& __rhs);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~promise();
+
+    // assignment
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    promise& operator=(promise&& __rhs) _NOEXCEPT
+        {
+            promise(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+    promise& operator=(const promise& __rhs) = delete;
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    promise& operator=(const promise& __rhs);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(promise& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // retrieving the result
+    future<_Rp> get_future();
+
+    // setting the result
+    void set_value(const _Rp& __r);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    void set_value(_Rp&& __r);
+#endif
+    void set_exception(exception_ptr __p);
+
+    // setting the result with deferred notification
+    void set_value_at_thread_exit(const _Rp& __r);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    void set_value_at_thread_exit(_Rp&& __r);
+#endif
+    void set_exception_at_thread_exit(exception_ptr __p);
+};
+
+template <class _Rp>
+promise<_Rp>::promise()
+    : __state_(new __assoc_state<_Rp>)
+{
+}
+
+template <class _Rp>
+template <class _Alloc>
+promise<_Rp>::promise(allocator_arg_t, const _Alloc& __a0)
+{
+    typedef __assoc_state_alloc<_Rp, _Alloc> _State;
+    typedef typename __allocator_traits_rebind<_Alloc, _State>::type _A2;
+    typedef __allocator_destructor<_A2> _D2;
+    _A2 __a(__a0);
+    unique_ptr<_State, _D2> __hold(__a.allocate(1), _D2(__a, 1));
+    ::new(static_cast<void*>(_CUDA_VSTD::addressof(*__hold.get()))) _State(__a0);
+    __state_ = _CUDA_VSTD::addressof(*__hold.release());
+}
+
+template <class _Rp>
+promise<_Rp>::~promise()
+{
+    if (__state_)
+    {
+        if (!__state_->__has_value() && __state_->use_count() > 1)
+            __state_->set_exception(make_exception_ptr(
+                      future_error(make_error_code(future_errc::broken_promise))
+                                                      ));
+        __state_->__release_shared();
+    }
+}
+
+template <class _Rp>
+future<_Rp>
+promise<_Rp>::get_future()
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    return future<_Rp>(__state_);
+}
+
+template <class _Rp>
+void
+promise<_Rp>::set_value(const _Rp& __r)
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_value(__r);
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp>
+void
+promise<_Rp>::set_value(_Rp&& __r)
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_value(_CUDA_VSTD::move(__r));
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp>
+void
+promise<_Rp>::set_exception(exception_ptr __p)
+{
+    _LIBCUDACXX_ASSERT( __p != nullptr, "promise::set_exception: received nullptr" );
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_exception(__p);
+}
+
+template <class _Rp>
+void
+promise<_Rp>::set_value_at_thread_exit(const _Rp& __r)
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_value_at_thread_exit(__r);
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp>
+void
+promise<_Rp>::set_value_at_thread_exit(_Rp&& __r)
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_value_at_thread_exit(_CUDA_VSTD::move(__r));
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Rp>
+void
+promise<_Rp>::set_exception_at_thread_exit(exception_ptr __p)
+{
+    _LIBCUDACXX_ASSERT( __p != nullptr, "promise::set_exception_at_thread_exit: received nullptr" );
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_exception_at_thread_exit(__p);
+}
+
+// promise<R&>
+
+template <class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE promise<_Rp&>
+{
+    __assoc_state<_Rp&>* __state_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit promise(nullptr_t) _NOEXCEPT : __state_(nullptr) {}
+
+    template <class> friend class packaged_task;
+
+public:
+    promise();
+    template <class _Allocator>
+        promise(allocator_arg_t, const _Allocator& __a);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    promise(promise&& __rhs) _NOEXCEPT
+        : __state_(__rhs.__state_) {__rhs.__state_ = nullptr;}
+    promise(const promise& __rhs) = delete;
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    promise(const promise& __rhs);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~promise();
+
+    // assignment
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    promise& operator=(promise&& __rhs) _NOEXCEPT
+        {
+            promise(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+    promise& operator=(const promise& __rhs) = delete;
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    promise& operator=(const promise& __rhs);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(promise& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // retrieving the result
+    future<_Rp&> get_future();
+
+    // setting the result
+    void set_value(_Rp& __r);
+    void set_exception(exception_ptr __p);
+
+    // setting the result with deferred notification
+    void set_value_at_thread_exit(_Rp&);
+    void set_exception_at_thread_exit(exception_ptr __p);
+};
+
+template <class _Rp>
+promise<_Rp&>::promise()
+    : __state_(new __assoc_state<_Rp&>)
+{
+}
+
+template <class _Rp>
+template <class _Alloc>
+promise<_Rp&>::promise(allocator_arg_t, const _Alloc& __a0)
+{
+    typedef __assoc_state_alloc<_Rp&, _Alloc> _State;
+    typedef typename __allocator_traits_rebind<_Alloc, _State>::type _A2;
+    typedef __allocator_destructor<_A2> _D2;
+    _A2 __a(__a0);
+    unique_ptr<_State, _D2> __hold(__a.allocate(1), _D2(__a, 1));
+    ::new(static_cast<void*>(_CUDA_VSTD::addressof(*__hold.get()))) _State(__a0);
+    __state_ = _CUDA_VSTD::addressof(*__hold.release());
+}
+
+template <class _Rp>
+promise<_Rp&>::~promise()
+{
+    if (__state_)
+    {
+        if (!__state_->__has_value() && __state_->use_count() > 1)
+            __state_->set_exception(make_exception_ptr(
+                      future_error(make_error_code(future_errc::broken_promise))
+                                                      ));
+        __state_->__release_shared();
+    }
+}
+
+template <class _Rp>
+future<_Rp&>
+promise<_Rp&>::get_future()
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    return future<_Rp&>(__state_);
+}
+
+template <class _Rp>
+void
+promise<_Rp&>::set_value(_Rp& __r)
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_value(__r);
+}
+
+template <class _Rp>
+void
+promise<_Rp&>::set_exception(exception_ptr __p)
+{
+    _LIBCUDACXX_ASSERT( __p != nullptr, "promise::set_exception: received nullptr" );
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_exception(__p);
+}
+
+template <class _Rp>
+void
+promise<_Rp&>::set_value_at_thread_exit(_Rp& __r)
+{
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_value_at_thread_exit(__r);
+}
+
+template <class _Rp>
+void
+promise<_Rp&>::set_exception_at_thread_exit(exception_ptr __p)
+{
+    _LIBCUDACXX_ASSERT( __p != nullptr, "promise::set_exception_at_thread_exit: received nullptr" );
+    if (__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    __state_->set_exception_at_thread_exit(__p);
+}
+
+// promise<void>
+
+template <>
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE promise<void>
+{
+    __assoc_sub_state* __state_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit promise(nullptr_t) _NOEXCEPT : __state_(nullptr) {}
+
+    template <class> friend class packaged_task;
+
+public:
+    promise();
+    template <class _Allocator>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        promise(allocator_arg_t, const _Allocator& __a);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    promise(promise&& __rhs) _NOEXCEPT
+        : __state_(__rhs.__state_) {__rhs.__state_ = nullptr;}
+    promise(const promise& __rhs) = delete;
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    promise(const promise& __rhs);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~promise();
+
+    // assignment
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    promise& operator=(promise&& __rhs) _NOEXCEPT
+        {
+            promise(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+    promise& operator=(const promise& __rhs) = delete;
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+private:
+    promise& operator=(const promise& __rhs);
+public:
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(promise& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // retrieving the result
+    future<void> get_future();
+
+    // setting the result
+    void set_value();
+    void set_exception(exception_ptr __p);
+
+    // setting the result with deferred notification
+    void set_value_at_thread_exit();
+    void set_exception_at_thread_exit(exception_ptr __p);
+};
+
+template <class _Alloc>
+promise<void>::promise(allocator_arg_t, const _Alloc& __a0)
+{
+    typedef __assoc_sub_state_alloc<_Alloc> _State;
+    typedef typename __allocator_traits_rebind<_Alloc, _State>::type _A2;
+    typedef __allocator_destructor<_A2> _D2;
+    _A2 __a(__a0);
+    unique_ptr<_State, _D2> __hold(__a.allocate(1), _D2(__a, 1));
+    ::new(static_cast<void*>(_CUDA_VSTD::addressof(*__hold.get()))) _State(__a0);
+    __state_ = _CUDA_VSTD::addressof(*__hold.release());
+}
+
+template <class _Rp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(promise<_Rp>& __x, promise<_Rp>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+template <class _Rp, class _Alloc>
+    struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator<promise<_Rp>, _Alloc>
+        : public true_type {};
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+// packaged_task
+
+template<class _Fp> class __packaged_task_base;
+
+template<class _Rp, class ..._ArgTypes>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __packaged_task_base<_Rp(_ArgTypes...)>
+{
+    __packaged_task_base(const __packaged_task_base&);
+    __packaged_task_base& operator=(const __packaged_task_base&);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __packaged_task_base() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual ~__packaged_task_base() {}
+    virtual void __move_to(__packaged_task_base*) _NOEXCEPT = 0;
+    virtual void destroy() = 0;
+    virtual void destroy_deallocate() = 0;
+    virtual _Rp operator()(_ArgTypes&& ...) = 0;
+};
+
+template<class _FD, class _Alloc, class _FB> class __packaged_task_func;
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __packaged_task_func<_Fp, _Alloc, _Rp(_ArgTypes...)>
+    : public  __packaged_task_base<_Rp(_ArgTypes...)>
+{
+    __compressed_pair<_Fp, _Alloc> __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __packaged_task_func(const _Fp& __f) : __f_(__f) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __packaged_task_func(_Fp&& __f) : __f_(_CUDA_VSTD::move(__f)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __packaged_task_func(const _Fp& __f, const _Alloc& __a)
+        : __f_(__f, __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __packaged_task_func(_Fp&& __f, const _Alloc& __a)
+        : __f_(_CUDA_VSTD::move(__f), __a) {}
+    virtual void __move_to(__packaged_task_base<_Rp(_ArgTypes...)>*) _NOEXCEPT;
+    virtual void destroy();
+    virtual void destroy_deallocate();
+    virtual _Rp operator()(_ArgTypes&& ... __args);
+};
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+void
+__packaged_task_func<_Fp, _Alloc, _Rp(_ArgTypes...)>::__move_to(
+                              __packaged_task_base<_Rp(_ArgTypes...)>* __p) _NOEXCEPT
+{
+    ::new (__p) __packaged_task_func(_CUDA_VSTD::move(__f_.first()), _CUDA_VSTD::move(__f_.second()));
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+void
+__packaged_task_func<_Fp, _Alloc, _Rp(_ArgTypes...)>::destroy()
+{
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+void
+__packaged_task_func<_Fp, _Alloc, _Rp(_ArgTypes...)>::destroy_deallocate()
+{
+    typedef typename __allocator_traits_rebind<_Alloc, __packaged_task_func>::type _Ap;
+    typedef allocator_traits<_Ap> _ATraits;
+    typedef pointer_traits<typename _ATraits::pointer> _PTraits;
+    _Ap __a(__f_.second());
+    __f_.~__compressed_pair<_Fp, _Alloc>();
+    __a.deallocate(_PTraits::pointer_to(*this), 1);
+}
+
+template<class _Fp, class _Alloc, class _Rp, class ..._ArgTypes>
+_Rp
+__packaged_task_func<_Fp, _Alloc, _Rp(_ArgTypes...)>::operator()(_ArgTypes&& ... __arg)
+{
+    return __invoke(__f_.first(), _CUDA_VSTD::forward<_ArgTypes>(__arg)...);
+}
+
+template <class _Callable> class __packaged_task_function;
+
+template<class _Rp, class ..._ArgTypes>
+class _LIBCUDACXX_AVAILABILITY_FUTURE __packaged_task_function<_Rp(_ArgTypes...)>
+{
+    typedef __packaged_task_base<_Rp(_ArgTypes...)> __base;
+    typename aligned_storage<3*sizeof(void*)>::type __buf_;
+    __base* __f_;
+
+public:
+    typedef _Rp result_type;
+
+    // construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __packaged_task_function() _NOEXCEPT : __f_(nullptr) {}
+    template<class _Fp>
+      __packaged_task_function(_Fp&& __f);
+    template<class _Fp, class _Alloc>
+      __packaged_task_function(allocator_arg_t, const _Alloc& __a, _Fp&& __f);
+
+    __packaged_task_function(__packaged_task_function&&) _NOEXCEPT;
+    __packaged_task_function& operator=(__packaged_task_function&&) _NOEXCEPT;
+
+    __packaged_task_function(const __packaged_task_function&) =  delete;
+    __packaged_task_function& operator=(const __packaged_task_function&) =  delete;
+
+    ~__packaged_task_function();
+
+    void swap(__packaged_task_function&) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Rp operator()(_ArgTypes...) const;
+};
+
+template<class _Rp, class ..._ArgTypes>
+__packaged_task_function<_Rp(_ArgTypes...)>::__packaged_task_function(__packaged_task_function&& __f) _NOEXCEPT
+{
+    if (__f.__f_ == nullptr)
+        __f_ = nullptr;
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__move_to(__f_);
+    }
+    else
+    {
+        __f_ = __f.__f_;
+        __f.__f_ = nullptr;
+    }
+}
+
+template<class _Rp, class ..._ArgTypes>
+template <class _Fp>
+__packaged_task_function<_Rp(_ArgTypes...)>::__packaged_task_function(_Fp&& __f)
+    : __f_(nullptr)
+{
+    typedef typename remove_reference<typename decay<_Fp>::type>::type _FR;
+    typedef __packaged_task_func<_FR, allocator<_FR>, _Rp(_ArgTypes...)> _FF;
+    if (sizeof(_FF) <= sizeof(__buf_))
+    {
+        __f_ = (__base*)&__buf_;
+        ::new (__f_) _FF(_CUDA_VSTD::forward<_Fp>(__f));
+    }
+    else
+    {
+        typedef allocator<_FF> _Ap;
+        _Ap __a;
+        typedef __allocator_destructor<_Ap> _Dp;
+        unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+        ::new (__hold.get()) _FF(_CUDA_VSTD::forward<_Fp>(__f), allocator<_FR>(__a));
+        __f_ = __hold.release();
+    }
+}
+
+template<class _Rp, class ..._ArgTypes>
+template <class _Fp, class _Alloc>
+__packaged_task_function<_Rp(_ArgTypes...)>::__packaged_task_function(
+                                  allocator_arg_t, const _Alloc& __a0, _Fp&& __f)
+    : __f_(nullptr)
+{
+    typedef typename remove_reference<typename decay<_Fp>::type>::type _FR;
+    typedef __packaged_task_func<_FR, _Alloc, _Rp(_ArgTypes...)> _FF;
+    if (sizeof(_FF) <= sizeof(__buf_))
+    {
+        __f_ = (__base*)&__buf_;
+        ::new (__f_) _FF(_CUDA_VSTD::forward<_Fp>(__f));
+    }
+    else
+    {
+        typedef typename __allocator_traits_rebind<_Alloc, _FF>::type _Ap;
+        _Ap __a(__a0);
+        typedef __allocator_destructor<_Ap> _Dp;
+        unique_ptr<__base, _Dp> __hold(__a.allocate(1), _Dp(__a, 1));
+        ::new (static_cast<void*>(_CUDA_VSTD::addressof(*__hold.get())))
+            _FF(_CUDA_VSTD::forward<_Fp>(__f), _Alloc(__a));
+        __f_ = _CUDA_VSTD::addressof(*__hold.release());
+    }
+}
+
+template<class _Rp, class ..._ArgTypes>
+__packaged_task_function<_Rp(_ArgTypes...)>&
+__packaged_task_function<_Rp(_ArgTypes...)>::operator=(__packaged_task_function&& __f) _NOEXCEPT
+{
+    if (__f_ == (__base*)&__buf_)
+        __f_->destroy();
+    else if (__f_)
+        __f_->destroy_deallocate();
+    __f_ = nullptr;
+    if (__f.__f_ == nullptr)
+        __f_ = nullptr;
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f_ = (__base*)&__buf_;
+        __f.__f_->__move_to(__f_);
+    }
+    else
+    {
+        __f_ = __f.__f_;
+        __f.__f_ = nullptr;
+    }
+    return *this;
+}
+
+template<class _Rp, class ..._ArgTypes>
+__packaged_task_function<_Rp(_ArgTypes...)>::~__packaged_task_function()
+{
+    if (__f_ == (__base*)&__buf_)
+        __f_->destroy();
+    else if (__f_)
+        __f_->destroy_deallocate();
+}
+
+template<class _Rp, class ..._ArgTypes>
+void
+__packaged_task_function<_Rp(_ArgTypes...)>::swap(__packaged_task_function& __f) _NOEXCEPT
+{
+    if (__f_ == (__base*)&__buf_ && __f.__f_ == (__base*)&__f.__buf_)
+    {
+        typename aligned_storage<sizeof(__buf_)>::type __tempbuf;
+        __base* __t = (__base*)&__tempbuf;
+        __f_->__move_to(__t);
+        __f_->destroy();
+        __f_ = nullptr;
+        __f.__f_->__move_to((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = nullptr;
+        __f_ = (__base*)&__buf_;
+        __t->__move_to((__base*)&__f.__buf_);
+        __t->destroy();
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f_ == (__base*)&__buf_)
+    {
+        __f_->__move_to((__base*)&__f.__buf_);
+        __f_->destroy();
+        __f_ = __f.__f_;
+        __f.__f_ = (__base*)&__f.__buf_;
+    }
+    else if (__f.__f_ == (__base*)&__f.__buf_)
+    {
+        __f.__f_->__move_to((__base*)&__buf_);
+        __f.__f_->destroy();
+        __f.__f_ = __f_;
+        __f_ = (__base*)&__buf_;
+    }
+    else
+        _CUDA_VSTD::swap(__f_, __f.__f_);
+}
+
+template<class _Rp, class ..._ArgTypes>
+inline
+_Rp
+__packaged_task_function<_Rp(_ArgTypes...)>::operator()(_ArgTypes... __arg) const
+{
+    return (*__f_)(_CUDA_VSTD::forward<_ArgTypes>(__arg)...);
+}
+
+template<class _Rp, class ..._ArgTypes>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE packaged_task<_Rp(_ArgTypes...)>
+{
+public:
+    typedef _Rp result_type; // extension
+
+private:
+    __packaged_task_function<result_type(_ArgTypes...)> __f_;
+    promise<result_type>                                __p_;
+
+public:
+    // construction and destruction
+    _LIBCUDACXX_INLINE_VISIBILITY
+    packaged_task() _NOEXCEPT : __p_(nullptr) {}
+    template <class _Fp,
+              class = typename enable_if
+              <
+                  !is_same<
+                      typename __uncvref<_Fp>::type,
+                      packaged_task
+                      >::value
+                  >::type
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit packaged_task(_Fp&& __f) : __f_(_CUDA_VSTD::forward<_Fp>(__f)) {}
+    template <class _Fp, class _Allocator,
+              class = typename enable_if
+              <
+                  !is_same<
+                      typename __uncvref<_Fp>::type,
+                      packaged_task
+                      >::value
+                  >::type
+              >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        packaged_task(allocator_arg_t, const _Allocator& __a, _Fp&& __f)
+             : __f_(allocator_arg, __a, _CUDA_VSTD::forward<_Fp>(__f)),
+               __p_(allocator_arg, __a) {}
+    // ~packaged_task() = default;
+
+    // no copy
+    packaged_task(const packaged_task&) = delete;
+    packaged_task& operator=(const packaged_task&) = delete;
+
+    // move support
+    _LIBCUDACXX_INLINE_VISIBILITY
+    packaged_task(packaged_task&& __other) _NOEXCEPT
+        : __f_(_CUDA_VSTD::move(__other.__f_)), __p_(_CUDA_VSTD::move(__other.__p_)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    packaged_task& operator=(packaged_task&& __other) _NOEXCEPT
+    {
+        __f_ = _CUDA_VSTD::move(__other.__f_);
+        __p_ = _CUDA_VSTD::move(__other.__p_);
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(packaged_task& __other) _NOEXCEPT
+    {
+        __f_.swap(__other.__f_);
+        __p_.swap(__other.__p_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __p_.__state_ != nullptr;}
+
+    // result retrieval
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future<result_type> get_future() {return __p_.get_future();}
+
+    // execution
+    void operator()(_ArgTypes... __args);
+    void make_ready_at_thread_exit(_ArgTypes... __args);
+
+    void reset();
+};
+
+template<class _Rp, class ..._ArgTypes>
+void
+packaged_task<_Rp(_ArgTypes...)>::operator()(_ArgTypes... __args)
+{
+    if (__p_.__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    if (__p_.__state_->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __p_.set_value(__f_(_CUDA_VSTD::forward<_ArgTypes>(__args)...));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __p_.set_exception(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class _Rp, class ..._ArgTypes>
+void
+packaged_task<_Rp(_ArgTypes...)>::make_ready_at_thread_exit(_ArgTypes... __args)
+{
+    if (__p_.__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    if (__p_.__state_->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __p_.set_value_at_thread_exit(__f_(_CUDA_VSTD::forward<_ArgTypes>(__args)...));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __p_.set_exception_at_thread_exit(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class _Rp, class ..._ArgTypes>
+void
+packaged_task<_Rp(_ArgTypes...)>::reset()
+{
+    if (!valid())
+        __throw_future_error(future_errc::no_state);
+    __p_ = promise<result_type>();
+}
+
+template<class ..._ArgTypes>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE packaged_task<void(_ArgTypes...)>
+{
+public:
+    typedef void result_type; // extension
+
+private:
+    __packaged_task_function<result_type(_ArgTypes...)> __f_;
+    promise<result_type>                                __p_;
+
+public:
+    // construction and destruction
+    _LIBCUDACXX_INLINE_VISIBILITY
+    packaged_task() _NOEXCEPT : __p_(nullptr) {}
+    template <class _Fp,
+              class = typename enable_if
+              <
+                  !is_same<
+                      typename __uncvref<_Fp>::type,
+                      packaged_task
+                      >::value
+                  >::type
+              >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit packaged_task(_Fp&& __f) : __f_(_CUDA_VSTD::forward<_Fp>(__f)) {}
+    template <class _Fp, class _Allocator,
+              class = typename enable_if
+              <
+                  !is_same<
+                      typename __uncvref<_Fp>::type,
+                      packaged_task
+                      >::value
+                  >::type
+              >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        packaged_task(allocator_arg_t, const _Allocator& __a, _Fp&& __f)
+             : __f_(allocator_arg, __a, _CUDA_VSTD::forward<_Fp>(__f)),
+               __p_(allocator_arg, __a) {}
+    // ~packaged_task() = default;
+
+    // no copy
+    packaged_task(const packaged_task&) = delete;
+    packaged_task& operator=(const packaged_task&) = delete;
+
+    // move support
+    _LIBCUDACXX_INLINE_VISIBILITY
+    packaged_task(packaged_task&& __other) _NOEXCEPT
+        : __f_(_CUDA_VSTD::move(__other.__f_)), __p_(_CUDA_VSTD::move(__other.__p_)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    packaged_task& operator=(packaged_task&& __other) _NOEXCEPT
+    {
+        __f_ = _CUDA_VSTD::move(__other.__f_);
+        __p_ = _CUDA_VSTD::move(__other.__p_);
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(packaged_task& __other) _NOEXCEPT
+    {
+        __f_.swap(__other.__f_);
+        __p_.swap(__other.__p_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __p_.__state_ != nullptr;}
+
+    // result retrieval
+    _LIBCUDACXX_INLINE_VISIBILITY
+    future<result_type> get_future() {return __p_.get_future();}
+
+    // execution
+    void operator()(_ArgTypes... __args);
+    void make_ready_at_thread_exit(_ArgTypes... __args);
+
+    void reset();
+};
+
+template<class ..._ArgTypes>
+void
+packaged_task<void(_ArgTypes...)>::operator()(_ArgTypes... __args)
+{
+    if (__p_.__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    if (__p_.__state_->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __f_(_CUDA_VSTD::forward<_ArgTypes>(__args)...);
+        __p_.set_value();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __p_.set_exception(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class ..._ArgTypes>
+void
+packaged_task<void(_ArgTypes...)>::make_ready_at_thread_exit(_ArgTypes... __args)
+{
+    if (__p_.__state_ == nullptr)
+        __throw_future_error(future_errc::no_state);
+    if (__p_.__state_->__has_value())
+        __throw_future_error(future_errc::promise_already_satisfied);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __f_(_CUDA_VSTD::forward<_ArgTypes>(__args)...);
+        __p_.set_value_at_thread_exit();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __p_.set_exception_at_thread_exit(current_exception());
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class ..._ArgTypes>
+void
+packaged_task<void(_ArgTypes...)>::reset()
+{
+    if (!valid())
+        __throw_future_error(future_errc::no_state);
+    __p_ = promise<result_type>();
+}
+
+template <class _Callable>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(packaged_task<_Callable>& __x, packaged_task<_Callable>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+template <class _Callable, class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator<packaged_task<_Callable>, _Alloc>
+    : public true_type {};
+
+template <class _Rp, class _Fp>
+future<_Rp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+__make_deferred_assoc_state(_Fp&& __f)
+#else
+__make_deferred_assoc_state(_Fp __f)
+#endif
+{
+    unique_ptr<__deferred_assoc_state<_Rp, _Fp>, __release_shared_count>
+        __h(new __deferred_assoc_state<_Rp, _Fp>(_CUDA_VSTD::forward<_Fp>(__f)));
+    return future<_Rp>(__h.get());
+}
+
+template <class _Rp, class _Fp>
+future<_Rp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+__make_async_assoc_state(_Fp&& __f)
+#else
+__make_async_assoc_state(_Fp __f)
+#endif
+{
+    unique_ptr<__async_assoc_state<_Rp, _Fp>, __release_shared_count>
+        __h(new __async_assoc_state<_Rp, _Fp>(_CUDA_VSTD::forward<_Fp>(__f)));
+    _CUDA_VSTD::thread(&__async_assoc_state<_Rp, _Fp>::__execute, __h.get()).detach();
+    return future<_Rp>(__h.get());
+}
+
+template <class _Fp, class... _Args>
+class __async_func
+{
+    tuple<_Fp, _Args...> __f_;
+
+public:
+    typedef typename __invoke_of<_Fp, _Args...>::type _Rp;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __async_func(_Fp&& __f, _Args&&... __args)
+        : __f_(_CUDA_VSTD::move(__f), _CUDA_VSTD::move(__args)...) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __async_func(__async_func&& __f) : __f_(_CUDA_VSTD::move(__f.__f_)) {}
+
+    _Rp operator()()
+    {
+        typedef typename __make_tuple_indices<1+sizeof...(_Args), 1>::type _Index;
+        return __execute(_Index());
+    }
+private:
+    template <size_t ..._Indices>
+    _Rp
+    __execute(__tuple_indices<_Indices...>)
+    {
+        return __invoke(_CUDA_VSTD::move(_CUDA_VSTD::get<0>(__f_)), _CUDA_VSTD::move(_CUDA_VSTD::get<_Indices>(__f_))...);
+    }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY bool __does_policy_contain(launch __policy, launch __value )
+{ return (int(__policy) & int(__value)) != 0; }
+
+template <class _Fp, class... _Args>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17
+future<typename __invoke_of<typename decay<_Fp>::type, typename decay<_Args>::type...>::type>
+async(launch __policy, _Fp&& __f, _Args&&... __args)
+{
+    typedef __async_func<typename decay<_Fp>::type, typename decay<_Args>::type...> _BF;
+    typedef typename _BF::_Rp _Rp;
+
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif
+        if (__does_policy_contain(__policy, launch::async))
+        return _CUDA_VSTD::__make_async_assoc_state<_Rp>(_BF(__decay_copy(_CUDA_VSTD::forward<_Fp>(__f)),
+                                                     __decay_copy(_CUDA_VSTD::forward<_Args>(__args))...));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch ( ... ) { if (__policy == launch::async) throw ; }
+#endif
+
+    if (__does_policy_contain(__policy, launch::deferred))
+        return _CUDA_VSTD::__make_deferred_assoc_state<_Rp>(_BF(__decay_copy(_CUDA_VSTD::forward<_Fp>(__f)),
+                                                        __decay_copy(_CUDA_VSTD::forward<_Args>(__args))...));
+    return future<_Rp>{};
+}
+
+template <class _Fp, class... _Args>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 inline _LIBCUDACXX_INLINE_VISIBILITY
+future<typename __invoke_of<typename decay<_Fp>::type, typename decay<_Args>::type...>::type>
+async(_Fp&& __f, _Args&&... __args)
+{
+    return _CUDA_VSTD::async(launch::any, _CUDA_VSTD::forward<_Fp>(__f),
+                                    _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+// shared_future
+
+template <class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS shared_future
+{
+    __assoc_state<_Rp>* __state_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future() _NOEXCEPT : __state_(nullptr) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(const shared_future& __rhs)  _NOEXCEPT : __state_(__rhs.__state_)
+        {if (__state_) __state_->__add_shared();}
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(future<_Rp>&& __f) _NOEXCEPT : __state_(__f.__state_)
+        {__f.__state_ = nullptr;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(shared_future&& __rhs) _NOEXCEPT : __state_(__rhs.__state_)
+        {__rhs.__state_ = nullptr;}
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~shared_future();
+    shared_future& operator=(const shared_future& __rhs) _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future& operator=(shared_future&& __rhs) _NOEXCEPT
+        {
+            shared_future(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+    // retrieving the value
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Rp& get() const {return __state_->copy();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(shared_future& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // functions to check state
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __state_ != nullptr;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const {__state_->wait();}
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+            {return __state_->wait_for(__rel_time);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+            {return __state_->wait_until(__abs_time);}
+};
+
+template <class _Rp>
+shared_future<_Rp>::~shared_future()
+{
+    if (__state_)
+        __state_->__release_shared();
+}
+
+template <class _Rp>
+shared_future<_Rp>&
+shared_future<_Rp>::operator=(const shared_future& __rhs) _NOEXCEPT
+{
+    if (__rhs.__state_)
+        __rhs.__state_->__add_shared();
+    if (__state_)
+        __state_->__release_shared();
+    __state_ = __rhs.__state_;
+    return *this;
+}
+
+template <class _Rp>
+class _LIBCUDACXX_TEMPLATE_VIS shared_future<_Rp&>
+{
+    __assoc_state<_Rp&>* __state_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future() _NOEXCEPT : __state_(nullptr) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(const shared_future& __rhs) : __state_(__rhs.__state_)
+        {if (__state_) __state_->__add_shared();}
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(future<_Rp&>&& __f) _NOEXCEPT : __state_(__f.__state_)
+        {__f.__state_ = nullptr;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(shared_future&& __rhs) _NOEXCEPT : __state_(__rhs.__state_)
+        {__rhs.__state_ = nullptr;}
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~shared_future();
+    shared_future& operator=(const shared_future& __rhs);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future& operator=(shared_future&& __rhs) _NOEXCEPT
+        {
+            shared_future(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+    // retrieving the value
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Rp& get() const {return __state_->copy();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(shared_future& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // functions to check state
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __state_ != nullptr;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const {__state_->wait();}
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+            {return __state_->wait_for(__rel_time);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+            {return __state_->wait_until(__abs_time);}
+};
+
+template <class _Rp>
+shared_future<_Rp&>::~shared_future()
+{
+    if (__state_)
+        __state_->__release_shared();
+}
+
+template <class _Rp>
+shared_future<_Rp&>&
+shared_future<_Rp&>::operator=(const shared_future& __rhs)
+{
+    if (__rhs.__state_)
+        __rhs.__state_->__add_shared();
+    if (__state_)
+        __state_->__release_shared();
+    __state_ = __rhs.__state_;
+    return *this;
+}
+
+template <>
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_FUTURE shared_future<void>
+{
+    __assoc_sub_state* __state_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future() _NOEXCEPT : __state_(nullptr) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(const shared_future& __rhs) : __state_(__rhs.__state_)
+        {if (__state_) __state_->__add_shared();}
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(future<void>&& __f) _NOEXCEPT : __state_(__f.__state_)
+        {__f.__state_ = nullptr;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future(shared_future&& __rhs) _NOEXCEPT : __state_(__rhs.__state_)
+        {__rhs.__state_ = nullptr;}
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~shared_future();
+    shared_future& operator=(const shared_future& __rhs);
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_future& operator=(shared_future&& __rhs) _NOEXCEPT
+        {
+            shared_future(std::move(__rhs)).swap(*this);
+            return *this;
+        }
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+    // retrieving the value
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void get() const {__state_->copy();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(shared_future& __rhs) _NOEXCEPT {_CUDA_VSTD::swap(__state_, __rhs.__state_);}
+
+    // functions to check state
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool valid() const _NOEXCEPT {return __state_ != nullptr;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const {__state_->wait();}
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_for(const chrono::duration<_Rep, _Period>& __rel_time) const
+            {return __state_->wait_for(__rel_time);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        future_status
+        wait_until(const chrono::time_point<_Clock, _Duration>& __abs_time) const
+            {return __state_->wait_until(__abs_time);}
+};
+
+template <class _Rp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(shared_future<_Rp>& __x, shared_future<_Rp>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+template <class _Rp>
+inline
+shared_future<_Rp>
+future<_Rp>::share() _NOEXCEPT
+{
+    return shared_future<_Rp>(_CUDA_VSTD::move(*this));
+}
+
+template <class _Rp>
+inline
+shared_future<_Rp&>
+future<_Rp&>::share() _NOEXCEPT
+{
+    return shared_future<_Rp&>(_CUDA_VSTD::move(*this));
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+inline
+shared_future<void>
+future<void>::share() _NOEXCEPT
+{
+    return shared_future<void>(_CUDA_VSTD::move(*this));
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+#endif  // _LIBCUDACXX_FUTURE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list
new file mode 100644
index 000000000000..d69d2f6a3fde
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/initializer_list
@@ -0,0 +1,117 @@
+// -*- C++ -*-
+//===----------------------- initializer_list -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_INITIALIZER_LIST
+#define _LIBCUDACXX_INITIALIZER_LIST
+
+/*
+    initializer_list synopsis
+
+namespace std
+{
+
+template<class E>
+class initializer_list
+{
+public:
+    typedef E        value_type;
+    typedef const E& reference;
+    typedef const E& const_reference;
+    typedef size_t   size_type;
+
+    typedef const E* iterator;
+    typedef const E* const_iterator;
+
+    initializer_list() noexcept; // constexpr in C++14
+
+    size_t   size()  const noexcept; // constexpr in C++14
+    const E* begin() const noexcept; // constexpr in C++14
+    const E* end()   const noexcept; // constexpr in C++14
+};
+
+template<class E> const E* begin(initializer_list<E> il) noexcept; // constexpr in C++14
+template<class E> const E* end(initializer_list<E> il) noexcept; // constexpr in C++14
+
+}  // std
+
+*/
+
+#include <__config>
+#include <cstddef>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+namespace std  // purposefully not versioned
+{
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Ep>
+class _LIBCUDACXX_TEMPLATE_VIS initializer_list
+{
+    const _Ep* __begin_;
+    size_t    __size_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    initializer_list(const _Ep* __b, size_t __s) _NOEXCEPT
+        : __begin_(__b),
+          __size_(__s)
+        {}
+public:
+    typedef _Ep        value_type;
+    typedef const _Ep& reference;
+    typedef const _Ep& const_reference;
+    typedef size_t    size_type;
+
+    typedef const _Ep* iterator;
+    typedef const _Ep* const_iterator;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    initializer_list() _NOEXCEPT : __begin_(nullptr), __size_(0) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    size_t    size()  const _NOEXCEPT {return __size_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const _Ep* begin() const _NOEXCEPT {return __begin_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const _Ep* end()   const _NOEXCEPT {return __begin_ + __size_;}
+};
+
+template<class _Ep>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Ep*
+begin(initializer_list<_Ep> __il) _NOEXCEPT
+{
+    return __il.begin();
+}
+
+template<class _Ep>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const _Ep*
+end(initializer_list<_Ep> __il) _NOEXCEPT
+{
+    return __il.end();
+}
+
+#endif  // !defined(_LIBCUDACXX_CXX03_LANG)
+
+}  // std
+
+#endif  // _LIBCUDACXX_INITIALIZER_LIST
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h
new file mode 100644
index 000000000000..c40180098d38
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/inttypes.h
@@ -0,0 +1,262 @@
+// -*- C++ -*-
+//===--------------------------- inttypes.h -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_INTTYPES_H
+// AIX system headers need inttypes.h to be re-enterable while _STD_TYPES_T
+// is defined until an inclusion of it without _STD_TYPES_T occurs, in which
+// case the header guard macro is defined.
+#if !defined(_AIX) || !defined(_STD_TYPES_T)
+#define _LIBCUDACXX_INTTYPES_H
+#endif // _STD_TYPES_T
+
+/*
+    inttypes.h synopsis
+
+This entire header is C99 / C++0X
+
+#include <stdint.h>  // <cinttypes> includes <cstdint>
+
+Macros:
+
+    PRId8
+    PRId16
+    PRId32
+    PRId64
+
+    PRIdLEAST8
+    PRIdLEAST16
+    PRIdLEAST32
+    PRIdLEAST64
+
+    PRIdFAST8
+    PRIdFAST16
+    PRIdFAST32
+    PRIdFAST64
+
+    PRIdMAX
+    PRIdPTR
+
+    PRIi8
+    PRIi16
+    PRIi32
+    PRIi64
+
+    PRIiLEAST8
+    PRIiLEAST16
+    PRIiLEAST32
+    PRIiLEAST64
+
+    PRIiFAST8
+    PRIiFAST16
+    PRIiFAST32
+    PRIiFAST64
+
+    PRIiMAX
+    PRIiPTR
+
+    PRIo8
+    PRIo16
+    PRIo32
+    PRIo64
+
+    PRIoLEAST8
+    PRIoLEAST16
+    PRIoLEAST32
+    PRIoLEAST64
+
+    PRIoFAST8
+    PRIoFAST16
+    PRIoFAST32
+    PRIoFAST64
+
+    PRIoMAX
+    PRIoPTR
+
+    PRIu8
+    PRIu16
+    PRIu32
+    PRIu64
+
+    PRIuLEAST8
+    PRIuLEAST16
+    PRIuLEAST32
+    PRIuLEAST64
+
+    PRIuFAST8
+    PRIuFAST16
+    PRIuFAST32
+    PRIuFAST64
+
+    PRIuMAX
+    PRIuPTR
+
+    PRIx8
+    PRIx16
+    PRIx32
+    PRIx64
+
+    PRIxLEAST8
+    PRIxLEAST16
+    PRIxLEAST32
+    PRIxLEAST64
+
+    PRIxFAST8
+    PRIxFAST16
+    PRIxFAST32
+    PRIxFAST64
+
+    PRIxMAX
+    PRIxPTR
+
+    PRIX8
+    PRIX16
+    PRIX32
+    PRIX64
+
+    PRIXLEAST8
+    PRIXLEAST16
+    PRIXLEAST32
+    PRIXLEAST64
+
+    PRIXFAST8
+    PRIXFAST16
+    PRIXFAST32
+    PRIXFAST64
+
+    PRIXMAX
+    PRIXPTR
+
+    SCNd8
+    SCNd16
+    SCNd32
+    SCNd64
+
+    SCNdLEAST8
+    SCNdLEAST16
+    SCNdLEAST32
+    SCNdLEAST64
+
+    SCNdFAST8
+    SCNdFAST16
+    SCNdFAST32
+    SCNdFAST64
+
+    SCNdMAX
+    SCNdPTR
+
+    SCNi8
+    SCNi16
+    SCNi32
+    SCNi64
+
+    SCNiLEAST8
+    SCNiLEAST16
+    SCNiLEAST32
+    SCNiLEAST64
+
+    SCNiFAST8
+    SCNiFAST16
+    SCNiFAST32
+    SCNiFAST64
+
+    SCNiMAX
+    SCNiPTR
+
+    SCNo8
+    SCNo16
+    SCNo32
+    SCNo64
+
+    SCNoLEAST8
+    SCNoLEAST16
+    SCNoLEAST32
+    SCNoLEAST64
+
+    SCNoFAST8
+    SCNoFAST16
+    SCNoFAST32
+    SCNoFAST64
+
+    SCNoMAX
+    SCNoPTR
+
+    SCNu8
+    SCNu16
+    SCNu32
+    SCNu64
+
+    SCNuLEAST8
+    SCNuLEAST16
+    SCNuLEAST32
+    SCNuLEAST64
+
+    SCNuFAST8
+    SCNuFAST16
+    SCNuFAST32
+    SCNuFAST64
+
+    SCNuMAX
+    SCNuPTR
+
+    SCNx8
+    SCNx16
+    SCNx32
+    SCNx64
+
+    SCNxLEAST8
+    SCNxLEAST16
+    SCNxLEAST32
+    SCNxLEAST64
+
+    SCNxFAST8
+    SCNxFAST16
+    SCNxFAST32
+    SCNxFAST64
+
+    SCNxMAX
+    SCNxPTR
+
+Types:
+
+    imaxdiv_t
+
+intmax_t  imaxabs(intmax_t j);
+imaxdiv_t imaxdiv(intmax_t numer, intmax_t denom);
+intmax_t  strtoimax(const char* restrict nptr, char** restrict endptr, int base);
+uintmax_t strtoumax(const char* restrict nptr, char** restrict endptr, int base);
+intmax_t  wcstoimax(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+uintmax_t wcstoumax(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+/* C99 stdlib (e.g. glibc < 2.18) does not provide format macros needed
+   for C++11 unless __STDC_FORMAT_MACROS is defined
+*/
+#if defined(__cplusplus) && !defined(__STDC_FORMAT_MACROS)
+#   define __STDC_FORMAT_MACROS
+#endif
+
+#include_next <inttypes.h>
+
+#ifdef __cplusplus
+
+#include <stdint.h>
+
+#undef imaxabs
+#undef imaxdiv
+
+#endif // __cplusplus
+
+#endif  // _LIBCUDACXX_INTTYPES_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip
new file mode 100644
index 000000000000..3684b01d69de
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iomanip
@@ -0,0 +1,670 @@
+// -*- C++ -*-
+//===--------------------------- iomanip ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_IOMANIP
+#define _LIBCUDACXX_IOMANIP
+
+/*
+    iomanip synopsis
+
+namespace std {
+
+// types T1, T2, ... are unspecified implementation types
+T1 resetiosflags(ios_base::fmtflags mask);
+T2 setiosflags (ios_base::fmtflags mask);
+T3 setbase(int base);
+template<charT> T4 setfill(charT c);
+T5 setprecision(int n);
+T6 setw(int n);
+template <class moneyT> T7 get_money(moneyT& mon, bool intl = false);
+template <class charT, class moneyT> T8 put_money(const moneyT& mon, bool intl = false);
+template <class charT> T9 get_time(struct tm* tmb, const charT* fmt);
+template <class charT> T10 put_time(const struct tm* tmb, const charT* fmt);
+
+template <class charT>
+  T11 quoted(const charT* s, charT delim=charT('"'), charT escape=charT('\\')); // C++14
+
+template <class charT, class traits, class Allocator>
+  T12 quoted(const basic_string<charT, traits, Allocator>& s,
+             charT delim=charT('"'), charT escape=charT('\\')); // C++14
+
+template <class charT, class traits, class Allocator>
+  T13 quoted(basic_string<charT, traits, Allocator>& s,
+             charT delim=charT('"'), charT escape=charT('\\')); // C++14
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__string>
+#include <istream>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// resetiosflags
+
+class __iom_t1
+{
+    ios_base::fmtflags __mask_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __iom_t1(ios_base::fmtflags __m) : __mask_(__m) {}
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t1& __x)
+    {
+        __is.unsetf(__x.__mask_);
+        return __is;
+    }
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t1& __x)
+    {
+        __os.unsetf(__x.__mask_);
+        return __os;
+    }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t1
+resetiosflags(ios_base::fmtflags __mask)
+{
+    return __iom_t1(__mask);
+}
+
+// setiosflags
+
+class __iom_t2
+{
+    ios_base::fmtflags __mask_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __iom_t2(ios_base::fmtflags __m) : __mask_(__m) {}
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t2& __x)
+    {
+        __is.setf(__x.__mask_);
+        return __is;
+    }
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t2& __x)
+    {
+        __os.setf(__x.__mask_);
+        return __os;
+    }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t2
+setiosflags(ios_base::fmtflags __mask)
+{
+    return __iom_t2(__mask);
+}
+
+// setbase
+
+class __iom_t3
+{
+    int __base_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __iom_t3(int __b) : __base_(__b) {}
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t3& __x)
+    {
+        __is.setf(__x.__base_ == 8  ? ios_base::oct :
+                  __x.__base_ == 10 ? ios_base::dec :
+                  __x.__base_ == 16 ? ios_base::hex :
+                  ios_base::fmtflags(0), ios_base::basefield);
+        return __is;
+    }
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t3& __x)
+    {
+        __os.setf(__x.__base_ == 8  ? ios_base::oct :
+                  __x.__base_ == 10 ? ios_base::dec :
+                  __x.__base_ == 16 ? ios_base::hex :
+                  ios_base::fmtflags(0), ios_base::basefield);
+        return __os;
+    }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t3
+setbase(int __base)
+{
+    return __iom_t3(__base);
+}
+
+// setfill
+
+template<class _CharT>
+class __iom_t4
+{
+    _CharT __fill_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __iom_t4(_CharT __c) : __fill_(__c) {}
+
+    template <class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t4& __x)
+    {
+        __os.fill(__x.__fill_);
+        return __os;
+    }
+};
+
+template<class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t4<_CharT>
+setfill(_CharT __c)
+{
+    return __iom_t4<_CharT>(__c);
+}
+
+// setprecision
+
+class __iom_t5
+{
+    int __n_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __iom_t5(int __n) : __n_(__n) {}
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t5& __x)
+    {
+        __is.precision(__x.__n_);
+        return __is;
+    }
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t5& __x)
+    {
+        __os.precision(__x.__n_);
+        return __os;
+    }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t5
+setprecision(int __n)
+{
+    return __iom_t5(__n);
+}
+
+// setw
+
+class __iom_t6
+{
+    int __n_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __iom_t6(int __n) : __n_(__n) {}
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t6& __x)
+    {
+        __is.width(__x.__n_);
+        return __is;
+    }
+
+    template <class _CharT, class _Traits>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t6& __x)
+    {
+        __os.width(__x.__n_);
+        return __os;
+    }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t6
+setw(int __n)
+{
+    return __iom_t6(__n);
+}
+
+// get_money
+
+template <class _MoneyT> class __iom_t7;
+
+template <class _CharT, class _Traits, class _MoneyT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t7<_MoneyT>& __x);
+
+template <class _MoneyT>
+class __iom_t7
+{
+    _MoneyT& __mon_;
+    bool __intl_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __iom_t7(_MoneyT& __mon, bool __intl)
+        : __mon_(__mon), __intl_(__intl) {}
+
+    template <class _CharT, class _Traits, class _Mp>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t7<_Mp>& __x);
+};
+
+template <class _CharT, class _Traits, class _MoneyT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t7<_MoneyT>& __x)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_istream<_CharT, _Traits>::sentry __s(__is);
+        if (__s)
+        {
+            typedef istreambuf_iterator<_CharT, _Traits> _Ip;
+            typedef money_get<_CharT, _Ip> _Fp;
+            ios_base::iostate __err = ios_base::goodbit;
+            const _Fp& __mf = use_facet<_Fp>(__is.getloc());
+            __mf.get(_Ip(__is), _Ip(), __x.__intl_, __is, __err, __x.__mon_);
+            __is.setstate(__err);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __is.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __is;
+}
+
+template <class _MoneyT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t7<_MoneyT>
+get_money(_MoneyT& __mon, bool __intl = false)
+{
+    return __iom_t7<_MoneyT>(__mon, __intl);
+}
+
+// put_money
+
+template <class _MoneyT> class __iom_t8;
+
+template <class _CharT, class _Traits, class _MoneyT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t8<_MoneyT>& __x);
+
+template <class _MoneyT>
+class __iom_t8
+{
+    const _MoneyT& __mon_;
+    bool __intl_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __iom_t8(const _MoneyT& __mon, bool __intl)
+        : __mon_(__mon), __intl_(__intl) {}
+
+    template <class _CharT, class _Traits, class _Mp>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t8<_Mp>& __x);
+};
+
+template <class _CharT, class _Traits, class _MoneyT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t8<_MoneyT>& __x)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_ostream<_CharT, _Traits>::sentry __s(__os);
+        if (__s)
+        {
+            typedef ostreambuf_iterator<_CharT, _Traits> _Op;
+            typedef money_put<_CharT, _Op> _Fp;
+            const _Fp& __mf = use_facet<_Fp>(__os.getloc());
+            if (__mf.put(_Op(__os), __x.__intl_, __os, __os.fill(), __x.__mon_).failed())
+                __os.setstate(ios_base::badbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __os.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __os;
+}
+
+template <class _MoneyT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t8<_MoneyT>
+put_money(const _MoneyT& __mon, bool __intl = false)
+{
+    return __iom_t8<_MoneyT>(__mon, __intl);
+}
+
+// get_time
+
+template <class _CharT> class __iom_t9;
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t9<_CharT>& __x);
+
+template <class _CharT>
+class __iom_t9
+{
+    tm* __tm_;
+    const _CharT* __fmt_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __iom_t9(tm* __tm, const _CharT* __fmt)
+        : __tm_(__tm), __fmt_(__fmt) {}
+
+    template <class _Cp, class _Traits>
+    friend
+    basic_istream<_Cp, _Traits>&
+    operator>>(basic_istream<_Cp, _Traits>& __is, const __iom_t9<_Cp>& __x);
+};
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, const __iom_t9<_CharT>& __x)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_istream<_CharT, _Traits>::sentry __s(__is);
+        if (__s)
+        {
+            typedef istreambuf_iterator<_CharT, _Traits> _Ip;
+            typedef time_get<_CharT, _Ip> _Fp;
+            ios_base::iostate __err = ios_base::goodbit;
+            const _Fp& __tf = use_facet<_Fp>(__is.getloc());
+            __tf.get(_Ip(__is), _Ip(), __is, __err, __x.__tm_,
+                     __x.__fmt_, __x.__fmt_ + _Traits::length(__x.__fmt_));
+            __is.setstate(__err);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __is.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __is;
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t9<_CharT>
+get_time(tm* __tm, const _CharT* __fmt)
+{
+    return __iom_t9<_CharT>(__tm, __fmt);
+}
+
+// put_time
+
+template <class _CharT> class __iom_t10;
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t10<_CharT>& __x);
+
+template <class _CharT>
+class __iom_t10
+{
+    const tm* __tm_;
+    const _CharT* __fmt_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __iom_t10(const tm* __tm, const _CharT* __fmt)
+        : __tm_(__tm), __fmt_(__fmt) {}
+
+    template <class _Cp, class _Traits>
+    friend
+    basic_ostream<_Cp, _Traits>&
+    operator<<(basic_ostream<_Cp, _Traits>& __os, const __iom_t10<_Cp>& __x);
+};
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const __iom_t10<_CharT>& __x)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_ostream<_CharT, _Traits>::sentry __s(__os);
+        if (__s)
+        {
+            typedef ostreambuf_iterator<_CharT, _Traits> _Op;
+            typedef time_put<_CharT, _Op> _Fp;
+            const _Fp& __tf = use_facet<_Fp>(__os.getloc());
+            if (__tf.put(_Op(__os), __os, __os.fill(), __x.__tm_,
+                         __x.__fmt_, __x.__fmt_ + _Traits::length(__x.__fmt_)).failed())
+                __os.setstate(ios_base::badbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __os.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __os;
+}
+
+template <class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__iom_t10<_CharT>
+put_time(const tm* __tm, const _CharT* __fmt)
+{
+    return __iom_t10<_CharT>(__tm, __fmt);
+}
+
+template <class _CharT, class _Traits, class _ForwardIterator>
+std::basic_ostream<_CharT, _Traits> &
+__quoted_output ( basic_ostream<_CharT, _Traits> &__os,
+        _ForwardIterator __first, _ForwardIterator __last, _CharT __delim, _CharT __escape )
+{
+    _CUDA_VSTD::basic_string<_CharT, _Traits> __str;
+    __str.push_back(__delim);
+    for ( ; __first != __last; ++ __first )
+    {
+        if (_Traits::eq (*__first, __escape) || _Traits::eq (*__first, __delim))
+            __str.push_back(__escape);
+        __str.push_back(*__first);
+    }
+    __str.push_back(__delim);
+    return __put_character_sequence(__os, __str.data(), __str.size());
+}
+
+template <class _CharT, class _Traits, class _String>
+basic_istream<_CharT, _Traits> &
+__quoted_input ( basic_istream<_CharT, _Traits> &__is, _String & __string, _CharT __delim, _CharT __escape )
+{
+    __string.clear ();
+    _CharT __c;
+    __is >> __c;
+    if ( __is.fail ())
+        return __is;
+
+    if (!_Traits::eq (__c, __delim))    // no delimiter, read the whole string
+    {
+        __is.unget ();
+        __is >> __string;
+        return __is;
+    }
+
+    __save_flags<_CharT, _Traits> sf(__is);
+    noskipws (__is);
+    while (true)
+        {
+        __is >> __c;
+        if ( __is.fail ())
+            break;
+        if (_Traits::eq (__c, __escape))
+        {
+            __is >> __c;
+            if ( __is.fail ())
+                break;
+        }
+        else if (_Traits::eq (__c, __delim))
+            break;
+        __string.push_back ( __c );
+        }
+    return __is;
+}
+
+
+template <class _CharT, class _Traits, class _Iter>
+basic_ostream<_CharT, _Traits>& operator<<(
+         basic_ostream<_CharT, _Traits>& __os,
+         const __quoted_output_proxy<_CharT, _Iter, _Traits> & __proxy)
+{
+    return __quoted_output (__os, __proxy.__first, __proxy.__last, __proxy.__delim, __proxy.__escape);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+struct __quoted_proxy
+{
+    basic_string<_CharT, _Traits, _Allocator> &__string;
+    _CharT  __delim;
+    _CharT  __escape;
+
+    __quoted_proxy(basic_string<_CharT, _Traits, _Allocator> &__s, _CharT __d, _CharT __e)
+    : __string(__s), __delim(__d), __escape(__e) {}
+};
+
+template <class _CharT, class _Traits, class _Allocator>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>& operator<<(
+        basic_ostream<_CharT, _Traits>& __os,
+        const __quoted_proxy<_CharT, _Traits, _Allocator> & __proxy)
+{
+    return __quoted_output (__os, __proxy.__string.cbegin (), __proxy.__string.cend (), __proxy.__delim, __proxy.__escape);
+}
+
+//  extractor for non-const basic_string& proxies
+template <class _CharT, class _Traits, class _Allocator>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>& operator>>(
+        basic_istream<_CharT, _Traits>& __is,
+        const __quoted_proxy<_CharT, _Traits, _Allocator> & __proxy)
+{
+    return __quoted_input ( __is, __proxy.__string, __proxy.__delim, __proxy.__escape );
+}
+
+
+template <class _CharT>
+_LIBCUDACXX_INLINE_VISIBILITY
+__quoted_output_proxy<_CharT, const _CharT *>
+quoted ( const _CharT *__s, _CharT __delim = _CharT('"'), _CharT __escape =_CharT('\\'))
+{
+    const _CharT *__end = __s;
+    while ( *__end ) ++__end;
+    return __quoted_output_proxy<_CharT, const _CharT *> ( __s, __end, __delim, __escape );
+}
+
+
+template <class _CharT, class _Traits, class _Allocator>
+_LIBCUDACXX_INLINE_VISIBILITY
+__quoted_output_proxy<_CharT, typename basic_string <_CharT, _Traits, _Allocator>::const_iterator>
+__quoted ( const basic_string <_CharT, _Traits, _Allocator> &__s, _CharT __delim = _CharT('"'), _CharT __escape=_CharT('\\'))
+{
+    return __quoted_output_proxy<_CharT,
+            typename basic_string <_CharT, _Traits, _Allocator>::const_iterator>
+                    ( __s.cbegin(), __s.cend (), __delim, __escape );
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+_LIBCUDACXX_INLINE_VISIBILITY
+__quoted_proxy<_CharT, _Traits, _Allocator>
+__quoted ( basic_string <_CharT, _Traits, _Allocator> &__s, _CharT __delim = _CharT('"'), _CharT __escape=_CharT('\\'))
+{
+    return __quoted_proxy<_CharT, _Traits, _Allocator>( __s, __delim, __escape );
+}
+
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template <class _CharT, class _Traits, class _Allocator>
+_LIBCUDACXX_INLINE_VISIBILITY
+__quoted_output_proxy<_CharT, typename basic_string <_CharT, _Traits, _Allocator>::const_iterator>
+quoted ( const basic_string <_CharT, _Traits, _Allocator> &__s, _CharT __delim = _CharT('"'), _CharT __escape=_CharT('\\'))
+{
+    return __quoted(__s, __delim, __escape);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+_LIBCUDACXX_INLINE_VISIBILITY
+__quoted_proxy<_CharT, _Traits, _Allocator>
+quoted ( basic_string <_CharT, _Traits, _Allocator> &__s, _CharT __delim = _CharT('"'), _CharT __escape=_CharT('\\'))
+{
+    return __quoted(__s, __delim, __escape);
+}
+
+template <class _CharT, class _Traits>
+__quoted_output_proxy<_CharT, const _CharT *, _Traits>
+quoted (basic_string_view <_CharT, _Traits> __sv,
+             _CharT __delim = _CharT('"'), _CharT __escape=_CharT('\\'))
+{
+    return __quoted_output_proxy<_CharT, const _CharT *, _Traits>
+         ( __sv.data(), __sv.data() + __sv.size(), __delim, __escape );
+}
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_IOMANIP
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ios b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ios
new file mode 100644
index 000000000000..9f389f2c035e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ios
@@ -0,0 +1,1066 @@
+// -*- C++ -*-
+//===---------------------------- ios -------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_IOS
+#define _LIBCUDACXX_IOS
+
+/*
+    ios synopsis
+
+#include <iosfwd>
+
+namespace std
+{
+
+typedef OFF_T streamoff;
+typedef SZ_T streamsize;
+template <class stateT> class fpos;
+
+class ios_base
+{
+public:
+    class failure;
+
+    typedef T1 fmtflags;
+    static constexpr fmtflags boolalpha;
+    static constexpr fmtflags dec;
+    static constexpr fmtflags fixed;
+    static constexpr fmtflags hex;
+    static constexpr fmtflags internal;
+    static constexpr fmtflags left;
+    static constexpr fmtflags oct;
+    static constexpr fmtflags right;
+    static constexpr fmtflags scientific;
+    static constexpr fmtflags showbase;
+    static constexpr fmtflags showpoint;
+    static constexpr fmtflags showpos;
+    static constexpr fmtflags skipws;
+    static constexpr fmtflags unitbuf;
+    static constexpr fmtflags uppercase;
+    static constexpr fmtflags adjustfield;
+    static constexpr fmtflags basefield;
+    static constexpr fmtflags floatfield;
+
+    typedef T2 iostate;
+    static constexpr iostate badbit;
+    static constexpr iostate eofbit;
+    static constexpr iostate failbit;
+    static constexpr iostate goodbit;
+
+    typedef T3 openmode;
+    static constexpr openmode app;
+    static constexpr openmode ate;
+    static constexpr openmode binary;
+    static constexpr openmode in;
+    static constexpr openmode out;
+    static constexpr openmode trunc;
+
+    typedef T4 seekdir;
+    static constexpr seekdir beg;
+    static constexpr seekdir cur;
+    static constexpr seekdir end;
+
+    class Init;
+
+    // 27.5.2.2 fmtflags state:
+    fmtflags flags() const;
+    fmtflags flags(fmtflags fmtfl);
+    fmtflags setf(fmtflags fmtfl);
+    fmtflags setf(fmtflags fmtfl, fmtflags mask);
+    void unsetf(fmtflags mask);
+
+    streamsize precision() const;
+    streamsize precision(streamsize prec);
+    streamsize width() const;
+    streamsize width(streamsize wide);
+
+    // 27.5.2.3 locales:
+    locale imbue(const locale& loc);
+    locale getloc() const;
+
+    // 27.5.2.5 storage:
+    static int xalloc();
+    long& iword(int index);
+    void*& pword(int index);
+
+    // destructor
+    virtual ~ios_base();
+
+    // 27.5.2.6 callbacks;
+    enum event { erase_event, imbue_event, copyfmt_event };
+    typedef void (*event_callback)(event, ios_base&, int index);
+    void register_callback(event_callback fn, int index);
+
+    ios_base(const ios_base&) = delete;
+    ios_base& operator=(const ios_base&) = delete;
+
+    static bool sync_with_stdio(bool sync = true);
+
+protected:
+    ios_base();
+};
+
+template <class charT, class traits = char_traits<charT> >
+class basic_ios
+    : public ios_base
+{
+public:
+    // types:
+    typedef charT char_type;
+    typedef typename traits::int_type int_type;  // removed in C++17
+    typedef typename traits::pos_type pos_type;  // removed in C++17
+    typedef typename traits::off_type off_type;  // removed in C++17
+    typedef traits traits_type;
+
+    operator unspecified-bool-type() const;
+    bool operator!() const;
+    iostate rdstate() const;
+    void clear(iostate state = goodbit);
+    void setstate(iostate state);
+    bool good() const;
+    bool eof() const;
+    bool fail() const;
+    bool bad() const;
+
+    iostate exceptions() const;
+    void exceptions(iostate except);
+
+    // 27.5.4.1 Constructor/destructor:
+    explicit basic_ios(basic_streambuf<charT,traits>* sb);
+    virtual ~basic_ios();
+
+    // 27.5.4.2 Members:
+    basic_ostream<charT,traits>* tie() const;
+    basic_ostream<charT,traits>* tie(basic_ostream<charT,traits>* tiestr);
+
+    basic_streambuf<charT,traits>* rdbuf() const;
+    basic_streambuf<charT,traits>* rdbuf(basic_streambuf<charT,traits>* sb);
+
+    basic_ios& copyfmt(const basic_ios& rhs);
+
+    char_type fill() const;
+    char_type fill(char_type ch);
+
+    locale imbue(const locale& loc);
+
+    char narrow(char_type c, char dfault) const;
+    char_type widen(char c) const;
+
+    basic_ios(const basic_ios& ) = delete;
+    basic_ios& operator=(const basic_ios&) = delete;
+
+protected:
+    basic_ios();
+    void init(basic_streambuf<charT,traits>* sb);
+    void move(basic_ios& rhs);
+    void swap(basic_ios& rhs) noexcept;
+    void set_rdbuf(basic_streambuf<charT, traits>* sb);
+};
+
+// 27.5.5, manipulators:
+ios_base& boolalpha (ios_base& str);
+ios_base& noboolalpha(ios_base& str);
+ios_base& showbase (ios_base& str);
+ios_base& noshowbase (ios_base& str);
+ios_base& showpoint (ios_base& str);
+ios_base& noshowpoint(ios_base& str);
+ios_base& showpos (ios_base& str);
+ios_base& noshowpos (ios_base& str);
+ios_base& skipws (ios_base& str);
+ios_base& noskipws (ios_base& str);
+ios_base& uppercase (ios_base& str);
+ios_base& nouppercase(ios_base& str);
+ios_base& unitbuf (ios_base& str);
+ios_base& nounitbuf (ios_base& str);
+
+// 27.5.5.2 adjustfield:
+ios_base& internal (ios_base& str);
+ios_base& left (ios_base& str);
+ios_base& right (ios_base& str);
+
+// 27.5.5.3 basefield:
+ios_base& dec (ios_base& str);
+ios_base& hex (ios_base& str);
+ios_base& oct (ios_base& str);
+
+// 27.5.5.4 floatfield:
+ios_base& fixed (ios_base& str);
+ios_base& scientific (ios_base& str);
+ios_base& hexfloat (ios_base& str);
+ios_base& defaultfloat(ios_base& str);
+
+// 27.5.5.5 error reporting:
+enum class io_errc
+{
+    stream = 1
+};
+
+concept_map ErrorCodeEnum<io_errc> { };
+error_code make_error_code(io_errc e) noexcept;
+error_condition make_error_condition(io_errc e) noexcept;
+storage-class-specifier const error_category& iostream_category() noexcept;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <iosfwd>
+#include <__locale>
+#include <system_error>
+
+#if !defined(_LIBCUDACXX_HAS_NO_ATOMIC_HEADER)
+#include <atomic>     // for __xindex_
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+typedef ptrdiff_t streamsize;
+
+class _LIBCUDACXX_TYPE_VIS ios_base
+{
+public:
+    class _LIBCUDACXX_EXCEPTION_ABI failure;
+
+    typedef unsigned int fmtflags;
+    static const fmtflags boolalpha   = 0x0001;
+    static const fmtflags dec         = 0x0002;
+    static const fmtflags fixed       = 0x0004;
+    static const fmtflags hex         = 0x0008;
+    static const fmtflags internal    = 0x0010;
+    static const fmtflags left        = 0x0020;
+    static const fmtflags oct         = 0x0040;
+    static const fmtflags right       = 0x0080;
+    static const fmtflags scientific  = 0x0100;
+    static const fmtflags showbase    = 0x0200;
+    static const fmtflags showpoint   = 0x0400;
+    static const fmtflags showpos     = 0x0800;
+    static const fmtflags skipws      = 0x1000;
+    static const fmtflags unitbuf     = 0x2000;
+    static const fmtflags uppercase   = 0x4000;
+    static const fmtflags adjustfield = left | right | internal;
+    static const fmtflags basefield   = dec | oct | hex;
+    static const fmtflags floatfield  = scientific | fixed;
+
+    typedef unsigned int iostate;
+    static const iostate badbit  = 0x1;
+    static const iostate eofbit  = 0x2;
+    static const iostate failbit = 0x4;
+    static const iostate goodbit = 0x0;
+
+    typedef unsigned int openmode;
+    static const openmode app    = 0x01;
+    static const openmode ate    = 0x02;
+    static const openmode binary = 0x04;
+    static const openmode in     = 0x08;
+    static const openmode out    = 0x10;
+    static const openmode trunc  = 0x20;
+
+    enum seekdir {beg, cur, end};
+
+#if _LIBCUDACXX_STD_VER <= 14
+    typedef iostate      io_state;
+    typedef openmode     open_mode;
+    typedef seekdir      seek_dir;
+
+    typedef _CUDA_VSTD::streamoff streamoff;
+    typedef _CUDA_VSTD::streampos streampos;
+#endif
+
+    class _LIBCUDACXX_TYPE_VIS Init;
+
+    // 27.5.2.2 fmtflags state:
+    _LIBCUDACXX_INLINE_VISIBILITY fmtflags flags() const;
+    _LIBCUDACXX_INLINE_VISIBILITY fmtflags flags(fmtflags __fmtfl);
+    _LIBCUDACXX_INLINE_VISIBILITY fmtflags setf(fmtflags __fmtfl);
+    _LIBCUDACXX_INLINE_VISIBILITY fmtflags setf(fmtflags __fmtfl, fmtflags __mask);
+    _LIBCUDACXX_INLINE_VISIBILITY void unsetf(fmtflags __mask);
+
+    _LIBCUDACXX_INLINE_VISIBILITY streamsize precision() const;
+    _LIBCUDACXX_INLINE_VISIBILITY streamsize precision(streamsize __prec);
+    _LIBCUDACXX_INLINE_VISIBILITY streamsize width() const;
+    _LIBCUDACXX_INLINE_VISIBILITY streamsize width(streamsize __wide);
+
+    // 27.5.2.3 locales:
+    locale imbue(const locale& __loc);
+    locale getloc() const;
+
+    // 27.5.2.5 storage:
+    static int xalloc();
+    long& iword(int __index);
+    void*& pword(int __index);
+
+    // destructor
+    virtual ~ios_base();
+
+    // 27.5.2.6 callbacks;
+    enum event { erase_event, imbue_event, copyfmt_event };
+    typedef void (*event_callback)(event, ios_base&, int __index);
+    void register_callback(event_callback __fn, int __index);
+
+private:
+    ios_base(const ios_base&); // = delete;
+    ios_base& operator=(const ios_base&); // = delete;
+
+public:
+    static bool sync_with_stdio(bool __sync = true);
+
+    _LIBCUDACXX_INLINE_VISIBILITY iostate rdstate() const;
+    void clear(iostate __state = goodbit);
+    _LIBCUDACXX_INLINE_VISIBILITY void setstate(iostate __state);
+
+    _LIBCUDACXX_INLINE_VISIBILITY bool good() const;
+    _LIBCUDACXX_INLINE_VISIBILITY bool eof() const;
+    _LIBCUDACXX_INLINE_VISIBILITY bool fail() const;
+    _LIBCUDACXX_INLINE_VISIBILITY bool bad() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY iostate exceptions() const;
+    _LIBCUDACXX_INLINE_VISIBILITY void exceptions(iostate __iostate);
+
+    void __set_badbit_and_consider_rethrow();
+    void __set_failbit_and_consider_rethrow();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __setstate_nothrow(iostate __state)
+    {
+        if (__rdbuf_)
+            __rdstate_ |= __state;
+        else
+            __rdstate_ |= __state | ios_base::badbit;
+    }
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ios_base() {// purposefully does no initialization
+               }
+
+    void init(void* __sb);
+    _LIBCUDACXX_INLINE_VISIBILITY void* rdbuf() const {return __rdbuf_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void rdbuf(void* __sb)
+    {
+        __rdbuf_ = __sb;
+        clear();
+    }
+
+    void __call_callbacks(event);
+    void copyfmt(const ios_base&);
+    void move(ios_base&);
+    void swap(ios_base&) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void set_rdbuf(void* __sb)
+    {
+        __rdbuf_ = __sb;
+    }
+
+private:
+    // All data members must be scalars
+    fmtflags        __fmtflags_;
+    streamsize      __precision_;
+    streamsize      __width_;
+    iostate         __rdstate_;
+    iostate         __exceptions_;
+    void*           __rdbuf_;
+    void*           __loc_;
+    event_callback* __fn_;
+    int*            __index_;
+    size_t          __event_size_;
+    size_t          __event_cap_;
+// TODO(EricWF): Enable this for both Clang and GCC. Currently it is only
+// enabled with clang.
+#if defined(_LIBCUDACXX_HAS_C_ATOMIC_IMP) && !defined(_LIBCUDACXX_HAS_NO_THREADS)
+    static atomic<int> __xindex_;
+#else
+    static int      __xindex_;
+#endif
+    long*           __iarray_;
+    size_t          __iarray_size_;
+    size_t          __iarray_cap_;
+    void**          __parray_;
+    size_t          __parray_size_;
+    size_t          __parray_cap_;
+};
+
+//enum class io_errc
+_LIBCUDACXX_DECLARE_STRONG_ENUM(io_errc)
+{
+    stream = 1
+};
+_LIBCUDACXX_DECLARE_STRONG_ENUM_EPILOG(io_errc)
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_code_enum<io_errc> : public true_type { };
+
+#ifdef _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_code_enum<io_errc::__lx> : public true_type { };
+#endif
+
+_LIBCUDACXX_FUNC_VIS
+const error_category& iostream_category() _NOEXCEPT;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+error_code
+make_error_code(io_errc __e) _NOEXCEPT
+{
+    return error_code(static_cast<int>(__e), iostream_category());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+error_condition
+make_error_condition(io_errc __e) _NOEXCEPT
+{
+    return error_condition(static_cast<int>(__e), iostream_category());
+}
+
+class _LIBCUDACXX_EXCEPTION_ABI ios_base::failure
+    : public system_error
+{
+public:
+    explicit failure(const string& __msg, const error_code& __ec = io_errc::stream);
+    explicit failure(const char* __msg, const error_code& __ec = io_errc::stream);
+    virtual ~failure() throw();
+};
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_failure(char const* __msg) {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw ios_base::failure(__msg);
+#else
+    ((void)__msg);
+    _CUDA_VSTD::abort();
+#endif
+}
+
+class _LIBCUDACXX_TYPE_VIS ios_base::Init
+{
+public:
+    Init();
+    ~Init();
+};
+
+// fmtflags
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base::fmtflags
+ios_base::flags() const
+{
+    return __fmtflags_;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base::fmtflags
+ios_base::flags(fmtflags __fmtfl)
+{
+    fmtflags __r = __fmtflags_;
+    __fmtflags_ = __fmtfl;
+    return __r;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base::fmtflags
+ios_base::setf(fmtflags __fmtfl)
+{
+    fmtflags __r = __fmtflags_;
+    __fmtflags_ |= __fmtfl;
+    return __r;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+ios_base::unsetf(fmtflags __mask)
+{
+    __fmtflags_ &= ~__mask;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base::fmtflags
+ios_base::setf(fmtflags __fmtfl, fmtflags __mask)
+{
+    fmtflags __r = __fmtflags_;
+    unsetf(__mask);
+    __fmtflags_ |= __fmtfl & __mask;
+    return __r;
+}
+
+// precision
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+streamsize
+ios_base::precision() const
+{
+    return __precision_;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+streamsize
+ios_base::precision(streamsize __prec)
+{
+    streamsize __r = __precision_;
+    __precision_ = __prec;
+    return __r;
+}
+
+// width
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+streamsize
+ios_base::width() const
+{
+    return __width_;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+streamsize
+ios_base::width(streamsize __wide)
+{
+    streamsize __r = __width_;
+    __width_ = __wide;
+    return __r;
+}
+
+// iostate
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base::iostate
+ios_base::rdstate() const
+{
+    return __rdstate_;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+ios_base::setstate(iostate __state)
+{
+    clear(__rdstate_ | __state);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+ios_base::good() const
+{
+    return __rdstate_ == 0;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+ios_base::eof() const
+{
+    return (__rdstate_ & eofbit) != 0;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+ios_base::fail() const
+{
+    return (__rdstate_ & (failbit | badbit)) != 0;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+ios_base::bad() const
+{
+    return (__rdstate_ & badbit) != 0;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base::iostate
+ios_base::exceptions() const
+{
+    return __exceptions_;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+ios_base::exceptions(iostate __iostate)
+{
+    __exceptions_ = __iostate;
+    clear(__rdstate_);
+}
+
+#if defined(_LIBCUDACXX_CXX03_LANG)
+struct _LIBCUDACXX_TYPE_VIS __cxx03_bool {
+  typedef void (__cxx03_bool::*__bool_type)();
+  void __true_value() {}
+};
+#endif
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_ios
+    : public ios_base
+{
+public:
+    // types:
+    typedef _CharT char_type;
+    typedef _Traits traits_type;
+
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    static_assert((is_same<_CharT, typename traits_type::char_type>::value),
+                  "traits_type::char_type must be the same type as CharT");
+
+  // __true_value will generate undefined references when linking unless
+  // we give it internal linkage.
+
+#if defined(_LIBCUDACXX_CXX03_LANG)
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator __cxx03_bool::__bool_type() const {
+        return !fail() ? &__cxx03_bool::__true_value : nullptr;
+    }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_EXPLICIT operator bool() const {return !fail();}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY bool operator!() const    {return  fail();}
+    _LIBCUDACXX_INLINE_VISIBILITY iostate rdstate() const   {return ios_base::rdstate();}
+    _LIBCUDACXX_INLINE_VISIBILITY void clear(iostate __state = goodbit) {ios_base::clear(__state);}
+    _LIBCUDACXX_INLINE_VISIBILITY void setstate(iostate __state) {ios_base::setstate(__state);}
+    _LIBCUDACXX_INLINE_VISIBILITY bool good() const {return ios_base::good();}
+    _LIBCUDACXX_INLINE_VISIBILITY bool eof() const  {return ios_base::eof();}
+    _LIBCUDACXX_INLINE_VISIBILITY bool fail() const {return ios_base::fail();}
+    _LIBCUDACXX_INLINE_VISIBILITY bool bad() const  {return ios_base::bad();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY iostate exceptions() const {return ios_base::exceptions();}
+    _LIBCUDACXX_INLINE_VISIBILITY void exceptions(iostate __iostate) {ios_base::exceptions(__iostate);}
+
+    // 27.5.4.1 Constructor/destructor:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ios(basic_streambuf<char_type,traits_type>* __sb);
+    virtual ~basic_ios();
+
+    // 27.5.4.2 Members:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<char_type, traits_type>* tie() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream<char_type, traits_type>* tie(basic_ostream<char_type, traits_type>* __tiestr);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_streambuf<char_type, traits_type>* rdbuf() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_streambuf<char_type, traits_type>* rdbuf(basic_streambuf<char_type, traits_type>* __sb);
+
+    basic_ios& copyfmt(const basic_ios& __rhs);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type fill() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type fill(char_type __ch);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    locale imbue(const locale& __loc);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char narrow(char_type __c, char __dfault) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type widen(char __c) const;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ios() {// purposefully does no initialization
+                }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void init(basic_streambuf<char_type, traits_type>* __sb);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void move(basic_ios& __rhs);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void move(basic_ios&& __rhs) {move(__rhs);}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_ios& __rhs) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void set_rdbuf(basic_streambuf<char_type, traits_type>* __sb);
+private:
+    basic_ostream<char_type, traits_type>* __tie_;
+    mutable int_type __fill_;
+};
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ios<_CharT, _Traits>::basic_ios(basic_streambuf<char_type,traits_type>* __sb)
+{
+    init(__sb);
+}
+
+template <class _CharT, class _Traits>
+basic_ios<_CharT, _Traits>::~basic_ios()
+{
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+basic_ios<_CharT, _Traits>::init(basic_streambuf<char_type, traits_type>* __sb)
+{
+    ios_base::init(__sb);
+    __tie_ = 0;
+    __fill_ = traits_type::eof();
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>*
+basic_ios<_CharT, _Traits>::tie() const
+{
+    return __tie_;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>*
+basic_ios<_CharT, _Traits>::tie(basic_ostream<char_type, traits_type>* __tiestr)
+{
+    basic_ostream<char_type, traits_type>* __r = __tie_;
+    __tie_ = __tiestr;
+    return __r;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_streambuf<_CharT, _Traits>*
+basic_ios<_CharT, _Traits>::rdbuf() const
+{
+    return static_cast<basic_streambuf<char_type, traits_type>*>(ios_base::rdbuf());
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_streambuf<_CharT, _Traits>*
+basic_ios<_CharT, _Traits>::rdbuf(basic_streambuf<char_type, traits_type>* __sb)
+{
+    basic_streambuf<char_type, traits_type>* __r = rdbuf();
+    ios_base::rdbuf(__sb);
+    return __r;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+locale
+basic_ios<_CharT, _Traits>::imbue(const locale& __loc)
+{
+    locale __r = getloc();
+    ios_base::imbue(__loc);
+    if (rdbuf())
+        rdbuf()->pubimbue(__loc);
+    return __r;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+char
+basic_ios<_CharT, _Traits>::narrow(char_type __c, char __dfault) const
+{
+    return use_facet<ctype<char_type> >(getloc()).narrow(__c, __dfault);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_CharT
+basic_ios<_CharT, _Traits>::widen(char __c) const
+{
+    return use_facet<ctype<char_type> >(getloc()).widen(__c);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_CharT
+basic_ios<_CharT, _Traits>::fill() const
+{
+    if (traits_type::eq_int_type(traits_type::eof(), __fill_))
+        __fill_ = widen(' ');
+    return __fill_;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_CharT
+basic_ios<_CharT, _Traits>::fill(char_type __ch)
+{
+    char_type __r = __fill_;
+    __fill_ = __ch;
+    return __r;
+}
+
+template <class _CharT, class _Traits>
+basic_ios<_CharT, _Traits>&
+basic_ios<_CharT, _Traits>::copyfmt(const basic_ios& __rhs)
+{
+    if (this != &__rhs)
+    {
+        __call_callbacks(erase_event);
+        ios_base::copyfmt(__rhs);
+        __tie_ = __rhs.__tie_;
+        __fill_ = __rhs.__fill_;
+        __call_callbacks(copyfmt_event);
+        exceptions(__rhs.exceptions());
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+basic_ios<_CharT, _Traits>::move(basic_ios& __rhs)
+{
+    ios_base::move(__rhs);
+    __tie_ = __rhs.__tie_;
+    __rhs.__tie_ = 0;
+    __fill_ = __rhs.__fill_;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+basic_ios<_CharT, _Traits>::swap(basic_ios& __rhs) _NOEXCEPT
+{
+    ios_base::swap(__rhs);
+    _CUDA_VSTD::swap(__tie_, __rhs.__tie_);
+    _CUDA_VSTD::swap(__fill_, __rhs.__fill_);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+basic_ios<_CharT, _Traits>::set_rdbuf(basic_streambuf<char_type, traits_type>* __sb)
+{
+    ios_base::set_rdbuf(__sb);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+boolalpha(ios_base& __str)
+{
+    __str.setf(ios_base::boolalpha);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+noboolalpha(ios_base& __str)
+{
+    __str.unsetf(ios_base::boolalpha);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+showbase(ios_base& __str)
+{
+    __str.setf(ios_base::showbase);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+noshowbase(ios_base& __str)
+{
+    __str.unsetf(ios_base::showbase);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+showpoint(ios_base& __str)
+{
+    __str.setf(ios_base::showpoint);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+noshowpoint(ios_base& __str)
+{
+    __str.unsetf(ios_base::showpoint);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+showpos(ios_base& __str)
+{
+    __str.setf(ios_base::showpos);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+noshowpos(ios_base& __str)
+{
+    __str.unsetf(ios_base::showpos);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+skipws(ios_base& __str)
+{
+    __str.setf(ios_base::skipws);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+noskipws(ios_base& __str)
+{
+    __str.unsetf(ios_base::skipws);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+uppercase(ios_base& __str)
+{
+    __str.setf(ios_base::uppercase);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+nouppercase(ios_base& __str)
+{
+    __str.unsetf(ios_base::uppercase);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+unitbuf(ios_base& __str)
+{
+    __str.setf(ios_base::unitbuf);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+nounitbuf(ios_base& __str)
+{
+    __str.unsetf(ios_base::unitbuf);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+internal(ios_base& __str)
+{
+    __str.setf(ios_base::internal, ios_base::adjustfield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+left(ios_base& __str)
+{
+    __str.setf(ios_base::left, ios_base::adjustfield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+right(ios_base& __str)
+{
+    __str.setf(ios_base::right, ios_base::adjustfield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+dec(ios_base& __str)
+{
+    __str.setf(ios_base::dec, ios_base::basefield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+hex(ios_base& __str)
+{
+    __str.setf(ios_base::hex, ios_base::basefield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+oct(ios_base& __str)
+{
+    __str.setf(ios_base::oct, ios_base::basefield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+fixed(ios_base& __str)
+{
+    __str.setf(ios_base::fixed, ios_base::floatfield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+scientific(ios_base& __str)
+{
+    __str.setf(ios_base::scientific, ios_base::floatfield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+hexfloat(ios_base& __str)
+{
+    __str.setf(ios_base::fixed | ios_base::scientific, ios_base::floatfield);
+    return __str;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+ios_base&
+defaultfloat(ios_base& __str)
+{
+    __str.unsetf(ios_base::floatfield);
+    return __str;
+}
+
+template <class _CharT, class _Traits>
+class __save_flags
+{
+    typedef basic_ios<_CharT, _Traits> __stream_type;
+    typedef typename __stream_type::fmtflags fmtflags;
+
+    __stream_type& __stream_;
+    fmtflags       __fmtflags_;
+    _CharT         __fill_;
+
+    __save_flags(const __save_flags&);
+    __save_flags& operator=(const __save_flags&);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __save_flags(__stream_type& __stream)
+        : __stream_(__stream),
+          __fmtflags_(__stream.flags()),
+          __fill_(__stream.fill())
+        {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__save_flags()
+    {
+        __stream_.flags(__fmtflags_);
+        __stream_.fill(__fill_);
+    }
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_IOS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd
new file mode 100644
index 000000000000..d10b5ecfae68
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iosfwd
@@ -0,0 +1,229 @@
+// -*- C++ -*-
+//===--------------------------- iosfwd -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_IOSFWD
+#define _LIBCUDACXX_IOSFWD
+
+/*
+    iosfwd synopsis
+
+namespace std
+{
+
+template<class charT> struct char_traits;
+template<>            struct char_traits<char>;
+template<>            struct char_traits<char8_t>;  // C++20
+template<>            struct char_traits<char16_t>;
+template<>            struct char_traits<char32_t>;
+template<>            struct char_traits<wchar_t>;
+
+template<class T>     class allocator;
+
+class ios_base;
+template <class charT, class traits = char_traits<charT> > class basic_ios;
+
+template <class charT, class traits = char_traits<charT> > class basic_streambuf;
+template <class charT, class traits = char_traits<charT> > class basic_istream;
+template <class charT, class traits = char_traits<charT> > class basic_ostream;
+template <class charT, class traits = char_traits<charT> > class basic_iostream;
+
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+    class basic_stringbuf;
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+    class basic_istringstream;
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+    class basic_ostringstream;
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+    class basic_stringstream;
+
+template <class charT, class traits = char_traits<charT> > class basic_filebuf;
+template <class charT, class traits = char_traits<charT> > class basic_ifstream;
+template <class charT, class traits = char_traits<charT> > class basic_ofstream;
+template <class charT, class traits = char_traits<charT> > class basic_fstream;
+
+template <class charT, class traits = char_traits<charT> > class istreambuf_iterator;
+template <class charT, class traits = char_traits<charT> > class ostreambuf_iterator;
+
+typedef basic_ios<char>              ios;
+typedef basic_ios<wchar_t>           wios;
+
+typedef basic_streambuf<char>        streambuf;
+typedef basic_istream<char>          istream;
+typedef basic_ostream<char>          ostream;
+typedef basic_iostream<char>         iostream;
+
+typedef basic_stringbuf<char>        stringbuf;
+typedef basic_istringstream<char>    istringstream;
+typedef basic_ostringstream<char>    ostringstream;
+typedef basic_stringstream<char>     stringstream;
+
+typedef basic_filebuf<char>          filebuf;
+typedef basic_ifstream<char>         ifstream;
+typedef basic_ofstream<char>         ofstream;
+typedef basic_fstream<char>          fstream;
+
+typedef basic_streambuf<wchar_t>     wstreambuf;
+typedef basic_istream<wchar_t>       wistream;
+typedef basic_ostream<wchar_t>       wostream;
+typedef basic_iostream<wchar_t>      wiostream;
+
+typedef basic_stringbuf<wchar_t>     wstringbuf;
+typedef basic_istringstream<wchar_t> wistringstream;
+typedef basic_ostringstream<wchar_t> wostringstream;
+typedef basic_stringstream<wchar_t>  wstringstream;
+
+typedef basic_filebuf<wchar_t>       wfilebuf;
+typedef basic_ifstream<wchar_t>      wifstream;
+typedef basic_ofstream<wchar_t>      wofstream;
+typedef basic_fstream<wchar_t>       wfstream;
+
+template <class state> class fpos;
+typedef fpos<char_traits<char>::state_type>    streampos;
+typedef fpos<char_traits<wchar_t>::state_type> wstreampos;
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <wchar.h>  // for mbstate_t
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+class _LIBCUDACXX_TYPE_VIS ios_base;
+
+template<class _CharT>  struct _LIBCUDACXX_TEMPLATE_VIS char_traits;
+template<> struct char_traits<char>;
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+template<> struct char_traits<char8_t>;
+#endif
+template<> struct char_traits<char16_t>;
+template<> struct char_traits<char32_t>;
+template<> struct char_traits<wchar_t>;
+
+template<class _Tp>     class _LIBCUDACXX_TEMPLATE_VIS allocator;
+
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_ios;
+
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_streambuf;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_istream;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_ostream;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_iostream;
+
+template <class _CharT, class _Traits = char_traits<_CharT>,
+          class _Allocator = allocator<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_stringbuf;
+template <class _CharT, class _Traits = char_traits<_CharT>,
+          class _Allocator = allocator<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_istringstream;
+template <class _CharT, class _Traits = char_traits<_CharT>,
+          class _Allocator = allocator<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_ostringstream;
+template <class _CharT, class _Traits = char_traits<_CharT>,
+          class _Allocator = allocator<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_stringstream;
+
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_filebuf;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_ifstream;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_ofstream;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_fstream;
+
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS istreambuf_iterator;
+template <class _CharT, class _Traits = char_traits<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS ostreambuf_iterator;
+
+typedef basic_ios<char>              ios;
+typedef basic_ios<wchar_t>           wios;
+
+typedef basic_streambuf<char>        streambuf;
+typedef basic_istream<char>          istream;
+typedef basic_ostream<char>          ostream;
+typedef basic_iostream<char>         iostream;
+
+typedef basic_stringbuf<char>        stringbuf;
+typedef basic_istringstream<char>    istringstream;
+typedef basic_ostringstream<char>    ostringstream;
+typedef basic_stringstream<char>     stringstream;
+
+typedef basic_filebuf<char>          filebuf;
+typedef basic_ifstream<char>         ifstream;
+typedef basic_ofstream<char>         ofstream;
+typedef basic_fstream<char>          fstream;
+
+typedef basic_streambuf<wchar_t>     wstreambuf;
+typedef basic_istream<wchar_t>       wistream;
+typedef basic_ostream<wchar_t>       wostream;
+typedef basic_iostream<wchar_t>      wiostream;
+
+typedef basic_stringbuf<wchar_t>     wstringbuf;
+typedef basic_istringstream<wchar_t> wistringstream;
+typedef basic_ostringstream<wchar_t> wostringstream;
+typedef basic_stringstream<wchar_t>  wstringstream;
+
+typedef basic_filebuf<wchar_t>       wfilebuf;
+typedef basic_ifstream<wchar_t>      wifstream;
+typedef basic_ofstream<wchar_t>      wofstream;
+typedef basic_fstream<wchar_t>       wfstream;
+
+#if !defined(_LIBCUDACXX_HAS_NO_WCHAR_H)
+template <class _State>             class _LIBCUDACXX_TEMPLATE_VIS fpos;
+typedef fpos<mbstate_t>    streampos;
+typedef fpos<mbstate_t>    wstreampos;
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+typedef fpos<mbstate_t>    u8streampos;
+#endif
+#ifndef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+typedef fpos<mbstate_t>    u16streampos;
+typedef fpos<mbstate_t>    u32streampos;
+#endif  // _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+#endif
+
+#if defined(_NEWLIB_VERSION)
+// On newlib, off_t is 'long int'
+typedef long int streamoff;         // for char_traits in <string>
+#else
+typedef long long streamoff;        // for char_traits in <string>
+#endif
+
+template <class _CharT,             // for <stdexcept>
+          class _Traits = char_traits<_CharT>,
+          class _Allocator = allocator<_CharT> >
+    class _LIBCUDACXX_TEMPLATE_VIS basic_string;
+typedef basic_string<char, char_traits<char>, allocator<char> > string;
+typedef basic_string<wchar_t, char_traits<wchar_t>, allocator<wchar_t> > wstring;
+
+
+// Include other forward declarations here
+template <class _Tp, class _Alloc = allocator<_Tp> >
+class _LIBCUDACXX_TEMPLATE_VIS vector;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_IOSFWD
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iostream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iostream
new file mode 100644
index 000000000000..e6f0d86dfdf0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iostream
@@ -0,0 +1,63 @@
+// -*- C++ -*-
+//===--------------------------- iostream ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_IOSTREAM
+#define _LIBCUDACXX_IOSTREAM
+
+/*
+    iostream synopsis
+
+#include <ios>
+#include <streambuf>
+#include <istream>
+#include <ostream>
+
+namespace std {
+
+extern istream cin;
+extern ostream cout;
+extern ostream cerr;
+extern ostream clog;
+extern wistream wcin;
+extern wostream wcout;
+extern wostream wcerr;
+extern wostream wclog;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <ios>
+#include <streambuf>
+#include <istream>
+#include <ostream>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#ifndef _LIBCUDACXX_HAS_NO_STDIN
+extern _LIBCUDACXX_FUNC_VIS istream cin;
+extern _LIBCUDACXX_FUNC_VIS wistream wcin;
+#endif
+#ifndef _LIBCUDACXX_HAS_NO_STDOUT
+extern _LIBCUDACXX_FUNC_VIS ostream cout;
+extern _LIBCUDACXX_FUNC_VIS wostream wcout;
+#endif
+extern _LIBCUDACXX_FUNC_VIS ostream cerr;
+extern _LIBCUDACXX_FUNC_VIS wostream wcerr;
+extern _LIBCUDACXX_FUNC_VIS ostream clog;
+extern _LIBCUDACXX_FUNC_VIS wostream wclog;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_IOSTREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/istream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/istream
new file mode 100644
index 000000000000..e4e860cadc88
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/istream
@@ -0,0 +1,1651 @@
+// -*- C++ -*-
+//===--------------------------- istream ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ISTREAM
+#define _LIBCUDACXX_ISTREAM
+
+/*
+    istream synopsis
+
+template <class charT, class traits = char_traits<charT> >
+class basic_istream
+    : virtual public basic_ios<charT,traits>
+{
+public:
+    // types (inherited from basic_ios (27.5.4)):
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // 27.7.1.1.1 Constructor/destructor:
+    explicit basic_istream(basic_streambuf<char_type, traits_type>* sb);
+    basic_istream(basic_istream&& rhs);
+    virtual ~basic_istream();
+
+    // 27.7.1.1.2 Assign/swap:
+    basic_istream& operator=(basic_istream&& rhs);
+    void swap(basic_istream& rhs);
+
+    // 27.7.1.1.3 Prefix/suffix:
+    class sentry;
+
+    // 27.7.1.2 Formatted input:
+    basic_istream& operator>>(basic_istream& (*pf)(basic_istream&));
+    basic_istream& operator>>(basic_ios<char_type, traits_type>&
+                              (*pf)(basic_ios<char_type, traits_type>&));
+    basic_istream& operator>>(ios_base& (*pf)(ios_base&));
+    basic_istream& operator>>(basic_streambuf<char_type, traits_type>* sb);
+    basic_istream& operator>>(bool& n);
+    basic_istream& operator>>(short& n);
+    basic_istream& operator>>(unsigned short& n);
+    basic_istream& operator>>(int& n);
+    basic_istream& operator>>(unsigned int& n);
+    basic_istream& operator>>(long& n);
+    basic_istream& operator>>(unsigned long& n);
+    basic_istream& operator>>(long long& n);
+    basic_istream& operator>>(unsigned long long& n);
+    basic_istream& operator>>(float& f);
+    basic_istream& operator>>(double& f);
+    basic_istream& operator>>(long double& f);
+    basic_istream& operator>>(void*& p);
+
+    // 27.7.1.3 Unformatted input:
+    streamsize gcount() const;
+    int_type get();
+    basic_istream& get(char_type& c);
+    basic_istream& get(char_type* s, streamsize n);
+    basic_istream& get(char_type* s, streamsize n, char_type delim);
+    basic_istream& get(basic_streambuf<char_type,traits_type>& sb);
+    basic_istream& get(basic_streambuf<char_type,traits_type>& sb, char_type delim);
+
+    basic_istream& getline(char_type* s, streamsize n);
+    basic_istream& getline(char_type* s, streamsize n, char_type delim);
+
+    basic_istream& ignore(streamsize n = 1, int_type delim = traits_type::eof());
+    int_type peek();
+    basic_istream& read (char_type* s, streamsize n);
+    streamsize readsome(char_type* s, streamsize n);
+
+    basic_istream& putback(char_type c);
+    basic_istream& unget();
+    int sync();
+
+    pos_type tellg();
+    basic_istream& seekg(pos_type);
+    basic_istream& seekg(off_type, ios_base::seekdir);
+protected:
+    basic_istream(const basic_istream& rhs) = delete;
+    basic_istream(basic_istream&& rhs);
+    // 27.7.2.1.2 Assign/swap:
+    basic_istream& operator=(const basic_istream& rhs) = delete;
+    basic_istream& operator=(basic_istream&& rhs);
+    void swap(basic_istream& rhs);
+};
+
+// 27.7.1.2.3 character extraction templates:
+template<class charT, class traits>
+  basic_istream<charT,traits>& operator>>(basic_istream<charT,traits>&, charT&);
+
+template<class traits>
+  basic_istream<char,traits>& operator>>(basic_istream<char,traits>&, unsigned char&);
+
+template<class traits>
+  basic_istream<char,traits>& operator>>(basic_istream<char,traits>&, signed char&);
+
+template<class charT, class traits>
+  basic_istream<charT,traits>& operator>>(basic_istream<charT,traits>&, charT*);
+
+template<class traits>
+  basic_istream<char,traits>& operator>>(basic_istream<char,traits>&, unsigned char*);
+
+template<class traits>
+  basic_istream<char,traits>& operator>>(basic_istream<char,traits>&, signed char*);
+
+template <class charT, class traits>
+  void
+  swap(basic_istream<charT, traits>& x, basic_istream<charT, traits>& y);
+
+typedef basic_istream<char> istream;
+typedef basic_istream<wchar_t> wistream;
+
+template <class charT, class traits = char_traits<charT> >
+class basic_iostream :
+    public basic_istream<charT,traits>,
+    public basic_ostream<charT,traits>
+{
+public:
+    // types:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // constructor/destructor
+    explicit basic_iostream(basic_streambuf<char_type, traits_type>* sb);
+    basic_iostream(basic_iostream&& rhs);
+    virtual ~basic_iostream();
+
+    // assign/swap
+    basic_iostream& operator=(basic_iostream&& rhs);
+    void swap(basic_iostream& rhs);
+};
+
+template <class charT, class traits>
+  void
+  swap(basic_iostream<charT, traits>& x, basic_iostream<charT, traits>& y);
+
+typedef basic_iostream<char> iostream;
+typedef basic_iostream<wchar_t> wiostream;
+
+template <class charT, class traits>
+  basic_istream<charT,traits>&
+  ws(basic_istream<charT,traits>& is);
+
+template <class charT, class traits, class T>
+  basic_istream<charT, traits>&
+  operator>>(basic_istream<charT, traits>&& is, T& x);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <version>
+#include <ostream>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_istream
+    : virtual public basic_ios<_CharT, _Traits>
+{
+    streamsize __gc_;
+public:
+    // types (inherited from basic_ios (27.5.4)):
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // 27.7.1.1.1 Constructor/destructor:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    explicit basic_istream(basic_streambuf<char_type, traits_type>* __sb) : __gc_(0)
+    { this->init(__sb); }
+    virtual ~basic_istream();
+protected:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream(basic_istream&& __rhs);
+
+    // 27.7.1.1.2 Assign/swap:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istream& operator=(basic_istream&& __rhs);
+#endif
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void swap(basic_istream& __rhs) {
+      _CUDA_VSTD::swap(__gc_, __rhs.__gc_);
+      basic_ios<char_type, traits_type>::swap(__rhs);
+    }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    basic_istream           (const basic_istream& __rhs) = delete;
+    basic_istream& operator=(const basic_istream& __rhs) = delete;
+#endif
+public:
+
+    // 27.7.1.1.3 Prefix/suffix:
+    class _LIBCUDACXX_TEMPLATE_VIS sentry;
+
+    // 27.7.1.2 Formatted input:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& operator>>(basic_istream& (*__pf)(basic_istream&))
+    { return __pf(*this); }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& operator>>(basic_ios<char_type, traits_type>&
+                              (*__pf)(basic_ios<char_type, traits_type>&))
+    { __pf(*this); return *this; }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& operator>>(ios_base& (*__pf)(ios_base&))
+    { __pf(*this); return *this; }
+
+    basic_istream& operator>>(basic_streambuf<char_type, traits_type>* __sb);
+    basic_istream& operator>>(bool& __n);
+    basic_istream& operator>>(short& __n);
+    basic_istream& operator>>(unsigned short& __n);
+    basic_istream& operator>>(int& __n);
+    basic_istream& operator>>(unsigned int& __n);
+    basic_istream& operator>>(long& __n);
+    basic_istream& operator>>(unsigned long& __n);
+    basic_istream& operator>>(long long& __n);
+    basic_istream& operator>>(unsigned long long& __n);
+    basic_istream& operator>>(float& __f);
+    basic_istream& operator>>(double& __f);
+    basic_istream& operator>>(long double& __f);
+    basic_istream& operator>>(void*& __p);
+
+    // 27.7.1.3 Unformatted input:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    streamsize gcount() const {return __gc_;}
+    int_type get();
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& get(char_type& __c) {
+      int_type __ch = get();
+      if (__ch != traits_type::eof())
+        __c = traits_type::to_char_type(__ch);
+      return *this;
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& get(char_type* __s, streamsize __n)
+    { return get(__s, __n, this->widen('\n')); }
+
+    basic_istream& get(char_type* __s, streamsize __n, char_type __dlm);
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& get(basic_streambuf<char_type, traits_type>& __sb)
+    { return get(__sb, this->widen('\n')); }
+
+    basic_istream& get(basic_streambuf<char_type, traits_type>& __sb, char_type __dlm);
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_istream& getline(char_type* __s, streamsize __n)
+    { return getline(__s, __n, this->widen('\n')); }
+
+    basic_istream& getline(char_type* __s, streamsize __n, char_type __dlm);
+
+    basic_istream& ignore(streamsize __n = 1, int_type __dlm = traits_type::eof());
+    int_type peek();
+    basic_istream& read (char_type* __s, streamsize __n);
+    streamsize readsome(char_type* __s, streamsize __n);
+
+    basic_istream& putback(char_type __c);
+    basic_istream& unget();
+    int sync();
+
+    pos_type tellg();
+    basic_istream& seekg(pos_type __pos);
+    basic_istream& seekg(off_type __off, ios_base::seekdir __dir);
+};
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_istream<_CharT, _Traits>::sentry
+{
+    bool __ok_;
+
+    sentry(const sentry&); // = delete;
+    sentry& operator=(const sentry&); // = delete;
+
+public:
+    explicit sentry(basic_istream<_CharT, _Traits>& __is, bool __noskipws = false);
+//    ~sentry() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+        _LIBCUDACXX_EXPLICIT
+        operator bool() const {return __ok_;}
+};
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>::sentry::sentry(basic_istream<_CharT, _Traits>& __is,
+                                               bool __noskipws)
+    : __ok_(false)
+{
+    if (__is.good())
+    {
+        if (__is.tie())
+            __is.tie()->flush();
+        if (!__noskipws && (__is.flags() & ios_base::skipws))
+        {
+            typedef istreambuf_iterator<_CharT, _Traits> _Ip;
+            const ctype<_CharT>& __ct = use_facet<ctype<_CharT> >(__is.getloc());
+            _Ip __i(__is);
+            _Ip __eof;
+            for (; __i != __eof; ++__i)
+                if (!__ct.is(__ct.space, *__i))
+                    break;
+            if (__i == __eof)
+                __is.setstate(ios_base::failbit | ios_base::eofbit);
+        }
+        __ok_ = __is.good();
+    }
+    else
+        __is.setstate(ios_base::failbit);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>::basic_istream(basic_istream&& __rhs)
+    : __gc_(__rhs.__gc_)
+{
+    __rhs.__gc_ = 0;
+    this->move(__rhs);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator=(basic_istream&& __rhs)
+{
+    swap(__rhs);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>::~basic_istream()
+{
+}
+
+template <class _Tp, class _CharT, class _Traits>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+__input_arithmetic(basic_istream<_CharT, _Traits>& __is, _Tp& __n) {
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __s(__is);
+    if (__s)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            typedef istreambuf_iterator<_CharT, _Traits> _Ip;
+            typedef num_get<_CharT, _Ip> _Fp;
+            use_facet<_Fp>(__is.getloc()).get(_Ip(__is), _Ip(), __is, __state, __n);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(unsigned short& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<unsigned short>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(unsigned int& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<unsigned int>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(long& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<long>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(unsigned long& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<unsigned long>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(long long& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<long long>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(unsigned long long& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<unsigned long long>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(float& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<float>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(double& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<double>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(long double& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<long double>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(bool& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<bool>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(void*& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic<void*>(*this, __n);
+}
+
+template <class _Tp, class _CharT, class _Traits>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+__input_arithmetic_with_numeric_limits(basic_istream<_CharT, _Traits>& __is, _Tp& __n) {
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __s(__is);
+    if (__s)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            typedef istreambuf_iterator<_CharT, _Traits> _Ip;
+            typedef num_get<_CharT, _Ip> _Fp;
+            long __temp;
+            use_facet<_Fp>(__is.getloc()).get(_Ip(__is), _Ip(), __is, __state, __temp);
+            if (__temp < numeric_limits<_Tp>::min())
+            {
+                __state |= ios_base::failbit;
+                __n = numeric_limits<_Tp>::min();
+            }
+            else if (__temp > numeric_limits<_Tp>::max())
+            {
+                __state |= ios_base::failbit;
+                __n = numeric_limits<_Tp>::max();
+            }
+            else
+            {
+                __n = static_cast<_Tp>(__temp);
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(short& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic_with_numeric_limits<short>(*this, __n);
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(int& __n)
+{
+    return _CUDA_VSTD::__input_arithmetic_with_numeric_limits<int>(*this, __n);
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+__input_c_string(basic_istream<_CharT, _Traits>& __is, _CharT* __p, size_t __n)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __sen(__is);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif
+            _CharT* __s = __p;
+            const ctype<_CharT>& __ct = use_facet<ctype<_CharT> >(__is.getloc());
+            while (__s != __p + (__n-1))
+            {
+                typename _Traits::int_type __i = __is.rdbuf()->sgetc();
+                if (_Traits::eq_int_type(__i, _Traits::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                _CharT __ch = _Traits::to_char_type(__i);
+                if (__ct.is(__ct.space, __ch))
+                    break;
+                *__s++ = __ch;
+                 __is.rdbuf()->sbumpc();
+            }
+            *__s = _CharT();
+            __is.width(0);
+            if (__s == __p)
+               __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+
+template<class _CharT, class _Traits, size_t _Np>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, _CharT (&__buf)[_Np])
+{
+    size_t __n = _Np;
+    if (__is.width() > 0)
+        __n = _CUDA_VSTD::min(size_t(__is.width()), _Np);
+    return _CUDA_VSTD::__input_c_string(__is, __buf, __n);
+}
+
+template<class _Traits, size_t _Np>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<char, _Traits>&
+operator>>(basic_istream<char, _Traits>& __is, unsigned char (&__buf)[_Np])
+{
+    return __is >> (char(&)[_Np])__buf;
+}
+
+template<class _Traits, size_t _Np>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<char, _Traits>&
+operator>>(basic_istream<char, _Traits>& __is, signed char (&__buf)[_Np])
+{
+    return __is >> (char(&)[_Np])__buf;
+}
+
+#else
+
+template<class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, _CharT* __s)
+{
+    streamsize __n = __is.width();
+    if (__n <= 0)
+        __n = numeric_limits<streamsize>::max() / sizeof(_CharT) - 1;
+    return _CUDA_VSTD::__input_c_string(__is, __s, size_t(__n));
+}
+
+template<class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<char, _Traits>&
+operator>>(basic_istream<char, _Traits>& __is, unsigned char* __s)
+{
+    return __is >> (char*)__s;
+}
+
+template<class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<char, _Traits>&
+operator>>(basic_istream<char, _Traits>& __is, signed char* __s)
+{
+    return __is >> (char*)__s;
+}
+
+#endif  // _LIBCUDACXX_STD_VER > 17
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, _CharT& __c)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __sen(__is);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif
+            typename _Traits::int_type __i = __is.rdbuf()->sbumpc();
+            if (_Traits::eq_int_type(__i, _Traits::eof()))
+                __state |= ios_base::eofbit | ios_base::failbit;
+            else
+                __c = _Traits::to_char_type(__i);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+template<class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<char, _Traits>&
+operator>>(basic_istream<char, _Traits>& __is, unsigned char& __c)
+{
+    return __is >> (char&)__c;
+}
+
+template<class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<char, _Traits>&
+operator>>(basic_istream<char, _Traits>& __is, signed char& __c)
+{
+    return __is >> (char&)__c;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::operator>>(basic_streambuf<char_type, traits_type>* __sb)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __s(*this, true);
+    if (__s)
+    {
+        if (__sb)
+        {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            try
+            {
+#endif // _LIBCUDACXX_NO_EXCEPTIONS
+                while (true)
+                {
+                    typename traits_type::int_type __i = this->rdbuf()->sgetc();
+                    if (traits_type::eq_int_type(__i, _Traits::eof()))
+                    {
+                       __state |= ios_base::eofbit;
+                       break;
+                    }
+                    if (traits_type::eq_int_type(
+                            __sb->sputc(traits_type::to_char_type(__i)),
+                            traits_type::eof()))
+                        break;
+                    ++__gc_;
+                    this->rdbuf()->sbumpc();
+                }
+                if (__gc_ == 0)
+                   __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            }
+            catch (...)
+            {
+                __state |= ios_base::badbit;
+                if (__gc_ == 0)
+                    __state |= ios_base::failbit;
+
+                this->__setstate_nothrow(__state);
+                if (this->exceptions() & ios_base::failbit || this->exceptions() & ios_base::badbit)
+                {
+                    throw;
+                }
+            }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        else
+        {
+            __state |= ios_base::failbit;
+        }
+        this->setstate(__state);
+    }
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+typename basic_istream<_CharT, _Traits>::int_type
+basic_istream<_CharT, _Traits>::get()
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    int_type __r = traits_type::eof();
+    sentry __s(*this, true);
+    if (__s)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif
+            __r = this->rdbuf()->sbumpc();
+            if (traits_type::eq_int_type(__r, traits_type::eof()))
+               __state |= ios_base::failbit | ios_base::eofbit;
+            else
+                __gc_ = 1;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            this->__setstate_nothrow(this->rdstate() | ios_base::badbit);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        this->setstate(__state);
+    }
+    return __r;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::get(char_type* __s, streamsize __n, char_type __dlm)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+        if (__n > 0)
+        {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            try
+            {
+#endif
+                while (__gc_ < __n-1)
+                {
+                    int_type __i = this->rdbuf()->sgetc();
+                    if (traits_type::eq_int_type(__i, traits_type::eof()))
+                    {
+                       __state |= ios_base::eofbit;
+                       break;
+                    }
+                    char_type __ch = traits_type::to_char_type(__i);
+                    if (traits_type::eq(__ch, __dlm))
+                        break;
+                    *__s++ = __ch;
+                    ++__gc_;
+                     this->rdbuf()->sbumpc();
+                }
+                if (__gc_ == 0)
+                   __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            }
+            catch (...)
+            {
+                __state |= ios_base::badbit;
+                this->__setstate_nothrow(__state);
+                if (this->exceptions() & ios_base::badbit)
+                {
+                    if (__n > 0)
+                        *__s = char_type();
+                    throw;
+                }
+            }
+#endif
+        }
+        else
+        {
+            __state |= ios_base::failbit;
+        }
+
+        if (__n > 0)
+            *__s = char_type();
+        this->setstate(__state);
+    }
+    if (__n > 0)
+        *__s = char_type();
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::get(basic_streambuf<char_type, traits_type>& __sb,
+                                    char_type __dlm)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            while (true)
+            {
+                typename traits_type::int_type __i = this->rdbuf()->sgetc();
+                if (traits_type::eq_int_type(__i, traits_type::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                char_type __ch = traits_type::to_char_type(__i);
+                if (traits_type::eq(__ch, __dlm))
+                    break;
+                if (traits_type::eq_int_type(__sb.sputc(__ch), traits_type::eof()))
+                    break;
+                ++__gc_;
+                this->rdbuf()->sbumpc();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            // according to the spec, exceptions here are caught but not rethrown
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        if (__gc_ == 0)
+           __state |= ios_base::failbit;
+        this->setstate(__state);
+    }
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::getline(char_type* __s, streamsize __n, char_type __dlm)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            while (true)
+            {
+                typename traits_type::int_type __i = this->rdbuf()->sgetc();
+                if (traits_type::eq_int_type(__i, traits_type::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                char_type __ch = traits_type::to_char_type(__i);
+                if (traits_type::eq(__ch, __dlm))
+                {
+                    this->rdbuf()->sbumpc();
+                    ++__gc_;
+                    break;
+                }
+                if (__gc_ >= __n-1)
+                {
+                    __state |= ios_base::failbit;
+                    break;
+                }
+                *__s++ = __ch;
+                this->rdbuf()->sbumpc();
+                ++__gc_;
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                if (__n > 0)
+                    *__s = char_type();
+                if (__gc_ == 0)
+                    __state |= ios_base::failbit;
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    if (__n > 0)
+        *__s = char_type();
+    if (__gc_ == 0)
+        __state |= ios_base::failbit;
+    this->setstate(__state);
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::ignore(streamsize __n, int_type __dlm)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (__n == numeric_limits<streamsize>::max())
+            {
+                while (true)
+                {
+                    typename traits_type::int_type __i = this->rdbuf()->sbumpc();
+                    if (traits_type::eq_int_type(__i, traits_type::eof()))
+                    {
+                       __state |= ios_base::eofbit;
+                       break;
+                    }
+                    ++__gc_;
+                    if (traits_type::eq_int_type(__i, __dlm))
+                        break;
+                }
+            }
+            else
+            {
+                while (__gc_ < __n)
+                {
+                    typename traits_type::int_type __i = this->rdbuf()->sbumpc();
+                    if (traits_type::eq_int_type(__i, traits_type::eof()))
+                    {
+                       __state |= ios_base::eofbit;
+                       break;
+                    }
+                    ++__gc_;
+                    if (traits_type::eq_int_type(__i, __dlm))
+                        break;
+                }
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->setstate(__state);
+    }
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+typename basic_istream<_CharT, _Traits>::int_type
+basic_istream<_CharT, _Traits>::peek()
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    int_type __r = traits_type::eof();
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            __r = this->rdbuf()->sgetc();
+            if (traits_type::eq_int_type(__r, traits_type::eof()))
+                __state |= ios_base::eofbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->setstate(__state);
+    }
+    return __r;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::read(char_type* __s, streamsize __n)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            __gc_ = this->rdbuf()->sgetn(__s, __n);
+            if (__gc_ != __n)
+                __state |= ios_base::failbit | ios_base::eofbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    else
+    {
+        __state |= ios_base::failbit;
+    }
+    this->setstate(__state);
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+streamsize
+basic_istream<_CharT, _Traits>::readsome(char_type* __s, streamsize __n)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    __gc_ = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            streamsize __c = this->rdbuf()->in_avail();
+            switch (__c)
+            {
+            case -1:
+                __state |= ios_base::eofbit;
+                break;
+            case 0:
+                break;
+            default:
+                __n = _CUDA_VSTD::min(__c, __n);
+                __gc_ = this->rdbuf()->sgetn(__s, __n);
+                if (__gc_ != __n)
+                    __state |= ios_base::failbit | ios_base::eofbit;
+                break;
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    else
+    {
+        __state |= ios_base::failbit;
+    }
+    this->setstate(__state);
+    return __gc_;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::putback(char_type __c)
+{
+    ios_base::iostate __state = this->rdstate() & ~ios_base::eofbit;
+    __gc_ = 0;
+    this->clear(__state);
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (this->rdbuf() == 0 || this->rdbuf()->sputbackc(__c) == traits_type::eof())
+                __state |= ios_base::badbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    else
+    {
+        __state |= ios_base::failbit;
+    }
+    this->setstate(__state);
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::unget()
+{
+    ios_base::iostate __state = this->rdstate() & ~ios_base::eofbit;
+    __gc_ = 0;
+    this->clear(__state);
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (this->rdbuf() == 0 || this->rdbuf()->sungetc() == traits_type::eof())
+                __state |= ios_base::badbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    else
+    {
+        __state |= ios_base::failbit;
+    }
+    this->setstate(__state);
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+int
+basic_istream<_CharT, _Traits>::sync()
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    int __r = 0;
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (this->rdbuf() == 0)
+                return -1;
+            if (this->rdbuf()->pubsync() == -1)
+            {
+                __state |= ios_base::badbit;
+                return -1;
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->setstate(__state);
+    }
+    return __r;
+}
+
+template<class _CharT, class _Traits>
+typename basic_istream<_CharT, _Traits>::pos_type
+basic_istream<_CharT, _Traits>::tellg()
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    pos_type __r(-1);
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __r = this->rdbuf()->pubseekoff(0, ios_base::cur, ios_base::in);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->setstate(__state);
+    }
+    return __r;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::seekg(pos_type __pos)
+{
+    ios_base::iostate __state = this->rdstate() & ~ios_base::eofbit;
+    this->clear(__state);
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (this->rdbuf()->pubseekpos(__pos, ios_base::in) == pos_type(-1))
+                __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->setstate(__state);
+    }
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+basic_istream<_CharT, _Traits>::seekg(off_type __off, ios_base::seekdir __dir)
+{
+    ios_base::iostate __state = this->rdstate() & ~ios_base::eofbit;
+    this->clear(__state);
+    sentry __sen(*this, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (this->rdbuf()->pubseekoff(__off, __dir, ios_base::in) == pos_type(-1))
+                __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            this->__setstate_nothrow(__state);
+            if (this->exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        this->setstate(__state);
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+ws(basic_istream<_CharT, _Traits>& __is)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __sen(__is, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            const ctype<_CharT>& __ct = use_facet<ctype<_CharT> >(__is.getloc());
+            while (true)
+            {
+                typename _Traits::int_type __i = __is.rdbuf()->sgetc();
+                if (_Traits::eq_int_type(__i, _Traits::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                if (!__ct.is(__ct.space, _Traits::to_char_type(__i)))
+                    break;
+                __is.rdbuf()->sbumpc();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>&& __is, _Tp&& __x)
+{
+    __is >> _CUDA_VSTD::forward<_Tp>(__x);
+    return __is;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_iostream
+    : public basic_istream<_CharT, _Traits>,
+      public basic_ostream<_CharT, _Traits>
+{
+public:
+    // types:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // constructor/destructor
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    explicit basic_iostream(basic_streambuf<char_type, traits_type>* __sb)
+      : basic_istream<_CharT, _Traits>(__sb)
+    {}
+
+    virtual ~basic_iostream();
+protected:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_iostream(basic_iostream&& __rhs);
+
+    // assign/swap
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_iostream& operator=(basic_iostream&& __rhs);
+#endif
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void swap(basic_iostream& __rhs)
+    { basic_istream<char_type, traits_type>::swap(__rhs); }
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_iostream<_CharT, _Traits>::basic_iostream(basic_iostream&& __rhs)
+    : basic_istream<_CharT, _Traits>(_CUDA_VSTD::move(__rhs))
+{
+}
+
+template <class _CharT, class _Traits>
+basic_iostream<_CharT, _Traits>&
+basic_iostream<_CharT, _Traits>::operator=(basic_iostream&& __rhs)
+{
+    swap(__rhs);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_iostream<_CharT, _Traits>::~basic_iostream()
+{
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           basic_string<_CharT, _Traits, _Allocator>& __str)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __sen(__is);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif
+            __str.clear();
+            streamsize __n = __is.width();
+            if (__n <= 0)
+                __n = __str.max_size();
+            if (__n <= 0)
+                __n = numeric_limits<streamsize>::max();
+            streamsize __c = 0;
+            const ctype<_CharT>& __ct = use_facet<ctype<_CharT> >(__is.getloc());
+            while (__c < __n)
+            {
+                typename _Traits::int_type __i = __is.rdbuf()->sgetc();
+                if (_Traits::eq_int_type(__i, _Traits::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                _CharT __ch = _Traits::to_char_type(__i);
+                if (__ct.is(__ct.space, __ch))
+                    break;
+                __str.push_back(__ch);
+                ++__c;
+                 __is.rdbuf()->sbumpc();
+            }
+            __is.width(0);
+            if (__c == 0)
+               __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str, _CharT __dlm)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __sen(__is, true);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif
+            __str.clear();
+            streamsize __extr = 0;
+            while (true)
+            {
+                typename _Traits::int_type __i = __is.rdbuf()->sbumpc();
+                if (_Traits::eq_int_type(__i, _Traits::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                ++__extr;
+                _CharT __ch = _Traits::to_char_type(__i);
+                if (_Traits::eq(__ch, __dlm))
+                    break;
+                __str.push_back(__ch);
+                if (__str.size() == __str.max_size())
+                {
+                    __state |= ios_base::failbit;
+                    break;
+                }
+            }
+            if (__extr == 0)
+               __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str)
+{
+    return getline(__is, __str, __is.widen('\n'));
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>&& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str, _CharT __dlm)
+{
+    return getline(__is, __str, __dlm);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>&& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str)
+{
+    return getline(__is, __str, __is.widen('\n'));
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, size_t _Size>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, bitset<_Size>& __x)
+{
+    ios_base::iostate __state = ios_base::goodbit;
+    typename basic_istream<_CharT, _Traits>::sentry __sen(__is);
+    if (__sen)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif
+            basic_string<_CharT, _Traits> __str;
+            const ctype<_CharT>& __ct = use_facet<ctype<_CharT> >(__is.getloc());
+            size_t __c = 0;
+            _CharT __zero = __ct.widen('0');
+            _CharT __one = __ct.widen('1');
+            while (__c < _Size)
+            {
+                typename _Traits::int_type __i = __is.rdbuf()->sgetc();
+                if (_Traits::eq_int_type(__i, _Traits::eof()))
+                {
+                   __state |= ios_base::eofbit;
+                   break;
+                }
+                _CharT __ch = _Traits::to_char_type(__i);
+                if (!_Traits::eq(__ch, __zero) && !_Traits::eq(__ch, __one))
+                    break;
+                __str.push_back(__ch);
+                ++__c;
+                 __is.rdbuf()->sbumpc();
+            }
+            __x = bitset<_Size>(__str);
+            if (_Size > 0 && __c == 0)
+               __state |= ios_base::failbit;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __state |= ios_base::badbit;
+            __is.__setstate_nothrow(__state);
+            if (__is.exceptions() & ios_base::badbit)
+            {
+                throw;
+            }
+        }
+#endif
+        __is.setstate(__state);
+    }
+    return __is;
+}
+
+#ifndef _LIBCUDACXX_DO_NOT_ASSUME_STREAMS_EXPLICIT_INSTANTIATION_IN_DYLIB
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_istream<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_istream<wchar_t>)
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_iostream<char>)
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_ISTREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iterator b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iterator
new file mode 100644
index 000000000000..95f9fe431453
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/iterator
@@ -0,0 +1,1943 @@
+// -*- C++ -*-
+//===-------------------------- iterator ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ITERATOR
+#define _LIBCUDACXX_ITERATOR
+
+/*
+    iterator synopsis
+
+namespace std
+{
+
+template<class Iterator>
+struct iterator_traits
+{
+    typedef typename Iterator::difference_type difference_type;
+    typedef typename Iterator::value_type value_type;
+    typedef typename Iterator::pointer pointer;
+    typedef typename Iterator::reference reference;
+    typedef typename Iterator::iterator_category iterator_category;
+};
+
+template<class T>
+struct iterator_traits<T*>
+{
+    typedef ptrdiff_t difference_type;
+    typedef T value_type;
+    typedef T* pointer;
+    typedef T& reference;
+    typedef random_access_iterator_tag iterator_category;
+};
+
+template<class Category, class T, class Distance = ptrdiff_t,
+         class Pointer = T*, class Reference = T&>
+struct iterator
+{
+    typedef T         value_type;
+    typedef Distance  difference_type;
+    typedef Pointer   pointer;
+    typedef Reference reference;
+    typedef Category  iterator_category;
+};
+
+struct input_iterator_tag  {};
+struct output_iterator_tag {};
+struct forward_iterator_tag       : public input_iterator_tag         {};
+struct bidirectional_iterator_tag : public forward_iterator_tag       {};
+struct random_access_iterator_tag : public bidirectional_iterator_tag {};
+
+// 27.4.3, iterator operations
+// extension: second argument not conforming to C++03
+template <class InputIterator>  // constexpr in C++17
+  constexpr void advance(InputIterator& i,
+             typename iterator_traits<InputIterator>::difference_type n);
+
+template <class InputIterator>  // constexpr in C++17
+  constexpr typename iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last);
+
+template <class InputIterator>  // constexpr in C++17
+  constexpr InputIterator next(InputIterator x,
+typename iterator_traits<InputIterator>::difference_type n = 1);
+
+template <class BidirectionalIterator>  // constexpr in C++17
+  constexpr BidirectionalIterator prev(BidirectionalIterator x,
+    typename iterator_traits<BidirectionalIterator>::difference_type n = 1);
+
+template <class Iterator>
+class reverse_iterator
+    : public iterator<typename iterator_traits<Iterator>::iterator_category,
+                      typename iterator_traits<Iterator>::value_type,
+                      typename iterator_traits<Iterator>::difference_type,
+                      typename iterator_traits<Iterator>::pointer,
+                      typename iterator_traits<Iterator>::reference>
+{
+protected:
+    Iterator current;
+public:
+    typedef Iterator                                            iterator_type;
+    typedef typename iterator_traits<Iterator>::difference_type difference_type;
+    typedef typename iterator_traits<Iterator>::reference       reference;
+    typedef typename iterator_traits<Iterator>::pointer         pointer;
+
+    constexpr reverse_iterator();
+    constexpr explicit reverse_iterator(Iterator x);
+    template <class U> constexpr reverse_iterator(const reverse_iterator<U>& u);
+    template <class U> constexpr reverse_iterator& operator=(const reverse_iterator<U>& u);
+    constexpr Iterator base() const;
+    constexpr reference operator*() const;
+    constexpr pointer   operator->() const;
+    constexpr reverse_iterator& operator++();
+    constexpr reverse_iterator  operator++(int);
+    constexpr reverse_iterator& operator--();
+    constexpr reverse_iterator  operator--(int);
+    constexpr reverse_iterator  operator+ (difference_type n) const;
+    constexpr reverse_iterator& operator+=(difference_type n);
+    constexpr reverse_iterator  operator- (difference_type n) const;
+    constexpr reverse_iterator& operator-=(difference_type n);
+    constexpr reference         operator[](difference_type n) const;
+};
+
+template <class Iterator1, class Iterator2>
+constexpr bool                          // constexpr in C++17
+operator==(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool                          // constexpr in C++17
+operator<(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool                          // constexpr in C++17
+operator!=(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool                          // constexpr in C++17
+operator>(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool                          // constexpr in C++17
+operator>=(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool                          // constexpr in C++17
+operator<=(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr auto
+operator-(const reverse_iterator<Iterator1>& x, const reverse_iterator<Iterator2>& y)
+-> decltype(__y.base() - __x.base());   // constexpr in C++17
+
+template <class Iterator>
+constexpr reverse_iterator<Iterator>
+operator+(typename reverse_iterator<Iterator>::difference_type n,
+          const reverse_iterator<Iterator>& x);   // constexpr in C++17
+
+template <class Iterator>
+constexpr reverse_iterator<Iterator> make_reverse_iterator(Iterator i); // C++14, constexpr in C++17
+
+template <class Container>
+class back_insert_iterator
+{
+protected:
+    Container* container;
+public:
+    typedef Container                   container_type;
+    typedef void                        value_type;
+    typedef void                        difference_type;
+    typedef void                        reference;
+    typedef void                        pointer;
+
+    explicit back_insert_iterator(Container& x);
+    back_insert_iterator& operator=(const typename Container::value_type& value);
+    back_insert_iterator& operator*();
+    back_insert_iterator& operator++();
+    back_insert_iterator  operator++(int);
+};
+
+template <class Container> back_insert_iterator<Container> back_inserter(Container& x);
+
+template <class Container>
+class front_insert_iterator
+{
+protected:
+    Container* container;
+public:
+    typedef Container                    container_type;
+    typedef void                         value_type;
+    typedef void                         difference_type;
+    typedef void                         reference;
+    typedef void                         pointer;
+
+    explicit front_insert_iterator(Container& x);
+    front_insert_iterator& operator=(const typename Container::value_type& value);
+    front_insert_iterator& operator*();
+    front_insert_iterator& operator++();
+    front_insert_iterator  operator++(int);
+};
+
+template <class Container> front_insert_iterator<Container> front_inserter(Container& x);
+
+template <class Container>
+class insert_iterator
+{
+protected:
+    Container* container;
+    typename Container::iterator iter;
+public:
+    typedef Container              container_type;
+    typedef void                   value_type;
+    typedef void                   difference_type;
+    typedef void                   reference;
+    typedef void                   pointer;
+
+    insert_iterator(Container& x, typename Container::iterator i);
+    insert_iterator& operator=(const typename Container::value_type& value);
+    insert_iterator& operator*();
+    insert_iterator& operator++();
+    insert_iterator& operator++(int);
+};
+
+template <class Container, class Iterator>
+insert_iterator<Container> inserter(Container& x, Iterator i);
+
+template <class Iterator>
+class move_iterator {
+public:
+    typedef Iterator                                              iterator_type;
+    typedef typename iterator_traits<Iterator>::difference_type   difference_type;
+    typedef Iterator                                              pointer;
+    typedef typename iterator_traits<Iterator>::value_type        value_type;
+    typedef typename iterator_traits<Iterator>::iterator_category iterator_category;
+    typedef value_type&&                                          reference;
+
+    constexpr move_iterator();  // all the constexprs are in C++17
+    constexpr explicit move_iterator(Iterator i);
+    template <class U>
+      constexpr move_iterator(const move_iterator<U>& u);
+    template <class U>
+      constexpr move_iterator& operator=(const move_iterator<U>& u);
+    constexpr iterator_type base() const;
+    constexpr reference operator*() const;
+    constexpr pointer operator->() const;
+    constexpr move_iterator& operator++();
+    constexpr move_iterator operator++(int);
+    constexpr move_iterator& operator--();
+    constexpr move_iterator operator--(int);
+    constexpr move_iterator operator+(difference_type n) const;
+    constexpr move_iterator& operator+=(difference_type n);
+    constexpr move_iterator operator-(difference_type n) const;
+    constexpr move_iterator& operator-=(difference_type n);
+    constexpr unspecified operator[](difference_type n) const;
+private:
+    Iterator current; // exposition only
+};
+
+template <class Iterator1, class Iterator2>
+constexpr bool   // constexpr in C++17
+operator==(const move_iterator<Iterator1>& x, const move_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool   // constexpr in C++17
+operator!=(const move_iterator<Iterator1>& x, const move_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool   // constexpr in C++17
+operator<(const move_iterator<Iterator1>& x, const move_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool   // constexpr in C++17
+operator<=(const move_iterator<Iterator1>& x, const move_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool   // constexpr in C++17
+operator>(const move_iterator<Iterator1>& x, const move_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr bool   // constexpr in C++17
+operator>=(const move_iterator<Iterator1>& x, const move_iterator<Iterator2>& y);
+
+template <class Iterator1, class Iterator2>
+constexpr auto   // constexpr in C++17
+operator-(const move_iterator<Iterator1>& x,
+          const move_iterator<Iterator2>& y) -> decltype(x.base() - y.base());
+
+template <class Iterator>
+constexpr move_iterator<Iterator> operator+(   // constexpr in C++17
+            typename move_iterator<Iterator>::difference_type n,
+            const move_iterator<Iterator>& x);
+
+template <class Iterator>   // constexpr in C++17
+constexpr  move_iterator<Iterator> make_move_iterator(const Iterator& i);
+
+
+template <class T, class charT = char, class traits = char_traits<charT>, class Distance = ptrdiff_t>
+class istream_iterator
+    : public iterator<input_iterator_tag, T, Distance, const T*, const T&>
+{
+public:
+    typedef charT char_type;
+    typedef traits traits_type;
+    typedef basic_istream<charT,traits> istream_type;
+
+    constexpr istream_iterator();
+    istream_iterator(istream_type& s);
+    istream_iterator(const istream_iterator& x);
+    ~istream_iterator();
+
+    const T& operator*() const;
+    const T* operator->() const;
+    istream_iterator& operator++();
+    istream_iterator  operator++(int);
+};
+
+template <class T, class charT, class traits, class Distance>
+bool operator==(const istream_iterator<T,charT,traits,Distance>& x,
+                const istream_iterator<T,charT,traits,Distance>& y);
+template <class T, class charT, class traits, class Distance>
+bool operator!=(const istream_iterator<T,charT,traits,Distance>& x,
+                const istream_iterator<T,charT,traits,Distance>& y);
+
+template <class T, class charT = char, class traits = char_traits<charT> >
+class ostream_iterator
+    : public iterator<output_iterator_tag, void, void, void ,void>
+{
+public:
+    typedef charT char_type;
+    typedef traits traits_type;
+    typedef basic_ostream<charT,traits> ostream_type;
+
+    ostream_iterator(ostream_type& s);
+    ostream_iterator(ostream_type& s, const charT* delimiter);
+    ostream_iterator(const ostream_iterator& x);
+    ~ostream_iterator();
+    ostream_iterator& operator=(const T& value);
+
+    ostream_iterator& operator*();
+    ostream_iterator& operator++();
+    ostream_iterator& operator++(int);
+};
+
+template<class charT, class traits = char_traits<charT> >
+class istreambuf_iterator
+    : public iterator<input_iterator_tag, charT,
+                      typename traits::off_type, unspecified,
+                      charT>
+{
+public:
+    typedef charT                         char_type;
+    typedef traits                        traits_type;
+    typedef typename traits::int_type     int_type;
+    typedef basic_streambuf<charT,traits> streambuf_type;
+    typedef basic_istream<charT,traits>   istream_type;
+
+    istreambuf_iterator() noexcept;
+    istreambuf_iterator(istream_type& s) noexcept;
+    istreambuf_iterator(streambuf_type* s) noexcept;
+    istreambuf_iterator(a-private-type) noexcept;
+
+    charT                operator*() const;
+    pointer operator->() const;
+    istreambuf_iterator& operator++();
+    a-private-type       operator++(int);
+
+    bool equal(const istreambuf_iterator& b) const;
+};
+
+template <class charT, class traits>
+bool operator==(const istreambuf_iterator<charT,traits>& a,
+                const istreambuf_iterator<charT,traits>& b);
+template <class charT, class traits>
+bool operator!=(const istreambuf_iterator<charT,traits>& a,
+                const istreambuf_iterator<charT,traits>& b);
+
+template <class charT, class traits = char_traits<charT> >
+class ostreambuf_iterator
+    : public iterator<output_iterator_tag, void, void, void, void>
+{
+public:
+    typedef charT                         char_type;
+    typedef traits                        traits_type;
+    typedef basic_streambuf<charT,traits> streambuf_type;
+    typedef basic_ostream<charT,traits>   ostream_type;
+
+    ostreambuf_iterator(ostream_type& s) noexcept;
+    ostreambuf_iterator(streambuf_type* s) noexcept;
+    ostreambuf_iterator& operator=(charT c);
+    ostreambuf_iterator& operator*();
+    ostreambuf_iterator& operator++();
+    ostreambuf_iterator& operator++(int);
+    bool failed() const noexcept;
+};
+
+template <class C> constexpr auto begin(C& c) -> decltype(c.begin());
+template <class C> constexpr auto begin(const C& c) -> decltype(c.begin());
+template <class C> constexpr auto end(C& c) -> decltype(c.end());
+template <class C> constexpr auto end(const C& c) -> decltype(c.end());
+template <class T, size_t N> constexpr T* begin(T (&array)[N]);
+template <class T, size_t N> constexpr T* end(T (&array)[N]);
+
+template <class C> auto constexpr cbegin(const C& c) -> decltype(std::begin(c));        // C++14
+template <class C> auto constexpr cend(const C& c) -> decltype(std::end(c));            // C++14
+template <class C> auto constexpr rbegin(C& c) -> decltype(c.rbegin());                 // C++14
+template <class C> auto constexpr rbegin(const C& c) -> decltype(c.rbegin());           // C++14
+template <class C> auto constexpr rend(C& c) -> decltype(c.rend());                     // C++14
+template <class C> constexpr auto rend(const C& c) -> decltype(c.rend());               // C++14
+template <class E> reverse_iterator<const E*> constexpr rbegin(initializer_list<E> il); // C++14
+template <class E> reverse_iterator<const E*> constexpr rend(initializer_list<E> il);   // C++14
+template <class T, size_t N> reverse_iterator<T*> constexpr rbegin(T (&array)[N]);      // C++14
+template <class T, size_t N> reverse_iterator<T*> constexpr rend(T (&array)[N]);        // C++14
+template <class C> constexpr auto crbegin(const C& c) -> decltype(std::rbegin(c));      // C++14
+template <class C> constexpr auto crend(const C& c) -> decltype(std::rend(c));          // C++14
+
+// 24.8, container access:
+template <class C> constexpr auto size(const C& c) -> decltype(c.size());         // C++17
+template <class T, size_t N> constexpr size_t size(const T (&array)[N]) noexcept; // C++17
+
+template <class C> constexpr auto ssize(const C& c)
+    -> common_type_t<ptrdiff_t, make_signed_t<decltype(c.size())>>;				       // C++20
+template <class T, ptrdiff_t> constexpr ptrdiff_t ssize(const T (&array)[N]) noexcept; // C++20
+
+template <class C> constexpr auto empty(const C& c) -> decltype(c.empty());       // C++17
+template <class T, size_t N> constexpr bool empty(const T (&array)[N]) noexcept;  // C++17
+template <class E> constexpr bool empty(initializer_list<E> il) noexcept;         // C++17
+template <class C> constexpr auto data(C& c) -> decltype(c.data());               // C++17
+template <class C> constexpr auto data(const C& c) -> decltype(c.data());         // C++17
+template <class T, size_t N> constexpr T* data(T (&array)[N]) noexcept;           // C++17
+template <class E> constexpr const E* data(initializer_list<E> il) noexcept;      // C++17
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <iosfwd> // for forward declarations of vector and string.
+#include <__functional_base>
+#include <type_traits>
+#include <cstddef>
+#include <initializer_list>
+#include <version>
+#ifdef __APPLE__
+#include <Availability.h>
+#endif
+#include <__debug>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+struct _LIBCUDACXX_TEMPLATE_VIS input_iterator_tag {};
+struct _LIBCUDACXX_TEMPLATE_VIS output_iterator_tag {};
+struct _LIBCUDACXX_TEMPLATE_VIS forward_iterator_tag       : public input_iterator_tag {};
+struct _LIBCUDACXX_TEMPLATE_VIS bidirectional_iterator_tag : public forward_iterator_tag {};
+struct _LIBCUDACXX_TEMPLATE_VIS random_access_iterator_tag : public bidirectional_iterator_tag {};
+
+template <class _Tp>
+struct __has_iterator_typedefs
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    template <class _Up> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER static __two __test(...);
+    template <class _Up> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER static char __test(typename std::__void_t<typename _Up::iterator_category>::type* = 0,
+    										typename std::__void_t<typename _Up::difference_type>::type* = 0,
+    										typename std::__void_t<typename _Up::value_type>::type* = 0,
+    										typename std::__void_t<typename _Up::reference>::type* = 0,
+    										typename std::__void_t<typename _Up::pointer>::type* = 0
+    										);
+public:
+    static const bool value = sizeof(__test<_Tp>(0,0,0,0,0)) == 1;
+};
+
+
+template <class _Tp>
+struct __has_iterator_category
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    template <class _Up> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER static __two __test(...);
+    template <class _Up> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER static char __test(typename _Up::iterator_category* = 0);
+public:
+    static const bool value = sizeof(__test<_Tp>(0)) == 1;
+};
+
+template <class _Iter, bool> struct __iterator_traits_impl {};
+
+template <class _Iter>
+struct __iterator_traits_impl<_Iter, true>
+{
+    typedef typename _Iter::difference_type   difference_type;
+    typedef typename _Iter::value_type        value_type;
+    typedef typename _Iter::pointer           pointer;
+    typedef typename _Iter::reference         reference;
+    typedef typename _Iter::iterator_category iterator_category;
+};
+
+template <class _Iter, bool> struct __iterator_traits {};
+
+template <class _Iter>
+struct __iterator_traits<_Iter, true>
+    :  __iterator_traits_impl
+      <
+        _Iter,
+        is_convertible<typename _Iter::iterator_category, input_iterator_tag>::value ||
+        is_convertible<typename _Iter::iterator_category, output_iterator_tag>::value
+      >
+{};
+
+// iterator_traits<Iterator> will only have the nested types if Iterator::iterator_category
+//    exists.  Else iterator_traits<Iterator> will be an empty class.  This is a
+//    conforming extension which allows some programs to compile and behave as
+//    the client expects instead of failing at compile time.
+
+template <class _Iter>
+struct _LIBCUDACXX_TEMPLATE_VIS iterator_traits
+    : __iterator_traits<_Iter, __has_iterator_typedefs<_Iter>::value> {};
+
+template<class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS iterator_traits<_Tp*>
+{
+    typedef ptrdiff_t difference_type;
+    typedef typename remove_cv<_Tp>::type value_type;
+    typedef _Tp* pointer;
+    typedef _Tp& reference;
+    typedef random_access_iterator_tag iterator_category;
+};
+
+template <class _Tp, class _Up, bool = __has_iterator_category<iterator_traits<_Tp> >::value>
+struct __has_iterator_category_convertible_to
+    : public integral_constant<bool, is_convertible<typename iterator_traits<_Tp>::iterator_category, _Up>::value>
+{};
+
+template <class _Tp, class _Up>
+struct __has_iterator_category_convertible_to<_Tp, _Up, false> : public false_type {};
+
+template <class _Tp>
+struct __is_input_iterator : public __has_iterator_category_convertible_to<_Tp, input_iterator_tag> {};
+
+template <class _Tp>
+struct __is_forward_iterator : public __has_iterator_category_convertible_to<_Tp, forward_iterator_tag> {};
+
+template <class _Tp>
+struct __is_bidirectional_iterator : public __has_iterator_category_convertible_to<_Tp, bidirectional_iterator_tag> {};
+
+template <class _Tp>
+struct __is_random_access_iterator : public __has_iterator_category_convertible_to<_Tp, random_access_iterator_tag> {};
+
+template <class _Tp>
+struct __is_exactly_input_iterator
+    : public integral_constant<bool,
+         __has_iterator_category_convertible_to<_Tp, input_iterator_tag>::value &&
+        !__has_iterator_category_convertible_to<_Tp, forward_iterator_tag>::value> {};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator>
+using __iter_value_type = typename iterator_traits<_InputIterator>::value_type;
+
+template<class _InputIterator>
+using __iter_key_type = remove_const_t<typename iterator_traits<_InputIterator>::value_type::first_type>;
+
+template<class _InputIterator>
+using __iter_mapped_type = typename iterator_traits<_InputIterator>::value_type::second_type;
+
+template<class _InputIterator>
+using __iter_to_alloc_type = pair<
+    add_const_t<typename iterator_traits<_InputIterator>::value_type::first_type>,
+    typename iterator_traits<_InputIterator>::value_type::second_type>;
+#endif
+
+template<class _Category, class _Tp, class _Distance = ptrdiff_t,
+         class _Pointer = _Tp*, class _Reference = _Tp&>
+struct _LIBCUDACXX_TEMPLATE_VIS iterator
+{
+    typedef _Tp        value_type;
+    typedef _Distance  difference_type;
+    typedef _Pointer   pointer;
+    typedef _Reference reference;
+    typedef _Category  iterator_category;
+};
+
+template <class _InputIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+void __advance(_InputIter& __i,
+             typename iterator_traits<_InputIter>::difference_type __n, input_iterator_tag)
+{
+    for (; __n > 0; --__n)
+        ++__i;
+}
+
+template <class _BiDirIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+void __advance(_BiDirIter& __i,
+             typename iterator_traits<_BiDirIter>::difference_type __n, bidirectional_iterator_tag)
+{
+    if (__n >= 0)
+        for (; __n > 0; --__n)
+            ++__i;
+    else
+        for (; __n < 0; ++__n)
+            --__i;
+}
+
+template <class _RandIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+void __advance(_RandIter& __i,
+             typename iterator_traits<_RandIter>::difference_type __n, random_access_iterator_tag)
+{
+   __i += __n;
+}
+
+template <class _InputIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+void advance(_InputIter& __i,
+             typename iterator_traits<_InputIter>::difference_type __n)
+{
+    _LIBCUDACXX_ASSERT(__n >= 0 || __is_bidirectional_iterator<_InputIter>::value,
+                       "Attempt to advance(it, -n) on a non-bidi iterator");
+    __advance(__i, __n, typename iterator_traits<_InputIter>::iterator_category());
+}
+
+template <class _InputIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+typename iterator_traits<_InputIter>::difference_type
+__distance(_InputIter __first, _InputIter __last, input_iterator_tag)
+{
+    typename iterator_traits<_InputIter>::difference_type __r(0);
+    for (; __first != __last; ++__first)
+        ++__r;
+    return __r;
+}
+
+template <class _RandIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+typename iterator_traits<_RandIter>::difference_type
+__distance(_RandIter __first, _RandIter __last, random_access_iterator_tag)
+{
+    return __last - __first;
+}
+
+template <class _InputIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+typename iterator_traits<_InputIter>::difference_type
+distance(_InputIter __first, _InputIter __last)
+{
+    return __distance(__first, __last, typename iterator_traits<_InputIter>::iterator_category());
+}
+
+template <class _InputIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+typename enable_if
+<
+    __is_input_iterator<_InputIter>::value,
+    _InputIter
+>::type
+next(_InputIter __x,
+     typename iterator_traits<_InputIter>::difference_type __n = 1)
+{
+    _LIBCUDACXX_ASSERT(__n >= 0 || __is_bidirectional_iterator<_InputIter>::value,
+                       "Attempt to next(it, -n) on a non-bidi iterator");
+
+    _CUDA_VSTD::advance(__x, __n);
+    return __x;
+}
+
+template <class _InputIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+typename enable_if
+<
+    __is_input_iterator<_InputIter>::value,
+    _InputIter
+>::type
+prev(_InputIter __x,
+     typename iterator_traits<_InputIter>::difference_type __n = 1)
+{
+    _LIBCUDACXX_ASSERT(__n <= 0 || __is_bidirectional_iterator<_InputIter>::value,
+                       "Attempt to prev(it, +n) on a non-bidi iterator");
+    _CUDA_VSTD::advance(__x, -__n);
+    return __x;
+}
+
+
+template <class _Tp, class = void>
+struct __is_stashing_iterator : false_type {};
+
+template <class _Tp>
+struct __is_stashing_iterator<_Tp, typename __void_t<typename _Tp::__stashing_iterator_tag>::type>
+  : true_type {};
+
+template <class _Iter>
+class _LIBCUDACXX_TEMPLATE_VIS reverse_iterator
+    : public iterator<typename iterator_traits<_Iter>::iterator_category,
+                      typename iterator_traits<_Iter>::value_type,
+                      typename iterator_traits<_Iter>::difference_type,
+                      typename iterator_traits<_Iter>::pointer,
+                      typename iterator_traits<_Iter>::reference>
+{
+private:
+    /*mutable*/ _Iter __t;  // no longer used as of LWG #2360, not removed due to ABI break
+
+    static_assert(!__is_stashing_iterator<_Iter>::value,
+      "The specified iterator type cannot be used with reverse_iterator; "
+      "Using stashing iterators with reverse_iterator causes undefined behavior");
+
+protected:
+    _Iter current;
+public:
+    typedef _Iter                                            iterator_type;
+    typedef typename iterator_traits<_Iter>::difference_type difference_type;
+    typedef typename iterator_traits<_Iter>::reference       reference;
+    typedef typename iterator_traits<_Iter>::pointer         pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator() : __t(), current() {}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    explicit reverse_iterator(_Iter __x) : __t(__x), current(__x) {}
+    template <class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+        reverse_iterator(const reverse_iterator<_Up>& __u) : __t(__u.base()), current(__u.base()) {}
+    template <class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+        reverse_iterator& operator=(const reverse_iterator<_Up>& __u)
+            { __t = current = __u.base(); return *this; }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    _Iter base() const {return current;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reference operator*() const {_Iter __tmp = current; return *--__tmp;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    pointer  operator->() const {return _CUDA_VSTD::addressof(operator*());}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator& operator++() {--current; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator  operator++(int) {reverse_iterator __tmp(*this); --current; return __tmp;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator& operator--() {++current; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator  operator--(int) {reverse_iterator __tmp(*this); ++current; return __tmp;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator  operator+ (difference_type __n) const {return reverse_iterator(current - __n);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator& operator+=(difference_type __n) {current -= __n; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator  operator- (difference_type __n) const {return reverse_iterator(current + __n);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reverse_iterator& operator-=(difference_type __n) {current += __n; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reference         operator[](difference_type __n) const {return *(*this + __n);}
+};
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator==(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __x.base() == __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator<(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __x.base() > __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator!=(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __x.base() != __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator>(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __x.base() < __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator>=(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __x.base() <= __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator<=(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __x.base() >= __y.base();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto
+operator-(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+-> decltype(__y.base() - __x.base())
+{
+    return __y.base() - __x.base();
+}
+#else
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename reverse_iterator<_Iter1>::difference_type
+operator-(const reverse_iterator<_Iter1>& __x, const reverse_iterator<_Iter2>& __y)
+{
+    return __y.base() - __x.base();
+}
+#endif
+
+template <class _Iter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+reverse_iterator<_Iter>
+operator+(typename reverse_iterator<_Iter>::difference_type __n, const reverse_iterator<_Iter>& __x)
+{
+    return reverse_iterator<_Iter>(__x.base() - __n);
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Iter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+reverse_iterator<_Iter> make_reverse_iterator(_Iter __i)
+{
+    return reverse_iterator<_Iter>(__i);
+}
+#endif
+
+template <class _Container>
+class _LIBCUDACXX_TEMPLATE_VIS back_insert_iterator
+    : public iterator<output_iterator_tag,
+                      void,
+                      void,
+                      void,
+                      void>
+{
+protected:
+    _Container* container;
+public:
+    typedef _Container container_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY explicit back_insert_iterator(_Container& __x) : container(_CUDA_VSTD::addressof(__x)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY back_insert_iterator& operator=(const typename _Container::value_type& __value_)
+        {container->push_back(__value_); return *this;}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY back_insert_iterator& operator=(typename _Container::value_type&& __value_)
+        {container->push_back(_CUDA_VSTD::move(__value_)); return *this;}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY back_insert_iterator& operator*()     {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY back_insert_iterator& operator++()    {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY back_insert_iterator  operator++(int) {return *this;}
+};
+
+template <class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+back_insert_iterator<_Container>
+back_inserter(_Container& __x)
+{
+    return back_insert_iterator<_Container>(__x);
+}
+
+template <class _Container>
+class _LIBCUDACXX_TEMPLATE_VIS front_insert_iterator
+    : public iterator<output_iterator_tag,
+                      void,
+                      void,
+                      void,
+                      void>
+{
+protected:
+    _Container* container;
+public:
+    typedef _Container container_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY explicit front_insert_iterator(_Container& __x) : container(_CUDA_VSTD::addressof(__x)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY front_insert_iterator& operator=(const typename _Container::value_type& __value_)
+        {container->push_front(__value_); return *this;}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY front_insert_iterator& operator=(typename _Container::value_type&& __value_)
+        {container->push_front(_CUDA_VSTD::move(__value_)); return *this;}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY front_insert_iterator& operator*()     {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY front_insert_iterator& operator++()    {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY front_insert_iterator  operator++(int) {return *this;}
+};
+
+template <class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+front_insert_iterator<_Container>
+front_inserter(_Container& __x)
+{
+    return front_insert_iterator<_Container>(__x);
+}
+
+template <class _Container>
+class _LIBCUDACXX_TEMPLATE_VIS insert_iterator
+    : public iterator<output_iterator_tag,
+                      void,
+                      void,
+                      void,
+                      void>
+{
+protected:
+    _Container* container;
+    typename _Container::iterator iter;
+public:
+    typedef _Container container_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY insert_iterator(_Container& __x, typename _Container::iterator __i)
+        : container(_CUDA_VSTD::addressof(__x)), iter(__i) {}
+    _LIBCUDACXX_INLINE_VISIBILITY insert_iterator& operator=(const typename _Container::value_type& __value_)
+        {iter = container->insert(iter, __value_); ++iter; return *this;}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY insert_iterator& operator=(typename _Container::value_type&& __value_)
+        {iter = container->insert(iter, _CUDA_VSTD::move(__value_)); ++iter; return *this;}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY insert_iterator& operator*()        {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY insert_iterator& operator++()       {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY insert_iterator& operator++(int)    {return *this;}
+};
+
+template <class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+insert_iterator<_Container>
+inserter(_Container& __x, typename _Container::iterator __i)
+{
+    return insert_iterator<_Container>(__x, __i);
+}
+
+template <class _Tp, class _CharT = char,
+          class _Traits = char_traits<_CharT>, class _Distance = ptrdiff_t>
+class _LIBCUDACXX_TEMPLATE_VIS istream_iterator
+    : public iterator<input_iterator_tag, _Tp, _Distance, const _Tp*, const _Tp&>
+{
+public:
+    typedef _CharT char_type;
+    typedef _Traits traits_type;
+    typedef basic_istream<_CharT,_Traits> istream_type;
+private:
+    istream_type* __in_stream_;
+    _Tp __value_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR istream_iterator() : __in_stream_(0), __value_() {}
+    _LIBCUDACXX_INLINE_VISIBILITY istream_iterator(istream_type& __s) : __in_stream_(_CUDA_VSTD::addressof(__s))
+        {
+            if (!(*__in_stream_ >> __value_))
+                __in_stream_ = 0;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY const _Tp& operator*() const {return __value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY const _Tp* operator->() const {return _CUDA_VSTD::addressof((operator*()));}
+    _LIBCUDACXX_INLINE_VISIBILITY istream_iterator& operator++()
+        {
+            if (!(*__in_stream_ >> __value_))
+                __in_stream_ = 0;
+            return *this;
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY istream_iterator  operator++(int)
+        {istream_iterator __t(*this); ++(*this); return __t;}
+
+    template <class _Up, class _CharU, class _TraitsU, class _DistanceU>
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool
+    operator==(const istream_iterator<_Up, _CharU, _TraitsU, _DistanceU>& __x,
+               const istream_iterator<_Up, _CharU, _TraitsU, _DistanceU>& __y);
+
+    template <class _Up, class _CharU, class _TraitsU, class _DistanceU>
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool
+    operator==(const istream_iterator<_Up, _CharU, _TraitsU, _DistanceU>& __x,
+               const istream_iterator<_Up, _CharU, _TraitsU, _DistanceU>& __y);
+};
+
+template <class _Tp, class _CharT, class _Traits, class _Distance>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const istream_iterator<_Tp, _CharT, _Traits, _Distance>& __x,
+           const istream_iterator<_Tp, _CharT, _Traits, _Distance>& __y)
+{
+    return __x.__in_stream_ == __y.__in_stream_;
+}
+
+template <class _Tp, class _CharT, class _Traits, class _Distance>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const istream_iterator<_Tp, _CharT, _Traits, _Distance>& __x,
+           const istream_iterator<_Tp, _CharT, _Traits, _Distance>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _CharT = char, class _Traits = char_traits<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS ostream_iterator
+    : public iterator<output_iterator_tag, void, void, void, void>
+{
+public:
+    typedef _CharT char_type;
+    typedef _Traits traits_type;
+    typedef basic_ostream<_CharT,_Traits> ostream_type;
+private:
+    ostream_type* __out_stream_;
+    const char_type* __delim_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY ostream_iterator(ostream_type& __s) _NOEXCEPT
+        : __out_stream_(_CUDA_VSTD::addressof(__s)), __delim_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY ostream_iterator(ostream_type& __s, const _CharT* __delimiter) _NOEXCEPT
+        : __out_stream_(_CUDA_VSTD::addressof(__s)), __delim_(__delimiter) {}
+    _LIBCUDACXX_INLINE_VISIBILITY ostream_iterator& operator=(const _Tp& __value_)
+        {
+            *__out_stream_ << __value_;
+            if (__delim_)
+                *__out_stream_ << __delim_;
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY ostream_iterator& operator*()     {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY ostream_iterator& operator++()    {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY ostream_iterator& operator++(int) {return *this;}
+};
+
+template<class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS istreambuf_iterator
+    : public iterator<input_iterator_tag, _CharT,
+                      typename _Traits::off_type, _CharT*,
+                      _CharT>
+{
+public:
+    typedef _CharT                          char_type;
+    typedef _Traits                         traits_type;
+    typedef typename _Traits::int_type      int_type;
+    typedef basic_streambuf<_CharT,_Traits> streambuf_type;
+    typedef basic_istream<_CharT,_Traits>   istream_type;
+private:
+    mutable streambuf_type* __sbuf_;
+
+    class __proxy
+    {
+        char_type __keep_;
+        streambuf_type* __sbuf_;
+        _LIBCUDACXX_INLINE_VISIBILITY __proxy(char_type __c, streambuf_type* __s)
+            : __keep_(__c), __sbuf_(__s) {}
+        friend class istreambuf_iterator;
+    public:
+        _LIBCUDACXX_INLINE_VISIBILITY char_type operator*() const {return __keep_;}
+    };
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __test_for_eof() const
+    {
+        if (__sbuf_ && traits_type::eq_int_type(__sbuf_->sgetc(), traits_type::eof()))
+            __sbuf_ = 0;
+        return __sbuf_ == 0;
+    }
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR istreambuf_iterator() _NOEXCEPT : __sbuf_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY istreambuf_iterator(istream_type& __s) _NOEXCEPT
+        : __sbuf_(__s.rdbuf()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY istreambuf_iterator(streambuf_type* __s) _NOEXCEPT
+        : __sbuf_(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY istreambuf_iterator(const __proxy& __p) _NOEXCEPT
+        : __sbuf_(__p.__sbuf_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY char_type  operator*() const
+        {return static_cast<char_type>(__sbuf_->sgetc());}
+    _LIBCUDACXX_INLINE_VISIBILITY istreambuf_iterator& operator++()
+        {
+            __sbuf_->sbumpc();
+            return *this;
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY __proxy              operator++(int)
+        {
+            return __proxy(__sbuf_->sbumpc(), __sbuf_);
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY bool equal(const istreambuf_iterator& __b) const
+        {return __test_for_eof() == __b.__test_for_eof();}
+};
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(const istreambuf_iterator<_CharT,_Traits>& __a,
+                const istreambuf_iterator<_CharT,_Traits>& __b)
+                {return __a.equal(__b);}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(const istreambuf_iterator<_CharT,_Traits>& __a,
+                const istreambuf_iterator<_CharT,_Traits>& __b)
+                {return !__a.equal(__b);}
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS ostreambuf_iterator
+    : public iterator<output_iterator_tag, void, void, void, void>
+{
+public:
+    typedef _CharT                          char_type;
+    typedef _Traits                         traits_type;
+    typedef basic_streambuf<_CharT,_Traits> streambuf_type;
+    typedef basic_ostream<_CharT,_Traits>   ostream_type;
+private:
+    streambuf_type* __sbuf_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY ostreambuf_iterator(ostream_type& __s) _NOEXCEPT
+        : __sbuf_(__s.rdbuf()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY ostreambuf_iterator(streambuf_type* __s) _NOEXCEPT
+        : __sbuf_(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY ostreambuf_iterator& operator=(_CharT __c)
+        {
+            if (__sbuf_ && traits_type::eq_int_type(__sbuf_->sputc(__c), traits_type::eof()))
+                __sbuf_ = 0;
+            return *this;
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY ostreambuf_iterator& operator*()     {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY ostreambuf_iterator& operator++()    {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY ostreambuf_iterator& operator++(int) {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY bool failed() const _NOEXCEPT {return __sbuf_ == 0;}
+
+    template <class _Ch, class _Tr>
+    friend
+    _LIBCUDACXX_HIDDEN _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER
+    ostreambuf_iterator<_Ch, _Tr>
+    __pad_and_output(ostreambuf_iterator<_Ch, _Tr> __s,
+                     const _Ch* __ob, const _Ch* __op, const _Ch* __oe,
+                     ios_base& __iob, _Ch __fl);
+};
+
+template <class _Iter>
+class _LIBCUDACXX_TEMPLATE_VIS move_iterator
+{
+private:
+    _Iter __i;
+public:
+    typedef _Iter                                            iterator_type;
+    typedef typename iterator_traits<iterator_type>::iterator_category iterator_category;
+    typedef typename iterator_traits<iterator_type>::value_type value_type;
+    typedef typename iterator_traits<iterator_type>::difference_type difference_type;
+    typedef iterator_type pointer;
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef typename iterator_traits<iterator_type>::reference __reference;
+    typedef typename conditional<
+            is_reference<__reference>::value,
+            typename remove_reference<__reference>::type&&,
+            __reference
+        >::type reference;
+#else
+    typedef typename iterator_traits<iterator_type>::reference reference;
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator() : __i() {}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    explicit move_iterator(_Iter __x) : __i(__x) {}
+    template <class _Up>
+      _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+      move_iterator(const move_iterator<_Up>& __u) : __i(__u.base()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 _Iter base() const {return __i;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reference operator*() const { return static_cast<reference>(*__i); }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    pointer  operator->() const { return __i;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator& operator++() {++__i; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator  operator++(int) {move_iterator __tmp(*this); ++__i; return __tmp;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator& operator--() {--__i; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator  operator--(int) {move_iterator __tmp(*this); --__i; return __tmp;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator  operator+ (difference_type __n) const {return move_iterator(__i + __n);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator& operator+=(difference_type __n) {__i += __n; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator  operator- (difference_type __n) const {return move_iterator(__i - __n);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    move_iterator& operator-=(difference_type __n) {__i -= __n; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+    reference operator[](difference_type __n) const { return _CUDA_VSTD::move(__i[__n]); }
+};
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator==(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() == __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator<(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() < __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator!=(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() != __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator>(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() > __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator>=(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() >= __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+bool
+operator<=(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() <= __y.base();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto
+operator-(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+-> decltype(__x.base() - __y.base())
+{
+    return __x.base() - __y.base();
+}
+#else
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename move_iterator<_Iter1>::difference_type
+operator-(const move_iterator<_Iter1>& __x, const move_iterator<_Iter2>& __y)
+{
+    return __x.base() - __y.base();
+}
+#endif
+
+template <class _Iter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+move_iterator<_Iter>
+operator+(typename move_iterator<_Iter>::difference_type __n, const move_iterator<_Iter>& __x)
+{
+    return move_iterator<_Iter>(__x.base() + __n);
+}
+
+template <class _Iter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+move_iterator<_Iter>
+make_move_iterator(_Iter __i)
+{
+    return move_iterator<_Iter>(__i);
+}
+
+// __wrap_iter
+
+template <class _Iter> class __wrap_iter;
+
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator==(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator<(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator!=(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator>(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator>=(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator<=(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+auto
+operator-(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+-> decltype(__x.base() - __y.base());
+#else
+template <class _Iter1, class _Iter2>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __wrap_iter<_Iter1>::difference_type
+operator-(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+#endif
+
+template <class _Iter>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+__wrap_iter<_Iter>
+operator+(typename __wrap_iter<_Iter>::difference_type, __wrap_iter<_Iter>) _NOEXCEPT;
+
+template <class _Ip, class _Op> _Op _LIBCUDACXX_INLINE_VISIBILITY copy(_Ip, _Ip, _Op);
+template <class _B1, class _B2> _B2 _LIBCUDACXX_INLINE_VISIBILITY copy_backward(_B1, _B1, _B2);
+template <class _Ip, class _Op> _Op _LIBCUDACXX_INLINE_VISIBILITY move(_Ip, _Ip, _Op);
+template <class _B1, class _B2> _B2 _LIBCUDACXX_INLINE_VISIBILITY move_backward(_B1, _B1, _B2);
+
+#if _LIBCUDACXX_DEBUG_LEVEL < 2
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+typename enable_if
+<
+    is_trivially_copy_assignable<_Tp>::value,
+    _Tp*
+>::type
+__unwrap_iter(__wrap_iter<_Tp*>);
+
+#else
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+typename enable_if
+<
+    is_trivially_copy_assignable<_Tp>::value,
+    __wrap_iter<_Tp*>
+>::type
+__unwrap_iter(__wrap_iter<_Tp*> __i);
+
+#endif
+
+template <class _Iter>
+class __wrap_iter
+{
+public:
+    typedef _Iter                                                      iterator_type;
+    typedef typename iterator_traits<iterator_type>::iterator_category iterator_category;
+    typedef typename iterator_traits<iterator_type>::value_type        value_type;
+    typedef typename iterator_traits<iterator_type>::difference_type   difference_type;
+    typedef typename iterator_traits<iterator_type>::pointer           pointer;
+    typedef typename iterator_traits<iterator_type>::reference         reference;
+private:
+    iterator_type __i;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter() _NOEXCEPT
+#if _LIBCUDACXX_STD_VER > 11
+                : __i{}
+#endif
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__insert_i(this);
+#endif
+    }
+    template <class _Up> _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+        __wrap_iter(const __wrap_iter<_Up>& __u,
+            typename enable_if<is_convertible<_Up, iterator_type>::value>::type* = 0) _NOEXCEPT
+            : __i(__u.base())
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__iterator_copy(this, &__u);
+#endif
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+    __wrap_iter(const __wrap_iter& __x)
+        : __i(__x.base())
+    {
+        __get_db()->__iterator_copy(this, &__x);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+    __wrap_iter& operator=(const __wrap_iter& __x)
+    {
+        if (this != &__x)
+        {
+            __get_db()->__iterator_copy(this, &__x);
+            __i = __x.__i;
+        }
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+    ~__wrap_iter()
+    {
+        __get_db()->__erase_i(this);
+    }
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG reference operator*() const _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to dereference a non-dereferenceable iterator");
+#endif
+        return *__i;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG pointer  operator->() const _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to dereference a non-dereferenceable iterator");
+#endif
+        return (pointer)_CUDA_VSTD::addressof(*__i);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter& operator++() _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to increment non-incrementable iterator");
+#endif
+        ++__i;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter  operator++(int) _NOEXCEPT
+        {__wrap_iter __tmp(*this); ++(*this); return __tmp;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter& operator--() _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__decrementable(this),
+                       "Attempted to decrement non-decrementable iterator");
+#endif
+        --__i;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter  operator--(int) _NOEXCEPT
+        {__wrap_iter __tmp(*this); --(*this); return __tmp;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter  operator+ (difference_type __n) const _NOEXCEPT
+        {__wrap_iter __w(*this); __w += __n; return __w;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter& operator+=(difference_type __n) _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__addable(this, __n),
+                   "Attempted to add/subtract iterator outside of valid range");
+#endif
+        __i += __n;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter  operator- (difference_type __n) const _NOEXCEPT
+        {return *this + (-__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter& operator-=(difference_type __n) _NOEXCEPT
+        {*this += -__n; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG reference    operator[](difference_type __n) const _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__subscriptable(this, __n),
+                   "Attempted to subscript iterator outside of valid range");
+#endif
+        return __i[__n];
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG iterator_type base() const _NOEXCEPT {return __i;}
+
+private:
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter(const void* __p, iterator_type __x) : __i(__x)
+    {
+        __get_db()->__insert_ic(this, __p);
+    }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG __wrap_iter(iterator_type __x) _NOEXCEPT : __i(__x) {}
+#endif
+
+    template <class _Up> friend class __wrap_iter;
+    template <class _CharT, class _Traits, class _Alloc> friend class basic_string;
+    template <class _Tp, class _Alloc> friend class vector;
+    template <class _Tp, size_t> friend class span;
+
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    bool
+    operator==(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    bool
+    operator<(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    bool
+    operator!=(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    bool
+    operator>(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    bool
+    operator>=(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    bool
+    operator<=(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    auto
+    operator-(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+    -> decltype(__x.base() - __y.base());
+#else
+    template <class _Iter1, class _Iter2>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    typename __wrap_iter<_Iter1>::difference_type
+    operator-(const __wrap_iter<_Iter1>&, const __wrap_iter<_Iter2>&) _NOEXCEPT;
+#endif
+
+    template <class _Iter1>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    __wrap_iter<_Iter1>
+    operator+(typename __wrap_iter<_Iter1>::difference_type, __wrap_iter<_Iter1>) _NOEXCEPT;
+
+    template <class _Ip, class _Op> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER friend _Op copy(_Ip, _Ip, _Op);
+    template <class _B1, class _B2> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER friend _B2 copy_backward(_B1, _B1, _B2);
+    template <class _Ip, class _Op> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER friend _Op move(_Ip, _Ip, _Op);
+    template <class _B1, class _B2> _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER friend _B2 move_backward(_B1, _B1, _B2);
+
+#if _LIBCUDACXX_DEBUG_LEVEL < 2
+    template <class _Tp>
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER _LIBCUDACXX_CONSTEXPR_IF_NODEBUG friend
+    typename enable_if
+    <
+        is_trivially_copy_assignable<_Tp>::value,
+        _Tp*
+    >::type
+    __unwrap_iter(__wrap_iter<_Tp*>);
+#else
+  template <class _Tp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+  typename enable_if
+  <
+      is_trivially_copy_assignable<_Tp>::value,
+      __wrap_iter<_Tp*>
+  >::type
+  __unwrap_iter(__wrap_iter<_Tp*> __i);
+#endif
+};
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator==(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+    return __x.base() == __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator<(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__less_than_comparable(&__x, &__y),
+                   "Attempted to compare incomparable iterators");
+#endif
+    return __x.base() < __y.base();
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator!=(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+    return !(__x == __y);
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator>(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+    return __y < __x;
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator>=(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+    return !(__x < __y);
+}
+
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator<=(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+    return !(__y < __x);
+}
+
+template <class _Iter1>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator!=(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter1>& __y) _NOEXCEPT
+{
+    return !(__x == __y);
+}
+
+template <class _Iter1>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator>(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter1>& __y) _NOEXCEPT
+{
+    return __y < __x;
+}
+
+template <class _Iter1>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator>=(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter1>& __y) _NOEXCEPT
+{
+    return !(__x < __y);
+}
+
+template <class _Iter1>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+bool
+operator<=(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter1>& __y) _NOEXCEPT
+{
+    return !(__y < __x);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+auto
+operator-(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+-> decltype(__x.base() - __y.base())
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__less_than_comparable(&__x, &__y),
+                   "Attempted to subtract incompatible iterators");
+#endif
+    return __x.base() - __y.base();
+}
+#else
+template <class _Iter1, class _Iter2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+typename __wrap_iter<_Iter1>::difference_type
+operator-(const __wrap_iter<_Iter1>& __x, const __wrap_iter<_Iter2>& __y) _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__less_than_comparable(&__x, &__y),
+                   "Attempted to subtract incompatible iterators");
+#endif
+    return __x.base() - __y.base();
+}
+#endif
+
+template <class _Iter>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_IF_NODEBUG
+__wrap_iter<_Iter>
+operator+(typename __wrap_iter<_Iter>::difference_type __n,
+          __wrap_iter<_Iter> __x) _NOEXCEPT
+{
+    __x += __n;
+    return __x;
+}
+
+template <class _Iter>
+struct __libcpp_is_trivial_iterator
+    : public _LIBCUDACXX_BOOL_CONSTANT(is_pointer<_Iter>::value) {};
+
+template <class _Iter>
+struct __libcpp_is_trivial_iterator<move_iterator<_Iter> >
+    : public _LIBCUDACXX_BOOL_CONSTANT(__libcpp_is_trivial_iterator<_Iter>::value) {};
+
+template <class _Iter>
+struct __libcpp_is_trivial_iterator<reverse_iterator<_Iter> >
+    : public _LIBCUDACXX_BOOL_CONSTANT(__libcpp_is_trivial_iterator<_Iter>::value) {};
+
+template <class _Iter>
+struct __libcpp_is_trivial_iterator<__wrap_iter<_Iter> >
+    : public _LIBCUDACXX_BOOL_CONSTANT(__libcpp_is_trivial_iterator<_Iter>::value) {};
+
+
+template <class _Tp, size_t _Np>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp*
+begin(_Tp (&__array)[_Np])
+{
+    return __array;
+}
+
+template <class _Tp, size_t _Np>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+_Tp*
+end(_Tp (&__array)[_Np])
+{
+    return __array + _Np;
+}
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto
+begin(_Cp& __c) -> decltype(__c.begin())
+{
+    return __c.begin();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto
+begin(const _Cp& __c) -> decltype(__c.begin())
+{
+    return __c.begin();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto
+end(_Cp& __c) -> decltype(__c.end())
+{
+    return __c.end();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto
+end(const _Cp& __c) -> decltype(__c.end())
+{
+    return __c.end();
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template <class _Tp, size_t _Np>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+reverse_iterator<_Tp*> rbegin(_Tp (&__array)[_Np])
+{
+    return reverse_iterator<_Tp*>(__array + _Np);
+}
+
+template <class _Tp, size_t _Np>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+reverse_iterator<_Tp*> rend(_Tp (&__array)[_Np])
+{
+    return reverse_iterator<_Tp*>(__array);
+}
+
+template <class _Ep>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+reverse_iterator<const _Ep*> rbegin(::std::initializer_list<_Ep> __il)
+{
+    return reverse_iterator<const _Ep*>(__il.end());
+}
+
+template <class _Ep>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+reverse_iterator<const _Ep*> rend(::std::initializer_list<_Ep> __il)
+{
+    return reverse_iterator<const _Ep*>(__il.begin());
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+auto cbegin(const _Cp& __c) -> decltype(_CUDA_VSTD::begin(__c))
+{
+    return _CUDA_VSTD::begin(__c);
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+auto cend(const _Cp& __c) -> decltype(_CUDA_VSTD::end(__c))
+{
+    return _CUDA_VSTD::end(__c);
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto rbegin(_Cp& __c) -> decltype(__c.rbegin())
+{
+    return __c.rbegin();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto rbegin(const _Cp& __c) -> decltype(__c.rbegin())
+{
+    return __c.rbegin();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto rend(_Cp& __c) -> decltype(__c.rend())
+{
+    return __c.rend();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto rend(const _Cp& __c) -> decltype(__c.rend())
+{
+    return __c.rend();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto crbegin(const _Cp& __c) -> decltype(_CUDA_VSTD::rbegin(__c))
+{
+    return _CUDA_VSTD::rbegin(__c);
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+auto crend(const _Cp& __c) -> decltype(_CUDA_VSTD::rend(__c))
+{
+    return _CUDA_VSTD::rend(__c);
+}
+
+#endif
+
+
+#else  // defined(_LIBCUDACXX_CXX03_LANG)
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename _Cp::iterator
+begin(_Cp& __c)
+{
+    return __c.begin();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename _Cp::const_iterator
+begin(const _Cp& __c)
+{
+    return __c.begin();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename _Cp::iterator
+end(_Cp& __c)
+{
+    return __c.end();
+}
+
+template <class _Cp>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename _Cp::const_iterator
+end(const _Cp& __c)
+{
+    return __c.end();
+}
+
+#endif  // !defined(_LIBCUDACXX_CXX03_LANG)
+
+#if _LIBCUDACXX_STD_VER > 14
+
+// #if _LIBCUDACXX_STD_VER > 11
+// template <>
+// struct _LIBCUDACXX_TEMPLATE_VIS plus<void>
+// {
+//     template <class _T1, class _T2>
+//     _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+//     auto operator()(_T1&& __t, _T2&& __u) const
+//     _NOEXCEPT_(noexcept(_CUDA_VSTD::forward<_T1>(__t) + _CUDA_VSTD::forward<_T2>(__u)))
+//     -> decltype        (_CUDA_VSTD::forward<_T1>(__t) + _CUDA_VSTD::forward<_T2>(__u))
+//         { return        _CUDA_VSTD::forward<_T1>(__t) + _CUDA_VSTD::forward<_T2>(__u); }
+//     typedef void is_transparent;
+// };
+// #endif
+
+template <class _Cont>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr auto size(const _Cont& __c)
+_NOEXCEPT_(noexcept(__c.size()))
+-> decltype        (__c.size())
+{ return            __c.size(); }
+
+template <class _Tp, size_t _Sz>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr size_t size(const _Tp (&)[_Sz]) noexcept { return _Sz; }
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Cont>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr auto ssize(const _Cont& __c)
+_NOEXCEPT_(noexcept(static_cast<common_type_t<ptrdiff_t, make_signed_t<decltype(__c.size())>>>(__c.size())))
+->                              common_type_t<ptrdiff_t, make_signed_t<decltype(__c.size())>>
+{ return            static_cast<common_type_t<ptrdiff_t, make_signed_t<decltype(__c.size())>>>(__c.size()); }
+
+template <class _Tp, ptrdiff_t _Sz>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr ptrdiff_t ssize(const _Tp (&)[_Sz]) noexcept { return _Sz; }
+#endif
+
+template <class _Cont>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+constexpr auto empty(const _Cont& __c)
+_NOEXCEPT_(noexcept(__c.empty()))
+-> decltype        (__c.empty())
+{ return            __c.empty(); }
+
+template <class _Tp, size_t _Sz>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool empty(const _Tp (&)[_Sz]) noexcept { return false; }
+
+template <class _Ep>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool empty(::std::initializer_list<_Ep> __il) noexcept { return __il.size() == 0; }
+
+template <class _Cont> constexpr
+_LIBCUDACXX_INLINE_VISIBILITY
+auto data(_Cont& __c)
+_NOEXCEPT_(noexcept(__c.data()))
+-> decltype        (__c.data())
+{ return            __c.data(); }
+
+template <class _Cont> constexpr
+_LIBCUDACXX_INLINE_VISIBILITY
+auto data(const _Cont& __c)
+_NOEXCEPT_(noexcept(__c.data()))
+-> decltype        (__c.data())
+{ return            __c.data(); }
+
+template <class _Tp, size_t _Sz>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr _Tp* data(_Tp (&__array)[_Sz]) noexcept { return __array; }
+
+template <class _Ep>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr const _Ep* data(::std::initializer_list<_Ep> __il) noexcept { return __il.begin(); }
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_ITERATOR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/latch b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/latch
new file mode 100644
index 000000000000..c2f92fbee977
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/latch
@@ -0,0 +1,129 @@
+// -*- C++ -*-
+//===--------------------------- latch -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_LATCH
+#define _LIBCUDACXX_LATCH
+
+/*
+    latch synopsis
+
+namespace std
+{
+
+  class latch
+  {
+  public:
+    constexpr explicit latch(ptrdiff_t __expected);
+    ~latch();
+
+    latch(const latch&) = delete;
+    latch& operator=(const latch&) = delete;
+
+    void count_down(ptrdiff_t __update = 1);
+    bool try_wait() const noexcept;
+    void wait() const;
+    void arrive_and_wait(ptrdiff_t __update = 1);
+
+  private:
+    ptrdiff_t __counter; // exposition only
+  };
+
+}
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__threading_support>
+#include <atomic>
+#include <cassert>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+# error <latch> is not supported on this single threaded system
+#endif
+
+#if _LIBCUDACXX_STD_VER < 11
+# error <latch> is requires C++11 or later
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+# if _LIBCUDACXX_CUDA_ABI_VERSION < 3
+#  define _LIBCUDACXX_LATCH_ALIGNMENT alignas(64)
+# else
+#  define _LIBCUDACXX_LATCH_ALIGNMENT
+# endif
+
+template<int _Sco = 0>
+class __latch_base
+{
+    _LIBCUDACXX_LATCH_ALIGNMENT __atomic_base<ptrdiff_t, _Sco> __counter;
+public:
+    inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    explicit __latch_base(ptrdiff_t __expected)
+        : __counter(__expected) { }
+
+    ~__latch_base() = default;
+    __latch_base(const __latch_base&) = delete;
+    __latch_base& operator=(const __latch_base&) = delete;
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void count_down(ptrdiff_t __update = 1)
+    {
+        assert(__update > 0);
+        auto const __old = __counter.fetch_sub(__update, memory_order_release);
+        assert(__old >= __update);
+        if(__old == __update)
+            __counter.notify_all();
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_wait() const noexcept
+    {
+        return __counter.load(memory_order_acquire) == 0;
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void wait() const
+    {
+        while(1) {
+            auto const __current = __counter.load(memory_order_acquire);
+            if(__current == 0)
+                return;
+            __counter.wait(__current, memory_order_relaxed)
+            ;
+        }
+    }
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void arrive_and_wait(ptrdiff_t __update = 1)
+    {
+        count_down(__update);
+        wait();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr ptrdiff_t max() noexcept
+    {
+        return numeric_limits<ptrdiff_t>::max();
+    }
+};
+
+using latch = __latch_base<>;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif //_LIBCUDACXX_LATCH
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits
new file mode 100644
index 000000000000..b89459074540
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits
@@ -0,0 +1,837 @@
+// -*- C++ -*-
+//===---------------------------- limits ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_LIMITS
+#define _LIBCUDACXX_LIMITS
+
+/*
+    limits synopsis
+
+namespace std
+{
+
+template<class T>
+class numeric_limits
+{
+public:
+    static constexpr bool is_specialized = false;
+    static constexpr T min() noexcept;
+    static constexpr T max() noexcept;
+    static constexpr T lowest() noexcept;
+
+    static constexpr int  digits = 0;
+    static constexpr int  digits10 = 0;
+    static constexpr int  max_digits10 = 0;
+    static constexpr bool is_signed = false;
+    static constexpr bool is_integer = false;
+    static constexpr bool is_exact = false;
+    static constexpr int  radix = 0;
+    static constexpr T epsilon() noexcept;
+    static constexpr T round_error() noexcept;
+
+    static constexpr int  min_exponent = 0;
+    static constexpr int  min_exponent10 = 0;
+    static constexpr int  max_exponent = 0;
+    static constexpr int  max_exponent10 = 0;
+
+    static constexpr bool has_infinity = false;
+    static constexpr bool has_quiet_NaN = false;
+    static constexpr bool has_signaling_NaN = false;
+    static constexpr float_denorm_style has_denorm = denorm_absent;
+    static constexpr bool has_denorm_loss = false;
+    static constexpr T infinity() noexcept;
+    static constexpr T quiet_NaN() noexcept;
+    static constexpr T signaling_NaN() noexcept;
+    static constexpr T denorm_min() noexcept;
+
+    static constexpr bool is_iec559 = false;
+    static constexpr bool is_bounded = false;
+    static constexpr bool is_modulo = false;
+
+    static constexpr bool traps = false;
+    static constexpr bool tinyness_before = false;
+    static constexpr float_round_style round_style = round_toward_zero;
+};
+
+enum float_round_style
+{
+    round_indeterminate       = -1,
+    round_toward_zero         =  0,
+    round_to_nearest          =  1,
+    round_toward_infinity     =  2,
+    round_toward_neg_infinity =  3
+};
+
+enum float_denorm_style
+{
+    denorm_indeterminate = -1,
+    denorm_absent = 0,
+    denorm_present = 1
+};
+
+template<> class numeric_limits<cv bool>;
+
+template<> class numeric_limits<cv char>;
+template<> class numeric_limits<cv signed char>;
+template<> class numeric_limits<cv unsigned char>;
+template<> class numeric_limits<cv wchar_t>;
+template<> class numeric_limits<cv char8_t>; // C++20
+template<> class numeric_limits<cv char16_t>;
+template<> class numeric_limits<cv char32_t>;
+
+template<> class numeric_limits<cv short>;
+template<> class numeric_limits<cv int>;
+template<> class numeric_limits<cv long>;
+template<> class numeric_limits<cv long long>;
+template<> class numeric_limits<cv unsigned short>;
+template<> class numeric_limits<cv unsigned int>;
+template<> class numeric_limits<cv unsigned long>;
+template<> class numeric_limits<cv unsigned long long>;
+
+template<> class numeric_limits<cv float>;
+template<> class numeric_limits<cv double>;
+template<> class numeric_limits<cv long double>;
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <type_traits>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
+#include "support/win32/limits_msvc_win32.h"
+#endif // _LIBCUDACXX_MSVCRT
+
+#if defined(__IBMCPP__)
+#include "support/ibm/limits.h"
+#endif // __IBMCPP__
+
+#ifndef __cuda_std__
+#include <version>
+#endif //__cuda_std__
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+enum float_round_style
+{
+    round_indeterminate       = -1,
+    round_toward_zero         =  0,
+    round_to_nearest          =  1,
+    round_toward_infinity     =  2,
+    round_toward_neg_infinity =  3
+};
+
+enum float_denorm_style
+{
+    denorm_indeterminate = -1,
+    denorm_absent = 0,
+    denorm_present = 1
+};
+
+template <class _Tp, bool = is_arithmetic<_Tp>::value>
+class __libcpp_numeric_limits
+{
+protected:
+    typedef _Tp type;
+
+    static _LIBCUDACXX_CONSTEXPR const  bool is_specialized = false;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return type();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return type();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return type();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  digits = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = false;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = 0;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return type();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return type();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = 0;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = false;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = false;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = false;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = denorm_absent;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = false;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return type();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return type();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return type();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return type();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = false;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = false;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = false;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = round_toward_zero;
+};
+
+template <class _Tp, int __digits, bool _IsSigned>
+struct __libcpp_compute_min
+{
+    static _LIBCUDACXX_CONSTEXPR const _Tp value = static_cast<_Tp>(_Tp(1) << __digits);
+};
+
+template <class _Tp, int __digits>
+struct __libcpp_compute_min<_Tp, __digits, false>
+{
+    static _LIBCUDACXX_CONSTEXPR const _Tp value = _Tp(0);
+};
+
+template <class _Tp>
+class __libcpp_numeric_limits<_Tp, true>
+{
+protected:
+    typedef _Tp type;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = true;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = type(-1) < type(0);
+    static _LIBCUDACXX_CONSTEXPR const int  digits = static_cast<int>(sizeof(type) * __CHAR_BIT__ - is_signed);
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = digits * 3 / 10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const type __min = __libcpp_compute_min<type, digits, is_signed>::value;
+    static _LIBCUDACXX_CONSTEXPR const type __max = is_signed ? type(type(~0) ^ __min) : type(~0);
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __min;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __max;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return min();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = true;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = 2;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return type(0);}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = 0;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = false;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = false;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = false;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = denorm_absent;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = false;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return type(0);}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = !_CUDA_VSTD::is_signed<_Tp>::value;
+
+#if defined(__i386__) || defined(__x86_64__) || defined(__pnacl__) || \
+    defined(__wasm__)
+    static _LIBCUDACXX_CONSTEXPR const bool traps = true;
+#else
+    static _LIBCUDACXX_CONSTEXPR const bool traps = false;
+#endif
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = false;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = round_toward_zero;
+};
+
+template <>
+class __libcpp_numeric_limits<bool, true>
+{
+protected:
+    typedef bool type;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = true;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = false;
+    static _LIBCUDACXX_CONSTEXPR const int  digits = 1;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const type __min = false;
+    static _LIBCUDACXX_CONSTEXPR const type __max = true;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __min;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __max;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return min();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = true;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = 2;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return type(0);}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = 0;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = 0;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = false;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = false;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = false;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = denorm_absent;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = false;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return type(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return type(0);}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = false;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = false;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = false;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = round_toward_zero;
+};
+
+template <>
+class __libcpp_numeric_limits<float, true>
+{
+protected:
+    typedef float type;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = true;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = true;
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __FLT_MANT_DIG__;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __FLT_DIG__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = 2+(digits * 30103l)/100000l;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __FLT_MIN__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __FLT_MAX__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return -max();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = false;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __FLT_RADIX__;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __FLT_EPSILON__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return 0.5F;}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __FLT_MIN_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __FLT_MIN_10_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __FLT_MAX_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __FLT_MAX_10_EXP__;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = true;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = true;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = true;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = denorm_present;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = false;
+#ifdef __CUDACC_RTC__
+    _LIBCUDACXX_INLINE_VISIBILITY static type infinity() _NOEXCEPT {return __builtin_huge_valf();}
+    _LIBCUDACXX_INLINE_VISIBILITY static type quiet_NaN() _NOEXCEPT {return __builtin_nanf("");}
+    _LIBCUDACXX_INLINE_VISIBILITY static type signaling_NaN() _NOEXCEPT {return __builtin_nansf("");}
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __builtin_huge_valf();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __builtin_nanf("");}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __builtin_nansf("");}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __FLT_DENORM_MIN__;}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = false;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = false;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = false;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = round_to_nearest;
+};
+
+template <>
+class __libcpp_numeric_limits<double, true>
+{
+protected:
+    typedef double type;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = true;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = true;
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __DBL_MANT_DIG__;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __DBL_DIG__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = 2+(digits * 30103l)/100000l;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __DBL_MIN__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __DBL_MAX__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return -max();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = false;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __FLT_RADIX__;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __DBL_EPSILON__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return 0.5;}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __DBL_MIN_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __DBL_MIN_10_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __DBL_MAX_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __DBL_MAX_10_EXP__;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = true;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = true;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = true;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = denorm_present;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = false;
+#ifdef __CUDACC_RTC__
+    _LIBCUDACXX_INLINE_VISIBILITY static type infinity() _NOEXCEPT {return __builtin_huge_val();}
+    _LIBCUDACXX_INLINE_VISIBILITY static type quiet_NaN() _NOEXCEPT {return __builtin_nan("");}
+    _LIBCUDACXX_INLINE_VISIBILITY static type signaling_NaN() _NOEXCEPT {return __builtin_nans("");}
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __builtin_huge_val();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __builtin_nan("");}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __builtin_nans("");}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __DBL_DENORM_MIN__;}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = false;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = false;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = false;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = round_to_nearest;
+};
+
+template <>
+class __libcpp_numeric_limits<long double, true>
+{
+#ifndef _LIBCUDACXX_HAS_NO_LONG_DOUBLE 
+protected:
+    typedef long double type;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = true;
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = true;
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __LDBL_MANT_DIG__;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __LDBL_DIG__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = 2+(digits * 30103l)/100000l;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __LDBL_MIN__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __LDBL_MAX__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return -max();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = false;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = false;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __FLT_RADIX__;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __LDBL_EPSILON__;}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return 0.5L;}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __LDBL_MIN_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __LDBL_MIN_10_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __LDBL_MAX_EXP__;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __LDBL_MAX_10_EXP__;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = true;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = true;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = true;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = denorm_present;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = false;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __builtin_huge_vall();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __builtin_nanl("");}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __builtin_nansl("");}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __LDBL_DENORM_MIN__;}
+
+#if (defined(__ppc__) || defined(__ppc64__))
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = false;
+#else
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = true;
+#endif
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = true;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = false;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = false;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = false;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = round_to_nearest;
+#endif
+};
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS numeric_limits
+    : private __libcpp_numeric_limits<typename remove_cv<_Tp>::type>
+{
+    typedef __libcpp_numeric_limits<typename remove_cv<_Tp>::type> __base;
+    typedef typename __base::type type;
+public:
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = __base::is_specialized;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __base::min();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __base::max();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return __base::lowest();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __base::digits;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __base::digits10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = __base::max_digits10;
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = __base::is_signed;
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = __base::is_integer;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = __base::is_exact;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __base::radix;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __base::epsilon();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return __base::round_error();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __base::min_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __base::min_exponent10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __base::max_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __base::max_exponent10;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = __base::has_infinity;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = __base::has_quiet_NaN;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = __base::has_signaling_NaN;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = __base::has_denorm;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = __base::has_denorm_loss;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __base::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __base::quiet_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __base::signaling_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __base::denorm_min();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = __base::is_iec559;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = __base::is_bounded;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = __base::is_modulo;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = __base::traps;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = __base::tinyness_before;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = __base::round_style;
+};
+
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_specialized;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::digits;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::max_digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_signed;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_integer;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_exact;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::radix;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::min_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::min_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::max_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<_Tp>::max_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::has_infinity;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::has_quiet_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::has_signaling_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_denorm_style numeric_limits<_Tp>::has_denorm;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::has_denorm_loss;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_iec559;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_bounded;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::is_modulo;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::traps;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<_Tp>::tinyness_before;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_round_style numeric_limits<_Tp>::round_style;
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS numeric_limits<const _Tp>
+    : private numeric_limits<_Tp>
+{
+    typedef numeric_limits<_Tp> __base;
+    typedef _Tp type;
+public:
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = __base::is_specialized;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __base::min();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __base::max();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return __base::lowest();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __base::digits;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __base::digits10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = __base::max_digits10;
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = __base::is_signed;
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = __base::is_integer;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = __base::is_exact;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __base::radix;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __base::epsilon();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return __base::round_error();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __base::min_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __base::min_exponent10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __base::max_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __base::max_exponent10;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = __base::has_infinity;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = __base::has_quiet_NaN;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = __base::has_signaling_NaN;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = __base::has_denorm;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = __base::has_denorm_loss;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __base::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __base::quiet_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __base::signaling_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __base::denorm_min();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = __base::is_iec559;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = __base::is_bounded;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = __base::is_modulo;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = __base::traps;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = __base::tinyness_before;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = __base::round_style;
+};
+
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_specialized;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::digits;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::max_digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_signed;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_integer;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_exact;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::radix;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::min_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::min_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::max_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const _Tp>::max_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::has_infinity;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::has_quiet_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::has_signaling_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_denorm_style numeric_limits<const _Tp>::has_denorm;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::has_denorm_loss;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_iec559;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_bounded;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::is_modulo;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::traps;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const _Tp>::tinyness_before;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_round_style numeric_limits<const _Tp>::round_style;
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS numeric_limits<volatile _Tp>
+    : private numeric_limits<_Tp>
+{
+    typedef numeric_limits<_Tp> __base;
+    typedef _Tp type;
+public:
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = __base::is_specialized;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __base::min();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __base::max();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return __base::lowest();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __base::digits;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __base::digits10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = __base::max_digits10;
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = __base::is_signed;
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = __base::is_integer;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = __base::is_exact;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __base::radix;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __base::epsilon();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return __base::round_error();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __base::min_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __base::min_exponent10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __base::max_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __base::max_exponent10;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = __base::has_infinity;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = __base::has_quiet_NaN;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = __base::has_signaling_NaN;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = __base::has_denorm;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = __base::has_denorm_loss;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __base::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __base::quiet_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __base::signaling_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __base::denorm_min();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = __base::is_iec559;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = __base::is_bounded;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = __base::is_modulo;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = __base::traps;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = __base::tinyness_before;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = __base::round_style;
+};
+
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_specialized;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::digits;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::max_digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_signed;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_integer;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_exact;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::radix;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::min_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::min_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::max_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<volatile _Tp>::max_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::has_infinity;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::has_quiet_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::has_signaling_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_denorm_style numeric_limits<volatile _Tp>::has_denorm;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::has_denorm_loss;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_iec559;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_bounded;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::is_modulo;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::traps;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<volatile _Tp>::tinyness_before;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_round_style numeric_limits<volatile _Tp>::round_style;
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS numeric_limits<const volatile _Tp>
+    : private numeric_limits<_Tp>
+{
+    typedef numeric_limits<_Tp> __base;
+    typedef _Tp type;
+public:
+    static _LIBCUDACXX_CONSTEXPR const bool is_specialized = __base::is_specialized;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type min() _NOEXCEPT {return __base::min();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type max() _NOEXCEPT {return __base::max();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type lowest() _NOEXCEPT {return __base::lowest();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  digits = __base::digits;
+    static _LIBCUDACXX_CONSTEXPR const int  digits10 = __base::digits10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_digits10 = __base::max_digits10;
+    static _LIBCUDACXX_CONSTEXPR const bool is_signed = __base::is_signed;
+    static _LIBCUDACXX_CONSTEXPR const bool is_integer = __base::is_integer;
+    static _LIBCUDACXX_CONSTEXPR const bool is_exact = __base::is_exact;
+    static _LIBCUDACXX_CONSTEXPR const int  radix = __base::radix;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type epsilon() _NOEXCEPT {return __base::epsilon();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type round_error() _NOEXCEPT {return __base::round_error();}
+
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent = __base::min_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  min_exponent10 = __base::min_exponent10;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent = __base::max_exponent;
+    static _LIBCUDACXX_CONSTEXPR const int  max_exponent10 = __base::max_exponent10;
+
+    static _LIBCUDACXX_CONSTEXPR const bool has_infinity = __base::has_infinity;
+    static _LIBCUDACXX_CONSTEXPR const bool has_quiet_NaN = __base::has_quiet_NaN;
+    static _LIBCUDACXX_CONSTEXPR const bool has_signaling_NaN = __base::has_signaling_NaN;
+    static _LIBCUDACXX_CONSTEXPR const float_denorm_style has_denorm = __base::has_denorm;
+    static _LIBCUDACXX_CONSTEXPR const bool has_denorm_loss = __base::has_denorm_loss;
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type infinity() _NOEXCEPT {return __base::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type quiet_NaN() _NOEXCEPT {return __base::quiet_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type signaling_NaN() _NOEXCEPT {return __base::signaling_NaN();}
+    _LIBCUDACXX_INLINE_VISIBILITY static _LIBCUDACXX_CONSTEXPR type denorm_min() _NOEXCEPT {return __base::denorm_min();}
+
+    static _LIBCUDACXX_CONSTEXPR const bool is_iec559 = __base::is_iec559;
+    static _LIBCUDACXX_CONSTEXPR const bool is_bounded = __base::is_bounded;
+    static _LIBCUDACXX_CONSTEXPR const bool is_modulo = __base::is_modulo;
+
+    static _LIBCUDACXX_CONSTEXPR const bool traps = __base::traps;
+    static _LIBCUDACXX_CONSTEXPR const bool tinyness_before = __base::tinyness_before;
+    static _LIBCUDACXX_CONSTEXPR const float_round_style round_style = __base::round_style;
+};
+
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_specialized;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::digits;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::max_digits10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_signed;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_integer;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_exact;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::radix;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::min_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::min_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::max_exponent;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const int numeric_limits<const volatile _Tp>::max_exponent10;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::has_infinity;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::has_quiet_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::has_signaling_NaN;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_denorm_style numeric_limits<const volatile _Tp>::has_denorm;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::has_denorm_loss;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_iec559;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_bounded;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::is_modulo;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::traps;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const bool numeric_limits<const volatile _Tp>::tinyness_before;
+template <class _Tp>
+    _LIBCUDACXX_CONSTEXPR const float_round_style numeric_limits<const volatile _Tp>::round_style;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_LIMITS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h
new file mode 100644
index 000000000000..6cb5334dee3c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/limits.h
@@ -0,0 +1,68 @@
+// -*- C++ -*-
+//===--------------------------- limits.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_LIMITS_H
+#define _LIBCUDACXX_LIMITS_H
+
+/*
+    limits.h synopsis
+
+Macros:
+
+    CHAR_BIT
+    SCHAR_MIN
+    SCHAR_MAX
+    UCHAR_MAX
+    CHAR_MIN
+    CHAR_MAX
+    MB_LEN_MAX
+    SHRT_MIN
+    SHRT_MAX
+    USHRT_MAX
+    INT_MIN
+    INT_MAX
+    UINT_MAX
+    LONG_MIN
+    LONG_MAX
+    ULONG_MAX
+    LLONG_MIN   // C99
+    LLONG_MAX   // C99
+    ULLONG_MAX  // C99
+
+*/
+
+#include <__config>
+
+#include <__pragma_push>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifndef __GNUC__
+#include_next <limits.h>
+#else
+// GCC header limits.h recursively includes itself through another header called
+// syslimits.h for some reason. This setup breaks down if we directly
+// #include_next GCC's limits.h (reasons not entirely clear to me). Therefore,
+// we manually re-create the necessary include sequence below:
+
+// Get the system limits.h defines (force recurse into the next level)
+#define _GCC_LIMITS_H_
+#define _GCC_NEXT_LIMITS_H
+#include_next <limits.h>
+
+// Get the ISO C defines
+#undef _GCC_LIMITS_H_
+#include_next <limits.h>
+#endif // __GNUC__
+
+#include <__pragma_pop>
+
+#endif  // _LIBCUDACXX_LIMITS_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/list b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/list
new file mode 100644
index 000000000000..c8e827a4e03e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/list
@@ -0,0 +1,2488 @@
+// -*- C++ -*-
+//===---------------------------- list ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_LIST
+#define _LIBCUDACXX_LIST
+
+/*
+    list synopsis
+
+namespace std
+{
+
+template <class T, class Alloc = allocator<T> >
+class list
+{
+public:
+
+    // types:
+    typedef T value_type;
+    typedef Alloc allocator_type;
+    typedef typename allocator_type::reference reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef typename allocator_type::pointer pointer;
+    typedef typename allocator_type::const_pointer const_pointer;
+    typedef implementation-defined iterator;
+    typedef implementation-defined const_iterator;
+    typedef implementation-defined size_type;
+    typedef implementation-defined difference_type;
+    typedef reverse_iterator<iterator> reverse_iterator;
+    typedef reverse_iterator<const_iterator> const_reverse_iterator;
+
+    list()
+        noexcept(is_nothrow_default_constructible<allocator_type>::value);
+    explicit list(const allocator_type& a);
+    explicit list(size_type n);
+    explicit list(size_type n, const allocator_type& a); // C++14
+    list(size_type n, const value_type& value);
+    list(size_type n, const value_type& value, const allocator_type& a);
+    template <class Iter>
+        list(Iter first, Iter last);
+    template <class Iter>
+        list(Iter first, Iter last, const allocator_type& a);
+    list(const list& x);
+    list(const list&, const allocator_type& a);
+    list(list&& x)
+        noexcept(is_nothrow_move_constructible<allocator_type>::value);
+    list(list&&, const allocator_type& a);
+    list(initializer_list<value_type>);
+    list(initializer_list<value_type>, const allocator_type& a);
+
+    ~list();
+
+    list& operator=(const list& x);
+    list& operator=(list&& x)
+        noexcept(
+             allocator_type::propagate_on_container_move_assignment::value &&
+             is_nothrow_move_assignable<allocator_type>::value);
+    list& operator=(initializer_list<value_type>);
+    template <class Iter>
+        void assign(Iter first, Iter last);
+    void assign(size_type n, const value_type& t);
+    void assign(initializer_list<value_type>);
+
+    allocator_type get_allocator() const noexcept;
+
+    iterator begin() noexcept;
+    const_iterator begin() const noexcept;
+    iterator end() noexcept;
+    const_iterator end() const noexcept;
+    reverse_iterator rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+    reverse_iterator rend() noexcept;
+    const_reverse_iterator rend() const noexcept;
+    const_iterator cbegin() const noexcept;
+    const_iterator cend() const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend() const noexcept;
+
+    reference front();
+    const_reference front() const;
+    reference back();
+    const_reference back() const;
+
+    bool empty() const noexcept;
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+
+    template <class... Args>
+        reference emplace_front(Args&&... args); // reference in C++17
+    void pop_front();
+    template <class... Args>
+        reference emplace_back(Args&&... args);  // reference in C++17
+    void pop_back();
+    void push_front(const value_type& x);
+    void push_front(value_type&& x);
+    void push_back(const value_type& x);
+    void push_back(value_type&& x);
+    template <class... Args>
+        iterator emplace(const_iterator position, Args&&... args);
+    iterator insert(const_iterator position, const value_type& x);
+    iterator insert(const_iterator position, value_type&& x);
+    iterator insert(const_iterator position, size_type n, const value_type& x);
+    template <class Iter>
+        iterator insert(const_iterator position, Iter first, Iter last);
+    iterator insert(const_iterator position, initializer_list<value_type> il);
+
+    iterator erase(const_iterator position);
+    iterator erase(const_iterator position, const_iterator last);
+
+    void resize(size_type sz);
+    void resize(size_type sz, const value_type& c);
+
+    void swap(list&)
+        noexcept(allocator_traits<allocator_type>::is_always_equal::value);  // C++17
+    void clear() noexcept;
+
+    void splice(const_iterator position, list& x);
+    void splice(const_iterator position, list&& x);
+    void splice(const_iterator position, list& x, const_iterator i);
+    void splice(const_iterator position, list&& x, const_iterator i);
+    void splice(const_iterator position, list& x, const_iterator first,
+                                                  const_iterator last);
+    void splice(const_iterator position, list&& x, const_iterator first,
+                                                  const_iterator last);
+
+    size_type remove(const value_type& value);       // void before C++20
+    template <class Pred>
+      size_type remove_if(Pred pred);                // void before C++20
+    size_type unique();                              // void before C++20
+    template <class BinaryPredicate>
+      size_type unique(BinaryPredicate binary_pred); // void before C++20
+    void merge(list& x);
+    void merge(list&& x);
+    template <class Compare>
+        void merge(list& x, Compare comp);
+    template <class Compare>
+        void merge(list&& x, Compare comp);
+    void sort();
+    template <class Compare>
+        void sort(Compare comp);
+    void reverse() noexcept;
+};
+
+
+template <class InputIterator, class Allocator = allocator<typename iterator_traits<InputIterator>::value_type>>
+    list(InputIterator, InputIterator, Allocator = Allocator())
+    -> list<typename iterator_traits<InputIterator>::value_type, Allocator>;  // C++17
+
+template <class T, class Alloc>
+    bool operator==(const list<T,Alloc>& x, const list<T,Alloc>& y);
+template <class T, class Alloc>
+    bool operator< (const list<T,Alloc>& x, const list<T,Alloc>& y);
+template <class T, class Alloc>
+    bool operator!=(const list<T,Alloc>& x, const list<T,Alloc>& y);
+template <class T, class Alloc>
+    bool operator> (const list<T,Alloc>& x, const list<T,Alloc>& y);
+template <class T, class Alloc>
+    bool operator>=(const list<T,Alloc>& x, const list<T,Alloc>& y);
+template <class T, class Alloc>
+    bool operator<=(const list<T,Alloc>& x, const list<T,Alloc>& y);
+
+template <class T, class Alloc>
+    void swap(list<T,Alloc>& x, list<T,Alloc>& y)
+         noexcept(noexcept(x.swap(y)));
+
+template <class T, class Allocator, class U>
+    void erase(list<T, Allocator>& c, const U& value);       // C++20
+template <class T, class Allocator, class Predicate>
+    void erase_if(list<T, Allocator>& c, Predicate pred);    // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+
+#include <memory>
+#include <limits>
+#include <initializer_list>
+#include <iterator>
+#include <algorithm>
+#include <type_traits>
+#include <version>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp, class _VoidPtr> struct __list_node;
+template <class _Tp, class _VoidPtr> struct __list_node_base;
+
+template <class _Tp, class _VoidPtr>
+struct __list_node_pointer_traits {
+  typedef typename __rebind_pointer<_VoidPtr, __list_node<_Tp, _VoidPtr> >::type
+        __node_pointer;
+  typedef typename __rebind_pointer<_VoidPtr, __list_node_base<_Tp, _VoidPtr> >::type
+        __base_pointer;
+
+#if defined(_LIBCUDACXX_ABI_LIST_REMOVE_NODE_POINTER_UB)
+  typedef __base_pointer __link_pointer;
+#else
+  typedef typename conditional<
+          is_pointer<_VoidPtr>::value,
+          __base_pointer,
+          __node_pointer
+  >::type __link_pointer;
+#endif
+
+  typedef typename conditional<
+          is_same<__link_pointer, __node_pointer>::value,
+          __base_pointer,
+          __node_pointer
+  >::type __non_link_pointer;
+
+  static _LIBCUDACXX_INLINE_VISIBILITY
+  __link_pointer __unsafe_link_pointer_cast(__link_pointer __p) {
+      return __p;
+  }
+
+  static _LIBCUDACXX_INLINE_VISIBILITY
+  __link_pointer __unsafe_link_pointer_cast(__non_link_pointer __p) {
+      return static_cast<__link_pointer>(static_cast<_VoidPtr>(__p));
+  }
+
+};
+
+template <class _Tp, class _VoidPtr>
+struct __list_node_base
+{
+    typedef __list_node_pointer_traits<_Tp, _VoidPtr> _NodeTraits;
+    typedef typename _NodeTraits::__node_pointer __node_pointer;
+    typedef typename _NodeTraits::__base_pointer __base_pointer;
+    typedef typename _NodeTraits::__link_pointer __link_pointer;
+
+    __link_pointer __prev_;
+    __link_pointer __next_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_node_base() : __prev_(_NodeTraits::__unsafe_link_pointer_cast(__self())),
+                         __next_(_NodeTraits::__unsafe_link_pointer_cast(__self())) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __base_pointer __self() {
+        return pointer_traits<__base_pointer>::pointer_to(*this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node_pointer __as_node() {
+        return static_cast<__node_pointer>(__self());
+    }
+};
+
+template <class _Tp, class _VoidPtr>
+struct __list_node
+    : public __list_node_base<_Tp, _VoidPtr>
+{
+    _Tp __value_;
+
+    typedef __list_node_base<_Tp, _VoidPtr> __base;
+    typedef typename __base::__link_pointer __link_pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __link_pointer __as_link() {
+        return static_cast<__link_pointer>(__base::__self());
+    }
+};
+
+template <class _Tp, class _Alloc = allocator<_Tp> > class _LIBCUDACXX_TEMPLATE_VIS list;
+template <class _Tp, class _Alloc> class __list_imp;
+template <class _Tp, class _VoidPtr> class _LIBCUDACXX_TEMPLATE_VIS __list_const_iterator;
+
+template <class _Tp, class _VoidPtr>
+class _LIBCUDACXX_TEMPLATE_VIS __list_iterator
+{
+    typedef __list_node_pointer_traits<_Tp, _VoidPtr> _NodeTraits;
+    typedef typename _NodeTraits::__link_pointer __link_pointer;
+
+    __link_pointer __ptr_;
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __list_iterator(__link_pointer __p, const void* __c) _NOEXCEPT
+        : __ptr_(__p)
+    {
+        __get_db()->__insert_ic(this, __c);
+    }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __list_iterator(__link_pointer __p) _NOEXCEPT : __ptr_(__p) {}
+#endif
+
+
+
+    template<class, class> friend class list;
+    template<class, class> friend class __list_imp;
+    template<class, class> friend class __list_const_iterator;
+public:
+    typedef bidirectional_iterator_tag       iterator_category;
+    typedef _Tp                              value_type;
+    typedef value_type&                      reference;
+    typedef typename __rebind_pointer<_VoidPtr, value_type>::type pointer;
+    typedef typename pointer_traits<pointer>::difference_type difference_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator() _NOEXCEPT : __ptr_(nullptr)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__insert_i(this);
+#endif
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator(const __list_iterator& __p)
+        : __ptr_(__p.__ptr_)
+    {
+        __get_db()->__iterator_copy(this, &__p);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__list_iterator()
+    {
+        __get_db()->__erase_i(this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator& operator=(const __list_iterator& __p)
+    {
+        if (this != &__p)
+        {
+            __get_db()->__iterator_copy(this, &__p);
+            __ptr_ = __p.__ptr_;
+        }
+        return *this;
+    }
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to dereference a non-dereferenceable list::iterator");
+#endif
+        return __ptr_->__as_node()->__value_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to dereference a non-dereferenceable list::iterator");
+#endif
+        return pointer_traits<pointer>::pointer_to(__ptr_->__as_node()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator& operator++()
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to increment non-incrementable list::iterator");
+#endif
+        __ptr_ = __ptr_->__next_;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator operator++(int) {__list_iterator __t(*this); ++(*this); return __t;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator& operator--()
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__decrementable(this),
+                       "Attempted to decrement non-decrementable list::iterator");
+#endif
+        __ptr_ = __ptr_->__prev_;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_iterator operator--(int) {__list_iterator __t(*this); --(*this); return __t;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __list_iterator& __x, const __list_iterator& __y)
+    {
+        return __x.__ptr_ == __y.__ptr_;
+    }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+     bool operator!=(const __list_iterator& __x, const __list_iterator& __y)
+        {return !(__x == __y);}
+};
+
+template <class _Tp, class _VoidPtr>
+class _LIBCUDACXX_TEMPLATE_VIS __list_const_iterator
+{
+    typedef __list_node_pointer_traits<_Tp, _VoidPtr> _NodeTraits;
+    typedef typename _NodeTraits::__link_pointer __link_pointer;
+
+    __link_pointer __ptr_;
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __list_const_iterator(__link_pointer __p, const void* __c) _NOEXCEPT
+        : __ptr_(__p)
+    {
+        __get_db()->__insert_ic(this, __c);
+    }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __list_const_iterator(__link_pointer __p) _NOEXCEPT : __ptr_(__p) {}
+#endif
+
+    template<class, class> friend class list;
+    template<class, class> friend class __list_imp;
+public:
+    typedef bidirectional_iterator_tag       iterator_category;
+    typedef _Tp                              value_type;
+    typedef const value_type&                reference;
+    typedef typename __rebind_pointer<_VoidPtr, const value_type>::type pointer;
+    typedef typename pointer_traits<pointer>::difference_type difference_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator() _NOEXCEPT : __ptr_(nullptr)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__insert_i(this);
+#endif
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator(const __list_iterator<_Tp, _VoidPtr>& __p) _NOEXCEPT
+        : __ptr_(__p.__ptr_)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__iterator_copy(this, &__p);
+#endif
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator(const __list_const_iterator& __p)
+        : __ptr_(__p.__ptr_)
+    {
+        __get_db()->__iterator_copy(this, &__p);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__list_const_iterator()
+    {
+        __get_db()->__erase_i(this);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator& operator=(const __list_const_iterator& __p)
+    {
+        if (this != &__p)
+        {
+            __get_db()->__iterator_copy(this, &__p);
+            __ptr_ = __p.__ptr_;
+        }
+        return *this;
+    }
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to dereference a non-dereferenceable list::const_iterator");
+#endif
+        return __ptr_->__as_node()->__value_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to dereference a non-dereferenceable list::const_iterator");
+#endif
+        return pointer_traits<pointer>::pointer_to(__ptr_->__as_node()->__value_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator& operator++()
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(this),
+                       "Attempted to increment non-incrementable list::const_iterator");
+#endif
+        __ptr_ = __ptr_->__next_;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator operator++(int) {__list_const_iterator __t(*this); ++(*this); return __t;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator& operator--()
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__decrementable(this),
+                       "Attempted to decrement non-decrementable list::const_iterator");
+#endif
+        __ptr_ = __ptr_->__prev_;
+        return *this;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_const_iterator operator--(int) {__list_const_iterator __t(*this); --(*this); return __t;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __list_const_iterator& __x, const __list_const_iterator& __y)
+    {
+        return __x.__ptr_ == __y.__ptr_;
+    }
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __list_const_iterator& __x, const __list_const_iterator& __y)
+        {return !(__x == __y);}
+};
+
+template <class _Tp, class _Alloc>
+class __list_imp
+{
+    __list_imp(const __list_imp&);
+    __list_imp& operator=(const __list_imp&);
+public:
+    typedef _Alloc                                                  allocator_type;
+    typedef allocator_traits<allocator_type>                        __alloc_traits;
+    typedef typename __alloc_traits::size_type                      size_type;
+protected:
+    typedef _Tp                                                     value_type;
+    typedef typename __alloc_traits::void_pointer                   __void_pointer;
+    typedef __list_iterator<value_type, __void_pointer>             iterator;
+    typedef __list_const_iterator<value_type, __void_pointer>       const_iterator;
+    typedef __list_node_base<value_type, __void_pointer>            __node_base;
+    typedef __list_node<value_type, __void_pointer>                 __node;
+    typedef typename __rebind_alloc_helper<__alloc_traits, __node>::type __node_allocator;
+    typedef allocator_traits<__node_allocator>                       __node_alloc_traits;
+    typedef typename __node_alloc_traits::pointer                    __node_pointer;
+    typedef typename __node_alloc_traits::pointer                    __node_const_pointer;
+    typedef __list_node_pointer_traits<value_type, __void_pointer> __node_pointer_traits;
+    typedef typename __node_pointer_traits::__link_pointer __link_pointer;
+    typedef __link_pointer __link_const_pointer;
+    typedef typename __alloc_traits::pointer                         pointer;
+    typedef typename __alloc_traits::const_pointer                   const_pointer;
+    typedef typename __alloc_traits::difference_type                 difference_type;
+
+    typedef typename __rebind_alloc_helper<__alloc_traits, __node_base>::type __node_base_allocator;
+    typedef typename allocator_traits<__node_base_allocator>::pointer __node_base_pointer;
+    static_assert((!is_same<allocator_type, __node_allocator>::value),
+                  "internal allocator type must differ from user-specified "
+                  "type; otherwise overload resolution breaks");
+
+    __node_base __end_;
+    __compressed_pair<size_type, __node_allocator> __size_alloc_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __link_pointer __end_as_link() const _NOEXCEPT {
+        return __node_pointer_traits::__unsafe_link_pointer_cast(
+                const_cast<__node_base&>(__end_).__self());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          size_type& __sz() _NOEXCEPT {return __size_alloc_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const size_type& __sz() const _NOEXCEPT
+        {return __size_alloc_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          __node_allocator& __node_alloc() _NOEXCEPT
+          {return __size_alloc_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const __node_allocator& __node_alloc() const _NOEXCEPT
+        {return __size_alloc_.second();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __node_alloc_max_size() const _NOEXCEPT {
+        return __node_alloc_traits::max_size(__node_alloc());
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __unlink_nodes(__link_pointer __f, __link_pointer __l) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_imp()
+        _NOEXCEPT_(is_nothrow_default_constructible<__node_allocator>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_imp(const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __list_imp(const __node_allocator& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __list_imp(__node_allocator&& __a) _NOEXCEPT;
+#endif
+    ~__list_imp();
+    void clear() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return __sz() == 0;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator begin() _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return iterator(__end_.__next_, this);
+#else
+        return iterator(__end_.__next_);
+#endif
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const  _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return const_iterator(__end_.__next_, this);
+#else
+        return const_iterator(__end_.__next_);
+#endif
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator end() _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return iterator(__end_as_link(), this);
+#else
+        return iterator(__end_as_link());
+#endif
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        return const_iterator(__end_as_link(), this);
+#else
+        return const_iterator(__end_as_link());
+#endif
+    }
+
+    void swap(__list_imp& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __list_imp& __c)
+        {__copy_assign_alloc(__c, integral_constant<bool,
+                      __node_alloc_traits::propagate_on_container_copy_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__list_imp& __c)
+        _NOEXCEPT_(
+            !__node_alloc_traits::propagate_on_container_move_assignment::value ||
+            is_nothrow_move_assignable<__node_allocator>::value)
+        {__move_assign_alloc(__c, integral_constant<bool,
+                      __node_alloc_traits::propagate_on_container_move_assignment::value>());}
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __list_imp& __c, true_type)
+        {
+            if (__node_alloc() != __c.__node_alloc())
+                clear();
+            __node_alloc() = __c.__node_alloc();
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __list_imp&, false_type)
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__list_imp& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<__node_allocator>::value)
+        {
+            __node_alloc() = _CUDA_VSTD::move(__c.__node_alloc());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__list_imp&, false_type)
+        _NOEXCEPT
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __invalidate_all_iterators() {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+      __get_db()->__invalidate_all(this);
+#endif
+    }
+};
+
+// Unlink nodes [__f, __l]
+template <class _Tp, class _Alloc>
+inline
+void
+__list_imp<_Tp, _Alloc>::__unlink_nodes(__link_pointer __f, __link_pointer __l)
+    _NOEXCEPT
+{
+    __f->__prev_->__next_ = __l->__next_;
+    __l->__next_->__prev_ = __f->__prev_;
+}
+
+template <class _Tp, class _Alloc>
+inline
+__list_imp<_Tp, _Alloc>::__list_imp()
+        _NOEXCEPT_(is_nothrow_default_constructible<__node_allocator>::value)
+    : __size_alloc_(0)
+{
+}
+
+template <class _Tp, class _Alloc>
+inline
+__list_imp<_Tp, _Alloc>::__list_imp(const allocator_type& __a)
+    : __size_alloc_(0, __node_allocator(__a))
+{
+}
+
+template <class _Tp, class _Alloc>
+inline __list_imp<_Tp, _Alloc>::__list_imp(const __node_allocator& __a)
+    : __size_alloc_(0, __a) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Alloc>
+inline __list_imp<_Tp, _Alloc>::__list_imp(__node_allocator&& __a) _NOEXCEPT
+    : __size_alloc_(0, std::move(__a)) {}
+#endif
+
+template <class _Tp, class _Alloc>
+__list_imp<_Tp, _Alloc>::~__list_imp() {
+  clear();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__erase_c(this);
+#endif
+}
+
+template <class _Tp, class _Alloc>
+void
+__list_imp<_Tp, _Alloc>::clear() _NOEXCEPT
+{
+    if (!empty())
+    {
+        __node_allocator& __na = __node_alloc();
+        __link_pointer __f = __end_.__next_;
+        __link_pointer __l = __end_as_link();
+        __unlink_nodes(__f, __l->__prev_);
+        __sz() = 0;
+        while (__f != __l)
+        {
+            __node_pointer __np = __f->__as_node();
+            __f = __f->__next_;
+            __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(__np->__value_));
+            __node_alloc_traits::deallocate(__na, __np, 1);
+        }
+        __invalidate_all_iterators();
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+__list_imp<_Tp, _Alloc>::swap(__list_imp& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value)
+#endif
+{
+    _LIBCUDACXX_ASSERT(__alloc_traits::propagate_on_container_swap::value ||
+                   this->__node_alloc() == __c.__node_alloc(),
+                   "list::swap: Either propagate_on_container_swap must be true"
+                   " or the allocators must compare equal");
+    using _CUDA_VSTD::swap;
+    __swap_allocator(__node_alloc(), __c.__node_alloc());
+    swap(__sz(), __c.__sz());
+    swap(__end_, __c.__end_);
+    if (__sz() == 0)
+        __end_.__next_ = __end_.__prev_ = __end_as_link();
+    else
+        __end_.__prev_->__next_ = __end_.__next_->__prev_ = __end_as_link();
+    if (__c.__sz() == 0)
+        __c.__end_.__next_ = __c.__end_.__prev_ = __c.__end_as_link();
+    else
+        __c.__end_.__prev_->__next_ = __c.__end_.__next_->__prev_ = __c.__end_as_link();
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __libcpp_db* __db = __get_db();
+    __c_node* __cn1 = __db->__find_c_and_lock(this);
+    __c_node* __cn2 = __db->__find_c(&__c);
+    std::swap(__cn1->beg_, __cn2->beg_);
+    std::swap(__cn1->end_, __cn2->end_);
+    std::swap(__cn1->cap_, __cn2->cap_);
+    for (__i_node** __p = __cn1->end_; __p != __cn1->beg_;)
+    {
+        --__p;
+        const_iterator* __i = static_cast<const_iterator*>((*__p)->__i_);
+        if (__i->__ptr_ == __c.__end_as_link())
+        {
+            __cn2->__add(*__p);
+            if (--__cn1->end_ != __p)
+                memmove(__p, __p+1, (__cn1->end_ - __p)*sizeof(__i_node*));
+        }
+        else
+            (*__p)->__c_ = __cn1;
+    }
+    for (__i_node** __p = __cn2->end_; __p != __cn2->beg_;)
+    {
+        --__p;
+        const_iterator* __i = static_cast<const_iterator*>((*__p)->__i_);
+        if (__i->__ptr_ == __end_as_link())
+        {
+            __cn1->__add(*__p);
+            if (--__cn2->end_ != __p)
+                memmove(__p, __p+1, (__cn2->end_ - __p)*sizeof(__i_node*));
+        }
+        else
+            (*__p)->__c_ = __cn2;
+    }
+    __db->unlock();
+#endif
+}
+
+template <class _Tp, class _Alloc /*= allocator<_Tp>*/>
+class _LIBCUDACXX_TEMPLATE_VIS list
+    : private __list_imp<_Tp, _Alloc>
+{
+    typedef __list_imp<_Tp, _Alloc> base;
+    typedef typename base::__node              __node;
+    typedef typename base::__node_allocator    __node_allocator;
+    typedef typename base::__node_pointer      __node_pointer;
+    typedef typename base::__node_alloc_traits __node_alloc_traits;
+    typedef typename base::__node_base         __node_base;
+    typedef typename base::__node_base_pointer __node_base_pointer;
+    typedef typename base::__link_pointer __link_pointer;
+
+public:
+    typedef _Tp                                      value_type;
+    typedef _Alloc                                   allocator_type;
+    static_assert((is_same<value_type, typename allocator_type::value_type>::value),
+                  "Invalid allocator::value_type");
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+    typedef typename base::pointer                   pointer;
+    typedef typename base::const_pointer             const_pointer;
+    typedef typename base::size_type                 size_type;
+    typedef typename base::difference_type           difference_type;
+    typedef typename base::iterator                  iterator;
+    typedef typename base::const_iterator            const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>         reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>   const_reverse_iterator;
+#if _LIBCUDACXX_STD_VER > 17
+    typedef size_type                                __remove_return_type;
+#else
+    typedef void                                     __remove_return_type;
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    list()
+        _NOEXCEPT_(is_nothrow_default_constructible<__node_allocator>::value)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__insert_c(this);
+#endif
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit list(const allocator_type& __a) : base(__a)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__insert_c(this);
+#endif
+    }
+    explicit list(size_type __n);
+#if _LIBCUDACXX_STD_VER > 11
+    explicit list(size_type __n, const allocator_type& __a);
+#endif
+    list(size_type __n, const value_type& __x);
+    list(size_type __n, const value_type& __x, const allocator_type& __a);
+    template <class _InpIter>
+        list(_InpIter __f, _InpIter __l,
+             typename enable_if<__is_input_iterator<_InpIter>::value>::type* = 0);
+    template <class _InpIter>
+        list(_InpIter __f, _InpIter __l, const allocator_type& __a,
+             typename enable_if<__is_input_iterator<_InpIter>::value>::type* = 0);
+
+    list(const list& __c);
+    list(const list& __c, const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    list& operator=(const list& __c);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    list(initializer_list<value_type> __il);
+    list(initializer_list<value_type> __il, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    list(list&& __c)
+        _NOEXCEPT_(is_nothrow_move_constructible<__node_allocator>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    list(list&& __c, const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    list& operator=(list&& __c)
+        _NOEXCEPT_(
+            __node_alloc_traits::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<__node_allocator>::value);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    list& operator=(initializer_list<value_type> __il)
+        {assign(__il.begin(), __il.end()); return *this;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(initializer_list<value_type> __il)
+        {assign(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template <class _InpIter>
+        void assign(_InpIter __f, _InpIter __l,
+             typename enable_if<__is_input_iterator<_InpIter>::value>::type* = 0);
+    void assign(size_type __n, const value_type& __x);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT     {return base::__sz();}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT         {return base::empty();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT
+        {
+            return std::min<size_type>(
+                base::__node_alloc_max_size(),
+                numeric_limits<difference_type >::max());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator begin() _NOEXCEPT        {return base::begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const _NOEXCEPT {return base::begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator end() _NOEXCEPT          {return base::end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const _NOEXCEPT {return base::end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return base::begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend()   const _NOEXCEPT {return base::end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rbegin() _NOEXCEPT
+            {return       reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin()  const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rend() _NOEXCEPT
+            {return       reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend()    const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend()   const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference front()
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "list::front called on empty list");
+        return base::__end_.__next_->__as_node()->__value_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference front() const
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "list::front called on empty list");
+        return base::__end_.__next_->__as_node()->__value_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference back()
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "list::back called on empty list");
+        return base::__end_.__prev_->__as_node()->__value_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference back() const
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "list::back called on empty list");
+        return base::__end_.__prev_->__as_node()->__value_;
+    }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    void push_front(value_type&& __x);
+    void push_back(value_type&& __x);
+
+    template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+       reference emplace_front(_Args&&... __args);
+#else
+       void      emplace_front(_Args&&... __args);
+#endif
+    template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+        reference emplace_back(_Args&&... __args);
+#else
+       void       emplace_back(_Args&&... __args);
+#endif
+    template <class... _Args>
+        iterator emplace(const_iterator __p, _Args&&... __args);
+
+    iterator insert(const_iterator __p, value_type&& __x);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, initializer_list<value_type> __il)
+        {return insert(__p, __il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    void push_front(const value_type& __x);
+    void push_back(const value_type& __x);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Arg>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __emplace_back(_Arg&& __arg) { emplace_back(_CUDA_VSTD::forward<_Arg>(__arg)); }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __emplace_back(value_type const& __arg) { push_back(__arg); }
+#endif
+
+    iterator insert(const_iterator __p, const value_type& __x);
+    iterator insert(const_iterator __p, size_type __n, const value_type& __x);
+    template <class _InpIter>
+        iterator insert(const_iterator __p, _InpIter __f, _InpIter __l,
+             typename enable_if<__is_input_iterator<_InpIter>::value>::type* = 0);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(list& __c)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__node_alloc_traits::propagate_on_container_swap::value ||
+                   __is_nothrow_swappable<__node_allocator>::value)
+#endif
+        {base::swap(__c);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {base::clear();}
+
+    void pop_front();
+    void pop_back();
+
+    iterator erase(const_iterator __p);
+    iterator erase(const_iterator __f, const_iterator __l);
+
+    void resize(size_type __n);
+    void resize(size_type __n, const value_type& __x);
+
+    void splice(const_iterator __p, list& __c);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void splice(const_iterator __p, list&& __c) {splice(__p, __c);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void splice(const_iterator __p, list&& __c, const_iterator __i)
+        {splice(__p, __c, __i);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void splice(const_iterator __p, list&& __c, const_iterator __f, const_iterator __l)
+        {splice(__p, __c, __f, __l);}
+#endif
+    void splice(const_iterator __p, list& __c, const_iterator __i);
+    void splice(const_iterator __p, list& __c, const_iterator __f, const_iterator __l);
+
+    __remove_return_type remove(const value_type& __x);
+    template <class _Pred> __remove_return_type remove_if(_Pred __pred);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __remove_return_type unique() { return unique(__equal_to<value_type>()); }
+    template <class _BinaryPred>
+        __remove_return_type unique(_BinaryPred __binary_pred);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(list& __c);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(list&& __c) {merge(__c);}
+
+    template <class _Comp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        void merge(list&& __c, _Comp __comp) {merge(__c, __comp);}
+#endif
+    template <class _Comp>
+        void merge(list& __c, _Comp __comp);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void sort();
+    template <class _Comp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void sort(_Comp __comp);
+
+    void reverse() _NOEXCEPT;
+
+    bool __invariants() const;
+
+    typedef __allocator_destructor<__node_allocator> __node_destructor;
+    typedef unique_ptr<__node, __node_destructor> __hold_pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hold_pointer __allocate_node(__node_allocator& __na) {
+      __node_pointer __p = __node_alloc_traits::allocate(__na, 1);
+      __p->__prev_ = nullptr;
+      return __hold_pointer(__p, __node_destructor(__na, 1));
+    }
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const;
+    bool __decrementable(const const_iterator* __i) const;
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const;
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const;
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __link_nodes  (__link_pointer __p, __link_pointer __f, __link_pointer __l);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __link_nodes_at_front(__link_pointer __f, __link_pointer __l);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __link_nodes_at_back (__link_pointer __f, __link_pointer __l);
+    iterator __iterator(size_type __n);
+    template <class _Comp>
+        static iterator __sort(iterator __f1, iterator __e2, size_type __n, _Comp& __comp);
+
+    void __move_assign(list& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<__node_allocator>::value);
+    void __move_assign(list& __c, false_type);
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Alloc = typename std::allocator<typename iterator_traits<_InputIterator>::value_type>,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+list(_InputIterator, _InputIterator)
+  -> list<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+
+template<class _InputIterator,
+         class _Alloc,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+list(_InputIterator, _InputIterator, _Alloc)
+  -> list<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+#endif
+
+// Link in nodes [__f, __l] just prior to __p
+template <class _Tp, class _Alloc>
+inline
+void
+list<_Tp, _Alloc>::__link_nodes(__link_pointer __p, __link_pointer __f, __link_pointer __l)
+{
+    __p->__prev_->__next_ = __f;
+    __f->__prev_ = __p->__prev_;
+    __p->__prev_ = __l;
+    __l->__next_ = __p;
+}
+
+// Link in nodes [__f, __l] at the front of the list
+template <class _Tp, class _Alloc>
+inline
+void
+list<_Tp, _Alloc>::__link_nodes_at_front(__link_pointer __f, __link_pointer __l)
+{
+    __f->__prev_ = base::__end_as_link();
+    __l->__next_ = base::__end_.__next_;
+    __l->__next_->__prev_ = __l;
+    base::__end_.__next_ = __f;
+}
+
+// Link in nodes [__f, __l] at the back of the list
+template <class _Tp, class _Alloc>
+inline
+void
+list<_Tp, _Alloc>::__link_nodes_at_back(__link_pointer __f, __link_pointer __l)
+{
+    __l->__next_ = base::__end_as_link();
+    __f->__prev_ = base::__end_.__prev_;
+    __f->__prev_->__next_ = __f;
+    base::__end_.__prev_ = __l;
+}
+
+
+template <class _Tp, class _Alloc>
+inline
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::__iterator(size_type __n)
+{
+    return __n <= base::__sz() / 2 ? _CUDA_VSTD::next(begin(), __n)
+                                   : _CUDA_VSTD::prev(end(), base::__sz() - __n);
+}
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(size_type __n)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __n > 0; --__n)
+#ifndef _LIBCUDACXX_CXX03_LANG
+        emplace_back();
+#else
+        push_back(value_type());
+#endif
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(size_type __n, const allocator_type& __a) : base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __n > 0; --__n)
+        emplace_back();
+}
+#endif
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(size_type __n, const value_type& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __n > 0; --__n)
+        push_back(__x);
+}
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(size_type __n, const value_type& __x, const allocator_type& __a)
+    : base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __n > 0; --__n)
+        push_back(__x);
+}
+
+template <class _Tp, class _Alloc>
+template <class _InpIter>
+list<_Tp, _Alloc>::list(_InpIter __f, _InpIter __l,
+                        typename enable_if<__is_input_iterator<_InpIter>::value>::type*)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __f != __l; ++__f)
+        __emplace_back(*__f);
+}
+
+template <class _Tp, class _Alloc>
+template <class _InpIter>
+list<_Tp, _Alloc>::list(_InpIter __f, _InpIter __l, const allocator_type& __a,
+                        typename enable_if<__is_input_iterator<_InpIter>::value>::type*)
+    : base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __f != __l; ++__f)
+        __emplace_back(*__f);
+}
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(const list& __c)
+    : base(__node_alloc_traits::select_on_container_copy_construction(
+          __c.__node_alloc())) {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (const_iterator __i = __c.begin(), __e = __c.end(); __i != __e; ++__i)
+        push_back(*__i);
+}
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(const list& __c, const allocator_type& __a)
+    : base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (const_iterator __i = __c.begin(), __e = __c.end(); __i != __e; ++__i)
+        push_back(*__i);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(initializer_list<value_type> __il, const allocator_type& __a)
+    : base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (typename initializer_list<value_type>::const_iterator __i = __il.begin(),
+            __e = __il.end(); __i != __e; ++__i)
+        push_back(*__i);
+}
+
+template <class _Tp, class _Alloc>
+list<_Tp, _Alloc>::list(initializer_list<value_type> __il)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (typename initializer_list<value_type>::const_iterator __i = __il.begin(),
+            __e = __il.end(); __i != __e; ++__i)
+        push_back(*__i);
+}
+
+template <class _Tp, class _Alloc>
+inline list<_Tp, _Alloc>::list(list&& __c)
+    _NOEXCEPT_(is_nothrow_move_constructible<__node_allocator>::value)
+    : base(_CUDA_VSTD::move(__c.__node_alloc())) {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    splice(end(), __c);
+}
+
+template <class _Tp, class _Alloc>
+inline
+list<_Tp, _Alloc>::list(list&& __c, const allocator_type& __a)
+    : base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__a == __c.get_allocator())
+        splice(end(), __c);
+    else
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__c.begin()), _Ip(__c.end()));
+    }
+}
+
+template <class _Tp, class _Alloc>
+inline
+list<_Tp, _Alloc>&
+list<_Tp, _Alloc>::operator=(list&& __c)
+        _NOEXCEPT_(
+            __node_alloc_traits::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<__node_allocator>::value)
+{
+    __move_assign(__c, integral_constant<bool,
+          __node_alloc_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::__move_assign(list& __c, false_type)
+{
+    if (base::__node_alloc() != __c.__node_alloc())
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__c.begin()), _Ip(__c.end()));
+    }
+    else
+        __move_assign(__c, true_type());
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::__move_assign(list& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<__node_allocator>::value)
+{
+    clear();
+    base::__move_assign_alloc(__c);
+    splice(end(), __c);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+inline
+list<_Tp, _Alloc>&
+list<_Tp, _Alloc>::operator=(const list& __c)
+{
+    if (this != &__c)
+    {
+        base::__copy_assign_alloc(__c);
+        assign(__c.begin(), __c.end());
+    }
+    return *this;
+}
+
+template <class _Tp, class _Alloc>
+template <class _InpIter>
+void
+list<_Tp, _Alloc>::assign(_InpIter __f, _InpIter __l,
+                          typename enable_if<__is_input_iterator<_InpIter>::value>::type*)
+{
+    iterator __i = begin();
+    iterator __e = end();
+    for (; __f != __l && __i != __e; ++__f, ++__i)
+        *__i = *__f;
+    if (__i == __e)
+        insert(__e, __f, __l);
+    else
+        erase(__i, __e);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+      __get_db()->__invalidate_all(this);
+#endif
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::assign(size_type __n, const value_type& __x)
+{
+    iterator __i = begin();
+    iterator __e = end();
+    for (; __n > 0 && __i != __e; --__n, ++__i)
+        *__i = __x;
+    if (__i == __e)
+        insert(__e, __n, __x);
+    else
+        erase(__i, __e);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+      __get_db()->__invalidate_all(this);
+#endif
+}
+
+template <class _Tp, class _Alloc>
+inline
+_Alloc
+list<_Tp, _Alloc>::get_allocator() const _NOEXCEPT
+{
+    return allocator_type(base::__node_alloc());
+}
+
+template <class _Tp, class _Alloc>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::insert(const_iterator __p, const value_type& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::insert(iterator, x) called with an iterator not"
+        " referring to this list");
+#endif
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+    __link_nodes(__p.__ptr_, __hold->__as_link(), __hold->__as_link());
+    ++base::__sz();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__hold.release()->__as_link(), this);
+#else
+    return iterator(__hold.release()->__as_link());
+#endif
+}
+
+template <class _Tp, class _Alloc>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::insert(const_iterator __p, size_type __n, const value_type& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::insert(iterator, n, x) called with an iterator not"
+        " referring to this list");
+    iterator __r(__p.__ptr_, this);
+#else
+    iterator __r(__p.__ptr_);
+#endif
+    if (__n > 0)
+    {
+        size_type __ds = 0;
+        __node_allocator& __na = base::__node_alloc();
+        __hold_pointer __hold = __allocate_node(__na);
+        __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+        ++__ds;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __r = iterator(__hold->__as_link(), this);
+#else
+        __r = iterator(__hold->__as_link());
+#endif
+        __hold.release();
+        iterator __e = __r;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (--__n; __n != 0; --__n, ++__e, ++__ds)
+            {
+                __hold.reset(__node_alloc_traits::allocate(__na, 1));
+                __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+                __e.__ptr_->__next_ = __hold->__as_link();
+                __hold->__prev_ = __e.__ptr_;
+                __hold.release();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            while (true)
+            {
+                __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(*__e));
+                __link_pointer __prev = __e.__ptr_->__prev_;
+                __node_alloc_traits::deallocate(__na, __e.__ptr_->__as_node(), 1);
+                if (__prev == 0)
+                    break;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                __e = iterator(__prev, this);
+#else
+                __e = iterator(__prev);
+#endif
+            }
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __link_nodes(__p.__ptr_, __r.__ptr_, __e.__ptr_);
+        base::__sz() += __ds;
+    }
+    return __r;
+}
+
+template <class _Tp, class _Alloc>
+template <class _InpIter>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::insert(const_iterator __p, _InpIter __f, _InpIter __l,
+             typename enable_if<__is_input_iterator<_InpIter>::value>::type*)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::insert(iterator, range) called with an iterator not"
+        " referring to this list");
+    iterator __r(__p.__ptr_, this);
+#else
+    iterator __r(__p.__ptr_);
+#endif
+    if (__f != __l)
+    {
+        size_type __ds = 0;
+        __node_allocator& __na = base::__node_alloc();
+        __hold_pointer __hold = __allocate_node(__na);
+        __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), *__f);
+        ++__ds;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __r = iterator(__hold.get()->__as_link(), this);
+#else
+        __r = iterator(__hold.get()->__as_link());
+#endif
+        __hold.release();
+        iterator __e = __r;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (++__f; __f != __l; ++__f, (void) ++__e, (void) ++__ds)
+            {
+                __hold.reset(__node_alloc_traits::allocate(__na, 1));
+                __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), *__f);
+                __e.__ptr_->__next_ = __hold.get()->__as_link();
+                __hold->__prev_ = __e.__ptr_;
+                __hold.release();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            while (true)
+            {
+                __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(*__e));
+                __link_pointer __prev = __e.__ptr_->__prev_;
+                __node_alloc_traits::deallocate(__na, __e.__ptr_->__as_node(), 1);
+                if (__prev == 0)
+                    break;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                __e = iterator(__prev, this);
+#else
+                __e = iterator(__prev);
+#endif
+            }
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __link_nodes(__p.__ptr_, __r.__ptr_, __e.__ptr_);
+        base::__sz() += __ds;
+    }
+    return __r;
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::push_front(const value_type& __x)
+{
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+    __link_pointer __nl = __hold->__as_link();
+    __link_nodes_at_front(__nl, __nl);
+    ++base::__sz();
+    __hold.release();
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::push_back(const value_type& __x)
+{
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+    __link_nodes_at_back(__hold.get()->__as_link(), __hold.get()->__as_link());
+    ++base::__sz();
+    __hold.release();
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::push_front(value_type&& __x)
+{
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), _CUDA_VSTD::move(__x));
+    __link_nodes_at_front(__hold.get()->__as_link(), __hold.get()->__as_link());
+    ++base::__sz();
+    __hold.release();
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::push_back(value_type&& __x)
+{
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), _CUDA_VSTD::move(__x));
+    __link_nodes_at_back(__hold.get()->__as_link(), __hold.get()->__as_link());
+    ++base::__sz();
+    __hold.release();
+}
+
+template <class _Tp, class _Alloc>
+template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+typename list<_Tp, _Alloc>::reference
+#else
+void
+#endif
+list<_Tp, _Alloc>::emplace_front(_Args&&... __args)
+{
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), _CUDA_VSTD::forward<_Args>(__args)...);
+    __link_nodes_at_front(__hold.get()->__as_link(), __hold.get()->__as_link());
+    ++base::__sz();
+#if _LIBCUDACXX_STD_VER > 14
+    return __hold.release()->__value_;
+#else
+    __hold.release();
+#endif
+}
+
+template <class _Tp, class _Alloc>
+template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+typename list<_Tp, _Alloc>::reference
+#else
+void
+#endif
+list<_Tp, _Alloc>::emplace_back(_Args&&... __args)
+{
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), _CUDA_VSTD::forward<_Args>(__args)...);
+    __link_pointer __nl = __hold->__as_link();
+    __link_nodes_at_back(__nl, __nl);
+    ++base::__sz();
+#if _LIBCUDACXX_STD_VER > 14
+    return __hold.release()->__value_;
+#else
+    __hold.release();
+#endif
+}
+
+template <class _Tp, class _Alloc>
+template <class... _Args>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::emplace(const_iterator __p, _Args&&... __args)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::emplace(iterator, args...) called with an iterator not"
+        " referring to this list");
+#endif
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), _CUDA_VSTD::forward<_Args>(__args)...);
+    __link_pointer __nl = __hold.get()->__as_link();
+    __link_nodes(__p.__ptr_, __nl, __nl);
+    ++base::__sz();
+    __hold.release();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__nl, this);
+#else
+    return iterator(__nl);
+#endif
+}
+
+template <class _Tp, class _Alloc>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::insert(const_iterator __p, value_type&& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::insert(iterator, x) called with an iterator not"
+        " referring to this list");
+#endif
+    __node_allocator& __na = base::__node_alloc();
+    __hold_pointer __hold = __allocate_node(__na);
+    __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), _CUDA_VSTD::move(__x));
+    __link_pointer __nl = __hold->__as_link();
+    __link_nodes(__p.__ptr_, __nl, __nl);
+    ++base::__sz();
+    __hold.release();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__nl, this);
+#else
+    return iterator(__nl);
+#endif
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::pop_front()
+{
+    _LIBCUDACXX_ASSERT(!empty(), "list::pop_front() called with empty list");
+    __node_allocator& __na = base::__node_alloc();
+    __link_pointer __n = base::__end_.__next_;
+    base::__unlink_nodes(__n, __n);
+    --base::__sz();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __c_node* __c = __get_db()->__find_c_and_lock(this);
+    for (__i_node** __p = __c->end_; __p != __c->beg_; )
+    {
+        --__p;
+        iterator* __i = static_cast<iterator*>((*__p)->__i_);
+        if (__i->__ptr_ == __n)
+        {
+            (*__p)->__c_ = nullptr;
+            if (--__c->end_ != __p)
+                memmove(__p, __p+1, (__c->end_ - __p)*sizeof(__i_node*));
+        }
+    }
+    __get_db()->unlock();
+#endif
+    __node_pointer __np = __n->__as_node();
+    __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(__np->__value_));
+    __node_alloc_traits::deallocate(__na, __np, 1);
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::pop_back()
+{
+    _LIBCUDACXX_ASSERT(!empty(), "list::pop_back() called with empty list");
+    __node_allocator& __na = base::__node_alloc();
+    __link_pointer __n = base::__end_.__prev_;
+    base::__unlink_nodes(__n, __n);
+    --base::__sz();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __c_node* __c = __get_db()->__find_c_and_lock(this);
+    for (__i_node** __p = __c->end_; __p != __c->beg_; )
+    {
+        --__p;
+        iterator* __i = static_cast<iterator*>((*__p)->__i_);
+        if (__i->__ptr_ == __n)
+        {
+            (*__p)->__c_ = nullptr;
+            if (--__c->end_ != __p)
+                memmove(__p, __p+1, (__c->end_ - __p)*sizeof(__i_node*));
+        }
+    }
+    __get_db()->unlock();
+#endif
+    __node_pointer __np = __n->__as_node();
+    __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(__np->__value_));
+    __node_alloc_traits::deallocate(__na, __np, 1);
+}
+
+template <class _Tp, class _Alloc>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::erase(const_iterator __p)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::erase(iterator) called with an iterator not"
+        " referring to this list");
+#endif
+    _LIBCUDACXX_ASSERT(__p != end(),
+        "list::erase(iterator) called with a non-dereferenceable iterator");
+    __node_allocator& __na = base::__node_alloc();
+    __link_pointer __n = __p.__ptr_;
+    __link_pointer __r = __n->__next_;
+    base::__unlink_nodes(__n, __n);
+    --base::__sz();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __c_node* __c = __get_db()->__find_c_and_lock(this);
+    for (__i_node** __ip = __c->end_; __ip != __c->beg_; )
+    {
+        --__ip;
+        iterator* __i = static_cast<iterator*>((*__ip)->__i_);
+        if (__i->__ptr_ == __n)
+        {
+            (*__ip)->__c_ = nullptr;
+            if (--__c->end_ != __ip)
+                memmove(__ip, __ip+1, (__c->end_ - __ip)*sizeof(__i_node*));
+        }
+    }
+    __get_db()->unlock();
+#endif
+    __node_pointer __np = __n->__as_node();
+    __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(__np->__value_));
+    __node_alloc_traits::deallocate(__na, __np, 1);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__r, this);
+#else
+    return iterator(__r);
+#endif
+}
+
+template <class _Tp, class _Alloc>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::erase(const_iterator __f, const_iterator __l)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__f) == this,
+        "list::erase(iterator, iterator) called with an iterator not"
+        " referring to this list");
+   _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__l) == this,
+        "list::erase(iterator, iterator) called with an iterator not"
+        " referring to this list");
+#endif
+    if (__f != __l)
+    {
+        __node_allocator& __na = base::__node_alloc();
+        base::__unlink_nodes(__f.__ptr_, __l.__ptr_->__prev_);
+        while (__f != __l)
+        {
+            __link_pointer __n = __f.__ptr_;
+            ++__f;
+            --base::__sz();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            __c_node* __c = __get_db()->__find_c_and_lock(this);
+            for (__i_node** __p = __c->end_; __p != __c->beg_; )
+            {
+                --__p;
+                iterator* __i = static_cast<iterator*>((*__p)->__i_);
+                if (__i->__ptr_ == __n)
+                {
+                    (*__p)->__c_ = nullptr;
+                    if (--__c->end_ != __p)
+                        memmove(__p, __p+1, (__c->end_ - __p)*sizeof(__i_node*));
+                }
+            }
+            __get_db()->unlock();
+#endif
+            __node_pointer __np = __n->__as_node();
+            __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(__np->__value_));
+            __node_alloc_traits::deallocate(__na, __np, 1);
+        }
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(__l.__ptr_, this);
+#else
+    return iterator(__l.__ptr_);
+#endif
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::resize(size_type __n)
+{
+    if (__n < base::__sz())
+        erase(__iterator(__n), end());
+    else if (__n > base::__sz())
+    {
+        __n -= base::__sz();
+        size_type __ds = 0;
+        __node_allocator& __na = base::__node_alloc();
+        __hold_pointer __hold = __allocate_node(__na);
+        __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_));
+        ++__ds;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        iterator __r = iterator(__hold.release()->__as_link(), this);
+#else
+        iterator __r = iterator(__hold.release()->__as_link());
+#endif
+        iterator __e = __r;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (--__n; __n != 0; --__n, ++__e, ++__ds)
+            {
+                __hold.reset(__node_alloc_traits::allocate(__na, 1));
+                __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_));
+                __e.__ptr_->__next_ = __hold.get()->__as_link();
+                __hold->__prev_ = __e.__ptr_;
+                __hold.release();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            while (true)
+            {
+                __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(*__e));
+                __link_pointer __prev = __e.__ptr_->__prev_;
+                __node_alloc_traits::deallocate(__na, __e.__ptr_->__as_node(), 1);
+                if (__prev == 0)
+                    break;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                __e = iterator(__prev, this);
+#else
+                __e = iterator(__prev);
+#endif
+            }
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __link_nodes_at_back(__r.__ptr_, __e.__ptr_);
+        base::__sz() += __ds;
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::resize(size_type __n, const value_type& __x)
+{
+    if (__n < base::__sz())
+        erase(__iterator(__n), end());
+    else if (__n > base::__sz())
+    {
+        __n -= base::__sz();
+        size_type __ds = 0;
+        __node_allocator& __na = base::__node_alloc();
+        __hold_pointer __hold = __allocate_node(__na);
+        __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+        ++__ds;
+        __link_pointer __nl = __hold.release()->__as_link();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        iterator __r = iterator(__nl, this);
+#else
+        iterator __r = iterator(__nl);
+#endif
+        iterator __e = __r;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (--__n; __n != 0; --__n, ++__e, ++__ds)
+            {
+                __hold.reset(__node_alloc_traits::allocate(__na, 1));
+                __node_alloc_traits::construct(__na, _CUDA_VSTD::addressof(__hold->__value_), __x);
+                __e.__ptr_->__next_ = __hold.get()->__as_link();
+                __hold->__prev_ = __e.__ptr_;
+                __hold.release();
+            }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            while (true)
+            {
+                __node_alloc_traits::destroy(__na, _CUDA_VSTD::addressof(*__e));
+                __link_pointer __prev = __e.__ptr_->__prev_;
+                __node_alloc_traits::deallocate(__na, __e.__ptr_->__as_node(), 1);
+                if (__prev == 0)
+                    break;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                __e = iterator(__prev, this);
+#else
+                __e = iterator(__prev);
+#endif
+            }
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        __link_nodes(base::__end_as_link(), __r.__ptr_, __e.__ptr_);
+        base::__sz() += __ds;
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::splice(const_iterator __p, list& __c)
+{
+    _LIBCUDACXX_ASSERT(this != &__c,
+                   "list::splice(iterator, list) called with this == &list");
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::splice(iterator, list) called with an iterator not"
+        " referring to this list");
+#endif
+    if (!__c.empty())
+    {
+        __link_pointer __f = __c.__end_.__next_;
+        __link_pointer __l = __c.__end_.__prev_;
+        base::__unlink_nodes(__f, __l);
+        __link_nodes(__p.__ptr_, __f, __l);
+        base::__sz() += __c.__sz();
+        __c.__sz() = 0;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        if (&__c != this) {
+            __libcpp_db* __db = __get_db();
+            __c_node* __cn1 = __db->__find_c_and_lock(this);
+            __c_node* __cn2 = __db->__find_c(&__c);
+            for (__i_node** __ip = __cn2->end_; __ip != __cn2->beg_;)
+            {
+                --__ip;
+                iterator* __i = static_cast<iterator*>((*__ip)->__i_);
+                if (__i->__ptr_ != __c.__end_as_link())
+                {
+                    __cn1->__add(*__ip);
+                    (*__ip)->__c_ = __cn1;
+                    if (--__cn2->end_ != __ip)
+                        memmove(__ip, __ip+1, (__cn2->end_ - __ip)*sizeof(__i_node*));
+                }
+            }
+            __db->unlock();
+        }
+#endif
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::splice(const_iterator __p, list& __c, const_iterator __i)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::splice(iterator, list, iterator) called with first iterator not"
+        " referring to this list");
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__i) == &__c,
+        "list::splice(iterator, list, iterator) called with second iterator not"
+        " referring to list argument");
+    _LIBCUDACXX_ASSERT(__get_const_db()->__dereferenceable(&__i),
+        "list::splice(iterator, list, iterator) called with second iterator not"
+        " derefereceable");
+#endif
+    if (__p.__ptr_ != __i.__ptr_ && __p.__ptr_ != __i.__ptr_->__next_)
+    {
+        __link_pointer __f = __i.__ptr_;
+        base::__unlink_nodes(__f, __f);
+        __link_nodes(__p.__ptr_, __f, __f);
+        --__c.__sz();
+        ++base::__sz();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        if (&__c != this) {
+            __libcpp_db* __db = __get_db();
+            __c_node* __cn1 = __db->__find_c_and_lock(this);
+            __c_node* __cn2 = __db->__find_c(&__c);
+            for (__i_node** __ip = __cn2->end_; __ip != __cn2->beg_;)
+            {
+                --__ip;
+                iterator* __j = static_cast<iterator*>((*__ip)->__i_);
+                if (__j->__ptr_ == __f)
+                {
+                    __cn1->__add(*__ip);
+                    (*__ip)->__c_ = __cn1;
+                    if (--__cn2->end_ != __ip)
+                        memmove(__ip, __ip+1, (__cn2->end_ - __ip)*sizeof(__i_node*));
+                }
+            }
+            __db->unlock();
+        }
+#endif
+    }
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::splice(const_iterator __p, list& __c, const_iterator __f, const_iterator __l)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+        "list::splice(iterator, list, iterator, iterator) called with first iterator not"
+        " referring to this list");
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__f) == &__c,
+        "list::splice(iterator, list, iterator, iterator) called with second iterator not"
+        " referring to list argument");
+    if (this == &__c)
+    {
+        for (const_iterator __i = __f; __i != __l; ++__i)
+            _LIBCUDACXX_ASSERT(__i != __p,
+                           "list::splice(iterator, list, iterator, iterator)"
+                           " called with the first iterator within the range"
+                           " of the second and third iterators");
+    }
+#endif
+    if (__f != __l)
+    {
+        __link_pointer __first = __f.__ptr_;
+        --__l;
+        __link_pointer __last = __l.__ptr_;
+        if (this != &__c)
+        {
+            size_type __s = _CUDA_VSTD::distance(__f, __l) + 1;
+            __c.__sz() -= __s;
+            base::__sz() += __s;
+        }
+        base::__unlink_nodes(__first, __last);
+        __link_nodes(__p.__ptr_, __first, __last);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        if (&__c != this) {
+            __libcpp_db* __db = __get_db();
+            __c_node* __cn1 = __db->__find_c_and_lock(this);
+            __c_node* __cn2 = __db->__find_c(&__c);
+            for (__i_node** __ip = __cn2->end_; __ip != __cn2->beg_;)
+            {
+                --__ip;
+                iterator* __j = static_cast<iterator*>((*__ip)->__i_);
+                for (__link_pointer __k = __f.__ptr_;
+                                              __k != __l.__ptr_; __k = __k->__next_)
+                {
+                    if (__j->__ptr_ == __k)
+                    {
+                        __cn1->__add(*__ip);
+                        (*__ip)->__c_ = __cn1;
+                        if (--__cn2->end_ != __ip)
+                            memmove(__ip, __ip+1, (__cn2->end_ - __ip)*sizeof(__i_node*));
+                    }
+                }
+            }
+            __db->unlock();
+        }
+#endif
+    }
+}
+
+template <class _Tp, class _Alloc>
+typename list<_Tp, _Alloc>::__remove_return_type
+list<_Tp, _Alloc>::remove(const value_type& __x)
+{
+    list<_Tp, _Alloc> __deleted_nodes(get_allocator()); // collect the nodes we're removing
+    for (const_iterator __i = begin(), __e = end(); __i != __e;)
+    {
+        if (*__i == __x)
+        {
+            const_iterator __j = _CUDA_VSTD::next(__i);
+            for (; __j != __e && *__j == __x; ++__j)
+                ;
+            __deleted_nodes.splice(__deleted_nodes.end(), *this, __i, __j);
+            __i = __j;
+            if (__i != __e)
+                ++__i;
+        }
+        else
+            ++__i;
+    }
+
+    return (__remove_return_type) __deleted_nodes.size();
+}
+
+template <class _Tp, class _Alloc>
+template <class _Pred>
+typename list<_Tp, _Alloc>::__remove_return_type
+list<_Tp, _Alloc>::remove_if(_Pred __pred)
+{
+    list<_Tp, _Alloc> __deleted_nodes(get_allocator()); // collect the nodes we're removing
+    for (iterator __i = begin(), __e = end(); __i != __e;)
+    {
+        if (__pred(*__i))
+        {
+            iterator __j = _CUDA_VSTD::next(__i);
+            for (; __j != __e && __pred(*__j); ++__j)
+                ;
+            __deleted_nodes.splice(__deleted_nodes.end(), *this, __i, __j);
+            __i = __j;
+            if (__i != __e)
+                ++__i;
+        }
+        else
+            ++__i;
+    }
+
+    return (__remove_return_type) __deleted_nodes.size();
+}
+
+template <class _Tp, class _Alloc>
+template <class _BinaryPred>
+typename list<_Tp, _Alloc>::__remove_return_type
+list<_Tp, _Alloc>::unique(_BinaryPred __binary_pred)
+{
+    list<_Tp, _Alloc> __deleted_nodes(get_allocator()); // collect the nodes we're removing
+    for (iterator __i = begin(), __e = end(); __i != __e;)
+    {
+        iterator __j = _CUDA_VSTD::next(__i);
+        for (; __j != __e && __binary_pred(*__i, *__j); ++__j)
+            ;
+        if (++__i != __j) {
+            __deleted_nodes.splice(__deleted_nodes.end(), *this, __i, __j);
+            __i = __j;
+            }
+    }
+    
+    return (__remove_return_type) __deleted_nodes.size();
+}
+
+template <class _Tp, class _Alloc>
+inline
+void
+list<_Tp, _Alloc>::merge(list& __c)
+{
+    merge(__c, __less<value_type>());
+}
+
+template <class _Tp, class _Alloc>
+template <class _Comp>
+void
+list<_Tp, _Alloc>::merge(list& __c, _Comp __comp)
+{
+    if (this != _CUDA_VSTD::addressof(__c))
+    {
+        iterator __f1 = begin();
+        iterator __e1 = end();
+        iterator __f2 = __c.begin();
+        iterator __e2 = __c.end();
+        while (__f1 != __e1 && __f2 != __e2)
+        {
+            if (__comp(*__f2, *__f1))
+            {
+                size_type __ds = 1;
+                iterator __m2 = _CUDA_VSTD::next(__f2);
+                for (; __m2 != __e2 && __comp(*__m2, *__f1); ++__m2, ++__ds)
+                    ;
+                base::__sz() += __ds;
+                __c.__sz() -= __ds;
+                __link_pointer __f = __f2.__ptr_;
+                __link_pointer __l = __m2.__ptr_->__prev_;
+                __f2 = __m2;
+                base::__unlink_nodes(__f, __l);
+                __m2 = _CUDA_VSTD::next(__f1);
+                __link_nodes(__f1.__ptr_, __f, __l);
+                __f1 = __m2;
+            }
+            else
+                ++__f1;
+        }
+        splice(__e1, __c);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __libcpp_db* __db = __get_db();
+        __c_node* __cn1 = __db->__find_c_and_lock(this);
+        __c_node* __cn2 = __db->__find_c(&__c);
+        for (__i_node** __p = __cn2->end_; __p != __cn2->beg_;)
+        {
+            --__p;
+            iterator* __i = static_cast<iterator*>((*__p)->__i_);
+            if (__i->__ptr_ != __c.__end_as_link())
+            {
+                __cn1->__add(*__p);
+                (*__p)->__c_ = __cn1;
+                if (--__cn2->end_ != __p)
+                    memmove(__p, __p+1, (__cn2->end_ - __p)*sizeof(__i_node*));
+            }
+        }
+        __db->unlock();
+#endif
+    }
+}
+
+template <class _Tp, class _Alloc>
+inline
+void
+list<_Tp, _Alloc>::sort()
+{
+    sort(__less<value_type>());
+}
+
+template <class _Tp, class _Alloc>
+template <class _Comp>
+inline
+void
+list<_Tp, _Alloc>::sort(_Comp __comp)
+{
+    __sort(begin(), end(), base::__sz(), __comp);
+}
+
+template <class _Tp, class _Alloc>
+template <class _Comp>
+typename list<_Tp, _Alloc>::iterator
+list<_Tp, _Alloc>::__sort(iterator __f1, iterator __e2, size_type __n, _Comp& __comp)
+{
+    switch (__n)
+    {
+    case 0:
+    case 1:
+        return __f1;
+    case 2:
+        if (__comp(*--__e2, *__f1))
+        {
+            __link_pointer __f = __e2.__ptr_;
+            base::__unlink_nodes(__f, __f);
+            __link_nodes(__f1.__ptr_, __f, __f);
+            return __e2;
+        }
+        return __f1;
+    }
+    size_type __n2 = __n / 2;
+    iterator __e1 = _CUDA_VSTD::next(__f1, __n2);
+    iterator  __r = __f1 = __sort(__f1, __e1, __n2, __comp);
+    iterator __f2 = __e1 = __sort(__e1, __e2, __n - __n2, __comp);
+    if (__comp(*__f2, *__f1))
+    {
+        iterator __m2 = _CUDA_VSTD::next(__f2);
+        for (; __m2 != __e2 && __comp(*__m2, *__f1); ++__m2)
+            ;
+        __link_pointer __f = __f2.__ptr_;
+        __link_pointer __l = __m2.__ptr_->__prev_;
+        __r = __f2;
+        __e1 = __f2 = __m2;
+        base::__unlink_nodes(__f, __l);
+        __m2 = _CUDA_VSTD::next(__f1);
+        __link_nodes(__f1.__ptr_, __f, __l);
+        __f1 = __m2;
+    }
+    else
+        ++__f1;
+    while (__f1 != __e1 && __f2 != __e2)
+    {
+        if (__comp(*__f2, *__f1))
+        {
+            iterator __m2 = _CUDA_VSTD::next(__f2);
+            for (; __m2 != __e2 && __comp(*__m2, *__f1); ++__m2)
+                ;
+            __link_pointer __f = __f2.__ptr_;
+            __link_pointer __l = __m2.__ptr_->__prev_;
+            if (__e1 == __f2)
+                __e1 = __m2;
+            __f2 = __m2;
+            base::__unlink_nodes(__f, __l);
+            __m2 = _CUDA_VSTD::next(__f1);
+            __link_nodes(__f1.__ptr_, __f, __l);
+            __f1 = __m2;
+        }
+        else
+            ++__f1;
+    }
+    return __r;
+}
+
+template <class _Tp, class _Alloc>
+void
+list<_Tp, _Alloc>::reverse() _NOEXCEPT
+{
+    if (base::__sz() > 1)
+    {
+        iterator __e = end();
+        for (iterator __i = begin(); __i.__ptr_ != __e.__ptr_;)
+        {
+            _CUDA_VSTD::swap(__i.__ptr_->__prev_, __i.__ptr_->__next_);
+            __i.__ptr_ = __i.__ptr_->__prev_;
+        }
+        _CUDA_VSTD::swap(__e.__ptr_->__prev_, __e.__ptr_->__next_);
+    }
+}
+
+template <class _Tp, class _Alloc>
+bool
+list<_Tp, _Alloc>::__invariants() const
+{
+    return size() == _CUDA_VSTD::distance(begin(), end());
+}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+template <class _Tp, class _Alloc>
+bool
+list<_Tp, _Alloc>::__dereferenceable(const const_iterator* __i) const
+{
+    return __i->__ptr_ != this->__end_as_link();
+}
+
+template <class _Tp, class _Alloc>
+bool
+list<_Tp, _Alloc>::__decrementable(const const_iterator* __i) const
+{
+    return !empty() &&  __i->__ptr_ != base::__end_.__next_;
+}
+
+template <class _Tp, class _Alloc>
+bool
+list<_Tp, _Alloc>::__addable(const const_iterator*, ptrdiff_t) const
+{
+    return false;
+}
+
+template <class _Tp, class _Alloc>
+bool
+list<_Tp, _Alloc>::__subscriptable(const const_iterator*, ptrdiff_t) const
+{
+    return false;
+}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const list<_Tp, _Alloc>& __x, const list<_Tp, _Alloc>& __y)
+{
+    return __x.size() == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const list<_Tp, _Alloc>& __x, const list<_Tp, _Alloc>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const list<_Tp, _Alloc>& __x, const list<_Tp, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const list<_Tp, _Alloc>& __x, const list<_Tp, _Alloc>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const list<_Tp, _Alloc>& __x, const list<_Tp, _Alloc>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const list<_Tp, _Alloc>& __x, const list<_Tp, _Alloc>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(list<_Tp, _Alloc>& __x, list<_Tp, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(list<_Tp, _Allocator>& __c, _Predicate __pred)
+{ __c.remove_if(__pred); }
+
+template <class _Tp, class _Allocator, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase(list<_Tp, _Allocator>& __c, const _Up& __v)
+{ _CUDA_VSTD::erase_if(__c, [&](auto& __elem) { return __elem == __v; }); }
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_LIST
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale
new file mode 100644
index 000000000000..5132d7db3bcf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale
@@ -0,0 +1,4353 @@
+// -*- C++ -*-
+//===-------------------------- locale ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_LOCALE
+#define _LIBCUDACXX_LOCALE
+
+/*
+    locale synopsis
+
+namespace std
+{
+
+class locale
+{
+public:
+    // types:
+    class facet;
+    class id;
+
+    typedef int category;
+    static const category // values assigned here are for exposition only
+        none     = 0x000,
+        collate  = 0x010,
+        ctype    = 0x020,
+        monetary = 0x040,
+        numeric  = 0x080,
+        time     = 0x100,
+        messages = 0x200,
+        all = collate | ctype | monetary | numeric | time | messages;
+
+    // construct/copy/destroy:
+    locale() noexcept;
+    locale(const locale& other) noexcept;
+    explicit locale(const char* std_name);
+    explicit locale(const string& std_name);
+    locale(const locale& other, const char* std_name, category);
+    locale(const locale& other, const string& std_name, category);
+    template <class Facet> locale(const locale& other, Facet* f);
+    locale(const locale& other, const locale& one, category);
+
+    ~locale(); // not virtual
+
+    const locale& operator=(const locale& other) noexcept;
+
+    template <class Facet> locale combine(const locale& other) const;
+
+    // locale operations:
+    basic_string<char> name() const;
+    bool operator==(const locale& other) const;
+    bool operator!=(const locale& other) const;
+    template <class charT, class Traits, class Allocator>
+      bool operator()(const basic_string<charT,Traits,Allocator>& s1,
+                      const basic_string<charT,Traits,Allocator>& s2) const;
+
+    // global locale objects:
+    static locale global(const locale&);
+    static const locale& classic();
+};
+
+template <class Facet> const Facet& use_facet(const locale&);
+template <class Facet> bool has_facet(const locale&) noexcept;
+
+// 22.3.3, convenience interfaces:
+template <class charT> bool isspace (charT c, const locale& loc);
+template <class charT> bool isprint (charT c, const locale& loc);
+template <class charT> bool iscntrl (charT c, const locale& loc);
+template <class charT> bool isupper (charT c, const locale& loc);
+template <class charT> bool islower (charT c, const locale& loc);
+template <class charT> bool isalpha (charT c, const locale& loc);
+template <class charT> bool isdigit (charT c, const locale& loc);
+template <class charT> bool ispunct (charT c, const locale& loc);
+template <class charT> bool isxdigit(charT c, const locale& loc);
+template <class charT> bool isalnum (charT c, const locale& loc);
+template <class charT> bool isgraph (charT c, const locale& loc);
+template <class charT> charT toupper(charT c, const locale& loc);
+template <class charT> charT tolower(charT c, const locale& loc);
+
+template<class Codecvt, class Elem = wchar_t,
+         class Wide_alloc = allocator<Elem>,
+         class Byte_alloc = allocator<char>>
+class wstring_convert
+{
+public:
+    typedef basic_string<char, char_traits<char>, Byte_alloc> byte_string;
+    typedef basic_string<Elem, char_traits<Elem>, Wide_alloc> wide_string;
+    typedef typename Codecvt::state_type                      state_type;
+    typedef typename wide_string::traits_type::int_type       int_type;
+
+    explicit wstring_convert(Codecvt* pcvt = new Codecvt);          // explicit in C++14
+    wstring_convert(Codecvt* pcvt, state_type state);
+    explicit wstring_convert(const byte_string& byte_err,           // explicit in C++14
+                    const wide_string& wide_err = wide_string());
+    wstring_convert(const wstring_convert&) = delete;               // C++14
+    wstring_convert & operator=(const wstring_convert &) = delete;  // C++14
+    ~wstring_convert();
+
+    wide_string from_bytes(char byte);
+    wide_string from_bytes(const char* ptr);
+    wide_string from_bytes(const byte_string& str);
+    wide_string from_bytes(const char* first, const char* last);
+
+    byte_string to_bytes(Elem wchar);
+    byte_string to_bytes(const Elem* wptr);
+    byte_string to_bytes(const wide_string& wstr);
+    byte_string to_bytes(const Elem* first, const Elem* last);
+
+    size_t converted() const; // noexcept in C++14
+    state_type state() const;
+};
+
+template <class Codecvt, class Elem = wchar_t, class Tr = char_traits<Elem>>
+class wbuffer_convert
+    : public basic_streambuf<Elem, Tr>
+{
+public:
+    typedef typename Tr::state_type state_type;
+
+    explicit wbuffer_convert(streambuf* bytebuf = 0, Codecvt* pcvt = new Codecvt,
+                    state_type state = state_type());       // explicit in C++14
+    wbuffer_convert(const wbuffer_convert&) = delete;               // C++14
+    wbuffer_convert & operator=(const wbuffer_convert &) = delete;  // C++14
+    ~wbuffer_convert();                                             // C++14
+
+    streambuf* rdbuf() const;
+    streambuf* rdbuf(streambuf* bytebuf);
+
+    state_type state() const;
+};
+
+// 22.4.1 and 22.4.1.3, ctype:
+class ctype_base;
+template <class charT> class ctype;
+template <> class ctype<char>; // specialization
+template <class charT> class ctype_byname;
+template <> class ctype_byname<char>; // specialization
+
+class codecvt_base;
+template <class internT, class externT, class stateT> class codecvt;
+template <class internT, class externT, class stateT> class codecvt_byname;
+
+// 22.4.2 and 22.4.3, numeric:
+template <class charT, class InputIterator> class num_get;
+template <class charT, class OutputIterator> class num_put;
+template <class charT> class numpunct;
+template <class charT> class numpunct_byname;
+
+// 22.4.4, col lation:
+template <class charT> class collate;
+template <class charT> class collate_byname;
+
+// 22.4.5, date and time:
+class time_base;
+template <class charT, class InputIterator> class time_get;
+template <class charT, class InputIterator> class time_get_byname;
+template <class charT, class OutputIterator> class time_put;
+template <class charT, class OutputIterator> class time_put_byname;
+
+// 22.4.6, money:
+class money_base;
+template <class charT, class InputIterator> class money_get;
+template <class charT, class OutputIterator> class money_put;
+template <class charT, bool Intl> class moneypunct;
+template <class charT, bool Intl> class moneypunct_byname;
+
+// 22.4.7, message retrieval:
+class messages_base;
+template <class charT> class messages;
+template <class charT> class messages_byname;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__locale>
+#include <__debug>
+#include <algorithm>
+#include <memory>
+#include <ios>
+#include <streambuf>
+#include <iterator>
+#include <limits>
+#include <version>
+#ifndef __APPLE__
+#include <cstdarg>
+#endif
+#include <cstdlib>
+#include <ctime>
+#include <cstdio>
+#ifdef _LIBCUDACXX_HAS_CATOPEN
+#include <nl_types.h>
+#endif
+
+#ifdef __APPLE__
+#include <Availability.h>
+#endif
+
+#ifdef _LIBCUDACXX_LOCALE__L_EXTENSIONS
+#include <__bsd_locale_defaults.h>
+#else
+#include <__bsd_locale_fallbacks.h>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if defined(__APPLE__) || defined(__FreeBSD__)
+#  define _LIBCUDACXX_GET_C_LOCALE 0
+#elif defined(__CloudABI__) || defined(__NetBSD__)
+#  define _LIBCUDACXX_GET_C_LOCALE LC_C_LOCALE
+#else
+#  define _LIBCUDACXX_GET_C_LOCALE __cloc()
+   // Get the C locale object
+   _LIBCUDACXX_FUNC_VIS locale_t __cloc();
+#define __cloc_defined
+#endif
+
+// __scan_keyword
+// Scans [__b, __e) until a match is found in the basic_strings range
+//  [__kb, __ke) or until it can be shown that there is no match in [__kb, __ke).
+//  __b will be incremented (visibly), consuming CharT until a match is found
+//  or proved to not exist.  A keyword may be "", in which will match anything.
+//  If one keyword is a prefix of another, and the next CharT in the input
+//  might match another keyword, the algorithm will attempt to find the longest
+//  matching keyword.  If the longer matching keyword ends up not matching, then
+//  no keyword match is found.  If no keyword match is found, __ke is returned
+//  and failbit is set in __err.
+//  Else an iterator pointing to the matching keyword is found.  If more than
+//  one keyword matches, an iterator to the first matching keyword is returned.
+//  If on exit __b == __e, eofbit is set in __err.  If __case_sensitive is false,
+//  __ct is used to force to lower case before comparing characters.
+//  Examples:
+//  Keywords:  "a", "abb"
+//  If the input is "a", the first keyword matches and eofbit is set.
+//  If the input is "abc", no match is found and "ab" are consumed.
+template <class _InputIterator, class _ForwardIterator, class _Ctype>
+_LIBCUDACXX_HIDDEN
+_ForwardIterator
+__scan_keyword(_InputIterator& __b, _InputIterator __e,
+               _ForwardIterator __kb, _ForwardIterator __ke,
+               const _Ctype& __ct, ios_base::iostate& __err,
+               bool __case_sensitive = true)
+{
+    typedef typename iterator_traits<_InputIterator>::value_type _CharT;
+    size_t __nkw = static_cast<size_t>(_CUDA_VSTD::distance(__kb, __ke));
+    const unsigned char __doesnt_match = '\0';
+    const unsigned char __might_match = '\1';
+    const unsigned char __does_match = '\2';
+    unsigned char __statbuf[100];
+    unsigned char* __status = __statbuf;
+    unique_ptr<unsigned char, void(*)(void*)> __stat_hold(0, free);
+    if (__nkw > sizeof(__statbuf))
+    {
+        __status = (unsigned char*)malloc(__nkw);
+        if (__status == 0)
+            __throw_bad_alloc();
+        __stat_hold.reset(__status);
+    }
+    size_t __n_might_match = __nkw;  // At this point, any keyword might match
+    size_t __n_does_match = 0;       // but none of them definitely do
+    // Initialize all statuses to __might_match, except for "" keywords are __does_match
+    unsigned char* __st = __status;
+    for (_ForwardIterator __ky = __kb; __ky != __ke; ++__ky, (void) ++__st)
+    {
+        if (!__ky->empty())
+            *__st = __might_match;
+        else
+        {
+            *__st = __does_match;
+            --__n_might_match;
+            ++__n_does_match;
+        }
+    }
+    // While there might be a match, test keywords against the next CharT
+    for (size_t __indx = 0; __b != __e && __n_might_match > 0; ++__indx)
+    {
+        // Peek at the next CharT but don't consume it
+        _CharT __c = *__b;
+        if (!__case_sensitive)
+            __c = __ct.toupper(__c);
+        bool __consume = false;
+        // For each keyword which might match, see if the __indx character is __c
+        // If a match if found, consume __c
+        // If a match is found, and that is the last character in the keyword,
+        //    then that keyword matches.
+        // If the keyword doesn't match this character, then change the keyword
+        //    to doesn't match
+        __st = __status;
+        for (_ForwardIterator __ky = __kb; __ky != __ke; ++__ky, (void) ++__st)
+        {
+            if (*__st == __might_match)
+            {
+                _CharT __kc = (*__ky)[__indx];
+                if (!__case_sensitive)
+                    __kc = __ct.toupper(__kc);
+                if (__c == __kc)
+                {
+                    __consume = true;
+                    if (__ky->size() == __indx+1)
+                    {
+                        *__st = __does_match;
+                        --__n_might_match;
+                        ++__n_does_match;
+                    }
+                }
+                else
+                {
+                    *__st = __doesnt_match;
+                    --__n_might_match;
+                }
+            }
+        }
+        // consume if we matched a character
+        if (__consume)
+        {
+            ++__b;
+            // If we consumed a character and there might be a matched keyword that
+            //   was marked matched on a previous iteration, then such keywords
+            //   which are now marked as not matching.
+            if (__n_might_match + __n_does_match > 1)
+            {
+                __st = __status;
+                for (_ForwardIterator __ky = __kb; __ky != __ke; ++__ky, (void) ++__st)
+                {
+                    if (*__st == __does_match && __ky->size() != __indx+1)
+                    {
+                        *__st = __doesnt_match;
+                        --__n_does_match;
+                    }
+                }
+            }
+        }
+    }
+    // We've exited the loop because we hit eof and/or we have no more "might matches".
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    // Return the first matching result
+    for (__st = __status; __kb != __ke; ++__kb, (void) ++__st)
+        if (*__st == __does_match)
+            break;
+    if (__kb == __ke)
+        __err |= ios_base::failbit;
+    return __kb;
+}
+
+struct _LIBCUDACXX_TYPE_VIS __num_get_base
+{
+    static const int __num_get_buf_sz = 40;
+
+    static int __get_base(ios_base&);
+    static const char __src[33];
+};
+
+_LIBCUDACXX_FUNC_VIS
+void __check_grouping(const string& __grouping, unsigned* __g, unsigned* __g_end,
+                      ios_base::iostate& __err);
+
+template <class _CharT>
+struct __num_get
+    : protected __num_get_base
+{
+    static string __stage2_float_prep(ios_base& __iob, _CharT* __atoms, _CharT& __decimal_point,
+                                      _CharT& __thousands_sep);
+
+    static int __stage2_float_loop(_CharT __ct, bool& __in_units, char& __exp,
+                                   char* __a, char*& __a_end,
+                                   _CharT __decimal_point, _CharT __thousands_sep,
+                                   const string& __grouping, unsigned* __g,
+                                   unsigned*& __g_end, unsigned& __dc, _CharT* __atoms);
+#ifndef _LIBCUDACXX_ABI_OPTIMIZED_LOCALE_NUM_GET
+    static string __stage2_int_prep(ios_base& __iob, _CharT* __atoms, _CharT& __thousands_sep);
+    static int __stage2_int_loop(_CharT __ct, int __base, char* __a, char*& __a_end,
+                  unsigned& __dc, _CharT __thousands_sep, const string& __grouping,
+                  unsigned* __g, unsigned*& __g_end, _CharT* __atoms);
+
+#else
+    static string __stage2_int_prep(ios_base& __iob, _CharT& __thousands_sep)
+    {
+        locale __loc = __iob.getloc();
+        const numpunct<_CharT>& __np = use_facet<numpunct<_CharT> >(__loc);
+        __thousands_sep = __np.thousands_sep();
+        return __np.grouping();
+    }
+
+    const _CharT* __do_widen(ios_base& __iob, _CharT* __atoms) const
+    {
+      return __do_widen_p(__iob, __atoms);
+    }
+
+
+    static int __stage2_int_loop(_CharT __ct, int __base, char* __a, char*& __a_end,
+                  unsigned& __dc, _CharT __thousands_sep, const string& __grouping,
+                  unsigned* __g, unsigned*& __g_end, const _CharT* __atoms);
+private:
+    template<typename T>
+    const T* __do_widen_p(ios_base& __iob, T* __atoms) const
+    {
+      locale __loc = __iob.getloc();
+      use_facet<ctype<T> >(__loc).widen(__src, __src + 26, __atoms);
+      return __atoms;
+    }
+
+    const char* __do_widen_p(ios_base& __iob, char* __atoms) const
+    {
+      (void)__iob;
+      (void)__atoms;
+      return __src;
+    }
+#endif
+};
+
+#ifndef _LIBCUDACXX_ABI_OPTIMIZED_LOCALE_NUM_GET
+template <class _CharT>
+string
+__num_get<_CharT>::__stage2_int_prep(ios_base& __iob, _CharT* __atoms, _CharT& __thousands_sep)
+{
+    locale __loc = __iob.getloc();
+    use_facet<ctype<_CharT> >(__loc).widen(__src, __src + 26, __atoms);
+    const numpunct<_CharT>& __np = use_facet<numpunct<_CharT> >(__loc);
+    __thousands_sep = __np.thousands_sep();
+    return __np.grouping();
+}
+#endif
+
+template <class _CharT>
+string
+__num_get<_CharT>::__stage2_float_prep(ios_base& __iob, _CharT* __atoms, _CharT& __decimal_point,
+                    _CharT& __thousands_sep)
+{
+    locale __loc = __iob.getloc();
+    use_facet<ctype<_CharT> >(__loc).widen(__src, __src + 32, __atoms);
+    const numpunct<_CharT>& __np = use_facet<numpunct<_CharT> >(__loc);
+    __decimal_point = __np.decimal_point();
+    __thousands_sep = __np.thousands_sep();
+    return __np.grouping();
+}
+
+template <class _CharT>
+int
+#ifndef _LIBCUDACXX_ABI_OPTIMIZED_LOCALE_NUM_GET
+__num_get<_CharT>::__stage2_int_loop(_CharT __ct, int __base, char* __a, char*& __a_end,
+                  unsigned& __dc, _CharT __thousands_sep, const string& __grouping,
+                  unsigned* __g, unsigned*& __g_end, _CharT* __atoms)
+#else
+__num_get<_CharT>::__stage2_int_loop(_CharT __ct, int __base, char* __a, char*& __a_end,
+                  unsigned& __dc, _CharT __thousands_sep, const string& __grouping,
+                  unsigned* __g, unsigned*& __g_end, const _CharT* __atoms)
+
+#endif
+{
+    if (__a_end == __a && (__ct == __atoms[24] || __ct == __atoms[25]))
+    {
+        *__a_end++ = __ct == __atoms[24] ? '+' : '-';
+        __dc = 0;
+        return 0;
+    }
+    if (__grouping.size() != 0 && __ct == __thousands_sep)
+    {
+        if (__g_end-__g < __num_get_buf_sz)
+        {
+            *__g_end++ = __dc;
+            __dc = 0;
+        }
+        return 0;
+    }
+    ptrdiff_t __f = find(__atoms, __atoms + 26, __ct) - __atoms;
+    if (__f >= 24)
+        return -1;
+    switch (__base)
+    {
+    case 8:
+    case 10:
+        if (__f >= __base)
+            return -1;
+        break;
+    case 16:
+        if (__f < 22)
+            break;
+        if (__a_end != __a && __a_end - __a <= 2 && __a_end[-1] == '0')
+        {
+            __dc = 0;
+            *__a_end++ = __src[__f];
+            return 0;
+        }
+        return -1;
+    }
+    *__a_end++ = __src[__f];
+    ++__dc;
+    return 0;
+}
+
+template <class _CharT>
+int
+__num_get<_CharT>::__stage2_float_loop(_CharT __ct, bool& __in_units, char& __exp, char* __a, char*& __a_end,
+                    _CharT __decimal_point, _CharT __thousands_sep, const string& __grouping,
+                    unsigned* __g, unsigned*& __g_end, unsigned& __dc, _CharT* __atoms)
+{
+    if (__ct == __decimal_point)
+    {
+        if (!__in_units)
+            return -1;
+        __in_units = false;
+        *__a_end++ = '.';
+        if (__grouping.size() != 0 && __g_end-__g < __num_get_buf_sz)
+            *__g_end++ = __dc;
+        return 0;
+    }
+    if (__ct == __thousands_sep && __grouping.size() != 0)
+    {
+        if (!__in_units)
+            return -1;
+        if (__g_end-__g < __num_get_buf_sz)
+        {
+            *__g_end++ = __dc;
+            __dc = 0;
+        }
+        return 0;
+    }
+    ptrdiff_t __f = find(__atoms, __atoms + 32, __ct) - __atoms;
+    if (__f >= 32)
+        return -1;
+    char __x = __src[__f];
+    if (__x == '-' || __x == '+')
+    {
+        if (__a_end == __a || (__a_end[-1] & 0x5F) == (__exp & 0x7F))
+        {
+            *__a_end++ = __x;
+            return 0;
+        }
+        return -1;
+    }
+    if (__x == 'x' || __x == 'X')
+        __exp = 'P';
+    else if ((__x & 0x5F) == __exp)
+    {
+        __exp |= (char) 0x80;
+        if (__in_units)
+        {
+            __in_units = false;
+            if (__grouping.size() != 0 && __g_end-__g < __num_get_buf_sz)
+                *__g_end++ = __dc;
+        }
+    }
+    *__a_end++ = __x;
+    if (__f >= 22)
+        return 0;
+    ++__dc;
+    return 0;
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(struct _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __num_get<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(struct _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __num_get<wchar_t>)
+
+template <class _CharT, class _InputIterator = istreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS num_get
+    : public locale::facet,
+      private __num_get<_CharT>
+{
+public:
+    typedef _CharT char_type;
+    typedef _InputIterator iter_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit num_get(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, bool& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, long& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, long long& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, unsigned short& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, unsigned int& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, unsigned long& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, unsigned long long& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, float& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, double& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, long double& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, void*& __v) const
+    {
+        return do_get(__b, __e, __iob, __err, __v);
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~num_get() {}
+
+    template <class _Fp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    iter_type __do_get_floating_point
+                            (iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, _Fp& __v) const;
+
+    template <class _Signed>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    iter_type __do_get_signed
+                            (iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, _Signed& __v) const;
+
+    template <class _Unsigned>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    iter_type __do_get_unsigned
+                            (iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, _Unsigned& __v) const;
+
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, bool& __v) const;
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, long& __v) const
+    { return this->__do_get_signed ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, long long& __v) const
+    { return this->__do_get_signed ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, unsigned short& __v) const
+    { return this->__do_get_unsigned ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, unsigned int& __v) const
+    { return this->__do_get_unsigned ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, unsigned long& __v) const
+    { return this->__do_get_unsigned ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, unsigned long long& __v) const
+    { return this->__do_get_unsigned ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, float& __v) const
+    { return this->__do_get_floating_point ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, double& __v) const
+    { return this->__do_get_floating_point ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, long double& __v) const
+    { return this->__do_get_floating_point ( __b, __e, __iob, __err, __v ); }
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, void*& __v) const;
+};
+
+template <class _CharT, class _InputIterator>
+locale::id
+num_get<_CharT, _InputIterator>::id;
+
+template <class _Tp>
+_LIBCUDACXX_HIDDEN _Tp
+__num_get_signed_integral(const char* __a, const char* __a_end,
+                          ios_base::iostate& __err, int __base)
+{
+    if (__a != __a_end)
+    {
+        typename remove_reference<decltype(errno)>::type __save_errno = errno;
+        errno = 0;
+        char *__p2;
+        long long __ll = strtoll_l(__a, &__p2, __base, _LIBCUDACXX_GET_C_LOCALE);
+        typename remove_reference<decltype(errno)>::type __current_errno = errno;
+        if (__current_errno == 0)
+            errno = __save_errno;
+        if (__p2 != __a_end)
+        {
+            __err = ios_base::failbit;
+            return 0;
+        }
+        else if (__current_errno == ERANGE         ||
+                 __ll < numeric_limits<_Tp>::min() ||
+                 numeric_limits<_Tp>::max() < __ll)
+        {
+            __err = ios_base::failbit;
+            if (__ll > 0)
+                return numeric_limits<_Tp>::max();
+            else
+                return numeric_limits<_Tp>::min();
+        }
+        return static_cast<_Tp>(__ll);
+    }
+    __err = ios_base::failbit;
+    return 0;
+}
+
+template <class _Tp>
+_LIBCUDACXX_HIDDEN _Tp
+__num_get_unsigned_integral(const char* __a, const char* __a_end,
+                            ios_base::iostate& __err, int __base)
+{
+    if (__a != __a_end)
+    {
+        const bool __negate = *__a == '-';
+        if (__negate && ++__a == __a_end) {
+          __err = ios_base::failbit;
+          return 0;
+        }
+        typename remove_reference<decltype(errno)>::type __save_errno = errno;
+        errno = 0;
+        char *__p2;
+        unsigned long long __ll = strtoull_l(__a, &__p2, __base, _LIBCUDACXX_GET_C_LOCALE);
+        typename remove_reference<decltype(errno)>::type __current_errno = errno;
+        if (__current_errno == 0)
+            errno = __save_errno;
+        if (__p2 != __a_end)
+        {
+            __err = ios_base::failbit;
+            return 0;
+        }
+        else if (__current_errno == ERANGE || numeric_limits<_Tp>::max() < __ll)
+        {
+            __err = ios_base::failbit;
+            return numeric_limits<_Tp>::max();
+        }
+        _Tp __res = static_cast<_Tp>(__ll);
+        if (__negate) __res = -__res;
+        return __res;
+    }
+    __err = ios_base::failbit;
+    return 0;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __do_strtod(const char* __a, char** __p2);
+
+template <>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+float __do_strtod<float>(const char* __a, char** __p2) {
+    return strtof_l(__a, __p2, _LIBCUDACXX_GET_C_LOCALE);
+}
+
+template <>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+double __do_strtod<double>(const char* __a, char** __p2) {
+    return strtod_l(__a, __p2, _LIBCUDACXX_GET_C_LOCALE);
+}
+
+template <>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+long double __do_strtod<long double>(const char* __a, char** __p2) {
+    return strtold_l(__a, __p2, _LIBCUDACXX_GET_C_LOCALE);
+}
+
+template <class _Tp>
+_LIBCUDACXX_HIDDEN
+_Tp
+__num_get_float(const char* __a, const char* __a_end, ios_base::iostate& __err)
+{
+    if (__a != __a_end)
+    {
+        typename remove_reference<decltype(errno)>::type __save_errno = errno;
+        errno = 0;
+        char *__p2;
+        _Tp __ld = __do_strtod<_Tp>(__a, &__p2);
+        typename remove_reference<decltype(errno)>::type __current_errno = errno;
+        if (__current_errno == 0)
+            errno = __save_errno;
+        if (__p2 != __a_end)
+        {
+            __err = ios_base::failbit;
+            return 0;
+        }
+        else if (__current_errno == ERANGE)
+            __err = ios_base::failbit;
+        return __ld;
+    }
+    __err = ios_base::failbit;
+    return 0;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+num_get<_CharT, _InputIterator>::do_get(iter_type __b, iter_type __e,
+                                        ios_base& __iob,
+                                        ios_base::iostate& __err,
+                                        bool& __v) const
+{
+    if ((__iob.flags() & ios_base::boolalpha) == 0)
+    {
+        long __lv = -1;
+        __b = do_get(__b, __e, __iob, __err, __lv);
+        switch (__lv)
+        {
+        case 0:
+            __v = false;
+            break;
+        case 1:
+            __v = true;
+            break;
+        default:
+            __v = true;
+            __err = ios_base::failbit;
+            break;
+        }
+        return __b;
+    }
+    const ctype<_CharT>& __ct = use_facet<ctype<_CharT> >(__iob.getloc());
+    const numpunct<_CharT>& __np = use_facet<numpunct<_CharT> >(__iob.getloc());
+    typedef typename numpunct<_CharT>::string_type string_type;
+    const string_type __names[2] = {__np.truename(), __np.falsename()};
+    const string_type* __i = __scan_keyword(__b, __e, __names, __names+2,
+                                            __ct, __err);
+    __v = __i == __names;
+    return __b;
+}
+
+// signed
+
+template <class _CharT, class _InputIterator>
+template <class _Signed>
+_InputIterator
+num_get<_CharT, _InputIterator>::__do_get_signed(iter_type __b, iter_type __e,
+                                        ios_base& __iob,
+                                        ios_base::iostate& __err,
+                                        _Signed& __v) const
+{
+    // Stage 1
+    int __base = this->__get_base(__iob);
+    // Stage 2
+    char_type __thousands_sep;
+    const int __atoms_size = 26;
+#ifdef _LIBCUDACXX_ABI_OPTIMIZED_LOCALE_NUM_GET
+    char_type __atoms1[__atoms_size];
+    const char_type *__atoms = this->__do_widen(__iob, __atoms1);
+    string __grouping = this->__stage2_int_prep(__iob, __thousands_sep);
+#else
+    char_type __atoms[__atoms_size];
+    string __grouping = this->__stage2_int_prep(__iob, __atoms, __thousands_sep);
+#endif
+    string __buf;
+    __buf.resize(__buf.capacity());
+    char* __a = &__buf[0];
+    char* __a_end = __a;
+    unsigned __g[__num_get_base::__num_get_buf_sz];
+    unsigned* __g_end = __g;
+    unsigned __dc = 0;
+    for (; __b != __e; ++__b)
+    {
+        if (__a_end == __a + __buf.size())
+        {
+            size_t __tmp = __buf.size();
+            __buf.resize(2*__buf.size());
+            __buf.resize(__buf.capacity());
+            __a = &__buf[0];
+            __a_end = __a + __tmp;
+        }
+        if (this->__stage2_int_loop(*__b, __base, __a, __a_end, __dc,
+                                    __thousands_sep, __grouping, __g, __g_end,
+                                    __atoms))
+            break;
+    }
+    if (__grouping.size() != 0 && __g_end-__g < __num_get_base::__num_get_buf_sz)
+        *__g_end++ = __dc;
+    // Stage 3
+    __v = __num_get_signed_integral<_Signed>(__a, __a_end, __err, __base);
+    // Digit grouping checked
+    __check_grouping(__grouping, __g, __g_end, __err);
+    // EOF checked
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+// unsigned
+
+template <class _CharT, class _InputIterator>
+template <class _Unsigned>
+_InputIterator
+num_get<_CharT, _InputIterator>::__do_get_unsigned(iter_type __b, iter_type __e,
+                                        ios_base& __iob,
+                                        ios_base::iostate& __err,
+                                        _Unsigned& __v) const
+{
+    // Stage 1
+    int __base = this->__get_base(__iob);
+    // Stage 2
+    char_type __thousands_sep;
+    const int __atoms_size = 26;
+#ifdef _LIBCUDACXX_ABI_OPTIMIZED_LOCALE_NUM_GET
+    char_type __atoms1[__atoms_size];
+    const char_type *__atoms = this->__do_widen(__iob, __atoms1);
+    string __grouping = this->__stage2_int_prep(__iob, __thousands_sep);
+#else
+    char_type __atoms[__atoms_size];
+    string __grouping = this->__stage2_int_prep(__iob, __atoms, __thousands_sep);
+#endif
+    string __buf;
+    __buf.resize(__buf.capacity());
+    char* __a = &__buf[0];
+    char* __a_end = __a;
+    unsigned __g[__num_get_base::__num_get_buf_sz];
+    unsigned* __g_end = __g;
+    unsigned __dc = 0;
+    for (; __b != __e; ++__b)
+    {
+        if (__a_end == __a + __buf.size())
+        {
+            size_t __tmp = __buf.size();
+            __buf.resize(2*__buf.size());
+            __buf.resize(__buf.capacity());
+            __a = &__buf[0];
+            __a_end = __a + __tmp;
+        }
+        if (this->__stage2_int_loop(*__b, __base, __a, __a_end, __dc,
+                                    __thousands_sep, __grouping, __g, __g_end,
+                                    __atoms))
+            break;
+    }
+    if (__grouping.size() != 0 && __g_end-__g < __num_get_base::__num_get_buf_sz)
+        *__g_end++ = __dc;
+    // Stage 3
+    __v = __num_get_unsigned_integral<_Unsigned>(__a, __a_end, __err, __base);
+    // Digit grouping checked
+    __check_grouping(__grouping, __g, __g_end, __err);
+    // EOF checked
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+// floating point
+
+template <class _CharT, class _InputIterator>
+template <class _Fp>
+_InputIterator
+num_get<_CharT, _InputIterator>::__do_get_floating_point(iter_type __b, iter_type __e,
+                                        ios_base& __iob,
+                                        ios_base::iostate& __err,
+                                        _Fp& __v) const
+{
+    // Stage 1, nothing to do
+    // Stage 2
+    char_type __atoms[32];
+    char_type __decimal_point;
+    char_type __thousands_sep;
+    string __grouping = this->__stage2_float_prep(__iob, __atoms,
+                                                  __decimal_point,
+                                                  __thousands_sep);
+    string __buf;
+    __buf.resize(__buf.capacity());
+    char* __a = &__buf[0];
+    char* __a_end = __a;
+    unsigned __g[__num_get_base::__num_get_buf_sz];
+    unsigned* __g_end = __g;
+    unsigned __dc = 0;
+    bool __in_units = true;
+    char __exp = 'E';
+    for (; __b != __e; ++__b)
+    {
+        if (__a_end == __a + __buf.size())
+        {
+            size_t __tmp = __buf.size();
+            __buf.resize(2*__buf.size());
+            __buf.resize(__buf.capacity());
+            __a = &__buf[0];
+            __a_end = __a + __tmp;
+        }
+        if (this->__stage2_float_loop(*__b, __in_units, __exp, __a, __a_end,
+                                      __decimal_point, __thousands_sep,
+                                      __grouping, __g, __g_end,
+                                      __dc, __atoms))
+            break;
+    }
+    if (__grouping.size() != 0 && __in_units && __g_end-__g < __num_get_base::__num_get_buf_sz)
+        *__g_end++ = __dc;
+    // Stage 3
+    __v = __num_get_float<_Fp>(__a, __a_end, __err);
+    // Digit grouping checked
+    __check_grouping(__grouping, __g, __g_end, __err);
+    // EOF checked
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+num_get<_CharT, _InputIterator>::do_get(iter_type __b, iter_type __e,
+                                        ios_base& __iob,
+                                        ios_base::iostate& __err,
+                                        void*& __v) const
+{
+    // Stage 1
+    int __base = 16;
+    // Stage 2
+    char_type __atoms[26];
+    char_type __thousands_sep = 0;
+    string __grouping;
+    use_facet<ctype<_CharT> >(__iob.getloc()).widen(__num_get_base::__src,
+                                                    __num_get_base::__src + 26, __atoms);
+    string __buf;
+    __buf.resize(__buf.capacity());
+    char* __a = &__buf[0];
+    char* __a_end = __a;
+    unsigned __g[__num_get_base::__num_get_buf_sz];
+    unsigned* __g_end = __g;
+    unsigned __dc = 0;
+    for (; __b != __e; ++__b)
+    {
+        if (__a_end == __a + __buf.size())
+        {
+            size_t __tmp = __buf.size();
+            __buf.resize(2*__buf.size());
+            __buf.resize(__buf.capacity());
+            __a = &__buf[0];
+            __a_end = __a + __tmp;
+        }
+        if (this->__stage2_int_loop(*__b, __base, __a, __a_end, __dc,
+                                    __thousands_sep, __grouping,
+                                    __g, __g_end, __atoms))
+            break;
+    }
+    // Stage 3
+    __buf.resize(__a_end - __a);
+    if (__libcpp_sscanf_l(__buf.c_str(), _LIBCUDACXX_GET_C_LOCALE, "%p", &__v) != 1)
+        __err = ios_base::failbit;
+    // EOF checked
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS num_get<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS num_get<wchar_t>)
+
+struct _LIBCUDACXX_TYPE_VIS __num_put_base
+{
+protected:
+    static void __format_int(char* __fmt, const char* __len, bool __signd,
+                             ios_base::fmtflags __flags);
+    static bool __format_float(char* __fmt, const char* __len,
+                               ios_base::fmtflags __flags);
+    static char* __identify_padding(char* __nb, char* __ne,
+                                    const ios_base& __iob);
+};
+
+template <class _CharT>
+struct __num_put
+    : protected __num_put_base
+{
+    static void __widen_and_group_int(char* __nb, char* __np, char* __ne,
+                                      _CharT* __ob, _CharT*& __op, _CharT*& __oe,
+                                      const locale& __loc);
+    static void __widen_and_group_float(char* __nb, char* __np, char* __ne,
+                                        _CharT* __ob, _CharT*& __op, _CharT*& __oe,
+                                        const locale& __loc);
+};
+
+template <class _CharT>
+void
+__num_put<_CharT>::__widen_and_group_int(char* __nb, char* __np, char* __ne,
+                                         _CharT* __ob, _CharT*& __op, _CharT*& __oe,
+                                         const locale& __loc)
+{
+    const ctype<_CharT>&    __ct = use_facet<ctype<_CharT> >   (__loc);
+    const numpunct<_CharT>& __npt = use_facet<numpunct<_CharT> >(__loc);
+    string __grouping = __npt.grouping();
+    if (__grouping.empty())
+    {
+        __ct.widen(__nb, __ne, __ob);
+        __oe = __ob + (__ne - __nb);
+    }
+    else
+    {
+        __oe = __ob;
+        char* __nf = __nb;
+        if (*__nf == '-' || *__nf == '+')
+            *__oe++ = __ct.widen(*__nf++);
+        if (__ne - __nf >= 2 && __nf[0] == '0' && (__nf[1] == 'x' ||
+                                                   __nf[1] == 'X'))
+        {
+            *__oe++ = __ct.widen(*__nf++);
+            *__oe++ = __ct.widen(*__nf++);
+        }
+        reverse(__nf, __ne);
+        _CharT __thousands_sep = __npt.thousands_sep();
+        unsigned __dc = 0;
+        unsigned __dg = 0;
+        for (char* __p = __nf; __p < __ne; ++__p)
+        {
+            if (static_cast<unsigned>(__grouping[__dg]) > 0 &&
+                __dc == static_cast<unsigned>(__grouping[__dg]))
+            {
+                *__oe++ = __thousands_sep;
+                __dc = 0;
+                if (__dg < __grouping.size()-1)
+                    ++__dg;
+            }
+            *__oe++ = __ct.widen(*__p);
+            ++__dc;
+        }
+        reverse(__ob + (__nf - __nb), __oe);
+    }
+    if (__np == __ne)
+        __op = __oe;
+    else
+        __op = __ob + (__np - __nb);
+}
+
+template <class _CharT>
+void
+__num_put<_CharT>::__widen_and_group_float(char* __nb, char* __np, char* __ne,
+                                           _CharT* __ob, _CharT*& __op, _CharT*& __oe,
+                                           const locale& __loc)
+{
+    const ctype<_CharT>&    __ct = use_facet<ctype<_CharT> >   (__loc);
+    const numpunct<_CharT>& __npt = use_facet<numpunct<_CharT> >(__loc);
+    string __grouping = __npt.grouping();
+    __oe = __ob;
+    char* __nf = __nb;
+    if (*__nf == '-' || *__nf == '+')
+        *__oe++ = __ct.widen(*__nf++);
+    char* __ns;
+    if (__ne - __nf >= 2 && __nf[0] == '0' && (__nf[1] == 'x' ||
+                                               __nf[1] == 'X'))
+    {
+        *__oe++ = __ct.widen(*__nf++);
+        *__oe++ = __ct.widen(*__nf++);
+        for (__ns = __nf; __ns < __ne; ++__ns)
+            if (!isxdigit_l(*__ns, _LIBCUDACXX_GET_C_LOCALE))
+                break;
+    }
+    else
+    {
+        for (__ns = __nf; __ns < __ne; ++__ns)
+            if (!isdigit_l(*__ns, _LIBCUDACXX_GET_C_LOCALE))
+                break;
+    }
+    if (__grouping.empty())
+    {
+        __ct.widen(__nf, __ns, __oe);
+        __oe += __ns - __nf;
+    }
+    else
+    {
+        reverse(__nf, __ns);
+        _CharT __thousands_sep = __npt.thousands_sep();
+        unsigned __dc = 0;
+        unsigned __dg = 0;
+        for (char* __p = __nf; __p < __ns; ++__p)
+        {
+            if (__grouping[__dg] > 0 && __dc == static_cast<unsigned>(__grouping[__dg]))
+            {
+                *__oe++ = __thousands_sep;
+                __dc = 0;
+                if (__dg < __grouping.size()-1)
+                    ++__dg;
+            }
+            *__oe++ = __ct.widen(*__p);
+            ++__dc;
+        }
+        reverse(__ob + (__nf - __nb), __oe);
+    }
+    for (__nf = __ns; __nf < __ne; ++__nf)
+    {
+        if (*__nf == '.')
+        {
+            *__oe++ = __npt.decimal_point();
+            ++__nf;
+            break;
+        }
+        else
+            *__oe++ = __ct.widen(*__nf);
+    }
+    __ct.widen(__nf, __ne, __oe);
+    __oe += __ne - __nf;
+    if (__np == __ne)
+        __op = __oe;
+    else
+        __op = __ob + (__np - __nb);
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(struct _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __num_put<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(struct _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __num_put<wchar_t>)
+
+template <class _CharT, class _OutputIterator = ostreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS num_put
+    : public locale::facet,
+      private __num_put<_CharT>
+{
+public:
+    typedef _CharT char_type;
+    typedef _OutputIterator iter_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit num_put(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  bool __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  long __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  long long __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  unsigned long __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  unsigned long long __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  double __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  long double __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  const void* __v) const
+    {
+        return do_put(__s, __iob, __fl, __v);
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~num_put() {}
+
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             bool __v) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             long __v) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             long long __v) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             unsigned long) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             unsigned long long) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             double __v) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             long double __v) const;
+    virtual iter_type do_put(iter_type __s, ios_base& __iob, char_type __fl,
+                             const void* __v) const;
+};
+
+template <class _CharT, class _OutputIterator>
+locale::id
+num_put<_CharT, _OutputIterator>::id;
+
+template <class _CharT, class _OutputIterator>
+_LIBCUDACXX_HIDDEN
+_OutputIterator
+__pad_and_output(_OutputIterator __s,
+                 const _CharT* __ob, const _CharT* __op, const _CharT* __oe,
+                 ios_base& __iob, _CharT __fl)
+{
+    streamsize __sz = __oe - __ob;
+    streamsize __ns = __iob.width();
+    if (__ns > __sz)
+        __ns -= __sz;
+    else
+        __ns = 0;
+    for (;__ob < __op; ++__ob, ++__s)
+        *__s = *__ob;
+    for (; __ns; --__ns, ++__s)
+        *__s = __fl;
+    for (; __ob < __oe; ++__ob, ++__s)
+        *__s = *__ob;
+    __iob.width(0);
+    return __s;
+}
+
+template <class _CharT, class _Traits>
+_LIBCUDACXX_HIDDEN
+ostreambuf_iterator<_CharT, _Traits>
+__pad_and_output(ostreambuf_iterator<_CharT, _Traits> __s,
+                 const _CharT* __ob, const _CharT* __op, const _CharT* __oe,
+                 ios_base& __iob, _CharT __fl)
+{
+    if (__s.__sbuf_ == nullptr)
+        return __s;
+    streamsize __sz = __oe - __ob;
+    streamsize __ns = __iob.width();
+    if (__ns > __sz)
+        __ns -= __sz;
+    else
+        __ns = 0;
+    streamsize __np = __op - __ob;
+    if (__np > 0)
+    {
+        if (__s.__sbuf_->sputn(__ob, __np) != __np)
+        {
+            __s.__sbuf_ = nullptr;
+            return __s;
+        }
+    }
+    if (__ns > 0)
+    {
+        basic_string<_CharT, _Traits> __sp(__ns, __fl);
+        if (__s.__sbuf_->sputn(__sp.data(), __ns) != __ns)
+        {
+            __s.__sbuf_ = nullptr;
+            return __s;
+        }
+    }
+    __np = __oe - __op;
+    if (__np > 0)
+    {
+        if (__s.__sbuf_->sputn(__op, __np) != __np)
+        {
+            __s.__sbuf_ = nullptr;
+            return __s;
+        }
+    }
+    __iob.width(0);
+    return __s;
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, bool __v) const
+{
+    if ((__iob.flags() & ios_base::boolalpha) == 0)
+        return do_put(__s, __iob, __fl, (unsigned long)__v);
+    const numpunct<char_type>& __np = use_facet<numpunct<char_type> >(__iob.getloc());
+    typedef typename numpunct<char_type>::string_type string_type;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    string_type __tmp(__v ? __np.truename() : __np.falsename());
+    string_type __nm = _CUDA_VSTD::move(__tmp);
+#else
+    string_type __nm = __v ? __np.truename() : __np.falsename();
+#endif
+    for (typename string_type::iterator __i = __nm.begin(); __i != __nm.end(); ++__i, ++__s)
+        *__s = *__i;
+    return __s;
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, long __v) const
+{
+    // Stage 1 - Get number in narrow char
+    char __fmt[6] = {'%', 0};
+    const char* __len = "l";
+    this->__format_int(__fmt+1, __len, true, __iob.flags());
+    const unsigned __nbuf = (numeric_limits<long>::digits / 3)
+                          + ((numeric_limits<long>::digits % 3) != 0)
+                          + ((__iob.flags() & ios_base::showbase) != 0)
+                          + 2;
+    char __nar[__nbuf];
+    int __nc = __libcpp_snprintf_l(__nar, sizeof(__nar), _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    char* __ne = __nar + __nc;
+    char* __np = this->__identify_padding(__nar, __ne, __iob);
+    // Stage 2 - Widen __nar while adding thousands separators
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    this->__widen_and_group_int(__nar, __np, __ne, __o, __op, __oe, __iob.getloc());
+    // [__o, __oe) contains thousands_sep'd wide number
+    // Stage 3 & 4
+    return __pad_and_output(__s, __o, __op, __oe, __iob, __fl);
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, long long __v) const
+{
+    // Stage 1 - Get number in narrow char
+    char __fmt[8] = {'%', 0};
+    const char* __len = "ll";
+    this->__format_int(__fmt+1, __len, true, __iob.flags());
+    const unsigned __nbuf = (numeric_limits<long long>::digits / 3)
+                          + ((numeric_limits<long long>::digits % 3) != 0)
+                          + ((__iob.flags() & ios_base::showbase) != 0)
+                          + 2;
+    char __nar[__nbuf];
+    int __nc = __libcpp_snprintf_l(__nar, sizeof(__nar), _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    char* __ne = __nar + __nc;
+    char* __np = this->__identify_padding(__nar, __ne, __iob);
+    // Stage 2 - Widen __nar while adding thousands separators
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    this->__widen_and_group_int(__nar, __np, __ne, __o, __op, __oe, __iob.getloc());
+    // [__o, __oe) contains thousands_sep'd wide number
+    // Stage 3 & 4
+    return __pad_and_output(__s, __o, __op, __oe, __iob, __fl);
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, unsigned long __v) const
+{
+    // Stage 1 - Get number in narrow char
+    char __fmt[6] = {'%', 0};
+    const char* __len = "l";
+    this->__format_int(__fmt+1, __len, false, __iob.flags());
+    const unsigned __nbuf = (numeric_limits<unsigned long>::digits / 3)
+                          + ((numeric_limits<unsigned long>::digits % 3) != 0)
+                          + ((__iob.flags() & ios_base::showbase) != 0)
+                          + 1;
+    char __nar[__nbuf];
+    int __nc = __libcpp_snprintf_l(__nar, sizeof(__nar), _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    char* __ne = __nar + __nc;
+    char* __np = this->__identify_padding(__nar, __ne, __iob);
+    // Stage 2 - Widen __nar while adding thousands separators
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    this->__widen_and_group_int(__nar, __np, __ne, __o, __op, __oe, __iob.getloc());
+    // [__o, __oe) contains thousands_sep'd wide number
+    // Stage 3 & 4
+    return __pad_and_output(__s, __o, __op, __oe, __iob, __fl);
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, unsigned long long __v) const
+{
+    // Stage 1 - Get number in narrow char
+    char __fmt[8] = {'%', 0};
+    const char* __len = "ll";
+    this->__format_int(__fmt+1, __len, false, __iob.flags());
+    const unsigned __nbuf = (numeric_limits<unsigned long long>::digits / 3)
+                          + ((numeric_limits<unsigned long long>::digits % 3) != 0)
+                          + ((__iob.flags() & ios_base::showbase) != 0)
+                          + 1;
+    char __nar[__nbuf];
+    int __nc = __libcpp_snprintf_l(__nar, sizeof(__nar), _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    char* __ne = __nar + __nc;
+    char* __np = this->__identify_padding(__nar, __ne, __iob);
+    // Stage 2 - Widen __nar while adding thousands separators
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    this->__widen_and_group_int(__nar, __np, __ne, __o, __op, __oe, __iob.getloc());
+    // [__o, __oe) contains thousands_sep'd wide number
+    // Stage 3 & 4
+    return __pad_and_output(__s, __o, __op, __oe, __iob, __fl);
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, double __v) const
+{
+    // Stage 1 - Get number in narrow char
+    char __fmt[8] = {'%', 0};
+    const char* __len = "";
+    bool __specify_precision = this->__format_float(__fmt+1, __len, __iob.flags());
+    const unsigned __nbuf = 30;
+    char __nar[__nbuf];
+    char* __nb = __nar;
+    int __nc;
+    if (__specify_precision)
+        __nc = __libcpp_snprintf_l(__nb, __nbuf, _LIBCUDACXX_GET_C_LOCALE, __fmt,
+                                   (int)__iob.precision(), __v);
+    else
+        __nc = __libcpp_snprintf_l(__nb, __nbuf, _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    unique_ptr<char, void(*)(void*)> __nbh(0, free);
+    if (__nc > static_cast<int>(__nbuf-1))
+    {
+        if (__specify_precision)
+            __nc = __libcpp_asprintf_l(&__nb, _LIBCUDACXX_GET_C_LOCALE, __fmt, (int)__iob.precision(), __v);
+        else
+            __nc = __libcpp_asprintf_l(&__nb, _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+        if (__nb == 0)
+            __throw_bad_alloc();
+        __nbh.reset(__nb);
+    }
+    char* __ne = __nb + __nc;
+    char* __np = this->__identify_padding(__nb, __ne, __iob);
+    // Stage 2 - Widen __nar while adding thousands separators
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __ob = __o;
+    unique_ptr<char_type, void(*)(void*)> __obh(0, free);
+    if (__nb != __nar)
+    {
+        __ob = (char_type*)malloc(2*static_cast<size_t>(__nc)*sizeof(char_type));
+        if (__ob == 0)
+            __throw_bad_alloc();
+        __obh.reset(__ob);
+    }
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    this->__widen_and_group_float(__nb, __np, __ne, __ob, __op, __oe, __iob.getloc());
+    // [__o, __oe) contains thousands_sep'd wide number
+    // Stage 3 & 4
+    __s = __pad_and_output(__s, __ob, __op, __oe, __iob, __fl);
+    return __s;
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, long double __v) const
+{
+    // Stage 1 - Get number in narrow char
+    char __fmt[8] = {'%', 0};
+    const char* __len = "L";
+    bool __specify_precision = this->__format_float(__fmt+1, __len, __iob.flags());
+    const unsigned __nbuf = 30;
+    char __nar[__nbuf];
+    char* __nb = __nar;
+    int __nc;
+    if (__specify_precision)
+        __nc = __libcpp_snprintf_l(__nb, __nbuf, _LIBCUDACXX_GET_C_LOCALE, __fmt,
+                                   (int)__iob.precision(), __v);
+    else
+        __nc = __libcpp_snprintf_l(__nb, __nbuf, _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    unique_ptr<char, void(*)(void*)> __nbh(0, free);
+    if (__nc > static_cast<int>(__nbuf-1))
+    {
+        if (__specify_precision)
+            __nc = __libcpp_asprintf_l(&__nb, _LIBCUDACXX_GET_C_LOCALE, __fmt, (int)__iob.precision(), __v);
+        else
+            __nc = __libcpp_asprintf_l(&__nb, _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+        if (__nb == 0)
+            __throw_bad_alloc();
+        __nbh.reset(__nb);
+    }
+    char* __ne = __nb + __nc;
+    char* __np = this->__identify_padding(__nb, __ne, __iob);
+    // Stage 2 - Widen __nar while adding thousands separators
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __ob = __o;
+    unique_ptr<char_type, void(*)(void*)> __obh(0, free);
+    if (__nb != __nar)
+    {
+        __ob = (char_type*)malloc(2*static_cast<size_t>(__nc)*sizeof(char_type));
+        if (__ob == 0)
+            __throw_bad_alloc();
+        __obh.reset(__ob);
+    }
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    this->__widen_and_group_float(__nb, __np, __ne, __ob, __op, __oe, __iob.getloc());
+    // [__o, __oe) contains thousands_sep'd wide number
+    // Stage 3 & 4
+    __s = __pad_and_output(__s, __ob, __op, __oe, __iob, __fl);
+    return __s;
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+num_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base& __iob,
+                                         char_type __fl, const void* __v) const
+{
+    // Stage 1 - Get pointer in narrow char
+    char __fmt[6] = "%p";
+    const unsigned __nbuf = 20;
+    char __nar[__nbuf];
+    int __nc = __libcpp_snprintf_l(__nar, sizeof(__nar), _LIBCUDACXX_GET_C_LOCALE, __fmt, __v);
+    char* __ne = __nar + __nc;
+    char* __np = this->__identify_padding(__nar, __ne, __iob);
+    // Stage 2 - Widen __nar
+    char_type __o[2*(__nbuf-1) - 1];
+    char_type* __op;  // pad here
+    char_type* __oe;  // end of output
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    __ct.widen(__nar, __ne, __o);
+    __oe = __o + (__ne - __nar);
+    if (__np == __ne)
+        __op = __oe;
+    else
+        __op = __o + (__np - __nar);
+    // [__o, __oe) contains wide number
+    // Stage 3 & 4
+    return __pad_and_output(__s, __o, __op, __oe, __iob, __fl);
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS num_put<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS num_put<wchar_t>)
+
+template <class _CharT, class _InputIterator>
+_LIBCUDACXX_HIDDEN
+int
+__get_up_to_n_digits(_InputIterator& __b, _InputIterator __e,
+                     ios_base::iostate& __err, const ctype<_CharT>& __ct, int __n)
+{
+    // Precondition:  __n >= 1
+    if (__b == __e)
+    {
+        __err |= ios_base::eofbit | ios_base::failbit;
+        return 0;
+    }
+    // get first digit
+    _CharT __c = *__b;
+    if (!__ct.is(ctype_base::digit, __c))
+    {
+        __err |= ios_base::failbit;
+        return 0;
+    }
+    int __r = __ct.narrow(__c, 0) - '0';
+    for (++__b, (void) --__n; __b != __e && __n > 0; ++__b, (void) --__n)
+    {
+        // get next digit
+        __c = *__b;
+        if (!__ct.is(ctype_base::digit, __c))
+            return __r;
+        __r = __r * 10 + __ct.narrow(__c, 0) - '0';
+    }
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __r;
+}
+
+class _LIBCUDACXX_TYPE_VIS time_base
+{
+public:
+    enum dateorder {no_order, dmy, mdy, ymd, ydm};
+};
+
+template <class _CharT>
+class _LIBCUDACXX_TEMPLATE_VIS __time_get_c_storage
+{
+protected:
+    typedef basic_string<_CharT> string_type;
+
+    virtual const string_type* __weeks() const;
+    virtual const string_type* __months() const;
+    virtual const string_type* __am_pm() const;
+    virtual const string_type& __c() const;
+    virtual const string_type& __r() const;
+    virtual const string_type& __x() const;
+    virtual const string_type& __X() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__time_get_c_storage() {}
+};
+
+template <> _LIBCUDACXX_FUNC_VIS const string* __time_get_c_storage<char>::__weeks() const;
+template <> _LIBCUDACXX_FUNC_VIS const string* __time_get_c_storage<char>::__months() const;
+template <> _LIBCUDACXX_FUNC_VIS const string* __time_get_c_storage<char>::__am_pm() const;
+template <> _LIBCUDACXX_FUNC_VIS const string& __time_get_c_storage<char>::__c() const;
+template <> _LIBCUDACXX_FUNC_VIS const string& __time_get_c_storage<char>::__r() const;
+template <> _LIBCUDACXX_FUNC_VIS const string& __time_get_c_storage<char>::__x() const;
+template <> _LIBCUDACXX_FUNC_VIS const string& __time_get_c_storage<char>::__X() const;
+
+template <> _LIBCUDACXX_FUNC_VIS const wstring* __time_get_c_storage<wchar_t>::__weeks() const;
+template <> _LIBCUDACXX_FUNC_VIS const wstring* __time_get_c_storage<wchar_t>::__months() const;
+template <> _LIBCUDACXX_FUNC_VIS const wstring* __time_get_c_storage<wchar_t>::__am_pm() const;
+template <> _LIBCUDACXX_FUNC_VIS const wstring& __time_get_c_storage<wchar_t>::__c() const;
+template <> _LIBCUDACXX_FUNC_VIS const wstring& __time_get_c_storage<wchar_t>::__r() const;
+template <> _LIBCUDACXX_FUNC_VIS const wstring& __time_get_c_storage<wchar_t>::__x() const;
+template <> _LIBCUDACXX_FUNC_VIS const wstring& __time_get_c_storage<wchar_t>::__X() const;
+
+template <class _CharT, class _InputIterator = istreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS time_get
+    : public locale::facet,
+      public time_base,
+      private __time_get_c_storage<_CharT>
+{
+public:
+    typedef _CharT                  char_type;
+    typedef _InputIterator          iter_type;
+    typedef time_base::dateorder    dateorder;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_get(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    dateorder date_order() const
+    {
+        return this->do_date_order();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get_time(iter_type __b, iter_type __e, ios_base& __iob,
+                       ios_base::iostate& __err, tm* __tm) const
+    {
+        return do_get_time(__b, __e, __iob, __err, __tm);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get_date(iter_type __b, iter_type __e, ios_base& __iob,
+                       ios_base::iostate& __err, tm* __tm) const
+    {
+        return do_get_date(__b, __e, __iob, __err, __tm);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get_weekday(iter_type __b, iter_type __e, ios_base& __iob,
+                          ios_base::iostate& __err, tm* __tm) const
+    {
+        return do_get_weekday(__b, __e, __iob, __err, __tm);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get_monthname(iter_type __b, iter_type __e, ios_base& __iob,
+                            ios_base::iostate& __err, tm* __tm) const
+    {
+        return do_get_monthname(__b, __e, __iob, __err, __tm);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get_year(iter_type __b, iter_type __e, ios_base& __iob,
+                       ios_base::iostate& __err, tm* __tm) const
+    {
+        return do_get_year(__b, __e, __iob, __err, __tm);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, tm *__tm,
+                  char __fmt, char __mod = 0) const
+    {
+        return do_get(__b, __e, __iob, __err, __tm, __fmt, __mod);
+    }
+
+    iter_type get(iter_type __b, iter_type __e, ios_base& __iob,
+                  ios_base::iostate& __err, tm* __tm,
+                  const char_type* __fmtb, const char_type* __fmte) const;
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~time_get() {}
+
+    virtual dateorder do_date_order() const;
+    virtual iter_type do_get_time(iter_type __b, iter_type __e, ios_base& __iob,
+                                  ios_base::iostate& __err, tm* __tm) const;
+    virtual iter_type do_get_date(iter_type __b, iter_type __e, ios_base& __iob,
+                                  ios_base::iostate& __err, tm* __tm) const;
+    virtual iter_type do_get_weekday(iter_type __b, iter_type __e, ios_base& __iob,
+                                     ios_base::iostate& __err, tm* __tm) const;
+    virtual iter_type do_get_monthname(iter_type __b, iter_type __e, ios_base& __iob,
+                                       ios_base::iostate& __err, tm* __tm) const;
+    virtual iter_type do_get_year(iter_type __b, iter_type __e, ios_base& __iob,
+                                  ios_base::iostate& __err, tm* __tm) const;
+    virtual iter_type do_get(iter_type __b, iter_type __e, ios_base& __iob,
+                             ios_base::iostate& __err, tm* __tm,
+                             char __fmt, char __mod) const;
+private:
+    void __get_white_space(iter_type& __b, iter_type __e,
+                           ios_base::iostate& __err, const ctype<char_type>& __ct) const;
+    void __get_percent(iter_type& __b, iter_type __e, ios_base::iostate& __err,
+                       const ctype<char_type>& __ct) const;
+
+    void __get_weekdayname(int& __m,
+                           iter_type& __b, iter_type __e,
+                           ios_base::iostate& __err,
+                           const ctype<char_type>& __ct) const;
+    void __get_monthname(int& __m,
+                         iter_type& __b, iter_type __e,
+                         ios_base::iostate& __err,
+                         const ctype<char_type>& __ct) const;
+    void __get_day(int& __d,
+                   iter_type& __b, iter_type __e,
+                   ios_base::iostate& __err,
+                   const ctype<char_type>& __ct) const;
+    void __get_month(int& __m,
+                     iter_type& __b, iter_type __e,
+                     ios_base::iostate& __err,
+                     const ctype<char_type>& __ct) const;
+    void __get_year(int& __y,
+                   iter_type& __b, iter_type __e,
+                   ios_base::iostate& __err,
+                   const ctype<char_type>& __ct) const;
+    void __get_year4(int& __y,
+                    iter_type& __b, iter_type __e,
+                    ios_base::iostate& __err,
+                    const ctype<char_type>& __ct) const;
+    void __get_hour(int& __d,
+                    iter_type& __b, iter_type __e,
+                    ios_base::iostate& __err,
+                    const ctype<char_type>& __ct) const;
+    void __get_12_hour(int& __h,
+                       iter_type& __b, iter_type __e,
+                       ios_base::iostate& __err,
+                       const ctype<char_type>& __ct) const;
+    void __get_am_pm(int& __h,
+                     iter_type& __b, iter_type __e,
+                     ios_base::iostate& __err,
+                     const ctype<char_type>& __ct) const;
+    void __get_minute(int& __m,
+                      iter_type& __b, iter_type __e,
+                      ios_base::iostate& __err,
+                      const ctype<char_type>& __ct) const;
+    void __get_second(int& __s,
+                      iter_type& __b, iter_type __e,
+                      ios_base::iostate& __err,
+                      const ctype<char_type>& __ct) const;
+    void __get_weekday(int& __w,
+                       iter_type& __b, iter_type __e,
+                       ios_base::iostate& __err,
+                       const ctype<char_type>& __ct) const;
+    void __get_day_year_num(int& __w,
+                            iter_type& __b, iter_type __e,
+                            ios_base::iostate& __err,
+                            const ctype<char_type>& __ct) const;
+};
+
+template <class _CharT, class _InputIterator>
+locale::id
+time_get<_CharT, _InputIterator>::id;
+
+// time_get primitives
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_weekdayname(int& __w,
+                                                    iter_type& __b, iter_type __e,
+                                                    ios_base::iostate& __err,
+                                                    const ctype<char_type>& __ct) const
+{
+    // Note:  ignoring case comes from the POSIX strptime spec
+    const string_type* __wk = this->__weeks();
+    ptrdiff_t __i = __scan_keyword(__b, __e, __wk, __wk+14, __ct, __err, false) - __wk;
+    if (__i < 14)
+        __w = __i % 7;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_monthname(int& __m,
+                                                  iter_type& __b, iter_type __e,
+                                                  ios_base::iostate& __err,
+                                                  const ctype<char_type>& __ct) const
+{
+    // Note:  ignoring case comes from the POSIX strptime spec
+    const string_type* __month = this->__months();
+    ptrdiff_t __i = __scan_keyword(__b, __e, __month, __month+24, __ct, __err, false) - __month;
+    if (__i < 24)
+        __m = __i % 12;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_day(int& __d,
+                                            iter_type& __b, iter_type __e,
+                                            ios_base::iostate& __err,
+                                            const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 2);
+    if (!(__err & ios_base::failbit) && 1 <= __t && __t <= 31)
+        __d = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_month(int& __m,
+                                              iter_type& __b, iter_type __e,
+                                              ios_base::iostate& __err,
+                                              const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 2) - 1;
+    if (!(__err & ios_base::failbit) && __t <= 11)
+        __m = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_year(int& __y,
+                                             iter_type& __b, iter_type __e,
+                                             ios_base::iostate& __err,
+                                             const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 4);
+    if (!(__err & ios_base::failbit))
+    {
+        if (__t < 69)
+            __t += 2000;
+        else if (69 <= __t && __t <= 99)
+            __t += 1900;
+        __y = __t - 1900;
+    }
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_year4(int& __y,
+                                              iter_type& __b, iter_type __e,
+                                              ios_base::iostate& __err,
+                                              const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 4);
+    if (!(__err & ios_base::failbit))
+        __y = __t - 1900;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_hour(int& __h,
+                                             iter_type& __b, iter_type __e,
+                                             ios_base::iostate& __err,
+                                             const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 2);
+    if (!(__err & ios_base::failbit) && __t <= 23)
+        __h = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_12_hour(int& __h,
+                                                iter_type& __b, iter_type __e,
+                                                ios_base::iostate& __err,
+                                                const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 2);
+    if (!(__err & ios_base::failbit) && 1 <= __t && __t <= 12)
+        __h = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_minute(int& __m,
+                                               iter_type& __b, iter_type __e,
+                                               ios_base::iostate& __err,
+                                               const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 2);
+    if (!(__err & ios_base::failbit) && __t <= 59)
+        __m = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_second(int& __s,
+                                               iter_type& __b, iter_type __e,
+                                               ios_base::iostate& __err,
+                                               const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 2);
+    if (!(__err & ios_base::failbit) && __t <= 60)
+        __s = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_weekday(int& __w,
+                                                iter_type& __b, iter_type __e,
+                                                ios_base::iostate& __err,
+                                                const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 1);
+    if (!(__err & ios_base::failbit) && __t <= 6)
+        __w = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_day_year_num(int& __d,
+                                                     iter_type& __b, iter_type __e,
+                                                     ios_base::iostate& __err,
+                                                     const ctype<char_type>& __ct) const
+{
+    int __t = __get_up_to_n_digits(__b, __e, __err, __ct, 3);
+    if (!(__err & ios_base::failbit) && __t <= 365)
+        __d = __t;
+    else
+        __err |= ios_base::failbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_white_space(iter_type& __b, iter_type __e,
+                                                    ios_base::iostate& __err,
+                                                    const ctype<char_type>& __ct) const
+{
+    for (; __b != __e && __ct.is(ctype_base::space, *__b); ++__b)
+        ;
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_am_pm(int& __h,
+                                              iter_type& __b, iter_type __e,
+                                              ios_base::iostate& __err,
+                                              const ctype<char_type>& __ct) const
+{
+    const string_type* __ap = this->__am_pm();
+    if (__ap[0].size() + __ap[1].size() == 0)
+    {
+        __err |= ios_base::failbit;
+        return;
+    }
+    ptrdiff_t __i = __scan_keyword(__b, __e, __ap, __ap+2, __ct, __err, false) - __ap;
+    if (__i == 0 && __h == 12)
+        __h = 0;
+    else if (__i == 1 && __h < 12)
+        __h += 12;
+}
+
+template <class _CharT, class _InputIterator>
+void
+time_get<_CharT, _InputIterator>::__get_percent(iter_type& __b, iter_type __e,
+                                                ios_base::iostate& __err,
+                                                const ctype<char_type>& __ct) const
+{
+    if (__b == __e)
+    {
+        __err |= ios_base::eofbit | ios_base::failbit;
+        return;
+    }
+    if (__ct.narrow(*__b, 0) != '%')
+        __err |= ios_base::failbit;
+    else if(++__b == __e)
+        __err |= ios_base::eofbit;
+}
+
+// time_get end primitives
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::get(iter_type __b, iter_type __e,
+                                      ios_base& __iob,
+                                      ios_base::iostate& __err, tm* __tm,
+                                      const char_type* __fmtb, const char_type* __fmte) const
+{
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    __err = ios_base::goodbit;
+    while (__fmtb != __fmte && __err == ios_base::goodbit)
+    {
+        if (__b == __e)
+        {
+            __err = ios_base::failbit;
+            break;
+        }
+        if (__ct.narrow(*__fmtb, 0) == '%')
+        {
+            if (++__fmtb == __fmte)
+            {
+                __err = ios_base::failbit;
+                break;
+            }
+            char __cmd = __ct.narrow(*__fmtb, 0);
+            char __opt = '\0';
+            if (__cmd == 'E' || __cmd == '0')
+            {
+                if (++__fmtb == __fmte)
+                {
+                    __err = ios_base::failbit;
+                    break;
+                }
+                __opt = __cmd;
+                __cmd = __ct.narrow(*__fmtb, 0);
+            }
+            __b = do_get(__b, __e, __iob, __err, __tm, __cmd, __opt);
+            ++__fmtb;
+        }
+        else if (__ct.is(ctype_base::space, *__fmtb))
+        {
+            for (++__fmtb; __fmtb != __fmte && __ct.is(ctype_base::space, *__fmtb); ++__fmtb)
+                ;
+            for (        ;    __b != __e    && __ct.is(ctype_base::space, *__b);    ++__b)
+                ;
+        }
+        else if (__ct.toupper(*__b) == __ct.toupper(*__fmtb))
+        {
+            ++__b;
+            ++__fmtb;
+        }
+        else
+            __err = ios_base::failbit;
+    }
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+template <class _CharT, class _InputIterator>
+typename time_get<_CharT, _InputIterator>::dateorder
+time_get<_CharT, _InputIterator>::do_date_order() const
+{
+    return mdy;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::do_get_time(iter_type __b, iter_type __e,
+                                              ios_base& __iob,
+                                              ios_base::iostate& __err,
+                                              tm* __tm) const
+{
+    const char_type __fmt[] = {'%', 'H', ':', '%', 'M', ':', '%', 'S'};
+    return get(__b, __e, __iob, __err, __tm, __fmt, __fmt + sizeof(__fmt)/sizeof(__fmt[0]));
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::do_get_date(iter_type __b, iter_type __e,
+                                              ios_base& __iob,
+                                              ios_base::iostate& __err,
+                                              tm* __tm) const
+{
+    const string_type& __fmt = this->__x();
+    return get(__b, __e, __iob, __err, __tm, __fmt.data(), __fmt.data() + __fmt.size());
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::do_get_weekday(iter_type __b, iter_type __e,
+                                                 ios_base& __iob,
+                                                 ios_base::iostate& __err,
+                                                 tm* __tm) const
+{
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    __get_weekdayname(__tm->tm_wday, __b, __e, __err, __ct);
+    return __b;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::do_get_monthname(iter_type __b, iter_type __e,
+                                                   ios_base& __iob,
+                                                   ios_base::iostate& __err,
+                                                   tm* __tm) const
+{
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    __get_monthname(__tm->tm_mon, __b, __e, __err, __ct);
+    return __b;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::do_get_year(iter_type __b, iter_type __e,
+                                              ios_base& __iob,
+                                              ios_base::iostate& __err,
+                                              tm* __tm) const
+{
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    __get_year(__tm->tm_year, __b, __e, __err, __ct);
+    return __b;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+time_get<_CharT, _InputIterator>::do_get(iter_type __b, iter_type __e,
+                                         ios_base& __iob,
+                                         ios_base::iostate& __err, tm* __tm,
+                                         char __fmt, char) const
+{
+    __err = ios_base::goodbit;
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    switch (__fmt)
+    {
+    case 'a':
+    case 'A':
+        __get_weekdayname(__tm->tm_wday, __b, __e, __err, __ct);
+        break;
+    case 'b':
+    case 'B':
+    case 'h':
+        __get_monthname(__tm->tm_mon, __b, __e, __err, __ct);
+        break;
+    case 'c':
+        {
+        const string_type& __fm = this->__c();
+        __b = get(__b, __e, __iob, __err, __tm, __fm.data(), __fm.data() + __fm.size());
+        }
+        break;
+    case 'd':
+    case 'e':
+        __get_day(__tm->tm_mday, __b, __e, __err, __ct);
+        break;
+    case 'D':
+        {
+        const char_type __fm[] = {'%', 'm', '/', '%', 'd', '/', '%', 'y'};
+        __b = get(__b, __e, __iob, __err, __tm, __fm, __fm + sizeof(__fm)/sizeof(__fm[0]));
+        }
+        break;
+    case 'F':
+        {
+        const char_type __fm[] = {'%', 'Y', '-', '%', 'm', '-', '%', 'd'};
+        __b = get(__b, __e, __iob, __err, __tm, __fm, __fm + sizeof(__fm)/sizeof(__fm[0]));
+        }
+        break;
+    case 'H':
+        __get_hour(__tm->tm_hour, __b, __e, __err, __ct);
+        break;
+    case 'I':
+        __get_12_hour(__tm->tm_hour, __b, __e, __err, __ct);
+        break;
+    case 'j':
+        __get_day_year_num(__tm->tm_yday, __b, __e, __err, __ct);
+        break;
+    case 'm':
+        __get_month(__tm->tm_mon, __b, __e, __err, __ct);
+        break;
+    case 'M':
+        __get_minute(__tm->tm_min, __b, __e, __err, __ct);
+        break;
+    case 'n':
+    case 't':
+        __get_white_space(__b, __e, __err, __ct);
+        break;
+    case 'p':
+        __get_am_pm(__tm->tm_hour, __b, __e, __err, __ct);
+        break;
+    case 'r':
+        {
+        const char_type __fm[] = {'%', 'I', ':', '%', 'M', ':', '%', 'S', ' ', '%', 'p'};
+        __b = get(__b, __e, __iob, __err, __tm, __fm, __fm + sizeof(__fm)/sizeof(__fm[0]));
+        }
+        break;
+    case 'R':
+        {
+        const char_type __fm[] = {'%', 'H', ':', '%', 'M'};
+        __b = get(__b, __e, __iob, __err, __tm, __fm, __fm + sizeof(__fm)/sizeof(__fm[0]));
+        }
+        break;
+    case 'S':
+        __get_second(__tm->tm_sec, __b, __e, __err, __ct);
+        break;
+    case 'T':
+        {
+        const char_type __fm[] = {'%', 'H', ':', '%', 'M', ':', '%', 'S'};
+        __b = get(__b, __e, __iob, __err, __tm, __fm, __fm + sizeof(__fm)/sizeof(__fm[0]));
+        }
+        break;
+    case 'w':
+        __get_weekday(__tm->tm_wday, __b, __e, __err, __ct);
+        break;
+    case 'x':
+        return do_get_date(__b, __e, __iob, __err, __tm);
+    case 'X':
+        {
+        const string_type& __fm = this->__X();
+        __b = get(__b, __e, __iob, __err, __tm, __fm.data(), __fm.data() + __fm.size());
+        }
+        break;
+    case 'y':
+        __get_year(__tm->tm_year, __b, __e, __err, __ct);
+        break;
+    case 'Y':
+        __get_year4(__tm->tm_year, __b, __e, __err, __ct);
+        break;
+    case '%':
+        __get_percent(__b, __e, __err, __ct);
+        break;
+    default:
+        __err |= ios_base::failbit;
+    }
+    return __b;
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_get<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_get<wchar_t>)
+
+class _LIBCUDACXX_TYPE_VIS __time_get
+{
+protected:
+    locale_t __loc_;
+
+    __time_get(const char* __nm);
+    __time_get(const string& __nm);
+    ~__time_get();
+};
+
+template <class _CharT>
+class _LIBCUDACXX_TEMPLATE_VIS __time_get_storage
+    : public __time_get
+{
+protected:
+    typedef basic_string<_CharT> string_type;
+
+    string_type __weeks_[14];
+    string_type __months_[24];
+    string_type __am_pm_[2];
+    string_type __c_;
+    string_type __r_;
+    string_type __x_;
+    string_type __X_;
+
+    explicit __time_get_storage(const char* __nm);
+    explicit __time_get_storage(const string& __nm);
+
+    _LIBCUDACXX_INLINE_VISIBILITY ~__time_get_storage() {}
+
+    time_base::dateorder __do_date_order() const;
+
+private:
+    void init(const ctype<_CharT>&);
+    string_type __analyze(char __fmt, const ctype<_CharT>&);
+};
+
+#define _LIBCUDACXX_TIME_GET_STORAGE_EXPLICIT_INSTANTIATION(_CharT) \
+template <> _LIBCUDACXX_FUNC_VIS time_base::dateorder __time_get_storage<_CharT>::__do_date_order() const; \
+template <> _LIBCUDACXX_FUNC_VIS __time_get_storage<_CharT>::__time_get_storage(const char*); \
+template <> _LIBCUDACXX_FUNC_VIS __time_get_storage<_CharT>::__time_get_storage(const string&); \
+template <> _LIBCUDACXX_FUNC_VIS void __time_get_storage<_CharT>::init(const ctype<_CharT>&); \
+template <> _LIBCUDACXX_FUNC_VIS __time_get_storage<_CharT>::string_type __time_get_storage<_CharT>::__analyze(char, const ctype<_CharT>&); \
+extern template _LIBCUDACXX_FUNC_VIS time_base::dateorder __time_get_storage<_CharT>::__do_date_order() const; \
+extern template _LIBCUDACXX_FUNC_VIS __time_get_storage<_CharT>::__time_get_storage(const char*); \
+extern template _LIBCUDACXX_FUNC_VIS __time_get_storage<_CharT>::__time_get_storage(const string&); \
+extern template _LIBCUDACXX_FUNC_VIS void __time_get_storage<_CharT>::init(const ctype<_CharT>&); \
+extern template _LIBCUDACXX_FUNC_VIS __time_get_storage<_CharT>::string_type __time_get_storage<_CharT>::__analyze(char, const ctype<_CharT>&); \
+/**/
+
+_LIBCUDACXX_TIME_GET_STORAGE_EXPLICIT_INSTANTIATION(char)
+_LIBCUDACXX_TIME_GET_STORAGE_EXPLICIT_INSTANTIATION(wchar_t)
+#undef _LIBCUDACXX_TIME_GET_STORAGE_EXPLICIT_INSTANTIATION
+
+template <class _CharT, class _InputIterator = istreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS time_get_byname
+    : public time_get<_CharT, _InputIterator>,
+      private __time_get_storage<_CharT>
+{
+public:
+    typedef time_base::dateorder    dateorder;
+    typedef _InputIterator          iter_type;
+    typedef _CharT                  char_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_get_byname(const char* __nm, size_t __refs = 0)
+        : time_get<_CharT, _InputIterator>(__refs),
+          __time_get_storage<_CharT>(__nm) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_get_byname(const string& __nm, size_t __refs = 0)
+        : time_get<_CharT, _InputIterator>(__refs),
+          __time_get_storage<_CharT>(__nm) {}
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~time_get_byname() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual dateorder do_date_order() const {return this->__do_date_order();}
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type* __weeks() const  {return this->__weeks_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type* __months() const {return this->__months_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type* __am_pm() const  {return this->__am_pm_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type& __c() const      {return this->__c_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type& __r() const      {return this->__r_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type& __x() const      {return this->__x_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual const string_type& __X() const      {return this->__X_;}
+};
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_get_byname<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_get_byname<wchar_t>)
+
+class _LIBCUDACXX_TYPE_VIS __time_put
+{
+    locale_t __loc_;
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY __time_put() : __loc_(_LIBCUDACXX_GET_C_LOCALE) {}
+    __time_put(const char* __nm);
+    __time_put(const string& __nm);
+    ~__time_put();
+    void __do_put(char* __nb, char*& __ne, const tm* __tm,
+                  char __fmt, char __mod) const;
+    void __do_put(wchar_t* __wb, wchar_t*& __we, const tm* __tm,
+                  char __fmt, char __mod) const;
+};
+
+template <class _CharT, class _OutputIterator = ostreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS time_put
+    : public locale::facet,
+      private __time_put
+{
+public:
+    typedef _CharT char_type;
+    typedef _OutputIterator iter_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_put(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl, const tm* __tm,
+                  const char_type* __pb, const char_type* __pe) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, ios_base& __iob, char_type __fl,
+                  const tm* __tm, char __fmt, char __mod = 0) const
+    {
+        return do_put(__s, __iob, __fl, __tm, __fmt, __mod);
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~time_put() {}
+    virtual iter_type do_put(iter_type __s, ios_base&, char_type, const tm* __tm,
+                             char __fmt, char __mod) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_put(const char* __nm, size_t __refs)
+        : locale::facet(__refs),
+          __time_put(__nm) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_put(const string& __nm, size_t __refs)
+        : locale::facet(__refs),
+          __time_put(__nm) {}
+};
+
+template <class _CharT, class _OutputIterator>
+locale::id
+time_put<_CharT, _OutputIterator>::id;
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+time_put<_CharT, _OutputIterator>::put(iter_type __s, ios_base& __iob,
+                                       char_type __fl, const tm* __tm,
+                                       const char_type* __pb,
+                                       const char_type* __pe) const
+{
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__iob.getloc());
+    for (; __pb != __pe; ++__pb)
+    {
+        if (__ct.narrow(*__pb, 0) == '%')
+        {
+            if (++__pb == __pe)
+            {
+                *__s++ = __pb[-1];
+                break;
+            }
+            char __mod = 0;
+            char __fmt = __ct.narrow(*__pb, 0);
+            if (__fmt == 'E' || __fmt == 'O')
+            {
+                if (++__pb == __pe)
+                {
+                    *__s++ = __pb[-2];
+                    *__s++ = __pb[-1];
+                    break;
+                }
+                __mod = __fmt;
+                __fmt = __ct.narrow(*__pb, 0);
+            }
+            __s = do_put(__s, __iob, __fl, __tm, __fmt, __mod);
+        }
+        else
+            *__s++ = *__pb;
+    }
+    return __s;
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+time_put<_CharT, _OutputIterator>::do_put(iter_type __s, ios_base&,
+                                          char_type, const tm* __tm,
+                                          char __fmt, char __mod) const
+{
+    char_type __nar[100];
+    char_type* __nb = __nar;
+    char_type* __ne = __nb + 100;
+    __do_put(__nb, __ne, __tm, __fmt, __mod);
+    return _CUDA_VSTD::copy(__nb, __ne, __s);
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_put<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_put<wchar_t>)
+
+template <class _CharT, class _OutputIterator = ostreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS time_put_byname
+    : public time_put<_CharT, _OutputIterator>
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_put_byname(const char* __nm, size_t __refs = 0)
+        : time_put<_CharT, _OutputIterator>(__nm, __refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit time_put_byname(const string& __nm, size_t __refs = 0)
+        : time_put<_CharT, _OutputIterator>(__nm, __refs) {}
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~time_put_byname() {}
+};
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_put_byname<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS time_put_byname<wchar_t>)
+
+// money_base
+
+class _LIBCUDACXX_TYPE_VIS money_base
+{
+public:
+    enum part {none, space, symbol, sign, value};
+    struct pattern {char field[4];};
+
+    _LIBCUDACXX_INLINE_VISIBILITY money_base() {}
+};
+
+// moneypunct
+
+template <class _CharT, bool _International = false>
+class _LIBCUDACXX_TEMPLATE_VIS moneypunct
+    : public locale::facet,
+      public money_base
+{
+public:
+    typedef _CharT                  char_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit moneypunct(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY char_type   decimal_point() const {return do_decimal_point();}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type   thousands_sep() const {return do_thousands_sep();}
+    _LIBCUDACXX_INLINE_VISIBILITY string      grouping()      const {return do_grouping();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type curr_symbol()   const {return do_curr_symbol();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type positive_sign() const {return do_positive_sign();}
+    _LIBCUDACXX_INLINE_VISIBILITY string_type negative_sign() const {return do_negative_sign();}
+    _LIBCUDACXX_INLINE_VISIBILITY int         frac_digits()   const {return do_frac_digits();}
+    _LIBCUDACXX_INLINE_VISIBILITY pattern     pos_format()    const {return do_pos_format();}
+    _LIBCUDACXX_INLINE_VISIBILITY pattern     neg_format()    const {return do_neg_format();}
+
+    static locale::id id;
+    static const bool intl = _International;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~moneypunct() {}
+
+    virtual char_type   do_decimal_point() const {return numeric_limits<char_type>::max();}
+    virtual char_type   do_thousands_sep() const {return numeric_limits<char_type>::max();}
+    virtual string      do_grouping()      const {return string();}
+    virtual string_type do_curr_symbol()   const {return string_type();}
+    virtual string_type do_positive_sign() const {return string_type();}
+    virtual string_type do_negative_sign() const {return string_type(1, '-');}
+    virtual int         do_frac_digits()   const {return 0;}
+    virtual pattern     do_pos_format()    const
+        {pattern __p = {{symbol, sign, none, value}}; return __p;}
+    virtual pattern     do_neg_format()    const
+        {pattern __p = {{symbol, sign, none, value}}; return __p;}
+};
+
+template <class _CharT, bool _International>
+locale::id
+moneypunct<_CharT, _International>::id;
+
+template <class _CharT, bool _International>
+const bool
+moneypunct<_CharT, _International>::intl;
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct<char, false>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct<char, true>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct<wchar_t, false>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct<wchar_t, true>)
+
+// moneypunct_byname
+
+template <class _CharT, bool _International = false>
+class _LIBCUDACXX_TEMPLATE_VIS moneypunct_byname
+    : public moneypunct<_CharT, _International>
+{
+public:
+    typedef money_base::pattern  pattern;
+    typedef _CharT                  char_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit moneypunct_byname(const char* __nm, size_t __refs = 0)
+        : moneypunct<_CharT, _International>(__refs) {init(__nm);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit moneypunct_byname(const string& __nm, size_t __refs = 0)
+        : moneypunct<_CharT, _International>(__refs) {init(__nm.c_str());}
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~moneypunct_byname() {}
+
+    virtual char_type   do_decimal_point() const {return __decimal_point_;}
+    virtual char_type   do_thousands_sep() const {return __thousands_sep_;}
+    virtual string      do_grouping()      const {return __grouping_;}
+    virtual string_type do_curr_symbol()   const {return __curr_symbol_;}
+    virtual string_type do_positive_sign() const {return __positive_sign_;}
+    virtual string_type do_negative_sign() const {return __negative_sign_;}
+    virtual int         do_frac_digits()   const {return __frac_digits_;}
+    virtual pattern     do_pos_format()    const {return __pos_format_;}
+    virtual pattern     do_neg_format()    const {return __neg_format_;}
+
+private:
+    char_type   __decimal_point_;
+    char_type   __thousands_sep_;
+    string      __grouping_;
+    string_type __curr_symbol_;
+    string_type __positive_sign_;
+    string_type __negative_sign_;
+    int         __frac_digits_;
+    pattern     __pos_format_;
+    pattern     __neg_format_;
+
+    void init(const char*);
+};
+
+template<> _LIBCUDACXX_FUNC_VIS void moneypunct_byname<char, false>::init(const char*);
+template<> _LIBCUDACXX_FUNC_VIS void moneypunct_byname<char, true>::init(const char*);
+template<> _LIBCUDACXX_FUNC_VIS void moneypunct_byname<wchar_t, false>::init(const char*);
+template<> _LIBCUDACXX_FUNC_VIS void moneypunct_byname<wchar_t, true>::init(const char*);
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct_byname<char, false>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct_byname<char, true>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct_byname<wchar_t, false>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS moneypunct_byname<wchar_t, true>)
+
+// money_get
+
+template <class _CharT>
+class __money_get
+{
+protected:
+    typedef _CharT                  char_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __money_get() {}
+
+    static void __gather_info(bool __intl, const locale& __loc,
+                              money_base::pattern& __pat, char_type& __dp,
+                              char_type& __ts, string& __grp,
+                              string_type& __sym, string_type& __psn,
+                              string_type& __nsn, int& __fd);
+};
+
+template <class _CharT>
+void
+__money_get<_CharT>::__gather_info(bool __intl, const locale& __loc,
+                                   money_base::pattern& __pat, char_type& __dp,
+                                   char_type& __ts, string& __grp,
+                                   string_type& __sym, string_type& __psn,
+                                   string_type& __nsn, int& __fd)
+{
+    if (__intl)
+    {
+        const moneypunct<char_type, true>& __mp =
+            use_facet<moneypunct<char_type, true> >(__loc);
+        __pat = __mp.neg_format();
+        __nsn = __mp.negative_sign();
+        __psn = __mp.positive_sign();
+        __dp = __mp.decimal_point();
+        __ts = __mp.thousands_sep();
+        __grp = __mp.grouping();
+        __sym = __mp.curr_symbol();
+        __fd = __mp.frac_digits();
+    }
+    else
+    {
+        const moneypunct<char_type, false>& __mp =
+            use_facet<moneypunct<char_type, false> >(__loc);
+        __pat = __mp.neg_format();
+        __nsn = __mp.negative_sign();
+        __psn = __mp.positive_sign();
+        __dp = __mp.decimal_point();
+        __ts = __mp.thousands_sep();
+        __grp = __mp.grouping();
+        __sym = __mp.curr_symbol();
+        __fd = __mp.frac_digits();
+    }
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __money_get<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __money_get<wchar_t>)
+
+template <class _CharT, class _InputIterator = istreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS money_get
+    : public locale::facet,
+      private __money_get<_CharT>
+{
+public:
+    typedef _CharT                  char_type;
+    typedef _InputIterator          iter_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit money_get(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, bool __intl, ios_base& __iob,
+                  ios_base::iostate& __err, long double& __v) const
+    {
+        return do_get(__b, __e, __intl, __iob, __err, __v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type get(iter_type __b, iter_type __e, bool __intl, ios_base& __iob,
+                  ios_base::iostate& __err, string_type& __v) const
+    {
+        return do_get(__b, __e, __intl, __iob, __err, __v);
+    }
+
+    static locale::id id;
+
+protected:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~money_get() {}
+
+    virtual iter_type do_get(iter_type __b, iter_type __e, bool __intl,
+                             ios_base& __iob, ios_base::iostate& __err,
+                             long double& __v) const;
+    virtual iter_type do_get(iter_type __b, iter_type __e, bool __intl,
+                             ios_base& __iob, ios_base::iostate& __err,
+                             string_type& __v) const;
+
+private:
+    static bool __do_get(iter_type& __b, iter_type __e,
+                         bool __intl, const locale& __loc,
+                         ios_base::fmtflags __flags, ios_base::iostate& __err,
+                         bool& __neg, const ctype<char_type>& __ct,
+                         unique_ptr<char_type, void(*)(void*)>& __wb,
+                         char_type*& __wn, char_type* __we);
+};
+
+template <class _CharT, class _InputIterator>
+locale::id
+money_get<_CharT, _InputIterator>::id;
+
+_LIBCUDACXX_FUNC_VIS void __do_nothing(void*);
+
+template <class _Tp>
+_LIBCUDACXX_HIDDEN
+void
+__double_or_nothing(unique_ptr<_Tp, void(*)(void*)>& __b, _Tp*& __n, _Tp*& __e)
+{
+    bool __owns = __b.get_deleter() != __do_nothing;
+    size_t __cur_cap = static_cast<size_t>(__e-__b.get()) * sizeof(_Tp);
+    size_t __new_cap = __cur_cap < numeric_limits<size_t>::max() / 2 ?
+                       2 * __cur_cap : numeric_limits<size_t>::max();
+    if (__new_cap == 0)
+        __new_cap = sizeof(_Tp);
+    size_t __n_off = static_cast<size_t>(__n - __b.get());
+    _Tp* __t = (_Tp*)realloc(__owns ? __b.get() : 0, __new_cap);
+    if (__t == 0)
+        __throw_bad_alloc();
+    if (__owns)
+        __b.release();
+    __b = unique_ptr<_Tp, void(*)(void*)>(__t, free);
+    __new_cap /= sizeof(_Tp);
+    __n = __b.get() + __n_off;
+    __e = __b.get() + __new_cap;
+}
+
+// true == success
+template <class _CharT, class _InputIterator>
+bool
+money_get<_CharT, _InputIterator>::__do_get(iter_type& __b, iter_type __e,
+                                            bool __intl, const locale& __loc,
+                                            ios_base::fmtflags __flags,
+                                            ios_base::iostate& __err,
+                                            bool& __neg,
+                                            const ctype<char_type>& __ct,
+                                            unique_ptr<char_type, void(*)(void*)>& __wb,
+                                            char_type*& __wn, char_type* __we)
+{
+    const unsigned __bz = 100;
+    unsigned __gbuf[__bz];
+    unique_ptr<unsigned, void(*)(void*)> __gb(__gbuf, __do_nothing);
+    unsigned* __gn = __gb.get();
+    unsigned* __ge = __gn + __bz;
+    money_base::pattern __pat;
+    char_type __dp;
+    char_type __ts;
+    string __grp;
+    string_type __sym;
+    string_type __psn;
+    string_type __nsn;
+    // Capture the spaces read into money_base::{space,none} so they
+    // can be compared to initial spaces in __sym.
+    string_type __spaces;
+    int __fd;
+    __money_get<_CharT>::__gather_info(__intl, __loc, __pat, __dp, __ts, __grp,
+                                       __sym, __psn, __nsn, __fd);
+    const string_type* __trailing_sign = 0;
+    __wn = __wb.get();
+    for (unsigned __p = 0; __p < 4 && __b != __e; ++__p)
+    {
+        switch (__pat.field[__p])
+        {
+        case money_base::space:
+            if (__p != 3)
+            {
+                if (__ct.is(ctype_base::space, *__b))
+                    __spaces.push_back(*__b++);
+                else
+                {
+                    __err |= ios_base::failbit;
+                    return false;
+                }
+            }
+            _LIBCUDACXX_FALLTHROUGH();
+        case money_base::none:
+            if (__p != 3)
+            {
+                while (__b != __e && __ct.is(ctype_base::space, *__b))
+                    __spaces.push_back(*__b++);
+            }
+            break;
+        case money_base::sign:
+            if (__psn.size() + __nsn.size() > 0)
+            {
+                if (__psn.size() == 0 || __nsn.size() == 0)
+                {   // sign is optional
+                    if (__psn.size() > 0)
+                    {   // __nsn.size() == 0
+                        if (*__b == __psn[0])
+                        {
+                            ++__b;
+                            if (__psn.size() > 1)
+                                __trailing_sign = &__psn;
+                        }
+                        else
+                            __neg = true;
+                    }
+                    else if (*__b == __nsn[0])  // __nsn.size() > 0 &&  __psn.size() == 0
+                    {
+                        ++__b;
+                        __neg = true;
+                        if (__nsn.size() > 1)
+                            __trailing_sign = &__nsn;
+                    }
+                }
+                else  // sign is required
+                {
+                    if (*__b == __psn[0])
+                    {
+                        ++__b;
+                        if (__psn.size() > 1)
+                            __trailing_sign = &__psn;
+                    }
+                    else if (*__b == __nsn[0])
+                    {
+                        ++__b;
+                        __neg = true;
+                        if (__nsn.size() > 1)
+                            __trailing_sign = &__nsn;
+                    }
+                    else
+                    {
+                        __err |= ios_base::failbit;
+                        return false;
+                    }
+                }
+            }
+            break;
+        case money_base::symbol:
+            {
+            bool __more_needed = __trailing_sign ||
+                                 (__p < 2)       ||
+                                 (__p == 2 && __pat.field[3] != static_cast<char>(money_base::none));
+            bool __sb = (__flags & ios_base::showbase) != 0;
+            if (__sb || __more_needed)
+            {
+                typename string_type::const_iterator __sym_space_end = __sym.begin();
+                if (__p > 0 && (__pat.field[__p - 1] == money_base::none ||
+                                __pat.field[__p - 1] == money_base::space)) {
+                    // Match spaces we've already read against spaces at
+                    // the beginning of __sym.
+                    while (__sym_space_end != __sym.end() &&
+                           __ct.is(ctype_base::space, *__sym_space_end))
+                        ++__sym_space_end;
+                    const size_t __num_spaces = __sym_space_end - __sym.begin();
+                    if (__num_spaces > __spaces.size() ||
+                        !equal(__spaces.end() - __num_spaces, __spaces.end(),
+                               __sym.begin())) {
+                        // No match. Put __sym_space_end back at the
+                        // beginning of __sym, which will prevent a
+                        // match in the next loop.
+                        __sym_space_end = __sym.begin();
+                    }
+                }
+                typename string_type::const_iterator __sym_curr_char = __sym_space_end;
+                while (__sym_curr_char != __sym.end() && __b != __e &&
+                       *__b == *__sym_curr_char) {
+                    ++__b;
+                    ++__sym_curr_char;
+                }
+                if (__sb && __sym_curr_char != __sym.end())
+                {
+                    __err |= ios_base::failbit;
+                    return false;
+                }
+            }
+            }
+            break;
+        case money_base::value:
+            {
+            unsigned __ng = 0;
+            for (; __b != __e; ++__b)
+            {
+                char_type __c = *__b;
+                if (__ct.is(ctype_base::digit, __c))
+                {
+                    if (__wn == __we)
+                        __double_or_nothing(__wb, __wn, __we);
+                    *__wn++ = __c;
+                    ++__ng;
+                }
+                else if (__grp.size() > 0 && __ng > 0 && __c == __ts)
+                {
+                    if (__gn == __ge)
+                        __double_or_nothing(__gb, __gn, __ge);
+                    *__gn++ = __ng;
+                    __ng = 0;
+                }
+                else
+                    break;
+            }
+            if (__gb.get() != __gn && __ng > 0)
+            {
+                if (__gn == __ge)
+                    __double_or_nothing(__gb, __gn, __ge);
+                *__gn++ = __ng;
+            }
+            if (__fd > 0)
+            {
+                if (__b == __e || *__b != __dp)
+                {
+                    __err |= ios_base::failbit;
+                    return false;
+                }
+                for (++__b; __fd > 0; --__fd, ++__b)
+                {
+                    if (__b == __e || !__ct.is(ctype_base::digit, *__b))
+                    {
+                        __err |= ios_base::failbit;
+                        return false;
+                    }
+                    if (__wn == __we)
+                        __double_or_nothing(__wb, __wn, __we);
+                    *__wn++ = *__b;
+                }
+            }
+            if (__wn == __wb.get())
+            {
+                __err |= ios_base::failbit;
+                return false;
+            }
+            }
+            break;
+        }
+    }
+    if (__trailing_sign)
+    {
+        for (unsigned __i = 1; __i < __trailing_sign->size(); ++__i, ++__b)
+        {
+            if (__b == __e || *__b != (*__trailing_sign)[__i])
+            {
+                __err |= ios_base::failbit;
+                return false;
+            }
+        }
+    }
+    if (__gb.get() != __gn)
+    {
+        ios_base::iostate __et = ios_base::goodbit;
+        __check_grouping(__grp, __gb.get(), __gn, __et);
+        if (__et)
+        {
+            __err |= ios_base::failbit;
+            return false;
+        }
+    }
+    return true;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+money_get<_CharT, _InputIterator>::do_get(iter_type __b, iter_type __e,
+                                          bool __intl, ios_base& __iob,
+                                          ios_base::iostate& __err,
+                                          long double& __v) const
+{
+    const int __bz = 100;
+    char_type __wbuf[__bz];
+    unique_ptr<char_type, void(*)(void*)> __wb(__wbuf, __do_nothing);
+    char_type* __wn;
+    char_type* __we = __wbuf + __bz;
+    locale __loc = __iob.getloc();
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__loc);
+    bool __neg = false;
+    if (__do_get(__b, __e, __intl, __loc, __iob.flags(), __err, __neg, __ct,
+                 __wb, __wn, __we))
+    {
+        const char __src[] = "0123456789";
+        char_type __atoms[sizeof(__src)-1];
+        __ct.widen(__src, __src + (sizeof(__src)-1), __atoms);
+        char __nbuf[__bz];
+        char* __nc = __nbuf;
+        unique_ptr<char, void(*)(void*)> __h(0, free);
+        if (__wn - __wb.get() > __bz-2)
+        {
+            __h.reset((char*)malloc(static_cast<size_t>(__wn - __wb.get() + 2)));
+            if (__h.get() == 0)
+                __throw_bad_alloc();
+            __nc = __h.get();
+        }
+        if (__neg)
+            *__nc++ = '-';
+        for (const char_type* __w = __wb.get(); __w < __wn; ++__w, ++__nc)
+            *__nc = __src[find(__atoms, _CUDA_VSTD::end(__atoms), *__w) - __atoms];
+        *__nc = char();
+        if (sscanf(__nbuf, "%Lf", &__v) != 1)
+            __throw_runtime_error("money_get error");
+    }
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+template <class _CharT, class _InputIterator>
+_InputIterator
+money_get<_CharT, _InputIterator>::do_get(iter_type __b, iter_type __e,
+                                          bool __intl, ios_base& __iob,
+                                          ios_base::iostate& __err,
+                                          string_type& __v) const
+{
+    const int __bz = 100;
+    char_type __wbuf[__bz];
+    unique_ptr<char_type, void(*)(void*)> __wb(__wbuf, __do_nothing);
+    char_type* __wn;
+    char_type* __we = __wbuf + __bz;
+    locale __loc = __iob.getloc();
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__loc);
+    bool __neg = false;
+    if (__do_get(__b, __e, __intl, __loc, __iob.flags(), __err, __neg, __ct,
+                 __wb, __wn, __we))
+    {
+        __v.clear();
+        if (__neg)
+            __v.push_back(__ct.widen('-'));
+        char_type __z = __ct.widen('0');
+        char_type* __w;
+        for (__w = __wb.get(); __w < __wn-1; ++__w)
+            if (*__w != __z)
+                break;
+        __v.append(__w, __wn);
+    }
+    if (__b == __e)
+        __err |= ios_base::eofbit;
+    return __b;
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS money_get<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS money_get<wchar_t>)
+
+// money_put
+
+template <class _CharT>
+class __money_put
+{
+protected:
+    typedef _CharT                  char_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY __money_put() {}
+
+    static void __gather_info(bool __intl, bool __neg, const locale& __loc,
+                              money_base::pattern& __pat, char_type& __dp,
+                              char_type& __ts, string& __grp,
+                              string_type& __sym, string_type& __sn,
+                              int& __fd);
+    static void __format(char_type* __mb, char_type*& __mi, char_type*& __me,
+                         ios_base::fmtflags __flags,
+                         const char_type* __db, const char_type* __de,
+                         const ctype<char_type>& __ct, bool __neg,
+                         const money_base::pattern& __pat, char_type __dp,
+                         char_type __ts, const string& __grp,
+                         const string_type& __sym, const string_type& __sn,
+                         int __fd);
+};
+
+template <class _CharT>
+void
+__money_put<_CharT>::__gather_info(bool __intl, bool __neg, const locale& __loc,
+                                   money_base::pattern& __pat, char_type& __dp,
+                                   char_type& __ts, string& __grp,
+                                   string_type& __sym, string_type& __sn,
+                                   int& __fd)
+{
+    if (__intl)
+    {
+        const moneypunct<char_type, true>& __mp =
+            use_facet<moneypunct<char_type, true> >(__loc);
+        if (__neg)
+        {
+            __pat = __mp.neg_format();
+            __sn = __mp.negative_sign();
+        }
+        else
+        {
+            __pat = __mp.pos_format();
+            __sn = __mp.positive_sign();
+        }
+        __dp = __mp.decimal_point();
+        __ts = __mp.thousands_sep();
+        __grp = __mp.grouping();
+        __sym = __mp.curr_symbol();
+        __fd = __mp.frac_digits();
+    }
+    else
+    {
+        const moneypunct<char_type, false>& __mp =
+            use_facet<moneypunct<char_type, false> >(__loc);
+        if (__neg)
+        {
+            __pat = __mp.neg_format();
+            __sn = __mp.negative_sign();
+        }
+        else
+        {
+            __pat = __mp.pos_format();
+            __sn = __mp.positive_sign();
+        }
+        __dp = __mp.decimal_point();
+        __ts = __mp.thousands_sep();
+        __grp = __mp.grouping();
+        __sym = __mp.curr_symbol();
+        __fd = __mp.frac_digits();
+    }
+}
+
+template <class _CharT>
+void
+__money_put<_CharT>::__format(char_type* __mb, char_type*& __mi, char_type*& __me,
+                              ios_base::fmtflags __flags,
+                              const char_type* __db, const char_type* __de,
+                              const ctype<char_type>& __ct, bool __neg,
+                              const money_base::pattern& __pat, char_type __dp,
+                              char_type __ts, const string& __grp,
+                              const string_type& __sym, const string_type& __sn,
+                              int __fd)
+{
+    __me = __mb;
+    for (unsigned __p = 0; __p < 4; ++__p)
+    {
+        switch (__pat.field[__p])
+        {
+        case money_base::none:
+            __mi = __me;
+            break;
+        case money_base::space:
+            __mi = __me;
+            *__me++ = __ct.widen(' ');
+            break;
+        case money_base::sign:
+            if (!__sn.empty())
+                *__me++ = __sn[0];
+            break;
+        case money_base::symbol:
+            if (!__sym.empty() && (__flags & ios_base::showbase))
+                __me = _CUDA_VSTD::copy(__sym.begin(), __sym.end(), __me);
+            break;
+        case money_base::value:
+            {
+            // remember start of value so we can reverse it
+            char_type* __t = __me;
+            // find beginning of digits
+            if (__neg)
+                ++__db;
+            // find end of digits
+            const char_type* __d;
+            for (__d = __db; __d < __de; ++__d)
+                if (!__ct.is(ctype_base::digit, *__d))
+                    break;
+            // print fractional part
+            if (__fd > 0)
+            {
+                int __f;
+                for (__f = __fd; __d > __db && __f > 0; --__f)
+                    *__me++ = *--__d;
+                char_type __z = __f > 0 ? __ct.widen('0') : char_type();
+                for (; __f > 0; --__f)
+                    *__me++ = __z;
+                *__me++ = __dp;
+            }
+            // print units part
+            if (__d == __db)
+            {
+                *__me++ = __ct.widen('0');
+            }
+            else
+            {
+                unsigned __ng = 0;
+                unsigned __ig = 0;
+                unsigned __gl = __grp.empty() ? numeric_limits<unsigned>::max()
+                                              : static_cast<unsigned>(__grp[__ig]);
+                while (__d != __db)
+                {
+                    if (__ng == __gl)
+                    {
+                        *__me++ = __ts;
+                        __ng = 0;
+                        if (++__ig < __grp.size())
+                            __gl = __grp[__ig] == numeric_limits<char>::max() ?
+                                        numeric_limits<unsigned>::max() :
+                                        static_cast<unsigned>(__grp[__ig]);
+                    }
+                    *__me++ = *--__d;
+                    ++__ng;
+                }
+            }
+            // reverse it
+            reverse(__t, __me);
+            }
+            break;
+        }
+    }
+    // print rest of sign, if any
+    if (__sn.size() > 1)
+        __me = _CUDA_VSTD::copy(__sn.begin()+1, __sn.end(), __me);
+    // set alignment
+    if ((__flags & ios_base::adjustfield) == ios_base::left)
+        __mi = __me;
+    else if ((__flags & ios_base::adjustfield) != ios_base::internal)
+        __mi = __mb;
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __money_put<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __money_put<wchar_t>)
+
+template <class _CharT, class _OutputIterator = ostreambuf_iterator<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS money_put
+    : public locale::facet,
+      private __money_put<_CharT>
+{
+public:
+    typedef _CharT                  char_type;
+    typedef _OutputIterator         iter_type;
+    typedef basic_string<char_type> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit money_put(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, bool __intl, ios_base& __iob, char_type __fl,
+                  long double __units) const
+    {
+        return do_put(__s, __intl, __iob, __fl, __units);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iter_type put(iter_type __s, bool __intl, ios_base& __iob, char_type __fl,
+                  const string_type& __digits) const
+    {
+        return do_put(__s, __intl, __iob, __fl, __digits);
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~money_put() {}
+
+    virtual iter_type do_put(iter_type __s, bool __intl, ios_base& __iob,
+                             char_type __fl, long double __units) const;
+    virtual iter_type do_put(iter_type __s, bool __intl, ios_base& __iob,
+                             char_type __fl, const string_type& __digits) const;
+};
+
+template <class _CharT, class _OutputIterator>
+locale::id
+money_put<_CharT, _OutputIterator>::id;
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+money_put<_CharT, _OutputIterator>::do_put(iter_type __s, bool __intl,
+                                           ios_base& __iob, char_type __fl,
+                                           long double __units) const
+{
+    // convert to char
+    const size_t __bs = 100;
+    char __buf[__bs];
+    char* __bb = __buf;
+    char_type __digits[__bs];
+    char_type* __db = __digits;
+    size_t __n = static_cast<size_t>(snprintf(__bb, __bs, "%.0Lf", __units));
+    unique_ptr<char, void(*)(void*)> __hn(0, free);
+    unique_ptr<char_type, void(*)(void*)> __hd(0, free);
+    // secure memory for digit storage
+    if (__n > __bs-1)
+    {
+        __n = static_cast<size_t>(__libcpp_asprintf_l(&__bb, _LIBCUDACXX_GET_C_LOCALE, "%.0Lf", __units));
+        if (__bb == 0)
+            __throw_bad_alloc();
+        __hn.reset(__bb);
+        __hd.reset((char_type*)malloc(__n * sizeof(char_type)));
+        if (__hd == nullptr)
+            __throw_bad_alloc();
+        __db = __hd.get();
+    }
+    // gather info
+    locale __loc = __iob.getloc();
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__loc);
+    __ct.widen(__bb, __bb + __n, __db);
+    bool __neg = __n > 0 && __bb[0] == '-';
+    money_base::pattern __pat;
+    char_type __dp;
+    char_type __ts;
+    string __grp;
+    string_type __sym;
+    string_type __sn;
+    int __fd;
+    this->__gather_info(__intl, __neg, __loc, __pat, __dp, __ts, __grp, __sym, __sn, __fd);
+    // secure memory for formatting
+    char_type __mbuf[__bs];
+    char_type* __mb = __mbuf;
+    unique_ptr<char_type, void(*)(void*)> __hw(0, free);
+    size_t __exn = static_cast<int>(__n) > __fd ?
+                   (__n - static_cast<size_t>(__fd)) * 2 + __sn.size() +
+                    __sym.size() + static_cast<size_t>(__fd) + 1
+                 : __sn.size() + __sym.size() + static_cast<size_t>(__fd) + 2;
+    if (__exn > __bs)
+    {
+        __hw.reset((char_type*)malloc(__exn * sizeof(char_type)));
+        __mb = __hw.get();
+        if (__mb == 0)
+            __throw_bad_alloc();
+    }
+    // format
+    char_type* __mi;
+    char_type* __me;
+    this->__format(__mb, __mi, __me, __iob.flags(),
+                   __db, __db + __n, __ct,
+                   __neg, __pat, __dp, __ts, __grp, __sym, __sn, __fd);
+    return __pad_and_output(__s, __mb, __mi, __me, __iob, __fl);
+}
+
+template <class _CharT, class _OutputIterator>
+_OutputIterator
+money_put<_CharT, _OutputIterator>::do_put(iter_type __s, bool __intl,
+                                           ios_base& __iob, char_type __fl,
+                                           const string_type& __digits) const
+{
+    // gather info
+    locale __loc = __iob.getloc();
+    const ctype<char_type>& __ct = use_facet<ctype<char_type> >(__loc);
+    bool __neg = __digits.size() > 0 && __digits[0] == __ct.widen('-');
+    money_base::pattern __pat;
+    char_type __dp;
+    char_type __ts;
+    string __grp;
+    string_type __sym;
+    string_type __sn;
+    int __fd;
+    this->__gather_info(__intl, __neg, __loc, __pat, __dp, __ts, __grp, __sym, __sn, __fd);
+    // secure memory for formatting
+    char_type __mbuf[100];
+    char_type* __mb = __mbuf;
+    unique_ptr<char_type, void(*)(void*)> __h(0, free);
+    size_t __exn = static_cast<int>(__digits.size()) > __fd ?
+                   (__digits.size() - static_cast<size_t>(__fd)) * 2 +
+                    __sn.size() + __sym.size() + static_cast<size_t>(__fd) + 1
+                 : __sn.size() + __sym.size() + static_cast<size_t>(__fd) + 2;
+    if (__exn > 100)
+    {
+        __h.reset((char_type*)malloc(__exn * sizeof(char_type)));
+        __mb = __h.get();
+        if (__mb == 0)
+            __throw_bad_alloc();
+    }
+    // format
+    char_type* __mi;
+    char_type* __me;
+    this->__format(__mb, __mi, __me, __iob.flags(),
+                   __digits.data(), __digits.data() + __digits.size(), __ct,
+                   __neg, __pat, __dp, __ts, __grp, __sym, __sn, __fd);
+    return __pad_and_output(__s, __mb, __mi, __me, __iob, __fl);
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS money_put<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS money_put<wchar_t>)
+
+// messages
+
+class _LIBCUDACXX_TYPE_VIS messages_base
+{
+public:
+    typedef ptrdiff_t catalog;
+
+    _LIBCUDACXX_INLINE_VISIBILITY messages_base() {}
+};
+
+template <class _CharT>
+class _LIBCUDACXX_TEMPLATE_VIS messages
+    : public locale::facet,
+      public messages_base
+{
+public:
+    typedef _CharT               char_type;
+    typedef basic_string<_CharT> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit messages(size_t __refs = 0)
+        : locale::facet(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    catalog open(const basic_string<char>& __nm, const locale& __loc) const
+    {
+        return do_open(__nm, __loc);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    string_type get(catalog __c, int __set, int __msgid,
+                    const string_type& __dflt) const
+    {
+        return do_get(__c, __set, __msgid, __dflt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void close(catalog __c) const
+    {
+        do_close(__c);
+    }
+
+    static locale::id id;
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~messages() {}
+
+    virtual catalog do_open(const basic_string<char>&, const locale&) const;
+    virtual string_type do_get(catalog, int __set, int __msgid,
+                               const string_type& __dflt) const;
+    virtual void do_close(catalog) const;
+};
+
+template <class _CharT>
+locale::id
+messages<_CharT>::id;
+
+template <class _CharT>
+typename messages<_CharT>::catalog
+messages<_CharT>::do_open(const basic_string<char>& __nm, const locale&) const
+{
+#ifdef _LIBCUDACXX_HAS_CATOPEN
+    catalog __cat = (catalog)catopen(__nm.c_str(), NL_CAT_LOCALE);
+    if (__cat != -1)
+        __cat = static_cast<catalog>((static_cast<size_t>(__cat) >> 1));
+    return __cat;
+#else // !_LIBCUDACXX_HAS_CATOPEN
+    _LIBCUDACXX_UNUSED_VAR(__nm);
+    return -1;
+#endif // _LIBCUDACXX_HAS_CATOPEN
+}
+
+template <class _CharT>
+typename messages<_CharT>::string_type
+messages<_CharT>::do_get(catalog __c, int __set, int __msgid,
+                         const string_type& __dflt) const
+{
+#ifdef _LIBCUDACXX_HAS_CATOPEN
+    string __ndflt;
+    __narrow_to_utf8<sizeof(char_type)*__CHAR_BIT__>()(back_inserter(__ndflt),
+                                                       __dflt.c_str(),
+                                                       __dflt.c_str() + __dflt.size());
+    if (__c != -1)
+        __c <<= 1;
+    nl_catd __cat = (nl_catd)__c;
+    char* __n = catgets(__cat, __set, __msgid, __ndflt.c_str());
+    string_type __w;
+    __widen_from_utf8<sizeof(char_type)*__CHAR_BIT__>()(back_inserter(__w),
+                                                        __n, __n + strlen(__n));
+    return __w;
+#else // !_LIBCUDACXX_HAS_CATOPEN
+    _LIBCUDACXX_UNUSED_VAR(__c);
+    _LIBCUDACXX_UNUSED_VAR(__set);
+    _LIBCUDACXX_UNUSED_VAR(__msgid);
+    return __dflt;
+#endif // _LIBCUDACXX_HAS_CATOPEN
+}
+
+template <class _CharT>
+void
+messages<_CharT>::do_close(catalog __c) const
+{
+#ifdef _LIBCUDACXX_HAS_CATOPEN
+    if (__c != -1)
+        __c <<= 1;
+    nl_catd __cat = (nl_catd)__c;
+    catclose(__cat);
+#else // !_LIBCUDACXX_HAS_CATOPEN
+    _LIBCUDACXX_UNUSED_VAR(__c);
+#endif // _LIBCUDACXX_HAS_CATOPEN
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS messages<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS messages<wchar_t>)
+
+template <class _CharT>
+class _LIBCUDACXX_TEMPLATE_VIS messages_byname
+    : public messages<_CharT>
+{
+public:
+    typedef messages_base::catalog catalog;
+    typedef basic_string<_CharT> string_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit messages_byname(const char*, size_t __refs = 0)
+        : messages<_CharT>(__refs) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit messages_byname(const string&, size_t __refs = 0)
+        : messages<_CharT>(__refs) {}
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~messages_byname() {}
+};
+
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS messages_byname<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE2(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS messages_byname<wchar_t>)
+
+template<class _Codecvt, class _Elem = wchar_t,
+         class _Wide_alloc = allocator<_Elem>,
+         class _Byte_alloc = allocator<char> >
+class _LIBCUDACXX_TEMPLATE_VIS wstring_convert
+{
+public:
+    typedef basic_string<char, char_traits<char>, _Byte_alloc>   byte_string;
+    typedef basic_string<_Elem, char_traits<_Elem>, _Wide_alloc> wide_string;
+    typedef typename _Codecvt::state_type                        state_type;
+    typedef typename wide_string::traits_type::int_type          int_type;
+
+private:
+    byte_string __byte_err_string_;
+    wide_string __wide_err_string_;
+    _Codecvt* __cvtptr_;
+    state_type __cvtstate_;
+    size_t __cvtcount_;
+
+    wstring_convert(const wstring_convert& __wc);
+    wstring_convert& operator=(const wstring_convert& __wc);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_EXPLICIT_AFTER_CXX11 wstring_convert(_Codecvt* __pcvt = new _Codecvt);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    wstring_convert(_Codecvt* __pcvt, state_type __state);
+    _LIBCUDACXX_EXPLICIT_AFTER_CXX11 wstring_convert(const byte_string& __byte_err,
+                    const wide_string& __wide_err = wide_string());
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    wstring_convert(wstring_convert&& __wc);
+#endif
+    ~wstring_convert();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    wide_string from_bytes(char __byte)
+        {return from_bytes(&__byte, &__byte+1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    wide_string from_bytes(const char* __ptr)
+        {return from_bytes(__ptr, __ptr + char_traits<char>::length(__ptr));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    wide_string from_bytes(const byte_string& __str)
+        {return from_bytes(__str.data(), __str.data() + __str.size());}
+    wide_string from_bytes(const char* __first, const char* __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    byte_string to_bytes(_Elem __wchar)
+        {return to_bytes(&__wchar, &__wchar+1);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    byte_string to_bytes(const _Elem* __wptr)
+        {return to_bytes(__wptr, __wptr + char_traits<_Elem>::length(__wptr));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    byte_string to_bytes(const wide_string& __wstr)
+        {return to_bytes(__wstr.data(), __wstr.data() + __wstr.size());}
+    byte_string to_bytes(const _Elem* __first, const _Elem* __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t converted() const _NOEXCEPT {return __cvtcount_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    state_type state() const {return __cvtstate_;}
+};
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+inline
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::
+    wstring_convert(_Codecvt* __pcvt)
+        : __cvtptr_(__pcvt), __cvtstate_(), __cvtcount_(0)
+{
+}
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+inline
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::
+    wstring_convert(_Codecvt* __pcvt, state_type __state)
+        : __cvtptr_(__pcvt), __cvtstate_(__state), __cvtcount_(0)
+{
+}
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::
+    wstring_convert(const byte_string& __byte_err, const wide_string& __wide_err)
+        : __byte_err_string_(__byte_err), __wide_err_string_(__wide_err),
+          __cvtstate_(), __cvtcount_(0)
+{
+    __cvtptr_ = new _Codecvt;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+inline
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::
+    wstring_convert(wstring_convert&& __wc)
+        : __byte_err_string_(_CUDA_VSTD::move(__wc.__byte_err_string_)),
+          __wide_err_string_(_CUDA_VSTD::move(__wc.__wide_err_string_)),
+          __cvtptr_(__wc.__cvtptr_),
+          __cvtstate_(__wc.__cvtstate_), __cvtcount_(__wc.__cvtcount_)
+{
+    __wc.__cvtptr_ = nullptr;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::~wstring_convert()
+{
+    delete __cvtptr_;
+}
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+typename wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::wide_string
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::
+    from_bytes(const char* __frm, const char* __frm_end)
+{
+    __cvtcount_ = 0;
+    if (__cvtptr_ != nullptr)
+    {
+        wide_string __ws(2*(__frm_end - __frm), _Elem());
+        if (__frm != __frm_end)
+            __ws.resize(__ws.capacity());
+        codecvt_base::result __r = codecvt_base::ok;
+        state_type __st = __cvtstate_;
+        if (__frm != __frm_end)
+        {
+            _Elem* __to = &__ws[0];
+            _Elem* __to_end = __to + __ws.size();
+            const char* __frm_nxt;
+            do
+            {
+                _Elem* __to_nxt;
+                __r = __cvtptr_->in(__st, __frm, __frm_end, __frm_nxt,
+                                          __to, __to_end, __to_nxt);
+                __cvtcount_ += __frm_nxt - __frm;
+                if (__frm_nxt == __frm)
+                {
+                    __r = codecvt_base::error;
+                }
+                else if (__r == codecvt_base::noconv)
+                {
+                    __ws.resize(__to - &__ws[0]);
+                    // This only gets executed if _Elem is char
+                    __ws.append((const _Elem*)__frm, (const _Elem*)__frm_end);
+                    __frm = __frm_nxt;
+                    __r = codecvt_base::ok;
+                }
+                else if (__r == codecvt_base::ok)
+                {
+                    __ws.resize(__to_nxt - &__ws[0]);
+                    __frm = __frm_nxt;
+                }
+                else if (__r == codecvt_base::partial)
+                {
+                    ptrdiff_t __s = __to_nxt - &__ws[0];
+                    __ws.resize(2 * __s);
+                    __to = &__ws[0] + __s;
+                    __to_end = &__ws[0] + __ws.size();
+                    __frm = __frm_nxt;
+                }
+            } while (__r == codecvt_base::partial && __frm_nxt < __frm_end);
+        }
+        if (__r == codecvt_base::ok)
+            return __ws;
+    }
+
+    if (__wide_err_string_.empty())
+        __throw_range_error("wstring_convert: from_bytes error");
+
+    return __wide_err_string_;
+}
+
+template<class _Codecvt, class _Elem, class _Wide_alloc, class _Byte_alloc>
+typename wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::byte_string
+wstring_convert<_Codecvt, _Elem, _Wide_alloc, _Byte_alloc>::
+    to_bytes(const _Elem* __frm, const _Elem* __frm_end)
+{
+    __cvtcount_ = 0;
+    if (__cvtptr_ != nullptr)
+    {
+        byte_string __bs(2*(__frm_end - __frm), char());
+        if (__frm != __frm_end)
+            __bs.resize(__bs.capacity());
+        codecvt_base::result __r = codecvt_base::ok;
+        state_type __st = __cvtstate_;
+        if (__frm != __frm_end)
+        {
+            char* __to = &__bs[0];
+            char* __to_end = __to + __bs.size();
+            const _Elem* __frm_nxt;
+            do
+            {
+                char* __to_nxt;
+                __r = __cvtptr_->out(__st, __frm, __frm_end, __frm_nxt,
+                                           __to, __to_end, __to_nxt);
+                __cvtcount_ += __frm_nxt - __frm;
+                if (__frm_nxt == __frm)
+                {
+                    __r = codecvt_base::error;
+                }
+                else if (__r == codecvt_base::noconv)
+                {
+                    __bs.resize(__to - &__bs[0]);
+                    // This only gets executed if _Elem is char
+                    __bs.append((const char*)__frm, (const char*)__frm_end);
+                    __frm = __frm_nxt;
+                    __r = codecvt_base::ok;
+                }
+                else if (__r == codecvt_base::ok)
+                {
+                    __bs.resize(__to_nxt - &__bs[0]);
+                    __frm = __frm_nxt;
+                }
+                else if (__r == codecvt_base::partial)
+                {
+                    ptrdiff_t __s = __to_nxt - &__bs[0];
+                    __bs.resize(2 * __s);
+                    __to = &__bs[0] + __s;
+                    __to_end = &__bs[0] + __bs.size();
+                    __frm = __frm_nxt;
+                }
+            } while (__r == codecvt_base::partial && __frm_nxt < __frm_end);
+        }
+        if (__r == codecvt_base::ok)
+        {
+            size_t __s = __bs.size();
+            __bs.resize(__bs.capacity());
+            char* __to = &__bs[0] + __s;
+            char* __to_end = __to + __bs.size();
+            do
+            {
+                char* __to_nxt;
+                __r = __cvtptr_->unshift(__st, __to, __to_end, __to_nxt);
+                if (__r == codecvt_base::noconv)
+                {
+                    __bs.resize(__to - &__bs[0]);
+                    __r = codecvt_base::ok;
+                }
+                else if (__r == codecvt_base::ok)
+                {
+                    __bs.resize(__to_nxt - &__bs[0]);
+                }
+                else if (__r == codecvt_base::partial)
+                {
+                    ptrdiff_t __sp = __to_nxt - &__bs[0];
+                    __bs.resize(2 * __sp);
+                    __to = &__bs[0] + __sp;
+                    __to_end = &__bs[0] + __bs.size();
+                }
+            } while (__r == codecvt_base::partial);
+            if (__r == codecvt_base::ok)
+                return __bs;
+        }
+    }
+
+    if (__byte_err_string_.empty())
+        __throw_range_error("wstring_convert: to_bytes error");
+
+    return __byte_err_string_;
+}
+
+template <class _Codecvt, class _Elem = wchar_t, class _Tr = char_traits<_Elem> >
+class _LIBCUDACXX_TEMPLATE_VIS wbuffer_convert
+    : public basic_streambuf<_Elem, _Tr>
+{
+public:
+    // types:
+    typedef _Elem                          char_type;
+    typedef _Tr                            traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef typename _Codecvt::state_type  state_type;
+
+private:
+    char*       __extbuf_;
+    const char* __extbufnext_;
+    const char* __extbufend_;
+    char __extbuf_min_[8];
+    size_t __ebs_;
+    char_type* __intbuf_;
+    size_t __ibs_;
+    streambuf* __bufptr_;
+    _Codecvt* __cv_;
+    state_type __st_;
+    ios_base::openmode __cm_;
+    bool __owns_eb_;
+    bool __owns_ib_;
+    bool __always_noconv_;
+
+    wbuffer_convert(const wbuffer_convert&);
+    wbuffer_convert& operator=(const wbuffer_convert&);
+public:
+    _LIBCUDACXX_EXPLICIT_AFTER_CXX11 wbuffer_convert(streambuf* __bytebuf = 0,
+            _Codecvt* __pcvt = new _Codecvt, state_type __state = state_type());
+    ~wbuffer_convert();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    streambuf* rdbuf() const {return __bufptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    streambuf* rdbuf(streambuf* __bytebuf)
+    {
+        streambuf* __r = __bufptr_;
+        __bufptr_ = __bytebuf;
+        return __r;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    state_type state() const {return __st_;}
+
+protected:
+    virtual int_type underflow();
+    virtual int_type pbackfail(int_type __c = traits_type::eof());
+    virtual int_type overflow (int_type __c = traits_type::eof());
+    virtual basic_streambuf<char_type, traits_type>* setbuf(char_type* __s,
+                                                            streamsize __n);
+    virtual pos_type seekoff(off_type __off, ios_base::seekdir __way,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type __sp,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+    virtual int sync();
+
+private:
+    bool __read_mode();
+    void __write_mode();
+    wbuffer_convert* __close();
+};
+
+template <class _Codecvt, class _Elem, class _Tr>
+wbuffer_convert<_Codecvt, _Elem, _Tr>::
+    wbuffer_convert(streambuf* __bytebuf, _Codecvt* __pcvt, state_type __state)
+    : __extbuf_(0),
+      __extbufnext_(0),
+      __extbufend_(0),
+      __ebs_(0),
+      __intbuf_(0),
+      __ibs_(0),
+      __bufptr_(__bytebuf),
+      __cv_(__pcvt),
+      __st_(__state),
+      __cm_(0),
+      __owns_eb_(false),
+      __owns_ib_(false),
+      __always_noconv_(__cv_ ? __cv_->always_noconv() : false)
+{
+    setbuf(0, 4096);
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+wbuffer_convert<_Codecvt, _Elem, _Tr>::~wbuffer_convert()
+{
+    __close();
+    delete __cv_;
+    if (__owns_eb_)
+        delete [] __extbuf_;
+    if (__owns_ib_)
+        delete [] __intbuf_;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+typename wbuffer_convert<_Codecvt, _Elem, _Tr>::int_type
+wbuffer_convert<_Codecvt, _Elem, _Tr>::underflow()
+{
+    if (__cv_ == 0 || __bufptr_ == 0)
+        return traits_type::eof();
+    bool __initial = __read_mode();
+    char_type __1buf;
+    if (this->gptr() == 0)
+        this->setg(&__1buf, &__1buf+1, &__1buf+1);
+    const size_t __unget_sz = __initial ? 0 : min<size_t>((this->egptr() - this->eback()) / 2, 4);
+    int_type __c = traits_type::eof();
+    if (this->gptr() == this->egptr())
+    {
+        memmove(this->eback(), this->egptr() - __unget_sz, __unget_sz * sizeof(char_type));
+        if (__always_noconv_)
+        {
+            streamsize __nmemb = static_cast<streamsize>(this->egptr() - this->eback() - __unget_sz);
+            __nmemb = __bufptr_->sgetn((char*)this->eback() + __unget_sz, __nmemb);
+            if (__nmemb != 0)
+            {
+                this->setg(this->eback(),
+                           this->eback() + __unget_sz,
+                           this->eback() + __unget_sz + __nmemb);
+                __c = *this->gptr();
+            }
+        }
+        else
+        {
+             _LIBCUDACXX_ASSERT(!(__extbufnext_ == NULL && (__extbufend_ != __extbufnext_)), "underflow moving from NULL" );
+             if (__extbufend_ != __extbufnext_)
+                memmove(__extbuf_, __extbufnext_, __extbufend_ - __extbufnext_);
+            __extbufnext_ = __extbuf_ + (__extbufend_ - __extbufnext_);
+            __extbufend_ = __extbuf_ + (__extbuf_ == __extbuf_min_ ? sizeof(__extbuf_min_) : __ebs_);
+            streamsize __nmemb = _CUDA_VSTD::min(static_cast<streamsize>(this->egptr() - this->eback() - __unget_sz),
+                                 static_cast<streamsize>(__extbufend_ - __extbufnext_));
+            codecvt_base::result __r;
+            // FIXME: Do we ever need to restore the state here?
+            //state_type __svs = __st_;
+            streamsize __nr = __bufptr_->sgetn(const_cast<char*>(__extbufnext_), __nmemb);
+            if (__nr != 0)
+            {
+                __extbufend_ = __extbufnext_ + __nr;
+                char_type*  __inext;
+                __r = __cv_->in(__st_, __extbuf_, __extbufend_, __extbufnext_,
+                                       this->eback() + __unget_sz,
+                                       this->egptr(), __inext);
+                if (__r == codecvt_base::noconv)
+                {
+                    this->setg((char_type*)__extbuf_, (char_type*)__extbuf_,
+                               (char_type*) const_cast<char *>(__extbufend_));
+                    __c = *this->gptr();
+                }
+                else if (__inext != this->eback() + __unget_sz)
+                {
+                    this->setg(this->eback(), this->eback() + __unget_sz, __inext);
+                    __c = *this->gptr();
+                }
+            }
+        }
+    }
+    else
+        __c = *this->gptr();
+    if (this->eback() == &__1buf)
+        this->setg(0, 0, 0);
+    return __c;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+typename wbuffer_convert<_Codecvt, _Elem, _Tr>::int_type
+wbuffer_convert<_Codecvt, _Elem, _Tr>::pbackfail(int_type __c)
+{
+    if (__cv_ != 0 && __bufptr_ != 0 && this->eback() < this->gptr())
+    {
+        if (traits_type::eq_int_type(__c, traits_type::eof()))
+        {
+            this->gbump(-1);
+            return traits_type::not_eof(__c);
+        }
+        if (traits_type::eq(traits_type::to_char_type(__c), this->gptr()[-1]))
+        {
+            this->gbump(-1);
+            *this->gptr() = traits_type::to_char_type(__c);
+            return __c;
+        }
+    }
+    return traits_type::eof();
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+typename wbuffer_convert<_Codecvt, _Elem, _Tr>::int_type
+wbuffer_convert<_Codecvt, _Elem, _Tr>::overflow(int_type __c)
+{
+    if (__cv_ == 0 || __bufptr_ == 0)
+        return traits_type::eof();
+    __write_mode();
+    char_type __1buf;
+    char_type* __pb_save = this->pbase();
+    char_type* __epb_save = this->epptr();
+    if (!traits_type::eq_int_type(__c, traits_type::eof()))
+    {
+        if (this->pptr() == 0)
+            this->setp(&__1buf, &__1buf+1);
+        *this->pptr() = traits_type::to_char_type(__c);
+        this->pbump(1);
+    }
+    if (this->pptr() != this->pbase())
+    {
+        if (__always_noconv_)
+        {
+            streamsize __nmemb = static_cast<streamsize>(this->pptr() - this->pbase());
+            if (__bufptr_->sputn((const char*)this->pbase(), __nmemb) != __nmemb)
+                return traits_type::eof();
+        }
+        else
+        {
+            char* __extbe = __extbuf_;
+            codecvt_base::result __r;
+            do
+            {
+                const char_type* __e;
+                __r = __cv_->out(__st_, this->pbase(), this->pptr(), __e,
+                                        __extbuf_, __extbuf_ + __ebs_, __extbe);
+                if (__e == this->pbase())
+                    return traits_type::eof();
+                if (__r == codecvt_base::noconv)
+                {
+                    streamsize __nmemb = static_cast<size_t>(this->pptr() - this->pbase());
+                    if (__bufptr_->sputn((const char*)this->pbase(), __nmemb) != __nmemb)
+                        return traits_type::eof();
+                }
+                else if (__r == codecvt_base::ok || __r == codecvt_base::partial)
+                {
+                    streamsize __nmemb = static_cast<size_t>(__extbe - __extbuf_);
+                    if (__bufptr_->sputn(__extbuf_, __nmemb) != __nmemb)
+                        return traits_type::eof();
+                    if (__r == codecvt_base::partial)
+                    {
+                        this->setp(const_cast<char_type *>(__e), this->pptr());
+                        this->__pbump(this->epptr() - this->pbase());
+                    }
+                }
+                else
+                    return traits_type::eof();
+            } while (__r == codecvt_base::partial);
+        }
+        this->setp(__pb_save, __epb_save);
+    }
+    return traits_type::not_eof(__c);
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+basic_streambuf<_Elem, _Tr>*
+wbuffer_convert<_Codecvt, _Elem, _Tr>::setbuf(char_type* __s, streamsize __n)
+{
+    this->setg(0, 0, 0);
+    this->setp(0, 0);
+    if (__owns_eb_)
+        delete [] __extbuf_;
+    if (__owns_ib_)
+        delete [] __intbuf_;
+    __ebs_ = __n;
+    if (__ebs_ > sizeof(__extbuf_min_))
+    {
+        if (__always_noconv_ && __s)
+        {
+            __extbuf_ = (char*)__s;
+            __owns_eb_ = false;
+        }
+        else
+        {
+            __extbuf_ = new char[__ebs_];
+            __owns_eb_ = true;
+        }
+    }
+    else
+    {
+        __extbuf_ = __extbuf_min_;
+        __ebs_ = sizeof(__extbuf_min_);
+        __owns_eb_ = false;
+    }
+    if (!__always_noconv_)
+    {
+        __ibs_ = max<streamsize>(__n, sizeof(__extbuf_min_));
+        if (__s && __ibs_ >= sizeof(__extbuf_min_))
+        {
+            __intbuf_ = __s;
+            __owns_ib_ = false;
+        }
+        else
+        {
+            __intbuf_ = new char_type[__ibs_];
+            __owns_ib_ = true;
+        }
+    }
+    else
+    {
+        __ibs_ = 0;
+        __intbuf_ = 0;
+        __owns_ib_ = false;
+    }
+    return this;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+typename wbuffer_convert<_Codecvt, _Elem, _Tr>::pos_type
+wbuffer_convert<_Codecvt, _Elem, _Tr>::seekoff(off_type __off, ios_base::seekdir __way,
+                                        ios_base::openmode __om)
+{
+    int __width = __cv_->encoding();
+    if (__cv_ == 0 || __bufptr_ == 0 || (__width <= 0 && __off != 0) || sync())
+        return pos_type(off_type(-1));
+    // __width > 0 || __off == 0, now check __way
+    if (__way != ios_base::beg && __way != ios_base::cur && __way != ios_base::end)
+        return pos_type(off_type(-1));
+    pos_type __r = __bufptr_->pubseekoff(__width * __off, __way, __om);
+    __r.state(__st_);
+    return __r;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+typename wbuffer_convert<_Codecvt, _Elem, _Tr>::pos_type
+wbuffer_convert<_Codecvt, _Elem, _Tr>::seekpos(pos_type __sp, ios_base::openmode __wch)
+{
+    if (__cv_ == 0 || __bufptr_ == 0 || sync())
+        return pos_type(off_type(-1));
+    if (__bufptr_->pubseekpos(__sp, __wch) == pos_type(off_type(-1)))
+        return pos_type(off_type(-1));
+    return __sp;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+int
+wbuffer_convert<_Codecvt, _Elem, _Tr>::sync()
+{
+    if (__cv_ == 0 || __bufptr_ == 0)
+        return 0;
+    if (__cm_ & ios_base::out)
+    {
+        if (this->pptr() != this->pbase())
+            if (overflow() == traits_type::eof())
+                return -1;
+        codecvt_base::result __r;
+        do
+        {
+            char* __extbe;
+            __r = __cv_->unshift(__st_, __extbuf_, __extbuf_ + __ebs_, __extbe);
+            streamsize __nmemb = static_cast<streamsize>(__extbe - __extbuf_);
+            if (__bufptr_->sputn(__extbuf_, __nmemb) != __nmemb)
+                return -1;
+        } while (__r == codecvt_base::partial);
+        if (__r == codecvt_base::error)
+            return -1;
+        if (__bufptr_->pubsync())
+            return -1;
+    }
+    else if (__cm_ & ios_base::in)
+    {
+        off_type __c;
+        if (__always_noconv_)
+            __c = this->egptr() - this->gptr();
+        else
+        {
+            int __width = __cv_->encoding();
+            __c = __extbufend_ - __extbufnext_;
+            if (__width > 0)
+                __c += __width * (this->egptr() - this->gptr());
+            else
+            {
+                if (this->gptr() != this->egptr())
+                {
+                    reverse(this->gptr(), this->egptr());
+                    codecvt_base::result __r;
+                    const char_type* __e = this->gptr();
+                    char* __extbe;
+                    do
+                    {
+                        __r = __cv_->out(__st_, __e, this->egptr(), __e,
+                                         __extbuf_, __extbuf_ + __ebs_, __extbe);
+                        switch (__r)
+                        {
+                        case codecvt_base::noconv:
+                            __c += this->egptr() - this->gptr();
+                            break;
+                        case codecvt_base::ok:
+                        case codecvt_base::partial:
+                            __c += __extbe - __extbuf_;
+                            break;
+                        default:
+                            return -1;
+                        }
+                    } while (__r == codecvt_base::partial);
+                }
+            }
+        }
+        if (__bufptr_->pubseekoff(-__c, ios_base::cur, __cm_) == pos_type(off_type(-1)))
+            return -1;
+        this->setg(0, 0, 0);
+        __cm_ = 0;
+    }
+    return 0;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+bool
+wbuffer_convert<_Codecvt, _Elem, _Tr>::__read_mode()
+{
+    if (!(__cm_ & ios_base::in))
+    {
+        this->setp(0, 0);
+        if (__always_noconv_)
+            this->setg((char_type*)__extbuf_,
+                       (char_type*)__extbuf_ + __ebs_,
+                       (char_type*)__extbuf_ + __ebs_);
+        else
+            this->setg(__intbuf_, __intbuf_ + __ibs_, __intbuf_ + __ibs_);
+        __cm_ = ios_base::in;
+        return true;
+    }
+    return false;
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+void
+wbuffer_convert<_Codecvt, _Elem, _Tr>::__write_mode()
+{
+    if (!(__cm_ & ios_base::out))
+    {
+        this->setg(0, 0, 0);
+        if (__ebs_ > sizeof(__extbuf_min_))
+        {
+            if (__always_noconv_)
+                this->setp((char_type*)__extbuf_,
+                           (char_type*)__extbuf_ + (__ebs_ - 1));
+            else
+                this->setp(__intbuf_, __intbuf_ + (__ibs_ - 1));
+        }
+        else
+            this->setp(0, 0);
+        __cm_ = ios_base::out;
+    }
+}
+
+template <class _Codecvt, class _Elem, class _Tr>
+wbuffer_convert<_Codecvt, _Elem, _Tr>*
+wbuffer_convert<_Codecvt, _Elem, _Tr>::__close()
+{
+    wbuffer_convert* __rt = 0;
+    if (__cv_ != 0 && __bufptr_ != 0)
+    {
+        __rt = this;
+        if ((__cm_ & ios_base::out) && sync())
+            __rt = 0;
+    }
+    return __rt;
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_LOCALE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h
new file mode 100644
index 000000000000..d156f3186087
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/locale.h
@@ -0,0 +1,44 @@
+// -*- C++ -*-
+//===---------------------------- locale.h --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_LOCALE_H
+#define _LIBCUDACXX_LOCALE_H
+
+/*
+    locale.h synopsis
+
+Macros:
+
+    LC_ALL
+    LC_COLLATE
+    LC_CTYPE
+    LC_MONETARY
+    LC_NUMERIC
+    LC_TIME
+
+Types:
+
+    lconv
+
+Functions:
+
+   setlocale
+   localeconv
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <locale.h>
+
+#endif  // _LIBCUDACXX_LOCALE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/map b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/map
new file mode 100644
index 000000000000..a3a17599b6db
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/map
@@ -0,0 +1,2246 @@
+// -*- C++ -*-
+//===----------------------------- map ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_MAP
+#define _LIBCUDACXX_MAP
+
+/*
+
+    map synopsis
+
+namespace std
+{
+
+template <class Key, class T, class Compare = less<Key>,
+          class Allocator = allocator<pair<const Key, T>>>
+class map
+{
+public:
+    // types:
+    typedef Key                                      key_type;
+    typedef T                                        mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+    typedef Compare                                  key_compare;
+    typedef Allocator                                allocator_type;
+    typedef typename allocator_type::reference       reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef typename allocator_type::pointer         pointer;
+    typedef typename allocator_type::const_pointer   const_pointer;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+    typedef unspecified                              node_type;              // C++17
+    typedef INSERT_RETURN_TYPE<iterator, node_type>  insert_return_type;     // C++17
+
+    class value_compare
+        : public binary_function<value_type, value_type, bool>
+    {
+        friend class map;
+    protected:
+        key_compare comp;
+
+        value_compare(key_compare c);
+    public:
+        bool operator()(const value_type& x, const value_type& y) const;
+    };
+
+    // construct/copy/destroy:
+    map()
+        noexcept(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value);
+    explicit map(const key_compare& comp);
+    map(const key_compare& comp, const allocator_type& a);
+    template <class InputIterator>
+        map(InputIterator first, InputIterator last,
+            const key_compare& comp = key_compare());
+    template <class InputIterator>
+        map(InputIterator first, InputIterator last,
+            const key_compare& comp, const allocator_type& a);
+    map(const map& m);
+    map(map&& m)
+        noexcept(
+            is_nothrow_move_constructible<allocator_type>::value &&
+            is_nothrow_move_constructible<key_compare>::value);
+    explicit map(const allocator_type& a);
+    map(const map& m, const allocator_type& a);
+    map(map&& m, const allocator_type& a);
+    map(initializer_list<value_type> il, const key_compare& comp = key_compare());
+    map(initializer_list<value_type> il, const key_compare& comp, const allocator_type& a);
+    template <class InputIterator>
+        map(InputIterator first, InputIterator last, const allocator_type& a)
+            : map(first, last, Compare(), a) {}  // C++14
+    map(initializer_list<value_type> il, const allocator_type& a)
+        : map(il, Compare(), a) {}  // C++14
+   ~map();
+
+    map& operator=(const map& m);
+    map& operator=(map&& m)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<key_compare>::value);
+    map& operator=(initializer_list<value_type> il);
+
+    // iterators:
+          iterator begin() noexcept;
+    const_iterator begin() const noexcept;
+          iterator end() noexcept;
+    const_iterator end()   const noexcept;
+
+          reverse_iterator rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+          reverse_iterator rend() noexcept;
+    const_reverse_iterator rend()   const noexcept;
+
+    const_iterator         cbegin()  const noexcept;
+    const_iterator         cend()    const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend()   const noexcept;
+
+    // capacity:
+    bool      empty()    const noexcept;
+    size_type size()     const noexcept;
+    size_type max_size() const noexcept;
+
+    // element access:
+    mapped_type& operator[](const key_type& k);
+    mapped_type& operator[](key_type&& k);
+
+          mapped_type& at(const key_type& k);
+    const mapped_type& at(const key_type& k) const;
+
+    // modifiers:
+    template <class... Args>
+        pair<iterator, bool> emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    pair<iterator, bool> insert(const value_type& v);
+    pair<iterator, bool> insert(      value_type&& v);                                // C++17
+    template <class P>
+        pair<iterator, bool> insert(P&& p);
+    iterator insert(const_iterator position, const value_type& v);
+    iterator insert(const_iterator position,       value_type&& v);                   // C++17
+    template <class P>
+        iterator insert(const_iterator position, P&& p);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type> il);
+
+    node_type extract(const_iterator position);                                       // C++17
+    node_type extract(const key_type& x);                                             // C++17
+    insert_return_type insert(node_type&& nh);                                        // C++17
+    iterator insert(const_iterator hint, node_type&& nh);                             // C++17
+
+    template <class... Args>
+        pair<iterator, bool> try_emplace(const key_type& k, Args&&... args);          // C++17
+    template <class... Args>
+        pair<iterator, bool> try_emplace(key_type&& k, Args&&... args);               // C++17
+    template <class... Args>
+        iterator try_emplace(const_iterator hint, const key_type& k, Args&&... args); // C++17
+    template <class... Args>
+        iterator try_emplace(const_iterator hint, key_type&& k, Args&&... args);      // C++17
+    template <class M>
+        pair<iterator, bool> insert_or_assign(const key_type& k, M&& obj);            // C++17
+    template <class M>
+        pair<iterator, bool> insert_or_assign(key_type&& k, M&& obj);                 // C++17
+    template <class M>
+        iterator insert_or_assign(const_iterator hint, const key_type& k, M&& obj);   // C++17
+    template <class M>
+        iterator insert_or_assign(const_iterator hint, key_type&& k, M&& obj);        // C++17
+
+    iterator  erase(const_iterator position);
+    iterator  erase(iterator position); // C++14
+    size_type erase(const key_type& k);
+    iterator  erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class C2>
+      void merge(map<Key, T, C2, Allocator>& source);         // C++17
+    template<class C2>
+      void merge(map<Key, T, C2, Allocator>&& source);        // C++17
+    template<class C2>
+      void merge(multimap<Key, T, C2, Allocator>& source);    // C++17
+    template<class C2>
+      void merge(multimap<Key, T, C2, Allocator>&& source);   // C++17
+
+    void swap(map& m)
+        noexcept(allocator_traits<allocator_type>::is_always_equal::value &&
+            is_nothrow_swappable<key_compare>::value); // C++17
+
+    // observers:
+    allocator_type get_allocator() const noexcept;
+    key_compare    key_comp()      const;
+    value_compare  value_comp()    const;
+
+    // map operations:
+          iterator find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    template<typename K>
+        iterator find(const K& x);              // C++14
+    template<typename K>
+        const_iterator find(const K& x) const;  // C++14
+    template<typename K>
+      size_type count(const K& x) const;        // C++14
+    size_type      count(const key_type& k) const;
+        bool contains(const key_type& x) const; // C++20
+          iterator lower_bound(const key_type& k);
+    const_iterator lower_bound(const key_type& k) const;
+    template<typename K>
+        iterator lower_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator lower_bound(const K& x) const;  // C++14
+
+          iterator upper_bound(const key_type& k);
+    const_iterator upper_bound(const key_type& k) const;
+    template<typename K>
+        iterator upper_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator upper_bound(const K& x) const;  // C++14
+
+    pair<iterator,iterator>             equal_range(const key_type& k);
+    pair<const_iterator,const_iterator> equal_range(const key_type& k) const;
+    template<typename K>
+        pair<iterator,iterator>             equal_range(const K& x);        // C++14
+    template<typename K>
+        pair<const_iterator,const_iterator> equal_range(const K& x) const;  // C++14
+};
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator==(const map<Key, T, Compare, Allocator>& x,
+           const map<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator< (const map<Key, T, Compare, Allocator>& x,
+           const map<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator!=(const map<Key, T, Compare, Allocator>& x,
+           const map<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator> (const map<Key, T, Compare, Allocator>& x,
+           const map<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator>=(const map<Key, T, Compare, Allocator>& x,
+           const map<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator<=(const map<Key, T, Compare, Allocator>& x,
+           const map<Key, T, Compare, Allocator>& y);
+
+// specialized algorithms:
+template <class Key, class T, class Compare, class Allocator>
+void
+swap(map<Key, T, Compare, Allocator>& x, map<Key, T, Compare, Allocator>& y)
+    noexcept(noexcept(x.swap(y)));
+
+template <class Key, class T, class Compare, class Allocator, class Predicate>
+  void erase_if(map<Key, T, Compare, Allocator>& c, Predicate pred);  // C++20
+
+
+template <class Key, class T, class Compare = less<Key>,
+          class Allocator = allocator<pair<const Key, T>>>
+class multimap
+{
+public:
+    // types:
+    typedef Key                                      key_type;
+    typedef T                                        mapped_type;
+    typedef pair<const key_type,mapped_type>         value_type;
+    typedef Compare                                  key_compare;
+    typedef Allocator                                allocator_type;
+    typedef typename allocator_type::reference       reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+    typedef typename allocator_type::pointer         pointer;
+    typedef typename allocator_type::const_pointer   const_pointer;
+
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+    typedef unspecified                              node_type;              // C++17
+
+    class value_compare
+        : public binary_function<value_type,value_type,bool>
+    {
+        friend class multimap;
+    protected:
+        key_compare comp;
+        value_compare(key_compare c);
+    public:
+        bool operator()(const value_type& x, const value_type& y) const;
+    };
+
+    // construct/copy/destroy:
+    multimap()
+        noexcept(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value);
+    explicit multimap(const key_compare& comp);
+    multimap(const key_compare& comp, const allocator_type& a);
+    template <class InputIterator>
+        multimap(InputIterator first, InputIterator last, const key_compare& comp);
+    template <class InputIterator>
+        multimap(InputIterator first, InputIterator last, const key_compare& comp,
+                 const allocator_type& a);
+    multimap(const multimap& m);
+    multimap(multimap&& m)
+        noexcept(
+            is_nothrow_move_constructible<allocator_type>::value &&
+            is_nothrow_move_constructible<key_compare>::value);
+    explicit multimap(const allocator_type& a);
+    multimap(const multimap& m, const allocator_type& a);
+    multimap(multimap&& m, const allocator_type& a);
+    multimap(initializer_list<value_type> il, const key_compare& comp = key_compare());
+    multimap(initializer_list<value_type> il, const key_compare& comp,
+             const allocator_type& a);
+    template <class InputIterator>
+        multimap(InputIterator first, InputIterator last, const allocator_type& a)
+            : multimap(first, last, Compare(), a) {} // C++14
+    multimap(initializer_list<value_type> il, const allocator_type& a)
+        : multimap(il, Compare(), a) {} // C++14
+    ~multimap();
+
+    multimap& operator=(const multimap& m);
+    multimap& operator=(multimap&& m)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<key_compare>::value);
+    multimap& operator=(initializer_list<value_type> il);
+
+    // iterators:
+          iterator begin() noexcept;
+    const_iterator begin() const noexcept;
+          iterator end() noexcept;
+    const_iterator end()   const noexcept;
+
+          reverse_iterator rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+          reverse_iterator rend() noexcept;
+    const_reverse_iterator rend()   const noexcept;
+
+    const_iterator         cbegin()  const noexcept;
+    const_iterator         cend()    const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend()   const noexcept;
+
+    // capacity:
+    bool      empty()    const noexcept;
+    size_type size()     const noexcept;
+    size_type max_size() const noexcept;
+
+    // modifiers:
+    template <class... Args>
+        iterator emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    iterator insert(const value_type& v);
+    iterator insert(      value_type&& v);                                            // C++17
+    template <class P>
+        iterator insert(P&& p);
+    iterator insert(const_iterator position, const value_type& v);
+    iterator insert(const_iterator position,       value_type&& v);                   // C++17
+    template <class P>
+        iterator insert(const_iterator position, P&& p);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type> il);
+
+    node_type extract(const_iterator position);                                       // C++17
+    node_type extract(const key_type& x);                                             // C++17
+    iterator insert(node_type&& nh);                                                  // C++17
+    iterator insert(const_iterator hint, node_type&& nh);                             // C++17
+
+    iterator  erase(const_iterator position);
+    iterator  erase(iterator position); // C++14
+    size_type erase(const key_type& k);
+    iterator  erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class C2>
+      void merge(multimap<Key, T, C2, Allocator>& source);    // C++17
+    template<class C2>
+      void merge(multimap<Key, T, C2, Allocator>&& source);   // C++17
+    template<class C2>
+      void merge(map<Key, T, C2, Allocator>& source);         // C++17
+    template<class C2>
+      void merge(map<Key, T, C2, Allocator>&& source);        // C++17
+
+    void swap(multimap& m)
+        noexcept(allocator_traits<allocator_type>::is_always_equal::value &&
+            is_nothrow_swappable<key_compare>::value); // C++17
+
+    // observers:
+    allocator_type get_allocator() const noexcept;
+    key_compare    key_comp()      const;
+    value_compare  value_comp()    const;
+
+    // map operations:
+          iterator find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    template<typename K>
+        iterator find(const K& x);              // C++14
+    template<typename K>
+        const_iterator find(const K& x) const;  // C++14
+    template<typename K>
+      size_type count(const K& x) const;        // C++14
+    size_type      count(const key_type& k) const;
+        bool contains(const key_type& x) const; // C++20
+          iterator lower_bound(const key_type& k);
+    const_iterator lower_bound(const key_type& k) const;
+    template<typename K>
+        iterator lower_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator lower_bound(const K& x) const;  // C++14
+
+          iterator upper_bound(const key_type& k);
+    const_iterator upper_bound(const key_type& k) const;
+    template<typename K>
+        iterator upper_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator upper_bound(const K& x) const;  // C++14
+
+    pair<iterator,iterator>             equal_range(const key_type& k);
+    pair<const_iterator,const_iterator> equal_range(const key_type& k) const;
+    template<typename K>
+        pair<iterator,iterator>             equal_range(const K& x);        // C++14
+    template<typename K>
+        pair<const_iterator,const_iterator> equal_range(const K& x) const;  // C++14
+};
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator==(const multimap<Key, T, Compare, Allocator>& x,
+           const multimap<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator< (const multimap<Key, T, Compare, Allocator>& x,
+           const multimap<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator!=(const multimap<Key, T, Compare, Allocator>& x,
+           const multimap<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator> (const multimap<Key, T, Compare, Allocator>& x,
+           const multimap<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator>=(const multimap<Key, T, Compare, Allocator>& x,
+           const multimap<Key, T, Compare, Allocator>& y);
+
+template <class Key, class T, class Compare, class Allocator>
+bool
+operator<=(const multimap<Key, T, Compare, Allocator>& x,
+           const multimap<Key, T, Compare, Allocator>& y);
+
+// specialized algorithms:
+template <class Key, class T, class Compare, class Allocator>
+void
+swap(multimap<Key, T, Compare, Allocator>& x,
+     multimap<Key, T, Compare, Allocator>& y)
+    noexcept(noexcept(x.swap(y)));
+
+template <class Key, class T, class Compare, class Allocator, class Predicate>
+  void erase_if(multimap<Key, T, Compare, Allocator>& c, Predicate pred);  // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__tree>
+#include <__node_handle>
+#include <iterator>
+#include <memory>
+#include <utility>
+#include <functional>
+#include <initializer_list>
+#include <type_traits>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Key, class _CP, class _Compare,
+          bool = is_empty<_Compare>::value && !__libcpp_is_final<_Compare>::value>
+class __map_value_compare
+    : private _Compare
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_value_compare()
+        _NOEXCEPT_(is_nothrow_default_constructible<_Compare>::value)
+        : _Compare() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_value_compare(_Compare c)
+        _NOEXCEPT_(is_nothrow_copy_constructible<_Compare>::value)
+        : _Compare(c) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Compare& key_comp() const _NOEXCEPT {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _CP& __x, const _CP& __y) const
+        {return static_cast<const _Compare&>(*this)(__x.__get_value().first, __y.__get_value().first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _CP& __x, const _Key& __y) const
+        {return static_cast<const _Compare&>(*this)(__x.__get_value().first, __y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Key& __x, const _CP& __y) const
+        {return static_cast<const _Compare&>(*this)(__x, __y.__get_value().first);}
+    void swap(__map_value_compare&__y)
+        _NOEXCEPT_(__is_nothrow_swappable<_Compare>::value)
+    {
+      using _CUDA_VSTD::swap;
+      swap(static_cast<_Compare&>(*this), static_cast<_Compare&>(__y));
+    }
+
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value, bool>::type
+    operator () ( const _K2& __x, const _CP& __y ) const
+        {return static_cast<const _Compare&>(*this) (__x, __y.__get_value().first);}
+
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value, bool>::type
+    operator () (const _CP& __x, const _K2& __y) const
+        {return static_cast<const _Compare&>(*this) (__x.__get_value().first, __y);}
+#endif
+};
+
+template <class _Key, class _CP, class _Compare>
+class __map_value_compare<_Key, _CP, _Compare, false>
+{
+    _Compare comp;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_value_compare()
+        _NOEXCEPT_(is_nothrow_default_constructible<_Compare>::value)
+        : comp() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_value_compare(_Compare c)
+        _NOEXCEPT_(is_nothrow_copy_constructible<_Compare>::value)
+        : comp(c) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Compare& key_comp() const _NOEXCEPT {return comp;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _CP& __x, const _CP& __y) const
+        {return comp(__x.__get_value().first, __y.__get_value().first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _CP& __x, const _Key& __y) const
+        {return comp(__x.__get_value().first, __y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Key& __x, const _CP& __y) const
+        {return comp(__x, __y.__get_value().first);}
+    void swap(__map_value_compare&__y)
+        _NOEXCEPT_(__is_nothrow_swappable<_Compare>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(comp, __y.comp);
+    }
+
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value, bool>::type
+    operator () ( const _K2& __x, const _CP& __y ) const
+        {return comp (__x, __y.__get_value().first);}
+
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value, bool>::type
+    operator () (const _CP& __x, const _K2& __y) const
+        {return comp (__x.__get_value().first, __y);}
+#endif
+};
+
+template <class _Key, class _CP, class _Compare, bool __b>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__map_value_compare<_Key, _CP, _Compare, __b>& __x,
+     __map_value_compare<_Key, _CP, _Compare, __b>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Allocator>
+class __map_node_destructor
+{
+    typedef _Allocator                          allocator_type;
+    typedef allocator_traits<allocator_type>    __alloc_traits;
+
+public:
+    typedef typename __alloc_traits::pointer    pointer;
+
+private:
+    allocator_type& __na_;
+
+    __map_node_destructor& operator=(const __map_node_destructor&);
+
+public:
+    bool __first_constructed;
+    bool __second_constructed;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __map_node_destructor(allocator_type& __na) _NOEXCEPT
+        : __na_(__na),
+          __first_constructed(false),
+          __second_constructed(false)
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_node_destructor(__tree_node_destructor<allocator_type>&& __x) _NOEXCEPT
+        : __na_(__x.__na_),
+          __first_constructed(__x.__value_constructed),
+          __second_constructed(__x.__value_constructed)
+        {
+            __x.__value_constructed = false;
+        }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p) _NOEXCEPT
+    {
+        if (__second_constructed)
+            __alloc_traits::destroy(__na_, _CUDA_VSTD::addressof(__p->__value_.__get_value().second));
+        if (__first_constructed)
+            __alloc_traits::destroy(__na_, _CUDA_VSTD::addressof(__p->__value_.__get_value().first));
+        if (__p)
+            __alloc_traits::deallocate(__na_, __p, 1);
+    }
+};
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+    class map;
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+    class multimap;
+template <class _TreeIterator> class __map_const_iterator;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp>
+struct __value_type
+{
+    typedef _Key                                     key_type;
+    typedef _Tp                                      mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+    typedef pair<key_type&, mapped_type&>            __nc_ref_pair_type;
+    typedef pair<key_type&&, mapped_type&&>          __nc_rref_pair_type;
+
+private:
+    value_type __cc;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type& __get_value()
+    {
+#if _LIBCUDACXX_STD_VER > 14
+        return *_CUDA_VSTD::launder(_CUDA_VSTD::addressof(__cc));
+#else
+        return __cc;
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type& __get_value() const
+    {
+#if _LIBCUDACXX_STD_VER > 14
+        return *_CUDA_VSTD::launder(_CUDA_VSTD::addressof(__cc));
+#else
+        return __cc;
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __nc_ref_pair_type __ref()
+    {
+        value_type& __v = __get_value();
+        return __nc_ref_pair_type(const_cast<key_type&>(__v.first), __v.second);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __nc_rref_pair_type __move()
+    {
+        value_type& __v = __get_value();
+        return __nc_rref_pair_type(
+            _CUDA_VSTD::move(const_cast<key_type&>(__v.first)),
+            _CUDA_VSTD::move(__v.second));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_type& operator=(const __value_type& __v)
+    {
+        __ref() = __v.__get_value();
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_type& operator=(__value_type&& __v)
+    {
+        __ref() = __v.__move();
+        return *this;
+    }
+
+    template <class _ValueTp,
+              class = typename enable_if<
+                    __is_same_uncvref<_ValueTp, value_type>::value
+                 >::type
+             >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __value_type& operator=(_ValueTp&& __v)
+    {
+        __ref() = _CUDA_VSTD::forward<_ValueTp>(__v);
+        return *this;
+    }
+
+private:
+    __value_type() _LIBCUDACXX_EQUAL_DELETE;
+    ~__value_type() _LIBCUDACXX_EQUAL_DELETE;
+    __value_type(const __value_type& __v) _LIBCUDACXX_EQUAL_DELETE;
+    __value_type(__value_type&& __v) _LIBCUDACXX_EQUAL_DELETE;
+};
+
+#else
+
+template <class _Key, class _Tp>
+struct __value_type
+{
+    typedef _Key                                     key_type;
+    typedef _Tp                                      mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+
+private:
+    value_type __cc;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type& __get_value() { return __cc; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type& __get_value() const { return __cc; }
+
+private:
+   __value_type();
+   __value_type(__value_type const&);
+   __value_type& operator=(__value_type const&);
+   ~__value_type();
+};
+
+#endif // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+struct __extract_key_value_types;
+
+template <class _Key, class _Tp>
+struct __extract_key_value_types<__value_type<_Key, _Tp> >
+{
+  typedef _Key const __key_type;
+  typedef _Tp        __mapped_type;
+};
+
+template <class _TreeIterator>
+class _LIBCUDACXX_TEMPLATE_VIS __map_iterator
+{
+    typedef typename _TreeIterator::_NodeTypes                   _NodeTypes;
+    typedef typename _TreeIterator::__pointer_traits             __pointer_traits;
+
+    _TreeIterator __i_;
+
+public:
+    typedef bidirectional_iterator_tag                           iterator_category;
+    typedef typename _NodeTypes::__map_value_type                value_type;
+    typedef typename _TreeIterator::difference_type              difference_type;
+    typedef value_type&                                          reference;
+    typedef typename _NodeTypes::__map_value_type_pointer        pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_iterator() _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_iterator(_TreeIterator __i) _NOEXCEPT : __i_(__i) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return __i_->__get_value();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return pointer_traits<pointer>::pointer_to(__i_->__get_value());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_iterator& operator++() {++__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_iterator operator++(int)
+    {
+        __map_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_iterator& operator--() {--__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_iterator operator--(int)
+    {
+        __map_iterator __t(*this);
+        --(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __map_iterator& __x, const __map_iterator& __y)
+        {return __x.__i_ == __y.__i_;}
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __map_iterator& __x, const __map_iterator& __y)
+        {return __x.__i_ != __y.__i_;}
+
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __map_const_iterator;
+};
+
+template <class _TreeIterator>
+class _LIBCUDACXX_TEMPLATE_VIS __map_const_iterator
+{
+    typedef typename _TreeIterator::_NodeTypes                   _NodeTypes;
+    typedef typename _TreeIterator::__pointer_traits             __pointer_traits;
+
+    _TreeIterator __i_;
+
+public:
+    typedef bidirectional_iterator_tag                           iterator_category;
+    typedef typename _NodeTypes::__map_value_type                value_type;
+    typedef typename _TreeIterator::difference_type              difference_type;
+    typedef const value_type&                                    reference;
+    typedef typename _NodeTypes::__const_map_value_type_pointer  pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator() _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator(_TreeIterator __i) _NOEXCEPT : __i_(__i) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator(__map_iterator<
+        typename _TreeIterator::__non_const_iterator> __i) _NOEXCEPT
+        : __i_(__i.__i_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return __i_->__get_value();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return pointer_traits<pointer>::pointer_to(__i_->__get_value());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator& operator++() {++__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator operator++(int)
+    {
+        __map_const_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator& operator--() {--__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __map_const_iterator operator--(int)
+    {
+        __map_const_iterator __t(*this);
+        --(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const __map_const_iterator& __x, const __map_const_iterator& __y)
+        {return __x.__i_ == __y.__i_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const __map_const_iterator& __x, const __map_const_iterator& __y)
+        {return __x.__i_ != __y.__i_;}
+
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+    template <class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS __tree_const_iterator;
+};
+
+template <class _Key, class _Tp, class _Compare = less<_Key>,
+          class _Allocator = allocator<pair<const _Key, _Tp> > >
+class _LIBCUDACXX_TEMPLATE_VIS map
+{
+public:
+    // types:
+    typedef _Key                                     key_type;
+    typedef _Tp                                      mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+    typedef typename __identity<_Compare>::type      key_compare;
+    typedef typename __identity<_Allocator>::type    allocator_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+    class _LIBCUDACXX_TEMPLATE_VIS value_compare
+        : public binary_function<value_type, value_type, bool>
+    {
+        friend class map;
+    protected:
+        key_compare comp;
+
+        _LIBCUDACXX_INLINE_VISIBILITY value_compare(key_compare c) : comp(c) {}
+    public:
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator()(const value_type& __x, const value_type& __y) const
+            {return comp(__x.first, __y.first);}
+    };
+
+private:
+
+    typedef _CUDA_VSTD::__value_type<key_type, mapped_type>             __value_type;
+    typedef __map_value_compare<key_type, __value_type, key_compare> __vc;
+    typedef typename __rebind_alloc_helper<allocator_traits<allocator_type>,
+                                                 __value_type>::type __allocator_type;
+    typedef __tree<__value_type, __vc, __allocator_type>   __base;
+    typedef typename __base::__node_traits                 __node_traits;
+    typedef allocator_traits<allocator_type>               __alloc_traits;
+
+    __base __tree_;
+
+public:
+    typedef typename __alloc_traits::pointer               pointer;
+    typedef typename __alloc_traits::const_pointer         const_pointer;
+    typedef typename __alloc_traits::size_type             size_type;
+    typedef typename __alloc_traits::difference_type       difference_type;
+    typedef __map_iterator<typename __base::iterator>             iterator;
+    typedef __map_const_iterator<typename __base::const_iterator> const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>               reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>         const_reverse_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __map_node_handle<typename __base::__node, allocator_type> node_type;
+    typedef __insert_return_type<iterator, node_type> insert_return_type;
+#endif
+
+    template <class _Key2, class _Value2, class _Comp2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class _Key2, class _Value2, class _Comp2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map()
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(__vc(key_compare())) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit map(const key_compare& __comp)
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(__vc(__comp)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit map(const key_compare& __comp, const allocator_type& __a)
+        : __tree_(__vc(__comp), typename __base::allocator_type(__a)) {}
+
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        map(_InputIterator __f, _InputIterator __l,
+            const key_compare& __comp = key_compare())
+        : __tree_(__vc(__comp))
+        {
+            insert(__f, __l);
+        }
+
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        map(_InputIterator __f, _InputIterator __l,
+            const key_compare& __comp, const allocator_type& __a)
+        : __tree_(__vc(__comp), typename __base::allocator_type(__a))
+        {
+            insert(__f, __l);
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(_InputIterator __f, _InputIterator __l, const allocator_type& __a)
+        : map(__f, __l, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(const map& __m)
+        : __tree_(__m.__tree_)
+        {
+            insert(__m.begin(), __m.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map& operator=(const map& __m)
+        {
+#ifndef _LIBCUDACXX_CXX03_LANG
+            __tree_ = __m.__tree_;
+#else
+            if (this != &__m) {
+                __tree_.clear();
+                __tree_.value_comp() = __m.__tree_.value_comp();
+                __tree_.__copy_assign_alloc(__m.__tree_);
+                insert(__m.begin(), __m.end());
+            }
+#endif
+            return *this;
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(map&& __m)
+        _NOEXCEPT_(is_nothrow_move_constructible<__base>::value)
+        : __tree_(_CUDA_VSTD::move(__m.__tree_))
+        {
+        }
+
+    map(map&& __m, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map& operator=(map&& __m)
+        _NOEXCEPT_(is_nothrow_move_assignable<__base>::value)
+        {
+            __tree_ = _CUDA_VSTD::move(__m.__tree_);
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(initializer_list<value_type> __il, const key_compare& __comp = key_compare())
+        : __tree_(__vc(__comp))
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(initializer_list<value_type> __il, const key_compare& __comp, const allocator_type& __a)
+        : __tree_(__vc(__comp), typename __base::allocator_type(__a))
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(initializer_list<value_type> __il, const allocator_type& __a)
+        : map(__il, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map& operator=(initializer_list<value_type> __il)
+        {
+            __tree_.__assign_unique(__il.begin(), __il.end());
+            return *this;
+        }
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit map(const allocator_type& __a)
+        : __tree_(typename __base::allocator_type(__a))
+        {
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    map(const map& __m, const allocator_type& __a)
+        : __tree_(__m.__tree_.value_comp(), typename __base::allocator_type(__a))
+        {
+            insert(__m.begin(), __m.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~map() {
+        static_assert(sizeof(__diagnose_non_const_comparator<_Key, _Compare>()), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator begin() _NOEXCEPT {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator end() _NOEXCEPT {return __tree_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT {return __tree_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rbegin() _NOEXCEPT {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rend() _NOEXCEPT
+            {return       reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend() const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend() const _NOEXCEPT {return rend();}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const _NOEXCEPT {return __tree_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT {return __tree_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __tree_.max_size();}
+
+    mapped_type& operator[](const key_type& __k);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    mapped_type& operator[](key_type&& __k);
+#endif
+
+          mapped_type& at(const key_type& __k);
+    const mapped_type& at(const key_type& __k) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT {return allocator_type(__tree_.__alloc());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_compare    key_comp()      const {return __tree_.value_comp().key_comp();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_compare  value_comp()    const {return value_compare(__tree_.value_comp().key_comp());}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> emplace(_Args&& ...__args) {
+        return __tree_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator emplace_hint(const_iterator __p, _Args&& ...__args) {
+        return __tree_.__emplace_hint_unique(__p.__i_, _CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> insert(_Pp&& __p)
+            {return __tree_.__insert_unique(_CUDA_VSTD::forward<_Pp>(__p));}
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert(const_iterator __pos, _Pp&& __p)
+            {return __tree_.__insert_unique(__pos.__i_, _CUDA_VSTD::forward<_Pp>(__p));}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+        insert(const value_type& __v) {return __tree_.__insert_unique(__v);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator
+        insert(const_iterator __p, const value_type& __v)
+            {return __tree_.__insert_unique(__p.__i_, __v);}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool>
+    insert(value_type&& __v) {return __tree_.__insert_unique(_CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p,  value_type&& __v)
+    {return __tree_.__insert_unique(__p.__i_, _CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+#endif
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __f, _InputIterator __l)
+        {
+            for (const_iterator __e = cend(); __f != __l; ++__f)
+                insert(__e.__i_, *__f);
+        }
+
+#if _LIBCUDACXX_STD_VER > 14
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> try_emplace(const key_type& __k, _Args&&... __args)
+    {
+        return __tree_.__emplace_unique_key_args(__k,
+            _CUDA_VSTD::piecewise_construct,
+            _CUDA_VSTD::forward_as_tuple(__k),
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...));
+    }
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> try_emplace(key_type&& __k, _Args&&... __args)
+    {
+        return __tree_.__emplace_unique_key_args(__k,
+            _CUDA_VSTD::piecewise_construct,
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__k)),
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...));
+    }
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator try_emplace(const_iterator __h, const key_type& __k, _Args&&... __args)
+    {
+        return __tree_.__emplace_hint_unique_key_args(__h.__i_, __k,
+            _CUDA_VSTD::piecewise_construct,
+            _CUDA_VSTD::forward_as_tuple(__k),
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...));
+    }
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator try_emplace(const_iterator __h, key_type&& __k, _Args&&... __args)
+    {
+        return __tree_.__emplace_hint_unique_key_args(__h.__i_, __k,
+            _CUDA_VSTD::piecewise_construct,
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__k)),
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...));
+    }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> insert_or_assign(const key_type& __k, _Vp&& __v)
+    {
+        iterator __p = lower_bound(__k);
+        if ( __p != end() && !key_comp()(__k, __p->first))
+        {
+            __p->second = _CUDA_VSTD::forward<_Vp>(__v);
+            return _CUDA_VSTD::make_pair(__p, false);
+        }
+        return _CUDA_VSTD::make_pair(emplace_hint(__p, __k, _CUDA_VSTD::forward<_Vp>(__v)), true);
+    }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> insert_or_assign(key_type&& __k, _Vp&& __v)
+    {
+        iterator __p = lower_bound(__k);
+        if ( __p != end() && !key_comp()(__k, __p->first))
+        {
+            __p->second = _CUDA_VSTD::forward<_Vp>(__v);
+            return _CUDA_VSTD::make_pair(__p, false);
+        }
+        return _CUDA_VSTD::make_pair(emplace_hint(__p, _CUDA_VSTD::move(__k), _CUDA_VSTD::forward<_Vp>(__v)), true);
+    }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert_or_assign(const_iterator __h, const key_type& __k, _Vp&& __v)
+     {
+        iterator __p = lower_bound(__k);
+        if ( __p != end() && !key_comp()(__k, __p->first))
+        {
+            __p->second = _CUDA_VSTD::forward<_Vp>(__v);
+            return __p;
+        }
+        return emplace_hint(__h, __k, _CUDA_VSTD::forward<_Vp>(__v));
+     }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert_or_assign(const_iterator __h, key_type&& __k, _Vp&& __v)
+     {
+        iterator __p = lower_bound(__k);
+        if ( __p != end() && !key_comp()(__k, __p->first))
+        {
+            __p->second = _CUDA_VSTD::forward<_Vp>(__v);
+            return __p;
+        }
+        return emplace_hint(__h, _CUDA_VSTD::move(__k), _CUDA_VSTD::forward<_Vp>(__v));
+     }
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __p) {return __tree_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(iterator __p)       {return __tree_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k)
+        {return __tree_.__erase_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator  erase(const_iterator __f, const_iterator __l)
+        {return __tree_.erase(__f.__i_, __l.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__tree_.clear();}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    insert_return_type insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to map::insert()");
+        return __tree_.template __node_handle_insert_unique<
+            node_type, insert_return_type>(_CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to map::insert()");
+        return __tree_.template __node_handle_insert_unique<node_type>(
+            __hint.__i_, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__it.__i_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(map<key_type, mapped_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(map<key_type, mapped_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multimap<key_type, mapped_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multimap<key_type, mapped_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(map& __m)
+        _NOEXCEPT_(__is_nothrow_swappable<__base>::value)
+        {__tree_.swap(__m.__tree_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator find(const key_type& __k)             {return __tree_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __tree_.find(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    find(const _K2& __k)                           {return __tree_.find(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    find(const _K2& __k) const                     {return __tree_.find(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type      count(const key_type& __k) const
+        {return __tree_.__count_unique(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,size_type>::type
+    count(const _K2& __k) const {return __tree_.__count_multi(__k);}
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool contains(const key_type& __k) const {return find(__k) != end();}
+#endif // _LIBCUDACXX_STD_VER > 17
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator lower_bound(const key_type& __k)
+        {return __tree_.lower_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator lower_bound(const key_type& __k) const
+        {return __tree_.lower_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    lower_bound(const _K2& __k)       {return __tree_.lower_bound(__k);}
+
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    lower_bound(const _K2& __k) const {return __tree_.lower_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator upper_bound(const key_type& __k)
+        {return __tree_.upper_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator upper_bound(const key_type& __k) const
+        {return __tree_.upper_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    upper_bound(const _K2& __k)       {return __tree_.upper_bound(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    upper_bound(const _K2& __k) const {return __tree_.upper_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator,iterator> equal_range(const key_type& __k)
+        {return __tree_.__equal_range_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator,const_iterator> equal_range(const key_type& __k) const
+        {return __tree_.__equal_range_unique(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,pair<iterator,iterator>>::type
+    equal_range(const _K2& __k)       {return __tree_.__equal_range_multi(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,pair<const_iterator,const_iterator>>::type
+    equal_range(const _K2& __k) const {return __tree_.__equal_range_multi(__k);}
+#endif
+
+private:
+    typedef typename __base::__node                    __node;
+    typedef typename __base::__node_allocator          __node_allocator;
+    typedef typename __base::__node_pointer            __node_pointer;
+    typedef typename __base::__node_base_pointer       __node_base_pointer;
+    typedef typename __base::__parent_pointer          __parent_pointer;
+
+    typedef __map_node_destructor<__node_allocator> _Dp;
+    typedef unique_ptr<__node, _Dp> __node_holder;
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    __node_holder __construct_node_with_key(const key_type& __k);
+#endif
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator, class _Compare = less<__iter_key_type<_InputIterator>>,
+         class _Allocator = allocator<__iter_to_alloc_type<_InputIterator>>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+map(_InputIterator, _InputIterator, _Compare = _Compare(), _Allocator = _Allocator())
+  -> map<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>, _Compare, _Allocator>;
+
+template<class _Key, class _Tp, class _Compare = less<remove_const_t<_Key>>,
+         class _Allocator = allocator<pair<const _Key, _Tp>>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+map(initializer_list<pair<_Key, _Tp>>, _Compare = _Compare(), _Allocator = _Allocator())
+  -> map<remove_const_t<_Key>, _Tp, _Compare, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+map(_InputIterator, _InputIterator, _Allocator)
+  -> map<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+         less<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+map(initializer_list<pair<_Key, _Tp>>, _Allocator)
+  -> map<remove_const_t<_Key>, _Tp, less<remove_const_t<_Key>>, _Allocator>;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+map<_Key, _Tp, _Compare, _Allocator>::map(map&& __m, const allocator_type& __a)
+    : __tree_(_CUDA_VSTD::move(__m.__tree_), typename __base::allocator_type(__a))
+{
+    if (__a != __m.get_allocator())
+    {
+        const_iterator __e = cend();
+        while (!__m.empty())
+            __tree_.__insert_unique(__e.__i_,
+                    __m.__tree_.remove(__m.begin().__i_)->__value_.__move());
+    }
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+_Tp&
+map<_Key, _Tp, _Compare, _Allocator>::operator[](const key_type& __k)
+{
+    return __tree_.__emplace_unique_key_args(__k,
+        _CUDA_VSTD::piecewise_construct,
+        _CUDA_VSTD::forward_as_tuple(__k),
+        _CUDA_VSTD::forward_as_tuple()).first->__get_value().second;
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+_Tp&
+map<_Key, _Tp, _Compare, _Allocator>::operator[](key_type&& __k)
+{
+    return __tree_.__emplace_unique_key_args(__k,
+        _CUDA_VSTD::piecewise_construct,
+        _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__k)),
+        _CUDA_VSTD::forward_as_tuple()).first->__get_value().second;
+}
+
+#else // _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+typename map<_Key, _Tp, _Compare, _Allocator>::__node_holder
+map<_Key, _Tp, _Compare, _Allocator>::__construct_node_with_key(const key_type& __k)
+{
+    __node_allocator& __na = __tree_.__node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _CUDA_VSTD::addressof(__h->__value_.__get_value().first), __k);
+    __h.get_deleter().__first_constructed = true;
+    __node_traits::construct(__na, _CUDA_VSTD::addressof(__h->__value_.__get_value().second));
+    __h.get_deleter().__second_constructed = true;
+    return _LIBCUDACXX_EXPLICIT_MOVE(__h);  // explicitly moved for C++03
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+_Tp&
+map<_Key, _Tp, _Compare, _Allocator>::operator[](const key_type& __k)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __tree_.__find_equal(__parent, __k);
+    __node_pointer __r = static_cast<__node_pointer>(__child);
+    if (__child == nullptr)
+    {
+        __node_holder __h = __construct_node_with_key(__k);
+        __tree_.__insert_node_at(__parent, __child, static_cast<__node_base_pointer>(__h.get()));
+        __r = __h.release();
+    }
+    return __r->__value_.__get_value().second;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+_Tp&
+map<_Key, _Tp, _Compare, _Allocator>::at(const key_type& __k)
+{
+    __parent_pointer __parent;
+    __node_base_pointer& __child = __tree_.__find_equal(__parent, __k);
+    if (__child == nullptr)
+        __throw_out_of_range("map::at:  key not found");
+    return static_cast<__node_pointer>(__child)->__value_.__get_value().second;
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+const _Tp&
+map<_Key, _Tp, _Compare, _Allocator>::at(const key_type& __k) const
+{
+    __parent_pointer __parent;
+    __node_base_pointer __child = __tree_.__find_equal(__parent, __k);
+    if (__child == nullptr)
+        __throw_out_of_range("map::at:  key not found");
+    return static_cast<__node_pointer>(__child)->__value_.__get_value().second;
+}
+
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const map<_Key, _Tp, _Compare, _Allocator>& __x,
+           const map<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return __x.size() == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const map<_Key, _Tp, _Compare, _Allocator>& __x,
+           const map<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const map<_Key, _Tp, _Compare, _Allocator>& __x,
+           const map<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const map<_Key, _Tp, _Compare, _Allocator>& __x,
+           const map<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const map<_Key, _Tp, _Compare, _Allocator>& __x,
+           const map<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const map<_Key, _Tp, _Compare, _Allocator>& __x,
+           const map<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(map<_Key, _Tp, _Compare, _Allocator>& __x,
+     map<_Key, _Tp, _Compare, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Key, class _Tp, class _Compare, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(map<_Key, _Tp, _Compare, _Allocator>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+
+template <class _Key, class _Tp, class _Compare = less<_Key>,
+          class _Allocator = allocator<pair<const _Key, _Tp> > >
+class _LIBCUDACXX_TEMPLATE_VIS multimap
+{
+public:
+    // types:
+    typedef _Key                                     key_type;
+    typedef _Tp                                      mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+    typedef typename __identity<_Compare>::type      key_compare;
+    typedef typename __identity<_Allocator>::type    allocator_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+    class _LIBCUDACXX_TEMPLATE_VIS value_compare
+        : public binary_function<value_type, value_type, bool>
+    {
+        friend class multimap;
+    protected:
+        key_compare comp;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        value_compare(key_compare c) : comp(c) {}
+    public:
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator()(const value_type& __x, const value_type& __y) const
+            {return comp(__x.first, __y.first);}
+    };
+
+private:
+
+    typedef _CUDA_VSTD::__value_type<key_type, mapped_type>             __value_type;
+    typedef __map_value_compare<key_type, __value_type, key_compare> __vc;
+    typedef typename __rebind_alloc_helper<allocator_traits<allocator_type>,
+                                                 __value_type>::type __allocator_type;
+    typedef __tree<__value_type, __vc, __allocator_type>            __base;
+    typedef typename __base::__node_traits                          __node_traits;
+    typedef allocator_traits<allocator_type>                        __alloc_traits;
+
+    __base __tree_;
+
+public:
+    typedef typename __alloc_traits::pointer               pointer;
+    typedef typename __alloc_traits::const_pointer         const_pointer;
+    typedef typename __alloc_traits::size_type             size_type;
+    typedef typename __alloc_traits::difference_type       difference_type;
+    typedef __map_iterator<typename __base::iterator>      iterator;
+    typedef __map_const_iterator<typename __base::const_iterator> const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>               reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>         const_reverse_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __map_node_handle<typename __base::__node, allocator_type> node_type;
+#endif
+
+    template <class _Key2, class _Value2, class _Comp2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS map;
+    template <class _Key2, class _Value2, class _Comp2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS multimap;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap()
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(__vc(key_compare())) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit multimap(const key_compare& __comp)
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(__vc(__comp)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit multimap(const key_compare& __comp, const allocator_type& __a)
+        : __tree_(__vc(__comp), typename __base::allocator_type(__a)) {}
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        multimap(_InputIterator __f, _InputIterator __l,
+            const key_compare& __comp = key_compare())
+        : __tree_(__vc(__comp))
+        {
+            insert(__f, __l);
+        }
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        multimap(_InputIterator __f, _InputIterator __l,
+            const key_compare& __comp, const allocator_type& __a)
+        : __tree_(__vc(__comp), typename __base::allocator_type(__a))
+        {
+            insert(__f, __l);
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(_InputIterator __f, _InputIterator __l, const allocator_type& __a)
+        : multimap(__f, __l, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(const multimap& __m)
+        : __tree_(__m.__tree_.value_comp(),
+          __alloc_traits::select_on_container_copy_construction(__m.__tree_.__alloc()))
+        {
+            insert(__m.begin(), __m.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap& operator=(const multimap& __m)
+        {
+#ifndef _LIBCUDACXX_CXX03_LANG
+            __tree_ = __m.__tree_;
+#else
+            if (this != &__m) {
+                __tree_.clear();
+                __tree_.value_comp() = __m.__tree_.value_comp();
+                __tree_.__copy_assign_alloc(__m.__tree_);
+                insert(__m.begin(), __m.end());
+            }
+#endif
+            return *this;
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(multimap&& __m)
+        _NOEXCEPT_(is_nothrow_move_constructible<__base>::value)
+        : __tree_(_CUDA_VSTD::move(__m.__tree_))
+        {
+        }
+
+    multimap(multimap&& __m, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap& operator=(multimap&& __m)
+        _NOEXCEPT_(is_nothrow_move_assignable<__base>::value)
+        {
+            __tree_ = _CUDA_VSTD::move(__m.__tree_);
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(initializer_list<value_type> __il, const key_compare& __comp = key_compare())
+        : __tree_(__vc(__comp))
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(initializer_list<value_type> __il, const key_compare& __comp, const allocator_type& __a)
+        : __tree_(__vc(__comp), typename __base::allocator_type(__a))
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(initializer_list<value_type> __il, const allocator_type& __a)
+        : multimap(__il, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap& operator=(initializer_list<value_type> __il)
+        {
+            __tree_.__assign_multi(__il.begin(), __il.end());
+            return *this;
+        }
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit multimap(const allocator_type& __a)
+        : __tree_(typename __base::allocator_type(__a))
+        {
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multimap(const multimap& __m, const allocator_type& __a)
+        : __tree_(__m.__tree_.value_comp(), typename __base::allocator_type(__a))
+        {
+            insert(__m.begin(), __m.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~multimap() {
+        static_assert(sizeof(__diagnose_non_const_comparator<_Key, _Compare>()), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator begin() _NOEXCEPT {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator end() _NOEXCEPT {return __tree_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT {return __tree_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rbegin() _NOEXCEPT {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rend() _NOEXCEPT {return reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend() const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin()  const _NOEXCEPT {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend() const _NOEXCEPT {return rend();}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return __tree_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT {return __tree_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __tree_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT {return allocator_type(__tree_.__alloc());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_compare    key_comp() const {return __tree_.value_comp().key_comp();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_compare  value_comp() const
+        {return value_compare(__tree_.value_comp().key_comp());}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator emplace(_Args&& ...__args) {
+        return __tree_.__emplace_multi(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class ..._Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator emplace_hint(const_iterator __p, _Args&& ...__args) {
+        return __tree_.__emplace_hint_multi(__p.__i_, _CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert(_Pp&& __p)
+            {return __tree_.__insert_multi(_CUDA_VSTD::forward<_Pp>(__p));}
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert(const_iterator __pos, _Pp&& __p)
+            {return __tree_.__insert_multi(__pos.__i_, _CUDA_VSTD::forward<_Pp>(__p));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(value_type&& __v)
+        {return __tree_.__insert_multi(_CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, value_type&& __v)
+        {return __tree_.__insert_multi(__p.__i_, _CUDA_VSTD::move(__v));}
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const value_type& __v) {return __tree_.__insert_multi(__v);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, const value_type& __v)
+            {return __tree_.__insert_multi(__p.__i_, __v);}
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __f, _InputIterator __l)
+        {
+            for (const_iterator __e = cend(); __f != __l; ++__f)
+                __tree_.__insert_multi(__e.__i_, *__f);
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __p) {return __tree_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(iterator __p)       {return __tree_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __tree_.__erase_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator  erase(const_iterator __f, const_iterator __l)
+        {return __tree_.erase(__f.__i_, __l.__i_);}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to multimap::insert()");
+        return __tree_.template __node_handle_insert_multi<node_type>(
+            _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to multimap::insert()");
+        return __tree_.template __node_handle_insert_multi<node_type>(
+            __hint.__i_, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __tree_.template __node_handle_extract<node_type>(
+            __it.__i_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multimap<key_type, mapped_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multimap<key_type, mapped_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(map<key_type, mapped_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(map<key_type, mapped_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__tree_.clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(multimap& __m)
+        _NOEXCEPT_(__is_nothrow_swappable<__base>::value)
+        {__tree_.swap(__m.__tree_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator find(const key_type& __k)             {return __tree_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __tree_.find(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    find(const _K2& __k)                           {return __tree_.find(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    find(const _K2& __k) const                     {return __tree_.find(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type      count(const key_type& __k) const
+        {return __tree_.__count_multi(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,size_type>::type
+    count(const _K2& __k) const {return __tree_.__count_multi(__k);}
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool contains(const key_type& __k) const {return find(__k) != end();}
+#endif // _LIBCUDACXX_STD_VER > 17
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator lower_bound(const key_type& __k)
+        {return __tree_.lower_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator lower_bound(const key_type& __k) const
+            {return __tree_.lower_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    lower_bound(const _K2& __k)       {return __tree_.lower_bound(__k);}
+
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    lower_bound(const _K2& __k) const {return __tree_.lower_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator upper_bound(const key_type& __k)
+            {return __tree_.upper_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator upper_bound(const key_type& __k) const
+            {return __tree_.upper_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    upper_bound(const _K2& __k)       {return __tree_.upper_bound(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    upper_bound(const _K2& __k) const {return __tree_.upper_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator,iterator>             equal_range(const key_type& __k)
+            {return __tree_.__equal_range_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator,const_iterator> equal_range(const key_type& __k) const
+            {return __tree_.__equal_range_multi(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,pair<iterator,iterator>>::type
+    equal_range(const _K2& __k)       {return __tree_.__equal_range_multi(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,pair<const_iterator,const_iterator>>::type
+    equal_range(const _K2& __k) const {return __tree_.__equal_range_multi(__k);}
+#endif
+
+private:
+    typedef typename __base::__node                    __node;
+    typedef typename __base::__node_allocator          __node_allocator;
+    typedef typename __base::__node_pointer            __node_pointer;
+
+    typedef __map_node_destructor<__node_allocator> _Dp;
+    typedef unique_ptr<__node, _Dp> __node_holder;
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator, class _Compare = less<__iter_key_type<_InputIterator>>,
+         class _Allocator = allocator<__iter_to_alloc_type<_InputIterator>>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+multimap(_InputIterator, _InputIterator, _Compare = _Compare(), _Allocator = _Allocator())
+  -> multimap<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>, _Compare, _Allocator>;
+
+template<class _Key, class _Tp, class _Compare = less<remove_const_t<_Key>>,
+         class _Allocator = allocator<pair<const _Key, _Tp>>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+multimap(initializer_list<pair<_Key, _Tp>>, _Compare = _Compare(), _Allocator = _Allocator())
+  -> multimap<remove_const_t<_Key>, _Tp, _Compare, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+multimap(_InputIterator, _InputIterator, _Allocator)
+  -> multimap<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+         less<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+multimap(initializer_list<pair<_Key, _Tp>>, _Allocator)
+  -> multimap<remove_const_t<_Key>, _Tp, less<remove_const_t<_Key>>, _Allocator>;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+multimap<_Key, _Tp, _Compare, _Allocator>::multimap(multimap&& __m, const allocator_type& __a)
+    : __tree_(_CUDA_VSTD::move(__m.__tree_), typename __base::allocator_type(__a))
+{
+    if (__a != __m.get_allocator())
+    {
+        const_iterator __e = cend();
+        while (!__m.empty())
+            __tree_.__insert_multi(__e.__i_,
+                    _CUDA_VSTD::move(__m.__tree_.remove(__m.begin().__i_)->__value_.__move()));
+    }
+}
+#endif
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+           const multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return __x.size() == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+           const multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+           const multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+           const multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+           const multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+           const multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Key, class _Tp, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(multimap<_Key, _Tp, _Compare, _Allocator>& __x,
+     multimap<_Key, _Tp, _Compare, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Key, class _Tp, class _Compare, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(multimap<_Key, _Tp, _Compare, _Allocator>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_MAP
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/math.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/math.h
new file mode 100644
index 000000000000..ad26e367cff7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/math.h
@@ -0,0 +1,1574 @@
+// -*- C++ -*-
+//===---------------------------- math.h ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_MATH_H
+#define _LIBCUDACXX_MATH_H
+
+/*
+    math.h synopsis
+
+Macros:
+
+    HUGE_VAL
+    HUGE_VALF               // C99
+    HUGE_VALL               // C99
+    INFINITY                // C99
+    NAN                     // C99
+    FP_INFINITE             // C99
+    FP_NAN                  // C99
+    FP_NORMAL               // C99
+    FP_SUBNORMAL            // C99
+    FP_ZERO                 // C99
+    FP_FAST_FMA             // C99
+    FP_FAST_FMAF            // C99
+    FP_FAST_FMAL            // C99
+    FP_ILOGB0               // C99
+    FP_ILOGBNAN             // C99
+    MATH_ERRNO              // C99
+    MATH_ERREXCEPT          // C99
+    math_errhandling        // C99
+
+Types:
+
+    float_t                 // C99
+    double_t                // C99
+
+// C90
+
+floating_point abs(floating_point x);
+
+floating_point acos (arithmetic x);
+float          acosf(float x);
+long double    acosl(long double x);
+
+floating_point asin (arithmetic x);
+float          asinf(float x);
+long double    asinl(long double x);
+
+floating_point atan (arithmetic x);
+float          atanf(float x);
+long double    atanl(long double x);
+
+floating_point atan2 (arithmetic y, arithmetic x);
+float          atan2f(float y, float x);
+long double    atan2l(long double y, long double x);
+
+floating_point ceil (arithmetic x);
+float          ceilf(float x);
+long double    ceill(long double x);
+
+floating_point cos (arithmetic x);
+float          cosf(float x);
+long double    cosl(long double x);
+
+floating_point cosh (arithmetic x);
+float          coshf(float x);
+long double    coshl(long double x);
+
+floating_point exp (arithmetic x);
+float          expf(float x);
+long double    expl(long double x);
+
+floating_point fabs (arithmetic x);
+float          fabsf(float x);
+long double    fabsl(long double x);
+
+floating_point floor (arithmetic x);
+float          floorf(float x);
+long double    floorl(long double x);
+
+floating_point fmod (arithmetic x, arithmetic y);
+float          fmodf(float x, float y);
+long double    fmodl(long double x, long double y);
+
+floating_point frexp (arithmetic value, int* exp);
+float          frexpf(float value, int* exp);
+long double    frexpl(long double value, int* exp);
+
+floating_point ldexp (arithmetic value, int exp);
+float          ldexpf(float value, int exp);
+long double    ldexpl(long double value, int exp);
+
+floating_point log (arithmetic x);
+float          logf(float x);
+long double    logl(long double x);
+
+floating_point log10 (arithmetic x);
+float          log10f(float x);
+long double    log10l(long double x);
+
+floating_point modf (floating_point value, floating_point* iptr);
+float          modff(float value, float* iptr);
+long double    modfl(long double value, long double* iptr);
+
+floating_point pow (arithmetic x, arithmetic y);
+float          powf(float x, float y);
+long double    powl(long double x, long double y);
+
+floating_point sin (arithmetic x);
+float          sinf(float x);
+long double    sinl(long double x);
+
+floating_point sinh (arithmetic x);
+float          sinhf(float x);
+long double    sinhl(long double x);
+
+floating_point sqrt (arithmetic x);
+float          sqrtf(float x);
+long double    sqrtl(long double x);
+
+floating_point tan (arithmetic x);
+float          tanf(float x);
+long double    tanl(long double x);
+
+floating_point tanh (arithmetic x);
+float          tanhf(float x);
+long double    tanhl(long double x);
+
+//  C99
+
+bool signbit(arithmetic x);
+
+int fpclassify(arithmetic x);
+
+bool isfinite(arithmetic x);
+bool isinf(arithmetic x);
+bool isnan(arithmetic x);
+bool isnormal(arithmetic x);
+
+bool isgreater(arithmetic x, arithmetic y);
+bool isgreaterequal(arithmetic x, arithmetic y);
+bool isless(arithmetic x, arithmetic y);
+bool islessequal(arithmetic x, arithmetic y);
+bool islessgreater(arithmetic x, arithmetic y);
+bool isunordered(arithmetic x, arithmetic y);
+
+floating_point acosh (arithmetic x);
+float          acoshf(float x);
+long double    acoshl(long double x);
+
+floating_point asinh (arithmetic x);
+float          asinhf(float x);
+long double    asinhl(long double x);
+
+floating_point atanh (arithmetic x);
+float          atanhf(float x);
+long double    atanhl(long double x);
+
+floating_point cbrt (arithmetic x);
+float          cbrtf(float x);
+long double    cbrtl(long double x);
+
+floating_point copysign (arithmetic x, arithmetic y);
+float          copysignf(float x, float y);
+long double    copysignl(long double x, long double y);
+
+floating_point erf (arithmetic x);
+float          erff(float x);
+long double    erfl(long double x);
+
+floating_point erfc (arithmetic x);
+float          erfcf(float x);
+long double    erfcl(long double x);
+
+floating_point exp2 (arithmetic x);
+float          exp2f(float x);
+long double    exp2l(long double x);
+
+floating_point expm1 (arithmetic x);
+float          expm1f(float x);
+long double    expm1l(long double x);
+
+floating_point fdim (arithmetic x, arithmetic y);
+float          fdimf(float x, float y);
+long double    fdiml(long double x, long double y);
+
+floating_point fma (arithmetic x, arithmetic y, arithmetic z);
+float          fmaf(float x, float y, float z);
+long double    fmal(long double x, long double y, long double z);
+
+floating_point fmax (arithmetic x, arithmetic y);
+float          fmaxf(float x, float y);
+long double    fmaxl(long double x, long double y);
+
+floating_point fmin (arithmetic x, arithmetic y);
+float          fminf(float x, float y);
+long double    fminl(long double x, long double y);
+
+floating_point hypot (arithmetic x, arithmetic y);
+float          hypotf(float x, float y);
+long double    hypotl(long double x, long double y);
+
+int ilogb (arithmetic x);
+int ilogbf(float x);
+int ilogbl(long double x);
+
+floating_point lgamma (arithmetic x);
+float          lgammaf(float x);
+long double    lgammal(long double x);
+
+long long llrint (arithmetic x);
+long long llrintf(float x);
+long long llrintl(long double x);
+
+long long llround (arithmetic x);
+long long llroundf(float x);
+long long llroundl(long double x);
+
+floating_point log1p (arithmetic x);
+float          log1pf(float x);
+long double    log1pl(long double x);
+
+floating_point log2 (arithmetic x);
+float          log2f(float x);
+long double    log2l(long double x);
+
+floating_point logb (arithmetic x);
+float          logbf(float x);
+long double    logbl(long double x);
+
+long lrint (arithmetic x);
+long lrintf(float x);
+long lrintl(long double x);
+
+long lround (arithmetic x);
+long lroundf(float x);
+long lroundl(long double x);
+
+double      nan (const char* str);
+float       nanf(const char* str);
+long double nanl(const char* str);
+
+floating_point nearbyint (arithmetic x);
+float          nearbyintf(float x);
+long double    nearbyintl(long double x);
+
+floating_point nextafter (arithmetic x, arithmetic y);
+float          nextafterf(float x, float y);
+long double    nextafterl(long double x, long double y);
+
+floating_point nexttoward (arithmetic x, long double y);
+float          nexttowardf(float x, long double y);
+long double    nexttowardl(long double x, long double y);
+
+floating_point remainder (arithmetic x, arithmetic y);
+float          remainderf(float x, float y);
+long double    remainderl(long double x, long double y);
+
+floating_point remquo (arithmetic x, arithmetic y, int* pquo);
+float          remquof(float x, float y, int* pquo);
+long double    remquol(long double x, long double y, int* pquo);
+
+floating_point rint (arithmetic x);
+float          rintf(float x);
+long double    rintl(long double x);
+
+floating_point round (arithmetic x);
+float          roundf(float x);
+long double    roundl(long double x);
+
+floating_point scalbln (arithmetic x, long ex);
+float          scalblnf(float x, long ex);
+long double    scalblnl(long double x, long ex);
+
+floating_point scalbn (arithmetic x, int ex);
+float          scalbnf(float x, int ex);
+long double    scalbnl(long double x, int ex);
+
+floating_point tgamma (arithmetic x);
+float          tgammaf(float x);
+long double    tgammal(long double x);
+
+floating_point trunc (arithmetic x);
+float          truncf(float x);
+long double    truncl(long double x);
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#define _LIBCUDACXX_STDLIB_INCLUDE_NEXT
+#include <stdlib.h>
+
+#include_next <math.h>
+
+#ifdef __cplusplus
+
+// We support including .h headers inside 'extern "C"' contexts, so switch
+// back to C++ linkage before including these C++ headers.
+extern "C++" {
+
+#include <type_traits>
+#include <limits>
+
+// signbit
+
+#ifdef signbit
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_signbit(_A1 __lcpp_x) _NOEXCEPT
+{
+    return signbit(__lcpp_x);
+}
+
+#undef signbit
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_floating_point<_A1>::value, bool>::type
+signbit(_A1 __lcpp_x) _NOEXCEPT
+{
+    return __libcpp_signbit((typename std::__promote<_A1>::type)__lcpp_x);
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_integral<_A1>::value && std::is_signed<_A1>::value, bool>::type
+signbit(_A1 __lcpp_x) _NOEXCEPT
+{ return __lcpp_x < 0; }
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_integral<_A1>::value && !std::is_signed<_A1>::value, bool>::type
+signbit(_A1) _NOEXCEPT
+{ return false; }
+
+#elif defined(_LIBCUDACXX_MSVCRT)
+
+template <typename _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_floating_point<_A1>::value, bool>::type
+signbit(_A1 __lcpp_x) _NOEXCEPT
+{
+  return ::signbit(static_cast<typename std::__promote<_A1>::type>(__lcpp_x));
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_integral<_A1>::value && std::is_signed<_A1>::value, bool>::type
+signbit(_A1 __lcpp_x) _NOEXCEPT
+{ return __lcpp_x < 0; }
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_integral<_A1>::value && !std::is_signed<_A1>::value, bool>::type
+signbit(_A1) _NOEXCEPT
+{ return false; }
+
+#endif  // signbit
+
+// fpclassify
+
+#ifdef fpclassify
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+int
+__libcpp_fpclassify(_A1 __lcpp_x) _NOEXCEPT
+{
+    return fpclassify(__lcpp_x);
+}
+
+#undef fpclassify
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_floating_point<_A1>::value, int>::type
+fpclassify(_A1 __lcpp_x) _NOEXCEPT
+{
+    return __libcpp_fpclassify((typename std::__promote<_A1>::type)__lcpp_x);
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, int>::type
+fpclassify(_A1 __lcpp_x) _NOEXCEPT
+{ return __lcpp_x == 0 ? FP_ZERO : FP_NORMAL; }
+
+#elif defined(_LIBCUDACXX_MSVCRT)
+
+template <typename _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_floating_point<_A1>::value, bool>::type
+fpclassify(_A1 __lcpp_x) _NOEXCEPT
+{
+  return ::fpclassify(static_cast<typename std::__promote<_A1>::type>(__lcpp_x));
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, int>::type
+fpclassify(_A1 __lcpp_x) _NOEXCEPT
+{ return __lcpp_x == 0 ? FP_ZERO : FP_NORMAL; }
+
+#endif  // fpclassify
+
+// isfinite
+
+#ifdef isfinite
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isfinite(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isfinite(__lcpp_x);
+}
+
+#undef isfinite
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_arithmetic<_A1>::value && std::numeric_limits<_A1>::has_infinity,
+    bool>::type
+isfinite(_A1 __lcpp_x) _NOEXCEPT
+{
+    return __libcpp_isfinite((typename std::__promote<_A1>::type)__lcpp_x);
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_arithmetic<_A1>::value && !std::numeric_limits<_A1>::has_infinity,
+    bool>::type
+isfinite(_A1) _NOEXCEPT
+{ return true; }
+
+#endif  // isfinite
+
+// isinf
+
+#ifdef isinf
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isinf(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isinf(__lcpp_x);
+}
+
+#undef isinf
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_arithmetic<_A1>::value && std::numeric_limits<_A1>::has_infinity,
+    bool>::type
+isinf(_A1 __lcpp_x) _NOEXCEPT
+{
+    return __libcpp_isinf((typename std::__promote<_A1>::type)__lcpp_x);
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<
+    std::is_arithmetic<_A1>::value && !std::numeric_limits<_A1>::has_infinity,
+    bool>::type
+isinf(_A1) _NOEXCEPT
+{ return false; }
+
+#ifdef _LIBCUDACXX_PREFERRED_OVERLOAD
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isinf(float __lcpp_x) _NOEXCEPT { return __libcpp_isinf(__lcpp_x); }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+bool
+isinf(double __lcpp_x) _NOEXCEPT { return __libcpp_isinf(__lcpp_x); }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isinf(long double __lcpp_x) _NOEXCEPT { return __libcpp_isinf(__lcpp_x); }
+#endif
+
+#endif  // isinf
+
+// isnan
+
+#ifdef isnan
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isnan(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isnan(__lcpp_x);
+}
+
+#undef isnan
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_floating_point<_A1>::value, bool>::type
+isnan(_A1 __lcpp_x) _NOEXCEPT
+{
+    return __libcpp_isnan((typename std::__promote<_A1>::type)__lcpp_x);
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, bool>::type
+isnan(_A1) _NOEXCEPT
+{ return false; }
+
+#ifdef _LIBCUDACXX_PREFERRED_OVERLOAD
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isnan(float __lcpp_x) _NOEXCEPT { return __libcpp_isnan(__lcpp_x); }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+bool
+isnan(double __lcpp_x) _NOEXCEPT { return __libcpp_isnan(__lcpp_x); }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+isnan(long double __lcpp_x) _NOEXCEPT { return __libcpp_isnan(__lcpp_x); }
+#endif
+
+#endif  // isnan
+
+// isnormal
+
+#ifdef isnormal
+
+template <class _A1>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isnormal(_A1 __lcpp_x) _NOEXCEPT
+{
+    return isnormal(__lcpp_x);
+}
+
+#undef isnormal
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_floating_point<_A1>::value, bool>::type
+isnormal(_A1 __lcpp_x) _NOEXCEPT
+{
+    return __libcpp_isnormal((typename std::__promote<_A1>::type)__lcpp_x);
+}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, bool>::type
+isnormal(_A1 __lcpp_x) _NOEXCEPT
+{ return __lcpp_x != 0; }
+
+#endif  // isnormal
+
+// isgreater
+
+#ifdef isgreater
+
+template <class _A1, class _A2>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isgreater(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    return isgreater(__lcpp_x, __lcpp_y);
+}
+
+#undef isgreater
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    bool
+>::type
+isgreater(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type type;
+    return __libcpp_isgreater((type)__lcpp_x, (type)__lcpp_y);
+}
+
+#endif  // isgreater
+
+// isgreaterequal
+
+#ifdef isgreaterequal
+
+template <class _A1, class _A2>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isgreaterequal(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    return isgreaterequal(__lcpp_x, __lcpp_y);
+}
+
+#undef isgreaterequal
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    bool
+>::type
+isgreaterequal(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type type;
+    return __libcpp_isgreaterequal((type)__lcpp_x, (type)__lcpp_y);
+}
+
+#endif  // isgreaterequal
+
+// isless
+
+#ifdef isless
+
+template <class _A1, class _A2>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isless(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    return isless(__lcpp_x, __lcpp_y);
+}
+
+#undef isless
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    bool
+>::type
+isless(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type type;
+    return __libcpp_isless((type)__lcpp_x, (type)__lcpp_y);
+}
+
+#endif  // isless
+
+// islessequal
+
+#ifdef islessequal
+
+template <class _A1, class _A2>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_islessequal(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    return islessequal(__lcpp_x, __lcpp_y);
+}
+
+#undef islessequal
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    bool
+>::type
+islessequal(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type type;
+    return __libcpp_islessequal((type)__lcpp_x, (type)__lcpp_y);
+}
+
+#endif  // islessequal
+
+// islessgreater
+
+#ifdef islessgreater
+
+template <class _A1, class _A2>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_islessgreater(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    return islessgreater(__lcpp_x, __lcpp_y);
+}
+
+#undef islessgreater
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    bool
+>::type
+islessgreater(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type type;
+    return __libcpp_islessgreater((type)__lcpp_x, (type)__lcpp_y);
+}
+
+#endif  // islessgreater
+
+// isunordered
+
+#ifdef isunordered
+
+template <class _A1, class _A2>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+__libcpp_isunordered(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    return isunordered(__lcpp_x, __lcpp_y);
+}
+
+#undef isunordered
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    bool
+>::type
+isunordered(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type type;
+    return __libcpp_isunordered((type)__lcpp_x, (type)__lcpp_y);
+}
+
+#endif  // isunordered
+
+// abs
+
+#undef abs
+#undef labs
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+#undef llabs
+#endif
+
+// MSVCRT already has the correct prototype in <stdlib.h> if __cplusplus is defined
+#if !defined(_LIBCUDACXX_MSVCRT) && !defined(__sun__) && !defined(_AIX)
+inline _LIBCUDACXX_INLINE_VISIBILITY long abs(long __x) _NOEXCEPT {
+  return ::labs(__x);
+}
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+inline _LIBCUDACXX_INLINE_VISIBILITY long long abs(long long __x) _NOEXCEPT {
+  return ::llabs(__x);
+}
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+#endif // !defined(_LIBCUDACXX_MSVCRT) && !defined(__sun__) && !defined(_AIX)
+
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float abs(float __lcpp_x) _NOEXCEPT {
+  return ::fabsf(__lcpp_x);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY double abs(double __lcpp_x) _NOEXCEPT {
+  return ::fabs(__lcpp_x);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long double
+abs(long double __lcpp_x) _NOEXCEPT {
+  return ::fabsl(__lcpp_x);
+}
+#endif // !(defined(_AIX) || defined(__sun__))
+
+// div
+
+#undef div
+#undef ldiv
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+#undef lldiv
+#endif
+
+// MSVCRT already has the correct prototype in <stdlib.h> if __cplusplus is defined
+#if !defined(_LIBCUDACXX_MSVCRT) && !defined(__sun__) && !defined(_AIX)
+inline _LIBCUDACXX_INLINE_VISIBILITY ldiv_t div(long __x, long __y) _NOEXCEPT {
+  return ::ldiv(__x, __y);
+}
+#ifndef _LIBCUDACXX_HAS_NO_LONG_LONG
+inline _LIBCUDACXX_INLINE_VISIBILITY lldiv_t div(long long __x,
+                                             long long __y) _NOEXCEPT {
+  return ::lldiv(__x, __y);
+}
+#endif // _LIBCUDACXX_HAS_NO_LONG_LONG
+#endif // _LIBCUDACXX_MSVCRT / __sun__ / _AIX
+
+// acos
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       acos(float __lcpp_x) _NOEXCEPT       {return ::acosf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double acos(long double __lcpp_x) _NOEXCEPT {return ::acosl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+acos(_A1 __lcpp_x) _NOEXCEPT {return ::acos((double)__lcpp_x);}
+
+// asin
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       asin(float __lcpp_x) _NOEXCEPT       {return ::asinf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double asin(long double __lcpp_x) _NOEXCEPT {return ::asinl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+asin(_A1 __lcpp_x) _NOEXCEPT {return ::asin((double)__lcpp_x);}
+
+// atan
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       atan(float __lcpp_x) _NOEXCEPT       {return ::atanf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double atan(long double __lcpp_x) _NOEXCEPT {return ::atanl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+atan(_A1 __lcpp_x) _NOEXCEPT {return ::atan((double)__lcpp_x);}
+
+// atan2
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       atan2(float __lcpp_y, float __lcpp_x) _NOEXCEPT             {return ::atan2f(__lcpp_y, __lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double atan2(long double __lcpp_y, long double __lcpp_x) _NOEXCEPT {return ::atan2l(__lcpp_y, __lcpp_x);}
+#endif
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+atan2(_A1 __lcpp_y, _A2 __lcpp_x) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::atan2((__result_type)__lcpp_y, (__result_type)__lcpp_x);
+}
+
+// ceil
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       ceil(float __lcpp_x) _NOEXCEPT       {return ::ceilf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double ceil(long double __lcpp_x) _NOEXCEPT {return ::ceill(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+ceil(_A1 __lcpp_x) _NOEXCEPT {return ::ceil((double)__lcpp_x);}
+
+// cos
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       cos(float __lcpp_x) _NOEXCEPT       {return ::cosf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double cos(long double __lcpp_x) _NOEXCEPT {return ::cosl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+cos(_A1 __lcpp_x) _NOEXCEPT {return ::cos((double)__lcpp_x);}
+
+// cosh
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       cosh(float __lcpp_x) _NOEXCEPT       {return ::coshf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double cosh(long double __lcpp_x) _NOEXCEPT {return ::coshl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+cosh(_A1 __lcpp_x) _NOEXCEPT {return ::cosh((double)__lcpp_x);}
+
+// exp
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       exp(float __lcpp_x) _NOEXCEPT       {return ::expf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double exp(long double __lcpp_x) _NOEXCEPT {return ::expl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+exp(_A1 __lcpp_x) _NOEXCEPT {return ::exp((double)__lcpp_x);}
+
+// fabs
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       fabs(float __lcpp_x) _NOEXCEPT       {return ::fabsf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double fabs(long double __lcpp_x) _NOEXCEPT {return ::fabsl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+fabs(_A1 __lcpp_x) _NOEXCEPT {return ::fabs((double)__lcpp_x);}
+
+// floor
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       floor(float __lcpp_x) _NOEXCEPT       {return ::floorf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double floor(long double __lcpp_x) _NOEXCEPT {return ::floorl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+floor(_A1 __lcpp_x) _NOEXCEPT {return ::floor((double)__lcpp_x);}
+
+// fmod
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       fmod(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::fmodf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double fmod(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::fmodl(__lcpp_x, __lcpp_y);}
+#endif
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+fmod(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::fmod((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// frexp
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       frexp(float __lcpp_x, int* __lcpp_e) _NOEXCEPT       {return ::frexpf(__lcpp_x, __lcpp_e);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double frexp(long double __lcpp_x, int* __lcpp_e) _NOEXCEPT {return ::frexpl(__lcpp_x, __lcpp_e);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+frexp(_A1 __lcpp_x, int* __lcpp_e) _NOEXCEPT {return ::frexp((double)__lcpp_x, __lcpp_e);}
+
+// ldexp
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       ldexp(float __lcpp_x, int __lcpp_e) _NOEXCEPT       {return ::ldexpf(__lcpp_x, __lcpp_e);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double ldexp(long double __lcpp_x, int __lcpp_e) _NOEXCEPT {return ::ldexpl(__lcpp_x, __lcpp_e);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+ldexp(_A1 __lcpp_x, int __lcpp_e) _NOEXCEPT {return ::ldexp((double)__lcpp_x, __lcpp_e);}
+
+// log
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       log(float __lcpp_x) _NOEXCEPT       {return ::logf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double log(long double __lcpp_x) _NOEXCEPT {return ::logl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+log(_A1 __lcpp_x) _NOEXCEPT {return ::log((double)__lcpp_x);}
+
+// log10
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       log10(float __lcpp_x) _NOEXCEPT       {return ::log10f(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double log10(long double __lcpp_x) _NOEXCEPT {return ::log10l(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+log10(_A1 __lcpp_x) _NOEXCEPT {return ::log10((double)__lcpp_x);}
+
+// modf
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       modf(float __lcpp_x, float* __lcpp_y) _NOEXCEPT             {return ::modff(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double modf(long double __lcpp_x, long double* __lcpp_y) _NOEXCEPT {return ::modfl(__lcpp_x, __lcpp_y);}
+#endif
+
+// pow
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       pow(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::powf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double pow(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::powl(__lcpp_x, __lcpp_y);}
+#endif
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+pow(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::pow((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// sin
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       sin(float __lcpp_x) _NOEXCEPT       {return ::sinf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double sin(long double __lcpp_x) _NOEXCEPT {return ::sinl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+sin(_A1 __lcpp_x) _NOEXCEPT {return ::sin((double)__lcpp_x);}
+
+// sinh
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       sinh(float __lcpp_x) _NOEXCEPT       {return ::sinhf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double sinh(long double __lcpp_x) _NOEXCEPT {return ::sinhl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+sinh(_A1 __lcpp_x) _NOEXCEPT {return ::sinh((double)__lcpp_x);}
+
+// sqrt
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       sqrt(float __lcpp_x) _NOEXCEPT       {return ::sqrtf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double sqrt(long double __lcpp_x) _NOEXCEPT {return ::sqrtl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+sqrt(_A1 __lcpp_x) _NOEXCEPT {return ::sqrt((double)__lcpp_x);}
+
+// tan
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       tan(float __lcpp_x) _NOEXCEPT       {return ::tanf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double tan(long double __lcpp_x) _NOEXCEPT {return ::tanl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+tan(_A1 __lcpp_x) _NOEXCEPT {return ::tan((double)__lcpp_x);}
+
+// tanh
+
+#if !(defined(_AIX) || defined(__sun__))
+inline _LIBCUDACXX_INLINE_VISIBILITY float       tanh(float __lcpp_x) _NOEXCEPT       {return ::tanhf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double tanh(long double __lcpp_x) _NOEXCEPT {return ::tanhl(__lcpp_x);}
+#endif
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+tanh(_A1 __lcpp_x) _NOEXCEPT {return ::tanh((double)__lcpp_x);}
+
+// acosh
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       acosh(float __lcpp_x) _NOEXCEPT       {return ::acoshf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double acosh(long double __lcpp_x) _NOEXCEPT {return ::acoshl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+acosh(_A1 __lcpp_x) _NOEXCEPT {return ::acosh((double)__lcpp_x);}
+
+// asinh
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       asinh(float __lcpp_x) _NOEXCEPT       {return ::asinhf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double asinh(long double __lcpp_x) _NOEXCEPT {return ::asinhl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+asinh(_A1 __lcpp_x) _NOEXCEPT {return ::asinh((double)__lcpp_x);}
+
+// atanh
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       atanh(float __lcpp_x) _NOEXCEPT       {return ::atanhf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double atanh(long double __lcpp_x) _NOEXCEPT {return ::atanhl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+atanh(_A1 __lcpp_x) _NOEXCEPT {return ::atanh((double)__lcpp_x);}
+
+// cbrt
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       cbrt(float __lcpp_x) _NOEXCEPT       {return ::cbrtf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double cbrt(long double __lcpp_x) _NOEXCEPT {return ::cbrtl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+cbrt(_A1 __lcpp_x) _NOEXCEPT {return ::cbrt((double)__lcpp_x);}
+
+// copysign
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float copysign(float __lcpp_x,
+                                                float __lcpp_y) _NOEXCEPT {
+  return ::copysignf(__lcpp_x, __lcpp_y);
+}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double
+copysign(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {
+  return ::copysignl(__lcpp_x, __lcpp_y);
+}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+copysign(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::copysign((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// erf
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       erf(float __lcpp_x) _NOEXCEPT       {return ::erff(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double erf(long double __lcpp_x) _NOEXCEPT {return ::erfl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+erf(_A1 __lcpp_x) _NOEXCEPT {return ::erf((double)__lcpp_x);}
+
+// erfc
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       erfc(float __lcpp_x) _NOEXCEPT       {return ::erfcf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double erfc(long double __lcpp_x) _NOEXCEPT {return ::erfcl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+erfc(_A1 __lcpp_x) _NOEXCEPT {return ::erfc((double)__lcpp_x);}
+
+// exp2
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       exp2(float __lcpp_x) _NOEXCEPT       {return ::exp2f(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double exp2(long double __lcpp_x) _NOEXCEPT {return ::exp2l(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+exp2(_A1 __lcpp_x) _NOEXCEPT {return ::exp2((double)__lcpp_x);}
+
+// expm1
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       expm1(float __lcpp_x) _NOEXCEPT       {return ::expm1f(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double expm1(long double __lcpp_x) _NOEXCEPT {return ::expm1l(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+expm1(_A1 __lcpp_x) _NOEXCEPT {return ::expm1((double)__lcpp_x);}
+
+// fdim
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       fdim(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::fdimf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double fdim(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::fdiml(__lcpp_x, __lcpp_y);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+fdim(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::fdim((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// fma
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       fma(float __lcpp_x, float __lcpp_y, float __lcpp_z) _NOEXCEPT                   {return ::fmaf(__lcpp_x, __lcpp_y, __lcpp_z);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double fma(long double __lcpp_x, long double __lcpp_y, long double __lcpp_z) _NOEXCEPT {return ::fmal(__lcpp_x, __lcpp_y, __lcpp_z);}
+
+template <class _A1, class _A2, class _A3>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value &&
+    std::is_arithmetic<_A3>::value,
+    std::__promote<_A1, _A2, _A3>
+>::type
+fma(_A1 __lcpp_x, _A2 __lcpp_y, _A3 __lcpp_z) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2, _A3>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value &&
+                     std::_IsSame<_A3, __result_type>::value)), "");
+    return ::fma((__result_type)__lcpp_x, (__result_type)__lcpp_y, (__result_type)__lcpp_z);
+}
+
+// fmax
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       fmax(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::fmaxf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double fmax(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::fmaxl(__lcpp_x, __lcpp_y);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+fmax(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::fmax((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// fmin
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       fmin(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::fminf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double fmin(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::fminl(__lcpp_x, __lcpp_y);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+fmin(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::fmin((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// hypot
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       hypot(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::hypotf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double hypot(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::hypotl(__lcpp_x, __lcpp_y);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+hypot(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::hypot((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// ilogb
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int ilogb(float __lcpp_x) _NOEXCEPT       {return ::ilogbf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY int ilogb(long double __lcpp_x) _NOEXCEPT {return ::ilogbl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, int>::type
+ilogb(_A1 __lcpp_x) _NOEXCEPT {return ::ilogb((double)__lcpp_x);}
+
+// lgamma
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       lgamma(float __lcpp_x) _NOEXCEPT       {return ::lgammaf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double lgamma(long double __lcpp_x) _NOEXCEPT {return ::lgammal(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+lgamma(_A1 __lcpp_x) _NOEXCEPT {return ::lgamma((double)__lcpp_x);}
+
+// llrint
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long long llrint(float __lcpp_x) _NOEXCEPT       {return ::llrintf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long long llrint(long double __lcpp_x) _NOEXCEPT {return ::llrintl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, long long>::type
+llrint(_A1 __lcpp_x) _NOEXCEPT {return ::llrint((double)__lcpp_x);}
+
+// llround
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long long llround(float __lcpp_x) _NOEXCEPT       {return ::llroundf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long long llround(long double __lcpp_x) _NOEXCEPT {return ::llroundl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, long long>::type
+llround(_A1 __lcpp_x) _NOEXCEPT {return ::llround((double)__lcpp_x);}
+
+// log1p
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       log1p(float __lcpp_x) _NOEXCEPT       {return ::log1pf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double log1p(long double __lcpp_x) _NOEXCEPT {return ::log1pl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+log1p(_A1 __lcpp_x) _NOEXCEPT {return ::log1p((double)__lcpp_x);}
+
+// log2
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       log2(float __lcpp_x) _NOEXCEPT       {return ::log2f(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double log2(long double __lcpp_x) _NOEXCEPT {return ::log2l(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+log2(_A1 __lcpp_x) _NOEXCEPT {return ::log2((double)__lcpp_x);}
+
+// logb
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       logb(float __lcpp_x) _NOEXCEPT       {return ::logbf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double logb(long double __lcpp_x) _NOEXCEPT {return ::logbl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+logb(_A1 __lcpp_x) _NOEXCEPT {return ::logb((double)__lcpp_x);}
+
+// lrint
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long lrint(float __lcpp_x) _NOEXCEPT       {return ::lrintf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long lrint(long double __lcpp_x) _NOEXCEPT {return ::lrintl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, long>::type
+lrint(_A1 __lcpp_x) _NOEXCEPT {return ::lrint((double)__lcpp_x);}
+
+// lround
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long lround(float __lcpp_x) _NOEXCEPT       {return ::lroundf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long lround(long double __lcpp_x) _NOEXCEPT {return ::lroundl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, long>::type
+lround(_A1 __lcpp_x) _NOEXCEPT {return ::lround((double)__lcpp_x);}
+
+// nan
+
+// nearbyint
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       nearbyint(float __lcpp_x) _NOEXCEPT       {return ::nearbyintf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double nearbyint(long double __lcpp_x) _NOEXCEPT {return ::nearbyintl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+nearbyint(_A1 __lcpp_x) _NOEXCEPT {return ::nearbyint((double)__lcpp_x);}
+
+// nextafter
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       nextafter(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::nextafterf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double nextafter(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::nextafterl(__lcpp_x, __lcpp_y);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+nextafter(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::nextafter((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// nexttoward
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       nexttoward(float __lcpp_x, long double __lcpp_y) _NOEXCEPT       {return ::nexttowardf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double nexttoward(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::nexttowardl(__lcpp_x, __lcpp_y);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+nexttoward(_A1 __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::nexttoward((double)__lcpp_x, __lcpp_y);}
+
+// remainder
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       remainder(float __lcpp_x, float __lcpp_y) _NOEXCEPT             {return ::remainderf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double remainder(long double __lcpp_x, long double __lcpp_y) _NOEXCEPT {return ::remainderl(__lcpp_x, __lcpp_y);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+remainder(_A1 __lcpp_x, _A2 __lcpp_y) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::remainder((__result_type)__lcpp_x, (__result_type)__lcpp_y);
+}
+
+// remquo
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       remquo(float __lcpp_x, float __lcpp_y, int* __lcpp_z) _NOEXCEPT             {return ::remquof(__lcpp_x, __lcpp_y, __lcpp_z);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double remquo(long double __lcpp_x, long double __lcpp_y, int* __lcpp_z) _NOEXCEPT {return ::remquol(__lcpp_x, __lcpp_y, __lcpp_z);}
+
+template <class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::_EnableIf
+<
+    std::is_arithmetic<_A1>::value &&
+    std::is_arithmetic<_A2>::value,
+    std::__promote<_A1, _A2>
+>::type
+remquo(_A1 __lcpp_x, _A2 __lcpp_y, int* __lcpp_z) _NOEXCEPT
+{
+    typedef typename std::__promote<_A1, _A2>::type __result_type;
+    static_assert((!(std::_IsSame<_A1, __result_type>::value &&
+                     std::_IsSame<_A2, __result_type>::value)), "");
+    return ::remquo((__result_type)__lcpp_x, (__result_type)__lcpp_y, __lcpp_z);
+}
+
+// rint
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       rint(float __lcpp_x) _NOEXCEPT       {return ::rintf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double rint(long double __lcpp_x) _NOEXCEPT {return ::rintl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+rint(_A1 __lcpp_x) _NOEXCEPT {return ::rint((double)__lcpp_x);}
+
+// round
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       round(float __lcpp_x) _NOEXCEPT       {return ::roundf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double round(long double __lcpp_x) _NOEXCEPT {return ::roundl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+round(_A1 __lcpp_x) _NOEXCEPT {return ::round((double)__lcpp_x);}
+
+// scalbln
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       scalbln(float __lcpp_x, long __lcpp_y) _NOEXCEPT       {return ::scalblnf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double scalbln(long double __lcpp_x, long __lcpp_y) _NOEXCEPT {return ::scalblnl(__lcpp_x, __lcpp_y);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+scalbln(_A1 __lcpp_x, long __lcpp_y) _NOEXCEPT {return ::scalbln((double)__lcpp_x, __lcpp_y);}
+
+// scalbn
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       scalbn(float __lcpp_x, int __lcpp_y) _NOEXCEPT       {return ::scalbnf(__lcpp_x, __lcpp_y);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double scalbn(long double __lcpp_x, int __lcpp_y) _NOEXCEPT {return ::scalbnl(__lcpp_x, __lcpp_y);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+scalbn(_A1 __lcpp_x, int __lcpp_y) _NOEXCEPT {return ::scalbn((double)__lcpp_x, __lcpp_y);}
+
+// tgamma
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       tgamma(float __lcpp_x) _NOEXCEPT       {return ::tgammaf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double tgamma(long double __lcpp_x) _NOEXCEPT {return ::tgammal(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+tgamma(_A1 __lcpp_x) _NOEXCEPT {return ::tgamma((double)__lcpp_x);}
+
+// trunc
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float       trunc(float __lcpp_x) _NOEXCEPT       {return ::truncf(__lcpp_x);}
+inline _LIBCUDACXX_INLINE_VISIBILITY long double trunc(long double __lcpp_x) _NOEXCEPT {return ::truncl(__lcpp_x);}
+
+template <class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename std::enable_if<std::is_integral<_A1>::value, double>::type
+trunc(_A1 __lcpp_x) _NOEXCEPT {return ::trunc((double)__lcpp_x);}
+
+} // extern "C++"
+
+#endif // __cplusplus
+
+#else // _LIBCUDACXX_MATH_H
+
+// This include lives outside the header guard in order to support an MSVC
+// extension which allows users to do:
+//
+// #define _USE_MATH_DEFINES
+// #include <math.h>
+//
+// and receive the definitions of mathematical constants, even if <math.h>
+// has previously been included.
+#if defined(_LIBCUDACXX_MSVCRT) && defined(_USE_MATH_DEFINES)
+#include_next <math.h>
+#endif
+
+#endif  // _LIBCUDACXX_MATH_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/memory b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/memory
new file mode 100644
index 000000000000..c029fac6d9cc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/memory
@@ -0,0 +1,5443 @@
+// -*- C++ -*-
+//===-------------------------- memory ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_MEMORY
+#define _LIBCUDACXX_MEMORY
+
+/*
+    memory synopsis
+
+namespace std
+{
+
+struct allocator_arg_t { };
+inline constexpr allocator_arg_t allocator_arg = allocator_arg_t();
+
+template <class T, class Alloc> struct uses_allocator;
+
+template <class Ptr>
+struct pointer_traits
+{
+    typedef Ptr pointer;
+    typedef <details> element_type;
+    typedef <details> difference_type;
+
+    template <class U> using rebind = <details>;
+
+    static pointer pointer_to(<details>);
+};
+
+template <class T>
+struct pointer_traits<T*>
+{
+    typedef T* pointer;
+    typedef T element_type;
+    typedef ptrdiff_t difference_type;
+
+    template <class U> using rebind = U*;
+
+    static pointer pointer_to(<details>) noexcept; // constexpr in C++20
+};
+
+template <class T> constexpr T* to_address(T* p) noexcept; // C++20
+template <class Ptr> auto to_address(const Ptr& p) noexcept; // C++20
+
+template <class Alloc>
+struct allocator_traits
+{
+    typedef Alloc                        allocator_type;
+    typedef typename allocator_type::value_type
+                                         value_type;
+
+    typedef Alloc::pointer | value_type* pointer;
+    typedef Alloc::const_pointer
+          | pointer_traits<pointer>::rebind<const value_type>
+                                         const_pointer;
+    typedef Alloc::void_pointer
+          | pointer_traits<pointer>::rebind<void>
+                                         void_pointer;
+    typedef Alloc::const_void_pointer
+          | pointer_traits<pointer>::rebind<const void>
+                                         const_void_pointer;
+    typedef Alloc::difference_type
+          | pointer_traits<pointer>::difference_type
+                                         difference_type;
+    typedef Alloc::size_type
+          | make_unsigned<difference_type>::type
+                                         size_type;
+    typedef Alloc::propagate_on_container_copy_assignment
+          | false_type                   propagate_on_container_copy_assignment;
+    typedef Alloc::propagate_on_container_move_assignment
+          | false_type                   propagate_on_container_move_assignment;
+    typedef Alloc::propagate_on_container_swap
+          | false_type                   propagate_on_container_swap;
+    typedef Alloc::is_always_equal
+          | is_empty                     is_always_equal;
+
+    template <class T> using rebind_alloc  = Alloc::rebind<U>::other | Alloc<T, Args...>;
+    template <class T> using rebind_traits = allocator_traits<rebind_alloc<T>>;
+
+    static pointer allocate(allocator_type& a, size_type n);                          // [[nodiscard]] in C++20
+    static pointer allocate(allocator_type& a, size_type n, const_void_pointer hint); // [[nodiscard]] in C++20
+
+    static void deallocate(allocator_type& a, pointer p, size_type n) noexcept;
+
+    template <class T, class... Args>
+        static void construct(allocator_type& a, T* p, Args&&... args);
+
+    template <class T>
+        static void destroy(allocator_type& a, T* p);
+
+    static size_type max_size(const allocator_type& a); // noexcept in C++14
+
+    static allocator_type
+        select_on_container_copy_construction(const allocator_type& a);
+};
+
+template <>
+class allocator<void>
+{
+public:
+    typedef void*                                 pointer;
+    typedef const void*                           const_pointer;
+    typedef void                                  value_type;
+
+    template <class _Up> struct rebind {typedef allocator<_Up> other;};
+};
+
+template <class T>
+class allocator
+{
+public:
+    typedef size_t                                size_type;
+    typedef ptrdiff_t                             difference_type;
+    typedef T*                                    pointer;
+    typedef const T*                              const_pointer;
+    typedef typename add_lvalue_reference<T>::type       reference;
+    typedef typename add_lvalue_reference<const T>::type const_reference;
+    typedef T                                     value_type;
+
+    template <class U> struct rebind {typedef allocator<U> other;};
+
+    constexpr allocator() noexcept;                      // constexpr in C++20
+    constexpr allocator(const allocator&) noexcept;      // constexpr in C++20
+    template <class U>
+      constexpr allocator(const allocator<U>&) noexcept; // constexpr in C++20
+    ~allocator();
+    pointer address(reference x) const noexcept;
+    const_pointer address(const_reference x) const noexcept;
+    pointer allocate(size_type, allocator<void>::const_pointer hint = 0);
+    void deallocate(pointer p, size_type n) noexcept;
+    size_type max_size() const noexcept;
+    template<class U, class... Args>
+        void construct(U* p, Args&&... args);
+    template <class U>
+        void destroy(U* p);
+};
+
+template <class T, class U>
+bool operator==(const allocator<T>&, const allocator<U>&) noexcept;
+
+template <class T, class U>
+bool operator!=(const allocator<T>&, const allocator<U>&) noexcept;
+
+template <class OutputIterator, class T>
+class raw_storage_iterator
+    : public iterator<output_iterator_tag,
+                      T,                               // purposefully not C++03
+                      ptrdiff_t,                       // purposefully not C++03
+                      T*,                              // purposefully not C++03
+                      raw_storage_iterator&>           // purposefully not C++03
+{
+public:
+    explicit raw_storage_iterator(OutputIterator x);
+    raw_storage_iterator& operator*();
+    raw_storage_iterator& operator=(const T& element);
+    raw_storage_iterator& operator++();
+    raw_storage_iterator  operator++(int);
+};
+
+template <class T> pair<T*,ptrdiff_t> get_temporary_buffer(ptrdiff_t n) noexcept;
+template <class T> void               return_temporary_buffer(T* p) noexcept;
+
+template <class T> T* addressof(T& r) noexcept;
+template <class T> T* addressof(const T&& r) noexcept = delete;
+
+template <class InputIterator, class ForwardIterator>
+ForwardIterator
+uninitialized_copy(InputIterator first, InputIterator last, ForwardIterator result);
+
+template <class InputIterator, class Size, class ForwardIterator>
+ForwardIterator
+uninitialized_copy_n(InputIterator first, Size n, ForwardIterator result);
+
+template <class ForwardIterator, class T>
+void uninitialized_fill(ForwardIterator first, ForwardIterator last, const T& x);
+
+template <class ForwardIterator, class Size, class T>
+ForwardIterator
+uninitialized_fill_n(ForwardIterator first, Size n, const T& x);
+
+template <class T>
+void destroy_at(T* location);
+
+template <class ForwardIterator>
+ void destroy(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class Size>
+ ForwardIterator destroy_n(ForwardIterator first, Size n);
+
+template <class InputIterator, class ForwardIterator>
+ ForwardIterator uninitialized_move(InputIterator first, InputIterator last, ForwardIterator result);
+
+template <class InputIterator, class Size, class ForwardIterator>
+ pair<InputIterator,ForwardIterator> uninitialized_move_n(InputIterator first, Size n, ForwardIterator result);
+
+template <class ForwardIterator>
+ void uninitialized_value_construct(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class Size>
+ ForwardIterator uninitialized_value_construct_n(ForwardIterator first, Size n);
+
+template <class ForwardIterator>
+ void uninitialized_default_construct(ForwardIterator first, ForwardIterator last);
+
+template <class ForwardIterator, class Size>
+ ForwardIterator uninitialized_default_construct_n(ForwardIterator first, Size n);
+
+template <class Y> struct auto_ptr_ref {};      // deprecated in C++11, removed in C++17
+
+template<class X>
+class auto_ptr                                  // deprecated in C++11, removed in C++17
+{
+public:
+    typedef X element_type;
+
+    explicit auto_ptr(X* p =0) throw();
+    auto_ptr(auto_ptr&) throw();
+    template<class Y> auto_ptr(auto_ptr<Y>&) throw();
+    auto_ptr& operator=(auto_ptr&) throw();
+    template<class Y> auto_ptr& operator=(auto_ptr<Y>&) throw();
+    auto_ptr& operator=(auto_ptr_ref<X> r) throw();
+    ~auto_ptr() throw();
+
+    typename add_lvalue_reference<X>::type operator*() const throw();
+    X* operator->() const throw();
+    X* get() const throw();
+    X* release() throw();
+    void reset(X* p =0) throw();
+
+    auto_ptr(auto_ptr_ref<X>) throw();
+    template<class Y> operator auto_ptr_ref<Y>() throw();
+    template<class Y> operator auto_ptr<Y>() throw();
+};
+
+template <class T>
+struct default_delete
+{
+    constexpr default_delete() noexcept = default;
+    template <class U> default_delete(const default_delete<U>&) noexcept;
+
+    void operator()(T*) const noexcept;
+};
+
+template <class T>
+struct default_delete<T[]>
+{
+    constexpr default_delete() noexcept = default;
+    void operator()(T*) const noexcept;
+    template <class U> void operator()(U*) const = delete;
+};
+
+template <class T, class D = default_delete<T>>
+class unique_ptr
+{
+public:
+    typedef see below pointer;
+    typedef T element_type;
+    typedef D deleter_type;
+
+    // constructors
+    constexpr unique_ptr() noexcept;
+    explicit unique_ptr(pointer p) noexcept;
+    unique_ptr(pointer p, see below d1) noexcept;
+    unique_ptr(pointer p, see below d2) noexcept;
+    unique_ptr(unique_ptr&& u) noexcept;
+    unique_ptr(nullptr_t) noexcept : unique_ptr() { }
+    template <class U, class E>
+        unique_ptr(unique_ptr<U, E>&& u) noexcept;
+    template <class U>
+        unique_ptr(auto_ptr<U>&& u) noexcept;       // removed in C++17
+
+    // destructor
+    ~unique_ptr();
+
+    // assignment
+    unique_ptr& operator=(unique_ptr&& u) noexcept;
+    template <class U, class E> unique_ptr& operator=(unique_ptr<U, E>&& u) noexcept;
+    unique_ptr& operator=(nullptr_t) noexcept;
+
+    // observers
+    typename add_lvalue_reference<T>::type operator*() const;
+    pointer operator->() const noexcept;
+    pointer get() const noexcept;
+    deleter_type& get_deleter() noexcept;
+    const deleter_type& get_deleter() const noexcept;
+    explicit operator bool() const noexcept;
+
+    // modifiers
+    pointer release() noexcept;
+    void reset(pointer p = pointer()) noexcept;
+    void swap(unique_ptr& u) noexcept;
+};
+
+template <class T, class D>
+class unique_ptr<T[], D>
+{
+public:
+    typedef implementation-defined pointer;
+    typedef T element_type;
+    typedef D deleter_type;
+
+    // constructors
+    constexpr unique_ptr() noexcept;
+    explicit unique_ptr(pointer p) noexcept;
+    unique_ptr(pointer p, see below d) noexcept;
+    unique_ptr(pointer p, see below d) noexcept;
+    unique_ptr(unique_ptr&& u) noexcept;
+    unique_ptr(nullptr_t) noexcept : unique_ptr() { }
+
+    // destructor
+    ~unique_ptr();
+
+    // assignment
+    unique_ptr& operator=(unique_ptr&& u) noexcept;
+    unique_ptr& operator=(nullptr_t) noexcept;
+
+    // observers
+    T& operator[](size_t i) const;
+    pointer get() const noexcept;
+    deleter_type& get_deleter() noexcept;
+    const deleter_type& get_deleter() const noexcept;
+    explicit operator bool() const noexcept;
+
+    // modifiers
+    pointer release() noexcept;
+    void reset(pointer p = pointer()) noexcept;
+    void reset(nullptr_t) noexcept;
+    template <class U> void reset(U) = delete;
+    void swap(unique_ptr& u) noexcept;
+};
+
+template <class T, class D>
+    void swap(unique_ptr<T, D>& x, unique_ptr<T, D>& y) noexcept;
+
+template <class T1, class D1, class T2, class D2>
+    bool operator==(const unique_ptr<T1, D1>& x, const unique_ptr<T2, D2>& y);
+template <class T1, class D1, class T2, class D2>
+    bool operator!=(const unique_ptr<T1, D1>& x, const unique_ptr<T2, D2>& y);
+template <class T1, class D1, class T2, class D2>
+    bool operator<(const unique_ptr<T1, D1>& x, const unique_ptr<T2, D2>& y);
+template <class T1, class D1, class T2, class D2>
+    bool operator<=(const unique_ptr<T1, D1>& x, const unique_ptr<T2, D2>& y);
+template <class T1, class D1, class T2, class D2>
+    bool operator>(const unique_ptr<T1, D1>& x, const unique_ptr<T2, D2>& y);
+template <class T1, class D1, class T2, class D2>
+    bool operator>=(const unique_ptr<T1, D1>& x, const unique_ptr<T2, D2>& y);
+
+template <class T, class D>
+    bool operator==(const unique_ptr<T, D>& x, nullptr_t) noexcept;
+template <class T, class D>
+    bool operator==(nullptr_t, const unique_ptr<T, D>& y) noexcept;
+template <class T, class D>
+    bool operator!=(const unique_ptr<T, D>& x, nullptr_t) noexcept;
+template <class T, class D>
+    bool operator!=(nullptr_t, const unique_ptr<T, D>& y) noexcept;
+
+template <class T, class D>
+    bool operator<(const unique_ptr<T, D>& x, nullptr_t);
+template <class T, class D>
+    bool operator<(nullptr_t, const unique_ptr<T, D>& y);
+template <class T, class D>
+    bool operator<=(const unique_ptr<T, D>& x, nullptr_t);
+template <class T, class D>
+    bool operator<=(nullptr_t, const unique_ptr<T, D>& y);
+template <class T, class D>
+    bool operator>(const unique_ptr<T, D>& x, nullptr_t);
+template <class T, class D>
+    bool operator>(nullptr_t, const unique_ptr<T, D>& y);
+template <class T, class D>
+    bool operator>=(const unique_ptr<T, D>& x, nullptr_t);
+template <class T, class D>
+    bool operator>=(nullptr_t, const unique_ptr<T, D>& y);
+
+class bad_weak_ptr
+    : public std::exception
+{
+    bad_weak_ptr() noexcept;
+};
+
+template<class T, class... Args> unique_ptr<T> make_unique(Args&&... args);     // C++14
+template<class T>                unique_ptr<T> make_unique(size_t n);           // C++14
+template<class T, class... Args> unspecified   make_unique(Args&&...) = delete; // C++14, T == U[N]
+
+template<class E, class T, class Y, class D>
+    basic_ostream<E, T>& operator<< (basic_ostream<E, T>& os, unique_ptr<Y, D> const& p);
+
+template<class T>
+class shared_ptr
+{
+public:
+    typedef T element_type;
+    typedef weak_ptr<T> weak_type; // C++17
+
+    // constructors:
+    constexpr shared_ptr() noexcept;
+    template<class Y> explicit shared_ptr(Y* p);
+    template<class Y, class D> shared_ptr(Y* p, D d);
+    template<class Y, class D, class A> shared_ptr(Y* p, D d, A a);
+    template <class D> shared_ptr(nullptr_t p, D d);
+    template <class D, class A> shared_ptr(nullptr_t p, D d, A a);
+    template<class Y> shared_ptr(const shared_ptr<Y>& r, T *p) noexcept;
+    shared_ptr(const shared_ptr& r) noexcept;
+    template<class Y> shared_ptr(const shared_ptr<Y>& r) noexcept;
+    shared_ptr(shared_ptr&& r) noexcept;
+    template<class Y> shared_ptr(shared_ptr<Y>&& r) noexcept;
+    template<class Y> explicit shared_ptr(const weak_ptr<Y>& r);
+    template<class Y> shared_ptr(auto_ptr<Y>&& r);          // removed in C++17
+    template <class Y, class D> shared_ptr(unique_ptr<Y, D>&& r);
+    shared_ptr(nullptr_t) : shared_ptr() { }
+
+    // destructor:
+    ~shared_ptr();
+
+    // assignment:
+    shared_ptr& operator=(const shared_ptr& r) noexcept;
+    template<class Y> shared_ptr& operator=(const shared_ptr<Y>& r) noexcept;
+    shared_ptr& operator=(shared_ptr&& r) noexcept;
+    template<class Y> shared_ptr& operator=(shared_ptr<Y>&& r);
+    template<class Y> shared_ptr& operator=(auto_ptr<Y>&& r); // removed in C++17
+    template <class Y, class D> shared_ptr& operator=(unique_ptr<Y, D>&& r);
+
+    // modifiers:
+    void swap(shared_ptr& r) noexcept;
+    void reset() noexcept;
+    template<class Y> void reset(Y* p);
+    template<class Y, class D> void reset(Y* p, D d);
+    template<class Y, class D, class A> void reset(Y* p, D d, A a);
+
+    // observers:
+    T* get() const noexcept;
+    T& operator*() const noexcept;
+    T* operator->() const noexcept;
+    long use_count() const noexcept;
+    bool unique() const noexcept;
+    explicit operator bool() const noexcept;
+    template<class U> bool owner_before(shared_ptr<U> const& b) const noexcept;
+    template<class U> bool owner_before(weak_ptr<U> const& b) const noexcept;
+};
+
+// shared_ptr comparisons:
+template<class T, class U>
+    bool operator==(shared_ptr<T> const& a, shared_ptr<U> const& b) noexcept;
+template<class T, class U>
+    bool operator!=(shared_ptr<T> const& a, shared_ptr<U> const& b) noexcept;
+template<class T, class U>
+    bool operator<(shared_ptr<T> const& a, shared_ptr<U> const& b) noexcept;
+template<class T, class U>
+    bool operator>(shared_ptr<T> const& a, shared_ptr<U> const& b) noexcept;
+template<class T, class U>
+    bool operator<=(shared_ptr<T> const& a, shared_ptr<U> const& b) noexcept;
+template<class T, class U>
+    bool operator>=(shared_ptr<T> const& a, shared_ptr<U> const& b) noexcept;
+
+template <class T>
+    bool operator==(const shared_ptr<T>& x, nullptr_t) noexcept;
+template <class T>
+    bool operator==(nullptr_t, const shared_ptr<T>& y) noexcept;
+template <class T>
+    bool operator!=(const shared_ptr<T>& x, nullptr_t) noexcept;
+template <class T>
+    bool operator!=(nullptr_t, const shared_ptr<T>& y) noexcept;
+template <class T>
+    bool operator<(const shared_ptr<T>& x, nullptr_t) noexcept;
+template <class T>
+bool operator<(nullptr_t, const shared_ptr<T>& y) noexcept;
+template <class T>
+    bool operator<=(const shared_ptr<T>& x, nullptr_t) noexcept;
+template <class T>
+    bool operator<=(nullptr_t, const shared_ptr<T>& y) noexcept;
+template <class T>
+    bool operator>(const shared_ptr<T>& x, nullptr_t) noexcept;
+template <class T>
+    bool operator>(nullptr_t, const shared_ptr<T>& y) noexcept;
+template <class T>
+    bool operator>=(const shared_ptr<T>& x, nullptr_t) noexcept;
+template <class T>
+    bool operator>=(nullptr_t, const shared_ptr<T>& y) noexcept;
+
+// shared_ptr specialized algorithms:
+template<class T> void swap(shared_ptr<T>& a, shared_ptr<T>& b) noexcept;
+
+// shared_ptr casts:
+template<class T, class U>
+    shared_ptr<T> static_pointer_cast(shared_ptr<U> const& r) noexcept;
+template<class T, class U>
+    shared_ptr<T> dynamic_pointer_cast(shared_ptr<U> const& r) noexcept;
+template<class T, class U>
+    shared_ptr<T> const_pointer_cast(shared_ptr<U> const& r) noexcept;
+
+// shared_ptr I/O:
+template<class E, class T, class Y>
+    basic_ostream<E, T>& operator<< (basic_ostream<E, T>& os, shared_ptr<Y> const& p);
+
+// shared_ptr get_deleter:
+template<class D, class T> D* get_deleter(shared_ptr<T> const& p) noexcept;
+
+template<class T, class... Args>
+    shared_ptr<T> make_shared(Args&&... args);
+template<class T, class A, class... Args>
+    shared_ptr<T> allocate_shared(const A& a, Args&&... args);
+
+template<class T>
+class weak_ptr
+{
+public:
+    typedef T element_type;
+
+    // constructors
+    constexpr weak_ptr() noexcept;
+    template<class Y> weak_ptr(shared_ptr<Y> const& r) noexcept;
+    weak_ptr(weak_ptr const& r) noexcept;
+    template<class Y> weak_ptr(weak_ptr<Y> const& r) noexcept;
+    weak_ptr(weak_ptr&& r) noexcept;                      // C++14
+    template<class Y> weak_ptr(weak_ptr<Y>&& r) noexcept; // C++14
+
+    // destructor
+    ~weak_ptr();
+
+    // assignment
+    weak_ptr& operator=(weak_ptr const& r) noexcept;
+    template<class Y> weak_ptr& operator=(weak_ptr<Y> const& r) noexcept;
+    template<class Y> weak_ptr& operator=(shared_ptr<Y> const& r) noexcept;
+    weak_ptr& operator=(weak_ptr&& r) noexcept;                      // C++14
+    template<class Y> weak_ptr& operator=(weak_ptr<Y>&& r) noexcept; // C++14
+
+    // modifiers
+    void swap(weak_ptr& r) noexcept;
+    void reset() noexcept;
+
+    // observers
+    long use_count() const noexcept;
+    bool expired() const noexcept;
+    shared_ptr<T> lock() const noexcept;
+    template<class U> bool owner_before(shared_ptr<U> const& b) const noexcept;
+    template<class U> bool owner_before(weak_ptr<U> const& b) const noexcept;
+};
+
+// weak_ptr specialized algorithms:
+template<class T> void swap(weak_ptr<T>& a, weak_ptr<T>& b) noexcept;
+
+// class owner_less:
+template<class T> struct owner_less;
+
+template<class T>
+struct owner_less<shared_ptr<T> >
+    : binary_function<shared_ptr<T>, shared_ptr<T>, bool>
+{
+    typedef bool result_type;
+    bool operator()(shared_ptr<T> const&, shared_ptr<T> const&) const noexcept;
+    bool operator()(shared_ptr<T> const&, weak_ptr<T> const&) const noexcept;
+    bool operator()(weak_ptr<T> const&, shared_ptr<T> const&) const noexcept;
+};
+
+template<class T>
+struct owner_less<weak_ptr<T> >
+    : binary_function<weak_ptr<T>, weak_ptr<T>, bool>
+{
+    typedef bool result_type;
+    bool operator()(weak_ptr<T> const&, weak_ptr<T> const&) const noexcept;
+    bool operator()(shared_ptr<T> const&, weak_ptr<T> const&) const noexcept;
+    bool operator()(weak_ptr<T> const&, shared_ptr<T> const&) const noexcept;
+};
+
+template <>  // Added in C++14
+struct owner_less<void>
+{
+    template <class _Tp, class _Up>
+    bool operator()( shared_ptr<_Tp> const& __x, shared_ptr<_Up> const& __y) const noexcept;
+    template <class _Tp, class _Up>
+    bool operator()( shared_ptr<_Tp> const& __x,   weak_ptr<_Up> const& __y) const noexcept;
+    template <class _Tp, class _Up>
+    bool operator()(   weak_ptr<_Tp> const& __x, shared_ptr<_Up> const& __y) const noexcept;
+    template <class _Tp, class _Up>
+    bool operator()(   weak_ptr<_Tp> const& __x,   weak_ptr<_Up> const& __y) const noexcept;
+
+    typedef void is_transparent;
+};
+
+template<class T>
+class enable_shared_from_this
+{
+protected:
+    constexpr enable_shared_from_this() noexcept;
+    enable_shared_from_this(enable_shared_from_this const&) noexcept;
+    enable_shared_from_this& operator=(enable_shared_from_this const&) noexcept;
+    ~enable_shared_from_this();
+public:
+    shared_ptr<T> shared_from_this();
+    shared_ptr<T const> shared_from_this() const;
+};
+
+template<class T>
+    bool atomic_is_lock_free(const shared_ptr<T>* p);
+template<class T>
+    shared_ptr<T> atomic_load(const shared_ptr<T>* p);
+template<class T>
+    shared_ptr<T> atomic_load_explicit(const shared_ptr<T>* p, memory_order mo);
+template<class T>
+    void atomic_store(shared_ptr<T>* p, shared_ptr<T> r);
+template<class T>
+    void atomic_store_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo);
+template<class T>
+    shared_ptr<T> atomic_exchange(shared_ptr<T>* p, shared_ptr<T> r);
+template<class T>
+    shared_ptr<T>
+    atomic_exchange_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo);
+template<class T>
+    bool
+    atomic_compare_exchange_weak(shared_ptr<T>* p, shared_ptr<T>* v, shared_ptr<T> w);
+template<class T>
+    bool
+    atomic_compare_exchange_strong( shared_ptr<T>* p, shared_ptr<T>* v, shared_ptr<T> w);
+template<class T>
+    bool
+    atomic_compare_exchange_weak_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
+                                          shared_ptr<T> w, memory_order success,
+                                          memory_order failure);
+template<class T>
+    bool
+    atomic_compare_exchange_strong_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
+                                            shared_ptr<T> w, memory_order success,
+                                            memory_order failure);
+// Hash support
+template <class T> struct hash;
+template <class T, class D> struct hash<unique_ptr<T, D> >;
+template <class T> struct hash<shared_ptr<T> >;
+
+template <class T, class Alloc>
+  inline constexpr bool uses_allocator_v = uses_allocator<T, Alloc>::value;
+
+// Pointer safety
+enum class pointer_safety { relaxed, preferred, strict };
+void declare_reachable(void *p);
+template <class T> T *undeclare_reachable(T *p);
+void declare_no_pointers(char *p, size_t n);
+void undeclare_no_pointers(char *p, size_t n);
+pointer_safety get_pointer_safety() noexcept;
+
+void* align(size_t alignment, size_t size, void*& ptr, size_t& space);
+
+}  // std
+
+*/
+#ifndef __cuda_std__
+#include <__config>
+#include <type_traits>
+#include <typeinfo>
+#include <cstddef>
+#include <cstdint>
+#include <new>
+#include <utility>
+#include <limits>
+#include <iterator>
+#include <__functional_base>
+#include <iosfwd>
+#include <tuple>
+#include <stdexcept>
+#include <cstring>
+#if !defined(_LIBCUDACXX_HAS_NO_ATOMIC_HEADER)
+#  include <atomic>
+#endif
+#include <version>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _ValueType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ValueType __libcpp_relaxed_load(_ValueType const* __value) {
+#if !defined(_LIBCUDACXX_HAS_NO_THREADS) && \
+    defined(__ATOMIC_RELAXED) &&        \
+    (__has_builtin(__atomic_load_n) || defined(_LIBCUDACXX_COMPILER_GCC))
+    return __atomic_load_n(__value, __ATOMIC_RELAXED);
+#else
+    return *__value;
+#endif
+}
+
+template <class _ValueType>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ValueType __libcpp_acquire_load(_ValueType const* __value) {
+#if !defined(_LIBCUDACXX_HAS_NO_THREADS) && \
+    defined(__ATOMIC_ACQUIRE) &&        \
+    (__has_builtin(__atomic_load_n) || defined(_LIBCUDACXX_COMPILER_GCC))
+    return __atomic_load_n(__value, __ATOMIC_ACQUIRE);
+#else
+    return *__value;
+#endif
+}
+
+// addressof moved to <type_traits>
+
+template <class _Tp> class allocator;
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS allocator<void>
+{
+public:
+    typedef void*             pointer;
+    typedef const void*       const_pointer;
+    typedef void              value_type;
+
+    template <class _Up> struct rebind {typedef allocator<_Up> other;};
+};
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS allocator<const void>
+{
+public:
+    typedef const void*       pointer;
+    typedef const void*       const_pointer;
+    typedef const void        value_type;
+
+    template <class _Up> struct rebind {typedef allocator<_Up> other;};
+};
+
+// pointer_traits
+
+template <class _Tp, class = void>
+struct __has_element_type : false_type {};
+
+template <class _Tp>
+struct __has_element_type<_Tp,
+              typename __void_t<typename _Tp::element_type>::type> : true_type {};
+
+template <class _Ptr, bool = __has_element_type<_Ptr>::value>
+struct __pointer_traits_element_type;
+
+template <class _Ptr>
+struct __pointer_traits_element_type<_Ptr, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Ptr::element_type type;
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <template <class, class...> class _Sp, class _Tp, class ..._Args>
+struct __pointer_traits_element_type<_Sp<_Tp, _Args...>, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Sp<_Tp, _Args...>::element_type type;
+};
+
+template <template <class, class...> class _Sp, class _Tp, class ..._Args>
+struct __pointer_traits_element_type<_Sp<_Tp, _Args...>, false>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;
+};
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <template <class> class _Sp, class _Tp>
+struct __pointer_traits_element_type<_Sp<_Tp>, true>
+{
+    typedef typename _Sp<_Tp>::element_type type;
+};
+
+template <template <class> class _Sp, class _Tp>
+struct __pointer_traits_element_type<_Sp<_Tp>, false>
+{
+    typedef _Tp type;
+};
+
+template <template <class, class> class _Sp, class _Tp, class _A0>
+struct __pointer_traits_element_type<_Sp<_Tp, _A0>, true>
+{
+    typedef typename _Sp<_Tp, _A0>::element_type type;
+};
+
+template <template <class, class> class _Sp, class _Tp, class _A0>
+struct __pointer_traits_element_type<_Sp<_Tp, _A0>, false>
+{
+    typedef _Tp type;
+};
+
+template <template <class, class, class> class _Sp, class _Tp, class _A0, class _A1>
+struct __pointer_traits_element_type<_Sp<_Tp, _A0, _A1>, true>
+{
+    typedef typename _Sp<_Tp, _A0, _A1>::element_type type;
+};
+
+template <template <class, class, class> class _Sp, class _Tp, class _A0, class _A1>
+struct __pointer_traits_element_type<_Sp<_Tp, _A0, _A1>, false>
+{
+    typedef _Tp type;
+};
+
+template <template <class, class, class, class> class _Sp, class _Tp, class _A0,
+                                                           class _A1, class _A2>
+struct __pointer_traits_element_type<_Sp<_Tp, _A0, _A1, _A2>, true>
+{
+    typedef typename _Sp<_Tp, _A0, _A1, _A2>::element_type type;
+};
+
+template <template <class, class, class, class> class _Sp, class _Tp, class _A0,
+                                                           class _A1, class _A2>
+struct __pointer_traits_element_type<_Sp<_Tp, _A0, _A1, _A2>, false>
+{
+    typedef _Tp type;
+};
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Tp, class = void>
+struct __has_difference_type : false_type {};
+
+template <class _Tp>
+struct __has_difference_type<_Tp,
+            typename __void_t<typename _Tp::difference_type>::type> : true_type {};
+
+template <class _Ptr, bool = __has_difference_type<_Ptr>::value>
+struct __pointer_traits_difference_type
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE ptrdiff_t type;
+};
+
+template <class _Ptr>
+struct __pointer_traits_difference_type<_Ptr, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Ptr::difference_type type;
+};
+
+template <class _Tp, class _Up>
+struct __has_rebind
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    template <class _Xp> static __two __test(...);
+    template <class _Xp> static char __test(typename _Xp::template rebind<_Up>* = 0);
+public:
+    static const bool value = sizeof(__test<_Tp>(0)) == 1;
+};
+
+template <class _Tp, class _Up, bool = __has_rebind<_Tp, _Up>::value>
+struct __pointer_traits_rebind
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Tp::template rebind<_Up> type;
+#else
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Tp::template rebind<_Up>::other type;
+#endif
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <template <class, class...> class _Sp, class _Tp, class ..._Args, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _Args...>, _Up, true>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Sp<_Tp, _Args...>::template rebind<_Up> type;
+#else
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Sp<_Tp, _Args...>::template rebind<_Up>::other type;
+#endif
+};
+
+template <template <class, class...> class _Sp, class _Tp, class ..._Args, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _Args...>, _Up, false>
+{
+    typedef _Sp<_Up, _Args...> type;
+};
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <template <class> class _Sp, class _Tp, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp>, _Up, true>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef typename _Sp<_Tp>::template rebind<_Up> type;
+#else
+    typedef typename _Sp<_Tp>::template rebind<_Up>::other type;
+#endif
+};
+
+template <template <class> class _Sp, class _Tp, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp>, _Up, false>
+{
+    typedef _Sp<_Up> type;
+};
+
+template <template <class, class> class _Sp, class _Tp, class _A0, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _A0>, _Up, true>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef typename _Sp<_Tp, _A0>::template rebind<_Up> type;
+#else
+    typedef typename _Sp<_Tp, _A0>::template rebind<_Up>::other type;
+#endif
+};
+
+template <template <class, class> class _Sp, class _Tp, class _A0, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _A0>, _Up, false>
+{
+    typedef _Sp<_Up, _A0> type;
+};
+
+template <template <class, class, class> class _Sp, class _Tp, class _A0,
+                                         class _A1, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _A0, _A1>, _Up, true>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef typename _Sp<_Tp, _A0, _A1>::template rebind<_Up> type;
+#else
+    typedef typename _Sp<_Tp, _A0, _A1>::template rebind<_Up>::other type;
+#endif
+};
+
+template <template <class, class, class> class _Sp, class _Tp, class _A0,
+                                         class _A1, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _A0, _A1>, _Up, false>
+{
+    typedef _Sp<_Up, _A0, _A1> type;
+};
+
+template <template <class, class, class, class> class _Sp, class _Tp, class _A0,
+                                                class _A1, class _A2, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _A0, _A1, _A2>, _Up, true>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef typename _Sp<_Tp, _A0, _A1, _A2>::template rebind<_Up> type;
+#else
+    typedef typename _Sp<_Tp, _A0, _A1, _A2>::template rebind<_Up>::other type;
+#endif
+};
+
+template <template <class, class, class, class> class _Sp, class _Tp, class _A0,
+                                                class _A1, class _A2, class _Up>
+struct __pointer_traits_rebind<_Sp<_Tp, _A0, _A1, _A2>, _Up, false>
+{
+    typedef _Sp<_Up, _A0, _A1, _A2> type;
+};
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Ptr>
+struct _LIBCUDACXX_TEMPLATE_VIS pointer_traits
+{
+    typedef _Ptr                                                     pointer;
+    typedef typename __pointer_traits_element_type<pointer>::type    element_type;
+    typedef typename __pointer_traits_difference_type<pointer>::type difference_type;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Up> using rebind = typename __pointer_traits_rebind<pointer, _Up>::type;
+#else
+    template <class _Up> struct rebind
+        {typedef typename __pointer_traits_rebind<pointer, _Up>::type other;};
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+private:
+    struct __nat {};
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static pointer pointer_to(typename conditional<is_void<element_type>::value,
+                                           __nat, element_type>::type& __r)
+        {return pointer::pointer_to(__r);}
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS pointer_traits<_Tp*>
+{
+    typedef _Tp*      pointer;
+    typedef _Tp       element_type;
+    typedef ptrdiff_t difference_type;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Up> using rebind = _Up*;
+#else
+    template <class _Up> struct rebind {typedef _Up* other;};
+#endif
+
+private:
+    struct __nat {};
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    static pointer pointer_to(typename conditional<is_void<element_type>::value,
+                                      __nat, element_type>::type& __r) _NOEXCEPT
+        {return _CUDA_VSTD::addressof(__r);}
+};
+
+template <class _From, class _To>
+struct __rebind_pointer {
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef typename pointer_traits<_From>::template rebind<_To>        type;
+#else
+    typedef typename pointer_traits<_From>::template rebind<_To>::other type;
+#endif
+};
+
+// allocator_traits
+
+template <class _Tp, class = void>
+struct __has_pointer_type : false_type {};
+
+template <class _Tp>
+struct __has_pointer_type<_Tp,
+          typename __void_t<typename _Tp::pointer>::type> : true_type {};
+
+namespace __pointer_type_imp
+{
+
+template <class _Tp, class _Dp, bool = __has_pointer_type<_Dp>::value>
+struct __pointer_type
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Dp::pointer type;
+};
+
+template <class _Tp, class _Dp>
+struct __pointer_type<_Tp, _Dp, false>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Tp* type;
+};
+
+}  // __pointer_type_imp
+
+template <class _Tp, class _Dp>
+struct __pointer_type
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __pointer_type_imp::__pointer_type<_Tp, typename remove_reference<_Dp>::type>::type type;
+};
+
+template <class _Tp, class = void>
+struct __has_const_pointer : false_type {};
+
+template <class _Tp>
+struct __has_const_pointer<_Tp,
+            typename __void_t<typename _Tp::const_pointer>::type> : true_type {};
+
+template <class _Tp, class _Ptr, class _Alloc, bool = __has_const_pointer<_Alloc>::value>
+struct __const_pointer
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::const_pointer type;
+};
+
+template <class _Tp, class _Ptr, class _Alloc>
+struct __const_pointer<_Tp, _Ptr, _Alloc, false>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename pointer_traits<_Ptr>::template rebind<const _Tp> type;
+#else
+    typedef typename pointer_traits<_Ptr>::template rebind<const _Tp>::other type;
+#endif
+};
+
+template <class _Tp, class = void>
+struct __has_void_pointer : false_type {};
+
+template <class _Tp>
+struct __has_void_pointer<_Tp,
+               typename __void_t<typename _Tp::void_pointer>::type> : true_type {};
+
+template <class _Ptr, class _Alloc, bool = __has_void_pointer<_Alloc>::value>
+struct __void_pointer
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::void_pointer type;
+};
+
+template <class _Ptr, class _Alloc>
+struct __void_pointer<_Ptr, _Alloc, false>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename pointer_traits<_Ptr>::template rebind<void> type;
+#else
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename pointer_traits<_Ptr>::template rebind<void>::other type;
+#endif
+};
+
+template <class _Tp, class = void>
+struct __has_const_void_pointer : false_type {};
+
+template <class _Tp>
+struct __has_const_void_pointer<_Tp,
+            typename __void_t<typename _Tp::const_void_pointer>::type> : true_type {};
+
+template <class _Ptr, class _Alloc, bool = __has_const_void_pointer<_Alloc>::value>
+struct __const_void_pointer
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::const_void_pointer type;
+};
+
+template <class _Ptr, class _Alloc>
+struct __const_void_pointer<_Ptr, _Alloc, false>
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename pointer_traits<_Ptr>::template rebind<const void> type;
+#else
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename pointer_traits<_Ptr>::template rebind<const void>::other type;
+#endif
+};
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp*
+__to_raw_pointer(_Tp* __p) _NOEXCEPT
+{
+    return __p;
+}
+
+#if _LIBCUDACXX_STD_VER <= 17
+template <class _Pointer>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename pointer_traits<_Pointer>::element_type*
+__to_raw_pointer(_Pointer __p) _NOEXCEPT
+{
+    return _CUDA_VSTD::__to_raw_pointer(__p.operator->());
+}
+#else
+template <class _Pointer>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__to_raw_pointer(const _Pointer& __p) _NOEXCEPT
+-> decltype(pointer_traits<_Pointer>::to_address(__p))
+{
+    return pointer_traits<_Pointer>::to_address(__p);
+}
+
+template <class _Pointer, class... _None>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__to_raw_pointer(const _Pointer& __p, _None...) _NOEXCEPT
+{
+    return _CUDA_VSTD::__to_raw_pointer(__p.operator->());
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY constexpr
+_Tp*
+to_address(_Tp* __p) _NOEXCEPT
+{
+    static_assert(!is_function_v<_Tp>, "_Tp is a function type");
+    return __p;
+}
+
+template <class _Pointer>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+to_address(const _Pointer& __p) _NOEXCEPT
+{
+    return _CUDA_VSTD::__to_raw_pointer(__p);
+}
+#endif
+
+template <class _Tp, class = void>
+struct __has_size_type : false_type {};
+
+template <class _Tp>
+struct __has_size_type<_Tp,
+               typename __void_t<typename _Tp::size_type>::type> : true_type {};
+
+template <class _Alloc, class _DiffType, bool = __has_size_type<_Alloc>::value>
+struct __size_type
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename make_unsigned<_DiffType>::type type;
+};
+
+template <class _Alloc, class _DiffType>
+struct __size_type<_Alloc, _DiffType, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::size_type type;
+};
+
+template <class _Tp, class = void>
+struct __has_propagate_on_container_copy_assignment : false_type {};
+
+template <class _Tp>
+struct __has_propagate_on_container_copy_assignment<_Tp,
+    typename __void_t<typename _Tp::propagate_on_container_copy_assignment>::type>
+        : true_type {};
+
+template <class _Alloc, bool = __has_propagate_on_container_copy_assignment<_Alloc>::value>
+struct __propagate_on_container_copy_assignment
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE false_type type;
+};
+
+template <class _Alloc>
+struct __propagate_on_container_copy_assignment<_Alloc, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::propagate_on_container_copy_assignment type;
+};
+
+template <class _Tp, class = void>
+struct __has_propagate_on_container_move_assignment : false_type {};
+
+template <class _Tp>
+struct __has_propagate_on_container_move_assignment<_Tp,
+           typename __void_t<typename _Tp::propagate_on_container_move_assignment>::type>
+               : true_type {};
+
+template <class _Alloc, bool = __has_propagate_on_container_move_assignment<_Alloc>::value>
+struct __propagate_on_container_move_assignment
+{
+    typedef false_type type;
+};
+
+template <class _Alloc>
+struct __propagate_on_container_move_assignment<_Alloc, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::propagate_on_container_move_assignment type;
+};
+
+template <class _Tp, class = void>
+struct __has_propagate_on_container_swap : false_type {};
+
+template <class _Tp>
+struct __has_propagate_on_container_swap<_Tp,
+           typename __void_t<typename _Tp::propagate_on_container_swap>::type>
+               : true_type {};
+
+template <class _Alloc, bool = __has_propagate_on_container_swap<_Alloc>::value>
+struct __propagate_on_container_swap
+{
+    typedef false_type type;
+};
+
+template <class _Alloc>
+struct __propagate_on_container_swap<_Alloc, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::propagate_on_container_swap type;
+};
+
+template <class _Tp, class = void>
+struct __has_is_always_equal : false_type {};
+
+template <class _Tp>
+struct __has_is_always_equal<_Tp,
+           typename __void_t<typename _Tp::is_always_equal>::type>
+               : true_type {};
+
+template <class _Alloc, bool = __has_is_always_equal<_Alloc>::value>
+struct __is_always_equal
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _CUDA_VSTD::is_empty<_Alloc>::type type;
+};
+
+template <class _Alloc>
+struct __is_always_equal<_Alloc, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::is_always_equal type;
+};
+
+template <class _Tp, class _Up, bool = __has_rebind<_Tp, _Up>::value>
+struct __has_rebind_other
+{
+private:
+    struct __two {char __lx; char __lxx;};
+    template <class _Xp> static __two __test(...);
+    template <class _Xp> static char __test(typename _Xp::template rebind<_Up>::other* = 0);
+public:
+    static const bool value = sizeof(__test<_Tp>(0)) == 1;
+};
+
+template <class _Tp, class _Up>
+struct __has_rebind_other<_Tp, _Up, false>
+{
+    static const bool value = false;
+};
+
+template <class _Tp, class _Up, bool = __has_rebind_other<_Tp, _Up>::value>
+struct __allocator_traits_rebind
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Tp::template rebind<_Up>::other type;
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <template <class, class...> class _Alloc, class _Tp, class ..._Args, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _Args...>, _Up, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc<_Tp, _Args...>::template rebind<_Up>::other type;
+};
+
+template <template <class, class...> class _Alloc, class _Tp, class ..._Args, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _Args...>, _Up, false>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Alloc<_Up, _Args...> type;
+};
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <template <class> class _Alloc, class _Tp, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp>, _Up, true>
+{
+    typedef typename _Alloc<_Tp>::template rebind<_Up>::other type;
+};
+
+template <template <class> class _Alloc, class _Tp, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp>, _Up, false>
+{
+    typedef _Alloc<_Up> type;
+};
+
+template <template <class, class> class _Alloc, class _Tp, class _A0, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _A0>, _Up, true>
+{
+    typedef typename _Alloc<_Tp, _A0>::template rebind<_Up>::other type;
+};
+
+template <template <class, class> class _Alloc, class _Tp, class _A0, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _A0>, _Up, false>
+{
+    typedef _Alloc<_Up, _A0> type;
+};
+
+template <template <class, class, class> class _Alloc, class _Tp, class _A0,
+                                         class _A1, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _A0, _A1>, _Up, true>
+{
+    typedef typename _Alloc<_Tp, _A0, _A1>::template rebind<_Up>::other type;
+};
+
+template <template <class, class, class> class _Alloc, class _Tp, class _A0,
+                                         class _A1, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _A0, _A1>, _Up, false>
+{
+    typedef _Alloc<_Up, _A0, _A1> type;
+};
+
+template <template <class, class, class, class> class _Alloc, class _Tp, class _A0,
+                                                class _A1, class _A2, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _A0, _A1, _A2>, _Up, true>
+{
+    typedef typename _Alloc<_Tp, _A0, _A1, _A2>::template rebind<_Up>::other type;
+};
+
+template <template <class, class, class, class> class _Alloc, class _Tp, class _A0,
+                                                class _A1, class _A2, class _Up>
+struct __allocator_traits_rebind<_Alloc<_Tp, _A0, _A1, _A2>, _Up, false>
+{
+    typedef _Alloc<_Up, _A0, _A1, _A2> type;
+};
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Alloc, class _SizeType, class _ConstVoidPtr>
+auto
+__has_allocate_hint_test(_Alloc&& __a, _SizeType&& __sz, _ConstVoidPtr&& __p)
+    -> decltype((void)__a.allocate(__sz, __p), true_type());
+
+template <class _Alloc, class _SizeType, class _ConstVoidPtr>
+auto
+__has_allocate_hint_test(const _Alloc& __a, _SizeType&& __sz, _ConstVoidPtr&& __p)
+    -> false_type;
+
+template <class _Alloc, class _SizeType, class _ConstVoidPtr>
+struct __has_allocate_hint
+    : integral_constant<bool,
+        is_same<
+            decltype(_CUDA_VSTD::__has_allocate_hint_test(declval<_Alloc>(),
+                                          declval<_SizeType>(),
+                                          declval<_ConstVoidPtr>())),
+            true_type>::value>
+{
+};
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Alloc, class _SizeType, class _ConstVoidPtr>
+struct __has_allocate_hint
+    : true_type
+{
+};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+
+template <class _Alloc, class _Tp, class ..._Args>
+decltype(_CUDA_VSTD::declval<_Alloc>().construct(_CUDA_VSTD::declval<_Tp*>(),
+                                           _CUDA_VSTD::declval<_Args>()...),
+                                           true_type())
+__has_construct_test(_Alloc&& __a, _Tp* __p, _Args&& ...__args);
+
+template <class _Alloc, class _Pointer, class ..._Args>
+false_type
+__has_construct_test(const _Alloc& __a, _Pointer&& __p, _Args&& ...__args);
+
+template <class _Alloc, class _Pointer, class ..._Args>
+struct __has_construct
+    : integral_constant<bool,
+        is_same<
+            decltype(_CUDA_VSTD::__has_construct_test(declval<_Alloc>(),
+                                          declval<_Pointer>(),
+                                          declval<_Args>()...)),
+            true_type>::value>
+{
+};
+
+template <class _Alloc, class _Pointer>
+auto
+__has_destroy_test(_Alloc&& __a, _Pointer&& __p)
+    -> decltype(__a.destroy(__p), true_type());
+
+template <class _Alloc, class _Pointer>
+auto
+__has_destroy_test(const _Alloc& __a, _Pointer&& __p)
+    -> false_type;
+
+template <class _Alloc, class _Pointer>
+struct __has_destroy
+    : integral_constant<bool,
+        is_same<
+            decltype(_CUDA_VSTD::__has_destroy_test(declval<_Alloc>(),
+                                        declval<_Pointer>())),
+            true_type>::value>
+{
+};
+
+template <class _Alloc>
+auto
+__has_max_size_test(_Alloc&& __a)
+    -> decltype(__a.max_size(), true_type());
+
+template <class _Alloc>
+auto
+__has_max_size_test(const volatile _Alloc& __a)
+    -> false_type;
+
+template <class _Alloc>
+struct __has_max_size
+    : integral_constant<bool,
+        is_same<
+            decltype(_CUDA_VSTD::__has_max_size_test(declval<_Alloc&>())),
+            true_type>::value>
+{
+};
+
+template <class _Alloc>
+auto
+__has_select_on_container_copy_construction_test(_Alloc&& __a)
+    -> decltype(__a.select_on_container_copy_construction(), true_type());
+
+template <class _Alloc>
+auto
+__has_select_on_container_copy_construction_test(const volatile _Alloc& __a)
+    -> false_type;
+
+template <class _Alloc>
+struct __has_select_on_container_copy_construction
+    : integral_constant<bool,
+        is_same<
+            decltype(_CUDA_VSTD::__has_select_on_container_copy_construction_test(declval<_Alloc&>())),
+            true_type>::value>
+{
+};
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Alloc, class _Pointer, class _Tp, class = void>
+struct __has_construct : std::false_type {};
+
+template <class _Alloc, class _Pointer, class _Tp>
+struct __has_construct<_Alloc, _Pointer, _Tp, typename __void_t<
+    decltype(_CUDA_VSTD::declval<_Alloc>().construct(_CUDA_VSTD::declval<_Pointer>(), _CUDA_VSTD::declval<_Tp>()))
+>::type> : std::true_type {};
+
+template <class _Alloc, class _Pointer, class = void>
+struct __has_destroy : false_type {};
+
+template <class _Alloc, class _Pointer>
+struct __has_destroy<_Alloc, _Pointer, typename __void_t<
+    decltype(_CUDA_VSTD::declval<_Alloc>().destroy(_CUDA_VSTD::declval<_Pointer>()))
+>::type> : std::true_type {};
+
+template <class _Alloc>
+struct __has_max_size
+    : true_type
+{
+};
+
+template <class _Alloc>
+struct __has_select_on_container_copy_construction
+    : false_type
+{
+};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Alloc, class _Ptr, bool = __has_difference_type<_Alloc>::value>
+struct __alloc_traits_difference_type
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename pointer_traits<_Ptr>::difference_type type;
+};
+
+template <class _Alloc, class _Ptr>
+struct __alloc_traits_difference_type<_Alloc, _Ptr, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Alloc::difference_type type;
+};
+
+template <class _Tp>
+struct __is_default_allocator : false_type {};
+
+template <class _Tp>
+struct __is_default_allocator<_CUDA_VSTD::allocator<_Tp> > : true_type {};
+
+
+
+template <class _Alloc,
+    bool = __has_construct<_Alloc, typename _Alloc::value_type*,  typename _Alloc::value_type&&>::value && !__is_default_allocator<_Alloc>::value
+    >
+struct __is_cpp17_move_insertable;
+template <class _Alloc>
+struct __is_cpp17_move_insertable<_Alloc, true> : std::true_type {};
+template <class _Alloc>
+struct __is_cpp17_move_insertable<_Alloc, false> : std::is_move_constructible<typename _Alloc::value_type> {};
+
+template <class _Alloc,
+    bool = __has_construct<_Alloc, typename _Alloc::value_type*, const typename _Alloc::value_type&>::value && !__is_default_allocator<_Alloc>::value
+    >
+struct __is_cpp17_copy_insertable;
+template <class _Alloc>
+struct __is_cpp17_copy_insertable<_Alloc, true> : __is_cpp17_move_insertable<_Alloc> {};
+template <class _Alloc>
+struct __is_cpp17_copy_insertable<_Alloc, false> : integral_constant<bool,
+    std::is_copy_constructible<typename _Alloc::value_type>::value &&
+    __is_cpp17_move_insertable<_Alloc>::value>
+  {};
+
+
+
+template <class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS allocator_traits
+{
+    typedef _Alloc                              allocator_type;
+    typedef typename allocator_type::value_type value_type;
+
+    typedef typename __pointer_type<value_type, allocator_type>::type pointer;
+    typedef typename __const_pointer<value_type, pointer, allocator_type>::type const_pointer;
+    typedef typename __void_pointer<pointer, allocator_type>::type void_pointer;
+    typedef typename __const_void_pointer<pointer, allocator_type>::type const_void_pointer;
+
+    typedef typename __alloc_traits_difference_type<allocator_type, pointer>::type difference_type;
+    typedef typename __size_type<allocator_type, difference_type>::type size_type;
+
+    typedef typename __propagate_on_container_copy_assignment<allocator_type>::type
+                     propagate_on_container_copy_assignment;
+    typedef typename __propagate_on_container_move_assignment<allocator_type>::type
+                     propagate_on_container_move_assignment;
+    typedef typename __propagate_on_container_swap<allocator_type>::type
+                     propagate_on_container_swap;
+    typedef typename __is_always_equal<allocator_type>::type
+                     is_always_equal;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Tp> using rebind_alloc =
+                  typename __allocator_traits_rebind<allocator_type, _Tp>::type;
+    template <class _Tp> using rebind_traits = allocator_traits<rebind_alloc<_Tp> >;
+#else  // _LIBCUDACXX_CXX03_LANG
+    template <class _Tp> struct rebind_alloc
+        {typedef typename __allocator_traits_rebind<allocator_type, _Tp>::type other;};
+    template <class _Tp> struct rebind_traits
+        {typedef allocator_traits<typename rebind_alloc<_Tp>::other> other;};
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    static pointer allocate(allocator_type& __a, size_type __n)
+        {return __a.allocate(__n);}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    static pointer allocate(allocator_type& __a, size_type __n, const_void_pointer __hint)
+        {return __allocate(__a, __n, __hint,
+            __has_allocate_hint<allocator_type, size_type, const_void_pointer>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void deallocate(allocator_type& __a, pointer __p, size_type __n) _NOEXCEPT
+        {__a.deallocate(__p, __n);}
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void construct(allocator_type& __a, _Tp* __p, _Args&&... __args)
+            {__construct(__has_construct<allocator_type, _Tp*, _Args...>(),
+                         __a, __p, _CUDA_VSTD::forward<_Args>(__args)...);}
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void construct(allocator_type&, _Tp* __p)
+            {
+                ::new ((void*)__p) _Tp();
+            }
+    template <class _Tp, class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void construct(allocator_type& __a, _Tp* __p, const _A0& __a0)
+            {
+                __construct(__has_construct<allocator_type, _Tp*, const _A0&>(),
+                            __a, __p, __a0);
+            }
+    template <class _Tp, class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void construct(allocator_type&, _Tp* __p, const _A0& __a0,
+                              const _A1& __a1)
+            {
+                ::new ((void*)__p) _Tp(__a0, __a1);
+            }
+    template <class _Tp, class _A0, class _A1, class _A2>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void construct(allocator_type&, _Tp* __p, const _A0& __a0,
+                              const _A1& __a1, const _A2& __a2)
+            {
+                ::new ((void*)__p) _Tp(__a0, __a1, __a2);
+            }
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void destroy(allocator_type& __a, _Tp* __p)
+            {__destroy(__has_destroy<allocator_type, _Tp*>(), __a, __p);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type max_size(const allocator_type& __a) _NOEXCEPT
+        {return __max_size(__has_max_size<const allocator_type>(), __a);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static allocator_type
+        select_on_container_copy_construction(const allocator_type& __a)
+            {return __select_on_container_copy_construction(
+                __has_select_on_container_copy_construction<const allocator_type>(),
+                __a);}
+
+    template <class _Ptr>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        void
+        __construct_forward_with_exception_guarantees(allocator_type& __a, _Ptr __begin1, _Ptr __end1, _Ptr& __begin2)
+        {
+            static_assert(__is_cpp17_move_insertable<allocator_type>::value,
+              "The specified type does not meet the requirements of Cpp17MoveInsertible");
+            for (; __begin1 != __end1; ++__begin1, (void) ++__begin2)
+              construct(__a, _CUDA_VSTD::__to_raw_pointer(__begin2),
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+                        _CUDA_VSTD::move(*__begin1)
+#else
+                        _CUDA_VSTD::move_if_noexcept(*__begin1)
+#endif
+                        );
+        }
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            (__is_default_allocator<allocator_type>::value
+                || !__has_construct<allocator_type, _Tp*, _Tp>::value) &&
+             is_trivially_move_constructible<_Tp>::value,
+            void
+        >::type
+        __construct_forward_with_exception_guarantees(allocator_type&, _Tp* __begin1, _Tp* __end1, _Tp*& __begin2)
+        {
+            ptrdiff_t _Np = __end1 - __begin1;
+            if (_Np > 0)
+            {
+                _CUDA_VSTD::memcpy(__begin2, __begin1, _Np * sizeof(_Tp));
+                __begin2 += _Np;
+            }
+        }
+
+    template <class _Iter, class _Ptr>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        void
+        __construct_range_forward(allocator_type& __a, _Iter __begin1, _Iter __end1, _Ptr& __begin2)
+        {
+            for (; __begin1 != __end1; ++__begin1, (void) ++__begin2)
+                construct(__a, _CUDA_VSTD::__to_raw_pointer(__begin2), *__begin1);
+        }
+
+    template <class _SourceTp, class _DestTp,
+              class _RawSourceTp = typename remove_const<_SourceTp>::type,
+              class _RawDestTp = typename remove_const<_DestTp>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            is_trivially_move_constructible<_DestTp>::value &&
+            is_same<_RawSourceTp, _RawDestTp>::value &&
+            (__is_default_allocator<allocator_type>::value ||
+             !__has_construct<allocator_type, _DestTp*, _SourceTp&>::value),
+            void
+        >::type
+        __construct_range_forward(allocator_type&, _SourceTp* __begin1, _SourceTp* __end1, _DestTp*& __begin2)
+        {
+            ptrdiff_t _Np = __end1 - __begin1;
+            if (_Np > 0)
+            {
+                _CUDA_VSTD::memcpy(const_cast<_RawDestTp*>(__begin2), __begin1, _Np * sizeof(_DestTp));
+                __begin2 += _Np;
+            }
+        }
+
+    template <class _Ptr>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        void
+        __construct_backward_with_exception_guarantees(allocator_type& __a, _Ptr __begin1, _Ptr __end1, _Ptr& __end2)
+        {
+            static_assert(__is_cpp17_move_insertable<allocator_type>::value,
+              "The specified type does not meet the requirements of Cpp17MoveInsertable");
+            while (__end1 != __begin1)
+            {
+              construct(__a, _CUDA_VSTD::__to_raw_pointer(__end2 - 1),
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+                        _CUDA_VSTD::move(*--__end1)
+#else
+                        _CUDA_VSTD::move_if_noexcept(*--__end1)
+#endif
+                        );
+              --__end2;
+            }
+        }
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            (__is_default_allocator<allocator_type>::value
+                || !__has_construct<allocator_type, _Tp*, _Tp>::value) &&
+             is_trivially_move_constructible<_Tp>::value,
+            void
+        >::type
+        __construct_backward_with_exception_guarantees(allocator_type&, _Tp* __begin1, _Tp* __end1, _Tp*& __end2)
+        {
+            ptrdiff_t _Np = __end1 - __begin1;
+            __end2 -= _Np;
+            if (_Np > 0)
+                _CUDA_VSTD::memcpy(__end2, __begin1, _Np * sizeof(_Tp));
+        }
+
+private:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static pointer __allocate(allocator_type& __a, size_type __n,
+        const_void_pointer __hint, true_type)
+        {return __a.allocate(__n, __hint);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static pointer __allocate(allocator_type& __a, size_type __n,
+        const_void_pointer, false_type)
+        {return __a.allocate(__n);}
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void __construct(true_type, allocator_type& __a, _Tp* __p, _Args&&... __args)
+            {__a.construct(__p, _CUDA_VSTD::forward<_Args>(__args)...);}
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void __construct(false_type, allocator_type&, _Tp* __p, _Args&&... __args)
+            {
+                ::new ((void*)__p) _Tp(_CUDA_VSTD::forward<_Args>(__args)...);
+            }
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+    template <class _Tp, class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void __construct(true_type, allocator_type& __a, _Tp* __p,
+                                const _A0& __a0)
+            {__a.construct(__p, __a0);}
+    template <class _Tp, class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void __construct(false_type, allocator_type&, _Tp* __p,
+                                const _A0& __a0)
+            {
+                ::new ((void*)__p) _Tp(__a0);
+            }
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void __destroy(true_type, allocator_type& __a, _Tp* __p)
+            {__a.destroy(__p);}
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static void __destroy(false_type, allocator_type&, _Tp* __p)
+            {
+                __p->~_Tp();
+            }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type __max_size(true_type, const allocator_type& __a) _NOEXCEPT
+            {return __a.max_size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type __max_size(false_type, const allocator_type&) _NOEXCEPT
+            {return numeric_limits<size_type>::max() / sizeof(value_type);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static allocator_type
+        __select_on_container_copy_construction(true_type, const allocator_type& __a)
+            {return __a.select_on_container_copy_construction();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static allocator_type
+        __select_on_container_copy_construction(false_type, const allocator_type& __a)
+            {return __a;}
+};
+
+template <class _Traits, class _Tp>
+struct __rebind_alloc_helper
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename _Traits::template rebind_alloc<_Tp>        type;
+#else
+    typedef typename _Traits::template rebind_alloc<_Tp>::other type;
+#endif
+};
+
+// allocator
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS allocator
+{
+public:
+    typedef size_t            size_type;
+    typedef ptrdiff_t         difference_type;
+    typedef _Tp*              pointer;
+    typedef const _Tp*        const_pointer;
+    typedef _Tp&              reference;
+    typedef const _Tp&        const_reference;
+    typedef _Tp               value_type;
+
+    typedef true_type propagate_on_container_move_assignment;
+    typedef true_type is_always_equal;
+
+    template <class _Up> struct rebind {typedef allocator<_Up> other;};
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    allocator() _NOEXCEPT {}
+
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    allocator(const allocator<_Up>&) _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY pointer address(reference __x) const _NOEXCEPT
+        {return _CUDA_VSTD::addressof(__x);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_pointer address(const_reference __x) const _NOEXCEPT
+        {return _CUDA_VSTD::addressof(__x);}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    pointer allocate(size_type __n, allocator<void>::const_pointer = 0)
+        {
+        if (__n > max_size())
+            __throw_length_error("allocator<T>::allocate(size_t n)"
+                                 " 'n' exceeds maximum supported size");
+        return static_cast<pointer>(_CUDA_VSTD::__libcpp_allocate(__n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp)));
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY void deallocate(pointer __p, size_type __n) _NOEXCEPT
+        {_CUDA_VSTD::__libcpp_deallocate((void*)__p, __n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp));}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type max_size() const _NOEXCEPT
+        {return size_type(~0) / sizeof(_Tp);}
+#if !defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCUDACXX_HAS_NO_VARIADICS)
+    template <class _Up, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(_Up* __p, _Args&&... __args)
+        {
+            ::new((void*)__p) _Up(_CUDA_VSTD::forward<_Args>(__args)...);
+        }
+#else  // !defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCUDACXX_HAS_NO_VARIADICS)
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p)
+        {
+            ::new((void*)__p) _Tp();
+        }
+# if defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES)
+
+    template <class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, _A0& __a0)
+        {
+            ::new((void*)__p) _Tp(__a0);
+        }
+    template <class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, const _A0& __a0)
+        {
+            ::new((void*)__p) _Tp(__a0);
+        }
+# endif  // defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES)
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, _A0& __a0, _A1& __a1)
+        {
+            ::new((void*)__p) _Tp(__a0, __a1);
+        }
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, const _A0& __a0, _A1& __a1)
+        {
+            ::new((void*)__p) _Tp(__a0, __a1);
+        }
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, _A0& __a0, const _A1& __a1)
+        {
+            ::new((void*)__p) _Tp(__a0, __a1);
+        }
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, const _A0& __a0, const _A1& __a1)
+        {
+            ::new((void*)__p) _Tp(__a0, __a1);
+        }
+#endif  // !defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCUDACXX_HAS_NO_VARIADICS)
+    _LIBCUDACXX_INLINE_VISIBILITY void destroy(pointer __p) {__p->~_Tp();}
+};
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS allocator<const _Tp>
+{
+public:
+    typedef size_t            size_type;
+    typedef ptrdiff_t         difference_type;
+    typedef const _Tp*        pointer;
+    typedef const _Tp*        const_pointer;
+    typedef const _Tp&        reference;
+    typedef const _Tp&        const_reference;
+    typedef const _Tp         value_type;
+
+    typedef true_type propagate_on_container_move_assignment;
+    typedef true_type is_always_equal;
+
+    template <class _Up> struct rebind {typedef allocator<_Up> other;};
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    allocator() _NOEXCEPT {}
+
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+    allocator(const allocator<_Up>&) _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY const_pointer address(const_reference __x) const _NOEXCEPT
+        {return _CUDA_VSTD::addressof(__x);}
+    _LIBCUDACXX_INLINE_VISIBILITY pointer allocate(size_type __n, allocator<void>::const_pointer = 0)
+    {
+        if (__n > max_size())
+            __throw_length_error("allocator<const T>::allocate(size_t n)"
+                                 " 'n' exceeds maximum supported size");
+        return static_cast<pointer>(_CUDA_VSTD::__libcpp_allocate(__n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp)));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY void deallocate(pointer __p, size_type __n) _NOEXCEPT
+        {_CUDA_VSTD::__libcpp_deallocate((void*) const_cast<_Tp *>(__p), __n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp));}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type max_size() const _NOEXCEPT
+        {return size_type(~0) / sizeof(_Tp);}
+#if !defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCUDACXX_HAS_NO_VARIADICS)
+    template <class _Up, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(_Up* __p, _Args&&... __args)
+        {
+            ::new((void*)__p) _Up(_CUDA_VSTD::forward<_Args>(__args)...);
+        }
+#else  // !defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCUDACXX_HAS_NO_VARIADICS)
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp();
+        }
+# if defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES)
+
+    template <class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, _A0& __a0)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp(__a0);
+        }
+    template <class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, const _A0& __a0)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp(__a0);
+        }
+# endif  // defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES)
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, _A0& __a0, _A1& __a1)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp(__a0, __a1);
+        }
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, const _A0& __a0, _A1& __a1)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp(__a0, __a1);
+        }
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, _A0& __a0, const _A1& __a1)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp(__a0, __a1);
+        }
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void
+        construct(pointer __p, const _A0& __a0, const _A1& __a1)
+        {
+            ::new((void*) const_cast<_Tp *>(__p)) _Tp(__a0, __a1);
+        }
+#endif  // !defined(_LIBCUDACXX_HAS_NO_RVALUE_REFERENCES) && !defined(_LIBCUDACXX_HAS_NO_VARIADICS)
+    _LIBCUDACXX_INLINE_VISIBILITY void destroy(pointer __p) {__p->~_Tp();}
+};
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(const allocator<_Tp>&, const allocator<_Up>&) _NOEXCEPT {return true;}
+
+template <class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(const allocator<_Tp>&, const allocator<_Up>&) _NOEXCEPT {return false;}
+
+template <class _OutputIterator, class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS raw_storage_iterator
+    : public iterator<output_iterator_tag,
+                      _Tp,                                         // purposefully not C++03
+                      ptrdiff_t,                                   // purposefully not C++03
+                      _Tp*,                                        // purposefully not C++03
+                      raw_storage_iterator<_OutputIterator, _Tp>&> // purposefully not C++03
+{
+private:
+    _OutputIterator __x_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit raw_storage_iterator(_OutputIterator __x) : __x_(__x) {}
+    _LIBCUDACXX_INLINE_VISIBILITY raw_storage_iterator& operator*() {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY raw_storage_iterator& operator=(const _Tp& __element)
+        {::new(_CUDA_VSTD::addressof(*__x_)) _Tp(__element); return *this;}
+#if _LIBCUDACXX_STD_VER >= 14
+    _LIBCUDACXX_INLINE_VISIBILITY raw_storage_iterator& operator=(_Tp&& __element)
+        {::new(_CUDA_VSTD::addressof(*__x_)) _Tp(_CUDA_VSTD::move(__element)); return *this;}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY raw_storage_iterator& operator++() {++__x_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY raw_storage_iterator  operator++(int)
+        {raw_storage_iterator __t(*this); ++__x_; return __t;}
+#if _LIBCUDACXX_STD_VER >= 14
+    _LIBCUDACXX_INLINE_VISIBILITY _OutputIterator base() const { return __x_; }
+#endif
+};
+
+template <class _Tp>
+_LIBCUDACXX_NODISCARD_EXT _LIBCUDACXX_NO_CFI
+pair<_Tp*, ptrdiff_t>
+get_temporary_buffer(ptrdiff_t __n) _NOEXCEPT
+{
+    pair<_Tp*, ptrdiff_t> __r(0, 0);
+    const ptrdiff_t __m = (~ptrdiff_t(0) ^
+                           ptrdiff_t(ptrdiff_t(1) << (sizeof(ptrdiff_t) * __CHAR_BIT__ - 1)))
+                           / sizeof(_Tp);
+    if (__n > __m)
+        __n = __m;
+    while (__n > 0)
+    {
+#if !defined(_LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION)
+    if (__is_overaligned_for_new(_LIBCUDACXX_ALIGNOF(_Tp)))
+        {
+            std::align_val_t __al =
+                std::align_val_t(std::alignment_of<_Tp>::value);
+            __r.first = static_cast<_Tp*>(::operator new(
+                __n * sizeof(_Tp), __al, nothrow));
+        } else {
+            __r.first = static_cast<_Tp*>(::operator new(
+                __n * sizeof(_Tp), nothrow));
+        }
+#else
+    if (__is_overaligned_for_new(_LIBCUDACXX_ALIGNOF(_Tp)))
+        {
+            // Since aligned operator new is unavailable, return an empty
+            // buffer rather than one with invalid alignment.
+            return __r;
+        }
+
+        __r.first = static_cast<_Tp*>(::operator new(__n * sizeof(_Tp), nothrow));
+#endif
+
+        if (__r.first)
+        {
+            __r.second = __n;
+            break;
+        }
+        __n /= 2;
+    }
+    return __r;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void return_temporary_buffer(_Tp* __p) _NOEXCEPT
+{
+  _CUDA_VSTD::__libcpp_deallocate_unsized((void*)__p, _LIBCUDACXX_ALIGNOF(_Tp));
+}
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+template <class _Tp>
+struct _LIBCUDACXX_DEPRECATED_IN_CXX11 auto_ptr_ref
+{
+    _Tp* __ptr_;
+};
+
+template<class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 auto_ptr
+{
+private:
+    _Tp* __ptr_;
+public:
+    typedef _Tp element_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY explicit auto_ptr(_Tp* __p = 0) throw() : __ptr_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY auto_ptr(auto_ptr& __p) throw() : __ptr_(__p.release()) {}
+    template<class _Up> _LIBCUDACXX_INLINE_VISIBILITY auto_ptr(auto_ptr<_Up>& __p) throw()
+        : __ptr_(__p.release()) {}
+    _LIBCUDACXX_INLINE_VISIBILITY auto_ptr& operator=(auto_ptr& __p) throw()
+        {reset(__p.release()); return *this;}
+    template<class _Up> _LIBCUDACXX_INLINE_VISIBILITY auto_ptr& operator=(auto_ptr<_Up>& __p) throw()
+        {reset(__p.release()); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY auto_ptr& operator=(auto_ptr_ref<_Tp> __p) throw()
+        {reset(__p.__ptr_); return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY ~auto_ptr() throw() {delete __ptr_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _Tp& operator*() const throw()
+        {return *__ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _Tp* operator->() const throw() {return __ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _Tp* get() const throw() {return __ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _Tp* release() throw()
+    {
+        _Tp* __t = __ptr_;
+        __ptr_ = 0;
+        return __t;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY void reset(_Tp* __p = 0) throw()
+    {
+        if (__ptr_ != __p)
+            delete __ptr_;
+        __ptr_ = __p;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY auto_ptr(auto_ptr_ref<_Tp> __p) throw() : __ptr_(__p.__ptr_) {}
+    template<class _Up> _LIBCUDACXX_INLINE_VISIBILITY operator auto_ptr_ref<_Up>() throw()
+        {auto_ptr_ref<_Up> __t; __t.__ptr_ = release(); return __t;}
+    template<class _Up> _LIBCUDACXX_INLINE_VISIBILITY operator auto_ptr<_Up>() throw()
+        {return auto_ptr<_Up>(release());}
+};
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_DEPRECATED_IN_CXX11 auto_ptr<void>
+{
+public:
+    typedef void element_type;
+};
+#endif
+
+template <class _Tp, int _Idx,
+          bool _CanBeEmptyBase =
+              is_empty<_Tp>::value && !__libcpp_is_final<_Tp>::value>
+struct __compressed_pair_elem {
+  typedef _Tp _ParamT;
+  typedef _Tp& reference;
+  typedef const _Tp& const_reference;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr __compressed_pair_elem() : __value_() {}
+
+  template <class _Up, class = typename enable_if<
+      !is_same<__compressed_pair_elem, typename decay<_Up>::type>::value
+  >::type>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr explicit
+  __compressed_pair_elem(_Up&& __u)
+      : __value_(_CUDA_VSTD::forward<_Up>(__u))
+    {
+    }
+
+  template <class... _Args, size_t... _Indexes>
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+  __compressed_pair_elem(piecewise_construct_t, tuple<_Args...> __args,
+                         __tuple_indices<_Indexes...>)
+      : __value_(_CUDA_VSTD::forward<_Args>(_CUDA_VSTD::get<_Indexes>(__args))...) {}
+#else
+  _LIBCUDACXX_INLINE_VISIBILITY __compressed_pair_elem() : __value_() {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __compressed_pair_elem(_ParamT __p) : __value_(std::forward<_ParamT>(__p)) {}
+#endif
+
+  _LIBCUDACXX_INLINE_VISIBILITY reference __get() _NOEXCEPT { return __value_; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const_reference __get() const _NOEXCEPT { return __value_; }
+
+private:
+  _Tp __value_;
+};
+
+template <class _Tp, int _Idx>
+struct __compressed_pair_elem<_Tp, _Idx, true> : private _Tp {
+  typedef _Tp _ParamT;
+  typedef _Tp& reference;
+  typedef const _Tp& const_reference;
+  typedef _Tp __value_type;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr __compressed_pair_elem() = default;
+
+  template <class _Up, class = typename enable_if<
+        !is_same<__compressed_pair_elem, typename decay<_Up>::type>::value
+  >::type>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr explicit
+  __compressed_pair_elem(_Up&& __u)
+      : __value_type(_CUDA_VSTD::forward<_Up>(__u))
+  {}
+
+  template <class... _Args, size_t... _Indexes>
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+  __compressed_pair_elem(piecewise_construct_t, tuple<_Args...> __args,
+                         __tuple_indices<_Indexes...>)
+      : __value_type(_CUDA_VSTD::forward<_Args>(_CUDA_VSTD::get<_Indexes>(__args))...) {}
+#else
+  _LIBCUDACXX_INLINE_VISIBILITY __compressed_pair_elem() : __value_type() {}
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __compressed_pair_elem(_ParamT __p)
+      : __value_type(std::forward<_ParamT>(__p)) {}
+#endif
+
+  _LIBCUDACXX_INLINE_VISIBILITY reference __get() _NOEXCEPT { return *this; }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const_reference __get() const _NOEXCEPT { return *this; }
+};
+
+// Tag used to construct the second element of the compressed pair.
+struct __second_tag {};
+
+template <class _T1, class _T2>
+class __compressed_pair : private __compressed_pair_elem<_T1, 0>,
+                          private __compressed_pair_elem<_T2, 1> {
+  typedef _LIBCUDACXX_NODEBUG_TYPE __compressed_pair_elem<_T1, 0> _Base1;
+  typedef _LIBCUDACXX_NODEBUG_TYPE __compressed_pair_elem<_T2, 1> _Base2;
+
+  // NOTE: This static assert should never fire because __compressed_pair
+  // is *almost never* used in a scenario where it's possible for T1 == T2.
+  // (The exception is std::function where it is possible that the function
+  //  object and the allocator have the same type).
+  static_assert((!is_same<_T1, _T2>::value),
+    "__compressed_pair cannot be instantated when T1 and T2 are the same type; "
+    "The current implementation is NOT ABI-compatible with the previous "
+    "implementation for this configuration");
+
+public:
+#ifndef _LIBCUDACXX_CXX03_LANG
+  template <bool _Dummy = true,
+      class = typename enable_if<
+          __dependent_type<is_default_constructible<_T1>, _Dummy>::value &&
+          __dependent_type<is_default_constructible<_T2>, _Dummy>::value
+      >::type
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr __compressed_pair() {}
+
+  template <class _Tp, typename enable_if<!is_same<typename decay<_Tp>::type,
+                                                   __compressed_pair>::value,
+                                          bool>::type = true>
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr explicit
+  __compressed_pair(_Tp&& __t)
+      : _Base1(std::forward<_Tp>(__t)), _Base2() {}
+
+  template <class _Tp>
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr
+  __compressed_pair(__second_tag, _Tp&& __t)
+      : _Base1(), _Base2(std::forward<_Tp>(__t)) {}
+
+  template <class _U1, class _U2>
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr
+  __compressed_pair(_U1&& __t1, _U2&& __t2)
+      : _Base1(std::forward<_U1>(__t1)), _Base2(std::forward<_U2>(__t2)) {}
+
+  template <class... _Args1, class... _Args2>
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+  __compressed_pair(piecewise_construct_t __pc, tuple<_Args1...> __first_args,
+                    tuple<_Args2...> __second_args)
+      : _Base1(__pc, _CUDA_VSTD::move(__first_args),
+               typename __make_tuple_indices<sizeof...(_Args1)>::type()),
+        _Base2(__pc, _CUDA_VSTD::move(__second_args),
+               typename __make_tuple_indices<sizeof...(_Args2)>::type()) {}
+
+#else
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __compressed_pair() {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY explicit
+  __compressed_pair(_T1 __t1) : _Base1(_CUDA_VSTD::forward<_T1>(__t1)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __compressed_pair(__second_tag, _T2 __t2)
+      : _Base1(), _Base2(_CUDA_VSTD::forward<_T2>(__t2)) {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  __compressed_pair(_T1 __t1, _T2 __t2)
+      : _Base1(_CUDA_VSTD::forward<_T1>(__t1)), _Base2(_CUDA_VSTD::forward<_T2>(__t2)) {}
+#endif
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename _Base1::reference first() _NOEXCEPT {
+    return static_cast<_Base1&>(*this).__get();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename _Base1::const_reference first() const _NOEXCEPT {
+    return static_cast<_Base1 const&>(*this).__get();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename _Base2::reference second() _NOEXCEPT {
+    return static_cast<_Base2&>(*this).__get();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename _Base2::const_reference second() const _NOEXCEPT {
+    return static_cast<_Base2 const&>(*this).__get();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void swap(__compressed_pair& __x)
+    _NOEXCEPT_(__is_nothrow_swappable<_T1>::value &&
+               __is_nothrow_swappable<_T2>::value)
+  {
+    using std::swap;
+    swap(first(), __x.first());
+    swap(second(), __x.second());
+  }
+};
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void swap(__compressed_pair<_T1, _T2>& __x, __compressed_pair<_T1, _T2>& __y)
+    _NOEXCEPT_(__is_nothrow_swappable<_T1>::value &&
+               __is_nothrow_swappable<_T2>::value) {
+  __x.swap(__y);
+}
+
+// default_delete
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS default_delete {
+    static_assert(!is_function<_Tp>::value,
+                  "default_delete cannot be instantiated for function types");
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr default_delete() _NOEXCEPT = default;
+#else
+  _LIBCUDACXX_INLINE_VISIBILITY default_delete() {}
+#endif
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  default_delete(const default_delete<_Up>&,
+                 typename enable_if<is_convertible<_Up*, _Tp*>::value>::type* =
+                     0) _NOEXCEPT {}
+
+  _LIBCUDACXX_INLINE_VISIBILITY void operator()(_Tp* __ptr) const _NOEXCEPT {
+    static_assert(sizeof(_Tp) > 0,
+                  "default_delete can not delete incomplete type");
+    static_assert(!is_void<_Tp>::value,
+                  "default_delete can not delete incomplete type");
+    delete __ptr;
+  }
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS default_delete<_Tp[]> {
+private:
+  template <class _Up>
+  struct _EnableIfConvertible
+      : enable_if<is_convertible<_Up(*)[], _Tp(*)[]>::value> {};
+
+public:
+#ifndef _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr default_delete() _NOEXCEPT = default;
+#else
+  _LIBCUDACXX_INLINE_VISIBILITY default_delete() {}
+#endif
+
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  default_delete(const default_delete<_Up[]>&,
+                 typename _EnableIfConvertible<_Up>::type* = 0) _NOEXCEPT {}
+
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename _EnableIfConvertible<_Up>::type
+  operator()(_Up* __ptr) const _NOEXCEPT {
+    static_assert(sizeof(_Tp) > 0,
+                  "default_delete can not delete incomplete type");
+    static_assert(!is_void<_Tp>::value,
+                  "default_delete can not delete void type");
+    delete[] __ptr;
+  }
+};
+
+template <class _Deleter>
+struct __unique_ptr_deleter_sfinae {
+  static_assert(!is_reference<_Deleter>::value, "incorrect specialization");
+  typedef const _Deleter& __lval_ref_type;
+  typedef _Deleter&& __good_rval_ref_type;
+  typedef true_type __enable_rval_overload;
+};
+
+template <class _Deleter>
+struct __unique_ptr_deleter_sfinae<_Deleter const&> {
+  typedef const _Deleter& __lval_ref_type;
+  typedef const _Deleter&& __bad_rval_ref_type;
+  typedef false_type __enable_rval_overload;
+};
+
+template <class _Deleter>
+struct __unique_ptr_deleter_sfinae<_Deleter&> {
+  typedef _Deleter& __lval_ref_type;
+  typedef _Deleter&& __bad_rval_ref_type;
+  typedef false_type __enable_rval_overload;
+};
+
+template <class _Tp, class _Dp = default_delete<_Tp> >
+class _LIBCUDACXX_TEMPLATE_VIS unique_ptr {
+public:
+  typedef _Tp element_type;
+  typedef _Dp deleter_type;
+  typedef _LIBCUDACXX_NODEBUG_TYPE typename __pointer_type<_Tp, deleter_type>::type pointer;
+
+  static_assert(!is_rvalue_reference<deleter_type>::value,
+                "the specified deleter type cannot be an rvalue reference");
+
+private:
+  __compressed_pair<pointer, deleter_type> __ptr_;
+
+  struct __nat { int __for_bool_; };
+
+  typedef _LIBCUDACXX_NODEBUG_TYPE __unique_ptr_deleter_sfinae<_Dp> _DeleterSFINAE;
+
+  template <bool _Dummy>
+  using _LValRefType _LIBCUDACXX_NODEBUG_TYPE =
+      typename __dependent_type<_DeleterSFINAE, _Dummy>::__lval_ref_type;
+
+  template <bool _Dummy>
+  using _GoodRValRefType _LIBCUDACXX_NODEBUG_TYPE =
+      typename __dependent_type<_DeleterSFINAE, _Dummy>::__good_rval_ref_type;
+
+  template <bool _Dummy>
+  using _BadRValRefType _LIBCUDACXX_NODEBUG_TYPE  =
+      typename __dependent_type<_DeleterSFINAE, _Dummy>::__bad_rval_ref_type;
+
+  template <bool _Dummy, class _Deleter = typename __dependent_type<
+                             __identity<deleter_type>, _Dummy>::type>
+  using _EnableIfDeleterDefaultConstructible _LIBCUDACXX_NODEBUG_TYPE =
+      typename enable_if<is_default_constructible<_Deleter>::value &&
+                         !is_pointer<_Deleter>::value>::type;
+
+  template <class _ArgType>
+  using _EnableIfDeleterConstructible _LIBCUDACXX_NODEBUG_TYPE  =
+      typename enable_if<is_constructible<deleter_type, _ArgType>::value>::type;
+
+  template <class _UPtr, class _Up>
+  using _EnableIfMoveConvertible _LIBCUDACXX_NODEBUG_TYPE  = typename enable_if<
+      is_convertible<typename _UPtr::pointer, pointer>::value &&
+      !is_array<_Up>::value
+  >::type;
+
+  template <class _UDel>
+  using _EnableIfDeleterConvertible _LIBCUDACXX_NODEBUG_TYPE  = typename enable_if<
+      (is_reference<_Dp>::value && is_same<_Dp, _UDel>::value) ||
+      (!is_reference<_Dp>::value && is_convertible<_UDel, _Dp>::value)
+    >::type;
+
+  template <class _UDel>
+  using _EnableIfDeleterAssignable = typename enable_if<
+      is_assignable<_Dp&, _UDel&&>::value
+    >::type;
+
+public:
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterDefaultConstructible<_Dummy> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_CONSTEXPR unique_ptr() _NOEXCEPT : __ptr_(pointer()) {}
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterDefaultConstructible<_Dummy> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_CONSTEXPR unique_ptr(nullptr_t) _NOEXCEPT : __ptr_(pointer()) {}
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterDefaultConstructible<_Dummy> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit unique_ptr(pointer __p) _NOEXCEPT : __ptr_(__p) {}
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_LValRefType<_Dummy> > >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(pointer __p, _LValRefType<_Dummy> __d) _NOEXCEPT
+      : __ptr_(__p, __d) {}
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_GoodRValRefType<_Dummy> > >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(pointer __p, _GoodRValRefType<_Dummy> __d) _NOEXCEPT
+      : __ptr_(__p, _CUDA_VSTD::move(__d)) {
+    static_assert(!is_reference<deleter_type>::value,
+                  "rvalue deleter bound to reference");
+  }
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_BadRValRefType<_Dummy> > >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(pointer __p, _BadRValRefType<_Dummy> __d) = delete;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(unique_ptr&& __u) _NOEXCEPT
+      : __ptr_(__u.release(), _CUDA_VSTD::forward<deleter_type>(__u.get_deleter())) {
+  }
+
+  template <class _Up, class _Ep,
+      class = _EnableIfMoveConvertible<unique_ptr<_Up, _Ep>, _Up>,
+      class = _EnableIfDeleterConvertible<_Ep>
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(unique_ptr<_Up, _Ep>&& __u) _NOEXCEPT
+      : __ptr_(__u.release(), _CUDA_VSTD::forward<_Ep>(__u.get_deleter())) {}
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(auto_ptr<_Up>&& __p,
+             typename enable_if<is_convertible<_Up*, _Tp*>::value &&
+                                    is_same<_Dp, default_delete<_Tp> >::value,
+                                __nat>::type = __nat()) _NOEXCEPT
+      : __ptr_(__p.release()) {}
+#endif
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr& operator=(unique_ptr&& __u) _NOEXCEPT {
+    reset(__u.release());
+    __ptr_.second() = _CUDA_VSTD::forward<deleter_type>(__u.get_deleter());
+    return *this;
+  }
+
+  template <class _Up, class _Ep,
+      class = _EnableIfMoveConvertible<unique_ptr<_Up, _Ep>, _Up>,
+      class = _EnableIfDeleterAssignable<_Ep>
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr& operator=(unique_ptr<_Up, _Ep>&& __u) _NOEXCEPT {
+    reset(__u.release());
+    __ptr_.second() = _CUDA_VSTD::forward<_Ep>(__u.get_deleter());
+    return *this;
+  }
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+  template <class _Up>
+  _LIBCUDACXX_INLINE_VISIBILITY
+      typename enable_if<is_convertible<_Up*, _Tp*>::value &&
+                             is_same<_Dp, default_delete<_Tp> >::value,
+                         unique_ptr&>::type
+      operator=(auto_ptr<_Up> __p) {
+    reset(__p.release());
+    return *this;
+  }
+#endif
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+  unique_ptr(unique_ptr const&) = delete;
+  unique_ptr& operator=(unique_ptr const&) = delete;
+#endif
+
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  ~unique_ptr() { reset(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr& operator=(nullptr_t) _NOEXCEPT {
+    reset();
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename add_lvalue_reference<_Tp>::type
+  operator*() const {
+    return *__ptr_.first();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  pointer operator->() const _NOEXCEPT {
+    return __ptr_.first();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  pointer get() const _NOEXCEPT {
+    return __ptr_.first();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  deleter_type& get_deleter() _NOEXCEPT {
+    return __ptr_.second();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const deleter_type& get_deleter() const _NOEXCEPT {
+    return __ptr_.second();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT {
+    return __ptr_.first() != nullptr;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  pointer release() _NOEXCEPT {
+    pointer __t = __ptr_.first();
+    __ptr_.first() = pointer();
+    return __t;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void reset(pointer __p = pointer()) _NOEXCEPT {
+    pointer __tmp = __ptr_.first();
+    __ptr_.first() = __p;
+    if (__tmp)
+      __ptr_.second()(__tmp);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void swap(unique_ptr& __u) _NOEXCEPT {
+    __ptr_.swap(__u.__ptr_);
+  }
+};
+
+
+template <class _Tp, class _Dp>
+class _LIBCUDACXX_TEMPLATE_VIS unique_ptr<_Tp[], _Dp> {
+public:
+  typedef _Tp element_type;
+  typedef _Dp deleter_type;
+  typedef typename __pointer_type<_Tp, deleter_type>::type pointer;
+
+private:
+  __compressed_pair<pointer, deleter_type> __ptr_;
+
+  template <class _From>
+  struct _CheckArrayPointerConversion : is_same<_From, pointer> {};
+
+  template <class _FromElem>
+  struct _CheckArrayPointerConversion<_FromElem*>
+      : integral_constant<bool,
+          is_same<_FromElem*, pointer>::value ||
+            (is_same<pointer, element_type*>::value &&
+             is_convertible<_FromElem(*)[], element_type(*)[]>::value)
+      >
+  {};
+
+  typedef __unique_ptr_deleter_sfinae<_Dp> _DeleterSFINAE;
+
+  template <bool _Dummy>
+  using _LValRefType _LIBCUDACXX_NODEBUG_TYPE =
+      typename __dependent_type<_DeleterSFINAE, _Dummy>::__lval_ref_type;
+
+  template <bool _Dummy>
+  using _GoodRValRefType _LIBCUDACXX_NODEBUG_TYPE =
+      typename __dependent_type<_DeleterSFINAE, _Dummy>::__good_rval_ref_type;
+
+  template <bool _Dummy>
+  using _BadRValRefType _LIBCUDACXX_NODEBUG_TYPE =
+      typename __dependent_type<_DeleterSFINAE, _Dummy>::__bad_rval_ref_type;
+
+  template <bool _Dummy, class _Deleter = typename __dependent_type<
+                             __identity<deleter_type>, _Dummy>::type>
+  using _EnableIfDeleterDefaultConstructible _LIBCUDACXX_NODEBUG_TYPE  =
+      typename enable_if<is_default_constructible<_Deleter>::value &&
+                         !is_pointer<_Deleter>::value>::type;
+
+  template <class _ArgType>
+  using _EnableIfDeleterConstructible _LIBCUDACXX_NODEBUG_TYPE  =
+      typename enable_if<is_constructible<deleter_type, _ArgType>::value>::type;
+
+  template <class _Pp>
+  using _EnableIfPointerConvertible _LIBCUDACXX_NODEBUG_TYPE  = typename enable_if<
+      _CheckArrayPointerConversion<_Pp>::value
+  >::type;
+
+  template <class _UPtr, class _Up,
+        class _ElemT = typename _UPtr::element_type>
+  using _EnableIfMoveConvertible _LIBCUDACXX_NODEBUG_TYPE  = typename enable_if<
+      is_array<_Up>::value &&
+      is_same<pointer, element_type*>::value &&
+      is_same<typename _UPtr::pointer, _ElemT*>::value &&
+      is_convertible<_ElemT(*)[], element_type(*)[]>::value
+    >::type;
+
+  template <class _UDel>
+  using _EnableIfDeleterConvertible _LIBCUDACXX_NODEBUG_TYPE  = typename enable_if<
+      (is_reference<_Dp>::value && is_same<_Dp, _UDel>::value) ||
+      (!is_reference<_Dp>::value && is_convertible<_UDel, _Dp>::value)
+    >::type;
+
+  template <class _UDel>
+  using _EnableIfDeleterAssignable _LIBCUDACXX_NODEBUG_TYPE  = typename enable_if<
+      is_assignable<_Dp&, _UDel&&>::value
+    >::type;
+
+public:
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterDefaultConstructible<_Dummy> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_CONSTEXPR unique_ptr() _NOEXCEPT : __ptr_(pointer()) {}
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterDefaultConstructible<_Dummy> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_CONSTEXPR unique_ptr(nullptr_t) _NOEXCEPT : __ptr_(pointer()) {}
+
+  template <class _Pp, bool _Dummy = true,
+            class = _EnableIfDeleterDefaultConstructible<_Dummy>,
+            class = _EnableIfPointerConvertible<_Pp> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  explicit unique_ptr(_Pp __p) _NOEXCEPT
+      : __ptr_(__p) {}
+
+  template <class _Pp, bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_LValRefType<_Dummy> >,
+            class = _EnableIfPointerConvertible<_Pp> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(_Pp __p, _LValRefType<_Dummy> __d) _NOEXCEPT
+      : __ptr_(__p, __d) {}
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_LValRefType<_Dummy> > >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(nullptr_t, _LValRefType<_Dummy> __d) _NOEXCEPT
+      : __ptr_(nullptr, __d) {}
+
+  template <class _Pp, bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_GoodRValRefType<_Dummy> >,
+            class = _EnableIfPointerConvertible<_Pp> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(_Pp __p, _GoodRValRefType<_Dummy> __d) _NOEXCEPT
+      : __ptr_(__p, _CUDA_VSTD::move(__d)) {
+    static_assert(!is_reference<deleter_type>::value,
+                  "rvalue deleter bound to reference");
+  }
+
+  template <bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_GoodRValRefType<_Dummy> > >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(nullptr_t, _GoodRValRefType<_Dummy> __d) _NOEXCEPT
+      : __ptr_(nullptr, _CUDA_VSTD::move(__d)) {
+    static_assert(!is_reference<deleter_type>::value,
+                  "rvalue deleter bound to reference");
+  }
+
+  template <class _Pp, bool _Dummy = true,
+            class = _EnableIfDeleterConstructible<_BadRValRefType<_Dummy> >,
+            class = _EnableIfPointerConvertible<_Pp> >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(_Pp __p, _BadRValRefType<_Dummy> __d) = delete;
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(unique_ptr&& __u) _NOEXCEPT
+      : __ptr_(__u.release(), _CUDA_VSTD::forward<deleter_type>(__u.get_deleter())) {
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr& operator=(unique_ptr&& __u) _NOEXCEPT {
+    reset(__u.release());
+    __ptr_.second() = _CUDA_VSTD::forward<deleter_type>(__u.get_deleter());
+    return *this;
+  }
+
+  template <class _Up, class _Ep,
+      class = _EnableIfMoveConvertible<unique_ptr<_Up, _Ep>, _Up>,
+      class = _EnableIfDeleterConvertible<_Ep>
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr(unique_ptr<_Up, _Ep>&& __u) _NOEXCEPT
+      : __ptr_(__u.release(), _CUDA_VSTD::forward<_Ep>(__u.get_deleter())) {
+  }
+
+  template <class _Up, class _Ep,
+      class = _EnableIfMoveConvertible<unique_ptr<_Up, _Ep>, _Up>,
+      class = _EnableIfDeleterAssignable<_Ep>
+  >
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr&
+  operator=(unique_ptr<_Up, _Ep>&& __u) _NOEXCEPT {
+    reset(__u.release());
+    __ptr_.second() = _CUDA_VSTD::forward<_Ep>(__u.get_deleter());
+    return *this;
+  }
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+  unique_ptr(unique_ptr const&) = delete;
+  unique_ptr& operator=(unique_ptr const&) = delete;
+#endif
+
+public:
+  _LIBCUDACXX_INLINE_VISIBILITY
+  ~unique_ptr() { reset(); }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  unique_ptr& operator=(nullptr_t) _NOEXCEPT {
+    reset();
+    return *this;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename add_lvalue_reference<_Tp>::type
+  operator[](size_t __i) const {
+    return __ptr_.first()[__i];
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  pointer get() const _NOEXCEPT {
+    return __ptr_.first();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  deleter_type& get_deleter() _NOEXCEPT {
+    return __ptr_.second();
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  const deleter_type& get_deleter() const _NOEXCEPT {
+    return __ptr_.second();
+  }
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT {
+    return __ptr_.first() != nullptr;
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  pointer release() _NOEXCEPT {
+    pointer __t = __ptr_.first();
+    __ptr_.first() = pointer();
+    return __t;
+  }
+
+  template <class _Pp>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  typename enable_if<
+      _CheckArrayPointerConversion<_Pp>::value
+  >::type
+  reset(_Pp __p) _NOEXCEPT {
+    pointer __tmp = __ptr_.first();
+    __ptr_.first() = __p;
+    if (__tmp)
+      __ptr_.second()(__tmp);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void reset(nullptr_t = nullptr) _NOEXCEPT {
+    pointer __tmp = __ptr_.first();
+    __ptr_.first() = nullptr;
+    if (__tmp)
+      __ptr_.second()(__tmp);
+  }
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void swap(unique_ptr& __u) _NOEXCEPT {
+    __ptr_.swap(__u.__ptr_);
+  }
+
+};
+
+template <class _Tp, class _Dp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<
+    __is_swappable<_Dp>::value,
+    void
+>::type
+swap(unique_ptr<_Tp, _Dp>& __x, unique_ptr<_Tp, _Dp>& __y) _NOEXCEPT {__x.swap(__y);}
+
+template <class _T1, class _D1, class _T2, class _D2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const unique_ptr<_T1, _D1>& __x, const unique_ptr<_T2, _D2>& __y) {return __x.get() == __y.get();}
+
+template <class _T1, class _D1, class _T2, class _D2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const unique_ptr<_T1, _D1>& __x, const unique_ptr<_T2, _D2>& __y) {return !(__x == __y);}
+
+template <class _T1, class _D1, class _T2, class _D2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const unique_ptr<_T1, _D1>& __x, const unique_ptr<_T2, _D2>& __y)
+{
+    typedef typename unique_ptr<_T1, _D1>::pointer _P1;
+    typedef typename unique_ptr<_T2, _D2>::pointer _P2;
+    typedef typename common_type<_P1, _P2>::type _Vp;
+    return less<_Vp>()(__x.get(), __y.get());
+}
+
+template <class _T1, class _D1, class _T2, class _D2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const unique_ptr<_T1, _D1>& __x, const unique_ptr<_T2, _D2>& __y) {return __y < __x;}
+
+template <class _T1, class _D1, class _T2, class _D2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const unique_ptr<_T1, _D1>& __x, const unique_ptr<_T2, _D2>& __y) {return !(__y < __x);}
+
+template <class _T1, class _D1, class _T2, class _D2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const unique_ptr<_T1, _D1>& __x, const unique_ptr<_T2, _D2>& __y) {return !(__x < __y);}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const unique_ptr<_T1, _D1>& __x, nullptr_t) _NOEXCEPT
+{
+    return !__x;
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(nullptr_t, const unique_ptr<_T1, _D1>& __x) _NOEXCEPT
+{
+    return !__x;
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const unique_ptr<_T1, _D1>& __x, nullptr_t) _NOEXCEPT
+{
+    return static_cast<bool>(__x);
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(nullptr_t, const unique_ptr<_T1, _D1>& __x) _NOEXCEPT
+{
+    return static_cast<bool>(__x);
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const unique_ptr<_T1, _D1>& __x, nullptr_t)
+{
+    typedef typename unique_ptr<_T1, _D1>::pointer _P1;
+    return less<_P1>()(__x.get(), nullptr);
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(nullptr_t, const unique_ptr<_T1, _D1>& __x)
+{
+    typedef typename unique_ptr<_T1, _D1>::pointer _P1;
+    return less<_P1>()(nullptr, __x.get());
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const unique_ptr<_T1, _D1>& __x, nullptr_t)
+{
+    return nullptr < __x;
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(nullptr_t, const unique_ptr<_T1, _D1>& __x)
+{
+    return __x < nullptr;
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const unique_ptr<_T1, _D1>& __x, nullptr_t)
+{
+    return !(nullptr < __x);
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(nullptr_t, const unique_ptr<_T1, _D1>& __x)
+{
+    return !(__x < nullptr);
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const unique_ptr<_T1, _D1>& __x, nullptr_t)
+{
+    return !(__x < nullptr);
+}
+
+template <class _T1, class _D1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(nullptr_t, const unique_ptr<_T1, _D1>& __x)
+{
+    return !(nullptr < __x);
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template<class _Tp>
+struct __unique_if
+{
+    typedef unique_ptr<_Tp> __unique_single;
+};
+
+template<class _Tp>
+struct __unique_if<_Tp[]>
+{
+    typedef unique_ptr<_Tp[]> __unique_array_unknown_bound;
+};
+
+template<class _Tp, size_t _Np>
+struct __unique_if<_Tp[_Np]>
+{
+    typedef void __unique_array_known_bound;
+};
+
+template<class _Tp, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __unique_if<_Tp>::__unique_single
+make_unique(_Args&&... __args)
+{
+    return unique_ptr<_Tp>(new _Tp(_CUDA_VSTD::forward<_Args>(__args)...));
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename __unique_if<_Tp>::__unique_array_unknown_bound
+make_unique(size_t __n)
+{
+    typedef typename remove_extent<_Tp>::type _Up;
+    return unique_ptr<_Tp>(new _Up[__n]());
+}
+
+template<class _Tp, class... _Args>
+    typename __unique_if<_Tp>::__unique_array_known_bound
+    make_unique(_Args&&...) = delete;
+
+#endif  // _LIBCUDACXX_STD_VER > 11
+
+template <class _Tp, class _Dp>
+#ifdef _LIBCUDACXX_CXX03_LANG
+struct _LIBCUDACXX_TEMPLATE_VIS hash<unique_ptr<_Tp, _Dp> >
+#else
+struct _LIBCUDACXX_TEMPLATE_VIS hash<__enable_hash_helper<
+    unique_ptr<_Tp, _Dp>, typename unique_ptr<_Tp, _Dp>::pointer> >
+#endif
+{
+    typedef unique_ptr<_Tp, _Dp> argument_type;
+    typedef size_t               result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator()(const argument_type& __ptr) const
+    {
+        typedef typename argument_type::pointer pointer;
+        return hash<pointer>()(__ptr.get());
+    }
+};
+
+struct __destruct_n
+{
+private:
+    size_t __size_;
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY void __process(_Tp* __p, false_type) _NOEXCEPT
+        {for (size_t __i = 0; __i < __size_; ++__i, ++__p) __p->~_Tp();}
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY void __process(_Tp*, true_type) _NOEXCEPT
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void __incr(false_type) _NOEXCEPT
+        {++__size_;}
+    _LIBCUDACXX_INLINE_VISIBILITY void __incr(true_type) _NOEXCEPT
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void __set(size_t __s, false_type) _NOEXCEPT
+        {__size_ = __s;}
+    _LIBCUDACXX_INLINE_VISIBILITY void __set(size_t, true_type) _NOEXCEPT
+        {}
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit __destruct_n(size_t __s) _NOEXCEPT
+        : __size_(__s) {}
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY void __incr(_Tp*) _NOEXCEPT
+        {__incr(integral_constant<bool, is_trivially_destructible<_Tp>::value>());}
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY void __set(size_t __s, _Tp*) _NOEXCEPT
+        {__set(__s, integral_constant<bool, is_trivially_destructible<_Tp>::value>());}
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY void operator()(_Tp* __p) _NOEXCEPT
+        {__process(__p, integral_constant<bool, is_trivially_destructible<_Tp>::value>());}
+};
+
+template <class _Alloc>
+class __allocator_destructor
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE allocator_traits<_Alloc> __alloc_traits;
+public:
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __alloc_traits::pointer pointer;
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __alloc_traits::size_type size_type;
+private:
+    _Alloc& __alloc_;
+    size_type __s_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY __allocator_destructor(_Alloc& __a, size_type __s)
+             _NOEXCEPT
+        : __alloc_(__a), __s_(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p) _NOEXCEPT
+        {__alloc_traits::deallocate(__alloc_, __p, __s_);}
+};
+
+template <class _InputIterator, class _ForwardIterator>
+_ForwardIterator
+uninitialized_copy(_InputIterator __f, _InputIterator __l, _ForwardIterator __r)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    _ForwardIterator __s = __r;
+    try
+    {
+#endif
+        for (; __f != __l; ++__f, (void) ++__r)
+            ::new (static_cast<void*>(_CUDA_VSTD::addressof(*__r))) value_type(*__f);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        for (; __s != __r; ++__s)
+            __s->~value_type();
+        throw;
+    }
+#endif
+    return __r;
+}
+
+template <class _InputIterator, class _Size, class _ForwardIterator>
+_ForwardIterator
+uninitialized_copy_n(_InputIterator __f, _Size __n, _ForwardIterator __r)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    _ForwardIterator __s = __r;
+    try
+    {
+#endif
+        for (; __n > 0; ++__f, (void) ++__r, (void) --__n)
+            ::new (static_cast<void*>(_CUDA_VSTD::addressof(*__r))) value_type(*__f);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        for (; __s != __r; ++__s)
+            __s->~value_type();
+        throw;
+    }
+#endif
+    return __r;
+}
+
+template <class _ForwardIterator, class _Tp>
+void
+uninitialized_fill(_ForwardIterator __f, _ForwardIterator __l, const _Tp& __x)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    _ForwardIterator __s = __f;
+    try
+    {
+#endif
+        for (; __f != __l; ++__f)
+            ::new (static_cast<void*>(_CUDA_VSTD::addressof(*__f))) value_type(__x);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        for (; __s != __f; ++__s)
+            __s->~value_type();
+        throw;
+    }
+#endif
+}
+
+template <class _ForwardIterator, class _Size, class _Tp>
+_ForwardIterator
+uninitialized_fill_n(_ForwardIterator __f, _Size __n, const _Tp& __x)
+{
+    typedef typename iterator_traits<_ForwardIterator>::value_type value_type;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    _ForwardIterator __s = __f;
+    try
+    {
+#endif
+        for (; __n > 0; ++__f, (void) --__n)
+            ::new (static_cast<void*>(_CUDA_VSTD::addressof(*__f))) value_type(__x);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        for (; __s != __f; ++__s)
+            __s->~value_type();
+        throw;
+    }
+#endif
+    return __f;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void destroy_at(_Tp* __loc) {
+    _LIBCUDACXX_ASSERT(__loc, "null pointer given to destroy_at");
+    __loc->~_Tp();
+}
+
+template <class _ForwardIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void destroy(_ForwardIterator __first, _ForwardIterator __last) {
+    for (; __first != __last; ++__first)
+        _CUDA_VSTD::destroy_at(_CUDA_VSTD::addressof(*__first));
+}
+
+template <class _ForwardIterator, class _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator destroy_n(_ForwardIterator __first, _Size __n) {
+    for (; __n > 0; (void)++__first, --__n)
+        _CUDA_VSTD::destroy_at(_CUDA_VSTD::addressof(*__first));
+    return __first;
+}
+
+template <class _ForwardIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void uninitialized_default_construct(_ForwardIterator __first, _ForwardIterator __last) {
+    using _Vt = typename iterator_traits<_ForwardIterator>::value_type;
+    auto __idx = __first;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try {
+#endif
+    for (; __idx != __last; ++__idx)
+        ::new((void*)_CUDA_VSTD::addressof(*__idx)) _Vt;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    } catch (...) {
+        _CUDA_VSTD::destroy(__first, __idx);
+        throw;
+    }
+#endif
+}
+
+template <class _ForwardIterator, class _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator uninitialized_default_construct_n(_ForwardIterator __first, _Size __n) {
+    using _Vt = typename iterator_traits<_ForwardIterator>::value_type;
+    auto __idx = __first;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try {
+#endif
+    for (; __n > 0; (void)++__idx, --__n)
+        ::new((void*)_CUDA_VSTD::addressof(*__idx)) _Vt;
+    return __idx;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    } catch (...) {
+        _CUDA_VSTD::destroy(__first, __idx);
+        throw;
+    }
+#endif
+}
+
+
+template <class _ForwardIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void uninitialized_value_construct(_ForwardIterator __first, _ForwardIterator __last) {
+    using _Vt = typename iterator_traits<_ForwardIterator>::value_type;
+    auto __idx = __first;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try {
+#endif
+    for (; __idx != __last; ++__idx)
+        ::new((void*)_CUDA_VSTD::addressof(*__idx)) _Vt();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    } catch (...) {
+        _CUDA_VSTD::destroy(__first, __idx);
+        throw;
+    }
+#endif
+}
+
+template <class _ForwardIterator, class _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIterator uninitialized_value_construct_n(_ForwardIterator __first, _Size __n) {
+    using _Vt = typename iterator_traits<_ForwardIterator>::value_type;
+    auto __idx = __first;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try {
+#endif
+    for (; __n > 0; (void)++__idx, --__n)
+        ::new((void*)_CUDA_VSTD::addressof(*__idx)) _Vt();
+    return __idx;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    } catch (...) {
+        _CUDA_VSTD::destroy(__first, __idx);
+        throw;
+    }
+#endif
+}
+
+
+template <class _InputIt, class _ForwardIt>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_ForwardIt uninitialized_move(_InputIt __first, _InputIt __last, _ForwardIt __first_res) {
+    using _Vt = typename iterator_traits<_ForwardIt>::value_type;
+    auto __idx = __first_res;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try {
+#endif
+    for (; __first != __last; (void)++__idx, ++__first)
+        ::new((void*)_CUDA_VSTD::addressof(*__idx)) _Vt(std::move(*__first));
+    return __idx;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    } catch (...) {
+        _CUDA_VSTD::destroy(__first_res, __idx);
+        throw;
+    }
+#endif
+}
+
+template <class _InputIt, class _Size, class _ForwardIt>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+pair<_InputIt, _ForwardIt>
+uninitialized_move_n(_InputIt __first, _Size __n, _ForwardIt __first_res) {
+    using _Vt = typename iterator_traits<_ForwardIt>::value_type;
+    auto __idx = __first_res;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try {
+#endif
+    for (; __n > 0; ++__idx, (void)++__first, --__n)
+        ::new((void*)_CUDA_VSTD::addressof(*__idx)) _Vt(std::move(*__first));
+    return {__first, __idx};
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    } catch (...) {
+        _CUDA_VSTD::destroy(__first_res, __idx);
+        throw;
+    }
+#endif
+}
+
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+// NOTE: Relaxed and acq/rel atomics (for increment and decrement respectively)
+// should be sufficient for thread safety.
+// See https://bugs.llvm.org/show_bug.cgi?id=22803
+#if defined(__clang__) && __has_builtin(__atomic_add_fetch)          \
+                       && defined(__ATOMIC_RELAXED)                  \
+                       && defined(__ATOMIC_ACQ_REL)
+#   define _LIBCUDACXX_HAS_BUILTIN_ATOMIC_SUPPORT
+#elif defined(_LIBCUDACXX_COMPILER_GCC)
+#   define _LIBCUDACXX_HAS_BUILTIN_ATOMIC_SUPPORT
+#endif
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _Tp
+__libcpp_atomic_refcount_increment(_Tp& __t) _NOEXCEPT
+{
+#if defined(_LIBCUDACXX_HAS_BUILTIN_ATOMIC_SUPPORT) && !defined(_LIBCUDACXX_HAS_NO_THREADS)
+    return __atomic_add_fetch(&__t, 1, __ATOMIC_RELAXED);
+#else
+    return __t += 1;
+#endif
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _Tp
+__libcpp_atomic_refcount_decrement(_Tp& __t) _NOEXCEPT
+{
+#if defined(_LIBCUDACXX_HAS_BUILTIN_ATOMIC_SUPPORT) && !defined(_LIBCUDACXX_HAS_NO_THREADS)
+    return __atomic_add_fetch(&__t, -1, __ATOMIC_ACQ_REL);
+#else
+    return __t -= 1;
+#endif
+}
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_weak_ptr
+    : public std::exception
+{
+public:
+    virtual ~bad_weak_ptr() _NOEXCEPT;
+    virtual const char* what() const  _NOEXCEPT;
+};
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_bad_weak_ptr()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw bad_weak_ptr();
+#else
+    _CUDA_VSTD::abort();
+#endif
+}
+
+template<class _Tp> class _LIBCUDACXX_TEMPLATE_VIS weak_ptr;
+
+class _LIBCUDACXX_TYPE_VIS __shared_count
+{
+    __shared_count(const __shared_count&);
+    __shared_count& operator=(const __shared_count&);
+
+protected:
+    long __shared_owners_;
+    virtual ~__shared_count();
+private:
+    virtual void __on_zero_shared() _NOEXCEPT = 0;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __shared_count(long __refs = 0) _NOEXCEPT
+        : __shared_owners_(__refs) {}
+
+#if defined(_LIBCUDACXX_BUILDING_LIBRARY) && \
+    defined(_LIBCUDACXX_DEPRECATED_ABI_LEGACY_LIBRARY_DEFINITIONS_FOR_INLINE_FUNCTIONS)
+    void __add_shared() _NOEXCEPT;
+    bool __release_shared() _NOEXCEPT;
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_shared() _NOEXCEPT {
+      __libcpp_atomic_refcount_increment(__shared_owners_);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __release_shared() _NOEXCEPT {
+      if (__libcpp_atomic_refcount_decrement(__shared_owners_) == -1) {
+        __on_zero_shared();
+        return true;
+      }
+      return false;
+    }
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    long use_count() const _NOEXCEPT {
+        return __libcpp_relaxed_load(&__shared_owners_) + 1;
+    }
+};
+
+class _LIBCUDACXX_TYPE_VIS __shared_weak_count
+    : private __shared_count
+{
+    long __shared_weak_owners_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __shared_weak_count(long __refs = 0) _NOEXCEPT
+        : __shared_count(__refs),
+          __shared_weak_owners_(__refs) {}
+protected:
+    virtual ~__shared_weak_count();
+
+public:
+#if defined(_LIBCUDACXX_BUILDING_LIBRARY) && \
+    defined(_LIBCUDACXX_DEPRECATED_ABI_LEGACY_LIBRARY_DEFINITIONS_FOR_INLINE_FUNCTIONS)
+    void __add_shared() _NOEXCEPT;
+    void __add_weak() _NOEXCEPT;
+    void __release_shared() _NOEXCEPT;
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_shared() _NOEXCEPT {
+      __shared_count::__add_shared();
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_weak() _NOEXCEPT {
+      __libcpp_atomic_refcount_increment(__shared_weak_owners_);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __release_shared() _NOEXCEPT {
+      if (__shared_count::__release_shared())
+        __release_weak();
+    }
+#endif
+    void __release_weak() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    long use_count() const _NOEXCEPT {return __shared_count::use_count();}
+    __shared_weak_count* lock() _NOEXCEPT;
+
+    // Define the function out only if we build static libc++ without RTTI.
+    // Otherwise we may break clients who need to compile their projects with
+    // -fno-rtti and yet link against a libc++.dylib compiled
+    // without -fno-rtti.
+#if !defined(_LIBCUDACXX_NO_RTTI) || !defined(_LIBCUDACXX_BUILD_STATIC)
+    virtual const void* __get_deleter(const type_info&) const _NOEXCEPT;
+#endif
+private:
+    virtual void __on_zero_shared_weak() _NOEXCEPT = 0;
+};
+
+template <class _Tp, class _Dp, class _Alloc>
+class __shared_ptr_pointer
+    : public __shared_weak_count
+{
+    __compressed_pair<__compressed_pair<_Tp, _Dp>, _Alloc> __data_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __shared_ptr_pointer(_Tp __p, _Dp __d, _Alloc __a)
+        :  __data_(__compressed_pair<_Tp, _Dp>(__p, _CUDA_VSTD::move(__d)), _CUDA_VSTD::move(__a)) {}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    virtual const void* __get_deleter(const type_info&) const _NOEXCEPT;
+#endif
+
+private:
+    virtual void __on_zero_shared() _NOEXCEPT;
+    virtual void __on_zero_shared_weak() _NOEXCEPT;
+};
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template <class _Tp, class _Dp, class _Alloc>
+const void*
+__shared_ptr_pointer<_Tp, _Dp, _Alloc>::__get_deleter(const type_info& __t) const _NOEXCEPT
+{
+    return __t == typeid(_Dp) ? _CUDA_VSTD::addressof(__data_.first().second()) : nullptr;
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template <class _Tp, class _Dp, class _Alloc>
+void
+__shared_ptr_pointer<_Tp, _Dp, _Alloc>::__on_zero_shared() _NOEXCEPT
+{
+    __data_.first().second()(__data_.first().first());
+    __data_.first().second().~_Dp();
+}
+
+template <class _Tp, class _Dp, class _Alloc>
+void
+__shared_ptr_pointer<_Tp, _Dp, _Alloc>::__on_zero_shared_weak() _NOEXCEPT
+{
+    typedef typename __allocator_traits_rebind<_Alloc, __shared_ptr_pointer>::type _Al;
+    typedef allocator_traits<_Al> _ATraits;
+    typedef pointer_traits<typename _ATraits::pointer> _PTraits;
+
+    _Al __a(__data_.second());
+    __data_.second().~_Alloc();
+    __a.deallocate(_PTraits::pointer_to(*this), 1);
+}
+
+template <class _Tp, class _Alloc>
+class __shared_ptr_emplace
+    : public __shared_weak_count
+{
+    __compressed_pair<_Alloc, _Tp> __data_;
+public:
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __shared_ptr_emplace(_Alloc __a)
+        :  __data_(_CUDA_VSTD::move(__a)) {}
+
+    template <class ..._Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __shared_ptr_emplace(_Alloc __a, _Args&& ...__args)
+            :  __data_(piecewise_construct, _CUDA_VSTD::forward_as_tuple(__a),
+                   _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...)) {}
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __shared_ptr_emplace(_Alloc __a)
+        :  __data_(__a) {}
+
+    template <class _A0>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __shared_ptr_emplace(_Alloc __a, _A0& __a0)
+            :  __data_(__a, _Tp(__a0)) {}
+
+    template <class _A0, class _A1>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __shared_ptr_emplace(_Alloc __a, _A0& __a0, _A1& __a1)
+            :  __data_(__a, _Tp(__a0, __a1)) {}
+
+    template <class _A0, class _A1, class _A2>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __shared_ptr_emplace(_Alloc __a, _A0& __a0, _A1& __a1, _A2& __a2)
+            :  __data_(__a, _Tp(__a0, __a1, __a2)) {}
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+private:
+    virtual void __on_zero_shared() _NOEXCEPT;
+    virtual void __on_zero_shared_weak() _NOEXCEPT;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp* get() _NOEXCEPT {return _CUDA_VSTD::addressof(__data_.second());}
+};
+
+template <class _Tp, class _Alloc>
+void
+__shared_ptr_emplace<_Tp, _Alloc>::__on_zero_shared() _NOEXCEPT
+{
+    __data_.second().~_Tp();
+}
+
+template <class _Tp, class _Alloc>
+void
+__shared_ptr_emplace<_Tp, _Alloc>::__on_zero_shared_weak() _NOEXCEPT
+{
+    typedef typename __allocator_traits_rebind<_Alloc, __shared_ptr_emplace>::type _Al;
+    typedef allocator_traits<_Al> _ATraits;
+    typedef pointer_traits<typename _ATraits::pointer> _PTraits;
+    _Al __a(__data_.first());
+    __data_.first().~_Alloc();
+    __a.deallocate(_PTraits::pointer_to(*this), 1);
+}
+
+struct __shared_ptr_dummy_rebind_allocator_type;
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS allocator<__shared_ptr_dummy_rebind_allocator_type>
+{
+public:
+    template <class _Other>
+    struct rebind
+    {
+        typedef allocator<_Other> other;
+    };
+};
+
+template<class _Tp> class _LIBCUDACXX_TEMPLATE_VIS enable_shared_from_this;
+
+template<class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS shared_ptr
+{
+public:
+    typedef _Tp element_type;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef weak_ptr<_Tp> weak_type;
+#endif
+private:
+    element_type*      __ptr_;
+    __shared_weak_count* __cntrl_;
+
+    struct __nat {int __for_bool_;};
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR shared_ptr() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR shared_ptr(nullptr_t) _NOEXCEPT;
+    template<class _Yp>
+        explicit shared_ptr(_Yp* __p,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat());
+    template<class _Yp, class _Dp>
+        shared_ptr(_Yp* __p, _Dp __d,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat());
+    template<class _Yp, class _Dp, class _Alloc>
+        shared_ptr(_Yp* __p, _Dp __d, _Alloc __a,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat());
+    template <class _Dp> shared_ptr(nullptr_t __p, _Dp __d);
+    template <class _Dp, class _Alloc> shared_ptr(nullptr_t __p, _Dp __d, _Alloc __a);
+    template<class _Yp> _LIBCUDACXX_INLINE_VISIBILITY shared_ptr(const shared_ptr<_Yp>& __r, element_type* __p) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_ptr(const shared_ptr& __r) _NOEXCEPT;
+    template<class _Yp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        shared_ptr(const shared_ptr<_Yp>& __r,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat())
+                       _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_ptr(shared_ptr&& __r) _NOEXCEPT;
+    template<class _Yp> _LIBCUDACXX_INLINE_VISIBILITY  shared_ptr(shared_ptr<_Yp>&& __r,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat())
+                       _NOEXCEPT;
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template<class _Yp> explicit shared_ptr(const weak_ptr<_Yp>& __r,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type= __nat());
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template<class _Yp>
+        shared_ptr(auto_ptr<_Yp>&& __r,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat());
+#else
+    template<class _Yp>
+        shared_ptr(auto_ptr<_Yp> __r,
+                   typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type = __nat());
+#endif
+#endif
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template <class _Yp, class _Dp>
+        shared_ptr(unique_ptr<_Yp, _Dp>&&,
+                   typename enable_if
+                   <
+                       !is_lvalue_reference<_Dp>::value &&
+                       !is_array<_Yp>::value &&
+                       is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+                       __nat
+                   >::type = __nat());
+    template <class _Yp, class _Dp>
+        shared_ptr(unique_ptr<_Yp, _Dp>&&,
+                   typename enable_if
+                   <
+                       is_lvalue_reference<_Dp>::value &&
+                       !is_array<_Yp>::value &&
+                       is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+                       __nat
+                   >::type = __nat());
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    template <class _Yp, class _Dp>
+        shared_ptr(unique_ptr<_Yp, _Dp>,
+                   typename enable_if
+                   <
+                       !is_lvalue_reference<_Dp>::value &&
+                       !is_array<_Yp>::value &&
+                       is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+                       __nat
+                   >::type = __nat());
+    template <class _Yp, class _Dp>
+        shared_ptr(unique_ptr<_Yp, _Dp>,
+                   typename enable_if
+                   <
+                       is_lvalue_reference<_Dp>::value &&
+                       !is_array<_Yp>::value &&
+                       is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+                       __nat
+                   >::type = __nat());
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+    ~shared_ptr();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_ptr& operator=(const shared_ptr& __r) _NOEXCEPT;
+    template<class _Yp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            shared_ptr&
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(const shared_ptr<_Yp>& __r) _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_ptr& operator=(shared_ptr&& __r) _NOEXCEPT;
+    template<class _Yp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            shared_ptr<_Tp>&
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(shared_ptr<_Yp>&& __r);
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+    template<class _Yp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            !is_array<_Yp>::value &&
+            is_convertible<_Yp*, element_type*>::value,
+            shared_ptr
+        >::type&
+        operator=(auto_ptr<_Yp>&& __r);
+#endif
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+    template<class _Yp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            !is_array<_Yp>::value &&
+            is_convertible<_Yp*, element_type*>::value,
+            shared_ptr&
+        >::type
+        operator=(auto_ptr<_Yp> __r);
+#endif
+#endif
+    template <class _Yp, class _Dp>
+        typename enable_if
+        <
+            !is_array<_Yp>::value &&
+            is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+            shared_ptr&
+        >::type
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(unique_ptr<_Yp, _Dp>&& __r);
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(unique_ptr<_Yp, _Dp> __r);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(shared_ptr& __r) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() _NOEXCEPT;
+    template<class _Yp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            void
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        reset(_Yp* __p);
+    template<class _Yp, class _Dp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            void
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        reset(_Yp* __p, _Dp __d);
+    template<class _Yp, class _Dp, class _Alloc>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            void
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        reset(_Yp* __p, _Dp __d, _Alloc __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    element_type* get() const _NOEXCEPT {return __ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename add_lvalue_reference<element_type>::type operator*() const _NOEXCEPT
+        {return *__ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    element_type* operator->() const _NOEXCEPT {return __ptr_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    long use_count() const _NOEXCEPT {return __cntrl_ ? __cntrl_->use_count() : 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool unique() const _NOEXCEPT {return use_count() == 1;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_EXPLICIT operator bool() const _NOEXCEPT {return get() != 0;}
+    template <class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool owner_before(shared_ptr<_Up> const& __p) const _NOEXCEPT
+        {return __cntrl_ < __p.__cntrl_;}
+    template <class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool owner_before(weak_ptr<_Up> const& __p) const _NOEXCEPT
+        {return __cntrl_ < __p.__cntrl_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool
+    __owner_equivalent(const shared_ptr& __p) const
+        {return __cntrl_ == __p.__cntrl_;}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+    template <class _Dp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        _Dp* __get_deleter() const _NOEXCEPT
+            {return static_cast<_Dp*>(__cntrl_
+                    ? const_cast<void *>(__cntrl_->__get_deleter(typeid(_Dp)))
+                      : nullptr);}
+#endif  // _LIBCUDACXX_NO_RTTI
+
+    template<class ..._Args>
+        static
+        shared_ptr<_Tp>
+        make_shared(_Args&& ...__args);
+
+    template<class _Alloc, class ..._Args>
+        static
+        shared_ptr<_Tp>
+        allocate_shared(const _Alloc& __a, _Args&& ...__args);
+
+private:
+    template <class _Yp, bool = is_function<_Yp>::value>
+        struct __shared_ptr_default_allocator
+        {
+            typedef allocator<_Yp> type;
+        };
+
+    template <class _Yp>
+        struct __shared_ptr_default_allocator<_Yp, true>
+        {
+            typedef allocator<__shared_ptr_dummy_rebind_allocator_type> type;
+        };
+
+    template <class _Yp, class _OrigPtr>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if<is_convertible<_OrigPtr*,
+                                          const enable_shared_from_this<_Yp>*
+        >::value,
+            void>::type
+        __enable_weak_this(const enable_shared_from_this<_Yp>* __e,
+                           _OrigPtr* __ptr) _NOEXCEPT
+        {
+            typedef typename remove_cv<_Yp>::type _RawYp;
+            if (__e && __e->__weak_this_.expired())
+            {
+                __e->__weak_this_ = shared_ptr<_RawYp>(*this,
+                    const_cast<_RawYp*>(static_cast<const _Yp*>(__ptr)));
+            }
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY void __enable_weak_this(...) _NOEXCEPT {}
+
+    template <class _Up> friend class _LIBCUDACXX_TEMPLATE_VIS shared_ptr;
+    template <class _Up> friend class _LIBCUDACXX_TEMPLATE_VIS weak_ptr;
+};
+
+
+template<class _Tp>
+inline
+_LIBCUDACXX_CONSTEXPR
+shared_ptr<_Tp>::shared_ptr() _NOEXCEPT
+    : __ptr_(0),
+      __cntrl_(0)
+{
+}
+
+template<class _Tp>
+inline
+_LIBCUDACXX_CONSTEXPR
+shared_ptr<_Tp>::shared_ptr(nullptr_t) _NOEXCEPT
+    : __ptr_(0),
+      __cntrl_(0)
+{
+}
+
+template<class _Tp>
+template<class _Yp>
+shared_ptr<_Tp>::shared_ptr(_Yp* __p,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+    : __ptr_(__p)
+{
+    unique_ptr<_Yp> __hold(__p);
+    typedef typename __shared_ptr_default_allocator<_Yp>::type _AllocT;
+    typedef __shared_ptr_pointer<_Yp*, default_delete<_Yp>, _AllocT > _CntrlBlk;
+    __cntrl_ = new _CntrlBlk(__p, default_delete<_Yp>(), _AllocT());
+    __hold.release();
+    __enable_weak_this(__p, __p);
+}
+
+template<class _Tp>
+template<class _Yp, class _Dp>
+shared_ptr<_Tp>::shared_ptr(_Yp* __p, _Dp __d,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+    : __ptr_(__p)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typedef typename __shared_ptr_default_allocator<_Yp>::type _AllocT;
+        typedef __shared_ptr_pointer<_Yp*, _Dp, _AllocT > _CntrlBlk;
+        __cntrl_ = new _CntrlBlk(__p, __d, _AllocT());
+        __enable_weak_this(__p, __p);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __d(__p);
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class _Tp>
+template<class _Dp>
+shared_ptr<_Tp>::shared_ptr(nullptr_t __p, _Dp __d)
+    : __ptr_(0)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typedef typename __shared_ptr_default_allocator<_Tp>::type _AllocT;
+        typedef __shared_ptr_pointer<nullptr_t, _Dp, _AllocT > _CntrlBlk;
+        __cntrl_ = new _CntrlBlk(__p, __d, _AllocT());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __d(__p);
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class _Tp>
+template<class _Yp, class _Dp, class _Alloc>
+shared_ptr<_Tp>::shared_ptr(_Yp* __p, _Dp __d, _Alloc __a,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+    : __ptr_(__p)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typedef __shared_ptr_pointer<_Yp*, _Dp, _Alloc> _CntrlBlk;
+        typedef typename __allocator_traits_rebind<_Alloc, _CntrlBlk>::type _A2;
+        typedef __allocator_destructor<_A2> _D2;
+        _A2 __a2(__a);
+        unique_ptr<_CntrlBlk, _D2> __hold2(__a2.allocate(1), _D2(__a2, 1));
+        ::new(static_cast<void*>(_CUDA_VSTD::addressof(*__hold2.get())))
+            _CntrlBlk(__p, __d, __a);
+        __cntrl_ = _CUDA_VSTD::addressof(*__hold2.release());
+        __enable_weak_this(__p, __p);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __d(__p);
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class _Tp>
+template<class _Dp, class _Alloc>
+shared_ptr<_Tp>::shared_ptr(nullptr_t __p, _Dp __d, _Alloc __a)
+    : __ptr_(0)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typedef __shared_ptr_pointer<nullptr_t, _Dp, _Alloc> _CntrlBlk;
+        typedef typename __allocator_traits_rebind<_Alloc, _CntrlBlk>::type _A2;
+        typedef __allocator_destructor<_A2> _D2;
+        _A2 __a2(__a);
+        unique_ptr<_CntrlBlk, _D2> __hold2(__a2.allocate(1), _D2(__a2, 1));
+        ::new(static_cast<void*>(_CUDA_VSTD::addressof(*__hold2.get())))
+            _CntrlBlk(__p, __d, __a);
+        __cntrl_ = _CUDA_VSTD::addressof(*__hold2.release());
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __d(__p);
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+shared_ptr<_Tp>::shared_ptr(const shared_ptr<_Yp>& __r, element_type *__p) _NOEXCEPT
+    : __ptr_(__p),
+      __cntrl_(__r.__cntrl_)
+{
+    if (__cntrl_)
+        __cntrl_->__add_shared();
+}
+
+template<class _Tp>
+inline
+shared_ptr<_Tp>::shared_ptr(const shared_ptr& __r) _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    if (__cntrl_)
+        __cntrl_->__add_shared();
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+shared_ptr<_Tp>::shared_ptr(const shared_ptr<_Yp>& __r,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+         _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    if (__cntrl_)
+        __cntrl_->__add_shared();
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+inline
+shared_ptr<_Tp>::shared_ptr(shared_ptr&& __r) _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    __r.__ptr_ = 0;
+    __r.__cntrl_ = 0;
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+shared_ptr<_Tp>::shared_ptr(shared_ptr<_Yp>&& __r,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+         _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    __r.__ptr_ = 0;
+    __r.__cntrl_ = 0;
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+template<class _Tp>
+template<class _Yp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+shared_ptr<_Tp>::shared_ptr(auto_ptr<_Yp>&& __r,
+#else
+shared_ptr<_Tp>::shared_ptr(auto_ptr<_Yp> __r,
+#endif
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+    : __ptr_(__r.get())
+{
+    typedef __shared_ptr_pointer<_Yp*, default_delete<_Yp>, allocator<_Yp> > _CntrlBlk;
+    __cntrl_ = new _CntrlBlk(__r.get(), default_delete<_Yp>(), allocator<_Yp>());
+    __enable_weak_this(__r.get(), __r.get());
+    __r.release();
+}
+#endif
+
+template<class _Tp>
+template <class _Yp, class _Dp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+shared_ptr<_Tp>::shared_ptr(unique_ptr<_Yp, _Dp>&& __r,
+#else
+shared_ptr<_Tp>::shared_ptr(unique_ptr<_Yp, _Dp> __r,
+#endif
+                            typename enable_if
+                            <
+                                !is_lvalue_reference<_Dp>::value &&
+                                !is_array<_Yp>::value &&
+                                is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+                                __nat
+                            >::type)
+    : __ptr_(__r.get())
+{
+#if _LIBCUDACXX_STD_VER > 11
+    if (__ptr_ == nullptr)
+        __cntrl_ = nullptr;
+    else
+#endif
+    {
+        typedef typename __shared_ptr_default_allocator<_Yp>::type _AllocT;
+        typedef __shared_ptr_pointer<_Yp*, _Dp, _AllocT > _CntrlBlk;
+        __cntrl_ = new _CntrlBlk(__r.get(), __r.get_deleter(), _AllocT());
+        __enable_weak_this(__r.get(), __r.get());
+    }
+    __r.release();
+}
+
+template<class _Tp>
+template <class _Yp, class _Dp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+shared_ptr<_Tp>::shared_ptr(unique_ptr<_Yp, _Dp>&& __r,
+#else
+shared_ptr<_Tp>::shared_ptr(unique_ptr<_Yp, _Dp> __r,
+#endif
+                            typename enable_if
+                            <
+                                is_lvalue_reference<_Dp>::value &&
+                                !is_array<_Yp>::value &&
+                                is_convertible<typename unique_ptr<_Yp, _Dp>::pointer, element_type*>::value,
+                                __nat
+                            >::type)
+    : __ptr_(__r.get())
+{
+#if _LIBCUDACXX_STD_VER > 11
+    if (__ptr_ == nullptr)
+        __cntrl_ = nullptr;
+    else
+#endif
+    {
+        typedef typename __shared_ptr_default_allocator<_Yp>::type _AllocT;
+        typedef __shared_ptr_pointer<_Yp*,
+                                     reference_wrapper<typename remove_reference<_Dp>::type>,
+                                     _AllocT > _CntrlBlk;
+        __cntrl_ = new _CntrlBlk(__r.get(), ref(__r.get_deleter()), _AllocT());
+        __enable_weak_this(__r.get(), __r.get());
+    }
+    __r.release();
+}
+
+template<class _Tp>
+template<class ..._Args>
+shared_ptr<_Tp>
+shared_ptr<_Tp>::make_shared(_Args&& ...__args)
+{
+    static_assert( is_constructible<_Tp, _Args...>::value, "Can't construct object in make_shared" );
+    typedef __shared_ptr_emplace<_Tp, allocator<_Tp> > _CntrlBlk;
+    typedef allocator<_CntrlBlk> _A2;
+    typedef __allocator_destructor<_A2> _D2;
+    _A2 __a2;
+    unique_ptr<_CntrlBlk, _D2> __hold2(__a2.allocate(1), _D2(__a2, 1));
+    ::new(__hold2.get()) _CntrlBlk(__a2, _CUDA_VSTD::forward<_Args>(__args)...);
+    shared_ptr<_Tp> __r;
+    __r.__ptr_ = __hold2.get()->get();
+    __r.__cntrl_ = __hold2.release();
+    __r.__enable_weak_this(__r.__ptr_, __r.__ptr_);
+    return __r;
+}
+
+template<class _Tp>
+template<class _Alloc, class ..._Args>
+shared_ptr<_Tp>
+shared_ptr<_Tp>::allocate_shared(const _Alloc& __a, _Args&& ...__args)
+{
+    static_assert( is_constructible<_Tp, _Args...>::value, "Can't construct object in allocate_shared" );
+    typedef __shared_ptr_emplace<_Tp, _Alloc> _CntrlBlk;
+    typedef typename __allocator_traits_rebind<_Alloc, _CntrlBlk>::type _A2;
+    typedef __allocator_destructor<_A2> _D2;
+    _A2 __a2(__a);
+    unique_ptr<_CntrlBlk, _D2> __hold2(__a2.allocate(1), _D2(__a2, 1));
+    ::new(static_cast<void*>(_CUDA_VSTD::addressof(*__hold2.get())))
+        _CntrlBlk(__a, _CUDA_VSTD::forward<_Args>(__args)...);
+    shared_ptr<_Tp> __r;
+    __r.__ptr_ = __hold2.get()->get();
+    __r.__cntrl_ = _CUDA_VSTD::addressof(*__hold2.release());
+    __r.__enable_weak_this(__r.__ptr_, __r.__ptr_);
+    return __r;
+}
+
+template<class _Tp>
+shared_ptr<_Tp>::~shared_ptr()
+{
+    if (__cntrl_)
+        __cntrl_->__release_shared();
+}
+
+template<class _Tp>
+inline
+shared_ptr<_Tp>&
+shared_ptr<_Tp>::operator=(const shared_ptr& __r) _NOEXCEPT
+{
+    shared_ptr(__r).swap(*this);
+    return *this;
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    shared_ptr<_Tp>&
+>::type
+shared_ptr<_Tp>::operator=(const shared_ptr<_Yp>& __r) _NOEXCEPT
+{
+    shared_ptr(__r).swap(*this);
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+inline
+shared_ptr<_Tp>&
+shared_ptr<_Tp>::operator=(shared_ptr&& __r) _NOEXCEPT
+{
+    shared_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    shared_ptr<_Tp>&
+>::type
+shared_ptr<_Tp>::operator=(shared_ptr<_Yp>&& __r)
+{
+    shared_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    !is_array<_Yp>::value &&
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    shared_ptr<_Tp>
+>::type&
+shared_ptr<_Tp>::operator=(auto_ptr<_Yp>&& __r)
+{
+    shared_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+#endif
+
+template<class _Tp>
+template <class _Yp, class _Dp>
+inline
+typename enable_if
+<
+    !is_array<_Yp>::value &&
+    is_convertible<typename unique_ptr<_Yp, _Dp>::pointer,
+                   typename shared_ptr<_Tp>::element_type*>::value,
+    shared_ptr<_Tp>&
+>::type
+shared_ptr<_Tp>::operator=(unique_ptr<_Yp, _Dp>&& __r)
+{
+    shared_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+
+#else  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+#if _LIBCUDACXX_STD_VER <= 14 || defined(_LIBCUDACXX_ENABLE_CXX17_REMOVED_AUTO_PTR)
+template<class _Tp>
+template<class _Yp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_array<_Yp>::value &&
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    shared_ptr<_Tp>&
+>::type
+shared_ptr<_Tp>::operator=(auto_ptr<_Yp> __r)
+{
+    shared_ptr(__r).swap(*this);
+    return *this;
+}
+#endif
+
+template<class _Tp>
+template <class _Yp, class _Dp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_array<_Yp>::value &&
+    is_convertible<typename unique_ptr<_Yp, _Dp>::pointer,
+                   typename shared_ptr<_Tp>::element_type*>::value,
+    shared_ptr<_Tp>&
+>::type
+shared_ptr<_Tp>::operator=(unique_ptr<_Yp, _Dp> __r)
+{
+    shared_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+inline
+void
+shared_ptr<_Tp>::swap(shared_ptr& __r) _NOEXCEPT
+{
+    _CUDA_VSTD::swap(__ptr_, __r.__ptr_);
+    _CUDA_VSTD::swap(__cntrl_, __r.__cntrl_);
+}
+
+template<class _Tp>
+inline
+void
+shared_ptr<_Tp>::reset() _NOEXCEPT
+{
+    shared_ptr().swap(*this);
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    void
+>::type
+shared_ptr<_Tp>::reset(_Yp* __p)
+{
+    shared_ptr(__p).swap(*this);
+}
+
+template<class _Tp>
+template<class _Yp, class _Dp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    void
+>::type
+shared_ptr<_Tp>::reset(_Yp* __p, _Dp __d)
+{
+    shared_ptr(__p, __d).swap(*this);
+}
+
+template<class _Tp>
+template<class _Yp, class _Dp, class _Alloc>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, typename shared_ptr<_Tp>::element_type*>::value,
+    void
+>::type
+shared_ptr<_Tp>::reset(_Yp* __p, _Dp __d, _Alloc __a)
+{
+    shared_ptr(__p, __d, __a).swap(*this);
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+template<class _Tp, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_array<_Tp>::value,
+    shared_ptr<_Tp>
+>::type
+make_shared(_Args&& ...__args)
+{
+    return shared_ptr<_Tp>::make_shared(_CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template<class _Tp, class _Alloc, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_array<_Tp>::value,
+    shared_ptr<_Tp>
+>::type
+allocate_shared(const _Alloc& __a, _Args&& ...__args)
+{
+    return shared_ptr<_Tp>::allocate_shared(__a, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+make_shared()
+{
+    return shared_ptr<_Tp>::make_shared();
+}
+
+template<class _Tp, class _A0>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+make_shared(_A0& __a0)
+{
+    return shared_ptr<_Tp>::make_shared(__a0);
+}
+
+template<class _Tp, class _A0, class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+make_shared(_A0& __a0, _A1& __a1)
+{
+    return shared_ptr<_Tp>::make_shared(__a0, __a1);
+}
+
+template<class _Tp, class _A0, class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+make_shared(_A0& __a0, _A1& __a1, _A2& __a2)
+{
+    return shared_ptr<_Tp>::make_shared(__a0, __a1, __a2);
+}
+
+template<class _Tp, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+allocate_shared(const _Alloc& __a)
+{
+    return shared_ptr<_Tp>::allocate_shared(__a);
+}
+
+template<class _Tp, class _Alloc, class _A0>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+allocate_shared(const _Alloc& __a, _A0& __a0)
+{
+    return shared_ptr<_Tp>::allocate_shared(__a, __a0);
+}
+
+template<class _Tp, class _Alloc, class _A0, class _A1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+allocate_shared(const _Alloc& __a, _A0& __a0, _A1& __a1)
+{
+    return shared_ptr<_Tp>::allocate_shared(__a, __a0, __a1);
+}
+
+template<class _Tp, class _Alloc, class _A0, class _A1, class _A2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+shared_ptr<_Tp>
+allocate_shared(const _Alloc& __a, _A0& __a0, _A1& __a1, _A2& __a2)
+{
+    return shared_ptr<_Tp>::allocate_shared(__a, __a0, __a1, __a2);
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const shared_ptr<_Tp>& __x, const shared_ptr<_Up>& __y) _NOEXCEPT
+{
+    return __x.get() == __y.get();
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const shared_ptr<_Tp>& __x, const shared_ptr<_Up>& __y) _NOEXCEPT
+{
+    return !(__x == __y);
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const shared_ptr<_Tp>& __x, const shared_ptr<_Up>& __y) _NOEXCEPT
+{
+#if _LIBCUDACXX_STD_VER <= 11
+    typedef typename common_type<_Tp*, _Up*>::type _Vp;
+    return less<_Vp>()(__x.get(), __y.get());
+#else
+    return less<>()(__x.get(), __y.get());
+#endif
+
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const shared_ptr<_Tp>& __x, const shared_ptr<_Up>& __y) _NOEXCEPT
+{
+    return __y < __x;
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const shared_ptr<_Tp>& __x, const shared_ptr<_Up>& __y) _NOEXCEPT
+{
+    return !(__y < __x);
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const shared_ptr<_Tp>& __x, const shared_ptr<_Up>& __y) _NOEXCEPT
+{
+    return !(__x < __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const shared_ptr<_Tp>& __x, nullptr_t) _NOEXCEPT
+{
+    return !__x;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(nullptr_t, const shared_ptr<_Tp>& __x) _NOEXCEPT
+{
+    return !__x;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const shared_ptr<_Tp>& __x, nullptr_t) _NOEXCEPT
+{
+    return static_cast<bool>(__x);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(nullptr_t, const shared_ptr<_Tp>& __x) _NOEXCEPT
+{
+    return static_cast<bool>(__x);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const shared_ptr<_Tp>& __x, nullptr_t) _NOEXCEPT
+{
+    return less<_Tp*>()(__x.get(), nullptr);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(nullptr_t, const shared_ptr<_Tp>& __x) _NOEXCEPT
+{
+    return less<_Tp*>()(nullptr, __x.get());
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const shared_ptr<_Tp>& __x, nullptr_t) _NOEXCEPT
+{
+    return nullptr < __x;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(nullptr_t, const shared_ptr<_Tp>& __x) _NOEXCEPT
+{
+    return __x < nullptr;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const shared_ptr<_Tp>& __x, nullptr_t) _NOEXCEPT
+{
+    return !(nullptr < __x);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(nullptr_t, const shared_ptr<_Tp>& __x) _NOEXCEPT
+{
+    return !(__x < nullptr);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const shared_ptr<_Tp>& __x, nullptr_t) _NOEXCEPT
+{
+    return !(__x < nullptr);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(nullptr_t, const shared_ptr<_Tp>& __x) _NOEXCEPT
+{
+    return !(nullptr < __x);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(shared_ptr<_Tp>& __x, shared_ptr<_Tp>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_array<_Tp>::value && !is_array<_Up>::value,
+    shared_ptr<_Tp>
+>::type
+static_pointer_cast(const shared_ptr<_Up>& __r) _NOEXCEPT
+{
+    return shared_ptr<_Tp>(__r, static_cast<_Tp*>(__r.get()));
+}
+
+template<class _Tp, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_array<_Tp>::value && !is_array<_Up>::value,
+    shared_ptr<_Tp>
+>::type
+dynamic_pointer_cast(const shared_ptr<_Up>& __r) _NOEXCEPT
+{
+    _Tp* __p = dynamic_cast<_Tp*>(__r.get());
+    return __p ? shared_ptr<_Tp>(__r, __p) : shared_ptr<_Tp>();
+}
+
+template<class _Tp, class _Up>
+typename enable_if
+<
+    is_array<_Tp>::value == is_array<_Up>::value,
+    shared_ptr<_Tp>
+>::type
+const_pointer_cast(const shared_ptr<_Up>& __r) _NOEXCEPT
+{
+    typedef typename remove_extent<_Tp>::type _RTp;
+    return shared_ptr<_Tp>(__r, const_cast<_RTp*>(__r.get()));
+}
+
+#ifndef _LIBCUDACXX_NO_RTTI
+
+template<class _Dp, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Dp*
+get_deleter(const shared_ptr<_Tp>& __p) _NOEXCEPT
+{
+    return __p.template __get_deleter<_Dp>();
+}
+
+#endif  // _LIBCUDACXX_NO_RTTI
+
+template<class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS weak_ptr
+{
+public:
+    typedef _Tp element_type;
+private:
+    element_type*        __ptr_;
+    __shared_weak_count* __cntrl_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR weak_ptr() _NOEXCEPT;
+    template<class _Yp> _LIBCUDACXX_INLINE_VISIBILITY weak_ptr(shared_ptr<_Yp> const& __r,
+                   typename enable_if<is_convertible<_Yp*, _Tp*>::value, __nat*>::type = 0)
+                        _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    weak_ptr(weak_ptr const& __r) _NOEXCEPT;
+    template<class _Yp> _LIBCUDACXX_INLINE_VISIBILITY weak_ptr(weak_ptr<_Yp> const& __r,
+                   typename enable_if<is_convertible<_Yp*, _Tp*>::value, __nat*>::type = 0)
+                         _NOEXCEPT;
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    _LIBCUDACXX_INLINE_VISIBILITY
+    weak_ptr(weak_ptr&& __r) _NOEXCEPT;
+    template<class _Yp> _LIBCUDACXX_INLINE_VISIBILITY weak_ptr(weak_ptr<_Yp>&& __r,
+                   typename enable_if<is_convertible<_Yp*, _Tp*>::value, __nat*>::type = 0)
+                         _NOEXCEPT;
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+    ~weak_ptr();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    weak_ptr& operator=(weak_ptr const& __r) _NOEXCEPT;
+    template<class _Yp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            weak_ptr&
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(weak_ptr<_Yp> const& __r) _NOEXCEPT;
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    weak_ptr& operator=(weak_ptr&& __r) _NOEXCEPT;
+    template<class _Yp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            weak_ptr&
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(weak_ptr<_Yp>&& __r) _NOEXCEPT;
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+    template<class _Yp>
+        typename enable_if
+        <
+            is_convertible<_Yp*, element_type*>::value,
+            weak_ptr&
+        >::type
+        _LIBCUDACXX_INLINE_VISIBILITY
+        operator=(shared_ptr<_Yp> const& __r) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(weak_ptr& __r) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    long use_count() const _NOEXCEPT
+        {return __cntrl_ ? __cntrl_->use_count() : 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool expired() const _NOEXCEPT
+        {return __cntrl_ == 0 || __cntrl_->use_count() == 0;}
+    shared_ptr<_Tp> lock() const _NOEXCEPT;
+    template<class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool owner_before(const shared_ptr<_Up>& __r) const _NOEXCEPT
+        {return __cntrl_ < __r.__cntrl_;}
+    template<class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool owner_before(const weak_ptr<_Up>& __r) const _NOEXCEPT
+        {return __cntrl_ < __r.__cntrl_;}
+
+    template <class _Up> friend class _LIBCUDACXX_TEMPLATE_VIS weak_ptr;
+    template <class _Up> friend class _LIBCUDACXX_TEMPLATE_VIS shared_ptr;
+};
+
+template<class _Tp>
+inline
+_LIBCUDACXX_CONSTEXPR
+weak_ptr<_Tp>::weak_ptr() _NOEXCEPT
+    : __ptr_(0),
+      __cntrl_(0)
+{
+}
+
+template<class _Tp>
+inline
+weak_ptr<_Tp>::weak_ptr(weak_ptr const& __r) _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    if (__cntrl_)
+        __cntrl_->__add_weak();
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+weak_ptr<_Tp>::weak_ptr(shared_ptr<_Yp> const& __r,
+                        typename enable_if<is_convertible<_Yp*, _Tp*>::value, __nat*>::type)
+                         _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    if (__cntrl_)
+        __cntrl_->__add_weak();
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+weak_ptr<_Tp>::weak_ptr(weak_ptr<_Yp> const& __r,
+                        typename enable_if<is_convertible<_Yp*, _Tp*>::value, __nat*>::type)
+         _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    if (__cntrl_)
+        __cntrl_->__add_weak();
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+inline
+weak_ptr<_Tp>::weak_ptr(weak_ptr&& __r) _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    __r.__ptr_ = 0;
+    __r.__cntrl_ = 0;
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+weak_ptr<_Tp>::weak_ptr(weak_ptr<_Yp>&& __r,
+                        typename enable_if<is_convertible<_Yp*, _Tp*>::value, __nat*>::type)
+         _NOEXCEPT
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_)
+{
+    __r.__ptr_ = 0;
+    __r.__cntrl_ = 0;
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+weak_ptr<_Tp>::~weak_ptr()
+{
+    if (__cntrl_)
+        __cntrl_->__release_weak();
+}
+
+template<class _Tp>
+inline
+weak_ptr<_Tp>&
+weak_ptr<_Tp>::operator=(weak_ptr const& __r) _NOEXCEPT
+{
+    weak_ptr(__r).swap(*this);
+    return *this;
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, _Tp*>::value,
+    weak_ptr<_Tp>&
+>::type
+weak_ptr<_Tp>::operator=(weak_ptr<_Yp> const& __r) _NOEXCEPT
+{
+    weak_ptr(__r).swap(*this);
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+inline
+weak_ptr<_Tp>&
+weak_ptr<_Tp>::operator=(weak_ptr&& __r) _NOEXCEPT
+{
+    weak_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, _Tp*>::value,
+    weak_ptr<_Tp>&
+>::type
+weak_ptr<_Tp>::operator=(weak_ptr<_Yp>&& __r) _NOEXCEPT
+{
+    weak_ptr(_CUDA_VSTD::move(__r)).swap(*this);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template<class _Tp>
+template<class _Yp>
+inline
+typename enable_if
+<
+    is_convertible<_Yp*, _Tp*>::value,
+    weak_ptr<_Tp>&
+>::type
+weak_ptr<_Tp>::operator=(shared_ptr<_Yp> const& __r) _NOEXCEPT
+{
+    weak_ptr(__r).swap(*this);
+    return *this;
+}
+
+template<class _Tp>
+inline
+void
+weak_ptr<_Tp>::swap(weak_ptr& __r) _NOEXCEPT
+{
+    _CUDA_VSTD::swap(__ptr_, __r.__ptr_);
+    _CUDA_VSTD::swap(__cntrl_, __r.__cntrl_);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(weak_ptr<_Tp>& __x, weak_ptr<_Tp>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+template<class _Tp>
+inline
+void
+weak_ptr<_Tp>::reset() _NOEXCEPT
+{
+    weak_ptr().swap(*this);
+}
+
+template<class _Tp>
+template<class _Yp>
+shared_ptr<_Tp>::shared_ptr(const weak_ptr<_Yp>& __r,
+                            typename enable_if<is_convertible<_Yp*, element_type*>::value, __nat>::type)
+    : __ptr_(__r.__ptr_),
+      __cntrl_(__r.__cntrl_ ? __r.__cntrl_->lock() : __r.__cntrl_)
+{
+    if (__cntrl_ == 0)
+        __throw_bad_weak_ptr();
+}
+
+template<class _Tp>
+shared_ptr<_Tp>
+weak_ptr<_Tp>::lock() const _NOEXCEPT
+{
+    shared_ptr<_Tp> __r;
+    __r.__cntrl_ = __cntrl_ ? __cntrl_->lock() : __cntrl_;
+    if (__r.__cntrl_)
+        __r.__ptr_ = __ptr_;
+    return __r;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp = void> struct owner_less;
+#else
+template <class _Tp> struct owner_less;
+#endif
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS owner_less<shared_ptr<_Tp> >
+    : binary_function<shared_ptr<_Tp>, shared_ptr<_Tp>, bool>
+{
+    typedef bool result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(shared_ptr<_Tp> const& __x, shared_ptr<_Tp> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(shared_ptr<_Tp> const& __x,   weak_ptr<_Tp> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(  weak_ptr<_Tp> const& __x, shared_ptr<_Tp> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS owner_less<weak_ptr<_Tp> >
+    : binary_function<weak_ptr<_Tp>, weak_ptr<_Tp>, bool>
+{
+    typedef bool result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(  weak_ptr<_Tp> const& __x,   weak_ptr<_Tp> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(shared_ptr<_Tp> const& __x,   weak_ptr<_Tp> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(  weak_ptr<_Tp> const& __x, shared_ptr<_Tp> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS owner_less<void>
+{
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()( shared_ptr<_Tp> const& __x, shared_ptr<_Up> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()( shared_ptr<_Tp> const& __x,   weak_ptr<_Up> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(   weak_ptr<_Tp> const& __x, shared_ptr<_Up> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(   weak_ptr<_Tp> const& __x,   weak_ptr<_Up> const& __y) const _NOEXCEPT
+        {return __x.owner_before(__y);}
+    typedef void is_transparent;
+};
+#endif
+
+template<class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS enable_shared_from_this
+{
+    mutable weak_ptr<_Tp> __weak_this_;
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    enable_shared_from_this() _NOEXCEPT {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    enable_shared_from_this(enable_shared_from_this const&) _NOEXCEPT {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    enable_shared_from_this& operator=(enable_shared_from_this const&) _NOEXCEPT
+        {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~enable_shared_from_this() {}
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_ptr<_Tp> shared_from_this()
+        {return shared_ptr<_Tp>(__weak_this_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_ptr<_Tp const> shared_from_this() const
+        {return shared_ptr<const _Tp>(__weak_this_);}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    weak_ptr<_Tp> weak_from_this() _NOEXCEPT
+       { return __weak_this_; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    weak_ptr<const _Tp> weak_from_this() const _NOEXCEPT
+        { return __weak_this_; }
+#endif // _LIBCUDACXX_STD_VER > 14
+
+    template <class _Up> friend class shared_ptr;
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<shared_ptr<_Tp> >
+{
+    typedef shared_ptr<_Tp>      argument_type;
+    typedef size_t               result_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator()(const argument_type& __ptr) const _NOEXCEPT
+    {
+        return hash<_Tp*>()(__ptr.get());
+    }
+};
+
+template<class _CharT, class _Traits, class _Yp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, shared_ptr<_Yp> const& __p);
+
+
+#if !defined(_LIBCUDACXX_HAS_NO_ATOMIC_HEADER)
+
+class _LIBCUDACXX_TYPE_VIS __sp_mut
+{
+    void* __lx;
+public:
+    void lock() _NOEXCEPT;
+    void unlock() _NOEXCEPT;
+
+private:
+    _LIBCUDACXX_CONSTEXPR __sp_mut(void*) _NOEXCEPT;
+    __sp_mut(const __sp_mut&);
+    __sp_mut& operator=(const __sp_mut&);
+
+    friend _LIBCUDACXX_FUNC_VIS __sp_mut& __get_sp_mut(const void*);
+};
+
+_LIBCUDACXX_FUNC_VIS _LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+__sp_mut& __get_sp_mut(const void*);
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+atomic_is_lock_free(const shared_ptr<_Tp>*)
+{
+    return false;
+}
+
+template <class _Tp>
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+shared_ptr<_Tp>
+atomic_load(const shared_ptr<_Tp>* __p)
+{
+    __sp_mut& __m = __get_sp_mut(__p);
+    __m.lock();
+    shared_ptr<_Tp> __q = *__p;
+    __m.unlock();
+    return __q;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+shared_ptr<_Tp>
+atomic_load_explicit(const shared_ptr<_Tp>* __p, memory_order)
+{
+    return atomic_load(__p);
+}
+
+template <class _Tp>
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+void
+atomic_store(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r)
+{
+    __sp_mut& __m = __get_sp_mut(__p);
+    __m.lock();
+    __p->swap(__r);
+    __m.unlock();
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+void
+atomic_store_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r, memory_order)
+{
+    atomic_store(__p, __r);
+}
+
+template <class _Tp>
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+shared_ptr<_Tp>
+atomic_exchange(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r)
+{
+    __sp_mut& __m = __get_sp_mut(__p);
+    __m.lock();
+    __p->swap(__r);
+    __m.unlock();
+    return __r;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+shared_ptr<_Tp>
+atomic_exchange_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r, memory_order)
+{
+    return atomic_exchange(__p, __r);
+}
+
+template <class _Tp>
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+bool
+atomic_compare_exchange_strong(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, shared_ptr<_Tp> __w)
+{
+    shared_ptr<_Tp> __temp;
+    __sp_mut& __m = __get_sp_mut(__p);
+    __m.lock();
+    if (__p->__owner_equivalent(*__v))
+    {
+        _CUDA_VSTD::swap(__temp, *__p);
+        *__p = __w;
+        __m.unlock();
+        return true;
+    }
+    _CUDA_VSTD::swap(__temp, *__v);
+    *__v = *__p;
+    __m.unlock();
+    return false;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+bool
+atomic_compare_exchange_weak(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, shared_ptr<_Tp> __w)
+{
+    return atomic_compare_exchange_strong(__p, __v, __w);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+bool
+atomic_compare_exchange_strong_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v,
+                                        shared_ptr<_Tp> __w, memory_order, memory_order)
+{
+    return atomic_compare_exchange_strong(__p, __v, __w);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_ATOMIC_SHARED_PTR
+bool
+atomic_compare_exchange_weak_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v,
+                                      shared_ptr<_Tp> __w, memory_order, memory_order)
+{
+    return atomic_compare_exchange_weak(__p, __v, __w);
+}
+
+#endif  // !defined(_LIBCUDACXX_HAS_NO_ATOMIC_HEADER)
+
+//enum class
+#if defined(_LIBCUDACXX_ABI_POINTER_SAFETY_ENUM_TYPE)
+# ifndef _LIBCUDACXX_CXX03_LANG
+enum class pointer_safety : unsigned char {
+  relaxed,
+  preferred,
+  strict
+};
+# endif
+#else
+struct _LIBCUDACXX_TYPE_VIS pointer_safety
+{
+    enum __lx
+    {
+        relaxed,
+        preferred,
+        strict
+    };
+
+    __lx __v_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer_safety() : __v_() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer_safety(__lx __v) : __v_(__v) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator int() const {return __v_;}
+};
+#endif
+
+#if !defined(_LIBCUDACXX_ABI_POINTER_SAFETY_ENUM_TYPE) && \
+    defined(_LIBCUDACXX_BUILDING_LIBRARY)
+_LIBCUDACXX_FUNC_VIS pointer_safety get_pointer_safety() _NOEXCEPT;
+#else
+// This function is only offered in C++03 under ABI v1.
+# if !defined(_LIBCUDACXX_ABI_POINTER_SAFETY_ENUM_TYPE) || !defined(_LIBCUDACXX_CXX03_LANG)
+inline _LIBCUDACXX_INLINE_VISIBILITY
+pointer_safety get_pointer_safety() _NOEXCEPT {
+  return pointer_safety::relaxed;
+}
+# endif
+#endif
+
+
+_LIBCUDACXX_FUNC_VIS void declare_reachable(void* __p);
+_LIBCUDACXX_FUNC_VIS void declare_no_pointers(char* __p, size_t __n);
+_LIBCUDACXX_FUNC_VIS void undeclare_no_pointers(char* __p, size_t __n);
+_LIBCUDACXX_FUNC_VIS void* __undeclare_reachable(void* __p);
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+undeclare_reachable(_Tp* __p)
+{
+    return static_cast<_Tp*>(__undeclare_reachable(__p));
+}
+
+_LIBCUDACXX_FUNC_VIS void* align(size_t __align, size_t __sz, void*& __ptr, size_t& __space);
+
+// --- Helper for container swap --
+template <typename _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __swap_allocator(_Alloc & __a1, _Alloc & __a2)
+#if _LIBCUDACXX_STD_VER >= 14
+    _NOEXCEPT
+#else
+    _NOEXCEPT_(__is_nothrow_swappable<_Alloc>::value)
+#endif
+{
+    __swap_allocator(__a1, __a2,
+      integral_constant<bool, _CUDA_VSTD::allocator_traits<_Alloc>::propagate_on_container_swap::value>());
+}
+
+template <typename _Alloc>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __swap_allocator(_Alloc & __a1, _Alloc & __a2, true_type)
+#if _LIBCUDACXX_STD_VER >= 14
+    _NOEXCEPT
+#else
+    _NOEXCEPT_(__is_nothrow_swappable<_Alloc>::value)
+#endif
+{
+    using _CUDA_VSTD::swap;
+    swap(__a1, __a2);
+}
+
+template <typename _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __swap_allocator(_Alloc &, _Alloc &, false_type) _NOEXCEPT {}
+
+template <typename _Alloc, typename _Traits=allocator_traits<_Alloc> >
+struct __noexcept_move_assign_container : public integral_constant<bool,
+    _Traits::propagate_on_container_move_assignment::value
+#if _LIBCUDACXX_STD_VER > 14
+        || _Traits::is_always_equal::value
+#else
+        && is_nothrow_move_assignable<_Alloc>::value
+#endif
+    > {};
+
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+template <class _Tp, class _Alloc>
+struct __temp_value {
+    typedef allocator_traits<_Alloc> _Traits;
+
+    typename aligned_storage<sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp)>::type __v;
+    _Alloc &__a;
+
+    _Tp *__addr() { return reinterpret_cast<_Tp *>(addressof(__v)); }
+    _Tp &   get() { return *__addr(); }
+
+    template<class... _Args>
+    _LIBCUDACXX_NO_CFI
+    __temp_value(_Alloc &__alloc, _Args&& ... __args) : __a(__alloc) {
+      _Traits::construct(__a, reinterpret_cast<_Tp*>(addressof(__v)),
+                         _CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    ~__temp_value() { _Traits::destroy(__a, __addr()); }
+    };
+#endif
+
+template<typename _Alloc, typename = void, typename = void>
+struct __is_allocator : false_type {};
+
+template<typename _Alloc>
+struct __is_allocator<_Alloc,
+       typename __void_t<typename _Alloc::value_type>::type,
+       typename __void_t<decltype(_CUDA_VSTD::declval<_Alloc&>().allocate(size_t(0)))>::type
+     >
+   : true_type {};
+
+// __builtin_new_allocator -- A non-templated helper for allocating and
+// deallocating memory using __builtin_operator_new and
+// __builtin_operator_delete. It should be used in preference to
+// `std::allocator<T>` to avoid additional instantiations.
+struct __builtin_new_allocator {
+  struct __builtin_new_deleter {
+    typedef void* pointer_type;
+
+    _LIBCUDACXX_CONSTEXPR explicit __builtin_new_deleter(size_t __size, size_t __align)
+        : __size_(__size), __align_(__align) {}
+
+    void operator()(void* p) const _NOEXCEPT {
+        std::__libcpp_deallocate(p, __size_, __align_);
+    }
+
+   private:
+    size_t __size_;
+    size_t __align_;
+  };
+
+  typedef unique_ptr<void, __builtin_new_deleter> __holder_t;
+
+  static __holder_t __allocate_bytes(size_t __s, size_t __align) {
+      return __holder_t(std::__libcpp_allocate(__s, __align),
+                     __builtin_new_deleter(__s, __align));
+  }
+
+  static void __deallocate_bytes(void* __p, size_t __s,
+                                 size_t __align) _NOEXCEPT {
+      std::__libcpp_deallocate(__p, __s, __align);
+  }
+
+  template <class _Tp>
+  _LIBCUDACXX_NODEBUG _LIBCUDACXX_ALWAYS_INLINE
+  static __holder_t __allocate_type(size_t __n) {
+      return __allocate_bytes(__n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp));
+  }
+
+  template <class _Tp>
+  _LIBCUDACXX_NODEBUG _LIBCUDACXX_ALWAYS_INLINE
+  static void __deallocate_type(void* __p, size_t __n) _NOEXCEPT {
+      __deallocate_bytes(__p, __n * sizeof(_Tp), _LIBCUDACXX_ALIGNOF(_Tp));
+  }
+};
+
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_HAS_PARALLEL_ALGORITHMS) && _LIBCUDACXX_STD_VER >= 17
+#   include <__pstl_memory>
+#endif
+
+#endif  // _LIBCUDACXX_MEMORY
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/module.modulemap b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/module.modulemap
new file mode 100644
index 000000000000..31d39ddf7c8a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/module.modulemap
@@ -0,0 +1,612 @@
+// define the module for __config outside of the top level 'std' module
+// since __config may be included from C headers which may create an
+// include cycle.
+module std_config [system] [extern_c] {
+    header "__config"
+}
+
+module std [system] {
+  export std_config
+  // FIXME: The standard does not require that each of these submodules
+  // re-exports its imported modules. We should provide an alternative form of
+  // export that issues a warning if a name from the submodule is used, and
+  // use that to provide a 'strict mode' for libc++.
+
+  // Deprecated C-compatibility headers. These can all be included from within
+  // an 'extern "C"' context.
+  module depr [extern_c] {
+    // <assert.h> provided by C library.
+    module ctype_h {
+      header "ctype.h"
+      export *
+    }
+    module errno_h {
+      header "errno.h"
+      export *
+    }
+    module fenv_h {
+      header "fenv.h"
+      export *
+    }
+    // <float.h> provided by compiler or C library.
+    module inttypes_h {
+      header "inttypes.h"
+      export stdint_h
+      export *
+    }
+    // <iso646.h> provided by compiler.
+    // <limits.h> provided by compiler or C library.
+    module locale_h {
+      header "locale.h"
+      export *
+    }
+    module math_h {
+      header "math.h"
+      export *
+    }
+    module setjmp_h {
+      header "setjmp.h"
+      export *
+    }
+    // FIXME: <stdalign.h> is missing.
+    // <signal.h> provided by C library.
+    // <stdarg.h> provided by compiler.
+    // <stdbool.h> provided by compiler.
+    module stddef_h {
+      // <stddef.h>'s __need_* macros require textual inclusion.
+      textual header "stddef.h"
+    }
+    module stdint_h {
+      header "stdint.h"
+      export *
+      // FIXME: This module only exists on OS X and for some reason the
+      // wildcard above doesn't export it.
+      export Darwin.C.stdint
+    }
+    module stdio_h {
+      // <stdio.h>'s __need_* macros require textual inclusion.
+      textual header "stdio.h"
+      export *
+      export Darwin.C.stdio
+    }
+    module stdlib_h {
+      // <stdlib.h>'s __need_* macros require textual inclusion.
+      textual header "stdlib.h"
+      export *
+    }
+    module string_h {
+      header "string.h"
+      export *
+    }
+    // FIXME: <uchar.h> is missing.
+    // <time.h> provided by C library.
+    module wchar_h {
+      // <wchar.h>'s __need_* macros require textual inclusion.
+      textual header "wchar.h"
+      export *
+    }
+    module wctype_h {
+      header "wctype.h"
+      export *
+    }
+  }
+
+  // <complex.h> and <tgmath.h> are not C headers in any real sense, do not
+  // allow their use in extern "C" contexts.
+  module complex_h {
+    header "complex.h"
+    export ccomplex
+    export *
+  }
+  module tgmath_h {
+    header "tgmath.h"
+    export ccomplex
+    export cmath
+    export *
+  }
+
+  // C compatibility headers.
+  module compat {
+    module cassert {
+      // <cassert>'s use of NDEBUG requires textual inclusion.
+      textual header "cassert"
+    }
+    module ccomplex {
+      header "ccomplex"
+      export complex
+      export *
+    }
+    module cctype {
+      header "cctype"
+      export *
+    }
+    module cerrno {
+      header "cerrno"
+      export *
+    }
+    module cfenv {
+      header "cfenv"
+      export *
+    }
+    module cfloat {
+      header "cfloat"
+      export *
+    }
+    module cinttypes {
+      header "cinttypes"
+      export cstdint
+      export *
+    }
+    module ciso646 {
+      header "ciso646"
+      export *
+    }
+    module climits {
+      header "climits"
+      export *
+    }
+    module clocale {
+      header "clocale"
+      export *
+    }
+    module cmath {
+      header "cmath"
+      export *
+    }
+    module csetjmp {
+      header "csetjmp"
+      export *
+    }
+    module csignal {
+      header "csignal"
+      export *
+    }
+    // FIXME: <cstdalign> is missing.
+    module cstdarg {
+      header "cstdarg"
+      export *
+    }
+    module cstdbool {
+      header "cstdbool"
+      export *
+    }
+    module cstddef {
+      header "cstddef"
+      export *
+    }
+    module cstdint {
+      header "cstdint"
+      export depr.stdint_h
+      export *
+    }
+    module cstdio {
+      header "cstdio"
+      export *
+    }
+    module cstdlib {
+      header "cstdlib"
+      export *
+    }
+    module cstring {
+      header "cstring"
+      export *
+    }
+    module ctgmath {
+      header "ctgmath"
+      export ccomplex
+      export cmath
+      export *
+    }
+    module ctime {
+      header "ctime"
+      export *
+    }
+    // FIXME: <cuchar> is missing.
+    module cwchar {
+      header "cwchar"
+      export depr.stdio_h
+      export *
+    }
+    module cwctype {
+      header "cwctype"
+      export *
+    }
+  }
+
+  module algorithm {
+    header "algorithm"
+    export initializer_list
+    export *
+  }
+  module any {
+    header "any"
+    export *
+  }
+  module array {
+    header "array"
+    export initializer_list
+    export *
+  }
+  module atomic {
+    header "atomic"
+    export *
+  }
+  module bit {
+    header "bit"
+    export *
+  }
+  module bitset {
+    header "bitset"
+    export string
+    export iosfwd
+    export *
+  }
+  // No submodule for cassert. It fundamentally needs repeated, textual inclusion.
+  module charconv {
+    header "charconv"
+    export *
+  }
+  module chrono {
+    header "chrono"
+    export *
+  }
+  module codecvt {
+    header "codecvt"
+    export *
+  }
+  module compare {
+    header "compare"
+    export *
+  }
+  module complex {
+    header "complex"
+    export *
+  }
+  module condition_variable {
+    header "condition_variable"
+    export *
+  }
+  module deque {
+    header "deque"
+    export initializer_list
+    export *
+  }
+  module exception {
+    header "exception"
+    export *
+  }
+  module execution {
+    header "execution"
+    export *
+  }
+  module filesystem {
+    header "filesystem"
+    export *
+  }
+  module forward_list {
+    header "forward_list"
+    export initializer_list
+    export *
+  }
+  module fstream {
+    header "fstream"
+    export *
+  }
+  module functional {
+    header "functional"
+    export *
+  }
+  module future {
+    header "future"
+    export *
+  }
+  module initializer_list {
+    header "initializer_list"
+    export *
+  }
+  module iomanip {
+    header "iomanip"
+    export *
+  }
+  module ios {
+    header "ios"
+    export iosfwd
+    export *
+  }
+  module iosfwd {
+    header "iosfwd"
+    export *
+  }
+  module iostream {
+    header "iostream"
+    export ios
+    export streambuf
+    export istream
+    export ostream
+    export *
+  }
+  module istream {
+    header "istream"
+    // FIXME: should re-export ios, streambuf?
+    export *
+  }
+  module iterator {
+    header "iterator"
+    export *
+  }
+  module limits {
+    header "limits"
+    export *
+  }
+  module list {
+    header "list"
+    export initializer_list
+    export *
+  }
+  module locale {
+    header "locale"
+    export *
+  }
+  module map {
+    header "map"
+    export initializer_list
+    export *
+  }
+  module memory {
+    header "memory"
+    export *
+  }
+  module mutex {
+    header "mutex"
+    export *
+  }
+  module new {
+    header "new"
+    export *
+  }
+  module numeric {
+    header "numeric"
+    export *
+  }
+  module optional {
+    header "optional"
+    export *
+  }
+  module ostream {
+    header "ostream"
+    // FIXME: should re-export ios, streambuf?
+    export *
+  }
+  module queue {
+    header "queue"
+    export initializer_list
+    export *
+  }
+  module random {
+    header "random"
+    export initializer_list
+    export *
+  }
+  module ratio {
+    header "ratio"
+    export *
+  }
+  module regex {
+    header "regex"
+    export initializer_list
+    export *
+  }
+  module scoped_allocator {
+    header "scoped_allocator"
+    export *
+  }
+  module set {
+    header "set"
+    export initializer_list
+    export *
+  }
+  module sstream {
+    header "sstream"
+    // FIXME: should re-export istream, ostream, ios, streambuf, string?
+    export *
+  }
+  module stack {
+    header "stack"
+    export initializer_list
+    export *
+  }
+  module stdexcept {
+    header "stdexcept"
+    export *
+  }
+  module streambuf {
+    header "streambuf"
+    export *
+  }
+  module string {
+    header "string"
+    export initializer_list
+    export string_view
+    export __string
+    export *
+  }
+  module string_view {
+    header "string_view"
+    export initializer_list
+    export __string
+    export *
+  }
+  module strstream {
+    header "strstream"
+    export *
+  }
+  module system_error {
+    header "system_error"
+    export *
+  }
+  module thread {
+    header "thread"
+    export *
+  }
+  module tuple {
+    header "tuple"
+    export *
+  }
+  module type_traits {
+    header "type_traits"
+    export *
+  }
+  module typeindex {
+    header "typeindex"
+    export *
+  }
+  module typeinfo {
+    header "typeinfo"
+    export *
+  }
+  module unordered_map {
+    header "unordered_map"
+    export initializer_list
+    export *
+  }
+  module unordered_set {
+    header "unordered_set"
+    export initializer_list
+    export *
+  }
+  module utility {
+    header "utility"
+    export initializer_list
+    export *
+  }
+  module valarray {
+    header "valarray"
+    export initializer_list
+    export *
+  }
+  module variant {
+    header "variant"
+    export *
+  }
+  module vector {
+    header "vector"
+    export initializer_list
+    export *
+  }
+  module version {
+    header "version"
+    export *
+  }
+
+  // FIXME: These should be private.
+  module __bit_reference { header "__bit_reference" export * }
+  module __debug { header "__debug" export * }
+  module __errc { header "__errc" export * }
+  module __functional_base { header "__functional_base" export * }
+  module __hash_table { header "__hash_table" export * }
+  module __locale { header "__locale" export * }
+  module __mutex_base { header "__mutex_base" export * }
+  module __split_buffer { header "__split_buffer" export * }
+  module __sso_allocator { header "__sso_allocator" export * }
+  module __std_stream { header "__std_stream" export * }
+  module __string { header "__string" export * }
+  module __tree { header "__tree" export * }
+  module __tuple { header "__tuple" export * }
+  module __undef_macros { header "__undef_macros" export * }
+  module __node_handle { header "__node_handle" export * }
+
+  module experimental {
+    requires cplusplus11
+
+    module algorithm {
+      header "experimental/algorithm"
+      export *
+    }
+     module coroutine {
+      requires coroutines
+      header "experimental/coroutine"
+      export *
+    }
+    module deque {
+      header "experimental/deque"
+      export *
+    }
+    module filesystem {
+      header "experimental/filesystem"
+      export *
+    }
+    module forward_list {
+      header "experimental/forward_list"
+      export *
+    }
+    module functional {
+      header "experimental/functional"
+      export *
+    }
+    module iterator {
+      header "experimental/iterator"
+      export *
+    }
+    module list {
+      header "experimental/list"
+      export *
+    }
+    module map {
+      header "experimental/map"
+      export *
+    }
+    module memory_resource {
+      header "experimental/memory_resource"
+      export *
+    }
+    module propagate_const {
+      header "experimental/propagate_const"
+      export *
+    }
+    module regex {
+      header "experimental/regex"
+      export *
+    }
+    module simd {
+      header "experimental/simd"
+      export *
+    }
+    module set {
+      header "experimental/set"
+      export *
+    }
+    module span {
+      header "span"
+      export *
+    }
+    module string {
+      header "experimental/string"
+      export *
+    }
+    module type_traits {
+      header "experimental/type_traits"
+      export *
+    }
+    module unordered_map {
+      header "experimental/unordered_map"
+      export *
+    }
+    module unordered_set {
+      header "experimental/unordered_set"
+      export *
+    }
+    module utility {
+      header "experimental/utility"
+      export *
+    }
+    module vector {
+      header "experimental/vector"
+      export *
+    }
+    // FIXME these should be private
+    module __memory {
+      header "experimental/__memory"
+      export *
+    }
+  } // end experimental
+}
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/mutex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/mutex
new file mode 100644
index 000000000000..7a454e5609e7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/mutex
@@ -0,0 +1,711 @@
+// -*- C++ -*-
+//===--------------------------- mutex ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_MUTEX
+#define _LIBCUDACXX_MUTEX
+
+/*
+    mutex synopsis
+
+namespace std
+{
+
+class mutex
+{
+public:
+     constexpr mutex() noexcept;
+     ~mutex();
+
+    mutex(const mutex&) = delete;
+    mutex& operator=(const mutex&) = delete;
+
+    void lock();
+    bool try_lock();
+    void unlock();
+
+    typedef pthread_mutex_t* native_handle_type;
+    native_handle_type native_handle();
+};
+
+class recursive_mutex
+{
+public:
+     recursive_mutex();
+     ~recursive_mutex();
+
+    recursive_mutex(const recursive_mutex&) = delete;
+    recursive_mutex& operator=(const recursive_mutex&) = delete;
+
+    void lock();
+    bool try_lock() noexcept;
+    void unlock();
+
+    typedef pthread_mutex_t* native_handle_type;
+    native_handle_type native_handle();
+};
+
+class timed_mutex
+{
+public:
+     timed_mutex();
+     ~timed_mutex();
+
+    timed_mutex(const timed_mutex&) = delete;
+    timed_mutex& operator=(const timed_mutex&) = delete;
+
+    void lock();
+    bool try_lock();
+    template <class Rep, class Period>
+        bool try_lock_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool try_lock_until(const chrono::time_point<Clock, Duration>& abs_time);
+    void unlock();
+};
+
+class recursive_timed_mutex
+{
+public:
+     recursive_timed_mutex();
+     ~recursive_timed_mutex();
+
+    recursive_timed_mutex(const recursive_timed_mutex&) = delete;
+    recursive_timed_mutex& operator=(const recursive_timed_mutex&) = delete;
+
+    void lock();
+    bool try_lock() noexcept;
+    template <class Rep, class Period>
+        bool try_lock_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool try_lock_until(const chrono::time_point<Clock, Duration>& abs_time);
+    void unlock();
+};
+
+struct defer_lock_t { explicit defer_lock_t() = default; };
+struct try_to_lock_t { explicit try_to_lock_t() = default; };
+struct adopt_lock_t { explicit adopt_lock_t() = default; };
+
+inline constexpr defer_lock_t  defer_lock{};
+inline constexpr try_to_lock_t try_to_lock{};
+inline constexpr adopt_lock_t  adopt_lock{};
+
+template <class Mutex>
+class lock_guard
+{
+public:
+    typedef Mutex mutex_type;
+
+    explicit lock_guard(mutex_type& m);
+    lock_guard(mutex_type& m, adopt_lock_t);
+    ~lock_guard();
+
+    lock_guard(lock_guard const&) = delete;
+    lock_guard& operator=(lock_guard const&) = delete;
+};
+
+template <class... MutexTypes>
+class scoped_lock // C++17
+{
+public:
+    using mutex_type = Mutex;  // If MutexTypes... consists of the single type Mutex
+
+    explicit scoped_lock(MutexTypes&... m);
+    scoped_lock(adopt_lock_t, MutexTypes&... m);
+    ~scoped_lock();
+    scoped_lock(scoped_lock const&) = delete;
+    scoped_lock& operator=(scoped_lock const&) = delete;
+private:
+    tuple<MutexTypes&...> pm; // exposition only
+};
+
+template <class Mutex>
+class unique_lock
+{
+public:
+    typedef Mutex mutex_type;
+    unique_lock() noexcept;
+    explicit unique_lock(mutex_type& m);
+    unique_lock(mutex_type& m, defer_lock_t) noexcept;
+    unique_lock(mutex_type& m, try_to_lock_t);
+    unique_lock(mutex_type& m, adopt_lock_t);
+    template <class Clock, class Duration>
+        unique_lock(mutex_type& m, const chrono::time_point<Clock, Duration>& abs_time);
+    template <class Rep, class Period>
+        unique_lock(mutex_type& m, const chrono::duration<Rep, Period>& rel_time);
+    ~unique_lock();
+
+    unique_lock(unique_lock const&) = delete;
+    unique_lock& operator=(unique_lock const&) = delete;
+
+    unique_lock(unique_lock&& u) noexcept;
+    unique_lock& operator=(unique_lock&& u) noexcept;
+
+    void lock();
+    bool try_lock();
+
+    template <class Rep, class Period>
+        bool try_lock_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool try_lock_until(const chrono::time_point<Clock, Duration>& abs_time);
+
+    void unlock();
+
+    void swap(unique_lock& u) noexcept;
+    mutex_type* release() noexcept;
+
+    bool owns_lock() const noexcept;
+    explicit operator bool () const noexcept;
+    mutex_type* mutex() const noexcept;
+};
+
+template <class Mutex>
+  void swap(unique_lock<Mutex>& x, unique_lock<Mutex>& y) noexcept;
+
+template <class L1, class L2, class... L3>
+  int try_lock(L1&, L2&, L3&...);
+template <class L1, class L2, class... L3>
+  void lock(L1&, L2&, L3&...);
+
+struct once_flag
+{
+    constexpr once_flag() noexcept;
+
+    once_flag(const once_flag&) = delete;
+    once_flag& operator=(const once_flag&) = delete;
+};
+
+template<class Callable, class ...Args>
+  void call_once(once_flag& flag, Callable&& func, Args&&... args);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__mutex_base>
+#include <cstdint>
+#include <functional>
+#include <memory>
+#ifndef _LIBCUDACXX_CXX03_LANG
+#include <tuple>
+#endif
+#include <version>
+#include <__threading_support>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#ifndef _LIBCUDACXX_HAS_NO_THREADS
+
+class _LIBCUDACXX_TYPE_VIS recursive_mutex
+{
+    __libcpp_recursive_mutex_t __m_;
+
+public:
+     recursive_mutex();
+     ~recursive_mutex();
+
+private:
+    recursive_mutex(const recursive_mutex&); // = delete;
+    recursive_mutex& operator=(const recursive_mutex&); // = delete;
+
+public:
+    void lock();
+    bool try_lock() _NOEXCEPT;
+    void unlock()  _NOEXCEPT;
+
+    typedef __libcpp_recursive_mutex_t* native_handle_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    native_handle_type native_handle() {return &__m_;}
+};
+
+class _LIBCUDACXX_TYPE_VIS timed_mutex
+{
+    mutex              __m_;
+    condition_variable __cv_;
+    bool               __locked_;
+public:
+     timed_mutex();
+     ~timed_mutex();
+
+private:
+    timed_mutex(const timed_mutex&); // = delete;
+    timed_mutex& operator=(const timed_mutex&); // = delete;
+
+public:
+    void lock();
+    bool try_lock() _NOEXCEPT;
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool try_lock_for(const chrono::duration<_Rep, _Period>& __d)
+            {return try_lock_until(chrono::steady_clock::now() + __d);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        bool try_lock_until(const chrono::time_point<_Clock, _Duration>& __t);
+    void unlock() _NOEXCEPT;
+};
+
+template <class _Clock, class _Duration>
+bool
+timed_mutex::try_lock_until(const chrono::time_point<_Clock, _Duration>& __t)
+{
+    using namespace chrono;
+    unique_lock<mutex> __lk(__m_);
+    bool no_timeout = _Clock::now() < __t;
+    while (no_timeout && __locked_)
+        no_timeout = __cv_.wait_until(__lk, __t) == cv_status::no_timeout;
+    if (!__locked_)
+    {
+        __locked_ = true;
+        return true;
+    }
+    return false;
+}
+
+class _LIBCUDACXX_TYPE_VIS recursive_timed_mutex
+{
+    mutex              __m_;
+    condition_variable __cv_;
+    size_t             __count_;
+    __thread_id        __id_;
+public:
+     recursive_timed_mutex();
+     ~recursive_timed_mutex();
+
+private:
+    recursive_timed_mutex(const recursive_timed_mutex&); // = delete;
+    recursive_timed_mutex& operator=(const recursive_timed_mutex&); // = delete;
+
+public:
+    void lock();
+    bool try_lock() _NOEXCEPT;
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool try_lock_for(const chrono::duration<_Rep, _Period>& __d)
+            {return try_lock_until(chrono::steady_clock::now() + __d);}
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        bool try_lock_until(const chrono::time_point<_Clock, _Duration>& __t);
+    void unlock() _NOEXCEPT;
+};
+
+template <class _Clock, class _Duration>
+bool
+recursive_timed_mutex::try_lock_until(const chrono::time_point<_Clock, _Duration>& __t)
+{
+    using namespace chrono;
+    __thread_id __id = this_thread::get_id();
+    unique_lock<mutex> lk(__m_);
+    if (__id == __id_)
+    {
+        if (__count_ == numeric_limits<size_t>::max())
+            return false;
+        ++__count_;
+        return true;
+    }
+    bool no_timeout = _Clock::now() < __t;
+    while (no_timeout && __count_ != 0)
+        no_timeout = __cv_.wait_until(lk, __t) == cv_status::no_timeout;
+    if (__count_ == 0)
+    {
+        __count_ = 1;
+        __id_ = __id;
+        return true;
+    }
+    return false;
+}
+
+template <class _L0, class _L1>
+int
+try_lock(_L0& __l0, _L1& __l1)
+{
+    unique_lock<_L0> __u0(__l0, try_to_lock);
+    if (__u0.owns_lock())
+    {
+        if (__l1.try_lock())
+        {
+            __u0.release();
+            return -1;
+        }
+        else
+            return 1;
+    }
+    return 0;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _L0, class _L1, class _L2, class... _L3>
+int
+try_lock(_L0& __l0, _L1& __l1, _L2& __l2, _L3&... __l3)
+{
+    int __r = 0;
+    unique_lock<_L0> __u0(__l0, try_to_lock);
+    if (__u0.owns_lock())
+    {
+        __r = try_lock(__l1, __l2, __l3...);
+        if (__r == -1)
+            __u0.release();
+        else
+            ++__r;
+    }
+    return __r;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _L0, class _L1>
+void
+lock(_L0& __l0, _L1& __l1)
+{
+    while (true)
+    {
+        {
+            unique_lock<_L0> __u0(__l0);
+            if (__l1.try_lock())
+            {
+                __u0.release();
+                break;
+            }
+        }
+        __libcpp_thread_yield();
+        {
+            unique_lock<_L1> __u1(__l1);
+            if (__l0.try_lock())
+            {
+                __u1.release();
+                break;
+            }
+        }
+        __libcpp_thread_yield();
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _L0, class _L1, class _L2, class ..._L3>
+void
+__lock_first(int __i, _L0& __l0, _L1& __l1, _L2& __l2, _L3& ...__l3)
+{
+    while (true)
+    {
+        switch (__i)
+        {
+        case 0:
+            {
+                unique_lock<_L0> __u0(__l0);
+                __i = try_lock(__l1, __l2, __l3...);
+                if (__i == -1)
+                {
+                    __u0.release();
+                    return;
+                }
+            }
+            ++__i;
+            __libcpp_thread_yield();
+            break;
+        case 1:
+            {
+                unique_lock<_L1> __u1(__l1);
+                __i = try_lock(__l2, __l3..., __l0);
+                if (__i == -1)
+                {
+                    __u1.release();
+                    return;
+                }
+            }
+            if (__i == sizeof...(_L3) + 1)
+                __i = 0;
+            else
+                __i += 2;
+            __libcpp_thread_yield();
+            break;
+        default:
+            __lock_first(__i - 2, __l2, __l3..., __l0, __l1);
+            return;
+        }
+    }
+}
+
+template <class _L0, class _L1, class _L2, class ..._L3>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+lock(_L0& __l0, _L1& __l1, _L2& __l2, _L3& ...__l3)
+{
+    __lock_first(0, __l0, __l1, __l2, __l3...);
+}
+
+template <class _L0>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __unlock(_L0& __l0) {
+    __l0.unlock();
+}
+
+template <class _L0, class _L1>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __unlock(_L0& __l0, _L1& __l1) {
+    __l0.unlock();
+    __l1.unlock();
+}
+
+template <class _L0, class _L1, class _L2, class ..._L3>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void __unlock(_L0& __l0, _L1& __l1, _L2& __l2, _L3&... __l3) {
+    __l0.unlock();
+    __l1.unlock();
+    _CUDA_VSTD::__unlock(__l2, __l3...);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class ..._Mutexes>
+class _LIBCUDACXX_TEMPLATE_VIS scoped_lock;
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS scoped_lock<> {
+public:
+    explicit scoped_lock() {}
+    ~scoped_lock() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit scoped_lock(adopt_lock_t) {}
+
+    scoped_lock(scoped_lock const&) = delete;
+    scoped_lock& operator=(scoped_lock const&) = delete;
+};
+
+template <class _Mutex>
+class _LIBCUDACXX_TEMPLATE_VIS _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(scoped_lockable) scoped_lock<_Mutex> {
+public:
+    typedef _Mutex  mutex_type;
+private:
+    mutex_type& __m_;
+public:
+    explicit scoped_lock(mutex_type & __m) _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(acquire_capability(__m))
+        : __m_(__m) {__m_.lock();}
+
+    ~scoped_lock() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(release_capability()) {__m_.unlock();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit scoped_lock(adopt_lock_t, mutex_type& __m) _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(requires_capability(__m))
+        : __m_(__m) {}
+
+    scoped_lock(scoped_lock const&) = delete;
+    scoped_lock& operator=(scoped_lock const&) = delete;
+};
+
+template <class ..._MArgs>
+class _LIBCUDACXX_TEMPLATE_VIS scoped_lock
+{
+    static_assert(sizeof...(_MArgs) > 1, "At least 2 lock types required");
+    typedef tuple<_MArgs&...> _MutexTuple;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit scoped_lock(_MArgs&... __margs)
+      : __t_(__margs...)
+    {
+        _CUDA_VSTD::lock(__margs...);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    scoped_lock(adopt_lock_t, _MArgs&... __margs)
+        : __t_(__margs...)
+    {
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~scoped_lock() {
+        typedef typename __make_tuple_indices<sizeof...(_MArgs)>::type _Indices;
+        __unlock_unpack(_Indices{}, __t_);
+    }
+
+    scoped_lock(scoped_lock const&) = delete;
+    scoped_lock& operator=(scoped_lock const&) = delete;
+
+private:
+    template <size_t ..._Indx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __unlock_unpack(__tuple_indices<_Indx...>, _MutexTuple& __mt) {
+        _CUDA_VSTD::__unlock(_CUDA_VSTD::get<_Indx>(__mt)...);
+    }
+
+    _MutexTuple __t_;
+};
+
+#endif // _LIBCUDACXX_STD_VER > 14
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+struct _LIBCUDACXX_TEMPLATE_VIS once_flag;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Callable, class... _Args>
+_LIBCUDACXX_INLINE_VISIBILITY
+void call_once(once_flag&, _Callable&&, _Args&&...);
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template<class _Callable>
+_LIBCUDACXX_INLINE_VISIBILITY
+void call_once(once_flag&, _Callable&);
+
+template<class _Callable>
+_LIBCUDACXX_INLINE_VISIBILITY
+void call_once(once_flag&, const _Callable&);
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+struct _LIBCUDACXX_TEMPLATE_VIS once_flag
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR
+        once_flag() _NOEXCEPT : __state_(0) {}
+
+#if defined(_LIBCUDACXX_ABI_MICROSOFT)
+   typedef uintptr_t _State_type;
+#else
+   typedef unsigned long _State_type;
+#endif
+
+
+private:
+    once_flag(const once_flag&); // = delete;
+    once_flag& operator=(const once_flag&); // = delete;
+
+    _State_type __state_;
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template<class _Callable, class... _Args>
+    friend
+    void call_once(once_flag&, _Callable&&, _Args&&...);
+#else  // _LIBCUDACXX_CXX03_LANG
+    template<class _Callable>
+    friend
+    void call_once(once_flag&, _Callable&);
+
+    template<class _Callable>
+    friend
+    void call_once(once_flag&, const _Callable&);
+#endif  // _LIBCUDACXX_CXX03_LANG
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Fp>
+class __call_once_param
+{
+    _Fp& __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __call_once_param(_Fp& __f) : __f_(__f) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()()
+    {
+        typedef typename __make_tuple_indices<tuple_size<_Fp>::value, 1>::type _Index;
+        __execute(_Index());
+    }
+
+private:
+    template <size_t ..._Indices>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __execute(__tuple_indices<_Indices...>)
+    {
+        __invoke(_CUDA_VSTD::get<0>(_CUDA_VSTD::move(__f_)), _CUDA_VSTD::get<_Indices>(_CUDA_VSTD::move(__f_))...);
+    }
+};
+
+#else
+
+template <class _Fp>
+class __call_once_param
+{
+    _Fp& __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __call_once_param(_Fp& __f) : __f_(__f) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()()
+    {
+        __f_();
+    }
+};
+
+#endif
+
+template <class _Fp>
+void
+__call_once_proxy(void* __vp)
+{
+    __call_once_param<_Fp>* __p = static_cast<__call_once_param<_Fp>*>(__vp);
+    (*__p)();
+}
+
+_LIBCUDACXX_FUNC_VIS void __call_once(volatile once_flag::_State_type&, void*,
+                                  void (*)(void*));
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _Callable, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+call_once(once_flag& __flag, _Callable&& __func, _Args&&... __args)
+{
+    if (__libcpp_acquire_load(&__flag.__state_) != ~once_flag::_State_type(0))
+    {
+        typedef tuple<_Callable&&, _Args&&...> _Gp;
+        _Gp __f(_CUDA_VSTD::forward<_Callable>(__func), _CUDA_VSTD::forward<_Args>(__args)...);
+        __call_once_param<_Gp> __p(__f);
+        __call_once(__flag.__state_, &__p, &__call_once_proxy<_Gp>);
+    }
+}
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template<class _Callable>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+call_once(once_flag& __flag, _Callable& __func)
+{
+    if (__libcpp_acquire_load(&__flag.__state_) != ~once_flag::_State_type(0))
+    {
+        __call_once_param<_Callable> __p(__func);
+        __call_once(__flag.__state_, &__p, &__call_once_proxy<_Callable>);
+    }
+}
+
+template<class _Callable>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+call_once(once_flag& __flag, const _Callable& __func)
+{
+    if (__libcpp_acquire_load(&__flag.__state_) != ~once_flag::_State_type(0))
+    {
+        __call_once_param<const _Callable> __p(__func);
+        __call_once(__flag.__state_, &__p, &__call_once_proxy<const _Callable>);
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_MUTEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/new b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/new
new file mode 100644
index 000000000000..8e4dbb571003
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/new
@@ -0,0 +1,378 @@
+// -*- C++ -*-
+//===----------------------------- new ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_NEW
+#define _LIBCUDACXX_NEW
+
+/*
+    new synopsis
+
+namespace std
+{
+
+class bad_alloc
+    : public exception
+{
+public:
+    bad_alloc() noexcept;
+    bad_alloc(const bad_alloc&) noexcept;
+    bad_alloc& operator=(const bad_alloc&) noexcept;
+    virtual const char* what() const noexcept;
+};
+
+class bad_array_new_length : public bad_alloc // C++14
+{
+public:
+    bad_array_new_length() noexcept;
+};
+
+enum class align_val_t : size_t {}; // C++17
+
+struct destroying_delete_t { // C++20
+  explicit destroying_delete_t() = default;
+};
+inline constexpr destroying_delete_t destroying_delete{}; // C++20
+
+struct nothrow_t { explicit nothrow_t() = default; };
+extern const nothrow_t nothrow;
+typedef void (*new_handler)();
+new_handler set_new_handler(new_handler new_p) noexcept;
+new_handler get_new_handler() noexcept;
+
+// 21.6.4, pointer optimization barrier
+template <class T> constexpr T* launder(T* p) noexcept; // C++17
+}  // std
+
+void* operator new(std::size_t size);                                   // replaceable, nodiscard in C++2a
+void* operator new(std::size_t size, std::align_val_t alignment);       // replaceable, C++17, nodiscard in C++2a
+void* operator new(std::size_t size, const std::nothrow_t&) noexcept;   // replaceable, nodiscard in C++2a
+void* operator new(std::size_t size, std::align_val_t alignment,
+                   const std::nothrow_t&) noexcept;                     // replaceable, C++17, nodiscard in C++2a
+void  operator delete(void* ptr) noexcept;                              // replaceable
+void  operator delete(void* ptr, std::size_t size) noexcept;            // replaceable, C++14
+void  operator delete(void* ptr, std::align_val_t alignment) noexcept;  // replaceable, C++17
+void  operator delete(void* ptr, std::size_t size,
+                      std::align_val_t alignment) noexcept;             // replaceable, C++17
+void  operator delete(void* ptr, const std::nothrow_t&) noexcept;       // replaceable
+void  operator delete(void* ptr, std:align_val_t alignment,
+                      const std::nothrow_t&) noexcept;                  // replaceable, C++17
+
+void* operator new[](std::size_t size);                                 // replaceable, nodiscard in C++2a
+void* operator new[](std::size_t size,
+                     std::align_val_t alignment) noexcept;              // replaceable, C++17, nodiscard in C++2a
+void* operator new[](std::size_t size, const std::nothrow_t&) noexcept; // replaceable, nodiscard in C++2a
+void* operator new[](std::size_t size, std::align_val_t alignment,
+                     const std::nothrow_t&) noexcept;                   // replaceable, C++17, nodiscard in C++2a
+void  operator delete[](void* ptr) noexcept;                            // replaceable
+void  operator delete[](void* ptr, std::size_t size) noexcept;          // replaceable, C++14
+void  operator delete[](void* ptr,
+                        std::align_val_t alignment) noexcept;           // replaceable, C++17
+void  operator delete[](void* ptr, std::size_t size,
+                        std::align_val_t alignment) noexcept;           // replaceable, C++17
+void  operator delete[](void* ptr, const std::nothrow_t&) noexcept;     // replaceable
+void  operator delete[](void* ptr, std::align_val_t alignment,
+                        const std::nothrow_t&) noexcept;                // replaceable, C++17
+
+void* operator new  (std::size_t size, void* ptr) noexcept;             // nodiscard in C++2a
+void* operator new[](std::size_t size, void* ptr) noexcept;             // nodiscard in C++2a
+void  operator delete  (void* ptr, void*) noexcept;
+void  operator delete[](void* ptr, void*) noexcept;
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <exception>
+#include <type_traits>
+#include <cstddef>
+#include <version>
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+#include <cstdlib>
+#endif
+#if defined(_LIBCUDACXX_ABI_VCRUNTIME)
+#include <new.h>
+#endif
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if !defined(__cpp_sized_deallocation) || __cpp_sized_deallocation  < 201309L
+#define _LIBCUDACXX_HAS_NO_LANGUAGE_SIZED_DEALLOCATION
+#endif
+
+#if !defined(_LIBCUDACXX_BUILDING_LIBRARY) && _LIBCUDACXX_STD_VER < 14 && \
+    defined(_LIBCUDACXX_HAS_NO_LANGUAGE_SIZED_DEALLOCATION)
+# define _LIBCUDACXX_HAS_NO_LIBRARY_SIZED_DEALLOCATION
+#endif
+
+#if defined(_LIBCUDACXX_HAS_NO_LIBRARY_SIZED_DEALLOCATION) || \
+    defined(_LIBCUDACXX_HAS_NO_LANGUAGE_SIZED_DEALLOCATION)
+# define _LIBCUDACXX_HAS_NO_SIZED_DEALLOCATION
+#endif
+
+#if !__has_builtin(__builtin_operator_new) || \
+   __has_builtin(__builtin_operator_new) < 201802L
+#define _LIBCUDACXX_HAS_NO_BUILTIN_OVERLOADED_OPERATOR_NEW_DELETE
+#endif
+
+namespace std  // purposefully not using versioning namespace
+{
+
+#if !defined(_LIBCUDACXX_ABI_VCRUNTIME)
+struct _LIBCUDACXX_TYPE_VIS nothrow_t { explicit nothrow_t() = default; };
+extern _LIBCUDACXX_FUNC_VIS const nothrow_t nothrow;
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_alloc
+    : public exception
+{
+public:
+    bad_alloc() _NOEXCEPT;
+    virtual ~bad_alloc() _NOEXCEPT;
+    virtual const char* what() const _NOEXCEPT;
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_array_new_length
+    : public bad_alloc
+{
+public:
+    bad_array_new_length() _NOEXCEPT;
+    virtual ~bad_array_new_length() _NOEXCEPT;
+    virtual const char* what() const _NOEXCEPT;
+};
+
+typedef void (*new_handler)();
+_LIBCUDACXX_FUNC_VIS new_handler set_new_handler(new_handler) _NOEXCEPT;
+_LIBCUDACXX_FUNC_VIS new_handler get_new_handler() _NOEXCEPT;
+
+#endif // !_LIBCUDACXX_ABI_VCRUNTIME
+
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS void __throw_bad_alloc();  // not in C++ spec
+
+#if !defined(_LIBCUDACXX_HAS_NO_LIBRARY_ALIGNED_ALLOCATION) && \
+    !defined(_LIBCUDACXX_ABI_VCRUNTIME)
+#ifndef _LIBCUDACXX_CXX03_LANG
+enum class _LIBCUDACXX_ENUM_VIS align_val_t : size_t { };
+#else
+enum align_val_t { __zero = 0, __max = (size_t)-1 };
+#endif
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17
+// Enable the declaration even if the compiler doesn't support the language
+// feature.
+struct destroying_delete_t {
+  explicit destroying_delete_t() = default;
+};
+_LIBCUDACXX_INLINE_VAR constexpr destroying_delete_t destroying_delete{};
+#endif // _LIBCUDACXX_STD_VER > 17
+
+}  // std
+
+#if defined(_LIBCUDACXX_CXX03_LANG)
+#define _THROW_BAD_ALLOC throw(std::bad_alloc)
+#else
+#define _THROW_BAD_ALLOC
+#endif
+
+#if !defined(_LIBCUDACXX_ABI_VCRUNTIME)
+
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new(std::size_t __sz) _THROW_BAD_ALLOC;
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new(std::size_t __sz, const std::nothrow_t&) _NOEXCEPT _LIBCUDACXX_NOALIAS;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete(void* __p) _NOEXCEPT;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete(void* __p, const std::nothrow_t&) _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_LIBRARY_SIZED_DEALLOCATION
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS _LIBCUDACXX_AVAILABILITY_SIZED_NEW_DELETE void  operator delete(void* __p, std::size_t __sz) _NOEXCEPT;
+#endif
+
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new[](std::size_t __sz) _THROW_BAD_ALLOC;
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new[](std::size_t __sz, const std::nothrow_t&) _NOEXCEPT _LIBCUDACXX_NOALIAS;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete[](void* __p) _NOEXCEPT;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete[](void* __p, const std::nothrow_t&) _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_LIBRARY_SIZED_DEALLOCATION
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS _LIBCUDACXX_AVAILABILITY_SIZED_NEW_DELETE void  operator delete[](void* __p, std::size_t __sz) _NOEXCEPT;
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_LIBRARY_ALIGNED_ALLOCATION
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new(std::size_t __sz, std::align_val_t) _THROW_BAD_ALLOC;
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new(std::size_t __sz, std::align_val_t, const std::nothrow_t&) _NOEXCEPT _LIBCUDACXX_NOALIAS;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete(void* __p, std::align_val_t) _NOEXCEPT;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete(void* __p, std::align_val_t, const std::nothrow_t&) _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_LIBRARY_SIZED_DEALLOCATION
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS _LIBCUDACXX_AVAILABILITY_SIZED_NEW_DELETE void  operator delete(void* __p, std::size_t __sz, std::align_val_t) _NOEXCEPT;
+#endif
+
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new[](std::size_t __sz, std::align_val_t) _THROW_BAD_ALLOC;
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_OVERRIDABLE_FUNC_VIS void* operator new[](std::size_t __sz, std::align_val_t, const std::nothrow_t&) _NOEXCEPT _LIBCUDACXX_NOALIAS;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete[](void* __p, std::align_val_t) _NOEXCEPT;
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS void  operator delete[](void* __p, std::align_val_t, const std::nothrow_t&) _NOEXCEPT;
+#ifndef _LIBCUDACXX_HAS_NO_LIBRARY_SIZED_DEALLOCATION
+_LIBCUDACXX_OVERRIDABLE_FUNC_VIS _LIBCUDACXX_AVAILABILITY_SIZED_NEW_DELETE void  operator delete[](void* __p, std::size_t __sz, std::align_val_t) _NOEXCEPT;
+#endif
+#endif
+
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 inline _LIBCUDACXX_INLINE_VISIBILITY void* operator new  (std::size_t, void* __p) _NOEXCEPT {return __p;}
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 inline _LIBCUDACXX_INLINE_VISIBILITY void* operator new[](std::size_t, void* __p) _NOEXCEPT {return __p;}
+inline _LIBCUDACXX_INLINE_VISIBILITY void  operator delete  (void*, void*) _NOEXCEPT {}
+inline _LIBCUDACXX_INLINE_VISIBILITY void  operator delete[](void*, void*) _NOEXCEPT {}
+
+#endif // !_LIBCUDACXX_ABI_VCRUNTIME
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+_LIBCUDACXX_CONSTEXPR inline _LIBCUDACXX_INLINE_VISIBILITY bool __is_overaligned_for_new(size_t __align) _NOEXCEPT {
+#ifdef __STDCPP_DEFAULT_NEW_ALIGNMENT__
+  return __align > __STDCPP_DEFAULT_NEW_ALIGNMENT__;
+#else
+  return __align > alignment_of<max_align_t>::value;
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void *__libcpp_allocate(size_t __size, size_t __align) {
+#ifndef _LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION
+  if (__is_overaligned_for_new(__align)) {
+    const align_val_t __align_val = static_cast<align_val_t>(__align);
+# ifdef _LIBCUDACXX_HAS_NO_BUILTIN_OVERLOADED_OPERATOR_NEW_DELETE
+    return ::operator new(__size, __align_val);
+# else
+    return __builtin_operator_new(__size, __align_val);
+# endif
+  }
+#else
+  ((void)__align);
+#endif
+#ifdef _LIBCUDACXX_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE
+  return ::operator new(__size);
+#else
+  return __builtin_operator_new(__size);
+#endif
+}
+
+struct _DeallocateCaller {
+  static inline _LIBCUDACXX_INLINE_VISIBILITY
+  void __do_deallocate_handle_size_align(void *__ptr, size_t __size, size_t __align) {
+#if defined(_LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION)
+    ((void)__align);
+    return __do_deallocate_handle_size(__ptr, __size);
+#else
+    if (__is_overaligned_for_new(__align)) {
+      const align_val_t __align_val = static_cast<align_val_t>(__align);
+      return __do_deallocate_handle_size(__ptr, __size, __align_val);
+    } else {
+      return __do_deallocate_handle_size(__ptr, __size);
+    }
+#endif
+  }
+
+  static inline _LIBCUDACXX_INLINE_VISIBILITY
+  void __do_deallocate_handle_align(void *__ptr, size_t __align) {
+#if defined(_LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION)
+    ((void)__align);
+    return __do_call(__ptr);
+#else
+    if (__is_overaligned_for_new(__align)) {
+      const align_val_t __align_val = static_cast<align_val_t>(__align);
+      return __do_call(__ptr, __align_val);
+    } else {
+      return __do_call(__ptr);
+    }
+#endif
+  }
+
+ private:
+  static inline void __do_deallocate_handle_size(void *__ptr, size_t __size) {
+#ifdef _LIBCUDACXX_HAS_NO_SIZED_DEALLOCATION
+    ((void)__size);
+    return __do_call(__ptr);
+#else
+    return __do_call(__ptr, __size);
+#endif
+  }
+
+#ifndef _LIBCUDACXX_HAS_NO_ALIGNED_ALLOCATION
+  static inline void __do_deallocate_handle_size(void *__ptr, size_t __size, align_val_t __align) {
+#ifdef _LIBCUDACXX_HAS_NO_SIZED_DEALLOCATION
+    ((void)__size);
+    return __do_call(__ptr, __align);
+#else
+    return __do_call(__ptr, __size, __align);
+#endif
+  }
+#endif
+
+private:
+  template <class _A1, class _A2>
+  static inline void __do_call(void *__ptr, _A1 __a1, _A2 __a2) {
+#if defined(_LIBCUDACXX_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE) || \
+    defined(_LIBCUDACXX_HAS_NO_BUILTIN_OVERLOADED_OPERATOR_NEW_DELETE)
+    return ::operator delete(__ptr, __a1, __a2);
+#else
+    return __builtin_operator_delete(__ptr, __a1, __a2);
+#endif
+  }
+
+  template <class _A1>
+  static inline void __do_call(void *__ptr, _A1 __a1) {
+#if defined(_LIBCUDACXX_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE) || \
+    defined(_LIBCUDACXX_HAS_NO_BUILTIN_OVERLOADED_OPERATOR_NEW_DELETE)
+    return ::operator delete(__ptr, __a1);
+#else
+    return __builtin_operator_delete(__ptr, __a1);
+#endif
+  }
+
+  static inline void __do_call(void *__ptr) {
+#ifdef _LIBCUDACXX_HAS_NO_BUILTIN_OPERATOR_NEW_DELETE
+    return ::operator delete(__ptr);
+#else
+    return __builtin_operator_delete(__ptr);
+#endif
+  }
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void __libcpp_deallocate(void* __ptr, size_t __size, size_t __align) {
+  _DeallocateCaller::__do_deallocate_handle_size_align(__ptr, __size, __align);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY void __libcpp_deallocate_unsized(void* __ptr, size_t __align) {
+  _DeallocateCaller::__do_deallocate_handle_align(__ptr, __align);
+}
+
+template <class _Tp>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 inline
+_LIBCUDACXX_CONSTEXPR _Tp* __launder(_Tp* __p) _NOEXCEPT
+{
+    static_assert (!(is_function<_Tp>::value), "can't launder functions" );
+    static_assert (!(is_same<void, typename remove_cv<_Tp>::type>::value), "can't launder cv-void" );
+#if defined(_LIBCUDACXX_LAUNDER)
+    return _LIBCUDACXX_LAUNDER(__p);
+#else
+    return __p;
+#endif // defined(_LIBCUDACXX_LAUNDER)
+}
+
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp>
+_LIBCUDACXX_NODISCARD_AFTER_CXX17 inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _Tp* launder(_Tp* __p) noexcept
+{
+    return _CUDA_VSTD::__launder(__p);
+}
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_NEW
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/numeric b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/numeric
new file mode 100644
index 000000000000..3584f29774fe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/numeric
@@ -0,0 +1,593 @@
+// -*- C++ -*-
+//===---------------------------- numeric ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_NUMERIC
+#define _LIBCUDACXX_NUMERIC
+
+/*
+    numeric synopsis
+
+namespace std
+{
+
+template <class InputIterator, class T>
+    T
+    accumulate(InputIterator first, InputIterator last, T init);
+
+template <class InputIterator, class T, class BinaryOperation>
+    T
+    accumulate(InputIterator first, InputIterator last, T init, BinaryOperation binary_op);
+
+template<class InputIterator>
+    typename iterator_traits<InputIterator>::value_type
+    reduce(InputIterator first, InputIterator last);  // C++17
+
+template<class InputIterator, class T>
+    T
+    reduce(InputIterator first, InputIterator last, T init);  // C++17
+
+template<class InputIterator, class T, class BinaryOperation>
+    T
+    reduce(InputIterator first, InputIterator last, T init, BinaryOperation binary_op);  // C++17
+
+template <class InputIterator1, class InputIterator2, class T>
+    T
+    inner_product(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, T init);
+
+template <class InputIterator1, class InputIterator2, class T, class BinaryOperation1, class BinaryOperation2>
+    T
+    inner_product(InputIterator1 first1, InputIterator1 last1, InputIterator2 first2,
+                  T init, BinaryOperation1 binary_op1, BinaryOperation2 binary_op2);
+
+
+template<class InputIterator1, class InputIterator2, class T>
+    T
+    transform_reduce(InputIterator1 first1, InputIterator1 last1,
+                     InputIterator2 first2, T init);  // C++17
+
+template<class InputIterator1, class InputIterator2, class T, class BinaryOperation1, class BinaryOperation2>
+    T
+    transform_reduce(InputIterator1 first1, InputIterator1 last1,
+                     InputIterator2 first2, T init,
+                     BinaryOperation1 binary_op1, BinaryOperation2 binary_op2);  // C++17
+
+template<class InputIterator, class T, class BinaryOperation, class UnaryOperation>
+    T
+    transform_reduce(InputIterator first, InputIterator last, T init,
+                     BinaryOperation binary_op, UnaryOperation unary_op);  // C++17
+
+template <class InputIterator, class OutputIterator>
+    OutputIterator
+    partial_sum(InputIterator first, InputIterator last, OutputIterator result);
+
+template <class InputIterator, class OutputIterator, class BinaryOperation>
+    OutputIterator
+    partial_sum(InputIterator first, InputIterator last, OutputIterator result, BinaryOperation binary_op);
+
+template<class InputIterator, class OutputIterator, class T>
+    OutputIterator
+    exclusive_scan(InputIterator first, InputIterator last,
+                   OutputIterator result, T init); // C++17
+
+template<class InputIterator, class OutputIterator, class T, class BinaryOperation>
+    OutputIterator
+    exclusive_scan(InputIterator first, InputIterator last,
+                   OutputIterator result, T init, BinaryOperation binary_op); // C++17
+
+template<class InputIterator, class OutputIterator>
+    OutputIterator
+    inclusive_scan(InputIterator first, InputIterator last, OutputIterator result);  // C++17
+
+template<class InputIterator, class OutputIterator, class BinaryOperation>
+    OutputIterator
+    inclusive_scan(InputIterator first, InputIterator last,
+                   OutputIterator result, BinaryOperation binary_op);  // C++17
+
+template<class InputIterator, class OutputIterator, class BinaryOperation, class T>
+    OutputIterator
+    inclusive_scan(InputIterator first, InputIterator last,
+                   OutputIterator result, BinaryOperation binary_op, T init);  // C++17
+
+template<class InputIterator, class OutputIterator, class T,
+         class BinaryOperation, class UnaryOperation>
+    OutputIterator
+    transform_exclusive_scan(InputIterator first, InputIterator last,
+                             OutputIterator result, T init,
+                             BinaryOperation binary_op, UnaryOperation unary_op);  // C++17
+
+template<class InputIterator, class OutputIterator,
+         class BinaryOperation, class UnaryOperation>
+    OutputIterator
+    transform_inclusive_scan(InputIterator first, InputIterator last,
+                             OutputIterator result,
+                             BinaryOperation binary_op, UnaryOperation unary_op);  // C++17
+
+template<class InputIterator, class OutputIterator,
+         class BinaryOperation, class UnaryOperation, class T>
+    OutputIterator
+    transform_inclusive_scan(InputIterator first, InputIterator last,
+                             OutputIterator result,
+                             BinaryOperation binary_op, UnaryOperation unary_op,
+                             T init);  // C++17
+
+template <class InputIterator, class OutputIterator>
+    OutputIterator
+    adjacent_difference(InputIterator first, InputIterator last, OutputIterator result);
+
+template <class InputIterator, class OutputIterator, class BinaryOperation>
+    OutputIterator
+    adjacent_difference(InputIterator first, InputIterator last, OutputIterator result, BinaryOperation binary_op);
+
+template <class ForwardIterator, class T>
+    void iota(ForwardIterator first, ForwardIterator last, T value);
+
+template <class M, class N>
+    constexpr common_type_t<M,N> gcd(M m, N n);    // C++17
+
+template <class M, class N>
+    constexpr common_type_t<M,N> lcm(M m, N n);    // C++17
+
+integer         midpoint(integer a, integer b);                  // C++20
+pointer         midpoint(pointer a, pointer b);                  // C++20
+floating_point  midpoint(floating_point a, floating_point b);    // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <iterator>
+#include <limits> // for numeric_limits
+#include <functional>
+#include <cmath> // for isnormal
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _InputIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+accumulate(_InputIterator __first, _InputIterator __last, _Tp __init)
+{
+    for (; __first != __last; ++__first)
+        __init = __init + *__first;
+    return __init;
+}
+
+template <class _InputIterator, class _Tp, class _BinaryOperation>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+accumulate(_InputIterator __first, _InputIterator __last, _Tp __init, _BinaryOperation __binary_op)
+{
+    for (; __first != __last; ++__first)
+        __init = __binary_op(__init, *__first);
+    return __init;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _InputIterator, class _Tp, class _BinaryOp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+reduce(_InputIterator __first, _InputIterator __last, _Tp __init, _BinaryOp __b)
+{
+    for (; __first != __last; ++__first)
+        __init = __b(__init, *__first);
+    return __init;
+}
+
+template <class _InputIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+reduce(_InputIterator __first, _InputIterator __last, _Tp __init)
+{
+    return _CUDA_VSTD::reduce(__first, __last, __init, _CUDA_VSTD::plus<>());
+}
+
+template <class _InputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename iterator_traits<_InputIterator>::value_type
+reduce(_InputIterator __first, _InputIterator __last)
+{
+    return _CUDA_VSTD::reduce(__first, __last,
+       typename iterator_traits<_InputIterator>::value_type{});
+}
+#endif
+
+template <class _InputIterator1, class _InputIterator2, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+inner_product(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2, _Tp __init)
+{
+    for (; __first1 != __last1; ++__first1, (void) ++__first2)
+        __init = __init + *__first1 * *__first2;
+    return __init;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _Tp, class _BinaryOperation1, class _BinaryOperation2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+inner_product(_InputIterator1 __first1, _InputIterator1 __last1, _InputIterator2 __first2,
+              _Tp __init, _BinaryOperation1 __binary_op1, _BinaryOperation2 __binary_op2)
+{
+    for (; __first1 != __last1; ++__first1, (void) ++__first2)
+        __init = __binary_op1(__init, __binary_op2(*__first1, *__first2));
+    return __init;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _InputIterator, class _Tp, class _BinaryOp, class _UnaryOp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+transform_reduce(_InputIterator __first, _InputIterator __last,
+           _Tp __init,  _BinaryOp __b, _UnaryOp __u)
+{
+    for (; __first != __last; ++__first)
+        __init = __b(__init, __u(*__first));
+    return __init;
+}
+
+template <class _InputIterator1, class _InputIterator2,
+          class _Tp, class _BinaryOp1, class _BinaryOp2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+transform_reduce(_InputIterator1 __first1, _InputIterator1 __last1,
+                 _InputIterator2 __first2, _Tp __init,  _BinaryOp1 __b1, _BinaryOp2 __b2)
+{
+    for (; __first1 != __last1; ++__first1, (void) ++__first2)
+        __init = __b1(__init, __b2(*__first1, *__first2));
+    return __init;
+}
+
+template <class _InputIterator1, class _InputIterator2, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp
+transform_reduce(_InputIterator1 __first1, _InputIterator1 __last1,
+                 _InputIterator2 __first2, _Tp __init)
+{
+    return _CUDA_VSTD::transform_reduce(__first1, __last1, __first2, _CUDA_VSTD::move(__init),
+                                   _CUDA_VSTD::plus<>(), _CUDA_VSTD::multiplies<>());
+}
+#endif
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+partial_sum(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    if (__first != __last)
+    {
+        typename iterator_traits<_InputIterator>::value_type __t(*__first);
+        *__result = __t;
+        for (++__first, (void) ++__result; __first != __last; ++__first, (void) ++__result)
+        {
+            __t = __t + *__first;
+            *__result = __t;
+        }
+    }
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _BinaryOperation>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+partial_sum(_InputIterator __first, _InputIterator __last, _OutputIterator __result,
+              _BinaryOperation __binary_op)
+{
+    if (__first != __last)
+    {
+        typename iterator_traits<_InputIterator>::value_type __t(*__first);
+        *__result = __t;
+        for (++__first, (void) ++__result; __first != __last; ++__first, (void) ++__result)
+        {
+            __t = __binary_op(__t, *__first);
+            *__result = __t;
+        }
+    }
+    return __result;
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _InputIterator, class _OutputIterator, class _Tp, class _BinaryOp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+exclusive_scan(_InputIterator __first, _InputIterator __last,
+               _OutputIterator __result, _Tp __init, _BinaryOp __b)
+{
+    if (__first != __last)
+    {
+        _Tp __saved = __init;
+        do
+        {
+            __init = __b(__init, *__first);
+            *__result = __saved;
+            __saved = __init;
+            ++__result;
+        } while (++__first != __last);
+    }
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+exclusive_scan(_InputIterator __first, _InputIterator __last,
+               _OutputIterator __result, _Tp __init)
+{
+    return _CUDA_VSTD::exclusive_scan(__first, __last, __result, __init, _CUDA_VSTD::plus<>());
+}
+
+template <class _InputIterator, class _OutputIterator, class _BinaryOp, class _Tp>
+_OutputIterator inclusive_scan(_InputIterator __first, _InputIterator __last,
+                               _OutputIterator __result, _BinaryOp __b,  _Tp __init)
+{
+    for (; __first != __last; ++__first, (void) ++__result) {
+        __init = __b(__init, *__first);
+        *__result = __init;
+        }
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _BinaryOp>
+_OutputIterator inclusive_scan(_InputIterator __first, _InputIterator __last,
+                               _OutputIterator __result, _BinaryOp __b)
+{
+    if (__first != __last) {
+        typename std::iterator_traits<_InputIterator>::value_type __init = *__first;
+        *__result++ = __init;
+        if (++__first != __last)
+            return _CUDA_VSTD::inclusive_scan(__first, __last, __result, __b, __init);
+        }
+
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator>
+_OutputIterator inclusive_scan(_InputIterator __first, _InputIterator __last,
+                               _OutputIterator __result)
+{
+    return _CUDA_VSTD::inclusive_scan(__first, __last, __result, std::plus<>());
+}
+
+template <class _InputIterator, class _OutputIterator, class _Tp,
+          class _BinaryOp, class _UnaryOp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+transform_exclusive_scan(_InputIterator __first, _InputIterator __last,
+                           _OutputIterator __result, _Tp __init,
+                           _BinaryOp __b, _UnaryOp __u)
+{
+    if (__first != __last)
+    {
+        _Tp __saved = __init;
+        do
+        {
+            __init = __b(__init, __u(*__first));
+            *__result = __saved;
+            __saved = __init;
+            ++__result;
+        } while (++__first != __last);
+    }
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _Tp, class _BinaryOp, class _UnaryOp>
+_OutputIterator transform_inclusive_scan(_InputIterator __first, _InputIterator __last,
+                           _OutputIterator __result, _BinaryOp __b, _UnaryOp __u, _Tp __init)
+{
+    for (; __first != __last; ++__first, (void) ++__result) {
+        __init = __b(__init, __u(*__first));
+        *__result = __init;
+        }
+
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _BinaryOp, class _UnaryOp>
+_OutputIterator transform_inclusive_scan(_InputIterator __first, _InputIterator __last,
+                               _OutputIterator __result, _BinaryOp __b, _UnaryOp __u)
+{
+    if (__first != __last) {
+        typename std::iterator_traits<_InputIterator>::value_type __init = __u(*__first);
+        *__result++ = __init;
+        if (++__first != __last)
+            return _CUDA_VSTD::transform_inclusive_scan(__first, __last, __result, __b, __u, __init);
+        }
+
+    return __result;
+}
+#endif
+
+template <class _InputIterator, class _OutputIterator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+adjacent_difference(_InputIterator __first, _InputIterator __last, _OutputIterator __result)
+{
+    if (__first != __last)
+    {
+        typename iterator_traits<_InputIterator>::value_type __t1(*__first);
+        *__result = __t1;
+        for (++__first, (void) ++__result; __first != __last; ++__first, (void) ++__result)
+        {
+            typename iterator_traits<_InputIterator>::value_type __t2(*__first);
+            *__result = __t2 - __t1;
+            __t1 = _CUDA_VSTD::move(__t2);
+        }
+    }
+    return __result;
+}
+
+template <class _InputIterator, class _OutputIterator, class _BinaryOperation>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+adjacent_difference(_InputIterator __first, _InputIterator __last, _OutputIterator __result,
+                      _BinaryOperation __binary_op)
+{
+    if (__first != __last)
+    {
+        typename iterator_traits<_InputIterator>::value_type __t1(*__first);
+        *__result = __t1;
+        for (++__first, (void) ++__result; __first != __last; ++__first, (void) ++__result)
+        {
+            typename iterator_traits<_InputIterator>::value_type __t2(*__first);
+            *__result = __binary_op(__t2, __t1);
+            __t1 = _CUDA_VSTD::move(__t2);
+        }
+    }
+    return __result;
+}
+
+template <class _ForwardIterator, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+iota(_ForwardIterator __first, _ForwardIterator __last, _Tp __value_)
+{
+    for (; __first != __last; ++__first, (void) ++__value_)
+        *__first = __value_;
+}
+
+
+#if _LIBCUDACXX_STD_VER > 14
+template <typename _Result, typename _Source, bool _IsSigned = is_signed<_Source>::value> struct __ct_abs;
+
+template <typename _Result, typename _Source>
+struct __ct_abs<_Result, _Source, true> {
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    _Result operator()(_Source __t) const noexcept
+    {
+    if (__t >= 0) return __t;
+    if (__t == numeric_limits<_Source>::min()) return -static_cast<_Result>(__t);
+    return -__t;
+    }
+};
+
+template <typename _Result, typename _Source>
+struct __ct_abs<_Result, _Source, false> {
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    _Result operator()(_Source __t) const noexcept { return __t; }
+};
+
+
+template<class _Tp>
+_LIBCUDACXX_CONSTEXPR _LIBCUDACXX_HIDDEN
+_Tp __gcd(_Tp __m, _Tp __n)
+{
+    static_assert((!is_signed<_Tp>::value), "");
+    return __n == 0 ? __m : _CUDA_VSTD::__gcd<_Tp>(__n, __m % __n);
+}
+
+
+template<class _Tp, class _Up>
+_LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+common_type_t<_Tp,_Up>
+gcd(_Tp __m, _Up __n)
+{
+    static_assert((is_integral<_Tp>::value && is_integral<_Up>::value), "Arguments to gcd must be integer types");
+    static_assert((!is_same<typename remove_cv<_Tp>::type, bool>::value), "First argument to gcd cannot be bool" );
+    static_assert((!is_same<typename remove_cv<_Up>::type, bool>::value), "Second argument to gcd cannot be bool" );
+    using _Rp = common_type_t<_Tp,_Up>;
+    using _Wp = make_unsigned_t<_Rp>;
+    return static_cast<_Rp>(_CUDA_VSTD::__gcd(
+        static_cast<_Wp>(__ct_abs<_Rp, _Tp>()(__m)),
+        static_cast<_Wp>(__ct_abs<_Rp, _Up>()(__n))));
+}
+
+template<class _Tp, class _Up>
+_LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+common_type_t<_Tp,_Up>
+lcm(_Tp __m, _Up __n)
+{
+    static_assert((is_integral<_Tp>::value && is_integral<_Up>::value), "Arguments to lcm must be integer types");
+    static_assert((!is_same<typename remove_cv<_Tp>::type, bool>::value), "First argument to lcm cannot be bool" );
+    static_assert((!is_same<typename remove_cv<_Up>::type, bool>::value), "Second argument to lcm cannot be bool" );
+    if (__m == 0 || __n == 0)
+        return 0;
+
+    using _Rp = common_type_t<_Tp,_Up>;
+    _Rp __val1 = __ct_abs<_Rp, _Tp>()(__m) / _CUDA_VSTD::gcd(__m, __n);
+    _Rp __val2 = __ct_abs<_Rp, _Up>()(__n);
+    _LIBCUDACXX_ASSERT((numeric_limits<_Rp>::max() / __val1 > __val2), "Overflow in lcm");
+    return __val1 * __val2;
+}
+
+#endif /* _LIBCUDACXX_STD_VER > 14 */
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+enable_if_t<is_integral_v<_Tp> && !is_same_v<bool, _Tp> && !is_null_pointer_v<_Tp>, _Tp>
+midpoint(_Tp __a, _Tp __b) noexcept
+_LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+{
+    using _Up = std::make_unsigned_t<_Tp>;
+
+    int __sign = 1;
+    _Up __m = __a;
+    _Up __M = __b;
+    if (__a > __b)
+    {
+        __sign = -1;
+        __m = __b;
+        __M = __a;
+    }
+     return __a + __sign * _Tp(_Up(__M-__m) >> 1);
+}
+
+
+template <class _TPtr>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+enable_if_t<is_pointer_v<_TPtr>
+             && is_object_v<remove_pointer_t<_TPtr>>
+             && ! is_void_v<remove_pointer_t<_TPtr>>
+             && (sizeof(remove_pointer_t<_TPtr>) > 0), _TPtr>
+midpoint(_TPtr __a, _TPtr __b) noexcept
+{
+    return __a + _CUDA_VSTD::midpoint(ptrdiff_t(0), __b - __a);
+}
+
+
+template <typename _Tp>
+constexpr int __sign(_Tp __val) {
+    return (_Tp(0) < __val) - (__val < _Tp(0));
+}
+
+template <typename _Fp>
+constexpr _Fp __fp_abs(_Fp __f) { return __f >= 0 ? __f : -__f; }
+
+template <class _Fp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+enable_if_t<is_floating_point_v<_Fp>, _Fp>
+midpoint(_Fp __a, _Fp __b) noexcept
+{
+	constexpr _Fp __lo = numeric_limits<_Fp>::min()*2;
+	constexpr _Fp __hi = numeric_limits<_Fp>::max()/2;
+    return __fp_abs(__a) <= __hi && __fp_abs(__b) <= __hi ?  // typical case: overflow is impossible
+      (__a + __b)/2 :                                        // always correctly rounded
+      __fp_abs(__a) < __lo ? __a + __b/2 :                   // not safe to halve a
+      __fp_abs(__a) < __lo ? __a/2 + __b :                   // not safe to halve b
+      __a/2 + __b/2;                                         // otherwise correctly rounded
+}
+
+#endif // _LIBCUDACXX_STD_VER > 17
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#if defined(_LIBCUDACXX_HAS_PARALLEL_ALGORITHMS) && _LIBCUDACXX_STD_VER >= 17
+#   include <__pstl_numeric>
+#endif
+
+#endif  // _LIBCUDACXX_NUMERIC
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/optional b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/optional
new file mode 100644
index 000000000000..47b406a33864
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/optional
@@ -0,0 +1,1420 @@
+// -*- C++ -*-
+//===-------------------------- optional ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_OPTIONAL
+#define _LIBCUDACXX_OPTIONAL
+
+/*
+    optional synopsis
+
+// C++1z
+
+namespace std {
+  // 23.6.3, optional for object types
+  template <class T> class optional;
+
+  // 23.6.4, no-value state indicator
+  struct nullopt_t{see below };
+  inline constexpr nullopt_t nullopt(unspecified );
+
+  // 23.6.5, class bad_optional_access
+  class bad_optional_access;
+
+  // 23.6.6, relational operators
+  template <class T, class U>
+  constexpr bool operator==(const optional<T>&, const optional<U>&);
+  template <class T, class U>
+  constexpr bool operator!=(const optional<T>&, const optional<U>&);
+  template <class T, class U>
+  constexpr bool operator<(const optional<T>&, const optional<U>&);
+  template <class T, class U>
+  constexpr bool operator>(const optional<T>&, const optional<U>&);
+  template <class T, class U>
+  constexpr bool operator<=(const optional<T>&, const optional<U>&);
+  template <class T, class U>
+  constexpr bool operator>=(const optional<T>&, const optional<U>&);
+
+  // 23.6.7 comparison with nullopt
+  template <class T> constexpr bool operator==(const optional<T>&, nullopt_t) noexcept;
+  template <class T> constexpr bool operator==(nullopt_t, const optional<T>&) noexcept;
+  template <class T> constexpr bool operator!=(const optional<T>&, nullopt_t) noexcept;
+  template <class T> constexpr bool operator!=(nullopt_t, const optional<T>&) noexcept;
+  template <class T> constexpr bool operator<(const optional<T>&, nullopt_t) noexcept;
+  template <class T> constexpr bool operator<(nullopt_t, const optional<T>&) noexcept;
+  template <class T> constexpr bool operator<=(const optional<T>&, nullopt_t) noexcept;
+  template <class T> constexpr bool operator<=(nullopt_t, const optional<T>&) noexcept;
+  template <class T> constexpr bool operator>(const optional<T>&, nullopt_t) noexcept;
+  template <class T> constexpr bool operator>(nullopt_t, const optional<T>&) noexcept;
+  template <class T> constexpr bool operator>=(const optional<T>&, nullopt_t) noexcept;
+  template <class T> constexpr bool operator>=(nullopt_t, const optional<T>&) noexcept;
+
+  // 23.6.8, comparison with T
+  template <class T, class U> constexpr bool operator==(const optional<T>&, const U&);
+  template <class T, class U> constexpr bool operator==(const T&, const optional<U>&);
+  template <class T, class U> constexpr bool operator!=(const optional<T>&, const U&);
+  template <class T, class U> constexpr bool operator!=(const T&, const optional<U>&);
+  template <class T, class U> constexpr bool operator<(const optional<T>&, const U&);
+  template <class T, class U> constexpr bool operator<(const T&, const optional<U>&);
+  template <class T, class U> constexpr bool operator<=(const optional<T>&, const U&);
+  template <class T, class U> constexpr bool operator<=(const T&, const optional<U>&);
+  template <class T, class U> constexpr bool operator>(const optional<T>&, const U&);
+  template <class T, class U> constexpr bool operator>(const T&, const optional<U>&);
+  template <class T, class U> constexpr bool operator>=(const optional<T>&, const U&);
+  template <class T, class U> constexpr bool operator>=(const T&, const optional<U>&);
+
+  // 23.6.9, specialized algorithms
+  template <class T> void swap(optional<T>&, optional<T>&) noexcept(see below );
+  template <class T> constexpr optional<see below > make_optional(T&&);
+  template <class T, class... Args>
+    constexpr optional<T> make_optional(Args&&... args);
+  template <class T, class U, class... Args>
+    constexpr optional<T> make_optional(initializer_list<U> il, Args&&... args);
+
+  // 23.6.10, hash support
+  template <class T> struct hash;
+  template <class T> struct hash<optional<T>>;
+
+  template <class T> class optional {
+  public:
+    using value_type = T;
+
+    // 23.6.3.1, constructors
+    constexpr optional() noexcept;
+    constexpr optional(nullopt_t) noexcept;
+    optional(const optional &);
+    optional(optional &&) noexcept(see below);
+    template <class... Args> constexpr explicit optional(in_place_t, Args &&...);
+    template <class U, class... Args>
+      constexpr explicit optional(in_place_t, initializer_list<U>, Args &&...);
+    template <class U = T>
+      constexpr EXPLICIT optional(U &&);
+    template <class U>
+      constexpr EXPLICIT optional(const optional<U> &);
+    template <class U>
+      constexpr EXPLICIT optional(optional<U> &&);
+
+    // 23.6.3.2, destructor
+    ~optional();
+
+    // 23.6.3.3, assignment
+    optional &operator=(nullopt_t) noexcept;
+    optional &operator=(const optional &);                // constexpr in C++20
+    optional &operator=(optional &&) noexcept(see below); // constexpr in C++20
+    template <class U = T> optional &operator=(U &&);
+    template <class U> optional &operator=(const optional<U> &);
+    template <class U> optional &operator=(optional<U> &&);
+    template <class... Args> T& emplace(Args &&...);
+    template <class U, class... Args>
+      T& emplace(initializer_list<U>, Args &&...);
+
+    // 23.6.3.4, swap
+    void swap(optional &) noexcept(see below );
+
+    // 23.6.3.5, observers
+    constexpr T const *operator->() const;
+    constexpr T *operator->();
+    constexpr T const &operator*() const &;
+    constexpr T &operator*() &;
+    constexpr T &&operator*() &&;
+    constexpr const T &&operator*() const &&;
+    constexpr explicit operator bool() const noexcept;
+    constexpr bool has_value() const noexcept;
+    constexpr T const &value() const &;
+    constexpr T &value() &;
+    constexpr T &&value() &&;
+    constexpr const T &&value() const &&;
+    template <class U> constexpr T value_or(U &&) const &;
+    template <class U> constexpr T value_or(U &&) &&;
+
+    // 23.6.3.6, modifiers
+    void reset() noexcept;
+
+  private:
+    T *val; // exposition only
+  };
+
+template<class T>
+  optional(T) -> optional<T>;
+
+} // namespace std
+
+*/
+
+#include <__config>
+#include <__debug>
+#include <__functional_base>
+#include <functional>
+#include <initializer_list>
+#include <new>
+#include <stdexcept>
+#include <type_traits>
+#include <utility>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+namespace std  // purposefully not using versioning namespace
+{
+
+class _LIBCUDACXX_EXCEPTION_ABI _LIBCUDACXX_AVAILABILITY_BAD_OPTIONAL_ACCESS bad_optional_access
+    : public exception
+{
+public:
+    // Get the key function ~bad_optional_access() into the dylib
+    virtual ~bad_optional_access() _NOEXCEPT;
+    virtual const char* what() const _NOEXCEPT;
+};
+
+}  // std
+
+#if _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+_LIBCUDACXX_NORETURN
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS
+void __throw_bad_optional_access() {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        throw bad_optional_access();
+#else
+        _CUDA_VSTD::abort();
+#endif
+}
+
+struct nullopt_t
+{
+    struct __secret_tag { _LIBCUDACXX_INLINE_VISIBILITY explicit __secret_tag() = default; };
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr explicit nullopt_t(__secret_tag, __secret_tag) noexcept {}
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr nullopt_t nullopt{nullopt_t::__secret_tag{}, nullopt_t::__secret_tag{}};
+
+template <class _Tp, bool = is_trivially_destructible<_Tp>::value>
+struct __optional_destruct_base;
+
+template <class _Tp>
+struct __optional_destruct_base<_Tp, false>
+{
+    typedef _Tp value_type;
+    static_assert(is_object_v<value_type>,
+        "instantiation of optional with a non-object type is undefined behavior");
+    union
+    {
+        char __null_state_;
+        value_type __val_;
+    };
+    bool __engaged_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__optional_destruct_base()
+    {
+        if (__engaged_)
+            __val_.~value_type();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr __optional_destruct_base() noexcept
+        :  __null_state_(),
+           __engaged_(false) {}
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit __optional_destruct_base(in_place_t, _Args&&... __args)
+        :  __val_(_CUDA_VSTD::forward<_Args>(__args)...),
+           __engaged_(true) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() noexcept
+    {
+        if (__engaged_)
+        {
+            __val_.~value_type();
+            __engaged_ = false;
+        }
+    }
+};
+
+template <class _Tp>
+struct __optional_destruct_base<_Tp, true>
+{
+    typedef _Tp value_type;
+    static_assert(is_object_v<value_type>,
+        "instantiation of optional with a non-object type is undefined behavior");
+    union
+    {
+        char __null_state_;
+        value_type __val_;
+    };
+    bool __engaged_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr __optional_destruct_base() noexcept
+        :  __null_state_(),
+           __engaged_(false) {}
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit __optional_destruct_base(in_place_t, _Args&&... __args)
+        :  __val_(_CUDA_VSTD::forward<_Args>(__args)...),
+           __engaged_(true) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() noexcept
+    {
+        if (__engaged_)
+        {
+            __engaged_ = false;
+        }
+    }
+};
+
+template <class _Tp, bool = is_reference<_Tp>::value>
+struct __optional_storage_base : __optional_destruct_base<_Tp>
+{
+    using __base = __optional_destruct_base<_Tp>;
+    using value_type = _Tp;
+    using __base::__base;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr bool has_value() const noexcept
+    {
+        return this->__engaged_;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr value_type& __get() & noexcept
+    {
+        return this->__val_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr const value_type& __get() const& noexcept
+    {
+        return this->__val_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr value_type&& __get() && noexcept
+    {
+        return _CUDA_VSTD::move(this->__val_);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr const value_type&& __get() const&& noexcept
+    {
+        return _CUDA_VSTD::move(this->__val_);
+    }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __construct(_Args&&... __args)
+    {
+        _LIBCUDACXX_ASSERT(!has_value(), "__construct called for engaged __optional_storage");
+        ::new((void*)_CUDA_VSTD::addressof(this->__val_)) value_type(_CUDA_VSTD::forward<_Args>(__args)...);
+        this->__engaged_ = true;
+    }
+
+    template <class _That>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __construct_from(_That&& __opt)
+    {
+        if (__opt.has_value())
+            __construct(_CUDA_VSTD::forward<_That>(__opt).__get());
+    }
+
+    template <class _That>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __assign_from(_That&& __opt)
+    {
+        if (this->__engaged_ == __opt.has_value())
+        {
+            if (this->__engaged_)
+                this->__val_ = _CUDA_VSTD::forward<_That>(__opt).__get();
+        }
+        else
+        {
+            if (this->__engaged_)
+                this->reset();
+            else
+                __construct(_CUDA_VSTD::forward<_That>(__opt).__get());
+        }
+    }
+};
+
+// optional<T&> is currently required ill-formed, however it may to be in the
+// future. For this reason it has already been implemented to ensure we can
+// make the change in an ABI compatible manner.
+template <class _Tp>
+struct __optional_storage_base<_Tp, true>
+{
+    using value_type = _Tp;
+    using __raw_type = remove_reference_t<_Tp>;
+    __raw_type* __value_;
+
+    template <class _Up>
+    static constexpr bool __can_bind_reference() {
+        using _RawUp = typename remove_reference<_Up>::type;
+        using _UpPtr = _RawUp*;
+        using _RawTp = typename remove_reference<_Tp>::type;
+        using _TpPtr = _RawTp*;
+        using _CheckLValueArg = integral_constant<bool,
+            (is_lvalue_reference<_Up>::value && is_convertible<_UpPtr, _TpPtr>::value)
+        ||  is_same<_RawUp, reference_wrapper<_RawTp>>::value
+        ||  is_same<_RawUp, reference_wrapper<typename remove_const<_RawTp>::type>>::value
+        >;
+        return (is_lvalue_reference<_Tp>::value && _CheckLValueArg::value)
+            || (is_rvalue_reference<_Tp>::value && !is_lvalue_reference<_Up>::value &&
+                is_convertible<_UpPtr, _TpPtr>::value);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr __optional_storage_base() noexcept
+        :  __value_(nullptr) {}
+
+    template <class _UArg>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit __optional_storage_base(in_place_t, _UArg&& __uarg)
+        :  __value_(_CUDA_VSTD::addressof(__uarg))
+    {
+      static_assert(__can_bind_reference<_UArg>(),
+        "Attempted to construct a reference element in tuple from a "
+        "possible temporary");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() noexcept { __value_ = nullptr; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr bool has_value() const noexcept
+      { return __value_ != nullptr; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr value_type& __get() const& noexcept
+      { return *__value_; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr value_type&& __get() const&& noexcept
+      { return _CUDA_VSTD::forward<value_type>(*__value_); }
+
+    template <class _UArg>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __construct(_UArg&& __val)
+    {
+        _LIBCUDACXX_ASSERT(!has_value(), "__construct called for engaged __optional_storage");
+        static_assert(__can_bind_reference<_UArg>(),
+            "Attempted to construct a reference element in tuple from a "
+            "possible temporary");
+        __value_ = _CUDA_VSTD::addressof(__val);
+    }
+
+    template <class _That>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __construct_from(_That&& __opt)
+    {
+        if (__opt.has_value())
+            __construct(_CUDA_VSTD::forward<_That>(__opt).__get());
+    }
+
+    template <class _That>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __assign_from(_That&& __opt)
+    {
+        if (has_value() == __opt.has_value())
+        {
+            if (has_value())
+                *__value_ = _CUDA_VSTD::forward<_That>(__opt).__get();
+        }
+        else
+        {
+            if (has_value())
+                reset();
+            else
+                __construct(_CUDA_VSTD::forward<_That>(__opt).__get());
+        }
+    }
+};
+
+template <class _Tp, bool = is_trivially_copy_constructible<_Tp>::value>
+struct __optional_copy_base : __optional_storage_base<_Tp>
+{
+    using __optional_storage_base<_Tp>::__optional_storage_base;
+};
+
+template <class _Tp>
+struct __optional_copy_base<_Tp, false> : __optional_storage_base<_Tp>
+{
+    using __optional_storage_base<_Tp>::__optional_storage_base;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_base() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_base(const __optional_copy_base& __opt)
+    {
+        this->__construct_from(__opt);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_base(__optional_copy_base&&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_base& operator=(const __optional_copy_base&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_base& operator=(__optional_copy_base&&) = default;
+};
+
+template <class _Tp, bool = is_trivially_move_constructible<_Tp>::value>
+struct __optional_move_base : __optional_copy_base<_Tp>
+{
+    using __optional_copy_base<_Tp>::__optional_copy_base;
+};
+
+template <class _Tp>
+struct __optional_move_base<_Tp, false> : __optional_copy_base<_Tp>
+{
+    using value_type = _Tp;
+    using __optional_copy_base<_Tp>::__optional_copy_base;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_base() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_base(const __optional_move_base&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_base(__optional_move_base&& __opt)
+        noexcept(is_nothrow_move_constructible_v<value_type>)
+    {
+        this->__construct_from(_CUDA_VSTD::move(__opt));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_base& operator=(const __optional_move_base&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_base& operator=(__optional_move_base&&) = default;
+};
+
+template <class _Tp, bool =
+    is_trivially_destructible<_Tp>::value &&
+    is_trivially_copy_constructible<_Tp>::value &&
+    is_trivially_copy_assignable<_Tp>::value>
+struct __optional_copy_assign_base : __optional_move_base<_Tp>
+{
+    using __optional_move_base<_Tp>::__optional_move_base;
+};
+
+template <class _Tp>
+struct __optional_copy_assign_base<_Tp, false> : __optional_move_base<_Tp>
+{
+    using __optional_move_base<_Tp>::__optional_move_base;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_assign_base() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_assign_base(const __optional_copy_assign_base&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_assign_base(__optional_copy_assign_base&&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_assign_base& operator=(const __optional_copy_assign_base& __opt)
+    {
+        this->__assign_from(__opt);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_copy_assign_base& operator=(__optional_copy_assign_base&&) = default;
+};
+
+template <class _Tp, bool =
+    is_trivially_destructible<_Tp>::value &&
+    is_trivially_move_constructible<_Tp>::value &&
+    is_trivially_move_assignable<_Tp>::value>
+struct __optional_move_assign_base : __optional_copy_assign_base<_Tp>
+{
+    using __optional_copy_assign_base<_Tp>::__optional_copy_assign_base;
+};
+
+template <class _Tp>
+struct __optional_move_assign_base<_Tp, false> : __optional_copy_assign_base<_Tp>
+{
+    using value_type = _Tp;
+    using __optional_copy_assign_base<_Tp>::__optional_copy_assign_base;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_assign_base() = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_assign_base(const __optional_move_assign_base& __opt) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_assign_base(__optional_move_assign_base&&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_assign_base& operator=(const __optional_move_assign_base&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __optional_move_assign_base& operator=(__optional_move_assign_base&& __opt)
+        noexcept(is_nothrow_move_assignable_v<value_type> &&
+                 is_nothrow_move_constructible_v<value_type>)
+    {
+        this->__assign_from(_CUDA_VSTD::move(__opt));
+        return *this;
+    }
+};
+
+template <class _Tp>
+using __optional_sfinae_ctor_base_t = __sfinae_ctor_base<
+    is_copy_constructible<_Tp>::value,
+    is_move_constructible<_Tp>::value
+>;
+
+template <class _Tp>
+using __optional_sfinae_assign_base_t = __sfinae_assign_base<
+    (is_copy_constructible<_Tp>::value && is_copy_assignable<_Tp>::value),
+    (is_move_constructible<_Tp>::value && is_move_assignable<_Tp>::value)
+>;
+
+template <class _Tp>
+class optional
+    : private __optional_move_assign_base<_Tp>
+    , private __optional_sfinae_ctor_base_t<_Tp>
+    , private __optional_sfinae_assign_base_t<_Tp>
+{
+    using __base = __optional_move_assign_base<_Tp>;
+public:
+    using value_type = _Tp;
+
+private:
+     // Disable the reference extension using this static assert.
+    static_assert(!is_same_v<__uncvref_t<value_type>, in_place_t>,
+        "instantiation of optional with in_place_t is ill-formed");
+    static_assert(!is_same_v<__uncvref_t<value_type>, nullopt_t>,
+        "instantiation of optional with nullopt_t is ill-formed");
+    static_assert(!is_reference_v<value_type>,
+        "instantiation of optional with a reference type is ill-formed");
+    static_assert(is_destructible_v<value_type>,
+        "instantiation of optional with a non-destructible type is ill-formed");
+    static_assert(!is_array_v<value_type>,
+        "instantiation of optional with an array type is ill-formed");
+
+    // LWG2756: conditionally explicit conversion from _Up
+    struct _CheckOptionalArgsConstructor {
+      template <class _Up>
+      static constexpr bool __enable_implicit() {
+          return is_constructible_v<_Tp, _Up&&> &&
+                 is_convertible_v<_Up&&, _Tp>;
+      }
+
+      template <class _Up>
+      static constexpr bool __enable_explicit() {
+          return is_constructible_v<_Tp, _Up&&> &&
+                 !is_convertible_v<_Up&&, _Tp>;
+      }
+    };
+    template <class _Up>
+    using _CheckOptionalArgsCtor = _If<
+        _IsNotSame<__uncvref_t<_Up>, in_place_t>::value &&
+        _IsNotSame<__uncvref_t<_Up>, optional>::value,
+        _CheckOptionalArgsConstructor,
+        __check_tuple_constructor_fail
+    >;
+    template <class _QualUp>
+    struct _CheckOptionalLikeConstructor {
+      template <class _Up, class _Opt = optional<_Up>>
+      using __check_constructible_from_opt = _Or<
+          is_constructible<_Tp, _Opt&>,
+          is_constructible<_Tp, _Opt const&>,
+          is_constructible<_Tp, _Opt&&>,
+          is_constructible<_Tp, _Opt const&&>,
+          is_convertible<_Opt&, _Tp>,
+          is_convertible<_Opt const&, _Tp>,
+          is_convertible<_Opt&&, _Tp>,
+          is_convertible<_Opt const&&, _Tp>
+      >;
+      template <class _Up, class _Opt = optional<_Up>>
+      using __check_assignable_from_opt = _Or<
+          is_assignable<_Tp&, _Opt&>,
+          is_assignable<_Tp&, _Opt const&>,
+          is_assignable<_Tp&, _Opt&&>,
+          is_assignable<_Tp&, _Opt const&&>
+      >;
+      template <class _Up, class _QUp = _QualUp>
+      static constexpr bool __enable_implicit() {
+          return is_convertible<_QUp, _Tp>::value &&
+              !__check_constructible_from_opt<_Up>::value;
+      }
+      template <class _Up, class _QUp = _QualUp>
+      static constexpr bool __enable_explicit() {
+          return !is_convertible<_QUp, _Tp>::value &&
+              !__check_constructible_from_opt<_Up>::value;
+      }
+      template <class _Up, class _QUp = _QualUp>
+      static constexpr bool __enable_assign() {
+          // Construction and assignability of _Qup to _Tp has already been
+          // checked.
+          return !__check_constructible_from_opt<_Up>::value &&
+              !__check_assignable_from_opt<_Up>::value;
+      }
+    };
+
+    template <class _Up, class _QualUp>
+    using _CheckOptionalLikeCtor = _If<
+      _And<
+         _IsNotSame<_Up, _Tp>,
+          is_constructible<_Tp, _QualUp>
+      >::value,
+      _CheckOptionalLikeConstructor<_QualUp>,
+      __check_tuple_constructor_fail
+    >;
+    template <class _Up, class _QualUp>
+    using _CheckOptionalLikeAssign = _If<
+      _And<
+          _IsNotSame<_Up, _Tp>,
+          is_constructible<_Tp, _QualUp>,
+          is_assignable<_Tp&, _QualUp>
+      >::value,
+      _CheckOptionalLikeConstructor<_QualUp>,
+      __check_tuple_constructor_fail
+    >;
+public:
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr optional() noexcept {}
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr optional(const optional&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr optional(optional&&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr optional(nullopt_t) noexcept {}
+
+    template <class _InPlaceT, class... _Args, class = _EnableIf<
+          _And<
+              _IsSame<_InPlaceT, in_place_t>,
+              is_constructible<value_type, _Args...>
+            >::value
+        >
+    >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit optional(_InPlaceT, _Args&&... __args)
+        : __base(in_place, _CUDA_VSTD::forward<_Args>(__args)...) {}
+
+    template <class _Up, class... _Args, class = _EnableIf<
+        is_constructible_v<value_type, initializer_list<_Up>&, _Args...>>
+    >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit optional(in_place_t, initializer_list<_Up> __il, _Args&&... __args)
+        : __base(in_place, __il, _CUDA_VSTD::forward<_Args>(__args)...) {}
+
+    template <class _Up = value_type, _EnableIf<
+        _CheckOptionalArgsCtor<_Up>::template __enable_implicit<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr optional(_Up&& __v)
+        : __base(in_place, _CUDA_VSTD::forward<_Up>(__v)) {}
+
+    template <class _Up, _EnableIf<
+        _CheckOptionalArgsCtor<_Up>::template __enable_explicit<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit optional(_Up&& __v)
+        : __base(in_place, _CUDA_VSTD::forward<_Up>(__v)) {}
+
+    // LWG2756: conditionally explicit conversion from const optional<_Up>&
+    template <class _Up, _EnableIf<
+        _CheckOptionalLikeCtor<_Up, _Up const&>::template __enable_implicit<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    optional(const optional<_Up>& __v)
+    {
+        this->__construct_from(__v);
+    }
+    template <class _Up, _EnableIf<
+        _CheckOptionalLikeCtor<_Up, _Up const&>::template __enable_explicit<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit optional(const optional<_Up>& __v)
+    {
+        this->__construct_from(__v);
+    }
+
+    // LWG2756: conditionally explicit conversion from optional<_Up>&&
+    template <class _Up, _EnableIf<
+        _CheckOptionalLikeCtor<_Up, _Up &&>::template __enable_implicit<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    optional(optional<_Up>&& __v)
+    {
+        this->__construct_from(_CUDA_VSTD::move(__v));
+    }
+    template <class _Up, _EnableIf<
+        _CheckOptionalLikeCtor<_Up, _Up &&>::template __enable_explicit<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit optional(optional<_Up>&& __v)
+    {
+        this->__construct_from(_CUDA_VSTD::move(__v));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    optional& operator=(nullopt_t) noexcept
+    {
+        reset();
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY optional& operator=(const optional&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY optional& operator=(optional&&) = default;
+
+    // LWG2756
+    template <class _Up = value_type,
+              class = _EnableIf<
+                      _And<
+                          _IsNotSame<__uncvref_t<_Up>, optional>,
+                          _Or<
+                              _IsNotSame<__uncvref_t<_Up>, value_type>,
+                              _Not<is_scalar<value_type>>
+                          >,
+                          is_constructible<value_type, _Up>,
+                          is_assignable<value_type&, _Up>
+                      >::value>
+             >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    optional&
+    operator=(_Up&& __v)
+    {
+        if (this->has_value())
+            this->__get() = _CUDA_VSTD::forward<_Up>(__v);
+        else
+            this->__construct(_CUDA_VSTD::forward<_Up>(__v));
+        return *this;
+    }
+
+    // LWG2756
+    template <class _Up, _EnableIf<
+        _CheckOptionalLikeAssign<_Up, _Up const&>::template __enable_assign<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    optional&
+    operator=(const optional<_Up>& __v)
+    {
+        this->__assign_from(__v);
+        return *this;
+    }
+
+    // LWG2756
+    template <class _Up, _EnableIf<
+        _CheckOptionalLikeCtor<_Up, _Up &&>::template __enable_assign<_Up>()
+    , int> = 0>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    optional&
+    operator=(optional<_Up>&& __v)
+    {
+        this->__assign_from(_CUDA_VSTD::move(__v));
+        return *this;
+    }
+
+    template <class... _Args,
+              class = _EnableIf
+                      <
+                          is_constructible_v<value_type, _Args...>
+                      >
+             >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp &
+    emplace(_Args&&... __args)
+    {
+        reset();
+        this->__construct(_CUDA_VSTD::forward<_Args>(__args)...);
+        return this->__get();
+    }
+
+    template <class _Up, class... _Args,
+              class = _EnableIf
+                      <
+                          is_constructible_v<value_type, initializer_list<_Up>&, _Args...>
+                      >
+             >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp &
+    emplace(initializer_list<_Up> __il, _Args&&... __args)
+    {
+        reset();
+        this->__construct(__il, _CUDA_VSTD::forward<_Args>(__args)...);
+        return this->__get();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(optional& __opt)
+        noexcept(is_nothrow_move_constructible_v<value_type> &&
+                 is_nothrow_swappable_v<value_type>)
+    {
+        if (this->has_value() == __opt.has_value())
+        {
+            using _CUDA_VSTD::swap;
+            if (this->has_value())
+                swap(this->__get(), __opt.__get());
+        }
+        else
+        {
+            if (this->has_value())
+            {
+                __opt.__construct(_CUDA_VSTD::move(this->__get()));
+                reset();
+            }
+            else
+            {
+                this->__construct(_CUDA_VSTD::move(__opt.__get()));
+                __opt.reset();
+            }
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    add_pointer_t<value_type const>
+    operator->() const
+    {
+        _LIBCUDACXX_ASSERT(this->has_value(), "optional operator-> called for disengaged value");
+#if defined(_LIBCUDACXX_ADDRESSOF)
+        return _CUDA_VSTD::addressof(this->__get());
+#else
+        return __operator_arrow(__has_operator_addressof<value_type>{}, this->__get());
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    add_pointer_t<value_type>
+    operator->()
+    {
+        _LIBCUDACXX_ASSERT(this->has_value(), "optional operator-> called for disengaged value");
+#if defined(_LIBCUDACXX_ADDRESSOF)
+        return _CUDA_VSTD::addressof(this->__get());
+#else
+        return __operator_arrow(__has_operator_addressof<value_type>{}, this->__get());
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    const value_type&
+    operator*() const&
+    {
+        _LIBCUDACXX_ASSERT(this->has_value(), "optional operator* called for disengaged value");
+        return this->__get();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    value_type&
+    operator*() &
+    {
+        _LIBCUDACXX_ASSERT(this->has_value(), "optional operator* called for disengaged value");
+        return this->__get();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    value_type&&
+    operator*() &&
+    {
+        _LIBCUDACXX_ASSERT(this->has_value(), "optional operator* called for disengaged value");
+        return _CUDA_VSTD::move(this->__get());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    const value_type&&
+    operator*() const&&
+    {
+        _LIBCUDACXX_ASSERT(this->has_value(), "optional operator* called for disengaged value");
+        return _CUDA_VSTD::move(this->__get());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr explicit operator bool() const noexcept { return has_value(); }
+
+    using __base::has_value;
+    using __base::__get;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS
+    constexpr value_type const& value() const&
+    {
+        if (!this->has_value())
+            __throw_bad_optional_access();
+        return this->__get();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS
+    constexpr value_type& value() &
+    {
+        if (!this->has_value())
+            __throw_bad_optional_access();
+        return this->__get();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS
+    constexpr value_type&& value() &&
+    {
+        if (!this->has_value())
+            __throw_bad_optional_access();
+        return _CUDA_VSTD::move(this->__get());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_AVAILABILITY_THROW_BAD_OPTIONAL_ACCESS
+    constexpr value_type const&& value() const&&
+    {
+        if (!this->has_value())
+            __throw_bad_optional_access();
+        return _CUDA_VSTD::move(this->__get());
+    }
+
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr value_type value_or(_Up&& __v) const&
+    {
+        static_assert(is_copy_constructible_v<value_type>,
+                      "optional<T>::value_or: T must be copy constructible");
+        static_assert(is_convertible_v<_Up, value_type>,
+                      "optional<T>::value_or: U must be convertible to T");
+        return this->has_value() ? this->__get() :
+                                  static_cast<value_type>(_CUDA_VSTD::forward<_Up>(__v));
+    }
+
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr value_type value_or(_Up&& __v) &&
+    {
+        static_assert(is_move_constructible_v<value_type>,
+                      "optional<T>::value_or: T must be move constructible");
+        static_assert(is_convertible_v<_Up, value_type>,
+                      "optional<T>::value_or: U must be convertible to T");
+        return this->has_value() ? _CUDA_VSTD::move(this->__get()) :
+                                  static_cast<value_type>(_CUDA_VSTD::forward<_Up>(__v));
+    }
+
+    using __base::reset;
+
+private:
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _Up*
+    __operator_arrow(true_type, _Up& __x)
+    {
+        return _CUDA_VSTD::addressof(__x);
+    }
+
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr _Up*
+    __operator_arrow(false_type, _Up& __x)
+    {
+        return &__x;
+    }
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class T>
+    optional(T) -> optional<T>;
+#endif
+
+// Comparisons between optionals
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() ==
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator==(const optional<_Tp>& __x, const optional<_Up>& __y)
+{
+    if (static_cast<bool>(__x) != static_cast<bool>(__y))
+        return false;
+    if (!static_cast<bool>(__x))
+        return true;
+    return *__x == *__y;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() !=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator!=(const optional<_Tp>& __x, const optional<_Up>& __y)
+{
+    if (static_cast<bool>(__x) != static_cast<bool>(__y))
+        return true;
+    if (!static_cast<bool>(__x))
+        return false;
+    return *__x != *__y;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() <
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator<(const optional<_Tp>& __x, const optional<_Up>& __y)
+{
+    if (!static_cast<bool>(__y))
+        return false;
+    if (!static_cast<bool>(__x))
+        return true;
+    return *__x < *__y;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() >
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator>(const optional<_Tp>& __x, const optional<_Up>& __y)
+{
+    if (!static_cast<bool>(__x))
+        return false;
+    if (!static_cast<bool>(__y))
+        return true;
+    return *__x > *__y;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() <=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator<=(const optional<_Tp>& __x, const optional<_Up>& __y)
+{
+    if (!static_cast<bool>(__x))
+        return true;
+    if (!static_cast<bool>(__y))
+        return false;
+    return *__x <= *__y;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() >=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator>=(const optional<_Tp>& __x, const optional<_Up>& __y)
+{
+    if (!static_cast<bool>(__y))
+        return true;
+    if (!static_cast<bool>(__x))
+        return false;
+    return *__x >= *__y;
+}
+
+// Comparisons with nullopt
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator==(const optional<_Tp>& __x, nullopt_t) noexcept
+{
+    return !static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator==(nullopt_t, const optional<_Tp>& __x) noexcept
+{
+    return !static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator!=(const optional<_Tp>& __x, nullopt_t) noexcept
+{
+    return static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator!=(nullopt_t, const optional<_Tp>& __x) noexcept
+{
+    return static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator<(const optional<_Tp>&, nullopt_t) noexcept
+{
+    return false;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator<(nullopt_t, const optional<_Tp>& __x) noexcept
+{
+    return static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator<=(const optional<_Tp>& __x, nullopt_t) noexcept
+{
+    return !static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator<=(nullopt_t, const optional<_Tp>&) noexcept
+{
+    return true;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator>(const optional<_Tp>& __x, nullopt_t) noexcept
+{
+    return static_cast<bool>(__x);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator>(nullopt_t, const optional<_Tp>&) noexcept
+{
+    return false;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator>=(const optional<_Tp>&, nullopt_t) noexcept
+{
+    return true;
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+bool
+operator>=(nullopt_t, const optional<_Tp>& __x) noexcept
+{
+    return !static_cast<bool>(__x);
+}
+
+// Comparisons with T
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() ==
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator==(const optional<_Tp>& __x, const _Up& __v)
+{
+    return static_cast<bool>(__x) ? *__x == __v : false;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() ==
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator==(const _Tp& __v, const optional<_Up>& __x)
+{
+    return static_cast<bool>(__x) ? __v == *__x : false;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() !=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator!=(const optional<_Tp>& __x, const _Up& __v)
+{
+    return static_cast<bool>(__x) ? *__x != __v : true;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() !=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator!=(const _Tp& __v, const optional<_Up>& __x)
+{
+    return static_cast<bool>(__x) ? __v != *__x : true;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() <
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator<(const optional<_Tp>& __x, const _Up& __v)
+{
+    return static_cast<bool>(__x) ? *__x < __v : true;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() <
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator<(const _Tp& __v, const optional<_Up>& __x)
+{
+    return static_cast<bool>(__x) ? __v < *__x : false;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() <=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator<=(const optional<_Tp>& __x, const _Up& __v)
+{
+    return static_cast<bool>(__x) ? *__x <= __v : true;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() <=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator<=(const _Tp& __v, const optional<_Up>& __x)
+{
+    return static_cast<bool>(__x) ? __v <= *__x : false;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() >
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator>(const optional<_Tp>& __x, const _Up& __v)
+{
+    return static_cast<bool>(__x) ? *__x > __v : false;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() >
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator>(const _Tp& __v, const optional<_Up>& __x)
+{
+    return static_cast<bool>(__x) ? __v > *__x : true;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() >=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator>=(const optional<_Tp>& __x, const _Up& __v)
+{
+    return static_cast<bool>(__x) ? *__x >= __v : false;
+}
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_EnableIf<
+    is_convertible_v<decltype(_CUDA_VSTD::declval<const _Tp&>() >=
+        _CUDA_VSTD::declval<const _Up&>()), bool>,
+    bool
+>
+operator>=(const _Tp& __v, const optional<_Up>& __x)
+{
+    return static_cast<bool>(__x) ? __v >= *__x : true;
+}
+
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_EnableIf<
+    is_move_constructible_v<_Tp> && is_swappable_v<_Tp>,
+    void
+>
+swap(optional<_Tp>& __x, optional<_Tp>& __y) noexcept(noexcept(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+optional<decay_t<_Tp>> make_optional(_Tp&& __v)
+{
+    return optional<decay_t<_Tp>>(_CUDA_VSTD::forward<_Tp>(__v));
+}
+
+template <class _Tp, class... _Args>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+optional<_Tp> make_optional(_Args&&... __args)
+{
+    return optional<_Tp>(in_place, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _Tp, class _Up, class... _Args>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+optional<_Tp> make_optional(initializer_list<_Up> __il,  _Args&&... __args)
+{
+    return optional<_Tp>(in_place, __il, _CUDA_VSTD::forward<_Args>(__args)...);
+}
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<
+    __enable_hash_helper<optional<_Tp>, remove_const_t<_Tp>>
+>
+{
+    typedef optional<_Tp> argument_type;
+    typedef size_t        result_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator()(const argument_type& __opt) const
+    {
+        return static_cast<bool>(__opt) ? hash<remove_const_t<_Tp>>()(*__opt) : 0;
+    }
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_OPTIONAL
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ostream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ostream
new file mode 100644
index 000000000000..3111c10ceb42
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ostream
@@ -0,0 +1,1107 @@
+// -*- C++ -*-
+//===-------------------------- ostream -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_OSTREAM
+#define _LIBCUDACXX_OSTREAM
+
+/*
+    ostream synopsis
+
+template <class charT, class traits = char_traits<charT> >
+class basic_ostream
+    : virtual public basic_ios<charT,traits>
+{
+public:
+    // types (inherited from basic_ios (27.5.4)):
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // 27.7.2.2 Constructor/destructor:
+    explicit basic_ostream(basic_streambuf<char_type,traits>* sb);
+    basic_ostream(basic_ostream&& rhs);
+    virtual ~basic_ostream();
+
+    // 27.7.2.3 Assign/swap
+    basic_ostream& operator=(const basic_ostream& rhs) = delete; // C++14
+    basic_ostream& operator=(basic_ostream&& rhs);
+    void swap(basic_ostream& rhs);
+
+    // 27.7.2.4 Prefix/suffix:
+    class sentry;
+
+    // 27.7.2.6 Formatted output:
+    basic_ostream& operator<<(basic_ostream& (*pf)(basic_ostream&));
+    basic_ostream& operator<<(basic_ios<charT, traits>& (*pf)(basic_ios<charT,traits>&));
+    basic_ostream& operator<<(ios_base& (*pf)(ios_base&));
+    basic_ostream& operator<<(bool n);
+    basic_ostream& operator<<(short n);
+    basic_ostream& operator<<(unsigned short n);
+    basic_ostream& operator<<(int n);
+    basic_ostream& operator<<(unsigned int n);
+    basic_ostream& operator<<(long n);
+    basic_ostream& operator<<(unsigned long n);
+    basic_ostream& operator<<(long long n);
+    basic_ostream& operator<<(unsigned long long n);
+    basic_ostream& operator<<(float f);
+    basic_ostream& operator<<(double f);
+    basic_ostream& operator<<(long double f);
+    basic_ostream& operator<<(const void* p);
+    basic_ostream& operator<<(basic_streambuf<char_type,traits>* sb);
+    basic_ostream& operator<<(nullptr_t);
+
+    // 27.7.2.7 Unformatted output:
+    basic_ostream& put(char_type c);
+    basic_ostream& write(const char_type* s, streamsize n);
+    basic_ostream& flush();
+
+    // 27.7.2.5 seeks:
+    pos_type tellp();
+    basic_ostream& seekp(pos_type);
+    basic_ostream& seekp(off_type, ios_base::seekdir);
+protected:
+    basic_ostream(const basic_ostream& rhs) = delete;
+    basic_ostream(basic_ostream&& rhs);
+    // 27.7.3.3 Assign/swap
+    basic_ostream& operator=(basic_ostream& rhs) = delete;
+    basic_ostream& operator=(const basic_ostream&& rhs);
+    void swap(basic_ostream& rhs);
+};
+
+// 27.7.2.6.4 character inserters
+
+template<class charT, class traits>
+  basic_ostream<charT,traits>& operator<<(basic_ostream<charT,traits>&, charT);
+
+template<class charT, class traits>
+  basic_ostream<charT,traits>& operator<<(basic_ostream<charT,traits>&, char);
+
+template<class traits>
+  basic_ostream<char,traits>& operator<<(basic_ostream<char,traits>&, char);
+
+// signed and unsigned
+
+template<class traits>
+  basic_ostream<char,traits>& operator<<(basic_ostream<char,traits>&, signed char);
+
+template<class traits>
+  basic_ostream<char,traits>& operator<<(basic_ostream<char,traits>&, unsigned char);
+
+// NTBS
+template<class charT, class traits>
+  basic_ostream<charT,traits>& operator<<(basic_ostream<charT,traits>&, const charT*);
+
+template<class charT, class traits>
+  basic_ostream<charT,traits>& operator<<(basic_ostream<charT,traits>&, const char*);
+
+template<class traits>
+  basic_ostream<char,traits>& operator<<(basic_ostream<char,traits>&, const char*);
+
+// signed and unsigned
+template<class traits>
+basic_ostream<char,traits>& operator<<(basic_ostream<char,traits>&, const signed char*);
+
+template<class traits>
+  basic_ostream<char,traits>& operator<<(basic_ostream<char,traits>&, const unsigned char*);
+
+// swap:
+template <class charT, class traits>
+  void swap(basic_ostream<charT, traits>& x, basic_ostream<charT, traits>& y);
+
+template <class charT, class traits>
+  basic_ostream<charT,traits>& endl(basic_ostream<charT,traits>& os);
+
+template <class charT, class traits>
+  basic_ostream<charT,traits>& ends(basic_ostream<charT,traits>& os);
+
+template <class charT, class traits>
+  basic_ostream<charT,traits>& flush(basic_ostream<charT,traits>& os);
+
+// rvalue stream insertion
+template <class charT, class traits, class T>
+  basic_ostream<charT, traits>&
+  operator<<(basic_ostream<charT, traits>&& os, const T& x);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <ios>
+#include <streambuf>
+#include <locale>
+#include <iterator>
+#include <bitset>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_ostream
+    : virtual public basic_ios<_CharT, _Traits>
+{
+public:
+    // types (inherited from basic_ios (27.5.4)):
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    // 27.7.2.2 Constructor/destructor:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    explicit basic_ostream(basic_streambuf<char_type, traits_type>* __sb)
+    { this->init(__sb); }
+    virtual ~basic_ostream();
+protected:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream(basic_ostream&& __rhs);
+
+    // 27.7.2.3 Assign/swap
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream& operator=(basic_ostream&& __rhs);
+#endif
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void swap(basic_ostream& __rhs)
+    { basic_ios<char_type, traits_type>::swap(__rhs); }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    basic_ostream           (const basic_ostream& __rhs) = delete;
+    basic_ostream& operator=(const basic_ostream& __rhs) = delete;
+#else
+    basic_ostream           (const basic_ostream& __rhs); // not defined
+    basic_ostream& operator=(const basic_ostream& __rhs); // not defined
+#endif
+public:
+
+    // 27.7.2.4 Prefix/suffix:
+    class _LIBCUDACXX_TEMPLATE_VIS sentry;
+
+    // 27.7.2.6 Formatted output:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_ostream& operator<<(basic_ostream& (*__pf)(basic_ostream&))
+    { return __pf(*this); }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_ostream& operator<<(basic_ios<char_type, traits_type>&
+                              (*__pf)(basic_ios<char_type,traits_type>&))
+    { __pf(*this); return *this; }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_ostream& operator<<(ios_base& (*__pf)(ios_base&))
+    { __pf(*this); return *this; }
+
+    basic_ostream& operator<<(bool __n);
+    basic_ostream& operator<<(short __n);
+    basic_ostream& operator<<(unsigned short __n);
+    basic_ostream& operator<<(int __n);
+    basic_ostream& operator<<(unsigned int __n);
+    basic_ostream& operator<<(long __n);
+    basic_ostream& operator<<(unsigned long __n);
+    basic_ostream& operator<<(long long __n);
+    basic_ostream& operator<<(unsigned long long __n);
+    basic_ostream& operator<<(float __f);
+    basic_ostream& operator<<(double __f);
+    basic_ostream& operator<<(long double __f);
+    basic_ostream& operator<<(const void* __p);
+    basic_ostream& operator<<(basic_streambuf<char_type, traits_type>* __sb);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream& operator<<(nullptr_t)
+    { return *this << "nullptr"; }
+
+    // 27.7.2.7 Unformatted output:
+    basic_ostream& put(char_type __c);
+    basic_ostream& write(const char_type* __s, streamsize __n);
+    basic_ostream& flush();
+
+    // 27.7.2.5 seeks:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    pos_type tellp();
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_ostream& seekp(pos_type __pos);
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_ostream& seekp(off_type __off, ios_base::seekdir __dir);
+
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostream() {}  // extension, intentially does not initialize
+};
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_ostream<_CharT, _Traits>::sentry
+{
+    bool __ok_;
+    basic_ostream<_CharT, _Traits>& __os_;
+
+    sentry(const sentry&); // = delete;
+    sentry& operator=(const sentry&); // = delete;
+
+public:
+    explicit sentry(basic_ostream<_CharT, _Traits>& __os);
+    ~sentry();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+        _LIBCUDACXX_EXPLICIT
+        operator bool() const {return __ok_;}
+};
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>::sentry::sentry(basic_ostream<_CharT, _Traits>& __os)
+    : __ok_(false),
+      __os_(__os)
+{
+    if (__os.good())
+    {
+        if (__os.tie())
+            __os.tie()->flush();
+        __ok_ = true;
+    }
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>::sentry::~sentry()
+{
+    if (__os_.rdbuf() && __os_.good() && (__os_.flags() & ios_base::unitbuf)
+                      && !uncaught_exception())
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            if (__os_.rdbuf()->pubsync() == -1)
+                __os_.setstate(ios_base::badbit);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>::basic_ostream(basic_ostream&& __rhs)
+{
+    this->move(__rhs);
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator=(basic_ostream&& __rhs)
+{
+    swap(__rhs);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>::~basic_ostream()
+{
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(basic_streambuf<char_type, traits_type>* __sb)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            if (__sb)
+            {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+                try
+                {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+                    typedef istreambuf_iterator<_CharT, _Traits> _Ip;
+                    typedef ostreambuf_iterator<_CharT, _Traits> _Op;
+                    _Ip __i(__sb);
+                    _Ip __eof;
+                    _Op __o(*this);
+                    size_t __c = 0;
+                    for (; __i != __eof; ++__i, ++__o, ++__c)
+                    {
+                        *__o = *__i;
+                        if (__o.failed())
+                            break;
+                    }
+                    if (__c == 0)
+                        this->setstate(ios_base::failbit);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+                }
+                catch (...)
+                {
+                    this->__set_failbit_and_consider_rethrow();
+                }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            }
+            else
+                this->setstate(ios_base::badbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(bool __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(short __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            ios_base::fmtflags __flags = ios_base::flags() & ios_base::basefield;
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(),
+                        __flags == ios_base::oct || __flags == ios_base::hex ?
+                        static_cast<long>(static_cast<unsigned short>(__n))  :
+                        static_cast<long>(__n)).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(unsigned short __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), static_cast<unsigned long>(__n)).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(int __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            ios_base::fmtflags __flags = ios_base::flags() & ios_base::basefield;
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(),
+                        __flags == ios_base::oct || __flags == ios_base::hex ?
+                        static_cast<long>(static_cast<unsigned int>(__n))  :
+                        static_cast<long>(__n)).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(unsigned int __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), static_cast<unsigned long>(__n)).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(long __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(unsigned long __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(long long __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(unsigned long long __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(float __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), static_cast<double>(__n)).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(double __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(long double __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::operator<<(const void* __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef num_put<char_type, ostreambuf_iterator<char_type, traits_type> > _Fp;
+            const _Fp& __f = use_facet<_Fp>(this->getloc());
+            if (__f.put(*this, *this, this->fill(), __n).failed())
+                this->setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+__put_character_sequence(basic_ostream<_CharT, _Traits>& __os,
+                          const _CharT* __str, size_t __len)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_ostream<_CharT, _Traits>::sentry __s(__os);
+        if (__s)
+        {
+            typedef ostreambuf_iterator<_CharT, _Traits> _Ip;
+            if (__pad_and_output(_Ip(__os),
+                                 __str,
+                                 (__os.flags() & ios_base::adjustfield) == ios_base::left ?
+                                     __str + __len :
+                                     __str,
+                                 __str + __len,
+                                 __os,
+                                 __os.fill()).failed())
+                __os.setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __os.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __os;
+}
+
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, _CharT __c)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, &__c, 1);
+}
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, char __cn)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_ostream<_CharT, _Traits>::sentry __s(__os);
+        if (__s)
+        {
+            _CharT __c = __os.widen(__cn);
+            typedef ostreambuf_iterator<_CharT, _Traits> _Ip;
+            if (__pad_and_output(_Ip(__os),
+                                 &__c,
+                                 (__os.flags() & ios_base::adjustfield) == ios_base::left ?
+                                     &__c + 1 :
+                                     &__c,
+                                 &__c + 1,
+                                 __os,
+                                 __os.fill()).failed())
+                __os.setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __os.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __os;
+}
+
+template<class _Traits>
+basic_ostream<char, _Traits>&
+operator<<(basic_ostream<char, _Traits>& __os, char __c)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, &__c, 1);
+}
+
+template<class _Traits>
+basic_ostream<char, _Traits>&
+operator<<(basic_ostream<char, _Traits>& __os, signed char __c)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, (char *) &__c, 1);
+}
+
+template<class _Traits>
+basic_ostream<char, _Traits>&
+operator<<(basic_ostream<char, _Traits>& __os, unsigned char __c)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, (char *) &__c, 1);
+}
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const _CharT* __str)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, __str, _Traits::length(__str));
+}
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const char* __strn)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        typename basic_ostream<_CharT, _Traits>::sentry __s(__os);
+        if (__s)
+        {
+            typedef ostreambuf_iterator<_CharT, _Traits> _Ip;
+            size_t __len = char_traits<char>::length(__strn);
+            const int __bs = 100;
+            _CharT __wbb[__bs];
+            _CharT* __wb = __wbb;
+            unique_ptr<_CharT, void(*)(void*)> __h(0, free);
+            if (__len > __bs)
+            {
+                __wb = (_CharT*)malloc(__len*sizeof(_CharT));
+                if (__wb == 0)
+                    __throw_bad_alloc();
+                __h.reset(__wb);
+            }
+            for (_CharT* __p = __wb; *__strn != '\0'; ++__strn, ++__p)
+                *__p = __os.widen(*__strn);
+            if (__pad_and_output(_Ip(__os),
+                                 __wb,
+                                 (__os.flags() & ios_base::adjustfield) == ios_base::left ?
+                                     __wb + __len :
+                                     __wb,
+                                 __wb + __len,
+                                 __os,
+                                 __os.fill()).failed())
+                __os.setstate(ios_base::badbit | ios_base::failbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        __os.__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return __os;
+}
+
+template<class _Traits>
+basic_ostream<char, _Traits>&
+operator<<(basic_ostream<char, _Traits>& __os, const char* __str)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, __str, _Traits::length(__str));
+}
+
+template<class _Traits>
+basic_ostream<char, _Traits>&
+operator<<(basic_ostream<char, _Traits>& __os, const signed char* __str)
+{
+    const char *__s = (const char *) __str;
+    return _CUDA_VSTD::__put_character_sequence(__os, __s, _Traits::length(__s));
+}
+
+template<class _Traits>
+basic_ostream<char, _Traits>&
+operator<<(basic_ostream<char, _Traits>& __os, const unsigned char* __str)
+{
+    const char *__s = (const char *) __str;
+    return _CUDA_VSTD::__put_character_sequence(__os, __s, _Traits::length(__s));
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::put(char_type __c)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __s(*this);
+        if (__s)
+        {
+            typedef ostreambuf_iterator<_CharT, _Traits> _Op;
+            _Op __o(*this);
+            *__o = __c;
+            if (__o.failed())
+                this->setstate(ios_base::badbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::write(const char_type* __s, streamsize __n)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        sentry __sen(*this);
+        if (__sen && __n)
+        {
+            if (this->rdbuf()->sputn(__s, __n) != __n)
+                this->setstate(ios_base::badbit);
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::flush()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        if (this->rdbuf())
+        {
+            sentry __s(*this);
+            if (__s)
+            {
+                if (this->rdbuf()->pubsync() == -1)
+                    this->setstate(ios_base::badbit);
+            }
+        }
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        this->__set_badbit_and_consider_rethrow();
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+typename basic_ostream<_CharT, _Traits>::pos_type
+basic_ostream<_CharT, _Traits>::tellp()
+{
+    if (this->fail())
+        return pos_type(-1);
+    return this->rdbuf()->pubseekoff(0, ios_base::cur, ios_base::out);
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::seekp(pos_type __pos)
+{
+    sentry __s(*this);
+    if (!this->fail())
+    {
+        if (this->rdbuf()->pubseekpos(__pos, ios_base::out) == pos_type(-1))
+            this->setstate(ios_base::failbit);
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+basic_ostream<_CharT, _Traits>::seekp(off_type __off, ios_base::seekdir __dir)
+{
+    sentry __s(*this);
+    if (!this->fail())
+    {
+        if (this->rdbuf()->pubseekoff(__off, __dir, ios_base::out) == pos_type(-1))
+            this->setstate(ios_base::failbit);
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+endl(basic_ostream<_CharT, _Traits>& __os)
+{
+    __os.put(__os.widen('\n'));
+    __os.flush();
+    return __os;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+ends(basic_ostream<_CharT, _Traits>& __os)
+{
+    __os.put(_CharT());
+    return __os;
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+flush(basic_ostream<_CharT, _Traits>& __os)
+{
+    __os.flush();
+    return __os;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Stream, class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    !is_lvalue_reference<_Stream>::value &&
+    is_base_of<ios_base, _Stream>::value,
+    _Stream&&
+>::type
+operator<<(_Stream&& __os, const _Tp& __x)
+{
+    __os << __x;
+    return _CUDA_VSTD::move(__os);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const basic_string<_CharT, _Traits, _Allocator>& __str)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, __str.data(), __str.size());
+}
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           basic_string_view<_CharT, _Traits> __sv)
+{
+    return _CUDA_VSTD::__put_character_sequence(__os, __sv.data(), __sv.size());
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const error_code& __ec)
+{
+    return __os << __ec.category().name() << ':' << __ec.value();
+}
+
+template<class _CharT, class _Traits, class _Yp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, shared_ptr<_Yp> const& __p)
+{
+    return __os << __p.get();
+}
+
+template<class _CharT, class _Traits, class _Yp, class _Dp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    is_same<void, typename __void_t<decltype((declval<basic_ostream<_CharT, _Traits>&>() << declval<typename unique_ptr<_Yp, _Dp>::pointer>()))>::type>::value,
+    basic_ostream<_CharT, _Traits>&
+>::type
+operator<<(basic_ostream<_CharT, _Traits>& __os, unique_ptr<_Yp, _Dp> const& __p)
+{
+    return __os << __p.get();
+}
+
+template <class _CharT, class _Traits, size_t _Size>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const bitset<_Size>& __x)
+{
+    return __os << __x.template to_string<_CharT, _Traits>
+                        (use_facet<ctype<_CharT> >(__os.getloc()).widen('0'),
+                         use_facet<ctype<_CharT> >(__os.getloc()).widen('1'));
+}
+
+#ifndef _LIBCUDACXX_DO_NOT_ASSUME_STREAMS_EXPLICIT_INSTANTIATION_IN_DYLIB
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_ostream<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_ostream<wchar_t>)
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_OSTREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/queue b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/queue
new file mode 100644
index 000000000000..25676b7e2d60
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/queue
@@ -0,0 +1,803 @@
+// -*- C++ -*-
+//===--------------------------- queue ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_QUEUE
+#define _LIBCUDACXX_QUEUE
+
+/*
+    queue synopsis
+
+namespace std
+{
+
+template <class T, class Container = deque<T>>
+class queue
+{
+public:
+    typedef Container                                container_type;
+    typedef typename container_type::value_type      value_type;
+    typedef typename container_type::reference       reference;
+    typedef typename container_type::const_reference const_reference;
+    typedef typename container_type::size_type       size_type;
+
+protected:
+    container_type c;
+
+public:
+    queue() = default;
+    ~queue() = default;
+
+    queue(const queue& q) = default;
+    queue(queue&& q) = default;
+
+    queue& operator=(const queue& q) = default;
+    queue& operator=(queue&& q) = default;
+
+    explicit queue(const container_type& c);
+    explicit queue(container_type&& c)
+    template <class Alloc>
+        explicit queue(const Alloc& a);
+    template <class Alloc>
+        queue(const container_type& c, const Alloc& a);
+    template <class Alloc>
+        queue(container_type&& c, const Alloc& a);
+    template <class Alloc>
+        queue(const queue& q, const Alloc& a);
+    template <class Alloc>
+        queue(queue&& q, const Alloc& a);
+
+    bool      empty() const;
+    size_type size() const;
+
+    reference       front();
+    const_reference front() const;
+    reference       back();
+    const_reference back() const;
+
+    void push(const value_type& v);
+    void push(value_type&& v);
+    template <class... Args> reference emplace(Args&&... args); // reference in C++17
+    void pop();
+
+    void swap(queue& q) noexcept(is_nothrow_swappable_v<Container>)
+};
+
+template<class Container>
+  queue(Container) -> queue<typename Container::value_type, Container>; // C++17
+
+template<class Container, class Allocator>
+  queue(Container, Allocator) -> queue<typename Container::value_type, Container>; // C++17
+
+template <class T, class Container>
+  bool operator==(const queue<T, Container>& x,const queue<T, Container>& y);
+
+template <class T, class Container>
+  bool operator< (const queue<T, Container>& x,const queue<T, Container>& y);
+
+template <class T, class Container>
+  bool operator!=(const queue<T, Container>& x,const queue<T, Container>& y);
+
+template <class T, class Container>
+  bool operator> (const queue<T, Container>& x,const queue<T, Container>& y);
+
+template <class T, class Container>
+  bool operator>=(const queue<T, Container>& x,const queue<T, Container>& y);
+
+template <class T, class Container>
+  bool operator<=(const queue<T, Container>& x,const queue<T, Container>& y);
+
+template <class T, class Container>
+  void swap(queue<T, Container>& x, queue<T, Container>& y)
+  noexcept(noexcept(x.swap(y)));
+
+template <class T, class Container = vector<T>,
+          class Compare = less<typename Container::value_type>>
+class priority_queue
+{
+public:
+    typedef Container                                container_type;
+    typedef typename container_type::value_type      value_type;
+    typedef typename container_type::reference       reference;
+    typedef typename container_type::const_reference const_reference;
+    typedef typename container_type::size_type       size_type;
+
+protected:
+    container_type c;
+    Compare comp;
+
+public:
+    priority_queue() = default;
+    ~priority_queue() = default;
+
+    priority_queue(const priority_queue& q) = default;
+    priority_queue(priority_queue&& q) = default;
+
+    priority_queue& operator=(const priority_queue& q) = default;
+    priority_queue& operator=(priority_queue&& q) = default;
+
+    explicit priority_queue(const Compare& comp);
+    priority_queue(const Compare& comp, const container_type& c);
+    explicit priority_queue(const Compare& comp, container_type&& c);
+    template <class InputIterator>
+        priority_queue(InputIterator first, InputIterator last,
+                       const Compare& comp = Compare());
+    template <class InputIterator>
+        priority_queue(InputIterator first, InputIterator last,
+                       const Compare& comp, const container_type& c);
+    template <class InputIterator>
+        priority_queue(InputIterator first, InputIterator last,
+                       const Compare& comp, container_type&& c);
+    template <class Alloc>
+        explicit priority_queue(const Alloc& a);
+    template <class Alloc>
+        priority_queue(const Compare& comp, const Alloc& a);
+    template <class Alloc>
+        priority_queue(const Compare& comp, const container_type& c,
+                       const Alloc& a);
+    template <class Alloc>
+        priority_queue(const Compare& comp, container_type&& c,
+                       const Alloc& a);
+    template <class Alloc>
+        priority_queue(const priority_queue& q, const Alloc& a);
+    template <class Alloc>
+        priority_queue(priority_queue&& q, const Alloc& a);
+
+    bool            empty() const;
+    size_type       size() const;
+    const_reference top() const;
+
+    void push(const value_type& v);
+    void push(value_type&& v);
+    template <class... Args> void emplace(Args&&... args);
+    void pop();
+
+    void swap(priority_queue& q)
+        noexcept(is_nothrow_swappable_v<Container> &&
+                 is_nothrow_swappable_v<Comp>)
+};
+
+template <class Compare, class Container>
+priority_queue(Compare, Container)
+    -> priority_queue<typename Container::value_type, Container, Compare>; // C++17
+
+template<class InputIterator,
+         class Compare = less<typename iterator_traits<InputIterator>::value_type>,
+         class Container = vector<typename iterator_traits<InputIterator>::value_type>>
+priority_queue(InputIterator, InputIterator, Compare = Compare(), Container = Container())
+    -> priority_queue<typename iterator_traits<InputIterator>::value_type, Container, Compare>; // C++17
+
+template<class Compare, class Container, class Allocator>
+priority_queue(Compare, Container, Allocator)
+    -> priority_queue<typename Container::value_type, Container, Compare>; // C++17
+
+template <class T, class Container, class Compare>
+  void swap(priority_queue<T, Container, Compare>& x,
+            priority_queue<T, Container, Compare>& y)
+            noexcept(noexcept(x.swap(y)));
+
+}  // std
+
+*/
+
+#include <__config>
+#include <deque>
+#include <vector>
+#include <functional>
+#include <algorithm>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp, class _Container = deque<_Tp> > class _LIBCUDACXX_TEMPLATE_VIS queue;
+
+template <class _Tp, class _Container>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y);
+
+template <class _Tp, class _Container>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y);
+
+template <class _Tp, class _Container /*= deque<_Tp>*/>
+class _LIBCUDACXX_TEMPLATE_VIS queue
+{
+public:
+    typedef _Container                               container_type;
+    typedef typename container_type::value_type      value_type;
+    typedef typename container_type::reference       reference;
+    typedef typename container_type::const_reference const_reference;
+    typedef typename container_type::size_type       size_type;
+    static_assert((is_same<_Tp, value_type>::value), "" );
+
+protected:
+    container_type c;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    queue()
+        _NOEXCEPT_(is_nothrow_default_constructible<container_type>::value)
+        : c() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    queue(const queue& __q) : c(__q.c) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    queue& operator=(const queue& __q) {c = __q.c; return *this;}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    queue(queue&& __q)
+        _NOEXCEPT_(is_nothrow_move_constructible<container_type>::value)
+        : c(_CUDA_VSTD::move(__q.c)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    queue& operator=(queue&& __q)
+        _NOEXCEPT_(is_nothrow_move_assignable<container_type>::value)
+        {c = _CUDA_VSTD::move(__q.c); return *this;}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit queue(const container_type& __c)  : c(__c) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit queue(container_type&& __c) : c(_CUDA_VSTD::move(__c)) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit queue(const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0)
+            : c(__a) {}
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        queue(const queue& __q, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0)
+            : c(__q.c, __a) {}
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        queue(const container_type& __c, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0)
+            : c(__c, __a) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        queue(container_type&& __c, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0)
+            : c(_CUDA_VSTD::move(__c), __a) {}
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        queue(queue&& __q, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0)
+            : c(_CUDA_VSTD::move(__q.c), __a) {}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const {return c.empty();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const  {return c.size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference       front()       {return c.front();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference front() const {return c.front();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference       back()        {return c.back();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference back() const  {return c.back();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void push(const value_type& __v) {c.push_back(__v);}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void push(value_type&& __v)      {c.push_back(_CUDA_VSTD::move(__v));}
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+#if _LIBCUDACXX_STD_VER > 14
+        decltype(auto) emplace(_Args&&... __args)
+            { return c.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);}
+#else
+        void     emplace(_Args&&... __args)
+            {        c.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);}
+#endif
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pop() {c.pop_front();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(queue& __q)
+        _NOEXCEPT_(__is_nothrow_swappable<container_type>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(c, __q.c);
+    }
+
+    template <class _T1, class _C1>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool
+    operator==(const queue<_T1, _C1>& __x,const queue<_T1, _C1>& __y);
+
+    template <class _T1, class _C1>
+    friend
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool
+    operator< (const queue<_T1, _C1>& __x,const queue<_T1, _C1>& __y);
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _Container,
+         class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type
+>
+queue(_Container)
+    -> queue<typename _Container::value_type, _Container>;
+
+template<class _Container,
+         class _Alloc,
+         class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type,
+         class = typename enable_if< __is_allocator<_Alloc>::value, nullptr_t>::type
+>
+queue(_Container, _Alloc)
+    -> queue<typename _Container::value_type, _Container>;
+#endif
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y)
+{
+    return __x.c == __y.c;
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y)
+{
+    return __x.c < __y.c;
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const queue<_Tp, _Container>& __x,const queue<_Tp, _Container>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<
+    __is_swappable<_Container>::value,
+    void
+>::type
+swap(queue<_Tp, _Container>& __x, queue<_Tp, _Container>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Tp, class _Container, class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator<queue<_Tp, _Container>, _Alloc>
+    : public uses_allocator<_Container, _Alloc>
+{
+};
+
+template <class _Tp, class _Container = vector<_Tp>,
+          class _Compare = less<typename _Container::value_type> >
+class _LIBCUDACXX_TEMPLATE_VIS priority_queue
+{
+public:
+    typedef _Container                               container_type;
+    typedef _Compare                                 value_compare;
+    typedef typename container_type::value_type      value_type;
+    typedef typename container_type::reference       reference;
+    typedef typename container_type::const_reference const_reference;
+    typedef typename container_type::size_type       size_type;
+    static_assert((is_same<_Tp, value_type>::value), "" );
+
+protected:
+    container_type c;
+    value_compare comp;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    priority_queue()
+        _NOEXCEPT_(is_nothrow_default_constructible<container_type>::value &&
+                   is_nothrow_default_constructible<value_compare>::value)
+        : c(), comp() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    priority_queue(const priority_queue& __q) : c(__q.c), comp(__q.comp) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    priority_queue& operator=(const priority_queue& __q)
+        {c = __q.c; comp = __q.comp; return *this;}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    priority_queue(priority_queue&& __q)
+        _NOEXCEPT_(is_nothrow_move_constructible<container_type>::value &&
+                   is_nothrow_move_constructible<value_compare>::value)
+        : c(_CUDA_VSTD::move(__q.c)), comp(_CUDA_VSTD::move(__q.comp)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    priority_queue& operator=(priority_queue&& __q)
+        _NOEXCEPT_(is_nothrow_move_assignable<container_type>::value &&
+                   is_nothrow_move_assignable<value_compare>::value)
+        {c = _CUDA_VSTD::move(__q.c); comp = _CUDA_VSTD::move(__q.comp); return *this;}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit priority_queue(const value_compare& __comp)
+        : c(), comp(__comp) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    priority_queue(const value_compare& __comp, const container_type& __c);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit priority_queue(const value_compare& __comp, container_type&& __c);
+#endif
+    template <class _InputIter>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(_InputIter __f, _InputIter __l,
+                       const value_compare& __comp = value_compare());
+    template <class _InputIter>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(_InputIter __f, _InputIter __l,
+                       const value_compare& __comp, const container_type& __c);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _InputIter>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(_InputIter __f, _InputIter __l,
+                       const value_compare& __comp, container_type&& __c);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit priority_queue(const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0);
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(const value_compare& __comp, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0);
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(const value_compare& __comp, const container_type& __c,
+                       const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0);
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(const priority_queue& __q, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(const value_compare& __comp, container_type&& __c,
+                       const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0);
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        priority_queue(priority_queue&& __q, const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool            empty() const {return c.empty();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type       size() const  {return c.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference top() const   {return c.front();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void push(const value_type& __v);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void push(value_type&& __v);
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void emplace(_Args&&... __args);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pop();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(priority_queue& __q)
+        _NOEXCEPT_(__is_nothrow_swappable<container_type>::value &&
+                   __is_nothrow_swappable<value_compare>::value);
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template <class _Compare,
+          class _Container,
+          class = typename enable_if<!__is_allocator<_Compare>::value, nullptr_t>::type,
+          class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type
+>
+priority_queue(_Compare, _Container)
+    -> priority_queue<typename _Container::value_type, _Container, _Compare>;
+
+template<class _InputIterator,
+         class _Compare   = less<typename iterator_traits<_InputIterator>::value_type>,
+         class _Container = vector<typename iterator_traits<_InputIterator>::value_type>,
+         class = typename enable_if< __is_input_iterator<_InputIterator>::value, nullptr_t>::type,
+         class = typename enable_if<!__is_allocator<_Compare>::value, nullptr_t>::type,
+         class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type
+>
+priority_queue(_InputIterator, _InputIterator, _Compare = _Compare(), _Container = _Container())
+    -> priority_queue<typename iterator_traits<_InputIterator>::value_type, _Container, _Compare>;
+
+template<class _Compare,
+         class _Container,
+         class _Alloc,
+         class = typename enable_if<!__is_allocator<_Compare>::value, nullptr_t>::type,
+         class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type,
+         class = typename enable_if< __is_allocator<_Alloc>::value, nullptr_t>::type
+>
+priority_queue(_Compare, _Container, _Alloc)
+    -> priority_queue<typename _Container::value_type, _Container, _Compare>;
+#endif
+
+template <class _Tp, class _Container, class _Compare>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const _Compare& __comp,
+                                                          const container_type& __c)
+    : c(__c),
+      comp(__comp)
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const value_compare& __comp,
+                                                          container_type&& __c)
+    : c(_CUDA_VSTD::move(__c)),
+      comp(__comp)
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+template <class _InputIter>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(_InputIter __f, _InputIter __l,
+                                                          const value_compare& __comp)
+    : c(__f, __l),
+      comp(__comp)
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+template <class _Tp, class _Container, class _Compare>
+template <class _InputIter>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(_InputIter __f, _InputIter __l,
+                                                          const value_compare& __comp,
+                                                          const container_type& __c)
+    : c(__c),
+      comp(__comp)
+{
+    c.insert(c.end(), __f, __l);
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+template <class _InputIter>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(_InputIter __f, _InputIter __l,
+                                                          const value_compare& __comp,
+                                                          container_type&& __c)
+    : c(_CUDA_VSTD::move(__c)),
+      comp(__comp)
+{
+    c.insert(c.end(), __f, __l);
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+template <class _Alloc>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type*)
+    : c(__a)
+{
+}
+
+template <class _Tp, class _Container, class _Compare>
+template <class _Alloc>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const value_compare& __comp,
+                                                          const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type*)
+    : c(__a),
+      comp(__comp)
+{
+}
+
+template <class _Tp, class _Container, class _Compare>
+template <class _Alloc>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const value_compare& __comp,
+                                                          const container_type& __c,
+                                                          const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type*)
+    : c(__c, __a),
+      comp(__comp)
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+template <class _Tp, class _Container, class _Compare>
+template <class _Alloc>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const priority_queue& __q,
+                                                          const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type*)
+    : c(__q.c, __a),
+      comp(__q.comp)
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+template <class _Alloc>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(const value_compare& __comp,
+                                                          container_type&& __c,
+                                                          const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type*)
+    : c(_CUDA_VSTD::move(__c), __a),
+      comp(__comp)
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+template <class _Tp, class _Container, class _Compare>
+template <class _Alloc>
+inline
+priority_queue<_Tp, _Container, _Compare>::priority_queue(priority_queue&& __q,
+                                                          const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type*)
+    : c(_CUDA_VSTD::move(__q.c), __a),
+      comp(_CUDA_VSTD::move(__q.comp))
+{
+    _CUDA_VSTD::make_heap(c.begin(), c.end(), comp);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+inline
+void
+priority_queue<_Tp, _Container, _Compare>::push(const value_type& __v)
+{
+    c.push_back(__v);
+    _CUDA_VSTD::push_heap(c.begin(), c.end(), comp);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+inline
+void
+priority_queue<_Tp, _Container, _Compare>::push(value_type&& __v)
+{
+    c.push_back(_CUDA_VSTD::move(__v));
+    _CUDA_VSTD::push_heap(c.begin(), c.end(), comp);
+}
+
+template <class _Tp, class _Container, class _Compare>
+template <class... _Args>
+inline
+void
+priority_queue<_Tp, _Container, _Compare>::emplace(_Args&&... __args)
+{
+    c.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);
+    _CUDA_VSTD::push_heap(c.begin(), c.end(), comp);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Container, class _Compare>
+inline
+void
+priority_queue<_Tp, _Container, _Compare>::pop()
+{
+    _CUDA_VSTD::pop_heap(c.begin(), c.end(), comp);
+    c.pop_back();
+}
+
+template <class _Tp, class _Container, class _Compare>
+inline
+void
+priority_queue<_Tp, _Container, _Compare>::swap(priority_queue& __q)
+        _NOEXCEPT_(__is_nothrow_swappable<container_type>::value &&
+                   __is_nothrow_swappable<value_compare>::value)
+{
+    using _CUDA_VSTD::swap;
+    swap(c, __q.c);
+    swap(comp, __q.comp);
+}
+
+template <class _Tp, class _Container, class _Compare>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<
+    __is_swappable<_Container>::value
+    && __is_swappable<_Compare>::value,
+    void
+>::type
+swap(priority_queue<_Tp, _Container, _Compare>& __x,
+     priority_queue<_Tp, _Container, _Compare>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Tp, class _Container, class _Compare, class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator<priority_queue<_Tp, _Container, _Compare>, _Alloc>
+    : public uses_allocator<_Container, _Alloc>
+{
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_QUEUE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/random b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/random
new file mode 100644
index 000000000000..6a6c483dcccf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/random
@@ -0,0 +1,6745 @@
+// -*- C++ -*-
+//===--------------------------- random -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_RANDOM
+#define _LIBCUDACXX_RANDOM
+
+/*
+    random synopsis
+
+#include <initializer_list>
+
+namespace std
+{
+
+// Engines
+
+template <class UIntType, UIntType a, UIntType c, UIntType m>
+class linear_congruential_engine
+{
+public:
+    // types
+    typedef UIntType result_type;
+
+    // engine characteristics
+    static constexpr result_type multiplier = a;
+    static constexpr result_type increment = c;
+    static constexpr result_type modulus = m;
+    static constexpr result_type min() { return c == 0u ? 1u: 0u;}
+    static constexpr result_type max() { return m - 1u;}
+    static constexpr result_type default_seed = 1u;
+
+    // constructors and seeding functions
+    explicit linear_congruential_engine(result_type s = default_seed);
+    template<class Sseq> explicit linear_congruential_engine(Sseq& q);
+    void seed(result_type s = default_seed);
+    template<class Sseq> void seed(Sseq& q);
+
+    // generating functions
+    result_type operator()();
+    void discard(unsigned long long z);
+};
+
+template <class UIntType, UIntType a, UIntType c, UIntType m>
+bool
+operator==(const linear_congruential_engine<UIntType, a, c, m>& x,
+           const linear_congruential_engine<UIntType, a, c, m>& y);
+
+template <class UIntType, UIntType a, UIntType c, UIntType m>
+bool
+operator!=(const linear_congruential_engine<UIntType, a, c, m>& x,
+           const linear_congruential_engine<UIntType, a, c, m>& y);
+
+template <class charT, class traits,
+          class UIntType, UIntType a, UIntType c, UIntType m>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os,
+           const linear_congruential_engine<UIntType, a, c, m>& x);
+
+template <class charT, class traits,
+          class UIntType, UIntType a, UIntType c, UIntType m>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is,
+           linear_congruential_engine<UIntType, a, c, m>& x);
+
+template <class UIntType, size_t w, size_t n, size_t m, size_t r,
+          UIntType a, size_t u, UIntType d, size_t s,
+          UIntType b, size_t t, UIntType c, size_t l, UIntType f>
+class mersenne_twister_engine
+{
+public:
+    // types
+    typedef UIntType result_type;
+
+    // engine characteristics
+    static constexpr size_t word_size = w;
+    static constexpr size_t state_size = n;
+    static constexpr size_t shift_size = m;
+    static constexpr size_t mask_bits = r;
+    static constexpr result_type xor_mask = a;
+    static constexpr size_t tempering_u = u;
+    static constexpr result_type tempering_d = d;
+    static constexpr size_t tempering_s = s;
+    static constexpr result_type tempering_b = b;
+    static constexpr size_t tempering_t = t;
+    static constexpr result_type tempering_c = c;
+    static constexpr size_t tempering_l = l;
+    static constexpr result_type initialization_multiplier = f;
+    static constexpr result_type min () { return 0; }
+    static constexpr result_type max() { return 2^w - 1; }
+    static constexpr result_type default_seed = 5489u;
+
+    // constructors and seeding functions
+    explicit mersenne_twister_engine(result_type value = default_seed);
+    template<class Sseq> explicit mersenne_twister_engine(Sseq& q);
+    void seed(result_type value = default_seed);
+    template<class Sseq> void seed(Sseq& q);
+
+    // generating functions
+    result_type operator()();
+    void discard(unsigned long long z);
+};
+
+template <class UIntType, size_t w, size_t n, size_t m, size_t r,
+          UIntType a, size_t u, UIntType d, size_t s,
+          UIntType b, size_t t, UIntType c, size_t l, UIntType f>
+bool
+operator==(
+    const mersenne_twister_engine<UIntType, w, n, m, r, a, u, d, s, b, t, c, l, f>& x,
+    const mersenne_twister_engine<UIntType, w, n, m, r, a, u, d, s, b, t, c, l, f>& y);
+
+template <class UIntType, size_t w, size_t n, size_t m, size_t r,
+          UIntType a, size_t u, UIntType d, size_t s,
+          UIntType b, size_t t, UIntType c, size_t l, UIntType f>
+bool
+operator!=(
+    const mersenne_twister_engine<UIntType, w, n, m, r, a, u, d, s, b, t, c, l, f>& x,
+    const mersenne_twister_engine<UIntType, w, n, m, r, a, u, d, s, b, t, c, l, f>& y);
+
+template <class charT, class traits,
+          class UIntType, size_t w, size_t n, size_t m, size_t r,
+          UIntType a, size_t u, UIntType d, size_t s,
+          UIntType b, size_t t, UIntType c, size_t l, UIntType f>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os,
+           const mersenne_twister_engine<UIntType, w, n, m, r, a, u, d, s, b, t, c, l, f>& x);
+
+template <class charT, class traits,
+          class UIntType, size_t w, size_t n, size_t m, size_t r,
+          UIntType a, size_t u, UIntType d, size_t s,
+          UIntType b, size_t t, UIntType c, size_t l, UIntType f>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is,
+           mersenne_twister_engine<UIntType, w, n, m, r, a, u, d, s, b, t, c, l, f>& x);
+
+template<class UIntType, size_t w, size_t s, size_t r>
+class subtract_with_carry_engine
+{
+public:
+    // types
+    typedef UIntType result_type;
+
+    // engine characteristics
+    static constexpr size_t word_size = w;
+    static constexpr size_t short_lag = s;
+    static constexpr size_t long_lag = r;
+    static constexpr result_type min() { return 0; }
+    static constexpr result_type max() { return m-1; }
+    static constexpr result_type default_seed = 19780503u;
+
+    // constructors and seeding functions
+    explicit subtract_with_carry_engine(result_type value = default_seed);
+    template<class Sseq> explicit subtract_with_carry_engine(Sseq& q);
+    void seed(result_type value = default_seed);
+    template<class Sseq> void seed(Sseq& q);
+
+    // generating functions
+    result_type operator()();
+    void discard(unsigned long long z);
+};
+
+template<class UIntType, size_t w, size_t s, size_t r>
+bool
+operator==(
+    const subtract_with_carry_engine<UIntType, w, s, r>& x,
+    const subtract_with_carry_engine<UIntType, w, s, r>& y);
+
+template<class UIntType, size_t w, size_t s, size_t r>
+bool
+operator!=(
+    const subtract_with_carry_engine<UIntType, w, s, r>& x,
+    const subtract_with_carry_engine<UIntType, w, s, r>& y);
+
+template <class charT, class traits,
+          class UIntType, size_t w, size_t s, size_t r>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os,
+           const subtract_with_carry_engine<UIntType, w, s, r>& x);
+
+template <class charT, class traits,
+          class UIntType, size_t w, size_t s, size_t r>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is,
+           subtract_with_carry_engine<UIntType, w, s, r>& x);
+
+template<class Engine, size_t p, size_t r>
+class discard_block_engine
+{
+public:
+    // types
+    typedef typename Engine::result_type result_type;
+
+    // engine characteristics
+    static constexpr size_t block_size = p;
+    static constexpr size_t used_block = r;
+    static constexpr result_type min() { return Engine::min(); }
+    static constexpr result_type max() { return Engine::max(); }
+
+    // constructors and seeding functions
+    discard_block_engine();
+    explicit discard_block_engine(const Engine& e);
+    explicit discard_block_engine(Engine&& e);
+    explicit discard_block_engine(result_type s);
+    template<class Sseq> explicit discard_block_engine(Sseq& q);
+    void seed();
+    void seed(result_type s);
+    template<class Sseq> void seed(Sseq& q);
+
+    // generating functions
+    result_type operator()();
+    void discard(unsigned long long z);
+
+    // property functions
+    const Engine& base() const noexcept;
+};
+
+template<class Engine, size_t p, size_t r>
+bool
+operator==(
+    const discard_block_engine<Engine, p, r>& x,
+    const discard_block_engine<Engine, p, r>& y);
+
+template<class Engine, size_t p, size_t r>
+bool
+operator!=(
+    const discard_block_engine<Engine, p, r>& x,
+    const discard_block_engine<Engine, p, r>& y);
+
+template <class charT, class traits,
+          class Engine, size_t p, size_t r>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os,
+           const discard_block_engine<Engine, p, r>& x);
+
+template <class charT, class traits,
+          class Engine, size_t p, size_t r>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is,
+           discard_block_engine<Engine, p, r>& x);
+
+template<class Engine, size_t w, class UIntType>
+class independent_bits_engine
+{
+public:
+    // types
+    typedef UIntType result_type;
+
+    // engine characteristics
+    static constexpr result_type min() { return 0; }
+    static constexpr result_type max() { return 2^w - 1; }
+
+    // constructors and seeding functions
+    independent_bits_engine();
+    explicit independent_bits_engine(const Engine& e);
+    explicit independent_bits_engine(Engine&& e);
+    explicit independent_bits_engine(result_type s);
+    template<class Sseq> explicit independent_bits_engine(Sseq& q);
+    void seed();
+    void seed(result_type s);
+    template<class Sseq> void seed(Sseq& q);
+
+    // generating functions
+    result_type operator()(); void discard(unsigned long long z);
+
+    // property functions
+    const Engine& base() const noexcept;
+};
+
+template<class Engine, size_t w, class UIntType>
+bool
+operator==(
+    const independent_bits_engine<Engine, w, UIntType>& x,
+    const independent_bits_engine<Engine, w, UIntType>& y);
+
+template<class Engine, size_t w, class UIntType>
+bool
+operator!=(
+    const independent_bits_engine<Engine, w, UIntType>& x,
+    const independent_bits_engine<Engine, w, UIntType>& y);
+
+template <class charT, class traits,
+          class Engine, size_t w, class UIntType>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os,
+           const independent_bits_engine<Engine, w, UIntType>& x);
+
+template <class charT, class traits,
+          class Engine, size_t w, class UIntType>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is,
+           independent_bits_engine<Engine, w, UIntType>& x);
+
+template<class Engine, size_t k>
+class shuffle_order_engine
+{
+public:
+    // types
+    typedef typename Engine::result_type result_type;
+
+    // engine characteristics
+    static constexpr size_t table_size = k;
+    static constexpr result_type min() { return Engine::min; }
+    static constexpr result_type max() { return Engine::max; }
+
+    // constructors and seeding functions
+    shuffle_order_engine();
+    explicit shuffle_order_engine(const Engine& e);
+    explicit shuffle_order_engine(Engine&& e);
+    explicit shuffle_order_engine(result_type s);
+    template<class Sseq> explicit shuffle_order_engine(Sseq& q);
+    void seed();
+    void seed(result_type s);
+    template<class Sseq> void seed(Sseq& q);
+
+    // generating functions
+    result_type operator()();
+    void discard(unsigned long long z);
+
+    // property functions
+    const Engine& base() const noexcept;
+};
+
+template<class Engine, size_t k>
+bool
+operator==(
+    const shuffle_order_engine<Engine, k>& x,
+    const shuffle_order_engine<Engine, k>& y);
+
+template<class Engine, size_t k>
+bool
+operator!=(
+    const shuffle_order_engine<Engine, k>& x,
+    const shuffle_order_engine<Engine, k>& y);
+
+template <class charT, class traits,
+          class Engine, size_t k>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os,
+           const shuffle_order_engine<Engine, k>& x);
+
+template <class charT, class traits,
+          class Engine, size_t k>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is,
+           shuffle_order_engine<Engine, k>& x);
+
+typedef linear_congruential_engine<uint_fast32_t, 16807, 0, 2147483647>
+                                                                   minstd_rand0;
+typedef linear_congruential_engine<uint_fast32_t, 48271, 0, 2147483647>
+                                                                    minstd_rand;
+typedef mersenne_twister_engine<uint_fast32_t, 32, 624, 397, 31,
+                                0x9908b0df,
+                                11, 0xffffffff,
+                                7,  0x9d2c5680,
+                                15, 0xefc60000,
+                                18, 1812433253>                         mt19937;
+typedef mersenne_twister_engine<uint_fast64_t, 64, 312, 156, 31,
+                                0xb5026f5aa96619e9,
+                                29, 0x5555555555555555,
+                                17, 0x71d67fffeda60000,
+                                37, 0xfff7eee000000000,
+                                43, 6364136223846793005>             mt19937_64;
+typedef subtract_with_carry_engine<uint_fast32_t, 24, 10, 24>     ranlux24_base;
+typedef subtract_with_carry_engine<uint_fast64_t, 48,  5, 12>     ranlux48_base;
+typedef discard_block_engine<ranlux24_base, 223, 23>                   ranlux24;
+typedef discard_block_engine<ranlux48_base, 389, 11>                   ranlux48;
+typedef shuffle_order_engine<minstd_rand0, 256>                         knuth_b;
+typedef minstd_rand                                       default_random_engine;
+
+// Generators
+
+class random_device
+{
+public:
+    // types
+    typedef unsigned int result_type;
+
+    // generator characteristics
+    static constexpr result_type min() { return numeric_limits<result_type>::min(); }
+    static constexpr result_type max() { return numeric_limits<result_type>::max(); }
+
+    // constructors
+    explicit random_device(const string& token = "/dev/urandom");
+
+    // generating functions
+    result_type operator()();
+
+    // property functions
+    double entropy() const noexcept;
+
+    // no copy functions
+    random_device(const random_device& ) = delete;
+    void operator=(const random_device& ) = delete;
+};
+
+// Utilities
+
+class seed_seq
+{
+public:
+    // types
+    typedef uint_least32_t result_type;
+
+    // constructors
+    seed_seq();
+    template<class T>
+        seed_seq(initializer_list<T> il);
+    template<class InputIterator>
+        seed_seq(InputIterator begin, InputIterator end);
+
+    // generating functions
+    template<class RandomAccessIterator>
+        void generate(RandomAccessIterator begin, RandomAccessIterator end);
+
+    // property functions
+    size_t size() const;
+    template<class OutputIterator>
+        void param(OutputIterator dest) const;
+
+    // no copy functions
+    seed_seq(const seed_seq&) = delete;
+    void operator=(const seed_seq& ) = delete;
+};
+
+template<class RealType, size_t bits, class URNG>
+    RealType generate_canonical(URNG& g);
+
+// Distributions
+
+template<class IntType = int>
+class uniform_int_distribution
+{
+public:
+    // types
+    typedef IntType result_type;
+
+    class param_type
+    {
+    public:
+        typedef uniform_int_distribution distribution_type;
+
+        explicit param_type(IntType a = 0,
+                                    IntType b = numeric_limits<IntType>::max());
+
+        result_type a() const;
+        result_type b() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit uniform_int_distribution(IntType a = 0,
+                                    IntType b = numeric_limits<IntType>::max());
+    explicit uniform_int_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type a() const;
+    result_type b() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const uniform_int_distribution& x,
+                           const uniform_int_distribution& y);
+    friend bool operator!=(const uniform_int_distribution& x,
+                           const uniform_int_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const uniform_int_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               uniform_int_distribution& x);
+};
+
+template<class RealType = double>
+class uniform_real_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef uniform_real_distribution distribution_type;
+
+        explicit param_type(RealType a = 0,
+                            RealType b = 1);
+
+        result_type a() const;
+        result_type b() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit uniform_real_distribution(RealType a = 0.0, RealType b = 1.0);
+    explicit uniform_real_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type a() const;
+    result_type b() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const uniform_real_distribution& x,
+                           const uniform_real_distribution& y);
+    friend bool operator!=(const uniform_real_distribution& x,
+                           const uniform_real_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const uniform_real_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               uniform_real_distribution& x);
+};
+
+class bernoulli_distribution
+{
+public:
+    // types
+    typedef bool result_type;
+
+    class param_type
+    {
+    public:
+        typedef bernoulli_distribution distribution_type;
+
+        explicit param_type(double p = 0.5);
+
+        double p() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit bernoulli_distribution(double p = 0.5);
+    explicit bernoulli_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    double p() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const bernoulli_distribution& x,
+                           const bernoulli_distribution& y);
+    friend bool operator!=(const bernoulli_distribution& x,
+                           const bernoulli_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const bernoulli_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               bernoulli_distribution& x);
+};
+
+template<class IntType = int>
+class binomial_distribution
+{
+public:
+    // types
+    typedef IntType result_type;
+
+    class param_type
+    {
+    public:
+        typedef binomial_distribution distribution_type;
+
+        explicit param_type(IntType t = 1, double p = 0.5);
+
+        IntType t() const;
+        double p() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit binomial_distribution(IntType t = 1, double p = 0.5);
+    explicit binomial_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    IntType t() const;
+    double p() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const binomial_distribution& x,
+                           const binomial_distribution& y);
+    friend bool operator!=(const binomial_distribution& x,
+                           const binomial_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const binomial_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               binomial_distribution& x);
+};
+
+template<class IntType = int>
+class geometric_distribution
+{
+public:
+    // types
+    typedef IntType result_type;
+
+    class param_type
+    {
+    public:
+        typedef geometric_distribution distribution_type;
+
+        explicit param_type(double p = 0.5);
+
+        double p() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit geometric_distribution(double p = 0.5);
+    explicit geometric_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    double p() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const geometric_distribution& x,
+                           const geometric_distribution& y);
+    friend bool operator!=(const geometric_distribution& x,
+                           const geometric_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const geometric_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               geometric_distribution& x);
+};
+
+template<class IntType = int>
+class negative_binomial_distribution
+{
+public:
+    // types
+    typedef IntType result_type;
+
+    class param_type
+    {
+    public:
+        typedef negative_binomial_distribution distribution_type;
+
+        explicit param_type(result_type k = 1, double p = 0.5);
+
+        result_type k() const;
+        double p() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit negative_binomial_distribution(result_type k = 1, double p = 0.5);
+    explicit negative_binomial_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type k() const;
+    double p() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const negative_binomial_distribution& x,
+                           const negative_binomial_distribution& y);
+    friend bool operator!=(const negative_binomial_distribution& x,
+                           const negative_binomial_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const negative_binomial_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               negative_binomial_distribution& x);
+};
+
+template<class IntType = int>
+class poisson_distribution
+{
+public:
+    // types
+    typedef IntType result_type;
+
+    class param_type
+    {
+    public:
+        typedef poisson_distribution distribution_type;
+
+        explicit param_type(double mean = 1.0);
+
+        double mean() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit poisson_distribution(double mean = 1.0);
+    explicit poisson_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    double mean() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const poisson_distribution& x,
+                           const poisson_distribution& y);
+    friend bool operator!=(const poisson_distribution& x,
+                           const poisson_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const poisson_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               poisson_distribution& x);
+};
+
+template<class RealType = double>
+class exponential_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef exponential_distribution distribution_type;
+
+        explicit param_type(result_type lambda = 1.0);
+
+        result_type lambda() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit exponential_distribution(result_type lambda = 1.0);
+    explicit exponential_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type lambda() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const exponential_distribution& x,
+                           const exponential_distribution& y);
+    friend bool operator!=(const exponential_distribution& x,
+                           const exponential_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const exponential_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               exponential_distribution& x);
+};
+
+template<class RealType = double>
+class gamma_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef gamma_distribution distribution_type;
+
+        explicit param_type(result_type alpha = 1, result_type beta = 1);
+
+        result_type alpha() const;
+        result_type beta() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit gamma_distribution(result_type alpha = 1, result_type beta = 1);
+    explicit gamma_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type alpha() const;
+    result_type beta() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const gamma_distribution& x,
+                           const gamma_distribution& y);
+    friend bool operator!=(const gamma_distribution& x,
+                           const gamma_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const gamma_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               gamma_distribution& x);
+};
+
+template<class RealType = double>
+class weibull_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef weibull_distribution distribution_type;
+
+        explicit param_type(result_type alpha = 1, result_type beta = 1);
+
+        result_type a() const;
+        result_type b() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit weibull_distribution(result_type a = 1, result_type b = 1);
+    explicit weibull_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type a() const;
+    result_type b() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const weibull_distribution& x,
+                           const weibull_distribution& y);
+    friend bool operator!=(const weibull_distribution& x,
+                           const weibull_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const weibull_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               weibull_distribution& x);
+};
+
+template<class RealType = double>
+class extreme_value_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef extreme_value_distribution distribution_type;
+
+        explicit param_type(result_type a = 0, result_type b = 1);
+
+        result_type a() const;
+        result_type b() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit extreme_value_distribution(result_type a = 0, result_type b = 1);
+    explicit extreme_value_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type a() const;
+    result_type b() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const extreme_value_distribution& x,
+                           const extreme_value_distribution& y);
+    friend bool operator!=(const extreme_value_distribution& x,
+                           const extreme_value_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const extreme_value_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               extreme_value_distribution& x);
+};
+
+template<class RealType = double>
+class normal_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef normal_distribution distribution_type;
+
+        explicit param_type(result_type mean = 0, result_type stddev = 1);
+
+        result_type mean() const;
+        result_type stddev() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructors and reset functions
+    explicit normal_distribution(result_type mean = 0, result_type stddev = 1);
+    explicit normal_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type mean() const;
+    result_type stddev() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const normal_distribution& x,
+                           const normal_distribution& y);
+    friend bool operator!=(const normal_distribution& x,
+                           const normal_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const normal_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               normal_distribution& x);
+};
+
+template<class RealType = double>
+class lognormal_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef lognormal_distribution distribution_type;
+
+        explicit param_type(result_type m = 0, result_type s = 1);
+
+        result_type m() const;
+        result_type s() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit lognormal_distribution(result_type m = 0, result_type s = 1);
+    explicit lognormal_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type m() const;
+    result_type s() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const lognormal_distribution& x,
+                           const lognormal_distribution& y);
+    friend bool operator!=(const lognormal_distribution& x,
+                           const lognormal_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const lognormal_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               lognormal_distribution& x);
+};
+
+template<class RealType = double>
+class chi_squared_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef chi_squared_distribution distribution_type;
+
+        explicit param_type(result_type n = 1);
+
+        result_type n() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit chi_squared_distribution(result_type n = 1);
+    explicit chi_squared_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type n() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const chi_squared_distribution& x,
+                           const chi_squared_distribution& y);
+    friend bool operator!=(const chi_squared_distribution& x,
+                           const chi_squared_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const chi_squared_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               chi_squared_distribution& x);
+};
+
+template<class RealType = double>
+class cauchy_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef cauchy_distribution distribution_type;
+
+        explicit param_type(result_type a = 0, result_type b = 1);
+
+        result_type a() const;
+        result_type b() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit cauchy_distribution(result_type a = 0, result_type b = 1);
+    explicit cauchy_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type a() const;
+    result_type b() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const cauchy_distribution& x,
+                           const cauchy_distribution& y);
+    friend bool operator!=(const cauchy_distribution& x,
+                           const cauchy_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const cauchy_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               cauchy_distribution& x);
+};
+
+template<class RealType = double>
+class fisher_f_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef fisher_f_distribution distribution_type;
+
+        explicit param_type(result_type m = 1, result_type n = 1);
+
+        result_type m() const;
+        result_type n() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit fisher_f_distribution(result_type m = 1, result_type n = 1);
+    explicit fisher_f_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type m() const;
+    result_type n() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const fisher_f_distribution& x,
+                           const fisher_f_distribution& y);
+    friend bool operator!=(const fisher_f_distribution& x,
+                           const fisher_f_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const fisher_f_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               fisher_f_distribution& x);
+};
+
+template<class RealType = double>
+class student_t_distribution
+{
+public:
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef student_t_distribution distribution_type;
+
+        explicit param_type(result_type n = 1);
+
+        result_type n() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    explicit student_t_distribution(result_type n = 1);
+    explicit student_t_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    result_type n() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const student_t_distribution& x,
+                           const student_t_distribution& y);
+    friend bool operator!=(const student_t_distribution& x,
+                           const student_t_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const student_t_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               student_t_distribution& x);
+};
+
+template<class IntType = int>
+class discrete_distribution
+{
+public:
+    // types
+    typedef IntType result_type;
+
+    class param_type
+    {
+    public:
+        typedef discrete_distribution distribution_type;
+
+        param_type();
+        template<class InputIterator>
+            param_type(InputIterator firstW, InputIterator lastW);
+        param_type(initializer_list<double> wl);
+        template<class UnaryOperation>
+            param_type(size_t nw, double xmin, double xmax, UnaryOperation fw);
+
+        vector<double> probabilities() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    discrete_distribution();
+    template<class InputIterator>
+        discrete_distribution(InputIterator firstW, InputIterator lastW);
+    discrete_distribution(initializer_list<double> wl);
+    template<class UnaryOperation>
+        discrete_distribution(size_t nw, double xmin, double xmax,
+                              UnaryOperation fw);
+    explicit discrete_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    vector<double> probabilities() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const discrete_distribution& x,
+                           const discrete_distribution& y);
+    friend bool operator!=(const discrete_distribution& x,
+                           const discrete_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const discrete_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               discrete_distribution& x);
+};
+
+template<class RealType = double>
+class piecewise_constant_distribution
+{
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef piecewise_constant_distribution distribution_type;
+
+        param_type();
+        template<class InputIteratorB, class InputIteratorW>
+            param_type(InputIteratorB firstB, InputIteratorB lastB,
+                       InputIteratorW firstW);
+        template<class UnaryOperation>
+            param_type(initializer_list<result_type> bl, UnaryOperation fw);
+        template<class UnaryOperation>
+            param_type(size_t nw, result_type xmin, result_type xmax,
+                       UnaryOperation fw);
+
+        vector<result_type> intervals() const;
+        vector<result_type> densities() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    piecewise_constant_distribution();
+    template<class InputIteratorB, class InputIteratorW>
+        piecewise_constant_distribution(InputIteratorB firstB,
+                                        InputIteratorB lastB,
+                                        InputIteratorW firstW);
+    template<class UnaryOperation>
+        piecewise_constant_distribution(initializer_list<result_type> bl,
+                                        UnaryOperation fw);
+    template<class UnaryOperation>
+        piecewise_constant_distribution(size_t nw, result_type xmin,
+                                        result_type xmax, UnaryOperation fw);
+    explicit piecewise_constant_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    vector<result_type> intervals() const;
+    vector<result_type> densities() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const piecewise_constant_distribution& x,
+                           const piecewise_constant_distribution& y);
+    friend bool operator!=(const piecewise_constant_distribution& x,
+                           const piecewise_constant_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const piecewise_constant_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               piecewise_constant_distribution& x);
+};
+
+template<class RealType = double>
+class piecewise_linear_distribution
+{
+    // types
+    typedef RealType result_type;
+
+    class param_type
+    {
+    public:
+        typedef piecewise_linear_distribution distribution_type;
+
+        param_type();
+        template<class InputIteratorB, class InputIteratorW>
+            param_type(InputIteratorB firstB, InputIteratorB lastB,
+                       InputIteratorW firstW);
+        template<class UnaryOperation>
+            param_type(initializer_list<result_type> bl, UnaryOperation fw);
+        template<class UnaryOperation>
+            param_type(size_t nw, result_type xmin, result_type xmax,
+                       UnaryOperation fw);
+
+        vector<result_type> intervals() const;
+        vector<result_type> densities() const;
+
+        friend bool operator==(const param_type& x, const param_type& y);
+        friend bool operator!=(const param_type& x, const param_type& y);
+    };
+
+    // constructor and reset functions
+    piecewise_linear_distribution();
+    template<class InputIteratorB, class InputIteratorW>
+        piecewise_linear_distribution(InputIteratorB firstB,
+                                      InputIteratorB lastB,
+                                      InputIteratorW firstW);
+
+    template<class UnaryOperation>
+        piecewise_linear_distribution(initializer_list<result_type> bl,
+                                      UnaryOperation fw);
+
+    template<class UnaryOperation>
+        piecewise_linear_distribution(size_t nw, result_type xmin,
+                                      result_type xmax, UnaryOperation fw);
+
+    explicit piecewise_linear_distribution(const param_type& parm);
+    void reset();
+
+    // generating functions
+    template<class URNG> result_type operator()(URNG& g);
+    template<class URNG> result_type operator()(URNG& g, const param_type& parm);
+
+    // property functions
+    vector<result_type> intervals() const;
+    vector<result_type> densities() const;
+
+    param_type param() const;
+    void param(const param_type& parm);
+
+    result_type min() const;
+    result_type max() const;
+
+    friend bool operator==(const piecewise_linear_distribution& x,
+                           const piecewise_linear_distribution& y);
+    friend bool operator!=(const piecewise_linear_distribution& x,
+                           const piecewise_linear_distribution& y);
+
+    template <class charT, class traits>
+    friend
+    basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os,
+               const piecewise_linear_distribution& x);
+
+    template <class charT, class traits>
+    friend
+    basic_istream<charT, traits>&
+    operator>>(basic_istream<charT, traits>& is,
+               piecewise_linear_distribution& x);
+};
+
+} // std
+*/
+
+#include <__config>
+#include <cstddef>
+#include <cstdint>
+#include <cmath>
+#include <type_traits>
+#include <initializer_list>
+#include <limits>
+#include <algorithm>
+#include <numeric>
+#include <vector>
+#include <string>
+#include <istream>
+#include <ostream>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// __is_seed_sequence
+
+template <class _Sseq, class _Engine>
+struct __is_seed_sequence
+{
+    static _LIBCUDACXX_CONSTEXPR const bool value =
+              !is_convertible<_Sseq, typename _Engine::result_type>::value &&
+              !is_same<typename remove_cv<_Sseq>::type, _Engine>::value;
+};
+
+// linear_congruential_engine
+
+template <unsigned long long __a, unsigned long long __c,
+          unsigned long long __m, unsigned long long _Mp,
+          bool _MightOverflow = (__a != 0 && __m != 0 && __m-1 > (_Mp-__c)/__a)>
+struct __lce_ta;
+
+// 64
+
+template <unsigned long long __a, unsigned long long __c, unsigned long long __m>
+struct __lce_ta<__a, __c, __m, (unsigned long long)(~0), true>
+{
+    typedef unsigned long long result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        // Schrage's algorithm
+        const result_type __q = __m / __a;
+        const result_type __r = __m % __a;
+        const result_type __t0 = __a * (__x % __q);
+        const result_type __t1 = __r * (__x / __q);
+        __x = __t0 + (__t0 < __t1) * __m - __t1;
+        __x += __c - (__x >= __m - __c) * __m;
+        return __x;
+    }
+};
+
+template <unsigned long long __a, unsigned long long __m>
+struct __lce_ta<__a, 0, __m, (unsigned long long)(~0), true>
+{
+    typedef unsigned long long result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        // Schrage's algorithm
+        const result_type __q = __m / __a;
+        const result_type __r = __m % __a;
+        const result_type __t0 = __a * (__x % __q);
+        const result_type __t1 = __r * (__x / __q);
+        __x = __t0 + (__t0 < __t1) * __m - __t1;
+        return __x;
+    }
+};
+
+template <unsigned long long __a, unsigned long long __c, unsigned long long __m>
+struct __lce_ta<__a, __c, __m, (unsigned long long)(~0), false>
+{
+    typedef unsigned long long result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        return (__a * __x + __c) % __m;
+    }
+};
+
+template <unsigned long long __a, unsigned long long __c>
+struct __lce_ta<__a, __c, 0, (unsigned long long)(~0), false>
+{
+    typedef unsigned long long result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        return __a * __x + __c;
+    }
+};
+
+// 32
+
+template <unsigned long long _Ap, unsigned long long _Cp, unsigned long long _Mp>
+struct __lce_ta<_Ap, _Cp, _Mp, unsigned(~0), true>
+{
+    typedef unsigned result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        const result_type __a = static_cast<result_type>(_Ap);
+        const result_type __c = static_cast<result_type>(_Cp);
+        const result_type __m = static_cast<result_type>(_Mp);
+        // Schrage's algorithm
+        const result_type __q = __m / __a;
+        const result_type __r = __m % __a;
+        const result_type __t0 = __a * (__x % __q);
+        const result_type __t1 = __r * (__x / __q);
+        __x = __t0 + (__t0 < __t1) * __m - __t1;
+        __x += __c - (__x >= __m - __c) * __m;
+        return __x;
+    }
+};
+
+template <unsigned long long _Ap, unsigned long long _Mp>
+struct __lce_ta<_Ap, 0, _Mp, unsigned(~0), true>
+{
+    typedef unsigned result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        const result_type __a = static_cast<result_type>(_Ap);
+        const result_type __m = static_cast<result_type>(_Mp);
+        // Schrage's algorithm
+        const result_type __q = __m / __a;
+        const result_type __r = __m % __a;
+        const result_type __t0 = __a * (__x % __q);
+        const result_type __t1 = __r * (__x / __q);
+        __x = __t0 + (__t0 < __t1) * __m - __t1;
+        return __x;
+    }
+};
+
+template <unsigned long long _Ap, unsigned long long _Cp, unsigned long long _Mp>
+struct __lce_ta<_Ap, _Cp, _Mp, unsigned(~0), false>
+{
+    typedef unsigned result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        const result_type __a = static_cast<result_type>(_Ap);
+        const result_type __c = static_cast<result_type>(_Cp);
+        const result_type __m = static_cast<result_type>(_Mp);
+        return (__a * __x + __c) % __m;
+    }
+};
+
+template <unsigned long long _Ap, unsigned long long _Cp>
+struct __lce_ta<_Ap, _Cp, 0, unsigned(~0), false>
+{
+    typedef unsigned result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        const result_type __a = static_cast<result_type>(_Ap);
+        const result_type __c = static_cast<result_type>(_Cp);
+        return __a * __x + __c;
+    }
+};
+
+// 16
+
+template <unsigned long long __a, unsigned long long __c, unsigned long long __m, bool __b>
+struct __lce_ta<__a, __c, __m, (unsigned short)(~0), __b>
+{
+    typedef unsigned short result_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type next(result_type __x)
+    {
+        return static_cast<result_type>(__lce_ta<__a, __c, __m, unsigned(~0)>::next(__x));
+    }
+};
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+class _LIBCUDACXX_TEMPLATE_VIS linear_congruential_engine;
+
+template <class _CharT, class _Traits,
+          class _Up, _Up _Ap, _Up _Cp, _Up _Np>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const linear_congruential_engine<_Up, _Ap, _Cp, _Np>&);
+
+template <class _CharT, class _Traits,
+          class _Up, _Up _Ap, _Up _Cp, _Up _Np>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           linear_congruential_engine<_Up, _Ap, _Cp, _Np>& __x);
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+class _LIBCUDACXX_TEMPLATE_VIS linear_congruential_engine
+{
+public:
+    // types
+    typedef _UIntType result_type;
+
+private:
+    result_type __x_;
+
+    static _LIBCUDACXX_CONSTEXPR const result_type _Mp = result_type(~0);
+
+    static_assert(__m == 0 || __a < __m, "linear_congruential_engine invalid parameters");
+    static_assert(__m == 0 || __c < __m, "linear_congruential_engine invalid parameters");
+public:
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = __c == 0u ? 1u: 0u;
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = __m - 1u;
+    static_assert(_Min < _Max,           "linear_congruential_engine invalid parameters");
+
+    // engine characteristics
+    static _LIBCUDACXX_CONSTEXPR const result_type multiplier = __a;
+    static _LIBCUDACXX_CONSTEXPR const result_type increment = __c;
+    static _LIBCUDACXX_CONSTEXPR const result_type modulus = __m;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() {return _Min;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() {return _Max;}
+    static _LIBCUDACXX_CONSTEXPR const result_type default_seed = 1u;
+
+    // constructors and seeding functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit linear_congruential_engine(result_type __s = default_seed)
+        {seed(__s);}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit linear_congruential_engine(_Sseq& __q,
+        typename enable_if<__is_seed_sequence<_Sseq, linear_congruential_engine>::value>::type* = 0)
+        {seed(__q);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(result_type __s = default_seed)
+        {seed(integral_constant<bool, __m == 0>(),
+              integral_constant<bool, __c == 0>(), __s);}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_seed_sequence<_Sseq, linear_congruential_engine>::value,
+            void
+        >::type
+        seed(_Sseq& __q)
+            {__seed(__q, integral_constant<unsigned,
+                1 + (__m == 0 ? (sizeof(result_type) * __CHAR_BIT__ - 1)/32
+                             :  (__m > 0x100000000ull))>());}
+
+    // generating functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator()()
+        {return __x_ = static_cast<result_type>(__lce_ta<__a, __c, __m, _Mp>::next(__x_));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void discard(unsigned long long __z) {for (; __z; --__z) operator()();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const linear_congruential_engine& __x,
+                    const linear_congruential_engine& __y)
+        {return __x.__x_ == __y.__x_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const linear_congruential_engine& __x,
+                    const linear_congruential_engine& __y)
+        {return !(__x == __y);}
+
+private:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(true_type, true_type, result_type __s) {__x_ = __s == 0 ? 1 : __s;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(true_type, false_type, result_type __s) {__x_ = __s;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(false_type, true_type, result_type __s) {__x_ = __s % __m == 0 ?
+                                                                 1 : __s % __m;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(false_type, false_type, result_type __s) {__x_ = __s % __m;}
+
+    template<class _Sseq>
+        void __seed(_Sseq& __q, integral_constant<unsigned, 1>);
+    template<class _Sseq>
+        void __seed(_Sseq& __q, integral_constant<unsigned, 2>);
+
+    template <class _CharT, class _Traits,
+              class _Up, _Up _Ap, _Up _Cp, _Up _Np>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const linear_congruential_engine<_Up, _Ap, _Cp, _Np>&);
+
+    template <class _CharT, class _Traits,
+              class _Up, _Up _Ap, _Up _Cp, _Up _Np>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               linear_congruential_engine<_Up, _Ap, _Cp, _Np>& __x);
+};
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+    _LIBCUDACXX_CONSTEXPR const typename linear_congruential_engine<_UIntType, __a, __c, __m>::result_type
+    linear_congruential_engine<_UIntType, __a, __c, __m>::multiplier;
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+    _LIBCUDACXX_CONSTEXPR const typename linear_congruential_engine<_UIntType, __a, __c, __m>::result_type
+    linear_congruential_engine<_UIntType, __a, __c, __m>::increment;
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+    _LIBCUDACXX_CONSTEXPR const typename linear_congruential_engine<_UIntType, __a, __c, __m>::result_type
+    linear_congruential_engine<_UIntType, __a, __c, __m>::modulus;
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+    _LIBCUDACXX_CONSTEXPR const typename linear_congruential_engine<_UIntType, __a, __c, __m>::result_type
+    linear_congruential_engine<_UIntType, __a, __c, __m>::default_seed;
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+template<class _Sseq>
+void
+linear_congruential_engine<_UIntType, __a, __c, __m>::__seed(_Sseq& __q,
+                                                 integral_constant<unsigned, 1>)
+{
+    const unsigned __k = 1;
+    uint32_t __ar[__k+3];
+    __q.generate(__ar, __ar + __k + 3);
+    result_type __s = static_cast<result_type>(__ar[3] % __m);
+    __x_ = __c == 0 && __s == 0 ? result_type(1) : __s;
+}
+
+template <class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+template<class _Sseq>
+void
+linear_congruential_engine<_UIntType, __a, __c, __m>::__seed(_Sseq& __q,
+                                                 integral_constant<unsigned, 2>)
+{
+    const unsigned __k = 2;
+    uint32_t __ar[__k+3];
+    __q.generate(__ar, __ar + __k + 3);
+    result_type __s = static_cast<result_type>((__ar[3] +
+                                              ((uint64_t)__ar[4] << 32)) % __m);
+    __x_ = __c == 0 && __s == 0 ? result_type(1) : __s;
+}
+
+template <class _CharT, class _Traits,
+          class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const linear_congruential_engine<_UIntType, __a, __c, __m>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left);
+    __os.fill(__os.widen(' '));
+    return __os << __x.__x_;
+}
+
+template <class _CharT, class _Traits,
+          class _UIntType, _UIntType __a, _UIntType __c, _UIntType __m>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           linear_congruential_engine<_UIntType, __a, __c, __m>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    _UIntType __t;
+    __is >> __t;
+    if (!__is.fail())
+        __x.__x_ = __t;
+    return __is;
+}
+
+typedef linear_congruential_engine<uint_fast32_t, 16807, 0, 2147483647>
+                                                                   minstd_rand0;
+typedef linear_congruential_engine<uint_fast32_t, 48271, 0, 2147483647>
+                                                                    minstd_rand;
+typedef minstd_rand                                       default_random_engine;
+// mersenne_twister_engine
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+class _LIBCUDACXX_TEMPLATE_VIS mersenne_twister_engine;
+
+template <class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+bool
+operator==(const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __x,
+           const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __y);
+
+template <class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __x,
+           const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __y);
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __x);
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                   _Bp, _Tp, _Cp, _Lp, _Fp>& __x);
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+class _LIBCUDACXX_TEMPLATE_VIS mersenne_twister_engine
+{
+public:
+    // types
+    typedef _UIntType result_type;
+
+private:
+    result_type __x_[__n];
+    size_t      __i_;
+
+    static_assert(  0 <  __m, "mersenne_twister_engine invalid parameters");
+    static_assert(__m <= __n, "mersenne_twister_engine invalid parameters");
+    static _LIBCUDACXX_CONSTEXPR const result_type _Dt = numeric_limits<result_type>::digits;
+    static_assert(__w <= _Dt, "mersenne_twister_engine invalid parameters");
+    static_assert(  2 <= __w, "mersenne_twister_engine invalid parameters");
+    static_assert(__r <= __w, "mersenne_twister_engine invalid parameters");
+    static_assert(__u <= __w, "mersenne_twister_engine invalid parameters");
+    static_assert(__s <= __w, "mersenne_twister_engine invalid parameters");
+    static_assert(__t <= __w, "mersenne_twister_engine invalid parameters");
+    static_assert(__l <= __w, "mersenne_twister_engine invalid parameters");
+public:
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = 0;
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = __w == _Dt ? result_type(~0) :
+                                                      (result_type(1) << __w) - result_type(1);
+    static_assert(_Min < _Max, "mersenne_twister_engine invalid parameters");
+    static_assert(__a <= _Max, "mersenne_twister_engine invalid parameters");
+    static_assert(__b <= _Max, "mersenne_twister_engine invalid parameters");
+    static_assert(__c <= _Max, "mersenne_twister_engine invalid parameters");
+    static_assert(__d <= _Max, "mersenne_twister_engine invalid parameters");
+    static_assert(__f <= _Max, "mersenne_twister_engine invalid parameters");
+
+    // engine characteristics
+    static _LIBCUDACXX_CONSTEXPR const size_t word_size = __w;
+    static _LIBCUDACXX_CONSTEXPR const size_t state_size = __n;
+    static _LIBCUDACXX_CONSTEXPR const size_t shift_size = __m;
+    static _LIBCUDACXX_CONSTEXPR const size_t mask_bits = __r;
+    static _LIBCUDACXX_CONSTEXPR const result_type xor_mask = __a;
+    static _LIBCUDACXX_CONSTEXPR const size_t tempering_u = __u;
+    static _LIBCUDACXX_CONSTEXPR const result_type tempering_d = __d;
+    static _LIBCUDACXX_CONSTEXPR const size_t tempering_s = __s;
+    static _LIBCUDACXX_CONSTEXPR const result_type tempering_b = __b;
+    static _LIBCUDACXX_CONSTEXPR const size_t tempering_t = __t;
+    static _LIBCUDACXX_CONSTEXPR const result_type tempering_c = __c;
+    static _LIBCUDACXX_CONSTEXPR const size_t tempering_l = __l;
+    static _LIBCUDACXX_CONSTEXPR const result_type initialization_multiplier = __f;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() { return _Min; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() { return _Max; }
+    static _LIBCUDACXX_CONSTEXPR const result_type default_seed = 5489u;
+
+    // constructors and seeding functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit mersenne_twister_engine(result_type __sd = default_seed)
+        {seed(__sd);}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit mersenne_twister_engine(_Sseq& __q,
+        typename enable_if<__is_seed_sequence<_Sseq, mersenne_twister_engine>::value>::type* = 0)
+        {seed(__q);}
+    void seed(result_type __sd = default_seed);
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_seed_sequence<_Sseq, mersenne_twister_engine>::value,
+            void
+        >::type
+        seed(_Sseq& __q)
+            {__seed(__q, integral_constant<unsigned, 1 + (__w - 1) / 32>());}
+
+    // generating functions
+    result_type operator()();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void discard(unsigned long long __z) {for (; __z; --__z) operator()();}
+
+    template <class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+              _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+              _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+    friend
+    bool
+    operator==(const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                             _Bp, _Tp, _Cp, _Lp, _Fp>& __x,
+               const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                             _Bp, _Tp, _Cp, _Lp, _Fp>& __y);
+
+    template <class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+              _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+              _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+    friend
+    bool
+    operator!=(const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                             _Bp, _Tp, _Cp, _Lp, _Fp>& __x,
+               const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                             _Bp, _Tp, _Cp, _Lp, _Fp>& __y);
+
+    template <class _CharT, class _Traits,
+              class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+              _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+              _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                             _Bp, _Tp, _Cp, _Lp, _Fp>& __x);
+
+    template <class _CharT, class _Traits,
+              class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+              _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+              _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                       _Bp, _Tp, _Cp, _Lp, _Fp>& __x);
+private:
+
+    template<class _Sseq>
+        void __seed(_Sseq& __q, integral_constant<unsigned, 1>);
+    template<class _Sseq>
+        void __seed(_Sseq& __q, integral_constant<unsigned, 2>);
+
+    template <size_t __count>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            __count < __w,
+            result_type
+        >::type
+        __lshift(result_type __x) {return (__x << __count) & _Max;}
+
+    template <size_t __count>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            (__count >= __w),
+            result_type
+        >::type
+        __lshift(result_type) {return result_type(0);}
+
+    template <size_t __count>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            __count < _Dt,
+            result_type
+        >::type
+        __rshift(result_type __x) {return __x >> __count;}
+
+    template <size_t __count>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            (__count >= _Dt),
+            result_type
+        >::type
+        __rshift(result_type) {return result_type(0);}
+};
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::word_size;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::state_size;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::shift_size;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::mask_bits;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const typename mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::result_type
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::xor_mask;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_u;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const typename mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::result_type
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_d;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_s;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const typename mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::result_type
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_b;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_t;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const typename mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::result_type
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_c;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const size_t
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::tempering_l;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const typename mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::result_type
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::initialization_multiplier;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+    _LIBCUDACXX_CONSTEXPR const typename mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::result_type
+    mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b, __t, __c, __l, __f>::default_seed;
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+void
+mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b,
+    __t, __c, __l, __f>::seed(result_type __sd)
+    _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+{   // __w >= 2
+    __x_[0] = __sd & _Max;
+    for (size_t __i = 1; __i < __n; ++__i)
+        __x_[__i] = (__f * (__x_[__i-1] ^ __rshift<__w - 2>(__x_[__i-1])) + __i) & _Max;
+    __i_ = 0;
+}
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+template<class _Sseq>
+void
+mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b,
+    __t, __c, __l, __f>::__seed(_Sseq& __q, integral_constant<unsigned, 1>)
+{
+    const unsigned __k = 1;
+    uint32_t __ar[__n * __k];
+    __q.generate(__ar, __ar + __n * __k);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __x_[__i] = static_cast<result_type>(__ar[__i] & _Max);
+    const result_type __mask = __r == _Dt ? result_type(~0) :
+                                       (result_type(1) << __r) - result_type(1);
+    __i_ = 0;
+    if ((__x_[0] & ~__mask) == 0)
+    {
+        for (size_t __i = 1; __i < __n; ++__i)
+            if (__x_[__i] != 0)
+                return;
+        __x_[0] = result_type(1) << (__w - 1);
+    }
+}
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+template<class _Sseq>
+void
+mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b,
+    __t, __c, __l, __f>::__seed(_Sseq& __q, integral_constant<unsigned, 2>)
+{
+    const unsigned __k = 2;
+    uint32_t __ar[__n * __k];
+    __q.generate(__ar, __ar + __n * __k);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __x_[__i] = static_cast<result_type>(
+            (__ar[2 * __i] + ((uint64_t)__ar[2 * __i + 1] << 32)) & _Max);
+    const result_type __mask = __r == _Dt ? result_type(~0) :
+                                       (result_type(1) << __r) - result_type(1);
+    __i_ = 0;
+    if ((__x_[0] & ~__mask) == 0)
+    {
+        for (size_t __i = 1; __i < __n; ++__i)
+            if (__x_[__i] != 0)
+                return;
+        __x_[0] = result_type(1) << (__w - 1);
+    }
+}
+
+template <class _UIntType, size_t __w, size_t __n, size_t __m, size_t __r,
+          _UIntType __a, size_t __u, _UIntType __d, size_t __s,
+          _UIntType __b, size_t __t, _UIntType __c, size_t __l, _UIntType __f>
+_UIntType
+mersenne_twister_engine<_UIntType, __w, __n, __m, __r, __a, __u, __d, __s, __b,
+    __t, __c, __l, __f>::operator()()
+{
+    const size_t __j = (__i_ + 1) % __n;
+    const result_type __mask = __r == _Dt ? result_type(~0) :
+                                       (result_type(1) << __r) - result_type(1);
+    const result_type _Yp = (__x_[__i_] & ~__mask) | (__x_[__j] & __mask);
+    const size_t __k = (__i_ + __m) % __n;
+    __x_[__i_] = __x_[__k] ^ __rshift<1>(_Yp) ^ (__a * (_Yp & 1));
+    result_type __z = __x_[__i_] ^ (__rshift<__u>(__x_[__i_]) & __d);
+    __i_ = __j;
+    __z ^= __lshift<__s>(__z) & __b;
+    __z ^= __lshift<__t>(__z) & __c;
+    return __z ^ __rshift<__l>(__z);
+}
+
+template <class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+bool
+operator==(const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __x,
+           const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __y)
+{
+    if (__x.__i_ == __y.__i_)
+        return _CUDA_VSTD::equal(__x.__x_, __x.__x_ + _Np, __y.__x_);
+    if (__x.__i_ == 0 || __y.__i_ == 0)
+    {
+        size_t __j = _CUDA_VSTD::min(_Np - __x.__i_, _Np - __y.__i_);
+        if (!_CUDA_VSTD::equal(__x.__x_ + __x.__i_, __x.__x_ + __x.__i_ + __j,
+                         __y.__x_ + __y.__i_))
+            return false;
+        if (__x.__i_ == 0)
+            return _CUDA_VSTD::equal(__x.__x_ + __j, __x.__x_ + _Np, __y.__x_);
+        return _CUDA_VSTD::equal(__x.__x_, __x.__x_ + (_Np - __j), __y.__x_ + __j);
+    }
+    if (__x.__i_ < __y.__i_)
+    {
+        size_t __j = _Np - __y.__i_;
+        if (!_CUDA_VSTD::equal(__x.__x_ + __x.__i_, __x.__x_ + (__x.__i_ + __j),
+                         __y.__x_ + __y.__i_))
+            return false;
+        if (!_CUDA_VSTD::equal(__x.__x_ + (__x.__i_ + __j), __x.__x_ + _Np,
+                         __y.__x_))
+            return false;
+        return _CUDA_VSTD::equal(__x.__x_, __x.__x_ + __x.__i_,
+                           __y.__x_ + (_Np - (__x.__i_ + __j)));
+    }
+    size_t __j = _Np - __x.__i_;
+    if (!_CUDA_VSTD::equal(__y.__x_ + __y.__i_, __y.__x_ + (__y.__i_ + __j),
+                     __x.__x_ + __x.__i_))
+        return false;
+    if (!_CUDA_VSTD::equal(__y.__x_ + (__y.__i_ + __j), __y.__x_ + _Np,
+                     __x.__x_))
+        return false;
+    return _CUDA_VSTD::equal(__y.__x_, __y.__x_ + __y.__i_,
+                       __x.__x_ + (_Np - (__y.__i_ + __j)));
+}
+
+template <class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __x,
+           const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                         _Bp, _Tp, _Cp, _Lp, _Fp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.__x_[__x.__i_];
+    for (size_t __j = __x.__i_ + 1; __j < _Np; ++__j)
+        __os << __sp << __x.__x_[__j];
+    for (size_t __j = 0; __j < __x.__i_; ++__j)
+        __os << __sp << __x.__x_[__j];
+    return __os;
+}
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Np, size_t _Mp, size_t _Rp,
+          _UInt _Ap, size_t _Up, _UInt _Dp, size_t _Sp,
+          _UInt _Bp, size_t _Tp, _UInt _Cp, size_t _Lp, _UInt _Fp>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           mersenne_twister_engine<_UInt, _Wp, _Np, _Mp, _Rp, _Ap, _Up, _Dp, _Sp,
+                                   _Bp, _Tp, _Cp, _Lp, _Fp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    _UInt __t[_Np];
+    for (size_t __i = 0; __i < _Np; ++__i)
+        __is >> __t[__i];
+    if (!__is.fail())
+    {
+        for (size_t __i = 0; __i < _Np; ++__i)
+            __x.__x_[__i] = __t[__i];
+        __x.__i_ = 0;
+    }
+    return __is;
+}
+
+typedef mersenne_twister_engine<uint_fast32_t, 32, 624, 397, 31,
+                                0x9908b0df, 11, 0xffffffff,
+                                7,  0x9d2c5680,
+                                15, 0xefc60000,
+                                18, 1812433253>                         mt19937;
+typedef mersenne_twister_engine<uint_fast64_t, 64, 312, 156, 31,
+                                0xb5026f5aa96619e9ULL, 29, 0x5555555555555555ULL,
+                                17, 0x71d67fffeda60000ULL,
+                                37, 0xfff7eee000000000ULL,
+                                43, 6364136223846793005ULL>          mt19937_64;
+
+// subtract_with_carry_engine
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+class _LIBCUDACXX_TEMPLATE_VIS subtract_with_carry_engine;
+
+template<class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+bool
+operator==(
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x,
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __y);
+
+template<class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x,
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __y);
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x);
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x);
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+class _LIBCUDACXX_TEMPLATE_VIS subtract_with_carry_engine
+{
+public:
+    // types
+    typedef _UIntType result_type;
+
+private:
+    result_type __x_[__r];
+    result_type  __c_;
+    size_t      __i_;
+
+    static _LIBCUDACXX_CONSTEXPR const result_type _Dt = numeric_limits<result_type>::digits;
+    static_assert(  0 <  __w, "subtract_with_carry_engine invalid parameters");
+    static_assert(__w <= _Dt, "subtract_with_carry_engine invalid parameters");
+    static_assert(  0 <  __s, "subtract_with_carry_engine invalid parameters");
+    static_assert(__s <  __r, "subtract_with_carry_engine invalid parameters");
+public:
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = 0;
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = __w == _Dt ? result_type(~0) :
+                                                      (result_type(1) << __w) - result_type(1);
+    static_assert(_Min < _Max, "subtract_with_carry_engine invalid parameters");
+
+    // engine characteristics
+    static _LIBCUDACXX_CONSTEXPR const size_t word_size = __w;
+    static _LIBCUDACXX_CONSTEXPR const size_t short_lag = __s;
+    static _LIBCUDACXX_CONSTEXPR const size_t long_lag = __r;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() { return _Min; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() { return _Max; }
+    static _LIBCUDACXX_CONSTEXPR const result_type default_seed = 19780503u;
+
+    // constructors and seeding functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit subtract_with_carry_engine(result_type __sd = default_seed)
+        {seed(__sd);}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit subtract_with_carry_engine(_Sseq& __q,
+        typename enable_if<__is_seed_sequence<_Sseq, subtract_with_carry_engine>::value>::type* = 0)
+        {seed(__q);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(result_type __sd = default_seed)
+        {seed(__sd, integral_constant<unsigned, 1 + (__w - 1) / 32>());}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_seed_sequence<_Sseq, subtract_with_carry_engine>::value,
+            void
+        >::type
+        seed(_Sseq& __q)
+            {__seed(__q, integral_constant<unsigned, 1 + (__w - 1) / 32>());}
+
+    // generating functions
+    result_type operator()();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void discard(unsigned long long __z) {for (; __z; --__z) operator()();}
+
+    template<class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+    friend
+    bool
+    operator==(
+        const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x,
+        const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __y);
+
+    template<class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+    friend
+    bool
+    operator!=(
+        const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x,
+        const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __y);
+
+    template <class _CharT, class _Traits,
+              class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x);
+
+    template <class _CharT, class _Traits,
+              class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x);
+
+private:
+
+    void seed(result_type __sd, integral_constant<unsigned, 1>);
+    void seed(result_type __sd, integral_constant<unsigned, 2>);
+    template<class _Sseq>
+        void __seed(_Sseq& __q, integral_constant<unsigned, 1>);
+    template<class _Sseq>
+        void __seed(_Sseq& __q, integral_constant<unsigned, 2>);
+};
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+    _LIBCUDACXX_CONSTEXPR const size_t subtract_with_carry_engine<_UIntType, __w, __s, __r>::word_size;
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+    _LIBCUDACXX_CONSTEXPR const size_t subtract_with_carry_engine<_UIntType, __w, __s, __r>::short_lag;
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+    _LIBCUDACXX_CONSTEXPR const size_t subtract_with_carry_engine<_UIntType, __w, __s, __r>::long_lag;
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+    _LIBCUDACXX_CONSTEXPR const typename subtract_with_carry_engine<_UIntType, __w, __s, __r>::result_type
+    subtract_with_carry_engine<_UIntType, __w, __s, __r>::default_seed;
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+void
+subtract_with_carry_engine<_UIntType, __w, __s, __r>::seed(result_type __sd,
+        integral_constant<unsigned, 1>)
+{
+    linear_congruential_engine<result_type, 40014u, 0u, 2147483563u>
+        __e(__sd == 0u ? default_seed : __sd);
+    for (size_t __i = 0; __i < __r; ++__i)
+        __x_[__i] = static_cast<result_type>(__e() & _Max);
+    __c_ = __x_[__r-1] == 0;
+    __i_ = 0;
+}
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+void
+subtract_with_carry_engine<_UIntType, __w, __s, __r>::seed(result_type __sd,
+        integral_constant<unsigned, 2>)
+{
+    linear_congruential_engine<result_type, 40014u, 0u, 2147483563u>
+        __e(__sd == 0u ? default_seed : __sd);
+    for (size_t __i = 0; __i < __r; ++__i)
+    {
+        result_type __e0 = __e();
+        __x_[__i] = static_cast<result_type>(
+                                    (__e0 + ((uint64_t)__e() << 32)) & _Max);
+    }
+    __c_ = __x_[__r-1] == 0;
+    __i_ = 0;
+}
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+template<class _Sseq>
+void
+subtract_with_carry_engine<_UIntType, __w, __s, __r>::__seed(_Sseq& __q,
+        integral_constant<unsigned, 1>)
+{
+    const unsigned __k = 1;
+    uint32_t __ar[__r * __k];
+    __q.generate(__ar, __ar + __r * __k);
+    for (size_t __i = 0; __i < __r; ++__i)
+        __x_[__i] = static_cast<result_type>(__ar[__i] & _Max);
+    __c_ = __x_[__r-1] == 0;
+    __i_ = 0;
+}
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+template<class _Sseq>
+void
+subtract_with_carry_engine<_UIntType, __w, __s, __r>::__seed(_Sseq& __q,
+        integral_constant<unsigned, 2>)
+{
+    const unsigned __k = 2;
+    uint32_t __ar[__r * __k];
+    __q.generate(__ar, __ar + __r * __k);
+    for (size_t __i = 0; __i < __r; ++__i)
+        __x_[__i] = static_cast<result_type>(
+                  (__ar[2 * __i] + ((uint64_t)__ar[2 * __i + 1] << 32)) & _Max);
+    __c_ = __x_[__r-1] == 0;
+    __i_ = 0;
+}
+
+template<class _UIntType, size_t __w, size_t __s, size_t __r>
+_UIntType
+subtract_with_carry_engine<_UIntType, __w, __s, __r>::operator()()
+{
+    const result_type& __xs = __x_[(__i_ + (__r - __s)) % __r];
+    result_type& __xr = __x_[__i_];
+    result_type __new_c = __c_ == 0 ? __xs < __xr : __xs != 0 ? __xs <= __xr : 1;
+    __xr = (__xs - __xr - __c_) & _Max;
+    __c_ = __new_c;
+    __i_ = (__i_ + 1) % __r;
+    return __xr;
+}
+
+template<class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+bool
+operator==(
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x,
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __y)
+{
+    if (__x.__c_ != __y.__c_)
+        return false;
+    if (__x.__i_ == __y.__i_)
+        return _CUDA_VSTD::equal(__x.__x_, __x.__x_ + _Rp, __y.__x_);
+    if (__x.__i_ == 0 || __y.__i_ == 0)
+    {
+        size_t __j = _CUDA_VSTD::min(_Rp - __x.__i_, _Rp - __y.__i_);
+        if (!_CUDA_VSTD::equal(__x.__x_ + __x.__i_, __x.__x_ + __x.__i_ + __j,
+                         __y.__x_ + __y.__i_))
+            return false;
+        if (__x.__i_ == 0)
+            return _CUDA_VSTD::equal(__x.__x_ + __j, __x.__x_ + _Rp, __y.__x_);
+        return _CUDA_VSTD::equal(__x.__x_, __x.__x_ + (_Rp - __j), __y.__x_ + __j);
+    }
+    if (__x.__i_ < __y.__i_)
+    {
+        size_t __j = _Rp - __y.__i_;
+        if (!_CUDA_VSTD::equal(__x.__x_ + __x.__i_, __x.__x_ + (__x.__i_ + __j),
+                         __y.__x_ + __y.__i_))
+            return false;
+        if (!_CUDA_VSTD::equal(__x.__x_ + (__x.__i_ + __j), __x.__x_ + _Rp,
+                         __y.__x_))
+            return false;
+        return _CUDA_VSTD::equal(__x.__x_, __x.__x_ + __x.__i_,
+                           __y.__x_ + (_Rp - (__x.__i_ + __j)));
+    }
+    size_t __j = _Rp - __x.__i_;
+    if (!_CUDA_VSTD::equal(__y.__x_ + __y.__i_, __y.__x_ + (__y.__i_ + __j),
+                     __x.__x_ + __x.__i_))
+        return false;
+    if (!_CUDA_VSTD::equal(__y.__x_ + (__y.__i_ + __j), __y.__x_ + _Rp,
+                     __x.__x_))
+        return false;
+    return _CUDA_VSTD::equal(__y.__x_, __y.__x_ + __y.__i_,
+                       __x.__x_ + (_Rp - (__y.__i_ + __j)));
+}
+
+template<class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x,
+    const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.__x_[__x.__i_];
+    for (size_t __j = __x.__i_ + 1; __j < _Rp; ++__j)
+        __os << __sp << __x.__x_[__j];
+    for (size_t __j = 0; __j < __x.__i_; ++__j)
+        __os << __sp << __x.__x_[__j];
+    __os << __sp << __x.__c_;
+    return __os;
+}
+
+template <class _CharT, class _Traits,
+          class _UInt, size_t _Wp, size_t _Sp, size_t _Rp>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           subtract_with_carry_engine<_UInt, _Wp, _Sp, _Rp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    _UInt __t[_Rp+1];
+    for (size_t __i = 0; __i < _Rp+1; ++__i)
+        __is >> __t[__i];
+    if (!__is.fail())
+    {
+        for (size_t __i = 0; __i < _Rp; ++__i)
+            __x.__x_[__i] = __t[__i];
+        __x.__c_ = __t[_Rp];
+        __x.__i_ = 0;
+    }
+    return __is;
+}
+
+typedef subtract_with_carry_engine<uint_fast32_t, 24, 10, 24>     ranlux24_base;
+typedef subtract_with_carry_engine<uint_fast64_t, 48,  5, 12>     ranlux48_base;
+
+// discard_block_engine
+
+template<class _Engine, size_t __p, size_t __r>
+class _LIBCUDACXX_TEMPLATE_VIS discard_block_engine
+{
+    _Engine __e_;
+    int     __n_;
+
+    static_assert(  0 <  __r, "discard_block_engine invalid parameters");
+    static_assert(__r <= __p, "discard_block_engine invalid parameters");
+    static_assert(__r <= INT_MAX, "discard_block_engine invalid parameters");
+public:
+    // types
+    typedef typename _Engine::result_type result_type;
+
+    // engine characteristics
+    static _LIBCUDACXX_CONSTEXPR const size_t block_size = __p;
+    static _LIBCUDACXX_CONSTEXPR const size_t used_block = __r;
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    static const result_type _Min = _Engine::_Min;
+    static const result_type _Max = _Engine::_Max;
+#else
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = _Engine::min();
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = _Engine::max();
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() { return _Engine::min(); }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() { return _Engine::max(); }
+
+    // constructors and seeding functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    discard_block_engine() : __n_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit discard_block_engine(const _Engine& __e)
+        : __e_(__e), __n_(0) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit discard_block_engine(_Engine&& __e)
+        : __e_(_CUDA_VSTD::move(__e)), __n_(0) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit discard_block_engine(result_type __sd) : __e_(__sd), __n_(0) {}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit discard_block_engine(_Sseq& __q,
+        typename enable_if<__is_seed_sequence<_Sseq, discard_block_engine>::value &&
+                           !is_convertible<_Sseq, _Engine>::value>::type* = 0)
+        : __e_(__q), __n_(0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed() {__e_.seed(); __n_ = 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(result_type __sd) {__e_.seed(__sd); __n_ = 0;}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_seed_sequence<_Sseq, discard_block_engine>::value,
+            void
+        >::type
+        seed(_Sseq& __q) {__e_.seed(__q); __n_ = 0;}
+
+    // generating functions
+    result_type operator()();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void discard(unsigned long long __z) {for (; __z; --__z) operator()();}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Engine& base() const _NOEXCEPT {return __e_;}
+
+    template<class _Eng, size_t _Pp, size_t _Rp>
+    friend
+    bool
+    operator==(
+        const discard_block_engine<_Eng, _Pp, _Rp>& __x,
+        const discard_block_engine<_Eng, _Pp, _Rp>& __y);
+
+    template<class _Eng, size_t _Pp, size_t _Rp>
+    friend
+    bool
+    operator!=(
+        const discard_block_engine<_Eng, _Pp, _Rp>& __x,
+        const discard_block_engine<_Eng, _Pp, _Rp>& __y);
+
+    template <class _CharT, class _Traits,
+              class _Eng, size_t _Pp, size_t _Rp>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const discard_block_engine<_Eng, _Pp, _Rp>& __x);
+
+    template <class _CharT, class _Traits,
+              class _Eng, size_t _Pp, size_t _Rp>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               discard_block_engine<_Eng, _Pp, _Rp>& __x);
+};
+
+template<class _Engine, size_t __p, size_t __r>
+    _LIBCUDACXX_CONSTEXPR const size_t discard_block_engine<_Engine, __p, __r>::block_size;
+
+template<class _Engine, size_t __p, size_t __r>
+    _LIBCUDACXX_CONSTEXPR const size_t discard_block_engine<_Engine, __p, __r>::used_block;
+
+template<class _Engine, size_t __p, size_t __r>
+typename discard_block_engine<_Engine, __p, __r>::result_type
+discard_block_engine<_Engine, __p, __r>::operator()()
+{
+    if (__n_ >= static_cast<int>(__r))
+    {
+        __e_.discard(__p - __r);
+        __n_ = 0;
+    }
+    ++__n_;
+    return __e_();
+}
+
+template<class _Eng, size_t _Pp, size_t _Rp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const discard_block_engine<_Eng, _Pp, _Rp>& __x,
+           const discard_block_engine<_Eng, _Pp, _Rp>& __y)
+{
+    return __x.__n_ == __y.__n_ && __x.__e_ == __y.__e_;
+}
+
+template<class _Eng, size_t _Pp, size_t _Rp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const discard_block_engine<_Eng, _Pp, _Rp>& __x,
+           const discard_block_engine<_Eng, _Pp, _Rp>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _CharT, class _Traits,
+          class _Eng, size_t _Pp, size_t _Rp>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const discard_block_engine<_Eng, _Pp, _Rp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    return __os << __x.__e_ << __sp << __x.__n_;
+}
+
+template <class _CharT, class _Traits,
+          class _Eng, size_t _Pp, size_t _Rp>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           discard_block_engine<_Eng, _Pp, _Rp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    _Eng __e;
+    int __n;
+    __is >> __e >> __n;
+    if (!__is.fail())
+    {
+        __x.__e_ = __e;
+        __x.__n_ = __n;
+    }
+    return __is;
+}
+
+typedef discard_block_engine<ranlux24_base, 223, 23> ranlux24;
+typedef discard_block_engine<ranlux48_base, 389, 11> ranlux48;
+
+// independent_bits_engine
+
+template<class _Engine, size_t __w, class _UIntType>
+class _LIBCUDACXX_TEMPLATE_VIS independent_bits_engine
+{
+    template <class _UInt, _UInt _R0, size_t _Wp, size_t _Mp>
+    class __get_n
+    {
+        static _LIBCUDACXX_CONSTEXPR const size_t _Dt = numeric_limits<_UInt>::digits;
+        static _LIBCUDACXX_CONSTEXPR const size_t _Np = _Wp / _Mp + (_Wp % _Mp != 0);
+        static _LIBCUDACXX_CONSTEXPR const size_t _W0 = _Wp / _Np;
+        static _LIBCUDACXX_CONSTEXPR const _UInt _Y0 = _W0 >= _Dt ? 0 : (_R0 >> _W0) << _W0;
+    public:
+        static _LIBCUDACXX_CONSTEXPR const size_t value = _R0 - _Y0 > _Y0 / _Np ? _Np + 1 : _Np;
+    };
+public:
+    // types
+    typedef _UIntType result_type;
+
+private:
+    _Engine __e_;
+
+    static _LIBCUDACXX_CONSTEXPR const result_type _Dt = numeric_limits<result_type>::digits;
+    static_assert(  0 <  __w, "independent_bits_engine invalid parameters");
+    static_assert(__w <= _Dt, "independent_bits_engine invalid parameters");
+
+    typedef typename _Engine::result_type _Engine_result_type;
+    typedef typename conditional
+        <
+            sizeof(_Engine_result_type) <= sizeof(result_type),
+                result_type,
+                _Engine_result_type
+        >::type _Working_result_type;
+#ifdef _LIBCUDACXX_CXX03_LANG
+    static const _Working_result_type _Rp = _Engine::_Max - _Engine::_Min
+                                          + _Working_result_type(1);
+#else
+    static _LIBCUDACXX_CONSTEXPR const _Working_result_type _Rp = _Engine::max() - _Engine::min()
+                                                            + _Working_result_type(1);
+#endif
+    static _LIBCUDACXX_CONSTEXPR const size_t __m = __log2<_Working_result_type, _Rp>::value;
+    static _LIBCUDACXX_CONSTEXPR const size_t __n = __get_n<_Working_result_type, _Rp, __w, __m>::value;
+    static _LIBCUDACXX_CONSTEXPR const size_t __w0 = __w / __n;
+    static _LIBCUDACXX_CONSTEXPR const size_t __n0 = __n - __w % __n;
+    static _LIBCUDACXX_CONSTEXPR const size_t _WDt = numeric_limits<_Working_result_type>::digits;
+    static _LIBCUDACXX_CONSTEXPR const size_t _EDt = numeric_limits<_Engine_result_type>::digits;
+    static _LIBCUDACXX_CONSTEXPR const _Working_result_type __y0 = __w0 >= _WDt ? 0 :
+                                                               (_Rp >> __w0) << __w0;
+    static _LIBCUDACXX_CONSTEXPR const _Working_result_type __y1 = __w0 >= _WDt - 1 ? 0 :
+                                                               (_Rp >> (__w0+1)) << (__w0+1);
+    static _LIBCUDACXX_CONSTEXPR const _Engine_result_type __mask0 = __w0 > 0 ?
+                                _Engine_result_type(~0) >> (_EDt - __w0) :
+                                _Engine_result_type(0);
+    static _LIBCUDACXX_CONSTEXPR const _Engine_result_type __mask1 = __w0 < _EDt - 1 ?
+                                _Engine_result_type(~0) >> (_EDt - (__w0 + 1)) :
+                                _Engine_result_type(~0);
+public:
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = 0;
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = __w == _Dt ? result_type(~0) :
+                                                      (result_type(1) << __w) - result_type(1);
+    static_assert(_Min < _Max, "independent_bits_engine invalid parameters");
+
+    // engine characteristics
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() { return _Min; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() { return _Max; }
+
+    // constructors and seeding functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    independent_bits_engine() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit independent_bits_engine(const _Engine& __e)
+        : __e_(__e) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit independent_bits_engine(_Engine&& __e)
+        : __e_(_CUDA_VSTD::move(__e)) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit independent_bits_engine(result_type __sd) : __e_(__sd) {}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit independent_bits_engine(_Sseq& __q,
+        typename enable_if<__is_seed_sequence<_Sseq, independent_bits_engine>::value &&
+                           !is_convertible<_Sseq, _Engine>::value>::type* = 0)
+         : __e_(__q) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed() {__e_.seed();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(result_type __sd) {__e_.seed(__sd);}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_seed_sequence<_Sseq, independent_bits_engine>::value,
+            void
+        >::type
+        seed(_Sseq& __q) {__e_.seed(__q);}
+
+    // generating functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator()() {return __eval(integral_constant<bool, _Rp != 0>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void discard(unsigned long long __z) {for (; __z; --__z) operator()();}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Engine& base() const _NOEXCEPT {return __e_;}
+
+    template<class _Eng, size_t _Wp, class _UInt>
+    friend
+    bool
+    operator==(
+        const independent_bits_engine<_Eng, _Wp, _UInt>& __x,
+        const independent_bits_engine<_Eng, _Wp, _UInt>& __y);
+
+    template<class _Eng, size_t _Wp, class _UInt>
+    friend
+    bool
+    operator!=(
+        const independent_bits_engine<_Eng, _Wp, _UInt>& __x,
+        const independent_bits_engine<_Eng, _Wp, _UInt>& __y);
+
+    template <class _CharT, class _Traits,
+              class _Eng, size_t _Wp, class _UInt>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const independent_bits_engine<_Eng, _Wp, _UInt>& __x);
+
+    template <class _CharT, class _Traits,
+              class _Eng, size_t _Wp, class _UInt>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               independent_bits_engine<_Eng, _Wp, _UInt>& __x);
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type __eval(false_type);
+    result_type __eval(true_type);
+
+    template <size_t __count>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            __count < _Dt,
+            result_type
+        >::type
+        __lshift(result_type __x) {return __x << __count;}
+
+    template <size_t __count>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static
+        typename enable_if
+        <
+            (__count >= _Dt),
+            result_type
+        >::type
+        __lshift(result_type) {return result_type(0);}
+};
+
+template<class _Engine, size_t __w, class _UIntType>
+inline
+_UIntType
+independent_bits_engine<_Engine, __w, _UIntType>::__eval(false_type)
+{
+    return static_cast<result_type>(__e_() & __mask0);
+}
+
+template<class _Engine, size_t __w, class _UIntType>
+_UIntType
+independent_bits_engine<_Engine, __w, _UIntType>::__eval(true_type)
+{
+    result_type _Sp = 0;
+    for (size_t __k = 0; __k < __n0; ++__k)
+    {
+        _Engine_result_type __u;
+        do
+        {
+            __u = __e_() - _Engine::min();
+        } while (__u >= __y0);
+        _Sp = static_cast<result_type>(__lshift<__w0>(_Sp) + (__u & __mask0));
+    }
+    for (size_t __k = __n0; __k < __n; ++__k)
+    {
+        _Engine_result_type __u;
+        do
+        {
+            __u = __e_() - _Engine::min();
+        } while (__u >= __y1);
+        _Sp = static_cast<result_type>(__lshift<__w0+1>(_Sp) + (__u & __mask1));
+    }
+    return _Sp;
+}
+
+template<class _Eng, size_t _Wp, class _UInt>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(
+    const independent_bits_engine<_Eng, _Wp, _UInt>& __x,
+    const independent_bits_engine<_Eng, _Wp, _UInt>& __y)
+{
+    return __x.base() == __y.base();
+}
+
+template<class _Eng, size_t _Wp, class _UInt>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(
+    const independent_bits_engine<_Eng, _Wp, _UInt>& __x,
+    const independent_bits_engine<_Eng, _Wp, _UInt>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _CharT, class _Traits,
+          class _Eng, size_t _Wp, class _UInt>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const independent_bits_engine<_Eng, _Wp, _UInt>& __x)
+{
+    return __os << __x.base();
+}
+
+template <class _CharT, class _Traits,
+          class _Eng, size_t _Wp, class _UInt>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           independent_bits_engine<_Eng, _Wp, _UInt>& __x)
+{
+    _Eng __e;
+    __is >> __e;
+    if (!__is.fail())
+        __x.__e_ = __e;
+    return __is;
+}
+
+// shuffle_order_engine
+
+template <uint64_t _Xp, uint64_t _Yp>
+struct __ugcd
+{
+    static _LIBCUDACXX_CONSTEXPR const uint64_t value = __ugcd<_Yp, _Xp % _Yp>::value;
+};
+
+template <uint64_t _Xp>
+struct __ugcd<_Xp, 0>
+{
+    static _LIBCUDACXX_CONSTEXPR const uint64_t value = _Xp;
+};
+
+template <uint64_t _Np, uint64_t _Dp>
+class __uratio
+{
+    static_assert(_Dp != 0, "__uratio divide by 0");
+    static _LIBCUDACXX_CONSTEXPR const uint64_t __gcd = __ugcd<_Np, _Dp>::value;
+public:
+    static _LIBCUDACXX_CONSTEXPR const uint64_t num = _Np / __gcd;
+    static _LIBCUDACXX_CONSTEXPR const uint64_t den = _Dp / __gcd;
+
+    typedef __uratio<num, den> type;
+};
+
+template<class _Engine, size_t __k>
+class _LIBCUDACXX_TEMPLATE_VIS shuffle_order_engine
+{
+    static_assert(0 < __k, "shuffle_order_engine invalid parameters");
+public:
+    // types
+    typedef typename _Engine::result_type result_type;
+
+private:
+    _Engine __e_;
+    result_type _V_[__k];
+    result_type _Y_;
+
+public:
+    // engine characteristics
+    static _LIBCUDACXX_CONSTEXPR const size_t table_size = __k;
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    static const result_type _Min = _Engine::_Min;
+    static const result_type _Max = _Engine::_Max;
+#else
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = _Engine::min();
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = _Engine::max();
+#endif
+    static_assert(_Min < _Max, "shuffle_order_engine invalid parameters");
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() { return _Min; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() { return _Max; }
+
+    static _LIBCUDACXX_CONSTEXPR const unsigned long long _Rp = _Max - _Min + 1ull;
+
+    // constructors and seeding functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shuffle_order_engine() {__init();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit shuffle_order_engine(const _Engine& __e)
+        : __e_(__e) {__init();}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit shuffle_order_engine(_Engine&& __e)
+        : __e_(_CUDA_VSTD::move(__e)) {__init();}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit shuffle_order_engine(result_type __sd) : __e_(__sd) {__init();}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit shuffle_order_engine(_Sseq& __q,
+        typename enable_if<__is_seed_sequence<_Sseq, shuffle_order_engine>::value &&
+                           !is_convertible<_Sseq, _Engine>::value>::type* = 0)
+         : __e_(__q) {__init();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed() {__e_.seed(); __init();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void seed(result_type __sd) {__e_.seed(__sd); __init();}
+    template<class _Sseq>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_seed_sequence<_Sseq, shuffle_order_engine>::value,
+            void
+        >::type
+        seed(_Sseq& __q) {__e_.seed(__q); __init();}
+
+    // generating functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator()() {return __eval(integral_constant<bool, _Rp != 0>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void discard(unsigned long long __z) {for (; __z; --__z) operator()();}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Engine& base() const _NOEXCEPT {return __e_;}
+
+private:
+    template<class _Eng, size_t _Kp>
+    friend
+    bool
+    operator==(
+        const shuffle_order_engine<_Eng, _Kp>& __x,
+        const shuffle_order_engine<_Eng, _Kp>& __y);
+
+    template<class _Eng, size_t _Kp>
+    friend
+    bool
+    operator!=(
+        const shuffle_order_engine<_Eng, _Kp>& __x,
+        const shuffle_order_engine<_Eng, _Kp>& __y);
+
+    template <class _CharT, class _Traits,
+              class _Eng, size_t _Kp>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const shuffle_order_engine<_Eng, _Kp>& __x);
+
+    template <class _CharT, class _Traits,
+              class _Eng, size_t _Kp>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               shuffle_order_engine<_Eng, _Kp>& __x);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __init()
+    {
+        for (size_t __i = 0; __i < __k; ++__i)
+            _V_[__i] = __e_();
+        _Y_ = __e_();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type __eval(false_type) {return __eval2(integral_constant<bool, __k & 1>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type __eval(true_type) {return __eval(__uratio<__k, _Rp>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type __eval2(false_type) {return __eval(__uratio<__k/2, 0x8000000000000000ull>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type __eval2(true_type) {return __evalf<__k, 0>();}
+
+    template <uint64_t _Np, uint64_t _Dp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            (__uratio<_Np, _Dp>::num > 0xFFFFFFFFFFFFFFFFull / (_Max - _Min)),
+            result_type
+        >::type
+        __eval(__uratio<_Np, _Dp>)
+            {return __evalf<__uratio<_Np, _Dp>::num, __uratio<_Np, _Dp>::den>();}
+
+    template <uint64_t _Np, uint64_t _Dp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __uratio<_Np, _Dp>::num <= 0xFFFFFFFFFFFFFFFFull / (_Max - _Min),
+            result_type
+        >::type
+        __eval(__uratio<_Np, _Dp>)
+        {
+            const size_t __j = static_cast<size_t>(__uratio<_Np, _Dp>::num * (_Y_ - _Min)
+                                                   / __uratio<_Np, _Dp>::den);
+            _Y_ = _V_[__j];
+            _V_[__j] = __e_();
+            return _Y_;
+        }
+
+    template <uint64_t __n, uint64_t __d>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type __evalf()
+        {
+            const double _Fp = __d == 0 ?
+                __n / (2. * 0x8000000000000000ull) :
+                __n / (double)__d;
+            const size_t __j = static_cast<size_t>(_Fp * (_Y_ - _Min));
+            _Y_ = _V_[__j];
+            _V_[__j] = __e_();
+            return _Y_;
+        }
+};
+
+template<class _Engine, size_t __k>
+    _LIBCUDACXX_CONSTEXPR const size_t shuffle_order_engine<_Engine, __k>::table_size;
+
+template<class _Eng, size_t _Kp>
+bool
+operator==(
+    const shuffle_order_engine<_Eng, _Kp>& __x,
+    const shuffle_order_engine<_Eng, _Kp>& __y)
+{
+    return __x._Y_ == __y._Y_ && _CUDA_VSTD::equal(__x._V_, __x._V_ + _Kp, __y._V_) &&
+           __x.__e_ == __y.__e_;
+}
+
+template<class _Eng, size_t _Kp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(
+    const shuffle_order_engine<_Eng, _Kp>& __x,
+    const shuffle_order_engine<_Eng, _Kp>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _CharT, class _Traits,
+          class _Eng, size_t _Kp>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const shuffle_order_engine<_Eng, _Kp>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.__e_ << __sp << __x._V_[0];
+    for (size_t __i = 1; __i < _Kp; ++__i)
+        __os << __sp << __x._V_[__i];
+    return __os << __sp << __x._Y_;
+}
+
+template <class _CharT, class _Traits,
+          class _Eng, size_t _Kp>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           shuffle_order_engine<_Eng, _Kp>& __x)
+{
+    typedef typename shuffle_order_engine<_Eng, _Kp>::result_type result_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    _Eng __e;
+    result_type _Vp[_Kp+1];
+    __is >> __e;
+    for (size_t __i = 0; __i < _Kp+1; ++__i)
+        __is >> _Vp[__i];
+    if (!__is.fail())
+    {
+        __x.__e_ = __e;
+        for (size_t __i = 0; __i < _Kp; ++__i)
+            __x._V_[__i] = _Vp[__i];
+        __x._Y_ = _Vp[_Kp];
+    }
+    return __is;
+}
+
+typedef shuffle_order_engine<minstd_rand0, 256>                         knuth_b;
+
+// random_device
+
+class _LIBCUDACXX_TYPE_VIS random_device
+{
+#ifdef _LIBCUDACXX_USING_DEV_RANDOM
+    int __f_;
+#endif // defined(_LIBCUDACXX_USING_DEV_RANDOM)
+public:
+    // types
+    typedef unsigned result_type;
+
+    // generator characteristics
+    static _LIBCUDACXX_CONSTEXPR const result_type _Min = 0;
+    static _LIBCUDACXX_CONSTEXPR const result_type _Max = 0xFFFFFFFFu;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type min() { return _Min;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static _LIBCUDACXX_CONSTEXPR result_type max() { return _Max;}
+
+    // constructors
+    explicit random_device(const string& __token = "/dev/urandom");
+    ~random_device();
+
+    // generating functions
+    result_type operator()();
+
+    // property functions
+    double entropy() const _NOEXCEPT;
+
+private:
+    // no copy functions
+    random_device(const random_device&); // = delete;
+    random_device& operator=(const random_device&); // = delete;
+};
+
+// seed_seq
+
+class _LIBCUDACXX_TEMPLATE_VIS seed_seq
+{
+public:
+    // types
+    typedef uint32_t result_type;
+
+private:
+    vector<result_type> __v_;
+
+    template<class _InputIterator>
+        void init(_InputIterator __first, _InputIterator __last);
+public:
+    // constructors
+    _LIBCUDACXX_INLINE_VISIBILITY
+    seed_seq() _NOEXCEPT {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template<class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        seed_seq(initializer_list<_Tp> __il) {init(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template<class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        seed_seq(_InputIterator __first, _InputIterator __last)
+             {init(__first, __last);}
+
+    // generating functions
+    template<class _RandomAccessIterator>
+        void generate(_RandomAccessIterator __first, _RandomAccessIterator __last);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const _NOEXCEPT {return __v_.size();}
+    template<class _OutputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void param(_OutputIterator __dest) const
+            {_CUDA_VSTD::copy(__v_.begin(), __v_.end(), __dest);}
+
+private:
+    // no copy functions
+    seed_seq(const seed_seq&); // = delete;
+    void operator=(const seed_seq&); // = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static result_type _Tp(result_type __x) {return __x ^ (__x >> 27);}
+};
+
+template<class _InputIterator>
+void
+seed_seq::init(_InputIterator __first, _InputIterator __last)
+{
+    for (_InputIterator __s = __first; __s != __last; ++__s)
+        __v_.push_back(*__s & 0xFFFFFFFF);
+}
+
+template<class _RandomAccessIterator>
+void
+seed_seq::generate(_RandomAccessIterator __first, _RandomAccessIterator __last)
+{
+    if (__first != __last)
+    {
+        _CUDA_VSTD::fill(__first, __last, 0x8b8b8b8b);
+        const size_t __n = static_cast<size_t>(__last - __first);
+        const size_t __s = __v_.size();
+        const size_t __t = (__n >= 623) ? 11
+                         : (__n >= 68) ? 7
+                         : (__n >= 39) ? 5
+                         : (__n >= 7)  ? 3
+                         : (__n - 1) / 2;
+        const size_t __p = (__n - __t) / 2;
+        const size_t __q = __p + __t;
+        const size_t __m = _CUDA_VSTD::max(__s + 1, __n);
+        // __k = 0;
+        {
+            result_type __r = 1664525 * _Tp(__first[0] ^ __first[__p]
+                                                      ^  __first[__n - 1]);
+            __first[__p] += __r;
+            __r += __s;
+            __first[__q] += __r;
+            __first[0] = __r;
+        }
+        for (size_t __k = 1; __k <= __s; ++__k)
+        {
+            const size_t __kmodn = __k % __n;
+            const size_t __kpmodn = (__k + __p) % __n;
+            result_type __r = 1664525 * _Tp(__first[__kmodn] ^ __first[__kpmodn]
+                                           ^ __first[(__k - 1) % __n]);
+            __first[__kpmodn] += __r;
+            __r +=  __kmodn + __v_[__k-1];
+            __first[(__k + __q) % __n] += __r;
+            __first[__kmodn] = __r;
+        }
+        for (size_t __k = __s + 1; __k < __m; ++__k)
+        {
+            const size_t __kmodn = __k % __n;
+            const size_t __kpmodn = (__k + __p) % __n;
+            result_type __r = 1664525 * _Tp(__first[__kmodn] ^ __first[__kpmodn]
+                                           ^ __first[(__k - 1) % __n]);
+            __first[__kpmodn] += __r;
+            __r +=  __kmodn;
+            __first[(__k + __q) % __n] += __r;
+            __first[__kmodn] = __r;
+        }
+        for (size_t __k = __m; __k < __m + __n; ++__k)
+        {
+            const size_t __kmodn = __k % __n;
+            const size_t __kpmodn = (__k + __p) % __n;
+            result_type __r = 1566083941 * _Tp(__first[__kmodn] +
+                                              __first[__kpmodn] +
+                                              __first[(__k - 1) % __n]);
+            __first[__kpmodn] ^= __r;
+            __r -= __kmodn;
+            __first[(__k + __q) % __n] ^= __r;
+            __first[__kmodn] = __r;
+        }
+    }
+}
+
+// generate_canonical
+
+template<class _RealType, size_t __bits, class _URNG>
+_RealType
+generate_canonical(_URNG& __g)
+{
+    const size_t _Dt = numeric_limits<_RealType>::digits;
+    const size_t __b = _Dt < __bits ? _Dt : __bits;
+#ifdef _LIBCUDACXX_CXX03_LANG
+    const size_t __logR = __log2<uint64_t, _URNG::_Max - _URNG::_Min + uint64_t(1)>::value;
+#else
+    const size_t __logR = __log2<uint64_t, _URNG::max() - _URNG::min() + uint64_t(1)>::value;
+#endif
+    const size_t __k = __b / __logR + (__b % __logR != 0) + (__b == 0);
+    const _RealType _Rp = static_cast<_RealType>(_URNG::max() - _URNG::min()) + _RealType(1);
+    _RealType __base = _Rp;
+    _RealType _Sp = __g() - _URNG::min();
+    for (size_t __i = 1; __i < __k; ++__i, __base *= _Rp)
+        _Sp += (__g() - _URNG::min()) * __base;
+    return _Sp / __base;
+}
+
+// uniform_int_distribution
+
+// in <algorithm>
+
+template <class _CharT, class _Traits, class _IT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const uniform_int_distribution<_IT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    return __os << __x.a() << __sp << __x.b();
+}
+
+template <class _CharT, class _Traits, class _IT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           uniform_int_distribution<_IT>& __x)
+{
+    typedef uniform_int_distribution<_IT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __a;
+    result_type __b;
+    __is >> __a >> __b;
+    if (!__is.fail())
+        __x.param(param_type(__a, __b));
+    return __is;
+}
+
+// uniform_real_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS uniform_real_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __a_;
+        result_type __b_;
+    public:
+        typedef uniform_real_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __a = 0,
+                            result_type __b = 1)
+            : __a_(__a), __b_(__b) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type a() const {return __a_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type b() const {return __b_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__a_ == __y.__a_ && __x.__b_ == __y.__b_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit uniform_real_distribution(result_type __a = 0, result_type __b = 1)
+        : __p_(param_type(__a, __b)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit uniform_real_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> _LIBCUDACXX_INLINE_VISIBILITY result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type a() const {return __p_.a();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type b() const {return __p_.b();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return a();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return b();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const uniform_real_distribution& __x,
+                        const uniform_real_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const uniform_real_distribution& __x,
+                        const uniform_real_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template<class _RealType>
+template<class _URNG>
+inline
+typename uniform_real_distribution<_RealType>::result_type
+uniform_real_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    return (__p.b() - __p.a())
+        * _CUDA_VSTD::generate_canonical<_RealType, numeric_limits<_RealType>::digits>(__g)
+        + __p.a();
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const uniform_real_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    return __os << __x.a() << __sp << __x.b();
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           uniform_real_distribution<_RT>& __x)
+{
+    typedef uniform_real_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __a;
+    result_type __b;
+    __is >> __a >> __b;
+    if (!__is.fail())
+        __x.param(param_type(__a, __b));
+    return __is;
+}
+
+// bernoulli_distribution
+
+class _LIBCUDACXX_TEMPLATE_VIS bernoulli_distribution
+{
+public:
+    // types
+    typedef bool result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        double __p_;
+    public:
+        typedef bernoulli_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(double __p = 0.5) : __p_(__p) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        double p() const {return __p_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__p_ == __y.__p_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit bernoulli_distribution(double __p = 0.5)
+        : __p_(param_type(__p)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit bernoulli_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> _LIBCUDACXX_INLINE_VISIBILITY result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    double p() const {return __p_.p();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return false;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return true;}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const bernoulli_distribution& __x,
+                        const bernoulli_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const bernoulli_distribution& __x,
+                        const bernoulli_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template<class _URNG>
+inline
+bernoulli_distribution::result_type
+bernoulli_distribution::operator()(_URNG& __g, const param_type& __p)
+{
+    uniform_real_distribution<double> __gen;
+    return __gen(__g) < __p.p();
+}
+
+template <class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const bernoulli_distribution& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    return __os << __x.p();
+}
+
+template <class _CharT, class _Traits>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is, bernoulli_distribution& __x)
+{
+    typedef bernoulli_distribution _Eng;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    double __p;
+    __is >> __p;
+    if (!__is.fail())
+        __x.param(param_type(__p));
+    return __is;
+}
+
+// binomial_distribution
+
+template<class _IntType = int>
+class _LIBCUDACXX_TEMPLATE_VIS binomial_distribution
+{
+public:
+    // types
+    typedef _IntType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __t_;
+        double __p_;
+        double __pr_;
+        double __odds_ratio_;
+        result_type __r0_;
+    public:
+        typedef binomial_distribution distribution_type;
+
+        explicit param_type(result_type __t = 1, double __p = 0.5);
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type t() const {return __t_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        double p() const {return __p_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__t_ == __y.__t_ && __x.__p_ == __y.__p_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+
+        friend class binomial_distribution;
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit binomial_distribution(result_type __t = 1, double __p = 0.5)
+        : __p_(param_type(__t, __p)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit binomial_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type t() const {return __p_.t();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    double p() const {return __p_.p();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return t();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const binomial_distribution& __x,
+                        const binomial_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const binomial_distribution& __x,
+                        const binomial_distribution& __y)
+        {return !(__x == __y);}
+};
+
+#ifndef _LIBCUDACXX_MSVCRT
+extern "C" double lgamma_r(double, int *);
+#endif
+
+inline _LIBCUDACXX_INLINE_VISIBILITY double __libcpp_lgamma(double __d) {
+#if defined(_LIBCUDACXX_MSVCRT)
+  return lgamma(__d);
+#else
+  int __sign;
+  return lgamma_r(__d, &__sign);
+#endif
+}
+
+template<class _IntType>
+binomial_distribution<_IntType>::param_type::param_type(const result_type __t, const double __p)
+    : __t_(__t), __p_(__p)
+{
+    if (0 < __p_ && __p_ < 1)
+    {
+        __r0_ = static_cast<result_type>((__t_ + 1) * __p_);
+        __pr_ = _CUDA_VSTD::exp(__libcpp_lgamma(__t_ + 1.) -
+                           __libcpp_lgamma(__r0_ + 1.) -
+                           __libcpp_lgamma(__t_ - __r0_ + 1.) + __r0_ * _CUDA_VSTD::log(__p_) +
+                           (__t_ - __r0_) * _CUDA_VSTD::log(1 - __p_));
+        __odds_ratio_ = __p_ / (1 - __p_);
+    }
+}
+
+// Reference: Kemp, C.D. (1986). `A modal method for generating binomial
+//           variables', Commun. Statist. - Theor. Meth. 15(3), 805-813.
+template<class _IntType>
+template<class _URNG>
+_IntType
+binomial_distribution<_IntType>::operator()(_URNG& __g, const param_type& __pr)
+{
+    if (__pr.__t_ == 0 || __pr.__p_ == 0)
+        return 0;
+    if (__pr.__p_ == 1)
+        return __pr.__t_;
+    uniform_real_distribution<double> __gen;
+    double __u = __gen(__g) - __pr.__pr_;
+    if (__u < 0)
+        return __pr.__r0_;
+    double __pu = __pr.__pr_;
+    double __pd = __pu;
+    result_type __ru = __pr.__r0_;
+    result_type __rd = __ru;
+    while (true)
+    {
+        if (__rd >= 1)
+        {
+            __pd *= __rd / (__pr.__odds_ratio_ * (__pr.__t_ - __rd + 1));
+            __u -= __pd;
+            if (__u < 0)
+                return __rd - 1;
+        }
+        if ( __rd != 0 )
+            --__rd;
+        ++__ru;
+        if (__ru <= __pr.__t_)
+        {
+            __pu *= (__pr.__t_ - __ru + 1) * __pr.__odds_ratio_ / __ru;
+            __u -= __pu;
+            if (__u < 0)
+                return __ru;
+        }
+    }
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const binomial_distribution<_IntType>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    return __os << __x.t() << __sp << __x.p();
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           binomial_distribution<_IntType>& __x)
+{
+    typedef binomial_distribution<_IntType> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __t;
+    double __p;
+    __is >> __t >> __p;
+    if (!__is.fail())
+        __x.param(param_type(__t, __p));
+    return __is;
+}
+
+// exponential_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS exponential_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __lambda_;
+    public:
+        typedef exponential_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __lambda = 1) : __lambda_(__lambda) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type lambda() const {return __lambda_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__lambda_ == __y.__lambda_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit exponential_distribution(result_type __lambda = 1)
+        : __p_(param_type(__lambda)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit exponential_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type lambda() const {return __p_.lambda();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const exponential_distribution& __x,
+                        const exponential_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const exponential_distribution& __x,
+                        const exponential_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _RealType>
+template<class _URNG>
+_RealType
+exponential_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    return -_CUDA_VSTD::log
+                  (
+                      result_type(1) -
+                      _CUDA_VSTD::generate_canonical<result_type,
+                                       numeric_limits<result_type>::digits>(__g)
+                  )
+                  / __p.lambda();
+}
+
+template <class _CharT, class _Traits, class _RealType>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const exponential_distribution<_RealType>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    return __os << __x.lambda();
+}
+
+template <class _CharT, class _Traits, class _RealType>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           exponential_distribution<_RealType>& __x)
+{
+    typedef exponential_distribution<_RealType> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __lambda;
+    __is >> __lambda;
+    if (!__is.fail())
+        __x.param(param_type(__lambda));
+    return __is;
+}
+
+// normal_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS normal_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __mean_;
+        result_type __stddev_;
+    public:
+        typedef normal_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __mean = 0, result_type __stddev = 1)
+            : __mean_(__mean), __stddev_(__stddev) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type mean() const {return __mean_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type stddev() const {return __stddev_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__mean_ == __y.__mean_ && __x.__stddev_ == __y.__stddev_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+    result_type _V_;
+    bool _V_hot_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit normal_distribution(result_type __mean = 0, result_type __stddev = 1)
+        : __p_(param_type(__mean, __stddev)), _V_hot_(false) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit normal_distribution(const param_type& __p)
+        : __p_(__p), _V_hot_(false) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {_V_hot_ = false;}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type mean() const {return __p_.mean();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type stddev() const {return __p_.stddev();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return -numeric_limits<result_type>::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const normal_distribution& __x,
+                        const normal_distribution& __y)
+        {return __x.__p_ == __y.__p_ && __x._V_hot_ == __y._V_hot_ &&
+                (!__x._V_hot_ || __x._V_ == __y._V_);}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const normal_distribution& __x,
+                        const normal_distribution& __y)
+        {return !(__x == __y);}
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const normal_distribution<_RT>& __x);
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               normal_distribution<_RT>& __x);
+};
+
+template <class _RealType>
+template<class _URNG>
+_RealType
+normal_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    result_type _Up;
+    if (_V_hot_)
+    {
+        _V_hot_ = false;
+        _Up = _V_;
+    }
+    else
+    {
+        uniform_real_distribution<result_type> _Uni(-1, 1);
+        result_type __u;
+        result_type __v;
+        result_type __s;
+        do
+        {
+            __u = _Uni(__g);
+            __v = _Uni(__g);
+            __s = __u * __u + __v * __v;
+        } while (__s > 1 || __s == 0);
+        result_type _Fp = _CUDA_VSTD::sqrt(-2 * _CUDA_VSTD::log(__s) / __s);
+        _V_ = __v * _Fp;
+        _V_hot_ = true;
+        _Up = __u * _Fp;
+    }
+    return _Up * __p.stddev() + __p.mean();
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const normal_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.mean() << __sp << __x.stddev() << __sp << __x._V_hot_;
+    if (__x._V_hot_)
+        __os << __sp << __x._V_;
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           normal_distribution<_RT>& __x)
+{
+    typedef normal_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __mean;
+    result_type __stddev;
+    result_type _Vp = 0;
+    bool _V_hot = false;
+    __is >> __mean >> __stddev >> _V_hot;
+    if (_V_hot)
+        __is >> _Vp;
+    if (!__is.fail())
+    {
+        __x.param(param_type(__mean, __stddev));
+        __x._V_hot_ = _V_hot;
+        __x._V_ = _Vp;
+    }
+    return __is;
+}
+
+// lognormal_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS lognormal_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        normal_distribution<result_type> __nd_;
+    public:
+        typedef lognormal_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __m = 0, result_type __s = 1)
+            : __nd_(__m, __s) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type m() const {return __nd_.mean();}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type s() const {return __nd_.stddev();}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__nd_ == __y.__nd_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+        friend class lognormal_distribution;
+
+        template <class _CharT, class _Traits, class _RT>
+        friend
+        basic_ostream<_CharT, _Traits>&
+        operator<<(basic_ostream<_CharT, _Traits>& __os,
+                   const lognormal_distribution<_RT>& __x);
+
+        template <class _CharT, class _Traits, class _RT>
+        friend
+        basic_istream<_CharT, _Traits>&
+        operator>>(basic_istream<_CharT, _Traits>& __is,
+                   lognormal_distribution<_RT>& __x);
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit lognormal_distribution(result_type __m = 0, result_type __s = 1)
+        : __p_(param_type(__m, __s)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit lognormal_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {__p_.__nd_.reset();}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g, const param_type& __p)
+        {return _CUDA_VSTD::exp(const_cast<normal_distribution<result_type>&>(__p.__nd_)(__g));}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type m() const {return __p_.m();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type s() const {return __p_.s();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const lognormal_distribution& __x,
+                        const lognormal_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const lognormal_distribution& __x,
+                        const lognormal_distribution& __y)
+        {return !(__x == __y);}
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const lognormal_distribution<_RT>& __x);
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               lognormal_distribution<_RT>& __x);
+};
+
+template <class _CharT, class _Traits, class _RT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const lognormal_distribution<_RT>& __x)
+{
+    return __os << __x.__p_.__nd_;
+}
+
+template <class _CharT, class _Traits, class _RT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           lognormal_distribution<_RT>& __x)
+{
+    return __is >> __x.__p_.__nd_;
+}
+
+// poisson_distribution
+
+template<class _IntType = int>
+class _LIBCUDACXX_TEMPLATE_VIS poisson_distribution
+{
+public:
+    // types
+    typedef _IntType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        double __mean_;
+        double __s_;
+        double __d_;
+        double __l_;
+        double __omega_;
+        double __c0_;
+        double __c1_;
+        double __c2_;
+        double __c3_;
+        double __c_;
+
+    public:
+        typedef poisson_distribution distribution_type;
+
+        explicit param_type(double __mean = 1.0);
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        double mean() const {return __mean_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__mean_ == __y.__mean_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+
+        friend class poisson_distribution;
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit poisson_distribution(double __mean = 1.0) : __p_(__mean) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit poisson_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    double mean() const {return __p_.mean();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::max();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const poisson_distribution& __x,
+                        const poisson_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const poisson_distribution& __x,
+                        const poisson_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template<class _IntType>
+poisson_distribution<_IntType>::param_type::param_type(double __mean)
+    : __mean_(__mean)
+{
+    if (__mean_ < 10)
+    {
+        __s_ = 0;
+        __d_ = 0;
+        __l_ = _CUDA_VSTD::exp(-__mean_);
+        __omega_ = 0;
+        __c3_ = 0;
+        __c2_ = 0;
+        __c1_ = 0;
+        __c0_ = 0;
+        __c_ = 0;
+    }
+    else
+    {
+        __s_ = _CUDA_VSTD::sqrt(__mean_);
+        __d_ = 6 * __mean_ * __mean_;
+        __l_ = static_cast<result_type>(__mean_ - 1.1484);
+        __omega_ = .3989423 / __s_;
+        double __b1_ = .4166667E-1 / __mean_;
+        double __b2_ = .3 * __b1_ * __b1_;
+        __c3_ = .1428571 * __b1_ * __b2_;
+        __c2_ = __b2_ - 15. * __c3_;
+        __c1_ = __b1_ - 6. * __b2_ + 45. * __c3_;
+        __c0_ = 1. - __b1_ + 3. * __b2_ - 15. * __c3_;
+        __c_ = .1069 / __mean_;
+    }
+}
+
+template <class _IntType>
+template<class _URNG>
+_IntType
+poisson_distribution<_IntType>::operator()(_URNG& __urng, const param_type& __pr)
+{
+    result_type __x;
+    uniform_real_distribution<double> __urd;
+    if (__pr.__mean_ < 10)
+    {
+         __x = 0;
+        for (double __p = __urd(__urng); __p > __pr.__l_; ++__x)
+            __p *= __urd(__urng);
+    }
+    else
+    {
+        double __difmuk;
+        double __g = __pr.__mean_ + __pr.__s_ * normal_distribution<double>()(__urng);
+        double __u;
+        if (__g > 0)
+        {
+            __x = static_cast<result_type>(__g);
+            if (__x >= __pr.__l_)
+                return __x;
+            __difmuk = __pr.__mean_ - __x;
+            __u = __urd(__urng);
+            if (__pr.__d_ * __u >= __difmuk * __difmuk * __difmuk)
+                return __x;
+        }
+        exponential_distribution<double> __edist;
+        for (bool __using_exp_dist = false; true; __using_exp_dist = true)
+        {
+            double __e;
+            if (__using_exp_dist || __g < 0)
+            {
+                double __t;
+                do
+                {
+                    __e = __edist(__urng);
+                    __u = __urd(__urng);
+                    __u += __u - 1;
+                    __t = 1.8 + (__u < 0 ? -__e : __e);
+                } while (__t <= -.6744);
+                __x = __pr.__mean_ + __pr.__s_ * __t;
+                __difmuk = __pr.__mean_ - __x;
+                __using_exp_dist = true;
+            }
+            double __px;
+            double __py;
+            if (__x < 10)
+            {
+                const double __fac[] = {1, 1, 2, 6, 24, 120, 720, 5040,
+                                             40320, 362880};
+                __px = -__pr.__mean_;
+                __py = _CUDA_VSTD::pow(__pr.__mean_, (double)__x) / __fac[__x];
+            }
+            else
+            {
+                double __del = .8333333E-1 / __x;
+                __del -= 4.8 * __del * __del * __del;
+                double __v = __difmuk / __x;
+                if (_CUDA_VSTD::abs(__v) > 0.25)
+                    __px = __x * _CUDA_VSTD::log(1 + __v) - __difmuk - __del;
+                else
+                    __px = __x * __v * __v * (((((((.1250060 * __v + -.1384794) *
+                           __v + .1421878) * __v + -.1661269) * __v + .2000118) *
+                           __v + -.2500068) * __v + .3333333) * __v + -.5) - __del;
+                __py = .3989423 / _CUDA_VSTD::sqrt(__x);
+            }
+            double __r = (0.5 - __difmuk) / __pr.__s_;
+            double __r2 = __r * __r;
+            double __fx = -0.5 * __r2;
+            double __fy = __pr.__omega_ * (((__pr.__c3_ * __r2 + __pr.__c2_) *
+                                        __r2 + __pr.__c1_) * __r2 + __pr.__c0_);
+            if (__using_exp_dist)
+            {
+                if (__pr.__c_ * _CUDA_VSTD::abs(__u) <= __py * _CUDA_VSTD::exp(__px + __e) -
+                                                   __fy * _CUDA_VSTD::exp(__fx + __e))
+                    break;
+            }
+            else
+            {
+                if (__fy - __u * __fy <= __py * _CUDA_VSTD::exp(__px - __fx))
+                    break;
+            }
+        }
+    }
+    return __x;
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const poisson_distribution<_IntType>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    return __os << __x.mean();
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           poisson_distribution<_IntType>& __x)
+{
+    typedef poisson_distribution<_IntType> _Eng;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    double __mean;
+    __is >> __mean;
+    if (!__is.fail())
+        __x.param(param_type(__mean));
+    return __is;
+}
+
+// weibull_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS weibull_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __a_;
+        result_type __b_;
+    public:
+        typedef weibull_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __a = 1, result_type __b = 1)
+            : __a_(__a), __b_(__b) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type a() const {return __a_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type b() const {return __b_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__a_ == __y.__a_ && __x.__b_ == __y.__b_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit weibull_distribution(result_type __a = 1, result_type __b = 1)
+        : __p_(param_type(__a, __b)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit weibull_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g, const param_type& __p)
+        {return __p.b() *
+            _CUDA_VSTD::pow(exponential_distribution<result_type>()(__g), 1/__p.a());}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type a() const {return __p_.a();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type b() const {return __p_.b();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const weibull_distribution& __x,
+                        const weibull_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const weibull_distribution& __x,
+                        const weibull_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const weibull_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.a() << __sp << __x.b();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           weibull_distribution<_RT>& __x)
+{
+    typedef weibull_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __a;
+    result_type __b;
+    __is >> __a >> __b;
+    if (!__is.fail())
+        __x.param(param_type(__a, __b));
+    return __is;
+}
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS extreme_value_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __a_;
+        result_type __b_;
+    public:
+        typedef extreme_value_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __a = 0, result_type __b = 1)
+            : __a_(__a), __b_(__b) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type a() const {return __a_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type b() const {return __b_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__a_ == __y.__a_ && __x.__b_ == __y.__b_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit extreme_value_distribution(result_type __a = 0, result_type __b = 1)
+        : __p_(param_type(__a, __b)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit extreme_value_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type a() const {return __p_.a();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type b() const {return __p_.b();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return -numeric_limits<result_type>::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const extreme_value_distribution& __x,
+                        const extreme_value_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const extreme_value_distribution& __x,
+                        const extreme_value_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template<class _RealType>
+template<class _URNG>
+_RealType
+extreme_value_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    return __p.a() - __p.b() *
+         _CUDA_VSTD::log(-_CUDA_VSTD::log(1-uniform_real_distribution<result_type>()(__g)));
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const extreme_value_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.a() << __sp << __x.b();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           extreme_value_distribution<_RT>& __x)
+{
+    typedef extreme_value_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __a;
+    result_type __b;
+    __is >> __a >> __b;
+    if (!__is.fail())
+        __x.param(param_type(__a, __b));
+    return __is;
+}
+
+// gamma_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS gamma_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __alpha_;
+        result_type __beta_;
+    public:
+        typedef gamma_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __alpha = 1, result_type __beta = 1)
+            : __alpha_(__alpha), __beta_(__beta) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type alpha() const {return __alpha_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type beta() const {return __beta_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__alpha_ == __y.__alpha_ && __x.__beta_ == __y.__beta_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit gamma_distribution(result_type __alpha = 1, result_type __beta = 1)
+        : __p_(param_type(__alpha, __beta)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit gamma_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type alpha() const {return __p_.alpha();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type beta() const {return __p_.beta();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const gamma_distribution& __x,
+                        const gamma_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const gamma_distribution& __x,
+                        const gamma_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _RealType>
+template<class _URNG>
+_RealType
+gamma_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    result_type __a = __p.alpha();
+    uniform_real_distribution<result_type> __gen(0, 1);
+    exponential_distribution<result_type> __egen;
+    result_type __x;
+    if (__a == 1)
+        __x = __egen(__g);
+    else if (__a > 1)
+    {
+        const result_type __b = __a - 1;
+        const result_type __c = 3 * __a - result_type(0.75);
+        while (true)
+        {
+            const result_type __u = __gen(__g);
+            const result_type __v = __gen(__g);
+            const result_type __w = __u * (1 - __u);
+            if (__w != 0)
+            {
+                const result_type __y = _CUDA_VSTD::sqrt(__c / __w) *
+                                        (__u - result_type(0.5));
+                __x = __b + __y;
+                if (__x >= 0)
+                {
+                    const result_type __z = 64 * __w * __w * __w * __v * __v;
+                    if (__z <= 1 - 2 * __y * __y / __x)
+                        break;
+                    if (_CUDA_VSTD::log(__z) <= 2 * (__b * _CUDA_VSTD::log(__x / __b) - __y))
+                        break;
+                }
+            }
+        }
+    }
+    else  // __a < 1
+    {
+        while (true)
+        {
+            const result_type __u = __gen(__g);
+            const result_type __es = __egen(__g);
+            if (__u <= 1 - __a)
+            {
+                __x = _CUDA_VSTD::pow(__u, 1 / __a);
+                if (__x <= __es)
+                    break;
+            }
+            else
+            {
+                const result_type __e = -_CUDA_VSTD::log((1-__u)/__a);
+                __x = _CUDA_VSTD::pow(1 - __a + __a * __e, 1 / __a);
+                if (__x <= __e + __es)
+                    break;
+            }
+        }
+    }
+    return __x * __p.beta();
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const gamma_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.alpha() << __sp << __x.beta();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           gamma_distribution<_RT>& __x)
+{
+    typedef gamma_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __alpha;
+    result_type __beta;
+    __is >> __alpha >> __beta;
+    if (!__is.fail())
+        __x.param(param_type(__alpha, __beta));
+    return __is;
+}
+
+// negative_binomial_distribution
+
+template<class _IntType = int>
+class _LIBCUDACXX_TEMPLATE_VIS negative_binomial_distribution
+{
+public:
+    // types
+    typedef _IntType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __k_;
+        double __p_;
+    public:
+        typedef negative_binomial_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __k = 1, double __p = 0.5)
+            : __k_(__k), __p_(__p) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type k() const {return __k_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        double p() const {return __p_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__k_ == __y.__k_ && __x.__p_ == __y.__p_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit negative_binomial_distribution(result_type __k = 1, double __p = 0.5)
+        : __p_(__k, __p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit negative_binomial_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type k() const {return __p_.k();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    double p() const {return __p_.p();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::max();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const negative_binomial_distribution& __x,
+                        const negative_binomial_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const negative_binomial_distribution& __x,
+                        const negative_binomial_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _IntType>
+template<class _URNG>
+_IntType
+negative_binomial_distribution<_IntType>::operator()(_URNG& __urng, const param_type& __pr)
+{
+    result_type __k = __pr.k();
+    double __p = __pr.p();
+    if (__k <= 21 * __p)
+    {
+        bernoulli_distribution __gen(__p);
+        result_type __f = 0;
+        result_type __s = 0;
+        while (__s < __k)
+        {
+            if (__gen(__urng))
+                ++__s;
+            else
+                ++__f;
+        }
+        return __f;
+    }
+    return poisson_distribution<result_type>(gamma_distribution<double>
+                                            (__k, (1-__p)/__p)(__urng))(__urng);
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const negative_binomial_distribution<_IntType>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    return __os << __x.k() << __sp << __x.p();
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           negative_binomial_distribution<_IntType>& __x)
+{
+    typedef negative_binomial_distribution<_IntType> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __k;
+    double __p;
+    __is >> __k >> __p;
+    if (!__is.fail())
+        __x.param(param_type(__k, __p));
+    return __is;
+}
+
+// geometric_distribution
+
+template<class _IntType = int>
+class _LIBCUDACXX_TEMPLATE_VIS geometric_distribution
+{
+public:
+    // types
+    typedef _IntType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        double __p_;
+    public:
+        typedef geometric_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(double __p = 0.5) : __p_(__p) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        double p() const {return __p_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__p_ == __y.__p_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructors and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit geometric_distribution(double __p = 0.5) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit geometric_distribution(const param_type& __p) : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g, const param_type& __p)
+        {return negative_binomial_distribution<result_type>(1, __p.p())(__g);}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    double p() const {return __p_.p();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::max();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const geometric_distribution& __x,
+                        const geometric_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const geometric_distribution& __x,
+                        const geometric_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _CharT, class _Traits, class _IntType>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const geometric_distribution<_IntType>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    return __os << __x.p();
+}
+
+template <class _CharT, class _Traits, class _IntType>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           geometric_distribution<_IntType>& __x)
+{
+    typedef geometric_distribution<_IntType> _Eng;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    double __p;
+    __is >> __p;
+    if (!__is.fail())
+        __x.param(param_type(__p));
+    return __is;
+}
+
+// chi_squared_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS chi_squared_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __n_;
+    public:
+        typedef chi_squared_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __n = 1) : __n_(__n) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type n() const {return __n_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__n_ == __y.__n_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit chi_squared_distribution(result_type __n = 1)
+        : __p_(param_type(__n)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit chi_squared_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g, const param_type& __p)
+        {return gamma_distribution<result_type>(__p.n() / 2, 2)(__g);}
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type n() const {return __p_.n();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const chi_squared_distribution& __x,
+                        const chi_squared_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const chi_squared_distribution& __x,
+                        const chi_squared_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const chi_squared_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    __os << __x.n();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           chi_squared_distribution<_RT>& __x)
+{
+    typedef chi_squared_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __n;
+    __is >> __n;
+    if (!__is.fail())
+        __x.param(param_type(__n));
+    return __is;
+}
+
+// cauchy_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS cauchy_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __a_;
+        result_type __b_;
+    public:
+        typedef cauchy_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __a = 0, result_type __b = 1)
+            : __a_(__a), __b_(__b) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type a() const {return __a_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type b() const {return __b_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__a_ == __y.__a_ && __x.__b_ == __y.__b_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit cauchy_distribution(result_type __a = 0, result_type __b = 1)
+        : __p_(param_type(__a, __b)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit cauchy_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> _LIBCUDACXX_INLINE_VISIBILITY result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type a() const {return __p_.a();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type b() const {return __p_.b();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return -numeric_limits<result_type>::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const cauchy_distribution& __x,
+                        const cauchy_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const cauchy_distribution& __x,
+                        const cauchy_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _RealType>
+template<class _URNG>
+inline
+_RealType
+cauchy_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    uniform_real_distribution<result_type> __gen;
+    // purposefully let tan arg get as close to pi/2 as it wants, tan will return a finite
+    return __p.a() + __p.b() * _CUDA_VSTD::tan(3.1415926535897932384626433832795 * __gen(__g));
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const cauchy_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.a() << __sp << __x.b();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           cauchy_distribution<_RT>& __x)
+{
+    typedef cauchy_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __a;
+    result_type __b;
+    __is >> __a >> __b;
+    if (!__is.fail())
+        __x.param(param_type(__a, __b));
+    return __is;
+}
+
+// fisher_f_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS fisher_f_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __m_;
+        result_type __n_;
+    public:
+        typedef fisher_f_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __m = 1, result_type __n = 1)
+            : __m_(__m), __n_(__n) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type m() const {return __m_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type n() const {return __n_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__m_ == __y.__m_ && __x.__n_ == __y.__n_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit fisher_f_distribution(result_type __m = 1, result_type __n = 1)
+        : __p_(param_type(__m, __n)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit fisher_f_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type m() const {return __p_.m();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type n() const {return __p_.n();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const fisher_f_distribution& __x,
+                        const fisher_f_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const fisher_f_distribution& __x,
+                        const fisher_f_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _RealType>
+template<class _URNG>
+_RealType
+fisher_f_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    gamma_distribution<result_type> __gdm(__p.m() * result_type(.5));
+    gamma_distribution<result_type> __gdn(__p.n() * result_type(.5));
+    return __p.n() * __gdm(__g) / (__p.m() * __gdn(__g));
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const fisher_f_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    __os << __x.m() << __sp << __x.n();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           fisher_f_distribution<_RT>& __x)
+{
+    typedef fisher_f_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __m;
+    result_type __n;
+    __is >> __m >> __n;
+    if (!__is.fail())
+        __x.param(param_type(__m, __n));
+    return __is;
+}
+
+// student_t_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS student_t_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        result_type __n_;
+    public:
+        typedef student_t_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit param_type(result_type __n = 1) : __n_(__n) {}
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type n() const {return __n_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__n_ == __y.__n_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+    };
+
+private:
+    param_type __p_;
+    normal_distribution<result_type> __nd_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit student_t_distribution(result_type __n = 1)
+        : __p_(param_type(__n)) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit student_t_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {__nd_.reset();}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type n() const {return __p_.n();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return -numeric_limits<result_type>::infinity();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return numeric_limits<result_type>::infinity();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const student_t_distribution& __x,
+                        const student_t_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const student_t_distribution& __x,
+                        const student_t_distribution& __y)
+        {return !(__x == __y);}
+};
+
+template <class _RealType>
+template<class _URNG>
+_RealType
+student_t_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    gamma_distribution<result_type> __gd(__p.n() * .5, 2);
+    return __nd_(__g) * _CUDA_VSTD::sqrt(__p.n()/__gd(__g));
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const student_t_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    __os << __x.n();
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           student_t_distribution<_RT>& __x)
+{
+    typedef student_t_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    typedef typename _Eng::param_type param_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    result_type __n;
+    __is >> __n;
+    if (!__is.fail())
+        __x.param(param_type(__n));
+    return __is;
+}
+
+// discrete_distribution
+
+template<class _IntType = int>
+class _LIBCUDACXX_TEMPLATE_VIS discrete_distribution
+{
+public:
+    // types
+    typedef _IntType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        vector<double> __p_;
+    public:
+        typedef discrete_distribution distribution_type;
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        param_type() {}
+        template<class _InputIterator>
+            _LIBCUDACXX_INLINE_VISIBILITY
+            param_type(_InputIterator __f, _InputIterator __l)
+            : __p_(__f, __l) {__init();}
+#ifndef _LIBCUDACXX_CXX03_LANG
+        _LIBCUDACXX_INLINE_VISIBILITY
+        param_type(initializer_list<double> __wl)
+            : __p_(__wl.begin(), __wl.end()) {__init();}
+#endif  // _LIBCUDACXX_CXX03_LANG
+        template<class _UnaryOperation>
+            param_type(size_t __nw, double __xmin, double __xmax,
+                       _UnaryOperation __fw);
+
+        vector<double> probabilities() const;
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__p_ == __y.__p_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+
+    private:
+        void __init();
+
+        friend class discrete_distribution;
+
+        template <class _CharT, class _Traits, class _IT>
+        friend
+        basic_ostream<_CharT, _Traits>&
+        operator<<(basic_ostream<_CharT, _Traits>& __os,
+                   const discrete_distribution<_IT>& __x);
+
+        template <class _CharT, class _Traits, class _IT>
+        friend
+        basic_istream<_CharT, _Traits>&
+        operator>>(basic_istream<_CharT, _Traits>& __is,
+                   discrete_distribution<_IT>& __x);
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    discrete_distribution() {}
+    template<class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        discrete_distribution(_InputIterator __f, _InputIterator __l)
+            : __p_(__f, __l) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    discrete_distribution(initializer_list<double> __wl)
+        : __p_(__wl) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    template<class _UnaryOperation>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        discrete_distribution(size_t __nw, double __xmin, double __xmax,
+                              _UnaryOperation __fw)
+        : __p_(__nw, __xmin, __xmax, __fw) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit discrete_distribution(const param_type& __p)
+        : __p_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector<double> probabilities() const {return __p_.probabilities();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return __p_.__p_.size();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const discrete_distribution& __x,
+                        const discrete_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const discrete_distribution& __x,
+                        const discrete_distribution& __y)
+        {return !(__x == __y);}
+
+    template <class _CharT, class _Traits, class _IT>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const discrete_distribution<_IT>& __x);
+
+    template <class _CharT, class _Traits, class _IT>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               discrete_distribution<_IT>& __x);
+};
+
+template<class _IntType>
+template<class _UnaryOperation>
+discrete_distribution<_IntType>::param_type::param_type(size_t __nw,
+                                                        double __xmin,
+                                                        double __xmax,
+                                                        _UnaryOperation __fw)
+{
+    if (__nw > 1)
+    {
+        __p_.reserve(__nw - 1);
+        double __d = (__xmax - __xmin) / __nw;
+        double __d2 = __d / 2;
+        for (size_t __k = 0; __k < __nw; ++__k)
+            __p_.push_back(__fw(__xmin + __k * __d + __d2));
+        __init();
+    }
+}
+
+template<class _IntType>
+void
+discrete_distribution<_IntType>::param_type::__init()
+{
+    if (!__p_.empty())
+    {
+        if (__p_.size() > 1)
+        {
+            double __s = _CUDA_VSTD::accumulate(__p_.begin(), __p_.end(), 0.0);
+            for (_CUDA_VSTD::vector<double>::iterator __i = __p_.begin(), __e = __p_.end();
+                                                                       __i < __e; ++__i)
+                *__i /= __s;
+            vector<double> __t(__p_.size() - 1);
+            _CUDA_VSTD::partial_sum(__p_.begin(), __p_.end() - 1, __t.begin());
+            swap(__p_, __t);
+        }
+        else
+        {
+            __p_.clear();
+            __p_.shrink_to_fit();
+        }
+    }
+}
+
+template<class _IntType>
+vector<double>
+discrete_distribution<_IntType>::param_type::probabilities() const
+{
+    size_t __n = __p_.size();
+    _CUDA_VSTD::vector<double> __p(__n+1);
+    _CUDA_VSTD::adjacent_difference(__p_.begin(), __p_.end(), __p.begin());
+    if (__n > 0)
+        __p[__n] = 1 - __p_[__n-1];
+    else
+        __p[0] = 1;
+    return __p;
+}
+
+template<class _IntType>
+template<class _URNG>
+_IntType
+discrete_distribution<_IntType>::operator()(_URNG& __g, const param_type& __p)
+{
+    uniform_real_distribution<double> __gen;
+    return static_cast<_IntType>(
+           _CUDA_VSTD::upper_bound(__p.__p_.begin(), __p.__p_.end(), __gen(__g)) -
+                                                              __p.__p_.begin());
+}
+
+template <class _CharT, class _Traits, class _IT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const discrete_distribution<_IT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    size_t __n = __x.__p_.__p_.size();
+    __os << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__p_[__i];
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _IT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           discrete_distribution<_IT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    size_t __n;
+    __is >> __n;
+    vector<double> __p(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __p[__i];
+    if (!__is.fail())
+        swap(__x.__p_.__p_, __p);
+    return __is;
+}
+
+// piecewise_constant_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS piecewise_constant_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        vector<result_type> __b_;
+        vector<result_type> __densities_;
+        vector<result_type> __areas_;
+    public:
+        typedef piecewise_constant_distribution distribution_type;
+
+        param_type();
+        template<class _InputIteratorB, class _InputIteratorW>
+            param_type(_InputIteratorB __fB, _InputIteratorB __lB,
+                       _InputIteratorW __fW);
+#ifndef _LIBCUDACXX_CXX03_LANG
+        template<class _UnaryOperation>
+            param_type(initializer_list<result_type> __bl, _UnaryOperation __fw);
+#endif  // _LIBCUDACXX_CXX03_LANG
+        template<class _UnaryOperation>
+            param_type(size_t __nw, result_type __xmin, result_type __xmax,
+                       _UnaryOperation __fw);
+        param_type(param_type const&) = default;
+        param_type & operator=(const param_type& __rhs);
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        vector<result_type> intervals() const {return __b_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        vector<result_type> densities() const {return __densities_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__densities_ == __y.__densities_ && __x.__b_ == __y.__b_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+
+    private:
+        void __init();
+
+        friend class piecewise_constant_distribution;
+
+        template <class _CharT, class _Traits, class _RT>
+        friend
+        basic_ostream<_CharT, _Traits>&
+        operator<<(basic_ostream<_CharT, _Traits>& __os,
+                   const piecewise_constant_distribution<_RT>& __x);
+
+        template <class _CharT, class _Traits, class _RT>
+        friend
+        basic_istream<_CharT, _Traits>&
+        operator>>(basic_istream<_CharT, _Traits>& __is,
+                   piecewise_constant_distribution<_RT>& __x);
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    piecewise_constant_distribution() {}
+    template<class _InputIteratorB, class _InputIteratorW>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        piecewise_constant_distribution(_InputIteratorB __fB,
+                                        _InputIteratorB __lB,
+                                        _InputIteratorW __fW)
+        : __p_(__fB, __lB, __fW) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template<class _UnaryOperation>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        piecewise_constant_distribution(initializer_list<result_type> __bl,
+                                        _UnaryOperation __fw)
+        : __p_(__bl, __fw) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template<class _UnaryOperation>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        piecewise_constant_distribution(size_t __nw, result_type __xmin,
+                                        result_type __xmax, _UnaryOperation __fw)
+        : __p_(__nw, __xmin, __xmax, __fw) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit piecewise_constant_distribution(const param_type& __p)
+        : __p_(__p) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector<result_type> intervals() const {return __p_.intervals();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector<result_type> densities() const {return __p_.densities();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return __p_.__b_.front();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return __p_.__b_.back();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const piecewise_constant_distribution& __x,
+                        const piecewise_constant_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const piecewise_constant_distribution& __x,
+                           const piecewise_constant_distribution& __y)
+        {return !(__x == __y);}
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const piecewise_constant_distribution<_RT>& __x);
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               piecewise_constant_distribution<_RT>& __x);
+};
+
+template<class _RealType>
+typename piecewise_constant_distribution<_RealType>::param_type &
+piecewise_constant_distribution<_RealType>::param_type::operator=
+                                                       (const param_type& __rhs)
+{
+//  These can throw
+    __b_.reserve        (__rhs.__b_.size ());
+    __densities_.reserve(__rhs.__densities_.size());
+    __areas_.reserve    (__rhs.__areas_.size());
+
+//  These can not throw
+    __b_         = __rhs.__b_;
+    __densities_ = __rhs.__densities_;
+    __areas_     =  __rhs.__areas_;
+    return *this;
+}
+
+template<class _RealType>
+void
+piecewise_constant_distribution<_RealType>::param_type::__init()
+{
+    // __densities_ contains non-normalized areas
+    result_type __total_area = _CUDA_VSTD::accumulate(__densities_.begin(),
+                                                __densities_.end(),
+                                                result_type());
+    for (size_t __i = 0; __i < __densities_.size(); ++__i)
+        __densities_[__i] /= __total_area;
+    // __densities_ contains normalized areas
+    __areas_.assign(__densities_.size(), result_type());
+    _CUDA_VSTD::partial_sum(__densities_.begin(), __densities_.end() - 1,
+                                                          __areas_.begin() + 1);
+    // __areas_ contains partial sums of normalized areas: [0, __densities_ - 1]
+    __densities_.back() = 1 - __areas_.back();  // correct round off error
+    for (size_t __i = 0; __i < __densities_.size(); ++__i)
+        __densities_[__i] /= (__b_[__i+1] - __b_[__i]);
+    // __densities_ now contains __densities_
+}
+
+template<class _RealType>
+piecewise_constant_distribution<_RealType>::param_type::param_type()
+    : __b_(2),
+      __densities_(1, 1.0),
+      __areas_(1, 0.0)
+{
+    __b_[1] = 1;
+}
+
+template<class _RealType>
+template<class _InputIteratorB, class _InputIteratorW>
+piecewise_constant_distribution<_RealType>::param_type::param_type(
+        _InputIteratorB __fB, _InputIteratorB __lB, _InputIteratorW __fW)
+    : __b_(__fB, __lB)
+{
+    if (__b_.size() < 2)
+    {
+        __b_.resize(2);
+        __b_[0] = 0;
+        __b_[1] = 1;
+        __densities_.assign(1, 1.0);
+        __areas_.assign(1, 0.0);
+    }
+    else
+    {
+        __densities_.reserve(__b_.size() - 1);
+        for (size_t __i = 0; __i < __b_.size() - 1; ++__i, ++__fW)
+            __densities_.push_back(*__fW);
+        __init();
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _RealType>
+template<class _UnaryOperation>
+piecewise_constant_distribution<_RealType>::param_type::param_type(
+        initializer_list<result_type> __bl, _UnaryOperation __fw)
+    : __b_(__bl.begin(), __bl.end())
+{
+    if (__b_.size() < 2)
+    {
+        __b_.resize(2);
+        __b_[0] = 0;
+        __b_[1] = 1;
+        __densities_.assign(1, 1.0);
+        __areas_.assign(1, 0.0);
+    }
+    else
+    {
+        __densities_.reserve(__b_.size() - 1);
+        for (size_t __i = 0; __i < __b_.size() - 1; ++__i)
+            __densities_.push_back(__fw((__b_[__i+1] + __b_[__i])*.5));
+        __init();
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template<class _RealType>
+template<class _UnaryOperation>
+piecewise_constant_distribution<_RealType>::param_type::param_type(
+        size_t __nw, result_type __xmin, result_type __xmax, _UnaryOperation __fw)
+    : __b_(__nw == 0 ? 2 : __nw + 1)
+{
+    size_t __n = __b_.size() - 1;
+    result_type __d = (__xmax - __xmin) / __n;
+    __densities_.reserve(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+    {
+        __b_[__i] = __xmin + __i * __d;
+        __densities_.push_back(__fw(__b_[__i] + __d*.5));
+    }
+    __b_[__n] = __xmax;
+    __init();
+}
+
+template<class _RealType>
+template<class _URNG>
+_RealType
+piecewise_constant_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    typedef uniform_real_distribution<result_type> _Gen;
+    result_type __u = _Gen()(__g);
+    ptrdiff_t __k = _CUDA_VSTD::upper_bound(__p.__areas_.begin(), __p.__areas_.end(),
+                                      __u) - __p.__areas_.begin() - 1;
+    return (__u - __p.__areas_[__k]) / __p.__densities_[__k] + __p.__b_[__k];
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const piecewise_constant_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    size_t __n = __x.__p_.__b_.size();
+    __os << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__b_[__i];
+    __n = __x.__p_.__densities_.size();
+    __os << __sp << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__densities_[__i];
+    __n = __x.__p_.__areas_.size();
+    __os << __sp << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__areas_[__i];
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           piecewise_constant_distribution<_RT>& __x)
+{
+    typedef piecewise_constant_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    size_t __n;
+    __is >> __n;
+    vector<result_type> __b(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __b[__i];
+    __is >> __n;
+    vector<result_type> __densities(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __densities[__i];
+    __is >> __n;
+    vector<result_type> __areas(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __areas[__i];
+    if (!__is.fail())
+    {
+        swap(__x.__p_.__b_, __b);
+        swap(__x.__p_.__densities_, __densities);
+        swap(__x.__p_.__areas_, __areas);
+    }
+    return __is;
+}
+
+// piecewise_linear_distribution
+
+template<class _RealType = double>
+class _LIBCUDACXX_TEMPLATE_VIS piecewise_linear_distribution
+{
+public:
+    // types
+    typedef _RealType result_type;
+
+    class _LIBCUDACXX_TEMPLATE_VIS param_type
+    {
+        vector<result_type> __b_;
+        vector<result_type> __densities_;
+        vector<result_type> __areas_;
+    public:
+        typedef piecewise_linear_distribution distribution_type;
+
+        param_type();
+        template<class _InputIteratorB, class _InputIteratorW>
+            param_type(_InputIteratorB __fB, _InputIteratorB __lB,
+                       _InputIteratorW __fW);
+#ifndef _LIBCUDACXX_CXX03_LANG
+        template<class _UnaryOperation>
+            param_type(initializer_list<result_type> __bl, _UnaryOperation __fw);
+#endif  // _LIBCUDACXX_CXX03_LANG
+        template<class _UnaryOperation>
+            param_type(size_t __nw, result_type __xmin, result_type __xmax,
+                       _UnaryOperation __fw);
+        param_type(param_type const&) = default;
+        param_type & operator=(const param_type& __rhs);
+
+        _LIBCUDACXX_INLINE_VISIBILITY
+        vector<result_type> intervals() const {return __b_;}
+        _LIBCUDACXX_INLINE_VISIBILITY
+        vector<result_type> densities() const {return __densities_;}
+
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator==(const param_type& __x, const param_type& __y)
+            {return __x.__densities_ == __y.__densities_ && __x.__b_ == __y.__b_;}
+        friend _LIBCUDACXX_INLINE_VISIBILITY
+            bool operator!=(const param_type& __x, const param_type& __y)
+            {return !(__x == __y);}
+
+    private:
+        void __init();
+
+        friend class piecewise_linear_distribution;
+
+        template <class _CharT, class _Traits, class _RT>
+        friend
+        basic_ostream<_CharT, _Traits>&
+        operator<<(basic_ostream<_CharT, _Traits>& __os,
+                   const piecewise_linear_distribution<_RT>& __x);
+
+        template <class _CharT, class _Traits, class _RT>
+        friend
+        basic_istream<_CharT, _Traits>&
+        operator>>(basic_istream<_CharT, _Traits>& __is,
+                   piecewise_linear_distribution<_RT>& __x);
+    };
+
+private:
+    param_type __p_;
+
+public:
+    // constructor and reset functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    piecewise_linear_distribution() {}
+    template<class _InputIteratorB, class _InputIteratorW>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        piecewise_linear_distribution(_InputIteratorB __fB,
+                                      _InputIteratorB __lB,
+                                      _InputIteratorW __fW)
+        : __p_(__fB, __lB, __fW) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template<class _UnaryOperation>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        piecewise_linear_distribution(initializer_list<result_type> __bl,
+                                      _UnaryOperation __fw)
+        : __p_(__bl, __fw) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template<class _UnaryOperation>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        piecewise_linear_distribution(size_t __nw, result_type __xmin,
+                                      result_type __xmax, _UnaryOperation __fw)
+        : __p_(__nw, __xmin, __xmax, __fw) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit piecewise_linear_distribution(const param_type& __p)
+        : __p_(__p) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reset() {}
+
+    // generating functions
+    template<class _URNG>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        result_type operator()(_URNG& __g)
+        {return (*this)(__g, __p_);}
+    template<class _URNG> result_type operator()(_URNG& __g, const param_type& __p);
+
+    // property functions
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector<result_type> intervals() const {return __p_.intervals();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector<result_type> densities() const {return __p_.densities();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    param_type param() const {return __p_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void param(const param_type& __p) {__p_ = __p;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const {return __p_.__b_.front();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const {return __p_.__b_.back();}
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const piecewise_linear_distribution& __x,
+                        const piecewise_linear_distribution& __y)
+        {return __x.__p_ == __y.__p_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const piecewise_linear_distribution& __x,
+                        const piecewise_linear_distribution& __y)
+        {return !(__x == __y);}
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_ostream<_CharT, _Traits>&
+    operator<<(basic_ostream<_CharT, _Traits>& __os,
+               const piecewise_linear_distribution<_RT>& __x);
+
+    template <class _CharT, class _Traits, class _RT>
+    friend
+    basic_istream<_CharT, _Traits>&
+    operator>>(basic_istream<_CharT, _Traits>& __is,
+               piecewise_linear_distribution<_RT>& __x);
+};
+
+template<class _RealType>
+typename piecewise_linear_distribution<_RealType>::param_type &
+piecewise_linear_distribution<_RealType>::param_type::operator=
+                                                       (const param_type& __rhs)
+{
+//  These can throw
+    __b_.reserve        (__rhs.__b_.size ());
+    __densities_.reserve(__rhs.__densities_.size());
+    __areas_.reserve    (__rhs.__areas_.size());
+
+//  These can not throw
+    __b_         = __rhs.__b_;
+    __densities_ = __rhs.__densities_;
+    __areas_     =  __rhs.__areas_;
+    return *this;
+}
+
+
+template<class _RealType>
+void
+piecewise_linear_distribution<_RealType>::param_type::__init()
+{
+    __areas_.assign(__densities_.size() - 1, result_type());
+    result_type _Sp = 0;
+    for (size_t __i = 0; __i < __areas_.size(); ++__i)
+    {
+        __areas_[__i] = (__densities_[__i+1] + __densities_[__i]) *
+                        (__b_[__i+1] - __b_[__i]) * .5;
+        _Sp += __areas_[__i];
+    }
+    for (size_t __i = __areas_.size(); __i > 1;)
+    {
+        --__i;
+        __areas_[__i] = __areas_[__i-1] / _Sp;
+    }
+    __areas_[0] = 0;
+    for (size_t __i = 1; __i < __areas_.size(); ++__i)
+        __areas_[__i] += __areas_[__i-1];
+    for (size_t __i = 0; __i < __densities_.size(); ++__i)
+        __densities_[__i] /= _Sp;
+}
+
+template<class _RealType>
+piecewise_linear_distribution<_RealType>::param_type::param_type()
+    : __b_(2),
+      __densities_(2, 1.0),
+      __areas_(1, 0.0)
+{
+    __b_[1] = 1;
+}
+
+template<class _RealType>
+template<class _InputIteratorB, class _InputIteratorW>
+piecewise_linear_distribution<_RealType>::param_type::param_type(
+        _InputIteratorB __fB, _InputIteratorB __lB, _InputIteratorW __fW)
+    : __b_(__fB, __lB)
+{
+    if (__b_.size() < 2)
+    {
+        __b_.resize(2);
+        __b_[0] = 0;
+        __b_[1] = 1;
+        __densities_.assign(2, 1.0);
+        __areas_.assign(1, 0.0);
+    }
+    else
+    {
+        __densities_.reserve(__b_.size());
+        for (size_t __i = 0; __i < __b_.size(); ++__i, ++__fW)
+            __densities_.push_back(*__fW);
+        __init();
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _RealType>
+template<class _UnaryOperation>
+piecewise_linear_distribution<_RealType>::param_type::param_type(
+        initializer_list<result_type> __bl, _UnaryOperation __fw)
+    : __b_(__bl.begin(), __bl.end())
+{
+    if (__b_.size() < 2)
+    {
+        __b_.resize(2);
+        __b_[0] = 0;
+        __b_[1] = 1;
+        __densities_.assign(2, 1.0);
+        __areas_.assign(1, 0.0);
+    }
+    else
+    {
+        __densities_.reserve(__b_.size());
+        for (size_t __i = 0; __i < __b_.size(); ++__i)
+            __densities_.push_back(__fw(__b_[__i]));
+        __init();
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template<class _RealType>
+template<class _UnaryOperation>
+piecewise_linear_distribution<_RealType>::param_type::param_type(
+        size_t __nw, result_type __xmin, result_type __xmax, _UnaryOperation __fw)
+    : __b_(__nw == 0 ? 2 : __nw + 1)
+{
+    size_t __n = __b_.size() - 1;
+    result_type __d = (__xmax - __xmin) / __n;
+    __densities_.reserve(__b_.size());
+    for (size_t __i = 0; __i < __n; ++__i)
+    {
+        __b_[__i] = __xmin + __i * __d;
+        __densities_.push_back(__fw(__b_[__i]));
+    }
+    __b_[__n] = __xmax;
+    __densities_.push_back(__fw(__b_[__n]));
+    __init();
+}
+
+template<class _RealType>
+template<class _URNG>
+_RealType
+piecewise_linear_distribution<_RealType>::operator()(_URNG& __g, const param_type& __p)
+{
+    typedef uniform_real_distribution<result_type> _Gen;
+    result_type __u = _Gen()(__g);
+    ptrdiff_t __k = _CUDA_VSTD::upper_bound(__p.__areas_.begin(), __p.__areas_.end(),
+                                      __u) - __p.__areas_.begin() - 1;
+    __u -= __p.__areas_[__k];
+    const result_type __dk = __p.__densities_[__k];
+    const result_type __dk1 = __p.__densities_[__k+1];
+    const result_type __deltad = __dk1 - __dk;
+    const result_type __bk = __p.__b_[__k];
+    if (__deltad == 0)
+        return __u / __dk + __bk;
+    const result_type __bk1 = __p.__b_[__k+1];
+    const result_type __deltab = __bk1 - __bk;
+    return (__bk * __dk1 - __bk1 * __dk +
+        _CUDA_VSTD::sqrt(__deltab * (__deltab * __dk * __dk + 2 * __deltad * __u))) /
+        __deltad;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const piecewise_linear_distribution<_RT>& __x)
+{
+    __save_flags<_CharT, _Traits> __lx(__os);
+    __os.flags(ios_base::dec | ios_base::left | ios_base::fixed |
+               ios_base::scientific);
+    _CharT __sp = __os.widen(' ');
+    __os.fill(__sp);
+    size_t __n = __x.__p_.__b_.size();
+    __os << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__b_[__i];
+    __n = __x.__p_.__densities_.size();
+    __os << __sp << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__densities_[__i];
+    __n = __x.__p_.__areas_.size();
+    __os << __sp << __n;
+    for (size_t __i = 0; __i < __n; ++__i)
+        __os << __sp << __x.__p_.__areas_[__i];
+    return __os;
+}
+
+template <class _CharT, class _Traits, class _RT>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           piecewise_linear_distribution<_RT>& __x)
+{
+    typedef piecewise_linear_distribution<_RT> _Eng;
+    typedef typename _Eng::result_type result_type;
+    __save_flags<_CharT, _Traits> __lx(__is);
+    __is.flags(ios_base::dec | ios_base::skipws);
+    size_t __n;
+    __is >> __n;
+    vector<result_type> __b(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __b[__i];
+    __is >> __n;
+    vector<result_type> __densities(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __densities[__i];
+    __is >> __n;
+    vector<result_type> __areas(__n);
+    for (size_t __i = 0; __i < __n; ++__i)
+        __is >> __areas[__i];
+    if (!__is.fail())
+    {
+        swap(__x.__p_.__b_, __b);
+        swap(__x.__p_.__densities_, __densities);
+        swap(__x.__p_.__areas_, __areas);
+    }
+    return __is;
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_RANDOM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ratio b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ratio
new file mode 100644
index 000000000000..8c40ceca8fb2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/ratio
@@ -0,0 +1,533 @@
+// -*- C++ -*-
+//===---------------------------- ratio -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_RATIO
+#define _LIBCUDACXX_RATIO
+
+/*
+    ratio synopsis
+
+namespace std
+{
+
+template <intmax_t N, intmax_t D = 1>
+class ratio
+{
+public:
+    static constexpr intmax_t num;
+    static constexpr intmax_t den;
+    typedef ratio<num, den> type;
+};
+
+// ratio arithmetic
+template <class R1, class R2> using ratio_add = ...;
+template <class R1, class R2> using ratio_subtract = ...;
+template <class R1, class R2> using ratio_multiply = ...;
+template <class R1, class R2> using ratio_divide = ...;
+
+// ratio comparison
+template <class R1, class R2> struct ratio_equal;
+template <class R1, class R2> struct ratio_not_equal;
+template <class R1, class R2> struct ratio_less;
+template <class R1, class R2> struct ratio_less_equal;
+template <class R1, class R2> struct ratio_greater;
+template <class R1, class R2> struct ratio_greater_equal;
+
+// convenience SI typedefs
+typedef ratio<1, 1000000000000000000000000> yocto;  // not supported
+typedef ratio<1,    1000000000000000000000> zepto;  // not supported
+typedef ratio<1,       1000000000000000000> atto;
+typedef ratio<1,          1000000000000000> femto;
+typedef ratio<1,             1000000000000> pico;
+typedef ratio<1,                1000000000> nano;
+typedef ratio<1,                   1000000> micro;
+typedef ratio<1,                      1000> milli;
+typedef ratio<1,                       100> centi;
+typedef ratio<1,                        10> deci;
+typedef ratio<                       10, 1> deca;
+typedef ratio<                      100, 1> hecto;
+typedef ratio<                     1000, 1> kilo;
+typedef ratio<                  1000000, 1> mega;
+typedef ratio<               1000000000, 1> giga;
+typedef ratio<            1000000000000, 1> tera;
+typedef ratio<         1000000000000000, 1> peta;
+typedef ratio<      1000000000000000000, 1> exa;
+typedef ratio<   1000000000000000000000, 1> zetta;  // not supported
+typedef ratio<1000000000000000000000000, 1> yotta;  // not supported
+
+  // 20.11.5, ratio comparison
+  template <class R1, class R2> inline constexpr bool ratio_equal_v
+    = ratio_equal<R1, R2>::value;                                       // C++17
+  template <class R1, class R2> inline constexpr bool ratio_not_equal_v
+    = ratio_not_equal<R1, R2>::value;                                   // C++17
+  template <class R1, class R2> inline constexpr bool ratio_less_v
+    = ratio_less<R1, R2>::value;                                        // C++17
+  template <class R1, class R2> inline constexpr bool ratio_less_equal_v
+    = ratio_less_equal<R1, R2>::value;                                  // C++17
+  template <class R1, class R2> inline constexpr bool ratio_greater_v
+    = ratio_greater<R1, R2>::value;                                     // C++17
+  template <class R1, class R2> inline constexpr bool ratio_greater_equal_v
+    = ratio_greater_equal<R1, R2>::value;                               // C++17
+}
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <cstdint>
+#include <climits>
+#include <type_traits>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// __static_gcd
+
+template <intmax_t _Xp, intmax_t _Yp>
+struct __static_gcd
+{
+    static const intmax_t value = __static_gcd<_Yp, _Xp % _Yp>::value;
+};
+
+template <intmax_t _Xp>
+struct __static_gcd<_Xp, 0>
+{
+    static const intmax_t value = _Xp;
+};
+
+template <>
+struct __static_gcd<0, 0>
+{
+    static const intmax_t value = 1;
+};
+
+// __static_lcm
+
+template <intmax_t _Xp, intmax_t _Yp>
+struct __static_lcm
+{
+    static const intmax_t value = _Xp / __static_gcd<_Xp, _Yp>::value * _Yp;
+};
+
+template <intmax_t _Xp>
+struct __static_abs
+{
+    static const intmax_t value = _Xp < 0 ? -_Xp : _Xp;
+};
+
+template <intmax_t _Xp>
+struct __static_sign
+{
+    static const intmax_t value = _Xp == 0 ? 0 : (_Xp < 0 ? -1 : 1);
+};
+
+template <intmax_t _Xp, intmax_t _Yp, intmax_t = __static_sign<_Yp>::value>
+class __ll_add;
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_add<_Xp, _Yp, 1>
+{
+    static const intmax_t min = (1LL << (sizeof(intmax_t) * CHAR_BIT - 1)) + 1;
+    static const intmax_t max = -min;
+
+    static_assert(_Xp <= max - _Yp, "overflow in __ll_add");
+public:
+    static const intmax_t value = _Xp + _Yp;
+};
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_add<_Xp, _Yp, 0>
+{
+public:
+    static const intmax_t value = _Xp;
+};
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_add<_Xp, _Yp, -1>
+{
+    static const intmax_t min = (1LL << (sizeof(intmax_t) * CHAR_BIT - 1)) + 1;
+    static const intmax_t max = -min;
+
+    static_assert(min - _Yp <= _Xp, "overflow in __ll_add");
+public:
+    static const intmax_t value = _Xp + _Yp;
+};
+
+template <intmax_t _Xp, intmax_t _Yp, intmax_t = __static_sign<_Yp>::value>
+class __ll_sub;
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_sub<_Xp, _Yp, 1>
+{
+    static const intmax_t min = (1LL << (sizeof(intmax_t) * CHAR_BIT - 1)) + 1;
+    static const intmax_t max = -min;
+
+    static_assert(min + _Yp <= _Xp, "overflow in __ll_sub");
+public:
+    static const intmax_t value = _Xp - _Yp;
+};
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_sub<_Xp, _Yp, 0>
+{
+public:
+    static const intmax_t value = _Xp;
+};
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_sub<_Xp, _Yp, -1>
+{
+    static const intmax_t min = (1LL << (sizeof(intmax_t) * CHAR_BIT - 1)) + 1;
+    static const intmax_t max = -min;
+
+    static_assert(_Xp <= max + _Yp, "overflow in __ll_sub");
+public:
+    static const intmax_t value = _Xp - _Yp;
+};
+
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_mul
+{
+    static const intmax_t nan = (1LL << (sizeof(intmax_t) * CHAR_BIT - 1));
+    static const intmax_t min = nan + 1;
+    static const intmax_t max = -min;
+    static const intmax_t __a_x = __static_abs<_Xp>::value;
+    static const intmax_t __a_y = __static_abs<_Yp>::value;
+
+    static_assert(_Xp != nan && _Yp != nan && __a_x <= max / __a_y, "overflow in __ll_mul");
+public:
+    static const intmax_t value = _Xp * _Yp;
+};
+
+template <intmax_t _Yp>
+class __ll_mul<0, _Yp>
+{
+public:
+    static const intmax_t value = 0;
+};
+
+template <intmax_t _Xp>
+class __ll_mul<_Xp, 0>
+{
+public:
+    static const intmax_t value = 0;
+};
+
+template <>
+class __ll_mul<0, 0>
+{
+public:
+    static const intmax_t value = 0;
+};
+
+// Not actually used but left here in case needed in future maintenance
+template <intmax_t _Xp, intmax_t _Yp>
+class __ll_div
+{
+    static const intmax_t nan = (1LL << (sizeof(intmax_t) * CHAR_BIT - 1));
+    static const intmax_t min = nan + 1;
+    static const intmax_t max = -min;
+
+    static_assert(_Xp != nan && _Yp != nan && _Yp != 0, "overflow in __ll_div");
+public:
+    static const intmax_t value = _Xp / _Yp;
+};
+
+template <intmax_t _Num, intmax_t _Den = 1>
+class _LIBCUDACXX_TEMPLATE_VIS ratio
+{
+    static_assert(__static_abs<_Num>::value >= 0, "ratio numerator is out of range");
+    static_assert(_Den != 0, "ratio divide by 0");
+    static_assert(__static_abs<_Den>::value >  0, "ratio denominator is out of range");
+    static _LIBCUDACXX_CONSTEXPR const intmax_t __na = __static_abs<_Num>::value;
+    static _LIBCUDACXX_CONSTEXPR const intmax_t __da = __static_abs<_Den>::value;
+    static _LIBCUDACXX_CONSTEXPR const intmax_t __s = __static_sign<_Num>::value * __static_sign<_Den>::value;
+    static _LIBCUDACXX_CONSTEXPR const intmax_t __gcd = __static_gcd<__na, __da>::value;
+public:
+    static _LIBCUDACXX_CONSTEXPR const intmax_t num = __s * __na / __gcd;
+    static _LIBCUDACXX_CONSTEXPR const intmax_t den = __da / __gcd;
+
+    typedef ratio<num, den> type;
+};
+
+template <intmax_t _Num, intmax_t _Den>
+_LIBCUDACXX_CONSTEXPR const intmax_t ratio<_Num, _Den>::num;
+
+template <intmax_t _Num, intmax_t _Den>
+_LIBCUDACXX_CONSTEXPR const intmax_t ratio<_Num, _Den>::den;
+
+template <class _Tp>                    struct __is_ratio                     : false_type {};
+template <intmax_t _Num, intmax_t _Den> struct __is_ratio<ratio<_Num, _Den> > : true_type  {};
+
+typedef ratio<1LL, 1000000000000000000LL> atto;
+typedef ratio<1LL,    1000000000000000LL> femto;
+typedef ratio<1LL,       1000000000000LL> pico;
+typedef ratio<1LL,          1000000000LL> nano;
+typedef ratio<1LL,             1000000LL> micro;
+typedef ratio<1LL,                1000LL> milli;
+typedef ratio<1LL,                 100LL> centi;
+typedef ratio<1LL,                  10LL> deci;
+typedef ratio<                 10LL, 1LL> deca;
+typedef ratio<                100LL, 1LL> hecto;
+typedef ratio<               1000LL, 1LL> kilo;
+typedef ratio<            1000000LL, 1LL> mega;
+typedef ratio<         1000000000LL, 1LL> giga;
+typedef ratio<      1000000000000LL, 1LL> tera;
+typedef ratio<   1000000000000000LL, 1LL> peta;
+typedef ratio<1000000000000000000LL, 1LL> exa;
+
+template <class _R1, class _R2>
+struct __ratio_multiply
+{
+//private:
+    static const intmax_t __gcd_n1_d2 = __static_gcd<_R1::num, _R2::den>::value;
+    static const intmax_t __gcd_d1_n2 = __static_gcd<_R1::den, _R2::num>::value;
+public:
+    typedef typename ratio
+        <
+            __ll_mul<_R1::num / __gcd_n1_d2, _R2::num / __gcd_d1_n2>::value,
+            __ll_mul<_R2::den / __gcd_n1_d2, _R1::den / __gcd_d1_n2>::value
+        >::type type;
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2> using ratio_multiply
+                                    = typename __ratio_multiply<_R1, _R2>::type;
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_multiply
+    : public __ratio_multiply<_R1, _R2>::type {};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct __ratio_divide
+{
+//private:
+    static const intmax_t __gcd_n1_n2 = __static_gcd<_R1::num, _R2::num>::value;
+    static const intmax_t __gcd_d1_d2 = __static_gcd<_R1::den, _R2::den>::value;
+public:
+    typedef typename ratio
+        <
+            __ll_mul<_R1::num / __gcd_n1_n2, _R2::den / __gcd_d1_d2>::value,
+            __ll_mul<_R2::num / __gcd_n1_n2, _R1::den / __gcd_d1_d2>::value
+        >::type type;
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2> using ratio_divide
+                                      = typename __ratio_divide<_R1, _R2>::type;
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_divide
+    : public __ratio_divide<_R1, _R2>::type {};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct __ratio_add
+{
+//private:
+    static const intmax_t __gcd_n1_n2 = __static_gcd<_R1::num, _R2::num>::value;
+    static const intmax_t __gcd_d1_d2 = __static_gcd<_R1::den, _R2::den>::value;
+public:
+    typedef typename ratio_multiply
+        <
+            ratio<__gcd_n1_n2, _R1::den / __gcd_d1_d2>,
+            ratio
+            <
+                __ll_add
+                <
+                    __ll_mul<_R1::num / __gcd_n1_n2, _R2::den / __gcd_d1_d2>::value,
+                    __ll_mul<_R2::num / __gcd_n1_n2, _R1::den / __gcd_d1_d2>::value
+                >::value,
+                _R2::den
+            >
+        >::type type;
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2> using ratio_add
+                                         = typename __ratio_add<_R1, _R2>::type;
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_add
+    : public __ratio_add<_R1, _R2>::type {};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct __ratio_subtract
+{
+//private:
+    static const intmax_t __gcd_n1_n2 = __static_gcd<_R1::num, _R2::num>::value;
+    static const intmax_t __gcd_d1_d2 = __static_gcd<_R1::den, _R2::den>::value;
+public:
+    typedef typename ratio_multiply
+        <
+            ratio<__gcd_n1_n2, _R1::den / __gcd_d1_d2>,
+            ratio
+            <
+                __ll_sub
+                <
+                    __ll_mul<_R1::num / __gcd_n1_n2, _R2::den / __gcd_d1_d2>::value,
+                    __ll_mul<_R2::num / __gcd_n1_n2, _R1::den / __gcd_d1_d2>::value
+                >::value,
+                _R2::den
+            >
+        >::type type;
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2> using ratio_subtract
+                                    = typename __ratio_subtract<_R1, _R2>::type;
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_subtract
+    : public __ratio_subtract<_R1, _R2>::type {};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+// ratio_equal
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_equal
+    : public _LIBCUDACXX_BOOL_CONSTANT((_R1::num == _R2::num && _R1::den == _R2::den)) {};
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_not_equal
+    : public _LIBCUDACXX_BOOL_CONSTANT((!ratio_equal<_R1, _R2>::value)) {};
+
+// ratio_less
+
+template <class _R1, class _R2, bool _Odd = false,
+          intmax_t _Q1 = _R1::num / _R1::den, intmax_t _M1 = _R1::num % _R1::den,
+          intmax_t _Q2 = _R2::num / _R2::den, intmax_t _M2 = _R2::num % _R2::den>
+struct __ratio_less1
+{
+    static const bool value = _Odd ? _Q2 < _Q1 : _Q1 < _Q2;
+};
+
+template <class _R1, class _R2, bool _Odd, intmax_t _Qp>
+struct __ratio_less1<_R1, _R2, _Odd, _Qp, 0, _Qp, 0>
+{
+    static const bool value = false;
+};
+
+template <class _R1, class _R2, bool _Odd, intmax_t _Qp, intmax_t _M2>
+struct __ratio_less1<_R1, _R2, _Odd, _Qp, 0, _Qp, _M2>
+{
+    static const bool value = !_Odd;
+};
+
+template <class _R1, class _R2, bool _Odd, intmax_t _Qp, intmax_t _M1>
+struct __ratio_less1<_R1, _R2, _Odd, _Qp, _M1, _Qp, 0>
+{
+    static const bool value = _Odd;
+};
+
+template <class _R1, class _R2, bool _Odd, intmax_t _Qp, intmax_t _M1,
+                                                        intmax_t _M2>
+struct __ratio_less1<_R1, _R2, _Odd, _Qp, _M1, _Qp, _M2>
+{
+    static const bool value = __ratio_less1<ratio<_R1::den, _M1>,
+                                            ratio<_R2::den, _M2>, !_Odd>::value;
+};
+
+template <class _R1, class _R2, intmax_t _S1 = __static_sign<_R1::num>::value,
+                                intmax_t _S2 = __static_sign<_R2::num>::value>
+struct __ratio_less
+{
+    static const bool value = _S1 < _S2;
+};
+
+template <class _R1, class _R2>
+struct __ratio_less<_R1, _R2, 1LL, 1LL>
+{
+    static const bool value = __ratio_less1<_R1, _R2>::value;
+};
+
+template <class _R1, class _R2>
+struct __ratio_less<_R1, _R2, -1LL, -1LL>
+{
+    static const bool value = __ratio_less1<ratio<-_R2::num, _R2::den>, ratio<-_R1::num, _R1::den> >::value;
+};
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_less
+    : public _LIBCUDACXX_BOOL_CONSTANT((__ratio_less<_R1, _R2>::value)) {};
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_less_equal
+    : public _LIBCUDACXX_BOOL_CONSTANT((!ratio_less<_R2, _R1>::value)) {};
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_greater
+    : public _LIBCUDACXX_BOOL_CONSTANT((ratio_less<_R2, _R1>::value)) {};
+
+template <class _R1, class _R2>
+struct _LIBCUDACXX_TEMPLATE_VIS ratio_greater_equal
+    : public _LIBCUDACXX_BOOL_CONSTANT((!ratio_less<_R1, _R2>::value)) {};
+
+template <class _R1, class _R2>
+struct __ratio_gcd
+{
+    typedef ratio<__static_gcd<_R1::num, _R2::num>::value,
+                  __static_lcm<_R1::den, _R2::den>::value> type;
+};
+
+#if _LIBCUDACXX_STD_VER > 14 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _R1, class _R2>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool ratio_equal_v
+    = ratio_equal<_R1, _R2>::value;
+
+template <class _R1, class _R2>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool ratio_not_equal_v
+    = ratio_not_equal<_R1, _R2>::value;
+
+template <class _R1, class _R2>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool ratio_less_v
+    = ratio_less<_R1, _R2>::value;
+
+template <class _R1, class _R2>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool ratio_less_equal_v
+    = ratio_less_equal<_R1, _R2>::value;
+
+template <class _R1, class _R2>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool ratio_greater_v
+    = ratio_greater<_R1, _R2>::value;
+
+template <class _R1, class _R2>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool ratio_greater_equal_v
+    = ratio_greater_equal<_R1, _R2>::value;
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_RATIO
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/regex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/regex
new file mode 100644
index 000000000000..10d879cb0eb0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/regex
@@ -0,0 +1,6662 @@
+// -*- C++ -*-
+//===--------------------------- regex ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_REGEX
+#define _LIBCUDACXX_REGEX
+
+/*
+    regex synopsis
+
+#include <initializer_list>
+
+namespace std
+{
+
+namespace regex_constants
+{
+
+emum syntax_option_type
+{
+    icase      = unspecified,
+    nosubs     = unspecified,
+    optimize   = unspecified,
+    collate    = unspecified,
+    ECMAScript = unspecified,
+    basic      = unspecified,
+    extended   = unspecified,
+    awk        = unspecified,
+    grep       = unspecified,
+    egrep      = unspecified
+};
+
+constexpr syntax_option_type operator~(syntax_option_type f);
+constexpr syntax_option_type operator&(syntax_option_type lhs, syntax_option_type rhs);
+constexpr syntax_option_type operator|(syntax_option_type lhs, syntax_option_type rhs);
+
+enum match_flag_type
+{
+    match_default     = 0,
+    match_not_bol     = unspecified,
+    match_not_eol     = unspecified,
+    match_not_bow     = unspecified,
+    match_not_eow     = unspecified,
+    match_any         = unspecified,
+    match_not_null    = unspecified,
+    match_continuous  = unspecified,
+    match_prev_avail  = unspecified,
+    format_default    = 0,
+    format_sed        = unspecified,
+    format_no_copy    = unspecified,
+    format_first_only = unspecified
+};
+
+constexpr match_flag_type operator~(match_flag_type f);
+constexpr match_flag_type operator&(match_flag_type lhs, match_flag_type rhs);
+constexpr match_flag_type operator|(match_flag_type lhs, match_flag_type rhs);
+
+enum error_type
+{
+    error_collate    = unspecified,
+    error_ctype      = unspecified,
+    error_escape     = unspecified,
+    error_backref    = unspecified,
+    error_brack      = unspecified,
+    error_paren      = unspecified,
+    error_brace      = unspecified,
+    error_badbrace   = unspecified,
+    error_range      = unspecified,
+    error_space      = unspecified,
+    error_badrepeat  = unspecified,
+    error_complexity = unspecified,
+    error_stack      = unspecified
+};
+
+}  // regex_constants
+
+class regex_error
+    : public runtime_error
+{
+public:
+    explicit regex_error(regex_constants::error_type ecode);
+    regex_constants::error_type code() const;
+};
+
+template <class charT>
+struct regex_traits
+{
+public:
+    typedef charT                   char_type;
+    typedef basic_string<char_type> string_type;
+    typedef locale                  locale_type;
+    typedef /bitmask_type/          char_class_type;
+
+    regex_traits();
+
+    static size_t length(const char_type* p);
+    charT translate(charT c) const;
+    charT translate_nocase(charT c) const;
+    template <class ForwardIterator>
+        string_type
+        transform(ForwardIterator first, ForwardIterator last) const;
+    template <class ForwardIterator>
+        string_type
+        transform_primary( ForwardIterator first, ForwardIterator last) const;
+    template <class ForwardIterator>
+        string_type
+        lookup_collatename(ForwardIterator first, ForwardIterator last) const;
+    template <class ForwardIterator>
+        char_class_type
+        lookup_classname(ForwardIterator first, ForwardIterator last,
+                         bool icase = false) const;
+    bool isctype(charT c, char_class_type f) const;
+    int value(charT ch, int radix) const;
+    locale_type imbue(locale_type l);
+    locale_type getloc()const;
+};
+
+template <class charT, class traits = regex_traits<charT>>
+class basic_regex
+{
+public:
+    // types:
+    typedef charT                               value_type;
+    typedef traits                              traits_type;
+    typedef typename traits::string_type        string_type;
+    typedef regex_constants::syntax_option_type flag_type;
+    typedef typename traits::locale_type        locale_type;
+
+    // constants:
+    static constexpr regex_constants::syntax_option_type icase = regex_constants::icase;
+    static constexpr regex_constants::syntax_option_type nosubs = regex_constants::nosubs;
+    static constexpr regex_constants::syntax_option_type optimize = regex_constants::optimize;
+    static constexpr regex_constants::syntax_option_type collate = regex_constants::collate;
+    static constexpr regex_constants::syntax_option_type ECMAScript = regex_constants::ECMAScript;
+    static constexpr regex_constants::syntax_option_type basic = regex_constants::basic;
+    static constexpr regex_constants::syntax_option_type extended = regex_constants::extended;
+    static constexpr regex_constants::syntax_option_type awk = regex_constants::awk;
+    static constexpr regex_constants::syntax_option_type grep = regex_constants::grep;
+    static constexpr regex_constants::syntax_option_type egrep = regex_constants::egrep;
+
+    // construct/copy/destroy:
+    basic_regex();
+    explicit basic_regex(const charT* p, flag_type f = regex_constants::ECMAScript);
+    basic_regex(const charT* p, size_t len, flag_type f = regex_constants::ECMAScript);
+    basic_regex(const basic_regex&);
+    basic_regex(basic_regex&&) noexcept;
+    template <class ST, class SA>
+        explicit basic_regex(const basic_string<charT, ST, SA>& p,
+                             flag_type f = regex_constants::ECMAScript);
+    template <class ForwardIterator>
+        basic_regex(ForwardIterator first, ForwardIterator last,
+                    flag_type f = regex_constants::ECMAScript);
+    basic_regex(initializer_list<charT>, flag_type = regex_constants::ECMAScript);
+
+    ~basic_regex();
+
+    basic_regex& operator=(const basic_regex&);
+    basic_regex& operator=(basic_regex&&) noexcept;
+    basic_regex& operator=(const charT* ptr);
+    basic_regex& operator=(initializer_list<charT> il);
+    template <class ST, class SA>
+        basic_regex& operator=(const basic_string<charT, ST, SA>& p);
+
+    // assign:
+    basic_regex& assign(const basic_regex& that);
+    basic_regex& assign(basic_regex&& that) noexcept;
+    basic_regex& assign(const charT* ptr,           flag_type f = regex_constants::ECMAScript);
+    basic_regex& assign(const charT* p, size_t len, flag_type f = regex_constants::ECMAScript);
+    template <class string_traits, class A>
+        basic_regex& assign(const basic_string<charT, string_traits, A>& s,
+                                                    flag_type f = regex_constants::ECMAScript);
+    template <class InputIterator>
+        basic_regex& assign(InputIterator first, InputIterator last,
+                                                    flag_type f = regex_constants::ECMAScript);
+    basic_regex& assign(initializer_list<charT>,    flag_type f = regex_constants::ECMAScript);
+
+    // const operations:
+    unsigned mark_count() const;
+    flag_type flags() const;
+
+    // locale:
+    locale_type imbue(locale_type loc);
+    locale_type getloc() const;
+
+    // swap:
+    void swap(basic_regex&);
+};
+
+template<class ForwardIterator>
+basic_regex(ForwardIterator, ForwardIterator,
+            regex_constants::syntax_option_type = regex_constants::ECMAScript)
+    -> basic_regex<typename iterator_traits<ForwardIterator>::value_type>; // C++17
+
+typedef basic_regex<char>    regex;
+typedef basic_regex<wchar_t> wregex;
+
+template <class charT, class traits>
+    void swap(basic_regex<charT, traits>& e1, basic_regex<charT, traits>& e2);
+
+template <class BidirectionalIterator>
+class sub_match
+    : public pair<BidirectionalIterator, BidirectionalIterator>
+{
+public:
+    typedef typename iterator_traits<BidirectionalIterator>::value_type value_type;
+    typedef typename iterator_traits<BidirectionalIterator>::difference_type difference_type;
+    typedef BidirectionalIterator                                      iterator;
+    typedef basic_string<value_type>                                string_type;
+
+    bool matched;
+
+    constexpr sub_match();
+
+    difference_type length() const;
+    operator string_type() const;
+    string_type str() const;
+
+    int compare(const sub_match& s) const;
+    int compare(const string_type& s) const;
+    int compare(const value_type* s) const;
+};
+
+typedef sub_match<const char*>             csub_match;
+typedef sub_match<const wchar_t*>          wcsub_match;
+typedef sub_match<string::const_iterator>  ssub_match;
+typedef sub_match<wstring::const_iterator> wssub_match;
+
+template <class BiIter>
+    bool
+    operator==(const sub_match<BiIter>& lhs, const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator!=(const sub_match<BiIter>& lhs, const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator<(const sub_match<BiIter>& lhs, const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator<=(const sub_match<BiIter>& lhs, const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator>=(const sub_match<BiIter>& lhs, const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator>(const sub_match<BiIter>& lhs, const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator==(const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator!=(const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator<(const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& lhs,
+              const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator>(const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& lhs,
+              const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool operator>=(const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& lhs,
+                    const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator<=(const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator==(const sub_match<BiIter>& lhs,
+               const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator!=(const sub_match<BiIter>& lhs,
+               const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator<(const sub_match<BiIter>& lhs,
+              const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool operator>(const sub_match<BiIter>& lhs,
+                   const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator>=(const sub_match<BiIter>& lhs,
+               const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& rhs);
+
+template <class BiIter, class ST, class SA>
+    bool
+    operator<=(const sub_match<BiIter>& lhs,
+               const basic_string<typename iterator_traits<BiIter>::value_type, ST, SA>& rhs);
+
+template <class BiIter>
+    bool
+    operator==(typename iterator_traits<BiIter>::value_type const* lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator!=(typename iterator_traits<BiIter>::value_type const* lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator<(typename iterator_traits<BiIter>::value_type const* lhs,
+              const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator>(typename iterator_traits<BiIter>::value_type const* lhs,
+              const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator>=(typename iterator_traits<BiIter>::value_type const* lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator<=(typename iterator_traits<BiIter>::value_type const* lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator==(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const* rhs);
+
+template <class BiIter>
+    bool
+    operator!=(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const* rhs);
+
+template <class BiIter>
+    bool
+    operator<(const sub_match<BiIter>& lhs,
+              typename iterator_traits<BiIter>::value_type const* rhs);
+
+template <class BiIter>
+    bool
+    operator>(const sub_match<BiIter>& lhs,
+              typename iterator_traits<BiIter>::value_type const* rhs);
+
+template <class BiIter>
+    bool
+    operator>=(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const* rhs);
+
+template <class BiIter>
+    bool
+    operator<=(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const* rhs);
+
+template <class BiIter>
+    bool
+    operator==(typename iterator_traits<BiIter>::value_type const& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator!=(typename iterator_traits<BiIter>::value_type const& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator<(typename iterator_traits<BiIter>::value_type const& lhs,
+              const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator>(typename iterator_traits<BiIter>::value_type const& lhs,
+              const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator>=(typename iterator_traits<BiIter>::value_type const& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator<=(typename iterator_traits<BiIter>::value_type const& lhs,
+               const sub_match<BiIter>& rhs);
+
+template <class BiIter>
+    bool
+    operator==(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const& rhs);
+
+template <class BiIter>
+    bool
+    operator!=(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const& rhs);
+
+template <class BiIter>
+    bool
+    operator<(const sub_match<BiIter>& lhs,
+              typename iterator_traits<BiIter>::value_type const& rhs);
+
+template <class BiIter>
+    bool
+    operator>(const sub_match<BiIter>& lhs,
+              typename iterator_traits<BiIter>::value_type const& rhs);
+
+template <class BiIter>
+    bool
+    operator>=(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const& rhs);
+
+template <class BiIter>
+    bool
+    operator<=(const sub_match<BiIter>& lhs,
+               typename iterator_traits<BiIter>::value_type const& rhs);
+
+template <class charT, class ST, class BiIter>
+    basic_ostream<charT, ST>&
+    operator<<(basic_ostream<charT, ST>& os, const sub_match<BiIter>& m);
+
+template <class BidirectionalIterator,
+          class Allocator = allocator<sub_match<BidirectionalIterator>>>
+class match_results
+{
+public:
+    typedef sub_match<BidirectionalIterator>                  value_type;
+    typedef const value_type&                                 const_reference;
+    typedef value_type&                                       reference;
+    typedef /implementation-defined/                          const_iterator;
+    typedef const_iterator                                    iterator;
+    typedef typename iterator_traits<BidirectionalIterator>::difference_type difference_type;
+    typedef typename allocator_traits<Allocator>::size_type   size_type;
+    typedef Allocator                                         allocator_type;
+    typedef typename iterator_traits<BidirectionalIterator>::value_type char_type;
+    typedef basic_string<char_type>                           string_type;
+
+    // construct/copy/destroy:
+    explicit match_results(const Allocator& a = Allocator());
+    match_results(const match_results& m);
+    match_results(match_results&& m) noexcept;
+    match_results& operator=(const match_results& m);
+    match_results& operator=(match_results&& m);
+    ~match_results();
+
+    bool ready() const;
+
+    // size:
+    size_type size() const;
+    size_type max_size() const;
+    bool empty() const;
+
+    // element access:
+    difference_type length(size_type sub = 0) const;
+    difference_type position(size_type sub = 0) const;
+    string_type str(size_type sub = 0) const;
+    const_reference operator[](size_type n) const;
+
+    const_reference prefix() const;
+    const_reference suffix() const;
+
+    const_iterator begin() const;
+    const_iterator end() const;
+    const_iterator cbegin() const;
+    const_iterator cend() const;
+
+    // format:
+    template <class OutputIter>
+        OutputIter
+        format(OutputIter out, const char_type* fmt_first,
+               const char_type* fmt_last,
+               regex_constants::match_flag_type flags = regex_constants::format_default) const;
+    template <class OutputIter, class ST, class SA>
+        OutputIter
+        format(OutputIter out, const basic_string<char_type, ST, SA>& fmt,
+               regex_constants::match_flag_type flags = regex_constants::format_default) const;
+    template <class ST, class SA>
+        basic_string<char_type, ST, SA>
+        format(const basic_string<char_type, ST, SA>& fmt,
+               regex_constants::match_flag_type flags = regex_constants::format_default) const;
+    string_type
+        format(const char_type* fmt,
+               regex_constants::match_flag_type flags = regex_constants::format_default) const;
+
+    // allocator:
+    allocator_type get_allocator() const;
+
+    // swap:
+    void swap(match_results& that);
+};
+
+typedef match_results<const char*>             cmatch;
+typedef match_results<const wchar_t*>          wcmatch;
+typedef match_results<string::const_iterator>  smatch;
+typedef match_results<wstring::const_iterator> wsmatch;
+
+template <class BidirectionalIterator, class Allocator>
+    bool
+    operator==(const match_results<BidirectionalIterator, Allocator>& m1,
+               const match_results<BidirectionalIterator, Allocator>& m2);
+
+template <class BidirectionalIterator, class Allocator>
+    bool
+    operator!=(const match_results<BidirectionalIterator, Allocator>& m1,
+               const match_results<BidirectionalIterator, Allocator>& m2);
+
+template <class BidirectionalIterator, class Allocator>
+    void
+    swap(match_results<BidirectionalIterator, Allocator>& m1,
+         match_results<BidirectionalIterator, Allocator>& m2);
+
+template <class BidirectionalIterator, class Allocator, class charT, class traits>
+    bool
+    regex_match(BidirectionalIterator first, BidirectionalIterator last,
+                match_results<BidirectionalIterator, Allocator>& m,
+                const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class BidirectionalIterator, class charT, class traits>
+    bool
+    regex_match(BidirectionalIterator first, BidirectionalIterator last,
+                const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class charT, class Allocator, class traits>
+    bool
+    regex_match(const charT* str, match_results<const charT*, Allocator>& m,
+                const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class ST, class SA, class Allocator, class charT, class traits>
+    bool
+    regex_match(const basic_string<charT, ST, SA>& s,
+                match_results<typename basic_string<charT, ST, SA>::const_iterator, Allocator>& m,
+                const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class ST, class SA, class Allocator, class charT, class traits>
+    bool
+    regex_match(const basic_string<charT, ST, SA>&& s,
+                match_results<typename basic_string<charT, ST, SA>::const_iterator, Allocator>& m,
+                const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default) = delete; // C++14
+
+template <class charT, class traits>
+    bool
+    regex_match(const charT* str, const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class ST, class SA, class charT, class traits>
+    bool
+    regex_match(const basic_string<charT, ST, SA>& s,
+                const basic_regex<charT, traits>& e,
+                regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class BidirectionalIterator, class Allocator, class charT, class traits>
+    bool
+    regex_search(BidirectionalIterator first, BidirectionalIterator last,
+                 match_results<BidirectionalIterator, Allocator>& m,
+                 const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class BidirectionalIterator, class charT, class traits>
+    bool
+    regex_search(BidirectionalIterator first, BidirectionalIterator last,
+                 const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class charT, class Allocator, class traits>
+    bool
+    regex_search(const charT* str, match_results<const charT*, Allocator>& m,
+                 const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class charT, class traits>
+    bool
+    regex_search(const charT* str, const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class ST, class SA, class charT, class traits>
+    bool
+    regex_search(const basic_string<charT, ST, SA>& s,
+                 const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class ST, class SA, class Allocator, class charT, class traits>
+    bool
+    regex_search(const basic_string<charT, ST, SA>& s,
+                 match_results<typename basic_string<charT, ST, SA>::const_iterator, Allocator>& m,
+                 const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class ST, class SA, class Allocator, class charT, class traits>
+    bool
+    regex_search(const basic_string<charT, ST, SA>&& s,
+                 match_results<typename basic_string<charT, ST, SA>::const_iterator, Allocator>& m,
+                 const basic_regex<charT, traits>& e,
+                 regex_constants::match_flag_type flags = regex_constants::match_default) = delete; // C++14
+
+template <class OutputIterator, class BidirectionalIterator,
+          class traits, class charT, class ST, class SA>
+    OutputIterator
+    regex_replace(OutputIterator out,
+                  BidirectionalIterator first, BidirectionalIterator last,
+                  const basic_regex<charT, traits>& e,
+                  const basic_string<charT, ST, SA>& fmt,
+                  regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class OutputIterator, class BidirectionalIterator,
+          class traits, class charT>
+    OutputIterator
+    regex_replace(OutputIterator out,
+                  BidirectionalIterator first, BidirectionalIterator last,
+                  const basic_regex<charT, traits>& e, const charT* fmt,
+                  regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class traits, class charT, class ST, class SA, class FST, class FSA>>
+    basic_string<charT, ST, SA>
+    regex_replace(const basic_string<charT, ST, SA>& s,
+                  const basic_regex<charT, traits>& e,
+                  const basic_string<charT, FST, FSA>& fmt,
+                  regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class traits, class charT, class ST, class SA>
+    basic_string<charT, ST, SA>
+    regex_replace(const basic_string<charT, ST, SA>& s,
+                  const basic_regex<charT, traits>& e, const charT* fmt,
+                  regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class traits, class charT, class ST, class SA>
+    basic_string<charT>
+    regex_replace(const charT* s,
+                  const basic_regex<charT, traits>& e,
+                  const basic_string<charT, ST, SA>& fmt,
+                  regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class traits, class charT>
+    basic_string<charT>
+    regex_replace(const charT* s,
+                  const basic_regex<charT, traits>& e,
+                  const charT* fmt,
+                  regex_constants::match_flag_type flags = regex_constants::match_default);
+
+template <class BidirectionalIterator,
+          class charT = typename iterator_traits< BidirectionalIterator>::value_type,
+          class traits = regex_traits<charT>>
+class regex_iterator
+{
+public:
+    typedef basic_regex<charT, traits>           regex_type;
+    typedef match_results<BidirectionalIterator> value_type;
+    typedef ptrdiff_t                            difference_type;
+    typedef const value_type*                    pointer;
+    typedef const value_type&                    reference;
+    typedef forward_iterator_tag                 iterator_category;
+
+    regex_iterator();
+    regex_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                   const regex_type& re,
+                   regex_constants::match_flag_type m = regex_constants::match_default);
+    regex_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                   const regex_type&& __re,
+                   regex_constants::match_flag_type __m
+                                     = regex_constants::match_default) = delete; // C++14
+    regex_iterator(const regex_iterator&);
+    regex_iterator& operator=(const regex_iterator&);
+
+    bool operator==(const regex_iterator&) const;
+    bool operator!=(const regex_iterator&) const;
+
+    const value_type& operator*() const;
+    const value_type* operator->() const;
+
+    regex_iterator& operator++();
+    regex_iterator operator++(int);
+};
+
+typedef regex_iterator<const char*>             cregex_iterator;
+typedef regex_iterator<const wchar_t*>          wcregex_iterator;
+typedef regex_iterator<string::const_iterator>  sregex_iterator;
+typedef regex_iterator<wstring::const_iterator> wsregex_iterator;
+
+template <class BidirectionalIterator,
+          class charT = typename iterator_traits< BidirectionalIterator>::value_type,
+          class traits = regex_traits<charT>>
+class regex_token_iterator
+{
+public:
+    typedef basic_regex<charT, traits>       regex_type;
+    typedef sub_match<BidirectionalIterator> value_type;
+    typedef ptrdiff_t                        difference_type;
+    typedef const value_type*                pointer;
+    typedef const value_type&                reference;
+    typedef forward_iterator_tag             iterator_category;
+
+    regex_token_iterator();
+    regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                         const regex_type& re, int submatch = 0,
+                         regex_constants::match_flag_type m = regex_constants::match_default);
+    regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                         const regex_type&& re, int submatch = 0,
+                         regex_constants::match_flag_type m = regex_constants::match_default) = delete; // C++14
+    regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                         const regex_type& re, const vector<int>& submatches,
+                         regex_constants::match_flag_type m = regex_constants::match_default);
+    regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                         const regex_type&& re, const vector<int>& submatches,
+                         regex_constants::match_flag_type m = regex_constants::match_default) = delete; // C++14
+    regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                         const regex_type& re, initializer_list<int> submatches,
+                         regex_constants::match_flag_type m = regex_constants::match_default);
+    regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                         const regex_type&& re, initializer_list<int> submatches,
+                         regex_constants::match_flag_type m = regex_constants::match_default) = delete; // C++14
+    template <size_t N>
+        regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                             const regex_type& re, const int (&submatches)[N],
+                             regex_constants::match_flag_type m = regex_constants::match_default);
+    template <size_t N>
+        regex_token_iterator(BidirectionalIterator a, BidirectionalIterator b,
+                             const regex_type& re, const int (&submatches)[N],
+                             regex_constants::match_flag_type m = regex_constants::match_default) = delete // C++14;
+    regex_token_iterator(const regex_token_iterator&);
+    regex_token_iterator& operator=(const regex_token_iterator&);
+
+    bool operator==(const regex_token_iterator&) const;
+    bool operator!=(const regex_token_iterator&) const;
+
+    const value_type& operator*() const;
+    const value_type* operator->() const;
+
+    regex_token_iterator& operator++();
+    regex_token_iterator operator++(int);
+};
+
+typedef regex_token_iterator<const char*>             cregex_token_iterator;
+typedef regex_token_iterator<const wchar_t*>          wcregex_token_iterator;
+typedef regex_token_iterator<string::const_iterator>  sregex_token_iterator;
+typedef regex_token_iterator<wstring::const_iterator> wsregex_token_iterator;
+
+} // std
+*/
+
+#include <__config>
+#include <stdexcept>
+#include <__locale>
+#include <initializer_list>
+#include <utility>
+#include <iterator>
+#include <string>
+#include <memory>
+#include <vector>
+#include <deque>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+#define _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR 4096
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+namespace regex_constants
+{
+
+// syntax_option_type
+
+enum syntax_option_type
+{
+    icase      = 1 << 0,
+    nosubs     = 1 << 1,
+    optimize   = 1 << 2,
+    collate    = 1 << 3,
+#ifdef _LIBCUDACXX_ABI_REGEX_CONSTANTS_NONZERO
+    ECMAScript = 1 << 9,
+#else
+    ECMAScript = 0,
+#endif
+    basic      = 1 << 4,
+    extended   = 1 << 5,
+    awk        = 1 << 6,
+    grep       = 1 << 7,
+    egrep      = 1 << 8
+};
+
+inline _LIBCUDACXX_CONSTEXPR
+syntax_option_type __get_grammar(syntax_option_type __g)
+{
+#ifdef _LIBCUDACXX_ABI_REGEX_CONSTANTS_NONZERO
+    return static_cast<syntax_option_type>(__g & 0x3F0);
+#else
+    return static_cast<syntax_option_type>(__g & 0x1F0);
+#endif
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+syntax_option_type
+operator~(syntax_option_type __x)
+{
+    return syntax_option_type(~int(__x) & 0x1FF);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+syntax_option_type
+operator&(syntax_option_type __x, syntax_option_type __y)
+{
+    return syntax_option_type(int(__x) & int(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+syntax_option_type
+operator|(syntax_option_type __x, syntax_option_type __y)
+{
+    return syntax_option_type(int(__x) | int(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+syntax_option_type
+operator^(syntax_option_type __x, syntax_option_type __y)
+{
+    return syntax_option_type(int(__x) ^ int(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+syntax_option_type&
+operator&=(syntax_option_type& __x, syntax_option_type __y)
+{
+    __x = __x & __y;
+    return __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+syntax_option_type&
+operator|=(syntax_option_type& __x, syntax_option_type __y)
+{
+    __x = __x | __y;
+    return __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+syntax_option_type&
+operator^=(syntax_option_type& __x, syntax_option_type __y)
+{
+    __x = __x ^ __y;
+    return __x;
+}
+
+// match_flag_type
+
+enum match_flag_type
+{
+    match_default     = 0,
+    match_not_bol     = 1 << 0,
+    match_not_eol     = 1 << 1,
+    match_not_bow     = 1 << 2,
+    match_not_eow     = 1 << 3,
+    match_any         = 1 << 4,
+    match_not_null    = 1 << 5,
+    match_continuous  = 1 << 6,
+    match_prev_avail  = 1 << 7,
+    format_default    = 0,
+    format_sed        = 1 << 8,
+    format_no_copy    = 1 << 9,
+    format_first_only = 1 << 10,
+    __no_update_pos   = 1 << 11,
+    __full_match      = 1 << 12
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+match_flag_type
+operator~(match_flag_type __x)
+{
+    return match_flag_type(~int(__x) & 0x0FFF);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+match_flag_type
+operator&(match_flag_type __x, match_flag_type __y)
+{
+    return match_flag_type(int(__x) & int(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+match_flag_type
+operator|(match_flag_type __x, match_flag_type __y)
+{
+    return match_flag_type(int(__x) | int(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR
+match_flag_type
+operator^(match_flag_type __x, match_flag_type __y)
+{
+    return match_flag_type(int(__x) ^ int(__y));
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+match_flag_type&
+operator&=(match_flag_type& __x, match_flag_type __y)
+{
+    __x = __x & __y;
+    return __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+match_flag_type&
+operator|=(match_flag_type& __x, match_flag_type __y)
+{
+    __x = __x | __y;
+    return __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+match_flag_type&
+operator^=(match_flag_type& __x, match_flag_type __y)
+{
+    __x = __x ^ __y;
+    return __x;
+}
+
+enum error_type
+{
+    error_collate = 1,
+    error_ctype,
+    error_escape,
+    error_backref,
+    error_brack,
+    error_paren,
+    error_brace,
+    error_badbrace,
+    error_range,
+    error_space,
+    error_badrepeat,
+    error_complexity,
+    error_stack,
+    __re_err_grammar,
+    __re_err_empty,
+    __re_err_unknown
+};
+
+}  // regex_constants
+
+class _LIBCUDACXX_EXCEPTION_ABI regex_error
+    : public runtime_error
+{
+    regex_constants::error_type __code_;
+public:
+    explicit regex_error(regex_constants::error_type __ecode);
+    virtual ~regex_error() throw();
+     _LIBCUDACXX_INLINE_VISIBILITY
+    regex_constants::error_type code() const {return __code_;}
+};
+
+template <regex_constants::error_type _Ev>
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_regex_error()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw regex_error(_Ev);
+#else
+    _CUDA_VSTD::abort();
+#endif
+}
+
+template <class _CharT>
+struct _LIBCUDACXX_TEMPLATE_VIS regex_traits
+{
+public:
+    typedef _CharT                  char_type;
+    typedef basic_string<char_type> string_type;
+    typedef locale                  locale_type;
+    typedef ctype_base::mask        char_class_type;
+
+    static const char_class_type __regex_word = ctype_base::__regex_word;
+private:
+    locale __loc_;
+    const ctype<char_type>* __ct_;
+    const collate<char_type>* __col_;
+
+public:
+    regex_traits();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_t length(const char_type* __p)
+        {return char_traits<char_type>::length(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char_type translate(char_type __c) const {return __c;}
+    char_type translate_nocase(char_type __c) const;
+    template <class _ForwardIterator>
+        string_type
+        transform(_ForwardIterator __f, _ForwardIterator __l) const;
+    template <class _ForwardIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        string_type
+        transform_primary( _ForwardIterator __f, _ForwardIterator __l) const
+            {return __transform_primary(__f, __l, char_type());}
+    template <class _ForwardIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        string_type
+        lookup_collatename(_ForwardIterator __f, _ForwardIterator __l) const
+            {return __lookup_collatename(__f, __l, char_type());}
+    template <class _ForwardIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        char_class_type
+        lookup_classname(_ForwardIterator __f, _ForwardIterator __l,
+                         bool __icase = false) const
+            {return __lookup_classname(__f, __l, __icase, char_type());}
+    bool isctype(char_type __c, char_class_type __m) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int value(char_type __ch, int __radix) const
+        {return __regex_traits_value(__ch, __radix);}
+    locale_type imbue(locale_type __l);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    locale_type getloc()const {return __loc_;}
+
+private:
+    void __init();
+
+    template <class _ForwardIterator>
+        string_type
+        __transform_primary(_ForwardIterator __f, _ForwardIterator __l, char) const;
+    template <class _ForwardIterator>
+        string_type
+        __transform_primary(_ForwardIterator __f, _ForwardIterator __l, wchar_t) const;
+
+    template <class _ForwardIterator>
+        string_type
+        __lookup_collatename(_ForwardIterator __f, _ForwardIterator __l, char) const;
+    template <class _ForwardIterator>
+        string_type
+        __lookup_collatename(_ForwardIterator __f, _ForwardIterator __l, wchar_t) const;
+
+    template <class _ForwardIterator>
+        char_class_type
+        __lookup_classname(_ForwardIterator __f, _ForwardIterator __l,
+                           bool __icase, char) const;
+    template <class _ForwardIterator>
+        char_class_type
+        __lookup_classname(_ForwardIterator __f, _ForwardIterator __l,
+                           bool __icase, wchar_t) const;
+
+    static int __regex_traits_value(unsigned char __ch, int __radix);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int __regex_traits_value(char __ch, int __radix) const
+        {return __regex_traits_value(static_cast<unsigned char>(__ch), __radix);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int __regex_traits_value(wchar_t __ch, int __radix) const;
+};
+
+template <class _CharT>
+const typename regex_traits<_CharT>::char_class_type
+regex_traits<_CharT>::__regex_word;
+
+template <class _CharT>
+regex_traits<_CharT>::regex_traits()
+{
+    __init();
+}
+
+template <class _CharT>
+typename regex_traits<_CharT>::char_type
+regex_traits<_CharT>::translate_nocase(char_type __c) const
+{
+    return __ct_->tolower(__c);
+}
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::string_type
+regex_traits<_CharT>::transform(_ForwardIterator __f, _ForwardIterator __l) const
+{
+    string_type __s(__f, __l);
+    return __col_->transform(__s.data(), __s.data() + __s.size());
+}
+
+template <class _CharT>
+void
+regex_traits<_CharT>::__init()
+{
+    __ct_ = &use_facet<ctype<char_type> >(__loc_);
+    __col_ = &use_facet<collate<char_type> >(__loc_);
+}
+
+template <class _CharT>
+typename regex_traits<_CharT>::locale_type
+regex_traits<_CharT>::imbue(locale_type __l)
+{
+    locale __r = __loc_;
+    __loc_ = __l;
+    __init();
+    return __r;
+}
+
+// transform_primary is very FreeBSD-specific
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::string_type
+regex_traits<_CharT>::__transform_primary(_ForwardIterator __f,
+                                          _ForwardIterator __l, char) const
+{
+    const string_type __s(__f, __l);
+    string_type __d = __col_->transform(__s.data(), __s.data() + __s.size());
+    switch (__d.size())
+    {
+    case 1:
+        break;
+    case 12:
+        __d[11] = __d[3];
+        break;
+    default:
+        __d.clear();
+        break;
+    }
+    return __d;
+}
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::string_type
+regex_traits<_CharT>::__transform_primary(_ForwardIterator __f,
+                                          _ForwardIterator __l, wchar_t) const
+{
+    const string_type __s(__f, __l);
+    string_type __d = __col_->transform(__s.data(), __s.data() + __s.size());
+    switch (__d.size())
+    {
+    case 1:
+        break;
+    case 3:
+        __d[2] = __d[0];
+        break;
+    default:
+        __d.clear();
+        break;
+    }
+    return __d;
+}
+
+// lookup_collatename is very FreeBSD-specific
+
+_LIBCUDACXX_FUNC_VIS string __get_collation_name(const char* __s);
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::string_type
+regex_traits<_CharT>::__lookup_collatename(_ForwardIterator __f,
+                                           _ForwardIterator __l, char) const
+{
+    string_type __s(__f, __l);
+    string_type __r;
+    if (!__s.empty())
+    {
+        __r = __get_collation_name(__s.c_str());
+        if (__r.empty() && __s.size() <= 2)
+        {
+            __r = __col_->transform(__s.data(), __s.data() + __s.size());
+            if (__r.size() == 1 || __r.size() == 12)
+                __r = __s;
+            else
+                __r.clear();
+        }
+    }
+    return __r;
+}
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::string_type
+regex_traits<_CharT>::__lookup_collatename(_ForwardIterator __f,
+                                           _ForwardIterator __l, wchar_t) const
+{
+    string_type __s(__f, __l);
+    string __n;
+    __n.reserve(__s.size());
+    for (typename string_type::const_iterator __i = __s.begin(), __e = __s.end();
+                                                              __i != __e; ++__i)
+    {
+        if (static_cast<unsigned>(*__i) >= 127)
+            return string_type();
+        __n.push_back(char(*__i));
+    }
+    string_type __r;
+    if (!__s.empty())
+    {
+        __n = __get_collation_name(__n.c_str());
+        if (!__n.empty())
+            __r.assign(__n.begin(), __n.end());
+        else if (__s.size() <= 2)
+        {
+            __r = __col_->transform(__s.data(), __s.data() + __s.size());
+            if (__r.size() == 1 || __r.size() == 3)
+                __r = __s;
+            else
+                __r.clear();
+        }
+    }
+    return __r;
+}
+
+// lookup_classname
+
+regex_traits<char>::char_class_type _LIBCUDACXX_FUNC_VIS
+__get_classname(const char* __s, bool __icase);
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::char_class_type
+regex_traits<_CharT>::__lookup_classname(_ForwardIterator __f,
+                                         _ForwardIterator __l,
+                                         bool __icase, char) const
+{
+    string_type __s(__f, __l);
+    __ct_->tolower(&__s[0], &__s[0] + __s.size());
+    return __get_classname(__s.c_str(), __icase);
+}
+
+template <class _CharT>
+template <class _ForwardIterator>
+typename regex_traits<_CharT>::char_class_type
+regex_traits<_CharT>::__lookup_classname(_ForwardIterator __f,
+                                         _ForwardIterator __l,
+                                         bool __icase, wchar_t) const
+{
+    string_type __s(__f, __l);
+    __ct_->tolower(&__s[0], &__s[0] + __s.size());
+    string __n;
+    __n.reserve(__s.size());
+    for (typename string_type::const_iterator __i = __s.begin(), __e = __s.end();
+                                                              __i != __e; ++__i)
+    {
+        if (static_cast<unsigned>(*__i) >= 127)
+            return char_class_type();
+        __n.push_back(char(*__i));
+    }
+    return __get_classname(__n.c_str(), __icase);
+}
+
+template <class _CharT>
+bool
+regex_traits<_CharT>::isctype(char_type __c, char_class_type __m) const
+{
+    if (__ct_->is(__m, __c))
+        return true;
+    return (__c == '_' && (__m & __regex_word));
+}
+
+template <class _CharT>
+int
+regex_traits<_CharT>::__regex_traits_value(unsigned char __ch, int __radix)
+{
+    if ((__ch & 0xF8u) == 0x30)  // '0' <= __ch && __ch <= '7'
+        return __ch - '0';
+    if (__radix != 8)
+    {
+        if ((__ch & 0xFEu) == 0x38)  // '8' <= __ch && __ch <= '9'
+            return __ch - '0';
+        if (__radix == 16)
+        {
+            __ch |= 0x20;  // tolower
+            if ('a' <= __ch && __ch <= 'f')
+                return __ch - ('a' - 10);
+        }
+    }
+    return -1;
+}
+
+template <class _CharT>
+inline
+int
+regex_traits<_CharT>::__regex_traits_value(wchar_t __ch, int __radix) const
+{
+    return __regex_traits_value(static_cast<unsigned char>(__ct_->narrow(__ch, char_type())), __radix);
+}
+
+template <class _CharT> class __node;
+
+template <class _BidirectionalIterator> class _LIBCUDACXX_TEMPLATE_VIS sub_match;
+
+template <class _BidirectionalIterator,
+          class _Allocator = allocator<sub_match<_BidirectionalIterator> > >
+class _LIBCUDACXX_TEMPLATE_VIS match_results;
+
+template <class _CharT>
+struct __state
+{
+    enum
+    {
+        __end_state = -1000,
+        __consume_input,  // -999
+        __begin_marked_expr, // -998
+        __end_marked_expr,   // -997
+        __pop_state,           // -996
+        __accept_and_consume,  // -995
+        __accept_but_not_consume,  // -994
+        __reject,                  // -993
+        __split,
+        __repeat
+    };
+
+    int __do_;
+    const _CharT* __first_;
+    const _CharT* __current_;
+    const _CharT* __last_;
+    vector<sub_match<const _CharT*> > __sub_matches_;
+    vector<pair<size_t, const _CharT*> > __loop_data_;
+    const __node<_CharT>* __node_;
+    regex_constants::match_flag_type __flags_;
+    bool __at_first_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __state()
+        : __do_(0), __first_(nullptr), __current_(nullptr), __last_(nullptr),
+          __node_(nullptr), __flags_() {}
+};
+
+// __node
+
+template <class _CharT>
+class __node
+{
+    __node(const __node&);
+    __node& operator=(const __node&);
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual ~__node() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual void __exec(__state&) const {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    virtual void __exec_split(bool, __state&) const {}
+};
+
+// __end_state
+
+template <class _CharT>
+class __end_state
+    : public __node<_CharT>
+{
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __end_state() {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__end_state<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__end_state;
+}
+
+// __has_one_state
+
+template <class _CharT>
+class __has_one_state
+    : public __node<_CharT>
+{
+    __node<_CharT>* __first_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __has_one_state(__node<_CharT>* __s)
+        : __first_(__s) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node<_CharT>*  first() const {return __first_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __node<_CharT>*& first()       {return __first_;}
+};
+
+// __owns_one_state
+
+template <class _CharT>
+class __owns_one_state
+    : public __has_one_state<_CharT>
+{
+    typedef __has_one_state<_CharT> base;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __owns_one_state(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual ~__owns_one_state();
+};
+
+template <class _CharT>
+__owns_one_state<_CharT>::~__owns_one_state()
+{
+    delete this->first();
+}
+
+// __empty_state
+
+template <class _CharT>
+class __empty_state
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __empty_state(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__empty_state<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__accept_but_not_consume;
+    __s.__node_ = this->first();
+}
+
+// __empty_non_own_state
+
+template <class _CharT>
+class __empty_non_own_state
+    : public __has_one_state<_CharT>
+{
+    typedef __has_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __empty_non_own_state(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__empty_non_own_state<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__accept_but_not_consume;
+    __s.__node_ = this->first();
+}
+
+// __repeat_one_loop
+
+template <class _CharT>
+class __repeat_one_loop
+    : public __has_one_state<_CharT>
+{
+    typedef __has_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __repeat_one_loop(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__repeat_one_loop<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__repeat;
+    __s.__node_ = this->first();
+}
+
+// __owns_two_states
+
+template <class _CharT>
+class __owns_two_states
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    base* __second_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __owns_two_states(__node<_CharT>* __s1, base* __s2)
+        : base(__s1), __second_(__s2) {}
+
+    virtual ~__owns_two_states();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    base*  second() const {return __second_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    base*& second()       {return __second_;}
+};
+
+template <class _CharT>
+__owns_two_states<_CharT>::~__owns_two_states()
+{
+    delete __second_;
+}
+
+// __loop
+
+template <class _CharT>
+class __loop
+    : public __owns_two_states<_CharT>
+{
+    typedef __owns_two_states<_CharT> base;
+
+    size_t __min_;
+    size_t __max_;
+    unsigned __loop_id_;
+    unsigned __mexp_begin_;
+    unsigned __mexp_end_;
+    bool __greedy_;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __loop(unsigned __loop_id,
+                          __node<_CharT>* __s1, __owns_one_state<_CharT>* __s2,
+                          unsigned __mexp_begin, unsigned __mexp_end,
+                          bool __greedy = true,
+                          size_t __min = 0,
+                          size_t __max = numeric_limits<size_t>::max())
+        : base(__s1, __s2), __min_(__min), __max_(__max), __loop_id_(__loop_id),
+          __mexp_begin_(__mexp_begin), __mexp_end_(__mexp_end),
+          __greedy_(__greedy) {}
+
+    virtual void __exec(__state& __s) const;
+    virtual void __exec_split(bool __second, __state& __s) const;
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __init_repeat(__state& __s) const
+    {
+        __s.__loop_data_[__loop_id_].second = __s.__current_;
+        for (size_t __i = __mexp_begin_-1; __i != __mexp_end_-1; ++__i)
+        {
+            __s.__sub_matches_[__i].first = __s.__last_;
+            __s.__sub_matches_[__i].second = __s.__last_;
+            __s.__sub_matches_[__i].matched = false;
+        }
+    }
+};
+
+template <class _CharT>
+void
+__loop<_CharT>::__exec(__state& __s) const
+{
+    if (__s.__do_ == __state::__repeat)
+    {
+        bool __do_repeat = ++__s.__loop_data_[__loop_id_].first < __max_;
+        bool __do_alt = __s.__loop_data_[__loop_id_].first >= __min_;
+        if (__do_repeat && __do_alt &&
+                               __s.__loop_data_[__loop_id_].second == __s.__current_)
+            __do_repeat = false;
+        if (__do_repeat && __do_alt)
+            __s.__do_ = __state::__split;
+        else if (__do_repeat)
+        {
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__node_ = this->first();
+            __init_repeat(__s);
+        }
+        else
+        {
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__node_ = this->second();
+        }
+    }
+    else
+    {
+        __s.__loop_data_[__loop_id_].first = 0;
+        bool __do_repeat = 0 < __max_;
+        bool __do_alt = 0 >= __min_;
+        if (__do_repeat && __do_alt)
+            __s.__do_ = __state::__split;
+        else if (__do_repeat)
+        {
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__node_ = this->first();
+            __init_repeat(__s);
+        }
+        else
+        {
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__node_ = this->second();
+        }
+    }
+}
+
+template <class _CharT>
+void
+__loop<_CharT>::__exec_split(bool __second, __state& __s) const
+{
+    __s.__do_ = __state::__accept_but_not_consume;
+    if (__greedy_ != __second)
+    {
+        __s.__node_ = this->first();
+        __init_repeat(__s);
+    }
+    else
+        __s.__node_ = this->second();
+}
+
+// __alternate
+
+template <class _CharT>
+class __alternate
+    : public __owns_two_states<_CharT>
+{
+    typedef __owns_two_states<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __alternate(__owns_one_state<_CharT>* __s1,
+                         __owns_one_state<_CharT>* __s2)
+        : base(__s1, __s2) {}
+
+    virtual void __exec(__state& __s) const;
+    virtual void __exec_split(bool __second, __state& __s) const;
+};
+
+template <class _CharT>
+void
+__alternate<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__split;
+}
+
+template <class _CharT>
+void
+__alternate<_CharT>::__exec_split(bool __second, __state& __s) const
+{
+    __s.__do_ = __state::__accept_but_not_consume;
+    if (__second)
+        __s.__node_ = this->second();
+    else
+        __s.__node_ = this->first();
+}
+
+// __begin_marked_subexpression
+
+template <class _CharT>
+class __begin_marked_subexpression
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    unsigned __mexp_;
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __begin_marked_subexpression(unsigned __mexp, __node<_CharT>* __s)
+        : base(__s), __mexp_(__mexp) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__begin_marked_subexpression<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__accept_but_not_consume;
+    __s.__sub_matches_[__mexp_-1].first = __s.__current_;
+    __s.__node_ = this->first();
+}
+
+// __end_marked_subexpression
+
+template <class _CharT>
+class __end_marked_subexpression
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    unsigned __mexp_;
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __end_marked_subexpression(unsigned __mexp, __node<_CharT>* __s)
+        : base(__s), __mexp_(__mexp) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__end_marked_subexpression<_CharT>::__exec(__state& __s) const
+{
+    __s.__do_ = __state::__accept_but_not_consume;
+    __s.__sub_matches_[__mexp_-1].second = __s.__current_;
+    __s.__sub_matches_[__mexp_-1].matched = true;
+    __s.__node_ = this->first();
+}
+
+// __back_ref
+
+template <class _CharT>
+class __back_ref
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    unsigned __mexp_;
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __back_ref(unsigned __mexp, __node<_CharT>* __s)
+        : base(__s), __mexp_(__mexp) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__back_ref<_CharT>::__exec(__state& __s) const
+{
+    if (__mexp_ > __s.__sub_matches_.size())
+        __throw_regex_error<regex_constants::error_backref>();
+    sub_match<const _CharT*>& __sm = __s.__sub_matches_[__mexp_-1];
+    if (__sm.matched)
+    {
+        ptrdiff_t __len = __sm.second - __sm.first;
+        if (__s.__last_ - __s.__current_ >= __len &&
+            _CUDA_VSTD::equal(__sm.first, __sm.second, __s.__current_))
+        {
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__current_ += __len;
+            __s.__node_ = this->first();
+        }
+        else
+        {
+            __s.__do_ = __state::__reject;
+            __s.__node_ = nullptr;
+        }
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __back_ref_icase
+
+template <class _CharT, class _Traits>
+class __back_ref_icase
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    _Traits __traits_;
+    unsigned __mexp_;
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __back_ref_icase(const _Traits& __traits, unsigned __mexp,
+                              __node<_CharT>* __s)
+        : base(__s), __traits_(__traits), __mexp_(__mexp) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT, class _Traits>
+void
+__back_ref_icase<_CharT, _Traits>::__exec(__state& __s) const
+{
+    sub_match<const _CharT*>& __sm = __s.__sub_matches_[__mexp_-1];
+    if (__sm.matched)
+    {
+        ptrdiff_t __len = __sm.second - __sm.first;
+        if (__s.__last_ - __s.__current_ >= __len)
+        {
+            for (ptrdiff_t __i = 0; __i < __len; ++__i)
+            {
+                if (__traits_.translate_nocase(__sm.first[__i]) !=
+                                __traits_.translate_nocase(__s.__current_[__i]))
+                    goto __not_equal;
+            }
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__current_ += __len;
+            __s.__node_ = this->first();
+        }
+        else
+        {
+            __s.__do_ = __state::__reject;
+            __s.__node_ = nullptr;
+        }
+    }
+    else
+    {
+__not_equal:
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __back_ref_collate
+
+template <class _CharT, class _Traits>
+class __back_ref_collate
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    _Traits __traits_;
+    unsigned __mexp_;
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __back_ref_collate(const _Traits& __traits, unsigned __mexp,
+                              __node<_CharT>* __s)
+        : base(__s), __traits_(__traits), __mexp_(__mexp) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT, class _Traits>
+void
+__back_ref_collate<_CharT, _Traits>::__exec(__state& __s) const
+{
+    sub_match<const _CharT*>& __sm = __s.__sub_matches_[__mexp_-1];
+    if (__sm.matched)
+    {
+        ptrdiff_t __len = __sm.second - __sm.first;
+        if (__s.__last_ - __s.__current_ >= __len)
+        {
+            for (ptrdiff_t __i = 0; __i < __len; ++__i)
+            {
+                if (__traits_.translate(__sm.first[__i]) !=
+                                       __traits_.translate(__s.__current_[__i]))
+                    goto __not_equal;
+            }
+            __s.__do_ = __state::__accept_but_not_consume;
+            __s.__current_ += __len;
+            __s.__node_ = this->first();
+        }
+        else
+        {
+            __s.__do_ = __state::__reject;
+            __s.__node_ = nullptr;
+        }
+    }
+    else
+    {
+__not_equal:
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __word_boundary
+
+template <class _CharT, class _Traits>
+class __word_boundary
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    _Traits __traits_;
+    bool __invert_;
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __word_boundary(const _Traits& __traits, bool __invert,
+                             __node<_CharT>* __s)
+        : base(__s), __traits_(__traits), __invert_(__invert) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT, class _Traits>
+void
+__word_boundary<_CharT, _Traits>::__exec(__state& __s) const
+{
+    bool __is_word_b = false;
+    if (__s.__first_ != __s.__last_)
+    {
+        if (__s.__current_ == __s.__last_)
+        {
+            if (!(__s.__flags_ & regex_constants::match_not_eow))
+            {
+                _CharT __c = __s.__current_[-1];
+                __is_word_b = __c == '_' ||
+                              __traits_.isctype(__c, ctype_base::alnum);
+            }
+        }
+        else if (__s.__current_ == __s.__first_ &&
+                !(__s.__flags_ & regex_constants::match_prev_avail))
+        {
+            if (!(__s.__flags_ & regex_constants::match_not_bow))
+            {
+                _CharT __c = *__s.__current_;
+                __is_word_b = __c == '_' ||
+                              __traits_.isctype(__c, ctype_base::alnum);
+            }
+        }
+        else
+        {
+            _CharT __c1 = __s.__current_[-1];
+            _CharT __c2 = *__s.__current_;
+            bool __is_c1_b = __c1 == '_' ||
+                             __traits_.isctype(__c1, ctype_base::alnum);
+            bool __is_c2_b = __c2 == '_' ||
+                             __traits_.isctype(__c2, ctype_base::alnum);
+            __is_word_b = __is_c1_b != __is_c2_b;
+        }
+    }
+    if (__is_word_b != __invert_)
+    {
+        __s.__do_ = __state::__accept_but_not_consume;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __l_anchor
+
+template <class _CharT>
+class __l_anchor
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __l_anchor(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__l_anchor<_CharT>::__exec(__state& __s) const
+{
+    if (__s.__at_first_ && __s.__current_ == __s.__first_ &&
+        !(__s.__flags_ & regex_constants::match_not_bol))
+    {
+        __s.__do_ = __state::__accept_but_not_consume;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __r_anchor
+
+template <class _CharT>
+class __r_anchor
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __r_anchor(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__r_anchor<_CharT>::__exec(__state& __s) const
+{
+    if (__s.__current_ == __s.__last_ &&
+        !(__s.__flags_ & regex_constants::match_not_eol))
+    {
+        __s.__do_ = __state::__accept_but_not_consume;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __match_any
+
+template <class _CharT>
+class __match_any
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __match_any(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__match_any<_CharT>::__exec(__state& __s) const
+{
+    if (__s.__current_ != __s.__last_ && *__s.__current_ != 0)
+    {
+        __s.__do_ = __state::__accept_and_consume;
+        ++__s.__current_;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __match_any_but_newline
+
+template <class _CharT>
+class __match_any_but_newline
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __match_any_but_newline(__node<_CharT>* __s)
+        : base(__s) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <> _LIBCUDACXX_FUNC_VIS void __match_any_but_newline<char>::__exec(__state&) const;
+template <> _LIBCUDACXX_FUNC_VIS void __match_any_but_newline<wchar_t>::__exec(__state&) const;
+
+// __match_char
+
+template <class _CharT>
+class __match_char
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    _CharT __c_;
+
+    __match_char(const __match_char&);
+    __match_char& operator=(const __match_char&);
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __match_char(_CharT __c, __node<_CharT>* __s)
+        : base(__s), __c_(__c) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT>
+void
+__match_char<_CharT>::__exec(__state& __s) const
+{
+    if (__s.__current_ != __s.__last_ && *__s.__current_ == __c_)
+    {
+        __s.__do_ = __state::__accept_and_consume;
+        ++__s.__current_;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __match_char_icase
+
+template <class _CharT, class _Traits>
+class __match_char_icase
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    _Traits __traits_;
+    _CharT __c_;
+
+    __match_char_icase(const __match_char_icase&);
+    __match_char_icase& operator=(const __match_char_icase&);
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __match_char_icase(const _Traits& __traits, _CharT __c, __node<_CharT>* __s)
+        : base(__s), __traits_(__traits), __c_(__traits.translate_nocase(__c)) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT, class _Traits>
+void
+__match_char_icase<_CharT, _Traits>::__exec(__state& __s) const
+{
+    if (__s.__current_ != __s.__last_ &&
+        __traits_.translate_nocase(*__s.__current_) == __c_)
+    {
+        __s.__do_ = __state::__accept_and_consume;
+        ++__s.__current_;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __match_char_collate
+
+template <class _CharT, class _Traits>
+class __match_char_collate
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    _Traits __traits_;
+    _CharT __c_;
+
+    __match_char_collate(const __match_char_collate&);
+    __match_char_collate& operator=(const __match_char_collate&);
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __match_char_collate(const _Traits& __traits, _CharT __c, __node<_CharT>* __s)
+        : base(__s), __traits_(__traits), __c_(__traits.translate(__c)) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT, class _Traits>
+void
+__match_char_collate<_CharT, _Traits>::__exec(__state& __s) const
+{
+    if (__s.__current_ != __s.__last_ &&
+        __traits_.translate(*__s.__current_) == __c_)
+    {
+        __s.__do_ = __state::__accept_and_consume;
+        ++__s.__current_;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+// __bracket_expression
+
+template <class _CharT, class _Traits>
+class __bracket_expression
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+    typedef typename _Traits::string_type string_type;
+
+    _Traits __traits_;
+    vector<_CharT> __chars_;
+    vector<_CharT> __neg_chars_;
+    vector<pair<string_type, string_type> > __ranges_;
+    vector<pair<_CharT, _CharT> > __digraphs_;
+    vector<string_type> __equivalences_;
+    typename regex_traits<_CharT>::char_class_type __mask_;
+    typename regex_traits<_CharT>::char_class_type __neg_mask_;
+    bool __negate_;
+    bool __icase_;
+    bool __collate_;
+    bool __might_have_digraph_;
+
+    __bracket_expression(const __bracket_expression&);
+    __bracket_expression& operator=(const __bracket_expression&);
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __bracket_expression(const _Traits& __traits, __node<_CharT>* __s,
+                                 bool __negate, bool __icase, bool __collate)
+        : base(__s), __traits_(__traits), __mask_(), __neg_mask_(),
+          __negate_(__negate), __icase_(__icase), __collate_(__collate),
+          __might_have_digraph_(__traits_.getloc().name() != "C") {}
+
+    virtual void __exec(__state&) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __negated() const {return __negate_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_char(_CharT __c)
+        {
+            if (__icase_)
+                __chars_.push_back(__traits_.translate_nocase(__c));
+            else if (__collate_)
+                __chars_.push_back(__traits_.translate(__c));
+            else
+                __chars_.push_back(__c);
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_neg_char(_CharT __c)
+        {
+            if (__icase_)
+                __neg_chars_.push_back(__traits_.translate_nocase(__c));
+            else if (__collate_)
+                __neg_chars_.push_back(__traits_.translate(__c));
+            else
+                __neg_chars_.push_back(__c);
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_range(string_type __b, string_type __e)
+        {
+            if (__collate_)
+            {
+                if (__icase_)
+                {
+                    for (size_t __i = 0; __i < __b.size(); ++__i)
+                        __b[__i] = __traits_.translate_nocase(__b[__i]);
+                    for (size_t __i = 0; __i < __e.size(); ++__i)
+                        __e[__i] = __traits_.translate_nocase(__e[__i]);
+                }
+                else
+                {
+                    for (size_t __i = 0; __i < __b.size(); ++__i)
+                        __b[__i] = __traits_.translate(__b[__i]);
+                    for (size_t __i = 0; __i < __e.size(); ++__i)
+                        __e[__i] = __traits_.translate(__e[__i]);
+                }
+                __ranges_.push_back(make_pair(
+                                  __traits_.transform(__b.begin(), __b.end()),
+                                  __traits_.transform(__e.begin(), __e.end())));
+            }
+            else
+            {
+                if (__b.size() != 1 || __e.size() != 1)
+                    __throw_regex_error<regex_constants::error_range>();
+                if (__icase_)
+                {
+                    __b[0] = __traits_.translate_nocase(__b[0]);
+                    __e[0] = __traits_.translate_nocase(__e[0]);
+                }
+                __ranges_.push_back(make_pair(_CUDA_VSTD::move(__b), _CUDA_VSTD::move(__e)));
+            }
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_digraph(_CharT __c1, _CharT __c2)
+        {
+            if (__icase_)
+                __digraphs_.push_back(make_pair(__traits_.translate_nocase(__c1),
+                                                __traits_.translate_nocase(__c2)));
+            else if (__collate_)
+                __digraphs_.push_back(make_pair(__traits_.translate(__c1),
+                                                __traits_.translate(__c2)));
+            else
+                __digraphs_.push_back(make_pair(__c1, __c2));
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_equivalence(const string_type& __s)
+        {__equivalences_.push_back(__s);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_class(typename regex_traits<_CharT>::char_class_type __mask)
+        {__mask_ |= __mask;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __add_neg_class(typename regex_traits<_CharT>::char_class_type __mask)
+        {__neg_mask_ |= __mask;}
+};
+
+template <class _CharT, class _Traits>
+void
+__bracket_expression<_CharT, _Traits>::__exec(__state& __s) const
+{
+    bool __found = false;
+    unsigned __consumed = 0;
+    if (__s.__current_ != __s.__last_)
+    {
+        ++__consumed;
+        if (__might_have_digraph_)
+        {
+            const _CharT* __next = _CUDA_VSTD::next(__s.__current_);
+            if (__next != __s.__last_)
+            {
+                pair<_CharT, _CharT> __ch2(*__s.__current_, *__next);
+                if (__icase_)
+                {
+                    __ch2.first = __traits_.translate_nocase(__ch2.first);
+                    __ch2.second = __traits_.translate_nocase(__ch2.second);
+                }
+                else if (__collate_)
+                {
+                    __ch2.first = __traits_.translate(__ch2.first);
+                    __ch2.second = __traits_.translate(__ch2.second);
+                }
+                if (!__traits_.lookup_collatename(&__ch2.first, &__ch2.first+2).empty())
+                {
+                    // __ch2 is a digraph in this locale
+                    ++__consumed;
+                    for (size_t __i = 0; __i < __digraphs_.size(); ++__i)
+                    {
+                        if (__ch2 == __digraphs_[__i])
+                        {
+                            __found = true;
+                            goto __exit;
+                        }
+                    }
+                    if (__collate_ && !__ranges_.empty())
+                    {
+                        string_type __s2 = __traits_.transform(&__ch2.first,
+                                                               &__ch2.first + 2);
+                        for (size_t __i = 0; __i < __ranges_.size(); ++__i)
+                        {
+                            if (__ranges_[__i].first <= __s2 &&
+                                __s2 <= __ranges_[__i].second)
+                            {
+                                __found = true;
+                                goto __exit;
+                            }
+                        }
+                    }
+                    if (!__equivalences_.empty())
+                    {
+                        string_type __s2 = __traits_.transform_primary(&__ch2.first,
+                                                                       &__ch2.first + 2);
+                        for (size_t __i = 0; __i < __equivalences_.size(); ++__i)
+                        {
+                            if (__s2 == __equivalences_[__i])
+                            {
+                                __found = true;
+                                goto __exit;
+                            }
+                        }
+                    }
+                    if (__traits_.isctype(__ch2.first, __mask_) &&
+                        __traits_.isctype(__ch2.second, __mask_))
+                    {
+                        __found = true;
+                        goto __exit;
+                    }
+                    if (!__traits_.isctype(__ch2.first, __neg_mask_) &&
+                        !__traits_.isctype(__ch2.second, __neg_mask_))
+                    {
+                        __found = true;
+                        goto __exit;
+                    }
+                    goto __exit;
+                }
+            }
+        }
+        // test *__s.__current_ as not a digraph
+        _CharT __ch = *__s.__current_;
+        if (__icase_)
+            __ch = __traits_.translate_nocase(__ch);
+        else if (__collate_)
+            __ch = __traits_.translate(__ch);
+        for (size_t __i = 0; __i < __chars_.size(); ++__i)
+        {
+            if (__ch == __chars_[__i])
+            {
+                __found = true;
+                goto __exit;
+            }
+        }
+        // When there's at least one of __neg_chars_ and __neg_mask_, the set
+        // of "__found" chars is
+        //   union(complement(union(__neg_chars_, __neg_mask_)),
+        //         other cases...)
+        //
+        // It doesn't make sense to check this when there are no __neg_chars_
+        // and no __neg_mask_.
+        if (!(__neg_mask_ == 0 && __neg_chars_.empty()))
+        {
+            const bool __in_neg_mask = __traits_.isctype(__ch, __neg_mask_);
+          const bool __in_neg_chars =
+              std::find(__neg_chars_.begin(), __neg_chars_.end(), __ch) !=
+              __neg_chars_.end();
+          if (!(__in_neg_mask || __in_neg_chars))
+          {
+            __found = true;
+            goto __exit;
+          }
+        }
+        if (!__ranges_.empty())
+        {
+            string_type __s2 = __collate_ ?
+                                   __traits_.transform(&__ch, &__ch + 1) :
+                                   string_type(1, __ch);
+            for (size_t __i = 0; __i < __ranges_.size(); ++__i)
+            {
+                if (__ranges_[__i].first <= __s2 && __s2 <= __ranges_[__i].second)
+                {
+                    __found = true;
+                    goto __exit;
+                }
+            }
+        }
+        if (!__equivalences_.empty())
+        {
+            string_type __s2 = __traits_.transform_primary(&__ch, &__ch + 1);
+            for (size_t __i = 0; __i < __equivalences_.size(); ++__i)
+            {
+                if (__s2 == __equivalences_[__i])
+                {
+                    __found = true;
+                    goto __exit;
+                }
+            }
+        }
+        if (__traits_.isctype(__ch, __mask_))
+        {
+            __found = true;
+            goto __exit;
+        }
+    }
+    else
+        __found = __negate_;  // force reject
+__exit:
+    if (__found != __negate_)
+    {
+        __s.__do_ = __state::__accept_and_consume;
+        __s.__current_ += __consumed;
+        __s.__node_ = this->first();
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+template <class _CharT, class _Traits> class __lookahead;
+
+template <class _CharT, class _Traits = regex_traits<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS basic_regex
+{
+public:
+    // types:
+    typedef _CharT                              value_type;
+    typedef _Traits                             traits_type;
+    typedef typename _Traits::string_type       string_type;
+    typedef regex_constants::syntax_option_type flag_type;
+    typedef typename _Traits::locale_type       locale_type;
+
+private:
+    _Traits   __traits_;
+    flag_type __flags_;
+    unsigned __marked_count_;
+    unsigned __loop_count_;
+    int __open_count_;
+    shared_ptr<__empty_state<_CharT> > __start_;
+    __owns_one_state<_CharT>* __end_;
+
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+    typedef _CUDA_VSTD::__node<_CharT> __node;
+
+public:
+    // constants:
+    static const regex_constants::syntax_option_type icase = regex_constants::icase;
+    static const regex_constants::syntax_option_type nosubs = regex_constants::nosubs;
+    static const regex_constants::syntax_option_type optimize = regex_constants::optimize;
+    static const regex_constants::syntax_option_type collate = regex_constants::collate;
+    static const regex_constants::syntax_option_type ECMAScript = regex_constants::ECMAScript;
+    static const regex_constants::syntax_option_type basic = regex_constants::basic;
+    static const regex_constants::syntax_option_type extended = regex_constants::extended;
+    static const regex_constants::syntax_option_type awk = regex_constants::awk;
+    static const regex_constants::syntax_option_type grep = regex_constants::grep;
+    static const regex_constants::syntax_option_type egrep = regex_constants::egrep;
+
+    // construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex()
+        : __flags_(regex_constants::ECMAScript), __marked_count_(0), __loop_count_(0), __open_count_(0),
+          __end_(0)
+        {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_regex(const value_type* __p, flag_type __f = regex_constants::ECMAScript)
+        : __flags_(__f), __marked_count_(0), __loop_count_(0), __open_count_(0),
+          __end_(0)
+        {
+        if (__get_grammar(__flags_) == 0) __flags_ |= regex_constants::ECMAScript;
+        __parse(__p, __p + __traits_.length(__p));
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex(const value_type* __p, size_t __len, flag_type __f = regex_constants::ECMAScript)
+        : __flags_(__f), __marked_count_(0), __loop_count_(0), __open_count_(0),
+          __end_(0)
+        {
+        if (__get_grammar(__flags_) == 0) __flags_ |= regex_constants::ECMAScript;
+        __parse(__p, __p + __len);
+        }
+
+//     basic_regex(const basic_regex&) = default;
+//     basic_regex(basic_regex&&) = default;
+    template <class _ST, class _SA>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit basic_regex(const basic_string<value_type, _ST, _SA>& __p,
+                             flag_type __f = regex_constants::ECMAScript)
+        : __flags_(__f), __marked_count_(0), __loop_count_(0), __open_count_(0),
+          __end_(0)
+        {
+        if (__get_grammar(__flags_) == 0) __flags_ |= regex_constants::ECMAScript;
+        __parse(__p.begin(), __p.end());
+        }
+
+    template <class _ForwardIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_regex(_ForwardIterator __first, _ForwardIterator __last,
+                    flag_type __f = regex_constants::ECMAScript)
+        : __flags_(__f), __marked_count_(0), __loop_count_(0), __open_count_(0),
+          __end_(0)
+        {
+        if (__get_grammar(__flags_) == 0) __flags_ |= regex_constants::ECMAScript;
+        __parse(__first, __last);
+        }
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex(initializer_list<value_type> __il,
+                flag_type __f = regex_constants::ECMAScript)
+        : __flags_(__f), __marked_count_(0), __loop_count_(0), __open_count_(0),
+          __end_(0)
+        {
+        if (__get_grammar(__flags_) == 0) __flags_ |= regex_constants::ECMAScript;
+        __parse(__il.begin(), __il.end());
+        }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+//    ~basic_regex() = default;
+
+//     basic_regex& operator=(const basic_regex&) = default;
+//     basic_regex& operator=(basic_regex&&) = default;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& operator=(const value_type* __p)
+        {return assign(__p);}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& operator=(initializer_list<value_type> __il)
+        {return assign(__il);}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    template <class _ST, class _SA>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_regex& operator=(const basic_string<value_type, _ST, _SA>& __p)
+        {return assign(__p);}
+
+    // assign:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& assign(const basic_regex& __that)
+        {return *this = __that;}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& assign(basic_regex&& __that) _NOEXCEPT
+        {return *this = _CUDA_VSTD::move(__that);}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& assign(const value_type* __p, flag_type __f = regex_constants::ECMAScript)
+        {return assign(__p, __p + __traits_.length(__p), __f);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& assign(const value_type* __p, size_t __len, flag_type __f = regex_constants::ECMAScript)
+        {return assign(__p, __p + __len, __f);}
+    template <class _ST, class _SA>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_regex& assign(const basic_string<value_type, _ST, _SA>& __s,
+                            flag_type __f = regex_constants::ECMAScript)
+            {return assign(__s.begin(), __s.end(), __f);}
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+             __is_input_iterator  <_InputIterator>::value &&
+            !__is_forward_iterator<_InputIterator>::value,
+            basic_regex&
+        >::type
+        assign(_InputIterator __first, _InputIterator __last,
+                            flag_type __f = regex_constants::ECMAScript)
+        {
+            basic_string<_CharT> __t(__first, __last);
+            return assign(__t.begin(), __t.end(), __f);
+        }
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __member_init(flag_type __f)
+    {
+        __flags_ = __f;
+        __marked_count_ = 0;
+        __loop_count_ = 0;
+        __open_count_ = 0;
+        __end_ = nullptr;
+    }
+public:
+
+    template <class _ForwardIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value,
+            basic_regex&
+        >::type
+        assign(_ForwardIterator __first, _ForwardIterator __last,
+                            flag_type __f = regex_constants::ECMAScript)
+        {
+            return assign(basic_regex(__first, __last, __f));
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_regex& assign(initializer_list<value_type> __il,
+                        flag_type __f = regex_constants::ECMAScript)
+        {return assign(__il.begin(), __il.end(), __f);}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    // const operations:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned mark_count() const {return __marked_count_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    flag_type flags() const {return __flags_;}
+
+    // locale:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    locale_type imbue(locale_type __loc)
+    {
+        __member_init(ECMAScript);
+        __start_.reset();
+        return __traits_.imbue(__loc);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    locale_type getloc() const {return __traits_.getloc();}
+
+    // swap:
+    void swap(basic_regex& __r);
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unsigned __loop_count() const {return __loop_count_;}
+
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_basic_reg_exp(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_RE_expression(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_simple_RE(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_nondupl_RE(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_one_char_or_coll_elem_RE(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_Back_open_paren(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_Back_close_paren(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_Back_open_brace(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_Back_close_brace(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_BACKREF(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_ORD_CHAR(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_QUOTED_CHAR(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_RE_dupl_symbol(_ForwardIterator __first, _ForwardIterator __last,
+                               __owns_one_state<_CharT>* __s,
+                               unsigned __mexp_begin, unsigned __mexp_end);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_ERE_dupl_symbol(_ForwardIterator __first, _ForwardIterator __last,
+                                __owns_one_state<_CharT>* __s,
+                                unsigned __mexp_begin, unsigned __mexp_end);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_bracket_expression(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_follow_list(_ForwardIterator __first, _ForwardIterator __last,
+                            __bracket_expression<_CharT, _Traits>* __ml);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_expression_term(_ForwardIterator __first, _ForwardIterator __last,
+                                __bracket_expression<_CharT, _Traits>* __ml);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_equivalence_class(_ForwardIterator __first, _ForwardIterator __last,
+                                  __bracket_expression<_CharT, _Traits>* __ml);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_character_class(_ForwardIterator __first, _ForwardIterator __last,
+                                __bracket_expression<_CharT, _Traits>* __ml);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_collating_symbol(_ForwardIterator __first, _ForwardIterator __last,
+                                 basic_string<_CharT>& __col_sym);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_DUP_COUNT(_ForwardIterator __first, _ForwardIterator __last, int& __c);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_extended_reg_exp(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_ERE_branch(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_ERE_expression(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_one_char_or_coll_elem_ERE(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_ORD_CHAR_ERE(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_QUOTED_CHAR_ERE(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_ecma_exp(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_alternative(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_term(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_assertion(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_atom(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_atom_escape(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_decimal_escape(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_character_class_escape(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_character_escape(_ForwardIterator __first, _ForwardIterator __last,
+                                 basic_string<_CharT>* __str = nullptr);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_pattern_character(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_grep(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_egrep(_ForwardIterator __first, _ForwardIterator __last);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_class_escape(_ForwardIterator __first, _ForwardIterator __last,
+                          basic_string<_CharT>& __str,
+                          __bracket_expression<_CharT, _Traits>* __ml);
+    template <class _ForwardIterator>
+        _ForwardIterator
+        __parse_awk_escape(_ForwardIterator __first, _ForwardIterator __last,
+                          basic_string<_CharT>* __str = nullptr);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __push_l_anchor();
+    void __push_r_anchor();
+    void __push_match_any();
+    void __push_match_any_but_newline();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __push_greedy_inf_repeat(size_t __min, __owns_one_state<_CharT>* __s,
+                                  unsigned __mexp_begin = 0, unsigned __mexp_end = 0)
+        {__push_loop(__min, numeric_limits<size_t>::max(), __s,
+                     __mexp_begin, __mexp_end);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __push_nongreedy_inf_repeat(size_t __min, __owns_one_state<_CharT>* __s,
+                                  unsigned __mexp_begin = 0, unsigned __mexp_end = 0)
+        {__push_loop(__min, numeric_limits<size_t>::max(), __s,
+                     __mexp_begin, __mexp_end, false);}
+    void __push_loop(size_t __min, size_t __max, __owns_one_state<_CharT>* __s,
+                     size_t __mexp_begin = 0, size_t __mexp_end = 0,
+                     bool __greedy = true);
+    __bracket_expression<_CharT, _Traits>* __start_matching_list(bool __negate);
+    void __push_char(value_type __c);
+    void __push_back_ref(int __i);
+    void __push_alternation(__owns_one_state<_CharT>* __sa,
+                            __owns_one_state<_CharT>* __sb);
+    void __push_begin_marked_subexpression();
+    void __push_end_marked_subexpression(unsigned);
+    void __push_empty();
+    void __push_word_boundary(bool);
+    void __push_lookahead(const basic_regex&, bool, unsigned);
+
+    template <class _Allocator>
+        bool
+        __search(const _CharT* __first, const _CharT* __last,
+                 match_results<const _CharT*, _Allocator>& __m,
+                 regex_constants::match_flag_type __flags) const;
+
+    template <class _Allocator>
+        bool
+        __match_at_start(const _CharT* __first, const _CharT* __last,
+                 match_results<const _CharT*, _Allocator>& __m,
+                 regex_constants::match_flag_type __flags, bool) const;
+    template <class _Allocator>
+        bool
+        __match_at_start_ecma(const _CharT* __first, const _CharT* __last,
+                 match_results<const _CharT*, _Allocator>& __m,
+                 regex_constants::match_flag_type __flags, bool) const;
+    template <class _Allocator>
+        bool
+        __match_at_start_posix_nosubs(const _CharT* __first, const _CharT* __last,
+                 match_results<const _CharT*, _Allocator>& __m,
+                 regex_constants::match_flag_type __flags, bool) const;
+    template <class _Allocator>
+        bool
+        __match_at_start_posix_subs(const _CharT* __first, const _CharT* __last,
+                 match_results<const _CharT*, _Allocator>& __m,
+                 regex_constants::match_flag_type __flags, bool) const;
+
+    template <class _Bp, class _Ap, class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(_Bp, _Bp, match_results<_Bp, _Ap>&, const basic_regex<_Cp, _Tp>&,
+                 regex_constants::match_flag_type);
+
+    template <class _Ap, class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(const _Cp*, const _Cp*, match_results<const _Cp*, _Ap>&,
+                 const basic_regex<_Cp, _Tp>&, regex_constants::match_flag_type);
+
+    template <class _Bp, class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(_Bp, _Bp, const basic_regex<_Cp, _Tp>&,
+                 regex_constants::match_flag_type);
+
+    template <class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(const _Cp*, const _Cp*,
+                 const basic_regex<_Cp, _Tp>&, regex_constants::match_flag_type);
+
+    template <class _Cp, class _Ap, class _Tp>
+    friend
+    bool
+    regex_search(const _Cp*, match_results<const _Cp*, _Ap>&, const basic_regex<_Cp, _Tp>&,
+                 regex_constants::match_flag_type);
+
+    template <class _ST, class _SA, class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(const basic_string<_Cp, _ST, _SA>& __s,
+                 const basic_regex<_Cp, _Tp>& __e,
+                 regex_constants::match_flag_type __flags);
+
+    template <class _ST, class _SA, class _Ap, class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(const basic_string<_Cp, _ST, _SA>& __s,
+                 match_results<typename basic_string<_Cp, _ST, _SA>::const_iterator, _Ap>&,
+                 const basic_regex<_Cp, _Tp>& __e,
+                 regex_constants::match_flag_type __flags);
+
+    template <class _Iter, class _Ap, class _Cp, class _Tp>
+    friend
+    bool
+    regex_search(__wrap_iter<_Iter> __first,
+                 __wrap_iter<_Iter> __last,
+                 match_results<__wrap_iter<_Iter>, _Ap>& __m,
+                 const basic_regex<_Cp, _Tp>& __e,
+                 regex_constants::match_flag_type __flags);
+
+    template <class, class> friend class __lookahead;
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template <class _ForwardIterator,
+          class = typename enable_if<__is_forward_iterator<_ForwardIterator>::value, nullptr_t>::type
+>
+basic_regex(_ForwardIterator, _ForwardIterator,
+            regex_constants::syntax_option_type = regex_constants::ECMAScript)
+    -> basic_regex<typename iterator_traits<_ForwardIterator>::value_type>;
+#endif
+
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::icase;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::nosubs;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::optimize;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::collate;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::ECMAScript;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::basic;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::extended;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::awk;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::grep;
+template <class _CharT, class _Traits>
+    const regex_constants::syntax_option_type basic_regex<_CharT, _Traits>::egrep;
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::swap(basic_regex& __r)
+{
+    using _CUDA_VSTD::swap;
+    swap(__traits_, __r.__traits_);
+    swap(__flags_, __r.__flags_);
+    swap(__marked_count_, __r.__marked_count_);
+    swap(__loop_count_, __r.__loop_count_);
+    swap(__open_count_, __r.__open_count_);
+    swap(__start_, __r.__start_);
+    swap(__end_, __r.__end_);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_regex<_CharT, _Traits>& __x, basic_regex<_CharT, _Traits>& __y)
+{
+    return __x.swap(__y);
+}
+
+// __lookahead
+
+template <class _CharT, class _Traits>
+class __lookahead
+    : public __owns_one_state<_CharT>
+{
+    typedef __owns_one_state<_CharT> base;
+
+    basic_regex<_CharT, _Traits> __exp_;
+    unsigned __mexp_;
+    bool __invert_;
+
+    __lookahead(const __lookahead&);
+    __lookahead& operator=(const __lookahead&);
+public:
+    typedef _CUDA_VSTD::__state<_CharT> __state;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __lookahead(const basic_regex<_CharT, _Traits>& __exp, bool __invert, __node<_CharT>* __s, unsigned __mexp)
+        : base(__s), __exp_(__exp), __mexp_(__mexp), __invert_(__invert) {}
+
+    virtual void __exec(__state&) const;
+};
+
+template <class _CharT, class _Traits>
+void
+__lookahead<_CharT, _Traits>::__exec(__state& __s) const
+{
+    match_results<const _CharT*> __m;
+    __m.__init(1 + __exp_.mark_count(), __s.__current_, __s.__last_);
+    bool __matched = __exp_.__match_at_start_ecma(
+        __s.__current_, __s.__last_,
+        __m,
+        (__s.__flags_ | regex_constants::match_continuous) &
+        ~regex_constants::__full_match,
+        __s.__at_first_ && __s.__current_ == __s.__first_);
+    if (__matched != __invert_)
+    {
+        __s.__do_ = __state::__accept_but_not_consume;
+        __s.__node_ = this->first();
+        for (unsigned __i = 1; __i < __m.size(); ++__i) {
+            __s.__sub_matches_[__mexp_ + __i - 1] = __m.__matches_[__i];
+        }
+    }
+    else
+    {
+        __s.__do_ = __state::__reject;
+        __s.__node_ = nullptr;
+    }
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse(_ForwardIterator __first,
+                                      _ForwardIterator __last)
+{
+    {
+        unique_ptr<__node> __h(new __end_state<_CharT>);
+        __start_.reset(new __empty_state<_CharT>(__h.get()));
+        __h.release();
+        __end_ = __start_.get();
+    }
+    switch (__get_grammar(__flags_))
+    {
+    case ECMAScript:
+        __first = __parse_ecma_exp(__first, __last);
+        break;
+    case basic:
+        __first = __parse_basic_reg_exp(__first, __last);
+        break;
+    case extended:
+    case awk:
+        __first = __parse_extended_reg_exp(__first, __last);
+        break;
+    case grep:
+        __first = __parse_grep(__first, __last);
+        break;
+    case egrep:
+        __first = __parse_egrep(__first, __last);
+        break;
+    default:
+        __throw_regex_error<regex_constants::__re_err_grammar>();
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_basic_reg_exp(_ForwardIterator __first,
+                                                    _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        if (*__first == '^')
+        {
+            __push_l_anchor();
+            ++__first;
+        }
+        if (__first != __last)
+        {
+            __first = __parse_RE_expression(__first, __last);
+            if (__first != __last)
+            {
+                _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+                if (__temp == __last && *__first == '$')
+                {
+                    __push_r_anchor();
+                    ++__first;
+                }
+            }
+        }
+        if (__first != __last)
+            __throw_regex_error<regex_constants::__re_err_empty>();
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_extended_reg_exp(_ForwardIterator __first,
+                                                       _ForwardIterator __last)
+{
+    __owns_one_state<_CharT>* __sa = __end_;
+    _ForwardIterator __temp = __parse_ERE_branch(__first, __last);
+    if (__temp == __first)
+        __throw_regex_error<regex_constants::__re_err_empty>();
+    __first = __temp;
+    while (__first != __last && *__first == '|')
+    {
+        __owns_one_state<_CharT>* __sb = __end_;
+        __temp = __parse_ERE_branch(++__first, __last);
+        if (__temp == __first)
+            __throw_regex_error<regex_constants::__re_err_empty>();
+        __push_alternation(__sa, __sb);
+        __first = __temp;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_ERE_branch(_ForwardIterator __first,
+                                                 _ForwardIterator __last)
+{
+    _ForwardIterator __temp = __parse_ERE_expression(__first, __last);
+    if (__temp == __first)
+        __throw_regex_error<regex_constants::__re_err_empty>();
+    do
+    {
+        __first = __temp;
+        __temp = __parse_ERE_expression(__first, __last);
+    } while (__temp != __first);
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_ERE_expression(_ForwardIterator __first,
+                                                     _ForwardIterator __last)
+{
+    __owns_one_state<_CharT>* __e = __end_;
+    unsigned __mexp_begin = __marked_count_;
+    _ForwardIterator __temp = __parse_one_char_or_coll_elem_ERE(__first, __last);
+    if (__temp == __first && __temp != __last)
+    {
+        switch (*__temp)
+        {
+        case '^':
+            __push_l_anchor();
+            ++__temp;
+            break;
+        case '$':
+            __push_r_anchor();
+            ++__temp;
+            break;
+        case '(':
+            __push_begin_marked_subexpression();
+            unsigned __temp_count = __marked_count_;
+            ++__open_count_;
+            __temp = __parse_extended_reg_exp(++__temp, __last);
+            if (__temp == __last || *__temp != ')')
+                __throw_regex_error<regex_constants::error_paren>();
+            __push_end_marked_subexpression(__temp_count);
+            --__open_count_;
+            ++__temp;
+            break;
+        }
+    }
+    if (__temp != __first)
+        __temp = __parse_ERE_dupl_symbol(__temp, __last, __e, __mexp_begin+1,
+                                         __marked_count_+1);
+    __first = __temp;
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_RE_expression(_ForwardIterator __first,
+                                                    _ForwardIterator __last)
+{
+    while (true)
+    {
+        _ForwardIterator __temp = __parse_simple_RE(__first, __last);
+        if (__temp == __first)
+            break;
+        __first = __temp;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_simple_RE(_ForwardIterator __first,
+                                                _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        __owns_one_state<_CharT>* __e = __end_;
+        unsigned __mexp_begin = __marked_count_;
+        _ForwardIterator __temp = __parse_nondupl_RE(__first, __last);
+        if (__temp != __first)
+            __first = __parse_RE_dupl_symbol(__temp, __last, __e,
+                                             __mexp_begin+1, __marked_count_+1);
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_nondupl_RE(_ForwardIterator __first,
+                                                 _ForwardIterator __last)
+{
+    _ForwardIterator __temp = __first;
+    __first = __parse_one_char_or_coll_elem_RE(__first, __last);
+    if (__temp == __first)
+    {
+        __temp = __parse_Back_open_paren(__first, __last);
+        if (__temp != __first)
+        {
+            __push_begin_marked_subexpression();
+            unsigned __temp_count = __marked_count_;
+            __first = __parse_RE_expression(__temp, __last);
+            __temp = __parse_Back_close_paren(__first, __last);
+            if (__temp == __first)
+                __throw_regex_error<regex_constants::error_paren>();
+            __push_end_marked_subexpression(__temp_count);
+            __first = __temp;
+        }
+        else
+            __first = __parse_BACKREF(__first, __last);
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_one_char_or_coll_elem_RE(
+                                                       _ForwardIterator __first,
+                                                       _ForwardIterator __last)
+{
+    _ForwardIterator __temp = __parse_ORD_CHAR(__first, __last);
+    if (__temp == __first)
+    {
+        __temp = __parse_QUOTED_CHAR(__first, __last);
+        if (__temp == __first)
+        {
+            if (__temp != __last && *__temp == '.')
+            {
+                __push_match_any();
+                ++__temp;
+            }
+            else
+                __temp = __parse_bracket_expression(__first, __last);
+        }
+    }
+    __first = __temp;
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_one_char_or_coll_elem_ERE(
+                                                       _ForwardIterator __first,
+                                                       _ForwardIterator __last)
+{
+    _ForwardIterator __temp = __parse_ORD_CHAR_ERE(__first, __last);
+    if (__temp == __first)
+    {
+        __temp = __parse_QUOTED_CHAR_ERE(__first, __last);
+        if (__temp == __first)
+        {
+            if (__temp != __last && *__temp == '.')
+            {
+                __push_match_any();
+                ++__temp;
+            }
+            else
+                __temp = __parse_bracket_expression(__first, __last);
+        }
+    }
+    __first = __temp;
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_Back_open_paren(_ForwardIterator __first,
+                                                      _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\' && *__temp == '(')
+                __first = ++__temp;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_Back_close_paren(_ForwardIterator __first,
+                                                       _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\' && *__temp == ')')
+                __first = ++__temp;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_Back_open_brace(_ForwardIterator __first,
+                                                      _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\' && *__temp == '{')
+                __first = ++__temp;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_Back_close_brace(_ForwardIterator __first,
+                                                       _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\' && *__temp == '}')
+                __first = ++__temp;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_BACKREF(_ForwardIterator __first,
+                                              _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\')
+            {
+                int __val = __traits_.value(*__temp, 10);
+                if (__val >= 1 && __val <= 9)
+                {
+                    __push_back_ref(__val);
+                    __first = ++__temp;
+                }
+            }
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_ORD_CHAR(_ForwardIterator __first,
+                                               _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp == __last && *__first == '$')
+            return __first;
+        // Not called inside a bracket
+        if (*__first == '.' || *__first == '\\' || *__first == '[')
+            return __first;
+        __push_char(*__first);
+        ++__first;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_ORD_CHAR_ERE(_ForwardIterator __first,
+                                                   _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        switch (*__first)
+        {
+        case '^':
+        case '.':
+        case '[':
+        case '$':
+        case '(':
+        case '|':
+        case '*':
+        case '+':
+        case '?':
+        case '{':
+        case '\\':
+            break;
+        case ')':
+            if (__open_count_ == 0)
+            {
+                __push_char(*__first);
+                ++__first;
+            }
+            break;
+        default:
+            __push_char(*__first);
+            ++__first;
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_QUOTED_CHAR(_ForwardIterator __first,
+                                                  _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\')
+            {
+                switch (*__temp)
+                {
+                case '^':
+                case '.':
+                case '*':
+                case '[':
+                case '$':
+                case '\\':
+                    __push_char(*__temp);
+                    __first = ++__temp;
+                    break;
+                }
+            }
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_QUOTED_CHAR_ERE(_ForwardIterator __first,
+                                                      _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        if (__temp != __last)
+        {
+            if (*__first == '\\')
+            {
+                switch (*__temp)
+                {
+                case '^':
+                case '.':
+                case '*':
+                case '[':
+                case '$':
+                case '\\':
+                case '(':
+                case ')':
+                case '|':
+                case '+':
+                case '?':
+                case '{':
+                case '}':
+                    __push_char(*__temp);
+                    __first = ++__temp;
+                    break;
+                default:
+                    if (__get_grammar(__flags_) == awk)
+                        __first = __parse_awk_escape(++__first, __last);
+                    break;
+                }
+            }
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_RE_dupl_symbol(_ForwardIterator __first,
+                                                     _ForwardIterator __last,
+                                                     __owns_one_state<_CharT>* __s,
+                                                     unsigned __mexp_begin,
+                                                     unsigned __mexp_end)
+{
+    if (__first != __last)
+    {
+        if (*__first == '*')
+        {
+            __push_greedy_inf_repeat(0, __s, __mexp_begin, __mexp_end);
+            ++__first;
+        }
+        else
+        {
+            _ForwardIterator __temp = __parse_Back_open_brace(__first, __last);
+            if (__temp != __first)
+            {
+                int __min = 0;
+                __first = __temp;
+                __temp = __parse_DUP_COUNT(__first, __last, __min);
+                if (__temp == __first)
+                    __throw_regex_error<regex_constants::error_badbrace>();
+                __first = __temp;
+                if (__first == __last)
+                    __throw_regex_error<regex_constants::error_brace>();
+                if (*__first != ',')
+                {
+                    __temp = __parse_Back_close_brace(__first, __last);
+                    if (__temp == __first)
+                        __throw_regex_error<regex_constants::error_brace>();
+                    __push_loop(__min, __min, __s, __mexp_begin, __mexp_end,
+                                    true);
+                    __first = __temp;
+                }
+                else
+                {
+                    ++__first;  // consume ','
+                    int __max = -1;
+                    __first = __parse_DUP_COUNT(__first, __last, __max);
+                    __temp = __parse_Back_close_brace(__first, __last);
+                    if (__temp == __first)
+                        __throw_regex_error<regex_constants::error_brace>();
+                    if (__max == -1)
+                        __push_greedy_inf_repeat(__min, __s, __mexp_begin, __mexp_end);
+                    else
+                    {
+                        if (__max < __min)
+                            __throw_regex_error<regex_constants::error_badbrace>();
+                        __push_loop(__min, __max, __s, __mexp_begin, __mexp_end,
+                                    true);
+                    }
+                    __first = __temp;
+                }
+            }
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_ERE_dupl_symbol(_ForwardIterator __first,
+                                                      _ForwardIterator __last,
+                                                      __owns_one_state<_CharT>* __s,
+                                                      unsigned __mexp_begin,
+                                                      unsigned __mexp_end)
+{
+    if (__first != __last)
+    {
+        unsigned __grammar = __get_grammar(__flags_);
+        switch (*__first)
+        {
+        case '*':
+            ++__first;
+            if (__grammar == ECMAScript && __first != __last && *__first == '?')
+            {
+                ++__first;
+                __push_nongreedy_inf_repeat(0, __s, __mexp_begin, __mexp_end);
+            }
+            else
+                __push_greedy_inf_repeat(0, __s, __mexp_begin, __mexp_end);
+            break;
+        case '+':
+            ++__first;
+            if (__grammar == ECMAScript && __first != __last && *__first == '?')
+            {
+                ++__first;
+                __push_nongreedy_inf_repeat(1, __s, __mexp_begin, __mexp_end);
+            }
+            else
+                __push_greedy_inf_repeat(1, __s, __mexp_begin, __mexp_end);
+            break;
+        case '?':
+            ++__first;
+            if (__grammar == ECMAScript && __first != __last && *__first == '?')
+            {
+                ++__first;
+                __push_loop(0, 1, __s, __mexp_begin, __mexp_end, false);
+            }
+            else
+                __push_loop(0, 1, __s, __mexp_begin, __mexp_end);
+            break;
+        case '{':
+            {
+                int __min;
+                _ForwardIterator __temp = __parse_DUP_COUNT(++__first, __last, __min);
+                if (__temp == __first)
+                    __throw_regex_error<regex_constants::error_badbrace>();
+                __first = __temp;
+                if (__first == __last)
+                    __throw_regex_error<regex_constants::error_brace>();
+                switch (*__first)
+                {
+                case '}':
+                    ++__first;
+                    if (__grammar == ECMAScript && __first != __last && *__first == '?')
+                    {
+                        ++__first;
+                        __push_loop(__min, __min, __s, __mexp_begin, __mexp_end, false);
+                    }
+                    else
+                        __push_loop(__min, __min, __s, __mexp_begin, __mexp_end);
+                    break;
+                case ',':
+                    ++__first;
+                    if (__first == __last)
+                        __throw_regex_error<regex_constants::error_badbrace>();
+                    if (*__first == '}')
+                    {
+                        ++__first;
+                        if (__grammar == ECMAScript && __first != __last && *__first == '?')
+                        {
+                            ++__first;
+                            __push_nongreedy_inf_repeat(__min, __s, __mexp_begin, __mexp_end);
+                        }
+                        else
+                            __push_greedy_inf_repeat(__min, __s, __mexp_begin, __mexp_end);
+                    }
+                    else
+                    {
+                        int __max = -1;
+                        __temp = __parse_DUP_COUNT(__first, __last, __max);
+                        if (__temp == __first)
+                            __throw_regex_error<regex_constants::error_brace>();
+                        __first = __temp;
+                        if (__first == __last || *__first != '}')
+                            __throw_regex_error<regex_constants::error_brace>();
+                        ++__first;
+                        if (__max < __min)
+                            __throw_regex_error<regex_constants::error_badbrace>();
+                        if (__grammar == ECMAScript && __first != __last && *__first == '?')
+                        {
+                            ++__first;
+                            __push_loop(__min, __max, __s, __mexp_begin, __mexp_end, false);
+                        }
+                        else
+                            __push_loop(__min, __max, __s, __mexp_begin, __mexp_end);
+                    }
+                    break;
+                default:
+                    __throw_regex_error<regex_constants::error_badbrace>();
+                }
+            }
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_bracket_expression(_ForwardIterator __first,
+                                                         _ForwardIterator __last)
+{
+    if (__first != __last && *__first == '[')
+    {
+        ++__first;
+        if (__first == __last)
+            __throw_regex_error<regex_constants::error_brack>();
+        bool __negate = false;
+        if (*__first == '^')
+        {
+            ++__first;
+            __negate = true;
+        }
+        __bracket_expression<_CharT, _Traits>* __ml = __start_matching_list(__negate);
+        // __ml owned by *this
+        if (__first == __last)
+            __throw_regex_error<regex_constants::error_brack>();
+        if (__get_grammar(__flags_) != ECMAScript && *__first == ']')
+        {
+            __ml->__add_char(']');
+            ++__first;
+        }
+        __first = __parse_follow_list(__first, __last, __ml);
+        if (__first == __last)
+            __throw_regex_error<regex_constants::error_brack>();
+        if (*__first == '-')
+        {
+            __ml->__add_char('-');
+            ++__first;
+        }
+        if (__first == __last || *__first != ']')
+            __throw_regex_error<regex_constants::error_brack>();
+        ++__first;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_follow_list(_ForwardIterator __first,
+                                    _ForwardIterator __last,
+                                    __bracket_expression<_CharT, _Traits>* __ml)
+{
+    if (__first != __last)
+    {
+        while (true)
+        {
+            _ForwardIterator __temp = __parse_expression_term(__first, __last,
+                                                              __ml);
+            if (__temp == __first)
+                break;
+            __first = __temp;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_expression_term(_ForwardIterator __first,
+                                    _ForwardIterator __last,
+                                    __bracket_expression<_CharT, _Traits>* __ml)
+{
+    if (__first != __last && *__first != ']')
+    {
+        _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+        basic_string<_CharT> __start_range;
+        if (__temp != __last && *__first == '[')
+        {
+            if (*__temp == '=')
+                return __parse_equivalence_class(++__temp, __last, __ml);
+            else if (*__temp == ':')
+                return __parse_character_class(++__temp, __last, __ml);
+            else if (*__temp == '.')
+                __first = __parse_collating_symbol(++__temp, __last, __start_range);
+        }
+        unsigned __grammar = __get_grammar(__flags_);
+        if (__start_range.empty())
+        {
+            if ((__grammar == ECMAScript || __grammar == awk) && *__first == '\\')
+            {
+                if (__grammar == ECMAScript)
+                    __first = __parse_class_escape(++__first, __last, __start_range, __ml);
+                else
+                    __first = __parse_awk_escape(++__first, __last, &__start_range);
+            }
+            else
+            {
+                __start_range = *__first;
+                ++__first;
+            }
+        }
+        if (__first != __last && *__first != ']')
+        {
+            __temp = _CUDA_VSTD::next(__first);
+            if (__temp != __last && *__first == '-' && *__temp != ']')
+            {
+                // parse a range
+                basic_string<_CharT> __end_range;
+                __first = __temp;
+                ++__temp;
+                if (__temp != __last && *__first == '[' && *__temp == '.')
+                    __first = __parse_collating_symbol(++__temp, __last, __end_range);
+                else
+                {
+                    if ((__grammar == ECMAScript || __grammar == awk) && *__first == '\\')
+                    {
+                        if (__grammar == ECMAScript)
+                            __first = __parse_class_escape(++__first, __last,
+                                                           __end_range, __ml);
+                        else
+                            __first = __parse_awk_escape(++__first, __last,
+                                                         &__end_range);
+                    }
+                    else
+                    {
+                        __end_range = *__first;
+                        ++__first;
+                    }
+                }
+                __ml->__add_range(_CUDA_VSTD::move(__start_range), _CUDA_VSTD::move(__end_range));
+            }
+            else if (!__start_range.empty())
+            {
+                if (__start_range.size() == 1)
+                    __ml->__add_char(__start_range[0]);
+                else
+                    __ml->__add_digraph(__start_range[0], __start_range[1]);
+            }
+        }
+        else if (!__start_range.empty())
+        {
+            if (__start_range.size() == 1)
+                __ml->__add_char(__start_range[0]);
+            else
+                __ml->__add_digraph(__start_range[0], __start_range[1]);
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_class_escape(_ForwardIterator __first,
+                          _ForwardIterator __last,
+                          basic_string<_CharT>& __str,
+                          __bracket_expression<_CharT, _Traits>* __ml)
+{
+    if (__first == __last)
+        __throw_regex_error<regex_constants::error_escape>();
+    switch (*__first)
+    {
+    case 0:
+        __str = *__first;
+        return ++__first;
+    case 'b':
+        __str = _CharT(8);
+        return ++__first;
+    case 'd':
+        __ml->__add_class(ctype_base::digit);
+        return ++__first;
+    case 'D':
+        __ml->__add_neg_class(ctype_base::digit);
+        return ++__first;
+    case 's':
+        __ml->__add_class(ctype_base::space);
+        return ++__first;
+    case 'S':
+        __ml->__add_neg_class(ctype_base::space);
+        return ++__first;
+    case 'w':
+        __ml->__add_class(ctype_base::alnum);
+        __ml->__add_char('_');
+        return ++__first;
+    case 'W':
+        __ml->__add_neg_class(ctype_base::alnum);
+        __ml->__add_neg_char('_');
+        return ++__first;
+    }
+    __first = __parse_character_escape(__first, __last, &__str);
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_awk_escape(_ForwardIterator __first,
+                          _ForwardIterator __last,
+                          basic_string<_CharT>* __str)
+{
+    if (__first == __last)
+        __throw_regex_error<regex_constants::error_escape>();
+    switch (*__first)
+    {
+    case '\\':
+    case '"':
+    case '/':
+        if (__str)
+            *__str = *__first;
+        else
+            __push_char(*__first);
+        return ++__first;
+    case 'a':
+        if (__str)
+            *__str = _CharT(7);
+        else
+            __push_char(_CharT(7));
+        return ++__first;
+    case 'b':
+        if (__str)
+            *__str = _CharT(8);
+        else
+            __push_char(_CharT(8));
+        return ++__first;
+    case 'f':
+        if (__str)
+            *__str = _CharT(0xC);
+        else
+            __push_char(_CharT(0xC));
+        return ++__first;
+    case 'n':
+        if (__str)
+            *__str = _CharT(0xA);
+        else
+            __push_char(_CharT(0xA));
+        return ++__first;
+    case 'r':
+        if (__str)
+            *__str = _CharT(0xD);
+        else
+            __push_char(_CharT(0xD));
+        return ++__first;
+    case 't':
+        if (__str)
+            *__str = _CharT(0x9);
+        else
+            __push_char(_CharT(0x9));
+        return ++__first;
+    case 'v':
+        if (__str)
+            *__str = _CharT(0xB);
+        else
+            __push_char(_CharT(0xB));
+        return ++__first;
+    }
+    if ('0' <= *__first && *__first <= '7')
+    {
+        unsigned __val = *__first - '0';
+        if (++__first != __last && ('0' <= *__first && *__first <= '7'))
+        {
+            __val = 8 * __val + *__first - '0';
+            if (++__first != __last && ('0' <= *__first && *__first <= '7'))
+                __val = 8 * __val + *__first++ - '0';
+        }
+        if (__str)
+            *__str = _CharT(__val);
+        else
+            __push_char(_CharT(__val));
+    }
+    else
+        __throw_regex_error<regex_constants::error_escape>();
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_equivalence_class(_ForwardIterator __first,
+                                    _ForwardIterator __last,
+                                    __bracket_expression<_CharT, _Traits>* __ml)
+{
+    // Found [=
+    //   This means =] must exist
+    value_type _Equal_close[2] = {'=', ']'};
+    _ForwardIterator __temp = _CUDA_VSTD::search(__first, __last, _Equal_close,
+                                                            _Equal_close+2);
+    if (__temp == __last)
+        __throw_regex_error<regex_constants::error_brack>();
+    // [__first, __temp) contains all text in [= ... =]
+    string_type __collate_name =
+        __traits_.lookup_collatename(__first, __temp);
+    if (__collate_name.empty())
+        __throw_regex_error<regex_constants::error_collate>();
+    string_type __equiv_name =
+        __traits_.transform_primary(__collate_name.begin(),
+                                    __collate_name.end());
+    if (!__equiv_name.empty())
+        __ml->__add_equivalence(__equiv_name);
+    else
+    {
+        switch (__collate_name.size())
+        {
+        case 1:
+            __ml->__add_char(__collate_name[0]);
+            break;
+        case 2:
+            __ml->__add_digraph(__collate_name[0], __collate_name[1]);
+            break;
+        default:
+            __throw_regex_error<regex_constants::error_collate>();
+        }
+    }
+    __first = _CUDA_VSTD::next(__temp, 2);
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_character_class(_ForwardIterator __first,
+                                    _ForwardIterator __last,
+                                    __bracket_expression<_CharT, _Traits>* __ml)
+{
+    // Found [:
+    //   This means :] must exist
+    value_type _Colon_close[2] = {':', ']'};
+    _ForwardIterator __temp = _CUDA_VSTD::search(__first, __last, _Colon_close,
+                                                            _Colon_close+2);
+    if (__temp == __last)
+        __throw_regex_error<regex_constants::error_brack>();
+    // [__first, __temp) contains all text in [: ... :]
+    typedef typename _Traits::char_class_type char_class_type;
+    char_class_type __class_type =
+        __traits_.lookup_classname(__first, __temp, __flags_ & icase);
+    if (__class_type == 0)
+        __throw_regex_error<regex_constants::error_ctype>();
+    __ml->__add_class(__class_type);
+    __first = _CUDA_VSTD::next(__temp, 2);
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_collating_symbol(_ForwardIterator __first,
+                                                _ForwardIterator __last,
+                                                basic_string<_CharT>& __col_sym)
+{
+    // Found [.
+    //   This means .] must exist
+    value_type _Dot_close[2] = {'.', ']'};
+    _ForwardIterator __temp = _CUDA_VSTD::search(__first, __last, _Dot_close,
+                                                            _Dot_close+2);
+    if (__temp == __last)
+        __throw_regex_error<regex_constants::error_brack>();
+    // [__first, __temp) contains all text in [. ... .]
+    __col_sym = __traits_.lookup_collatename(__first, __temp);
+    switch (__col_sym.size())
+    {
+    case 1:
+    case 2:
+        break;
+    default:
+        __throw_regex_error<regex_constants::error_collate>();
+    }
+    __first = _CUDA_VSTD::next(__temp, 2);
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_DUP_COUNT(_ForwardIterator __first,
+                                                _ForwardIterator __last,
+                                                int& __c)
+{
+    if (__first != __last )
+    {
+        int __val = __traits_.value(*__first, 10);
+        if ( __val != -1 )
+        {
+            __c = __val;
+            for (++__first;
+                 __first != __last && ( __val = __traits_.value(*__first, 10)) != -1;
+                 ++__first)
+            {
+                if (__c >= std::numeric_limits<int>::max() / 10)
+                    __throw_regex_error<regex_constants::error_badbrace>();
+                __c *= 10;
+                __c += __val;
+            }
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_ecma_exp(_ForwardIterator __first,
+                                               _ForwardIterator __last)
+{
+    __owns_one_state<_CharT>* __sa = __end_;
+    _ForwardIterator __temp = __parse_alternative(__first, __last);
+    if (__temp == __first)
+        __push_empty();
+    __first = __temp;
+    while (__first != __last && *__first == '|')
+    {
+        __owns_one_state<_CharT>* __sb = __end_;
+        __temp = __parse_alternative(++__first, __last);
+        if (__temp == __first)
+            __push_empty();
+        __push_alternation(__sa, __sb);
+        __first = __temp;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_alternative(_ForwardIterator __first,
+                                                  _ForwardIterator __last)
+{
+    while (true)
+    {
+        _ForwardIterator __temp = __parse_term(__first, __last);
+        if (__temp == __first)
+            break;
+        __first = __temp;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_term(_ForwardIterator __first,
+                                           _ForwardIterator __last)
+{
+    _ForwardIterator __temp = __parse_assertion(__first, __last);
+    if (__temp == __first)
+    {
+        __owns_one_state<_CharT>* __e = __end_;
+        unsigned __mexp_begin = __marked_count_;
+        __temp = __parse_atom(__first, __last);
+        if (__temp != __first)
+            __first = __parse_ERE_dupl_symbol(__temp, __last, __e,
+                                              __mexp_begin+1, __marked_count_+1);
+    }
+    else
+        __first = __temp;
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_assertion(_ForwardIterator __first,
+                                                _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        switch (*__first)
+        {
+        case '^':
+            __push_l_anchor();
+            ++__first;
+            break;
+        case '$':
+            __push_r_anchor();
+            ++__first;
+            break;
+        case '\\':
+            {
+                _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+                if (__temp != __last)
+                {
+                    if (*__temp == 'b')
+                    {
+                        __push_word_boundary(false);
+                        __first = ++__temp;
+                    }
+                    else if (*__temp == 'B')
+                    {
+                        __push_word_boundary(true);
+                        __first = ++__temp;
+                    }
+                }
+            }
+            break;
+        case '(':
+            {
+                _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+                if (__temp != __last && *__temp == '?')
+                {
+                    if (++__temp != __last)
+                    {
+                        switch (*__temp)
+                        {
+                        case '=':
+                            {
+                                basic_regex __exp;
+                                __exp.__flags_ = __flags_;
+                                __temp = __exp.__parse(++__temp, __last);
+                                unsigned __mexp = __exp.__marked_count_;
+                                __push_lookahead(_CUDA_VSTD::move(__exp), false, __marked_count_);
+                                __marked_count_ += __mexp;
+                                if (__temp == __last || *__temp != ')')
+                                    __throw_regex_error<regex_constants::error_paren>();
+                                __first = ++__temp;
+                            }
+                            break;
+                        case '!':
+                            {
+                                basic_regex __exp;
+                                __exp.__flags_ = __flags_;
+                                __temp = __exp.__parse(++__temp, __last);
+                                unsigned __mexp = __exp.__marked_count_;
+                                __push_lookahead(_CUDA_VSTD::move(__exp), true, __marked_count_);
+                                __marked_count_ += __mexp;
+                                if (__temp == __last || *__temp != ')')
+                                    __throw_regex_error<regex_constants::error_paren>();
+                                __first = ++__temp;
+                            }
+                            break;
+                        }
+                    }
+                }
+            }
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_atom(_ForwardIterator __first,
+                                           _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        switch (*__first)
+        {
+        case '.':
+            __push_match_any_but_newline();
+            ++__first;
+            break;
+        case '\\':
+            __first = __parse_atom_escape(__first, __last);
+            break;
+        case '[':
+            __first = __parse_bracket_expression(__first, __last);
+            break;
+        case '(':
+            {
+                ++__first;
+                if (__first == __last)
+                    __throw_regex_error<regex_constants::error_paren>();
+                _ForwardIterator __temp = _CUDA_VSTD::next(__first);
+                if (__temp != __last && *__first == '?' && *__temp == ':')
+                {
+                    ++__open_count_;
+                    __first = __parse_ecma_exp(++__temp, __last);
+                    if (__first == __last || *__first != ')')
+                        __throw_regex_error<regex_constants::error_paren>();
+                    --__open_count_;
+                    ++__first;
+                }
+                else
+                {
+                    __push_begin_marked_subexpression();
+                    unsigned __temp_count = __marked_count_;
+                    ++__open_count_;
+                    __first = __parse_ecma_exp(__first, __last);
+                    if (__first == __last || *__first != ')')
+                        __throw_regex_error<regex_constants::error_paren>();
+                    __push_end_marked_subexpression(__temp_count);
+                    --__open_count_;
+                    ++__first;
+                }
+            }
+            break;
+        case '*':
+        case '+':
+        case '?':
+        case '{':
+            __throw_regex_error<regex_constants::error_badrepeat>();
+            break;
+        default:
+            __first = __parse_pattern_character(__first, __last);
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_atom_escape(_ForwardIterator __first,
+                                                  _ForwardIterator __last)
+{
+    if (__first != __last && *__first == '\\')
+    {
+        _ForwardIterator __t1 = _CUDA_VSTD::next(__first);
+        if (__t1 == __last)
+            __throw_regex_error<regex_constants::error_escape>();
+
+        _ForwardIterator __t2 = __parse_decimal_escape(__t1, __last);
+        if (__t2 != __t1)
+            __first = __t2;
+        else
+        {
+            __t2 = __parse_character_class_escape(__t1, __last);
+            if (__t2 != __t1)
+                __first = __t2;
+            else
+            {
+                __t2 = __parse_character_escape(__t1, __last);
+                if (__t2 != __t1)
+                    __first = __t2;
+            }
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_decimal_escape(_ForwardIterator __first,
+                                                     _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        if (*__first == '0')
+        {
+            __push_char(_CharT());
+            ++__first;
+        }
+        else if ('1' <= *__first && *__first <= '9')
+        {
+            unsigned __v = *__first - '0';
+            for (++__first;
+                    __first != __last && '0' <= *__first && *__first <= '9'; ++__first)
+                {
+                if (__v >= std::numeric_limits<unsigned>::max() / 10)
+                    __throw_regex_error<regex_constants::error_backref>();
+                __v = 10 * __v + *__first - '0';
+                }
+            if (__v == 0 || __v > mark_count())
+                __throw_regex_error<regex_constants::error_backref>();
+            __push_back_ref(__v);
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_character_class_escape(_ForwardIterator __first,
+                                                             _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        __bracket_expression<_CharT, _Traits>* __ml;
+        switch (*__first)
+        {
+        case 'd':
+            __ml = __start_matching_list(false);
+            __ml->__add_class(ctype_base::digit);
+            ++__first;
+            break;
+        case 'D':
+            __ml = __start_matching_list(true);
+            __ml->__add_class(ctype_base::digit);
+            ++__first;
+            break;
+        case 's':
+            __ml = __start_matching_list(false);
+            __ml->__add_class(ctype_base::space);
+            ++__first;
+            break;
+        case 'S':
+            __ml = __start_matching_list(true);
+            __ml->__add_class(ctype_base::space);
+            ++__first;
+            break;
+        case 'w':
+            __ml = __start_matching_list(false);
+            __ml->__add_class(ctype_base::alnum);
+            __ml->__add_char('_');
+            ++__first;
+            break;
+        case 'W':
+            __ml = __start_matching_list(true);
+            __ml->__add_class(ctype_base::alnum);
+            __ml->__add_char('_');
+            ++__first;
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_character_escape(_ForwardIterator __first,
+                                                    _ForwardIterator __last,
+                                                    basic_string<_CharT>* __str)
+{
+    if (__first != __last)
+    {
+        _ForwardIterator __t;
+        unsigned __sum = 0;
+        int __hd;
+        switch (*__first)
+        {
+        case 'f':
+            if (__str)
+                *__str = _CharT(0xC);
+            else
+                __push_char(_CharT(0xC));
+            ++__first;
+            break;
+        case 'n':
+            if (__str)
+                *__str = _CharT(0xA);
+            else
+                __push_char(_CharT(0xA));
+            ++__first;
+            break;
+        case 'r':
+            if (__str)
+                *__str = _CharT(0xD);
+            else
+                __push_char(_CharT(0xD));
+            ++__first;
+            break;
+        case 't':
+            if (__str)
+                *__str = _CharT(0x9);
+            else
+                __push_char(_CharT(0x9));
+            ++__first;
+            break;
+        case 'v':
+            if (__str)
+                *__str = _CharT(0xB);
+            else
+                __push_char(_CharT(0xB));
+            ++__first;
+            break;
+        case 'c':
+            if ((__t = _CUDA_VSTD::next(__first)) != __last)
+            {
+                if (('A' <= *__t && *__t <= 'Z') ||
+                    ('a' <= *__t && *__t <= 'z'))
+                {
+                    if (__str)
+                        *__str = _CharT(*__t % 32);
+                    else
+                        __push_char(_CharT(*__t % 32));
+                    __first = ++__t;
+                }
+                else
+                    __throw_regex_error<regex_constants::error_escape>();
+            }
+            else
+                __throw_regex_error<regex_constants::error_escape>();
+            break;
+        case 'u':
+            ++__first;
+            if (__first == __last)
+                __throw_regex_error<regex_constants::error_escape>();
+            __hd = __traits_.value(*__first, 16);
+            if (__hd == -1)
+                __throw_regex_error<regex_constants::error_escape>();
+            __sum = 16 * __sum + static_cast<unsigned>(__hd);
+            ++__first;
+            if (__first == __last)
+                __throw_regex_error<regex_constants::error_escape>();
+            __hd = __traits_.value(*__first, 16);
+            if (__hd == -1)
+                __throw_regex_error<regex_constants::error_escape>();
+            __sum = 16 * __sum + static_cast<unsigned>(__hd);
+            // drop through
+        case 'x':
+            ++__first;
+            if (__first == __last)
+                __throw_regex_error<regex_constants::error_escape>();
+            __hd = __traits_.value(*__first, 16);
+            if (__hd == -1)
+                __throw_regex_error<regex_constants::error_escape>();
+            __sum = 16 * __sum + static_cast<unsigned>(__hd);
+            ++__first;
+            if (__first == __last)
+                __throw_regex_error<regex_constants::error_escape>();
+            __hd = __traits_.value(*__first, 16);
+            if (__hd == -1)
+                __throw_regex_error<regex_constants::error_escape>();
+            __sum = 16 * __sum + static_cast<unsigned>(__hd);
+            if (__str)
+                *__str = _CharT(__sum);
+            else
+                __push_char(_CharT(__sum));
+            ++__first;
+            break;
+        case '0':
+            if (__str)
+                *__str = _CharT(0);
+            else
+                __push_char(_CharT(0));
+            ++__first;
+            break;
+        default:
+            if (*__first != '_' && !__traits_.isctype(*__first, ctype_base::alnum))
+            {
+                if (__str)
+                    *__str = *__first;
+                else
+                    __push_char(*__first);
+                ++__first;
+            }
+            else
+                __throw_regex_error<regex_constants::error_escape>();
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_pattern_character(_ForwardIterator __first,
+                                                        _ForwardIterator __last)
+{
+    if (__first != __last)
+    {
+        switch (*__first)
+        {
+        case '^':
+        case '$':
+        case '\\':
+        case '.':
+        case '*':
+        case '+':
+        case '?':
+        case '(':
+        case ')':
+        case '[':
+        case ']':
+        case '{':
+        case '}':
+        case '|':
+            break;
+        default:
+            __push_char(*__first);
+            ++__first;
+            break;
+        }
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_grep(_ForwardIterator __first,
+                                           _ForwardIterator __last)
+{
+    __owns_one_state<_CharT>* __sa = __end_;
+    _ForwardIterator __t1 = _CUDA_VSTD::find(__first, __last, _CharT('\n'));
+    if (__t1 != __first)
+        __parse_basic_reg_exp(__first, __t1);
+    else
+        __push_empty();
+    __first = __t1;
+    if (__first != __last)
+        ++__first;
+    while (__first != __last)
+    {
+        __t1 = _CUDA_VSTD::find(__first, __last, _CharT('\n'));
+        __owns_one_state<_CharT>* __sb = __end_;
+        if (__t1 != __first)
+            __parse_basic_reg_exp(__first, __t1);
+        else
+            __push_empty();
+        __push_alternation(__sa, __sb);
+        __first = __t1;
+        if (__first != __last)
+            ++__first;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+template <class _ForwardIterator>
+_ForwardIterator
+basic_regex<_CharT, _Traits>::__parse_egrep(_ForwardIterator __first,
+                                            _ForwardIterator __last)
+{
+    __owns_one_state<_CharT>* __sa = __end_;
+    _ForwardIterator __t1 = _CUDA_VSTD::find(__first, __last, _CharT('\n'));
+    if (__t1 != __first)
+        __parse_extended_reg_exp(__first, __t1);
+    else
+        __push_empty();
+    __first = __t1;
+    if (__first != __last)
+        ++__first;
+    while (__first != __last)
+    {
+        __t1 = _CUDA_VSTD::find(__first, __last, _CharT('\n'));
+        __owns_one_state<_CharT>* __sb = __end_;
+        if (__t1 != __first)
+            __parse_extended_reg_exp(__first, __t1);
+        else
+            __push_empty();
+        __push_alternation(__sa, __sb);
+        __first = __t1;
+        if (__first != __last)
+            ++__first;
+    }
+    return __first;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_loop(size_t __min, size_t __max,
+        __owns_one_state<_CharT>* __s, size_t __mexp_begin, size_t __mexp_end,
+        bool __greedy)
+{
+    unique_ptr<__empty_state<_CharT> > __e1(new __empty_state<_CharT>(__end_->first()));
+    __end_->first() = nullptr;
+    unique_ptr<__loop<_CharT> > __e2(new __loop<_CharT>(__loop_count_,
+                __s->first(), __e1.get(), __mexp_begin, __mexp_end, __greedy,
+                __min, __max));
+    __s->first() = nullptr;
+    __e1.release();
+    __end_->first() = new __repeat_one_loop<_CharT>(__e2.get());
+    __end_ = __e2->second();
+    __s->first() = __e2.release();
+    ++__loop_count_;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_char(value_type __c)
+{
+    if (flags() & icase)
+        __end_->first() = new __match_char_icase<_CharT, _Traits>
+                                              (__traits_, __c, __end_->first());
+    else if (flags() & collate)
+        __end_->first() = new __match_char_collate<_CharT, _Traits>
+                                              (__traits_, __c, __end_->first());
+    else
+        __end_->first() = new __match_char<_CharT>(__c, __end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_begin_marked_subexpression()
+{
+    if (!(__flags_ & nosubs))
+    {
+        __end_->first() =
+                new __begin_marked_subexpression<_CharT>(++__marked_count_,
+                                                         __end_->first());
+        __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+    }
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_end_marked_subexpression(unsigned __sub)
+{
+    if (!(__flags_ & nosubs))
+    {
+        __end_->first() =
+                new __end_marked_subexpression<_CharT>(__sub, __end_->first());
+        __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+    }
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_l_anchor()
+{
+    __end_->first() = new __l_anchor<_CharT>(__end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_r_anchor()
+{
+    __end_->first() = new __r_anchor<_CharT>(__end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_match_any()
+{
+    __end_->first() = new __match_any<_CharT>(__end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_match_any_but_newline()
+{
+    __end_->first() = new __match_any_but_newline<_CharT>(__end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_empty()
+{
+    __end_->first() = new __empty_state<_CharT>(__end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_word_boundary(bool __invert)
+{
+    __end_->first() = new __word_boundary<_CharT, _Traits>(__traits_, __invert,
+                                                           __end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_back_ref(int __i)
+{
+    if (flags() & icase)
+        __end_->first() = new __back_ref_icase<_CharT, _Traits>
+                                              (__traits_, __i, __end_->first());
+    else if (flags() & collate)
+        __end_->first() = new __back_ref_collate<_CharT, _Traits>
+                                              (__traits_, __i, __end_->first());
+    else
+        __end_->first() = new __back_ref<_CharT>(__i, __end_->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_alternation(__owns_one_state<_CharT>* __sa,
+                                                 __owns_one_state<_CharT>* __ea)
+{
+    __sa->first() = new __alternate<_CharT>(
+                         static_cast<__owns_one_state<_CharT>*>(__sa->first()),
+                         static_cast<__owns_one_state<_CharT>*>(__ea->first()));
+    __ea->first() = nullptr;
+    __ea->first() = new __empty_state<_CharT>(__end_->first());
+    __end_->first() = nullptr;
+    __end_->first() = new __empty_non_own_state<_CharT>(__ea->first());
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__ea->first());
+}
+
+template <class _CharT, class _Traits>
+__bracket_expression<_CharT, _Traits>*
+basic_regex<_CharT, _Traits>::__start_matching_list(bool __negate)
+{
+    __bracket_expression<_CharT, _Traits>* __r =
+        new __bracket_expression<_CharT, _Traits>(__traits_, __end_->first(),
+                                                  __negate, __flags_ & icase,
+                                                  __flags_ & collate);
+    __end_->first() = __r;
+    __end_ = __r;
+    return __r;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_regex<_CharT, _Traits>::__push_lookahead(const basic_regex& __exp,
+                                               bool __invert,
+                                               unsigned __mexp)
+{
+    __end_->first() = new __lookahead<_CharT, _Traits>(__exp, __invert,
+                                                           __end_->first(), __mexp);
+    __end_ = static_cast<__owns_one_state<_CharT>*>(__end_->first());
+}
+
+typedef basic_regex<char>    regex;
+typedef basic_regex<wchar_t> wregex;
+
+// sub_match
+
+template <class _BidirectionalIterator>
+class _LIBCUDACXX_TEMPLATE_VIS sub_match
+    : public pair<_BidirectionalIterator, _BidirectionalIterator>
+{
+public:
+    typedef _BidirectionalIterator                              iterator;
+    typedef typename iterator_traits<iterator>::value_type      value_type;
+    typedef typename iterator_traits<iterator>::difference_type difference_type;
+    typedef basic_string<value_type>                            string_type;
+
+    bool matched;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR sub_match() : matched() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    difference_type length() const
+        {return matched ? _CUDA_VSTD::distance(this->first, this->second) : 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    string_type str() const
+        {return matched ? string_type(this->first, this->second) : string_type();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator string_type() const
+        {return str();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(const sub_match& __s) const
+        {return str().compare(__s.str());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(const string_type& __s) const
+        {return str().compare(__s);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(const value_type* __s) const
+        {return str().compare(__s);}
+};
+
+typedef sub_match<const char*>             csub_match;
+typedef sub_match<const wchar_t*>          wcsub_match;
+typedef sub_match<string::const_iterator>  ssub_match;
+typedef sub_match<wstring::const_iterator> wssub_match;
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const sub_match<_BiIter>& __x, const sub_match<_BiIter>& __y)
+{
+    return __x.compare(__y) == 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const sub_match<_BiIter>& __x, const sub_match<_BiIter>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const sub_match<_BiIter>& __x, const sub_match<_BiIter>& __y)
+{
+    return __x.compare(__y) < 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const sub_match<_BiIter>& __x, const sub_match<_BiIter>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const sub_match<_BiIter>& __x, const sub_match<_BiIter>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const sub_match<_BiIter>& __x, const sub_match<_BiIter>& __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __x,
+           const sub_match<_BiIter>& __y)
+{
+    return __y.compare(typename sub_match<_BiIter>::string_type(__x.data(), __x.size())) == 0;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __x,
+          const sub_match<_BiIter>& __y)
+{
+    return __y.compare(typename sub_match<_BiIter>::string_type(__x.data(), __x.size())) > 0;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __x,
+          const sub_match<_BiIter>& __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>=(const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __x,
+                const sub_match<_BiIter>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const sub_match<_BiIter>& __x,
+           const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __y)
+{
+    return __x.compare(typename sub_match<_BiIter>::string_type(__y.data(), __y.size())) == 0;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const sub_match<_BiIter>& __x,
+           const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const sub_match<_BiIter>& __x,
+          const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __y)
+{
+    return __x.compare(typename sub_match<_BiIter>::string_type(__y.data(), __y.size())) < 0;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>(const sub_match<_BiIter>& __x,
+               const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const sub_match<_BiIter>& __x,
+           const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const sub_match<_BiIter>& __x,
+           const basic_string<typename iterator_traits<_BiIter>::value_type, _ST, _SA>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(typename iterator_traits<_BiIter>::value_type const* __x,
+           const sub_match<_BiIter>& __y)
+{
+    return __y.compare(__x) == 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(typename iterator_traits<_BiIter>::value_type const* __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(typename iterator_traits<_BiIter>::value_type const* __x,
+          const sub_match<_BiIter>& __y)
+{
+    return __y.compare(__x) > 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(typename iterator_traits<_BiIter>::value_type const* __x,
+          const sub_match<_BiIter>& __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(typename iterator_traits<_BiIter>::value_type const* __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(typename iterator_traits<_BiIter>::value_type const* __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const* __y)
+{
+    return __x.compare(__y) == 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const* __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const sub_match<_BiIter>& __x,
+          typename iterator_traits<_BiIter>::value_type const* __y)
+{
+    return __x.compare(__y) < 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const sub_match<_BiIter>& __x,
+          typename iterator_traits<_BiIter>::value_type const* __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const* __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const* __y)
+{
+    return !(__y < __x);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(typename iterator_traits<_BiIter>::value_type const& __x,
+           const sub_match<_BiIter>& __y)
+{
+    typedef basic_string<typename iterator_traits<_BiIter>::value_type> string_type;
+    return __y.compare(string_type(1, __x)) == 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(typename iterator_traits<_BiIter>::value_type const& __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(typename iterator_traits<_BiIter>::value_type const& __x,
+          const sub_match<_BiIter>& __y)
+{
+    typedef basic_string<typename iterator_traits<_BiIter>::value_type> string_type;
+    return __y.compare(string_type(1, __x)) > 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(typename iterator_traits<_BiIter>::value_type const& __x,
+          const sub_match<_BiIter>& __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(typename iterator_traits<_BiIter>::value_type const& __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(typename iterator_traits<_BiIter>::value_type const& __x,
+           const sub_match<_BiIter>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const& __y)
+{
+    typedef basic_string<typename iterator_traits<_BiIter>::value_type> string_type;
+    return __x.compare(string_type(1, __y)) == 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const sub_match<_BiIter>& __x,
+          typename iterator_traits<_BiIter>::value_type const& __y)
+{
+    typedef basic_string<typename iterator_traits<_BiIter>::value_type> string_type;
+    return __x.compare(string_type(1, __y)) < 0;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>(const sub_match<_BiIter>& __x,
+          typename iterator_traits<_BiIter>::value_type const& __y)
+{
+    return __y < __x;
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const sub_match<_BiIter>& __x,
+           typename iterator_traits<_BiIter>::value_type const& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _CharT, class _ST, class _BiIter>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _ST>&
+operator<<(basic_ostream<_CharT, _ST>& __os, const sub_match<_BiIter>& __m)
+{
+    return __os << __m.str();
+}
+
+template <class _BidirectionalIterator, class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS match_results
+{
+public:
+    typedef _Allocator                                        allocator_type;
+    typedef sub_match<_BidirectionalIterator>                 value_type;
+private:
+    typedef vector<value_type, allocator_type>                __container_type;
+
+    __container_type  __matches_;
+    value_type __unmatched_;
+    value_type __prefix_;
+    value_type __suffix_;
+    bool       __ready_;
+public:
+    _BidirectionalIterator __position_start_;
+    typedef const value_type&                                 const_reference;
+    typedef value_type&                                       reference;
+    typedef typename __container_type::const_iterator         const_iterator;
+    typedef const_iterator                                    iterator;
+    typedef typename iterator_traits<_BidirectionalIterator>::difference_type difference_type;
+    typedef typename allocator_traits<allocator_type>::size_type size_type;
+    typedef typename iterator_traits<_BidirectionalIterator>::value_type char_type;
+    typedef basic_string<char_type>                           string_type;
+
+    // construct/copy/destroy:
+    explicit match_results(const allocator_type& __a = allocator_type());
+//    match_results(const match_results&) = default;
+//    match_results& operator=(const match_results&) = default;
+//    match_results(match_results&& __m) = default;
+//    match_results& operator=(match_results&& __m) = default;
+//    ~match_results() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool ready() const {return __ready_;}
+
+    // size:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT {return __matches_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __matches_.max_size();}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return size() == 0;}
+
+    // element access:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    difference_type length(size_type __sub = 0) const
+        {
+        _LIBCUDACXX_ASSERT(ready(), "match_results::length() called when not ready");
+        return (*this)[__sub].length();
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    difference_type position(size_type __sub = 0) const
+        {
+        _LIBCUDACXX_ASSERT(ready(), "match_results::position() called when not ready");
+        return _CUDA_VSTD::distance(__position_start_, (*this)[__sub].first);
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    string_type str(size_type __sub = 0) const
+        {
+        _LIBCUDACXX_ASSERT(ready(), "match_results::str() called when not ready");
+        return (*this)[__sub].str();
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference operator[](size_type __n) const
+        {
+        _LIBCUDACXX_ASSERT(ready(), "match_results::operator[]() called when not ready");
+        return __n < __matches_.size() ? __matches_[__n] : __unmatched_;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference prefix() const
+        {
+        _LIBCUDACXX_ASSERT(ready(), "match_results::prefix() called when not ready");
+        return __prefix_;
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference suffix() const
+        {
+        _LIBCUDACXX_ASSERT(ready(), "match_results::suffix() called when not ready");
+        return __suffix_;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const {return empty() ? __matches_.end() : __matches_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const {return __matches_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const {return empty() ? __matches_.end() : __matches_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const {return __matches_.end();}
+
+    // format:
+    template <class _OutputIter>
+        _OutputIter
+        format(_OutputIter __output_iter, const char_type* __fmt_first,
+               const char_type* __fmt_last,
+               regex_constants::match_flag_type __flags = regex_constants::format_default) const;
+    template <class _OutputIter, class _ST, class _SA>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        _OutputIter
+        format(_OutputIter __output_iter, const basic_string<char_type, _ST, _SA>& __fmt,
+               regex_constants::match_flag_type __flags = regex_constants::format_default) const
+            {return format(__output_iter, __fmt.data(), __fmt.data() + __fmt.size(), __flags);}
+    template <class _ST, class _SA>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string<char_type, _ST, _SA>
+        format(const basic_string<char_type, _ST, _SA>& __fmt,
+               regex_constants::match_flag_type __flags = regex_constants::format_default) const
+        {
+            basic_string<char_type, _ST, _SA> __r;
+            format(back_inserter(__r), __fmt.data(), __fmt.data() + __fmt.size(),
+                   __flags);
+            return __r;
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    string_type
+        format(const char_type* __fmt,
+               regex_constants::match_flag_type __flags = regex_constants::format_default) const
+        {
+            string_type __r;
+            format(back_inserter(__r), __fmt,
+                   __fmt + char_traits<char_type>::length(__fmt), __flags);
+            return __r;
+        }
+
+    // allocator:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const {return __matches_.get_allocator();}
+
+    // swap:
+    void swap(match_results& __m);
+
+    template <class _Bp, class _Ap>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void __assign(_BidirectionalIterator __f, _BidirectionalIterator __l,
+                      const match_results<_Bp, _Ap>& __m, bool __no_update_pos)
+    {
+        _Bp __mf = __m.prefix().first;
+        __matches_.resize(__m.size());
+        for (size_type __i = 0; __i < __matches_.size(); ++__i)
+        {
+            __matches_[__i].first = _CUDA_VSTD::next(__f, _CUDA_VSTD::distance(__mf, __m[__i].first));
+            __matches_[__i].second = _CUDA_VSTD::next(__f, _CUDA_VSTD::distance(__mf, __m[__i].second));
+            __matches_[__i].matched = __m[__i].matched;
+        }
+        __unmatched_.first   = __l;
+        __unmatched_.second  = __l;
+        __unmatched_.matched = false;
+        __prefix_.first = _CUDA_VSTD::next(__f, _CUDA_VSTD::distance(__mf, __m.prefix().first));
+        __prefix_.second = _CUDA_VSTD::next(__f, _CUDA_VSTD::distance(__mf, __m.prefix().second));
+        __prefix_.matched = __m.prefix().matched;
+        __suffix_.first = _CUDA_VSTD::next(__f, _CUDA_VSTD::distance(__mf, __m.suffix().first));
+        __suffix_.second = _CUDA_VSTD::next(__f, _CUDA_VSTD::distance(__mf, __m.suffix().second));
+        __suffix_.matched = __m.suffix().matched;
+        if (!__no_update_pos)
+            __position_start_ = __prefix_.first;
+        __ready_ = __m.ready();
+    }
+
+private:
+    void __init(unsigned __s,
+                _BidirectionalIterator __f, _BidirectionalIterator __l,
+                bool __no_update_pos = false);
+
+    template <class, class> friend class basic_regex;
+
+    template <class _Bp, class _Ap, class _Cp, class _Tp>
+    friend
+    bool
+    regex_match(_Bp, _Bp, match_results<_Bp, _Ap>&, const basic_regex<_Cp, _Tp>&,
+                regex_constants::match_flag_type);
+
+    template <class _Bp, class _Ap>
+    friend
+    bool
+    operator==(const match_results<_Bp, _Ap>&, const match_results<_Bp, _Ap>&);
+
+    template <class, class> friend class __lookahead;
+};
+
+template <class _BidirectionalIterator, class _Allocator>
+match_results<_BidirectionalIterator, _Allocator>::match_results(
+        const allocator_type& __a)
+    : __matches_(__a),
+      __unmatched_(),
+      __prefix_(),
+      __suffix_(),
+      __ready_(false),
+      __position_start_()
+{
+}
+
+template <class _BidirectionalIterator, class _Allocator>
+void
+match_results<_BidirectionalIterator, _Allocator>::__init(unsigned __s,
+                         _BidirectionalIterator __f, _BidirectionalIterator __l,
+                         bool __no_update_pos)
+{
+    __unmatched_.first   = __l;
+    __unmatched_.second  = __l;
+    __unmatched_.matched = false;
+    __matches_.assign(__s, __unmatched_);
+    __prefix_.first      = __f;
+    __prefix_.second     = __f;
+    __prefix_.matched    = false;
+    __suffix_ = __unmatched_;
+    if (!__no_update_pos)
+        __position_start_ = __prefix_.first;
+    __ready_ = true;
+}
+
+template <class _BidirectionalIterator, class _Allocator>
+template <class _OutputIter>
+_OutputIter
+match_results<_BidirectionalIterator, _Allocator>::format(_OutputIter __output_iter,
+        const char_type* __fmt_first, const char_type* __fmt_last,
+        regex_constants::match_flag_type __flags) const
+{
+    _LIBCUDACXX_ASSERT(ready(), "match_results::format() called when not ready");
+    if (__flags & regex_constants::format_sed)
+    {
+        for (; __fmt_first != __fmt_last; ++__fmt_first)
+        {
+            if (*__fmt_first == '&')
+                __output_iter = _CUDA_VSTD::copy(__matches_[0].first, __matches_[0].second,
+                                   __output_iter);
+            else if (*__fmt_first == '\\' && __fmt_first + 1 != __fmt_last)
+            {
+                ++__fmt_first;
+                if ('0' <= *__fmt_first && *__fmt_first <= '9')
+                {
+                    size_t __i = *__fmt_first - '0';
+                    __output_iter = _CUDA_VSTD::copy((*this)[__i].first,
+                                        (*this)[__i].second, __output_iter);
+                }
+                else
+                {
+                    *__output_iter = *__fmt_first;
+                    ++__output_iter;
+                }
+            }
+            else
+            {
+                *__output_iter = *__fmt_first;
+                ++__output_iter;
+            }
+        }
+    }
+    else
+    {
+        for (; __fmt_first != __fmt_last; ++__fmt_first)
+        {
+            if (*__fmt_first == '$' && __fmt_first + 1 != __fmt_last)
+            {
+                switch (__fmt_first[1])
+                {
+                case '$':
+                    *__output_iter = *++__fmt_first;
+                    ++__output_iter;
+                    break;
+                case '&':
+                    ++__fmt_first;
+                    __output_iter = _CUDA_VSTD::copy(__matches_[0].first, __matches_[0].second,
+                                       __output_iter);
+                    break;
+                case '`':
+                    ++__fmt_first;
+                    __output_iter = _CUDA_VSTD::copy(__prefix_.first, __prefix_.second, __output_iter);
+                    break;
+                case '\'':
+                    ++__fmt_first;
+                    __output_iter = _CUDA_VSTD::copy(__suffix_.first, __suffix_.second, __output_iter);
+                    break;
+                default:
+                    if ('0' <= __fmt_first[1] && __fmt_first[1] <= '9')
+                    {
+                        ++__fmt_first;
+                        size_t __idx = *__fmt_first - '0';
+                        if (__fmt_first + 1 != __fmt_last &&
+                            '0' <= __fmt_first[1] && __fmt_first[1] <= '9')
+                        {
+                            ++__fmt_first;
+                            if (__idx >= std::numeric_limits<size_t>::max() / 10)
+                                __throw_regex_error<regex_constants::error_escape>();
+                            __idx = 10 * __idx + *__fmt_first - '0';
+                        }
+                        __output_iter = _CUDA_VSTD::copy((*this)[__idx].first,
+                                            (*this)[__idx].second, __output_iter);
+                    }
+                    else
+                    {
+                        *__output_iter = *__fmt_first;
+                        ++__output_iter;
+                    }
+                    break;
+                }
+            }
+            else
+            {
+                *__output_iter = *__fmt_first;
+                ++__output_iter;
+            }
+        }
+    }
+    return __output_iter;
+}
+
+template <class _BidirectionalIterator, class _Allocator>
+void
+match_results<_BidirectionalIterator, _Allocator>::swap(match_results& __m)
+{
+    using _CUDA_VSTD::swap;
+    swap(__matches_, __m.__matches_);
+    swap(__unmatched_, __m.__unmatched_);
+    swap(__prefix_, __m.__prefix_);
+    swap(__suffix_, __m.__suffix_);
+    swap(__position_start_, __m.__position_start_);
+    swap(__ready_, __m.__ready_);
+}
+
+typedef match_results<const char*>             cmatch;
+typedef match_results<const wchar_t*>          wcmatch;
+typedef match_results<string::const_iterator>  smatch;
+typedef match_results<wstring::const_iterator> wsmatch;
+
+template <class _BidirectionalIterator, class _Allocator>
+bool
+operator==(const match_results<_BidirectionalIterator, _Allocator>& __x,
+           const match_results<_BidirectionalIterator, _Allocator>& __y)
+{
+    if (__x.__ready_ != __y.__ready_)
+        return false;
+    if (!__x.__ready_)
+        return true;
+    return __x.__matches_ == __y.__matches_ &&
+           __x.__prefix_ == __y.__prefix_ &&
+           __x.__suffix_ == __y.__suffix_;
+}
+
+template <class _BidirectionalIterator, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const match_results<_BidirectionalIterator, _Allocator>& __x,
+           const match_results<_BidirectionalIterator, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _BidirectionalIterator, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(match_results<_BidirectionalIterator, _Allocator>& __x,
+     match_results<_BidirectionalIterator, _Allocator>& __y)
+{
+    __x.swap(__y);
+}
+
+// regex_search
+
+template <class _CharT, class _Traits>
+template <class _Allocator>
+bool
+basic_regex<_CharT, _Traits>::__match_at_start_ecma(
+        const _CharT* __first, const _CharT* __last,
+        match_results<const _CharT*, _Allocator>& __m,
+        regex_constants::match_flag_type __flags, bool __at_first) const
+{
+    vector<__state> __states;
+    __node* __st = __start_.get();
+    if (__st)
+    {
+        sub_match<const _CharT*> __unmatched;
+        __unmatched.first   = __last;
+        __unmatched.second  = __last;
+        __unmatched.matched = false;
+
+        __states.push_back(__state());
+        __states.back().__do_ = 0;
+        __states.back().__first_ = __first;
+        __states.back().__current_ = __first;
+        __states.back().__last_ = __last;
+        __states.back().__sub_matches_.resize(mark_count(), __unmatched);
+        __states.back().__loop_data_.resize(__loop_count());
+        __states.back().__node_ = __st;
+        __states.back().__flags_ = __flags;
+        __states.back().__at_first_ = __at_first;
+        int __counter = 0;
+        int __length = __last - __first;
+        do
+        {
+            ++__counter;
+            if (__counter % _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR == 0 &&
+                __counter / _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR >= __length)
+              __throw_regex_error<regex_constants::error_complexity>();
+            __state& __s = __states.back();
+            if (__s.__node_)
+                __s.__node_->__exec(__s);
+            switch (__s.__do_)
+            {
+            case __state::__end_state:
+                if ((__flags & regex_constants::match_not_null) &&
+                    __s.__current_ == __first)
+                {
+                  __states.pop_back();
+                  break;
+                }
+                if ((__flags & regex_constants::__full_match) &&
+                    __s.__current_ != __last)
+                {
+                  __states.pop_back();
+                  break;
+                }
+                __m.__matches_[0].first = __first;
+                __m.__matches_[0].second = _CUDA_VSTD::next(__first, __s.__current_ - __first);
+                __m.__matches_[0].matched = true;
+                for (unsigned __i = 0; __i < __s.__sub_matches_.size(); ++__i)
+                    __m.__matches_[__i+1] = __s.__sub_matches_[__i];
+                return true;
+            case __state::__accept_and_consume:
+            case __state::__repeat:
+            case __state::__accept_but_not_consume:
+                break;
+            case __state::__split:
+                {
+                __state __snext = __s;
+                __s.__node_->__exec_split(true, __s);
+                __snext.__node_->__exec_split(false, __snext);
+                __states.push_back(_CUDA_VSTD::move(__snext));
+                }
+                break;
+            case __state::__reject:
+                __states.pop_back();
+                break;
+            default:
+                __throw_regex_error<regex_constants::__re_err_unknown>();
+                break;
+
+            }
+        } while (!__states.empty());
+    }
+    return false;
+}
+
+template <class _CharT, class _Traits>
+template <class _Allocator>
+bool
+basic_regex<_CharT, _Traits>::__match_at_start_posix_nosubs(
+        const _CharT* __first, const _CharT* __last,
+        match_results<const _CharT*, _Allocator>& __m,
+        regex_constants::match_flag_type __flags, bool __at_first) const
+{
+    deque<__state> __states;
+    ptrdiff_t __highest_j = 0;
+    ptrdiff_t _Np = _CUDA_VSTD::distance(__first, __last);
+    __node* __st = __start_.get();
+    if (__st)
+    {
+        __states.push_back(__state());
+        __states.back().__do_ = 0;
+        __states.back().__first_ = __first;
+        __states.back().__current_ = __first;
+        __states.back().__last_ = __last;
+        __states.back().__loop_data_.resize(__loop_count());
+        __states.back().__node_ = __st;
+        __states.back().__flags_ = __flags;
+        __states.back().__at_first_ = __at_first;
+        bool __matched = false;
+        int __counter = 0;
+        int __length = __last - __first;
+        do
+        {
+            ++__counter;
+            if (__counter % _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR == 0 &&
+                __counter / _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR >= __length)
+              __throw_regex_error<regex_constants::error_complexity>();
+            __state& __s = __states.back();
+            if (__s.__node_)
+                __s.__node_->__exec(__s);
+            switch (__s.__do_)
+            {
+            case __state::__end_state:
+                if ((__flags & regex_constants::match_not_null) &&
+                    __s.__current_ == __first)
+                {
+                  __states.pop_back();
+                  break;
+                }
+                if ((__flags & regex_constants::__full_match) &&
+                    __s.__current_ != __last)
+                {
+                  __states.pop_back();
+                  break;
+                }
+                if (!__matched || __highest_j < __s.__current_ - __s.__first_)
+                    __highest_j = __s.__current_ - __s.__first_;
+                __matched = true;
+                if (__highest_j == _Np)
+                    __states.clear();
+                else
+                    __states.pop_back();
+                break;
+            case __state::__consume_input:
+                break;
+            case __state::__accept_and_consume:
+                __states.push_front(_CUDA_VSTD::move(__s));
+                __states.pop_back();
+                break;
+            case __state::__repeat:
+            case __state::__accept_but_not_consume:
+                break;
+            case __state::__split:
+                {
+                __state __snext = __s;
+                __s.__node_->__exec_split(true, __s);
+                __snext.__node_->__exec_split(false, __snext);
+                __states.push_back(_CUDA_VSTD::move(__snext));
+                }
+                break;
+            case __state::__reject:
+                __states.pop_back();
+                break;
+            default:
+                __throw_regex_error<regex_constants::__re_err_unknown>();
+                break;
+            }
+        } while (!__states.empty());
+        if (__matched)
+        {
+            __m.__matches_[0].first = __first;
+            __m.__matches_[0].second = _CUDA_VSTD::next(__first, __highest_j);
+            __m.__matches_[0].matched = true;
+            return true;
+        }
+    }
+    return false;
+}
+
+template <class _CharT, class _Traits>
+template <class _Allocator>
+bool
+basic_regex<_CharT, _Traits>::__match_at_start_posix_subs(
+        const _CharT* __first, const _CharT* __last,
+        match_results<const _CharT*, _Allocator>& __m,
+        regex_constants::match_flag_type __flags, bool __at_first) const
+{
+    vector<__state> __states;
+    __state __best_state;
+    ptrdiff_t __j = 0;
+    ptrdiff_t __highest_j = 0;
+    ptrdiff_t _Np = _CUDA_VSTD::distance(__first, __last);
+    __node* __st = __start_.get();
+    if (__st)
+    {
+        sub_match<const _CharT*> __unmatched;
+        __unmatched.first   = __last;
+        __unmatched.second  = __last;
+        __unmatched.matched = false;
+
+        __states.push_back(__state());
+        __states.back().__do_ = 0;
+        __states.back().__first_ = __first;
+        __states.back().__current_ = __first;
+        __states.back().__last_ = __last;
+        __states.back().__sub_matches_.resize(mark_count(), __unmatched);
+        __states.back().__loop_data_.resize(__loop_count());
+        __states.back().__node_ = __st;
+        __states.back().__flags_ = __flags;
+        __states.back().__at_first_ = __at_first;
+        const _CharT* __current = __first;
+        bool __matched = false;
+        int __counter = 0;
+        int __length = __last - __first;
+        do
+        {
+            ++__counter;
+            if (__counter % _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR == 0 &&
+                __counter / _LIBCUDACXX_REGEX_COMPLEXITY_FACTOR >= __length)
+              __throw_regex_error<regex_constants::error_complexity>();
+            __state& __s = __states.back();
+            if (__s.__node_)
+                __s.__node_->__exec(__s);
+            switch (__s.__do_)
+            {
+            case __state::__end_state:
+                if ((__flags & regex_constants::match_not_null) &&
+                    __s.__current_ == __first)
+                {
+                  __states.pop_back();
+                  break;
+                }
+                if ((__flags & regex_constants::__full_match) &&
+                    __s.__current_ != __last)
+                {
+                  __states.pop_back();
+                  break;
+                }
+                if (!__matched || __highest_j < __s.__current_ - __s.__first_)
+                {
+                    __highest_j = __s.__current_ - __s.__first_;
+                    __best_state = __s;
+                }
+                __matched = true;
+                if (__highest_j == _Np)
+                    __states.clear();
+                else
+                    __states.pop_back();
+                break;
+            case __state::__accept_and_consume:
+                __j += __s.__current_ - __current;
+                __current = __s.__current_;
+                break;
+            case __state::__repeat:
+            case __state::__accept_but_not_consume:
+                break;
+            case __state::__split:
+                {
+                __state __snext = __s;
+                __s.__node_->__exec_split(true, __s);
+                __snext.__node_->__exec_split(false, __snext);
+                __states.push_back(_CUDA_VSTD::move(__snext));
+                }
+                break;
+            case __state::__reject:
+                __states.pop_back();
+                break;
+            default:
+                __throw_regex_error<regex_constants::__re_err_unknown>();
+                break;
+            }
+        } while (!__states.empty());
+        if (__matched)
+        {
+            __m.__matches_[0].first = __first;
+            __m.__matches_[0].second = _CUDA_VSTD::next(__first, __highest_j);
+            __m.__matches_[0].matched = true;
+            for (unsigned __i = 0; __i < __best_state.__sub_matches_.size(); ++__i)
+                __m.__matches_[__i+1] = __best_state.__sub_matches_[__i];
+            return true;
+        }
+    }
+    return false;
+}
+
+template <class _CharT, class _Traits>
+template <class _Allocator>
+bool
+basic_regex<_CharT, _Traits>::__match_at_start(
+        const _CharT* __first, const _CharT* __last,
+        match_results<const _CharT*, _Allocator>& __m,
+        regex_constants::match_flag_type __flags, bool __at_first) const
+{
+    if (__get_grammar(__flags_) == ECMAScript)
+        return __match_at_start_ecma(__first, __last, __m, __flags, __at_first);
+    if (mark_count() == 0)
+        return __match_at_start_posix_nosubs(__first, __last, __m, __flags, __at_first);
+    return __match_at_start_posix_subs(__first, __last, __m, __flags, __at_first);
+}
+
+template <class _CharT, class _Traits>
+template <class _Allocator>
+bool
+basic_regex<_CharT, _Traits>::__search(
+        const _CharT* __first, const _CharT* __last,
+        match_results<const _CharT*, _Allocator>& __m,
+        regex_constants::match_flag_type __flags) const
+{
+    __m.__init(1 + mark_count(), __first, __last,
+                                    __flags & regex_constants::__no_update_pos);
+    if (__match_at_start(__first, __last, __m, __flags,
+                                    !(__flags & regex_constants::__no_update_pos)))
+    {
+        __m.__prefix_.second = __m[0].first;
+        __m.__prefix_.matched = __m.__prefix_.first != __m.__prefix_.second;
+        __m.__suffix_.first = __m[0].second;
+        __m.__suffix_.matched = __m.__suffix_.first != __m.__suffix_.second;
+        return true;
+    }
+    if (__first != __last && !(__flags & regex_constants::match_continuous))
+    {
+        __flags |= regex_constants::match_prev_avail;
+        for (++__first; __first != __last; ++__first)
+        {
+            __m.__matches_.assign(__m.size(), __m.__unmatched_);
+            if (__match_at_start(__first, __last, __m, __flags, false))
+            {
+                __m.__prefix_.second = __m[0].first;
+                __m.__prefix_.matched = __m.__prefix_.first != __m.__prefix_.second;
+                __m.__suffix_.first = __m[0].second;
+                __m.__suffix_.matched = __m.__suffix_.first != __m.__suffix_.second;
+                return true;
+            }
+            __m.__matches_.assign(__m.size(), __m.__unmatched_);
+        }
+    }
+    __m.__matches_.clear();
+    return false;
+}
+
+template <class _BidirectionalIterator, class _Allocator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(_BidirectionalIterator __first, _BidirectionalIterator __last,
+             match_results<_BidirectionalIterator, _Allocator>& __m,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    int __offset = (__flags & regex_constants::match_prev_avail) ? 1 : 0;
+    basic_string<_CharT> __s(_CUDA_VSTD::prev(__first, __offset), __last);
+    match_results<const _CharT*> __mc;
+    bool __r = __e.__search(__s.data() + __offset, __s.data() + __s.size(), __mc, __flags);
+    __m.__assign(__first, __last, __mc, __flags & regex_constants::__no_update_pos);
+    return __r;
+}
+
+template <class _Iter, class _Allocator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(__wrap_iter<_Iter> __first,
+             __wrap_iter<_Iter> __last,
+             match_results<__wrap_iter<_Iter>, _Allocator>& __m,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    match_results<const _CharT*> __mc;
+    bool __r = __e.__search(__first.base(), __last.base(), __mc, __flags);
+    __m.__assign(__first, __last, __mc, __flags & regex_constants::__no_update_pos);
+    return __r;
+}
+
+template <class _Allocator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(const _CharT* __first, const _CharT* __last,
+             match_results<const _CharT*, _Allocator>& __m,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return __e.__search(__first, __last, __m, __flags);
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(_BidirectionalIterator __first, _BidirectionalIterator __last,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    basic_string<_CharT> __s(__first, __last);
+    match_results<const _CharT*> __mc;
+    return __e.__search(__s.data(), __s.data() + __s.size(), __mc, __flags);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(const _CharT* __first, const _CharT* __last,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    match_results<const _CharT*> __mc;
+    return __e.__search(__first, __last, __mc, __flags);
+}
+
+template <class _CharT, class _Allocator, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(const _CharT* __str, match_results<const _CharT*, _Allocator>& __m,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return __e.__search(__str, __str + _Traits::length(__str), __m, __flags);
+}
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(const _CharT* __str, const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    match_results<const _CharT*> __m;
+    return _CUDA_VSTD::regex_search(__str, __m, __e, __flags);
+}
+
+template <class _ST, class _SA, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(const basic_string<_CharT, _ST, _SA>& __s,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    match_results<const _CharT*> __mc;
+    return __e.__search(__s.data(), __s.data() + __s.size(), __mc, __flags);
+}
+
+template <class _ST, class _SA, class _Allocator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_search(const basic_string<_CharT, _ST, _SA>& __s,
+             match_results<typename basic_string<_CharT, _ST, _SA>::const_iterator, _Allocator>& __m,
+             const basic_regex<_CharT, _Traits>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    match_results<const _CharT*> __mc;
+    bool __r = __e.__search(__s.data(), __s.data() + __s.size(), __mc, __flags);
+    __m.__assign(__s.begin(), __s.end(), __mc, __flags & regex_constants::__no_update_pos);
+    return __r;
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _ST, class _SA, class _Ap, class _Cp, class _Tp>
+bool
+regex_search(const basic_string<_Cp, _ST, _SA>&& __s,
+             match_results<typename basic_string<_Cp, _ST, _SA>::const_iterator, _Ap>&,
+             const basic_regex<_Cp, _Tp>& __e,
+             regex_constants::match_flag_type __flags = regex_constants::match_default) = delete;
+#endif
+
+// regex_match
+
+template <class _BidirectionalIterator, class _Allocator, class _CharT, class _Traits>
+bool
+regex_match(_BidirectionalIterator __first, _BidirectionalIterator __last,
+            match_results<_BidirectionalIterator, _Allocator>& __m,
+            const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    bool __r = _CUDA_VSTD::regex_search(
+        __first, __last, __m, __e,
+        __flags | regex_constants::match_continuous |
+        regex_constants::__full_match);
+    if (__r)
+    {
+        __r = !__m.suffix().matched;
+        if (!__r)
+            __m.__matches_.clear();
+    }
+    return __r;
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_match(_BidirectionalIterator __first, _BidirectionalIterator __last,
+            const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    match_results<_BidirectionalIterator> __m;
+    return _CUDA_VSTD::regex_match(__first, __last, __m, __e, __flags);
+}
+
+template <class _CharT, class _Allocator, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_match(const _CharT* __str, match_results<const _CharT*, _Allocator>& __m,
+            const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return _CUDA_VSTD::regex_match(__str, __str + _Traits::length(__str), __m, __e, __flags);
+}
+
+template <class _ST, class _SA, class _Allocator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_match(const basic_string<_CharT, _ST, _SA>& __s,
+            match_results<typename basic_string<_CharT, _ST, _SA>::const_iterator, _Allocator>& __m,
+            const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return _CUDA_VSTD::regex_match(__s.begin(), __s.end(), __m, __e, __flags);
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _ST, class _SA, class _Allocator, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_match(const basic_string<_CharT, _ST, _SA>&& __s,
+            match_results<typename basic_string<_CharT, _ST, _SA>::const_iterator, _Allocator>& __m,
+            const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default) = delete;
+#endif
+
+template <class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_match(const _CharT* __str, const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return _CUDA_VSTD::regex_match(__str, __str + _Traits::length(__str), __e, __flags);
+}
+
+template <class _ST, class _SA, class _CharT, class _Traits>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+regex_match(const basic_string<_CharT, _ST, _SA>& __s,
+            const basic_regex<_CharT, _Traits>& __e,
+            regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return _CUDA_VSTD::regex_match(__s.begin(), __s.end(), __e, __flags);
+}
+
+// regex_iterator
+
+template <class _BidirectionalIterator,
+          class _CharT = typename iterator_traits<_BidirectionalIterator>::value_type,
+          class _Traits = regex_traits<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS regex_iterator
+{
+public:
+    typedef basic_regex<_CharT, _Traits>          regex_type;
+    typedef match_results<_BidirectionalIterator> value_type;
+    typedef ptrdiff_t                             difference_type;
+    typedef const value_type*                     pointer;
+    typedef const value_type&                     reference;
+    typedef forward_iterator_tag                  iterator_category;
+
+private:
+    _BidirectionalIterator           __begin_;
+    _BidirectionalIterator           __end_;
+    const regex_type*                __pregex_;
+    regex_constants::match_flag_type __flags_;
+    value_type                       __match_;
+
+public:
+    regex_iterator();
+    regex_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                   const regex_type& __re,
+                   regex_constants::match_flag_type __m
+                                              = regex_constants::match_default);
+#if _LIBCUDACXX_STD_VER > 11
+    regex_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                   const regex_type&& __re,
+                   regex_constants::match_flag_type __m
+                                     = regex_constants::match_default) = delete;
+#endif
+
+    bool operator==(const regex_iterator& __x) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const regex_iterator& __x) const {return !(*this == __x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return  __match_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const  {return _CUDA_VSTD::addressof(__match_);}
+
+    regex_iterator& operator++();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    regex_iterator operator++(int)
+    {
+        regex_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+};
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_iterator<_BidirectionalIterator, _CharT, _Traits>::regex_iterator()
+    : __begin_(), __end_(), __pregex_(nullptr), __flags_(), __match_()
+{
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                   const regex_type& __re, regex_constants::match_flag_type __m)
+    : __begin_(__a),
+      __end_(__b),
+      __pregex_(_CUDA_VSTD::addressof(__re)),
+      __flags_(__m)
+{
+    _CUDA_VSTD::regex_search(__begin_, __end_, __match_, *__pregex_, __flags_);
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+bool
+regex_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    operator==(const regex_iterator& __x) const
+{
+    if (__match_.empty() && __x.__match_.empty())
+        return true;
+    if (__match_.empty() || __x.__match_.empty())
+        return false;
+    return __begin_ == __x.__begin_       &&
+           __end_ == __x.__end_           &&
+           __pregex_ == __x.__pregex_     &&
+           __flags_ == __x.__flags_       &&
+           __match_[0] == __x.__match_[0];
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_iterator<_BidirectionalIterator, _CharT, _Traits>&
+regex_iterator<_BidirectionalIterator, _CharT, _Traits>::operator++()
+{
+    __flags_ |= regex_constants::__no_update_pos;
+    _BidirectionalIterator __start = __match_[0].second;
+    if (__match_[0].first == __match_[0].second)
+    {
+        if (__start == __end_)
+        {
+            __match_ = value_type();
+            return *this;
+        }
+        else if (_CUDA_VSTD::regex_search(__start, __end_, __match_, *__pregex_,
+                                    __flags_ | regex_constants::match_not_null |
+                                    regex_constants::match_continuous))
+            return *this;
+        else
+            ++__start;
+    }
+    __flags_ |= regex_constants::match_prev_avail;
+    if (!_CUDA_VSTD::regex_search(__start, __end_, __match_, *__pregex_, __flags_))
+        __match_ = value_type();
+    return *this;
+}
+
+typedef regex_iterator<const char*>             cregex_iterator;
+typedef regex_iterator<const wchar_t*>          wcregex_iterator;
+typedef regex_iterator<string::const_iterator>  sregex_iterator;
+typedef regex_iterator<wstring::const_iterator> wsregex_iterator;
+
+// regex_token_iterator
+
+template <class _BidirectionalIterator,
+          class _CharT = typename iterator_traits<_BidirectionalIterator>::value_type,
+          class _Traits = regex_traits<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS regex_token_iterator
+{
+public:
+    typedef basic_regex<_CharT, _Traits>      regex_type;
+    typedef sub_match<_BidirectionalIterator> value_type;
+    typedef ptrdiff_t                         difference_type;
+    typedef const value_type*                 pointer;
+    typedef const value_type&                 reference;
+    typedef forward_iterator_tag              iterator_category;
+
+private:
+    typedef regex_iterator<_BidirectionalIterator, _CharT, _Traits> _Position;
+
+    _Position         __position_;
+    const value_type* __result_;
+    value_type        __suffix_;
+    ptrdiff_t         __n_;
+    vector<int>       __subs_;
+
+public:
+    regex_token_iterator();
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type& __re, int __submatch = 0,
+                         regex_constants::match_flag_type __m =
+                                                regex_constants::match_default);
+#if _LIBCUDACXX_STD_VER > 11
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type&& __re, int __submatch = 0,
+                         regex_constants::match_flag_type __m =
+                                       regex_constants::match_default) = delete;
+#endif
+
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type& __re, const vector<int>& __submatches,
+                         regex_constants::match_flag_type __m =
+                                                regex_constants::match_default);
+#if _LIBCUDACXX_STD_VER > 11
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type&& __re, const vector<int>& __submatches,
+                         regex_constants::match_flag_type __m =
+                                     regex_constants::match_default) = delete;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type& __re,
+                         initializer_list<int> __submatches,
+                         regex_constants::match_flag_type __m =
+                                                regex_constants::match_default);
+
+#if _LIBCUDACXX_STD_VER > 11
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type&& __re,
+                         initializer_list<int> __submatches,
+                         regex_constants::match_flag_type __m =
+                                       regex_constants::match_default) = delete;
+#endif
+#endif  // _LIBCUDACXX_CXX03_LANG
+    template <size_t _Np>
+        regex_token_iterator(_BidirectionalIterator __a,
+                             _BidirectionalIterator __b,
+                             const regex_type& __re,
+                             const int (&__submatches)[_Np],
+                             regex_constants::match_flag_type __m =
+                                                regex_constants::match_default);
+#if _LIBCUDACXX_STD_VER > 11
+    template <std::size_t _Np>
+        regex_token_iterator(_BidirectionalIterator __a,
+                             _BidirectionalIterator __b,
+                             const regex_type&& __re,
+                             const int (&__submatches)[_Np],
+                             regex_constants::match_flag_type __m =
+                                      regex_constants::match_default) = delete;
+#endif
+
+    regex_token_iterator(const regex_token_iterator&);
+    regex_token_iterator& operator=(const regex_token_iterator&);
+
+    bool operator==(const regex_token_iterator& __x) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const regex_token_iterator& __x) const {return !(*this == __x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type& operator*() const {return *__result_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type* operator->() const {return __result_;}
+
+    regex_token_iterator& operator++();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    regex_token_iterator operator++(int)
+    {
+        regex_token_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+private:
+    void __init(_BidirectionalIterator __a, _BidirectionalIterator __b);
+    void __establish_result () {
+        if (__subs_[__n_] == -1)
+            __result_ = &__position_->prefix();
+        else
+            __result_ = &(*__position_)[__subs_[__n_]];
+        }
+};
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_token_iterator()
+    : __result_(nullptr),
+      __suffix_(),
+      __n_(0)
+{
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+void
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    __init(_BidirectionalIterator __a, _BidirectionalIterator __b)
+{
+    if (__position_ != _Position())
+        __establish_result ();
+    else if (__subs_[__n_] == -1)
+    {
+        __suffix_.matched = true;
+        __suffix_.first = __a;
+        __suffix_.second = __b;
+        __result_ = &__suffix_;
+    }
+    else
+        __result_ = nullptr;
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type& __re, int __submatch,
+                         regex_constants::match_flag_type __m)
+    : __position_(__a, __b, __re, __m),
+      __n_(0),
+      __subs_(1, __submatch)
+{
+    __init(__a, __b);
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type& __re, const vector<int>& __submatches,
+                         regex_constants::match_flag_type __m)
+    : __position_(__a, __b, __re, __m),
+      __n_(0),
+      __subs_(__submatches)
+{
+    __init(__a, __b);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                         const regex_type& __re,
+                         initializer_list<int> __submatches,
+                         regex_constants::match_flag_type __m)
+    : __position_(__a, __b, __re, __m),
+      __n_(0),
+      __subs_(__submatches)
+{
+    __init(__a, __b);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+template <size_t _Np>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_token_iterator(_BidirectionalIterator __a, _BidirectionalIterator __b,
+                             const regex_type& __re,
+                             const int (&__submatches)[_Np],
+                             regex_constants::match_flag_type __m)
+    : __position_(__a, __b, __re, __m),
+      __n_(0),
+      __subs_(begin(__submatches), end(__submatches))
+{
+    __init(__a, __b);
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    regex_token_iterator(const regex_token_iterator& __x)
+    : __position_(__x.__position_),
+      __result_(__x.__result_),
+      __suffix_(__x.__suffix_),
+      __n_(__x.__n_),
+      __subs_(__x.__subs_)
+{
+    if (__x.__result_ == &__x.__suffix_)
+        __result_ = &__suffix_;
+    else if ( __result_ != nullptr )
+        __establish_result ();
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>&
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    operator=(const regex_token_iterator& __x)
+{
+    if (this != &__x)
+    {
+        __position_ = __x.__position_;
+        if (__x.__result_ == &__x.__suffix_)
+            __result_ = &__suffix_;
+        else
+            __result_ = __x.__result_;
+        __suffix_ = __x.__suffix_;
+        __n_ = __x.__n_;
+        __subs_ = __x.__subs_;
+
+        if ( __result_ != nullptr && __result_ != &__suffix_ )
+            __establish_result();
+    }
+    return *this;
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+bool
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::
+    operator==(const regex_token_iterator& __x) const
+{
+    if (__result_ == nullptr && __x.__result_ == nullptr)
+        return true;
+    if (__result_ == &__suffix_ && __x.__result_ == &__x.__suffix_ &&
+            __suffix_ == __x.__suffix_)
+        return true;
+    if (__result_ == nullptr || __x.__result_ == nullptr)
+        return false;
+    if (__result_ == &__suffix_ || __x.__result_ == &__x.__suffix_)
+        return false;
+    return __position_ == __x.__position_ && __n_ == __x.__n_ &&
+           __subs_ == __x.__subs_;
+}
+
+template <class _BidirectionalIterator, class _CharT, class _Traits>
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>&
+regex_token_iterator<_BidirectionalIterator, _CharT, _Traits>::operator++()
+{
+    _Position __prev = __position_;
+    if (__result_ == &__suffix_)
+        __result_ = nullptr;
+    else if (static_cast<size_t>(__n_ + 1) < __subs_.size())
+    {
+        ++__n_;
+        __establish_result();
+    }
+    else
+    {
+        __n_ = 0;
+        ++__position_;
+        if (__position_ != _Position())
+            __establish_result();
+        else
+        {
+            if (_CUDA_VSTD::find(__subs_.begin(), __subs_.end(), -1) != __subs_.end()
+                && __prev->suffix().length() != 0)
+            {
+                __suffix_.matched = true;
+                __suffix_.first = __prev->suffix().first;
+                __suffix_.second = __prev->suffix().second;
+                __result_ = &__suffix_;
+            }
+            else
+                __result_ = nullptr;
+        }
+    }
+    return *this;
+}
+
+typedef regex_token_iterator<const char*>             cregex_token_iterator;
+typedef regex_token_iterator<const wchar_t*>          wcregex_token_iterator;
+typedef regex_token_iterator<string::const_iterator>  sregex_token_iterator;
+typedef regex_token_iterator<wstring::const_iterator> wsregex_token_iterator;
+
+// regex_replace
+
+template <class _OutputIterator, class _BidirectionalIterator,
+          class _Traits, class _CharT>
+_OutputIterator
+regex_replace(_OutputIterator __output_iter,
+              _BidirectionalIterator __first, _BidirectionalIterator __last,
+              const basic_regex<_CharT, _Traits>& __e, const _CharT* __fmt,
+              regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    typedef regex_iterator<_BidirectionalIterator, _CharT, _Traits> _Iter;
+    _Iter __i(__first, __last, __e, __flags);
+    _Iter __eof;
+    if (__i == __eof)
+    {
+        if (!(__flags & regex_constants::format_no_copy))
+            __output_iter = _CUDA_VSTD::copy(__first, __last, __output_iter);
+    }
+    else
+    {
+        sub_match<_BidirectionalIterator> __lm;
+        for (size_t __len = char_traits<_CharT>::length(__fmt); __i != __eof; ++__i)
+        {
+            if (!(__flags & regex_constants::format_no_copy))
+                __output_iter = _CUDA_VSTD::copy(__i->prefix().first, __i->prefix().second, __output_iter);
+            __output_iter = __i->format(__output_iter, __fmt, __fmt + __len, __flags);
+            __lm = __i->suffix();
+            if (__flags & regex_constants::format_first_only)
+                break;
+        }
+        if (!(__flags & regex_constants::format_no_copy))
+            __output_iter = _CUDA_VSTD::copy(__lm.first, __lm.second, __output_iter);
+    }
+    return __output_iter;
+}
+
+template <class _OutputIterator, class _BidirectionalIterator,
+          class _Traits, class _CharT, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_OutputIterator
+regex_replace(_OutputIterator __output_iter,
+              _BidirectionalIterator __first, _BidirectionalIterator __last,
+              const basic_regex<_CharT, _Traits>& __e,
+              const basic_string<_CharT, _ST, _SA>& __fmt,
+              regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    return _CUDA_VSTD::regex_replace(__output_iter, __first, __last, __e, __fmt.c_str(), __flags);
+}
+
+template <class _Traits, class _CharT, class _ST, class _SA, class _FST,
+          class _FSA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _ST, _SA>
+regex_replace(const basic_string<_CharT, _ST, _SA>& __s,
+              const basic_regex<_CharT, _Traits>& __e,
+              const basic_string<_CharT, _FST, _FSA>& __fmt,
+              regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    basic_string<_CharT, _ST, _SA> __r;
+    _CUDA_VSTD::regex_replace(back_inserter(__r), __s.begin(), __s.end(), __e,
+                        __fmt.c_str(), __flags);
+    return __r;
+}
+
+template <class _Traits, class _CharT, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _ST, _SA>
+regex_replace(const basic_string<_CharT, _ST, _SA>& __s,
+              const basic_regex<_CharT, _Traits>& __e, const _CharT* __fmt,
+              regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    basic_string<_CharT, _ST, _SA> __r;
+    _CUDA_VSTD::regex_replace(back_inserter(__r), __s.begin(), __s.end(), __e,
+                        __fmt, __flags);
+    return __r;
+}
+
+template <class _Traits, class _CharT, class _ST, class _SA>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT>
+regex_replace(const _CharT* __s,
+              const basic_regex<_CharT, _Traits>& __e,
+              const basic_string<_CharT, _ST, _SA>& __fmt,
+              regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    basic_string<_CharT> __r;
+    _CUDA_VSTD::regex_replace(back_inserter(__r), __s,
+                        __s + char_traits<_CharT>::length(__s), __e,
+                        __fmt.c_str(), __flags);
+    return __r;
+}
+
+template <class _Traits, class _CharT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT>
+regex_replace(const _CharT* __s,
+              const basic_regex<_CharT, _Traits>& __e,
+              const _CharT* __fmt,
+              regex_constants::match_flag_type __flags = regex_constants::match_default)
+{
+    basic_string<_CharT> __r;
+    _CUDA_VSTD::regex_replace(back_inserter(__r), __s,
+                        __s + char_traits<_CharT>::length(__s), __e,
+                        __fmt, __flags);
+    return __r;
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_REGEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator
new file mode 100644
index 000000000000..ddf1e85ed1e0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/scoped_allocator
@@ -0,0 +1,683 @@
+// -*- C++ -*-
+//===-------------------------- scoped_allocator --------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SCOPED_ALLOCATOR
+#define _LIBCUDACXX_SCOPED_ALLOCATOR
+
+/*
+    scoped_allocator synopsis
+
+namespace std
+{
+
+template <class OuterAlloc, class... InnerAllocs>
+class scoped_allocator_adaptor : public OuterAlloc
+{
+    typedef allocator_traits<OuterAlloc> OuterTraits; // exposition only
+    scoped_allocator_adaptor<InnerAllocs...> inner;   // exposition only
+public:
+
+    typedef OuterAlloc outer_allocator_type;
+    typedef see below inner_allocator_type;
+
+    typedef typename OuterTraits::value_type value_type;
+    typedef typename OuterTraits::size_type size_type;
+    typedef typename OuterTraits::difference_type difference_type;
+    typedef typename OuterTraits::pointer pointer;
+    typedef typename OuterTraits::const_pointer const_pointer;
+    typedef typename OuterTraits::void_pointer void_pointer;
+    typedef typename OuterTraits::const_void_pointer const_void_pointer;
+
+    typedef see below propagate_on_container_copy_assignment;
+    typedef see below propagate_on_container_move_assignment;
+    typedef see below propagate_on_container_swap;
+    typedef see below is_always_equal;
+
+    template <class Tp>
+        struct rebind
+        {
+            typedef scoped_allocator_adaptor<
+                OuterTraits::template rebind_alloc<Tp>, InnerAllocs...> other;
+        };
+
+    scoped_allocator_adaptor();
+    template <class OuterA2>
+        scoped_allocator_adaptor(OuterA2&& outerAlloc,
+                                 const InnerAllocs&... innerAllocs) noexcept;
+    scoped_allocator_adaptor(const scoped_allocator_adaptor& other) noexcept;
+    scoped_allocator_adaptor(scoped_allocator_adaptor&& other) noexcept;
+    template <class OuterA2>
+        scoped_allocator_adaptor(const scoped_allocator_adaptor<OuterA2, InnerAllocs...>& other) noexcept;
+    template <class OuterA2>
+        scoped_allocator_adaptor(const scoped_allocator_adaptor<OuterA2, InnerAllocs...>&& other) noexcept;
+
+    scoped_allocator_adaptor& operator=(const scoped_allocator_adaptor&) = default;
+    scoped_allocator_adaptor& operator=(scoped_allocator_adaptor&&) = default;
+    ~scoped_allocator_adaptor();
+
+    inner_allocator_type& inner_allocator() noexcept;
+    const inner_allocator_type& inner_allocator() const noexcept;
+
+    outer_allocator_type& outer_allocator() noexcept;
+    const outer_allocator_type& outer_allocator() const noexcept;
+
+    pointer allocate(size_type n);                           // [[nodiscard]] in C++20
+    pointer allocate(size_type n, const_void_pointer hint);  // [[nodiscard]] in C++20
+    void deallocate(pointer p, size_type n) noexcept;
+
+    size_type max_size() const;
+    template <class T, class... Args> void construct(T* p, Args&& args);
+    template <class T1, class T2, class... Args1, class... Args2>
+        void construct(pair<T1, T2>* p, piecewise_construct t, tuple<Args1...> x,
+                       tuple<Args2...> y);
+    template <class T1, class T2>
+        void construct(pair<T1, T2>* p);
+    template <class T1, class T2, class U, class V>
+        void construct(pair<T1, T2>* p, U&& x, V&& y);
+    template <class T1, class T2, class U, class V>
+        void construct(pair<T1, T2>* p, const pair<U, V>& x);
+    template <class T1, class T2, class U, class V>
+        void construct(pair<T1, T2>* p, pair<U, V>&& x);
+    template <class T> void destroy(T* p);
+
+    template <class T> void destroy(T* p) noexcept;
+
+    scoped_allocator_adaptor select_on_container_copy_construction() const noexcept;
+};
+
+template <class OuterA1, class OuterA2, class... InnerAllocs>
+    bool
+    operator==(const scoped_allocator_adaptor<OuterA1, InnerAllocs...>& a,
+               const scoped_allocator_adaptor<OuterA2, InnerAllocs...>& b) noexcept;
+
+template <class OuterA1, class OuterA2, class... InnerAllocs>
+    bool
+    operator!=(const scoped_allocator_adaptor<OuterA1, InnerAllocs...>& a,
+               const scoped_allocator_adaptor<OuterA2, InnerAllocs...>& b) noexcept;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <memory>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+
+// scoped_allocator_adaptor
+
+template <class ..._Allocs>
+class scoped_allocator_adaptor;
+
+template <class ..._Allocs> struct __get_poc_copy_assignment;
+
+template <class _A0>
+struct __get_poc_copy_assignment<_A0>
+{
+    static const bool value = allocator_traits<_A0>::
+                              propagate_on_container_copy_assignment::value;
+};
+
+template <class _A0, class ..._Allocs>
+struct __get_poc_copy_assignment<_A0, _Allocs...>
+{
+    static const bool value =
+        allocator_traits<_A0>::propagate_on_container_copy_assignment::value ||
+        __get_poc_copy_assignment<_Allocs...>::value;
+};
+
+template <class ..._Allocs> struct __get_poc_move_assignment;
+
+template <class _A0>
+struct __get_poc_move_assignment<_A0>
+{
+    static const bool value = allocator_traits<_A0>::
+                              propagate_on_container_move_assignment::value;
+};
+
+template <class _A0, class ..._Allocs>
+struct __get_poc_move_assignment<_A0, _Allocs...>
+{
+    static const bool value =
+        allocator_traits<_A0>::propagate_on_container_move_assignment::value ||
+        __get_poc_move_assignment<_Allocs...>::value;
+};
+
+template <class ..._Allocs> struct __get_poc_swap;
+
+template <class _A0>
+struct __get_poc_swap<_A0>
+{
+    static const bool value = allocator_traits<_A0>::
+                              propagate_on_container_swap::value;
+};
+
+template <class _A0, class ..._Allocs>
+struct __get_poc_swap<_A0, _Allocs...>
+{
+    static const bool value =
+        allocator_traits<_A0>::propagate_on_container_swap::value ||
+        __get_poc_swap<_Allocs...>::value;
+};
+
+template <class ..._Allocs> struct __get_is_always_equal;
+
+template <class _A0>
+struct __get_is_always_equal<_A0>
+{
+    static const bool value = allocator_traits<_A0>::is_always_equal::value;
+};
+
+template <class _A0, class ..._Allocs>
+struct __get_is_always_equal<_A0, _Allocs...>
+{
+    static const bool value =
+        allocator_traits<_A0>::is_always_equal::value &&
+        __get_is_always_equal<_Allocs...>::value;
+};
+
+template <class ..._Allocs>
+class __scoped_allocator_storage;
+
+template <class _OuterAlloc, class... _InnerAllocs>
+class __scoped_allocator_storage<_OuterAlloc, _InnerAllocs...>
+    : public _OuterAlloc
+{
+    typedef _OuterAlloc outer_allocator_type;
+protected:
+    typedef scoped_allocator_adaptor<_InnerAllocs...> inner_allocator_type;
+
+private:
+    inner_allocator_type __inner_;
+
+protected:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __scoped_allocator_storage() _NOEXCEPT {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(_OuterA2&& __outerAlloc,
+                                   const _InnerAllocs& ...__innerAllocs) _NOEXCEPT
+            : outer_allocator_type(_CUDA_VSTD::forward<_OuterA2>(__outerAlloc)),
+              __inner_(__innerAllocs...) {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, const _OuterA2&>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(
+            const __scoped_allocator_storage<_OuterA2, _InnerAllocs...>& __other) _NOEXCEPT
+            : outer_allocator_type(__other.outer_allocator()),
+              __inner_(__other.inner_allocator()) {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(
+            __scoped_allocator_storage<_OuterA2, _InnerAllocs...>&& __other) _NOEXCEPT
+            : outer_allocator_type(_CUDA_VSTD::move(__other.outer_allocator())),
+              __inner_(_CUDA_VSTD::move(__other.inner_allocator())) {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(_OuterA2&& __o,
+                                   const inner_allocator_type& __i) _NOEXCEPT
+            : outer_allocator_type(_CUDA_VSTD::forward<_OuterA2>(__o)),
+              __inner_(__i)
+        {
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inner_allocator_type& inner_allocator() _NOEXCEPT             {return __inner_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const inner_allocator_type& inner_allocator() const _NOEXCEPT {return __inner_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    outer_allocator_type& outer_allocator() _NOEXCEPT
+        {return static_cast<outer_allocator_type&>(*this);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const outer_allocator_type& outer_allocator() const _NOEXCEPT
+        {return static_cast<const outer_allocator_type&>(*this);}
+
+    scoped_allocator_adaptor<outer_allocator_type, _InnerAllocs...>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    select_on_container_copy_construction() const _NOEXCEPT
+        {
+            return scoped_allocator_adaptor<outer_allocator_type, _InnerAllocs...>
+            (
+                allocator_traits<outer_allocator_type>::
+                    select_on_container_copy_construction(outer_allocator()),
+                allocator_traits<inner_allocator_type>::
+                    select_on_container_copy_construction(inner_allocator())
+            );
+        }
+
+    template <class...> friend class __scoped_allocator_storage;
+};
+
+template <class _OuterAlloc>
+class __scoped_allocator_storage<_OuterAlloc>
+    : public _OuterAlloc
+{
+    typedef _OuterAlloc outer_allocator_type;
+protected:
+    typedef scoped_allocator_adaptor<_OuterAlloc> inner_allocator_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __scoped_allocator_storage() _NOEXCEPT {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(_OuterA2&& __outerAlloc) _NOEXCEPT
+            : outer_allocator_type(_CUDA_VSTD::forward<_OuterA2>(__outerAlloc)) {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, const _OuterA2&>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(
+            const __scoped_allocator_storage<_OuterA2>& __other) _NOEXCEPT
+            : outer_allocator_type(__other.outer_allocator()) {}
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __scoped_allocator_storage(
+            __scoped_allocator_storage<_OuterA2>&& __other) _NOEXCEPT
+            : outer_allocator_type(_CUDA_VSTD::move(__other.outer_allocator())) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inner_allocator_type& inner_allocator() _NOEXCEPT
+        {return static_cast<inner_allocator_type&>(*this);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const inner_allocator_type& inner_allocator() const _NOEXCEPT
+        {return static_cast<const inner_allocator_type&>(*this);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    outer_allocator_type& outer_allocator() _NOEXCEPT
+        {return static_cast<outer_allocator_type&>(*this);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const outer_allocator_type& outer_allocator() const _NOEXCEPT
+        {return static_cast<const outer_allocator_type&>(*this);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    scoped_allocator_adaptor<outer_allocator_type>
+    select_on_container_copy_construction() const _NOEXCEPT
+        {return scoped_allocator_adaptor<outer_allocator_type>(
+            allocator_traits<outer_allocator_type>::
+                select_on_container_copy_construction(outer_allocator())
+        );}
+
+    __scoped_allocator_storage(const outer_allocator_type& __o,
+                               const inner_allocator_type& __i) _NOEXCEPT;
+
+    template <class...> friend class __scoped_allocator_storage;
+};
+
+// __outermost
+
+template <class _Alloc>
+decltype(declval<_Alloc>().outer_allocator(), true_type())
+__has_outer_allocator_test(_Alloc&& __a);
+
+template <class _Alloc>
+false_type
+__has_outer_allocator_test(const volatile _Alloc& __a);
+
+template <class _Alloc>
+struct __has_outer_allocator
+    : public common_type
+             <
+                 decltype(__has_outer_allocator_test(declval<_Alloc&>()))
+             >::type
+{
+};
+
+template <class _Alloc, bool = __has_outer_allocator<_Alloc>::value>
+struct __outermost
+{
+    typedef _Alloc type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    type& operator()(type& __a) const _NOEXCEPT {return __a;}
+};
+
+template <class _Alloc>
+struct __outermost<_Alloc, true>
+{
+    typedef typename remove_reference
+                     <
+                        decltype(_CUDA_VSTD::declval<_Alloc>().outer_allocator())
+                     >::type                                    _OuterAlloc;
+    typedef typename __outermost<_OuterAlloc>::type             type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    type& operator()(_Alloc& __a) const _NOEXCEPT
+        {return __outermost<_OuterAlloc>()(__a.outer_allocator());}
+};
+
+template <class _OuterAlloc, class... _InnerAllocs>
+class _LIBCUDACXX_TEMPLATE_VIS scoped_allocator_adaptor<_OuterAlloc, _InnerAllocs...>
+    : public __scoped_allocator_storage<_OuterAlloc, _InnerAllocs...>
+{
+    typedef __scoped_allocator_storage<_OuterAlloc, _InnerAllocs...> base;
+    typedef allocator_traits<_OuterAlloc>             _OuterTraits;
+public:
+    typedef _OuterAlloc                               outer_allocator_type;
+    typedef typename base::inner_allocator_type       inner_allocator_type;
+    typedef typename _OuterTraits::size_type          size_type;
+    typedef typename _OuterTraits::difference_type    difference_type;
+    typedef typename _OuterTraits::pointer            pointer;
+    typedef typename _OuterTraits::const_pointer      const_pointer;
+    typedef typename _OuterTraits::void_pointer       void_pointer;
+    typedef typename _OuterTraits::const_void_pointer const_void_pointer;
+
+    typedef integral_constant
+            <
+                bool,
+                __get_poc_copy_assignment<outer_allocator_type,
+                                          _InnerAllocs...>::value
+            > propagate_on_container_copy_assignment;
+    typedef integral_constant
+            <
+                bool,
+                __get_poc_move_assignment<outer_allocator_type,
+                                          _InnerAllocs...>::value
+            > propagate_on_container_move_assignment;
+    typedef integral_constant
+            <
+                bool,
+                __get_poc_swap<outer_allocator_type, _InnerAllocs...>::value
+            > propagate_on_container_swap;
+    typedef integral_constant
+            <
+                bool,
+                __get_is_always_equal<outer_allocator_type, _InnerAllocs...>::value
+            > is_always_equal;
+
+    template <class _Tp>
+    struct rebind
+    {
+        typedef scoped_allocator_adaptor
+        <
+            typename _OuterTraits::template rebind_alloc<_Tp>, _InnerAllocs...
+        > other;
+    };
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    scoped_allocator_adaptor() _NOEXCEPT {}
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        scoped_allocator_adaptor(_OuterA2&& __outerAlloc,
+                                 const _InnerAllocs& ...__innerAllocs) _NOEXCEPT
+            : base(_CUDA_VSTD::forward<_OuterA2>(__outerAlloc), __innerAllocs...) {}
+    // scoped_allocator_adaptor(const scoped_allocator_adaptor& __other) = default;
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, const _OuterA2&>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        scoped_allocator_adaptor(
+            const scoped_allocator_adaptor<_OuterA2, _InnerAllocs...>& __other) _NOEXCEPT
+                : base(__other) {}
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        scoped_allocator_adaptor(
+            scoped_allocator_adaptor<_OuterA2, _InnerAllocs...>&& __other) _NOEXCEPT
+                : base(_CUDA_VSTD::move(__other)) {}
+
+    // scoped_allocator_adaptor& operator=(const scoped_allocator_adaptor&) = default;
+    // scoped_allocator_adaptor& operator=(scoped_allocator_adaptor&&) = default;
+    // ~scoped_allocator_adaptor() = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inner_allocator_type& inner_allocator() _NOEXCEPT
+        {return base::inner_allocator();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const inner_allocator_type& inner_allocator() const _NOEXCEPT
+        {return base::inner_allocator();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    outer_allocator_type& outer_allocator() _NOEXCEPT
+        {return base::outer_allocator();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const outer_allocator_type& outer_allocator() const _NOEXCEPT
+        {return base::outer_allocator();}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    pointer allocate(size_type __n)
+        {return allocator_traits<outer_allocator_type>::
+            allocate(outer_allocator(), __n);}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    pointer allocate(size_type __n, const_void_pointer __hint)
+        {return allocator_traits<outer_allocator_type>::
+            allocate(outer_allocator(), __n, __hint);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void deallocate(pointer __p, size_type __n) _NOEXCEPT
+        {allocator_traits<outer_allocator_type>::
+            deallocate(outer_allocator(), __p, __n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const
+        {return allocator_traits<outer_allocator_type>::max_size(outer_allocator());}
+
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void construct(_Tp* __p, _Args&& ...__args)
+            {__construct(__uses_alloc_ctor<_Tp, inner_allocator_type&, _Args...>(),
+                         __p, _CUDA_VSTD::forward<_Args>(__args)...);}
+
+    template <class _T1, class _T2, class... _Args1, class... _Args2>
+    void construct(pair<_T1, _T2>* __p, piecewise_construct_t,
+                       tuple<_Args1...> __x, tuple<_Args2...> __y)
+    {
+        typedef __outermost<outer_allocator_type> _OM;
+        allocator_traits<typename _OM::type>::construct(
+            _OM()(outer_allocator()), __p, piecewise_construct
+          , __transform_tuple(
+              typename __uses_alloc_ctor<
+                  _T1, inner_allocator_type&, _Args1...
+              >::type()
+            , _CUDA_VSTD::move(__x)
+            , typename __make_tuple_indices<sizeof...(_Args1)>::type{}
+          )
+          , __transform_tuple(
+              typename __uses_alloc_ctor<
+                  _T2, inner_allocator_type&, _Args2...
+              >::type()
+            , _CUDA_VSTD::move(__y)
+            , typename __make_tuple_indices<sizeof...(_Args2)>::type{}
+          )
+        );
+    }
+
+    template <class _T1, class _T2>
+    void construct(pair<_T1, _T2>* __p)
+    { construct(__p, piecewise_construct, tuple<>{}, tuple<>{}); }
+
+    template <class _T1, class _T2, class _Up, class _Vp>
+    void construct(pair<_T1, _T2>* __p, _Up&& __x, _Vp&& __y) {
+        construct(__p, piecewise_construct,
+                  _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Up>(__x)),
+                  _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Vp>(__y)));
+    }
+
+    template <class _T1, class _T2, class _Up, class _Vp>
+    void construct(pair<_T1, _T2>* __p, const pair<_Up, _Vp>& __x) {
+        construct(__p, piecewise_construct,
+                  _CUDA_VSTD::forward_as_tuple(__x.first),
+                  _CUDA_VSTD::forward_as_tuple(__x.second));
+    }
+
+    template <class _T1, class _T2, class _Up, class _Vp>
+    void construct(pair<_T1, _T2>* __p, pair<_Up, _Vp>&& __x) {
+        construct(__p, piecewise_construct,
+                  _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Up>(__x.first)),
+                  _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Vp>(__x.second)));
+    }
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void destroy(_Tp* __p)
+            {
+                typedef __outermost<outer_allocator_type> _OM;
+                allocator_traits<typename _OM::type>::
+                                         destroy(_OM()(outer_allocator()), __p);
+            }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    scoped_allocator_adaptor select_on_container_copy_construction() const _NOEXCEPT
+        {return base::select_on_container_copy_construction();}
+
+private:
+
+
+    template <class _OuterA2,
+              class = typename enable_if<
+                        is_constructible<outer_allocator_type, _OuterA2>::value
+                      >::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    scoped_allocator_adaptor(_OuterA2&& __o,
+                             const inner_allocator_type& __i) _NOEXCEPT
+        : base(_CUDA_VSTD::forward<_OuterA2>(__o), __i) {}
+
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void __construct(integral_constant<int, 0>, _Tp* __p, _Args&& ...__args)
+            {
+                typedef __outermost<outer_allocator_type> _OM;
+                allocator_traits<typename _OM::type>::construct
+                (
+                    _OM()(outer_allocator()),
+                    __p,
+                    _CUDA_VSTD::forward<_Args>(__args)...
+                );
+            }
+
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void __construct(integral_constant<int, 1>, _Tp* __p, _Args&& ...__args)
+            {
+                typedef __outermost<outer_allocator_type> _OM;
+                allocator_traits<typename _OM::type>::construct
+                (
+                    _OM()(outer_allocator()),
+                    __p, allocator_arg, inner_allocator(),
+                    _CUDA_VSTD::forward<_Args>(__args)...
+                );
+            }
+
+    template <class _Tp, class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void __construct(integral_constant<int, 2>, _Tp* __p, _Args&& ...__args)
+            {
+                typedef __outermost<outer_allocator_type> _OM;
+                allocator_traits<typename _OM::type>::construct
+                (
+                    _OM()(outer_allocator()),
+                    __p,
+                    _CUDA_VSTD::forward<_Args>(__args)...,
+                    inner_allocator()
+                );
+            }
+
+    template <class ..._Args, size_t ..._Idx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple<_Args&&...>
+    __transform_tuple(integral_constant<int, 0>, tuple<_Args...>&& __t,
+                      __tuple_indices<_Idx...>)
+    {
+        return _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::get<_Idx>(_CUDA_VSTD::move(__t))...);
+    }
+
+    template <class ..._Args, size_t ..._Idx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple<allocator_arg_t, inner_allocator_type&, _Args&&...>
+    __transform_tuple(integral_constant<int, 1>, tuple<_Args...> && __t,
+                      __tuple_indices<_Idx...>)
+    {
+        using _Tup = tuple<allocator_arg_t, inner_allocator_type&, _Args&&...>;
+        return _Tup(allocator_arg, inner_allocator(),
+                    _CUDA_VSTD::get<_Idx>(_CUDA_VSTD::move(__t))...);
+    }
+
+    template <class ..._Args, size_t ..._Idx>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple<_Args&&..., inner_allocator_type&>
+    __transform_tuple(integral_constant<int, 2>, tuple<_Args...> && __t,
+                      __tuple_indices<_Idx...>)
+    {
+        using _Tup = tuple<_Args&&..., inner_allocator_type&>;
+        return _Tup(_CUDA_VSTD::get<_Idx>(_CUDA_VSTD::move(__t))..., inner_allocator());
+    }
+
+    template <class...> friend class __scoped_allocator_storage;
+};
+
+template <class _OuterA1, class _OuterA2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const scoped_allocator_adaptor<_OuterA1>& __a,
+           const scoped_allocator_adaptor<_OuterA2>& __b) _NOEXCEPT
+{
+    return __a.outer_allocator() == __b.outer_allocator();
+}
+
+template <class _OuterA1, class _OuterA2, class _InnerA0, class... _InnerAllocs>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const scoped_allocator_adaptor<_OuterA1, _InnerA0, _InnerAllocs...>& __a,
+           const scoped_allocator_adaptor<_OuterA2, _InnerA0, _InnerAllocs...>& __b) _NOEXCEPT
+{
+    return __a.outer_allocator() == __b.outer_allocator() &&
+           __a.inner_allocator() == __b.inner_allocator();
+}
+
+template <class _OuterA1, class _OuterA2, class... _InnerAllocs>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const scoped_allocator_adaptor<_OuterA1, _InnerAllocs...>& __a,
+           const scoped_allocator_adaptor<_OuterA2, _InnerAllocs...>& __b) _NOEXCEPT
+{
+    return !(__a == __b);
+}
+
+#endif  // !defined(_LIBCUDACXX_CXX03_LANG)
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_SCOPED_ALLOCATOR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore
new file mode 100644
index 000000000000..bc51f5f4a5ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/semaphore
@@ -0,0 +1,447 @@
+// -*- C++ -*-
+//===--------------------------- semaphore --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SEMAPHORE
+#define _LIBCUDACXX_SEMAPHORE
+
+/*
+    semaphore synopsis
+
+namespace std {
+
+template<ptrdiff_t least_max_value = implementation-defined>
+class counting_semaphore
+{
+public:
+static constexpr ptrdiff_t max() noexcept;
+
+constexpr explicit counting_semaphore(ptrdiff_t desired);
+~counting_semaphore();
+
+counting_semaphore(const counting_semaphore&) = delete;
+counting_semaphore& operator=(const counting_semaphore&) = delete;
+
+void release(ptrdiff_t __update = 1);
+void acquire();
+bool try_acquire() noexcept;
+template<class Rep, class Period>
+    bool try_acquire_for(const chrono::duration<Rep, Period>& __rel_time);
+template<class Clock, class Duration>
+    bool try_acquire_until(const chrono::time_point<Clock, Duration>& __abs_time);
+
+private:
+ptrdiff_t counter; // exposition only
+};
+
+using binary_semaphore = counting_semaphore<1>;
+
+}
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__threading_support>
+#include <atomic>
+#include <cassert>
+#include <__pragma_push>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+# error <semaphore> is not supported on this single threaded system
+#endif
+
+#if _LIBCUDACXX_STD_VER < 11
+# error <semaphore> is requires C++11 or later
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template<int _Sco, ptrdiff_t __least_max_value>
+class __atomic_semaphore_base
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __fetch_sub_if_slow(ptrdiff_t __old)
+    {
+        while (__old != 0) {
+            if (__count.compare_exchange_weak(__old, __old - 1, memory_order_acquire, memory_order_relaxed))
+                return true;
+        }
+        return false;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __fetch_sub_if()
+    {
+        ptrdiff_t __old = __count.load(memory_order_acquire);
+        if (__old == 0)
+            return false;
+        if(__count.compare_exchange_weak(__old, __old - 1, memory_order_acquire, memory_order_relaxed))
+            return true;
+        return __fetch_sub_if_slow(__old); // fail only if not __available
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __wait_slow()
+    {
+        while (1) {
+            ptrdiff_t const __old = __count.load(memory_order_acquire);
+            if(__old != 0)
+                break;
+            __count.wait(__old, memory_order_relaxed);
+        }
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __acquire_slow_timed(chrono::nanoseconds const& __rel_time)
+    {
+        return __libcpp_thread_poll_with_backoff([this]() {
+            ptrdiff_t const __old = __count.load(memory_order_acquire);
+            return __old != 0 && __fetch_sub_if_slow(__old);
+        }, __rel_time);
+    }
+    __atomic_base<ptrdiff_t, _Sco> __count;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr ptrdiff_t max() noexcept
+    {
+        return numeric_limits<ptrdiff_t>::max();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __atomic_semaphore_base(ptrdiff_t __count) noexcept : __count(__count) { }
+
+    ~__atomic_semaphore_base() = default;
+
+    __atomic_semaphore_base(__atomic_semaphore_base const&) = delete;
+    __atomic_semaphore_base& operator=(__atomic_semaphore_base const&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void release(ptrdiff_t __update = 1)
+    {
+        __count.fetch_add(__update, memory_order_release);
+        if(__update > 1)
+            __count.notify_all();
+        else
+            __count.notify_one();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void acquire()
+    {
+        while (!try_acquire())
+            __wait_slow();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire() noexcept
+    {
+        return __fetch_sub_if();
+    }
+
+    template <class Clock, class Duration>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire_until(chrono::time_point<Clock, Duration> const& __abs_time)
+    {
+        if (try_acquire())
+            return true;
+        else
+            return __acquire_slow_timed(__abs_time - Clock::now());
+    }
+
+    template <class Rep, class Period>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire_for(chrono::duration<Rep, Period> const& __rel_time)
+    {
+
+        if (try_acquire())
+            return true;
+        else
+            return __acquire_slow_timed(__rel_time);
+    }
+};
+
+#ifndef _LIBCUDACXX_USE_NATIVE_SEMAPHORES
+
+template<int _Sco>
+class __atomic_semaphore_base<_Sco, 1> {
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __acquire_slow_timed(chrono::nanoseconds const& __rel_time)
+    {
+        return __libcpp_thread_poll_with_backoff([this]() {
+            return try_acquire();
+        }, __rel_time);
+    }
+    __atomic_base<int, _Sco> __available;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr ptrdiff_t max() noexcept { return 1; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    __atomic_semaphore_base(ptrdiff_t __available) : __available(__available) { }
+
+    ~__atomic_semaphore_base() = default;
+
+    __atomic_semaphore_base(__atomic_semaphore_base const&) = delete;
+    __atomic_semaphore_base& operator=(__atomic_semaphore_base const&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void release(ptrdiff_t __update = 1)
+    {
+        assert(__update == 1);
+        __available.store(1, memory_order_release);
+        __available.notify_one();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void acquire()
+    {
+        while (!try_acquire())
+            __available.wait(0, memory_order_relaxed);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire() noexcept
+    {
+        return 1 == __available.exchange(0, memory_order_acquire);
+    }
+
+    template <class Clock, class Duration>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire_until(chrono::time_point<Clock, Duration> const& __abs_time)
+    {
+        if (try_acquire())
+            return true;
+        else
+            return __acquire_slow_timed(__abs_time - Clock::now());
+    }
+
+    template <class Rep, class Period>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire_for(chrono::duration<Rep, Period> const& __rel_time)
+    {
+        if (try_acquire())
+            return true;
+        else
+            return __acquire_slow_timed(__rel_time);
+    }
+};
+
+#else
+
+template<int _Sco>
+class __sem_semaphore_base {
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __backfill(bool __success)
+    {
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_BACK_BUFFER
+        if(__success) {
+            auto const __back_amount = __backbuffer.fetch_sub(2, memory_order_acquire);
+            bool const __post_one = __back_amount > 0;
+            bool const __post_two = __back_amount > 1;
+            auto const __success = (!__post_one || __libcpp_semaphore_post(&__semaphore)) &&
+                                 (!__post_two || __libcpp_semaphore_post(&__semaphore));
+            assert(__success);
+            if(!__post_one || !__post_two)
+                __backbuffer.fetch_add(!__post_one ? 2 : 1, memory_order_relaxed);
+        }
+#endif
+        return __success;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __try_acquire_fast()
+    {
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+
+        ptrdiff_t __old;
+        __libcpp_thread_poll_with_backoff([&]() {
+            __old = __frontbuffer.load(memory_order_relaxed);
+            return 0 != (__old >> 32);
+        }, chrono::microseconds(5));
+
+        // always steal if you can
+        while(__old >> 32)
+            if(__frontbuffer.compare_exchange_weak(__old, __old - (1ll << 32), memory_order_acquire))
+                return true;
+        // record we're waiting
+        __old = __frontbuffer.fetch_add(1ll, memory_order_release);
+        // ALWAYS steal if you can!
+        while(__old >> 32)
+            if(__frontbuffer.compare_exchange_weak(__old, __old - (1ll << 32), memory_order_acquire))
+                break;
+        // not going to wait after all
+        if(__old >> 32)
+            return __try_done(true);
+#endif
+        // the wait has begun...
+        return false;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __try_done(bool __success)
+    {
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+        // record we're NOT waiting
+        __frontbuffer.fetch_sub(1ll, memory_order_release);
+#endif
+        return __backfill(__success);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __release_slow(ptrdiff_t __post_amount)
+    {
+    #ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_BACK_BUFFER
+        bool const __post_one = __post_amount > 0;
+        bool const __post_two = __post_amount > 1;
+        if(__post_amount > 2)
+            __backbuffer.fetch_add(__post_amount - 2, memory_order_acq_rel);
+        auto const __success = (!__post_one || __libcpp_semaphore_post(&__semaphore)) &&
+                             (!__post_two || __libcpp_semaphore_post(&__semaphore));
+        assert(__success);
+    #else
+        for(; __post_amount; --__post_amount) {
+            auto const __success = __libcpp_semaphore_post(&__semaphore);
+            assert(__success);
+        }
+    #endif
+    }
+
+    __libcpp_semaphore_t __semaphore;
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+    __atomic_base<ptrdiff_t, _Sco> __frontbuffer;
+#endif
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_BACK_BUFFER
+    __atomic_base<ptrdiff_t, _Sco> __backbuffer;
+#endif
+
+public:
+    static constexpr ptrdiff_t max() noexcept {
+        return _LIBCUDACXX_SEMAPHORE_MAX;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __sem_semaphore_base(ptrdiff_t __count = 0) : __semaphore()
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+    , __frontbuffer(__count << 32)
+#endif
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_BACK_BUFFER
+    , __backbuffer(0)
+#endif
+    {
+        assert(__count <= max());
+        auto const __success =
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+            __libcpp_semaphore_init(&__semaphore, 0);
+#else
+            __libcpp_semaphore_init(&__semaphore, __count);
+#endif
+        assert(__success);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~__sem_semaphore_base() {
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+        assert(0 == (__frontbuffer.load(memory_order_relaxed) & ~0u));
+#endif
+        auto const __success = __libcpp_semaphore_destroy(&__semaphore);
+        assert(__success);
+    }
+
+    __sem_semaphore_base(const __sem_semaphore_base&) = delete;
+    __sem_semaphore_base& operator=(const __sem_semaphore_base&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void release(ptrdiff_t __update = 1)
+    {
+#ifndef _LIBCUDACXX_HAS_NO_SEMAPHORE_FRONT_BUFFER
+        // boldly assume the semaphore is taken but uncontended
+        ptrdiff_t __old = 0;
+        // try to fast-release as long as it's uncontended
+        while(0 == (__old & ~0ul))
+            if(__frontbuffer.compare_exchange_weak(__old, __old + (__update << 32), memory_order_acq_rel))
+                return;
+#endif
+        // slow-release it is
+        __release_slow(__update);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void acquire()
+    {
+        if(!__try_acquire_fast())
+            __try_done(__libcpp_semaphore_wait(&__semaphore));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire() noexcept
+    {
+        return try_acquire_for(chrono::nanoseconds(0));
+    }
+
+    template <class Clock, class Duration>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire_until(chrono::time_point<Clock, Duration> const& __abs_time)
+    {
+        auto const current = max(Clock::now(), __abs_time);
+        return try_acquire_for(chrono::duration_cast<chrono::nanoseconds>(__abs_time - current));
+    }
+
+    template <class Rep, class Period>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool try_acquire_for(chrono::duration<Rep, Period> const& __rel_time)
+    {
+        return __try_acquire_fast() ||
+               __try_done(__libcpp_semaphore_wait_timed(&__semaphore, __rel_time));
+    }
+};
+
+#endif //_LIBCUDACXX_HAS_NO_SEMAPHORES
+
+template<ptrdiff_t __least_max_value, int _Sco>
+using __semaphore_base =
+#ifdef _LIBCUDACXX_USE_NATIVE_SEMAPHORES
+    typename conditional<__least_max_value <= __sem_semaphore_base<_Sco>::max(),
+                        __sem_semaphore_base<_Sco>,
+                        __atomic_semaphore_base<_Sco, __least_max_value>>::type
+#else
+    __atomic_semaphore_base<_Sco, __least_max_value>
+#endif
+    ;
+
+template<ptrdiff_t __least_max_value = INT_MAX>
+class counting_semaphore : public __semaphore_base<__least_max_value, 0>
+{
+    static_assert(__least_max_value <= __semaphore_base<__least_max_value, 0>::max(), "");
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    counting_semaphore(ptrdiff_t __count = 0) : __semaphore_base<__least_max_value, 0>(__count) { }
+    ~counting_semaphore() = default;
+
+    counting_semaphore(const counting_semaphore&) = delete;
+    counting_semaphore& operator=(const counting_semaphore&) = delete;
+};
+
+using binary_semaphore = counting_semaphore<1>;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif //_LIBCUDACXX_SEMAPHORE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/set b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/set
new file mode 100644
index 000000000000..d7192078da02
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/set
@@ -0,0 +1,1493 @@
+// -*- C++ -*-
+//===---------------------------- set -------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SET
+#define _LIBCUDACXX_SET
+
+/*
+
+    set synopsis
+
+namespace std
+{
+
+template <class Key, class Compare = less<Key>,
+          class Allocator = allocator<Key>>
+class set
+{
+public:
+    // types:
+    typedef Key                                      key_type;
+    typedef key_type                                 value_type;
+    typedef Compare                                  key_compare;
+    typedef key_compare                              value_compare;
+    typedef Allocator                                allocator_type;
+    typedef typename allocator_type::reference       reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+    typedef typename allocator_type::pointer         pointer;
+    typedef typename allocator_type::const_pointer   const_pointer;
+
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+    typedef unspecified                              node_type;               // C++17
+    typedef INSERT_RETURN_TYPE<iterator, node_type>  insert_return_type;      // C++17
+
+    // construct/copy/destroy:
+    set()
+        noexcept(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value);
+    explicit set(const value_compare& comp);
+    set(const value_compare& comp, const allocator_type& a);
+    template <class InputIterator>
+        set(InputIterator first, InputIterator last,
+            const value_compare& comp = value_compare());
+    template <class InputIterator>
+        set(InputIterator first, InputIterator last, const value_compare& comp,
+            const allocator_type& a);
+    set(const set& s);
+    set(set&& s)
+        noexcept(
+            is_nothrow_move_constructible<allocator_type>::value &&
+            is_nothrow_move_constructible<key_compare>::value);
+    explicit set(const allocator_type& a);
+    set(const set& s, const allocator_type& a);
+    set(set&& s, const allocator_type& a);
+    set(initializer_list<value_type> il, const value_compare& comp = value_compare());
+    set(initializer_list<value_type> il, const value_compare& comp,
+        const allocator_type& a);
+    template <class InputIterator>
+        set(InputIterator first, InputIterator last, const allocator_type& a)
+            : set(first, last, Compare(), a) {}  // C++14
+    set(initializer_list<value_type> il, const allocator_type& a)
+        : set(il, Compare(), a) {}  // C++14
+    ~set();
+
+    set& operator=(const set& s);
+    set& operator=(set&& s)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<key_compare>::value);
+    set& operator=(initializer_list<value_type> il);
+
+    // iterators:
+          iterator begin() noexcept;
+    const_iterator begin() const noexcept;
+          iterator end() noexcept;
+    const_iterator end()   const noexcept;
+
+          reverse_iterator rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+          reverse_iterator rend() noexcept;
+    const_reverse_iterator rend()   const noexcept;
+
+    const_iterator         cbegin()  const noexcept;
+    const_iterator         cend()    const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend()   const noexcept;
+
+    // capacity:
+    bool      empty()    const noexcept;
+    size_type size()     const noexcept;
+    size_type max_size() const noexcept;
+
+    // modifiers:
+    template <class... Args>
+        pair<iterator, bool> emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    pair<iterator,bool> insert(const value_type& v);
+    pair<iterator,bool> insert(value_type&& v);
+    iterator insert(const_iterator position, const value_type& v);
+    iterator insert(const_iterator position, value_type&& v);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type> il);
+
+    node_type extract(const_iterator position);                                       // C++17
+    node_type extract(const key_type& x);                                             // C++17
+    insert_return_type insert(node_type&& nh);                                        // C++17
+    iterator insert(const_iterator hint, node_type&& nh);                             // C++17
+
+    iterator  erase(const_iterator position);
+    iterator  erase(iterator position);  // C++14
+    size_type erase(const key_type& k);
+    iterator  erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class C2>
+      void merge(set<Key, C2, Allocator>& source);         // C++17
+    template<class C2>
+      void merge(set<Key, C2, Allocator>&& source);        // C++17
+    template<class C2>
+      void merge(multiset<Key, C2, Allocator>& source);    // C++17
+    template<class C2>
+      void merge(multiset<Key, C2, Allocator>&& source);   // C++17
+
+    void swap(set& s)
+        noexcept(
+            __is_nothrow_swappable<key_compare>::value &&
+            (!allocator_type::propagate_on_container_swap::value ||
+             __is_nothrow_swappable<allocator_type>::value));
+
+    // observers:
+    allocator_type get_allocator() const noexcept;
+    key_compare    key_comp()      const;
+    value_compare  value_comp()    const;
+
+    // set operations:
+          iterator find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    template<typename K>
+        iterator find(const K& x);
+    template<typename K>
+        const_iterator find(const K& x) const;  // C++14
+    template<typename K>
+        size_type count(const K& x) const;        // C++14
+    size_type      count(const key_type& k) const;
+        bool contains(const key_type& x) const; // C++20
+          iterator lower_bound(const key_type& k);
+    const_iterator lower_bound(const key_type& k) const;
+    template<typename K>
+        iterator lower_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator lower_bound(const K& x) const;  // C++14
+
+          iterator upper_bound(const key_type& k);
+    const_iterator upper_bound(const key_type& k) const;
+    template<typename K>
+        iterator upper_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator upper_bound(const K& x) const;  // C++14
+    pair<iterator,iterator>             equal_range(const key_type& k);
+    pair<const_iterator,const_iterator> equal_range(const key_type& k) const;
+    template<typename K>
+        pair<iterator,iterator>             equal_range(const K& x);        // C++14
+    template<typename K>
+        pair<const_iterator,const_iterator> equal_range(const K& x) const;  // C++14
+};
+
+template <class Key, class Compare, class Allocator>
+bool
+operator==(const set<Key, Compare, Allocator>& x,
+           const set<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator< (const set<Key, Compare, Allocator>& x,
+           const set<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator!=(const set<Key, Compare, Allocator>& x,
+           const set<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator> (const set<Key, Compare, Allocator>& x,
+           const set<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator>=(const set<Key, Compare, Allocator>& x,
+           const set<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator<=(const set<Key, Compare, Allocator>& x,
+           const set<Key, Compare, Allocator>& y);
+
+// specialized algorithms:
+template <class Key, class Compare, class Allocator>
+void
+swap(set<Key, Compare, Allocator>& x, set<Key, Compare, Allocator>& y)
+    noexcept(noexcept(x.swap(y)));
+
+template <class Key, class Compare, class Allocator, class Predicate>
+  void erase_if(set<Key, Compare, Allocator>& c, Predicate pred);  // C++20
+
+template <class Key, class Compare = less<Key>,
+          class Allocator = allocator<Key>>
+class multiset
+{
+public:
+    // types:
+    typedef Key                                      key_type;
+    typedef key_type                                 value_type;
+    typedef Compare                                  key_compare;
+    typedef key_compare                              value_compare;
+    typedef Allocator                                allocator_type;
+    typedef typename allocator_type::reference       reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+    typedef typename allocator_type::pointer         pointer;
+    typedef typename allocator_type::const_pointer   const_pointer;
+
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+    typedef unspecified                              node_type;               // C++17
+
+    // construct/copy/destroy:
+    multiset()
+        noexcept(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value);
+    explicit multiset(const value_compare& comp);
+    multiset(const value_compare& comp, const allocator_type& a);
+    template <class InputIterator>
+        multiset(InputIterator first, InputIterator last,
+                 const value_compare& comp = value_compare());
+    template <class InputIterator>
+        multiset(InputIterator first, InputIterator last,
+                 const value_compare& comp, const allocator_type& a);
+    multiset(const multiset& s);
+    multiset(multiset&& s)
+        noexcept(
+            is_nothrow_move_constructible<allocator_type>::value &&
+            is_nothrow_move_constructible<key_compare>::value);
+    explicit multiset(const allocator_type& a);
+    multiset(const multiset& s, const allocator_type& a);
+    multiset(multiset&& s, const allocator_type& a);
+    multiset(initializer_list<value_type> il, const value_compare& comp = value_compare());
+    multiset(initializer_list<value_type> il, const value_compare& comp,
+             const allocator_type& a);
+    template <class InputIterator>
+        multiset(InputIterator first, InputIterator last, const allocator_type& a)
+            : set(first, last, Compare(), a) {}  // C++14
+    multiset(initializer_list<value_type> il, const allocator_type& a)
+        : set(il, Compare(), a) {}  // C++14
+    ~multiset();
+
+    multiset& operator=(const multiset& s);
+    multiset& operator=(multiset&& s)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<key_compare>::value);
+    multiset& operator=(initializer_list<value_type> il);
+
+    // iterators:
+          iterator begin() noexcept;
+    const_iterator begin() const noexcept;
+          iterator end() noexcept;
+    const_iterator end()   const noexcept;
+
+          reverse_iterator rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+          reverse_iterator rend() noexcept;
+    const_reverse_iterator rend()   const noexcept;
+
+    const_iterator         cbegin()  const noexcept;
+    const_iterator         cend()    const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend()   const noexcept;
+
+    // capacity:
+    bool      empty()    const noexcept;
+    size_type size()     const noexcept;
+    size_type max_size() const noexcept;
+
+    // modifiers:
+    template <class... Args>
+        iterator emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    iterator insert(const value_type& v);
+    iterator insert(value_type&& v);
+    iterator insert(const_iterator position, const value_type& v);
+    iterator insert(const_iterator position, value_type&& v);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type> il);
+
+    node_type extract(const_iterator position);                                       // C++17
+    node_type extract(const key_type& x);                                             // C++17
+    iterator insert(node_type&& nh);                                                  // C++17
+    iterator insert(const_iterator hint, node_type&& nh);                             // C++17
+
+    iterator  erase(const_iterator position);
+    iterator  erase(iterator position);  // C++14
+    size_type erase(const key_type& k);
+    iterator  erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class C2>
+      void merge(multiset<Key, C2, Allocator>& source);    // C++17
+    template<class C2>
+      void merge(multiset<Key, C2, Allocator>&& source);   // C++17
+    template<class C2>
+      void merge(set<Key, C2, Allocator>& source);         // C++17
+    template<class C2>
+      void merge(set<Key, C2, Allocator>&& source);        // C++17
+
+    void swap(multiset& s)
+        noexcept(
+            __is_nothrow_swappable<key_compare>::value &&
+            (!allocator_type::propagate_on_container_swap::value ||
+             __is_nothrow_swappable<allocator_type>::value));
+
+    // observers:
+    allocator_type get_allocator() const noexcept;
+    key_compare    key_comp()      const;
+    value_compare  value_comp()    const;
+
+    // set operations:
+          iterator find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    template<typename K>
+        iterator find(const K& x);
+    template<typename K>
+        const_iterator find(const K& x) const;  // C++14
+    template<typename K>
+        size_type count(const K& x) const;      // C++14
+    size_type      count(const key_type& k) const;
+        bool contains(const key_type& x) const; // C++20
+          iterator lower_bound(const key_type& k);
+    const_iterator lower_bound(const key_type& k) const;
+    template<typename K>
+        iterator lower_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator lower_bound(const K& x) const;  // C++14
+
+          iterator upper_bound(const key_type& k);
+    const_iterator upper_bound(const key_type& k) const;
+    template<typename K>
+        iterator upper_bound(const K& x);              // C++14
+    template<typename K>
+        const_iterator upper_bound(const K& x) const;  // C++14
+
+    pair<iterator,iterator>             equal_range(const key_type& k);
+    pair<const_iterator,const_iterator> equal_range(const key_type& k) const;
+    template<typename K>
+        pair<iterator,iterator>             equal_range(const K& x);        // C++14
+    template<typename K>
+        pair<const_iterator,const_iterator> equal_range(const K& x) const;  // C++14
+};
+
+template <class Key, class Compare, class Allocator>
+bool
+operator==(const multiset<Key, Compare, Allocator>& x,
+           const multiset<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator< (const multiset<Key, Compare, Allocator>& x,
+           const multiset<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator!=(const multiset<Key, Compare, Allocator>& x,
+           const multiset<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator> (const multiset<Key, Compare, Allocator>& x,
+           const multiset<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator>=(const multiset<Key, Compare, Allocator>& x,
+           const multiset<Key, Compare, Allocator>& y);
+
+template <class Key, class Compare, class Allocator>
+bool
+operator<=(const multiset<Key, Compare, Allocator>& x,
+           const multiset<Key, Compare, Allocator>& y);
+
+// specialized algorithms:
+template <class Key, class Compare, class Allocator>
+void
+swap(multiset<Key, Compare, Allocator>& x, multiset<Key, Compare, Allocator>& y)
+    noexcept(noexcept(x.swap(y)));
+
+template <class Key, class Compare, class Allocator, class Predicate>
+  void erase_if(multiset<Key, Compare, Allocator>& c, Predicate pred);  // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__tree>
+#include <__node_handle>
+#include <functional>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Key, class _Compare, class _Allocator>
+class multiset;
+
+template <class _Key, class _Compare = less<_Key>,
+          class _Allocator = allocator<_Key> >
+class _LIBCUDACXX_TEMPLATE_VIS set
+{
+public:
+    // types:
+    typedef _Key                                     key_type;
+    typedef key_type                                 value_type;
+    typedef _Compare                                 key_compare;
+    typedef key_compare                              value_compare;
+    typedef typename __identity<_Allocator>::type    allocator_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+private:
+    typedef __tree<value_type, value_compare, allocator_type> __base;
+    typedef allocator_traits<allocator_type>                  __alloc_traits;
+    typedef typename __base::__node_holder                    __node_holder;
+
+    __base __tree_;
+
+public:
+    typedef typename __base::pointer               pointer;
+    typedef typename __base::const_pointer         const_pointer;
+    typedef typename __base::size_type             size_type;
+    typedef typename __base::difference_type       difference_type;
+    typedef typename __base::const_iterator        iterator;
+    typedef typename __base::const_iterator        const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>       reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator> const_reverse_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __set_node_handle<typename __base::__node, allocator_type> node_type;
+    typedef __insert_return_type<iterator, node_type> insert_return_type;
+#endif
+
+    template <class _Key2, class _Compare2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS set;
+    template <class _Key2, class _Compare2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS multiset;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set()
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(value_compare()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit set(const value_compare& __comp)
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(__comp) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit set(const value_compare& __comp, const allocator_type& __a)
+        : __tree_(__comp, __a) {}
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        set(_InputIterator __f, _InputIterator __l,
+            const value_compare& __comp = value_compare())
+        : __tree_(__comp)
+        {
+            insert(__f, __l);
+        }
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        set(_InputIterator __f, _InputIterator __l, const value_compare& __comp,
+            const allocator_type& __a)
+        : __tree_(__comp, __a)
+        {
+            insert(__f, __l);
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+        template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        set(_InputIterator __f, _InputIterator __l, const allocator_type& __a)
+            : set(__f, __l, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set(const set& __s)
+        : __tree_(__s.__tree_)
+        {
+            insert(__s.begin(), __s.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set& operator=(const set& __s)
+        {
+            __tree_ = __s.__tree_;
+            return *this;
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set(set&& __s)
+        _NOEXCEPT_(is_nothrow_move_constructible<__base>::value)
+        : __tree_(_CUDA_VSTD::move(__s.__tree_)) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit set(const allocator_type& __a)
+        : __tree_(__a) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set(const set& __s, const allocator_type& __a)
+        : __tree_(__s.__tree_.value_comp(), __a)
+        {
+            insert(__s.begin(), __s.end());
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    set(set&& __s, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set(initializer_list<value_type> __il, const value_compare& __comp = value_compare())
+        : __tree_(__comp)
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set(initializer_list<value_type> __il, const value_compare& __comp,
+        const allocator_type& __a)
+        : __tree_(__comp, __a)
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set(initializer_list<value_type> __il, const allocator_type& __a)
+        : set(__il, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set& operator=(initializer_list<value_type> __il)
+        {
+            __tree_.__assign_unique(__il.begin(), __il.end());
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    set& operator=(set&& __s)
+        _NOEXCEPT_(is_nothrow_move_assignable<__base>::value)
+        {
+            __tree_ = _CUDA_VSTD::move(__s.__tree_);
+            return *this;
+        }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~set() {
+        static_assert(sizeof(__diagnose_non_const_comparator<_Key, _Compare>()), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator begin() _NOEXCEPT       {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator end() _NOEXCEPT         {return __tree_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()   const _NOEXCEPT {return __tree_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rbegin() _NOEXCEPT
+            {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rend() _NOEXCEPT
+            {return reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend() const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin()  const _NOEXCEPT {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend() const _NOEXCEPT {return rend();}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return __tree_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT {return __tree_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __tree_.max_size();}
+
+    // modifiers:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> emplace(_Args&&... __args)
+            {return __tree_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...);}
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator emplace_hint(const_iterator __p, _Args&&... __args)
+            {return __tree_.__emplace_hint_unique(__p, _CUDA_VSTD::forward<_Args>(__args)...);}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator,bool> insert(const value_type& __v)
+        {return __tree_.__insert_unique(__v);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, const value_type& __v)
+        {return __tree_.__insert_unique(__p, __v);}
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __f, _InputIterator __l)
+        {
+            for (const_iterator __e = cend(); __f != __l; ++__f)
+                __tree_.__insert_unique(__e, *__f);
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator,bool> insert(value_type&& __v)
+        {return __tree_.__insert_unique(_CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, value_type&& __v)
+        {return __tree_.__insert_unique(__p, _CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator  erase(const_iterator __p) {return __tree_.erase(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k)
+        {return __tree_.__erase_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator  erase(const_iterator __f, const_iterator __l)
+        {return __tree_.erase(__f, __l);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__tree_.clear();}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    insert_return_type insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to set::insert()");
+        return __tree_.template __node_handle_insert_unique<
+            node_type, insert_return_type>(_CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to set::insert()");
+        return __tree_.template __node_handle_insert_unique<node_type>(
+            __hint, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__it);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(set<key_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(set<key_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multiset<key_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multiset<key_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_unique(__source.__tree_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(set& __s) _NOEXCEPT_(__is_nothrow_swappable<__base>::value)
+        {__tree_.swap(__s.__tree_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT {return __tree_.__alloc();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_compare    key_comp()      const {return __tree_.value_comp();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_compare  value_comp()    const {return __tree_.value_comp();}
+
+    // set operations:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator find(const key_type& __k)             {return __tree_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __tree_.find(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    find(const _K2& __k)                           {return __tree_.find(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    find(const _K2& __k) const                     {return __tree_.find(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type      count(const key_type& __k) const
+        {return __tree_.__count_unique(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,size_type>::type
+    count(const _K2& __k) const                    {return __tree_.__count_multi(__k);}
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool contains(const key_type& __k) const {return find(__k) != end();}
+#endif // _LIBCUDACXX_STD_VER > 17
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator lower_bound(const key_type& __k)
+        {return __tree_.lower_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator lower_bound(const key_type& __k) const
+        {return __tree_.lower_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    lower_bound(const _K2& __k)       {return __tree_.lower_bound(__k);}
+
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    lower_bound(const _K2& __k) const {return __tree_.lower_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator upper_bound(const key_type& __k)
+        {return __tree_.upper_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator upper_bound(const key_type& __k) const
+        {return __tree_.upper_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,iterator>::type
+    upper_bound(const _K2& __k)       {return __tree_.upper_bound(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    upper_bound(const _K2& __k) const {return __tree_.upper_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator,iterator> equal_range(const key_type& __k)
+        {return __tree_.__equal_range_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator,const_iterator> equal_range(const key_type& __k) const
+        {return __tree_.__equal_range_unique(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,pair<iterator,iterator>>::type
+    equal_range(const _K2& __k)       {return __tree_.__equal_range_multi(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,pair<const_iterator,const_iterator>>::type
+    equal_range(const _K2& __k) const {return __tree_.__equal_range_multi(__k);}
+#endif
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Compare = less<typename iterator_traits<_InputIterator>::value_type>,
+         class _Allocator = allocator<typename iterator_traits<_InputIterator>::value_type>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>>
+set(_InputIterator, _InputIterator, _Compare = _Compare(), _Allocator = _Allocator())
+  -> set<typename iterator_traits<_InputIterator>::value_type, _Compare, _Allocator>;
+
+template<class _Key, class _Compare = less<_Key>,
+         class _Allocator = allocator<_Key>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>>
+set(initializer_list<_Key>, _Compare = _Compare(), _Allocator = _Allocator())
+  -> set<_Key, _Compare, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+set(_InputIterator, _InputIterator, _Allocator)
+  -> set<typename iterator_traits<_InputIterator>::value_type,
+         less<typename iterator_traits<_InputIterator>::value_type>, _Allocator>;
+
+template<class _Key, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+set(initializer_list<_Key>, _Allocator)
+  -> set<_Key, less<_Key>, _Allocator>;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Compare, class _Allocator>
+set<_Key, _Compare, _Allocator>::set(set&& __s, const allocator_type& __a)
+    : __tree_(_CUDA_VSTD::move(__s.__tree_), __a)
+{
+    if (__a != __s.get_allocator())
+    {
+        const_iterator __e = cend();
+        while (!__s.empty())
+            insert(__e, _CUDA_VSTD::move(__s.__tree_.remove(__s.begin())->__value_));
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const set<_Key, _Compare, _Allocator>& __x,
+           const set<_Key, _Compare, _Allocator>& __y)
+{
+    return __x.size() == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const set<_Key, _Compare, _Allocator>& __x,
+           const set<_Key, _Compare, _Allocator>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const set<_Key, _Compare, _Allocator>& __x,
+           const set<_Key, _Compare, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const set<_Key, _Compare, _Allocator>& __x,
+           const set<_Key, _Compare, _Allocator>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const set<_Key, _Compare, _Allocator>& __x,
+           const set<_Key, _Compare, _Allocator>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const set<_Key, _Compare, _Allocator>& __x,
+           const set<_Key, _Compare, _Allocator>& __y)
+{
+    return !(__y < __x);
+}
+
+// specialized algorithms:
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(set<_Key, _Compare, _Allocator>& __x,
+     set<_Key, _Compare, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Key, class _Compare, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(set<_Key, _Compare, _Allocator>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+template <class _Key, class _Compare = less<_Key>,
+          class _Allocator = allocator<_Key> >
+class _LIBCUDACXX_TEMPLATE_VIS multiset
+{
+public:
+    // types:
+    typedef _Key                                      key_type;
+    typedef key_type                                 value_type;
+    typedef _Compare                                  key_compare;
+    typedef key_compare                              value_compare;
+    typedef typename __identity<_Allocator>::type    allocator_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+private:
+    typedef __tree<value_type, value_compare, allocator_type> __base;
+    typedef allocator_traits<allocator_type>                  __alloc_traits;
+    typedef typename __base::__node_holder                    __node_holder;
+
+    __base __tree_;
+
+public:
+    typedef typename __base::pointer               pointer;
+    typedef typename __base::const_pointer         const_pointer;
+    typedef typename __base::size_type             size_type;
+    typedef typename __base::difference_type       difference_type;
+    typedef typename __base::const_iterator        iterator;
+    typedef typename __base::const_iterator        const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>       reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator> const_reverse_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __set_node_handle<typename __base::__node, allocator_type> node_type;
+#endif
+
+    template <class _Key2, class _Compare2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS set;
+    template <class _Key2, class _Compare2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS multiset;
+
+    // construct/copy/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset()
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_default_constructible<key_compare>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(value_compare()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit multiset(const value_compare& __comp)
+        _NOEXCEPT_(
+            is_nothrow_default_constructible<allocator_type>::value &&
+            is_nothrow_copy_constructible<key_compare>::value)
+        : __tree_(__comp) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit multiset(const value_compare& __comp, const allocator_type& __a)
+        : __tree_(__comp, __a) {}
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        multiset(_InputIterator __f, _InputIterator __l,
+                 const value_compare& __comp = value_compare())
+        : __tree_(__comp)
+        {
+            insert(__f, __l);
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+        template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        multiset(_InputIterator __f, _InputIterator __l, const allocator_type& __a)
+            : multiset(__f, __l, key_compare(), __a) {}
+#endif
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        multiset(_InputIterator __f, _InputIterator __l,
+                 const value_compare& __comp, const allocator_type& __a)
+        : __tree_(__comp, __a)
+        {
+            insert(__f, __l);
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset(const multiset& __s)
+        : __tree_(__s.__tree_.value_comp(),
+          __alloc_traits::select_on_container_copy_construction(__s.__tree_.__alloc()))
+        {
+            insert(__s.begin(), __s.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset& operator=(const multiset& __s)
+        {
+            __tree_ = __s.__tree_;
+            return *this;
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset(multiset&& __s)
+        _NOEXCEPT_(is_nothrow_move_constructible<__base>::value)
+        : __tree_(_CUDA_VSTD::move(__s.__tree_)) {}
+
+    multiset(multiset&& __s, const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit multiset(const allocator_type& __a)
+        : __tree_(__a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset(const multiset& __s, const allocator_type& __a)
+        : __tree_(__s.__tree_.value_comp(), __a)
+        {
+            insert(__s.begin(), __s.end());
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset(initializer_list<value_type> __il, const value_compare& __comp = value_compare())
+        : __tree_(__comp)
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset(initializer_list<value_type> __il, const value_compare& __comp,
+        const allocator_type& __a)
+        : __tree_(__comp, __a)
+        {
+            insert(__il.begin(), __il.end());
+        }
+
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset(initializer_list<value_type> __il, const allocator_type& __a)
+        : multiset(__il, key_compare(), __a) {}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset& operator=(initializer_list<value_type> __il)
+        {
+            __tree_.__assign_multi(__il.begin(), __il.end());
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    multiset& operator=(multiset&& __s)
+        _NOEXCEPT_(is_nothrow_move_assignable<__base>::value)
+        {
+            __tree_ = _CUDA_VSTD::move(__s.__tree_);
+            return *this;
+        }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~multiset() {
+        static_assert(sizeof(__diagnose_non_const_comparator<_Key, _Compare>()), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator begin() _NOEXCEPT       {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT {return __tree_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          iterator end() _NOEXCEPT         {return __tree_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()   const _NOEXCEPT {return __tree_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rbegin() _NOEXCEPT
+            {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+          reverse_iterator rend() _NOEXCEPT
+            {return       reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend() const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin()  const _NOEXCEPT {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend() const _NOEXCEPT {return rend();}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return __tree_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT {return __tree_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __tree_.max_size();}
+
+    // modifiers:
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator emplace(_Args&&... __args)
+            {return __tree_.__emplace_multi(_CUDA_VSTD::forward<_Args>(__args)...);}
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator emplace_hint(const_iterator __p, _Args&&... __args)
+            {return __tree_.__emplace_hint_multi(__p, _CUDA_VSTD::forward<_Args>(__args)...);}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const value_type& __v)
+        {return __tree_.__insert_multi(__v);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, const value_type& __v)
+        {return __tree_.__insert_multi(__p, __v);}
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __f, _InputIterator __l)
+        {
+            for (const_iterator __e = cend(); __f != __l; ++__f)
+                __tree_.__insert_multi(__e, *__f);
+        }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(value_type&& __v)
+        {return __tree_.__insert_multi(_CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, value_type&& __v)
+        {return __tree_.__insert_multi(__p, _CUDA_VSTD::move(__v));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator  erase(const_iterator __p) {return __tree_.erase(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __tree_.__erase_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator  erase(const_iterator __f, const_iterator __l)
+        {return __tree_.erase(__f, __l);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__tree_.clear();}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to multiset::insert()");
+        return __tree_.template __node_handle_insert_multi<node_type>(
+            _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to multiset::insert()");
+        return __tree_.template __node_handle_insert_multi<node_type>(
+            __hint, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __tree_.template __node_handle_extract<node_type>(__it);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multiset<key_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(multiset<key_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(set<key_type, _Compare2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+    template <class _Compare2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(set<key_type, _Compare2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __tree_.__node_handle_merge_multi(__source.__tree_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(multiset& __s)
+        _NOEXCEPT_(__is_nothrow_swappable<__base>::value)
+        {__tree_.swap(__s.__tree_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT {return __tree_.__alloc();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_compare    key_comp()      const {return __tree_.value_comp();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_compare  value_comp()    const {return __tree_.value_comp();}
+
+    // set operations:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator find(const key_type& __k)             {return __tree_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __tree_.find(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,iterator>::type
+    find(const _K2& __k)                           {return __tree_.find(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    find(const _K2& __k) const                     {return __tree_.find(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type      count(const key_type& __k) const
+        {return __tree_.__count_multi(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename enable_if<__is_transparent<_Compare, _K2>::value,size_type>::type
+    count(const _K2& __k) const            {return __tree_.__count_multi(__k);}
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool contains(const key_type& __k) const {return find(__k) != end();}
+#endif // _LIBCUDACXX_STD_VER > 17
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator lower_bound(const key_type& __k)
+        {return __tree_.lower_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator lower_bound(const key_type& __k) const
+            {return __tree_.lower_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,iterator>::type
+    lower_bound(const _K2& __k)       {return __tree_.lower_bound(__k);}
+
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    lower_bound(const _K2& __k) const {return __tree_.lower_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator upper_bound(const key_type& __k)
+            {return __tree_.upper_bound(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator upper_bound(const key_type& __k) const
+            {return __tree_.upper_bound(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,iterator>::type
+    upper_bound(const _K2& __k)       {return __tree_.upper_bound(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,const_iterator>::type
+    upper_bound(const _K2& __k) const {return __tree_.upper_bound(__k);}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator,iterator>             equal_range(const key_type& __k)
+            {return __tree_.__equal_range_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator,const_iterator> equal_range(const key_type& __k) const
+            {return __tree_.__equal_range_multi(__k);}
+#if _LIBCUDACXX_STD_VER > 11
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,pair<iterator,iterator>>::type
+    equal_range(const _K2& __k)       {return __tree_.__equal_range_multi(__k);}
+    template <typename _K2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    typename _CUDA_VSTD::enable_if<_CUDA_VSTD::__is_transparent<_Compare, _K2>::value,pair<const_iterator,const_iterator>>::type
+    equal_range(const _K2& __k) const {return __tree_.__equal_range_multi(__k);}
+#endif
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Compare = less<typename iterator_traits<_InputIterator>::value_type>,
+         class _Allocator = allocator<typename iterator_traits<_InputIterator>::value_type>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>>
+multiset(_InputIterator, _InputIterator, _Compare = _Compare(), _Allocator = _Allocator())
+  -> multiset<typename iterator_traits<_InputIterator>::value_type, _Compare, _Allocator>;
+
+template<class _Key, class _Compare = less<_Key>,
+         class _Allocator = allocator<_Key>,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>,
+         class = _EnableIf<!__is_allocator<_Compare>::value, void>>
+multiset(initializer_list<_Key>, _Compare = _Compare(), _Allocator = _Allocator())
+  -> multiset<_Key, _Compare, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+multiset(_InputIterator, _InputIterator, _Allocator)
+  -> multiset<typename iterator_traits<_InputIterator>::value_type,
+         less<typename iterator_traits<_InputIterator>::value_type>, _Allocator>;
+
+template<class _Key, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value, void>>
+multiset(initializer_list<_Key>, _Allocator)
+  -> multiset<_Key, less<_Key>, _Allocator>;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Compare, class _Allocator>
+multiset<_Key, _Compare, _Allocator>::multiset(multiset&& __s, const allocator_type& __a)
+    : __tree_(_CUDA_VSTD::move(__s.__tree_), __a)
+{
+    if (__a != __s.get_allocator())
+    {
+        const_iterator __e = cend();
+        while (!__s.empty())
+            insert(__e, _CUDA_VSTD::move(__s.__tree_.remove(__s.begin())->__value_));
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const multiset<_Key, _Compare, _Allocator>& __x,
+           const multiset<_Key, _Compare, _Allocator>& __y)
+{
+    return __x.size() == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const multiset<_Key, _Compare, _Allocator>& __x,
+           const multiset<_Key, _Compare, _Allocator>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const multiset<_Key, _Compare, _Allocator>& __x,
+           const multiset<_Key, _Compare, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const multiset<_Key, _Compare, _Allocator>& __x,
+           const multiset<_Key, _Compare, _Allocator>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const multiset<_Key, _Compare, _Allocator>& __x,
+           const multiset<_Key, _Compare, _Allocator>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const multiset<_Key, _Compare, _Allocator>& __x,
+           const multiset<_Key, _Compare, _Allocator>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Key, class _Compare, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(multiset<_Key, _Compare, _Allocator>& __x,
+     multiset<_Key, _Compare, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Key, class _Compare, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(multiset<_Key, _Compare, _Allocator>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_SET
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h
new file mode 100644
index 000000000000..d609c1668413
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/setjmp.h
@@ -0,0 +1,44 @@
+// -*- C++ -*-
+//===--------------------------- setjmp.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SETJMP_H
+#define _LIBCUDACXX_SETJMP_H
+
+/*
+    setjmp.h synopsis
+
+Macros:
+
+    setjmp
+
+Types:
+
+    jmp_buf
+
+void longjmp(jmp_buf env, int val);
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <setjmp.h>
+
+#ifdef __cplusplus
+
+#ifndef setjmp
+#define setjmp(env) setjmp(env)
+#endif
+
+#endif // __cplusplus
+
+#endif  // _LIBCUDACXX_SETJMP_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex
new file mode 100644
index 000000000000..1b01302f26ab
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/shared_mutex
@@ -0,0 +1,508 @@
+// -*- C++ -*-
+//===------------------------ shared_mutex --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SHARED_MUTEX
+#define _LIBCUDACXX_SHARED_MUTEX
+
+/*
+    shared_mutex synopsis
+
+// C++1y
+
+namespace std
+{
+
+class shared_mutex      // C++17
+{
+public:
+    shared_mutex();
+    ~shared_mutex();
+
+    shared_mutex(const shared_mutex&) = delete;
+    shared_mutex& operator=(const shared_mutex&) = delete;
+
+    // Exclusive ownership
+    void lock(); // blocking
+    bool try_lock();
+    void unlock();
+
+    // Shared ownership
+    void lock_shared(); // blocking
+    bool try_lock_shared();
+    void unlock_shared();
+
+    typedef implementation-defined native_handle_type; // See 30.2.3
+    native_handle_type native_handle(); // See 30.2.3
+};
+
+class shared_timed_mutex
+{
+public:
+    shared_timed_mutex();
+    ~shared_timed_mutex();
+
+    shared_timed_mutex(const shared_timed_mutex&) = delete;
+    shared_timed_mutex& operator=(const shared_timed_mutex&) = delete;
+
+    // Exclusive ownership
+    void lock(); // blocking
+    bool try_lock();
+    template <class Rep, class Period>
+        bool try_lock_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool try_lock_until(const chrono::time_point<Clock, Duration>& abs_time);
+    void unlock();
+
+    // Shared ownership
+    void lock_shared(); // blocking
+    bool try_lock_shared();
+    template <class Rep, class Period>
+        bool
+        try_lock_shared_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool
+        try_lock_shared_until(const chrono::time_point<Clock, Duration>& abs_time);
+    void unlock_shared();
+};
+
+template <class Mutex>
+class shared_lock
+{
+public:
+    typedef Mutex mutex_type;
+
+    // Shared locking
+    shared_lock() noexcept;
+    explicit shared_lock(mutex_type& m); // blocking
+    shared_lock(mutex_type& m, defer_lock_t) noexcept;
+    shared_lock(mutex_type& m, try_to_lock_t);
+    shared_lock(mutex_type& m, adopt_lock_t);
+    template <class Clock, class Duration>
+        shared_lock(mutex_type& m,
+                    const chrono::time_point<Clock, Duration>& abs_time);
+    template <class Rep, class Period>
+        shared_lock(mutex_type& m,
+                    const chrono::duration<Rep, Period>& rel_time);
+    ~shared_lock();
+
+    shared_lock(shared_lock const&) = delete;
+    shared_lock& operator=(shared_lock const&) = delete;
+
+    shared_lock(shared_lock&& u) noexcept;
+    shared_lock& operator=(shared_lock&& u) noexcept;
+
+    void lock(); // blocking
+    bool try_lock();
+    template <class Rep, class Period>
+        bool try_lock_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool try_lock_until(const chrono::time_point<Clock, Duration>& abs_time);
+    void unlock();
+
+    // Setters
+    void swap(shared_lock& u) noexcept;
+    mutex_type* release() noexcept;
+
+    // Getters
+    bool owns_lock() const noexcept;
+    explicit operator bool () const noexcept;
+    mutex_type* mutex() const noexcept;
+};
+
+template <class Mutex>
+    void swap(shared_lock<Mutex>& x, shared_lock<Mutex>& y) noexcept;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <version>
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+#if _LIBCUDACXX_STD_VER > 11 || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+
+#include <__mutex_base>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+#error <shared_mutex> is not supported on this single threaded system
+#else // !_LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+struct _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_SHARED_MUTEX _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(capability("shared_mutex"))
+__shared_mutex_base
+{
+    mutex               __mut_;
+    condition_variable  __gate1_;
+    condition_variable  __gate2_;
+    unsigned            __state_;
+
+    static const unsigned __write_entered_ = 1U << (sizeof(unsigned)*__CHAR_BIT__ - 1);
+    static const unsigned __n_readers_ = ~__write_entered_;
+
+    __shared_mutex_base();
+    _LIBCUDACXX_INLINE_VISIBILITY ~__shared_mutex_base() = default;
+
+    __shared_mutex_base(const __shared_mutex_base&) = delete;
+    __shared_mutex_base& operator=(const __shared_mutex_base&) = delete;
+
+    // Exclusive ownership
+    void lock() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(acquire_capability()); // blocking
+    bool try_lock() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(try_acquire_capability(true));
+    void unlock() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(release_capability());
+
+    // Shared ownership
+    void lock_shared() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(acquire_shared_capability()); // blocking
+    bool try_lock_shared() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(try_acquire_shared_capability(true));
+    void unlock_shared() _LIBCUDACXX_THREAD_SAFETY_ANNOTATION(release_shared_capability());
+
+//     typedef implementation-defined native_handle_type; // See 30.2.3
+//     native_handle_type native_handle(); // See 30.2.3
+};
+
+
+#if _LIBCUDACXX_STD_VER > 14
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_SHARED_MUTEX shared_mutex
+{
+    __shared_mutex_base __base;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY shared_mutex() : __base() {}
+    _LIBCUDACXX_INLINE_VISIBILITY ~shared_mutex() = default;
+
+    shared_mutex(const shared_mutex&) = delete;
+    shared_mutex& operator=(const shared_mutex&) = delete;
+
+    // Exclusive ownership
+    _LIBCUDACXX_INLINE_VISIBILITY void lock()     { return __base.lock(); }
+    _LIBCUDACXX_INLINE_VISIBILITY bool try_lock() { return __base.try_lock(); }
+    _LIBCUDACXX_INLINE_VISIBILITY void unlock()   { return __base.unlock(); }
+
+    // Shared ownership
+    _LIBCUDACXX_INLINE_VISIBILITY void lock_shared()     { return __base.lock_shared(); }
+    _LIBCUDACXX_INLINE_VISIBILITY bool try_lock_shared() { return __base.try_lock_shared(); }
+    _LIBCUDACXX_INLINE_VISIBILITY void unlock_shared()   { return __base.unlock_shared(); }
+
+//     typedef __shared_mutex_base::native_handle_type native_handle_type;
+//     _LIBCUDACXX_INLINE_VISIBILITY native_handle_type native_handle() { return __base::unlock_shared(); }
+};
+#endif
+
+
+class _LIBCUDACXX_TYPE_VIS _LIBCUDACXX_AVAILABILITY_SHARED_MUTEX shared_timed_mutex
+{
+    __shared_mutex_base __base;
+public:
+    shared_timed_mutex();
+    _LIBCUDACXX_INLINE_VISIBILITY ~shared_timed_mutex() = default;
+
+    shared_timed_mutex(const shared_timed_mutex&) = delete;
+    shared_timed_mutex& operator=(const shared_timed_mutex&) = delete;
+
+    // Exclusive ownership
+    void lock();
+    bool try_lock();
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool
+        try_lock_for(const chrono::duration<_Rep, _Period>& __rel_time)
+        {
+            return try_lock_until(chrono::steady_clock::now() + __rel_time);
+        }
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        bool
+        try_lock_until(const chrono::time_point<_Clock, _Duration>& __abs_time);
+    void unlock();
+
+    // Shared ownership
+    void lock_shared();
+    bool try_lock_shared();
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool
+        try_lock_shared_for(const chrono::duration<_Rep, _Period>& __rel_time)
+        {
+            return try_lock_shared_until(chrono::steady_clock::now() + __rel_time);
+        }
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        bool
+        try_lock_shared_until(const chrono::time_point<_Clock, _Duration>& __abs_time);
+    void unlock_shared();
+};
+
+template <class _Clock, class _Duration>
+bool
+shared_timed_mutex::try_lock_until(
+                        const chrono::time_point<_Clock, _Duration>& __abs_time)
+{
+    unique_lock<mutex> __lk(__base.__mut_);
+    if (__base.__state_ & __base.__write_entered_)
+    {
+        while (true)
+        {
+            cv_status __status = __base.__gate1_.wait_until(__lk, __abs_time);
+            if ((__base.__state_ & __base.__write_entered_) == 0)
+                break;
+            if (__status == cv_status::timeout)
+                return false;
+        }
+    }
+    __base.__state_ |= __base.__write_entered_;
+    if (__base.__state_ & __base.__n_readers_)
+    {
+        while (true)
+        {
+            cv_status __status = __base.__gate2_.wait_until(__lk, __abs_time);
+            if ((__base.__state_ & __base.__n_readers_) == 0)
+                break;
+            if (__status == cv_status::timeout)
+            {
+                __base.__state_ &= ~__base.__write_entered_;
+                __base.__gate1_.notify_all();
+                return false;
+            }
+        }
+    }
+    return true;
+}
+
+template <class _Clock, class _Duration>
+bool
+shared_timed_mutex::try_lock_shared_until(
+                        const chrono::time_point<_Clock, _Duration>& __abs_time)
+{
+    unique_lock<mutex> __lk(__base.__mut_);
+    if ((__base.__state_ & __base.__write_entered_) || (__base.__state_ & __base.__n_readers_) == __base.__n_readers_)
+    {
+        while (true)
+        {
+            cv_status status = __base.__gate1_.wait_until(__lk, __abs_time);
+            if ((__base.__state_ & __base.__write_entered_) == 0 &&
+                                       (__base.__state_ & __base.__n_readers_) < __base.__n_readers_)
+                break;
+            if (status == cv_status::timeout)
+                return false;
+        }
+    }
+    unsigned __num_readers = (__base.__state_ & __base.__n_readers_) + 1;
+    __base.__state_ &= ~__base.__n_readers_;
+    __base.__state_ |= __num_readers;
+    return true;
+}
+
+template <class _Mutex>
+class shared_lock
+{
+public:
+    typedef _Mutex mutex_type;
+
+private:
+    mutex_type* __m_;
+    bool __owns_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_lock() _NOEXCEPT
+        : __m_(nullptr),
+          __owns_(false)
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit shared_lock(mutex_type& __m)
+        : __m_(_CUDA_VSTD::addressof(__m)),
+          __owns_(true)
+        {__m_->lock_shared();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_lock(mutex_type& __m, defer_lock_t) _NOEXCEPT
+        : __m_(_CUDA_VSTD::addressof(__m)),
+          __owns_(false)
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_lock(mutex_type& __m, try_to_lock_t)
+        : __m_(_CUDA_VSTD::addressof(__m)),
+          __owns_(__m.try_lock_shared())
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_lock(mutex_type& __m, adopt_lock_t)
+        : __m_(_CUDA_VSTD::addressof(__m)),
+          __owns_(true)
+        {}
+
+    template <class _Clock, class _Duration>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        shared_lock(mutex_type& __m,
+                    const chrono::time_point<_Clock, _Duration>& __abs_time)
+            : __m_(_CUDA_VSTD::addressof(__m)),
+              __owns_(__m.try_lock_shared_until(__abs_time))
+            {}
+
+    template <class _Rep, class _Period>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        shared_lock(mutex_type& __m,
+                    const chrono::duration<_Rep, _Period>& __rel_time)
+            : __m_(_CUDA_VSTD::addressof(__m)),
+              __owns_(__m.try_lock_shared_for(__rel_time))
+            {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~shared_lock()
+    {
+        if (__owns_)
+            __m_->unlock_shared();
+    }
+
+    shared_lock(shared_lock const&) = delete;
+    shared_lock& operator=(shared_lock const&) = delete;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_lock(shared_lock&& __u) _NOEXCEPT
+        : __m_(__u.__m_),
+          __owns_(__u.__owns_)
+        {
+            __u.__m_ = nullptr;
+            __u.__owns_ = false;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    shared_lock& operator=(shared_lock&& __u) _NOEXCEPT
+    {
+        if (__owns_)
+            __m_->unlock_shared();
+        __m_ = nullptr;
+        __owns_ = false;
+        __m_ = __u.__m_;
+        __owns_ = __u.__owns_;
+        __u.__m_ = nullptr;
+        __u.__owns_ = false;
+        return *this;
+    }
+
+    void lock();
+    bool try_lock();
+    template <class Rep, class Period>
+        bool try_lock_for(const chrono::duration<Rep, Period>& rel_time);
+    template <class Clock, class Duration>
+        bool try_lock_until(const chrono::time_point<Clock, Duration>& abs_time);
+    void unlock();
+
+    // Setters
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(shared_lock& __u) _NOEXCEPT
+    {
+        _CUDA_VSTD::swap(__m_, __u.__m_);
+        _CUDA_VSTD::swap(__owns_, __u.__owns_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mutex_type* release() _NOEXCEPT
+    {
+        mutex_type* __m = __m_;
+        __m_ = nullptr;
+        __owns_ = false;
+        return __m;
+    }
+
+    // Getters
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool owns_lock() const _NOEXCEPT {return __owns_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit operator bool () const _NOEXCEPT {return __owns_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mutex_type* mutex() const _NOEXCEPT {return __m_;}
+};
+
+template <class _Mutex>
+void
+shared_lock<_Mutex>::lock()
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "shared_lock::lock: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "shared_lock::lock: already locked");
+    __m_->lock_shared();
+    __owns_ = true;
+}
+
+template <class _Mutex>
+bool
+shared_lock<_Mutex>::try_lock()
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "shared_lock::try_lock: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "shared_lock::try_lock: already locked");
+    __owns_ = __m_->try_lock_shared();
+    return __owns_;
+}
+
+template <class _Mutex>
+template <class _Rep, class _Period>
+bool
+shared_lock<_Mutex>::try_lock_for(const chrono::duration<_Rep, _Period>& __d)
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "shared_lock::try_lock_for: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "shared_lock::try_lock_for: already locked");
+    __owns_ = __m_->try_lock_shared_for(__d);
+    return __owns_;
+}
+
+template <class _Mutex>
+template <class _Clock, class _Duration>
+bool
+shared_lock<_Mutex>::try_lock_until(const chrono::time_point<_Clock, _Duration>& __t)
+{
+    if (__m_ == nullptr)
+        __throw_system_error(EPERM, "shared_lock::try_lock_until: references null mutex");
+    if (__owns_)
+        __throw_system_error(EDEADLK, "shared_lock::try_lock_until: already locked");
+    __owns_ = __m_->try_lock_shared_until(__t);
+    return __owns_;
+}
+
+template <class _Mutex>
+void
+shared_lock<_Mutex>::unlock()
+{
+    if (!__owns_)
+        __throw_system_error(EPERM, "shared_lock::unlock: not locked");
+    __m_->unlock_shared();
+    __owns_ = false;
+}
+
+template <class _Mutex>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(shared_lock<_Mutex>& __x, shared_lock<_Mutex>& __y) _NOEXCEPT
+    {__x.swap(__y);}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // !_LIBCUDACXX_HAS_NO_THREADS
+
+#endif  // _LIBCUDACXX_STD_VER > 11
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_SHARED_MUTEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/span b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/span
new file mode 100644
index 000000000000..c1d2f9357f16
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/span
@@ -0,0 +1,591 @@
+// -*- C++ -*-
+//===------------------------------ span ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===---------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SPAN
+#define _LIBCUDACXX_SPAN
+
+/*
+    span synopsis
+
+namespace std {
+
+// constants
+inline constexpr size_t dynamic_extent = numeric_limits<size_t>::max();
+
+// [views.span], class template span
+template <class ElementType, size_t Extent = dynamic_extent>
+    class span;
+
+// [span.objectrep], views of object representation
+template <class ElementType, size_t Extent>
+    span<const byte, ((Extent == dynamic_extent) ? dynamic_extent :
+        (sizeof(ElementType) * Extent))> as_bytes(span<ElementType, Extent> s) noexcept;
+
+template <class ElementType, size_t Extent>
+    span<      byte, ((Extent == dynamic_extent) ? dynamic_extent :
+        (sizeof(ElementType) * Extent))> as_writable_bytes(span<ElementType, Extent> s) noexcept;
+
+
+namespace std {
+template <class ElementType, size_t Extent = dynamic_extent>
+class span {
+public:
+    // constants and types
+    using element_type = ElementType;
+    using value_type = remove_cv_t<ElementType>;
+    using index_type = size_t;
+    using difference_type = ptrdiff_t;
+    using pointer = element_type*;
+    using const_pointer = const element_type*;
+    using reference = element_type&;
+    using const_reference = const element_type&;
+    using iterator = implementation-defined;
+    using const_iterator = implementation-defined;
+    using reverse_iterator = std::reverse_iterator<iterator>;
+    using const_reverse_iterator = std::reverse_iterator<const_iterator>;
+    static constexpr index_type extent = Extent;
+
+    // [span.cons], span constructors, copy, assignment, and destructor
+    constexpr span() noexcept;
+    constexpr span(pointer ptr, index_type count);
+    constexpr span(pointer firstElem, pointer lastElem);
+    template <size_t N>
+        constexpr span(element_type (&arr)[N]) noexcept;
+    template <size_t N>
+        constexpr span(array<value_type, N>& arr) noexcept;
+    template <size_t N>
+        constexpr span(const array<value_type, N>& arr) noexcept;
+    template <class Container>
+        constexpr span(Container& cont);
+    template <class Container>
+        constexpr span(const Container& cont);
+    constexpr span(const span& other) noexcept = default;
+    template <class OtherElementType, size_t OtherExtent>
+        constexpr span(const span<OtherElementType, OtherExtent>& s) noexcept;
+    ~span() noexcept = default;
+    constexpr span& operator=(const span& other) noexcept = default;
+
+    // [span.sub], span subviews
+    template <size_t Count>
+        constexpr span<element_type, Count> first() const;
+    template <size_t Count>
+        constexpr span<element_type, Count> last() const;
+    template <size_t Offset, size_t Count = dynamic_extent>
+        constexpr span<element_type, see below> subspan() const;
+
+    constexpr span<element_type, dynamic_extent> first(index_type count) const;
+    constexpr span<element_type, dynamic_extent> last(index_type count) const;
+    constexpr span<element_type, dynamic_extent> subspan(index_type offset, index_type count = dynamic_extent) const;
+
+    // [span.obs], span observers
+    constexpr index_type size() const noexcept;
+    constexpr index_type size_bytes() const noexcept;
+    constexpr bool empty() const noexcept;
+
+    // [span.elem], span element access
+    constexpr reference operator[](index_type idx) const;
+    constexpr reference front() const;
+    constexpr reference back() const;
+    constexpr pointer data() const noexcept;
+
+    // [span.iterators], span iterator support
+    constexpr iterator begin() const noexcept;
+    constexpr iterator end() const noexcept;
+    constexpr const_iterator cbegin() const noexcept;
+    constexpr const_iterator cend() const noexcept;
+    constexpr reverse_iterator rbegin() const noexcept;
+    constexpr reverse_iterator rend() const noexcept;
+    constexpr const_reverse_iterator crbegin() const noexcept;
+    constexpr const_reverse_iterator crend() const noexcept;
+
+private:
+    pointer data_;     // exposition only
+    index_type size_;  // exposition only
+};
+
+template<class T, size_t N>
+    span(T (&)[N]) -> span<T, N>;
+
+template<class T, size_t N>
+    span(array<T, N>&) -> span<T, N>;
+
+template<class T, size_t N>
+    span(const array<T, N>&) -> span<const T, N>;
+
+template<class Container>
+    span(Container&) -> span<typename Container::value_type>;
+
+template<class Container>
+    span(const Container&) -> span<const typename Container::value_type>;
+
+} // namespace std
+
+*/
+
+#include <__config>
+#include <cstddef>      // for ptrdiff_t
+#include <iterator>     // for iterators
+#include <array>        // for array
+#include <type_traits>  // for remove_cv, etc
+#include <cstddef>      // for byte
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 17
+
+inline constexpr size_t dynamic_extent = numeric_limits<size_t>::max();
+template <typename _Tp, size_t _Extent = dynamic_extent> class span;
+
+
+template <class _Tp>
+struct __is_span_impl : public false_type {};
+
+template <class _Tp, size_t _Extent>
+struct __is_span_impl<span<_Tp, _Extent>> : public true_type {};
+
+template <class _Tp>
+struct __is_span : public __is_span_impl<remove_cv_t<_Tp>> {};
+
+template <class _Tp>
+struct __is_std_array_impl : public false_type {};
+
+template <class _Tp, size_t _Sz>
+struct __is_std_array_impl<array<_Tp, _Sz>> : public true_type {};
+
+template <class _Tp>
+struct __is_std_array : public __is_std_array_impl<remove_cv_t<_Tp>> {};
+
+template <class _Tp, class _ElementType, class = void>
+struct __is_span_compatible_container : public false_type {};
+
+template <class _Tp, class _ElementType>
+struct __is_span_compatible_container<_Tp, _ElementType,
+        void_t<
+        // is not a specialization of span
+            typename enable_if<!__is_span<_Tp>::value, nullptr_t>::type,
+        // is not a specialization of array
+            typename enable_if<!__is_std_array<_Tp>::value, nullptr_t>::type,
+        // is_array_v<Container> is false,
+            typename enable_if<!is_array_v<_Tp>, nullptr_t>::type,
+        // data(cont) and size(cont) are well formed
+            decltype(data(declval<_Tp>())),
+            decltype(size(declval<_Tp>())),
+        // remove_pointer_t<decltype(data(cont))>(*)[] is convertible to ElementType(*)[]
+            typename enable_if<
+                is_convertible_v<remove_pointer_t<decltype(data(declval<_Tp &>()))>(*)[],
+                                 _ElementType(*)[]>,
+                nullptr_t>::type
+        >>
+    : public true_type {};
+
+
+template <typename _Tp, size_t _Extent>
+class _LIBCUDACXX_TEMPLATE_VIS span {
+public:
+//  constants and types
+    using element_type           = _Tp;
+    using value_type             = remove_cv_t<_Tp>;
+    using index_type             = size_t;
+    using difference_type        = ptrdiff_t;
+    using pointer                = _Tp *;
+    using const_pointer          = const _Tp *;
+    using reference              = _Tp &;
+    using const_reference        = const _Tp &;
+    using iterator               =  __wrap_iter<pointer>;
+    using const_iterator         =  __wrap_iter<const_pointer>;
+    using reverse_iterator       = _CUDA_VSTD::reverse_iterator<iterator>;
+    using const_reverse_iterator = _CUDA_VSTD::reverse_iterator<const_iterator>;
+
+    static constexpr index_type extent = _Extent;
+
+// [span.cons], span constructors, copy, assignment, and destructor
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span() noexcept : __data{nullptr}
+    { static_assert(_Extent == 0, "Can't default construct a statically sized span with size > 0"); }
+
+    constexpr span           (const span&) noexcept = default;
+    constexpr span& operator=(const span&) noexcept = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(pointer __ptr, index_type __count) : __data{__ptr}
+        { (void)__count; _LIBCUDACXX_ASSERT(_Extent == __count, "size mismatch in span's constructor (ptr, len)"); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(pointer __f, pointer __l) : __data{__f}
+        { (void)__l;     _LIBCUDACXX_ASSERT(_Extent == distance(__f, __l), "size mismatch in span's constructor (ptr, ptr)"); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(element_type (&__arr)[_Extent])          noexcept : __data{__arr} {}
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(      array<value_type, _Extent>& __arr) noexcept : __data{__arr.data()} {}
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(const array<value_type, _Extent>& __arr) noexcept : __data{__arr.data()} {}
+
+    template <class _OtherElementType>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        constexpr span(const span<_OtherElementType, _Extent>& __other,
+                       enable_if_t<
+                          is_convertible_v<_OtherElementType(*)[], element_type (*)[]>,
+                          nullptr_t> = nullptr)
+        : __data{__other.data()} {}
+
+    template <class _OtherElementType>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        constexpr span(const span<_OtherElementType, dynamic_extent>& __other,
+                       enable_if_t<
+                          is_convertible_v<_OtherElementType(*)[], element_type (*)[]>,
+                          nullptr_t> = nullptr) noexcept
+        : __data{__other.data()} { _LIBCUDACXX_ASSERT(_Extent == __other.size(), "size mismatch in span's constructor (other span)"); }
+
+
+//  ~span() noexcept = default;
+
+    template <size_t _Count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, _Count> first() const noexcept
+    {
+        static_assert(_Count <= _Extent, "Count out of range in span::first()");
+        return {data(), _Count};
+    }
+
+    template <size_t _Count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, _Count> last() const noexcept
+    {
+        static_assert(_Count <= _Extent, "Count out of range in span::last()");
+        return {data() + size() - _Count, _Count};
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, dynamic_extent> first(index_type __count) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__count <= size(), "Count out of range in span::first(count)");
+        return {data(), __count};
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, dynamic_extent> last(index_type __count) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__count <= size(), "Count out of range in span::last(count)");
+        return {data() + size() - __count, __count};
+    }
+
+    template <size_t _Offset, size_t _Count = dynamic_extent>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr auto subspan() const noexcept
+        -> span<element_type, _Count != dynamic_extent ? _Count : _Extent - _Offset>
+    {
+        static_assert(_Offset <= _Extent, "Offset out of range in span::subspan()");
+        return {data() + _Offset, _Count == dynamic_extent ? size() - _Offset : _Count};
+    }
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, dynamic_extent>
+       subspan(index_type __offset, index_type __count = dynamic_extent) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__offset <= size(), "Offset out of range in span::subspan(offset, count)");
+        _LIBCUDACXX_ASSERT(__count  <= size() || __count == dynamic_extent, "Count out of range in span::subspan(offset, count)");
+        if (__count == dynamic_extent)
+            return {data() + __offset, size() - __offset};
+        _LIBCUDACXX_ASSERT(__offset <= size() - __count, "count + offset out of range in span::subspan(offset, count)");
+        return {data() + __offset, __count};
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr index_type size()       const noexcept { return _Extent; }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr index_type size_bytes() const noexcept { return _Extent * sizeof(element_type); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr bool empty()            const noexcept { return _Extent == 0; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reference operator[](index_type __idx) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__idx >= 0 && __idx < size(), "span<T,N>[] index out of bounds");
+        return __data[__idx];
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reference front() const noexcept
+    {
+        static_assert(_Extent > 0, "span<T,N>[].front() on empty span");
+        return __data[0];
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reference back() const noexcept
+    {
+        static_assert(_Extent > 0, "span<T,N>[].back() on empty span");
+        return __data[size()-1];
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr pointer data()                         const noexcept { return __data; }
+
+// [span.iter], span iterator support
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr iterator                 begin() const noexcept { return iterator(data()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr iterator                   end() const noexcept { return iterator(data() + size()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_iterator          cbegin() const noexcept { return const_iterator(data()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_iterator            cend() const noexcept { return const_iterator(data() + size()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reverse_iterator        rbegin() const noexcept { return reverse_iterator(end()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reverse_iterator          rend() const noexcept { return reverse_iterator(begin()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_reverse_iterator crbegin() const noexcept { return const_reverse_iterator(cend()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_reverse_iterator   crend() const noexcept { return const_reverse_iterator(cbegin()); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr void swap(span &__other) noexcept
+    {
+        pointer __p = __data;
+        __data = __other.__data;
+        __other.__data = __p;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY span<const byte, _Extent * sizeof(element_type)> __as_bytes() const noexcept
+    { return {reinterpret_cast<const byte *>(data()), size_bytes()}; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY span<byte, _Extent * sizeof(element_type)> __as_writable_bytes() const noexcept
+    { return {reinterpret_cast<byte *>(data()), size_bytes()}; }
+
+private:
+    pointer    __data;
+
+};
+
+
+template <typename _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS span<_Tp, dynamic_extent> {
+private:
+
+public:
+//  constants and types
+    using element_type           = _Tp;
+    using value_type             = remove_cv_t<_Tp>;
+    using index_type             = size_t;
+    using difference_type        = ptrdiff_t;
+    using pointer                = _Tp *;
+    using const_pointer          = const _Tp *;
+    using reference              = _Tp &;
+    using const_reference        = const _Tp &;
+    using iterator               =  __wrap_iter<pointer>;
+    using const_iterator         =  __wrap_iter<const_pointer>;
+    using reverse_iterator       = _CUDA_VSTD::reverse_iterator<iterator>;
+    using const_reverse_iterator = _CUDA_VSTD::reverse_iterator<const_iterator>;
+
+    static constexpr index_type extent = dynamic_extent;
+
+// [span.cons], span constructors, copy, assignment, and destructor
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span() noexcept : __data{nullptr}, __size{0} {}
+
+    constexpr span           (const span&) noexcept = default;
+    constexpr span& operator=(const span&) noexcept = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(pointer __ptr, index_type __count) : __data{__ptr}, __size{__count} {}
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr span(pointer __f, pointer __l) : __data{__f}, __size{static_cast<size_t>(distance(__f, __l))} {}
+
+    template <size_t _Sz>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span(element_type (&__arr)[_Sz])          noexcept : __data{__arr}, __size{_Sz} {}
+
+    template <size_t _Sz>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span(array<value_type, _Sz>& __arr)       noexcept : __data{__arr.data()}, __size{_Sz} {}
+
+    template <size_t _Sz>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span(const array<value_type, _Sz>& __arr) noexcept : __data{__arr.data()}, __size{_Sz} {}
+
+    template <class _Container>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        constexpr span(      _Container& __c,
+            enable_if_t<__is_span_compatible_container<_Container, _Tp>::value, nullptr_t> = nullptr)
+        : __data{_CUDA_VSTD::data(__c)}, __size{(index_type) _CUDA_VSTD::size(__c)} {}
+
+    template <class _Container>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        constexpr span(const _Container& __c,
+            enable_if_t<__is_span_compatible_container<const _Container, _Tp>::value, nullptr_t> = nullptr)
+        : __data{_CUDA_VSTD::data(__c)}, __size{(index_type) _CUDA_VSTD::size(__c)} {}
+
+
+    template <class _OtherElementType, size_t _OtherExtent>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        constexpr span(const span<_OtherElementType, _OtherExtent>& __other,
+                       enable_if_t<
+                          is_convertible_v<_OtherElementType(*)[], element_type (*)[]>,
+                          nullptr_t> = nullptr) noexcept
+        : __data{__other.data()}, __size{__other.size()} {}
+
+//    ~span() noexcept = default;
+
+    template <size_t _Count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, _Count> first() const noexcept
+    {
+        _LIBCUDACXX_ASSERT(_Count <= size(), "Count out of range in span::first()");
+        return {data(), _Count};
+    }
+
+    template <size_t _Count>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, _Count> last() const noexcept
+    {
+        _LIBCUDACXX_ASSERT(_Count <= size(), "Count out of range in span::last()");
+        return {data() + size() - _Count, _Count};
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, dynamic_extent> first(index_type __count) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__count <= size(), "Count out of range in span::first(count)");
+        return {data(), __count};
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<element_type, dynamic_extent> last (index_type __count) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__count <= size(), "Count out of range in span::last(count)");
+        return {data() + size() - __count, __count};
+    }
+
+    template <size_t _Offset, size_t _Count = dynamic_extent>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr span<_Tp, dynamic_extent> subspan() const noexcept
+    {
+        _LIBCUDACXX_ASSERT(_Offset <= size(), "Offset out of range in span::subspan()");
+        _LIBCUDACXX_ASSERT(_Count == dynamic_extent || _Offset + _Count <= size(), "Count out of range in span::subspan()");
+        return {data() + _Offset, _Count == dynamic_extent ? size() - _Offset : _Count};
+    }
+
+    constexpr span<element_type, dynamic_extent>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    subspan(index_type __offset, index_type __count = dynamic_extent) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__offset <= size(), "Offset out of range in span::subspan(offset, count)");
+        _LIBCUDACXX_ASSERT(__count  <= size() || __count == dynamic_extent, "count out of range in span::subspan(offset, count)");
+        if (__count == dynamic_extent)
+            return {data() + __offset, size() - __offset};
+        _LIBCUDACXX_ASSERT(__offset <= size() - __count, "Offset + count out of range in span::subspan(offset, count)");
+        return {data() + __offset, __count};
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr index_type size()       const noexcept { return __size; }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr index_type size_bytes() const noexcept { return __size * sizeof(element_type); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr bool empty()            const noexcept { return __size == 0; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reference operator[](index_type __idx) const noexcept
+    {
+        _LIBCUDACXX_ASSERT(__idx >= 0 && __idx < size(), "span<T>[] index out of bounds");
+        return __data[__idx];
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reference front() const noexcept
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "span<T>[].front() on empty span");
+        return __data[0];
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reference back() const noexcept
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "span<T>[].back() on empty span");
+        return __data[size()-1];
+    }
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr pointer data()                         const noexcept { return __data; }
+
+// [span.iter], span iterator support
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr iterator                 begin() const noexcept { return iterator(data()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr iterator                   end() const noexcept { return iterator(data() + size()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_iterator          cbegin() const noexcept { return const_iterator(data()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_iterator            cend() const noexcept { return const_iterator(data() + size()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reverse_iterator        rbegin() const noexcept { return reverse_iterator(end()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr reverse_iterator          rend() const noexcept { return reverse_iterator(begin()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_reverse_iterator crbegin() const noexcept { return const_reverse_iterator(cend()); }
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr const_reverse_iterator   crend() const noexcept { return const_reverse_iterator(cbegin()); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY constexpr void swap(span &__other) noexcept
+    {
+        pointer __p = __data;
+        __data = __other.__data;
+        __other.__data = __p;
+
+        index_type __sz = __size;
+        __size = __other.__size;
+        __other.__size = __sz;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY span<const byte, dynamic_extent> __as_bytes() const noexcept
+    { return {reinterpret_cast<const byte *>(data()), size_bytes()}; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY span<byte, dynamic_extent> __as_writable_bytes() const noexcept
+    { return {reinterpret_cast<byte *>(data()), size_bytes()}; }
+
+private:
+    pointer    __data;
+    index_type __size;
+};
+
+//  tuple interface
+template <class _Tp, size_t _Size>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<span<_Tp, _Size>>
+    : public integral_constant<size_t, _Size> {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<span<_Tp, dynamic_extent>>; // declared but not defined
+
+
+template <size_t _Ip, class _Tp, size_t _Size>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, span<_Tp, _Size>>
+{
+    static_assert( dynamic_extent != _Size, "std::tuple_element<> not supported for std::span<T, dynamic_extent>");
+    static_assert(_Ip < _Size, "Index out of bounds in std::tuple_element<> (std::span)");
+    typedef _Tp type;
+};
+
+template <size_t _Ip, class _Tp, size_t _Size>
+_LIBCUDACXX_INLINE_VISIBILITY constexpr
+_Tp&
+get(span<_Tp, _Size> __s) noexcept
+{
+    static_assert( dynamic_extent != _Size, "std::get<> not supported for std::span<T, dynamic_extent>");
+    static_assert(_Ip < _Size, "Index out of bounds in std::get<> (std::span)");
+    return __s[_Ip];
+}
+
+
+//  as_bytes & as_writable_bytes
+template <class _Tp, size_t _Extent>
+_LIBCUDACXX_INLINE_VISIBILITY
+auto as_bytes(span<_Tp, _Extent> __s) noexcept
+-> decltype(__s.__as_bytes())
+{ return    __s.__as_bytes(); }
+
+template <class _Tp, size_t _Extent>
+_LIBCUDACXX_INLINE_VISIBILITY
+auto as_writable_bytes(span<_Tp, _Extent> __s) noexcept
+-> enable_if_t<!is_const_v<_Tp>, decltype(__s.__as_writable_bytes())>
+{ return __s.__as_writable_bytes(); }
+
+template <class _Tp, size_t _Extent>
+_LIBCUDACXX_INLINE_VISIBILITY
+constexpr void swap(span<_Tp, _Extent> &__lhs, span<_Tp, _Extent> &__rhs) noexcept
+{ __lhs.swap(__rhs); }
+
+
+//  Deduction guides
+template<class _Tp, size_t _Sz>
+    span(_Tp (&)[_Sz]) -> span<_Tp, _Sz>;
+
+template<class _Tp, size_t _Sz>
+    span(array<_Tp, _Sz>&) -> span<_Tp, _Sz>;
+
+template<class _Tp, size_t _Sz>
+    span(const array<_Tp, _Sz>&) -> span<const _Tp, _Sz>;
+
+template<class _Container>
+    span(_Container&) -> span<typename _Container::value_type>;
+
+template<class _Container>
+    span(const _Container&) -> span<const typename _Container::value_type>;
+
+#endif // _LIBCUDACXX_STD_VER > 17
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // _LIBCUDACXX_SPAN
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/sstream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/sstream
new file mode 100644
index 000000000000..c58fbc323115
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/sstream
@@ -0,0 +1,985 @@
+// -*- C++ -*-
+//===--------------------------- sstream ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SSTREAM
+#define _LIBCUDACXX_SSTREAM
+
+/*
+    sstream synopsis
+
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+class basic_stringbuf
+    : public basic_streambuf<charT, traits>
+{
+public:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef Allocator                      allocator_type;
+
+    // 27.8.1.1 Constructors:
+    explicit basic_stringbuf(ios_base::openmode which = ios_base::in | ios_base::out);
+    explicit basic_stringbuf(const basic_string<char_type, traits_type, allocator_type>& str,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    basic_stringbuf(basic_stringbuf&& rhs);
+
+    // 27.8.1.2 Assign and swap:
+    basic_stringbuf& operator=(basic_stringbuf&& rhs);
+    void swap(basic_stringbuf& rhs);
+
+    // 27.8.1.3 Get and set:
+    basic_string<char_type, traits_type, allocator_type> str() const;
+    void str(const basic_string<char_type, traits_type, allocator_type>& s);
+
+protected:
+    // 27.8.1.4 Overridden virtual functions:
+    virtual int_type underflow();
+    virtual int_type pbackfail(int_type c = traits_type::eof());
+    virtual int_type overflow (int_type c = traits_type::eof());
+    virtual basic_streambuf<char_type, traits_type>* setbuf(char_type*, streamsize);
+    virtual pos_type seekoff(off_type off, ios_base::seekdir way,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type sp,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+};
+
+template <class charT, class traits, class Allocator>
+  void swap(basic_stringbuf<charT, traits, Allocator>& x,
+            basic_stringbuf<charT, traits, Allocator>& y);
+
+typedef basic_stringbuf<char>    stringbuf;
+typedef basic_stringbuf<wchar_t> wstringbuf;
+
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+class basic_istringstream
+    : public basic_istream<charT, traits>
+{
+public:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef Allocator                      allocator_type;
+
+    // 27.8.2.1 Constructors:
+    explicit basic_istringstream(ios_base::openmode which = ios_base::in);
+    explicit basic_istringstream(const basic_string<char_type, traits_type,allocator_type>& str,
+                                 ios_base::openmode which = ios_base::in);
+    basic_istringstream(basic_istringstream&& rhs);
+
+    // 27.8.2.2 Assign and swap:
+    basic_istringstream& operator=(basic_istringstream&& rhs);
+    void swap(basic_istringstream& rhs);
+
+    // 27.8.2.3 Members:
+    basic_stringbuf<char_type, traits_type, allocator_type>* rdbuf() const;
+    basic_string<char_type, traits_type, allocator_type> str() const;
+    void str(const basic_string<char_type, traits_type, allocator_type>& s);
+};
+
+template <class charT, class traits, class Allocator>
+  void swap(basic_istringstream<charT, traits, Allocator>& x,
+            basic_istringstream<charT, traits, Allocator>& y);
+
+typedef basic_istringstream<char>    istringstream;
+typedef basic_istringstream<wchar_t> wistringstream;
+
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+class basic_ostringstream
+    : public basic_ostream<charT, traits>
+{
+public:
+    // types:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef Allocator                      allocator_type;
+
+    // 27.8.3.1 Constructors/destructor:
+    explicit basic_ostringstream(ios_base::openmode which = ios_base::out);
+    explicit basic_ostringstream(const basic_string<char_type, traits_type, allocator_type>& str,
+                                 ios_base::openmode which = ios_base::out);
+    basic_ostringstream(basic_ostringstream&& rhs);
+
+    // 27.8.3.2 Assign/swap:
+    basic_ostringstream& operator=(basic_ostringstream&& rhs);
+    void swap(basic_ostringstream& rhs);
+
+    // 27.8.3.3 Members:
+    basic_stringbuf<char_type, traits_type, allocator_type>* rdbuf() const;
+    basic_string<char_type, traits_type, allocator_type> str() const;
+    void str(const basic_string<char_type, traits_type, allocator_type>& s);
+};
+
+template <class charT, class traits, class Allocator>
+  void swap(basic_ostringstream<charT, traits, Allocator>& x,
+            basic_ostringstream<charT, traits, Allocator>& y);
+
+typedef basic_ostringstream<char>    ostringstream;
+typedef basic_ostringstream<wchar_t> wostringstream;
+
+template <class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+class basic_stringstream
+    : public basic_iostream<charT, traits>
+{
+public:
+    // types:
+    typedef charT                          char_type;
+    typedef traits                         traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef Allocator                      allocator_type;
+
+    // constructors/destructor
+    explicit basic_stringstream(ios_base::openmode which = ios_base::out|ios_base::in);
+    explicit basic_stringstream(const basic_string<char_type, traits_type, allocator_type>& str,
+                                ios_base::openmode which = ios_base::out|ios_base::in);
+    basic_stringstream(basic_stringstream&& rhs);
+
+    // 27.8.5.1 Assign/swap:
+    basic_stringstream& operator=(basic_stringstream&& rhs);
+    void swap(basic_stringstream& rhs);
+
+    // Members:
+    basic_stringbuf<char_type, traits_type, allocator_type>* rdbuf() const;
+    basic_string<char_type, traits_type, allocator_type> str() const;
+    void str(const basic_string<char_type, traits_type, allocator_type>& str);
+};
+
+template <class charT, class traits, class Allocator>
+  void swap(basic_stringstream<charT, traits, Allocator>& x,
+            basic_stringstream<charT, traits, Allocator>& y);
+
+typedef basic_stringstream<char>    stringstream;
+typedef basic_stringstream<wchar_t> wstringstream;
+
+}  // std
+
+*/
+
+#include <__config>
+#include <ostream>
+#include <istream>
+#include <string>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// basic_stringbuf
+
+template <class _CharT, class _Traits, class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS basic_stringbuf
+    : public basic_streambuf<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef _Allocator                     allocator_type;
+
+    typedef basic_string<char_type, traits_type, allocator_type> string_type;
+
+private:
+
+    string_type __str_;
+    mutable char_type* __hm_;
+    ios_base::openmode __mode_;
+
+public:
+    // 27.8.1.1 Constructors:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_stringbuf(ios_base::openmode __wch = ios_base::in | ios_base::out);
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_stringbuf(const string_type& __s,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    basic_stringbuf(basic_stringbuf&& __rhs);
+
+    // 27.8.1.2 Assign and swap:
+    basic_stringbuf& operator=(basic_stringbuf&& __rhs);
+#endif
+    void swap(basic_stringbuf& __rhs);
+
+    // 27.8.1.3 Get and set:
+    string_type str() const;
+    void str(const string_type& __s);
+
+protected:
+    // 27.8.1.4 Overridden virtual functions:
+    virtual int_type underflow();
+    virtual int_type pbackfail(int_type __c = traits_type::eof());
+    virtual int_type overflow (int_type __c = traits_type::eof());
+    virtual pos_type seekoff(off_type __off, ios_base::seekdir __way,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    virtual pos_type seekpos(pos_type __sp,
+                             ios_base::openmode __wch = ios_base::in | ios_base::out);
+};
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>::basic_stringbuf(ios_base::openmode __wch)
+    : __hm_(0),
+      __mode_(__wch)
+{
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>::basic_stringbuf(const string_type& __s,
+                             ios_base::openmode __wch)
+    : __str_(__s.get_allocator()),
+      __hm_(0),
+      __mode_(__wch)
+{
+    str(__s);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>::basic_stringbuf(basic_stringbuf&& __rhs)
+    : __mode_(__rhs.__mode_)
+{
+    char_type* __p = const_cast<char_type*>(__rhs.__str_.data());
+    ptrdiff_t __binp = -1;
+    ptrdiff_t __ninp = -1;
+    ptrdiff_t __einp = -1;
+    if (__rhs.eback() != nullptr)
+    {
+        __binp = __rhs.eback() - __p;
+        __ninp = __rhs.gptr() - __p;
+        __einp = __rhs.egptr() - __p;
+    }
+    ptrdiff_t __bout = -1;
+    ptrdiff_t __nout = -1;
+    ptrdiff_t __eout = -1;
+    if (__rhs.pbase() != nullptr)
+    {
+        __bout = __rhs.pbase() - __p;
+        __nout = __rhs.pptr() - __p;
+        __eout = __rhs.epptr() - __p;
+    }
+    ptrdiff_t __hm = __rhs.__hm_ == nullptr ? -1 : __rhs.__hm_ - __p;
+    __str_ = _CUDA_VSTD::move(__rhs.__str_);
+    __p = const_cast<char_type*>(__str_.data());
+    if (__binp != -1)
+        this->setg(__p + __binp, __p + __ninp, __p + __einp);
+    if (__bout != -1)
+    {
+        this->setp(__p + __bout, __p + __eout);
+        this->__pbump(__nout);
+    }
+    __hm_ = __hm == -1 ? nullptr : __p + __hm;
+    __p = const_cast<char_type*>(__rhs.__str_.data());
+    __rhs.setg(__p, __p, __p);
+    __rhs.setp(__p, __p);
+    __rhs.__hm_ = __p;
+    this->pubimbue(__rhs.getloc());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>&
+basic_stringbuf<_CharT, _Traits, _Allocator>::operator=(basic_stringbuf&& __rhs)
+{
+    char_type* __p = const_cast<char_type*>(__rhs.__str_.data());
+    ptrdiff_t __binp = -1;
+    ptrdiff_t __ninp = -1;
+    ptrdiff_t __einp = -1;
+    if (__rhs.eback() != nullptr)
+    {
+        __binp = __rhs.eback() - __p;
+        __ninp = __rhs.gptr() - __p;
+        __einp = __rhs.egptr() - __p;
+    }
+    ptrdiff_t __bout = -1;
+    ptrdiff_t __nout = -1;
+    ptrdiff_t __eout = -1;
+    if (__rhs.pbase() != nullptr)
+    {
+        __bout = __rhs.pbase() - __p;
+        __nout = __rhs.pptr() - __p;
+        __eout = __rhs.epptr() - __p;
+    }
+    ptrdiff_t __hm = __rhs.__hm_ == nullptr ? -1 : __rhs.__hm_ - __p;
+    __str_ = _CUDA_VSTD::move(__rhs.__str_);
+    __p = const_cast<char_type*>(__str_.data());
+    if (__binp != -1)
+        this->setg(__p + __binp, __p + __ninp, __p + __einp);
+    else
+        this->setg(nullptr, nullptr, nullptr);
+    if (__bout != -1)
+    {
+        this->setp(__p + __bout, __p + __eout);
+        this->__pbump(__nout);
+    }
+    else
+        this->setp(nullptr, nullptr);
+
+    __hm_ = __hm == -1 ? nullptr : __p + __hm;
+    __mode_ = __rhs.__mode_;
+    __p = const_cast<char_type*>(__rhs.__str_.data());
+    __rhs.setg(__p, __p, __p);
+    __rhs.setp(__p, __p);
+    __rhs.__hm_ = __p;
+    this->pubimbue(__rhs.getloc());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_stringbuf<_CharT, _Traits, _Allocator>::swap(basic_stringbuf& __rhs)
+{
+    char_type* __p = const_cast<char_type*>(__rhs.__str_.data());
+    ptrdiff_t __rbinp = -1;
+    ptrdiff_t __rninp = -1;
+    ptrdiff_t __reinp = -1;
+    if (__rhs.eback() != nullptr)
+    {
+        __rbinp = __rhs.eback() - __p;
+        __rninp = __rhs.gptr() - __p;
+        __reinp = __rhs.egptr() - __p;
+    }
+    ptrdiff_t __rbout = -1;
+    ptrdiff_t __rnout = -1;
+    ptrdiff_t __reout = -1;
+    if (__rhs.pbase() != nullptr)
+    {
+        __rbout = __rhs.pbase() - __p;
+        __rnout = __rhs.pptr() - __p;
+        __reout = __rhs.epptr() - __p;
+    }
+    ptrdiff_t __rhm = __rhs.__hm_ == nullptr ? -1 : __rhs.__hm_ - __p;
+    __p = const_cast<char_type*>(__str_.data());
+    ptrdiff_t __lbinp = -1;
+    ptrdiff_t __lninp = -1;
+    ptrdiff_t __leinp = -1;
+    if (this->eback() != nullptr)
+    {
+        __lbinp = this->eback() - __p;
+        __lninp = this->gptr() - __p;
+        __leinp = this->egptr() - __p;
+    }
+    ptrdiff_t __lbout = -1;
+    ptrdiff_t __lnout = -1;
+    ptrdiff_t __leout = -1;
+    if (this->pbase() != nullptr)
+    {
+        __lbout = this->pbase() - __p;
+        __lnout = this->pptr() - __p;
+        __leout = this->epptr() - __p;
+    }
+    ptrdiff_t __lhm = __hm_ == nullptr ? -1 : __hm_ - __p;
+    _CUDA_VSTD::swap(__mode_, __rhs.__mode_);
+    __str_.swap(__rhs.__str_);
+    __p = const_cast<char_type*>(__str_.data());
+    if (__rbinp != -1)
+        this->setg(__p + __rbinp, __p + __rninp, __p + __reinp);
+    else
+        this->setg(nullptr, nullptr, nullptr);
+    if (__rbout != -1)
+    {
+        this->setp(__p + __rbout, __p + __reout);
+        this->__pbump(__rnout);
+    }
+    else
+        this->setp(nullptr, nullptr);
+    __hm_ = __rhm == -1 ? nullptr : __p + __rhm;
+    __p = const_cast<char_type*>(__rhs.__str_.data());
+    if (__lbinp != -1)
+        __rhs.setg(__p + __lbinp, __p + __lninp, __p + __leinp);
+    else
+        __rhs.setg(nullptr, nullptr, nullptr);
+    if (__lbout != -1)
+    {
+        __rhs.setp(__p + __lbout, __p + __leout);
+        __rhs.__pbump(__lnout);
+    }
+    else
+        __rhs.setp(nullptr, nullptr);
+    __rhs.__hm_ = __lhm == -1 ? nullptr : __p + __lhm;
+    locale __tl = __rhs.getloc();
+    __rhs.pubimbue(this->getloc());
+    this->pubimbue(__tl);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_stringbuf<_CharT, _Traits, _Allocator>& __x,
+     basic_stringbuf<_CharT, _Traits, _Allocator>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>::str() const
+{
+    if (__mode_ & ios_base::out)
+    {
+        if (__hm_ < this->pptr())
+            __hm_ = this->pptr();
+        return string_type(this->pbase(), __hm_, __str_.get_allocator());
+    }
+    else if (__mode_ & ios_base::in)
+        return string_type(this->eback(), this->egptr(), __str_.get_allocator());
+    return string_type(__str_.get_allocator());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_stringbuf<_CharT, _Traits, _Allocator>::str(const string_type& __s)
+{
+    __str_ = __s;
+    __hm_ = 0;
+    if (__mode_ & ios_base::in)
+    {
+        __hm_ = const_cast<char_type*>(__str_.data()) + __str_.size();
+        this->setg(const_cast<char_type*>(__str_.data()),
+                   const_cast<char_type*>(__str_.data()),
+                   __hm_);
+    }
+    if (__mode_ & ios_base::out)
+    {
+        typename string_type::size_type __sz = __str_.size();
+        __hm_ = const_cast<char_type*>(__str_.data()) + __sz;
+        __str_.resize(__str_.capacity());
+        this->setp(const_cast<char_type*>(__str_.data()),
+                   const_cast<char_type*>(__str_.data()) + __str_.size());
+        if (__mode_ & (ios_base::app | ios_base::ate))
+        {
+            while (__sz > INT_MAX)
+            {
+                this->pbump(INT_MAX);
+                __sz -= INT_MAX;
+            }
+            if (__sz > 0)
+                this->pbump(__sz);
+        }
+    }
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_stringbuf<_CharT, _Traits, _Allocator>::int_type
+basic_stringbuf<_CharT, _Traits, _Allocator>::underflow()
+{
+    if (__hm_ < this->pptr())
+        __hm_ = this->pptr();
+    if (__mode_ & ios_base::in)
+    {
+        if (this->egptr() < __hm_)
+            this->setg(this->eback(), this->gptr(), __hm_);
+        if (this->gptr() < this->egptr())
+            return traits_type::to_int_type(*this->gptr());
+    }
+    return traits_type::eof();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_stringbuf<_CharT, _Traits, _Allocator>::int_type
+basic_stringbuf<_CharT, _Traits, _Allocator>::pbackfail(int_type __c)
+{
+    if (__hm_ < this->pptr())
+        __hm_ = this->pptr();
+    if (this->eback() < this->gptr())
+    {
+        if (traits_type::eq_int_type(__c, traits_type::eof()))
+        {
+            this->setg(this->eback(), this->gptr()-1, __hm_);
+            return traits_type::not_eof(__c);
+        }
+        if ((__mode_ & ios_base::out) ||
+            traits_type::eq(traits_type::to_char_type(__c), this->gptr()[-1]))
+        {
+            this->setg(this->eback(), this->gptr()-1, __hm_);
+            *this->gptr() = traits_type::to_char_type(__c);
+            return __c;
+        }
+    }
+    return traits_type::eof();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_stringbuf<_CharT, _Traits, _Allocator>::int_type
+basic_stringbuf<_CharT, _Traits, _Allocator>::overflow(int_type __c)
+{
+    if (!traits_type::eq_int_type(__c, traits_type::eof()))
+    {
+        ptrdiff_t __ninp = this->gptr()  - this->eback();
+        if (this->pptr() == this->epptr())
+        {
+            if (!(__mode_ & ios_base::out))
+                return traits_type::eof();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            try
+            {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+                ptrdiff_t __nout = this->pptr()  - this->pbase();
+                ptrdiff_t __hm = __hm_ - this->pbase();
+                __str_.push_back(char_type());
+                __str_.resize(__str_.capacity());
+                char_type* __p = const_cast<char_type*>(__str_.data());
+                this->setp(__p, __p + __str_.size());
+                this->__pbump(__nout);
+                __hm_ = this->pbase() + __hm;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+            }
+            catch (...)
+            {
+                return traits_type::eof();
+            }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        __hm_ = _CUDA_VSTD::max(this->pptr() + 1, __hm_);
+        if (__mode_ & ios_base::in)
+        {
+            char_type* __p = const_cast<char_type*>(__str_.data());
+            this->setg(__p, __p + __ninp, __hm_);
+        }
+        return this->sputc(traits_type::to_char_type(__c));
+    }
+    return traits_type::not_eof(__c);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_stringbuf<_CharT, _Traits, _Allocator>::pos_type
+basic_stringbuf<_CharT, _Traits, _Allocator>::seekoff(off_type __off,
+                                                      ios_base::seekdir __way,
+                                                      ios_base::openmode __wch)
+{
+    if (__hm_ < this->pptr())
+        __hm_ = this->pptr();
+    if ((__wch & (ios_base::in | ios_base::out)) == 0)
+        return pos_type(-1);
+    if ((__wch & (ios_base::in | ios_base::out)) == (ios_base::in | ios_base::out)
+        && __way == ios_base::cur)
+        return pos_type(-1);
+    const ptrdiff_t __hm = __hm_ == nullptr ? 0 : __hm_ - __str_.data();
+    off_type __noff;
+    switch (__way)
+    {
+    case ios_base::beg:
+        __noff = 0;
+        break;
+    case ios_base::cur:
+        if (__wch & ios_base::in)
+            __noff = this->gptr() - this->eback();
+        else
+            __noff = this->pptr() - this->pbase();
+        break;
+    case ios_base::end:
+        __noff = __hm;
+        break;
+    default:
+        return pos_type(-1);
+    }
+    __noff += __off;
+    if (__noff < 0 || __hm < __noff)
+        return pos_type(-1);
+    if (__noff != 0)
+    {
+        if ((__wch & ios_base::in) && this->gptr() == 0)
+            return pos_type(-1);
+        if ((__wch & ios_base::out) && this->pptr() == 0)
+            return pos_type(-1);
+    }
+    if (__wch & ios_base::in)
+        this->setg(this->eback(), this->eback() + __noff, __hm_);
+    if (__wch & ios_base::out)
+    {
+        this->setp(this->pbase(), this->epptr());
+        this->pbump(__noff);
+    }
+    return pos_type(__noff);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_stringbuf<_CharT, _Traits, _Allocator>::pos_type
+basic_stringbuf<_CharT, _Traits, _Allocator>::seekpos(pos_type __sp,
+                                                      ios_base::openmode __wch)
+{
+    return seekoff(__sp, ios_base::beg, __wch);
+}
+
+// basic_istringstream
+
+template <class _CharT, class _Traits, class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS basic_istringstream
+    : public basic_istream<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef _Allocator                     allocator_type;
+
+    typedef basic_string<char_type, traits_type, allocator_type> string_type;
+
+private:
+    basic_stringbuf<char_type, traits_type, allocator_type> __sb_;
+
+public:
+    // 27.8.2.1 Constructors:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_istringstream(ios_base::openmode __wch = ios_base::in);
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_istringstream(const string_type& __s,
+                                 ios_base::openmode __wch = ios_base::in);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_istringstream(basic_istringstream&& __rhs);
+
+    // 27.8.2.2 Assign and swap:
+    basic_istringstream& operator=(basic_istringstream&& __rhs);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_istringstream& __rhs);
+
+    // 27.8.2.3 Members:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_stringbuf<char_type, traits_type, allocator_type>* rdbuf() const;
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    string_type str() const;
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void str(const string_type& __s);
+};
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_istringstream<_CharT, _Traits, _Allocator>::basic_istringstream(ios_base::openmode __wch)
+    : basic_istream<_CharT, _Traits>(&__sb_),
+      __sb_(__wch | ios_base::in)
+{
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_istringstream<_CharT, _Traits, _Allocator>::basic_istringstream(const string_type& __s,
+                                                                      ios_base::openmode __wch)
+    : basic_istream<_CharT, _Traits>(&__sb_),
+      __sb_(__s, __wch | ios_base::in)
+{
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_istringstream<_CharT, _Traits, _Allocator>::basic_istringstream(basic_istringstream&& __rhs)
+    : basic_istream<_CharT, _Traits>(_CUDA_VSTD::move(__rhs)),
+      __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+{
+    basic_istream<_CharT, _Traits>::set_rdbuf(&__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_istringstream<_CharT, _Traits, _Allocator>&
+basic_istringstream<_CharT, _Traits, _Allocator>::operator=(basic_istringstream&& __rhs)
+{
+    basic_istream<char_type, traits_type>::operator=(_CUDA_VSTD::move(__rhs));
+    __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+void basic_istringstream<_CharT, _Traits, _Allocator>::swap(basic_istringstream& __rhs)
+{
+    basic_istream<char_type, traits_type>::swap(__rhs);
+    __sb_.swap(__rhs.__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_istringstream<_CharT, _Traits, _Allocator>& __x,
+     basic_istringstream<_CharT, _Traits, _Allocator>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>*
+basic_istringstream<_CharT, _Traits, _Allocator>::rdbuf() const
+{
+    return const_cast<basic_stringbuf<char_type, traits_type, allocator_type>*>(&__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+basic_istringstream<_CharT, _Traits, _Allocator>::str() const
+{
+    return __sb_.str();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void basic_istringstream<_CharT, _Traits, _Allocator>::str(const string_type& __s)
+{
+    __sb_.str(__s);
+}
+
+// basic_ostringstream
+
+template <class _CharT, class _Traits, class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS basic_ostringstream
+    : public basic_ostream<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef _Allocator                     allocator_type;
+
+    typedef basic_string<char_type, traits_type, allocator_type> string_type;
+
+private:
+    basic_stringbuf<char_type, traits_type, allocator_type> __sb_;
+
+public:
+    // 27.8.2.1 Constructors:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ostringstream(ios_base::openmode __wch = ios_base::out);
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_ostringstream(const string_type& __s,
+                                 ios_base::openmode __wch = ios_base::out);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_ostringstream(basic_ostringstream&& __rhs);
+
+    // 27.8.2.2 Assign and swap:
+    basic_ostringstream& operator=(basic_ostringstream&& __rhs);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_ostringstream& __rhs);
+
+    // 27.8.2.3 Members:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_stringbuf<char_type, traits_type, allocator_type>* rdbuf() const;
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    string_type str() const;
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void str(const string_type& __s);
+};
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_ostringstream<_CharT, _Traits, _Allocator>::basic_ostringstream(ios_base::openmode __wch)
+    : basic_ostream<_CharT, _Traits>(&__sb_),
+      __sb_(__wch | ios_base::out)
+{
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_ostringstream<_CharT, _Traits, _Allocator>::basic_ostringstream(const string_type& __s,
+                                                                      ios_base::openmode __wch)
+    : basic_ostream<_CharT, _Traits>(&__sb_),
+      __sb_(__s, __wch | ios_base::out)
+{
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_ostringstream<_CharT, _Traits, _Allocator>::basic_ostringstream(basic_ostringstream&& __rhs)
+    : basic_ostream<_CharT, _Traits>(_CUDA_VSTD::move(__rhs)),
+      __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+{
+    basic_ostream<_CharT, _Traits>::set_rdbuf(&__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_ostringstream<_CharT, _Traits, _Allocator>&
+basic_ostringstream<_CharT, _Traits, _Allocator>::operator=(basic_ostringstream&& __rhs)
+{
+    basic_ostream<char_type, traits_type>::operator=(_CUDA_VSTD::move(__rhs));
+    __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_ostringstream<_CharT, _Traits, _Allocator>::swap(basic_ostringstream& __rhs)
+{
+    basic_ostream<char_type, traits_type>::swap(__rhs);
+    __sb_.swap(__rhs.__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_ostringstream<_CharT, _Traits, _Allocator>& __x,
+     basic_ostringstream<_CharT, _Traits, _Allocator>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>*
+basic_ostringstream<_CharT, _Traits, _Allocator>::rdbuf() const
+{
+    return const_cast<basic_stringbuf<char_type, traits_type, allocator_type>*>(&__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+basic_ostringstream<_CharT, _Traits, _Allocator>::str() const
+{
+    return __sb_.str();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_ostringstream<_CharT, _Traits, _Allocator>::str(const string_type& __s)
+{
+    __sb_.str(__s);
+}
+
+// basic_stringstream
+
+template <class _CharT, class _Traits, class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS basic_stringstream
+    : public basic_iostream<_CharT, _Traits>
+{
+public:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+    typedef _Allocator                     allocator_type;
+
+    typedef basic_string<char_type, traits_type, allocator_type> string_type;
+
+private:
+    basic_stringbuf<char_type, traits_type, allocator_type> __sb_;
+
+public:
+    // 27.8.2.1 Constructors:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_stringstream(ios_base::openmode __wch = ios_base::in | ios_base::out);
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    explicit basic_stringstream(const string_type& __s,
+                                ios_base::openmode __wch = ios_base::in | ios_base::out);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_stringstream(basic_stringstream&& __rhs);
+
+    // 27.8.2.2 Assign and swap:
+    basic_stringstream& operator=(basic_stringstream&& __rhs);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_stringstream& __rhs);
+
+    // 27.8.2.3 Members:
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_stringbuf<char_type, traits_type, allocator_type>* rdbuf() const;
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    string_type str() const;
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    void str(const string_type& __s);
+};
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringstream<_CharT, _Traits, _Allocator>::basic_stringstream(ios_base::openmode __wch)
+    : basic_iostream<_CharT, _Traits>(&__sb_),
+      __sb_(__wch)
+{
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringstream<_CharT, _Traits, _Allocator>::basic_stringstream(const string_type& __s,
+                                                                    ios_base::openmode __wch)
+    : basic_iostream<_CharT, _Traits>(&__sb_),
+      __sb_(__s, __wch)
+{
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringstream<_CharT, _Traits, _Allocator>::basic_stringstream(basic_stringstream&& __rhs)
+    : basic_iostream<_CharT, _Traits>(_CUDA_VSTD::move(__rhs)),
+      __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+{
+    basic_istream<_CharT, _Traits>::set_rdbuf(&__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringstream<_CharT, _Traits, _Allocator>&
+basic_stringstream<_CharT, _Traits, _Allocator>::operator=(basic_stringstream&& __rhs)
+{
+    basic_iostream<char_type, traits_type>::operator=(_CUDA_VSTD::move(__rhs));
+    __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_stringstream<_CharT, _Traits, _Allocator>::swap(basic_stringstream& __rhs)
+{
+    basic_iostream<char_type, traits_type>::swap(__rhs);
+    __sb_.swap(__rhs.__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_stringstream<_CharT, _Traits, _Allocator>& __x,
+     basic_stringstream<_CharT, _Traits, _Allocator>& __y)
+{
+    __x.swap(__y);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_stringbuf<_CharT, _Traits, _Allocator>*
+basic_stringstream<_CharT, _Traits, _Allocator>::rdbuf() const
+{
+    return const_cast<basic_stringbuf<char_type, traits_type, allocator_type>*>(&__sb_);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+basic_stringstream<_CharT, _Traits, _Allocator>::str() const
+{
+    return __sb_.str();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_stringstream<_CharT, _Traits, _Allocator>::str(const string_type& __s)
+{
+    __sb_.str(__s);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_SSTREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stack b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stack
new file mode 100644
index 000000000000..96355ba680a6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stack
@@ -0,0 +1,321 @@
+// -*- C++ -*-
+//===---------------------------- stack -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STACK
+#define _LIBCUDACXX_STACK
+
+/*
+    stack synopsis
+
+namespace std
+{
+
+template <class T, class Container = deque<T>>
+class stack
+{
+public:
+    typedef Container                                container_type;
+    typedef typename container_type::value_type      value_type;
+    typedef typename container_type::reference       reference;
+    typedef typename container_type::const_reference const_reference;
+    typedef typename container_type::size_type       size_type;
+
+protected:
+    container_type c;
+
+public:
+    stack() = default;
+    ~stack() = default;
+
+    stack(const stack& q) = default;
+    stack(stack&& q) = default;
+
+    stack& operator=(const stack& q) = default;
+    stack& operator=(stack&& q) = default;
+
+    explicit stack(const container_type& c);
+    explicit stack(container_type&& c);
+    template <class Alloc> explicit stack(const Alloc& a);
+    template <class Alloc> stack(const container_type& c, const Alloc& a);
+    template <class Alloc> stack(container_type&& c, const Alloc& a);
+    template <class Alloc> stack(const stack& c, const Alloc& a);
+    template <class Alloc> stack(stack&& c, const Alloc& a);
+
+    bool empty() const;
+    size_type size() const;
+    reference top();
+    const_reference top() const;
+
+    void push(const value_type& x);
+    void push(value_type&& x);
+    template <class... Args> reference emplace(Args&&... args); // reference in C++17
+    void pop();
+
+    void swap(stack& c) noexcept(is_nothrow_swappable_v<Container>)
+};
+
+template<class Container>
+  stack(Container) -> stack<typename Container::value_type, Container>;  // C++17
+
+template<class Container, class Allocator>
+  stack(Container, Allocator) -> stack<typename Container::value_type, Container>; // C++17
+
+template <class T, class Container>
+  bool operator==(const stack<T, Container>& x, const stack<T, Container>& y);
+template <class T, class Container>
+  bool operator< (const stack<T, Container>& x, const stack<T, Container>& y);
+template <class T, class Container>
+  bool operator!=(const stack<T, Container>& x, const stack<T, Container>& y);
+template <class T, class Container>
+  bool operator> (const stack<T, Container>& x, const stack<T, Container>& y);
+template <class T, class Container>
+  bool operator>=(const stack<T, Container>& x, const stack<T, Container>& y);
+template <class T, class Container>
+  bool operator<=(const stack<T, Container>& x, const stack<T, Container>& y);
+
+template <class T, class Container>
+  void swap(stack<T, Container>& x, stack<T, Container>& y)
+  noexcept(noexcept(x.swap(y)));
+
+}  // std
+
+*/
+
+#include <__config>
+#include <deque>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp, class _Container = deque<_Tp> > class _LIBCUDACXX_TEMPLATE_VIS stack;
+
+template <class _Tp, class _Container>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y);
+
+template <class _Tp, class _Container>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y);
+
+template <class _Tp, class _Container /*= deque<_Tp>*/>
+class _LIBCUDACXX_TEMPLATE_VIS stack
+{
+public:
+    typedef _Container                               container_type;
+    typedef typename container_type::value_type      value_type;
+    typedef typename container_type::reference       reference;
+    typedef typename container_type::const_reference const_reference;
+    typedef typename container_type::size_type       size_type;
+    static_assert((is_same<_Tp, value_type>::value), "" );
+
+protected:
+    container_type c;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    stack()
+        _NOEXCEPT_(is_nothrow_default_constructible<container_type>::value)
+        : c() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    stack(const stack& __q) : c(__q.c) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    stack& operator=(const stack& __q) {c = __q.c; return *this;}
+
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    stack(stack&& __q)
+        _NOEXCEPT_(is_nothrow_move_constructible<container_type>::value)
+        : c(_CUDA_VSTD::move(__q.c)) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    stack& operator=(stack&& __q)
+        _NOEXCEPT_(is_nothrow_move_assignable<container_type>::value)
+        {c = _CUDA_VSTD::move(__q.c); return *this;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit stack(container_type&& __c) : c(_CUDA_VSTD::move(__c)) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit stack(const container_type& __c) : c(__c) {}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit stack(const _Alloc& __a,
+                       typename enable_if<uses_allocator<container_type,
+                                                         _Alloc>::value>::type* = 0)
+            : c(__a) {}
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        stack(const container_type& __c, const _Alloc& __a,
+              typename enable_if<uses_allocator<container_type,
+                                                _Alloc>::value>::type* = 0)
+            : c(__c, __a) {}
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        stack(const stack& __s, const _Alloc& __a,
+              typename enable_if<uses_allocator<container_type,
+                                                _Alloc>::value>::type* = 0)
+            : c(__s.c, __a) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        stack(container_type&& __c, const _Alloc& __a,
+              typename enable_if<uses_allocator<container_type,
+                                                _Alloc>::value>::type* = 0)
+            : c(_CUDA_VSTD::move(__c), __a) {}
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        stack(stack&& __s, const _Alloc& __a,
+              typename enable_if<uses_allocator<container_type,
+                                                _Alloc>::value>::type* = 0)
+            : c(_CUDA_VSTD::move(__s.c), __a) {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty()     const      {return c.empty();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const      {return c.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference top()             {return c.back();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference top() const {return c.back();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void push(const value_type& __v) {c.push_back(__v);}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void push(value_type&& __v) {c.push_back(_CUDA_VSTD::move(__v));}
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+#if _LIBCUDACXX_STD_VER > 14
+        decltype(auto) emplace(_Args&&... __args)
+        { return c.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);}
+#else
+        void      emplace(_Args&&... __args)
+        {        c.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);}
+#endif
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pop() {c.pop_back();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(stack& __s)
+        _NOEXCEPT_(__is_nothrow_swappable<container_type>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(c, __s.c);
+    }
+
+    template <class T1, class _C1>
+    friend
+    bool
+    operator==(const stack<T1, _C1>& __x, const stack<T1, _C1>& __y);
+
+    template <class T1, class _C1>
+    friend
+    bool
+    operator< (const stack<T1, _C1>& __x, const stack<T1, _C1>& __y);
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _Container,
+         class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type
+>
+stack(_Container)
+    -> stack<typename _Container::value_type, _Container>;
+
+template<class _Container,
+         class _Alloc,
+         class = typename enable_if<!__is_allocator<_Container>::value, nullptr_t>::type,
+         class = typename enable_if< __is_allocator<_Alloc>::value, nullptr_t>::type
+         >
+stack(_Container, _Alloc)
+    -> stack<typename _Container::value_type, _Container>;
+#endif
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y)
+{
+    return __x.c == __y.c;
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y)
+{
+    return __x.c < __y.c;
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const stack<_Tp, _Container>& __x, const stack<_Tp, _Container>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, class _Container>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if<
+    __is_swappable<_Container>::value,
+    void
+>::type
+swap(stack<_Tp, _Container>& __x, stack<_Tp, _Container>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Tp, class _Container, class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator<stack<_Tp, _Container>, _Alloc>
+    : public uses_allocator<_Container, _Alloc>
+{
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_STACK
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h
new file mode 100644
index 000000000000..274609ad0d92
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdbool.h
@@ -0,0 +1,38 @@
+// -*- C++ -*-
+//===--------------------------- stdbool.h --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef _LIBCUDACXX_STDBOOL_H
+#define _LIBCUDACXX_STDBOOL_H
+
+
+/*
+    stdbool.h synopsis
+
+Macros:
+
+    __bool_true_false_are_defined
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <stdbool.h>
+
+#ifdef __cplusplus
+#undef bool
+#undef true
+#undef false
+#undef __bool_true_false_are_defined
+#define __bool_true_false_are_defined 1
+#endif
+
+#endif  // _LIBCUDACXX_STDBOOL_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h
new file mode 100644
index 000000000000..fe0a0e9d8a11
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stddef.h
@@ -0,0 +1,66 @@
+// -*- C++ -*-
+//===--------------------------- stddef.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__need_ptrdiff_t) || defined(__need_size_t) || \
+    defined(__need_wchar_t) || defined(__need_NULL) || defined(__need_wint_t)
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <stddef.h>
+
+#elif !defined(_LIBCUDACXX_STDDEF_H)
+#define _LIBCUDACXX_STDDEF_H
+
+/*
+    stddef.h synopsis
+
+Macros:
+
+    offsetof(type,member-designator)
+    NULL
+
+Types:
+
+    ptrdiff_t
+    size_t
+    max_align_t
+    nullptr_t
+
+*/
+
+#include <__config>
+
+#include <__pragma_push>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <stddef.h>
+
+#ifdef __cplusplus
+
+extern "C++" {
+#include <__nullptr>
+using std::nullptr_t;
+}
+
+// Re-use the compiler's <stddef.h> max_align_t where possible.
+#if !defined(__CLANG_MAX_ALIGN_T_DEFINED) && !defined(_GCC_MAX_ALIGN_T) && \
+    !defined(__DEFINED_max_align_t) && !defined(__NetBSD__)
+typedef long double max_align_t;
+#endif
+
+#endif
+
+#include <__pragma_pop>
+
+#endif  // _LIBCUDACXX_STDDEF_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept
new file mode 100644
index 000000000000..8b03be60c4aa
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdexcept
@@ -0,0 +1,321 @@
+// -*- C++ -*-
+//===--------------------------- stdexcept --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STDEXCEPT
+#define _LIBCUDACXX_STDEXCEPT
+
+/*
+    stdexcept synopsis
+
+namespace std
+{
+
+class logic_error;
+    class domain_error;
+    class invalid_argument;
+    class length_error;
+    class out_of_range;
+class runtime_error;
+    class range_error;
+    class overflow_error;
+    class underflow_error;
+
+for each class xxx_error:
+
+class xxx_error : public exception // at least indirectly
+{
+public:
+    explicit xxx_error(const string& what_arg);
+    explicit xxx_error(const char*   what_arg);
+
+    virtual const char* what() const noexcept // returns what_arg
+};
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <exception>
+#include <iosfwd>  // for string forward decl
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+#include <cstdlib>
+#endif
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+class _LIBCUDACXX_HIDDEN __libcpp_refstring
+{
+    const char* __imp_;
+
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER bool __uses_refcount() const;
+public:
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER explicit __libcpp_refstring(const char* __msg);
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER __libcpp_refstring(const __libcpp_refstring& __s) _NOEXCEPT;
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER __libcpp_refstring& operator=(const __libcpp_refstring& __s) _NOEXCEPT;
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER ~__libcpp_refstring();
+
+    _LIBCUDACXX_EXECUTION_SPACE_SPECIFIER const char* c_str() const _NOEXCEPT {return __imp_;}
+};
+#endif // !_LIBCUDACXX_ABI_VCRUNTIME
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#ifdef _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+_LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+#else
+namespace std  // purposefully not versioned
+{
+#endif //_LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+
+class _LIBCUDACXX_EXCEPTION_ABI logic_error
+    : public exception
+{
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+private:
+    _CUDA_VSTD::__libcpp_refstring __imp_;
+public:
+    explicit logic_error(const string&);
+    explicit logic_error(const char*);
+
+    logic_error(const logic_error&) _NOEXCEPT;
+    logic_error& operator=(const logic_error&) _NOEXCEPT;
+
+    virtual ~logic_error() _NOEXCEPT;
+
+    virtual const char* what() const _NOEXCEPT;
+#else
+public:
+    explicit logic_error(const _CUDA_VSTD::string&); // Symbol uses versioned std::string
+    _LIBCUDACXX_INLINE_VISIBILITY explicit logic_error(const char* __s) : exception(__s) {}
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI runtime_error
+    : public exception
+{
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+private:
+    _CUDA_VSTD::__libcpp_refstring __imp_;
+public:
+    explicit runtime_error(const string&);
+    explicit runtime_error(const char*);
+
+    runtime_error(const runtime_error&) _NOEXCEPT;
+    runtime_error& operator=(const runtime_error&) _NOEXCEPT;
+
+    virtual ~runtime_error() _NOEXCEPT;
+
+    virtual const char* what() const _NOEXCEPT;
+#else
+public:
+   explicit runtime_error(const _CUDA_VSTD::string&); // Symbol uses versioned std::string
+   _LIBCUDACXX_INLINE_VISIBILITY explicit runtime_error(const char* __s) : exception(__s) {}
+#endif // _LIBCUDACXX_ABI_VCRUNTIME
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI domain_error
+    : public logic_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit domain_error(const string& __s) : logic_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit domain_error(const char* __s)   : logic_error(__s) {}
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~domain_error() _NOEXCEPT;
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI invalid_argument
+    : public logic_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit invalid_argument(const string& __s) : logic_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit invalid_argument(const char* __s)   : logic_error(__s) {}
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~invalid_argument() _NOEXCEPT;
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI length_error
+    : public logic_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit length_error(const string& __s) : logic_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit length_error(const char* __s)   : logic_error(__s) {}
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~length_error() _NOEXCEPT;
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI out_of_range
+    : public logic_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit out_of_range(const string& __s) : logic_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit out_of_range(const char* __s)   : logic_error(__s) {}
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~out_of_range() _NOEXCEPT;
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI range_error
+    : public runtime_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit range_error(const string& __s) : runtime_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit range_error(const char* __s)   : runtime_error(__s) {}
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~range_error() _NOEXCEPT;
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI overflow_error
+    : public runtime_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit overflow_error(const string& __s) : runtime_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit overflow_error(const char* __s)   : runtime_error(__s) {}
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~overflow_error() _NOEXCEPT;
+#endif
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI underflow_error
+    : public runtime_error
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY explicit underflow_error(const string& __s) : runtime_error(__s) {}
+    _LIBCUDACXX_INLINE_VISIBILITY explicit underflow_error(const char* __s)   : runtime_error(__s) {}
+
+#ifndef _LIBCUDACXX_ABI_VCRUNTIME
+    virtual ~underflow_error() _NOEXCEPT;
+#endif
+};
+
+#ifdef _LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+_LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+#else
+}
+#endif //_LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+#endif //__cuda_std__
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// in the dylib
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS void __throw_runtime_error(const char*);
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_logic_error(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw logic_error(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_domain_error(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw domain_error(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_invalid_argument(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw invalid_argument(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_length_error(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw length_error(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_out_of_range(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw out_of_range(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_range_error(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw range_error(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_overflow_error(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw overflow_error(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_underflow_error(const char*__msg)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw underflow_error(__msg);
+#else
+    ((void)__msg);
+    _LIBCUDACXX_UNREACHABLE();
+#endif
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_STDEXCEPT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h
new file mode 100644
index 000000000000..6a95b3362a51
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdint.h
@@ -0,0 +1,129 @@
+// -*- C++ -*-
+//===---------------------------- stdint.h --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STDINT_H
+// AIX system headers need stdint.h to be re-enterable while _STD_TYPES_T
+// is defined until an inclusion of it without _STD_TYPES_T occurs, in which
+// case the header guard macro is defined.
+#if !defined(_AIX) || !defined(_STD_TYPES_T)
+#define _LIBCUDACXX_STDINT_H
+#endif // _STD_TYPES_T
+
+/*
+    stdint.h synopsis
+
+Macros:
+
+    INT8_MIN
+    INT16_MIN
+    INT32_MIN
+    INT64_MIN
+
+    INT8_MAX
+    INT16_MAX
+    INT32_MAX
+    INT64_MAX
+
+    UINT8_MAX
+    UINT16_MAX
+    UINT32_MAX
+    UINT64_MAX
+
+    INT_LEAST8_MIN
+    INT_LEAST16_MIN
+    INT_LEAST32_MIN
+    INT_LEAST64_MIN
+
+    INT_LEAST8_MAX
+    INT_LEAST16_MAX
+    INT_LEAST32_MAX
+    INT_LEAST64_MAX
+
+    UINT_LEAST8_MAX
+    UINT_LEAST16_MAX
+    UINT_LEAST32_MAX
+    UINT_LEAST64_MAX
+
+    INT_FAST8_MIN
+    INT_FAST16_MIN
+    INT_FAST32_MIN
+    INT_FAST64_MIN
+
+    INT_FAST8_MAX
+    INT_FAST16_MAX
+    INT_FAST32_MAX
+    INT_FAST64_MAX
+
+    UINT_FAST8_MAX
+    UINT_FAST16_MAX
+    UINT_FAST32_MAX
+    UINT_FAST64_MAX
+
+    INTPTR_MIN
+    INTPTR_MAX
+    UINTPTR_MAX
+
+    INTMAX_MIN
+    INTMAX_MAX
+
+    UINTMAX_MAX
+
+    PTRDIFF_MIN
+    PTRDIFF_MAX
+
+    SIG_ATOMIC_MIN
+    SIG_ATOMIC_MAX
+
+    SIZE_MAX
+
+    WCHAR_MIN
+    WCHAR_MAX
+
+    WINT_MIN
+    WINT_MAX
+
+    INT8_C(value)
+    INT16_C(value)
+    INT32_C(value)
+    INT64_C(value)
+
+    UINT8_C(value)
+    UINT16_C(value)
+    UINT32_C(value)
+    UINT64_C(value)
+
+    INTMAX_C(value)
+    UINTMAX_C(value)
+
+*/
+
+#include <__config>
+
+#include <__pragma_push>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+/* C99 stdlib (e.g. glibc < 2.18) does not provide macros needed
+   for C++11 unless __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS
+   are defined
+*/
+#if defined(__cplusplus) && !defined(__STDC_LIMIT_MACROS)
+#   define __STDC_LIMIT_MACROS
+#endif
+#if defined(__cplusplus) && !defined(__STDC_CONSTANT_MACROS)
+#   define __STDC_CONSTANT_MACROS
+#endif
+
+#include_next <stdint.h>
+
+#include <__pragma_pop>
+
+#endif  // _LIBCUDACXX_STDINT_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h
new file mode 100644
index 000000000000..e793e93de9df
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdio.h
@@ -0,0 +1,119 @@
+// -*- C++ -*-
+//===---------------------------- stdio.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__need_FILE) || defined(__need___FILE)
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <stdio.h>
+
+#elif !defined(_LIBCUDACXX_STDIO_H)
+#define _LIBCUDACXX_STDIO_H
+
+/*
+    stdio.h synopsis
+
+Macros:
+
+    BUFSIZ
+    EOF
+    FILENAME_MAX
+    FOPEN_MAX
+    L_tmpnam
+    NULL
+    SEEK_CUR
+    SEEK_END
+    SEEK_SET
+    TMP_MAX
+    _IOFBF
+    _IOLBF
+    _IONBF
+    stderr
+    stdin
+    stdout
+
+Types:
+
+FILE
+fpos_t
+size_t
+
+int remove(const char* filename);
+int rename(const char* old, const char* new);
+FILE* tmpfile(void);
+char* tmpnam(char* s);
+int fclose(FILE* stream);
+int fflush(FILE* stream);
+FILE* fopen(const char* restrict filename, const char* restrict mode);
+FILE* freopen(const char* restrict filename, const char * restrict mode,
+              FILE * restrict stream);
+void setbuf(FILE* restrict stream, char* restrict buf);
+int setvbuf(FILE* restrict stream, char* restrict buf, int mode, size_t size);
+int fprintf(FILE* restrict stream, const char* restrict format, ...);
+int fscanf(FILE* restrict stream, const char * restrict format, ...);
+int printf(const char* restrict format, ...);
+int scanf(const char* restrict format, ...);
+int snprintf(char* restrict s, size_t n, const char* restrict format, ...);    // C99
+int sprintf(char* restrict s, const char* restrict format, ...);
+int sscanf(const char* restrict s, const char* restrict format, ...);
+int vfprintf(FILE* restrict stream, const char* restrict format, va_list arg);
+int vfscanf(FILE* restrict stream, const char* restrict format, va_list arg);  // C99
+int vprintf(const char* restrict format, va_list arg);
+int vscanf(const char* restrict format, va_list arg);                          // C99
+int vsnprintf(char* restrict s, size_t n, const char* restrict format,         // C99
+              va_list arg);
+int vsprintf(char* restrict s, const char* restrict format, va_list arg);
+int vsscanf(const char* restrict s, const char* restrict format, va_list arg); // C99
+int fgetc(FILE* stream);
+char* fgets(char* restrict s, int n, FILE* restrict stream);
+int fputc(int c, FILE* stream);
+int fputs(const char* restrict s, FILE* restrict stream);
+int getc(FILE* stream);
+int getchar(void);
+char* gets(char* s);  // removed in C++14
+int putc(int c, FILE* stream);
+int putchar(int c);
+int puts(const char* s);
+int ungetc(int c, FILE* stream);
+size_t fread(void* restrict ptr, size_t size, size_t nmemb,
+             FILE* restrict stream);
+size_t fwrite(const void* restrict ptr, size_t size, size_t nmemb,
+              FILE* restrict stream);
+int fgetpos(FILE* restrict stream, fpos_t* restrict pos);
+int fseek(FILE* stream, long offset, int whence);
+int fsetpos(FILE*stream, const fpos_t* pos);
+long ftell(FILE* stream);
+void rewind(FILE* stream);
+void clearerr(FILE* stream);
+int feof(FILE* stream);
+int ferror(FILE* stream);
+void perror(const char* s);
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <stdio.h>
+
+#ifdef __cplusplus
+
+#undef getc
+#undef putc
+#undef clearerr
+#undef feof
+#undef ferror
+
+#endif
+
+#endif  // _LIBCUDACXX_STDIO_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h
new file mode 100644
index 000000000000..a980e7642c50
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/stdlib.h
@@ -0,0 +1,103 @@
+// -*- C++ -*-
+//===--------------------------- stdlib.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__need_malloc_and_calloc) || defined(_LIBCUDACXX_STDLIB_INCLUDE_NEXT)
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if defined(_LIBCUDACXX_STDLIB_INCLUDE_NEXT)
+#undef _LIBCUDACXX_STDLIB_INCLUDE_NEXT
+#endif
+
+#include_next <stdlib.h>
+
+#elif !defined(_LIBCUDACXX_STDLIB_H)
+#define _LIBCUDACXX_STDLIB_H
+
+/*
+    stdlib.h synopsis
+
+Macros:
+
+    EXIT_FAILURE
+    EXIT_SUCCESS
+    MB_CUR_MAX
+    NULL
+    RAND_MAX
+
+Types:
+
+    size_t
+    div_t
+    ldiv_t
+    lldiv_t                                                               // C99
+
+double    atof (const char* nptr);
+int       atoi (const char* nptr);
+long      atol (const char* nptr);
+long long atoll(const char* nptr);                                        // C99
+double             strtod  (const char* restrict nptr, char** restrict endptr);
+float              strtof  (const char* restrict nptr, char** restrict endptr); // C99
+long double        strtold (const char* restrict nptr, char** restrict endptr); // C99
+long               strtol  (const char* restrict nptr, char** restrict endptr, int base);
+long long          strtoll (const char* restrict nptr, char** restrict endptr, int base); // C99
+unsigned long      strtoul (const char* restrict nptr, char** restrict endptr, int base);
+unsigned long long strtoull(const char* restrict nptr, char** restrict endptr, int base); // C99
+int rand(void);
+void srand(unsigned int seed);
+void* calloc(size_t nmemb, size_t size);
+void free(void* ptr);
+void* malloc(size_t size);
+void* realloc(void* ptr, size_t size);
+void abort(void);
+int atexit(void (*func)(void));
+void exit(int status);
+void _Exit(int status);
+char* getenv(const char* name);
+int system(const char* string);
+void* bsearch(const void* key, const void* base, size_t nmemb, size_t size,
+              int (*compar)(const void *, const void *));
+void qsort(void* base, size_t nmemb, size_t size,
+           int (*compar)(const void *, const void *));
+int         abs(      int j);
+long        abs(     long j);
+long long   abs(long long j);                                             // C++0X
+long       labs(     long j);
+long long llabs(long long j);                                             // C99
+div_t     div(      int numer,       int denom);
+ldiv_t    div(     long numer,      long denom);
+lldiv_t   div(long long numer, long long denom);                          // C++0X
+ldiv_t   ldiv(     long numer,      long denom);
+lldiv_t lldiv(long long numer, long long denom);                          // C99
+int mblen(const char* s, size_t n);
+int mbtowc(wchar_t* restrict pwc, const char* restrict s, size_t n);
+int wctomb(char* s, wchar_t wchar);
+size_t mbstowcs(wchar_t* restrict pwcs, const char* restrict s, size_t n);
+size_t wcstombs(char* restrict s, const wchar_t* restrict pwcs, size_t n);
+int at_quick_exit(void (*func)(void))                                     // C++11
+void quick_exit(int status);                                              // C++11
+void *aligned_alloc(size_t alignment, size_t size);                       // C11
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <stdlib.h>
+
+#ifdef __cplusplus
+#include <math.h>
+#endif  // __cplusplus
+
+#endif  // _LIBCUDACXX_STDLIB_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf
new file mode 100644
index 000000000000..8a305dc2ad5f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/streambuf
@@ -0,0 +1,500 @@
+// -*- C++ -*-
+//===------------------------- streambuf ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STEAMBUF
+#define _LIBCUDACXX_STEAMBUF
+
+/*
+    streambuf synopsis
+
+namespace std
+{
+
+template <class charT, class traits = char_traits<charT> >
+class basic_streambuf
+{
+public:
+    // types:
+    typedef charT char_type;
+    typedef traits traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    virtual ~basic_streambuf();
+
+    // 27.6.2.2.1 locales:
+    locale pubimbue(const locale& loc);
+    locale getloc() const;
+
+    // 27.6.2.2.2 buffer and positioning:
+    basic_streambuf* pubsetbuf(char_type* s, streamsize n);
+    pos_type pubseekoff(off_type off, ios_base::seekdir way,
+                        ios_base::openmode which = ios_base::in | ios_base::out);
+    pos_type pubseekpos(pos_type sp,
+                        ios_base::openmode which = ios_base::in | ios_base::out);
+    int pubsync();
+
+    // Get and put areas:
+    // 27.6.2.2.3 Get area:
+    streamsize in_avail();
+    int_type snextc();
+    int_type sbumpc();
+    int_type sgetc();
+    streamsize sgetn(char_type* s, streamsize n);
+
+    // 27.6.2.2.4 Putback:
+    int_type sputbackc(char_type c);
+    int_type sungetc();
+
+    // 27.6.2.2.5 Put area:
+    int_type sputc(char_type c);
+    streamsize sputn(const char_type* s, streamsize n);
+
+protected:
+    basic_streambuf();
+    basic_streambuf(const basic_streambuf& rhs);
+    basic_streambuf& operator=(const basic_streambuf& rhs);
+    void swap(basic_streambuf& rhs);
+
+    // 27.6.2.3.2 Get area:
+    char_type* eback() const;
+    char_type* gptr() const;
+    char_type* egptr() const;
+    void gbump(int n);
+    void setg(char_type* gbeg, char_type* gnext, char_type* gend);
+
+    // 27.6.2.3.3 Put area:
+    char_type* pbase() const;
+    char_type* pptr() const;
+    char_type* epptr() const;
+    void pbump(int n);
+    void setp(char_type* pbeg, char_type* pend);
+
+    // 27.6.2.4 virtual functions:
+    // 27.6.2.4.1 Locales:
+    virtual void imbue(const locale& loc);
+
+    // 27.6.2.4.2 Buffer management and positioning:
+    virtual basic_streambuf* setbuf(char_type* s, streamsize n);
+    virtual pos_type seekoff(off_type off, ios_base::seekdir way,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type sp,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual int sync();
+
+    // 27.6.2.4.3 Get area:
+    virtual streamsize showmanyc();
+    virtual streamsize xsgetn(char_type* s, streamsize n);
+    virtual int_type underflow();
+    virtual int_type uflow();
+
+    // 27.6.2.4.4 Putback:
+    virtual int_type pbackfail(int_type c = traits_type::eof());
+
+    // 27.6.2.4.5 Put area:
+    virtual streamsize xsputn(const char_type* s, streamsize n);
+    virtual int_type overflow (int_type c = traits_type::eof());
+};
+
+}  // std
+
+*/
+
+#include <__config>
+#include <iosfwd>
+#include <ios>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _CharT, class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS basic_streambuf
+{
+public:
+    // types:
+    typedef _CharT                         char_type;
+    typedef _Traits                        traits_type;
+    typedef typename traits_type::int_type int_type;
+    typedef typename traits_type::pos_type pos_type;
+    typedef typename traits_type::off_type off_type;
+
+    static_assert((is_same<_CharT, typename traits_type::char_type>::value),
+                  "traits_type::char_type must be the same type as CharT");
+
+    virtual ~basic_streambuf();
+
+    // 27.6.2.2.1 locales:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    locale pubimbue(const locale& __loc) {
+        imbue(__loc);
+        locale __r = __loc_;
+        __loc_ = __loc;
+        return __r;
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    locale getloc() const { return __loc_; }
+
+    // 27.6.2.2.2 buffer and positioning:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    basic_streambuf* pubsetbuf(char_type* __s, streamsize __n)
+    { return setbuf(__s, __n); }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    pos_type pubseekoff(off_type __off, ios_base::seekdir __way,
+                        ios_base::openmode __which = ios_base::in | ios_base::out)
+    { return seekoff(__off, __way, __which); }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    pos_type pubseekpos(pos_type __sp,
+                        ios_base::openmode __which = ios_base::in | ios_base::out)
+    { return seekpos(__sp, __which); }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int pubsync() { return sync(); }
+
+    // Get and put areas:
+    // 27.6.2.2.3 Get area:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    streamsize in_avail() {
+        if (__ninp_ < __einp_)
+            return static_cast<streamsize>(__einp_ - __ninp_);
+        return showmanyc();
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int_type snextc() {
+        if (sbumpc() == traits_type::eof())
+            return traits_type::eof();
+        return sgetc();
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int_type sbumpc() {
+        if (__ninp_ == __einp_)
+            return uflow();
+        return traits_type::to_int_type(*__ninp_++);
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int_type sgetc() {
+        if (__ninp_ == __einp_)
+            return underflow();
+        return traits_type::to_int_type(*__ninp_);
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    streamsize sgetn(char_type* __s, streamsize __n)
+    { return xsgetn(__s, __n); }
+
+    // 27.6.2.2.4 Putback:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int_type sputbackc(char_type __c) {
+        if (__binp_ == __ninp_ || !traits_type::eq(__c, __ninp_[-1]))
+            return pbackfail(traits_type::to_int_type(__c));
+        return traits_type::to_int_type(*--__ninp_);
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int_type sungetc() {
+        if (__binp_ == __ninp_)
+          return pbackfail();
+        return traits_type::to_int_type(*--__ninp_);
+    }
+
+    // 27.6.2.2.5 Put area:
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    int_type sputc(char_type __c) {
+        if (__nout_ == __eout_)
+            return overflow(traits_type::to_int_type(__c));
+        *__nout_++ = __c;
+        return traits_type::to_int_type(__c);
+    }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    streamsize sputn(const char_type* __s, streamsize __n)
+    { return xsputn(__s, __n); }
+
+protected:
+    basic_streambuf();
+    basic_streambuf(const basic_streambuf& __rhs);
+    basic_streambuf& operator=(const basic_streambuf& __rhs);
+    void swap(basic_streambuf& __rhs);
+
+    // 27.6.2.3.2 Get area:
+    _LIBCUDACXX_INLINE_VISIBILITY char_type* eback() const {return __binp_;}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type* gptr()  const {return __ninp_;}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type* egptr() const {return __einp_;}
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void gbump(int __n) { __ninp_ += __n; }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void setg(char_type* __gbeg, char_type* __gnext, char_type* __gend) {
+        __binp_ = __gbeg;
+        __ninp_ = __gnext;
+        __einp_ = __gend;
+    }
+
+    // 27.6.2.3.3 Put area:
+    _LIBCUDACXX_INLINE_VISIBILITY char_type* pbase() const {return __bout_;}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type* pptr()  const {return __nout_;}
+    _LIBCUDACXX_INLINE_VISIBILITY char_type* epptr() const {return __eout_;}
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void pbump(int __n) { __nout_ += __n; }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __pbump(streamsize __n) { __nout_ += __n; }
+
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    void setp(char_type* __pbeg, char_type* __pend) {
+        __bout_ = __nout_ = __pbeg;
+        __eout_ = __pend;
+    }
+
+    // 27.6.2.4 virtual functions:
+    // 27.6.2.4.1 Locales:
+    virtual void imbue(const locale& __loc);
+
+    // 27.6.2.4.2 Buffer management and positioning:
+    virtual basic_streambuf* setbuf(char_type* __s, streamsize __n);
+    virtual pos_type seekoff(off_type __off, ios_base::seekdir __way,
+                             ios_base::openmode __which = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type __sp,
+                             ios_base::openmode __which = ios_base::in | ios_base::out);
+    virtual int sync();
+
+    // 27.6.2.4.3 Get area:
+    virtual streamsize showmanyc();
+    virtual streamsize xsgetn(char_type* __s, streamsize __n);
+    virtual int_type underflow();
+    virtual int_type uflow();
+
+    // 27.6.2.4.4 Putback:
+    virtual int_type pbackfail(int_type __c = traits_type::eof());
+
+    // 27.6.2.4.5 Put area:
+    virtual streamsize xsputn(const char_type* __s, streamsize __n);
+    virtual int_type overflow(int_type __c = traits_type::eof());
+
+private:
+    locale __loc_;
+    char_type* __binp_;
+    char_type* __ninp_;
+    char_type* __einp_;
+    char_type* __bout_;
+    char_type* __nout_;
+    char_type* __eout_;
+};
+
+template <class _CharT, class _Traits>
+basic_streambuf<_CharT, _Traits>::~basic_streambuf()
+{
+}
+
+template <class _CharT, class _Traits>
+basic_streambuf<_CharT, _Traits>::basic_streambuf()
+    : __binp_(0),
+      __ninp_(0),
+      __einp_(0),
+      __bout_(0),
+      __nout_(0),
+      __eout_(0)
+{
+}
+
+template <class _CharT, class _Traits>
+basic_streambuf<_CharT, _Traits>::basic_streambuf(const basic_streambuf& __sb)
+    : __loc_(__sb.__loc_),
+      __binp_(__sb.__binp_),
+      __ninp_(__sb.__ninp_),
+      __einp_(__sb.__einp_),
+      __bout_(__sb.__bout_),
+      __nout_(__sb.__nout_),
+      __eout_(__sb.__eout_)
+{
+}
+
+template <class _CharT, class _Traits>
+basic_streambuf<_CharT, _Traits>&
+basic_streambuf<_CharT, _Traits>::operator=(const basic_streambuf& __sb)
+{
+    __loc_ = __sb.__loc_;
+    __binp_ = __sb.__binp_;
+    __ninp_ = __sb.__ninp_;
+    __einp_ = __sb.__einp_;
+    __bout_ = __sb.__bout_;
+    __nout_ = __sb.__nout_;
+    __eout_ = __sb.__eout_;
+    return *this;
+}
+
+template <class _CharT, class _Traits>
+void
+basic_streambuf<_CharT, _Traits>::swap(basic_streambuf& __sb)
+{
+    _CUDA_VSTD::swap(__loc_, __sb.__loc_);
+    _CUDA_VSTD::swap(__binp_, __sb.__binp_);
+    _CUDA_VSTD::swap(__ninp_, __sb.__ninp_);
+    _CUDA_VSTD::swap(__einp_, __sb.__einp_);
+    _CUDA_VSTD::swap(__bout_, __sb.__bout_);
+    _CUDA_VSTD::swap(__nout_, __sb.__nout_);
+    _CUDA_VSTD::swap(__eout_, __sb.__eout_);
+}
+
+template <class _CharT, class _Traits>
+void
+basic_streambuf<_CharT, _Traits>::imbue(const locale&)
+{
+}
+
+template <class _CharT, class _Traits>
+basic_streambuf<_CharT, _Traits>*
+basic_streambuf<_CharT, _Traits>::setbuf(char_type*, streamsize)
+{
+    return this;
+}
+
+template <class _CharT, class _Traits>
+typename basic_streambuf<_CharT, _Traits>::pos_type
+basic_streambuf<_CharT, _Traits>::seekoff(off_type, ios_base::seekdir,
+                                          ios_base::openmode)
+{
+    return pos_type(off_type(-1));
+}
+
+template <class _CharT, class _Traits>
+typename basic_streambuf<_CharT, _Traits>::pos_type
+basic_streambuf<_CharT, _Traits>::seekpos(pos_type, ios_base::openmode)
+{
+    return pos_type(off_type(-1));
+}
+
+template <class _CharT, class _Traits>
+int
+basic_streambuf<_CharT, _Traits>::sync()
+{
+    return 0;
+}
+
+template <class _CharT, class _Traits>
+streamsize
+basic_streambuf<_CharT, _Traits>::showmanyc()
+{
+    return 0;
+}
+
+template <class _CharT, class _Traits>
+streamsize
+basic_streambuf<_CharT, _Traits>::xsgetn(char_type* __s, streamsize __n)
+{
+    const int_type __eof = traits_type::eof();
+    int_type __c;
+    streamsize __i = 0;
+    while(__i < __n)
+    {
+        if (__ninp_ < __einp_)
+        {
+            const streamsize __len = _CUDA_VSTD::min(static_cast<streamsize>(INT_MAX),
+                                _CUDA_VSTD::min(__einp_ - __ninp_, __n - __i));
+            traits_type::copy(__s, __ninp_, __len);
+            __s +=  __len;
+            __i +=  __len;
+            this->gbump(__len);
+        }
+        else if ((__c = uflow()) != __eof)
+        {
+            *__s = traits_type::to_char_type(__c);
+            ++__s;
+            ++__i;
+        }
+        else
+            break;
+    }
+    return __i;
+}
+
+template <class _CharT, class _Traits>
+typename basic_streambuf<_CharT, _Traits>::int_type
+basic_streambuf<_CharT, _Traits>::underflow()
+{
+    return traits_type::eof();
+}
+
+template <class _CharT, class _Traits>
+typename basic_streambuf<_CharT, _Traits>::int_type
+basic_streambuf<_CharT, _Traits>::uflow()
+{
+    if (underflow() == traits_type::eof())
+        return traits_type::eof();
+    return traits_type::to_int_type(*__ninp_++);
+}
+
+template <class _CharT, class _Traits>
+typename basic_streambuf<_CharT, _Traits>::int_type
+basic_streambuf<_CharT, _Traits>::pbackfail(int_type)
+{
+    return traits_type::eof();
+}
+
+template <class _CharT, class _Traits>
+streamsize
+basic_streambuf<_CharT, _Traits>::xsputn(const char_type* __s, streamsize __n)
+{
+    streamsize __i = 0;
+    int_type __eof = traits_type::eof();
+    while( __i < __n)
+    {
+        if (__nout_ >= __eout_)
+        {
+            if (overflow(traits_type::to_int_type(*__s)) == __eof)
+                break;
+            ++__s;
+            ++__i;
+        }
+        else
+        {
+            streamsize __chunk_size = _CUDA_VSTD::min(__eout_ - __nout_, __n - __i);
+            traits_type::copy(__nout_, __s, __chunk_size);
+            __nout_ += __chunk_size;
+            __s     += __chunk_size;
+            __i     += __chunk_size;
+        }
+    }
+    return __i;
+}
+
+template <class _CharT, class _Traits>
+typename basic_streambuf<_CharT, _Traits>::int_type
+basic_streambuf<_CharT, _Traits>::overflow(int_type)
+{
+    return traits_type::eof();
+}
+
+#ifndef _LIBCUDACXX_DO_NOT_ASSUME_STREAMS_EXPLICIT_INSTANTIATION_IN_DYLIB
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_streambuf<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_streambuf<wchar_t>)
+
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_ios<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_ios<wchar_t>)
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_STEAMBUF
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string
new file mode 100644
index 000000000000..b4f6c2c18f12
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string
@@ -0,0 +1,4365 @@
+// -*- C++ -*-
+//===--------------------------- string -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STRING
+#define _LIBCUDACXX_STRING
+
+/*
+    string synopsis
+
+namespace std
+{
+
+template <class stateT>
+class fpos
+{
+private:
+    stateT st;
+public:
+    fpos(streamoff = streamoff());
+
+    operator streamoff() const;
+
+    stateT state() const;
+    void state(stateT);
+
+    fpos& operator+=(streamoff);
+    fpos  operator+ (streamoff) const;
+    fpos& operator-=(streamoff);
+    fpos  operator- (streamoff) const;
+};
+
+template <class stateT> streamoff operator-(const fpos<stateT>& x, const fpos<stateT>& y);
+
+template <class stateT> bool operator==(const fpos<stateT>& x, const fpos<stateT>& y);
+template <class stateT> bool operator!=(const fpos<stateT>& x, const fpos<stateT>& y);
+
+template <class charT>
+struct char_traits
+{
+    typedef charT     char_type;
+    typedef ...       int_type;
+    typedef streamoff off_type;
+    typedef streampos pos_type;
+    typedef mbstate_t state_type;
+
+    static void assign(char_type& c1, const char_type& c2) noexcept;
+    static constexpr bool eq(char_type c1, char_type c2) noexcept;
+    static constexpr bool lt(char_type c1, char_type c2) noexcept;
+
+    static int              compare(const char_type* s1, const char_type* s2, size_t n);
+    static size_t           length(const char_type* s);
+    static const char_type* find(const char_type* s, size_t n, const char_type& a);
+    static char_type*       move(char_type* s1, const char_type* s2, size_t n);
+    static char_type*       copy(char_type* s1, const char_type* s2, size_t n);
+    static char_type*       assign(char_type* s, size_t n, char_type a);
+
+    static constexpr int_type  not_eof(int_type c) noexcept;
+    static constexpr char_type to_char_type(int_type c) noexcept;
+    static constexpr int_type  to_int_type(char_type c) noexcept;
+    static constexpr bool      eq_int_type(int_type c1, int_type c2) noexcept;
+    static constexpr int_type  eof() noexcept;
+};
+
+template <> struct char_traits<char>;
+template <> struct char_traits<wchar_t>;
+
+template<class charT, class traits = char_traits<charT>, class Allocator = allocator<charT> >
+class basic_string
+{
+public:
+// types:
+    typedef traits traits_type;
+    typedef typename traits_type::char_type value_type;
+    typedef Allocator allocator_type;
+    typedef typename allocator_type::size_type size_type;
+    typedef typename allocator_type::difference_type difference_type;
+    typedef typename allocator_type::reference reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef typename allocator_type::pointer pointer;
+    typedef typename allocator_type::const_pointer const_pointer;
+    typedef implementation-defined iterator;
+    typedef implementation-defined const_iterator;
+    typedef std::reverse_iterator<iterator> reverse_iterator;
+    typedef std::reverse_iterator<const_iterator> const_reverse_iterator;
+
+    static const size_type npos = -1;
+
+    basic_string()
+        noexcept(is_nothrow_default_constructible<allocator_type>::value);
+    explicit basic_string(const allocator_type& a);
+    basic_string(const basic_string& str);
+    basic_string(basic_string&& str)
+        noexcept(is_nothrow_move_constructible<allocator_type>::value);
+    basic_string(const basic_string& str, size_type pos,
+                 const allocator_type& a = allocator_type());
+    basic_string(const basic_string& str, size_type pos, size_type n,
+                 const Allocator& a = Allocator());
+    template<class T>
+        basic_string(const T& t, size_type pos, size_type n, const Allocator& a = Allocator()); // C++17
+    template <class T>
+        explicit basic_string(const T& t, const Allocator& a = Allocator()); // C++17
+    basic_string(const value_type* s, const allocator_type& a = allocator_type());
+    basic_string(const value_type* s, size_type n, const allocator_type& a = allocator_type());
+    basic_string(size_type n, value_type c, const allocator_type& a = allocator_type());
+    template<class InputIterator>
+        basic_string(InputIterator begin, InputIterator end,
+                     const allocator_type& a = allocator_type());
+    basic_string(initializer_list<value_type>, const Allocator& = Allocator());
+    basic_string(const basic_string&, const Allocator&);
+    basic_string(basic_string&&, const Allocator&);
+
+    ~basic_string();
+
+    operator basic_string_view<charT, traits>() const noexcept;
+
+    basic_string& operator=(const basic_string& str);
+    template <class T>
+        basic_string& operator=(const T& t); // C++17
+    basic_string& operator=(basic_string&& str)
+        noexcept(
+             allocator_type::propagate_on_container_move_assignment::value ||
+             allocator_type::is_always_equal::value ); // C++17
+    basic_string& operator=(const value_type* s);
+    basic_string& operator=(value_type c);
+    basic_string& operator=(initializer_list<value_type>);
+
+    iterator       begin() noexcept;
+    const_iterator begin() const noexcept;
+    iterator       end() noexcept;
+    const_iterator end() const noexcept;
+
+    reverse_iterator       rbegin() noexcept;
+    const_reverse_iterator rbegin() const noexcept;
+    reverse_iterator       rend() noexcept;
+    const_reverse_iterator rend() const noexcept;
+
+    const_iterator         cbegin() const noexcept;
+    const_iterator         cend() const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend() const noexcept;
+
+    size_type size() const noexcept;
+    size_type length() const noexcept;
+    size_type max_size() const noexcept;
+    size_type capacity() const noexcept;
+
+    void resize(size_type n, value_type c);
+    void resize(size_type n);
+
+    void reserve(size_type res_arg = 0);
+    void shrink_to_fit();
+    void clear() noexcept;
+    bool empty() const noexcept;
+
+    const_reference operator[](size_type pos) const;
+    reference       operator[](size_type pos);
+
+    const_reference at(size_type n) const;
+    reference       at(size_type n);
+
+    basic_string& operator+=(const basic_string& str);
+    template <class T>
+        basic_string& operator+=(const T& t);              // C++17
+    basic_string& operator+=(const value_type* s);
+    basic_string& operator+=(value_type c);
+    basic_string& operator+=(initializer_list<value_type>);
+
+    basic_string& append(const basic_string& str);
+    template <class T>
+        basic_string& append(const T& t);                 // C++17
+    basic_string& append(const basic_string& str, size_type pos, size_type n=npos); //C++14
+    template <class T>
+        basic_string& append(const T& t, size_type pos, size_type n=npos); // C++17
+    basic_string& append(const value_type* s, size_type n);
+    basic_string& append(const value_type* s);
+    basic_string& append(size_type n, value_type c);
+    template<class InputIterator>
+        basic_string& append(InputIterator first, InputIterator last);
+    basic_string& append(initializer_list<value_type>);
+
+    void push_back(value_type c);
+    void pop_back();
+    reference       front();
+    const_reference front() const;
+    reference       back();
+    const_reference back() const;
+
+    basic_string& assign(const basic_string& str);
+    template <class T>
+        basic_string& assign(const T& t);  // C++17
+    basic_string& assign(basic_string&& str);
+    basic_string& assign(const basic_string& str, size_type pos, size_type n=npos); // C++14
+    template <class T>
+        basic_string& assign(const T& t, size_type pos, size_type n=npos); // C++17
+    basic_string& assign(const value_type* s, size_type n);
+    basic_string& assign(const value_type* s);
+    basic_string& assign(size_type n, value_type c);
+    template<class InputIterator>
+        basic_string& assign(InputIterator first, InputIterator last);
+    basic_string& assign(initializer_list<value_type>);
+
+    basic_string& insert(size_type pos1, const basic_string& str);
+    template <class T>
+        basic_string& insert(size_type pos1, const T& t);
+    basic_string& insert(size_type pos1, const basic_string& str,
+                         size_type pos2, size_type n);
+    template <class T>
+        basic_string& insert(size_type pos1, const T& t, size_type pos2, size_type n); // C++17
+    basic_string& insert(size_type pos, const value_type* s, size_type n=npos); //C++14
+    basic_string& insert(size_type pos, const value_type* s);
+    basic_string& insert(size_type pos, size_type n, value_type c);
+    iterator      insert(const_iterator p, value_type c);
+    iterator      insert(const_iterator p, size_type n, value_type c);
+    template<class InputIterator>
+        iterator insert(const_iterator p, InputIterator first, InputIterator last);
+    iterator      insert(const_iterator p, initializer_list<value_type>);
+
+    basic_string& erase(size_type pos = 0, size_type n = npos);
+    iterator      erase(const_iterator position);
+    iterator      erase(const_iterator first, const_iterator last);
+
+    basic_string& replace(size_type pos1, size_type n1, const basic_string& str);
+    template <class T>
+    basic_string& replace(size_type pos1, size_type n1, const T& t);  // C++17
+    basic_string& replace(size_type pos1, size_type n1, const basic_string& str,
+                          size_type pos2, size_type n2=npos); // C++14
+    template <class T>
+        basic_string& replace(size_type pos1, size_type n1, const T& t,
+                              size_type pos2, size_type n); // C++17
+    basic_string& replace(size_type pos, size_type n1, const value_type* s, size_type n2);
+    basic_string& replace(size_type pos, size_type n1, const value_type* s);
+    basic_string& replace(size_type pos, size_type n1, size_type n2, value_type c);
+    basic_string& replace(const_iterator i1, const_iterator i2, const basic_string& str);
+    template <class T>
+        basic_string& replace(const_iterator i1, const_iterator i2, const T& t);  // C++17
+    basic_string& replace(const_iterator i1, const_iterator i2, const value_type* s, size_type n);
+    basic_string& replace(const_iterator i1, const_iterator i2, const value_type* s);
+    basic_string& replace(const_iterator i1, const_iterator i2, size_type n, value_type c);
+    template<class InputIterator>
+        basic_string& replace(const_iterator i1, const_iterator i2, InputIterator j1, InputIterator j2);
+    basic_string& replace(const_iterator i1, const_iterator i2, initializer_list<value_type>);
+
+    size_type copy(value_type* s, size_type n, size_type pos = 0) const;
+    basic_string substr(size_type pos = 0, size_type n = npos) const;
+
+    void swap(basic_string& str)
+        noexcept(allocator_traits<allocator_type>::propagate_on_container_swap::value ||
+                 allocator_traits<allocator_type>::is_always_equal::value);  // C++17
+
+    const value_type* c_str() const noexcept;
+    const value_type* data() const noexcept;
+          value_type* data()       noexcept;   // C++17
+
+    allocator_type get_allocator() const noexcept;
+
+    size_type find(const basic_string& str, size_type pos = 0) const noexcept;
+    template <class T>
+        size_type find(const T& t, size_type pos = 0) const;  // C++17
+    size_type find(const value_type* s, size_type pos, size_type n) const noexcept;
+    size_type find(const value_type* s, size_type pos = 0) const noexcept;
+    size_type find(value_type c, size_type pos = 0) const noexcept;
+
+    size_type rfind(const basic_string& str, size_type pos = npos) const noexcept;
+    template <class T>
+        size_type rfind(const T& t, size_type pos = npos) const;  // C++17
+    size_type rfind(const value_type* s, size_type pos, size_type n) const noexcept;
+    size_type rfind(const value_type* s, size_type pos = npos) const noexcept;
+    size_type rfind(value_type c, size_type pos = npos) const noexcept;
+
+    size_type find_first_of(const basic_string& str, size_type pos = 0) const noexcept;
+    template <class T>
+        size_type find_first_of(const T& t, size_type pos = 0) const; // C++17
+    size_type find_first_of(const value_type* s, size_type pos, size_type n) const noexcept;
+    size_type find_first_of(const value_type* s, size_type pos = 0) const noexcept;
+    size_type find_first_of(value_type c, size_type pos = 0) const noexcept;
+
+    size_type find_last_of(const basic_string& str, size_type pos = npos) const noexcept;
+    template <class T>
+        size_type find_last_of(const T& t, size_type pos = npos) const noexcept;  // C++17
+    size_type find_last_of(const value_type* s, size_type pos, size_type n) const noexcept;
+    size_type find_last_of(const value_type* s, size_type pos = npos) const noexcept;
+    size_type find_last_of(value_type c, size_type pos = npos) const noexcept;
+
+    size_type find_first_not_of(const basic_string& str, size_type pos = 0) const noexcept;
+    template <class T>
+        size_type find_first_not_of(const T& t, size_type pos = 0) const; // C++17
+    size_type find_first_not_of(const value_type* s, size_type pos, size_type n) const noexcept;
+    size_type find_first_not_of(const value_type* s, size_type pos = 0) const noexcept;
+    size_type find_first_not_of(value_type c, size_type pos = 0) const noexcept;
+
+    size_type find_last_not_of(const basic_string& str, size_type pos = npos) const noexcept;
+    template <class T>
+        size_type find_last_not_of(const T& t, size_type pos = npos) const; // C++17
+    size_type find_last_not_of(const value_type* s, size_type pos, size_type n) const noexcept;
+    size_type find_last_not_of(const value_type* s, size_type pos = npos) const noexcept;
+    size_type find_last_not_of(value_type c, size_type pos = npos) const noexcept;
+
+    int compare(const basic_string& str) const noexcept;
+    template <class T>
+        int compare(const T& t) const noexcept;  // C++17
+    int compare(size_type pos1, size_type n1, const basic_string& str) const;
+    template <class T>
+        int compare(size_type pos1, size_type n1, const T& t) const;  // C++17
+    int compare(size_type pos1, size_type n1, const basic_string& str,
+                size_type pos2, size_type n2=npos) const; // C++14
+    template <class T>
+        int compare(size_type pos1, size_type n1, const T& t,
+                    size_type pos2, size_type n2=npos) const; // C++17
+    int compare(const value_type* s) const noexcept;
+    int compare(size_type pos1, size_type n1, const value_type* s) const;
+    int compare(size_type pos1, size_type n1, const value_type* s, size_type n2) const;
+
+    bool starts_with(basic_string_view<charT, traits> sv) const noexcept; // C++2a
+    bool starts_with(charT c) const noexcept;                             // C++2a
+    bool starts_with(const charT* s) const;                               // C++2a
+    bool ends_with(basic_string_view<charT, traits> sv) const noexcept;   // C++2a
+    bool ends_with(charT c) const noexcept;                               // C++2a
+    bool ends_with(const charT* s) const;                                 // C++2a
+
+    bool __invariants() const;
+};
+
+template<class InputIterator,
+         class Allocator = allocator<typename iterator_traits<InputIterator>::value_type>>
+basic_string(InputIterator, InputIterator, Allocator = Allocator())
+   -> basic_string<typename iterator_traits<InputIterator>::value_type,
+                  char_traits<typename iterator_traits<InputIterator>::value_type>,
+                  Allocator>;   // C++17
+
+template<class charT, class traits, class Allocator>
+basic_string<charT, traits, Allocator>
+operator+(const basic_string<charT, traits, Allocator>& lhs,
+          const basic_string<charT, traits, Allocator>& rhs);
+
+template<class charT, class traits, class Allocator>
+basic_string<charT, traits, Allocator>
+operator+(const charT* lhs , const basic_string<charT,traits,Allocator>&rhs);
+
+template<class charT, class traits, class Allocator>
+basic_string<charT, traits, Allocator>
+operator+(charT lhs, const basic_string<charT,traits,Allocator>& rhs);
+
+template<class charT, class traits, class Allocator>
+basic_string<charT, traits, Allocator>
+operator+(const basic_string<charT, traits, Allocator>& lhs, const charT* rhs);
+
+template<class charT, class traits, class Allocator>
+basic_string<charT, traits, Allocator>
+operator+(const basic_string<charT, traits, Allocator>& lhs, charT rhs);
+
+template<class charT, class traits, class Allocator>
+bool operator==(const basic_string<charT, traits, Allocator>& lhs,
+                const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator==(const charT* lhs, const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator==(const basic_string<charT,traits,Allocator>& lhs, const charT* rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator!=(const basic_string<charT,traits,Allocator>& lhs,
+                const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator!=(const charT* lhs, const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator!=(const basic_string<charT, traits, Allocator>& lhs, const charT* rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator< (const basic_string<charT, traits, Allocator>& lhs,
+                const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator< (const basic_string<charT, traits, Allocator>& lhs, const charT* rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator< (const charT* lhs, const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator> (const basic_string<charT, traits, Allocator>& lhs,
+                const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator> (const basic_string<charT, traits, Allocator>& lhs, const charT* rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator> (const charT* lhs, const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator<=(const basic_string<charT, traits, Allocator>& lhs,
+                const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator<=(const basic_string<charT, traits, Allocator>& lhs, const charT* rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator<=(const charT* lhs, const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator>=(const basic_string<charT, traits, Allocator>& lhs,
+                const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator>=(const basic_string<charT, traits, Allocator>& lhs, const charT* rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+bool operator>=(const charT* lhs, const basic_string<charT, traits, Allocator>& rhs) noexcept;
+
+template<class charT, class traits, class Allocator>
+void swap(basic_string<charT, traits, Allocator>& lhs,
+          basic_string<charT, traits, Allocator>& rhs)
+            noexcept(noexcept(lhs.swap(rhs)));
+
+template<class charT, class traits, class Allocator>
+basic_istream<charT, traits>&
+operator>>(basic_istream<charT, traits>& is, basic_string<charT, traits, Allocator>& str);
+
+template<class charT, class traits, class Allocator>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& os, const basic_string<charT, traits, Allocator>& str);
+
+template<class charT, class traits, class Allocator>
+basic_istream<charT, traits>&
+getline(basic_istream<charT, traits>& is, basic_string<charT, traits, Allocator>& str,
+        charT delim);
+
+template<class charT, class traits, class Allocator>
+basic_istream<charT, traits>&
+getline(basic_istream<charT, traits>& is, basic_string<charT, traits, Allocator>& str);
+
+template<class charT, class traits, class Allocator, class U>
+void erase(basic_string<charT, traits, Allocator>& c, const U& value); // C++20
+template<class charT, class traits, class Allocator, class Predicate>
+void erase_if(basic_string<charT, traits, Allocator>& c, Predicate pred); // C++20
+
+typedef basic_string<char>    string;
+typedef basic_string<wchar_t> wstring;
+typedef basic_string<char16_t> u16string;
+typedef basic_string<char32_t> u32string;
+
+int                stoi  (const string& str, size_t* idx = 0, int base = 10);
+long               stol  (const string& str, size_t* idx = 0, int base = 10);
+unsigned long      stoul (const string& str, size_t* idx = 0, int base = 10);
+long long          stoll (const string& str, size_t* idx = 0, int base = 10);
+unsigned long long stoull(const string& str, size_t* idx = 0, int base = 10);
+
+float       stof (const string& str, size_t* idx = 0);
+double      stod (const string& str, size_t* idx = 0);
+long double stold(const string& str, size_t* idx = 0);
+
+string to_string(int val);
+string to_string(unsigned val);
+string to_string(long val);
+string to_string(unsigned long val);
+string to_string(long long val);
+string to_string(unsigned long long val);
+string to_string(float val);
+string to_string(double val);
+string to_string(long double val);
+
+int                stoi  (const wstring& str, size_t* idx = 0, int base = 10);
+long               stol  (const wstring& str, size_t* idx = 0, int base = 10);
+unsigned long      stoul (const wstring& str, size_t* idx = 0, int base = 10);
+long long          stoll (const wstring& str, size_t* idx = 0, int base = 10);
+unsigned long long stoull(const wstring& str, size_t* idx = 0, int base = 10);
+
+float       stof (const wstring& str, size_t* idx = 0);
+double      stod (const wstring& str, size_t* idx = 0);
+long double stold(const wstring& str, size_t* idx = 0);
+
+wstring to_wstring(int val);
+wstring to_wstring(unsigned val);
+wstring to_wstring(long val);
+wstring to_wstring(unsigned long val);
+wstring to_wstring(long long val);
+wstring to_wstring(unsigned long long val);
+wstring to_wstring(float val);
+wstring to_wstring(double val);
+wstring to_wstring(long double val);
+
+template <> struct hash<string>;
+template <> struct hash<u16string>;
+template <> struct hash<u32string>;
+template <> struct hash<wstring>;
+
+basic_string<char>     operator "" s( const char *str,     size_t len ); // C++14
+basic_string<wchar_t>  operator "" s( const wchar_t *str,  size_t len ); // C++14
+basic_string<char16_t> operator "" s( const char16_t *str, size_t len ); // C++14
+basic_string<char32_t> operator "" s( const char32_t *str, size_t len ); // C++14
+
+}  // std
+
+*/
+
+#include <__config>
+#include <string_view>
+#include <iosfwd>
+#include <cstring>
+#include <cstdio>  // For EOF.
+#include <cwchar>
+#include <algorithm>
+#include <iterator>
+#include <utility>
+#include <memory>
+#include <stdexcept>
+#include <type_traits>
+#include <initializer_list>
+#include <__functional_base>
+#include <version>
+#ifndef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+#include <cstdint>
+#endif
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// fpos
+
+template <class _StateT>
+class _LIBCUDACXX_TEMPLATE_VIS fpos
+{
+private:
+    _StateT __st_;
+    streamoff __off_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY fpos(streamoff __off = streamoff()) : __st_(), __off_(__off) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY operator streamoff() const {return __off_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY _StateT state() const {return __st_;}
+    _LIBCUDACXX_INLINE_VISIBILITY void state(_StateT __st) {__st_ = __st;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY fpos& operator+=(streamoff __off) {__off_ += __off; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY fpos  operator+ (streamoff __off) const {fpos __t(*this); __t += __off; return __t;}
+    _LIBCUDACXX_INLINE_VISIBILITY fpos& operator-=(streamoff __off) {__off_ -= __off; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY fpos  operator- (streamoff __off) const {fpos __t(*this); __t -= __off; return __t;}
+};
+
+template <class _StateT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+streamoff operator-(const fpos<_StateT>& __x, const fpos<_StateT>& __y)
+    {return streamoff(__x) - streamoff(__y);}
+
+template <class _StateT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(const fpos<_StateT>& __x, const fpos<_StateT>& __y)
+    {return streamoff(__x) == streamoff(__y);}
+
+template <class _StateT>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(const fpos<_StateT>& __x, const fpos<_StateT>& __y)
+    {return streamoff(__x) != streamoff(__y);}
+
+// basic_string
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __x,
+          const basic_string<_CharT, _Traits, _Allocator>& __y);
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const _CharT* __x, const basic_string<_CharT,_Traits,_Allocator>& __y);
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(_CharT __x, const basic_string<_CharT,_Traits,_Allocator>& __y);
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __x, const _CharT* __y);
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __x, _CharT __y);
+
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS string operator+<char, char_traits<char>, allocator<char> >(char const*, string const&))
+
+template <bool>
+class _LIBCUDACXX_TEMPLATE_VIS __basic_string_common
+{
+protected:
+    _LIBCUDACXX_NORETURN void __throw_length_error() const;
+    _LIBCUDACXX_NORETURN void __throw_out_of_range() const;
+};
+
+template <bool __b>
+void
+__basic_string_common<__b>::__throw_length_error() const
+{
+    _CUDA_VSTD::__throw_length_error("basic_string");
+}
+
+template <bool __b>
+void
+__basic_string_common<__b>::__throw_out_of_range() const
+{
+    _CUDA_VSTD::__throw_out_of_range("basic_string");
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __basic_string_common<true>)
+
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+template <class _Iter>
+struct __libcpp_string_gets_noexcept_iterator_impl : public true_type {};
+#elif defined(_LIBCUDACXX_HAS_NO_NOEXCEPT)
+template <class _Iter>
+struct __libcpp_string_gets_noexcept_iterator_impl : public false_type {};
+#else
+template <class _Iter, bool = __is_forward_iterator<_Iter>::value>
+struct __libcpp_string_gets_noexcept_iterator_impl : public _LIBCUDACXX_BOOL_CONSTANT((
+    noexcept(++(declval<_Iter&>())) &&
+    is_nothrow_assignable<_Iter&, _Iter>::value &&
+    noexcept(declval<_Iter>() == declval<_Iter>()) &&
+    noexcept(*declval<_Iter>())
+)) {};
+
+template <class _Iter>
+struct __libcpp_string_gets_noexcept_iterator_impl<_Iter, false> : public false_type {};
+#endif
+
+
+template <class _Iter>
+struct __libcpp_string_gets_noexcept_iterator
+    : public _LIBCUDACXX_BOOL_CONSTANT(__libcpp_is_trivial_iterator<_Iter>::value || __libcpp_string_gets_noexcept_iterator_impl<_Iter>::value) {};
+
+template <class _CharT, class _Traits, class _Tp>
+struct __can_be_converted_to_string_view : public _LIBCUDACXX_BOOL_CONSTANT(
+    ( is_convertible<const _Tp&, basic_string_view<_CharT, _Traits> >::value &&
+     !is_convertible<const _Tp&, const _CharT*>::value)) {};
+
+#ifdef _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+template <class _CharT, size_t = sizeof(_CharT)>
+struct __padding
+{
+    unsigned char __xx[sizeof(_CharT)-1];
+};
+
+template <class _CharT>
+struct __padding<_CharT, 1>
+{
+};
+
+#endif  // _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+template<class _CharT, class _Traits, class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS basic_string
+    : private __basic_string_common<true>
+{
+public:
+    typedef basic_string                                 __self;
+    typedef basic_string_view<_CharT, _Traits>           __self_view;
+    typedef _Traits                                      traits_type;
+    typedef _CharT                                       value_type;
+    typedef _Allocator                                   allocator_type;
+    typedef allocator_traits<allocator_type>             __alloc_traits;
+    typedef typename __alloc_traits::size_type           size_type;
+    typedef typename __alloc_traits::difference_type     difference_type;
+    typedef value_type&                                  reference;
+    typedef const value_type&                            const_reference;
+    typedef typename __alloc_traits::pointer             pointer;
+    typedef typename __alloc_traits::const_pointer       const_pointer;
+
+    static_assert((!is_array<value_type>::value), "Character type of basic_string must not be an array");
+    static_assert(( is_standard_layout<value_type>::value), "Character type of basic_string must be standard-layout");
+    static_assert(( is_trivial<value_type>::value), "Character type of basic_string must be trivial");
+    static_assert(( is_same<_CharT, typename traits_type::char_type>::value),
+                  "traits_type::char_type must be the same type as CharT");
+    static_assert(( is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+#if defined(_LIBCUDACXX_RAW_ITERATORS)
+    typedef pointer                                      iterator;
+    typedef const_pointer                                const_iterator;
+#else  // defined(_LIBCUDACXX_RAW_ITERATORS)
+    typedef __wrap_iter<pointer>                         iterator;
+    typedef __wrap_iter<const_pointer>                   const_iterator;
+#endif  // defined(_LIBCUDACXX_RAW_ITERATORS)
+    typedef _CUDA_VSTD::reverse_iterator<iterator>             reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>       const_reverse_iterator;
+
+private:
+
+#ifdef _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+    struct __long
+    {
+        pointer   __data_;
+        size_type __size_;
+        size_type __cap_;
+    };
+
+#ifdef _LIBCUDACXX_BIG_ENDIAN
+    static const size_type __short_mask = 0x01;
+    static const size_type __long_mask  = 0x1ul;
+#else  // _LIBCUDACXX_BIG_ENDIAN
+    static const size_type __short_mask = 0x80;
+    static const size_type __long_mask  = ~(size_type(~0) >> 1);
+#endif  // _LIBCUDACXX_BIG_ENDIAN
+
+    enum {__min_cap = (sizeof(__long) - 1)/sizeof(value_type) > 2 ?
+                      (sizeof(__long) - 1)/sizeof(value_type) : 2};
+
+    struct __short
+    {
+        value_type __data_[__min_cap];
+        struct
+            : __padding<value_type>
+        {
+            unsigned char __size_;
+        };
+    };
+
+#else
+
+    struct __long
+    {
+        size_type __cap_;
+        size_type __size_;
+        pointer   __data_;
+    };
+
+#ifdef _LIBCUDACXX_BIG_ENDIAN
+    static const size_type __short_mask = 0x80;
+    static const size_type __long_mask  = ~(size_type(~0) >> 1);
+#else  // _LIBCUDACXX_BIG_ENDIAN
+    static const size_type __short_mask = 0x01;
+    static const size_type __long_mask  = 0x1ul;
+#endif  // _LIBCUDACXX_BIG_ENDIAN
+
+    enum {__min_cap = (sizeof(__long) - 1)/sizeof(value_type) > 2 ?
+                      (sizeof(__long) - 1)/sizeof(value_type) : 2};
+
+    struct __short
+    {
+        union
+        {
+            unsigned char __size_;
+            value_type __lx;
+        };
+        value_type __data_[__min_cap];
+    };
+
+#endif  // _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+    union __ulx{__long __lx; __short __lxx;};
+
+    enum {__n_words = sizeof(__ulx) / sizeof(size_type)};
+
+    struct __raw
+    {
+        size_type __words[__n_words];
+    };
+
+    struct __rep
+    {
+        union
+        {
+            __long  __l;
+            __short __s;
+            __raw   __r;
+        };
+    };
+
+    __compressed_pair<__rep, allocator_type> __r_;
+
+public:
+    static const size_type npos = static_cast<size_type>(-1);
+
+    _LIBCUDACXX_INLINE_VISIBILITY basic_string()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value);
+
+    _LIBCUDACXX_INLINE_VISIBILITY explicit basic_string(const allocator_type& __a)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_copy_constructible<allocator_type>::value);
+#else
+        _NOEXCEPT;
+#endif
+
+    basic_string(const basic_string& __str);
+    basic_string(const basic_string& __str, const allocator_type& __a);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(basic_string&& __str)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value);
+#else
+        _NOEXCEPT;
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(basic_string&& __str, const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template <class = typename enable_if<__is_allocator<_Allocator>::value, nullptr_t>::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(const _CharT* __s) {
+      _LIBCUDACXX_ASSERT(__s != nullptr, "basic_string(const char*) detected nullptr");
+      __init(__s, traits_type::length(__s));
+#   if _LIBCUDACXX_DEBUG_LEVEL >= 2
+      __get_db()->__insert_c(this);
+#   endif
+    }
+
+    template <class = typename enable_if<__is_allocator<_Allocator>::value, nullptr_t>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string(const _CharT* __s, const _Allocator& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(const _CharT* __s, size_type __n);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(const _CharT* __s, size_type __n, const _Allocator& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(size_type __n, _CharT __c);
+
+    template <class = typename enable_if<__is_allocator<_Allocator>::value, nullptr_t>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string(size_type __n, _CharT __c, const _Allocator& __a);
+
+    basic_string(const basic_string& __str, size_type __pos, size_type __n,
+                 const _Allocator& __a = _Allocator());
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(const basic_string& __str, size_type __pos,
+                 const _Allocator& __a = _Allocator());
+
+    template<class _Tp, class = typename enable_if<__can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value, void>::type>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        basic_string(const _Tp& __t, size_type __pos, size_type __n,
+                              const allocator_type& __a = allocator_type());
+
+    template<class _Tp, class = typename enable_if<__can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value, void>::type>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        explicit basic_string(const _Tp& __t);
+
+    template<class _Tp, class = typename enable_if<__can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value, void>::type>
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        explicit basic_string(const _Tp& __t, const allocator_type& __a);
+
+    template<class _InputIterator, class = typename enable_if<__is_input_iterator<_InputIterator>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string(_InputIterator __first, _InputIterator __last);
+    template<class _InputIterator, class = typename enable_if<__is_input_iterator<_InputIterator>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        basic_string(_InputIterator __first, _InputIterator __last, const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(initializer_list<_CharT> __il);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string(initializer_list<_CharT> __il, const _Allocator& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    inline ~basic_string();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator __self_view() const _NOEXCEPT { return __self_view(data(), size()); }
+
+    basic_string& operator=(const basic_string& __str);
+
+    template <class _Tp, class = typename enable_if<__can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value, void>::type>
+    basic_string& operator=(const _Tp& __t)
+        {__self_view __sv = __t; return assign(__sv);}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& operator=(basic_string&& __str)
+        _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value));
+     _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& operator=(initializer_list<value_type> __il) {return assign(__il.begin(), __il.size());}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY basic_string& operator=(const value_type* __s) {return assign(__s);}
+    basic_string& operator=(value_type __c);
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator begin() _NOEXCEPT
+        {return iterator(this, __get_pointer());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT
+        {return const_iterator(this, __get_pointer());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator end() _NOEXCEPT
+        {return iterator(this, __get_pointer() + size());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT
+        {return const_iterator(this, __get_pointer() + size());}
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator begin() _NOEXCEPT
+        {return iterator(__get_pointer());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT
+        {return const_iterator(__get_pointer());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator end() _NOEXCEPT
+        {return iterator(__get_pointer() + size());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end() const _NOEXCEPT
+        {return const_iterator(__get_pointer() + size());}
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator rbegin() _NOEXCEPT
+        {return reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator rend() _NOEXCEPT
+        {return reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend() const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT
+        {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend() const _NOEXCEPT
+        {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT
+        {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend() const _NOEXCEPT
+        {return rend();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY size_type size() const _NOEXCEPT
+        {return __is_long() ? __get_long_size() : __get_short_size();}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type length() const _NOEXCEPT {return size();}
+    _LIBCUDACXX_INLINE_VISIBILITY size_type max_size() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY size_type capacity() const _NOEXCEPT
+        {return (__is_long() ? __get_long_cap()
+                             : static_cast<size_type>(__min_cap)) - 1;}
+
+    void resize(size_type __n, value_type __c);
+    _LIBCUDACXX_INLINE_VISIBILITY void resize(size_type __n) {resize(__n, value_type());}
+
+    void reserve(size_type __res_arg);
+    _LIBCUDACXX_INLINE_VISIBILITY void __resize_default_init(size_type __n);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reserve() _NOEXCEPT {reserve(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void shrink_to_fit() _NOEXCEPT {reserve();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT;
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT {return size() == 0;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference operator[](size_type __pos) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY reference       operator[](size_type __pos)       _NOEXCEPT;
+
+    const_reference at(size_type __n) const;
+    reference       at(size_type __n);
+
+    _LIBCUDACXX_INLINE_VISIBILITY basic_string& operator+=(const basic_string& __str) {return append(__str);}
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                                            operator+=(const _Tp& __t)            {__self_view __sv = __t; return append(__sv);}
+    _LIBCUDACXX_INLINE_VISIBILITY basic_string& operator+=(const value_type* __s)     {return append(__s);}
+    _LIBCUDACXX_INLINE_VISIBILITY basic_string& operator+=(value_type __c)            {push_back(__c); return *this;}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY basic_string& operator+=(initializer_list<value_type> __il) {return append(__il);}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& append(const basic_string& __str);
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  append(const _Tp& __t) { __self_view __sv = __t; return append(__sv.data(), __sv.size()); }
+    basic_string& append(const basic_string& __str, size_type __pos, size_type __n=npos);
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  append(const _Tp& __t, size_type __pos, size_type __n=npos);
+    basic_string& append(const value_type* __s, size_type __n);
+    basic_string& append(const value_type* __s);
+    basic_string& append(size_type __n, value_type __c);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __append_default_init(size_type __n);
+
+    template <class _ForwardIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    basic_string& __append_forward_unsafe(_ForwardIterator, _ForwardIterator);
+    template<class _InputIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __is_exactly_input_iterator<_InputIterator>::value
+                || !__libcpp_string_gets_noexcept_iterator<_InputIterator>::value,
+            basic_string&
+        >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    append(_InputIterator __first, _InputIterator __last) {
+      const basic_string __temp (__first, __last, __alloc());
+      append(__temp.data(), __temp.size());
+      return *this;
+    }
+    template<class _ForwardIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value
+                && __libcpp_string_gets_noexcept_iterator<_ForwardIterator>::value,
+            basic_string&
+        >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    append(_ForwardIterator __first, _ForwardIterator __last) {
+      return __append_forward_unsafe(__first, __last);
+    }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& append(initializer_list<value_type> __il) {return append(__il.begin(), __il.size());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    void push_back(value_type __c);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pop_back();
+    _LIBCUDACXX_INLINE_VISIBILITY reference       front() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference front() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY reference       back() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference back() const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                 assign(const _Tp & __t) { __self_view __sv = __t; return assign(__sv.data(), __sv.size()); }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& assign(const basic_string& __str) { return *this = __str; }
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& assign(basic_string&& __str)
+        _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value))
+        {*this = _CUDA_VSTD::move(__str); return *this;}
+#endif
+    basic_string& assign(const basic_string& __str, size_type __pos, size_type __n=npos);
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  assign(const _Tp & __t, size_type __pos, size_type __n=npos);
+    basic_string& assign(const value_type* __s, size_type __n);
+    basic_string& assign(const value_type* __s);
+    basic_string& assign(size_type __n, value_type __c);
+    template<class _InputIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+           __is_exactly_input_iterator<_InputIterator>::value
+                || !__libcpp_string_gets_noexcept_iterator<_InputIterator>::value,
+            basic_string&
+        >::type
+        assign(_InputIterator __first, _InputIterator __last);
+    template<class _ForwardIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value
+                 && __libcpp_string_gets_noexcept_iterator<_ForwardIterator>::value,
+            basic_string&
+        >::type
+        assign(_ForwardIterator __first, _ForwardIterator __last);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& assign(initializer_list<value_type> __il) {return assign(__il.begin(), __il.size());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& insert(size_type __pos1, const basic_string& __str);
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                 insert(size_type __pos1, const _Tp& __t)
+    { __self_view __sv = __t; return insert(__pos1, __sv.data(), __sv.size()); }
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  insert(size_type __pos1, const _Tp& __t, size_type __pos2, size_type __n=npos);
+    basic_string& insert(size_type __pos1, const basic_string& __str, size_type __pos2, size_type __n=npos);
+    basic_string& insert(size_type __pos, const value_type* __s, size_type __n);
+    basic_string& insert(size_type __pos, const value_type* __s);
+    basic_string& insert(size_type __pos, size_type __n, value_type __c);
+    iterator      insert(const_iterator __pos, value_type __c);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator      insert(const_iterator __pos, size_type __n, value_type __c);
+    template<class _InputIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+           __is_exactly_input_iterator<_InputIterator>::value
+                || !__libcpp_string_gets_noexcept_iterator<_InputIterator>::value,
+            iterator
+        >::type
+        insert(const_iterator __pos, _InputIterator __first, _InputIterator __last);
+    template<class _ForwardIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value
+                 && __libcpp_string_gets_noexcept_iterator<_ForwardIterator>::value,
+            iterator
+        >::type
+        insert(const_iterator __pos, _ForwardIterator __first, _ForwardIterator __last);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __pos, initializer_list<value_type> __il)
+                    {return insert(__pos, __il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    basic_string& erase(size_type __pos = 0, size_type __n = npos);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator      erase(const_iterator __pos);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator      erase(const_iterator __first, const_iterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& replace(size_type __pos1, size_type __n1, const basic_string& __str);
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  replace(size_type __pos1, size_type __n1, const _Tp& __t) { __self_view __sv = __t; return replace(__pos1, __n1, __sv.data(), __sv.size()); }
+    basic_string& replace(size_type __pos1, size_type __n1, const basic_string& __str, size_type __pos2, size_type __n2=npos);
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  replace(size_type __pos1, size_type __n1, const _Tp& __t, size_type __pos2, size_type __n2=npos);
+    basic_string& replace(size_type __pos, size_type __n1, const value_type* __s, size_type __n2);
+    basic_string& replace(size_type __pos, size_type __n1, const value_type* __s);
+    basic_string& replace(size_type __pos, size_type __n1, size_type __n2, value_type __c);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& replace(const_iterator __i1, const_iterator __i2, const basic_string& __str);
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            basic_string&
+        >::type
+                  replace(const_iterator __i1, const_iterator __i2, const _Tp& __t) { __self_view __sv = __t; return replace(__i1 - begin(), __i2 - __i1, __sv); }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& replace(const_iterator __i1, const_iterator __i2, const value_type* __s, size_type __n);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& replace(const_iterator __i1, const_iterator __i2, const value_type* __s);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& replace(const_iterator __i1, const_iterator __i2, size_type __n, value_type __c);
+    template<class _InputIterator>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __is_input_iterator<_InputIterator>::value,
+            basic_string&
+        >::type
+        replace(const_iterator __i1, const_iterator __i2, _InputIterator __j1, _InputIterator __j2);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string& replace(const_iterator __i1, const_iterator __i2, initializer_list<value_type> __il)
+        {return replace(__i1, __i2, __il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    size_type copy(value_type* __s, size_type __n, size_type __pos = 0) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string substr(size_type __pos = 0, size_type __n = npos) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_string& __str)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type* c_str() const _NOEXCEPT {return data();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type* data() const _NOEXCEPT  {return _CUDA_VSTD::__to_raw_pointer(__get_pointer());}
+#if _LIBCUDACXX_STD_VER > 14 || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type* data()             _NOEXCEPT  {return _CUDA_VSTD::__to_raw_pointer(__get_pointer());}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT {return __alloc();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find(const basic_string& __str, size_type __pos = 0) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            size_type
+        >::type
+              find(const _Tp& __t, size_type __pos = 0) const;
+    size_type find(const value_type* __s, size_type __pos, size_type __n) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find(const value_type* __s, size_type __pos = 0) const _NOEXCEPT;
+    size_type find(value_type __c, size_type __pos = 0) const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type rfind(const basic_string& __str, size_type __pos = npos) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            size_type
+        >::type
+              rfind(const _Tp& __t, size_type __pos = npos) const;
+    size_type rfind(const value_type* __s, size_type __pos, size_type __n) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type rfind(const value_type* __s, size_type __pos = npos) const _NOEXCEPT;
+    size_type rfind(value_type __c, size_type __pos = npos) const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(const basic_string& __str, size_type __pos = 0) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            size_type
+        >::type
+              find_first_of(const _Tp& __t, size_type __pos = 0) const;
+    size_type find_first_of(const value_type* __s, size_type __pos, size_type __n) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(const value_type* __s, size_type __pos = 0) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(value_type __c, size_type __pos = 0) const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(const basic_string& __str, size_type __pos = npos) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            size_type
+        >::type
+              find_last_of(const _Tp& __t, size_type __pos = npos) const;
+    size_type find_last_of(const value_type* __s, size_type __pos, size_type __n) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(const value_type* __s, size_type __pos = npos) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(value_type __c, size_type __pos = npos) const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(const basic_string& __str, size_type __pos = 0) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            size_type
+        >::type
+              find_first_not_of(const _Tp &__t, size_type __pos = 0) const;
+    size_type find_first_not_of(const value_type* __s, size_type __pos, size_type __n) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(const value_type* __s, size_type __pos = 0) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(value_type __c, size_type __pos = 0) const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(const basic_string& __str, size_type __pos = npos) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            size_type
+        >::type
+              find_last_not_of(const _Tp& __t, size_type __pos = npos) const;
+    size_type find_last_not_of(const value_type* __s, size_type __pos, size_type __n) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(const value_type* __s, size_type __pos = npos) const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(value_type __c, size_type __pos = npos) const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(const basic_string& __str) const _NOEXCEPT;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            int
+        >::type
+        compare(const _Tp &__t) const;
+
+    template <class _Tp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            int
+        >::type
+         compare(size_type __pos1, size_type __n1, const _Tp& __t) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(size_type __pos1, size_type __n1, const basic_string& __str) const;
+    int compare(size_type __pos1, size_type __n1, const basic_string& __str, size_type __pos2, size_type __n2=npos) const;
+
+    template <class _Tp>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+            int
+        >::type
+        compare(size_type __pos1, size_type __n1, const _Tp& __t, size_type __pos2, size_type __n2=npos) const;
+    int compare(const value_type* __s) const _NOEXCEPT;
+    int compare(size_type __pos1, size_type __n1, const value_type* __s) const;
+    int compare(size_type __pos1, size_type __n1, const value_type* __s, size_type __n2) const;
+
+#if _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool starts_with(__self_view __sv) const _NOEXCEPT
+    { return __self_view(data(), size()).starts_with(__sv); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool starts_with(value_type __c) const _NOEXCEPT
+    { return !empty() && _Traits::eq(front(), __c); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool starts_with(const value_type* __s) const _NOEXCEPT
+    { return starts_with(__self_view(__s)); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool ends_with(__self_view __sv) const _NOEXCEPT
+    { return __self_view(data(), size()).ends_with( __sv); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool ends_with(value_type __c) const _NOEXCEPT
+    { return !empty() && _Traits::eq(back(), __c); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool ends_with(const value_type* __s) const _NOEXCEPT
+    { return ends_with(__self_view(__s)); }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY bool __invariants() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY void __clear_and_shrink() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool __is_long() const _NOEXCEPT
+        {return bool(__r_.first().__s.__size_ & __short_mask);}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const;
+    bool __decrementable(const const_iterator* __i) const;
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const;
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const;
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type& __alloc() _NOEXCEPT
+        {return __r_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const allocator_type& __alloc() const _NOEXCEPT
+        {return __r_.second();}
+
+#ifdef _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_short_size(size_type __s) _NOEXCEPT
+#   ifdef _LIBCUDACXX_BIG_ENDIAN
+        {__r_.first().__s.__size_ = (unsigned char)(__s << 1);}
+#   else
+        {__r_.first().__s.__size_ = (unsigned char)(__s);}
+#   endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __get_short_size() const _NOEXCEPT
+#   ifdef _LIBCUDACXX_BIG_ENDIAN
+        {return __r_.first().__s.__size_ >> 1;}
+#   else
+        {return __r_.first().__s.__size_;}
+#   endif
+
+#else  // _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_short_size(size_type __s) _NOEXCEPT
+#   ifdef _LIBCUDACXX_BIG_ENDIAN
+        {__r_.first().__s.__size_ = (unsigned char)(__s);}
+#   else
+        {__r_.first().__s.__size_ = (unsigned char)(__s << 1);}
+#   endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __get_short_size() const _NOEXCEPT
+#   ifdef _LIBCUDACXX_BIG_ENDIAN
+        {return __r_.first().__s.__size_;}
+#   else
+        {return __r_.first().__s.__size_ >> 1;}
+#   endif
+
+#endif  // _LIBCUDACXX_ABI_ALTERNATE_STRING_LAYOUT
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_long_size(size_type __s) _NOEXCEPT
+        {__r_.first().__l.__size_ = __s;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __get_long_size() const _NOEXCEPT
+        {return __r_.first().__l.__size_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_size(size_type __s) _NOEXCEPT
+        {if (__is_long()) __set_long_size(__s); else __set_short_size(__s);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_long_cap(size_type __s) _NOEXCEPT
+        {__r_.first().__l.__cap_  = __long_mask | __s;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __get_long_cap() const _NOEXCEPT
+        {return __r_.first().__l.__cap_ & size_type(~__long_mask);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __set_long_pointer(pointer __p) _NOEXCEPT
+        {__r_.first().__l.__data_ = __p;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer __get_long_pointer() _NOEXCEPT
+        {return __r_.first().__l.__data_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_pointer __get_long_pointer() const _NOEXCEPT
+        {return __r_.first().__l.__data_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer __get_short_pointer() _NOEXCEPT
+        {return pointer_traits<pointer>::pointer_to(__r_.first().__s.__data_[0]);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_pointer __get_short_pointer() const _NOEXCEPT
+        {return pointer_traits<const_pointer>::pointer_to(__r_.first().__s.__data_[0]);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer __get_pointer() _NOEXCEPT
+        {return __is_long() ? __get_long_pointer() : __get_short_pointer();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_pointer __get_pointer() const _NOEXCEPT
+        {return __is_long() ? __get_long_pointer() : __get_short_pointer();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __zero() _NOEXCEPT
+        {
+            size_type (&__a)[__n_words] = __r_.first().__r.__words;
+            for (unsigned __i = 0; __i < __n_words; ++__i)
+                __a[__i] = 0;
+        }
+
+    template <size_type __a> static
+        _LIBCUDACXX_INLINE_VISIBILITY
+        size_type __align_it(size_type __s) _NOEXCEPT
+            {return (__s + (__a-1)) & ~(__a-1);}
+    enum {__alignment = 16};
+    static _LIBCUDACXX_INLINE_VISIBILITY
+    size_type __recommend(size_type __s) _NOEXCEPT
+        {
+        if (__s < __min_cap) return static_cast<size_type>(__min_cap) - 1;
+        size_type __guess = __align_it<sizeof(value_type) < __alignment ?
+                     __alignment/sizeof(value_type) : 1 > (__s+1) - 1;
+        if (__guess == __min_cap) ++__guess;
+        return __guess;
+        }
+
+    inline
+    void __init(const value_type* __s, size_type __sz, size_type __reserve);
+    inline
+    void __init(const value_type* __s, size_type __sz);
+    inline
+    void __init(size_type __n, value_type __c);
+
+    template <class _InputIterator>
+    inline
+    typename enable_if
+    <
+        __is_exactly_input_iterator<_InputIterator>::value,
+        void
+    >::type
+    __init(_InputIterator __first, _InputIterator __last);
+
+    template <class _ForwardIterator>
+    inline
+    typename enable_if
+    <
+        __is_forward_iterator<_ForwardIterator>::value,
+        void
+    >::type
+    __init(_ForwardIterator __first, _ForwardIterator __last);
+
+    void __grow_by(size_type __old_cap, size_type __delta_cap, size_type __old_sz,
+                   size_type __n_copy,  size_type __n_del,     size_type __n_add = 0);
+    void __grow_by_and_replace(size_type __old_cap, size_type __delta_cap, size_type __old_sz,
+                               size_type __n_copy,  size_type __n_del,
+                               size_type __n_add, const value_type* __p_new_stuff);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __erase_to_end(size_type __pos);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const basic_string& __str)
+        {__copy_assign_alloc(__str, integral_constant<bool,
+                      __alloc_traits::propagate_on_container_copy_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const basic_string& __str, true_type)
+        {
+            if (__alloc() == __str.__alloc())
+                __alloc() = __str.__alloc();
+            else
+            {
+                if (!__str.__is_long())
+                {
+                    __clear_and_shrink();
+                    __alloc() = __str.__alloc();
+                }
+                else
+                {
+                    allocator_type __a = __str.__alloc();
+                    pointer __p = __alloc_traits::allocate(__a, __str.__get_long_cap());
+                    __clear_and_shrink();
+                    __alloc() = _CUDA_VSTD::move(__a);
+                    __set_long_pointer(__p);
+                    __set_long_cap(__str.__get_long_cap());
+                    __set_long_size(__str.size());
+                }
+            }
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const basic_string&, false_type) _NOEXCEPT
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign(basic_string& __str, false_type)
+        _NOEXCEPT_(__alloc_traits::is_always_equal::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign(basic_string& __str, true_type)
+#if _LIBCUDACXX_STD_VER > 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value);
+#endif
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void
+    __move_assign_alloc(basic_string& __str)
+        _NOEXCEPT_(
+            !__alloc_traits::propagate_on_container_move_assignment::value ||
+            is_nothrow_move_assignable<allocator_type>::value)
+    {__move_assign_alloc(__str, integral_constant<bool,
+                      __alloc_traits::propagate_on_container_move_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(basic_string& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+        {
+            __alloc() = _CUDA_VSTD::move(__c.__alloc());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(basic_string&, false_type)
+        _NOEXCEPT
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY void __invalidate_all_iterators();
+    _LIBCUDACXX_INLINE_VISIBILITY void __invalidate_iterators_past(size_type);
+
+    friend basic_string operator+<>(const basic_string&, const basic_string&);
+    friend basic_string operator+<>(const value_type*, const basic_string&);
+    friend basic_string operator+<>(value_type, const basic_string&);
+    friend basic_string operator+<>(const basic_string&, const value_type*);
+    friend basic_string operator+<>(const basic_string&, value_type);
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _CharT = typename iterator_traits<_InputIterator>::value_type,
+         class _Allocator = allocator<_CharT>,
+         class = typename enable_if<__is_input_iterator<_InputIterator>::value, void>::type,
+         class = typename enable_if<__is_allocator<_Allocator>::value, void>::type
+         >
+basic_string(_InputIterator, _InputIterator, _Allocator = _Allocator())
+  -> basic_string<_CharT, char_traits<_CharT>, _Allocator>;
+
+template<class _CharT,
+         class _Traits,
+         class _Allocator = allocator<_CharT>,
+         class = typename enable_if<__is_allocator<_Allocator>::value, void>::type
+         >
+explicit basic_string(basic_string_view<_CharT, _Traits>, const _Allocator& = _Allocator())
+  -> basic_string<_CharT, _Traits, _Allocator>;
+
+template<class _CharT,
+         class _Traits,
+         class _Allocator = allocator<_CharT>,
+         class = typename enable_if<__is_allocator<_Allocator>::value, void>::type,
+         class _Sz = typename allocator_traits<_Allocator>::size_type
+         >
+basic_string(basic_string_view<_CharT, _Traits>, _Sz, _Sz, const _Allocator& = _Allocator())
+  -> basic_string<_CharT, _Traits, _Allocator>;
+#endif
+
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::__invalidate_all_iterators()
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__invalidate_all(this);
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::__invalidate_iterators_past(size_type
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+                                                                        __pos
+#endif
+                                                                      )
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __c_node* __c = __get_db()->__find_c_and_lock(this);
+    if (__c)
+    {
+        const_pointer __new_last = __get_pointer() + __pos;
+        for (__i_node** __p = __c->end_; __p != __c->beg_; )
+        {
+            --__p;
+            const_iterator* __i = static_cast<const_iterator*>((*__p)->__i_);
+            if (__i->base() > __new_last)
+            {
+                (*__p)->__c_ = nullptr;
+                if (--__c->end_ != __p)
+                    memmove(__p, __p+1, (__c->end_ - __p)*sizeof(__i_node*));
+            }
+        }
+        __get_db()->unlock();
+    }
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string()
+    _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __zero();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const allocator_type& __a)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_copy_constructible<allocator_type>::value)
+#else
+        _NOEXCEPT
+#endif
+: __r_(__second_tag(), __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __zero();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void basic_string<_CharT, _Traits, _Allocator>::__init(const value_type* __s,
+                                                       size_type __sz,
+                                                       size_type __reserve)
+{
+    if (__reserve > max_size())
+        this->__throw_length_error();
+    pointer __p;
+    if (__reserve < __min_cap)
+    {
+        __set_short_size(__sz);
+        __p = __get_short_pointer();
+    }
+    else
+    {
+        size_type __cap = __recommend(__reserve);
+        __p = __alloc_traits::allocate(__alloc(), __cap+1);
+        __set_long_pointer(__p);
+        __set_long_cap(__cap+1);
+        __set_long_size(__sz);
+    }
+    traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p), __s, __sz);
+    traits_type::assign(__p[__sz], value_type());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::__init(const value_type* __s, size_type __sz)
+{
+    if (__sz > max_size())
+        this->__throw_length_error();
+    pointer __p;
+    if (__sz < __min_cap)
+    {
+        __set_short_size(__sz);
+        __p = __get_short_pointer();
+    }
+    else
+    {
+        size_type __cap = __recommend(__sz);
+        __p = __alloc_traits::allocate(__alloc(), __cap+1);
+        __set_long_pointer(__p);
+        __set_long_cap(__cap+1);
+        __set_long_size(__sz);
+    }
+    traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p), __s, __sz);
+    traits_type::assign(__p[__sz], value_type());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const _CharT* __s, const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "basic_string(const char*, allocator) detected nullptr");
+    __init(__s, traits_type::length(__s));
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const _CharT* __s, size_type __n)
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "basic_string(const char*, n) detected nullptr");
+    __init(__s, __n);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const _CharT* __s, size_type __n, const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "basic_string(const char*, n, allocator) detected nullptr");
+    __init(__s, __n);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const basic_string& __str)
+    : __r_(__second_tag(), __alloc_traits::select_on_container_copy_construction(__str.__alloc()))
+{
+    if (!__str.__is_long())
+        __r_.first().__r = __str.__r_.first().__r;
+    else
+        __init(_CUDA_VSTD::__to_raw_pointer(__str.__get_long_pointer()), __str.__get_long_size());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(
+    const basic_string& __str, const allocator_type& __a)
+    : __r_(__second_tag(), __a)
+{
+    if (!__str.__is_long())
+        __r_.first().__r = __str.__r_.first().__r;
+    else
+        __init(_CUDA_VSTD::__to_raw_pointer(__str.__get_long_pointer()), __str.__get_long_size());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(basic_string&& __str)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value)
+#else
+        _NOEXCEPT
+#endif
+    : __r_(_CUDA_VSTD::move(__str.__r_))
+{
+    __str.__zero();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    if (__is_long())
+        __get_db()->swap(this, &__str);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(basic_string&& __str, const allocator_type& __a)
+    : __r_(__second_tag(), __a)
+{
+    if (__str.__is_long() && __a != __str.__alloc()) // copy, not move
+        __init(_CUDA_VSTD::__to_raw_pointer(__str.__get_long_pointer()), __str.__get_long_size());
+    else
+    {
+        __r_.first().__r = __str.__r_.first().__r;
+        __str.__zero();
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    if (__is_long())
+        __get_db()->swap(this, &__str);
+#endif
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::__init(size_type __n, value_type __c)
+{
+    if (__n > max_size())
+        this->__throw_length_error();
+    pointer __p;
+    if (__n < __min_cap)
+    {
+        __set_short_size(__n);
+        __p = __get_short_pointer();
+    }
+    else
+    {
+        size_type __cap = __recommend(__n);
+        __p = __alloc_traits::allocate(__alloc(), __cap+1);
+        __set_long_pointer(__p);
+        __set_long_cap(__cap+1);
+        __set_long_size(__n);
+    }
+    traits_type::assign(_CUDA_VSTD::__to_raw_pointer(__p), __n, __c);
+    traits_type::assign(__p[__n], value_type());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(size_type __n, _CharT __c)
+{
+    __init(__n, __c);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(size_type __n, _CharT __c, const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    __init(__n, __c);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const basic_string& __str,
+                                                        size_type __pos, size_type __n,
+                                                        const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    size_type __str_sz = __str.size();
+    if (__pos > __str_sz)
+        this->__throw_out_of_range();
+    __init(__str.data() + __pos, _CUDA_VSTD::min(__n, __str_sz - __pos));
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const basic_string& __str, size_type __pos,
+                                                        const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    size_type __str_sz = __str.size();
+    if (__pos > __str_sz)
+        this->__throw_out_of_range();
+    __init(__str.data() + __pos, __str_sz - __pos);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp, class>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(
+             const _Tp& __t, size_type __pos, size_type __n, const allocator_type& __a)
+    : __r_(__second_tag(), __a)
+{
+    __self_view __sv0 = __t;
+    __self_view __sv = __sv0.substr(__pos, __n);
+    __init(__sv.data(), __sv.size());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp, class>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const _Tp & __t)
+{
+    __self_view __sv = __t;
+    __init(__sv.data(), __sv.size());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp, class>
+basic_string<_CharT, _Traits, _Allocator>::basic_string(const _Tp & __t, const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    __self_view __sv = __t;
+    __init(__sv.data(), __sv.size());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _InputIterator>
+typename enable_if
+<
+    __is_exactly_input_iterator<_InputIterator>::value,
+    void
+>::type
+basic_string<_CharT, _Traits, _Allocator>::__init(_InputIterator __first, _InputIterator __last)
+{
+    __zero();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    for (; __first != __last; ++__first)
+        push_back(*__first);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        if (__is_long())
+            __alloc_traits::deallocate(__alloc(), __get_long_pointer(), __get_long_cap());
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value,
+    void
+>::type
+basic_string<_CharT, _Traits, _Allocator>::__init(_ForwardIterator __first, _ForwardIterator __last)
+{
+    size_type __sz = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__sz > max_size())
+        this->__throw_length_error();
+    pointer __p;
+    if (__sz < __min_cap)
+    {
+        __set_short_size(__sz);
+        __p = __get_short_pointer();
+    }
+    else
+    {
+        size_type __cap = __recommend(__sz);
+        __p = __alloc_traits::allocate(__alloc(), __cap+1);
+        __set_long_pointer(__p);
+        __set_long_cap(__cap+1);
+        __set_long_size(__sz);
+    }
+    for (; __first != __last; ++__first, (void) ++__p)
+        traits_type::assign(*__p, *__first);
+    traits_type::assign(*__p, value_type());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _InputIterator, class>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(_InputIterator __first, _InputIterator __last)
+{
+    __init(__first, __last);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _InputIterator, class>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(_InputIterator __first, _InputIterator __last,
+                                                        const allocator_type& __a)
+    : __r_(__second_tag(), __a)
+{
+    __init(__first, __last);
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>::basic_string(
+    initializer_list<_CharT> __il)
+{
+    __init(__il.begin(), __il.end());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+
+basic_string<_CharT, _Traits, _Allocator>::basic_string(
+    initializer_list<_CharT> __il, const _Allocator& __a)
+    : __r_(__second_tag(), __a)
+{
+    __init(__il.begin(), __il.end());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>::~basic_string()
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__erase_c(this);
+#endif
+    if (__is_long())
+        __alloc_traits::deallocate(__alloc(), __get_long_pointer(), __get_long_cap());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::__grow_by_and_replace
+    (size_type __old_cap, size_type __delta_cap, size_type __old_sz,
+     size_type __n_copy,  size_type __n_del,     size_type __n_add, const value_type* __p_new_stuff)
+{
+    size_type __ms = max_size();
+    if (__delta_cap > __ms - __old_cap - 1)
+        this->__throw_length_error();
+    pointer __old_p = __get_pointer();
+    size_type __cap = __old_cap < __ms / 2 - __alignment ?
+                          __recommend(_CUDA_VSTD::max(__old_cap + __delta_cap, 2 * __old_cap)) :
+                          __ms - 1;
+    pointer __p = __alloc_traits::allocate(__alloc(), __cap+1);
+    __invalidate_all_iterators();
+    if (__n_copy != 0)
+        traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p),
+                          _CUDA_VSTD::__to_raw_pointer(__old_p), __n_copy);
+    if (__n_add != 0)
+        traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p) + __n_copy, __p_new_stuff, __n_add);
+    size_type __sec_cp_sz = __old_sz - __n_del - __n_copy;
+    if (__sec_cp_sz != 0)
+        traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p) + __n_copy + __n_add,
+                          _CUDA_VSTD::__to_raw_pointer(__old_p) + __n_copy + __n_del, __sec_cp_sz);
+    if (__old_cap+1 != __min_cap)
+        __alloc_traits::deallocate(__alloc(), __old_p, __old_cap+1);
+    __set_long_pointer(__p);
+    __set_long_cap(__cap+1);
+    __old_sz = __n_copy + __n_add + __sec_cp_sz;
+    __set_long_size(__old_sz);
+    traits_type::assign(__p[__old_sz], value_type());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::__grow_by(size_type __old_cap, size_type __delta_cap, size_type __old_sz,
+                                                     size_type __n_copy,  size_type __n_del,     size_type __n_add)
+{
+    size_type __ms = max_size();
+    if (__delta_cap > __ms - __old_cap)
+        this->__throw_length_error();
+    pointer __old_p = __get_pointer();
+    size_type __cap = __old_cap < __ms / 2 - __alignment ?
+                          __recommend(_CUDA_VSTD::max(__old_cap + __delta_cap, 2 * __old_cap)) :
+                          __ms - 1;
+    pointer __p = __alloc_traits::allocate(__alloc(), __cap+1);
+    __invalidate_all_iterators();
+    if (__n_copy != 0)
+        traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p),
+                          _CUDA_VSTD::__to_raw_pointer(__old_p), __n_copy);
+    size_type __sec_cp_sz = __old_sz - __n_del - __n_copy;
+    if (__sec_cp_sz != 0)
+        traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__p) + __n_copy + __n_add,
+                          _CUDA_VSTD::__to_raw_pointer(__old_p) + __n_copy + __n_del,
+                          __sec_cp_sz);
+    if (__old_cap+1 != __min_cap)
+        __alloc_traits::deallocate(__alloc(), __old_p, __old_cap+1);
+    __set_long_pointer(__p);
+    __set_long_cap(__cap+1);
+}
+
+// assign
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::assign(const value_type* __s, size_type __n)
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::assign received nullptr");
+    size_type __cap = capacity();
+    if (__cap >= __n)
+    {
+        value_type* __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+        traits_type::move(__p, __s, __n);
+        traits_type::assign(__p[__n], value_type());
+        __set_size(__n);
+        __invalidate_iterators_past(__n);
+    }
+    else
+    {
+        size_type __sz = size();
+        __grow_by_and_replace(__cap, __n - __cap, __sz, 0, __sz, __n, __s);
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::assign(size_type __n, value_type __c)
+{
+    size_type __cap = capacity();
+    if (__cap < __n)
+    {
+        size_type __sz = size();
+        __grow_by(__cap, __n - __cap, __sz, 0, __sz);
+    }
+    else
+        __invalidate_iterators_past(__n);
+    value_type* __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+    traits_type::assign(__p, __n, __c);
+    traits_type::assign(__p[__n], value_type());
+    __set_size(__n);
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::operator=(value_type __c)
+{
+    pointer __p;
+    if (__is_long())
+    {
+        __p = __get_long_pointer();
+        __set_long_size(1);
+    }
+    else
+    {
+        __p = __get_short_pointer();
+        __set_short_size(1);
+    }
+    traits_type::assign(*__p, __c);
+    traits_type::assign(*++__p, value_type());
+    __invalidate_iterators_past(1);
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::operator=(const basic_string& __str)
+{
+    if (this != &__str)
+    {
+        __copy_assign_alloc(__str);
+        assign(__str.data(), __str.size());
+    }
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::__move_assign(basic_string& __str, false_type)
+    _NOEXCEPT_(__alloc_traits::is_always_equal::value)
+{
+    if (__alloc() != __str.__alloc())
+        assign(__str);
+    else
+        __move_assign(__str, true_type());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::__move_assign(basic_string& __str, true_type)
+#if _LIBCUDACXX_STD_VER > 14
+    _NOEXCEPT
+#else
+    _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+#endif
+{
+    __clear_and_shrink();
+    __r_.first() = __str.__r_.first();
+    __move_assign_alloc(__str);
+    __str.__zero();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::operator=(basic_string&& __str)
+    _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value))
+{
+    __move_assign(__str, integral_constant<bool,
+          __alloc_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+#endif
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _InputIterator>
+typename enable_if
+<
+     __is_exactly_input_iterator <_InputIterator>::value
+          || !__libcpp_string_gets_noexcept_iterator<_InputIterator>::value,
+    basic_string<_CharT, _Traits, _Allocator>&
+>::type
+basic_string<_CharT, _Traits, _Allocator>::assign(_InputIterator __first, _InputIterator __last)
+{
+    const basic_string __temp(__first, __last, __alloc());
+    assign(__temp.data(), __temp.size());
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value
+         && __libcpp_string_gets_noexcept_iterator<_ForwardIterator>::value,
+    basic_string<_CharT, _Traits, _Allocator>&
+>::type
+basic_string<_CharT, _Traits, _Allocator>::assign(_ForwardIterator __first, _ForwardIterator __last)
+{
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    size_type __cap = capacity();
+    if (__cap < __n)
+    {
+        size_type __sz = size();
+        __grow_by(__cap, __n - __cap, __sz, 0, __sz);
+    }
+    else
+        __invalidate_iterators_past(__n);
+    pointer __p = __get_pointer();
+    for (; __first != __last; ++__first, ++__p)
+        traits_type::assign(*__p, *__first);
+    traits_type::assign(*__p, value_type());
+    __set_size(__n);
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::assign(const basic_string& __str, size_type __pos, size_type __n)
+{
+    size_type __sz = __str.size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    return assign(__str.data() + __pos, _CUDA_VSTD::min(__n, __sz - __pos));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    basic_string<_CharT, _Traits, _Allocator>&
+>::type
+basic_string<_CharT, _Traits, _Allocator>::assign(const _Tp & __t, size_type __pos, size_type __n)
+{
+    __self_view __sv = __t;
+    size_type __sz = __sv.size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    return assign(__sv.data() + __pos, _CUDA_VSTD::min(__n, __sz - __pos));
+}
+
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::assign(const value_type* __s)
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::assign received nullptr");
+    return assign(__s, traits_type::length(__s));
+}
+
+// append
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::append(const value_type* __s, size_type __n)
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::append received nullptr");
+    size_type __cap = capacity();
+    size_type __sz = size();
+    if (__cap - __sz >= __n)
+    {
+        if (__n)
+        {
+            value_type* __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+            traits_type::copy(__p + __sz, __s, __n);
+            __sz += __n;
+            __set_size(__sz);
+            traits_type::assign(__p[__sz], value_type());
+        }
+    }
+    else
+        __grow_by_and_replace(__cap, __sz + __n - __cap, __sz, __sz, 0, __n, __s);
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::append(size_type __n, value_type __c)
+{
+    if (__n)
+    {
+        size_type __cap = capacity();
+        size_type __sz = size();
+        if (__cap - __sz < __n)
+            __grow_by(__cap, __sz + __n - __cap, __sz, __sz, 0);
+        pointer __p = __get_pointer();
+        traits_type::assign(_CUDA_VSTD::__to_raw_pointer(__p) + __sz, __n, __c);
+        __sz += __n;
+        __set_size(__sz);
+        traits_type::assign(__p[__sz], value_type());
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline void
+basic_string<_CharT, _Traits, _Allocator>::__append_default_init(size_type __n)
+{
+    if (__n)
+    {
+        size_type __cap = capacity();
+        size_type __sz = size();
+        if (__cap - __sz < __n)
+            __grow_by(__cap, __sz + __n - __cap, __sz, __sz, 0);
+        pointer __p = __get_pointer();
+        __sz += __n;
+        __set_size(__sz);
+        traits_type::assign(__p[__sz], value_type());
+    }
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::push_back(value_type __c)
+{
+    bool __is_short = !__is_long();
+    size_type __cap;
+    size_type __sz;
+    if (__is_short)
+    {
+        __cap = __min_cap - 1;
+        __sz = __get_short_size();
+    }
+    else
+    {
+        __cap = __get_long_cap() - 1;
+        __sz = __get_long_size();
+    }
+    if (__sz == __cap)
+    {
+        __grow_by(__cap, 1, __sz, __sz, 0);
+        __is_short = !__is_long();
+    }
+    pointer __p;
+    if (__is_short)
+    {
+        __p = __get_short_pointer() + __sz;
+        __set_short_size(__sz+1);
+    }
+    else
+    {
+        __p = __get_long_pointer() + __sz;
+        __set_long_size(__sz+1);
+    }
+    traits_type::assign(*__p, __c);
+    traits_type::assign(*++__p, value_type());
+}
+
+template <class _Tp>
+bool __ptr_in_range (const _Tp* __p, const _Tp* __first, const _Tp* __last)
+{
+    return __first <= __p && __p < __last;
+}
+
+template <class _Tp1, class _Tp2>
+bool __ptr_in_range (const _Tp1*, const _Tp2*, const _Tp2*)
+{
+    return false;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _ForwardIterator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::__append_forward_unsafe(
+    _ForwardIterator __first, _ForwardIterator __last)
+{
+    static_assert(__is_forward_iterator<_ForwardIterator>::value,
+                  "function requires a ForwardIterator");
+    size_type __sz = size();
+    size_type __cap = capacity();
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__n)
+    {
+        typedef typename iterator_traits<_ForwardIterator>::reference _CharRef;
+        _CharRef __tmp_ref = *__first;
+        if (__ptr_in_range(_CUDA_VSTD::addressof(__tmp_ref), data(), data() + size()))
+        {
+            const basic_string __temp (__first, __last, __alloc());
+            append(__temp.data(), __temp.size());
+        }
+        else
+        {
+            if (__cap - __sz < __n)
+                __grow_by(__cap, __sz + __n - __cap, __sz, __sz, 0);
+            pointer __p = __get_pointer() + __sz;
+            for (; __first != __last; ++__p, ++__first)
+                traits_type::assign(*__p, *__first);
+            traits_type::assign(*__p, value_type());
+            __set_size(__sz + __n);
+        }
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::append(const basic_string& __str)
+{
+    return append(__str.data(), __str.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::append(const basic_string& __str, size_type __pos, size_type __n)
+{
+    size_type __sz = __str.size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    return append(__str.data() + __pos, _CUDA_VSTD::min(__n, __sz - __pos));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+    typename enable_if
+    <
+        __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+        basic_string<_CharT, _Traits, _Allocator>&
+    >::type
+basic_string<_CharT, _Traits, _Allocator>::append(const _Tp & __t, size_type __pos, size_type __n)
+{
+    __self_view __sv = __t;
+    size_type __sz = __sv.size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    return append(__sv.data() + __pos, _CUDA_VSTD::min(__n, __sz - __pos));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::append(const value_type* __s)
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::append received nullptr");
+    return append(__s, traits_type::length(__s));
+}
+
+// insert
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::insert(size_type __pos, const value_type* __s, size_type __n)
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::insert received nullptr");
+    size_type __sz = size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    size_type __cap = capacity();
+    if (__cap - __sz >= __n)
+    {
+        if (__n)
+        {
+            value_type* __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+            size_type __n_move = __sz - __pos;
+            if (__n_move != 0)
+            {
+                if (__p + __pos <= __s && __s < __p + __sz)
+                    __s += __n;
+                traits_type::move(__p + __pos + __n, __p + __pos, __n_move);
+            }
+            traits_type::move(__p + __pos, __s, __n);
+            __sz += __n;
+            __set_size(__sz);
+            traits_type::assign(__p[__sz], value_type());
+        }
+    }
+    else
+        __grow_by_and_replace(__cap, __sz + __n - __cap, __sz, __pos, 0, __n, __s);
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::insert(size_type __pos, size_type __n, value_type __c)
+{
+    size_type __sz = size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    if (__n)
+    {
+        size_type __cap = capacity();
+        value_type* __p;
+        if (__cap - __sz >= __n)
+        {
+            __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+            size_type __n_move = __sz - __pos;
+            if (__n_move != 0)
+                traits_type::move(__p + __pos + __n, __p + __pos, __n_move);
+        }
+        else
+        {
+            __grow_by(__cap, __sz + __n - __cap, __sz, __pos, 0, __n);
+            __p = _CUDA_VSTD::__to_raw_pointer(__get_long_pointer());
+        }
+        traits_type::assign(__p + __pos, __n, __c);
+        __sz += __n;
+        __set_size(__sz);
+        traits_type::assign(__p[__sz], value_type());
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _InputIterator>
+typename enable_if
+<
+   __is_exactly_input_iterator<_InputIterator>::value
+        || !__libcpp_string_gets_noexcept_iterator<_InputIterator>::value,
+   typename basic_string<_CharT, _Traits, _Allocator>::iterator
+>::type
+basic_string<_CharT, _Traits, _Allocator>::insert(const_iterator __pos, _InputIterator __first, _InputIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__pos) == this,
+        "string::insert(iterator, range) called with an iterator not"
+        " referring to this string");
+#endif
+    const basic_string __temp(__first, __last, __alloc());
+    return insert(__pos, __temp.data(), __temp.data() + __temp.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value
+        && __libcpp_string_gets_noexcept_iterator<_ForwardIterator>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::iterator
+>::type
+basic_string<_CharT, _Traits, _Allocator>::insert(const_iterator __pos, _ForwardIterator __first, _ForwardIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__pos) == this,
+        "string::insert(iterator, range) called with an iterator not"
+        " referring to this string");
+#endif
+    size_type __ip = static_cast<size_type>(__pos - begin());
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__n)
+    {
+        typedef typename iterator_traits<_ForwardIterator>::reference _CharRef;
+        _CharRef __tmp_char = *__first;
+        if (__ptr_in_range(_CUDA_VSTD::addressof(__tmp_char), data(), data() + size()))
+        {
+            const basic_string __temp(__first, __last, __alloc());
+            return insert(__pos, __temp.data(), __temp.data() + __temp.size());
+        }
+
+        size_type __sz = size();
+        size_type __cap = capacity();
+        value_type* __p;
+        if (__cap - __sz >= __n)
+        {
+            __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+            size_type __n_move = __sz - __ip;
+            if (__n_move != 0)
+                traits_type::move(__p + __ip + __n, __p + __ip, __n_move);
+        }
+        else
+        {
+            __grow_by(__cap, __sz + __n - __cap, __sz, __ip, 0, __n);
+            __p = _CUDA_VSTD::__to_raw_pointer(__get_long_pointer());
+        }
+        __sz += __n;
+        __set_size(__sz);
+        traits_type::assign(__p[__sz], value_type());
+        for (__p += __ip; __first != __last; ++__p, ++__first)
+            traits_type::assign(*__p, *__first);
+    }
+    return begin() + __ip;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::insert(size_type __pos1, const basic_string& __str)
+{
+    return insert(__pos1, __str.data(), __str.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::insert(size_type __pos1, const basic_string& __str,
+                                                  size_type __pos2, size_type __n)
+{
+    size_type __str_sz = __str.size();
+    if (__pos2 > __str_sz)
+        this->__throw_out_of_range();
+    return insert(__pos1, __str.data() + __pos2, _CUDA_VSTD::min(__n, __str_sz - __pos2));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    basic_string<_CharT, _Traits, _Allocator>&
+>::type
+basic_string<_CharT, _Traits, _Allocator>::insert(size_type __pos1, const _Tp& __t,
+                                                  size_type __pos2, size_type __n)
+{
+    __self_view __sv = __t;
+    size_type __str_sz = __sv.size();
+    if (__pos2 > __str_sz)
+        this->__throw_out_of_range();
+    return insert(__pos1, __sv.data() + __pos2, _CUDA_VSTD::min(__n, __str_sz - __pos2));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::insert(size_type __pos, const value_type* __s)
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::insert received nullptr");
+    return insert(__pos, __s, traits_type::length(__s));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::iterator
+basic_string<_CharT, _Traits, _Allocator>::insert(const_iterator __pos, value_type __c)
+{
+    size_type __ip = static_cast<size_type>(__pos - begin());
+    size_type __sz = size();
+    size_type __cap = capacity();
+    value_type* __p;
+    if (__cap == __sz)
+    {
+        __grow_by(__cap, 1, __sz, __ip, 0, 1);
+        __p = _CUDA_VSTD::__to_raw_pointer(__get_long_pointer());
+    }
+    else
+    {
+        __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+        size_type __n_move = __sz - __ip;
+        if (__n_move != 0)
+            traits_type::move(__p + __ip + 1, __p + __ip, __n_move);
+    }
+    traits_type::assign(__p[__ip], __c);
+    traits_type::assign(__p[++__sz], value_type());
+    __set_size(__sz);
+    return begin() + static_cast<difference_type>(__ip);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::iterator
+basic_string<_CharT, _Traits, _Allocator>::insert(const_iterator __pos, size_type __n, value_type __c)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__pos) == this,
+        "string::insert(iterator, n, value) called with an iterator not"
+        " referring to this string");
+#endif
+    difference_type __p = __pos - begin();
+    insert(static_cast<size_type>(__p), __n, __c);
+    return begin() + __p;
+}
+
+// replace
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(size_type __pos, size_type __n1, const value_type* __s, size_type __n2)
+    _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+{
+    _LIBCUDACXX_ASSERT(__n2 == 0 || __s != nullptr, "string::replace received nullptr");
+    size_type __sz = size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    __n1 = _CUDA_VSTD::min(__n1, __sz - __pos);
+    size_type __cap = capacity();
+    if (__cap - __sz + __n1 >= __n2)
+    {
+        value_type* __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+        if (__n1 != __n2)
+        {
+            size_type __n_move = __sz - __pos - __n1;
+            if (__n_move != 0)
+            {
+                if (__n1 > __n2)
+                {
+                    traits_type::move(__p + __pos, __s, __n2);
+                    traits_type::move(__p + __pos + __n2, __p + __pos + __n1, __n_move);
+                    goto __finish;
+                }
+                if (__p + __pos < __s && __s < __p + __sz)
+                {
+                    if (__p + __pos + __n1 <= __s)
+                        __s += __n2 - __n1;
+                    else // __p + __pos < __s < __p + __pos + __n1
+                    {
+                        traits_type::move(__p + __pos, __s, __n1);
+                        __pos += __n1;
+                        __s += __n2;
+                        __n2 -= __n1;
+                        __n1 = 0;
+                    }
+                }
+                traits_type::move(__p + __pos + __n2, __p + __pos + __n1, __n_move);
+            }
+        }
+        traits_type::move(__p + __pos, __s, __n2);
+__finish:
+// __sz += __n2 - __n1; in this and the below function below can cause unsigned integer overflow,
+// but this is a safe operation, so we disable the check.
+        __sz += __n2 - __n1;
+        __set_size(__sz);
+        __invalidate_iterators_past(__sz);
+        traits_type::assign(__p[__sz], value_type());
+    }
+    else
+        __grow_by_and_replace(__cap, __sz - __n1 + __n2 - __cap, __sz, __pos, __n1, __n2, __s);
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(size_type __pos, size_type __n1, size_type __n2, value_type __c)
+    _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+{
+    size_type __sz = size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    __n1 = _CUDA_VSTD::min(__n1, __sz - __pos);
+    size_type __cap = capacity();
+    value_type* __p;
+    if (__cap - __sz + __n1 >= __n2)
+    {
+        __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+        if (__n1 != __n2)
+        {
+            size_type __n_move = __sz - __pos - __n1;
+            if (__n_move != 0)
+                traits_type::move(__p + __pos + __n2, __p + __pos + __n1, __n_move);
+        }
+    }
+    else
+    {
+        __grow_by(__cap, __sz - __n1 + __n2 - __cap, __sz, __pos, __n1, __n2);
+        __p = _CUDA_VSTD::__to_raw_pointer(__get_long_pointer());
+    }
+    traits_type::assign(__p + __pos, __n2, __c);
+    __sz += __n2 - __n1;
+    __set_size(__sz);
+    __invalidate_iterators_past(__sz);
+    traits_type::assign(__p[__sz], value_type());
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template<class _InputIterator>
+typename enable_if
+<
+    __is_input_iterator<_InputIterator>::value,
+    basic_string<_CharT, _Traits, _Allocator>&
+>::type
+basic_string<_CharT, _Traits, _Allocator>::replace(const_iterator __i1, const_iterator __i2,
+                                                   _InputIterator __j1, _InputIterator __j2)
+{
+    const basic_string __temp(__j1, __j2, __alloc());
+    return this->replace(__i1, __i2, __temp);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(size_type __pos1, size_type __n1, const basic_string& __str)
+{
+    return replace(__pos1, __n1, __str.data(), __str.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(size_type __pos1, size_type __n1, const basic_string& __str,
+                                                   size_type __pos2, size_type __n2)
+{
+    size_type __str_sz = __str.size();
+    if (__pos2 > __str_sz)
+        this->__throw_out_of_range();
+    return replace(__pos1, __n1, __str.data() + __pos2, _CUDA_VSTD::min(__n2, __str_sz - __pos2));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    basic_string<_CharT, _Traits, _Allocator>&
+>::type
+basic_string<_CharT, _Traits, _Allocator>::replace(size_type __pos1, size_type __n1, const _Tp& __t,
+                                                   size_type __pos2, size_type __n2)
+{
+    __self_view __sv = __t;
+    size_type __str_sz = __sv.size();
+    if (__pos2 > __str_sz)
+        this->__throw_out_of_range();
+    return replace(__pos1, __n1, __sv.data() + __pos2, _CUDA_VSTD::min(__n2, __str_sz - __pos2));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(size_type __pos, size_type __n1, const value_type* __s)
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::replace received nullptr");
+    return replace(__pos, __n1, __s, traits_type::length(__s));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(const_iterator __i1, const_iterator __i2, const basic_string& __str)
+{
+    return replace(static_cast<size_type>(__i1 - begin()), static_cast<size_type>(__i2 - __i1),
+                   __str.data(), __str.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(const_iterator __i1, const_iterator __i2, const value_type* __s, size_type __n)
+{
+    return replace(static_cast<size_type>(__i1 - begin()), static_cast<size_type>(__i2 - __i1), __s, __n);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(const_iterator __i1, const_iterator __i2, const value_type* __s)
+{
+    return replace(static_cast<size_type>(__i1 - begin()), static_cast<size_type>(__i2 - __i1), __s);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::replace(const_iterator __i1, const_iterator __i2, size_type __n, value_type __c)
+{
+    return replace(static_cast<size_type>(__i1 - begin()), static_cast<size_type>(__i2 - __i1), __n, __c);
+}
+
+// erase
+
+template <class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>&
+basic_string<_CharT, _Traits, _Allocator>::erase(size_type __pos, size_type __n)
+{
+    size_type __sz = size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    if (__n)
+    {
+        value_type* __p = _CUDA_VSTD::__to_raw_pointer(__get_pointer());
+        __n = _CUDA_VSTD::min(__n, __sz - __pos);
+        size_type __n_move = __sz - __pos - __n;
+        if (__n_move != 0)
+            traits_type::move(__p + __pos, __p + __pos + __n, __n_move);
+        __sz -= __n;
+        __set_size(__sz);
+        __invalidate_iterators_past(__sz);
+        traits_type::assign(__p[__sz], value_type());
+    }
+    return *this;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::iterator
+basic_string<_CharT, _Traits, _Allocator>::erase(const_iterator __pos)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__pos) == this,
+        "string::erase(iterator) called with an iterator not"
+        " referring to this string");
+#endif
+    _LIBCUDACXX_ASSERT(__pos != end(),
+        "string::erase(iterator) called with a non-dereferenceable iterator");
+    iterator __b = begin();
+    size_type __r = static_cast<size_type>(__pos - __b);
+    erase(__r, 1);
+    return __b + static_cast<difference_type>(__r);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::iterator
+basic_string<_CharT, _Traits, _Allocator>::erase(const_iterator __first, const_iterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__first) == this,
+        "string::erase(iterator,  iterator) called with an iterator not"
+        " referring to this string");
+#endif
+    _LIBCUDACXX_ASSERT(__first <= __last, "string::erase(first, last) called with invalid range");
+    iterator __b = begin();
+    size_type __r = static_cast<size_type>(__first - __b);
+    erase(__r, static_cast<size_type>(__last - __first));
+    return __b + static_cast<difference_type>(__r);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::pop_back()
+{
+    _LIBCUDACXX_ASSERT(!empty(), "string::pop_back(): string is already empty");
+    size_type __sz;
+    if (__is_long())
+    {
+        __sz = __get_long_size() - 1;
+        __set_long_size(__sz);
+        traits_type::assign(*(__get_long_pointer() + __sz), value_type());
+    }
+    else
+    {
+        __sz = __get_short_size() - 1;
+        __set_short_size(__sz);
+        traits_type::assign(*(__get_short_pointer() + __sz), value_type());
+    }
+    __invalidate_iterators_past(__sz);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::clear() _NOEXCEPT
+{
+    __invalidate_all_iterators();
+    if (__is_long())
+    {
+        traits_type::assign(*__get_long_pointer(), value_type());
+        __set_long_size(0);
+    }
+    else
+    {
+        traits_type::assign(*__get_short_pointer(), value_type());
+        __set_short_size(0);
+    }
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::__erase_to_end(size_type __pos)
+{
+    if (__is_long())
+    {
+        traits_type::assign(*(__get_long_pointer() + __pos), value_type());
+        __set_long_size(__pos);
+    }
+    else
+    {
+        traits_type::assign(*(__get_short_pointer() + __pos), value_type());
+        __set_short_size(__pos);
+    }
+    __invalidate_iterators_past(__pos);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::resize(size_type __n, value_type __c)
+{
+    size_type __sz = size();
+    if (__n > __sz)
+        append(__n - __sz, __c);
+    else
+        __erase_to_end(__n);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline void
+basic_string<_CharT, _Traits, _Allocator>::__resize_default_init(size_type __n)
+{
+    size_type __sz = size();
+    if (__n > __sz) {
+       __append_default_init(__n - __sz);
+    } else
+        __erase_to_end(__n);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::max_size() const _NOEXCEPT
+{
+    size_type __m = __alloc_traits::max_size(__alloc());
+#ifdef _LIBCUDACXX_BIG_ENDIAN
+    return (__m <= ~__long_mask ? __m : __m/2) - __alignment;
+#else
+    return __m - __alignment;
+#endif
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+void
+basic_string<_CharT, _Traits, _Allocator>::reserve(size_type __res_arg)
+{
+    if (__res_arg > max_size())
+        this->__throw_length_error();
+    size_type __cap = capacity();
+    size_type __sz = size();
+    __res_arg = _CUDA_VSTD::max(__res_arg, __sz);
+    __res_arg = __recommend(__res_arg);
+    if (__res_arg != __cap)
+    {
+        pointer __new_data, __p;
+        bool __was_long, __now_long;
+        if (__res_arg == __min_cap - 1)
+        {
+            __was_long = true;
+            __now_long = false;
+            __new_data = __get_short_pointer();
+            __p = __get_long_pointer();
+        }
+        else
+        {
+            if (__res_arg > __cap)
+                __new_data = __alloc_traits::allocate(__alloc(), __res_arg+1);
+            else
+            {
+            #ifndef _LIBCUDACXX_NO_EXCEPTIONS
+                try
+                {
+            #endif  // _LIBCUDACXX_NO_EXCEPTIONS
+                    __new_data = __alloc_traits::allocate(__alloc(), __res_arg+1);
+            #ifndef _LIBCUDACXX_NO_EXCEPTIONS
+                }
+                catch (...)
+                {
+                    return;
+                }
+            #else  // _LIBCUDACXX_NO_EXCEPTIONS
+                if (__new_data == nullptr)
+                    return;
+            #endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            }
+            __now_long = true;
+            __was_long = __is_long();
+            __p = __get_pointer();
+        }
+        traits_type::copy(_CUDA_VSTD::__to_raw_pointer(__new_data),
+                          _CUDA_VSTD::__to_raw_pointer(__p), size()+1);
+        if (__was_long)
+            __alloc_traits::deallocate(__alloc(), __p, __cap+1);
+        if (__now_long)
+        {
+            __set_long_cap(__res_arg+1);
+            __set_long_size(__sz);
+            __set_long_pointer(__new_data);
+        }
+        else
+            __set_short_size(__sz);
+        __invalidate_all_iterators();
+    }
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::const_reference
+basic_string<_CharT, _Traits, _Allocator>::operator[](size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__pos <= size(), "string index out of bounds");
+    return *(data() + __pos);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::reference
+basic_string<_CharT, _Traits, _Allocator>::operator[](size_type __pos) _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__pos <= size(), "string index out of bounds");
+    return *(__get_pointer() + __pos);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::const_reference
+basic_string<_CharT, _Traits, _Allocator>::at(size_type __n) const
+{
+    if (__n >= size())
+        this->__throw_out_of_range();
+    return (*this)[__n];
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::reference
+basic_string<_CharT, _Traits, _Allocator>::at(size_type __n)
+{
+    if (__n >= size())
+        this->__throw_out_of_range();
+    return (*this)[__n];
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::reference
+basic_string<_CharT, _Traits, _Allocator>::front() _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(!empty(), "string::front(): string is empty");
+    return *__get_pointer();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::const_reference
+basic_string<_CharT, _Traits, _Allocator>::front() const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(!empty(), "string::front(): string is empty");
+    return *data();
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::reference
+basic_string<_CharT, _Traits, _Allocator>::back() _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(!empty(), "string::back(): string is empty");
+    return *(__get_pointer() + size() - 1);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::const_reference
+basic_string<_CharT, _Traits, _Allocator>::back() const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(!empty(), "string::back(): string is empty");
+    return *(data() + size() - 1);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::copy(value_type* __s, size_type __n, size_type __pos) const
+{
+    size_type __sz = size();
+    if (__pos > __sz)
+        this->__throw_out_of_range();
+    size_type __rlen = _CUDA_VSTD::min(__n, __sz - __pos);
+    traits_type::copy(__s, data() + __pos, __rlen);
+    return __rlen;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>
+basic_string<_CharT, _Traits, _Allocator>::substr(size_type __pos, size_type __n) const
+{
+    return basic_string(*this, __pos, __n, __alloc());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::swap(basic_string& __str)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value)
+#endif
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    if (!__is_long())
+        __get_db()->__invalidate_all(this);
+    if (!__str.__is_long())
+        __get_db()->__invalidate_all(&__str);
+    __get_db()->swap(this, &__str);
+#endif
+    _LIBCUDACXX_ASSERT(
+        __alloc_traits::propagate_on_container_swap::value ||
+        __alloc_traits::is_always_equal::value ||
+        __alloc() == __str.__alloc(), "swapping non-equal allocators");
+    _CUDA_VSTD::swap(__r_.first(), __str.__r_.first());
+    __swap_allocator(__alloc(), __str.__alloc());
+}
+
+// find
+
+template <class _Traits>
+struct _LIBCUDACXX_HIDDEN __traits_eq
+{
+    typedef typename _Traits::char_type char_type;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const char_type& __x, const char_type& __y) _NOEXCEPT
+        {return _Traits::eq(__x, __y);}
+};
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find(const value_type* __s,
+                                                size_type __pos,
+                                                size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::find(): received nullptr");
+    return __str_find<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, __n);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find(const basic_string& __str,
+                                                size_type __pos) const _NOEXCEPT
+{
+    return __str_find<value_type, size_type, traits_type, npos>
+        (data(), size(), __str.data(), __pos, __str.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type
+>::type
+basic_string<_CharT, _Traits, _Allocator>::find(const _Tp &__t,
+                                                size_type __pos) const
+{
+    __self_view __sv = __t;
+    return __str_find<value_type, size_type, traits_type, npos>
+        (data(), size(), __sv.data(), __pos, __sv.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find(const value_type* __s,
+                                                size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::find(): received nullptr");
+    return __str_find<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, traits_type::length(__s));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find(value_type __c,
+                                                size_type __pos) const _NOEXCEPT
+{
+    return __str_find<value_type, size_type, traits_type, npos>
+        (data(), size(), __c, __pos);
+}
+
+// rfind
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::rfind(const value_type* __s,
+                                                 size_type __pos,
+                                                 size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::rfind(): received nullptr");
+    return __str_rfind<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, __n);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::rfind(const basic_string& __str,
+                                                 size_type __pos) const _NOEXCEPT
+{
+    return __str_rfind<value_type, size_type, traits_type, npos>
+        (data(), size(), __str.data(), __pos, __str.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type
+>::type
+basic_string<_CharT, _Traits, _Allocator>::rfind(const _Tp& __t,
+                                                size_type __pos) const
+{
+    __self_view __sv = __t;
+    return __str_rfind<value_type, size_type, traits_type, npos>
+        (data(), size(), __sv.data(), __pos, __sv.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::rfind(const value_type* __s,
+                                                 size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::rfind(): received nullptr");
+    return __str_rfind<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, traits_type::length(__s));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::rfind(value_type __c,
+                                                 size_type __pos) const _NOEXCEPT
+{
+    return __str_rfind<value_type, size_type, traits_type, npos>
+        (data(), size(), __c, __pos);
+}
+
+// find_first_of
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_of(const value_type* __s,
+                                                         size_type __pos,
+                                                         size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::find_first_of(): received nullptr");
+    return __str_find_first_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, __n);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_of(const basic_string& __str,
+                                                         size_type __pos) const _NOEXCEPT
+{
+    return __str_find_first_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __str.data(), __pos, __str.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type
+>::type
+basic_string<_CharT, _Traits, _Allocator>::find_first_of(const _Tp& __t,
+                                                size_type __pos) const
+{
+    __self_view __sv = __t;
+    return __str_find_first_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __sv.data(), __pos, __sv.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_of(const value_type* __s,
+                                                         size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::find_first_of(): received nullptr");
+    return __str_find_first_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, traits_type::length(__s));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_of(value_type __c,
+                                                         size_type __pos) const _NOEXCEPT
+{
+    return find(__c, __pos);
+}
+
+// find_last_of
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_of(const value_type* __s,
+                                                        size_type __pos,
+                                                        size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::find_last_of(): received nullptr");
+    return __str_find_last_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, __n);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_of(const basic_string& __str,
+                                                        size_type __pos) const _NOEXCEPT
+{
+    return __str_find_last_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __str.data(), __pos, __str.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type
+>::type
+basic_string<_CharT, _Traits, _Allocator>::find_last_of(const _Tp& __t,
+                                                size_type __pos) const
+{
+    __self_view __sv = __t;
+    return __str_find_last_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __sv.data(), __pos, __sv.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_of(const value_type* __s,
+                                                        size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::find_last_of(): received nullptr");
+    return __str_find_last_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, traits_type::length(__s));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_of(value_type __c,
+                                                        size_type __pos) const _NOEXCEPT
+{
+    return rfind(__c, __pos);
+}
+
+// find_first_not_of
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_not_of(const value_type* __s,
+                                                             size_type __pos,
+                                                             size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::find_first_not_of(): received nullptr");
+    return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, __n);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_not_of(const basic_string& __str,
+                                                             size_type __pos) const _NOEXCEPT
+{
+    return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __str.data(), __pos, __str.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type
+>::type
+basic_string<_CharT, _Traits, _Allocator>::find_first_not_of(const _Tp& __t,
+                                                size_type __pos) const
+{
+    __self_view __sv = __t;
+    return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __sv.data(), __pos, __sv.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_not_of(const value_type* __s,
+                                                             size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::find_first_not_of(): received nullptr");
+    return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, traits_type::length(__s));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_first_not_of(value_type __c,
+                                                             size_type __pos) const _NOEXCEPT
+{
+    return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __c, __pos);
+}
+
+// find_last_not_of
+
+template<class _CharT, class _Traits, class _Allocator>
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_not_of(const value_type* __s,
+                                                            size_type __pos,
+                                                            size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string::find_last_not_of(): received nullptr");
+    return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, __n);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_not_of(const basic_string& __str,
+                                                            size_type __pos) const _NOEXCEPT
+{
+    return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __str.data(), __pos, __str.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type
+>::type
+basic_string<_CharT, _Traits, _Allocator>::find_last_not_of(const _Tp& __t,
+                                                size_type __pos) const
+{
+    __self_view __sv = __t;
+    return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __sv.data(), __pos, __sv.size());
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_not_of(const value_type* __s,
+                                                            size_type __pos) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::find_last_not_of(): received nullptr");
+    return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __s, __pos, traits_type::length(__s));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+typename basic_string<_CharT, _Traits, _Allocator>::size_type
+basic_string<_CharT, _Traits, _Allocator>::find_last_not_of(value_type __c,
+                                                            size_type __pos) const _NOEXCEPT
+{
+    return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+        (data(), size(), __c, __pos);
+}
+
+// compare
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    int
+>::type
+basic_string<_CharT, _Traits, _Allocator>::compare(const _Tp& __t) const
+{
+    __self_view __sv = __t;
+    size_t __lhs_sz = size();
+    size_t __rhs_sz = __sv.size();
+    int __result = traits_type::compare(data(), __sv.data(),
+                                        _CUDA_VSTD::min(__lhs_sz, __rhs_sz));
+    if (__result != 0)
+        return __result;
+    if (__lhs_sz < __rhs_sz)
+        return -1;
+    if (__lhs_sz > __rhs_sz)
+        return 1;
+    return 0;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+int
+basic_string<_CharT, _Traits, _Allocator>::compare(const basic_string& __str) const _NOEXCEPT
+{
+    return compare(__self_view(__str));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+int
+basic_string<_CharT, _Traits, _Allocator>::compare(size_type __pos1,
+                                                   size_type __n1,
+                                                   const value_type* __s,
+                                                   size_type __n2) const
+{
+    _LIBCUDACXX_ASSERT(__n2 == 0 || __s != nullptr, "string::compare(): received nullptr");
+    size_type __sz = size();
+    if (__pos1 > __sz || __n2 == npos)
+        this->__throw_out_of_range();
+    size_type __rlen = _CUDA_VSTD::min(__n1, __sz - __pos1);
+    int __r = traits_type::compare(data() + __pos1, __s, _CUDA_VSTD::min(__rlen, __n2));
+    if (__r == 0)
+    {
+        if (__rlen < __n2)
+            __r = -1;
+        else if (__rlen > __n2)
+            __r = 1;
+    }
+    return __r;
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    int
+>::type
+basic_string<_CharT, _Traits, _Allocator>::compare(size_type __pos1,
+                                                   size_type __n1,
+                                                   const _Tp& __t) const
+{
+    __self_view __sv = __t;
+    return compare(__pos1, __n1, __sv.data(), __sv.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+inline
+int
+basic_string<_CharT, _Traits, _Allocator>::compare(size_type __pos1,
+                                                   size_type __n1,
+                                                   const basic_string& __str) const
+{
+    return compare(__pos1, __n1, __str.data(), __str.size());
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+template <class _Tp>
+typename enable_if
+<
+    __can_be_converted_to_string_view<_CharT, _Traits, _Tp>::value,
+    int
+>::type
+basic_string<_CharT, _Traits, _Allocator>::compare(size_type __pos1,
+                                                   size_type __n1,
+                                                   const _Tp& __t,
+                                                   size_type __pos2,
+                                                   size_type __n2) const
+{
+    __self_view __sv = __t;
+    return __self_view(*this).substr(__pos1, __n1).compare(__sv.substr(__pos2, __n2));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+int
+basic_string<_CharT, _Traits, _Allocator>::compare(size_type __pos1,
+                                                   size_type __n1,
+                                                   const basic_string& __str,
+                                                   size_type __pos2,
+                                                   size_type __n2) const
+{
+        return compare(__pos1, __n1, __self_view(__str), __pos2, __n2);
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+int
+basic_string<_CharT, _Traits, _Allocator>::compare(const value_type* __s) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::compare(): received nullptr");
+    return compare(0, npos, __s, traits_type::length(__s));
+}
+
+template <class _CharT, class _Traits, class _Allocator>
+int
+basic_string<_CharT, _Traits, _Allocator>::compare(size_type __pos1,
+                                                   size_type __n1,
+                                                   const value_type* __s) const
+{
+    _LIBCUDACXX_ASSERT(__s != nullptr, "string::compare(): received nullptr");
+    return compare(__pos1, __n1, __s, traits_type::length(__s));
+}
+
+// __invariants
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+bool
+basic_string<_CharT, _Traits, _Allocator>::__invariants() const
+{
+    if (size() > capacity())
+        return false;
+    if (capacity() < __min_cap - 1)
+        return false;
+    if (data() == 0)
+        return false;
+    if (data()[size()] != value_type(0))
+        return false;
+    return true;
+}
+
+// __clear_and_shrink
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+void
+basic_string<_CharT, _Traits, _Allocator>::__clear_and_shrink() _NOEXCEPT
+{
+    clear();
+    if(__is_long())
+    {
+        __alloc_traits::deallocate(__alloc(), __get_long_pointer(), capacity() + 1);
+        __set_long_cap(0);
+        __set_short_size(0);
+    }
+}
+
+// operator==
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    size_t __lhs_sz = __lhs.size();
+    return __lhs_sz == __rhs.size() && _Traits::compare(__lhs.data(),
+                                                        __rhs.data(),
+                                                        __lhs_sz) == 0;
+}
+
+template<class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const basic_string<char, char_traits<char>, _Allocator>& __lhs,
+           const basic_string<char, char_traits<char>, _Allocator>& __rhs) _NOEXCEPT
+{
+    size_t __lhs_sz = __lhs.size();
+    if (__lhs_sz != __rhs.size())
+        return false;
+    const char* __lp = __lhs.data();
+    const char* __rp = __rhs.data();
+    if (__lhs.__is_long())
+        return char_traits<char>::compare(__lp, __rp, __lhs_sz) == 0;
+    for (; __lhs_sz != 0; --__lhs_sz, ++__lp, ++__rp)
+        if (*__lp != *__rp)
+            return false;
+    return true;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const _CharT* __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    typedef basic_string<_CharT, _Traits, _Allocator> _String;
+    _LIBCUDACXX_ASSERT(__lhs != nullptr, "operator==(char*, basic_string): received nullptr");
+    size_t __lhs_len = _Traits::length(__lhs);
+    if (__lhs_len != __rhs.size()) return false;
+    return __rhs.compare(0, _String::npos, __lhs, __lhs_len) == 0;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const basic_string<_CharT,_Traits,_Allocator>& __lhs,
+           const _CharT* __rhs) _NOEXCEPT
+{
+    typedef basic_string<_CharT, _Traits, _Allocator> _String;
+    _LIBCUDACXX_ASSERT(__rhs != nullptr, "operator==(basic_string, char*): received nullptr");
+    size_t __rhs_len = _Traits::length(__rhs);
+    if (__rhs_len != __lhs.size()) return false;
+    return __lhs.compare(0, _String::npos, __rhs, __rhs_len) == 0;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const basic_string<_CharT,_Traits,_Allocator>& __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return !(__lhs == __rhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const _CharT* __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return !(__lhs == __rhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const _CharT* __rhs) _NOEXCEPT
+{
+    return !(__lhs == __rhs);
+}
+
+// operator<
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) < 0;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const _CharT* __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) < 0;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const _CharT* __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return __rhs.compare(__lhs) > 0;
+}
+
+// operator>
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return __rhs < __lhs;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const _CharT* __rhs) _NOEXCEPT
+{
+    return __rhs < __lhs;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const _CharT* __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return __rhs < __lhs;
+}
+
+// operator<=
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return !(__rhs < __lhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const _CharT* __rhs) _NOEXCEPT
+{
+    return !(__rhs < __lhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const _CharT* __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return !(__rhs < __lhs);
+}
+
+// operator>=
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return !(__lhs < __rhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+           const _CharT* __rhs) _NOEXCEPT
+{
+    return !(__lhs < __rhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const _CharT* __lhs,
+           const basic_string<_CharT, _Traits, _Allocator>& __rhs) _NOEXCEPT
+{
+    return !(__lhs < __rhs);
+}
+
+// operator +
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __lhs,
+          const basic_string<_CharT, _Traits, _Allocator>& __rhs)
+{
+    basic_string<_CharT, _Traits, _Allocator> __r(__lhs.get_allocator());
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __lhs_sz = __lhs.size();
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __rhs_sz = __rhs.size();
+    __r.__init(__lhs.data(), __lhs_sz, __lhs_sz + __rhs_sz);
+    __r.append(__rhs.data(), __rhs_sz);
+    return __r;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const _CharT* __lhs , const basic_string<_CharT,_Traits,_Allocator>& __rhs)
+{
+    basic_string<_CharT, _Traits, _Allocator> __r(__rhs.get_allocator());
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __lhs_sz = _Traits::length(__lhs);
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __rhs_sz = __rhs.size();
+    __r.__init(__lhs, __lhs_sz, __lhs_sz + __rhs_sz);
+    __r.append(__rhs.data(), __rhs_sz);
+    return __r;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(_CharT __lhs, const basic_string<_CharT,_Traits,_Allocator>& __rhs)
+{
+    basic_string<_CharT, _Traits, _Allocator> __r(__rhs.get_allocator());
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __rhs_sz = __rhs.size();
+    __r.__init(&__lhs, 1, 1 + __rhs_sz);
+    __r.append(__rhs.data(), __rhs_sz);
+    return __r;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __lhs, const _CharT* __rhs)
+{
+    basic_string<_CharT, _Traits, _Allocator> __r(__lhs.get_allocator());
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __lhs_sz = __lhs.size();
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __rhs_sz = _Traits::length(__rhs);
+    __r.__init(__lhs.data(), __lhs_sz, __lhs_sz + __rhs_sz);
+    __r.append(__rhs, __rhs_sz);
+    return __r;
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __lhs, _CharT __rhs)
+{
+    basic_string<_CharT, _Traits, _Allocator> __r(__lhs.get_allocator());
+    typename basic_string<_CharT, _Traits, _Allocator>::size_type __lhs_sz = __lhs.size();
+    __r.__init(__lhs.data(), __lhs_sz, __lhs_sz + 1);
+    __r.push_back(__rhs);
+    return __r;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(basic_string<_CharT, _Traits, _Allocator>&& __lhs, const basic_string<_CharT, _Traits, _Allocator>& __rhs)
+{
+    return _CUDA_VSTD::move(__lhs.append(__rhs));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const basic_string<_CharT, _Traits, _Allocator>& __lhs, basic_string<_CharT, _Traits, _Allocator>&& __rhs)
+{
+    return _CUDA_VSTD::move(__rhs.insert(0, __lhs));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(basic_string<_CharT, _Traits, _Allocator>&& __lhs, basic_string<_CharT, _Traits, _Allocator>&& __rhs)
+{
+    return _CUDA_VSTD::move(__lhs.append(__rhs));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(const _CharT* __lhs , basic_string<_CharT,_Traits,_Allocator>&& __rhs)
+{
+    return _CUDA_VSTD::move(__rhs.insert(0, __lhs));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(_CharT __lhs, basic_string<_CharT,_Traits,_Allocator>&& __rhs)
+{
+    __rhs.insert(__rhs.begin(), __lhs);
+    return _CUDA_VSTD::move(__rhs);
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(basic_string<_CharT, _Traits, _Allocator>&& __lhs, const _CharT* __rhs)
+{
+    return _CUDA_VSTD::move(__lhs.append(__rhs));
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_string<_CharT, _Traits, _Allocator>
+operator+(basic_string<_CharT, _Traits, _Allocator>&& __lhs, _CharT __rhs)
+{
+    __lhs.push_back(__rhs);
+    return _CUDA_VSTD::move(__lhs);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+// swap
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(basic_string<_CharT, _Traits, _Allocator>& __lhs,
+     basic_string<_CharT, _Traits, _Allocator>& __rhs)
+     _NOEXCEPT_(_NOEXCEPT_(__lhs.swap(__rhs)))
+{
+    __lhs.swap(__rhs);
+}
+
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+typedef basic_string<char8_t> u8string;
+#endif
+
+#ifndef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+typedef basic_string<char16_t> u16string;
+typedef basic_string<char32_t> u32string;
+#endif  // _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+
+_LIBCUDACXX_FUNC_VIS int                stoi  (const string& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS long               stol  (const string& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS unsigned long      stoul (const string& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS long long          stoll (const string& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS unsigned long long stoull(const string& __str, size_t* __idx = 0, int __base = 10);
+
+_LIBCUDACXX_FUNC_VIS float       stof (const string& __str, size_t* __idx = 0);
+_LIBCUDACXX_FUNC_VIS double      stod (const string& __str, size_t* __idx = 0);
+_LIBCUDACXX_FUNC_VIS long double stold(const string& __str, size_t* __idx = 0);
+
+_LIBCUDACXX_FUNC_VIS string to_string(int __val);
+_LIBCUDACXX_FUNC_VIS string to_string(unsigned __val);
+_LIBCUDACXX_FUNC_VIS string to_string(long __val);
+_LIBCUDACXX_FUNC_VIS string to_string(unsigned long __val);
+_LIBCUDACXX_FUNC_VIS string to_string(long long __val);
+_LIBCUDACXX_FUNC_VIS string to_string(unsigned long long __val);
+_LIBCUDACXX_FUNC_VIS string to_string(float __val);
+_LIBCUDACXX_FUNC_VIS string to_string(double __val);
+_LIBCUDACXX_FUNC_VIS string to_string(long double __val);
+
+_LIBCUDACXX_FUNC_VIS int                stoi  (const wstring& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS long               stol  (const wstring& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS unsigned long      stoul (const wstring& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS long long          stoll (const wstring& __str, size_t* __idx = 0, int __base = 10);
+_LIBCUDACXX_FUNC_VIS unsigned long long stoull(const wstring& __str, size_t* __idx = 0, int __base = 10);
+
+_LIBCUDACXX_FUNC_VIS float       stof (const wstring& __str, size_t* __idx = 0);
+_LIBCUDACXX_FUNC_VIS double      stod (const wstring& __str, size_t* __idx = 0);
+_LIBCUDACXX_FUNC_VIS long double stold(const wstring& __str, size_t* __idx = 0);
+
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(int __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(unsigned __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(long __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(unsigned long __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(long long __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(unsigned long long __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(float __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(double __val);
+_LIBCUDACXX_FUNC_VIS wstring to_wstring(long double __val);
+
+template<class _CharT, class _Traits, class _Allocator>
+    const typename basic_string<_CharT, _Traits, _Allocator>::size_type
+                   basic_string<_CharT, _Traits, _Allocator>::npos;
+
+template <class _CharT, class _Allocator>
+struct _LIBCUDACXX_TEMPLATE_VIS
+    hash<basic_string<_CharT, char_traits<_CharT>, _Allocator> >
+    : public unary_function<
+          basic_string<_CharT, char_traits<_CharT>, _Allocator>, size_t>
+{
+    size_t
+    operator()(const basic_string<_CharT, char_traits<_CharT>, _Allocator>& __val) const _NOEXCEPT
+    { return __do_string_hash(__val.data(), __val.data() + __val.size()); }
+};
+
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           const basic_string<_CharT, _Traits, _Allocator>& __str);
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_istream<_CharT, _Traits>&
+operator>>(basic_istream<_CharT, _Traits>& __is,
+           basic_string<_CharT, _Traits, _Allocator>& __str);
+
+template<class _CharT, class _Traits, class _Allocator>
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str, _CharT __dlm);
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>&& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str, _CharT __dlm);
+
+template<class _CharT, class _Traits, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+basic_istream<_CharT, _Traits>&
+getline(basic_istream<_CharT, _Traits>&& __is,
+        basic_string<_CharT, _Traits, _Allocator>& __str);
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 17
+template<class _CharT, class _Traits, class _Allocator, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase(basic_string<_CharT, _Traits, _Allocator>& __str, const _Up& __v)
+{ __str.erase(_CUDA_VSTD::remove(__str.begin(), __str.end(), __v), __str.end()); }
+
+template<class _CharT, class _Traits, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(basic_string<_CharT, _Traits, _Allocator>& __str, _Predicate __pred)
+{ __str.erase(_CUDA_VSTD::remove_if(__str.begin(), __str.end(), __pred), __str.end()); }
+#endif
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+template<class _CharT, class _Traits, class _Allocator>
+bool
+basic_string<_CharT, _Traits, _Allocator>::__dereferenceable(const const_iterator* __i) const
+{
+    return this->data() <= _CUDA_VSTD::__to_raw_pointer(__i->base()) &&
+           _CUDA_VSTD::__to_raw_pointer(__i->base()) < this->data() + this->size();
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+bool
+basic_string<_CharT, _Traits, _Allocator>::__decrementable(const const_iterator* __i) const
+{
+    return this->data() < _CUDA_VSTD::__to_raw_pointer(__i->base()) &&
+           _CUDA_VSTD::__to_raw_pointer(__i->base()) <= this->data() + this->size();
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+bool
+basic_string<_CharT, _Traits, _Allocator>::__addable(const const_iterator* __i, ptrdiff_t __n) const
+{
+    const value_type* __p = _CUDA_VSTD::__to_raw_pointer(__i->base()) + __n;
+    return this->data() <= __p && __p <= this->data() + this->size();
+}
+
+template<class _CharT, class _Traits, class _Allocator>
+bool
+basic_string<_CharT, _Traits, _Allocator>::__subscriptable(const const_iterator* __i, ptrdiff_t __n) const
+{
+    const value_type* __p = _CUDA_VSTD::__to_raw_pointer(__i->base()) + __n;
+    return this->data() <= __p && __p < this->data() + this->size();
+}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_string<char>)
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS basic_string<wchar_t>)
+
+#if _LIBCUDACXX_STD_VER > 11
+// Literal suffixes for basic_string [basic.string.literals]
+inline namespace literals
+{
+  inline namespace string_literals
+  {
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string<char> operator "" s( const char *__str, size_t __len )
+    {
+        return basic_string<char> (__str, __len);
+    }
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string<wchar_t> operator "" s( const wchar_t *__str, size_t __len )
+    {
+        return basic_string<wchar_t> (__str, __len);
+    }
+
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string<char8_t> operator "" s(const char8_t *__str, size_t __len) _NOEXCEPT
+    {
+        return basic_string<char8_t> (__str, __len);
+    }
+#endif
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string<char16_t> operator "" s( const char16_t *__str, size_t __len )
+    {
+        return basic_string<char16_t> (__str, __len);
+    }
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string<char32_t> operator "" s( const char32_t *__str, size_t __len )
+    {
+        return basic_string<char32_t> (__str, __len);
+    }
+  }
+}
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_STRING
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string.h
new file mode 100644
index 000000000000..f05914478546
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string.h
@@ -0,0 +1,109 @@
+// -*- C++ -*-
+//===--------------------------- string.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STRING_H
+#define _LIBCUDACXX_STRING_H
+
+/*
+    string.h synopsis
+
+Macros:
+
+    NULL
+
+Types:
+
+    size_t
+
+void* memcpy(void* restrict s1, const void* restrict s2, size_t n);
+void* memmove(void* s1, const void* s2, size_t n);
+char* strcpy (char* restrict s1, const char* restrict s2);
+char* strncpy(char* restrict s1, const char* restrict s2, size_t n);
+char* strcat (char* restrict s1, const char* restrict s2);
+char* strncat(char* restrict s1, const char* restrict s2, size_t n);
+int memcmp(const void* s1, const void* s2, size_t n);
+int strcmp (const char* s1, const char* s2);
+int strncmp(const char* s1, const char* s2, size_t n);
+int strcoll(const char* s1, const char* s2);
+size_t strxfrm(char* restrict s1, const char* restrict s2, size_t n);
+const void* memchr(const void* s, int c, size_t n);
+      void* memchr(      void* s, int c, size_t n);
+const char* strchr(const char* s, int c);
+      char* strchr(      char* s, int c);
+size_t strcspn(const char* s1, const char* s2);
+const char* strpbrk(const char* s1, const char* s2);
+      char* strpbrk(      char* s1, const char* s2);
+const char* strrchr(const char* s, int c);
+      char* strrchr(      char* s, int c);
+size_t strspn(const char* s1, const char* s2);
+const char* strstr(const char* s1, const char* s2);
+      char* strstr(      char* s1, const char* s2);
+char* strtok(char* restrict s1, const char* restrict s2);
+void* memset(void* s, int c, size_t n);
+char* strerror(int errnum);
+size_t strlen(const char* s);
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <string.h>
+
+// MSVCRT, GNU libc and its derivates may already have the correct prototype in
+// <string.h>. This macro can be defined by users if their C library provides
+// the right signature.
+#if defined(__CORRECT_ISO_CPP_STRING_H_PROTO) || defined(_LIBCUDACXX_MSVCRT) || \
+    defined(__sun__) || defined(_STRING_H_CPLUSPLUS_98_CONFORMANCE_)
+#define _LIBCUDACXX_STRING_H_HAS_CONST_OVERLOADS
+#endif
+
+#if defined(__cplusplus) && !defined(_LIBCUDACXX_STRING_H_HAS_CONST_OVERLOADS) && defined(_LIBCUDACXX_PREFERRED_OVERLOAD)
+extern "C++" {
+inline _LIBCUDACXX_INLINE_VISIBILITY
+char* __libcpp_strchr(const char* __s, int __c) {return (char*)strchr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const char* strchr(const char* __s, int __c) {return __libcpp_strchr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      char* strchr(      char* __s, int __c) {return __libcpp_strchr(__s, __c);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+char* __libcpp_strpbrk(const char* __s1, const char* __s2) {return (char*)strpbrk(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const char* strpbrk(const char* __s1, const char* __s2) {return __libcpp_strpbrk(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      char* strpbrk(      char* __s1, const char* __s2) {return __libcpp_strpbrk(__s1, __s2);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+char* __libcpp_strrchr(const char* __s, int __c) {return (char*)strrchr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const char* strrchr(const char* __s, int __c) {return __libcpp_strrchr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      char* strrchr(      char* __s, int __c) {return __libcpp_strrchr(__s, __c);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void* __libcpp_memchr(const void* __s, int __c, size_t __n) {return (void*)memchr(__s, __c, __n);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const void* memchr(const void* __s, int __c, size_t __n) {return __libcpp_memchr(__s, __c, __n);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      void* memchr(      void* __s, int __c, size_t __n) {return __libcpp_memchr(__s, __c, __n);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+char* __libcpp_strstr(const char* __s1, const char* __s2) {return (char*)strstr(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const char* strstr(const char* __s1, const char* __s2) {return __libcpp_strstr(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      char* strstr(      char* __s1, const char* __s2) {return __libcpp_strstr(__s1, __s2);}
+}
+#endif
+
+#endif  // _LIBCUDACXX_STRING_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string_view b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string_view
new file mode 100644
index 000000000000..0d5dec6ace88
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/string_view
@@ -0,0 +1,840 @@
+// -*- C++ -*-
+//===------------------------ string_view ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STRING_VIEW
+#define _LIBCUDACXX_STRING_VIEW
+
+/*
+string_view synopsis
+
+namespace std {
+
+    // 7.2, Class template basic_string_view
+    template<class charT, class traits = char_traits<charT>>
+        class basic_string_view;
+
+    // 7.9, basic_string_view non-member comparison functions
+    template<class charT, class traits>
+    constexpr bool operator==(basic_string_view<charT, traits> x,
+                              basic_string_view<charT, traits> y) noexcept;
+    template<class charT, class traits>
+    constexpr bool operator!=(basic_string_view<charT, traits> x,
+                              basic_string_view<charT, traits> y) noexcept;
+    template<class charT, class traits>
+    constexpr bool operator< (basic_string_view<charT, traits> x,
+                                 basic_string_view<charT, traits> y) noexcept;
+    template<class charT, class traits>
+    constexpr bool operator> (basic_string_view<charT, traits> x,
+                              basic_string_view<charT, traits> y) noexcept;
+    template<class charT, class traits>
+    constexpr bool operator<=(basic_string_view<charT, traits> x,
+                                 basic_string_view<charT, traits> y) noexcept;
+    template<class charT, class traits>
+    constexpr bool operator>=(basic_string_view<charT, traits> x,
+                              basic_string_view<charT, traits> y) noexcept;
+    // see below, sufficient additional overloads of comparison functions
+
+    // 7.10, Inserters and extractors
+    template<class charT, class traits>
+      basic_ostream<charT, traits>&
+        operator<<(basic_ostream<charT, traits>& os,
+                   basic_string_view<charT, traits> str);
+
+    // basic_string_view typedef names
+    typedef basic_string_view<char> string_view;
+    typedef basic_string_view<char16_t> u16string_view;
+    typedef basic_string_view<char32_t> u32string_view;
+    typedef basic_string_view<wchar_t> wstring_view;
+
+    template<class charT, class traits = char_traits<charT>>
+    class basic_string_view {
+      public:
+      // types
+      typedef traits traits_type;
+      typedef charT value_type;
+      typedef charT* pointer;
+      typedef const charT* const_pointer;
+      typedef charT& reference;
+      typedef const charT& const_reference;
+      typedef implementation-defined const_iterator;
+      typedef const_iterator iterator;
+      typedef reverse_iterator<const_iterator> const_reverse_iterator;
+      typedef const_reverse_iterator reverse_iterator;
+      typedef size_t size_type;
+      typedef ptrdiff_t difference_type;
+      static constexpr size_type npos = size_type(-1);
+
+      // 7.3, basic_string_view constructors and assignment operators
+      constexpr basic_string_view() noexcept;
+      constexpr basic_string_view(const basic_string_view&) noexcept = default;
+      basic_string_view& operator=(const basic_string_view&) noexcept = default;
+      template<class Allocator>
+      constexpr basic_string_view(const charT* str);
+      constexpr basic_string_view(const charT* str, size_type len);
+
+      // 7.4, basic_string_view iterator support
+      constexpr const_iterator begin() const noexcept;
+      constexpr const_iterator end() const noexcept;
+      constexpr const_iterator cbegin() const noexcept;
+      constexpr const_iterator cend() const noexcept;
+      const_reverse_iterator rbegin() const noexcept;
+      const_reverse_iterator rend() const noexcept;
+      const_reverse_iterator crbegin() const noexcept;
+      const_reverse_iterator crend() const noexcept;
+
+      // 7.5, basic_string_view capacity
+      constexpr size_type size() const noexcept;
+      constexpr size_type length() const noexcept;
+      constexpr size_type max_size() const noexcept;
+      constexpr bool empty() const noexcept;
+
+      // 7.6, basic_string_view element access
+      constexpr const_reference operator[](size_type pos) const;
+      constexpr const_reference at(size_type pos) const;
+      constexpr const_reference front() const;
+      constexpr const_reference back() const;
+      constexpr const_pointer data() const noexcept;
+
+      // 7.7, basic_string_view modifiers
+      constexpr void remove_prefix(size_type n);
+      constexpr void remove_suffix(size_type n);
+      constexpr void swap(basic_string_view& s) noexcept;
+
+      size_type copy(charT* s, size_type n, size_type pos = 0) const;
+
+      constexpr basic_string_view substr(size_type pos = 0, size_type n = npos) const;
+      constexpr int compare(basic_string_view s) const noexcept;
+      constexpr int compare(size_type pos1, size_type n1, basic_string_view s) const;
+      constexpr int compare(size_type pos1, size_type n1,
+                            basic_string_view s, size_type pos2, size_type n2) const;
+      constexpr int compare(const charT* s) const;
+      constexpr int compare(size_type pos1, size_type n1, const charT* s) const;
+      constexpr int compare(size_type pos1, size_type n1,
+                            const charT* s, size_type n2) const;
+      constexpr size_type find(basic_string_view s, size_type pos = 0) const noexcept;
+      constexpr size_type find(charT c, size_type pos = 0) const noexcept;
+      constexpr size_type find(const charT* s, size_type pos, size_type n) const;
+      constexpr size_type find(const charT* s, size_type pos = 0) const;
+      constexpr size_type rfind(basic_string_view s, size_type pos = npos) const noexcept;
+      constexpr size_type rfind(charT c, size_type pos = npos) const noexcept;
+      constexpr size_type rfind(const charT* s, size_type pos, size_type n) const;
+      constexpr size_type rfind(const charT* s, size_type pos = npos) const;
+      constexpr size_type find_first_of(basic_string_view s, size_type pos = 0) const noexcept;
+      constexpr size_type find_first_of(charT c, size_type pos = 0) const noexcept;
+      constexpr size_type find_first_of(const charT* s, size_type pos, size_type n) const;
+      constexpr size_type find_first_of(const charT* s, size_type pos = 0) const;
+      constexpr size_type find_last_of(basic_string_view s, size_type pos = npos) const noexcept;
+      constexpr size_type find_last_of(charT c, size_type pos = npos) const noexcept;
+      constexpr size_type find_last_of(const charT* s, size_type pos, size_type n) const;
+      constexpr size_type find_last_of(const charT* s, size_type pos = npos) const;
+      constexpr size_type find_first_not_of(basic_string_view s, size_type pos = 0) const noexcept;
+      constexpr size_type find_first_not_of(charT c, size_type pos = 0) const noexcept;
+      constexpr size_type find_first_not_of(const charT* s, size_type pos, size_type n) const;
+      constexpr size_type find_first_not_of(const charT* s, size_type pos = 0) const;
+      constexpr size_type find_last_not_of(basic_string_view s, size_type pos = npos) const noexcept;
+      constexpr size_type find_last_not_of(charT c, size_type pos = npos) const noexcept;
+      constexpr size_type find_last_not_of(const charT* s, size_type pos, size_type n) const;
+      constexpr size_type find_last_not_of(const charT* s, size_type pos = npos) const;
+
+      constexpr bool starts_with(basic_string_view s) const noexcept; // C++2a
+      constexpr bool starts_with(charT c) const noexcept;             // C++2a
+      constexpr bool starts_with(const charT* s) const;               // C++2a
+      constexpr bool ends_with(basic_string_view s) const noexcept;   // C++2a
+      constexpr bool ends_with(charT c) const noexcept;               // C++2a
+      constexpr bool ends_with(const charT* s) const;                 // C++2a
+
+     private:
+      const_pointer data_;  // exposition only
+      size_type     size_;  // exposition only
+    };
+
+  // 7.11, Hash support
+  template <class T> struct hash;
+  template <> struct hash<string_view>;
+  template <> struct hash<u16string_view>;
+  template <> struct hash<u32string_view>;
+  template <> struct hash<wstring_view>;
+
+  constexpr basic_string_view<char>     operator "" sv( const char *str,     size_t len ) noexcept;
+  constexpr basic_string_view<wchar_t>  operator "" sv( const wchar_t *str,  size_t len ) noexcept;
+  constexpr basic_string_view<char16_t> operator "" sv( const char16_t *str, size_t len ) noexcept;
+  constexpr basic_string_view<char32_t> operator "" sv( const char32_t *str, size_t len ) noexcept;
+
+}  // namespace std
+
+
+*/
+
+#include <__config>
+#include <__string>
+#include <iosfwd>
+#include <algorithm>
+#include <iterator>
+#include <limits>
+#include <stdexcept>
+#include <version>
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template<class _CharT, class _Traits = char_traits<_CharT> >
+class _LIBCUDACXX_TEMPLATE_VIS basic_string_view {
+public:
+    // types
+    typedef _Traits                                    traits_type;
+    typedef _CharT                                     value_type;
+    typedef _CharT*                                    pointer;
+    typedef const _CharT*                              const_pointer;
+    typedef _CharT&                                    reference;
+    typedef const _CharT&                              const_reference;
+    typedef const_pointer                              const_iterator; // See [string.view.iterators]
+    typedef const_iterator                             iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>    const_reverse_iterator;
+    typedef const_reverse_iterator                     reverse_iterator;
+    typedef size_t                                     size_type;
+    typedef ptrdiff_t                                  difference_type;
+    static _LIBCUDACXX_CONSTEXPR const size_type npos = -1; // size_type(-1);
+
+    static_assert((!is_array<value_type>::value), "Character type of basic_string_view must not be an array");
+    static_assert(( is_standard_layout<value_type>::value), "Character type of basic_string_view must be standard-layout");
+    static_assert(( is_trivial<value_type>::value), "Character type of basic_string_view must be trivial");
+    static_assert((is_same<_CharT, typename traits_type::char_type>::value),
+                  "traits_type::char_type must be the same type as CharT");
+
+    // [string.view.cons], construct/copy
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string_view() _NOEXCEPT : __data (nullptr), __size(0) {}
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string_view(const basic_string_view&) _NOEXCEPT = default;
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string_view& operator=(const basic_string_view&) _NOEXCEPT = default;
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string_view(const _CharT* __s, size_type __len) _NOEXCEPT
+        : __data(__s), __size(__len)
+    {
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_ASSERT(__len == 0 || __s != nullptr, "string_view::string_view(_CharT *, size_t): received nullptr");
+#endif
+    }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string_view(const _CharT* __s)
+        : __data(__s), __size(std::__char_traits_length_checked<_Traits>(__s)) {}
+
+    // [string.view.iterators], iterators
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const _NOEXCEPT { return cbegin(); }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const _NOEXCEPT { return cend(); }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT { return __data; }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend()   const _NOEXCEPT { return __data + __size; }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin()   const _NOEXCEPT { return const_reverse_iterator(cend()); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend()     const _NOEXCEPT { return const_reverse_iterator(cbegin()); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin()  const _NOEXCEPT { return const_reverse_iterator(cend()); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX14 _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend()    const _NOEXCEPT { return const_reverse_iterator(cbegin()); }
+
+    // [string.view.capacity], capacity
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size()     const _NOEXCEPT { return __size; }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    size_type length()   const _NOEXCEPT { return __size; }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT { return numeric_limits<size_type>::max(); }
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    bool empty()         const _NOEXCEPT { return __size == 0; }
+
+    // [string.view.access], element access
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference operator[](size_type __pos) const _NOEXCEPT { return __data[__pos]; }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference at(size_type __pos) const
+    {
+        return __pos >= size()
+            ? (__throw_out_of_range("string_view::at"), __data[0])
+            : __data[__pos];
+    }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference front() const _NOEXCEPT
+    {
+        return _LIBCUDACXX_ASSERT(!empty(), "string_view::front(): string is empty"), __data[0];
+    }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference back() const _NOEXCEPT
+    {
+        return _LIBCUDACXX_ASSERT(!empty(), "string_view::back(): string is empty"), __data[__size-1];
+    }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    const_pointer data() const _NOEXCEPT { return __data; }
+
+    // [string.view.modifiers], modifiers:
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    void remove_prefix(size_type __n) _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__n <= size(), "remove_prefix() can't remove more than size()");
+        __data += __n;
+        __size -= __n;
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    void remove_suffix(size_type __n) _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__n <= size(), "remove_suffix() can't remove more than size()");
+        __size -= __n;
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(basic_string_view& __other) _NOEXCEPT
+    {
+        const value_type *__p = __data;
+        __data = __other.__data;
+        __other.__data = __p;
+
+        size_type __sz = __size;
+        __size = __other.__size;
+        __other.__size = __sz;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type copy(_CharT* __s, size_type __n, size_type __pos = 0) const
+    {
+        if (__pos > size())
+            __throw_out_of_range("string_view::copy");
+        size_type __rlen = _CUDA_VSTD::min(__n, size() - __pos);
+        _Traits::copy(__s, data() + __pos, __rlen);
+        return __rlen;
+    }
+
+    _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+    basic_string_view substr(size_type __pos = 0, size_type __n = npos) const
+    {
+        return __pos > size()
+            ? (__throw_out_of_range("string_view::substr"), basic_string_view())
+            : basic_string_view(data() + __pos, _CUDA_VSTD::min(__n, size() - __pos));
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 int compare(basic_string_view __sv) const _NOEXCEPT
+    {
+        size_type __rlen = _CUDA_VSTD::min( size(), __sv.size());
+        int __retval = _Traits::compare(data(), __sv.data(), __rlen);
+        if ( __retval == 0 ) // first __rlen chars matched
+            __retval = size() == __sv.size() ? 0 : ( size() < __sv.size() ? -1 : 1 );
+        return __retval;
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(size_type __pos1, size_type __n1, basic_string_view __sv) const
+    {
+        return substr(__pos1, __n1).compare(__sv);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(                       size_type __pos1, size_type __n1,
+                basic_string_view __sv, size_type __pos2, size_type __n2) const
+    {
+        return substr(__pos1, __n1).compare(__sv.substr(__pos2, __n2));
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(const _CharT* __s) const _NOEXCEPT
+    {
+        return compare(basic_string_view(__s));
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(size_type __pos1, size_type __n1, const _CharT* __s) const
+    {
+        return substr(__pos1, __n1).compare(basic_string_view(__s));
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    int compare(size_type __pos1, size_type __n1, const _CharT* __s, size_type __n2) const
+    {
+        return substr(__pos1, __n1).compare(basic_string_view(__s, __n2));
+    }
+
+    // find
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find(basic_string_view __s, size_type __pos = 0) const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__s.size() == 0 || __s.data() != nullptr, "string_view::find(): received nullptr");
+        return __str_find<value_type, size_type, traits_type, npos>
+            (data(), size(), __s.data(), __pos, __s.size());
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find(_CharT __c, size_type __pos = 0) const _NOEXCEPT
+    {
+        return __str_find<value_type, size_type, traits_type, npos>
+            (data(), size(), __c, __pos);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find(const _CharT* __s, size_type __pos, size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string_view::find(): received nullptr");
+        return __str_find<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, __n);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find(const _CharT* __s, size_type __pos = 0) const
+    {
+        _LIBCUDACXX_ASSERT(__s != nullptr, "string_view::find(): received nullptr");
+        return __str_find<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, traits_type::length(__s));
+    }
+
+    // rfind
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type rfind(basic_string_view __s, size_type __pos = npos) const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__s.size() == 0 || __s.data() != nullptr, "string_view::find(): received nullptr");
+        return __str_rfind<value_type, size_type, traits_type, npos>
+            (data(), size(), __s.data(), __pos, __s.size());
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type rfind(_CharT __c, size_type __pos = npos) const _NOEXCEPT
+    {
+        return __str_rfind<value_type, size_type, traits_type, npos>
+            (data(), size(), __c, __pos);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type rfind(const _CharT* __s, size_type __pos, size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string_view::rfind(): received nullptr");
+        return __str_rfind<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, __n);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type rfind(const _CharT* __s, size_type __pos=npos) const
+    {
+        _LIBCUDACXX_ASSERT(__s != nullptr, "string_view::rfind(): received nullptr");
+        return __str_rfind<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, traits_type::length(__s));
+    }
+
+    // find_first_of
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(basic_string_view __s, size_type __pos = 0) const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__s.size() == 0 || __s.data() != nullptr, "string_view::find_first_of(): received nullptr");
+        return __str_find_first_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s.data(), __pos, __s.size());
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(_CharT __c, size_type __pos = 0) const _NOEXCEPT
+    { return find(__c, __pos); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(const _CharT* __s, size_type __pos, size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string_view::find_first_of(): received nullptr");
+        return __str_find_first_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, __n);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_of(const _CharT* __s, size_type __pos=0) const
+    {
+        _LIBCUDACXX_ASSERT(__s != nullptr, "string_view::find_first_of(): received nullptr");
+        return __str_find_first_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, traits_type::length(__s));
+    }
+
+    // find_last_of
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(basic_string_view __s, size_type __pos=npos) const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__s.size() == 0 || __s.data() != nullptr, "string_view::find_last_of(): received nullptr");
+        return __str_find_last_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s.data(), __pos, __s.size());
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(_CharT __c, size_type __pos = npos) const _NOEXCEPT
+    { return rfind(__c, __pos); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(const _CharT* __s, size_type __pos, size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string_view::find_last_of(): received nullptr");
+        return __str_find_last_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, __n);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_of(const _CharT* __s, size_type __pos=npos) const
+    {
+        _LIBCUDACXX_ASSERT(__s != nullptr, "string_view::find_last_of(): received nullptr");
+        return __str_find_last_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, traits_type::length(__s));
+    }
+
+    // find_first_not_of
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(basic_string_view __s, size_type __pos=0) const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__s.size() == 0 || __s.data() != nullptr, "string_view::find_first_not_of(): received nullptr");
+        return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s.data(), __pos, __s.size());
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(_CharT __c, size_type __pos=0) const _NOEXCEPT
+    {
+        return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __c, __pos);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(const _CharT* __s, size_type __pos, size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string_view::find_first_not_of(): received nullptr");
+        return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, __n);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_first_not_of(const _CharT* __s, size_type __pos=0) const
+    {
+        _LIBCUDACXX_ASSERT(__s != nullptr, "string_view::find_first_not_of(): received nullptr");
+        return __str_find_first_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, traits_type::length(__s));
+    }
+
+    // find_last_not_of
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(basic_string_view __s, size_type __pos=npos) const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(__s.size() == 0 || __s.data() != nullptr, "string_view::find_last_not_of(): received nullptr");
+        return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s.data(), __pos, __s.size());
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(_CharT __c, size_type __pos=npos) const _NOEXCEPT
+    {
+        return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __c, __pos);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(const _CharT* __s, size_type __pos, size_type __n) const
+    {
+        _LIBCUDACXX_ASSERT(__n == 0 || __s != nullptr, "string_view::find_last_not_of(): received nullptr");
+        return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, __n);
+    }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    size_type find_last_not_of(const _CharT* __s, size_type __pos=npos) const
+    {
+        _LIBCUDACXX_ASSERT(__s != nullptr, "string_view::find_last_not_of(): received nullptr");
+        return __str_find_last_not_of<value_type, size_type, traits_type, npos>
+            (data(), size(), __s, __pos, traits_type::length(__s));
+    }
+
+#if _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool starts_with(basic_string_view __s) const _NOEXCEPT
+    { return size() >= __s.size() && compare(0, __s.size(), __s) == 0; }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool starts_with(value_type __c) const _NOEXCEPT
+    { return !empty() && _Traits::eq(front(), __c); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool starts_with(const value_type* __s) const _NOEXCEPT
+    { return starts_with(basic_string_view(__s)); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool ends_with(basic_string_view __s) const _NOEXCEPT
+    { return size() >= __s.size() && compare(size() - __s.size(), npos, __s) == 0; }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool ends_with(value_type __c) const _NOEXCEPT
+    { return !empty() && _Traits::eq(back(), __c); }
+
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+    bool ends_with(const value_type* __s) const _NOEXCEPT
+    { return ends_with(basic_string_view(__s)); }
+#endif
+
+private:
+    const   value_type* __data;
+    size_type           __size;
+};
+
+
+// [string.view.comparison]
+// operator ==
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(basic_string_view<_CharT, _Traits> __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    if ( __lhs.size() != __rhs.size()) return false;
+    return __lhs.compare(__rhs) == 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(basic_string_view<_CharT, _Traits> __lhs,
+                typename common_type<basic_string_view<_CharT, _Traits> >::type __rhs) _NOEXCEPT
+{
+    if ( __lhs.size() != __rhs.size()) return false;
+    return __lhs.compare(__rhs) == 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator==(typename common_type<basic_string_view<_CharT, _Traits> >::type __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    if ( __lhs.size() != __rhs.size()) return false;
+    return __lhs.compare(__rhs) == 0;
+}
+
+
+// operator !=
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(basic_string_view<_CharT, _Traits> __lhs, basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    if ( __lhs.size() != __rhs.size())
+        return true;
+    return __lhs.compare(__rhs) != 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(basic_string_view<_CharT, _Traits> __lhs,
+                typename common_type<basic_string_view<_CharT, _Traits> >::type __rhs) _NOEXCEPT
+{
+    if ( __lhs.size() != __rhs.size())
+        return true;
+    return __lhs.compare(__rhs) != 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator!=(typename common_type<basic_string_view<_CharT, _Traits> >::type __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    if ( __lhs.size() != __rhs.size())
+        return true;
+    return __lhs.compare(__rhs) != 0;
+}
+
+
+// operator <
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<(basic_string_view<_CharT, _Traits> __lhs, basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) < 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<(basic_string_view<_CharT, _Traits> __lhs,
+                typename common_type<basic_string_view<_CharT, _Traits> >::type __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) < 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<(typename common_type<basic_string_view<_CharT, _Traits> >::type __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) < 0;
+}
+
+
+// operator >
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator> (basic_string_view<_CharT, _Traits> __lhs, basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) > 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>(basic_string_view<_CharT, _Traits> __lhs,
+                typename common_type<basic_string_view<_CharT, _Traits> >::type __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) > 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>(typename common_type<basic_string_view<_CharT, _Traits> >::type __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) > 0;
+}
+
+
+// operator <=
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<=(basic_string_view<_CharT, _Traits> __lhs, basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) <= 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<=(basic_string_view<_CharT, _Traits> __lhs,
+                typename common_type<basic_string_view<_CharT, _Traits> >::type __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) <= 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator<=(typename common_type<basic_string_view<_CharT, _Traits> >::type __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) <= 0;
+}
+
+
+// operator >=
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>=(basic_string_view<_CharT, _Traits> __lhs, basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) >= 0;
+}
+
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>=(basic_string_view<_CharT, _Traits> __lhs,
+                typename common_type<basic_string_view<_CharT, _Traits> >::type __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) >= 0;
+}
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+bool operator>=(typename common_type<basic_string_view<_CharT, _Traits> >::type __lhs,
+                basic_string_view<_CharT, _Traits> __rhs) _NOEXCEPT
+{
+    return __lhs.compare(__rhs) >= 0;
+}
+
+
+template<class _CharT, class _Traits>
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os,
+           basic_string_view<_CharT, _Traits> __str);
+
+typedef basic_string_view<char>     string_view;
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+typedef basic_string_view<char8_t>  u8string_view;
+#endif
+typedef basic_string_view<char16_t> u16string_view;
+typedef basic_string_view<char32_t> u32string_view;
+typedef basic_string_view<wchar_t>  wstring_view;
+
+// [string.view.hash]
+template<class _CharT>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<basic_string_view<_CharT, char_traits<_CharT> > >
+    : public unary_function<basic_string_view<_CharT, char_traits<_CharT> >, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const basic_string_view<_CharT, char_traits<_CharT> > __val) const _NOEXCEPT {
+        return __do_string_hash(__val.data(), __val.data() + __val.size());
+    }
+};
+
+
+#if _LIBCUDACXX_STD_VER > 11
+inline namespace literals
+{
+  inline namespace string_view_literals
+  {
+    inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    basic_string_view<char> operator "" sv(const char *__str, size_t __len) _NOEXCEPT
+    {
+        return basic_string_view<char> (__str, __len);
+    }
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    basic_string_view<wchar_t> operator "" sv(const wchar_t *__str, size_t __len) _NOEXCEPT
+    {
+        return basic_string_view<wchar_t> (__str, __len);
+    }
+
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+    inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    basic_string_view<char8_t> operator "" sv(const char8_t *__str, size_t __len) _NOEXCEPT
+    {
+        return basic_string_view<char8_t> (__str, __len);
+    }
+#endif
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    basic_string_view<char16_t> operator "" sv(const char16_t *__str, size_t __len) _NOEXCEPT
+    {
+        return basic_string_view<char16_t> (__str, __len);
+    }
+
+    inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    basic_string_view<char32_t> operator "" sv(const char32_t *__str, size_t __len) _NOEXCEPT
+    {
+        return basic_string_view<char32_t> (__str, __len);
+    }
+  }
+}
+#endif
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif // _LIBCUDACXX_STRING_VIEW
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/strstream b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/strstream
new file mode 100644
index 000000000000..9c865fb244ad
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/strstream
@@ -0,0 +1,399 @@
+// -*- C++ -*-
+//===--------------------------- strstream --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_STRSTREAM
+#define _LIBCUDACXX_STRSTREAM
+
+/*
+    strstream synopsis
+
+class strstreambuf
+    : public basic_streambuf<char>
+{
+public:
+    explicit strstreambuf(streamsize alsize_arg = 0);
+    strstreambuf(void* (*palloc_arg)(size_t), void (*pfree_arg)(void*));
+    strstreambuf(char* gnext_arg, streamsize n, char* pbeg_arg = 0);
+    strstreambuf(const char* gnext_arg, streamsize n);
+
+    strstreambuf(signed char* gnext_arg, streamsize n, signed char* pbeg_arg = 0);
+    strstreambuf(const signed char* gnext_arg, streamsize n);
+    strstreambuf(unsigned char* gnext_arg, streamsize n, unsigned char* pbeg_arg = 0);
+    strstreambuf(const unsigned char* gnext_arg, streamsize n);
+
+    strstreambuf(strstreambuf&& rhs);
+    strstreambuf& operator=(strstreambuf&& rhs);
+
+    virtual ~strstreambuf();
+
+    void swap(strstreambuf& rhs);
+
+    void freeze(bool freezefl = true);
+    char* str();
+    int pcount() const;
+
+protected:
+    virtual int_type overflow (int_type c = EOF);
+    virtual int_type pbackfail(int_type c = EOF);
+    virtual int_type underflow();
+    virtual pos_type seekoff(off_type off, ios_base::seekdir way,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type sp,
+                             ios_base::openmode which = ios_base::in | ios_base::out);
+    virtual streambuf* setbuf(char* s, streamsize n);
+
+private:
+    typedef T1 strstate;                // exposition only
+    static const strstate allocated;    // exposition only
+    static const strstate constant;     // exposition only
+    static const strstate dynamic;      // exposition only
+    static const strstate frozen;       // exposition only
+    strstate strmode;                   // exposition only
+    streamsize alsize;                  // exposition only
+    void* (*palloc)(size_t);            // exposition only
+    void (*pfree)(void*);               // exposition only
+};
+
+class istrstream
+    : public basic_istream<char>
+{
+public:
+    explicit istrstream(const char* s);
+    explicit istrstream(char* s);
+    istrstream(const char* s, streamsize n);
+    istrstream(char* s, streamsize n);
+
+    virtual ~istrstream();
+
+    strstreambuf* rdbuf() const;
+    char *str();
+
+private:
+    strstreambuf sb; // exposition only
+};
+
+class ostrstream
+    : public basic_ostream<char>
+{
+public:
+    ostrstream();
+    ostrstream(char* s, int n, ios_base::openmode mode = ios_base::out);
+
+    virtual ~ostrstream();
+
+    strstreambuf* rdbuf() const;
+    void freeze(bool freezefl = true);
+    char* str();
+    int pcount() const;
+
+private:
+    strstreambuf sb; // exposition only
+};
+
+class strstream
+    : public basic_iostream<char>
+{
+public:
+    // Types
+    typedef char                        char_type;
+    typedef char_traits<char>::int_type int_type;
+    typedef char_traits<char>::pos_type pos_type;
+    typedef char_traits<char>::off_type off_type;
+
+    // constructors/destructor
+    strstream();
+    strstream(char* s, int n, ios_base::openmode mode = ios_base::in | ios_base::out);
+
+    virtual ~strstream();
+
+    // Members:
+    strstreambuf* rdbuf() const;
+    void freeze(bool freezefl = true);
+    int pcount() const;
+    char* str();
+
+private:
+    strstreambuf sb; // exposition only
+};
+
+}  // std
+
+*/
+
+#include <__config>
+#include <ostream>
+#include <istream>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+class _LIBCUDACXX_TYPE_VIS strstreambuf
+    : public streambuf
+{
+public:
+    explicit strstreambuf(streamsize __alsize = 0);
+    strstreambuf(void* (*__palloc)(size_t), void (*__pfree)(void*));
+    strstreambuf(char* __gnext, streamsize __n, char* __pbeg = 0);
+    strstreambuf(const char* __gnext, streamsize __n);
+
+    strstreambuf(signed char* __gnext, streamsize __n, signed char* __pbeg = 0);
+    strstreambuf(const signed char* __gnext, streamsize __n);
+    strstreambuf(unsigned char* __gnext, streamsize __n, unsigned char* __pbeg = 0);
+    strstreambuf(const unsigned char* __gnext, streamsize __n);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstreambuf(strstreambuf&& __rhs);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstreambuf& operator=(strstreambuf&& __rhs);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    virtual ~strstreambuf();
+
+    void swap(strstreambuf& __rhs);
+
+    void freeze(bool __freezefl = true);
+    char* str();
+    int pcount() const;
+
+protected:
+    virtual int_type overflow (int_type __c = EOF);
+    virtual int_type pbackfail(int_type __c = EOF);
+    virtual int_type underflow();
+    virtual pos_type seekoff(off_type __off, ios_base::seekdir __way,
+                             ios_base::openmode __which = ios_base::in | ios_base::out);
+    virtual pos_type seekpos(pos_type __sp,
+                             ios_base::openmode __which = ios_base::in | ios_base::out);
+
+private:
+    typedef unsigned __mode_type;
+    static const __mode_type __allocated = 0x01;
+    static const __mode_type __constant  = 0x02;
+    static const __mode_type __dynamic   = 0x04;
+    static const __mode_type __frozen    = 0x08;
+    static const streamsize    __default_alsize = 4096;
+
+    __mode_type __strmode_;
+    streamsize __alsize_;
+    void* (*__palloc_)(size_t);
+    void (*__pfree_)(void*);
+
+    void __init(char* __gnext, streamsize __n, char* __pbeg);
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+strstreambuf::strstreambuf(strstreambuf&& __rhs)
+    : streambuf(__rhs),
+      __strmode_(__rhs.__strmode_),
+      __alsize_(__rhs.__alsize_),
+      __palloc_(__rhs.__palloc_),
+      __pfree_(__rhs.__pfree_)
+{
+    __rhs.setg(nullptr, nullptr, nullptr);
+    __rhs.setp(nullptr, nullptr);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+strstreambuf&
+strstreambuf::operator=(strstreambuf&& __rhs)
+{
+    if (eback() && (__strmode_ & __allocated) != 0 && (__strmode_ & __frozen) == 0)
+    {
+        if (__pfree_)
+            __pfree_(eback());
+        else
+            delete [] eback();
+    }
+    streambuf::operator=(__rhs);
+    __strmode_ = __rhs.__strmode_;
+    __alsize_ = __rhs.__alsize_;
+    __palloc_ = __rhs.__palloc_;
+    __pfree_ = __rhs.__pfree_;
+    __rhs.setg(nullptr, nullptr, nullptr);
+    __rhs.setp(nullptr, nullptr);
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+class _LIBCUDACXX_TYPE_VIS istrstream
+    : public istream
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit istrstream(const char* __s)
+        : istream(&__sb_), __sb_(__s, 0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit istrstream(char* __s)
+        : istream(&__sb_), __sb_(__s, 0) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    istrstream(const char* __s, streamsize __n)
+        : istream(&__sb_), __sb_(__s, __n) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    istrstream(char* __s, streamsize __n)
+        : istream(&__sb_), __sb_(__s, __n) {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    istrstream(istrstream&& __rhs)
+        : istream(_CUDA_VSTD::move(__rhs)),
+          __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+    {
+        istream::set_rdbuf(&__sb_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    istrstream& operator=(istrstream&& __rhs)
+    {
+        istream::operator=(_CUDA_VSTD::move(__rhs));
+        __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+        return *this;
+    }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    virtual ~istrstream();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(istrstream& __rhs)
+    {
+        istream::swap(__rhs);
+        __sb_.swap(__rhs.__sb_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstreambuf* rdbuf() const {return const_cast<strstreambuf*>(&__sb_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char *str() {return __sb_.str();}
+
+private:
+    strstreambuf __sb_;
+};
+
+class _LIBCUDACXX_TYPE_VIS ostrstream
+    : public ostream
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ostrstream()
+        : ostream(&__sb_) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ostrstream(char* __s, int __n, ios_base::openmode __mode = ios_base::out)
+        : ostream(&__sb_),
+          __sb_(__s, __n, __s + (__mode & ios::app ? strlen(__s) : 0))
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ostrstream(ostrstream&& __rhs)
+        : ostream(_CUDA_VSTD::move(__rhs)),
+          __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+    {
+        ostream::set_rdbuf(&__sb_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ostrstream& operator=(ostrstream&& __rhs)
+    {
+        ostream::operator=(_CUDA_VSTD::move(__rhs));
+        __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+        return *this;
+    }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    virtual ~ostrstream();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(ostrstream& __rhs)
+    {
+        ostream::swap(__rhs);
+        __sb_.swap(__rhs.__sb_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstreambuf* rdbuf() const {return const_cast<strstreambuf*>(&__sb_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void freeze(bool __freezefl = true) {__sb_.freeze(__freezefl);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char* str()         {return __sb_.str();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int pcount() const  {return __sb_.pcount();}
+
+private:
+    strstreambuf __sb_; // exposition only
+};
+
+class _LIBCUDACXX_TYPE_VIS strstream
+    : public iostream
+{
+public:
+    // Types
+    typedef char                        char_type;
+    typedef char_traits<char>::int_type int_type;
+    typedef char_traits<char>::pos_type pos_type;
+    typedef char_traits<char>::off_type off_type;
+
+    // constructors/destructor
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstream()
+        : iostream(&__sb_) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstream(char* __s, int __n, ios_base::openmode __mode = ios_base::in | ios_base::out)
+        : iostream(&__sb_),
+          __sb_(__s, __n, __s + (__mode & ios::app ? strlen(__s) : 0))
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstream(strstream&& __rhs)
+        : iostream(_CUDA_VSTD::move(__rhs)),
+          __sb_(_CUDA_VSTD::move(__rhs.__sb_))
+    {
+        iostream::set_rdbuf(&__sb_);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstream& operator=(strstream&& __rhs)
+    {
+        iostream::operator=(_CUDA_VSTD::move(__rhs));
+        __sb_ = _CUDA_VSTD::move(__rhs.__sb_);
+        return *this;
+    }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    virtual ~strstream();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(strstream& __rhs)
+    {
+        iostream::swap(__rhs);
+        __sb_.swap(__rhs.__sb_);
+    }
+
+    // Members:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    strstreambuf* rdbuf() const {return const_cast<strstreambuf*>(&__sb_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void freeze(bool __freezefl = true) {__sb_.freeze(__freezefl);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int pcount() const {return __sb_.pcount();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    char* str()        {return __sb_.str();}
+
+private:
+    strstreambuf __sb_; // exposition only
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_STRSTREAM
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/android/locale_bionic.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/android/locale_bionic.h
new file mode 100644
index 000000000000..88a399a77c69
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/android/locale_bionic.h
@@ -0,0 +1,67 @@
+// -*- C++ -*-
+//===------------------- support/android/locale_bionic.h ------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_ANDROID_LOCALE_BIONIC_H
+#define _LIBCUDACXX_SUPPORT_ANDROID_LOCALE_BIONIC_H
+
+#if defined(__BIONIC__)
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdlib.h>
+#include <xlocale.h>
+
+#ifdef __cplusplus
+}
+#endif
+
+#if defined(__ANDROID__)
+
+#include <android/api-level.h>
+#include <android/ndk-version.h>
+#include <support/xlocale/__posix_l_fallback.h>
+// In NDK versions later than 16, locale-aware functions are provided by
+// legacy_stdlib_inlines.h
+#if __NDK_MAJOR__ <= 16
+#if __ANDROID_API__ < 21
+#include <support/xlocale/__strtonum_fallback.h>
+#elif __ANDROID_API__ < 26
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float strtof_l(const char* __nptr, char** __endptr,
+                                                locale_t) {
+  return ::strtof(__nptr, __endptr);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY double strtod_l(const char* __nptr,
+                                                 char** __endptr, locale_t) {
+  return ::strtod(__nptr, __endptr);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long strtol_l(const char* __nptr, char** __endptr,
+                                               int __base, locale_t) {
+  return ::strtol(__nptr, __endptr, __base);
+}
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif // __ANDROID_API__ < 26
+
+#endif // __NDK_MAJOR__ <= 16
+#endif // defined(__ANDROID__)
+
+#endif // defined(__BIONIC__)
+#endif // _LIBCUDACXX_SUPPORT_ANDROID_LOCALE_BIONIC_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_base.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_base.h
new file mode 100644
index 000000000000..5c431cc9c561
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_base.h
@@ -0,0 +1,192 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ATOMIC_BASE_H
+#define _LIBCUDACXX_ATOMIC_BASE_H
+
+#include "cxx_atomic.h"
+
+// Guard ifdef for lock free query in case it is assigned elsewhere (MSVC/CUDA)
+#ifndef _LIBCUDACXX_ATOMIC_IS_LOCK_FREE
+#define _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(__x) __atomic_is_lock_free(__x, 0)
+#endif
+
+_LIBCUDACXX_INLINE_VISIBILITY inline _LIBCUDACXX_CONSTEXPR int __cxx_atomic_order_to_int(memory_order __order) {
+  // Avoid switch statement to make this a constexpr.
+  return __order == memory_order_relaxed ? __ATOMIC_RELAXED:
+         (__order == memory_order_acquire ? __ATOMIC_ACQUIRE:
+          (__order == memory_order_release ? __ATOMIC_RELEASE:
+           (__order == memory_order_seq_cst ? __ATOMIC_SEQ_CST:
+            (__order == memory_order_acq_rel ? __ATOMIC_ACQ_REL:
+              __ATOMIC_CONSUME))));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY inline _LIBCUDACXX_CONSTEXPR int __cxx_atomic_failure_order_to_int(memory_order __order) {
+  // Avoid switch statement to make this a constexpr.
+  return __order == memory_order_relaxed ? __ATOMIC_RELAXED:
+         (__order == memory_order_acquire ? __ATOMIC_ACQUIRE:
+          (__order == memory_order_release ? __ATOMIC_RELAXED:
+           (__order == memory_order_seq_cst ? __ATOMIC_SEQ_CST:
+            (__order == memory_order_acq_rel ? __ATOMIC_ACQUIRE:
+              __ATOMIC_CONSUME))));
+}
+
+template <typename _Tp, typename _Up>
+inline void __cxx_atomic_init(volatile _Tp* __a,  _Up __val) {
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  __cxx_atomic_assign_volatile(*__a_tmp, __val);
+}
+
+template <typename _Tp, typename _Up>
+inline void __cxx_atomic_init(_Tp* __a,  _Up __val) {
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  *__a_tmp = __val;
+}
+
+inline
+void __cxx_atomic_thread_fence(memory_order __order) {
+  __atomic_thread_fence(__cxx_atomic_order_to_int(__order));
+}
+
+inline
+void __cxx_atomic_signal_fence(memory_order __order) {
+  __atomic_signal_fence(__cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp, typename _Up>
+inline void __cxx_atomic_store(_Tp* __a,  _Up __val,
+                        memory_order __order) {
+  auto __v_temp = __cxx_atomic_wrap_to_base(__a, __val);
+  __atomic_store(__cxx_atomic_unwrap(__a), &__v_temp, __cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp>
+inline auto __cxx_atomic_load(const _Tp* __a,
+                       memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __ret = __cxx_atomic_base_temporary(__a);
+  __atomic_load(__cxx_atomic_unwrap(__a), &__ret, __cxx_atomic_order_to_int(__order));
+  return *__cxx_get_underlying_atomic(&__ret);
+}
+
+template <typename _Tp, typename _Up>
+inline auto __cxx_atomic_exchange(_Tp* __a, _Up __val,
+                          memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __v_temp = __cxx_atomic_wrap_to_base(__a, __val);
+  auto __ret = __cxx_atomic_base_temporary(__a);
+  __atomic_exchange(__cxx_atomic_unwrap(__a), &__v_temp, &__ret, __cxx_atomic_order_to_int(__order));
+  return *__cxx_get_underlying_atomic(&__ret);
+}
+
+template <typename _Tp, typename _Up>
+inline bool __cxx_atomic_compare_exchange_strong(
+    _Tp* __a, _Up* __expected, _Up __value, memory_order __success,
+    memory_order __failure) {
+  (void)__expected;
+  return __atomic_compare_exchange(__cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a)),
+                                   __expected, &__value, false,
+                                   __cxx_atomic_order_to_int(__success),
+                                   __cxx_atomic_failure_order_to_int(__failure));
+}
+
+template <typename _Tp, typename _Up>
+inline bool __cxx_atomic_compare_exchange_weak(
+    _Tp* __a, _Up* __expected, _Up __value, memory_order __success,
+    memory_order __failure) {
+  (void)__expected;
+  return __atomic_compare_exchange(__cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a)),
+                                   __expected, &__value, true,
+                                   __cxx_atomic_order_to_int(__success),
+                                   __cxx_atomic_failure_order_to_int(__failure));
+}
+
+template <typename _Tp>
+struct __atomic_ptr_inc { enum {value = 1}; };
+
+template <typename _Tp>
+struct __atomic_ptr_inc<_Tp*> { enum {value = sizeof(_Tp)}; };
+
+// FIXME: Haven't figured out what the spec says about using arrays with
+// atomic_fetch_add. Force a failure rather than creating bad behavior.
+template <typename _Tp>
+struct __atomic_ptr_inc<_Tp[]> { };
+template <typename _Tp, int n>
+struct __atomic_ptr_inc<_Tp[n]> { };
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_add(_Tp* __a, _Td __delta,
+                           memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  constexpr auto __skip_v = __atomic_ptr_inc<__cxx_atomic_underlying_t<_Tp>>::value;
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  return __atomic_fetch_add(__a_tmp, __delta * __skip_v,
+                            __cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_sub(_Tp* __a, _Td __delta,
+                           memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  constexpr auto __skip_v = __atomic_ptr_inc<__cxx_atomic_underlying_t<_Tp>>::value;
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  return __atomic_fetch_sub(__a_tmp, __delta * __skip_v,
+                            __cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_and(_Tp* __a, _Td __pattern,
+                            memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  return __atomic_fetch_and(__a_tmp, __pattern,
+                            __cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_or(_Tp* __a, _Td __pattern,
+                          memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  return __atomic_fetch_or(__a_tmp, __pattern,
+                           __cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_xor(_Tp* __a, _Td __pattern,
+                           memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __a_tmp = __cxx_get_underlying_atomic(__cxx_atomic_unwrap(__a));
+  return __atomic_fetch_xor(__a_tmp, __pattern,
+                            __cxx_atomic_order_to_int(__order));
+}
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_max(_Tp* __a, _Td __val,
+                           memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __expected = __cxx_atomic_load(__a, memory_order_relaxed);
+  auto __desired = __expected > __val ? __expected : __val;
+
+  while(__desired == __val &&
+          !__cxx_atomic_compare_exchange_strong(__a, &__expected, __desired, __order, __order)) {
+      __desired = __expected > __val ? __expected : __val;
+  }
+
+  return __expected;
+}
+
+template <typename _Tp, typename _Td>
+inline auto __cxx_atomic_fetch_min(_Tp* __a, _Td __val,
+                           memory_order __order) -> __cxx_atomic_underlying_t<_Tp> {
+  auto __expected = __cxx_atomic_load(__a, memory_order_relaxed);
+  auto __desired = __expected < __val ? __expected : __val;
+
+  while(__desired == __val &&
+          !__cxx_atomic_compare_exchange_strong(__a, &__expected, __desired, __order, __order)) {
+      __desired = __expected < __val ? __expected : __val;
+  }
+
+  return __expected;
+}
+
+#endif // _LIBCUDACXX_ATOMIC_BASE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_c11.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_c11.h
new file mode 100644
index 000000000000..3e4b341485f6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_c11.h
@@ -0,0 +1,183 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// Atomics for C11
+
+template <typename _Tp>
+struct __cxx_atomic_base_impl {
+
+  _LIBCUDACXX_INLINE_VISIBILITY
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __cxx_atomic_base_impl() _NOEXCEPT = default;
+#else
+    __cxx_atomic_base_impl() _NOEXCEPT : __a_value() {}
+#endif // _LIBCUDACXX_CXX03_LANG
+  _LIBCUDACXX_CONSTEXPR explicit __cxx_atomic_base_impl(_Tp value) _NOEXCEPT
+    : __a_value(value) {}
+  _LIBCUDACXX_DISABLE_EXTENSION_WARNING _Atomic(_Tp) __a_value;
+};
+
+#ifndef _LIBCUDACXX_ATOMIC_IS_LOCK_FREE
+#define _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(__x) __c11_atomic_is_lock_free(__x, 0)
+#endif
+
+_LIBCUDACXX_INLINE_VISIBILITY inline
+void __cxx_atomic_thread_fence(memory_order __order) _NOEXCEPT {
+    __c11_atomic_thread_fence(static_cast<__memory_order_underlying_t>(__order));
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY inline
+void __cxx_atomic_signal_fence(memory_order __order) _NOEXCEPT {
+    __c11_atomic_signal_fence(static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_init(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __val) _NOEXCEPT {
+    __c11_atomic_init(&__a->__a_value, __val);
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_init(__cxx_atomic_base_impl<_Tp> * __a, _Tp __val) _NOEXCEPT {
+    __c11_atomic_init(&__a->__a_value, __val);
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_store(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __val, memory_order __order) _NOEXCEPT {
+    __c11_atomic_store(&__a->__a_value, __val, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __cxx_atomic_store(__cxx_atomic_base_impl<_Tp> * __a, _Tp __val, memory_order __order) _NOEXCEPT {
+    __c11_atomic_store(&__a->__a_value, __val, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_load(__cxx_atomic_base_impl<_Tp> const volatile* __a, memory_order __order) _NOEXCEPT {
+    using __ptr_type = typename remove_const<decltype(__a->__a_value)>::type*;
+    return __c11_atomic_load(const_cast<__ptr_type>(&__a->__a_value), static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_load(__cxx_atomic_base_impl<_Tp> const* __a, memory_order __order) _NOEXCEPT {
+    using __ptr_type = typename remove_const<decltype(__a->__a_value)>::type*;
+    return __c11_atomic_load(const_cast<__ptr_type>(&__a->__a_value), static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_exchange(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __value, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_exchange(&__a->__a_value, __value, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_exchange(__cxx_atomic_base_impl<_Tp> * __a, _Tp __value, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_exchange(&__a->__a_value, __value, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_strong(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp* __expected, _Tp __value, memory_order __success, memory_order __failure) _NOEXCEPT {
+    return __c11_atomic_compare_exchange_strong(&__a->__a_value, __expected, __value, static_cast<__memory_order_underlying_t>(__success), static_cast<__memory_order_underlying_t>(__failure));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_strong(__cxx_atomic_base_impl<_Tp> * __a, _Tp* __expected, _Tp __value, memory_order __success, memory_order __failure) _NOEXCEPT {
+    return __c11_atomic_compare_exchange_strong(&__a->__a_value, __expected, __value, static_cast<__memory_order_underlying_t>(__success), static_cast<__memory_order_underlying_t>(__failure));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_weak(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp* __expected, _Tp __value, memory_order __success, memory_order __failure) _NOEXCEPT {
+    return __c11_atomic_compare_exchange_weak(&__a->__a_value, __expected, __value, static_cast<__memory_order_underlying_t>(__success), static_cast<__memory_order_underlying_t>(__failure));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+bool __cxx_atomic_compare_exchange_weak(__cxx_atomic_base_impl<_Tp> * __a, _Tp* __expected, _Tp __value, memory_order __success, memory_order __failure) _NOEXCEPT {
+    return __c11_atomic_compare_exchange_weak(&__a->__a_value, __expected, __value,  static_cast<__memory_order_underlying_t>(__success), static_cast<__memory_order_underlying_t>(__failure));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_add(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_add(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_add(__cxx_atomic_base_impl<_Tp> * __a, _Tp __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_add(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp* __cxx_atomic_fetch_add(__cxx_atomic_base_impl<_Tp*> volatile* __a, ptrdiff_t __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_add(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp* __cxx_atomic_fetch_add(__cxx_atomic_base_impl<_Tp*> * __a, ptrdiff_t __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_add(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_sub(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_sub(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_sub(__cxx_atomic_base_impl<_Tp> * __a, _Tp __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_sub(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp* __cxx_atomic_fetch_sub(__cxx_atomic_base_impl<_Tp*> volatile* __a, ptrdiff_t __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_sub(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp* __cxx_atomic_fetch_sub(__cxx_atomic_base_impl<_Tp*> * __a, ptrdiff_t __delta, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_sub(&__a->__a_value, __delta, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_and(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __pattern, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_and(&__a->__a_value, __pattern, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_and(__cxx_atomic_base_impl<_Tp> * __a, _Tp __pattern, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_and(&__a->__a_value, __pattern, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_or(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __pattern, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_or(&__a->__a_value, __pattern, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_or(__cxx_atomic_base_impl<_Tp> * __a, _Tp __pattern, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_or(&__a->__a_value, __pattern, static_cast<__memory_order_underlying_t>(__order));
+}
+
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_xor(__cxx_atomic_base_impl<_Tp> volatile* __a, _Tp __pattern, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_xor(&__a->__a_value, __pattern, static_cast<__memory_order_underlying_t>(__order));
+}
+template<class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp __cxx_atomic_fetch_xor(__cxx_atomic_base_impl<_Tp> * __a, _Tp __pattern, memory_order __order) _NOEXCEPT {
+    return __c11_atomic_fetch_xor(&__a->__a_value, __pattern, static_cast<__memory_order_underlying_t>(__order));
+}
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda.h
new file mode 100644
index 000000000000..35a7cd274b46
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda.h
@@ -0,0 +1,482 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__CUDA_MINIMUM_ARCH__) && ((!defined(_MSC_VER) && __CUDA_MINIMUM_ARCH__ < 600) || (defined(_MSC_VER) && __CUDA_MINIMUM_ARCH__ < 700))
+#  error "CUDA atomics are only supported for sm_60 and up on *nix and sm_70 and up on Windows."
+#endif
+
+#if !defined(__CLANG_ATOMIC_BOOL_LOCK_FREE) && !defined(__GCC_ATOMIC_BOOL_LOCK_FREE)
+#define ATOMIC_BOOL_LOCK_FREE      2
+#define ATOMIC_CHAR_LOCK_FREE      2
+#define ATOMIC_CHAR16_T_LOCK_FREE  2
+#define ATOMIC_CHAR32_T_LOCK_FREE  2
+#define ATOMIC_WCHAR_T_LOCK_FREE   2
+#define ATOMIC_SHORT_LOCK_FREE     2
+#define ATOMIC_INT_LOCK_FREE       2
+#define ATOMIC_LONG_LOCK_FREE      2
+#define ATOMIC_LLONG_LOCK_FREE     2
+#define ATOMIC_POINTER_LOCK_FREE   2
+#endif //!defined(__CLANG_ATOMIC_BOOL_LOCK_FREE) && !defined(__GCC_ATOMIC_BOOL_LOCK_FREE)
+
+#ifndef __ATOMIC_RELAXED
+#define __ATOMIC_RELAXED 0
+#define __ATOMIC_CONSUME 1
+#define __ATOMIC_ACQUIRE 2
+#define __ATOMIC_RELEASE 3
+#define __ATOMIC_ACQ_REL 4
+#define __ATOMIC_SEQ_CST 5
+#endif //__ATOMIC_RELAXED
+
+inline __host__ __device__ int __stronger_order_cuda(int __a, int __b) {
+    int const __max = __a > __b ? __a : __b;
+    if(__max != __ATOMIC_RELEASE)
+        return __max;
+    static int const __xform[] = {
+        __ATOMIC_RELEASE,
+        __ATOMIC_ACQ_REL,
+        __ATOMIC_ACQ_REL,
+        __ATOMIC_RELEASE };
+    return __xform[__a < __b ? __a : __b];
+}
+
+// pre-define lock free query for heterogeneous compatibility
+#ifndef _LIBCUDACXX_ATOMIC_IS_LOCK_FREE
+#define _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(__x) (__x <= 8)
+#endif
+
+// Wrap host atomic implementations into a sub-namespace
+namespace __host {
+#if defined(_LIBCUDACXX_COMPILER_MSVC)
+#  include "atomic_msvc.h"
+#elif defined (_LIBCUDACXX_HAS_GCC_ATOMIC_IMP)
+#  include "atomic_gcc.h"
+#elif defined (_LIBCUDACXX_HAS_C11_ATOMIC_IMP)
+//TODO
+// #  include "atomic_c11.h"
+#elif defined(_LIBCUDACXX_COMPILER_NVRTC)
+#  include "atomic_nvrtc.h"
+#endif
+}
+
+#include "atomic_cuda_generated.h"
+#include "atomic_cuda_derived.h"
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline
+ void __cxx_atomic_thread_fence(memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            __atomic_thread_fence_cuda(static_cast<__memory_order_underlying_t>(__order), __thread_scope_system_tag());
+        ),
+        NV_IS_HOST, (
+            __host::__cxx_atomic_thread_fence(__order);
+        )
+    )
+}
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline
+ void __cxx_atomic_signal_fence(memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            __atomic_signal_fence_cuda(static_cast<__memory_order_underlying_t>(__order));
+        ),
+        NV_IS_HOST, (
+            __host::__cxx_atomic_signal_fence(__order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref = false>
+struct __cxx_atomic_base_heterogeneous_impl {
+    __cxx_atomic_base_heterogeneous_impl() noexcept = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit
+      __cxx_atomic_base_heterogeneous_impl(_Tp __value) : __a_value(__value) {
+    }
+
+    __host::__cxx_atomic_base_impl<_Tp, _Sco> __a_value;
+};
+
+template <typename _Tp, int _Sco>
+struct __cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, true> {
+    __cxx_atomic_base_heterogeneous_impl() noexcept = default;
+
+    static_assert(sizeof(_Tp) >= 4, "atomic_ref does not support 1 or 2 byte types");
+    static_assert(sizeof(_Tp) <= 8, "atomic_ref does not support types larger than 8 bytes");
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit
+      __cxx_atomic_base_heterogeneous_impl(_Tp& __value) : __a_value(__value) {
+    }
+
+    __host::__cxx_atomic_ref_base_impl<_Tp, _Sco> __a_value;
+};
+
+template <typename _Tp, int _Sco, bool _Ref>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp* __cxx_get_underlying_device_atomic(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> * __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(&__a->__a_value);
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+volatile _Tp* __cxx_get_underlying_device_atomic(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(&__a->__a_value);
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Tp* __cxx_get_underlying_device_atomic(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> const* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(&__a->__a_value);
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const volatile _Tp* __cxx_get_underlying_device_atomic(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> const volatile* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(&__a->__a_value);
+}
+
+template <typename _Tp, int _Sco>
+struct __cxx_atomic_base_small_impl {
+    __cxx_atomic_base_small_impl() noexcept = default;
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit
+      __cxx_atomic_base_small_impl(_Tp __value) : __a_value(__value) {
+    }
+
+    __cxx_atomic_base_heterogeneous_impl<uint32_t, _Sco, false> __a_value;
+};
+
+template <typename _Tp>
+using __cxx_small_proxy = typename conditional<sizeof(_Tp) == 1,
+                                               uint8_t,
+                                               typename conditional<sizeof(_Tp) == 2,
+                                                                    uint16_t,
+                                                                    void>::type >::type;
+
+template <typename _Tp, int _Sco>
+using __cxx_atomic_base_impl = typename conditional<sizeof(_Tp) < 4,
+                                    __cxx_atomic_base_small_impl<_Tp, _Sco>,
+                                    __cxx_atomic_base_heterogeneous_impl<_Tp, _Sco> >::type;
+
+
+template <typename _Tp, int _Sco>
+using __cxx_atomic_ref_base_impl = __cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, true>;
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ void __cxx_atomic_init(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __val) {
+    alignas(_Tp) auto __tmp = __val;
+    __cxx_atomic_assign_volatile(*__cxx_get_underlying_device_atomic(__a), __tmp);
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ void __cxx_atomic_store(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __val, memory_order __order) {
+    alignas(_Tp) auto __tmp = __val;
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            __atomic_store_n_cuda(__cxx_get_underlying_device_atomic(__a), __tmp, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            __host::__cxx_atomic_store(&__a->__a_value, __tmp, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_load(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> const volatile* __a, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_load_n_cuda(__cxx_get_underlying_device_atomic(__a), static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_load(&__a->__a_value, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_exchange(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __val, memory_order __order) {
+    alignas(_Tp) auto __tmp = __val;
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_exchange_n_cuda(__cxx_get_underlying_device_atomic(__a), __tmp, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_exchange(&__a->__a_value, __tmp, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ bool __cxx_atomic_compare_exchange_strong(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp* __expected, _Tp __val, memory_order __success, memory_order __failure) {
+    alignas(_Tp) auto __tmp = *__expected;
+    bool __result = false;
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            alignas(_Tp) auto __tmp_v = __val;
+            __result = __atomic_compare_exchange_cuda(__cxx_get_underlying_device_atomic(__a), &__tmp, &__tmp_v, false, static_cast<__memory_order_underlying_t>(__success), static_cast<__memory_order_underlying_t>(__failure), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            __result = __host::__cxx_atomic_compare_exchange_strong(&__a->__a_value, &__tmp, __val, __success, __failure);
+        )
+    )
+    *__expected = __tmp;
+    return __result;
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ bool __cxx_atomic_compare_exchange_weak(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp* __expected, _Tp __val, memory_order __success, memory_order __failure) {
+    alignas(_Tp) auto __tmp = *__expected;
+    bool __result = false;
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            alignas(_Tp) auto __tmp_v = __val;
+            __result = __atomic_compare_exchange_cuda(__cxx_get_underlying_device_atomic(__a), &__tmp, &__tmp_v, true, static_cast<__memory_order_underlying_t>(__success), static_cast<__memory_order_underlying_t>(__failure), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            __result = __host::__cxx_atomic_compare_exchange_weak(&__a->__a_value, &__tmp, __val, __success, __failure);
+        )
+    )
+    *__expected = __tmp;
+    return __result;
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_add(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __delta, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_add_cuda(__cxx_get_underlying_device_atomic(__a), __delta, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_add(&__a->__a_value, __delta, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp* __cxx_atomic_fetch_add(__cxx_atomic_base_heterogeneous_impl<_Tp*, _Sco, _Ref> volatile* __a, ptrdiff_t __delta, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_add_cuda(__cxx_get_underlying_device_atomic(__a), __delta, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_add(&__a->__a_value, __delta, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_sub(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __delta, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_sub_cuda(__cxx_get_underlying_device_atomic(__a), __delta, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_sub(&__a->__a_value, __delta, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp* __cxx_atomic_fetch_sub(__cxx_atomic_base_heterogeneous_impl<_Tp*, _Sco, _Ref> volatile* __a, ptrdiff_t __delta, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_sub_cuda(__cxx_get_underlying_device_atomic(__a), __delta, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_sub(&__a->__a_value, __delta, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_and(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __pattern, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_and_cuda(__cxx_get_underlying_device_atomic(__a), __pattern, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_and(&__a->__a_value, __pattern, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_or(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __pattern, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_or_cuda(__cxx_get_underlying_device_atomic(__a), __pattern, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_or(&__a->__a_value, __pattern, __order);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_xor(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Tp __pattern, memory_order __order) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_xor_cuda(__cxx_get_underlying_device_atomic(__a), __pattern, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ),
+        NV_IS_HOST, (
+            return __host::__cxx_atomic_fetch_xor(&__a->__a_value, __pattern, __order);
+        )
+    )
+}
+
+template <typename _Tp, typename _Delta, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_max(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Delta __val, memory_order __order) {
+    NV_IF_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_max_cuda(__cxx_get_underlying_device_atomic(__a), __val, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ), (
+            return __host::__cxx_atomic_fetch_max(&__a->__a_value, __val, __order);
+        )
+    )
+}
+
+template <typename _Tp, typename _Delta, int _Sco, bool _Ref>
+__host__ __device__
+ _Tp __cxx_atomic_fetch_min(__cxx_atomic_base_heterogeneous_impl<_Tp, _Sco, _Ref> volatile* __a, _Delta __val, memory_order __order) {
+    NV_IF_TARGET(
+        NV_IS_DEVICE, (
+            return __atomic_fetch_min_cuda(__cxx_get_underlying_device_atomic(__a), __val, static_cast<__memory_order_underlying_t>(__order), __scope_tag<_Sco>());
+        ), (
+            return __host::__cxx_atomic_fetch_min(&__a->__a_value, __val, __order);
+        )
+    )
+}
+
+template<class _Tp>
+__host__ __device__ inline uint32_t __cxx_small_to_32(_Tp __val) {
+    __cxx_small_proxy<_Tp> __temp = 0;
+    memcpy(&__temp, &__val, sizeof(_Tp));
+    return __temp;
+}
+
+template<class _Tp>
+__host__ __device__ inline _Tp __cxx_small_from_32(uint32_t __val) {
+    __cxx_small_proxy<_Tp> __temp = static_cast<__cxx_small_proxy<_Tp>>(__val);
+    _Tp __result;
+    memcpy(&__result, &__temp, sizeof(_Tp));
+    return __result;
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline void __cxx_atomic_init(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __val) {
+    __cxx_atomic_init(&__a->__a_value, __cxx_small_to_32(__val));
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline void __cxx_atomic_store(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __val, memory_order __order) {
+    __cxx_atomic_store(&__a->__a_value, __cxx_small_to_32(__val), __order);
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_load(__cxx_atomic_base_small_impl<_Tp, _Sco> const volatile* __a, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_load(&__a->__a_value, __order));
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_exchange(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __value, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_exchange(&__a->__a_value, __cxx_small_to_32(__value), __order));
+}
+__host__ __device__
+inline int __cuda_memcmp(void const * __lhs, void const * __rhs, size_t __count) {
+    NV_DISPATCH_TARGET(
+        NV_IS_DEVICE, (
+            auto __lhs_c = reinterpret_cast<unsigned char const *>(__lhs);
+            auto __rhs_c = reinterpret_cast<unsigned char const *>(__rhs);
+            while (__count--) {
+                auto const __lhs_v = *__lhs_c++;
+                auto const __rhs_v = *__rhs_c++;
+                if (__lhs_v < __rhs_v) { return -1; }
+                if (__lhs_v > __rhs_v) { return 1; }
+            }
+            return 0;
+        ),
+        NV_IS_HOST, (
+            return memcmp(__lhs, __rhs, __count);
+        )
+    )
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline bool __cxx_atomic_compare_exchange_weak(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp* __expected, _Tp __value, memory_order __success, memory_order __failure) {
+    auto __temp = __cxx_small_to_32(*__expected);
+    auto const __ret = __cxx_atomic_compare_exchange_weak(&__a->__a_value, &__temp, __cxx_small_to_32(__value), __success, __failure);
+    auto const __actual = __cxx_small_from_32<_Tp>(__temp);
+    if(!__ret) {
+        if(0 == __cuda_memcmp(&__actual, __expected, sizeof(_Tp)))
+            __cxx_atomic_fetch_and(&__a->__a_value, (1u << (8*sizeof(_Tp))) - 1, memory_order_relaxed);
+        else
+            *__expected = __actual;
+    }
+    return __ret;
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline bool __cxx_atomic_compare_exchange_strong(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp* __expected, _Tp __value, memory_order __success, memory_order __failure) {
+    auto const __old = *__expected;
+    while(1) {
+        if(__cxx_atomic_compare_exchange_weak(__a, __expected, __value, __success, __failure))
+            return true;
+        if(0 != __cuda_memcmp(&__old, __expected, sizeof(_Tp)))
+            return false;
+    }
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_add(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __delta, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_add(&__a->__a_value, __cxx_small_to_32(__delta), __order));
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_sub(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __delta, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_sub(&__a->__a_value, __cxx_small_to_32(__delta), __order));
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_and(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __pattern, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_and(&__a->__a_value, __cxx_small_to_32(__pattern), __order));
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_or(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __pattern, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_or(&__a->__a_value, __cxx_small_to_32(__pattern), __order));
+}
+
+template <typename _Tp, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_xor(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Tp __pattern, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_xor(&__a->__a_value, __cxx_small_to_32(__pattern), __order));
+}
+
+template <typename _Tp, typename _Delta, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_max(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Delta __val, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_max(&__a->__a_value, __cxx_small_to_32(__val), __order));
+}
+
+template <typename _Tp, typename _Delta, int _Sco>
+__host__ __device__ inline _Tp __cxx_atomic_fetch_min(__cxx_atomic_base_small_impl<_Tp, _Sco> volatile* __a, _Delta __val, memory_order __order) {
+    return __cxx_small_from_32<_Tp>(__cxx_atomic_fetch_min(&__a->__a_value, __cxx_small_to_32(__val), __order));
+}
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_derived.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_derived.h
new file mode 100644
index 000000000000..78afe9131264
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_derived.h
@@ -0,0 +1,143 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+template<class _Type, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type) <= 2, int>::type = 0>
+bool __device__ __atomic_compare_exchange_cuda(_Type volatile *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, _Scope __s) {
+
+    auto const __aligned = (uint32_t*)((intptr_t)__ptr & ~(sizeof(uint32_t) - 1));
+    auto const __offset = uint32_t((intptr_t)__ptr & (sizeof(uint32_t) - 1)) * 8;
+    auto const __mask = ((1 << sizeof(_Type)*8) - 1) << __offset;
+
+    uint32_t __old = *__expected << __offset;
+    uint32_t __old_value;
+    while (1) {
+        __old_value = (__old & __mask) >> __offset;
+        if (__old_value != *__expected)
+            break;
+        uint32_t const __attempt = (__old & ~__mask) | (*__desired << __offset);
+        if (__atomic_compare_exchange_cuda(__aligned, &__old, &__attempt, true, __success_memorder, __failure_memorder, __s))
+            return true;
+    }
+    *__expected = __old_value;
+    return false;
+}
+
+template<class _Type, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+void __device__ __atomic_exchange_cuda(_Type volatile *__ptr, _Type *__val, _Type *__ret, int __memorder, _Scope __s) {
+
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    while(!__atomic_compare_exchange_cuda(__ptr, &__expected, __val, true, __memorder, __memorder, __s))
+        ;
+    *__ret = __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __device__ __atomic_fetch_add_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected + __val;
+    while(!__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s))
+        __desired = __expected + __val;
+    return __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __host__ __device__ __atomic_fetch_max_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected > __val ? __expected : __val;
+
+    while(__desired == __val &&
+            !__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s)) {
+        __desired = __expected > __val ? __expected : __val;
+    }
+
+    return __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __host__ __device__ __atomic_fetch_min_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected < __val ? __expected : __val;
+
+    while(__desired != __val &&
+            !__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s)) {
+        __desired = __expected < __val ? __expected : __val;
+    }
+
+    return __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __device__ __atomic_fetch_sub_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected - __val;
+    while(!__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s))
+        __desired = __expected - __val;
+    return __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __device__ __atomic_fetch_and_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected & __val;
+    while(!__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s))
+        __desired = __expected & __val;
+    return __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __device__ __atomic_fetch_xor_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected ^ __val;
+    while(!__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s))
+        __desired = __expected ^ __val;
+    return __expected;
+}
+
+template<class _Type, class _Delta, class _Scope, typename _CUDA_VSTD::enable_if<sizeof(_Type)<=2, int>::type = 0>
+_Type __device__ __atomic_fetch_or_cuda(_Type volatile *__ptr, _Delta __val, int __memorder, _Scope __s) {
+
+    _Type __expected = __atomic_load_n_cuda(__ptr, __ATOMIC_RELAXED, __s);
+    _Type __desired = __expected | __val;
+    while(!__atomic_compare_exchange_cuda(__ptr, &__expected, &__desired, true, __memorder, __memorder, __s))
+        __desired = __expected | __val;
+    return __expected;
+}
+
+template<class _Type, class _Scope>
+_Type __device__ __atomic_load_n_cuda(const _Type volatile *__ptr, int __memorder, _Scope __s) {
+    _Type __ret;
+    __atomic_load_cuda(__ptr, &__ret, __memorder, __s);
+    return __ret;
+}
+
+template<class _Type, class _Scope>
+void __device__ __atomic_store_n_cuda(_Type volatile *__ptr, _Type __val, int __memorder, _Scope __s) {
+    __atomic_store_cuda(__ptr, &__val, __memorder, __s);
+}
+
+template<class _Type, class _Scope>
+bool __device__ __atomic_compare_exchange_n_cuda(_Type volatile *__ptr, _Type *__expected, _Type __desired, bool __weak, int __success_memorder, int __failure_memorder, _Scope __s) {
+    return __atomic_compare_exchange_cuda(__ptr, __expected, &__desired, __weak, __success_memorder, __failure_memorder, __s);
+}
+
+template<class _Type, class _Scope>
+_Type __device__ __atomic_exchange_n_cuda(_Type volatile *__ptr, _Type __val, int __memorder, _Scope __s) {
+    _Type __ret;
+    __atomic_exchange_cuda(__ptr, &__val, &__ret, __memorder, __s);
+    return __ret;
+}
+
+static inline __device__ void __atomic_signal_fence_cuda(int) {
+    asm volatile("":::"memory");
+}
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_generated.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_generated.h
new file mode 100644
index 000000000000..f4d8cd52c617
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_cuda_generated.h
@@ -0,0 +1,2644 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+
+static inline __device__ void __cuda_membar_block() { asm volatile("membar.cta;":::"memory"); }
+static inline __device__ void __cuda_fence_acq_rel_block() { asm volatile("fence.acq_rel.cta;":::"memory"); }
+static inline __device__ void __cuda_fence_sc_block() { asm volatile("fence.sc.cta;":::"memory"); }
+static inline __device__ void __atomic_thread_fence_cuda(int __memorder, __thread_scope_block_tag) {
+  NV_DISPATCH_TARGET(
+    NV_PROVIDES_SM_70, (
+      switch (__memorder) {
+        case __ATOMIC_SEQ_CST: __cuda_fence_sc_block(); break;
+        case __ATOMIC_CONSUME:
+        case __ATOMIC_ACQUIRE:
+        case __ATOMIC_ACQ_REL:
+        case __ATOMIC_RELEASE: __cuda_fence_acq_rel_block(); break;
+        case __ATOMIC_RELAXED: break;
+        default: assert(0);
+      }
+    ),
+    NV_IS_DEVICE, (
+      switch (__memorder) {
+        case __ATOMIC_SEQ_CST:
+        case __ATOMIC_CONSUME:
+        case __ATOMIC_ACQUIRE:
+        case __ATOMIC_ACQ_REL:
+        case __ATOMIC_RELEASE: __cuda_membar_block(); break;
+        case __ATOMIC_RELAXED: break;
+        default: assert(0);
+      }
+    )
+  )
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.acquire.cta.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.relaxed.cta.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.volatile.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_load_cuda(const volatile _Type *__ptr, _Type *__ret, int __memorder, __thread_scope_block_tag) {
+    uint32_t __tmp = 0;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_acquire_32_block(__ptr, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_load_relaxed_32_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_volatile_32_block(__ptr, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELAXED: __cuda_load_volatile_32_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 4);
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.acquire.cta.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.relaxed.cta.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.volatile.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_load_cuda(const volatile _Type *__ptr, _Type *__ret, int __memorder, __thread_scope_block_tag) {
+    uint64_t __tmp = 0;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_acquire_64_block(__ptr, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_load_relaxed_64_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_volatile_64_block(__ptr, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELAXED: __cuda_load_volatile_64_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 8);
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_relaxed_32_block(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.relaxed.cta.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_release_32_block(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.release.cta.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_volatile_32_block(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.volatile.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_store_cuda(volatile _Type *__ptr, _Type *__val, int __memorder, __thread_scope_block_tag) {
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, __val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE: __cuda_store_release_32_block(__ptr, __tmp); break;
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_RELAXED: __cuda_store_relaxed_32_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE:
+          case __ATOMIC_SEQ_CST: __cuda_membar_block();
+          case __ATOMIC_RELAXED: __cuda_store_volatile_32_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_relaxed_64_block(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.relaxed.cta.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_release_64_block(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.release.cta.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_volatile_64_block(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.volatile.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_store_cuda(volatile _Type *__ptr, _Type *__val, int __memorder, __thread_scope_block_tag) {
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, __val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE: __cuda_store_release_64_block(__ptr, __tmp); break;
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_RELAXED: __cuda_store_relaxed_64_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE:
+          case __ATOMIC_SEQ_CST: __cuda_membar_block();
+          case __ATOMIC_RELAXED: __cuda_store_volatile_64_block(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acq_rel.cta.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acquire.cta.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.relaxed.cta.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.release.cta.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.cta.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ bool __atomic_compare_exchange_cuda(volatile _Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, __thread_scope_block_tag) {
+    uint32_t __tmp = 0, __old = 0, __old_tmp;
+    memcpy(&__tmp, __desired, 4);
+    memcpy(&__old, __expected, 4);
+    __old_tmp = __old;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_32_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_32_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_compare_exchange_release_32_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_32_block(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_32_block(__ptr, __old, __old_tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_compare_exchange_volatile_32_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_32_block(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    bool const __ret = __old == __old_tmp;
+    memcpy(__expected, &__old, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acq_rel.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acquire.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.relaxed.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.release.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_exchange_cuda(volatile _Type *__ptr, _Type *__val, _Type *__ret, int __memorder, __thread_scope_block_tag) {
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, __val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_exchange_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_exchange_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_exchange_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 4);
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acq_rel.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acquire.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.relaxed.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.release.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_add_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_add_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acq_rel.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acquire.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.relaxed.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.release.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_and_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_and_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_and_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_and_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acq_rel.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acquire.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.relaxed.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.release.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_max_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_max_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_max_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_max_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acq_rel.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acquire.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.relaxed.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.release.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_min_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_min_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_min_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_min_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acq_rel.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acquire.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.relaxed.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.release.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_or_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_or_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_or_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_or_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acq_rel.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acquire.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.relaxed.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.release.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.cta.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_sub_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_sub_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_sub_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_sub_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acq_rel_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acq_rel.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acquire_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acquire.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_relaxed_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.relaxed.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_release_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.release.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_volatile_32_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.cta.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_xor_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_acquire_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_xor_acq_rel_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_xor_release_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_relaxed_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_volatile_32_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_xor_volatile_32_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_volatile_32_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acq_rel.cta.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acquire.cta.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.relaxed.cta.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.release.cta.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.cta.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ bool __atomic_compare_exchange_cuda(volatile _Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, __thread_scope_block_tag) {
+    uint64_t __tmp = 0, __old = 0, __old_tmp;
+    memcpy(&__tmp, __desired, 8);
+    memcpy(&__old, __expected, 8);
+    __old_tmp = __old;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_64_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_64_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_compare_exchange_release_64_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_64_block(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_64_block(__ptr, __old, __old_tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_compare_exchange_volatile_64_block(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_64_block(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    bool const __ret = __old == __old_tmp;
+    memcpy(__expected, &__old, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acq_rel.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acquire.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.relaxed.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.release.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_exchange_cuda(volatile _Type *__ptr, _Type *__val, _Type *__ret, int __memorder, __thread_scope_block_tag) {
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, __val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_exchange_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_exchange_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_exchange_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 8);
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acq_rel.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acquire.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.relaxed.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.release.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_add_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acq_rel.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acquire.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.relaxed.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.release.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_and_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_and_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_and_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_and_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acq_rel.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acquire.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.relaxed.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.release.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_max_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_max_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_max_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_max_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acq_rel.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acquire.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.relaxed.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.release.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_min_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_min_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_min_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_min_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acq_rel.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acquire.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.relaxed.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.release.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_or_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_or_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_or_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_or_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acq_rel.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acquire.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.relaxed.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.release.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.cta.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_sub_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_sub_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_sub_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_sub_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acq_rel_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acq_rel.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acquire_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acquire.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_relaxed_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.relaxed.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_release_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.release.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_volatile_64_block(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.cta.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_xor_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_block_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_xor_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_xor_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_relaxed_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_xor_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _Type>
+__device__ _Type* __atomic_fetch_add_cuda(_Type *volatile *__ptr, ptrdiff_t __val, int __memorder, __thread_scope_block_tag) {
+    _Type* __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    __tmp *= sizeof(_Type);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_block(__ptr, __tmp, __tmp); break;
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _Type>
+__device__ _Type* __atomic_fetch_sub_cuda(_Type *volatile *__ptr, ptrdiff_t __val, int __memorder, __thread_scope_block_tag) {
+    _Type* __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    __tmp = -__tmp;
+    __tmp *= sizeof(_Type);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_block(__ptr, __tmp, __tmp); break;
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_block();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); __cuda_membar_block(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_block(); __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_block(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+static inline __device__ void __cuda_membar_device() { asm volatile("membar.gl;":::"memory"); }
+static inline __device__ void __cuda_fence_acq_rel_device() { asm volatile("fence.acq_rel.gpu;":::"memory"); }
+static inline __device__ void __cuda_fence_sc_device() { asm volatile("fence.sc.gpu;":::"memory"); }
+static inline __device__ void __atomic_thread_fence_cuda(int __memorder, __thread_scope_device_tag) {
+  NV_DISPATCH_TARGET(
+    NV_PROVIDES_SM_70, (
+      switch (__memorder) {
+        case __ATOMIC_SEQ_CST: __cuda_fence_sc_device(); break;
+        case __ATOMIC_CONSUME:
+        case __ATOMIC_ACQUIRE:
+        case __ATOMIC_ACQ_REL:
+        case __ATOMIC_RELEASE: __cuda_fence_acq_rel_device(); break;
+        case __ATOMIC_RELAXED: break;
+        default: assert(0);
+      }
+    ),
+    NV_IS_DEVICE, (
+      switch (__memorder) {
+        case __ATOMIC_SEQ_CST:
+        case __ATOMIC_CONSUME:
+        case __ATOMIC_ACQUIRE:
+        case __ATOMIC_ACQ_REL:
+        case __ATOMIC_RELEASE: __cuda_membar_device(); break;
+        case __ATOMIC_RELAXED: break;
+        default: assert(0);
+      }
+    )
+  )
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.acquire.gpu.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.relaxed.gpu.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.volatile.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_load_cuda(const volatile _Type *__ptr, _Type *__ret, int __memorder, __thread_scope_device_tag) {
+    uint32_t __tmp = 0;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_acquire_32_device(__ptr, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_load_relaxed_32_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_volatile_32_device(__ptr, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELAXED: __cuda_load_volatile_32_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 4);
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.acquire.gpu.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.relaxed.gpu.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.volatile.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_load_cuda(const volatile _Type *__ptr, _Type *__ret, int __memorder, __thread_scope_device_tag) {
+    uint64_t __tmp = 0;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_acquire_64_device(__ptr, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_load_relaxed_64_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_volatile_64_device(__ptr, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELAXED: __cuda_load_volatile_64_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 8);
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_relaxed_32_device(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.relaxed.gpu.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_release_32_device(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.release.gpu.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_volatile_32_device(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.volatile.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_store_cuda(volatile _Type *__ptr, _Type *__val, int __memorder, __thread_scope_device_tag) {
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, __val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE: __cuda_store_release_32_device(__ptr, __tmp); break;
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_RELAXED: __cuda_store_relaxed_32_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE:
+          case __ATOMIC_SEQ_CST: __cuda_membar_device();
+          case __ATOMIC_RELAXED: __cuda_store_volatile_32_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_relaxed_64_device(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.relaxed.gpu.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_release_64_device(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.release.gpu.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_volatile_64_device(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.volatile.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_store_cuda(volatile _Type *__ptr, _Type *__val, int __memorder, __thread_scope_device_tag) {
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, __val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE: __cuda_store_release_64_device(__ptr, __tmp); break;
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_RELAXED: __cuda_store_relaxed_64_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE:
+          case __ATOMIC_SEQ_CST: __cuda_membar_device();
+          case __ATOMIC_RELAXED: __cuda_store_volatile_64_device(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acq_rel.gpu.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acquire.gpu.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.relaxed.gpu.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.release.gpu.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.gpu.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ bool __atomic_compare_exchange_cuda(volatile _Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, __thread_scope_device_tag) {
+    uint32_t __tmp = 0, __old = 0, __old_tmp;
+    memcpy(&__tmp, __desired, 4);
+    memcpy(&__old, __expected, 4);
+    __old_tmp = __old;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_32_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_32_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_compare_exchange_release_32_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_32_device(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_32_device(__ptr, __old, __old_tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_compare_exchange_volatile_32_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_32_device(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    bool const __ret = __old == __old_tmp;
+    memcpy(__expected, &__old, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acq_rel.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acquire.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.relaxed.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.release.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_exchange_cuda(volatile _Type *__ptr, _Type *__val, _Type *__ret, int __memorder, __thread_scope_device_tag) {
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, __val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_exchange_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_exchange_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_exchange_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 4);
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acq_rel.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acquire.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.relaxed.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.release.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_add_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_add_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acq_rel.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acquire.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.relaxed.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.release.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_and_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_and_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_and_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_and_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acq_rel.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acquire.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.relaxed.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.release.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_max_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_max_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_max_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_max_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acq_rel.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acquire.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.relaxed.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.release.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_min_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_min_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_min_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_min_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acq_rel.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acquire.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.relaxed.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.release.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_or_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_or_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_or_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_or_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acq_rel.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acquire.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.relaxed.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.release.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.gpu.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_sub_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_sub_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_sub_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_sub_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acq_rel_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acq_rel.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acquire_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acquire.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_relaxed_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.relaxed.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_release_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.release.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_volatile_32_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.gpu.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_xor_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_acquire_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_xor_acq_rel_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_xor_release_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_relaxed_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_volatile_32_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_xor_volatile_32_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_volatile_32_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acq_rel.gpu.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acquire.gpu.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.relaxed.gpu.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.release.gpu.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.gpu.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ bool __atomic_compare_exchange_cuda(volatile _Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, __thread_scope_device_tag) {
+    uint64_t __tmp = 0, __old = 0, __old_tmp;
+    memcpy(&__tmp, __desired, 8);
+    memcpy(&__old, __expected, 8);
+    __old_tmp = __old;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_64_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_64_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_compare_exchange_release_64_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_64_device(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_64_device(__ptr, __old, __old_tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_compare_exchange_volatile_64_device(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_64_device(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    bool const __ret = __old == __old_tmp;
+    memcpy(__expected, &__old, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acq_rel.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acquire.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.relaxed.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.release.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_exchange_cuda(volatile _Type *__ptr, _Type *__val, _Type *__ret, int __memorder, __thread_scope_device_tag) {
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, __val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_exchange_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_exchange_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_exchange_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 8);
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acq_rel.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acquire.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.relaxed.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.release.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_add_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acq_rel.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acquire.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.relaxed.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.release.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_and_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_and_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_and_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_and_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acq_rel.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acquire.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.relaxed.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.release.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_max_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_max_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_max_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_max_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acq_rel.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acquire.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.relaxed.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.release.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_min_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_min_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_min_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_min_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acq_rel.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acquire.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.relaxed.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.release.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_or_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_or_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_or_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_or_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acq_rel.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acquire.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.relaxed.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.release.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.gpu.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_sub_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_sub_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_sub_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_sub_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acq_rel_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acq_rel.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acquire_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acquire.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_relaxed_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.relaxed.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_release_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.release.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_volatile_64_device(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.gpu.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_xor_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_device_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_xor_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_xor_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_relaxed_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_xor_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _Type>
+__device__ _Type* __atomic_fetch_add_cuda(_Type *volatile *__ptr, ptrdiff_t __val, int __memorder, __thread_scope_device_tag) {
+    _Type* __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    __tmp *= sizeof(_Type);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_device(__ptr, __tmp, __tmp); break;
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _Type>
+__device__ _Type* __atomic_fetch_sub_cuda(_Type *volatile *__ptr, ptrdiff_t __val, int __memorder, __thread_scope_device_tag) {
+    _Type* __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    __tmp = -__tmp;
+    __tmp *= sizeof(_Type);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_device(__ptr, __tmp, __tmp); break;
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_device();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); __cuda_membar_device(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_device(); __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_device(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+static inline __device__ void __cuda_membar_system() { asm volatile("membar.sys;":::"memory"); }
+static inline __device__ void __cuda_fence_acq_rel_system() { asm volatile("fence.acq_rel.sys;":::"memory"); }
+static inline __device__ void __cuda_fence_sc_system() { asm volatile("fence.sc.sys;":::"memory"); }
+static inline __device__ void __atomic_thread_fence_cuda(int __memorder, __thread_scope_system_tag) {
+  NV_DISPATCH_TARGET(
+    NV_PROVIDES_SM_70, (
+      switch (__memorder) {
+        case __ATOMIC_SEQ_CST: __cuda_fence_sc_system(); break;
+        case __ATOMIC_CONSUME:
+        case __ATOMIC_ACQUIRE:
+        case __ATOMIC_ACQ_REL:
+        case __ATOMIC_RELEASE: __cuda_fence_acq_rel_system(); break;
+        case __ATOMIC_RELAXED: break;
+        default: assert(0);
+      }
+    ),
+    NV_IS_DEVICE, (
+      switch (__memorder) {
+        case __ATOMIC_SEQ_CST:
+        case __ATOMIC_CONSUME:
+        case __ATOMIC_ACQUIRE:
+        case __ATOMIC_ACQ_REL:
+        case __ATOMIC_RELEASE: __cuda_membar_system(); break;
+        case __ATOMIC_RELAXED: break;
+        default: assert(0);
+      }
+    )
+  )
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.acquire.sys.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.relaxed.sys.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.volatile.b32 %0,[%1];" : "=r"(__dst) : "l"(__ptr) : "memory"); }
+template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_load_cuda(const volatile _Type *__ptr, _Type *__ret, int __memorder, __thread_scope_system_tag) {
+    uint32_t __tmp = 0;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_acquire_32_system(__ptr, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_load_relaxed_32_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_volatile_32_system(__ptr, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELAXED: __cuda_load_volatile_32_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 4);
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.acquire.sys.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.relaxed.sys.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_load_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst) {asm volatile("ld.volatile.b64 %0,[%1];" : "=l"(__dst) : "l"(__ptr) : "memory"); }
+template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_load_cuda(const volatile _Type *__ptr, _Type *__ret, int __memorder, __thread_scope_system_tag) {
+    uint64_t __tmp = 0;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_acquire_64_system(__ptr, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_load_relaxed_64_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_load_volatile_64_system(__ptr, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELAXED: __cuda_load_volatile_64_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 8);
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_relaxed_32_system(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.relaxed.sys.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_release_32_system(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.release.sys.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_volatile_32_system(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.volatile.b32 [%0], %1;" :: "l"(__ptr),"r"(__src) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_store_cuda(volatile _Type *__ptr, _Type *__val, int __memorder, __thread_scope_system_tag) {
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, __val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE: __cuda_store_release_32_system(__ptr, __tmp); break;
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_RELAXED: __cuda_store_relaxed_32_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE:
+          case __ATOMIC_SEQ_CST: __cuda_membar_system();
+          case __ATOMIC_RELAXED: __cuda_store_volatile_32_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+}
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_relaxed_64_system(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.relaxed.sys.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_release_64_system(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.release.sys.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _CUDA_A, class _CUDA_B> static inline __device__ void __cuda_store_volatile_64_system(_CUDA_A __ptr, _CUDA_B __src) { asm volatile("st.volatile.b64 [%0], %1;" :: "l"(__ptr),"l"(__src) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_store_cuda(volatile _Type *__ptr, _Type *__val, int __memorder, __thread_scope_system_tag) {
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, __val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE: __cuda_store_release_64_system(__ptr, __tmp); break;
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_RELAXED: __cuda_store_relaxed_64_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_RELEASE:
+          case __ATOMIC_SEQ_CST: __cuda_membar_system();
+          case __ATOMIC_RELAXED: __cuda_store_volatile_64_system(__ptr, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acq_rel.sys.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acquire.sys.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.relaxed.sys.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.release.sys.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.sys.b32 %0,[%1],%2,%3;" : "=r"(__dst) : "l"(__ptr),"r"(__cmp),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ bool __atomic_compare_exchange_cuda(volatile _Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, __thread_scope_system_tag) {
+    uint32_t __tmp = 0, __old = 0, __old_tmp;
+    memcpy(&__tmp, __desired, 4);
+    memcpy(&__old, __expected, 4);
+    __old_tmp = __old;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_32_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_32_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_compare_exchange_release_32_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_32_system(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_32_system(__ptr, __old, __old_tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_compare_exchange_volatile_32_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_32_system(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    bool const __ret = __old == __old_tmp;
+    memcpy(__expected, &__old, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acq_rel.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acquire.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.relaxed.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.release.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ void __atomic_exchange_cuda(volatile _Type *__ptr, _Type *__val, _Type *__ret, int __memorder, __thread_scope_system_tag) {
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, __val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_exchange_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_exchange_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_exchange_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 4);
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acq_rel.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acquire.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.relaxed.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.release.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_add_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_add_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acq_rel.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acquire.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.relaxed.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.release.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_and_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_and_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_and_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_and_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acq_rel.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acquire.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.relaxed.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.release.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_max_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_max_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_max_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_max_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acq_rel.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acquire.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.relaxed.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.release.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_min_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_min_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_min_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_min_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acq_rel.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acquire.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.relaxed.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.release.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_or_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_or_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_or_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_or_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acq_rel.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acquire.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.relaxed.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.release.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.sys.u32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_sub_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_sub_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_sub_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_sub_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acq_rel_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acq_rel.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acquire_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acquire.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_relaxed_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.relaxed.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_release_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.release.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_volatile_32_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.sys.b32 %0,[%1],%2;" : "=r"(__dst) : "l"(__ptr),"r"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==4, int>::type = 0>
+__device__ _Type __atomic_fetch_xor_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint32_t __tmp = 0;
+    memcpy(&__tmp, &__val, 4);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_acquire_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_xor_acq_rel_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_xor_release_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_relaxed_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_volatile_32_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_xor_volatile_32_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_volatile_32_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 4);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acq_rel.sys.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.acquire.sys.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.relaxed.sys.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.release.sys.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> static inline __device__ void __cuda_compare_exchange_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op) { asm volatile("atom.cas.sys.b64 %0,[%1],%2,%3;" : "=l"(__dst) : "l"(__ptr),"l"(__cmp),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ bool __atomic_compare_exchange_cuda(volatile _Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, __thread_scope_system_tag) {
+    uint64_t __tmp = 0, __old = 0, __old_tmp;
+    memcpy(&__tmp, __desired, 8);
+    memcpy(&__old, __expected, 8);
+    __old_tmp = __old;
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_64_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_64_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_compare_exchange_release_64_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_64_system(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_64_system(__ptr, __old, __old_tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_compare_exchange_volatile_64_system(__ptr, __old, __old_tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_64_system(__ptr, __old, __old_tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    bool const __ret = __old == __old_tmp;
+    memcpy(__expected, &__old, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acq_rel.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.acquire.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.relaxed.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.release.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_exchange_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.exch.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ void __atomic_exchange_cuda(volatile _Type *__ptr, _Type *__val, _Type *__ret, int __memorder, __thread_scope_system_tag) {
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, __val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_exchange_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_exchange_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_exchange_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_exchange_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_exchange_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(__ret, &__tmp, 8);
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acq_rel.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.acquire.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.relaxed.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.release.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_add_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.add.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_add_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acq_rel.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.acquire.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.relaxed.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.release.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_and_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.and.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_and_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_and_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_and_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_and_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_and_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_and_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acq_rel.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.acquire.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.relaxed.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.release.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_max_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.max.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_max_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_max_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_max_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_max_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_max_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_max_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acq_rel.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.acquire.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.relaxed.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.release.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_min_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.min.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_min_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_min_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_min_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_min_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_min_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_min_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acq_rel.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.acquire.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.relaxed.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.release.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_or_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.or.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_or_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_or_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_or_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_or_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_or_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_or_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acq_rel.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.acquire.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.relaxed.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.release.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_sub_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { __op = -__op;
+asm volatile("atom.add.sys.u64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_sub_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_sub_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_sub_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_sub_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_sub_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_sub_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acq_rel_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acq_rel.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_acquire_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.acquire.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_relaxed_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.relaxed.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_release_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.release.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _CUDA_A, class _CUDA_B, class _CUDA_C> static inline __device__ void __cuda_fetch_xor_volatile_64_system(_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op) { asm volatile("atom.xor.sys.b64 %0,[%1],%2;" : "=l"(__dst) : "l"(__ptr),"l"(__op) : "memory"); }
+template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==8, int>::type = 0>
+__device__ _Type __atomic_fetch_xor_cuda(volatile _Type *__ptr, _Type __val, int __memorder, __thread_scope_system_tag) {
+    _Type __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_xor_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_xor_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_relaxed_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_xor_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_xor_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_xor_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _Type>
+__device__ _Type* __atomic_fetch_add_cuda(_Type *volatile *__ptr, ptrdiff_t __val, int __memorder, __thread_scope_system_tag) {
+    _Type* __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    __tmp *= sizeof(_Type);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_system(__ptr, __tmp, __tmp); break;
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
+template<class _Type>
+__device__ _Type* __atomic_fetch_sub_cuda(_Type *volatile *__ptr, ptrdiff_t __val, int __memorder, __thread_scope_system_tag) {
+    _Type* __ret;
+    uint64_t __tmp = 0;
+    memcpy(&__tmp, &__val, 8);
+    __tmp = -__tmp;
+    __tmp *= sizeof(_Type);
+    NV_DISPATCH_TARGET(
+      NV_PROVIDES_SM_70, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST: __cuda_fence_sc_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_system(__ptr, __tmp, __tmp); break;
+        }
+      ),
+      NV_IS_DEVICE, (
+        switch (__memorder) {
+          case __ATOMIC_SEQ_CST:
+          case __ATOMIC_ACQ_REL: __cuda_membar_system();
+          case __ATOMIC_CONSUME:
+          case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); __cuda_membar_system(); break;
+          case __ATOMIC_RELEASE: __cuda_membar_system(); __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); break;
+          case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_system(__ptr, __tmp, __tmp); break;
+          default: assert(0);
+        }
+      )
+    )
+    memcpy(&__ret, &__tmp, 8);
+    return __ret;
+}
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_gcc.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_gcc.h
new file mode 100644
index 000000000000..22e976b802fd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_gcc.h
@@ -0,0 +1,16 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ATOMIC_GCC_H
+#define _LIBCUDACXX_ATOMIC_GCC_H
+
+#include "atomic_base.h"
+
+#endif // _LIBCUDACXX_ATOMIC_GCC_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_msvc.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_msvc.h
new file mode 100644
index 000000000000..b99b8015ca1d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_msvc.h
@@ -0,0 +1,449 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _MSC_VER
+#error "This file is only for CL.EXE's benefit"
+#endif
+
+#define _Compiler_barrier() _ReadWriteBarrier()
+
+#if defined(_M_ARM) || defined(_M_ARM64)
+    #define _Memory_barrier()             __dmb(0xB) // inner shared data memory barrier
+    #define _Compiler_or_memory_barrier() _Memory_barrier()
+#elif defined(_M_IX86) || defined(_M_X64)
+    #define _Memory_barrier()             __faststorefence()
+    // x86/x64 hardware only emits memory barriers inside _Interlocked intrinsics
+    #define _Compiler_or_memory_barrier() _Compiler_barrier()
+#else // ^^^ x86/x64 / unsupported hardware vvv
+    #error Unsupported hardware
+#endif // hardware
+
+// MSVC Does not have compiler intrinsics for lock-free checking
+#ifndef _LIBCUDACXX_ATOMIC_IS_LOCK_FREE
+#define _LIBCUDACXX_ATOMIC_IS_LOCK_FREE(__x) (__x <= 8)
+#endif
+
+inline int __stronger_order_msvc(int __a, int __b) {
+    int const __max = __a > __b ? __a : __b;
+    if(__max != __ATOMIC_RELEASE)
+        return __max;
+    static int const __xform[] = {
+        __ATOMIC_RELEASE,
+        __ATOMIC_ACQ_REL,
+        __ATOMIC_ACQ_REL,
+        __ATOMIC_RELEASE };
+    return __xform[__a < __b ? __a : __b];
+}
+
+static inline void __atomic_signal_fence(int __memorder) {
+    if (__memorder != __ATOMIC_RELAXED)
+        _Compiler_barrier();
+}
+
+static inline void __atomic_thread_fence(int __memorder) {
+    if (__memorder != __ATOMIC_RELAXED)
+        _Memory_barrier();
+}
+
+template <typename _Type, size_t _Size>
+using _enable_if_sized_as = typename enable_if<sizeof(_Type) == _Size, int>::type;
+
+template<class _Type, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_load_relaxed(const volatile _Type *__ptr, _Type *__ret) {
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    __int8 __tmp = *(const volatile __int8 *)__ptr;
+#else
+    __int8 __tmp = __iso_volatile_load8((const volatile __int8 *)__ptr);
+#endif
+    *__ret = reinterpret_cast<_Type&>(__tmp);
+}
+template<class _Type, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_load_relaxed(const volatile _Type *__ptr, _Type *__ret) {
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    __int16 __tmp = *(const volatile __int16 *)__ptr;
+#else
+    __int16 __tmp = __iso_volatile_load16((const volatile __int16 *)__ptr);
+#endif
+    *__ret = reinterpret_cast<_Type&>(__tmp);
+}
+template<class _Type, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_load_relaxed(const volatile _Type *__ptr, _Type *__ret) {
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    __int32 __tmp = *(const volatile __int32 *)__ptr;
+#else
+    __int32 __tmp = __iso_volatile_load32((const volatile __int32 *)__ptr);
+#endif
+    *__ret = reinterpret_cast<_Type&>(__tmp);
+}
+template<class _Type, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_load_relaxed(const volatile _Type *__ptr, _Type *__ret) {
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    __int64 __tmp = *(const volatile __int64 *)__ptr;
+#else
+    __int64 __tmp = __iso_volatile_load64((const volatile __int64 *)__ptr);
+#endif
+    *__ret = reinterpret_cast<_Type&>(__tmp);
+}
+
+template<class _Type>
+void __atomic_load(const volatile _Type *__ptr, _Type *__ret, int __memorder) {
+    switch (__memorder) {
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: __atomic_load_relaxed(__ptr, __ret); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: __atomic_load_relaxed(__ptr, __ret); break;
+    default: assert(0);
+    }
+}
+
+template<class _Type, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_store_relaxed(volatile _Type *__ptr, _Type *__val) {
+    auto __t = reinterpret_cast<__int8 *>(__val);
+    auto __d = reinterpret_cast<volatile __int8 *>(__ptr);
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    (void)_InterlockedExchange8(__d, *__t);
+#else
+    __iso_volatile_store8(__d, *__t);
+#endif
+}
+template<class _Type, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_store_relaxed(volatile _Type *__ptr, _Type *__val) {
+    auto __t = reinterpret_cast<__int16 *>(__val);
+    auto __d = reinterpret_cast<volatile __int16 *>(__ptr);
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    (void)_InterlockedExchange16(__d, *__t);
+#else
+    __iso_volatile_store16(__d, *__t);
+#endif
+}
+template<class _Type, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_store_relaxed(volatile _Type *__ptr, _Type *__val) {
+    auto __t = reinterpret_cast<__int32 *>(__val);
+    auto __d = reinterpret_cast<volatile __int32 *>(__ptr);
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    // int cannot be converted to long?...
+    (void)_InterlockedExchange(reinterpret_cast<volatile long *>(__d), *__t);
+#else
+    __iso_volatile_store32(__d, *__t);
+#endif
+}
+template<class _Type, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_store_relaxed(volatile _Type *__ptr, _Type *__val) {
+    auto __t = reinterpret_cast<__int64 *>(__val);
+    auto __d = reinterpret_cast<volatile __int64 *>(__ptr);
+#ifdef _LIBCUDACXX_MSVC_HAS_NO_ISO_INTRIN
+    (void)_InterlockedExchange64(__d, *__t);
+#else
+    __iso_volatile_store64(__d, *__t);
+#endif
+}
+
+template<class _Type>
+void __atomic_store(volatile _Type *__ptr, _Type *__val, int __memorder) {
+    switch (__memorder) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_store_relaxed(__ptr, __val); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_RELAXED: __atomic_store_relaxed(__ptr, __val); break;
+    default: assert(0);
+    }
+}
+
+template<class _Type, _enable_if_sized_as<_Type,1> = 0>
+bool __atomic_compare_exchange_relaxed(const volatile _Type *__ptr, _Type *__expected, const _Type *__desired) {
+    auto __tmp_desired = reinterpret_cast<const char&>(*__desired);
+    auto __tmp_expected = reinterpret_cast<char&>(*__expected);
+    auto const __old = _InterlockedCompareExchange8((volatile char *)__ptr, __tmp_desired, __tmp_expected);
+    if(__old == __tmp_expected)
+        return true;
+    *__expected = reinterpret_cast<const _Type&>(__old);
+    return false;
+}
+template<class _Type, _enable_if_sized_as<_Type,2> = 0>
+bool __atomic_compare_exchange_relaxed(const volatile _Type *__ptr, _Type *__expected, const _Type *__desired) {
+    auto __tmp_desired = reinterpret_cast<const short&>(*__desired);
+    auto __tmp_expected = reinterpret_cast<short&>(*__expected);
+    auto const __old = _InterlockedCompareExchange16((volatile short *)__ptr, __tmp_desired, __tmp_expected);
+    if(__old == __tmp_expected)
+        return true;
+    *__expected = reinterpret_cast<const _Type&>(__old);
+    return false;
+}
+template<class _Type, _enable_if_sized_as<_Type,4> = 0>
+bool __atomic_compare_exchange_relaxed(const volatile _Type *__ptr, _Type *__expected, const _Type *__desired) {
+    auto __tmp_desired = reinterpret_cast<const long&>(*__desired);
+    auto __tmp_expected = reinterpret_cast<long&>(*__expected);
+    auto const __old = _InterlockedCompareExchange((volatile long *)__ptr, __tmp_desired, __tmp_expected);
+    if(__old == __tmp_expected)
+        return true;
+    *__expected = reinterpret_cast<const _Type&>(__old);
+    return false;
+}
+template<class _Type, _enable_if_sized_as<_Type,8> = 0>
+bool __atomic_compare_exchange_relaxed(const volatile _Type *__ptr, _Type *__expected, const _Type *__desired) {
+    auto __tmp_desired = reinterpret_cast<const __int64&>(*__desired);
+    auto __tmp_expected = reinterpret_cast<__int64&>(*__expected);
+    auto const __old = _InterlockedCompareExchange64((volatile __int64 *)__ptr, __tmp_desired, __tmp_expected);
+    if(__old == __tmp_expected)
+        return true;
+    *__expected = reinterpret_cast<const _Type&>(__old);
+    return false;
+}
+template<class _Type>
+bool __atomic_compare_exchange(_Type volatile *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder) {
+    bool success = false;
+    switch (__stronger_order_msvc(__success_memorder, __failure_memorder)) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); break;
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: success = __atomic_compare_exchange_relaxed(__ptr, __expected, __desired); break;
+    default: assert(0);
+    }
+    return success;
+}
+
+template<class _Type, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_exchange_relaxed(const volatile _Type *__ptr, const _Type *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchange8((volatile char *)__ptr, reinterpret_cast<char const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_exchange_relaxed(const volatile _Type *__ptr, const _Type *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchange16((volatile short *)__ptr, reinterpret_cast<short const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_exchange_relaxed(const volatile _Type *__ptr, const _Type *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchange((volatile long *)__ptr, reinterpret_cast<long const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_exchange_relaxed(const volatile _Type *__ptr, const _Type *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchange64((volatile __int64 *)__ptr, reinterpret_cast<__int64 const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type>
+void __atomic_exchange(_Type volatile *__ptr, const _Type *__val, _Type *__ret, int __memorder) {
+    switch (__memorder) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_exchange_relaxed(__ptr, __val, __ret);break;
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: __atomic_exchange_relaxed(__ptr, __val, __ret); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_exchange_relaxed(__ptr, __val, __ret); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: __atomic_exchange_relaxed(__ptr, __val, __ret); break;
+    default: assert(0);
+    }
+}
+
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_fetch_add_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchangeAdd8((volatile char *)__ptr, reinterpret_cast<char const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_fetch_add_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchangeAdd16((volatile short *)__ptr, reinterpret_cast<short const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_fetch_add_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchangeAdd((volatile long *)__ptr, reinterpret_cast<long const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_fetch_add_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedExchangeAdd64((volatile __int64 *)__ptr, reinterpret_cast<__int64 const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta>
+_Type __atomic_fetch_add(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    alignas(_Type) unsigned char __buf[sizeof(_Type)] = {};
+    auto* __dest = reinterpret_cast<_Type*>(__buf);
+
+    switch (__memorder) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_add_relaxed(__ptr, &__val, __dest);break;
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: __atomic_fetch_add_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_add_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: __atomic_fetch_add_relaxed(__ptr, &__val, __dest); break;
+    default: assert(0);
+    }
+    return *__dest;
+}
+template<class _Type, class _Delta>
+_Type __atomic_fetch_sub(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    return __atomic_fetch_add(__ptr, 0-__val, __memorder);
+}
+
+
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_fetch_and_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedAnd8((volatile char *)__ptr, reinterpret_cast<char const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_fetch_and_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedAnd16((volatile short *)__ptr, reinterpret_cast<short const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_fetch_and_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedAnd((volatile long *)__ptr, reinterpret_cast<long const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_fetch_and_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedAnd64((volatile __int64 *)__ptr, reinterpret_cast<__int64 const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta>
+_Type __atomic_fetch_and(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    alignas(_Type) unsigned char __buf[sizeof(_Type)] = {};
+    auto* __dest = reinterpret_cast<_Type*>(__buf);
+
+    switch (__memorder) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_and_relaxed(__ptr, &__val, __dest);break;
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: __atomic_fetch_and_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_and_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: __atomic_fetch_and_relaxed(__ptr, &__val, __dest); break;
+    default: assert(0);
+    }
+    return *__dest;
+}
+
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_fetch_xor_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedXor8((volatile char *)__ptr, reinterpret_cast<char const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_fetch_xor_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedXor16((volatile short *)__ptr, reinterpret_cast<short const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_fetch_xor_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedXor((volatile long *)__ptr, reinterpret_cast<long const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_fetch_xor_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedXor64((volatile __int64 *)__ptr, reinterpret_cast<__int64 const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta>
+_Type __atomic_fetch_xor(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    alignas(_Type) unsigned char __buf[sizeof(_Type)] = {};
+    auto* __dest = reinterpret_cast<_Type*>(__buf);
+
+    switch (__memorder) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_xor_relaxed(__ptr, &__val, __dest);break;
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: __atomic_fetch_xor_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_xor_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: __atomic_fetch_xor_relaxed(__ptr, &__val, __dest); break;
+    default: assert(0);
+    }
+    return *__dest;
+}
+
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,1> = 0>
+void __atomic_fetch_or_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedOr8((volatile char *)__ptr, reinterpret_cast<char const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,2> = 0>
+void __atomic_fetch_or_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedOr16((volatile short *)__ptr, reinterpret_cast<short const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,4> = 0>
+void __atomic_fetch_or_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedOr((volatile long *)__ptr, reinterpret_cast<long const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta, _enable_if_sized_as<_Type,8> = 0>
+void __atomic_fetch_or_relaxed(const volatile _Type *__ptr, const _Delta *__val, _Type *__ret) {
+    auto const __old = _InterlockedOr64((volatile __int64 *)__ptr, reinterpret_cast<__int64 const&>(*__val));
+    *__ret = reinterpret_cast<_Type const&>(__old);
+}
+template<class _Type, class _Delta>
+_Type __atomic_fetch_or(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    alignas(_Type) unsigned char __buf[sizeof(_Type)] = {};
+    auto* __dest = reinterpret_cast<_Type*>(__buf);
+
+    switch (__memorder) {
+    case __ATOMIC_RELEASE: _Compiler_or_memory_barrier(); __atomic_fetch_or_relaxed(__ptr, &__val, __dest);break;
+    case __ATOMIC_ACQ_REL: _Compiler_or_memory_barrier(); _LIBCUDACXX_FALLTHROUGH();
+    case __ATOMIC_CONSUME:
+    case __ATOMIC_ACQUIRE: __atomic_fetch_or_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_SEQ_CST: _Memory_barrier(); __atomic_fetch_or_relaxed(__ptr, &__val, __dest); _Compiler_or_memory_barrier(); break;
+    case __ATOMIC_RELAXED: __atomic_fetch_or_relaxed(__ptr, &__val, __dest); break;
+    default: assert(0);
+    }
+    return *__dest;
+}
+
+template<class _Type>
+_Type __atomic_load_n(const _Type volatile *__ptr, int __memorder) {
+    alignas(_Type) unsigned char __buf[sizeof(_Type)] = {};
+    auto* __dest = reinterpret_cast<_Type*>(__buf);
+
+    __atomic_load(__ptr, __dest, __memorder);
+    return *__dest;
+}
+
+template<class _Type>
+void __atomic_store_n(_Type volatile *__ptr, _Type __val, int __memorder) {
+    __atomic_store(__ptr, &__val, __memorder);
+}
+
+template<class _Type>
+bool __atomic_compare_exchange_n(_Type volatile *__ptr, _Type *__expected, _Type __desired, bool __weak, int __success_memorder, int __failure_memorder) {
+    return __atomic_compare_exchange(__ptr, __expected, &__desired, __weak, __success_memorder, __failure_memorder);
+}
+
+template<class _Type>
+_Type __atomic_exchange_n(_Type volatile *__ptr, _Type __val, int __memorder) {
+    alignas(_Type) unsigned char __buf[sizeof(_Type)] = {};
+    auto* __dest = reinterpret_cast<_Type*>(__buf);
+
+    __atomic_exchange(__ptr, &__val, __dest, __memorder);
+    return *__dest;
+}
+
+template<class _Type, class _Delta>
+_Type __atomic_fetch_max(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    _Type __expected = __atomic_load_n(__ptr, __ATOMIC_RELAXED);
+    _Type __desired = __expected < __val ? __expected : __val;
+    while(__desired == __val &&
+          !__atomic_compare_exchange_n(__ptr, &__expected, __desired, __memorder, __memorder)) {
+        __desired = __expected > __val ? __expected : __val;
+    }
+    return __expected;
+}
+
+template<class _Type, class _Delta>
+_Type __atomic_fetch_min(_Type volatile *__ptr, _Delta __val, int __memorder) {
+    _Type __expected = __atomic_load_n(__ptr, __ATOMIC_RELAXED);
+    _Type __desired = __expected < __val ? __expected : __val;
+    while(__desired != __val &&
+          !__atomic_compare_exchange_n(__ptr, &__expected, __desired, __memorder, __memorder)) {
+        __desired = __expected < __val ? __expected : __val;
+    }
+    return __expected;
+}
+
+#include "atomic_base.h"
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_nvrtc.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_nvrtc.h
new file mode 100644
index 000000000000..e1e42603cf38
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_nvrtc.h
@@ -0,0 +1,16 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_ATOMIC_NVRTC_H
+#define _LIBCUDACXX_ATOMIC_NVRTC_H
+
+#include "cxx_atomic.h"
+
+#endif // _LIBCUDACXX_ATOMIC_NVRTC_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_scopes.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_scopes.h
new file mode 100644
index 000000000000..89d73bb3f8cf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/atomic_scopes.h
@@ -0,0 +1,46 @@
+#ifndef __LIBCUDACXX_ATOMIC_SCOPES_H
+#define __LIBCUDACXX_ATOMIC_SCOPES_H
+
+#ifndef __ATOMIC_BLOCK
+#define __ATOMIC_SYSTEM 0 // 0 indicates default
+#define __ATOMIC_DEVICE 1
+#define __ATOMIC_BLOCK 2
+#define __ATOMIC_THREAD 10
+#endif //__ATOMIC_BLOCK
+
+enum thread_scope {
+    thread_scope_system = __ATOMIC_SYSTEM,
+    thread_scope_device = __ATOMIC_DEVICE,
+    thread_scope_block = __ATOMIC_BLOCK,
+    thread_scope_thread = __ATOMIC_THREAD
+};
+
+#define _LIBCUDACXX_ATOMIC_SCOPE_TYPE ::cuda::thread_scope
+#define _LIBCUDACXX_ATOMIC_SCOPE_DEFAULT ::cuda::thread_scope::system
+
+struct __thread_scope_thread_tag { };
+struct __thread_scope_block_tag { };
+struct __thread_scope_device_tag { };
+struct __thread_scope_system_tag { };
+
+template<int _Scope>  struct __scope_enum_to_tag { };
+/* This would be the implementation once an actual thread-scope backend exists.
+template<> struct __scope_enum_to_tag<(int)thread_scope_thread> {
+    using type = __thread_scope_thread_tag; };
+Until then: */
+template<> struct __scope_enum_to_tag<(int)thread_scope_thread> {
+    using type = __thread_scope_block_tag; };
+template<> struct __scope_enum_to_tag<(int)thread_scope_block> {
+    using type = __thread_scope_block_tag; };
+template<> struct __scope_enum_to_tag<(int)thread_scope_device> {
+    using type = __thread_scope_device_tag; };
+template<> struct __scope_enum_to_tag<(int)thread_scope_system> {
+    using type = __thread_scope_system_tag; };
+
+template <int _Scope>
+_LIBCUDACXX_INLINE_VISIBILITY auto constexpr __scope_tag() ->
+        typename __scope_enum_to_tag<_Scope>::type {
+    return typename __scope_enum_to_tag<_Scope>::type();
+}
+
+#endif // __LIBCUDACXX_ATOMIC_SCOPES_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/cxx_atomic.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/cxx_atomic.h
new file mode 100644
index 000000000000..fcbc23eb196c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/atomic/cxx_atomic.h
@@ -0,0 +1,160 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_CXX_ATOMIC_H
+#define _LIBCUDACXX_CXX_ATOMIC_H
+
+template <typename _Tp, int _Sco>
+struct __cxx_atomic_base_impl {
+  using __underlying_t = _Tp;
+  using __temporary_t = __cxx_atomic_base_impl<_Tp, _Sco>;
+  using __wrap_t = __cxx_atomic_base_impl<_Tp, _Sco>;
+
+  static constexpr int __sco = _Sco;
+
+#if !defined(_LIBCUDACXX_COMPILER_GCC) || (__GNUC__ >= 5)
+  static_assert(is_trivially_copyable<_Tp>::value,
+    "std::atomic<Tp> requires that 'Tp' be a trivially copyable type");
+#endif
+
+  _LIBCUDACXX_CONSTEXPR
+  __cxx_atomic_base_impl() _NOEXCEPT = default;
+
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit
+  __cxx_atomic_base_impl(_Tp value) _NOEXCEPT : __a_value(value) {}
+
+  _ALIGNAS(sizeof(_Tp)) _Tp __a_value;
+};
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp* __cxx_get_underlying_atomic(__cxx_atomic_base_impl<_Tp, _Sco> * __a) _NOEXCEPT {
+  return &__a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+volatile _Tp* __cxx_get_underlying_atomic(__cxx_atomic_base_impl<_Tp, _Sco> volatile* __a) _NOEXCEPT {
+  return &__a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Tp* __cxx_get_underlying_atomic(__cxx_atomic_base_impl<_Tp, _Sco> const* __a) _NOEXCEPT {
+  return &__a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const volatile _Tp* __cxx_get_underlying_atomic(__cxx_atomic_base_impl<_Tp, _Sco> const volatile* __a) _NOEXCEPT {
+  return &__a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+__cxx_atomic_base_impl<_Tp, _Sco>* __cxx_atomic_unwrap(__cxx_atomic_base_impl<_Tp, _Sco>* __a) _NOEXCEPT {
+  return __a;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+volatile __cxx_atomic_base_impl<_Tp, _Sco>* __cxx_atomic_unwrap(__cxx_atomic_base_impl<_Tp, _Sco> volatile* __a) _NOEXCEPT {
+  return __a;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const __cxx_atomic_base_impl<_Tp, _Sco>* __cxx_atomic_unwrap(__cxx_atomic_base_impl<_Tp, _Sco> const* __a) _NOEXCEPT {
+  return __a;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const volatile __cxx_atomic_base_impl<_Tp, _Sco>* __cxx_atomic_unwrap(__cxx_atomic_base_impl<_Tp, _Sco> const volatile* __a) _NOEXCEPT {
+  return __a;
+}
+
+template <typename _Tp, int _Sco>
+struct __cxx_atomic_ref_base_impl {
+  using __underlying_t = _Tp;
+  using __temporary_t = _Tp;
+  using __wrap_t = _Tp;
+
+  static constexpr int __sco = _Sco;
+
+#if !defined(_LIBCUDACXX_COMPILER_GCC) || (__GNUC__ >= 5)
+  static_assert(is_trivially_copyable<_Tp>::value,
+    "std::atomic_ref<Tp> requires that 'Tp' be a trivially copyable type");
+#endif
+
+  _LIBCUDACXX_CONSTEXPR
+  __cxx_atomic_ref_base_impl() _NOEXCEPT = delete;
+
+  _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR explicit
+  __cxx_atomic_ref_base_impl(_Tp& value) _NOEXCEPT : __a_value(&value) {}
+
+  _Tp* __a_value;
+};
+
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp* __cxx_get_underlying_atomic(__cxx_atomic_ref_base_impl<_Tp, _Sco>* __a) _NOEXCEPT {
+  return __a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+volatile _Tp* __cxx_get_underlying_atomic(__cxx_atomic_ref_base_impl<_Tp, _Sco> volatile* __a) _NOEXCEPT {
+  return __a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Tp* __cxx_get_underlying_atomic(__cxx_atomic_ref_base_impl<_Tp, _Sco> const* __a) _NOEXCEPT {
+  return __a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const volatile _Tp* __cxx_get_underlying_atomic(__cxx_atomic_ref_base_impl<_Tp, _Sco> const volatile* __a) _NOEXCEPT {
+  return __a->__a_value;
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp* __cxx_atomic_unwrap(__cxx_atomic_ref_base_impl<_Tp, _Sco>* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(__a);
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+volatile _Tp* __cxx_atomic_unwrap(__cxx_atomic_ref_base_impl<_Tp, _Sco> volatile* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(__a);
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const _Tp* __cxx_atomic_unwrap(__cxx_atomic_ref_base_impl<_Tp, _Sco> const* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(__a);
+}
+template <typename _Tp, int _Sco>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+const volatile _Tp* __cxx_atomic_unwrap(__cxx_atomic_ref_base_impl<_Tp, _Sco> const volatile* __a) _NOEXCEPT {
+  return __cxx_get_underlying_atomic(__a);
+}
+
+template <typename _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp* __cxx_get_underlying_atomic(_Tp* __a) _NOEXCEPT {
+  return __a;
+}
+
+template <typename _Tp, typename _Up>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+auto __cxx_atomic_wrap_to_base(_Tp*, _Up __val) _NOEXCEPT -> typename _Tp::__wrap_t {
+  return typename _Tp::__wrap_t(__val);
+}
+template <typename _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+auto __cxx_atomic_base_temporary(_Tp*) _NOEXCEPT -> typename _Tp::__temporary_t {
+  return typename _Tp::__temporary_t();
+}
+
+template <typename _Tp>
+using __cxx_atomic_underlying_t = typename _Tp::__underlying_t;
+
+#endif //_LIBCUDACXX_CXX_ATOMIC_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/fuchsia/xlocale.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/fuchsia/xlocale.h
new file mode 100644
index 000000000000..596ddb49bc5f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/fuchsia/xlocale.h
@@ -0,0 +1,22 @@
+// -*- C++ -*-
+//===------------------- support/fuchsia/xlocale.h ------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_FUCHSIA_XLOCALE_H
+#define _LIBCUDACXX_SUPPORT_FUCHSIA_XLOCALE_H
+
+#if defined(__Fuchsia__)
+
+#include <cstdlib>
+#include <cwchar>
+#include <support/xlocale/__posix_l_fallback.h>
+#include <support/xlocale/__strtonum_fallback.h>
+
+#endif // defined(__Fuchsia__)
+
+#endif // _LIBCUDACXX_SUPPORT_FUCHSIA_XLOCALE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/limits.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/limits.h
new file mode 100644
index 000000000000..db2461204ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/limits.h
@@ -0,0 +1,98 @@
+// -*- C++ -*-
+//===--------------------- support/ibm/limits.h ---------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_IBM_LIMITS_H
+#define _LIBCUDACXX_SUPPORT_IBM_LIMITS_H
+
+#if !defined(_AIX) // Linux
+#include <math.h> // for HUGE_VAL, HUGE_VALF, HUGE_VALL, and NAN
+
+static const unsigned int _QNAN_F = 0x7fc00000;
+#define NANF (*((float *)(&_QNAN_F)))
+static const unsigned int _QNAN_LDBL128[4] = {0x7ff80000, 0x0, 0x0, 0x0};
+#define NANL (*((long double *)(&_QNAN_LDBL128)))
+static const unsigned int _SNAN_F= 0x7f855555;
+#define NANSF (*((float *)(&_SNAN_F)))
+static const unsigned int _SNAN_D[2] = {0x7ff55555, 0x55555555};
+#define NANS (*((double *)(&_SNAN_D)))
+static const unsigned int _SNAN_LDBL128[4] = {0x7ff55555, 0x55555555, 0x0, 0x0};
+#define NANSL (*((long double *)(&_SNAN_LDBL128)))
+
+#define __builtin_huge_val()     HUGE_VAL
+#define __builtin_huge_valf()    HUGE_VALF
+#define __builtin_huge_vall()    HUGE_VALL
+#define __builtin_nan(__dummy)   NAN
+#define __builtin_nanf(__dummy)  NANF
+#define __builtin_nanl(__dummy)  NANL
+#define __builtin_nans(__dummy)  NANS
+#define __builtin_nansf(__dummy) NANSF
+#define __builtin_nansl(__dummy) NANSL
+
+#else
+
+#include <math.h>
+#include <float.h> // limit constants
+
+#define __builtin_huge_val()     HUGE_VAL  //0x7ff0000000000000
+#define __builtin_huge_valf()    HUGE_VALF //0x7f800000
+#define __builtin_huge_vall()    HUGE_VALL //0x7ff0000000000000
+#define __builtin_nan(__dummy)   nan(__dummy) //0x7ff8000000000000
+#define __builtin_nanf(__dummy)  nanf(__dummy) // 0x7ff80000
+#define __builtin_nanl(__dummy)  nanl(__dummy) //0x7ff8000000000000
+#define __builtin_nans(__dummy)  DBL_SNAN //0x7ff5555555555555
+#define __builtin_nansf(__dummy) FLT_SNAN //0x7f855555
+#define __builtin_nansl(__dummy) DBL_SNAN //0x7ff5555555555555
+
+#define __FLT_MANT_DIG__   FLT_MANT_DIG
+#define __FLT_DIG__        FLT_DIG
+#define __FLT_RADIX__      FLT_RADIX
+#define __FLT_MIN_EXP__    FLT_MIN_EXP
+#define __FLT_MIN_10_EXP__ FLT_MIN_10_EXP
+#define __FLT_MAX_EXP__    FLT_MAX_EXP
+#define __FLT_MAX_10_EXP__ FLT_MAX_10_EXP
+#define __FLT_MIN__        FLT_MIN
+#define __FLT_MAX__        FLT_MAX
+#define __FLT_EPSILON__    FLT_EPSILON
+// predefined by XLC on LoP
+#define __FLT_DENORM_MIN__ 1.40129846e-45F
+
+#define __DBL_MANT_DIG__   DBL_MANT_DIG
+#define __DBL_DIG__        DBL_DIG
+#define __DBL_MIN_EXP__    DBL_MIN_EXP
+#define __DBL_MIN_10_EXP__ DBL_MIN_10_EXP
+#define __DBL_MAX_EXP__    DBL_MAX_EXP
+#define __DBL_MAX_10_EXP__ DBL_MAX_10_EXP
+#define __DBL_MIN__        DBL_MIN
+#define __DBL_MAX__        DBL_MAX
+#define __DBL_EPSILON__    DBL_EPSILON
+// predefined by XLC on LoP
+#define __DBL_DENORM_MIN__ 4.9406564584124654e-324
+
+#define __LDBL_MANT_DIG__   LDBL_MANT_DIG
+#define __LDBL_DIG__        LDBL_DIG
+#define __LDBL_MIN_EXP__    LDBL_MIN_EXP
+#define __LDBL_MIN_10_EXP__ LDBL_MIN_10_EXP
+#define __LDBL_MAX_EXP__    LDBL_MAX_EXP
+#define __LDBL_MAX_10_EXP__ LDBL_MAX_10_EXP
+#define __LDBL_MIN__        LDBL_MIN
+#define __LDBL_MAX__        LDBL_MAX
+#define __LDBL_EPSILON__    LDBL_EPSILON
+// predefined by XLC on LoP
+#if __LONGDOUBLE128
+#define __LDBL_DENORM_MIN__ 4.94065645841246544176568792868221e-324L
+#else
+#define __LDBL_DENORM_MIN__ 4.9406564584124654e-324L
+#endif
+
+// predefined by XLC on LoP
+#define __CHAR_BIT__    8
+
+#endif // _AIX
+
+#endif // _LIBCUDACXX_SUPPORT_IBM_LIMITS_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/locale_mgmt_aix.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/locale_mgmt_aix.h
new file mode 100644
index 000000000000..1528f4443e45
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/locale_mgmt_aix.h
@@ -0,0 +1,84 @@
+// -*- C++ -*-
+//===------------------- support/ibm/locale_mgmt_aix.h --------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_IBM_LOCALE_MGMT_AIX_H
+#define _LIBCUDACXX_SUPPORT_IBM_LOCALE_MGMT_AIX_H
+
+#if defined(_AIX)
+#include "cstdlib"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if !defined(_AIX71)
+// AIX 7.1 and higher has these definitions.  Definitions and stubs
+// are provied here as a temporary workaround on AIX 6.1.
+
+#define LC_COLLATE_MASK         1
+#define LC_CTYPE_MASK           2
+#define LC_MESSAGES_MASK        4
+#define LC_MONETARY_MASK        8
+#define LC_NUMERIC_MASK         16
+#define LC_TIME_MASK            32
+#define LC_ALL_MASK             (LC_COLLATE_MASK | LC_CTYPE_MASK | \
+                                 LC_MESSAGES_MASK | LC_MONETARY_MASK |\
+                                 LC_NUMERIC_MASK | LC_TIME_MASK)
+
+typedef void* locale_t;
+
+// The following are stubs.  They are not supported on AIX 6.1.
+static inline
+locale_t newlocale(int category_mask, const char *locale, locale_t base)
+{
+  _LC_locale_t *newloc, *loc;
+  if ((loc = (_LC_locale_t *)__xopen_locale(locale)) == NULL)
+  {
+    errno = EINVAL;
+    return (locale_t)0;
+  }
+  if ((newloc = (_LC_locale_t *)calloc(1, sizeof(_LC_locale_t))) == NULL)
+  {
+    errno = ENOMEM;
+    return (locale_t)0;
+  }
+  if (!base)
+    base = (_LC_locale_t *)__xopen_locale("C");
+  memcpy(newloc, base, sizeof (_LC_locale_t));
+  if (category_mask & LC_COLLATE_MASK)
+    newloc->lc_collate = loc->lc_collate;
+  if (category_mask & LC_CTYPE_MASK)
+    newloc->lc_ctype = loc->lc_ctype;
+  //if (category_mask & LC_MESSAGES_MASK)
+  //  newloc->lc_messages = loc->lc_messages;
+  if (category_mask & LC_MONETARY_MASK)
+    newloc->lc_monetary = loc->lc_monetary;
+  if (category_mask & LC_TIME_MASK)
+    newloc->lc_time = loc->lc_time;
+  if (category_mask & LC_NUMERIC_MASK)
+    newloc->lc_numeric = loc->lc_numeric;
+  return (locale_t)newloc;
+}
+static inline
+void freelocale(locale_t locobj)
+{
+  free(locobj);
+}
+static inline
+locale_t uselocale(locale_t newloc)
+{
+  return (locale_t)0;
+}
+#endif // !defined(_AIX71)
+
+#ifdef __cplusplus
+}
+#endif
+#endif // defined(_AIX)
+#endif // _LIBCUDACXX_SUPPORT_IBM_LOCALE_MGMT_AIX_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/support.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/support.h
new file mode 100644
index 000000000000..162611a04f06
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/support.h
@@ -0,0 +1,53 @@
+// -*- C++ -*-
+//===----------------------- support/ibm/support.h ----------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_IBM_SUPPORT_H
+#define _LIBCUDACXX_SUPPORT_IBM_SUPPORT_H
+
+extern "builtin" int __popcnt4(unsigned int);
+extern "builtin" int __popcnt8(unsigned long long);
+extern "builtin" unsigned int __cnttz4(unsigned int);
+extern "builtin" unsigned int __cnttz8(unsigned long long);
+extern "builtin" unsigned int __cntlz4(unsigned int);
+extern "builtin" unsigned int __cntlz8(unsigned long long);
+
+// Builtin functions for counting population
+#define __builtin_popcount(x) __popcnt4(x)
+#define __builtin_popcountll(x) __popcnt8(x)
+#if defined(__64BIT__)
+#define __builtin_popcountl(x) __builtin_popcountll(x)
+#else
+#define __builtin_popcountl(x) __builtin_popcount(x)
+#endif
+
+// Builtin functions for counting trailing zeros
+#define __builtin_ctz(x) __cnttz4(x)
+#define __builtin_ctzll(x) __cnttz8(x)
+#if defined(__64BIT__)
+#define __builtin_ctzl(x) __builtin_ctzll(x)
+#else
+#define __builtin_ctzl(x) __builtin_ctz(x)
+#endif
+
+// Builtin functions for counting leading zeros
+#define __builtin_clz(x) __cntlz4(x)
+#define __builtin_clzll(x) __cntlz8(x)
+#if defined(__64BIT__)
+#define __builtin_clzl(x) __builtin_clzll(x)
+#else
+#define __builtin_clzl(x) __builtin_clz(x)
+#endif
+
+#if defined(__64BIT__)
+#define __SIZE_WIDTH__ 64
+#else
+#define __SIZE_WIDTH__ 32
+#endif
+
+#endif // _LIBCUDACXX_SUPPORT_IBM_SUPPORT_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/xlocale.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/xlocale.h
new file mode 100644
index 000000000000..8f44448b2ea7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/ibm/xlocale.h
@@ -0,0 +1,270 @@
+// -*- C++ -*-
+//===--------------------- support/ibm/xlocale.h -------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_IBM_XLOCALE_H
+#define _LIBCUDACXX_SUPPORT_IBM_XLOCALE_H
+#include <support/ibm/locale_mgmt_aix.h>
+
+#if defined(_AIX)
+#include "cstdlib"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if !defined(_AIX71)
+// AIX 7.1 and higher has these definitions.  Definitions and stubs
+// are provied here as a temporary workaround on AIX 6.1.
+static inline
+int isalnum_l(int c, locale_t locale)
+{
+  return __xisalnum(locale, c);
+}
+static inline
+int isalpha_l(int c, locale_t locale)
+{
+  return __xisalpha(locale, c);
+}
+static inline
+int isblank_l(int c, locale_t locale)
+{
+  return __xisblank(locale, c);
+}
+static inline
+int iscntrl_l(int c, locale_t locale)
+{
+  return __xiscntrl(locale, c);
+}
+static inline
+int isdigit_l(int c, locale_t locale)
+{
+  return __xisdigit(locale, c);
+}
+static inline
+int isgraph_l(int c, locale_t locale)
+{
+  return __xisgraph(locale, c);
+}
+static inline
+int islower_l(int c, locale_t locale)
+{
+  return __xislower(locale, c);
+}
+static inline
+int isprint_l(int c, locale_t locale)
+{
+  return __xisprint(locale, c);
+}
+
+static inline
+int ispunct_l(int c, locale_t locale)
+{
+  return __xispunct(locale, c);
+}
+static inline
+int isspace_l(int c, locale_t locale)
+{
+  return __xisspace(locale, c);
+}
+static inline
+int isupper_l(int c, locale_t locale)
+{
+  return __xisupper(locale, c);
+}
+
+static inline
+int isxdigit_l(int c, locale_t locale)
+{
+  return __xisxdigit(locale, c);
+}
+
+static inline
+int iswalnum_l(wchar_t wc, locale_t locale)
+{
+  return __xiswalnum(locale, wc); 
+}
+
+static inline
+int iswalpha_l(wchar_t wc, locale_t locale)
+{
+  return __xiswalpha(locale, wc);
+}
+
+static inline
+int iswblank_l(wchar_t wc, locale_t locale)
+{
+  return __xiswblank(locale, wc);
+}
+
+static inline
+int iswcntrl_l(wchar_t wc, locale_t locale)
+{
+  return __xiswcntrl(locale, wc);
+}
+
+static inline
+int iswdigit_l(wchar_t wc, locale_t locale)
+{
+  return __xiswdigit(locale, wc);
+}
+
+static inline
+int iswgraph_l(wchar_t wc, locale_t locale)
+{
+  return __xiswgraph(locale, wc);
+}
+
+static inline
+int iswlower_l(wchar_t wc, locale_t locale)
+{
+  return __xiswlower(locale, wc);
+}
+
+static inline
+int iswprint_l(wchar_t wc, locale_t locale)
+{
+  return __xiswprint(locale, wc);
+}
+
+static inline
+int iswpunct_l(wchar_t wc, locale_t locale)
+{
+  return __xiswpunct(locale, wc);
+}
+
+static inline
+int iswspace_l(wchar_t wc, locale_t locale)
+{
+  return __xiswspace(locale, wc);
+}
+
+static inline
+int iswupper_l(wchar_t wc, locale_t locale)
+{
+  return __xiswupper(locale, wc);
+}
+
+static inline
+int iswxdigit_l(wchar_t wc, locale_t locale)
+{
+  return __xiswxdigit(locale, wc);
+}
+
+static inline
+int iswctype_l(wint_t wc, wctype_t desc, locale_t locale)
+{
+  return __xiswctype(locale, wc, desc); 
+}
+
+static inline
+int toupper_l(int c, locale_t locale)
+{
+  return __xtoupper(locale, c);
+}
+static inline
+int tolower_l(int c, locale_t locale)
+{
+  return __xtolower(locale, c);
+}
+static inline
+wint_t towupper_l(wint_t wc, locale_t locale)
+{
+  return __xtowupper(locale, wc);
+}
+static inline
+wint_t towlower_l(wint_t wc, locale_t locale)
+{
+  return __xtowlower(locale, wc);
+}
+
+static inline
+int strcoll_l(const char *__s1, const char *__s2, locale_t locale)
+{
+  return __xstrcoll(locale, __s1, __s2);
+}
+static inline
+int wcscoll_l(const wchar_t *__s1, const wchar_t *__s2, locale_t locale)
+{
+  return __xwcscoll(locale, __s1, __s2);
+}
+static inline
+size_t strxfrm_l(char *__s1, const char *__s2, size_t __n, locale_t locale)
+{
+  return __xstrxfrm(locale, __s1, __s2, __n);
+}
+
+static inline
+size_t wcsxfrm_l(wchar_t *__ws1, const wchar_t *__ws2, size_t __n,
+    locale_t locale)
+{
+  return __xwcsxfrm(locale, __ws1, __ws2, __n);
+}
+#endif // !defined(_AIX71)
+
+// strftime_l() is defined by POSIX. However, AIX 7.1 does not have it
+// implemented yet.
+static inline
+size_t strftime_l(char *__s, size_t __size, const char *__fmt,
+                  const struct tm *__tm, locale_t locale) {
+  return __xstrftime(locale, __s, __size, __fmt, __tm);
+}
+
+// The following are not POSIX routines.  These are quick-and-dirty hacks
+// to make things pretend to work
+static inline
+long long strtoll_l(const char *__nptr, char **__endptr,
+    int __base, locale_t locale) {
+  return strtoll(__nptr, __endptr, __base);
+}
+static inline
+long strtol_l(const char *__nptr, char **__endptr,
+    int __base, locale_t locale) {
+  return strtol(__nptr, __endptr, __base);
+}
+static inline
+long double strtold_l(const char *__nptr, char **__endptr,
+    locale_t locale) {
+  return strtold(__nptr, __endptr);
+}
+static inline
+unsigned long long strtoull_l(const char *__nptr, char **__endptr,
+    int __base, locale_t locale) {
+  return strtoull(__nptr, __endptr, __base);
+}
+static inline
+unsigned long strtoul_l(const char *__nptr, char **__endptr,
+    int __base, locale_t locale) {
+  return strtoul(__nptr, __endptr, __base);
+}
+
+static inline
+int vasprintf(char **strp, const char *fmt, va_list ap)
+{
+  const size_t buff_size = 256;
+  int str_size;
+  if ((*strp = (char *)malloc(buff_size)) == NULL)
+  {
+    return -1;
+  }
+  if ((str_size = vsnprintf(*strp, buff_size, fmt,  ap)) >= buff_size)
+  {
+    if ((*strp = (char *)realloc(*strp, str_size + 1)) == NULL)
+    {
+      return -1;
+    }
+    str_size = vsnprintf(*strp, str_size + 1, fmt,  ap);
+  }
+  return str_size;
+}  
+
+#ifdef __cplusplus
+}
+#endif
+#endif // defined(_AIX)
+#endif // _LIBCUDACXX_SUPPORT_IBM_XLOCALE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/musl/xlocale.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/musl/xlocale.h
new file mode 100644
index 000000000000..1b7a6883bce6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/musl/xlocale.h
@@ -0,0 +1,57 @@
+// -*- C++ -*-
+//===------------------- support/musl/xlocale.h ------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// This adds support for the extended locale functions that are currently
+// missing from the Musl C library.
+//
+// This only works when the specified locale is "C" or "POSIX", but that's
+// about as good as we can do without implementing full xlocale support
+// in Musl.
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_MUSL_XLOCALE_H
+#define _LIBCUDACXX_SUPPORT_MUSL_XLOCALE_H
+
+#include <cstdlib>
+#include <cwchar>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+static inline long long strtoll_l(const char *nptr, char **endptr, int base,
+                                  locale_t) {
+  return strtoll(nptr, endptr, base);
+}
+
+static inline unsigned long long strtoull_l(const char *nptr, char **endptr,
+                                            int base, locale_t) {
+  return strtoull(nptr, endptr, base);
+}
+
+static inline long long wcstoll_l(const wchar_t *nptr, wchar_t **endptr,
+                                  int base, locale_t) {
+  return wcstoll(nptr, endptr, base);
+}
+
+static inline unsigned long long wcstoull_l(const wchar_t *nptr,
+                                            wchar_t **endptr, int base,
+                                            locale_t) {
+  return wcstoull(nptr, endptr, base);
+}
+
+static inline long double wcstold_l(const wchar_t *nptr, wchar_t **endptr,
+                                    locale_t) {
+  return wcstold(nptr, endptr);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // _LIBCUDACXX_SUPPORT_MUSL_XLOCALE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/newlib/xlocale.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/newlib/xlocale.h
new file mode 100644
index 000000000000..8be7c6b43c4f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/newlib/xlocale.h
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_NEWLIB_XLOCALE_H
+#define _LIBCUDACXX_SUPPORT_NEWLIB_XLOCALE_H
+
+#if defined(_NEWLIB_VERSION)
+
+#include <cstdlib>
+#include <clocale>
+#include <cwctype>
+#include <ctype.h>
+#if !defined(__NEWLIB__) || __NEWLIB__ < 2 || \
+    __NEWLIB__ == 2 && __NEWLIB_MINOR__ < 5
+#include <support/xlocale/__nop_locale_mgmt.h>
+#include <support/xlocale/__posix_l_fallback.h>
+#include <support/xlocale/__strtonum_fallback.h>
+#endif
+
+#endif // _NEWLIB_VERSION
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/floatingpoint.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/floatingpoint.h
new file mode 100644
index 000000000000..5f1628fbe4f5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/floatingpoint.h
@@ -0,0 +1,13 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#define atof sun_atof
+#define strtod sun_strtod
+#include_next "floatingpoint.h"
+#undef atof
+#undef strtod
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/wchar.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/wchar.h
new file mode 100644
index 000000000000..9dc9ac3f006c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/wchar.h
@@ -0,0 +1,46 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#define iswalpha sun_iswalpha
+#define iswupper sun_iswupper
+#define iswlower sun_iswlower
+#define iswdigit sun_iswdigit
+#define iswxdigit sun_iswxdigit
+#define iswalnum sun_iswalnum
+#define iswspace sun_iswspace
+#define iswpunct sun_iswpunct
+#define iswprint sun_iswprint
+#define iswgraph sun_iswgraph
+#define iswcntrl sun_iswcntrl
+#define iswctype sun_iswctype
+#define towlower sun_towlower
+#define towupper sun_towupper
+#define wcswcs sun_wcswcs
+#define wcswidth sun_wcswidth
+#define wcwidth sun_wcwidth
+#define wctype sun_wctype
+#define _WCHAR_T 1
+#include_next "wchar.h"
+#undef iswalpha 
+#undef iswupper
+#undef iswlower
+#undef iswdigit
+#undef iswxdigit
+#undef iswalnum
+#undef iswspace
+#undef iswpunct
+#undef iswprint
+#undef iswgraph
+#undef iswcntrl
+#undef iswctype
+#undef towlower
+#undef towupper
+#undef wcswcs
+#undef wcswidth
+#undef wcwidth
+#undef wctype
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/xlocale.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/xlocale.h
new file mode 100644
index 000000000000..05131f0272ef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/solaris/xlocale.h
@@ -0,0 +1,76 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+////////////////////////////////////////////////////////////////////////////////
+// Minimal xlocale implementation for Solaris.  This implements the subset of
+// the xlocale APIs that libc++ depends on.
+////////////////////////////////////////////////////////////////////////////////
+#ifndef __XLOCALE_H_INCLUDED
+#define __XLOCALE_H_INCLUDED
+
+#include <stdlib.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+int snprintf_l(char *__s, size_t __n, locale_t __l, const char *__format, ...);
+int asprintf_l(char **__s, locale_t __l, const char *__format, ...);
+
+int sscanf_l(const char *__s, locale_t __l, const char *__format, ...);
+
+int toupper_l(int __c, locale_t __l);
+int tolower_l(int __c, locale_t __l);
+
+struct lconv *localeconv(void);
+struct lconv *localeconv_l(locale_t __l);
+
+// FIXME: These are quick-and-dirty hacks to make things pretend to work
+static inline
+long long strtoll_l(const char *__nptr, char **__endptr,
+    int __base, locale_t __loc) {
+  return strtoll(__nptr, __endptr, __base);
+}
+static inline
+long strtol_l(const char *__nptr, char **__endptr,
+    int __base, locale_t __loc) {
+  return strtol(__nptr, __endptr, __base);
+}
+static inline
+unsigned long long strtoull_l(const char *__nptr, char **__endptr,
+    int __base, locale_t __loc) {
+  return strtoull(__nptr, __endptr, __base);
+}
+static inline
+unsigned long strtoul_l(const char *__nptr, char **__endptr,
+    int __base, locale_t __loc) {
+  return strtoul(__nptr, __endptr, __base);
+}
+static inline
+float strtof_l(const char *__nptr, char **__endptr,
+    locale_t __loc) {
+  return strtof(__nptr, __endptr);
+}
+static inline
+double strtod_l(const char *__nptr, char **__endptr,
+    locale_t __loc) {
+  return strtod(__nptr, __endptr);
+}
+static inline
+long double strtold_l(const char *__nptr, char **__endptr,
+    locale_t __loc) {
+  return strtold(__nptr, __endptr);
+}
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__nop_locale_mgmt.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__nop_locale_mgmt.h
new file mode 100644
index 000000000000..12502e8f735b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__nop_locale_mgmt.h
@@ -0,0 +1,51 @@
+// -*- C++ -*-
+//===------------  support/xlocale/__nop_locale_mgmt.h -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_XLOCALE_NOP_LOCALE_MGMT_H
+#define _LIBCUDACXX_SUPPORT_XLOCALE_NOP_LOCALE_MGMT_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Patch over lack of extended locale support
+typedef void *locale_t;
+static inline locale_t duplocale(locale_t) {
+  return NULL;
+}
+
+static inline void freelocale(locale_t) {
+}
+
+static inline locale_t newlocale(int, const char *, locale_t) {
+  return NULL;
+}
+
+static inline locale_t uselocale(locale_t) {
+  return NULL;
+}
+
+#define LC_COLLATE_MASK  (1 << LC_COLLATE)
+#define LC_CTYPE_MASK    (1 << LC_CTYPE)
+#define LC_MESSAGES_MASK (1 << LC_MESSAGES)
+#define LC_MONETARY_MASK (1 << LC_MONETARY)
+#define LC_NUMERIC_MASK  (1 << LC_NUMERIC)
+#define LC_TIME_MASK     (1 << LC_TIME)
+#define LC_ALL_MASK (LC_COLLATE_MASK|\
+                     LC_CTYPE_MASK|\
+                     LC_MONETARY_MASK|\
+                     LC_NUMERIC_MASK|\
+                     LC_TIME_MASK|\
+                     LC_MESSAGES_MASK)
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#endif // _LIBCUDACXX_SUPPORT_XLOCALE_NOP_LOCALE_MGMT_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__posix_l_fallback.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__posix_l_fallback.h
new file mode 100644
index 000000000000..9aff01bc278e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__posix_l_fallback.h
@@ -0,0 +1,164 @@
+// -*- C++ -*-
+//===--------------- support/xlocale/__posix_l_fallback.h -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// These are reimplementations of some extended locale functions ( *_l ) that
+// are normally part of POSIX.  This shared implementation provides parts of the
+// extended locale support for libc's that normally don't have any (like
+// Android's bionic and Newlib).
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_XLOCALE_POSIX_L_FALLBACK_H
+#define _LIBCUDACXX_SUPPORT_XLOCALE_POSIX_L_FALLBACK_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isalnum_l(int c, locale_t) {
+  return ::isalnum(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isalpha_l(int c, locale_t) {
+  return ::isalpha(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isblank_l(int c, locale_t) {
+  return ::isblank(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iscntrl_l(int c, locale_t) {
+  return ::iscntrl(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isdigit_l(int c, locale_t) {
+  return ::isdigit(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isgraph_l(int c, locale_t) {
+  return ::isgraph(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int islower_l(int c, locale_t) {
+  return ::islower(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isprint_l(int c, locale_t) {
+  return ::isprint(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int ispunct_l(int c, locale_t) {
+  return ::ispunct(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isspace_l(int c, locale_t) {
+  return ::isspace(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isupper_l(int c, locale_t) {
+  return ::isupper(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int isxdigit_l(int c, locale_t) {
+  return ::isxdigit(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswalnum_l(wint_t c, locale_t) {
+  return ::iswalnum(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswalpha_l(wint_t c, locale_t) {
+  return ::iswalpha(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswblank_l(wint_t c, locale_t) {
+  return ::iswblank(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswcntrl_l(wint_t c, locale_t) {
+  return ::iswcntrl(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswdigit_l(wint_t c, locale_t) {
+  return ::iswdigit(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswgraph_l(wint_t c, locale_t) {
+  return ::iswgraph(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswlower_l(wint_t c, locale_t) {
+  return ::iswlower(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswprint_l(wint_t c, locale_t) {
+  return ::iswprint(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswpunct_l(wint_t c, locale_t) {
+  return ::iswpunct(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswspace_l(wint_t c, locale_t) {
+  return ::iswspace(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswupper_l(wint_t c, locale_t) {
+  return ::iswupper(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int iswxdigit_l(wint_t c, locale_t) {
+  return ::iswxdigit(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int toupper_l(int c, locale_t) {
+  return ::toupper(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int tolower_l(int c, locale_t) {
+  return ::tolower(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY wint_t towupper_l(wint_t c, locale_t) {
+  return ::towupper(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY wint_t towlower_l(wint_t c, locale_t) {
+  return ::towlower(c);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int strcoll_l(const char *s1, const char *s2,
+                                               locale_t) {
+  return ::strcoll(s1, s2);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY size_t strxfrm_l(char *dest, const char *src,
+                                                  size_t n, locale_t) {
+  return ::strxfrm(dest, src, n);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY size_t strftime_l(char *s, size_t max,
+                                                   const char *format,
+                                                   const struct tm *tm, locale_t) {
+  return ::strftime(s, max, format, tm);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY int wcscoll_l(const wchar_t *ws1,
+                                               const wchar_t *ws2, locale_t) {
+  return ::wcscoll(ws1, ws2);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY size_t wcsxfrm_l(wchar_t *dest, const wchar_t *src,
+                                                  size_t n, locale_t) {
+  return ::wcsxfrm(dest, src, n);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // _LIBCUDACXX_SUPPORT_XLOCALE_POSIX_L_FALLBACK_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__strtonum_fallback.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__strtonum_fallback.h
new file mode 100644
index 000000000000..18f77d600aa1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/support/xlocale/__strtonum_fallback.h
@@ -0,0 +1,66 @@
+// -*- C++ -*-
+//===-------------- support/xlocale/__strtonum_fallback.h -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// These are reimplementations of some extended locale functions ( *_l ) that
+// aren't part of POSIX.  They are widely available though (GLIBC, BSD, maybe
+// others).  The unifying aspect in this case is that all of these functions
+// convert strings to some numeric type.
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SUPPORT_XLOCALE_STRTONUM_FALLBACK_H
+#define _LIBCUDACXX_SUPPORT_XLOCALE_STRTONUM_FALLBACK_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+inline _LIBCUDACXX_INLINE_VISIBILITY float strtof_l(const char *nptr,
+                                                char **endptr, locale_t) {
+  return ::strtof(nptr, endptr);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY double strtod_l(const char *nptr,
+                                                 char **endptr, locale_t) {
+  return ::strtod(nptr, endptr);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long double strtold_l(const char *nptr,
+                                                       char **endptr, locale_t) {
+  return ::strtold(nptr, endptr);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long long
+strtoll_l(const char *nptr, char **endptr, int base, locale_t) {
+  return ::strtoll(nptr, endptr, base);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY unsigned long long
+strtoull_l(const char *nptr, char **endptr, int base, locale_t) {
+  return ::strtoull(nptr, endptr, base);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long long
+wcstoll_l(const wchar_t *nptr, wchar_t **endptr, int base, locale_t) {
+  return ::wcstoll(nptr, endptr, base);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY unsigned long long
+wcstoull_l(const wchar_t *nptr, wchar_t **endptr, int base, locale_t) {
+  return ::wcstoull(nptr, endptr, base);
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY long double wcstold_l(const wchar_t *nptr,
+                                                       wchar_t **endptr, locale_t) {
+  return ::wcstold(nptr, endptr);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // _LIBCUDACXX_SUPPORT_XLOCALE_STRTONUM_FALLBACK_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/system_error b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/system_error
new file mode 100644
index 000000000000..d224a078f95b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/system_error
@@ -0,0 +1,486 @@
+// -*- C++ -*-
+//===---------------------------- system_error ----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_SYSTEM_ERROR
+#define _LIBCUDACXX_SYSTEM_ERROR
+
+/*
+    system_error synopsis
+
+namespace std
+{
+
+class error_category
+{
+public:
+    virtual ~error_category() noexcept;
+
+    constexpr error_category();
+    error_category(const error_category&) = delete;
+    error_category& operator=(const error_category&) = delete;
+
+    virtual const char* name() const noexcept = 0;
+    virtual error_condition default_error_condition(int ev) const noexcept;
+    virtual bool equivalent(int code, const error_condition& condition) const noexcept;
+    virtual bool equivalent(const error_code& code, int condition) const noexcept;
+    virtual string message(int ev) const = 0;
+
+    bool operator==(const error_category& rhs) const noexcept;
+    bool operator!=(const error_category& rhs) const noexcept;
+    bool operator<(const error_category& rhs) const noexcept;
+};
+
+const error_category& generic_category() noexcept;
+const error_category& system_category() noexcept;
+
+template <class T> struct is_error_code_enum
+    : public false_type {};
+
+template <class T> struct is_error_condition_enum
+    : public false_type {};
+
+template <class _Tp>
+inline constexpr size_t is_error_condition_enum_v = is_error_condition_enum<_Tp>::value; // C++17
+
+template <class _Tp>
+inline constexpr size_t is_error_code_enum_v = is_error_code_enum<_Tp>::value; // C++17
+
+class error_code
+{
+public:
+    // constructors:
+    error_code() noexcept;
+    error_code(int val, const error_category& cat) noexcept;
+    template <class ErrorCodeEnum>
+        error_code(ErrorCodeEnum e) noexcept;
+
+    // modifiers:
+    void assign(int val, const error_category& cat) noexcept;
+    template <class ErrorCodeEnum>
+        error_code& operator=(ErrorCodeEnum e) noexcept;
+    void clear() noexcept;
+
+    // observers:
+    int value() const noexcept;
+    const error_category& category() const noexcept;
+    error_condition default_error_condition() const noexcept;
+    string message() const;
+    explicit operator bool() const noexcept;
+};
+
+// non-member functions:
+bool operator<(const error_code& lhs, const error_code& rhs) noexcept;
+template <class charT, class traits>
+    basic_ostream<charT,traits>&
+    operator<<(basic_ostream<charT,traits>& os, const error_code& ec);
+
+class error_condition
+{
+public:
+    // constructors:
+    error_condition() noexcept;
+    error_condition(int val, const error_category& cat) noexcept;
+    template <class ErrorConditionEnum>
+        error_condition(ErrorConditionEnum e) noexcept;
+
+    // modifiers:
+    void assign(int val, const error_category& cat) noexcept;
+    template <class ErrorConditionEnum>
+        error_condition& operator=(ErrorConditionEnum e) noexcept;
+    void clear() noexcept;
+
+    // observers:
+    int value() const noexcept;
+    const error_category& category() const noexcept;
+    string message() const noexcept;
+    explicit operator bool() const noexcept;
+};
+
+bool operator<(const error_condition& lhs, const error_condition& rhs) noexcept;
+
+class system_error
+    : public runtime_error
+{
+public:
+    system_error(error_code ec, const string& what_arg);
+    system_error(error_code ec, const char* what_arg);
+    system_error(error_code ec);
+    system_error(int ev, const error_category& ecat, const string& what_arg);
+    system_error(int ev, const error_category& ecat, const char* what_arg);
+    system_error(int ev, const error_category& ecat);
+
+    const error_code& code() const noexcept;
+    const char* what() const noexcept;
+};
+
+template <> struct is_error_condition_enum<errc>
+    : true_type { }
+
+error_code make_error_code(errc e) noexcept;
+error_condition make_error_condition(errc e) noexcept;
+
+// Comparison operators:
+bool operator==(const error_code& lhs, const error_code& rhs) noexcept;
+bool operator==(const error_code& lhs, const error_condition& rhs) noexcept;
+bool operator==(const error_condition& lhs, const error_code& rhs) noexcept;
+bool operator==(const error_condition& lhs, const error_condition& rhs) noexcept;
+bool operator!=(const error_code& lhs, const error_code& rhs) noexcept;
+bool operator!=(const error_code& lhs, const error_condition& rhs) noexcept;
+bool operator!=(const error_condition& lhs, const error_code& rhs) noexcept;
+bool operator!=(const error_condition& lhs, const error_condition& rhs) noexcept;
+
+template <> struct hash<std::error_code>;
+template <> struct hash<std::error_condition>;
+
+}  // std
+
+*/
+
+#include <__errc>
+#include <type_traits>
+#include <stdexcept>
+#include <__functional_base>
+#include <string>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// is_error_code_enum
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_code_enum
+    : public false_type {};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t is_error_code_enum_v = is_error_code_enum<_Tp>::value;
+#endif
+
+// is_error_condition_enum
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_condition_enum
+    : public false_type {};
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t is_error_condition_enum_v = is_error_condition_enum<_Tp>::value;
+#endif
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_condition_enum<errc>
+    : true_type { };
+
+#ifdef _LIBCUDACXX_HAS_NO_STRONG_ENUMS
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS is_error_condition_enum<errc::__lx>
+    : true_type { };
+#endif
+
+class _LIBCUDACXX_TYPE_VIS error_condition;
+class _LIBCUDACXX_TYPE_VIS error_code;
+
+// class error_category
+
+class _LIBCUDACXX_HIDDEN __do_message;
+
+class _LIBCUDACXX_TYPE_VIS error_category
+{
+public:
+    virtual ~error_category() _NOEXCEPT;
+
+#if defined(_LIBCUDACXX_BUILDING_LIBRARY) && \
+    defined(_LIBCUDACXX_DEPRECATED_ABI_LEGACY_LIBRARY_DEFINITIONS_FOR_INLINE_FUNCTIONS)
+    error_category() _NOEXCEPT;
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 error_category() _NOEXCEPT _LIBCUDACXX_DEFAULT
+#endif
+private:
+    error_category(const error_category&);// = delete;
+    error_category& operator=(const error_category&);// = delete;
+
+public:
+    virtual const char* name() const _NOEXCEPT = 0;
+    virtual error_condition default_error_condition(int __ev) const _NOEXCEPT;
+    virtual bool equivalent(int __code, const error_condition& __condition) const _NOEXCEPT;
+    virtual bool equivalent(const error_code& __code, int __condition) const _NOEXCEPT;
+    virtual string message(int __ev) const = 0;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const error_category& __rhs) const _NOEXCEPT {return this == &__rhs;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const error_category& __rhs) const _NOEXCEPT {return !(*this == __rhs);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator< (const error_category& __rhs) const _NOEXCEPT {return this < &__rhs;}
+
+    friend class _LIBCUDACXX_HIDDEN __do_message;
+};
+
+class _LIBCUDACXX_HIDDEN __do_message
+    : public error_category
+{
+public:
+    virtual string message(int ev) const;
+};
+
+_LIBCUDACXX_FUNC_VIS const error_category& generic_category() _NOEXCEPT;
+_LIBCUDACXX_FUNC_VIS const error_category& system_category() _NOEXCEPT;
+
+class _LIBCUDACXX_TYPE_VIS error_condition
+{
+    int __val_;
+    const error_category* __cat_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    error_condition() _NOEXCEPT : __val_(0), __cat_(&generic_category()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    error_condition(int __val, const error_category& __cat) _NOEXCEPT
+        : __val_(__val), __cat_(&__cat) {}
+
+    template <class _Ep>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        error_condition(_Ep __e,
+              typename enable_if<is_error_condition_enum<_Ep>::value>::type* = 0
+                                                                     ) _NOEXCEPT
+            {*this = make_error_condition(__e);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(int __val, const error_category& __cat) _NOEXCEPT
+    {
+        __val_ = __val;
+        __cat_ = &__cat;
+    }
+
+    template <class _Ep>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            is_error_condition_enum<_Ep>::value,
+            error_condition&
+        >::type
+        operator=(_Ep __e) _NOEXCEPT
+            {*this = make_error_condition(__e); return *this;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT
+    {
+        __val_ = 0;
+        __cat_ = &generic_category();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int value() const _NOEXCEPT {return __val_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const error_category& category() const _NOEXCEPT {return *__cat_;}
+    string message() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+        _LIBCUDACXX_EXPLICIT
+        operator bool() const _NOEXCEPT {return __val_ != 0;}
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+error_condition
+make_error_condition(errc __e) _NOEXCEPT
+{
+    return error_condition(static_cast<int>(__e), generic_category());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const error_condition& __x, const error_condition& __y) _NOEXCEPT
+{
+    return __x.category() < __y.category()
+        || (__x.category() == __y.category() && __x.value() < __y.value());
+}
+
+// error_code
+
+class _LIBCUDACXX_TYPE_VIS error_code
+{
+    int __val_;
+    const error_category* __cat_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    error_code() _NOEXCEPT : __val_(0), __cat_(&system_category()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    error_code(int __val, const error_category& __cat) _NOEXCEPT
+        : __val_(__val), __cat_(&__cat) {}
+
+    template <class _Ep>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        error_code(_Ep __e,
+                   typename enable_if<is_error_code_enum<_Ep>::value>::type* = 0
+                                                                     ) _NOEXCEPT
+            {*this = make_error_code(__e);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(int __val, const error_category& __cat) _NOEXCEPT
+    {
+        __val_ = __val;
+        __cat_ = &__cat;
+    }
+
+    template <class _Ep>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            is_error_code_enum<_Ep>::value,
+            error_code&
+        >::type
+        operator=(_Ep __e) _NOEXCEPT
+            {*this = make_error_code(__e); return *this;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT
+    {
+        __val_ = 0;
+        __cat_ = &system_category();
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int value() const _NOEXCEPT {return __val_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const error_category& category() const _NOEXCEPT {return *__cat_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    error_condition default_error_condition() const _NOEXCEPT
+        {return __cat_->default_error_condition(__val_);}
+
+    string message() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+        _LIBCUDACXX_EXPLICIT
+        operator bool() const _NOEXCEPT {return __val_ != 0;}
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+error_code
+make_error_code(errc __e) _NOEXCEPT
+{
+    return error_code(static_cast<int>(__e), generic_category());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<(const error_code& __x, const error_code& __y) _NOEXCEPT
+{
+    return __x.category() < __y.category()
+        || (__x.category() == __y.category() && __x.value() < __y.value());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const error_code& __x, const error_code& __y) _NOEXCEPT
+{
+    return __x.category() == __y.category() && __x.value() == __y.value();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const error_code& __x, const error_condition& __y) _NOEXCEPT
+{
+    return __x.category().equivalent(__x.value(), __y)
+        || __y.category().equivalent(__x, __y.value());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const error_condition& __x, const error_code& __y) _NOEXCEPT
+{
+    return __y == __x;
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const error_condition& __x, const error_condition& __y) _NOEXCEPT
+{
+    return __x.category() == __y.category() && __x.value() == __y.value();
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const error_code& __x, const error_code& __y) _NOEXCEPT
+{return !(__x == __y);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const error_code& __x, const error_condition& __y) _NOEXCEPT
+{return !(__x == __y);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const error_condition& __x, const error_code& __y) _NOEXCEPT
+{return !(__x == __y);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const error_condition& __x, const error_condition& __y) _NOEXCEPT
+{return !(__x == __y);}
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<error_code>
+    : public unary_function<error_code, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const error_code& __ec) const _NOEXCEPT
+    {
+        return static_cast<size_t>(__ec.value());
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<error_condition>
+    : public unary_function<error_condition, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const error_condition& __ec) const _NOEXCEPT
+    {
+        return static_cast<size_t>(__ec.value());
+    }
+};
+
+// system_error
+
+class _LIBCUDACXX_TYPE_VIS system_error
+    : public runtime_error
+{
+    error_code __ec_;
+public:
+    system_error(error_code __ec, const string& __what_arg);
+    system_error(error_code __ec, const char* __what_arg);
+    system_error(error_code __ec);
+    system_error(int __ev, const error_category& __ecat, const string& __what_arg);
+    system_error(int __ev, const error_category& __ecat, const char* __what_arg);
+    system_error(int __ev, const error_category& __ecat);
+    ~system_error() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const error_code& code() const _NOEXCEPT {return __ec_;}
+
+private:
+    static string __init(const error_code&, string);
+};
+
+_LIBCUDACXX_NORETURN _LIBCUDACXX_FUNC_VIS
+void __throw_system_error(int ev, const char* what_arg);
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_SYSTEM_ERROR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h
new file mode 100644
index 000000000000..a2f03bc906ec
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tgmath.h
@@ -0,0 +1,36 @@
+// -*- C++ -*-
+//===-------------------------- tgmath.h ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_TGMATH_H
+#define _LIBCUDACXX_TGMATH_H
+
+/*
+    tgmath.h synopsis
+
+#include <ctgmath>
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef __cplusplus
+
+#include <ctgmath>
+
+#else  // __cplusplus
+
+#include_next <tgmath.h>
+
+#endif  // __cplusplus
+
+#endif  // _LIBCUDACXX_TGMATH_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/thread b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/thread
new file mode 100644
index 000000000000..82696ba44c35
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/thread
@@ -0,0 +1,420 @@
+// -*- C++ -*-
+//===--------------------------- thread -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_THREAD
+#define _LIBCUDACXX_THREAD
+
+/*
+
+    thread synopsis
+
+namespace std
+{
+
+class thread
+{
+public:
+    class id;
+    typedef pthread_t native_handle_type;
+
+    thread() noexcept;
+    template <class F, class ...Args> explicit thread(F&& f, Args&&... args);
+    ~thread();
+
+    thread(const thread&) = delete;
+    thread(thread&& t) noexcept;
+
+    thread& operator=(const thread&) = delete;
+    thread& operator=(thread&& t) noexcept;
+
+    void swap(thread& t) noexcept;
+
+    bool joinable() const noexcept;
+    void join();
+    void detach();
+    id get_id() const noexcept;
+    native_handle_type native_handle();
+
+    static unsigned hardware_concurrency() noexcept;
+};
+
+void swap(thread& x, thread& y) noexcept;
+
+class thread::id
+{
+public:
+    id() noexcept;
+};
+
+bool operator==(thread::id x, thread::id y) noexcept;
+bool operator!=(thread::id x, thread::id y) noexcept;
+bool operator< (thread::id x, thread::id y) noexcept;
+bool operator<=(thread::id x, thread::id y) noexcept;
+bool operator> (thread::id x, thread::id y) noexcept;
+bool operator>=(thread::id x, thread::id y) noexcept;
+
+template<class charT, class traits>
+basic_ostream<charT, traits>&
+operator<<(basic_ostream<charT, traits>& out, thread::id id);
+
+namespace this_thread
+{
+
+thread::id get_id() noexcept;
+
+void yield() noexcept;
+
+template <class Clock, class Duration>
+void sleep_until(const chrono::time_point<Clock, Duration>& abs_time);
+
+template <class Rep, class Period>
+void sleep_for(const chrono::duration<Rep, Period>& rel_time);
+
+}  // this_thread
+
+}  // std
+
+*/
+
+#include <__config>
+#include <iosfwd>
+#include <__functional_base>
+#include <type_traits>
+#include <cstddef>
+#include <functional>
+#include <memory>
+#include <system_error>
+#include <chrono>
+#include <__mutex_base>
+#ifndef _LIBCUDACXX_CXX03_LANG
+#include <tuple>
+#endif
+#include <__threading_support>
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+#ifdef _LIBCUDACXX_HAS_NO_THREADS
+#error <thread> is not supported on this single threaded system
+#else // !_LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Tp> class __thread_specific_ptr;
+class _LIBCUDACXX_TYPE_VIS __thread_struct;
+class _LIBCUDACXX_HIDDEN __thread_struct_imp;
+class __assoc_sub_state;
+
+_LIBCUDACXX_FUNC_VIS __thread_specific_ptr<__thread_struct>& __thread_local_data();
+
+class _LIBCUDACXX_TYPE_VIS __thread_struct
+{
+    __thread_struct_imp* __p_;
+
+    __thread_struct(const __thread_struct&);
+    __thread_struct& operator=(const __thread_struct&);
+public:
+    __thread_struct();
+    ~__thread_struct();
+
+    void notify_all_at_thread_exit(condition_variable*, mutex*);
+    void __make_ready_at_thread_exit(__assoc_sub_state*);
+};
+
+template <class _Tp>
+class __thread_specific_ptr
+{
+    __libcpp_tls_key __key_;
+
+     // Only __thread_local_data() may construct a __thread_specific_ptr
+     // and only with _Tp == __thread_struct.
+    static_assert((is_same<_Tp, __thread_struct>::value), "");
+    __thread_specific_ptr();
+    friend _LIBCUDACXX_FUNC_VIS __thread_specific_ptr<__thread_struct>& __thread_local_data();
+
+    __thread_specific_ptr(const __thread_specific_ptr&);
+    __thread_specific_ptr& operator=(const __thread_specific_ptr&);
+
+    _LIBCUDACXX_HIDDEN static void _LIBCUDACXX_TLS_DESTRUCTOR_CC __at_thread_exit(void*);
+
+public:
+    typedef _Tp* pointer;
+
+    ~__thread_specific_ptr();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer get() const {return static_cast<_Tp*>(__libcpp_tls_get(__key_));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator*() const {return *get();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return get();}
+    void set_pointer(pointer __p);
+};
+
+template <class _Tp>
+void _LIBCUDACXX_TLS_DESTRUCTOR_CC
+__thread_specific_ptr<_Tp>::__at_thread_exit(void* __p)
+{
+    delete static_cast<pointer>(__p);
+}
+
+template <class _Tp>
+__thread_specific_ptr<_Tp>::__thread_specific_ptr()
+{
+  int __ec =
+      __libcpp_tls_create(&__key_, &__thread_specific_ptr::__at_thread_exit);
+  if (__ec)
+    __throw_system_error(__ec, "__thread_specific_ptr construction failed");
+}
+
+template <class _Tp>
+__thread_specific_ptr<_Tp>::~__thread_specific_ptr()
+{
+    // __thread_specific_ptr is only created with a static storage duration
+    // so this destructor is only invoked during program termination. Invoking
+    // pthread_key_delete(__key_) may prevent other threads from deleting their
+    // thread local data. For this reason we leak the key.
+}
+
+template <class _Tp>
+void
+__thread_specific_ptr<_Tp>::set_pointer(pointer __p)
+{
+    _LIBCUDACXX_ASSERT(get() == nullptr,
+                   "Attempting to overwrite thread local data");
+    __libcpp_tls_set(__key_, __p);
+}
+
+template<>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<__thread_id>
+    : public unary_function<__thread_id, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(__thread_id __v) const _NOEXCEPT
+    {
+        return hash<__libcpp_thread_id>()(__v.__id_);
+    }
+};
+
+template<class _CharT, class _Traits>
+_LIBCUDACXX_INLINE_VISIBILITY
+basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, __thread_id __id)
+{return __os << __id.__id_;}
+
+class _LIBCUDACXX_TYPE_VIS thread
+{
+    __libcpp_thread_t __t_;
+
+    thread(const thread&);
+    thread& operator=(const thread&);
+public:
+    typedef __thread_id id;
+    typedef __libcpp_thread_t native_handle_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    thread() _NOEXCEPT : __t_(_LIBCUDACXX_NULL_THREAD) {}
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Fp, class ..._Args,
+              class = typename enable_if
+              <
+                   !is_same<typename __uncvref<_Fp>::type, thread>::value
+              >::type
+             >
+        _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+        explicit thread(_Fp&& __f, _Args&&... __args);
+#else  // _LIBCUDACXX_CXX03_LANG
+    template <class _Fp>
+    _LIBCUDACXX_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS
+    explicit thread(_Fp __f);
+#endif
+    ~thread();
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    thread(thread&& __t) _NOEXCEPT : __t_(__t.__t_) {__t.__t_ = _LIBCUDACXX_NULL_THREAD;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    thread& operator=(thread&& __t) _NOEXCEPT;
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(thread& __t) _NOEXCEPT {_CUDA_VSTD::swap(__t_, __t.__t_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool joinable() const _NOEXCEPT {return !__libcpp_thread_isnull(&__t_);}
+    void join();
+    void detach();
+    _LIBCUDACXX_INLINE_VISIBILITY
+    id get_id() const _NOEXCEPT {return __libcpp_thread_get_id(&__t_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    native_handle_type native_handle() _NOEXCEPT {return __t_;}
+
+    static unsigned hardware_concurrency() _NOEXCEPT;
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _TSp, class _Fp, class ..._Args, size_t ..._Indices>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__thread_execute(tuple<_TSp, _Fp, _Args...>& __t, __tuple_indices<_Indices...>)
+{
+    __invoke(_CUDA_VSTD::move(_CUDA_VSTD::get<1>(__t)), _CUDA_VSTD::move(_CUDA_VSTD::get<_Indices>(__t))...);
+}
+
+template <class _Fp>
+void* __thread_proxy(void* __vp)
+{
+    // _Fp = std::tuple< unique_ptr<__thread_struct>, Functor, Args...>
+    std::unique_ptr<_Fp> __p(static_cast<_Fp*>(__vp));
+    __thread_local_data().set_pointer(_CUDA_VSTD::get<0>(*__p).release());
+    typedef typename __make_tuple_indices<tuple_size<_Fp>::value, 2>::type _Index;
+    __thread_execute(*__p, _Index());
+    return nullptr;
+}
+
+template <class _Fp, class ..._Args,
+          class
+         >
+thread::thread(_Fp&& __f, _Args&&... __args)
+{
+    typedef unique_ptr<__thread_struct> _TSPtr;
+    _TSPtr __tsp(new __thread_struct);
+    typedef tuple<_TSPtr, typename decay<_Fp>::type, typename decay<_Args>::type...> _Gp;
+    _CUDA_VSTD::unique_ptr<_Gp> __p(
+            new _Gp(std::move(__tsp),
+                    __decay_copy(_CUDA_VSTD::forward<_Fp>(__f)),
+                    __decay_copy(_CUDA_VSTD::forward<_Args>(__args))...));
+    int __ec = __libcpp_thread_create(&__t_, &__thread_proxy<_Gp>, __p.get());
+    if (__ec == 0)
+        __p.release();
+    else
+        __throw_system_error(__ec, "thread constructor failed");
+}
+
+inline
+thread&
+thread::operator=(thread&& __t) _NOEXCEPT
+{
+    if (!__libcpp_thread_isnull(&__t_))
+        terminate();
+    __t_ = __t.__t_;
+    __t.__t_ = _LIBCUDACXX_NULL_THREAD;
+    return *this;
+}
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Fp>
+struct __thread_invoke_pair {
+    // This type is used to pass memory for thread local storage and a functor
+    // to a newly created thread because std::pair doesn't work with
+    // std::unique_ptr in C++03.
+    __thread_invoke_pair(_Fp& __f) : __tsp_(new __thread_struct), __fn_(__f) {}
+    unique_ptr<__thread_struct> __tsp_;
+    _Fp __fn_;
+};
+
+template <class _Fp>
+void* __thread_proxy_cxx03(void* __vp)
+{
+    std::unique_ptr<_Fp> __p(static_cast<_Fp*>(__vp));
+    __thread_local_data().set_pointer(__p->__tsp_.release());
+    (__p->__fn_)();
+    return nullptr;
+}
+
+template <class _Fp>
+thread::thread(_Fp __f)
+{
+
+    typedef __thread_invoke_pair<_Fp> _InvokePair;
+    typedef std::unique_ptr<_InvokePair> _PairPtr;
+    _PairPtr __pp(new _InvokePair(__f));
+    int __ec = __libcpp_thread_create(&__t_, &__thread_proxy_cxx03<_InvokePair>, __pp.get());
+    if (__ec == 0)
+        __pp.release();
+    else
+        __throw_system_error(__ec, "thread constructor failed");
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void swap(thread& __x, thread& __y) _NOEXCEPT {__x.swap(__y);}
+
+namespace this_thread
+{
+
+_LIBCUDACXX_FUNC_VIS void sleep_for(const chrono::nanoseconds& __ns);
+
+template <class _Rep, class _Period>
+void
+sleep_for(const chrono::duration<_Rep, _Period>& __d)
+{
+    using namespace chrono;
+    if (__d > duration<_Rep, _Period>::zero())
+    {
+#if defined(_LIBCUDACXX_COMPILER_GCC) && (__powerpc__ || __POWERPC__)
+    //  GCC's long double const folding is incomplete for IBM128 long doubles.
+        _LIBCUDACXX_CONSTEXPR duration<long double> _Max = nanoseconds::max();
+#else
+        _LIBCUDACXX_CONSTEXPR duration<long double> _Max = duration<long double>(ULLONG_MAX/1000000000ULL) ;
+#endif
+        nanoseconds __ns;
+        if (__d < _Max)
+        {
+            __ns = duration_cast<nanoseconds>(__d);
+            if (__ns < __d)
+                ++__ns;
+        }
+        else
+            __ns = nanoseconds::max();
+        sleep_for(__ns);
+    }
+}
+
+template <class _Clock, class _Duration>
+void
+sleep_until(const chrono::time_point<_Clock, _Duration>& __t)
+{
+    using namespace chrono;
+    mutex __mut;
+    condition_variable __cv;
+    unique_lock<mutex> __lk(__mut);
+    while (_Clock::now() < __t)
+        __cv.wait_until(__lk, __t);
+}
+
+template <class _Duration>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+sleep_until(const chrono::time_point<chrono::steady_clock, _Duration>& __t)
+{
+    using namespace chrono;
+    sleep_for(__t - steady_clock::now());
+}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void yield() _NOEXCEPT {__libcpp_thread_yield();}
+
+}  // this_thread
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif // !_LIBCUDACXX_HAS_NO_THREADS
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_THREAD
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tuple b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tuple
new file mode 100644
index 000000000000..6ba6cd75e728
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/tuple
@@ -0,0 +1,1547 @@
+// -*- C++ -*-
+//===--------------------------- tuple ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_TUPLE
+#define _LIBCUDACXX_TUPLE
+
+/*
+    tuple synopsis
+
+namespace std
+{
+
+template <class... T>
+class tuple {
+public:
+    explicit(see-below) constexpr tuple();
+    explicit(see-below) tuple(const T&...);  // constexpr in C++14
+    template <class... U>
+        explicit(see-below) tuple(U&&...);  // constexpr in C++14
+    tuple(const tuple&) = default;
+    tuple(tuple&&) = default;
+    template <class... U>
+        explicit(see-below) tuple(const tuple<U...>&);  // constexpr in C++14
+    template <class... U>
+        explicit(see-below) tuple(tuple<U...>&&);  // constexpr in C++14
+    template <class U1, class U2>
+        explicit(see-below) tuple(const pair<U1, U2>&); // iff sizeof...(T) == 2 // constexpr in C++14
+    template <class U1, class U2>
+        explicit(see-below) tuple(pair<U1, U2>&&); // iff sizeof...(T) == 2  // constexpr in C++14
+
+    // allocator-extended constructors
+    template <class Alloc>
+        tuple(allocator_arg_t, const Alloc& a);
+    template <class Alloc>
+        explicit(see-below) tuple(allocator_arg_t, const Alloc& a, const T&...);
+    template <class Alloc, class... U>
+        explicit(see-below) tuple(allocator_arg_t, const Alloc& a, U&&...);
+    template <class Alloc>
+        tuple(allocator_arg_t, const Alloc& a, const tuple&);
+    template <class Alloc>
+        tuple(allocator_arg_t, const Alloc& a, tuple&&);
+    template <class Alloc, class... U>
+        explicit(see-below) tuple(allocator_arg_t, const Alloc& a, const tuple<U...>&);
+    template <class Alloc, class... U>
+        explicit(see-below) tuple(allocator_arg_t, const Alloc& a, tuple<U...>&&);
+    template <class Alloc, class U1, class U2>
+        explicit(see-below) tuple(allocator_arg_t, const Alloc& a, const pair<U1, U2>&);
+    template <class Alloc, class U1, class U2>
+        explicit(see-below) tuple(allocator_arg_t, const Alloc& a, pair<U1, U2>&&);
+
+    tuple& operator=(const tuple&);
+    tuple&
+        operator=(tuple&&) noexcept(AND(is_nothrow_move_assignable<T>::value ...));
+    template <class... U>
+        tuple& operator=(const tuple<U...>&);
+    template <class... U>
+        tuple& operator=(tuple<U...>&&);
+    template <class U1, class U2>
+        tuple& operator=(const pair<U1, U2>&); // iff sizeof...(T) == 2
+    template <class U1, class U2>
+        tuple& operator=(pair<U1, U2>&&); // iff sizeof...(T) == 2
+
+    void swap(tuple&) noexcept(AND(swap(declval<T&>(), declval<T&>())...));
+};
+
+template <class ...T>
+tuple(T...) -> tuple<T...>;                                         // since C++17
+template <class T1, class T2>
+tuple(pair<T1, T2>) -> tuple<T1, T2>;                               // since C++17
+template <class Alloc, class ...T>
+tuple(allocator_arg_t, Alloc, T...) -> tuple<T...>;                 // since C++17
+template <class Alloc, class T1, class T2>
+tuple(allocator_arg_t, Alloc, pair<T1, T2>) -> tuple<T1, T2>;       // since C++17
+template <class Alloc, class ...T>
+tuple(allocator_arg_t, Alloc, tuple<T...>) -> tuple<T...>;          // since C++17
+
+inline constexpr unspecified ignore;
+
+template <class... T> tuple<V...>  make_tuple(T&&...); // constexpr in C++14
+template <class... T> tuple<ATypes...> forward_as_tuple(T&&...) noexcept; // constexpr in C++14
+template <class... T> tuple<T&...> tie(T&...) noexcept; // constexpr in C++14
+template <class... Tuples> tuple<CTypes...> tuple_cat(Tuples&&... tpls); // constexpr in C++14
+
+// [tuple.apply], calling a function with a tuple of arguments:
+template <class F, class Tuple>
+  constexpr decltype(auto) apply(F&& f, Tuple&& t); // C++17
+template <class T, class Tuple>
+  constexpr T make_from_tuple(Tuple&& t); // C++17
+
+// 20.4.1.4, tuple helper classes:
+template <class T> struct tuple_size; // undefined
+template <class... T> struct tuple_size<tuple<T...>>;
+template <class T>
+ inline constexpr size_t tuple_size_v = tuple_size<T>::value; // C++17
+template <size_t I, class T> struct tuple_element; // undefined
+template <size_t I, class... T> struct tuple_element<I, tuple<T...>>;
+template <size_t I, class T>
+  using tuple_element_t = typename tuple_element <I, T>::type; // C++14
+
+// 20.4.1.5, element access:
+template <size_t I, class... T>
+    typename tuple_element<I, tuple<T...>>::type&
+    get(tuple<T...>&) noexcept; // constexpr in C++14
+template <size_t I, class... T>
+    const typename tuple_element<I, tuple<T...>>::type&
+    get(const tuple<T...>&) noexcept; // constexpr in C++14
+template <size_t I, class... T>
+    typename tuple_element<I, tuple<T...>>::type&&
+    get(tuple<T...>&&) noexcept; // constexpr in C++14
+template <size_t I, class... T>
+    const typename tuple_element<I, tuple<T...>>::type&&
+    get(const tuple<T...>&&) noexcept; // constexpr in C++14
+
+template <class T1, class... T>
+    constexpr T1& get(tuple<T...>&) noexcept;  // C++14
+template <class T1, class... T>
+    constexpr const T1& get(const tuple<T...>&) noexcept;   // C++14
+template <class T1, class... T>
+    constexpr T1&& get(tuple<T...>&&) noexcept;   // C++14
+template <class T1, class... T>
+    constexpr const T1&& get(const tuple<T...>&&) noexcept;   // C++14
+
+// 20.4.1.6, relational operators:
+template<class... T, class... U> bool operator==(const tuple<T...>&, const tuple<U...>&); // constexpr in C++14
+template<class... T, class... U> bool operator<(const tuple<T...>&, const tuple<U...>&);  // constexpr in C++14
+template<class... T, class... U> bool operator!=(const tuple<T...>&, const tuple<U...>&); // constexpr in C++14
+template<class... T, class... U> bool operator>(const tuple<T...>&, const tuple<U...>&);  // constexpr in C++14
+template<class... T, class... U> bool operator<=(const tuple<T...>&, const tuple<U...>&); // constexpr in C++14
+template<class... T, class... U> bool operator>=(const tuple<T...>&, const tuple<U...>&); // constexpr in C++14
+
+template <class... Types, class Alloc>
+  struct uses_allocator<tuple<Types...>, Alloc>;
+
+template <class... Types>
+  void
+  swap(tuple<Types...>& x, tuple<Types...>& y) noexcept(noexcept(x.swap(y)));
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__tuple>
+#include <cstddef>
+#include <type_traits>
+#include <__functional_base>
+#include <utility>
+#include <version>
+#include <__pragma_push>
+#endif // __cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+
+// __tuple_leaf
+
+template <size_t _Ip, class _Hp,
+          bool=is_empty<_Hp>::value && !__libcpp_is_final<_Hp>::value
+         >
+class __tuple_leaf;
+
+template <size_t _Ip, class _Hp, bool _Ep>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void swap(__tuple_leaf<_Ip, _Hp, _Ep>& __x, __tuple_leaf<_Ip, _Hp, _Ep>& __y)
+    _NOEXCEPT_(__is_nothrow_swappable<_Hp>::value)
+{
+    swap(__x.get(), __y.get());
+}
+
+template <size_t _Ip, class _Hp, bool>
+class __tuple_leaf
+{
+    _Hp __value_;
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr bool __can_bind_reference() {
+#if __has_keyword(__reference_binds_to_temporary)
+      return !__reference_binds_to_temporary(_Hp, _Tp);
+#else
+      return true;
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY __tuple_leaf& operator=(const __tuple_leaf&);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR __tuple_leaf()
+             _NOEXCEPT_(is_nothrow_default_constructible<_Hp>::value) : __value_()
+       {static_assert(!is_reference<_Hp>::value,
+              "Attempted to default construct a reference element in a tuple");}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf(integral_constant<int, 0>, const _Alloc&)
+            : __value_()
+        {static_assert(!is_reference<_Hp>::value,
+              "Attempted to default construct a reference element in a tuple");}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf(integral_constant<int, 1>, const _Alloc& __a)
+            : __value_(allocator_arg_t(), __a)
+        {static_assert(!is_reference<_Hp>::value,
+              "Attempted to default construct a reference element in a tuple");}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf(integral_constant<int, 2>, const _Alloc& __a)
+            : __value_(__a)
+        {static_assert(!is_reference<_Hp>::value,
+              "Attempted to default construct a reference element in a tuple");}
+
+    template <class _Tp,
+              _EnableIf<
+                  _And<
+                      _IsNotSame<__uncvref_t<_Tp>, __tuple_leaf>,
+                      is_constructible<_Hp, _Tp>
+                    >::value,
+                  bool
+                > = false
+            >
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit __tuple_leaf(_Tp&& __t) _NOEXCEPT_((is_nothrow_constructible<_Hp, _Tp>::value))
+            : __value_(_CUDA_VSTD::forward<_Tp>(__t))
+        {static_assert(__can_bind_reference<_Tp&&>(),
+       "Attempted construction of reference element binds to a temporary whose lifetime has ended");}
+
+    template <class _Tp, class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit __tuple_leaf(integral_constant<int, 0>, const _Alloc&, _Tp&& __t)
+            : __value_(_CUDA_VSTD::forward<_Tp>(__t))
+        {static_assert(__can_bind_reference<_Tp&&>(),
+       "Attempted construction of reference element binds to a temporary whose lifetime has ended");}
+
+    template <class _Tp, class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit __tuple_leaf(integral_constant<int, 1>, const _Alloc& __a, _Tp&& __t)
+            : __value_(allocator_arg_t(), __a, _CUDA_VSTD::forward<_Tp>(__t))
+        {static_assert(!is_reference<_Hp>::value,
+            "Attempted to uses-allocator construct a reference element in a tuple");}
+
+    template <class _Tp, class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit __tuple_leaf(integral_constant<int, 2>, const _Alloc& __a, _Tp&& __t)
+            : __value_(_CUDA_VSTD::forward<_Tp>(__t), __a)
+        {static_assert(!is_reference<_Hp>::value,
+           "Attempted to uses-allocator construct a reference element in a tuple");}
+
+    __tuple_leaf(const __tuple_leaf& __t) = default;
+    __tuple_leaf(__tuple_leaf&& __t) = default;
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf&
+        operator=(_Tp&& __t) _NOEXCEPT_((is_nothrow_assignable<_Hp&, _Tp>::value))
+        {
+            __value_ = _CUDA_VSTD::forward<_Tp>(__t);
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int swap(__tuple_leaf& __t) _NOEXCEPT_(__is_nothrow_swappable<__tuple_leaf>::value)
+    {
+        _CUDA_VSTD::swap(*this, __t);
+        return 0;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11       _Hp& get()       _NOEXCEPT {return __value_;}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const _Hp& get() const _NOEXCEPT {return __value_;}
+};
+
+template <size_t _Ip, class _Hp>
+class __tuple_leaf<_Ip, _Hp, true>
+    : private _Hp
+{
+
+    _LIBCUDACXX_INLINE_VISIBILITY __tuple_leaf& operator=(const __tuple_leaf&);
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR __tuple_leaf()
+             _NOEXCEPT_(is_nothrow_default_constructible<_Hp>::value) {}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf(integral_constant<int, 0>, const _Alloc&) {}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf(integral_constant<int, 1>, const _Alloc& __a)
+            : _Hp(allocator_arg_t(), __a) {}
+
+    template <class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf(integral_constant<int, 2>, const _Alloc& __a)
+            : _Hp(__a) {}
+
+    template <class _Tp,
+              _EnableIf<
+                  _And<
+                    _IsNotSame<__uncvref_t<_Tp>, __tuple_leaf>,
+                    is_constructible<_Hp, _Tp>
+                  >::value,
+                  bool
+                > = false
+            >
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit __tuple_leaf(_Tp&& __t) _NOEXCEPT_((is_nothrow_constructible<_Hp, _Tp>::value))
+            : _Hp(_CUDA_VSTD::forward<_Tp>(__t)) {}
+
+    template <class _Tp, class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit __tuple_leaf(integral_constant<int, 0>, const _Alloc&, _Tp&& __t)
+            : _Hp(_CUDA_VSTD::forward<_Tp>(__t)) {}
+
+    template <class _Tp, class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit __tuple_leaf(integral_constant<int, 1>, const _Alloc& __a, _Tp&& __t)
+            : _Hp(allocator_arg_t(), __a, _CUDA_VSTD::forward<_Tp>(__t)) {}
+
+    template <class _Tp, class _Alloc>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit __tuple_leaf(integral_constant<int, 2>, const _Alloc& __a, _Tp&& __t)
+            : _Hp(_CUDA_VSTD::forward<_Tp>(__t), __a) {}
+
+    __tuple_leaf(__tuple_leaf const &) = default;
+    __tuple_leaf(__tuple_leaf &&) = default;
+
+    template <class _Tp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_leaf&
+        operator=(_Tp&& __t) _NOEXCEPT_((is_nothrow_assignable<_Hp&, _Tp>::value))
+        {
+            _Hp::operator=(_CUDA_VSTD::forward<_Tp>(__t));
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    int
+    swap(__tuple_leaf& __t) _NOEXCEPT_(__is_nothrow_swappable<__tuple_leaf>::value)
+    {
+        _CUDA_VSTD::swap(*this, __t);
+        return 0;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11       _Hp& get()       _NOEXCEPT {return static_cast<_Hp&>(*this);}
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 const _Hp& get() const _NOEXCEPT {return static_cast<const _Hp&>(*this);}
+};
+
+template <class ..._Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __swallow(_Tp&&...) _NOEXCEPT {}
+
+template <class _Tp>
+struct __all_default_constructible;
+
+template <class ..._Tp>
+struct __all_default_constructible<__tuple_types<_Tp...>>
+    : __all<is_default_constructible<_Tp>::value...>
+{ };
+
+// __tuple_impl
+
+template<class _Indx, class ..._Tp> struct __tuple_impl;
+
+template<size_t ..._Indx, class ..._Tp>
+struct _LIBCUDACXX_DECLSPEC_EMPTY_BASES __tuple_impl<__tuple_indices<_Indx...>, _Tp...>
+    : public __tuple_leaf<_Indx, _Tp>...
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _LIBCUDACXX_CONSTEXPR __tuple_impl()
+        _NOEXCEPT_(__all<is_nothrow_default_constructible<_Tp>::value...>::value) {}
+
+    // Handle non-allocator, full initialization
+    template <size_t ..._Uf, class ..._Tf,
+              size_t ..._Ul, class ..._Tl, class ..._Up,
+              _EnableIf<sizeof...(_Up) == sizeof...(_Tp), bool> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit
+        __tuple_impl(__tuple_indices<_Uf...>, __tuple_types<_Tf...>,
+                     __tuple_indices<_Ul...>, __tuple_types<_Tl...>,
+                     _Up&&... __u)
+                     _NOEXCEPT_((__all<is_nothrow_constructible<_Tf, _Up>::value...>::value)) :
+            __tuple_leaf<_Uf, _Tf>(_CUDA_VSTD::forward<_Up>(__u))...
+            {}
+
+    // Handle non-allocator, partial default initialization
+    template <size_t ..._Uf, class ..._Tf,
+              size_t ..._Ul, class ..._Tl, class ..._Up,
+              _EnableIf<(sizeof...(_Up) > 0) && (sizeof...(_Up) < sizeof...(_Tp)), bool> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit
+        __tuple_impl(__tuple_indices<_Uf...>, __tuple_types<_Tf...>,
+                     __tuple_indices<_Ul...>, __tuple_types<_Tl...>,
+                     _Up&&... __u)
+                     _NOEXCEPT_((__all<is_nothrow_constructible<_Tf, _Up>::value...>::value &&
+                                 __all<is_nothrow_default_constructible<_Tl>::value...>::value)) :
+            __tuple_leaf<_Uf, _Tf>(_CUDA_VSTD::forward<_Up>(__u))...,
+            __tuple_leaf<_Ul, _Tl>()...
+            {}
+
+    // Handle non-allocator, full default initialization
+    template <size_t ..._Uf, class ..._Tf,
+              size_t ..._Ul, class ..._Tl, class ..._Up,
+              _EnableIf<sizeof...(_Up) == 0, bool> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit
+        __tuple_impl(__tuple_indices<_Uf...>, __tuple_types<_Tf...>,
+                     __tuple_indices<_Ul...>, __tuple_types<_Tl...>,
+                     _Up&&...)
+                     _NOEXCEPT_((__all<is_nothrow_default_constructible<_Tl>::value...>::value)) :
+            __tuple_leaf<_Ul, _Tl>()...
+            {}
+
+
+    // Handle allocator aware, full initialization
+    template <class _Alloc, size_t ..._Uf, class ..._Tf,
+              size_t ..._Ul, class ..._Tl, class ..._Up,
+              _EnableIf<sizeof...(_Up) == sizeof...(_Tp), bool> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit
+        __tuple_impl(allocator_arg_t, const _Alloc& __a,
+                     __tuple_indices<_Uf...>, __tuple_types<_Tf...>,
+                     __tuple_indices<_Ul...>, __tuple_types<_Tl...>,
+                     _Up&&... __u) :
+            __tuple_leaf<_Uf, _Tf>(__uses_alloc_ctor<_Tf, _Alloc, _Up>(), __a,
+                    _CUDA_VSTD::forward<_Up>(__u))...
+            {}
+
+    // Handle allocator aware, partial default initialization
+    template <class _Alloc, size_t ..._Uf, class ..._Tf,
+              size_t ..._Ul, class ..._Tl, class ..._Up,
+              _EnableIf<(sizeof...(_Up) > 0) && (sizeof...(_Up) < sizeof...(_Tp)), bool> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit
+        __tuple_impl(allocator_arg_t, const _Alloc& __a,
+                     __tuple_indices<_Uf...>, __tuple_types<_Tf...>,
+                     __tuple_indices<_Ul...>, __tuple_types<_Tl...>,
+                     _Up&&... __u) :
+            __tuple_leaf<_Uf, _Tf>(__uses_alloc_ctor<_Tf, _Alloc, _Up>(), __a,
+                    _CUDA_VSTD::forward<_Up>(__u))...,
+            __tuple_leaf<_Ul, _Tl>(__uses_alloc_ctor<_Tl, _Alloc>(), __a)...
+            {}
+
+    // Handle allocator aware, full default initialization
+    template <class _Alloc, size_t ..._Uf, class ..._Tf,
+              size_t ..._Ul, class ..._Tl, class ..._Up,
+              _EnableIf<sizeof...(_Up) == 0, bool> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit
+        __tuple_impl(allocator_arg_t, const _Alloc& __a,
+                     __tuple_indices<_Uf...>, __tuple_types<_Tf...>,
+                     __tuple_indices<_Ul...>, __tuple_types<_Tl...>,
+                     _Up&&...) :
+            __tuple_leaf<_Ul, _Tl>(__uses_alloc_ctor<_Tl, _Alloc>(), __a)...
+            {}
+
+    template <class _Tuple,
+              _EnableIf<
+                    __tuple_constructible<_Tuple, tuple<_Tp...> >::value,
+                    bool
+                > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        __tuple_impl(_Tuple&& __t) _NOEXCEPT_((__all<is_nothrow_constructible<_Tp, typename tuple_element<_Indx,
+                                       typename __make_tuple_types<_Tuple>::type>::type>::value...>::value))
+            : __tuple_leaf<_Indx, _Tp>(_CUDA_VSTD::forward<typename tuple_element<_Indx,
+                                       typename __make_tuple_types<_Tuple>::type>::type>(_CUDA_VSTD::get<_Indx>(__t)))...
+            {}
+
+    template <class _Alloc, class _Tuple,
+              _EnableIf<
+                    __tuple_constructible<_Tuple, tuple<_Tp...> >::value,
+                    bool
+                > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        __tuple_impl(allocator_arg_t, const _Alloc& __a, _Tuple&& __t)
+            : __tuple_leaf<_Indx, _Tp>(__uses_alloc_ctor<_Tp, _Alloc, typename tuple_element<_Indx,
+                                       typename __make_tuple_types<_Tuple>::type>::type>(), __a,
+                                       _CUDA_VSTD::forward<typename tuple_element<_Indx,
+                                       typename __make_tuple_types<_Tuple>::type>::type>(_CUDA_VSTD::get<_Indx>(__t)))...
+            {}
+
+    template <class _Tuple>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        typename enable_if
+        <
+            __tuple_assignable<_Tuple, tuple<_Tp...> >::value,
+            __tuple_impl&
+        >::type
+        operator=(_Tuple&& __t) _NOEXCEPT_((__all<is_nothrow_assignable<_Tp&, typename tuple_element<_Indx,
+                                       typename __make_tuple_types<_Tuple>::type>::type>::value...>::value))
+        {
+            __swallow(__tuple_leaf<_Indx, _Tp>::operator=(_CUDA_VSTD::forward<typename tuple_element<_Indx,
+                                       typename __make_tuple_types<_Tuple>::type>::type>(_CUDA_VSTD::get<_Indx>(__t)))...);
+            return *this;
+        }
+
+    __tuple_impl(const __tuple_impl&) = default;
+    __tuple_impl(__tuple_impl&&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tuple_impl&
+    operator=(const __tuple_impl& __t) _NOEXCEPT_((__all<is_nothrow_copy_assignable<_Tp>::value...>::value))
+    {
+        __swallow(__tuple_leaf<_Indx, _Tp>::operator=(static_cast<const __tuple_leaf<_Indx, _Tp>&>(__t).get())...);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __tuple_impl&
+    operator=(__tuple_impl&& __t) _NOEXCEPT_((__all<is_nothrow_move_assignable<_Tp>::value...>::value))
+    {
+        __swallow(__tuple_leaf<_Indx, _Tp>::operator=(_CUDA_VSTD::forward<_Tp>(static_cast<__tuple_leaf<_Indx, _Tp>&>(__t).get()))...);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(__tuple_impl& __t)
+        _NOEXCEPT_(__all<__is_nothrow_swappable<_Tp>::value...>::value)
+    {
+        __swallow(__tuple_leaf<_Indx, _Tp>::swap(static_cast<__tuple_leaf<_Indx, _Tp>&>(__t))...);
+    }
+};
+
+
+
+template <class ..._Tp>
+class _LIBCUDACXX_TEMPLATE_VIS tuple
+{
+    typedef __tuple_impl<typename __make_tuple_indices<sizeof...(_Tp)>::type, _Tp...> _BaseT;
+
+    _BaseT __base_;
+
+#if defined(_LIBCUDACXX_ENABLE_TUPLE_IMPLICIT_REDUCED_ARITY_EXTENSION)
+    static constexpr bool _EnableImplicitReducedArityExtension = true;
+#else
+    static constexpr bool _EnableImplicitReducedArityExtension = false;
+#endif
+
+    template <class ..._Args>
+    struct _PackExpandsToThisTuple : false_type {};
+
+    template <class _Arg>
+    struct _PackExpandsToThisTuple<_Arg>
+        : is_same<typename __uncvref<_Arg>::type, tuple> {};
+
+    template <bool _MaybeEnable, class _Dummy = void>
+    struct _CheckArgsConstructor : __check_tuple_constructor_fail {};
+
+    template <class _Dummy>
+    struct _CheckArgsConstructor<true, _Dummy>
+    {
+        template <int&...>
+        struct __enable_implicit_default {
+            static constexpr bool value =
+                __all<__is_implicitly_default_constructible<_Tp>::value...>::value;
+        };
+
+        template <int&...>
+        struct __enable_explicit_default {
+            static constexpr bool value =
+                __all<is_default_constructible<_Tp>::value ...>::value &&
+                !__enable_implicit_default<>::value;
+        };
+
+        template <class ..._Args>
+        struct __enable_explicit {
+            static constexpr bool value =
+                __tuple_constructible<
+                    tuple<_Args...>,
+                    typename __make_tuple_types<tuple,
+                             sizeof...(_Args) < sizeof...(_Tp) ?
+                                 sizeof...(_Args) :
+                                 sizeof...(_Tp)>::type
+                >::value &&
+                !__tuple_convertible<
+                    tuple<_Args...>,
+                    typename __make_tuple_types<tuple,
+                             sizeof...(_Args) < sizeof...(_Tp) ?
+                                 sizeof...(_Args) :
+                                 sizeof...(_Tp)>::type
+                >::value &&
+                __all_default_constructible<
+                    typename __make_tuple_types<tuple, sizeof...(_Tp),
+                             sizeof...(_Args) < sizeof...(_Tp) ?
+                                 sizeof...(_Args) :
+                                 sizeof...(_Tp)>::type
+                >::value;
+        };
+
+        template <class ..._Args>
+        struct __enable_implicit {
+            static constexpr bool value =
+               __tuple_constructible<
+                    tuple<_Args...>,
+                    typename __make_tuple_types<tuple,
+                             sizeof...(_Args) < sizeof...(_Tp) ?
+                                 sizeof...(_Args) :
+                                 sizeof...(_Tp)>::type
+                >::value &&
+                __tuple_convertible<
+                    tuple<_Args...>,
+                    typename __make_tuple_types<tuple,
+                             sizeof...(_Args) < sizeof...(_Tp) ?
+                                 sizeof...(_Args) :
+                                 sizeof...(_Tp)>::type
+                >::value &&
+                __all_default_constructible<
+                    typename __make_tuple_types<tuple, sizeof...(_Tp),
+                             sizeof...(_Args) < sizeof...(_Tp) ?
+                                 sizeof...(_Args) :
+                                 sizeof...(_Tp)>::type
+                >::value;
+        };
+    };
+
+    template <bool _MaybeEnable,
+              bool = sizeof...(_Tp) == 1,
+              class _Dummy = void>
+    struct _CheckTupleLikeConstructor : __check_tuple_constructor_fail {};
+
+    template <class _Dummy>
+    struct _CheckTupleLikeConstructor<true, false, _Dummy>
+    {
+        template <class _Tuple>
+        struct __enable_implicit {
+            static constexpr bool value =
+                   __tuple_constructible<_Tuple, tuple>::value
+                && __tuple_convertible<_Tuple, tuple>::value;
+        };
+
+        template <class _Tuple>
+        struct __enable_explicit {
+            static constexpr bool value =
+                   __tuple_constructible<_Tuple, tuple>::value
+               && !__tuple_convertible<_Tuple, tuple>::value;
+        };
+    };
+
+    template <class _Dummy>
+    struct _CheckTupleLikeConstructor<true, true, _Dummy>
+    {
+        // This trait is used to disable the tuple-like constructor when
+        // the UTypes... constructor should be selected instead.
+        // See LWG issue #2549.
+        template <class _Tuple>
+        using _PreferTupleLikeConstructor = _Or<
+            // Don't attempt the two checks below if the tuple we are given
+            // has the same type as this tuple.
+            _IsSame<__uncvref_t<_Tuple>, tuple>,
+            _Lazy<_And,
+                _Not<is_constructible<_Tp..., _Tuple>>,
+                _Not<is_convertible<_Tuple, _Tp...>>
+            >
+        >;
+
+        template <class _Tuple>
+        struct __enable_implicit {
+            static constexpr bool value =
+                _And<
+                    __tuple_constructible<_Tuple, tuple>,
+                    __tuple_convertible<_Tuple, tuple>,
+                    _PreferTupleLikeConstructor<_Tuple>
+                >::value;
+        };
+
+        template <class _Tuple>
+        struct __enable_explicit {
+            static constexpr bool value =
+                _And<
+                    __tuple_constructible<_Tuple, tuple>,
+                    _PreferTupleLikeConstructor<_Tuple>,
+                    _Not<__tuple_convertible<_Tuple, tuple>>
+                >::value;
+        };
+    };
+
+    template <class _Tuple, bool _DisableIfLValue>
+    using _EnableImplicitTupleLikeConstructor = _EnableIf<
+                         _CheckTupleLikeConstructor<
+                             __tuple_like_with_size<_Tuple, sizeof...(_Tp)>::value
+                             && !_PackExpandsToThisTuple<_Tuple>::value
+                             && (!is_lvalue_reference<_Tuple>::value || !_DisableIfLValue)
+                         >::template __enable_implicit<_Tuple>::value,
+                         bool
+                      >;
+
+    template <class _Tuple, bool _DisableIfLValue>
+    using _EnableExplicitTupleLikeConstructor = _EnableIf<
+                         _CheckTupleLikeConstructor<
+                             __tuple_like_with_size<_Tuple, sizeof...(_Tp)>::value
+                             && !_PackExpandsToThisTuple<_Tuple>::value
+                             && (!is_lvalue_reference<_Tuple>::value || !_DisableIfLValue)
+                         >::template __enable_explicit<_Tuple>::value,
+                         bool
+                      >;
+
+    template <size_t _Jp, class ..._Up> friend
+        _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+        typename tuple_element<_Jp, tuple<_Up...> >::type& get(tuple<_Up...>&) _NOEXCEPT;
+    template <size_t _Jp, class ..._Up> friend
+        _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+        const typename tuple_element<_Jp, tuple<_Up...> >::type& get(const tuple<_Up...>&) _NOEXCEPT;
+    template <size_t _Jp, class ..._Up> friend
+        _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+        typename tuple_element<_Jp, tuple<_Up...> >::type&& get(tuple<_Up...>&&) _NOEXCEPT;
+    template <size_t _Jp, class ..._Up> friend
+        _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+        const typename tuple_element<_Jp, tuple<_Up...> >::type&& get(const tuple<_Up...>&&) _NOEXCEPT;
+public:
+
+    template <bool _Dummy = true, _EnableIf<
+        _CheckArgsConstructor<_Dummy>::template __enable_implicit_default<>::value
+    , bool> = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    tuple()
+        _NOEXCEPT_(__all<is_nothrow_default_constructible<_Tp>::value...>::value) {}
+
+    template <bool _Dummy = true, _EnableIf<
+        _CheckArgsConstructor<_Dummy>::template __enable_explicit_default<>::value
+    , bool> = false>
+    explicit _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    tuple()
+        _NOEXCEPT_(__all<is_nothrow_default_constructible<_Tp>::value...>::value) {}
+
+    tuple(tuple const&) = default;
+    tuple(tuple&&) = default;
+
+    template <class _AllocArgT, class _Alloc, _EnableIf<
+             _CheckArgsConstructor<_IsSame<allocator_arg_t, _AllocArgT>::value >::template __enable_implicit_default<>::value
+      , bool> = false
+    >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple(_AllocArgT, _Alloc const& __a)
+      : __base_(allocator_arg_t(), __a,
+                    __tuple_indices<>(), __tuple_types<>(),
+                    typename __make_tuple_indices<sizeof...(_Tp), 0>::type(),
+                    __tuple_types<_Tp...>()) {}
+
+    template <class _AllocArgT, class _Alloc, _EnableIf<
+             _CheckArgsConstructor<_IsSame<allocator_arg_t, _AllocArgT>::value>::template __enable_explicit_default<>::value
+      , bool> = false
+    >
+    explicit _LIBCUDACXX_INLINE_VISIBILITY
+    tuple(_AllocArgT, _Alloc const& __a)
+      : __base_(allocator_arg_t(), __a,
+                    __tuple_indices<>(), __tuple_types<>(),
+                    typename __make_tuple_indices<sizeof...(_Tp), 0>::type(),
+                    __tuple_types<_Tp...>()) {}
+
+    template <bool _Dummy = true,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                            _Dummy
+                         >::template __enable_implicit<_Tp const&...>::value,
+                         bool
+                      > = false
+        >
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    tuple(const _Tp& ... __t) _NOEXCEPT_((__all<is_nothrow_copy_constructible<_Tp>::value...>::value))
+        : __base_(typename __make_tuple_indices<sizeof...(_Tp)>::type(),
+                typename __make_tuple_types<tuple, sizeof...(_Tp)>::type(),
+                typename __make_tuple_indices<0>::type(),
+                typename __make_tuple_types<tuple, 0>::type(),
+                __t...
+               ) {}
+
+    template <bool _Dummy = true,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                            _Dummy
+                         >::template __enable_explicit<_Tp const&...>::value,
+                         bool
+                      > = false
+        >
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    explicit tuple(const _Tp& ... __t) _NOEXCEPT_((__all<is_nothrow_copy_constructible<_Tp>::value...>::value))
+        : __base_(typename __make_tuple_indices<sizeof...(_Tp)>::type(),
+                typename __make_tuple_types<tuple, sizeof...(_Tp)>::type(),
+                typename __make_tuple_indices<0>::type(),
+                typename __make_tuple_types<tuple, 0>::type(),
+                __t...
+               ) {}
+
+    template <class _Alloc, bool _Dummy = true,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                            _Dummy
+                         >::template __enable_implicit<_Tp const&...>::value,
+                         bool
+                      > = false
+        >
+      _LIBCUDACXX_INLINE_VISIBILITY
+      tuple(allocator_arg_t, const _Alloc& __a, const _Tp& ... __t)
+        : __base_(allocator_arg_t(), __a,
+                typename __make_tuple_indices<sizeof...(_Tp)>::type(),
+                typename __make_tuple_types<tuple, sizeof...(_Tp)>::type(),
+                typename __make_tuple_indices<0>::type(),
+                typename __make_tuple_types<tuple, 0>::type(),
+                __t...
+               ) {}
+
+    template <class _Alloc, bool _Dummy = true,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                            _Dummy
+                         >::template __enable_explicit<_Tp const&...>::value,
+                         bool
+                      > = false
+        >
+      _LIBCUDACXX_INLINE_VISIBILITY
+      explicit
+      tuple(allocator_arg_t, const _Alloc& __a, const _Tp& ... __t)
+        : __base_(allocator_arg_t(), __a,
+                typename __make_tuple_indices<sizeof...(_Tp)>::type(),
+                typename __make_tuple_types<tuple, sizeof...(_Tp)>::type(),
+                typename __make_tuple_indices<0>::type(),
+                typename __make_tuple_types<tuple, 0>::type(),
+                __t...
+               ) {}
+
+#if defined(_LIBCUDACXX_NO_TUPLE_NOEXCEPT)
+    template <class ..._Vp>
+    using __base_noexcept_constructible = false_type;
+#else
+    template <class ..._Vp>
+    using __base_noexcept_constructible = is_nothrow_constructible<
+                _BaseT,
+                _Vp...
+            >;
+#endif // defined(_LIBCUDACXX_NO_TUPLE_NOEXCEPT)
+
+    template <class ..._Up,
+              bool _PackIsTuple = _PackExpandsToThisTuple<_Up...>::value,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                             sizeof...(_Up) == sizeof...(_Tp)
+                             && !_PackIsTuple
+                         >::template __enable_implicit<_Up...>::value ||
+                        _CheckArgsConstructor<
+                            _EnableImplicitReducedArityExtension
+                            && sizeof...(_Up) < sizeof...(_Tp)
+                            && !_PackIsTuple
+                         >::template __enable_implicit<_Up...>::value,
+                         bool
+                      > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        tuple(_Up&&... __u)
+            _NOEXCEPT_((
+                __base_noexcept_constructible<
+                    typename __make_tuple_indices<sizeof...(_Up)>::type,
+                    typename __make_tuple_types<tuple, sizeof...(_Up)>::type,
+                    typename __make_tuple_indices<sizeof...(_Tp), sizeof...(_Up)>::type,
+                    typename __make_tuple_types<tuple, sizeof...(_Tp), sizeof...(_Up)>::type,
+                    _Up...
+                >::value
+            ))
+            : __base_(typename __make_tuple_indices<sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Up)>::type(),
+                    typename __make_tuple_indices<sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    _CUDA_VSTD::forward<_Up>(__u)...) {}
+
+    template <class ..._Up,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                             sizeof...(_Up) <= sizeof...(_Tp)
+                             && !_PackExpandsToThisTuple<_Up...>::value
+                         >::template __enable_explicit<_Up...>::value ||
+                         _CheckArgsConstructor<
+                            !_EnableImplicitReducedArityExtension
+                            && sizeof...(_Up) < sizeof...(_Tp)
+                            && !_PackExpandsToThisTuple<_Up...>::value
+                         >::template __enable_implicit<_Up...>::value,
+                         bool
+                      > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit
+        tuple(_Up&&... __u)
+            _NOEXCEPT_((
+                __base_noexcept_constructible<
+                    typename __make_tuple_indices<sizeof...(_Up)>::type,
+                    typename __make_tuple_types<tuple, sizeof...(_Up)>::type,
+                    typename __make_tuple_indices<sizeof...(_Tp), sizeof...(_Up)>::type,
+                    typename __make_tuple_types<tuple, sizeof...(_Tp), sizeof...(_Up)>::type,
+                    _Up...
+                >::value
+            ))
+            : __base_(typename __make_tuple_indices<sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Up)>::type(),
+                    typename __make_tuple_indices<sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    _CUDA_VSTD::forward<_Up>(__u)...) {}
+
+    template <class _Alloc, class ..._Up,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                             sizeof...(_Up) == sizeof...(_Tp) &&
+                             !_PackExpandsToThisTuple<_Up...>::value
+                         >::template __enable_implicit<_Up...>::value,
+                         bool
+                      > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        tuple(allocator_arg_t, const _Alloc& __a, _Up&&... __u)
+            : __base_(allocator_arg_t(), __a,
+                    typename __make_tuple_indices<sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Up)>::type(),
+                    typename __make_tuple_indices<sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    _CUDA_VSTD::forward<_Up>(__u)...) {}
+
+    template <class _Alloc, class ..._Up,
+              _EnableIf
+                      <
+                         _CheckArgsConstructor<
+                             sizeof...(_Up) == sizeof...(_Tp) &&
+                             !_PackExpandsToThisTuple<_Up...>::value
+                         >::template __enable_explicit<_Up...>::value,
+                         bool
+                      > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit
+        tuple(allocator_arg_t, const _Alloc& __a, _Up&&... __u)
+            : __base_(allocator_arg_t(), __a,
+                    typename __make_tuple_indices<sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Up)>::type(),
+                    typename __make_tuple_indices<sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    typename __make_tuple_types<tuple, sizeof...(_Tp), sizeof...(_Up)>::type(),
+                    _CUDA_VSTD::forward<_Up>(__u)...) {}
+
+    template <class _Tuple, _EnableImplicitTupleLikeConstructor<_Tuple, true> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        tuple(_Tuple&& __t) _NOEXCEPT_((is_nothrow_constructible<_BaseT, _Tuple>::value))
+            : __base_(_CUDA_VSTD::forward<_Tuple>(__t)) {}
+
+    template <class _Tuple, _EnableImplicitTupleLikeConstructor<const _Tuple&, false> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        tuple(const _Tuple& __t) _NOEXCEPT_((is_nothrow_constructible<_BaseT, const _Tuple&>::value))
+            : __base_(__t) {}
+    template <class _Tuple, _EnableExplicitTupleLikeConstructor<_Tuple, true> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit
+        tuple(_Tuple&& __t) _NOEXCEPT_((is_nothrow_constructible<_BaseT, _Tuple>::value))
+            : __base_(_CUDA_VSTD::forward<_Tuple>(__t)) {}
+
+    template <class _Tuple, _EnableExplicitTupleLikeConstructor<const _Tuple&, false> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+        explicit
+        tuple(const _Tuple& __t) _NOEXCEPT_((is_nothrow_constructible<_BaseT, const _Tuple&>::value))
+            : __base_(__t) {}
+
+    template <typename _Tuple>
+    using _EnableIfTupleLikeConstructor = typename enable_if<
+            _CheckTupleLikeConstructor<
+                __tuple_like_with_size<_Tuple, sizeof...(_Tp)>::value
+            >::template __enable_implicit<_Tuple>::value,
+            bool
+        >::type;
+
+    template <typename _Tuple>
+    using _EnableIfTupleLikeExplicitConstructor = typename enable_if<
+            _CheckTupleLikeConstructor<
+                __tuple_like_with_size<_Tuple, sizeof...(_Tp)>::value
+            >::template __enable_explicit<_Tuple>::value,
+            bool
+        >::type;
+
+    template <class _Alloc, class _Tuple,
+        _EnableIfTupleLikeConstructor<_Tuple> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        tuple(allocator_arg_t, const _Alloc& __a, _Tuple&& __t)
+            : __base_(allocator_arg_t(), __a, _CUDA_VSTD::forward<_Tuple>(__t)) {}
+
+    template <class _Alloc, class _Tuple,
+              _EnableIfTupleLikeExplicitConstructor<_Tuple> = false>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        explicit
+        tuple(allocator_arg_t, const _Alloc& __a, _Tuple&& __t)
+            : __base_(allocator_arg_t(), __a, _CUDA_VSTD::forward<_Tuple>(__t)) {}
+
+    using _CanCopyAssign = __all<is_copy_assignable<_Tp>::value...>;
+    using _CanMoveAssign = __all<is_move_assignable<_Tp>::value...>;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple& operator=(typename conditional<_CanCopyAssign::value, tuple, __nat>::type const& __t)
+        _NOEXCEPT_((__all<is_nothrow_copy_assignable<_Tp>::value...>::value))
+    {
+        __base_.operator=(__t.__base_);
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    tuple& operator=(typename conditional<_CanMoveAssign::value, tuple, __nat>::type&& __t)
+        _NOEXCEPT_((__all<is_nothrow_move_assignable<_Tp>::value...>::value))
+    {
+        __base_.operator=(static_cast<_BaseT&&>(__t.__base_));
+        return *this;
+    }
+
+    template <class _Tuple,
+              _EnableIf
+                      <
+                         __tuple_assignable<_Tuple, tuple>::value,
+                         bool
+                      > = false
+             >
+        _LIBCUDACXX_INLINE_VISIBILITY
+        tuple&
+        operator=(_Tuple&& __t) _NOEXCEPT_((is_nothrow_assignable<_BaseT&, _Tuple>::value))
+        {
+            __base_.operator=(_CUDA_VSTD::forward<_Tuple>(__t));
+            return *this;
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(tuple& __t) _NOEXCEPT_(__all<__is_nothrow_swappable<_Tp>::value...>::value)
+        {__base_.swap(__t.__base_);}
+};
+
+template <>
+class _LIBCUDACXX_TEMPLATE_VIS tuple<>
+{
+public:
+    _LIBCUDACXX_CONSTEXPR tuple() _NOEXCEPT = default;
+    template <class _Alloc>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        tuple(allocator_arg_t, const _Alloc&) _NOEXCEPT {}
+    template <class _Alloc>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        tuple(allocator_arg_t, const _Alloc&, const tuple&) _NOEXCEPT {}
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        tuple(array<_Up, 0>) _NOEXCEPT {}
+    template <class _Alloc, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+        tuple(allocator_arg_t, const _Alloc&, array<_Up, 0>) _NOEXCEPT {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(tuple&) _NOEXCEPT {}
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template <class ..._Tp>
+tuple(_Tp...) -> tuple<_Tp...>;
+template <class _Tp1, class _Tp2>
+tuple(pair<_Tp1, _Tp2>) -> tuple<_Tp1, _Tp2>;
+template <class _Alloc, class ..._Tp>
+tuple(allocator_arg_t, _Alloc, _Tp...) -> tuple<_Tp...>;
+template <class _Alloc, class _Tp1, class _Tp2>
+tuple(allocator_arg_t, _Alloc, pair<_Tp1, _Tp2>) -> tuple<_Tp1, _Tp2>;
+template <class _Alloc, class ..._Tp>
+tuple(allocator_arg_t, _Alloc, tuple<_Tp...>) -> tuple<_Tp...>;
+#endif
+
+template <class ..._Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __all<__is_swappable<_Tp>::value...>::value,
+    void
+>::type
+swap(tuple<_Tp...>& __t, tuple<_Tp...>& __u)
+                 _NOEXCEPT_(__all<__is_nothrow_swappable<_Tp>::value...>::value)
+    {__t.swap(__u);}
+
+// get
+
+template <size_t _Ip, class ..._Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, tuple<_Tp...> >::type&
+get(tuple<_Tp...>& __t) _NOEXCEPT
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename tuple_element<_Ip, tuple<_Tp...> >::type type;
+    return static_cast<__tuple_leaf<_Ip, type>&>(__t.__base_).get();
+}
+
+template <size_t _Ip, class ..._Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, tuple<_Tp...> >::type&
+get(const tuple<_Tp...>& __t) _NOEXCEPT
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename tuple_element<_Ip, tuple<_Tp...> >::type type;
+    return static_cast<const __tuple_leaf<_Ip, type>&>(__t.__base_).get();
+}
+
+template <size_t _Ip, class ..._Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, tuple<_Tp...> >::type&&
+get(tuple<_Tp...>&& __t) _NOEXCEPT
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename tuple_element<_Ip, tuple<_Tp...> >::type type;
+    return static_cast<type&&>(
+             static_cast<__tuple_leaf<_Ip, type>&&>(__t.__base_).get());
+}
+
+template <size_t _Ip, class ..._Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, tuple<_Tp...> >::type&&
+get(const tuple<_Tp...>&& __t) _NOEXCEPT
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename tuple_element<_Ip, tuple<_Tp...> >::type type;
+    return static_cast<const type&&>(
+             static_cast<const __tuple_leaf<_Ip, type>&&>(__t.__base_).get());
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+
+namespace __find_detail {
+
+static constexpr size_t __not_found = ~size_t(0);
+static constexpr size_t __ambiguous = __not_found - 1;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr size_t __find_idx_return(size_t __curr_i, size_t __res, bool __matches) {
+    return !__matches ? __res :
+        (__res == __not_found ? __curr_i : __ambiguous);
+}
+
+template <size_t _Nx>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr size_t __find_idx(size_t __i, const bool (&__matches)[_Nx]) {
+  return __i == _Nx ? __not_found :
+      __find_idx_return(__i, __find_idx(__i + 1, __matches), __matches[__i]);
+}
+
+template <class _T1, class ..._Args>
+struct __find_exactly_one_checked {
+    static constexpr bool __matches[sizeof...(_Args)] = {is_same<_T1, _Args>::value...};
+    static constexpr size_t value = __find_detail::__find_idx(0, __matches);
+    static_assert(value != __not_found, "type not found in type list" );
+    static_assert(value != __ambiguous, "type occurs more than once in type list");
+};
+
+template <class _T1>
+struct __find_exactly_one_checked<_T1> {
+    static_assert(!is_same<_T1, _T1>::value, "type not in empty type list");
+};
+
+} // namespace __find_detail;
+
+template <typename _T1, typename... _Args>
+struct __find_exactly_one_t
+    : public __find_detail::__find_exactly_one_checked<_T1, _Args...> {
+};
+
+template <class _T1, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1& get(tuple<_Args...>& __tup) noexcept
+{
+    return _CUDA_VSTD::get<__find_exactly_one_t<_T1, _Args...>::value>(__tup);
+}
+
+template <class _T1, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 const& get(tuple<_Args...> const& __tup) noexcept
+{
+    return _CUDA_VSTD::get<__find_exactly_one_t<_T1, _Args...>::value>(__tup);
+}
+
+template <class _T1, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1&& get(tuple<_Args...>&& __tup) noexcept
+{
+    return _CUDA_VSTD::get<__find_exactly_one_t<_T1, _Args...>::value>(_CUDA_VSTD::move(__tup));
+}
+
+template <class _T1, class... _Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 const&& get(tuple<_Args...> const&& __tup) noexcept
+{
+    return _CUDA_VSTD::get<__find_exactly_one_t<_T1, _Args...>::value>(_CUDA_VSTD::move(__tup));
+}
+
+#endif
+
+// tie
+
+template <class ..._Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+tuple<_Tp&...>
+tie(_Tp&... __t) _NOEXCEPT
+{
+    return tuple<_Tp&...>(__t...);
+}
+
+template <class _Up>
+struct __ignore_t
+{
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const __ignore_t& operator=(_Tp&&) const {return *this;}
+};
+
+namespace {
+#ifdef __CUDA_ARCH__
+  _LIBCUDACXX_INLINE_VAR __constant__ __ignore_t<unsigned char> ignore = __ignore_t<unsigned char>();
+#else
+  _LIBCUDACXX_INLINE_VAR constexpr __ignore_t<unsigned char> ignore = __ignore_t<unsigned char>();
+#endif
+}
+
+template <class... _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+tuple<typename __unwrap_ref_decay<_Tp>::type...>
+make_tuple(_Tp&&... __t)
+{
+    return tuple<typename __unwrap_ref_decay<_Tp>::type...>(_CUDA_VSTD::forward<_Tp>(__t)...);
+}
+
+template <class... _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+tuple<_Tp&&...>
+forward_as_tuple(_Tp&&... __t) _NOEXCEPT
+{
+    return tuple<_Tp&&...>(_CUDA_VSTD::forward<_Tp>(__t)...);
+}
+
+template <size_t _Ip>
+struct __tuple_equal
+{
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _Tp& __x, const _Up& __y)
+    {
+        return __tuple_equal<_Ip - 1>()(__x, __y) && _CUDA_VSTD::get<_Ip-1>(__x) == _CUDA_VSTD::get<_Ip-1>(__y);
+    }
+};
+
+template <>
+struct __tuple_equal<0>
+{
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _Tp&, const _Up&)
+    {
+        return true;
+    }
+};
+
+template <class ..._Tp, class ..._Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator==(const tuple<_Tp...>& __x, const tuple<_Up...>& __y)
+{
+    static_assert (sizeof...(_Tp) == sizeof...(_Up), "Can't compare tuples of different sizes");
+    return __tuple_equal<sizeof...(_Tp)>()(__x, __y);
+}
+
+template <class ..._Tp, class ..._Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator!=(const tuple<_Tp...>& __x, const tuple<_Up...>& __y)
+{
+    return !(__x == __y);
+}
+
+template <size_t _Ip>
+struct __tuple_less
+{
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _Tp& __x, const _Up& __y)
+    {
+        const size_t __idx = tuple_size<_Tp>::value - _Ip;
+        if (_CUDA_VSTD::get<__idx>(__x) < _CUDA_VSTD::get<__idx>(__y))
+            return true;
+        if (_CUDA_VSTD::get<__idx>(__y) < _CUDA_VSTD::get<__idx>(__x))
+            return false;
+        return __tuple_less<_Ip-1>()(__x, __y);
+    }
+};
+
+template <>
+struct __tuple_less<0>
+{
+    template <class _Tp, class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    bool operator()(const _Tp&, const _Up&)
+    {
+        return false;
+    }
+};
+
+template <class ..._Tp, class ..._Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator<(const tuple<_Tp...>& __x, const tuple<_Up...>& __y)
+{
+    static_assert (sizeof...(_Tp) == sizeof...(_Up), "Can't compare tuples of different sizes");
+    return __tuple_less<sizeof...(_Tp)>()(__x, __y);
+}
+
+template <class ..._Tp, class ..._Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator>(const tuple<_Tp...>& __x, const tuple<_Up...>& __y)
+{
+    return __y < __x;
+}
+
+template <class ..._Tp, class ..._Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator>=(const tuple<_Tp...>& __x, const tuple<_Up...>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class ..._Tp, class ..._Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator<=(const tuple<_Tp...>& __x, const tuple<_Up...>& __y)
+{
+    return !(__y < __x);
+}
+
+// tuple_cat
+
+template <class _Tp, class _Up> struct __tuple_cat_type;
+
+template <class ..._Ttypes, class ..._Utypes>
+struct __tuple_cat_type<tuple<_Ttypes...>, __tuple_types<_Utypes...> >
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE tuple<_Ttypes..., _Utypes...> type;
+};
+
+template <class _ResultTuple, bool _Is_Tuple0TupleLike, class ..._Tuples>
+struct __tuple_cat_return_1
+{
+};
+
+template <class ..._Types, class _Tuple0>
+struct __tuple_cat_return_1<tuple<_Types...>, true, _Tuple0>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __tuple_cat_type<tuple<_Types...>,
+            typename __make_tuple_types<typename __uncvref<_Tuple0>::type>::type>::type
+                                                                           type;
+};
+
+template <class ..._Types, class _Tuple0, class _Tuple1, class ..._Tuples>
+struct __tuple_cat_return_1<tuple<_Types...>, true, _Tuple0, _Tuple1, _Tuples...>
+    : public __tuple_cat_return_1<
+                 typename __tuple_cat_type<
+                     tuple<_Types...>,
+                     typename __make_tuple_types<typename __uncvref<_Tuple0>::type>::type
+                 >::type,
+                 __tuple_like<typename remove_reference<_Tuple1>::type>::value,
+                 _Tuple1, _Tuples...>
+{
+};
+
+template <class ..._Tuples> struct __tuple_cat_return;
+
+template <class _Tuple0, class ..._Tuples>
+struct __tuple_cat_return<_Tuple0, _Tuples...>
+    : public __tuple_cat_return_1<tuple<>,
+         __tuple_like<typename remove_reference<_Tuple0>::type>::value, _Tuple0,
+                                                                     _Tuples...>
+{
+};
+
+template <>
+struct __tuple_cat_return<>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE tuple<> type;
+};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+tuple<>
+tuple_cat()
+{
+    return tuple<>();
+}
+
+template <class _Rp, class _Indices, class _Tuple0, class ..._Tuples>
+struct __tuple_cat_return_ref_imp;
+
+template <class ..._Types, size_t ..._I0, class _Tuple0>
+struct __tuple_cat_return_ref_imp<tuple<_Types...>, __tuple_indices<_I0...>, _Tuple0>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tuple0>::type _T0;
+    typedef tuple<_Types..., typename __apply_cv<_Tuple0,
+                          typename tuple_element<_I0, _T0>::type>::type&&...> type;
+};
+
+template <class ..._Types, size_t ..._I0, class _Tuple0, class _Tuple1, class ..._Tuples>
+struct __tuple_cat_return_ref_imp<tuple<_Types...>, __tuple_indices<_I0...>,
+                                  _Tuple0, _Tuple1, _Tuples...>
+    : public __tuple_cat_return_ref_imp<
+         tuple<_Types..., typename __apply_cv<_Tuple0,
+               typename tuple_element<_I0,
+                  typename remove_reference<_Tuple0>::type>::type>::type&&...>,
+         typename __make_tuple_indices<tuple_size<typename
+                                 remove_reference<_Tuple1>::type>::value>::type,
+         _Tuple1, _Tuples...>
+{
+};
+
+template <class _Tuple0, class ..._Tuples>
+struct __tuple_cat_return_ref
+    : public __tuple_cat_return_ref_imp<tuple<>,
+               typename __make_tuple_indices<
+                        tuple_size<typename remove_reference<_Tuple0>::type>::value
+               >::type, _Tuple0, _Tuples...>
+{
+};
+
+template <class _Types, class _I0, class _J0>
+struct __tuple_cat;
+
+template <class ..._Types, size_t ..._I0, size_t ..._J0>
+struct __tuple_cat<tuple<_Types...>, __tuple_indices<_I0...>, __tuple_indices<_J0...> >
+{
+    template <class _Tuple0>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    typename __tuple_cat_return_ref<tuple<_Types...>&&, _Tuple0&&>::type
+    operator()(tuple<_Types...> __t, _Tuple0&& __t0)
+    {
+        (void)__t;
+        return forward_as_tuple(_CUDA_VSTD::forward<_Types>(_CUDA_VSTD::get<_I0>(__t))...,
+                                      _CUDA_VSTD::get<_J0>(_CUDA_VSTD::forward<_Tuple0>(__t0))...);
+    }
+
+    template <class _Tuple0, class _Tuple1, class ..._Tuples>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    typename __tuple_cat_return_ref<tuple<_Types...>&&, _Tuple0&&, _Tuple1&&, _Tuples&&...>::type
+    operator()(tuple<_Types...> __t, _Tuple0&& __t0, _Tuple1&& __t1, _Tuples&& ...__tpls)
+    {
+        (void)__t;
+        typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tuple0>::type _T0;
+        typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tuple1>::type _T1;
+        return __tuple_cat<
+           tuple<_Types..., typename __apply_cv<_Tuple0, typename tuple_element<_J0, _T0>::type>::type&&...>,
+           typename __make_tuple_indices<sizeof ...(_Types) + tuple_size<_T0>::value>::type,
+           typename __make_tuple_indices<tuple_size<_T1>::value>::type>()
+                           (forward_as_tuple(
+                              _CUDA_VSTD::forward<_Types>(_CUDA_VSTD::get<_I0>(__t))...,
+                              _CUDA_VSTD::get<_J0>(_CUDA_VSTD::forward<_Tuple0>(__t0))...
+                            ),
+                            _CUDA_VSTD::forward<_Tuple1>(__t1),
+                            _CUDA_VSTD::forward<_Tuples>(__tpls)...);
+    }
+};
+
+template <class _Tuple0, class... _Tuples>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename __tuple_cat_return<_Tuple0, _Tuples...>::type
+tuple_cat(_Tuple0&& __t0, _Tuples&&... __tpls)
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tuple0>::type _T0;
+    return __tuple_cat<tuple<>, __tuple_indices<>,
+                  typename __make_tuple_indices<tuple_size<_T0>::value>::type>()
+                  (tuple<>(), _CUDA_VSTD::forward<_Tuple0>(__t0),
+                                            _CUDA_VSTD::forward<_Tuples>(__tpls)...);
+}
+
+template <class ..._Tp, class _Alloc>
+struct _LIBCUDACXX_TEMPLATE_VIS uses_allocator<tuple<_Tp...>, _Alloc>
+    : true_type {};
+
+template <class _T1, class _T2>
+template <class... _Args1, class... _Args2, size_t ..._I1, size_t ..._I2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+pair<_T1, _T2>::pair(piecewise_construct_t,
+                     tuple<_Args1...>& __first_args, tuple<_Args2...>& __second_args,
+                     __tuple_indices<_I1...>, __tuple_indices<_I2...>)
+    :  first(_CUDA_VSTD::forward<_Args1>(_CUDA_VSTD::get<_I1>( __first_args))...),
+      second(_CUDA_VSTD::forward<_Args2>(_CUDA_VSTD::get<_I2>(__second_args))...)
+{
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t tuple_size_v = tuple_size<_Tp>::value;
+
+#define _LIBCUDACXX_NOEXCEPT_RETURN(...) noexcept(noexcept(__VA_ARGS__)) { return __VA_ARGS__; }
+
+template <class _Fn, class _Tuple, size_t ..._Id>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr decltype(auto) __apply_tuple_impl(_Fn && __f, _Tuple && __t,
+                                            __tuple_indices<_Id...>)
+_LIBCUDACXX_NOEXCEPT_RETURN(
+    _CUDA_VSTD::__invoke_constexpr(
+        _CUDA_VSTD::forward<_Fn>(__f),
+        _CUDA_VSTD::get<_Id>(_CUDA_VSTD::forward<_Tuple>(__t))...)
+)
+
+template <class _Fn, class _Tuple>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr decltype(auto) apply(_Fn && __f, _Tuple && __t)
+_LIBCUDACXX_NOEXCEPT_RETURN(
+    _CUDA_VSTD::__apply_tuple_impl(
+        _CUDA_VSTD::forward<_Fn>(__f), _CUDA_VSTD::forward<_Tuple>(__t),
+        typename __make_tuple_indices<tuple_size_v<remove_reference_t<_Tuple>>>::type{})
+)
+
+template <class _Tp, class _Tuple, size_t... _Idx>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _Tp __make_from_tuple_impl(_Tuple&& __t, __tuple_indices<_Idx...>)
+_LIBCUDACXX_NOEXCEPT_RETURN(
+    _Tp(_CUDA_VSTD::get<_Idx>(_CUDA_VSTD::forward<_Tuple>(__t))...)
+)
+
+template <class _Tp, class _Tuple>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _Tp make_from_tuple(_Tuple&& __t)
+_LIBCUDACXX_NOEXCEPT_RETURN(
+    _CUDA_VSTD::__make_from_tuple_impl<_Tp>(_CUDA_VSTD::forward<_Tuple>(__t),
+        typename __make_tuple_indices<tuple_size_v<remove_reference_t<_Tuple>>>::type{})
+)
+
+#undef _LIBCUDACXX_NOEXCEPT_RETURN
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+#endif // !defined(_LIBCUDACXX_CXX03_LANG)
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_TUPLE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits
new file mode 100644
index 000000000000..9d1e6f991680
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/type_traits
@@ -0,0 +1,4898 @@
+// -*- C++ -*-
+//===------------------------ type_traits ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_TYPE_TRAITS
+#define _LIBCUDACXX_TYPE_TRAITS
+
+/*
+    type_traits synopsis
+
+namespace std
+{
+
+    // helper class:
+    template <class T, T v> struct integral_constant;
+    typedef integral_constant<bool, true>  true_type;   // C++11
+    typedef integral_constant<bool, false> false_type;  // C++11
+
+    template <bool B>                                   // C++14
+    using bool_constant = integral_constant<bool, B>;   // C++14
+    typedef bool_constant<true> true_type;              // C++14
+    typedef bool_constant<false> false_type;            // C++14
+
+    // helper traits
+    template <bool, class T = void> struct enable_if;
+    template <bool, class T, class F> struct conditional;
+
+    // Primary classification traits:
+    template <class T> struct is_void;
+    template <class T> struct is_null_pointer;  // C++14
+    template <class T> struct is_integral;
+    template <class T> struct is_floating_point;
+    template <class T> struct is_array;
+    template <class T> struct is_pointer;
+    template <class T> struct is_lvalue_reference;
+    template <class T> struct is_rvalue_reference;
+    template <class T> struct is_member_object_pointer;
+    template <class T> struct is_member_function_pointer;
+    template <class T> struct is_enum;
+    template <class T> struct is_union;
+    template <class T> struct is_class;
+    template <class T> struct is_function;
+
+    // Secondary classification traits:
+    template <class T> struct is_reference;
+    template <class T> struct is_arithmetic;
+    template <class T> struct is_fundamental;
+    template <class T> struct is_member_pointer;
+    template <class T> struct is_scalar;
+    template <class T> struct is_object;
+    template <class T> struct is_compound;
+
+    // Const-volatile properties and transformations:
+    template <class T> struct is_const;
+    template <class T> struct is_volatile;
+    template <class T> struct remove_const;
+    template <class T> struct remove_volatile;
+    template <class T> struct remove_cv;
+    template <class T> struct add_const;
+    template <class T> struct add_volatile;
+    template <class T> struct add_cv;
+
+    // Reference transformations:
+    template <class T> struct remove_reference;
+    template <class T> struct add_lvalue_reference;
+    template <class T> struct add_rvalue_reference;
+
+    // Pointer transformations:
+    template <class T> struct remove_pointer;
+    template <class T> struct add_pointer;
+
+    template<class T> struct type_identity;                     // C++20
+    template<class T>
+      using type_identity_t = typename type_identity<T>::type;  // C++20
+
+    // Integral properties:
+    template <class T> struct is_signed;
+    template <class T> struct is_unsigned;
+    template <class T> struct make_signed;
+    template <class T> struct make_unsigned;
+
+    // Array properties and transformations:
+    template <class T> struct rank;
+    template <class T, unsigned I = 0> struct extent;
+    template <class T> struct remove_extent;
+    template <class T> struct remove_all_extents;
+
+    template <class T> struct is_bounded_array;                 // C++20
+    template <class T> struct is_unbounded_array;               // C++20
+
+    // Member introspection:
+    template <class T> struct is_pod;
+    template <class T> struct is_trivial;
+    template <class T> struct is_trivially_copyable;
+    template <class T> struct is_standard_layout;
+    template <class T> struct is_literal_type;
+    template <class T> struct is_empty;
+    template <class T> struct is_polymorphic;
+    template <class T> struct is_abstract;
+    template <class T> struct is_final; // C++14
+    template <class T> struct is_aggregate; // C++17
+
+    template <class T, class... Args> struct is_constructible;
+    template <class T>                struct is_default_constructible;
+    template <class T>                struct is_copy_constructible;
+    template <class T>                struct is_move_constructible;
+    template <class T, class U>       struct is_assignable;
+    template <class T>                struct is_copy_assignable;
+    template <class T>                struct is_move_assignable;
+    template <class T, class U>       struct is_swappable_with;       // C++17
+    template <class T>                struct is_swappable;            // C++17
+    template <class T>                struct is_destructible;
+
+    template <class T, class... Args> struct is_trivially_constructible;
+    template <class T>                struct is_trivially_default_constructible;
+    template <class T>                struct is_trivially_copy_constructible;
+    template <class T>                struct is_trivially_move_constructible;
+    template <class T, class U>       struct is_trivially_assignable;
+    template <class T>                struct is_trivially_copy_assignable;
+    template <class T>                struct is_trivially_move_assignable;
+    template <class T>                struct is_trivially_destructible;
+
+    template <class T, class... Args> struct is_nothrow_constructible;
+    template <class T>                struct is_nothrow_default_constructible;
+    template <class T>                struct is_nothrow_copy_constructible;
+    template <class T>                struct is_nothrow_move_constructible;
+    template <class T, class U>       struct is_nothrow_assignable;
+    template <class T>                struct is_nothrow_copy_assignable;
+    template <class T>                struct is_nothrow_move_assignable;
+    template <class T, class U>       struct is_nothrow_swappable_with; // C++17
+    template <class T>                struct is_nothrow_swappable;      // C++17
+    template <class T>                struct is_nothrow_destructible;
+
+    template <class T> struct has_virtual_destructor;
+
+    template<class T> struct has_unique_object_representations;         // C++17
+
+    // Relationships between types:
+    template <class T, class U> struct is_same;
+    template <class Base, class Derived> struct is_base_of;
+
+    template <class From, class To> struct is_convertible;
+    template <typename From, typename To> struct is_nothrow_convertible;                  // C++20
+    template <typename From, typename To> inline constexpr bool is_nothrow_convertible_v; // C++20
+
+    template <class Fn, class... ArgTypes> struct is_invocable;
+    template <class R, class Fn, class... ArgTypes> struct is_invocable_r;
+
+    template <class Fn, class... ArgTypes> struct is_nothrow_invocable;
+    template <class R, class Fn, class... ArgTypes> struct is_nothrow_invocable_r;
+
+    // Alignment properties and transformations:
+    template <class T> struct alignment_of;
+    template <size_t Len, size_t Align = most_stringent_alignment_requirement>
+        struct aligned_storage;
+    template <size_t Len, class... Types> struct aligned_union;
+    template <class T> struct remove_cvref; // C++20
+
+    template <class T> struct decay;
+    template <class... T> struct common_type;
+    template <class T> struct underlying_type;
+    template <class> class result_of; // undefined
+    template <class Fn, class... ArgTypes> class result_of<Fn(ArgTypes...)>;
+    template <class Fn, class... ArgTypes> struct invoke_result;  // C++17
+
+    // const-volatile modifications:
+    template <class T>
+      using remove_const_t    = typename remove_const<T>::type;  // C++14
+    template <class T>
+      using remove_volatile_t = typename remove_volatile<T>::type;  // C++14
+    template <class T>
+      using remove_cv_t       = typename remove_cv<T>::type;  // C++14
+    template <class T>
+      using add_const_t       = typename add_const<T>::type;  // C++14
+    template <class T>
+      using add_volatile_t    = typename add_volatile<T>::type;  // C++14
+    template <class T>
+      using add_cv_t          = typename add_cv<T>::type;  // C++14
+
+    // reference modifications:
+    template <class T>
+      using remove_reference_t     = typename remove_reference<T>::type;  // C++14
+    template <class T>
+      using add_lvalue_reference_t = typename add_lvalue_reference<T>::type;  // C++14
+    template <class T>
+      using add_rvalue_reference_t = typename add_rvalue_reference<T>::type;  // C++14
+
+    // sign modifications:
+    template <class T>
+      using make_signed_t   = typename make_signed<T>::type;  // C++14
+    template <class T>
+      using make_unsigned_t = typename make_unsigned<T>::type;  // C++14
+
+    // array modifications:
+    template <class T>
+      using remove_extent_t      = typename remove_extent<T>::type;  // C++14
+    template <class T>
+      using remove_all_extents_t = typename remove_all_extents<T>::type;  // C++14
+
+    template <class T>
+      inline constexpr bool is_bounded_array_v
+        = is_bounded_array<T>::value;                                     // C++20
+      inline constexpr bool is_unbounded_array_v
+        = is_unbounded_array<T>::value;                                   // C++20
+
+    // pointer modifications:
+    template <class T>
+      using remove_pointer_t = typename remove_pointer<T>::type;  // C++14
+    template <class T>
+      using add_pointer_t    = typename add_pointer<T>::type;  // C++14
+
+    // other transformations:
+    template <size_t Len, std::size_t Align=default-alignment>
+      using aligned_storage_t = typename aligned_storage<Len,Align>::type;  // C++14
+    template <std::size_t Len, class... Types>
+      using aligned_union_t   = typename aligned_union<Len,Types...>::type;  // C++14
+    template <class T>
+      using remove_cvref_t    = typename remove_cvref<T>::type;  // C++20
+    template <class T>
+      using decay_t           = typename decay<T>::type;  // C++14
+    template <bool b, class T=void>
+      using enable_if_t       = typename enable_if<b,T>::type;  // C++14
+    template <bool b, class T, class F>
+      using conditional_t     = typename conditional<b,T,F>::type;  // C++14
+    template <class... T>
+      using common_type_t     = typename common_type<T...>::type;  // C++14
+    template <class T>
+      using underlying_type_t = typename underlying_type<T>::type;  // C++14
+    template <class T>
+      using result_of_t       = typename result_of<T>::type;  // C++14
+    template <class Fn, class... ArgTypes>
+      using invoke_result_t   = typename invoke_result<Fn, ArgTypes...>::type;  // C++17
+
+    template <class...>
+      using void_t = void;   // C++17
+
+      // See C++14 20.10.4.1, primary type categories
+      template <class T> inline constexpr bool is_void_v
+        = is_void<T>::value;                                             // C++17
+      template <class T> inline constexpr bool is_null_pointer_v
+        = is_null_pointer<T>::value;                                     // C++17
+      template <class T> inline constexpr bool is_integral_v
+        = is_integral<T>::value;                                         // C++17
+      template <class T> inline constexpr bool is_floating_point_v
+        = is_floating_point<T>::value;                                   // C++17
+      template <class T> inline constexpr bool is_array_v
+        = is_array<T>::value;                                            // C++17
+      template <class T> inline constexpr bool is_pointer_v
+        = is_pointer<T>::value;                                          // C++17
+      template <class T> inline constexpr bool is_lvalue_reference_v
+        = is_lvalue_reference<T>::value;                                 // C++17
+      template <class T> inline constexpr bool is_rvalue_reference_v
+        = is_rvalue_reference<T>::value;                                 // C++17
+      template <class T> inline constexpr bool is_member_object_pointer_v
+        = is_member_object_pointer<T>::value;                            // C++17
+      template <class T> inline constexpr bool is_member_function_pointer_v
+        = is_member_function_pointer<T>::value;                          // C++17
+      template <class T> inline constexpr bool is_enum_v
+        = is_enum<T>::value;                                             // C++17
+      template <class T> inline constexpr bool is_union_v
+        = is_union<T>::value;                                            // C++17
+      template <class T> inline constexpr bool is_class_v
+        = is_class<T>::value;                                            // C++17
+      template <class T> inline constexpr bool is_function_v
+        = is_function<T>::value;                                         // C++17
+
+      // See C++14 20.10.4.2, composite type categories
+      template <class T> inline constexpr bool is_reference_v
+        = is_reference<T>::value;                                        // C++17
+      template <class T> inline constexpr bool is_arithmetic_v
+        = is_arithmetic<T>::value;                                       // C++17
+      template <class T> inline constexpr bool is_fundamental_v
+        = is_fundamental<T>::value;                                      // C++17
+      template <class T> inline constexpr bool is_object_v
+        = is_object<T>::value;                                           // C++17
+      template <class T> inline constexpr bool is_scalar_v
+        = is_scalar<T>::value;                                           // C++17
+      template <class T> inline constexpr bool is_compound_v
+        = is_compound<T>::value;                                         // C++17
+      template <class T> inline constexpr bool is_member_pointer_v
+        = is_member_pointer<T>::value;                                   // C++17
+
+      // See C++14 20.10.4.3, type properties
+      template <class T> inline constexpr bool is_const_v
+        = is_const<T>::value;                                            // C++17
+      template <class T> inline constexpr bool is_volatile_v
+        = is_volatile<T>::value;                                         // C++17
+      template <class T> inline constexpr bool is_trivial_v
+        = is_trivial<T>::value;                                          // C++17
+      template <class T> inline constexpr bool is_trivially_copyable_v
+        = is_trivially_copyable<T>::value;                               // C++17
+      template <class T> inline constexpr bool is_standard_layout_v
+        = is_standard_layout<T>::value;                                  // C++17
+      template <class T> inline constexpr bool is_pod_v
+        = is_pod<T>::value;                                              // C++17
+      template <class T> inline constexpr bool is_literal_type_v
+        = is_literal_type<T>::value;                                     // C++17
+      template <class T> inline constexpr bool is_empty_v
+        = is_empty<T>::value;                                            // C++17
+      template <class T> inline constexpr bool is_polymorphic_v
+        = is_polymorphic<T>::value;                                      // C++17
+      template <class T> inline constexpr bool is_abstract_v
+        = is_abstract<T>::value;                                         // C++17
+      template <class T> inline constexpr bool is_final_v
+        = is_final<T>::value;                                            // C++17
+      template <class T> inline constexpr bool is_aggregate_v
+        = is_aggregate<T>::value;                                        // C++17
+      template <class T> inline constexpr bool is_signed_v
+        = is_signed<T>::value;                                           // C++17
+      template <class T> inline constexpr bool is_unsigned_v
+        = is_unsigned<T>::value;                                         // C++17
+      template <class T, class... Args> inline constexpr bool is_constructible_v
+        = is_constructible<T, Args...>::value;                           // C++17
+      template <class T> inline constexpr bool is_default_constructible_v
+        = is_default_constructible<T>::value;                            // C++17
+      template <class T> inline constexpr bool is_copy_constructible_v
+        = is_copy_constructible<T>::value;                               // C++17
+      template <class T> inline constexpr bool is_move_constructible_v
+        = is_move_constructible<T>::value;                               // C++17
+      template <class T, class U> inline constexpr bool is_assignable_v
+        = is_assignable<T, U>::value;                                    // C++17
+      template <class T> inline constexpr bool is_copy_assignable_v
+        = is_copy_assignable<T>::value;                                  // C++17
+      template <class T> inline constexpr bool is_move_assignable_v
+        = is_move_assignable<T>::value;                                  // C++17
+      template <class T, class U> inline constexpr bool is_swappable_with_v
+        = is_swappable_with<T, U>::value;                                // C++17
+      template <class T> inline constexpr bool is_swappable_v
+        = is_swappable<T>::value;                                        // C++17
+      template <class T> inline constexpr bool is_destructible_v
+        = is_destructible<T>::value;                                     // C++17
+      template <class T, class... Args> inline constexpr bool is_trivially_constructible_v
+        = is_trivially_constructible<T, Args...>::value;                 // C++17
+      template <class T> inline constexpr bool is_trivially_default_constructible_v
+        = is_trivially_default_constructible<T>::value;                  // C++17
+      template <class T> inline constexpr bool is_trivially_copy_constructible_v
+        = is_trivially_copy_constructible<T>::value;                     // C++17
+      template <class T> inline constexpr bool is_trivially_move_constructible_v
+        = is_trivially_move_constructible<T>::value;                     // C++17
+      template <class T, class U> inline constexpr bool is_trivially_assignable_v
+        = is_trivially_assignable<T, U>::value;                          // C++17
+      template <class T> inline constexpr bool is_trivially_copy_assignable_v
+        = is_trivially_copy_assignable<T>::value;                        // C++17
+      template <class T> inline constexpr bool is_trivially_move_assignable_v
+        = is_trivially_move_assignable<T>::value;                        // C++17
+      template <class T> inline constexpr bool is_trivially_destructible_v
+        = is_trivially_destructible<T>::value;                           // C++17
+      template <class T, class... Args> inline constexpr bool is_nothrow_constructible_v
+        = is_nothrow_constructible<T, Args...>::value;                   // C++17
+      template <class T> inline constexpr bool is_nothrow_default_constructible_v
+        = is_nothrow_default_constructible<T>::value;                    // C++17
+      template <class T> inline constexpr bool is_nothrow_copy_constructible_v
+        = is_nothrow_copy_constructible<T>::value;                       // C++17
+      template <class T> inline constexpr bool is_nothrow_move_constructible_v
+        = is_nothrow_move_constructible<T>::value;                       // C++17
+      template <class T, class U> inline constexpr bool is_nothrow_assignable_v
+        = is_nothrow_assignable<T, U>::value;                            // C++17
+      template <class T> inline constexpr bool is_nothrow_copy_assignable_v
+        = is_nothrow_copy_assignable<T>::value;                          // C++17
+      template <class T> inline constexpr bool is_nothrow_move_assignable_v
+        = is_nothrow_move_assignable<T>::value;                          // C++17
+      template <class T, class U> inline constexpr bool is_nothrow_swappable_with_v
+        = is_nothrow_swappable_with<T, U>::value;                       // C++17
+      template <class T> inline constexpr bool is_nothrow_swappable_v
+        = is_nothrow_swappable<T>::value;                               // C++17
+      template <class T> inline constexpr bool is_nothrow_destructible_v
+        = is_nothrow_destructible<T>::value;                             // C++17
+      template <class T> inline constexpr bool has_virtual_destructor_v
+        = has_virtual_destructor<T>::value;                              // C++17
+      template<class T> inline constexpr bool has_unique_object_representations_v // C++17
+        = has_unique_object_representations<T>::value;
+
+      // See C++14 20.10.5, type property queries
+      template <class T> inline constexpr size_t alignment_of_v
+        = alignment_of<T>::value;                                        // C++17
+      template <class T> inline constexpr size_t rank_v
+        = rank<T>::value;                                                // C++17
+      template <class T, unsigned I = 0> inline constexpr size_t extent_v
+        = extent<T, I>::value;                                           // C++17
+
+      // See C++14 20.10.6, type relations
+      template <class T, class U> inline constexpr bool is_same_v
+        = is_same<T, U>::value;                                          // C++17
+      template <class Base, class Derived> inline constexpr bool is_base_of_v
+        = is_base_of<Base, Derived>::value;                              // C++17
+      template <class From, class To> inline constexpr bool is_convertible_v
+        = is_convertible<From, To>::value;                               // C++17
+      template <class Fn, class... ArgTypes> inline constexpr bool is_invocable_v
+        = is_invocable<Fn, ArgTypes...>::value;                          // C++17
+      template <class R, class Fn, class... ArgTypes> inline constexpr bool is_invocable_r_v
+        = is_invocable_r<R, Fn, ArgTypes...>::value;                     // C++17
+      template <class Fn, class... ArgTypes> inline constexpr bool is_nothrow_invocable_v
+        = is_nothrow_invocable<Fn, ArgTypes...>::value;                  // C++17
+      template <class R, class Fn, class... ArgTypes> inline constexpr bool is_nothrow_invocable_r_v
+        = is_nothrow_invocable_r<R, Fn, ArgTypes...>::value;             // C++17
+
+      // [meta.logical], logical operator traits:
+      template<class... B> struct conjunction;                           // C++17
+      template<class... B>
+        inline constexpr bool conjunction_v = conjunction<B...>::value;  // C++17
+      template<class... B> struct disjunction;                           // C++17
+      template<class... B>
+        inline constexpr bool disjunction_v = disjunction<B...>::value;  // C++17
+      template<class B> struct negation;                                 // C++17
+      template<class B>
+        inline constexpr bool negation_v = negation<B>::value;           // C++17
+
+}
+
+*/
+#ifndef __cuda_std__
+#include <__config>
+#include <cstddef>
+#include <version>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _T1, class _T2> struct _LIBCUDACXX_TEMPLATE_VIS pair;
+template <class _Tp> class _LIBCUDACXX_TEMPLATE_VIS reference_wrapper;
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS hash;
+
+
+template <class _Tp, _Tp __v>
+struct _LIBCUDACXX_TEMPLATE_VIS integral_constant
+{
+  static _LIBCUDACXX_CONSTEXPR const _Tp      value = __v;
+  typedef _Tp               value_type;
+  typedef integral_constant type;
+  _LIBCUDACXX_INLINE_VISIBILITY
+  _LIBCUDACXX_CONSTEXPR operator value_type() const _NOEXCEPT {return value;}
+#if _LIBCUDACXX_STD_VER > 11
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr value_type operator ()() const _NOEXCEPT {return value;}
+#endif
+};
+
+template <class _Tp, _Tp __v>
+_LIBCUDACXX_CONSTEXPR const _Tp integral_constant<_Tp, __v>::value;
+
+#if _LIBCUDACXX_STD_VER > 11
+template <bool __b>
+using bool_constant = integral_constant<bool, __b>;
+#define _LIBCUDACXX_BOOL_CONSTANT(__b) bool_constant<(__b)>
+#else
+#define _LIBCUDACXX_BOOL_CONSTANT(__b) integral_constant<bool,(__b)>
+#endif
+
+typedef _LIBCUDACXX_BOOL_CONSTANT(true)  true_type;
+typedef _LIBCUDACXX_BOOL_CONSTANT(false) false_type;
+
+template <bool _Val>
+using _BoolConstant _LIBCUDACXX_NODEBUG_TYPE = integral_constant<bool, _Val>;
+
+template <bool> struct _MetaBase;
+template <>
+struct _MetaBase<true> {
+  template <class _Tp, class _Up>
+  using _SelectImpl _LIBCUDACXX_NODEBUG_TYPE = _Tp;
+  template <template <class...> class _FirstFn, template <class...> class, class ..._Args>
+  using _SelectApplyImpl _LIBCUDACXX_NODEBUG_TYPE = _FirstFn<_Args...>;
+  template <class _First, class...>
+  using _FirstImpl _LIBCUDACXX_NODEBUG_TYPE = _First;
+  template <class, class _Second, class...>
+  using _SecondImpl _LIBCUDACXX_NODEBUG_TYPE = _Second;
+  template <class _Tp = void>
+  using _EnableIfImpl _LIBCUDACXX_NODEBUG_TYPE = _Tp;
+  template <class _Result, class _First, class ..._Rest>
+  using _OrImpl _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase<_First::value != true && sizeof...(_Rest) != 0>::template _OrImpl<_First, _Rest...>;
+  template <class _Result, class _First, class ..._Rest>
+  using _AndImpl _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase<_First::value == true && sizeof...(_Rest) != 0>::template _AndImpl<_First, _Rest...>;
+};
+
+template <>
+struct _MetaBase<false> {
+  template <class _Tp, class _Up>
+  using _SelectImpl _LIBCUDACXX_NODEBUG_TYPE = _Up;
+  template <template <class...> class, template <class...> class _SecondFn, class ..._Args>
+  using _SelectApplyImpl _LIBCUDACXX_NODEBUG_TYPE = _SecondFn<_Args...>;
+  template <class _Result, class ...>
+  using _OrImpl _LIBCUDACXX_NODEBUG_TYPE = _Result;
+  template <class _Result, class ...>
+  using _AndImpl _LIBCUDACXX_NODEBUG_TYPE = _Result;
+};
+template <bool _Cond, class _Ret = void>
+using _EnableIf _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase<_Cond>::template _EnableIfImpl<_Ret>;
+template <bool _Cond, class _IfRes, class _ElseRes>
+using _If _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase<_Cond>::template _SelectImpl<_IfRes, _ElseRes>;
+template <class ..._Rest>
+using _Or _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase< sizeof...(_Rest) != 0 >::template _OrImpl<false_type, _Rest...>;
+template <class ..._Rest>
+using _And _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase< sizeof...(_Rest) != 0 >::template _AndImpl<true_type, _Rest...>;
+template <class _Pred>
+struct _Not : _BoolConstant<!_Pred::value> {};
+template <class ..._Args>
+using _FirstType _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase<(sizeof...(_Args) >= 1)>::template _FirstImpl<_Args...>;
+template <class ..._Args>
+using _SecondType _LIBCUDACXX_NODEBUG_TYPE = typename _MetaBase<(sizeof...(_Args) >= 2)>::template _SecondImpl<_Args...>;
+
+template <template <class...> class _Func, class ..._Args>
+struct _Lazy : _Func<_Args...> {};
+
+// Member detector base
+
+template <template <class...> class _Templ, class ..._Args, class = _Templ<_Args...> >
+_LIBCUDACXX_INLINE_VISIBILITY
+true_type __sfinae_test_impl(int);
+template <template <class...> class, class ...>
+_LIBCUDACXX_INLINE_VISIBILITY
+false_type __sfinae_test_impl(...);
+
+template <template <class ...> class _Templ, class ..._Args>
+using _IsValidExpansion _LIBCUDACXX_NODEBUG_TYPE = decltype(std::__sfinae_test_impl<_Templ, _Args...>(0));
+
+template <class>
+struct __void_t { typedef void type; };
+
+template <class _Tp>
+struct __identity { typedef _Tp type; };
+
+template <class _Tp, bool>
+struct _LIBCUDACXX_TEMPLATE_VIS __dependent_type : public _Tp {};
+
+template <bool _Bp, class _If, class _Then>
+    struct _LIBCUDACXX_TEMPLATE_VIS conditional {typedef _If type;};
+template <class _If, class _Then>
+    struct _LIBCUDACXX_TEMPLATE_VIS conditional<false, _If, _Then> {typedef _Then type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <bool _Bp, class _If, class _Then> using conditional_t = typename conditional<_Bp, _If, _Then>::type;
+#endif
+
+template <bool, class _Tp = void> struct _LIBCUDACXX_TEMPLATE_VIS enable_if {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS enable_if<true, _Tp> {typedef _Tp type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <bool _Bp, class _Tp = void> using enable_if_t = typename enable_if<_Bp, _Tp>::type;
+#endif
+
+// is_same
+
+template <class _Tp, class _Up> struct _LIBCUDACXX_TEMPLATE_VIS is_same           : public false_type {};
+template <class _Tp>            struct _LIBCUDACXX_TEMPLATE_VIS is_same<_Tp, _Tp> : public true_type {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_same_v
+    = is_same<_Tp, _Up>::value;
+#endif
+
+template <class _Tp, class _Up>
+using _IsSame = _BoolConstant<
+#if defined(_LIBCUDACXX_IS_SAME) && !defined(_LIBCUDACXX_USE_IS_SAME_FALLBACK)
+    _LIBCUDACXX_IS_SAME(_Tp, _Up)
+#else
+    _CUDA_VSTD::is_same<_Tp, _Up>::value
+#endif // defined(_LIBCUDACXX_IS_SAME) && !defined(_LIBCUDACXX_USE_IS_SAME_FALLBACK)
+>;
+
+template <class _Tp, class _Up>
+using _IsNotSame = _BoolConstant<
+#if defined(_LIBCUDACXX_IS_SAME) && !defined(_LIBCUDACXX_USE_IS_SAME_FALLBACK)
+    !_LIBCUDACXX_IS_SAME(_Tp, _Up)
+#else
+    !_CUDA_VSTD::is_same<_Tp, _Up>::value
+#endif // defined(_LIBCUDACXX_IS_SAME) && !defined(_LIBCUDACXX_USE_IS_SAME_FALLBACK)
+>;
+
+// addressof
+#if defined(_LIBCUDACXX_ADDRESSOF)
+
+template <class _Tp>
+inline _LIBCUDACXX_CONSTEXPR_AFTER_CXX14
+_LIBCUDACXX_NO_CFI _LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+addressof(_Tp& __x) _NOEXCEPT
+{
+    return _LIBCUDACXX_ADDRESSOF(__x);
+}
+
+#else
+
+template <class _Tp>
+inline _LIBCUDACXX_NO_CFI _LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+addressof(_Tp& __x) _NOEXCEPT
+{
+  return reinterpret_cast<_Tp *>(
+      const_cast<char *>(&reinterpret_cast<const volatile char &>(__x)));
+}
+
+#endif // defined(_LIBCUDACXX_ADDRESSOF)
+
+#if defined(_LIBCUDACXX_HAS_OBJC_ARC) && !defined(_LIBCUDACXX_PREDEFINED_OBJC_ARC_ADDRESSOF)
+// Objective-C++ Automatic Reference Counting uses qualified pointers
+// that require special addressof() signatures. When
+// _LIBCUDACXX_PREDEFINED_OBJC_ARC_ADDRESSOF is defined, the compiler
+// itself is providing these definitions. Otherwise, we provide them.
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__strong _Tp*
+addressof(__strong _Tp& __x) _NOEXCEPT
+{
+  return &__x;
+}
+
+#ifdef _LIBCUDACXX_HAS_OBJC_ARC_WEAK
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__weak _Tp*
+addressof(__weak _Tp& __x) _NOEXCEPT
+{
+  return &__x;
+}
+#endif
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__autoreleasing _Tp*
+addressof(__autoreleasing _Tp& __x) _NOEXCEPT
+{
+  return &__x;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__unsafe_unretained _Tp*
+addressof(__unsafe_unretained _Tp& __x) _NOEXCEPT
+{
+  return &__x;
+}
+#endif
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+template <class _Tp> _Tp* addressof(const _Tp&&) noexcept = delete;
+#endif
+
+struct __two {char __lx[2];};
+
+// helper class:
+
+// is_const
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_const            : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_const<_Tp const> : public true_type {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_const_v
+    = is_const<_Tp>::value;
+#endif
+
+// is_volatile
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_volatile               : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_volatile<_Tp volatile> : public true_type {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_volatile_v
+    = is_volatile<_Tp>::value;
+#endif
+
+// remove_const
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_const            {typedef _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_const<const _Tp> {typedef _Tp type;};
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_const_t = typename remove_const<_Tp>::type;
+#endif
+
+// remove_volatile
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_volatile               {typedef _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_volatile<volatile _Tp> {typedef _Tp type;};
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_volatile_t = typename remove_volatile<_Tp>::type;
+#endif
+
+// remove_cv
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_cv
+{typedef typename remove_volatile<typename remove_const<_Tp>::type>::type type;};
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_cv_t = typename remove_cv<_Tp>::type;
+#endif
+
+// is_void
+
+template <class _Tp> struct __libcpp_is_void       : public false_type {};
+template <>          struct __libcpp_is_void<void> : public true_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_void
+    : public __libcpp_is_void<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_void_v
+    = is_void<_Tp>::value;
+#endif
+
+// __is_nullptr_t
+
+template <class _Tp> struct __is_nullptr_t_impl       : public false_type {};
+template <>          struct __is_nullptr_t_impl<nullptr_t> : public true_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS __is_nullptr_t
+    : public __is_nullptr_t_impl<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_null_pointer
+    : public __is_nullptr_t_impl<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_null_pointer_v
+    = is_null_pointer<_Tp>::value;
+#endif
+#endif
+
+// is_integral
+
+template <class _Tp> struct __libcpp_is_integral                     : public false_type {};
+template <>          struct __libcpp_is_integral<bool>               : public true_type {};
+template <>          struct __libcpp_is_integral<char>               : public true_type {};
+template <>          struct __libcpp_is_integral<signed char>        : public true_type {};
+template <>          struct __libcpp_is_integral<unsigned char>      : public true_type {};
+template <>          struct __libcpp_is_integral<wchar_t>            : public true_type {};
+#ifndef _LIBCUDACXX_NO_HAS_CHAR8_T
+template <>          struct __libcpp_is_integral<char8_t>            : public true_type {};
+#endif
+#ifndef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+template <>          struct __libcpp_is_integral<char16_t>           : public true_type {};
+template <>          struct __libcpp_is_integral<char32_t>           : public true_type {};
+#endif  // _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+template <>          struct __libcpp_is_integral<short>              : public true_type {};
+template <>          struct __libcpp_is_integral<unsigned short>     : public true_type {};
+template <>          struct __libcpp_is_integral<int>                : public true_type {};
+template <>          struct __libcpp_is_integral<unsigned int>       : public true_type {};
+template <>          struct __libcpp_is_integral<long>               : public true_type {};
+template <>          struct __libcpp_is_integral<unsigned long>      : public true_type {};
+template <>          struct __libcpp_is_integral<long long>          : public true_type {};
+template <>          struct __libcpp_is_integral<unsigned long long> : public true_type {};
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+template <>          struct __libcpp_is_integral<__int128_t>         : public true_type {};
+template <>          struct __libcpp_is_integral<__uint128_t>        : public true_type {};
+#endif
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_integral
+    : public __libcpp_is_integral<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_integral_v
+    = is_integral<_Tp>::value;
+#endif
+
+// is_floating_point
+
+template <class _Tp> struct __libcpp_is_floating_point              : public false_type {};
+template <>          struct __libcpp_is_floating_point<float>       : public true_type {};
+template <>          struct __libcpp_is_floating_point<double>      : public true_type {};
+template <>          struct __libcpp_is_floating_point<long double> : public true_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_floating_point
+    : public __libcpp_is_floating_point<typename remove_cv<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_floating_point_v
+    = is_floating_point<_Tp>::value;
+#endif
+
+// is_array
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_array
+    : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_array<_Tp[]>
+    : public true_type {};
+template <class _Tp, size_t _Np> struct _LIBCUDACXX_TEMPLATE_VIS is_array<_Tp[_Np]>
+    : public true_type {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_array_v
+    = is_array<_Tp>::value;
+#endif
+
+// is_pointer
+
+template <class _Tp> struct __libcpp_is_pointer       : public false_type {};
+template <class _Tp> struct __libcpp_is_pointer<_Tp*> : public true_type {};
+
+template <class _Tp> struct __libcpp_remove_objc_qualifiers { typedef _Tp type; };
+#if defined(_LIBCUDACXX_HAS_OBJC_ARC)
+template <class _Tp> struct __libcpp_remove_objc_qualifiers<_Tp __strong> { typedef _Tp type; };
+template <class _Tp> struct __libcpp_remove_objc_qualifiers<_Tp __weak> { typedef _Tp type; };
+template <class _Tp> struct __libcpp_remove_objc_qualifiers<_Tp __autoreleasing> { typedef _Tp type; };
+template <class _Tp> struct __libcpp_remove_objc_qualifiers<_Tp __unsafe_unretained> { typedef _Tp type; };
+#endif
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_pointer
+    : public __libcpp_is_pointer<typename __libcpp_remove_objc_qualifiers<typename remove_cv<_Tp>::type>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_pointer_v
+    = is_pointer<_Tp>::value;
+#endif
+
+// is_reference
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_lvalue_reference       : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_lvalue_reference<_Tp&> : public true_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_rvalue_reference        : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_rvalue_reference<_Tp&&> : public true_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_reference        : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_reference<_Tp&>  : public true_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_reference<_Tp&&> : public true_type {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_reference_v
+    = is_reference<_Tp>::value;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_lvalue_reference_v
+    = is_lvalue_reference<_Tp>::value;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_rvalue_reference_v
+    = is_rvalue_reference<_Tp>::value;
+#endif
+// is_union
+
+#if defined(_LIBCUDACXX_IS_UNION) && !defined(_LIBCUDACXX_USE_IS_UNION_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_union
+    : public integral_constant<bool, _LIBCUDACXX_IS_UNION(_Tp)> {};
+
+#else
+
+template <class _Tp> struct __libcpp_union : public false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_union
+    : public __libcpp_union<typename remove_cv<_Tp>::type> {};
+
+#endif // defined(_LIBCUDACXX_IS_UNION) && !defined(_LIBCUDACXX_USE_IS_UNION_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_union_v
+    = is_union<_Tp>::value;
+#endif
+
+// is_class
+
+#if defined(_LIBCUDACXX_IS_CLASS) && !defined(_LIBCUDACXX_USE_IS_CLASS_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_class
+    : public integral_constant<bool, _LIBCUDACXX_IS_CLASS(_Tp)> {};
+
+#else
+
+namespace __is_class_imp
+{
+template <class _Tp> char  __test(int _Tp::*);
+template <class _Tp> __two __test(...);
+}
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_class
+    : public integral_constant<bool, sizeof(__is_class_imp::__test<_Tp>(0)) == 1 && !is_union<_Tp>::value> {};
+
+#endif // defined(_LIBCUDACXX_IS_CLASS) && !defined(_LIBCUDACXX_USE_IS_CLASS_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_class_v
+    = is_class<_Tp>::value;
+#endif
+
+// is_function
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_function
+    : public _BoolConstant<
+#if defined(_LIBCUDACXX_IS_FUNCTION) && !defined(_LIBCUDACXX_USE_IS_FUNCTION_FALLBACK)
+    _LIBCUDACXX_IS_FUNCTION(_Tp)
+#else
+    !(is_reference<_Tp>::value || is_const<const _Tp>::value)
+#endif // defined(_LIBCUDACXX_IS_FUNCTION) && !defined(_LIBCUDACXX_USE_IS_FUNCTION_FALLBACK)
+    > {};
+
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_function_v
+    = is_function<_Tp>::value;
+#endif
+
+template <class _Tp> struct __libcpp_is_member_pointer {
+  enum {
+    __is_member = false,
+    __is_func = false,
+    __is_obj = false
+  };
+};
+template <class _Tp, class _Up> struct __libcpp_is_member_pointer<_Tp _Up::*> {
+  enum {
+    __is_member = true,
+    __is_func = is_function<_Tp>::value,
+    __is_obj = !__is_func,
+  };
+};
+
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_member_function_pointer
+    : public _BoolConstant< __libcpp_is_member_pointer<typename remove_cv<_Tp>::type>::__is_func > {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_member_function_pointer_v
+    = is_member_function_pointer<_Tp>::value;
+#endif
+
+// is_member_pointer
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_member_pointer
+ : public _BoolConstant< __libcpp_is_member_pointer<typename remove_cv<_Tp>::type>::__is_member > {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_member_pointer_v
+    = is_member_pointer<_Tp>::value;
+#endif
+
+// is_member_object_pointer
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_member_object_pointer
+    : public _BoolConstant< __libcpp_is_member_pointer<typename remove_cv<_Tp>::type>::__is_obj >  {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_member_object_pointer_v
+    = is_member_object_pointer<_Tp>::value;
+#endif
+
+// is_enum
+
+#if defined(_LIBCUDACXX_IS_ENUM) && !defined(_LIBCUDACXX_USE_IS_ENUM_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_enum
+    : public integral_constant<bool, _LIBCUDACXX_IS_ENUM(_Tp)> {};
+
+#else
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_enum
+    : public integral_constant<bool, !is_void<_Tp>::value             &&
+                                     !is_integral<_Tp>::value         &&
+                                     !is_floating_point<_Tp>::value   &&
+                                     !is_array<_Tp>::value            &&
+                                     !is_pointer<_Tp>::value          &&
+                                     !is_reference<_Tp>::value        &&
+                                     !is_member_pointer<_Tp>::value   &&
+                                     !is_union<_Tp>::value            &&
+                                     !is_class<_Tp>::value            &&
+                                     !is_function<_Tp>::value         > {};
+
+#endif // defined(_LIBCUDACXX_IS_ENUM) && !defined(_LIBCUDACXX_USE_IS_ENUM_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_enum_v
+    = is_enum<_Tp>::value;
+#endif
+
+// is_arithmetic
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_arithmetic
+    : public integral_constant<bool, is_integral<_Tp>::value      ||
+                                     is_floating_point<_Tp>::value> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_arithmetic_v
+    = is_arithmetic<_Tp>::value;
+#endif
+
+// is_fundamental
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_fundamental
+    : public integral_constant<bool, is_void<_Tp>::value        ||
+                                     __is_nullptr_t<_Tp>::value ||
+                                     is_arithmetic<_Tp>::value> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_fundamental_v
+    = is_fundamental<_Tp>::value;
+#endif
+
+// is_scalar
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_scalar
+    : public integral_constant<bool, is_arithmetic<_Tp>::value     ||
+                                     is_member_pointer<_Tp>::value ||
+                                     is_pointer<_Tp>::value        ||
+                                     __is_nullptr_t<_Tp>::value    ||
+                                     is_enum<_Tp>::value           > {};
+
+template <> struct _LIBCUDACXX_TEMPLATE_VIS is_scalar<nullptr_t> : public true_type {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_scalar_v
+    = is_scalar<_Tp>::value;
+#endif
+
+// is_object
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_object
+    : public integral_constant<bool, is_scalar<_Tp>::value ||
+                                     is_array<_Tp>::value  ||
+                                     is_union<_Tp>::value  ||
+                                     is_class<_Tp>::value  > {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_object_v
+    = is_object<_Tp>::value;
+#endif
+
+// is_compound
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_compound
+    : public integral_constant<bool, !is_fundamental<_Tp>::value> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_compound_v
+    = is_compound<_Tp>::value;
+#endif
+
+
+// __is_referenceable  [defns.referenceable]
+
+struct __is_referenceable_impl {
+    template <class _Tp> _LIBCUDACXX_INLINE_VISIBILITY static _Tp& __test(int);
+    template <class _Tp> _LIBCUDACXX_INLINE_VISIBILITY static __two __test(...);
+};
+
+template <class _Tp>
+struct __is_referenceable : integral_constant<bool,
+    _IsNotSame<decltype(__is_referenceable_impl::__test<_Tp>(0)), __two>::value> {};
+
+
+// add_const
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS add_const {
+  typedef _LIBCUDACXX_NODEBUG_TYPE const _Tp type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using add_const_t = typename add_const<_Tp>::type;
+#endif
+
+// add_volatile
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS add_volatile {
+  typedef _LIBCUDACXX_NODEBUG_TYPE volatile _Tp type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using add_volatile_t = typename add_volatile<_Tp>::type;
+#endif
+
+// add_cv
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS add_cv {
+  typedef _LIBCUDACXX_NODEBUG_TYPE const volatile _Tp type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using add_cv_t = typename add_cv<_Tp>::type;
+#endif
+
+// remove_reference
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_reference        {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_reference<_Tp&>  {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_reference<_Tp&&> {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_reference_t = typename remove_reference<_Tp>::type;
+#endif
+
+// add_lvalue_reference
+
+template <class _Tp, bool = __is_referenceable<_Tp>::value> struct __add_lvalue_reference_impl            { typedef _LIBCUDACXX_NODEBUG_TYPE _Tp  type; };
+template <class _Tp                                       > struct __add_lvalue_reference_impl<_Tp, true> { typedef _LIBCUDACXX_NODEBUG_TYPE _Tp& type; };
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS add_lvalue_reference
+{typedef _LIBCUDACXX_NODEBUG_TYPE typename  __add_lvalue_reference_impl<_Tp>::type type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using add_lvalue_reference_t = typename add_lvalue_reference<_Tp>::type;
+#endif
+
+template <class _Tp, bool = __is_referenceable<_Tp>::value> struct __add_rvalue_reference_impl            { typedef _LIBCUDACXX_NODEBUG_TYPE  _Tp   type; };
+template <class _Tp                                       > struct __add_rvalue_reference_impl<_Tp, true> { typedef _LIBCUDACXX_NODEBUG_TYPE  _Tp&& type; };
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS add_rvalue_reference
+{typedef _LIBCUDACXX_NODEBUG_TYPE  typename __add_rvalue_reference_impl<_Tp>::type type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using add_rvalue_reference_t = typename add_rvalue_reference<_Tp>::type;
+#endif
+
+// Suppress deprecation notice for volatile-qualified return type resulting
+// from volatile-qualified types _Tp.
+_LIBCUDACXX_SUPPRESS_DEPRECATED_PUSH
+template <class _Tp> _LIBCUDACXX_INLINE_VISIBILITY _Tp&& __declval(int);
+template <class _Tp> _LIBCUDACXX_INLINE_VISIBILITY _Tp   __declval(long);
+_LIBCUDACXX_SUPPRESS_DEPRECATED_POP
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+decltype(_CUDA_VSTD::__declval<_Tp>(0))
+declval() _NOEXCEPT;
+
+// __uncvref
+
+template <class _Tp>
+struct __uncvref  {
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_cv<typename remove_reference<_Tp>::type>::type type;
+};
+
+template <class _Tp>
+struct __unconstref {
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_const<typename remove_reference<_Tp>::type>::type type;
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp>
+using __uncvref_t _LIBCUDACXX_NODEBUG_TYPE = typename __uncvref<_Tp>::type;
+#endif
+
+// __is_same_uncvref
+
+template <class _Tp, class _Up>
+struct __is_same_uncvref : _IsSame<typename __uncvref<_Tp>::type,
+                                   typename __uncvref<_Up>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11
+// remove_cvref - same as __uncvref
+template <class _Tp>
+struct remove_cvref : public __uncvref<_Tp> {};
+
+template <class _Tp> using remove_cvref_t = typename remove_cvref<_Tp>::type;
+#endif
+
+
+struct __any
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __any(...);
+};
+
+// remove_pointer
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_pointer                      {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_pointer<_Tp*>                {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_pointer<_Tp* const>          {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_pointer<_Tp* volatile>       {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_pointer<_Tp* const volatile> {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_pointer_t = typename remove_pointer<_Tp>::type;
+#endif
+
+// add_pointer
+
+template <class _Tp,
+        bool = __is_referenceable<_Tp>::value ||
+                _IsSame<typename remove_cv<_Tp>::type, void>::value>
+struct __add_pointer_impl
+    {typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tp>::type* type;};
+template <class _Tp> struct __add_pointer_impl<_Tp, false>
+    {typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type;};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS add_pointer
+    {typedef _LIBCUDACXX_NODEBUG_TYPE typename __add_pointer_impl<_Tp>::type type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using add_pointer_t = typename add_pointer<_Tp>::type;
+#endif
+
+// type_identity
+#if _LIBCUDACXX_STD_VER > 11
+template<class _Tp> struct type_identity { typedef _Tp type; };
+template<class _Tp> using type_identity_t = typename type_identity<_Tp>::type;
+#endif
+
+// is_signed
+
+template <class _Tp, bool = is_integral<_Tp>::value>
+struct __libcpp_is_signed_impl : public _LIBCUDACXX_BOOL_CONSTANT(_Tp(-1) < _Tp(0)) {};
+
+template <class _Tp>
+struct __libcpp_is_signed_impl<_Tp, false> : public true_type {};  // floating point
+
+template <class _Tp, bool = is_arithmetic<_Tp>::value>
+struct __libcpp_is_signed : public __libcpp_is_signed_impl<_Tp> {};
+
+template <class _Tp> struct __libcpp_is_signed<_Tp, false> : public false_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_signed : public __libcpp_is_signed<_Tp> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_signed_v
+    = is_signed<_Tp>::value;
+#endif
+
+// is_unsigned
+
+template <class _Tp, bool = is_integral<_Tp>::value>
+struct __libcpp_is_unsigned_impl : public _LIBCUDACXX_BOOL_CONSTANT(_Tp(0) < _Tp(-1)) {};
+
+template <class _Tp>
+struct __libcpp_is_unsigned_impl<_Tp, false> : public false_type {};  // floating point
+
+template <class _Tp, bool = is_arithmetic<_Tp>::value>
+struct __libcpp_is_unsigned : public __libcpp_is_unsigned_impl<_Tp> {};
+
+template <class _Tp> struct __libcpp_is_unsigned<_Tp, false> : public false_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_unsigned : public __libcpp_is_unsigned<_Tp> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_unsigned_v
+    = is_unsigned<_Tp>::value;
+#endif
+
+// rank
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS rank
+    : public integral_constant<size_t, 0> {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS rank<_Tp[]>
+    : public integral_constant<size_t, rank<_Tp>::value + 1> {};
+template <class _Tp, size_t _Np> struct _LIBCUDACXX_TEMPLATE_VIS rank<_Tp[_Np]>
+    : public integral_constant<size_t, rank<_Tp>::value + 1> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR size_t rank_v
+    = rank<_Tp>::value;
+#endif
+
+// extent
+
+template <class _Tp, unsigned _Ip = 0> struct _LIBCUDACXX_TEMPLATE_VIS extent
+    : public integral_constant<size_t, 0> {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS extent<_Tp[], 0>
+    : public integral_constant<size_t, 0> {};
+template <class _Tp, unsigned _Ip> struct _LIBCUDACXX_TEMPLATE_VIS extent<_Tp[], _Ip>
+    : public integral_constant<size_t, extent<_Tp, _Ip-1>::value> {};
+template <class _Tp, size_t _Np> struct _LIBCUDACXX_TEMPLATE_VIS extent<_Tp[_Np], 0>
+    : public integral_constant<size_t, _Np> {};
+template <class _Tp, size_t _Np, unsigned _Ip> struct _LIBCUDACXX_TEMPLATE_VIS extent<_Tp[_Np], _Ip>
+    : public integral_constant<size_t, extent<_Tp, _Ip-1>::value> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, unsigned _Ip = 0>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR size_t extent_v
+    = extent<_Tp, _Ip>::value;
+#endif
+
+// remove_extent
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_extent
+    {typedef _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_extent<_Tp[]>
+    {typedef _Tp type;};
+template <class _Tp, size_t _Np> struct _LIBCUDACXX_TEMPLATE_VIS remove_extent<_Tp[_Np]>
+    {typedef _Tp type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_extent_t = typename remove_extent<_Tp>::type;
+#endif
+
+// remove_all_extents
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_all_extents
+    {typedef _Tp type;};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS remove_all_extents<_Tp[]>
+    {typedef typename remove_all_extents<_Tp>::type type;};
+template <class _Tp, size_t _Np> struct _LIBCUDACXX_TEMPLATE_VIS remove_all_extents<_Tp[_Np]>
+    {typedef typename remove_all_extents<_Tp>::type type;};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using remove_all_extents_t = typename remove_all_extents<_Tp>::type;
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11
+// is_bounded_array
+
+template <class>                 struct _LIBCUDACXX_TEMPLATE_VIS is_bounded_array           : false_type {};
+template <class _Tp, size_t _Np> struct _LIBCUDACXX_TEMPLATE_VIS is_bounded_array<_Tp[_Np]> : true_type {};
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR
+bool is_bounded_array_v  = is_bounded_array<_Tp>::value;
+
+// is_unbounded_array
+
+template <class>     struct _LIBCUDACXX_TEMPLATE_VIS is_unbounded_array        : false_type {};
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_unbounded_array<_Tp[]> : true_type {};
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR
+bool is_unbounded_array_v  = is_unbounded_array<_Tp>::value;
+#endif
+
+// decay
+
+template <class _Up, bool>
+struct __decay {
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_cv<_Up>::type type;
+};
+
+template <class _Up>
+struct __decay<_Up, true> {
+public:
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename conditional
+                     <
+                         is_array<_Up>::value,
+                         typename remove_extent<_Up>::type*,
+                         typename conditional
+                         <
+                              is_function<_Up>::value,
+                              typename add_pointer<_Up>::type,
+                              typename remove_cv<_Up>::type
+                         >::type
+                     >::type type;
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS decay
+{
+private:
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tp>::type _Up;
+public:
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __decay<_Up, __is_referenceable<_Up>::value>::type type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using decay_t = typename decay<_Tp>::type;
+#endif
+
+// is_abstract
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_abstract
+    : public integral_constant<bool, __is_abstract(_Tp)> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_abstract_v
+    = is_abstract<_Tp>::value;
+#endif
+
+// is_final
+
+#if defined(_LIBCUDACXX_IS_FINAL)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS
+__libcpp_is_final : public integral_constant<bool, _LIBCUDACXX_IS_FINAL(_Tp)> {};
+
+#else
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS
+__libcpp_is_final : public false_type {};
+
+#endif // defined(_LIBCUDACXX_IS_FINAL)
+
+#if defined(_LIBCUDACXX_IS_FINAL)  && _LIBCUDACXX_STD_VER > 11
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS
+is_final : public integral_constant<bool, _LIBCUDACXX_IS_FINAL(_Tp)> {};
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_final_v
+    = is_final<_Tp>::value;
+#endif
+
+// is_aggregate
+#if _LIBCUDACXX_STD_VER > 11 && defined(_LIBCUDACXX_IS_AGGREGATE)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS
+is_aggregate : public integral_constant<bool, _LIBCUDACXX_IS_AGGREGATE(_Tp)> {};
+
+#if !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_aggregate_v
+    = is_aggregate<_Tp>::value;
+#endif
+
+#endif // _LIBCUDACXX_STD_VER > 11 && defined(_LIBCUDACXX_IS_AGGREGATE)
+
+// is_base_of
+
+#if defined(_LIBCUDACXX_IS_BASE_OF) && !defined(_LIBCUDACXX_USE_IS_BASE_OF_FALLBACK)
+
+template <class _Bp, class _Dp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_base_of
+    : public integral_constant<bool, _LIBCUDACXX_IS_BASE_OF(_Bp, _Dp)> {};
+
+#else  // defined(_LIBCUDACXX_IS_BASE_OF) && !defined(_LIBCUDACXX_USE_IS_BASE_OF_FALLBACK)
+
+namespace __is_base_of_imp
+{
+template <class _Tp>
+struct _Dst
+{
+    _Dst(const volatile _Tp &);
+};
+template <class _Tp>
+struct _Src
+{
+    operator const volatile _Tp &();
+    template <class _Up> operator const _Dst<_Up> &();
+};
+template <size_t> struct __one { typedef char type; };
+template <class _Bp, class _Dp> typename __one<sizeof(_Dst<_Bp>(declval<_Src<_Dp> >()))>::type __test(int);
+template <class _Bp, class _Dp> __two __test(...);
+}
+
+template <class _Bp, class _Dp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_base_of
+    : public integral_constant<bool, is_class<_Bp>::value &&
+                                     sizeof(__is_base_of_imp::__test<_Bp, _Dp>(0)) == 2> {};
+
+#endif  // defined(_LIBCUDACXX_IS_BASE_OF) && !defined(_LIBCUDACXX_USE_IS_BASE_OF_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Bp, class _Dp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_base_of_v
+    = is_base_of<_Bp, _Dp>::value;
+#endif
+
+// is_convertible
+
+#if defined(_LIBCUDACXX_IS_CONVERTIBLE_TO) && !defined(_LIBCUDACXX_USE_IS_CONVERTIBLE_FALLBACK)
+
+template <class _T1, class _T2> struct _LIBCUDACXX_TEMPLATE_VIS is_convertible
+    : public integral_constant<bool, _LIBCUDACXX_IS_CONVERTIBLE_TO(_T1, _T2) &&
+                                     !is_abstract<_T2>::value> {};
+
+#else  // defined(_LIBCUDACXX_IS_CONVERTIBLE_TO) && !defined(_LIBCUDACXX_USE_IS_CONVERTIBLE_FALLBACK)
+
+namespace __is_convertible_imp
+{
+template <class _Tp> _LIBCUDACXX_INLINE_VISIBILITY void  __test_convert(_Tp);
+
+template <class _From, class _To, class = void>
+struct __is_convertible_test : public false_type {};
+
+template <class _From, class _To>
+struct __is_convertible_test<_From, _To,
+    decltype(_CUDA_VSTD::__is_convertible_imp::__test_convert<_To>(_CUDA_VSTD::declval<_From>()))> : public true_type
+{};
+
+template <class _Tp, bool _IsArray =    is_array<_Tp>::value,
+                     bool _IsFunction = is_function<_Tp>::value,
+                     bool _IsVoid =     is_void<_Tp>::value>
+                     struct __is_array_function_or_void                          {enum {value = 0};};
+template <class _Tp> struct __is_array_function_or_void<_Tp, true, false, false> {enum {value = 1};};
+template <class _Tp> struct __is_array_function_or_void<_Tp, false, true, false> {enum {value = 2};};
+template <class _Tp> struct __is_array_function_or_void<_Tp, false, false, true> {enum {value = 3};};
+}
+
+template <class _Tp,
+    unsigned = __is_convertible_imp::__is_array_function_or_void<typename remove_reference<_Tp>::type>::value>
+struct __is_convertible_check
+{
+    static const size_t __v = 0;
+};
+
+template <class _Tp>
+struct __is_convertible_check<_Tp, 0>
+{
+    static const size_t __v = sizeof(_Tp);
+};
+
+template <class _T1, class _T2,
+    unsigned _T1_is_array_function_or_void = __is_convertible_imp::__is_array_function_or_void<_T1>::value,
+    unsigned _T2_is_array_function_or_void = __is_convertible_imp::__is_array_function_or_void<_T2>::value>
+struct __is_convertible_fallback
+    : public integral_constant<bool,
+        __is_convertible_imp::__is_convertible_test<_T1, _T2>::value
+    >
+{};
+
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 0, 1> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 1, 1> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 2, 1> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 3, 1> : public false_type {};
+
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 0, 2> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 1, 2> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 2, 2> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 3, 2> : public false_type {};
+
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 0, 3> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 1, 3> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 2, 3> : public false_type {};
+template <class _T1, class _T2> struct __is_convertible_fallback<_T1, _T2, 3, 3> : public true_type {};
+
+template <class _T1, class _T2> struct _LIBCUDACXX_TEMPLATE_VIS is_convertible
+    : public __is_convertible_fallback<_T1, _T2>
+{
+    static const size_t __complete_check1 = __is_convertible_check<_T1>::__v;
+    static const size_t __complete_check2 = __is_convertible_check<_T2>::__v;
+};
+
+#endif // defined(_LIBCUDACXX_IS_CONVERTIBLE_TO) && !defined(_LIBCUDACXX_USE_IS_CONVERTIBLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _From, class _To>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_convertible_v
+    = is_convertible<_From, _To>::value;
+#endif
+
+// is_nothrow_convertible
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template <typename _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY static void __test_noexcept(_Tp) noexcept;
+
+template<typename _Fm, typename _To>
+_LIBCUDACXX_INLINE_VISIBILITY static bool_constant<noexcept(__test_noexcept<_To>(declval<_Fm>()))>
+__is_nothrow_convertible_test();
+
+template <typename _Fm, typename _To>
+struct __is_nothrow_convertible_helper: decltype(__is_nothrow_convertible_test<_Fm, _To>())
+{ };
+
+template <typename _Fm, typename _To>
+struct is_nothrow_convertible : _Or<
+    _And<is_void<_To>, is_void<_Fm>>,
+    _Lazy<_And, is_convertible<_Fm, _To>, __is_nothrow_convertible_helper<_Fm, _To>>
+>::type { };
+
+template <typename _Fm, typename _To>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_nothrow_convertible_v = is_nothrow_convertible<_Fm, _To>::value;
+
+#endif // _LIBCUDACXX_STD_VER > 11
+
+// is_empty
+
+#if defined(_LIBCUDACXX_IS_EMPTY) && !defined(_LIBCUDACXX_USE_IS_EMPTY_FALLBACK)
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_empty
+    : public integral_constant<bool, _LIBCUDACXX_IS_EMPTY(_Tp)> {};
+
+#else
+
+template <class _Tp>
+struct __is_empty1
+    : public _Tp
+{
+    double __lx;
+};
+
+struct __is_empty2
+{
+    double __lx;
+};
+
+template <class _Tp, bool = is_class<_Tp>::value>
+struct __libcpp_empty : public integral_constant<bool, sizeof(__is_empty1<_Tp>) == sizeof(__is_empty2)> {};
+
+template <class _Tp> struct __libcpp_empty<_Tp, false> : public false_type {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_empty : public __libcpp_empty<_Tp> {};
+
+#endif // defined(_LIBCUDACXX_IS_EMPTY) && !defined(_LIBCUDACXX_USE_IS_EMPTY_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_empty_v
+    = is_empty<_Tp>::value;
+#endif
+
+// is_polymorphic
+
+#if defined(_LIBCUDACXX_IS_POLYMORPHIC) && !defined(_LIBCUDACXX_USE_IS_POLYMORPHIC_FALLBACK)
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_polymorphic
+    : public integral_constant<bool, _LIBCUDACXX_IS_POLYMORPHIC(_Tp)> {};
+
+#else
+
+template<typename _Tp> char &__is_polymorphic_impl(
+    typename enable_if<sizeof((_Tp*)dynamic_cast<const volatile void*>(declval<_Tp*>())) != 0,
+                       int>::type);
+template<typename _Tp> __two &__is_polymorphic_impl(...);
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_polymorphic
+    : public integral_constant<bool, sizeof(__is_polymorphic_impl<_Tp>(0)) == 1> {};
+
+#endif // defined(_LIBCUDACXX_IS_POLYMORPHIC) && !defined(_LIBCUDACXX_USE_IS_POLYMORPHIC_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_polymorphic_v
+    = is_polymorphic<_Tp>::value;
+#endif
+
+// has_virtual_destructor
+
+#if defined(_LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_VIRTUAL_DESTRUCTOR_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS has_virtual_destructor
+    : public integral_constant<bool, _LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR(_Tp)> {};
+
+#else
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS has_virtual_destructor
+    : public false_type {};
+
+#endif // defined(_LIBCUDACXX_HAS_VIRTUAL_DESTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_VIRTUAL_DESTRUCTOR_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool has_virtual_destructor_v
+    = has_virtual_destructor<_Tp>::value;
+#endif
+
+// has_unique_object_representations
+
+#if _LIBCUDACXX_STD_VER > 11 && defined(_LIBCUDACXX_HAS_UNIQUE_OBJECT_REPRESENTATIONS)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS has_unique_object_representations
+    : public integral_constant<bool,
+       _LIBCUDACXX_HAS_UNIQUE_OBJECT_REPRESENTATIONS(remove_cv_t<remove_all_extents_t<_Tp>>)> {};
+
+#if !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool has_unique_object_representations_v
+    = has_unique_object_representations<_Tp>::value;
+#endif
+
+#endif // _LIBCUDACXX_STD_VER > 11 && defined(_LIBCUDACXX_HAS_UNIQUE_OBJECT_REPRESENTATIONS)
+
+// alignment_of
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS alignment_of
+    : public integral_constant<size_t, _LIBCUDACXX_ALIGNOF(_Tp)> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR size_t alignment_of_v
+    = alignment_of<_Tp>::value;
+#endif
+
+// aligned_storage
+
+template <class _Hp, class _Tp>
+struct __type_list
+{
+    typedef _Hp _Head;
+    typedef _Tp _Tail;
+};
+
+struct __nat
+{
+#ifndef _LIBCUDACXX_CXX03_LANG
+    __nat() = delete;
+    __nat(const __nat&) = delete;
+    __nat& operator=(const __nat&) = delete;
+    ~__nat() = delete;
+#endif
+};
+
+template <class _Tp>
+struct __align_type
+{
+    static const size_t value = _LIBCUDACXX_PREFERRED_ALIGNOF(_Tp);
+    typedef _Tp type;
+};
+
+struct __struct_double {long double __lx;};
+struct __struct_double4 {double __lx[4];};
+
+typedef
+    __type_list<__align_type<unsigned char>,
+    __type_list<__align_type<unsigned short>,
+    __type_list<__align_type<unsigned int>,
+    __type_list<__align_type<unsigned long>,
+    __type_list<__align_type<unsigned long long>,
+    __type_list<__align_type<double>,
+    __type_list<__align_type<long double>,
+    __type_list<__align_type<__struct_double>,
+    __type_list<__align_type<__struct_double4>,
+    __type_list<__align_type<int*>,
+    __nat
+    > > > > > > > > > > __all_types;
+
+template <size_t _Align>
+struct _ALIGNAS(_Align) __fallback_overaligned {};
+
+template <class _TL, size_t _Align> struct __find_pod;
+
+template <class _Hp, size_t _Align>
+struct __find_pod<__type_list<_Hp, __nat>, _Align>
+{
+    typedef typename conditional<
+                             _Align == _Hp::value,
+                             typename _Hp::type,
+                             __fallback_overaligned<_Align>
+                         >::type type;
+};
+
+template <class _Hp, class _Tp, size_t _Align>
+struct __find_pod<__type_list<_Hp, _Tp>, _Align>
+{
+    typedef typename conditional<
+                             _Align == _Hp::value,
+                             typename _Hp::type,
+                             typename __find_pod<_Tp, _Align>::type
+                         >::type type;
+};
+
+template <class _TL, size_t _Len> struct __find_max_align;
+
+template <class _Hp, size_t _Len>
+struct __find_max_align<__type_list<_Hp, __nat>, _Len> : public integral_constant<size_t, _Hp::value> {};
+
+template <size_t _Len, size_t _A1, size_t _A2>
+struct __select_align
+{
+private:
+    static const size_t __min = _A2 < _A1 ? _A2 : _A1;
+    static const size_t __max = _A1 < _A2 ? _A2 : _A1;
+public:
+    static const size_t value = _Len < __max ? __min : __max;
+};
+
+template <class _Hp, class _Tp, size_t _Len>
+struct __find_max_align<__type_list<_Hp, _Tp>, _Len>
+    : public integral_constant<size_t, __select_align<_Len, _Hp::value, __find_max_align<_Tp, _Len>::value>::value> {};
+
+template <size_t _Len, size_t _Align = __find_max_align<__all_types, _Len>::value>
+struct _LIBCUDACXX_TEMPLATE_VIS aligned_storage
+{
+    typedef typename __find_pod<__all_types, _Align>::type _Aligner;
+    union type
+    {
+        _Aligner __align;
+        unsigned char __data[(_Len + _Align - 1)/_Align * _Align];
+    };
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <size_t _Len, size_t _Align = __find_max_align<__all_types, _Len>::value>
+    using aligned_storage_t = typename aligned_storage<_Len, _Align>::type;
+#endif
+
+#define _CREATE_ALIGNED_STORAGE_SPECIALIZATION(n) \
+template <size_t _Len>\
+struct _LIBCUDACXX_TEMPLATE_VIS aligned_storage<_Len, n>\
+{\
+    struct _ALIGNAS(n) type\
+    {\
+        unsigned char __lx[(_Len + n - 1)/n * n];\
+    };\
+}
+
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x1);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x2);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x4);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x8);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x10);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x20);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x40);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x80);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x100);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x200);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x400);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x800);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x1000);
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x2000);
+// PE/COFF does not support alignment beyond 8192 (=0x2000)
+#if !defined(_LIBCUDACXX_OBJECT_FORMAT_COFF)
+_CREATE_ALIGNED_STORAGE_SPECIALIZATION(0x4000);
+#endif // !defined(_LIBCUDACXX_OBJECT_FORMAT_COFF)
+
+#undef _CREATE_ALIGNED_STORAGE_SPECIALIZATION
+
+
+// aligned_union
+
+template <size_t _I0, size_t ..._In>
+struct __static_max;
+
+template <size_t _I0>
+struct __static_max<_I0>
+{
+    static const size_t value = _I0;
+};
+
+template <size_t _I0, size_t _I1, size_t ..._In>
+struct __static_max<_I0, _I1, _In...>
+{
+    static const size_t value = _I0 >= _I1 ? __static_max<_I0, _In...>::value :
+                                             __static_max<_I1, _In...>::value;
+};
+
+template <size_t _Len, class _Type0, class ..._Types>
+struct aligned_union
+{
+    static const size_t alignment_value = __static_max<_LIBCUDACXX_PREFERRED_ALIGNOF(_Type0),
+                                                       _LIBCUDACXX_PREFERRED_ALIGNOF(_Types)...>::value;
+    static const size_t __len = __static_max<_Len, sizeof(_Type0),
+                                             sizeof(_Types)...>::value;
+    typedef typename aligned_storage<__len, alignment_value>::type type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <size_t _Len, class ..._Types> using aligned_union_t = typename aligned_union<_Len, _Types...>::type;
+#endif
+
+template <class _Tp>
+struct __numeric_type
+{
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static void __test(...);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static float __test(float);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(char);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(int);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(unsigned);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(long);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(unsigned long);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(long long);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(unsigned long long);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static double __test(double);
+   _LIBCUDACXX_INLINE_VISIBILITY
+   static long double __test(long double);
+
+   typedef decltype(__test(declval<_Tp>())) type;
+   static const bool value = _IsNotSame<type, void>::value;
+};
+
+template <>
+struct __numeric_type<void>
+{
+   static const bool value = true;
+};
+
+// __promote
+
+template <class _A1, class _A2 = void, class _A3 = void,
+          bool = __numeric_type<_A1>::value &&
+                 __numeric_type<_A2>::value &&
+                 __numeric_type<_A3>::value>
+class __promote_imp
+{
+public:
+    static const bool value = false;
+};
+
+template <class _A1, class _A2, class _A3>
+class __promote_imp<_A1, _A2, _A3, true>
+{
+private:
+    typedef typename __promote_imp<_A1>::type __type1;
+    typedef typename __promote_imp<_A2>::type __type2;
+    typedef typename __promote_imp<_A3>::type __type3;
+public:
+    typedef decltype(__type1() + __type2() + __type3()) type;
+    static const bool value = true;
+};
+
+template <class _A1, class _A2>
+class __promote_imp<_A1, _A2, void, true>
+{
+private:
+    typedef typename __promote_imp<_A1>::type __type1;
+    typedef typename __promote_imp<_A2>::type __type2;
+public:
+    typedef decltype(__type1() + __type2()) type;
+    static const bool value = true;
+};
+
+template <class _A1>
+class __promote_imp<_A1, void, void, true>
+{
+public:
+    typedef typename __numeric_type<_A1>::type type;
+    static const bool value = true;
+};
+
+template <class _A1, class _A2 = void, class _A3 = void>
+class __promote : public __promote_imp<_A1, _A2, _A3> {};
+
+// make_signed / make_unsigned
+
+typedef
+    __type_list<signed char,
+    __type_list<signed short,
+    __type_list<signed int,
+    __type_list<signed long,
+    __type_list<signed long long,
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+    __type_list<__int128_t,
+#endif
+    __nat
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+    >
+#endif
+    > > > > > __signed_types;
+
+typedef
+    __type_list<unsigned char,
+    __type_list<unsigned short,
+    __type_list<unsigned int,
+    __type_list<unsigned long,
+    __type_list<unsigned long long,
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+    __type_list<__uint128_t,
+#endif
+    __nat
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+    >
+#endif
+    > > > > > __unsigned_types;
+
+template <class _TypeList, size_t _Size, bool = _Size <= sizeof(typename _TypeList::_Head)> struct __find_first;
+
+template <class _Hp, class _Tp, size_t _Size>
+struct __find_first<__type_list<_Hp, _Tp>, _Size, true>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Hp type;
+};
+
+template <class _Hp, class _Tp, size_t _Size>
+struct __find_first<__type_list<_Hp, _Tp>, _Size, false>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename __find_first<_Tp, _Size>::type type;
+};
+
+template <class _Tp, class _Up, bool = is_const<typename remove_reference<_Tp>::type>::value,
+                             bool = is_volatile<typename remove_reference<_Tp>::type>::value>
+struct __apply_cv
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _Up type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp, _Up, true, false>
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE const _Up type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp, _Up, false, true>
+{
+    typedef volatile _Up type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp, _Up, true, true>
+{
+    typedef const volatile _Up type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp&, _Up, false, false>
+{
+    typedef _Up& type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp&, _Up, true, false>
+{
+    typedef const _Up& type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp&, _Up, false, true>
+{
+    typedef volatile _Up& type;
+};
+
+template <class _Tp, class _Up>
+struct __apply_cv<_Tp&, _Up, true, true>
+{
+    typedef const volatile _Up& type;
+};
+
+template <class _Tp, bool = is_integral<_Tp>::value || is_enum<_Tp>::value>
+struct __make_signed {};
+
+template <class _Tp>
+struct __make_signed<_Tp, true>
+{
+    typedef typename __find_first<__signed_types, sizeof(_Tp)>::type type;
+};
+
+template <> struct __make_signed<bool,               true> {};
+template <> struct __make_signed<  signed short,     true> {typedef short     type;};
+template <> struct __make_signed<unsigned short,     true> {typedef short     type;};
+template <> struct __make_signed<  signed int,       true> {typedef int       type;};
+template <> struct __make_signed<unsigned int,       true> {typedef int       type;};
+template <> struct __make_signed<  signed long,      true> {typedef long      type;};
+template <> struct __make_signed<unsigned long,      true> {typedef long      type;};
+template <> struct __make_signed<  signed long long, true> {typedef long long type;};
+template <> struct __make_signed<unsigned long long, true> {typedef long long type;};
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+template <> struct __make_signed<__int128_t,         true> {typedef __int128_t type;};
+template <> struct __make_signed<__uint128_t,        true> {typedef __int128_t type;};
+#endif
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS make_signed
+{
+    typedef typename __apply_cv<_Tp, typename __make_signed<typename remove_cv<_Tp>::type>::type>::type type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using make_signed_t = typename make_signed<_Tp>::type;
+#endif
+
+template <class _Tp, bool = is_integral<_Tp>::value || is_enum<_Tp>::value>
+struct __make_unsigned {};
+
+template <class _Tp>
+struct __make_unsigned<_Tp, true>
+{
+    typedef typename __find_first<__unsigned_types, sizeof(_Tp)>::type type;
+};
+
+template <> struct __make_unsigned<bool,               true> {};
+template <> struct __make_unsigned<  signed short,     true> {typedef unsigned short     type;};
+template <> struct __make_unsigned<unsigned short,     true> {typedef unsigned short     type;};
+template <> struct __make_unsigned<  signed int,       true> {typedef unsigned int       type;};
+template <> struct __make_unsigned<unsigned int,       true> {typedef unsigned int       type;};
+template <> struct __make_unsigned<  signed long,      true> {typedef unsigned long      type;};
+template <> struct __make_unsigned<unsigned long,      true> {typedef unsigned long      type;};
+template <> struct __make_unsigned<  signed long long, true> {typedef unsigned long long type;};
+template <> struct __make_unsigned<unsigned long long, true> {typedef unsigned long long type;};
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+template <> struct __make_unsigned<__int128_t,         true> {typedef __uint128_t        type;};
+template <> struct __make_unsigned<__uint128_t,        true> {typedef __uint128_t        type;};
+#endif
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS make_unsigned
+{
+    typedef typename __apply_cv<_Tp, typename __make_unsigned<typename remove_cv<_Tp>::type>::type>::type type;
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using make_unsigned_t = typename make_unsigned<_Tp>::type;
+#endif
+
+template <class _Tp, class _Up, class = void>
+struct __common_type2_imp {};
+
+template <class _Tp, class _Up>
+struct __common_type2_imp<_Tp, _Up,
+                          typename __void_t<decltype(
+                                            true ? _CUDA_VSTD::declval<_Tp>() : _CUDA_VSTD::declval<_Up>()
+                                            )>::type>
+{
+  typedef _LIBCUDACXX_NODEBUG_TYPE typename decay<decltype(
+                         true ? _CUDA_VSTD::declval<_Tp>() : _CUDA_VSTD::declval<_Up>()
+                         )>::type type;
+};
+
+template <class, class = void>
+struct __common_type_impl {};
+
+// Clang provides variadic templates in C++03 as an extension.
+#if !defined(_LIBCUDACXX_CXX03_LANG) || defined(__clang__)
+# define _LIBCUDACXX_OPTIONAL_PACK(...) , __VA_ARGS__
+template <class... Tp>
+struct __common_types;
+template <class... _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS common_type;
+#else
+# define _LIBCUDACXX_OPTIONAL_PACK(...)
+struct __no_arg;
+template <class _Tp, class _Up, class = __no_arg>
+struct __common_types;
+template <class _Tp = __no_arg, class _Up = __no_arg, class _Vp = __no_arg,
+          class _Unused = __no_arg>
+struct common_type {
+  static_assert(sizeof(_Unused) == 0,
+                "common_type accepts at most 3 arguments in C++03");
+};
+#endif // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Up>
+struct __common_type_impl<
+    __common_types<_Tp, _Up>,
+    typename __void_t<typename common_type<_Tp, _Up>::type>::type>
+{
+  typedef typename common_type<_Tp, _Up>::type type;
+};
+
+template <class _Tp, class _Up, class _Vp _LIBCUDACXX_OPTIONAL_PACK(class... _Rest)>
+struct __common_type_impl<
+    __common_types<_Tp, _Up, _Vp _LIBCUDACXX_OPTIONAL_PACK(_Rest...)>,
+    typename __void_t<typename common_type<_Tp, _Up>::type>::type>
+    : __common_type_impl<__common_types<typename common_type<_Tp, _Up>::type,
+                                        _Vp _LIBCUDACXX_OPTIONAL_PACK(_Rest...)> > {
+};
+
+// bullet 1 - sizeof...(Tp) == 0
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS common_type<> {};
+
+// bullet 2 - sizeof...(Tp) == 1
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS common_type<_Tp>
+    : public common_type<_Tp, _Tp> {};
+
+// bullet 3 - sizeof...(Tp) == 2
+
+template <class _Tp, class _Up>
+struct _LIBCUDACXX_TEMPLATE_VIS common_type<_Tp, _Up>
+    : conditional<
+        _IsSame<_Tp, typename decay<_Tp>::type>::value && _IsSame<_Up, typename decay<_Up>::type>::value,
+        __common_type2_imp<_Tp, _Up>,
+        common_type<typename decay<_Tp>::type, typename decay<_Up>::type>
+    >::type
+{};
+
+// bullet 4 - sizeof...(Tp) > 2
+
+template <class _Tp, class _Up, class _Vp _LIBCUDACXX_OPTIONAL_PACK(class... _Rest)>
+struct _LIBCUDACXX_TEMPLATE_VIS
+    common_type<_Tp, _Up, _Vp _LIBCUDACXX_OPTIONAL_PACK(_Rest...)>
+    : __common_type_impl<
+          __common_types<_Tp, _Up, _Vp _LIBCUDACXX_OPTIONAL_PACK(_Rest...)> > {};
+
+#undef _LIBCUDACXX_OPTIONAL_PACK
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class ..._Tp> using common_type_t = typename common_type<_Tp...>::type;
+#endif
+
+// is_assignable
+
+template<typename, typename _Tp> struct __select_2nd { typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type; };
+
+template <class _Tp, class _Arg>
+_LIBCUDACXX_INLINE_VISIBILITY
+typename __select_2nd<decltype((_CUDA_VSTD::declval<_Tp>() = _CUDA_VSTD::declval<_Arg>())), true_type>::type
+__is_assignable_test(int);
+
+#if defined(_LIBCUDACXX_IS_ASSIGNABLE) && !defined(_LIBCUDACXX_USE_IS_ASSIGNABLE_FALLBACK)
+
+template <class _T1, class _T2> struct _LIBCUDACXX_TEMPLATE_VIS is_assignable
+    : public integral_constant<bool, _LIBCUDACXX_IS_ASSIGNABLE(_T1, _T2)> {};
+
+#else
+
+template <class, class>
+_LIBCUDACXX_INLINE_VISIBILITY
+false_type __is_assignable_test(...);
+
+template <class _Tp, class _Arg, bool = is_void<_Tp>::value || is_void<_Arg>::value>
+struct __is_assignable_imp
+    : public decltype((_CUDA_VSTD::__is_assignable_test<_Tp, _Arg>(0))) {};
+
+template <class _Tp, class _Arg>
+struct __is_assignable_imp<_Tp, _Arg, true>
+    : public false_type
+{
+};
+
+template <class _Tp, class _Arg>
+struct is_assignable
+    : public __is_assignable_imp<_Tp, _Arg> {};
+
+#endif // defined(_LIBCUDACXX_IS_ASSIGNABLE) && !defined(_LIBCUDACXX_USE_IS_ASSIGNABLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class _Arg>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_assignable_v
+    = is_assignable<_Tp, _Arg>::value;
+#endif
+
+// is_copy_assignable
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_copy_assignable
+    : public is_assignable<typename add_lvalue_reference<_Tp>::type,
+                  typename add_lvalue_reference<typename add_const<_Tp>::type>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_copy_assignable_v
+    = is_copy_assignable<_Tp>::value;
+#endif
+
+// is_move_assignable
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_move_assignable
+    : public is_assignable<typename add_lvalue_reference<_Tp>::type,
+                           typename add_rvalue_reference<_Tp>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_move_assignable_v
+    = is_move_assignable<_Tp>::value;
+#endif
+
+// is_destructible
+
+//  if it's a reference, return true
+//  if it's a function, return false
+//  if it's   void,     return false
+//  if it's an array of unknown bound, return false
+//  Otherwise, return "std::declval<_Up&>().~_Up()" is well-formed
+//    where _Up is remove_all_extents<_Tp>::type
+
+#if defined(_LIBCUDACXX_IS_DESTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_DESTRUCTIBLE_FALLBACK)
+
+template <class _Tp>
+struct is_destructible
+   : public _CUDA_VSTD::integral_constant<bool, _LIBCUDACXX_IS_DESTRUCTIBLE(_Tp)> {};
+
+#else
+
+template <class>
+struct __is_destructible_apply { typedef int type; };
+
+template <typename _Tp>
+struct __is_destructor_wellformed {
+    template <typename _Tp1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static char  __test (
+        typename __is_destructible_apply<decltype(_CUDA_VSTD::declval<_Tp1&>().~_Tp1())>::type
+    );
+
+    template <typename _Tp1>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static __two __test (...);
+
+    static const bool value = sizeof(__test<_Tp>(12)) == sizeof(char);
+};
+
+template <class _Tp, bool>
+struct __destructible_imp;
+
+template <class _Tp>
+struct __destructible_imp<_Tp, false>
+   : public _CUDA_VSTD::integral_constant<bool,
+        __is_destructor_wellformed<typename _CUDA_VSTD::remove_all_extents<_Tp>::type>::value> {};
+
+template <class _Tp>
+struct __destructible_imp<_Tp, true>
+    : public _CUDA_VSTD::true_type {};
+
+template <class _Tp, bool>
+struct __destructible_false;
+
+template <class _Tp>
+struct __destructible_false<_Tp, false> : public __destructible_imp<_Tp, _CUDA_VSTD::is_reference<_Tp>::value> {};
+
+template <class _Tp>
+struct __destructible_false<_Tp, true> : public _CUDA_VSTD::false_type {};
+
+template <class _Tp>
+struct is_destructible
+    : public __destructible_false<_Tp, _CUDA_VSTD::is_function<_Tp>::value> {};
+
+template <class _Tp>
+struct is_destructible<_Tp[]>
+    : public _CUDA_VSTD::false_type {};
+
+template <>
+struct is_destructible<void>
+    : public _CUDA_VSTD::false_type {};
+
+#endif // defined(_LIBCUDACXX_IS_DESTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_DESTRUCTIBLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_destructible_v
+    = is_destructible<_Tp>::value;
+#endif
+
+// move
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename remove_reference<_Tp>::type&&
+move(_Tp&& __t) _NOEXCEPT
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE typename remove_reference<_Tp>::type _Up;
+    return static_cast<_Up&&>(__t);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp&&
+forward(typename remove_reference<_Tp>::type& __t) _NOEXCEPT
+{
+    return static_cast<_Tp&&>(__t);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+_Tp&&
+forward(typename remove_reference<_Tp>::type&& __t) _NOEXCEPT
+{
+    static_assert(!is_lvalue_reference<_Tp>::value,
+                  "can not forward an rvalue as an lvalue");
+    return static_cast<_Tp&&>(__t);
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename decay<_Tp>::type
+__decay_copy(_Tp&& __t)
+{
+    return _CUDA_VSTD::forward<_Tp>(__t);
+}
+
+template <class _MP, bool _IsMemberFunctionPtr, bool _IsMemberObjectPtr>
+struct __member_pointer_traits_imp
+{
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) const volatile, true, false>
+{
+    typedef _Class const volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) const volatile, true, false>
+{
+    typedef _Class const volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+#if defined(_LIBCUDACXX_HAS_REFERENCE_QUALIFIED_FUNCTIONS)
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) &, true, false>
+{
+    typedef _Class& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) &, true, false>
+{
+    typedef _Class& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) const&, true, false>
+{
+    typedef _Class const& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) const&, true, false>
+{
+    typedef _Class const& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) volatile&, true, false>
+{
+    typedef _Class volatile& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) volatile&, true, false>
+{
+    typedef _Class volatile& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) const volatile&, true, false>
+{
+    typedef _Class const volatile& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) const volatile&, true, false>
+{
+    typedef _Class const volatile& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) &&, true, false>
+{
+    typedef _Class&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) &&, true, false>
+{
+    typedef _Class&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) const&&, true, false>
+{
+    typedef _Class const&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) const&&, true, false>
+{
+    typedef _Class const&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) volatile&&, true, false>
+{
+    typedef _Class volatile&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) volatile&&, true, false>
+{
+    typedef _Class volatile&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param...) const volatile&&, true, false>
+{
+    typedef _Class const volatile&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param...);
+};
+
+template <class _Rp, class _Class, class ..._Param>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_Param..., ...) const volatile&&, true, false>
+{
+    typedef _Class const volatile&& _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_Param..., ...);
+};
+
+#endif // defined(_LIBCUDACXX_HAS_REFERENCE_QUALIFIED_FUNCTIONS)
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) ();
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(...), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (...);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, ...), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, ...);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, ...), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, ...);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1, class _P2>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, _P2), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, _P2);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1, class _P2>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, _P2, ...), true, false>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, _P2, ...);
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)() const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) ();
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(...) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (...);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, ...) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, ...);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, ...) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, ...);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1, class _P2>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, _P2) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, _P2);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1, class _P2>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, _P2, ...) const, true, false>
+{
+    typedef _Class const _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, _P2, ...);
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)() volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) ();
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(...) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (...);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, ...) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, ...);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, ...) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, ...);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1, class _P2>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, _P2) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, _P2);
+};
+
+template <class _Rp, class _Class, class _P0, class _P1, class _P2>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0, _P1, _P2, ...) volatile, true, false>
+{
+    typedef _Class volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0, _P1, _P2, ...);
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)() const volatile, true, false>
+{
+    typedef _Class const volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) ();
+};
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(...) const volatile, true, false>
+{
+    typedef _Class const volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (...);
+};
+
+template <class _Rp, class _Class, class _P0>
+struct __member_pointer_traits_imp<_Rp (_Class::*)(_P0) const volatile, true, false>
+{
+    typedef _Class const volatile _ClassType;
+    typedef _Rp _ReturnType;
+    typedef _Rp (_FnType) (_P0);
+};
+
+#endif
+
+
+template <class _Rp, class _Class>
+struct __member_pointer_traits_imp<_Rp _Class::*, false, true>
+{
+    typedef _Class _ClassType;
+    typedef _Rp _ReturnType;
+};
+
+template <class _MP>
+struct __member_pointer_traits
+    : public __member_pointer_traits_imp<typename remove_cv<_MP>::type,
+                    is_member_function_pointer<_MP>::value,
+                    is_member_object_pointer<_MP>::value>
+{
+//     typedef ... _ClassType;
+//     typedef ... _ReturnType;
+//     typedef ... _FnType;
+};
+
+
+template <class _DecayedFp>
+struct __member_pointer_class_type {};
+
+template <class _Ret, class _ClassType>
+struct __member_pointer_class_type<_Ret _ClassType::*> {
+  typedef _ClassType type;
+};
+
+// result_of
+
+template <class _Callable> class result_of;
+
+#ifdef _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Fn, bool, bool>
+class __result_of
+{
+};
+
+template <class _Fn>
+class __result_of<_Fn(), true, false>
+{
+public:
+    typedef decltype(declval<_Fn>()()) type;
+};
+
+template <class _Fn, class _A0>
+class __result_of<_Fn(_A0), true, false>
+{
+public:
+    typedef decltype(declval<_Fn>()(declval<_A0>())) type;
+};
+
+template <class _Fn, class _A0, class _A1>
+class __result_of<_Fn(_A0, _A1), true, false>
+{
+public:
+    typedef decltype(declval<_Fn>()(declval<_A0>(), declval<_A1>())) type;
+};
+
+template <class _Fn, class _A0, class _A1, class _A2>
+class __result_of<_Fn(_A0, _A1, _A2), true, false>
+{
+public:
+    typedef decltype(declval<_Fn>()(declval<_A0>(), declval<_A1>(), declval<_A2>())) type;
+};
+
+template <class _MP, class _Tp, bool _IsMemberFunctionPtr>
+struct __result_of_mp;
+
+// member function pointer
+
+template <class _MP, class _Tp>
+struct __result_of_mp<_MP, _Tp, true>
+    : public __identity<typename __member_pointer_traits<_MP>::_ReturnType>
+{
+};
+
+// member data pointer
+
+template <class _MP, class _Tp, bool>
+struct __result_of_mdp;
+
+template <class _Rp, class _Class, class _Tp>
+struct __result_of_mdp<_Rp _Class::*, _Tp, false>
+{
+    typedef typename __apply_cv<decltype(*_CUDA_VSTD::declval<_Tp>()), _Rp>::type& type;
+};
+
+template <class _Rp, class _Class, class _Tp>
+struct __result_of_mdp<_Rp _Class::*, _Tp, true>
+{
+    typedef typename __apply_cv<_Tp, _Rp>::type& type;
+};
+
+template <class _Rp, class _Class, class _Tp>
+struct __result_of_mp<_Rp _Class::*, _Tp, false>
+    : public __result_of_mdp<_Rp _Class::*, _Tp,
+            is_base_of<_Class, typename remove_reference<_Tp>::type>::value>
+{
+};
+
+
+
+template <class _Fn, class _Tp>
+class __result_of<_Fn(_Tp), false, true>  // _Fn must be member pointer
+    : public __result_of_mp<typename remove_reference<_Fn>::type,
+                            _Tp,
+                            is_member_function_pointer<typename remove_reference<_Fn>::type>::value>
+{
+};
+
+template <class _Fn, class _Tp, class _A0>
+class __result_of<_Fn(_Tp, _A0), false, true>  // _Fn must be member pointer
+    : public __result_of_mp<typename remove_reference<_Fn>::type,
+                            _Tp,
+                            is_member_function_pointer<typename remove_reference<_Fn>::type>::value>
+{
+};
+
+template <class _Fn, class _Tp, class _A0, class _A1>
+class __result_of<_Fn(_Tp, _A0, _A1), false, true>  // _Fn must be member pointer
+    : public __result_of_mp<typename remove_reference<_Fn>::type,
+                            _Tp,
+                            is_member_function_pointer<typename remove_reference<_Fn>::type>::value>
+{
+};
+
+template <class _Fn, class _Tp, class _A0, class _A1, class _A2>
+class __result_of<_Fn(_Tp, _A0, _A1, _A2), false, true>  // _Fn must be member pointer
+    : public __result_of_mp<typename remove_reference<_Fn>::type,
+                            _Tp,
+                            is_member_function_pointer<typename remove_reference<_Fn>::type>::value>
+{
+};
+
+// result_of
+
+template <class _Fn>
+class _LIBCUDACXX_TEMPLATE_VIS result_of<_Fn()>
+    : public __result_of<_Fn(),
+                         is_class<typename remove_reference<_Fn>::type>::value ||
+                         is_function<typename remove_pointer<typename remove_reference<_Fn>::type>::type>::value,
+                         is_member_pointer<typename remove_reference<_Fn>::type>::value
+                        >
+{
+};
+
+template <class _Fn, class _A0>
+class _LIBCUDACXX_TEMPLATE_VIS result_of<_Fn(_A0)>
+    : public __result_of<_Fn(_A0),
+                         is_class<typename remove_reference<_Fn>::type>::value ||
+                         is_function<typename remove_pointer<typename remove_reference<_Fn>::type>::type>::value,
+                         is_member_pointer<typename remove_reference<_Fn>::type>::value
+                        >
+{
+};
+
+template <class _Fn, class _A0, class _A1>
+class _LIBCUDACXX_TEMPLATE_VIS result_of<_Fn(_A0, _A1)>
+    : public __result_of<_Fn(_A0, _A1),
+                         is_class<typename remove_reference<_Fn>::type>::value ||
+                         is_function<typename remove_pointer<typename remove_reference<_Fn>::type>::type>::value,
+                         is_member_pointer<typename remove_reference<_Fn>::type>::value
+                        >
+{
+};
+
+template <class _Fn, class _A0, class _A1, class _A2>
+class _LIBCUDACXX_TEMPLATE_VIS result_of<_Fn(_A0, _A1, _A2)>
+    : public __result_of<_Fn(_A0, _A1, _A2),
+                         is_class<typename remove_reference<_Fn>::type>::value ||
+                         is_function<typename remove_pointer<typename remove_reference<_Fn>::type>::type>::value,
+                         is_member_pointer<typename remove_reference<_Fn>::type>::value
+                        >
+{
+};
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+// template <class T, class... Args> struct is_constructible;
+
+namespace __is_construct
+{
+struct __nat {};
+}
+
+// FIXME: This logic isn't awesome.
+#if !defined(_LIBCUDACXX_CXX03_LANG) && (!defined(_LIBCUDACXX_IS_CONSTRUCTIBLE) || \
+    defined(_LIBCUDACXX_TESTING_FALLBACK_IS_CONSTRUCTIBLE) || \
+    defined(_LIBCUDACXX_USE_IS_CONSTRUCTIBLE_FALLBACK))
+
+template <class _Tp, class... _Args>
+struct __libcpp_is_constructible;
+
+template <class _To, class _From>
+struct __is_invalid_base_to_derived_cast {
+  static_assert(is_reference<_To>::value, "Wrong specialization");
+  using _RawFrom = __uncvref_t<_From>;
+  using _RawTo = __uncvref_t<_To>;
+  static const bool value = _And<
+        _IsNotSame<_RawFrom, _RawTo>,
+        is_base_of<_RawFrom, _RawTo>,
+        _Not<__libcpp_is_constructible<_RawTo, _From>>
+  >::value;
+};
+
+template <class _To, class _From>
+struct __is_invalid_lvalue_to_rvalue_cast : false_type {
+  static_assert(is_reference<_To>::value, "Wrong specialization");
+};
+
+template <class _ToRef, class _FromRef>
+struct __is_invalid_lvalue_to_rvalue_cast<_ToRef&&, _FromRef&> {
+  using _RawFrom = __uncvref_t<_FromRef>;
+  using _RawTo = __uncvref_t<_ToRef>;
+  static const bool value = _And<
+      _Not<is_function<_RawTo>>,
+      _Or<
+        _IsSame<_RawFrom, _RawTo>,
+        is_base_of<_RawTo, _RawFrom>>
+    >::value;
+};
+
+struct __is_constructible_helper
+{
+    template <class _To>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __eat(_To);
+
+    // This overload is needed to work around a Clang bug that disallows
+    // static_cast<T&&>(e) for non-reference-compatible types.
+    // Example: static_cast<int&&>(declval<double>());
+    // NOTE: The static_cast implementation below is required to support
+    //  classes with explicit conversion operators.
+    template <class _To, class _From,
+              class = decltype(__eat<_To>(_CUDA_VSTD::declval<_From>()))>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static true_type __test_cast(int);
+
+    template <class _To, class _From,
+              class = decltype(static_cast<_To>(_CUDA_VSTD::declval<_From>()))>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static integral_constant<bool,
+        !__is_invalid_base_to_derived_cast<_To, _From>::value &&
+        !__is_invalid_lvalue_to_rvalue_cast<_To, _From>::value
+    > __test_cast(long);
+
+    template <class, class>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static false_type __test_cast(...);
+
+    template <class _Tp, class ..._Args,
+        class = decltype(_Tp(_CUDA_VSTD::declval<_Args>()...))>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static true_type __test_nary(int);
+    template <class _Tp, class...>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static false_type __test_nary(...);
+
+    template <class _Tp, class _A0, class = decltype(::new _Tp(_CUDA_VSTD::declval<_A0>()))>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static is_destructible<_Tp> __test_unary(int);
+    template <class, class>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static false_type __test_unary(...);
+};
+
+template <class _Tp, bool = is_void<_Tp>::value>
+struct __is_default_constructible
+    : decltype(__is_constructible_helper::__test_nary<_Tp>(0))
+{};
+
+template <class _Tp>
+struct __is_default_constructible<_Tp, true> : false_type {};
+
+template <class _Tp>
+struct __is_default_constructible<_Tp[], false> : false_type {};
+
+template <class _Tp, size_t _Nx>
+struct __is_default_constructible<_Tp[_Nx], false>
+    : __is_default_constructible<typename remove_all_extents<_Tp>::type>  {};
+
+template <class _Tp, class... _Args>
+struct __libcpp_is_constructible
+{
+  static_assert(sizeof...(_Args) > 1, "Wrong specialization");
+  typedef decltype(__is_constructible_helper::__test_nary<_Tp, _Args...>(0))
+      type;
+};
+
+template <class _Tp>
+struct __libcpp_is_constructible<_Tp> : __is_default_constructible<_Tp> {};
+
+template <class _Tp, class _A0>
+struct __libcpp_is_constructible<_Tp, _A0>
+    : public decltype(__is_constructible_helper::__test_unary<_Tp, _A0>(0))
+{};
+
+template <class _Tp, class _A0>
+struct __libcpp_is_constructible<_Tp&, _A0>
+    : public decltype(__is_constructible_helper::
+    __test_cast<_Tp&, _A0>(0))
+{};
+
+template <class _Tp, class _A0>
+struct __libcpp_is_constructible<_Tp&&, _A0>
+    : public decltype(__is_constructible_helper::
+    __test_cast<_Tp&&, _A0>(0))
+{};
+
+#endif
+
+#if defined(_LIBCUDACXX_IS_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_CONSTRUCTIBLE_FALLBACK)
+template <class _Tp, class ..._Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible
+    : public integral_constant<bool, _LIBCUDACXX_IS_CONSTRUCTIBLE(_Tp, _Args...)>
+    {};
+#elif !defined(_LIBCUDACXX_CXX03_LANG)
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible
+    : public __libcpp_is_constructible<_Tp, _Args...>::type {};
+#else
+// template <class T> struct is_constructible0;
+
+//      main is_constructible0 test
+
+template <class _Tp>
+decltype((_Tp(), true_type()))
+__is_constructible0_test(_Tp&);
+
+false_type
+__is_constructible0_test(__any);
+
+template <class _Tp, class _A0>
+decltype((_Tp(_CUDA_VSTD::declval<_A0>()), true_type()))
+__is_constructible1_test(_Tp&, _A0&);
+
+template <class _A0>
+false_type
+__is_constructible1_test(__any, _A0&);
+
+template <class _Tp, class _A0, class _A1>
+decltype((_Tp(_CUDA_VSTD::declval<_A0>(), _CUDA_VSTD::declval<_A1>()), true_type()))
+__is_constructible2_test(_Tp&, _A0&, _A1&);
+
+template <class _A0, class _A1>
+false_type
+__is_constructible2_test(__any, _A0&, _A1&);
+
+template <class _Tp, class _A0, class _A1, class _A2>
+decltype((_Tp(_CUDA_VSTD::declval<_A0>(), _CUDA_VSTD::declval<_A1>(), _CUDA_VSTD::declval<_A2>()), true_type()))
+__is_constructible3_test(_Tp&, _A0&, _A1&, _A2&);
+
+template <class _A0, class _A1, class _A2>
+false_type
+__is_constructible3_test(__any, _A0&, _A1&, _A2&);
+
+template <bool, class _Tp>
+struct __is_constructible0_imp // false, _Tp is not a scalar
+    : public common_type
+             <
+                 decltype(__is_constructible0_test(declval<_Tp&>()))
+             >::type
+    {};
+
+template <bool, class _Tp, class _A0>
+struct __is_constructible1_imp // false, _Tp is not a scalar
+    : public common_type
+             <
+                 decltype(__is_constructible1_test(declval<_Tp&>(), declval<_A0&>()))
+             >::type
+    {};
+
+template <bool, class _Tp, class _A0, class _A1>
+struct __is_constructible2_imp // false, _Tp is not a scalar
+    : public common_type
+             <
+                 decltype(__is_constructible2_test(declval<_Tp&>(), declval<_A0>(), declval<_A1>()))
+             >::type
+    {};
+
+template <bool, class _Tp, class _A0, class _A1, class _A2>
+struct __is_constructible3_imp // false, _Tp is not a scalar
+    : public common_type
+             <
+                 decltype(__is_constructible3_test(declval<_Tp&>(), declval<_A0>(), declval<_A1>(), declval<_A2>()))
+             >::type
+    {};
+
+//      handle scalars and reference types
+
+//      Scalars are default constructible, references are not
+
+template <class _Tp>
+struct __is_constructible0_imp<true, _Tp>
+    : public is_scalar<_Tp>
+    {};
+
+template <class _Tp, class _A0>
+struct __is_constructible1_imp<true, _Tp, _A0>
+    : public is_convertible<_A0, _Tp>
+    {};
+
+template <class _Tp, class _A0, class _A1>
+struct __is_constructible2_imp<true, _Tp, _A0, _A1>
+    : public false_type
+    {};
+
+template <class _Tp, class _A0, class _A1, class _A2>
+struct __is_constructible3_imp<true, _Tp, _A0, _A1, _A2>
+    : public false_type
+    {};
+
+//      Treat scalars and reference types separately
+
+template <bool, class _Tp>
+struct __is_constructible0_void_check
+    : public __is_constructible0_imp<is_scalar<_Tp>::value || is_reference<_Tp>::value,
+                                _Tp>
+    {};
+
+template <bool, class _Tp, class _A0>
+struct __is_constructible1_void_check
+    : public __is_constructible1_imp<is_scalar<_Tp>::value || is_reference<_Tp>::value,
+                                _Tp, _A0>
+    {};
+
+template <bool, class _Tp, class _A0, class _A1>
+struct __is_constructible2_void_check
+    : public __is_constructible2_imp<is_scalar<_Tp>::value || is_reference<_Tp>::value,
+                                _Tp, _A0, _A1>
+    {};
+
+template <bool, class _Tp, class _A0, class _A1, class _A2>
+struct __is_constructible3_void_check
+    : public __is_constructible3_imp<is_scalar<_Tp>::value || is_reference<_Tp>::value,
+                                _Tp, _A0, _A1, _A2>
+    {};
+
+//      If any of T or Args is void, is_constructible should be false
+
+template <class _Tp>
+struct __is_constructible0_void_check<true, _Tp>
+    : public false_type
+    {};
+
+template <class _Tp, class _A0>
+struct __is_constructible1_void_check<true, _Tp, _A0>
+    : public false_type
+    {};
+
+template <class _Tp, class _A0, class _A1>
+struct __is_constructible2_void_check<true, _Tp, _A0, _A1>
+    : public false_type
+    {};
+
+template <class _Tp, class _A0, class _A1, class _A2>
+struct __is_constructible3_void_check<true, _Tp, _A0, _A1, _A2>
+    : public false_type
+    {};
+
+//      is_constructible entry point
+
+template <class _Tp, class _A0 = __is_construct::__nat,
+                     class _A1 = __is_construct::__nat,
+                     class _A2 = __is_construct::__nat>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible
+    : public __is_constructible3_void_check<is_void<_Tp>::value
+                                        || is_abstract<_Tp>::value
+                                        || is_function<_Tp>::value
+                                        || is_void<_A0>::value
+                                        || is_void<_A1>::value
+                                        || is_void<_A2>::value,
+                                           _Tp, _A0, _A1, _A2>
+    {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible<_Tp, __is_construct::__nat, __is_construct::__nat>
+    : public __is_constructible0_void_check<is_void<_Tp>::value
+                                        || is_abstract<_Tp>::value
+                                        || is_function<_Tp>::value,
+                                           _Tp>
+    {};
+
+template <class _Tp, class _A0>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible<_Tp, _A0, __is_construct::__nat>
+    : public __is_constructible1_void_check<is_void<_Tp>::value
+                                        || is_abstract<_Tp>::value
+                                        || is_function<_Tp>::value
+                                        || is_void<_A0>::value,
+                                           _Tp, _A0>
+    {};
+
+template <class _Tp, class _A0, class _A1>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible<_Tp, _A0, _A1, __is_construct::__nat>
+    : public __is_constructible2_void_check<is_void<_Tp>::value
+                                        || is_abstract<_Tp>::value
+                                        || is_function<_Tp>::value
+                                        || is_void<_A0>::value
+                                        || is_void<_A1>::value,
+                                           _Tp, _A0, _A1>
+    {};
+
+//      Array types are default constructible if their element type
+//      is default constructible
+
+template <class _Ap, size_t _Np>
+struct __is_constructible0_imp<false, _Ap[_Np]>
+    : public is_constructible<typename remove_all_extents<_Ap>::type>
+    {};
+
+template <class _Ap, size_t _Np, class _A0>
+struct __is_constructible1_imp<false, _Ap[_Np], _A0>
+    : public false_type
+    {};
+
+template <class _Ap, size_t _Np, class _A0, class _A1>
+struct __is_constructible2_imp<false, _Ap[_Np], _A0, _A1>
+    : public false_type
+    {};
+
+template <class _Ap, size_t _Np, class _A0, class _A1, class _A2>
+struct __is_constructible3_imp<false, _Ap[_Np], _A0, _A1, _A2>
+    : public false_type
+    {};
+
+//      Incomplete array types are not constructible
+
+template <class _Ap>
+struct __is_constructible0_imp<false, _Ap[]>
+    : public false_type
+    {};
+
+template <class _Ap, class _A0>
+struct __is_constructible1_imp<false, _Ap[], _A0>
+    : public false_type
+    {};
+
+template <class _Tp, class _A0>
+struct __libcpp_is_constructible<_Tp&, _A0>
+    : public decltype(__is_constructible_helper::
+    __test_cast<_Tp&, _A0>(0))
+{};
+
+template <class _Tp, class _A0>
+struct __libcpp_is_constructible<_Tp&&, _A0>
+    : public decltype(__is_constructible_helper::
+    __test_cast<_Tp&&, _A0>(0))
+{};
+
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_constructible
+    : public __libcpp_is_constructible<_Tp, _Args...>::type {};
+
+#endif  // defined(_LIBCUDACXX_IS_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_CONSTRUCTIBLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class ..._Args>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_constructible_v
+    = is_constructible<_Tp, _Args...>::value;
+#endif
+
+// is_default_constructible
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_default_constructible
+    : public is_constructible<_Tp>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_default_constructible_v
+    = is_default_constructible<_Tp>::value;
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+// First of all, we can't implement this check in C++03 mode because the {}
+// default initialization syntax isn't valid.
+// Second, we implement the trait in a funny manner with two defaulted template
+// arguments to workaround Clang's PR43454.
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __test_implicit_default_constructible(_Tp);
+
+template <class _Tp, class = void, bool = is_default_constructible<_Tp>::value>
+struct __is_implicitly_default_constructible
+    : false_type
+{ };
+
+template <class _Tp>
+struct __is_implicitly_default_constructible<_Tp, decltype(__test_implicit_default_constructible<_Tp const&>({})), true>
+    : true_type
+{ };
+
+template <class _Tp>
+struct __is_implicitly_default_constructible<_Tp, decltype(__test_implicit_default_constructible<_Tp const&>({})), false>
+    : false_type
+{ };
+#endif // !C++03
+
+// is_copy_constructible
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_copy_constructible
+    : public is_constructible<_Tp,
+                  typename add_lvalue_reference<typename add_const<_Tp>::type>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_copy_constructible_v
+    = is_copy_constructible<_Tp>::value;
+#endif
+
+// is_move_constructible
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_move_constructible
+    : public is_constructible<_Tp, typename add_rvalue_reference<_Tp>::type>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_move_constructible_v
+    = is_move_constructible<_Tp>::value;
+#endif
+
+// is_trivially_constructible
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+#if defined(_LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_CONSTRUCTIBLE_FALLBACK)
+
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible
+    : integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(_Tp, _Args...)>
+{
+};
+
+#else
+
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible
+    : false_type
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp>
+#if defined(_LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_TRIVIAL_CONSTRUCTOR_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_TRIVIAL_CONSTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_TRIVIAL_CONSTRUCTOR_FALLBACK)
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, _Tp&&>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, const _Tp&>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, _Tp&>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+#endif // defined(_LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_CONSTRUCTIBLE_FALLBACK)
+
+#else  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Tp, class _A0 = __is_construct::__nat,
+                     class _A1 = __is_construct::__nat>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible
+    : false_type
+{
+};
+
+#if defined(_LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_CONSTRUCTIBLE_FALLBACK)
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, __is_construct::__nat,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(_Tp)>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, _Tp,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(_Tp, _Tp)>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, const _Tp&,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(_Tp, const _Tp&)>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, _Tp&,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE(_Tp, _Tp&)>
+{
+};
+
+#else
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, __is_construct::__nat,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, _Tp,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, const _Tp&,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_constructible<_Tp, _Tp&,
+                                                       __is_construct::__nat>
+    : integral_constant<bool, is_scalar<_Tp>::value>
+{
+};
+
+#endif // defined(_LIBCUDACXX_IS_TRIVIALLY_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_CONSTRUCTIBLE_FALLBACK)
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class... _Args>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_constructible_v
+    = is_trivially_constructible<_Tp, _Args...>::value;
+#endif
+
+// is_trivially_default_constructible
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_default_constructible
+    : public is_trivially_constructible<_Tp>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_default_constructible_v
+    = is_trivially_default_constructible<_Tp>::value;
+#endif
+
+// is_trivially_copy_constructible
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_copy_constructible
+    : public is_trivially_constructible<_Tp, typename add_lvalue_reference<const _Tp>::type>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_copy_constructible_v
+    = is_trivially_copy_constructible<_Tp>::value;
+#endif
+
+// is_trivially_move_constructible
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_move_constructible
+    : public is_trivially_constructible<_Tp, typename add_rvalue_reference<_Tp>::type>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_move_constructible_v
+    = is_trivially_move_constructible<_Tp>::value;
+#endif
+
+// is_trivially_assignable
+
+#if defined(_LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_ASSIGNABLE_FALLBACK)
+
+template <class _Tp, class _Arg>
+struct is_trivially_assignable
+    : integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE(_Tp, _Arg)>
+{
+};
+
+#else
+
+template <class _Tp, class _Arg>
+struct is_trivially_assignable
+    : public false_type {};
+
+template <class _Tp>
+struct is_trivially_assignable<_Tp&, _Tp>
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+
+template <class _Tp>
+struct is_trivially_assignable<_Tp&, _Tp&>
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+
+template <class _Tp>
+struct is_trivially_assignable<_Tp&, const _Tp&>
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+
+template <class _Tp>
+struct is_trivially_assignable<_Tp&, _Tp&&>
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+
+#endif // defined(_LIBCUDACXX_IS_TRIVIALLY_ASSIGNABLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_ASSIGNABLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class _Arg>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_assignable_v
+    = is_trivially_assignable<_Tp, _Arg>::value;
+#endif
+
+// is_trivially_copy_assignable
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_copy_assignable
+    : public is_trivially_assignable<typename add_lvalue_reference<_Tp>::type,
+                  typename add_lvalue_reference<typename add_const<_Tp>::type>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_copy_assignable_v
+    = is_trivially_copy_assignable<_Tp>::value;
+#endif
+
+// is_trivially_move_assignable
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_move_assignable
+    : public is_trivially_assignable<typename add_lvalue_reference<_Tp>::type,
+                                     typename add_rvalue_reference<_Tp>::type>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_move_assignable_v
+    = is_trivially_move_assignable<_Tp>::value;
+#endif
+
+// is_trivially_destructible
+
+#if defined(_LIBCUDACXX_IS_TRIVIALLY_DESTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_DESTRUCTIBLE_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_destructible
+    : public integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_DESTRUCTIBLE(_Tp)> {};
+
+#elif defined(_LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_TRIVIAL_DESTRUCTOR_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_destructible
+    : public integral_constant<bool, is_destructible<_Tp>::value && _LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR(_Tp)> {};
+
+#else
+
+template <class _Tp> struct __libcpp_trivial_destructor
+    : public integral_constant<bool, is_scalar<_Tp>::value ||
+                                     is_reference<_Tp>::value> {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_destructible
+    : public __libcpp_trivial_destructor<typename remove_all_extents<_Tp>::type> {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_destructible<_Tp[]>
+    : public false_type {};
+
+#endif // defined(_LIBCUDACXX_HAS_TRIVIAL_DESTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_TRIVIAL_DESTRUCTOR_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_destructible_v
+    = is_trivially_destructible<_Tp>::value;
+#endif
+
+// is_nothrow_constructible
+
+#if defined(_LIBCUDACXX_IS_NOTHROW_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_NOTHROW_CONSTRUCTIBLE_FALLBACK)
+
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible
+    : public integral_constant<bool, _LIBCUDACXX_IS_NOTHROW_CONSTRUCTIBLE(_Tp, _Args...)>
+{
+};
+
+#else
+
+#ifndef _LIBCUDACXX_HAS_NO_VARIADICS
+
+#if !defined(_LIBCUDACXX_HAS_NO_NOEXCEPT)
+
+template <bool, bool, class _Tp, class... _Args> struct __libcpp_is_nothrow_constructible;
+
+template <class _Tp, class... _Args>
+struct __libcpp_is_nothrow_constructible</*is constructible*/true, /*is reference*/false, _Tp, _Args...>
+    : public integral_constant<bool, noexcept(_Tp(declval<_Args>()...))>
+{
+};
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+void __implicit_conversion_to(_Tp) noexcept { }
+
+template <class _Tp, class _Arg>
+struct __libcpp_is_nothrow_constructible</*is constructible*/true, /*is reference*/true, _Tp, _Arg>
+    : public integral_constant<bool, noexcept(__implicit_conversion_to<_Tp>(declval<_Arg>()))>
+{
+};
+
+template <class _Tp, bool _IsReference, class... _Args>
+struct __libcpp_is_nothrow_constructible</*is constructible*/false, _IsReference, _Tp, _Args...>
+    : public false_type
+{
+};
+
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible
+    : __libcpp_is_nothrow_constructible<is_constructible<_Tp, _Args...>::value, is_reference<_Tp>::value, _Tp, _Args...>
+{
+};
+
+template <class _Tp, size_t _Ns>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp[_Ns]>
+    : __libcpp_is_nothrow_constructible<is_constructible<_Tp>::value, is_reference<_Tp>::value, _Tp>
+{
+};
+
+#else
+
+template <class _Tp, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible
+    : false_type
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_CONSTRUCTOR_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_CONSTRUCTOR_FALLBACK)
+{
+};
+
+template <class _Tp>
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, _Tp&&>
+#else
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, _Tp>
+#endif
+#if defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_COPY(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, const _Tp&>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_COPY(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, _Tp&>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_COPY(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+{
+};
+
+#endif // !defined(_LIBCUDACXX_HAS_NO_NOEXCEPT)
+
+#else // _LIBCUDACXX_HAS_NO_VARIADICS
+
+template <class _Tp, class _A0 = __is_construct::__nat,
+                     class _A1 = __is_construct::__nat>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible
+    : false_type
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, __is_construct::__nat,
+                                                       __is_construct::__nat>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_CONSTRUCTOR_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_CONSTRUCTOR) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_CONSTRUCTOR_FALLBACK)
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, _Tp,
+                                                       __is_construct::__nat>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_COPY(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, const _Tp&,
+                                                       __is_construct::__nat>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_COPY(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_constructible<_Tp, _Tp&,
+                                                       __is_construct::__nat>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_COPY(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value>
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_COPY) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_COPY_FALLBACK)
+{
+};
+
+#endif  // _LIBCUDACXX_HAS_NO_VARIADICS
+
+#endif // defined(_LIBCUDACXX_IS_NOTHROW_CONSTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_NOTHROW_CONSTRUCTIBLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class ..._Args>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_constructible_v
+    = is_nothrow_constructible<_Tp, _Args...>::value;
+#endif
+
+// is_nothrow_default_constructible
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_default_constructible
+    : public is_nothrow_constructible<_Tp>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_default_constructible_v
+    = is_nothrow_default_constructible<_Tp>::value;
+#endif
+
+// is_nothrow_copy_constructible
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_copy_constructible
+    : public is_nothrow_constructible<_Tp,
+                  typename add_lvalue_reference<typename add_const<_Tp>::type>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_copy_constructible_v
+    = is_nothrow_copy_constructible<_Tp>::value;
+#endif
+
+// is_nothrow_move_constructible
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_move_constructible
+    : public is_nothrow_constructible<_Tp, typename add_rvalue_reference<_Tp>::type>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_move_constructible_v
+    = is_nothrow_move_constructible<_Tp>::value;
+#endif
+
+// is_nothrow_assignable
+
+#if defined(_LIBCUDACXX_IS_NOTHROW_ASSIGNABLE) && !defined(_LIBCUDACXX_USE_IS_NOTHROW_ASSIGNABLE_FALLBACK)
+
+template <class _Tp, class _Arg>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_assignable
+    : public integral_constant<bool, _LIBCUDACXX_IS_NOTHROW_ASSIGNABLE(_Tp, _Arg)> {};
+
+#elif !defined(_LIBCUDACXX_HAS_NO_NOEXCEPT) && !defined(_LIBCUDACXX_HAS_NO_NOEXCEPT_SFINAE)
+
+template <bool, class _Tp, class _Arg> struct __libcpp_is_nothrow_assignable;
+
+template <class _Tp, class _Arg>
+struct __libcpp_is_nothrow_assignable<false, _Tp, _Arg>
+    : public false_type
+{
+};
+
+template <class _Tp, class _Arg>
+struct __libcpp_is_nothrow_assignable<true, _Tp, _Arg>
+    : public integral_constant<bool, noexcept(_CUDA_VSTD::declval<_Tp>() = _CUDA_VSTD::declval<_Arg>()) >
+{
+};
+
+template <class _Tp, class _Arg>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_assignable
+    : public __libcpp_is_nothrow_assignable<is_assignable<_Tp, _Arg>::value, _Tp, _Arg>
+{
+};
+
+#else
+
+template <class _Tp, class _Arg>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_assignable
+    : public false_type {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_assignable<_Tp&, _Tp>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_ASSIGN(_Tp)> {};
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_assignable<_Tp&, _Tp&>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_ASSIGN(_Tp)> {};
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_assignable<_Tp&, const _Tp&>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_ASSIGN(_Tp)> {};
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+
+#ifndef _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+template <class _Tp>
+struct is_nothrow_assignable<_Tp&, _Tp&&>
+#if defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+    : integral_constant<bool, _LIBCUDACXX_HAS_NOTHROW_ASSIGN(_Tp)> {};
+#else
+    : integral_constant<bool, is_scalar<_Tp>::value> {};
+#endif // defined(_LIBCUDACXX_HAS_NOTHROW_ASSIGN) && !defined(_LIBCUDACXX_USE_HAS_NOTHROW_ASSIGN_FALLBACK)
+
+#endif // _LIBCUDACXX_HAS_NO_RVALUE_REFERENCES
+
+#endif // !defined(_LIBCUDACXX_HAS_NO_NOEXCEPT)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp, class _Arg>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_assignable_v
+    = is_nothrow_assignable<_Tp, _Arg>::value;
+#endif
+
+// is_nothrow_copy_assignable
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_copy_assignable
+    : public is_nothrow_assignable<typename add_lvalue_reference<_Tp>::type,
+                  typename add_lvalue_reference<typename add_const<_Tp>::type>::type> {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_copy_assignable_v
+    = is_nothrow_copy_assignable<_Tp>::value;
+#endif
+
+// is_nothrow_move_assignable
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_move_assignable
+    : public is_nothrow_assignable<typename add_lvalue_reference<_Tp>::type,
+                                     typename add_rvalue_reference<_Tp>::type>
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_move_assignable_v
+    = is_nothrow_move_assignable<_Tp>::value;
+#endif
+
+// is_nothrow_destructible
+
+#if defined(_LIBCUDACXX_IS_NOTHROW_DESTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_NOTHROW_DESTRUCTIBLE_FALLBACK)
+
+template <class _Tp>
+struct is_nothrow_destructible
+   : public _CUDA_VSTD::integral_constant<bool, _LIBCUDACXX_IS_NOTHROW_DESTRUCTIBLE(_Tp)> {};
+
+#elif !defined(_LIBCUDACXX_HAS_NO_NOEXCEPT)
+
+template <bool, class _Tp> struct __libcpp_is_nothrow_destructible;
+
+template <class _Tp>
+struct __libcpp_is_nothrow_destructible<false, _Tp>
+    : public false_type
+{
+};
+
+template <class _Tp>
+struct __libcpp_is_nothrow_destructible<true, _Tp>
+    : public integral_constant<bool, noexcept(_CUDA_VSTD::declval<_Tp>().~_Tp()) >
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_destructible
+    : public __libcpp_is_nothrow_destructible<is_destructible<_Tp>::value, _Tp>
+{
+};
+
+template <class _Tp, size_t _Ns>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_destructible<_Tp[_Ns]>
+    : public is_nothrow_destructible<_Tp>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_destructible<_Tp&>
+    : public true_type
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_destructible<_Tp&&>
+    : public true_type
+{
+};
+
+#else
+
+template <class _Tp> struct __libcpp_nothrow_destructor
+    : public integral_constant<bool, is_scalar<_Tp>::value ||
+                                     is_reference<_Tp>::value> {};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_destructible
+    : public __libcpp_nothrow_destructor<typename remove_all_extents<_Tp>::type> {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_destructible<_Tp[]>
+    : public false_type {};
+
+#endif // defined(_LIBCUDACXX_IS_NOTHROW_DESTRUCTIBLE) && !defined(_LIBCUDACXX_USE_IS_NOTHROW_DESTRUCTIBLE_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_nothrow_destructible_v
+    = is_nothrow_destructible<_Tp>::value;
+#endif
+
+// is_pod
+
+#if defined(_LIBCUDACXX_IS_POD) && !defined(_LIBCUDACXX_USE_IS_POD_FALLBACK)
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_pod
+    : public integral_constant<bool, _LIBCUDACXX_IS_POD(_Tp)> {};
+
+#else
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_pod
+    : public integral_constant<bool, is_trivially_default_constructible<_Tp>::value   &&
+                                     is_trivially_copy_constructible<_Tp>::value      &&
+                                     is_trivially_copy_assignable<_Tp>::value    &&
+                                     is_trivially_destructible<_Tp>::value> {};
+
+#endif // defined(_LIBCUDACXX_IS_POD) && !defined(_LIBCUDACXX_USE_IS_POD_FALLBACK)
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_pod_v
+    = is_pod<_Tp>::value;
+#endif
+
+// is_literal_type;
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_literal_type
+#if defined(_LIBCUDACXX_IS_LITERAL) && !defined(_LIBCUDACXX_USE_IS_LITERAL_FALLBACK)
+    : public integral_constant<bool, _LIBCUDACXX_IS_LITERAL(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<typename remove_all_extents<_Tp>::type>::value ||
+                              is_reference<typename remove_all_extents<_Tp>::type>::value>
+#endif // defined(_LIBCUDACXX_IS_LITERAL) && !defined(_LIBCUDACXX_USE_IS_LITERAL_FALLBACK)
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_literal_type_v
+    = is_literal_type<_Tp>::value;
+#endif
+
+// is_standard_layout;
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_standard_layout
+#if defined(_LIBCUDACXX_IS_STANDARD_LAYOUT) && !defined(_LIBCUDACXX_USE_IS_STANDARD_LAYOUT_FALLBACK)
+    : public integral_constant<bool, _LIBCUDACXX_IS_STANDARD_LAYOUT(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<typename remove_all_extents<_Tp>::type>::value>
+#endif // defined(_LIBCUDACXX_IS_STANDARD_LAYOUT) && !defined(_LIBCUDACXX_USE_IS_STANDARD_LAYOUT_FALLBACK)
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_standard_layout_v
+    = is_standard_layout<_Tp>::value;
+#endif
+
+// is_trivially_copyable;
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivially_copyable
+#if defined(_LIBCUDACXX_IS_TRIVIALLY_COPYABLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_COPYABLE_FALLBACK)
+    : public integral_constant<bool, _LIBCUDACXX_IS_TRIVIALLY_COPYABLE(_Tp)>
+#else
+    : integral_constant<bool, is_scalar<typename remove_all_extents<_Tp>::type>::value>
+#endif // defined(_LIBCUDACXX_IS_TRIVIALLY_COPYABLE) && !defined(_LIBCUDACXX_USE_IS_TRIVIALLY_COPYABLE_FALLBACK)
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivially_copyable_v
+    = is_trivially_copyable<_Tp>::value;
+#endif
+
+// is_trivial;
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS is_trivial
+#if defined(_LIBCUDACXX_IS_TRIVIAL) && !defined(_LIBCUDACXX_USE_IS_TRIVIAL_FALLBACK)
+    : public integral_constant<bool, _LIBCUDACXX_IS_TRIVIAL(_Tp)>
+#else
+    : integral_constant<bool, is_trivially_copyable<_Tp>::value &&
+                               is_trivially_default_constructible<_Tp>::value>
+#endif // defined(_LIBCUDACXX_IS_TRIVIAL) && !defined(_LIBCUDACXX_USE_IS_TRIVIAL_FALLBACK)
+    {};
+
+#if _LIBCUDACXX_STD_VER > 11 && !defined(_LIBCUDACXX_HAS_NO_VARIABLE_TEMPLATES)
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR _LIBCUDACXX_CONSTEXPR bool is_trivial_v
+    = is_trivial<_Tp>::value;
+#endif
+
+template <class _Tp> struct __is_reference_wrapper_impl : public false_type {};
+template <class _Tp> struct __is_reference_wrapper_impl<reference_wrapper<_Tp> > : public true_type {};
+template <class _Tp> struct __is_reference_wrapper
+    : public __is_reference_wrapper_impl<typename remove_cv<_Tp>::type> {};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Fp, class _A0,
+         class _DecayFp = typename decay<_Fp>::type,
+         class _DecayA0 = typename decay<_A0>::type,
+         class _ClassT = typename __member_pointer_class_type<_DecayFp>::type>
+using __enable_if_bullet1 = typename enable_if
+    <
+        is_member_function_pointer<_DecayFp>::value
+        && is_base_of<_ClassT, _DecayA0>::value
+    >::type;
+
+template <class _Fp, class _A0,
+         class _DecayFp = typename decay<_Fp>::type,
+         class _DecayA0 = typename decay<_A0>::type>
+using __enable_if_bullet2 = typename enable_if
+    <
+        is_member_function_pointer<_DecayFp>::value
+        && __is_reference_wrapper<_DecayA0>::value
+    >::type;
+
+template <class _Fp, class _A0,
+         class _DecayFp = typename decay<_Fp>::type,
+         class _DecayA0 = typename decay<_A0>::type,
+         class _ClassT = typename __member_pointer_class_type<_DecayFp>::type>
+using __enable_if_bullet3 = typename enable_if
+    <
+        is_member_function_pointer<_DecayFp>::value
+        && !is_base_of<_ClassT, _DecayA0>::value
+        && !__is_reference_wrapper<_DecayA0>::value
+    >::type;
+
+template <class _Fp, class _A0,
+         class _DecayFp = typename decay<_Fp>::type,
+         class _DecayA0 = typename decay<_A0>::type,
+         class _ClassT = typename __member_pointer_class_type<_DecayFp>::type>
+using __enable_if_bullet4 = typename enable_if
+    <
+        is_member_object_pointer<_DecayFp>::value
+        && is_base_of<_ClassT, _DecayA0>::value
+    >::type;
+
+template <class _Fp, class _A0,
+         class _DecayFp = typename decay<_Fp>::type,
+         class _DecayA0 = typename decay<_A0>::type>
+using __enable_if_bullet5 = typename enable_if
+    <
+        is_member_object_pointer<_DecayFp>::value
+        && __is_reference_wrapper<_DecayA0>::value
+    >::type;
+
+template <class _Fp, class _A0,
+         class _DecayFp = typename decay<_Fp>::type,
+         class _DecayA0 = typename decay<_A0>::type,
+         class _ClassT = typename __member_pointer_class_type<_DecayFp>::type>
+using __enable_if_bullet6 = typename enable_if
+    <
+        is_member_object_pointer<_DecayFp>::value
+        && !is_base_of<_ClassT, _DecayA0>::value
+        && !__is_reference_wrapper<_DecayA0>::value
+    >::type;
+
+// __invoke forward declarations
+
+// fall back - none of the bullets
+
+#define _LIBCUDACXX_INVOKE_RETURN(...) \
+    noexcept(noexcept(__VA_ARGS__)) -> decltype(__VA_ARGS__) \
+    { return __VA_ARGS__; }
+
+template <class ..._Args>
+_LIBCUDACXX_INLINE_VISIBILITY
+auto __invoke(__any, _Args&& ...__args) -> __nat;
+
+template <class ..._Args>
+_LIBCUDACXX_INLINE_VISIBILITY
+auto __invoke_constexpr(__any, _Args&& ...__args) -> __nat;
+
+// bullets 1, 2 and 3
+
+template <class _Fp, class _A0, class ..._Args,
+          class = __enable_if_bullet1<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _A0&& __a0, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN((_CUDA_VSTD::forward<_A0>(__a0).*__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+template <class _Fp, class _A0, class ..._Args,
+          class = __enable_if_bullet1<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _A0&& __a0, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN((_CUDA_VSTD::forward<_A0>(__a0).*__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+template <class _Fp, class _A0, class ..._Args,
+          class = __enable_if_bullet2<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _A0&& __a0, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN((__a0.get().*__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+template <class _Fp, class _A0, class ..._Args,
+          class = __enable_if_bullet2<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _A0&& __a0, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN((__a0.get().*__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+template <class _Fp, class _A0, class ..._Args,
+          class = __enable_if_bullet3<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _A0&& __a0, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN(((*_CUDA_VSTD::forward<_A0>(__a0)).*__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+template <class _Fp, class _A0, class ..._Args,
+          class = __enable_if_bullet3<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _A0&& __a0, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN(((*_CUDA_VSTD::forward<_A0>(__a0)).*__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+// bullets 4, 5 and 6
+
+template <class _Fp, class _A0,
+          class = __enable_if_bullet4<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _A0&& __a0)
+_LIBCUDACXX_INVOKE_RETURN(_CUDA_VSTD::forward<_A0>(__a0).*__f)
+
+template <class _Fp, class _A0,
+          class = __enable_if_bullet4<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _A0&& __a0)
+_LIBCUDACXX_INVOKE_RETURN(_CUDA_VSTD::forward<_A0>(__a0).*__f)
+
+template <class _Fp, class _A0,
+          class = __enable_if_bullet5<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _A0&& __a0)
+_LIBCUDACXX_INVOKE_RETURN(__a0.get().*__f)
+
+template <class _Fp, class _A0,
+          class = __enable_if_bullet5<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _A0&& __a0)
+_LIBCUDACXX_INVOKE_RETURN(__a0.get().*__f)
+
+template <class _Fp, class _A0,
+          class = __enable_if_bullet6<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _A0&& __a0)
+_LIBCUDACXX_INVOKE_RETURN((*_CUDA_VSTD::forward<_A0>(__a0)).*__f)
+
+template <class _Fp, class _A0,
+          class = __enable_if_bullet6<_Fp, _A0>>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _A0&& __a0)
+_LIBCUDACXX_INVOKE_RETURN((*_CUDA_VSTD::forward<_A0>(__a0)).*__f)
+
+// bullet 7
+
+template <class _Fp, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto
+__invoke(_Fp&& __f, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN(_CUDA_VSTD::forward<_Fp>(__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+template <class _Fp, class ..._Args>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_CONSTEXPR auto
+__invoke_constexpr(_Fp&& __f, _Args&& ...__args)
+_LIBCUDACXX_INVOKE_RETURN(_CUDA_VSTD::forward<_Fp>(__f)(_CUDA_VSTD::forward<_Args>(__args)...))
+
+#undef _LIBCUDACXX_INVOKE_RETURN
+
+// __invokable
+template <class _Ret, class _Fp, class ..._Args>
+struct __invokable_r
+{
+  template <class _XFp, class ..._XArgs>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static auto __try_call(int) -> decltype(
+    _CUDA_VSTD::__invoke(_CUDA_VSTD::declval<_XFp>(), _CUDA_VSTD::declval<_XArgs>()...));
+  template <class _XFp, class ..._XArgs>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  static __nat __try_call(...);
+
+  // FIXME: Check that _Ret, _Fp, and _Args... are all complete types, cv void,
+  // or incomplete array types as required by the standard.
+  using _Result = decltype(__try_call<_Fp, _Args...>(0));
+
+  using type =
+  typename conditional<
+      _IsNotSame<_Result, __nat>::value,
+      typename conditional<
+          is_void<_Ret>::value,
+          true_type,
+          is_convertible<_Result, _Ret>
+      >::type,
+      false_type
+  >::type;
+  static const bool value = type::value;
+};
+template <class _Fp, class ..._Args>
+using __invokable = __invokable_r<void, _Fp, _Args...>;
+
+template <bool _IsInvokable, bool _IsCVVoid, class _Ret, class _Fp, class ..._Args>
+struct __nothrow_invokable_r_imp {
+  static const bool value = false;
+};
+
+template <class _Ret, class _Fp, class ..._Args>
+struct __nothrow_invokable_r_imp<true, false, _Ret, _Fp, _Args...>
+{
+    typedef __nothrow_invokable_r_imp _ThisT;
+
+    template <class _Tp>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static void __test_noexcept(_Tp) noexcept;
+
+    static const bool value = noexcept(_ThisT::__test_noexcept<_Ret>(
+        _CUDA_VSTD::__invoke(_CUDA_VSTD::declval<_Fp>(), _CUDA_VSTD::declval<_Args>()...)));
+};
+
+template <class _Ret, class _Fp, class ..._Args>
+struct __nothrow_invokable_r_imp<true, true, _Ret, _Fp, _Args...>
+{
+    static const bool value = noexcept(
+        _CUDA_VSTD::__invoke(_CUDA_VSTD::declval<_Fp>(), _CUDA_VSTD::declval<_Args>()...));
+};
+
+template <class _Ret, class _Fp, class ..._Args>
+using __nothrow_invokable_r =
+    __nothrow_invokable_r_imp<
+            __invokable_r<_Ret, _Fp, _Args...>::value,
+            is_void<_Ret>::value,
+            _Ret, _Fp, _Args...
+    >;
+
+template <class _Fp, class ..._Args>
+using __nothrow_invokable =
+    __nothrow_invokable_r_imp<
+            __invokable<_Fp, _Args...>::value,
+            true, void, _Fp, _Args...
+    >;
+
+template <class _Fp, class ..._Args>
+struct __invoke_of
+    : public enable_if<
+        __invokable<_Fp, _Args...>::value,
+        typename __invokable_r<void, _Fp, _Args...>::_Result>
+{
+};
+
+// result_of
+
+template <class _Fp, class ..._Args>
+class _LIBCUDACXX_TEMPLATE_VIS result_of<_Fp(_Args...)>
+    : public __invoke_of<_Fp, _Args...>
+{
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using result_of_t = typename result_of<_Tp>::type;
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11
+
+// invoke_result
+
+template <class _Fn, class... _Args>
+struct _LIBCUDACXX_TEMPLATE_VIS invoke_result
+    : __invoke_of<_Fn, _Args...>
+{
+};
+
+template <class _Fn, class... _Args>
+using invoke_result_t = typename invoke_result<_Fn, _Args...>::type;
+
+// is_invocable
+
+template <class _Fn, class ..._Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_invocable
+    : integral_constant<bool, __invokable<_Fn, _Args...>::value> {};
+
+template <class _Ret, class _Fn, class ..._Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_invocable_r
+    : integral_constant<bool, __invokable_r<_Ret, _Fn, _Args...>::value> {};
+
+template <class _Fn, class ..._Args>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_invocable_v
+    = is_invocable<_Fn, _Args...>::value;
+
+template <class _Ret, class _Fn, class ..._Args>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_invocable_r_v
+    = is_invocable_r<_Ret, _Fn, _Args...>::value;
+
+// is_nothrow_invocable
+
+template <class _Fn, class ..._Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_invocable
+    : integral_constant<bool, __nothrow_invokable<_Fn, _Args...>::value> {};
+
+template <class _Ret, class _Fn, class ..._Args>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_invocable_r
+    : integral_constant<bool, __nothrow_invokable_r<_Ret, _Fn, _Args...>::value> {};
+
+template <class _Fn, class ..._Args>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_nothrow_invocable_v
+    = is_nothrow_invocable<_Fn, _Args...>::value;
+
+template <class _Ret, class _Fn, class ..._Args>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_nothrow_invocable_r_v
+    = is_nothrow_invocable_r<_Ret, _Fn, _Args...>::value;
+
+#endif // _LIBCUDACXX_STD_VER > 11
+
+#endif  // !defined(_LIBCUDACXX_CXX03_LANG)
+
+template <class _Tp> struct __is_swappable;
+template <class _Tp> struct __is_nothrow_swappable;
+
+// swap, swap_ranges
+
+template <class _ForwardIterator1, class _ForwardIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator2
+swap_ranges(_ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2);
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+#ifndef _LIBCUDACXX_CXX03_LANG
+typename enable_if
+<
+    is_move_constructible<_Tp>::value &&
+    is_move_assignable<_Tp>::value
+>::type
+#else
+void
+#endif
+_LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+swap(_Tp& __x, _Tp& __y) _NOEXCEPT_(is_nothrow_move_constructible<_Tp>::value &&
+                                    is_nothrow_move_assignable<_Tp>::value)
+{
+    _Tp __t(_CUDA_VSTD::move(__x));
+    __x = _CUDA_VSTD::move(__y);
+    __y = _CUDA_VSTD::move(__t);
+}
+
+template<class _Tp, size_t _Np>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+typename enable_if<
+    __is_swappable<_Tp>::value
+>::type
+swap(_Tp (&__a)[_Np], _Tp (&__b)[_Np]) _NOEXCEPT_(__is_nothrow_swappable<_Tp>::value)
+{
+    _CUDA_VSTD::swap_ranges(__a, __a + _Np, __b);
+}
+
+template <class _ForwardIterator1, class _ForwardIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_ForwardIterator2
+swap_ranges(_ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2)
+{
+    for(; __first1 != __last1; ++__first1, (void) ++__first2)
+        swap(*__first1, *__first2);
+    return __first2;
+}
+
+// iter_swap
+
+template <class _ForwardIterator1, class _ForwardIterator2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+void
+iter_swap(_ForwardIterator1 __a, _ForwardIterator2 __b)
+    //                                  _NOEXCEPT_(_NOEXCEPT_(swap(*__a, *__b)))
+               _NOEXCEPT_(_NOEXCEPT_(swap(*_CUDA_VSTD::declval<_ForwardIterator1>(),
+                                          *_CUDA_VSTD::declval<_ForwardIterator2>())))
+{
+    swap(*__a, *__b);
+}
+
+// __swappable
+
+namespace __detail
+{
+// ALL generic swap overloads MUST already have a declaration available at this point.
+
+template <class _Tp, class _Up = _Tp,
+          bool _NotVoid = !is_void<_Tp>::value && !is_void<_Up>::value>
+struct __swappable_with
+{
+    template <class _LHS, class _RHS>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static decltype(swap(_CUDA_VSTD::declval<_LHS>(), _CUDA_VSTD::declval<_RHS>()))
+    __test_swap(int);
+    template <class, class>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static __nat __test_swap(long);
+
+    // Extra parens are needed for the C++03 definition of decltype.
+    typedef decltype((__test_swap<_Tp, _Up>(0))) __swap1;
+    typedef decltype((__test_swap<_Up, _Tp>(0))) __swap2;
+
+    static const bool value = _IsNotSame<__swap1, __nat>::value
+                           && _IsNotSame<__swap2, __nat>::value;
+};
+
+template <class _Tp, class _Up>
+struct __swappable_with<_Tp, _Up,  false> : false_type {};
+
+template <class _Tp, class _Up = _Tp, bool _Swappable = __swappable_with<_Tp, _Up>::value>
+struct __nothrow_swappable_with {
+  static const bool value =
+#ifndef _LIBCUDACXX_HAS_NO_NOEXCEPT
+      noexcept(swap(_CUDA_VSTD::declval<_Tp>(), _CUDA_VSTD::declval<_Up>()))
+  &&  noexcept(swap(_CUDA_VSTD::declval<_Up>(), _CUDA_VSTD::declval<_Tp>()));
+#else
+      false;
+#endif
+};
+
+template <class _Tp, class _Up>
+struct __nothrow_swappable_with<_Tp, _Up, false> : false_type {};
+
+}  // __detail
+
+template <class _Tp>
+struct __is_swappable
+    : public integral_constant<bool, __detail::__swappable_with<_Tp&>::value>
+{
+};
+
+template <class _Tp>
+struct __is_nothrow_swappable
+    : public integral_constant<bool, __detail::__nothrow_swappable_with<_Tp&>::value>
+{
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template <class _Tp, class _Up>
+struct _LIBCUDACXX_TEMPLATE_VIS is_swappable_with
+    : public integral_constant<bool, __detail::__swappable_with<_Tp, _Up>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_swappable
+    : public conditional<
+        __is_referenceable<_Tp>::value,
+        is_swappable_with<
+            typename add_lvalue_reference<_Tp>::type,
+            typename add_lvalue_reference<_Tp>::type>,
+        false_type
+    >::type
+{
+};
+
+template <class _Tp, class _Up>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_swappable_with
+    : public integral_constant<bool, __detail::__nothrow_swappable_with<_Tp, _Up>::value>
+{
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS is_nothrow_swappable
+    : public conditional<
+        __is_referenceable<_Tp>::value,
+        is_nothrow_swappable_with<
+            typename add_lvalue_reference<_Tp>::type,
+            typename add_lvalue_reference<_Tp>::type>,
+        false_type
+    >::type
+{
+};
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_swappable_with_v
+    = is_swappable_with<_Tp, _Up>::value;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_swappable_v
+    = is_swappable<_Tp>::value;
+
+template <class _Tp, class _Up>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_nothrow_swappable_with_v
+    = is_nothrow_swappable_with<_Tp, _Up>::value;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool is_nothrow_swappable_v
+    = is_nothrow_swappable<_Tp>::value;
+
+#endif // _LIBCUDACXX_STD_VER > 11
+
+#if defined(_LIBCUDACXX_UNDERLYING_TYPE) && !defined(_LIBCUDACXX_USE_UNDERLYING_TYPE_FALLBACK)
+template <class _Tp, bool = is_enum<_Tp>::value> struct __underlying_type_impl;
+
+template <class _Tp>
+struct __underlying_type_impl<_Tp, false> {};
+
+template <class _Tp>
+struct __underlying_type_impl<_Tp, true>
+{
+    typedef _LIBCUDACXX_UNDERLYING_TYPE(_Tp) type;
+};
+
+template <class _Tp>
+struct underlying_type : __underlying_type_impl<_Tp, is_enum<_Tp>::value> {};
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp> using underlying_type_t = typename underlying_type<_Tp>::type;
+#endif
+
+#else
+
+template <class _Tp, bool _Support = false>
+struct underlying_type
+{
+    static_assert(_Support, "The underyling_type trait requires compiler "
+                            "support. Either no such support exists or "
+                            "libc++ does not know how to use it.");
+};
+
+#endif // defined(_LIBCUDACXX_UNDERLYING_TYPE) && !defined(_LIBCUDACXX_USE_UNDERLYING_TYPE_FALLBACK)
+
+
+template <class _Tp, bool = is_enum<_Tp>::value>
+struct __sfinae_underlying_type
+{
+    typedef typename underlying_type<_Tp>::type type;
+    typedef decltype(((type)1) + 0) __promoted_type;
+};
+
+template <class _Tp>
+struct __sfinae_underlying_type<_Tp, false> {};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+int __convert_to_integral(int __val) { return __val; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+unsigned __convert_to_integral(unsigned __val) { return __val; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+long __convert_to_integral(long __val) { return __val; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+unsigned long __convert_to_integral(unsigned long __val) { return __val; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+long long __convert_to_integral(long long __val) { return __val; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+unsigned long long __convert_to_integral(unsigned long long __val) {return __val; }
+
+template<typename _Fp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename enable_if<is_floating_point<_Fp>::value, long long>::type
+ __convert_to_integral(_Fp __val) { return __val; }
+
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+__int128_t __convert_to_integral(__int128_t __val) { return __val; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+__uint128_t __convert_to_integral(__uint128_t __val) { return __val; }
+#endif
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+typename __sfinae_underlying_type<_Tp>::__promoted_type
+__convert_to_integral(_Tp __val) { return __val; }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+struct __has_operator_addressof_member_imp
+{
+    template <class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static auto __test(int)
+            -> typename __select_2nd<decltype(_CUDA_VSTD::declval<_Up>().operator&()), true_type>::type;
+    template <class>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static auto __test(long) -> false_type;
+
+    static const bool value = decltype(__test<_Tp>(0))::value;
+};
+
+template <class _Tp>
+struct __has_operator_addressof_free_imp
+{
+    template <class _Up>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static auto __test(int)
+            -> typename __select_2nd<decltype(operator&(_CUDA_VSTD::declval<_Up>())), true_type>::type;
+    template <class>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        static auto __test(long) -> false_type;
+
+    static const bool value = decltype(__test<_Tp>(0))::value;
+};
+
+template <class _Tp>
+struct __has_operator_addressof
+    : public integral_constant<bool, __has_operator_addressof_member_imp<_Tp>::value
+                                  || __has_operator_addressof_free_imp<_Tp>::value>
+{};
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template <class...> using void_t = void;
+
+template <class... _Args>
+struct conjunction
+#if defined(_LIBCUDACXX_HAS_NO_LOGICAL_METAFUNCTION_ALIASES)
+    : _MetaBase< sizeof...(_Args) != 0 >::template _AndImpl<true_type, _Args...>
+#else
+    : _And<_Args...>
+#endif
+    {};
+template<class... _Args>
+_LIBCUDACXX_INLINE_VAR constexpr bool conjunction_v
+    = conjunction<_Args...>::value;
+
+template <class... _Args>
+struct disjunction
+#if defined(_LIBCUDACXX_HAS_NO_LOGICAL_METAFUNCTION_ALIASES)
+    : _MetaBase< sizeof...(_Args) != 0 >::template _OrImpl<false_type, _Args...>
+#else
+    : _Or<_Args...>
+#endif
+    {};
+template<class... _Args>
+_LIBCUDACXX_INLINE_VAR constexpr bool disjunction_v
+    = disjunction<_Args...>::value;
+
+template <class _Tp>
+struct negation : _Not<_Tp> {};
+template<class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr bool negation_v
+    = negation<_Tp>::value;
+#endif  // _LIBCUDACXX_STD_VER > 11
+
+// These traits are used in __tree and __hash_table
+#ifndef _LIBCUDACXX_CXX03_LANG
+struct __extract_key_fail_tag {};
+struct __extract_key_self_tag {};
+struct __extract_key_first_tag {};
+
+template <class _ValTy, class _Key,
+          class _RawValTy = typename __unconstref<_ValTy>::type>
+struct __can_extract_key
+    : conditional<_IsSame<_RawValTy, _Key>::value, __extract_key_self_tag,
+                  __extract_key_fail_tag>::type {};
+
+template <class _Pair, class _Key, class _First, class _Second>
+struct __can_extract_key<_Pair, _Key, pair<_First, _Second>>
+    : conditional<_IsSame<typename remove_const<_First>::type, _Key>::value,
+                  __extract_key_first_tag, __extract_key_fail_tag>::type {};
+
+// __can_extract_map_key uses true_type/false_type instead of the tags.
+// It returns true if _Key != _ContainerValueTy (the container is a map not a set)
+// and _ValTy == _Key.
+template <class _ValTy, class _Key, class _ContainerValueTy,
+          class _RawValTy = typename __unconstref<_ValTy>::type>
+struct __can_extract_map_key
+    : integral_constant<bool, _IsSame<_RawValTy, _Key>::value> {};
+
+// This specialization returns __extract_key_fail_tag for non-map containers
+// because _Key == _ContainerValueTy
+template <class _ValTy, class _Key, class _RawValTy>
+struct __can_extract_map_key<_ValTy, _Key, _Key, _RawValTy>
+    : false_type {};
+
+#endif
+
+#if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED)
+#if (defined(__cuda_std__) && _LIBCUDACXX_STD_VER >= 11) || _LIBCUDACXX_STD_VER > 17
+_LIBCUDACXX_INLINE_VISIBILITY
+inline constexpr bool is_constant_evaluated() noexcept {
+  return _LIBCUDACXX_IS_CONSTANT_EVALUATED();
+}
+#endif
+
+inline _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+bool __libcpp_is_constant_evaluated() _NOEXCEPT { return _LIBCUDACXX_IS_CONSTANT_EVALUATED(); }
+#else
+inline _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+bool __libcpp_is_constant_evaluated() _NOEXCEPT { return false; }
+#endif // defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED)
+
+template <class _CharT>
+using _IsCharLikeType = _And<is_standard_layout<_CharT>, is_trivial<_CharT> >;
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 11
+// std::byte
+#ifdef _LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+_LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+#else
+namespace std  // purposefully not versioned
+{
+#endif //_LIBCUDACXX_BEGIN_NAMESPACE_STD_NOVERSION
+template <class _Integer>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr typename enable_if<is_integral_v<_Integer>, byte>::type &
+  operator<<=(byte& __lhs, _Integer __shift) noexcept
+  { return __lhs = __lhs << __shift; }
+
+template <class _Integer>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr typename enable_if<is_integral_v<_Integer>, byte>::type
+  operator<< (byte  __lhs, _Integer __shift) noexcept
+  { return static_cast<byte>(static_cast<unsigned char>(static_cast<unsigned int>(__lhs) << __shift)); }
+
+template <class _Integer>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr typename enable_if<is_integral_v<_Integer>, byte>::type &
+  operator>>=(byte& __lhs, _Integer __shift) noexcept
+  { return __lhs = __lhs >> __shift; }
+
+template <class _Integer>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr typename enable_if<is_integral_v<_Integer>, byte>::type
+  operator>> (byte  __lhs, _Integer __shift) noexcept
+  { return static_cast<byte>(static_cast<unsigned char>(static_cast<unsigned int>(__lhs) >> __shift)); }
+
+template <class _Integer>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr typename enable_if<is_integral_v<_Integer>, _Integer>::type
+  to_integer(byte __b) noexcept { return static_cast<_Integer>(__b); }
+
+#ifdef _LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+_LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+#else
+}
+#endif //_LIBCUDACXX_END_NAMESPACE_STD_NOVERSION
+#endif // _LIBCUDACXX_STD_VER > 11
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_TYPE_TRAITS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex
new file mode 100644
index 000000000000..59954c0383d9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeindex
@@ -0,0 +1,102 @@
+// -*- C++ -*-
+//===-------------------------- typeindex ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_TYPEINDEX
+#define _LIBCUDACXX_TYPEINDEX
+
+/*
+
+    typeindex synopsis
+
+namespace std
+{
+
+class type_index
+{
+public:
+    type_index(const type_info& rhs) noexcept;
+
+    bool operator==(const type_index& rhs) const noexcept;
+    bool operator!=(const type_index& rhs) const noexcept;
+    bool operator< (const type_index& rhs) const noexcept;
+    bool operator<=(const type_index& rhs) const noexcept;
+    bool operator> (const type_index& rhs) const noexcept;
+    bool operator>=(const type_index& rhs) const noexcept;
+
+    size_t hash_code() const noexcept;
+    const char* name() const noexcept;
+};
+
+template <>
+struct hash<type_index>
+    : public unary_function<type_index, size_t>
+{
+    size_t operator()(type_index index) const noexcept;
+};
+
+}  // std
+
+*/
+
+#include <__config>
+#include <typeinfo>
+#include <__functional_base>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+class _LIBCUDACXX_TEMPLATE_VIS type_index
+{
+    const type_info* __t_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    type_index(const type_info& __y) _NOEXCEPT : __t_(&__y) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const type_index& __y) const _NOEXCEPT
+        {return *__t_ == *__y.__t_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const type_index& __y) const _NOEXCEPT
+        {return *__t_ != *__y.__t_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator< (const type_index& __y) const _NOEXCEPT
+        {return  __t_->before(*__y.__t_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator<=(const type_index& __y) const _NOEXCEPT
+        {return !__y.__t_->before(*__t_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator> (const type_index& __y) const _NOEXCEPT
+        {return  __y.__t_->before(*__t_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator>=(const type_index& __y) const _NOEXCEPT
+        {return !__t_->before(*__y.__t_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t hash_code() const _NOEXCEPT {return __t_->hash_code();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char* name() const _NOEXCEPT {return __t_->name();}
+};
+
+template <class _Tp> struct _LIBCUDACXX_TEMPLATE_VIS hash;
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<type_index>
+    : public unary_function<type_index, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(type_index __index) const _NOEXCEPT
+        {return __index.hash_code();}
+};
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_TYPEINDEX
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo
new file mode 100644
index 000000000000..366031fd91d9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/typeinfo
@@ -0,0 +1,350 @@
+// -*- C++ -*-
+//===-------------------------- typeinfo ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __LIBCUDACXX_TYPEINFO
+#define __LIBCUDACXX_TYPEINFO
+
+/*
+
+    typeinfo synopsis
+
+namespace std {
+
+class type_info
+{
+public:
+    virtual ~type_info();
+
+    bool operator==(const type_info& rhs) const noexcept;
+    bool operator!=(const type_info& rhs) const noexcept;
+
+    bool before(const type_info& rhs) const noexcept;
+    size_t hash_code() const noexcept;
+    const char* name() const noexcept;
+
+    type_info(const type_info& rhs) = delete;
+    type_info& operator=(const type_info& rhs) = delete;
+};
+
+class bad_cast
+    : public exception
+{
+public:
+    bad_cast() noexcept;
+    bad_cast(const bad_cast&) noexcept;
+    bad_cast& operator=(const bad_cast&) noexcept;
+    virtual const char* what() const noexcept;
+};
+
+class bad_typeid
+    : public exception
+{
+public:
+    bad_typeid() noexcept;
+    bad_typeid(const bad_typeid&) noexcept;
+    bad_typeid& operator=(const bad_typeid&) noexcept;
+    virtual const char* what() const noexcept;
+};
+
+}  // std
+
+*/
+
+#include <__config>
+#include <exception>
+#include <cstddef>
+#include <cstdint>
+#ifdef _LIBCUDACXX_NO_EXCEPTIONS
+#include <cstdlib>
+#endif
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if defined(_LIBCUDACXX_ABI_VCRUNTIME)
+#include <vcruntime_typeinfo.h>
+#else
+
+namespace std  // purposefully not using versioning namespace
+{
+
+
+#if defined(_LIBCUDACXX_ABI_MICROSOFT)
+
+class _LIBCUDACXX_EXCEPTION_ABI type_info
+{
+    type_info& operator=(const type_info&);
+    type_info(const type_info&);
+
+    mutable struct {
+      const char *__undecorated_name;
+      const char __decorated_name[1];
+    } __data;
+
+    int __compare(const type_info &__rhs) const _NOEXCEPT;
+
+public:
+    _LIBCUDACXX_AVAILABILITY_TYPEINFO_VTABLE
+    virtual ~type_info();
+
+    const char *name() const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool before(const type_info& __arg) const _NOEXCEPT {
+      return __compare(__arg) < 0;
+    }
+
+    size_t hash_code() const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const type_info& __arg) const _NOEXCEPT {
+      return __compare(__arg) == 0;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const type_info& __arg) const _NOEXCEPT
+    { return !operator==(__arg); }
+};
+
+#else // !defined(_LIBCUDACXX_ABI_MICROSOFT)
+
+// ========================================================================== //
+//                           Implementations
+// ========================================================================== //
+// ------------------------------------------------------------------------- //
+//                               Unique
+// ------------------------------------------------------------------------- //
+// This implementation of type_info assumes a unique copy of the RTTI for a
+// given type inside a program. This is a valid assumption when abiding to
+// Itanium ABI (http://itanium-cxx-abi.github.io/cxx-abi/abi.html#vtable-components).
+// Under this assumption, we can always compare the addresses of the type names
+// to implement equality-comparison of type_infos instead of having to perform
+// a deep string comparison.
+// -------------------------------------------------------------------------- //
+//                             NonUnique
+// -------------------------------------------------------------------------- //
+// This implementation of type_info does not assume there is always a unique
+// copy of the RTTI for a given type inside a program. For various reasons
+// the linker may have failed to merge every copy of a types RTTI
+// (For example: -Bsymbolic or llvm.org/PR37398). Under this assumption, two
+// type_infos are equal if their addresses are equal or if a deep string
+// comparison is equal.
+// -------------------------------------------------------------------------- //
+//                          NonUniqueARMRTTIBit
+// -------------------------------------------------------------------------- //
+// This implementation of type_info does not assume always a unique copy of
+// the RTTI for a given type inside a program. It packs the pointer to the
+// type name into a uintptr_t and reserves the high bit of that pointer (which
+// is assumed to be free for use under the ABI in use) to represent whether
+// that specific copy of the RTTI can be assumed unique inside the program.
+// To implement equality-comparison of type_infos, we check whether BOTH
+// type_infos are guaranteed unique, and if so, we simply compare the addresses
+// of their type names instead of doing a deep string comparison, which is
+// faster. If at least one of the type_infos can't guarantee uniqueness, we
+// have no choice but to fall back to a deep string comparison.
+//
+// This implementation is specific to ARM64 on Apple platforms.
+//
+// Note that the compiler is the one setting (or unsetting) the high bit of
+// the pointer when it constructs the type_info, depending on whether it can
+// guarantee uniqueness for that specific type_info.
+
+struct __type_info_implementations {
+  struct __string_impl_base {
+    typedef const char* __type_name_t;
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    _LIBCUDACXX_CONSTEXPR static const char* __type_name_to_string(__type_name_t __v) _NOEXCEPT {
+      return __v;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    _LIBCUDACXX_CONSTEXPR static __type_name_t __string_to_type_name(const char* __v) _NOEXCEPT {
+      return __v;
+    }
+  };
+
+  struct __unique_impl : __string_impl_base {
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static size_t __hash(__type_name_t __v) _NOEXCEPT {
+      return reinterpret_cast<size_t>(__v);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static bool __eq(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      return __lhs == __rhs;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static bool __lt(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      return __lhs < __rhs;
+    }
+  };
+
+  struct __non_unique_impl : __string_impl_base {
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static size_t __hash(__type_name_t __ptr) _NOEXCEPT {
+      size_t __hash = 5381;
+      while (unsigned char __c = static_cast<unsigned char>(*__ptr++))
+        __hash = (__hash * 33) ^ __c;
+      return __hash;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static bool __eq(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      return __lhs == __rhs || __builtin_strcmp(__lhs, __rhs) == 0;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static bool __lt(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      return __builtin_strcmp(__lhs, __rhs) < 0;
+    }
+  };
+
+  struct __non_unique_arm_rtti_bit_impl {
+    typedef uintptr_t __type_name_t;
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static const char* __type_name_to_string(__type_name_t __v) _NOEXCEPT {
+      return reinterpret_cast<const char*>(__v &
+          ~__non_unique_rtti_bit::value);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static __type_name_t __string_to_type_name(const char* __v) _NOEXCEPT {
+      return reinterpret_cast<__type_name_t>(__v);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static size_t __hash(__type_name_t __v) _NOEXCEPT {
+      if (__is_type_name_unique(__v))
+        return reinterpret_cast<size_t>(__v);
+      return __non_unique_impl::__hash(__type_name_to_string(__v));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static bool __eq(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      if (__lhs == __rhs)
+        return true;
+      if (__is_type_name_unique(__lhs, __rhs))
+        return false;
+      return __builtin_strcmp(__type_name_to_string(__lhs), __type_name_to_string(__rhs)) == 0;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_ALWAYS_INLINE
+    static bool __lt(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      if (__is_type_name_unique(__lhs, __rhs))
+        return __lhs < __rhs;
+      return __builtin_strcmp(__type_name_to_string(__lhs), __type_name_to_string(__rhs)) < 0;
+    }
+
+   private:
+    // The unique bit is the top bit. It is expected that __type_name_t is 64 bits when
+    // this implementation is actually used.
+    typedef std::integral_constant<__type_name_t,
+      (1ULL << ((__CHAR_BIT__ * sizeof(__type_name_t)) - 1))> __non_unique_rtti_bit;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static bool __is_type_name_unique(__type_name_t __lhs) _NOEXCEPT {
+      return !(__lhs & __non_unique_rtti_bit::value);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static bool __is_type_name_unique(__type_name_t __lhs, __type_name_t __rhs) _NOEXCEPT {
+      return !((__lhs & __rhs) & __non_unique_rtti_bit::value);
+    }
+  };
+
+  typedef
+#if defined(__APPLE__) && defined(__LP64__) && !defined(__x86_64__)
+    __non_unique_arm_rtti_bit_impl
+#elif _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT == 0
+    __non_unique_impl
+#elif _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT == 1
+    __unique_impl
+#else
+#   error invalid configuration for _LIBCUDACXX_HAS_MERGED_TYPEINFO_NAMES_DEFAULT
+#endif
+     __impl;
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI type_info
+{
+  type_info& operator=(const type_info&);
+  type_info(const type_info&);
+
+ protected:
+    typedef __type_info_implementations::__impl __impl;
+
+    __impl::__type_name_t __type_name;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit type_info(const char* __n)
+      : __type_name(__impl::__string_to_type_name(__n)) {}
+
+public:
+    _LIBCUDACXX_AVAILABILITY_TYPEINFO_VTABLE
+    virtual ~type_info();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const char* name() const _NOEXCEPT
+    {
+      return __impl::__type_name_to_string(__type_name);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool before(const type_info& __arg) const _NOEXCEPT
+    {
+      return __impl::__lt(__type_name, __arg.__type_name);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t hash_code() const _NOEXCEPT
+    {
+      return __impl::__hash(__type_name);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator==(const type_info& __arg) const _NOEXCEPT
+    {
+      return __impl::__eq(__type_name, __arg.__type_name);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator!=(const type_info& __arg) const _NOEXCEPT
+    { return !operator==(__arg); }
+};
+#endif // defined(_LIBCUDACXX_ABI_MICROSOFT)
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_cast
+    : public exception
+{
+ public:
+  bad_cast() _NOEXCEPT;
+  virtual ~bad_cast() _NOEXCEPT;
+  virtual const char* what() const _NOEXCEPT;
+};
+
+class _LIBCUDACXX_EXCEPTION_ABI bad_typeid
+    : public exception
+{
+ public:
+  bad_typeid() _NOEXCEPT;
+  virtual ~bad_typeid() _NOEXCEPT;
+  virtual const char* what() const _NOEXCEPT;
+};
+
+}  // std
+
+#endif // defined(_LIBCUDACXX_ABI_VCRUNTIME)
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+_LIBCUDACXX_NORETURN inline _LIBCUDACXX_INLINE_VISIBILITY
+void __throw_bad_cast()
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    throw bad_cast();
+#else
+    _CUDA_VSTD::abort();
+#endif
+}
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // __LIBCUDACXX_TYPEINFO
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map
new file mode 100644
index 000000000000..df310bc99e74
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_map
@@ -0,0 +1,2445 @@
+// -*- C++ -*-
+//===-------------------------- unordered_map -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_UNORDERED_MAP
+#define _LIBCUDACXX_UNORDERED_MAP
+
+/*
+
+    unordered_map synopsis
+
+#include <initializer_list>
+
+namespace std
+{
+
+template <class Key, class T, class Hash = hash<Key>, class Pred = equal_to<Key>,
+          class Alloc = allocator<pair<const Key, T>>>
+class unordered_map
+{
+public:
+    // types
+    typedef Key                                                        key_type;
+    typedef T                                                          mapped_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef pair<const key_type, mapped_type>                          value_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+    typedef /unspecified/ local_iterator;
+    typedef /unspecified/ const_local_iterator;
+
+    typedef unspecified                             node_type;            // C++17
+    typedef INSERT_RETURN_TYPE<iterator, node_type> insert_return_type;   // C++17
+
+    unordered_map()
+        noexcept(
+            is_nothrow_default_constructible<hasher>::value &&
+            is_nothrow_default_constructible<key_equal>::value &&
+            is_nothrow_default_constructible<allocator_type>::value);
+    explicit unordered_map(size_type n, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        unordered_map(InputIterator f, InputIterator l,
+                      size_type n = 0, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    explicit unordered_map(const allocator_type&);
+    unordered_map(const unordered_map&);
+    unordered_map(const unordered_map&, const Allocator&);
+    unordered_map(unordered_map&&)
+        noexcept(
+            is_nothrow_move_constructible<hasher>::value &&
+            is_nothrow_move_constructible<key_equal>::value &&
+            is_nothrow_move_constructible<allocator_type>::value);
+    unordered_map(unordered_map&&, const Allocator&);
+    unordered_map(initializer_list<value_type>, size_type n = 0,
+                  const hasher& hf = hasher(), const key_equal& eql = key_equal(),
+                  const allocator_type& a = allocator_type());
+    unordered_map(size_type n, const allocator_type& a)
+      : unordered_map(n, hasher(), key_equal(), a) {}  // C++14
+    unordered_map(size_type n, const hasher& hf, const allocator_type& a)
+      : unordered_map(n, hf, key_equal(), a) {}  // C++14
+    template <class InputIterator>
+      unordered_map(InputIterator f, InputIterator l, size_type n, const allocator_type& a)
+      : unordered_map(f, l, n, hasher(), key_equal(), a) {}  // C++14
+    template <class InputIterator>
+      unordered_map(InputIterator f, InputIterator l, size_type n, const hasher& hf,
+        const allocator_type& a)
+      : unordered_map(f, l, n, hf, key_equal(), a) {}  // C++14
+    unordered_map(initializer_list<value_type> il, size_type n, const allocator_type& a)
+      : unordered_map(il, n, hasher(), key_equal(), a) {}  // C++14
+    unordered_map(initializer_list<value_type> il, size_type n, const hasher& hf,
+      const allocator_type& a)
+      : unordered_map(il, n, hf, key_equal(), a) {}  // C++14
+    ~unordered_map();
+    unordered_map& operator=(const unordered_map&);
+    unordered_map& operator=(unordered_map&&)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<hasher>::value &&
+            is_nothrow_move_assignable<key_equal>::value);
+    unordered_map& operator=(initializer_list<value_type>);
+
+    allocator_type get_allocator() const noexcept;
+
+    bool      empty() const noexcept;
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+
+    iterator       begin() noexcept;
+    iterator       end() noexcept;
+    const_iterator begin()  const noexcept;
+    const_iterator end()    const noexcept;
+    const_iterator cbegin() const noexcept;
+    const_iterator cend()   const noexcept;
+
+    template <class... Args>
+        pair<iterator, bool> emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    pair<iterator, bool> insert(const value_type& obj);
+    template <class P>
+        pair<iterator, bool> insert(P&& obj);
+    iterator insert(const_iterator hint, const value_type& obj);
+    template <class P>
+        iterator insert(const_iterator hint, P&& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type>);
+
+    node_type extract(const_iterator position);                                       // C++17
+    node_type extract(const key_type& x);                                             // C++17
+    insert_return_type insert(node_type&& nh);                                        // C++17
+    iterator           insert(const_iterator hint, node_type&& nh);                   // C++17
+
+    template <class... Args>
+        pair<iterator, bool> try_emplace(const key_type& k, Args&&... args);          // C++17
+    template <class... Args>
+        pair<iterator, bool> try_emplace(key_type&& k, Args&&... args);               // C++17
+    template <class... Args>
+        iterator try_emplace(const_iterator hint, const key_type& k, Args&&... args); // C++17
+    template <class... Args>
+        iterator try_emplace(const_iterator hint, key_type&& k, Args&&... args);      // C++17
+    template <class M>
+        pair<iterator, bool> insert_or_assign(const key_type& k, M&& obj);            // C++17
+    template <class M>
+        pair<iterator, bool> insert_or_assign(key_type&& k, M&& obj);                 // C++17
+    template <class M>
+        iterator insert_or_assign(const_iterator hint, const key_type& k, M&& obj);   // C++17
+    template <class M>
+        iterator insert_or_assign(const_iterator hint, key_type&& k, M&& obj);        // C++17
+
+    iterator erase(const_iterator position);
+    iterator erase(iterator position);  // C++14
+    size_type erase(const key_type& k);
+    iterator erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class H2, class P2>
+      void merge(unordered_map<Key, T, H2, P2, Allocator>& source);         // C++17
+    template<class H2, class P2>
+      void merge(unordered_map<Key, T, H2, P2, Allocator>&& source);        // C++17
+    template<class H2, class P2>
+      void merge(unordered_multimap<Key, T, H2, P2, Allocator>& source);    // C++17
+    template<class H2, class P2>
+      void merge(unordered_multimap<Key, T, H2, P2, Allocator>&& source);   // C++17
+
+    void swap(unordered_map&)
+        noexcept(
+            (!allocator_type::propagate_on_container_swap::value ||
+             __is_nothrow_swappable<allocator_type>::value) &&
+            __is_nothrow_swappable<hasher>::value &&
+            __is_nothrow_swappable<key_equal>::value);
+
+    hasher hash_function() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    bool contains(const key_type& k) const; // C++20
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    mapped_type& operator[](const key_type& k);
+    mapped_type& operator[](key_type&& k);
+
+    mapped_type&       at(const key_type& k);
+    const mapped_type& at(const key_type& k) const;
+
+    size_type bucket_count() const noexcept;
+    size_type max_bucket_count() const noexcept;
+
+    size_type bucket_size(size_type n) const;
+    size_type bucket(const key_type& k) const;
+
+    local_iterator       begin(size_type n);
+    local_iterator       end(size_type n);
+    const_local_iterator begin(size_type n) const;
+    const_local_iterator end(size_type n) const;
+    const_local_iterator cbegin(size_type n) const;
+    const_local_iterator cend(size_type n) const;
+
+    float load_factor() const noexcept;
+    float max_load_factor() const noexcept;
+    void max_load_factor(float z);
+    void rehash(size_type n);
+    void reserve(size_type n);
+};
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x,
+              unordered_map<Key, T, Hash, Pred, Alloc>& y)
+              noexcept(noexcept(x.swap(y)));
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const unordered_map<Key, T, Hash, Pred, Alloc>& x,
+               const unordered_map<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const unordered_map<Key, T, Hash, Pred, Alloc>& x,
+               const unordered_map<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash = hash<Key>, class Pred = equal_to<Key>,
+          class Alloc = allocator<pair<const Key, T>>>
+class unordered_multimap
+{
+public:
+    // types
+    typedef Key                                                        key_type;
+    typedef T                                                          mapped_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef pair<const key_type, mapped_type>                          value_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+    typedef /unspecified/ local_iterator;
+    typedef /unspecified/ const_local_iterator;
+
+    typedef unspecified node_type;    // C++17
+
+    unordered_multimap()
+        noexcept(
+            is_nothrow_default_constructible<hasher>::value &&
+            is_nothrow_default_constructible<key_equal>::value &&
+            is_nothrow_default_constructible<allocator_type>::value);
+    explicit unordered_multimap(size_type n, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        unordered_multimap(InputIterator f, InputIterator l,
+                      size_type n = 0, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    explicit unordered_multimap(const allocator_type&);
+    unordered_multimap(const unordered_multimap&);
+    unordered_multimap(const unordered_multimap&, const Allocator&);
+    unordered_multimap(unordered_multimap&&)
+        noexcept(
+            is_nothrow_move_constructible<hasher>::value &&
+            is_nothrow_move_constructible<key_equal>::value &&
+            is_nothrow_move_constructible<allocator_type>::value);
+    unordered_multimap(unordered_multimap&&, const Allocator&);
+    unordered_multimap(initializer_list<value_type>, size_type n = 0,
+                  const hasher& hf = hasher(), const key_equal& eql = key_equal(),
+                  const allocator_type& a = allocator_type());
+    unordered_multimap(size_type n, const allocator_type& a)
+      : unordered_multimap(n, hasher(), key_equal(), a) {}  // C++14
+    unordered_multimap(size_type n, const hasher& hf, const allocator_type& a)
+      : unordered_multimap(n, hf, key_equal(), a) {}  // C++14
+    template <class InputIterator>
+      unordered_multimap(InputIterator f, InputIterator l, size_type n, const allocator_type& a)
+      : unordered_multimap(f, l, n, hasher(), key_equal(), a) {}  // C++14
+    template <class InputIterator>
+      unordered_multimap(InputIterator f, InputIterator l, size_type n, const hasher& hf,
+        const allocator_type& a)
+      : unordered_multimap(f, l, n, hf, key_equal(), a) {}  // C++14
+    unordered_multimap(initializer_list<value_type> il, size_type n, const allocator_type& a)
+      : unordered_multimap(il, n, hasher(), key_equal(), a) {}  // C++14
+    unordered_multimap(initializer_list<value_type> il, size_type n, const hasher& hf,
+      const allocator_type& a)
+      : unordered_multimap(il, n, hf, key_equal(), a) {}  // C++14
+    ~unordered_multimap();
+    unordered_multimap& operator=(const unordered_multimap&);
+    unordered_multimap& operator=(unordered_multimap&&)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<hasher>::value &&
+            is_nothrow_move_assignable<key_equal>::value);
+    unordered_multimap& operator=(initializer_list<value_type>);
+
+    allocator_type get_allocator() const noexcept;
+
+    bool      empty() const noexcept;
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+
+    iterator       begin() noexcept;
+    iterator       end() noexcept;
+    const_iterator begin()  const noexcept;
+    const_iterator end()    const noexcept;
+    const_iterator cbegin() const noexcept;
+    const_iterator cend()   const noexcept;
+
+    template <class... Args>
+        iterator emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    iterator insert(const value_type& obj);
+    template <class P>
+        iterator insert(P&& obj);
+    iterator insert(const_iterator hint, const value_type& obj);
+    template <class P>
+        iterator insert(const_iterator hint, P&& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type>);
+
+    node_type extract(const_iterator position);                // C++17
+    node_type extract(const key_type& x);                      // C++17
+    iterator insert(node_type&& nh);                           // C++17
+    iterator insert(const_iterator hint, node_type&& nh);      // C++17
+
+    iterator erase(const_iterator position);
+    iterator erase(iterator position);  // C++14
+    size_type erase(const key_type& k);
+    iterator erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class H2, class P2>
+      void merge(unordered_multimap<Key, T, H2, P2, Allocator>& source);    // C++17
+    template<class H2, class P2>
+      void merge(unordered_multimap<Key, T, H2, P2, Allocator>&& source);   // C++17
+    template<class H2, class P2>
+      void merge(unordered_map<Key, T, H2, P2, Allocator>& source);         // C++17
+    template<class H2, class P2>
+      void merge(unordered_map<Key, T, H2, P2, Allocator>&& source);        // C++17
+
+    void swap(unordered_multimap&)
+        noexcept(
+            (!allocator_type::propagate_on_container_swap::value ||
+             __is_nothrow_swappable<allocator_type>::value) &&
+            __is_nothrow_swappable<hasher>::value &&
+            __is_nothrow_swappable<key_equal>::value);
+
+    hasher hash_function() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    bool contains(const key_type& k) const; // C++20
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    size_type bucket_count() const noexcept;
+    size_type max_bucket_count() const noexcept;
+
+    size_type bucket_size(size_type n) const;
+    size_type bucket(const key_type& k) const;
+
+    local_iterator       begin(size_type n);
+    local_iterator       end(size_type n);
+    const_local_iterator begin(size_type n) const;
+    const_local_iterator end(size_type n) const;
+    const_local_iterator cbegin(size_type n) const;
+    const_local_iterator cend(size_type n) const;
+
+    float load_factor() const noexcept;
+    float max_load_factor() const noexcept;
+    void max_load_factor(float z);
+    void rehash(size_type n);
+    void reserve(size_type n);
+};
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x,
+              unordered_multimap<Key, T, Hash, Pred, Alloc>& y)
+              noexcept(noexcept(x.swap(y)));
+
+template <class K, class T, class H, class P, class A, class Predicate>
+    void erase_if(unordered_set<K, T, H, P, A>& c, Predicate pred);       // C++20
+
+template <class K, class T, class H, class P, class A, class Predicate>
+    void erase_if(unordered_multiset<K, T, H, P, A>& c, Predicate pred);  // C++20
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const unordered_multimap<Key, T, Hash, Pred, Alloc>& x,
+               const unordered_multimap<Key, T, Hash, Pred, Alloc>& y);
+
+template <class Key, class T, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const unordered_multimap<Key, T, Hash, Pred, Alloc>& x,
+               const unordered_multimap<Key, T, Hash, Pred, Alloc>& y);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <__hash_table>
+#include <__node_handle>
+#include <functional>
+#include <stdexcept>
+#include <tuple>
+#include <version>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Key, class _Cp, class _Hash,
+          bool = is_empty<_Hash>::value && !__libcpp_is_final<_Hash>::value>
+class __unordered_map_hasher
+    : private _Hash
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_hasher()
+        _NOEXCEPT_(is_nothrow_default_constructible<_Hash>::value)
+        : _Hash() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_hasher(const _Hash& __h)
+        _NOEXCEPT_(is_nothrow_copy_constructible<_Hash>::value)
+        : _Hash(__h) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Hash& hash_function() const _NOEXCEPT {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const _Cp& __x) const
+        {return static_cast<const _Hash&>(*this)(__x.__get_value().first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const _Key& __x) const
+        {return static_cast<const _Hash&>(*this)(__x);}
+    void swap(__unordered_map_hasher&__y)
+        _NOEXCEPT_(__is_nothrow_swappable<_Hash>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(static_cast<_Hash&>(*this), static_cast<_Hash&>(__y));
+    }
+};
+
+template <class _Key, class _Cp, class _Hash>
+class __unordered_map_hasher<_Key, _Cp, _Hash, false>
+{
+    _Hash __hash_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_hasher()
+        _NOEXCEPT_(is_nothrow_default_constructible<_Hash>::value)
+        : __hash_() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_hasher(const _Hash& __h)
+        _NOEXCEPT_(is_nothrow_copy_constructible<_Hash>::value)
+        : __hash_(__h) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Hash& hash_function() const _NOEXCEPT {return __hash_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const _Cp& __x) const
+        {return __hash_(__x.__get_value().first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const _Key& __x) const
+        {return __hash_(__x);}
+    void swap(__unordered_map_hasher&__y)
+        _NOEXCEPT_(__is_nothrow_swappable<_Hash>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(__hash_, __y.__hash_);
+    }
+};
+
+template <class _Key, class _Cp, class _Hash, bool __b>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__unordered_map_hasher<_Key, _Cp, _Hash, __b>& __x,
+     __unordered_map_hasher<_Key, _Cp, _Hash, __b>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Key, class _Cp, class _Pred,
+          bool = is_empty<_Pred>::value && !__libcpp_is_final<_Pred>::value>
+class __unordered_map_equal
+    : private _Pred
+{
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_equal()
+        _NOEXCEPT_(is_nothrow_default_constructible<_Pred>::value)
+        : _Pred() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_equal(const _Pred& __p)
+        _NOEXCEPT_(is_nothrow_copy_constructible<_Pred>::value)
+        : _Pred(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Pred& key_eq() const _NOEXCEPT {return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Cp& __x, const _Cp& __y) const
+        {return static_cast<const _Pred&>(*this)(__x.__get_value().first, __y.__get_value().first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Cp& __x, const _Key& __y) const
+        {return static_cast<const _Pred&>(*this)(__x.__get_value().first, __y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Key& __x, const _Cp& __y) const
+        {return static_cast<const _Pred&>(*this)(__x, __y.__get_value().first);}
+    void swap(__unordered_map_equal&__y)
+        _NOEXCEPT_(__is_nothrow_swappable<_Pred>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(static_cast<_Pred&>(*this), static_cast<_Pred&>(__y));
+    }
+};
+
+template <class _Key, class _Cp, class _Pred>
+class __unordered_map_equal<_Key, _Cp, _Pred, false>
+{
+    _Pred __pred_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_equal()
+        _NOEXCEPT_(is_nothrow_default_constructible<_Pred>::value)
+        : __pred_() {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __unordered_map_equal(const _Pred& __p)
+        _NOEXCEPT_(is_nothrow_copy_constructible<_Pred>::value)
+        : __pred_(__p) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const _Pred& key_eq() const _NOEXCEPT {return __pred_;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Cp& __x, const _Cp& __y) const
+        {return __pred_(__x.__get_value().first, __y.__get_value().first);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Cp& __x, const _Key& __y) const
+        {return __pred_(__x.__get_value().first, __y);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    bool operator()(const _Key& __x, const _Cp& __y) const
+        {return __pred_(__x, __y.__get_value().first);}
+    void swap(__unordered_map_equal&__y)
+        _NOEXCEPT_(__is_nothrow_swappable<_Pred>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(__pred_, __y.__pred_);
+    }
+};
+
+template <class _Key, class _Cp, class _Pred, bool __b>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(__unordered_map_equal<_Key, _Cp, _Pred, __b>& __x,
+     __unordered_map_equal<_Key, _Cp, _Pred, __b>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+template <class _Alloc>
+class __hash_map_node_destructor
+{
+    typedef _Alloc                              allocator_type;
+    typedef allocator_traits<allocator_type>    __alloc_traits;
+
+public:
+
+    typedef typename __alloc_traits::pointer       pointer;
+private:
+
+    allocator_type& __na_;
+
+    __hash_map_node_destructor& operator=(const __hash_map_node_destructor&);
+
+public:
+    bool __first_constructed;
+    bool __second_constructed;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __hash_map_node_destructor(allocator_type& __na) _NOEXCEPT
+        : __na_(__na),
+          __first_constructed(false),
+          __second_constructed(false)
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_node_destructor(__hash_node_destructor<allocator_type>&& __x)
+        _NOEXCEPT
+        : __na_(__x.__na_),
+          __first_constructed(__x.__value_constructed),
+          __second_constructed(__x.__value_constructed)
+        {
+            __x.__value_constructed = false;
+        }
+#else  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_node_destructor(const __hash_node_destructor<allocator_type>& __x)
+        : __na_(__x.__na_),
+          __first_constructed(__x.__value_constructed),
+          __second_constructed(__x.__value_constructed)
+        {
+            const_cast<bool&>(__x.__value_constructed) = false;
+        }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator()(pointer __p) _NOEXCEPT
+    {
+        if (__second_constructed)
+            __alloc_traits::destroy(__na_, _CUDA_VSTD::addressof(__p->__value_.__get_value().second));
+        if (__first_constructed)
+            __alloc_traits::destroy(__na_, _CUDA_VSTD::addressof(__p->__value_.__get_value().first));
+        if (__p)
+            __alloc_traits::deallocate(__na_, __p, 1);
+    }
+};
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Key, class _Tp>
+struct __hash_value_type
+{
+    typedef _Key                                     key_type;
+    typedef _Tp                                      mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+    typedef pair<key_type&, mapped_type&>            __nc_ref_pair_type;
+    typedef pair<key_type&&, mapped_type&&>          __nc_rref_pair_type;
+
+private:
+    value_type __cc;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type& __get_value()
+    {
+#if _LIBCUDACXX_STD_VER > 14
+        return *_CUDA_VSTD::launder(_CUDA_VSTD::addressof(__cc));
+#else
+        return __cc;
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type& __get_value() const
+    {
+#if _LIBCUDACXX_STD_VER > 14
+        return *_CUDA_VSTD::launder(_CUDA_VSTD::addressof(__cc));
+#else
+        return __cc;
+#endif
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __nc_ref_pair_type __ref()
+    {
+        value_type& __v = __get_value();
+        return __nc_ref_pair_type(const_cast<key_type&>(__v.first), __v.second);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __nc_rref_pair_type __move()
+    {
+        value_type& __v = __get_value();
+        return __nc_rref_pair_type(
+            _CUDA_VSTD::move(const_cast<key_type&>(__v.first)),
+            _CUDA_VSTD::move(__v.second));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_value_type& operator=(const __hash_value_type& __v)
+    {
+        __ref() = __v.__get_value();
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_value_type& operator=(__hash_value_type&& __v)
+    {
+        __ref() = __v.__move();
+        return *this;
+    }
+
+    template <class _ValueTp,
+              class = typename enable_if<
+                    __is_same_uncvref<_ValueTp, value_type>::value
+                 >::type
+             >
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_value_type& operator=(_ValueTp&& __v)
+    {
+        __ref() = _CUDA_VSTD::forward<_ValueTp>(__v);
+        return *this;
+    }
+
+private:
+    __hash_value_type(const __hash_value_type& __v) = delete;
+    __hash_value_type(__hash_value_type&& __v) = delete;
+    template <class ..._Args>
+    explicit __hash_value_type(_Args&& ...__args) = delete;
+
+    ~__hash_value_type() = delete;
+};
+
+#else
+
+template <class _Key, class _Tp>
+struct __hash_value_type
+{
+    typedef _Key                                     key_type;
+    typedef _Tp                                      mapped_type;
+    typedef pair<const key_type, mapped_type>        value_type;
+
+private:
+    value_type __cc;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type& __get_value() { return __cc; }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type& __get_value() const { return __cc; }
+
+private:
+   ~__hash_value_type();
+};
+
+#endif
+
+template <class _HashIterator>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_map_iterator
+{
+    _HashIterator __i_;
+
+    typedef  __hash_node_types_from_iterator<_HashIterator> _NodeTypes;
+
+public:
+    typedef forward_iterator_tag                                 iterator_category;
+    typedef typename _NodeTypes::__map_value_type                value_type;
+    typedef typename _NodeTypes::difference_type                 difference_type;
+    typedef value_type&                                          reference;
+    typedef typename _NodeTypes::__map_value_type_pointer       pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_iterator() _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_iterator(_HashIterator __i) _NOEXCEPT : __i_(__i) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return __i_->__get_value();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return pointer_traits<pointer>::pointer_to(__i_->__get_value());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_iterator& operator++() {++__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_iterator operator++(int)
+    {
+        __hash_map_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const __hash_map_iterator& __x, const __hash_map_iterator& __y)
+        {return __x.__i_ == __y.__i_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const __hash_map_iterator& __x, const __hash_map_iterator& __y)
+        {return __x.__i_ != __y.__i_;}
+
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator;
+};
+
+template <class _HashIterator>
+class _LIBCUDACXX_TEMPLATE_VIS __hash_map_const_iterator
+{
+    _HashIterator __i_;
+
+    typedef  __hash_node_types_from_iterator<_HashIterator> _NodeTypes;
+
+public:
+    typedef forward_iterator_tag                                 iterator_category;
+    typedef typename _NodeTypes::__map_value_type                value_type;
+    typedef typename _NodeTypes::difference_type                 difference_type;
+    typedef const value_type&                                    reference;
+    typedef typename _NodeTypes::__const_map_value_type_pointer  pointer;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator() _NOEXCEPT {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator(_HashIterator __i) _NOEXCEPT : __i_(__i) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator(
+            __hash_map_iterator<typename _HashIterator::__non_const_iterator> __i)
+                 _NOEXCEPT
+                : __i_(__i.__i_) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference operator*() const {return __i_->__get_value();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer operator->() const {return pointer_traits<pointer>::pointer_to(__i_->__get_value());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator& operator++() {++__i_; return *this;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __hash_map_const_iterator operator++(int)
+    {
+        __hash_map_const_iterator __t(*this);
+        ++(*this);
+        return __t;
+    }
+
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator==(const __hash_map_const_iterator& __x, const __hash_map_const_iterator& __y)
+        {return __x.__i_ == __y.__i_;}
+    friend _LIBCUDACXX_INLINE_VISIBILITY
+        bool operator!=(const __hash_map_const_iterator& __x, const __hash_map_const_iterator& __y)
+        {return __x.__i_ != __y.__i_;}
+
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class, class, class, class, class> friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_iterator;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS __hash_const_local_iterator;
+};
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+class unordered_multimap;
+
+template <class _Key, class _Tp, class _Hash = hash<_Key>, class _Pred = equal_to<_Key>,
+          class _Alloc = allocator<pair<const _Key, _Tp> > >
+class _LIBCUDACXX_TEMPLATE_VIS unordered_map
+{
+public:
+    // types
+    typedef _Key                                           key_type;
+    typedef _Tp                                            mapped_type;
+    typedef typename __identity<_Hash>::type               hasher;
+    typedef typename __identity<_Pred>::type               key_equal;
+    typedef typename __identity<_Alloc>::type              allocator_type;
+    typedef pair<const key_type, mapped_type>              value_type;
+    typedef value_type&                                    reference;
+    typedef const value_type&                              const_reference;
+    static_assert((is_same<value_type, typename allocator_type::value_type>::value),
+                  "Invalid allocator::value_type");
+
+private:
+    typedef __hash_value_type<key_type, mapped_type>                 __value_type;
+    typedef __unordered_map_hasher<key_type, __value_type, hasher>   __hasher;
+    typedef __unordered_map_equal<key_type, __value_type, key_equal> __key_equal;
+    typedef typename __rebind_alloc_helper<allocator_traits<allocator_type>,
+                                                 __value_type>::type __allocator_type;
+
+    typedef __hash_table<__value_type, __hasher,
+                         __key_equal,  __allocator_type>   __table;
+
+    __table __table_;
+
+    typedef typename __table::_NodeTypes                   _NodeTypes;
+    typedef typename __table::__node_pointer               __node_pointer;
+    typedef typename __table::__node_const_pointer         __node_const_pointer;
+    typedef typename __table::__node_traits                __node_traits;
+    typedef typename __table::__node_allocator             __node_allocator;
+    typedef typename __table::__node                       __node;
+    typedef __hash_map_node_destructor<__node_allocator>   _Dp;
+    typedef unique_ptr<__node, _Dp>                         __node_holder;
+    typedef allocator_traits<allocator_type>               __alloc_traits;
+
+    static_assert((is_same<typename __table::__container_value_type, value_type>::value), "");
+    static_assert((is_same<typename __table::__node_value_type, __value_type>::value), "");
+public:
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+    typedef typename __table::size_type              size_type;
+    typedef typename __table::difference_type        difference_type;
+
+    typedef __hash_map_iterator<typename __table::iterator>       iterator;
+    typedef __hash_map_const_iterator<typename __table::const_iterator> const_iterator;
+    typedef __hash_map_iterator<typename __table::local_iterator> local_iterator;
+    typedef __hash_map_const_iterator<typename __table::const_local_iterator> const_local_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __map_node_handle<__node, allocator_type> node_type;
+    typedef __insert_return_type<iterator, node_type> insert_return_type;
+#endif
+
+    template <class _Key2, class _Tp2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class _Key2, class _Tp2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map()
+        _NOEXCEPT_(is_nothrow_default_constructible<__table>::value)
+        {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            __get_db()->__insert_c(this);
+#endif
+        }
+    explicit unordered_map(size_type __n, const hasher& __hf = hasher(),
+                           const key_equal& __eql = key_equal());
+    unordered_map(size_type __n, const hasher& __hf,
+                  const key_equal& __eql,
+                  const allocator_type& __a);
+    template <class _InputIterator>
+        unordered_map(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        unordered_map(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        unordered_map(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf,
+                      const key_equal& __eql,
+                      const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit unordered_map(const allocator_type& __a);
+    unordered_map(const unordered_map& __u);
+    unordered_map(const unordered_map& __u, const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map(unordered_map&& __u)
+        _NOEXCEPT_(is_nothrow_move_constructible<__table>::value);
+    unordered_map(unordered_map&& __u, const allocator_type& __a);
+    unordered_map(initializer_list<value_type> __il);
+    unordered_map(initializer_list<value_type> __il, size_type __n,
+                  const hasher& __hf = hasher(), const key_equal& __eql = key_equal());
+    unordered_map(initializer_list<value_type> __il, size_type __n,
+                  const hasher& __hf, const key_equal& __eql,
+                  const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map(size_type __n, const allocator_type& __a)
+      : unordered_map(__n, hasher(), key_equal(), __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map(size_type __n, const hasher& __hf, const allocator_type& __a)
+      : unordered_map(__n, __hf, key_equal(), __a) {}
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+      unordered_map(_InputIterator __first, _InputIterator __last, size_type __n, const allocator_type& __a)
+      : unordered_map(__first, __last, __n, hasher(), key_equal(), __a) {}
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+      unordered_map(_InputIterator __first, _InputIterator __last, size_type __n, const hasher& __hf,
+        const allocator_type& __a)
+      : unordered_map(__first, __last, __n, __hf, key_equal(), __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map(initializer_list<value_type> __il, size_type __n, const allocator_type& __a)
+      : unordered_map(__il, __n, hasher(), key_equal(), __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map(initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+      const allocator_type& __a)
+      : unordered_map(__il, __n, __hf, key_equal(), __a) {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~unordered_map() {
+        static_assert(sizeof(__diagnose_unordered_container_requirements<_Key, _Hash, _Pred>(0)), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map& operator=(const unordered_map& __u)
+    {
+#ifndef _LIBCUDACXX_CXX03_LANG
+        __table_ = __u.__table_;
+#else
+        if (this != &__u) {
+            __table_.clear();
+            __table_.hash_function() = __u.__table_.hash_function();
+            __table_.key_eq() = __u.__table_.key_eq();
+            __table_.max_load_factor() = __u.__table_.max_load_factor();
+            __table_.__copy_assign_alloc(__u.__table_);
+            insert(__u.begin(), __u.end());
+        }
+#endif
+        return *this;
+    }
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map& operator=(unordered_map&& __u)
+        _NOEXCEPT_(is_nothrow_move_assignable<__table>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_map& operator=(initializer_list<value_type> __il);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const _NOEXCEPT {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const _NOEXCEPT {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend()   const _NOEXCEPT {return __table_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> insert(const value_type& __x)
+        {return __table_.__insert_unique(__x);}
+
+    iterator insert(const_iterator __p, const value_type& __x) {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+            "unordered_map::insert(const_iterator, const value_type&) called with an iterator not"
+            " referring to this unordered_map");
+#else
+        ((void)__p);
+#endif
+        return insert(__x).first;
+    }
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> insert(value_type&& __x)
+        {return __table_.__insert_unique(_CUDA_VSTD::move(__x));}
+
+    iterator insert(const_iterator __p, value_type&& __x) {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+            "unordered_map::insert(const_iterator, const value_type&) called with an iterator not"
+            " referring to this unordered_map");
+#else
+        ((void)__p);
+#endif
+        return __table_.__insert_unique(_CUDA_VSTD::move(__x)).first;
+    }
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> insert(_Pp&& __x)
+            {return __table_.__insert_unique(_CUDA_VSTD::forward<_Pp>(__x));}
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert(const_iterator __p, _Pp&& __x)
+        {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+                "unordered_map::insert(const_iterator, value_type&&) called with an iterator not"
+                " referring to this unordered_map");
+#else
+          ((void)__p);
+#endif
+            return insert(_CUDA_VSTD::forward<_Pp>(__x)).first;
+        }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> emplace(_Args&&... __args) {
+        return __table_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator emplace_hint(const_iterator __p, _Args&&... __args) {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+            "unordered_map::emplace_hint(const_iterator, args...) called with an iterator not"
+            " referring to this unordered_map");
+#else
+          ((void)__p);
+#endif
+        return __table_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...).first;
+    }
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 14
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> try_emplace(const key_type& __k, _Args&&... __args)
+    {
+        return __table_.__emplace_unique_key_args(__k, _CUDA_VSTD::piecewise_construct,
+            _CUDA_VSTD::forward_as_tuple(__k),
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...));
+    }
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> try_emplace(key_type&& __k, _Args&&... __args)
+    {
+        return __table_.__emplace_unique_key_args(__k, _CUDA_VSTD::piecewise_construct,
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::move(__k)),
+            _CUDA_VSTD::forward_as_tuple(_CUDA_VSTD::forward<_Args>(__args)...));
+    }
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator try_emplace(const_iterator __h, const key_type& __k, _Args&&... __args)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__h) == this,
+            "unordered_map::try_emplace(const_iterator, key, args...) called with an iterator not"
+            " referring to this unordered_map");
+#else
+        ((void)__h);
+#endif
+        return try_emplace(__k, _CUDA_VSTD::forward<_Args>(__args)...).first;
+    }
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator try_emplace(const_iterator __h, key_type&& __k, _Args&&... __args)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__h) == this,
+            "unordered_map::try_emplace(const_iterator, key, args...) called with an iterator not"
+            " referring to this unordered_map");
+#else
+        ((void)__h);
+#endif
+        return try_emplace(_CUDA_VSTD::move(__k), _CUDA_VSTD::forward<_Args>(__args)...).first;
+    }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> insert_or_assign(const key_type& __k, _Vp&& __v)
+    {
+        pair<iterator, bool> __res = __table_.__emplace_unique_key_args(__k,
+            __k, _CUDA_VSTD::forward<_Vp>(__v));
+        if (!__res.second) {
+            __res.first->second = _CUDA_VSTD::forward<_Vp>(__v);
+        }
+        return __res;
+    }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> insert_or_assign(key_type&& __k, _Vp&& __v)
+    {
+        pair<iterator, bool> __res = __table_.__emplace_unique_key_args(__k,
+            _CUDA_VSTD::move(__k), _CUDA_VSTD::forward<_Vp>(__v));
+        if (!__res.second) {
+            __res.first->second = _CUDA_VSTD::forward<_Vp>(__v);
+        }
+        return __res;
+    }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert_or_assign(const_iterator, const key_type& __k, _Vp&& __v)
+     {
+          // FIXME: Add debug mode checking for the iterator input
+          return insert_or_assign(__k, _CUDA_VSTD::forward<_Vp>(__v)).first;
+     }
+
+    template <class _Vp>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator insert_or_assign(const_iterator, key_type&& __k, _Vp&& __v)
+     {
+        // FIXME: Add debug mode checking for the iterator input
+        return insert_or_assign(_CUDA_VSTD::move(__k), _CUDA_VSTD::forward<_Vp>(__v)).first;
+     }
+#endif // _LIBCUDACXX_STD_VER > 14
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __p) {return __table_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(iterator __p)       {return __table_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __first, const_iterator __last)
+        {return __table_.erase(__first.__i_, __last.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+        void clear() _NOEXCEPT {__table_.clear();}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    insert_return_type insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_map::insert()");
+        return __table_.template __node_handle_insert_unique<
+            node_type, insert_return_type>(_CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_map::insert()");
+        return __table_.template __node_handle_insert_unique<node_type>(
+            __hint.__i_, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __table_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __table_.template __node_handle_extract<node_type>(
+            __it.__i_);
+    }
+
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_map<key_type, mapped_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_unique(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_map<key_type, mapped_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_unique(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multimap<key_type, mapped_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_unique(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multimap<key_type, mapped_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_unique(__source.__table_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(unordered_map& __u)
+        _NOEXCEPT_(__is_nothrow_swappable<__table>::value)
+        { __table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_function() const
+        {return __table_.hash_function().hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const
+        {return __table_.key_eq().key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_unique(__k);}
+    #if _LIBCUDACXX_STD_VER > 17
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool contains(const key_type& __k) const {return find(__k) != end();}
+    #endif // _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_unique(__k);}
+
+    mapped_type& operator[](const key_type& __k);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    mapped_type& operator[](key_type&& __k);
+#endif
+
+    mapped_type&       at(const key_type& __k);
+    const mapped_type& at(const key_type& __k) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const _NOEXCEPT {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const _NOEXCEPT {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_size(size_type __n) const
+        {return __table_.bucket_size(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket(const key_type& __k) const {return __table_.bucket(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       begin(size_type __n)        {return __table_.begin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       end(size_type __n)          {return __table_.end(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator begin(size_type __n) const  {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator end(size_type __n) const    {return __table_.cend(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cbegin(size_type __n) const {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cend(size_type __n) const   {return __table_.cend(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float load_factor() const _NOEXCEPT {return __table_.load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float max_load_factor() const _NOEXCEPT {return __table_.max_load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void max_load_factor(float __mlf) {__table_.max_load_factor(__mlf);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void rehash(size_type __n) {__table_.rehash(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reserve(size_type __n) {__table_.reserve(__n);}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const
+        {return __table_.__dereferenceable(&__i->__i_);}
+    bool __decrementable(const const_iterator* __i) const
+        {return __table_.__decrementable(&__i->__i_);}
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(&__i->__i_, __n);}
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(&__i->__i_, __n);}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+private:
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    __node_holder __construct_node_with_key(const key_type& __k);
+#endif
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Hash = hash<__iter_key_type<_InputIterator>>,
+         class _Pred = equal_to<__iter_key_type<_InputIterator>>,
+         class _Allocator = allocator<__iter_to_alloc_type<_InputIterator>>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type = 0,
+              _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_map<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>, _Hash, _Pred, _Allocator>;
+
+template<class _Key, class _Tp, class _Hash = hash<remove_const_t<_Key>>,
+         class _Pred = equal_to<remove_const_t<_Key>>,
+         class _Allocator = allocator<pair<const _Key, _Tp>>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(initializer_list<pair<_Key, _Tp>>, typename allocator_traits<_Allocator>::size_type = 0,
+              _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_map<remove_const_t<_Key>, _Tp, _Hash, _Pred, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_map<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+                   hash<__iter_key_type<_InputIterator>>, equal_to<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(_InputIterator, _InputIterator, _Allocator)
+  -> unordered_map<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+                   hash<__iter_key_type<_InputIterator>>, equal_to<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _InputIterator, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_map<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+                   _Hash, equal_to<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(initializer_list<pair<_Key, _Tp>>, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_map<remove_const_t<_Key>, _Tp,
+                   hash<remove_const_t<_Key>>,
+                   equal_to<remove_const_t<_Key>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(initializer_list<pair<_Key, _Tp>>, _Allocator)
+  -> unordered_map<remove_const_t<_Key>, _Tp,
+                   hash<remove_const_t<_Key>>,
+                   equal_to<remove_const_t<_Key>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_map(initializer_list<pair<_Key, _Tp>>, typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_map<remove_const_t<_Key>, _Tp, _Hash,
+                   equal_to<remove_const_t<_Key>>, _Allocator>;
+#endif
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        size_type __n, const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        size_type __n, const hasher& __hf, const key_equal& __eql,
+        const allocator_type& __a)
+    : __table_(__hf, __eql, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        const allocator_type& __a)
+    : __table_(typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        _InputIterator __first, _InputIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        const unordered_map& __u)
+    : __table_(__u.__table_)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        const unordered_map& __u, const allocator_type& __a)
+    : __table_(__u.__table_, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        unordered_map&& __u)
+    _NOEXCEPT_(is_nothrow_move_constructible<__table>::value)
+    : __table_(_CUDA_VSTD::move(__u.__table_))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        unordered_map&& __u, const allocator_type& __a)
+    : __table_(_CUDA_VSTD::move(__u.__table_), typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__a != __u.get_allocator())
+    {
+        iterator __i = __u.begin();
+        while (__u.size() != 0) {
+            __table_.__emplace_unique(
+                __u.__table_.remove((__i++).__i_)->__value_.__move());
+        }
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    else
+        __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        initializer_list<value_type> __il)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::operator=(unordered_map&& __u)
+    _NOEXCEPT_(is_nothrow_move_assignable<__table>::value)
+{
+    __table_ = _CUDA_VSTD::move(__u.__table_);
+    return *this;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::operator=(
+        initializer_list<value_type> __il)
+{
+    __table_.__assign_unique(__il.begin(), __il.end());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                       _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_unique(*__first);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+_Tp&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::operator[](const key_type& __k)
+{
+    return __table_.__emplace_unique_key_args(__k,
+        std::piecewise_construct, std::forward_as_tuple(__k),
+                                  std::forward_as_tuple()).first->__get_value().second;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+_Tp&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::operator[](key_type&& __k)
+{
+    return __table_.__emplace_unique_key_args(__k,
+        std::piecewise_construct, std::forward_as_tuple(std::move(__k)),
+                                  std::forward_as_tuple()).first->__get_value().second;
+}
+#else // _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+typename unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::__node_holder
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::__construct_node_with_key(const key_type& __k)
+{
+    __node_allocator& __na = __table_.__node_alloc();
+    __node_holder __h(__node_traits::allocate(__na, 1), _Dp(__na));
+    __node_traits::construct(__na, _CUDA_VSTD::addressof(__h->__value_.__get_value().first), __k);
+    __h.get_deleter().__first_constructed = true;
+    __node_traits::construct(__na, _CUDA_VSTD::addressof(__h->__value_.__get_value().second));
+    __h.get_deleter().__second_constructed = true;
+    return _LIBCUDACXX_EXPLICIT_MOVE(__h);  // explicitly moved for C++03
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+_Tp&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::operator[](const key_type& __k)
+{
+    iterator __i = find(__k);
+    if (__i != end())
+        return __i->second;
+    __node_holder __h = __construct_node_with_key(__k);
+    pair<iterator, bool> __r = __table_.__node_insert_unique(__h.get());
+    __h.release();
+    return __r.first->second;
+}
+
+#endif  // _LIBCUDACXX_CXX03_MODE
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+_Tp&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::at(const key_type& __k)
+{
+    iterator __i = find(__k);
+    if (__i == end())
+        __throw_out_of_range("unordered_map::at: key not found");
+    return __i->second;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+const _Tp&
+unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::at(const key_type& __k) const
+{
+    const_iterator __i = find(__k);
+    if (__i == end())
+        __throw_out_of_range("unordered_map::at: key not found");
+    return __i->second;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+     unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(), __ey = __y.end();
+            __i != __ex; ++__i)
+    {
+        const_iterator __j = __y.find(__i->first);
+        if (__j == __ey || !(*__i == *__j))
+            return false;
+    }
+    return true;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Key, class _Tp, class _Hash = hash<_Key>, class _Pred = equal_to<_Key>,
+          class _Alloc = allocator<pair<const _Key, _Tp> > >
+class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap
+{
+public:
+    // types
+    typedef _Key                                           key_type;
+    typedef _Tp                                            mapped_type;
+    typedef typename __identity<_Hash>::type               hasher;
+    typedef typename __identity<_Pred>::type               key_equal;
+    typedef typename __identity<_Alloc>::type              allocator_type;
+    typedef pair<const key_type, mapped_type>              value_type;
+    typedef value_type&                                    reference;
+    typedef const value_type&                              const_reference;
+    static_assert((is_same<value_type, typename allocator_type::value_type>::value),
+                  "Invalid allocator::value_type");
+
+private:
+    typedef __hash_value_type<key_type, mapped_type>                 __value_type;
+    typedef __unordered_map_hasher<key_type, __value_type, hasher>   __hasher;
+    typedef __unordered_map_equal<key_type, __value_type, key_equal> __key_equal;
+    typedef typename __rebind_alloc_helper<allocator_traits<allocator_type>,
+                                                 __value_type>::type __allocator_type;
+
+    typedef __hash_table<__value_type, __hasher,
+                         __key_equal,  __allocator_type>   __table;
+
+    __table __table_;
+
+    typedef typename __table::_NodeTypes                   _NodeTypes;
+    typedef typename __table::__node_traits                __node_traits;
+    typedef typename __table::__node_allocator             __node_allocator;
+    typedef typename __table::__node                       __node;
+    typedef __hash_map_node_destructor<__node_allocator>   _Dp;
+    typedef unique_ptr<__node, _Dp>                         __node_holder;
+    typedef allocator_traits<allocator_type>               __alloc_traits;
+    static_assert((is_same<typename __node_traits::size_type,
+                          typename __alloc_traits::size_type>::value),
+                 "Allocator uses different size_type for different types");
+public:
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+    typedef typename __table::size_type              size_type;
+    typedef typename __table::difference_type        difference_type;
+
+    typedef __hash_map_iterator<typename __table::iterator>       iterator;
+    typedef __hash_map_const_iterator<typename __table::const_iterator> const_iterator;
+    typedef __hash_map_iterator<typename __table::local_iterator> local_iterator;
+    typedef __hash_map_const_iterator<typename __table::const_local_iterator> const_local_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __map_node_handle<__node, allocator_type> node_type;
+#endif
+
+    template <class _Key2, class _Tp2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_map;
+    template <class _Key2, class _Tp2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multimap;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap()
+        _NOEXCEPT_(is_nothrow_default_constructible<__table>::value)
+        {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            __get_db()->__insert_c(this);
+#endif
+        }
+    explicit unordered_multimap(size_type __n, const hasher& __hf = hasher(),
+                                const key_equal& __eql = key_equal());
+    unordered_multimap(size_type __n, const hasher& __hf,
+                                const key_equal& __eql,
+                                const allocator_type& __a);
+    template <class _InputIterator>
+        unordered_multimap(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        unordered_multimap(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        unordered_multimap(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf,
+                      const key_equal& __eql,
+                      const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit unordered_multimap(const allocator_type& __a);
+    unordered_multimap(const unordered_multimap& __u);
+    unordered_multimap(const unordered_multimap& __u, const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap(unordered_multimap&& __u)
+        _NOEXCEPT_(is_nothrow_move_constructible<__table>::value);
+    unordered_multimap(unordered_multimap&& __u, const allocator_type& __a);
+    unordered_multimap(initializer_list<value_type> __il);
+    unordered_multimap(initializer_list<value_type> __il, size_type __n,
+                       const hasher& __hf = hasher(),
+                       const key_equal& __eql = key_equal());
+    unordered_multimap(initializer_list<value_type> __il, size_type __n,
+                       const hasher& __hf, const key_equal& __eql,
+                       const allocator_type& __a);
+#endif  // _LIBCUDACXX_CXX03_LANG
+#if _LIBCUDACXX_STD_VER > 11
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap(size_type __n, const allocator_type& __a)
+      : unordered_multimap(__n, hasher(), key_equal(), __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap(size_type __n, const hasher& __hf, const allocator_type& __a)
+      : unordered_multimap(__n, __hf, key_equal(), __a) {}
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+      unordered_multimap(_InputIterator __first, _InputIterator __last, size_type __n, const allocator_type& __a)
+      : unordered_multimap(__first, __last, __n, hasher(), key_equal(), __a) {}
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+      unordered_multimap(_InputIterator __first, _InputIterator __last, size_type __n, const hasher& __hf,
+        const allocator_type& __a)
+      : unordered_multimap(__first, __last, __n, __hf, key_equal(), __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap(initializer_list<value_type> __il, size_type __n, const allocator_type& __a)
+      : unordered_multimap(__il, __n, hasher(), key_equal(), __a) {}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap(initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+      const allocator_type& __a)
+      : unordered_multimap(__il, __n, __hf, key_equal(), __a) {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~unordered_multimap() {
+        static_assert(sizeof(__diagnose_unordered_container_requirements<_Key, _Hash, _Pred>(0)), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap& operator=(const unordered_multimap& __u)
+    {
+#ifndef _LIBCUDACXX_CXX03_LANG
+        __table_ = __u.__table_;
+#else
+        if (this != &__u) {
+            __table_.clear();
+            __table_.hash_function() = __u.__table_.hash_function();
+            __table_.key_eq() = __u.__table_.key_eq();
+            __table_.max_load_factor() = __u.__table_.max_load_factor();
+            __table_.__copy_assign_alloc(__u.__table_);
+            insert(__u.begin(), __u.end());
+        }
+#endif
+        return *this;
+    }
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap& operator=(unordered_multimap&& __u)
+        _NOEXCEPT_(is_nothrow_move_assignable<__table>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multimap& operator=(initializer_list<value_type> __il);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const _NOEXCEPT {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const _NOEXCEPT {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend()   const _NOEXCEPT {return __table_.end();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const value_type& __x) {return __table_.__insert_multi(__x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, const value_type& __x)
+        {return __table_.__insert_multi(__p.__i_, __x);}
+
+    template <class _InputIterator>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(_InputIterator __first, _InputIterator __last);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(value_type&& __x) {return __table_.__insert_multi(_CUDA_VSTD::move(__x));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, value_type&& __x)
+        {return __table_.__insert_multi(__p.__i_, _CUDA_VSTD::move(__x));}
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(_Pp&& __x)
+        {return __table_.__insert_multi(_CUDA_VSTD::forward<_Pp>(__x));}
+
+    template <class _Pp,
+              class = typename enable_if<is_constructible<value_type, _Pp>::value>::type>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, _Pp&& __x)
+        {return __table_.__insert_multi(__p.__i_, _CUDA_VSTD::forward<_Pp>(__x));}
+
+    template <class... _Args>
+    iterator emplace(_Args&&... __args) {
+        return __table_.__emplace_multi(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+
+    template <class... _Args>
+    iterator emplace_hint(const_iterator __p, _Args&&... __args) {
+        return __table_.__emplace_hint_multi(__p.__i_, _CUDA_VSTD::forward<_Args>(__args)...);
+    }
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __p) {return __table_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(iterator __p)       {return __table_.erase(__p.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __first, const_iterator __last)
+        {return __table_.erase(__first.__i_, __last.__i_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__table_.clear();}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_multimap::insert()");
+        return __table_.template __node_handle_insert_multi<node_type>(
+            _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_multimap::insert()");
+        return __table_.template __node_handle_insert_multi<node_type>(
+            __hint.__i_, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __table_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __table_.template __node_handle_extract<node_type>(
+            __it.__i_);
+    }
+
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multimap<key_type, mapped_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multimap<key_type, mapped_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_map<key_type, mapped_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_map<key_type, mapped_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(unordered_multimap& __u)
+        _NOEXCEPT_(__is_nothrow_swappable<__table>::value)
+        {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_function() const
+        {return __table_.hash_function().hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const
+        {return __table_.key_eq().key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_multi(__k);}
+    #if _LIBCUDACXX_STD_VER > 17
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool contains(const key_type& __k) const {return find(__k) != end();}
+    #endif // _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_multi(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const _NOEXCEPT {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const _NOEXCEPT
+        {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_size(size_type __n) const
+        {return __table_.bucket_size(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket(const key_type& __k) const {return __table_.bucket(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       begin(size_type __n)        {return __table_.begin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       end(size_type __n)          {return __table_.end(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator begin(size_type __n) const  {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator end(size_type __n) const    {return __table_.cend(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cbegin(size_type __n) const {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cend(size_type __n) const   {return __table_.cend(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float load_factor() const _NOEXCEPT {return __table_.load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float max_load_factor() const _NOEXCEPT {return __table_.max_load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void max_load_factor(float __mlf) {__table_.max_load_factor(__mlf);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void rehash(size_type __n) {__table_.rehash(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reserve(size_type __n) {__table_.reserve(__n);}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const
+        {return __table_.__dereferenceable(&__i->__i_);}
+    bool __decrementable(const const_iterator* __i) const
+        {return __table_.__decrementable(&__i->__i_);}
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(&__i->__i_, __n);}
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(&__i->__i_, __n);}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Hash = hash<__iter_key_type<_InputIterator>>,
+         class _Pred = equal_to<__iter_key_type<_InputIterator>>,
+         class _Allocator = allocator<__iter_to_alloc_type<_InputIterator>>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type = 0,
+                   _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_multimap<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>, _Hash, _Pred, _Allocator>;
+
+template<class _Key, class _Tp, class _Hash = hash<remove_const_t<_Key>>,
+         class _Pred = equal_to<remove_const_t<_Key>>,
+         class _Allocator = allocator<pair<const _Key, _Tp>>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(initializer_list<pair<_Key, _Tp>>, typename allocator_traits<_Allocator>::size_type = 0,
+                   _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_multimap<remove_const_t<_Key>, _Tp, _Hash, _Pred, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_multimap<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+                        hash<__iter_key_type<_InputIterator>>, equal_to<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(_InputIterator, _InputIterator, _Allocator)
+  -> unordered_multimap<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+                        hash<__iter_key_type<_InputIterator>>, equal_to<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _InputIterator, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_multimap<__iter_key_type<_InputIterator>, __iter_mapped_type<_InputIterator>,
+                        _Hash, equal_to<__iter_key_type<_InputIterator>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(initializer_list<pair<_Key, _Tp>>, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_multimap<remove_const_t<_Key>, _Tp,
+                        hash<remove_const_t<_Key>>,
+                        equal_to<remove_const_t<_Key>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(initializer_list<pair<_Key, _Tp>>, _Allocator)
+  -> unordered_multimap<remove_const_t<_Key>, _Tp,
+                        hash<remove_const_t<_Key>>,
+                        equal_to<remove_const_t<_Key>>, _Allocator>;
+
+template<class _Key, class _Tp, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multimap(initializer_list<pair<_Key, _Tp>>, typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_multimap<remove_const_t<_Key>, _Tp, _Hash,
+                        equal_to<remove_const_t<_Key>>, _Allocator>;
+#endif
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        size_type __n, const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        size_type __n, const hasher& __hf, const key_equal& __eql,
+        const allocator_type& __a)
+    : __table_(__hf, __eql, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        _InputIterator __first, _InputIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        const allocator_type& __a)
+    : __table_(typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        const unordered_multimap& __u)
+    : __table_(__u.__table_)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        const unordered_multimap& __u, const allocator_type& __a)
+    : __table_(__u.__table_, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        unordered_multimap&& __u)
+    _NOEXCEPT_(is_nothrow_move_constructible<__table>::value)
+    : __table_(_CUDA_VSTD::move(__u.__table_))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        unordered_multimap&& __u, const allocator_type& __a)
+    : __table_(_CUDA_VSTD::move(__u.__table_), typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__a != __u.get_allocator())
+    {
+        iterator __i = __u.begin();
+        while (__u.size() != 0)
+        {
+            __table_.__insert_multi(
+                __u.__table_.remove((__i++).__i_)->__value_.__move());
+        }
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    else
+        __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        initializer_list<value_type> __il)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_multimap(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, typename __table::allocator_type(__a))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>&
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::operator=(unordered_multimap&& __u)
+    _NOEXCEPT_(is_nothrow_move_assignable<__table>::value)
+{
+    __table_ = _CUDA_VSTD::move(__u.__table_);
+    return *this;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>&
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::operator=(
+        initializer_list<value_type> __il)
+{
+    __table_.__assign_multi(__il.begin(), __il.end());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                            _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_multi(*__first);
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+     unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    typedef pair<const_iterator, const_iterator> _EqRng;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(); __i != __ex;)
+    {
+        _EqRng __xeq = __x.equal_range(__i->first);
+        _EqRng __yeq = __y.equal_range(__i->first);
+        if (_CUDA_VSTD::distance(__xeq.first, __xeq.second) !=
+            _CUDA_VSTD::distance(__yeq.first, __yeq.second) ||
+                  !_CUDA_VSTD::is_permutation(__xeq.first, __xeq.second, __yeq.first))
+            return false;
+        __i = __xeq.second;
+    }
+    return true;
+}
+
+template <class _Key, class _Tp, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __x,
+           const unordered_multimap<_Key, _Tp, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_UNORDERED_MAP
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set
new file mode 100644
index 000000000000..c636958e096c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/unordered_set
@@ -0,0 +1,1680 @@
+// -*- C++ -*-
+//===-------------------------- unordered_set -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_UNORDERED_SET
+#define _LIBCUDACXX_UNORDERED_SET
+
+/*
+
+    unordered_set synopsis
+
+#include <initializer_list>
+
+namespace std
+{
+
+template <class Value, class Hash = hash<Value>, class Pred = equal_to<Value>,
+          class Alloc = allocator<Value>>
+class unordered_set
+{
+public:
+    // types
+    typedef Value                                                      key_type;
+    typedef key_type                                                   value_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+    typedef /unspecified/ local_iterator;
+    typedef /unspecified/ const_local_iterator;
+
+    typedef unspecified node_type unspecified;                            // C++17
+    typedef INSERT_RETURN_TYPE<iterator, node_type> insert_return_type;   // C++17
+
+    unordered_set()
+        noexcept(
+            is_nothrow_default_constructible<hasher>::value &&
+            is_nothrow_default_constructible<key_equal>::value &&
+            is_nothrow_default_constructible<allocator_type>::value);
+    explicit unordered_set(size_type n, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        unordered_set(InputIterator f, InputIterator l,
+                      size_type n = 0, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    explicit unordered_set(const allocator_type&);
+    unordered_set(const unordered_set&);
+    unordered_set(const unordered_set&, const Allocator&);
+    unordered_set(unordered_set&&)
+        noexcept(
+            is_nothrow_move_constructible<hasher>::value &&
+            is_nothrow_move_constructible<key_equal>::value &&
+            is_nothrow_move_constructible<allocator_type>::value);
+    unordered_set(unordered_set&&, const Allocator&);
+    unordered_set(initializer_list<value_type>, size_type n = 0,
+                  const hasher& hf = hasher(), const key_equal& eql = key_equal(),
+                  const allocator_type& a = allocator_type());
+    unordered_set(size_type n, const allocator_type& a); // C++14
+    unordered_set(size_type n, const hasher& hf, const allocator_type& a); // C++14
+    template <class InputIterator>
+      unordered_set(InputIterator f, InputIterator l, size_type n, const allocator_type& a); // C++14
+    template <class InputIterator>
+      unordered_set(InputIterator f, InputIterator l, size_type n,
+                    const hasher& hf,  const allocator_type& a); // C++14
+    unordered_set(initializer_list<value_type> il, size_type n, const allocator_type& a); // C++14
+    unordered_set(initializer_list<value_type> il, size_type n,
+                  const hasher& hf,  const allocator_type& a); // C++14
+    ~unordered_set();
+    unordered_set& operator=(const unordered_set&);
+    unordered_set& operator=(unordered_set&&)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<hasher>::value &&
+            is_nothrow_move_assignable<key_equal>::value);
+    unordered_set& operator=(initializer_list<value_type>);
+
+    allocator_type get_allocator() const noexcept;
+
+    bool      empty() const noexcept;
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+
+    iterator       begin() noexcept;
+    iterator       end() noexcept;
+    const_iterator begin()  const noexcept;
+    const_iterator end()    const noexcept;
+    const_iterator cbegin() const noexcept;
+    const_iterator cend()   const noexcept;
+
+    template <class... Args>
+        pair<iterator, bool> emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    pair<iterator, bool> insert(const value_type& obj);
+    pair<iterator, bool> insert(value_type&& obj);
+    iterator insert(const_iterator hint, const value_type& obj);
+    iterator insert(const_iterator hint, value_type&& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type>);
+
+    node_type extract(const_iterator position);                       // C++17
+    node_type extract(const key_type& x);                             // C++17
+    insert_return_type insert(node_type&& nh);                        // C++17
+    iterator           insert(const_iterator hint, node_type&& nh);   // C++17
+
+    iterator erase(const_iterator position);
+    iterator erase(iterator position);  // C++14
+    size_type erase(const key_type& k);
+    iterator erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class H2, class P2>
+      void merge(unordered_set<Key, H2, P2, Allocator>& source);         // C++17
+    template<class H2, class P2>
+      void merge(unordered_set<Key, H2, P2, Allocator>&& source);        // C++17
+    template<class H2, class P2>
+      void merge(unordered_multiset<Key, H2, P2, Allocator>& source);    // C++17
+    template<class H2, class P2>
+      void merge(unordered_multiset<Key, H2, P2, Allocator>&& source);   // C++17
+
+    void swap(unordered_set&)
+       noexcept(allocator_traits<Allocator>::is_always_equal::value &&
+                 noexcept(swap(declval<hasher&>(), declval<hasher&>())) &&
+                 noexcept(swap(declval<key_equal&>(), declval<key_equal&>()))); // C++17
+
+    hasher hash_function() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    bool contains(const key_type& k) const; // C++20
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    size_type bucket_count() const noexcept;
+    size_type max_bucket_count() const noexcept;
+
+    size_type bucket_size(size_type n) const;
+    size_type bucket(const key_type& k) const;
+
+    local_iterator       begin(size_type n);
+    local_iterator       end(size_type n);
+    const_local_iterator begin(size_type n) const;
+    const_local_iterator end(size_type n) const;
+    const_local_iterator cbegin(size_type n) const;
+    const_local_iterator cend(size_type n) const;
+
+    float load_factor() const noexcept;
+    float max_load_factor() const noexcept;
+    void max_load_factor(float z);
+    void rehash(size_type n);
+    void reserve(size_type n);
+};
+
+template <class Value, class Hash, class Pred, class Alloc>
+    void swap(unordered_set<Value, Hash, Pred, Alloc>& x,
+              unordered_set<Value, Hash, Pred, Alloc>& y)
+              noexcept(noexcept(x.swap(y)));
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const unordered_set<Value, Hash, Pred, Alloc>& x,
+               const unordered_set<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const unordered_set<Value, Hash, Pred, Alloc>& x,
+               const unordered_set<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash = hash<Value>, class Pred = equal_to<Value>,
+          class Alloc = allocator<Value>>
+class unordered_multiset
+{
+public:
+    // types
+    typedef Value                                                      key_type;
+    typedef key_type                                                   value_type;
+    typedef Hash                                                       hasher;
+    typedef Pred                                                       key_equal;
+    typedef Alloc                                                      allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    typedef typename allocator_traits<allocator_type>::pointer         pointer;
+    typedef typename allocator_traits<allocator_type>::const_pointer   const_pointer;
+    typedef typename allocator_traits<allocator_type>::size_type       size_type;
+    typedef typename allocator_traits<allocator_type>::difference_type difference_type;
+
+    typedef /unspecified/ iterator;
+    typedef /unspecified/ const_iterator;
+    typedef /unspecified/ local_iterator;
+    typedef /unspecified/ const_local_iterator;
+
+    typedef unspecified node_type unspecified;   // C++17
+
+    unordered_multiset()
+        noexcept(
+            is_nothrow_default_constructible<hasher>::value &&
+            is_nothrow_default_constructible<key_equal>::value &&
+            is_nothrow_default_constructible<allocator_type>::value);
+    explicit unordered_multiset(size_type n, const hasher& hf = hasher(),
+                           const key_equal& eql = key_equal(),
+                           const allocator_type& a = allocator_type());
+    template <class InputIterator>
+        unordered_multiset(InputIterator f, InputIterator l,
+                      size_type n = 0, const hasher& hf = hasher(),
+                      const key_equal& eql = key_equal(),
+                      const allocator_type& a = allocator_type());
+    explicit unordered_multiset(const allocator_type&);
+    unordered_multiset(const unordered_multiset&);
+    unordered_multiset(const unordered_multiset&, const Allocator&);
+    unordered_multiset(unordered_multiset&&)
+        noexcept(
+            is_nothrow_move_constructible<hasher>::value &&
+            is_nothrow_move_constructible<key_equal>::value &&
+            is_nothrow_move_constructible<allocator_type>::value);
+    unordered_multiset(unordered_multiset&&, const Allocator&);
+    unordered_multiset(initializer_list<value_type>, size_type n = /see below/,
+                  const hasher& hf = hasher(), const key_equal& eql = key_equal(),
+                  const allocator_type& a = allocator_type());
+    unordered_multiset(size_type n, const allocator_type& a); // C++14
+    unordered_multiset(size_type n, const hasher& hf, const allocator_type& a); // C++14
+    template <class InputIterator>
+      unordered_multiset(InputIterator f, InputIterator l, size_type n, const allocator_type& a); // C++14
+    template <class InputIterator>
+      unordered_multiset(InputIterator f, InputIterator l, size_type n,
+                         const hasher& hf, const allocator_type& a); // C++14
+    unordered_multiset(initializer_list<value_type> il, size_type n, const allocator_type& a); // C++14
+    unordered_multiset(initializer_list<value_type> il, size_type n,
+                       const hasher& hf,  const allocator_type& a); // C++14
+    ~unordered_multiset();
+    unordered_multiset& operator=(const unordered_multiset&);
+    unordered_multiset& operator=(unordered_multiset&&)
+        noexcept(
+            allocator_type::propagate_on_container_move_assignment::value &&
+            is_nothrow_move_assignable<allocator_type>::value &&
+            is_nothrow_move_assignable<hasher>::value &&
+            is_nothrow_move_assignable<key_equal>::value);
+    unordered_multiset& operator=(initializer_list<value_type>);
+
+    allocator_type get_allocator() const noexcept;
+
+    bool      empty() const noexcept;
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+
+    iterator       begin() noexcept;
+    iterator       end() noexcept;
+    const_iterator begin()  const noexcept;
+    const_iterator end()    const noexcept;
+    const_iterator cbegin() const noexcept;
+    const_iterator cend()   const noexcept;
+
+    template <class... Args>
+        iterator emplace(Args&&... args);
+    template <class... Args>
+        iterator emplace_hint(const_iterator position, Args&&... args);
+    iterator insert(const value_type& obj);
+    iterator insert(value_type&& obj);
+    iterator insert(const_iterator hint, const value_type& obj);
+    iterator insert(const_iterator hint, value_type&& obj);
+    template <class InputIterator>
+        void insert(InputIterator first, InputIterator last);
+    void insert(initializer_list<value_type>);
+
+    node_type extract(const_iterator position);             // C++17
+    node_type extract(const key_type& x);                   // C++17
+    iterator insert(node_type&& nh);                        // C++17
+    iterator insert(const_iterator hint, node_type&& nh);   // C++17
+
+    iterator erase(const_iterator position);
+    iterator erase(iterator position);  // C++14
+    size_type erase(const key_type& k);
+    iterator erase(const_iterator first, const_iterator last);
+    void clear() noexcept;
+
+    template<class H2, class P2>
+      void merge(unordered_multiset<Key, H2, P2, Allocator>& source);    // C++17
+    template<class H2, class P2>
+      void merge(unordered_multiset<Key, H2, P2, Allocator>&& source);   // C++17
+    template<class H2, class P2>
+      void merge(unordered_set<Key, H2, P2, Allocator>& source);         // C++17
+    template<class H2, class P2>
+      void merge(unordered_set<Key, H2, P2, Allocator>&& source);        // C++17
+
+    void swap(unordered_multiset&)
+       noexcept(allocator_traits<Allocator>::is_always_equal::value &&
+                 noexcept(swap(declval<hasher&>(), declval<hasher&>())) &&
+                 noexcept(swap(declval<key_equal&>(), declval<key_equal&>()))); // C++17
+
+    hasher hash_function() const;
+    key_equal key_eq() const;
+
+    iterator       find(const key_type& k);
+    const_iterator find(const key_type& k) const;
+    size_type count(const key_type& k) const;
+    bool contains(const key_type& k) const; // C++20
+    pair<iterator, iterator>             equal_range(const key_type& k);
+    pair<const_iterator, const_iterator> equal_range(const key_type& k) const;
+
+    size_type bucket_count() const noexcept;
+    size_type max_bucket_count() const noexcept;
+
+    size_type bucket_size(size_type n) const;
+    size_type bucket(const key_type& k) const;
+
+    local_iterator       begin(size_type n);
+    local_iterator       end(size_type n);
+    const_local_iterator begin(size_type n) const;
+    const_local_iterator end(size_type n) const;
+    const_local_iterator cbegin(size_type n) const;
+    const_local_iterator cend(size_type n) const;
+
+    float load_factor() const noexcept;
+    float max_load_factor() const noexcept;
+    void max_load_factor(float z);
+    void rehash(size_type n);
+    void reserve(size_type n);
+};
+
+template <class Value, class Hash, class Pred, class Alloc>
+    void swap(unordered_multiset<Value, Hash, Pred, Alloc>& x,
+              unordered_multiset<Value, Hash, Pred, Alloc>& y)
+              noexcept(noexcept(x.swap(y)));
+
+template <class K, class T, class H, class P, class A, class Predicate>
+    void erase_if(unordered_set<K, T, H, P, A>& c, Predicate pred);       // C++20
+
+template <class K, class T, class H, class P, class A, class Predicate>
+    void erase_if(unordered_multiset<K, T, H, P, A>& c, Predicate pred);  // C++20
+
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator==(const unordered_multiset<Value, Hash, Pred, Alloc>& x,
+               const unordered_multiset<Value, Hash, Pred, Alloc>& y);
+
+template <class Value, class Hash, class Pred, class Alloc>
+    bool
+    operator!=(const unordered_multiset<Value, Hash, Pred, Alloc>& x,
+               const unordered_multiset<Value, Hash, Pred, Alloc>& y);
+}  // std
+
+*/
+
+#include <__config>
+#include <__hash_table>
+#include <__node_handle>
+#include <functional>
+#include <version>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+class unordered_multiset;
+
+template <class _Value, class _Hash = hash<_Value>, class _Pred = equal_to<_Value>,
+          class _Alloc = allocator<_Value> >
+class _LIBCUDACXX_TEMPLATE_VIS unordered_set
+{
+public:
+    // types
+    typedef _Value                                                     key_type;
+    typedef key_type                                                   value_type;
+    typedef typename __identity<_Hash>::type                           hasher;
+    typedef typename __identity<_Pred>::type                           key_equal;
+    typedef typename __identity<_Alloc>::type                          allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    static_assert((is_same<value_type, typename allocator_type::value_type>::value),
+                  "Invalid allocator::value_type");
+
+private:
+    typedef __hash_table<value_type, hasher, key_equal, allocator_type> __table;
+
+    __table __table_;
+
+public:
+    typedef typename __table::pointer         pointer;
+    typedef typename __table::const_pointer   const_pointer;
+    typedef typename __table::size_type       size_type;
+    typedef typename __table::difference_type difference_type;
+
+    typedef typename __table::const_iterator       iterator;
+    typedef typename __table::const_iterator       const_iterator;
+    typedef typename __table::const_local_iterator local_iterator;
+    typedef typename __table::const_local_iterator const_local_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __set_node_handle<typename __table::__node, allocator_type> node_type;
+    typedef __insert_return_type<iterator, node_type> insert_return_type;
+#endif
+
+    template <class _Value2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_set;
+    template <class _Value2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multiset;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set()
+        _NOEXCEPT_(is_nothrow_default_constructible<__table>::value)
+        {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            __get_db()->__insert_c(this);
+#endif
+        }
+    explicit unordered_set(size_type __n, const hasher& __hf = hasher(),
+                           const key_equal& __eql = key_equal());
+#if _LIBCUDACXX_STD_VER > 11
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set(size_type __n, const allocator_type& __a)
+        : unordered_set(__n, hasher(), key_equal(), __a) {}
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set(size_type __n, const hasher& __hf, const allocator_type& __a)
+        : unordered_set(__n, __hf, key_equal(), __a) {}
+#endif
+    unordered_set(size_type __n, const hasher& __hf, const key_equal& __eql,
+                  const allocator_type& __a);
+    template <class _InputIterator>
+        unordered_set(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        unordered_set(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        unordered_set(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf, const key_equal& __eql,
+                      const allocator_type& __a);
+#if _LIBCUDACXX_STD_VER > 11
+    template <class _InputIterator>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+        unordered_set(_InputIterator __first, _InputIterator __last,
+                    size_type __n, const allocator_type& __a)
+            : unordered_set(__first, __last, __n, hasher(), key_equal(), __a) {}
+    template <class _InputIterator>
+        unordered_set(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf, const allocator_type& __a)
+            : unordered_set(__first, __last, __n, __hf, key_equal(), __a) {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit unordered_set(const allocator_type& __a);
+    unordered_set(const unordered_set& __u);
+    unordered_set(const unordered_set& __u, const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set(unordered_set&& __u)
+        _NOEXCEPT_(is_nothrow_move_constructible<__table>::value);
+    unordered_set(unordered_set&& __u, const allocator_type& __a);
+    unordered_set(initializer_list<value_type> __il);
+    unordered_set(initializer_list<value_type> __il, size_type __n,
+                  const hasher& __hf = hasher(),
+                  const key_equal& __eql = key_equal());
+    unordered_set(initializer_list<value_type> __il, size_type __n,
+                  const hasher& __hf, const key_equal& __eql,
+                  const allocator_type& __a);
+#if _LIBCUDACXX_STD_VER > 11
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set(initializer_list<value_type> __il, size_type __n,
+                                                      const allocator_type& __a)
+        : unordered_set(__il, __n, hasher(), key_equal(), __a) {}
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set(initializer_list<value_type> __il, size_type __n,
+                                  const hasher& __hf, const allocator_type& __a)
+        : unordered_set(__il, __n, __hf, key_equal(), __a) {}
+#endif
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~unordered_set() {
+        static_assert(sizeof(__diagnose_unordered_container_requirements<_Value, _Hash, _Pred>(0)), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set& operator=(const unordered_set& __u)
+    {
+        __table_ = __u.__table_;
+        return *this;
+    }
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set& operator=(unordered_set&& __u)
+        _NOEXCEPT_(is_nothrow_move_assignable<__table>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_set& operator=(initializer_list<value_type> __il);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const _NOEXCEPT {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const _NOEXCEPT {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend()   const _NOEXCEPT {return __table_.end();}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair<iterator, bool> emplace(_Args&&... __args)
+            {return __table_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...);}
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        iterator emplace_hint(const_iterator __p, _Args&&... __args)
+        {
+            _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+                "unordered_set::emplace_hint(const_iterator, args...) called with an iterator not"
+                " referring to this unordered_set");
+            return __table_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...).first;
+        }
+#else
+        iterator emplace_hint(const_iterator, _Args&&... __args)
+            {return __table_.__emplace_unique(_CUDA_VSTD::forward<_Args>(__args)...).first;}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> insert(value_type&& __x)
+        {return __table_.__insert_unique(_CUDA_VSTD::move(__x));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    iterator insert(const_iterator __p, value_type&& __x)
+        {
+            _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+                "unordered_set::insert(const_iterator, value_type&&) called with an iterator not"
+                " referring to this unordered_set");
+            return insert(_CUDA_VSTD::move(__x)).first;
+        }
+#else
+    iterator insert(const_iterator, value_type&& __x)
+        {return insert(_CUDA_VSTD::move(__x)).first;}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, bool> insert(const value_type& __x)
+        {return __table_.__insert_unique(__x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    iterator insert(const_iterator __p, const value_type& __x)
+        {
+            _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__p) == this,
+                "unordered_set::insert(const_iterator, const value_type&) called with an iterator not"
+                " referring to this unordered_set");
+            return insert(__x).first;
+        }
+#else
+    iterator insert(const_iterator, const value_type& __x)
+        {return insert(__x).first;}
+#endif
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __p) {return __table_.erase(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __first, const_iterator __last)
+        {return __table_.erase(__first, __last);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__table_.clear();}
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    insert_return_type insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_set::insert()");
+        return __table_.template __node_handle_insert_unique<
+            node_type, insert_return_type>(_CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __h, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_set::insert()");
+        return __table_.template __node_handle_insert_unique<node_type>(
+            __h, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __table_.template __node_handle_extract<node_type>(__key);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __it)
+    {
+        return __table_.template __node_handle_extract<node_type>(__it);
+    }
+
+    template<class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_set<key_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __table_.__node_handle_merge_unique(__source.__table_);
+    }
+    template<class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_set<key_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __table_.__node_handle_merge_unique(__source.__table_);
+    }
+    template<class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multiset<key_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __table_.__node_handle_merge_unique(__source.__table_);
+    }
+    template<class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multiset<key_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        __table_.__node_handle_merge_unique(__source.__table_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(unordered_set& __u)
+        _NOEXCEPT_(__is_nothrow_swappable<__table>::value)
+        {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_function() const {return __table_.hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const {return __table_.key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_unique(__k);}
+    #if _LIBCUDACXX_STD_VER > 17
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool contains(const key_type& __k) const {return find(__k) != end();}
+    #endif // _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_unique(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_unique(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const _NOEXCEPT {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const _NOEXCEPT {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_size(size_type __n) const {return __table_.bucket_size(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket(const key_type& __k) const {return __table_.bucket(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       begin(size_type __n)        {return __table_.begin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       end(size_type __n)          {return __table_.end(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator begin(size_type __n) const  {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator end(size_type __n) const    {return __table_.cend(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cbegin(size_type __n) const {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cend(size_type __n) const   {return __table_.cend(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float load_factor() const _NOEXCEPT {return __table_.load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float max_load_factor() const _NOEXCEPT {return __table_.max_load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void max_load_factor(float __mlf) {__table_.max_load_factor(__mlf);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void rehash(size_type __n) {__table_.rehash(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reserve(size_type __n) {__table_.reserve(__n);}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const
+        {return __table_.__dereferenceable(__i);}
+    bool __decrementable(const const_iterator* __i) const
+        {return __table_.__decrementable(__i);}
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(__i, __n);}
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(__i, __n);}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Hash = hash<__iter_value_type<_InputIterator>>,
+         class _Pred = equal_to<__iter_value_type<_InputIterator>>,
+         class _Allocator = allocator<__iter_value_type<_InputIterator>>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_set(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type = 0,
+              _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_set<__iter_value_type<_InputIterator>, _Hash, _Pred, _Allocator>;
+
+template<class _Tp, class _Hash = hash<_Tp>,
+         class _Pred = equal_to<_Tp>,
+         class _Allocator = allocator<_Tp>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_set(initializer_list<_Tp>, typename allocator_traits<_Allocator>::size_type = 0,
+              _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_set<_Tp, _Hash, _Pred, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_set(_InputIterator, _InputIterator,
+              typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_set<__iter_value_type<_InputIterator>,
+                   hash<__iter_value_type<_InputIterator>>,
+                   equal_to<__iter_value_type<_InputIterator>>,
+                   _Allocator>;
+
+template<class _InputIterator, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_set(_InputIterator, _InputIterator,
+              typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_set<__iter_value_type<_InputIterator>, _Hash,
+                   equal_to<__iter_value_type<_InputIterator>>,
+                   _Allocator>;
+
+template<class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_set(initializer_list<_Tp>, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_set<_Tp, hash<_Tp>, equal_to<_Tp>, _Allocator>;
+
+template<class _Tp, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_set(initializer_list<_Tp>, typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_set<_Tp, _Hash, equal_to<_Tp>, _Allocator>;
+#endif
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        _InputIterator __first, _InputIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        const allocator_type& __a)
+    : __table_(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        const unordered_set& __u)
+    : __table_(__u.__table_)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        const unordered_set& __u, const allocator_type& __a)
+    : __table_(__u.__table_, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        unordered_set&& __u)
+    _NOEXCEPT_(is_nothrow_move_constructible<__table>::value)
+    : __table_(_CUDA_VSTD::move(__u.__table_))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        unordered_set&& __u, const allocator_type& __a)
+    : __table_(_CUDA_VSTD::move(__u.__table_), __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__a != __u.get_allocator())
+    {
+        iterator __i = __u.begin();
+        while (__u.size() != 0)
+            __table_.__insert_unique(_CUDA_VSTD::move(__u.__table_.remove(__i++)->__value_));
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    else
+        __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        initializer_list<value_type> __il)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_set<_Value, _Hash, _Pred, _Alloc>::unordered_set(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_set<_Value, _Hash, _Pred, _Alloc>&
+unordered_set<_Value, _Hash, _Pred, _Alloc>::operator=(unordered_set&& __u)
+    _NOEXCEPT_(is_nothrow_move_assignable<__table>::value)
+{
+    __table_ = _CUDA_VSTD::move(__u.__table_);
+    return *this;
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_set<_Value, _Hash, _Pred, _Alloc>&
+unordered_set<_Value, _Hash, _Pred, _Alloc>::operator=(
+        initializer_list<value_type> __il)
+{
+    __table_.__assign_unique(__il.begin(), __il.end());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+unordered_set<_Value, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                    _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_unique(*__first);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(unordered_set<_Value, _Hash, _Pred, _Alloc>& __x,
+     unordered_set<_Value, _Hash, _Pred, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Value, class _Hash, class _Pred, class _Alloc, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(unordered_set<_Value, _Hash, _Pred, _Alloc>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const unordered_set<_Value, _Hash, _Pred, _Alloc>& __x,
+           const unordered_set<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename unordered_set<_Value, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(), __ey = __y.end();
+            __i != __ex; ++__i)
+    {
+        const_iterator __j = __y.find(*__i);
+        if (__j == __ey || !(*__i == *__j))
+            return false;
+    }
+    return true;
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const unordered_set<_Value, _Hash, _Pred, _Alloc>& __x,
+           const unordered_set<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Value, class _Hash = hash<_Value>, class _Pred = equal_to<_Value>,
+          class _Alloc = allocator<_Value> >
+class _LIBCUDACXX_TEMPLATE_VIS unordered_multiset
+{
+public:
+    // types
+    typedef _Value                                                     key_type;
+    typedef key_type                                                   value_type;
+    typedef typename __identity<_Hash>::type                           hasher;
+    typedef typename __identity<_Pred>::type                           key_equal;
+    typedef typename __identity<_Alloc>::type                          allocator_type;
+    typedef value_type&                                                reference;
+    typedef const value_type&                                          const_reference;
+    static_assert((is_same<value_type, typename allocator_type::value_type>::value),
+                  "Invalid allocator::value_type");
+
+private:
+    typedef __hash_table<value_type, hasher, key_equal, allocator_type> __table;
+
+    __table __table_;
+
+public:
+    typedef typename __table::pointer         pointer;
+    typedef typename __table::const_pointer   const_pointer;
+    typedef typename __table::size_type       size_type;
+    typedef typename __table::difference_type difference_type;
+
+    typedef typename __table::const_iterator       iterator;
+    typedef typename __table::const_iterator       const_iterator;
+    typedef typename __table::const_local_iterator local_iterator;
+    typedef typename __table::const_local_iterator const_local_iterator;
+
+#if _LIBCUDACXX_STD_VER > 14
+    typedef __set_node_handle<typename __table::__node, allocator_type> node_type;
+#endif
+
+    template <class _Value2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_set;
+    template <class _Value2, class _Hash2, class _Pred2, class _Alloc2>
+        friend class _LIBCUDACXX_TEMPLATE_VIS unordered_multiset;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset()
+        _NOEXCEPT_(is_nothrow_default_constructible<__table>::value)
+        {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            __get_db()->__insert_c(this);
+#endif
+        }
+    explicit unordered_multiset(size_type __n, const hasher& __hf = hasher(),
+                                const key_equal& __eql = key_equal());
+    unordered_multiset(size_type __n, const hasher& __hf,
+                       const key_equal& __eql, const allocator_type& __a);
+#if _LIBCUDACXX_STD_VER > 11
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(size_type __n, const allocator_type& __a)
+        : unordered_multiset(__n, hasher(), key_equal(), __a) {}
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(size_type __n, const hasher& __hf, const allocator_type& __a)
+        : unordered_multiset(__n, __hf, key_equal(), __a) {}
+#endif
+    template <class _InputIterator>
+        unordered_multiset(_InputIterator __first, _InputIterator __last);
+    template <class _InputIterator>
+        unordered_multiset(_InputIterator __first, _InputIterator __last,
+                      size_type __n, const hasher& __hf = hasher(),
+                      const key_equal& __eql = key_equal());
+    template <class _InputIterator>
+        unordered_multiset(_InputIterator __first, _InputIterator __last,
+                      size_type __n , const hasher& __hf,
+                      const key_equal& __eql, const allocator_type& __a);
+#if _LIBCUDACXX_STD_VER > 11
+    template <class _InputIterator>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(_InputIterator __first, _InputIterator __last,
+                       size_type __n, const allocator_type& __a)
+        : unordered_multiset(__first, __last, __n, hasher(), key_equal(), __a) {}
+    template <class _InputIterator>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(_InputIterator __first, _InputIterator __last,
+                       size_type __n, const hasher& __hf, const allocator_type& __a)
+        : unordered_multiset(__first, __last, __n, __hf, key_equal(), __a) {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit unordered_multiset(const allocator_type& __a);
+    unordered_multiset(const unordered_multiset& __u);
+    unordered_multiset(const unordered_multiset& __u, const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(unordered_multiset&& __u)
+        _NOEXCEPT_(is_nothrow_move_constructible<__table>::value);
+    unordered_multiset(unordered_multiset&& __u, const allocator_type& __a);
+    unordered_multiset(initializer_list<value_type> __il);
+    unordered_multiset(initializer_list<value_type> __il, size_type __n,
+                       const hasher& __hf = hasher(),
+                       const key_equal& __eql = key_equal());
+    unordered_multiset(initializer_list<value_type> __il, size_type __n,
+                       const hasher& __hf, const key_equal& __eql,
+                       const allocator_type& __a);
+#if _LIBCUDACXX_STD_VER > 11
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(initializer_list<value_type> __il, size_type __n, const allocator_type& __a)
+      : unordered_multiset(__il, __n, hasher(), key_equal(), __a) {}
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset(initializer_list<value_type> __il, size_type __n, const hasher& __hf, const allocator_type& __a)
+      : unordered_multiset(__il, __n, __hf, key_equal(), __a) {}
+#endif
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~unordered_multiset() {
+        static_assert(sizeof(__diagnose_unordered_container_requirements<_Value, _Hash, _Pred>(0)), "");
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset& operator=(const unordered_multiset& __u)
+    {
+        __table_ = __u.__table_;
+        return *this;
+    }
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    unordered_multiset& operator=(unordered_multiset&& __u)
+        _NOEXCEPT_(is_nothrow_move_assignable<__table>::value);
+    unordered_multiset& operator=(initializer_list<value_type> __il);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT
+        {return allocator_type(__table_.__node_alloc());}
+
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool      empty() const _NOEXCEPT {return __table_.size() == 0;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT  {return __table_.size();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_size() const _NOEXCEPT {return __table_.max_size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       begin() _NOEXCEPT        {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       end() _NOEXCEPT          {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin()  const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()    const _NOEXCEPT {return __table_.end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cbegin() const _NOEXCEPT {return __table_.begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator cend()   const _NOEXCEPT {return __table_.end();}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator emplace(_Args&&... __args)
+            {return __table_.__emplace_multi(_CUDA_VSTD::forward<_Args>(__args)...);}
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        iterator emplace_hint(const_iterator __p, _Args&&... __args)
+            {return __table_.__emplace_hint_multi(__p, _CUDA_VSTD::forward<_Args>(__args)...);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(value_type&& __x) {return __table_.__insert_multi(_CUDA_VSTD::move(__x));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, value_type&& __x)
+        {return __table_.__insert_multi(__p, _CUDA_VSTD::move(__x));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void insert(initializer_list<value_type> __il)
+        {insert(__il.begin(), __il.end());}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const value_type& __x) {return __table_.__insert_multi(__x);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __p, const value_type& __x)
+        {return __table_.__insert_multi(__p, __x);}
+
+    template <class _InputIterator>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        void insert(_InputIterator __first, _InputIterator __last);
+
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_multiset::insert()");
+        return __table_.template __node_handle_insert_multi<node_type>(
+            _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __hint, node_type&& __nh)
+    {
+        _LIBCUDACXX_ASSERT(__nh.empty() || __nh.get_allocator() == get_allocator(),
+            "node_type with incompatible allocator passed to unordered_multiset::insert()");
+        return __table_.template __node_handle_insert_multi<node_type>(
+            __hint, _CUDA_VSTD::move(__nh));
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(const_iterator __position)
+    {
+        return __table_.template __node_handle_extract<node_type>(
+            __position);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY
+    node_type extract(key_type const& __key)
+    {
+        return __table_.template __node_handle_extract<node_type>(__key);
+    }
+
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multiset<key_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_multiset<key_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_set<key_type, _H2, _P2, allocator_type>& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+    template <class _H2, class _P2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void merge(unordered_set<key_type, _H2, _P2, allocator_type>&& __source)
+    {
+        _LIBCUDACXX_ASSERT(__source.get_allocator() == get_allocator(),
+                       "merging container with incompatible allocator");
+        return __table_.__node_handle_merge_multi(__source.__table_);
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __p) {return __table_.erase(__p);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type erase(const key_type& __k) {return __table_.__erase_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator erase(const_iterator __first, const_iterator __last)
+        {return __table_.erase(__first, __last);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__table_.clear();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(unordered_multiset& __u)
+        _NOEXCEPT_(__is_nothrow_swappable<__table>::value)
+        {__table_.swap(__u.__table_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    hasher hash_function() const {return __table_.hash_function();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    key_equal key_eq() const {return __table_.key_eq();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       find(const key_type& __k)       {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator find(const key_type& __k) const {return __table_.find(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type count(const key_type& __k) const {return __table_.__count_multi(__k);}
+    #if _LIBCUDACXX_STD_VER > 17
+        _LIBCUDACXX_INLINE_VISIBILITY
+        bool contains(const key_type& __k) const {return find(__k) != end();}
+    #endif // _LIBCUDACXX_STD_VER > 17
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<iterator, iterator>             equal_range(const key_type& __k)
+        {return __table_.__equal_range_multi(__k);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair<const_iterator, const_iterator> equal_range(const key_type& __k) const
+        {return __table_.__equal_range_multi(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_count() const _NOEXCEPT {return __table_.bucket_count();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type max_bucket_count() const _NOEXCEPT {return __table_.max_bucket_count();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket_size(size_type __n) const {return __table_.bucket_size(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type bucket(const key_type& __k) const {return __table_.bucket(__k);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       begin(size_type __n)        {return __table_.begin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    local_iterator       end(size_type __n)          {return __table_.end(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator begin(size_type __n) const  {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator end(size_type __n) const    {return __table_.cend(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cbegin(size_type __n) const {return __table_.cbegin(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_local_iterator cend(size_type __n) const   {return __table_.cend(__n);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float load_factor() const _NOEXCEPT {return __table_.load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    float max_load_factor() const _NOEXCEPT {return __table_.max_load_factor();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void max_load_factor(float __mlf) {__table_.max_load_factor(__mlf);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void rehash(size_type __n) {__table_.rehash(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void reserve(size_type __n) {__table_.reserve(__n);}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const
+        {return __table_.__dereferenceable(__i);}
+    bool __decrementable(const const_iterator* __i) const
+        {return __table_.__decrementable(__i);}
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(__i, __n);}
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const
+        {return __table_.__addable(__i, __n);}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Hash = hash<__iter_value_type<_InputIterator>>,
+         class _Pred = equal_to<__iter_value_type<_InputIterator>>,
+         class _Allocator = allocator<__iter_value_type<_InputIterator>>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multiset(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type = 0,
+              _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_multiset<__iter_value_type<_InputIterator>, _Hash, _Pred, _Allocator>;
+
+template<class _Tp, class _Hash = hash<_Tp>,
+         class _Pred = equal_to<_Tp>, class _Allocator = allocator<_Tp>,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<!__is_allocator<_Pred>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multiset(initializer_list<_Tp>, typename allocator_traits<_Allocator>::size_type = 0,
+              _Hash = _Hash(), _Pred = _Pred(), _Allocator = _Allocator())
+  -> unordered_multiset<_Tp, _Hash, _Pred, _Allocator>;
+
+template<class _InputIterator, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multiset(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_multiset<__iter_value_type<_InputIterator>,
+                   hash<__iter_value_type<_InputIterator>>,
+                   equal_to<__iter_value_type<_InputIterator>>,
+                   _Allocator>;
+
+template<class _InputIterator, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multiset(_InputIterator, _InputIterator, typename allocator_traits<_Allocator>::size_type,
+              _Hash, _Allocator)
+  -> unordered_multiset<__iter_value_type<_InputIterator>, _Hash,
+                   equal_to<__iter_value_type<_InputIterator>>,
+                   _Allocator>;
+
+template<class _Tp, class _Allocator,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multiset(initializer_list<_Tp>, typename allocator_traits<_Allocator>::size_type, _Allocator)
+  -> unordered_multiset<_Tp, hash<_Tp>, equal_to<_Tp>, _Allocator>;
+
+template<class _Tp, class _Hash, class _Allocator,
+         class = _EnableIf<!__is_allocator<_Hash>::value>,
+         class = _EnableIf<!is_integral<_Hash>::value>,
+         class = _EnableIf<__is_allocator<_Allocator>::value>>
+unordered_multiset(initializer_list<_Tp>, typename allocator_traits<_Allocator>::size_type, _Hash, _Allocator)
+  -> unordered_multiset<_Tp, _Hash, equal_to<_Tp>, _Allocator>;
+#endif
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        size_type __n, const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        size_type __n, const hasher& __hf, const key_equal& __eql,
+        const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        _InputIterator __first, _InputIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        _InputIterator __first, _InputIterator __last, size_type __n,
+        const hasher& __hf, const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__first, __last);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        const allocator_type& __a)
+    : __table_(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        const unordered_multiset& __u)
+    : __table_(__u.__table_)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        const unordered_multiset& __u, const allocator_type& __a)
+    : __table_(__u.__table_, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__u.bucket_count());
+    insert(__u.begin(), __u.end());
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        unordered_multiset&& __u)
+    _NOEXCEPT_(is_nothrow_move_constructible<__table>::value)
+    : __table_(_CUDA_VSTD::move(__u.__table_))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        unordered_multiset&& __u, const allocator_type& __a)
+    : __table_(_CUDA_VSTD::move(__u.__table_), __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__a != __u.get_allocator())
+    {
+        iterator __i = __u.begin();
+        while (__u.size() != 0)
+            __table_.__insert_multi(_CUDA_VSTD::move(__u.__table_.remove(__i++)->__value_));
+    }
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    else
+        __get_db()->swap(this, &__u);
+#endif
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        initializer_list<value_type> __il)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql)
+    : __table_(__hf, __eql)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::unordered_multiset(
+        initializer_list<value_type> __il, size_type __n, const hasher& __hf,
+        const key_equal& __eql, const allocator_type& __a)
+    : __table_(__hf, __eql, __a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    __table_.rehash(__n);
+    insert(__il.begin(), __il.end());
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>&
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::operator=(
+        unordered_multiset&& __u)
+    _NOEXCEPT_(is_nothrow_move_assignable<__table>::value)
+{
+    __table_ = _CUDA_VSTD::move(__u.__table_);
+    return *this;
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>&
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::operator=(
+        initializer_list<value_type> __il)
+{
+    __table_.__assign_multi(__il.begin(), __il.end());
+    return *this;
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+template <class _InputIterator>
+inline
+void
+unordered_multiset<_Value, _Hash, _Pred, _Alloc>::insert(_InputIterator __first,
+                                                         _InputIterator __last)
+{
+    for (; __first != __last; ++__first)
+        __table_.__insert_multi(*__first);
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __x,
+     unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Value, class _Hash, class _Pred, class _Alloc, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __c, _Predicate __pred)
+{ __libcpp_erase_if_container(__c, __pred); }
+#endif
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+bool
+operator==(const unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __x,
+           const unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    if (__x.size() != __y.size())
+        return false;
+    typedef typename unordered_multiset<_Value, _Hash, _Pred, _Alloc>::const_iterator
+                                                                 const_iterator;
+    typedef pair<const_iterator, const_iterator> _EqRng;
+    for (const_iterator __i = __x.begin(), __ex = __x.end(); __i != __ex;)
+    {
+        _EqRng __xeq = __x.equal_range(*__i);
+        _EqRng __yeq = __y.equal_range(*__i);
+        if (_CUDA_VSTD::distance(__xeq.first, __xeq.second) !=
+            _CUDA_VSTD::distance(__yeq.first, __yeq.second) ||
+                  !_CUDA_VSTD::is_permutation(__xeq.first, __xeq.second, __yeq.first))
+            return false;
+        __i = __xeq.second;
+    }
+    return true;
+}
+
+template <class _Value, class _Hash, class _Pred, class _Alloc>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __x,
+           const unordered_multiset<_Value, _Hash, _Pred, _Alloc>& __y)
+{
+    return !(__x == __y);
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#endif  // _LIBCUDACXX_UNORDERED_SET
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/utility b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/utility
new file mode 100644
index 000000000000..e9c9606249f3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/utility
@@ -0,0 +1,1650 @@
+// -*- C++ -*-
+//===-------------------------- utility -----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_UTILITY
+#define _LIBCUDACXX_UTILITY
+
+/*
+    utility synopsis
+
+#include <initializer_list>
+
+namespace std
+{
+
+template <class T>
+    void
+    swap(T& a, T& b);
+
+namespace rel_ops
+{
+    template<class T> bool operator!=(const T&, const T&);
+    template<class T> bool operator> (const T&, const T&);
+    template<class T> bool operator<=(const T&, const T&);
+    template<class T> bool operator>=(const T&, const T&);
+}
+
+template<class T>
+void
+swap(T& a, T& b) noexcept(is_nothrow_move_constructible<T>::value &&
+                          is_nothrow_move_assignable<T>::value);
+
+template <class T, size_t N>
+void
+swap(T (&a)[N], T (&b)[N]) noexcept(noexcept(swap(*a, *b)));
+
+template <class T> T&& forward(typename remove_reference<T>::type& t) noexcept;  // constexpr in C++14
+template <class T> T&& forward(typename remove_reference<T>::type&& t) noexcept; // constexpr in C++14
+
+template <class T> typename remove_reference<T>::type&& move(T&&) noexcept;      // constexpr in C++14
+
+template <class T>
+    typename conditional
+    <
+        !is_nothrow_move_constructible<T>::value && is_copy_constructible<T>::value,
+        const T&,
+        T&&
+    >::type
+    move_if_noexcept(T& x) noexcept; // constexpr in C++14
+
+template <class T> constexpr add_const_t<T>& as_const(T& t) noexcept;      // C++17
+template <class T>                      void as_const(const T&&) = delete; // C++17
+
+template <class T> typename add_rvalue_reference<T>::type declval() noexcept;
+
+template <class T1, class T2>
+struct pair
+{
+    typedef T1 first_type;
+    typedef T2 second_type;
+
+    T1 first;
+    T2 second;
+
+    pair(const pair&) = default;
+    pair(pair&&) = default;
+    explicit(see-below) constexpr pair();
+    explicit(see-below) pair(const T1& x, const T2& y);                          // constexpr in C++14
+    template <class U, class V> explicit(see-below) pair(U&& x, V&& y);          // constexpr in C++14
+    template <class U, class V> explicit(see-below) pair(const pair<U, V>& p);   // constexpr in C++14
+    template <class U, class V> explicit(see-below) pair(pair<U, V>&& p);        // constexpr in C++14
+    template <class... Args1, class... Args2>
+        pair(piecewise_construct_t, tuple<Args1...> first_args,
+             tuple<Args2...> second_args);
+
+    template <class U, class V> pair& operator=(const pair<U, V>& p);
+    pair& operator=(pair&& p) noexcept(is_nothrow_move_assignable<T1>::value &&
+                                       is_nothrow_move_assignable<T2>::value);
+    template <class U, class V> pair& operator=(pair<U, V>&& p);
+
+    void swap(pair& p) noexcept(is_nothrow_swappable_v<T1> &&
+                                is_nothrow_swappable_v<T2>);
+};
+
+template <class T1, class T2> bool operator==(const pair<T1,T2>&, const pair<T1,T2>&); // constexpr in C++14
+template <class T1, class T2> bool operator!=(const pair<T1,T2>&, const pair<T1,T2>&); // constexpr in C++14
+template <class T1, class T2> bool operator< (const pair<T1,T2>&, const pair<T1,T2>&); // constexpr in C++14
+template <class T1, class T2> bool operator> (const pair<T1,T2>&, const pair<T1,T2>&); // constexpr in C++14
+template <class T1, class T2> bool operator>=(const pair<T1,T2>&, const pair<T1,T2>&); // constexpr in C++14
+template <class T1, class T2> bool operator<=(const pair<T1,T2>&, const pair<T1,T2>&); // constexpr in C++14
+
+template <class T1, class T2> pair<V1, V2> make_pair(T1&&, T2&&);   // constexpr in C++14
+template <class T1, class T2>
+void
+swap(pair<T1, T2>& x, pair<T1, T2>& y) noexcept(noexcept(x.swap(y)));
+
+struct piecewise_construct_t { explicit piecewise_construct_t() = default; };
+inline constexpr piecewise_construct_t piecewise_construct = piecewise_construct_t();
+
+template <class T> struct tuple_size;
+template <size_t I, class T> struct tuple_element;
+
+template <class T1, class T2> struct tuple_size<pair<T1, T2> >;
+template <class T1, class T2> struct tuple_element<0, pair<T1, T2> >;
+template <class T1, class T2> struct tuple_element<1, pair<T1, T2> >;
+
+template<size_t I, class T1, class T2>
+    typename tuple_element<I, pair<T1, T2> >::type&
+    get(pair<T1, T2>&) noexcept; // constexpr in C++14
+
+template<size_t I, class T1, class T2>
+    const typename tuple_element<I, pair<T1, T2> >::type&
+    get(const pair<T1, T2>&) noexcept; // constexpr in C++14
+
+template<size_t I, class T1, class T2>
+    typename tuple_element<I, pair<T1, T2> >::type&&
+    get(pair<T1, T2>&&) noexcept; // constexpr in C++14
+
+template<size_t I, class T1, class T2>
+    const typename tuple_element<I, pair<T1, T2> >::type&&
+    get(const pair<T1, T2>&&) noexcept; // constexpr in C++14
+
+template<class T1, class T2>
+    constexpr T1& get(pair<T1, T2>&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr const T1& get(const pair<T1, T2>&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr T1&& get(pair<T1, T2>&&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr const T1&& get(const pair<T1, T2>&&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr T1& get(pair<T2, T1>&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr const T1& get(const pair<T2, T1>&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr T1&& get(pair<T2, T1>&&) noexcept; // C++14
+
+template<class T1, class T2>
+    constexpr const T1&& get(const pair<T2, T1>&&) noexcept; // C++14
+
+// C++14
+
+template<class T, T... I>
+struct integer_sequence
+{
+    typedef T value_type;
+
+    static constexpr size_t size() noexcept;
+};
+
+template<size_t... I>
+  using index_sequence = integer_sequence<size_t, I...>;
+
+template<class T, T N>
+  using make_integer_sequence = integer_sequence<T, 0, 1, ..., N-1>;
+template<size_t N>
+  using make_index_sequence = make_integer_sequence<size_t, N>;
+
+template<class... T>
+  using index_sequence_for = make_index_sequence<sizeof...(T)>;
+
+template<class T, class U=T>
+    T exchange(T& obj, U&& new_value);
+
+// 20.2.7, in-place construction // C++17
+struct in_place_t {
+  explicit in_place_t() = default;
+};
+inline constexpr in_place_t in_place{};
+template <class T>
+  struct in_place_type_t {
+    explicit in_place_type_t() = default;
+  };
+template <class T>
+  inline constexpr in_place_type_t<T> in_place_type{};
+template <size_t I>
+  struct in_place_index_t {
+    explicit in_place_index_t() = default;
+  };
+template <size_t I>
+  inline constexpr in_place_index_t<I> in_place_index{};
+
+}  // std
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__tuple>
+#include <type_traits>
+#include <initializer_list>
+#include <cstddef>
+#include <cstring>
+#include <cstdint>
+#include <version>
+#include <__debug>
+#include <__pragma_push>
+#endif //__cuda_std__
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+// This header is not currently supported. It exists to allow for the use of
+// `unary_function`, needed by `<cuda/std/functional>`, which happens to be defined here.
+
+#ifndef __cuda_std__
+
+namespace rel_ops
+{
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const _Tp& __x, const _Tp& __y)
+{
+    return !(__x == __y);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const _Tp& __x, const _Tp& __y)
+{
+    return __y < __x;
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const _Tp& __x, const _Tp& __y)
+{
+    return !(__y < __x);
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const _Tp& __x, const _Tp& __y)
+{
+    return !(__x < __y);
+}
+
+}  // rel_ops
+
+// swap_ranges is defined in <type_traits>`
+
+// swap is defined in <type_traits>
+
+// move_if_noexcept
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+#ifndef _LIBCUDACXX_CXX03_LANG
+typename conditional
+<
+    !is_nothrow_move_constructible<_Tp>::value && is_copy_constructible<_Tp>::value,
+    const _Tp&,
+    _Tp&&
+>::type
+#else  // _LIBCUDACXX_CXX03_LANG
+const _Tp&
+#endif
+move_if_noexcept(_Tp& __x) _NOEXCEPT
+{
+    return _CUDA_VSTD::move(__x);
+}
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Tp> constexpr add_const_t<_Tp>& as_const(_Tp& __t) noexcept { return __t; }
+template <class _Tp>                        void as_const(const _Tp&&) = delete;
+#endif
+
+#endif //__cuda_std__
+
+struct _LIBCUDACXX_TEMPLATE_VIS piecewise_construct_t { explicit piecewise_construct_t() = default; };
+#if defined(_LIBCUDACXX_CXX03_LANG) || defined(_LIBCUDACXX_BUILDING_LIBRARY)
+extern _LIBCUDACXX_EXPORTED_FROM_ABI const piecewise_construct_t piecewise_construct;// = piecewise_construct_t();
+#else
+/* _LIBCUDACXX_INLINE_VAR */ constexpr piecewise_construct_t piecewise_construct = piecewise_construct_t();
+#endif
+
+#if defined(_LIBCUDACXX_DEPRECATED_ABI_DISABLE_PAIR_TRIVIAL_COPY_CTOR)
+template <class, class>
+struct __non_trivially_copyable_base {
+  _LIBCUDACXX_CONSTEXPR _LIBCUDACXX_INLINE_VISIBILITY
+  __non_trivially_copyable_base() _NOEXCEPT {}
+  _LIBCUDACXX_CONSTEXPR_AFTER_CXX11 _LIBCUDACXX_INLINE_VISIBILITY
+  __non_trivially_copyable_base(__non_trivially_copyable_base const&) _NOEXCEPT {}
+};
+#endif
+
+template <class _T1, class _T2>
+struct _LIBCUDACXX_TEMPLATE_VIS pair
+#if defined(_LIBCUDACXX_DEPRECATED_ABI_DISABLE_PAIR_TRIVIAL_COPY_CTOR)
+: private __non_trivially_copyable_base<_T1, _T2>
+#endif
+{
+    typedef _T1 first_type;
+    typedef _T2 second_type;
+
+    _T1 first;
+    _T2 second;
+
+#if !defined(_LIBCUDACXX_CXX03_LANG)
+    pair(pair const&) = default;
+    pair(pair&&) = default;
+#else
+  // Use the implicitly declared copy constructor in C++03
+#endif
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair() : first(), second() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair(_T1 const& __t1, _T2 const& __t2) : first(__t1), second(__t2) {}
+
+    template <class _U1, class _U2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair(const pair<_U1, _U2>& __p) : first(__p.first), second(__p.second) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair& operator=(pair const& __p) {
+        first = __p.first;
+        second = __p.second;
+        return *this;
+    }
+#else
+    template <bool _Val>
+    using _EnableB _LIBCUDACXX_NODEBUG_TYPE = typename enable_if<_Val, bool>::type;
+
+    struct _CheckArgs {
+      template <int&...>
+      struct __enable_implicit_default {
+          static constexpr bool value =
+                 __is_implicitly_default_constructible<_T1>::value
+              && __is_implicitly_default_constructible<_T2>::value;
+      };
+
+      template <int&...>
+      struct __enable_explicit_default {
+          static constexpr bool value =
+                 is_default_constructible<_T1>::value
+              && is_default_constructible<_T2>::value
+              && !__enable_implicit_default<>::value;
+      };
+
+      template <class _U1, class _U2>
+      struct __enable_explicit {
+          static constexpr bool value =
+                 is_constructible<first_type, _U1>::value
+              && is_constructible<second_type, _U2>::value
+              && (!is_convertible<_U1, first_type>::value
+                  || !is_convertible<_U2, second_type>::value);
+      };
+
+      template <class _U1, class _U2>
+      struct __enable_implicit {
+          static constexpr bool value =
+                 is_constructible<first_type, _U1>::value
+              && is_constructible<second_type, _U2>::value
+              && is_convertible<_U1, first_type>::value
+              && is_convertible<_U2, second_type>::value;
+      };
+    };
+
+    template <bool _MaybeEnable>
+    using _CheckArgsDep _LIBCUDACXX_NODEBUG_TYPE = typename conditional<
+      _MaybeEnable, _CheckArgs, __check_tuple_constructor_fail>::type;
+
+    struct _CheckTupleLikeConstructor {
+        template <class _Tuple>
+        struct __enable_implicit {
+            static constexpr bool value =
+                __tuple_convertible<_Tuple, pair>::value;
+        };
+
+        template <class _Tuple>
+        struct __enable_explicit {
+            static constexpr bool value =
+                   __tuple_constructible<_Tuple, pair>::value
+               && !__tuple_convertible<_Tuple, pair>::value;
+        };
+
+        template <class _Tuple>
+        struct __enable_assign {
+            static constexpr bool value =
+                __tuple_assignable<_Tuple, pair>::value;
+        };
+    };
+
+    template <class _Tuple>
+    using _CheckTLC _LIBCUDACXX_NODEBUG_TYPE = typename conditional<
+        __tuple_like_with_size<_Tuple, 2>::value
+            && !is_same<typename decay<_Tuple>::type, pair>::value,
+        _CheckTupleLikeConstructor,
+        __check_tuple_constructor_fail
+    >::type;
+
+    template<bool _Dummy = true, _EnableB<
+            _CheckArgsDep<_Dummy>::template __enable_explicit_default<>::value
+    > = false>
+    explicit _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    pair() _NOEXCEPT_(is_nothrow_default_constructible<first_type>::value &&
+                      is_nothrow_default_constructible<second_type>::value)
+        : first(), second() {}
+
+    template<bool _Dummy = true, _EnableB<
+            _CheckArgsDep<_Dummy>::template __enable_implicit_default<>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    pair() _NOEXCEPT_(is_nothrow_default_constructible<first_type>::value &&
+                      is_nothrow_default_constructible<second_type>::value)
+        : first(), second() {}
+
+    template <bool _Dummy = true, _EnableB<
+             _CheckArgsDep<_Dummy>::template __enable_explicit<
+                    typename remove_reference<_T1>::type const&,
+                    typename remove_reference<_T2>::type const&
+                >::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    explicit pair(_T1 const& __t1, _T2 const& __t2)
+        _NOEXCEPT_(is_nothrow_copy_constructible<first_type>::value &&
+                   is_nothrow_copy_constructible<second_type>::value)
+        : first(__t1), second(__t2) {}
+
+    template<bool _Dummy = true, _EnableB<
+            _CheckArgsDep<_Dummy>::template __enable_implicit<
+                    typename remove_reference<_T1>::type const&,
+                    typename remove_reference<_T2>::type const&
+                >::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    pair(_T1 const& __t1, _T2 const& __t2)
+        _NOEXCEPT_(is_nothrow_copy_constructible<first_type>::value &&
+                   is_nothrow_copy_constructible<second_type>::value)
+        : first(__t1), second(__t2) {}
+
+    template<class _U1, class _U2, _EnableB<
+             _CheckArgs::template __enable_explicit<_U1, _U2>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    explicit pair(_U1&& __u1, _U2&& __u2)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _U1>::value &&
+                    is_nothrow_constructible<second_type, _U2>::value))
+        : first(_CUDA_VSTD::forward<_U1>(__u1)), second(_CUDA_VSTD::forward<_U2>(__u2)) {}
+
+    template<class _U1, class _U2, _EnableB<
+            _CheckArgs::template __enable_implicit<_U1, _U2>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    pair(_U1&& __u1, _U2&& __u2)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _U1>::value &&
+                    is_nothrow_constructible<second_type, _U2>::value))
+        : first(_CUDA_VSTD::forward<_U1>(__u1)), second(_CUDA_VSTD::forward<_U2>(__u2)) {}
+
+    template<class _U1, class _U2, _EnableB<
+            _CheckArgs::template __enable_explicit<_U1 const&, _U2 const&>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    explicit pair(pair<_U1, _U2> const& __p)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _U1 const&>::value &&
+                    is_nothrow_constructible<second_type, _U2 const&>::value))
+        : first(__p.first), second(__p.second) {}
+
+    template<class _U1, class _U2, _EnableB<
+            _CheckArgs::template __enable_implicit<_U1 const&, _U2 const&>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    pair(pair<_U1, _U2> const& __p)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _U1 const&>::value &&
+                    is_nothrow_constructible<second_type, _U2 const&>::value))
+        : first(__p.first), second(__p.second) {}
+
+    template<class _U1, class _U2, _EnableB<
+            _CheckArgs::template __enable_explicit<_U1, _U2>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    explicit pair(pair<_U1, _U2>&&__p)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _U1&&>::value &&
+                    is_nothrow_constructible<second_type, _U2&&>::value))
+        : first(_CUDA_VSTD::forward<_U1>(__p.first)), second(_CUDA_VSTD::forward<_U2>(__p.second)) {}
+
+    template<class _U1, class _U2, _EnableB<
+            _CheckArgs::template __enable_implicit<_U1, _U2>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    pair(pair<_U1, _U2>&& __p)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _U1&&>::value &&
+                    is_nothrow_constructible<second_type, _U2&&>::value))
+        : first(_CUDA_VSTD::forward<_U1>(__p.first)), second(_CUDA_VSTD::forward<_U2>(__p.second)) {}
+
+    template<class _Tuple, _EnableB<
+            _CheckTLC<_Tuple>::template __enable_explicit<_Tuple>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    explicit pair(_Tuple&& __p)
+        : first(_CUDA_VSTD::get<0>(_CUDA_VSTD::forward<_Tuple>(__p))),
+          second(_CUDA_VSTD::get<1>(_CUDA_VSTD::forward<_Tuple>(__p))) {}
+
+    template<class _Tuple, _EnableB<
+            _CheckTLC<_Tuple>::template __enable_implicit<_Tuple>::value
+    > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    pair(_Tuple&& __p)
+        : first(_CUDA_VSTD::get<0>(_CUDA_VSTD::forward<_Tuple>(__p))),
+          second(_CUDA_VSTD::get<1>(_CUDA_VSTD::forward<_Tuple>(__p))) {}
+
+    template <class... _Args1, class... _Args2>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair(piecewise_construct_t __pc,
+         tuple<_Args1...> __first_args, tuple<_Args2...> __second_args)
+        _NOEXCEPT_((is_nothrow_constructible<first_type, _Args1...>::value &&
+                    is_nothrow_constructible<second_type, _Args2...>::value))
+        : pair(__pc, __first_args, __second_args,
+                typename __make_tuple_indices<sizeof...(_Args1)>::type(),
+                typename __make_tuple_indices<sizeof...(_Args2) >::type()) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair& operator=(typename conditional<
+                        is_copy_assignable<first_type>::value &&
+                        is_copy_assignable<second_type>::value,
+                    pair, __nat>::type const& __p)
+        _NOEXCEPT_(is_nothrow_copy_assignable<first_type>::value &&
+                   is_nothrow_copy_assignable<second_type>::value)
+    {
+        first = __p.first;
+        second = __p.second;
+        return *this;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair& operator=(typename conditional<
+                        is_move_assignable<first_type>::value &&
+                        is_move_assignable<second_type>::value,
+                    pair, __nat>::type&& __p)
+        _NOEXCEPT_(is_nothrow_move_assignable<first_type>::value &&
+                   is_nothrow_move_assignable<second_type>::value)
+    {
+        first = _CUDA_VSTD::forward<first_type>(__p.first);
+        second = _CUDA_VSTD::forward<second_type>(__p.second);
+        return *this;
+    }
+
+    template <class _Tuple, _EnableB<
+            _CheckTLC<_Tuple>::template __enable_assign<_Tuple>::value
+     > = false>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pair& operator=(_Tuple&& __p) {
+        first = _CUDA_VSTD::get<0>(_CUDA_VSTD::forward<_Tuple>(__p));
+        second = _CUDA_VSTD::get<1>(_CUDA_VSTD::forward<_Tuple>(__p));
+        return *this;
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void
+    swap(pair& __p) _NOEXCEPT_(__is_nothrow_swappable<first_type>::value &&
+                               __is_nothrow_swappable<second_type>::value)
+    {
+        using _CUDA_VSTD::swap;
+        swap(first,  __p.first);
+        swap(second, __p.second);
+    }
+private:
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class... _Args1, class... _Args2, size_t... _I1, size_t... _I2>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        pair(piecewise_construct_t,
+             tuple<_Args1...>& __first_args, tuple<_Args2...>& __second_args,
+             __tuple_indices<_I1...>, __tuple_indices<_I2...>);
+#endif
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _T1, class _T2>
+pair(_T1, _T2) -> pair<_T1, _T2>;
+#endif // _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator==(const pair<_T1,_T2>& __x, const pair<_T1,_T2>& __y)
+{
+    return __x.first == __y.first && __x.second == __y.second;
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator!=(const pair<_T1,_T2>& __x, const pair<_T1,_T2>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator< (const pair<_T1,_T2>& __x, const pair<_T1,_T2>& __y)
+{
+    return __x.first < __y.first || (!(__y.first < __x.first) && __x.second < __y.second);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator> (const pair<_T1,_T2>& __x, const pair<_T1,_T2>& __y)
+{
+    return __y < __x;
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator>=(const pair<_T1,_T2>& __x, const pair<_T1,_T2>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+bool
+operator<=(const pair<_T1,_T2>& __x, const pair<_T1,_T2>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_swappable<_T1>::value &&
+    __is_swappable<_T2>::value,
+    void
+>::type
+swap(pair<_T1, _T2>& __x, pair<_T1, _T2>& __y)
+                     _NOEXCEPT_((__is_nothrow_swappable<_T1>::value &&
+                                 __is_nothrow_swappable<_T2>::value))
+{
+    __x.swap(__y);
+}
+
+template <class _Tp>
+struct __unwrap_reference { typedef _LIBCUDACXX_NODEBUG_TYPE _Tp type; };
+
+template <class _Tp>
+struct __unwrap_reference<reference_wrapper<_Tp> > { typedef _LIBCUDACXX_NODEBUG_TYPE _Tp& type; };
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp>
+struct unwrap_reference : __unwrap_reference<_Tp> { };
+
+template <class _Tp>
+struct unwrap_ref_decay : unwrap_reference<typename decay<_Tp>::type> { };
+#endif // > C++17
+
+template <class _Tp>
+struct __unwrap_ref_decay
+#if _LIBCUDACXX_STD_VER > 17
+    : unwrap_ref_decay<_Tp>
+#else
+    : __unwrap_reference<typename decay<_Tp>::type>
+#endif
+{ };
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+pair<typename __unwrap_ref_decay<_T1>::type, typename __unwrap_ref_decay<_T2>::type>
+make_pair(_T1&& __t1, _T2&& __t2)
+{
+    return pair<typename __unwrap_ref_decay<_T1>::type, typename __unwrap_ref_decay<_T2>::type>
+               (_CUDA_VSTD::forward<_T1>(__t1), _CUDA_VSTD::forward<_T2>(__t2));
+}
+
+#else  // _LIBCUDACXX_CXX03_LANG
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+pair<_T1,_T2>
+make_pair(_T1 __x, _T2 __y)
+{
+    return pair<_T1, _T2>(__x, __y);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _T1, class _T2>
+  struct _LIBCUDACXX_TEMPLATE_VIS tuple_size<pair<_T1, _T2> >
+    : public integral_constant<size_t, 2> {};
+
+template <size_t _Ip, class _T1, class _T2>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<_Ip, pair<_T1, _T2> >
+{
+    static_assert(_Ip < 2, "Index out of bounds in std::tuple_element<std::pair<T1, T2>>");
+};
+
+template <class _T1, class _T2>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<0, pair<_T1, _T2> >
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _T1 type;
+};
+
+template <class _T1, class _T2>
+struct _LIBCUDACXX_TEMPLATE_VIS tuple_element<1, pair<_T1, _T2> >
+{
+    typedef _LIBCUDACXX_NODEBUG_TYPE _T2 type;
+};
+
+template <size_t _Ip> struct __get_pair;
+
+template <>
+struct __get_pair<0>
+{
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    _T1&
+    get(pair<_T1, _T2>& __p) _NOEXCEPT {return __p.first;}
+
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const _T1&
+    get(const pair<_T1, _T2>& __p) _NOEXCEPT {return __p.first;}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    _T1&&
+    get(pair<_T1, _T2>&& __p) _NOEXCEPT {return _CUDA_VSTD::forward<_T1>(__p.first);}
+
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const _T1&&
+    get(const pair<_T1, _T2>&& __p) _NOEXCEPT {return _CUDA_VSTD::forward<const _T1>(__p.first);}
+#endif  // _LIBCUDACXX_CXX03_LANG
+};
+
+template <>
+struct __get_pair<1>
+{
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    _T2&
+    get(pair<_T1, _T2>& __p) _NOEXCEPT {return __p.second;}
+
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const _T2&
+    get(const pair<_T1, _T2>& __p) _NOEXCEPT {return __p.second;}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    _T2&&
+    get(pair<_T1, _T2>&& __p) _NOEXCEPT {return _CUDA_VSTD::forward<_T2>(__p.second);}
+
+    template <class _T1, class _T2>
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+    const _T2&&
+    get(const pair<_T1, _T2>&& __p) _NOEXCEPT {return _CUDA_VSTD::forward<const _T2>(__p.second);}
+#endif  // _LIBCUDACXX_CXX03_LANG
+};
+
+template <size_t _Ip, class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, pair<_T1, _T2> >::type&
+get(pair<_T1, _T2>& __p) _NOEXCEPT
+{
+    return __get_pair<_Ip>::get(__p);
+}
+
+template <size_t _Ip, class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, pair<_T1, _T2> >::type&
+get(const pair<_T1, _T2>& __p) _NOEXCEPT
+{
+    return __get_pair<_Ip>::get(__p);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <size_t _Ip, class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+typename tuple_element<_Ip, pair<_T1, _T2> >::type&&
+get(pair<_T1, _T2>&& __p) _NOEXCEPT
+{
+    return __get_pair<_Ip>::get(_CUDA_VSTD::move(__p));
+}
+
+template <size_t _Ip, class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX11
+const typename tuple_element<_Ip, pair<_T1, _T2> >::type&&
+get(const pair<_T1, _T2>&& __p) _NOEXCEPT
+{
+    return __get_pair<_Ip>::get(_CUDA_VSTD::move(__p));
+}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 & get(pair<_T1, _T2>& __p) _NOEXCEPT
+{
+    return __get_pair<0>::get(__p);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 const & get(pair<_T1, _T2> const& __p) _NOEXCEPT
+{
+    return __get_pair<0>::get(__p);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 && get(pair<_T1, _T2>&& __p) _NOEXCEPT
+{
+    return __get_pair<0>::get(_CUDA_VSTD::move(__p));
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 const && get(pair<_T1, _T2> const&& __p) _NOEXCEPT
+{
+    return __get_pair<0>::get(_CUDA_VSTD::move(__p));
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 & get(pair<_T2, _T1>& __p) _NOEXCEPT
+{
+    return __get_pair<1>::get(__p);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 const & get(pair<_T2, _T1> const& __p) _NOEXCEPT
+{
+    return __get_pair<1>::get(__p);
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 && get(pair<_T2, _T1>&& __p) _NOEXCEPT
+{
+    return __get_pair<1>::get(_CUDA_VSTD::move(__p));
+}
+
+template <class _T1, class _T2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _T1 const && get(pair<_T2, _T1> const&& __p) _NOEXCEPT
+{
+    return __get_pair<1>::get(_CUDA_VSTD::move(__p));
+}
+
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template<class _Tp, _Tp... _Ip>
+struct _LIBCUDACXX_TEMPLATE_VIS integer_sequence
+{
+    typedef _Tp value_type;
+    static_assert( is_integral<_Tp>::value,
+                  "std::integer_sequence can only be instantiated with an integral type" );
+    static
+    _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr
+    size_t
+    size() noexcept { return sizeof...(_Ip); }
+};
+
+template<size_t... _Ip>
+    using index_sequence = integer_sequence<size_t, _Ip...>;
+
+#if __has_builtin(__make_integer_seq) && !defined(_LIBCUDACXX_TESTING_FALLBACK_MAKE_INTEGER_SEQUENCE)
+
+template <class _Tp, _Tp _Ep>
+using __make_integer_sequence _LIBCUDACXX_NODEBUG_TYPE = __make_integer_seq<integer_sequence, _Tp, _Ep>;
+
+#else
+
+template<typename _Tp, _Tp _Np> using __make_integer_sequence_unchecked _LIBCUDACXX_NODEBUG_TYPE  =
+  typename __detail::__make<_Np>::type::template __convert<integer_sequence, _Tp>;
+
+template <class _Tp, _Tp _Ep>
+struct __make_integer_sequence_checked
+{
+    static_assert(is_integral<_Tp>::value,
+                  "std::make_integer_sequence can only be instantiated with an integral type" );
+    static_assert(0 <= _Ep, "std::make_integer_sequence must have a non-negative sequence length");
+    // Workaround GCC bug by preventing bad installations when 0 <= _Ep
+    // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68929
+    typedef _LIBCUDACXX_NODEBUG_TYPE  __make_integer_sequence_unchecked<_Tp, 0 <= _Ep ? _Ep : 0> type;
+};
+
+template <class _Tp, _Tp _Ep>
+using __make_integer_sequence _LIBCUDACXX_NODEBUG_TYPE = typename __make_integer_sequence_checked<_Tp, _Ep>::type;
+
+#endif
+
+template<class _Tp, _Tp _Np>
+    using make_integer_sequence = __make_integer_sequence<_Tp, _Np>;
+
+template<size_t _Np>
+    using make_index_sequence = make_integer_sequence<size_t, _Np>;
+
+template<class... _Tp>
+    using index_sequence_for = make_index_sequence<sizeof...(_Tp)>;
+
+#endif  // _LIBCUDACXX_STD_VER > 11
+
+#if _LIBCUDACXX_STD_VER > 11
+template<class _T1, class _T2 = _T1>
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR_AFTER_CXX17
+_T1 exchange(_T1& __obj, _T2 && __new_value)
+{
+    _T1 __old_value = _CUDA_VSTD::move(__obj);
+    __obj = _CUDA_VSTD::forward<_T2>(__new_value);
+    return __old_value;
+}
+#endif  // _LIBCUDACXX_STD_VER > 11
+
+#if _LIBCUDACXX_STD_VER > 14
+
+struct _LIBCUDACXX_TYPE_VIS in_place_t {
+    explicit in_place_t() = default;
+};
+_LIBCUDACXX_INLINE_VAR constexpr in_place_t in_place{};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS in_place_type_t {
+    explicit in_place_type_t() = default;
+};
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr in_place_type_t<_Tp> in_place_type{};
+
+template <size_t _Idx>
+struct _LIBCUDACXX_TYPE_VIS in_place_index_t {
+    explicit in_place_index_t() = default;
+};
+template <size_t _Idx>
+_LIBCUDACXX_INLINE_VAR constexpr in_place_index_t<_Idx> in_place_index{};
+
+template <class _Tp> struct __is_inplace_type_imp : false_type {};
+template <class _Tp> struct __is_inplace_type_imp<in_place_type_t<_Tp>> : true_type {};
+
+template <class _Tp>
+using __is_inplace_type = __is_inplace_type_imp<__uncvref_t<_Tp>>;
+
+template <class _Tp> struct __is_inplace_index_imp : false_type {};
+template <size_t _Idx> struct __is_inplace_index_imp<in_place_index_t<_Idx>> : true_type {};
+
+template <class _Tp>
+using __is_inplace_index = __is_inplace_index_imp<__uncvref_t<_Tp>>;
+
+#endif // _LIBCUDACXX_STD_VER > 14
+
+template <class _Arg, class _Result>
+struct _LIBCUDACXX_TEMPLATE_VIS unary_function
+{
+    typedef _Arg    argument_type;
+    typedef _Result result_type;
+};
+
+#ifndef __cuda_std__
+
+template <class _Size>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Size
+__loadword(const void* __p)
+{
+    _Size __r;
+    std::memcpy(&__r, __p, sizeof(__r));
+    return __r;
+}
+
+// We use murmur2 when size_t is 32 bits, and cityhash64 when size_t
+// is 64 bits.  This is because cityhash64 uses 64bit x 64bit
+// multiplication, which can be very slow on 32-bit systems.
+template <class _Size, size_t = sizeof(_Size)*__CHAR_BIT__>
+struct __murmur2_or_cityhash;
+
+template <class _Size>
+struct __murmur2_or_cityhash<_Size, 32>
+{
+    inline _Size operator()(const void* __key, _Size __len)
+         _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK;
+};
+
+// murmur2
+template <class _Size>
+_Size
+__murmur2_or_cityhash<_Size, 32>::operator()(const void* __key, _Size __len)
+{
+    const _Size __m = 0x5bd1e995;
+    const _Size __r = 24;
+    _Size __h = __len;
+    const unsigned char* __data = static_cast<const unsigned char*>(__key);
+    for (; __len >= 4; __data += 4, __len -= 4)
+    {
+        _Size __k = __loadword<_Size>(__data);
+        __k *= __m;
+        __k ^= __k >> __r;
+        __k *= __m;
+        __h *= __m;
+        __h ^= __k;
+    }
+    switch (__len)
+    {
+    case 3:
+        __h ^= __data[2] << 16;
+        _LIBCUDACXX_FALLTHROUGH();
+    case 2:
+        __h ^= __data[1] << 8;
+        _LIBCUDACXX_FALLTHROUGH();
+    case 1:
+        __h ^= __data[0];
+        __h *= __m;
+    }
+    __h ^= __h >> 13;
+    __h *= __m;
+    __h ^= __h >> 15;
+    return __h;
+}
+
+template <class _Size>
+struct __murmur2_or_cityhash<_Size, 64>
+{
+    inline _Size operator()(const void* __key, _Size __len)  _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK;
+
+ private:
+  // Some primes between 2^63 and 2^64.
+  static const _Size __k0 = 0xc3a5c85c97cb3127ULL;
+  static const _Size __k1 = 0xb492b66fbe98f273ULL;
+  static const _Size __k2 = 0x9ae16a3b2f90404fULL;
+  static const _Size __k3 = 0xc949d7c7509e6557ULL;
+
+  static _Size __rotate(_Size __val, int __shift) {
+    return __shift == 0 ? __val : ((__val >> __shift) | (__val << (64 - __shift)));
+  }
+
+  static _Size __rotate_by_at_least_1(_Size __val, int __shift) {
+    return (__val >> __shift) | (__val << (64 - __shift));
+  }
+
+  static _Size __shift_mix(_Size __val) {
+    return __val ^ (__val >> 47);
+  }
+
+  static _Size __hash_len_16(_Size __u, _Size __v)
+     _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+  {
+    const _Size __mul = 0x9ddfea08eb382d69ULL;
+    _Size __a = (__u ^ __v) * __mul;
+    __a ^= (__a >> 47);
+    _Size __b = (__v ^ __a) * __mul;
+    __b ^= (__b >> 47);
+    __b *= __mul;
+    return __b;
+  }
+
+  static _Size __hash_len_0_to_16(const char* __s, _Size __len)
+     _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+  {
+    if (__len > 8) {
+      const _Size __a = __loadword<_Size>(__s);
+      const _Size __b = __loadword<_Size>(__s + __len - 8);
+      return __hash_len_16(__a, __rotate_by_at_least_1(__b + __len, __len)) ^ __b;
+    }
+    if (__len >= 4) {
+      const uint32_t __a = __loadword<uint32_t>(__s);
+      const uint32_t __b = __loadword<uint32_t>(__s + __len - 4);
+      return __hash_len_16(__len + (__a << 3), __b);
+    }
+    if (__len > 0) {
+      const unsigned char __a = __s[0];
+      const unsigned char __b = __s[__len >> 1];
+      const unsigned char __c = __s[__len - 1];
+      const uint32_t __y = static_cast<uint32_t>(__a) +
+                           (static_cast<uint32_t>(__b) << 8);
+      const uint32_t __z = __len + (static_cast<uint32_t>(__c) << 2);
+      return __shift_mix(__y * __k2 ^ __z * __k3) * __k2;
+    }
+    return __k2;
+  }
+
+  static _Size __hash_len_17_to_32(const char *__s, _Size __len)
+     _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+  {
+    const _Size __a = __loadword<_Size>(__s) * __k1;
+    const _Size __b = __loadword<_Size>(__s + 8);
+    const _Size __c = __loadword<_Size>(__s + __len - 8) * __k2;
+    const _Size __d = __loadword<_Size>(__s + __len - 16) * __k0;
+    return __hash_len_16(__rotate(__a - __b, 43) + __rotate(__c, 30) + __d,
+                         __a + __rotate(__b ^ __k3, 20) - __c + __len);
+  }
+
+  // Return a 16-byte hash for 48 bytes.  Quick and dirty.
+  // Callers do best to use "random-looking" values for a and b.
+  static pair<_Size, _Size> __weak_hash_len_32_with_seeds(
+      _Size __w, _Size __x, _Size __y, _Size __z, _Size __a, _Size __b)
+        _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+  {
+    __a += __w;
+    __b = __rotate(__b + __a + __z, 21);
+    const _Size __c = __a;
+    __a += __x;
+    __a += __y;
+    __b += __rotate(__a, 44);
+    return pair<_Size, _Size>(__a + __z, __b + __c);
+  }
+
+  // Return a 16-byte hash for s[0] ... s[31], a, and b.  Quick and dirty.
+  static pair<_Size, _Size> __weak_hash_len_32_with_seeds(
+      const char* __s, _Size __a, _Size __b)
+    _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+  {
+    return __weak_hash_len_32_with_seeds(__loadword<_Size>(__s),
+                                         __loadword<_Size>(__s + 8),
+                                         __loadword<_Size>(__s + 16),
+                                         __loadword<_Size>(__s + 24),
+                                         __a,
+                                         __b);
+  }
+
+  // Return an 8-byte hash for 33 to 64 bytes.
+  static _Size __hash_len_33_to_64(const char *__s, size_t __len)
+    _LIBCUDACXX_DISABLE_UBSAN_UNSIGNED_INTEGER_CHECK
+  {
+    _Size __z = __loadword<_Size>(__s + 24);
+    _Size __a = __loadword<_Size>(__s) +
+                (__len + __loadword<_Size>(__s + __len - 16)) * __k0;
+    _Size __b = __rotate(__a + __z, 52);
+    _Size __c = __rotate(__a, 37);
+    __a += __loadword<_Size>(__s + 8);
+    __c += __rotate(__a, 7);
+    __a += __loadword<_Size>(__s + 16);
+    _Size __vf = __a + __z;
+    _Size __vs = __b + __rotate(__a, 31) + __c;
+    __a = __loadword<_Size>(__s + 16) + __loadword<_Size>(__s + __len - 32);
+    __z += __loadword<_Size>(__s + __len - 8);
+    __b = __rotate(__a + __z, 52);
+    __c = __rotate(__a, 37);
+    __a += __loadword<_Size>(__s + __len - 24);
+    __c += __rotate(__a, 7);
+    __a += __loadword<_Size>(__s + __len - 16);
+    _Size __wf = __a + __z;
+    _Size __ws = __b + __rotate(__a, 31) + __c;
+    _Size __r = __shift_mix((__vf + __ws) * __k2 + (__wf + __vs) * __k0);
+    return __shift_mix(__r * __k0 + __vs) * __k2;
+  }
+};
+
+// cityhash64
+template <class _Size>
+_Size
+__murmur2_or_cityhash<_Size, 64>::operator()(const void* __key, _Size __len)
+{
+  const char* __s = static_cast<const char*>(__key);
+  if (__len <= 32) {
+    if (__len <= 16) {
+      return __hash_len_0_to_16(__s, __len);
+    } else {
+      return __hash_len_17_to_32(__s, __len);
+    }
+  } else if (__len <= 64) {
+    return __hash_len_33_to_64(__s, __len);
+  }
+
+  // For strings over 64 bytes we hash the end first, and then as we
+  // loop we keep 56 bytes of state: v, w, x, y, and z.
+  _Size __x = __loadword<_Size>(__s + __len - 40);
+  _Size __y = __loadword<_Size>(__s + __len - 16) +
+              __loadword<_Size>(__s + __len - 56);
+  _Size __z = __hash_len_16(__loadword<_Size>(__s + __len - 48) + __len,
+                          __loadword<_Size>(__s + __len - 24));
+  pair<_Size, _Size> __v = __weak_hash_len_32_with_seeds(__s + __len - 64, __len, __z);
+  pair<_Size, _Size> __w = __weak_hash_len_32_with_seeds(__s + __len - 32, __y + __k1, __x);
+  __x = __x * __k1 + __loadword<_Size>(__s);
+
+  // Decrease len to the nearest multiple of 64, and operate on 64-byte chunks.
+  __len = (__len - 1) & ~static_cast<_Size>(63);
+  do {
+    __x = __rotate(__x + __y + __v.first + __loadword<_Size>(__s + 8), 37) * __k1;
+    __y = __rotate(__y + __v.second + __loadword<_Size>(__s + 48), 42) * __k1;
+    __x ^= __w.second;
+    __y += __v.first + __loadword<_Size>(__s + 40);
+    __z = __rotate(__z + __w.first, 33) * __k1;
+    __v = __weak_hash_len_32_with_seeds(__s, __v.second * __k1, __x + __w.first);
+    __w = __weak_hash_len_32_with_seeds(__s + 32, __z + __w.second,
+                                        __y + __loadword<_Size>(__s + 16));
+    std::swap(__z, __x);
+    __s += 64;
+    __len -= 64;
+  } while (__len != 0);
+  return __hash_len_16(
+      __hash_len_16(__v.first, __w.first) + __shift_mix(__y) * __k1 + __z,
+      __hash_len_16(__v.second, __w.second) + __x);
+}
+
+template <class _Tp, size_t = sizeof(_Tp) / sizeof(size_t)>
+struct __scalar_hash;
+
+template <class _Tp>
+struct __scalar_hash<_Tp, 0>
+    : public unary_function<_Tp, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp __v) const _NOEXCEPT
+    {
+        union
+        {
+            _Tp    __t;
+            size_t __a;
+        } __u;
+        __u.__a = 0;
+        __u.__t = __v;
+        return __u.__a;
+    }
+};
+
+template <class _Tp>
+struct __scalar_hash<_Tp, 1>
+    : public unary_function<_Tp, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp __v) const _NOEXCEPT
+    {
+        union
+        {
+            _Tp    __t;
+            size_t __a;
+        } __u;
+        __u.__t = __v;
+        return __u.__a;
+    }
+};
+
+template <class _Tp>
+struct __scalar_hash<_Tp, 2>
+    : public unary_function<_Tp, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp __v) const _NOEXCEPT
+    {
+        union
+        {
+            _Tp __t;
+            struct
+            {
+                size_t __a;
+                size_t __b;
+            } __s;
+        } __u;
+        __u.__t = __v;
+        return __murmur2_or_cityhash<size_t>()(&__u, sizeof(__u));
+    }
+};
+
+template <class _Tp>
+struct __scalar_hash<_Tp, 3>
+    : public unary_function<_Tp, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp __v) const _NOEXCEPT
+    {
+        union
+        {
+            _Tp __t;
+            struct
+            {
+                size_t __a;
+                size_t __b;
+                size_t __c;
+            } __s;
+        } __u;
+        __u.__t = __v;
+        return __murmur2_or_cityhash<size_t>()(&__u, sizeof(__u));
+    }
+};
+
+template <class _Tp>
+struct __scalar_hash<_Tp, 4>
+    : public unary_function<_Tp, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp __v) const _NOEXCEPT
+    {
+        union
+        {
+            _Tp __t;
+            struct
+            {
+                size_t __a;
+                size_t __b;
+                size_t __c;
+                size_t __d;
+            } __s;
+        } __u;
+        __u.__t = __v;
+        return __murmur2_or_cityhash<size_t>()(&__u, sizeof(__u));
+    }
+};
+
+struct _PairT {
+  size_t first;
+  size_t second;
+};
+
+_LIBCUDACXX_INLINE_VISIBILITY
+inline size_t __hash_combine(size_t __lhs, size_t __rhs) _NOEXCEPT {
+    typedef __scalar_hash<_PairT> _HashT;
+    const _PairT __p = {__lhs, __rhs};
+    return _HashT()(__p);
+}
+
+template<class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<_Tp*>
+    : public unary_function<_Tp*, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp* __v) const _NOEXCEPT
+    {
+        union
+        {
+            _Tp* __t;
+            size_t __a;
+        } __u;
+        __u.__t = __v;
+        return __murmur2_or_cityhash<size_t>()(&__u, sizeof(__u));
+    }
+};
+
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<bool>
+    : public unary_function<bool, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(bool __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<char>
+    : public unary_function<char, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(char __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<signed char>
+    : public unary_function<signed char, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(signed char __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned char>
+    : public unary_function<unsigned char, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned char __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<char16_t>
+    : public unary_function<char16_t, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(char16_t __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<char32_t>
+    : public unary_function<char32_t, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(char32_t __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+#endif  // _LIBCUDACXX_HAS_NO_UNICODE_CHARS
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<wchar_t>
+    : public unary_function<wchar_t, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(wchar_t __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<short>
+    : public unary_function<short, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(short __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned short>
+    : public unary_function<unsigned short, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned short __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<int>
+    : public unary_function<int, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(int __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned int>
+    : public unary_function<unsigned int, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned int __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<long>
+    : public unary_function<long, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(long __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned long>
+    : public unary_function<unsigned long, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(unsigned long __v) const _NOEXCEPT {return static_cast<size_t>(__v);}
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<long long>
+    : public __scalar_hash<long long>
+{
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<unsigned long long>
+    : public __scalar_hash<unsigned long long>
+{
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_INT128
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<__int128_t>
+    : public __scalar_hash<__int128_t>
+{
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<__uint128_t>
+    : public __scalar_hash<__uint128_t>
+{
+};
+
+#endif
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<float>
+    : public __scalar_hash<float>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(float __v) const _NOEXCEPT
+    {
+        // -0.0 and 0.0 should return same hash
+       if (__v == 0.0f)
+           return 0;
+        return __scalar_hash<float>::operator()(__v);
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<double>
+    : public __scalar_hash<double>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(double __v) const _NOEXCEPT
+    {
+        // -0.0 and 0.0 should return same hash
+       if (__v == 0.0)
+           return 0;
+        return __scalar_hash<double>::operator()(__v);
+    }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<long double>
+    : public __scalar_hash<long double>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(long double __v) const _NOEXCEPT
+    {
+        // -0.0 and 0.0 should return same hash
+        if (__v == 0.0L)
+            return 0;
+#if defined(__i386__)
+        // Zero out padding bits
+        union
+        {
+            long double __t;
+            struct
+            {
+                size_t __a;
+                size_t __b;
+                size_t __c;
+                size_t __d;
+            } __s;
+        } __u;
+        __u.__s.__a = 0;
+        __u.__s.__b = 0;
+        __u.__s.__c = 0;
+        __u.__s.__d = 0;
+        __u.__t = __v;
+        return __u.__s.__a ^ __u.__s.__b ^ __u.__s.__c ^ __u.__s.__d;
+#elif defined(__x86_64__)
+        // Zero out padding bits
+        union
+        {
+            long double __t;
+            struct
+            {
+                size_t __a;
+                size_t __b;
+            } __s;
+        } __u;
+        __u.__s.__a = 0;
+        __u.__s.__b = 0;
+        __u.__t = __v;
+        return __u.__s.__a ^ __u.__s.__b;
+#else
+        return __scalar_hash<long double>::operator()(__v);
+#endif
+    }
+};
+
+#if _LIBCUDACXX_STD_VER > 11
+
+template <class _Tp, bool = is_enum<_Tp>::value>
+struct _LIBCUDACXX_TEMPLATE_VIS __enum_hash
+    : public unary_function<_Tp, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(_Tp __v) const _NOEXCEPT
+    {
+        typedef typename underlying_type<_Tp>::type type;
+        return hash<type>{}(static_cast<type>(__v));
+    }
+};
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS __enum_hash<_Tp, false> {
+    __enum_hash() = delete;
+    __enum_hash(__enum_hash const&) = delete;
+    __enum_hash& operator=(__enum_hash const&) = delete;
+};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS hash : public __enum_hash<_Tp>
+{
+};
+#endif
+
+#if _LIBCUDACXX_STD_VER > 14
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<nullptr_t>
+  : public unary_function<nullptr_t, size_t>
+{
+  _LIBCUDACXX_INLINE_VISIBILITY
+  size_t operator()(nullptr_t) const _NOEXCEPT {
+    return 662607004ull;
+  }
+};
+#endif
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Key, class _Hash>
+using __check_hash_requirements _LIBCUDACXX_NODEBUG_TYPE  = integral_constant<bool,
+    is_copy_constructible<_Hash>::value &&
+    is_move_constructible<_Hash>::value &&
+    __invokable_r<size_t, _Hash, _Key const&>::value
+>;
+
+template <class _Key, class _Hash = std::hash<_Key> >
+using __has_enabled_hash _LIBCUDACXX_NODEBUG_TYPE = integral_constant<bool,
+    __check_hash_requirements<_Key, _Hash>::value &&
+    is_default_constructible<_Hash>::value
+>;
+
+#if _LIBCUDACXX_STD_VER > 14
+template <class _Type, class>
+using __enable_hash_helper_imp _LIBCUDACXX_NODEBUG_TYPE  = _Type;
+
+template <class _Type, class ..._Keys>
+using __enable_hash_helper _LIBCUDACXX_NODEBUG_TYPE  = __enable_hash_helper_imp<_Type,
+  typename enable_if<__all<__has_enabled_hash<_Keys>::value...>::value>::type
+>;
+#else
+template <class _Type, class ...>
+using __enable_hash_helper _LIBCUDACXX_NODEBUG_TYPE = _Type;
+#endif
+
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+#endif // __cuda_std__
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+#ifndef __cuda_std__
+#include <__pragma_pop>
+#endif //__cuda_std__
+
+#endif  // _LIBCUDACXX_UTILITY
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/valarray b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/valarray
new file mode 100644
index 000000000000..0d02d4ab2eed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/valarray
@@ -0,0 +1,4930 @@
+// -*- C++ -*-
+//===-------------------------- valarray ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_VALARRAY
+#define _LIBCUDACXX_VALARRAY
+
+/*
+    valarray synopsis
+
+namespace std
+{
+
+template<class T>
+class valarray
+{
+public:
+    typedef T value_type;
+
+    // construct/destroy:
+    valarray();
+    explicit valarray(size_t n);
+    valarray(const value_type& x, size_t n);
+    valarray(const value_type* px, size_t n);
+    valarray(const valarray& v);
+    valarray(valarray&& v) noexcept;
+    valarray(const slice_array<value_type>& sa);
+    valarray(const gslice_array<value_type>& ga);
+    valarray(const mask_array<value_type>& ma);
+    valarray(const indirect_array<value_type>& ia);
+    valarray(initializer_list<value_type> il);
+    ~valarray();
+
+    // assignment:
+    valarray& operator=(const valarray& v);
+    valarray& operator=(valarray&& v) noexcept;
+    valarray& operator=(initializer_list<value_type> il);
+    valarray& operator=(const value_type& x);
+    valarray& operator=(const slice_array<value_type>& sa);
+    valarray& operator=(const gslice_array<value_type>& ga);
+    valarray& operator=(const mask_array<value_type>& ma);
+    valarray& operator=(const indirect_array<value_type>& ia);
+
+    // element access:
+    const value_type& operator[](size_t i) const;
+    value_type&       operator[](size_t i);
+
+    // subset operations:
+    valarray                   operator[](slice s) const;
+    slice_array<value_type>    operator[](slice s);
+    valarray                   operator[](const gslice& gs) const;
+    gslice_array<value_type>   operator[](const gslice& gs);
+    valarray                   operator[](const valarray<bool>& vb) const;
+    mask_array<value_type>     operator[](const valarray<bool>& vb);
+    valarray                   operator[](const valarray<size_t>& vs) const;
+    indirect_array<value_type> operator[](const valarray<size_t>& vs);
+
+    // unary operators:
+    valarray       operator+() const;
+    valarray       operator-() const;
+    valarray       operator~() const;
+    valarray<bool> operator!() const;
+
+    // computed assignment:
+    valarray& operator*= (const value_type& x);
+    valarray& operator/= (const value_type& x);
+    valarray& operator%= (const value_type& x);
+    valarray& operator+= (const value_type& x);
+    valarray& operator-= (const value_type& x);
+    valarray& operator^= (const value_type& x);
+    valarray& operator&= (const value_type& x);
+    valarray& operator|= (const value_type& x);
+    valarray& operator<<=(const value_type& x);
+    valarray& operator>>=(const value_type& x);
+
+    valarray& operator*= (const valarray& v);
+    valarray& operator/= (const valarray& v);
+    valarray& operator%= (const valarray& v);
+    valarray& operator+= (const valarray& v);
+    valarray& operator-= (const valarray& v);
+    valarray& operator^= (const valarray& v);
+    valarray& operator|= (const valarray& v);
+    valarray& operator&= (const valarray& v);
+    valarray& operator<<=(const valarray& v);
+    valarray& operator>>=(const valarray& v);
+
+    // member functions:
+    void swap(valarray& v) noexcept;
+
+    size_t size() const;
+
+    value_type sum() const;
+    value_type min() const;
+    value_type max() const;
+
+    valarray shift (int i) const;
+    valarray cshift(int i) const;
+    valarray apply(value_type f(value_type)) const;
+    valarray apply(value_type f(const value_type&)) const;
+    void resize(size_t n, value_type x = value_type());
+};
+
+class slice
+{
+public:
+    slice();
+    slice(size_t start, size_t size, size_t stride);
+
+    size_t start()  const;
+    size_t size()   const;
+    size_t stride() const;
+};
+
+template <class T>
+class slice_array
+{
+public:
+    typedef T value_type;
+
+    const slice_array& operator=(const slice_array& sa) const;
+    void operator=  (const valarray<value_type>& v) const;
+    void operator*= (const valarray<value_type>& v) const;
+    void operator/= (const valarray<value_type>& v) const;
+    void operator%= (const valarray<value_type>& v) const;
+    void operator+= (const valarray<value_type>& v) const;
+    void operator-= (const valarray<value_type>& v) const;
+    void operator^= (const valarray<value_type>& v) const;
+    void operator&= (const valarray<value_type>& v) const;
+    void operator|= (const valarray<value_type>& v) const;
+    void operator<<=(const valarray<value_type>& v) const;
+    void operator>>=(const valarray<value_type>& v) const;
+
+    void operator=(const value_type& x) const;
+
+    slice_array() = delete;
+};
+
+class gslice
+{
+public:
+    gslice();
+    gslice(size_t start, const valarray<size_t>& size,
+                         const valarray<size_t>& stride);
+
+    size_t           start()  const;
+    valarray<size_t> size()   const;
+    valarray<size_t> stride() const;
+};
+
+template <class T>
+class gslice_array
+{
+public:
+    typedef T value_type;
+
+    void operator=  (const valarray<value_type>& v) const;
+    void operator*= (const valarray<value_type>& v) const;
+    void operator/= (const valarray<value_type>& v) const;
+    void operator%= (const valarray<value_type>& v) const;
+    void operator+= (const valarray<value_type>& v) const;
+    void operator-= (const valarray<value_type>& v) const;
+    void operator^= (const valarray<value_type>& v) const;
+    void operator&= (const valarray<value_type>& v) const;
+    void operator|= (const valarray<value_type>& v) const;
+    void operator<<=(const valarray<value_type>& v) const;
+    void operator>>=(const valarray<value_type>& v) const;
+
+    gslice_array(const gslice_array& ga);
+    ~gslice_array();
+    const gslice_array& operator=(const gslice_array& ga) const;
+    void operator=(const value_type& x) const;
+
+    gslice_array() = delete;
+};
+
+template <class T>
+class mask_array
+{
+public:
+    typedef T value_type;
+
+    void operator=  (const valarray<value_type>& v) const;
+    void operator*= (const valarray<value_type>& v) const;
+    void operator/= (const valarray<value_type>& v) const;
+    void operator%= (const valarray<value_type>& v) const;
+    void operator+= (const valarray<value_type>& v) const;
+    void operator-= (const valarray<value_type>& v) const;
+    void operator^= (const valarray<value_type>& v) const;
+    void operator&= (const valarray<value_type>& v) const;
+    void operator|= (const valarray<value_type>& v) const;
+    void operator<<=(const valarray<value_type>& v) const;
+    void operator>>=(const valarray<value_type>& v) const;
+
+    mask_array(const mask_array& ma);
+    ~mask_array();
+    const mask_array& operator=(const mask_array& ma) const;
+    void operator=(const value_type& x) const;
+
+    mask_array() = delete;
+};
+
+template <class T>
+class indirect_array
+{
+public:
+    typedef T value_type;
+
+    void operator=  (const valarray<value_type>& v) const;
+    void operator*= (const valarray<value_type>& v) const;
+    void operator/= (const valarray<value_type>& v) const;
+    void operator%= (const valarray<value_type>& v) const;
+    void operator+= (const valarray<value_type>& v) const;
+    void operator-= (const valarray<value_type>& v) const;
+    void operator^= (const valarray<value_type>& v) const;
+    void operator&= (const valarray<value_type>& v) const;
+    void operator|= (const valarray<value_type>& v) const;
+    void operator<<=(const valarray<value_type>& v) const;
+    void operator>>=(const valarray<value_type>& v) const;
+
+    indirect_array(const indirect_array& ia);
+    ~indirect_array();
+    const indirect_array& operator=(const indirect_array& ia) const;
+    void operator=(const value_type& x) const;
+
+    indirect_array() = delete;
+};
+
+template<class T> void swap(valarray<T>& x, valarray<T>& y) noexcept;
+
+template<class T> valarray<T> operator* (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator* (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator* (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator/ (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator/ (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator/ (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator% (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator% (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator% (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator+ (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator+ (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator+ (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator- (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator- (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator- (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator^ (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator^ (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator^ (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator& (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator& (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator& (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator| (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator| (const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator| (const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator<<(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator<<(const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator<<(const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> operator>>(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> operator>>(const valarray<T>& x, const T& y);
+template<class T> valarray<T> operator>>(const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator&&(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator&&(const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator&&(const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator||(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator||(const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator||(const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator==(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator==(const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator==(const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator!=(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator!=(const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator!=(const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator< (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator< (const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator< (const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator> (const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator> (const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator> (const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator<=(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator<=(const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator<=(const T& x, const valarray<T>& y);
+
+template<class T> valarray<bool> operator>=(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<bool> operator>=(const valarray<T>& x, const T& y);
+template<class T> valarray<bool> operator>=(const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> abs (const valarray<T>& x);
+template<class T> valarray<T> acos (const valarray<T>& x);
+template<class T> valarray<T> asin (const valarray<T>& x);
+template<class T> valarray<T> atan (const valarray<T>& x);
+
+template<class T> valarray<T> atan2(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> atan2(const valarray<T>& x, const T& y);
+template<class T> valarray<T> atan2(const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> cos (const valarray<T>& x);
+template<class T> valarray<T> cosh (const valarray<T>& x);
+template<class T> valarray<T> exp (const valarray<T>& x);
+template<class T> valarray<T> log (const valarray<T>& x);
+template<class T> valarray<T> log10(const valarray<T>& x);
+
+template<class T> valarray<T> pow(const valarray<T>& x, const valarray<T>& y);
+template<class T> valarray<T> pow(const valarray<T>& x, const T& y);
+template<class T> valarray<T> pow(const T& x, const valarray<T>& y);
+
+template<class T> valarray<T> sin (const valarray<T>& x);
+template<class T> valarray<T> sinh (const valarray<T>& x);
+template<class T> valarray<T> sqrt (const valarray<T>& x);
+template<class T> valarray<T> tan (const valarray<T>& x);
+template<class T> valarray<T> tanh (const valarray<T>& x);
+
+template <class T> unspecified1 begin(valarray<T>& v);
+template <class T> unspecified2 begin(const valarray<T>& v);
+template <class T> unspecified1 end(valarray<T>& v);
+template <class T> unspecified2 end(const valarray<T>& v);
+
+}  // std
+
+*/
+
+#include <__config>
+#include <cstddef>
+#include <cmath>
+#include <initializer_list>
+#include <algorithm>
+#include <functional>
+#include <new>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template<class _Tp> class _LIBCUDACXX_TEMPLATE_VIS valarray;
+
+class _LIBCUDACXX_TEMPLATE_VIS slice
+{
+    size_t __start_;
+    size_t __size_;
+    size_t __stride_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    slice()
+        : __start_(0),
+          __size_(0),
+          __stride_(0)
+          {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    slice(size_t __start, size_t __size, size_t __stride)
+        : __start_(__start),
+          __size_(__size),
+          __stride_(__stride)
+          {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY size_t start()  const {return __start_;}
+    _LIBCUDACXX_INLINE_VISIBILITY size_t size()   const {return __size_;}
+    _LIBCUDACXX_INLINE_VISIBILITY size_t stride() const {return __stride_;}
+};
+
+template <class _Tp> class _LIBCUDACXX_TEMPLATE_VIS slice_array;
+class _LIBCUDACXX_TYPE_VIS gslice;
+template <class _Tp> class _LIBCUDACXX_TEMPLATE_VIS gslice_array;
+template <class _Tp> class _LIBCUDACXX_TEMPLATE_VIS mask_array;
+template <class _Tp> class _LIBCUDACXX_TEMPLATE_VIS indirect_array;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+begin(valarray<_Tp>& __v);
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+const _Tp*
+begin(const valarray<_Tp>& __v);
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+end(valarray<_Tp>& __v);
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VISIBILITY
+const _Tp*
+end(const valarray<_Tp>& __v);
+
+template <class _Op, class _A0>
+struct _UnaryOp
+{
+    typedef typename _Op::result_type result_type;
+    typedef typename _A0::value_type value_type;
+
+    _Op __op_;
+    _A0 __a0_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _UnaryOp(const _Op& __op, const _A0& __a0) : __op_(__op), __a0_(__a0) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const {return __op_(__a0_[__i]);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __a0_.size();}
+};
+
+template <class _Op, class _A0, class _A1>
+struct _BinaryOp
+{
+    typedef typename _Op::result_type result_type;
+    typedef typename _A0::value_type value_type;
+
+    _Op __op_;
+    _A0 __a0_;
+    _A1 __a1_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _BinaryOp(const _Op& __op, const _A0& __a0, const _A1& __a1)
+        : __op_(__op), __a0_(__a0), __a1_(__a1) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type operator[](size_t __i) const {return __op_(__a0_[__i], __a1_[__i]);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __a0_.size();}
+};
+
+template <class _Tp>
+class __scalar_expr
+{
+public:
+    typedef _Tp        value_type;
+    typedef const _Tp& result_type;
+private:
+    const value_type& __t_;
+    size_t __s_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __scalar_expr(const value_type& __t, size_t __s) : __t_(__t), __s_(__s) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t) const {return __t_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __s_;}
+};
+
+template <class _Tp>
+struct __unary_plus : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return +__x;}
+};
+
+template <class _Tp>
+struct __bit_not  : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return ~__x;}
+};
+
+template <class _Tp>
+struct __bit_shift_left : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x << __y;}
+};
+
+template <class _Tp>
+struct __bit_shift_right : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return __x >> __y;}
+};
+
+template <class _Tp, class _Fp>
+struct __apply_expr   : unary_function<_Tp, _Tp>
+{
+private:
+    _Fp __f_;
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __apply_expr(_Fp __f) : __f_(__f) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return __f_(__x);}
+};
+
+template <class _Tp>
+struct __abs_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return abs(__x);}
+};
+
+template <class _Tp>
+struct __acos_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return acos(__x);}
+};
+
+template <class _Tp>
+struct __asin_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return asin(__x);}
+};
+
+template <class _Tp>
+struct __atan_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return atan(__x);}
+};
+
+template <class _Tp>
+struct __atan2_expr : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return atan2(__x, __y);}
+};
+
+template <class _Tp>
+struct __cos_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return cos(__x);}
+};
+
+template <class _Tp>
+struct __cosh_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return cosh(__x);}
+};
+
+template <class _Tp>
+struct __exp_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return exp(__x);}
+};
+
+template <class _Tp>
+struct __log_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return log(__x);}
+};
+
+template <class _Tp>
+struct __log10_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return log10(__x);}
+};
+
+template <class _Tp>
+struct __pow_expr : binary_function<_Tp, _Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x, const _Tp& __y) const
+        {return pow(__x, __y);}
+};
+
+template <class _Tp>
+struct __sin_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return sin(__x);}
+};
+
+template <class _Tp>
+struct __sinh_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return sinh(__x);}
+};
+
+template <class _Tp>
+struct __sqrt_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return sqrt(__x);}
+};
+
+template <class _Tp>
+struct __tan_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return tan(__x);}
+};
+
+template <class _Tp>
+struct __tanh_expr : unary_function<_Tp, _Tp>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _Tp operator()(const _Tp& __x) const
+        {return tanh(__x);}
+};
+
+template <class _ValExpr>
+class __slice_expr
+{
+    typedef typename remove_reference<_ValExpr>::type  _RmExpr;
+public:
+    typedef typename _RmExpr::value_type value_type;
+    typedef value_type result_type;
+
+private:
+    _ValExpr __expr_;
+    size_t __start_;
+    size_t __size_;
+    size_t __stride_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __slice_expr(const slice& __sl, const _RmExpr& __e)
+        : __expr_(__e),
+          __start_(__sl.start()),
+          __size_(__sl.size()),
+          __stride_(__sl.stride())
+        {}
+public:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const
+        {return __expr_[__start_ + __i * __stride_];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __size_;}
+
+    template <class> friend class __val_expr;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS valarray;
+};
+
+template <class _ValExpr>
+class __mask_expr;
+
+template <class _ValExpr>
+class __indirect_expr;
+
+template <class _ValExpr>
+class __shift_expr
+{
+    typedef typename remove_reference<_ValExpr>::type  _RmExpr;
+public:
+    typedef typename _RmExpr::value_type value_type;
+    typedef value_type result_type;
+
+private:
+    _ValExpr __expr_;
+    size_t __size_;
+    ptrdiff_t __ul_;
+    ptrdiff_t __sn_;
+    ptrdiff_t __n_;
+    static const ptrdiff_t _Np = static_cast<ptrdiff_t>(
+                                    sizeof(ptrdiff_t) * __CHAR_BIT__ - 1);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __shift_expr(int __n, const _RmExpr& __e)
+        : __expr_(__e),
+          __size_(__e.size()),
+          __n_(__n)
+        {
+            ptrdiff_t __neg_n = static_cast<ptrdiff_t>(__n_ >> _Np);
+            __sn_ = __neg_n | static_cast<ptrdiff_t>(static_cast<size_t>(-__n_) >> _Np);
+            __ul_ = ((__size_ - __n_) & ~__neg_n) | ((__n_ + 1) & __neg_n);
+        }
+public:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __j) const
+        {
+            ptrdiff_t __i = static_cast<ptrdiff_t>(__j);
+            ptrdiff_t __m = (__sn_ * __i - __ul_) >> _Np;
+            return (__expr_[(__i + __n_) & __m] & __m) | (value_type() & ~__m);
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __size_;}
+
+    template <class> friend class __val_expr;
+};
+
+template <class _ValExpr>
+class __cshift_expr
+{
+    typedef typename remove_reference<_ValExpr>::type  _RmExpr;
+public:
+    typedef typename _RmExpr::value_type value_type;
+    typedef value_type result_type;
+
+private:
+    _ValExpr __expr_;
+    size_t __size_;
+    size_t __m_;
+    size_t __o1_;
+    size_t __o2_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __cshift_expr(int __n, const _RmExpr& __e)
+        : __expr_(__e),
+          __size_(__e.size())
+        {
+            __n %= static_cast<int>(__size_);
+            if (__n >= 0)
+            {
+                __m_ = __size_ - __n;
+                __o1_ = __n;
+                __o2_ = __n - __size_;
+            }
+            else
+            {
+                __m_ = -__n;
+                __o1_ = __n + __size_;
+                __o2_ = __n;
+            }
+        }
+public:
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const
+        {
+            if (__i < __m_)
+                return __expr_[__i + __o1_];
+            return __expr_[__i + __o2_];
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __size_;}
+
+    template <class> friend class __val_expr;
+};
+
+template<class _ValExpr>
+class __val_expr;
+
+template<class _ValExpr>
+struct __is_val_expr : false_type {};
+
+template<class _ValExpr>
+struct __is_val_expr<__val_expr<_ValExpr> > : true_type {};
+
+template<class _Tp>
+struct __is_val_expr<valarray<_Tp> > : true_type {};
+
+template<class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS valarray
+{
+public:
+    typedef _Tp value_type;
+    typedef _Tp result_type;
+
+private:
+    value_type* __begin_;
+    value_type* __end_;
+
+public:
+    // construct/destroy:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray() : __begin_(0), __end_(0) {}
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    explicit valarray(size_t __n);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray(const value_type& __x, size_t __n);
+    valarray(const value_type* __p, size_t __n);
+    valarray(const valarray& __v);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray(valarray&& __v) _NOEXCEPT;
+    valarray(initializer_list<value_type> __il);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    valarray(const slice_array<value_type>& __sa);
+    valarray(const gslice_array<value_type>& __ga);
+    valarray(const mask_array<value_type>& __ma);
+    valarray(const indirect_array<value_type>& __ia);
+    inline _LIBCUDACXX_HIDE_FROM_ABI_AFTER_V1
+    ~valarray();
+
+    // assignment:
+    valarray& operator=(const valarray& __v);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(valarray&& __v) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(initializer_list<value_type>);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(const slice_array<value_type>& __sa);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(const gslice_array<value_type>& __ga);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(const mask_array<value_type>& __ma);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator=(const indirect_array<value_type>& __ia);
+    template <class _ValExpr>
+        _LIBCUDACXX_INLINE_VISIBILITY
+        valarray& operator=(const __val_expr<_ValExpr>& __v);
+
+    // element access:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type& operator[](size_t __i) const {return __begin_[__i];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type&       operator[](size_t __i)       {return __begin_[__i];}
+
+    // subset operations:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__slice_expr<const valarray&> >    operator[](slice __s) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    slice_array<value_type>                       operator[](slice __s);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__indirect_expr<const valarray&> > operator[](const gslice& __gs) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice_array<value_type>   operator[](const gslice& __gs);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__indirect_expr<const valarray&> > operator[](gslice&& __gs) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice_array<value_type>                      operator[](gslice&& __gs);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__mask_expr<const valarray&> >     operator[](const valarray<bool>& __vb) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mask_array<value_type>                        operator[](const valarray<bool>& __vb);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__mask_expr<const valarray&> >     operator[](valarray<bool>&& __vb) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mask_array<value_type>                        operator[](valarray<bool>&& __vb);
+#endif  // _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__indirect_expr<const valarray&> > operator[](const valarray<size_t>& __vs) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    indirect_array<value_type>                    operator[](const valarray<size_t>& __vs);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__indirect_expr<const valarray&> > operator[](valarray<size_t>&& __vs) const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    indirect_array<value_type>                    operator[](valarray<size_t>&& __vs);
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    // unary operators:
+    valarray       operator+() const;
+    valarray       operator-() const;
+    valarray       operator~() const;
+    valarray<bool> operator!() const;
+
+    // computed assignment:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator*= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator/= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator%= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator+= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator-= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator^= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator&= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator|= (const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator<<=(const value_type& __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray& operator>>=(const value_type& __x);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator*= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator/= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator%= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator+= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator-= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator^= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator|= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator&= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator<<= (const _Expr& __v);
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        valarray&
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator>>= (const _Expr& __v);
+
+    // member functions:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void swap(valarray& __v) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return static_cast<size_t>(__end_ - __begin_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type sum() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type min() const;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type max() const;
+
+    valarray shift (int __i) const;
+    valarray cshift(int __i) const;
+    valarray apply(value_type __f(value_type)) const;
+    valarray apply(value_type __f(const value_type&)) const;
+    void     resize(size_t __n, value_type __x = value_type());
+
+private:
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS valarray;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS slice_array;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS gslice_array;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS mask_array;
+    template <class> friend class __mask_expr;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS indirect_array;
+    template <class> friend class __indirect_expr;
+    template <class> friend class __val_expr;
+
+    template <class _Up>
+    friend
+    _Up*
+    begin(valarray<_Up>& __v);
+
+    template <class _Up>
+    friend
+    const _Up*
+    begin(const valarray<_Up>& __v);
+
+    template <class _Up>
+    friend
+    _Up*
+    end(valarray<_Up>& __v);
+
+    template <class _Up>
+    friend
+    const _Up*
+    end(const valarray<_Up>& __v);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __clear(size_t __capacity);
+    valarray& __assign_range(const value_type* __f, const value_type* __l);
+};
+
+_LIBCUDACXX_EXTERN_TEMPLATE(_LIBCUDACXX_FUNC_VIS void valarray<size_t>::resize(size_t, size_t))
+
+template <class _Op, class _Tp>
+struct _UnaryOp<_Op, valarray<_Tp> >
+{
+    typedef typename _Op::result_type result_type;
+    typedef _Tp value_type;
+
+    _Op __op_;
+    const valarray<_Tp>& __a0_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _UnaryOp(const _Op& __op, const valarray<_Tp>& __a0) : __op_(__op), __a0_(__a0) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const {return __op_(__a0_[__i]);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __a0_.size();}
+};
+
+template <class _Op, class _Tp, class _A1>
+struct _BinaryOp<_Op, valarray<_Tp>, _A1>
+{
+    typedef typename _Op::result_type result_type;
+    typedef _Tp value_type;
+
+    _Op __op_;
+    const valarray<_Tp>& __a0_;
+    _A1 __a1_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _BinaryOp(const _Op& __op, const valarray<_Tp>& __a0, const _A1& __a1)
+        : __op_(__op), __a0_(__a0), __a1_(__a1) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type operator[](size_t __i) const {return __op_(__a0_[__i], __a1_[__i]);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __a0_.size();}
+};
+
+template <class _Op, class _A0, class _Tp>
+struct _BinaryOp<_Op, _A0, valarray<_Tp> >
+{
+    typedef typename _Op::result_type result_type;
+    typedef _Tp value_type;
+
+    _Op __op_;
+    _A0 __a0_;
+    const valarray<_Tp>& __a1_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _BinaryOp(const _Op& __op, const _A0& __a0, const valarray<_Tp>& __a1)
+        : __op_(__op), __a0_(__a0), __a1_(__a1) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type operator[](size_t __i) const {return __op_(__a0_[__i], __a1_[__i]);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __a0_.size();}
+};
+
+template <class _Op, class _Tp>
+struct _BinaryOp<_Op, valarray<_Tp>, valarray<_Tp> >
+{
+    typedef typename _Op::result_type result_type;
+    typedef _Tp value_type;
+
+    _Op __op_;
+    const valarray<_Tp>& __a0_;
+    const valarray<_Tp>& __a1_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    _BinaryOp(const _Op& __op, const valarray<_Tp>& __a0, const valarray<_Tp>& __a1)
+        : __op_(__op), __a0_(__a0), __a1_(__a1) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type operator[](size_t __i) const {return __op_(__a0_[__i], __a1_[__i]);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __a0_.size();}
+};
+
+// slice_array
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS slice_array
+{
+public:
+    typedef _Tp value_type;
+
+private:
+    value_type* __vp_;
+    size_t __size_;
+    size_t __stride_;
+
+public:
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator*=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator/=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator%=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator+=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator-=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator^=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator&=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator|=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator<<=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator>>=(const _Expr& __v) const;
+
+    slice_array(slice_array const&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const slice_array& operator=(const slice_array& __sa) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator=(const value_type& __x) const;
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    slice_array(const slice& __sl, const valarray<value_type>& __v)
+        : __vp_(const_cast<value_type*>(__v.__begin_ + __sl.start())),
+          __size_(__sl.size()),
+          __stride_(__sl.stride())
+        {}
+
+    template <class> friend class valarray;
+    template <class> friend class sliceExpr;
+};
+
+template <class _Tp>
+inline
+const slice_array<_Tp>&
+slice_array<_Tp>::operator=(const slice_array& __sa) const
+{
+    value_type* __t = __vp_;
+    const value_type* __s = __sa.__vp_;
+    for (size_t __n = __size_; __n; --__n, __t += __stride_, __s += __sa.__stride_)
+        *__t = *__s;
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t = __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator*=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t *= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator/=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t /= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator%=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t %= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator+=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t += __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator-=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t -= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator^=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t ^= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator&=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t &= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator|=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t |= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator<<=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t <<= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+slice_array<_Tp>::operator>>=(const _Expr& __v) const
+{
+    value_type* __t = __vp_;
+    for (size_t __i = 0; __i < __size_; ++__i, __t += __stride_)
+        *__t >>= __v[__i];
+}
+
+template <class _Tp>
+inline
+void
+slice_array<_Tp>::operator=(const value_type& __x) const
+{
+    value_type* __t = __vp_;
+    for (size_t __n = __size_; __n; --__n, __t += __stride_)
+        *__t = __x;
+}
+
+// gslice
+
+class _LIBCUDACXX_TYPE_VIS gslice
+{
+    valarray<size_t> __size_;
+    valarray<size_t> __stride_;
+    valarray<size_t> __1d_;
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice() {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice(size_t __start, const valarray<size_t>& __size,
+                           const valarray<size_t>& __stride)
+        : __size_(__size),
+          __stride_(__stride)
+        {__init(__start);}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice(size_t __start, const valarray<size_t>&  __size,
+                                 valarray<size_t>&& __stride)
+        : __size_(__size),
+          __stride_(move(__stride))
+        {__init(__start);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice(size_t __start,       valarray<size_t>&& __size,
+                           const valarray<size_t>&  __stride)
+        : __size_(move(__size)),
+          __stride_(__stride)
+        {__init(__start);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    gslice(size_t __start,       valarray<size_t>&& __size,
+                                 valarray<size_t>&& __stride)
+        : __size_(move(__size)),
+          __stride_(move(__stride))
+        {__init(__start);}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t           start()  const {return __1d_.size() ? __1d_[0] : 0;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray<size_t> size()   const {return __size_;}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    valarray<size_t> stride() const {return __stride_;}
+
+private:
+    void __init(size_t __start);
+
+    template <class> friend class gslice_array;
+    template <class> friend class valarray;
+    template <class> friend class __val_expr;
+};
+
+// gslice_array
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS gslice_array
+{
+public:
+    typedef _Tp value_type;
+
+private:
+    value_type*      __vp_;
+    valarray<size_t> __1d_;
+
+public:
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator*=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator/=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator%=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator+=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator-=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator^=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator&=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator|=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator<<=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator>>=(const _Expr& __v) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const gslice_array& operator=(const gslice_array& __ga) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator=(const value_type& __x) const;
+
+    gslice_array(const gslice_array&) = default;
+
+private:
+    gslice_array(const gslice& __gs, const valarray<value_type>& __v)
+        : __vp_(const_cast<value_type*>(__v.__begin_)),
+          __1d_(__gs.__1d_)
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    gslice_array(gslice&& __gs, const valarray<value_type>& __v)
+        : __vp_(const_cast<value_type*>(__v.__begin_)),
+          __1d_(move(__gs.__1d_))
+        {}
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template <class> friend class valarray;
+};
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] = __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator*=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] *= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator/=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] /= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator%=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] %= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator+=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] += __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator-=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] -= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator^=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] ^= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator&=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] &= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator|=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] |= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator<<=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] <<= __v[__j];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+gslice_array<_Tp>::operator>>=(const _Expr& __v) const
+{
+    typedef const size_t* _Ip;
+    size_t __j = 0;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i, ++__j)
+        __vp_[*__i] >>= __v[__j];
+}
+
+template <class _Tp>
+inline
+const gslice_array<_Tp>&
+gslice_array<_Tp>::operator=(const gslice_array& __ga) const
+{
+    typedef const size_t* _Ip;
+    const value_type* __s = __ga.__vp_;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_, __j = __ga.__1d_.__begin_;
+            __i != __e; ++__i, ++__j)
+        __vp_[*__i] = __s[*__j];
+    return *this;
+}
+
+template <class _Tp>
+inline
+void
+gslice_array<_Tp>::operator=(const value_type& __x) const
+{
+    typedef const size_t* _Ip;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i)
+        __vp_[*__i] = __x;
+}
+
+// mask_array
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS mask_array
+{
+public:
+    typedef _Tp value_type;
+
+private:
+    value_type*      __vp_;
+    valarray<size_t> __1d_;
+
+public:
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator*=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator/=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator%=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator+=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator-=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator^=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator&=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator|=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator<<=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator>>=(const _Expr& __v) const;
+
+    mask_array(const mask_array&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const mask_array& operator=(const mask_array& __ma) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator=(const value_type& __x) const;
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    mask_array(const valarray<bool>& __vb, const valarray<value_type>& __v)
+        : __vp_(const_cast<value_type*>(__v.__begin_)),
+          __1d_(static_cast<size_t>(count(__vb.__begin_, __vb.__end_, true)))
+          {
+              size_t __j = 0;
+              for (size_t __i = 0; __i < __vb.size(); ++__i)
+                  if (__vb[__i])
+                      __1d_[__j++] = __i;
+          }
+
+    template <class> friend class valarray;
+};
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] = __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator*=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] *= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator/=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] /= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator%=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] %= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator+=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] += __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator-=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] -= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator^=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] ^= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator&=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] &= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator|=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] |= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator<<=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] <<= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+mask_array<_Tp>::operator>>=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] >>= __v[__i];
+}
+
+template <class _Tp>
+inline
+const mask_array<_Tp>&
+mask_array<_Tp>::operator=(const mask_array& __ma) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] = __ma.__vp_[__1d_[__i]];
+    return *this;
+}
+
+template <class _Tp>
+inline
+void
+mask_array<_Tp>::operator=(const value_type& __x) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] = __x;
+}
+
+template <class _ValExpr>
+class __mask_expr
+{
+    typedef typename remove_reference<_ValExpr>::type  _RmExpr;
+public:
+    typedef typename _RmExpr::value_type value_type;
+    typedef value_type result_type;
+
+private:
+    _ValExpr __expr_;
+    valarray<size_t> __1d_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __mask_expr(const valarray<bool>& __vb, const _RmExpr& __e)
+        : __expr_(__e),
+          __1d_(static_cast<size_t>(count(__vb.__begin_, __vb.__end_, true)))
+          {
+              size_t __j = 0;
+              for (size_t __i = 0; __i < __vb.size(); ++__i)
+                  if (__vb[__i])
+                      __1d_[__j++] = __i;
+          }
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const
+        {return __expr_[__1d_[__i]];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __1d_.size();}
+
+    template <class> friend class __val_expr;
+    template <class> friend class valarray;
+};
+
+// indirect_array
+
+template <class _Tp>
+class _LIBCUDACXX_TEMPLATE_VIS indirect_array
+{
+public:
+    typedef _Tp value_type;
+
+private:
+    value_type*      __vp_;
+    valarray<size_t> __1d_;
+
+public:
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator*=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator/=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator%=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator+=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator-=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator^=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator&=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator|=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator<<=(const _Expr& __v) const;
+
+    template <class _Expr>
+    typename enable_if
+    <
+        __is_val_expr<_Expr>::value,
+        void
+    >::type
+    _LIBCUDACXX_INLINE_VISIBILITY
+    operator>>=(const _Expr& __v) const;
+
+    indirect_array(const indirect_array&) = default;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const indirect_array& operator=(const indirect_array& __ia) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void operator=(const value_type& __x) const;
+
+private:
+     _LIBCUDACXX_INLINE_VISIBILITY
+   indirect_array(const valarray<size_t>& __ia, const valarray<value_type>& __v)
+        : __vp_(const_cast<value_type*>(__v.__begin_)),
+          __1d_(__ia)
+        {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    indirect_array(valarray<size_t>&& __ia, const valarray<value_type>& __v)
+        : __vp_(const_cast<value_type*>(__v.__begin_)),
+          __1d_(move(__ia))
+        {}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+    template <class> friend class valarray;
+};
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] = __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator*=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] *= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator/=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] /= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator%=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] %= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator+=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] += __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator-=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] -= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator^=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] ^= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator&=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] &= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator|=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] |= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator<<=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] <<= __v[__i];
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    void
+>::type
+indirect_array<_Tp>::operator>>=(const _Expr& __v) const
+{
+    size_t __n = __1d_.size();
+    for (size_t __i = 0; __i < __n; ++__i)
+        __vp_[__1d_[__i]] >>= __v[__i];
+}
+
+template <class _Tp>
+inline
+const indirect_array<_Tp>&
+indirect_array<_Tp>::operator=(const indirect_array& __ia) const
+{
+    typedef const size_t* _Ip;
+    const value_type* __s = __ia.__vp_;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_, __j = __ia.__1d_.__begin_;
+            __i != __e; ++__i, ++__j)
+        __vp_[*__i] = __s[*__j];
+    return *this;
+}
+
+template <class _Tp>
+inline
+void
+indirect_array<_Tp>::operator=(const value_type& __x) const
+{
+    typedef const size_t* _Ip;
+    for (_Ip __i = __1d_.__begin_, __e = __1d_.__end_; __i != __e; ++__i)
+        __vp_[*__i] = __x;
+}
+
+template <class _ValExpr>
+class __indirect_expr
+{
+    typedef typename remove_reference<_ValExpr>::type  _RmExpr;
+public:
+    typedef typename _RmExpr::value_type value_type;
+    typedef value_type result_type;
+
+private:
+    _ValExpr __expr_;
+    valarray<size_t> __1d_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __indirect_expr(const valarray<size_t>& __ia, const _RmExpr& __e)
+        : __expr_(__e),
+          __1d_(__ia)
+          {}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __indirect_expr(valarray<size_t>&& __ia, const _RmExpr& __e)
+        : __expr_(__e),
+          __1d_(move(__ia))
+          {}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const
+        {return __expr_[__1d_[__i]];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __1d_.size();}
+
+    template <class> friend class __val_expr;
+    template <class> friend class _LIBCUDACXX_TEMPLATE_VIS valarray;
+};
+
+template<class _ValExpr>
+class __val_expr
+{
+    typedef typename remove_reference<_ValExpr>::type  _RmExpr;
+
+    _ValExpr __expr_;
+public:
+    typedef typename _RmExpr::value_type value_type;
+    typedef typename _RmExpr::result_type result_type;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    explicit __val_expr(const _RmExpr& __e) : __expr_(__e) {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type operator[](size_t __i) const
+        {return __expr_[__i];}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__slice_expr<_ValExpr> > operator[](slice __s) const
+    {
+        typedef __slice_expr<_ValExpr> _NewExpr;
+        return __val_expr< _NewExpr >(_NewExpr(__s, __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__indirect_expr<_ValExpr> > operator[](const gslice& __gs) const
+    {
+        typedef __indirect_expr<_ValExpr> _NewExpr;
+        return __val_expr<_NewExpr >(_NewExpr(__gs.__1d_, __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__mask_expr<_ValExpr> > operator[](const valarray<bool>& __vb) const
+    {
+        typedef __mask_expr<_ValExpr> _NewExpr;
+        return __val_expr< _NewExpr >( _NewExpr(__vb, __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__indirect_expr<_ValExpr> > operator[](const valarray<size_t>& __vs) const
+    {
+        typedef __indirect_expr<_ValExpr> _NewExpr;
+        return __val_expr< _NewExpr >(_NewExpr(__vs, __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<_UnaryOp<__unary_plus<value_type>, _ValExpr> >
+    operator+() const
+    {
+        typedef _UnaryOp<__unary_plus<value_type>, _ValExpr> _NewExpr;
+        return __val_expr<_NewExpr>(_NewExpr(__unary_plus<value_type>(), __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<_UnaryOp<negate<value_type>, _ValExpr> >
+    operator-() const
+    {
+        typedef _UnaryOp<negate<value_type>, _ValExpr> _NewExpr;
+        return __val_expr<_NewExpr>(_NewExpr(negate<value_type>(), __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<_UnaryOp<__bit_not<value_type>, _ValExpr> >
+    operator~() const
+    {
+        typedef _UnaryOp<__bit_not<value_type>, _ValExpr> _NewExpr;
+        return __val_expr<_NewExpr>(_NewExpr(__bit_not<value_type>(), __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<_UnaryOp<logical_not<value_type>, _ValExpr> >
+    operator!() const
+    {
+        typedef _UnaryOp<logical_not<value_type>, _ValExpr> _NewExpr;
+        return __val_expr<_NewExpr>(_NewExpr(logical_not<value_type>(), __expr_));
+    }
+
+    operator valarray<result_type>() const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t size() const {return __expr_.size();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type sum() const
+    {
+        size_t __n = __expr_.size();
+        result_type __r = __n ? __expr_[0] : result_type();
+        for (size_t __i = 1; __i < __n; ++__i)
+            __r += __expr_[__i];
+        return __r;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type min() const
+    {
+        size_t __n = size();
+        result_type __r = __n ? (*this)[0] : result_type();
+        for (size_t __i = 1; __i < __n; ++__i)
+        {
+            result_type __x = __expr_[__i];
+            if (__x < __r)
+                __r = __x;
+        }
+        return __r;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    result_type max() const
+    {
+        size_t __n = size();
+        result_type __r = __n ? (*this)[0] : result_type();
+        for (size_t __i = 1; __i < __n; ++__i)
+        {
+            result_type __x = __expr_[__i];
+            if (__r < __x)
+                __r = __x;
+        }
+        return __r;
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__shift_expr<_ValExpr> > shift (int __i) const
+        {return __val_expr<__shift_expr<_ValExpr> >(__shift_expr<_ValExpr>(__i, __expr_));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<__cshift_expr<_ValExpr> > cshift(int __i) const
+        {return __val_expr<__cshift_expr<_ValExpr> >(__cshift_expr<_ValExpr>(__i, __expr_));}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<_UnaryOp<__apply_expr<value_type, value_type(*)(value_type)>, _ValExpr> >
+    apply(value_type __f(value_type)) const
+    {
+        typedef __apply_expr<value_type, value_type(*)(value_type)> _Op;
+        typedef _UnaryOp<_Op, _ValExpr> _NewExpr;
+        return __val_expr<_NewExpr>(_NewExpr(_Op(__f), __expr_));
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __val_expr<_UnaryOp<__apply_expr<value_type, value_type(*)(const value_type&)>, _ValExpr> >
+    apply(value_type __f(const value_type&)) const
+    {
+        typedef __apply_expr<value_type, value_type(*)(const value_type&)> _Op;
+        typedef _UnaryOp<_Op, _ValExpr> _NewExpr;
+        return __val_expr<_NewExpr>(_NewExpr(_Op(__f), __expr_));
+    }
+};
+
+template<class _ValExpr>
+__val_expr<_ValExpr>::operator valarray<__val_expr::result_type>() const
+{
+    valarray<result_type> __r;
+    size_t __n = __expr_.size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<result_type*>(
+                    _CUDA_VSTD::__libcpp_allocate(__n * sizeof(result_type), _LIBCUDACXX_ALIGNOF(result_type)));
+        for (size_t __i = 0; __i != __n; ++__r.__end_, ++__i)
+            ::new (__r.__end_) result_type(__expr_[__i]);
+    }
+    return __r;
+}
+
+// valarray
+
+template <class _Tp>
+inline
+valarray<_Tp>::valarray(size_t __n)
+    : __begin_(0),
+      __end_(0)
+{
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+            _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (size_t __n_left = __n; __n_left; --__n_left, ++__end_)
+                ::new (__end_) value_type();
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>::valarray(const value_type& __x, size_t __n)
+    : __begin_(0),
+      __end_(0)
+{
+    resize(__n, __x);
+}
+
+template <class _Tp>
+valarray<_Tp>::valarray(const value_type* __p, size_t __n)
+    : __begin_(0),
+      __end_(0)
+{
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+            _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (size_t __n_left = __n; __n_left; ++__end_, ++__p, --__n_left)
+                ::new (__end_) value_type(*__p);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp>
+valarray<_Tp>::valarray(const valarray& __v)
+    : __begin_(0),
+      __end_(0)
+{
+    if (__v.size())
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+            _CUDA_VSTD::__libcpp_allocate(__v.size() * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (value_type* __p = __v.__begin_; __p != __v.__end_; ++__end_, ++__p)
+                ::new (__end_) value_type(*__p);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__v.size());
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+valarray<_Tp>::valarray(valarray&& __v) _NOEXCEPT
+    : __begin_(__v.__begin_),
+      __end_(__v.__end_)
+{
+    __v.__begin_ = __v.__end_ = nullptr;
+}
+
+template <class _Tp>
+valarray<_Tp>::valarray(initializer_list<value_type> __il)
+    : __begin_(0),
+      __end_(0)
+{
+    const size_t __n = __il.size();
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+_CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            size_t __n_left = __n;
+            for (const value_type* __p = __il.begin(); __n_left; ++__end_, ++__p, --__n_left)
+                ::new (__end_) value_type(*__p);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+valarray<_Tp>::valarray(const slice_array<value_type>& __sa)
+    : __begin_(0),
+      __end_(0)
+{
+    const size_t __n = __sa.__size_;
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+          _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            size_t __n_left = __n;
+            for (const value_type* __p = __sa.__vp_; __n_left; ++__end_, __p += __sa.__stride_, --__n_left)
+                ::new (__end_) value_type(*__p);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp>
+valarray<_Tp>::valarray(const gslice_array<value_type>& __ga)
+    : __begin_(0),
+      __end_(0)
+{
+    const size_t __n = __ga.__1d_.size();
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            typedef const size_t* _Ip;
+            const value_type* __s = __ga.__vp_;
+            for (_Ip __i = __ga.__1d_.__begin_, __e = __ga.__1d_.__end_;
+                    __i != __e; ++__i, ++__end_)
+                ::new (__end_) value_type(__s[*__i]);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp>
+valarray<_Tp>::valarray(const mask_array<value_type>& __ma)
+    : __begin_(0),
+      __end_(0)
+{
+    const size_t __n = __ma.__1d_.size();
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            typedef const size_t* _Ip;
+            const value_type* __s = __ma.__vp_;
+            for (_Ip __i = __ma.__1d_.__begin_, __e = __ma.__1d_.__end_;
+                    __i != __e; ++__i, ++__end_)
+                ::new (__end_) value_type(__s[*__i]);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp>
+valarray<_Tp>::valarray(const indirect_array<value_type>& __ia)
+    : __begin_(0),
+      __end_(0)
+{
+    const size_t __n = __ia.__1d_.size();
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            typedef const size_t* _Ip;
+            const value_type* __s = __ia.__vp_;
+            for (_Ip __i = __ia.__1d_.__begin_, __e = __ia.__1d_.__end_;
+                    __i != __e; ++__i, ++__end_)
+                ::new (__end_) value_type(__s[*__i]);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>::~valarray()
+{
+    __clear(size());
+}
+
+template <class _Tp>
+valarray<_Tp>&
+valarray<_Tp>::__assign_range(const value_type* __f, const value_type* __l)
+{
+    size_t __n = __l - __f;
+    if (size() != __n)
+    {
+        __clear(size());
+        __begin_ = static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        __end_ = __begin_ + __n;
+        _CUDA_VSTD::uninitialized_copy(__f, __l, __begin_);
+    } else {
+        _CUDA_VSTD::copy(__f, __l, __begin_);
+    }
+    return *this;
+}
+
+template <class _Tp>
+valarray<_Tp>&
+valarray<_Tp>::operator=(const valarray& __v)
+{
+    if (this != &__v)
+        return __assign_range(__v.__begin_, __v.__end_);
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(valarray&& __v) _NOEXCEPT
+{
+    __clear(size());
+    __begin_ = __v.__begin_;
+    __end_ = __v.__end_;
+    __v.__begin_ = nullptr;
+    __v.__end_ = nullptr;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(initializer_list<value_type> __il)
+{
+    return __assign_range(__il.begin(), __il.end());
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(const value_type& __x)
+{
+    _CUDA_VSTD::fill(__begin_, __end_, __x);
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(const slice_array<value_type>& __sa)
+{
+    value_type* __t = __begin_;
+    const value_type* __s = __sa.__vp_;
+    for (size_t __n = __sa.__size_; __n; --__n, __s += __sa.__stride_, ++__t)
+        *__t = *__s;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(const gslice_array<value_type>& __ga)
+{
+    typedef const size_t* _Ip;
+    value_type* __t = __begin_;
+    const value_type* __s = __ga.__vp_;
+    for (_Ip __i = __ga.__1d_.__begin_, __e = __ga.__1d_.__end_;
+                    __i != __e; ++__i, ++__t)
+        *__t = __s[*__i];
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(const mask_array<value_type>& __ma)
+{
+    typedef const size_t* _Ip;
+    value_type* __t = __begin_;
+    const value_type* __s = __ma.__vp_;
+    for (_Ip __i = __ma.__1d_.__begin_, __e = __ma.__1d_.__end_;
+                    __i != __e; ++__i, ++__t)
+        *__t = __s[*__i];
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(const indirect_array<value_type>& __ia)
+{
+    typedef const size_t* _Ip;
+    value_type* __t = __begin_;
+    const value_type* __s = __ia.__vp_;
+    for (_Ip __i = __ia.__1d_.__begin_, __e = __ia.__1d_.__end_;
+                    __i != __e; ++__i, ++__t)
+        *__t = __s[*__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _ValExpr>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator=(const __val_expr<_ValExpr>& __v)
+{
+    size_t __n = __v.size();
+    if (size() != __n)
+        resize(__n);
+    value_type* __t = __begin_;
+    for (size_t __i = 0; __i != __n; ++__t, ++__i)
+        *__t = result_type(__v[__i]);
+    return *this;
+}
+
+template <class _Tp>
+inline
+__val_expr<__slice_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](slice __s) const
+{
+    return __val_expr<__slice_expr<const valarray&> >(__slice_expr<const valarray&>(__s, *this));
+}
+
+template <class _Tp>
+inline
+slice_array<_Tp>
+valarray<_Tp>::operator[](slice __s)
+{
+    return slice_array<value_type>(__s, *this);
+}
+
+template <class _Tp>
+inline
+__val_expr<__indirect_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](const gslice& __gs) const
+{
+    return __val_expr<__indirect_expr<const valarray&> >(__indirect_expr<const valarray&>(__gs.__1d_, *this));
+}
+
+template <class _Tp>
+inline
+gslice_array<_Tp>
+valarray<_Tp>::operator[](const gslice& __gs)
+{
+    return gslice_array<value_type>(__gs, *this);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+__val_expr<__indirect_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](gslice&& __gs) const
+{
+    return __val_expr<__indirect_expr<const valarray&> >(__indirect_expr<const valarray&>(move(__gs.__1d_), *this));
+}
+
+template <class _Tp>
+inline
+gslice_array<_Tp>
+valarray<_Tp>::operator[](gslice&& __gs)
+{
+    return gslice_array<value_type>(move(__gs), *this);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+__val_expr<__mask_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](const valarray<bool>& __vb) const
+{
+    return __val_expr<__mask_expr<const valarray&> >(__mask_expr<const valarray&>(__vb, *this));
+}
+
+template <class _Tp>
+inline
+mask_array<_Tp>
+valarray<_Tp>::operator[](const valarray<bool>& __vb)
+{
+    return mask_array<value_type>(__vb, *this);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+__val_expr<__mask_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](valarray<bool>&& __vb) const
+{
+    return __val_expr<__mask_expr<const valarray&> >(__mask_expr<const valarray&>(move(__vb), *this));
+}
+
+template <class _Tp>
+inline
+mask_array<_Tp>
+valarray<_Tp>::operator[](valarray<bool>&& __vb)
+{
+    return mask_array<value_type>(move(__vb), *this);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+__val_expr<__indirect_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](const valarray<size_t>& __vs) const
+{
+    return __val_expr<__indirect_expr<const valarray&> >(__indirect_expr<const valarray&>(__vs, *this));
+}
+
+template <class _Tp>
+inline
+indirect_array<_Tp>
+valarray<_Tp>::operator[](const valarray<size_t>& __vs)
+{
+    return indirect_array<value_type>(__vs, *this);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+inline
+__val_expr<__indirect_expr<const valarray<_Tp>&> >
+valarray<_Tp>::operator[](valarray<size_t>&& __vs) const
+{
+    return __val_expr<__indirect_expr<const valarray&> >(__indirect_expr<const valarray&>(move(__vs), *this));
+}
+
+template <class _Tp>
+inline
+indirect_array<_Tp>
+valarray<_Tp>::operator[](valarray<size_t>&& __vs)
+{
+    return indirect_array<value_type>(move(__vs), *this);
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::operator+() const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
+            ::new (__r.__end_) value_type(+*__p);
+    }
+    return __r;
+}
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::operator-() const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
+            ::new (__r.__end_) value_type(-*__p);
+    }
+    return __r;
+}
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::operator~() const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
+            ::new (__r.__end_) value_type(~*__p);
+    }
+    return __r;
+}
+
+template <class _Tp>
+valarray<bool>
+valarray<_Tp>::operator!() const
+{
+    valarray<bool> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<bool*>(_CUDA_VSTD::__libcpp_allocate(__n * sizeof(bool), _LIBCUDACXX_ALIGNOF(bool)));
+        for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
+            ::new (__r.__end_) bool(!*__p);
+    }
+    return __r;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator*=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p *= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator/=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p /= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator%=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p %= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator+=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p += __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator-=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p -= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator^=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p ^= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator&=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p &= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator|=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p |= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator<<=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p <<= __x;
+    return *this;
+}
+
+template <class _Tp>
+inline
+valarray<_Tp>&
+valarray<_Tp>::operator>>=(const value_type& __x)
+{
+    for (value_type* __p = __begin_; __p != __end_; ++__p)
+        *__p >>= __x;
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator*=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t *= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator/=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t /= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator%=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t %= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator+=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t += __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator-=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t -= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator^=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t ^= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator|=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t |= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator&=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t &= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator<<=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t <<= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+template <class _Expr>
+inline
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    valarray<_Tp>&
+>::type
+valarray<_Tp>::operator>>=(const _Expr& __v)
+{
+    size_t __i = 0;
+    for (value_type* __t = __begin_; __t != __end_ ; ++__t, ++__i)
+        *__t >>= __v[__i];
+    return *this;
+}
+
+template <class _Tp>
+inline
+void
+valarray<_Tp>::swap(valarray& __v) _NOEXCEPT
+{
+    _CUDA_VSTD::swap(__begin_, __v.__begin_);
+    _CUDA_VSTD::swap(__end_, __v.__end_);
+}
+
+template <class _Tp>
+inline
+_Tp
+valarray<_Tp>::sum() const
+{
+    if (__begin_ == __end_)
+        return value_type();
+    const value_type* __p = __begin_;
+    _Tp __r = *__p;
+    for (++__p; __p != __end_; ++__p)
+        __r += *__p;
+    return __r;
+}
+
+template <class _Tp>
+inline
+_Tp
+valarray<_Tp>::min() const
+{
+    if (__begin_ == __end_)
+        return value_type();
+    return *_CUDA_VSTD::min_element(__begin_, __end_);
+}
+
+template <class _Tp>
+inline
+_Tp
+valarray<_Tp>::max() const
+{
+    if (__begin_ == __end_)
+        return value_type();
+    return *_CUDA_VSTD::max_element(__begin_, __end_);
+}
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::shift(int __i) const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        const value_type* __sb;
+        value_type* __tb;
+        value_type* __te;
+        if (__i >= 0)
+        {
+            __i = _CUDA_VSTD::min(__i, static_cast<int>(__n));
+            __sb = __begin_ + __i;
+            __tb = __r.__begin_;
+            __te = __r.__begin_ + (__n - __i);
+        }
+        else
+        {
+            __i = _CUDA_VSTD::min(-__i, static_cast<int>(__n));
+            __sb = __begin_;
+            __tb = __r.__begin_ + __i;
+            __te = __r.__begin_ + __n;
+        }
+        for (; __r.__end_ != __tb; ++__r.__end_)
+            ::new (__r.__end_) value_type();
+        for (; __r.__end_ != __te; ++__r.__end_, ++__sb)
+            ::new (__r.__end_) value_type(*__sb);
+        for (__te = __r.__begin_ + __n; __r.__end_ != __te; ++__r.__end_)
+            ::new (__r.__end_) value_type();
+    }
+    return __r;
+}
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::cshift(int __i) const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        __i %= static_cast<int>(__n);
+        const value_type* __m = __i >= 0 ? __begin_ + __i : __end_ + __i;
+        for (const value_type* __s = __m; __s != __end_; ++__r.__end_, ++__s)
+            ::new (__r.__end_) value_type(*__s);
+        for (const value_type* __s = __begin_; __s != __m; ++__r.__end_, ++__s)
+            ::new (__r.__end_) value_type(*__s);
+    }
+    return __r;
+}
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::apply(value_type __f(value_type)) const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
+            ::new (__r.__end_) value_type(__f(*__p));
+    }
+    return __r;
+}
+
+template <class _Tp>
+valarray<_Tp>
+valarray<_Tp>::apply(value_type __f(const value_type&)) const
+{
+    valarray<value_type> __r;
+    size_t __n = size();
+    if (__n)
+    {
+        __r.__begin_ =
+            __r.__end_ =
+                static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+        for (const value_type* __p = __begin_; __n; ++__r.__end_, ++__p, --__n)
+            ::new (__r.__end_) value_type(__f(*__p));
+    }
+    return __r;
+}
+
+template <class _Tp>
+inline
+void valarray<_Tp>::__clear(size_t __capacity)
+{
+  if (__begin_ != nullptr)
+  {
+    while (__end_ != __begin_)
+      (--__end_)->~value_type();
+    _CUDA_VSTD::__libcpp_deallocate(__begin_, __capacity * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type));
+    __begin_ = __end_ = nullptr;
+  }
+}
+
+template <class _Tp>
+void
+valarray<_Tp>::resize(size_t __n, value_type __x)
+{
+    __clear(size());
+    if (__n)
+    {
+        __begin_ = __end_ = static_cast<value_type*>(
+           _CUDA_VSTD::__libcpp_allocate(__n * sizeof(value_type), _LIBCUDACXX_ALIGNOF(value_type)));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            for (size_t __n_left = __n; __n_left; --__n_left, ++__end_)
+                ::new (__end_) value_type(__x);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            __clear(__n);
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template<class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(valarray<_Tp>& __x, valarray<_Tp>& __y) _NOEXCEPT
+{
+    __x.swap(__y);
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<multiplies<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator*(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<multiplies<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(multiplies<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<multiplies<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator*(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<multiplies<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(multiplies<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<multiplies<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator*(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<multiplies<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(multiplies<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<divides<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator/(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<divides<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(divides<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<divides<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator/(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<divides<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(divides<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<divides<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator/(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<divides<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(divides<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<modulus<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator%(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<modulus<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(modulus<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<modulus<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator%(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<modulus<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(modulus<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<modulus<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator%(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<modulus<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(modulus<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<plus<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator+(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<plus<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(plus<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<plus<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator+(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<plus<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(plus<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<plus<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator+(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<plus<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(plus<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<minus<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator-(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<minus<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(minus<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<minus<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator-(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<minus<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(minus<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<minus<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator-(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<minus<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(minus<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<bit_xor<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator^(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<bit_xor<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(bit_xor<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<bit_xor<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator^(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<bit_xor<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(bit_xor<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<bit_xor<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator^(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<bit_xor<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(bit_xor<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<bit_and<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator&(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<bit_and<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(bit_and<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<bit_and<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator&(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<bit_and<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(bit_and<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<bit_and<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator&(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<bit_and<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(bit_and<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<bit_or<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator|(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<bit_or<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(bit_or<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<bit_or<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator|(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<bit_or<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(bit_or<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<bit_or<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator|(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<bit_or<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(bit_or<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<__bit_shift_left<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator<<(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<__bit_shift_left<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(__bit_shift_left<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__bit_shift_left<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator<<(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__bit_shift_left<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(__bit_shift_left<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__bit_shift_left<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator<<(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__bit_shift_left<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__bit_shift_left<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<__bit_shift_right<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator>>(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<__bit_shift_right<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(__bit_shift_right<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__bit_shift_right<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator>>(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__bit_shift_right<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(__bit_shift_right<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__bit_shift_right<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator>>(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__bit_shift_right<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__bit_shift_right<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<logical_and<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator&&(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<logical_and<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(logical_and<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<logical_and<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator&&(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<logical_and<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(logical_and<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<logical_and<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator&&(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<logical_and<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(logical_and<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<logical_or<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator||(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<logical_or<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(logical_or<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<logical_or<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator||(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<logical_or<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(logical_or<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<logical_or<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator||(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<logical_or<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(logical_or<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<equal_to<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator==(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<equal_to<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(equal_to<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<equal_to<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator==(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<equal_to<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(equal_to<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<equal_to<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator==(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<equal_to<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(equal_to<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<not_equal_to<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator!=(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<not_equal_to<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(not_equal_to<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<not_equal_to<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator!=(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<not_equal_to<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(not_equal_to<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<not_equal_to<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator!=(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<not_equal_to<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(not_equal_to<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<less<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator<(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<less<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(less<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<less<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator<(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<less<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(less<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<less<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator<(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<less<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(less<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<greater<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator>(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<greater<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(greater<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<greater<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator>(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<greater<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(greater<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<greater<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator>(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<greater<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(greater<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<less_equal<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator<=(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<less_equal<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(less_equal<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<less_equal<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator<=(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<less_equal<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(less_equal<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<less_equal<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator<=(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<less_equal<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(less_equal<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<greater_equal<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+operator>=(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<greater_equal<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(greater_equal<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<greater_equal<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+operator>=(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<greater_equal<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(greater_equal<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<greater_equal<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+operator>=(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<greater_equal<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(greater_equal<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__abs_expr<typename _Expr::value_type>, _Expr> >
+>::type
+abs(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__abs_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__abs_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__acos_expr<typename _Expr::value_type>, _Expr> >
+>::type
+acos(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__acos_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__acos_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__asin_expr<typename _Expr::value_type>, _Expr> >
+>::type
+asin(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__asin_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__asin_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__atan_expr<typename _Expr::value_type>, _Expr> >
+>::type
+atan(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__atan_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__atan_expr<value_type>(), __x));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<__atan2_expr<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+atan2(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<__atan2_expr<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(__atan2_expr<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__atan2_expr<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+atan2(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__atan2_expr<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(__atan2_expr<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__atan2_expr<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+atan2(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__atan2_expr<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__atan2_expr<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__cos_expr<typename _Expr::value_type>, _Expr> >
+>::type
+cos(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__cos_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__cos_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__cosh_expr<typename _Expr::value_type>, _Expr> >
+>::type
+cosh(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__cosh_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__cosh_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__exp_expr<typename _Expr::value_type>, _Expr> >
+>::type
+exp(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__exp_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__exp_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__log_expr<typename _Expr::value_type>, _Expr> >
+>::type
+log(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__log_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__log_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__log10_expr<typename _Expr::value_type>, _Expr> >
+>::type
+log10(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__log10_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__log10_expr<value_type>(), __x));
+}
+
+template<class _Expr1, class _Expr2>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr1>::value && __is_val_expr<_Expr2>::value,
+    __val_expr<_BinaryOp<__pow_expr<typename _Expr1::value_type>, _Expr1, _Expr2> >
+>::type
+pow(const _Expr1& __x, const _Expr2& __y)
+{
+    typedef typename _Expr1::value_type value_type;
+    typedef _BinaryOp<__pow_expr<value_type>, _Expr1, _Expr2> _Op;
+    return __val_expr<_Op>(_Op(__pow_expr<value_type>(), __x, __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__pow_expr<typename _Expr::value_type>,
+               _Expr, __scalar_expr<typename _Expr::value_type> > >
+>::type
+pow(const _Expr& __x, const typename _Expr::value_type& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__pow_expr<value_type>, _Expr, __scalar_expr<value_type> > _Op;
+    return __val_expr<_Op>(_Op(__pow_expr<value_type>(),
+                           __x, __scalar_expr<value_type>(__y, __x.size())));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_BinaryOp<__pow_expr<typename _Expr::value_type>,
+               __scalar_expr<typename _Expr::value_type>, _Expr> >
+>::type
+pow(const typename _Expr::value_type& __x, const _Expr& __y)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _BinaryOp<__pow_expr<value_type>, __scalar_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__pow_expr<value_type>(),
+                           __scalar_expr<value_type>(__x, __y.size()), __y));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__sin_expr<typename _Expr::value_type>, _Expr> >
+>::type
+sin(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__sin_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__sin_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__sinh_expr<typename _Expr::value_type>, _Expr> >
+>::type
+sinh(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__sinh_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__sinh_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__sqrt_expr<typename _Expr::value_type>, _Expr> >
+>::type
+sqrt(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__sqrt_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__sqrt_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__tan_expr<typename _Expr::value_type>, _Expr> >
+>::type
+tan(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__tan_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__tan_expr<value_type>(), __x));
+}
+
+template<class _Expr>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename enable_if
+<
+    __is_val_expr<_Expr>::value,
+    __val_expr<_UnaryOp<__tanh_expr<typename _Expr::value_type>, _Expr> >
+>::type
+tanh(const _Expr& __x)
+{
+    typedef typename _Expr::value_type value_type;
+    typedef _UnaryOp<__tanh_expr<value_type>, _Expr> _Op;
+    return __val_expr<_Op>(_Op(__tanh_expr<value_type>(), __x));
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+begin(valarray<_Tp>& __v)
+{
+    return __v.__begin_;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+const _Tp*
+begin(const valarray<_Tp>& __v)
+{
+    return __v.__begin_;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_Tp*
+end(valarray<_Tp>& __v)
+{
+    return __v.__end_;
+}
+
+template <class _Tp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+const _Tp*
+end(const valarray<_Tp>& __v)
+{
+    return __v.__end_;
+}
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_VALARRAY
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/variant b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/variant
new file mode 100644
index 000000000000..217a6804d54f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/variant
@@ -0,0 +1,1668 @@
+// -*- C++ -*-
+//===------------------------------ variant -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_VARIANT
+#define _LIBCUDACXX_VARIANT
+
+/*
+   variant synopsis
+
+namespace std {
+
+  // 20.7.2, class template variant
+  template <class... Types>
+  class variant {
+  public:
+
+    // 20.7.2.1, constructors
+    constexpr variant() noexcept(see below);
+    variant(const variant&);                // constexpr in C++20
+    variant(variant&&) noexcept(see below); // constexpr in C++20
+
+    template <class T> constexpr variant(T&&) noexcept(see below);
+
+    template <class T, class... Args>
+    constexpr explicit variant(in_place_type_t<T>, Args&&...);
+
+    template <class T, class U, class... Args>
+    constexpr explicit variant(
+        in_place_type_t<T>, initializer_list<U>, Args&&...);
+
+    template <size_t I, class... Args>
+    constexpr explicit variant(in_place_index_t<I>, Args&&...);
+
+    template <size_t I, class U, class... Args>
+    constexpr explicit variant(
+        in_place_index_t<I>, initializer_list<U>, Args&&...);
+
+    // 20.7.2.2, destructor
+    ~variant();
+
+    // 20.7.2.3, assignment
+    variant& operator=(const variant&);                // constexpr in C++20
+    variant& operator=(variant&&) noexcept(see below); // constexpr in C++20
+
+    template <class T> variant& operator=(T&&) noexcept(see below);
+
+    // 20.7.2.4, modifiers
+    template <class T, class... Args>
+    T& emplace(Args&&...);
+
+    template <class T, class U, class... Args>
+    T& emplace(initializer_list<U>, Args&&...);
+
+    template <size_t I, class... Args>
+    variant_alternative_t<I, variant>& emplace(Args&&...);
+
+    template <size_t I, class U, class...  Args>
+    variant_alternative_t<I, variant>& emplace(initializer_list<U>, Args&&...);
+
+    // 20.7.2.5, value status
+    constexpr bool valueless_by_exception() const noexcept;
+    constexpr size_t index() const noexcept;
+
+    // 20.7.2.6, swap
+    void swap(variant&) noexcept(see below);
+  };
+
+  // 20.7.3, variant helper classes
+  template <class T> struct variant_size; // undefined
+
+  template <class T>
+  inline constexpr size_t variant_size_v = variant_size<T>::value;
+
+  template <class T> struct variant_size<const T>;
+  template <class T> struct variant_size<volatile T>;
+  template <class T> struct variant_size<const volatile T>;
+
+  template <class... Types>
+  struct variant_size<variant<Types...>>;
+
+  template <size_t I, class T> struct variant_alternative; // undefined
+
+  template <size_t I, class T>
+  using variant_alternative_t = typename variant_alternative<I, T>::type;
+
+  template <size_t I, class T> struct variant_alternative<I, const T>;
+  template <size_t I, class T> struct variant_alternative<I, volatile T>;
+  template <size_t I, class T> struct variant_alternative<I, const volatile T>;
+
+  template <size_t I, class... Types>
+  struct variant_alternative<I, variant<Types...>>;
+
+  inline constexpr size_t variant_npos = -1;
+
+  // 20.7.4, value access
+  template <class T, class... Types>
+  constexpr bool holds_alternative(const variant<Types...>&) noexcept;
+
+  template <size_t I, class... Types>
+  constexpr variant_alternative_t<I, variant<Types...>>&
+  get(variant<Types...>&);
+
+  template <size_t I, class... Types>
+  constexpr variant_alternative_t<I, variant<Types...>>&&
+  get(variant<Types...>&&);
+
+  template <size_t I, class... Types>
+  constexpr variant_alternative_t<I, variant<Types...>> const&
+  get(const variant<Types...>&);
+
+  template <size_t I, class... Types>
+  constexpr variant_alternative_t<I, variant<Types...>> const&&
+  get(const variant<Types...>&&);
+
+  template <class T, class...  Types>
+  constexpr T& get(variant<Types...>&);
+
+  template <class T, class... Types>
+  constexpr T&& get(variant<Types...>&&);
+
+  template <class T, class... Types>
+  constexpr const T& get(const variant<Types...>&);
+
+  template <class T, class... Types>
+  constexpr const T&& get(const variant<Types...>&&);
+
+  template <size_t I, class... Types>
+  constexpr add_pointer_t<variant_alternative_t<I, variant<Types...>>>
+  get_if(variant<Types...>*) noexcept;
+
+  template <size_t I, class... Types>
+  constexpr add_pointer_t<const variant_alternative_t<I, variant<Types...>>>
+  get_if(const variant<Types...>*) noexcept;
+
+  template <class T, class... Types>
+  constexpr add_pointer_t<T>
+  get_if(variant<Types...>*) noexcept;
+
+  template <class T, class... Types>
+  constexpr add_pointer_t<const T>
+  get_if(const variant<Types...>*) noexcept;
+
+  // 20.7.5, relational operators
+  template <class... Types>
+  constexpr bool operator==(const variant<Types...>&, const variant<Types...>&);
+
+  template <class... Types>
+  constexpr bool operator!=(const variant<Types...>&, const variant<Types...>&);
+
+  template <class... Types>
+  constexpr bool operator<(const variant<Types...>&, const variant<Types...>&);
+
+  template <class... Types>
+  constexpr bool operator>(const variant<Types...>&, const variant<Types...>&);
+
+  template <class... Types>
+  constexpr bool operator<=(const variant<Types...>&, const variant<Types...>&);
+
+  template <class... Types>
+  constexpr bool operator>=(const variant<Types...>&, const variant<Types...>&);
+
+  // 20.7.6, visitation
+  template <class Visitor, class... Variants>
+  constexpr see below visit(Visitor&&, Variants&&...);
+
+  // 20.7.7, class monostate
+  struct monostate;
+
+  // 20.7.8, monostate relational operators
+  constexpr bool operator<(monostate, monostate) noexcept;
+  constexpr bool operator>(monostate, monostate) noexcept;
+  constexpr bool operator<=(monostate, monostate) noexcept;
+  constexpr bool operator>=(monostate, monostate) noexcept;
+  constexpr bool operator==(monostate, monostate) noexcept;
+  constexpr bool operator!=(monostate, monostate) noexcept;
+
+  // 20.7.9, specialized algorithms
+  template <class... Types>
+  void swap(variant<Types...>&, variant<Types...>&) noexcept(see below);
+
+  // 20.7.10, class bad_variant_access
+  class bad_variant_access;
+
+  // 20.7.11, hash support
+  template <class T> struct hash;
+  template <class... Types> struct hash<variant<Types...>>;
+  template <> struct hash<monostate>;
+
+} // namespace std
+
+*/
+
+#include <__config>
+#include <__tuple>
+#include <array>
+#include <exception>
+#include <functional>
+#include <initializer_list>
+#include <new>
+#include <tuple>
+#include <type_traits>
+#include <utility>
+#include <limits>
+#include <version>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+namespace std { // explicitly not using versioning namespace
+
+class _LIBCUDACXX_EXCEPTION_ABI _LIBCUDACXX_AVAILABILITY_BAD_VARIANT_ACCESS bad_variant_access : public exception {
+public:
+  virtual const char* what() const _NOEXCEPT;
+};
+
+} // namespace std
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+#if _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_NORETURN
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+void __throw_bad_variant_access() {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        throw bad_variant_access();
+#else
+        _CUDA_VSTD::abort();
+#endif
+}
+
+template <class... _Types>
+class _LIBCUDACXX_TEMPLATE_VIS variant;
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_size;
+
+template <class _Tp>
+_LIBCUDACXX_INLINE_VAR constexpr size_t variant_size_v = variant_size<_Tp>::value;
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_size<const _Tp> : variant_size<_Tp> {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_size<volatile _Tp> : variant_size<_Tp> {};
+
+template <class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_size<const volatile _Tp>
+    : variant_size<_Tp> {};
+
+template <class... _Types>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_size<variant<_Types...>>
+    : integral_constant<size_t, sizeof...(_Types)> {};
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_alternative;
+
+template <size_t _Ip, class _Tp>
+using variant_alternative_t = typename variant_alternative<_Ip, _Tp>::type;
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_alternative<_Ip, const _Tp>
+    : add_const<variant_alternative_t<_Ip, _Tp>> {};
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_alternative<_Ip, volatile _Tp>
+    : add_volatile<variant_alternative_t<_Ip, _Tp>> {};
+
+template <size_t _Ip, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_alternative<_Ip, const volatile _Tp>
+    : add_cv<variant_alternative_t<_Ip, _Tp>> {};
+
+template <size_t _Ip, class... _Types>
+struct _LIBCUDACXX_TEMPLATE_VIS variant_alternative<_Ip, variant<_Types...>> {
+  static_assert(_Ip < sizeof...(_Types), "Index out of bounds in std::variant_alternative<>");
+  using type = __type_pack_element<_Ip, _Types...>;
+};
+
+_LIBCUDACXX_INLINE_VAR constexpr size_t variant_npos = static_cast<size_t>(-1);
+
+constexpr int __choose_index_type(unsigned int __num_elem) {
+  if (__num_elem < std::numeric_limits<unsigned char>::max())
+    return 0;
+  if (__num_elem < std::numeric_limits<unsigned short>::max())
+    return 1;
+  return 2;
+}
+
+template <size_t _NumAlts>
+using __variant_index_t =
+#ifndef _LIBCUDACXX_ABI_VARIANT_INDEX_TYPE_OPTIMIZATION
+  unsigned int;
+#else
+  std::tuple_element_t<
+      __choose_index_type(_NumAlts),
+      std::tuple<unsigned char, unsigned short, unsigned int>
+  >;
+#endif
+
+template <class _IndexType>
+constexpr _IndexType __variant_npos = static_cast<_IndexType>(-1);
+
+namespace __find_detail {
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr size_t __find_index() {
+  constexpr bool __matches[] = {is_same_v<_Tp, _Types>...};
+  size_t __result = __not_found;
+  for (size_t __i = 0; __i < sizeof...(_Types); ++__i) {
+    if (__matches[__i]) {
+      if (__result != __not_found) {
+        return __ambiguous;
+      }
+      __result = __i;
+    }
+  }
+  return __result;
+}
+
+template <size_t _Index>
+struct __find_unambiguous_index_sfinae_impl
+    : integral_constant<size_t, _Index> {};
+
+template <>
+struct __find_unambiguous_index_sfinae_impl<__not_found> {};
+
+template <>
+struct __find_unambiguous_index_sfinae_impl<__ambiguous> {};
+
+template <class _Tp, class... _Types>
+struct __find_unambiguous_index_sfinae
+    : __find_unambiguous_index_sfinae_impl<__find_index<_Tp, _Types...>()> {};
+
+} // namespace __find_detail
+
+namespace __variant_detail {
+
+struct __valueless_t {};
+
+enum class _Trait { _TriviallyAvailable, _Available, _Unavailable };
+
+template <typename _Tp,
+          template <typename> class _IsTriviallyAvailable,
+          template <typename> class _IsAvailable>
+constexpr _Trait __trait =
+    _IsTriviallyAvailable<_Tp>::value
+        ? _Trait::_TriviallyAvailable
+        : _IsAvailable<_Tp>::value ? _Trait::_Available : _Trait::_Unavailable;
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr _Trait __common_trait(initializer_list<_Trait> __traits) {
+  _Trait __result = _Trait::_TriviallyAvailable;
+  for (_Trait __t : __traits) {
+    if (static_cast<int>(__t) > static_cast<int>(__result)) {
+      __result = __t;
+    }
+  }
+  return __result;
+}
+
+template <typename... _Types>
+struct __traits {
+  static constexpr _Trait __copy_constructible_trait =
+      __common_trait({__trait<_Types,
+                              is_trivially_copy_constructible,
+                              is_copy_constructible>...});
+
+  static constexpr _Trait __move_constructible_trait =
+      __common_trait({__trait<_Types,
+                              is_trivially_move_constructible,
+                              is_move_constructible>...});
+
+  static constexpr _Trait __copy_assignable_trait = __common_trait(
+      {__copy_constructible_trait,
+       __trait<_Types, is_trivially_copy_assignable, is_copy_assignable>...});
+
+  static constexpr _Trait __move_assignable_trait = __common_trait(
+      {__move_constructible_trait,
+       __trait<_Types, is_trivially_move_assignable, is_move_assignable>...});
+
+  static constexpr _Trait __destructible_trait = __common_trait(
+      {__trait<_Types, is_trivially_destructible, is_destructible>...});
+};
+
+namespace __access {
+
+struct __union {
+  template <class _Vp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto&& __get_alt(_Vp&& __v, in_place_index_t<0>) {
+    return _CUDA_VSTD::forward<_Vp>(__v).__head;
+  }
+
+  template <class _Vp, size_t _Ip>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto&& __get_alt(_Vp&& __v, in_place_index_t<_Ip>) {
+    return __get_alt(_CUDA_VSTD::forward<_Vp>(__v).__tail, in_place_index<_Ip - 1>);
+  }
+};
+
+struct __base {
+  template <size_t _Ip, class _Vp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto&& __get_alt(_Vp&& __v) {
+    return __union::__get_alt(_CUDA_VSTD::forward<_Vp>(__v).__data,
+                              in_place_index<_Ip>);
+  }
+};
+
+struct __variant {
+  template <size_t _Ip, class _Vp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto&& __get_alt(_Vp&& __v) {
+    return __base::__get_alt<_Ip>(_CUDA_VSTD::forward<_Vp>(__v).__impl);
+  }
+};
+
+} // namespace __access
+
+namespace __visitation {
+
+struct __base {
+  template <class _Visitor, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr decltype(auto)
+  __visit_alt_at(size_t __index, _Visitor&& __visitor, _Vs&&... __vs) {
+    constexpr auto __fdiagonal =
+        __make_fdiagonal<_Visitor&&,
+                         decltype(_CUDA_VSTD::forward<_Vs>(__vs).__as_base())...>();
+    return __fdiagonal[__index](_CUDA_VSTD::forward<_Visitor>(__visitor),
+                                _CUDA_VSTD::forward<_Vs>(__vs).__as_base()...);
+  }
+
+  template <class _Visitor, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr decltype(auto) __visit_alt(_Visitor&& __visitor,
+                                              _Vs&&... __vs) {
+    constexpr auto __fmatrix =
+        __make_fmatrix<_Visitor&&,
+                       decltype(_CUDA_VSTD::forward<_Vs>(__vs).__as_base())...>();
+    return __at(__fmatrix, __vs.index()...)(
+        _CUDA_VSTD::forward<_Visitor>(__visitor),
+        _CUDA_VSTD::forward<_Vs>(__vs).__as_base()...);
+  }
+
+private:
+  template <class _Tp>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr const _Tp& __at(const _Tp& __elem) { return __elem; }
+
+  template <class _Tp, size_t _Np, typename... _Indices>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto&& __at(const array<_Tp, _Np>& __elems,
+                               size_t __index, _Indices... __indices) {
+    return __at(__elems[__index], __indices...);
+  }
+
+  template <class _Fp, class... _Fs>
+  static constexpr void __std_visit_visitor_return_type_check() {
+    static_assert(
+        __all<is_same_v<_Fp, _Fs>...>::value,
+        "`std::visit` requires the visitor to have a single return type.");
+  }
+
+  template <class... _Fs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_farray(_Fs&&... __fs) {
+    __std_visit_visitor_return_type_check<__uncvref_t<_Fs>...>();
+    using __result = array<common_type_t<__uncvref_t<_Fs>...>, sizeof...(_Fs)>;
+    return __result{{_CUDA_VSTD::forward<_Fs>(__fs)...}};
+  }
+
+  template <std::size_t... _Is>
+  struct __dispatcher {
+    template <class _Fp, class... _Vs>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    static constexpr decltype(auto) __dispatch(_Fp __f, _Vs... __vs) {
+        return __invoke_constexpr(
+            static_cast<_Fp>(__f),
+            __access::__base::__get_alt<_Is>(static_cast<_Vs>(__vs))...);
+    }
+  };
+
+  template <class _Fp, class... _Vs, size_t... _Is>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_dispatch(index_sequence<_Is...>) {
+    return __dispatcher<_Is...>::template __dispatch<_Fp, _Vs...>;
+  }
+
+  template <size_t _Ip, class _Fp, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_fdiagonal_impl() {
+    return __make_dispatch<_Fp, _Vs...>(
+        index_sequence<(__identity<_Vs>{}, _Ip)...>{});
+  }
+
+  template <class _Fp, class... _Vs, size_t... _Is>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_fdiagonal_impl(index_sequence<_Is...>) {
+    return __base::__make_farray(__make_fdiagonal_impl<_Is, _Fp, _Vs...>()...);
+  }
+
+  template <class _Fp, class _Vp, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_fdiagonal() {
+    constexpr size_t _Np = __uncvref_t<_Vp>::__size();
+    static_assert(__all<(_Np == __uncvref_t<_Vs>::__size())...>::value);
+    return __make_fdiagonal_impl<_Fp, _Vp, _Vs...>(make_index_sequence<_Np>{});
+  }
+
+  template <class _Fp, class... _Vs, size_t... _Is>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_fmatrix_impl(index_sequence<_Is...> __is) {
+    return __make_dispatch<_Fp, _Vs...>(__is);
+  }
+
+  template <class _Fp, class... _Vs, size_t... _Is, size_t... _Js, class... _Ls>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_fmatrix_impl(index_sequence<_Is...>,
+                                            index_sequence<_Js...>,
+                                            _Ls... __ls) {
+    return __base::__make_farray(__make_fmatrix_impl<_Fp, _Vs...>(
+        index_sequence<_Is..., _Js>{}, __ls...)...);
+  }
+
+  template <class _Fp, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_fmatrix() {
+    return __make_fmatrix_impl<_Fp, _Vs...>(
+        index_sequence<>{}, make_index_sequence<__uncvref_t<_Vs>::__size()>{}...);
+  }
+};
+
+struct __variant {
+  template <class _Visitor, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr decltype(auto)
+  __visit_alt_at(size_t __index, _Visitor&& __visitor, _Vs&&... __vs) {
+    return __base::__visit_alt_at(__index,
+                                  _CUDA_VSTD::forward<_Visitor>(__visitor),
+                                  _CUDA_VSTD::forward<_Vs>(__vs).__impl...);
+  }
+
+  template <class _Visitor, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr decltype(auto) __visit_alt(_Visitor&& __visitor,
+                                              _Vs&&... __vs) {
+    return __base::__visit_alt(_CUDA_VSTD::forward<_Visitor>(__visitor),
+                               _CUDA_VSTD::forward<_Vs>(__vs).__impl...);
+  }
+
+  template <class _Visitor, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr decltype(auto)
+  __visit_value_at(size_t __index, _Visitor&& __visitor, _Vs&&... __vs) {
+    return __visit_alt_at(
+        __index,
+        __make_value_visitor(_CUDA_VSTD::forward<_Visitor>(__visitor)),
+        _CUDA_VSTD::forward<_Vs>(__vs)...);
+  }
+
+  template <class _Visitor, class... _Vs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr decltype(auto) __visit_value(_Visitor&& __visitor,
+                                                _Vs&&... __vs) {
+    return __visit_alt(
+        __make_value_visitor(_CUDA_VSTD::forward<_Visitor>(__visitor)),
+        _CUDA_VSTD::forward<_Vs>(__vs)...);
+  }
+
+private:
+  template <class _Visitor, class... _Values>
+  static constexpr void __std_visit_exhaustive_visitor_check() {
+    static_assert(is_invocable_v<_Visitor, _Values...>,
+                  "`std::visit` requires the visitor to be exhaustive.");
+  }
+
+  template <class _Visitor>
+  struct __value_visitor {
+    template <class... _Alts>
+    inline _LIBCUDACXX_INLINE_VISIBILITY
+    constexpr decltype(auto) operator()(_Alts&&... __alts) const {
+      __std_visit_exhaustive_visitor_check<
+          _Visitor,
+          decltype((_CUDA_VSTD::forward<_Alts>(__alts).__value))...>();
+      return __invoke_constexpr(_CUDA_VSTD::forward<_Visitor>(__visitor),
+                                _CUDA_VSTD::forward<_Alts>(__alts).__value...);
+    }
+    _Visitor&& __visitor;
+  };
+
+  template <class _Visitor>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr auto __make_value_visitor(_Visitor&& __visitor) {
+    return __value_visitor<_Visitor>{_CUDA_VSTD::forward<_Visitor>(__visitor)};
+  }
+};
+
+} // namespace __visitation
+
+template <size_t _Index, class _Tp>
+struct _LIBCUDACXX_TEMPLATE_VIS __alt {
+  using __value_type = _Tp;
+
+  template <class... _Args>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr __alt(in_place_t, _Args&&... __args)
+      : __value(_CUDA_VSTD::forward<_Args>(__args)...) {}
+
+  __value_type __value;
+};
+
+template <_Trait _DestructibleTrait, size_t _Index, class... _Types>
+union _LIBCUDACXX_TEMPLATE_VIS __union;
+
+template <_Trait _DestructibleTrait, size_t _Index>
+union _LIBCUDACXX_TEMPLATE_VIS __union<_DestructibleTrait, _Index> {};
+
+#define _LIBCUDACXX_VARIANT_UNION(destructible_trait, destructor)                  \
+  template <size_t _Index, class _Tp, class... _Types>                         \
+  union _LIBCUDACXX_TEMPLATE_VIS __union<destructible_trait,                      \
+                                      _Index,                                  \
+                                      _Tp,                                     \
+                                      _Types...> {                             \
+  public:                                                                      \
+    inline _LIBCUDACXX_INLINE_VISIBILITY                                           \
+    explicit constexpr __union(__valueless_t) noexcept : __dummy{} {}          \
+                                                                               \
+    template <class... _Args>                                                  \
+    inline _LIBCUDACXX_INLINE_VISIBILITY                                           \
+    explicit constexpr __union(in_place_index_t<0>, _Args&&... __args)         \
+        : __head(in_place, _CUDA_VSTD::forward<_Args>(__args)...) {}                \
+                                                                               \
+    template <size_t _Ip, class... _Args>                                      \
+    inline _LIBCUDACXX_INLINE_VISIBILITY                                           \
+    explicit constexpr __union(in_place_index_t<_Ip>, _Args&&... __args)       \
+        : __tail(in_place_index<_Ip - 1>, _CUDA_VSTD::forward<_Args>(__args)...) {} \
+                                                                               \
+    __union(const __union&) = default;                                         \
+    __union(__union&&) = default;                                              \
+                                                                               \
+    destructor                                                                 \
+                                                                               \
+    __union& operator=(const __union&) = default;                              \
+    __union& operator=(__union&&) = default;                                   \
+                                                                               \
+  private:                                                                     \
+    char __dummy;                                                              \
+    __alt<_Index, _Tp> __head;                                                 \
+    __union<destructible_trait, _Index + 1, _Types...> __tail;                 \
+                                                                               \
+    friend struct __access::__union;                                           \
+  }
+
+_LIBCUDACXX_VARIANT_UNION(_Trait::_TriviallyAvailable, ~__union() = default;);
+_LIBCUDACXX_VARIANT_UNION(_Trait::_Available, ~__union() {});
+_LIBCUDACXX_VARIANT_UNION(_Trait::_Unavailable, ~__union() = delete;);
+
+#undef _LIBCUDACXX_VARIANT_UNION
+
+template <_Trait _DestructibleTrait, class... _Types>
+class _LIBCUDACXX_TEMPLATE_VIS __base {
+public:
+  using __index_t = __variant_index_t<sizeof...(_Types)>;
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr __base(__valueless_t tag) noexcept
+      : __data(tag), __index(__variant_npos<__index_t>) {}
+
+  template <size_t _Ip, class... _Args>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr __base(in_place_index_t<_Ip>, _Args&&... __args)
+      :
+        __data(in_place_index<_Ip>, _CUDA_VSTD::forward<_Args>(__args)...),
+        __index(_Ip) {}
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr bool valueless_by_exception() const noexcept {
+    return index() == variant_npos;
+  }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr size_t index() const noexcept {
+    return __index == __variant_npos<__index_t> ? variant_npos : __index;
+  }
+
+protected:
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr auto&& __as_base() & { return *this; }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr auto&& __as_base() && { return _CUDA_VSTD::move(*this); }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr auto&& __as_base() const & { return *this; }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr auto&& __as_base() const && { return _CUDA_VSTD::move(*this); }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static constexpr size_t __size() { return sizeof...(_Types); }
+
+  __union<_DestructibleTrait, 0, _Types...> __data;
+  __index_t __index;
+
+  friend struct __access::__base;
+  friend struct __visitation::__base;
+};
+
+template <class _Traits, _Trait = _Traits::__destructible_trait>
+class _LIBCUDACXX_TEMPLATE_VIS __destructor;
+
+#define _LIBCUDACXX_VARIANT_DESTRUCTOR(destructible_trait, destructor, destroy)    \
+  template <class... _Types>                                                   \
+  class _LIBCUDACXX_TEMPLATE_VIS __destructor<__traits<_Types...>,                 \
+                                           destructible_trait>                 \
+      : public __base<destructible_trait, _Types...> {                         \
+    using __base_type = __base<destructible_trait, _Types...>;                 \
+    using __index_t = typename __base_type::__index_t;                         \
+                                                                               \
+  public:                                                                      \
+    using __base_type::__base_type;                                            \
+    using __base_type::operator=;                                              \
+                                                                               \
+    __destructor(const __destructor&) = default;                               \
+    __destructor(__destructor&&) = default;                                    \
+    destructor                                                                 \
+    __destructor& operator=(const __destructor&) = default;                    \
+    __destructor& operator=(__destructor&&) = default;                         \
+                                                                               \
+  protected:                                                                   \
+    inline _LIBCUDACXX_INLINE_VISIBILITY                                           \
+    destroy                                                                    \
+  }
+
+_LIBCUDACXX_VARIANT_DESTRUCTOR(
+    _Trait::_TriviallyAvailable,
+    ~__destructor() = default;,
+    void __destroy() noexcept { this->__index = __variant_npos<__index_t>; });
+
+_LIBCUDACXX_VARIANT_DESTRUCTOR(
+    _Trait::_Available,
+    ~__destructor() { __destroy(); },
+    void __destroy() noexcept {
+      if (!this->valueless_by_exception()) {
+        __visitation::__base::__visit_alt(
+            [](auto& __alt) noexcept {
+              using __alt_type = __uncvref_t<decltype(__alt)>;
+              __alt.~__alt_type();
+            },
+            *this);
+      }
+      this->__index = __variant_npos<__index_t>;
+    });
+
+_LIBCUDACXX_VARIANT_DESTRUCTOR(
+    _Trait::_Unavailable,
+    ~__destructor() = delete;,
+    void __destroy() noexcept = delete;);
+
+#undef _LIBCUDACXX_VARIANT_DESTRUCTOR
+
+template <class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS __constructor : public __destructor<_Traits> {
+  using __base_type = __destructor<_Traits>;
+
+public:
+  using __base_type::__base_type;
+  using __base_type::operator=;
+
+protected:
+  template <size_t _Ip, class _Tp, class... _Args>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static _Tp& __construct_alt(__alt<_Ip, _Tp>& __a, _Args&&... __args) {
+    ::new ((void*)_CUDA_VSTD::addressof(__a))
+        __alt<_Ip, _Tp>(in_place, _CUDA_VSTD::forward<_Args>(__args)...);
+    return __a.__value;
+  }
+
+  template <class _Rhs>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  static void __generic_construct(__constructor& __lhs, _Rhs&& __rhs) {
+    __lhs.__destroy();
+    if (!__rhs.valueless_by_exception()) {
+      __visitation::__base::__visit_alt_at(
+          __rhs.index(),
+          [](auto& __lhs_alt, auto&& __rhs_alt) {
+            __construct_alt(
+                __lhs_alt,
+                _CUDA_VSTD::forward<decltype(__rhs_alt)>(__rhs_alt).__value);
+          },
+          __lhs, _CUDA_VSTD::forward<_Rhs>(__rhs));
+      __lhs.__index = __rhs.index();
+    }
+  }
+};
+
+template <class _Traits, _Trait = _Traits::__move_constructible_trait>
+class _LIBCUDACXX_TEMPLATE_VIS __move_constructor;
+
+#define _LIBCUDACXX_VARIANT_MOVE_CONSTRUCTOR(move_constructible_trait,             \
+                                         move_constructor)                     \
+  template <class... _Types>                                                   \
+  class _LIBCUDACXX_TEMPLATE_VIS __move_constructor<__traits<_Types...>,          \
+                                                 move_constructible_trait>     \
+      : public __constructor<__traits<_Types...>> {                            \
+    using __base_type = __constructor<__traits<_Types...>>;                    \
+                                                                               \
+  public:                                                                      \
+    using __base_type::__base_type;                                            \
+    using __base_type::operator=;                                              \
+                                                                               \
+    __move_constructor(const __move_constructor&) = default;                   \
+    move_constructor                                                           \
+    ~__move_constructor() = default;                                           \
+    __move_constructor& operator=(const __move_constructor&) = default;        \
+    __move_constructor& operator=(__move_constructor&&) = default;             \
+  }
+
+_LIBCUDACXX_VARIANT_MOVE_CONSTRUCTOR(
+    _Trait::_TriviallyAvailable,
+    __move_constructor(__move_constructor&& __that) = default;);
+
+_LIBCUDACXX_VARIANT_MOVE_CONSTRUCTOR(
+    _Trait::_Available,
+    __move_constructor(__move_constructor&& __that) noexcept(
+        __all<is_nothrow_move_constructible_v<_Types>...>::value)
+        : __move_constructor(__valueless_t{}) {
+      this->__generic_construct(*this, _CUDA_VSTD::move(__that));
+    });
+
+_LIBCUDACXX_VARIANT_MOVE_CONSTRUCTOR(
+    _Trait::_Unavailable,
+    __move_constructor(__move_constructor&&) = delete;);
+
+#undef _LIBCUDACXX_VARIANT_MOVE_CONSTRUCTOR
+
+template <class _Traits, _Trait = _Traits::__copy_constructible_trait>
+class _LIBCUDACXX_TEMPLATE_VIS __copy_constructor;
+
+#define _LIBCUDACXX_VARIANT_COPY_CONSTRUCTOR(copy_constructible_trait,             \
+                                         copy_constructor)                     \
+  template <class... _Types>                                                   \
+  class _LIBCUDACXX_TEMPLATE_VIS __copy_constructor<__traits<_Types...>,          \
+                                                 copy_constructible_trait>     \
+      : public __move_constructor<__traits<_Types...>> {                       \
+    using __base_type = __move_constructor<__traits<_Types...>>;               \
+                                                                               \
+  public:                                                                      \
+    using __base_type::__base_type;                                            \
+    using __base_type::operator=;                                              \
+                                                                               \
+    copy_constructor                                                           \
+    __copy_constructor(__copy_constructor&&) = default;                        \
+    ~__copy_constructor() = default;                                           \
+    __copy_constructor& operator=(const __copy_constructor&) = default;        \
+    __copy_constructor& operator=(__copy_constructor&&) = default;             \
+  }
+
+_LIBCUDACXX_VARIANT_COPY_CONSTRUCTOR(
+    _Trait::_TriviallyAvailable,
+    __copy_constructor(const __copy_constructor& __that) = default;);
+
+_LIBCUDACXX_VARIANT_COPY_CONSTRUCTOR(
+    _Trait::_Available,
+    __copy_constructor(const __copy_constructor& __that)
+        : __copy_constructor(__valueless_t{}) {
+      this->__generic_construct(*this, __that);
+    });
+
+_LIBCUDACXX_VARIANT_COPY_CONSTRUCTOR(
+    _Trait::_Unavailable,
+    __copy_constructor(const __copy_constructor&) = delete;);
+
+#undef _LIBCUDACXX_VARIANT_COPY_CONSTRUCTOR
+
+template <class _Traits>
+class _LIBCUDACXX_TEMPLATE_VIS __assignment : public __copy_constructor<_Traits> {
+  using __base_type = __copy_constructor<_Traits>;
+
+public:
+  using __base_type::__base_type;
+  using __base_type::operator=;
+
+  template <size_t _Ip, class... _Args>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  auto& __emplace(_Args&&... __args) {
+    this->__destroy();
+    auto& __res = this->__construct_alt(__access::__base::__get_alt<_Ip>(*this),
+                          _CUDA_VSTD::forward<_Args>(__args)...);
+    this->__index = _Ip;
+    return __res;
+  }
+
+protected:
+  template <size_t _Ip, class _Tp, class _Arg>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  void __assign_alt(__alt<_Ip, _Tp>& __a, _Arg&& __arg) {
+    if (this->index() == _Ip) {
+      __a.__value = _CUDA_VSTD::forward<_Arg>(__arg);
+    } else {
+      struct {
+        void operator()(true_type) const {
+          __this->__emplace<_Ip>(_CUDA_VSTD::forward<_Arg>(__arg));
+        }
+        void operator()(false_type) const {
+          __this->__emplace<_Ip>(_Tp(_CUDA_VSTD::forward<_Arg>(__arg)));
+        }
+        __assignment* __this;
+        _Arg&& __arg;
+      } __impl{this, _CUDA_VSTD::forward<_Arg>(__arg)};
+      __impl(bool_constant<is_nothrow_constructible_v<_Tp, _Arg> ||
+                           !is_nothrow_move_constructible_v<_Tp>>{});
+    }
+  }
+
+  template <class _That>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  void __generic_assign(_That&& __that) {
+    if (this->valueless_by_exception() && __that.valueless_by_exception()) {
+      // do nothing.
+    } else if (__that.valueless_by_exception()) {
+      this->__destroy();
+    } else {
+      __visitation::__base::__visit_alt_at(
+          __that.index(),
+          [this](auto& __this_alt, auto&& __that_alt) {
+            this->__assign_alt(
+                __this_alt,
+                _CUDA_VSTD::forward<decltype(__that_alt)>(__that_alt).__value);
+          },
+          *this, _CUDA_VSTD::forward<_That>(__that));
+    }
+  }
+};
+
+template <class _Traits, _Trait = _Traits::__move_assignable_trait>
+class _LIBCUDACXX_TEMPLATE_VIS __move_assignment;
+
+#define _LIBCUDACXX_VARIANT_MOVE_ASSIGNMENT(move_assignable_trait,                 \
+                                        move_assignment)                       \
+  template <class... _Types>                                                   \
+  class _LIBCUDACXX_TEMPLATE_VIS __move_assignment<__traits<_Types...>,           \
+                                                move_assignable_trait>         \
+      : public __assignment<__traits<_Types...>> {                             \
+    using __base_type = __assignment<__traits<_Types...>>;                     \
+                                                                               \
+  public:                                                                      \
+    using __base_type::__base_type;                                            \
+    using __base_type::operator=;                                              \
+                                                                               \
+    __move_assignment(const __move_assignment&) = default;                     \
+    __move_assignment(__move_assignment&&) = default;                          \
+    ~__move_assignment() = default;                                            \
+    __move_assignment& operator=(const __move_assignment&) = default;          \
+    move_assignment                                                            \
+  }
+
+_LIBCUDACXX_VARIANT_MOVE_ASSIGNMENT(
+    _Trait::_TriviallyAvailable,
+    __move_assignment& operator=(__move_assignment&& __that) = default;);
+
+_LIBCUDACXX_VARIANT_MOVE_ASSIGNMENT(
+    _Trait::_Available,
+    __move_assignment& operator=(__move_assignment&& __that) noexcept(
+        __all<(is_nothrow_move_constructible_v<_Types> &&
+               is_nothrow_move_assignable_v<_Types>)...>::value) {
+      this->__generic_assign(_CUDA_VSTD::move(__that));
+      return *this;
+    });
+
+_LIBCUDACXX_VARIANT_MOVE_ASSIGNMENT(
+    _Trait::_Unavailable,
+    __move_assignment& operator=(__move_assignment&&) = delete;);
+
+#undef _LIBCUDACXX_VARIANT_MOVE_ASSIGNMENT
+
+template <class _Traits, _Trait = _Traits::__copy_assignable_trait>
+class _LIBCUDACXX_TEMPLATE_VIS __copy_assignment;
+
+#define _LIBCUDACXX_VARIANT_COPY_ASSIGNMENT(copy_assignable_trait,                 \
+                                        copy_assignment)                       \
+  template <class... _Types>                                                   \
+  class _LIBCUDACXX_TEMPLATE_VIS __copy_assignment<__traits<_Types...>,           \
+                                                copy_assignable_trait>         \
+      : public __move_assignment<__traits<_Types...>> {                        \
+    using __base_type = __move_assignment<__traits<_Types...>>;                \
+                                                                               \
+  public:                                                                      \
+    using __base_type::__base_type;                                            \
+    using __base_type::operator=;                                              \
+                                                                               \
+    __copy_assignment(const __copy_assignment&) = default;                     \
+    __copy_assignment(__copy_assignment&&) = default;                          \
+    ~__copy_assignment() = default;                                            \
+    copy_assignment                                                            \
+    __copy_assignment& operator=(__copy_assignment&&) = default;               \
+  }
+
+_LIBCUDACXX_VARIANT_COPY_ASSIGNMENT(
+    _Trait::_TriviallyAvailable,
+    __copy_assignment& operator=(const __copy_assignment& __that) = default;);
+
+_LIBCUDACXX_VARIANT_COPY_ASSIGNMENT(
+    _Trait::_Available,
+    __copy_assignment& operator=(const __copy_assignment& __that) {
+      this->__generic_assign(__that);
+      return *this;
+    });
+
+_LIBCUDACXX_VARIANT_COPY_ASSIGNMENT(
+    _Trait::_Unavailable,
+    __copy_assignment& operator=(const __copy_assignment&) = delete;);
+
+#undef _LIBCUDACXX_VARIANT_COPY_ASSIGNMENT
+
+template <class... _Types>
+class _LIBCUDACXX_TEMPLATE_VIS __impl
+    : public __copy_assignment<__traits<_Types...>> {
+  using __base_type = __copy_assignment<__traits<_Types...>>;
+
+public:
+  using __base_type::__base_type;
+  using __base_type::operator=;
+
+  template <size_t _Ip, class _Arg>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  void __assign(_Arg&& __arg) {
+    this->__assign_alt(__access::__base::__get_alt<_Ip>(*this),
+                       _CUDA_VSTD::forward<_Arg>(__arg));
+  }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  void __swap(__impl& __that)  {
+    if (this->valueless_by_exception() && __that.valueless_by_exception()) {
+      // do nothing.
+    } else if (this->index() == __that.index()) {
+      __visitation::__base::__visit_alt_at(
+          this->index(),
+          [](auto& __this_alt, auto& __that_alt) {
+            using _CUDA_VSTD::swap;
+            swap(__this_alt.__value, __that_alt.__value);
+          },
+          *this,
+          __that);
+    } else {
+      __impl* __lhs = this;
+      __impl* __rhs = _CUDA_VSTD::addressof(__that);
+      if (__lhs->__move_nothrow() && !__rhs->__move_nothrow()) {
+        _CUDA_VSTD::swap(__lhs, __rhs);
+      }
+      __impl __tmp(_CUDA_VSTD::move(*__rhs));
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+      // EXTENSION: When the move construction of `__lhs` into `__rhs` throws
+      // and `__tmp` is nothrow move constructible then we move `__tmp` back
+      // into `__rhs` and provide the strong exception safety guarantee.
+      try {
+        this->__generic_construct(*__rhs, _CUDA_VSTD::move(*__lhs));
+      } catch (...) {
+        if (__tmp.__move_nothrow()) {
+          this->__generic_construct(*__rhs, _CUDA_VSTD::move(__tmp));
+        }
+        throw;
+      }
+#else
+      this->__generic_construct(*__rhs, _CUDA_VSTD::move(*__lhs));
+#endif
+      this->__generic_construct(*__lhs, _CUDA_VSTD::move(__tmp));
+    }
+  }
+
+private:
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  bool __move_nothrow() const {
+    constexpr bool __results[] = {is_nothrow_move_constructible_v<_Types>...};
+    return this->valueless_by_exception() || __results[this->index()];
+  }
+};
+
+struct __no_narrowing_check {
+  template <class _Dest, class _Source>
+  using _Apply = __identity<_Dest>;
+};
+
+struct __narrowing_check {
+  template <class _Dest>
+  static auto __test_impl(_Dest (&&)[1]) -> __identity<_Dest>;
+  template <class _Dest, class _Source>
+  using _Apply _LIBCUDACXX_NODEBUG_TYPE = decltype(__test_impl<_Dest>({std::declval<_Source>()}));
+};
+
+template <class _Dest, class _Source>
+using __check_for_narrowing _LIBCUDACXX_NODEBUG_TYPE =
+  typename _If<
+#ifdef _LIBCUDACXX_ENABLE_NARROWING_CONVERSIONS_IN_VARIANT
+    false &&
+#endif
+    is_arithmetic<_Dest>::value,
+    __narrowing_check,
+    __no_narrowing_check
+  >::template _Apply<_Dest, _Source>;
+
+template <class _Tp, size_t _Idx>
+struct __overload {
+  template <class _Up>
+  auto operator()(_Tp, _Up&&) const -> __check_for_narrowing<_Tp, _Up>;
+};
+
+template <class _Tp, size_t>
+struct __overload_bool  {
+  template <class _Up, class _Ap = __uncvref_t<_Up>>
+  auto operator()(bool, _Up&&) const
+      -> enable_if_t<is_same_v<_Ap, bool>, __identity<_Tp>>;
+};
+
+template <size_t _Idx>
+struct __overload<bool, _Idx> : __overload_bool<bool, _Idx> {};
+template <size_t _Idx>
+struct __overload<bool const, _Idx> : __overload_bool<bool const, _Idx> {};
+template <size_t _Idx>
+struct __overload<bool volatile, _Idx> : __overload_bool<bool volatile, _Idx> {};
+template <size_t _Idx>
+struct __overload<bool const volatile, _Idx> : __overload_bool<bool const volatile, _Idx> {};
+
+template <class ..._Bases>
+struct __all_overloads : _Bases... {
+  void operator()() const;
+  using _Bases::operator()...;
+};
+
+template <class IdxSeq>
+struct __make_overloads_imp;
+
+template <size_t ..._Idx>
+struct __make_overloads_imp<__tuple_indices<_Idx...> > {
+  template <class ..._Types>
+  using _Apply _LIBCUDACXX_NODEBUG_TYPE = __all_overloads<__overload<_Types, _Idx>...>;
+};
+
+template <class ..._Types>
+using _MakeOverloads _LIBCUDACXX_NODEBUG_TYPE = typename __make_overloads_imp<
+    __make_indices_imp<sizeof...(_Types), 0> >::template _Apply<_Types...>;
+
+template <class _Tp, class... _Types>
+using __best_match_t =
+    typename invoke_result_t<_MakeOverloads<_Types...>, _Tp, _Tp>::type;
+
+} // __variant_detail
+
+template <class... _Types>
+class _LIBCUDACXX_TEMPLATE_VIS variant
+    : private __sfinae_ctor_base<
+          __all<is_copy_constructible_v<_Types>...>::value,
+          __all<is_move_constructible_v<_Types>...>::value>,
+      private __sfinae_assign_base<
+          __all<(is_copy_constructible_v<_Types> &&
+                 is_copy_assignable_v<_Types>)...>::value,
+          __all<(is_move_constructible_v<_Types> &&
+                 is_move_assignable_v<_Types>)...>::value> {
+  static_assert(0 < sizeof...(_Types),
+                "variant must consist of at least one alternative.");
+
+  static_assert(__all<!is_array_v<_Types>...>::value,
+                "variant can not have an array type as an alternative.");
+
+  static_assert(__all<!is_reference_v<_Types>...>::value,
+                "variant can not have a reference type as an alternative.");
+
+  static_assert(__all<!is_void_v<_Types>...>::value,
+                "variant can not have a void type as an alternative.");
+
+  using __first_type = variant_alternative_t<0, variant>;
+
+public:
+  template <bool _Dummy = true,
+            enable_if_t<__dependent_type<is_default_constructible<__first_type>,
+                                         _Dummy>::value,
+                        int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr variant() noexcept(is_nothrow_default_constructible_v<__first_type>)
+      : __impl(in_place_index<0>) {}
+
+  variant(const variant&) = default;
+  variant(variant&&) = default;
+
+  template <
+      class _Arg,
+      enable_if_t<!is_same_v<__uncvref_t<_Arg>, variant>, int> = 0,
+      enable_if_t<!__is_inplace_type<__uncvref_t<_Arg>>::value, int> = 0,
+      enable_if_t<!__is_inplace_index<__uncvref_t<_Arg>>::value, int> = 0,
+      class _Tp = __variant_detail::__best_match_t<_Arg, _Types...>,
+      size_t _Ip =
+          __find_detail::__find_unambiguous_index_sfinae<_Tp, _Types...>::value,
+      enable_if_t<is_constructible_v<_Tp, _Arg>, int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr variant(_Arg&& __arg) noexcept(
+      is_nothrow_constructible_v<_Tp, _Arg>)
+      : __impl(in_place_index<_Ip>, _CUDA_VSTD::forward<_Arg>(__arg)) {}
+
+  template <size_t _Ip, class... _Args,
+            class = enable_if_t<(_Ip < sizeof...(_Types)), int>,
+            class _Tp = variant_alternative_t<_Ip, variant<_Types...>>,
+            enable_if_t<is_constructible_v<_Tp, _Args...>, int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr variant(
+      in_place_index_t<_Ip>,
+      _Args&&... __args) noexcept(is_nothrow_constructible_v<_Tp, _Args...>)
+      : __impl(in_place_index<_Ip>, _CUDA_VSTD::forward<_Args>(__args)...) {}
+
+  template <
+      size_t _Ip,
+      class _Up,
+      class... _Args,
+      enable_if_t<(_Ip < sizeof...(_Types)), int> = 0,
+      class _Tp = variant_alternative_t<_Ip, variant<_Types...>>,
+      enable_if_t<is_constructible_v<_Tp, initializer_list<_Up>&, _Args...>,
+                  int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr variant(
+      in_place_index_t<_Ip>,
+      initializer_list<_Up> __il,
+      _Args&&... __args) noexcept(
+      is_nothrow_constructible_v<_Tp, initializer_list<_Up>&, _Args...>)
+      : __impl(in_place_index<_Ip>, __il, _CUDA_VSTD::forward<_Args>(__args)...) {}
+
+  template <
+      class _Tp,
+      class... _Args,
+      size_t _Ip =
+          __find_detail::__find_unambiguous_index_sfinae<_Tp, _Types...>::value,
+      enable_if_t<is_constructible_v<_Tp, _Args...>, int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr variant(in_place_type_t<_Tp>, _Args&&... __args) noexcept(
+      is_nothrow_constructible_v<_Tp, _Args...>)
+      : __impl(in_place_index<_Ip>, _CUDA_VSTD::forward<_Args>(__args)...) {}
+
+  template <
+      class _Tp,
+      class _Up,
+      class... _Args,
+      size_t _Ip =
+          __find_detail::__find_unambiguous_index_sfinae<_Tp, _Types...>::value,
+      enable_if_t<is_constructible_v<_Tp, initializer_list<_Up>&, _Args...>,
+                  int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  explicit constexpr variant(
+      in_place_type_t<_Tp>,
+      initializer_list<_Up> __il,
+      _Args&&... __args) noexcept(
+      is_nothrow_constructible_v<_Tp, initializer_list< _Up>&, _Args...>)
+      : __impl(in_place_index<_Ip>, __il, _CUDA_VSTD::forward<_Args>(__args)...) {}
+
+  ~variant() = default;
+
+  variant& operator=(const variant&) = default;
+  variant& operator=(variant&&) = default;
+
+  template <
+      class _Arg,
+      enable_if_t<!is_same_v<__uncvref_t<_Arg>, variant>, int> = 0,
+      class _Tp = __variant_detail::__best_match_t<_Arg, _Types...>,
+      size_t _Ip =
+          __find_detail::__find_unambiguous_index_sfinae<_Tp, _Types...>::value,
+      enable_if_t<is_assignable_v<_Tp&, _Arg> && is_constructible_v<_Tp, _Arg>,
+                  int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  variant& operator=(_Arg&& __arg) noexcept(
+      is_nothrow_assignable_v<_Tp&, _Arg> &&
+      is_nothrow_constructible_v<_Tp, _Arg>) {
+    __impl.template __assign<_Ip>(_CUDA_VSTD::forward<_Arg>(__arg));
+    return *this;
+  }
+
+  template <
+      size_t _Ip,
+      class... _Args,
+      enable_if_t<(_Ip < sizeof...(_Types)), int> = 0,
+      class _Tp = variant_alternative_t<_Ip, variant<_Types...>>,
+      enable_if_t<is_constructible_v<_Tp, _Args...>, int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  _Tp& emplace(_Args&&... __args) {
+    return __impl.template __emplace<_Ip>(_CUDA_VSTD::forward<_Args>(__args)...);
+  }
+
+  template <
+      size_t _Ip,
+      class _Up,
+      class... _Args,
+      enable_if_t<(_Ip < sizeof...(_Types)), int> = 0,
+      class _Tp = variant_alternative_t<_Ip, variant<_Types...>>,
+      enable_if_t<is_constructible_v<_Tp, initializer_list<_Up>&, _Args...>,
+                  int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  _Tp& emplace(initializer_list<_Up> __il, _Args&&... __args) {
+    return __impl.template __emplace<_Ip>(__il, _CUDA_VSTD::forward<_Args>(__args)...);
+  }
+
+  template <
+      class _Tp,
+      class... _Args,
+      size_t _Ip =
+          __find_detail::__find_unambiguous_index_sfinae<_Tp, _Types...>::value,
+      enable_if_t<is_constructible_v<_Tp, _Args...>, int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  _Tp& emplace(_Args&&... __args) {
+    return __impl.template __emplace<_Ip>(_CUDA_VSTD::forward<_Args>(__args)...);
+  }
+
+  template <
+      class _Tp,
+      class _Up,
+      class... _Args,
+      size_t _Ip =
+          __find_detail::__find_unambiguous_index_sfinae<_Tp, _Types...>::value,
+      enable_if_t<is_constructible_v<_Tp, initializer_list<_Up>&, _Args...>,
+                  int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  _Tp& emplace(initializer_list<_Up> __il, _Args&&... __args) {
+    return __impl.template __emplace<_Ip>(__il, _CUDA_VSTD::forward<_Args>(__args)...);
+  }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr bool valueless_by_exception() const noexcept {
+    return __impl.valueless_by_exception();
+  }
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  constexpr size_t index() const noexcept { return __impl.index(); }
+
+  template <
+      bool _Dummy = true,
+      enable_if_t<
+          __all<(
+              __dependent_type<is_move_constructible<_Types>, _Dummy>::value &&
+              __dependent_type<is_swappable<_Types>, _Dummy>::value)...>::value,
+          int> = 0>
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  void swap(variant& __that) noexcept(
+      __all<(is_nothrow_move_constructible_v<_Types> &&
+             is_nothrow_swappable_v<_Types>)...>::value) {
+    __impl.__swap(__that.__impl);
+  }
+
+private:
+  __variant_detail::__impl<_Types...> __impl;
+
+  friend struct __variant_detail::__access::__variant;
+  friend struct __variant_detail::__visitation::__variant;
+};
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool __holds_alternative(const variant<_Types...>& __v) noexcept {
+  return __v.index() == _Ip;
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool holds_alternative(const variant<_Types...>& __v) noexcept {
+  return __holds_alternative<__find_exactly_one_t<_Tp, _Types...>::value>(__v);
+}
+
+template <size_t _Ip, class _Vp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr auto&& __generic_get(_Vp&& __v) {
+  using __variant_detail::__access::__variant;
+  if (!__holds_alternative<_Ip>(__v)) {
+    __throw_bad_variant_access();
+  }
+  return __variant::__get_alt<_Ip>(_CUDA_VSTD::forward<_Vp>(__v)).__value;
+}
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr variant_alternative_t<_Ip, variant<_Types...>>& get(
+    variant<_Types...>& __v) {
+  static_assert(_Ip < sizeof...(_Types));
+  static_assert(!is_void_v<variant_alternative_t<_Ip, variant<_Types...>>>);
+  return __generic_get<_Ip>(__v);
+}
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr variant_alternative_t<_Ip, variant<_Types...>>&& get(
+    variant<_Types...>&& __v) {
+  static_assert(_Ip < sizeof...(_Types));
+  static_assert(!is_void_v<variant_alternative_t<_Ip, variant<_Types...>>>);
+  return __generic_get<_Ip>(_CUDA_VSTD::move(__v));
+}
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr const variant_alternative_t<_Ip, variant<_Types...>>& get(
+    const variant<_Types...>& __v) {
+  static_assert(_Ip < sizeof...(_Types));
+  static_assert(!is_void_v<variant_alternative_t<_Ip, variant<_Types...>>>);
+  return __generic_get<_Ip>(__v);
+}
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr const variant_alternative_t<_Ip, variant<_Types...>>&& get(
+    const variant<_Types...>&& __v) {
+  static_assert(_Ip < sizeof...(_Types));
+  static_assert(!is_void_v<variant_alternative_t<_Ip, variant<_Types...>>>);
+  return __generic_get<_Ip>(_CUDA_VSTD::move(__v));
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr _Tp& get(variant<_Types...>& __v) {
+  static_assert(!is_void_v<_Tp>);
+  return _CUDA_VSTD::get<__find_exactly_one_t<_Tp, _Types...>::value>(__v);
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr _Tp&& get(variant<_Types...>&& __v) {
+  static_assert(!is_void_v<_Tp>);
+  return _CUDA_VSTD::get<__find_exactly_one_t<_Tp, _Types...>::value>(
+      _CUDA_VSTD::move(__v));
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr const _Tp& get(const variant<_Types...>& __v) {
+  static_assert(!is_void_v<_Tp>);
+  return _CUDA_VSTD::get<__find_exactly_one_t<_Tp, _Types...>::value>(__v);
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr const _Tp&& get(const variant<_Types...>&& __v) {
+  static_assert(!is_void_v<_Tp>);
+  return _CUDA_VSTD::get<__find_exactly_one_t<_Tp, _Types...>::value>(
+      _CUDA_VSTD::move(__v));
+}
+
+template <size_t _Ip, class _Vp>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr auto* __generic_get_if(_Vp* __v) noexcept {
+  using __variant_detail::__access::__variant;
+  return __v && __holds_alternative<_Ip>(*__v)
+             ? _CUDA_VSTD::addressof(__variant::__get_alt<_Ip>(*__v).__value)
+             : nullptr;
+}
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr add_pointer_t<variant_alternative_t<_Ip, variant<_Types...>>>
+get_if(variant<_Types...>* __v) noexcept {
+  static_assert(_Ip < sizeof...(_Types));
+  static_assert(!is_void_v<variant_alternative_t<_Ip, variant<_Types...>>>);
+  return __generic_get_if<_Ip>(__v);
+}
+
+template <size_t _Ip, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr add_pointer_t<const variant_alternative_t<_Ip, variant<_Types...>>>
+get_if(const variant<_Types...>* __v) noexcept {
+  static_assert(_Ip < sizeof...(_Types));
+  static_assert(!is_void_v<variant_alternative_t<_Ip, variant<_Types...>>>);
+  return __generic_get_if<_Ip>(__v);
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr add_pointer_t<_Tp>
+get_if(variant<_Types...>* __v) noexcept {
+  static_assert(!is_void_v<_Tp>);
+  return _CUDA_VSTD::get_if<__find_exactly_one_t<_Tp, _Types...>::value>(__v);
+}
+
+template <class _Tp, class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr add_pointer_t<const _Tp>
+get_if(const variant<_Types...>* __v) noexcept {
+  static_assert(!is_void_v<_Tp>);
+  return _CUDA_VSTD::get_if<__find_exactly_one_t<_Tp, _Types...>::value>(__v);
+}
+
+template <class _Operator>
+struct __convert_to_bool {
+  template <class _T1, class _T2>
+  _LIBCUDACXX_INLINE_VISIBILITY constexpr bool operator()(_T1 && __t1, _T2&& __t2) const {
+    static_assert(std::is_convertible<decltype(_Operator{}(_CUDA_VSTD::forward<_T1>(__t1), _CUDA_VSTD::forward<_T2>(__t2))), bool>::value,
+        "the relational operator does not return a type which is implicitly convertible to bool");
+    return _Operator{}(_CUDA_VSTD::forward<_T1>(__t1), _CUDA_VSTD::forward<_T2>(__t2));
+  }
+};
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(const variant<_Types...>& __lhs,
+                          const variant<_Types...>& __rhs) {
+  using __variant_detail::__visitation::__variant;
+  if (__lhs.index() != __rhs.index()) return false;
+  if (__lhs.valueless_by_exception()) return true;
+  return __variant::__visit_value_at(__lhs.index(), __convert_to_bool<equal_to<>>{}, __lhs, __rhs);
+}
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(const variant<_Types...>& __lhs,
+                          const variant<_Types...>& __rhs) {
+  using __variant_detail::__visitation::__variant;
+  if (__lhs.index() != __rhs.index()) return true;
+  if (__lhs.valueless_by_exception()) return false;
+  return __variant::__visit_value_at(
+      __lhs.index(), __convert_to_bool<not_equal_to<>>{}, __lhs, __rhs);
+}
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<(const variant<_Types...>& __lhs,
+                         const variant<_Types...>& __rhs) {
+  using __variant_detail::__visitation::__variant;
+  if (__rhs.valueless_by_exception()) return false;
+  if (__lhs.valueless_by_exception()) return true;
+  if (__lhs.index() < __rhs.index()) return true;
+  if (__lhs.index() > __rhs.index()) return false;
+  return __variant::__visit_value_at(__lhs.index(), __convert_to_bool<less<>>{}, __lhs, __rhs);
+}
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>(const variant<_Types...>& __lhs,
+                         const variant<_Types...>& __rhs) {
+  using __variant_detail::__visitation::__variant;
+  if (__lhs.valueless_by_exception()) return false;
+  if (__rhs.valueless_by_exception()) return true;
+  if (__lhs.index() > __rhs.index()) return true;
+  if (__lhs.index() < __rhs.index()) return false;
+  return __variant::__visit_value_at(__lhs.index(), __convert_to_bool<greater<>>{}, __lhs, __rhs);
+}
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(const variant<_Types...>& __lhs,
+                          const variant<_Types...>& __rhs) {
+  using __variant_detail::__visitation::__variant;
+  if (__lhs.valueless_by_exception()) return true;
+  if (__rhs.valueless_by_exception()) return false;
+  if (__lhs.index() < __rhs.index()) return true;
+  if (__lhs.index() > __rhs.index()) return false;
+  return __variant::__visit_value_at(
+      __lhs.index(), __convert_to_bool<less_equal<>>{}, __lhs, __rhs);
+}
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(const variant<_Types...>& __lhs,
+                          const variant<_Types...>& __rhs) {
+  using __variant_detail::__visitation::__variant;
+  if (__rhs.valueless_by_exception()) return true;
+  if (__lhs.valueless_by_exception()) return false;
+  if (__lhs.index() > __rhs.index()) return true;
+  if (__lhs.index() < __rhs.index()) return false;
+  return __variant::__visit_value_at(
+      __lhs.index(), __convert_to_bool<greater_equal<>>{}, __lhs, __rhs);
+}
+
+template <class _Visitor, class... _Vs>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+_LIBCUDACXX_AVAILABILITY_THROW_BAD_VARIANT_ACCESS
+constexpr decltype(auto) visit(_Visitor&& __visitor, _Vs&&... __vs) {
+  using __variant_detail::__visitation::__variant;
+  bool __results[] = {__vs.valueless_by_exception()...};
+  for (bool __result : __results) {
+    if (__result) {
+      __throw_bad_variant_access();
+    }
+  }
+  return __variant::__visit_value(_CUDA_VSTD::forward<_Visitor>(__visitor),
+                                  _CUDA_VSTD::forward<_Vs>(__vs)...);
+}
+
+struct _LIBCUDACXX_TEMPLATE_VIS monostate {};
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<(monostate, monostate) noexcept { return false; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>(monostate, monostate) noexcept { return false; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator<=(monostate, monostate) noexcept { return true; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator>=(monostate, monostate) noexcept { return true; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator==(monostate, monostate) noexcept { return true; }
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+constexpr bool operator!=(monostate, monostate) noexcept { return false; }
+
+template <class... _Types>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+auto swap(variant<_Types...>& __lhs,
+          variant<_Types...>& __rhs) noexcept(noexcept(__lhs.swap(__rhs)))
+    -> decltype(__lhs.swap(__rhs)) {
+  __lhs.swap(__rhs);
+}
+
+template <class... _Types>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<
+    __enable_hash_helper<variant<_Types...>, remove_const_t<_Types>...>> {
+  using argument_type = variant<_Types...>;
+  using result_type = size_t;
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  result_type operator()(const argument_type& __v) const {
+    using __variant_detail::__visitation::__variant;
+    size_t __res =
+        __v.valueless_by_exception()
+               ? 299792458 // Random value chosen by the universe upon creation
+               : __variant::__visit_alt(
+                     [](const auto& __alt) {
+                       using __alt_type = __uncvref_t<decltype(__alt)>;
+                       using __value_type = remove_const_t<
+                         typename __alt_type::__value_type>;
+                       return hash<__value_type>{}(__alt.__value);
+                     },
+                     __v);
+    return __hash_combine(__res, hash<size_t>{}(__v.index()));
+  }
+};
+
+template <>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<monostate> {
+  using argument_type = monostate;
+  using result_type = size_t;
+
+  inline _LIBCUDACXX_INLINE_VISIBILITY
+  result_type operator()(const argument_type&) const _NOEXCEPT {
+    return 66740831; // return a fundamentally attractive random value.
+  }
+};
+
+#endif  // _LIBCUDACXX_STD_VER > 14
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_VARIANT
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/vector b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/vector
new file mode 100644
index 000000000000..0bcc09bf42b6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/vector
@@ -0,0 +1,3406 @@
+// -*- C++ -*-
+//===------------------------------ vector --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_VECTOR
+#define _LIBCUDACXX_VECTOR
+
+/*
+    vector synopsis
+
+namespace std
+{
+
+template <class T, class Allocator = allocator<T> >
+class vector
+{
+public:
+    typedef T                                        value_type;
+    typedef Allocator                                allocator_type;
+    typedef typename allocator_type::reference       reference;
+    typedef typename allocator_type::const_reference const_reference;
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+    typedef typename allocator_type::pointer         pointer;
+    typedef typename allocator_type::const_pointer   const_pointer;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+
+    vector()
+        noexcept(is_nothrow_default_constructible<allocator_type>::value);
+    explicit vector(const allocator_type&);
+    explicit vector(size_type n);
+    explicit vector(size_type n, const allocator_type&); // C++14
+    vector(size_type n, const value_type& value, const allocator_type& = allocator_type());
+    template <class InputIterator>
+        vector(InputIterator first, InputIterator last, const allocator_type& = allocator_type());
+    vector(const vector& x);
+    vector(vector&& x)
+        noexcept(is_nothrow_move_constructible<allocator_type>::value);
+    vector(initializer_list<value_type> il);
+    vector(initializer_list<value_type> il, const allocator_type& a);
+    ~vector();
+    vector& operator=(const vector& x);
+    vector& operator=(vector&& x)
+        noexcept(
+             allocator_type::propagate_on_container_move_assignment::value ||
+             allocator_type::is_always_equal::value); // C++17
+    vector& operator=(initializer_list<value_type> il);
+    template <class InputIterator>
+        void assign(InputIterator first, InputIterator last);
+    void assign(size_type n, const value_type& u);
+    void assign(initializer_list<value_type> il);
+
+    allocator_type get_allocator() const noexcept;
+
+    iterator               begin() noexcept;
+    const_iterator         begin()   const noexcept;
+    iterator               end() noexcept;
+    const_iterator         end()     const noexcept;
+
+    reverse_iterator       rbegin() noexcept;
+    const_reverse_iterator rbegin()  const noexcept;
+    reverse_iterator       rend() noexcept;
+    const_reverse_iterator rend()    const noexcept;
+
+    const_iterator         cbegin()  const noexcept;
+    const_iterator         cend()    const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend()   const noexcept;
+
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+    size_type capacity() const noexcept;
+    bool empty() const noexcept;
+    void reserve(size_type n);
+    void shrink_to_fit() noexcept;
+
+    reference       operator[](size_type n);
+    const_reference operator[](size_type n) const;
+    reference       at(size_type n);
+    const_reference at(size_type n) const;
+
+    reference       front();
+    const_reference front() const;
+    reference       back();
+    const_reference back() const;
+
+    value_type*       data() noexcept;
+    const value_type* data() const noexcept;
+
+    void push_back(const value_type& x);
+    void push_back(value_type&& x);
+    template <class... Args>
+        reference emplace_back(Args&&... args); // reference in C++17
+    void pop_back();
+
+    template <class... Args> iterator emplace(const_iterator position, Args&&... args);
+    iterator insert(const_iterator position, const value_type& x);
+    iterator insert(const_iterator position, value_type&& x);
+    iterator insert(const_iterator position, size_type n, const value_type& x);
+    template <class InputIterator>
+        iterator insert(const_iterator position, InputIterator first, InputIterator last);
+    iterator insert(const_iterator position, initializer_list<value_type> il);
+
+    iterator erase(const_iterator position);
+    iterator erase(const_iterator first, const_iterator last);
+
+    void clear() noexcept;
+
+    void resize(size_type sz);
+    void resize(size_type sz, const value_type& c);
+
+    void swap(vector&)
+        noexcept(allocator_traits<allocator_type>::propagate_on_container_swap::value ||
+                 allocator_traits<allocator_type>::is_always_equal::value);  // C++17
+
+    bool __invariants() const;
+};
+
+template <class Allocator = allocator<T> >
+class vector<bool, Allocator>
+{
+public:
+    typedef bool                                     value_type;
+    typedef Allocator                                allocator_type;
+    typedef implementation-defined                   iterator;
+    typedef implementation-defined                   const_iterator;
+    typedef typename allocator_type::size_type       size_type;
+    typedef typename allocator_type::difference_type difference_type;
+    typedef iterator                                 pointer;
+    typedef const_iterator                           const_pointer;
+    typedef std::reverse_iterator<iterator>          reverse_iterator;
+    typedef std::reverse_iterator<const_iterator>    const_reverse_iterator;
+
+    class reference
+    {
+    public:
+        reference(const reference&) noexcept;
+        operator bool() const noexcept;
+        reference& operator=(const bool x) noexcept;
+        reference& operator=(const reference& x) noexcept;
+        iterator operator&() const noexcept;
+        void flip() noexcept;
+    };
+
+    class const_reference
+    {
+    public:
+        const_reference(const reference&) noexcept;
+        operator bool() const noexcept;
+        const_iterator operator&() const noexcept;
+    };
+
+    vector()
+        noexcept(is_nothrow_default_constructible<allocator_type>::value);
+    explicit vector(const allocator_type&);
+    explicit vector(size_type n, const allocator_type& a = allocator_type()); // C++14
+    vector(size_type n, const value_type& value, const allocator_type& = allocator_type());
+    template <class InputIterator>
+        vector(InputIterator first, InputIterator last, const allocator_type& = allocator_type());
+    vector(const vector& x);
+    vector(vector&& x)
+        noexcept(is_nothrow_move_constructible<allocator_type>::value);
+    vector(initializer_list<value_type> il);
+    vector(initializer_list<value_type> il, const allocator_type& a);
+    ~vector();
+    vector& operator=(const vector& x);
+    vector& operator=(vector&& x)
+        noexcept(
+             allocator_type::propagate_on_container_move_assignment::value ||
+             allocator_type::is_always_equal::value); // C++17
+    vector& operator=(initializer_list<value_type> il);
+    template <class InputIterator>
+        void assign(InputIterator first, InputIterator last);
+    void assign(size_type n, const value_type& u);
+    void assign(initializer_list<value_type> il);
+
+    allocator_type get_allocator() const noexcept;
+
+    iterator               begin() noexcept;
+    const_iterator         begin()   const noexcept;
+    iterator               end() noexcept;
+    const_iterator         end()     const noexcept;
+
+    reverse_iterator       rbegin() noexcept;
+    const_reverse_iterator rbegin()  const noexcept;
+    reverse_iterator       rend() noexcept;
+    const_reverse_iterator rend()    const noexcept;
+
+    const_iterator         cbegin()  const noexcept;
+    const_iterator         cend()    const noexcept;
+    const_reverse_iterator crbegin() const noexcept;
+    const_reverse_iterator crend()   const noexcept;
+
+    size_type size() const noexcept;
+    size_type max_size() const noexcept;
+    size_type capacity() const noexcept;
+    bool empty() const noexcept;
+    void reserve(size_type n);
+    void shrink_to_fit() noexcept;
+
+    reference       operator[](size_type n);
+    const_reference operator[](size_type n) const;
+    reference       at(size_type n);
+    const_reference at(size_type n) const;
+
+    reference       front();
+    const_reference front() const;
+    reference       back();
+    const_reference back() const;
+
+    void push_back(const value_type& x);
+    template <class... Args> reference emplace_back(Args&&... args);  // C++14; reference in C++17
+    void pop_back();
+
+    template <class... Args> iterator emplace(const_iterator position, Args&&... args);  // C++14
+    iterator insert(const_iterator position, const value_type& x);
+    iterator insert(const_iterator position, size_type n, const value_type& x);
+    template <class InputIterator>
+        iterator insert(const_iterator position, InputIterator first, InputIterator last);
+    iterator insert(const_iterator position, initializer_list<value_type> il);
+
+    iterator erase(const_iterator position);
+    iterator erase(const_iterator first, const_iterator last);
+
+    void clear() noexcept;
+
+    void resize(size_type sz);
+    void resize(size_type sz, value_type x);
+
+    void swap(vector&)
+        noexcept(allocator_traits<allocator_type>::propagate_on_container_swap::value ||
+                 allocator_traits<allocator_type>::is_always_equal::value);  // C++17
+    void flip() noexcept;
+
+    bool __invariants() const;
+};
+
+template <class InputIterator, class Allocator = allocator<typename iterator_traits<InputIterator>::value_type>>
+   vector(InputIterator, InputIterator, Allocator = Allocator())
+   -> vector<typename iterator_traits<InputIterator>::value_type, Allocator>;
+
+template <class Allocator> struct hash<std::vector<bool, Allocator>>;
+
+template <class T, class Allocator> bool operator==(const vector<T,Allocator>& x, const vector<T,Allocator>& y);
+template <class T, class Allocator> bool operator< (const vector<T,Allocator>& x, const vector<T,Allocator>& y);
+template <class T, class Allocator> bool operator!=(const vector<T,Allocator>& x, const vector<T,Allocator>& y);
+template <class T, class Allocator> bool operator> (const vector<T,Allocator>& x, const vector<T,Allocator>& y);
+template <class T, class Allocator> bool operator>=(const vector<T,Allocator>& x, const vector<T,Allocator>& y);
+template <class T, class Allocator> bool operator<=(const vector<T,Allocator>& x, const vector<T,Allocator>& y);
+
+template <class T, class Allocator>
+void swap(vector<T,Allocator>& x, vector<T,Allocator>& y)
+    noexcept(noexcept(x.swap(y)));
+
+template <class T, class Allocator, class U>
+    void erase(vector<T, Allocator>& c, const U& value);       // C++20
+template <class T, class Allocator, class Predicate>
+    void erase_if(vector<T, Allocator>& c, Predicate pred);    // C++20
+
+}  // std
+
+*/
+
+#include <__config>
+#include <iosfwd> // for forward declaration of vector
+#include <__bit_reference>
+#include <type_traits>
+#include <climits>
+#include <limits>
+#include <initializer_list>
+#include <memory>
+#include <stdexcept>
+#include <algorithm>
+#include <cstring>
+#include <version>
+#include <__split_buffer>
+#include <__functional_base>
+
+#include <__debug>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+_LIBCUDACXX_PUSH_MACROS
+#include <__undef_macros>
+
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+
+template <bool>
+class _LIBCUDACXX_TEMPLATE_VIS __vector_base_common
+{
+protected:
+    _LIBCUDACXX_INLINE_VISIBILITY __vector_base_common() {}
+    _LIBCUDACXX_NORETURN void __throw_length_error() const;
+    _LIBCUDACXX_NORETURN void __throw_out_of_range() const;
+};
+
+template <bool __b>
+void
+__vector_base_common<__b>::__throw_length_error() const
+{
+    _CUDA_VSTD::__throw_length_error("vector");
+}
+
+template <bool __b>
+void
+__vector_base_common<__b>::__throw_out_of_range() const
+{
+    _CUDA_VSTD::__throw_out_of_range("vector");
+}
+
+_LIBCUDACXX_EXTERN_TEMPLATE(class _LIBCUDACXX_EXTERN_TEMPLATE_TYPE_VIS __vector_base_common<true>)
+
+template <class _Tp, class _Allocator>
+class __vector_base
+    : protected __vector_base_common<true>
+{
+public:
+    typedef _Allocator                               allocator_type;
+    typedef allocator_traits<allocator_type>         __alloc_traits;
+    typedef typename __alloc_traits::size_type       size_type;
+protected:
+    typedef _Tp                                      value_type;
+    typedef value_type&                              reference;
+    typedef const value_type&                        const_reference;
+    typedef typename __alloc_traits::difference_type difference_type;
+    typedef typename __alloc_traits::pointer         pointer;
+    typedef typename __alloc_traits::const_pointer   const_pointer;
+    typedef pointer                                  iterator;
+    typedef const_pointer                            const_iterator;
+
+    pointer                                         __begin_;
+    pointer                                         __end_;
+    __compressed_pair<pointer, allocator_type> __end_cap_;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type& __alloc() _NOEXCEPT
+        {return __end_cap_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const allocator_type& __alloc() const _NOEXCEPT
+        {return __end_cap_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    pointer& __end_cap() _NOEXCEPT
+        {return __end_cap_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const pointer& __end_cap() const _NOEXCEPT
+        {return __end_cap_.first();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __vector_base()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY __vector_base(const allocator_type& __a);
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY __vector_base(allocator_type&& __a) _NOEXCEPT;
+#endif
+    ~__vector_base();
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__destruct_at_end(__begin_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type capacity() const _NOEXCEPT
+        {return static_cast<size_type>(__end_cap() - __begin_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __destruct_at_end(pointer __new_last) _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __vector_base& __c)
+        {__copy_assign_alloc(__c, integral_constant<bool,
+                      __alloc_traits::propagate_on_container_copy_assignment::value>());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__vector_base& __c)
+        _NOEXCEPT_(
+            !__alloc_traits::propagate_on_container_move_assignment::value ||
+            is_nothrow_move_assignable<allocator_type>::value)
+        {__move_assign_alloc(__c, integral_constant<bool,
+                      __alloc_traits::propagate_on_container_move_assignment::value>());}
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __vector_base& __c, true_type)
+        {
+            if (__alloc() != __c.__alloc())
+            {
+                clear();
+                __alloc_traits::deallocate(__alloc(), __begin_, capacity());
+                __begin_ = __end_ = __end_cap() = nullptr;
+            }
+            __alloc() = __c.__alloc();
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const __vector_base&, false_type)
+        {}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__vector_base& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+        {
+            __alloc() = _CUDA_VSTD::move(__c.__alloc());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(__vector_base&, false_type)
+        _NOEXCEPT
+        {}
+};
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+__vector_base<_Tp, _Allocator>::__destruct_at_end(pointer __new_last) _NOEXCEPT
+{
+    pointer __soon_to_be_end = __end_;
+    while (__new_last != __soon_to_be_end)
+        __alloc_traits::destroy(__alloc(), _CUDA_VSTD::__to_raw_pointer(--__soon_to_be_end));
+    __end_ = __new_last;
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__vector_base<_Tp, _Allocator>::__vector_base()
+        _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+    : __begin_(nullptr),
+      __end_(nullptr),
+      __end_cap_(nullptr)
+{
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__vector_base<_Tp, _Allocator>::__vector_base(const allocator_type& __a)
+    : __begin_(nullptr),
+      __end_(nullptr),
+      __end_cap_(nullptr, __a)
+{
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+__vector_base<_Tp, _Allocator>::__vector_base(allocator_type&& __a) _NOEXCEPT
+    : __begin_(nullptr),
+      __end_(nullptr),
+      __end_cap_(nullptr, std::move(__a)) {}
+#endif
+
+template <class _Tp, class _Allocator>
+__vector_base<_Tp, _Allocator>::~__vector_base()
+{
+    if (__begin_ != nullptr)
+    {
+        clear();
+        __alloc_traits::deallocate(__alloc(), __begin_, capacity());
+    }
+}
+
+template <class _Tp, class _Allocator /* = allocator<_Tp> */>
+class _LIBCUDACXX_TEMPLATE_VIS vector
+    : private __vector_base<_Tp, _Allocator>
+{
+private:
+    typedef __vector_base<_Tp, _Allocator>           __base;
+    typedef allocator<_Tp>                           __default_allocator_type;
+public:
+    typedef vector                                   __self;
+    typedef _Tp                                      value_type;
+    typedef _Allocator                               allocator_type;
+    typedef typename __base::__alloc_traits          __alloc_traits;
+    typedef typename __base::reference               reference;
+    typedef typename __base::const_reference         const_reference;
+    typedef typename __base::size_type               size_type;
+    typedef typename __base::difference_type         difference_type;
+    typedef typename __base::pointer                 pointer;
+    typedef typename __base::const_pointer           const_pointer;
+    typedef __wrap_iter<pointer>                     iterator;
+    typedef __wrap_iter<const_pointer>               const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>         reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>   const_reverse_iterator;
+
+    static_assert((is_same<typename allocator_type::value_type, value_type>::value),
+                  "Allocator::value_type must be same type as value_type");
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector() _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+        {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+            __get_db()->__insert_c(this);
+#endif
+        }
+    _LIBCUDACXX_INLINE_VISIBILITY explicit vector(const allocator_type& __a)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_copy_constructible<allocator_type>::value)
+#else
+        _NOEXCEPT
+#endif
+        : __base(__a)
+    {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__insert_c(this);
+#endif
+    }
+    explicit vector(size_type __n);
+#if _LIBCUDACXX_STD_VER > 11
+    explicit vector(size_type __n, const allocator_type& __a);
+#endif
+    vector(size_type __n, const value_type& __x);
+    vector(size_type __n, const value_type& __x, const allocator_type& __a);
+    template <class _InputIterator>
+        vector(_InputIterator __first,
+               typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                                 !__is_forward_iterator<_InputIterator>::value &&
+                                 is_constructible<
+                                    value_type,
+                                    typename iterator_traits<_InputIterator>::reference>::value,
+                                 _InputIterator>::type __last);
+    template <class _InputIterator>
+        vector(_InputIterator __first, _InputIterator __last, const allocator_type& __a,
+               typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                                 !__is_forward_iterator<_InputIterator>::value &&
+                                 is_constructible<
+                                    value_type,
+                                    typename iterator_traits<_InputIterator>::reference>::value>::type* = 0);
+    template <class _ForwardIterator>
+        vector(_ForwardIterator __first,
+               typename enable_if<__is_forward_iterator<_ForwardIterator>::value &&
+                                 is_constructible<
+                                    value_type,
+                                    typename iterator_traits<_ForwardIterator>::reference>::value,
+                                 _ForwardIterator>::type __last);
+    template <class _ForwardIterator>
+        vector(_ForwardIterator __first, _ForwardIterator __last, const allocator_type& __a,
+               typename enable_if<__is_forward_iterator<_ForwardIterator>::value &&
+                                 is_constructible<
+                                    value_type,
+                                    typename iterator_traits<_ForwardIterator>::reference>::value>::type* = 0);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    ~vector()
+    {
+        __annotate_delete();
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->__erase_c(this);
+#endif
+    }
+
+    vector(const vector& __x);
+    vector(const vector& __x, const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector& operator=(const vector& __x);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector(initializer_list<value_type> __il);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector(initializer_list<value_type> __il, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector(vector&& __x)
+#if _LIBCUDACXX_STD_VER > 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value);
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector(vector&& __x, const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector& operator=(vector&& __x)
+        _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value));
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector& operator=(initializer_list<value_type> __il)
+        {assign(__il.begin(), __il.end()); return *this;}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+    template <class _InputIterator>
+        typename enable_if
+        <
+             __is_input_iterator  <_InputIterator>::value &&
+            !__is_forward_iterator<_InputIterator>::value &&
+            is_constructible<
+                 value_type,
+                 typename iterator_traits<_InputIterator>::reference>::value,
+            void
+        >::type
+        assign(_InputIterator __first, _InputIterator __last);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value &&
+            is_constructible<
+                 value_type,
+                 typename iterator_traits<_ForwardIterator>::reference>::value,
+            void
+        >::type
+        assign(_ForwardIterator __first, _ForwardIterator __last);
+
+    void assign(size_type __n, const_reference __u);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(initializer_list<value_type> __il)
+        {assign(__il.begin(), __il.end());}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    allocator_type get_allocator() const _NOEXCEPT
+        {return this->__alloc();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY iterator               begin() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator         begin()   const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY iterator               end() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY const_iterator         end()     const _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator       rbegin() _NOEXCEPT
+        {return       reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin()  const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator       rend() _NOEXCEPT
+        {return       reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend()    const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator         cbegin()  const _NOEXCEPT
+        {return begin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator         cend()    const _NOEXCEPT
+        {return end();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT
+        {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend()   const _NOEXCEPT
+        {return rend();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT
+        {return static_cast<size_type>(this->__end_ - this->__begin_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type capacity() const _NOEXCEPT
+        {return __base::capacity();}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT
+        {return this->__begin_ == this->__end_;}
+    size_type max_size() const _NOEXCEPT;
+    void reserve(size_type __n);
+    void shrink_to_fit() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference       operator[](size_type __n) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference operator[](size_type __n) const _NOEXCEPT;
+    reference       at(size_type __n);
+    const_reference at(size_type __n) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference       front() _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "front() called for empty vector");
+        return *this->__begin_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference front() const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "front() called for empty vector");
+        return *this->__begin_;
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY reference       back() _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "back() called for empty vector");
+        return *(this->__end_ - 1);
+    }
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference back()  const _NOEXCEPT
+    {
+        _LIBCUDACXX_ASSERT(!empty(), "back() called for empty vector");
+        return *(this->__end_ - 1);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    value_type*       data() _NOEXCEPT
+        {return _CUDA_VSTD::__to_raw_pointer(this->__begin_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const value_type* data() const _NOEXCEPT
+        {return _CUDA_VSTD::__to_raw_pointer(this->__begin_);}
+
+#ifdef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __emplace_back(const value_type& __x) { push_back(__x); }
+#else
+    template <class _Arg>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __emplace_back(_Arg&& __arg) {
+      emplace_back(_CUDA_VSTD::forward<_Arg>(__arg));
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY void push_back(const_reference __x);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY void push_back(value_type&& __x);
+
+    template <class... _Args>
+        _LIBCUDACXX_INLINE_VISIBILITY
+#if _LIBCUDACXX_STD_VER > 14
+        reference emplace_back(_Args&&... __args);
+#else
+        void      emplace_back(_Args&&... __args);
+#endif
+#endif // !_LIBCUDACXX_CXX03_LANG
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void pop_back();
+
+    iterator insert(const_iterator __position, const_reference __x);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    iterator insert(const_iterator __position, value_type&& __x);
+    template <class... _Args>
+        iterator emplace(const_iterator __position, _Args&&... __args);
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+    iterator insert(const_iterator __position, size_type __n, const_reference __x);
+    template <class _InputIterator>
+        typename enable_if
+        <
+             __is_input_iterator  <_InputIterator>::value &&
+            !__is_forward_iterator<_InputIterator>::value &&
+            is_constructible<
+                 value_type,
+                 typename iterator_traits<_InputIterator>::reference>::value,
+            iterator
+        >::type
+        insert(const_iterator __position, _InputIterator __first, _InputIterator __last);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value &&
+            is_constructible<
+                 value_type,
+                 typename iterator_traits<_ForwardIterator>::reference>::value,
+            iterator
+        >::type
+        insert(const_iterator __position, _ForwardIterator __first, _ForwardIterator __last);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __position, initializer_list<value_type> __il)
+        {return insert(__position, __il.begin(), __il.end());}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY iterator erase(const_iterator __position);
+    iterator erase(const_iterator __first, const_iterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT
+    {
+        size_type __old_size = size();
+        __base::clear();
+        __annotate_shrink(__old_size);
+        __invalidate_all_iterators();
+    }
+
+    void resize(size_type __sz);
+    void resize(size_type __sz, const_reference __x);
+
+    void swap(vector&)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value);
+#endif
+
+    bool __invariants() const;
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+    bool __dereferenceable(const const_iterator* __i) const;
+    bool __decrementable(const const_iterator* __i) const;
+    bool __addable(const const_iterator* __i, ptrdiff_t __n) const;
+    bool __subscriptable(const const_iterator* __i, ptrdiff_t __n) const;
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY void __invalidate_all_iterators();
+    _LIBCUDACXX_INLINE_VISIBILITY void __invalidate_iterators_past(pointer __new_last);
+    void __vallocate(size_type __n);
+    void __vdeallocate() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY size_type __recommend(size_type __new_size) const;
+    void __construct_at_end(size_type __n);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __construct_at_end(size_type __n, const_reference __x);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value,
+            void
+        >::type
+        __construct_at_end(_ForwardIterator __first, _ForwardIterator __last, size_type __n);
+    void __append(size_type __n);
+    void __append(size_type __n, const_reference __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator       __make_iter(pointer __p) _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator __make_iter(const_pointer __p) const _NOEXCEPT;
+    void __swap_out_circular_buffer(__split_buffer<value_type, allocator_type&>& __v);
+    pointer __swap_out_circular_buffer(__split_buffer<value_type, allocator_type&>& __v, pointer __p);
+    void __move_range(pointer __from_s, pointer __from_e, pointer __to);
+    void __move_assign(vector& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value);
+    void __move_assign(vector& __c, false_type)
+        _NOEXCEPT_(__alloc_traits::is_always_equal::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __destruct_at_end(pointer __new_last) _NOEXCEPT
+    {
+        __invalidate_iterators_past(__new_last);
+        size_type __old_size = size();
+        __base::__destruct_at_end(__new_last);
+        __annotate_shrink(__old_size);
+    }
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline void __push_back_slow_path(_Up&& __x);
+
+    template <class... _Args>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline void __emplace_back_slow_path(_Args&&... __args);
+#else
+    template <class _Up>
+    _LIBCUDACXX_INLINE_VISIBILITY
+    inline void __push_back_slow_path(_Up& __x);
+#endif
+
+    // The following functions are no-ops outside of AddressSanitizer mode.
+    // We call annotatations only for the default Allocator because other allocators
+    // may not meet the AddressSanitizer alignment constraints.
+    // See the documentation for __sanitizer_annotate_contiguous_container for more details.
+#ifndef _LIBCUDACXX_HAS_NO_ASAN
+    void __annotate_contiguous_container(const void *__beg, const void *__end,
+                                         const void *__old_mid,
+                                         const void *__new_mid) const
+    {
+
+      if (__beg && is_same<allocator_type, __default_allocator_type>::value)
+        __sanitizer_annotate_contiguous_container(__beg, __end, __old_mid, __new_mid);
+    }
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __annotate_contiguous_container(const void*, const void*, const void*,
+                                         const void*) const _NOEXCEPT {}
+#endif
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __annotate_new(size_type __current_size) const _NOEXCEPT {
+      __annotate_contiguous_container(data(), data() + capacity(),
+                                      data() + capacity(), data() + __current_size);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __annotate_delete() const _NOEXCEPT {
+      __annotate_contiguous_container(data(), data() + capacity(),
+                                      data() + size(), data() + capacity());
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __annotate_increase(size_type __n) const _NOEXCEPT
+    {
+      __annotate_contiguous_container(data(), data() + capacity(),
+                                      data() + size(), data() + size() + __n);
+    }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __annotate_shrink(size_type __old_size) const _NOEXCEPT
+    {
+      __annotate_contiguous_container(data(), data() + capacity(),
+                                      data() + __old_size, data() + size());
+    }
+
+  struct _ConstructTransaction {
+    explicit _ConstructTransaction(vector &__v, size_type __n)
+      : __v_(__v),  __pos_(__v.__end_), __new_end_(__v.__end_ + __n) {
+#ifndef _LIBCUDACXX_HAS_NO_ASAN
+      __v_.__annotate_increase(__n);
+#endif
+    }
+    ~_ConstructTransaction() {
+      __v_.__end_ = __pos_;
+#ifndef _LIBCUDACXX_HAS_NO_ASAN
+      if (__pos_ != __new_end_) {
+        __v_.__annotate_shrink(__new_end_ - __v_.__begin_);
+      }
+#endif
+    }
+
+    vector &__v_;
+    pointer __pos_;
+    const_pointer const __new_end_;
+
+  private:
+    _ConstructTransaction(_ConstructTransaction const&) = delete;
+    _ConstructTransaction& operator=(_ConstructTransaction const&) = delete;
+  };
+
+  template <class ..._Args>
+  _LIBCUDACXX_INLINE_VISIBILITY
+  void __construct_one_at_end(_Args&& ...__args) {
+    _ConstructTransaction __tx(*this, 1);
+    __alloc_traits::construct(this->__alloc(), _CUDA_VSTD::__to_raw_pointer(__tx.__pos_),
+        _CUDA_VSTD::forward<_Args>(__args)...);
+    ++__tx.__pos_;
+  }
+};
+
+#ifndef _LIBCUDACXX_HAS_NO_DEDUCTION_GUIDES
+template<class _InputIterator,
+         class _Alloc = typename std::allocator<typename iterator_traits<_InputIterator>::value_type>,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+vector(_InputIterator, _InputIterator)
+  -> vector<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+
+template<class _InputIterator,
+         class _Alloc,
+         class = typename enable_if<__is_allocator<_Alloc>::value, void>::type
+         >
+vector(_InputIterator, _InputIterator, _Alloc)
+  -> vector<typename iterator_traits<_InputIterator>::value_type, _Alloc>;
+#endif
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__swap_out_circular_buffer(__split_buffer<value_type, allocator_type&>& __v)
+{
+
+    __annotate_delete();
+    __alloc_traits::__construct_backward_with_exception_guarantees(
+        this->__alloc(), this->__begin_, this->__end_, __v.__begin_);
+    _CUDA_VSTD::swap(this->__begin_, __v.__begin_);
+    _CUDA_VSTD::swap(this->__end_, __v.__end_);
+    _CUDA_VSTD::swap(this->__end_cap(), __v.__end_cap());
+    __v.__first_ = __v.__begin_;
+    __annotate_new(size());
+    __invalidate_all_iterators();
+}
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::pointer
+vector<_Tp, _Allocator>::__swap_out_circular_buffer(__split_buffer<value_type, allocator_type&>& __v, pointer __p)
+{
+    __annotate_delete();
+    pointer __r = __v.__begin_;
+    __alloc_traits::__construct_backward_with_exception_guarantees(
+        this->__alloc(), this->__begin_, __p, __v.__begin_);
+    __alloc_traits::__construct_forward_with_exception_guarantees(
+        this->__alloc(), __p, this->__end_, __v.__end_);
+    _CUDA_VSTD::swap(this->__begin_, __v.__begin_);
+    _CUDA_VSTD::swap(this->__end_, __v.__end_);
+    _CUDA_VSTD::swap(this->__end_cap(), __v.__end_cap());
+    __v.__first_ = __v.__begin_;
+    __annotate_new(size());
+    __invalidate_all_iterators();
+    return __r;
+}
+
+//  Allocate space for __n objects
+//  throws length_error if __n > max_size()
+//  throws (probably bad_alloc) if memory run out
+//  Precondition:  __begin_ == __end_ == __end_cap() == 0
+//  Precondition:  __n > 0
+//  Postcondition:  capacity() == __n
+//  Postcondition:  size() == 0
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__vallocate(size_type __n)
+{
+    if (__n > max_size())
+        this->__throw_length_error();
+    this->__begin_ = this->__end_ = __alloc_traits::allocate(this->__alloc(), __n);
+    this->__end_cap() = this->__begin_ + __n;
+    __annotate_new(0);
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__vdeallocate() _NOEXCEPT
+{
+    if (this->__begin_ != nullptr)
+    {
+        clear();
+        __alloc_traits::deallocate(this->__alloc(), this->__begin_, capacity());
+        this->__begin_ = this->__end_ = this->__end_cap() = nullptr;
+    }
+}
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::size_type
+vector<_Tp, _Allocator>::max_size() const _NOEXCEPT
+{
+    return _CUDA_VSTD::min<size_type>(__alloc_traits::max_size(this->__alloc()),
+                                 numeric_limits<difference_type>::max());
+}
+
+//  Precondition:  __new_size > capacity()
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::size_type
+vector<_Tp, _Allocator>::__recommend(size_type __new_size) const
+{
+    const size_type __ms = max_size();
+    if (__new_size > __ms)
+        this->__throw_length_error();
+    const size_type __cap = capacity();
+    if (__cap >= __ms / 2)
+        return __ms;
+    return _CUDA_VSTD::max<size_type>(2*__cap, __new_size);
+}
+
+//  Default constructs __n objects starting at __end_
+//  throws if construction throws
+//  Precondition:  __n > 0
+//  Precondition:  size() + __n <= capacity()
+//  Postcondition:  size() == size() + __n
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__construct_at_end(size_type __n)
+{
+    _ConstructTransaction __tx(*this, __n);
+    for (; __tx.__pos_ != __tx.__new_end_; ++__tx.__pos_) {
+      __alloc_traits::construct(this->__alloc(), _CUDA_VSTD::__to_raw_pointer(__tx.__pos_));
+    }
+}
+
+//  Copy constructs __n objects starting at __end_ from __x
+//  throws if construction throws
+//  Precondition:  __n > 0
+//  Precondition:  size() + __n <= capacity()
+//  Postcondition:  size() == old size() + __n
+//  Postcondition:  [i] == __x for all i in [size() - __n, __n)
+template <class _Tp, class _Allocator>
+inline
+void
+vector<_Tp, _Allocator>::__construct_at_end(size_type __n, const_reference __x)
+{
+    _ConstructTransaction __tx(*this, __n);
+    for (; __tx.__pos_ != __tx.__new_end_; ++__tx.__pos_) {
+        __alloc_traits::construct(this->__alloc(), _CUDA_VSTD::__to_raw_pointer(__tx.__pos_), __x);
+    }
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value,
+    void
+>::type
+vector<_Tp, _Allocator>::__construct_at_end(_ForwardIterator __first, _ForwardIterator __last, size_type __n)
+{
+    _ConstructTransaction __tx(*this, __n);
+    __alloc_traits::__construct_range_forward(this->__alloc(), __first, __last, __tx.__pos_);
+}
+
+//  Default constructs __n objects starting at __end_
+//  throws if construction throws
+//  Postcondition:  size() == size() + __n
+//  Exception safety: strong.
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__append(size_type __n)
+{
+    if (static_cast<size_type>(this->__end_cap() - this->__end_) >= __n)
+        this->__construct_at_end(__n);
+    else
+    {
+        allocator_type& __a = this->__alloc();
+        __split_buffer<value_type, allocator_type&> __v(__recommend(size() + __n), size(), __a);
+        __v.__construct_at_end(__n);
+        __swap_out_circular_buffer(__v);
+    }
+}
+
+//  Default constructs __n objects starting at __end_
+//  throws if construction throws
+//  Postcondition:  size() == size() + __n
+//  Exception safety: strong.
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__append(size_type __n, const_reference __x)
+{
+    if (static_cast<size_type>(this->__end_cap() - this->__end_) >= __n)
+        this->__construct_at_end(__n, __x);
+    else
+    {
+        allocator_type& __a = this->__alloc();
+        __split_buffer<value_type, allocator_type&> __v(__recommend(size() + __n), size(), __a);
+        __v.__construct_at_end(__n, __x);
+        __swap_out_circular_buffer(__v);
+    }
+}
+
+template <class _Tp, class _Allocator>
+vector<_Tp, _Allocator>::vector(size_type __n)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n);
+    }
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Tp, class _Allocator>
+vector<_Tp, _Allocator>::vector(size_type __n, const allocator_type& __a)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n);
+    }
+}
+#endif
+
+template <class _Tp, class _Allocator>
+vector<_Tp, _Allocator>::vector(size_type __n, const value_type& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n, __x);
+    }
+}
+
+template <class _Tp, class _Allocator>
+vector<_Tp, _Allocator>::vector(size_type __n, const value_type& __x, const allocator_type& __a)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n, __x);
+    }
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIterator>
+vector<_Tp, _Allocator>::vector(_InputIterator __first,
+       typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                         !__is_forward_iterator<_InputIterator>::value &&
+                         is_constructible<
+                            value_type,
+                            typename iterator_traits<_InputIterator>::reference>::value,
+                          _InputIterator>::type __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __first != __last; ++__first)
+        __emplace_back(*__first);
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIterator>
+vector<_Tp, _Allocator>::vector(_InputIterator __first, _InputIterator __last, const allocator_type& __a,
+       typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                         !__is_forward_iterator<_InputIterator>::value &&
+                         is_constructible<
+                            value_type,
+                            typename iterator_traits<_InputIterator>::reference>::value>::type*)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    for (; __first != __last; ++__first)
+        __emplace_back(*__first);
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+vector<_Tp, _Allocator>::vector(_ForwardIterator __first,
+                                typename enable_if<__is_forward_iterator<_ForwardIterator>::value &&
+                                is_constructible<
+                                   value_type,
+                                   typename iterator_traits<_ForwardIterator>::reference>::value,
+                                                   _ForwardIterator>::type __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__first, __last, __n);
+    }
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+vector<_Tp, _Allocator>::vector(_ForwardIterator __first, _ForwardIterator __last, const allocator_type& __a,
+                                typename enable_if<__is_forward_iterator<_ForwardIterator>::value &&
+                                is_constructible<
+                                   value_type,
+                                   typename iterator_traits<_ForwardIterator>::reference>::value>::type*)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__first, __last, __n);
+    }
+}
+
+template <class _Tp, class _Allocator>
+vector<_Tp, _Allocator>::vector(const vector& __x)
+    : __base(__alloc_traits::select_on_container_copy_construction(__x.__alloc()))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    size_type __n = __x.size();
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__x.__begin_, __x.__end_, __n);
+    }
+}
+
+template <class _Tp, class _Allocator>
+vector<_Tp, _Allocator>::vector(const vector& __x, const allocator_type& __a)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    size_type __n = __x.size();
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__x.__begin_, __x.__end_, __n);
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<_Tp, _Allocator>::vector(vector&& __x)
+#if _LIBCUDACXX_STD_VER > 14
+        _NOEXCEPT
+#else
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value)
+#endif
+    : __base(_CUDA_VSTD::move(__x.__alloc()))
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+    __get_db()->swap(this, &__x);
+#endif
+    this->__begin_ = __x.__begin_;
+    this->__end_ = __x.__end_;
+    this->__end_cap() = __x.__end_cap();
+    __x.__begin_ = __x.__end_ = __x.__end_cap() = nullptr;
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<_Tp, _Allocator>::vector(vector&& __x, const allocator_type& __a)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__a == __x.__alloc())
+    {
+        this->__begin_ = __x.__begin_;
+        this->__end_ = __x.__end_;
+        this->__end_cap() = __x.__end_cap();
+        __x.__begin_ = __x.__end_ = __x.__end_cap() = nullptr;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+        __get_db()->swap(this, &__x);
+#endif
+    }
+    else
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__x.begin()), _Ip(__x.end()));
+    }
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<_Tp, _Allocator>::vector(initializer_list<value_type> __il)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__il.size() > 0)
+    {
+        __vallocate(__il.size());
+        __construct_at_end(__il.begin(), __il.end(), __il.size());
+    }
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<_Tp, _Allocator>::vector(initializer_list<value_type> __il, const allocator_type& __a)
+    : __base(__a)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__insert_c(this);
+#endif
+    if (__il.size() > 0)
+    {
+        __vallocate(__il.size());
+        __construct_at_end(__il.begin(), __il.end(), __il.size());
+    }
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<_Tp, _Allocator>&
+vector<_Tp, _Allocator>::operator=(vector&& __x)
+    _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value))
+{
+    __move_assign(__x, integral_constant<bool,
+          __alloc_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__move_assign(vector& __c, false_type)
+    _NOEXCEPT_(__alloc_traits::is_always_equal::value)
+{
+    if (__base::__alloc() != __c.__alloc())
+    {
+        typedef move_iterator<iterator> _Ip;
+        assign(_Ip(__c.begin()), _Ip(__c.end()));
+    }
+    else
+        __move_assign(__c, true_type());
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__move_assign(vector& __c, true_type)
+    _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+{
+    __vdeallocate();
+    __base::__move_assign_alloc(__c); // this can throw
+    this->__begin_ = __c.__begin_;
+    this->__end_ = __c.__end_;
+    this->__end_cap() = __c.__end_cap();
+    __c.__begin_ = __c.__end_ = __c.__end_cap() = nullptr;
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->swap(this, &__c);
+#endif
+}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<_Tp, _Allocator>&
+vector<_Tp, _Allocator>::operator=(const vector& __x)
+{
+    if (this != &__x)
+    {
+        __base::__copy_assign_alloc(__x);
+        assign(__x.__begin_, __x.__end_);
+    }
+    return *this;
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIterator>
+typename enable_if
+<
+     __is_input_iterator  <_InputIterator>::value &&
+    !__is_forward_iterator<_InputIterator>::value &&
+    is_constructible<
+       _Tp,
+       typename iterator_traits<_InputIterator>::reference>::value,
+    void
+>::type
+vector<_Tp, _Allocator>::assign(_InputIterator __first, _InputIterator __last)
+{
+    clear();
+    for (; __first != __last; ++__first)
+        __emplace_back(*__first);
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value &&
+    is_constructible<
+       _Tp,
+       typename iterator_traits<_ForwardIterator>::reference>::value,
+    void
+>::type
+vector<_Tp, _Allocator>::assign(_ForwardIterator __first, _ForwardIterator __last)
+{
+    size_type __new_size = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__new_size <= capacity())
+    {
+        _ForwardIterator __mid = __last;
+        bool __growing = false;
+        if (__new_size > size())
+        {
+            __growing = true;
+            __mid =  __first;
+            _CUDA_VSTD::advance(__mid, size());
+        }
+        pointer __m = _CUDA_VSTD::copy(__first, __mid, this->__begin_);
+        if (__growing)
+            __construct_at_end(__mid, __last, __new_size - size());
+        else
+            this->__destruct_at_end(__m);
+    }
+    else
+    {
+        __vdeallocate();
+        __vallocate(__recommend(__new_size));
+        __construct_at_end(__first, __last, __new_size);
+    }
+    __invalidate_all_iterators();
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::assign(size_type __n, const_reference __u)
+{
+    if (__n <= capacity())
+    {
+        size_type __s = size();
+        _CUDA_VSTD::fill_n(this->__begin_, _CUDA_VSTD::min(__n, __s), __u);
+        if (__n > __s)
+            __construct_at_end(__n - __s, __u);
+        else
+            this->__destruct_at_end(this->__begin_ + __n);
+    }
+    else
+    {
+        __vdeallocate();
+        __vallocate(__recommend(static_cast<size_type>(__n)));
+        __construct_at_end(__n, __u);
+    }
+    __invalidate_all_iterators();
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::__make_iter(pointer __p) _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return iterator(this, __p);
+#else
+    return iterator(__p);
+#endif
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::const_iterator
+vector<_Tp, _Allocator>::__make_iter(const_pointer __p) const _NOEXCEPT
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    return const_iterator(this, __p);
+#else
+    return const_iterator(__p);
+#endif
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::begin() _NOEXCEPT
+{
+    return __make_iter(this->__begin_);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::const_iterator
+vector<_Tp, _Allocator>::begin() const _NOEXCEPT
+{
+    return __make_iter(this->__begin_);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::end() _NOEXCEPT
+{
+    return __make_iter(this->__end_);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::const_iterator
+vector<_Tp, _Allocator>::end() const _NOEXCEPT
+{
+    return __make_iter(this->__end_);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::reference
+vector<_Tp, _Allocator>::operator[](size_type __n) _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n < size(), "vector[] index out of bounds");
+    return this->__begin_[__n];
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::const_reference
+vector<_Tp, _Allocator>::operator[](size_type __n) const _NOEXCEPT
+{
+    _LIBCUDACXX_ASSERT(__n < size(), "vector[] index out of bounds");
+    return this->__begin_[__n];
+}
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::reference
+vector<_Tp, _Allocator>::at(size_type __n)
+{
+    if (__n >= size())
+        this->__throw_out_of_range();
+    return this->__begin_[__n];
+}
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::const_reference
+vector<_Tp, _Allocator>::at(size_type __n) const
+{
+    if (__n >= size())
+        this->__throw_out_of_range();
+    return this->__begin_[__n];
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::reserve(size_type __n)
+{
+    if (__n > capacity())
+    {
+        allocator_type& __a = this->__alloc();
+        __split_buffer<value_type, allocator_type&> __v(__n, size(), __a);
+        __swap_out_circular_buffer(__v);
+    }
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::shrink_to_fit() _NOEXCEPT
+{
+    if (capacity() > size())
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            allocator_type& __a = this->__alloc();
+            __split_buffer<value_type, allocator_type&> __v(size(), size(), __a);
+            __swap_out_circular_buffer(__v);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Tp, class _Allocator>
+template <class _Up>
+void
+#ifndef _LIBCUDACXX_CXX03_LANG
+vector<_Tp, _Allocator>::__push_back_slow_path(_Up&& __x)
+#else
+vector<_Tp, _Allocator>::__push_back_slow_path(_Up& __x)
+#endif
+{
+    allocator_type& __a = this->__alloc();
+    __split_buffer<value_type, allocator_type&> __v(__recommend(size() + 1), size(), __a);
+    // __v.push_back(_CUDA_VSTD::forward<_Up>(__x));
+    __alloc_traits::construct(__a, _CUDA_VSTD::__to_raw_pointer(__v.__end_), _CUDA_VSTD::forward<_Up>(__x));
+    __v.__end_++;
+    __swap_out_circular_buffer(__v);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+vector<_Tp, _Allocator>::push_back(const_reference __x)
+{
+    if (this->__end_ != this->__end_cap())
+    {
+        __construct_one_at_end(__x);
+    }
+    else
+        __push_back_slow_path(__x);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+vector<_Tp, _Allocator>::push_back(value_type&& __x)
+{
+    if (this->__end_ < this->__end_cap())
+    {
+        __construct_one_at_end(_CUDA_VSTD::move(__x));
+    }
+    else
+        __push_back_slow_path(_CUDA_VSTD::move(__x));
+}
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+void
+vector<_Tp, _Allocator>::__emplace_back_slow_path(_Args&&... __args)
+{
+    allocator_type& __a = this->__alloc();
+    __split_buffer<value_type, allocator_type&> __v(__recommend(size() + 1), size(), __a);
+//    __v.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);
+    __alloc_traits::construct(__a, _CUDA_VSTD::__to_raw_pointer(__v.__end_), _CUDA_VSTD::forward<_Args>(__args)...);
+    __v.__end_++;
+    __swap_out_circular_buffer(__v);
+}
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+inline
+#if _LIBCUDACXX_STD_VER > 14
+typename vector<_Tp, _Allocator>::reference
+#else
+void
+#endif
+vector<_Tp, _Allocator>::emplace_back(_Args&&... __args)
+{
+    if (this->__end_ < this->__end_cap())
+    {
+        __construct_one_at_end(_CUDA_VSTD::forward<_Args>(__args)...);
+    }
+    else
+        __emplace_back_slow_path(_CUDA_VSTD::forward<_Args>(__args)...);
+#if _LIBCUDACXX_STD_VER > 14
+    return this->back();
+#endif
+}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+inline
+void
+vector<_Tp, _Allocator>::pop_back()
+{
+    _LIBCUDACXX_ASSERT(!empty(), "vector::pop_back called for empty vector");
+    this->__destruct_at_end(this->__end_ - 1);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::erase(const_iterator __position)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::erase(iterator) called with an iterator not"
+        " referring to this vector");
+#endif
+    _LIBCUDACXX_ASSERT(__position != end(),
+        "vector::erase(iterator) called with a non-dereferenceable iterator");
+    difference_type __ps = __position - cbegin();
+    pointer __p = this->__begin_ + __ps;
+    this->__destruct_at_end(_CUDA_VSTD::move(__p + 1, this->__end_, __p));
+    this->__invalidate_iterators_past(__p-1);
+    iterator __r = __make_iter(__p);
+    return __r;
+}
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::erase(const_iterator __first, const_iterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__first) == this,
+        "vector::erase(iterator,  iterator) called with an iterator not"
+        " referring to this vector");
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__last) == this,
+        "vector::erase(iterator,  iterator) called with an iterator not"
+        " referring to this vector");
+#endif
+    _LIBCUDACXX_ASSERT(__first <= __last, "vector::erase(first, last) called with invalid range");
+    pointer __p = this->__begin_ + (__first - begin());
+    if (__first != __last) {
+        this->__destruct_at_end(_CUDA_VSTD::move(__p + (__last - __first), this->__end_, __p));
+        this->__invalidate_iterators_past(__p - 1);
+    }
+    iterator __r = __make_iter(__p);
+    return __r;
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::__move_range(pointer __from_s, pointer __from_e, pointer __to)
+{
+    pointer __old_last = this->__end_;
+    difference_type __n = __old_last - __to;
+    {
+      pointer __i = __from_s + __n;
+      _ConstructTransaction __tx(*this, __from_e - __i);
+      for (; __i < __from_e; ++__i, ++__tx.__pos_) {
+          __alloc_traits::construct(this->__alloc(),
+                                    _CUDA_VSTD::__to_raw_pointer(__tx.__pos_),
+                                    _CUDA_VSTD::move(*__i));
+      }
+    }
+    _CUDA_VSTD::move_backward(__from_s, __from_s + __n, __old_last);
+}
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::insert(const_iterator __position, const_reference __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::insert(iterator, x) called with an iterator not"
+        " referring to this vector");
+#endif
+    pointer __p = this->__begin_ + (__position - begin());
+    if (this->__end_ < this->__end_cap())
+    {
+        if (__p == this->__end_)
+        {
+            __construct_one_at_end(__x);
+        }
+        else
+        {
+            __move_range(__p, this->__end_, __p + 1);
+            const_pointer __xr = pointer_traits<const_pointer>::pointer_to(__x);
+            if (__p <= __xr && __xr < this->__end_)
+                ++__xr;
+            *__p = *__xr;
+        }
+    }
+    else
+    {
+        allocator_type& __a = this->__alloc();
+        __split_buffer<value_type, allocator_type&> __v(__recommend(size() + 1), __p - this->__begin_, __a);
+        __v.push_back(__x);
+        __p = __swap_out_circular_buffer(__v, __p);
+    }
+    return __make_iter(__p);
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::insert(const_iterator __position, value_type&& __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::insert(iterator, x) called with an iterator not"
+        " referring to this vector");
+#endif
+    pointer __p = this->__begin_ + (__position - begin());
+    if (this->__end_ < this->__end_cap())
+    {
+        if (__p == this->__end_)
+        {
+            __construct_one_at_end(_CUDA_VSTD::move(__x));
+        }
+        else
+        {
+            __move_range(__p, this->__end_, __p + 1);
+            *__p = _CUDA_VSTD::move(__x);
+        }
+    }
+    else
+    {
+        allocator_type& __a = this->__alloc();
+        __split_buffer<value_type, allocator_type&> __v(__recommend(size() + 1), __p - this->__begin_, __a);
+        __v.push_back(_CUDA_VSTD::move(__x));
+        __p = __swap_out_circular_buffer(__v, __p);
+    }
+    return __make_iter(__p);
+}
+
+template <class _Tp, class _Allocator>
+template <class... _Args>
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::emplace(const_iterator __position, _Args&&... __args)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::emplace(iterator, x) called with an iterator not"
+        " referring to this vector");
+#endif
+    pointer __p = this->__begin_ + (__position - begin());
+    if (this->__end_ < this->__end_cap())
+    {
+        if (__p == this->__end_)
+        {
+            __construct_one_at_end(_CUDA_VSTD::forward<_Args>(__args)...);
+        }
+        else
+        {
+            __temp_value<value_type, _Allocator> __tmp(this->__alloc(), _CUDA_VSTD::forward<_Args>(__args)...);
+            __move_range(__p, this->__end_, __p + 1);
+            *__p = _CUDA_VSTD::move(__tmp.get());
+        }
+    }
+    else
+    {
+        allocator_type& __a = this->__alloc();
+        __split_buffer<value_type, allocator_type&> __v(__recommend(size() + 1), __p - this->__begin_, __a);
+        __v.emplace_back(_CUDA_VSTD::forward<_Args>(__args)...);
+        __p = __swap_out_circular_buffer(__v, __p);
+    }
+    return __make_iter(__p);
+}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+template <class _Tp, class _Allocator>
+typename vector<_Tp, _Allocator>::iterator
+vector<_Tp, _Allocator>::insert(const_iterator __position, size_type __n, const_reference __x)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::insert(iterator, n, x) called with an iterator not"
+        " referring to this vector");
+#endif
+    pointer __p = this->__begin_ + (__position - begin());
+    if (__n > 0)
+    {
+        if (__n <= static_cast<size_type>(this->__end_cap() - this->__end_))
+        {
+            size_type __old_n = __n;
+            pointer __old_last = this->__end_;
+            if (__n > static_cast<size_type>(this->__end_ - __p))
+            {
+                size_type __cx = __n - (this->__end_ - __p);
+                __construct_at_end(__cx, __x);
+                __n -= __cx;
+            }
+            if (__n > 0)
+            {
+                __move_range(__p, __old_last, __p + __old_n);
+                const_pointer __xr = pointer_traits<const_pointer>::pointer_to(__x);
+                if (__p <= __xr && __xr < this->__end_)
+                    __xr += __old_n;
+                _CUDA_VSTD::fill_n(__p, __n, *__xr);
+            }
+        }
+        else
+        {
+            allocator_type& __a = this->__alloc();
+            __split_buffer<value_type, allocator_type&> __v(__recommend(size() + __n), __p - this->__begin_, __a);
+            __v.__construct_at_end(__n, __x);
+            __p = __swap_out_circular_buffer(__v, __p);
+        }
+    }
+    return __make_iter(__p);
+}
+
+template <class _Tp, class _Allocator>
+template <class _InputIterator>
+typename enable_if
+<
+     __is_input_iterator  <_InputIterator>::value &&
+    !__is_forward_iterator<_InputIterator>::value &&
+    is_constructible<
+       _Tp,
+       typename iterator_traits<_InputIterator>::reference>::value,
+    typename vector<_Tp, _Allocator>::iterator
+>::type
+vector<_Tp, _Allocator>::insert(const_iterator __position, _InputIterator __first, _InputIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::insert(iterator, range) called with an iterator not"
+        " referring to this vector");
+#endif
+    difference_type __off = __position - begin();
+    pointer __p = this->__begin_ + __off;
+    allocator_type& __a = this->__alloc();
+    pointer __old_last = this->__end_;
+    for (; this->__end_ != this->__end_cap() && __first != __last; ++__first)
+    {
+        __construct_one_at_end(*__first);
+    }
+    __split_buffer<value_type, allocator_type&> __v(__a);
+    if (__first != __last)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            __v.__construct_at_end(__first, __last);
+            difference_type __old_size = __old_last - this->__begin_;
+            difference_type __old_p = __p - this->__begin_;
+            reserve(__recommend(size() + __v.size()));
+            __p = this->__begin_ + __old_p;
+            __old_last = this->__begin_ + __old_size;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            erase(__make_iter(__old_last), end());
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    __p = _CUDA_VSTD::rotate(__p, __old_last, this->__end_);
+    insert(__make_iter(__p), make_move_iterator(__v.begin()),
+                                    make_move_iterator(__v.end()));
+    return begin() + __off;
+}
+
+template <class _Tp, class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value &&
+    is_constructible<
+       _Tp,
+       typename iterator_traits<_ForwardIterator>::reference>::value,
+    typename vector<_Tp, _Allocator>::iterator
+>::type
+vector<_Tp, _Allocator>::insert(const_iterator __position, _ForwardIterator __first, _ForwardIterator __last)
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    _LIBCUDACXX_ASSERT(__get_const_db()->__find_c_from_i(&__position) == this,
+        "vector::insert(iterator, range) called with an iterator not"
+        " referring to this vector");
+#endif
+    pointer __p = this->__begin_ + (__position - begin());
+    difference_type __n = _CUDA_VSTD::distance(__first, __last);
+    if (__n > 0)
+    {
+        if (__n <= this->__end_cap() - this->__end_)
+        {
+            size_type __old_n = __n;
+            pointer __old_last = this->__end_;
+            _ForwardIterator __m = __last;
+            difference_type __dx = this->__end_ - __p;
+            if (__n > __dx)
+            {
+                __m = __first;
+                difference_type __diff = this->__end_ - __p;
+                _CUDA_VSTD::advance(__m, __diff);
+                __construct_at_end(__m, __last, __n - __diff);
+                __n = __dx;
+            }
+            if (__n > 0)
+            {
+                __move_range(__p, __old_last, __p + __old_n);
+                _CUDA_VSTD::copy(__first, __m, __p);
+            }
+        }
+        else
+        {
+            allocator_type& __a = this->__alloc();
+            __split_buffer<value_type, allocator_type&> __v(__recommend(size() + __n), __p - this->__begin_, __a);
+            __v.__construct_at_end(__first, __last);
+            __p = __swap_out_circular_buffer(__v, __p);
+        }
+    }
+    return __make_iter(__p);
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::resize(size_type __sz)
+{
+    size_type __cs = size();
+    if (__cs < __sz)
+        this->__append(__sz - __cs);
+    else if (__cs > __sz)
+        this->__destruct_at_end(this->__begin_ + __sz);
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::resize(size_type __sz, const_reference __x)
+{
+    size_type __cs = size();
+    if (__cs < __sz)
+        this->__append(__sz - __cs, __x);
+    else if (__cs > __sz)
+        this->__destruct_at_end(this->__begin_ + __sz);
+}
+
+template <class _Tp, class _Allocator>
+void
+vector<_Tp, _Allocator>::swap(vector& __x)
+#if _LIBCUDACXX_STD_VER >= 14
+    _NOEXCEPT
+#else
+    _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                __is_nothrow_swappable<allocator_type>::value)
+#endif
+{
+    _LIBCUDACXX_ASSERT(__alloc_traits::propagate_on_container_swap::value ||
+                   this->__alloc() == __x.__alloc(),
+                   "vector::swap: Either propagate_on_container_swap must be true"
+                   " or the allocators must compare equal");
+    _CUDA_VSTD::swap(this->__begin_, __x.__begin_);
+    _CUDA_VSTD::swap(this->__end_, __x.__end_);
+    _CUDA_VSTD::swap(this->__end_cap(), __x.__end_cap());
+    __swap_allocator(this->__alloc(), __x.__alloc(),
+        integral_constant<bool,__alloc_traits::propagate_on_container_swap::value>());
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->swap(this, &__x);
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+}
+
+template <class _Tp, class _Allocator>
+bool
+vector<_Tp, _Allocator>::__invariants() const
+{
+    if (this->__begin_ == nullptr)
+    {
+        if (this->__end_ != nullptr || this->__end_cap() != nullptr)
+            return false;
+    }
+    else
+    {
+        if (this->__begin_ > this->__end_)
+            return false;
+        if (this->__begin_ == this->__end_cap())
+            return false;
+        if (this->__end_ > this->__end_cap())
+            return false;
+    }
+    return true;
+}
+
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+template <class _Tp, class _Allocator>
+bool
+vector<_Tp, _Allocator>::__dereferenceable(const const_iterator* __i) const
+{
+    return this->__begin_ <= __i->base() && __i->base() < this->__end_;
+}
+
+template <class _Tp, class _Allocator>
+bool
+vector<_Tp, _Allocator>::__decrementable(const const_iterator* __i) const
+{
+    return this->__begin_ < __i->base() && __i->base() <= this->__end_;
+}
+
+template <class _Tp, class _Allocator>
+bool
+vector<_Tp, _Allocator>::__addable(const const_iterator* __i, ptrdiff_t __n) const
+{
+    const_pointer __p = __i->base() + __n;
+    return this->__begin_ <= __p && __p <= this->__end_;
+}
+
+template <class _Tp, class _Allocator>
+bool
+vector<_Tp, _Allocator>::__subscriptable(const const_iterator* __i, ptrdiff_t __n) const
+{
+    const_pointer __p = __i->base() + __n;
+    return this->__begin_ <= __p && __p < this->__end_;
+}
+
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+vector<_Tp, _Allocator>::__invalidate_all_iterators()
+{
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+    __get_db()->__invalidate_all(this);
+#endif  // _LIBCUDACXX_DEBUG_LEVEL >= 2
+}
+
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+vector<_Tp, _Allocator>::__invalidate_iterators_past(pointer __new_last) {
+#if _LIBCUDACXX_DEBUG_LEVEL >= 2
+  __c_node* __c = __get_db()->__find_c_and_lock(this);
+  for (__i_node** __p = __c->end_; __p != __c->beg_; ) {
+    --__p;
+    const_iterator* __i = static_cast<const_iterator*>((*__p)->__i_);
+    if (__i->base() > __new_last) {
+      (*__p)->__c_ = nullptr;
+      if (--__c->end_ != __p)
+        memmove(__p, __p+1, (__c->end_ - __p)*sizeof(__i_node*));
+    }
+  }
+  __get_db()->unlock();
+#else
+  ((void)__new_last);
+#endif
+}
+
+// vector<bool>
+
+template <class _Allocator> class vector<bool, _Allocator>;
+
+template <class _Allocator> struct hash<vector<bool, _Allocator> >;
+
+template <class _Allocator>
+struct __has_storage_type<vector<bool, _Allocator> >
+{
+    static const bool value = true;
+};
+
+template <class _Allocator>
+class _LIBCUDACXX_TEMPLATE_VIS vector<bool, _Allocator>
+    : private __vector_base_common<true>
+{
+public:
+    typedef vector                                   __self;
+    typedef bool                                     value_type;
+    typedef _Allocator                               allocator_type;
+    typedef allocator_traits<allocator_type>         __alloc_traits;
+    typedef typename __alloc_traits::size_type       size_type;
+    typedef typename __alloc_traits::difference_type difference_type;
+    typedef size_type __storage_type;
+    typedef __bit_iterator<vector, false>            pointer;
+    typedef __bit_iterator<vector, true>             const_pointer;
+    typedef pointer                                  iterator;
+    typedef const_pointer                            const_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<iterator>         reverse_iterator;
+    typedef _CUDA_VSTD::reverse_iterator<const_iterator>   const_reverse_iterator;
+
+private:
+    typedef typename __rebind_alloc_helper<__alloc_traits, __storage_type>::type __storage_allocator;
+    typedef allocator_traits<__storage_allocator>    __storage_traits;
+    typedef typename __storage_traits::pointer       __storage_pointer;
+    typedef typename __storage_traits::const_pointer __const_storage_pointer;
+
+    __storage_pointer                                      __begin_;
+    size_type                                              __size_;
+    __compressed_pair<size_type, __storage_allocator> __cap_alloc_;
+public:
+    typedef __bit_reference<vector>                  reference;
+    typedef __bit_const_reference<vector>            const_reference;
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type& __cap() _NOEXCEPT
+        {return __cap_alloc_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const size_type& __cap() const _NOEXCEPT
+        {return __cap_alloc_.first();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    __storage_allocator& __alloc() _NOEXCEPT
+        {return __cap_alloc_.second();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const __storage_allocator& __alloc() const _NOEXCEPT
+        {return __cap_alloc_.second();}
+
+    static const unsigned __bits_per_word = static_cast<unsigned>(sizeof(__storage_type) * CHAR_BIT);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type __internal_cap_to_external(size_type __n) _NOEXCEPT
+        {return __n * __bits_per_word;}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type __external_cap_to_internal(size_type __n) _NOEXCEPT
+        {return (__n - 1) / __bits_per_word + 1;}
+
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector() _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value);
+
+    _LIBCUDACXX_INLINE_VISIBILITY explicit vector(const allocator_type& __a)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_copy_constructible<allocator_type>::value);
+#else
+        _NOEXCEPT;
+#endif
+    ~vector();
+    explicit vector(size_type __n);
+#if _LIBCUDACXX_STD_VER > 11
+    explicit vector(size_type __n, const allocator_type& __a);
+#endif
+    vector(size_type __n, const value_type& __v);
+    vector(size_type __n, const value_type& __v, const allocator_type& __a);
+    template <class _InputIterator>
+        vector(_InputIterator __first, _InputIterator __last,
+               typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                                 !__is_forward_iterator<_InputIterator>::value>::type* = 0);
+    template <class _InputIterator>
+        vector(_InputIterator __first, _InputIterator __last, const allocator_type& __a,
+               typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                                 !__is_forward_iterator<_InputIterator>::value>::type* = 0);
+    template <class _ForwardIterator>
+        vector(_ForwardIterator __first, _ForwardIterator __last,
+               typename enable_if<__is_forward_iterator<_ForwardIterator>::value>::type* = 0);
+    template <class _ForwardIterator>
+        vector(_ForwardIterator __first, _ForwardIterator __last, const allocator_type& __a,
+               typename enable_if<__is_forward_iterator<_ForwardIterator>::value>::type* = 0);
+
+    vector(const vector& __v);
+    vector(const vector& __v, const allocator_type& __a);
+    vector& operator=(const vector& __v);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    vector(initializer_list<value_type> __il);
+    vector(initializer_list<value_type> __il, const allocator_type& __a);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector(vector&& __v)
+#if _LIBCUDACXX_STD_VER > 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value);
+#endif
+    vector(vector&& __v, const allocator_type& __a);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector& operator=(vector&& __v)
+        _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value));
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    vector& operator=(initializer_list<value_type> __il)
+        {assign(__il.begin(), __il.end()); return *this;}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+    template <class _InputIterator>
+        typename enable_if
+        <
+            __is_input_iterator<_InputIterator>::value &&
+           !__is_forward_iterator<_InputIterator>::value,
+           void
+        >::type
+        assign(_InputIterator __first, _InputIterator __last);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value,
+           void
+        >::type
+        assign(_ForwardIterator __first, _ForwardIterator __last);
+
+    void assign(size_type __n, const value_type& __x);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void assign(initializer_list<value_type> __il)
+        {assign(__il.begin(), __il.end());}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY allocator_type get_allocator() const _NOEXCEPT
+        {return allocator_type(this->__alloc());}
+
+    size_type max_size() const _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type capacity() const _NOEXCEPT
+        {return __internal_cap_to_external(__cap());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_type size() const _NOEXCEPT
+        {return __size_;}
+    _LIBCUDACXX_NODISCARD_AFTER_CXX17 _LIBCUDACXX_INLINE_VISIBILITY
+    bool empty() const _NOEXCEPT
+        {return __size_ == 0;}
+    void reserve(size_type __n);
+    void shrink_to_fit() _NOEXCEPT;
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator begin() _NOEXCEPT
+        {return __make_iter(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator begin() const _NOEXCEPT
+        {return __make_iter(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator end() _NOEXCEPT
+        {return __make_iter(__size_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator end()   const _NOEXCEPT
+        {return __make_iter(__size_);}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator rbegin() _NOEXCEPT
+        {return       reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rbegin() const _NOEXCEPT
+        {return const_reverse_iterator(end());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reverse_iterator rend() _NOEXCEPT
+        {return       reverse_iterator(begin());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator rend()   const _NOEXCEPT
+        {return const_reverse_iterator(begin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator         cbegin()  const _NOEXCEPT
+        {return __make_iter(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator         cend()    const _NOEXCEPT
+        {return __make_iter(__size_);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crbegin() const _NOEXCEPT
+        {return rbegin();}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reverse_iterator crend()   const _NOEXCEPT
+        {return rend();}
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference       operator[](size_type __n)       {return __make_ref(__n);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference operator[](size_type __n) const {return __make_ref(__n);}
+    reference       at(size_type __n);
+    const_reference at(size_type __n) const;
+
+    _LIBCUDACXX_INLINE_VISIBILITY reference       front()       {return __make_ref(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference front() const {return __make_ref(0);}
+    _LIBCUDACXX_INLINE_VISIBILITY reference       back()        {return __make_ref(__size_ - 1);}
+    _LIBCUDACXX_INLINE_VISIBILITY const_reference back()  const {return __make_ref(__size_ - 1);}
+
+    void push_back(const value_type& __x);
+#if _LIBCUDACXX_STD_VER > 11
+    template <class... _Args>
+#if _LIBCUDACXX_STD_VER > 14
+    _LIBCUDACXX_INLINE_VISIBILITY reference emplace_back(_Args&&... __args)
+#else
+    _LIBCUDACXX_INLINE_VISIBILITY void      emplace_back(_Args&&... __args)
+#endif
+    {
+        push_back ( value_type ( _CUDA_VSTD::forward<_Args>(__args)... ));
+#if _LIBCUDACXX_STD_VER > 14
+        return this->back();
+#endif
+    }
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY void pop_back() {--__size_;}
+
+#if _LIBCUDACXX_STD_VER > 11
+    template <class... _Args>
+   _LIBCUDACXX_INLINE_VISIBILITY iterator emplace(const_iterator position, _Args&&... __args)
+        { return insert ( position, value_type ( _CUDA_VSTD::forward<_Args>(__args)... )); }
+#endif
+
+    iterator insert(const_iterator __position, const value_type& __x);
+    iterator insert(const_iterator __position, size_type __n, const value_type& __x);
+    iterator insert(const_iterator __position, size_type __n, const_reference __x);
+    template <class _InputIterator>
+        typename enable_if
+        <
+             __is_input_iterator  <_InputIterator>::value &&
+            !__is_forward_iterator<_InputIterator>::value,
+            iterator
+        >::type
+        insert(const_iterator __position, _InputIterator __first, _InputIterator __last);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value,
+            iterator
+        >::type
+        insert(const_iterator __position, _ForwardIterator __first, _ForwardIterator __last);
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator insert(const_iterator __position, initializer_list<value_type> __il)
+        {return insert(__position, __il.begin(), __il.end());}
+#endif
+
+    _LIBCUDACXX_INLINE_VISIBILITY iterator erase(const_iterator __position);
+    iterator erase(const_iterator __first, const_iterator __last);
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void clear() _NOEXCEPT {__size_ = 0;}
+
+    void swap(vector&)
+#if _LIBCUDACXX_STD_VER >= 14
+        _NOEXCEPT;
+#else
+        _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                    __is_nothrow_swappable<allocator_type>::value);
+#endif
+    static void swap(reference __x, reference __y) _NOEXCEPT { _CUDA_VSTD::swap(__x, __y); }
+
+    void resize(size_type __sz, value_type __x = false);
+    void flip() _NOEXCEPT;
+
+    bool __invariants() const;
+
+private:
+    _LIBCUDACXX_INLINE_VISIBILITY void __invalidate_all_iterators();
+    void __vallocate(size_type __n);
+    void __vdeallocate() _NOEXCEPT;
+    _LIBCUDACXX_INLINE_VISIBILITY
+    static size_type __align_it(size_type __new_size) _NOEXCEPT
+        {return __new_size + (__bits_per_word-1) & ~((size_type)__bits_per_word-1);}
+    _LIBCUDACXX_INLINE_VISIBILITY  size_type __recommend(size_type __new_size) const;
+    _LIBCUDACXX_INLINE_VISIBILITY void __construct_at_end(size_type __n, bool __x);
+    template <class _ForwardIterator>
+        typename enable_if
+        <
+            __is_forward_iterator<_ForwardIterator>::value,
+            void
+        >::type
+        __construct_at_end(_ForwardIterator __first, _ForwardIterator __last);
+    void __append(size_type __n, const_reference __x);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    reference __make_ref(size_type __pos) _NOEXCEPT
+        {return reference(__begin_ + __pos / __bits_per_word, __storage_type(1) << __pos % __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_reference __make_ref(size_type __pos) const _NOEXCEPT
+        {return const_reference(__begin_ + __pos / __bits_per_word, __storage_type(1) << __pos % __bits_per_word);}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __make_iter(size_type __pos) _NOEXCEPT
+        {return iterator(__begin_ + __pos / __bits_per_word, static_cast<unsigned>(__pos % __bits_per_word));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    const_iterator __make_iter(size_type __pos) const _NOEXCEPT
+        {return const_iterator(__begin_ + __pos / __bits_per_word, static_cast<unsigned>(__pos % __bits_per_word));}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    iterator __const_iterator_cast(const_iterator __p) _NOEXCEPT
+        {return begin() + (__p - cbegin());}
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const vector& __v)
+        {__copy_assign_alloc(__v, integral_constant<bool,
+                      __storage_traits::propagate_on_container_copy_assignment::value>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const vector& __c, true_type)
+        {
+            if (__alloc() != __c.__alloc())
+                __vdeallocate();
+            __alloc() = __c.__alloc();
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __copy_assign_alloc(const vector&, false_type)
+        {}
+
+    void __move_assign(vector& __c, false_type);
+    void __move_assign(vector& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value);
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(vector& __c)
+        _NOEXCEPT_(
+            !__storage_traits::propagate_on_container_move_assignment::value ||
+            is_nothrow_move_assignable<allocator_type>::value)
+        {__move_assign_alloc(__c, integral_constant<bool,
+                      __storage_traits::propagate_on_container_move_assignment::value>());}
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(vector& __c, true_type)
+        _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+        {
+            __alloc() = _CUDA_VSTD::move(__c.__alloc());
+        }
+
+    _LIBCUDACXX_INLINE_VISIBILITY
+    void __move_assign_alloc(vector&, false_type)
+        _NOEXCEPT
+        {}
+
+    size_t __hash_code() const _NOEXCEPT;
+
+    friend class __bit_reference<vector>;
+    friend class __bit_const_reference<vector>;
+    friend class __bit_iterator<vector, false>;
+    friend class __bit_iterator<vector, true>;
+    friend struct __bit_array<vector>;
+    friend struct _LIBCUDACXX_TEMPLATE_VIS hash<vector>;
+};
+
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+vector<bool, _Allocator>::__invalidate_all_iterators()
+{
+}
+
+//  Allocate space for __n objects
+//  throws length_error if __n > max_size()
+//  throws (probably bad_alloc) if memory run out
+//  Precondition:  __begin_ == __end_ == __cap() == 0
+//  Precondition:  __n > 0
+//  Postcondition:  capacity() == __n
+//  Postcondition:  size() == 0
+template <class _Allocator>
+void
+vector<bool, _Allocator>::__vallocate(size_type __n)
+{
+    if (__n > max_size())
+        this->__throw_length_error();
+    __n = __external_cap_to_internal(__n);
+    this->__begin_ = __storage_traits::allocate(this->__alloc(), __n);
+    this->__size_ = 0;
+    this->__cap() = __n;
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::__vdeallocate() _NOEXCEPT
+{
+    if (this->__begin_ != nullptr)
+    {
+        __storage_traits::deallocate(this->__alloc(), this->__begin_, __cap());
+        __invalidate_all_iterators();
+        this->__begin_ = nullptr;
+        this->__size_ = this->__cap() = 0;
+    }
+}
+
+template <class _Allocator>
+typename vector<bool, _Allocator>::size_type
+vector<bool, _Allocator>::max_size() const _NOEXCEPT
+{
+    size_type __amax = __storage_traits::max_size(__alloc());
+    size_type __nmax = numeric_limits<size_type>::max() / 2;  // end() >= begin(), always
+    if (__nmax / __bits_per_word <= __amax)
+        return __nmax;
+    return __internal_cap_to_external(__amax);
+}
+
+//  Precondition:  __new_size > capacity()
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<bool, _Allocator>::size_type
+vector<bool, _Allocator>::__recommend(size_type __new_size) const
+{
+    const size_type __ms = max_size();
+    if (__new_size > __ms)
+        this->__throw_length_error();
+    const size_type __cap = capacity();
+    if (__cap >= __ms / 2)
+        return __ms;
+    return _CUDA_VSTD::max(2*__cap, __align_it(__new_size));
+}
+
+//  Default constructs __n objects starting at __end_
+//  Precondition:  __n > 0
+//  Precondition:  size() + __n <= capacity()
+//  Postcondition:  size() == size() + __n
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+vector<bool, _Allocator>::__construct_at_end(size_type __n, bool __x)
+{
+    size_type __old_size = this->__size_;
+    this->__size_ += __n;
+    if (__old_size == 0 || ((__old_size - 1) / __bits_per_word) != ((this->__size_ - 1) / __bits_per_word))
+    {
+        if (this->__size_ <= __bits_per_word)
+            this->__begin_[0] = __storage_type(0);
+        else
+            this->__begin_[(this->__size_ - 1) / __bits_per_word] = __storage_type(0);
+    }
+    _CUDA_VSTD::fill_n(__make_iter(__old_size), __n, __x);
+}
+
+template <class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value,
+    void
+>::type
+vector<bool, _Allocator>::__construct_at_end(_ForwardIterator __first, _ForwardIterator __last)
+{
+    size_type __old_size = this->__size_;
+    this->__size_ += _CUDA_VSTD::distance(__first, __last);
+    if (__old_size == 0 || ((__old_size - 1) / __bits_per_word) != ((this->__size_ - 1) / __bits_per_word))
+    {
+        if (this->__size_ <= __bits_per_word)
+            this->__begin_[0] = __storage_type(0);
+        else
+            this->__begin_[(this->__size_ - 1) / __bits_per_word] = __storage_type(0);
+    }
+    _CUDA_VSTD::copy(__first, __last, __make_iter(__old_size));
+}
+
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<bool, _Allocator>::vector()
+    _NOEXCEPT_(is_nothrow_default_constructible<allocator_type>::value)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0)
+{
+}
+
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<bool, _Allocator>::vector(const allocator_type& __a)
+#if _LIBCUDACXX_STD_VER <= 14
+        _NOEXCEPT_(is_nothrow_copy_constructible<allocator_type>::value)
+#else
+        _NOEXCEPT
+#endif
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, static_cast<__storage_allocator>(__a))
+{
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(size_type __n)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0)
+{
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n, false);
+    }
+}
+
+#if _LIBCUDACXX_STD_VER > 11
+template <class _Allocator>
+vector<bool, _Allocator>::vector(size_type __n, const allocator_type& __a)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, static_cast<__storage_allocator>(__a))
+{
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n, false);
+    }
+}
+#endif
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(size_type __n, const value_type& __x)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0)
+{
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n, __x);
+    }
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(size_type __n, const value_type& __x, const allocator_type& __a)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, static_cast<__storage_allocator>(__a))
+{
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__n, __x);
+    }
+}
+
+template <class _Allocator>
+template <class _InputIterator>
+vector<bool, _Allocator>::vector(_InputIterator __first, _InputIterator __last,
+       typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                         !__is_forward_iterator<_InputIterator>::value>::type*)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0)
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        for (; __first != __last; ++__first)
+            push_back(*__first);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        if (__begin_ != nullptr)
+            __storage_traits::deallocate(__alloc(), __begin_, __cap());
+        __invalidate_all_iterators();
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _Allocator>
+template <class _InputIterator>
+vector<bool, _Allocator>::vector(_InputIterator __first, _InputIterator __last, const allocator_type& __a,
+       typename enable_if<__is_input_iterator  <_InputIterator>::value &&
+                         !__is_forward_iterator<_InputIterator>::value>::type*)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, static_cast<__storage_allocator>(__a))
+{
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    try
+    {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+        for (; __first != __last; ++__first)
+            push_back(*__first);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    catch (...)
+    {
+        if (__begin_ != nullptr)
+            __storage_traits::deallocate(__alloc(), __begin_, __cap());
+        __invalidate_all_iterators();
+        throw;
+    }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+}
+
+template <class _Allocator>
+template <class _ForwardIterator>
+vector<bool, _Allocator>::vector(_ForwardIterator __first, _ForwardIterator __last,
+                                typename enable_if<__is_forward_iterator<_ForwardIterator>::value>::type*)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0)
+{
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__first, __last);
+    }
+}
+
+template <class _Allocator>
+template <class _ForwardIterator>
+vector<bool, _Allocator>::vector(_ForwardIterator __first, _ForwardIterator __last, const allocator_type& __a,
+                                typename enable_if<__is_forward_iterator<_ForwardIterator>::value>::type*)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, static_cast<__storage_allocator>(__a))
+{
+    size_type __n = static_cast<size_type>(_CUDA_VSTD::distance(__first, __last));
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__first, __last);
+    }
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(initializer_list<value_type> __il)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0)
+{
+    size_type __n = static_cast<size_type>(__il.size());
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__il.begin(), __il.end());
+    }
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(initializer_list<value_type> __il, const allocator_type& __a)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, static_cast<__storage_allocator>(__a))
+{
+    size_type __n = static_cast<size_type>(__il.size());
+    if (__n > 0)
+    {
+        __vallocate(__n);
+        __construct_at_end(__il.begin(), __il.end());
+    }
+}
+
+#endif  // _LIBCUDACXX_CXX03_LANG
+
+template <class _Allocator>
+vector<bool, _Allocator>::~vector()
+{
+    if (__begin_ != nullptr)
+        __storage_traits::deallocate(__alloc(), __begin_, __cap());
+    __invalidate_all_iterators();
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(const vector& __v)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, __storage_traits::select_on_container_copy_construction(__v.__alloc()))
+{
+    if (__v.size() > 0)
+    {
+        __vallocate(__v.size());
+        __construct_at_end(__v.begin(), __v.end());
+    }
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(const vector& __v, const allocator_type& __a)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, __a)
+{
+    if (__v.size() > 0)
+    {
+        __vallocate(__v.size());
+        __construct_at_end(__v.begin(), __v.end());
+    }
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>&
+vector<bool, _Allocator>::operator=(const vector& __v)
+{
+    if (this != &__v)
+    {
+        __copy_assign_alloc(__v);
+        if (__v.__size_)
+        {
+            if (__v.__size_ > capacity())
+            {
+                __vdeallocate();
+                __vallocate(__v.__size_);
+            }
+            _CUDA_VSTD::copy(__v.__begin_, __v.__begin_ + __external_cap_to_internal(__v.__size_), __begin_);
+        }
+        __size_ = __v.__size_;
+    }
+    return *this;
+}
+
+#ifndef _LIBCUDACXX_CXX03_LANG
+
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY vector<bool, _Allocator>::vector(vector&& __v)
+#if _LIBCUDACXX_STD_VER > 14
+    _NOEXCEPT
+#else
+    _NOEXCEPT_(is_nothrow_move_constructible<allocator_type>::value)
+#endif
+    : __begin_(__v.__begin_),
+      __size_(__v.__size_),
+      __cap_alloc_(std::move(__v.__cap_alloc_)) {
+    __v.__begin_ = nullptr;
+    __v.__size_ = 0;
+    __v.__cap() = 0;
+}
+
+template <class _Allocator>
+vector<bool, _Allocator>::vector(vector&& __v, const allocator_type& __a)
+    : __begin_(nullptr),
+      __size_(0),
+      __cap_alloc_(0, __a)
+{
+    if (__a == allocator_type(__v.__alloc()))
+    {
+        this->__begin_ = __v.__begin_;
+        this->__size_ = __v.__size_;
+        this->__cap() = __v.__cap();
+        __v.__begin_ = nullptr;
+        __v.__cap() = __v.__size_ = 0;
+    }
+    else if (__v.size() > 0)
+    {
+        __vallocate(__v.size());
+        __construct_at_end(__v.begin(), __v.end());
+    }
+}
+
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+vector<bool, _Allocator>&
+vector<bool, _Allocator>::operator=(vector&& __v)
+    _NOEXCEPT_((__noexcept_move_assign_container<_Allocator, __alloc_traits>::value))
+{
+    __move_assign(__v, integral_constant<bool,
+          __storage_traits::propagate_on_container_move_assignment::value>());
+    return *this;
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::__move_assign(vector& __c, false_type)
+{
+    if (__alloc() != __c.__alloc())
+        assign(__c.begin(), __c.end());
+    else
+        __move_assign(__c, true_type());
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::__move_assign(vector& __c, true_type)
+    _NOEXCEPT_(is_nothrow_move_assignable<allocator_type>::value)
+{
+    __vdeallocate();
+    __move_assign_alloc(__c);
+    this->__begin_ = __c.__begin_;
+    this->__size_ = __c.__size_;
+    this->__cap() = __c.__cap();
+    __c.__begin_ = nullptr;
+    __c.__cap() = __c.__size_ = 0;
+}
+
+#endif  // !_LIBCUDACXX_CXX03_LANG
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::assign(size_type __n, const value_type& __x)
+{
+    __size_ = 0;
+    if (__n > 0)
+    {
+        size_type __c = capacity();
+        if (__n <= __c)
+            __size_ = __n;
+        else
+        {
+            vector __v(__alloc());
+            __v.reserve(__recommend(__n));
+            __v.__size_ = __n;
+            swap(__v);
+        }
+        _CUDA_VSTD::fill_n(begin(), __n, __x);
+    }
+  __invalidate_all_iterators();
+}
+
+template <class _Allocator>
+template <class _InputIterator>
+typename enable_if
+<
+    __is_input_iterator<_InputIterator>::value &&
+   !__is_forward_iterator<_InputIterator>::value,
+   void
+>::type
+vector<bool, _Allocator>::assign(_InputIterator __first, _InputIterator __last)
+{
+    clear();
+    for (; __first != __last; ++__first)
+        push_back(*__first);
+}
+
+template <class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value,
+   void
+>::type
+vector<bool, _Allocator>::assign(_ForwardIterator __first, _ForwardIterator __last)
+{
+    clear();
+    difference_type __ns = _CUDA_VSTD::distance(__first, __last);
+    _LIBCUDACXX_ASSERT(__ns >= 0, "invalid range specified");
+    const size_t __n = static_cast<size_type>(__ns);
+    if (__n)
+    {
+        if (__n > capacity())
+        {
+            __vdeallocate();
+            __vallocate(__n);
+        }
+        __construct_at_end(__first, __last);
+    }
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::reserve(size_type __n)
+{
+    if (__n > capacity())
+    {
+        vector __v(this->__alloc());
+        __v.__vallocate(__n);
+        __v.__construct_at_end(this->begin(), this->end());
+        swap(__v);
+        __invalidate_all_iterators();
+    }
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::shrink_to_fit() _NOEXCEPT
+{
+    if (__external_cap_to_internal(size()) > __cap())
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            vector(*this, allocator_type(__alloc())).swap(*this);
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+}
+
+template <class _Allocator>
+typename vector<bool, _Allocator>::reference
+vector<bool, _Allocator>::at(size_type __n)
+{
+    if (__n >= size())
+        this->__throw_out_of_range();
+    return (*this)[__n];
+}
+
+template <class _Allocator>
+typename vector<bool, _Allocator>::const_reference
+vector<bool, _Allocator>::at(size_type __n) const
+{
+    if (__n >= size())
+        this->__throw_out_of_range();
+    return (*this)[__n];
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::push_back(const value_type& __x)
+{
+    if (this->__size_ == this->capacity())
+        reserve(__recommend(this->__size_ + 1));
+    ++this->__size_;
+    back() = __x;
+}
+
+template <class _Allocator>
+typename vector<bool, _Allocator>::iterator
+vector<bool, _Allocator>::insert(const_iterator __position, const value_type& __x)
+{
+    iterator __r;
+    if (size() < capacity())
+    {
+        const_iterator __old_end = end();
+        ++__size_;
+        _CUDA_VSTD::copy_backward(__position, __old_end, end());
+        __r = __const_iterator_cast(__position);
+    }
+    else
+    {
+        vector __v(__alloc());
+        __v.reserve(__recommend(__size_ + 1));
+        __v.__size_ = __size_ + 1;
+        __r = _CUDA_VSTD::copy(cbegin(), __position, __v.begin());
+        _CUDA_VSTD::copy_backward(__position, cend(), __v.end());
+        swap(__v);
+    }
+    *__r = __x;
+    return __r;
+}
+
+template <class _Allocator>
+typename vector<bool, _Allocator>::iterator
+vector<bool, _Allocator>::insert(const_iterator __position, size_type __n, const value_type& __x)
+{
+    iterator __r;
+    size_type __c = capacity();
+    if (__n <= __c && size() <= __c - __n)
+    {
+        const_iterator __old_end = end();
+        __size_ += __n;
+        _CUDA_VSTD::copy_backward(__position, __old_end, end());
+        __r = __const_iterator_cast(__position);
+    }
+    else
+    {
+        vector __v(__alloc());
+        __v.reserve(__recommend(__size_ + __n));
+        __v.__size_ = __size_ + __n;
+        __r = _CUDA_VSTD::copy(cbegin(), __position, __v.begin());
+        _CUDA_VSTD::copy_backward(__position, cend(), __v.end());
+        swap(__v);
+    }
+    _CUDA_VSTD::fill_n(__r, __n, __x);
+    return __r;
+}
+
+template <class _Allocator>
+template <class _InputIterator>
+typename enable_if
+<
+     __is_input_iterator  <_InputIterator>::value &&
+    !__is_forward_iterator<_InputIterator>::value,
+    typename vector<bool, _Allocator>::iterator
+>::type
+vector<bool, _Allocator>::insert(const_iterator __position, _InputIterator __first, _InputIterator __last)
+{
+    difference_type __off = __position - begin();
+    iterator __p = __const_iterator_cast(__position);
+    iterator __old_end = end();
+    for (; size() != capacity() && __first != __last; ++__first)
+    {
+        ++this->__size_;
+        back() = *__first;
+    }
+    vector __v(__alloc());
+    if (__first != __last)
+    {
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        try
+        {
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+            __v.assign(__first, __last);
+            difference_type __old_size = static_cast<difference_type>(__old_end - begin());
+            difference_type __old_p = __p - begin();
+            reserve(__recommend(size() + __v.size()));
+            __p = begin() + __old_p;
+            __old_end = begin() + __old_size;
+#ifndef _LIBCUDACXX_NO_EXCEPTIONS
+        }
+        catch (...)
+        {
+            erase(__old_end, end());
+            throw;
+        }
+#endif  // _LIBCUDACXX_NO_EXCEPTIONS
+    }
+    __p = _CUDA_VSTD::rotate(__p, __old_end, end());
+    insert(__p, __v.begin(), __v.end());
+    return begin() + __off;
+}
+
+template <class _Allocator>
+template <class _ForwardIterator>
+typename enable_if
+<
+    __is_forward_iterator<_ForwardIterator>::value,
+    typename vector<bool, _Allocator>::iterator
+>::type
+vector<bool, _Allocator>::insert(const_iterator __position, _ForwardIterator __first, _ForwardIterator __last)
+{
+    const difference_type __n_signed = _CUDA_VSTD::distance(__first, __last);
+    _LIBCUDACXX_ASSERT(__n_signed >= 0, "invalid range specified");
+    const size_type __n = static_cast<size_type>(__n_signed);
+    iterator __r;
+    size_type __c = capacity();
+    if (__n <= __c && size() <= __c - __n)
+    {
+        const_iterator __old_end = end();
+        __size_ += __n;
+        _CUDA_VSTD::copy_backward(__position, __old_end, end());
+        __r = __const_iterator_cast(__position);
+    }
+    else
+    {
+        vector __v(__alloc());
+        __v.reserve(__recommend(__size_ + __n));
+        __v.__size_ = __size_ + __n;
+        __r = _CUDA_VSTD::copy(cbegin(), __position, __v.begin());
+        _CUDA_VSTD::copy_backward(__position, cend(), __v.end());
+        swap(__v);
+    }
+    _CUDA_VSTD::copy(__first, __last, __r);
+    return __r;
+}
+
+template <class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+typename vector<bool, _Allocator>::iterator
+vector<bool, _Allocator>::erase(const_iterator __position)
+{
+    iterator __r = __const_iterator_cast(__position);
+    _CUDA_VSTD::copy(__position + 1, this->cend(), __r);
+    --__size_;
+    return __r;
+}
+
+template <class _Allocator>
+typename vector<bool, _Allocator>::iterator
+vector<bool, _Allocator>::erase(const_iterator __first, const_iterator __last)
+{
+    iterator __r = __const_iterator_cast(__first);
+    difference_type __d = __last - __first;
+    _CUDA_VSTD::copy(__last, this->cend(), __r);
+    __size_ -= __d;
+    return __r;
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::swap(vector& __x)
+#if _LIBCUDACXX_STD_VER >= 14
+    _NOEXCEPT
+#else
+    _NOEXCEPT_(!__alloc_traits::propagate_on_container_swap::value ||
+                __is_nothrow_swappable<allocator_type>::value)
+#endif
+{
+    _CUDA_VSTD::swap(this->__begin_, __x.__begin_);
+    _CUDA_VSTD::swap(this->__size_, __x.__size_);
+    _CUDA_VSTD::swap(this->__cap(), __x.__cap());
+    __swap_allocator(this->__alloc(), __x.__alloc(),
+        integral_constant<bool, __alloc_traits::propagate_on_container_swap::value>());
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::resize(size_type __sz, value_type __x)
+{
+    size_type __cs = size();
+    if (__cs < __sz)
+    {
+        iterator __r;
+        size_type __c = capacity();
+        size_type __n = __sz - __cs;
+        if (__n <= __c && __cs <= __c - __n)
+        {
+            __r = end();
+            __size_ += __n;
+        }
+        else
+        {
+            vector __v(__alloc());
+            __v.reserve(__recommend(__size_ + __n));
+            __v.__size_ = __size_ + __n;
+            __r = _CUDA_VSTD::copy(cbegin(), cend(), __v.begin());
+            swap(__v);
+        }
+        _CUDA_VSTD::fill_n(__r, __n, __x);
+    }
+    else
+        __size_ = __sz;
+}
+
+template <class _Allocator>
+void
+vector<bool, _Allocator>::flip() _NOEXCEPT
+{
+    // do middle whole words
+    size_type __n = __size_;
+    __storage_pointer __p = __begin_;
+    for (; __n >= __bits_per_word; ++__p, __n -= __bits_per_word)
+        *__p = ~*__p;
+    // do last partial word
+    if (__n > 0)
+    {
+        __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __storage_type __b = *__p & __m;
+        *__p &= ~__m;
+        *__p |= ~__b & __m;
+    }
+}
+
+template <class _Allocator>
+bool
+vector<bool, _Allocator>::__invariants() const
+{
+    if (this->__begin_ == nullptr)
+    {
+        if (this->__size_ != 0 || this->__cap() != 0)
+            return false;
+    }
+    else
+    {
+        if (this->__cap() == 0)
+            return false;
+        if (this->__size_ > this->capacity())
+            return false;
+    }
+    return true;
+}
+
+template <class _Allocator>
+size_t
+vector<bool, _Allocator>::__hash_code() const _NOEXCEPT
+{
+    size_t __h = 0;
+    // do middle whole words
+    size_type __n = __size_;
+    __storage_pointer __p = __begin_;
+    for (; __n >= __bits_per_word; ++__p, __n -= __bits_per_word)
+        __h ^= *__p;
+    // do last partial word
+    if (__n > 0)
+    {
+        const __storage_type __m = ~__storage_type(0) >> (__bits_per_word - __n);
+        __h ^= *__p & __m;
+    }
+    return __h;
+}
+
+template <class _Allocator>
+struct _LIBCUDACXX_TEMPLATE_VIS hash<vector<bool, _Allocator> >
+    : public unary_function<vector<bool, _Allocator>, size_t>
+{
+    _LIBCUDACXX_INLINE_VISIBILITY
+    size_t operator()(const vector<bool, _Allocator>& __vec) const _NOEXCEPT
+        {return __vec.__hash_code();}
+};
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator==(const vector<_Tp, _Allocator>& __x, const vector<_Tp, _Allocator>& __y)
+{
+    const typename vector<_Tp, _Allocator>::size_type __sz = __x.size();
+    return __sz == __y.size() && _CUDA_VSTD::equal(__x.begin(), __x.end(), __y.begin());
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator!=(const vector<_Tp, _Allocator>& __x, const vector<_Tp, _Allocator>& __y)
+{
+    return !(__x == __y);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator< (const vector<_Tp, _Allocator>& __x, const vector<_Tp, _Allocator>& __y)
+{
+    return _CUDA_VSTD::lexicographical_compare(__x.begin(), __x.end(), __y.begin(), __y.end());
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator> (const vector<_Tp, _Allocator>& __x, const vector<_Tp, _Allocator>& __y)
+{
+    return __y < __x;
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator>=(const vector<_Tp, _Allocator>& __x, const vector<_Tp, _Allocator>& __y)
+{
+    return !(__x < __y);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+bool
+operator<=(const vector<_Tp, _Allocator>& __x, const vector<_Tp, _Allocator>& __y)
+{
+    return !(__y < __x);
+}
+
+template <class _Tp, class _Allocator>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void
+swap(vector<_Tp, _Allocator>& __x, vector<_Tp, _Allocator>& __y)
+    _NOEXCEPT_(_NOEXCEPT_(__x.swap(__y)))
+{
+    __x.swap(__y);
+}
+
+#if _LIBCUDACXX_STD_VER > 17
+template <class _Tp, class _Allocator, class _Up>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase(vector<_Tp, _Allocator>& __c, const _Up& __v)
+{ __c.erase(_CUDA_VSTD::remove(__c.begin(), __c.end(), __v), __c.end()); }
+
+template <class _Tp, class _Allocator, class _Predicate>
+inline _LIBCUDACXX_INLINE_VISIBILITY
+void erase_if(vector<_Tp, _Allocator>& __c, _Predicate __pred)
+{ __c.erase(_CUDA_VSTD::remove_if(__c.begin(), __c.end(), __pred), __c.end()); }
+#endif
+
+_LIBCUDACXX_END_NAMESPACE_STD
+
+_LIBCUDACXX_POP_MACROS
+
+#endif  // _LIBCUDACXX_VECTOR
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/version b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/version
new file mode 100644
index 000000000000..736c4343f571
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/version
@@ -0,0 +1,242 @@
+// -*- C++ -*-
+//===--------------------------- version ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_VERSIONH
+#define _LIBCUDACXX_VERSIONH
+
+/*
+  version synopsis
+
+Macro name                                              Value   Headers
+__cpp_lib_addressof_constexpr                           201603L <memory>
+__cpp_lib_allocator_traits_is_always_equal              201411L <memory> <scoped_allocator> <string>
+                                                                <deque> <forward_list> <list>
+                                                                <vector> <map> <set>
+                                                                <unordered_map> <unordered_set>
+__cpp_lib_any                                           201606L <any>
+__cpp_lib_apply                                         201603L <tuple>
+__cpp_lib_array_constexpr                               201603L <iterator> <array>
+__cpp_lib_as_const                                      201510L <utility>
+__cpp_lib_atomic_is_always_lock_free                    201603L <atomic>
+__cpp_lib_atomic_ref                                    201806L <atomic>
+__cpp_lib_bind_front                                    201811L <functional>
+__cpp_lib_bit_cast                                      201806L <bit>
+__cpp_lib_bool_constant                                 201505L <type_traits>
+__cpp_lib_boyer_moore_searcher                          201603L <functional>
+__cpp_lib_byte                                          201603L <cstddef>
+__cpp_lib_char8_t                                       201811L <atomic> <filesystem> <istream>
+                                                                <limits> <locale> <ostream>
+                                                                <string> <string_view>
+__cpp_lib_chrono                                        201611L <chrono>
+__cpp_lib_chrono_udls                                   201304L <chrono>
+__cpp_lib_clamp                                         201603L <algorithm>
+__cpp_lib_complex_udls                                  201309L <complex>
+__cpp_lib_concepts                                      201806L <concepts>
+__cpp_lib_constexpr_misc                                201811L <array> <functional> <iterator>
+                                                                <string_view> <tuple> <utility>
+__cpp_lib_constexpr_swap_algorithms                     201806L <algorithm>
+__cpp_lib_destroying_delete                             201806L <new>
+__cpp_lib_enable_shared_from_this                       201603L <memory>
+__cpp_lib_erase_if                                      201811L <string> <deque> <forward_list>
+                                                                <list> <vector> <map>
+                                                                <set> <unordered_map> <unordered_set>
+__cpp_lib_exchange_function                             201304L <utility>
+__cpp_lib_execution                                     201603L <execution>
+__cpp_lib_filesystem                                    201703L <filesystem>
+__cpp_lib_gcd_lcm                                       201606L <numeric>
+__cpp_lib_generic_associative_lookup                    201304L <map> <set>
+__cpp_lib_generic_unordered_lookup                      201811L <unordered_map> <unordered_set>
+__cpp_lib_hardware_interference_size                    201703L <new>
+__cpp_lib_has_unique_object_representations             201606L <type_traits>
+__cpp_lib_hypot                                         201603L <cmath>
+__cpp_lib_incomplete_container_elements                 201505L <forward_list> <list> <vector>
+__cpp_lib_integer_sequence                              201304L <utility>
+__cpp_lib_integral_constant_callable                    201304L <type_traits>
+__cpp_lib_interpolate                                   201902L <numeric>
+__cpp_lib_invoke                                        201411L <functional>
+__cpp_lib_is_aggregate                                  201703L <type_traits>
+__cpp_lib_is_constant_evaluated                         201811L <type_traits>
+__cpp_lib_is_final                                      201402L <type_traits>
+__cpp_lib_is_invocable                                  201703L <type_traits>
+__cpp_lib_is_null_pointer                               201309L <type_traits>
+__cpp_lib_is_swappable                                  201603L <type_traits>
+__cpp_lib_launder                                       201606L <new>
+__cpp_lib_list_remove_return_type                       201806L <forward_list> <list>
+__cpp_lib_logical_traits                                201510L <type_traits>
+__cpp_lib_make_from_tuple                               201606L <tuple>
+__cpp_lib_make_reverse_iterator                         201402L <iterator>
+__cpp_lib_make_unique                                   201304L <memory>
+__cpp_lib_map_try_emplace                               201411L <map>
+__cpp_lib_math_special_functions                        201603L <cmath>
+__cpp_lib_memory_resource                               201603L <memory_resource>
+__cpp_lib_node_extract                                  201606L <map> <set> <unordered_map>
+                                                                <unordered_set>
+__cpp_lib_nonmember_container_access                    201411L <iterator> <array> <deque>
+                                                                <forward_list> <list> <map>
+                                                                <regex> <set> <string>
+                                                                <unordered_map> <unordered_set> <vector>
+__cpp_lib_not_fn                                        201603L <functional>
+__cpp_lib_null_iterators                                201304L <iterator>
+__cpp_lib_optional                                      201606L <optional>
+__cpp_lib_parallel_algorithm                            201603L <algorithm> <numeric>
+__cpp_lib_quoted_string_io                              201304L <iomanip>
+__cpp_lib_ranges                                        201811L <algorithm> <functional> <iterator>
+                                                                <memory> <ranges>
+__cpp_lib_raw_memory_algorithms                         201606L <memory>
+__cpp_lib_result_of_sfinae                              201210L <functional> <type_traits>
+__cpp_lib_robust_nonmodifying_seq_ops                   201304L <algorithm>
+__cpp_lib_sample                                        201603L <algorithm>
+__cpp_lib_scoped_lock                                   201703L <mutex>
+__cpp_lib_shared_mutex                                  201505L <shared_mutex>
+__cpp_lib_shared_ptr_arrays                             201611L <memory>
+__cpp_lib_shared_ptr_weak_type                          201606L <memory>
+__cpp_lib_shared_timed_mutex                            201402L <shared_mutex>
+__cpp_lib_string_udls                                   201304L <string>
+__cpp_lib_string_view                                   201606L <string> <string_view>
+__cpp_lib_three_way_comparison                          201711L <compare>
+__cpp_lib_to_chars                                      201611L <utility>
+__cpp_lib_transformation_trait_aliases                  201304L <type_traits>
+__cpp_lib_transparent_operators                         201510L <functional>
+                                                        201210L // C++14
+__cpp_lib_tuple_element_t                               201402L <tuple>
+__cpp_lib_tuples_by_type                                201304L <utility> <tuple>
+__cpp_lib_type_trait_variable_templates                 201510L <type_traits>
+__cpp_lib_uncaught_exceptions                           201411L <exception>
+__cpp_lib_unordered_map_try_emplace                     201411L <unordered_map>
+__cpp_lib_variant                                       201606L <variant>
+__cpp_lib_void_t                                        201411L <type_traits>
+
+*/
+
+#ifndef __cuda_std__
+#include <__config>
+#include <__pragma_push>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#if _LIBCUDACXX_STD_VER > 11
+# define __cpp_lib_chrono_udls                          201304L
+# define __cpp_lib_complex_udls                         201309L
+# define __cpp_lib_exchange_function                    201304L
+# define __cpp_lib_generic_associative_lookup           201304L
+# define __cpp_lib_integer_sequence                     201304L
+# define __cpp_lib_integral_constant_callable           201304L
+# define __cpp_lib_is_final                             201402L
+# define __cpp_lib_is_null_pointer                      201309L
+# define __cpp_lib_make_reverse_iterator                201402L
+# define __cpp_lib_make_unique                          201304L
+# define __cpp_lib_null_iterators                       201304L
+# define __cpp_lib_quoted_string_io                     201304L
+# define __cpp_lib_result_of_sfinae                     201210L
+# define __cpp_lib_robust_nonmodifying_seq_ops          201304L
+# if !defined(_LIBCUDACXX_HAS_NO_THREADS)
+#   define __cpp_lib_shared_timed_mutex                 201402L
+# endif
+# define __cpp_lib_string_udls                          201304L
+# define __cpp_lib_transformation_trait_aliases         201304L
+# define __cpp_lib_transparent_operators                201210L
+# define __cpp_lib_tuple_element_t                      201402L
+# define __cpp_lib_tuples_by_type                       201304L
+#endif
+
+#if _LIBCUDACXX_STD_VER > 14
+# if defined(_LIBCUDACXX_ADDRESSOF)
+#   define __cpp_lib_addressof_constexpr                201603L
+# endif
+# define __cpp_lib_allocator_traits_is_always_equal     201411L
+# define __cpp_lib_any                                  201606L
+# define __cpp_lib_apply                                201603L
+# define __cpp_lib_array_constexpr                      201603L
+# define __cpp_lib_as_const                             201510L
+# if !defined(_LIBCUDACXX_HAS_NO_THREADS)
+#   define __cpp_lib_atomic_is_always_lock_free         201603L
+# endif
+# define __cpp_lib_bool_constant                        201505L
+// # define __cpp_lib_boyer_moore_searcher                 201603L
+# define __cpp_lib_byte                                 201603L
+# define __cpp_lib_chrono                               201611L
+# define __cpp_lib_clamp                                201603L
+# define __cpp_lib_enable_shared_from_this              201603L
+// # define __cpp_lib_execution                            201603L
+# define __cpp_lib_filesystem                           201703L
+# define __cpp_lib_gcd_lcm                              201606L
+# define __cpp_lib_hardware_interference_size           201703L
+# if defined(_LIBCUDACXX_HAS_UNIQUE_OBJECT_REPRESENTATIONS)
+#   define __cpp_lib_has_unique_object_representations  201606L
+# endif
+# define __cpp_lib_hypot                                201603L
+# define __cpp_lib_incomplete_container_elements        201505L
+# define __cpp_lib_invoke                               201411L
+# if !defined(_LIBCUDACXX_HAS_NO_IS_AGGREGATE)
+#   define __cpp_lib_is_aggregate                       201703L
+# endif
+# define __cpp_lib_is_invocable                         201703L
+# define __cpp_lib_is_swappable                         201603L
+# define __cpp_lib_launder                              201606L
+# define __cpp_lib_logical_traits                       201510L
+# define __cpp_lib_make_from_tuple                      201606L
+# define __cpp_lib_map_try_emplace                      201411L
+// # define __cpp_lib_math_special_functions               201603L
+// # define __cpp_lib_memory_resource                      201603L
+# define __cpp_lib_node_extract                         201606L
+# define __cpp_lib_nonmember_container_access           201411L
+# define __cpp_lib_not_fn                               201603L
+# define __cpp_lib_optional                             201606L
+// # define __cpp_lib_parallel_algorithm                   201603L
+# define __cpp_lib_raw_memory_algorithms                201606L
+# define __cpp_lib_sample                               201603L
+# define __cpp_lib_scoped_lock                          201703L
+# if !defined(_LIBCUDACXX_HAS_NO_THREADS)
+#   define __cpp_lib_shared_mutex                       201505L
+# endif
+// # define __cpp_lib_shared_ptr_arrays                    201611L
+# define __cpp_lib_shared_ptr_weak_type                 201606L
+# define __cpp_lib_string_view                          201606L
+// # define __cpp_lib_to_chars                             201611L
+# undef  __cpp_lib_transparent_operators
+# define __cpp_lib_transparent_operators                201510L
+# define __cpp_lib_type_trait_variable_templates        201510L
+# define __cpp_lib_uncaught_exceptions                  201411L
+# define __cpp_lib_unordered_map_try_emplace            201411L
+# define __cpp_lib_variant                              201606L
+# define __cpp_lib_void_t                               201411L
+#endif
+
+#if _LIBCUDACXX_STD_VER > 17
+# if !defined(_LIBCUDACXX_HAS_NO_THREADS)
+// #   define __cpp_lib_atomic_ref                         201806L
+# endif
+// # define __cpp_lib_bind_front                           201811L
+// # define __cpp_lib_bit_cast                             201806L
+# if !defined(_LIBCUDACXX_NO_HAS_CHAR8_T)
+#   define __cpp_lib_char8_t                            201811L
+# endif
+// # define __cpp_lib_concepts                             201806L
+// # define __cpp_lib_constexpr_misc                       201811L
+// # define __cpp_lib_constexpr_swap_algorithms            201806L
+# if _LIBCUDACXX_STD_VER > 17 && defined(__cpp_impl_destroying_delete) && __cpp_impl_destroying_delete >= 201806L
+#   define __cpp_lib_destroying_delete                  201806L
+# endif
+# define __cpp_lib_erase_if                             201811L
+// # define __cpp_lib_generic_unordered_lookup             201811L
+# define __cpp_lib_interpolate                          201902L
+# if defined(_LIBCUDACXX_IS_CONSTANT_EVALUATED)
+#   define __cpp_lib_is_constant_evaluated              201811L
+# endif
+// # define __cpp_lib_list_remove_return_type              201806L
+// # define __cpp_lib_ranges                               201811L
+// # define __cpp_lib_three_way_comparison                 201711L
+#endif
+
+#include <__pragma_pop>
+#endif
+
+#endif // _LIBCUDACXX_VERSIONH
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h
new file mode 100644
index 000000000000..999837c38932
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wchar.h
@@ -0,0 +1,181 @@
+// -*- C++ -*-
+//===--------------------------- wchar.h ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__need_wint_t) || defined(__need_mbstate_t)
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <wchar.h>
+
+#elif !defined(_LIBCUDACXX_WCHAR_H)
+#define _LIBCUDACXX_WCHAR_H
+
+/*
+    wchar.h synopsis
+
+Macros:
+
+    NULL
+    WCHAR_MAX
+    WCHAR_MIN
+    WEOF
+
+Types:
+
+    mbstate_t
+    size_t
+    tm
+    wint_t
+
+int fwprintf(FILE* restrict stream, const wchar_t* restrict format, ...);
+int fwscanf(FILE* restrict stream, const wchar_t* restrict format, ...);
+int swprintf(wchar_t* restrict s, size_t n, const wchar_t* restrict format, ...);
+int swscanf(const wchar_t* restrict s, const wchar_t* restrict format, ...);
+int vfwprintf(FILE* restrict stream, const wchar_t* restrict format, va_list arg);
+int vfwscanf(FILE* restrict stream, const wchar_t* restrict format, va_list arg);  // C99
+int vswprintf(wchar_t* restrict s, size_t n, const wchar_t* restrict format, va_list arg);
+int vswscanf(const wchar_t* restrict s, const wchar_t* restrict format, va_list arg);  // C99
+int vwprintf(const wchar_t* restrict format, va_list arg);
+int vwscanf(const wchar_t* restrict format, va_list arg);  // C99
+int wprintf(const wchar_t* restrict format, ...);
+int wscanf(const wchar_t* restrict format, ...);
+wint_t fgetwc(FILE* stream);
+wchar_t* fgetws(wchar_t* restrict s, int n, FILE* restrict stream);
+wint_t fputwc(wchar_t c, FILE* stream);
+int fputws(const wchar_t* restrict s, FILE* restrict stream);
+int fwide(FILE* stream, int mode);
+wint_t getwc(FILE* stream);
+wint_t getwchar();
+wint_t putwc(wchar_t c, FILE* stream);
+wint_t putwchar(wchar_t c);
+wint_t ungetwc(wint_t c, FILE* stream);
+double wcstod(const wchar_t* restrict nptr, wchar_t** restrict endptr);
+float wcstof(const wchar_t* restrict nptr, wchar_t** restrict endptr);         // C99
+long double wcstold(const wchar_t* restrict nptr, wchar_t** restrict endptr);  // C99
+long wcstol(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+long long wcstoll(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);  // C99
+unsigned long wcstoul(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);
+unsigned long long wcstoull(const wchar_t* restrict nptr, wchar_t** restrict endptr, int base);  // C99
+wchar_t* wcscpy(wchar_t* restrict s1, const wchar_t* restrict s2);
+wchar_t* wcsncpy(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+wchar_t* wcscat(wchar_t* restrict s1, const wchar_t* restrict s2);
+wchar_t* wcsncat(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+int wcscmp(const wchar_t* s1, const wchar_t* s2);
+int wcscoll(const wchar_t* s1, const wchar_t* s2);
+int wcsncmp(const wchar_t* s1, const wchar_t* s2, size_t n);
+size_t wcsxfrm(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+const wchar_t* wcschr(const wchar_t* s, wchar_t c);
+      wchar_t* wcschr(      wchar_t* s, wchar_t c);
+size_t wcscspn(const wchar_t* s1, const wchar_t* s2);
+size_t wcslen(const wchar_t* s);
+const wchar_t* wcspbrk(const wchar_t* s1, const wchar_t* s2);
+      wchar_t* wcspbrk(      wchar_t* s1, const wchar_t* s2);
+const wchar_t* wcsrchr(const wchar_t* s, wchar_t c);
+      wchar_t* wcsrchr(      wchar_t* s, wchar_t c);
+size_t wcsspn(const wchar_t* s1, const wchar_t* s2);
+const wchar_t* wcsstr(const wchar_t* s1, const wchar_t* s2);
+      wchar_t* wcsstr(      wchar_t* s1, const wchar_t* s2);
+wchar_t* wcstok(wchar_t* restrict s1, const wchar_t* restrict s2, wchar_t** restrict ptr);
+const wchar_t* wmemchr(const wchar_t* s, wchar_t c, size_t n);
+      wchar_t* wmemchr(      wchar_t* s, wchar_t c, size_t n);
+int wmemcmp(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+wchar_t* wmemcpy(wchar_t* restrict s1, const wchar_t* restrict s2, size_t n);
+wchar_t* wmemmove(wchar_t* s1, const wchar_t* s2, size_t n);
+wchar_t* wmemset(wchar_t* s, wchar_t c, size_t n);
+size_t wcsftime(wchar_t* restrict s, size_t maxsize, const wchar_t* restrict format,
+                const tm* restrict timeptr);
+wint_t btowc(int c);
+int wctob(wint_t c);
+int mbsinit(const mbstate_t* ps);
+size_t mbrlen(const char* restrict s, size_t n, mbstate_t* restrict ps);
+size_t mbrtowc(wchar_t* restrict pwc, const char* restrict s, size_t n, mbstate_t* restrict ps);
+size_t wcrtomb(char* restrict s, wchar_t wc, mbstate_t* restrict ps);
+size_t mbsrtowcs(wchar_t* restrict dst, const char** restrict src, size_t len,
+                 mbstate_t* restrict ps);
+size_t wcsrtombs(char* restrict dst, const wchar_t** restrict src, size_t len,
+                 mbstate_t* restrict ps);
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#ifdef __cplusplus
+#define __CORRECT_ISO_CPP_WCHAR_H_PROTO
+#endif
+
+#include_next <wchar.h>
+
+// Determine whether we have const-correct overloads for wcschr and friends.
+#if defined(_WCHAR_H_CPLUSPLUS_98_CONFORMANCE_)
+#  define _LIBCUDACXX_WCHAR_H_HAS_CONST_OVERLOADS 1
+#elif defined(__GLIBC_PREREQ)
+#  if __GLIBC_PREREQ(2, 10)
+#    define _LIBCUDACXX_WCHAR_H_HAS_CONST_OVERLOADS 1
+#  endif
+#elif defined(_LIBCUDACXX_MSVCRT)
+#  if defined(_CRT_CONST_CORRECT_OVERLOADS)
+#    define _LIBCUDACXX_WCHAR_H_HAS_CONST_OVERLOADS 1
+#  endif
+#endif
+
+#if defined(__cplusplus) && !defined(_LIBCUDACXX_WCHAR_H_HAS_CONST_OVERLOADS) && defined(_LIBCUDACXX_PREFERRED_OVERLOAD)
+extern "C++" {
+inline _LIBCUDACXX_INLINE_VISIBILITY
+wchar_t* __libcpp_wcschr(const wchar_t* __s, wchar_t __c) {return (wchar_t*)wcschr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const wchar_t* wcschr(const wchar_t* __s, wchar_t __c) {return __libcpp_wcschr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      wchar_t* wcschr(      wchar_t* __s, wchar_t __c) {return __libcpp_wcschr(__s, __c);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+wchar_t* __libcpp_wcspbrk(const wchar_t* __s1, const wchar_t* __s2) {return (wchar_t*)wcspbrk(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const wchar_t* wcspbrk(const wchar_t* __s1, const wchar_t* __s2) {return __libcpp_wcspbrk(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      wchar_t* wcspbrk(      wchar_t* __s1, const wchar_t* __s2) {return __libcpp_wcspbrk(__s1, __s2);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+wchar_t* __libcpp_wcsrchr(const wchar_t* __s, wchar_t __c) {return (wchar_t*)wcsrchr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const wchar_t* wcsrchr(const wchar_t* __s, wchar_t __c) {return __libcpp_wcsrchr(__s, __c);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      wchar_t* wcsrchr(      wchar_t* __s, wchar_t __c) {return __libcpp_wcsrchr(__s, __c);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+wchar_t* __libcpp_wcsstr(const wchar_t* __s1, const wchar_t* __s2) {return (wchar_t*)wcsstr(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const wchar_t* wcsstr(const wchar_t* __s1, const wchar_t* __s2) {return __libcpp_wcsstr(__s1, __s2);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      wchar_t* wcsstr(      wchar_t* __s1, const wchar_t* __s2) {return __libcpp_wcsstr(__s1, __s2);}
+
+inline _LIBCUDACXX_INLINE_VISIBILITY
+wchar_t* __libcpp_wmemchr(const wchar_t* __s, wchar_t __c, size_t __n) {return (wchar_t*)wmemchr(__s, __c, __n);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+const wchar_t* wmemchr(const wchar_t* __s, wchar_t __c, size_t __n) {return __libcpp_wmemchr(__s, __c, __n);}
+inline _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_PREFERRED_OVERLOAD
+      wchar_t* wmemchr(      wchar_t* __s, wchar_t __c, size_t __n) {return __libcpp_wmemchr(__s, __c, __n);}
+}
+#endif
+
+#if defined(__cplusplus) && defined(_LIBCUDACXX_MSVCRT_LIKE)
+extern "C" {
+size_t mbsnrtowcs(wchar_t *__restrict dst, const char **__restrict src,
+                  size_t nmc, size_t len, mbstate_t *__restrict ps);
+size_t wcsnrtombs(char *__restrict dst, const wchar_t **__restrict src,
+                  size_t nwc, size_t len, mbstate_t *__restrict ps);
+}  // extern "C++"
+#endif  // __cplusplus && _LIBCUDACXX_MSVCRT
+
+#endif  // _LIBCUDACXX_WCHAR_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h
new file mode 100644
index 000000000000..ad4a88a38aeb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/detail/libcxx/include/wctype.h
@@ -0,0 +1,78 @@
+// -*- C++ -*-
+//===--------------------------- wctype.h ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCUDACXX_WCTYPE_H
+#define _LIBCUDACXX_WCTYPE_H
+
+/*
+    wctype.h synopsis
+
+Macros:
+
+    WEOF
+
+Types:
+
+    wint_t
+    wctrans_t
+    wctype_t
+
+int iswalnum(wint_t wc);
+int iswalpha(wint_t wc);
+int iswblank(wint_t wc);  // C99
+int iswcntrl(wint_t wc);
+int iswdigit(wint_t wc);
+int iswgraph(wint_t wc);
+int iswlower(wint_t wc);
+int iswprint(wint_t wc);
+int iswpunct(wint_t wc);
+int iswspace(wint_t wc);
+int iswupper(wint_t wc);
+int iswxdigit(wint_t wc);
+int iswctype(wint_t wc, wctype_t desc);
+wctype_t wctype(const char* property);
+wint_t towlower(wint_t wc);
+wint_t towupper(wint_t wc);
+wint_t towctrans(wint_t wc, wctrans_t desc);
+wctrans_t wctrans(const char* property);
+
+*/
+
+#include <__config>
+
+#if defined(_LIBCUDACXX_USE_PRAGMA_GCC_SYSTEM_HEADER)
+#pragma GCC system_header
+#endif
+
+#include_next <wctype.h>
+
+#ifdef __cplusplus
+
+#undef iswalnum
+#undef iswalpha
+#undef iswblank
+#undef iswcntrl
+#undef iswdigit
+#undef iswgraph
+#undef iswlower
+#undef iswprint
+#undef iswpunct
+#undef iswspace
+#undef iswupper
+#undef iswxdigit
+#undef iswctype
+#undef wctype
+#undef towlower
+#undef towupper
+#undef towctrans
+#undef wctrans
+
+#endif  // __cplusplus
+
+#endif  // _LIBCUDACXX_WCTYPE_H
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/functional b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/functional
new file mode 100644
index 000000000000..4a4df5213c5d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/functional
@@ -0,0 +1,32 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_FUNCTIONAL
+#define _CUDA_FUNCTIONAL
+
+#include "type_traits"
+#include "version"
+#include "iterator"
+// #include "memory"
+// #include "tuple"
+#include "utility"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/__functional_base"
+
+#include "detail/libcxx/include/functional"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_FUNCTIONAL
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/initializer_list b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/initializer_list
new file mode 100644
index 000000000000..f9e6e2af70b0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/initializer_list
@@ -0,0 +1,25 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_INITIALIZER_LIST
+#define _CUDA_INITIALIZER_LIST
+
+#include "detail/__config"
+
+#if !defined(_LIBCUDACXX_COMPILER_NVRTC)
+#include <initializer_list>
+#endif
+
+_LIBCUDACXX_BEGIN_NAMESPACE_STD
+    using ::std::initializer_list;
+_LIBCUDACXX_END_NAMESPACE_STD
+
+
+#endif //_CUDA_INITIALIZER_LIST
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/iterator b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/iterator
new file mode 100644
index 000000000000..7cf78ea3857e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/iterator
@@ -0,0 +1,29 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_ITERATOR
+#define _CUDA_ITERATOR
+
+#include "type_traits"
+
+#if !defined(_LIBCUDACXX_COMPILER_NVRTC)
+#include <initializer_list>
+#endif
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/iosfwd"
+#include "detail/libcxx/include/iterator"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_ITERATOR
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/latch b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/latch
new file mode 100644
index 000000000000..0bb4c4f27ae6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/latch
@@ -0,0 +1,40 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 700
+#  error "CUDA synchronization primitives are only supported for sm_70 and up."
+#endif
+
+#ifndef _CUDA_LATCH
+#define _CUDA_LATCH
+
+#include "atomic"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/latch"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+template<thread_scope _Sco>
+class latch : public std::__latch_base<_Sco> {
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    latch(std::ptrdiff_t __count)
+        : std::__latch_base<_Sco>(__count) {
+    }
+};
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_LATCH
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/limits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/limits
new file mode 100644
index 000000000000..c79cff1860b8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/limits
@@ -0,0 +1,29 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_LIMITS
+#define _CUDA_LIMITS
+
+#ifdef __CUDACC_RTC__
+    #include "climits"
+#endif
+#include "type_traits"
+#include "version"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/limits"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_LIMITS
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ratio b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ratio
new file mode 100644
index 000000000000..459382d3609f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/ratio
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_RATIO
+#define _CUDA_RATIO
+
+#include "cstdint"
+#include "climits"
+#include "type_traits"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/ratio"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_RATIO
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/semaphore b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/semaphore
new file mode 100644
index 000000000000..7a02b4e332bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/semaphore
@@ -0,0 +1,47 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 700
+#  error "CUDA synchronization primitives are only supported for sm_70 and up."
+#endif
+
+#ifndef _CUDA_SEMAPHORE
+#define _CUDA_SEMAPHORE
+
+#include "atomic"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/semaphore"
+
+_LIBCUDACXX_BEGIN_NAMESPACE_CUDA
+
+template<thread_scope _Sco, ptrdiff_t __least_max_value = INT_MAX>
+class counting_semaphore : public std::__semaphore_base<__least_max_value, _Sco>
+{
+    static_assert(__least_max_value <= std::__semaphore_base<__least_max_value, _Sco>::max(), "");
+public:
+    _LIBCUDACXX_INLINE_VISIBILITY _LIBCUDACXX_CONSTEXPR
+    counting_semaphore(ptrdiff_t __count = 0) : std::__semaphore_base<__least_max_value, _Sco>(__count) { }
+    ~counting_semaphore() = default;
+
+    counting_semaphore(const counting_semaphore&) = delete;
+    counting_semaphore& operator=(const counting_semaphore&) = delete;
+};
+
+template<thread_scope _Sco>
+using binary_semaphore = counting_semaphore<_Sco, 1>;
+
+_LIBCUDACXX_END_NAMESPACE_CUDA
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_SEMAPHORE
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/tuple b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/tuple
new file mode 100644
index 000000000000..5161b8e0aa0d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/tuple
@@ -0,0 +1,33 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_TUPLE
+#define _CUDA_TUPLE
+
+#include "cstdint"
+#include "climits"
+
+#include "type_traits"
+
+#include "detail/libcxx/include/__tuple"
+
+#include "functional"
+#include "utility"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/tuple"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_TUPLE
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/type_traits b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/type_traits
new file mode 100644
index 000000000000..e0b073018796
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/type_traits
@@ -0,0 +1,28 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_TYPE_TRAITS
+#define _CUDA_TYPE_TRAITS
+
+#ifndef __CUDACC_RTC__
+    #include <type_traits>
+#endif //__CUDACC_RTC__
+
+#include "cstddef"
+#include "version"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/type_traits"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_TYPE_TRAITS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/utility b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/utility
new file mode 100644
index 000000000000..0a567ab6143a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/utility
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_UTILITY
+#define _CUDA_UTILITY
+
+#include "type_traits"
+#include "version"
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/__tuple"
+#include "detail/libcxx/include/utility"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_UTILITY
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/version b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/version
new file mode 100644
index 000000000000..fd84a8db18b9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/cuda/std/version
@@ -0,0 +1,21 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_VERSION
+#define _CUDA_VERSION
+
+#include "detail/__config"
+
+#include "detail/__pragma_push"
+
+#include "detail/libcxx/include/version"
+
+#include "detail/__pragma_pop"
+
+#endif //_CUDA_VERSION
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__preprocessor b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__preprocessor
new file mode 100644
index 000000000000..d1553a6963ef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__preprocessor
@@ -0,0 +1,117 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#if defined(__GNUC__)
+#pragma GCC system_header
+#endif
+
+// For all compilers and dialects this header defines:
+//  _NV_EVAL
+//  _NV_IF
+//  _NV_CONCAT_EVAL
+// For C++11 and up it defines:
+//  _NV_STRIP_PAREN
+//  _NV_DISPATCH_N_ARY
+//  _NV_FIRST_ARG
+//  _NV_REMOVE_PAREN
+
+#if defined(_NV_TARGET_CPP11)
+#  define _NV_EVAL1(...) __VA_ARGS__
+#  define _NV_EVAL(...) _NV_EVAL1(__VA_ARGS__)
+#else
+#  define _NV_EVAL1(x) x
+#  define _NV_EVAL(x) _NV_EVAL1(x)
+#endif // C++11
+
+#define _NV_CONCAT_EVAL1(l, r) _NV_EVAL(l ## r)
+#define _NV_CONCAT_EVAL(l, r) _NV_CONCAT_EVAL1(l, r)
+
+#define _NV_IF_0(t, f) f
+#define _NV_IF_1(t, f) t
+
+#define _NV_IF_BIT(b) _NV_EVAL(_NV_IF_##b)
+#define _NV_IF__EVAL(fn, t, f) _NV_EVAL(fn(t, f))
+#define _NV_IF_EVAL(cond, t, f) _NV_IF__EVAL(_NV_IF_BIT(cond), t, f)
+
+#define _NV_IF1(cond, t, f) _NV_IF_EVAL(cond, t, f)
+#define _NV_IF(cond, t, f) _NV_IF1(_NV_EVAL(cond), _NV_EVAL(t), _NV_EVAL(f))
+
+#if defined(_NV_TARGET_CPP11)
+
+// The below mechanisms were derived from: https://gustedt.wordpress.com/2010/06/08/detect-empty-macro-arguments/
+
+#define _NV_ARG32(...) _NV_EVAL(_NV_ARG32_0(__VA_ARGS__))
+#define _NV_ARG32_0(                                                                         \
+    _0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15,                    \
+    _16, _17, _18, _19, _20, _21, _22, _23, _24, _25, _26, _27, _28, _29, _30, _31, ...) _31
+
+#define _NV_HAS_COMMA(...) _NV_ARG32(__VA_ARGS__,   \
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, \
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0)
+
+#define _NV_TRIGGER_PARENTHESIS_(...) ,
+
+#define _NV_ISEMPTY(...)                                                    \
+    _NV_ISEMPTY0(                                                           \
+          /* test if there is just one argument, eventually an empty        \
+             one */                                                         \
+          _NV_EVAL(_NV_HAS_COMMA(__VA_ARGS__)),                                       \
+          /* test if _TRIGGER_PARENTHESIS_ together with the argument       \
+             adds a comma */                                                \
+          _NV_EVAL(_NV_HAS_COMMA(_NV_TRIGGER_PARENTHESIS_ __VA_ARGS__)),              \
+          /* test if the argument together with a parenthesis               \
+             adds a comma */                                                \
+          _NV_EVAL(_NV_HAS_COMMA(__VA_ARGS__ (/*empty*/))),                           \
+          /* test if placing it between _TRIGGER_PARENTHESIS_ and the       \
+             parenthesis adds a comma */                                    \
+          _NV_EVAL(_NV_HAS_COMMA(_NV_TRIGGER_PARENTHESIS_ __VA_ARGS__ (/*empty*/)))   \
+          )
+
+#define _NV_PASTE5(_0, _1, _2, _3, _4) _0 ## _1 ## _2 ## _3 ## _4
+#define _NV_ISEMPTY0(_0, _1, _2, _3) _NV_HAS_COMMA(_NV_PASTE5(_NV_IS_EMPTY_CASE_, _0, _1, _2, _3))
+#define _NV_IS_EMPTY_CASE_0001 ,
+
+
+#define _NV_REMOVE_PAREN(...) _NV_REMOVE_PAREN1(__VA_ARGS__)
+#define _NV_REMOVE_PAREN1(...) _NV_STRIP_PAREN(_NV_IF(_NV_TEST_PAREN(__VA_ARGS__), (_NV_STRIP_PAREN(__VA_ARGS__)), (__VA_ARGS__)))
+
+#define _NV_STRIP_PAREN2(...) __VA_ARGS__
+#define _NV_STRIP_PAREN1(...) _NV_STRIP_PAREN2 __VA_ARGS__
+#define _NV_STRIP_PAREN(...) _NV_STRIP_PAREN1(__VA_ARGS__)
+
+#define _NV_TEST_PAREN(...) _NV_TEST_PAREN1(__VA_ARGS__)
+#define _NV_TEST_PAREN1(...) _NV_TEST_PAREN2(_NV_TEST_PAREN_DUMMY __VA_ARGS__)
+#define _NV_TEST_PAREN2(...) _NV_TEST_PAREN3(_NV_CONCAT_EVAL(_, __VA_ARGS__))
+#define _NV_TEST_PAREN3(...) _NV_EVAL(_NV_FIRST_ARG(__VA_ARGS__))
+
+#define __NV_PAREN_YES 1
+#define __NV_PAREN_NO 0
+
+#define _NV_TEST_PAREN_DUMMY(...) _NV_PAREN_YES
+#define __NV_TEST_PAREN_DUMMY     __NV_PAREN_NO,
+
+#define _NV_FIRST_ARG1(x, ...) x
+#define _NV_FIRST_ARG(x, ...) _NV_FIRST_ARG1(x)
+
+#define _NV_REMOVE_FIRST_ARGS1(...) __VA_ARGS__
+#define _NV_REMOVE_FIRST_ARGS(x, ...) _NV_REMOVE_FIRST_ARGS1(__VA_ARGS__)
+
+#define _NV_NUM_ARGS(...) _NV_NUM_ARGS0(__VA_ARGS__)
+#define _NV_NUM_ARGS0(...) _NV_EVAL(_NV_NUM_ARGS1(__VA_ARGS__))
+#define _NV_NUM_ARGS1(...) _NV_IF(_NV_ISEMPTY(__VA_ARGS__), 0, _NV_NUM_ARGS2(__VA_ARGS__))
+#define _NV_NUM_ARGS2(...) _NV_ARG32(__VA_ARGS__,   \
+    31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16, \
+    15,14,13,12,11,10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0)
+
+#define _NV_DISPATCH_N_IMPL1(name, ...) _NV_EVAL(name(__VA_ARGS__))
+#define _NV_DISPATCH_N_IMPL0(depth, name, ...) _NV_DISPATCH_N_IMPL1(_NV_CONCAT_EVAL(name, depth), __VA_ARGS__)
+#define _NV_DISPATCH_N_IMPL(name, ...) _NV_DISPATCH_N_IMPL0(_NV_NUM_ARGS(__VA_ARGS__), name, __VA_ARGS__)
+#define _NV_DISPATCH_N_ARY(name, ...) _NV_DISPATCH_N_IMPL(name, __VA_ARGS__)
+
+#endif // C++11
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__target_macros b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__target_macros
new file mode 100644
index 000000000000..90132918051d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/detail/__target_macros
@@ -0,0 +1,493 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _NV__TARGET_MACROS
+#define _NV__TARGET_MACROS
+
+#include "__preprocessor"
+
+#if defined(__GNUC__)
+#pragma GCC system_header
+#endif
+
+#  define _NV_TARGET_ARCH_TO_SELECTOR_350 nv::target::sm_35
+#  define _NV_TARGET_ARCH_TO_SELECTOR_370 nv::target::sm_37
+#  define _NV_TARGET_ARCH_TO_SELECTOR_500 nv::target::sm_50
+#  define _NV_TARGET_ARCH_TO_SELECTOR_520 nv::target::sm_52
+#  define _NV_TARGET_ARCH_TO_SELECTOR_530 nv::target::sm_53
+#  define _NV_TARGET_ARCH_TO_SELECTOR_600 nv::target::sm_60
+#  define _NV_TARGET_ARCH_TO_SELECTOR_610 nv::target::sm_61
+#  define _NV_TARGET_ARCH_TO_SELECTOR_620 nv::target::sm_62
+#  define _NV_TARGET_ARCH_TO_SELECTOR_700 nv::target::sm_70
+#  define _NV_TARGET_ARCH_TO_SELECTOR_720 nv::target::sm_72
+#  define _NV_TARGET_ARCH_TO_SELECTOR_750 nv::target::sm_75
+#  define _NV_TARGET_ARCH_TO_SELECTOR_800 nv::target::sm_80
+#  define _NV_TARGET_ARCH_TO_SELECTOR_860 nv::target::sm_86
+#  define _NV_TARGET_ARCH_TO_SELECTOR_870 nv::target::sm_87
+#  define _NV_TARGET_ARCH_TO_SELECTOR_900 nv::target::sm_90
+
+#  define _NV_TARGET_ARCH_TO_SM_350 35
+#  define _NV_TARGET_ARCH_TO_SM_370 37
+#  define _NV_TARGET_ARCH_TO_SM_500 50
+#  define _NV_TARGET_ARCH_TO_SM_520 52
+#  define _NV_TARGET_ARCH_TO_SM_530 53
+#  define _NV_TARGET_ARCH_TO_SM_600 60
+#  define _NV_TARGET_ARCH_TO_SM_610 61
+#  define _NV_TARGET_ARCH_TO_SM_620 62
+#  define _NV_TARGET_ARCH_TO_SM_700 70
+#  define _NV_TARGET_ARCH_TO_SM_720 72
+#  define _NV_TARGET_ARCH_TO_SM_750 75
+#  define _NV_TARGET_ARCH_TO_SM_800 80
+#  define _NV_TARGET_ARCH_TO_SM_860 86
+#  define _NV_TARGET_ARCH_TO_SM_870 87
+#  define _NV_TARGET_ARCH_TO_SM_900 90
+
+#if defined(_NV_COMPILER_NVCXX)
+
+#  define _NV_TARGET_VAL_SM_35 nv::target::sm_35
+#  define _NV_TARGET_VAL_SM_37 nv::target::sm_37
+#  define _NV_TARGET_VAL_SM_50 nv::target::sm_50
+#  define _NV_TARGET_VAL_SM_52 nv::target::sm_52
+#  define _NV_TARGET_VAL_SM_53 nv::target::sm_53
+#  define _NV_TARGET_VAL_SM_60 nv::target::sm_60
+#  define _NV_TARGET_VAL_SM_61 nv::target::sm_61
+#  define _NV_TARGET_VAL_SM_62 nv::target::sm_62
+#  define _NV_TARGET_VAL_SM_70 nv::target::sm_70
+#  define _NV_TARGET_VAL_SM_72 nv::target::sm_72
+#  define _NV_TARGET_VAL_SM_75 nv::target::sm_75
+#  define _NV_TARGET_VAL_SM_80 nv::target::sm_80
+#  define _NV_TARGET_VAL_SM_86 nv::target::sm_86
+#  define _NV_TARGET_VAL_SM_87 nv::target::sm_87
+#  define _NV_TARGET_VAL_SM_90 nv::target::sm_90
+
+#  define _NV_TARGET___NV_IS_HOST nv::target::is_host
+#  define _NV_TARGET___NV_IS_DEVICE nv::target::is_device
+
+#  define _NV_TARGET___NV_ANY_TARGET (nv::target::any_target)
+#  define _NV_TARGET___NV_NO_TARGET (nv::target::no_target)
+
+#  if defined(NV_TARGET_SM_INTEGER_LIST)
+#    define NV_TARGET_MINIMUM_SM_SELECTOR _NV_FIRST_ARG(NV_TARGET_SM_SELECTOR_LIST)
+#    define NV_TARGET_MINIMUM_SM_INTEGER _NV_FIRST_ARG(NV_TARGET_SM_INTEGER_LIST)
+#    define __CUDA_MINIMUM_ARCH__ _NV_CONCAT_EVAL(_NV_FIRST_ARG(NV_TARGET_SM_INTEGER_LIST), 0)
+#  endif
+
+#  define _NV_TARGET_PROVIDES(q)   nv::target::provides(q)
+#  define _NV_TARGET_IS_EXACTLY(q) nv::target::is_exactly(q)
+
+#elif defined(_NV_COMPILER_NVCC) || defined (_NV_COMPILER_CLANG_CUDA)
+
+#  define _NV_TARGET_VAL_SM_35 350
+#  define _NV_TARGET_VAL_SM_37 370
+#  define _NV_TARGET_VAL_SM_50 500
+#  define _NV_TARGET_VAL_SM_52 520
+#  define _NV_TARGET_VAL_SM_53 530
+#  define _NV_TARGET_VAL_SM_60 600
+#  define _NV_TARGET_VAL_SM_61 610
+#  define _NV_TARGET_VAL_SM_62 620
+#  define _NV_TARGET_VAL_SM_70 700
+#  define _NV_TARGET_VAL_SM_72 720
+#  define _NV_TARGET_VAL_SM_75 750
+#  define _NV_TARGET_VAL_SM_80 800
+#  define _NV_TARGET_VAL_SM_86 860
+#  define _NV_TARGET_VAL_SM_87 870
+#  define _NV_TARGET_VAL_SM_90 900
+
+#  if defined(__CUDA_ARCH__)
+#    define _NV_TARGET_VAL __CUDA_ARCH__
+#    define NV_TARGET_MINIMUM_SM_SELECTOR _NV_CONCAT_EVAL(_NV_TARGET_ARCH_TO_SELECTOR_, __CUDA_ARCH__)
+#    define NV_TARGET_MINIMUM_SM_INTEGER _NV_CONCAT_EVAL(_NV_TARGET_ARCH_TO_SM_, __CUDA_ARCH__)
+#    define __CUDA_MINIMUM_ARCH__ __CUDA_ARCH__
+#  endif
+
+#  if defined(__CUDA_ARCH__)
+#    define _NV_TARGET_IS_HOST   0
+#    define _NV_TARGET_IS_DEVICE 1
+#  else
+#    define _NV_TARGET_IS_HOST   1
+#    define _NV_TARGET_IS_DEVICE 0
+#  endif
+
+#  if defined(_NV_TARGET_VAL)
+#    define _NV_DEVICE_CHECK(q) (q)
+#  else
+#    define _NV_DEVICE_CHECK(q) (0)
+#  endif
+
+#  define _NV_TARGET_PROVIDES(q)   _NV_DEVICE_CHECK(_NV_TARGET_VAL >= q)
+#  define _NV_TARGET_IS_EXACTLY(q) _NV_DEVICE_CHECK(_NV_TARGET_VAL == q)
+
+// NVCC/NVCXX not being used, only host dispatches allowed
+#else
+
+#  define _NV_COMPILER_NVCC
+
+#  define _NV_TARGET_VAL_SM_35 350
+#  define _NV_TARGET_VAL_SM_37 370
+#  define _NV_TARGET_VAL_SM_50 500
+#  define _NV_TARGET_VAL_SM_52 520
+#  define _NV_TARGET_VAL_SM_53 530
+#  define _NV_TARGET_VAL_SM_60 600
+#  define _NV_TARGET_VAL_SM_61 610
+#  define _NV_TARGET_VAL_SM_62 620
+#  define _NV_TARGET_VAL_SM_70 700
+#  define _NV_TARGET_VAL_SM_72 720
+#  define _NV_TARGET_VAL_SM_75 750
+#  define _NV_TARGET_VAL_SM_80 800
+#  define _NV_TARGET_VAL_SM_86 860
+#  define _NV_TARGET_VAL_SM_87 870
+#  define _NV_TARGET_VAL_SM_90 900
+
+#  define _NV_TARGET_VAL 0
+
+#  define _NV_TARGET_IS_HOST   1
+#  define _NV_TARGET_IS_DEVICE 0
+
+#  define _NV_DEVICE_CHECK(q) (false)
+
+#  define _NV_TARGET_PROVIDES(q)   _NV_DEVICE_CHECK(_NV_TARGET_VAL >= q)
+#  define _NV_TARGET_IS_EXACTLY(q) _NV_DEVICE_CHECK(_NV_TARGET_VAL == q)
+
+#endif
+
+#define _NV_TARGET___NV_PROVIDES_SM_35 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_35))
+#define _NV_TARGET___NV_PROVIDES_SM_37 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_37))
+#define _NV_TARGET___NV_PROVIDES_SM_50 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_50))
+#define _NV_TARGET___NV_PROVIDES_SM_52 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_52))
+#define _NV_TARGET___NV_PROVIDES_SM_53 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_53))
+#define _NV_TARGET___NV_PROVIDES_SM_60 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_60))
+#define _NV_TARGET___NV_PROVIDES_SM_61 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_61))
+#define _NV_TARGET___NV_PROVIDES_SM_62 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_62))
+#define _NV_TARGET___NV_PROVIDES_SM_70 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_70))
+#define _NV_TARGET___NV_PROVIDES_SM_72 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_72))
+#define _NV_TARGET___NV_PROVIDES_SM_75 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_75))
+#define _NV_TARGET___NV_PROVIDES_SM_80 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_80))
+#define _NV_TARGET___NV_PROVIDES_SM_86 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_86))
+#define _NV_TARGET___NV_PROVIDES_SM_87 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_87))
+#define _NV_TARGET___NV_PROVIDES_SM_90 (_NV_TARGET_PROVIDES(_NV_TARGET_VAL_SM_90))
+
+#define _NV_TARGET___NV_IS_EXACTLY_SM_35 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_35))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_37 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_37))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_50 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_50))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_52 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_52))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_53 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_53))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_60 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_60))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_61 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_61))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_62 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_62))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_70 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_70))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_72 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_72))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_75 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_75))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_80 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_80))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_86 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_86))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_87 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_87))
+#define _NV_TARGET___NV_IS_EXACTLY_SM_90 (_NV_TARGET_IS_EXACTLY(_NV_TARGET_VAL_SM_90))
+
+#define NV_PROVIDES_SM_35   __NV_PROVIDES_SM_35
+#define NV_PROVIDES_SM_37   __NV_PROVIDES_SM_37
+#define NV_PROVIDES_SM_50   __NV_PROVIDES_SM_50
+#define NV_PROVIDES_SM_52   __NV_PROVIDES_SM_52
+#define NV_PROVIDES_SM_53   __NV_PROVIDES_SM_53
+#define NV_PROVIDES_SM_60   __NV_PROVIDES_SM_60
+#define NV_PROVIDES_SM_61   __NV_PROVIDES_SM_61
+#define NV_PROVIDES_SM_62   __NV_PROVIDES_SM_62
+#define NV_PROVIDES_SM_70   __NV_PROVIDES_SM_70
+#define NV_PROVIDES_SM_72   __NV_PROVIDES_SM_72
+#define NV_PROVIDES_SM_75   __NV_PROVIDES_SM_75
+#define NV_PROVIDES_SM_80   __NV_PROVIDES_SM_80
+#define NV_PROVIDES_SM_86   __NV_PROVIDES_SM_86
+#define NV_PROVIDES_SM_87   __NV_PROVIDES_SM_87
+#define NV_PROVIDES_SM_90   __NV_PROVIDES_SM_90
+
+#define NV_IS_EXACTLY_SM_35 __NV_IS_EXACTLY_SM_35
+#define NV_IS_EXACTLY_SM_37 __NV_IS_EXACTLY_SM_37
+#define NV_IS_EXACTLY_SM_50 __NV_IS_EXACTLY_SM_50
+#define NV_IS_EXACTLY_SM_52 __NV_IS_EXACTLY_SM_52
+#define NV_IS_EXACTLY_SM_53 __NV_IS_EXACTLY_SM_53
+#define NV_IS_EXACTLY_SM_60 __NV_IS_EXACTLY_SM_60
+#define NV_IS_EXACTLY_SM_61 __NV_IS_EXACTLY_SM_61
+#define NV_IS_EXACTLY_SM_62 __NV_IS_EXACTLY_SM_62
+#define NV_IS_EXACTLY_SM_70 __NV_IS_EXACTLY_SM_70
+#define NV_IS_EXACTLY_SM_72 __NV_IS_EXACTLY_SM_72
+#define NV_IS_EXACTLY_SM_75 __NV_IS_EXACTLY_SM_75
+#define NV_IS_EXACTLY_SM_80 __NV_IS_EXACTLY_SM_80
+#define NV_IS_EXACTLY_SM_86 __NV_IS_EXACTLY_SM_86
+#define NV_IS_EXACTLY_SM_87 __NV_IS_EXACTLY_SM_87
+#define NV_IS_EXACTLY_SM_90 __NV_IS_EXACTLY_SM_90
+
+#define NV_IS_HOST         __NV_IS_HOST
+#define NV_IS_DEVICE       __NV_IS_DEVICE
+
+#define NV_ANY_TARGET      __NV_ANY_TARGET
+#define NV_NO_TARGET       __NV_NO_TARGET
+
+// Platform invoke mechanisms
+#if defined(_NV_COMPILER_NVCXX)
+
+#  define _NV_ARCH_COND(q) (_NV_TARGET_##q)
+
+#  define _NV_BLOCK_EXPAND(...) _NV_REMOVE_PAREN(__VA_ARGS__)
+
+#  define _NV_TARGET_IF(cond, t, ...) \
+    (if target _NV_ARCH_COND(cond) {    \
+      _NV_BLOCK_EXPAND(t)        \
+    } else { _NV_BLOCK_EXPAND(__VA_ARGS__) })
+
+#elif defined(_NV_COMPILER_NVCC) || defined (_NV_COMPILER_CLANG_CUDA)
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_35)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_35 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_35 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_37)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_37 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_37 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_50)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_50 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_50 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_52)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_52 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_52 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_53)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_53 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_53 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_60)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_60 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_60 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_61)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_61 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_61 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_62)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_62 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_62 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_70)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_70 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_70 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_72)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_72 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_72 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_75)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_75 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_75 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_80)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_80 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_80 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_86)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_86 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_86 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_87)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_87 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_87 0
+#  endif
+
+#  if (_NV_TARGET___NV_IS_EXACTLY_SM_90)
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_90 1
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_EXACTLY_SM_90 0
+#  endif
+
+#  if (_NV_TARGET_IS_HOST)
+#    define _NV_TARGET_BOOL___NV_IS_HOST   1
+#    define _NV_TARGET_BOOL___NV_IS_DEVICE 0
+#  else
+#    define _NV_TARGET_BOOL___NV_IS_HOST   0
+#    define _NV_TARGET_BOOL___NV_IS_DEVICE 1
+#  endif
+
+#  define _NV_TARGET_BOOL___NV_ANY_TARGET 1
+#  define _NV_TARGET_BOOL___NV_NO_TARGET 0
+
+// NVCC Greater than stuff
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_35)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_35 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_35 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_37)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_37 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_37 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_50)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_50 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_50 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_52)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_52 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_52 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_53)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_53 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_53 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_60)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_60 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_60 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_61)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_61 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_61 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_62)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_62 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_62 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_70)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_70 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_70 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_72)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_72 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_72 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_75)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_75 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_75 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_80)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_80 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_80 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_86)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_86 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_86 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_87)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_87 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_87 0
+#  endif
+
+#  if (_NV_TARGET___NV_PROVIDES_SM_90)
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_90 1
+#  else
+#    define _NV_TARGET_BOOL___NV_PROVIDES_SM_90 0
+#  endif
+
+#  define _NV_ARCH_COND_CAT1(cond) _NV_TARGET_BOOL_##cond
+#  define _NV_ARCH_COND_CAT(cond) _NV_EVAL(_NV_ARCH_COND_CAT1(cond))
+
+#    define _NV_TARGET_EMPTY_PARAM ;
+
+#  if defined(_NV_TARGET_CPP11)
+
+#    define _NV_BLOCK_EXPAND(...) { _NV_REMOVE_PAREN(__VA_ARGS__) }
+#    define _NV_TARGET_IF(cond, t, ...) _NV_IF( _NV_ARCH_COND_CAT(cond), t, __VA_ARGS__)
+
+#  else // <C++11 fallback
+
+#    define _NV_BLOCK_EXPAND(x) { x }
+
+#    define _NV_TARGET_IF(cond, t)         _NV_IF(_NV_ARCH_COND_CAT(cond), t, _NV_TARGET_EMPTY_PARAM)
+#    define _NV_TARGET_IF_ELSE(cond, t, f) _NV_IF(_NV_ARCH_COND_CAT(cond), t, f)
+
+#  endif
+
+#endif // _NV_COMPILER_NVCC
+
+#if defined(_NV_TARGET_CPP11)
+
+#  define _NV_TARGET_DISPATCH_HANDLE0()
+#  define _NV_TARGET_DISPATCH_HANDLE2(q, fn)        _NV_TARGET_IF(q, fn)
+#  define _NV_TARGET_DISPATCH_HANDLE4(q, fn, ...)   _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE2(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE6(q, fn, ...)   _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE4(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE8(q, fn, ...)   _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE6(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE10(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE8(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE12(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE10(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE14(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE12(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE16(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE14(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE18(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE16(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE20(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE18(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE22(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE20(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE24(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE22(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE26(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE24(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE28(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE26(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE30(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE28(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE32(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE30(__VA_ARGS__))
+#  define _NV_TARGET_DISPATCH_HANDLE34(q, fn, ...)  _NV_TARGET_IF(q, fn, _NV_TARGET_DISPATCH_HANDLE32(__VA_ARGS__))
+
+#  define _NV_TARGET_DISPATCH(...) _NV_BLOCK_EXPAND(_NV_DISPATCH_N_ARY(_NV_TARGET_DISPATCH_HANDLE, __VA_ARGS__))
+
+// NV_IF_TARGET supports a false statement provided as a variadic macro
+#  define NV_IF_TARGET(cond, ...)    _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, __VA_ARGS__))
+#  define NV_IF_ELSE_TARGET(cond, t, f) _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, t, f))
+#  define NV_DISPATCH_TARGET(...)       _NV_TARGET_DISPATCH(__VA_ARGS__)
+
+#else // <C++11 fallback
+
+// NV_IF_TARGET does not support a fallback false statement in C++03 or C dialects
+#  define NV_IF_TARGET(cond, t)         _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, t))
+#  define NV_IF_ELSE_TARGET(cond, t, f) _NV_BLOCK_EXPAND(_NV_TARGET_IF_ELSE(cond, t, f))
+
+#endif
+
+#endif // _NV__TARGET_MACROS
diff --git a/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/target b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/target
new file mode 100644
index 000000000000..9788b6ad3692
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/dependencies/libcudacxx/include/nv/target
@@ -0,0 +1,207 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of libcu++, the C++ Standard Library for your entire system,
+// under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// This header contains a preview of a portability system that enables
+// CUDA C++ development with NVC++, NVCC, and supported host compilers.
+// These interfaces are not guaranteed to be stable.
+
+#ifndef __NV_TARGET_H
+#define __NV_TARGET_H
+
+#if defined(__NVCC__) || defined(__CUDACC_RTC__)
+#  define _NV_COMPILER_NVCC
+#elif defined(__NVCOMPILER) && __cplusplus >= 201103L
+#  define _NV_COMPILER_NVCXX
+#elif defined(__clang__) && defined(__CUDA__) && defined(__CUDA_ARCH__)
+// clang compiling CUDA code, device mode.
+#  define _NV_COMPILER_CLANG_CUDA
+#endif
+
+#if defined(__CUDACC_RTC__)
+#  define _NV_FUNCTION_ANNOTATION __device__
+#else
+#  define _NV_FUNCTION_ANNOTATION
+#endif
+
+#if defined(_NV_COMPILER_NVCXX)
+#  define _NV_BITSET_ATTRIBUTE [[nv::__target_bitset]]
+#else
+#  define _NV_BITSET_ATTRIBUTE
+#endif
+
+#if (!defined(__ibmxl__)) && \
+    ((defined(__cplusplus) && __cplusplus >= 201103L) || \
+     (defined(_MSC_VER) && _MSVC_LANG >= 201103L))
+#  define _NV_TARGET_CPP11
+#endif
+
+#if defined(_NV_TARGET_CPP11)
+
+namespace nv {
+  namespace target {
+    namespace detail {
+
+      typedef unsigned long long base_int_t;
+
+      // No host specialization
+      constexpr base_int_t all_hosts = 1;
+
+      // NVIDIA GPUs
+      constexpr base_int_t sm_35_bit = 1 << 1;
+      constexpr base_int_t sm_37_bit = 1 << 2;
+      constexpr base_int_t sm_50_bit = 1 << 3;
+      constexpr base_int_t sm_52_bit = 1 << 4;
+      constexpr base_int_t sm_53_bit = 1 << 5;
+      constexpr base_int_t sm_60_bit = 1 << 6;
+      constexpr base_int_t sm_61_bit = 1 << 7;
+      constexpr base_int_t sm_62_bit = 1 << 8;
+      constexpr base_int_t sm_70_bit = 1 << 9;
+      constexpr base_int_t sm_72_bit = 1 << 10;
+      constexpr base_int_t sm_75_bit = 1 << 11;
+      constexpr base_int_t sm_80_bit = 1 << 12;
+      constexpr base_int_t sm_86_bit = 1 << 13;
+      constexpr base_int_t sm_87_bit = 1 << 14;
+      constexpr base_int_t sm_90_bit = 1 << 15;
+      constexpr base_int_t all_devices =
+          sm_35_bit | sm_37_bit |
+          sm_50_bit | sm_52_bit | sm_53_bit |
+          sm_60_bit | sm_61_bit | sm_62_bit |
+          sm_70_bit | sm_72_bit | sm_75_bit |
+          sm_80_bit | sm_86_bit | sm_87_bit |
+          sm_90_bit;
+
+      // Store a set of targets as a set of bits
+      struct _NV_BITSET_ATTRIBUTE target_description {
+        base_int_t targets;
+        _NV_FUNCTION_ANNOTATION
+        constexpr target_description(base_int_t a) : targets(a) { }
+      };
+
+      // The type of the user-visible names of the NVIDIA GPU targets
+      enum class sm_selector : base_int_t {
+        sm_35 = 35, sm_37 = 37,
+        sm_50 = 50, sm_52 = 52, sm_53 = 53,
+        sm_60 = 60, sm_61 = 61, sm_62 = 62,
+        sm_70 = 70, sm_72 = 72, sm_75 = 75,
+        sm_80 = 80, sm_86 = 86, sm_87 = 87,
+        sm_90 = 90,
+      };
+      _NV_FUNCTION_ANNOTATION
+      constexpr base_int_t toint(sm_selector a) {
+        return static_cast<base_int_t>(a);
+      }
+      _NV_FUNCTION_ANNOTATION
+      constexpr base_int_t bitexact(sm_selector a) {
+        return toint(a) == 35 ? sm_35_bit :
+               toint(a) == 37 ? sm_37_bit :
+               toint(a) == 50 ? sm_50_bit :
+               toint(a) == 52 ? sm_52_bit :
+               toint(a) == 53 ? sm_53_bit :
+               toint(a) == 60 ? sm_60_bit :
+               toint(a) == 61 ? sm_61_bit :
+               toint(a) == 62 ? sm_62_bit :
+               toint(a) == 70 ? sm_70_bit :
+               toint(a) == 72 ? sm_72_bit :
+               toint(a) == 75 ? sm_75_bit :
+               toint(a) == 80 ? sm_80_bit :
+               toint(a) == 86 ? sm_86_bit :
+               toint(a) == 87 ? sm_87_bit :
+               toint(a) == 90 ? sm_90_bit : 0;
+      }
+      _NV_FUNCTION_ANNOTATION
+      constexpr base_int_t bitrounddown(sm_selector a) {
+        return toint(a) >= 90 ? sm_90_bit :
+               toint(a) >= 87 ? sm_87_bit :
+               toint(a) >= 86 ? sm_86_bit :
+               toint(a) >= 80 ? sm_80_bit :
+               toint(a) >= 75 ? sm_75_bit :
+               toint(a) >= 72 ? sm_72_bit :
+               toint(a) >= 70 ? sm_70_bit :
+               toint(a) >= 62 ? sm_62_bit :
+               toint(a) >= 61 ? sm_61_bit :
+               toint(a) >= 60 ? sm_60_bit :
+               toint(a) >= 53 ? sm_53_bit :
+               toint(a) >= 52 ? sm_52_bit :
+               toint(a) >= 50 ? sm_50_bit :
+               toint(a) >= 37 ? sm_37_bit :
+               toint(a) >= 35 ? sm_35_bit : 0;
+      }
+
+      // Public API for NVIDIA GPUs
+
+      _NV_FUNCTION_ANNOTATION
+      constexpr target_description is_exactly(sm_selector a) {
+        return target_description(bitexact(a));
+      }
+
+      _NV_FUNCTION_ANNOTATION
+      constexpr target_description provides(sm_selector a) {
+        return target_description(~(bitrounddown(a) - 1) & all_devices);
+      }
+
+      // Boolean operations on target sets
+
+      _NV_FUNCTION_ANNOTATION
+      constexpr target_description operator&&(target_description a,
+                                              target_description b) {
+        return target_description(a.targets & b.targets);
+      }
+
+      _NV_FUNCTION_ANNOTATION
+      constexpr target_description operator||(target_description a,
+                                              target_description b) {
+        return target_description(a.targets | b.targets);
+      }
+
+      _NV_FUNCTION_ANNOTATION
+      constexpr target_description operator!(target_description a) {
+        return target_description(~a.targets & (all_devices | all_hosts));
+      }
+    }
+
+    using detail::target_description;
+    using detail::sm_selector;
+
+    // The predicates for basic host/device selection
+    constexpr target_description is_host =
+      target_description(detail::all_hosts);
+    constexpr target_description is_device =
+      target_description(detail::all_devices);
+    constexpr target_description any_target =
+      target_description(detail::all_hosts | detail::all_devices);
+    constexpr target_description no_target =
+      target_description(0);
+
+    // The public names for NVIDIA GPU architectures
+    constexpr sm_selector sm_35 = sm_selector::sm_35;
+    constexpr sm_selector sm_37 = sm_selector::sm_37;
+    constexpr sm_selector sm_50 = sm_selector::sm_50;
+    constexpr sm_selector sm_52 = sm_selector::sm_52;
+    constexpr sm_selector sm_53 = sm_selector::sm_53;
+    constexpr sm_selector sm_60 = sm_selector::sm_60;
+    constexpr sm_selector sm_61 = sm_selector::sm_61;
+    constexpr sm_selector sm_62 = sm_selector::sm_62;
+    constexpr sm_selector sm_70 = sm_selector::sm_70;
+    constexpr sm_selector sm_72 = sm_selector::sm_72;
+    constexpr sm_selector sm_75 = sm_selector::sm_75;
+    constexpr sm_selector sm_80 = sm_selector::sm_80;
+    constexpr sm_selector sm_86 = sm_selector::sm_86;
+    constexpr sm_selector sm_87 = sm_selector::sm_87;
+    constexpr sm_selector sm_90 = sm_selector::sm_90;
+
+    using detail::is_exactly;
+    using detail::provides;
+  }
+}
+
+#endif // C++11
+
+#include "detail/__target_macros"
+
+#endif // __NV_TARGET_H
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/addressof.h b/thirdparty/manifold/thirdparty/thrust/thrust/addressof.h
new file mode 100644
index 000000000000..d21df0c76a5e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/addressof.h
@@ -0,0 +1,31 @@
+// Copyright (c) 2018 NVIDIA Corporation
+// Author: Bryce Adelstein Lelbach <brycelelbach@gmail.com>
+//
+// Distributed under the Boost Software License v1.0 (boost.org/LICENSE_1_0.txt)
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+#  include <thrust/detail/memory_wrapper.h>
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! Obtains the actual address of the object or function arg, even in presence of overloaded operator&.
+ */
+template <typename T>
+__host__ __device__
+T* addressof(T& arg) 
+{
+  return reinterpret_cast<T*>(
+    &const_cast<char&>(reinterpret_cast<const volatile char&>(arg))
+  );
+}
+
+///////////////////////////////////////////////////////////////////////////////
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/adjacent_difference.h b/thirdparty/manifold/thirdparty/thrust/thrust/adjacent_difference.h
new file mode 100644
index 000000000000..e8385c240dd9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/adjacent_difference.h
@@ -0,0 +1,244 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file adjacent_difference.h
+ *  \brief Compute difference between consecutive elements of a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup transformations Transformations
+ *  \{
+ */
+
+
+/*! \p adjacent_difference calculates the differences of adjacent elements in the
+ *  range <tt>[first, last)</tt>. That is, <tt>\*first</tt> is assigned to
+ *  <tt>\*result</tt>, and, for each iterator \p i in the range
+ *  <tt>[first + 1, last)</tt>, the difference of <tt>\*i</tt> and <tt>*(i - 1)</tt>
+ *  is assigned to <tt>\*(result + (i - first))</tt>.
+ *
+ *  This version of \p adjacent_difference uses <tt>operator-</tt> to calculate
+ *  differences.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \return The iterator <tt>result + (last - first)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \c x and \c y are objects of \p InputIterator's \c value_type, then \c x - \c is defined,
+ *          and \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types,
+ *          and the return type of <tt>x - y</tt> is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \remark Note that \p result is permitted to be the same iterator as \p first. This is
+ *          useful for computing differences "in place".
+ *
+ *  The following code snippet demonstrates how to use \p adjacent_difference to compute
+ *  the difference between adjacent elements of a range using the \p thrust::device execution policy:
+ *
+ *  \code
+ *  #include <thrust/adjacent_difference.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int h_data[8] = {1, 2, 1, 2, 1, 2, 1, 2};
+ *  thrust::device_vector<int> d_data(h_data, h_data + 8);
+ *  thrust::device_vector<int> d_result(8);
+ *
+ *  thrust::adjacent_difference(thrust::device, d_data.begin(), d_data.end(), d_result.begin());
+ *
+ *  // d_result is now [1, 1, -1, 1, -1, 1, -1, 1]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/adjacent_difference
+ *  \see inclusive_scan
+ */
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last, 
+                                   OutputIterator result);
+
+/*! \p adjacent_difference calculates the differences of adjacent elements in the
+ *  range <tt>[first, last)</tt>. That is, <tt>*first</tt> is assigned to
+ *  <tt>\*result</tt>, and, for each iterator \p i in the range
+ *  <tt>[first + 1, last)</tt>, <tt>binary_op(\*i, \*(i - 1))</tt> is assigned to
+ *  <tt>\*(result + (i - first))</tt>.
+ *  
+ *  This version of \p adjacent_difference uses the binary function \p binary_op to
+ *  calculate differences.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \param binary_op The binary function used to compute differences.
+ *  \return The iterator <tt>result + (last - first)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p BinaryFunction's \c first_argument_type and \c second_argument_type,
+ *          and \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam BinaryFunction's \c result_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *
+ *  \remark Note that \p result is permitted to be the same iterator as \p first. This is
+ *          useful for computing differences "in place".
+ *
+ *  The following code snippet demonstrates how to use \p adjacent_difference to compute
+ *  the sum between adjacent elements of a range using the \p thrust::device execution policy:
+ *
+ *  \code
+ *  #include <thrust/adjacent_difference.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int h_data[8] = {1, 2, 1, 2, 1, 2, 1, 2};
+ *  thrust::device_vector<int> d_data(h_data, h_data + 8);
+ *  thrust::device_vector<int> d_result(8);
+ *
+ *  thrust::adjacent_difference(thrust::device, d_data.begin(), d_data.end(), d_result.begin(), thrust::plus<int>());
+ *
+ *  // d_result is now [1, 3, 3, 3, 3, 3, 3, 3]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/adjacent_difference
+ *  \see inclusive_scan
+ */
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename BinaryFunction>
+__host__ __device__
+OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op);
+
+/*! \p adjacent_difference calculates the differences of adjacent elements in the
+ *  range <tt>[first, last)</tt>. That is, <tt>\*first</tt> is assigned to
+ *  <tt>\*result</tt>, and, for each iterator \p i in the range
+ *  <tt>[first + 1, last)</tt>, the difference of <tt>\*i</tt> and <tt>*(i - 1)</tt>
+ *  is assigned to <tt>\*(result + (i - first))</tt>.
+ *
+ *  This version of \p adjacent_difference uses <tt>operator-</tt> to calculate
+ *  differences.
+ *
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \return The iterator <tt>result + (last - first)</tt>
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \c x and \c y are objects of \p InputIterator's \c value_type, then \c x - \c is defined,
+ *          and \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types,
+ *          and the return type of <tt>x - y</tt> is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \remark Note that \p result is permitted to be the same iterator as \p first. This is
+ *          useful for computing differences "in place".
+ *
+ *  The following code snippet demonstrates how to use \p adjacent_difference to compute
+ *  the difference between adjacent elements of a range.
+ *
+ *  \code
+ *  #include <thrust/adjacent_difference.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  int h_data[8] = {1, 2, 1, 2, 1, 2, 1, 2};
+ *  thrust::device_vector<int> d_data(h_data, h_data + 8);
+ *  thrust::device_vector<int> d_result(8);
+ *
+ *  thrust::adjacent_difference(d_data.begin(), d_data.end(), d_result.begin());
+ *
+ *  // d_result is now [1, 1, -1, 1, -1, 1, -1, 1]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/adjacent_difference
+ *  \see inclusive_scan
+ */
+template <typename InputIterator, typename OutputIterator>
+OutputIterator adjacent_difference(InputIterator first, InputIterator last, 
+                                   OutputIterator result);
+
+/*! \p adjacent_difference calculates the differences of adjacent elements in the
+ *  range <tt>[first, last)</tt>. That is, <tt>*first</tt> is assigned to
+ *  <tt>\*result</tt>, and, for each iterator \p i in the range
+ *  <tt>[first + 1, last)</tt>, <tt>binary_op(\*i, \*(i - 1))</tt> is assigned to
+ *  <tt>\*(result + (i - first))</tt>.
+ *  
+ *  This version of \p adjacent_difference uses the binary function \p binary_op to
+ *  calculate differences.
+ *
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \param binary_op The binary function used to compute differences.
+ *  \return The iterator <tt>result + (last - first)</tt>
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p BinaryFunction's \c first_argument_type and \c second_argument_type,
+ *          and \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam BinaryFunction's \c result_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *
+ *  \remark Note that \p result is permitted to be the same iterator as \p first. This is
+ *          useful for computing differences "in place".
+ *
+ *  The following code snippet demonstrates how to use \p adjacent_difference to compute
+ *  the sum between adjacent elements of a range.
+ *
+ *  \code
+ *  #include <thrust/adjacent_difference.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  int h_data[8] = {1, 2, 1, 2, 1, 2, 1, 2};
+ *  thrust::device_vector<int> d_data(h_data, h_data + 8);
+ *  thrust::device_vector<int> d_result(8);
+ *
+ *  thrust::adjacent_difference(d_data.begin(), d_data.end(), d_result.begin(), thrust::plus<int>());
+ *
+ *  // d_result is now [1, 3, 3, 3, 3, 3, 3, 3]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/adjacent_difference
+ *  \see inclusive_scan
+ */
+template <typename InputIterator, typename OutputIterator, typename BinaryFunction>
+OutputIterator adjacent_difference(InputIterator first, InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op);
+
+/*! \}
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/adjacent_difference.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/advance.h b/thirdparty/manifold/thirdparty/thrust/thrust/advance.h
new file mode 100644
index 000000000000..a5162e203002
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/advance.h
@@ -0,0 +1,140 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file advance.h
+ *  \brief Advance an iterator by a given distance.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \p advance(i, n) increments the iterator \p i by the distance \p n.
+ *  If <tt>n > 0</tt> it is equivalent to executing <tt>++i</tt> \p n
+ *  times, and if <tt>n < 0</tt> it is equivalent to executing <tt>--i</tt>
+ *  \p n times. If <tt>n == 0</tt>, the call has no effect.
+ *
+ *  \param i The iterator to be advanced.
+ *  \param n The distance by which to advance the iterator.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Distance is an integral type that is convertible to \p InputIterator's distance type.
+ *
+ *  \pre \p n shall be negative only for bidirectional and random access iterators.
+ *
+ *  The following code snippet demonstrates how to use \p advance to increment
+ *  an iterator a given number of times.
+ *
+ *  \code
+ *  #include <thrust/advance.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> vec(13);
+ *  thrust::device_vector<int>::iterator iter = vec.begin();
+ *
+ *  thrust::advance(iter, 7);
+ *
+ *  // iter - vec.begin() == 7
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/advance
+ */
+template <typename InputIterator, typename Distance>
+__host__ __device__
+void advance(InputIterator& i, Distance n);
+
+/*! \p next(i, n) returns the \p n th successor of the iterator \p i.
+ *
+ *  \param i An iterator.
+ *  \param n The number of elements to advance.
+ *
+ *  \tparam InputIterator must meet the <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">InputIterator</a>.
+ *
+ *  \pre \p n shall be negative only for bidirectional and random access iterators.
+ *
+ *  The following code snippet demonstrates how to use \p next.
+ *
+ *  \code
+ *  #include <thrust/advance.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> vec(13);
+ *  thrust::device_vector<int>::iterator i0 = vec.begin();
+ *
+ *  auto i1 = thrust::next(i0);
+ *
+ *  // i0 - vec.begin() == 0
+ *  // i1 - vec.begin() == 1
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/next
+ */
+#if 0 // Doxygen only
+template <typename InputIterator, typename Distance>
+__host__ __device__
+InputIterator next(
+  InputIterator i
+, typename iterator_traits<InputIterator>::difference_type n = 1
+);
+#endif
+
+/*! \p prev(i, n) returns the \p n th predecessor of the iterator \p i.
+ *
+ *  \param i An iterator.
+ *  \param n The number of elements to descend.
+ *
+ *  \tparam BidirectionalIterator must meet the <a href="https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator">BidirectionalIterator</a>.
+ *
+ *  The following code snippet demonstrates how to use \p prev.
+ *
+ *  \code
+ *  #include <thrust/advance.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> vec(13);
+ *  thrust::device_vector<int>::iterator i0 = vec.end();
+ *
+ *  auto i1 = thrust::prev(i0);
+ *
+ *  // vec.end() - i0 == 0
+ *  // vec.end() - i1 == 1
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/prev
+ */
+#if 0 // Doxygen only
+template <typename BidirectionalIterator, typename Distance>
+__host__ __device__
+BidirectionalIterator prev(
+  BidirectionalIterator i
+, typename iterator_traits<BidirectionalIterator>::difference_type n = 1
+);
+#endif
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/advance.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/allocate_unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/allocate_unique.h
new file mode 100644
index 000000000000..ff10cb51cd19
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/allocate_unique.h
@@ -0,0 +1,443 @@
+// Copyright (c) 2018 NVIDIA Corporation
+// Author: Bryce Adelstein Lelbach <brycelelbach@gmail.com>
+//
+// Distributed under the Boost Software License v1.0 (boost.org/LICENSE_1_0.txt)
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/type_deduction.h>
+#include <thrust/detail/memory_algorithms.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+
+#include <utility>
+#include <thrust/detail/memory_wrapper.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// wg21.link/p0316r0
+
+///////////////////////////////////////////////////////////////////////////////
+
+namespace detail
+{
+
+template <typename Allocator, typename Pointer>
+void allocator_delete_impl(
+  Allocator const& alloc, Pointer p, std::false_type
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  if (nullptr != pointer_traits<Pointer>::get(p))
+  {
+    traits::destroy(alloc_T, thrust::raw_pointer_cast(p));
+    traits::deallocate(alloc_T, p, 1);
+  }
+}
+
+template <typename Allocator, typename Pointer>
+void allocator_delete_impl(
+  Allocator const& alloc, Pointer p, std::true_type
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  if (nullptr != pointer_traits<Pointer>::get(p))
+  {
+    traits::deallocate(alloc_T, p, 1);
+  }
+}
+
+} // namespace detail
+
+template <typename T, typename Allocator, bool Uninitialized = false>
+struct allocator_delete final
+{
+  using allocator_type
+    = typename std::remove_cv<
+        typename std::remove_reference<Allocator>::type
+      >::type::template rebind<T>::other;
+  using pointer = typename detail::allocator_traits<allocator_type>::pointer;
+
+  template <typename UAllocator>
+  allocator_delete(UAllocator&& other) noexcept
+    : alloc_(THRUST_FWD(other))
+  {}
+
+  template <typename U, typename UAllocator>
+  allocator_delete(
+      allocator_delete<U, UAllocator> const& other
+    ) noexcept
+    : alloc_(other.get_allocator())
+  {}
+  template <typename U, typename UAllocator>
+  allocator_delete(
+      allocator_delete<U, UAllocator>&& other
+    ) noexcept
+    : alloc_(std::move(other.get_allocator()))
+  {}
+
+  template <typename U, typename UAllocator>
+  allocator_delete& operator=(
+    allocator_delete<U, UAllocator> const& other
+  ) noexcept
+  {
+    alloc_ = other.get_allocator();
+    return *this;
+  }
+  template <typename U, typename UAllocator>
+  allocator_delete& operator=(
+    allocator_delete<U, UAllocator>&& other
+  ) noexcept
+  {
+    alloc_ = std::move(other.get_allocator());
+    return *this;
+  }
+
+  void operator()(pointer p)
+  {
+    std::integral_constant<bool, Uninitialized> ic;
+
+    detail::allocator_delete_impl(get_allocator(), p, ic);
+  }
+
+  allocator_type& get_allocator() noexcept { return alloc_; }
+  allocator_type const& get_allocator() const noexcept { return alloc_; }
+
+  void swap(allocator_delete& other) noexcept
+  {
+    using std::swap;
+    swap(alloc_, other.alloc_);
+  }
+
+private:
+  allocator_type alloc_;
+};
+
+template <typename T, typename Allocator>
+using uninitialized_allocator_delete = allocator_delete<T, Allocator, true>;
+
+namespace detail {
+
+template <typename Allocator, typename Pointer, typename Size>
+void array_allocator_delete_impl(
+  Allocator const& alloc, Pointer p, Size count, std::false_type
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  if (nullptr != pointer_traits<Pointer>::get(p))
+  {
+    destroy_n(alloc_T, p, count);
+    traits::deallocate(alloc_T, p, count);
+  }
+}
+
+template <typename Allocator, typename Pointer, typename Size>
+void array_allocator_delete_impl(
+  Allocator const& alloc, Pointer p, Size count, std::true_type
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  if (nullptr != pointer_traits<Pointer>::get(p))
+  {
+    traits::deallocate(alloc_T, p, count);
+  }
+}
+
+} // namespace detail
+
+template <typename T, typename Allocator, bool Uninitialized = false>
+struct array_allocator_delete final
+{
+  using allocator_type
+    = typename std::remove_cv<
+        typename std::remove_reference<Allocator>::type
+      >::type::template rebind<T>::other;
+  using pointer = typename detail::allocator_traits<allocator_type>::pointer;
+
+  template <typename UAllocator>
+  array_allocator_delete(UAllocator&& other, std::size_t n) noexcept
+    : alloc_(THRUST_FWD(other)), count_(n)
+  {}
+
+  template <typename U, typename UAllocator>
+  array_allocator_delete(
+      array_allocator_delete<U, UAllocator> const& other
+    ) noexcept
+    : alloc_(other.get_allocator()), count_(other.count_)
+  {}
+  template <typename U, typename UAllocator>
+  array_allocator_delete(
+      array_allocator_delete<U, UAllocator>&& other
+    ) noexcept
+    : alloc_(std::move(other.get_allocator())), count_(other.count_)
+  {}
+
+  template <typename U, typename UAllocator>
+  array_allocator_delete& operator=(
+    array_allocator_delete<U, UAllocator> const& other
+  ) noexcept
+  {
+    alloc_ = other.get_allocator();
+    count_ = other.count_;
+    return *this;
+  }
+  template <typename U, typename UAllocator>
+  array_allocator_delete& operator=(
+    array_allocator_delete<U, UAllocator>&& other
+  ) noexcept
+  {
+    alloc_ = std::move(other.get_allocator());
+    count_ = other.count_;
+    return *this;
+  }
+
+  void operator()(pointer p)
+  {
+    std::integral_constant<bool, Uninitialized> ic;
+
+    detail::array_allocator_delete_impl(get_allocator(), p, count_, ic);
+  }
+
+  allocator_type& get_allocator() noexcept { return alloc_; }
+  allocator_type const& get_allocator() const noexcept { return alloc_; }
+
+  void swap(array_allocator_delete& other) noexcept
+  {
+    using std::swap;
+    swap(alloc_, other.alloc_);
+    swap(count_, other.count_);
+  }
+
+private:
+  allocator_type alloc_;
+  std::size_t    count_;
+};
+  
+template <typename T, typename Allocator>
+using uninitialized_array_allocator_delete
+  = array_allocator_delete<T, Allocator, true>;
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename Pointer, typename Lambda>
+struct tagged_deleter : Lambda
+{
+  __host__ __device__
+  tagged_deleter(Lambda&& l) : Lambda(THRUST_FWD(l)) {}
+
+  using pointer = Pointer;
+};
+
+template <typename Pointer, typename Lambda>
+__host__ __device__
+tagged_deleter<Pointer, Lambda>
+make_tagged_deleter(Lambda&& l)
+{
+  return tagged_deleter<Pointer, Lambda>(THRUST_FWD(l));
+}
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T, typename Allocator, typename... Args>
+__host__
+std::unique_ptr<
+  T,
+  allocator_delete<
+    T
+  , typename detail::allocator_traits<
+      typename std::remove_cv<
+        typename std::remove_reference<Allocator>::type
+      >::type
+    >::template rebind_traits<T>::allocator_type
+  >
+>
+allocate_unique(
+  Allocator const& alloc, Args&&... args
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  auto hold_deleter = make_tagged_deleter<typename traits::pointer>(
+    [&alloc_T] (typename traits::pointer p) {
+      traits::deallocate(alloc_T, p, 1);
+    }
+  );
+  using hold_t = std::unique_ptr<T, decltype(hold_deleter)>;
+  auto hold = hold_t(traits::allocate(alloc_T, 1), hold_deleter);
+
+  traits::construct(
+    alloc_T, thrust::raw_pointer_cast(hold.get()), THRUST_FWD(args)...
+  );
+  auto deleter = allocator_delete<T, typename traits::allocator_type>(alloc);
+  return std::unique_ptr<T, decltype(deleter)>
+    (hold.release(), std::move(deleter));
+}
+
+template <typename T, typename Allocator>
+__host__
+std::unique_ptr<
+  T,
+  uninitialized_allocator_delete<
+    T
+  , typename detail::allocator_traits<
+      typename std::remove_cv<
+        typename std::remove_reference<Allocator>::type
+      >::type
+    >::template rebind_traits<T>::allocator_type
+  >
+>
+uninitialized_allocate_unique(
+  Allocator const& alloc
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  auto hold_deleter = make_tagged_deleter<typename traits::pointer>(
+    [&alloc_T] (typename traits::pointer p) {
+      traits::deallocate(alloc_T, p, 1);
+    }
+  );
+  using hold_t = std::unique_ptr<T, decltype(hold_deleter)>;
+  auto hold = hold_t(traits::allocate(alloc_T, 1), hold_deleter);
+
+  auto deleter = uninitialized_allocator_delete<
+    T, typename traits::allocator_type
+  >(alloc_T);
+  return std::unique_ptr<T, decltype(deleter)>
+    (hold.release(), std::move(deleter));
+}
+
+template <typename T, typename Allocator, typename Size, typename... Args>
+__host__
+std::unique_ptr<
+  T[],
+  array_allocator_delete<
+    T
+  , typename detail::allocator_traits<
+      typename std::remove_cv<
+        typename std::remove_reference<Allocator>::type
+      >::type
+    >::template rebind_traits<T>::allocator_type
+  >
+>
+allocate_unique_n(
+  Allocator const& alloc, Size n, Args&&... args
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  auto hold_deleter = make_tagged_deleter<typename traits::pointer>(
+    [n, &alloc_T] (typename traits::pointer p) {
+      traits::deallocate(alloc_T, p, n);
+    }
+  );
+  using hold_t = std::unique_ptr<T[], decltype(hold_deleter)>;
+  auto hold = hold_t(traits::allocate(alloc_T, n), hold_deleter);
+
+  uninitialized_construct_n_with_allocator(
+    alloc_T, hold.get(), n, THRUST_FWD(args)...
+  );
+  auto deleter = array_allocator_delete<
+    T, typename traits::allocator_type
+  >(alloc_T, n);
+  return std::unique_ptr<T[], decltype(deleter)>
+    (hold.release(), std::move(deleter));
+}
+
+template <typename T, typename Allocator, typename Size>
+__host__
+std::unique_ptr<
+  T[],
+  uninitialized_array_allocator_delete<
+    T
+  , typename detail::allocator_traits<
+      typename std::remove_cv<
+        typename std::remove_reference<Allocator>::type
+      >::type
+    >::template rebind_traits<T>::allocator_type
+  >
+>
+uninitialized_allocate_unique_n(
+  Allocator const& alloc, Size n
+)
+{
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  auto hold_deleter = make_tagged_deleter<typename traits::pointer>(
+    [n, &alloc_T] (typename traits::pointer p) {
+      traits::deallocate(alloc_T, p, n);
+    }
+  );
+  using hold_t = std::unique_ptr<T[], decltype(hold_deleter)>;
+  auto hold = hold_t(traits::allocate(alloc_T, n), hold_deleter);
+
+  auto deleter = uninitialized_array_allocator_delete<
+    T, typename traits::allocator_type
+  >(alloc_T, n);
+  return std::unique_ptr<T[], decltype(deleter)>
+    (hold.release(), std::move(deleter));
+}
+
+///////////////////////////////////////////////////////////////////////////////
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/async/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/async/copy.h
new file mode 100644
index 000000000000..a8edc7411e59
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/async/copy.h
@@ -0,0 +1,154 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Algorithms for asynchronously copying a range.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/select_system.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/system/detail/adl/async/copy.h>
+
+#include <thrust/event.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace async
+{
+
+/*! \cond
+ */
+
+namespace unimplemented
+{
+
+template <
+  typename FromPolicy, typename ToPolicy
+, typename ForwardIt, typename Sentinel, typename OutputIt
+>
+__host__
+event<FromPolicy>
+async_copy(
+  thrust::execution_policy<FromPolicy>& from_exec
+, thrust::execution_policy<ToPolicy>&   to_exec
+, ForwardIt first, Sentinel last, OutputIt output
+)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value)
+  , "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace copy_detail
+{
+
+using thrust::async::unimplemented::async_copy;
+
+struct copy_fn final
+{
+  template <
+    typename FromPolicy, typename ToPolicy
+  , typename ForwardIt, typename Sentinel, typename OutputIt
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<FromPolicy> const& from_exec
+  , thrust::detail::execution_policy_base<ToPolicy> const&   to_exec
+  , ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_copy(
+      thrust::detail::derived_cast(thrust::detail::strip_const(from_exec))
+    , thrust::detail::derived_cast(thrust::detail::strip_const(to_exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename OutputIt
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  )
+  THRUST_RETURNS(
+    copy_fn::call(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+      // Synthesize a suitable new execution policy, because we don't want to
+      // try and extract twice from the one we were passed.
+    , typename remove_cvref_t<
+        decltype(thrust::detail::derived_cast(thrust::detail::strip_const(exec)))
+      >::tag_type{}
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename OutputIt>
+  __host__
+  static auto call(ForwardIt&& first, Sentinel&& last, OutputIt&& output)
+  THRUST_RETURNS(
+    copy_fn::call(
+      thrust::detail::select_system(
+        typename thrust::iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , thrust::detail::select_system(
+        typename thrust::iterator_system<remove_cvref_t<OutputIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    )
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace copy_detail
+
+THRUST_INLINE_CONSTANT copy_detail::copy_fn copy{};
+
+/*! \endcond
+ */
+
+} // namespace async
+
+THRUST_NAMESPACE_END
+
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/async/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/async/for_each.h
new file mode 100644
index 000000000000..0d3b3a189050
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/async/for_each.h
@@ -0,0 +1,123 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Algorithms for asynchronously iterating over the elements of a range.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/select_system.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/system/detail/adl/async/for_each.h>
+
+#include <thrust/event.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace async
+{
+
+/*! \cond
+ */
+
+namespace unimplemented
+{
+
+template <
+  typename DerivedPolicy
+, typename ForwardIt, typename Sentinel, typename UnaryFunction
+>
+__host__
+event<DerivedPolicy>
+async_for_each(
+  thrust::execution_policy<DerivedPolicy>&, ForwardIt, Sentinel, UnaryFunction
+)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value)
+  , "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace for_each_detail
+{
+
+using thrust::async::unimplemented::async_for_each;
+
+struct for_each_fn final
+{
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename UnaryFunction
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , UnaryFunction&& f
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_for_each(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(f)
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename UnaryFunction>
+  __host__
+  static auto call(ForwardIt&& first, Sentinel&& last, UnaryFunction&& f)
+  THRUST_RETURNS(
+    for_each_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(f)
+    )
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace for_each_detail
+
+THRUST_INLINE_CONSTANT for_each_detail::for_each_fn for_each{};
+
+/*! \endcond
+ */
+
+} // namespace async
+
+THRUST_NAMESPACE_END
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/async/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/async/reduce.h
new file mode 100644
index 000000000000..8f4fe3133e6f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/async/reduce.h
@@ -0,0 +1,446 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Algorithms for asynchronously reducing a range to a single value.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/select_system.h>
+#include <thrust/type_traits/logical_metafunctions.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/type_traits/is_execution_policy.h>
+#include <thrust/system/detail/adl/async/reduce.h>
+
+#include <thrust/future.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace async
+{
+
+/*! \cond
+ */
+
+namespace unimplemented
+{
+
+template <
+  typename DerivedPolicy
+, typename ForwardIt, typename Sentinel, typename T, typename BinaryOp
+>
+__host__
+future<DerivedPolicy, T>
+async_reduce(
+  thrust::execution_policy<DerivedPolicy>&, ForwardIt, Sentinel, T, BinaryOp
+)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value)
+  , "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace reduce_detail
+{
+
+using thrust::async::unimplemented::async_reduce;
+
+struct reduce_fn final
+{
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename T, typename BinaryOp
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , T&& init
+  , BinaryOp&& op
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_reduce(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(init)
+    , THRUST_FWD(op)
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename T
+  >
+  __host__
+  static auto call4(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , T&& init
+  , thrust::true_type
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_reduce(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(init)
+    , thrust::plus<remove_cvref_t<T>>{}
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel
+  >
+  __host__
+  static auto
+  call3(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , thrust::true_type
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_reduce(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type{}
+    , thrust::plus<
+        remove_cvref_t<
+          typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+        >
+      >{}
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename T, typename BinaryOp>
+  __host__
+  static auto call4(ForwardIt&& first, Sentinel&& last,
+                    T&& init,
+                    BinaryOp&& op,
+                    thrust::false_type)
+  THRUST_RETURNS(
+    reduce_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(init)
+    , THRUST_FWD(op)
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename T>
+  __host__
+  static auto call3(ForwardIt&& first, Sentinel&& last,
+                    T&& init,
+                    thrust::false_type)
+  THRUST_RETURNS(
+    reduce_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(init)
+    , thrust::plus<remove_cvref_t<T>>{}
+    )
+  )
+
+  // MSVC WAR: MSVC gets angsty and eats all available RAM when we try to detect
+  // if T1 is an execution_policy by using SFINAE. Switching to a static
+  // dispatch pattern to prevent this.
+  template <typename T1, typename T2, typename T3>
+  __host__
+  static auto call(T1&& t1, T2&& t2, T3&& t3)
+  THRUST_RETURNS(
+    reduce_fn::call3(THRUST_FWD(t1), THRUST_FWD(t2), THRUST_FWD(t3),
+                     thrust::is_execution_policy<thrust::remove_cvref_t<T1>>{})
+  )
+
+  template <typename T1, typename T2, typename T3, typename T4>
+  __host__
+  static auto call(T1&& t1, T2&& t2, T3&& t3, T4&& t4)
+  THRUST_RETURNS(
+    reduce_fn::call4(THRUST_FWD(t1), THRUST_FWD(t2), THRUST_FWD(t3), THRUST_FWD(t4),
+                     thrust::is_execution_policy<thrust::remove_cvref_t<T1>>{})
+  )
+
+  template <typename ForwardIt, typename Sentinel>
+  __host__
+  static auto call(ForwardIt&& first, Sentinel&& last)
+  THRUST_RETURNS(
+    reduce_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type{}
+    , thrust::plus<
+        remove_cvref_t<
+          typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+        >
+      >{}
+    )
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace reduce_detail
+
+THRUST_INLINE_CONSTANT reduce_detail::reduce_fn reduce{};
+
+///////////////////////////////////////////////////////////////////////////////
+
+namespace unimplemented
+{
+
+template <
+  typename DerivedPolicy
+, typename ForwardIt, typename Sentinel, typename OutputIt
+, typename T, typename BinaryOp
+>
+__host__
+event<DerivedPolicy>
+async_reduce_into(
+  thrust::execution_policy<DerivedPolicy>&
+, ForwardIt, Sentinel, OutputIt, T, BinaryOp
+)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value)
+  , "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace reduce_into_detail
+{
+
+using thrust::async::unimplemented::async_reduce_into;
+
+struct reduce_into_fn final
+{
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename OutputIt
+  , typename T, typename BinaryOp
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , T&& init
+  , BinaryOp&& op
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_reduce_into(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , THRUST_FWD(init)
+    , THRUST_FWD(op)
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename OutputIt
+  , typename T
+  >
+  __host__
+  static auto call5(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , T&& init
+  , thrust::true_type
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_reduce_into(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , THRUST_FWD(init)
+    , thrust::plus<remove_cvref_t<T>>{}
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename OutputIt
+  >
+  __host__
+  static auto
+  call4(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , thrust::true_type
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_reduce_into(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type{}
+    , thrust::plus<
+        remove_cvref_t<
+          typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+        >
+      >{}
+    )
+  )
+
+  template <
+    typename ForwardIt, typename Sentinel, typename OutputIt
+  , typename T, typename BinaryOp
+  >
+  __host__
+  static auto call5(
+    ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , T&& init
+  , BinaryOp&& op
+  , thrust::false_type
+  )
+  THRUST_RETURNS(
+    reduce_into_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      , typename iterator_system<remove_cvref_t<OutputIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , THRUST_FWD(init)
+    , THRUST_FWD(op)
+    )
+  )
+
+  template <
+    typename ForwardIt, typename Sentinel, typename OutputIt
+  , typename T
+  >
+  __host__
+  static auto call4(
+    ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , T&& init
+  , thrust::false_type
+  )
+  THRUST_RETURNS(
+    reduce_into_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      , typename iterator_system<remove_cvref_t<OutputIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , THRUST_FWD(init)
+    , thrust::plus<remove_cvref_t<T>>{}
+    )
+  )
+
+  template <
+    typename ForwardIt, typename Sentinel, typename OutputIt
+  >
+  __host__
+  static auto call(
+    ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  )
+  THRUST_RETURNS(
+    reduce_into_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      , typename iterator_system<remove_cvref_t<OutputIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type{}
+    , thrust::plus<
+        remove_cvref_t<
+          typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+        >
+      >{}
+    )
+  )
+
+  // MSVC WAR: MSVC gets angsty and eats all available RAM when we try to detect
+  // if T1 is an execution_policy by using SFINAE. Switching to a static
+  // dispatch pattern to prevent this.
+  template <typename T1, typename T2, typename T3, typename T4>
+  __host__
+  static auto call(T1&& t1, T2&& t2, T3&& t3, T4&& t4)
+  THRUST_RETURNS(
+    reduce_into_fn::call4(
+      THRUST_FWD(t1), THRUST_FWD(t2), THRUST_FWD(t3), THRUST_FWD(t4),
+      thrust::is_execution_policy<thrust::remove_cvref_t<T1>>{})
+  )
+
+  template <typename T1, typename T2, typename T3, typename T4, typename T5>
+  __host__
+  static auto call(T1&& t1, T2&& t2, T3&& t3, T4&& t4, T5&& t5)
+  THRUST_RETURNS(
+    reduce_into_fn::call5(
+      THRUST_FWD(t1), THRUST_FWD(t2), THRUST_FWD(t3), THRUST_FWD(t4),
+      THRUST_FWD(t5), thrust::is_execution_policy<thrust::remove_cvref_t<T1>>{})
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace reduce_into_detail
+
+THRUST_INLINE_CONSTANT reduce_into_detail::reduce_into_fn reduce_into{};
+
+/*! \endcond
+ */
+
+} // namespace async
+
+THRUST_NAMESPACE_END
+
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/async/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/async/scan.h
new file mode 100644
index 000000000000..1bcf81257807
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/async/scan.h
@@ -0,0 +1,344 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file async/scan.h
+ *  \brief Functions for asynchronously computing prefix scans.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/select_system.h>
+#include <thrust/detail/static_assert.h>
+
+#include <thrust/system/detail/adl/async/scan.h>
+
+#include <thrust/type_traits/is_execution_policy.h>
+#include <thrust/type_traits/logical_metafunctions.h>
+#include <thrust/type_traits/remove_cvref.h>
+
+#include <thrust/future.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace async
+{
+
+// Fallback implementations used when no overloads are found via ADL:
+namespace unimplemented
+{
+
+template <typename DerivedPolicy,
+          typename ForwardIt,
+          typename Sentinel,
+          typename OutputIt,
+          typename BinaryOp>
+event<DerivedPolicy>
+async_inclusive_scan(thrust::execution_policy<DerivedPolicy>&,
+                     ForwardIt,
+                     Sentinel,
+                     OutputIt,
+                     BinaryOp)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value),
+    "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+template <typename DerivedPolicy,
+          typename ForwardIt,
+          typename Sentinel,
+          typename OutputIt,
+          typename InitialValueType,
+          typename BinaryOp>
+event<DerivedPolicy>
+async_exclusive_scan(thrust::execution_policy<DerivedPolicy>&,
+                     ForwardIt,
+                     Sentinel,
+                     OutputIt,
+                     InitialValueType,
+                     BinaryOp)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value),
+    "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace inclusive_scan_detail
+{
+
+// Include fallback implementation for ADL failures
+using thrust::async::unimplemented::async_inclusive_scan;
+
+// Implementation of the thrust::async::inclusive_scan CPO.
+struct inclusive_scan_fn final
+{
+  template <typename DerivedPolicy,
+            typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt,
+            typename BinaryOp>
+  auto
+  operator()(thrust::detail::execution_policy_base<DerivedPolicy> const& exec,
+             ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out,
+             BinaryOp&& op) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_inclusive_scan(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      THRUST_FWD(op)
+    )
+  )
+
+  template <typename DerivedPolicy,
+            typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt>
+  auto
+  operator()(thrust::detail::execution_policy_base<DerivedPolicy> const& exec,
+             ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_inclusive_scan(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      thrust::plus<>{}
+    )
+  )
+
+  template <typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt,
+            typename BinaryOp,
+            typename = std::enable_if_t<!is_execution_policy_v<remove_cvref_t<ForwardIt>>>>
+  auto operator()(ForwardIt&& first,
+                  Sentinel&& last,
+                  OutputIt&& out,
+                  BinaryOp&& op) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_inclusive_scan(
+      thrust::detail::select_system(
+        iterator_system_t<remove_cvref_t<ForwardIt>>{},
+        iterator_system_t<remove_cvref_t<OutputIt>>{}
+      ),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      THRUST_FWD(op)
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename OutputIt>
+  auto operator()(ForwardIt&& first, Sentinel&& last, OutputIt&& out) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_inclusive_scan(
+      thrust::detail::select_system(
+        iterator_system_t<remove_cvref_t<ForwardIt>>{},
+        iterator_system_t<remove_cvref_t<OutputIt>>{}
+      ),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      thrust::plus<>{}
+    )
+  )
+};
+
+} // namespace inclusive_scan_detail
+
+THRUST_INLINE_CONSTANT inclusive_scan_detail::inclusive_scan_fn inclusive_scan{};
+
+namespace exclusive_scan_detail
+{
+
+// Include fallback implementation for ADL failures
+using thrust::async::unimplemented::async_exclusive_scan;
+
+// Implementation of the thrust::async::exclusive_scan CPO.
+struct exclusive_scan_fn final
+{
+  template <typename DerivedPolicy,
+            typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt,
+            typename InitialValueType,
+            typename BinaryOp>
+  auto
+  operator()(thrust::detail::execution_policy_base<DerivedPolicy> const& exec,
+             ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out,
+             InitialValueType&& init,
+             BinaryOp&& op) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_exclusive_scan(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      THRUST_FWD(init),
+      THRUST_FWD(op)
+    )
+  )
+
+  template <typename DerivedPolicy,
+            typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt,
+            typename InitialValueType>
+  auto
+  operator()(thrust::detail::execution_policy_base<DerivedPolicy> const& exec,
+             ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out,
+             InitialValueType&& init) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_exclusive_scan(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      THRUST_FWD(init),
+      thrust::plus<>{}
+    )
+  )
+
+  template <typename DerivedPolicy,
+            typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt>
+  auto
+  operator()(thrust::detail::execution_policy_base<DerivedPolicy> const& exec,
+             ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_exclusive_scan(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      iterator_value_t<remove_cvref_t<ForwardIt>>{},
+      thrust::plus<>{}
+    )
+  )
+
+  template <typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt,
+            typename InitialValueType,
+            typename BinaryOp,
+            typename = std::enable_if_t<!is_execution_policy_v<remove_cvref_t<ForwardIt>>>>
+  auto
+  operator()(ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out,
+             InitialValueType&& init,
+             BinaryOp&& op) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_exclusive_scan(
+      thrust::detail::select_system(
+        iterator_system_t<remove_cvref_t<ForwardIt>>{},
+        iterator_system_t<remove_cvref_t<OutputIt>>{}
+      ),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      THRUST_FWD(init),
+      THRUST_FWD(op)
+    )
+  )
+
+  template <typename ForwardIt,
+            typename Sentinel,
+            typename OutputIt,
+            typename InitialValueType,
+            typename = std::enable_if_t<!is_execution_policy_v<remove_cvref_t<ForwardIt>>>>
+  auto
+  operator()(ForwardIt&& first,
+             Sentinel&& last,
+             OutputIt&& out,
+             InitialValueType&& init) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_exclusive_scan(
+      thrust::detail::select_system(
+        iterator_system_t<remove_cvref_t<ForwardIt>>{},
+        iterator_system_t<remove_cvref_t<OutputIt>>{}
+      ),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      THRUST_FWD(init),
+      thrust::plus<>{}
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename OutputIt>
+  auto operator()(ForwardIt&& first,
+                  Sentinel&& last,
+                  OutputIt&& out) const
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_exclusive_scan(
+      thrust::detail::select_system(
+        iterator_system_t<remove_cvref_t<ForwardIt>>{},
+        iterator_system_t<remove_cvref_t<OutputIt>>{}
+      ),
+      THRUST_FWD(first),
+      THRUST_FWD(last),
+      THRUST_FWD(out),
+      iterator_value_t<remove_cvref_t<ForwardIt>>{},
+      thrust::plus<>{}
+    )
+  )
+};
+
+} // namespace exclusive_scan_detail
+
+THRUST_INLINE_CONSTANT exclusive_scan_detail::exclusive_scan_fn exclusive_scan{};
+
+} // namespace async
+
+THRUST_NAMESPACE_END
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/async/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/async/sort.h
new file mode 100644
index 000000000000..8881793970bf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/async/sort.h
@@ -0,0 +1,280 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Algorithms for asynchronously sorting a range.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/select_system.h>
+#include <thrust/type_traits/logical_metafunctions.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/type_traits/is_execution_policy.h>
+#include <thrust/system/detail/adl/async/sort.h>
+
+#include <thrust/event.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace async
+{
+
+/*! \cond
+ */
+
+namespace unimplemented
+{
+
+template <
+  typename DerivedPolicy
+, typename ForwardIt, typename Sentinel, typename StrictWeakOrdering
+>
+__host__
+event<DerivedPolicy>
+async_stable_sort(
+  thrust::execution_policy<DerivedPolicy>&
+, ForwardIt, Sentinel, StrictWeakOrdering
+)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value)
+  , "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace stable_sort_detail
+{
+
+using thrust::async::unimplemented::async_stable_sort;
+
+struct stable_sort_fn final
+{
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename StrictWeakOrdering
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , StrictWeakOrdering&& comp
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_stable_sort(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(comp)
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_stable_sort(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , thrust::less<
+        typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+      >{}
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename StrictWeakOrdering>
+  __host__
+  static auto call(ForwardIt&& first, Sentinel&& last, StrictWeakOrdering&& comp)
+  THRUST_RETURNS(
+    stable_sort_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(comp)
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel>
+  __host__
+  static auto call(ForwardIt&& first, Sentinel&& last)
+  THRUST_RETURNS(
+    stable_sort_fn::call(
+      THRUST_FWD(first), THRUST_FWD(last)
+    , thrust::less<
+        typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+      >{}
+    )
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace stable_sort_detail
+
+THRUST_INLINE_CONSTANT stable_sort_detail::stable_sort_fn stable_sort{};
+
+namespace fallback
+{
+
+template <
+  typename DerivedPolicy
+, typename ForwardIt, typename Sentinel, typename StrictWeakOrdering
+>
+__host__
+event<DerivedPolicy>
+async_sort(
+  thrust::execution_policy<DerivedPolicy>& exec
+, ForwardIt&& first, Sentinel&& last, StrictWeakOrdering&& comp
+)
+{
+  return async_stable_sort(
+    thrust::detail::derived_cast(exec)
+  , THRUST_FWD(first), THRUST_FWD(last), THRUST_FWD(comp)
+  );
+}
+
+} // namespace fallback
+
+namespace sort_detail
+{
+
+using thrust::async::fallback::async_sort;
+
+struct sort_fn final
+{
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename StrictWeakOrdering
+  >
+  __host__
+  static auto call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , StrictWeakOrdering&& comp
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_sort(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(comp)
+    )
+  )
+
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel
+  >
+  __host__
+  static auto call3(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , thrust::true_type
+  )
+  THRUST_RETURNS(
+    sort_fn::call(
+      exec
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , thrust::less<
+        typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+      >{}
+    )
+  )
+
+  template <typename ForwardIt, typename Sentinel, typename StrictWeakOrdering>
+  __host__
+  static auto call3(ForwardIt&& first, Sentinel&& last,
+                    StrictWeakOrdering&& comp,
+                    thrust::false_type)
+  THRUST_RETURNS(
+    sort_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(comp)
+    )
+  )
+
+  // MSVC WAR: MSVC gets angsty and eats all available RAM when we try to detect
+  // if T1 is an execution_policy by using SFINAE. Switching to a static
+  // dispatch pattern to prevent this.
+  template <typename T1, typename T2, typename T3>
+  __host__
+  static auto call(T1&& t1, T2&& t2, T3&& t3)
+  THRUST_RETURNS(
+    sort_fn::call3(THRUST_FWD(t1), THRUST_FWD(t2), THRUST_FWD(t3),
+                   thrust::is_execution_policy<thrust::remove_cvref_t<T1>>{})
+  )
+
+  template <typename ForwardIt, typename Sentinel>
+  __host__
+  static auto call(ForwardIt&& first, Sentinel&& last)
+  THRUST_RETURNS(
+    sort_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , thrust::less<
+        typename iterator_traits<remove_cvref_t<ForwardIt>>::value_type
+      >{}
+    )
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace sort_detail
+
+THRUST_INLINE_CONSTANT sort_detail::sort_fn sort{};
+
+/*! \endcond
+ */
+
+} // namespace async
+
+THRUST_NAMESPACE_END
+
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/async/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/async/transform.h
new file mode 100644
index 000000000000..de72549bf01b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/async/transform.h
@@ -0,0 +1,138 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Algorithms for asynchronously transforming a range.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/select_system.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/system/detail/adl/async/transform.h>
+
+#include <thrust/event.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace async
+{
+
+/*! \cond
+ */
+
+namespace unimplemented
+{
+
+template <
+  typename DerivedPolicy
+, typename ForwardIt, typename Sentinel, typename OutputIt
+, typename UnaryOperation
+>
+__host__
+event<DerivedPolicy>
+async_transform(
+  thrust::execution_policy<DerivedPolicy>& exec
+, ForwardIt first, Sentinel last, OutputIt output, UnaryOperation op
+)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<ForwardIt, false>::value)
+  , "this algorithm is not implemented for the specified system"
+  );
+  return {};
+}
+
+} // namespace unimplemented
+
+namespace transform_detail
+{
+
+using thrust::async::unimplemented::async_transform;
+
+struct transform_fn final
+{
+  template <
+    typename DerivedPolicy
+  , typename ForwardIt, typename Sentinel, typename OutputIt
+  , typename UnaryOperation
+  >
+  __host__
+  static auto
+  call(
+    thrust::detail::execution_policy_base<DerivedPolicy> const& exec
+  , ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , UnaryOperation&& op
+  )
+  // ADL dispatch.
+  THRUST_RETURNS(
+    async_transform(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec))
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , THRUST_FWD(op)
+    )
+  )
+
+  template <
+    typename ForwardIt, typename Sentinel, typename OutputIt
+  , typename UnaryOperation
+  >
+  __host__
+  static auto call(
+    ForwardIt&& first, Sentinel&& last
+  , OutputIt&& output
+  , UnaryOperation&& op
+  )
+  THRUST_RETURNS(
+    transform_fn::call(
+      thrust::detail::select_system(
+        typename iterator_system<remove_cvref_t<ForwardIt>>::type{}
+      , typename iterator_system<remove_cvref_t<OutputIt>>::type{}
+      )
+    , THRUST_FWD(first), THRUST_FWD(last)
+    , THRUST_FWD(output)
+    , THRUST_FWD(op)
+    )
+  )
+
+  template <typename... Args>
+  THRUST_NODISCARD __host__
+  auto operator()(Args&&... args) const
+  THRUST_RETURNS(
+    call(THRUST_FWD(args)...)
+  )
+};
+
+} // namespace tranform_detail
+
+THRUST_INLINE_CONSTANT transform_detail::transform_fn transform{};
+
+/*! \endcond
+ */
+
+} // namespace async
+
+THRUST_NAMESPACE_END
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/binary_search.h b/thirdparty/manifold/thirdparty/thrust/thrust/binary_search.h
new file mode 100644
index 000000000000..7a4746e0b165
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/binary_search.h
@@ -0,0 +1,1899 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file binary_search.h
+ *  \brief Search for values in sorted ranges.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+
+/*! \addtogroup searching
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \addtogroup binary_search Binary Search
+ *  \ingroup searching
+ *  \{
+ */
+
+
+//////////////////////   
+// Scalar Functions //
+//////////////////////
+
+
+/*! \p lower_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the first position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p lower_bound uses <tt>operator<</tt> for comparison and returns
+ * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>*j < value</tt>. 
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return The furthermost iterator \c i, such that <tt>*i < value</tt>.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::lower_bound(thrust::device, input.begin(), input.end(), 0); // returns input.begin()
+ *  thrust::lower_bound(thrust::device, input.begin(), input.end(), 1); // returns input.begin() + 1
+ *  thrust::lower_bound(thrust::device, input.begin(), input.end(), 2); // returns input.begin() + 1
+ *  thrust::lower_bound(thrust::device, input.begin(), input.end(), 3); // returns input.begin() + 2
+ *  thrust::lower_bound(thrust::device, input.begin(), input.end(), 8); // returns input.begin() + 4
+ *  thrust::lower_bound(thrust::device, input.begin(), input.end(), 9); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const LessThanComparable &value);
+
+
+/*! \p lower_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the first position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p lower_bound uses <tt>operator<</tt> for comparison and returns
+ * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>*j < value</tt>. 
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return The furthermost iterator \c i, such that <tt>*i < value</tt>.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::lower_bound(input.begin(), input.end(), 0); // returns input.begin()
+ *  thrust::lower_bound(input.begin(), input.end(), 1); // returns input.begin() + 1
+ *  thrust::lower_bound(input.begin(), input.end(), 2); // returns input.begin() + 1
+ *  thrust::lower_bound(input.begin(), input.end(), 3); // returns input.begin() + 2
+ *  thrust::lower_bound(input.begin(), input.end(), 8); // returns input.begin() + 4
+ *  thrust::lower_bound(input.begin(), input.end(), 9); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class LessThanComparable>
+ForwardIterator lower_bound(ForwardIterator first, 
+                            ForwardIterator last,
+                            const LessThanComparable& value);
+
+
+/*! \p lower_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the first position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p lower_bound uses function object \c comp for comparison 
+ * and returns the furthermost iterator \c i in <tt>[first, last)</tt>
+ * such that, for every iterator \c j in <tt>[first, i)</tt>, 
+ * <tt>comp(*j, value)</tt> is \c true. 
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return The furthermost iterator \c i, such that <tt>comp(*i, value)</tt> is \c true.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::lower_bound(input.begin(), input.end(), 0, thrust::less<int>()); // returns input.begin()
+ *  thrust::lower_bound(input.begin(), input.end(), 1, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::lower_bound(input.begin(), input.end(), 2, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::lower_bound(input.begin(), input.end(), 3, thrust::less<int>()); // returns input.begin() + 2
+ *  thrust::lower_bound(input.begin(), input.end(), 8, thrust::less<int>()); // returns input.begin() + 4
+ *  thrust::lower_bound(input.begin(), input.end(), 9, thrust::less<int>()); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const T &value,
+                            StrictWeakOrdering comp);
+
+
+/*! \p lower_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the first position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p lower_bound uses function object \c comp for comparison 
+ * and returns the furthermost iterator \c i in <tt>[first, last)</tt>
+ * such that, for every iterator \c j in <tt>[first, i)</tt>, 
+ * <tt>comp(*j, value)</tt> is \c true. 
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return The furthermost iterator \c i, such that <tt>comp(*i, value)</tt> is \c true.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::lower_bound(input.begin(), input.end(), 0, thrust::less<int>()); // returns input.begin()
+ *  thrust::lower_bound(input.begin(), input.end(), 1, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::lower_bound(input.begin(), input.end(), 2, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::lower_bound(input.begin(), input.end(), 3, thrust::less<int>()); // returns input.begin() + 2
+ *  thrust::lower_bound(input.begin(), input.end(), 8, thrust::less<int>()); // returns input.begin() + 4
+ *  thrust::lower_bound(input.begin(), input.end(), 9, thrust::less<int>()); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class T, class StrictWeakOrdering>
+ForwardIterator lower_bound(ForwardIterator first,
+                            ForwardIterator last,
+                            const T& value, 
+                            StrictWeakOrdering comp);
+
+
+/*! \p upper_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the last position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p upper_bound uses <tt>operator<</tt> for comparison and returns
+ * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>value < *j</tt>
+ * is \c false.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return The furthermost iterator \c i, such that <tt>value < *i</tt> is \c false.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelism:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 0); // returns input.begin() + 1
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 1); // returns input.begin() + 1
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 2); // returns input.begin() + 2
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 3); // returns input.begin() + 2
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 8); // returns input.end()
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 9); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p lower_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const LessThanComparable &value);
+
+
+/*! \p upper_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the last position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p upper_bound uses <tt>operator<</tt> for comparison and returns
+ * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>value < *j</tt>
+ * is \c false.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return The furthermost iterator \c i, such that <tt>value < *i</tt> is \c false.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::upper_bound(input.begin(), input.end(), 0); // returns input.begin() + 1
+ *  thrust::upper_bound(input.begin(), input.end(), 1); // returns input.begin() + 1
+ *  thrust::upper_bound(input.begin(), input.end(), 2); // returns input.begin() + 2
+ *  thrust::upper_bound(input.begin(), input.end(), 3); // returns input.begin() + 2
+ *  thrust::upper_bound(input.begin(), input.end(), 8); // returns input.end()
+ *  thrust::upper_bound(input.begin(), input.end(), 9); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p lower_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class LessThanComparable>
+ForwardIterator upper_bound(ForwardIterator first, 
+                            ForwardIterator last,
+                            const LessThanComparable& value);
+
+
+/*! \p upper_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the last position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p upper_bound uses function object \c comp for comparison and returns
+ * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>comp(value, *j)</tt>
+ * is \c false.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return The furthermost iterator \c i, such that <tt>comp(value, *i)</tt> is \c false.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 0, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 1, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 2, thrust::less<int>()); // returns input.begin() + 2
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 3, thrust::less<int>()); // returns input.begin() + 2
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 8, thrust::less<int>()); // returns input.end()
+ *  thrust::upper_bound(thrust::device, input.begin(), input.end(), 9, thrust::less<int>()); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p lower_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const T &value,
+                            StrictWeakOrdering comp);
+
+/*! \p upper_bound is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * Specifically, it returns the last position where value could be
+ * inserted without violating the ordering. This version of 
+ * \p upper_bound uses function object \c comp for comparison and returns
+ * the furthermost iterator \c i in <tt>[first, last)</tt> such that,
+ * for every iterator \c j in <tt>[first, i)</tt>, <tt>comp(value, *j)</tt>
+ * is \c false.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return The furthermost iterator \c i, such that <tt>comp(value, *i)</tt> is \c false.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::upper_bound(input.begin(), input.end(), 0, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::upper_bound(input.begin(), input.end(), 1, thrust::less<int>()); // returns input.begin() + 1
+ *  thrust::upper_bound(input.begin(), input.end(), 2, thrust::less<int>()); // returns input.begin() + 2
+ *  thrust::upper_bound(input.begin(), input.end(), 3, thrust::less<int>()); // returns input.begin() + 2
+ *  thrust::upper_bound(input.begin(), input.end(), 8, thrust::less<int>()); // returns input.end()
+ *  thrust::upper_bound(input.begin(), input.end(), 9, thrust::less<int>()); // returns input.end()
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p lower_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class T, class StrictWeakOrdering>
+ForwardIterator upper_bound(ForwardIterator first,
+                            ForwardIterator last,
+                            const T& value, 
+                            StrictWeakOrdering comp);
+
+
+/*! \p binary_search is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.  Specifically, this version returns \c true if and only if 
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * <tt>*i < value</tt> and <tt>value < *i</tt> are both \c false.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 0); // returns true
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 1); // returns false
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 2); // returns true
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 3); // returns false
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 8); // returns true
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 9); // returns false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   ForwardIterator first, 
+                   ForwardIterator last,
+                   const LessThanComparable& value);
+
+
+/*! \p binary_search is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.  Specifically, this version returns \c true if and only if 
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * <tt>*i < value</tt> and <tt>value < *i</tt> are both \c false.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::binary_search(input.begin(), input.end(), 0); // returns true
+ *  thrust::binary_search(input.begin(), input.end(), 1); // returns false
+ *  thrust::binary_search(input.begin(), input.end(), 2); // returns true
+ *  thrust::binary_search(input.begin(), input.end(), 3); // returns false
+ *  thrust::binary_search(input.begin(), input.end(), 8); // returns true
+ *  thrust::binary_search(input.begin(), input.end(), 9); // returns false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <class ForwardIterator, class LessThanComparable>
+bool binary_search(ForwardIterator first, 
+                   ForwardIterator last,
+                   const LessThanComparable& value);
+
+
+/*! \p binary_search is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.  Specifically, this version returns \c true if and only if 
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * <tt>comp(*i, value)</tt> and <tt>comp(value, *i)</tt> are both \c false.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 0, thrust::less<int>()); // returns true
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 1, thrust::less<int>()); // returns false
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 2, thrust::less<int>()); // returns true
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 3, thrust::less<int>()); // returns false
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 8, thrust::less<int>()); // returns true
+ *  thrust::binary_search(thrust::device, input.begin(), input.end(), 9, thrust::less<int>()); // returns false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   ForwardIterator first,
+                   ForwardIterator last,
+                   const T& value, 
+                   StrictWeakOrdering comp);
+
+
+/*! \p binary_search is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. 
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.  Specifically, this version returns \c true if and only if 
+ * there exists an iterator \c i in <tt>[first, last)</tt> such that 
+ * <tt>comp(*i, value)</tt> and <tt>comp(value, *i)</tt> are both \c false.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return \c true if an equivalent element exists in <tt>[first, last)</tt>, otherwise \c false.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::binary_search(input.begin(), input.end(), 0, thrust::less<int>()); // returns true
+ *  thrust::binary_search(input.begin(), input.end(), 1, thrust::less<int>()); // returns false
+ *  thrust::binary_search(input.begin(), input.end(), 2, thrust::less<int>()); // returns true
+ *  thrust::binary_search(input.begin(), input.end(), 3, thrust::less<int>()); // returns false
+ *  thrust::binary_search(input.begin(), input.end(), 8, thrust::less<int>()); // returns true
+ *  thrust::binary_search(input.begin(), input.end(), 9, thrust::less<int>()); // returns false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <class ForwardIterator, class T, class StrictWeakOrdering>
+bool binary_search(ForwardIterator first,
+                   ForwardIterator last,
+                   const T& value, 
+                   StrictWeakOrdering comp);
+
+
+/*! \p equal_range is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * value returned by \p equal_range is essentially a combination of
+ * the values returned by \p lower_bound and \p upper_bound: it returns
+ * a \p pair of iterators \c i and \c j such that \c i is the first
+ * position where value could be inserted without violating the 
+ * ordering and \c j is the last position where value could be inserted
+ * without violating the ordering. It follows that every element in the
+ * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
+ * has this property. 
+ *
+ * This version of \p equal_range returns a \p pair of iterators 
+ * <tt>[i, j)</tt>, where \c i is the furthermost iterator in 
+ * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * <tt>[first, i)</tt>, <tt>*k < value</tt>.  \c j is the furthermost
+ * iterator in <tt>[first, last)</tt> such that, for every iterator 
+ * \c k in <tt>[first, j)</tt>, <tt>value < *k</tt> is \c false. 
+ * For every iterator \c k in <tt>[i, j)</tt>, neither 
+ * <tt>value < *k</tt> nor <tt>*k < value</tt> is \c true.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p equal_range
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 0); // returns [input.begin(), input.begin() + 1)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 1); // returns [input.begin() + 1, input.begin() + 1)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 2); // returns [input.begin() + 1, input.begin() + 2)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 3); // returns [input.begin() + 2, input.begin() + 2)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 8); // returns [input.begin() + 4, input.end)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 9); // returns [input.end(), input.end)
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal_range
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p binary_search
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable& value);
+
+
+/*! \p equal_range is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * value returned by \p equal_range is essentially a combination of
+ * the values returned by \p lower_bound and \p upper_bound: it returns
+ * a \p pair of iterators \c i and \c j such that \c i is the first
+ * position where value could be inserted without violating the 
+ * ordering and \c j is the last position where value could be inserted
+ * without violating the ordering. It follows that every element in the
+ * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
+ * has this property. 
+ *
+ * This version of \p equal_range returns a \p pair of iterators 
+ * <tt>[i, j)</tt>, where \c i is the furthermost iterator in 
+ * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * <tt>[first, i)</tt>, <tt>*k < value</tt>.  \c j is the furthermost
+ * iterator in <tt>[first, last)</tt> such that, for every iterator 
+ * \c k in <tt>[first, j)</tt>, <tt>value < *k</tt> is \c false. 
+ * For every iterator \c k in <tt>[i, j)</tt>, neither 
+ * <tt>value < *k</tt> nor <tt>*k < value</tt> is \c true.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam LessThanComparable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>. 
+ *
+ *  The following code snippet demonstrates how to use \p equal_range
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::equal_range(input.begin(), input.end(), 0); // returns [input.begin(), input.begin() + 1)
+ *  thrust::equal_range(input.begin(), input.end(), 1); // returns [input.begin() + 1, input.begin() + 1)
+ *  thrust::equal_range(input.begin(), input.end(), 2); // returns [input.begin() + 1, input.begin() + 2)
+ *  thrust::equal_range(input.begin(), input.end(), 3); // returns [input.begin() + 2, input.begin() + 2)
+ *  thrust::equal_range(input.begin(), input.end(), 8); // returns [input.begin() + 4, input.end)
+ *  thrust::equal_range(input.begin(), input.end(), 9); // returns [input.end(), input.end)
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal_range
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class LessThanComparable>
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable& value);
+
+
+/*! \p equal_range is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * value returned by \p equal_range is essentially a combination of
+ * the values returned by \p lower_bound and \p upper_bound: it returns
+ * a \p pair of iterators \c i and \c j such that \c i is the first
+ * position where value could be inserted without violating the 
+ * ordering and \c j is the last position where value could be inserted
+ * without violating the ordering. It follows that every element in the
+ * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
+ * has this property. 
+ *
+ * This version of \p equal_range returns a \p pair of iterators 
+ * <tt>[i, j)</tt>. \c i is the furthermost iterator in 
+ * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * <tt>[first, i)</tt>, <tt>comp(*k, value)</tt> is \c true.
+ * \c j is the furthermost iterator in <tt>[first, last)</tt> such
+ * that, for every iterator \c k in <tt>[first, last)</tt>, 
+ * <tt>comp(value, *k)</tt> is \c false. For every iterator \c k 
+ * in <tt>[i, j)</tt>, neither <tt>comp(value, *k)</tt> nor 
+ * <tt>comp(*k, value)</tt> is \c true.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p equal_range
+ *  to search for values in a ordered range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 0, thrust::less<int>()); // returns [input.begin(), input.begin() + 1)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 1, thrust::less<int>()); // returns [input.begin() + 1, input.begin() + 1)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 2, thrust::less<int>()); // returns [input.begin() + 1, input.begin() + 2)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 3, thrust::less<int>()); // returns [input.begin() + 2, input.begin() + 2)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 8, thrust::less<int>()); // returns [input.begin() + 4, input.end)
+ *  thrust::equal_range(thrust::device, input.begin(), input.end(), 9, thrust::less<int>()); // returns [input.end(), input.end)
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal_range
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p binary_search
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const T& value,
+            StrictWeakOrdering comp);
+
+
+/*! \p equal_range is a version of binary search: it attempts to find
+ * the element value in an ordered range <tt>[first, last)</tt>. The 
+ * value returned by \p equal_range is essentially a combination of
+ * the values returned by \p lower_bound and \p upper_bound: it returns
+ * a \p pair of iterators \c i and \c j such that \c i is the first
+ * position where value could be inserted without violating the 
+ * ordering and \c j is the last position where value could be inserted
+ * without violating the ordering. It follows that every element in the
+ * range <tt>[i, j)</tt> is equivalent to value, and that 
+ * <tt>[i, j)</tt> is the largest subrange of <tt>[first, last)</tt> that
+ * has this property. 
+ *
+ * This version of \p equal_range returns a \p pair of iterators 
+ * <tt>[i, j)</tt>. \c i is the furthermost iterator in 
+ * <tt>[first, last)</tt> such that, for every iterator \c k in 
+ * <tt>[first, i)</tt>, <tt>comp(*k, value)</tt> is \c true.
+ * \c j is the furthermost iterator in <tt>[first, last)</tt> such
+ * that, for every iterator \c k in <tt>[first, last)</tt>, 
+ * <tt>comp(value, *k)</tt> is \c false. For every iterator \c k 
+ * in <tt>[i, j)</tt>, neither <tt>comp(value, *k)</tt> nor 
+ * <tt>comp(*k, value)</tt> is \c true.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param value The value to be searched.
+ *  \param comp The comparison operator.
+ *  \return A \p pair of iterators <tt>[i, j)</tt> that define the range of equivalent elements.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam T is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p equal_range
+ *  to search for values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::equal_range(input.begin(), input.end(), 0, thrust::less<int>()); // returns [input.begin(), input.begin() + 1)
+ *  thrust::equal_range(input.begin(), input.end(), 1, thrust::less<int>()); // returns [input.begin() + 1, input.begin() + 1)
+ *  thrust::equal_range(input.begin(), input.end(), 2, thrust::less<int>()); // returns [input.begin() + 1, input.begin() + 2)
+ *  thrust::equal_range(input.begin(), input.end(), 3, thrust::less<int>()); // returns [input.begin() + 2, input.begin() + 2)
+ *  thrust::equal_range(input.begin(), input.end(), 8, thrust::less<int>()); // returns [input.begin() + 4, input.end)
+ *  thrust::equal_range(input.begin(), input.end(), 9, thrust::less<int>()); // returns [input.end(), input.end)
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal_range
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class T, class StrictWeakOrdering>
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(ForwardIterator first,
+            ForwardIterator last,
+            const T& value,
+            StrictWeakOrdering comp);
+
+
+/*! \addtogroup vectorized_binary_search Vectorized Searches
+ *  \ingroup binary_search
+ *  \{
+ */
+
+
+//////////////////////
+// Vector Functions //
+//////////////////////
+
+
+/*! \p lower_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of first position where value could
+ * be inserted without violating the ordering.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for multiple values in a ordered range using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::lower_bound(thrust::device,
+ *                      input.begin(), input.end(),
+ *                      values.begin(), values.end(),
+ *                      output.begin());
+ *
+ *  // output is now [0, 1, 1, 2, 4, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result);
+
+
+/*! \p lower_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of first position where value could
+ * be inserted without violating the ordering.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::lower_bound(input.begin(), input.end(),
+ *                      values.begin(), values.end(),
+ *                      output.begin());
+ *
+ *  // output is now [0, 1, 1, 2, 4, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class InputIterator, class OutputIterator>
+OutputIterator lower_bound(ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result);
+
+
+/*! \p lower_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of first position where value could
+ * be inserted without violating the ordering.  This version of 
+ * \p lower_bound uses function object \c comp for comparison.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param comp The comparison operator.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::lower_bound(input.begin(), input.end(),
+ *                      values.begin(), values.end(), 
+ *                      output.begin(),
+ *                      thrust::less<int>());
+ *
+ *  // output is now [0, 1, 1, 2, 4, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result,
+                           StrictWeakOrdering comp);
+
+
+/*! \p lower_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of first position where value could
+ * be inserted without violating the ordering.  This version of 
+ * \p lower_bound uses function object \c comp for comparison.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param comp The comparison operator.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p lower_bound
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::lower_bound(input.begin(), input.end(),
+ *                      values.begin(), values.end(), 
+ *                      output.begin(),
+ *                      thrust::less<int>());
+ *
+ *  // output is now [0, 1, 1, 2, 4, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class InputIterator, class OutputIterator, class StrictWeakOrdering>
+OutputIterator lower_bound(ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result,
+                           StrictWeakOrdering comp);
+
+
+/*! \p upper_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of last position where value could
+ * be inserted without violating the ordering.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for multiple values in a ordered range using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::upper_bound(thrust::device,
+ *                      input.begin(), input.end(),
+ *                      values.begin(), values.end(),
+ *                      output.begin());
+ *
+ *  // output is now [1, 1, 2, 2, 5, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result);
+
+
+/*! \p upper_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of last position where value could
+ * be inserted without violating the ordering.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::upper_bound(input.begin(), input.end(),
+ *                      values.begin(), values.end(),
+ *                      output.begin());
+ *
+ *  // output is now [1, 1, 2, 2, 5, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class InputIterator, class OutputIterator>
+OutputIterator upper_bound(ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result);
+
+
+/*! \p upper_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of first position where value could
+ * be inserted without violating the ordering.  This version of 
+ * \p upper_bound uses function object \c comp for comparison.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param comp The comparison operator.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for multiple values in a ordered range using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::upper_bound(thrust::device,
+ *                      input.begin(), input.end(),
+ *                      values.begin(), values.end(), 
+ *                      output.begin(),
+ *                      thrust::less<int>());
+ *
+ *  // output is now [1, 1, 2, 2, 5, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p lower_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result,
+                           StrictWeakOrdering comp);
+
+
+/*! \p upper_bound is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * Specifically, it returns the index of first position where value could
+ * be inserted without violating the ordering.  This version of 
+ * \p upper_bound uses function object \c comp for comparison.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param comp The comparison operator.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is comparable to \p ForwardIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and \c ForwardIterator's difference_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p upper_bound
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<unsigned int> output(6);
+ *
+ *  thrust::upper_bound(input.begin(), input.end(),
+ *                      values.begin(), values.end(), 
+ *                      output.begin(),
+ *                      thrust::less<int>());
+ *
+ *  // output is now [1, 1, 2, 2, 5, 5]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/upper_bound
+ *  \see \p lower_bound
+ *  \see \p equal_range
+ *  \see \p binary_search
+ */
+template <class ForwardIterator, class InputIterator, class OutputIterator, class StrictWeakOrdering>
+OutputIterator upper_bound(ForwardIterator first, 
+                           ForwardIterator last,
+                           InputIterator values_first, 
+                           InputIterator values_last,
+                           OutputIterator result,
+                           StrictWeakOrdering comp);
+
+
+/*! \p binary_search is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and bool is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for multiple values in a ordered range using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<bool> output(6);
+ *
+ *  thrust::binary_search(thrust::device,
+ *                        input.begin(), input.end(),
+ *                        values.begin(), values.end(),
+ *                        output.begin());
+ *
+ *  // output is now [true, false, true, false, true, false]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                             ForwardIterator first, 
+                             ForwardIterator last,
+                             InputIterator values_first, 
+                             InputIterator values_last,
+                             OutputIterator result);
+
+
+/*! \p binary_search is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and bool is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<bool> output(6);
+ *
+ *  thrust::binary_search(input.begin(), input.end(),
+ *                        values.begin(), values.end(),
+ *                        output.begin());
+ *
+ *  // output is now [true, false, true, false, true, false]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <class ForwardIterator, class InputIterator, class OutputIterator>
+OutputIterator binary_search(ForwardIterator first, 
+                             ForwardIterator last,
+                             InputIterator values_first, 
+                             InputIterator values_last,
+                             OutputIterator result);
+
+
+/*! \p binary_search is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.  This version of \p binary_search uses function object 
+ * \c comp for comparison.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param comp The comparison operator.
+ * 
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and bool is convertible to \c OutputIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for multiple values in a ordered range using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<bool> output(6);
+ *
+ *  thrust::binary_search(thrust::device,
+ *                        input.begin(), input.end(),
+ *                        values.begin(), values.end(),
+ *                        output.begin(),
+ *                        thrust::less<T>());
+ *
+ *  // output is now [true, false, true, false, true, false]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                             ForwardIterator first, 
+                             ForwardIterator last,
+                             InputIterator values_first, 
+                             InputIterator values_last,
+                             OutputIterator result,
+                             StrictWeakOrdering comp);
+
+
+/*! \p binary_search is a vectorized version of binary search: for each 
+ * iterator \c v in <tt>[values_first, values_last)</tt> it attempts to
+ * find the value <tt>*v</tt> in an ordered range <tt>[first, last)</tt>.
+ * It returns \c true if an element that is equivalent to \c value 
+ * is present in <tt>[first, last)</tt> and \c false if no such element
+ * exists.  This version of \p binary_search uses function object 
+ * \c comp for comparison.
+ *
+ *  \param first The beginning of the ordered sequence.
+ *  \param last The end of the ordered sequence.
+ *  \param values_first The beginning of the search values sequence.
+ *  \param values_last The end of the search values sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param comp The comparison operator.
+ * 
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *                        and \c InputIterator's \c value_type is <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThanComparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *                        and bool is convertible to \c OutputIterator's \c value_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p binary_search
+ *  to search for multiple values in a ordered range.
+ *
+ *  \code
+ *  #include <thrust/binary_search.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<int> input(5);
+ *
+ *  input[0] = 0;
+ *  input[1] = 2;
+ *  input[2] = 5;
+ *  input[3] = 7;
+ *  input[4] = 8;
+ *
+ *  thrust::device_vector<int> values(6);
+ *  values[0] = 0; 
+ *  values[1] = 1;
+ *  values[2] = 2;
+ *  values[3] = 3;
+ *  values[4] = 8;
+ *  values[5] = 9;
+ *
+ *  thrust::device_vector<bool> output(6);
+ *
+ *  thrust::binary_search(input.begin(), input.end(),
+ *                        values.begin(), values.end(),
+ *                        output.begin(),
+ *                        thrust::less<T>());
+ *
+ *  // output is now [true, false, true, false, true, false]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/binary_search
+ *  \see \p lower_bound
+ *  \see \p upper_bound
+ *  \see \p equal_range
+ */
+template <class ForwardIterator, class InputIterator, class OutputIterator, class StrictWeakOrdering>
+OutputIterator binary_search(ForwardIterator first, 
+                             ForwardIterator last,
+                             InputIterator values_first, 
+                             InputIterator values_last,
+                             OutputIterator result,
+                             StrictWeakOrdering comp);
+
+
+/*! \} // end vectorized_binary_search
+ */
+
+
+/*! \} // end binary_search
+ */
+
+
+/*! \} // end searching
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/binary_search.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/complex.h b/thirdparty/manifold/thirdparty/thrust/thrust/complex.h
new file mode 100644
index 000000000000..8c0be0d61d96
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/complex.h
@@ -0,0 +1,1047 @@
+/*
+ *  Copyright 2008-2019 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file complex.h
+ *  \brief Complex numbers
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <cmath>
+#include <complex>
+#include <sstream>
+#include <thrust/detail/type_traits.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+#  define THRUST_STD_COMPLEX_REAL(z) \
+    reinterpret_cast< \
+      const typename thrust::detail::remove_reference<decltype(z)>::type::value_type (&)[2] \
+    >(z)[0]
+#  define THRUST_STD_COMPLEX_IMAG(z) \
+    reinterpret_cast< \
+      const typename thrust::detail::remove_reference<decltype(z)>::type::value_type (&)[2] \
+    >(z)[1]
+#  define THRUST_STD_COMPLEX_DEVICE __device__
+#else
+#  define THRUST_STD_COMPLEX_REAL(z) (z).real()
+#  define THRUST_STD_COMPLEX_IMAG(z) (z).imag()
+#  define THRUST_STD_COMPLEX_DEVICE
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+/*
+ *  Calls to the standard math library from inside the thrust namespace
+ *  with real arguments require explicit scope otherwise they will fail
+ *  to resolve as it will find the equivalent complex function but then
+ *  fail to match the template, and give up looking for other scopes.
+ */
+
+
+/*! \addtogroup numerics
+ *  \{
+ */
+
+/*! \addtogroup complex_numbers Complex Numbers
+ *  \{
+ */
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T, std::size_t Align>
+struct complex_storage;
+
+#if THRUST_CPP_DIALECT >= 2011                                                    \
+  && (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                       \
+  && (THRUST_GCC_VERSION >= 40800)
+  // C++11 implementation, excluding GCC 4.7, which doesn't have `alignas`.
+  template <typename T, std::size_t Align>
+  struct complex_storage
+  {
+    struct alignas(Align) type { T x; T y; };
+  };
+#elif  (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC)                    \
+    || (   (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                 \
+        && (THRUST_GCC_VERSION < 40600))
+  // C++03 implementation for MSVC and GCC <= 4.5.
+  //
+  // We have to implement `aligned_type` with specializations for MSVC
+  // and GCC 4.2 and older because they require literals as arguments to
+  // their alignment attribute.
+
+  #if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC)
+    // MSVC implementation.
+    #define THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(X)                   \
+      template <typename T>                                                   \
+      struct complex_storage<T, X>                                            \
+      {                                                                       \
+        __declspec(align(X)) struct type { T x; T y; };                       \
+      };                                                                      \
+      /**/
+  #else
+    // GCC <= 4.2 implementation.
+    #define THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(X)                   \
+      template <typename T>                                                   \
+      struct complex_storage<T, X>                                            \
+      {                                                                       \
+        struct type { T x; T y; } __attribute__((aligned(X)));                \
+      };                                                                      \
+      /**/
+  #endif
+
+  // The primary template is a fallback, which doesn't specify any alignment.
+  // It's only used when T is very large and we're using an older compilers
+  // which we have to fully specialize each alignment case.
+  template <typename T, std::size_t Align>
+  struct complex_storage
+  {
+    T x; T y;
+  };
+
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(1);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(2);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(4);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(8);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(16);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(32);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(64);
+  THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION(128);
+
+  #undef THRUST_DEFINE_COMPLEX_STORAGE_SPECIALIZATION
+#else
+  // C++03 implementation for GCC > 4.5, Clang, PGI, ICPC, and xlC.
+  template <typename T, std::size_t Align>
+  struct complex_storage
+  {
+    struct type { T x; T y; } __attribute__((aligned(Align)));
+  };
+#endif
+
+} // end namespace detail
+
+/*! \endcond
+ */
+
+/*! \p complex is the Thrust equivalent to <tt>std::complex</tt>. It is
+ *  functionally identical to it, but can also be used in device code which
+ *  <tt>std::complex</tt> currently cannot.
+ *
+ *  \tparam T The type used to hold the real and imaginary parts. Should be
+ *  <tt>float</tt> or <tt>double</tt>. Others types are not supported.
+ *
+ */
+template <typename T>
+struct complex
+{
+public:
+
+  /*! \p value_type is the type of \p complex's real and imaginary parts.
+   */
+  typedef T value_type;
+
+
+
+  /* --- Constructors --- */
+
+  /*! Construct a complex number with an imaginary part of 0.
+   *
+   *  \param re The real part of the number.
+   */
+  __host__ __device__
+  complex(const T& re);
+
+  /*! Construct a complex number from its real and imaginary parts.
+   *
+   *  \param re The real part of the number.
+   *  \param im The imaginary part of the number.
+   */
+  __host__ __device__
+  complex(const T& re, const T& im);
+
+#if THRUST_CPP_DIALECT >= 2011
+  /*! Default construct a complex number.
+   */
+  complex() = default;
+
+  /*! This copy constructor copies from a \p complex with a type that is
+   *  convertible to this \p complex's \c value_type.
+   *
+   *  \param z The \p complex to copy from.
+   */
+  complex(const complex<T>& z) = default;
+#else
+  /*! Default construct a complex number.
+   */
+  __host__ __device__
+  complex();
+
+  /*! This copy constructor copies from a \p complex with a type that is
+   *  convertible to this \p complex's \c value_type.
+   *
+   *  \param z The \p complex to copy from.
+   */
+  __host__ __device__
+  complex(const complex<T>& z);
+#endif
+
+  /*! This converting copy constructor copies from a \p complex with a type
+   *  that is convertible to this \p complex's \c value_type.
+   *
+   *  \param z The \p complex to copy from.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex(const complex<U>& z);
+
+  /*! This converting copy constructor copies from a <tt>std::complex</tt> with
+   *  a type that is convertible to this \p complex's \c value_type.
+   *
+   *  \param z The \p complex to copy from.
+   */
+  __host__ THRUST_STD_COMPLEX_DEVICE
+  complex(const std::complex<T>& z);
+
+  /*! This converting copy constructor copies from a <tt>std::complex</tt> with
+   *  a type that is convertible to this \p complex's \c value_type.
+   *
+   *  \param z The \p complex to copy from.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ THRUST_STD_COMPLEX_DEVICE
+  complex(const std::complex<U>& z);
+
+
+
+  /* --- Assignment Operators --- */
+
+  /*! Assign `re` to the real part of this \p complex and set the imaginary part
+   *  to 0.
+   *
+   *  \param re The real part of the number.
+   */
+  __host__ __device__
+  complex& operator=(const T& re);
+
+#if THRUST_CPP_DIALECT >= 2011
+  /*! Assign `z.real()` and `z.imag()` to the real and imaginary parts of this
+   *  \p complex respectively.
+   *
+   *  \param z The \p complex to copy from.
+   */
+  complex& operator=(const complex<T>& z) = default;
+#else
+  /*! Assign `z.real()` and `z.imag()` to the real and imaginary parts of this
+   *  \p complex respectively.
+   *
+   *  \param z The \p complex to copy from.
+   */
+  __host__ __device__
+  complex& operator=(const complex<T>& z);
+#endif
+
+  /*! Assign `z.real()` and `z.imag()` to the real and imaginary parts of this
+   *  \p complex respectively.
+   *
+   *  \param z The \p complex to copy from.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex& operator=(const complex<U>& z);
+
+  /*! Assign `z.real()` and `z.imag()` to the real and imaginary parts of this
+   *  \p complex respectively.
+   *
+   *  \param z The \p complex to copy from.
+   */
+  __host__ THRUST_STD_COMPLEX_DEVICE
+  complex& operator=(const std::complex<T>& z);
+
+  /*! Assign `z.real()` and `z.imag()` to the real and imaginary parts of this
+   *  \p complex respectively.
+   *
+   *  \param z The \p complex to copy from.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ THRUST_STD_COMPLEX_DEVICE
+  complex& operator=(const std::complex<U>& z);
+
+
+  /* --- Compound Assignment Operators --- */
+
+  /*! Adds a \p complex to this \p complex and assigns the result to this
+   *  \p complex.
+   *
+   *  \param z The \p complex to be added.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator+=(const complex<U>& z);
+
+  /*! Subtracts a \p complex from this \p complex and assigns the result to
+   *  this \p complex.
+   *
+   *  \param z The \p complex to be subtracted.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator-=(const complex<U>& z);
+
+  /*! Multiplies this \p complex by another \p complex and assigns the result
+   *  to this \p complex.
+   *
+   *  \param z The \p complex to be multiplied.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator*=(const complex<U>& z);
+
+  /*! Divides this \p complex by another \p complex and assigns the result to
+   *  this \p complex.
+   *
+   *  \param z The \p complex to be divided.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator/=(const complex<U>& z);
+
+  /*! Adds a scalar to this \p complex and assigns the result to this
+   *  \p complex.
+   *
+   *  \param z The \p complex to be added.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator+=(const U& z);
+
+  /*! Subtracts a scalar from this \p complex and assigns the result to
+   *  this \p complex.
+   *
+   *  \param z The scalar to be subtracted.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator-=(const U& z);
+
+  /*! Multiplies this \p complex by a scalar and assigns the result
+   *  to this \p complex.
+   *
+   *  \param z The scalar to be multiplied.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator*=(const U& z);
+
+  /*! Divides this \p complex by a scalar and assigns the result to
+   *  this \p complex.
+   *
+   *  \param z The scalar to be divided.
+   *
+   *  \tparam U is convertible to \c value_type.
+   */
+  template <typename U>
+  __host__ __device__
+  complex<T>& operator/=(const U& z);
+
+
+
+  /* --- Getter functions ---
+   * The volatile ones are there to help for example
+   * with certain reductions optimizations
+   */
+
+  /*! Returns the real part of this \p complex.
+   */
+  __host__ __device__
+  T real() const volatile { return data.x; }
+
+  /*! Returns the imaginary part of this \p complex.
+   */
+  __host__ __device__
+  T imag() const volatile { return data.y; }
+
+  /*! Returns the real part of this \p complex.
+   */
+  __host__ __device__
+  T real() const { return data.x; }
+
+  /*! Returns the imaginary part of this \p complex.
+   */
+  __host__ __device__
+  T imag() const { return data.y; }
+
+
+
+  /* --- Setter functions ---
+   * The volatile ones are there to help for example
+   * with certain reductions optimizations
+   */
+
+  /*! Sets the real part of this \p complex.
+   *
+   *  \param re The new real part of this \p complex.
+   */
+  __host__ __device__
+  void real(T re) volatile { data.x = re; }
+
+  /*! Sets the imaginary part of this \p complex.
+   *
+   *  \param im The new imaginary part of this \p complex.e
+   */
+  __host__ __device__
+  void imag(T im) volatile { data.y = im; }
+
+  /*! Sets the real part of this \p complex.
+   *
+   *  \param re The new real part of this \p complex.
+   */
+  __host__ __device__
+  void real(T re) { data.x = re; }
+
+  /*! Sets the imaginary part of this \p complex.
+   *
+   *  \param im The new imaginary part of this \p complex.
+   */
+  __host__ __device__
+  void imag(T im) { data.y = im; }
+
+
+
+  /* --- Casting functions --- */
+
+  /*! Casts this \p complex to a <tt>std::complex</tt> of the same type.
+   */
+  __host__
+  operator std::complex<T>() const { return std::complex<T>(real(), imag()); }
+
+private:
+  typename detail::complex_storage<T, sizeof(T) * 2>::type data;
+};
+
+
+/* --- General Functions --- */
+
+/*! Returns the magnitude (also known as absolute value) of a \p complex.
+ *
+ *  \param z The \p complex from which to calculate the absolute value.
+ */
+template<typename T>
+__host__ __device__
+T abs(const complex<T>& z);
+
+/*! Returns the phase angle (also known as argument) in radians of a \p complex.
+ *
+ *  \param z The \p complex from which to calculate the phase angle.
+ */
+template <typename T>
+__host__ __device__
+T arg(const complex<T>& z);
+
+/*! Returns the square of the magnitude of a \p complex.
+ *
+ *  \param z The \p complex from which to calculate the norm.
+ */
+template <typename T>
+__host__ __device__
+T norm(const complex<T>& z);
+
+/*! Returns the complex conjugate of a \p complex.
+ *
+ *  \param z The \p complex from which to calculate the complex conjugate.
+ */
+template <typename T>
+__host__ __device__
+complex<T> conj(const complex<T>& z);
+
+/*! Returns a \p complex with the specified magnitude and phase.
+ *
+ *  \param m The magnitude of the returned \p complex.
+ *  \param theta The phase of the returned \p complex in radians.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+polar(const T0& m, const T1& theta = T1());
+
+/*! Returns the projection of a \p complex on the Riemann sphere.
+ *  For all finite \p complex it returns the argument. For \p complexs
+ *  with a non finite part returns (INFINITY,+/-0) where the sign of
+ *  the zero matches the sign of the imaginary part of the argument.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> proj(const T& z);
+
+
+
+/* --- Binary Arithmetic operators --- */
+
+/*! Adds two \p complex numbers.
+ *
+ *  The value types of the two \p complex types should be compatible and the
+ *  type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator+(const complex<T0>& x, const complex<T1>& y);
+
+/*! Adds a scalar to a \p complex number.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The \p complex.
+ *  \param y The scalar.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator+(const complex<T0>& x, const T1& y);
+
+/*! Adds a \p complex number to a scalar.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The scalar.
+ *  \param y The \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator+(const T0& x, const complex<T1>& y);
+
+/*! Subtracts two \p complex numbers.
+ *
+ *  The value types of the two \p complex types should be compatible and the
+ *  type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The first \p complex (minuend).
+ *  \param y The second \p complex (subtrahend).
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator-(const complex<T0>& x, const complex<T1>& y);
+
+/*! Subtracts a scalar from a \p complex number.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The \p complex (minuend).
+ *  \param y The scalar (subtrahend).
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator-(const complex<T0>& x, const T1& y);
+
+/*! Subtracts a \p complex number from a scalar.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The scalar (minuend).
+ *  \param y The \p complex (subtrahend).
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator-(const T0& x, const complex<T1>& y);
+
+/*! Multiplies two \p complex numbers.
+ *
+ *  The value types of the two \p complex types should be compatible and the
+ *  type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator*(const complex<T0>& x, const complex<T1>& y);
+
+/*! Multiplies a \p complex number by a scalar.
+ *
+ *  \param x The \p complex.
+ *  \param y The scalar.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator*(const complex<T0>& x, const T1& y);
+
+/*! Multiplies a scalar by a \p complex number.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The scalar.
+ *  \param y The \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator*(const T0& x, const complex<T1>& y);
+
+/*! Divides two \p complex numbers.
+ *
+ *  The value types of the two \p complex types should be compatible and the
+ *  type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The numerator (dividend).
+ *  \param y The denomimator (divisor).
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator/(const complex<T0>& x, const complex<T1>& y);
+
+/*! Divides a \p complex number by a scalar.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The complex numerator (dividend).
+ *  \param y The scalar denomimator (divisor).
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator/(const complex<T0>& x, const T1& y);
+
+/*! Divides a scalar by a \p complex number.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The scalar numerator (dividend).
+ *  \param y The complex denomimator (divisor).
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator/(const T0& x, const complex<T1>& y);
+
+
+
+/* --- Unary Arithmetic operators --- */
+
+/*! Unary plus, returns its \p complex argument.
+ *
+ *  \param y The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T>
+operator+(const complex<T>& y);
+
+/*! Unary minus, returns the additive inverse (negation) of its \p complex
+ * argument.
+ *
+ *  \param y The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T>
+operator-(const complex<T>& y);
+
+
+
+/* --- Exponential Functions --- */
+
+/*! Returns the complex exponential of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> exp(const complex<T>& z);
+
+/*! Returns the complex natural logarithm of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> log(const complex<T>& z);
+
+/*! Returns the complex base 10 logarithm of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> log10(const complex<T>& z);
+
+
+
+/* --- Power Functions --- */
+
+/*! Returns a \p complex number raised to another.
+ *
+ *  The value types of the two \p complex types should be compatible and the
+ *  type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The base.
+ *  \param y The exponent.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+pow(const complex<T0>& x, const complex<T1>& y);
+
+/*! Returns a \p complex number raised to a scalar.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The base.
+ *  \param y The exponent.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+pow(const complex<T0>& x, const T1& y);
+
+/*! Returns a scalar raised to a \p complex number.
+ *
+ *  The value type of the \p complex should be compatible with the scalar and
+ *  the type of the returned \p complex is the promoted type of the two arguments.
+ *
+ *  \param x The base.
+ *  \param y The exponent.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+pow(const T0& x, const complex<T1>& y);
+
+/*! Returns the complex square root of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> sqrt(const complex<T>& z);
+
+
+/* --- Trigonometric Functions --- */
+
+/*! Returns the complex cosine of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> cos(const complex<T>& z);
+
+/*! Returns the complex sine of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> sin(const complex<T>& z);
+
+/*! Returns the complex tangent of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> tan(const complex<T>& z);
+
+
+
+/* --- Hyperbolic Functions --- */
+
+/*! Returns the complex hyperbolic cosine of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> cosh(const complex<T>& z);
+
+/*! Returns the complex hyperbolic sine of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> sinh(const complex<T>& z);
+
+/*! Returns the complex hyperbolic tangent of a \p complex number.
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> tanh(const complex<T>& z);
+
+
+
+/* --- Inverse Trigonometric Functions --- */
+
+/*! Returns the complex arc cosine of a \p complex number.
+ *
+ *  The range of the real part of the result is [0, Pi] and
+ *  the range of the imaginary part is [-inf, +inf]
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> acos(const complex<T>& z);
+
+/*! Returns the complex arc sine of a \p complex number.
+ *
+ *  The range of the real part of the result is [-Pi/2, Pi/2] and
+ *  the range of the imaginary part is [-inf, +inf]
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> asin(const complex<T>& z);
+
+/*! Returns the complex arc tangent of a \p complex number.
+ *
+ *  The range of the real part of the result is [-Pi/2, Pi/2] and
+ *  the range of the imaginary part is [-inf, +inf]
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> atan(const complex<T>& z);
+
+
+
+/* --- Inverse Hyperbolic Functions --- */
+
+/*! Returns the complex inverse hyperbolic cosine of a \p complex number.
+ *
+ *  The range of the real part of the result is [0, +inf] and
+ *  the range of the imaginary part is [-Pi, Pi]
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> acosh(const complex<T>& z);
+
+/*! Returns the complex inverse hyperbolic sine of a \p complex number.
+ *
+ *  The range of the real part of the result is [-inf, +inf] and
+ *  the range of the imaginary part is [-Pi/2, Pi/2]
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> asinh(const complex<T>& z);
+
+/*! Returns the complex inverse hyperbolic tangent of a \p complex number.
+ *
+ *  The range of the real part of the result is [-inf, +inf] and
+ *  the range of the imaginary part is [-Pi/2, Pi/2]
+ *
+ *  \param z The \p complex argument.
+ */
+template <typename T>
+__host__ __device__
+complex<T> atanh(const complex<T>& z);
+
+
+
+/* --- Stream Operators --- */
+
+/*! Writes to an output stream a \p complex number in the form (real, imaginary).
+ *
+ *  \param os The output stream.
+ *  \param z The \p complex number to output.
+ */
+template <typename T, typename CharT, typename Traits>
+std::basic_ostream<CharT, Traits>&
+operator<<(std::basic_ostream<CharT, Traits>& os, const complex<T>& z);
+
+/*! Reads a \p complex number from an input stream.
+ *
+ *  The recognized formats are:
+ * - real
+ * - (real)
+ * - (real, imaginary)
+ *
+ * The values read must be convertible to the \p complex's \c value_type
+ *
+ *  \param is The input stream.
+ *  \param z The \p complex number to set.
+ */
+template <typename T, typename CharT, typename Traits>
+__host__
+std::basic_istream<CharT, Traits>&
+operator>>(std::basic_istream<CharT, Traits>& is, complex<T>& z);
+
+
+
+/* --- Equality Operators --- */
+
+/*! Returns true if two \p complex numbers are equal and false otherwise.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+bool operator==(const complex<T0>& x, const complex<T1>& y);
+
+/*! Returns true if two \p complex numbers are equal and false otherwise.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator==(const complex<T0>& x, const std::complex<T1>& y);
+
+/*! Returns true if two \p complex numbers are equal and false otherwise.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator==(const std::complex<T0>& x, const complex<T1>& y);
+
+/*! Returns true if the imaginary part of the \p complex number is zero and
+ *  the real part is equal to the scalar. Returns false otherwise.
+ *
+ *  \param x The scalar.
+ *  \param y The \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+bool operator==(const T0& x, const complex<T1>& y);
+
+/*! Returns true if the imaginary part of the \p complex number is zero and
+ *  the real part is equal to the scalar. Returns false otherwise.
+ *
+ *  \param x The \p complex.
+ *  \param y The scalar.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+bool operator==(const complex<T0>& x, const T1& y);
+
+/*! Returns true if two \p complex numbers are different and false otherwise.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+bool operator!=(const complex<T0>& x, const complex<T1>& y);
+
+/*! Returns true if two \p complex numbers are different and false otherwise.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator!=(const complex<T0>& x, const std::complex<T1>& y);
+
+/*! Returns true if two \p complex numbers are different and false otherwise.
+ *
+ *  \param x The first \p complex.
+ *  \param y The second \p complex.
+ */
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator!=(const std::complex<T0>& x, const complex<T1>& y);
+
+/*! Returns true if the imaginary part of the \p complex number is not zero or
+ *  the real part is different from the scalar. Returns false otherwise.
+ *
+ *  \param x The scalar.
+ *  \param y The \p complex.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+bool operator!=(const T0& x, const complex<T1>& y);
+
+/*! Returns true if the imaginary part of the \p complex number is not zero or
+ *  the real part is different from the scalar. Returns false otherwise.
+ *
+ *  \param x The \p complex.
+ *  \param y The scalar.
+ */
+template <typename T0, typename T1>
+__host__ __device__
+bool operator!=(const complex<T0>& x, const T1& y);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/complex/complex.inl>
+
+#undef THRUST_STD_COMPLEX_REAL
+#undef THRUST_STD_COMPLEX_IMAG
+#undef THRUST_STD_COMPLEX_DEVICE
+
+/*! \} // complex_numbers
+ */
+
+/*! \} // numerics
+ */
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/copy.h
new file mode 100644
index 000000000000..99d488174dd6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/copy.h
@@ -0,0 +1,512 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/copy.h
+ *  \brief Copies elements from one range to another
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+/*! \addtogroup copying
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p copy copies elements from the range [\p first, \p last) to the range
+ *  [\p result, \p result + (\p last - \p first)). That is, it performs
+ *  the assignments *\p result = *\p first, *(\p result + \c 1) = *(\p first + \c 1),
+ *  and so on. Generally, for every integer \c n from \c 0 to \p last - \p first, \p copy
+ *  performs the assignment *(\p result + \c n) = *(\p first + \c n). Unlike
+ *  \c std::copy, \p copy offers no guarantee on order of operation.  As a result,
+ *  calling \p copy with overlapping source and destination ranges has undefined
+ *  behavior.
+ *
+ *  The return value is \p result + (\p last - \p first).
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to copy.
+ *  \param last The end of the sequence to copy.
+ *  \param result The destination sequence.
+ *  \return The end of the destination sequence.
+ *  \see https://en.cppreference.com/w/cpp/algorithm/copy
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre \p result may be equal to \p first, but \p result shall not be in the range <tt>[first, last)</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p copy
+ *  to copy from one range to another using the \p thrust::device parallelization policy:
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  thrust::device_vector<int> vec0(100);
+ *  thrust::device_vector<int> vec1(100);
+ *  ...
+ *
+ *  thrust::copy(thrust::device, vec0.begin(), vec0.end(), vec1.begin());
+ *
+ *  // vec1 is now a copy of vec0
+ *  \endcode
+ */
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      OutputIterator result);
+
+
+/*! \p copy_n copies elements from the range <tt>[first, first + n)</tt> to the range
+ *  <tt>[result, result + n)</tt>. That is, it performs the assignments <tt>*result = *first, *(result + 1) = *(first + 1)</tt>,
+ *  and so on. Generally, for every integer \c i from \c 0 to \c n, \p copy
+ *  performs the assignment *(\p result + \c i) = *(\p first + \c i). Unlike
+ *  \c std::copy_n, \p copy_n offers no guarantee on order of operation. As a result,
+ *  calling \p copy_n with overlapping source and destination ranges has undefined
+ *  behavior.
+ *
+ *  The return value is \p result + \p n.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range to copy.
+ *  \param n The number of elements to copy.
+ *  \param result The beginning destination range.
+ *  \return The end of the destination range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam Size is an integral type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre \p result may be equal to \p first, but \p result shall not be in the range <tt>[first, first + n)</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p copy
+ *  to copy from one range to another using the \p thrust::device parallelization policy:
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  size_t n = 100;
+ *  thrust::device_vector<int> vec0(n);
+ *  thrust::device_vector<int> vec1(n);
+ *  ...
+ *  thrust::copy_n(thrust::device, vec0.begin(), n, vec1.begin());
+ *
+ *  // vec1 is now a copy of vec0
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/copy_n
+ *  \see thrust::copy
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Size, typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        InputIterator first,
+                        Size n,
+                        OutputIterator result);
+
+
+	
+/*! \p copy copies elements from the range [\p first, \p last) to the range
+ *  [\p result, \p result + (\p last - \p first)). That is, it performs
+ *  the assignments *\p result = *\p first, *(\p result + \c 1) = *(\p first + \c 1),
+ *  and so on. Generally, for every integer \c n from \c 0 to \p last - \p first, \p copy
+ *  performs the assignment *(\p result + \c n) = *(\p first + \c n). Unlike
+ *  \c std::copy, \p copy offers no guarantee on order of operation.  As a result,
+ *  calling \p copy with overlapping source and destination ranges has undefined
+ *  behavior.
+ *
+ *  The return value is \p result + (\p last - \p first).
+ *
+ *  \param first The beginning of the sequence to copy.
+ *  \param last The end of the sequence to copy.
+ *  \param result The destination sequence.
+ *  \return The end of the destination sequence.
+ *  \see https://en.cppreference.com/w/cpp/algorithm/copy
+ *
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre \p result may be equal to \p first, but \p result shall not be in the range <tt>[first, last)</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p copy
+ *  to copy from one range to another.
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *
+ *  thrust::device_vector<int> vec0(100);
+ *  thrust::device_vector<int> vec1(100);
+ *  ...
+ *
+ *  thrust::copy(vec0.begin(), vec0.end(),
+ *               vec1.begin());
+ *
+ *  // vec1 is now a copy of vec0
+ *  \endcode
+ */
+template<typename InputIterator, typename OutputIterator>
+  OutputIterator copy(InputIterator first,
+                      InputIterator last,
+                      OutputIterator result);
+
+/*! \p copy_n copies elements from the range <tt>[first, first + n)</tt> to the range
+ *  <tt>[result, result + n)</tt>. That is, it performs the assignments <tt>*result = *first, *(result + 1) = *(first + 1)</tt>,
+ *  and so on. Generally, for every integer \c i from \c 0 to \c n, \p copy
+ *  performs the assignment *(\p result + \c i) = *(\p first + \c i). Unlike
+ *  \c std::copy_n, \p copy_n offers no guarantee on order of operation. As a result,
+ *  calling \p copy_n with overlapping source and destination ranges has undefined
+ *  behavior.
+ *
+ *  The return value is \p result + \p n.
+ *
+ *  \param first The beginning of the range to copy.
+ *  \param n The number of elements to copy.
+ *  \param result The beginning destination range.
+ *  \return The end of the destination range.
+ *
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam Size is an integral type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre \p result may be equal to \p first, but \p result shall not be in the range <tt>[first, first + n)</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p copy
+ *  to copy from one range to another.
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  size_t n = 100;
+ *  thrust::device_vector<int> vec0(n);
+ *  thrust::device_vector<int> vec1(n);
+ *  ...
+ *  thrust::copy_n(vec0.begin(), n, vec1.begin());
+ *
+ *  // vec1 is now a copy of vec0
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/copy_n
+ *  \see thrust::copy
+ */
+template<typename InputIterator, typename Size, typename OutputIterator>
+  OutputIterator copy_n(InputIterator first,
+                        Size n,
+                        OutputIterator result);
+
+/*! \} // end copying
+ */
+
+/*! \addtogroup stream_compaction
+ *  \{
+ */
+
+
+/*! This version of \p copy_if copies elements from the range <tt>[first,last)</tt>
+ *  to a range beginning at \p result, except that any element which causes \p pred
+ *  to be \c false is not copied. \p copy_if is stable, meaning that the relative
+ *  order of elements that are copied is unchanged.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>,
+ *  \p copy_if performs the assignment <tt>*result = *(first+n)</tt> and \p result
+ *  is advanced one position if <tt>pred(*(first+n))</tt>. Otherwise, no assignment
+ *  occurs and \p result is not advanced.
+ *
+ *  The algorithm's execution is parallelized as determined by \p system.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence from which to copy.
+ *  \param last The end of the sequence from which to copy.
+ *  \param result The beginning of the sequence into which to copy.
+ *  \param pred The predicate to test on every value of the range <tt>[first, last)</tt>.
+ *  \return <tt>result + n</tt>, where \c n is equal to the number of times \p pred
+ *          evaluated to \c true in the range <tt>[first, last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *                        and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p copy_if to perform stream compaction
+ *  to copy even numbers to an output range using the \p thrust::host parallelization policy:
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[4];
+ *
+ *  thrust::copy_if(thrust::host, V, V + N, result, is_even());
+ *
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-2, 0, 0, 2}
+ *  \endcode
+ *
+ *  \see \c remove_copy_if
+ */
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+
+/*! This version of \p copy_if copies elements from the range <tt>[first,last)</tt>
+ *  to a range beginning at \p result, except that any element which causes \p pred
+ *  to \c false is not copied. \p copy_if is stable, meaning that the relative
+ *  order of elements that are copied is unchanged.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>,
+ *  \p copy_if performs the assignment <tt>*result = *(first+n)</tt> and \p result
+ *  is advanced one position if <tt>pred(*(first+n))</tt>. Otherwise, no assignment
+ *  occurs and \p result is not advanced.
+ *
+ *  \param first The beginning of the sequence from which to copy.
+ *  \param last The end of the sequence from which to copy.
+ *  \param result The beginning of the sequence into which to copy.
+ *  \param pred The predicate to test on every value of the range <tt>[first, last)</tt>.
+ *  \return <tt>result + n</tt>, where \c n is equal to the number of times \p pred
+ *          evaluated to \c true in the range <tt>[first, last)</tt>.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *                        and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p copy_if to perform stream compaction
+ *  to copy even numbers to an output range.
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[4];
+ *
+ *  thrust::copy_if(V, V + N, result, is_even());
+ *
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-2, 0, 0, 2}
+ *  \endcode
+ *
+ *  \see \c remove_copy_if
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator copy_if(InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+/*! This version of \p copy_if copies elements from the range <tt>[first,last)</tt>
+ *  to a range beginning at \p result, except that any element whose corresponding stencil
+ *  element causes \p pred to be \c false is not copied. \p copy_if is stable, meaning
+ *  that the relative order of elements that are copied is unchanged.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>,
+ *  \p copy_if performs the assignment <tt>*result = *(first+n)</tt> and \p result
+ *  is advanced one position if <tt>pred(*(stencil+n))</tt>. Otherwise, no assignment
+ *  occurs and \p result is not advanced.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence from which to copy.
+ *  \param last The end of the sequence from which to copy.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the sequence into which to copy.
+ *  \param pred The predicate to test on every value of the range <tt>[stencil, stencil + (last-first))</tt>.
+ *  \return <tt>result + n</tt>, where \c n is equal to the number of times \p pred
+ *          evaluated to \c true in the range <tt>[stencil, stencil + (last-first))</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *                         and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/OutputIterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *  \pre The ranges <tt>[stencil, stencil + (last - first))</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p copy_if to perform stream compaction
+ *  to copy numbers to an output range when corresponding stencil elements are even using the \p thrust::host execution policy:
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int N = 6;
+ *  int data[N]    = { 0, 1,  2, 3, 4, 5};
+ *  int stencil[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[4];
+ *
+ *  thrust::copy_if(thrust::host, data, data + N, stencil, result, is_even());
+ *
+ *  // data remains    = { 0, 1,  2, 3, 4, 5};
+ *  // stencil remains = {-2, 0, -1, 0, 1, 2};
+ *  // result is now     { 0, 1,  3, 5}
+ *  \endcode
+ *
+ *  \see \c remove_copy_if
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+/*! This version of \p copy_if copies elements from the range <tt>[first,last)</tt>
+ *  to a range beginning at \p result, except that any element whose corresponding stencil
+ *  element causes \p pred to be \c false is not copied. \p copy_if is stable, meaning
+ *  that the relative order of elements that are copied is unchanged.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>,
+ *  \p copy_if performs the assignment <tt>*result = *(first+n)</tt> and \p result
+ *  is advanced one position if <tt>pred(*(stencil+n))</tt>. Otherwise, no assignment
+ *  occurs and \p result is not advanced.
+ *
+ *  \param first The beginning of the sequence from which to copy.
+ *  \param last The end of the sequence from which to copy.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the sequence into which to copy.
+ *  \param pred The predicate to test on every value of the range <tt>[stencil, stencil + (last-first))</tt>.
+ *  \return <tt>result + n</tt>, where \c n is equal to the number of times \p pred
+ *          evaluated to \c true in the range <tt>[stencil, stencil + (last-first))</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *                         and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/OutputIterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *  \pre The ranges <tt>[stencil, stencil + (last - first))</tt> and <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p copy_if to perform stream compaction
+ *  to copy numbers to an output range when corresponding stencil elements are even:
+ *
+ *  \code
+ *  #include <thrust/copy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int N = 6;
+ *  int data[N]    = { 0, 1,  2, 3, 4, 5};
+ *  int stencil[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[4];
+ *
+ *  thrust::copy_if(data, data + N, stencil, result, is_even());
+ *
+ *  // data remains    = { 0, 1,  2, 3, 4, 5};
+ *  // stencil remains = {-2, 0, -1, 0, 1, 2};
+ *  // result is now     { 0, 1,  3, 5}
+ *  \endcode
+ *
+ *  \see \c remove_copy_if
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator copy_if(InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred);
+
+/*! \} // end stream_compaction
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/copy.h>
+#include <thrust/detail/copy_if.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/count.h b/thirdparty/manifold/thirdparty/thrust/thrust/count.h
new file mode 100644
index 000000000000..abf8b2d6c300
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/count.h
@@ -0,0 +1,231 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file count.h
+ *  \brief Counting elements in a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+/*! \addtogroup reductions
+ *  \ingroup algorithms
+ *  \{
+ */
+
+/*! \addtogroup counting
+ *  \ingroup reductions
+ *  \{
+ */
+
+
+/*! \p count finds the number of elements in <tt>[first,last)</tt> that are equal
+ *  to \p value. More precisely, \p count returns the number of iterators \c i in
+ *  <tt>[first, last)</tt> such that <tt>*i == value</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param value The value to be counted.
+ *  \return The number of elements equal to \p value.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be a model of must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam EqualityComparable must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a> and can be compared for equality with \c InputIterator's \c value_type
+ *
+ *  The following code snippet demonstrates how to use \p count to 
+ *  count the number of instances in a range of a value of interest using the \p thrust::device execution policy:
+ *
+ *  \code
+ *  #include <thrust/count.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  // put 3 1s in a device_vector
+ *  thrust::device_vector<int> vec(5,0);
+ *  vec[1] = 1;
+ *  vec[3] = 1;
+ *  vec[4] = 1;
+ *  
+ *  // count the 1s
+ *  int result = thrust::count(thrust::device, vec.begin(), vec.end(), 1);
+ *  // result == 3
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/count
+ */
+template<typename DerivedPolicy, typename InputIterator, typename EqualityComparable>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, const EqualityComparable& value);
+
+
+
+/*! \p count finds the number of elements in <tt>[first,last)</tt> that are equal
+ *  to \p value. More precisely, \p count returns the number of iterators \c i in
+ *  <tt>[first, last)</tt> such that <tt>*i == value</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param value The value to be counted.
+ *  \return The number of elements equal to \p value.
+ *
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be a model of must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam EqualityComparable must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a> and can be compared for equality with \c InputIterator's \c value_type
+ *
+ *  The following code snippet demonstrates how to use \p count to 
+ *  count the number of instances in a range of a value of interest.
+ *  \code
+ *  #include <thrust/count.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  // put 3 1s in a device_vector
+ *  thrust::device_vector<int> vec(5,0);
+ *  vec[1] = 1;
+ *  vec[3] = 1;
+ *  vec[4] = 1;
+ *  
+ *  // count the 1s
+ *  int result = thrust::count(vec.begin(), vec.end(), 1);
+ *  // result == 3
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/count
+ */
+template <typename InputIterator, typename EqualityComparable>
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count(InputIterator first, InputIterator last, const EqualityComparable& value);
+
+
+/*! \p count_if finds the number of elements in <tt>[first,last)</tt> for which 
+ *  a predicate is \c true. More precisely, \p count_if returns the number of iterators
+ *  \c i in <tt>[first, last)</tt> such that <tt>pred(*i) == true</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param pred The predicate.
+ *  \return The number of elements where \p pred is \c true.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c Predicate's \c argument_type.
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p count to
+ *  count the number of odd numbers in a range using the \p thrust::device execution policy:
+ *
+ *  \code
+ *  #include <thrust/count.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_odd
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int &x)
+ *    {
+ *      return x & 1;
+ *    }
+ *  };
+ *  ...
+ *  // fill a device_vector with even & odd numbers
+ *  thrust::device_vector<int> vec(5);
+ *  vec[0] = 0;
+ *  vec[1] = 1;
+ *  vec[2] = 2;
+ *  vec[3] = 3;
+ *  vec[4] = 4;
+ *
+ *  // count the odd elements in vec
+ *  int result = thrust::count_if(thrust::device, vec.begin(), vec.end(), is_odd());
+ *  // result == 2
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/count
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \p count_if finds the number of elements in <tt>[first,last)</tt> for which 
+ *  a predicate is \c true. More precisely, \p count_if returns the number of iterators
+ *  \c i in <tt>[first, last)</tt> such that <tt>pred(*i) == true</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param pred The predicate.
+ *  \return The number of elements where \p pred is \c true.
+ *
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c Predicate's \c argument_type.
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p count to
+ *  count the number of odd numbers in a range.
+ *  \code
+ *  #include <thrust/count.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  struct is_odd
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int &x)
+ *    {
+ *      return x & 1;
+ *    }
+ *  };
+ *  ...
+ *  // fill a device_vector with even & odd numbers
+ *  thrust::device_vector<int> vec(5);
+ *  vec[0] = 0;
+ *  vec[1] = 1;
+ *  vec[2] = 2;
+ *  vec[3] = 3;
+ *  vec[4] = 4;
+ *
+ *  // count the odd elements in vec
+ *  int result = thrust::count_if(vec.begin(), vec.end(), is_odd());
+ *  // result == 2
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/count
+ */
+template <typename InputIterator, typename Predicate>
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count_if(InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \} // end counting
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/count.h>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/adjacent_difference.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/adjacent_difference.inl
new file mode 100644
index 000000000000..844687cff4c9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/adjacent_difference.inl
@@ -0,0 +1,86 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/adjacent_difference.h>
+#include <thrust/system/detail/adl/adjacent_difference.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result)
+{
+  using thrust::system::detail::generic::adjacent_difference;
+
+  return adjacent_difference(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result);
+} // end adjacent_difference()
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename BinaryFunction>
+__host__ __device__
+OutputIterator adjacent_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::adjacent_difference;
+
+  return adjacent_difference(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, binary_op);
+} // end adjacent_difference()
+
+
+template <typename InputIterator, typename OutputIterator>
+OutputIterator adjacent_difference(InputIterator first, InputIterator last,
+                                   OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::adjacent_difference(select_system(system1, system2), first, last, result);
+} // end adjacent_difference()
+
+
+template <typename InputIterator, typename OutputIterator, typename BinaryFunction>
+OutputIterator adjacent_difference(InputIterator first, InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::adjacent_difference(select_system(system1, system2), first, last, result, binary_op);
+} // end adjacent_difference()
+
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/advance.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/advance.inl
new file mode 100644
index 000000000000..7b5f261bdced
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/advance.inl
@@ -0,0 +1,74 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/advance.h>
+#include <thrust/system/detail/generic/advance.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/has_nested_type.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_difference_type, difference_type)
+
+template <typename InputIterator, typename Distance>
+__host__ __device__
+void advance(InputIterator& i, Distance n)
+{
+  thrust::system::detail::generic::advance(i, n);
+}
+
+template <typename InputIterator>
+__host__ __device__
+InputIterator next(
+  InputIterator i
+, typename iterator_traits<InputIterator>::difference_type n = 1
+)
+{
+  thrust::system::detail::generic::advance(i, n);
+  return i;
+}
+
+template <typename BidirectionalIterator>
+__host__ __device__
+BidirectionalIterator prev(
+  BidirectionalIterator i
+, typename iterator_traits<BidirectionalIterator>::difference_type n = 1
+)
+{
+  thrust::system::detail::generic::advance(i, -n);
+  return i;
+}
+
+template <typename BidirectionalIterator>
+__host__ __device__
+typename detail::disable_if<
+  has_difference_type<iterator_traits<BidirectionalIterator> >::value
+, BidirectionalIterator
+>::type prev(
+  BidirectionalIterator i
+, typename detail::pointer_traits<BidirectionalIterator>::difference_type n = 1
+)
+{
+  thrust::system::detail::generic::advance(i, -n);
+  return i;
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/algorithm_wrapper.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/algorithm_wrapper.h
new file mode 100644
index 000000000000..c09b9a0a0b4d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/algorithm_wrapper.h
@@ -0,0 +1,27 @@
+/*
+ *  Copyright 2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+// When a compiler uses Thrust as part of its implementation of Standard C++
+// algorithms, a cycle of included files may result when Thrust code tries to
+// use a standard algorithm.  Having a macro that is defined only when Thrust
+// is including an algorithms-related header gives the compiler a chance to
+// detect and break the cycle of includes.
+
+#define THRUST_INCLUDING_ALGORITHMS_HEADER
+#include <algorithm>
+#undef  THRUST_INCLUDING_ALGORITHMS_HEADER
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/alignment.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/alignment.h
new file mode 100644
index 000000000000..08f73501e031
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/alignment.h
@@ -0,0 +1,230 @@
+/*
+ *  Copyright 2017 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file alignment.h
+ *  \brief Type-alignment utilities.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h> // For `integral_constant`.
+
+#include <cstddef> // For `std::size_t` and `std::max_align_t`.
+
+#if THRUST_CPP_DIALECT >= 2011
+    #include <type_traits> // For `std::alignment_of` and `std::aligned_storage`.
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+/// \p THRUST_ALIGNOF is a macro that takes a single type-id as a parameter,
+/// and returns the alignment requirement of the type in bytes.
+/// 
+/// It is an approximation of C++11's `alignof` operator.
+///
+/// Note: MSVC does not allow the builtin used to implement this to be placed
+/// inside of a `__declspec(align(#))` attribute. As a workaround, you can
+/// assign the result of \p THRUST_ALIGNOF to a variable and pass the variable
+/// as the argument to `__declspec(align(#))`.
+#if THRUST_CPP_DIALECT >= 2011
+    #define THRUST_ALIGNOF(x) alignof(x) 
+#else
+    #define THRUST_ALIGNOF(x) __alignof(x)
+#endif
+
+/// \p alignment_of provides the member constant `value` which is equal to the
+/// alignment requirement of the type `T`, as if obtained by a C++11 `alignof`
+/// expression.
+/// 
+/// It is an implementation of C++11's \p std::alignment_of.
+#if THRUST_CPP_DIALECT >= 2011
+    template <typename T>
+    using alignment_of = std::alignment_of<T>;
+#else
+    template <typename T>
+    struct alignment_of;
+
+    template <typename T, std::size_t size_diff>
+    struct alignment_of_helper
+    {
+        static const std::size_t value =
+            integral_constant<std::size_t, size_diff>::value;
+    };
+
+    template <typename T>
+    struct alignment_of_helper<T, 0>
+    {
+        static const std::size_t value = alignment_of<T>::value;
+    };
+
+    template <typename T>
+    struct alignment_of
+    {
+      private:
+        struct impl
+        {
+            T    x;
+            char c;
+        };
+
+      public:
+        static const std::size_t value =
+            alignment_of_helper<impl, sizeof(impl) - sizeof(T)>::value;
+    };
+#endif
+
+/// \p aligned_type provides the nested type `type`, which is a trivial
+/// type whose alignment requirement is a divisor of `Align`.
+///
+/// The behavior is undefined if `Align` is not a power of 2.
+template <std::size_t Align>
+struct aligned_type;
+
+#if THRUST_CPP_DIALECT >= 2011                                                     \
+  && (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                        \
+  && (THRUST_GCC_VERSION >= 40800)
+    // C++11 implementation, excluding GCC 4.7, which doesn't have `alignas`.
+    template <std::size_t Align>
+    struct aligned_type
+    {
+        struct alignas(Align) type {};
+    };
+#elif  (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC)                    \
+    || (   (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                 \
+        && (THRUST_GCC_VERSION < 40600))
+    // C++03 implementation for MSVC and GCC <= 4.5.
+    // 
+    // We have to implement `aligned_type` with specializations for MSVC
+    // and GCC 4.2.x and older because they require literals as arguments to 
+    // their alignment attribute.
+
+    #if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC)
+        // MSVC implementation.
+        #define THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(X)                  \
+            template <>                                                       \
+            struct aligned_type<X>                                            \
+            {                                                                 \
+                __declspec(align(X)) struct type {};                          \
+            };                                                                \
+            /**/
+    #else
+        // GCC <= 4.2 implementation.
+        #define THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(X)                  \
+            template <>                                                       \
+            struct aligned_type<X>                                            \
+            {                                                                 \
+                struct type {} __attribute__((aligned(X)));                   \
+            };                                                                \
+            /**/
+    #endif
+    
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(1);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(2);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(4);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(8);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(16);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(32);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(64);
+    THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION(128);
+
+    #undef THRUST_DEFINE_ALIGNED_TYPE_SPECIALIZATION
+#else
+    // C++03 implementation for GCC > 4.5, Clang, PGI, ICPC, and xlC.
+    template <std::size_t Align>
+    struct aligned_type
+    {
+        struct type {} __attribute__((aligned(Align)));
+    };
+#endif
+
+/// \p aligned_storage provides the nested type `type`, which is a trivial type
+/// suitable for use as uninitialized storage for any object whose size is at
+/// most `Len` bytes and whose alignment requirement is a divisor of `Align`.
+/// 
+/// The behavior is undefined if `Len` is 0 or `Align` is not a power of 2.
+///
+/// It is an implementation of C++11's \p std::aligned_storage.
+#if THRUST_CPP_DIALECT >= 2011
+    template <std::size_t Len, std::size_t Align>
+    using aligned_storage = std::aligned_storage<Len, Align>;
+#else
+    template <std::size_t Len, std::size_t Align>
+    struct aligned_storage
+    {
+        union type
+        {
+            unsigned char data[Len];
+            // We put this into the union in case the alignment requirement of
+            // an array of `unsigned char` of length `Len` is greater than
+            // `Align`.
+
+            typename aligned_type<Align>::type align;
+        };
+    };
+#endif
+
+/// \p max_align_t is a trivial type whose alignment requirement is at least as
+/// strict (as large) as that of every scalar type.
+///
+/// It is an implementation of C++11's \p std::max_align_t.
+#if THRUST_CPP_DIALECT >= 2011                                                     \
+  && (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                        \
+  && (THRUST_GCC_VERSION >= 40900)
+    // GCC 4.7 and 4.8 don't have `std::max_align_t`.
+    using max_align_t = std::max_align_t;
+#else
+    union max_align_t
+    {
+        // These cannot be private because C++03 POD types cannot have private
+        // data members.
+        char c;
+        short s;
+        int i;
+        long l;
+        float f;
+        double d;
+        long long ll;
+        long double ld;
+        void* p;
+    };
+#endif
+
+/// \p aligned_reinterpret_cast `reinterpret_cast`s \p u of type \p U to `void*`
+/// and then `reinterpret_cast`s the result to \p T. The indirection through
+/// `void*` suppresses compiler warnings when the alignment requirement of \p *u
+/// is less than the alignment requirement of \p *t. The caller of
+/// \p aligned_reinterpret_cast is responsible for ensuring that the alignment
+/// requirements are actually satisified.
+template <typename T, typename U>
+__host__ __device__
+T aligned_reinterpret_cast(U u)
+{
+  return reinterpret_cast<T>(reinterpret_cast<void*>(u));
+}
+
+__host__ __device__
+inline std::size_t aligned_storage_size(std::size_t n, std::size_t align)
+{
+  return ((n + align - 1) / align) * align;
+}
+
+} // end namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.h
new file mode 100644
index 000000000000..3a5af3661f39
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.h
@@ -0,0 +1,442 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// allocator_traits::rebind_alloc and allocator::rebind_traits are from libc++,
+// dual licensed under the MIT and the University of Illinois Open Source
+// Licenses.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/detail/type_traits/has_nested_type.h>
+#include <thrust/detail/type_traits/has_member_function.h>
+#include <thrust/detail/type_traits.h>
+
+#include <thrust/detail/memory_wrapper.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+// forward declaration for has_member_system
+template<typename Alloc> struct allocator_system;
+
+
+namespace allocator_traits_detail
+{
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_value_type, value_type)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_pointer, pointer)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_const_pointer, const_pointer)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_reference, reference)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_const_reference, const_reference)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_void_pointer, void_pointer)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_const_void_pointer, const_void_pointer)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_difference_type, difference_type)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_size_type, size_type)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_propagate_on_container_copy_assignment, propagate_on_container_copy_assignment)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_propagate_on_container_move_assignment, propagate_on_container_move_assignment)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_propagate_on_container_swap, propagate_on_container_swap)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_system_type, system_type)
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_is_always_equal, is_always_equal)
+__THRUST_DEFINE_HAS_MEMBER_FUNCTION(has_member_system_impl, system)
+
+template<typename Alloc, typename U>
+  struct has_rebind
+{
+  typedef char yes_type;
+  typedef int  no_type;
+
+  template<typename S>
+  static yes_type test(typename S::template rebind<U>::other*);
+  template<typename S>
+  static no_type  test(...);
+
+  static bool const value = sizeof(test<U>(0)) == sizeof(yes_type);
+
+  typedef thrust::detail::integral_constant<bool, value> type;
+};
+
+// The following fields of std::allocator have been deprecated (since C++17).
+// There's no way to detect it other than explicit specialization.
+#if THRUST_CPP_DIALECT >= 2017
+#define THRUST_SPECIALIZE_DEPRECATED(trait_name)                               \
+template <typename T>                                                          \
+struct trait_name<std::allocator<T>> : false_type {};
+
+THRUST_SPECIALIZE_DEPRECATED(has_is_always_equal)
+THRUST_SPECIALIZE_DEPRECATED(has_pointer)
+THRUST_SPECIALIZE_DEPRECATED(has_const_pointer)
+THRUST_SPECIALIZE_DEPRECATED(has_reference)
+THRUST_SPECIALIZE_DEPRECATED(has_const_reference)
+
+#undef THRUST_SPECIALIZE_DEPRECATED
+
+template<typename T, typename U>
+struct has_rebind<std::allocator<T>, U> : false_type {};
+#endif
+
+template<typename T>
+  struct nested_pointer
+{
+  typedef typename T::pointer type;
+};
+
+template<typename T>
+  struct nested_const_pointer
+{
+  typedef typename T::const_pointer type;
+};
+
+template<typename T>
+  struct nested_reference
+{
+  typedef typename T::reference type;
+};
+
+template<typename T>
+  struct nested_const_reference
+{
+  typedef typename T::const_reference type;
+};
+
+template<typename T>
+  struct nested_void_pointer
+{
+  typedef typename T::void_pointer type;
+};
+
+template<typename T>
+  struct nested_const_void_pointer
+{
+  typedef typename T::const_void_pointer type;
+};
+
+template<typename T>
+  struct nested_difference_type
+{
+  typedef typename T::difference_type type;
+};
+
+template<typename T>
+  struct nested_size_type
+{
+  typedef typename T::size_type type;
+};
+
+template<typename T>
+  struct nested_propagate_on_container_copy_assignment
+{
+  typedef typename T::propagate_on_container_copy_assignment type;
+};
+
+template<typename T>
+  struct nested_propagate_on_container_move_assignment
+{
+  typedef typename T::propagate_on_container_move_assignment type;
+};
+
+template<typename T>
+  struct nested_propagate_on_container_swap
+{
+  typedef typename T::propagate_on_container_swap type;
+};
+
+template<typename T>
+  struct nested_is_always_equal
+{
+  typedef typename T::is_always_equal type;
+};
+
+template<typename T>
+  struct nested_system_type
+{
+  typedef typename T::system_type type;
+};
+
+template<typename Alloc>
+  struct has_member_system
+{
+  typedef typename allocator_system<Alloc>::type system_type;
+
+  typedef typename has_member_system_impl<Alloc, system_type&(void)>::type type;
+  static const bool value = type::value;
+};
+
+template<class Alloc, class U, bool = has_rebind<Alloc, U>::value>
+  struct rebind_alloc
+{
+    typedef typename Alloc::template rebind<U>::other type;
+};
+
+#if THRUST_CPP_DIALECT >= 2011
+template<template<typename, typename...> class Alloc,
+         typename T, typename... Args, typename U>
+  struct rebind_alloc<Alloc<T, Args...>, U, true>
+{
+    typedef typename Alloc<T, Args...>::template rebind<U>::other type;
+};
+
+template<template<typename, typename...> class Alloc,
+         typename T, typename... Args, typename U>
+  struct rebind_alloc<Alloc<T, Args...>, U, false>
+{
+    typedef Alloc<U, Args...> type;
+};
+#else // C++03
+template <template <typename> class Alloc, typename T, typename U>
+  struct rebind_alloc<Alloc<T>, U, true>
+{
+    typedef typename Alloc<T>::template rebind<U>::other type;
+};
+
+template <template <typename> class Alloc, typename T, typename U>
+  struct rebind_alloc<Alloc<T>, U, false>
+{
+    typedef Alloc<U> type;
+};
+
+template<template<typename, typename> class Alloc,
+         typename T, typename A0, typename U>
+  struct rebind_alloc<Alloc<T, A0>, U, true>
+{
+    typedef typename Alloc<T, A0>::template rebind<U>::other type;
+};
+
+template<template<typename, typename> class Alloc,
+         typename T, typename A0, typename U>
+  struct rebind_alloc<Alloc<T, A0>, U, false>
+{
+    typedef Alloc<U, A0> type;
+};
+
+template<template<typename, typename, typename> class Alloc,
+         typename T, typename A0, typename A1, typename U>
+  struct rebind_alloc<Alloc<T, A0, A1>, U, true>
+{
+    typedef typename Alloc<T, A0, A1>::template rebind<U>::other type;
+};
+
+template<template<typename, typename, typename> class Alloc,
+         typename T, typename A0, typename A1, typename U>
+  struct rebind_alloc<Alloc<T, A0, A1>, U, false>
+{
+    typedef Alloc<U, A0, A1> type;
+};
+
+template<template<typename, typename, typename, typename> class Alloc,
+         typename T, typename A0, typename A1, typename A2, typename U>
+  struct rebind_alloc<Alloc<T, A0, A1, A2>, U, true>
+{
+    typedef typename Alloc<T, A0, A1, A2>::template rebind<U>::other type;
+};
+
+template<template<typename, typename, typename, typename> class Alloc,
+         typename T, typename A0, typename A1, typename A2, typename U>
+  struct rebind_alloc<Alloc<T, A0, A1, A2>, U, false>
+{
+    typedef Alloc<U, A0, A1, A2> type;
+};
+#endif
+
+} // end allocator_traits_detail
+
+
+template<typename Alloc>
+  struct allocator_traits
+{
+  typedef Alloc allocator_type;
+
+  typedef typename allocator_type::value_type value_type;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_pointer<allocator_type>::value,
+    allocator_traits_detail::nested_pointer<allocator_type>,
+    identity_<value_type*>
+  >::type pointer;
+
+  private:
+    template<typename T>
+      struct rebind_pointer
+    {
+      typedef typename pointer_traits<pointer>::template rebind<T>::other type;
+    };
+
+  public:
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_const_pointer<allocator_type>::value,
+    allocator_traits_detail::nested_const_pointer<allocator_type>,
+    rebind_pointer<const value_type>
+  >::type const_pointer;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_void_pointer<allocator_type>::value,
+    allocator_traits_detail::nested_void_pointer<allocator_type>,
+    rebind_pointer<void>
+  >::type void_pointer;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_const_void_pointer<allocator_type>::value,
+    allocator_traits_detail::nested_const_void_pointer<allocator_type>,
+    rebind_pointer<const void>
+  >::type const_void_pointer;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_difference_type<allocator_type>::value,
+    allocator_traits_detail::nested_difference_type<allocator_type>,
+    pointer_difference<pointer>
+  >::type difference_type;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_size_type<allocator_type>::value,
+    allocator_traits_detail::nested_size_type<allocator_type>,
+    make_unsigned<difference_type>
+  >::type size_type;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_propagate_on_container_copy_assignment<allocator_type>::value,
+    allocator_traits_detail::nested_propagate_on_container_copy_assignment<allocator_type>,
+    identity_<false_type>
+  >::type propagate_on_container_copy_assignment;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_propagate_on_container_move_assignment<allocator_type>::value,
+    allocator_traits_detail::nested_propagate_on_container_move_assignment<allocator_type>,
+    identity_<false_type>
+  >::type propagate_on_container_move_assignment;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_propagate_on_container_swap<allocator_type>::value,
+    allocator_traits_detail::nested_propagate_on_container_swap<allocator_type>,
+    identity_<false_type>
+  >::type propagate_on_container_swap;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_is_always_equal<allocator_type>::value,
+    allocator_traits_detail::nested_is_always_equal<allocator_type>,
+    is_empty<allocator_type>
+  >::type is_always_equal;
+
+  typedef typename eval_if<
+    allocator_traits_detail::has_system_type<allocator_type>::value,
+    allocator_traits_detail::nested_system_type<allocator_type>,
+    thrust::iterator_system<pointer>
+  >::type system_type;
+
+  // XXX rebind and rebind_traits are alias templates
+  //     and so are omitted while c++11 is unavailable
+
+#if THRUST_CPP_DIALECT >= 2011
+  template <typename U>
+  using rebind_alloc =
+    typename allocator_traits_detail::rebind_alloc<allocator_type, U>::type;
+
+  template <typename U>
+  using rebind_traits = allocator_traits<rebind_alloc<U>>;
+
+  // We define this nested type alias for compatibility with the C++03-style
+  // rebind_* mechanisms.
+  using other = allocator_traits;
+#else
+  template <typename U>
+  struct rebind_alloc
+  {
+    typedef typename
+      allocator_traits_detail::rebind_alloc<allocator_type, U>::type other;
+  };
+  template <typename U>
+  struct rebind_traits
+  {
+    typedef allocator_traits<typename rebind_alloc<U>::other> other;
+  };
+#endif
+
+  // Deprecated std::allocator typedefs that we need:
+  typedef typename thrust::detail::pointer_traits<pointer>::reference reference;
+  typedef typename thrust::detail::pointer_traits<const_pointer>::reference const_reference;
+
+  inline __host__ __device__
+  static pointer allocate(allocator_type &a, size_type n);
+
+  inline __host__ __device__
+  static pointer allocate(allocator_type &a, size_type n, const_void_pointer hint);
+
+  inline __host__ __device__
+  static void deallocate(allocator_type &a, pointer p, size_type n);
+
+  // XXX should probably change T* to pointer below and then relax later
+
+  template<typename T>
+  inline __host__ __device__ static void construct(allocator_type &a, T *p);
+  
+  template<typename T, typename Arg1>
+  inline __host__ __device__ static void construct(allocator_type &a, T *p, const Arg1 &arg1);
+
+#if THRUST_CPP_DIALECT >= 2011
+  template<typename T, typename... Args>
+  inline __host__ __device__ static void construct(allocator_type &a, T *p, Args&&... args);
+#endif
+
+  template<typename T>
+  inline __host__ __device__ static void destroy(allocator_type &a, T *p);
+
+  inline __host__ __device__
+  static size_type max_size(const allocator_type &a);
+}; // end allocator_traits
+
+
+// we consider a type an allocator if T::value_type exists
+// it doesn't make much sense (containers, which are not allocators, will fulfill this requirement),
+// but allocator_traits is specified to work for any type with that nested typedef
+template<typename T>
+  struct is_allocator
+    : allocator_traits_detail::has_value_type<T>
+{};
+
+
+// XXX consider moving this non-standard functionality inside allocator_traits
+template<typename Alloc>
+  struct allocator_system
+{
+  // the type of the allocator's system
+  typedef typename eval_if<
+    allocator_traits_detail::has_system_type<Alloc>::value,
+    allocator_traits_detail::nested_system_type<Alloc>,
+    thrust::iterator_system<
+      typename allocator_traits<Alloc>::pointer
+    >
+  >::type type;
+
+  // the type that get returns
+  typedef typename eval_if<
+    allocator_traits_detail::has_member_system<Alloc>::value, // if Alloc.system() exists
+    add_reference<type>,                                      // then get() needs to return a reference
+    identity_<type>                                           // else get() needs to return a value
+  >::type get_result_type;
+
+  __host__ __device__
+  inline static get_result_type get(Alloc &a);
+};
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/allocator_traits.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.inl
new file mode 100644
index 000000000000..275330094949
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/allocator_traits.inl
@@ -0,0 +1,465 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/type_traits/is_call_possible.h>
+#include <thrust/detail/integer_traits.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+  #include <thrust/detail/type_deduction.h>
+#endif
+
+#include <thrust/detail/memory_wrapper.h>
+#include <new>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+#if THRUST_CPP_DIALECT >= 2011
+
+// std::allocator's member functions are deprecated in C++17 and removed in
+// C++20, so we can't just use the generic implementation for allocator_traits
+// that calls the allocator's member functions.
+// Instead, specialize allocator_traits for std::allocator and defer to
+// std::allocator_traits<std::allocator> and let the STL do whatever it needs
+// to for the current c++ version. Manually forward the calls to suppress
+// host/device warnings.
+template <typename T>
+struct allocator_traits<std::allocator<T>>
+  : public std::allocator_traits<std::allocator<T>>
+{
+private:
+  using superclass = std::allocator_traits<std::allocator<T>>;
+
+public:
+  using allocator_type = typename superclass::allocator_type;
+  using value_type = typename superclass::value_type;
+  using pointer = typename superclass::pointer;
+  using const_pointer = typename superclass::const_pointer;
+  using void_pointer = typename superclass::void_pointer;
+  using const_void_pointer = typename superclass::const_void_pointer;
+  using difference_type = typename superclass::difference_type;
+  using size_type = typename superclass::size_type;
+  using propagate_on_container_swap = typename superclass::propagate_on_container_swap;
+  using propagate_on_container_copy_assignment =
+    typename superclass::propagate_on_container_copy_assignment;
+  using propagate_on_container_move_assignment =
+    typename superclass::propagate_on_container_move_assignment;
+
+  // std::allocator_traits added this in C++17, but thrust::allocator_traits defines
+  // it unconditionally.
+  using is_always_equal = typename eval_if<
+      allocator_traits_detail::has_is_always_equal<allocator_type>::value,
+      allocator_traits_detail::nested_is_always_equal<allocator_type>,
+      is_empty<allocator_type>
+    >::type;
+
+  // std::allocator_traits doesn't provide these, but
+  // thrust::detail::allocator_traits does. These used to be part of the
+  // std::allocator API but were deprecated in C++17.
+  using reference = typename thrust::detail::pointer_traits<pointer>::reference;
+  using const_reference = typename thrust::detail::pointer_traits<const_pointer>::reference;
+
+  template <typename U>
+  using rebind_alloc = std::allocator<U>;
+  template <typename U>
+  using rebind_traits = allocator_traits<std::allocator<U>>;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  static pointer allocate(allocator_type &a, size_type n)
+  {
+    return superclass::allocate(a, n);
+  }
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  static pointer allocate(allocator_type &a, size_type n, const_void_pointer hint)
+  {
+    return superclass::allocate(a, n, hint);
+  }
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  static void deallocate(allocator_type &a, pointer p, size_type n)
+  {
+    superclass::deallocate(a, p, n);
+  }
+
+  __thrust_exec_check_disable__
+  template <typename U, typename ...Args>
+  __host__ __device__
+  static void construct(allocator_type &a, U *p, Args&&... args)
+  {
+    superclass::construct(a, p, THRUST_FWD(args)...);
+  }
+
+  __thrust_exec_check_disable__
+  template <typename U>
+  __host__ __device__
+  static void destroy(allocator_type &a, U *p)
+  {
+    superclass::destroy(a, p);
+  }
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  static size_type max_size(const allocator_type &a)
+  {
+    return superclass::max_size(a);
+  }
+};
+
+#endif //  C++11
+
+namespace allocator_traits_detail
+{
+
+__THRUST_DEFINE_IS_CALL_POSSIBLE(has_member_allocate_with_hint_impl, allocate)
+
+template<typename Alloc>
+  class has_member_allocate_with_hint
+{
+  typedef typename allocator_traits<Alloc>::pointer            pointer;
+  typedef typename allocator_traits<Alloc>::size_type          size_type;
+  typedef typename allocator_traits<Alloc>::const_void_pointer const_void_pointer;
+
+  public:
+    typedef typename has_member_allocate_with_hint_impl<Alloc, pointer(size_type,const_void_pointer)>::type type;
+    static const bool value = type::value;
+};
+
+template<typename Alloc>
+__host__ __device__
+  typename enable_if<
+    has_member_allocate_with_hint<Alloc>::value,
+    typename allocator_traits<Alloc>::pointer
+  >::type
+    allocate(Alloc &a, typename allocator_traits<Alloc>::size_type n, typename allocator_traits<Alloc>::const_void_pointer hint)
+{
+  return a.allocate(n,hint);
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename disable_if<
+    has_member_allocate_with_hint<Alloc>::value,
+    typename allocator_traits<Alloc>::pointer
+  >::type
+    allocate(Alloc &a, typename allocator_traits<Alloc>::size_type n, typename allocator_traits<Alloc>::const_void_pointer)
+{
+  return a.allocate(n);
+}
+
+
+__THRUST_DEFINE_IS_CALL_POSSIBLE(has_member_construct1_impl, construct)
+
+template<typename Alloc, typename T>
+  struct has_member_construct1
+    : has_member_construct1_impl<Alloc, void(T*)>
+{};
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T>
+  inline __host__ __device__
+    typename enable_if<
+      has_member_construct1<Alloc,T>::value
+    >::type
+      construct(Alloc &a, T *p)
+{
+  a.construct(p);
+}
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T>
+  inline __host__ __device__
+    typename disable_if<
+      has_member_construct1<Alloc,T>::value
+    >::type
+      construct(Alloc &, T *p)
+{
+  ::new(static_cast<void*>(p)) T();
+}
+
+
+__THRUST_DEFINE_IS_CALL_POSSIBLE(has_member_construct2_impl, construct)
+
+template<typename Alloc, typename T, typename Arg1>
+  struct has_member_construct2
+    : has_member_construct2_impl<Alloc, void(T*,const Arg1 &)>
+{};
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T, typename Arg1>
+  inline __host__ __device__
+    typename enable_if<
+      has_member_construct2<Alloc,T,Arg1>::value
+    >::type
+      construct(Alloc &a, T *p, const Arg1 &arg1)
+{
+  a.construct(p,arg1);
+}
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T, typename Arg1>
+  inline __host__ __device__
+    typename disable_if<
+      has_member_construct2<Alloc,T,Arg1>::value
+    >::type
+      construct(Alloc &, T *p, const Arg1 &arg1)
+{
+  ::new(static_cast<void*>(p)) T(arg1);
+}
+
+#if THRUST_CPP_DIALECT >= 2011
+
+__THRUST_DEFINE_IS_CALL_POSSIBLE(has_member_constructN_impl, construct)
+
+template<typename Alloc, typename T, typename... Args>
+  struct has_member_constructN
+    : has_member_constructN_impl<Alloc, void(T*, Args...)>
+{};
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T, typename... Args>
+  inline __host__ __device__
+    typename enable_if<
+      has_member_constructN<Alloc, T, Args...>::value
+    >::type
+      construct(Alloc &a, T* p, Args&&... args)
+{
+  a.construct(p, THRUST_FWD(args)...);
+}
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T, typename... Args>
+  inline __host__ __device__
+    typename disable_if<
+      has_member_constructN<Alloc, T, Args...>::value
+    >::type
+      construct(Alloc &, T* p, Args&&... args)
+{
+  ::new(static_cast<void*>(p)) T(THRUST_FWD(args)...);
+}
+
+#endif
+
+__THRUST_DEFINE_IS_CALL_POSSIBLE(has_member_destroy_impl, destroy)
+
+template<typename Alloc, typename T>
+  struct has_member_destroy
+    : has_member_destroy_impl<Alloc, void(T*)>
+{};
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T>
+  inline __host__ __device__
+    typename enable_if<
+      has_member_destroy<Alloc,T>::value
+    >::type
+      destroy(Alloc &a, T *p)
+{
+  a.destroy(p);
+}
+
+__thrust_exec_check_disable__
+template<typename Alloc, typename T>
+  inline __host__ __device__
+    typename disable_if<
+      has_member_destroy<Alloc,T>::value
+    >::type
+      destroy(Alloc &, T *p)
+{
+  p->~T();
+}
+
+
+__THRUST_DEFINE_IS_CALL_POSSIBLE(has_member_max_size_impl, max_size)
+
+template<typename Alloc>
+  class has_member_max_size
+{
+  typedef typename allocator_traits<Alloc>::size_type size_type;
+
+  public:
+    typedef typename has_member_max_size_impl<Alloc, size_type(void)>::type type;
+    static const bool value = type::value;
+};
+
+template<typename Alloc>
+__host__ __device__
+  typename enable_if<
+    has_member_max_size<Alloc>::value,
+    typename allocator_traits<Alloc>::size_type
+  >::type
+    max_size(const Alloc &a)
+{
+  return a.max_size();
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename disable_if<
+    has_member_max_size<Alloc>::value,
+    typename allocator_traits<Alloc>::size_type
+  >::type
+    max_size(const Alloc &)
+{
+  typedef typename allocator_traits<Alloc>::size_type size_type;
+  return thrust::detail::integer_traits<size_type>::const_max;
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename enable_if<
+    has_member_system<Alloc>::value,
+    typename allocator_system<Alloc>::type &
+  >::type
+    system(Alloc &a)
+{
+  // return the allocator's system
+  return a.system();
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename disable_if<
+    has_member_system<Alloc>::value,
+    typename allocator_system<Alloc>::type
+  >::type
+    system(Alloc &)
+{
+  // return a copy of a value-initialized system
+  return typename allocator_system<Alloc>::type();
+}
+
+
+} // end allocator_traits_detail
+
+
+template<typename Alloc>
+__host__ __device__
+  typename allocator_traits<Alloc>::pointer
+    allocator_traits<Alloc>
+      ::allocate(Alloc &a, typename allocator_traits<Alloc>::size_type n)
+{
+  struct workaround_warnings
+  {
+    __thrust_exec_check_disable__
+    static __host__ __device__
+    typename allocator_traits<Alloc>::pointer
+      allocate(Alloc &a, typename allocator_traits<Alloc>::size_type n)
+    {
+      return a.allocate(n);
+    }
+  };
+
+  return workaround_warnings::allocate(a, n);
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename allocator_traits<Alloc>::pointer
+    allocator_traits<Alloc>
+      ::allocate(Alloc &a, typename allocator_traits<Alloc>::size_type n, typename allocator_traits<Alloc>::const_void_pointer hint)
+{
+  return allocator_traits_detail::allocate(a, n, hint);
+}
+
+template<typename Alloc>
+__host__ __device__
+  void allocator_traits<Alloc>
+    ::deallocate(Alloc &a, typename allocator_traits<Alloc>::pointer p, typename allocator_traits<Alloc>::size_type n)
+{
+  struct workaround_warnings
+  {
+    __thrust_exec_check_disable__
+    static __host__ __device__
+    void deallocate(Alloc &a, typename allocator_traits<Alloc>::pointer p, typename allocator_traits<Alloc>::size_type n)
+    {
+      return a.deallocate(p,n);
+    }
+  };
+
+  return workaround_warnings::deallocate(a,p,n);
+}
+
+template<typename Alloc>
+  template<typename T>
+  __host__ __device__
+    void allocator_traits<Alloc>
+      ::construct(allocator_type &a, T *p)
+{
+  return allocator_traits_detail::construct(a,p);
+}
+
+template<typename Alloc>
+  template<typename T, typename Arg1>
+  __host__ __device__
+    void allocator_traits<Alloc>
+      ::construct(allocator_type &a, T *p, const Arg1 &arg1)
+{
+  return allocator_traits_detail::construct(a,p,arg1);
+}
+
+#if THRUST_CPP_DIALECT >= 2011
+
+template<typename Alloc>
+  template<typename T, typename... Args>
+  __host__ __device__
+    void allocator_traits<Alloc>
+      ::construct(allocator_type &a, T *p, Args&&... args)
+{
+  return allocator_traits_detail::construct(a, p, THRUST_FWD(args)...);
+}
+
+#endif
+
+template<typename Alloc>
+  template<typename T>
+  __host__ __device__
+    void allocator_traits<Alloc>
+      ::destroy(allocator_type &a, T *p)
+{
+  return allocator_traits_detail::destroy(a,p);
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename allocator_traits<Alloc>::size_type
+    allocator_traits<Alloc>
+      ::max_size(const allocator_type &a)
+{
+  return allocator_traits_detail::max_size(a);
+}
+
+template<typename Alloc>
+__host__ __device__
+  typename allocator_system<Alloc>::get_result_type
+    allocator_system<Alloc>
+      ::get(Alloc &a)
+{
+  return allocator_traits_detail::system(a);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.h
new file mode 100644
index 000000000000..b3c2de324627
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.h
@@ -0,0 +1,46 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template<typename System, typename Allocator, typename InputIterator, typename Pointer>
+__host__ __device__
+  Pointer copy_construct_range(thrust::execution_policy<System> &from_system,
+                               Allocator &a,
+                               InputIterator first,
+                               InputIterator last,
+                               Pointer result);
+
+template<typename System, typename Allocator, typename InputIterator, typename Size, typename Pointer>
+__host__ __device__
+  Pointer copy_construct_range_n(thrust::execution_policy<System> &from_system,
+                                 Allocator &a,
+                                 InputIterator first,
+                                 Size n,
+                                 Pointer result);
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/copy_construct_range.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.inl
new file mode 100644
index 000000000000..a71cca1f7207
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/copy_construct_range.inl
@@ -0,0 +1,310 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/copy.h>
+#include <thrust/tuple.h>
+#include <thrust/advance.h>
+#include <thrust/distance.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/for_each.h>
+#include <thrust/detail/memory_wrapper.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace allocator_traits_detail
+{
+
+
+template<typename Allocator, typename InputType, typename OutputType>
+  struct copy_construct_with_allocator
+{
+  Allocator &a;
+
+  __host__ __device__
+  copy_construct_with_allocator(Allocator &a)
+    : a(a)
+  {}
+
+  template<typename Tuple>
+  inline __host__ __device__
+  void operator()(Tuple t)
+  {
+    const InputType &in = thrust::get<0>(t);
+    OutputType &out = thrust::get<1>(t);
+
+    allocator_traits<Allocator>::construct(a, &out, in);
+  }
+};
+
+
+// we need to use allocator_traits<Allocator>::construct() to
+// copy construct a T if either:
+// 1. Allocator has a 2-argument construct() member or
+// 2. T has a non-trivial copy constructor
+template<typename Allocator, typename T>
+  struct needs_copy_construct_via_allocator
+    : integral_constant<
+        bool,
+        (has_member_construct2<Allocator,T,T>::value || !has_trivial_copy_constructor<T>::value)
+      >
+{};
+
+
+// we know that std::allocator::construct's only effect is to call T's
+// copy constructor, so we needn't consider or use its construct() member for copy construction
+template<typename U, typename T>
+  struct needs_copy_construct_via_allocator<std::allocator<U>, T>
+    : integral_constant<
+        bool,
+        !has_trivial_copy_constructor<T>::value
+      >
+{};
+
+
+// XXX it's regrettable that this implementation is copied almost
+//     exactly from system::detail::generic::uninitialized_copy
+//     perhaps generic::uninitialized_copy could call this routine
+//     with a default allocator
+template<typename Allocator, typename FromSystem, typename ToSystem, typename InputIterator, typename Pointer>
+__host__ __device__
+  typename enable_if_convertible<
+    FromSystem,
+    ToSystem,
+    Pointer
+  >::type
+    uninitialized_copy_with_allocator(Allocator &a,
+                                      const thrust::execution_policy<FromSystem> &,
+                                      const thrust::execution_policy<ToSystem> &to_system,
+                                      InputIterator first,
+                                      InputIterator last,
+                                      Pointer result)
+{
+  // zip up the iterators
+  typedef thrust::tuple<InputIterator,Pointer> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple>  ZipIterator;
+
+  ZipIterator begin = thrust::make_zip_iterator(thrust::make_tuple(first,result));
+  ZipIterator end = begin;
+
+  // get a zip_iterator pointing to the end
+  const typename thrust::iterator_difference<InputIterator>::type n = thrust::distance(first,last);
+  thrust::advance(end,n);
+
+  // create a functor
+  typedef typename iterator_traits<InputIterator>::value_type InputType;
+  typedef typename iterator_traits<Pointer>::value_type       OutputType;
+
+  // do the for_each
+  // note we use to_system to dispatch the for_each
+  thrust::for_each(to_system, begin, end, copy_construct_with_allocator<Allocator,InputType,OutputType>(a));
+
+  // return the end of the output range
+  return thrust::get<1>(end.get_iterator_tuple());
+}
+
+
+// XXX it's regrettable that this implementation is copied almost
+//     exactly from system::detail::generic::uninitialized_copy_n
+//     perhaps generic::uninitialized_copy_n could call this routine
+//     with a default allocator
+template<typename Allocator, typename FromSystem, typename ToSystem, typename InputIterator, typename Size, typename Pointer>
+__host__ __device__
+  typename enable_if_convertible<
+    FromSystem,
+    ToSystem,
+    Pointer
+  >::type
+    uninitialized_copy_with_allocator_n(Allocator &a,
+                                        const thrust::execution_policy<FromSystem> &,
+                                        const thrust::execution_policy<ToSystem> &to_system,
+                                        InputIterator first,
+                                        Size n,
+                                        Pointer result)
+{
+  // zip up the iterators
+  typedef thrust::tuple<InputIterator,Pointer> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple>  ZipIterator;
+
+  ZipIterator begin = thrust::make_zip_iterator(thrust::make_tuple(first,result));
+
+  // create a functor
+  typedef typename iterator_traits<InputIterator>::value_type InputType;
+  typedef typename iterator_traits<Pointer>::value_type       OutputType;
+
+  // do the for_each_n
+  // note we use to_system to dispatch the for_each_n
+  ZipIterator end = thrust::for_each_n(to_system, begin, n, copy_construct_with_allocator<Allocator,InputType,OutputType>(a));
+
+  // return the end of the output range
+  return thrust::get<1>(end.get_iterator_tuple());
+}
+
+
+template<typename Allocator, typename FromSystem, typename ToSystem, typename InputIterator, typename Pointer>
+__host__ __device__
+  typename disable_if_convertible<
+    FromSystem,
+    ToSystem,
+    Pointer
+  >::type
+    uninitialized_copy_with_allocator(Allocator &,
+                                      const thrust::execution_policy<FromSystem> &from_system,
+                                      const thrust::execution_policy<ToSystem> &to_system,
+                                      InputIterator first,
+                                      InputIterator last,
+                                      Pointer result)
+{
+  // the systems aren't trivially interoperable
+  // just call two_system_copy and hope for the best
+  return thrust::detail::two_system_copy(from_system, to_system, first, last, result);
+} // end uninitialized_copy_with_allocator()
+
+
+template<typename Allocator, typename FromSystem, typename ToSystem, typename InputIterator, typename Size, typename Pointer>
+__host__ __device__
+  typename disable_if_convertible<
+    FromSystem,
+    ToSystem,
+    Pointer
+  >::type
+    uninitialized_copy_with_allocator_n(Allocator &,
+                                        const thrust::execution_policy<FromSystem> &from_system,
+                                        const thrust::execution_policy<ToSystem> &to_system,
+                                        InputIterator first,
+                                        Size n,
+                                        Pointer result)
+{
+  // the systems aren't trivially interoperable
+  // just call two_system_copy_n and hope for the best
+  return thrust::detail::two_system_copy_n(from_system, to_system, first, n, result);
+} // end uninitialized_copy_with_allocator_n()
+
+
+template<typename FromSystem, typename Allocator, typename InputIterator, typename Pointer>
+__host__ __device__
+  typename disable_if<
+    needs_copy_construct_via_allocator<
+      Allocator,
+      typename pointer_element<Pointer>::type
+    >::value,
+    Pointer
+  >::type
+    copy_construct_range(thrust::execution_policy<FromSystem> &from_system,
+                         Allocator &a,
+                         InputIterator first,
+                         InputIterator last,
+                         Pointer result)
+{
+  // just call two_system_copy
+  return thrust::detail::two_system_copy(from_system, allocator_system<Allocator>::get(a), first, last, result);
+}
+
+
+template<typename FromSystem, typename Allocator, typename InputIterator, typename Size, typename Pointer>
+__host__ __device__
+  typename disable_if<
+    needs_copy_construct_via_allocator<
+      Allocator,
+      typename pointer_element<Pointer>::type
+    >::value,
+    Pointer
+  >::type
+    copy_construct_range_n(thrust::execution_policy<FromSystem> &from_system,
+                           Allocator &a,
+                           InputIterator first,
+                           Size n,
+                           Pointer result)
+{
+  // just call two_system_copy_n
+  return thrust::detail::two_system_copy_n(from_system, allocator_system<Allocator>::get(a), first, n, result);
+}
+
+
+template<typename FromSystem, typename Allocator, typename InputIterator, typename Pointer>
+__host__ __device__
+  typename enable_if<
+    needs_copy_construct_via_allocator<
+      Allocator,
+      typename pointer_element<Pointer>::type
+    >::value,
+    Pointer
+  >::type
+    copy_construct_range(thrust::execution_policy<FromSystem> &from_system,
+                         Allocator &a,
+                         InputIterator first,
+                         InputIterator last,
+                         Pointer result)
+{
+  return uninitialized_copy_with_allocator(a, from_system, allocator_system<Allocator>::get(a), first, last, result);
+}
+
+
+template<typename FromSystem, typename Allocator, typename InputIterator, typename Size, typename Pointer>
+__host__ __device__
+  typename enable_if<
+    needs_copy_construct_via_allocator<
+      Allocator,
+      typename pointer_element<Pointer>::type
+    >::value,
+    Pointer
+  >::type
+    copy_construct_range_n(thrust::execution_policy<FromSystem> &from_system,
+                           Allocator &a,
+                           InputIterator first,
+                           Size n,
+                           Pointer result)
+{
+  return uninitialized_copy_with_allocator_n(a, from_system, allocator_system<Allocator>::get(a), first, n, result);
+}
+
+
+} // end allocator_traits_detail
+
+
+template<typename System, typename Allocator, typename InputIterator, typename Pointer>
+__host__ __device__
+  Pointer copy_construct_range(thrust::execution_policy<System> &from_system,
+                               Allocator &a,
+                               InputIterator first,
+                               InputIterator last,
+                               Pointer result)
+{
+  return allocator_traits_detail::copy_construct_range(from_system, a, first, last, result);
+}
+
+
+template<typename System, typename Allocator, typename InputIterator, typename Size, typename Pointer>
+__host__ __device__
+  Pointer copy_construct_range_n(thrust::execution_policy<System> &from_system,
+                                 Allocator &a,
+                                 InputIterator first,
+                                 Size n,
+                                 Pointer result)
+{
+  return allocator_traits_detail::copy_construct_range_n(from_system, a, first, n, result);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.h
new file mode 100644
index 000000000000..8b5026c05b34
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.h
@@ -0,0 +1,36 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+inline void default_construct_range(Allocator &a, Pointer p, Size n);
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/default_construct_range.inl>
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.inl
new file mode 100644
index 000000000000..6d26578fa230
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/default_construct_range.inl
@@ -0,0 +1,112 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/for_each.h>
+#include <thrust/uninitialized_fill.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace allocator_traits_detail
+{
+
+
+template<typename Allocator>
+  struct construct1_via_allocator
+{
+  Allocator &a;
+
+  __host__ __device__
+  construct1_via_allocator(Allocator &a)
+    : a(a)
+  {}
+
+  template<typename T>
+  inline __host__ __device__
+  void operator()(T &x)
+  {
+    allocator_traits<Allocator>::construct(a, &x);
+  }
+};
+
+
+// we need to construct T via the allocator if...
+template<typename Allocator, typename T>
+  struct needs_default_construct_via_allocator
+    : thrust::detail::or_<
+        has_member_construct1<Allocator,T>,               // if the Allocator does something interesting
+        thrust::detail::not_<has_trivial_constructor<T> > // or if T's default constructor does something interesting
+      >
+{};
+
+
+// we know that std::allocator::construct's only effect is to call T's
+// default constructor, so we needn't use it for default construction
+// unless T's constructor does something interesting
+template<typename U, typename T>
+  struct needs_default_construct_via_allocator<std::allocator<U>, T>
+    : thrust::detail::not_<has_trivial_constructor<T> >
+{};
+
+
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  typename enable_if<
+    needs_default_construct_via_allocator<
+      Allocator,
+      typename pointer_element<Pointer>::type
+    >::value
+  >::type
+    default_construct_range(Allocator &a, Pointer p, Size n)
+{
+  thrust::for_each_n(allocator_system<Allocator>::get(a), p, n, construct1_via_allocator<Allocator>(a));
+}
+
+
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  typename disable_if<
+    needs_default_construct_via_allocator<
+      Allocator,
+      typename pointer_element<Pointer>::type
+    >::value
+  >::type
+    default_construct_range(Allocator &a, Pointer p, Size n)
+{
+  thrust::uninitialized_fill_n(allocator_system<Allocator>::get(a), p, n, typename pointer_element<Pointer>::type());
+}
+
+
+} // end allocator_traits_detail
+
+
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  void default_construct_range(Allocator &a, Pointer p, Size n)
+{
+  return allocator_traits_detail::default_construct_range(a,p,n);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.h
new file mode 100644
index 000000000000..cfc7e3f6ef51
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.h
@@ -0,0 +1,33 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  inline void destroy_range(Allocator &a, Pointer p, Size n);
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/destroy_range.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.inl
new file mode 100644
index 000000000000..662177f3a72f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/destroy_range.inl
@@ -0,0 +1,167 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/allocator/destroy_range.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/for_each.h>
+#include <thrust/detail/memory_wrapper.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace allocator_traits_detail
+{
+
+
+// destroy_range has three cases:
+// if Allocator has an effectful member function destroy:
+//   1. destroy via the allocator
+// else
+//   2. if T has a non-trivial destructor, destroy the range without using the allocator
+//   3. if T has a trivial destructor, do a no-op
+
+template<typename Allocator, typename T>
+  struct has_effectful_member_destroy
+    : has_member_destroy<Allocator,T>
+{};
+
+// std::allocator::destroy's only effect is to invoke its argument's destructor
+template<typename U, typename T>
+  struct has_effectful_member_destroy<std::allocator<U>, T>
+    : thrust::detail::false_type
+{};
+
+// case 1: Allocator has an effectful 1-argument member function "destroy"
+template<typename Allocator, typename Pointer>
+  struct enable_if_destroy_range_case1
+    : thrust::detail::enable_if<
+        has_effectful_member_destroy<
+          Allocator,
+          typename pointer_element<Pointer>::type
+        >::value
+      >
+{};
+
+// case 2: Allocator has no member function "destroy", but T has a non-trivial destructor
+template<typename Allocator, typename Pointer>
+  struct enable_if_destroy_range_case2
+    : thrust::detail::enable_if<
+        !has_effectful_member_destroy<
+          Allocator,
+          typename pointer_element<Pointer>::type
+        >::value &&
+        !has_trivial_destructor<
+          typename pointer_element<Pointer>::type
+        >::value
+      >
+{};
+
+// case 3: Allocator has no member function "destroy", and T has a trivial destructor
+template<typename Allocator, typename Pointer>
+  struct enable_if_destroy_range_case3
+    : thrust::detail::enable_if<
+        !has_effectful_member_destroy<
+          Allocator,
+          typename pointer_element<Pointer>::type
+        >::value &&
+        has_trivial_destructor<
+          typename pointer_element<Pointer>::type
+        >::value
+      >
+{};
+
+
+
+template<typename Allocator>
+  struct destroy_via_allocator
+{
+  Allocator &a;
+
+  __host__ __device__
+  destroy_via_allocator(Allocator &a)
+    : a(a)
+  {}
+
+  template<typename T>
+  inline __host__ __device__
+  void operator()(T &x)
+  {
+    allocator_traits<Allocator>::destroy(a, &x);
+  }
+};
+
+
+// destroy_range case 1: destroy via allocator
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  typename enable_if_destroy_range_case1<Allocator,Pointer>::type
+    destroy_range(Allocator &a, Pointer p, Size n)
+{
+  thrust::for_each_n(allocator_system<Allocator>::get(a), p, n, destroy_via_allocator<Allocator>(a));
+}
+
+
+// we must prepare for His coming
+struct gozer
+{
+  __thrust_exec_check_disable__
+  template<typename T>
+  inline __host__ __device__
+  void operator()(T &x)
+  {
+    x.~T();
+  }
+};
+
+// destroy_range case 2: destroy without the allocator
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  typename enable_if_destroy_range_case2<Allocator,Pointer>::type
+    destroy_range(Allocator &a, Pointer p, Size n)
+{
+  thrust::for_each_n(allocator_system<Allocator>::get(a), p, n, gozer());
+}
+
+
+// destroy_range case 3: no-op
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  typename enable_if_destroy_range_case3<Allocator,Pointer>::type
+    destroy_range(Allocator &, Pointer, Size)
+{
+  // no op
+}
+
+
+} // end allocator_traits_detail
+
+
+template<typename Allocator, typename Pointer, typename Size>
+__host__ __device__
+  void destroy_range(Allocator &a, Pointer p, Size n)
+{
+  return allocator_traits_detail::destroy_range(a,p,n);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.h
new file mode 100644
index 000000000000..a7572cb2d2bc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.h
@@ -0,0 +1,35 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename Allocator, typename Pointer, typename Size, typename T>
+__host__ __device__
+inline void fill_construct_range(Allocator &a, Pointer p, Size n, const T &value);
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/fill_construct_range.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.inl
new file mode 100644
index 000000000000..876b5ddd2647
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/fill_construct_range.inl
@@ -0,0 +1,114 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/for_each.h>
+#include <thrust/uninitialized_fill.h>
+#include <thrust/detail/memory_wrapper.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace allocator_traits_detail
+{
+
+// fill_construct_range has 2 cases:
+// if Allocator has an effectful member function construct:
+//   1. construct via the allocator
+// else
+//   2. construct via uninitialized_fill
+
+template<typename Allocator, typename T, typename Arg1>
+  struct has_effectful_member_construct2
+    : has_member_construct2<Allocator,T,Arg1>
+{};
+
+// std::allocator::construct's only effect is to invoke placement new
+template<typename U, typename T, typename Arg1>
+  struct has_effectful_member_construct2<std::allocator<U>,T,Arg1>
+    : thrust::detail::false_type
+{};
+
+
+template<typename Allocator, typename Arg1>
+  struct construct2_via_allocator
+{
+  Allocator &a;
+  Arg1 arg;
+
+  __host__ __device__
+  construct2_via_allocator(Allocator &a, const Arg1 &arg)
+    : a(a), arg(arg)
+  {}
+
+  template<typename T>
+  inline __host__ __device__
+  void operator()(T &x)
+  {
+    allocator_traits<Allocator>::construct(a, &x, arg);
+  }
+};
+
+
+template<typename Allocator, typename Pointer, typename Size, typename T>
+__host__ __device__
+  typename enable_if<
+    has_effectful_member_construct2<
+      Allocator,
+      typename pointer_element<Pointer>::type,
+      T
+    >::value
+  >::type
+    fill_construct_range(Allocator &a, Pointer p, Size n, const T &value)
+{
+  thrust::for_each_n(allocator_system<Allocator>::get(a), p, n, construct2_via_allocator<Allocator,T>(a, value));
+}
+
+
+template<typename Allocator, typename Pointer, typename Size, typename T>
+__host__ __device__
+  typename disable_if<
+    has_effectful_member_construct2<
+      Allocator,
+      typename pointer_element<Pointer>::type,
+      T
+    >::value
+  >::type
+    fill_construct_range(Allocator &a, Pointer p, Size n, const T &value)
+{
+  thrust::uninitialized_fill_n(allocator_system<Allocator>::get(a), p, n, value);
+}
+
+
+} // end allocator_traits_detail
+
+
+template<typename Alloc, typename Pointer, typename Size, typename T>
+__host__ __device__
+  void fill_construct_range(Alloc &a, Pointer p, Size n, const T &value)
+{
+  return allocator_traits_detail::fill_construct_range(a,p,n,value);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.h
new file mode 100644
index 000000000000..af3d0fccbd74
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.h
@@ -0,0 +1,51 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/detail/allocator/tagged_allocator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template<typename T, typename System, typename Pointer>
+  class malloc_allocator
+    : public thrust::detail::tagged_allocator<
+               T, System, Pointer
+             >
+{
+  private:
+    typedef thrust::detail::tagged_allocator<
+      T, System, Pointer
+    > super_t;
+
+  public:
+    typedef typename super_t::pointer   pointer;
+    typedef typename super_t::size_type size_type;
+
+    pointer allocate(size_type cnt);
+
+    void deallocate(pointer p, size_type n);
+};
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/malloc_allocator.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.inl
new file mode 100644
index 000000000000..d03d33305cb3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/malloc_allocator.inl
@@ -0,0 +1,65 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/malloc_allocator.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/bad_alloc.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/malloc_and_free.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename T, typename System, typename Pointer>
+  typename malloc_allocator<T,System,Pointer>::pointer
+    malloc_allocator<T,System,Pointer>
+      ::allocate(typename malloc_allocator<T,System,Pointer>::size_type cnt)
+{
+  using thrust::system::detail::generic::select_system;
+
+  // XXX should use a hypothetical thrust::static_pointer_cast here
+  System system;
+
+  pointer result = thrust::malloc<T>(select_system(system), cnt);
+
+  if(result.get() == 0)
+  {
+    throw thrust::system::detail::bad_alloc("malloc_allocator::allocate: malloc failed");
+  } // end if
+
+  return result;
+} // end malloc_allocator::allocate()
+
+
+template<typename T, typename System, typename Pointer>
+  void malloc_allocator<T,System,Pointer>
+    ::deallocate(typename malloc_allocator<T,System,Pointer>::pointer p, typename malloc_allocator<T,System,Pointer>::size_type)
+{
+  using thrust::system::detail::generic::select_system;
+
+  System system;
+  thrust::free(select_system(system), p);
+} // end malloc_allocator
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/no_throw_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/no_throw_allocator.h
new file mode 100644
index 000000000000..a6c16985b11b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/no_throw_allocator.h
@@ -0,0 +1,72 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <nv/target>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template<typename BaseAllocator>
+  struct no_throw_allocator : BaseAllocator
+{
+  private:
+    typedef BaseAllocator super_t;
+  
+  public:
+    inline __host__ __device__
+    no_throw_allocator(const BaseAllocator &other = BaseAllocator())
+      : super_t(other)
+    {}
+
+    template<typename U>
+      struct rebind
+    {
+      typedef no_throw_allocator<typename super_t::template rebind<U>::other> other;
+    }; // end rebind
+
+    __host__ __device__
+    void deallocate(typename super_t::pointer p, typename super_t::size_type n)
+    {
+      NV_IF_TARGET(NV_IS_HOST, (
+        try
+        {
+          super_t::deallocate(p, n);
+        } // end try
+        catch(...)
+        {
+          // catch anything
+        } // end catch
+      ), (
+        super_t::deallocate(p, n);
+      ));
+    } // end deallocate()
+
+    inline __host__ __device__
+    bool operator==(no_throw_allocator const &other) { return super_t::operator==(other); }
+
+    inline __host__ __device__
+    bool operator!=(no_throw_allocator const &other) { return super_t::operator!=(other); }
+}; // end no_throw_allocator
+
+} // end detail
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.h
new file mode 100644
index 000000000000..804c4e42e511
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.h
@@ -0,0 +1,100 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template<typename T, typename Tag, typename Pointer> class tagged_allocator;
+
+template<typename Tag, typename Pointer>
+  class tagged_allocator<void, Tag, Pointer>
+{
+  public:
+    typedef void                                                                                 value_type;
+    typedef typename thrust::detail::pointer_traits<Pointer>::template rebind<void>::other       pointer;
+    typedef typename thrust::detail::pointer_traits<Pointer>::template rebind<const void>::other const_pointer;
+    typedef std::size_t                                                                          size_type;
+    typedef typename thrust::detail::pointer_traits<Pointer>::difference_type                    difference_type;
+    typedef Tag                                                                                  system_type;
+
+    template<typename U>
+      struct rebind
+    {
+      typedef tagged_allocator<U,Tag,Pointer> other;
+    }; // end rebind
+};
+
+template<typename T, typename Tag, typename Pointer>
+  class tagged_allocator
+{
+  public:
+    typedef T                                                                                 value_type;
+    typedef typename thrust::detail::pointer_traits<Pointer>::template rebind<T>::other       pointer;
+    typedef typename thrust::detail::pointer_traits<Pointer>::template rebind<const T>::other const_pointer;
+    typedef typename thrust::iterator_reference<pointer>::type                                reference;
+    typedef typename thrust::iterator_reference<const_pointer>::type                          const_reference;
+    typedef std::size_t                                                                       size_type;
+    typedef typename thrust::detail::pointer_traits<pointer>::difference_type                 difference_type;
+    typedef Tag                                                                               system_type;
+
+    template<typename U>
+      struct rebind
+    {
+      typedef tagged_allocator<U,Tag,Pointer> other;
+    }; // end rebind
+
+    __host__ __device__
+    inline tagged_allocator();
+
+    __host__ __device__
+    inline tagged_allocator(const tagged_allocator &);
+
+    template<typename U, typename OtherPointer>
+    __host__ __device__
+    inline tagged_allocator(const tagged_allocator<U, Tag, OtherPointer> &);
+
+    __host__ __device__
+    inline ~tagged_allocator();
+
+    __host__ __device__
+    pointer address(reference x) const;
+
+    __host__ __device__
+    const_pointer address(const_reference x) const;
+
+    size_type max_size() const;
+};
+
+template<typename T1, typename Pointer1, typename T2, typename Pointer2, typename Tag>
+__host__ __device__
+bool operator==(const tagged_allocator<T1,Pointer1,Tag> &, const tagged_allocator<T2,Pointer2,Tag> &);
+
+template<typename T1, typename Pointer1, typename T2, typename Pointer2, typename Tag>
+__host__ __device__
+bool operator!=(const tagged_allocator<T1,Pointer1,Tag> &, const tagged_allocator<T2,Pointer2,Tag> &);
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/tagged_allocator.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.inl
new file mode 100644
index 000000000000..bcd534cbceeb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/tagged_allocator.inl
@@ -0,0 +1,104 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/tagged_allocator.h>
+#include <limits>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename T, typename Tag, typename Pointer>
+  __host__ __device__
+  tagged_allocator<T,Tag,Pointer>
+    ::tagged_allocator()
+{}
+
+
+template<typename T, typename Tag, typename Pointer>
+  __host__ __device__
+  tagged_allocator<T,Tag,Pointer>
+    ::tagged_allocator(const tagged_allocator<T,Tag,Pointer> &)
+{}
+
+
+template<typename T, typename Tag, typename Pointer>
+  template<typename U, typename OtherPointer>
+    __host__ __device__
+    tagged_allocator<T,Tag,Pointer>
+      ::tagged_allocator(const tagged_allocator<U,Tag,OtherPointer> &)
+{}
+
+
+template<typename T, typename Tag, typename Pointer>
+  __host__ __device__
+  tagged_allocator<T,Tag,Pointer>
+    ::~tagged_allocator()
+{}
+
+
+template<typename T, typename Tag, typename Pointer>
+  __host__ __device__
+  typename tagged_allocator<T,Tag,Pointer>::pointer
+    tagged_allocator<T,Tag,Pointer>
+      ::address(reference x) const
+{
+  return &x;
+}
+
+
+template<typename T, typename Tag, typename Pointer>
+  __host__ __device__
+  typename tagged_allocator<T,Tag,Pointer>::const_pointer
+    tagged_allocator<T,Tag,Pointer>
+      ::address(const_reference x) const
+{
+  return &x;
+}
+
+
+template<typename T, typename Tag, typename Pointer>
+  typename tagged_allocator<T,Tag,Pointer>::size_type
+    tagged_allocator<T,Tag,Pointer>
+      ::max_size() const
+{
+  return (std::numeric_limits<size_type>::max)() / sizeof(T);
+}
+
+
+template<typename T1, typename Pointer1, typename T2, typename Pointer2, typename Tag>
+__host__ __device__
+bool operator==(const tagged_allocator<T1,Pointer1,Tag> &, const tagged_allocator<T2,Pointer2,Tag> &)
+{
+  return true;
+}
+
+
+template<typename T1, typename Pointer1, typename T2, typename Pointer2, typename Tag>
+__host__ __device__
+bool operator!=(const tagged_allocator<T1,Pointer1,Tag> &, const tagged_allocator<T2,Pointer2,Tag> &)
+{
+  return false;
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.h
new file mode 100644
index 000000000000..c8ef606258a0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.h
@@ -0,0 +1,84 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/tagged_allocator.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/pair.h>
+#include <thrust/memory.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+// XXX the pointer parameter given to tagged_allocator should be related to
+//     the type of the expression get_temporary_buffer(system, n).first
+//     without decltype, compromise on pointer<T,System>
+template<typename T, typename System>
+  class temporary_allocator
+    : public thrust::detail::tagged_allocator<
+               T, System, thrust::pointer<T,System>
+             >
+{
+  private:
+    typedef thrust::detail::tagged_allocator<
+      T, System, thrust::pointer<T,System>
+    > super_t;
+
+    System &m_system;
+
+  public:
+    typedef typename super_t::pointer   pointer;
+    typedef typename super_t::size_type size_type;
+
+    inline __host__ __device__
+    temporary_allocator(const temporary_allocator &other) :
+      super_t(),
+      m_system(other.m_system)
+    {}
+
+    inline __host__ __device__
+    explicit temporary_allocator(thrust::execution_policy<System> &system) :
+      super_t(),
+      m_system(thrust::detail::derived_cast(system))
+    {}
+
+    __host__ __device__
+    pointer allocate(size_type cnt);
+
+    __host__ __device__
+    void deallocate(pointer p, size_type n);
+
+    __host__ __device__
+    inline System &system()
+    {
+      return m_system;
+    } // end system()
+
+  private:
+    typedef thrust::pair<pointer, size_type> pointer_and_size;
+}; // end temporary_allocator
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/allocator/temporary_allocator.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.inl
new file mode 100644
index 000000000000..ef5d1afa503b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator/temporary_allocator.inl
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator/temporary_allocator.h>
+#include <thrust/detail/temporary_buffer.h>
+#include <thrust/system/detail/bad_alloc.h>
+#include <cassert>
+
+#include <nv/target>
+
+#if THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
+#if (defined(_NVHPC_CUDA) || defined(__CUDA_ARCH__))
+#include <thrust/system/cuda/detail/terminate.h>
+#endif // NVCC device pass or NVC++
+#endif // CUDA
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename T, typename System>
+__host__ __device__
+  typename temporary_allocator<T,System>::pointer
+    temporary_allocator<T,System>
+      ::allocate(typename temporary_allocator<T,System>::size_type cnt)
+{
+  pointer_and_size result = thrust::get_temporary_buffer<T>(system(), cnt);
+
+  // handle failure
+  if(result.second < cnt)
+  {
+    // deallocate and throw
+    // note that we pass cnt to deallocate, not a value derived from result.second
+    deallocate(result.first, cnt);
+
+#if THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
+    NV_IF_TARGET(NV_IS_HOST, (
+      throw thrust::system::detail::bad_alloc("temporary_buffer::allocate: get_temporary_buffer failed");
+    ), ( // NV_IS_DEVICE
+      thrust::system::cuda::detail::terminate_with_message("temporary_buffer::allocate: get_temporary_buffer failed");
+    ));
+#else
+    throw thrust::system::detail::bad_alloc("temporary_buffer::allocate: get_temporary_buffer failed");
+#endif
+  } // end if
+
+  return result.first;
+} // end temporary_allocator::allocate()
+
+
+template<typename T, typename System>
+__host__ __device__
+  void temporary_allocator<T,System>
+    ::deallocate(typename temporary_allocator<T,System>::pointer p, typename temporary_allocator<T,System>::size_type n)
+{
+  return thrust::return_temporary_buffer(system(), p, n);
+} // end temporary_allocator
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator_aware_execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator_aware_execution_policy.h
new file mode 100644
index 000000000000..eea93c035eab
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/allocator_aware_execution_policy.h
@@ -0,0 +1,101 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execute_with_allocator_fwd.h>
+#include <thrust/detail/alignment.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+  #include <type_traits>
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+namespace mr
+{
+
+template<typename T, class MR>
+class allocator;
+
+}
+
+namespace detail
+{
+
+template<template <typename> class ExecutionPolicyCRTPBase>
+struct allocator_aware_execution_policy
+{
+  template<typename MemoryResource>
+  struct execute_with_memory_resource_type
+  {
+    typedef thrust::detail::execute_with_allocator<
+      thrust::mr::allocator<
+        thrust::detail::max_align_t,
+        MemoryResource
+      >,
+      ExecutionPolicyCRTPBase
+    > type;
+  };
+
+  template<typename Allocator>
+  struct execute_with_allocator_type
+  {
+      typedef thrust::detail::execute_with_allocator<
+        Allocator,
+        ExecutionPolicyCRTPBase
+      > type;
+  };
+
+  template<typename MemoryResource>
+    typename execute_with_memory_resource_type<MemoryResource>::type
+      operator()(MemoryResource * mem_res) const
+  {
+    return typename execute_with_memory_resource_type<MemoryResource>::type(mem_res);
+  }
+
+  template<typename Allocator>
+    typename execute_with_allocator_type<Allocator&>::type
+      operator()(Allocator &alloc) const
+  {
+    return typename execute_with_allocator_type<Allocator&>::type(alloc);
+  }
+
+  template<typename Allocator>
+    typename execute_with_allocator_type<Allocator>::type
+      operator()(const Allocator &alloc) const
+  {
+    return typename execute_with_allocator_type<Allocator>::type(alloc);
+  }
+
+#if THRUST_CPP_DIALECT >= 2011
+  // just the rvalue overload
+  // perfect forwarding doesn't help, because a const reference has to be turned
+  // into a value by copying for the purpose of storing it in execute_with_allocator
+  template<typename Allocator,
+      typename std::enable_if<!std::is_lvalue_reference<Allocator>::value>::type * = nullptr>
+    typename execute_with_allocator_type<Allocator>::type
+      operator()(Allocator &&alloc) const
+  {
+    return typename execute_with_allocator_type<Allocator>::type(std::move(alloc));
+  }
+#endif
+};
+
+} // end namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/binary_search.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/binary_search.inl
new file mode 100644
index 000000000000..90350ced4bb3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/binary_search.inl
@@ -0,0 +1,480 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/binary_search.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/binary_search.h>
+#include <thrust/system/detail/adl/binary_search.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const LessThanComparable &value)
+{
+    using thrust::system::detail::generic::lower_bound;
+    return lower_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const T &value,
+                            StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::lower_bound;
+    return lower_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value, comp);
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const LessThanComparable &value)
+{
+    using thrust::system::detail::generic::upper_bound;
+    return upper_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const T &value,
+                            StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::upper_bound;
+    return upper_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value, comp);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   ForwardIterator first,
+                   ForwardIterator last,
+                   const LessThanComparable& value)
+{
+    using thrust::system::detail::generic::binary_search;
+    return binary_search(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+bool binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   ForwardIterator first,
+                   ForwardIterator last,
+                   const T& value,
+                   StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::binary_search;
+    return binary_search(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value, comp);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const T& value,
+            StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::equal_range;
+    return equal_range(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value, comp);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable& value)
+{
+    using thrust::system::detail::generic::equal_range;
+    return equal_range(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output)
+{
+    using thrust::system::detail::generic::lower_bound;
+    return lower_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, values_first, values_last, output);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator lower_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output,
+                           StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::lower_bound;
+    return lower_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, values_first, values_last, output, comp);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output)
+{
+    using thrust::system::detail::generic::upper_bound;
+    return upper_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, values_first, values_last, output);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator upper_bound(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output,
+                           StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::upper_bound;
+    return upper_bound(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, values_first, values_last, output, comp);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                             ForwardIterator first,
+                             ForwardIterator last,
+                             InputIterator values_first,
+                             InputIterator values_last,
+                             OutputIterator output)
+{
+    using thrust::system::detail::generic::binary_search;
+    return binary_search(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, values_first, values_last, output);
+}
+
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator binary_search(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                             ForwardIterator first,
+                             ForwardIterator last,
+                             InputIterator values_first,
+                             InputIterator values_last,
+                             OutputIterator output,
+                             StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::binary_search;
+    return binary_search(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, values_first, values_last, output, comp);
+}
+
+
+//////////////////////
+// Scalar Functions //
+//////////////////////
+
+template <typename ForwardIterator, typename LessThanComparable>
+ForwardIterator lower_bound(ForwardIterator first,
+                            ForwardIterator last,
+                            const LessThanComparable& value)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::lower_bound(select_system(system), first, last, value);
+}
+
+template <typename ForwardIterator, typename T, typename StrictWeakOrdering>
+ForwardIterator lower_bound(ForwardIterator first,
+                            ForwardIterator last,
+                            const T& value,
+                            StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::lower_bound(select_system(system), first, last, value, comp);
+}
+
+template <typename ForwardIterator, typename LessThanComparable>
+ForwardIterator upper_bound(ForwardIterator first,
+                            ForwardIterator last,
+                            const LessThanComparable& value)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::upper_bound(select_system(system), first, last, value);
+}
+
+template <typename ForwardIterator, typename T, typename StrictWeakOrdering>
+ForwardIterator upper_bound(ForwardIterator first,
+                            ForwardIterator last,
+                            const T& value,
+                            StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::upper_bound(select_system(system), first, last, value, comp);
+}
+
+template <typename ForwardIterator, typename LessThanComparable>
+bool binary_search(ForwardIterator first,
+                   ForwardIterator last,
+                   const LessThanComparable& value)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::binary_search(select_system(system), first, last, value);
+}
+
+template <typename ForwardIterator, typename T, typename StrictWeakOrdering>
+bool binary_search(ForwardIterator first,
+                   ForwardIterator last,
+                   const T& value,
+                   StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::binary_search(select_system(system), first, last, value, comp);
+}
+
+template <typename ForwardIterator, typename LessThanComparable>
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable& value)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::equal_range(select_system(system), first, last, value);
+}
+
+template <typename ForwardIterator, typename T, typename StrictWeakOrdering>
+thrust::pair<ForwardIterator, ForwardIterator>
+equal_range(ForwardIterator first,
+            ForwardIterator last,
+            const T& value,
+            StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+    System system;
+
+    return thrust::equal_range(select_system(system), first, last, value, comp);
+}
+
+//////////////////////
+// Vector Functions //
+//////////////////////
+
+template <typename ForwardIterator, typename InputIterator, typename OutputIterator>
+OutputIterator lower_bound(ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+    typedef typename thrust::iterator_system<InputIterator>::type   System2;
+    typedef typename thrust::iterator_system<OutputIterator>::type  System3;
+
+    System1 system1;
+    System2 system2;
+    System3 system3;
+
+    return thrust::lower_bound(select_system(system1,system2,system3), first, last, values_first, values_last, output);
+}
+
+template <typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+OutputIterator lower_bound(ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output,
+                           StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+    typedef typename thrust::iterator_system<InputIterator>::type   System2;
+    typedef typename thrust::iterator_system<OutputIterator>::type  System3;
+
+    System1 system1;
+    System2 system2;
+    System3 system3;
+
+    return thrust::lower_bound(select_system(system1,system2,system3), first, last, values_first, values_last, output, comp);
+}
+
+template <typename ForwardIterator, typename InputIterator, typename OutputIterator>
+OutputIterator upper_bound(ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+    typedef typename thrust::iterator_system<InputIterator>::type   System2;
+    typedef typename thrust::iterator_system<OutputIterator>::type  System3;
+
+    System1 system1;
+    System2 system2;
+    System3 system3;
+
+    return thrust::upper_bound(select_system(system1,system2,system3), first, last, values_first, values_last, output);
+}
+
+template <typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+OutputIterator upper_bound(ForwardIterator first,
+                           ForwardIterator last,
+                           InputIterator values_first,
+                           InputIterator values_last,
+                           OutputIterator output,
+                           StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+    typedef typename thrust::iterator_system<InputIterator>::type   System2;
+    typedef typename thrust::iterator_system<OutputIterator>::type  System3;
+
+    System1 system1;
+    System2 system2;
+    System3 system3;
+
+    return thrust::upper_bound(select_system(system1,system2,system3), first, last, values_first, values_last, output, comp);
+}
+
+template <typename ForwardIterator, typename InputIterator, typename OutputIterator>
+OutputIterator binary_search(ForwardIterator first,
+                             ForwardIterator last,
+                             InputIterator values_first,
+                             InputIterator values_last,
+                             OutputIterator output)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+    typedef typename thrust::iterator_system<InputIterator>::type   System2;
+    typedef typename thrust::iterator_system<OutputIterator>::type  System3;
+
+    System1 system1;
+    System2 system2;
+    System3 system3;
+
+    return thrust::binary_search(select_system(system1,system2,system3), first, last, values_first, values_last, output);
+}
+
+template <typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+OutputIterator binary_search(ForwardIterator first,
+                             ForwardIterator last,
+                             InputIterator values_first,
+                             InputIterator values_last,
+                             OutputIterator output,
+                             StrictWeakOrdering comp)
+{
+    using thrust::system::detail::generic::select_system;
+
+    typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+    typedef typename thrust::iterator_system<InputIterator>::type   System2;
+    typedef typename thrust::iterator_system<OutputIterator>::type  System3;
+
+    System1 system1;
+    System2 system2;
+    System3 system3;
+
+    return thrust::binary_search(select_system(system1,system2,system3), first, last, values_first, values_last, output, comp);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/caching_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/caching_allocator.h
new file mode 100644
index 000000000000..941f52755ca6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/caching_allocator.h
@@ -0,0 +1,47 @@
+/*
+ *  Copyright 2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/mr/allocator.h>
+#include <thrust/mr/disjoint_tls_pool.h>
+#include <thrust/mr/new.h>
+#include <thrust/mr/device_memory_resource.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+inline
+thrust::mr::allocator<
+    char,
+    thrust::mr::disjoint_unsynchronized_pool_resource<
+        thrust::device_memory_resource,
+        thrust::mr::new_delete_resource
+    >
+> single_device_tls_caching_allocator()
+{
+    return {
+        &thrust::mr::tls_disjoint_pool(
+            thrust::mr::get_global_resource<thrust::device_memory_resource>(),
+            thrust::mr::get_global_resource<thrust::mr::new_delete_resource>()
+        )
+    };
+}
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/arithmetic.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/arithmetic.h
new file mode 100644
index 000000000000..518f184503c6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/arithmetic.h
@@ -0,0 +1,303 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/c99math.h>
+#include <cfloat>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+
+  /* --- Binary Arithmetic Operators --- */
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator+(const complex<T0>& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x.real() + y.real(), x.imag() + y.imag());
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator+(const complex<T0>& x, const T1& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x.real() + y, x.imag());
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator+(const T0& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x + y.real(), y.imag());
+}
+
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator-(const complex<T0>& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x.real() - y.real(), x.imag() - y.imag());
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator-(const complex<T0>& x, const T1& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x.real() - y, x.imag());
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator-(const T0& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x - y.real(), -y.imag());
+}
+
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator*(const complex<T0>& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>( x.real() * y.real() - x.imag() * y.imag()
+			             , x.real() * y.imag() + x.imag() * y.real());
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator*(const complex<T0>& x, const T1& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x.real() * y, x.imag() * y);
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator*(const T0& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x * y.real(), x * y.imag());
+}
+
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator/(const complex<T0>& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+
+  // Find `abs` by ADL.
+  using std::abs;
+
+  T s = abs(y.real()) + abs(y.imag());
+
+  T oos = T(1.0) / s;
+
+  T ars = x.real() * oos;
+  T ais = x.imag() * oos;
+  T brs = y.real() * oos;
+  T bis = y.imag() * oos;
+
+  s = (brs * brs) + (bis * bis);
+
+  oos = T(1.0) / s;
+
+  complex<T> quot( ((ars * brs) + (ais * bis)) * oos
+                 , ((ais * brs) - (ars * bis)) * oos);
+  return quot;
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator/(const complex<T0>& x, const T1& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x.real() / y, x.imag() / y);
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+operator/(const T0& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return complex<T>(x) / y;
+}
+
+
+
+/* --- Unary Arithmetic Operators --- */
+
+template <typename T>
+__host__ __device__
+complex<T> operator+(const complex<T>& y)
+{
+  return y;
+}
+
+template <typename T>
+__host__ __device__
+complex<T> operator-(const complex<T>& y)
+{
+  return y * -T(1);
+}
+
+
+/* --- Other Basic Arithmetic Functions --- */
+
+// As std::hypot is only C++11 we have to use the C interface
+template <typename T>
+__host__ __device__
+T abs(const complex<T>& z)
+{
+  return hypot(z.real(), z.imag());
+}
+
+// XXX Why are we specializing here?
+namespace detail {
+namespace complex {
+
+__host__ __device__
+inline float abs(const thrust::complex<float>& z)
+{
+  return hypotf(z.real(),z.imag());
+}
+
+__host__ __device__
+inline double abs(const thrust::complex<double>& z)
+{
+  return hypot(z.real(),z.imag());
+}
+
+} // end namespace complex
+} // end namespace detail
+
+template <>
+__host__ __device__
+inline float abs(const complex<float>& z)
+{
+  return detail::complex::abs(z);
+}
+
+template <>
+__host__ __device__
+inline double abs(const complex<double>& z)
+{
+  return detail::complex::abs(z);
+}
+
+
+template <typename T>
+__host__ __device__
+T arg(const complex<T>& z)
+{
+  // Find `atan2` by ADL.
+  using std::atan2;
+  return atan2(z.imag(), z.real());
+}
+
+
+template <typename T>
+__host__ __device__
+complex<T> conj(const complex<T>& z)
+{
+  return complex<T>(z.real(), -z.imag());
+}
+
+
+template <typename T>
+__host__ __device__
+T norm(const complex<T>& z)
+{
+  return z.real() * z.real() + z.imag() * z.imag();
+}
+
+// XXX Why specialize these, we could just rely on ADL.
+template <>
+__host__ __device__
+inline float norm(const complex<float>& z)
+{
+  // Find `abs` and `sqrt` by ADL.
+  using std::abs;
+  using std::sqrt;
+
+  if (abs(z.real()) < sqrt(FLT_MIN) && abs(z.imag()) < sqrt(FLT_MIN))
+  {
+    float a = z.real() * 4.0f;
+    float b = z.imag() * 4.0f;
+    return (a * a + b * b) / 16.0f;
+  }
+
+  return z.real() * z.real() + z.imag() * z.imag();
+}
+
+template <>
+__host__ __device__
+inline double norm(const complex<double>& z)
+{
+  // Find `abs` and `sqrt` by ADL.
+  using std::abs;
+  using std::sqrt;
+
+  if (abs(z.real()) < sqrt(DBL_MIN) && abs(z.imag()) < sqrt(DBL_MIN))
+  {
+    double a = z.real() * 4.0;
+    double b = z.imag() * 4.0;
+    return (a * a + b * b) / 16.0;
+  }
+
+  return z.real() * z.real() + z.imag() * z.imag();
+}
+
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+polar(const T0& m, const T1& theta)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+
+  // Find `cos` and `sin` by ADL.
+  using std::cos;
+  using std::sin;
+
+  return complex<T>(m * cos(theta), m * sin(theta));
+}
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/c99math.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/c99math.h
new file mode 100644
index 000000000000..e735b850c73b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/c99math.h
@@ -0,0 +1,197 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <math.h>
+#include <cmath>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace complex
+{
+
+// Define basic arithmetic functions so we can use them without explicit scope
+// keeping the code as close as possible to FreeBSDs for ease of maintenance.
+// It also provides an easy way to support compilers with missing C99 functions.
+// When possible, just use the names in the global scope.
+// Some platforms define these as macros, others as free functions.
+// Avoid using the std:: form of these as nvcc may treat std::foo() as __host__ functions.
+
+using ::log;
+using ::acos;
+using ::asin;
+using ::sqrt;
+using ::sinh;
+using ::tan;
+using ::cos;
+using ::sin;
+using ::exp;
+using ::cosh;
+using ::atan;
+
+template <typename T>
+inline __host__ __device__ T infinity();
+
+template <>
+inline __host__ __device__ float infinity<float>()
+{
+  float res;
+  set_float_word(res, 0x7f800000);
+  return res;
+}
+
+
+template <>
+inline __host__ __device__ double infinity<double>()
+{
+  double res;
+  insert_words(res, 0x7ff00000,0);
+  return res;
+}
+
+#if defined _MSC_VER
+__host__ __device__ inline int isinf(float x){
+  return std::abs(x) == infinity<float>();
+}
+
+__host__ __device__ inline int isinf(double x){
+  return std::abs(x) == infinity<double>();
+}
+
+__host__ __device__ inline int isnan(float x){
+  return x != x;
+}
+
+__host__ __device__ inline int isnan(double x){
+  return x != x;
+}
+
+__host__ __device__ inline int signbit(float x){
+  return ((*((uint32_t *)&x)) & 0x80000000) != 0 ? 1 : 0;
+}
+
+__host__ __device__ inline int signbit(double x){
+  return ((*((uint64_t *)&x)) & 0x8000000000000000) != 0ull ? 1 : 0;
+}
+
+__host__ __device__ inline int isfinite(float x){
+  return !isnan(x) && !isinf(x);
+}
+
+__host__ __device__ inline int isfinite(double x){
+  return !isnan(x) && !isinf(x);
+}
+
+#else
+
+#  if defined(__CUDACC__) && !(defined(__CUDA__) && defined(__clang__)) && !defined(_NVHPC_CUDA)
+// NVCC implements at least some signature of these as functions not macros.
+using ::isinf;
+using ::isnan;
+using ::signbit;
+using ::isfinite;
+#  else
+// Some compilers do not provide these in the global scope, because they are
+// supposed to be macros. The versions in `std` are supposed to be functions.
+// Since we're not compiling with nvcc, it's safe to use the functions in std::
+using std::isinf;
+using std::isnan;
+using std::signbit;
+using std::isfinite;
+#  endif // __CUDACC__
+#endif // _MSC_VER
+
+using ::atanh;
+
+#if defined _MSC_VER
+
+__host__ __device__ inline double copysign(double x, double y){
+  uint32_t hx,hy;
+  get_high_word(hx,x);
+  get_high_word(hy,y);
+  set_high_word(x,(hx&0x7fffffff)|(hy&0x80000000));
+  return x;
+}
+
+__host__ __device__ inline float copysignf(float x, float y){
+  uint32_t ix,iy;
+  get_float_word(ix,x);
+  get_float_word(iy,y);
+  set_float_word(x,(ix&0x7fffffff)|(iy&0x80000000));
+  return x;
+}
+
+
+
+#if !defined(__CUDACC__) && !defined(_NVHPC_CUDA)
+
+// Simple approximation to log1p as Visual Studio is lacking one
+inline double log1p(double x){
+  double u = 1.0+x;
+  if(u == 1.0){
+    return x;
+  }else{
+    if(u > 2.0){
+      // Use normal log for large arguments
+      return log(u);
+    }else{
+      return log(u)*(x/(u-1.0));
+    }
+  }
+}
+
+inline float log1pf(float x){
+  float u = 1.0f+x;
+  if(u == 1.0f){
+    return x;
+  }else{
+    if(u > 2.0f){
+      // Use normal log for large arguments
+      return logf(u);
+    }else{
+      return logf(u)*(x/(u-1.0f));
+    }
+  }
+}
+
+#if _MSV_VER <= 1500
+#include <complex>
+
+inline float hypotf(float x, float y){
+	return abs(std::complex<float>(x,y));
+}
+
+inline double hypot(double x, double y){
+	return _hypot(x,y);
+}
+
+#endif // _MSC_VER <= 1500
+
+#endif // __CUDACC__
+
+#endif // _MSC_VER
+
+} // namespace complex
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrig.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrig.h
new file mode 100644
index 000000000000..4955ec5bfa5c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrig.h
@@ -0,0 +1,785 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2012 Stephen Montgomery-Smith <stephen@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/*
+ * Adapted from FreeBSD by Filipe Maia <filipe.c.maia@gmail.com>:
+ *    freebsd/lib/msun/src/catrig.c
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <cfloat>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{
+
+using thrust::complex;
+
+__host__ __device__
+inline void raise_inexact(){
+  const volatile float tiny = 7.888609052210118054117286e-31; /* 0x1p-100; */
+  // needs the volatile to prevent compiler from ignoring it
+  volatile float junk = 1 + tiny;
+  (void)junk;
+}
+
+__host__ __device__ inline complex<double> clog_for_large_values(complex<double> z);
+
+/*
+ * Testing indicates that all these functions are accurate up to 4 ULP.
+ * The functions casin(h) and cacos(h) are about 2.5 times slower than asinh.
+ * The functions catan(h) are a little under 2 times slower than atanh.
+ *
+ * The code for casinh, casin, cacos, and cacosh comes first.  The code is
+ * rather complicated, and the four functions are highly interdependent.
+ *
+ * The code for catanh and catan comes at the end.  It is much simpler than
+ * the other functions, and the code for these can be disconnected from the
+ * rest of the code.
+ */
+
+/*
+ *			================================
+ *			| casinh, casin, cacos, cacosh |
+ *			================================
+ */
+
+/*
+ * The algorithm is very close to that in "Implementing the complex arcsine
+ * and arccosine functions using exception handling" by T. E. Hull, Thomas F.
+ * Fairgrieve, and Ping Tak Peter Tang, published in ACM Transactions on
+ * Mathematical Software, Volume 23 Issue 3, 1997, Pages 299-335,
+ * http://dl.acm.org/citation.cfm?id=275324.
+ *
+ * Throughout we use the convention z = x + I*y.
+ *
+ * casinh(z) = sign(x)*log(A+sqrt(A*A-1)) + I*asin(B)
+ * where
+ * A = (|z+I| + |z-I|) / 2
+ * B = (|z+I| - |z-I|) / 2 = y/A
+ *
+ * These formulas become numerically unstable:
+ *   (a) for Re(casinh(z)) when z is close to the line segment [-I, I] (that
+ *       is, Re(casinh(z)) is close to 0);
+ *   (b) for Im(casinh(z)) when z is close to either of the intervals
+ *       [I, I*infinity) or (-I*infinity, -I] (that is, |Im(casinh(z))| is
+ *       close to PI/2).
+ *
+ * These numerical problems are overcome by defining
+ * f(a, b) = (hypot(a, b) - b) / 2 = a*a / (hypot(a, b) + b) / 2
+ * Then if A < A_crossover, we use
+ *   log(A + sqrt(A*A-1)) = log1p((A-1) + sqrt((A-1)*(A+1)))
+ *   A-1 = f(x, 1+y) + f(x, 1-y)
+ * and if B > B_crossover, we use
+ *   asin(B) = atan2(y, sqrt(A*A - y*y)) = atan2(y, sqrt((A+y)*(A-y)))
+ *   A-y = f(x, y+1) + f(x, y-1)
+ * where without loss of generality we have assumed that x and y are
+ * non-negative.
+ *
+ * Much of the difficulty comes because the intermediate computations may
+ * produce overflows or underflows.  This is dealt with in the paper by Hull
+ * et al by using exception handling.  We do this by detecting when
+ * computations risk underflow or overflow.  The hardest part is handling the
+ * underflows when computing f(a, b).
+ *
+ * Note that the function f(a, b) does not appear explicitly in the paper by
+ * Hull et al, but the idea may be found on pages 308 and 309.  Introducing the
+ * function f(a, b) allows us to concentrate many of the clever tricks in this
+ * paper into one function.
+ */
+
+/*
+ * Function f(a, b, hypot_a_b) = (hypot(a, b) - b) / 2.
+ * Pass hypot(a, b) as the third argument.
+ */
+__host__ __device__
+inline double
+f(double a, double b, double hypot_a_b)
+{
+  if (b < 0)
+    return ((hypot_a_b - b) / 2);
+  if (b == 0)
+    return (a / 2);
+  return (a * a / (hypot_a_b + b) / 2);
+}
+
+/*
+ * All the hard work is contained in this function.
+ * x and y are assumed positive or zero, and less than RECIP_EPSILON.
+ * Upon return:
+ * rx = Re(casinh(z)) = -Im(cacos(y + I*x)).
+ * B_is_usable is set to 1 if the value of B is usable.
+ * If B_is_usable is set to 0, sqrt_A2my2 = sqrt(A*A - y*y), and new_y = y.
+ * If returning sqrt_A2my2 has potential to result in an underflow, it is
+ * rescaled, and new_y is similarly rescaled.
+ */
+__host__ __device__
+inline void
+do_hard_work(double x, double y, double *rx, int *B_is_usable, double *B,
+			double *sqrt_A2my2, double *new_y)
+{
+  double R, S, A; /* A, B, R, and S are as in Hull et al. */
+  double Am1, Amy; /* A-1, A-y. */
+  const double A_crossover = 10; /* Hull et al suggest 1.5, but 10 works better */
+  const double FOUR_SQRT_MIN = 5.966672584960165394632772e-154; /* =0x1p-509; >= 4 * sqrt(DBL_MIN) */
+  const double B_crossover = 0.6417; /* suggested by Hull et al */
+
+  R = hypot(x, y + 1);		/* |z+I| */
+  S = hypot(x, y - 1);		/* |z-I| */
+
+  /* A = (|z+I| + |z-I|) / 2 */
+  A = (R + S) / 2;
+  /*
+   * Mathematically A >= 1.  There is a small chance that this will not
+   * be so because of rounding errors.  So we will make certain it is
+   * so.
+   */
+  if (A < 1)
+    A = 1;
+
+  if (A < A_crossover) {
+    /*
+     * Am1 = fp + fm, where fp = f(x, 1+y), and fm = f(x, 1-y).
+     * rx = log1p(Am1 + sqrt(Am1*(A+1)))
+     */
+    if (y == 1 && x < DBL_EPSILON * DBL_EPSILON / 128) {
+      /*
+       * fp is of order x^2, and fm = x/2.
+       * A = 1 (inexactly).
+       */
+      *rx = sqrt(x);
+    } else if (x >= DBL_EPSILON * fabs(y - 1)) {
+      /*
+       * Underflow will not occur because
+       * x >= DBL_EPSILON^2/128 >= FOUR_SQRT_MIN
+       */
+      Am1 = f(x, 1 + y, R) + f(x, 1 - y, S);
+      *rx = log1p(Am1 + sqrt(Am1 * (A + 1)));
+    } else if (y < 1) {
+      /*
+       * fp = x*x/(1+y)/4, fm = x*x/(1-y)/4, and
+       * A = 1 (inexactly).
+       */
+      *rx = x / sqrt((1 - y) * (1 + y));
+    } else {		/* if (y > 1) */
+      /*
+       * A-1 = y-1 (inexactly).
+       */
+      *rx = log1p((y - 1) + sqrt((y - 1) * (y + 1)));
+    }
+  } else {
+    *rx = log(A + sqrt(A * A - 1));
+  }
+
+  *new_y = y;
+
+  if (y < FOUR_SQRT_MIN) {
+    /*
+     * Avoid a possible underflow caused by y/A.  For casinh this
+     * would be legitimate, but will be picked up by invoking atan2
+     * later on.  For cacos this would not be legitimate.
+     */
+    *B_is_usable = 0;
+    *sqrt_A2my2 = A * (2 / DBL_EPSILON);
+    *new_y = y * (2 / DBL_EPSILON);
+    return;
+  }
+
+  /* B = (|z+I| - |z-I|) / 2 = y/A */
+  *B = y / A;
+  *B_is_usable = 1;
+
+  if (*B > B_crossover) {
+    *B_is_usable = 0;
+    /*
+     * Amy = fp + fm, where fp = f(x, y+1), and fm = f(x, y-1).
+     * sqrt_A2my2 = sqrt(Amy*(A+y))
+     */
+    if (y == 1 && x < DBL_EPSILON / 128) {
+      /*
+       * fp is of order x^2, and fm = x/2.
+       * A = 1 (inexactly).
+       */
+      *sqrt_A2my2 = sqrt(x) * sqrt((A + y) / 2);
+    } else if (x >= DBL_EPSILON * fabs(y - 1)) {
+      /*
+       * Underflow will not occur because
+       * x >= DBL_EPSILON/128 >= FOUR_SQRT_MIN
+       * and
+       * x >= DBL_EPSILON^2 >= FOUR_SQRT_MIN
+       */
+      Amy = f(x, y + 1, R) + f(x, y - 1, S);
+      *sqrt_A2my2 = sqrt(Amy * (A + y));
+    } else if (y > 1) {
+      /*
+       * fp = x*x/(y+1)/4, fm = x*x/(y-1)/4, and
+       * A = y (inexactly).
+       *
+       * y < RECIP_EPSILON.  So the following
+       * scaling should avoid any underflow problems.
+       */
+      *sqrt_A2my2 = x * (4 / DBL_EPSILON / DBL_EPSILON) * y /
+	sqrt((y + 1) * (y - 1));
+      *new_y = y * (4 / DBL_EPSILON / DBL_EPSILON);
+    } else {		/* if (y < 1) */
+      /*
+       * fm = 1-y >= DBL_EPSILON, fp is of order x^2, and
+       * A = 1 (inexactly).
+       */
+      *sqrt_A2my2 = sqrt((1 - y) * (1 + y));
+    }
+  }
+}
+
+/*
+ * casinh(z) = z + O(z^3)   as z -> 0
+ *
+ * casinh(z) = sign(x)*clog(sign(x)*z) + O(1/z^2)   as z -> infinity
+ * The above formula works for the imaginary part as well, because
+ * Im(casinh(z)) = sign(x)*atan2(sign(x)*y, fabs(x)) + O(y/z^3)
+ *    as z -> infinity, uniformly in y
+ */
+__host__ __device__ inline
+complex<double> casinh(complex<double> z)
+{
+  double x, y, ax, ay, rx, ry, B, sqrt_A2my2, new_y;
+  int B_is_usable;
+  complex<double> w;
+  const double RECIP_EPSILON = 1.0 / DBL_EPSILON;
+  const double m_ln2 = 6.9314718055994531e-1; /*  0x162e42fefa39ef.0p-53 */
+  x = z.real();
+  y = z.imag();
+  ax = fabs(x);
+  ay = fabs(y);
+
+  if (isnan(x) || isnan(y)) {
+    /* casinh(+-Inf + I*NaN) = +-Inf + I*NaN */
+    if (isinf(x))
+      return (complex<double>(x, y + y));
+    /* casinh(NaN + I*+-Inf) = opt(+-)Inf + I*NaN */
+    if (isinf(y))
+      return (complex<double>(y, x + x));
+    /* casinh(NaN + I*0) = NaN + I*0 */
+    if (y == 0)
+      return (complex<double>(x + x, y));
+    /*
+     * All other cases involving NaN return NaN + I*NaN.
+     * C99 leaves it optional whether to raise invalid if one of
+     * the arguments is not NaN, so we opt not to raise it.
+     */
+    return (complex<double>(x + 0.0 + (y + 0.0), x + 0.0 + (y + 0.0)));
+  }
+
+  if (ax > RECIP_EPSILON || ay > RECIP_EPSILON) {
+    /* clog...() will raise inexact unless x or y is infinite. */
+    if (signbit(x) == 0)
+      w = clog_for_large_values(z) + m_ln2;
+    else
+      w = clog_for_large_values(-z) + m_ln2;
+    return (complex<double>(copysign(w.real(), x), copysign(w.imag(), y)));
+  }
+
+  /* Avoid spuriously raising inexact for z = 0. */
+  if (x == 0 && y == 0)
+    return (z);
+
+  /* All remaining cases are inexact. */
+  raise_inexact();
+
+  const double SQRT_6_EPSILON = 3.6500241499888571e-8; /*  0x13988e1409212e.0p-77 */
+  if (ax < SQRT_6_EPSILON / 4 && ay < SQRT_6_EPSILON / 4)
+    return (z);
+
+  do_hard_work(ax, ay, &rx, &B_is_usable, &B, &sqrt_A2my2, &new_y);
+  if (B_is_usable)
+    ry = asin(B);
+  else
+    ry = atan2(new_y, sqrt_A2my2);
+  return (complex<double>(copysign(rx, x), copysign(ry, y)));
+}
+
+/*
+ * casin(z) = reverse(casinh(reverse(z)))
+ * where reverse(x + I*y) = y + I*x = I*conj(z).
+ */
+__host__ __device__ inline
+complex<double> casin(complex<double> z)
+{
+  complex<double> w = casinh(complex<double>(z.imag(), z.real()));
+
+  return (complex<double>(w.imag(), w.real()));
+}
+
+/*
+ * cacos(z) = PI/2 - casin(z)
+ * but do the computation carefully so cacos(z) is accurate when z is
+ * close to 1.
+ *
+ * cacos(z) = PI/2 - z + O(z^3)   as z -> 0
+ *
+ * cacos(z) = -sign(y)*I*clog(z) + O(1/z^2)   as z -> infinity
+ * The above formula works for the real part as well, because
+ * Re(cacos(z)) = atan2(fabs(y), x) + O(y/z^3)
+ *    as z -> infinity, uniformly in y
+ */
+__host__ __device__ inline
+complex<double> cacos(complex<double> z)
+{
+  double x, y, ax, ay, rx, ry, B, sqrt_A2mx2, new_x;
+  int sx, sy;
+  int B_is_usable;
+  complex<double> w;
+  const double pio2_hi = 1.5707963267948966e0; /*  0x1921fb54442d18.0p-52 */
+  const volatile double pio2_lo = 6.1232339957367659e-17;	/*  0x11a62633145c07.0p-106 */
+  const double m_ln2 = 6.9314718055994531e-1; /*  0x162e42fefa39ef.0p-53 */
+
+  x = z.real();
+  y = z.imag();
+  sx = signbit(x);
+  sy = signbit(y);
+  ax = fabs(x);
+  ay = fabs(y);
+
+  if (isnan(x) || isnan(y)) {
+    /* cacos(+-Inf + I*NaN) = NaN + I*opt(-)Inf */
+    if (isinf(x))
+      return (complex<double>(y + y, -infinity<double>()));
+    /* cacos(NaN + I*+-Inf) = NaN + I*-+Inf */
+    if (isinf(y))
+      return (complex<double>(x + x, -y));
+    /* cacos(0 + I*NaN) = PI/2 + I*NaN with inexact */
+    if (x == 0)
+      return (complex<double>(pio2_hi + pio2_lo, y + y));
+    /*
+     * All other cases involving NaN return NaN + I*NaN.
+     * C99 leaves it optional whether to raise invalid if one of
+     * the arguments is not NaN, so we opt not to raise it.
+     */
+    return (complex<double>(x + 0.0 + (y + 0), x + 0.0 + (y + 0)));
+  }
+
+  const double RECIP_EPSILON = 1.0 / DBL_EPSILON;
+  if (ax > RECIP_EPSILON || ay > RECIP_EPSILON) {
+    /* clog...() will raise inexact unless x or y is infinite. */
+    w = clog_for_large_values(z);
+    rx = fabs(w.imag());
+    ry = w.real() + m_ln2;
+    if (sy == 0)
+      ry = -ry;
+    return (complex<double>(rx, ry));
+  }
+
+  /* Avoid spuriously raising inexact for z = 1. */
+  if (x == 1.0 && y == 0.0)
+    return (complex<double>(0, -y));
+
+  /* All remaining cases are inexact. */
+  raise_inexact();
+
+  const double SQRT_6_EPSILON = 3.6500241499888571e-8; /*  0x13988e1409212e.0p-77 */
+  if (ax < SQRT_6_EPSILON / 4 && ay < SQRT_6_EPSILON / 4)
+    return (complex<double>(pio2_hi - (x - pio2_lo), -y));
+
+  do_hard_work(ay, ax, &ry, &B_is_usable, &B, &sqrt_A2mx2, &new_x);
+  if (B_is_usable) {
+    if (sx == 0)
+      rx = acos(B);
+    else
+      rx = acos(-B);
+  } else {
+    if (sx == 0)
+      rx = atan2(sqrt_A2mx2, new_x);
+    else
+      rx = atan2(sqrt_A2mx2, -new_x);
+  }
+  if (sy == 0)
+    ry = -ry;
+  return (complex<double>(rx, ry));
+}
+
+/*
+ * cacosh(z) = I*cacos(z) or -I*cacos(z)
+ * where the sign is chosen so Re(cacosh(z)) >= 0.
+ */
+__host__ __device__ inline
+complex<double> cacosh(complex<double> z)
+{
+  complex<double> w;
+  double rx, ry;
+
+  w = cacos(z);
+  rx = w.real();
+  ry = w.imag();
+  /* cacosh(NaN + I*NaN) = NaN + I*NaN */
+  if (isnan(rx) && isnan(ry))
+    return (complex<double>(ry, rx));
+  /* cacosh(NaN + I*+-Inf) = +Inf + I*NaN */
+  /* cacosh(+-Inf + I*NaN) = +Inf + I*NaN */
+  if (isnan(rx))
+    return (complex<double>(fabs(ry), rx));
+  /* cacosh(0 + I*NaN) = NaN + I*NaN */
+  if (isnan(ry))
+    return (complex<double>(ry, ry));
+  return (complex<double>(fabs(ry), copysign(rx, z.imag())));
+}
+
+/*
+ * Optimized version of clog() for |z| finite and larger than ~RECIP_EPSILON.
+ */
+__host__ __device__ inline
+complex<double> clog_for_large_values(complex<double> z)
+{
+  double x, y;
+  double ax, ay, t;
+  const double m_e = 2.7182818284590452e0; /*  0x15bf0a8b145769.0p-51 */
+
+  x = z.real();
+  y = z.imag();
+  ax = fabs(x);
+  ay = fabs(y);
+  if (ax < ay) {
+    t = ax;
+    ax = ay;
+    ay = t;
+  }
+
+  /*
+   * Avoid overflow in hypot() when x and y are both very large.
+   * Divide x and y by E, and then add 1 to the logarithm.  This depends
+   * on E being larger than sqrt(2).
+   * Dividing by E causes an insignificant loss of accuracy; however
+   * this method is still poor since it is uneccessarily slow.
+   */
+  if (ax > DBL_MAX / 2)
+    return (complex<double>(log(hypot(x / m_e, y / m_e)) + 1, atan2(y, x)));
+
+  /*
+   * Avoid overflow when x or y is large.  Avoid underflow when x or
+   * y is small.
+   */
+  const double QUARTER_SQRT_MAX = 5.966672584960165394632772e-154; /* = 0x1p509; <= sqrt(DBL_MAX) / 4 */
+  const double SQRT_MIN =	1.491668146240041348658193e-154; /* = 0x1p-511; >= sqrt(DBL_MIN) */
+  if (ax > QUARTER_SQRT_MAX || ay < SQRT_MIN)
+    return (complex<double>(log(hypot(x, y)), atan2(y, x)));
+
+  return (complex<double>(log(ax * ax + ay * ay) / 2, atan2(y, x)));
+}
+
+/*
+ *				=================
+ *				| catanh, catan |
+ *				=================
+ */
+
+/*
+   * sum_squares(x,y) = x*x + y*y (or just x*x if y*y would underflow).
+   * Assumes x*x and y*y will not overflow.
+   * Assumes x and y are finite.
+   * Assumes y is non-negative.
+   * Assumes fabs(x) >= DBL_EPSILON.
+   */
+__host__ __device__
+inline double sum_squares(double x, double y)
+{
+  const double SQRT_MIN =	1.491668146240041348658193e-154; /* = 0x1p-511; >= sqrt(DBL_MIN) */
+  /* Avoid underflow when y is small. */
+  if (y < SQRT_MIN)
+    return (x * x);
+
+  return (x * x + y * y);
+}
+
+/*
+ * real_part_reciprocal(x, y) = Re(1/(x+I*y)) = x/(x*x + y*y).
+ * Assumes x and y are not NaN, and one of x and y is larger than
+ * RECIP_EPSILON.  We avoid unwarranted underflow.  It is important to not use
+ * the code creal(1/z), because the imaginary part may produce an unwanted
+ * underflow.
+ * This is only called in a context where inexact is always raised before
+ * the call, so no effort is made to avoid or force inexact.
+ */
+__host__ __device__
+inline double real_part_reciprocal(double x, double y)
+{
+  double scale;
+  uint32_t hx, hy;
+  int32_t ix, iy;
+
+  /*
+   * This code is inspired by the C99 document n1124.pdf, Section G.5.1,
+   * example 2.
+   */
+  get_high_word(hx, x);
+  ix = hx & 0x7ff00000;
+  get_high_word(hy, y);
+  iy = hy & 0x7ff00000;
+  //#define	BIAS	(DBL_MAX_EXP - 1)
+  const int BIAS = DBL_MAX_EXP - 1;
+  /* XXX more guard digits are useful iff there is extra precision. */
+  //#define	CUTOFF	(DBL_MANT_DIG / 2 + 1)	/* just half or 1 guard digit */
+  const int CUTOFF = (DBL_MANT_DIG / 2 + 1);
+  if (ix - iy >= CUTOFF << 20 || isinf(x))
+    return (1 / x);		/* +-Inf -> +-0 is special */
+  if (iy - ix >= CUTOFF << 20)
+    return (x / y / y);	/* should avoid double div, but hard */
+  if (ix <= (BIAS + DBL_MAX_EXP / 2 - CUTOFF) << 20)
+    return (x / (x * x + y * y));
+  scale = 1;
+  set_high_word(scale, 0x7ff00000 - ix);	/* 2**(1-ilogb(x)) */
+  x *= scale;
+  y *= scale;
+  return (x / (x * x + y * y) * scale);
+}
+
+
+/*
+ * catanh(z) = log((1+z)/(1-z)) / 2
+ *           = log1p(4*x / |z-1|^2) / 4
+ *             + I * atan2(2*y, (1-x)*(1+x)-y*y) / 2
+ *
+ * catanh(z) = z + O(z^3)   as z -> 0
+ *
+ * catanh(z) = 1/z + sign(y)*I*PI/2 + O(1/z^3)   as z -> infinity
+ * The above formula works for the real part as well, because
+ * Re(catanh(z)) = x/|z|^2 + O(x/z^4)
+ *    as z -> infinity, uniformly in x
+ */
+#if THRUST_CPP_DIALECT >= 2011 || THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+__host__ __device__ inline
+complex<double> catanh(complex<double> z)
+{
+  double x, y, ax, ay, rx, ry;
+  const volatile double pio2_lo = 6.1232339957367659e-17; /*  0x11a62633145c07.0p-106 */
+  const double pio2_hi = 1.5707963267948966e0;/*  0x1921fb54442d18.0p-52 */
+
+
+  x = z.real();
+  y = z.imag();
+  ax = fabs(x);
+  ay = fabs(y);
+
+  /* This helps handle many cases. */
+  if (y == 0 && ax <= 1)
+    return (complex<double>(atanh(x), y));
+
+  /* To ensure the same accuracy as atan(), and to filter out z = 0. */
+  if (x == 0)
+    return (complex<double>(x, atan(y)));
+
+  if (isnan(x) || isnan(y)) {
+    /* catanh(+-Inf + I*NaN) = +-0 + I*NaN */
+    if (isinf(x))
+      return (complex<double>(copysign(0.0, x), y + y));
+    /* catanh(NaN + I*+-Inf) = sign(NaN)0 + I*+-PI/2 */
+    if (isinf(y))
+      return (complex<double>(copysign(0.0, x),
+			      copysign(pio2_hi + pio2_lo, y)));
+    /*
+     * All other cases involving NaN return NaN + I*NaN.
+     * C99 leaves it optional whether to raise invalid if one of
+     * the arguments is not NaN, so we opt not to raise it.
+     */
+    return (complex<double>(x + 0.0 + (y + 0), x + 0.0 + (y + 0)));
+  }
+
+  const double RECIP_EPSILON = 1.0 / DBL_EPSILON;
+  if (ax > RECIP_EPSILON || ay > RECIP_EPSILON)
+    return (complex<double>(real_part_reciprocal(x, y),
+			    copysign(pio2_hi + pio2_lo, y)));
+
+  const double SQRT_3_EPSILON = 2.5809568279517849e-8; /*  0x1bb67ae8584caa.0p-78 */
+  if (ax < SQRT_3_EPSILON / 2 && ay < SQRT_3_EPSILON / 2) {
+    /*
+     * z = 0 was filtered out above.  All other cases must raise
+     * inexact, but this is the only only that needs to do it
+     * explicitly.
+     */
+    raise_inexact();
+    return (z);
+  }
+
+  const double m_ln2 = 6.9314718055994531e-1; /*  0x162e42fefa39ef.0p-53 */
+  if (ax == 1 && ay < DBL_EPSILON)
+    rx = (m_ln2 - log(ay)) / 2;
+  else
+    rx = log1p(4 * ax / sum_squares(ax - 1, ay)) / 4;
+
+  if (ax == 1)
+    ry = atan2(2.0, -ay) / 2;
+  else if (ay < DBL_EPSILON)
+    ry = atan2(2 * ay, (1 - ax) * (1 + ax)) / 2;
+  else
+    ry = atan2(2 * ay, (1 - ax) * (1 + ax) - ay * ay) / 2;
+
+  return (complex<double>(copysign(rx, x), copysign(ry, y)));
+}
+
+/*
+ * catan(z) = reverse(catanh(reverse(z)))
+ * where reverse(x + I*y) = y + I*x = I*conj(z).
+ */
+__host__ __device__ inline
+complex<double>catan(complex<double> z)
+{
+  complex<double> w = catanh(complex<double>(z.imag(), z.real()));
+  return (complex<double>(w.imag(), w.real()));
+}
+
+#endif
+
+} // namespace complex
+
+} // namespace detail
+
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> acos(const complex<ValueType>& z){
+  const complex<ValueType> ret = thrust::asin(z);
+  const ValueType pi = ValueType(3.14159265358979323846);
+  return complex<ValueType>(pi/2 - ret.real(),-ret.imag());
+}
+
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> asin(const complex<ValueType>& z){
+  const complex<ValueType> i(0,1);
+  return -i*asinh(i*z);
+}
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> atan(const complex<ValueType>& z){
+  const complex<ValueType> i(0,1);
+  return -i*thrust::atanh(i*z);
+}
+
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> acosh(const complex<ValueType>& z){
+  thrust::complex<ValueType> ret((z.real() - z.imag()) * (z.real() + z.imag()) - ValueType(1.0),
+				 ValueType(2.0) * z.real() * z.imag());
+  ret = thrust::sqrt(ret);
+  if (z.real() < ValueType(0.0)){
+    ret = -ret;
+  }
+  ret += z;
+  ret = thrust::log(ret);
+  if (ret.real() < ValueType(0.0)){
+    ret = -ret;
+  }
+  return ret;
+}
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> asinh(const complex<ValueType>& z){
+  return thrust::log(thrust::sqrt(z*z+ValueType(1))+z);
+}
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> atanh(const complex<ValueType>& z){
+  ValueType imag2 = z.imag() *  z.imag();
+  ValueType n = ValueType(1.0) + z.real();
+  n = imag2 + n * n;
+
+  ValueType d = ValueType(1.0) - z.real();
+  d = imag2 + d * d;
+  complex<ValueType> ret(ValueType(0.25) * (std::log(n) - std::log(d)),0);
+
+  d = ValueType(1.0) -  z.real() * z.real() - imag2;
+
+  ret.imag(ValueType(0.5) * std::atan2(ValueType(2.0) * z.imag(), d));
+  return ret;
+}
+
+template <>
+__host__ __device__
+inline complex<double> acos(const complex<double>& z){
+  return detail::complex::cacos(z);
+}
+
+template <>
+__host__ __device__
+inline complex<double> asin(const complex<double>& z){
+  return detail::complex::casin(z);
+}
+
+#if THRUST_CPP_DIALECT >= 2011 || THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+template <>
+__host__ __device__
+inline complex<double> atan(const complex<double>& z){
+  return detail::complex::catan(z);
+}
+#endif
+
+template <>
+__host__ __device__
+inline complex<double> acosh(const complex<double>& z){
+  return detail::complex::cacosh(z);
+}
+
+
+template <>
+__host__ __device__
+inline complex<double> asinh(const complex<double>& z){
+  return detail::complex::casinh(z);
+}
+
+#if THRUST_CPP_DIALECT >= 2011 || THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+template <>
+__host__ __device__
+inline complex<double> atanh(const complex<double>& z){
+  return detail::complex::catanh(z);
+}
+#endif
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrigf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrigf.h
new file mode 100644
index 000000000000..c06791311a82
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/catrigf.h
@@ -0,0 +1,500 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2012 Stephen Montgomery-Smith <stephen@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/*
+ * Adapted from FreeBSD by Filipe Maia <filipe.c.maia@gmail.com>:
+ *    freebsd/lib/msun/src/catrig.c
+ */
+
+#pragma once
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <thrust/detail/config.h>
+#include <cfloat>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+  
+__host__ __device__ inline
+      complex<float> clog_for_large_values(complex<float> z);
+
+/*
+ * The algorithm is very close to that in "Implementing the complex arcsine
+ * and arccosine functions using exception handling" by T. E. Hull, Thomas F.
+ * Fairgrieve, and Ping Tak Peter Tang, published in ACM Transactions on
+ * Mathematical Software, Volume 23 Issue 3, 1997, Pages 299-335,
+ * http://dl.acm.org/citation.cfm?id=275324.
+ *
+ * See catrig.c for complete comments.
+ *
+ * XXX comments were removed automatically, and even short ones on the right
+ * of statements were removed (all of them), contrary to normal style.  Only
+ * a few comments on the right of declarations remain.
+ */
+
+__host__ __device__
+inline float
+f(float a, float b, float hypot_a_b)
+{
+  if (b < 0.0f)
+    return ((hypot_a_b - b) / 2.0f);
+  if (b == 0.0f)
+    return (a / 2.0f);
+  return (a * a / (hypot_a_b + b) / 2.0f);
+}
+
+/*
+ * All the hard work is contained in this function.
+ * x and y are assumed positive or zero, and less than RECIP_EPSILON.
+ * Upon return:
+ * rx = Re(casinh(z)) = -Im(cacos(y + I*x)).
+ * B_is_usable is set to 1 if the value of B is usable.
+ * If B_is_usable is set to 0, sqrt_A2my2 = sqrt(A*A - y*y), and new_y = y.
+ * If returning sqrt_A2my2 has potential to result in an underflow, it is
+ * rescaled, and new_y is similarly rescaled.
+ */
+__host__ __device__ 
+inline void
+do_hard_work(float x, float y, float *rx, int *B_is_usable, float *B,
+	     float *sqrt_A2my2, float *new_y)
+{
+  float R, S, A; /* A, B, R, and S are as in Hull et al. */
+  float Am1, Amy; /* A-1, A-y. */
+  const float A_crossover = 10; /* Hull et al suggest 1.5, but 10 works better */
+  const float FOUR_SQRT_MIN = 4.336808689942017736029811e-19f;; /* =0x1p-61; >= 4 * sqrt(FLT_MIN) */
+  const float B_crossover = 0.6417f; /* suggested by Hull et al */
+  R = hypotf(x, y + 1);
+  S = hypotf(x, y - 1);
+
+  A = (R + S) / 2;
+  if (A < 1)
+    A = 1;
+
+  if (A < A_crossover) {
+    if (y == 1 && x < FLT_EPSILON * FLT_EPSILON / 128) {
+      *rx = sqrtf(x);
+    } else if (x >= FLT_EPSILON * fabsf(y - 1)) {
+      Am1 = f(x, 1 + y, R) + f(x, 1 - y, S);
+      *rx = log1pf(Am1 + sqrtf(Am1 * (A + 1)));
+    } else if (y < 1) {
+      *rx = x / sqrtf((1 - y) * (1 + y));
+    } else {
+      *rx = log1pf((y - 1) + sqrtf((y - 1) * (y + 1)));
+    }
+  } else {
+    *rx = logf(A + sqrtf(A * A - 1));
+  }
+
+  *new_y = y;
+
+  if (y < FOUR_SQRT_MIN) {
+    *B_is_usable = 0;
+    *sqrt_A2my2 = A * (2 / FLT_EPSILON);
+    *new_y = y * (2 / FLT_EPSILON);
+    return;
+  }
+
+  *B = y / A;
+  *B_is_usable = 1;
+
+  if (*B > B_crossover) {
+    *B_is_usable = 0;
+    if (y == 1 && x < FLT_EPSILON / 128) {
+      *sqrt_A2my2 = sqrtf(x) * sqrtf((A + y) / 2);
+    } else if (x >= FLT_EPSILON * fabsf(y - 1)) {
+      Amy = f(x, y + 1, R) + f(x, y - 1, S);
+      *sqrt_A2my2 = sqrtf(Amy * (A + y));
+    } else if (y > 1) {
+      *sqrt_A2my2 = x * (4 / FLT_EPSILON / FLT_EPSILON) * y /
+	sqrtf((y + 1) * (y - 1));
+      *new_y = y * (4 / FLT_EPSILON / FLT_EPSILON);
+    } else {
+      *sqrt_A2my2 = sqrtf((1 - y) * (1 + y));
+    }
+  }
+
+}
+
+__host__ __device__ inline
+complex<float>
+casinhf(complex<float> z)
+{
+  float x, y, ax, ay, rx, ry, B, sqrt_A2my2, new_y;
+  int B_is_usable;
+  complex<float> w;
+  const float RECIP_EPSILON = 1.0f / FLT_EPSILON;
+  const float m_ln2 = 6.9314718055994531e-1f; /*  0x162e42fefa39ef.0p-53 */
+  x = z.real();
+  y = z.imag();
+  ax = fabsf(x);
+  ay = fabsf(y);
+
+  if (isnan(x) || isnan(y)) {
+    if (isinf(x))
+      return (complex<float>(x, y + y));
+    if (isinf(y))
+      return (complex<float>(y, x + x));
+    if (y == 0)
+      return (complex<float>(x + x, y));
+    return (complex<float>(x + 0.0f + (y + 0), x + 0.0f + (y + 0)));
+  }
+
+  if (ax > RECIP_EPSILON || ay > RECIP_EPSILON) {
+    if (signbit(x) == 0)
+      w = clog_for_large_values(z) + m_ln2;
+    else
+      w = clog_for_large_values(-z) + m_ln2;
+    return (complex<float>(copysignf(w.real(), x),
+			   copysignf(w.imag(), y)));
+  }
+
+  if (x == 0 && y == 0)
+    return (z);
+
+  raise_inexact();
+
+  const float SQRT_6_EPSILON = 8.4572793338e-4f;	/*  0xddb3d7.0p-34 */
+  if (ax < SQRT_6_EPSILON / 4 && ay < SQRT_6_EPSILON / 4)
+    return (z);
+
+  do_hard_work(ax, ay, &rx, &B_is_usable, &B, &sqrt_A2my2, &new_y);
+  if (B_is_usable)
+    ry = asinf(B);
+  else
+    ry = atan2f(new_y, sqrt_A2my2);
+  return (complex<float>(copysignf(rx, x), copysignf(ry, y)));
+}
+
+__host__ __device__ inline
+complex<float> casinf(complex<float> z)
+{
+  complex<float> w = casinhf(complex<float>(z.imag(), z.real()));
+
+  return (complex<float>(w.imag(), w.real()));
+}
+
+__host__ __device__ inline
+complex<float> cacosf(complex<float> z)
+{
+  float x, y, ax, ay, rx, ry, B, sqrt_A2mx2, new_x;
+  int sx, sy;
+  int B_is_usable;
+  complex<float> w;
+  const float pio2_hi = 1.5707963267948966e0f; /*  0x1921fb54442d18.0p-52 */
+  const volatile float pio2_lo = 6.1232339957367659e-17f;	/*  0x11a62633145c07.0p-106 */
+  const float m_ln2 = 6.9314718055994531e-1f; /*  0x162e42fefa39ef.0p-53 */
+
+  x = z.real();
+  y = z.imag();
+  sx = signbit(x);
+  sy = signbit(y);
+  ax = fabsf(x);
+  ay = fabsf(y);
+
+  if (isnan(x) || isnan(y)) {
+    if (isinf(x))
+      return (complex<float>(y + y, -infinity<float>()));
+    if (isinf(y))
+      return (complex<float>(x + x, -y));
+    if (x == 0)
+      return (complex<float>(pio2_hi + pio2_lo, y + y));
+    return (complex<float>(x + 0.0f + (y + 0), x + 0.0f + (y + 0)));
+  }
+
+  const float RECIP_EPSILON = 1.0f / FLT_EPSILON;
+  if (ax > RECIP_EPSILON || ay > RECIP_EPSILON) {
+    w = clog_for_large_values(z);
+    rx = fabsf(w.imag());
+    ry = w.real() + m_ln2;
+    if (sy == 0)
+      ry = -ry;
+    return (complex<float>(rx, ry));
+  }
+
+  if (x == 1 && y == 0)
+    return (complex<float>(0, -y));
+
+  raise_inexact();
+
+  const float SQRT_6_EPSILON = 8.4572793338e-4f;	/*  0xddb3d7.0p-34 */
+  if (ax < SQRT_6_EPSILON / 4 && ay < SQRT_6_EPSILON / 4)
+    return (complex<float>(pio2_hi - (x - pio2_lo), -y));
+
+  do_hard_work(ay, ax, &ry, &B_is_usable, &B, &sqrt_A2mx2, &new_x);
+  if (B_is_usable) {
+    if (sx == 0)
+      rx = acosf(B);
+    else
+      rx = acosf(-B);
+  } else {
+    if (sx == 0)
+      rx = atan2f(sqrt_A2mx2, new_x);
+    else
+      rx = atan2f(sqrt_A2mx2, -new_x);
+  }
+  if (sy == 0)
+    ry = -ry;
+  return (complex<float>(rx, ry));
+}
+
+__host__ __device__ inline
+complex<float> cacoshf(complex<float> z)
+{
+  complex<float> w;
+  float rx, ry;
+
+  w = cacosf(z);
+  rx = w.real();
+  ry = w.imag();
+  /* cacosh(NaN + I*NaN) = NaN + I*NaN */
+  if (isnan(rx) && isnan(ry))
+    return (complex<float>(ry, rx));
+  /* cacosh(NaN + I*+-Inf) = +Inf + I*NaN */
+  /* cacosh(+-Inf + I*NaN) = +Inf + I*NaN */
+  if (isnan(rx))
+    return (complex<float>(fabsf(ry), rx));
+  /* cacosh(0 + I*NaN) = NaN + I*NaN */
+  if (isnan(ry))
+    return (complex<float>(ry, ry));
+  return (complex<float>(fabsf(ry), copysignf(rx, z.imag())));
+}
+
+  /*
+   * Optimized version of clog() for |z| finite and larger than ~RECIP_EPSILON.
+   */
+__host__ __device__ inline
+complex<float> clog_for_large_values(complex<float> z)
+{
+  float x, y;
+  float ax, ay, t;
+  const float m_e = 2.7182818284590452e0f; /*  0x15bf0a8b145769.0p-51 */
+
+  x = z.real();
+  y = z.imag();
+  ax = fabsf(x);
+  ay = fabsf(y);
+  if (ax < ay) {
+    t = ax;
+    ax = ay;
+    ay = t;
+  }
+
+  if (ax > FLT_MAX / 2)
+    return (complex<float>(logf(hypotf(x / m_e, y / m_e)) + 1,
+			   atan2f(y, x)));
+
+  const float QUARTER_SQRT_MAX = 2.3058430092136939520000000e+18f; /* = 0x1p61; <= sqrt(FLT_MAX) / 4 */
+  const float SQRT_MIN =	1.084202172485504434007453e-19f; /* 0x1p-63; >= sqrt(FLT_MIN) */
+  if (ax > QUARTER_SQRT_MAX || ay < SQRT_MIN)
+    return (complex<float>(logf(hypotf(x, y)), atan2f(y, x)));
+
+  return (complex<float>(logf(ax * ax + ay * ay) / 2, atan2f(y, x)));
+}
+
+/*
+ *				=================
+ *				| catanh, catan |
+ *				=================
+ */
+
+/*
+ * sum_squares(x,y) = x*x + y*y (or just x*x if y*y would underflow).
+ * Assumes x*x and y*y will not overflow.
+ * Assumes x and y are finite.
+ * Assumes y is non-negative.
+ * Assumes fabsf(x) >= FLT_EPSILON.
+ */
+__host__ __device__
+inline float sum_squares(float x, float y)
+{
+  const float SQRT_MIN =	1.084202172485504434007453e-19f; /* 0x1p-63; >= sqrt(FLT_MIN) */
+  /* Avoid underflow when y is small. */
+  if (y < SQRT_MIN)
+    return (x * x);
+
+  return (x * x + y * y);
+}
+
+__host__ __device__
+inline float real_part_reciprocal(float x, float y)
+{
+  float scale;
+  uint32_t hx, hy;
+  int32_t ix, iy;
+
+  get_float_word(hx, x);
+  ix = hx & 0x7f800000;
+  get_float_word(hy, y);
+  iy = hy & 0x7f800000;
+  //#define	BIAS	(FLT_MAX_EXP - 1)
+  const int BIAS = FLT_MAX_EXP - 1;
+  //#define	CUTOFF	(FLT_MANT_DIG / 2 + 1)
+  const int CUTOFF = (FLT_MANT_DIG / 2 + 1);
+  if (ix - iy >= CUTOFF << 23 || isinf(x))
+    return (1 / x);
+  if (iy - ix >= CUTOFF << 23)
+    return (x / y / y);
+  if (ix <= (BIAS + FLT_MAX_EXP / 2 - CUTOFF) << 23)
+    return (x / (x * x + y * y));
+  set_float_word(scale, 0x7f800000 - ix);
+  x *= scale;
+  y *= scale;
+  return (x / (x * x + y * y) * scale);
+}
+
+#if THRUST_CPP_DIALECT >= 2011 || THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+__host__ __device__ inline
+complex<float> catanhf(complex<float> z)
+{
+  float x, y, ax, ay, rx, ry;
+  const volatile float pio2_lo = 6.1232339957367659e-17f; /*  0x11a62633145c07.0p-106 */
+  const float pio2_hi = 1.5707963267948966e0f;/*  0x1921fb54442d18.0p-52 */
+
+
+  x = z.real();
+  y = z.imag();
+  ax = fabsf(x);
+  ay = fabsf(y);
+
+
+  if (y == 0 && ax <= 1)
+    return (complex<float>(atanhf(x), y));
+
+  if (x == 0)
+    return (complex<float>(x, atanf(y)));
+
+  if (isnan(x) || isnan(y)) {
+    if (isinf(x))
+      return (complex<float>(copysignf(0, x), y + y));
+    if (isinf(y))
+      return (complex<float>(copysignf(0, x),
+			     copysignf(pio2_hi + pio2_lo, y)));
+    return (complex<float>(x + 0.0f + (y + 0.0f), x + 0.0f + (y + 0.0f)));
+  }
+
+  const float RECIP_EPSILON = 1.0f / FLT_EPSILON;
+  if (ax > RECIP_EPSILON || ay > RECIP_EPSILON)
+    return (complex<float>(real_part_reciprocal(x, y),
+			   copysignf(pio2_hi + pio2_lo, y)));
+
+  const float SQRT_3_EPSILON = 5.9801995673e-4f; /*  0x9cc471.0p-34 */
+  if (ax < SQRT_3_EPSILON / 2 && ay < SQRT_3_EPSILON / 2) {
+    raise_inexact();
+    return (z);
+  }
+
+  const float m_ln2 = 6.9314718056e-1f; /*  0xb17218.0p-24 */
+  if (ax == 1 && ay < FLT_EPSILON)
+    rx = (m_ln2 - logf(ay)) / 2;
+  else
+    rx = log1pf(4 * ax / sum_squares(ax - 1, ay)) / 4;
+
+  if (ax == 1)
+    ry = atan2f(2, -ay) / 2;
+  else if (ay < FLT_EPSILON)
+    ry = atan2f(2 * ay, (1 - ax) * (1 + ax)) / 2;
+  else
+    ry = atan2f(2 * ay, (1 - ax) * (1 + ax) - ay * ay) / 2;
+
+  return (complex<float>(copysignf(rx, x), copysignf(ry, y)));
+}
+
+__host__ __device__ inline
+complex<float>catanf(complex<float> z){
+  complex<float> w = catanhf(complex<float>(z.imag(), z.real()));
+  return (complex<float>(w.imag(), w.real()));
+}
+#endif
+
+} // namespace complex
+
+} // namespace detail
+
+
+template <>
+__host__ __device__
+inline complex<float> acos(const complex<float>& z){
+  return detail::complex::cacosf(z);
+}
+
+template <>
+__host__ __device__
+inline complex<float> asin(const complex<float>& z){
+  return detail::complex::casinf(z);
+}
+
+#if THRUST_CPP_DIALECT >= 2011 || THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+template <>
+__host__ __device__
+inline complex<float> atan(const complex<float>& z){
+  return detail::complex::catanf(z);
+}
+#endif
+
+template <>
+__host__ __device__
+inline complex<float> acosh(const complex<float>& z){
+  return detail::complex::cacoshf(z);
+}
+
+
+template <>
+__host__ __device__
+inline complex<float> asinh(const complex<float>& z){
+  return detail::complex::casinhf(z);
+}
+
+#if THRUST_CPP_DIALECT >= 2011 || THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+template <>
+__host__ __device__
+inline complex<float> atanh(const complex<float>& z){
+  return detail::complex::catanhf(z);
+}
+#endif
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccosh.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccosh.h
new file mode 100644
index 000000000000..722dfcd843de
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccosh.h
@@ -0,0 +1,215 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/s_ccosh.c
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+/*
+ * Hyperbolic cosine of a complex argument z = x + i y.
+ *
+ * cosh(z) = cosh(x+iy)
+ *         = cosh(x) cos(y) + i sinh(x) sin(y).
+ *
+ * Exceptional values are noted in the comments within the source code.
+ * These values and the return value were taken from n1124.pdf.
+ */
+      
+__host__ __device__ inline
+thrust::complex<double> ccosh(const thrust::complex<double>& z){
+  
+
+  const double huge = 8.98846567431157953864652595395e+307; // 0x1p1023
+  double x, y, h;
+  uint32_t hx, hy, ix, iy, lx, ly;
+
+  x = z.real();
+  y = z.imag();
+
+  extract_words(hx, lx, x);
+  extract_words(hy, ly, y);
+
+  ix = 0x7fffffff & hx;
+  iy = 0x7fffffff & hy;
+
+  /* Handle the nearly-non-exceptional cases where x and y are finite. */
+  if (ix < 0x7ff00000 && iy < 0x7ff00000) {
+    if ((iy | ly) == 0)
+      return (thrust::complex<double>(::cosh(x), x * y));
+    if (ix < 0x40360000)	/* small x: normal case */
+      return (thrust::complex<double>(::cosh(x) * ::cos(y), ::sinh(x) * ::sin(y)));
+
+    /* |x| >= 22, so cosh(x) ~= exp(|x|) */
+    if (ix < 0x40862e42) {
+      /* x < 710: exp(|x|) won't overflow */
+      h = ::exp(::fabs(x)) * 0.5;
+      return (thrust::complex<double>(h * cos(y), copysign(h, x) * sin(y)));
+    } else if (ix < 0x4096bbaa) {
+      /* x < 1455: scale to avoid overflow */
+      thrust::complex<double> z_;
+      z_ = ldexp_cexp(thrust::complex<double>(fabs(x), y), -1);
+      return (thrust::complex<double>(z_.real(), z_.imag() * copysign(1.0, x)));
+    } else {
+      /* x >= 1455: the result always overflows */
+      h = huge * x;
+      return (thrust::complex<double>(h * h * cos(y), h * sin(y)));
+    }
+  }
+
+  /*
+   * cosh(+-0 +- I Inf) = dNaN + I sign(d(+-0, dNaN))0.
+   * The sign of 0 in the result is unspecified.  Choice = normally
+   * the same as dNaN.  Raise the invalid floating-point exception.
+   *
+   * cosh(+-0 +- I NaN) = d(NaN) + I sign(d(+-0, NaN))0.
+   * The sign of 0 in the result is unspecified.  Choice = normally
+   * the same as d(NaN).
+   */
+  if ((ix | lx) == 0 && iy >= 0x7ff00000)
+    return (thrust::complex<double>(y - y, copysign(0.0, x * (y - y))));
+
+  /*
+   * cosh(+-Inf +- I 0) = +Inf + I (+-)(+-)0.
+   *
+   * cosh(NaN +- I 0)   = d(NaN) + I sign(d(NaN, +-0))0.
+   * The sign of 0 in the result is unspecified.
+   */
+  if ((iy | ly) == 0 && ix >= 0x7ff00000) {
+    if (((hx & 0xfffff) | lx) == 0)
+      return (thrust::complex<double>(x * x, copysign(0.0, x) * y));
+    return (thrust::complex<double>(x * x, copysign(0.0, (x + x) * y)));
+  }
+
+  /*
+   * cosh(x +- I Inf) = dNaN + I dNaN.
+   * Raise the invalid floating-point exception for finite nonzero x.
+   *
+   * cosh(x + I NaN) = d(NaN) + I d(NaN).
+   * Optionally raises the invalid floating-point exception for finite
+   * nonzero x.  Choice = don't raise (except for signaling NaNs).
+   */
+  if (ix < 0x7ff00000 && iy >= 0x7ff00000)
+    return (thrust::complex<double>(y - y, x * (y - y)));
+
+  /*
+   * cosh(+-Inf + I NaN)  = +Inf + I d(NaN).
+   *
+   * cosh(+-Inf +- I Inf) = +Inf + I dNaN.
+   * The sign of Inf in the result is unspecified.  Choice = always +.
+   * Raise the invalid floating-point exception.
+   *
+   * cosh(+-Inf + I y)   = +Inf cos(y) +- I Inf sin(y)
+   */
+  if (ix >= 0x7ff00000 && ((hx & 0xfffff) | lx) == 0) {
+    if (iy >= 0x7ff00000)
+      return (thrust::complex<double>(x * x, x * (y - y)));
+    return (thrust::complex<double>((x * x) * cos(y), x * sin(y)));
+  }
+
+  /*
+   * cosh(NaN + I NaN)  = d(NaN) + I d(NaN).
+   *
+   * cosh(NaN +- I Inf) = d(NaN) + I d(NaN).
+   * Optionally raises the invalid floating-point exception.
+   * Choice = raise.
+   *
+   * cosh(NaN + I y)    = d(NaN) + I d(NaN).
+   * Optionally raises the invalid floating-point exception for finite
+   * nonzero y.  Choice = don't raise (except for signaling NaNs).
+   */
+  return (thrust::complex<double>((x * x) * (y - y), (x + x) * (y - y)));
+}
+
+
+__host__ __device__ inline
+thrust::complex<double> ccos(const thrust::complex<double>& z){	
+  /* ccos(z) = ccosh(I * z) */
+  return (ccosh(thrust::complex<double>(-z.imag(), z.real())));
+}
+
+} // namespace complex
+
+} // namespace detail
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> cos(const complex<ValueType>& z){
+  const ValueType re = z.real();
+  const ValueType im = z.imag();
+  return complex<ValueType>(std::cos(re) * std::cosh(im), 
+			    -std::sin(re) * std::sinh(im));
+}
+  
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> cosh(const complex<ValueType>& z){
+  const ValueType re = z.real();
+  const ValueType im = z.imag();
+  return complex<ValueType>(std::cosh(re) * std::cos(im), 
+			    std::sinh(re) * std::sin(im));
+}
+
+template <>
+__host__ __device__
+inline thrust::complex<double> cos(const thrust::complex<double>& z){
+  return detail::complex::ccos(z);
+}
+
+template <>
+__host__ __device__
+inline thrust::complex<double> cosh(const thrust::complex<double>& z){
+  return detail::complex::ccosh(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccoshf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccoshf.h
new file mode 100644
index 000000000000..aa43f120826a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ccoshf.h
@@ -0,0 +1,143 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/s_ccoshf.c
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+      
+__host__ __device__ inline
+complex<float> ccoshf(const complex<float>& z){
+  float x, y, h;
+  uint32_t hx, hy, ix, iy;
+  const float huge = 1.70141183460469231731687303716e+38; //0x1p127;	
+  
+  
+  x = z.real();
+  y = z.imag();
+  
+  get_float_word(hx, x);
+  get_float_word(hy, y);
+  
+  ix = 0x7fffffff & hx;
+  iy = 0x7fffffff & hy;
+  if (ix < 0x7f800000 && iy < 0x7f800000) {
+    if (iy == 0){
+      return (complex<float>(coshf(x), x * y));
+    }
+    if (ix < 0x41100000){	/* small x: normal case */
+      return (complex<float>(coshf(x) * cosf(y), sinhf(x) * sinf(y)));
+    }
+    /* |x| >= 9, so cosh(x) ~= exp(|x|) */
+    if (ix < 0x42b17218) {
+      /* x < 88.7: expf(|x|) won't overflow */
+      h = expf(fabsf(x)) * 0.5f;
+      return (complex<float>(h * cosf(y), copysignf(h, x) * sinf(y)));
+    } else if (ix < 0x4340b1e7) {
+      /* x < 192.7: scale to avoid overflow */
+      thrust::complex<float> z_;
+      z_ = ldexp_cexpf(complex<float>(fabsf(x), y), -1);
+      return (complex<float>(z_.real(), z_.imag() * copysignf(1.0f, x)));
+    } else {
+      /* x >= 192.7: the result always overflows */
+      h = huge * x;
+      return (complex<float>(h * h * cosf(y), h * sinf(y)));
+    }
+  }
+  
+  if (ix == 0 && iy >= 0x7f800000){
+    return (complex<float>(y - y, copysignf(0.0f, x * (y - y))));
+  }
+  if (iy == 0 && ix >= 0x7f800000) {
+    if ((hx & 0x7fffff) == 0)
+      return (complex<float>(x * x, copysignf(0.0f, x) * y));
+    return (complex<float>(x * x, copysignf(0.0f, (x + x) * y)));
+  }
+  
+  if (ix < 0x7f800000 && iy >= 0x7f800000){
+    return (complex<float>(y - y, x * (y - y)));
+  }
+  
+  if (ix >= 0x7f800000 && (hx & 0x7fffff) == 0) {
+    if (iy >= 0x7f800000)
+      return (complex<float>(x * x, x * (y - y)));
+    return (complex<float>((x * x) * cosf(y), x * sinf(y)));
+  }
+  return (complex<float>((x * x) * (y - y), (x + x) * (y - y)));
+}
+  
+__host__ __device__ inline
+complex<float> ccosf(const complex<float>& z){	
+  return (ccoshf(complex<float>(-z.imag(), z.real())));
+}
+
+} // namespace complex
+
+} // namespace detail
+
+template <>
+__host__ __device__
+inline complex<float> cos(const complex<float>& z){
+  return detail::complex::ccosf(z);
+}
+  
+template <>
+__host__ __device__
+inline complex<float> cosh(const complex<float>& z){
+  return detail::complex::ccoshf(z);
+}
+  
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexp.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexp.h
new file mode 100644
index 000000000000..c0c8c07d2508
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexp.h
@@ -0,0 +1,185 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/s_cexp.c
+ *    lib/msun/src/k_exp.c
+ *
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+/*
+ * Compute exp(x), scaled to avoid spurious overflow.  An exponent is
+ * returned separately in 'expt'.
+ *
+ * Input:  ln(DBL_MAX) <= x < ln(2 * DBL_MAX / DBL_MIN_DENORM) ~= 1454.91
+ * Output: 2**1023 <= y < 2**1024
+ */
+__host__ __device__ inline
+	double frexp_exp(double x, int *expt){
+  const uint32_t k = 1799;		/* constant for reduction */
+  const double kln2 =  1246.97177782734161156;	/* k * ln2 */
+	
+  double exp_x;
+  uint32_t hx;
+	
+  /*
+   * We use exp(x) = exp(x - kln2) * 2**k, carefully chosen to
+   * minimize |exp(kln2) - 2**k|.  We also scale the exponent of
+   * exp_x to MAX_EXP so that the result can be multiplied by
+   * a tiny number without losing accuracy due to denormalization.
+   */
+  exp_x = exp(x - kln2);
+  get_high_word(hx, exp_x);
+  *expt = (hx >> 20) - (0x3ff + 1023) + k;
+  set_high_word(exp_x, (hx & 0xfffff) | ((0x3ff + 1023) << 20));
+  return (exp_x);
+}
+      
+      
+__host__ __device__ inline
+complex<double>	ldexp_cexp(complex<double> z, int expt){
+  double x, y, exp_x, scale1, scale2;
+  int ex_expt, half_expt;
+	
+  x = z.real();
+  y = z.imag();
+  exp_x = frexp_exp(x, &ex_expt);
+  expt += ex_expt;
+	
+  /*
+   * Arrange so that scale1 * scale2 == 2**expt.  We use this to
+   * compensate for scalbn being horrendously slow.
+   */
+  half_expt = expt / 2;
+  insert_words(scale1, (0x3ff + half_expt) << 20, 0);
+  half_expt = expt - half_expt;
+  insert_words(scale2, (0x3ff + half_expt) << 20, 0);
+	
+  return (complex<double>(cos(y) * exp_x * scale1 * scale2,
+			  sin(y) * exp_x * scale1 * scale2));
+}
+	
+
+__host__ __device__ inline
+complex<double> cexp(const complex<double>& z){
+  double x, y, exp_x;
+  uint32_t hx, hy, lx, ly;
+
+  const uint32_t
+    exp_ovfl  = 0x40862e42,			/* high bits of MAX_EXP * ln2 ~= 710 */
+    cexp_ovfl = 0x4096b8e4;			/* (MAX_EXP - MIN_DENORM_EXP) * ln2 */
+
+	  
+  x = z.real();
+  y = z.imag();
+	  
+  extract_words(hy, ly, y);
+  hy &= 0x7fffffff;
+	  
+  /* cexp(x + I 0) = exp(x) + I 0 */
+  if ((hy | ly) == 0)
+    return (complex<double>(exp(x), y));
+  extract_words(hx, lx, x);
+  /* cexp(0 + I y) = cos(y) + I sin(y) */
+  if (((hx & 0x7fffffff) | lx) == 0)
+    return (complex<double>(cos(y), sin(y)));
+	  
+  if (hy >= 0x7ff00000) {
+    if (lx != 0 || (hx & 0x7fffffff) != 0x7ff00000) {
+      /* cexp(finite|NaN +- I Inf|NaN) = NaN + I NaN */
+      return (complex<double>(y - y, y - y));
+    } else if (hx & 0x80000000) {
+      /* cexp(-Inf +- I Inf|NaN) = 0 + I 0 */
+      return (complex<double>(0.0, 0.0));
+    } else {
+      /* cexp(+Inf +- I Inf|NaN) = Inf + I NaN */
+      return (complex<double>(x, y - y));
+    }
+  }
+	  
+  if (hx >= exp_ovfl && hx <= cexp_ovfl) {
+    /*
+     * x is between 709.7 and 1454.3, so we must scale to avoid
+     * overflow in exp(x).
+     */
+    return (ldexp_cexp(z, 0));
+  } else {
+    /*
+     * Cases covered here:
+     *  -  x < exp_ovfl and exp(x) won't overflow (common case)
+     *  -  x > cexp_ovfl, so exp(x) * s overflows for all s > 0
+     *  -  x = +-Inf (generated by exp())
+     *  -  x = NaN (spurious inexact exception from y)
+     */
+    exp_x = std::exp(x);
+    return (complex<double>(exp_x * cos(y), exp_x * sin(y)));
+  }
+}
+	
+} // namespace complex
+ 
+} // namespace detail
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> exp(const complex<ValueType>& z){    
+  return polar(std::exp(z.real()),z.imag());
+}
+
+template <>
+__host__ __device__
+inline complex<double> exp(const complex<double>& z){    
+  return detail::complex::cexp(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexpf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexpf.h
new file mode 100644
index 000000000000..cae030fe7bbf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cexpf.h
@@ -0,0 +1,163 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2011 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/s_cexpf.c
+ *    lib/msun/src/k_exp.c
+ *
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{
+
+__host__ __device__ inline
+float frexp_expf(float x, int *expt){
+  const uint32_t k = 235;                 /* constant for reduction */
+  const float kln2 =  162.88958740F;       /* k * ln2 */
+	
+  // should this be a double instead?
+  float exp_x;
+  uint32_t hx;
+	
+  exp_x = expf(x - kln2);
+  get_float_word(hx, exp_x);
+  *expt = (hx >> 23) - (0x7f + 127) + k;
+  set_float_word(exp_x, (hx & 0x7fffff) | ((0x7f + 127) << 23));
+  return (exp_x);
+}
+      
+__host__ __device__ inline
+complex<float> 
+ldexp_cexpf(complex<float> z, int expt)
+{
+  float x, y, exp_x, scale1, scale2;
+  int ex_expt, half_expt;
+	
+  x = z.real();
+  y = z.imag();
+  exp_x = frexp_expf(x, &ex_expt);
+  expt += ex_expt;
+	
+  half_expt = expt / 2;
+  set_float_word(scale1, (0x7f + half_expt) << 23);
+  half_expt = expt - half_expt;
+  set_float_word(scale2, (0x7f + half_expt) << 23);
+	
+  return (complex<float>(std::cos(y) * exp_x * scale1 * scale2,
+			 std::sin(y) * exp_x * scale1 * scale2));
+}
+      
+__host__ __device__ inline
+complex<float> cexpf(const complex<float>& z){
+  float x, y, exp_x;
+  uint32_t hx, hy;
+
+  const uint32_t
+    exp_ovfl  = 0x42b17218,		/* MAX_EXP * ln2 ~= 88.722839355 */
+    cexp_ovfl = 0x43400074;		/* (MAX_EXP - MIN_DENORM_EXP) * ln2 */
+
+  x = z.real();
+  y = z.imag();
+
+  get_float_word(hy, y);
+  hy &= 0x7fffffff;
+
+  /* cexp(x + I 0) = exp(x) + I 0 */
+  if (hy == 0)
+    return (complex<float>(std::exp(x), y));
+  get_float_word(hx, x);
+  /* cexp(0 + I y) = cos(y) + I sin(y) */
+  if ((hx & 0x7fffffff) == 0){
+    return (complex<float>(std::cos(y), std::sin(y)));
+  }
+  if (hy >= 0x7f800000) {
+    if ((hx & 0x7fffffff) != 0x7f800000) {
+      /* cexp(finite|NaN +- I Inf|NaN) = NaN + I NaN */
+      return (complex<float>(y - y, y - y));
+    } else if (hx & 0x80000000) {
+      /* cexp(-Inf +- I Inf|NaN) = 0 + I 0 */
+      return (complex<float>(0.0, 0.0));
+    } else {
+      /* cexp(+Inf +- I Inf|NaN) = Inf + I NaN */
+      return (complex<float>(x, y - y));
+    }
+  }
+
+  if (hx >= exp_ovfl && hx <= cexp_ovfl) {
+    /*
+     * x is between 88.7 and 192, so we must scale to avoid
+     * overflow in expf(x).
+     */
+    return (ldexp_cexpf(z, 0));
+  } else {
+    /*
+     * Cases covered here:
+     *  -  x < exp_ovfl and exp(x) won't overflow (common case)
+     *  -  x > cexp_ovfl, so exp(x) * s overflows for all s > 0
+     *  -  x = +-Inf (generated by exp())
+     *  -  x = NaN (spurious inexact exception from y)
+     */
+    exp_x = std::exp(x);
+    return (complex<float>(exp_x * std::cos(y), exp_x * std::sin(y)));
+  }
+}
+
+} // namespace complex
+
+} // namespace detail
+
+template <>
+__host__ __device__
+inline complex<float> exp(const complex<float>& z){    
+  return detail::complex::cexpf(z);
+}    
+  
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clog.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clog.h
new file mode 100644
index 000000000000..b727121c395d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clog.h
@@ -0,0 +1,214 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2012 Stephen Montgomery-Smith <stephen@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSDs msun:*/
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{
+
+using thrust::complex;
+
+/* round down to 18 = 54/3 bits */
+__host__ __device__ inline
+double trim(double x){
+  uint32_t hi;
+  get_high_word(hi, x);
+  insert_words(x, hi &0xfffffff8, 0);
+  return x;
+}
+
+
+__host__ __device__ inline
+complex<double> clog(const complex<double>& z){
+
+  // Adapted from FreeBSDs msun
+  double x, y;
+  double ax, ay;
+  double x0, y0, x1, y1, x2, y2, t, hm1;
+  double val[12];
+  int i, sorted;
+  const double e = 2.7182818284590452354;
+
+  x = z.real();
+  y = z.imag();
+
+  /* Handle NaNs using the general formula to mix them right. */
+  if (x != x || y != y){
+    return (complex<double>(std::log(norm(z)), std::atan2(y, x)));
+  }
+
+  ax = std::abs(x);
+  ay = std::abs(y);
+  if (ax < ay) {
+    t = ax;
+    ax = ay;
+    ay = t;
+  }
+
+  /*
+   * To avoid unnecessary overflow, if x and y are very large, divide x
+   * and y by M_E, and then add 1 to the logarithm.  This depends on
+   * M_E being larger than sqrt(2).
+   * There is a potential loss of accuracy caused by dividing by M_E,
+   * but this case should happen extremely rarely.
+   */
+  //    if (ay > 5e307){
+  // For high values of ay -> hypotf(DBL_MAX,ay) = inf
+  // We expect that for values at or below ay = 5e307 this should not happen
+  if (ay > 5e307){
+    return (complex<double>(std::log(hypot(x / e, y / e)) + 1.0, std::atan2(y, x)));
+  }
+  if (ax == 1.) {
+    if (ay < 1e-150){
+      return (complex<double>((ay * 0.5) * ay, std::atan2(y, x)));
+    }
+    return (complex<double>(log1p(ay * ay) * 0.5, std::atan2(y, x)));
+  }
+
+  /*
+   * Because atan2 and hypot conform to C99, this also covers all the
+   * edge cases when x or y are 0 or infinite.
+   */
+  if (ax < 1e-50 || ay < 1e-50 || ax > 1e50 || ay > 1e50){
+    return (complex<double>(std::log(hypot(x, y)), std::atan2(y, x)));
+  }
+
+  /*
+   * From this point on, we don't need to worry about underflow or
+   * overflow in calculating ax*ax or ay*ay.
+   */
+
+  /* Some easy cases. */
+
+  if (ax >= 1.0){
+    return (complex<double>(log1p((ax-1)*(ax+1) + ay*ay) * 0.5, atan2(y, x)));
+  }
+
+  if (ax*ax + ay*ay <= 0.7){
+    return (complex<double>(std::log(ax*ax + ay*ay) * 0.5, std::atan2(y, x)));
+  }
+
+  /*
+   * Take extra care so that ULP of real part is small if hypot(x,y) is
+   * moderately close to 1.
+   */
+
+
+  x0 = trim(ax);
+  ax = ax-x0;
+  x1 = trim(ax);
+  x2 = ax-x1;
+  y0 = trim(ay);
+  ay = ay-y0;
+  y1 = trim(ay);
+  y2 = ay-y1;
+
+  val[0] = x0*x0;
+  val[1] = y0*y0;
+  val[2] = 2*x0*x1;
+  val[3] = 2*y0*y1;
+  val[4] = x1*x1;
+  val[5] = y1*y1;
+  val[6] = 2*x0*x2;
+  val[7] = 2*y0*y2;
+  val[8] = 2*x1*x2;
+  val[9] = 2*y1*y2;
+  val[10] = x2*x2;
+  val[11] = y2*y2;
+
+  /* Bubble sort. */
+
+  do {
+    sorted = 1;
+    for (i=0;i<11;i++) {
+      if (val[i] < val[i+1]) {
+	sorted = 0;
+	t = val[i];
+	val[i] = val[i+1];
+	val[i+1] = t;
+      }
+    }
+  } while (!sorted);
+
+  hm1 = -1;
+  for (i=0;i<12;i++){
+    hm1 += val[i];
+  }
+  return (complex<double>(0.5 * log1p(hm1), atan2(y, x)));
+}
+
+} // namespace complex
+
+} // namespace detail
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> log(const complex<ValueType>& z){
+  return complex<ValueType>(std::log(thrust::abs(z)),thrust::arg(z));
+}
+
+template <>
+__host__ __device__
+inline complex<double> log(const complex<double>& z){
+  return detail::complex::clog(z);
+}
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> log10(const complex<ValueType>& z){
+  // Using the explicit literal prevents compile time warnings in
+  // devices that don't support doubles
+  return thrust::log(z)/ValueType(2.30258509299404568402);
+}
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clogf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clogf.h
new file mode 100644
index 000000000000..c72370c42842
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/clogf.h
@@ -0,0 +1,200 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2012 Stephen Montgomery-Smith <stephen@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSDs msun:*/
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{
+
+using thrust::complex;
+
+/* round down to 8 = 24/3 bits */
+__host__ __device__ inline
+float trim(float x){
+  uint32_t hx;
+  get_float_word(hx, x);
+  hx &= 0xffff0000;
+  float ret;
+  set_float_word(ret,hx);
+  return ret;
+}
+
+
+__host__ __device__ inline
+complex<float> clogf(const complex<float>& z){
+
+  // Adapted from FreeBSDs msun
+  float x, y;
+  float ax, ay;
+  float x0, y0, x1, y1, x2, y2, t, hm1;
+  float val[12];
+  int i, sorted;
+  const float e = 2.7182818284590452354f;
+
+  x = z.real();
+  y = z.imag();
+
+  /* Handle NaNs using the general formula to mix them right. */
+  if (x != x || y != y){
+    return (complex<float>(std::log(norm(z)), std::atan2(y, x)));
+  }
+
+  ax = std::abs(x);
+  ay = std::abs(y);
+  if (ax < ay) {
+    t = ax;
+    ax = ay;
+    ay = t;
+  }
+
+  /*
+   * To avoid unnecessary overflow, if x and y are very large, divide x
+   * and y by M_E, and then add 1 to the logarithm.  This depends on
+   * M_E being larger than sqrt(2).
+   * There is a potential loss of accuracy caused by dividing by M_E,
+   * but this case should happen extremely rarely.
+   */
+  // For high values of ay -> hypotf(FLT_MAX,ay) = inf
+  // We expect that for values at or below ay = 1e34f this should not happen
+  if (ay > 1e34f){
+    return (complex<float>(std::log(hypotf(x / e, y / e)) + 1.0f, std::atan2(y, x)));
+  }
+  if (ax == 1.f) {
+    if (ay < 1e-19f){
+      return (complex<float>((ay * 0.5f) * ay, std::atan2(y, x)));
+    }
+    return (complex<float>(log1pf(ay * ay) * 0.5f, std::atan2(y, x)));
+  }
+
+  /*
+   * Because atan2 and hypot conform to C99, this also covers all the
+   * edge cases when x or y are 0 or infinite.
+   */
+  if (ax < 1e-6f || ay < 1e-6f || ax > 1e6f || ay > 1e6f){
+    return (complex<float>(std::log(hypotf(x, y)), std::atan2(y, x)));
+  }
+
+  /*
+   * From this point on, we don't need to worry about underflow or
+   * overflow in calculating ax*ax or ay*ay.
+   */
+
+  /* Some easy cases. */
+
+  if (ax >= 1.0f){
+    return (complex<float>(log1pf((ax-1.f)*(ax+1.f) + ay*ay) * 0.5f, atan2(y, x)));
+  }
+
+  if (ax*ax + ay*ay <= 0.7f){
+    return (complex<float>(std::log(ax*ax + ay*ay) * 0.5f, std::atan2(y, x)));
+  }
+
+  /*
+   * Take extra care so that ULP of real part is small if hypot(x,y) is
+   * moderately close to 1.
+   */
+
+
+  x0 = trim(ax);
+  ax = ax-x0;
+  x1 = trim(ax);
+  x2 = ax-x1;
+  y0 = trim(ay);
+  ay = ay-y0;
+  y1 = trim(ay);
+  y2 = ay-y1;
+
+  val[0] = x0*x0;
+  val[1] = y0*y0;
+  val[2] = 2*x0*x1;
+  val[3] = 2*y0*y1;
+  val[4] = x1*x1;
+  val[5] = y1*y1;
+  val[6] = 2*x0*x2;
+  val[7] = 2*y0*y2;
+  val[8] = 2*x1*x2;
+  val[9] = 2*y1*y2;
+  val[10] = x2*x2;
+  val[11] = y2*y2;
+
+  /* Bubble sort. */
+
+  do {
+    sorted = 1;
+    for (i=0;i<11;i++) {
+      if (val[i] < val[i+1]) {
+	sorted = 0;
+	t = val[i];
+	val[i] = val[i+1];
+	val[i+1] = t;
+      }
+    }
+  } while (!sorted);
+
+  hm1 = -1;
+  for (i=0;i<12;i++){
+    hm1 += val[i];
+  }
+  return (complex<float>(0.5f * log1pf(hm1), atan2(y, x)));
+}
+
+} // namespace complex
+
+} // namespace detail
+
+template <>
+__host__ __device__
+inline complex<float> log(const complex<float>& z){
+  return detail::complex::clogf(z);
+}
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/complex.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/complex.inl
new file mode 100644
index 000000000000..a00b81a4b01d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/complex.inl
@@ -0,0 +1,355 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/type_traits/is_trivially_relocatable.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/* --- Constructors --- */
+
+#if THRUST_CPP_DIALECT < 2011
+template <typename T>
+__host__ __device__
+complex<T>::complex()
+{
+  real(T());
+  imag(T());
+}
+#endif
+
+template <typename T>
+__host__ __device__
+complex<T>::complex(const T& re)
+#if THRUST_CPP_DIALECT >= 2011
+  // Initialize the storage in the member initializer list using C++ unicorn
+  // initialization. This allows `complex<T const>` to work.
+  : data{re, T()}
+{}
+#else
+{
+  real(re);
+  imag(T());
+}
+#endif
+
+
+template <typename T>
+__host__ __device__
+complex<T>::complex(const T& re, const T& im)
+#if THRUST_CPP_DIALECT >= 2011
+  // Initialize the storage in the member initializer list using C++ unicorn
+  // initialization. This allows `complex<T const>` to work.
+  : data{re, im}
+{}
+#else
+{
+  real(re);
+  imag(im);
+}
+#endif
+
+#if THRUST_CPP_DIALECT < 2011
+template <typename T>
+__host__ __device__
+complex<T>::complex(const complex<T>& z)
+{
+  real(z.real());
+  imag(z.imag());
+}
+#endif
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>::complex(const complex<U>& z)
+#if THRUST_CPP_DIALECT >= 2011
+  // Initialize the storage in the member initializer list using C++ unicorn
+  // initialization. This allows `complex<T const>` to work.
+  // We do a functional-style cast here to suppress conversion warnings.
+  : data{T(z.real()), T(z.imag())}
+{}
+#else
+{
+  real(T(z.real()));
+  imag(T(z.imag()));
+}
+#endif
+
+template <typename T>
+__host__ THRUST_STD_COMPLEX_DEVICE
+complex<T>::complex(const std::complex<T>& z)
+#if THRUST_CPP_DIALECT >= 2011
+  // Initialize the storage in the member initializer list using C++ unicorn
+  // initialization. This allows `complex<T const>` to work.
+  : data{THRUST_STD_COMPLEX_REAL(z), THRUST_STD_COMPLEX_IMAG(z)}
+{}
+#else
+{
+  real(THRUST_STD_COMPLEX_REAL(z));
+  imag(THRUST_STD_COMPLEX_IMAG(z));
+}
+#endif
+
+template <typename T>
+template <typename U>
+__host__ THRUST_STD_COMPLEX_DEVICE
+complex<T>::complex(const std::complex<U>& z)
+#if THRUST_CPP_DIALECT >= 2011
+  // Initialize the storage in the member initializer list using C++ unicorn
+  // initialization. This allows `complex<T const>` to work.
+  // We do a functional-style cast here to suppress conversion warnings.
+  : data{T(THRUST_STD_COMPLEX_REAL(z)), T(THRUST_STD_COMPLEX_IMAG(z))}
+{}
+#else
+{
+  real(T(THRUST_STD_COMPLEX_REAL(z)));
+  imag(T(THRUST_STD_COMPLEX_IMAG(z)));
+}
+#endif
+
+
+
+/* --- Assignment Operators --- */
+
+template <typename T>
+__host__ __device__
+complex<T>& complex<T>::operator=(const T& re)
+{
+  real(re);
+  imag(T());
+  return *this;
+}
+
+#if THRUST_CPP_DIALECT < 2011
+template <typename T>
+__host__ __device__
+complex<T>& complex<T>::operator=(const complex<T>& z)
+{
+  real(z.real());
+  imag(z.imag());
+  return *this;
+}
+#endif
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator=(const complex<U>& z)
+{
+  real(T(z.real()));
+  imag(T(z.imag()));
+  return *this;
+}
+
+template <typename T>
+__host__ THRUST_STD_COMPLEX_DEVICE
+complex<T>& complex<T>::operator=(const std::complex<T>& z)
+{
+  real(THRUST_STD_COMPLEX_REAL(z));
+  imag(THRUST_STD_COMPLEX_IMAG(z));
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ THRUST_STD_COMPLEX_DEVICE
+complex<T>& complex<T>::operator=(const std::complex<U>& z)
+{
+  real(T(THRUST_STD_COMPLEX_REAL(z)));
+  imag(T(THRUST_STD_COMPLEX_IMAG(z)));
+  return *this;
+}
+
+
+
+/* --- Compound Assignment Operators --- */
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator+=(const complex<U>& z)
+{
+  *this = *this + z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator-=(const complex<U>& z)
+{
+  *this = *this - z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator*=(const complex<U>& z)
+{
+  *this = *this * z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator/=(const complex<U>& z)
+{
+  *this = *this / z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator+=(const U& z)
+{
+  *this = *this + z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator-=(const U& z)
+{
+  *this = *this - z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator*=(const U& z)
+{
+  *this = *this * z;
+  return *this;
+}
+
+template <typename T>
+template <typename U>
+__host__ __device__
+complex<T>& complex<T>::operator/=(const U& z)
+{
+  *this = *this / z;
+  return *this;
+}
+
+
+
+/* --- Equality Operators --- */
+
+template <typename T0, typename T1>
+__host__ __device__
+bool operator==(const complex<T0>& x, const complex<T1>& y)
+{
+  return x.real() == y.real() && x.imag() == y.imag();
+}
+
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator==(const complex<T0>& x, const std::complex<T1>& y)
+{
+  return x.real() == THRUST_STD_COMPLEX_REAL(y) && x.imag() == THRUST_STD_COMPLEX_IMAG(y);
+}
+
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator==(const std::complex<T0>& x, const complex<T1>& y)
+{
+  return THRUST_STD_COMPLEX_REAL(x) == y.real() && THRUST_STD_COMPLEX_IMAG(x) == y.imag();
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+bool operator==(const T0& x, const complex<T1>& y)
+{
+  return x == y.real() && y.imag() == T1();
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+bool operator==(const complex<T0>& x, const T1& y)
+{
+  return x.real() == y && x.imag() == T1();
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+bool operator!=(const complex<T0>& x, const complex<T1>& y)
+{
+  return !(x == y);
+}
+
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator!=(const complex<T0>& x, const std::complex<T1>& y)
+{
+  return !(x == y);
+}
+
+template <typename T0, typename T1>
+__host__ THRUST_STD_COMPLEX_DEVICE
+bool operator!=(const std::complex<T0>& x, const complex<T1>& y)
+{
+  return !(x == y);
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+bool operator!=(const T0& x, const complex<T1>& y)
+{
+  return !(x == y);
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+bool operator!=(const complex<T0>& x, const T1& y)
+{
+  return !(x == y);
+}
+
+template <typename T>
+struct proclaim_trivially_relocatable<complex<T> > : thrust::true_type {};
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/complex/arithmetic.h>
+#include <thrust/detail/complex/cproj.h>
+#include <thrust/detail/complex/cexp.h>
+#include <thrust/detail/complex/cexpf.h>
+#include <thrust/detail/complex/clog.h>
+#include <thrust/detail/complex/clogf.h>
+#include <thrust/detail/complex/cpow.h>
+#include <thrust/detail/complex/ccosh.h>
+#include <thrust/detail/complex/ccoshf.h>
+#include <thrust/detail/complex/csinh.h>
+#include <thrust/detail/complex/csinhf.h>
+#include <thrust/detail/complex/ctanh.h>
+#include <thrust/detail/complex/ctanhf.h>
+#include <thrust/detail/complex/csqrt.h>
+#include <thrust/detail/complex/csqrtf.h>
+#include <thrust/detail/complex/catrig.h>
+#include <thrust/detail/complex/catrigf.h>
+#include <thrust/detail/complex/stream.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cpow.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cpow.h
new file mode 100644
index 000000000000..c204c451fff9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cpow.h
@@ -0,0 +1,57 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+pow(const complex<T0>& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return exp(log(complex<T>(x)) * complex<T>(y));
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+pow(const complex<T0>& x, const T1& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  return exp(log(complex<T>(x)) * T(y));
+}
+
+template <typename T0, typename T1>
+__host__ __device__
+complex<typename detail::promoted_numerical_type<T0, T1>::type>
+pow(const T0& x, const complex<T1>& y)
+{
+  typedef typename detail::promoted_numerical_type<T0, T1>::type T;
+  // Find `log` by ADL.
+  using std::log;
+  return exp(log(T(x)) * complex<T>(y));
+}
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cproj.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cproj.h
new file mode 100644
index 000000000000..7537c99fd75e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/cproj.h
@@ -0,0 +1,72 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{	 
+__host__ __device__
+inline complex<float> cprojf(const complex<float>& z){
+  if(!isinf(z.real()) && !isinf(z.imag())){
+    return z;
+  }else{
+    // std::numeric_limits<T>::infinity() doesn't run on the GPU
+    return complex<float>(infinity<float>(), copysignf(0.0, z.imag()));
+  }
+}
+  
+__host__ __device__
+inline complex<double> cproj(const complex<double>& z){
+  if(!isinf(z.real()) && !isinf(z.imag())){
+    return z;
+  }else{
+    // std::numeric_limits<T>::infinity() doesn't run on the GPU
+    return complex<double>(infinity<double>(), copysign(0.0, z.imag()));
+  }
+}
+
+}
+ 
+}
+
+template <typename T>
+__host__ __device__
+inline thrust::complex<T> proj(const thrust::complex<T>& z){
+  return detail::complex::cproj(z);
+}
+  
+
+template <>
+__host__ __device__
+inline thrust::complex<double> proj(const thrust::complex<double>& z){
+  return detail::complex::cproj(z);
+}
+  
+template <>
+__host__ __device__
+inline thrust::complex<float> proj(const thrust::complex<float>& z){
+  return detail::complex::cprojf(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinh.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinh.h
new file mode 100644
index 000000000000..b5a22af01e6b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinh.h
@@ -0,0 +1,207 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/s_csinh.c
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+
+__host__ __device__ inline
+complex<double> csinh(const complex<double>& z){
+  double x, y, h;
+  uint32_t hx, hy, ix, iy, lx, ly;
+  const double huge = 8.98846567431157953864652595395e+307; // 0x1p1023;
+
+  x = z.real();
+  y = z.imag();
+
+  extract_words(hx, lx, x);
+  extract_words(hy, ly, y);
+
+  ix = 0x7fffffff & hx;
+  iy = 0x7fffffff & hy;
+
+  /* Handle the nearly-non-exceptional cases where x and y are finite. */
+  if (ix < 0x7ff00000 && iy < 0x7ff00000) {
+    if ((iy | ly) == 0)
+      return (complex<double>(sinh(x), y));
+    if (ix < 0x40360000)	/* small x: normal case */
+      return (complex<double>(sinh(x) * cos(y), cosh(x) * sin(y)));
+
+    /* |x| >= 22, so cosh(x) ~= exp(|x|) */
+    if (ix < 0x40862e42) {
+      /* x < 710: exp(|x|) won't overflow */
+      h = exp(fabs(x)) * 0.5;
+      return (complex<double>(copysign(h, x) * cos(y), h * sin(y)));
+    } else if (ix < 0x4096bbaa) {
+      /* x < 1455: scale to avoid overflow */
+      complex<double> z_ = ldexp_cexp(complex<double>(fabs(x), y), -1);
+      return (complex<double>(z_.real() * copysign(1.0, x), z_.imag()));
+    } else {
+      /* x >= 1455: the result always overflows */
+      h = huge * x;
+      return (complex<double>(h * cos(y), h * h * sin(y)));
+    }
+  }
+
+  /*
+   * sinh(+-0 +- I Inf) = sign(d(+-0, dNaN))0 + I dNaN.
+   * The sign of 0 in the result is unspecified.  Choice = normally
+   * the same as dNaN.  Raise the invalid floating-point exception.
+   *
+   * sinh(+-0 +- I NaN) = sign(d(+-0, NaN))0 + I d(NaN).
+   * The sign of 0 in the result is unspecified.  Choice = normally
+   * the same as d(NaN).
+   */
+  if ((ix | lx) == 0 && iy >= 0x7ff00000)
+    return (complex<double>(copysign(0.0, x * (y - y)), y - y));
+
+  /*
+   * sinh(+-Inf +- I 0) = +-Inf + I +-0.
+   *
+   * sinh(NaN +- I 0)   = d(NaN) + I +-0.
+   */
+  if ((iy | ly) == 0 && ix >= 0x7ff00000) {
+    if (((hx & 0xfffff) | lx) == 0)
+      return (complex<double>(x, y));
+    return (complex<double>(x, copysign(0.0, y)));
+  }
+
+  /*
+   * sinh(x +- I Inf) = dNaN + I dNaN.
+   * Raise the invalid floating-point exception for finite nonzero x.
+   *
+   * sinh(x + I NaN) = d(NaN) + I d(NaN).
+   * Optionally raises the invalid floating-point exception for finite
+   * nonzero x.  Choice = don't raise (except for signaling NaNs).
+   */
+  if (ix < 0x7ff00000 && iy >= 0x7ff00000)
+    return (complex<double>(y - y, x * (y - y)));
+
+  /*
+   * sinh(+-Inf + I NaN)  = +-Inf + I d(NaN).
+   * The sign of Inf in the result is unspecified.  Choice = normally
+   * the same as d(NaN).
+   *
+   * sinh(+-Inf +- I Inf) = +Inf + I dNaN.
+   * The sign of Inf in the result is unspecified.  Choice = always +.
+   * Raise the invalid floating-point exception.
+   *
+   * sinh(+-Inf + I y)   = +-Inf cos(y) + I Inf sin(y)
+   */
+  if (ix >= 0x7ff00000 && ((hx & 0xfffff) | lx) == 0) {
+    if (iy >= 0x7ff00000)
+      return (complex<double>(x * x, x * (y - y)));
+    return (complex<double>(x * cos(y), infinity<double>() * sin(y)));
+  }
+
+  /*
+   * sinh(NaN + I NaN)  = d(NaN) + I d(NaN).
+   *
+   * sinh(NaN +- I Inf) = d(NaN) + I d(NaN).
+   * Optionally raises the invalid floating-point exception.
+   * Choice = raise.
+   *
+   * sinh(NaN + I y)    = d(NaN) + I d(NaN).
+   * Optionally raises the invalid floating-point exception for finite
+   * nonzero y.  Choice = don't raise (except for signaling NaNs).
+   */
+  return (complex<double>((x * x) * (y - y), (x + x) * (y - y)));
+}
+
+__host__ __device__ inline
+complex<double> csin(complex<double> z){
+  /* csin(z) = -I * csinh(I * z) */
+  z = csinh(complex<double>(-z.imag(), z.real()));
+  return (complex<double>(z.imag(), -z.real()));
+}
+
+} // namespace complex
+
+} // namespace detail
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> sin(const complex<ValueType>& z){
+  const ValueType re = z.real();
+  const ValueType im = z.imag();
+  return complex<ValueType>(std::sin(re) * std::cosh(im), 
+			    std::cos(re) * std::sinh(im));
+}
+
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> sinh(const complex<ValueType>& z){
+  const ValueType re = z.real();
+  const ValueType im = z.imag();
+  return complex<ValueType>(std::sinh(re) * std::cos(im), 
+			    std::cosh(re) * std::sin(im));
+}
+
+template <>
+__host__ __device__
+inline complex<double> sin(const complex<double>& z){
+  return detail::complex::csin(z);
+}
+
+template <>
+__host__ __device__
+inline complex<double> sinh(const complex<double>& z){
+  return detail::complex::csinh(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinhf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinhf.h
new file mode 100644
index 000000000000..d271081c6547
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csinhf.h
@@ -0,0 +1,144 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2005 Bruce D. Evans and Steven G. Kargl
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/s_csinhf.c
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+
+__host__ __device__ inline
+complex<float> csinhf(const complex<float>& z){
+
+  float x, y, h;
+  uint32_t hx, hy, ix, iy;
+
+  const float huge = 1.70141183460469231731687303716e+38; //0x1p127;
+
+  x = z.real();
+  y = z.imag();
+
+  get_float_word(hx, x);
+  get_float_word(hy, y);
+
+  ix = 0x7fffffff & hx;
+  iy = 0x7fffffff & hy;
+
+  if (ix < 0x7f800000 && iy < 0x7f800000) {
+    if (iy == 0)
+      return (complex<float>(sinhf(x), y));
+    if (ix < 0x41100000)	/* small x: normal case */
+      return (complex<float>(sinhf(x) * cosf(y), coshf(x) * sinf(y)));
+
+    /* |x| >= 9, so cosh(x) ~= exp(|x|) */
+    if (ix < 0x42b17218) {
+      /* x < 88.7: expf(|x|) won't overflow */
+      h = expf(fabsf(x)) * 0.5f;
+      return (complex<float>(copysignf(h, x) * cosf(y), h * sinf(y)));
+    } else if (ix < 0x4340b1e7) {
+      /* x < 192.7: scale to avoid overflow */
+      complex<float> z_ = ldexp_cexpf(complex<float>(fabsf(x), y), -1);
+      return (complex<float>(z_.real() * copysignf(1.0f, x), z_.imag()));
+    } else {
+      /* x >= 192.7: the result always overflows */
+      h = huge * x;
+      return (complex<float>(h * cosf(y), h * h * sinf(y)));
+    }
+  }
+
+  if (ix == 0 && iy >= 0x7f800000)
+    return (complex<float>(copysignf(0, x * (y - y)), y - y));
+
+  if (iy == 0 && ix >= 0x7f800000) {
+    if ((hx & 0x7fffff) == 0)
+      return (complex<float>(x, y));
+    return (complex<float>(x, copysignf(0.0f, y)));
+  }
+
+  if (ix < 0x7f800000 && iy >= 0x7f800000)
+    return (complex<float>(y - y, x * (y - y)));
+
+  if (ix >= 0x7f800000 && (hx & 0x7fffff) == 0) {
+    if (iy >= 0x7f800000)
+      return (complex<float>(x * x, x * (y - y)));
+    return (complex<float>(x * cosf(y), infinity<float>() * sinf(y)));
+  }
+
+  return (complex<float>((x * x) * (y - y), (x + x) * (y - y)));
+}
+
+__host__ __device__ inline
+complex<float> csinf(complex<float> z){
+  z = csinhf(complex<float>(-z.imag(), z.real()));
+  return (complex<float>(z.imag(), -z.real()));
+}
+      
+} // namespace complex
+
+} // namespace detail
+  
+template <>
+__host__ __device__
+inline complex<float> sin(const complex<float>& z){
+  return detail::complex::csinf(z);
+}
+
+template <>
+__host__ __device__
+inline complex<float> sinh(const complex<float>& z){
+  return detail::complex::csinhf(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrt.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrt.h
new file mode 100644
index 000000000000..eb4da5289f83
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrt.h
@@ -0,0 +1,154 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2007 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/*
+ * Adapted from FreeBSD by Filipe Maia <filipe.c.maia@gmail.com>:
+ *    freebsd/lib/msun/src/s_csqrt.c
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+
+__host__ __device__ inline
+complex<double> csqrt(const complex<double>& z){
+  complex<double> result;
+  double a, b;
+  double t;
+  int scale;
+
+  /* We risk spurious overflow for components >= DBL_MAX / (1 + sqrt(2)). */
+  const double THRESH = 7.446288774449766337959726e+307;
+
+  a = z.real();
+  b = z.imag();
+
+  /* Handle special cases. */
+  if (z == 0.0)
+    return (complex<double>(0.0, b));
+  if (isinf(b))
+    return (complex<double>(infinity<double>(), b));
+  if (isnan(a)) {
+    t = (b - b) / (b - b);	/* raise invalid if b is not a NaN */
+    return (complex<double>(a, t));	/* return NaN + NaN i */
+  }
+  if (isinf(a)) {
+    /*
+     * csqrt(inf + NaN i)  = inf +  NaN i
+     * csqrt(inf + y i)    = inf +  0 i
+     * csqrt(-inf + NaN i) = NaN +- inf i
+     * csqrt(-inf + y i)   = 0   +  inf i
+     */
+    if (signbit(a))
+      return (complex<double>(fabs(b - b), copysign(a, b)));
+    else
+      return (complex<double>(a, copysign(b - b, b)));
+  }
+  /*
+   * The remaining special case (b is NaN) is handled just fine by
+   * the normal code path below.
+   */
+
+  // DBL_MIN*2
+  const double low_thresh = 4.450147717014402766180465e-308;
+  scale = 0;
+
+  if (fabs(a) >= THRESH || fabs(b) >= THRESH) {
+    /* Scale to avoid overflow. */
+    a *= 0.25;
+    b *= 0.25;
+    scale = 1;
+  }else if (fabs(a) <= low_thresh && fabs(b) <= low_thresh) {
+    /* Scale to avoid underflow. */
+    a *= 4.0;
+    b *= 4.0;
+    scale = 2;
+  }
+	
+
+  /* Algorithm 312, CACM vol 10, Oct 1967. */
+  if (a >= 0.0) {
+    t = sqrt((a + hypot(a, b)) * 0.5);
+    result = complex<double>(t, b / (2 * t));
+  } else {
+    t = sqrt((-a + hypot(a, b)) * 0.5);
+    result = complex<double>(fabs(b) / (2 * t), copysign(t, b));
+  }
+
+  /* Rescale. */
+  if (scale == 1)
+    return (result * 2.0);
+  else if (scale == 2)
+    return (result * 0.5);
+  else
+    return (result);
+}
+      
+} // namespace complex
+
+} // namespace detail
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> sqrt(const complex<ValueType>& z){
+  return thrust::polar(std::sqrt(thrust::abs(z)),thrust::arg(z)/ValueType(2));
+}
+
+template <>
+__host__ __device__
+inline complex<double> sqrt(const complex<double>& z){
+  return detail::complex::csqrt(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrtf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrtf.h
new file mode 100644
index 000000000000..dba489a33a89
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/csqrtf.h
@@ -0,0 +1,149 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2007 David Schultz <das@FreeBSD.ORG>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/*
+ * Adapted from FreeBSD by Filipe Maia <filipe.c.maia@gmail.com>:
+ *    freebsd/lib/msun/src/s_csqrt.c
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+
+__host__ __device__ inline
+complex<float> csqrtf(const complex<float>& z){
+  float a = z.real(), b = z.imag();
+  float t;
+  int scale;
+  complex<float> result;
+
+  /* We risk spurious overflow for components >= FLT_MAX / (1 + sqrt(2)). */
+  const float THRESH = 1.40949553037932e+38f;
+
+  /* Handle special cases. */
+  if (z == 0.0f)
+    return (complex<float>(0, b));
+  if (isinf(b))
+    return (complex<float>(infinity<float>(), b));
+  if (isnan(a)) {
+    t = (b - b) / (b - b);	/* raise invalid if b is not a NaN */
+    return (complex<float>(a, t));	/* return NaN + NaN i */
+  }
+  if (isinf(a)) {
+    /*
+     * csqrtf(inf + NaN i)  = inf +  NaN i
+     * csqrtf(inf + y i)    = inf +  0 i
+     * csqrtf(-inf + NaN i) = NaN +- inf i
+     * csqrtf(-inf + y i)   = 0   +  inf i
+     */
+    if (signbit(a))
+      return (complex<float>(fabsf(b - b), copysignf(a, b)));
+    else
+      return (complex<float>(a, copysignf(b - b, b)));
+  }
+  /*
+   * The remaining special case (b is NaN) is handled just fine by
+   * the normal code path below.
+   */
+
+  /* 
+   * Unlike in the FreeBSD code we'll avoid using double precision as
+   * not all hardware supports it.
+   */
+
+  // FLT_MIN*2
+  const float low_thresh = 2.35098870164458e-38f;
+  scale = 0;
+
+  if (fabsf(a) >= THRESH || fabsf(b) >= THRESH) {
+    /* Scale to avoid overflow. */
+    a *= 0.25f;
+    b *= 0.25f;
+    scale = 1;
+  }else if (fabsf(a) <= low_thresh && fabsf(b) <= low_thresh) {
+    /* Scale to avoid underflow. */
+    a *= 4.f;
+    b *= 4.f;
+    scale = 2;
+  }
+
+  /* Algorithm 312, CACM vol 10, Oct 1967. */
+  if (a >= 0.0f) {
+    t = sqrtf((a + hypotf(a, b)) * 0.5f);
+    result = complex<float>(t, b / (2.0f * t));
+  } else {
+    t = sqrtf((-a + hypotf(a, b)) * 0.5f);
+    result = complex<float>(fabsf(b) / (2.0f * t), copysignf(t, b));
+  }
+
+  /* Rescale. */
+  if (scale == 1)
+    return (result * 2.0f);
+  else if (scale == 2)
+    return (result * 0.5f);
+  else
+    return (result);
+}      
+
+} // namespace complex
+
+} // namespace detail
+
+template <>
+__host__ __device__
+inline complex<float> sqrt(const complex<float>& z){
+  return detail::complex::csqrtf(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanh.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanh.h
new file mode 100644
index 000000000000..3275c0343cb7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanh.h
@@ -0,0 +1,202 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2011 David Schultz
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * Adapted from FreeBSD by Filipe Maia <filipe.c.maia@gmail.com>:
+ *    freebsd/lib/msun/src/s_ctanh.c
+ */
+
+/*
+ * Hyperbolic tangent of a complex argument z = x + i y.
+ *
+ * The algorithm is from:
+ *
+ *   W. Kahan.  Branch Cuts for Complex Elementary Functions or Much
+ *   Ado About Nothing's Sign Bit.  In The State of the Art in
+ *   Numerical Analysis, pp. 165 ff.  Iserles and Powell, eds., 1987.
+ *
+ * Method:
+ *
+ *   Let t    = tan(x)
+ *       beta = 1/cos^2(y)
+ *       s    = sinh(x)
+ *       rho  = cosh(x)
+ *
+ *   We have:
+ *
+ *   tanh(z) = sinh(z) / cosh(z)
+ *
+ *             sinh(x) cos(y) + i cosh(x) sin(y)
+ *           = ---------------------------------
+ *             cosh(x) cos(y) + i sinh(x) sin(y)
+ *
+ *             cosh(x) sinh(x) / cos^2(y) + i tan(y)
+ *           = -------------------------------------
+ *                    1 + sinh^2(x) / cos^2(y)
+ *
+ *             beta rho s + i t
+ *           = ----------------
+ *               1 + beta s^2
+ *
+ * Modifications:
+ *
+ *   I omitted the original algorithm's handling of overflow in tan(x) after
+ *   verifying with nearpi.c that this can't happen in IEEE single or double
+ *   precision.  I also handle large x differently.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+
+__host__ __device__ inline
+complex<double> ctanh(const complex<double>& z){
+  double x, y;
+  double t, beta, s, rho, denom;
+  uint32_t hx, ix, lx;
+
+  x = z.real();
+  y = z.imag();
+
+  extract_words(hx, lx, x);
+  ix = hx & 0x7fffffff;
+
+  /*
+   * ctanh(NaN + i 0) = NaN + i 0
+   *
+   * ctanh(NaN + i y) = NaN + i NaN		for y != 0
+   *
+   * The imaginary part has the sign of x*sin(2*y), but there's no
+   * special effort to get this right.
+   *
+   * ctanh(+-Inf +- i Inf) = +-1 +- 0
+   *
+   * ctanh(+-Inf + i y) = +-1 + 0 sin(2y)		for y finite
+   *
+   * The imaginary part of the sign is unspecified.  This special
+   * case is only needed to avoid a spurious invalid exception when
+   * y is infinite.
+   */
+  if (ix >= 0x7ff00000) {
+    if ((ix & 0xfffff) | lx)	/* x is NaN */
+      return (complex<double>(x, (y == 0 ? y : x * y)));
+    set_high_word(x, hx - 0x40000000);	/* x = copysign(1, x) */
+    return (complex<double>(x, copysign(0.0, isinf(y) ? y : sin(y) * cos(y))));
+  }
+
+  /*
+   * ctanh(x + i NAN) = NaN + i NaN
+   * ctanh(x +- i Inf) = NaN + i NaN
+   */
+  if (!isfinite(y))
+    return (complex<double>(y - y, y - y));
+
+  /*
+   * ctanh(+-huge + i +-y) ~= +-1 +- i 2sin(2y)/exp(2x), using the
+   * approximation sinh^2(huge) ~= exp(2*huge) / 4.
+   * We use a modified formula to avoid spurious overflow.
+   */
+  if (ix >= 0x40360000) {	/* x >= 22 */
+    double exp_mx = exp(-fabs(x));
+    return (complex<double>(copysign(1.0, x),
+			    4.0 * sin(y) * cos(y) * exp_mx * exp_mx));
+  }
+
+  /* Kahan's algorithm */
+  t = tan(y);
+  beta = 1.0 + t * t;	/* = 1 / cos^2(y) */
+  s = sinh(x);
+  rho = sqrt(1.0 + s * s);	/* = cosh(x) */
+  denom = 1.0 + beta * s * s;
+  return (complex<double>((beta * rho * s) / denom, t / denom));
+}
+
+__host__ __device__ inline
+complex<double> ctan(complex<double> z){
+  /* ctan(z) = -I * ctanh(I * z) */
+  z = ctanh(complex<double>(-z.imag(), z.real()));
+  return (complex<double>(z.imag(), -z.real()));
+}
+
+} // namespace complex
+
+} // namespace detail
+
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> tan(const complex<ValueType>& z){
+  return sin(z)/cos(z);
+}
+
+template <typename ValueType>
+__host__ __device__
+inline complex<ValueType> tanh(const complex<ValueType>& z){
+  // This implementation seems better than the simple sin/cos
+  return (thrust::exp(ValueType(2)*z)-ValueType(1))/
+    (thrust::exp(ValueType(2)*z)+ValueType(1));
+}
+
+template <>
+__host__ __device__
+inline complex<double> tan(const complex<double>& z){
+  return detail::complex::ctan(z);
+}
+  
+template <>
+__host__ __device__
+inline complex<double> tanh(const complex<double>& z){
+  return detail::complex::ctanh(z);
+}
+  
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanhf.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanhf.h
new file mode 100644
index 000000000000..221b5ce47cb0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/ctanhf.h
@@ -0,0 +1,126 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*-
+ * Copyright (c) 2011 David Schultz
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice unmodified, this list of conditions, and the following
+ *    disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * Adapted from FreeBSD by Filipe Maia, filipe.c.maia@gmail.com:
+ *    freebsd/lib/msun/src/s_ctanhf.c
+ */
+
+/*
+ * Hyperbolic tangent of a complex argument z.  See ctanh.c for details.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+#include <thrust/detail/complex/math_private.h>
+#include <cmath>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{		      	
+
+using thrust::complex;
+
+__host__ __device__ inline
+complex<float> ctanhf(const complex<float>& z){
+  float x, y;
+  float t, beta, s, rho, denom;
+  uint32_t hx, ix;
+
+  x = z.real();
+  y = z.imag();
+
+  get_float_word(hx, x);
+  ix = hx & 0x7fffffff;
+
+  if (ix >= 0x7f800000) {
+    if (ix & 0x7fffff)
+      return (complex<float>(x, (y == 0.0f ? y : x * y)));
+    set_float_word(x, hx - 0x40000000);
+    return (complex<float>(x,
+			   copysignf(0, isinf(y) ? y : sinf(y) * cosf(y))));
+  }
+
+  if (!isfinite(y))
+    return (complex<float>(y - y, y - y));
+
+  if (ix >= 0x41300000) {	/* x >= 11 */
+    float exp_mx = expf(-fabsf(x));
+    return (complex<float>(copysignf(1.0f, x),
+			   4.0f * sinf(y) * cosf(y) * exp_mx * exp_mx));
+  }
+
+  t = tanf(y);
+  beta = 1.0f + t * t;
+  s = sinhf(x);
+  rho = sqrtf(1.0f + s * s);
+  denom = 1.0f + beta * s * s;
+  return (complex<float>((beta * rho * s) / denom, t / denom));
+}
+
+  __host__ __device__ inline
+  complex<float> ctanf(complex<float> z){
+    z = ctanhf(complex<float>(-z.imag(), z.real()));
+    return (complex<float>(z.imag(), -z.real()));
+  }
+
+} // namespace complex
+
+} // namespace detail
+
+template <>
+__host__ __device__
+inline complex<float> tan(const complex<float>& z){
+  return detail::complex::ctanf(z);
+}
+
+template <>
+__host__ __device__
+inline complex<float> tanh(const complex<float>& z){
+  return detail::complex::ctanhf(z);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/math_private.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/math_private.h
new file mode 100644
index 000000000000..3a40c8e72918
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/math_private.h
@@ -0,0 +1,136 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*
+ * ====================================================
+ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+ *
+ * Developed at SunPro, a Sun Microsystems, Inc. business.
+ * Permission to use, copy, modify, and distribute this
+ * software is freely granted, provided that this notice
+ * is preserved.
+ * ====================================================
+ */
+
+/* adapted from FreeBSD:
+ *    lib/msun/src/math_private.h
+ */
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/complex.h>
+#include <thrust/detail/cstdint.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail{
+namespace complex{
+
+using thrust::complex;
+
+typedef union
+{
+  float value;
+  uint32_t word;
+} ieee_float_shape_type;
+  
+__host__ __device__
+inline void get_float_word(uint32_t & i, float d){
+  ieee_float_shape_type gf_u;
+  gf_u.value = (d);
+  (i) = gf_u.word;
+}
+
+__host__ __device__
+inline void get_float_word(int32_t & i, float d){
+  ieee_float_shape_type gf_u;
+  gf_u.value = (d);
+  (i) = gf_u.word;
+}
+
+__host__ __device__
+inline void set_float_word(float & d, uint32_t i){
+  ieee_float_shape_type sf_u;
+  sf_u.word = (i);
+  (d) = sf_u.value;
+}
+
+// Assumes little endian ordering
+typedef union
+{
+  double value;
+  struct
+  {
+    uint32_t lsw;
+    uint32_t msw;
+  } parts;
+  struct
+  {
+    uint64_t w;
+  } xparts;
+} ieee_double_shape_type;
+  
+__host__ __device__ inline
+void get_high_word(uint32_t & i,double d){
+  ieee_double_shape_type gh_u;
+  gh_u.value = (d);
+  (i) = gh_u.parts.msw;                                   
+}
+  
+/* Set the more significant 32 bits of a double from an int.  */
+__host__ __device__ inline
+void set_high_word(double & d, uint32_t v){
+  ieee_double_shape_type sh_u;
+  sh_u.value = (d);
+  sh_u.parts.msw = (v);
+  (d) = sh_u.value;
+}
+  
+  
+__host__ __device__ inline 
+void  insert_words(double & d, uint32_t ix0, uint32_t ix1){
+  ieee_double_shape_type iw_u;
+  iw_u.parts.msw = (ix0);
+  iw_u.parts.lsw = (ix1);
+  (d) = iw_u.value;
+}
+  
+/* Get two 32 bit ints from a double.  */
+__host__ __device__ inline
+void  extract_words(uint32_t & ix0,uint32_t & ix1, double d){
+  ieee_double_shape_type ew_u;
+  ew_u.value = (d);
+  (ix0) = ew_u.parts.msw;
+  (ix1) = ew_u.parts.lsw;
+}
+  
+/* Get two 32 bit ints from a double.  */
+__host__ __device__ inline
+void  extract_words(int32_t & ix0,int32_t & ix1, double d){
+  ieee_double_shape_type ew_u;
+  ew_u.value = (d);
+  (ix0) = ew_u.parts.msw;
+  (ix1) = ew_u.parts.lsw;
+}
+  
+} // namespace complex
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
+
+
+#include <thrust/detail/complex/c99math.h>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/stream.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/stream.h
new file mode 100644
index 000000000000..95434b41beb7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/complex/stream.h
@@ -0,0 +1,74 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *  Copyright 2013 Filipe RNC Maia
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/complex.h>
+
+THRUST_NAMESPACE_BEGIN
+template<typename ValueType,class charT, class traits>
+std::basic_ostream<charT, traits>& operator<<(std::basic_ostream<charT, traits>& os, const complex<ValueType>& z)
+{
+  os << '(' << z.real() << ',' << z.imag() << ')';
+  return os;
+}
+
+template<typename ValueType, typename charT, class traits>
+std::basic_istream<charT, traits>&
+operator>>(std::basic_istream<charT, traits>& is, complex<ValueType>& z)
+{
+  ValueType re, im;
+
+  charT ch;
+  is >> ch;
+
+  if(ch == '(')
+    {
+      is >> re >> ch;
+      if (ch == ',')
+        {
+          is >> im >> ch;
+          if (ch == ')')
+	    {
+	      z = complex<ValueType>(re, im);
+	    }
+          else
+	    {
+	      is.setstate(std::ios_base::failbit);
+	    }
+        }
+      else if (ch == ')')
+        {
+          z = re;
+        }
+      else
+        {
+          is.setstate(std::ios_base::failbit);
+        }
+    }
+  else
+    {
+      is.putback(ch);
+      is >> re;
+      z = re;
+    }
+  return is;
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config.h
new file mode 100644
index 000000000000..5a5573a410e6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config.h
@@ -0,0 +1,24 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+/*! \file config.h
+ *  \brief Defines platform configuration.
+ */
+
+#pragma once
+
+#include <thrust/version.h>
+#include <thrust/detail/config/config.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler.h
new file mode 100644
index 000000000000..e35652f6a41f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler.h
@@ -0,0 +1,189 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file compiler.h
+ *  \brief Compiler-specific configuration
+ */
+
+#pragma once
+
+// enumerate host compilers we know about
+#define THRUST_HOST_COMPILER_UNKNOWN 0
+#define THRUST_HOST_COMPILER_MSVC    1
+#define THRUST_HOST_COMPILER_GCC     2
+#define THRUST_HOST_COMPILER_CLANG   3
+#define THRUST_HOST_COMPILER_INTEL   4
+
+// enumerate device compilers we know about
+#define THRUST_DEVICE_COMPILER_UNKNOWN 0
+#define THRUST_DEVICE_COMPILER_MSVC    1
+#define THRUST_DEVICE_COMPILER_GCC     2
+#define THRUST_DEVICE_COMPILER_CLANG   3
+#define THRUST_DEVICE_COMPILER_NVCC    4
+
+// figure out which host compiler we're using
+// XXX we should move the definition of THRUST_DEPRECATED out of this logic
+#if   defined(_MSC_VER)
+#define THRUST_HOST_COMPILER THRUST_HOST_COMPILER_MSVC
+#define THRUST_MSVC_VERSION _MSC_VER
+#define THRUST_MSVC_VERSION_FULL _MSC_FULL_VER
+#elif defined(__ICC)
+#define THRUST_HOST_COMPILER THRUST_HOST_COMPILER_INTEL
+#elif defined(__clang__)
+#define THRUST_HOST_COMPILER THRUST_HOST_COMPILER_CLANG
+#define THRUST_CLANG_VERSION (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__)
+#elif defined(__GNUC__)
+#define THRUST_HOST_COMPILER THRUST_HOST_COMPILER_GCC
+#define THRUST_GCC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
+#if (THRUST_GCC_VERSION >= 50000)
+#define THRUST_MODERN_GCC
+#else
+#define THRUST_LEGACY_GCC
+#endif
+#else
+#define THRUST_HOST_COMPILER THRUST_HOST_COMPILER_UNKNOWN
+#endif // THRUST_HOST_COMPILER
+
+// figure out which device compiler we're using
+#if defined(__CUDACC__) || defined(_NVHPC_CUDA)
+#define THRUST_DEVICE_COMPILER THRUST_DEVICE_COMPILER_NVCC
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+#define THRUST_DEVICE_COMPILER THRUST_DEVICE_COMPILER_MSVC
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC
+#define THRUST_DEVICE_COMPILER THRUST_DEVICE_COMPILER_GCC
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG
+// CUDA-capable clang should behave similar to NVCC.
+#if defined(__CUDA__)
+#define THRUST_DEVICE_COMPILER THRUST_DEVICE_COMPILER_NVCC
+#else
+#define THRUST_DEVICE_COMPILER THRUST_DEVICE_COMPILER_CLANG
+#endif
+#else
+#define THRUST_DEVICE_COMPILER THRUST_DEVICE_COMPILER_UNKNOWN
+#endif
+
+// is the device compiler capable of compiling omp?
+#if defined(_OPENMP) || defined(_NVHPC_STDPAR_OPENMP)
+#define THRUST_DEVICE_COMPILER_IS_OMP_CAPABLE THRUST_TRUE
+#else
+#define THRUST_DEVICE_COMPILER_IS_OMP_CAPABLE THRUST_FALSE
+#endif // _OPENMP
+
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC) && !defined(__CUDA_ARCH__)
+  #define THRUST_DISABLE_MSVC_WARNING_BEGIN(x)                                \
+    __pragma(warning(push))                                                   \
+    __pragma(warning(disable : x))                                            \
+    /**/
+  #define THRUST_DISABLE_MSVC_WARNING_END(x)                                  \
+    __pragma(warning(pop))                                                    \
+    /**/
+#else
+  #define THRUST_DISABLE_MSVC_WARNING_BEGIN(x)
+  #define THRUST_DISABLE_MSVC_WARNING_END(x)
+#endif
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG) && !defined(__CUDA_ARCH__)
+  #define THRUST_IGNORE_CLANG_WARNING_IMPL(x)                                 \
+    THRUST_PP_STRINGIZE(clang diagnostic ignored x)                           \
+    /**/
+  #define THRUST_IGNORE_CLANG_WARNING(x)                                      \
+    THRUST_IGNORE_CLANG_WARNING_IMPL(THRUST_PP_STRINGIZE(x))                  \
+    /**/
+
+  #define THRUST_DISABLE_CLANG_WARNING_BEGIN(x)                               \
+    _Pragma("clang diagnostic push")                                          \
+    _Pragma(THRUST_IGNORE_CLANG_WARNING(x))                                   \
+    /**/
+  #define THRUST_DISABLE_CLANG_WARNING_END(x)                                 \
+    _Pragma("clang diagnostic pop")                                           \
+    /**/
+#else
+  #define THRUST_DISABLE_CLANG_WARNING_BEGIN(x)
+  #define THRUST_DISABLE_CLANG_WARNING_END(x)
+#endif
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC) && !defined(__CUDA_ARCH__)
+  #define THRUST_IGNORE_GCC_WARNING_IMPL(x)                                   \
+    THRUST_PP_STRINGIZE(GCC diagnostic ignored x)                             \
+    /**/
+  #define THRUST_IGNORE_GCC_WARNING(x)                                        \
+    THRUST_IGNORE_GCC_WARNING_IMPL(THRUST_PP_STRINGIZE(x))                    \
+    /**/
+
+  #define THRUST_DISABLE_GCC_WARNING_BEGIN(x)                                 \
+    _Pragma("GCC diagnostic push")                                            \
+    _Pragma(THRUST_IGNORE_GCC_WARNING(x))                                     \
+    /**/
+  #define THRUST_DISABLE_GCC_WARNING_END(x)                                   \
+    _Pragma("GCC diagnostic pop")                                             \
+    /**/
+#else
+  #define THRUST_DISABLE_GCC_WARNING_BEGIN(x)
+  #define THRUST_DISABLE_GCC_WARNING_END(x)
+#endif
+
+#define THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_BEGIN               \
+  THRUST_DISABLE_MSVC_WARNING_BEGIN(4244 4267)                                \
+  /**/
+#define THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_END                 \
+  THRUST_DISABLE_MSVC_WARNING_END(4244 4267)                                  \
+  /**/
+#define THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING(x)                  \
+  THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_BEGIN                     \
+  x;                                                                          \
+  THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_END                       \
+  /**/
+
+#define THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING_BEGIN               \
+  THRUST_DISABLE_MSVC_WARNING_BEGIN(4800)                                     \
+  /**/
+#define THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING_END                 \
+  THRUST_DISABLE_MSVC_WARNING_END(4800)                                       \
+  /**/
+#define THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING(x)                  \
+  THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING_BEGIN                     \
+  x;                                                                          \
+  THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING_END                       \
+  /**/
+
+#define THRUST_DISABLE_CLANG_SELF_ASSIGNMENT_WARNING_BEGIN                    \
+  THRUST_DISABLE_CLANG_WARNING_BEGIN(-Wself-assign)                           \
+  /**/
+#define THRUST_DISABLE_CLANG_SELF_ASSIGNMENT_WARNING_END                      \
+  THRUST_DISABLE_CLANG_WARNING_END(-Wself-assign)                             \
+  /**/
+#define THRUST_DISABLE_CLANG_SELF_ASSIGNMENT_WARNING(x)                       \
+  THRUST_DISABLE_CLANG_SELF_ASSIGNMENT_WARNING_BEGIN                          \
+  x;                                                                          \
+  THRUST_DISABLE_CLANG_SELF_ASSIGNMENT_WARNING_END                            \
+  /**/
+
+#define THRUST_DISABLE_CLANG_AND_GCC_INITIALIZER_REORDERING_WARNING_BEGIN     \
+  THRUST_DISABLE_CLANG_WARNING_BEGIN(-Wreorder)                               \
+  THRUST_DISABLE_GCC_WARNING_BEGIN(-Wreorder)                                 \
+  /**/
+#define THRUST_DISABLE_CLANG_AND_GCC_INITIALIZER_REORDERING_WARNING_END       \
+  THRUST_DISABLE_CLANG_WARNING_END(-Wreorder)                                 \
+  THRUST_DISABLE_GCC_WARNING_END(-Wreorder)                                   \
+  /**/
+#define THRUST_DISABLE_CLANG_AND_GCC_INITIALIZER_REORDERING_WARNING(x)        \
+  THRUST_DISABLE_CLANG_AND_GCC_INITIALIZER_REORDERING_WARNING_BEGIN           \
+  x;                                                                          \
+  THRUST_DISABLE_CLANG_AND_GCC_INITIALIZER_REORDERING_WARNING_END             \
+  /**/
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler_fence.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler_fence.h
new file mode 100644
index 000000000000..c379abaf364b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/compiler_fence.h
@@ -0,0 +1,62 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/preprocessor.h>
+
+// TODO: Enable this or remove this file once nvGRAPH/CUSP migrates off of it.
+//#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+//  #pragma message("warning: The functionality in this header is unsafe, deprecated, and will soon be removed. Use C++11 or C11 atomics instead.")
+//#else
+//  #warning The functionality in this header is unsafe, deprecated, and will soon be removed. Use C++11 or C11 atomics instead.
+//#endif
+
+// msvc case
+#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+
+#ifndef _DEBUG
+
+#include <intrin.h>
+#pragma intrinsic(_ReadWriteBarrier)
+#define __thrust_compiler_fence() _ReadWriteBarrier()
+#else
+
+#define __thrust_compiler_fence() do {} while (0)
+
+#endif // _DEBUG
+
+// gcc case
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC
+
+#if THRUST_GCC_VERSION >= 40200 // atomic built-ins were introduced ~4.2
+#define __thrust_compiler_fence() __sync_synchronize()
+#else
+// allow the code to compile without any guarantees
+#define __thrust_compiler_fence() do {} while (0)
+#endif // THRUST_GCC_VERSION
+
+// unknown case
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG
+#define __thrust_compiler_fence() __sync_synchronize()
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_UNKNOWN
+
+// allow the code to compile without any guarantees
+#define __thrust_compiler_fence() do {} while (0)
+
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/config.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/config.h
new file mode 100644
index 000000000000..797f6605b082
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/config.h
@@ -0,0 +1,40 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file config.h
+ *  \brief Defines platform configuration.
+ */
+
+#pragma once
+
+// NOTE: The order of these #includes matters.
+
+#include <thrust/detail/config/simple_defines.h>
+#include <thrust/detail/config/compiler.h>
+#include <thrust/detail/config/cpp_dialect.h>
+#include <thrust/detail/config/cpp_compatibility.h>
+#include <thrust/detail/config/deprecated.h>
+// host_system.h & device_system.h must be #included as early as possible
+// because other config headers depend on it
+#include <thrust/detail/config/host_system.h>
+#include <thrust/detail/config/device_system.h>
+#include <thrust/detail/config/host_device.h>
+#include <thrust/detail/config/debug.h>
+#include <thrust/detail/config/forceinline.h>
+#include <thrust/detail/config/exec_check_disable.h>
+#include <thrust/detail/config/global_workarounds.h>
+#include <thrust/detail/config/namespace.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_compatibility.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_compatibility.h
new file mode 100644
index 000000000000..18b9cbdcf599
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_compatibility.h
@@ -0,0 +1,101 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config/cpp_dialect.h>
+
+#include <cstddef>
+
+#ifndef __has_cpp_attribute
+#  define __has_cpp_attribute(X) 0
+#endif
+
+// Trailing return types seem to confuse Doxygen, and cause it to interpret
+// parts of the function's body as new function signatures.
+#if defined(THRUST_DOXYGEN)
+#  define THRUST_TRAILING_RETURN(...)
+#else
+#  define THRUST_TRAILING_RETURN(...) -> __VA_ARGS__
+#endif
+
+#if THRUST_CPP_DIALECT >= 2014 && __has_cpp_attribute(nodiscard)
+#  define THRUST_NODISCARD [[nodiscard]]
+#else
+#  define THRUST_NODISCARD
+#endif
+
+#if THRUST_CPP_DIALECT >= 2017 && __cpp_if_constexpr
+#  define THRUST_IF_CONSTEXPR if constexpr
+#else
+#  define THRUST_IF_CONSTEXPR if
+#endif
+
+// FIXME: Combine THRUST_INLINE_CONSTANT and
+// THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT into one macro when NVCC properly
+// supports `constexpr` globals in host and device code.
+#if defined(__CUDA_ARCH__) || defined(_NVHPC_CUDA)
+// FIXME: Add this when NVCC supports inline variables.
+//#  if   THRUST_CPP_DIALECT >= 2017
+//#    define THRUST_INLINE_CONSTANT                 inline constexpr
+//#    define THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT inline constexpr
+#  if THRUST_CPP_DIALECT >= 2011
+#    define THRUST_INLINE_CONSTANT                 static const __device__
+#    define THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT static constexpr
+#  else
+#    define THRUST_INLINE_CONSTANT                 static const __device__
+#    define THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT static const
+#  endif
+#else
+// FIXME: Add this when NVCC supports inline variables.
+//#  if   THRUST_CPP_DIALECT >= 2017
+//#    define THRUST_INLINE_CONSTANT                 inline constexpr
+//#    define THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT inline constexpr
+#  if THRUST_CPP_DIALECT >= 2011
+#    define THRUST_INLINE_CONSTANT                 static constexpr
+#    define THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT static constexpr
+#  else
+#    define THRUST_INLINE_CONSTANT                 static const
+#    define THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT static const
+#  endif
+#endif
+
+// These definitions were intended for internal use only and are now obsolete.
+// If you relied on them, consider porting your code to use the functionality
+// in libcu++'s <nv/target> header.
+// For a temporary workaround, define THRUST_PROVIDE_LEGACY_ARCH_MACROS to make
+// them available again. These should be considered deprecated and will be
+// fully removed in a future version.
+#ifdef THRUST_PROVIDE_LEGACY_ARCH_MACROS
+  #ifndef THRUST_IS_DEVICE_CODE
+    #if defined(_NVHPC_CUDA)
+      #define THRUST_IS_DEVICE_CODE __builtin_is_device_code()
+      #define THRUST_IS_HOST_CODE (!__builtin_is_device_code())
+      #define THRUST_INCLUDE_DEVICE_CODE 1
+      #define THRUST_INCLUDE_HOST_CODE 1
+    #elif defined(__CUDA_ARCH__)
+      #define THRUST_IS_DEVICE_CODE 1
+      #define THRUST_IS_HOST_CODE 0
+      #define THRUST_INCLUDE_DEVICE_CODE 1
+      #define THRUST_INCLUDE_HOST_CODE 0
+    #else
+      #define THRUST_IS_DEVICE_CODE 0
+      #define THRUST_IS_HOST_CODE 1
+      #define THRUST_INCLUDE_DEVICE_CODE 0
+      #define THRUST_INCLUDE_HOST_CODE 1
+    #endif
+  #endif
+#endif // THRUST_PROVIDE_LEGACY_ARCH_MACROS
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_dialect.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_dialect.h
new file mode 100644
index 000000000000..46b0caec7803
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/cpp_dialect.h
@@ -0,0 +1,140 @@
+/*
+ *  Copyright 2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file cpp_dialect.h
+ *  \brief Detect the version of the C++ standard used by the compiler.
+ */
+
+#pragma once
+
+#include <thrust/detail/config/compiler.h>
+
+// Deprecation warnings may be silenced by defining the following macros. These
+// may be combined.
+// - THRUST_IGNORE_DEPRECATED_CPP_DIALECT:
+//   Ignore all deprecated C++ dialects and outdated compilers.
+// - THRUST_IGNORE_DEPRECATED_CPP_11:
+//   Ignore deprecation warnings when compiling with C++11. C++03 and outdated
+//   compilers will still issue warnings.
+// - THRUST_IGNORE_DEPRECATED_COMPILER
+//   Ignore deprecation warnings when using deprecated compilers. Compiling
+//   with C++03 and C++11 will still issue warnings.
+
+// Check for the CUB opt-outs as well:
+#if !defined(THRUST_IGNORE_DEPRECATED_CPP_DIALECT) && \
+     defined(CUB_IGNORE_DEPRECATED_CPP_DIALECT)
+#  define    THRUST_IGNORE_DEPRECATED_CPP_DIALECT
+#endif
+#if !defined(THRUST_IGNORE_DEPRECATED_CPP_11) && \
+     defined(CUB_IGNORE_DEPRECATED_CPP_11)
+#  define    THRUST_IGNORE_DEPRECATED_CPP_11
+#endif
+#if !defined(THRUST_IGNORE_DEPRECATED_COMPILER) && \
+     defined(CUB_IGNORE_DEPRECATED_COMPILER)
+#  define    THRUST_IGNORE_DEPRECATED_COMPILER
+#endif
+
+#ifdef THRUST_IGNORE_DEPRECATED_CPP_DIALECT
+#  define THRUST_IGNORE_DEPRECATED_CPP_11
+#  define THRUST_IGNORE_DEPRECATED_COMPILER
+#endif
+
+// Define this to override the built-in detection.
+#ifndef THRUST_CPP_DIALECT
+
+// MSVC does not define __cplusplus correctly. _MSVC_LANG is used instead.
+// This macro is only defined in MSVC 2015U3+.
+#  ifdef _MSVC_LANG // Do not replace with THRUST_HOST_COMPILER test (see above)
+// MSVC2015 reports C++14 but lacks extended constexpr support. Treat as C++11.
+#    if THRUST_MSVC_VERSION < 1910 && _MSVC_LANG > 201103L /* MSVC < 2017 && CPP > 2011 */
+#      define THRUST_CPLUSPLUS 201103L /* Fix to 2011 */
+#    else
+#      define THRUST_CPLUSPLUS _MSVC_LANG /* We'll trust this for now. */
+#    endif // MSVC 2015 C++14 fix
+#  else
+#    define THRUST_CPLUSPLUS __cplusplus
+#  endif
+
+// Detect current dialect:
+#  if THRUST_CPLUSPLUS < 201103L
+#    define THRUST_CPP_DIALECT 2003
+#  elif THRUST_CPLUSPLUS < 201402L
+#    define THRUST_CPP_DIALECT 2011
+#  elif THRUST_CPLUSPLUS < 201703L
+#    define THRUST_CPP_DIALECT 2014
+#  elif THRUST_CPLUSPLUS == 201703L
+#    define THRUST_CPP_DIALECT 2017
+#  elif THRUST_CPLUSPLUS > 201703L // unknown, but is higher than 2017.
+#    define THRUST_CPP_DIALECT 2020
+#  endif
+
+#  undef THRUST_CPLUSPLUS // cleanup
+
+#endif // !THRUST_CPP_DIALECT
+
+// Define THRUST_COMPILER_DEPRECATION macro:
+#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+#  define THRUST_COMP_DEPR_IMPL(msg) \
+    __pragma(message(__FILE__ ":" THRUST_COMP_DEPR_IMPL0(__LINE__) ": warning: " #msg))
+#  define THRUST_COMP_DEPR_IMPL0(x) THRUST_COMP_DEPR_IMPL1(x)
+#  define THRUST_COMP_DEPR_IMPL1(x) #x
+#else // clang / gcc:
+#  define THRUST_COMP_DEPR_IMPL(msg) THRUST_COMP_DEPR_IMPL0(GCC warning #msg)
+#  define THRUST_COMP_DEPR_IMPL0(expr) _Pragma(#expr)
+#  define THRUST_COMP_DEPR_IMPL1 /* intentionally blank */
+#endif
+
+#define THRUST_COMPILER_DEPRECATION(REQ) \
+  THRUST_COMP_DEPR_IMPL(Thrust requires at least REQ. Define THRUST_IGNORE_DEPRECATED_CPP_DIALECT to suppress this message.)
+
+#define THRUST_COMPILER_DEPRECATION_SOFT(REQ, CUR) \
+  THRUST_COMP_DEPR_IMPL(Thrust requires at least REQ. CUR is deprecated but still supported. CUR support will be removed in a future release. Define THRUST_IGNORE_DEPRECATED_CPP_DIALECT to suppress this message.)
+
+#ifndef THRUST_IGNORE_DEPRECATED_COMPILER
+
+// Compiler checks:
+#  if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC && THRUST_GCC_VERSION < 50000
+     THRUST_COMPILER_DEPRECATION(GCC 5.0);
+#  elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG && THRUST_CLANG_VERSION < 70000
+     THRUST_COMPILER_DEPRECATION(Clang 7.0);
+#  elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC && THRUST_MSVC_VERSION < 1910
+     // <2017. Hard upgrade message:
+     THRUST_COMPILER_DEPRECATION(MSVC 2019 (19.20/16.0/14.20));
+#  elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC && THRUST_MSVC_VERSION < 1920
+     // >=2017, <2019. Soft deprecation message:
+     THRUST_COMPILER_DEPRECATION_SOFT(MSVC 2019 (19.20/16.0/14.20), MSVC 2017);
+#  endif
+
+#endif // THRUST_IGNORE_DEPRECATED_COMPILER
+
+#ifndef THRUST_IGNORE_DEPRECATED_DIALECT
+
+// Dialect checks:
+#  if THRUST_CPP_DIALECT < 2011
+     // <C++11. Hard upgrade message:
+     THRUST_COMPILER_DEPRECATION(C++14);
+#  elif THRUST_CPP_DIALECT == 2011 && !defined(THRUST_IGNORE_DEPRECATED_CPP_11)
+     // =C++11. Soft upgrade message:
+     THRUST_COMPILER_DEPRECATION_SOFT(C++14, C++11);
+#  endif
+
+#endif // THRUST_IGNORE_DEPRECATED_DIALECT
+
+#undef THRUST_COMPILER_DEPRECATION_SOFT
+#undef THRUST_COMPILER_DEPRECATION
+#undef THRUST_COMP_DEPR_IMPL
+#undef THRUST_COMP_DEPR_IMPL0
+#undef THRUST_COMP_DEPR_IMPL1
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/debug.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/debug.h
new file mode 100644
index 000000000000..16f65d67c905
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/debug.h
@@ -0,0 +1,32 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#ifndef THRUST_DEBUG
+#  ifndef NDEBUG
+#    if defined(DEBUG) || defined(_DEBUG)
+#      define THRUST_DEBUG 1
+#    endif // (DEBUG || _DEBUG)
+#  endif // NDEBUG
+#endif // THRUST_DEBUG
+
+#if THRUST_DEBUG
+#  ifndef __THRUST_SYNCHRONOUS
+#    define __THRUST_SYNCHRONOUS 1
+#  endif // __THRUST_SYNCHRONOUS
+#endif // THRUST_DEBUG
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/deprecated.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/deprecated.h
new file mode 100644
index 000000000000..05851c6768fc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/deprecated.h
@@ -0,0 +1,42 @@
+/*
+ *  Copyright 2018-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file deprecated.h
+ *  \brief Defines the THRUST_DEPRECATED macro
+ */
+
+#pragma once
+
+#include <thrust/detail/config/compiler.h>
+#include <thrust/detail/config/cpp_dialect.h>
+
+#if defined(CUB_IGNORE_DEPRECATED_API) && !defined(THRUST_IGNORE_DEPRECATED_API)
+#  define THRUST_IGNORE_DEPRECATED_API
+#endif
+
+#ifdef THRUST_IGNORE_DEPRECATED_API
+#  define THRUST_DEPRECATED
+#elif THRUST_CPP_DIALECT >= 2014
+#  define THRUST_DEPRECATED [[deprecated]]
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+#  define THRUST_DEPRECATED __declspec(deprecated)
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG
+#  define THRUST_DEPRECATED __attribute__((deprecated))
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC
+#  define THRUST_DEPRECATED __attribute__((deprecated))
+#else
+#  define THRUST_DEPRECATED
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/device_system.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/device_system.h
new file mode 100644
index 000000000000..29418c90308a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/device_system.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+// reserve 0 for undefined
+#define THRUST_DEVICE_SYSTEM_CUDA    1
+#define THRUST_DEVICE_SYSTEM_OMP     2
+#define THRUST_DEVICE_SYSTEM_TBB     3
+#define THRUST_DEVICE_SYSTEM_CPP     4
+
+#ifndef THRUST_DEVICE_SYSTEM
+#define THRUST_DEVICE_SYSTEM THRUST_DEVICE_SYSTEM_CUDA
+#endif // THRUST_DEVICE_SYSTEM
+
+#ifdef THRUST_DEVICE_BACKEND
+#  error THRUST_DEVICE_BACKEND is no longer supported; use THRUST_DEVICE_SYSTEM instead.
+#endif // THRUST_DEVICE_BACKEND
+
+#if THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
+#define __THRUST_DEVICE_SYSTEM_NAMESPACE cuda
+#elif THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_OMP
+#define __THRUST_DEVICE_SYSTEM_NAMESPACE omp
+#elif THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_TBB
+#define __THRUST_DEVICE_SYSTEM_NAMESPACE tbb
+#elif THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CPP
+#define __THRUST_DEVICE_SYSTEM_NAMESPACE cpp
+#endif
+
+#define __THRUST_DEVICE_SYSTEM_ROOT thrust/system/__THRUST_DEVICE_SYSTEM_NAMESPACE
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/exec_check_disable.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/exec_check_disable.h
new file mode 100644
index 000000000000..9b25b375dfbf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/exec_check_disable.h
@@ -0,0 +1,43 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file exec_check_disable.h
+ *  \brief Defines __thrust_exec_check_disable__
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #pragma nv_exec_check_disable is only recognized by NVCC.  Having a macro
+// expand to a #pragma (rather than _Pragma) only works with NVCC's compilation
+// model, not with other compilers.
+#if defined(__CUDACC__) && !defined(_NVHPC_CUDA) && \
+    !(defined(__CUDA__) && defined(__clang__))
+
+#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+#define __thrust_exec_check_disable__ __pragma("nv_exec_check_disable")
+#else // MSVC
+#define __thrust_exec_check_disable__ _Pragma("nv_exec_check_disable")
+#endif // MSVC
+
+#else
+
+#define __thrust_exec_check_disable__
+
+#endif
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/forceinline.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/forceinline.h
new file mode 100644
index 000000000000..b001fd4b1148
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/forceinline.h
@@ -0,0 +1,36 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file forceinline.h
+ *  \brief Defines __thrust_forceinline__
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#if defined(__CUDACC__) || defined(_NVHPC_CUDA)
+
+#define __thrust_forceinline__ __forceinline__
+
+#else
+
+// TODO add 
+
+#define __thrust_forceinline__
+
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/global_workarounds.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/global_workarounds.h
new file mode 100644
index 000000000000..9800f03593ac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/global_workarounds.h
@@ -0,0 +1,27 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config/compiler.h>
+
+// XXX workaround gcc 4.8+'s complaints about unused local typedefs by silencing them globally
+#if defined(THRUST_GCC_VERSION) && (THRUST_GCC_VERSION >= 40800)
+#  if defined(__NVCC__) && (CUDART_VERSION >= 6000)
+#    pragma GCC diagnostic ignored "-Wunused-local-typedefs"
+#  endif // nvcc & cuda 6+
+#endif // gcc 4.8
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_device.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_device.h
new file mode 100644
index 000000000000..5540f91260d8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_device.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file host_device.h
+ *  \brief Defines __host__ and __device__
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// since nvcc defines __host__ and __device__ for us,
+// and only nvcc knows what to do with __host__ and __device__,
+// define them to be the empty string for other compilers
+
+#if THRUST_DEVICE_COMPILER != THRUST_DEVICE_COMPILER_NVCC
+
+// since __host__ & __device__ might have already be defined, only
+// #define them if not defined already
+// XXX this will break if the client does #include <host_defines.h> later
+
+#ifndef __host__
+#define __host__
+#endif // __host__
+
+#ifndef __device__
+#define __device__
+#endif // __device__
+
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_system.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_system.h
new file mode 100644
index 000000000000..f216f649237b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/host_system.h
@@ -0,0 +1,41 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+// reserve 0 for undefined
+#define THRUST_HOST_SYSTEM_CPP    1
+#define THRUST_HOST_SYSTEM_OMP    2
+#define THRUST_HOST_SYSTEM_TBB    3
+
+#ifndef THRUST_HOST_SYSTEM
+#define THRUST_HOST_SYSTEM THRUST_HOST_SYSTEM_CPP
+#endif // THRUST_HOST_SYSTEM
+
+#ifdef THRUST_HOST_BACKEND
+#  error THRUST_HOST_BACKEND is no longer supported; use THRUST_HOST_SYSTEM instead.
+#endif // THRUST_HOST_BACKEND
+
+#if THRUST_HOST_SYSTEM == THRUST_HOST_SYSTEM_CPP
+#define __THRUST_HOST_SYSTEM_NAMESPACE cpp
+#elif THRUST_HOST_SYSTEM == THRUST_HOST_SYSTEM_OMP
+#define __THRUST_HOST_SYSTEM_NAMESPACE omp
+#elif THRUST_HOST_SYSTEM == THRUST_HOST_SYSTEM_TBB
+#define __THRUST_HOST_SYSTEM_NAMESPACE tbb
+#endif
+
+#define __THRUST_HOST_SYSTEM_ROOT thrust/system/__THRUST_HOST_SYSTEM_NAMESPACE
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/memory_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/memory_resource.h
new file mode 100644
index 000000000000..ab719c9bde9b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/memory_resource.h
@@ -0,0 +1,35 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <cstddef>
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/alignment.h>
+#include <thrust/detail/config/cpp_compatibility.h>
+
+#define THRUST_MR_DEFAULT_ALIGNMENT THRUST_ALIGNOF(THRUST_NS_QUALIFIER::detail::max_align_t)
+
+#if THRUST_CPP_DIALECT >= 2017
+#  if __has_include(<memory_resource>)
+#    define THRUST_MR_STD_MR_HEADER <memory_resource>
+#    define THRUST_MR_STD_MR_NS std::pmr
+#  elif __has_include(<experimental/memory_resource>)
+#    define THRUST_MR_STD_MR_HEADER <experimental/memory_resource>
+#    define THRUST_MR_STD_MR_NS std::experimental::pmr
+#  endif
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/namespace.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/namespace.h
new file mode 100644
index 000000000000..9c7904616935
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/namespace.h
@@ -0,0 +1,120 @@
+/*
+ *  Copyright 2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+/**
+ * \file namespace.h
+ * \brief Utilities that allow `thrust::` to be placed inside an
+ * application-specific namespace.
+ */
+
+/**
+ * \def THRUST_CUB_WRAPPED_NAMESPACE
+ * If defined, this value will be used as the name of a namespace that wraps the
+ * `thrust::` and `cub::` namespaces.
+ * This macro should not be used with any other Thrust namespace macros.
+ */
+#ifdef THRUST_CUB_WRAPPED_NAMESPACE
+#define THRUST_WRAPPED_NAMESPACE THRUST_CUB_WRAPPED_NAMESPACE
+#endif
+
+/**
+ * \def THRUST_WRAPPED_NAMESPACE
+ * If defined, this value will be used as the name of a namespace that wraps the
+ * `thrust::` namespace.
+ * If THRUST_CUB_WRAPPED_NAMESPACE is set, this will inherit that macro's value.
+ * This macro should not be used with any other Thrust namespace macros.
+ */
+#ifdef THRUST_WRAPPED_NAMESPACE
+#define THRUST_NS_PREFIX                                                       \
+  namespace THRUST_WRAPPED_NAMESPACE                                           \
+  {
+
+#define THRUST_NS_POSTFIX }
+
+#define THRUST_NS_QUALIFIER ::THRUST_WRAPPED_NAMESPACE::thrust
+#endif
+
+/**
+ * \def THRUST_NS_PREFIX
+ * This macro is inserted prior to all `namespace thrust { ... }` blocks. It is
+ * derived from THRUST_WRAPPED_NAMESPACE, if set, and will be empty otherwise.
+ * It may be defined by users, in which case THRUST_NS_PREFIX,
+ * THRUST_NS_POSTFIX, and THRUST_NS_QUALIFIER must all be set consistently.
+ */
+#ifndef THRUST_NS_PREFIX
+#define THRUST_NS_PREFIX
+#endif
+
+/**
+ * \def THRUST_NS_POSTFIX
+ * This macro is inserted following the closing braces of all
+ * `namespace thrust { ... }` block. It is defined appropriately when
+ * THRUST_WRAPPED_NAMESPACE is set, and will be empty otherwise. It may be
+ * defined by users, in which case THRUST_NS_PREFIX, THRUST_NS_POSTFIX, and
+ * THRUST_NS_QUALIFIER must all be set consistently.
+ */
+#ifndef THRUST_NS_POSTFIX
+#define THRUST_NS_POSTFIX
+#endif
+
+/**
+ * \def THRUST_NS_QUALIFIER
+ * This macro is used to qualify members of thrust:: when accessing them from
+ * outside of their namespace. By default, this is just `::thrust`, and will be
+ * set appropriately when THRUST_WRAPPED_NAMESPACE is defined. This macro may be
+ * defined by users, in which case THRUST_NS_PREFIX, THRUST_NS_POSTFIX, and
+ * THRUST_NS_QUALIFIER must all be set consistently.
+ */
+#ifndef THRUST_NS_QUALIFIER
+#define THRUST_NS_QUALIFIER ::thrust
+#endif
+
+/**
+ * \def THRUST_NAMESPACE_BEGIN
+ * This macro is used to open a `thrust::` namespace block, along with any
+ * enclosing namespaces requested by THRUST_WRAPPED_NAMESPACE, etc.
+ * This macro is defined by Thrust and may not be overridden.
+ */
+#define THRUST_NAMESPACE_BEGIN                                                 \
+  THRUST_NS_PREFIX                                                             \
+  namespace thrust                                                             \
+  {
+
+/**
+ * \def THRUST_NAMESPACE_END
+ * This macro is used to close a `thrust::` namespace block, along with any
+ * enclosing namespaces requested by THRUST_WRAPPED_NAMESPACE, etc.
+ * This macro is defined by Thrust and may not be overridden.
+ */
+#define THRUST_NAMESPACE_END                                                   \
+  } /* end namespace thrust */                                                 \
+  THRUST_NS_POSTFIX
+
+// The following is just here to add docs for the thrust namespace:
+
+THRUST_NS_PREFIX
+
+/*! \namespace thrust
+ *  \brief \p thrust is the top-level namespace which contains all Thrust
+ *         functions and types.
+ */
+namespace thrust
+{
+}
+
+THRUST_NS_POSTFIX
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/simple_defines.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/simple_defines.h
new file mode 100644
index 000000000000..e3ea2eb64e76
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/config/simple_defines.h
@@ -0,0 +1,30 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file simple_defines.h
+ *  \brief Primitive macros without dependencies.
+ */
+
+#pragma once
+
+#define THRUST_UNKNOWN 0
+#define THRUST_FALSE   0
+#define THRUST_TRUE    1
+
+#define THRUST_UNUSED_VAR(expr) do { (void)(expr); } while (0)
+
+#define THRUST_PREVENT_MACRO_SUBSTITUTION
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.h
new file mode 100644
index 000000000000..536c1c27c2e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.h
@@ -0,0 +1,235 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/iterator/detail/normal_iterator.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+struct copy_allocator_t {};
+
+// XXX parameter T is redundant with parameter Alloc
+template<typename T, typename Alloc>
+  class contiguous_storage
+{
+  private:
+    typedef thrust::detail::allocator_traits<Alloc> alloc_traits;
+
+  public:
+    typedef Alloc                                      allocator_type;
+    typedef T                                          value_type;
+    typedef typename alloc_traits::pointer             pointer;
+    typedef typename alloc_traits::const_pointer       const_pointer;
+    typedef typename alloc_traits::size_type           size_type;
+    typedef typename alloc_traits::difference_type     difference_type;
+    typedef typename alloc_traits::reference           reference;
+    typedef typename alloc_traits::const_reference     const_reference;
+
+    typedef thrust::detail::normal_iterator<pointer>       iterator;
+    typedef thrust::detail::normal_iterator<const_pointer> const_iterator;
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    explicit contiguous_storage(const allocator_type &alloc = allocator_type());
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    explicit contiguous_storage(size_type n, const allocator_type &alloc = allocator_type());
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    explicit contiguous_storage(copy_allocator_t, const contiguous_storage &other);
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    explicit contiguous_storage(copy_allocator_t, const contiguous_storage &other, size_type n);
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    ~contiguous_storage();
+
+    __host__ __device__
+    size_type size() const;
+
+    __host__ __device__
+    size_type max_size() const;
+
+    __host__ __device__
+    pointer data();
+
+    __host__ __device__
+    const_pointer data() const;
+
+    __host__ __device__
+    iterator begin();
+
+    __host__ __device__
+    const_iterator begin() const;
+
+    __host__ __device__
+    iterator end();
+
+    __host__ __device__
+    const_iterator end() const;
+
+    __host__ __device__
+    reference operator[](size_type n);
+
+    __host__ __device__
+    const_reference operator[](size_type n) const;
+
+    __host__ __device__
+    allocator_type get_allocator() const;
+
+    // note that allocate does *not* automatically call deallocate
+    __host__ __device__
+    void allocate(size_type n);
+
+    __host__ __device__
+    void deallocate();
+
+    __host__ __device__
+    void swap(contiguous_storage &x);
+
+    __host__ __device__
+    void default_construct_n(iterator first, size_type n);
+
+    __host__ __device__
+    void uninitialized_fill_n(iterator first, size_type n, const value_type &value);
+
+    template<typename InputIterator>
+    __host__ __device__
+    iterator uninitialized_copy(InputIterator first, InputIterator last, iterator result);
+
+    template<typename System, typename InputIterator>
+    __host__ __device__
+    iterator uninitialized_copy(thrust::execution_policy<System> &from_system,
+                                InputIterator first,
+                                InputIterator last,
+                                iterator result);
+
+    template<typename InputIterator, typename Size>
+    __host__ __device__
+    iterator uninitialized_copy_n(InputIterator first, Size n, iterator result);
+
+    template<typename System, typename InputIterator, typename Size>
+    __host__ __device__
+    iterator uninitialized_copy_n(thrust::execution_policy<System> &from_system,
+                                  InputIterator first,
+                                  Size n,
+                                  iterator result);
+
+    __host__ __device__
+    void destroy(iterator first, iterator last);
+
+    __host__ __device__
+    void deallocate_on_allocator_mismatch(const contiguous_storage &other);
+
+    __host__ __device__
+    void destroy_on_allocator_mismatch(const contiguous_storage &other,
+        iterator first, iterator last);
+
+    __host__ __device__
+    void set_allocator(const allocator_type &alloc);
+
+    __host__ __device__
+    bool is_allocator_not_equal(const allocator_type &alloc) const;
+
+    __host__ __device__
+    bool is_allocator_not_equal(const contiguous_storage &other) const;
+
+    __host__ __device__
+    void propagate_allocator(const contiguous_storage &other);
+
+#if THRUST_CPP_DIALECT >= 2011
+    __host__ __device__
+    void propagate_allocator(contiguous_storage &other);
+
+    // allow move assignment for a sane implementation of allocator propagation
+    // on move assignment
+    __host__ __device__
+    contiguous_storage &operator=(contiguous_storage &&other);
+#endif
+
+  private:
+    // XXX we could inherit from this to take advantage of empty base class optimization
+    allocator_type m_allocator;
+
+    iterator m_begin;
+
+    size_type m_size;
+
+    // disallow assignment
+    contiguous_storage &operator=(const contiguous_storage &x);
+
+    __host__ __device__
+    void swap_allocators(true_type, const allocator_type &);
+
+    __host__ __device__
+    void swap_allocators(false_type, allocator_type &);
+
+    __host__ __device__
+    bool is_allocator_not_equal_dispatch(true_type, const allocator_type &) const;
+
+    __host__ __device__
+    bool is_allocator_not_equal_dispatch(false_type, const allocator_type &) const;
+
+    __host__ __device__
+    void deallocate_on_allocator_mismatch_dispatch(true_type, const contiguous_storage &other);
+
+    __host__ __device__
+    void deallocate_on_allocator_mismatch_dispatch(false_type, const contiguous_storage &other);
+
+    __host__ __device__
+    void destroy_on_allocator_mismatch_dispatch(true_type, const contiguous_storage &other,
+        iterator first, iterator last);
+
+    __host__ __device__
+    void destroy_on_allocator_mismatch_dispatch(false_type, const contiguous_storage &other,
+        iterator first, iterator last);
+
+    __host__ __device__
+    void propagate_allocator_dispatch(true_type, const contiguous_storage &other);
+
+    __host__ __device__
+    void propagate_allocator_dispatch(false_type, const contiguous_storage &other);
+
+#if THRUST_CPP_DIALECT >= 2011
+    __host__ __device__
+    void propagate_allocator_dispatch(true_type, contiguous_storage &other);
+
+    __host__ __device__
+    void propagate_allocator_dispatch(false_type, contiguous_storage &other);
+#endif
+}; // end contiguous_storage
+
+} // end detail
+
+template<typename T, typename Alloc>
+__host__ __device__
+void swap(detail::contiguous_storage<T,Alloc> &lhs, detail::contiguous_storage<T,Alloc> &rhs);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/contiguous_storage.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.inl
new file mode 100644
index 000000000000..7ae8657f0811
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/contiguous_storage.inl
@@ -0,0 +1,550 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/contiguous_storage.h>
+#include <thrust/detail/swap.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/allocator/copy_construct_range.h>
+#include <thrust/detail/allocator/default_construct_range.h>
+#include <thrust/detail/allocator/destroy_range.h>
+#include <thrust/detail/allocator/fill_construct_range.h>
+
+#include <nv/target>
+
+#include <stdexcept> // for std::runtime_error
+#include <utility> // for use of std::swap in the WAR below
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+class allocator_mismatch_on_swap : public std::runtime_error
+{
+public:
+  allocator_mismatch_on_swap()
+    :std::runtime_error("swap called on containers with allocators that propagate on swap, but compare non-equal")
+  {
+  }
+};
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  contiguous_storage<T,Alloc>
+    ::contiguous_storage(const Alloc &alloc)
+      :m_allocator(alloc),
+       m_begin(pointer(static_cast<T*>(0))),
+       m_size(0)
+{
+  ;
+} // end contiguous_storage::contiguous_storage()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  contiguous_storage<T,Alloc>
+    ::contiguous_storage(size_type n, const Alloc &alloc)
+      :m_allocator(alloc),
+       m_begin(pointer(static_cast<T*>(0))),
+       m_size(0)
+{
+  allocate(n);
+} // end contiguous_storage::contiguous_storage()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  contiguous_storage<T,Alloc>
+    ::contiguous_storage(copy_allocator_t,
+        const contiguous_storage &other)
+      :m_allocator(other.m_allocator),
+       m_begin(pointer(static_cast<T*>(0))),
+       m_size(0)
+{
+} // end contiguous_storage::contiguous_storage()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  contiguous_storage<T,Alloc>
+    ::contiguous_storage(copy_allocator_t,
+        const contiguous_storage &other, size_type n)
+      :m_allocator(other.m_allocator),
+       m_begin(pointer(static_cast<T*>(0))),
+       m_size(0)
+{
+  allocate(n);
+} // end contiguous_storage::contiguous_storage()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  contiguous_storage<T,Alloc>
+    ::~contiguous_storage()
+{
+  deallocate();
+} // end contiguous_storage::~contiguous_storage()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::size_type
+    contiguous_storage<T,Alloc>
+      ::size() const
+{
+  return m_size;
+} // end contiguous_storage::size()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::size_type
+    contiguous_storage<T,Alloc>
+      ::max_size() const
+{
+  return alloc_traits::max_size(m_allocator);
+} // end contiguous_storage::max_size()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::iterator
+    contiguous_storage<T,Alloc>
+      ::begin()
+{
+  return m_begin;
+} // end contiguous_storage::begin()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::const_iterator
+    contiguous_storage<T,Alloc>
+      ::begin() const
+{
+  return m_begin;
+} // end contiguous_storage::begin()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::iterator
+    contiguous_storage<T,Alloc>
+      ::end()
+{
+  return m_begin + size();
+} // end contiguous_storage::end()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::const_iterator
+    contiguous_storage<T,Alloc>
+      ::end() const
+{
+  return m_begin + size();
+} // end contiguous_storage::end()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::pointer
+    contiguous_storage<T,Alloc>
+      ::data()
+{
+  return &*m_begin;
+} // end contiguous_storage::data()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::const_pointer
+    contiguous_storage<T,Alloc>
+      ::data() const
+{
+  return &*m_begin;
+} // end contiguous_storage::data()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::reference
+    contiguous_storage<T,Alloc>
+      ::operator[](size_type n)
+{
+  return m_begin[n];
+} // end contiguous_storage::operator[]()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::const_reference
+    contiguous_storage<T,Alloc>
+      ::operator[](size_type n) const
+{
+  return m_begin[n];
+} // end contiguous_storage::operator[]()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  typename contiguous_storage<T,Alloc>::allocator_type
+    contiguous_storage<T,Alloc>
+      ::get_allocator() const
+{
+  return m_allocator;
+} // end contiguous_storage::get_allocator()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::allocate(size_type n)
+{
+  if(n > 0)
+  {
+    m_begin = iterator(alloc_traits::allocate(m_allocator,n));
+    m_size = n;
+  } // end if
+  else
+  {
+    m_begin = iterator(pointer(static_cast<T*>(0)));
+    m_size = 0;
+  } // end else
+} // end contiguous_storage::allocate()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::deallocate()
+{
+  if(size() > 0)
+  {
+    alloc_traits::deallocate(m_allocator,m_begin.base(), size());
+    m_begin = iterator(pointer(static_cast<T*>(0)));
+    m_size = 0;
+  } // end if
+} // end contiguous_storage::deallocate()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::swap(contiguous_storage &x)
+{
+  thrust::swap(m_begin, x.m_begin);
+  thrust::swap(m_size, x.m_size);
+
+  swap_allocators(
+    integral_constant<
+      bool,
+      allocator_traits<Alloc>::propagate_on_container_swap::value
+    >(),
+    x.m_allocator);
+
+  thrust::swap(m_allocator, x.m_allocator);
+} // end contiguous_storage::swap()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::default_construct_n(iterator first, size_type n)
+{
+  default_construct_range(m_allocator, first.base(), n);
+} // end contiguous_storage::default_construct_n()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::uninitialized_fill_n(iterator first, size_type n, const value_type &x)
+{
+  fill_construct_range(m_allocator, first.base(), n, x);
+} // end contiguous_storage::uninitialized_fill()
+
+template<typename T, typename Alloc>
+  template<typename System, typename InputIterator>
+  __host__ __device__
+    typename contiguous_storage<T,Alloc>::iterator
+      contiguous_storage<T,Alloc>
+        ::uninitialized_copy(thrust::execution_policy<System> &from_system, InputIterator first, InputIterator last, iterator result)
+{
+  return iterator(copy_construct_range(from_system, m_allocator, first, last, result.base()));
+} // end contiguous_storage::uninitialized_copy()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+  __host__ __device__
+    typename contiguous_storage<T,Alloc>::iterator
+      contiguous_storage<T,Alloc>
+        ::uninitialized_copy(InputIterator first, InputIterator last, iterator result)
+{
+  // XXX assumes InputIterator's associated System is default-constructible
+  typename thrust::iterator_system<InputIterator>::type from_system;
+
+  return iterator(copy_construct_range(from_system, m_allocator, first, last, result.base()));
+} // end contiguous_storage::uninitialized_copy()
+
+template<typename T, typename Alloc>
+  template<typename System, typename InputIterator, typename Size>
+  __host__ __device__
+    typename contiguous_storage<T,Alloc>::iterator
+      contiguous_storage<T,Alloc>
+        ::uninitialized_copy_n(thrust::execution_policy<System> &from_system, InputIterator first, Size n, iterator result)
+{
+  return iterator(copy_construct_range_n(from_system, m_allocator, first, n, result.base()));
+} // end contiguous_storage::uninitialized_copy_n()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator, typename Size>
+  __host__ __device__
+    typename contiguous_storage<T,Alloc>::iterator
+      contiguous_storage<T,Alloc>
+        ::uninitialized_copy_n(InputIterator first, Size n, iterator result)
+{
+  // XXX assumes InputIterator's associated System is default-constructible
+  typename thrust::iterator_system<InputIterator>::type from_system;
+
+  return iterator(copy_construct_range_n(from_system, m_allocator, first, n, result.base()));
+} // end contiguous_storage::uninitialized_copy_n()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::destroy(iterator first, iterator last)
+{
+  destroy_range(m_allocator, first.base(), last - first);
+} // end contiguous_storage::destroy()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::deallocate_on_allocator_mismatch(const contiguous_storage &other)
+{
+  integral_constant<
+    bool,
+    allocator_traits<Alloc>::propagate_on_container_copy_assignment::value
+  > c;
+
+  deallocate_on_allocator_mismatch_dispatch(c, other);
+} // end contiguous_storage::deallocate_on_allocator_mismatch
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::destroy_on_allocator_mismatch(const contiguous_storage &other,
+        iterator first, iterator last)
+{
+  integral_constant<
+    bool,
+    allocator_traits<Alloc>::propagate_on_container_copy_assignment::value
+  > c;
+
+  destroy_on_allocator_mismatch_dispatch(c, other, first, last);
+} // end contiguous_storage::destroy_on_allocator_mismatch
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::set_allocator(const Alloc &alloc)
+{
+  m_allocator = alloc;
+} // end contiguous_storage::set_allocator()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  bool contiguous_storage<T,Alloc>
+    ::is_allocator_not_equal(const Alloc &alloc) const
+{
+  return is_allocator_not_equal_dispatch(
+    integral_constant<
+      bool,
+      allocator_traits<Alloc>::is_always_equal::value
+    >(),
+    alloc);
+} // end contiguous_storage::is_allocator_not_equal()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  bool contiguous_storage<T,Alloc>
+    ::is_allocator_not_equal(const contiguous_storage<T,Alloc> &other) const
+{
+  return is_allocator_not_equal(m_allocator, other.m_allocator);
+} // end contiguous_storage::is_allocator_not_equal()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::propagate_allocator(const contiguous_storage &other)
+{
+  integral_constant<
+    bool,
+    allocator_traits<Alloc>::propagate_on_container_copy_assignment::value
+  > c;
+
+  propagate_allocator_dispatch(c, other);
+} // end contiguous_storage::propagate_allocator()
+
+#if THRUST_CPP_DIALECT >= 2011
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::propagate_allocator(contiguous_storage &other)
+{
+  integral_constant<
+    bool,
+    allocator_traits<Alloc>::propagate_on_container_move_assignment::value
+  > c;
+
+  propagate_allocator_dispatch(c, other);
+} // end contiguous_storage::propagate_allocator()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  contiguous_storage<T,Alloc> &contiguous_storage<T,Alloc>
+    ::operator=(contiguous_storage &&other)
+{
+  if (size() > 0)
+  {
+    deallocate();
+  }
+  propagate_allocator(other);
+  m_begin = std::move(other.m_begin);
+  m_size = std::move(other.m_size);
+
+  other.m_begin = pointer(static_cast<T*>(0));
+  other.m_size = 0;
+
+  return *this;
+} // end contiguous_storage::propagate_allocator()
+#endif
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::swap_allocators(true_type, const Alloc &)
+{
+} // end contiguous_storage::swap_allocators()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::swap_allocators(false_type, Alloc &other)
+{
+  NV_IF_TARGET(NV_IS_DEVICE, (
+    // allocators must be equal when swapping containers with allocators that propagate on swap
+    assert(!is_allocator_not_equal(other));
+  ), (
+    if (is_allocator_not_equal(other))
+    {
+      throw allocator_mismatch_on_swap();
+    }
+  ));
+
+  thrust::swap(m_allocator, other);
+} // end contiguous_storage::swap_allocators()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  bool contiguous_storage<T,Alloc>
+    ::is_allocator_not_equal_dispatch(true_type /*is_always_equal*/, const Alloc &) const
+{
+  return false;
+} // end contiguous_storage::is_allocator_not_equal_dispatch()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  bool contiguous_storage<T,Alloc>
+    ::is_allocator_not_equal_dispatch(false_type /*!is_always_equal*/, const Alloc& other) const
+{
+  return m_allocator != other;
+} // end contiguous_storage::is_allocator_not_equal_dispatch()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::deallocate_on_allocator_mismatch_dispatch(true_type, const contiguous_storage &other)
+{
+  if (m_allocator != other.m_allocator)
+  {
+    deallocate();
+  }
+} // end contiguous_storage::deallocate_on_allocator_mismatch()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::deallocate_on_allocator_mismatch_dispatch(false_type, const contiguous_storage &)
+{
+} // end contiguous_storage::deallocate_on_allocator_mismatch()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::destroy_on_allocator_mismatch_dispatch(true_type, const contiguous_storage &other,
+        iterator first, iterator last)
+{
+  if (m_allocator != other.m_allocator)
+  {
+    destroy(first, last);
+  }
+} // end contiguous_storage::destroy_on_allocator_mismatch()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::destroy_on_allocator_mismatch_dispatch(false_type, const contiguous_storage &,
+        iterator, iterator)
+{
+} // end contiguous_storage::destroy_on_allocator_mismatch()
+
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::propagate_allocator_dispatch(true_type, const contiguous_storage &other)
+{
+  m_allocator = other.m_allocator;
+} // end contiguous_storage::propagate_allocator()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::propagate_allocator_dispatch(false_type, const contiguous_storage &)
+{
+} // end contiguous_storage::propagate_allocator()
+
+#if THRUST_CPP_DIALECT >= 2011
+__thrust_exec_check_disable__
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::propagate_allocator_dispatch(true_type, contiguous_storage &other)
+{
+  m_allocator = std::move(other.m_allocator);
+} // end contiguous_storage::propagate_allocator()
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void contiguous_storage<T,Alloc>
+    ::propagate_allocator_dispatch(false_type, contiguous_storage &)
+{
+} // end contiguous_storage::propagate_allocator()
+#endif
+
+} // end detail
+
+template<typename T, typename Alloc>
+__host__ __device__
+  void swap(detail::contiguous_storage<T,Alloc> &lhs, detail::contiguous_storage<T,Alloc> &rhs)
+{
+  lhs.swap(rhs);
+} // end swap()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.h
new file mode 100644
index 000000000000..d6c5bc805c4c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.h
@@ -0,0 +1,90 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename System,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(const thrust::detail::execution_policy_base<System> &system,
+                      InputIterator first,
+                      InputIterator last,
+                      OutputIterator result);
+
+template<typename System,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(const thrust::detail::execution_policy_base<System> &system,
+                        InputIterator first,
+                        Size n,
+                        OutputIterator result);
+
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator copy(InputIterator first,
+                      InputIterator last,
+                      OutputIterator result);
+
+template<typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+  OutputIterator copy_n(InputIterator first,
+                        Size n,
+                        OutputIterator result);
+
+
+namespace detail
+{
+
+
+template<typename FromSystem,
+         typename ToSystem,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator two_system_copy(const thrust::execution_policy<FromSystem> &from_system,
+                                 const thrust::execution_policy<ToSystem>   &two_system,
+                                 InputIterator first,
+                                 InputIterator last,
+                                 OutputIterator result);
+
+
+template<typename FromSystem,
+         typename ToSystem,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator two_system_copy_n(const thrust::execution_policy<FromSystem> &from_system,
+                                   const thrust::execution_policy<ToSystem>   &two_system,
+                                   InputIterator first,
+                                   Size n,
+                                   OutputIterator result);
+
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/copy.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.inl
new file mode 100644
index 000000000000..4d62798c705b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy.inl
@@ -0,0 +1,129 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/copy.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/copy.h>
+#include <thrust/system/detail/adl/copy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      OutputIterator result)
+{
+  using thrust::system::detail::generic::copy;
+  return copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result);
+} // end copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Size, typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        InputIterator first,
+                        Size n,
+                        OutputIterator result)
+{
+  using thrust::system::detail::generic::copy_n;
+  return copy_n(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, n, result);
+} // end copy_n()
+
+
+namespace detail
+{
+
+
+__thrust_exec_check_disable__ // because we might call e.g. std::ostream_iterator's constructor
+template<typename System1,
+         typename System2,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator two_system_copy(const thrust::execution_policy<System1> &system1,
+                                 const thrust::execution_policy<System2> &system2,
+                                 InputIterator first,
+                                 InputIterator last,
+                                 OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  return thrust::copy(select_system(thrust::detail::derived_cast(thrust::detail::strip_const(system1)), thrust::detail::derived_cast(thrust::detail::strip_const(system2))), first, last, result);
+} // end two_system_copy()
+
+
+__thrust_exec_check_disable__ // because we might call e.g. std::ostream_iterator's constructor
+template<typename System1,
+         typename System2,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator two_system_copy_n(const thrust::execution_policy<System1> &system1,
+                                   const thrust::execution_policy<System2> &system2,
+                                   InputIterator first,
+                                   Size n,
+                                   OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  return thrust::copy_n(select_system(thrust::detail::derived_cast(thrust::detail::strip_const(system1)), thrust::detail::derived_cast(thrust::detail::strip_const(system2))), first, n, result);
+} // end two_system_copy_n()
+
+
+} // end detail
+
+
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator copy(InputIterator first,
+                      InputIterator last,
+                      OutputIterator result)
+{
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::detail::two_system_copy(system1, system2, first, last, result);
+} // end copy()
+
+
+template<typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+  OutputIterator copy_n(InputIterator first,
+                        Size n,
+                        OutputIterator result)
+{
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::detail::two_system_copy_n(system1, system2, first, n, result);
+} // end copy_n()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.h
new file mode 100644
index 000000000000..32eb5e08323c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.h
@@ -0,0 +1,71 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator copy_if(InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator copy_if(InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/copy_if.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.inl
new file mode 100644
index 000000000000..952541c51806
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/copy_if.inl
@@ -0,0 +1,107 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/copy_if.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/copy_if.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/adl/copy_if.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred)
+{
+  using thrust::system::detail::generic::copy_if;
+  return copy_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, pred);
+} // end copy_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred)
+{
+  using thrust::system::detail::generic::copy_if;
+  return copy_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, result, pred);
+} // end copy_if()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator copy_if(InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::copy_if(select_system(system1,system2), first, last, result, pred);
+} // end copy_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator copy_if(InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::copy_if(select_system(system1,system2,system3), first, last, stencil, result, pred);
+} // end copy_if()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/count.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/count.h
new file mode 100644
index 000000000000..7c48bc546060
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/count.h
@@ -0,0 +1,60 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename EqualityComparable>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+          InputIterator first,
+          InputIterator last,
+          const EqualityComparable& value);
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+             InputIterator first,
+             InputIterator last,
+             Predicate pred);
+
+template <typename InputIterator,
+          typename EqualityComparable>
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count(InputIterator first,
+          InputIterator last,
+          const EqualityComparable& value);
+
+template <typename InputIterator,
+          typename Predicate>
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count_if(InputIterator first,
+             InputIterator last,
+             Predicate pred);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/count.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/count.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/count.inl
new file mode 100644
index 000000000000..5d1f628a93c4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/count.inl
@@ -0,0 +1,77 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/count.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/count.h>
+#include <thrust/system/detail/adl/count.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename EqualityComparable>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, const EqualityComparable& value)
+{
+  using thrust::system::detail::generic::count;
+  return count(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+} // end count()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    count_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::count_if;
+  return count_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end count_if()
+
+
+template <typename InputIterator, typename EqualityComparable>
+typename thrust::iterator_traits<InputIterator>::difference_type
+count(InputIterator first, InputIterator last, const EqualityComparable& value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::count(select_system(system), first, last, value);
+} // end count()
+
+
+template <typename InputIterator, typename Predicate>
+typename thrust::iterator_traits<InputIterator>::difference_type
+count_if(InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::count_if(select_system(system), first, last, pred);
+} // end count_if()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp11_required.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp11_required.h
new file mode 100644
index 000000000000..a7fb4fb12cdd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp11_required.h
@@ -0,0 +1,26 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config/cpp_dialect.h>
+
+#ifndef THRUST_CPP11_REQUIRED_NO_ERROR
+#  if THRUST_CPP_DIALECT < 2011 
+#    error C++11 is required for this Thrust feature; please upgrade your compiler or pass the appropriate -std=c++XX flag to it.
+#  endif
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp14_required.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp14_required.h
new file mode 100644
index 000000000000..083c8a1ad478
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/cpp14_required.h
@@ -0,0 +1,26 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config/cpp_dialect.h>
+
+#ifndef THRUST_CPP14_REQUIRED_NO_ERROR
+#  if THRUST_CPP_DIALECT < 2014
+#    error C++14 is required for this Thrust feature; please upgrade your compiler or pass the appropriate -std=c++14 flag to it.
+#  endif
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/cstdint.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/cstdint.h
new file mode 100644
index 000000000000..f41e11475765
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/cstdint.h
@@ -0,0 +1,83 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC) || \
+    (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG) || \
+    (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_INTEL)
+#include <stdint.h>
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC)
+
+#if (_MSC_VER < 1300)
+   typedef signed   char     int8_t;
+   typedef signed   short    int16_t;
+   typedef signed   int      int32_t;
+   typedef unsigned char     uint8_t;
+   typedef unsigned short    uint16_t;
+   typedef unsigned int      uint32_t;
+#else
+   typedef signed   __int8   int8_t;
+   typedef signed   __int16  int16_t;
+   typedef signed   __int32  int32_t;
+   typedef unsigned __int8   uint8_t;
+   typedef unsigned __int16  uint16_t;
+   typedef unsigned __int32  uint32_t;
+#endif
+typedef signed   __int64     int64_t;
+typedef unsigned __int64     uint64_t;
+
+#else
+
+typedef ::int8_t   int8_t;
+typedef ::int16_t  int16_t;
+typedef ::int32_t  int32_t;
+typedef ::int64_t  int64_t;
+typedef ::uint8_t  uint8_t;
+typedef ::uint16_t uint16_t;
+typedef ::uint32_t uint32_t;
+typedef ::uint64_t uint64_t;
+
+#endif
+
+
+// an oracle to tell us how to define intptr_t
+template<int word_size = sizeof(void*)> struct divine_intptr_t;
+template<int word_size = sizeof(void*)> struct divine_uintptr_t;
+
+// 32b platforms
+template<>  struct divine_intptr_t<4>  {  typedef thrust::detail::int32_t  type; };
+template<>  struct divine_uintptr_t<4> {  typedef thrust::detail::uint32_t type; };
+
+// 64b platforms
+template<>  struct divine_intptr_t<8>  { typedef thrust::detail::int64_t  type; };
+template<>  struct divine_uintptr_t<8> { typedef thrust::detail::uint64_t type; };
+
+typedef divine_intptr_t<>::type   intptr_t;
+typedef divine_uintptr_t<>::type  uintptr_t;
+
+} // end detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/dependencies_aware_execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/dependencies_aware_execution_policy.h
new file mode 100644
index 000000000000..a7567a3fab30
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/dependencies_aware_execution_policy.h
@@ -0,0 +1,106 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <tuple>
+
+#include <thrust/detail/execute_with_dependencies.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<template<typename> class ExecutionPolicyCRTPBase>
+struct dependencies_aware_execution_policy
+{
+    template<typename ...Dependencies>
+    __host__
+    thrust::detail::execute_with_dependencies<
+        ExecutionPolicyCRTPBase,
+        Dependencies...
+    >
+    after(Dependencies&& ...dependencies) const
+    {
+        return { capture_as_dependency(THRUST_FWD(dependencies))... };
+    }
+
+    template<typename ...Dependencies>
+    __host__
+    thrust::detail::execute_with_dependencies<
+        ExecutionPolicyCRTPBase,
+        Dependencies...
+    >
+    after(std::tuple<Dependencies...>& dependencies) const
+    {
+        return { capture_as_dependency(dependencies) };
+    }
+    template<typename ...Dependencies>
+    __host__
+    thrust::detail::execute_with_dependencies<
+        ExecutionPolicyCRTPBase,
+        Dependencies...
+    >
+    after(std::tuple<Dependencies...>&& dependencies) const
+    {
+        return { capture_as_dependency(std::move(dependencies)) };
+    }
+
+    template<typename ...Dependencies>
+    __host__
+    thrust::detail::execute_with_dependencies<
+        ExecutionPolicyCRTPBase,
+        Dependencies...
+    >
+    rebind_after(Dependencies&& ...dependencies) const
+    {
+        return { capture_as_dependency(THRUST_FWD(dependencies))... };
+    }
+
+    template<typename ...Dependencies>
+    __host__
+    thrust::detail::execute_with_dependencies<
+        ExecutionPolicyCRTPBase,
+        Dependencies...
+    >
+    rebind_after(std::tuple<Dependencies...>& dependencies) const
+    {
+        return { capture_as_dependency(dependencies) };
+    }
+    template<typename ...Dependencies>
+    __host__
+    thrust::detail::execute_with_dependencies<
+        ExecutionPolicyCRTPBase,
+        Dependencies...
+    >
+    rebind_after(std::tuple<Dependencies...>&& dependencies) const
+    {
+        return { capture_as_dependency(std::move(dependencies)) };
+    }
+};
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_delete.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_delete.inl
new file mode 100644
index 000000000000..87f73aad9023
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_delete.inl
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_delete.h>
+#include <thrust/device_free.h>
+#include <thrust/detail/allocator/destroy_range.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// define an empty allocator class to use below
+struct device_delete_allocator {};
+
+}
+
+template<typename T>
+  void device_delete(device_ptr<T> ptr,
+                     const size_t n)
+{
+  // we can use device_allocator to destroy the range
+  thrust::detail::device_delete_allocator a;
+  thrust::detail::destroy_range(a, ptr, n);
+  thrust::device_free(ptr);
+} // end device_delete()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_free.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_free.inl
new file mode 100644
index 000000000000..806802e16513
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_free.inl
@@ -0,0 +1,39 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_free.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/detail/malloc_and_free.h>
+
+THRUST_NAMESPACE_BEGIN
+
+void device_free(thrust::device_ptr<void> ptr)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef thrust::iterator_system< thrust::device_ptr<void> >::type system;
+
+  // XXX lower to select_system(system) here
+  system s;
+
+  thrust::free(s, ptr);
+} // end device_free()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_malloc.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_malloc.inl
new file mode 100644
index 000000000000..f4222f51da30
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_malloc.inl
@@ -0,0 +1,53 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_malloc.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/detail/malloc_and_free.h>
+
+THRUST_NAMESPACE_BEGIN
+
+thrust::device_ptr<void> device_malloc(const std::size_t n)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef thrust::iterator_system< thrust::device_ptr<void> >::type system;
+
+  // XXX lower to select_system(system) here
+  system s;
+
+  return thrust::device_ptr<void>(thrust::malloc(s, n).get());
+} // end device_malloc()
+
+
+template<typename T>
+  thrust::device_ptr<T> device_malloc(const std::size_t n)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef thrust::iterator_system< thrust::device_ptr<void> >::type system;
+
+  // XXX lower to select_system(system) here
+  system s;
+
+  return thrust::device_ptr<T>(thrust::malloc<T>(s,n).get());
+} // end device_malloc()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_new.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_new.inl
new file mode 100644
index 000000000000..c66e2cbff6cb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_new.inl
@@ -0,0 +1,56 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_new.h>
+#include <thrust/device_malloc.h>
+#include <thrust/uninitialized_fill.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename T>
+  device_ptr<T> device_new(device_ptr<void> p,
+                           const size_t n)
+{
+  // XXX TODO dispatch n null device constructors at p here
+  // in the meantime, dispatch 1 null host constructor here
+  // and dispatch n copy constructors
+  return device_new<T>(p, T(), n);
+} // end device_new()
+
+template<typename T>
+  device_ptr<T> device_new(device_ptr<void> p,
+                           const T &exemplar,
+                           const size_t n)
+{
+  device_ptr<T> result(reinterpret_cast<T*>(p.get()));
+
+  // run copy constructors at p here
+  thrust::uninitialized_fill(result, result + n, exemplar);
+
+  return result;
+} // end device_new()
+
+template<typename T>
+  device_ptr<T> device_new(const size_t n)
+{
+  // call placement new
+  return device_new<T>(thrust::device_malloc<T>(n));
+} // end device_new()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_ptr.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_ptr.inl
new file mode 100644
index 000000000000..361c61f33149
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/device_ptr.inl
@@ -0,0 +1,64 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/device_ptr.h>
+#include <thrust/device_reference.h>
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename T>
+  __host__ __device__
+  device_ptr<T> device_pointer_cast(T *ptr)
+{
+  return device_ptr<T>(ptr);
+} // end device_pointer_cast()
+
+template<typename T>
+  __host__ __device__
+  device_ptr<T> device_pointer_cast(const device_ptr<T> &ptr)
+{
+  return ptr;
+} // end device_pointer_cast()
+
+
+namespace detail
+{
+
+template<typename T>
+  struct is_device_ptr< thrust::device_ptr<T> >
+    : public true_type
+{
+}; // end is_device_ptr
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC) && (_MSC_VER <= 1400)
+// XXX WAR MSVC 2005 problem with correctly implementing
+//     pointer_raw_pointer for device_ptr by specializing it here
+template<typename T>
+  struct pointer_raw_pointer< thrust::device_ptr<T> >
+{
+  typedef typename device_ptr<T>::raw_pointer type;
+}; // end pointer_raw_pointer
+#endif
+
+
+} // end namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/distance.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/distance.inl
new file mode 100644
index 000000000000..6702c2b6f738
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/distance.inl
@@ -0,0 +1,35 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/advance.h>
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/distance.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename InputIterator>
+inline __host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last)
+{
+  return thrust::system::detail::generic::distance(first, last);
+} // end distance()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/equal.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/equal.inl
new file mode 100644
index 000000000000..e21ddfa5abb8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/equal.inl
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/equal.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/equal.h>
+#include <thrust/system/detail/adl/equal.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename System, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+bool equal(const thrust::detail::execution_policy_base<System> &system, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2)
+{
+  using thrust::system::detail::generic::equal;
+  return equal(thrust::detail::derived_cast(thrust::detail::strip_const(system)), first1, last1, first2);
+} // end equal()
+
+
+__thrust_exec_check_disable__
+template<typename System, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+bool equal(const thrust::detail::execution_policy_base<System> &system, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::equal;
+  return equal(thrust::detail::derived_cast(thrust::detail::strip_const(system)), first1, last1, first2, binary_pred);
+} // end equal()
+
+
+template <typename InputIterator1, typename InputIterator2>
+bool equal(InputIterator1 first1, InputIterator1 last1,
+           InputIterator2 first2)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::equal(select_system(system1,system2), first1, last1, first2);
+}
+
+
+template <typename InputIterator1, typename InputIterator2,
+          typename BinaryPredicate>
+bool equal(InputIterator1 first1, InputIterator1 last1,
+           InputIterator2 first2, BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::equal(select_system(system1,system2), first1, last1, first2, binary_pred);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/event_error.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/event_error.h
new file mode 100644
index 000000000000..b928e0650c25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/event_error.h
@@ -0,0 +1,164 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/// \file thrust/detail/event_error.h
+/// \brief \c thrust::future and thrust::future error handling types and codes.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/detail/type_traits.h>
+#include <thrust/system/error_code.h>
+
+#include <stdexcept>
+
+THRUST_NAMESPACE_BEGIN
+
+enum class event_errc
+{
+  unknown_event_error
+, no_state
+, no_content
+, last_event_error
+};
+
+/// \return <tt>error_code(static_cast<int>(e), event_category())</tt>
+inline error_code make_error_code(event_errc e);
+
+/// \return <tt>error_condition(static_cast<int>(e), event_category())</tt>.
+inline error_condition make_error_condition(event_errc e);
+
+struct event_error_category : error_category
+{
+  event_error_category() = default;
+
+  virtual char const* name() const
+  {
+    return "event";
+  }
+
+  virtual std::string message(int ev) const
+  {
+    switch (static_cast<event_errc>(ev))
+    {
+      case event_errc::no_state:
+      {
+        return "no_state: an operation that requires an event or future to have "
+               "a stream or content has been performed on a event or future "
+               "without either, e.g. a moved-from or default constructed event "
+               "or future (an event or future may have been consumed more than "
+               "once)";
+      }
+      case event_errc::no_content:
+      {
+        return "no_content: an operation that requires a future to have content "
+               "has been performed on future without any, e.g. a moved-from, "
+               "default constructed, or `thrust::new_stream` constructed future "
+               "(a future may have been consumed more than once)";
+      }
+      default:
+      {
+        return "unknown_event_error: an unknown error with a future "
+               "object has occurred";
+      }
+    };
+  }
+
+  virtual error_condition default_error_condition(int ev) const
+  {
+    if (
+         event_errc::last_event_error
+         >
+         static_cast<event_errc>(ev)
+       )
+      return make_error_condition(static_cast<event_errc>(ev));
+
+    return system_category().default_error_condition(ev);
+  }
+};
+
+/// Obtains a reference to the static error category object for the errors
+/// related to futures and promises. The object is required to override the
+/// virtual function error_category::name() to return a pointer to the string
+/// "event". It is used to identify error codes provided in the
+/// exceptions of type event_error.
+inline error_category const& event_category()
+{
+  static const event_error_category result;
+  return result;
+}
+
+namespace system
+{
+/// Specialization of \p is_error_code_enum for \p event_errc.
+template<> struct is_error_code_enum<event_errc> : true_type {};
+} // end system
+
+/// \return <tt>error_code(static_cast<int>(e), event_category())</tt>
+inline error_code make_error_code(event_errc e)
+{
+  return error_code(static_cast<int>(e), event_category());
+}
+
+/// \return <tt>error_condition(static_cast<int>(e), event_category())</tt>.
+inline error_condition make_error_condition(event_errc e)
+{
+  return error_condition(static_cast<int>(e), event_category());
+}
+
+struct event_error : std::logic_error
+{
+  __host__
+  explicit event_error(error_code ec)
+    : std::logic_error(ec.message()), ec_(ec)
+  {}
+
+  __host__
+  explicit event_error(event_errc e)
+    : event_error(make_error_code(e))
+  {}
+
+  __host__
+  error_code const& code() const noexcept
+  {
+    return ec_;
+  }
+
+  __host__
+  virtual ~event_error() noexcept {}
+
+private:
+  error_code ec_;
+};
+
+inline bool operator==(event_error const& lhs, event_error const& rhs) noexcept
+{
+  return lhs.code() == rhs.code();
+}
+
+inline bool operator<(event_error const& lhs, event_error const& rhs) noexcept
+{
+  return lhs.code() < rhs.code();
+}
+
+THRUST_NAMESPACE_END
+
+#endif // C++14
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator.h
new file mode 100644
index 000000000000..430fe739c3f4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator.h
@@ -0,0 +1,149 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/execute_with_allocator_fwd.h>
+#include <thrust/pair.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/integer_math.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template <
+    typename T
+  , typename Allocator
+  , template <typename> class BaseSystem
+>
+__host__
+thrust::pair<T*, std::ptrdiff_t>
+get_temporary_buffer(
+    thrust::detail::execute_with_allocator<Allocator, BaseSystem>& system
+  , std::ptrdiff_t n
+    )
+{
+  typedef typename thrust::detail::remove_reference<Allocator>::type naked_allocator;
+  typedef typename thrust::detail::allocator_traits<naked_allocator> alloc_traits;
+  typedef typename alloc_traits::void_pointer                        void_pointer;
+  typedef typename alloc_traits::size_type                           size_type;
+  typedef typename alloc_traits::value_type                          value_type;
+
+  // How many elements of type value_type do we need to accommodate n elements
+  // of type T?
+  size_type num_elements = divide_ri(sizeof(T) * n, sizeof(value_type));
+
+  void_pointer ptr = alloc_traits::allocate(system.get_allocator(), num_elements);
+
+  // Return the pointer and the number of elements of type T allocated.
+  return thrust::make_pair(thrust::reinterpret_pointer_cast<T*>(ptr),n);
+}
+
+template <
+    typename Pointer
+  , typename Allocator
+  , template <typename> class BaseSystem
+>
+__host__
+void
+return_temporary_buffer(
+    thrust::detail::execute_with_allocator<Allocator, BaseSystem>& system
+  , Pointer p
+  , std::ptrdiff_t n
+    )
+{
+  typedef typename thrust::detail::remove_reference<Allocator>::type naked_allocator;
+  typedef typename thrust::detail::allocator_traits<naked_allocator> alloc_traits;
+  typedef typename alloc_traits::pointer                             pointer;
+  typedef typename alloc_traits::size_type                           size_type;
+  typedef typename alloc_traits::value_type                          value_type;
+  typedef typename thrust::detail::pointer_traits<Pointer>::element_type T;
+
+  size_type num_elements = divide_ri(sizeof(T) * n, sizeof(value_type));
+
+  pointer to_ptr = thrust::reinterpret_pointer_cast<pointer>(p);
+  alloc_traits::deallocate(system.get_allocator(), to_ptr, num_elements);
+}
+
+#if THRUST_CPP_DIALECT >= 2011
+
+template <
+    typename T,
+    template <typename> class BaseSystem,
+    typename Allocator,
+    typename ...Dependencies
+>
+__host__
+thrust::pair<T*, std::ptrdiff_t>
+get_temporary_buffer(
+    thrust::detail::execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>& system,
+    std::ptrdiff_t n
+    )
+{
+  typedef typename thrust::detail::remove_reference<Allocator>::type naked_allocator;
+  typedef typename thrust::detail::allocator_traits<naked_allocator> alloc_traits;
+  typedef typename alloc_traits::void_pointer                        void_pointer;
+  typedef typename alloc_traits::size_type                           size_type;
+  typedef typename alloc_traits::value_type                          value_type;
+
+  // How many elements of type value_type do we need to accommodate n elements
+  // of type T?
+  size_type num_elements = divide_ri(sizeof(T) * n, sizeof(value_type));
+
+  void_pointer ptr = alloc_traits::allocate(system.get_allocator(), num_elements);
+
+  // Return the pointer and the number of elements of type T allocated.
+  return thrust::make_pair(thrust::reinterpret_pointer_cast<T*>(ptr),n);
+}
+
+template <
+    typename Pointer,
+    template <typename> class BaseSystem,
+    typename Allocator,
+    typename ...Dependencies
+>
+__host__
+void
+return_temporary_buffer(
+    thrust::detail::execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>& system,
+    Pointer p,
+    std::ptrdiff_t n
+    )
+{
+  typedef typename thrust::detail::remove_reference<Allocator>::type naked_allocator;
+  typedef typename thrust::detail::allocator_traits<naked_allocator> alloc_traits;
+  typedef typename alloc_traits::pointer                             pointer;
+  typedef typename alloc_traits::size_type                           size_type;
+  typedef typename alloc_traits::value_type                          value_type;
+  typedef typename thrust::detail::pointer_traits<Pointer>::element_type T;
+
+  size_type num_elements = divide_ri(sizeof(T) * n, sizeof(value_type));
+
+  pointer to_ptr = thrust::reinterpret_pointer_cast<pointer>(p);
+  alloc_traits::deallocate(system.get_allocator(), to_ptr, num_elements);
+}
+
+#endif
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator_fwd.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator_fwd.h
new file mode 100644
index 000000000000..1d5899a7db76
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_allocator_fwd.h
@@ -0,0 +1,106 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+  #include <thrust/detail/execute_with_dependencies.h>
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template <typename Allocator, template <typename> class BaseSystem>
+struct execute_with_allocator
+  : BaseSystem<execute_with_allocator<Allocator, BaseSystem> >
+{
+private:
+  typedef BaseSystem<execute_with_allocator<Allocator, BaseSystem> > super_t;
+
+  Allocator alloc;
+
+public:
+  __host__ __device__
+  execute_with_allocator(super_t const& super, Allocator alloc_)
+    : super_t(super), alloc(alloc_)
+  {}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  execute_with_allocator(Allocator alloc_)
+    : alloc(alloc_)
+  {}
+
+  typename remove_reference<Allocator>::type& get_allocator() { return alloc; }
+
+#if THRUST_CPP_DIALECT >= 2011
+  template<typename ...Dependencies>
+  __host__
+  execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>
+  after(Dependencies&& ...dependencies) const
+  {
+    return { alloc, capture_as_dependency(THRUST_FWD(dependencies))... };
+  }
+
+  template<typename ...Dependencies>
+  __host__
+  execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>
+  after(std::tuple<Dependencies...>& dependencies) const
+  {
+      return { alloc, capture_as_dependency(dependencies) };
+  }
+  template<typename ...Dependencies>
+  __host__
+  execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>
+  after(std::tuple<Dependencies...>&& dependencies) const
+  {
+      return { alloc, capture_as_dependency(std::move(dependencies)) };
+  }
+
+  template<typename ...Dependencies>
+  __host__
+  execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>
+  rebind_after(Dependencies&& ...dependencies) const
+  {
+    return { alloc, capture_as_dependency(THRUST_FWD(dependencies))... };
+  }
+
+  template<typename ...Dependencies>
+  __host__
+  execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>
+  rebind_after(std::tuple<Dependencies...>& dependencies) const
+  {
+      return { alloc, capture_as_dependency(dependencies) };
+  }
+  template<typename ...Dependencies>
+  __host__
+  execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>
+  rebind_after(std::tuple<Dependencies...>&& dependencies) const
+  {
+      return { alloc, capture_as_dependency(std::move(dependencies)) };
+  }
+#endif
+};
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_dependencies.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_dependencies.h
new file mode 100644
index 000000000000..ec54010b0520
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execute_with_dependencies.h
@@ -0,0 +1,267 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/detail/type_deduction.h>
+#include <thrust/type_traits/remove_cvref.h>
+
+#include <tuple>
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+struct capture_as_dependency_fn
+{
+  template<typename Dependency>
+  auto operator()(Dependency&& dependency) const
+  THRUST_DECLTYPE_RETURNS(capture_as_dependency(THRUST_FWD(dependency)))
+};
+
+// Default implementation: universal forwarding.
+template<typename Dependency>
+auto capture_as_dependency(Dependency&& dependency)
+THRUST_DECLTYPE_RETURNS(THRUST_FWD(dependency))
+
+template<typename... Dependencies>
+auto capture_as_dependency(std::tuple<Dependencies...>& dependencies)
+THRUST_DECLTYPE_RETURNS(
+  tuple_for_each(THRUST_FWD(dependencies), capture_as_dependency_fn{})
+)
+
+template<template<typename> class BaseSystem, typename... Dependencies>
+struct execute_with_dependencies
+    : BaseSystem<execute_with_dependencies<BaseSystem, Dependencies...>>
+{
+private:
+    using super_t = BaseSystem<execute_with_dependencies<BaseSystem, Dependencies...>>;
+
+    std::tuple<remove_cvref_t<Dependencies>...> dependencies;
+
+public:
+    __host__
+    execute_with_dependencies(super_t const &super, Dependencies && ...dependencies)
+        : super_t(super), dependencies(std::forward<Dependencies>(dependencies)...)
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_dependencies(super_t const &super, UDependencies && ...deps)
+        : super_t(super), dependencies(THRUST_FWD(deps)...)
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_dependencies(UDependencies && ...deps)
+        : dependencies(THRUST_FWD(deps)...)
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_dependencies(super_t const &super, std::tuple<UDependencies...>&& deps)
+        : super_t(super), dependencies(std::move(deps))
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_dependencies(std::tuple<UDependencies...>&& deps)
+        : dependencies(std::move(deps))
+    {
+    }
+
+    std::tuple<remove_cvref_t<Dependencies>...>
+    __host__
+    extract_dependencies() 
+    {
+        return std::move(dependencies);
+    }
+
+    // Rebinding.
+    template<typename ...UDependencies>
+    __host__
+    execute_with_dependencies<BaseSystem, UDependencies...>
+    rebind_after(UDependencies&& ...udependencies) const
+    {
+        return { capture_as_dependency(THRUST_FWD(udependencies))... };
+    }
+
+    // Rebinding.
+    template<typename ...UDependencies>
+    __host__
+    execute_with_dependencies<BaseSystem, UDependencies...>
+    rebind_after(std::tuple<UDependencies...>& udependencies) const
+    {
+        return { capture_as_dependency(udependencies) };
+    }
+    template<typename ...UDependencies>
+    __host__
+    execute_with_dependencies<BaseSystem, UDependencies...>
+    rebind_after(std::tuple<UDependencies...>&& udependencies) const
+    {
+        return { capture_as_dependency(std::move(udependencies)) };
+    }
+};
+
+template<
+    typename Allocator,
+    template<typename> class BaseSystem,
+    typename... Dependencies
+>
+struct execute_with_allocator_and_dependencies
+    : BaseSystem<
+        execute_with_allocator_and_dependencies<
+            Allocator,
+            BaseSystem,
+            Dependencies...
+        >
+    >
+{
+private:
+    using super_t = BaseSystem<
+        execute_with_allocator_and_dependencies<
+            Allocator,
+            BaseSystem,
+            Dependencies...
+        >
+    >;
+
+    std::tuple<remove_cvref_t<Dependencies>...> dependencies;
+    Allocator alloc;
+
+public:
+    template <typename... UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies(super_t const &super, Allocator a, UDependencies && ...deps)
+        : super_t(super), dependencies(THRUST_FWD(deps)...), alloc(a)
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies(Allocator a, UDependencies && ...deps)
+        : dependencies(THRUST_FWD(deps)...), alloc(a)
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies(super_t const &super, Allocator a, std::tuple<UDependencies...>&& deps)
+        : super_t(super), dependencies(std::move(deps)), alloc(a)
+    {
+    }
+
+    template <typename... UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies(Allocator a, std::tuple<UDependencies...>&& deps)
+        : dependencies(std::move(deps)), alloc(a)
+    {
+    }
+
+    std::tuple<remove_cvref_t<Dependencies>...>
+    __host__
+    extract_dependencies() 
+    {
+        return std::move(dependencies);
+    }
+
+    __host__
+    typename std::add_lvalue_reference<Allocator>::type
+    get_allocator()
+    {
+        return alloc;
+    }
+
+    // Rebinding.
+    template<typename ...UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies<Allocator, BaseSystem, UDependencies...>
+    rebind_after(UDependencies&& ...udependencies) const
+    {
+        return { alloc, capture_as_dependency(THRUST_FWD(udependencies))... };
+    }
+
+    // Rebinding.
+    template<typename ...UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies<Allocator, BaseSystem, UDependencies...>
+    rebind_after(std::tuple<UDependencies...>& udependencies) const
+    {
+        return { alloc, capture_as_dependency(udependencies) };
+    }
+    template<typename ...UDependencies>
+    __host__
+    execute_with_allocator_and_dependencies<Allocator, BaseSystem, UDependencies...>
+    rebind_after(std::tuple<UDependencies...>&& udependencies) const
+    {
+        return { alloc, capture_as_dependency(std::move(udependencies)) };
+    }
+};
+
+template<template<typename> class BaseSystem, typename ...Dependencies>
+__host__
+std::tuple<remove_cvref_t<Dependencies>...>
+extract_dependencies(thrust::detail::execute_with_dependencies<BaseSystem, Dependencies...>&& system)
+{
+    return std::move(system).extract_dependencies();
+}
+template<template<typename> class BaseSystem, typename ...Dependencies>
+__host__
+std::tuple<remove_cvref_t<Dependencies>...>
+extract_dependencies(thrust::detail::execute_with_dependencies<BaseSystem, Dependencies...>& system)
+{
+    return std::move(system).extract_dependencies();
+}
+
+template<typename Allocator, template<typename> class BaseSystem, typename ...Dependencies>
+__host__
+std::tuple<remove_cvref_t<Dependencies>...>
+extract_dependencies(thrust::detail::execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>&& system)
+{
+    return std::move(system).extract_dependencies();
+}
+template<typename Allocator, template<typename> class BaseSystem, typename ...Dependencies>
+__host__
+std::tuple<remove_cvref_t<Dependencies>...>
+extract_dependencies(thrust::detail::execute_with_allocator_and_dependencies<Allocator, BaseSystem, Dependencies...>& system)
+{
+    return std::move(system).extract_dependencies();
+}
+
+template<typename System>
+__host__
+std::tuple<>
+extract_dependencies(System &&)
+{
+    return std::tuple<>{};
+}
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execution_policy.h
new file mode 100644
index 000000000000..dcc11a770331
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/execution_policy.h
@@ -0,0 +1,76 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+struct execution_policy_marker {};
+
+// execution_policy_base serves as a guard against
+// inifinite recursion in thrust entry points:
+//
+// template<typename DerivedPolicy>
+// void foo(const thrust::detail::execution_policy_base<DerivedPolicy> &s)
+// {
+//   using thrust::system::detail::generic::foo;
+//
+//   foo(thrust::detail::derived_cast(thrust::detail::strip_const(s));
+// }
+//
+// foo is not recursive when
+// 1. DerivedPolicy is derived from thrust::execution_policy below
+// 2. generic::foo takes thrust::execution_policy as a parameter
+template<typename DerivedPolicy>
+struct execution_policy_base : execution_policy_marker {};
+
+
+template<typename DerivedPolicy>
+constexpr __host__ __device__
+execution_policy_base<DerivedPolicy> &strip_const(const execution_policy_base<DerivedPolicy> &x)
+{
+  return const_cast<execution_policy_base<DerivedPolicy>&>(x);
+}
+
+
+template<typename DerivedPolicy>
+constexpr __host__ __device__
+DerivedPolicy &derived_cast(execution_policy_base<DerivedPolicy> &x)
+{
+  return static_cast<DerivedPolicy&>(x);
+}
+
+
+template<typename DerivedPolicy>
+constexpr __host__ __device__
+const DerivedPolicy &derived_cast(const execution_policy_base<DerivedPolicy> &x)
+{
+  return static_cast<const DerivedPolicy&>(x);
+}
+
+} // end detail
+
+template<typename DerivedPolicy>
+  struct execution_policy
+    : thrust::detail::execution_policy_base<DerivedPolicy>
+{};
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/extrema.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/extrema.inl
new file mode 100644
index 000000000000..2c1750e7d481
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/extrema.inl
@@ -0,0 +1,169 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/extrema.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/extrema.h>
+#include <thrust/system/detail/adl/extrema.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator min_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last)
+{
+  using thrust::system::detail::generic::min_element;
+  return min_element(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end min_element()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator min_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last, BinaryPredicate comp)
+{
+  using thrust::system::detail::generic::min_element;
+  return min_element(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end min_element()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator max_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last)
+{
+  using thrust::system::detail::generic::max_element;
+  return max_element(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end max_element()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator max_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last, BinaryPredicate comp)
+{
+  using thrust::system::detail::generic::max_element;
+  return max_element(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end max_element()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last)
+{
+  using thrust::system::detail::generic::minmax_element;
+  return minmax_element(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end minmax_element()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last, BinaryPredicate comp)
+{
+  using thrust::system::detail::generic::minmax_element;
+  return minmax_element(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end minmax_element()
+
+
+template <typename ForwardIterator>
+ForwardIterator min_element(ForwardIterator first, ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::min_element(select_system(system), first, last);
+} // end min_element()
+
+
+template <typename ForwardIterator, typename BinaryPredicate>
+ForwardIterator min_element(ForwardIterator first, ForwardIterator last,
+                            BinaryPredicate comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::min_element(select_system(system), first, last, comp);
+} // end min_element()
+
+
+template <typename ForwardIterator>
+ForwardIterator max_element(ForwardIterator first, ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::max_element(select_system(system), first, last);
+} // end max_element()
+
+
+template <typename ForwardIterator, typename BinaryPredicate>
+ForwardIterator max_element(ForwardIterator first, ForwardIterator last,
+                            BinaryPredicate comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::max_element(select_system(system), first, last, comp);
+} // end max_element()
+
+
+template <typename ForwardIterator>
+thrust::pair<ForwardIterator,ForwardIterator>
+minmax_element(ForwardIterator first, ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::minmax_element(select_system(system), first, last);
+} // end minmax_element()
+
+
+template <typename ForwardIterator, typename BinaryPredicate>
+thrust::pair<ForwardIterator,ForwardIterator>
+minmax_element(ForwardIterator first, ForwardIterator last, BinaryPredicate comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::minmax_element(select_system(system), first, last, comp);
+} // end minmax_element()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/fill.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/fill.inl
new file mode 100644
index 000000000000..e68672bbebef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/fill.inl
@@ -0,0 +1,86 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/fill.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/fill.h>
+#include <thrust/system/detail/adl/fill.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void fill(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const T &value)
+{
+  using thrust::system::detail::generic::fill;
+  return fill(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+} // end fill()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename OutputIterator, typename Size, typename T>
+__host__ __device__
+  OutputIterator fill_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        OutputIterator first,
+                        Size n,
+                        const T &value)
+{
+  using thrust::system::detail::generic::fill_n;
+  return fill_n(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, n, value);
+} // end fill_n()
+
+
+template<typename ForwardIterator, typename T>
+__host__ __device__
+  void fill(ForwardIterator first,
+            ForwardIterator last,
+            const T &value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  thrust::fill(select_system(system), first, last, value);
+} // end fill()
+
+
+template<typename OutputIterator, typename Size, typename T>
+__host__ __device__
+  OutputIterator fill_n(OutputIterator first,
+                        Size n,
+                        const T &value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<OutputIterator>::type System;
+
+  System system;
+
+  return thrust::fill_n(select_system(system), first, n, value);
+} // end fill()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/find.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/find.inl
new file mode 100644
index 000000000000..5b494f61a60d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/find.inl
@@ -0,0 +1,108 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/find.h>
+#include <thrust/system/detail/adl/find.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+InputIterator find(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   const T& value)
+{
+  using thrust::system::detail::generic::find;
+  return find(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+} // end find()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  using thrust::system::detail::generic::find_if;
+  return find_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end find_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if_not(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          Predicate pred)
+{
+  using thrust::system::detail::generic::find_if_not;
+  return find_if_not(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end find_if_not()
+
+
+template <typename InputIterator, typename T>
+InputIterator find(InputIterator first,
+                   InputIterator last,
+                   const T& value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::find(select_system(system), first, last, value);
+}
+
+template <typename InputIterator, typename Predicate>
+InputIterator find_if(InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::find_if(select_system(system), first, last, pred);
+}
+
+template <typename InputIterator, typename Predicate>
+InputIterator find_if_not(InputIterator first,
+                          InputIterator last,
+                          Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::find_if_not(select_system(system), first, last, pred);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/for_each.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/for_each.inl
new file mode 100644
index 000000000000..4ba39c71a406
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/for_each.inl
@@ -0,0 +1,86 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/for_each.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/for_each.h>
+#include <thrust/system/detail/adl/for_each.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename UnaryFunction>
+__host__ __device__
+  InputIterator for_each(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator first,
+                         InputIterator last,
+                         UnaryFunction f)
+{
+  using thrust::system::detail::generic::for_each;
+
+  return for_each(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, f);
+}
+
+
+template<typename InputIterator,
+         typename UnaryFunction>
+InputIterator for_each(InputIterator first,
+                       InputIterator last,
+                       UnaryFunction f)
+{
+  using thrust::system::detail::generic::select_system;
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+  return thrust::for_each(select_system(system), first, last, f);
+} // end for_each()
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Size, typename UnaryFunction>
+__host__ __device__
+  InputIterator for_each_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first,
+                           Size n,
+                           UnaryFunction f)
+{
+  using thrust::system::detail::generic::for_each_n;
+
+  return for_each_n(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, n, f);
+} // end for_each_n()
+
+
+template<typename InputIterator,
+         typename Size,
+         typename UnaryFunction>
+InputIterator for_each_n(InputIterator first,
+                         Size n,
+                         UnaryFunction f)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+  return thrust::for_each_n(select_system(system), first, n, f);
+} // end for_each_n()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/function.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/function.h
new file mode 100644
index 000000000000..ba20507a5477
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/function.h
@@ -0,0 +1,161 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/raw_reference_cast.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template <typename Function, typename Result>
+struct wrapped_function
+{
+  // mutable because Function::operator() might be const
+  mutable Function m_f;
+
+  inline __host__ __device__
+  wrapped_function()
+      : m_f()
+  {}
+
+  inline __host__ __device__
+  wrapped_function(const Function& f)
+      : m_f(f)
+  {}
+
+  __thrust_exec_check_disable__
+  template <typename Argument>
+  __thrust_forceinline__ __host__ __device__
+  Result operator()(Argument& x) const
+  {
+    return static_cast<Result>(m_f(thrust::raw_reference_cast(x)));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument>
+  __thrust_forceinline__ __host__ __device__
+  Result operator()(const Argument& x) const
+  {
+    return static_cast<Result>(m_f(thrust::raw_reference_cast(x)));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  Result operator()(Argument1& x, Argument2& y) const
+  {
+    return static_cast<Result>(m_f(thrust::raw_reference_cast(x),
+                                   thrust::raw_reference_cast(y)));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  Result operator()(const Argument1& x, Argument2& y) const
+  {
+    return static_cast<Result>(m_f(thrust::raw_reference_cast(x),
+                                   thrust::raw_reference_cast(y)));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  Result operator()(const Argument1& x, const Argument2& y) const
+  {
+    return static_cast<Result>(m_f(thrust::raw_reference_cast(x),
+                                   thrust::raw_reference_cast(y)));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  Result operator()(Argument1& x, const Argument2& y) const
+  {
+    return static_cast<Result>(m_f(thrust::raw_reference_cast(x),
+                                   thrust::raw_reference_cast(y)));
+  }
+}; // end wrapped_function
+
+// Specialize for void return types:
+template <typename Function>
+struct wrapped_function<Function, void>
+{
+  // mutable because Function::operator() might be const
+  mutable Function m_f;
+  inline __host__ __device__
+  wrapped_function()
+    : m_f()
+  {}
+
+  inline __host__ __device__
+  wrapped_function(const Function& f)
+    : m_f(f)
+  {}
+
+  __thrust_exec_check_disable__
+  template <typename Argument>
+  __thrust_forceinline__ __host__ __device__
+  void operator()(Argument& x) const
+  {
+    m_f(thrust::raw_reference_cast(x));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument>
+  __thrust_forceinline__ __host__ __device__
+  void operator()(const Argument& x) const
+  {
+    m_f(thrust::raw_reference_cast(x));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  void operator()(Argument1& x, Argument2& y) const
+  {
+    m_f(thrust::raw_reference_cast(x), thrust::raw_reference_cast(y));
+  }
+
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  void operator()(const Argument1& x, Argument2& y) const
+  {
+    m_f(thrust::raw_reference_cast(x), thrust::raw_reference_cast(y));
+  }
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  void operator()(const Argument1& x, const Argument2& y) const
+  {
+    m_f(thrust::raw_reference_cast(x), thrust::raw_reference_cast(y));
+  }
+  __thrust_exec_check_disable__
+  template <typename Argument1, typename Argument2>
+  __thrust_forceinline__ __host__ __device__
+  void operator()(Argument1& x, const Argument2& y) const
+  {
+    m_f(thrust::raw_reference_cast(x), thrust::raw_reference_cast(y));
+  }
+}; // end wrapped_function
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional.inl
new file mode 100644
index 000000000000..bdf8e0415297
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional.inl
@@ -0,0 +1,126 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<typename Operation>
+  struct unary_traits_imp;
+
+template<typename Operation>
+  struct unary_traits_imp<Operation*>
+{
+  typedef Operation                         function_type;
+  typedef const function_type &             param_type;
+  typedef typename Operation::result_type   result_type;
+  typedef typename Operation::argument_type argument_type;
+}; // end unary_traits_imp
+
+template<typename Result, typename Argument>
+  struct unary_traits_imp<Result(*)(Argument)>
+{
+  typedef Result   (*function_type)(Argument);
+  typedef Result   (*param_type)(Argument);
+  typedef Result   result_type;
+  typedef Argument argument_type;
+}; // end unary_traits_imp
+
+template<typename Operation>
+  struct binary_traits_imp;
+
+template<typename Operation>
+  struct binary_traits_imp<Operation*>
+{
+  typedef Operation                                function_type;
+  typedef const function_type &                    param_type;
+  typedef typename Operation::result_type          result_type;
+  typedef typename Operation::first_argument_type  first_argument_type;
+  typedef typename Operation::second_argument_type second_argument_type;
+}; // end binary_traits_imp
+
+template<typename Result, typename Argument1, typename Argument2>
+  struct binary_traits_imp<Result(*)(Argument1, Argument2)>
+{
+  typedef Result (*function_type)(Argument1, Argument2);
+  typedef Result (*param_type)(Argument1, Argument2);
+  typedef Result result_type;
+  typedef Argument1 first_argument_type;
+  typedef Argument2 second_argument_type;
+}; // end binary_traits_imp
+
+} // end detail
+
+template<typename Operation>
+  struct unary_traits
+{
+  typedef typename detail::unary_traits_imp<Operation*>::function_type function_type;
+  typedef typename detail::unary_traits_imp<Operation*>::param_type    param_type;
+  typedef typename detail::unary_traits_imp<Operation*>::result_type   result_type;
+  typedef typename detail::unary_traits_imp<Operation*>::argument_type argument_type;
+}; // end unary_traits
+
+template<typename Result, typename Argument>
+  struct unary_traits<Result(*)(Argument)>
+{
+  typedef Result   (*function_type)(Argument);
+  typedef Result   (*param_type)(Argument);
+  typedef Result   result_type;
+  typedef Argument argument_type;
+}; // end unary_traits
+
+template<typename Operation>
+  struct binary_traits
+{
+  typedef typename detail::binary_traits_imp<Operation*>::function_type        function_type;
+  typedef typename detail::binary_traits_imp<Operation*>::param_type           param_type;
+  typedef typename detail::binary_traits_imp<Operation*>::result_type          result_type;
+  typedef typename detail::binary_traits_imp<Operation*>::first_argument_type  first_argument_type;
+  typedef typename detail::binary_traits_imp<Operation*>::second_argument_type second_argument_type;
+}; // end binary_traits
+
+template<typename Result, typename Argument1, typename Argument2>
+  struct binary_traits<Result(*)(Argument1, Argument2)>
+{
+  typedef Result (*function_type)(Argument1, Argument2);
+  typedef Result (*param_type)(Argument1, Argument2);
+  typedef Result result_type;
+  typedef Argument1 first_argument_type;
+  typedef Argument2 second_argument_type;
+}; // end binary_traits
+
+template<typename Predicate>
+  __host__ __device__
+  unary_negate<Predicate> not1(const Predicate &pred)
+{
+  return unary_negate<Predicate>(pred);
+} // end not1()
+
+template<typename BinaryPredicate>
+  __host__ __device__
+  binary_negate<BinaryPredicate> not2(const BinaryPredicate &pred)
+{
+  return binary_negate<BinaryPredicate>(pred);
+} // end not2()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.h
new file mode 100644
index 000000000000..cee0770a46b6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.h
@@ -0,0 +1,155 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// Portions of this code are derived from
+//
+// Manjunath Kudlur's Carbon library
+//
+// and
+//
+// Based on Boost.Phoenix v1.2
+// Copyright (c) 2001-2002 Joel de Guzman
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/tuple.h>
+#include <thrust/detail/functional/value.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/assignment_operator.h>
+#include <thrust/detail/raw_reference_cast.h>
+#include <thrust/detail/type_traits/result_of_adaptable_function.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+// eval_ref<T> is
+// - T when T is a subclass of thrust::reference
+// - T& otherwise
+// This is used to let thrust::references pass through actor evaluations.
+template <typename T>
+using eval_ref = typename std::conditional<
+  thrust::detail::is_wrapped_reference<T>::value, T, T&>::type;
+
+template<typename Action, typename Env>
+  struct apply_actor
+{
+  typedef typename Action::template result<Env>::type type;
+};
+
+template<typename Eval>
+  struct actor
+    : Eval
+{
+  typedef Eval eval_type;
+
+  __host__ __device__
+  constexpr actor();
+
+  __host__ __device__
+  actor(const Eval &base);
+
+  __host__ __device__
+  typename apply_actor<eval_type, thrust::null_type >::type
+  operator()(void) const;
+
+  template <typename... Ts>
+  __host__ __device__
+  typename apply_actor<eval_type, thrust::tuple<eval_ref<Ts>...>>::type
+  operator()(Ts&&... ts) const;
+
+  template<typename T>
+  __host__ __device__
+  typename assign_result<Eval,T>::type
+  operator=(const T &_1) const;
+}; // end actor
+
+// in general, as_actor should turn things into values
+template<typename T>
+  struct as_actor
+{
+  typedef value<T> type;
+
+  static inline __host__ __device__ type convert(const T &x)
+  {
+    return val(x);
+  } // end convert()
+}; // end as_actor
+
+// specialization for things which are already actors
+template<typename Eval>
+  struct as_actor<actor<Eval> >
+{
+  typedef actor<Eval> type;
+
+  static inline __host__ __device__ const type &convert(const actor<Eval> &x)
+  {
+    return x;
+  } // end convert()
+}; // end as_actor
+
+template<typename T>
+  typename as_actor<T>::type
+  __host__ __device__
+    make_actor(const T &x)
+{
+  return as_actor<T>::convert(x);
+} // end make_actor()
+
+} // end functional
+
+// provide specializations for result_of for nullary, unary, and binary invocations of actor
+template<typename Eval>
+  struct result_of_adaptable_function<
+    thrust::detail::functional::actor<Eval>()
+  >
+{
+  typedef typename thrust::detail::functional::apply_actor<
+    thrust::detail::functional::actor<Eval>,
+    thrust::null_type
+  >::type type;
+}; // end result_of
+
+template<typename Eval, typename Arg1>
+  struct result_of_adaptable_function<
+    thrust::detail::functional::actor<Eval>(Arg1)
+  >
+{
+  typedef typename thrust::detail::functional::apply_actor<
+    thrust::detail::functional::actor<Eval>,
+    thrust::tuple<Arg1>
+  >::type type;
+}; // end result_of
+
+template<typename Eval, typename Arg1, typename Arg2>
+  struct result_of_adaptable_function<
+    thrust::detail::functional::actor<Eval>(Arg1,Arg2)
+  >
+{
+  typedef typename thrust::detail::functional::apply_actor<
+    thrust::detail::functional::actor<Eval>,
+    thrust::tuple<Arg1,Arg2>
+  >::type type;
+}; // end result_of
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/functional/actor.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.inl
new file mode 100644
index 000000000000..e0bdebbbf3ea
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/actor.inl
@@ -0,0 +1,114 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// Portions of this code are derived from
+//
+// Manjunath Kudlur's Carbon library
+//
+// and
+//
+// Based on Boost.Phoenix v1.2
+// Copyright (c) 2001-2002 Joel de Guzman
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/assignment_operator.h>
+#include <thrust/functional.h>
+#include <thrust/type_traits/logical_metafunctions.h>
+
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+namespace functional
+{
+
+template<typename Eval>
+  __host__ __device__
+  constexpr actor<Eval>
+    ::actor()
+      : eval_type()
+{}
+
+template<typename Eval>
+  __host__ __device__
+  actor<Eval>
+    ::actor(const Eval &base)
+      : eval_type(base)
+{}
+
+template<typename Eval>
+  __host__ __device__
+  typename apply_actor<
+    typename actor<Eval>::eval_type,
+    typename thrust::null_type
+  >::type
+    actor<Eval>
+      ::operator()(void) const
+{
+  return eval_type::eval(thrust::null_type());
+} // end basic_environment::operator()
+
+// actor::operator() needs to construct a tuple of references to its
+// arguments. To make this work with thrust::reference<T>, we need to
+// detect thrust proxy references and store them as T rather than T&.
+// This check ensures that the forwarding references passed into
+// actor::operator() are either:
+// - T&& if and only if T is a thrust::reference<U>, or
+// - T& for any other types.
+// This struct provides a nicer diagnostic for when these conditions aren't
+// met.
+template <typename T>
+using actor_check_ref_type =
+  thrust::detail::integral_constant<bool,
+    ( std::is_lvalue_reference<T>::value ||
+      thrust::detail::is_wrapped_reference<T>::value )>;
+
+template <typename... Ts>
+using actor_check_ref_types =
+  thrust::conjunction<actor_check_ref_type<Ts>...>;
+
+template<typename Eval>
+template<typename... Ts>
+__host__ __device__
+typename apply_actor<typename actor<Eval>::eval_type,
+                     thrust::tuple<eval_ref<Ts>...>>::type
+actor<Eval>::operator()(Ts&&... ts) const
+{
+  static_assert(actor_check_ref_types<Ts...>::value,
+                "Actor evaluations only support rvalue references to "
+                "thrust::reference subclasses.");
+  using tuple_type = thrust::tuple<eval_ref<Ts>...>;
+  return eval_type::eval(tuple_type(THRUST_FWD(ts)...));
+} // end actor<Eval>::operator()
+
+template<typename Eval>
+  template<typename T>
+    __host__ __device__
+    typename assign_result<Eval,T>::type
+      actor<Eval>
+        ::operator=(const T& _1) const
+{
+  return do_assign(*this,_1);
+} // end actor::operator=()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/argument.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/argument.h
new file mode 100644
index 000000000000..aac29f53713e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/argument.h
@@ -0,0 +1,74 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// Portions of this code are derived from
+//
+// Manjunath Kudlur's Carbon library
+//
+// and
+//
+// Based on Boost.Phoenix v1.2
+// Copyright (c) 2001-2002 Joel de Guzman
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/tuple.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+template<unsigned int i, typename Env>
+  struct argument_helper
+{
+  typedef typename thrust::tuple_element<i,Env>::type type;
+};
+
+template<unsigned int i>
+  struct argument_helper<i,thrust::null_type>
+{
+  typedef thrust::null_type type;
+};
+
+
+template<unsigned int i>
+  class argument
+{
+  public:
+    template<typename Env>
+      struct result
+        : argument_helper<i,Env>
+    {
+    };
+
+    __host__ __device__
+    constexpr argument(){}
+
+    template<typename Env>
+    __host__ __device__
+    typename result<Env>::type eval(const Env &e) const
+    {
+      return thrust::get<i>(e);
+    } // end eval()
+}; // end argument
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/composite.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/composite.h
new file mode 100644
index 000000000000..41ee74739254
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/composite.h
@@ -0,0 +1,164 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// Portions of this code are derived from
+//
+// Manjunath Kudlur's Carbon library
+//
+// and
+//
+// Based on Boost.Phoenix v1.2
+// Copyright (c) 2001-2002 Joel de Guzman
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/functional/actor.h>
+#include <thrust/tuple.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+// XXX we should just take a single EvalTuple
+template<typename Eval0,
+         typename Eval1  = thrust::null_type,
+         typename Eval2  = thrust::null_type,
+         typename Eval3  = thrust::null_type,
+         typename Eval4  = thrust::null_type,
+         typename Eval5  = thrust::null_type,
+         typename Eval6  = thrust::null_type,
+         typename Eval7  = thrust::null_type,
+         typename Eval8  = thrust::null_type,
+         typename Eval9  = thrust::null_type,
+         typename Eval10 = thrust::null_type>
+  class composite;
+
+template<typename Eval0, typename Eval1>
+  class composite<
+    Eval0,
+    Eval1,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type
+  >
+{
+  public:
+    template<typename Env>
+      struct result
+    {
+      typedef typename Eval0::template result<
+        thrust::tuple<
+          typename Eval1::template result<Env>::type
+        >
+      >::type type;
+    };
+
+    __host__ __device__
+    composite(const Eval0 &e0, const Eval1 &e1)
+      : m_eval0(e0),
+        m_eval1(e1)
+    {}
+
+    template<typename Env>
+    __host__ __device__
+    typename result<Env>::type
+    eval(const Env &x) const
+    {
+      typename Eval1::template result<Env>::type result1 = m_eval1.eval(x);
+      return m_eval0.eval(thrust::tie(result1));
+    }
+
+  private:
+    Eval0 m_eval0;
+    Eval1 m_eval1;
+}; // end composite<Eval0,Eval1>
+
+template<typename Eval0, typename Eval1, typename Eval2>
+  class composite<
+    Eval0,
+    Eval1,
+    Eval2,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type,
+    thrust::null_type
+  >
+{
+  public:
+    template<typename Env>
+      struct result
+    {
+      typedef typename Eval0::template result<
+        thrust::tuple<
+          typename Eval1::template result<Env>::type,
+          typename Eval2::template result<Env>::type
+        >
+      >::type type;
+    };
+
+    __host__ __device__
+    composite(const Eval0 &e0, const Eval1 &e1, const Eval2 &e2)
+      : m_eval0(e0),
+        m_eval1(e1),
+        m_eval2(e2)
+    {}
+
+    template<typename Env>
+    __host__ __device__
+    typename result<Env>::type
+    eval(const Env &x) const
+    {
+      typename Eval1::template result<Env>::type result1 = m_eval1.eval(x);
+      typename Eval2::template result<Env>::type result2 = m_eval2.eval(x);
+      return m_eval0.eval(thrust::tie(result1,result2));
+    }
+
+  private:
+    Eval0 m_eval0;
+    Eval1 m_eval1;
+    Eval2 m_eval2;
+}; // end composite<Eval0,Eval1,Eval2>
+
+template<typename Eval0, typename Eval1>
+__host__ __device__
+  actor<composite<Eval0,Eval1> > compose(const Eval0 &e0, const Eval1 &e1)
+{
+  return actor<composite<Eval0,Eval1> >(composite<Eval0,Eval1>(e0,e1));
+}
+
+template<typename Eval0, typename Eval1, typename Eval2>
+__host__ __device__
+  actor<composite<Eval0,Eval1,Eval2> > compose(const Eval0 &e0, const Eval1 &e1, const Eval2 &e2)
+{
+  return actor<composite<Eval0,Eval1,Eval2> >(composite<Eval0,Eval1,Eval2>(e0,e1,e2));
+}
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators.h
new file mode 100644
index 000000000000..f86ea2052181
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators.h
@@ -0,0 +1,25 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/operators/arithmetic_operators.h>
+#include <thrust/detail/functional/operators/relational_operators.h>
+#include <thrust/detail/functional/operators/logical_operators.h>
+#include <thrust/detail/functional/operators/bitwise_operators.h>
+#include <thrust/detail/functional/operators/compound_assignment_operators.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/arithmetic_operators.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/arithmetic_operators.h
new file mode 100644
index 000000000000..443d307cbd60
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/arithmetic_operators.h
@@ -0,0 +1,436 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/operator_adaptors.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<thrust::negate<>>,
+    actor<Eval>
+  >
+>
+__host__ __device__
+operator-(const actor<Eval> &_1)
+{
+  return compose(transparent_unary_operator<thrust::negate<>>(), _1);
+} // end operator-()
+
+// there's no standard unary_plus functional, so roll an ad hoc one here
+struct unary_plus
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1) const
+  noexcept(noexcept(+THRUST_FWD(t1)))
+  THRUST_TRAILING_RETURN(decltype(+THRUST_FWD(t1)))
+  {
+    return +THRUST_FWD(t1);
+  }
+};
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<unary_plus>,
+    actor<Eval>
+  >
+>
+operator+(const actor<Eval> &_1)
+{
+  return compose(transparent_unary_operator<unary_plus>(), _1);
+} // end operator+()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::plus<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator+(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::plus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator+()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::plus<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator+(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::plus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator+()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::plus<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator+(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::plus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator+()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::minus<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator-(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::minus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator-()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::minus<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator-(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::minus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator-()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::minus<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator-(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::minus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator-()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::multiplies<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator*(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::multiplies<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator*()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::multiplies<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator*(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::multiplies<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator*()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::multiplies<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator*(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::multiplies<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator*()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::divides<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator/(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::divides<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator/()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::divides<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator/(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::divides<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator/()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::divides<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator/(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::divides<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator/()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::modulus<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator%(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::modulus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator%()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::modulus<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator%(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::modulus<void>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator%()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::modulus<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator%(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::modulus<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator%()
+
+// there's no standard prefix_increment functional, so roll an ad hoc one here
+struct prefix_increment
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1) const
+  noexcept(noexcept(++THRUST_FWD(t1)))
+  THRUST_TRAILING_RETURN(decltype(++THRUST_FWD(t1)))
+  {
+    return ++THRUST_FWD(t1);
+  }
+}; // end prefix_increment
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<prefix_increment>,
+    actor<Eval>
+  >
+>
+operator++(const actor<Eval> &_1)
+{
+  return compose(transparent_unary_operator<prefix_increment>(), _1);
+} // end operator++()
+
+
+// there's no standard postfix_increment functional, so roll an ad hoc one here
+struct postfix_increment
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1) const
+  noexcept(noexcept(THRUST_FWD(t1)++))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1)++))
+  {
+    return THRUST_FWD(t1)++;
+  }
+}; // end postfix_increment
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<postfix_increment>,
+    actor<Eval>
+  >
+>
+operator++(const actor<Eval> &_1, int)
+{
+  return compose(transparent_unary_operator<postfix_increment>(), _1);
+} // end operator++()
+
+
+// there's no standard prefix_decrement functional, so roll an ad hoc one here
+struct prefix_decrement
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1) const
+  noexcept(noexcept(--THRUST_FWD(t1)))
+  THRUST_TRAILING_RETURN(decltype(--THRUST_FWD(t1)))
+  {
+    return --THRUST_FWD(t1);
+  }
+}; // end prefix_decrement
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<prefix_decrement>,
+    actor<Eval>
+  >
+>
+operator--(const actor<Eval> &_1)
+{
+  return compose(transparent_unary_operator<prefix_decrement>(), _1);
+} // end operator--()
+
+
+// there's no standard postfix_decrement functional, so roll an ad hoc one here
+struct postfix_decrement
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1) const
+  noexcept(noexcept(THRUST_FWD(t1)--))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1)--))
+  {
+    return THRUST_FWD(t1)--;
+  }
+}; // end prefix_increment
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<postfix_decrement>,
+    actor<Eval>
+  >
+>
+operator--(const actor<Eval> &_1, int)
+{
+  return compose(transparent_unary_operator<postfix_decrement>(), _1);
+} // end operator--()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/assignment_operator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/assignment_operator.h
new file mode 100644
index 000000000000..870354b6f0b5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/assignment_operator.h
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/operator_adaptors.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// XXX WAR circular inclusion with this forward declaration
+template<typename,typename,typename> struct binary_function;
+
+namespace detail
+{
+namespace functional
+{
+
+// XXX WAR circular inclusion with this forward declaration
+template<typename> struct as_actor;
+
+// there's no standard assign functional, so roll an ad hoc one here
+struct assign
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) = THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) = THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) = THRUST_FWD(t2);
+  }
+};
+
+template<typename Eval, typename T>
+  struct assign_result
+{
+  typedef actor<
+    composite<
+      transparent_binary_operator<assign>,
+      actor<Eval>,
+      typename as_actor<T>::type
+    >
+  > type;
+}; // end assign_result
+
+template<typename Eval, typename T>
+  __host__ __device__
+    typename assign_result<Eval,T>::type
+      do_assign(const actor<Eval> &_1, const T &_2)
+{
+  return compose(transparent_binary_operator<assign>(),
+                 _1,
+                 as_actor<T>::convert(_2));
+} // end do_assign()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/bitwise_operators.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/bitwise_operators.h
new file mode 100644
index 000000000000..065cd15403c7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/bitwise_operators.h
@@ -0,0 +1,338 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/operator_adaptors.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_and<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator&(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_and<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_and<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator&(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_and<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_and<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator&(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_and<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_or<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator|(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_or<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_or<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator|(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_or<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_or<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator|(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_or<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_xor<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator^(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_xor<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator^()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_xor<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator^(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_xor<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator^()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_xor<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator^(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_xor<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator^()
+
+
+// there's no standard bit_not functional, so roll an ad hoc one here
+struct bit_not
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1) const
+  noexcept(noexcept(~THRUST_FWD(t1)))
+  THRUST_TRAILING_RETURN(decltype(~THRUST_FWD(t1)))
+  {
+    return ~THRUST_FWD(t1);
+  }
+}; // end prefix_increment
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<bit_not>,
+    actor<Eval>
+  >
+>
+__host__ __device__
+operator~(const actor<Eval> &_1)
+{
+  return compose(transparent_unary_operator<bit_not>(), _1);
+} // end operator~()
+
+// there's no standard bit_lshift functional, so roll an ad hoc one here
+struct bit_lshift
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) << THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) << THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) << THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_lshift>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator<<(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_lshift>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<<()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_lshift>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator<<(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_lshift>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<<()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_lshift>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator<<(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_lshift>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<<()
+
+// there's no standard bit_rshift functional, so roll an ad hoc one here
+struct bit_rshift
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) >> THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) >> THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) >> THRUST_FWD(t2);
+  }
+};
+
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_rshift>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator>>(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_rshift>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>>()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_rshift>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator>>(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_rshift>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>>()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_rshift>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator>>(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_rshift>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>>()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/compound_assignment_operators.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/compound_assignment_operators.h
new file mode 100644
index 000000000000..b5ba77fb48d9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/compound_assignment_operators.h
@@ -0,0 +1,512 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/operator_adaptors.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+// there's no standard plus_equal functional, so roll an ad hoc one here
+struct plus_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) += THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) += THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) += THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<plus_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator+=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<plus_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator+=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<plus_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator+=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<plus_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator+=()
+
+// there's no standard minus_equal functional, so roll an ad hoc one here
+struct minus_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) -= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) -= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) -= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<minus_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator-=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<minus_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator-=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<minus_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator-=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<minus_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator-=()
+
+// there's no standard multiplies_equal functional, so roll an ad hoc one here
+struct multiplies_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) *= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) *= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) *= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<multiplies_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator*=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<multiplies_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator*=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<multiplies_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator*=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<multiplies_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator*=()
+
+// there's no standard divides_equal functional, so roll an ad hoc one here
+struct divides_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) /= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) /= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) /= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<divides_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator/=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<divides_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator/=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<divides_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator/=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<divides_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator/=()
+
+// there's no standard modulus_equal functional, so roll an ad hoc one here
+struct modulus_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) %= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) %= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) %= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<modulus_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator%=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<modulus_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator%=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<modulus_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator%=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<modulus_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator%=()
+
+// there's no standard bit_and_equal functional, so roll an ad hoc one here
+struct bit_and_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) &= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) &= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) &= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_and_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator&=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_and_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_and_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator&=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_and_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&=()
+
+// there's no standard bit_or_equal functional, so roll an ad hoc one here
+struct bit_or_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) |= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) |= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) |= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_or_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator|=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_or_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_or_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator|=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_or_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|=()
+
+// there's no standard bit_xor_equal functional, so roll an ad hoc one here
+struct bit_xor_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) ^= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) ^= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) ^= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_xor_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator^=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_xor_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_xor_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator^=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_xor_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator|=()
+
+// there's no standard bit_lshift_equal functional, so roll an ad hoc one here
+struct bit_lshift_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) <<= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) <<= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) <<= THRUST_FWD(t2);
+  }
+};
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_lshift_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator<<=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_lshift_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<<=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_lshift_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator<<=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_lshift_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<<=()
+
+// there's no standard bit_rshift_equal functional, so roll an ad hoc one here
+struct bit_rshift_equal
+{
+  using is_transparent = void;
+
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t1) >>= THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1) >>= THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t1) >>= THRUST_FWD(t2);
+  }
+};
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_rshift_equal>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator>>=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<bit_rshift_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>>=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<bit_rshift_equal>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator>>=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<bit_rshift_equal>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>>=()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/logical_operators.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/logical_operators.h
new file mode 100644
index 000000000000..e1e4ff71994c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/logical_operators.h
@@ -0,0 +1,143 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/operator_adaptors.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::logical_and<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator&&(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::logical_and<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::logical_and<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator&&(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::logical_and<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::logical_and<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator&&(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::logical_and<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::logical_or<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator||(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::logical_or<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::logical_or<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator||(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::logical_or<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&&()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::logical_or<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator||(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::logical_or<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator&&()
+
+template<typename Eval>
+__host__ __device__
+actor<
+  composite<
+    transparent_unary_operator<thrust::logical_not<>>,
+    actor<Eval>
+  >
+>
+operator!(const actor<Eval> &_1)
+{
+  return compose(transparent_unary_operator<thrust::logical_not<>>(), _1);
+} // end operator!()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/operator_adaptors.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/operator_adaptors.h
new file mode 100644
index 000000000000..67326c5c1e55
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/operator_adaptors.h
@@ -0,0 +1,136 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/argument.h>
+#include <thrust/detail/type_deduction.h>
+#include <thrust/tuple.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/type_traits/void_t.h>
+
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+// Adapts a transparent unary functor from functional.h (e.g. thrust::negate<>)
+// into the Eval interface.
+template <typename UnaryFunctor>
+struct transparent_unary_operator
+{
+  template <typename>
+  using operator_type = UnaryFunctor;
+
+  template <typename Env>
+  using argument =
+  typename thrust::detail::eval_if<
+    thrust::tuple_size<Env>::value != 1,
+    thrust::detail::identity_<thrust::null_type>,
+    thrust::detail::functional::argument_helper<0, Env>
+  >::type;
+
+  template <typename Env>
+  struct result_type_impl
+  {
+    using type = decltype(
+      std::declval<UnaryFunctor>()(std::declval<argument<Env>>()));
+  };
+
+  template <typename Env>
+  using result_type =
+  typename thrust::detail::eval_if<
+    std::is_same<thrust::null_type, argument<Env>>::value,
+    thrust::detail::identity_<thrust::null_type>,
+    result_type_impl<Env>
+  >::type;
+
+  template <typename Env>
+  struct result
+  {
+    using op_type = UnaryFunctor;
+    using type = result_type<Env>;
+  };
+
+  template <typename Env>
+  __host__ __device__
+  result_type<Env> eval(Env&& e) const
+  THRUST_RETURNS(UnaryFunctor{}(thrust::get<0>(THRUST_FWD(e))))
+};
+
+
+// Adapts a transparent binary functor from functional.h (e.g. thrust::less<>)
+// into the Eval interface.
+template <typename BinaryFunctor>
+struct transparent_binary_operator
+{
+  template <typename>
+  using operator_type = BinaryFunctor;
+
+  template <typename Env>
+  using first_argument =
+    typename thrust::detail::eval_if<
+      thrust::tuple_size<Env>::value != 2,
+      thrust::detail::identity_<thrust::null_type>,
+      thrust::detail::functional::argument_helper<0, Env>
+    >::type;
+
+  template <typename Env>
+  using second_argument =
+    typename thrust::detail::eval_if<
+      thrust::tuple_size<Env>::value != 2,
+      thrust::detail::identity_<thrust::null_type>,
+      thrust::detail::functional::argument_helper<1, Env>
+    >::type;
+
+  template <typename Env>
+  struct result_type_impl
+  {
+    using type = decltype(
+      std::declval<BinaryFunctor>()(std::declval<first_argument<Env>>(),
+                                    std::declval<second_argument<Env>>()));
+  };
+
+  template <typename Env>
+  using result_type =
+    typename thrust::detail::eval_if<
+      (std::is_same<thrust::null_type, first_argument<Env>>::value ||
+       std::is_same<thrust::null_type, second_argument<Env>>::value),
+      thrust::detail::identity_<thrust::null_type>,
+      result_type_impl<Env>
+    >::type;
+
+  template <typename Env>
+  struct result
+  {
+    using op_type = BinaryFunctor;
+    using type = result_type<Env>;
+  };
+
+  template <typename Env>
+  __host__ __device__
+  result_type<Env> eval(Env&& e) const
+  THRUST_RETURNS(BinaryFunctor{}(thrust::get<0>(e), thrust::get<1>(e)))
+};
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/relational_operators.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/relational_operators.h
new file mode 100644
index 000000000000..6c58325e2251
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/operators/relational_operators.h
@@ -0,0 +1,322 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/composite.h>
+#include <thrust/detail/functional/operators/operator_adaptors.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::equal_to<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator==(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::equal_to<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator==()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::equal_to<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator==(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::equal_to<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator==()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::equal_to<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator==(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::equal_to<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator==()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::not_equal_to<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator!=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::not_equal_to<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator!=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::not_equal_to<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator!=(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::not_equal_to<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator!=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::not_equal_to<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator!=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::not_equal_to<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator!=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::greater<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator>(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::greater<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::greater<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator>(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::greater<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::greater<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator>(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::greater<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::less<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator<(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::less<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::less<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator<(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::less<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::less<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator<(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::less<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::greater_equal<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator>=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::greater_equal<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::greater_equal<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator>=(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::greater_equal<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::greater_equal<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator>=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::greater_equal<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator>=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::less_equal<>>,
+    actor<T1>,
+    typename as_actor<T2>::type
+  >
+>
+operator<=(const actor<T1> &_1, const T2 &_2)
+{
+  return compose(transparent_binary_operator<thrust::less_equal<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::less_equal<>>,
+    typename as_actor<T1>::type,
+    actor<T2>
+  >
+>
+operator<=(const T1 &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::less_equal<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<=()
+
+template<typename T1, typename T2>
+__host__ __device__
+actor<
+  composite<
+    transparent_binary_operator<thrust::less_equal<>>,
+    actor<T1>,
+    actor<T2>
+  >
+>
+operator<=(const actor<T1> &_1, const actor<T2> &_2)
+{
+  return compose(transparent_binary_operator<thrust::less_equal<>>(),
+                 make_actor(_1),
+                 make_actor(_2));
+} // end operator<=()
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/placeholder.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/placeholder.h
new file mode 100644
index 000000000000..e3c0835531f0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/placeholder.h
@@ -0,0 +1,38 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+#include <thrust/detail/functional/argument.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+template<unsigned int i>
+  struct placeholder
+{
+  typedef actor<argument<i> > type;
+};
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/value.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/value.h
new file mode 100644
index 000000000000..d6b1563b15c5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/functional/value.h
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// Portions of this code are derived from
+//
+// Manjunath Kudlur's Carbon library
+//
+// and
+//
+// Based on Boost.Phoenix v1.2
+// Copyright (c) 2001-2002 Joel de Guzman
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/functional/actor.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace functional
+{
+
+
+template<typename Eval> struct actor;
+
+
+template<typename T>
+  class value
+{
+  public:
+
+    template<typename Env>
+      struct result
+    {
+      typedef T type;
+    };
+
+    __host__ __device__
+    value(const T &arg)
+      : m_val(arg)
+    {}
+
+    template<typename Env>
+    __host__ __device__
+      T eval(const Env &) const
+    {
+      return m_val;
+    }
+
+  private:
+    T m_val;
+}; // end value
+
+template<typename T>
+__host__ __device__
+actor<value<T> > val(const T &x)
+{
+  return value<T>(x);
+} // end val()
+
+
+} // end functional
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/gather.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/gather.inl
new file mode 100644
index 000000000000..3812702f61c6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/gather.inl
@@ -0,0 +1,161 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/gather.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/gather.h>
+#include <thrust/system/detail/adl/gather.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        InputIterator                                               map_first,
+                        InputIterator                                               map_last,
+                        RandomAccessIterator                                        input_first,
+                        OutputIterator                                              result)
+{
+  using thrust::system::detail::generic::gather;
+  return gather(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), map_first, map_last, input_first, result);
+} // end gather()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1                                              map_first,
+                           InputIterator1                                              map_last,
+                           InputIterator2                                              stencil,
+                           RandomAccessIterator                                        input_first,
+                           OutputIterator                                              result)
+{
+  using thrust::system::detail::generic::gather_if;
+  return gather_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), map_first, map_last, stencil, input_first, result);
+} // end gather_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator gather_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1                                              map_first,
+                           InputIterator1                                              map_last,
+                           InputIterator2                                              stencil,
+                           RandomAccessIterator                                        input_first,
+                           OutputIterator                                              result,
+                           Predicate                                                   pred)
+{
+  using thrust::system::detail::generic::gather_if;
+  return gather_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), map_first, map_last, stencil, input_first, result, pred);
+} // end gather_if()
+
+
+template<typename InputIterator,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+  OutputIterator gather(InputIterator        map_first,
+                        InputIterator        map_last,
+                        RandomAccessIterator input_first,
+                        OutputIterator       result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type        System1;
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type       System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::gather(select_system(system1,system2,system3), map_first, map_last, input_first, result);
+} // end gather()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+  OutputIterator gather_if(InputIterator1       map_first,
+                           InputIterator1       map_last,
+                           InputIterator2       stencil,
+                           RandomAccessIterator input_first,
+                           OutputIterator       result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type       System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type       System2;
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator>::type       System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::gather_if(select_system(system1,system2,system3,system4), map_first, map_last, stencil, input_first, result);
+} // end gather_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator gather_if(InputIterator1       map_first,
+                           InputIterator1       map_last,
+                           InputIterator2       stencil,
+                           RandomAccessIterator input_first,
+                           OutputIterator       result,
+                           Predicate            pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type       System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type       System2;
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator>::type       System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::gather_if(select_system(system1,system2,system3,system4), map_first, map_last, stencil, input_first, result, pred);
+} // end gather_if()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/generate.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/generate.inl
new file mode 100644
index 000000000000..2ecb65d5843b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/generate.inl
@@ -0,0 +1,92 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/generate.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/generate.h>
+#include <thrust/system/detail/adl/generate.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Generator>
+__host__ __device__
+  void generate(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                Generator gen)
+{
+  using thrust::system::detail::generic::generate;
+  return generate(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, gen);
+} // end generate()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename OutputIterator,
+         typename Size,
+         typename Generator>
+__host__ __device__
+  OutputIterator generate_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            OutputIterator first,
+                            Size n,
+                            Generator gen)
+{
+  using thrust::system::detail::generic::generate_n;
+  return generate_n(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, n, gen);
+} // end generate_n()
+
+
+template<typename ForwardIterator,
+         typename Generator>
+  void generate(ForwardIterator first,
+                ForwardIterator last,
+                Generator gen)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::generate(select_system(system), first, last, gen);
+} // end generate()
+
+
+template<typename OutputIterator,
+         typename Size,
+         typename Generator>
+  OutputIterator generate_n(OutputIterator first,
+                            Size n,
+                            Generator gen)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<OutputIterator>::type System;
+
+  System system;
+
+  return thrust::generate_n(select_system(system), first, n, gen);
+} // end generate_n()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/get_iterator_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/get_iterator_value.h
new file mode 100644
index 000000000000..27e0a4e47a0b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/get_iterator_value.h
@@ -0,0 +1,55 @@
+#pragma once
+/*
+ *  Copyright 2008-2016 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/execution_policy.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/system/detail/generic/memory.h> // for get_value()
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail {
+
+// get_iterator_value specialization on iterators
+// --------------------------------------------------
+// it is okay to dereference iterator in the usual way
+template<typename DerivedPolicy, typename Iterator>
+__host__ __device__
+typename thrust::iterator_traits<Iterator>::value_type
+get_iterator_value(thrust::execution_policy<DerivedPolicy> &, Iterator it)
+{
+  return *it;
+} // get_iterator_value(exec,Iterator);
+
+// get_iterator_value specialization on pointer
+// ----------------------------------------------
+// we can't just dereference a pointer in the usual way, because
+// it may point to a location in the device memory. 
+// we use get_value(exec,pointer*) function
+// to perform a dereferencing consistent with the execution policy
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+typename thrust::detail::pointer_traits<Pointer*>::element_type 
+get_iterator_value(thrust::execution_policy<DerivedPolicy> &exec, Pointer* ptr)
+{
+  return get_value(derived_cast(exec),ptr);
+} // get_iterator_value(exec,Pointer*)
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/inner_product.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/inner_product.inl
new file mode 100644
index 000000000000..97cd2b0b52e6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/inner_product.inl
@@ -0,0 +1,101 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/inner_product.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/inner_product.h>
+#include <thrust/system/detail/adl/inner_product.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputType>
+__host__ __device__
+OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init)
+{
+  using thrust::system::detail::generic::inner_product;
+  return inner_product(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, init);
+} // end inner_product()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputType,
+         typename BinaryFunction1,
+         typename BinaryFunction2>
+__host__ __device__
+OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init,
+                         BinaryFunction1 binary_op1,
+                         BinaryFunction2 binary_op2)
+{
+  using thrust::system::detail::generic::inner_product;
+  return inner_product(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, init, binary_op1, binary_op2);
+} // end inner_product()
+
+
+template<typename InputIterator1, typename InputIterator2, typename OutputType>
+OutputType
+inner_product(InputIterator1 first1, InputIterator1 last1,
+              InputIterator2 first2, OutputType init)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::inner_product(select_system(system1,system2), first1, last1, first2, init);
+} // end inner_product()
+
+
+template<typename InputIterator1, typename InputIterator2, typename OutputType,
+         typename BinaryFunction1, typename BinaryFunction2>
+OutputType
+inner_product(InputIterator1 first1, InputIterator1 last1,
+              InputIterator2 first2, OutputType init,
+              BinaryFunction1 binary_op1, BinaryFunction2 binary_op2)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::inner_product(select_system(system1,system2), first1, last1, first2, init, binary_op1, binary_op2);
+} // end inner_product()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_math.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_math.h
new file mode 100644
index 000000000000..0f8c8aac1688
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_math.h
@@ -0,0 +1,152 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_deduction.h>
+
+#include <nv/target>
+
+#include <limits>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template <typename Integer>
+__host__ __device__ __thrust_forceinline__
+Integer clz(Integer x)
+{
+  Integer result;
+
+  NV_IF_TARGET(NV_IS_DEVICE, (
+    result = ::__clz(x);
+  ), (
+    int num_bits = 8 * sizeof(Integer);
+    int num_bits_minus_one = num_bits - 1;
+    result = num_bits;
+    for (int i = num_bits_minus_one; i >= 0; --i)
+    {
+      if ((Integer(1) << i) & x)
+      {
+        result = num_bits_minus_one - i;
+        break;
+      }
+    }
+  ));
+
+  return result;
+}
+
+template <typename Integer>
+__host__ __device__ __thrust_forceinline__
+bool is_power_of_2(Integer x)
+{
+  return 0 == (x & (x - 1));
+}
+
+template <typename Integer>
+__host__ __device__ __thrust_forceinline__
+bool is_odd(Integer x)
+{
+  return 1 & x;
+}
+
+template <typename Integer>
+__host__ __device__ __thrust_forceinline__
+Integer log2(Integer x)
+{
+  Integer num_bits = 8 * sizeof(Integer);
+  Integer num_bits_minus_one = num_bits - 1;
+
+  return num_bits_minus_one - clz(x);
+}
+
+
+template <typename Integer>
+__host__ __device__ __thrust_forceinline__
+Integer log2_ri(Integer x)
+{
+  Integer result = log2(x);
+
+  // This is where we round up to the nearest log.
+  if (!is_power_of_2(x))
+    ++result;
+
+  return result;
+}
+
+// x/y rounding towards +infinity for integers
+// Used to determine # of blocks/warps etc.
+template <typename Integer0, typename Integer1>
+__host__ __device__ __thrust_forceinline__
+#if THRUST_CPP_DIALECT >= 2011
+// FIXME: Should use common_type.
+auto divide_ri(Integer0 const x, Integer1 const y)
+THRUST_DECLTYPE_RETURNS((x + (y - 1)) / y)
+#else
+// FIXME: Should use common_type.
+Integer0 divide_ri(Integer0 const x, Integer1 const y)
+{
+  return (x + (y - 1)) / y;
+}
+#endif
+
+// x/y rounding towards zero for integers.
+// Used to determine # of blocks/warps etc.
+template <typename Integer0, typename Integer1>
+__host__ __device__ __thrust_forceinline__
+#if THRUST_CPP_DIALECT >= 2011
+auto divide_rz(Integer0 const x, Integer1 const y)
+THRUST_DECLTYPE_RETURNS(x / y)
+#else
+// FIXME: Should use common_type.
+Integer0 divide_rz(Integer0 const x, Integer1 const y)
+{
+  return x / y;
+}
+#endif
+
+// Round x towards infinity to the next multiple of y.
+template <typename Integer0, typename Integer1>
+__host__ __device__ __thrust_forceinline__
+#if THRUST_CPP_DIALECT >= 2011
+auto round_i(Integer0 const x, Integer1 const y)
+THRUST_DECLTYPE_RETURNS(y * divide_ri(x, y))
+#else
+Integer0 round_i(Integer0 const x, Integer1 const y)
+{
+  return y * divide_ri(x, y);
+}
+#endif
+
+// Round x towards 0 to the next multiple of y.
+template <typename Integer0, typename Integer1>
+__host__ __device__ __thrust_forceinline__
+#if THRUST_CPP_DIALECT >= 2011
+auto round_z(Integer0 const x, Integer1 const y)
+THRUST_DECLTYPE_RETURNS(y * divide_rz(x, y))
+#else
+Integer0 round_z(Integer0 const x, Integer1 const y)
+{
+  return y * divide_rz(x, y);
+}
+#endif
+
+} // end detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_traits.h
new file mode 100644
index 000000000000..853af20b860d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/integer_traits.h
@@ -0,0 +1,130 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <limits>
+#include <limits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<typename T>
+  class integer_traits
+{
+  public:
+    static constexpr bool is_integral = false;
+};
+
+template<typename T, T min_val, T max_val>
+  class integer_traits_base
+{
+  public:
+    static constexpr bool is_integral = true;
+    static constexpr T const_min = min_val;
+    static constexpr T const_max = max_val;
+};
+
+
+template<>
+  class integer_traits<bool>
+    : public std::numeric_limits<bool>,
+      public integer_traits_base<bool, false, true>
+{};
+
+
+template<>
+  class integer_traits<char>
+    : public std::numeric_limits<char>,
+      public integer_traits_base<char, CHAR_MIN, CHAR_MAX>
+{};
+
+
+template<>
+  class integer_traits<signed char>
+    : public std::numeric_limits<signed char>,
+      public integer_traits_base<signed char, SCHAR_MIN, SCHAR_MAX>
+{};
+
+
+template<>
+  class integer_traits<unsigned char>
+    : public std::numeric_limits<unsigned char>,
+      public integer_traits_base<unsigned char, 0, UCHAR_MAX>
+{};
+
+
+template<>
+  class integer_traits<short>
+    : public std::numeric_limits<short>,
+      public integer_traits_base<short, SHRT_MIN, SHRT_MAX>
+{};
+
+
+template<>
+  class integer_traits<unsigned short>
+    : public std::numeric_limits<unsigned short>,
+      public integer_traits_base<unsigned short, 0, USHRT_MAX>
+{};
+
+
+template<>
+  class integer_traits<int>
+    : public std::numeric_limits<int>,
+      public integer_traits_base<int, INT_MIN, INT_MAX>
+{};
+
+
+template<>
+  class integer_traits<unsigned int>
+    : public std::numeric_limits<unsigned int>,
+      public integer_traits_base<unsigned int, 0, UINT_MAX>
+{};
+
+
+template<>
+  class integer_traits<long>
+    : public std::numeric_limits<long>,
+      public integer_traits_base<long, LONG_MIN, LONG_MAX>
+{};
+
+
+template<>
+  class integer_traits<unsigned long>
+    : public std::numeric_limits<unsigned long>,
+      public integer_traits_base<unsigned long, 0, ULONG_MAX>
+{};
+
+
+template<>
+  class integer_traits<long long>
+    : public std::numeric_limits<long long>,
+      public integer_traits_base<long long, LLONG_MIN, LLONG_MAX>
+{};
+
+
+template<>
+  class integer_traits<unsigned long long>
+    : public std::numeric_limits<unsigned long long>,
+      public integer_traits_base<unsigned long long, 0, ULLONG_MAX>
+{};
+
+} // end detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/internal_functional.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/internal_functional.h
new file mode 100644
index 000000000000..a0c4056febae
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/internal_functional.h
@@ -0,0 +1,558 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file internal_functional.inl
+ *  \brief Non-public functionals used to implement algorithm internals.
+ */
+
+#pragma once
+
+#include <thrust/tuple.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/config.h>
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/tuple_of_iterator_references.h>
+#include <thrust/detail/raw_reference_cast.h>
+#include <thrust/detail/memory_wrapper.h> // for ::new
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// unary_negate does not need to know argument_type
+template<typename Predicate>
+struct unary_negate
+{
+  typedef bool result_type;
+
+  Predicate pred;
+
+  __host__ __device__
+  explicit unary_negate(const Predicate& pred) : pred(pred) {}
+
+  template <typename T>
+  __host__ __device__
+  bool operator()(const T& x)
+  {
+    return !bool(pred(x));
+  }
+};
+
+// binary_negate does not need to know first_argument_type or second_argument_type
+template<typename Predicate>
+struct binary_negate
+{
+  typedef bool result_type;
+
+  Predicate pred;
+
+  __host__ __device__
+  explicit binary_negate(const Predicate& pred) : pred(pred) {}
+
+  template <typename T1, typename T2>
+  __host__ __device__
+  bool operator()(const T1& x, const T2& y)
+  {
+    return !bool(pred(x,y));
+  }
+};
+
+template<typename Predicate>
+__host__ __device__
+thrust::detail::unary_negate<Predicate> not1(const Predicate &pred)
+{
+  return thrust::detail::unary_negate<Predicate>(pred);
+}
+
+template<typename Predicate>
+__host__ __device__
+thrust::detail::binary_negate<Predicate> not2(const Predicate &pred)
+{
+  return thrust::detail::binary_negate<Predicate>(pred);
+}
+
+
+// convert a predicate to a 0 or 1 integral value
+template<typename Predicate, typename IntegralType>
+struct predicate_to_integral
+{
+  Predicate pred;
+
+  __host__ __device__
+  explicit predicate_to_integral(const Predicate& pred) : pred(pred) {}
+
+  template <typename T>
+  __host__ __device__
+  IntegralType operator()(const T& x)
+  {
+    return pred(x) ? IntegralType(1) : IntegralType(0);
+  }
+};
+
+
+// note that detail::equal_to does not force conversion from T2 -> T1 as equal_to does
+template<typename T1>
+struct equal_to
+{
+  typedef bool result_type;
+
+  template <typename T2>
+  __host__ __device__
+  bool operator()(const T1& lhs, const T2& rhs) const
+  {
+    return lhs == rhs;
+  }
+};
+
+// note that equal_to_value does not force conversion from T2 -> T1 as equal_to does
+template<typename T2>
+struct equal_to_value
+{
+  T2 rhs;
+
+  __host__ __device__
+  equal_to_value(const T2& rhs) : rhs(rhs) {}
+
+  template <typename T1>
+  __host__ __device__
+  bool operator()(const T1& lhs) const
+  {
+    return lhs == rhs;
+  }
+};
+
+template<typename Predicate>
+struct tuple_binary_predicate
+{
+  typedef bool result_type;
+
+  __host__ __device__
+  tuple_binary_predicate(const Predicate& p) : pred(p) {}
+
+  template<typename Tuple>
+  __host__ __device__
+  bool operator()(const Tuple& t) const
+  {
+    return pred(thrust::get<0>(t), thrust::get<1>(t));
+  }
+
+  mutable Predicate pred;
+};
+
+template<typename Predicate>
+struct tuple_not_binary_predicate
+{
+  typedef bool result_type;
+
+  __host__ __device__
+  tuple_not_binary_predicate(const Predicate& p) : pred(p) {}
+
+  template<typename Tuple>
+  __host__ __device__
+  bool operator()(const Tuple& t) const
+  {
+    return !pred(thrust::get<0>(t), thrust::get<1>(t));
+  }
+
+  mutable Predicate pred;
+};
+
+template<typename Generator>
+  struct host_generate_functor
+{
+  typedef void result_type;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  host_generate_functor(Generator g)
+    : gen(g) {}
+
+  // operator() does not take an lvalue reference because some iterators
+  // produce temporary proxy references when dereferenced. for example,
+  // consider the temporary tuple of references produced by zip_iterator.
+  // such temporaries cannot bind to an lvalue reference.
+  //
+  // to WAR this, accept a const reference (which is bindable to a temporary),
+  // and const_cast in the implementation.
+  //
+  // XXX change to an rvalue reference upon c++0x (which either a named variable
+  //     or temporary can bind to)
+  template<typename T>
+  __host__
+  void operator()(const T &x)
+  {
+    // we have to be naughty and const_cast this to get it to work
+    T &lvalue = const_cast<T&>(x);
+
+    // this assigns correctly whether x is a true reference or proxy
+    lvalue = gen();
+  }
+
+  Generator gen;
+};
+
+template<typename Generator>
+  struct device_generate_functor
+{
+  typedef void result_type;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  device_generate_functor(Generator g)
+    : gen(g) {}
+
+  // operator() does not take an lvalue reference because some iterators
+  // produce temporary proxy references when dereferenced. for example,
+  // consider the temporary tuple of references produced by zip_iterator.
+  // such temporaries cannot bind to an lvalue reference.
+  //
+  // to WAR this, accept a const reference (which is bindable to a temporary),
+  // and const_cast in the implementation.
+  //
+  // XXX change to an rvalue reference upon c++0x (which either a named variable
+  //     or temporary can bind to)
+  template<typename T>
+  __host__ __device__
+  void operator()(const T &x)
+  {
+    // we have to be naughty and const_cast this to get it to work
+    T &lvalue = const_cast<T&>(x);
+
+    // this assigns correctly whether x is a true reference or proxy
+    lvalue = gen();
+  }
+
+  Generator gen;
+};
+
+template<typename System, typename Generator>
+  struct generate_functor
+    : thrust::detail::eval_if<
+        thrust::detail::is_convertible<System, thrust::host_system_tag>::value,
+        thrust::detail::identity_<host_generate_functor<Generator> >,
+        thrust::detail::identity_<device_generate_functor<Generator> >
+      >
+{};
+
+
+template<typename ResultType, typename BinaryFunction>
+  struct zipped_binary_op
+{
+  typedef ResultType result_type;
+
+  __host__ __device__
+  zipped_binary_op(BinaryFunction binary_op)
+    : m_binary_op(binary_op) {}
+
+  template<typename Tuple>
+  __host__ __device__
+  inline result_type operator()(Tuple t)
+  {
+    return m_binary_op(thrust::get<0>(t), thrust::get<1>(t));
+  }
+
+  BinaryFunction m_binary_op;
+};
+
+
+template<typename T>
+  struct is_non_const_reference
+    : thrust::detail::and_<
+        thrust::detail::not_<thrust::detail::is_const<T> >,
+        thrust::detail::or_<thrust::detail::is_reference<T>,
+                            thrust::detail::is_proxy_reference<T> >
+      >
+{};
+
+template<typename T> struct is_tuple_of_iterator_references : thrust::detail::false_type {};
+
+template<typename... Ts>
+  struct is_tuple_of_iterator_references<
+    thrust::detail::tuple_of_iterator_references<
+      Ts...
+    >
+  >
+    : thrust::detail::true_type
+{};
+
+// use this enable_if to avoid assigning to temporaries in the transform functors below
+// XXX revisit this problem with c++11 perfect forwarding
+template<typename T>
+  struct enable_if_non_const_reference_or_tuple_of_iterator_references
+    : thrust::detail::enable_if<
+        is_non_const_reference<T>::value || is_tuple_of_iterator_references<T>::value
+      >
+{};
+
+
+template<typename UnaryFunction>
+  struct unary_transform_functor
+{
+  typedef void result_type;
+
+  UnaryFunction f;
+
+  __host__ __device__
+  unary_transform_functor(UnaryFunction f)
+    : f(f)
+  {}
+
+  __thrust_exec_check_disable__
+  template<typename Tuple>
+  inline __host__ __device__
+  typename enable_if_non_const_reference_or_tuple_of_iterator_references<
+    typename thrust::tuple_element<1,Tuple>::type
+  >::type
+    operator()(Tuple t)
+  {
+    thrust::get<1>(t) = f(thrust::get<0>(t));
+  }
+};
+
+
+template<typename BinaryFunction>
+  struct binary_transform_functor
+{
+  BinaryFunction f;
+
+  __host__ __device__
+  binary_transform_functor(BinaryFunction f)
+    : f(f)
+  {}
+
+  __thrust_exec_check_disable__
+  template<typename Tuple>
+  inline __host__ __device__
+  typename enable_if_non_const_reference_or_tuple_of_iterator_references<
+    typename thrust::tuple_element<2,Tuple>::type
+  >::type
+    operator()(Tuple t)
+  {
+    thrust::get<2>(t) = f(thrust::get<0>(t), thrust::get<1>(t));
+  }
+};
+
+
+template<typename UnaryFunction, typename Predicate>
+struct unary_transform_if_functor
+{
+  UnaryFunction unary_op;
+  Predicate pred;
+
+  __host__ __device__
+  unary_transform_if_functor(UnaryFunction unary_op, Predicate pred)
+    : unary_op(unary_op), pred(pred)
+  {}
+
+  __thrust_exec_check_disable__
+  template<typename Tuple>
+  inline __host__ __device__
+  typename enable_if_non_const_reference_or_tuple_of_iterator_references<
+    typename thrust::tuple_element<1,Tuple>::type
+  >::type
+    operator()(Tuple t)
+  {
+    if(pred(thrust::get<0>(t)))
+    {
+      thrust::get<1>(t) = unary_op(thrust::get<0>(t));
+    }
+  }
+}; // end unary_transform_if_functor
+
+
+template<typename UnaryFunction, typename Predicate>
+struct unary_transform_if_with_stencil_functor
+{
+  UnaryFunction unary_op;
+  Predicate pred;
+
+  __host__ __device__
+  unary_transform_if_with_stencil_functor(UnaryFunction unary_op, Predicate pred)
+    : unary_op(unary_op), pred(pred)
+  {}
+
+  __thrust_exec_check_disable__
+  template<typename Tuple>
+  inline __host__ __device__
+  typename enable_if_non_const_reference_or_tuple_of_iterator_references<
+    typename thrust::tuple_element<2,Tuple>::type
+  >::type
+    operator()(Tuple t)
+  {
+    if(pred(thrust::get<1>(t)))
+      thrust::get<2>(t) = unary_op(thrust::get<0>(t));
+  }
+}; // end unary_transform_if_with_stencil_functor
+
+
+template<typename BinaryFunction, typename Predicate>
+struct binary_transform_if_functor
+{
+  BinaryFunction binary_op;
+  Predicate pred;
+
+  __host__ __device__
+  binary_transform_if_functor(BinaryFunction binary_op, Predicate pred)
+    : binary_op(binary_op), pred(pred) {}
+
+  __thrust_exec_check_disable__
+  template<typename Tuple>
+  inline __host__ __device__
+  typename enable_if_non_const_reference_or_tuple_of_iterator_references<
+    typename thrust::tuple_element<3,Tuple>::type
+  >::type
+    operator()(Tuple t)
+  {
+    if(pred(thrust::get<2>(t)))
+      thrust::get<3>(t) = binary_op(thrust::get<0>(t), thrust::get<1>(t));
+  }
+}; // end binary_transform_if_functor
+
+
+template<typename T>
+  struct host_destroy_functor
+{
+  __host__
+  void operator()(T &x) const
+  {
+    x.~T();
+  } // end operator()()
+}; // end host_destroy_functor
+
+
+template<typename T>
+  struct device_destroy_functor
+{
+  // add __host__ to allow the omp backend to compile with nvcc
+  __host__ __device__
+  void operator()(T &x) const
+  {
+    x.~T();
+  } // end operator()()
+}; // end device_destroy_functor
+
+
+template<typename System, typename T>
+  struct destroy_functor
+    : thrust::detail::eval_if<
+        thrust::detail::is_convertible<System, thrust::host_system_tag>::value,
+        thrust::detail::identity_<host_destroy_functor<T> >,
+        thrust::detail::identity_<device_destroy_functor<T> >
+      >
+{};
+
+
+template <typename T>
+struct fill_functor
+{
+  T exemplar;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  fill_functor(const T& _exemplar)
+    : exemplar(_exemplar) {}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  fill_functor(const fill_functor & other)
+    :exemplar(other.exemplar){}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  ~fill_functor() {}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  T operator()(void) const
+  {
+    return exemplar;
+  }
+};
+
+
+template<typename T>
+  struct uninitialized_fill_functor
+{
+  T exemplar;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  uninitialized_fill_functor(const T & x):exemplar(x){}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  uninitialized_fill_functor(const uninitialized_fill_functor & other)
+    :exemplar(other.exemplar){}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  ~uninitialized_fill_functor() {}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  void operator()(T &x)
+  {
+    ::new(static_cast<void*>(&x)) T(exemplar);
+  } // end operator()()
+}; // end uninitialized_fill_functor
+
+
+// this predicate tests two two-element tuples
+// we first use a Compare for the first element
+// if the first elements are equivalent, we use
+// < for the second elements
+template<typename Compare>
+  struct compare_first_less_second
+{
+  compare_first_less_second(Compare c)
+    : comp(c) {}
+
+  template<typename T1, typename T2>
+  __host__ __device__
+  bool operator()(T1 lhs, T2 rhs)
+  {
+    return comp(thrust::get<0>(lhs), thrust::get<0>(rhs)) || (!comp(thrust::get<0>(rhs), thrust::get<0>(lhs)) && thrust::get<1>(lhs) < thrust::get<1>(rhs));
+  }
+
+  Compare comp;
+}; // end compare_first_less_second
+
+
+template<typename Compare>
+  struct compare_first
+{
+  Compare comp;
+
+  __host__ __device__
+  compare_first(Compare comp)
+    : comp(comp)
+  {}
+
+  template<typename Tuple1, typename Tuple2>
+  __host__ __device__
+  bool operator()(const Tuple1 &x, const Tuple2 &y)
+  {
+    return comp(thrust::raw_reference_cast(thrust::get<0>(x)), thrust::raw_reference_cast(thrust::get<0>(y)));
+  }
+}; // end compare_first
+
+
+} // end namespace detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/logical.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/logical.inl
new file mode 100644
index 000000000000..3d39cac92689
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/logical.inl
@@ -0,0 +1,95 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/logical.h>
+#include <thrust/system/detail/adl/logical.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool all_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::all_of;
+  return all_of(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end all_of()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool any_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::any_of;
+  return any_of(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end any_of()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool none_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::none_of;
+  return none_of(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end none_of()
+
+
+template<typename InputIterator, typename Predicate>
+bool all_of(InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::all_of(select_system(system), first, last, pred);
+}
+
+
+template<typename InputIterator, typename Predicate>
+bool any_of(InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::any_of(select_system(system), first, last, pred);
+}
+
+
+template<typename InputIterator, typename Predicate>
+bool none_of(InputIterator first, InputIterator last, Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::none_of(select_system(system), first, last, pred);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/malloc_and_free.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/malloc_and_free.h
new file mode 100644
index 000000000000..143518893198
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/malloc_and_free.h
@@ -0,0 +1,83 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/pointer.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/system/detail/generic/memory.h>
+#include <thrust/system/detail/adl/malloc_and_free.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy>
+__host__ __device__
+pointer<void,DerivedPolicy> malloc(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, std::size_t n)
+{
+  using thrust::system::detail::generic::malloc;
+
+  // XXX should use a hypothetical thrust::static_pointer_cast here
+  void *raw_ptr = static_cast<void*>(thrust::raw_pointer_cast(malloc(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), n)));
+
+  return pointer<void,DerivedPolicy>(raw_ptr);
+}
+
+__thrust_exec_check_disable__
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+pointer<T,DerivedPolicy> malloc(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, std::size_t n)
+{
+  using thrust::system::detail::generic::malloc;
+
+  T *raw_ptr = static_cast<T*>(thrust::raw_pointer_cast(malloc<T>(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), n)));
+
+  return pointer<T,DerivedPolicy>(raw_ptr);
+}
+
+
+// XXX WAR nvbug 992955
+#if THRUST_DEVICE_COMPILER == THRUST_DEVICE_COMPILER_NVCC
+#if CUDART_VERSION < 5000
+
+// cudafe generates unqualified calls to free(int *volatile)
+// which get confused with thrust::free
+// spoof a thrust::free which simply maps to ::free
+inline __host__ __device__
+void free(int *volatile ptr)
+{
+  ::free(ptr);
+}
+
+#endif // CUDART_VERSION
+#endif // THRUST_DEVICE_COMPILER
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+void free(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, Pointer ptr)
+{
+  using thrust::system::detail::generic::free;
+
+  free(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), ptr);
+}
+
+// XXX consider another form of free which does not take a system argument and
+// instead infers the system from the pointer
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_algorithms.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_algorithms.h
new file mode 100644
index 000000000000..2f6b3a81d71c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_algorithms.h
@@ -0,0 +1,237 @@
+// Copyright (c) 2018 NVIDIA Corporation
+// Author: Bryce Adelstein Lelbach <brycelelbach@gmail.com>
+//
+// Distributed under the Boost Software License v1.0 (boost.org/LICENSE_1_0.txt)
+
+// TODO: These need to be turned into proper Thrust algorithms (dispatch layer,
+// backends, etc).
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/allocator/allocator_traits.h>
+#include <thrust/detail/memory_wrapper.h>
+#include <thrust/addressof.h>
+
+#include <nv/target>
+
+#include <utility>
+#include <new>
+
+
+THRUST_NAMESPACE_BEGIN
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T>
+__host__ __device__
+void destroy_at(T* location)
+{
+  location->~T();
+}
+
+template <typename Allocator, typename T>
+__host__ __device__
+void destroy_at(Allocator const& alloc, T* location)
+{
+  typedef typename detail::allocator_traits<
+    typename detail::remove_cv<
+      typename detail::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>::other traits;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  traits::destroy(alloc_T, location);
+}
+
+template <typename ForwardIt>
+__host__ __device__
+ForwardIt destroy(ForwardIt first, ForwardIt last)
+{
+  for (; first != last; ++first)
+    destroy_at(addressof(*first));
+
+  return first;
+}
+
+template <typename Allocator, typename ForwardIt>
+__host__ __device__
+ForwardIt destroy(Allocator const& alloc, ForwardIt first, ForwardIt last)
+{
+  typedef typename iterator_traits<ForwardIt>::value_type T;
+  typedef typename detail::allocator_traits<
+    typename detail::remove_cv<
+      typename detail::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>::other traits;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  for (; first != last; ++first)
+    destroy_at(alloc_T, addressof(*first));
+
+  return first;
+}
+
+template <typename ForwardIt, typename Size>
+__host__ __device__
+ForwardIt destroy_n(ForwardIt first, Size n)
+{
+  for (; n > 0; (void) ++first, --n)
+    destroy_at(addressof(*first));
+
+  return first;
+}
+
+template <typename Allocator, typename ForwardIt, typename Size>
+__host__ __device__
+ForwardIt destroy_n(Allocator const& alloc, ForwardIt first, Size n)
+{
+  typedef typename iterator_traits<ForwardIt>::value_type T;
+  typedef typename detail::allocator_traits<
+    typename detail::remove_cv<
+      typename detail::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>::other traits;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  for (; n > 0; (void) ++first, --n)
+    destroy_at(alloc_T, addressof(*first));
+
+  return first;
+}
+
+template <typename ForwardIt, typename... Args>
+__host__ __device__
+void uninitialized_construct(
+  ForwardIt first, ForwardIt last, Args const&... args
+)
+{
+  using T = typename iterator_traits<ForwardIt>::value_type;
+
+  ForwardIt current = first;
+
+  // No exceptions in CUDA.
+  NV_IF_TARGET(NV_IS_HOST, (
+    try {
+      for (; current != last; ++current)
+      {
+        ::new (static_cast<void*>(addressof(*current))) T(args...);
+      }
+    } catch (...) {
+      destroy(first, current);
+      throw;
+    }
+  ), (
+    for (; current != last; ++current)
+    {
+      ::new (static_cast<void*>(addressof(*current))) T(args...);
+    }
+  ));
+}
+
+template <typename Allocator, typename ForwardIt, typename... Args>
+void uninitialized_construct_with_allocator(
+  Allocator const& alloc, ForwardIt first, ForwardIt last, Args const&... args
+)
+{
+  using T = typename iterator_traits<ForwardIt>::value_type;
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  ForwardIt current = first;
+
+  // No exceptions in CUDA.
+  NV_IF_TARGET(NV_IS_HOST, (
+    try {
+      for (; current != last; ++current)
+      {
+        traits::construct(alloc_T, addressof(*current), args...);
+      }
+    } catch (...) {
+      destroy(alloc_T, first, current);
+      throw;
+    }
+  ), (
+    for (; current != last; ++current)
+    {
+      traits::construct(alloc_T, addressof(*current), args...);
+    }
+  ));
+}
+
+template <typename ForwardIt, typename Size, typename... Args>
+void uninitialized_construct_n(
+  ForwardIt first, Size n, Args const&... args
+)
+{
+  using T = typename iterator_traits<ForwardIt>::value_type;
+
+  ForwardIt current = first;
+
+  // No exceptions in CUDA.
+  NV_IF_TARGET(NV_IS_HOST, (
+    try {
+      for (; n > 0; ++current, --n)
+      {
+        ::new (static_cast<void*>(addressof(*current))) T(args...);
+      }
+    } catch (...) {
+      destroy(first, current);
+      throw;
+    }
+  ), (
+    for (; n > 0; ++current, --n)
+    {
+      ::new (static_cast<void*>(addressof(*current))) T(args...);
+    }
+  ));
+}
+
+template <typename Allocator, typename ForwardIt, typename Size, typename... Args>
+void uninitialized_construct_n_with_allocator(
+  Allocator const& alloc, ForwardIt first, Size n, Args const&... args
+)
+{
+  using T = typename iterator_traits<ForwardIt>::value_type;
+  using traits = typename detail::allocator_traits<
+    typename std::remove_cv<
+      typename std::remove_reference<Allocator>::type
+    >::type
+  >::template rebind_traits<T>;
+
+  typename traits::allocator_type alloc_T(alloc);
+
+  ForwardIt current = first;
+
+  // No exceptions in CUDA.
+  NV_IF_TARGET(NV_IS_HOST, (
+    try {
+      for (; n > 0; (void) ++current, --n)
+      {
+        traits::construct(alloc_T, addressof(*current), args...);
+      }
+    } catch (...) {
+      destroy(alloc_T, first, current);
+      throw;
+    }
+  ), (
+    for (; n > 0; (void) ++current, --n)
+    {
+      traits::construct(alloc_T, addressof(*current), args...);
+    }
+  ));
+}
+
+///////////////////////////////////////////////////////////////////////////////
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_wrapper.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_wrapper.h
new file mode 100644
index 000000000000..bfc9056fa15f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/memory_wrapper.h
@@ -0,0 +1,30 @@
+/*
+ *  Copyright 2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+// When a compiler uses Thrust as part of its implementation of Standard C++
+// algorithms, a cycle of included files may result when Thrust code tries to
+// use a standard algorithm.  Having a macro that is defined only when Thrust
+// is including an algorithms-related header gives the compiler a chance to
+// detect and break the cycle of includes.  (<memory> declares several standard
+// algorithms, including all of the uninitialized_* algorithms.  "_ALGORITHMS_"
+// in the macro name is meant generically, not as a specific reference to
+// the header <algorithms>.)
+
+#define THRUST_INCLUDING_ALGORITHMS_HEADER
+#include <memory>
+#undef  THRUST_INCLUDING_ALGORITHMS_HEADER
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/merge.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/merge.inl
new file mode 100644
index 000000000000..1595cc1a145e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/merge.inl
@@ -0,0 +1,221 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/merge.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/merge.h>
+#include <thrust/system/detail/adl/merge.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator merge(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result)
+{
+  using thrust::system::detail::generic::merge;
+  return merge(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result);
+} // end merge()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+  OutputIterator merge(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result,
+                       StrictWeakCompare comp)
+{
+  using thrust::system::detail::generic::merge;
+  return merge(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result, comp);
+} // end merge()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result)
+{
+  using thrust::system::detail::generic::merge_by_key;
+  return merge_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end merge_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2, typename Compare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result,
+                 Compare comp)
+{
+  using thrust::system::detail::generic::merge_by_key;
+  return merge_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end merge_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+  OutputIterator merge(InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result,
+                       StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::merge(select_system(system1,system2,system3), first1, last1, first2, last2, result, comp);
+} // end merge()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator merge(InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::merge(select_system(system1,system2,system3), first1, last1, first2, last2, result);
+} // end merge()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(InputIterator1 keys_first1,
+                 InputIterator1 keys_last1,
+                 InputIterator2 keys_first2,
+                 InputIterator2 keys_last2,
+                 InputIterator3 values_first1,
+                 InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result,
+                 StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::merge_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end merge_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(InputIterator1 keys_first1,
+                 InputIterator1 keys_last1,
+                 InputIterator2 keys_first2,
+                 InputIterator2 keys_last2,
+                 InputIterator3 values_first1,
+                 InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::merge_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end merge_by_key()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/minmax.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/minmax.h
new file mode 100644
index 000000000000..c565a74bdffb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/minmax.h
@@ -0,0 +1,51 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename T, typename BinaryPredicate>
+__host__ __device__
+  T min THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs, BinaryPredicate comp)
+{
+  return comp(rhs, lhs) ? rhs : lhs;
+} // end min()
+
+template<typename T>
+__host__ __device__
+  T min THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs)
+{
+  return rhs < lhs ? rhs : lhs;
+} // end min()
+
+template<typename T, typename BinaryPredicate>
+__host__ __device__
+  T max THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs, BinaryPredicate comp)
+{
+  return comp(lhs,rhs) ? rhs : lhs;
+} // end max()
+
+template<typename T>
+__host__ __device__
+  T max THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs)
+{
+  return lhs < rhs ? rhs : lhs;
+} // end max()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/mismatch.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/mismatch.inl
new file mode 100644
index 000000000000..16c579d80697
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/mismatch.inl
@@ -0,0 +1,89 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/mismatch.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/mismatch.h>
+#include <thrust/system/detail/adl/mismatch.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                                      InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2)
+{
+  using thrust::system::detail::generic::mismatch;
+  return mismatch(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2);
+} // end mismatch()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                                      InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2,
+                                                      BinaryPredicate pred)
+{
+  using thrust::system::detail::generic::mismatch;
+  return mismatch(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, pred);
+} // end mismatch()
+
+
+template<typename InputIterator1, typename InputIterator2>
+thrust::pair<InputIterator1, InputIterator2> mismatch(InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::mismatch(select_system(system1,system2), first1, last1, first2);
+} // end mismatch()
+
+
+template<typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+thrust::pair<InputIterator1, InputIterator2> mismatch(InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2,
+                                                      BinaryPredicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::mismatch(select_system(system1,system2), first1, last1, first2, pred);
+} // end mismatch()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/modern_gcc_required.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/modern_gcc_required.h
new file mode 100644
index 000000000000..a8c3d98ba996
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/modern_gcc_required.h
@@ -0,0 +1,26 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config/cpp_dialect.h>
+
+#ifndef THRUST_MODERN_GCC_REQUIRED_NO_ERROR
+#  if defined(THRUST_GCC_VERSION) && !defined(THRUST_MODERN_GCC)
+#    error GCC 5 or later is required for this Thrust feature; please upgrade your compiler.
+#  endif
+#endif
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/mpl/math.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/mpl/math.h
new file mode 100644
index 000000000000..bda98003cd58
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/mpl/math.h
@@ -0,0 +1,175 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file math.h
+ *  \brief Math-related metaprogramming functionality.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+namespace mpl
+{
+
+namespace math
+{
+
+namespace detail
+{
+
+// compute the log base-2 of an integer at compile time
+template <unsigned int N, unsigned int Cur>
+struct log2
+{
+    static const unsigned int value = log2<N / 2,Cur+1>::value;
+};
+
+template <unsigned int Cur>
+struct log2<1, Cur>
+{
+    static const unsigned int value = Cur;
+};
+
+template <unsigned int Cur>
+struct log2<0, Cur>
+{
+    // undefined
+};
+
+} // end namespace detail
+
+
+template <unsigned int N>
+struct log2
+{
+    static const unsigned int value = detail::log2<N,0>::value;
+};
+
+
+template <typename T, T lhs, T rhs>
+struct min
+{
+  static const T value = (lhs < rhs) ? lhs : rhs;
+};
+
+
+template <typename T, T lhs, T rhs>
+struct max
+{
+  static const T value = (!(lhs < rhs)) ? lhs : rhs;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct mul
+{
+  static const result_type value = x * y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct mod
+{
+  static const result_type value = x % y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct div
+{
+  static const result_type value = x / y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct geq
+{
+  static const bool value = x >= y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct lt
+{
+  static const bool value = x < y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct gt
+{
+  static const bool value = x > y;
+};
+
+
+template<bool x, bool y>
+  struct or_
+{
+  static const bool value = (x || y);
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct bit_and
+{
+  static const result_type value = x & y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct plus
+{
+  static const result_type value = x + y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct minus
+{
+  static const result_type value = x - y;
+};
+
+
+template<typename result_type, result_type x, result_type y>
+  struct equal
+{
+  static const bool value = x == y;
+};
+
+
+template<typename result_type, result_type x>
+  struct is_odd
+{
+  static const bool value = x & 1;
+};
+
+
+} // end namespace math
+
+} // end namespace mpl
+
+} // end namespace detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_traits.h
new file mode 100644
index 000000000000..e728adcaf3e8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_traits.h
@@ -0,0 +1,129 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <limits>
+
+//#include <stdint.h> // for intmax_t (not provided on MSVS 2005)
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// XXX good enough for the platforms we care about
+typedef long long intmax_t;
+
+template<typename Number>
+  struct is_signed
+    : integral_constant<bool, std::numeric_limits<Number>::is_signed>
+{}; // end is_signed
+
+
+template<typename T>
+  struct num_digits
+    : eval_if<
+        std::numeric_limits<T>::is_specialized,
+        integral_constant<
+          int,
+          std::numeric_limits<T>::digits
+        >,
+        integral_constant<
+          int,
+          sizeof(T) * std::numeric_limits<unsigned char>::digits - (is_signed<T>::value ? 1 : 0)  
+        >
+      >::type
+{}; // end num_digits
+
+
+template<typename Integer>
+  struct integer_difference
+    //: eval_if<
+    //    sizeof(Integer) >= sizeof(intmax_t),
+    //    eval_if<
+    //      is_signed<Integer>::value,
+    //      identity_<Integer>,
+    //      identity_<intmax_t>
+    //    >,
+    //    eval_if<
+    //      sizeof(Integer) < sizeof(std::ptrdiff_t),
+    //      identity_<std::ptrdiff_t>,
+    //      identity_<intmax_t>
+    //    >
+    //  >
+{
+  private:
+    // XXX workaround a pedantic warning in old versions of g++
+    //     which complains about &&ing with a constant value
+    template<bool x, bool y>
+      struct and_
+    {
+      static const bool value = false;
+    };
+
+    template<bool y>
+      struct and_<true,y>
+    {
+      static const bool value = y;
+    };
+
+  public:
+    typedef typename
+      eval_if<
+        and_<
+          std::numeric_limits<Integer>::is_signed,
+          // digits is the number of no-sign bits
+          (!std::numeric_limits<Integer>::is_bounded || (int(std::numeric_limits<Integer>::digits) + 1 >= num_digits<intmax_t>::value))
+        >::value,
+        identity_<Integer>,
+        eval_if<
+          int(std::numeric_limits<Integer>::digits) + 1 < num_digits<signed int>::value,
+          identity_<signed int>,
+          eval_if<
+            int(std::numeric_limits<Integer>::digits) + 1 < num_digits<signed long>::value,
+            identity_<signed long>,
+            identity_<intmax_t>
+          >
+        >
+      >::type type;
+}; // end integer_difference
+
+
+template<typename Number>
+  struct numeric_difference
+    : eval_if<
+      is_integral<Number>::value,
+      integer_difference<Number>,
+      identity_<Number>
+    >
+{}; // end numeric_difference
+
+
+template<typename Number>
+__host__ __device__
+typename numeric_difference<Number>::type
+numeric_distance(Number x, Number y)
+{
+  typedef typename numeric_difference<Number>::type difference_type;
+  return difference_type(y) - difference_type(x);
+} // end numeric_distance
+
+} // end detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_wrapper.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_wrapper.h
new file mode 100644
index 000000000000..9ebc6e23b711
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/numeric_wrapper.h
@@ -0,0 +1,27 @@
+/*
+ *  Copyright 2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+// When a compiler uses Thrust as part of its implementation of Standard C++
+// algorithms, a cycle of included files may result when Thrust code tries to
+// use a standard algorithm.  Having a macro that is defined only when Thrust
+// is including an algorithms-related header gives the compiler a chance to
+// detect and break the cycle of includes.
+
+#define THRUST_INCLUDING_ALGORITHMS_HEADER
+#include <numeric>
+#undef  THRUST_INCLUDING_ALGORITHMS_HEADER
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/overlapped_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/overlapped_copy.h
new file mode 100644
index 000000000000..418497de8db9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/overlapped_copy.h
@@ -0,0 +1,131 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/detail/copy.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/system/cpp/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator sequential_copy(InputIterator first,
+                                 InputIterator last,
+                                 OutputIterator result)
+{
+  for(; first != last; ++first, ++result)
+  {
+    *result = *first;
+  } // end for
+
+  return result;
+} // end sequential_copy()
+
+
+template<typename BidirectionalIterator1,
+         typename BidirectionalIterator2>
+  BidirectionalIterator2 sequential_copy_backward(BidirectionalIterator1 first,
+                                                  BidirectionalIterator1 last,
+                                                  BidirectionalIterator2 result)
+{
+  // yes, we preincrement
+  // the ranges are open on the right, i.e. [first, last)
+  while(first != last)
+  {
+    *--result = *--last;
+  } // end while
+
+  return result;
+} // end sequential_copy_backward()
+
+
+namespace dispatch
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  RandomAccessIterator2 overlapped_copy(thrust::system::cpp::detail::execution_policy<DerivedPolicy> &,
+                                        RandomAccessIterator1 first,
+                                        RandomAccessIterator1 last,
+                                        RandomAccessIterator2 result)
+{
+  if(first < last && first <= result && result < last)
+  {
+    // result lies in [first, last)
+    // it's safe to use std::copy_backward here
+    thrust::detail::sequential_copy_backward(first, last, result + (last - first));
+    result += (last - first);
+  } // end if
+  else
+  {
+    // result + (last - first) lies in [first, last)
+    // it's safe to use sequential_copy here
+    result = thrust::detail::sequential_copy(first, last, result);
+  } // end else
+
+  return result;
+} // end overlapped_copy()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  RandomAccessIterator2 overlapped_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                                        RandomAccessIterator1 first,
+                                        RandomAccessIterator1 last,
+                                        RandomAccessIterator2 result)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type value_type;
+
+  // make a temporary copy of [first,last), and copy into it first
+  thrust::detail::temporary_array<value_type, DerivedPolicy> temp(exec, first, last);
+  return thrust::copy(exec, temp.begin(), temp.end(), result);
+} // end overlapped_copy()
+
+} // end dispatch
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  RandomAccessIterator2 overlapped_copy(RandomAccessIterator1 first,
+                                        RandomAccessIterator1 last,
+                                        RandomAccessIterator2 result)
+{
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type System1;
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type System2;
+
+  typedef typename thrust::detail::minimum_system<System1, System2>::type System;
+
+  // XXX presumes System is default constructible
+  System system;
+
+  return thrust::detail::dispatch::overlapped_copy(system, first, last, result);
+} // end overlapped_copy()
+
+} // end detail
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/pair.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/pair.inl
new file mode 100644
index 000000000000..4b7dd6eb0fed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/pair.inl
@@ -0,0 +1,231 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/pair.h>
+#include <thrust/detail/swap.h>
+#include <thrust/tuple.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename T1, typename T2>
+  __host__ __device__
+  pair<T1,T2>
+    ::pair(void)
+      :first(),second()
+{
+  ;
+} // end pair::pair()
+
+
+template <typename T1, typename T2>
+  __host__ __device__
+  pair<T1,T2>
+    ::pair(const T1 &x, const T2 &y)
+      :first(x),second(y)
+{
+  ;
+} // end pair::pair()
+
+
+template <typename T1, typename T2>
+  template <typename U1, typename U2>
+    __host__ __device__
+    pair<T1,T2>
+      ::pair(const pair<U1,U2> &p)
+        :first(p.first),second(p.second)
+{
+  ;
+} // end pair::pair()
+
+
+template <typename T1, typename T2>
+  template <typename U1, typename U2>
+    __host__ __device__
+    pair<T1,T2>
+      ::pair(const std::pair<U1,U2> &p)
+        :first(p.first),second(p.second)
+{
+  ;
+} // end pair::pair()
+
+
+template<typename T1, typename T2>
+  inline __host__ __device__
+    void pair<T1,T2>
+      ::swap(thrust::pair<T1,T2> &p)
+{
+  using thrust::swap;
+
+  swap(first, p.first);
+  swap(second, p.second);
+} // end pair::swap()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator==(const pair<T1,T2> &x, const pair<T1,T2> &y)
+{
+  return x.first == y.first && x.second == y.second;
+} // end operator==()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator<(const pair<T1,T2> &x, const pair<T1,T2> &y)
+{
+  return x.first < y.first || (!(y.first < x.first) && x.second < y.second);
+} // end operator<()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator!=(const pair<T1,T2> &x, const pair<T1,T2> &y)
+{
+  return !(x == y);
+} // end operator==()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator>(const pair<T1,T2> &x, const pair<T1,T2> &y)
+{
+  return y < x;
+} // end operator<()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator<=(const pair<T1,T2> &x, const pair<T1,T2> &y)
+{
+  return !(y < x);
+} // end operator<=()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator>=(const pair<T1,T2> &x, const pair<T1,T2> &y)
+{
+  return !(x < y);
+} // end operator>=()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    void swap(pair<T1,T2> &x, pair<T1,T2> &y)
+{
+  return x.swap(y);
+} // end swap()
+
+
+template <typename T1, typename T2>
+  inline __host__ __device__
+    pair<T1,T2> make_pair(T1 x, T2 y)
+{
+  return pair<T1,T2>(x,y);
+} // end make_pair()
+
+
+// specializations of tuple_element for pair
+template<typename T1, typename T2>
+  struct tuple_element<0, pair<T1,T2>>
+{
+  typedef T1 type;
+}; // end tuple_element
+
+template<typename T1, typename T2>
+  struct tuple_element<1, pair<T1,T2>>
+{
+  typedef T2 type;
+}; // end tuple_element
+
+
+// specialization of tuple_size for pair
+template<typename T1, typename T2>
+  struct tuple_size<pair<T1,T2>>
+{
+  static const unsigned int value = 2;
+}; // end tuple_size
+
+
+
+namespace detail
+{
+
+
+template<int N, typename Pair> struct pair_get {};
+
+template<typename Pair>
+  struct pair_get<0, Pair>
+{
+  inline __host__ __device__
+    const typename tuple_element<0, Pair>::type &
+      operator()(const Pair &p) const
+  {
+    return p.first;
+  } // end operator()()
+
+  inline __host__ __device__
+    typename tuple_element<0, Pair>::type &
+      operator()(Pair &p) const
+  {
+    return p.first;
+  } // end operator()()
+}; // end pair_get
+
+
+template<typename Pair>
+  struct pair_get<1, Pair>
+{
+  inline __host__ __device__
+    const typename tuple_element<1, Pair>::type &
+      operator()(const Pair &p) const
+  {
+    return p.second;
+  } // end operator()()
+
+  inline __host__ __device__
+    typename tuple_element<1, Pair>::type &
+      operator()(Pair &p) const
+  {
+    return p.second;
+  } // end operator()()
+}; // end pair_get
+
+} // end detail
+
+
+
+template<unsigned int N, typename T1, typename T2>
+  inline __host__ __device__
+    typename tuple_element<N, pair<T1,T2> >::type &
+      get(pair<T1,T2> &p)
+{
+  return detail::pair_get<N, pair<T1,T2> >()(p);
+} // end get()
+
+template<unsigned int N, typename T1, typename T2>
+  inline __host__ __device__
+    const typename tuple_element<N, pair<T1,T2> >::type &
+      get(const pair<T1,T2> &p)
+{
+  return detail::pair_get<N, pair<T1,T2> >()(p);
+} // end get()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/partition.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/partition.inl
new file mode 100644
index 000000000000..5c51bca80950
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/partition.inl
@@ -0,0 +1,411 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/partition.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/partition.h>
+#include <thrust/system/detail/adl/partition.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::partition;
+  return partition(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end partition()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::partition;
+  return partition(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, pred);
+} // end partition()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred)
+{
+  using thrust::system::detail::generic::partition_copy;
+  return partition_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, out_true, out_false, pred);
+} // end partition_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator1 first,
+                   InputIterator1 last,
+                   InputIterator2 stencil,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred)
+{
+  using thrust::system::detail::generic::partition_copy;
+  return partition_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, out_true, out_false, pred);
+} // end partition_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred)
+{
+  using thrust::system::detail::generic::stable_partition;
+  return stable_partition(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end stable_partition()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred)
+{
+  using thrust::system::detail::generic::stable_partition;
+  return stable_partition(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, pred);
+} // end stable_partition()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  using thrust::system::detail::generic::stable_partition_copy;
+  return stable_partition_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, out_true, out_false, pred);
+} // end stable_partition_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  using thrust::system::detail::generic::stable_partition_copy;
+  return stable_partition_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, out_true, out_false, pred);
+} // end stable_partition_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename Predicate>
+__host__ __device__
+  ForwardIterator partition_point(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Predicate pred)
+{
+  using thrust::system::detail::generic::partition_point;
+  return partition_point(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end partition_point()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+  bool is_partitioned(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  using thrust::system::detail::generic::is_partitioned;
+  return is_partitioned(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end is_partitioned()
+
+
+template<typename ForwardIterator,
+         typename Predicate>
+  ForwardIterator partition(ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::partition(select_system(system), first, last, pred);
+} // end partition()
+
+
+template<typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+  ForwardIterator partition(ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+  typedef typename thrust::iterator_system<InputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::partition(select_system(system1,system2), first, last, stencil, pred);
+} // end partition()
+
+
+template<typename ForwardIterator,
+         typename Predicate>
+  ForwardIterator stable_partition(ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::stable_partition(select_system(system), first, last, pred);
+} // end stable_partition()
+
+
+template<typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+  ForwardIterator stable_partition(ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+  typedef typename thrust::iterator_system<InputIterator>::type   System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::stable_partition(select_system(system1,system2), first, last, stencil, pred);
+} // end stable_partition()
+
+
+template<typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(InputIterator first,
+                   InputIterator last,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type   System1;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::partition_copy(select_system(system1,system2,system3), first, last, out_true, out_false, pred);
+} // end partition_copy()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(InputIterator1 first,
+                   InputIterator1 last,
+                   InputIterator2 stencil,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator1>::type  System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::partition_copy(select_system(system1,system2,system3,system4), first, last, stencil, out_true, out_false, pred);
+} // end partition_copy()
+
+
+template<typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type   System1;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::stable_partition_copy(select_system(system1,system2,system3), first, last, out_true, out_false, pred);
+} // end stable_partition_copy()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type   System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type   System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type  System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type  System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::stable_partition_copy(select_system(system1,system2,system3,system4), first, last, stencil, out_true, out_false, pred);
+} // end stable_partition_copy()
+
+
+template<typename ForwardIterator, typename Predicate>
+  ForwardIterator partition_point(ForwardIterator first,
+                                  ForwardIterator last,
+                                  Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::partition_point(select_system(system), first, last, pred);
+} // end partition_point()
+
+
+template<typename InputIterator, typename Predicate>
+  bool is_partitioned(InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::is_partitioned(select_system(system), first, last, pred);
+} // end is_partitioned()
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.h
new file mode 100644
index 000000000000..aed1fcc2444f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.h
@@ -0,0 +1,255 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief A pointer to a variable which resides in memory associated with a
+ *  system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/detail/iterator_traversal_tags.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/reference_forward_declaration.h>
+#include <ostream>
+#include <cstddef>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename Element, typename Tag, typename Reference = use_default, typename Derived = use_default>
+class pointer;
+
+// Specialize `thrust::iterator_traits` to avoid problems with the name of
+// pointer's constructor shadowing its nested pointer type. We do this before
+// pointer is defined so the specialization is correctly used inside the
+// definition.
+template <typename Element, typename Tag, typename Reference, typename Derived>
+struct iterator_traits<thrust::pointer<Element, Tag, Reference, Derived>>
+{
+  using pointer           = thrust::pointer<Element, Tag, Reference, Derived>;
+  using iterator_category = typename pointer::iterator_category;
+  using value_type        = typename pointer::value_type;
+  using difference_type   = typename pointer::difference_type;
+  using reference         = typename pointer::reference;
+};
+
+THRUST_NAMESPACE_END
+
+namespace std
+{
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+struct iterator_traits<THRUST_NS_QUALIFIER::pointer<Element, Tag, Reference, Derived>>
+{
+  using pointer           = THRUST_NS_QUALIFIER::pointer<Element, Tag, Reference, Derived>;
+  using iterator_category = typename pointer::iterator_category;
+  using value_type        = typename pointer::value_type;
+  using difference_type   = typename pointer::difference_type;
+  using reference         = typename pointer::reference;
+};
+
+} // namespace std
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// this metafunction computes the type of iterator_adaptor thrust::pointer should inherit from
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  struct pointer_base
+{
+  // void pointers should have no element type
+  // note that we remove_cv from the Element type to get the value_type
+  typedef typename thrust::detail::eval_if<
+    thrust::detail::is_void<typename thrust::remove_cvref<Element>::type>::value,
+    thrust::detail::identity_<void>,
+    thrust::detail::remove_cv<Element>
+  >::type value_type;
+
+  // if no Derived type is given, just use pointer
+  typedef typename thrust::detail::eval_if<
+    thrust::detail::is_same<Derived,use_default>::value,
+    thrust::detail::identity_<pointer<Element,Tag,Reference,Derived> >,
+    thrust::detail::identity_<Derived>
+  >::type derived_type;
+
+  // void pointers should have no reference type
+  // if no Reference type is given, just use reference
+  typedef typename thrust::detail::eval_if<
+    thrust::detail::is_void<typename thrust::remove_cvref<Element>::type>::value,
+    thrust::detail::identity_<void>,
+    thrust::detail::eval_if<
+      thrust::detail::is_same<Reference,use_default>::value,
+      thrust::detail::identity_<reference<Element,derived_type> >,
+      thrust::detail::identity_<Reference>
+    >
+  >::type reference_type;
+
+  typedef thrust::iterator_adaptor<
+    derived_type,                        // pass along the type of our Derived class to iterator_adaptor
+    Element *,                           // we adapt a raw pointer
+    value_type,                          // the value type
+    Tag,                                 // system tag
+    thrust::random_access_traversal_tag, // pointers have random access traversal
+    reference_type,                      // pass along our Reference type
+    std::ptrdiff_t
+  > type;
+}; // end pointer_base
+
+
+} // end detail
+
+
+// the base type for all of thrust's tagged pointers.
+// for reasonable pointer-like semantics, derived types should reimplement the following:
+// 1. no-argument constructor
+// 2. constructor from OtherElement *
+// 3. constructor from OtherPointer related by convertibility
+// 4. constructor from OtherPointer to void
+// 5. assignment from OtherPointer related by convertibility
+// These should just call the corresponding members of pointer.
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  class pointer
+    : public thrust::detail::pointer_base<Element,Tag,Reference,Derived>::type
+{
+  private:
+    typedef typename thrust::detail::pointer_base<Element,Tag,Reference,Derived>::type         super_t;
+
+    typedef typename thrust::detail::pointer_base<Element,Tag,Reference,Derived>::derived_type derived_type;
+
+    // friend iterator_core_access to give it access to dereference
+    friend class thrust::iterator_core_access;
+
+    __host__ __device__
+    typename super_t::reference dereference() const;
+
+    // don't provide access to this part of super_t's interface
+    using super_t::base;
+    using typename super_t::base_type;
+
+  public:
+    typedef typename super_t::base_type raw_pointer;
+
+    // constructors
+
+    __host__ __device__
+    pointer();
+
+    // NOTE: This is needed so that Thrust smart pointers can be used in
+    // `std::unique_ptr`.
+    __host__ __device__
+    pointer(std::nullptr_t);
+
+    // OtherValue shall be convertible to Value
+    // XXX consider making the pointer implementation a template parameter which defaults to Element *
+    template<typename OtherElement>
+    __host__ __device__
+    explicit pointer(OtherElement *ptr);
+
+    // OtherPointer's element_type shall be convertible to Element
+    // OtherPointer's system shall be convertible to Tag
+    template<typename OtherPointer>
+    __host__ __device__
+    pointer(const OtherPointer &other,
+            typename thrust::detail::enable_if_pointer_is_convertible<
+              OtherPointer,
+              pointer<Element,Tag,Reference,Derived>
+            >::type * = 0);
+
+    // OtherPointer's element_type shall be void
+    // OtherPointer's system shall be convertible to Tag
+    template<typename OtherPointer>
+    __host__ __device__
+    explicit
+    pointer(const OtherPointer &other,
+            typename thrust::detail::enable_if_void_pointer_is_system_convertible<
+              OtherPointer,
+              pointer<Element,Tag,Reference,Derived>
+            >::type * = 0);
+
+    // assignment
+
+    // NOTE: This is needed so that Thrust smart pointers can be used in
+    // `std::unique_ptr`.
+    __host__ __device__
+    derived_type& operator=(std::nullptr_t);
+
+    // OtherPointer's element_type shall be convertible to Element
+    // OtherPointer's system shall be convertible to Tag
+    template<typename OtherPointer>
+    __host__ __device__
+    typename thrust::detail::enable_if_pointer_is_convertible<
+      OtherPointer,
+      pointer,
+      derived_type &
+    >::type
+    operator=(const OtherPointer &other);
+
+    // observers
+
+    __host__ __device__
+    Element *get() const;
+
+    __host__ __device__
+    Element *operator->() const;
+
+    // NOTE: This is needed so that Thrust smart pointers can be used in
+    // `std::unique_ptr`.
+    __host__ __device__
+    explicit operator bool() const;
+
+    __host__ __device__
+    static derived_type pointer_to(typename thrust::detail::pointer_traits_detail::pointer_to_param<Element>::type r)
+    {
+      return thrust::detail::pointer_traits<derived_type>::pointer_to(r);
+    }
+}; // end pointer
+
+// Output stream operator
+template<typename Element, typename Tag, typename Reference, typename Derived,
+         typename charT, typename traits>
+__host__
+std::basic_ostream<charT, traits> &
+operator<<(std::basic_ostream<charT, traits> &os,
+           const pointer<Element, Tag, Reference, Derived> &p);
+
+// NOTE: This is needed so that Thrust smart pointers can be used in
+// `std::unique_ptr`.
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator==(std::nullptr_t, pointer<Element, Tag, Reference, Derived> p);
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator==(pointer<Element, Tag, Reference, Derived> p, std::nullptr_t);
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator!=(std::nullptr_t, pointer<Element, Tag, Reference, Derived> p);
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator!=(pointer<Element, Tag, Reference, Derived> p, std::nullptr_t);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/pointer.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.inl
new file mode 100644
index 000000000000..de05ff20f0cc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/pointer.inl
@@ -0,0 +1,209 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/pointer.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  pointer<Element,Tag,Reference,Derived>
+    ::pointer()
+      : super_t(static_cast<Element*>(nullptr))
+{} // end pointer::pointer
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  pointer<Element,Tag,Reference,Derived>
+    ::pointer(std::nullptr_t)
+      : super_t(static_cast<Element*>(nullptr))
+{} // end pointer::pointer
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  template<typename OtherElement>
+    __host__ __device__
+    pointer<Element,Tag,Reference,Derived>
+      ::pointer(OtherElement *other)
+        : super_t(other)
+{} // end pointer::pointer
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  template<typename OtherPointer>
+    __host__ __device__
+    pointer<Element,Tag,Reference,Derived>
+      ::pointer(const OtherPointer &other,
+                typename thrust::detail::enable_if_pointer_is_convertible<
+                  OtherPointer,
+                  pointer<Element,Tag,Reference,Derived>
+                 >::type *)
+        : super_t(thrust::detail::pointer_traits<OtherPointer>::get(other))
+{} // end pointer::pointer
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  template<typename OtherPointer>
+    __host__ __device__
+    pointer<Element,Tag,Reference,Derived>
+      ::pointer(const OtherPointer &other,
+                typename thrust::detail::enable_if_void_pointer_is_system_convertible<
+                  OtherPointer,
+                  pointer<Element,Tag,Reference,Derived>
+                 >::type *)
+        : super_t(static_cast<Element *>(thrust::detail::pointer_traits<OtherPointer>::get(other)))
+{} // end pointer::pointer
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  typename pointer<Element,Tag,Reference,Derived>::derived_type &
+    pointer<Element,Tag,Reference,Derived>
+      ::operator=(decltype(nullptr))
+{
+  super_t::base_reference() = nullptr;
+  return static_cast<derived_type&>(*this);
+} // end pointer::operator=
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  template<typename OtherPointer>
+    __host__ __device__
+    typename thrust::detail::enable_if_pointer_is_convertible<
+      OtherPointer,
+      pointer<Element,Tag,Reference,Derived>,
+      typename pointer<Element,Tag,Reference,Derived>::derived_type &
+    >::type
+      pointer<Element,Tag,Reference,Derived>
+        ::operator=(const OtherPointer &other)
+{
+  super_t::base_reference() = thrust::detail::pointer_traits<OtherPointer>::get(other);
+  return static_cast<derived_type&>(*this);
+} // end pointer::operator=
+
+namespace detail
+{
+
+// Implementation for dereference() when Reference is Element&,
+// e.g. cuda's managed_memory_pointer
+template <typename Reference, typename Derived>
+__host__ __device__
+Reference pointer_dereference_impl(const Derived& ptr,
+                                   thrust::detail::true_type /* is_cpp_ref */)
+{
+  return *ptr.get();
+}
+
+// Implementation for pointers with proxy references:
+template <typename Reference, typename Derived>
+__host__ __device__
+Reference pointer_dereference_impl(const Derived& ptr,
+                                   thrust::detail::false_type /* is_cpp_ref */)
+{
+  return Reference(ptr);
+}
+
+} // namespace detail
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  typename pointer<Element,Tag,Reference,Derived>::super_t::reference
+  pointer<Element,Tag,Reference,Derived>
+    ::dereference() const
+{
+  // Need to handle cpp refs and fancy refs differently:
+  typedef typename super_t::reference RefT;
+  typedef typename thrust::detail::is_reference<RefT>::type IsCppRef;
+
+  const derived_type& derivedPtr = static_cast<const derived_type&>(*this);
+
+  return detail::pointer_dereference_impl<RefT>(derivedPtr, IsCppRef());
+} // end pointer::dereference
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  Element *pointer<Element,Tag,Reference,Derived>
+    ::get() const
+{
+  return super_t::base();
+} // end pointer::get
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  Element *pointer<Element,Tag,Reference,Derived>
+    ::operator->() const
+{
+  return super_t::base();
+} // end pointer::operator->
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived>
+  __host__ __device__
+  pointer<Element,Tag,Reference,Derived>
+    ::operator bool() const
+{
+  return bool(get());
+} // end pointer::operator bool
+
+
+template<typename Element, typename Tag, typename Reference, typename Derived,
+         typename charT, typename traits>
+__host__
+std::basic_ostream<charT, traits> &
+operator<<(std::basic_ostream<charT, traits> &os,
+           const pointer<Element, Tag, Reference, Derived> &p) {
+  return os << p.get();
+}
+
+// NOTE: These are needed so that Thrust smart pointers work with
+// `std::unique_ptr`.
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator==(std::nullptr_t, pointer<Element, Tag, Reference, Derived> p)
+{
+  return nullptr == p.get();
+}
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator==(pointer<Element, Tag, Reference, Derived> p, std::nullptr_t)
+{
+  return nullptr == p.get();
+}
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator!=(std::nullptr_t, pointer<Element, Tag, Reference, Derived> p)
+{
+  return !(nullptr == p);
+}
+
+template <typename Element, typename Tag, typename Reference, typename Derived>
+__host__ __device__
+bool operator!=(pointer<Element, Tag, Reference, Derived> p, std::nullptr_t)
+{
+  return !(nullptr == p);
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/preprocessor.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/preprocessor.h
new file mode 100644
index 000000000000..2e850c764dd8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/preprocessor.h
@@ -0,0 +1,1182 @@
+// Copyright (c) 2017-2018 NVIDIA Corporation
+// Copyright (c) 2014-2018 Bryce Adelstein Lelbach
+// Copyright (c) 2001-2015 Housemarque Oy (housemarque.com)
+// Copyright (c) 2007-2015 Hartmut Kaiser
+// Copyright (c)      2002 Peter Dimov and Multi Media Ltd
+//                         (`THRUST_CURRENT_FUNCTION`)
+//
+// Distributed under the Boost Software License v1.0 (boost.org/LICENSE_1_0.txt)
+
+#pragma once
+
+///////////////////////////////////////////////////////////////////////////////
+
+/// \def THRUST_PP_STRINGIZE(expr)
+/// \brief Stringizes the expression \a expr.
+///
+/// \par <b>Example</b>:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << THRUST_PP_STRINGIZE(foo) << "\n";
+/// }
+/// \endcode
+///
+/// The above code expands to:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << "foo" << "\n";
+/// }
+/// \endcode
+///
+#define THRUST_PP_STRINGIZE(expr) THRUST_PP_STRINGIZE_IMPL0(expr)
+#define THRUST_PP_STRINGIZE_IMPL0(expr) #expr
+
+///////////////////////////////////////////////////////////////////////////////
+
+/// \def THRUST_PP_CAT2(a, b)
+/// \brief Concatenates the tokens \a a and \b b.
+///
+/// \par <b>Example</b>:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << THRUST_PP_CAT2(1, THRUST_PP_CAT2(2, 3)) << "\n";
+/// }
+/// \endcode
+///
+/// The above code expands to:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << 123 << "\n";
+/// }
+/// \endcode
+///
+#define THRUST_PP_CAT2(a, b) THRUST_PP_CAT2_IMPL0(a, b)
+
+#if    defined(_MSC_VER)                                                      \
+  && (defined(__EDG__) || defined(__EDG_VERSION__))                         \
+  && (defined(__INTELLISENSE__) || __EDG_VERSION__ >= 308)
+  #define THRUST_PP_CAT2_IMPL0(a, b) THRUST_PP_CAT2_IMPL1(~, a ## b)
+  #define THRUST_PP_CAT2_IMPL1(p, res) res
+#else
+  #define THRUST_PP_CAT2_IMPL0(a, b) a ## b
+#endif
+
+#define THRUST_PP_CAT3(a, b, c)                                               \
+  THRUST_PP_CAT2(a,                                                           \
+    THRUST_PP_CAT2(b, c))                                                     \
+  /**/
+
+#define THRUST_PP_CAT4(a, b, c, d)                                            \
+  THRUST_PP_CAT2(a,                                                           \
+    THRUST_PP_CAT2(b,                                                         \
+      THRUST_PP_CAT2(c, d)))                                                  \
+  /**/
+
+#define THRUST_PP_CAT5(a, b, c, d, e)                                         \
+  THRUST_PP_CAT2(a,                                                           \
+    THRUST_PP_CAT2(b,                                                         \
+      THRUST_PP_CAT2(c,                                                       \
+        THRUST_PP_CAT2(d, e))))                                               \
+  /**/
+
+///////////////////////////////////////////////////////////////////////////////
+
+/// \def THRUST_PP_EXPAND(x)
+/// \brief Performs macro expansion on \a x.
+///
+/// \par <b>Example</b>:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// #define FOO_BAR() "foo_bar"
+/// #define BUZZ()     THRUST_PP_EXPAND(THRUST_PP_CAT2(FOO_, BAR)())
+///
+/// int main()
+/// {
+///   std::cout << BUZZ() << "\n";
+/// }
+/// \endcode
+///
+/// The above code expands to:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << "foo_bar" << "\n";
+/// }
+/// \endcode
+///
+#define THRUST_PP_EXPAND(x) THRUST_PP_EXPAND_IMPL0(x)
+#define THRUST_PP_EXPAND_IMPL0(x) x
+
+#define THRUST_PP_EXPAND_ARGS(...) THRUST_PP_EXPAND_ARGS_IMPL0(__VA_ARGS__)
+#define THRUST_PP_EXPAND_ARGS_IMPL0(...) __VA_ARGS__
+
+#define THRUST_PP_HEAD(x, ...) x
+
+#define THRUST_PP_TAIL(x, ...) __VA_ARGS__
+
+///////////////////////////////////////////////////////////////////////////////
+
+#define THRUST_PP_EMPTY()
+
+#define THRUST_PP_COMMA() ,
+
+///////////////////////////////////////////////////////////////////////////////
+
+#define THRUST_PP_INC(x) THRUST_PP_INC_IMPL0(x)
+
+#define THRUST_PP_INC_IMPL0(x) THRUST_PP_CAT2(THRUST_PP_INC_IMPL_TAG, x)
+
+#define THRUST_PP_INC_IMPL_TAG0 1
+#define THRUST_PP_INC_IMPL_TAG1 2
+#define THRUST_PP_INC_IMPL_TAG2 3
+#define THRUST_PP_INC_IMPL_TAG3 4
+#define THRUST_PP_INC_IMPL_TAG4 5
+#define THRUST_PP_INC_IMPL_TAG5 6
+#define THRUST_PP_INC_IMPL_TAG6 7
+#define THRUST_PP_INC_IMPL_TAG7 8
+#define THRUST_PP_INC_IMPL_TAG8 9
+#define THRUST_PP_INC_IMPL_TAG9 10
+#define THRUST_PP_INC_IMPL_TAG10 11
+#define THRUST_PP_INC_IMPL_TAG11 12
+#define THRUST_PP_INC_IMPL_TAG12 13
+#define THRUST_PP_INC_IMPL_TAG13 14
+#define THRUST_PP_INC_IMPL_TAG14 15
+#define THRUST_PP_INC_IMPL_TAG15 16
+#define THRUST_PP_INC_IMPL_TAG16 17
+#define THRUST_PP_INC_IMPL_TAG17 18
+#define THRUST_PP_INC_IMPL_TAG18 19
+#define THRUST_PP_INC_IMPL_TAG19 20
+#define THRUST_PP_INC_IMPL_TAG20 21
+#define THRUST_PP_INC_IMPL_TAG21 22
+#define THRUST_PP_INC_IMPL_TAG22 23
+#define THRUST_PP_INC_IMPL_TAG23 24
+#define THRUST_PP_INC_IMPL_TAG24 25
+#define THRUST_PP_INC_IMPL_TAG25 26
+#define THRUST_PP_INC_IMPL_TAG26 27
+#define THRUST_PP_INC_IMPL_TAG27 28
+#define THRUST_PP_INC_IMPL_TAG28 29
+#define THRUST_PP_INC_IMPL_TAG29 30
+#define THRUST_PP_INC_IMPL_TAG30 31
+#define THRUST_PP_INC_IMPL_TAG31 32
+#define THRUST_PP_INC_IMPL_TAG32 33
+#define THRUST_PP_INC_IMPL_TAG33 34
+#define THRUST_PP_INC_IMPL_TAG34 35
+#define THRUST_PP_INC_IMPL_TAG35 36
+#define THRUST_PP_INC_IMPL_TAG36 37
+#define THRUST_PP_INC_IMPL_TAG37 38
+#define THRUST_PP_INC_IMPL_TAG38 39
+#define THRUST_PP_INC_IMPL_TAG39 40
+#define THRUST_PP_INC_IMPL_TAG40 41
+#define THRUST_PP_INC_IMPL_TAG41 42
+#define THRUST_PP_INC_IMPL_TAG42 43
+#define THRUST_PP_INC_IMPL_TAG43 44
+#define THRUST_PP_INC_IMPL_TAG44 45
+#define THRUST_PP_INC_IMPL_TAG45 46
+#define THRUST_PP_INC_IMPL_TAG46 47
+#define THRUST_PP_INC_IMPL_TAG47 48
+#define THRUST_PP_INC_IMPL_TAG48 49
+#define THRUST_PP_INC_IMPL_TAG49 50
+#define THRUST_PP_INC_IMPL_TAG50 51
+#define THRUST_PP_INC_IMPL_TAG51 52
+#define THRUST_PP_INC_IMPL_TAG52 53
+#define THRUST_PP_INC_IMPL_TAG53 54
+#define THRUST_PP_INC_IMPL_TAG54 55
+#define THRUST_PP_INC_IMPL_TAG55 56
+#define THRUST_PP_INC_IMPL_TAG56 57
+#define THRUST_PP_INC_IMPL_TAG57 58
+#define THRUST_PP_INC_IMPL_TAG58 59
+#define THRUST_PP_INC_IMPL_TAG59 60
+#define THRUST_PP_INC_IMPL_TAG60 61
+#define THRUST_PP_INC_IMPL_TAG61 62
+#define THRUST_PP_INC_IMPL_TAG62 63
+#define THRUST_PP_INC_IMPL_TAG63 64
+#define THRUST_PP_INC_IMPL_TAG64 65
+#define THRUST_PP_INC_IMPL_TAG65 66
+#define THRUST_PP_INC_IMPL_TAG66 67
+#define THRUST_PP_INC_IMPL_TAG67 68
+#define THRUST_PP_INC_IMPL_TAG68 69
+#define THRUST_PP_INC_IMPL_TAG69 70
+#define THRUST_PP_INC_IMPL_TAG70 71
+#define THRUST_PP_INC_IMPL_TAG71 72
+#define THRUST_PP_INC_IMPL_TAG72 73
+#define THRUST_PP_INC_IMPL_TAG73 74
+#define THRUST_PP_INC_IMPL_TAG74 75
+#define THRUST_PP_INC_IMPL_TAG75 76
+#define THRUST_PP_INC_IMPL_TAG76 77
+#define THRUST_PP_INC_IMPL_TAG77 78
+#define THRUST_PP_INC_IMPL_TAG78 79
+#define THRUST_PP_INC_IMPL_TAG79 80
+#define THRUST_PP_INC_IMPL_TAG80 81
+#define THRUST_PP_INC_IMPL_TAG81 82
+#define THRUST_PP_INC_IMPL_TAG82 83
+#define THRUST_PP_INC_IMPL_TAG83 84
+#define THRUST_PP_INC_IMPL_TAG84 85
+#define THRUST_PP_INC_IMPL_TAG85 86
+#define THRUST_PP_INC_IMPL_TAG86 87
+#define THRUST_PP_INC_IMPL_TAG87 88
+#define THRUST_PP_INC_IMPL_TAG88 89
+#define THRUST_PP_INC_IMPL_TAG89 90
+#define THRUST_PP_INC_IMPL_TAG90 91
+#define THRUST_PP_INC_IMPL_TAG91 92
+#define THRUST_PP_INC_IMPL_TAG92 93
+#define THRUST_PP_INC_IMPL_TAG93 94
+#define THRUST_PP_INC_IMPL_TAG94 95
+#define THRUST_PP_INC_IMPL_TAG95 96
+#define THRUST_PP_INC_IMPL_TAG96 97
+#define THRUST_PP_INC_IMPL_TAG97 98
+#define THRUST_PP_INC_IMPL_TAG98 99
+#define THRUST_PP_INC_IMPL_TAG99 100
+#define THRUST_PP_INC_IMPL_TAG100 101
+#define THRUST_PP_INC_IMPL_TAG101 102
+#define THRUST_PP_INC_IMPL_TAG102 103
+#define THRUST_PP_INC_IMPL_TAG103 104
+#define THRUST_PP_INC_IMPL_TAG104 105
+#define THRUST_PP_INC_IMPL_TAG105 106
+#define THRUST_PP_INC_IMPL_TAG106 107
+#define THRUST_PP_INC_IMPL_TAG107 108
+#define THRUST_PP_INC_IMPL_TAG108 109
+#define THRUST_PP_INC_IMPL_TAG109 110
+#define THRUST_PP_INC_IMPL_TAG110 111
+#define THRUST_PP_INC_IMPL_TAG111 112
+#define THRUST_PP_INC_IMPL_TAG112 113
+#define THRUST_PP_INC_IMPL_TAG113 114
+#define THRUST_PP_INC_IMPL_TAG114 115
+#define THRUST_PP_INC_IMPL_TAG115 116
+#define THRUST_PP_INC_IMPL_TAG116 117
+#define THRUST_PP_INC_IMPL_TAG117 118
+#define THRUST_PP_INC_IMPL_TAG118 119
+#define THRUST_PP_INC_IMPL_TAG119 120
+#define THRUST_PP_INC_IMPL_TAG120 121
+#define THRUST_PP_INC_IMPL_TAG121 122
+#define THRUST_PP_INC_IMPL_TAG122 123
+#define THRUST_PP_INC_IMPL_TAG123 124
+#define THRUST_PP_INC_IMPL_TAG124 125
+#define THRUST_PP_INC_IMPL_TAG125 126
+#define THRUST_PP_INC_IMPL_TAG126 127
+#define THRUST_PP_INC_IMPL_TAG127 128
+#define THRUST_PP_INC_IMPL_TAG128 129
+#define THRUST_PP_INC_IMPL_TAG129 130
+#define THRUST_PP_INC_IMPL_TAG130 131
+#define THRUST_PP_INC_IMPL_TAG131 132
+#define THRUST_PP_INC_IMPL_TAG132 133
+#define THRUST_PP_INC_IMPL_TAG133 134
+#define THRUST_PP_INC_IMPL_TAG134 135
+#define THRUST_PP_INC_IMPL_TAG135 136
+#define THRUST_PP_INC_IMPL_TAG136 137
+#define THRUST_PP_INC_IMPL_TAG137 138
+#define THRUST_PP_INC_IMPL_TAG138 139
+#define THRUST_PP_INC_IMPL_TAG139 140
+#define THRUST_PP_INC_IMPL_TAG140 141
+#define THRUST_PP_INC_IMPL_TAG141 142
+#define THRUST_PP_INC_IMPL_TAG142 143
+#define THRUST_PP_INC_IMPL_TAG143 144
+#define THRUST_PP_INC_IMPL_TAG144 145
+#define THRUST_PP_INC_IMPL_TAG145 146
+#define THRUST_PP_INC_IMPL_TAG146 147
+#define THRUST_PP_INC_IMPL_TAG147 148
+#define THRUST_PP_INC_IMPL_TAG148 149
+#define THRUST_PP_INC_IMPL_TAG149 150
+#define THRUST_PP_INC_IMPL_TAG150 151
+#define THRUST_PP_INC_IMPL_TAG151 152
+#define THRUST_PP_INC_IMPL_TAG152 153
+#define THRUST_PP_INC_IMPL_TAG153 154
+#define THRUST_PP_INC_IMPL_TAG154 155
+#define THRUST_PP_INC_IMPL_TAG155 156
+#define THRUST_PP_INC_IMPL_TAG156 157
+#define THRUST_PP_INC_IMPL_TAG157 158
+#define THRUST_PP_INC_IMPL_TAG158 159
+#define THRUST_PP_INC_IMPL_TAG159 160
+#define THRUST_PP_INC_IMPL_TAG160 161
+#define THRUST_PP_INC_IMPL_TAG161 162
+#define THRUST_PP_INC_IMPL_TAG162 163
+#define THRUST_PP_INC_IMPL_TAG163 164
+#define THRUST_PP_INC_IMPL_TAG164 165
+#define THRUST_PP_INC_IMPL_TAG165 166
+#define THRUST_PP_INC_IMPL_TAG166 167
+#define THRUST_PP_INC_IMPL_TAG167 168
+#define THRUST_PP_INC_IMPL_TAG168 169
+#define THRUST_PP_INC_IMPL_TAG169 170
+#define THRUST_PP_INC_IMPL_TAG170 171
+#define THRUST_PP_INC_IMPL_TAG171 172
+#define THRUST_PP_INC_IMPL_TAG172 173
+#define THRUST_PP_INC_IMPL_TAG173 174
+#define THRUST_PP_INC_IMPL_TAG174 175
+#define THRUST_PP_INC_IMPL_TAG175 176
+#define THRUST_PP_INC_IMPL_TAG176 177
+#define THRUST_PP_INC_IMPL_TAG177 178
+#define THRUST_PP_INC_IMPL_TAG178 179
+#define THRUST_PP_INC_IMPL_TAG179 180
+#define THRUST_PP_INC_IMPL_TAG180 181
+#define THRUST_PP_INC_IMPL_TAG181 182
+#define THRUST_PP_INC_IMPL_TAG182 183
+#define THRUST_PP_INC_IMPL_TAG183 184
+#define THRUST_PP_INC_IMPL_TAG184 185
+#define THRUST_PP_INC_IMPL_TAG185 186
+#define THRUST_PP_INC_IMPL_TAG186 187
+#define THRUST_PP_INC_IMPL_TAG187 188
+#define THRUST_PP_INC_IMPL_TAG188 189
+#define THRUST_PP_INC_IMPL_TAG189 190
+#define THRUST_PP_INC_IMPL_TAG190 191
+#define THRUST_PP_INC_IMPL_TAG191 192
+#define THRUST_PP_INC_IMPL_TAG192 193
+#define THRUST_PP_INC_IMPL_TAG193 194
+#define THRUST_PP_INC_IMPL_TAG194 195
+#define THRUST_PP_INC_IMPL_TAG195 196
+#define THRUST_PP_INC_IMPL_TAG196 197
+#define THRUST_PP_INC_IMPL_TAG197 198
+#define THRUST_PP_INC_IMPL_TAG198 199
+#define THRUST_PP_INC_IMPL_TAG199 200
+#define THRUST_PP_INC_IMPL_TAG200 201
+#define THRUST_PP_INC_IMPL_TAG201 202
+#define THRUST_PP_INC_IMPL_TAG202 203
+#define THRUST_PP_INC_IMPL_TAG203 204
+#define THRUST_PP_INC_IMPL_TAG204 205
+#define THRUST_PP_INC_IMPL_TAG205 206
+#define THRUST_PP_INC_IMPL_TAG206 207
+#define THRUST_PP_INC_IMPL_TAG207 208
+#define THRUST_PP_INC_IMPL_TAG208 209
+#define THRUST_PP_INC_IMPL_TAG209 210
+#define THRUST_PP_INC_IMPL_TAG210 211
+#define THRUST_PP_INC_IMPL_TAG211 212
+#define THRUST_PP_INC_IMPL_TAG212 213
+#define THRUST_PP_INC_IMPL_TAG213 214
+#define THRUST_PP_INC_IMPL_TAG214 215
+#define THRUST_PP_INC_IMPL_TAG215 216
+#define THRUST_PP_INC_IMPL_TAG216 217
+#define THRUST_PP_INC_IMPL_TAG217 218
+#define THRUST_PP_INC_IMPL_TAG218 219
+#define THRUST_PP_INC_IMPL_TAG219 220
+#define THRUST_PP_INC_IMPL_TAG220 221
+#define THRUST_PP_INC_IMPL_TAG221 222
+#define THRUST_PP_INC_IMPL_TAG222 223
+#define THRUST_PP_INC_IMPL_TAG223 224
+#define THRUST_PP_INC_IMPL_TAG224 225
+#define THRUST_PP_INC_IMPL_TAG225 226
+#define THRUST_PP_INC_IMPL_TAG226 227
+#define THRUST_PP_INC_IMPL_TAG227 228
+#define THRUST_PP_INC_IMPL_TAG228 229
+#define THRUST_PP_INC_IMPL_TAG229 230
+#define THRUST_PP_INC_IMPL_TAG230 231
+#define THRUST_PP_INC_IMPL_TAG231 232
+#define THRUST_PP_INC_IMPL_TAG232 233
+#define THRUST_PP_INC_IMPL_TAG233 234
+#define THRUST_PP_INC_IMPL_TAG234 235
+#define THRUST_PP_INC_IMPL_TAG235 236
+#define THRUST_PP_INC_IMPL_TAG236 237
+#define THRUST_PP_INC_IMPL_TAG237 238
+#define THRUST_PP_INC_IMPL_TAG238 239
+#define THRUST_PP_INC_IMPL_TAG239 240
+#define THRUST_PP_INC_IMPL_TAG240 241
+#define THRUST_PP_INC_IMPL_TAG241 242
+#define THRUST_PP_INC_IMPL_TAG242 243
+#define THRUST_PP_INC_IMPL_TAG243 244
+#define THRUST_PP_INC_IMPL_TAG244 245
+#define THRUST_PP_INC_IMPL_TAG245 246
+#define THRUST_PP_INC_IMPL_TAG246 247
+#define THRUST_PP_INC_IMPL_TAG247 248
+#define THRUST_PP_INC_IMPL_TAG248 249
+#define THRUST_PP_INC_IMPL_TAG249 250
+#define THRUST_PP_INC_IMPL_TAG250 251
+#define THRUST_PP_INC_IMPL_TAG251 252
+#define THRUST_PP_INC_IMPL_TAG252 253
+#define THRUST_PP_INC_IMPL_TAG253 254
+#define THRUST_PP_INC_IMPL_TAG254 255
+#define THRUST_PP_INC_IMPL_TAG255 256
+#define THRUST_PP_INC_IMPL_TAG256 256
+
+#define THRUST_PP_DEC(x) THRUST_PP_DEC_IMPL0(x)
+
+#define THRUST_PP_DEC_IMPL0(x) THRUST_PP_CAT2(THRUST_PP_DEC_IMPL_TAG, x)
+
+#define THRUST_PP_DEC_IMPL_TAG0 0
+#define THRUST_PP_DEC_IMPL_TAG1 0
+#define THRUST_PP_DEC_IMPL_TAG2 1
+#define THRUST_PP_DEC_IMPL_TAG3 2
+#define THRUST_PP_DEC_IMPL_TAG4 3
+#define THRUST_PP_DEC_IMPL_TAG5 4
+#define THRUST_PP_DEC_IMPL_TAG6 5
+#define THRUST_PP_DEC_IMPL_TAG7 6
+#define THRUST_PP_DEC_IMPL_TAG8 7
+#define THRUST_PP_DEC_IMPL_TAG9 8
+#define THRUST_PP_DEC_IMPL_TAG10 9
+#define THRUST_PP_DEC_IMPL_TAG11 10
+#define THRUST_PP_DEC_IMPL_TAG12 11
+#define THRUST_PP_DEC_IMPL_TAG13 12
+#define THRUST_PP_DEC_IMPL_TAG14 13
+#define THRUST_PP_DEC_IMPL_TAG15 14
+#define THRUST_PP_DEC_IMPL_TAG16 15
+#define THRUST_PP_DEC_IMPL_TAG17 16
+#define THRUST_PP_DEC_IMPL_TAG18 17
+#define THRUST_PP_DEC_IMPL_TAG19 18
+#define THRUST_PP_DEC_IMPL_TAG20 19
+#define THRUST_PP_DEC_IMPL_TAG21 20
+#define THRUST_PP_DEC_IMPL_TAG22 21
+#define THRUST_PP_DEC_IMPL_TAG23 22
+#define THRUST_PP_DEC_IMPL_TAG24 23
+#define THRUST_PP_DEC_IMPL_TAG25 24
+#define THRUST_PP_DEC_IMPL_TAG26 25
+#define THRUST_PP_DEC_IMPL_TAG27 26
+#define THRUST_PP_DEC_IMPL_TAG28 27
+#define THRUST_PP_DEC_IMPL_TAG29 28
+#define THRUST_PP_DEC_IMPL_TAG30 29
+#define THRUST_PP_DEC_IMPL_TAG31 30
+#define THRUST_PP_DEC_IMPL_TAG32 31
+#define THRUST_PP_DEC_IMPL_TAG33 32
+#define THRUST_PP_DEC_IMPL_TAG34 33
+#define THRUST_PP_DEC_IMPL_TAG35 34
+#define THRUST_PP_DEC_IMPL_TAG36 35
+#define THRUST_PP_DEC_IMPL_TAG37 36
+#define THRUST_PP_DEC_IMPL_TAG38 37
+#define THRUST_PP_DEC_IMPL_TAG39 38
+#define THRUST_PP_DEC_IMPL_TAG40 39
+#define THRUST_PP_DEC_IMPL_TAG41 40
+#define THRUST_PP_DEC_IMPL_TAG42 41
+#define THRUST_PP_DEC_IMPL_TAG43 42
+#define THRUST_PP_DEC_IMPL_TAG44 43
+#define THRUST_PP_DEC_IMPL_TAG45 44
+#define THRUST_PP_DEC_IMPL_TAG46 45
+#define THRUST_PP_DEC_IMPL_TAG47 46
+#define THRUST_PP_DEC_IMPL_TAG48 47
+#define THRUST_PP_DEC_IMPL_TAG49 48
+#define THRUST_PP_DEC_IMPL_TAG50 49
+#define THRUST_PP_DEC_IMPL_TAG51 50
+#define THRUST_PP_DEC_IMPL_TAG52 51
+#define THRUST_PP_DEC_IMPL_TAG53 52
+#define THRUST_PP_DEC_IMPL_TAG54 53
+#define THRUST_PP_DEC_IMPL_TAG55 54
+#define THRUST_PP_DEC_IMPL_TAG56 55
+#define THRUST_PP_DEC_IMPL_TAG57 56
+#define THRUST_PP_DEC_IMPL_TAG58 57
+#define THRUST_PP_DEC_IMPL_TAG59 58
+#define THRUST_PP_DEC_IMPL_TAG60 59
+#define THRUST_PP_DEC_IMPL_TAG61 60
+#define THRUST_PP_DEC_IMPL_TAG62 61
+#define THRUST_PP_DEC_IMPL_TAG63 62
+#define THRUST_PP_DEC_IMPL_TAG64 63
+#define THRUST_PP_DEC_IMPL_TAG65 64
+#define THRUST_PP_DEC_IMPL_TAG66 65
+#define THRUST_PP_DEC_IMPL_TAG67 66
+#define THRUST_PP_DEC_IMPL_TAG68 67
+#define THRUST_PP_DEC_IMPL_TAG69 68
+#define THRUST_PP_DEC_IMPL_TAG70 69
+#define THRUST_PP_DEC_IMPL_TAG71 70
+#define THRUST_PP_DEC_IMPL_TAG72 71
+#define THRUST_PP_DEC_IMPL_TAG73 72
+#define THRUST_PP_DEC_IMPL_TAG74 73
+#define THRUST_PP_DEC_IMPL_TAG75 74
+#define THRUST_PP_DEC_IMPL_TAG76 75
+#define THRUST_PP_DEC_IMPL_TAG77 76
+#define THRUST_PP_DEC_IMPL_TAG78 77
+#define THRUST_PP_DEC_IMPL_TAG79 78
+#define THRUST_PP_DEC_IMPL_TAG80 79
+#define THRUST_PP_DEC_IMPL_TAG81 80
+#define THRUST_PP_DEC_IMPL_TAG82 81
+#define THRUST_PP_DEC_IMPL_TAG83 82
+#define THRUST_PP_DEC_IMPL_TAG84 83
+#define THRUST_PP_DEC_IMPL_TAG85 84
+#define THRUST_PP_DEC_IMPL_TAG86 85
+#define THRUST_PP_DEC_IMPL_TAG87 86
+#define THRUST_PP_DEC_IMPL_TAG88 87
+#define THRUST_PP_DEC_IMPL_TAG89 88
+#define THRUST_PP_DEC_IMPL_TAG90 89
+#define THRUST_PP_DEC_IMPL_TAG91 90
+#define THRUST_PP_DEC_IMPL_TAG92 91
+#define THRUST_PP_DEC_IMPL_TAG93 92
+#define THRUST_PP_DEC_IMPL_TAG94 93
+#define THRUST_PP_DEC_IMPL_TAG95 94
+#define THRUST_PP_DEC_IMPL_TAG96 95
+#define THRUST_PP_DEC_IMPL_TAG97 96
+#define THRUST_PP_DEC_IMPL_TAG98 97
+#define THRUST_PP_DEC_IMPL_TAG99 98
+#define THRUST_PP_DEC_IMPL_TAG100 99
+#define THRUST_PP_DEC_IMPL_TAG101 100
+#define THRUST_PP_DEC_IMPL_TAG102 101
+#define THRUST_PP_DEC_IMPL_TAG103 102
+#define THRUST_PP_DEC_IMPL_TAG104 103
+#define THRUST_PP_DEC_IMPL_TAG105 104
+#define THRUST_PP_DEC_IMPL_TAG106 105
+#define THRUST_PP_DEC_IMPL_TAG107 106
+#define THRUST_PP_DEC_IMPL_TAG108 107
+#define THRUST_PP_DEC_IMPL_TAG109 108
+#define THRUST_PP_DEC_IMPL_TAG110 109
+#define THRUST_PP_DEC_IMPL_TAG111 110
+#define THRUST_PP_DEC_IMPL_TAG112 111
+#define THRUST_PP_DEC_IMPL_TAG113 112
+#define THRUST_PP_DEC_IMPL_TAG114 113
+#define THRUST_PP_DEC_IMPL_TAG115 114
+#define THRUST_PP_DEC_IMPL_TAG116 115
+#define THRUST_PP_DEC_IMPL_TAG117 116
+#define THRUST_PP_DEC_IMPL_TAG118 117
+#define THRUST_PP_DEC_IMPL_TAG119 118
+#define THRUST_PP_DEC_IMPL_TAG120 119
+#define THRUST_PP_DEC_IMPL_TAG121 120
+#define THRUST_PP_DEC_IMPL_TAG122 121
+#define THRUST_PP_DEC_IMPL_TAG123 122
+#define THRUST_PP_DEC_IMPL_TAG124 123
+#define THRUST_PP_DEC_IMPL_TAG125 124
+#define THRUST_PP_DEC_IMPL_TAG126 125
+#define THRUST_PP_DEC_IMPL_TAG127 126
+#define THRUST_PP_DEC_IMPL_TAG128 127
+#define THRUST_PP_DEC_IMPL_TAG129 128
+#define THRUST_PP_DEC_IMPL_TAG130 129
+#define THRUST_PP_DEC_IMPL_TAG131 130
+#define THRUST_PP_DEC_IMPL_TAG132 131
+#define THRUST_PP_DEC_IMPL_TAG133 132
+#define THRUST_PP_DEC_IMPL_TAG134 133
+#define THRUST_PP_DEC_IMPL_TAG135 134
+#define THRUST_PP_DEC_IMPL_TAG136 135
+#define THRUST_PP_DEC_IMPL_TAG137 136
+#define THRUST_PP_DEC_IMPL_TAG138 137
+#define THRUST_PP_DEC_IMPL_TAG139 138
+#define THRUST_PP_DEC_IMPL_TAG140 139
+#define THRUST_PP_DEC_IMPL_TAG141 140
+#define THRUST_PP_DEC_IMPL_TAG142 141
+#define THRUST_PP_DEC_IMPL_TAG143 142
+#define THRUST_PP_DEC_IMPL_TAG144 143
+#define THRUST_PP_DEC_IMPL_TAG145 144
+#define THRUST_PP_DEC_IMPL_TAG146 145
+#define THRUST_PP_DEC_IMPL_TAG147 146
+#define THRUST_PP_DEC_IMPL_TAG148 147
+#define THRUST_PP_DEC_IMPL_TAG149 148
+#define THRUST_PP_DEC_IMPL_TAG150 149
+#define THRUST_PP_DEC_IMPL_TAG151 150
+#define THRUST_PP_DEC_IMPL_TAG152 151
+#define THRUST_PP_DEC_IMPL_TAG153 152
+#define THRUST_PP_DEC_IMPL_TAG154 153
+#define THRUST_PP_DEC_IMPL_TAG155 154
+#define THRUST_PP_DEC_IMPL_TAG156 155
+#define THRUST_PP_DEC_IMPL_TAG157 156
+#define THRUST_PP_DEC_IMPL_TAG158 157
+#define THRUST_PP_DEC_IMPL_TAG159 158
+#define THRUST_PP_DEC_IMPL_TAG160 159
+#define THRUST_PP_DEC_IMPL_TAG161 160
+#define THRUST_PP_DEC_IMPL_TAG162 161
+#define THRUST_PP_DEC_IMPL_TAG163 162
+#define THRUST_PP_DEC_IMPL_TAG164 163
+#define THRUST_PP_DEC_IMPL_TAG165 164
+#define THRUST_PP_DEC_IMPL_TAG166 165
+#define THRUST_PP_DEC_IMPL_TAG167 166
+#define THRUST_PP_DEC_IMPL_TAG168 167
+#define THRUST_PP_DEC_IMPL_TAG169 168
+#define THRUST_PP_DEC_IMPL_TAG170 169
+#define THRUST_PP_DEC_IMPL_TAG171 170
+#define THRUST_PP_DEC_IMPL_TAG172 171
+#define THRUST_PP_DEC_IMPL_TAG173 172
+#define THRUST_PP_DEC_IMPL_TAG174 173
+#define THRUST_PP_DEC_IMPL_TAG175 174
+#define THRUST_PP_DEC_IMPL_TAG176 175
+#define THRUST_PP_DEC_IMPL_TAG177 176
+#define THRUST_PP_DEC_IMPL_TAG178 177
+#define THRUST_PP_DEC_IMPL_TAG179 178
+#define THRUST_PP_DEC_IMPL_TAG180 179
+#define THRUST_PP_DEC_IMPL_TAG181 180
+#define THRUST_PP_DEC_IMPL_TAG182 181
+#define THRUST_PP_DEC_IMPL_TAG183 182
+#define THRUST_PP_DEC_IMPL_TAG184 183
+#define THRUST_PP_DEC_IMPL_TAG185 184
+#define THRUST_PP_DEC_IMPL_TAG186 185
+#define THRUST_PP_DEC_IMPL_TAG187 186
+#define THRUST_PP_DEC_IMPL_TAG188 187
+#define THRUST_PP_DEC_IMPL_TAG189 188
+#define THRUST_PP_DEC_IMPL_TAG190 189
+#define THRUST_PP_DEC_IMPL_TAG191 190
+#define THRUST_PP_DEC_IMPL_TAG192 191
+#define THRUST_PP_DEC_IMPL_TAG193 192
+#define THRUST_PP_DEC_IMPL_TAG194 193
+#define THRUST_PP_DEC_IMPL_TAG195 194
+#define THRUST_PP_DEC_IMPL_TAG196 195
+#define THRUST_PP_DEC_IMPL_TAG197 196
+#define THRUST_PP_DEC_IMPL_TAG198 197
+#define THRUST_PP_DEC_IMPL_TAG199 198
+#define THRUST_PP_DEC_IMPL_TAG200 199
+#define THRUST_PP_DEC_IMPL_TAG201 200
+#define THRUST_PP_DEC_IMPL_TAG202 201
+#define THRUST_PP_DEC_IMPL_TAG203 202
+#define THRUST_PP_DEC_IMPL_TAG204 203
+#define THRUST_PP_DEC_IMPL_TAG205 204
+#define THRUST_PP_DEC_IMPL_TAG206 205
+#define THRUST_PP_DEC_IMPL_TAG207 206
+#define THRUST_PP_DEC_IMPL_TAG208 207
+#define THRUST_PP_DEC_IMPL_TAG209 208
+#define THRUST_PP_DEC_IMPL_TAG210 209
+#define THRUST_PP_DEC_IMPL_TAG211 210
+#define THRUST_PP_DEC_IMPL_TAG212 211
+#define THRUST_PP_DEC_IMPL_TAG213 212
+#define THRUST_PP_DEC_IMPL_TAG214 213
+#define THRUST_PP_DEC_IMPL_TAG215 214
+#define THRUST_PP_DEC_IMPL_TAG216 215
+#define THRUST_PP_DEC_IMPL_TAG217 216
+#define THRUST_PP_DEC_IMPL_TAG218 217
+#define THRUST_PP_DEC_IMPL_TAG219 218
+#define THRUST_PP_DEC_IMPL_TAG220 219
+#define THRUST_PP_DEC_IMPL_TAG221 220
+#define THRUST_PP_DEC_IMPL_TAG222 221
+#define THRUST_PP_DEC_IMPL_TAG223 222
+#define THRUST_PP_DEC_IMPL_TAG224 223
+#define THRUST_PP_DEC_IMPL_TAG225 224
+#define THRUST_PP_DEC_IMPL_TAG226 225
+#define THRUST_PP_DEC_IMPL_TAG227 226
+#define THRUST_PP_DEC_IMPL_TAG228 227
+#define THRUST_PP_DEC_IMPL_TAG229 228
+#define THRUST_PP_DEC_IMPL_TAG230 229
+#define THRUST_PP_DEC_IMPL_TAG231 230
+#define THRUST_PP_DEC_IMPL_TAG232 231
+#define THRUST_PP_DEC_IMPL_TAG233 232
+#define THRUST_PP_DEC_IMPL_TAG234 233
+#define THRUST_PP_DEC_IMPL_TAG235 234
+#define THRUST_PP_DEC_IMPL_TAG236 235
+#define THRUST_PP_DEC_IMPL_TAG237 236
+#define THRUST_PP_DEC_IMPL_TAG238 237
+#define THRUST_PP_DEC_IMPL_TAG239 238
+#define THRUST_PP_DEC_IMPL_TAG240 239
+#define THRUST_PP_DEC_IMPL_TAG241 240
+#define THRUST_PP_DEC_IMPL_TAG242 241
+#define THRUST_PP_DEC_IMPL_TAG243 242
+#define THRUST_PP_DEC_IMPL_TAG244 243
+#define THRUST_PP_DEC_IMPL_TAG245 244
+#define THRUST_PP_DEC_IMPL_TAG246 245
+#define THRUST_PP_DEC_IMPL_TAG247 246
+#define THRUST_PP_DEC_IMPL_TAG248 247
+#define THRUST_PP_DEC_IMPL_TAG249 248
+#define THRUST_PP_DEC_IMPL_TAG250 249
+#define THRUST_PP_DEC_IMPL_TAG251 250
+#define THRUST_PP_DEC_IMPL_TAG252 251
+#define THRUST_PP_DEC_IMPL_TAG253 252
+#define THRUST_PP_DEC_IMPL_TAG254 253
+#define THRUST_PP_DEC_IMPL_TAG255 254
+#define THRUST_PP_DEC_IMPL_TAG256 255
+#define THRUST_PP_DEC_IMPL_TAG257 256
+
+#define THRUST_PP_BOOL(x) THRUST_PP_BOOL_IMPL0(x)
+
+#define THRUST_PP_BOOL_IMPL0(x) THRUST_PP_CAT2(THRUST_PP_BOOL_IMPL_TAG, x)
+
+#define THRUST_PP_BOOL_IMPL_TAG0 0
+#define THRUST_PP_BOOL_IMPL_TAG1 1
+#define THRUST_PP_BOOL_IMPL_TAG2 1
+#define THRUST_PP_BOOL_IMPL_TAG3 1
+#define THRUST_PP_BOOL_IMPL_TAG4 1
+#define THRUST_PP_BOOL_IMPL_TAG5 1
+#define THRUST_PP_BOOL_IMPL_TAG6 1
+#define THRUST_PP_BOOL_IMPL_TAG7 1
+#define THRUST_PP_BOOL_IMPL_TAG8 1
+#define THRUST_PP_BOOL_IMPL_TAG9 1
+#define THRUST_PP_BOOL_IMPL_TAG10 1
+#define THRUST_PP_BOOL_IMPL_TAG11 1
+#define THRUST_PP_BOOL_IMPL_TAG12 1
+#define THRUST_PP_BOOL_IMPL_TAG13 1
+#define THRUST_PP_BOOL_IMPL_TAG14 1
+#define THRUST_PP_BOOL_IMPL_TAG15 1
+#define THRUST_PP_BOOL_IMPL_TAG16 1
+#define THRUST_PP_BOOL_IMPL_TAG17 1
+#define THRUST_PP_BOOL_IMPL_TAG18 1
+#define THRUST_PP_BOOL_IMPL_TAG19 1
+#define THRUST_PP_BOOL_IMPL_TAG20 1
+#define THRUST_PP_BOOL_IMPL_TAG21 1
+#define THRUST_PP_BOOL_IMPL_TAG22 1
+#define THRUST_PP_BOOL_IMPL_TAG23 1
+#define THRUST_PP_BOOL_IMPL_TAG24 1
+#define THRUST_PP_BOOL_IMPL_TAG25 1
+#define THRUST_PP_BOOL_IMPL_TAG26 1
+#define THRUST_PP_BOOL_IMPL_TAG27 1
+#define THRUST_PP_BOOL_IMPL_TAG28 1
+#define THRUST_PP_BOOL_IMPL_TAG29 1
+#define THRUST_PP_BOOL_IMPL_TAG30 1
+#define THRUST_PP_BOOL_IMPL_TAG31 1
+#define THRUST_PP_BOOL_IMPL_TAG32 1
+#define THRUST_PP_BOOL_IMPL_TAG33 1
+#define THRUST_PP_BOOL_IMPL_TAG34 1
+#define THRUST_PP_BOOL_IMPL_TAG35 1
+#define THRUST_PP_BOOL_IMPL_TAG36 1
+#define THRUST_PP_BOOL_IMPL_TAG37 1
+#define THRUST_PP_BOOL_IMPL_TAG38 1
+#define THRUST_PP_BOOL_IMPL_TAG39 1
+#define THRUST_PP_BOOL_IMPL_TAG40 1
+#define THRUST_PP_BOOL_IMPL_TAG41 1
+#define THRUST_PP_BOOL_IMPL_TAG42 1
+#define THRUST_PP_BOOL_IMPL_TAG43 1
+#define THRUST_PP_BOOL_IMPL_TAG44 1
+#define THRUST_PP_BOOL_IMPL_TAG45 1
+#define THRUST_PP_BOOL_IMPL_TAG46 1
+#define THRUST_PP_BOOL_IMPL_TAG47 1
+#define THRUST_PP_BOOL_IMPL_TAG48 1
+#define THRUST_PP_BOOL_IMPL_TAG49 1
+#define THRUST_PP_BOOL_IMPL_TAG50 1
+#define THRUST_PP_BOOL_IMPL_TAG51 1
+#define THRUST_PP_BOOL_IMPL_TAG52 1
+#define THRUST_PP_BOOL_IMPL_TAG53 1
+#define THRUST_PP_BOOL_IMPL_TAG54 1
+#define THRUST_PP_BOOL_IMPL_TAG55 1
+#define THRUST_PP_BOOL_IMPL_TAG56 1
+#define THRUST_PP_BOOL_IMPL_TAG57 1
+#define THRUST_PP_BOOL_IMPL_TAG58 1
+#define THRUST_PP_BOOL_IMPL_TAG59 1
+#define THRUST_PP_BOOL_IMPL_TAG60 1
+#define THRUST_PP_BOOL_IMPL_TAG61 1
+#define THRUST_PP_BOOL_IMPL_TAG62 1
+#define THRUST_PP_BOOL_IMPL_TAG63 1
+#define THRUST_PP_BOOL_IMPL_TAG64 1
+#define THRUST_PP_BOOL_IMPL_TAG65 1
+#define THRUST_PP_BOOL_IMPL_TAG66 1
+#define THRUST_PP_BOOL_IMPL_TAG67 1
+#define THRUST_PP_BOOL_IMPL_TAG68 1
+#define THRUST_PP_BOOL_IMPL_TAG69 1
+#define THRUST_PP_BOOL_IMPL_TAG70 1
+#define THRUST_PP_BOOL_IMPL_TAG71 1
+#define THRUST_PP_BOOL_IMPL_TAG72 1
+#define THRUST_PP_BOOL_IMPL_TAG73 1
+#define THRUST_PP_BOOL_IMPL_TAG74 1
+#define THRUST_PP_BOOL_IMPL_TAG75 1
+#define THRUST_PP_BOOL_IMPL_TAG76 1
+#define THRUST_PP_BOOL_IMPL_TAG77 1
+#define THRUST_PP_BOOL_IMPL_TAG78 1
+#define THRUST_PP_BOOL_IMPL_TAG79 1
+#define THRUST_PP_BOOL_IMPL_TAG80 1
+#define THRUST_PP_BOOL_IMPL_TAG81 1
+#define THRUST_PP_BOOL_IMPL_TAG82 1
+#define THRUST_PP_BOOL_IMPL_TAG83 1
+#define THRUST_PP_BOOL_IMPL_TAG84 1
+#define THRUST_PP_BOOL_IMPL_TAG85 1
+#define THRUST_PP_BOOL_IMPL_TAG86 1
+#define THRUST_PP_BOOL_IMPL_TAG87 1
+#define THRUST_PP_BOOL_IMPL_TAG88 1
+#define THRUST_PP_BOOL_IMPL_TAG89 1
+#define THRUST_PP_BOOL_IMPL_TAG90 1
+#define THRUST_PP_BOOL_IMPL_TAG91 1
+#define THRUST_PP_BOOL_IMPL_TAG92 1
+#define THRUST_PP_BOOL_IMPL_TAG93 1
+#define THRUST_PP_BOOL_IMPL_TAG94 1
+#define THRUST_PP_BOOL_IMPL_TAG95 1
+#define THRUST_PP_BOOL_IMPL_TAG96 1
+#define THRUST_PP_BOOL_IMPL_TAG97 1
+#define THRUST_PP_BOOL_IMPL_TAG98 1
+#define THRUST_PP_BOOL_IMPL_TAG99 1
+#define THRUST_PP_BOOL_IMPL_TAG100 1
+#define THRUST_PP_BOOL_IMPL_TAG101 1
+#define THRUST_PP_BOOL_IMPL_TAG102 1
+#define THRUST_PP_BOOL_IMPL_TAG103 1
+#define THRUST_PP_BOOL_IMPL_TAG104 1
+#define THRUST_PP_BOOL_IMPL_TAG105 1
+#define THRUST_PP_BOOL_IMPL_TAG106 1
+#define THRUST_PP_BOOL_IMPL_TAG107 1
+#define THRUST_PP_BOOL_IMPL_TAG108 1
+#define THRUST_PP_BOOL_IMPL_TAG109 1
+#define THRUST_PP_BOOL_IMPL_TAG110 1
+#define THRUST_PP_BOOL_IMPL_TAG111 1
+#define THRUST_PP_BOOL_IMPL_TAG112 1
+#define THRUST_PP_BOOL_IMPL_TAG113 1
+#define THRUST_PP_BOOL_IMPL_TAG114 1
+#define THRUST_PP_BOOL_IMPL_TAG115 1
+#define THRUST_PP_BOOL_IMPL_TAG116 1
+#define THRUST_PP_BOOL_IMPL_TAG117 1
+#define THRUST_PP_BOOL_IMPL_TAG118 1
+#define THRUST_PP_BOOL_IMPL_TAG119 1
+#define THRUST_PP_BOOL_IMPL_TAG120 1
+#define THRUST_PP_BOOL_IMPL_TAG121 1
+#define THRUST_PP_BOOL_IMPL_TAG122 1
+#define THRUST_PP_BOOL_IMPL_TAG123 1
+#define THRUST_PP_BOOL_IMPL_TAG124 1
+#define THRUST_PP_BOOL_IMPL_TAG125 1
+#define THRUST_PP_BOOL_IMPL_TAG126 1
+#define THRUST_PP_BOOL_IMPL_TAG127 1
+#define THRUST_PP_BOOL_IMPL_TAG128 1
+#define THRUST_PP_BOOL_IMPL_TAG129 1
+#define THRUST_PP_BOOL_IMPL_TAG130 1
+#define THRUST_PP_BOOL_IMPL_TAG131 1
+#define THRUST_PP_BOOL_IMPL_TAG132 1
+#define THRUST_PP_BOOL_IMPL_TAG133 1
+#define THRUST_PP_BOOL_IMPL_TAG134 1
+#define THRUST_PP_BOOL_IMPL_TAG135 1
+#define THRUST_PP_BOOL_IMPL_TAG136 1
+#define THRUST_PP_BOOL_IMPL_TAG137 1
+#define THRUST_PP_BOOL_IMPL_TAG138 1
+#define THRUST_PP_BOOL_IMPL_TAG139 1
+#define THRUST_PP_BOOL_IMPL_TAG140 1
+#define THRUST_PP_BOOL_IMPL_TAG141 1
+#define THRUST_PP_BOOL_IMPL_TAG142 1
+#define THRUST_PP_BOOL_IMPL_TAG143 1
+#define THRUST_PP_BOOL_IMPL_TAG144 1
+#define THRUST_PP_BOOL_IMPL_TAG145 1
+#define THRUST_PP_BOOL_IMPL_TAG146 1
+#define THRUST_PP_BOOL_IMPL_TAG147 1
+#define THRUST_PP_BOOL_IMPL_TAG148 1
+#define THRUST_PP_BOOL_IMPL_TAG149 1
+#define THRUST_PP_BOOL_IMPL_TAG150 1
+#define THRUST_PP_BOOL_IMPL_TAG151 1
+#define THRUST_PP_BOOL_IMPL_TAG152 1
+#define THRUST_PP_BOOL_IMPL_TAG153 1
+#define THRUST_PP_BOOL_IMPL_TAG154 1
+#define THRUST_PP_BOOL_IMPL_TAG155 1
+#define THRUST_PP_BOOL_IMPL_TAG156 1
+#define THRUST_PP_BOOL_IMPL_TAG157 1
+#define THRUST_PP_BOOL_IMPL_TAG158 1
+#define THRUST_PP_BOOL_IMPL_TAG159 1
+#define THRUST_PP_BOOL_IMPL_TAG160 1
+#define THRUST_PP_BOOL_IMPL_TAG161 1
+#define THRUST_PP_BOOL_IMPL_TAG162 1
+#define THRUST_PP_BOOL_IMPL_TAG163 1
+#define THRUST_PP_BOOL_IMPL_TAG164 1
+#define THRUST_PP_BOOL_IMPL_TAG165 1
+#define THRUST_PP_BOOL_IMPL_TAG166 1
+#define THRUST_PP_BOOL_IMPL_TAG167 1
+#define THRUST_PP_BOOL_IMPL_TAG168 1
+#define THRUST_PP_BOOL_IMPL_TAG169 1
+#define THRUST_PP_BOOL_IMPL_TAG170 1
+#define THRUST_PP_BOOL_IMPL_TAG171 1
+#define THRUST_PP_BOOL_IMPL_TAG172 1
+#define THRUST_PP_BOOL_IMPL_TAG173 1
+#define THRUST_PP_BOOL_IMPL_TAG174 1
+#define THRUST_PP_BOOL_IMPL_TAG175 1
+#define THRUST_PP_BOOL_IMPL_TAG176 1
+#define THRUST_PP_BOOL_IMPL_TAG177 1
+#define THRUST_PP_BOOL_IMPL_TAG178 1
+#define THRUST_PP_BOOL_IMPL_TAG179 1
+#define THRUST_PP_BOOL_IMPL_TAG180 1
+#define THRUST_PP_BOOL_IMPL_TAG181 1
+#define THRUST_PP_BOOL_IMPL_TAG182 1
+#define THRUST_PP_BOOL_IMPL_TAG183 1
+#define THRUST_PP_BOOL_IMPL_TAG184 1
+#define THRUST_PP_BOOL_IMPL_TAG185 1
+#define THRUST_PP_BOOL_IMPL_TAG186 1
+#define THRUST_PP_BOOL_IMPL_TAG187 1
+#define THRUST_PP_BOOL_IMPL_TAG188 1
+#define THRUST_PP_BOOL_IMPL_TAG189 1
+#define THRUST_PP_BOOL_IMPL_TAG190 1
+#define THRUST_PP_BOOL_IMPL_TAG191 1
+#define THRUST_PP_BOOL_IMPL_TAG192 1
+#define THRUST_PP_BOOL_IMPL_TAG193 1
+#define THRUST_PP_BOOL_IMPL_TAG194 1
+#define THRUST_PP_BOOL_IMPL_TAG195 1
+#define THRUST_PP_BOOL_IMPL_TAG196 1
+#define THRUST_PP_BOOL_IMPL_TAG197 1
+#define THRUST_PP_BOOL_IMPL_TAG198 1
+#define THRUST_PP_BOOL_IMPL_TAG199 1
+#define THRUST_PP_BOOL_IMPL_TAG200 1
+#define THRUST_PP_BOOL_IMPL_TAG201 1
+#define THRUST_PP_BOOL_IMPL_TAG202 1
+#define THRUST_PP_BOOL_IMPL_TAG203 1
+#define THRUST_PP_BOOL_IMPL_TAG204 1
+#define THRUST_PP_BOOL_IMPL_TAG205 1
+#define THRUST_PP_BOOL_IMPL_TAG206 1
+#define THRUST_PP_BOOL_IMPL_TAG207 1
+#define THRUST_PP_BOOL_IMPL_TAG208 1
+#define THRUST_PP_BOOL_IMPL_TAG209 1
+#define THRUST_PP_BOOL_IMPL_TAG210 1
+#define THRUST_PP_BOOL_IMPL_TAG211 1
+#define THRUST_PP_BOOL_IMPL_TAG212 1
+#define THRUST_PP_BOOL_IMPL_TAG213 1
+#define THRUST_PP_BOOL_IMPL_TAG214 1
+#define THRUST_PP_BOOL_IMPL_TAG215 1
+#define THRUST_PP_BOOL_IMPL_TAG216 1
+#define THRUST_PP_BOOL_IMPL_TAG217 1
+#define THRUST_PP_BOOL_IMPL_TAG218 1
+#define THRUST_PP_BOOL_IMPL_TAG219 1
+#define THRUST_PP_BOOL_IMPL_TAG220 1
+#define THRUST_PP_BOOL_IMPL_TAG221 1
+#define THRUST_PP_BOOL_IMPL_TAG222 1
+#define THRUST_PP_BOOL_IMPL_TAG223 1
+#define THRUST_PP_BOOL_IMPL_TAG224 1
+#define THRUST_PP_BOOL_IMPL_TAG225 1
+#define THRUST_PP_BOOL_IMPL_TAG226 1
+#define THRUST_PP_BOOL_IMPL_TAG227 1
+#define THRUST_PP_BOOL_IMPL_TAG228 1
+#define THRUST_PP_BOOL_IMPL_TAG229 1
+#define THRUST_PP_BOOL_IMPL_TAG230 1
+#define THRUST_PP_BOOL_IMPL_TAG231 1
+#define THRUST_PP_BOOL_IMPL_TAG232 1
+#define THRUST_PP_BOOL_IMPL_TAG233 1
+#define THRUST_PP_BOOL_IMPL_TAG234 1
+#define THRUST_PP_BOOL_IMPL_TAG235 1
+#define THRUST_PP_BOOL_IMPL_TAG236 1
+#define THRUST_PP_BOOL_IMPL_TAG237 1
+#define THRUST_PP_BOOL_IMPL_TAG238 1
+#define THRUST_PP_BOOL_IMPL_TAG239 1
+#define THRUST_PP_BOOL_IMPL_TAG240 1
+#define THRUST_PP_BOOL_IMPL_TAG241 1
+#define THRUST_PP_BOOL_IMPL_TAG242 1
+#define THRUST_PP_BOOL_IMPL_TAG243 1
+#define THRUST_PP_BOOL_IMPL_TAG244 1
+#define THRUST_PP_BOOL_IMPL_TAG245 1
+#define THRUST_PP_BOOL_IMPL_TAG246 1
+#define THRUST_PP_BOOL_IMPL_TAG247 1
+#define THRUST_PP_BOOL_IMPL_TAG248 1
+#define THRUST_PP_BOOL_IMPL_TAG249 1
+#define THRUST_PP_BOOL_IMPL_TAG250 1
+#define THRUST_PP_BOOL_IMPL_TAG251 1
+#define THRUST_PP_BOOL_IMPL_TAG252 1
+#define THRUST_PP_BOOL_IMPL_TAG253 1
+#define THRUST_PP_BOOL_IMPL_TAG254 1
+#define THRUST_PP_BOOL_IMPL_TAG255 1
+#define THRUST_PP_BOOL_IMPL_TAG256 1
+
+///////////////////////////////////////////////////////////////////////////////
+
+#define THRUST_PP_IIF(bit, t, f) THRUST_PP_IIF_IMPL0(bit, t, f)
+
+#if defined(_MSC_VER)
+  #define THRUST_PP_IIF_IMPL0(bit, t, f)                                      \
+    THRUST_PP_IIF_IMPL1(THRUST_PP_CAT2(THRUST_PP_IIF_IMPL_TAG, bit(t, f)))    \
+    /**/
+  #define THRUST_PP_IIF_IMPL1(id) id
+#else
+  #define THRUST_PP_IIF_IMPL0(bit, t, f)                                      \
+    THRUST_PP_CAT2(THRUST_PP_IIF_IMPL_TAG, bit(t, f))
+    /**/
+#endif
+
+#define THRUST_PP_IIF_IMPL_TAG0(t, f) f
+#define THRUST_PP_IIF_IMPL_TAG1(t, f) t
+
+#if defined(__EDG__)
+  #define THRUST_PP_IF(cond, t, f) THRUST_PP_IF_IMPL0(cond, t, f)
+  #define THRUST_PP_IF_IMPL0(cond, t, f)                                      \
+    THRUST_PP_IIF(THRUST_PP_BOOL(cond), t, f)                                 \
+    /**/
+#else
+  #define THRUST_PP_IF(cond, t, f) THRUST_PP_IIF(THRUST_PP_BOOL(cond), t, f)
+#endif
+
+/// \def THRUST_COMMA_IF(cond)
+/// \brief If \a cond is true, expands to a comma. Otherwise, expands to nothing.
+///
+/// \par <b>Example</b>:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << THRUST_PP_STRINGIZE(THRUST_COMMA_IF(0)) << "\n"
+///             << THRUST_PP_STRINGIZE(THRUST_COMMA_IF(1)) << "\n";
+/// }
+/// \endcode
+///
+/// The above code expands to:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << ""  << "\n"
+///             << "," << "\n";
+/// }
+/// \endcode
+///
+#if defined(__EDG__)
+  #define THRUST_PP_COMMA_IF(cond) THRUST_PP_COMMA_IF_IMPL0(cond)
+  #define THRUST_PP_COMMA_IF_IMPL0(cond)                                      \
+    THRUST_PP_IF(cond, THRUST_PP_COMMA, THRUST_PP_EMPTY)()                    \
+    /**/
+#else
+  #define THRUST_PP_COMMA_IF(cond)                                            \
+    THRUST_PP_IF(cond, THRUST_PP_COMMA, THRUST_PP_EMPTY)()                    \
+    /**/
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+// http://gustedt.wordpress.com/2010/06/08/detect-empty-macro-arguments
+
+#define THRUST_PP_64TH_ARG(                                                   \
+     _1, _2, _3, _4, _5, _6, _7, _8, _9,_10,_11,_12,_13,_14,_15,_16           \
+  , _17,_18,_19,_20,_21,_22,_23,_24,_25,_26,_27,_28,_29,_30,_31,_32           \
+  , _33,_34,_35,_36,_37,_38,_39,_40,_41,_42,_43,_44,_45,_46,_47,_48           \
+  , _49,_50,_51,_52,_53,_54,_55,_56,_57,_58,_59,_60,_61,_62,_63,  N           \
+  , ...                                                                       \
+  ) N                                                                         \
+  /**/
+
+#define THRUST_PP_HAS_COMMA(...)                                              \
+  THRUST_PP_EXPAND(THRUST_PP_64TH_ARG(                                        \
+    __VA_ARGS__                                                               \
+  , 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1                                           \
+  , 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1                                           \
+  , 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1                                           \
+  , 1,1,1,1,1,1,1,1,1,1,1,1,1,1,0                                             \
+  ))                                                                          \
+  /**/
+
+#define THRUST_PP_TRIGGER_PAREN(...) ,
+
+#define THRUST_PP_IS_VARIADIC_NULLARY(...)                                    \
+  THRUST_PP_IS_VARIADIC_NULLARY_IMPL0(                                        \
+    /* Test if there is just one argument, eventually an empty one. */        \
+    THRUST_PP_HAS_COMMA(__VA_ARGS__),                                         \
+    /* Test if THRUST_PP_TRIGGER_PAREN together with the argument adds a */   \
+    /* comma. */                                                              \
+    THRUST_PP_HAS_COMMA(THRUST_PP_TRIGGER_PAREN __VA_ARGS__),                 \
+    /* Test if the argument together with a parenthesis adds a comma. */      \
+    THRUST_PP_HAS_COMMA(__VA_ARGS__ (/*empty*/)),                             \
+    /* Test if placing it between THRUST_PP_TRIGGER_PAREN and the */          \
+    /* parenthesis adds a comma. */                                           \
+    THRUST_PP_HAS_COMMA(THRUST_PP_TRIGGER_PAREN __VA_ARGS__ (/*empty*/))      \
+  )                                                                           \
+  /**/
+
+#define THRUST_PP_IS_VARIADIC_NULLARY_IMPL0(_0, _1, _2, _3)                   \
+  THRUST_PP_HAS_COMMA(                                                        \
+    THRUST_PP_CAT5(THRUST_PP_IS_VARIADIC_NULLARY_IMPL_TAG, _0, _1, _2, _3)    \
+  )                                                                           \
+
+#define THRUST_PP_IS_VARIADIC_NULLARY_IMPL_TAG0001 ,
+
+///////////////////////////////////////////////////////////////////////////////
+
+/// \def THRUST_PP_ARITY(...)
+/// \brief Returns the number of arguments that it was called with. Must be
+///        called with less than 64 arguments.
+///
+/// \par <b>Example</b>:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << THRUST_PP_ARITY()        << "\n"
+///             << THRUST_PP_ARITY(x)       << "\n"
+///             << THRUST_PP_ARITY(x, y)    << "\n"
+///             << THRUST_PP_ARITY(x, y, z) << "\n";
+/// }
+/// \endcode
+///
+/// The above code expands to:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << 0 << "\n"
+///             << 1 << "\n"
+///             << 2 << "\n"
+///             << 3 << "\n";
+/// }
+/// \endcode
+///
+#define THRUST_PP_ARITY(...)                                                  \
+  THRUST_PP_EXPAND(                                                           \
+    THRUST_PP_IF(                                                             \
+      THRUST_PP_IS_VARIADIC_NULLARY(__VA_ARGS__)                              \
+    , 0                                                                       \
+    , THRUST_PP_64TH_ARG(                                                     \
+        __VA_ARGS__                                                           \
+      , 63,62,61,60,59,58,57,56,55,54,53,52,51,50,49,48                       \
+      , 47,46,45,44,43,42,41,40,39,38,37,36,35,34,33,32                       \
+      , 31,30,29,28,27,26,25,24,23,22,21,20,19,18,17,16                       \
+      , 15,14,13,12,11,10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0                       \
+      )                                                                       \
+    )                                                                         \
+  )                                                                           \
+  /**/
+
+/// \def THRUST_PP_DISPATCH(basename, ...)
+/// \brief Expands to <tt>basenameN(...)</tt>, where <tt>N</tt> is the
+///        number of variadic arguments that \a THRUST_PP_DISPATCH was called
+///        with. This macro can be used to implement "macro overloading".
+///
+/// \par <b>Example</b>:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// #define PLUS(...) THRUST_PP_DISPATCH(PLUS, __VA_ARGS__)
+/// #define PLUS0()        0
+/// #define PLUS1(x)       x
+/// #define PLUS2(x, y)    x + y
+/// #define PLUS3(x, y, z) x + y + z
+///
+/// int main()
+/// {
+///   std::cout << PLUS()        << "\n"
+///             << PLUS(1)       << "\n"
+///             << PLUS(1, 2)    << "\n"
+///             << PLUS(1, 2, 3) << "\n";
+/// }
+/// \endcode
+///
+/// The above code expands to:
+///
+/// \code
+/// #include <thrust/detail/preprocessor.h>
+/// #include <iostream>
+///
+/// int main()
+/// {
+///   std::cout << 0         << "\n"
+///             << 1         << "\n"
+///             << 1 + 2     << "\n"
+///             << 1 + 2 + 3 << "\n";
+/// }
+/// \endcode
+///
+#define THRUST_PP_DISPATCH(basename, ...)                                     \
+  THRUST_PP_EXPAND(                                                           \
+    THRUST_PP_CAT2(                                                           \
+      basename,                                                               \
+      THRUST_PP_ARITY(__VA_ARGS__)                                            \
+    )(__VA_ARGS__)                                                            \
+  )                                                                           \
+  /**/
+
+///////////////////////////////////////////////////////////////////////////////
+
+/// \def THRUST_CURRENT_FUNCTION
+/// \brief The name of the current function as a string.
+///
+#if    defined(__GNUC__)                                                      \
+    || (defined(__MWERKS__) && (__MWERKS__ >= 0x3000))                        \
+    || (defined(__ICC) && (__ICC >= 600)) || defined(__ghs__)
+  #define THRUST_CURRENT_FUNCTION __PRETTY_FUNCTION__
+#elif defined(__DMC__) && (__DMC__ >= 0x810)
+  #define THRUST_CURRENT_FUNCTION __PRETTY_FUNCTION__
+#elif defined(__FUNCSIG__)
+  #define THRUST_CURRENT_FUNCTION __FUNCSIG__
+#elif    (defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 600))             \
+      || (defined(__IBMCTHRUST_PP__) && (__IBMCTHRUST_PP__ >= 500))
+  #define THRUST_CURRENT_FUNCTION __FUNCTION__
+#elif defined(__BORLANDC__) && (__BORLANDC__ >= 0x550)
+  #define THRUST_CURRENT_FUNCTION __FUNC__
+#elif defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901)
+  #define THRUST_CURRENT_FUNCTION __func__
+#elif defined(__cplusplus) && (__cplusplus >= 201103)
+  #define THRUST_CURRENT_FUNCTION __func__
+#else
+  #define THRUST_CURRENT_FUNCTION "(unknown)"
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/range/head_flags.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/range/head_flags.h
new file mode 100644
index 000000000000..b755840c9916
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/range/head_flags.h
@@ -0,0 +1,229 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/tuple.h>
+#include <thrust/functional.h>
+
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename RandomAccessIterator,
+         typename BinaryPredicate = thrust::equal_to<typename thrust::iterator_value<RandomAccessIterator>::type>,
+         typename ValueType = bool,
+         typename IndexType = typename thrust::iterator_difference<RandomAccessIterator>::type>
+  class head_flags_with_init
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type init_type;
+
+  // XXX WAR cudafe issue
+  //private:
+  public:
+    struct head_flag_functor
+    {
+      BinaryPredicate binary_pred; // this must be the first member for performance reasons
+      init_type init;
+      IndexType n;
+
+      typedef ValueType result_type;
+
+      __host__ __device__
+      head_flag_functor(init_type init, IndexType n)
+        : binary_pred(), init(init), n(n)
+      {}
+
+      __host__ __device__
+      head_flag_functor(init_type init, IndexType n, BinaryPredicate binary_pred)
+        : binary_pred(binary_pred), init(init), n(n)
+      {}
+
+      template<typename Tuple>
+      __host__ __device__ __thrust_forceinline__
+      result_type operator()(const Tuple &t)
+      {
+        const IndexType i = thrust::get<0>(t);
+
+        if(i == 0)
+        {
+          return !binary_pred(init, thrust::get<1>(t));
+        }
+
+        return !binary_pred(thrust::get<1>(t), thrust::get<2>(t));
+      }
+    };
+
+    typedef thrust::counting_iterator<IndexType> counting_iterator;
+
+  public:
+    typedef thrust::transform_iterator<
+      head_flag_functor,
+      thrust::zip_iterator<thrust::tuple<counting_iterator,RandomAccessIterator,RandomAccessIterator> >
+    > iterator;
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    head_flags_with_init(RandomAccessIterator first, RandomAccessIterator last, init_type init)
+      : m_begin(thrust::make_transform_iterator(thrust::make_zip_iterator(thrust::make_tuple(thrust::counting_iterator<IndexType>(0), first, first - 1)),
+                                                head_flag_functor(init, last - first))),
+        m_end(m_begin + (last - first))
+    {}
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    head_flags_with_init(RandomAccessIterator first, RandomAccessIterator last, init_type init, BinaryPredicate binary_pred)
+      : m_begin(thrust::make_transform_iterator(thrust::make_zip_iterator(thrust::make_tuple(thrust::counting_iterator<IndexType>(0), first, first - 1)),
+                                                head_flag_functor(init, last - first, binary_pred))),
+        m_end(m_begin + (last - first))
+    {}
+
+    __host__ __device__
+    iterator begin() const
+    {
+      return m_begin;
+    }
+
+    __host__ __device__
+    iterator end() const
+    {
+      return m_end;
+    }
+
+    template<typename OtherIndex>
+    __host__ __device__
+    typename iterator::reference operator[](OtherIndex i)
+    {
+      return *(begin() + i);
+    }
+
+  private:
+    iterator m_begin, m_end;
+};
+
+
+
+template<typename RandomAccessIterator,
+         typename BinaryPredicate = thrust::equal_to<typename thrust::iterator_value<RandomAccessIterator>::type>,
+         typename ValueType = bool,
+         typename IndexType = typename thrust::iterator_difference<RandomAccessIterator>::type>
+  class head_flags
+{
+  // XXX WAR cudafe issue
+  //private:
+  public:
+    struct head_flag_functor
+    {
+      BinaryPredicate binary_pred; // this must be the first member for performance reasons
+      IndexType n;
+
+      typedef ValueType result_type;
+
+      __host__ __device__
+      head_flag_functor(IndexType n)
+        : binary_pred(), n(n)
+      {}
+
+      __host__ __device__
+      head_flag_functor(IndexType n, BinaryPredicate binary_pred)
+        : binary_pred(binary_pred), n(n)
+      {}
+
+      template<typename Tuple>
+      __host__ __device__ __thrust_forceinline__
+      result_type operator()(const Tuple &t)
+      {
+        const IndexType i = thrust::get<0>(t);
+
+        // note that we do not dereference the tuple's 2nd element when i <= 0
+        // and therefore do not dereference a bad location at the boundary
+        return (i == 0 || !binary_pred(thrust::get<1>(t), thrust::get<2>(t)));
+      }
+    };
+
+    typedef thrust::counting_iterator<IndexType> counting_iterator;
+
+  public:
+    typedef thrust::transform_iterator<
+      head_flag_functor,
+      thrust::zip_iterator<thrust::tuple<counting_iterator,RandomAccessIterator,RandomAccessIterator> >
+    > iterator;
+
+    __host__ __device__
+    head_flags(RandomAccessIterator first, RandomAccessIterator last)
+      : m_begin(thrust::make_transform_iterator(thrust::make_zip_iterator(thrust::make_tuple(thrust::counting_iterator<IndexType>(0), first, first - 1)),
+                                                head_flag_functor(last - first))),
+        m_end(m_begin + (last - first))
+    {}
+
+    __host__ __device__
+    head_flags(RandomAccessIterator first, RandomAccessIterator last, BinaryPredicate binary_pred)
+      : m_begin(thrust::make_transform_iterator(thrust::make_zip_iterator(thrust::make_tuple(thrust::counting_iterator<IndexType>(0), first, first - 1)),
+                                                head_flag_functor(last - first, binary_pred))),
+        m_end(m_begin + (last - first))
+    {}
+
+    __host__ __device__
+    iterator begin() const
+    {
+      return m_begin;
+    }
+
+    __host__ __device__
+    iterator end() const
+    {
+      return m_end;
+    }
+
+    template<typename OtherIndex>
+    __host__ __device__
+    typename iterator::reference operator[](OtherIndex i)
+    {
+      return *(begin() + i);
+    }
+
+  private:
+    iterator m_begin, m_end;
+};
+
+
+template<typename RandomAccessIterator, typename BinaryPredicate>
+__host__ __device__
+head_flags<RandomAccessIterator, BinaryPredicate>
+  make_head_flags(RandomAccessIterator first, RandomAccessIterator last, BinaryPredicate binary_pred)
+{
+  return head_flags<RandomAccessIterator, BinaryPredicate>(first, last, binary_pred);
+}
+
+
+template<typename RandomAccessIterator>
+__host__ __device__
+head_flags<RandomAccessIterator>
+  make_head_flags(RandomAccessIterator first, RandomAccessIterator last)
+{
+  return head_flags<RandomAccessIterator>(first, last);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/range/tail_flags.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/range/tail_flags.h
new file mode 100644
index 000000000000..41ee5dd29cc8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/range/tail_flags.h
@@ -0,0 +1,133 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/tuple.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename RandomAccessIterator,
+         typename BinaryPredicate = thrust::equal_to<typename thrust::iterator_value<RandomAccessIterator>::type>,
+         typename ValueType = bool,
+         typename IndexType = typename thrust::iterator_difference<RandomAccessIterator>::type>
+  class tail_flags
+{
+  // XXX WAR cudafe bug
+  //private:
+  public:
+    struct tail_flag_functor
+    {
+      BinaryPredicate binary_pred; // this must be the first member for performance reasons
+      RandomAccessIterator iter;
+      IndexType n;
+
+      typedef ValueType result_type;
+
+      __host__ __device__
+      tail_flag_functor(RandomAccessIterator first, RandomAccessIterator last)
+        : binary_pred(), iter(first), n(last - first)
+      {}
+
+      __host__ __device__
+      tail_flag_functor(RandomAccessIterator first, RandomAccessIterator last, BinaryPredicate binary_pred)
+        : binary_pred(binary_pred), iter(first), n(last - first)
+      {}
+
+      __host__ __device__ __thrust_forceinline__
+      result_type operator()(const IndexType &i)
+      {
+        return (i == (n - 1) || !binary_pred(iter[i], iter[i+1]));
+      }
+    };
+
+    typedef thrust::counting_iterator<IndexType> counting_iterator;
+
+  public:
+    typedef thrust::transform_iterator<
+      tail_flag_functor,
+      counting_iterator
+    > iterator;
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    tail_flags(RandomAccessIterator first, RandomAccessIterator last)
+      : m_begin(thrust::make_transform_iterator(thrust::counting_iterator<IndexType>(0),
+                                                tail_flag_functor(first, last))),
+        m_end(m_begin + (last - first))
+    {}
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    tail_flags(RandomAccessIterator first, RandomAccessIterator last, BinaryPredicate binary_pred)
+      : m_begin(thrust::make_transform_iterator(thrust::counting_iterator<IndexType>(0),
+                                                tail_flag_functor(first, last, binary_pred))),
+        m_end(m_begin + (last - first))
+    {}
+
+    __host__ __device__
+    iterator begin() const
+    {
+      return m_begin;
+    }
+
+    __host__ __device__
+    iterator end() const
+    {
+      return m_end;
+    }
+
+    template<typename OtherIndex>
+    __host__ __device__
+    typename iterator::reference operator[](OtherIndex i)
+    {
+      return *(begin() + i);
+    }
+
+  private:
+    iterator m_begin, m_end;
+};
+
+
+template<typename RandomAccessIterator, typename BinaryPredicate>
+__host__ __device__
+tail_flags<RandomAccessIterator, BinaryPredicate>
+  make_tail_flags(RandomAccessIterator first, RandomAccessIterator last, BinaryPredicate binary_pred)
+{
+  return tail_flags<RandomAccessIterator, BinaryPredicate>(first, last, binary_pred);
+}
+
+
+template<typename RandomAccessIterator>
+__host__ __device__
+tail_flags<RandomAccessIterator>
+  make_tail_flags(RandomAccessIterator first, RandomAccessIterator last)
+{
+  return tail_flags<RandomAccessIterator>(first, last);
+}
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_pointer_cast.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_pointer_cast.h
new file mode 100644
index 000000000000..53a77861ef0e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_pointer_cast.h
@@ -0,0 +1,50 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename Pointer>
+__host__ __device__
+typename thrust::detail::pointer_traits<Pointer>::raw_pointer
+raw_pointer_cast(Pointer ptr)
+{
+  return thrust::detail::pointer_traits<Pointer>::get(ptr);
+}
+
+template <typename ToPointer, typename FromPointer>
+__host__ __device__
+ToPointer
+reinterpret_pointer_cast(FromPointer ptr)
+{
+  typedef typename thrust::detail::pointer_element<ToPointer>::type to_element;
+  return ToPointer(reinterpret_cast<to_element*>(thrust::raw_pointer_cast(ptr)));
+}
+
+template <typename ToPointer, typename FromPointer>
+__host__ __device__
+ToPointer
+static_pointer_cast(FromPointer ptr)
+{
+  typedef typename thrust::detail::pointer_element<ToPointer>::type to_element;
+  return ToPointer(static_cast<to_element*>(thrust::raw_pointer_cast(ptr)));
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_reference_cast.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_reference_cast.h
new file mode 100644
index 000000000000..eff45f0c273b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/raw_reference_cast.h
@@ -0,0 +1,332 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/type_traits/has_nested_type.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/tuple_transform.h>
+#include <thrust/iterator/detail/tuple_of_iterator_references.h>
+
+
+// the order of declarations and definitions in this file is totally goofy
+// this header defines raw_reference_cast, which has a few overloads towards the bottom of the file
+// raw_reference_cast depends on metafunctions such as is_unwrappable and raw_reference
+// we need to be sure that these metafunctions are completely defined (including specializations) before they are instantiated by raw_reference_cast
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(is_wrapped_reference, wrapped_reference_hint)
+
+
+// wrapped reference-like things which aren't strictly wrapped references
+// (e.g. tuples of wrapped references) are considered unwrappable
+template<typename T>
+  struct is_unwrappable
+    : is_wrapped_reference<T>
+{};
+
+
+// specialize is_unwrappable
+// a tuple is_unwrappable if any of its elements is_unwrappable
+template<typename... Ts>
+  struct is_unwrappable<
+    thrust::tuple<Ts...>
+  >
+    : or_<
+        is_unwrappable<Ts>...
+      >
+{};
+
+
+// specialize is_unwrappable
+// a tuple_of_iterator_references is_unwrappable if any of its elements is_unwrappable
+template<
+  typename... Ts
+>
+  struct is_unwrappable<
+    thrust::detail::tuple_of_iterator_references<Ts...>
+  >
+    : or_<
+        is_unwrappable<Ts>...
+      >
+{};
+
+
+template<typename T, typename Result = void>
+  struct enable_if_unwrappable
+    : enable_if<
+        is_unwrappable<T>::value,
+        Result
+      >
+{};
+
+
+namespace raw_reference_detail
+{
+
+
+template<typename T, typename Enable = void>
+  struct raw_reference_impl
+    : add_reference<T>
+{};
+
+
+template<typename T>
+  struct raw_reference_impl<
+    T,
+    typename thrust::detail::enable_if<
+      is_wrapped_reference<
+        typename remove_cv<T>::type
+      >::value
+    >::type
+  >
+{
+  typedef typename add_reference<
+    typename pointer_element<typename T::pointer>::type
+  >::type type;
+};
+
+
+} // end raw_reference_detail
+
+
+template<typename T>
+  struct raw_reference :
+    raw_reference_detail::raw_reference_impl<T>
+{};
+
+
+namespace raw_reference_detail
+{
+
+// unlike raw_reference,
+// raw_reference_tuple_helper needs to return a value
+// when it encounters one, rather than a reference
+// upon encountering tuple, recurse
+//
+// we want the following behavior:
+//  1. T                                -> T
+//  2. T&                               -> T&
+//  3. null_type                        -> null_type
+//  4. reference<T>                     -> T&
+//  5. tuple_of_iterator_references<T>  -> tuple_of_iterator_references<raw_reference_tuple_helper<T>::type>
+
+
+// wrapped references are unwrapped using raw_reference, otherwise, return T
+template<typename T>
+  struct raw_reference_tuple_helper
+    : eval_if<
+        is_unwrappable<
+          typename remove_cv<T>::type
+        >::value,
+        raw_reference<T>,
+        identity_<T>
+      >
+{};
+
+
+// recurse on tuples
+template <
+  typename... Ts
+>
+  struct raw_reference_tuple_helper<
+    thrust::tuple<Ts...>
+  >
+{
+  typedef thrust::tuple<
+    typename raw_reference_tuple_helper<Ts>::type...
+  > type;
+};
+
+
+template <
+  typename... Ts
+>
+  struct raw_reference_tuple_helper<
+    thrust::detail::tuple_of_iterator_references<Ts...>
+  >
+{
+  typedef thrust::detail::tuple_of_iterator_references<
+    typename raw_reference_tuple_helper<Ts>::type...
+  > type;
+};
+
+
+} // end raw_reference_detail
+
+
+// a couple of specializations of raw_reference for tuples follow
+
+
+// if a tuple "tuple_type" is_unwrappable,
+//   then the raw_reference of tuple_type is a tuple of its members' raw_references
+//   else the raw_reference of tuple_type is tuple_type &
+template <
+  typename... Ts
+>
+  struct raw_reference<
+    thrust::tuple<Ts...>
+  >
+{
+  private:
+    typedef thrust::tuple<Ts...> tuple_type;
+
+  public:
+    typedef typename eval_if<
+      is_unwrappable<tuple_type>::value,
+      raw_reference_detail::raw_reference_tuple_helper<tuple_type>,
+      add_reference<tuple_type>
+    >::type type;
+};
+
+
+template <
+  typename... Ts
+>
+  struct raw_reference<
+    thrust::detail::tuple_of_iterator_references<Ts...>
+  >
+{
+  private:
+    typedef detail::tuple_of_iterator_references<Ts...> tuple_type;
+
+  public:
+    typedef typename raw_reference_detail::raw_reference_tuple_helper<tuple_type>::type type;
+
+    // XXX figure out why is_unwrappable seems to be broken for tuple_of_iterator_references
+    //typedef typename eval_if<
+    //  is_unwrappable<tuple_type>::value,
+    //  raw_reference_detail::raw_reference_tuple_helper<tuple_type>,
+    //  add_reference<tuple_type>
+    //>::type type;
+};
+
+
+} // end detail
+
+
+// provide declarations of raw_reference_cast's overloads for raw_reference_caster below
+template<typename T>
+__host__ __device__
+typename detail::raw_reference<T>::type
+  raw_reference_cast(T &ref);
+
+
+template<typename T>
+__host__ __device__
+typename detail::raw_reference<const T>::type
+  raw_reference_cast(const T &ref);
+
+
+template<
+  typename... Ts
+>
+__host__ __device__
+typename detail::enable_if_unwrappable<
+  thrust::detail::tuple_of_iterator_references<Ts...>,
+  typename detail::raw_reference<
+    thrust::detail::tuple_of_iterator_references<Ts...>
+  >::type
+>::type
+raw_reference_cast(thrust::detail::tuple_of_iterator_references<Ts...> t);
+
+
+namespace detail
+{
+
+
+struct raw_reference_caster
+{
+  template<typename T>
+  __host__ __device__
+  typename detail::raw_reference<T>::type operator()(T &ref)
+  {
+    return thrust::raw_reference_cast(ref);
+  }
+
+  template<typename T>
+  __host__ __device__
+  typename detail::raw_reference<const T>::type operator()(const T &ref)
+  {
+    return thrust::raw_reference_cast(ref);
+  }
+
+  template<
+    typename... Ts
+  >
+  __host__ __device__
+  typename detail::raw_reference<
+    thrust::detail::tuple_of_iterator_references<Ts...>
+  >::type
+  operator()(thrust::detail::tuple_of_iterator_references<Ts...> t,
+             typename enable_if<
+               is_unwrappable<thrust::detail::tuple_of_iterator_references<Ts...> >::value
+             >::type * = 0)
+  {
+    return thrust::raw_reference_cast(t);
+  }
+}; // end raw_reference_caster
+
+
+} // end detail
+
+
+template<typename T>
+__host__ __device__
+typename detail::raw_reference<T>::type
+  raw_reference_cast(T &ref)
+{
+  return *thrust::raw_pointer_cast(&ref);
+} // end raw_reference_cast
+
+
+template<typename T>
+__host__ __device__
+typename detail::raw_reference<const T>::type
+  raw_reference_cast(const T &ref)
+{
+  return *thrust::raw_pointer_cast(&ref);
+} // end raw_reference_cast
+
+
+template<
+  typename... Ts
+>
+__host__ __device__
+typename detail::enable_if_unwrappable<
+  thrust::detail::tuple_of_iterator_references<Ts...>,
+  typename detail::raw_reference<
+    thrust::detail::tuple_of_iterator_references<Ts...>
+  >::type
+>::type
+raw_reference_cast(thrust::detail::tuple_of_iterator_references<Ts...> t)
+{
+  thrust::detail::raw_reference_caster f;
+
+  // note that we pass raw_reference_tuple_helper, not raw_reference as the unary metafunction
+  // the different way that raw_reference_tuple_helper unwraps tuples is important
+  return thrust::detail::tuple_host_device_transform<detail::raw_reference_detail::raw_reference_tuple_helper>(t, f);
+} // end raw_reference_cast
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/reduce.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reduce.inl
new file mode 100644
index 000000000000..448a4b38c4a6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reduce.inl
@@ -0,0 +1,276 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/reduce.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/reduce.h>
+#include <thrust/system/detail/generic/reduce_by_key.h>
+#include <thrust/system/detail/adl/reduce.h>
+#include <thrust/system/detail/adl/reduce_by_key.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::value_type
+    reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last)
+{
+  using thrust::system::detail::generic::reduce;
+  return reduce(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end reduce()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+  T reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+           InputIterator first,
+           InputIterator last,
+           T init)
+{
+  using thrust::system::detail::generic::reduce;
+  return reduce(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, init);
+} // end reduce()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename T,
+         typename BinaryFunction>
+__host__ __device__
+  T reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+           InputIterator first,
+           InputIterator last,
+           T init,
+           BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::reduce;
+  return reduce(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, init, binary_op);
+} // end reduce()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                InputIterator1 keys_first,
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output)
+{
+  using thrust::system::detail::generic::reduce_by_key;
+  return reduce_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, keys_output, values_output);
+} // end reduce_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                InputIterator1 keys_first,
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::reduce_by_key;
+  return reduce_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, keys_output, values_output, binary_pred);
+} // end reduce_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                InputIterator1 keys_first,
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred,
+                BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::reduce_by_key;
+  return reduce_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, keys_output, values_output, binary_pred, binary_op);
+} // end reduce_by_key()
+
+
+template<typename InputIterator>
+typename thrust::iterator_traits<InputIterator>::value_type
+  reduce(InputIterator first,
+         InputIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::reduce(select_system(system), first, last);
+}
+
+
+template<typename InputIterator,
+         typename T>
+   T reduce(InputIterator first,
+            InputIterator last,
+            T init)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::reduce(select_system(system), first, last, init);
+}
+
+
+template<typename InputIterator,
+         typename T,
+         typename BinaryFunction>
+   T reduce(InputIterator first,
+            InputIterator last,
+            T init,
+            BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::reduce(select_system(system), first, last, init, binary_op);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(InputIterator1 keys_first,
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::reduce_by_key(select_system(system1,system2,system3,system4), keys_first, keys_last, values_first, keys_output, values_output);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(InputIterator1 keys_first,
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::reduce_by_key(select_system(system1,system2,system3,system4), keys_first, keys_last, values_first, keys_output, values_output, binary_pred);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(InputIterator1 keys_first,
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred,
+                BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::reduce_by_key(select_system(system1,system2,system3,system4), keys_first, keys_last, values_first, keys_output, values_output, binary_pred, binary_op);
+}
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/reference.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reference.h
new file mode 100644
index 000000000000..5cc13625dafe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reference.h
@@ -0,0 +1,518 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A pointer to a variable which resides in memory associated with a
+ *  system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/reference_forward_declaration.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/memory.h>
+#include <thrust/system/detail/adl/get_value.h>
+#include <thrust/system/detail/adl/assign_value.h>
+#include <thrust/system/detail/adl/iter_swap.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <type_traits>
+#include <ostream>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+template <typename>
+struct is_wrapped_reference;
+}
+
+/*! \p reference acts as a reference-like wrapper for an object residing in
+ *  memory that a \p pointer refers to.
+ */
+template <typename Element, typename Pointer, typename Derived>
+class reference
+{
+private:
+  using derived_type = typename std::conditional<
+    std::is_same<Derived, use_default>::value, reference, Derived
+  >::type;
+
+public:
+  using pointer    = Pointer;
+  using value_type = typename thrust::remove_cvref<Element>::type;
+
+  reference(reference const&) = default;
+
+  reference(reference&&) = default;
+
+  /*! Construct a \p reference from another \p reference whose pointer type is
+   *  convertible to \p pointer. After this \p reference is constructed, it
+   *  shall refer to the same object as \p other.
+   *
+   *  \tparam OtherElement The element type of the other \p reference.
+   *  \tparam OtherPointer The pointer type of the other \p reference.
+   *  \tparam OtherDerived The derived type of the other \p reference.
+   *  \param  other        A \p reference to copy from.
+   */
+  template <typename OtherElement, typename OtherPointer, typename OtherDerived>
+  __host__ __device__
+  reference(
+    reference<OtherElement, OtherPointer, OtherDerived> const& other
+  /*! \cond
+   */
+  , typename std::enable_if<
+      std::is_convertible<
+        typename reference<OtherElement, OtherPointer, OtherDerived>::pointer
+      , pointer
+      >::value
+    >::type* = nullptr
+  /*! \endcond
+   */
+  )
+    : ptr(other.ptr)
+  {}
+
+  /*! Construct a \p reference that refers to an object pointed to by the given
+   *  \p pointer. After this \p reference is constructed, it shall refer to the
+   *  object pointed to by \p ptr.
+   *
+   *  \param ptr A \p pointer to construct from.
+   */
+  __host__ __device__
+  explicit reference(pointer const& p) : ptr(p) {}
+
+  /*! Assign the object referred to \p other to the object referred to by
+   *  this \p reference.
+   *
+   *  \param other The other \p reference to assign from.
+   *
+   *  \return <tt>*this</tt>.
+   */
+  __host__ __device__
+  derived_type& operator=(reference const& other)
+  {
+    assign_from(&other);
+    return derived();
+  }
+
+  /*! Assign the object referred to by this \p reference with the object
+   *  referred to by another \p reference whose pointer type is convertible to
+   *  \p pointer.
+   *
+   *  \tparam OtherElement The element type of the other \p reference.
+   *  \tparam OtherPointer The pointer type of the other \p reference.
+   *  \tparam OtherDerived The derived type of the other \p reference.
+   *  \param  other        The other \p reference to assign from.
+   *
+   *  \return <tt>*this</tt>.
+   */
+  template <typename OtherElement, typename OtherPointer, typename OtherDerived>
+  __host__ __device__
+  /*! \cond
+   */
+  typename std::enable_if<
+    std::is_convertible<
+      typename reference<OtherElement, OtherPointer, OtherDerived>::pointer
+    , pointer
+    >::value,
+  /*! \endcond
+   */
+    derived_type&
+  /*! \cond
+   */
+  >::type
+  /*! \endcond
+   */
+  operator=(reference<OtherElement, OtherPointer, OtherDerived> const& other)
+  {
+    assign_from(&other);
+    return derived();
+  }
+
+  /*! Assign \p rhs to the object referred to by this \p tagged_reference.
+   *
+   *  \param rhs The \p value_type to assign from.
+   *
+   *  \return <tt>*this</tt>.
+   */
+  __host__ __device__
+  derived_type& operator=(value_type const& rhs)
+  {
+    assign_from(&rhs);
+    return derived();
+  }
+
+  /*! Exchanges the value of the object referred to by this \p tagged_reference
+   *  with the object referred to by \p other.
+   *
+   *  \param other The \p tagged_reference to swap with.
+   */
+  __host__ __device__
+  void swap(derived_type& other)
+  {
+    // Avoid default-constructing a system; instead, just use a null pointer
+    // for dispatch. This assumes that `get_value` will not access any system
+    // state.
+    typename thrust::iterator_system<pointer>::type* system = nullptr;
+    swap(system, other);
+  }
+
+  __host__ __device__ pointer operator&() const { return ptr; }
+
+  // This is inherently hazardous, as it discards the strong type information
+  // about what system the object is on.
+  __host__ __device__ operator value_type() const
+  {
+    // Avoid default-constructing a system; instead, just use a null pointer
+    // for dispatch. This assumes that `get_value` will not access any system
+    // state.
+    typename thrust::iterator_system<pointer>::type* system = nullptr;
+    return convert_to_value_type(system);
+  }
+
+  __host__ __device__
+  derived_type& operator++()
+  {
+    // Sadly, this has to make a copy. The only mechanism we have for
+    // modifying the value, which may be in memory inaccessible to this
+    // system, is to get a copy of it, modify the copy, and then update it.
+    value_type tmp = *this;
+    ++tmp;
+    *this = tmp;
+    return derived();
+  }
+
+  __host__ __device__
+  value_type operator++(int)
+  {
+    value_type tmp = *this;
+    value_type result = tmp++;
+    *this = std::move(tmp);
+    return result;
+  }
+
+  derived_type& operator--()
+  {
+    // Sadly, this has to make a copy. The only mechanism we have for
+    // modifying the value, which may be in memory inaccessible to this
+    // system, is to get a copy of it, modify the copy, and then update it.
+    value_type tmp = *this;
+    --tmp;
+    *this = std::move(tmp);
+    return derived();
+  }
+
+  value_type operator--(int)
+  {
+    value_type tmp = *this;
+    value_type result = tmp--;
+    *this = std::move(tmp);
+    return derived();
+  }
+
+  __host__ __device__
+  derived_type& operator+=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp += rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator-=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp -= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator*=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp *= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator/=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp /= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator%=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp %= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator<<=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp <<= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator>>=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp >>= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator&=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp &= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator|=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp |= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+  derived_type& operator^=(value_type const& rhs)
+  {
+    value_type tmp = *this;
+    tmp ^= rhs;
+    *this = tmp;
+    return derived();
+  }
+
+private:
+  pointer const ptr;
+
+  // `thrust::detail::is_wrapped_reference` is a trait that indicates whether
+  // a type is a fancy reference. It detects such types by loooking for a
+  // nested `wrapped_reference_hint` type.
+  struct wrapped_reference_hint {};
+  template <typename>
+  friend struct thrust::detail::is_wrapped_reference;
+
+  template <typename OtherElement, typename OtherPointer, typename OtherDerived>
+  friend class reference;
+
+  __host__ __device__
+  derived_type& derived() { return static_cast<derived_type&>(*this); }
+
+  template<typename System>
+  __host__ __device__
+  value_type convert_to_value_type(System* system) const
+  {
+    using thrust::system::detail::generic::select_system;
+    return strip_const_get_value(select_system(*system));
+  }
+
+  template <typename System>
+  __host__ __device__
+  value_type strip_const_get_value(System const& system) const
+  {
+    System &non_const_system = const_cast<System&>(system);
+
+    using thrust::system::detail::generic::get_value;
+    return get_value(thrust::detail::derived_cast(non_const_system), ptr);
+  }
+
+  template <typename System0, typename System1, typename OtherPointer>
+  __host__ __device__
+  void assign_from(System0* system0, System1* system1, OtherPointer src)
+  {
+    using thrust::system::detail::generic::select_system;
+    strip_const_assign_value(select_system(*system0, *system1), src);
+  }
+
+  template <typename OtherPointer>
+  __host__ __device__
+  void assign_from(OtherPointer src)
+  {
+    // Avoid default-constructing systems; instead, just use a null pointer
+    // for dispatch. This assumes that `get_value` will not access any system
+    // state.
+    typename thrust::iterator_system<pointer>::type*      system0 = nullptr;
+    typename thrust::iterator_system<OtherPointer>::type* system1 = nullptr;
+    assign_from(system0, system1, src);
+  }
+
+  template <typename System, typename OtherPointer>
+  __host__ __device__
+  void strip_const_assign_value(System const& system, OtherPointer src)
+  {
+    System& non_const_system = const_cast<System&>(system);
+
+    using thrust::system::detail::generic::assign_value;
+    assign_value(thrust::detail::derived_cast(non_const_system), ptr, src);
+  }
+
+  template <typename System>
+  __host__ __device__
+  void swap(System* system, derived_type& other)
+  {
+    using thrust::system::detail::generic::select_system;
+    using thrust::system::detail::generic::iter_swap;
+
+    iter_swap(select_system(*system, *system), ptr, other.ptr);
+  }
+};
+
+template <typename Pointer, typename Derived>
+class reference<void, Pointer, Derived> {};
+
+template <typename Pointer, typename Derived>
+class reference<void const, Pointer, Derived> {};
+
+template <
+  typename Element, typename Pointer, typename Derived
+, typename CharT, typename Traits
+>
+std::basic_ostream<CharT, Traits>& operator<<(
+  std::basic_ostream<CharT, Traits>&os
+, reference<Element, Pointer, Derived> const& r
+) {
+  using value_type = typename reference<Element, Pointer, Derived>::value_type;
+  return os << static_cast<value_type>(r);
+}
+
+template <typename Element, typename Tag>
+class tagged_reference;
+
+/*! \p tagged_reference acts as a reference-like wrapper for an object residing
+ *  in memory associated with system \p Tag that a \p pointer refers to.
+ */
+template <typename Element, typename Tag>
+class tagged_reference
+  : public thrust::reference<
+      Element
+    , thrust::pointer<Element, Tag, tagged_reference<Element, Tag>>
+    , tagged_reference<Element, Tag>
+    >
+{
+private:
+  using base_type = thrust::reference<
+    Element
+  , thrust::pointer<Element, Tag, tagged_reference<Element, Tag>>
+  , tagged_reference<Element, Tag>
+  >;
+
+public:
+  using value_type = typename base_type::value_type;
+  using pointer    = typename base_type::pointer;
+
+  tagged_reference(tagged_reference const&) = default;
+
+  tagged_reference(tagged_reference&&) = default;
+
+  /*! Construct a \p tagged_reference from another \p tagged_reference whose
+   *  pointer type is convertible to \p pointer. After this \p tagged_reference
+   *  is constructed, it shall refer to the same object as \p other.
+   *
+   *  \tparam OtherElement The element type of the other \p tagged_reference.
+   *  \tparam OtherTag     The tag type of the other \p tagged_reference.
+   *  \param  other        A \p tagged_reference to copy from.
+   */
+  template <typename OtherElement, typename OtherTag>
+  __host__ __device__
+  tagged_reference(tagged_reference<OtherElement, OtherTag> const& other)
+    : base_type(other)
+  {}
+
+  /*! Construct a \p tagged_reference that refers to an object pointed to by
+   *  the given \p pointer. After this \p tagged_reference is constructed, it
+   *  shall refer to the object pointed to by \p ptr.
+   *
+   *  \param ptr A \p pointer to construct from.
+   */
+  __host__ __device__ explicit tagged_reference(pointer const& p)
+    : base_type(p)
+  {}
+
+  /*! Assign the object referred to \p other to the object referred to by
+   *  this \p tagged_reference.
+   *
+   *  \param other The other \p tagged_reference to assign from.
+   *
+   *  \return <tt>*this</tt>.
+   */
+  __host__ __device__
+  tagged_reference& operator=(tagged_reference const& other)
+  {
+    return base_type::operator=(other);
+  }
+
+  /*! Assign the object referred to by this \p tagged_reference with the object
+   *  referred to by another \p tagged_reference whose pointer type is
+   *  convertible to \p pointer.
+   *
+   *  \tparam OtherElement The element type of the other \p tagged_reference.
+   *  \tparam OtherTag     The tag type of the other \p tagged_reference.
+   *  \param  other        The other \p tagged_reference to assign from.
+   *
+   *  \return <tt>*this</tt>.
+   */
+  template <typename OtherElement, typename OtherTag>
+  __host__ __device__
+  tagged_reference&
+  operator=(tagged_reference<OtherElement, OtherTag> const& other)
+  {
+    return base_type::operator=(other);
+  }
+
+  /*! Assign \p rhs to the object referred to by this \p tagged_reference.
+   *
+   *  \param rhs The \p value_type to assign from.
+   *
+   *  \return <tt>*this</tt>.
+   */
+  __host__ __device__
+  tagged_reference& operator=(value_type const& rhs)
+  {
+    return base_type::operator=(rhs);
+  }
+};
+
+template <typename Tag>
+class tagged_reference<void, Tag> {};
+
+template <typename Tag>
+class tagged_reference<void const, Tag> {};
+
+/*! Exchanges the values of two objects referred to by \p tagged_reference.
+ *
+ *  \param x The first \p tagged_reference of interest.
+ *  \param y The second \p tagged_reference of interest.
+ */
+template <typename Element, typename Tag>
+__host__ __device__
+void swap(tagged_reference<Element, Tag>& x, tagged_reference<Element, Tag>& y)
+{
+  x.swap(y);
+}
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/reference_forward_declaration.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reference_forward_declaration.h
new file mode 100644
index 000000000000..6f2b999497ab
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reference_forward_declaration.h
@@ -0,0 +1,28 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/use_default.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename Element, typename Pointer, typename Derived = use_default>
+class reference;
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/remove.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/remove.inl
new file mode 100644
index 000000000000..7ccc0cc4656a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/remove.inl
@@ -0,0 +1,246 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/remove.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/remove.h>
+#include <thrust/system/detail/adl/remove.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  ForwardIterator remove(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last,
+                         const T &value)
+{
+  using thrust::system::detail::generic::remove;
+  return remove(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, value);
+} // end remove()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator remove_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator result,
+                             const T &value)
+{
+  using thrust::system::detail::generic::remove_copy;
+  return remove_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, value);
+} // end remove_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::remove_if;
+  return remove_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred);
+} // end remove_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  using thrust::system::detail::generic::remove_copy_if;
+  return remove_copy_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, pred);
+} // end remove_copy_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::remove_if;
+  return remove_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, pred);
+} // end remove_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  using thrust::system::detail::generic::remove_copy_if;
+  return remove_copy_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, result, pred);
+} // end remove_copy_if()
+
+
+template<typename ForwardIterator,
+         typename T>
+  ForwardIterator remove(ForwardIterator first,
+                         ForwardIterator last,
+                         const T &value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::remove(select_system(system), first, last, value);
+} // end remove()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename T>
+  OutputIterator remove_copy(InputIterator first,
+                             InputIterator last,
+                             OutputIterator result,
+                             const T &value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::remove_copy(select_system(system1,system2), first, last, result, value);
+} // end remove_copy()
+
+
+template<typename ForwardIterator,
+         typename Predicate>
+  ForwardIterator remove_if(ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::remove_if(select_system(system), first, last, pred);
+} // end remove_if()
+
+
+template<typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+  ForwardIterator remove_if(ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+  typedef typename thrust::iterator_system<InputIterator>::type   System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::remove_if(select_system(system1,system2), first, last, stencil, pred);
+} // end remove_if()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator remove_copy_if(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::remove_copy_if(select_system(system1,system2), first, last, result, pred);
+} // end remove_copy_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator remove_copy_if(InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::remove_copy_if(select_system(system1,system2,system3), first, last, stencil, result, pred);
+} // end remove_copy_if()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/replace.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/replace.inl
new file mode 100644
index 000000000000..629287bee083
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/replace.inl
@@ -0,0 +1,218 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/replace.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/replace.h>
+#include <thrust/system/detail/adl/replace.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void replace(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+               ForwardIterator first, ForwardIterator last,
+               const T &old_value,
+               const T &new_value)
+{
+  using thrust::system::detail::generic::replace;
+  return replace(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, old_value, new_value);
+} // end replace()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  ForwardIterator first, ForwardIterator last,
+                  Predicate pred,
+                  const T &new_value)
+{
+  using thrust::system::detail::generic::replace_if;
+  return replace_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, pred, new_value);
+} // end replace_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  ForwardIterator first, ForwardIterator last,
+                  InputIterator stencil,
+                  Predicate pred,
+                  const T &new_value)
+{
+  using thrust::system::detail::generic::replace_if;
+  return replace_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, pred, new_value);
+} // end replace_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename T>
+__host__ __device__
+  OutputIterator replace_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              InputIterator first, InputIterator last,
+                              OutputIterator result,
+                              const T &old_value,
+                              const T &new_value)
+{
+  using thrust::system::detail::generic::replace_copy;
+  return replace_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, old_value, new_value);
+} // end replace_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                 InputIterator first, InputIterator last,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value)
+{
+  using thrust::system::detail::generic::replace_copy_if;
+  return replace_copy_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, pred, new_value);
+} // end replace_copy_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                 InputIterator1 first, InputIterator1 last,
+                                 InputIterator2 stencil,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value)
+{
+  using thrust::system::detail::generic::replace_copy_if;
+  return replace_copy_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, result, pred, new_value);
+} // end replace_copy_if()
+
+
+template<typename InputIterator, typename OutputIterator, typename Predicate, typename T>
+  OutputIterator replace_copy_if(InputIterator first, InputIterator last,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::replace_copy_if(select_system(system1,system2), first, last, result, pred, new_value);
+} // end replace_copy_if()
+
+
+template<typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate, typename T>
+  OutputIterator replace_copy_if(InputIterator1 first, InputIterator1 last,
+                                 InputIterator2 stencil,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::replace_copy_if(select_system(system1,system2,system3), first, last, stencil, result, pred, new_value);
+} // end replace_copy_if()
+
+
+template<typename InputIterator, typename OutputIterator, typename T>
+  OutputIterator replace_copy(InputIterator first, InputIterator last,
+                              OutputIterator result,
+                              const T &old_value,
+                              const T &new_value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::replace_copy(select_system(system1,system2), first, last, result, old_value, new_value);
+} // end replace_copy()
+
+
+template<typename ForwardIterator, typename Predicate, typename T>
+  void replace_if(ForwardIterator first, ForwardIterator last,
+                  Predicate pred,
+                  const T &new_value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::replace_if(select_system(system), first, last, pred, new_value);
+} // end replace_if()
+
+
+template<typename ForwardIterator, typename InputIterator, typename Predicate, typename T>
+  void replace_if(ForwardIterator first, ForwardIterator last,
+                  InputIterator stencil,
+                  Predicate pred,
+                  const T &new_value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System1;
+  typedef typename thrust::iterator_system<InputIterator>::type   System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::replace_if(select_system(system1,system2), first, last, stencil, pred, new_value);
+} // end replace_if()
+
+
+template<typename ForwardIterator, typename T>
+  void replace(ForwardIterator first, ForwardIterator last,
+               const T &old_value,
+               const T &new_value)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::replace(select_system(system), first, last, old_value, new_value);
+} // end replace()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/reverse.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reverse.inl
new file mode 100644
index 000000000000..dc316d18f52f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/reverse.inl
@@ -0,0 +1,87 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/reverse.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/reverse.h>
+#include <thrust/system/detail/adl/reverse.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename BidirectionalIterator>
+__host__ __device__
+  void reverse(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+               BidirectionalIterator first,
+               BidirectionalIterator last)
+{
+  using thrust::system::detail::generic::reverse;
+  return reverse(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end reverse()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename BidirectionalIterator, typename OutputIterator>
+__host__ __device__
+  OutputIterator reverse_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              BidirectionalIterator first,
+                              BidirectionalIterator last,
+                              OutputIterator result)
+{
+  using thrust::system::detail::generic::reverse_copy;
+  return reverse_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result);
+} // end reverse_copy()
+
+
+template<typename BidirectionalIterator>
+  void reverse(BidirectionalIterator first,
+               BidirectionalIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<BidirectionalIterator>::type System;
+
+  System system;
+
+  return thrust::reverse(select_system(system), first, last);
+} // end reverse()
+
+
+template<typename BidirectionalIterator,
+         typename OutputIterator>
+  OutputIterator reverse_copy(BidirectionalIterator first,
+                              BidirectionalIterator last,
+                              OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<BidirectionalIterator>::type System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type        System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::reverse_copy(select_system(system1,system2), first, last, result);
+} // end reverse_copy()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/scan.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/scan.inl
new file mode 100644
index 000000000000..b781b0e280c1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/scan.inl
@@ -0,0 +1,522 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/scan.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/scan.h>
+#include <thrust/system/detail/generic/scan_by_key.h>
+#include <thrust/system/detail/adl/scan.h>
+#include <thrust/system/detail/adl/scan_by_key.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result)
+{
+  using thrust::system::detail::generic::inclusive_scan;
+  return inclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result);
+} // end inclusive_scan()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator inclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::inclusive_scan;
+  return inclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, binary_op);
+} // end inclusive_scan()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result)
+{
+  using thrust::system::detail::generic::exclusive_scan;
+  return exclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result);
+} // end exclusive_scan()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init)
+{
+  using thrust::system::detail::generic::exclusive_scan;
+  return exclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, init);
+} // end exclusive_scan()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init,
+                                AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::exclusive_scan;
+  return exclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, init, binary_op);
+} // end exclusive_scan()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result)
+{
+  using thrust::system::detail::generic::inclusive_scan_by_key;
+  return inclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result);
+} // end inclusive_scan_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::inclusive_scan_by_key;
+  return inclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result, binary_pred);
+} // end inclusive_scan_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::inclusive_scan_by_key;
+  return inclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result, binary_pred, binary_op);
+} // end inclusive_scan_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result)
+{
+  using thrust::system::detail::generic::exclusive_scan_by_key;
+  return exclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result);
+} // end exclusive_scan_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init)
+{
+  using thrust::system::detail::generic::exclusive_scan_by_key;
+  return exclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result, init);
+} // end exclusive_scan_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::exclusive_scan_by_key;
+  return exclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result, init, binary_pred);
+} // end exclusive_scan_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::exclusive_scan_by_key;
+  return exclusive_scan_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result, init, binary_pred, binary_op);
+} // end exclusive_scan_by_key()
+
+
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator inclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::inclusive_scan(select_system(system1,system2), first, last, result);
+} // end inclusive_scan()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename BinaryFunction>
+  OutputIterator inclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::inclusive_scan(select_system(system1,system2), first, last, result, binary_op);
+} // end inclusive_scan()
+
+
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator exclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::exclusive_scan(select_system(system1,system2), first, last, result);
+} // end exclusive_scan()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename T>
+  OutputIterator exclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::exclusive_scan(select_system(system1,system2), first, last, result, init);
+} // end exclusive_scan()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename T,
+         typename BinaryFunction>
+  OutputIterator exclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init,
+                                BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::exclusive_scan(select_system(system1,system2), first, last, result, init, binary_op);
+} // end exclusive_scan()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator inclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::inclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate>
+  OutputIterator inclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::inclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result, binary_pred);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+  OutputIterator inclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::inclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result, binary_pred, binary_op);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::exclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::exclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result, init);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::exclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result, init, binary_pred);
+}
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::exclusive_scan_by_key(select_system(system1,system2,system3), first1, last1, first2, result, init, binary_pred, binary_op);
+}
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/scatter.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/scatter.inl
new file mode 100644
index 000000000000..30dd611d11f0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/scatter.inl
@@ -0,0 +1,163 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/scatter.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/scatter.h>
+#include <thrust/system/detail/adl/scatter.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+               InputIterator1 first,
+               InputIterator1 last,
+               InputIterator2 map,
+               RandomAccessIterator output)
+{
+  using thrust::system::detail::generic::scatter;
+  return scatter(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, map, output);
+} // end scatter()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output)
+{
+  using thrust::system::detail::generic::scatter_if;
+  return scatter_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, map, stencil, output);
+} // end scatter_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator,
+         typename Predicate>
+__host__ __device__
+  void scatter_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output,
+                  Predicate pred)
+{
+  using thrust::system::detail::generic::scatter_if;
+  return scatter_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, map, stencil, output, pred);
+} // end scatter_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator>
+  void scatter(InputIterator1 first,
+               InputIterator1 last,
+               InputIterator2 map,
+               RandomAccessIterator output)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type       System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type       System2;
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::scatter(select_system(system1,system2,system3), first, last, map, output);
+} // end scatter()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator>
+  void scatter_if(InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type       System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type       System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type       System3;
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::scatter_if(select_system(system1,system2,system3,system4), first, last, map, stencil, output);
+} // end scatter_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator,
+         typename Predicate>
+  void scatter_if(InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output,
+                  Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type       System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type       System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type       System3;
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::scatter_if(select_system(system1,system2,system3,system4), first, last, map, stencil, output, pred);
+} // end scatter_if()
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/select_system.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/select_system.h
new file mode 100644
index 000000000000..968446162dd2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/select_system.h
@@ -0,0 +1,84 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/detail/type_deduction.h>
+#include <thrust/type_traits/remove_cvref.h>
+#include <thrust/system/detail/generic/select_system.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// We need a way to compute the return type of `select_system`, which is found
+// by using `thrust::system::detail::generic::select_system` and then making an
+// ADL call. We have no trait that defines the return type. With the
+// limitations of C++11 return type deduction, we need to be able to stick all
+// of that into `decltype`. So, we put the using statement into a detail
+// namespace, and then implement the generic dispatch function in that
+// namespace.
+
+namespace select_system_detail
+{
+
+using thrust::system::detail::generic::select_system;
+
+struct select_system_fn final
+{
+  __thrust_exec_check_disable__
+  template <typename DerivedPolicy0>
+  __host__ __device__
+  auto operator()(
+    thrust::detail::execution_policy_base<DerivedPolicy0> const& exec0
+  ) const
+  THRUST_DECLTYPE_RETURNS(
+    select_system(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec0))
+    )
+  )
+
+  __thrust_exec_check_disable__
+  template <typename DerivedPolicy0, typename DerivedPolicy1>
+  __host__ __device__
+  auto operator()(
+    thrust::detail::execution_policy_base<DerivedPolicy0> const& exec0
+  , thrust::detail::execution_policy_base<DerivedPolicy1> const& exec1
+  ) const
+  THRUST_DECLTYPE_RETURNS(
+    select_system(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec0))
+    , thrust::detail::derived_cast(thrust::detail::strip_const(exec1))
+    )
+  )
+};
+
+} // namespace select_system_detail
+
+THRUST_INLINE_CONSTANT select_system_detail::select_system_fn select_system{};
+
+} // detail
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/seq.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/seq.h
new file mode 100644
index 000000000000..ba18c2dbf515
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/seq.h
@@ -0,0 +1,52 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator_aware_execution_policy.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+struct seq_t : thrust::system::detail::sequential::execution_policy<seq_t>,
+  thrust::detail::allocator_aware_execution_policy<
+    thrust::system::detail::sequential::execution_policy>
+{
+  __host__ __device__
+  constexpr seq_t() : thrust::system::detail::sequential::execution_policy<seq_t>() {}
+
+  // allow any execution_policy to convert to seq_t
+  template<typename DerivedPolicy>
+  __host__ __device__
+  seq_t(const thrust::execution_policy<DerivedPolicy> &)
+    : thrust::system::detail::sequential::execution_policy<seq_t>()
+  {}
+};
+
+
+} // end detail
+
+
+THRUST_INLINE_CONSTANT detail::seq_t seq;
+
+
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/sequence.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/sequence.inl
new file mode 100644
index 000000000000..ffc9b968b97d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/sequence.inl
@@ -0,0 +1,114 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/sequence.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/sequence.h>
+#include <thrust/system/detail/adl/sequence.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  void sequence(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last)
+{
+  using thrust::system::detail::generic::sequence;
+  return sequence(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end sequence()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init)
+{
+  using thrust::system::detail::generic::sequence;
+  return sequence(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, init);
+} // end sequence()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init,
+                T step)
+{
+  using thrust::system::detail::generic::sequence;
+  return sequence(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, init, step);
+} // end sequence()
+
+
+template<typename ForwardIterator>
+  void sequence(ForwardIterator first,
+                ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::sequence(select_system(system), first, last);
+} // end sequence()
+
+
+template<typename ForwardIterator, typename T>
+  void sequence(ForwardIterator first,
+                ForwardIterator last,
+                T init)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::sequence(select_system(system), first, last, init);
+} // end sequence()
+
+
+template<typename ForwardIterator, typename T>
+  void sequence(ForwardIterator first,
+                ForwardIterator last,
+                T init,
+                T step)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::sequence(select_system(system), first, last, init, step);
+} // end sequence()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/set_operations.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/set_operations.inl
new file mode 100644
index 000000000000..7915f7b3eab6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/set_operations.inl
@@ -0,0 +1,865 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/set_operations.h>
+#include <thrust/system/detail/adl/set_operations.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              InputIterator1                                              first1,
+                              InputIterator1                                              last1,
+                              InputIterator2                                              first2,
+                              InputIterator2                                              last2,
+                              OutputIterator                                              result)
+{
+  using thrust::system::detail::generic::set_difference;
+  return set_difference(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result);
+} // end set_difference()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+OutputIterator set_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              InputIterator1                                              first1,
+                              InputIterator1                                              last1,
+                              InputIterator2                                              first2,
+                              InputIterator2                                              last2,
+                              OutputIterator                                              result,
+                              StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_difference;
+  return set_difference(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result, comp);
+} // end set_difference()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        InputIterator1                                              keys_first1,
+                        InputIterator1                                              keys_last1,
+                        InputIterator2                                              keys_first2,
+                        InputIterator2                                              keys_last2,
+                        InputIterator3                                              values_first1,
+                        InputIterator4                                              values_first2,
+                        OutputIterator1                                             keys_result,
+                        OutputIterator2                                             values_result)
+{
+  using thrust::system::detail::generic::set_difference_by_key;
+  return set_difference_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end set_difference_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        InputIterator1                                              keys_first1,
+                        InputIterator1                                              keys_last1,
+                        InputIterator2                                              keys_first2,
+                        InputIterator2                                              keys_last2,
+                        InputIterator3                                              values_first1,
+                        InputIterator4                                              values_first2,
+                        OutputIterator1                                             keys_result,
+                        OutputIterator2                                             values_result,
+                        StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_difference_by_key;
+  return set_difference_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end set_difference_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_intersection(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator1                                              first1,
+                                InputIterator1                                              last1,
+                                InputIterator2                                              first2,
+                                InputIterator2                                              last2,
+                                OutputIterator                                              result)
+{
+  using thrust::system::detail::generic::set_intersection;
+  return set_intersection(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result);
+} // end set_intersection()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+OutputIterator set_intersection(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator1                                              first1,
+                                InputIterator1                                              last1,
+                                InputIterator2                                              first2,
+                                InputIterator2                                              last2,
+                                OutputIterator                                              result,
+                                StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_intersection;
+  return set_intersection(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result, comp);
+} // end set_intersection()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_intersection_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator1                                              keys_first1,
+                          InputIterator1                                              keys_last1,
+                          InputIterator2                                              keys_first2,
+                          InputIterator2                                              keys_last2,
+                          InputIterator3                                              values_first1,
+                          OutputIterator1                                             keys_result,
+                          OutputIterator2                                             values_result)
+{
+  using thrust::system::detail::generic::set_intersection_by_key;
+  return set_intersection_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, keys_result, values_result);
+} // end set_intersection_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_intersection_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator1                                              keys_first1,
+                          InputIterator1                                              keys_last1,
+                          InputIterator2                                              keys_first2,
+                          InputIterator2                                              keys_last2,
+                          InputIterator3                                              values_first1,
+                          OutputIterator1                                             keys_result,
+                          OutputIterator2                                             values_result,
+                          StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_intersection_by_key;
+  return set_intersection_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, keys_result, values_result, comp);
+} // end set_intersection_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_symmetric_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                        InputIterator1                                              first1,
+                                        InputIterator1                                              last1,
+                                        InputIterator2                                              first2,
+                                        InputIterator2                                              last2,
+                                        OutputIterator                                              result)
+{
+  using thrust::system::detail::generic::set_symmetric_difference;
+  return set_symmetric_difference(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result);
+} // end set_symmetric_difference()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+OutputIterator set_symmetric_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                        InputIterator1                                              first1,
+                                        InputIterator1                                              last1,
+                                        InputIterator2                                              first2,
+                                        InputIterator2                                              last2,
+                                        OutputIterator                                              result,
+                                        StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_symmetric_difference;
+  return set_symmetric_difference(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result, comp);
+} // end set_symmetric_difference()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_symmetric_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  InputIterator1                                              keys_first1,
+                                  InputIterator1                                              keys_last1,
+                                  InputIterator2                                              keys_first2,
+                                  InputIterator2                                              keys_last2,
+                                  InputIterator3                                              values_first1,
+                                  InputIterator4                                              values_first2,
+                                  OutputIterator1                                             keys_result,
+                                  OutputIterator2                                             values_result)
+{
+  using thrust::system::detail::generic::set_symmetric_difference_by_key;
+  return set_symmetric_difference_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end set_symmetric_difference_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_symmetric_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  InputIterator1                                              keys_first1,
+                                  InputIterator1                                              keys_last1,
+                                  InputIterator2                                              keys_first2,
+                                  InputIterator2                                              keys_last2,
+                                  InputIterator3                                              values_first1,
+                                  InputIterator4                                              values_first2,
+                                  OutputIterator1                                             keys_result,
+                                  OutputIterator2                                             values_result,
+                                  StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_symmetric_difference_by_key;
+  return set_symmetric_difference_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end set_symmetric_difference_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_union(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1                                              first1,
+                         InputIterator1                                              last1,
+                         InputIterator2                                              first2,
+                         InputIterator2                                              last2,
+                         OutputIterator                                              result)
+{
+  using thrust::system::detail::generic::set_union;
+  return set_union(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result);
+} // end set_union()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+OutputIterator set_union(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1                                              first1,
+                         InputIterator1                                              last1,
+                         InputIterator2                                              first2,
+                         InputIterator2                                              last2,
+                         OutputIterator                                              result,
+                         StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_union;
+  return set_union(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, last2, result, comp);
+} // end set_union()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_union_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator1                                              keys_first1,
+                   InputIterator1                                              keys_last1,
+                   InputIterator2                                              keys_first2,
+                   InputIterator2                                              keys_last2,
+                   InputIterator3                                              values_first1,
+                   InputIterator4                                              values_first2,
+                   OutputIterator1                                             keys_result,
+                   OutputIterator2                                             values_result)
+{
+  using thrust::system::detail::generic::set_union_by_key;
+  return set_union_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end set_union_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_union_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator1                                              keys_first1,
+                   InputIterator1                                              keys_last1,
+                   InputIterator2                                              keys_first2,
+                   InputIterator2                                              keys_last2,
+                   InputIterator3                                              values_first1,
+                   InputIterator4                                              values_first2,
+                   OutputIterator1                                             keys_result,
+                   OutputIterator2                                             values_result,
+                   StrictWeakCompare                                           comp)
+{
+  using thrust::system::detail::generic::set_union_by_key;
+  return set_union_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end set_union_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+  OutputIterator set_difference(InputIterator1 first1,
+                                InputIterator1 last1,
+                                InputIterator2 first2,
+                                InputIterator2 last2,
+                                OutputIterator result,
+                                StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_difference(select_system(system1,system2,system3), first1, last1, first2, last2, result, comp);
+} // end set_difference()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_difference(InputIterator1 first1,
+                                InputIterator1 last1,
+                                InputIterator2 first2,
+                                InputIterator2 last2,
+                                OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_difference(select_system(system1,system2,system3), first1, last1, first2, last2, result);
+} // end set_difference()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_difference_by_key(InputIterator1 keys_first1,
+                          InputIterator1 keys_last1,
+                          InputIterator2 keys_first2,
+                          InputIterator2 keys_last2,
+                          InputIterator3 values_first1,
+                          InputIterator4 values_first2,
+                          OutputIterator1 keys_result,
+                          OutputIterator2 values_result,
+                          StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::set_difference_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end set_difference_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_difference_by_key(InputIterator1 keys_first1,
+                          InputIterator1 keys_last1,
+                          InputIterator2 keys_first2,
+                          InputIterator2 keys_last2,
+                          InputIterator3 values_first1,
+                          InputIterator4 values_first2,
+                          OutputIterator1 keys_result,
+                          OutputIterator2 values_result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::set_difference_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end set_difference_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+  OutputIterator set_intersection(InputIterator1 first1,
+                                  InputIterator1 last1,
+                                  InputIterator2 first2,
+                                  InputIterator2 last2,
+                                  OutputIterator result,
+                                  StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_intersection(select_system(system1,system2,system3), first1, last1, first2, last2, result, comp);
+} // end set_intersection()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_intersection(InputIterator1 first1,
+                                  InputIterator1 last1,
+                                  InputIterator2 first2,
+                                  InputIterator2 last2,
+                                  OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_intersection(select_system(system1,system2,system3), first1, last1, first2, last2, result);
+} // end set_intersection()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_intersection_by_key(InputIterator1 keys_first1,
+                            InputIterator1 keys_last1,
+                            InputIterator2 keys_first2,
+                            InputIterator2 keys_last2,
+                            InputIterator3 values_first1,
+                            OutputIterator1 keys_result,
+                            OutputIterator2 values_result,
+                            StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System4;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System5;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+
+  return thrust::set_intersection_by_key(select_system(system1,system2,system3,system4,system5), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, keys_result, values_result, comp);
+} // end set_intersection_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_intersection_by_key(InputIterator1 keys_first1,
+                            InputIterator1 keys_last1,
+                            InputIterator2 keys_first2,
+                            InputIterator2 keys_last2,
+                            InputIterator3 values_first1,
+                            OutputIterator1 keys_result,
+                            OutputIterator2 values_result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System4;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System5;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+
+  return thrust::set_intersection_by_key(select_system(system1,system2,system3,system4,system5), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, keys_result, values_result);
+} // end set_intersection_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+  OutputIterator set_symmetric_difference(InputIterator1 first1,
+                                          InputIterator1 last1,
+                                          InputIterator2 first2,
+                                          InputIterator2 last2,
+                                          OutputIterator result,
+                                          StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_symmetric_difference(select_system(system1,system2,system3), first1, last1, first2, last2, result, comp);
+} // end set_symmetric_difference()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_symmetric_difference(InputIterator1 first1,
+                                          InputIterator1 last1,
+                                          InputIterator2 first2,
+                                          InputIterator2 last2,
+                                          OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_symmetric_difference(select_system(system1,system2,system3), first1, last1, first2, last2, result);
+} // end set_symmetric_difference()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_symmetric_difference_by_key(InputIterator1 keys_first1,
+                                    InputIterator1 keys_last1,
+                                    InputIterator2 keys_first2,
+                                    InputIterator2 keys_last2,
+                                    InputIterator3 values_first1,
+                                    InputIterator4 values_first2,
+                                    OutputIterator1 keys_result,
+                                    OutputIterator2 values_result,
+                                    StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::set_symmetric_difference_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end set_symmetric_difference_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_symmetric_difference_by_key(InputIterator1 keys_first1,
+                                    InputIterator1 keys_last1,
+                                    InputIterator2 keys_first2,
+                                    InputIterator2 keys_last2,
+                                    InputIterator3 values_first1,
+                                    InputIterator4 values_first2,
+                                    OutputIterator1 keys_result,
+                                    OutputIterator2 values_result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::set_symmetric_difference_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end set_symmetric_difference_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+  OutputIterator set_union(InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           InputIterator2 last2,
+                           OutputIterator result,
+                           StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_union(select_system(system1,system2,system3), first1, last1, first2, last2, result, comp);
+} // end set_union()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_union(InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           InputIterator2 last2,
+                           OutputIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::set_union(select_system(system1,system2,system3), first1, last1, first2, last2, result);
+} // end set_union()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_union_by_key(InputIterator1 keys_first1,
+                     InputIterator1 keys_last1,
+                     InputIterator2 keys_first2,
+                     InputIterator2 keys_last2,
+                     InputIterator3 values_first1,
+                     InputIterator4 values_first2,
+                     OutputIterator1 keys_result,
+                     OutputIterator2 values_result,
+                     StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::set_union_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, comp);
+} // end set_union_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_union_by_key(InputIterator1 keys_first1,
+                     InputIterator1 keys_last1,
+                     InputIterator2 keys_first2,
+                     InputIterator2 keys_last2,
+                     InputIterator3 values_first1,
+                     InputIterator4 values_first2,
+                     OutputIterator1 keys_result,
+                     OutputIterator2 values_result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<InputIterator4>::type  System4;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System5;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System6;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+  System5 system5;
+  System6 system6;
+
+  return thrust::set_union_by_key(select_system(system1,system2,system3,system4,system5,system6), keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result);
+} // end set_union_by_key()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/shuffle.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/shuffle.inl
new file mode 100644
index 000000000000..48f5ba639294
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/shuffle.inl
@@ -0,0 +1,83 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/shuffle.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/shuffle.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename RandomIterator, typename URBG>
+__host__ __device__ void shuffle(
+    const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
+    RandomIterator first, RandomIterator last, URBG&& g) {
+  using thrust::system::detail::generic::shuffle;
+  return shuffle(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      first, last, g);
+}
+
+template <typename RandomIterator, typename URBG>
+__host__ __device__ void shuffle(RandomIterator first, RandomIterator last,
+                                 URBG&& g) {
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomIterator>::type System;
+  System system;
+
+  return thrust::shuffle(select_system(system), first, last, g);
+}
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy, typename RandomIterator,
+          typename OutputIterator, typename URBG>
+__host__ __device__ void shuffle_copy(
+    const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
+    RandomIterator first, RandomIterator last, OutputIterator result,
+    URBG&& g) {
+  using thrust::system::detail::generic::shuffle_copy;
+  return shuffle_copy(
+      thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+      first, last, result, g);
+}
+
+template <typename RandomIterator, typename OutputIterator, typename URBG>
+__host__ __device__ void shuffle_copy(RandomIterator first, RandomIterator last,
+                                      OutputIterator result, URBG&& g) {
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomIterator>::type System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::shuffle_copy(select_system(system1, system2), first, last,
+                              result, g);
+}
+
+THRUST_NAMESPACE_END
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/sort.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/sort.inl
new file mode 100644
index 000000000000..53f8bad93b1d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/sort.inl
@@ -0,0 +1,404 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/sort.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/sort.h>
+#include <thrust/system/detail/adl/sort.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename RandomAccessIterator>
+__host__ __device__
+  void sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last)
+{
+  using thrust::system::detail::generic::sort;
+  return sort(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end sort()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last,
+            StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::sort;
+  return sort(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end sort()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename RandomAccessIterator>
+__host__ __device__
+  void stable_sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last)
+{
+  using thrust::system::detail::generic::stable_sort;
+  return stable_sort(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end stable_sort()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last,
+                   StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::stable_sort;
+  return stable_sort(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end stable_sort()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first)
+{
+  using thrust::system::detail::generic::sort_by_key;
+  return sort_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first);
+} // end sort_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first,
+                   StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::sort_by_key;
+  return sort_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, comp);
+} // end sort_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void stable_sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first)
+{
+  using thrust::system::detail::generic::stable_sort_by_key;
+  return stable_sort_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first);
+} // end stable_sort_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first,
+                          StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::stable_sort_by_key;
+  return stable_sort_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, comp);
+} // end stable_sort_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  bool is_sorted(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last)
+{
+  using thrust::system::detail::generic::is_sorted;
+  return is_sorted(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end is_sorted()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename Compare>
+__host__ __device__
+  bool is_sorted(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 Compare comp)
+{
+  using thrust::system::detail::generic::is_sorted;
+  return is_sorted(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end is_sorted()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator is_sorted_until(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last)
+{
+  using thrust::system::detail::generic::is_sorted_until;
+  return is_sorted_until(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end is_sorted_until()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename Compare>
+__host__ __device__
+  ForwardIterator is_sorted_until(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Compare comp)
+{
+  using thrust::system::detail::generic::is_sorted_until;
+  return is_sorted_until(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, comp);
+} // end is_sorted_until()
+
+
+///////////////
+// Key Sorts //
+///////////////
+
+template<typename RandomAccessIterator>
+  void sort(RandomAccessIterator first,
+            RandomAccessIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System;
+
+  System system;
+
+  return thrust::sort(select_system(system), first, last);
+} // end sort()
+
+
+template<typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+  __host__ __device__
+  void sort(RandomAccessIterator first,
+            RandomAccessIterator last,
+            StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System;
+
+  System system;
+
+  return thrust::sort(select_system(system), first, last, comp);
+} // end sort()
+
+
+template<typename RandomAccessIterator>
+  void stable_sort(RandomAccessIterator first,
+                   RandomAccessIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System;
+
+  System system;
+
+  return thrust::stable_sort(select_system(system), first, last);
+} // end stable_sort()
+
+
+template<typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+  void stable_sort(RandomAccessIterator first,
+                   RandomAccessIterator last,
+                   StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator>::type System;
+
+  System system;
+
+  return thrust::stable_sort(select_system(system), first, last, comp);
+} // end stable_sort()
+
+
+
+/////////////////////
+// Key-Value Sorts //
+/////////////////////
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  void sort_by_key(RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator1>::type System1;
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::sort_by_key(select_system(system1,system2), keys_first, keys_last, values_first);
+} // end sort_by_key()
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+  void sort_by_key(RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first,
+                   StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator1>::type System1;
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::sort_by_key(select_system(system1,system2), keys_first, keys_last, values_first, comp);
+} // end sort_by_key()
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  void stable_sort_by_key(RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator1>::type System1;
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::stable_sort_by_key(select_system(system1,system2), keys_first, keys_last, values_first);
+} // end stable_sort_by_key()
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+  void stable_sort_by_key(RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first,
+                          StrictWeakOrdering comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator1>::type System1;
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::stable_sort_by_key(select_system(system1,system2), keys_first, keys_last, values_first, comp);
+} // end stable_sort_by_key()
+
+
+template<typename ForwardIterator>
+  bool is_sorted(ForwardIterator first,
+                 ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::is_sorted(select_system(system), first, last);
+} // end is_sorted()
+
+
+template<typename ForwardIterator,
+         typename Compare>
+  bool is_sorted(ForwardIterator first,
+                 ForwardIterator last,
+                 Compare comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::is_sorted(select_system(system), first, last, comp);
+} // end is_sorted()
+
+
+template<typename ForwardIterator>
+  ForwardIterator is_sorted_until(ForwardIterator first,
+                                  ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::is_sorted_until(select_system(system), first, last);
+} // end is_sorted_until()
+
+
+template<typename ForwardIterator,
+         typename Compare>
+  ForwardIterator is_sorted_until(ForwardIterator first,
+                                  ForwardIterator last,
+                                  Compare comp)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::is_sorted_until(select_system(system), first, last, comp);
+} // end is_sorted_until()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/static_assert.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/static_assert.h
new file mode 100644
index 000000000000..0e6132790555
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/static_assert.h
@@ -0,0 +1,91 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*
+ * (C) Copyright John Maddock 2000.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/preprocessor.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template <typename, bool x>
+struct depend_on_instantiation
+{
+  THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT bool value = x;
+};
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#  if THRUST_CPP_DIALECT >= 2017
+#    define THRUST_STATIC_ASSERT(B)        static_assert(B)
+#  else
+#    define THRUST_STATIC_ASSERT(B)        static_assert(B, "static assertion failed")
+#  endif
+#  define THRUST_STATIC_ASSERT_MSG(B, msg) static_assert(B, msg)
+
+#else // Older than C++11.
+
+// HP aCC cannot deal with missing names for template value parameters.
+template <bool x> struct STATIC_ASSERTION_FAILURE;
+
+template <> struct STATIC_ASSERTION_FAILURE<true> {};
+
+// HP aCC cannot deal with missing names for template value parameters.
+template <int x> struct static_assert_test {};
+
+#if    (  (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC)                  \
+       && (THRUST_GCC_VERSION >= 40800))                                      \
+    || (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG)
+  // Clang and GCC 4.8+ will complain about this typedef being unused unless we
+  // annotate it as such.
+#  define THRUST_STATIC_ASSERT(B)                                             \
+    typedef THRUST_NS_QUALIFIER::detail::static_assert_test<                  \
+      sizeof(THRUST_NS_QUALIFIER::detail::STATIC_ASSERTION_FAILURE<(bool)(B)>)\
+    >                                                                         \
+      THRUST_PP_CAT2(thrust_static_assert_typedef_, __LINE__)                 \
+      __attribute__((unused))                                                 \
+    /**/      
+#else
+#  define THRUST_STATIC_ASSERT(B)                                             \
+    typedef THRUST_NS_QUALIFIER::detail::static_assert_test<                  \
+      sizeof(THRUST_NS_QUALIFIER::detail::STATIC_ASSERTION_FAILURE<(bool)(B)>)\
+    >                                                                         \
+      THRUST_PP_CAT2(thrust_static_assert_typedef_, __LINE__)                 \
+    /**/      
+#endif
+
+#define THRUST_STATIC_ASSERT_MSG(B, msg) THRUST_STATIC_ASSERT(B)
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/static_map.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/static_map.h
new file mode 100644
index 000000000000..9f0d79e83e25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/static_map.h
@@ -0,0 +1,169 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+
+#include <thrust/detail/config.h>
+
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace static_map_detail
+{
+
+
+template<unsigned int k, unsigned int v>
+struct key_value
+{
+  static const unsigned int key = k;
+  static const unsigned int value = v;
+};
+
+
+template<typename Head, typename Tail = void>
+struct cons
+{
+  template<unsigned int key, unsigned int default_value>
+  struct static_get
+  {
+    static const unsigned int value = (key == Head::key) ? (Head::value) : Tail::template static_get<key,default_value>::value;
+  };
+
+
+  template<unsigned int default_value>
+  __host__ __device__
+  static unsigned int get(unsigned int key)
+  {
+    return (key == Head::key) ? (Head::value) : Tail::template get<default_value>(key);
+  }
+};
+
+
+template<typename Head>
+struct cons<Head,void>
+{
+  template<unsigned int key, unsigned int default_value>
+  struct static_get
+  {
+    static const unsigned int value = (key == Head::key) ? (Head::value) : default_value;
+  };
+
+  template<unsigned int default_value>
+  __host__ __device__
+  static unsigned int get(unsigned int key)
+  {
+    return (key == Head::key) ? (Head::value) : default_value;
+  }
+};
+
+
+template<unsigned int default_value,
+         unsigned int key0 = 0, unsigned int value0 = default_value,
+         unsigned int key1 = 0, unsigned int value1 = default_value,
+         unsigned int key2 = 0, unsigned int value2 = default_value,
+         unsigned int key3 = 0, unsigned int value3 = default_value,
+         unsigned int key4 = 0, unsigned int value4 = default_value,
+         unsigned int key5 = 0, unsigned int value5 = default_value,
+         unsigned int key6 = 0, unsigned int value6 = default_value,
+         unsigned int key7 = 0, unsigned int value7 = default_value>
+struct static_map
+{
+  typedef cons<
+    key_value<key0,value0>,
+    cons<
+      key_value<key1,value1>,
+      cons<
+        key_value<key2,value2>,
+        cons<
+          key_value<key3,value3>,
+          cons<
+            key_value<key4,value4>,
+            cons<
+              key_value<key5,value5>,
+              cons<
+                key_value<key6,value6>,
+                cons<
+                  key_value<key7,value7>
+                >
+              >
+            >
+          >
+        >
+      >
+    >
+  > impl;
+
+  template<unsigned int key>
+  struct static_get
+  {
+    static const unsigned int value = impl::template static_get<key,default_value>::value;
+  };
+
+  __host__ __device__
+  static unsigned int get(unsigned int key)
+  {
+    return impl::template get<default_value>(key);
+  }
+};
+
+
+} // end namespace static_map_detail
+
+
+template<unsigned int default_value,
+         unsigned int key0 = 0, unsigned int value0 = default_value,
+         unsigned int key1 = 0, unsigned int value1 = default_value,
+         unsigned int key2 = 0, unsigned int value2 = default_value,
+         unsigned int key3 = 0, unsigned int value3 = default_value,
+         unsigned int key4 = 0, unsigned int value4 = default_value,
+         unsigned int key5 = 0, unsigned int value5 = default_value,
+         unsigned int key6 = 0, unsigned int value6 = default_value,
+         unsigned int key7 = 0, unsigned int value7 = default_value>
+struct static_map
+  : static_map_detail::static_map<
+      default_value,
+      key0, value0,
+      key1, value1,
+      key2, value2,
+      key3, value3,
+      key4, value4,
+      key5, value5,
+      key6, value6,
+      key7, value7
+    >
+{};
+
+
+template<unsigned int key, typename StaticMap>
+struct static_lookup
+{
+  static const unsigned int value = StaticMap::template static_get<key>::value;
+};
+
+
+template<typename StaticMap>
+__host__ __device__
+unsigned int lookup(unsigned int key)
+{
+  return StaticMap::get(key);
+}
+
+
+} // end namespace detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.h
new file mode 100644
index 000000000000..305750f8ae78
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.h
@@ -0,0 +1,35 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+__thrust_exec_check_disable__
+template<typename Assignable1, typename Assignable2>
+__host__ __device__
+inline void swap(Assignable1 &a, Assignable2 &b)
+{
+  Assignable1 temp = a;
+  a = b;
+  b = temp;
+} // end swap()
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.inl
new file mode 100644
index 000000000000..196c34f41ee4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap.inl
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/swap.h>
+#include <thrust/detail/swap.h>
+#include <thrust/detail/swap_ranges.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap_ranges.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap_ranges.inl
new file mode 100644
index 000000000000..1f35c1ff3183
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/swap_ranges.inl
@@ -0,0 +1,64 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/swap.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/swap_ranges.h>
+#include <thrust/system/detail/adl/swap_ranges.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  ForwardIterator2 swap_ranges(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               ForwardIterator1 first1,
+                               ForwardIterator1 last1,
+                               ForwardIterator2 first2)
+{
+  using thrust::system::detail::generic::swap_ranges;
+  return swap_ranges(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2);
+} // end swap_ranges()
+
+
+template<typename ForwardIterator1,
+         typename ForwardIterator2>
+  ForwardIterator2 swap_ranges(ForwardIterator1 first1,
+                               ForwardIterator1 last1,
+                               ForwardIterator2 first2)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator1>::type System1;
+  typedef typename thrust::iterator_system<ForwardIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::swap_ranges(select_system(system1,system2), first1, last1, first2);
+} // end swap_ranges()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/tabulate.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tabulate.inl
new file mode 100644
index 000000000000..308be061faf2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tabulate.inl
@@ -0,0 +1,58 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/tabulate.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/tabulate.h>
+#include <thrust/system/detail/adl/tabulate.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename UnaryOperation>
+__host__ __device__
+  void tabulate(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                UnaryOperation unary_op)
+{
+  using thrust::system::detail::generic::tabulate;
+  return tabulate(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, unary_op);
+} // end tabulate()
+
+
+template<typename ForwardIterator, typename UnaryOperation>
+  void tabulate(ForwardIterator first,
+                ForwardIterator last,
+                UnaryOperation unary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::tabulate(select_system(system), first, last, unary_op);
+} // end tabulate()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.h
new file mode 100644
index 000000000000..cf4bc7d2d2e2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.h
@@ -0,0 +1,181 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file temporary_array.h
+ *  \brief Container-like class temporary storage inside algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+// Forward declare temporary_array, as it's used by the CUDA copy backend, which
+// is included in contiguous_storage's definition.
+template<typename T, typename System>
+  class temporary_array;
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/tagged_iterator.h>
+#include <thrust/detail/contiguous_storage.h>
+#include <thrust/detail/allocator/temporary_allocator.h>
+#include <thrust/detail/allocator/no_throw_allocator.h>
+#include <thrust/detail/memory_wrapper.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename T, typename System>
+  class temporary_array
+    : public contiguous_storage<
+               T,
+               no_throw_allocator<
+                 temporary_allocator<T,System>
+               >
+             >
+{
+  private:
+    typedef contiguous_storage<
+      T,
+      no_throw_allocator<
+        temporary_allocator<T,System>
+      >
+    > super_t;
+
+    // to help out the constructor
+    typedef no_throw_allocator<temporary_allocator<T,System> > alloc_type;
+
+  public:
+    typedef typename super_t::size_type size_type;
+
+    __host__ __device__
+    temporary_array(thrust::execution_policy<System> &system);
+
+    __host__ __device__
+    temporary_array(thrust::execution_policy<System> &system, size_type n);
+
+    // provide a kill-switch to explicitly avoid initialization
+    __host__ __device__
+    temporary_array(int uninit, thrust::execution_policy<System> &system, size_type n);
+
+    template<typename InputIterator>
+    __host__ __device__
+    temporary_array(thrust::execution_policy<System> &system,
+                    InputIterator first,
+                    size_type n);
+
+    template<typename InputIterator, typename InputSystem>
+    __host__ __device__
+    temporary_array(thrust::execution_policy<System> &system,
+                    thrust::execution_policy<InputSystem> &input_system,
+                    InputIterator first,
+                    size_type n);
+
+    template<typename InputIterator>
+    __host__ __device__
+    temporary_array(thrust::execution_policy<System> &system,
+                    InputIterator first,
+                    InputIterator last);
+
+    template<typename InputSystem, typename InputIterator>
+    __host__ __device__
+    temporary_array(thrust::execution_policy<System> &system,
+                    thrust::execution_policy<InputSystem> &input_system,
+                    InputIterator first,
+                    InputIterator last);
+
+    __host__ __device__
+    ~temporary_array();
+}; // end temporary_array
+
+
+// XXX eliminate this when we do ranges for real
+template<typename Iterator, typename System>
+  class tagged_iterator_range
+{
+  public:
+    typedef thrust::detail::tagged_iterator<Iterator,System> iterator;
+
+    template<typename Ignored1, typename Ignored2>
+    tagged_iterator_range(const Ignored1 &, const Ignored2 &, Iterator first, Iterator last)
+      : m_begin(first),
+        m_end(last)
+    {}
+
+    iterator begin(void) const { return m_begin; }
+    iterator end(void) const { return m_end; }
+
+  private:
+    iterator m_begin, m_end;
+};
+
+
+// if FromSystem is convertible to ToSystem, then just make a shallow
+// copy of the range. else, use a temporary_array
+// note that the resulting iterator is explicitly tagged with ToSystem either way
+template<typename Iterator, typename FromSystem, typename ToSystem>
+  struct move_to_system_base
+    : public eval_if<
+        is_convertible<
+          FromSystem,
+          ToSystem
+        >::value,
+        identity_<
+          tagged_iterator_range<Iterator,ToSystem>
+        >,
+        identity_<
+          temporary_array<
+            typename thrust::iterator_value<Iterator>::type,
+            ToSystem
+          >
+        >
+      >
+{};
+
+
+template<typename Iterator, typename FromSystem, typename ToSystem>
+  class move_to_system
+    : public move_to_system_base<
+        Iterator,
+        FromSystem,
+        ToSystem
+      >::type
+{
+  typedef typename move_to_system_base<Iterator,FromSystem,ToSystem>::type super_t;
+
+  public:
+    move_to_system(thrust::execution_policy<FromSystem> &from_system,
+                   thrust::execution_policy<ToSystem> &to_system,
+                   Iterator first,
+                   Iterator last)
+      : super_t(to_system, from_system, first, last) {}
+};
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/temporary_array.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.inl
new file mode 100644
index 000000000000..90b7279ac6d4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_array.inl
@@ -0,0 +1,168 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/temporary_array.h>
+#include <thrust/distance.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+namespace temporary_array_detail
+{
+
+
+template<typename T> struct avoid_initialization : thrust::detail::has_trivial_copy_constructor<T> {};
+
+
+template<typename T, typename TemporaryArray, typename Size>
+__host__ __device__
+typename thrust::detail::enable_if<
+  avoid_initialization<T>::value
+>::type
+  construct_values(TemporaryArray &,
+                   Size)
+{
+  // avoid the overhead of initialization
+} // end construct_values()
+
+
+template<typename T, typename TemporaryArray, typename Size>
+__host__ __device__
+typename thrust::detail::disable_if<
+  avoid_initialization<T>::value
+>::type
+  construct_values(TemporaryArray &a,
+                   Size n)
+{
+  a.default_construct_n(a.begin(), n);
+} // end construct_values()
+
+
+} // end temporary_array_detail
+
+
+template<typename T, typename System>
+__host__ __device__
+  temporary_array<T,System>
+    ::temporary_array(thrust::execution_policy<System> &system)
+      :super_t(alloc_type(temporary_allocator<T,System>(system)))
+{
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+__host__ __device__
+  temporary_array<T,System>
+    ::temporary_array(thrust::execution_policy<System> &system, size_type n)
+      :super_t(n, alloc_type(temporary_allocator<T,System>(system)))
+{
+  temporary_array_detail::construct_values<T>(*this, n);
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+__host__ __device__
+  temporary_array<T,System>
+    ::temporary_array(int, thrust::execution_policy<System> &system, size_type n)
+      :super_t(n, alloc_type(temporary_allocator<T,System>(system)))
+{
+  // avoid initialization
+  ;
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+  template<typename InputIterator>
+  __host__ __device__
+    temporary_array<T,System>
+      ::temporary_array(thrust::execution_policy<System> &system,
+                        InputIterator first,
+                        size_type n)
+        : super_t(alloc_type(temporary_allocator<T,System>(system)))
+{
+  super_t::allocate(n);
+
+  super_t::uninitialized_copy_n(system, first, n, super_t::begin());
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+  template<typename InputIterator, typename InputSystem>
+  __host__ __device__
+    temporary_array<T,System>
+      ::temporary_array(thrust::execution_policy<System> &system,
+                        thrust::execution_policy<InputSystem> &input_system,
+                        InputIterator first,
+                        size_type n)
+        : super_t(alloc_type(temporary_allocator<T,System>(system)))
+{
+  super_t::allocate(n);
+
+  super_t::uninitialized_copy_n(input_system, first, n, super_t::begin());
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+  template<typename InputIterator>
+  __host__ __device__
+    temporary_array<T,System>
+      ::temporary_array(thrust::execution_policy<System> &system,
+                        InputIterator first,
+                        InputIterator last)
+        : super_t(alloc_type(temporary_allocator<T,System>(system)))
+{
+  super_t::allocate(thrust::distance(first,last));
+
+  super_t::uninitialized_copy(system, first, last, super_t::begin());
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+  template<typename InputSystem, typename InputIterator>
+  __host__ __device__
+    temporary_array<T,System>
+      ::temporary_array(thrust::execution_policy<System> &system,
+                        thrust::execution_policy<InputSystem> &input_system,
+                        InputIterator first,
+                        InputIterator last)
+        : super_t(alloc_type(temporary_allocator<T,System>(system)))
+{
+  super_t::allocate(thrust::distance(first,last));
+
+  super_t::uninitialized_copy(input_system, first, last, super_t::begin());
+} // end temporary_array::temporary_array()
+
+
+template<typename T, typename System>
+__host__ __device__
+  temporary_array<T,System>
+    ::~temporary_array()
+{
+  // note that super_t::destroy will ignore trivial destructors automatically
+  super_t::destroy(super_t::begin(), super_t::end());
+} // end temporary_array::~temporary_array()
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_buffer.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_buffer.h
new file mode 100644
index 000000000000..be95e7180f5d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/temporary_buffer.h
@@ -0,0 +1,75 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+#include <thrust/detail/pointer.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/execute_with_allocator.h>
+#include <thrust/system/detail/generic/temporary_buffer.h>
+#include <thrust/system/detail/adl/temporary_buffer.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename T, typename DerivedPolicy, typename Pair>
+__host__ __device__
+  thrust::pair<thrust::pointer<T,DerivedPolicy>, typename thrust::pointer<T,DerivedPolicy>::difference_type>
+    down_cast_pair(Pair p)
+{
+  // XXX should use a hypothetical thrust::static_pointer_cast here
+  thrust::pointer<T,DerivedPolicy> ptr = thrust::pointer<T,DerivedPolicy>(static_cast<T*>(thrust::raw_pointer_cast(p.first)));
+
+  typedef thrust::pair<thrust::pointer<T,DerivedPolicy>, typename thrust::pointer<T,DerivedPolicy>::difference_type> result_type;
+  return result_type(ptr, p.second);
+} // end down_cast_pair()
+
+
+} // end detail
+
+
+__thrust_exec_check_disable__
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+  thrust::pair<thrust::pointer<T,DerivedPolicy>, typename thrust::pointer<T,DerivedPolicy>::difference_type>
+    get_temporary_buffer(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, typename thrust::pointer<T,DerivedPolicy>::difference_type n)
+{
+  using thrust::detail::get_temporary_buffer; // execute_with_allocator
+  using thrust::system::detail::generic::get_temporary_buffer;
+
+  return thrust::detail::down_cast_pair<T,DerivedPolicy>(get_temporary_buffer<T>(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), n));
+} // end get_temporary_buffer()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  void return_temporary_buffer(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, Pointer p, std::ptrdiff_t n)
+{
+  using thrust::detail::return_temporary_buffer; // execute_with_allocator
+  using thrust::system::detail::generic::return_temporary_buffer;
+
+  return return_temporary_buffer(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), p, n);
+} // end return_temporary_buffer()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform.inl
new file mode 100644
index 000000000000..62bafd35ef02
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform.inl
@@ -0,0 +1,247 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/transform.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/transform.h>
+#include <thrust/system/detail/adl/transform.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction>
+__host__ __device__
+  OutputIterator transform(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first, InputIterator last,
+                           OutputIterator result,
+                           UnaryFunction op)
+{
+  using thrust::system::detail::generic::transform;
+  return transform(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, op);
+} // end transform()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator transform(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1 first1, InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputIterator result,
+                           BinaryFunction op)
+{
+  using thrust::system::detail::generic::transform;
+  return transform(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, result, op);
+} // end transform()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               InputIterator first, InputIterator last,
+                               ForwardIterator result,
+                               UnaryFunction op,
+                               Predicate pred)
+{
+  using thrust::system::detail::generic::transform_if;
+  return transform_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, op, pred);
+} // end transform_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               InputIterator1 first, InputIterator1 last,
+                               InputIterator2 stencil,
+                               ForwardIterator result,
+                               UnaryFunction op,
+                               Predicate pred)
+{
+  using thrust::system::detail::generic::transform_if;
+  return transform_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, stencil, result, op, pred);
+} // end transform_if()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename ForwardIterator,
+         typename BinaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               InputIterator1 first1, InputIterator1 last1,
+                               InputIterator2 first2,
+                               InputIterator3 stencil,
+                               ForwardIterator result,
+                               BinaryFunction binary_op,
+                               Predicate pred)
+{
+  using thrust::system::detail::generic::transform_if;
+  return transform_if(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first1, last1, first2, stencil, result, binary_op, pred);
+} // end transform_if()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction>
+  OutputIterator transform(InputIterator first,
+                           InputIterator last,
+                           OutputIterator result,
+                           UnaryFunction op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::transform(select_system(system1,system2), first, last, result, op);
+} // end transform()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryFunction>
+  OutputIterator transform(InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputIterator result,
+                           BinaryFunction op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type System2;
+  typedef typename thrust::iterator_system<OutputIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::transform(select_system(system1,system2,system3), first1, last1, first2, result, op);
+} // end transform()
+
+
+template<typename InputIterator,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+  ForwardIterator transform_if(InputIterator first,
+                               InputIterator last,
+                               ForwardIterator result,
+                               UnaryFunction unary_op,
+                               Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type   System1;
+  typedef typename thrust::iterator_system<ForwardIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::transform_if(select_system(system1,system2), first, last, result, unary_op, pred);
+} // end transform_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+  ForwardIterator transform_if(InputIterator1 first,
+                               InputIterator1 last,
+                               InputIterator2 stencil,
+                               ForwardIterator result,
+                               UnaryFunction unary_op,
+                               Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<ForwardIterator>::type System3;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+
+  return thrust::transform_if(select_system(system1,system2,system3), first, last, stencil, result, unary_op, pred);
+} // end transform_if()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename ForwardIterator,
+         typename BinaryFunction,
+         typename Predicate>
+  ForwardIterator transform_if(InputIterator1 first1,
+                               InputIterator1 last1,
+                               InputIterator2 first2,
+                               InputIterator3 stencil,
+                               ForwardIterator result,
+                               BinaryFunction binary_op,
+                               Predicate pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<InputIterator3>::type  System3;
+  typedef typename thrust::iterator_system<ForwardIterator>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::transform_if(select_system(system1,system2,system3,system4), first1, last1, first2, stencil, result, binary_op, pred);
+} // end transform_if()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_reduce.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_reduce.inl
new file mode 100644
index 000000000000..702dd9f73b5a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_reduce.inl
@@ -0,0 +1,68 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/transform_reduce.h>
+#include <thrust/system/detail/adl/transform_reduce.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename UnaryFunction,
+         typename OutputType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputType transform_reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              InputIterator first,
+                              InputIterator last,
+                              UnaryFunction unary_op,
+                              OutputType init,
+                              BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::transform_reduce;
+  return transform_reduce(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, unary_op, init, binary_op);
+} // end transform_reduce()
+
+
+template<typename InputIterator,
+         typename UnaryFunction,
+         typename OutputType,
+         typename BinaryFunction>
+  OutputType transform_reduce(InputIterator first,
+                              InputIterator last,
+                              UnaryFunction unary_op,
+                              OutputType init,
+                              BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type System;
+
+  System system;
+
+  return thrust::transform_reduce(select_system(system), first, last, unary_op, init, binary_op);
+} // end transform_reduce()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_scan.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_scan.inl
new file mode 100644
index 000000000000..957001cefd6d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/transform_scan.inl
@@ -0,0 +1,117 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/scan.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/transform_scan.h>
+#include <thrust/system/detail/adl/transform_scan.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator transform_inclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::transform_inclusive_scan;
+  return transform_inclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, unary_op, binary_op);
+} // end transform_inclusive_scan()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename T,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator transform_exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          T init,
+                                          AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::transform_exclusive_scan;
+  return transform_exclusive_scan(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result, unary_op, init, binary_op);
+} // end transform_exclusive_scan()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename BinaryFunction>
+  OutputIterator transform_inclusive_scan(InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          BinaryFunction binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::transform_inclusive_scan(select_system(system1,system2), first, last, result, unary_op, binary_op);
+} // end transform_inclusive_scan()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename T,
+         typename AssociativeOperator>
+  OutputIterator transform_exclusive_scan(InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          T init,
+                                          AssociativeOperator binary_op)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::transform_exclusive_scan(select_system(system1,system2), first, last, result, unary_op, init, binary_op);
+} // end transform_exclusive_scan()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/trivial_sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/trivial_sequence.h
new file mode 100644
index 000000000000..2cf98e7872ab
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/trivial_sequence.h
@@ -0,0 +1,96 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file trivial_sequence.h
+ *  \brief Container-like class for wrapping sequences.  The wrapped
+ *         sequence always has trivial iterators, even when the input
+ *         sequence does not.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/type_traits/is_contiguous_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// never instantiated
+template<typename Iterator, typename DerivedPolicy, typename is_trivial> struct _trivial_sequence { };
+
+// trivial case
+template<typename Iterator, typename DerivedPolicy>
+struct _trivial_sequence<Iterator, DerivedPolicy, thrust::detail::true_type>
+{
+    typedef Iterator iterator_type;
+    Iterator first, last;
+
+    __host__ __device__
+    _trivial_sequence(thrust::execution_policy<DerivedPolicy> &, Iterator _first, Iterator _last) : first(_first), last(_last)
+    {
+    }
+
+    __host__ __device__
+    iterator_type begin() { return first; }
+
+    __host__ __device__
+    iterator_type end()   { return last; }
+};
+
+// non-trivial case
+template<typename Iterator, typename DerivedPolicy>
+struct _trivial_sequence<Iterator, DerivedPolicy, thrust::detail::false_type>
+{
+    typedef typename thrust::iterator_value<Iterator>::type iterator_value;
+    typedef typename thrust::detail::temporary_array<iterator_value, DerivedPolicy>::iterator iterator_type;
+    
+    thrust::detail::temporary_array<iterator_value, DerivedPolicy> buffer;
+
+    __host__ __device__
+    _trivial_sequence(thrust::execution_policy<DerivedPolicy> &exec, Iterator first, Iterator last)
+      : buffer(exec, first, last)
+    {
+    }
+
+    __host__ __device__
+    iterator_type begin() { return buffer.begin(); }
+
+    __host__ __device__
+    iterator_type end()   { return buffer.end(); }
+};
+
+template <typename Iterator, typename DerivedPolicy>
+struct trivial_sequence
+  : detail::_trivial_sequence<Iterator, DerivedPolicy, typename thrust::is_contiguous_iterator<Iterator>::type>
+{
+    typedef _trivial_sequence<Iterator, DerivedPolicy, typename thrust::is_contiguous_iterator<Iterator>::type> super_t;
+
+    __host__ __device__
+    trivial_sequence(thrust::execution_policy<DerivedPolicy> &exec, Iterator first, Iterator last) : super_t(exec, first, last) { }
+};
+
+} // end namespace detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple.inl
new file mode 100644
index 000000000000..f4930bf4bd6c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple.inl
@@ -0,0 +1,1004 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/swap.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// define null_type
+struct null_type {};
+
+// null_type comparisons
+__host__ __device__ inline
+bool operator==(const null_type&, const null_type&) { return true; }
+
+__host__ __device__ inline
+bool operator>=(const null_type&, const null_type&) { return true; }
+
+__host__ __device__ inline
+bool operator<=(const null_type&, const null_type&) { return true; }
+
+__host__ __device__ inline
+bool operator!=(const null_type&, const null_type&) { return false; }
+
+__host__ __device__ inline
+bool operator<(const null_type&, const null_type&) { return false; }
+
+__host__ __device__ inline
+bool operator>(const null_type&, const null_type&) { return false; }
+
+// forward declaration for tuple
+template <
+  class T0 = null_type, class T1 = null_type, class T2 = null_type,
+  class T3 = null_type, class T4 = null_type, class T5 = null_type,
+  class T6 = null_type, class T7 = null_type, class T8 = null_type,
+  class T9 = null_type>
+class tuple;
+
+
+template <size_t N, class T> struct tuple_element;
+
+template<size_t N, class T>
+  struct tuple_element_impl
+{
+  private:
+    typedef typename T::tail_type Next;
+
+  public:
+    /*! The result of this metafunction is returned in \c type.
+     */
+    typedef typename tuple_element_impl<N-1, Next>::type type;
+}; // end tuple_element
+
+template<class T>
+  struct tuple_element_impl<0,T>
+{
+  typedef typename T::head_type type;
+};
+
+template <size_t N, class T>
+  struct tuple_element<N, T const>
+{
+    using type = typename std::add_const<typename tuple_element<N, T>::type>::type;
+};
+
+template <size_t N, class T>
+struct tuple_element<N, T volatile>
+{
+    using type = typename std::add_volatile<typename tuple_element<N, T>::type>::type;
+};
+
+template <size_t N, class T>
+  struct tuple_element<N, T const volatile>
+{
+    using type = typename std::add_cv<typename tuple_element<N, T>::type>::type;
+};
+
+template <size_t N, class T>
+struct tuple_element{
+    using type = typename tuple_element_impl<N,T>::type;
+};
+
+// forward declaration of tuple_size
+template<class T> struct tuple_size;
+
+template<class T>
+  struct tuple_size<T const> : public tuple_size<T> {};
+
+template<class T>
+  struct tuple_size<T volatile> : public tuple_size<T> {};
+
+template<class T>
+  struct tuple_size<T const volatile> : public tuple_size<T> {};
+
+/*! This metafunction returns the number of elements
+ *  of a \p tuple type of interest.
+ *
+ *  \tparam T A \c tuple type of interest.
+ *
+ *  \see pair
+ *  \see tuple
+ */
+template<class T>
+  struct tuple_size
+{
+  /*! The result of this metafunction is returned in \c value.
+   */
+  static const int value = 1 + tuple_size<typename T::tail_type>::value;
+}; // end tuple_size
+
+
+// specializations for tuple_size
+template<>
+  struct tuple_size< tuple<> >
+{
+  static const int value = 0;
+}; // end tuple_size< tuple<> >
+
+template<>
+  struct tuple_size<null_type>
+{
+  static const int value = 0;
+}; // end tuple_size<null_type>
+
+
+
+// forward declaration of detail::cons
+namespace detail
+{
+
+template <class HT, class TT> struct cons;
+
+} // end detail
+
+
+// -- some traits classes for get functions
+template <class T> struct access_traits
+{
+  typedef const T& const_type;
+  typedef T& non_const_type;
+
+  typedef const typename thrust::detail::remove_cv<T>::type& parameter_type;
+
+// used as the tuple constructors parameter types
+// Rationale: non-reference tuple element types can be cv-qualified.
+// It should be possible to initialize such types with temporaries,
+// and when binding temporaries to references, the reference must
+// be non-volatile and const. 8.5.3. (5)
+}; // end access_traits
+
+template <class T> struct access_traits<T&>
+{
+  typedef T& const_type;
+  typedef T& non_const_type;
+
+  typedef T& parameter_type;
+}; // end access_traits<T&>
+
+// forward declarations of get()
+template<int N, class HT, class TT>
+__host__ __device__
+inline typename access_traits<
+                  typename tuple_element<N, detail::cons<HT, TT> >::type
+                >::non_const_type
+// XXX we probably don't need to do this for any compiler we care about -jph
+//get(cons<HT, TT>& c BOOST_APPEND_EXPLICIT_TEMPLATE_NON_TYPE(int, N));
+get(detail::cons<HT, TT>& c);
+
+template<int N, class HT, class TT>
+__host__ __device__
+inline typename access_traits<
+                  typename tuple_element<N, detail::cons<HT, TT> >::type
+                >::const_type
+// XXX we probably don't need to do this for any compiler we care about -jph
+//get(const cons<HT, TT>& c BOOST_APPEND_EXPLICIT_TEMPLATE_NON_TYPE(int, N));
+get(const detail::cons<HT, TT>& c);
+
+namespace detail
+{
+
+// -- generate error template, referencing to non-existing members of this
+// template is used to produce compilation errors intentionally
+template<class T>
+class generate_error;
+
+// - cons getters --------------------------------------------------------
+// called: get_class<N>::get<RETURN_TYPE>(aTuple)
+
+template< int N >
+struct get_class
+{
+  template<class RET, class HT, class TT >
+  __host__ __device__
+  inline static RET get(const cons<HT, TT>& t)
+  {
+    // XXX we may not need to deal with this for any compiler we care about -jph
+    //return get_class<N-1>::BOOST_NESTED_TEMPLATE get<RET>(t.tail);
+    return get_class<N-1>::template get<RET>(t.tail);
+
+    // gcc 4.3 couldn't compile this:
+    //return get_class<N-1>::get<RET>(t.tail);
+  }
+
+  template<class RET, class HT, class TT >
+  __host__ __device__
+  inline static RET get(cons<HT, TT>& t)
+  {
+    // XXX we may not need to deal with this for any compiler we care about -jph
+    //return get_class<N-1>::BOOST_NESTED_TEMPLATE get<RET>(t.tail);
+    return get_class<N-1>::template get<RET>(t.tail);
+
+    // gcc 4.3 couldn't compile this:
+    //return get_class<N-1>::get<RET>(t.tail);
+  }
+}; // end get_class
+
+template<>
+struct get_class<0>
+{
+  template<class RET, class HT, class TT>
+  __host__ __device__
+  inline static RET get(const cons<HT, TT>& t)
+  {
+    return t.head;
+  }
+
+  template<class RET, class HT, class TT>
+  __host__ __device__
+  inline static RET get(cons<HT, TT>& t)
+  {
+    return t.head;
+  }
+}; // get get_class<0>
+
+
+template <bool If, class Then, class Else> struct IF
+{
+  typedef Then RET;
+};
+
+template <class Then, class Else> struct IF<false, Then, Else>
+{
+  typedef Else RET;
+};
+
+//  These helper templates wrap void types and plain function types.
+//  The rationale is to allow one to write tuple types with those types
+//  as elements, even though it is not possible to instantiate such object.
+//  E.g: typedef tuple<void> some_type; // ok
+//  but: some_type x; // fails
+
+template <class T> class non_storeable_type
+{
+  __host__ __device__
+  non_storeable_type();
+};
+
+template <class T> struct wrap_non_storeable_type
+{
+  // XXX is_function looks complicated; punt for now -jph
+  //typedef typename IF<
+  //  ::thrust::detail::is_function<T>::value, non_storeable_type<T>, T
+  //>::RET type;
+
+  typedef T type;
+};
+
+template <> struct wrap_non_storeable_type<void>
+{
+  typedef non_storeable_type<void> type;
+};
+
+
+template <class HT, class TT>
+  struct cons
+{
+  typedef HT head_type;
+  typedef TT tail_type;
+
+  typedef typename
+    wrap_non_storeable_type<head_type>::type stored_head_type;
+
+  stored_head_type head;
+  tail_type tail;
+
+  inline __host__ __device__
+  typename access_traits<stored_head_type>::non_const_type
+  get_head() { return head; }
+
+  inline __host__ __device__
+  typename access_traits<tail_type>::non_const_type
+  get_tail() { return tail; }
+
+  inline __host__ __device__
+  typename access_traits<stored_head_type>::const_type
+  get_head() const { return head; }
+
+  inline __host__ __device__
+  typename access_traits<tail_type>::const_type
+  get_tail() const { return tail; }
+
+  inline __host__ __device__
+  cons(void) : head(), tail() {}
+  //  cons() : head(detail::default_arg<HT>::f()), tail() {}
+
+  // the argument for head is not strictly needed, but it prevents
+  // array type elements. This is good, since array type elements
+  // cannot be supported properly in any case (no assignment,
+  // copy works only if the tails are exactly the same type, ...)
+
+  inline __host__ __device__
+  cons(typename access_traits<stored_head_type>::parameter_type h,
+       const tail_type& t)
+    : head (h), tail(t) {}
+
+  template <class T1, class T2, class T3, class T4, class T5,
+            class T6, class T7, class T8, class T9, class T10>
+  inline __host__ __device__
+  cons( T1& t1, T2& t2, T3& t3, T4& t4, T5& t5,
+        T6& t6, T7& t7, T8& t8, T9& t9, T10& t10 )
+    : head (t1),
+      tail (t2, t3, t4, t5, t6, t7, t8, t9, t10, static_cast<const null_type&>(null_type()))
+      {}
+
+  template <class T2, class T3, class T4, class T5,
+            class T6, class T7, class T8, class T9, class T10>
+  inline __host__ __device__
+  cons( const null_type& /*t1*/, T2& t2, T3& t3, T4& t4, T5& t5,
+        T6& t6, T7& t7, T8& t8, T9& t9, T10& t10 )
+    : head (),
+      tail (t2, t3, t4, t5, t6, t7, t8, t9, t10, static_cast<const null_type&>(null_type()))
+      {}
+
+
+  template <class HT2, class TT2>
+  inline __host__ __device__
+  cons( const cons<HT2, TT2>& u ) : head(u.head), tail(u.tail) {}
+
+#if THRUST_CPP_DIALECT >= 2011
+  cons(const cons &) = default;
+#endif
+
+  __thrust_exec_check_disable__
+  template <class HT2, class TT2>
+  inline __host__ __device__
+  cons& operator=( const cons<HT2, TT2>& u ) {
+    head=u.head; tail=u.tail; return *this;
+  }
+
+  // must define assignment operator explicitly, implicit version is
+  // illformed if HT is a reference (12.8. (12))
+  __thrust_exec_check_disable__
+  inline __host__ __device__
+  cons& operator=(const cons& u) {
+    head = u.head; tail = u.tail;  return *this;
+  }
+
+  // XXX enable when we support std::pair -jph
+  //template <class T1, class T2>
+  //__host__ __device__
+  //cons& operator=( const std::pair<T1, T2>& u ) {
+  //  //BOOST_STATIC_ASSERT(length<cons>::value == 2); // check length = 2
+  //  head = u.first; tail.head = u.second; return *this;
+  //}
+
+  // get member functions (non-const and const)
+  template <int N>
+  __host__ __device__
+  typename access_traits<
+             typename tuple_element<N, cons<HT, TT> >::type
+           >::non_const_type
+  get() {
+    return thrust::get<N>(*this); // delegate to non-member get
+  }
+
+  template <int N>
+  __host__ __device__
+  typename access_traits<
+             typename tuple_element<N, cons<HT, TT> >::type
+           >::const_type
+  get() const {
+    return thrust::get<N>(*this); // delegate to non-member get
+  }
+
+  inline __host__ __device__
+  void swap(cons &c)
+  {
+    using thrust::swap;
+
+    swap(head, c.head);
+    tail.swap(c.tail);
+  }
+};
+
+template <class HT>
+  struct cons<HT, null_type>
+{
+  typedef HT head_type;
+  typedef null_type tail_type;
+  typedef cons<HT, null_type> self_type;
+
+  typedef typename
+    wrap_non_storeable_type<head_type>::type stored_head_type;
+  stored_head_type head;
+
+  typename access_traits<stored_head_type>::non_const_type
+  inline __host__ __device__
+  get_head() { return head; }
+
+  inline __host__ __device__
+  null_type get_tail() { return null_type(); }
+
+  inline __host__ __device__
+  typename access_traits<stored_head_type>::const_type
+  get_head() const { return head; }
+
+  inline __host__ __device__
+  null_type get_tail() const { return null_type(); }
+
+  inline __host__ __device__
+  cons() : head() {}
+
+  inline __host__ __device__
+  cons(typename access_traits<stored_head_type>::parameter_type h,
+       const null_type& = null_type())
+    : head (h) {}
+
+  template<class T1>
+  inline __host__ __device__
+  cons(T1& t1, const null_type&, const null_type&, const null_type&,
+       const null_type&, const null_type&, const null_type&,
+       const null_type&, const null_type&, const null_type&)
+  : head (t1) {}
+
+  inline __host__ __device__
+  cons(const null_type&,
+       const null_type&, const null_type&, const null_type&,
+       const null_type&, const null_type&, const null_type&,
+       const null_type&, const null_type&, const null_type&)
+  : head () {}
+
+  template <class HT2>
+  inline __host__ __device__
+  cons( const cons<HT2, null_type>& u ) : head(u.head) {}
+
+#if THRUST_CPP_DIALECT >= 2011
+  cons(const cons &) = default;
+#endif
+
+  __thrust_exec_check_disable__
+  template <class HT2>
+  inline __host__ __device__
+  cons& operator=(const cons<HT2, null_type>& u )
+  {
+    head = u.head;
+    return *this;
+  }
+
+  // must define assignment operator explicitly, implicit version
+  // is illformed if HT is a reference
+  inline __host__ __device__
+  cons& operator=(const cons& u) { head = u.head; return *this; }
+
+  template <int N>
+  inline __host__ __device__
+  typename access_traits<
+             typename tuple_element<N, self_type>::type
+            >::non_const_type
+  // XXX we probably don't need this for the compilers we care about -jph
+  //get(BOOST_EXPLICIT_TEMPLATE_NON_TYPE(int, N))
+  get(void)
+  {
+    return thrust::get<N>(*this);
+  }
+
+  template <int N>
+  inline __host__ __device__
+  typename access_traits<
+             typename tuple_element<N, self_type>::type
+           >::const_type
+  // XXX we probably don't need this for the compilers we care about -jph
+  //get(BOOST_EXPLICIT_TEMPLATE_NON_TYPE(int, N)) const
+  get(void) const
+  {
+    return thrust::get<N>(*this);
+  }
+
+  inline __host__ __device__
+  void swap(cons &c)
+  {
+    using thrust::swap;
+
+    swap(head, c.head);
+  }
+}; // end cons
+
+template <class T0, class T1, class T2, class T3, class T4,
+          class T5, class T6, class T7, class T8, class T9>
+  struct map_tuple_to_cons
+{
+  typedef cons<T0,
+               typename map_tuple_to_cons<T1, T2, T3, T4, T5,
+                                          T6, T7, T8, T9, null_type>::type
+              > type;
+}; // end map_tuple_to_cons
+
+// The empty tuple is a null_type
+template <>
+  struct map_tuple_to_cons<null_type, null_type, null_type, null_type, null_type, null_type, null_type, null_type, null_type, null_type>
+{
+  typedef null_type type;
+}; // end map_tuple_to_cons<...>
+
+
+
+// ---------------------------------------------------------------------------
+// The call_traits for make_tuple
+
+// Must be instantiated with plain or const plain types (not with references)
+
+// from template<class T> foo(const T& t) : make_tuple_traits<const T>::type
+// from template<class T> foo(T& t) : make_tuple_traits<T>::type
+
+// Conversions:
+// T -> T,
+// references -> compile_time_error
+// array -> const ref array
+
+
+template<class T>
+struct make_tuple_traits {
+  typedef T type;
+
+  // commented away, see below  (JJ)
+  //  typedef typename IF<
+  //  boost::is_function<T>::value,
+  //  T&,
+  //  T>::RET type;
+
+};
+
+// The is_function test was there originally for plain function types,
+// which can't be stored as such (we must either store them as references or
+// pointers). Such a type could be formed if make_tuple was called with a
+// reference to a function.
+// But this would mean that a const qualified function type was formed in
+// the make_tuple function and hence make_tuple can't take a function
+// reference as a parameter, and thus T can't be a function type.
+// So is_function test was removed.
+// (14.8.3. says that type deduction fails if a cv-qualified function type
+// is created. (It only applies for the case of explicitly specifying template
+// args, though?)) (JJ)
+
+template<class T>
+struct make_tuple_traits<T&> {
+  typedef typename
+     detail::generate_error<T&>::
+       do_not_use_with_reference_type error;
+};
+
+// Arrays can't be stored as plain types; convert them to references.
+// All arrays are converted to const. This is because make_tuple takes its
+// parameters as const T& and thus the knowledge of the potential
+// non-constness of actual argument is lost.
+template<class T, int n>  struct make_tuple_traits <T[n]> {
+  typedef const T (&type)[n];
+};
+
+template<class T, int n>
+struct make_tuple_traits<const T[n]> {
+  typedef const T (&type)[n];
+};
+
+template<class T, int n>  struct make_tuple_traits<volatile T[n]> {
+  typedef const volatile T (&type)[n];
+};
+
+template<class T, int n>
+struct make_tuple_traits<const volatile T[n]> {
+  typedef const volatile T (&type)[n];
+};
+
+// XXX enable these if we ever care about reference_wrapper -jph
+//template<class T>
+//struct make_tuple_traits<reference_wrapper<T> >{
+//  typedef T& type;
+//};
+//
+//template<class T>
+//struct make_tuple_traits<const reference_wrapper<T> >{
+//  typedef T& type;
+//};
+
+
+// a helper traits to make the make_tuple functions shorter (Vesa Karvonen's
+// suggestion)
+template <
+  class T0 = null_type, class T1 = null_type, class T2 = null_type,
+  class T3 = null_type, class T4 = null_type, class T5 = null_type,
+  class T6 = null_type, class T7 = null_type, class T8 = null_type,
+  class T9 = null_type
+>
+struct make_tuple_mapper {
+  typedef
+    tuple<typename make_tuple_traits<T0>::type,
+          typename make_tuple_traits<T1>::type,
+          typename make_tuple_traits<T2>::type,
+          typename make_tuple_traits<T3>::type,
+          typename make_tuple_traits<T4>::type,
+          typename make_tuple_traits<T5>::type,
+          typename make_tuple_traits<T6>::type,
+          typename make_tuple_traits<T7>::type,
+          typename make_tuple_traits<T8>::type,
+          typename make_tuple_traits<T9>::type> type;
+};
+
+} // end detail
+
+
+template<int N, class HT, class TT>
+__host__ __device__
+inline typename access_traits<
+                  typename tuple_element<N, detail::cons<HT, TT> >::type
+                >::non_const_type
+get(detail::cons<HT, TT>& c)
+{
+  //return detail::get_class<N>::BOOST_NESTED_TEMPLATE
+
+  // gcc 4.3 couldn't compile this:
+  //return detail::get_class<N>::
+
+  return detail::get_class<N>::template
+         get<
+           typename access_traits<
+             typename tuple_element<N, detail::cons<HT, TT> >::type
+           >::non_const_type,
+           HT,TT
+         >(c);
+}
+
+
+// get function for const cons-lists, returns a const reference to
+// the element. If the element is a reference, returns the reference
+// as such (that is, can return a non-const reference)
+template<int N, class HT, class TT>
+__host__ __device__
+inline typename access_traits<
+                  typename tuple_element<N, detail::cons<HT, TT> >::type
+                >::const_type
+get(const detail::cons<HT, TT>& c)
+{
+  //return detail::get_class<N>::BOOST_NESTED_TEMPLATE
+
+  // gcc 4.3 couldn't compile this:
+  //return detail::get_class<N>::
+
+  return detail::get_class<N>::template
+         get<
+           typename access_traits<
+             typename tuple_element<N, detail::cons<HT, TT> >::type
+           >::const_type,
+           HT,TT
+         >(c);
+}
+
+
+template<class T0>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0>::type
+    make_tuple(const T0& t0)
+{
+  typedef typename detail::make_tuple_mapper<T0>::type t;
+  return t(t0);
+} // end make_tuple()
+
+template<class T0, class T1>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1>::type
+    make_tuple(const T0& t0, const T1& t1)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1>::type t;
+  return t(t0,t1);
+} // end make_tuple()
+
+template<class T0, class T1, class T2>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2>::type t;
+  return t(t0,t1,t2);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3>::type t;
+  return t(t0,t1,t2,t3);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3, class T4>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3,T4>::type t;
+  return t(t0,t1,t2,t3,t4);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3, class T4, class T5>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3,T4,T5>::type t;
+  return t(t0,t1,t2,t3,t4,t5);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3,T4,T5,T6>::type t;
+  return t(t0,t1,t2,t3,t4,t5,t6);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6, class T7>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6, T7>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6, const T7& t7)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3,T4,T5,T6,T7>::type t;
+  return t(t0,t1,t2,t3,t4,t5,t6,t7);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6, class T7, class T8>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6, T7, T8>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6, const T7& t7, const T8& t8)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3,T4,T5,T6,T7,T8>::type t;
+  return t(t0,t1,t2,t3,t4,t5,t6,t7,t8);
+} // end make_tuple()
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6, class T7, class T8, class T9>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6, T7, T8, T9>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6, const T7& t7, const T8& t8, const T9& t9)
+{
+  typedef typename detail::make_tuple_mapper<T0,T1,T2,T3,T4,T5,T6,T7,T8,T9>::type t;
+  return t(t0,t1,t2,t3,t4,t5,t6,t7,t8,t9);
+} // end make_tuple()
+
+
+template<typename T0>
+__host__ __device__ inline
+tuple<T0&> tie(T0 &t0)
+{
+  return tuple<T0&>(t0);
+}
+
+template<typename T0,typename T1>
+__host__ __device__ inline
+tuple<T0&,T1&> tie(T0 &t0, T1 &t1)
+{
+  return tuple<T0&,T1&>(t0,t1);
+}
+
+template<typename T0,typename T1, typename T2>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&> tie(T0 &t0, T1 &t1, T2 &t2)
+{
+  return tuple<T0&,T1&,T2&>(t0,t1,t2);
+}
+
+template<typename T0,typename T1, typename T2, typename T3>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3)
+{
+  return tuple<T0&,T1&,T2&,T3&>(t0,t1,t2,t3);
+}
+
+template<typename T0,typename T1, typename T2, typename T3, typename T4>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4)
+{
+  return tuple<T0&,T1&,T2&,T3&,T4&>(t0,t1,t2,t3,t4);
+}
+
+template<typename T0,typename T1, typename T2, typename T3, typename T4, typename T5>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5)
+{
+  return tuple<T0&,T1&,T2&,T3&,T4&,T5&>(t0,t1,t2,t3,t4,t5);
+}
+
+template<typename T0,typename T1, typename T2, typename T3, typename T4, typename T5, typename T6>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6)
+{
+  return tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&>(t0,t1,t2,t3,t4,t5,t6);
+}
+
+template<typename T0,typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6, T7 &t7)
+{
+  return tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&>(t0,t1,t2,t3,t4,t5,t6,t7);
+}
+
+template<typename T0,typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&,T8&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6, T7 &t7, T8 &t8)
+{
+  return tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&,T8&>(t0,t1,t2,t3,t4,t5,t6,t7,t8);
+}
+
+template<typename T0,typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&,T8&,T9&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6, T7 &t7, T8 &t8, T9 &t9)
+{
+  return tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&,T8&,T9&>(t0,t1,t2,t3,t4,t5,t6,t7,t8,t9);
+}
+
+template<
+  typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9,
+  typename U0, typename U1, typename U2, typename U3, typename U4, typename U5, typename U6, typename U7, typename U8, typename U9
+>
+__host__ __device__ inline
+void swap(thrust::tuple<T0,T1,T2,T3,T4,T5,T6,T7,T8,T9> &x,
+          thrust::tuple<U0,U1,U2,U3,U4,U5,U6,U7,U8,U9> &y)
+{
+  return x.swap(y);
+}
+
+
+
+namespace detail
+{
+
+template<class T1, class T2>
+__host__ __device__
+inline bool eq(const T1& lhs, const T2& rhs) {
+  return lhs.get_head() == rhs.get_head() &&
+         eq(lhs.get_tail(), rhs.get_tail());
+}
+template<>
+__host__ __device__
+inline bool eq<null_type,null_type>(const null_type&, const null_type&) { return true; }
+
+template<class T1, class T2>
+__host__ __device__
+inline bool neq(const T1& lhs, const T2& rhs) {
+  return lhs.get_head() != rhs.get_head()  ||
+         neq(lhs.get_tail(), rhs.get_tail());
+}
+template<>
+__host__ __device__
+inline bool neq<null_type,null_type>(const null_type&, const null_type&) { return false; }
+
+template<class T1, class T2>
+__host__ __device__
+inline bool lt(const T1& lhs, const T2& rhs) {
+  return (lhs.get_head() < rhs.get_head())  ||
+            (!(rhs.get_head() < lhs.get_head()) &&
+             lt(lhs.get_tail(), rhs.get_tail()));
+}
+template<>
+__host__ __device__
+inline bool lt<null_type,null_type>(const null_type&, const null_type&) { return false; }
+
+template<class T1, class T2>
+__host__ __device__
+inline bool gt(const T1& lhs, const T2& rhs) {
+  return (lhs.get_head() > rhs.get_head())  ||
+            (!(rhs.get_head() > lhs.get_head()) &&
+             gt(lhs.get_tail(), rhs.get_tail()));
+}
+template<>
+__host__ __device__
+inline bool gt<null_type,null_type>(const null_type&, const null_type&) { return false; }
+
+template<class T1, class T2>
+__host__ __device__
+inline bool lte(const T1& lhs, const T2& rhs) {
+  return lhs.get_head() <= rhs.get_head()  &&
+          ( !(rhs.get_head() <= lhs.get_head()) ||
+            lte(lhs.get_tail(), rhs.get_tail()));
+}
+template<>
+__host__ __device__
+inline bool lte<null_type,null_type>(const null_type&, const null_type&) { return true; }
+
+template<class T1, class T2>
+__host__ __device__
+inline bool gte(const T1& lhs, const T2& rhs) {
+  return lhs.get_head() >= rhs.get_head()  &&
+          ( !(rhs.get_head() >= lhs.get_head()) ||
+            gte(lhs.get_tail(), rhs.get_tail()));
+}
+template<>
+__host__ __device__
+inline bool gte<null_type,null_type>(const null_type&, const null_type&) { return true; }
+
+} // end detail
+
+
+
+// equal ----
+
+template<class T1, class T2, class S1, class S2>
+__host__ __device__
+inline bool operator==(const detail::cons<T1, T2>& lhs, const detail::cons<S1, S2>& rhs)
+{
+  // XXX support this eventually -jph
+  //// check that tuple lengths are equal
+  //BOOST_STATIC_ASSERT(tuple_size<T2>::value == tuple_size<S2>::value);
+
+  return  detail::eq(lhs, rhs);
+} // end operator==()
+
+// not equal -----
+
+template<class T1, class T2, class S1, class S2>
+__host__ __device__
+inline bool operator!=(const detail::cons<T1, T2>& lhs, const detail::cons<S1, S2>& rhs)
+{
+  // XXX support this eventually -jph
+  //// check that tuple lengths are equal
+  //BOOST_STATIC_ASSERT(tuple_size<T2>::value == tuple_size<S2>::value);
+
+  return detail::neq(lhs, rhs);
+} // end operator!=()
+
+// <
+template<class T1, class T2, class S1, class S2>
+__host__ __device__
+inline bool operator<(const detail::cons<T1, T2>& lhs, const detail::cons<S1, S2>& rhs)
+{
+  // XXX support this eventually -jph
+  //// check that tuple lengths are equal
+  //BOOST_STATIC_ASSERT(tuple_size<T2>::value == tuple_size<S2>::value);
+
+  return detail::lt(lhs, rhs);
+} // end operator<()
+
+// >
+template<class T1, class T2, class S1, class S2>
+__host__ __device__
+inline bool operator>(const detail::cons<T1, T2>& lhs, const detail::cons<S1, S2>& rhs)
+{
+  // XXX support this eventually -jph
+  //// check that tuple lengths are equal
+  //BOOST_STATIC_ASSERT(tuple_size<T2>::value == tuple_size<S2>::value);
+
+  return detail::gt(lhs, rhs);
+} // end operator>()
+
+// <=
+template<class T1, class T2, class S1, class S2>
+__host__ __device__
+inline bool operator<=(const detail::cons<T1, T2>& lhs, const detail::cons<S1, S2>& rhs)
+{
+  // XXX support this eventually -jph
+  //// check that tuple lengths are equal
+  //BOOST_STATIC_ASSERT(tuple_size<T2>::value == tuple_size<S2>::value);
+
+  return detail::lte(lhs, rhs);
+} // end operator<=()
+
+// >=
+template<class T1, class T2, class S1, class S2>
+__host__ __device__
+inline bool operator>=(const detail::cons<T1, T2>& lhs, const detail::cons<S1, S2>& rhs)
+{
+  // XXX support this eventually -jph
+  //// check that tuple lengths are equal
+  //BOOST_STATIC_ASSERT(tuple_size<T2>::value == tuple_size<S2>::value);
+
+  return detail::gte(lhs, rhs);
+} // end operator>=()
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_algorithms.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_algorithms.h
new file mode 100644
index 000000000000..2e49f4281f3f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_algorithms.h
@@ -0,0 +1,110 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/detail/type_deduction.h>
+#include <thrust/type_traits/integer_sequence.h>
+
+#include <tuple>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename Tuple, std::size_t... Is>
+auto tuple_subset(Tuple&& t, index_sequence<Is...>)
+THRUST_DECLTYPE_RETURNS(std::make_tuple(std::get<Is>(THRUST_FWD(t))...));
+
+namespace detail
+{
+
+template <typename Tuple, typename F, std::size_t... Is>
+void tuple_for_each_impl(Tuple&& t, F&& f, index_sequence<Is...>)
+{
+  auto l = { (f(std::get<Is>(t)), 0)... };
+  THRUST_UNUSED_VAR(l);
+}
+
+template <typename Tuple, typename F, std::size_t... Is>
+auto tuple_transform_impl(Tuple&& t, F&& f, index_sequence<Is...>)
+THRUST_DECLTYPE_RETURNS(std::make_tuple(f(std::get<Is>(t))...));
+
+} // namespace detail
+
+template <typename... Ts, typename F>
+auto tuple_for_each(std::tuple<Ts...>& t, F&& f)
+THRUST_DECLTYPE_RETURNS(
+  detail::tuple_for_each_impl(
+    t
+  , THRUST_FWD(f)
+  , make_index_sequence<sizeof...(Ts)>{}
+  )
+);
+template <typename... Ts, typename F>
+auto tuple_for_each(std::tuple<Ts...> const& t, F&& f)
+THRUST_DECLTYPE_RETURNS(
+  detail::tuple_for_each_impl(
+    t
+  , THRUST_FWD(f)
+  , make_index_sequence<sizeof...(Ts)>{}
+  )
+);
+template <typename... Ts, typename F>
+auto tuple_for_each(std::tuple<Ts...>&& t, F&& f)
+THRUST_DECLTYPE_RETURNS(
+  detail::tuple_for_each_impl(
+    std::move(t)
+  , THRUST_FWD(f)
+  , make_index_sequence<sizeof...(Ts)>{}
+  )
+);
+
+template <typename... Ts, typename F>
+auto tuple_transform(std::tuple<Ts...>& t, F&& f)
+THRUST_DECLTYPE_RETURNS(
+  detail::tuple_transform_impl(
+    t
+  , THRUST_FWD(f)
+  , make_index_sequence<sizeof...(Ts)>{}
+  )
+);
+template <typename... Ts, typename F>
+auto tuple_transform(std::tuple<Ts...> const& t, F&& f)
+THRUST_DECLTYPE_RETURNS(
+  detail::tuple_transform_impl(
+    t
+  , THRUST_FWD(f)
+  , make_index_sequence<sizeof...(Ts)>{}
+  )
+);
+template <typename... Ts, typename F>
+auto tuple_transform(std::tuple<Ts...>&& t, F&& f)
+THRUST_DECLTYPE_RETURNS(
+  detail::tuple_transform_impl(
+    std::move(t)
+  , THRUST_FWD(f)
+  , make_index_sequence<sizeof...(Ts)>{}
+  )
+);
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_meta_transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_meta_transform.h
new file mode 100644
index 000000000000..285cae8b4c72
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_meta_transform.h
@@ -0,0 +1,58 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/tuple.h>
+#include <thrust/type_traits/integer_sequence.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// introduce an intermediate type tuple_meta_transform_WAR_NVCC
+// rather than directly specializing tuple_meta_transform with
+// default argument IndexSequence = thrust::make_index_sequence<thrust::tuple_size<Tuple>::value>
+// to workaround nvcc 11.0 compiler bug
+template<typename Tuple,
+         template<typename> class UnaryMetaFunction,
+         typename IndexSequence>
+  struct tuple_meta_transform_WAR_NVCC;
+
+template<typename Tuple,
+         template<typename> class UnaryMetaFunction,
+         size_t... Is>
+  struct tuple_meta_transform_WAR_NVCC<Tuple, UnaryMetaFunction, thrust::index_sequence<Is...>>
+{
+  typedef thrust::tuple<
+    typename UnaryMetaFunction<typename thrust::tuple_element<Is,Tuple>::type>::type...
+  > type;
+};
+
+template<typename Tuple,
+         template<typename> class UnaryMetaFunction>
+  struct tuple_meta_transform
+{
+  typedef typename tuple_meta_transform_WAR_NVCC<Tuple, UnaryMetaFunction, thrust::make_index_sequence<thrust::tuple_size<Tuple>::value>>::type type;
+};
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_transform.h
new file mode 100644
index 000000000000..1011d5179e31
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/tuple_transform.h
@@ -0,0 +1,84 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/tuple.h>
+#include <thrust/detail/tuple_meta_transform.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<typename Tuple,
+         template<typename> class UnaryMetaFunction,
+         typename UnaryFunction,
+         typename IndexSequence = thrust::make_index_sequence<thrust::tuple_size<Tuple>::value>>
+  struct tuple_transform_functor;
+
+
+template<typename Tuple,
+         template<typename> class UnaryMetaFunction,
+         typename UnaryFunction,
+         size_t... Is>
+  struct tuple_transform_functor<Tuple,UnaryMetaFunction,UnaryFunction,thrust::index_sequence<Is...>>
+{
+  static __host__
+  typename tuple_meta_transform<Tuple,UnaryMetaFunction>::type
+  do_it_on_the_host(const Tuple &t, UnaryFunction f)
+  {
+    typedef typename tuple_meta_transform<Tuple,UnaryMetaFunction>::type XfrmTuple;
+
+    return XfrmTuple(f(thrust::get<Is>(t))...);
+  }
+
+  static __host__ __device__
+  typename tuple_meta_transform<Tuple,UnaryMetaFunction>::type
+  do_it_on_the_host_or_device(const Tuple &t, UnaryFunction f)
+  {
+    typedef typename tuple_meta_transform<Tuple,UnaryMetaFunction>::type XfrmTuple;
+
+    return XfrmTuple(f(thrust::get<Is>(t))...);
+  }
+};
+
+
+template<template<typename> class UnaryMetaFunction,
+         typename Tuple,
+         typename UnaryFunction>
+typename tuple_meta_transform<Tuple,UnaryMetaFunction>::type
+tuple_host_transform(const Tuple &t, UnaryFunction f)
+{
+  return tuple_transform_functor<Tuple,UnaryMetaFunction,UnaryFunction>::do_it_on_the_host(t,f);
+}
+
+template<template<typename> class UnaryMetaFunction,
+         typename Tuple,
+         typename UnaryFunction>
+typename tuple_meta_transform<Tuple,UnaryMetaFunction>::type
+__host__ __device__
+tuple_host_device_transform(const Tuple &t, UnaryFunction f)
+{
+  return tuple_transform_functor<Tuple,UnaryMetaFunction,UnaryFunction>::do_it_on_the_host_or_device(t,f);
+}
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_deduction.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_deduction.h
new file mode 100644
index 000000000000..6f240711d0a1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_deduction.h
@@ -0,0 +1,90 @@
+// Copyright (c)      2018 NVIDIA Corporation
+//                         (Bryce Adelstein Lelbach <brycelelbach@gmail.com>)
+// Copyright (c) 2013-2018 Eric Niebler (`THRUST_RETURNS`, etc)
+// Copyright (c) 2016-2018 Casey Carter (`THRUST_RETURNS`, etc)
+//
+// Distributed under the Boost Software License v1.0 (boost.org/LICENSE_1_0.txt)
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/detail/preprocessor.h>
+
+#include <utility>
+#include <type_traits>
+
+///////////////////////////////////////////////////////////////////////////////
+
+/// \def THRUST_FWD(x)
+/// \brief Performs universal forwarding of a universal reference.
+///
+#define THRUST_FWD(x) ::std::forward<decltype(x)>(x)
+
+/// \def THRUST_MVCAP(x)
+/// \brief Capture `x` into a lambda by moving.
+///
+#define THRUST_MVCAP(x) x = ::std::move(x)
+
+/// \def THRUST_RETOF(invocable, ...)
+/// \brief Expands to the type returned by invoking an instance of the invocable
+///        type \a invocable with parameters of type \c __VA_ARGS__. Must
+///        be called with 1 or fewer parameters to the invocable.
+///
+#define THRUST_RETOF(...)   THRUST_PP_DISPATCH(THRUST_RETOF, __VA_ARGS__)
+#define THRUST_RETOF1(C)    decltype(::std::declval<C>()())
+#define THRUST_RETOF2(C, V) decltype(::std::declval<C>()(::std::declval<V>()))
+
+/// \def THRUST_RETURNS(...)
+/// \brief Expands to a function definition that returns the expression
+///        \c __VA_ARGS__.
+///
+#define THRUST_RETURNS(...)                                                   \
+  noexcept(noexcept(__VA_ARGS__))                                             \
+  { return (__VA_ARGS__); }                                                   \
+  /**/
+
+/// \def THRUST_DECLTYPE_RETURNS(...)
+/// \brief Expands to a function definition, including a trailing returning
+///        type, that returns the expression \c __VA_ARGS__.
+///
+// Trailing return types seem to confuse Doxygen, and cause it to interpret
+// parts of the function's body as new function signatures.
+#if defined(THRUST_DOXYGEN)
+  #define THRUST_DECLTYPE_RETURNS(...)                                        \
+  { return (__VA_ARGS__); }                                                   \
+  /**/
+#else
+  #define THRUST_DECLTYPE_RETURNS(...)                                        \
+    noexcept(noexcept(__VA_ARGS__))                                           \
+    -> decltype(__VA_ARGS__)                                                  \
+    { return (__VA_ARGS__); }                                                 \
+    /**/
+#endif
+
+/// \def THRUST_DECLTYPE_RETURNS_WITH_SFINAE_CONDITION(condition, ...)
+/// \brief Expands to a function definition, including a trailing returning
+///        type, that returns the expression \c __VA_ARGS__. It shall only
+///        participate in overload resolution if \c condition is \c true.
+///
+// Trailing return types seem to confuse Doxygen, and cause it to interpret
+// parts of the function's body as new function signatures.
+#if defined(THRUST_DOXYGEN)
+  #define THRUST_DECLTYPE_RETURNS(...)                                        \
+  { return (__VA_ARGS__); }                                                   \
+  /**/
+#else
+  #define THRUST_DECLTYPE_RETURNS_WITH_SFINAE_CONDITION(condition, ...)       \
+    noexcept(noexcept(__VA_ARGS__))                                           \
+    -> typename std::enable_if<condition, decltype(__VA_ARGS__)>::type        \
+    { return (__VA_ARGS__); }                                                 \
+    /**/
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits.h
new file mode 100644
index 000000000000..f25eaeaf050a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits.h
@@ -0,0 +1,717 @@
+/*
+ *  Copyright 2008-2022 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file type_traits.h
+ *  \brief Temporarily define some type traits
+ *         until nvcc can compile tr1::type_traits.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <cuda/std/type_traits>
+
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declaration of device_reference
+template<typename T> class device_reference;
+
+namespace detail
+{
+ /// helper classes [4.3].
+ template<typename T, T v>
+   struct integral_constant
+   {
+     THRUST_INLINE_INTEGRAL_MEMBER_CONSTANT T value = v;
+
+     typedef T                       value_type;
+     typedef integral_constant<T, v> type;
+
+     // We don't want to switch to std::integral_constant, because we want access
+     // to the C++14 operator(), but we'd like standard traits to interoperate
+     // with our version when tag dispatching.
+     integral_constant() = default;
+
+     integral_constant(integral_constant const&) = default;
+
+     integral_constant& operator=(integral_constant const&) = default;
+
+     constexpr __host__ __device__
+     integral_constant(std::integral_constant<T, v>) noexcept {}
+
+     constexpr __host__ __device__ operator value_type() const noexcept { return value; }
+     constexpr __host__ __device__ value_type operator()() const noexcept { return value; }
+   };
+ 
+ /// typedef for true_type
+ typedef integral_constant<bool, true>  true_type;
+
+ /// typedef for true_type
+ typedef integral_constant<bool, false> false_type;
+
+//template<typename T> struct is_integral : public std::tr1::is_integral<T> {};
+template<typename T> struct is_integral                           : public false_type {};
+template<>           struct is_integral<bool>                     : public true_type {};
+template<>           struct is_integral<char>                     : public true_type {};
+template<>           struct is_integral<signed char>              : public true_type {};
+template<>           struct is_integral<unsigned char>            : public true_type {};
+template<>           struct is_integral<short>                    : public true_type {};
+template<>           struct is_integral<unsigned short>           : public true_type {};
+template<>           struct is_integral<int>                      : public true_type {};
+template<>           struct is_integral<unsigned int>             : public true_type {};
+template<>           struct is_integral<long>                     : public true_type {};
+template<>           struct is_integral<unsigned long>            : public true_type {};
+template<>           struct is_integral<long long>                : public true_type {};
+template<>           struct is_integral<unsigned long long>       : public true_type {};
+template<>           struct is_integral<const bool>               : public true_type {};
+template<>           struct is_integral<const char>               : public true_type {};
+template<>           struct is_integral<const unsigned char>      : public true_type {};
+template<>           struct is_integral<const short>              : public true_type {};
+template<>           struct is_integral<const unsigned short>     : public true_type {};
+template<>           struct is_integral<const int>                : public true_type {};
+template<>           struct is_integral<const unsigned int>       : public true_type {};
+template<>           struct is_integral<const long>               : public true_type {};
+template<>           struct is_integral<const unsigned long>      : public true_type {};
+template<>           struct is_integral<const long long>          : public true_type {};
+template<>           struct is_integral<const unsigned long long> : public true_type {};
+
+template<typename T> struct is_floating_point              : public false_type {};
+template<>           struct is_floating_point<float>       : public true_type {};
+template<>           struct is_floating_point<double>      : public true_type {};
+template<>           struct is_floating_point<long double> : public true_type {};
+
+template<typename T> struct is_arithmetic               : public is_integral<T> {};
+template<>           struct is_arithmetic<float>        : public true_type {};
+template<>           struct is_arithmetic<double>       : public true_type {};
+template<>           struct is_arithmetic<const float>  : public true_type {};
+template<>           struct is_arithmetic<const double> : public true_type {};
+
+template<typename T> struct is_pointer      : public false_type {};
+template<typename T> struct is_pointer<T *> : public true_type  {};
+
+template<typename T> struct is_device_ptr  : public false_type {};
+
+template<typename T> struct is_void             : public false_type {};
+template<>           struct is_void<void>       : public true_type {};
+template<>           struct is_void<const void> : public true_type {};
+
+template<typename T> struct is_non_bool_integral       : public is_integral<T> {};
+template<>           struct is_non_bool_integral<bool> : public false_type {};
+
+template<typename T> struct is_non_bool_arithmetic       : public is_arithmetic<T> {};
+template<>           struct is_non_bool_arithmetic<bool> : public false_type {};
+
+template<typename T> struct is_pod
+   : public integral_constant<
+       bool,
+       is_void<T>::value || is_pointer<T>::value || is_arithmetic<T>::value
+#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC || \
+    THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_CLANG
+// use intrinsic type traits
+       || __is_pod(T)
+#elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC
+// only use the intrinsic for >= 4.3
+#if (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
+       || __is_pod(T)
+#endif // GCC VERSION
+#endif // THRUST_HOST_COMPILER
+     >
+ {};
+
+
+template <typename T> 
+struct has_trivial_constructor
+  : public integral_constant<bool, is_pod<T>::value || ::cuda::std::is_trivially_constructible<T>::value> 
+{};
+
+template<typename T> 
+struct has_trivial_copy_constructor
+  : public integral_constant<bool, is_pod<T>::value || ::cuda::std::is_trivially_copyable<T>::value>
+{};
+
+template<typename T> struct has_trivial_destructor : public is_pod<T> {};
+
+template<typename T> struct is_const          : public false_type {};
+template<typename T> struct is_const<const T> : public true_type {};
+
+template<typename T> struct is_volatile             : public false_type {};
+template<typename T> struct is_volatile<volatile T> : public true_type {};
+
+template<typename T>
+  struct add_const
+{
+  typedef T const type;
+}; // end add_const
+
+template<typename T>
+  struct remove_const
+{
+  typedef T type;
+}; // end remove_const
+
+template<typename T>
+  struct remove_const<const T>
+{
+  typedef T type;
+}; // end remove_const
+
+template<typename T>
+  struct add_volatile
+{
+  typedef volatile T type;
+}; // end add_volatile
+
+template<typename T>
+  struct remove_volatile
+{
+  typedef T type;
+}; // end remove_volatile
+
+template<typename T>
+  struct remove_volatile<volatile T>
+{
+  typedef T type;
+}; // end remove_volatile
+
+template<typename T>
+  struct add_cv
+{
+  typedef const volatile T type;
+}; // end add_cv
+
+template<typename T>
+  struct remove_cv
+{
+  typedef typename remove_const<typename remove_volatile<T>::type>::type type;
+}; // end remove_cv
+
+
+template<typename T> struct is_reference     : public false_type {};
+template<typename T> struct is_reference<T&> : public true_type {};
+
+template<typename T> struct is_proxy_reference  : public false_type {};
+
+template<typename T> struct is_device_reference                                : public false_type {};
+template<typename T> struct is_device_reference< thrust::device_reference<T> > : public true_type {};
+
+
+// NB: Careful with reference to void.
+template<typename _Tp, bool = (is_void<_Tp>::value || is_reference<_Tp>::value)>
+  struct __add_reference_helper
+  { typedef _Tp&    type; };
+
+template<typename _Tp>
+  struct __add_reference_helper<_Tp, true>
+  { typedef _Tp     type; };
+
+template<typename _Tp>
+  struct add_reference
+    : public __add_reference_helper<_Tp>{};
+
+template<typename T>
+  struct remove_reference
+{
+  typedef T type;
+}; // end remove_reference
+
+template<typename T>
+  struct remove_reference<T&>
+{
+  typedef T type;
+}; // end remove_reference
+
+template<typename T1, typename T2>
+  struct is_same
+    : public false_type
+{
+}; // end is_same
+
+template<typename T>
+  struct is_same<T,T>
+    : public true_type
+{
+}; // end is_same
+
+template<typename T1, typename T2>
+  struct lazy_is_same
+    : is_same<typename T1::type, typename T2::type>
+{
+}; // end lazy_is_same
+
+template<typename T1, typename T2>
+  struct is_different
+    : public true_type
+{
+}; // end is_different
+
+template<typename T>
+  struct is_different<T,T>
+    : public false_type
+{
+}; // end is_different
+
+template<typename T1, typename T2>
+  struct lazy_is_different
+    : is_different<typename T1::type, typename T2::type>
+{
+}; // end lazy_is_different
+
+#if THRUST_CPP_DIALECT >= 2011
+
+using std::is_convertible;
+
+#else
+
+namespace tt_detail
+{
+
+template<typename T>
+  struct is_int_or_cref
+{
+  typedef typename remove_reference<T>::type type_sans_ref;
+  static const bool value = (is_integral<T>::value
+                             || (is_integral<type_sans_ref>::value
+                                 && is_const<type_sans_ref>::value
+                                 && !is_volatile<type_sans_ref>::value));
+}; // end is_int_or_cref
+
+
+THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_BEGIN
+THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING_BEGIN
+
+template<typename From, typename To>
+  struct is_convertible_sfinae
+{
+  private:
+    typedef char                          yes;
+    typedef struct { char two_chars[2]; } no;
+
+    static inline yes   test(To) { return yes(); }
+    static inline no    test(...) { return no(); } 
+    static inline typename remove_reference<From>::type& from() { typename remove_reference<From>::type* ptr = 0; return *ptr; }
+
+  public:
+    static const bool value = sizeof(test(from())) == sizeof(yes);
+}; // end is_convertible_sfinae
+
+
+THRUST_DISABLE_MSVC_FORCING_VALUE_TO_BOOL_WARNING_END
+THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_END
+
+
+template<typename From, typename To>
+  struct is_convertible_needs_simple_test
+{
+  static const bool from_is_void      = is_void<From>::value;
+  static const bool to_is_void        = is_void<To>::value;
+  static const bool from_is_float     = is_floating_point<typename remove_reference<From>::type>::value;
+  static const bool to_is_int_or_cref = is_int_or_cref<To>::value;
+
+  static const bool value = (from_is_void || to_is_void || (from_is_float && to_is_int_or_cref));
+}; // end is_convertible_needs_simple_test
+
+
+template<typename From, typename To,
+         bool = is_convertible_needs_simple_test<From,To>::value>
+  struct is_convertible
+{
+  static const bool value = (is_void<To>::value
+                             || (is_int_or_cref<To>::value
+                                 && !is_void<From>::value));
+}; // end is_convertible
+
+
+template<typename From, typename To>
+  struct is_convertible<From, To, false>
+{
+  static const bool value = (is_convertible_sfinae<typename
+                             add_reference<From>::type, To>::value);
+}; // end is_convertible
+
+
+} // end tt_detail
+
+template<typename From, typename To>
+  struct is_convertible
+    : public integral_constant<bool, tt_detail::is_convertible<From, To>::value>
+{
+}; // end is_convertible
+
+#endif
+
+template<typename T1, typename T2>
+  struct is_one_convertible_to_the_other
+    : public integral_constant<
+        bool,
+        is_convertible<T1,T2>::value || is_convertible<T2,T1>::value
+      >
+{};
+
+
+// mpl stuff
+template<typename... Conditions>
+  struct or_;
+
+template <>
+  struct or_<>
+    : public integral_constant<
+        bool,
+        false_type::value  // identity for or_
+      >
+{
+}; // end or_
+
+template <typename Condition, typename... Conditions>
+  struct or_<Condition, Conditions...>
+    : public integral_constant<
+        bool,
+        Condition::value || or_<Conditions...>::value
+      >
+{
+}; // end or_
+
+template <typename... Conditions>
+  struct and_;
+
+template<>
+  struct and_<>
+    : public integral_constant<
+        bool,
+        true_type::value // identity for and_
+      >
+{
+}; // end and_
+
+template <typename Condition, typename... Conditions>
+  struct and_<Condition, Conditions...>
+    : public integral_constant<
+        bool,
+        Condition::value && and_<Conditions...>::value>
+{
+}; // end and_
+
+template <typename Boolean>
+  struct not_
+    : public integral_constant<bool, !Boolean::value>
+{
+}; // end not_
+
+template<bool B, class T, class F>
+struct conditional { typedef T type; };
+ 
+template<class T, class F>
+struct conditional<false, T, F> { typedef F type; };
+
+template <bool, typename Then, typename Else>
+  struct eval_if
+{
+}; // end eval_if
+
+template<typename Then, typename Else>
+  struct eval_if<true, Then, Else>
+{
+  typedef typename Then::type type;
+}; // end eval_if
+
+template<typename Then, typename Else>
+  struct eval_if<false, Then, Else>
+{
+  typedef typename Else::type type;
+}; // end eval_if
+
+template<typename T>
+//  struct identity
+//  XXX WAR nvcc's confusion with thrust::identity
+  struct identity_
+{
+  typedef T type;
+}; // end identity
+
+template<bool, typename T = void> struct enable_if {};
+template<typename T>              struct enable_if<true, T> {typedef T type;};
+
+template<bool, typename T> struct lazy_enable_if {};
+template<typename T>       struct lazy_enable_if<true, T> {typedef typename T::type type;};
+
+template<bool condition, typename T = void> struct disable_if : enable_if<!condition, T> {};
+template<bool condition, typename T>        struct lazy_disable_if : lazy_enable_if<!condition, T> {};
+
+
+template<typename T1, typename T2, typename T = void>
+  struct enable_if_convertible
+    : enable_if< is_convertible<T1,T2>::value, T >
+{};
+
+
+template<typename T1, typename T2, typename T = void>
+  struct disable_if_convertible
+    : disable_if< is_convertible<T1,T2>::value, T >
+{};
+
+
+template<typename T1, typename T2, typename Result = void>
+  struct enable_if_different
+    : enable_if<is_different<T1,T2>::value, Result>
+{};
+
+
+template<typename T>
+  struct is_numeric
+    : and_<
+        is_convertible<int,T>,
+        is_convertible<T,int>
+      >
+{
+}; // end is_numeric
+
+
+template<typename> struct is_reference_to_const             : false_type {};
+template<typename T> struct is_reference_to_const<const T&> : true_type {};
+
+
+// make_unsigned follows
+
+namespace tt_detail
+{
+
+template<typename T> struct make_unsigned_simple;
+
+template<> struct make_unsigned_simple<char>                   { typedef unsigned char          type; };
+template<> struct make_unsigned_simple<signed char>            { typedef unsigned char          type; };
+template<> struct make_unsigned_simple<unsigned char>          { typedef unsigned char          type; };
+template<> struct make_unsigned_simple<short>                  { typedef unsigned short         type; };
+template<> struct make_unsigned_simple<unsigned short>         { typedef unsigned short         type; };
+template<> struct make_unsigned_simple<int>                    { typedef unsigned int           type; };
+template<> struct make_unsigned_simple<unsigned int>           { typedef unsigned int           type; };
+template<> struct make_unsigned_simple<long int>               { typedef unsigned long int      type; };
+template<> struct make_unsigned_simple<unsigned long int>      { typedef unsigned long int      type; };
+template<> struct make_unsigned_simple<long long int>          { typedef unsigned long long int type; };
+template<> struct make_unsigned_simple<unsigned long long int> { typedef unsigned long long int type; };
+
+template<typename T>
+  struct make_unsigned_base
+{
+  // remove cv
+  typedef typename remove_cv<T>::type remove_cv_t;
+
+  // get the simple unsigned type
+  typedef typename make_unsigned_simple<remove_cv_t>::type unsigned_remove_cv_t;
+
+  // add back const, volatile, both, or neither to the simple result
+  typedef typename eval_if<
+    is_const<T>::value && is_volatile<T>::value,
+    // add cv back
+    add_cv<unsigned_remove_cv_t>,
+    // check const & volatile individually
+    eval_if<
+      is_const<T>::value,
+      // add c back
+      add_const<unsigned_remove_cv_t>,
+      eval_if<
+        is_volatile<T>::value,
+        // add v back
+        add_volatile<unsigned_remove_cv_t>,
+        // original type was neither cv, return the simple unsigned result
+        identity_<unsigned_remove_cv_t>
+      >
+    >
+  >::type type;
+};
+
+} // end tt_detail
+
+template<typename T>
+  struct make_unsigned
+    : tt_detail::make_unsigned_base<T>
+{};
+
+struct largest_available_float
+{
+  typedef double type;
+};
+
+// T1 wins if they are both the same size
+template<typename T1, typename T2>
+  struct larger_type
+    : thrust::detail::eval_if<
+        (sizeof(T2) > sizeof(T1)),
+        thrust::detail::identity_<T2>,
+        thrust::detail::identity_<T1>
+      >
+{};
+
+#if THRUST_CPP_DIALECT >= 2011
+
+using std::is_base_of;
+
+#else
+
+namespace is_base_of_ns
+{
+
+typedef char                          yes;
+typedef struct { char two_chars[2]; } no;
+
+template<typename Base, typename Derived>
+  struct host
+{
+  operator Base*() const;
+  operator Derived*();
+}; // end host
+
+template<typename Base, typename Derived>
+  struct impl
+{
+  template<typename T> static yes check(Derived *, T);
+  static no check(Base*, int);
+
+  static const bool value = sizeof(check(host<Base,Derived>(), int())) == sizeof(yes);
+}; // end impl
+
+} // end is_base_of_ns
+
+
+template<typename Base, typename Derived>
+  struct is_base_of
+    : integral_constant<
+        bool,
+        is_base_of_ns::impl<Base,Derived>::value
+      >
+{};
+
+#endif
+
+template<typename Base, typename Derived, typename Result = void>
+  struct enable_if_base_of
+    : enable_if<
+        is_base_of<Base,Derived>::value,
+        Result
+      >
+{};
+
+
+namespace is_assignable_ns
+{
+
+template<typename T1, typename T2>
+  class is_assignable
+{
+  typedef char                      yes_type;
+  typedef struct { char array[2]; } no_type;
+
+  template<typename T> static typename add_reference<T>::type declval();
+  
+  template<size_t> struct helper { typedef void * type; };
+
+  template<typename U1, typename U2> static yes_type test(typename helper<sizeof(declval<U1>() = declval<U2>())>::type);
+
+  template<typename,typename> static no_type test(...);
+
+  public:
+    static const bool value = sizeof(test<T1,T2>(0)) == 1;
+}; // end is_assignable
+
+} // end is_assignable_ns
+
+
+template<typename T1, typename T2>
+  struct is_assignable
+    : integral_constant<
+        bool,
+        is_assignable_ns::is_assignable<T1,T2>::value
+      >
+{};
+
+
+template<typename T>
+  struct is_copy_assignable
+    : is_assignable<
+        typename add_reference<T>::type,
+        typename add_reference<typename add_const<T>::type>::type
+      >
+{};
+
+
+template<typename T1, typename T2, typename Enable = void> struct promoted_numerical_type;
+
+template<typename T1, typename T2> 
+  struct promoted_numerical_type<T1,T2,typename enable_if<and_
+  <typename is_floating_point<T1>::type,typename is_floating_point<T2>::type>
+  ::value>::type>
+  {
+  typedef typename larger_type<T1,T2>::type type;
+  };
+
+template<typename T1, typename T2> 
+  struct promoted_numerical_type<T1,T2,typename enable_if<and_
+  <typename is_integral<T1>::type,typename is_floating_point<T2>::type>
+  ::value>::type>
+  {
+  typedef T2 type;
+  };
+
+template<typename T1, typename T2>
+  struct promoted_numerical_type<T1,T2,typename enable_if<and_
+  <typename is_floating_point<T1>::type, typename is_integral<T2>::type>
+  ::value>::type>
+  {
+  typedef T1 type;
+  };
+
+template<typename T>
+  struct is_empty_helper : public T
+  {
+  };
+
+struct is_empty_helper_base
+{
+};
+
+template<typename T>
+  struct is_empty : integral_constant<bool,
+    sizeof(is_empty_helper_base) == sizeof(is_empty_helper<T>)
+  >
+  {
+  };
+
+template <typename Invokable, typename... Args>
+using invoke_result_t =
+#if THRUST_CPP_DIALECT < 2017
+  typename ::cuda::std::result_of<Invokable(Args...)>::type;
+#else // 2017+
+  ::cuda::std::invoke_result_t<Invokable, Args...>;
+#endif
+
+template <class F, class... Us> 
+struct invoke_result
+{
+  using type = invoke_result_t<F, Us...>;
+};
+
+} // end detail
+
+using detail::integral_constant;
+using detail::true_type;
+using detail::false_type;
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/type_traits/has_trivial_assign.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/function_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/function_traits.h
new file mode 100644
index 000000000000..1098201366ef
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/function_traits.h
@@ -0,0 +1,97 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/has_nested_type.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward definitions for is_commutative
+template <typename T> struct plus;
+template <typename T> struct multiplies;
+template <typename T> struct minimum;
+template <typename T> struct maximum;
+template <typename T> struct logical_or;
+template <typename T> struct logical_and;
+template <typename T> struct bit_or;
+template <typename T> struct bit_and;
+template <typename T> struct bit_xor;
+
+namespace detail
+{
+
+
+// some metafunctions which check for the nested types of the adaptable functions
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_result_type, result_type)
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_argument_type, argument_type)
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_first_argument_type, first_argument_type)
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_second_argument_type, second_argument_type)
+
+
+template<typename AdaptableBinaryFunction>
+  struct result_type
+{
+  typedef typename AdaptableBinaryFunction::result_type type;
+};
+
+
+template<typename T>
+  struct is_adaptable_unary_function
+    : thrust::detail::and_<
+        has_result_type<T>,
+        has_argument_type<T>
+      >
+{};
+
+
+template<typename T>
+  struct is_adaptable_binary_function
+    : thrust::detail::and_<
+        has_result_type<T>,
+        thrust::detail::and_<
+          has_first_argument_type<T>,
+          has_second_argument_type<T>
+        >
+      >
+{};
+
+
+template<typename BinaryFunction>
+  struct is_commutative
+    : public thrust::detail::false_type
+{};
+
+template<typename T> struct is_commutative< typename thrust::plus<T>        > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::multiplies<T>  > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::minimum<T>     > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::maximum<T>     > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::logical_or<T>  > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::logical_and<T> > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::bit_or<T>      > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::bit_and<T>     > : public thrust::detail::is_arithmetic<T> {};
+template<typename T> struct is_commutative< typename thrust::bit_xor<T>     > : public thrust::detail::is_arithmetic<T> {};
+
+} // end namespace detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_member_function.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_member_function.h
new file mode 100644
index 000000000000..c33fe28f6545
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_member_function.h
@@ -0,0 +1,38 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/type_traits.h>
+
+#include <utility> // for std::declval
+
+#define __THRUST_DEFINE_HAS_MEMBER_FUNCTION(trait_name, member_function_name)  \
+  template <typename T, typename Signature, typename = void>                   \
+  struct trait_name : thrust::false_type                                       \
+  {};                                                                          \
+                                                                               \
+  template <typename T, typename ResultT, typename... Args>                    \
+  struct trait_name<T,                                                         \
+                    ResultT(Args...),                                          \
+                    typename thrust::detail::enable_if<                        \
+                      thrust::detail::is_same<ResultT, void>::value ||         \
+                      thrust::detail::is_convertible<                          \
+                        ResultT,                                               \
+                        decltype(std::declval<T>().member_function_name(       \
+                          std::declval<Args>()...))>::value>::type>            \
+      : thrust::true_type                                                      \
+  {};
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_nested_type.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_nested_type.h
new file mode 100644
index 000000000000..78bb4b7f57a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_nested_type.h
@@ -0,0 +1,32 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/type_traits.h>
+
+#define __THRUST_DEFINE_HAS_NESTED_TYPE(trait_name, nested_type_name) \
+template<typename T> \
+  struct trait_name  \
+{                    \
+  typedef char yes_type; \
+  typedef int  no_type;  \
+  template<typename S> static yes_type test(typename S::nested_type_name *); \
+  template<typename S> static no_type  test(...); \
+  static bool const value = sizeof(test<T>(0)) == sizeof(yes_type);\
+  typedef thrust::detail::integral_constant<bool, value> type;\
+};
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_trivial_assign.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_trivial_assign.h
new file mode 100644
index 000000000000..7222ce593f51
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/has_trivial_assign.h
@@ -0,0 +1,47 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file type_traits.h
+ *  \brief Temporarily define some type traits
+ *         until nvcc can compile tr1::type_traits.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+
+#include <cuda/std/type_traits>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<typename T> 
+struct has_trivial_assign
+  : public integral_constant<
+      bool, 
+      (is_pod<T>::value && !is_const<T>::value) 
+      || ::cuda::std::is_trivially_copy_assignable<T>::value
+    >
+{};
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_call_possible.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_call_possible.h
new file mode 100644
index 000000000000..58c1aca4df0f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_call_possible.h
@@ -0,0 +1,162 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/has_member_function.h>
+
+// inspired by Roman Perepelitsa's presentation from comp.lang.c++.moderated
+// based on the implementation here: http://www.rsdn.ru/forum/cpp/2759773.1.aspx
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace is_call_possible_detail
+{
+
+template<typename T> class void_exp_result {}; 
+
+template<typename T, typename U> 
+U const& operator,(U const&, void_exp_result<T>); 
+
+template<typename T, typename U> 
+U& operator,(U&, void_exp_result<T>); 
+
+template<typename src_type, typename dest_type> 
+struct clone_constness 
+{
+  typedef dest_type type; 
+}; 
+
+template<typename src_type, typename dest_type> 
+struct clone_constness<const src_type, dest_type> 
+{ 
+  typedef const dest_type type; 
+};
+
+} // end is_call_possible_detail
+} // end detail
+THRUST_NAMESPACE_END
+
+#define __THRUST_DEFINE_IS_CALL_POSSIBLE(trait_name, member_function_name)                                                                \
+__THRUST_DEFINE_HAS_MEMBER_FUNCTION(trait_name##_has_member, member_function_name)                                                        \
+                                                                                                                                          \
+template <typename T, typename Signature>                                                                                                 \
+struct trait_name                                                                                                                         \
+{                                                                                                                                         \
+  private:                                                                                                                                \
+    struct yes {};                                                                                                                        \
+    struct no { yes m[2]; };                                                                                                              \
+    struct derived : public T                                                                                                             \
+    {                                                                                                                                     \
+      using T::member_function_name;                                                                                                      \
+      no member_function_name(...) const;                                                                                                 \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    typedef typename thrust::detail::is_call_possible_detail::clone_constness<T, derived>::type derived_type;                             \
+                                                                                                                                          \
+    template<typename U, typename Result>                                                                                                 \
+    struct return_value_check                                                                                                             \
+    {                                                                                                                                     \
+      static yes deduce(Result);                                                                                                          \
+      static no deduce(...);                                                                                                              \
+      static no deduce(no);                                                                                                               \
+      static no deduce(thrust::detail::is_call_possible_detail::void_exp_result<T>);                                                      \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    template<typename U>                                                                                                                  \
+    struct return_value_check<U, void>                                                                                                    \
+    {                                                                                                                                     \
+      static yes deduce(...);                                                                                                             \
+      static no deduce(no);                                                                                                               \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    template<bool has_the_member_of_interest, typename F>                                                                                 \
+    struct impl                                                                                                                           \
+    {                                                                                                                                     \
+      static const bool value = false;                                                                                                    \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    template<typename Result, typename Arg>                                                                                               \
+    struct impl<true, Result(Arg)>                                                                                                        \
+    {                                                                                                                                     \
+      static typename add_reference<derived_type>::type test_me;                                                                          \
+      static typename add_reference<Arg>::type          arg;                                                                              \
+                                                                                                                                          \
+      static const bool value =                                                                                                           \
+        sizeof(                                                                                                                           \
+                return_value_check<T, Result>::deduce(                                                                                    \
+                  (test_me.member_function_name(arg), thrust::detail::is_call_possible_detail::void_exp_result<T>())                      \
+                )                                                                                                                         \
+              ) == sizeof(yes);                                                                                                           \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    template<typename Result, typename Arg1, typename Arg2>                                                                               \
+    struct impl<true, Result(Arg1,Arg2)>                                                                                                  \
+    {                                                                                                                                     \
+      static typename add_reference<derived_type>::type test_me;                                                                          \
+      static typename add_reference<Arg1>::type         arg1;                                                                             \
+      static typename add_reference<Arg2>::type         arg2;                                                                             \
+                                                                                                                                          \
+      static const bool value =                                                                                                           \
+        sizeof(                                                                                                                           \
+                return_value_check<T, Result>::deduce(                                                                                    \
+                  (test_me.member_function_name(arg1,arg2), thrust::detail::is_call_possible_detail::void_exp_result<T>())                \
+                )                                                                                                                         \
+              ) == sizeof(yes);                                                                                                           \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    template<typename Result, typename Arg1, typename Arg2, typename Arg3>                                                                \
+    struct impl<true, Result(Arg1,Arg2,Arg3)>                                                                                             \
+    {                                                                                                                                     \
+      static typename add_reference<derived_type>::type test_me;                                                                          \
+      static typename add_reference<Arg1>::type         arg1;                                                                             \
+      static typename add_reference<Arg2>::type         arg2;                                                                             \
+      static typename add_reference<Arg3>::type         arg3;                                                                             \
+                                                                                                                                          \
+      static const bool value =                                                                                                           \
+        sizeof(                                                                                                                           \
+                return_value_check<T, Result>::deduce(                                                                                    \
+                  (test_me.member_function_name(arg1,arg2,arg3), thrust::detail::is_call_possible_detail::void_exp_result<T>())           \
+                )                                                                                                                         \
+              ) == sizeof(yes);                                                                                                           \
+    };                                                                                                                                    \
+                                                                                                                                          \
+    template<typename Result, typename Arg1, typename Arg2, typename Arg3, typename Arg4>                                                 \
+    struct impl<true, Result(Arg1,Arg2,Arg3,Arg4)>                                                                                        \
+    {                                                                                                                                     \
+      static typename add_reference<derived_type>::type test_me;                                                                          \
+      static typename add_reference<Arg1>::type         arg1;                                                                             \
+      static typename add_reference<Arg2>::type         arg2;                                                                             \
+      static typename add_reference<Arg3>::type         arg3;                                                                             \
+      static typename add_reference<Arg4>::type         arg4;                                                                             \
+                                                                                                                                          \
+      static const bool value =                                                                                                           \
+        sizeof(                                                                                                                           \
+                return_value_check<T, Result>::deduce(                                                                                    \
+                  (test_me.member_function_name(arg1,arg2,arg3,arg4), thrust::detail::is_call_possible_detail::void_exp_result<T>())      \
+                )                                                                                                                         \
+              ) == sizeof(yes);                                                                                                           \
+    };                                                                                                                                    \
+                                                                                                                                          \
+  public:                                                                                                                                 \
+    static const bool value = impl<trait_name##_has_member<T,Signature>::value, Signature>::value;                                        \
+    typedef thrust::detail::integral_constant<bool,value> type;                                                                           \
+}; 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_metafunction_defined.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_metafunction_defined.h
new file mode 100644
index 000000000000..2c7a4be52df6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/is_metafunction_defined.h
@@ -0,0 +1,42 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits/has_nested_type.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(is_metafunction_defined, type)
+
+template<typename Metafunction>
+  struct enable_if_defined
+    : thrust::detail::lazy_enable_if<
+        is_metafunction_defined<Metafunction>::value,
+        Metafunction
+      >
+{};
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h
new file mode 100644
index 000000000000..210409d62690
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_discard_iterator.h
@@ -0,0 +1,39 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/discard_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template <typename Iterator>
+struct is_discard_iterator
+  : public thrust::detail::false_type
+{};
+
+template <typename System>
+struct is_discard_iterator< thrust::discard_iterator<System> >
+ : public thrust::detail::true_type
+{};
+
+} // end namespace detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h
new file mode 100644
index 000000000000..555b67400cd6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/iterator/is_output_iterator.h
@@ -0,0 +1,65 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/is_metafunction_defined.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/any_assign.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+
+template<typename T>
+  struct is_void_like
+    : thrust::detail::or_<
+        thrust::detail::is_void<T>,
+        thrust::detail::is_same<T,thrust::detail::any_assign>
+      >
+{}; // end is_void_like
+
+
+template<typename T>
+  struct lazy_is_void_like
+    : is_void_like<typename T::type>
+{}; // end lazy_is_void_like
+
+
+// XXX this meta function should first check that T is actually an iterator
+//
+//     if thrust::iterator_value<T> is defined and thrust::iterator_value<T>::type == void
+//       return false
+//     else
+//       return true
+template<typename T>
+  struct is_output_iterator
+    : eval_if<
+        is_metafunction_defined<thrust::iterator_value<T> >::value,
+        lazy_is_void_like<thrust::iterator_value<T> >,
+        thrust::detail::true_type
+      >::type
+{
+}; // end is_output_iterator
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/minimum_type.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/minimum_type.h
new file mode 100644
index 000000000000..2417e327df8e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/minimum_type.h
@@ -0,0 +1,163 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{ 
+
+namespace minimum_type_detail
+{
+
+//
+// Returns the minimum type or is empty
+// if T1 and T2 are unrelated.
+//
+template <typename T1, typename T2, bool GreaterEqual, bool LessEqual> struct minimum_type_impl {};
+  
+template <typename T1, typename T2>
+struct minimum_type_impl<T1,T2,true,false>
+{
+  typedef T2 type;
+}; // end minimum_type_impl
+
+template <typename T1, typename T2>
+struct minimum_type_impl<T1,T2,false,true>
+{
+  typedef T1 type;
+}; // end minimum_type_impl
+
+template <typename T1, typename T2>
+struct minimum_type_impl<T1,T2,true,true>
+{
+  typedef T1 type;
+}; // end minimum_type_impl
+
+template <typename T1, typename T2>
+struct primitive_minimum_type
+  : minimum_type_detail::minimum_type_impl<
+      T1,
+      T2,
+      THRUST_NS_QUALIFIER::detail::is_convertible<T1,T2>::value,
+      THRUST_NS_QUALIFIER::detail::is_convertible<T2,T1>::value
+    >
+{
+}; // end primitive_minimum_type
+
+// because some types are not convertible (even to themselves)
+// specialize primitive_minimum_type for when both types are identical
+template <typename T>
+struct primitive_minimum_type<T,T>
+{
+  typedef T type;
+}; // end primitive_minimum_type
+
+// XXX this belongs somewhere more general
+struct any_conversion
+{
+  template<typename T> operator T (void);
+};
+
+} // end minimum_type_detail
+
+template<typename T1,
+         typename T2  = minimum_type_detail::any_conversion,
+         typename T3  = minimum_type_detail::any_conversion,
+         typename T4  = minimum_type_detail::any_conversion,
+         typename T5  = minimum_type_detail::any_conversion,
+         typename T6  = minimum_type_detail::any_conversion,
+         typename T7  = minimum_type_detail::any_conversion,
+         typename T8  = minimum_type_detail::any_conversion,
+         typename T9  = minimum_type_detail::any_conversion,
+         typename T10 = minimum_type_detail::any_conversion,
+         typename T11 = minimum_type_detail::any_conversion,
+         typename T12 = minimum_type_detail::any_conversion,
+         typename T13 = minimum_type_detail::any_conversion,
+         typename T14 = minimum_type_detail::any_conversion,
+         typename T15 = minimum_type_detail::any_conversion,
+         typename T16 = minimum_type_detail::any_conversion>
+  struct minimum_type;
+
+// base case
+template<typename T1, typename T2>
+  struct minimum_type<T1,T2>
+    : minimum_type_detail::primitive_minimum_type<T1,T2>
+{};
+
+template<typename T1, typename T2>
+  struct lazy_minimum_type
+    : minimum_type<
+        typename T1::type,
+        typename T2::type
+      >
+{};
+
+// carefully avoid referring to a nested ::type which may not exist
+template<typename T1,  typename T2,  typename T3,  typename T4,
+         typename T5,  typename T6,  typename T7,  typename T8,
+         typename T9,  typename T10, typename T11, typename T12,
+         typename T13, typename T14, typename T15, typename T16>
+  struct minimum_type
+    : lazy_minimum_type<
+        lazy_minimum_type<
+          lazy_minimum_type<
+            minimum_type<
+              T1,T2
+            >,
+            minimum_type<
+              T3,T4
+            >
+          >,
+          lazy_minimum_type<
+            minimum_type<
+              T5,T6
+            >,
+            minimum_type<
+              T7,T8
+            >
+          >
+        >,
+        lazy_minimum_type<
+          lazy_minimum_type<
+            minimum_type<
+              T9,T10
+            >,
+            minimum_type<
+              T11,T12
+            >
+          >,
+          lazy_minimum_type<
+            minimum_type<
+              T13,T14
+            >,
+            minimum_type<
+              T15,T16
+            >
+          >
+        >
+      >
+{};
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/pointer_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/pointer_traits.h
new file mode 100644
index 000000000000..90a8bc29dec3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/pointer_traits.h
@@ -0,0 +1,395 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/is_metafunction_defined.h>
+#include <thrust/detail/type_traits/has_nested_type.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <cstddef>
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template<typename Ptr> struct pointer_element;
+
+template<template<typename> class Ptr, typename Arg>
+  struct pointer_element<Ptr<Arg> >
+{
+  typedef Arg type;
+};
+
+template<template<typename,typename> class Ptr, typename Arg1, typename Arg2>
+  struct pointer_element<Ptr<Arg1,Arg2> >
+{
+  typedef Arg1 type;
+};
+
+template<template<typename,typename,typename> class Ptr, typename Arg1, typename Arg2, typename Arg3>
+  struct pointer_element<Ptr<Arg1,Arg2,Arg3> >
+{
+  typedef Arg1 type;
+};
+
+template<template<typename,typename,typename,typename> class Ptr, typename Arg1, typename Arg2, typename Arg3, typename Arg4>
+  struct pointer_element<Ptr<Arg1,Arg2,Arg3,Arg4> >
+{
+  typedef Arg1 type;
+};
+
+template<template<typename,typename,typename,typename,typename> class Ptr, typename Arg1, typename Arg2, typename Arg3, typename Arg4, typename Arg5>
+  struct pointer_element<Ptr<Arg1,Arg2,Arg3,Arg4,Arg5> >
+{
+  typedef Arg1 type;
+};
+
+template<typename T>
+  struct pointer_element<T*>
+{
+  typedef T type;
+};
+
+template<typename Ptr>
+  struct pointer_difference
+{
+  typedef typename Ptr::difference_type type;
+};
+
+template<typename T>
+  struct pointer_difference<T*>
+{
+  typedef std::ptrdiff_t type;
+};
+
+template<typename Ptr, typename T> struct rebind_pointer;
+
+template<typename T, typename U>
+  struct rebind_pointer<T*,U>
+{
+  using type = U*;
+};
+
+// Rebind generic fancy pointers.
+template<template<typename, typename...> class Ptr, typename OldT, typename... Tail, typename T>
+  struct rebind_pointer<Ptr<OldT,Tail...>,T>
+{
+  using type = Ptr<T,Tail...>;
+};
+
+// Rebind `thrust::pointer`-like things with `thrust::reference`-like references.
+template<template<typename, typename, typename, typename...> class Ptr, typename OldT, typename Tag,
+         template<typename...> class Ref, typename... RefTail,
+         typename... PtrTail, typename T>
+  struct rebind_pointer<Ptr<OldT,Tag,Ref<OldT,RefTail...>,PtrTail...>,T>
+{
+//  static_assert(std::is_same<OldT, Tag>::value, "0");
+  using type = Ptr<T,Tag,Ref<T,RefTail...>,PtrTail...>;
+};
+
+// Rebind `thrust::pointer`-like things with `thrust::reference`-like references
+// and templated derived types.
+template<template<typename, typename, typename, typename...> class Ptr, typename OldT, typename Tag,
+         template<typename...> class Ref, typename... RefTail,
+         template<typename...> class DerivedPtr, typename... DerivedPtrTail,
+         typename T>
+  struct rebind_pointer<Ptr<OldT,Tag,Ref<OldT,RefTail...>,DerivedPtr<OldT,DerivedPtrTail...>>,T>
+{
+//  static_assert(std::is_same<OldT, Tag>::value, "1");
+  using type = Ptr<T,Tag,Ref<T,RefTail...>,DerivedPtr<T,DerivedPtrTail...>>;
+};
+
+// Rebind `thrust::pointer`-like things with native reference types.
+template<template<typename, typename, typename, typename...> class Ptr, typename OldT, typename Tag,
+         typename... PtrTail, typename T>
+  struct rebind_pointer<Ptr<OldT,Tag,typename std::add_lvalue_reference<OldT>::type,PtrTail...>,T>
+{
+//  static_assert(std::is_same<OldT, Tag>::value, "2");
+  using type = Ptr<T,Tag,typename std::add_lvalue_reference<T>::type,PtrTail...>;
+};
+
+// Rebind `thrust::pointer`-like things with native reference types and templated
+// derived types.
+template<template<typename, typename, typename, typename...> class Ptr, typename OldT, typename Tag,
+         template<typename...> class DerivedPtr, typename... DerivedPtrTail,
+         typename T>
+  struct rebind_pointer<Ptr<OldT,Tag,typename std::add_lvalue_reference<OldT>::type,DerivedPtr<OldT,DerivedPtrTail...>>,T>
+{
+//  static_assert(std::is_same<OldT, Tag>::value, "3");
+  using type = Ptr<T,Tag,typename std::add_lvalue_reference<T>::type,DerivedPtr<T,DerivedPtrTail...>>;
+};
+
+__THRUST_DEFINE_HAS_NESTED_TYPE(has_raw_pointer, raw_pointer)
+
+namespace pointer_traits_detail
+{
+
+template<typename Ptr, typename Enable = void> struct pointer_raw_pointer_impl {};
+
+template<typename T>
+  struct pointer_raw_pointer_impl<T*>
+{
+  typedef T* type;
+};
+
+template<typename Ptr>
+  struct pointer_raw_pointer_impl<Ptr, typename enable_if<has_raw_pointer<Ptr>::value>::type>
+{
+  typedef typename Ptr::raw_pointer type;
+};
+
+} // end pointer_traits_detail
+
+template<typename T>
+  struct pointer_raw_pointer
+    : pointer_traits_detail::pointer_raw_pointer_impl<T>
+{};
+
+namespace pointer_traits_detail
+{
+
+template<typename Void>
+  struct capture_address
+{
+  template<typename T>
+  __host__ __device__
+  capture_address(T &r)
+    : m_addr(&r)
+  {}
+
+  inline __host__ __device__
+  Void *operator&() const
+  {
+    return m_addr;
+  }
+
+  Void *m_addr;
+};
+
+// metafunction to compute the type of pointer_to's parameter below
+template<typename T>
+  struct pointer_to_param
+    : thrust::detail::eval_if<
+        thrust::detail::is_void<T>::value,
+        thrust::detail::identity_<capture_address<T> >,
+        thrust::detail::add_reference<T>
+      >
+{};
+
+}
+
+template<typename Ptr>
+  struct pointer_traits
+{
+  typedef Ptr                                    pointer;
+  typedef typename Ptr::reference                reference;
+  typedef typename pointer_element<Ptr>::type    element_type;
+  typedef typename pointer_difference<Ptr>::type difference_type;
+
+  template<typename U>
+    struct rebind
+  {
+    typedef typename rebind_pointer<Ptr,U>::type other;
+  };
+
+  __host__ __device__
+  inline static pointer pointer_to(typename pointer_traits_detail::pointer_to_param<element_type>::type r)
+  {
+    // XXX this is supposed to be pointer::pointer_to(&r); (i.e., call a static member function of pointer called pointer_to)
+    //     assume that pointer has a constructor from raw pointer instead
+
+    return pointer(&r);
+  }
+
+  // thrust additions follow
+  typedef typename pointer_raw_pointer<Ptr>::type raw_pointer;
+
+  __host__ __device__
+  inline static raw_pointer get(pointer ptr)
+  {
+    return ptr.get();
+  }
+};
+
+template<typename T>
+  struct pointer_traits<T*>
+{
+  typedef T*                                    pointer;
+  typedef T&                                    reference;
+  typedef T                                     element_type;
+  typedef typename pointer_difference<T*>::type difference_type;
+
+  template<typename U>
+    struct rebind
+  {
+    typedef U* other;
+  };
+
+  __host__ __device__
+  inline static pointer pointer_to(typename pointer_traits_detail::pointer_to_param<element_type>::type r)
+  {
+    return &r;
+  }
+
+  // thrust additions follow
+  typedef typename pointer_raw_pointer<T*>::type raw_pointer;
+
+  __host__ __device__
+  inline static raw_pointer get(pointer ptr)
+  {
+    return ptr;
+  }
+};
+
+template<>
+  struct pointer_traits<void*>
+{
+  typedef void*                                    pointer;
+  typedef void                                     reference;
+  typedef void                                     element_type;
+  typedef pointer_difference<void*>::type          difference_type;
+
+  template<typename U>
+    struct rebind
+  {
+    typedef U* other;
+  };
+
+  __host__ __device__
+  inline static pointer pointer_to(pointer_traits_detail::pointer_to_param<element_type>::type r)
+  {
+    return &r;
+  }
+
+  // thrust additions follow
+  typedef pointer_raw_pointer<void*>::type raw_pointer;
+
+  __host__ __device__
+  inline static raw_pointer get(pointer ptr)
+  {
+    return ptr;
+  }
+};
+
+template<>
+  struct pointer_traits<const void*>
+{
+  typedef const void*                           pointer;
+  typedef const void                            reference;
+  typedef const void                            element_type;
+  typedef pointer_difference<const void*>::type difference_type;
+
+  template<typename U>
+    struct rebind
+  {
+    typedef U* other;
+  };
+
+  __host__ __device__
+  inline static pointer pointer_to(pointer_traits_detail::pointer_to_param<element_type>::type r)
+  {
+    return &r;
+  }
+
+  // thrust additions follow
+  typedef pointer_raw_pointer<const void*>::type raw_pointer;
+
+  __host__ __device__
+  inline static raw_pointer get(pointer ptr)
+  {
+    return ptr;
+  }
+};
+
+template<typename FromPtr, typename ToPtr>
+  struct is_pointer_system_convertible
+    : thrust::detail::is_convertible<
+        typename iterator_system<FromPtr>::type,
+        typename iterator_system<ToPtr>::type
+      >
+{};
+
+template<typename FromPtr, typename ToPtr>
+  struct is_pointer_convertible
+    : thrust::detail::and_<
+        thrust::detail::is_convertible<
+          typename pointer_element<FromPtr>::type *,
+          typename pointer_element<ToPtr>::type *
+        >,
+        is_pointer_system_convertible<FromPtr, ToPtr>
+      >
+{};
+
+template<typename FromPtr, typename ToPtr>
+  struct is_void_pointer_system_convertible
+    : thrust::detail::and_<
+        thrust::detail::is_same<
+          typename pointer_element<FromPtr>::type,
+          void
+        >,
+        is_pointer_system_convertible<FromPtr, ToPtr>
+      >
+{};
+
+// this could be a lot better, but for our purposes, it's probably
+// sufficient just to check if pointer_raw_pointer<T> has meaning
+template<typename T>
+  struct is_thrust_pointer
+    : is_metafunction_defined<pointer_raw_pointer<T> >
+{};
+
+// avoid inspecting traits of the arguments if they aren't known to be pointers
+template<typename FromPtr, typename ToPtr>
+  struct lazy_is_pointer_convertible
+    : thrust::detail::eval_if<
+        is_thrust_pointer<FromPtr>::value && is_thrust_pointer<ToPtr>::value,
+        is_pointer_convertible<FromPtr,ToPtr>,
+        thrust::detail::identity_<thrust::detail::false_type>
+      >
+{};
+
+template<typename FromPtr, typename ToPtr>
+  struct lazy_is_void_pointer_system_convertible
+    : thrust::detail::eval_if<
+        is_thrust_pointer<FromPtr>::value && is_thrust_pointer<ToPtr>::value,
+        is_void_pointer_system_convertible<FromPtr,ToPtr>,
+        thrust::detail::identity_<thrust::detail::false_type>
+      >
+{};
+
+template<typename FromPtr, typename ToPtr, typename T = void>
+  struct enable_if_pointer_is_convertible
+    : thrust::detail::enable_if<
+        lazy_is_pointer_convertible<FromPtr,ToPtr>::type::value,
+        T
+      >
+{};
+
+template<typename FromPtr, typename ToPtr, typename T = void>
+  struct enable_if_void_pointer_is_system_convertible
+    : thrust::detail::enable_if<
+        lazy_is_void_pointer_system_convertible<FromPtr,ToPtr>::type::value,
+        T
+      >
+{};
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/result_of_adaptable_function.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/result_of_adaptable_function.h
new file mode 100644
index 000000000000..edf797f141bc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/type_traits/result_of_adaptable_function.h
@@ -0,0 +1,59 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/function_traits.h>
+
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+// Sets `type` to the result of the specified Signature invocation. If the
+// callable defines a `result_type` alias member, that type is used instead.
+// Use invoke_result / result_of when FuncType::result_type is not defined.
+template <typename Signature, typename Enable = void>
+struct result_of_adaptable_function
+{
+private:
+  template <typename Sig> struct impl;
+
+  template <typename F, typename... Args>
+  struct impl<F(Args...)>
+  {
+    using type = invoke_result_t<F, Args...>;
+  };
+
+public:
+  using type = typename impl<Signature>::type;
+};
+
+// specialization for invocations which define result_type
+template <typename Functor, typename... ArgTypes>
+struct result_of_adaptable_function<
+  Functor(ArgTypes...),
+  typename thrust::detail::enable_if<
+    thrust::detail::has_result_type<Functor>::value>::type>
+{
+  using type = typename Functor::result_type;
+};
+
+} // namespace detail
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_copy.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_copy.inl
new file mode 100644
index 000000000000..2778693ad3cd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_copy.inl
@@ -0,0 +1,95 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/uninitialized_copy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/uninitialized_copy.h>
+#include <thrust/system/detail/adl/uninitialized_copy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                     InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result)
+{
+  using thrust::system::detail::generic::uninitialized_copy;
+  return uninitialized_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, result);
+} // end uninitialized_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator, typename Size, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator first,
+                                       Size n,
+                                       ForwardIterator result)
+{
+  using thrust::system::detail::generic::uninitialized_copy_n;
+  return uninitialized_copy_n(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, n, result);
+} // end uninitialized_copy_n()
+
+
+template<typename InputIterator,
+         typename ForwardIterator>
+  ForwardIterator uninitialized_copy(InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type   System1;
+  typedef typename thrust::iterator_system<ForwardIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::uninitialized_copy(select_system(system1,system2), first, last, result);
+} // end uninitialized_copy()
+
+
+template<typename InputIterator,
+         typename Size,
+         typename ForwardIterator>
+  ForwardIterator uninitialized_copy_n(InputIterator first,
+                                       Size n,
+                                       ForwardIterator result)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type   System1;
+  typedef typename thrust::iterator_system<ForwardIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::uninitialized_copy_n(select_system(system1,system2), first, n, result);
+} // end uninitialized_copy_n()
+
+
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_fill.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_fill.inl
new file mode 100644
index 000000000000..e013dac7b078
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/uninitialized_fill.inl
@@ -0,0 +1,90 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/uninitialized_fill.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/uninitialized_fill.h>
+#include <thrust/system/detail/adl/uninitialized_fill.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void uninitialized_fill(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x)
+{
+  using thrust::system::detail::generic::uninitialized_fill;
+  return uninitialized_fill(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, x);
+} // end uninitialized_fill()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename ForwardIterator, typename Size, typename T>
+__host__ __device__
+  ForwardIterator uninitialized_fill_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       ForwardIterator first,
+                                       Size n,
+                                       const T &x)
+{
+  using thrust::system::detail::generic::uninitialized_fill_n;
+  return uninitialized_fill_n(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, n, x);
+} // end uninitialized_fill_n()
+
+
+template<typename ForwardIterator,
+         typename T>
+  void uninitialized_fill(ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  thrust::uninitialized_fill(select_system(system), first, last, x);
+} // end uninitialized_fill()
+
+
+template<typename ForwardIterator,
+         typename Size,
+         typename T>
+  ForwardIterator uninitialized_fill_n(ForwardIterator first,
+                                       Size n,
+                                       const T &x)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::uninitialized_fill_n(select_system(system), first, n, x);
+} // end uninitialized_fill_n()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/unique.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/unique.inl
new file mode 100644
index 000000000000..ac5475f0238d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/unique.inl
@@ -0,0 +1,393 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/unique.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/generic/select_system.h>
+#include <thrust/system/detail/generic/unique.h>
+#include <thrust/system/detail/generic/unique_by_key.h>
+#include <thrust/system/detail/adl/unique.h>
+#include <thrust/system/detail/adl/unique_by_key.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+ForwardIterator unique(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       ForwardIterator first,
+                       ForwardIterator last)
+{
+  using thrust::system::detail::generic::unique;
+  return unique(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end unique()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+ForwardIterator unique(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       ForwardIterator first,
+                       ForwardIterator last,
+                       BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::unique;
+  return unique(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, binary_pred);
+} // end unique()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator unique_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator output)
+{
+  using thrust::system::detail::generic::unique_copy;
+  return unique_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, output);
+} // end unique_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+OutputIterator unique_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator output,
+                           BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::unique_copy;
+  return unique_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, output, binary_pred);
+} // end unique_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+  unique_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator1 keys_first,
+                ForwardIterator1 keys_last,
+                ForwardIterator2 values_first)
+{
+  using thrust::system::detail::generic::unique_by_key;
+  return unique_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first);
+} // end unique_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+  unique_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator1 keys_first,
+                ForwardIterator1 keys_last,
+                ForwardIterator2 values_first,
+                BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::unique_by_key;
+  return unique_by_key(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, binary_pred);
+} // end unique_by_key()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  unique_by_key_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                     InputIterator1 keys_first,
+                     InputIterator1 keys_last,
+                     InputIterator2 values_first,
+                     OutputIterator1 keys_output,
+                     OutputIterator2 values_output)
+{
+  using thrust::system::detail::generic::unique_by_key_copy;
+  return unique_by_key_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, keys_output, values_output);
+} // end unique_by_key_copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  unique_by_key_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                     InputIterator1 keys_first,
+                     InputIterator1 keys_last,
+                     InputIterator2 values_first,
+                     OutputIterator1 keys_output,
+                     OutputIterator2 values_output,
+                     BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::unique_by_key_copy;
+  return unique_by_key_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, keys_output, values_output, binary_pred);
+} // end unique_by_key_copy()
+
+
+template<typename ForwardIterator>
+  ForwardIterator unique(ForwardIterator first,
+                         ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::unique(select_system(system), first, last);
+} // end unique()
+
+
+template<typename ForwardIterator,
+         typename BinaryPredicate>
+  ForwardIterator unique(ForwardIterator first,
+                         ForwardIterator last,
+                         BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::unique(select_system(system), first, last, binary_pred);
+} // end unique()
+
+
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator unique_copy(InputIterator first,
+                             InputIterator last,
+                             OutputIterator output)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::unique_copy(select_system(system1,system2), first, last, output);
+} // end unique_copy()
+
+
+template<typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+  OutputIterator unique_copy(InputIterator first,
+                             InputIterator last,
+                             OutputIterator output,
+                             BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator>::type  System1;
+  typedef typename thrust::iterator_system<OutputIterator>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::unique_copy(select_system(system1,system2), first, last, output, binary_pred);
+} // end unique_copy()
+
+
+template<typename ForwardIterator1,
+         typename ForwardIterator2>
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+    unique_by_key(ForwardIterator1 keys_first,
+                  ForwardIterator1 keys_last,
+                  ForwardIterator2 values_first)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator1>::type System1;
+  typedef typename thrust::iterator_system<ForwardIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::unique_by_key(select_system(system1,system2), keys_first, keys_last, values_first);
+} // end unique_by_key()
+
+
+template<typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+    unique_by_key(ForwardIterator1 keys_first,
+                  ForwardIterator1 keys_last,
+                  ForwardIterator2 values_first,
+                  BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator1>::type System1;
+  typedef typename thrust::iterator_system<ForwardIterator2>::type System2;
+
+  System1 system1;
+  System2 system2;
+
+  return thrust::unique_by_key(select_system(system1,system2), keys_first, keys_last, values_first, binary_pred);
+} // end unique_by_key()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(InputIterator1 keys_first,
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_output,
+                       OutputIterator2 values_output)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::unique_by_key_copy(select_system(system1,system2,system3,system4), keys_first, keys_last, values_first, keys_output, values_output);
+} // end unique_by_key_copy()
+
+
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(InputIterator1 keys_first,
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_output,
+                       OutputIterator2 values_output,
+                       BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<InputIterator1>::type  System1;
+  typedef typename thrust::iterator_system<InputIterator2>::type  System2;
+  typedef typename thrust::iterator_system<OutputIterator1>::type System3;
+  typedef typename thrust::iterator_system<OutputIterator2>::type System4;
+
+  System1 system1;
+  System2 system2;
+  System3 system3;
+  System4 system4;
+
+  return thrust::unique_by_key_copy(select_system(system1,system2,system3,system4), keys_first, keys_last, values_first, keys_output, values_output, binary_pred);
+} // end unique_by_key_copy()
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy,
+          typename ForwardIterator,
+          typename BinaryPredicate>
+__host__ __device__
+    typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::unique_count;
+  return unique_count(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, binary_pred);
+} // end unique_count()
+
+__thrust_exec_check_disable__
+template <typename DerivedPolicy,
+          typename ForwardIterator>
+__host__ __device__
+    typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last)
+{
+  using thrust::system::detail::generic::unique_count;
+  return unique_count(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
+} // end unique_count()
+
+__thrust_exec_check_disable__
+template <typename ForwardIterator,
+          typename BinaryPredicate>
+__host__ __device__
+    typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::unique_count(select_system(system), first, last, binary_pred);
+} // end unique_count()
+
+__thrust_exec_check_disable__
+template <typename ForwardIterator>
+__host__ __device__
+    typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(ForwardIterator first,
+                 ForwardIterator last)
+{
+  using thrust::system::detail::generic::select_system;
+
+  typedef typename thrust::iterator_system<ForwardIterator>::type System;
+
+  System system;
+
+  return thrust::unique_count(select_system(system), first, last);
+} // end unique_count()
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/use_default.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/use_default.h
new file mode 100644
index 000000000000..f25b6274c447
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/use_default.h
@@ -0,0 +1,26 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+struct use_default {};
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/util/align.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/util/align.h
new file mode 100644
index 000000000000..a3aa75bfe411
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/util/align.h
@@ -0,0 +1,60 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/cstdint.h>
+
+// functions to handle memory alignment
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+namespace util
+{
+
+
+template<typename T>
+__host__ __device__
+T *align_up(T * ptr, detail::uintptr_t bytes)
+{
+  return (T *) ( bytes * (((detail::uintptr_t) ptr + (bytes - 1)) / bytes) );
+}
+
+
+template<typename T>
+__host__ __device__
+T *align_down(T * ptr, detail::uintptr_t bytes)
+{
+  return (T *) ( bytes * (detail::uintptr_t(ptr) / bytes) );
+}
+
+
+template<typename T>
+__host__ __device__
+bool is_aligned(T * ptr, detail::uintptr_t bytes = sizeof(T))
+{
+  return detail::uintptr_t(ptr) % bytes == 0;
+}
+
+
+} // end namespace util
+} // end namespace detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.h
new file mode 100644
index 000000000000..df2edad1840d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.h
@@ -0,0 +1,603 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file vector_base.h
+ *  \brief Defines the interface to a base class for
+ *         host_vector & device_vector.
+ */
+
+#pragma once
+
+#include <thrust/iterator/detail/normal_iterator.h>
+#include <thrust/iterator/reverse_iterator.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/config.h>
+#include <thrust/detail/contiguous_storage.h>
+
+#include <initializer_list>
+#include <vector>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<typename T, typename Alloc>
+  class vector_base
+{
+  private:
+    typedef thrust::detail::contiguous_storage<T,Alloc> storage_type;
+
+  public:
+    // typedefs
+    typedef typename storage_type::value_type      value_type;
+    typedef typename storage_type::pointer         pointer;
+    typedef typename storage_type::const_pointer   const_pointer;
+    typedef typename storage_type::reference       reference;
+    typedef typename storage_type::const_reference const_reference;
+    typedef typename storage_type::size_type       size_type;
+    typedef typename storage_type::difference_type difference_type;
+    typedef typename storage_type::allocator_type  allocator_type;
+
+    typedef typename storage_type::iterator        iterator;
+    typedef typename storage_type::const_iterator  const_iterator;
+
+    typedef thrust::reverse_iterator<iterator>       reverse_iterator;
+    typedef thrust::reverse_iterator<const_iterator> const_reverse_iterator;
+
+    /*! This constructor creates an empty vector_base.
+     */
+    vector_base(void);
+
+    /*! This constructor creates an empty vector_base.
+     *  \param alloc The allocator to use by this vector_base.
+     */
+    explicit vector_base(const Alloc &alloc);
+
+    /*! This constructor creates a vector_base with default-constructed
+     *  elements.
+     *  \param n The number of elements to create.
+     */
+    explicit vector_base(size_type n);
+
+    /*! This constructor creates a vector_base with default-constructed
+     *  elements.
+     *  \param n The number of elements to create.
+     *  \param alloc The allocator to use by this vector_base.
+     */
+    explicit vector_base(size_type n, const Alloc &alloc);
+
+    /*! This constructor creates a vector_base with copies
+     *  of an exemplar element.
+     *  \param n The number of elements to initially create.
+     *  \param value An element to copy.
+     */
+    explicit vector_base(size_type n, const value_type &value);
+
+    /*! This constructor creates a vector_base with copies
+     *  of an exemplar element.
+     *  \param n The number of elements to initially create.
+     *  \param value An element to copy.
+     *  \param alloc The allocator to use by this vector_base.
+     */
+    explicit vector_base(size_type n, const value_type &value, const Alloc &alloc);
+
+    /*! Copy constructor copies from an exemplar vector_base.
+     *  \param v The vector_base to copy.
+     */
+    vector_base(const vector_base &v);
+
+    /*! Copy constructor copies from an exemplar vector_base.
+     *  \param v The vector_base to copy.
+     *  \param alloc The allocator to use by this vector_base.
+     */
+    vector_base(const vector_base &v, const Alloc &alloc);
+
+  #if THRUST_CPP_DIALECT >= 2011
+    /*! Move constructor moves from another vector_base.
+     *  \param v The vector_base to move.
+     */
+    vector_base(vector_base &&v);
+
+    // FIXME: the internal Thrust machinery in range_init doesn't work with move
+    // iterators, which is necessary for the following constructor to be implemented
+    // correctly
+    // vector_base(vector_base &&v, const Alloc &alloc);
+  #endif
+
+    /*! Copy assign operator copies from another vector_base.
+     *  \param v The vector_base to copy.
+     */
+    vector_base &operator=(const vector_base &v);
+
+    /*! Move assign operator moves from another vector_base.
+     *  \param v The vector_base to move.
+     */
+    vector_base &operator=(vector_base &&v);
+
+    /*! This constructor builds a \p vector_base from an intializer_list.
+     *  \param il The intializer_list.
+     */
+    vector_base(std::initializer_list<T> il);
+      
+    /*! This constructor builds a \p vector_base from an intializer_list.
+     *  \param il The intializer_list.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    vector_base(std::initializer_list<T> il, const Alloc &alloc);
+    
+    /*! Assign operator copies from an initializer_list
+     *  \param il The initializer_list.
+     */
+    vector_base &operator=(std::initializer_list<T> il);
+
+    /*! Copy constructor copies from an exemplar vector_base with different
+     *  type.
+     *  \param v The vector_base to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    vector_base(const vector_base<OtherT, OtherAlloc> &v);
+
+    /*! assign operator makes a copy of an exemplar vector_base with different
+     *  type.
+     *  \param v The vector_base to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    vector_base &operator=(const vector_base<OtherT,OtherAlloc> &v);
+
+    /*! Copy constructor copies from an exemplar std::vector.
+     *  \param v The std::vector to copy.
+     *  XXX TODO: Make this method redundant with a properly templatized constructor.
+     *            We would like to copy from a vector whose element type is anything
+     *            assignable to value_type.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    vector_base(const std::vector<OtherT, OtherAlloc> &v);
+
+    /*! assign operator makes a copy of an exemplar std::vector.
+     *  \param v The vector to copy.
+     *  XXX TODO: Templatize this assign on the type of the vector to copy from.
+     *            We would like to copy from a vector whose element type is anything
+     *            assignable to value_type.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    vector_base &operator=(const std::vector<OtherT,OtherAlloc> &v);
+
+    /*! This constructor builds a vector_base from a range.
+     *  \param first The beginning of the range.
+     *  \param last The end of the range.
+     */
+    template<typename InputIterator>
+    vector_base(InputIterator first, InputIterator last);
+
+    /*! This constructor builds a vector_base from a range.
+     *  \param first The beginning of the range.
+     *  \param last The end of the range.
+     *  \param alloc The allocator to use by this vector_base.
+     */
+    template<typename InputIterator>
+    vector_base(InputIterator first, InputIterator last, const Alloc &alloc);
+
+    /*! The destructor erases the elements.
+     */
+    ~vector_base(void);
+
+    /*! \brief Resizes this vector_base to the specified number of elements.
+     *  \param new_size Number of elements this vector_base should contain.
+     *  \throw std::length_error If n exceeds max_size9).
+     *
+     *  This method will resize this vector_base to the specified number of
+     *  elements. If the number is smaller than this vector_base's current
+     *  size this vector_base is truncated, otherwise this vector_base is
+     *  extended and new elements are default constructed.
+     */
+    void resize(size_type new_size);
+
+    /*! \brief Resizes this vector_base to the specified number of elements.
+     *  \param new_size Number of elements this vector_base should contain.
+     *  \param x Data with which new elements should be populated.
+     *  \throw std::length_error If n exceeds max_size().
+     *
+     *  This method will resize this vector_base to the specified number of
+     *  elements.  If the number is smaller than this vector_base's current
+     *  size this vector_base is truncated, otherwise this vector_base is
+     *  extended and new elements are populated with given data.
+     */
+    void resize(size_type new_size, const value_type &x);
+
+    /*! Returns the number of elements in this vector_base.
+     */
+    size_type size(void) const;
+
+    /*! Returns the size() of the largest possible vector_base.
+     *  \return The largest possible return value of size().
+     */
+    size_type max_size(void) const;
+
+    /*! \brief If n is less than or equal to capacity(), this call has no effect.
+     *         Otherwise, this method is a request for allocation of additional memory. If
+     *         the request is successful, then capacity() is greater than or equal to
+     *         n; otherwise, capacity() is unchanged. In either case, size() is unchanged.
+     *  \throw std::length_error If n exceeds max_size().
+     */
+    void reserve(size_type n);
+
+    /*! Returns the number of elements which have been reserved in this
+     *  vector_base.
+     */
+    size_type capacity(void) const;
+
+    /*! This method shrinks the capacity of this vector_base to exactly
+     *  fit its elements.
+     */
+    void shrink_to_fit(void);
+
+    /*! \brief Subscript access to the data contained in this vector_dev.
+     *  \param n The index of the element for which data should be accessed.
+     *  \return Read/write reference to data.
+     *
+     *  This operator allows for easy, array-style, data access.
+     *  Note that data access with this operator is unchecked and
+     *  out_of_range lookups are not defined.
+     */
+    reference operator[](size_type n);
+
+    /*! \brief Subscript read access to the data contained in this vector_dev.
+     *  \param n The index of the element for which data should be accessed.
+     *  \return Read reference to data.
+     *
+     *  This operator allows for easy, array-style, data access.
+     *  Note that data access with this operator is unchecked and
+     *  out_of_range lookups are not defined.
+     */
+    const_reference operator[](size_type n) const;
+
+    /*! This method returns an iterator pointing to the beginning of
+     *  this vector_base.
+     *  \return mStart
+     */
+    iterator begin(void);
+
+    /*! This method returns a const_iterator pointing to the beginning
+     *  of this vector_base.
+     *  \return mStart
+     */
+    const_iterator begin(void) const;
+
+    /*! This method returns a const_iterator pointing to the beginning
+     *  of this vector_base.
+     *  \return mStart
+     */
+    const_iterator cbegin(void) const;
+
+    /*! This method returns a reverse_iterator pointing to the beginning of
+     *  this vector_base's reversed sequence.
+     *  \return A reverse_iterator pointing to the beginning of this
+     *          vector_base's reversed sequence.
+     */
+    reverse_iterator rbegin(void);
+
+    /*! This method returns a const_reverse_iterator pointing to the beginning of
+     *  this vector_base's reversed sequence.
+     *  \return A const_reverse_iterator pointing to the beginning of this
+     *          vector_base's reversed sequence.
+     */
+    const_reverse_iterator rbegin(void) const;
+
+    /*! This method returns a const_reverse_iterator pointing to the beginning of
+     *  this vector_base's reversed sequence.
+     *  \return A const_reverse_iterator pointing to the beginning of this
+     *          vector_base's reversed sequence.
+     */
+    const_reverse_iterator crbegin(void) const;
+
+    /*! This method returns an iterator pointing to one element past the
+     *  last of this vector_base.
+     *  \return begin() + size().
+     */
+    iterator end(void);
+
+    /*! This method returns a const_iterator pointing to one element past the
+     *  last of this vector_base.
+     *  \return begin() + size().
+     */
+    const_iterator end(void) const;
+
+    /*! This method returns a const_iterator pointing to one element past the
+     *  last of this vector_base.
+     *  \return begin() + size().
+     */
+    const_iterator cend(void) const;
+
+    /*! This method returns a reverse_iterator pointing to one element past the
+     *  last of this vector_base's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    reverse_iterator rend(void);
+
+    /*! This method returns a const_reverse_iterator pointing to one element past the
+     *  last of this vector_base's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    const_reverse_iterator rend(void) const;
+
+    /*! This method returns a const_reverse_iterator pointing to one element past the
+     *  last of this vector_base's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    const_reverse_iterator crend(void) const;
+
+    /*! This method returns a const_reference referring to the first element of this
+     *  vector_base.
+     *  \return The first element of this vector_base.
+     */
+    const_reference front(void) const;
+
+    /*! This method returns a reference pointing to the first element of this
+     *  vector_base.
+     *  \return The first element of this vector_base.
+     */
+    reference front(void);
+
+    /*! This method returns a const reference pointing to the last element of
+     *  this vector_base.
+     *  \return The last element of this vector_base.
+     */
+    const_reference back(void) const;
+
+    /*! This method returns a reference referring to the last element of
+     *  this vector_dev.
+     *  \return The last element of this vector_base.
+     */
+    reference back(void);
+
+    /*! This method returns a pointer to this vector_base's first element.
+     *  \return A pointer to the first element of this vector_base.
+     */
+    pointer data(void);
+
+    /*! This method returns a const_pointer to this vector_base's first element.
+     *  \return a const_pointer to the first element of this vector_base.
+     */
+    const_pointer data(void) const;
+
+    /*! This method resizes this vector_base to 0.
+     */
+    void clear(void);
+
+    /*! This method returns true iff size() == 0.
+     *  \return true if size() == 0; false, otherwise.
+     */
+    bool empty(void) const;
+
+    /*! This method appends the given element to the end of this vector_base.
+     *  \param x The element to append.
+     */
+    void push_back(const value_type &x);
+
+    /*! This method erases the last element of this vector_base, invalidating
+     *  all iterators and references to it.
+     */
+    void pop_back(void);
+
+    /*! This method swaps the contents of this vector_base with another vector_base.
+     *  \param v The vector_base with which to swap.
+     */
+    void swap(vector_base &v);
+
+    /*! This method removes the element at position pos.
+     *  \param pos The position of the element of interest.
+     *  \return An iterator pointing to the new location of the element that followed the element
+     *          at position pos.
+     */
+    iterator erase(iterator pos);
+
+    /*! This method removes the range of elements [first,last) from this vector_base.
+     *  \param first The beginning of the range of elements to remove.
+     *  \param last The end of the range of elements to remove.
+     *  \return An iterator pointing to the new location of the element that followed the last
+     *          element in the sequence [first,last).
+     */
+    iterator erase(iterator first, iterator last);
+
+    /*! This method inserts a single copy of a given exemplar value at the
+     *  specified position in this vector_base.
+     *  \param position The insertion position.
+     *  \param x The exemplar element to copy & insert.
+     *  \return An iterator pointing to the newly inserted element.
+     */
+    iterator insert(iterator position, const T &x); 
+
+    /*! This method inserts a copy of an exemplar value to a range at the
+     *  specified position in this vector_base.
+     *  \param position The insertion position
+     *  \param n The number of insertions to perform.
+     *  \param x The value to replicate and insert.
+     */
+    void insert(iterator position, size_type n, const T &x);
+
+    /*! This method inserts a copy of an input range at the specified position
+     *  in this vector_base.
+     *  \param position The insertion position.
+     *  \param first The beginning of the range to copy.
+     *  \param last  The end of the range to copy.
+     *
+     *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator>Input Iterator</a>,
+     *                        and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>.
+     */
+    template<typename InputIterator>
+    void insert(iterator position, InputIterator first, InputIterator last);
+
+    /*! This version of \p assign replicates a given exemplar
+     *  \p n times into this vector_base.
+     *  \param n The number of times to copy \p x.
+     *  \param x The exemplar element to replicate.
+     */
+    void assign(size_type n, const T &x);
+
+    /*! This version of \p assign makes this vector_base a copy of a given input range.
+     *  \param first The beginning of the range to copy.
+     *  \param last  The end of the range to copy.
+     *
+     *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>.
+     */
+    template<typename InputIterator>
+    void assign(InputIterator first, InputIterator last);
+
+    /*! This method returns a copy of this vector's allocator.
+     *  \return A copy of the alloctor used by this vector.
+     */
+    allocator_type get_allocator(void) const;
+
+  protected:
+    // Our storage
+    storage_type m_storage;
+
+    // The size of this vector_base, in number of elements.
+    size_type m_size;
+
+  private:
+    // these methods resolve the ambiguity of the constructor template of form (Iterator, Iterator)
+    template<typename IteratorOrIntegralType>
+      void init_dispatch(IteratorOrIntegralType begin, IteratorOrIntegralType end, false_type); 
+
+    template<typename IteratorOrIntegralType>
+      void init_dispatch(IteratorOrIntegralType n, IteratorOrIntegralType value, true_type); 
+
+    template<typename InputIterator>
+      void range_init(InputIterator first, InputIterator last);
+
+    template<typename InputIterator>
+      void range_init(InputIterator first, InputIterator last, thrust::incrementable_traversal_tag);
+
+    template<typename ForwardIterator>
+      void range_init(ForwardIterator first, ForwardIterator last, thrust::random_access_traversal_tag);
+
+    void default_init(size_type n);
+
+    void fill_init(size_type n, const T &x);
+
+    // these methods resolve the ambiguity of the insert() template of form (iterator, InputIterator, InputIterator)
+    template<typename InputIteratorOrIntegralType>
+      void insert_dispatch(iterator position, InputIteratorOrIntegralType first, InputIteratorOrIntegralType last, false_type);
+
+    // these methods resolve the ambiguity of the insert() template of form (iterator, InputIterator, InputIterator)
+    template<typename InputIteratorOrIntegralType>
+      void insert_dispatch(iterator position, InputIteratorOrIntegralType n, InputIteratorOrIntegralType x, true_type);
+
+    // this method appends n default-constructed elements at the end
+    void append(size_type n);
+
+    // this method performs insertion from a fill value
+    void fill_insert(iterator position, size_type n, const T &x);
+
+    // this method performs insertion from a range
+    template<typename InputIterator>
+      void copy_insert(iterator position, InputIterator first, InputIterator last);
+
+    // these methods resolve the ambiguity of the assign() template of form (InputIterator, InputIterator)
+    template<typename InputIterator>
+      void assign_dispatch(InputIterator first, InputIterator last, false_type);
+
+    // these methods resolve the ambiguity of the assign() template of form (InputIterator, InputIterator)
+    template<typename Integral>
+      void assign_dispatch(Integral n, Integral x, true_type);
+
+    // this method performs assignment from a range
+    template<typename InputIterator>
+      void range_assign(InputIterator first, InputIterator last);
+
+    // this method performs assignment from a range of RandomAccessIterators
+    template<typename RandomAccessIterator>
+      void range_assign(RandomAccessIterator first, RandomAccessIterator last, thrust::random_access_traversal_tag);
+
+    // this method performs assignment from a range of InputIterators
+    template<typename InputIterator>
+      void range_assign(InputIterator first, InputIterator last, thrust::incrementable_traversal_tag);
+
+    // this method performs assignment from a fill value
+    void fill_assign(size_type n, const T &x);
+
+    // this method allocates new storage and construct copies the given range
+    template<typename ForwardIterator>
+    void allocate_and_copy(size_type requested_size,
+                           ForwardIterator first, ForwardIterator last,
+                           storage_type &new_storage);
+}; // end vector_base
+
+} // end detail
+
+/*! This function assigns the contents of vector a to vector b and the
+ *  contents of vector b to vector a.
+ *
+ *  \param a The first vector of interest. After completion, the contents
+ *           of b will be returned here.
+ *  \param b The second vector of interest. After completion, the contents
+ *           of a will be returned here.
+ */
+template<typename T, typename Alloc>
+  void swap(detail::vector_base<T,Alloc> &a,
+            detail::vector_base<T,Alloc> &b);
+
+
+/*! This operator allows comparison between two vectors.
+ *  \param lhs The first \p vector to compare.
+ *  \param rhs The second \p vector to compare.
+ *  \return \c true if and only if each corresponding element in either
+ *          \p vector equals the other; \c false, otherwise.
+ */
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator==(const detail::vector_base<T1,Alloc1>& lhs,
+                const detail::vector_base<T2,Alloc2>& rhs);
+    
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator==(const detail::vector_base<T1,Alloc1>& lhs,
+                const std::vector<T2,Alloc2>&         rhs);
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator==(const std::vector<T1,Alloc1>&         lhs,
+                const detail::vector_base<T2,Alloc2>& rhs);
+
+/*! This operator allows comparison between two vectors.
+ *  \param lhs The first \p vector to compare.
+ *  \param rhs The second \p vector to compare.
+ *  \return \c false if and only if each corresponding element in either
+ *          \p vector equals the other; \c true, otherwise.
+ */
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator!=(const detail::vector_base<T1,Alloc1>& lhs,
+                const detail::vector_base<T2,Alloc2>& rhs);
+    
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator!=(const detail::vector_base<T1,Alloc1>& lhs,
+                const std::vector<T2,Alloc2>&         rhs);
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator!=(const std::vector<T1,Alloc1>&         lhs,
+                const detail::vector_base<T2,Alloc2>& rhs);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/vector_base.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.inl b/thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.inl
new file mode 100644
index 000000000000..d9c29b0d291a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/detail/vector_base.inl
@@ -0,0 +1,1345 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/vector_base.h>
+#include <thrust/detail/copy.h>
+#include <thrust/detail/overlapped_copy.h>
+#include <thrust/equal.h>
+#include <thrust/distance.h>
+#include <thrust/advance.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/minmax.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/temporary_array.h>
+
+#include <stdexcept>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(void)
+      :m_storage(),
+       m_size(0)
+{
+  ;
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(const Alloc &alloc)
+      :m_storage(alloc),
+       m_size(0)
+{
+  ;
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(size_type n)
+      :m_storage(),
+       m_size(0)
+{
+  default_init(n);
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(size_type n, const Alloc &alloc)
+      :m_storage(alloc),
+       m_size(0)
+{
+  default_init(n);
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(size_type n, const value_type &value)
+      :m_storage(),
+       m_size(0)
+{
+  fill_init(n,value);
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(size_type n, const value_type &value, const Alloc &alloc)
+      :m_storage(alloc),
+       m_size(0)
+{
+  fill_init(n,value);
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(const vector_base &v)
+      :m_storage(copy_allocator_t(), v.m_storage),
+       m_size(0)
+{
+  range_init(v.begin(), v.end());
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(const vector_base &v, const Alloc &alloc)
+      :m_storage(alloc),
+       m_size(0)
+{
+  range_init(v.begin(), v.end());
+} // end vector_base::vector_base()
+
+#if THRUST_CPP_DIALECT >= 2011
+  template<typename T, typename Alloc>
+    vector_base<T,Alloc>
+      ::vector_base(vector_base &&v)
+        :m_storage(copy_allocator_t(), v.m_storage),
+         m_size(0)
+  {
+    *this = std::move(v);
+  } //end vector_base::vector_base()
+#endif
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc> &
+    vector_base<T,Alloc>
+      ::operator=(const vector_base &v)
+{
+  if(this != &v)
+  {
+    m_storage.destroy_on_allocator_mismatch(v.m_storage, begin(), end());
+    m_storage.deallocate_on_allocator_mismatch(v.m_storage);
+
+    m_storage.propagate_allocator(v.m_storage);
+
+    assign(v.begin(), v.end());
+  } // end if
+
+  return *this;
+} // end vector_base::operator=()
+
+#if THRUST_CPP_DIALECT >= 2011
+  template<typename T, typename Alloc>
+    vector_base<T,Alloc> &
+      vector_base<T,Alloc>
+        ::operator=(vector_base &&v)
+  {
+    m_storage.destroy(begin(), end());
+    m_storage = std::move(v.m_storage);
+    m_size = std::move(v.m_size);
+
+    v.m_storage = contiguous_storage<T,Alloc>(copy_allocator_t(), m_storage);
+    v.m_size = 0;
+
+    return *this;
+  } // end vector_base::operator=()
+#endif
+
+template<typename T, typename Alloc>
+  template<typename OtherT, typename OtherAlloc>
+    vector_base<T,Alloc>
+      ::vector_base(const vector_base<OtherT,OtherAlloc> &v)
+        :m_storage(),
+         m_size(0)
+{
+  range_init(v.begin(), v.end());
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  template<typename OtherT, typename OtherAlloc>
+    vector_base<T,Alloc> &
+      vector_base<T,Alloc>
+        ::operator=(const vector_base<OtherT,OtherAlloc> &v)
+{
+  assign(v.begin(), v.end());
+
+  return *this;
+} // end vector_base::operator=()
+
+template<typename T, typename Alloc>
+  template<typename OtherT, typename OtherAlloc>
+    vector_base<T,Alloc>
+      ::vector_base(const std::vector<OtherT,OtherAlloc> &v)
+        :m_storage(),
+         m_size(0)
+{
+  range_init(v.begin(), v.end());
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  template<typename OtherT, typename OtherAlloc>
+    vector_base<T,Alloc> &
+      vector_base<T,Alloc>
+        ::operator=(const std::vector<OtherT,OtherAlloc> &v)
+{
+  assign(v.begin(), v.end());
+
+  return *this;
+} // end vector_base::operator=()
+
+  template<typename T, typename Alloc>
+    vector_base<T,Alloc>
+      ::vector_base(std::initializer_list<T> il)
+        :m_storage(),
+         m_size(0)
+  {
+    range_init(il.begin(), il.end());
+  } // end vector_base::vector_base()
+
+  template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::vector_base(std::initializer_list<T> il, const Alloc &alloc)
+    :m_storage(alloc),
+      m_size(0)
+  {
+    range_init(il.begin(), il.end());
+  } // end vector_base::vector_base()
+
+  template<typename T, typename Alloc>
+    vector_base<T,Alloc> &
+      vector_base<T,Alloc>
+      ::operator=(std::initializer_list<T> il)
+  {
+    assign(il.begin(), il.end());
+
+    return *this;
+  } // end vector_base::operator=()
+
+template<typename T, typename Alloc>
+  template<typename IteratorOrIntegralType>
+    void vector_base<T,Alloc>
+      ::init_dispatch(IteratorOrIntegralType n,
+                      IteratorOrIntegralType value,
+                      true_type)
+{
+  fill_init(n,value);
+} // end vector_base::init_dispatch()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::default_init(size_type n)
+{
+  if(n > 0)
+  {
+    m_storage.allocate(n);
+    m_size = n;
+
+    m_storage.default_construct_n(begin(), size());
+  } // end if
+} // end vector_base::default_init()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::fill_init(size_type n, const T &x)
+{
+  if(n > 0)
+  {
+    m_storage.allocate(n);
+    m_size = n;
+
+    m_storage.uninitialized_fill_n(begin(), size(), x);
+  } // end if
+} // end vector_base::fill_init()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::init_dispatch(InputIterator first,
+                      InputIterator last,
+                      false_type)
+{
+  range_init(first, last);
+} // end vector_base::init_dispatch()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::range_init(InputIterator first,
+                   InputIterator last)
+{
+  range_init(first, last,
+    typename thrust::iterator_traversal<InputIterator>::type());
+} // end vector_base::range_init()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::range_init(InputIterator first,
+                   InputIterator last,
+                   thrust::incrementable_traversal_tag)
+{
+  for(; first != last; ++first)
+    push_back(*first);
+} // end vector_base::range_init()
+
+template<typename T, typename Alloc>
+  template<typename ForwardIterator>
+    void vector_base<T,Alloc>
+      ::range_init(ForwardIterator first,
+                   ForwardIterator last,
+                   thrust::random_access_traversal_tag)
+{
+  size_type new_size = thrust::distance(first, last);
+
+  allocate_and_copy(new_size, first, last, m_storage);
+  m_size    = new_size;
+} // end vector_base::range_init()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    vector_base<T,Alloc>
+      ::vector_base(InputIterator first,
+                    InputIterator last)
+        :m_storage(),
+         m_size(0)
+{
+  // check the type of InputIterator: if it's an integral type,
+  // we need to interpret this call as (size_type, value_type)
+  typedef thrust::detail::is_integral<InputIterator> Integer;
+
+  init_dispatch(first, last, Integer());
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    vector_base<T,Alloc>
+      ::vector_base(InputIterator first,
+                    InputIterator last,
+                    const Alloc &alloc)
+        :m_storage(alloc),
+         m_size(0)
+{
+  // check the type of InputIterator: if it's an integral type,
+  // we need to interpret this call as (size_type, value_type)
+  typedef thrust::detail::is_integral<InputIterator> Integer;
+
+  init_dispatch(first, last, Integer());
+} // end vector_base::vector_base()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::resize(size_type new_size)
+{
+  if(new_size < size())
+  {
+    iterator new_end = begin();
+    thrust::advance(new_end, new_size);
+    erase(new_end, end());
+  } // end if
+  else
+  {
+    append(new_size - size());
+  } // end else
+} // end vector_base::resize()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::resize(size_type new_size, const value_type &x)
+{
+  if(new_size < size())
+  {
+    iterator new_end = begin();
+    thrust::advance(new_end, new_size);
+    erase(new_end, end());
+  } // end if
+  else
+  {
+    insert(end(), new_size - size(), x);
+  } // end else
+} // end vector_base::resize()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::size_type
+    vector_base<T,Alloc>
+      ::size(void) const
+{
+  return m_size;
+} // end vector_base::size()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::size_type
+    vector_base<T,Alloc>
+      ::max_size(void) const
+{
+  return m_storage.max_size();
+} // end vector_base::max_size()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::reserve(size_type n)
+{
+  if(n > capacity())
+  {
+    // compute the new capacity after the allocation
+    size_type new_capacity = n;
+
+    // do not exceed maximum storage
+    new_capacity = thrust::min THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, max_size());
+
+    // create new storage
+    storage_type new_storage(copy_allocator_t(), m_storage, new_capacity);
+
+    // record how many constructors we invoke in the try block below
+    iterator new_end = new_storage.begin();
+
+    try
+    {
+      // construct copy all elements into the newly allocated storage
+      new_end = m_storage.uninitialized_copy(begin(), end(), new_storage.begin());
+    } // end try
+    catch(...)
+    {
+      // something went wrong, so destroy & deallocate the new storage
+      new_storage.destroy(new_storage.begin(), new_end);
+      new_storage.deallocate();
+
+      // rethrow
+      throw;
+    } // end catch
+
+    // call destructors on the elements in the old storage
+    m_storage.destroy(begin(), end());
+
+    // record the vector's new state
+    m_storage.swap(new_storage);
+  } // end if
+} // end vector_base::reserve()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::size_type
+    vector_base<T,Alloc>
+      ::capacity(void) const
+{
+  return m_storage.size();
+} // end vector_base::capacity()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::shrink_to_fit(void)
+{
+  // use the swap trick
+  vector_base(*this).swap(*this);
+} // end vector_base::shrink_to_fit()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::reference
+    vector_base<T,Alloc>
+      ::operator[](const size_type n)
+{
+  return m_storage[n];
+} // end vector_base::operator[]
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reference
+    vector_base<T,Alloc>
+      ::operator[](const size_type n) const
+{
+  return m_storage[n];
+} // end vector_base::operator[]
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::iterator
+    vector_base<T,Alloc>
+      ::begin(void)
+{
+  return m_storage.begin();
+} // end vector_base::begin()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_iterator
+    vector_base<T,Alloc>
+      ::begin(void) const
+{
+  return m_storage.begin();
+} // end vector_base::begin()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_iterator
+    vector_base<T,Alloc>
+      ::cbegin(void) const
+{
+  return begin();
+} // end vector_base::cbegin()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::reverse_iterator
+    vector_base<T,Alloc>
+      ::rbegin(void)
+{
+  return reverse_iterator(end());
+} // end vector_base::rbegin()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reverse_iterator
+    vector_base<T,Alloc>
+      ::rbegin(void) const
+{
+  return const_reverse_iterator(end());
+} // end vector_base::rbegin()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reverse_iterator
+    vector_base<T,Alloc>
+      ::crbegin(void) const
+{
+  return rbegin();
+} // end vector_base::crbegin()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::iterator
+    vector_base<T,Alloc>
+      ::end(void)
+{
+  iterator result = begin();
+  thrust::advance(result, size());
+  return result;
+} // end vector_base::end()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_iterator
+    vector_base<T,Alloc>
+      ::end(void) const
+{
+  const_iterator result = begin();
+  thrust::advance(result, size());
+  return result;
+} // end vector_base::end()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_iterator
+    vector_base<T,Alloc>
+      ::cend(void) const
+{
+  return end();
+} // end vector_base::cend()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::reverse_iterator
+    vector_base<T,Alloc>
+      ::rend(void)
+{
+  return reverse_iterator(begin());
+} // end vector_base::rend()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reverse_iterator
+    vector_base<T,Alloc>
+      ::rend(void) const
+{
+  return const_reverse_iterator(begin());
+} // end vector_base::rend()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reverse_iterator
+    vector_base<T,Alloc>
+      ::crend(void) const
+{
+  return rend();
+} // end vector_base::crend()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reference
+    vector_base<T,Alloc>
+      ::front(void) const
+{
+  return *begin();
+} // end vector_base::front()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::reference
+    vector_base<T,Alloc>
+      ::front(void)
+{
+  return *begin();
+} // end vector_base::front()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_reference
+    vector_base<T,Alloc>
+      ::back(void) const
+{
+  const_iterator ptr_to_back = end();
+  --ptr_to_back;
+  return *ptr_to_back;
+} // end vector_base::vector_base
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::reference
+    vector_base<T,Alloc>
+      ::back(void)
+{
+  iterator ptr_to_back = end();
+  --ptr_to_back;
+  return *ptr_to_back;
+} // end vector_base::vector_base
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::pointer
+    vector_base<T,Alloc>
+      ::data(void)
+{
+  return pointer(&front());
+} // end vector_base::data()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::const_pointer
+    vector_base<T,Alloc>
+      ::data(void) const
+{
+  return const_pointer(&front());
+} // end vector_base::data()
+
+template<typename T, typename Alloc>
+  vector_base<T,Alloc>
+    ::~vector_base(void)
+{
+  // destroy every living thing
+  if (!empty())
+    m_storage.destroy(begin(),end());
+} // end vector_base::~vector_base()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::clear(void)
+{
+  erase(begin(), end());
+} // end vector_base::~vector_dev()
+
+template<typename T, typename Alloc>
+  bool vector_base<T,Alloc>
+    ::empty(void) const
+{
+  return size() == 0;
+} // end vector_base::empty();
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::push_back(const value_type &x)
+{
+  insert(end(), x);
+} // end vector_base::push_back()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::pop_back(void)
+{
+  iterator e = end();
+  iterator ptr_to_back = e;
+  --ptr_to_back;
+  m_storage.destroy(ptr_to_back, e);
+  --m_size;
+} // end vector_base::pop_back()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::iterator vector_base<T,Alloc>
+    ::erase(iterator pos)
+{
+  iterator end = pos;
+  ++end;
+  return erase(pos,end);
+} // end vector_base::erase()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::iterator vector_base<T,Alloc>
+    ::erase(iterator first, iterator last)
+{
+  // overlap copy the range [last,end()) to first
+  // XXX this copy only potentially overlaps
+  iterator i = thrust::detail::overlapped_copy(last, end(), first);
+
+  // destroy everything after i
+  m_storage.destroy(i, end());
+
+  // modify our size
+  m_size -= (last - first);
+
+  // return an iterator pointing to the position of the first element
+  // following the erased range
+  return first;
+} // end vector_base::erase()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::swap(vector_base &v)
+{
+  thrust::swap(m_storage,  v.m_storage);
+  thrust::swap(m_size,     v.m_size);
+} // end vector_base::swap()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::assign(size_type n, const T &x)
+{
+  fill_assign(n, x);
+} // end vector_base::assign()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::assign(InputIterator first, InputIterator last)
+{
+  // we could have received assign(n, x), so disambiguate on the
+  // type of InputIterator
+  typedef typename thrust::detail::is_integral<InputIterator> integral;
+
+  assign_dispatch(first, last, integral());
+} // end vector_base::assign()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::allocator_type
+    vector_base<T,Alloc>
+      ::get_allocator(void) const
+{
+  return m_storage.get_allocator();
+} // end vector_base::get_allocator()
+
+template<typename T, typename Alloc>
+  typename vector_base<T,Alloc>::iterator
+    vector_base<T,Alloc>
+      ::insert(iterator position, const T &x)
+{
+  // find the index of the insertion
+  size_type index = thrust::distance(begin(), position);
+
+  // make the insertion
+  insert(position, 1, x);
+
+  // return an iterator pointing back to position
+  iterator result = begin();
+  thrust::advance(result, index);
+  return result;
+} // end vector_base::insert()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::insert(iterator position, size_type n, const T &x)
+{
+  fill_insert(position, n, x);
+} // end vector_base::insert()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::insert(iterator position, InputIterator first, InputIterator last)
+{
+  // we could have received insert(position, n, x), so disambiguate on the
+  // type of InputIterator
+  typedef typename thrust::detail::is_integral<InputIterator> integral;
+
+  insert_dispatch(position, first, last, integral());
+} // end vector_base::insert()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::assign_dispatch(InputIterator first, InputIterator last, false_type)
+{
+  range_assign(first, last);
+} // end vector_base::assign_dispatch()
+
+template<typename T, typename Alloc>
+  template<typename Integral>
+    void vector_base<T,Alloc>
+      ::assign_dispatch(Integral n, Integral x, true_type)
+{
+  fill_assign(n, x);
+} // end vector_base::assign_dispatch()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::insert_dispatch(iterator position, InputIterator first, InputIterator last, false_type)
+{
+  copy_insert(position, first, last);
+} // end vector_base::insert_dispatch()
+
+template<typename T, typename Alloc>
+  template<typename Integral>
+    void vector_base<T,Alloc>
+      ::insert_dispatch(iterator position, Integral n, Integral x, true_type)
+{
+  fill_insert(position, n, x);
+} // end vector_base::insert_dispatch()
+
+template<typename T, typename Alloc>
+  template<typename ForwardIterator>
+    void vector_base<T,Alloc>
+      ::copy_insert(iterator position,
+                    ForwardIterator first,
+                    ForwardIterator last)
+{
+  if(first != last)
+  {
+    // how many new elements will we create?
+    const size_type num_new_elements = thrust::distance(first, last);
+    if(capacity() - size() >= num_new_elements)
+    {
+      // we've got room for all of them
+      // how many existing elements will we displace?
+      const size_type num_displaced_elements = end() - position;
+      iterator old_end = end();
+
+      if(num_displaced_elements > num_new_elements)
+      {
+        // construct copy n displaced elements to new elements
+        // following the insertion
+        m_storage.uninitialized_copy(end() - num_new_elements, end(), end());
+
+        // extend the size
+        m_size += num_new_elements;
+
+        // copy num_displaced_elements - num_new_elements elements to existing elements
+        // this copy overlaps
+        const size_type copy_length = (old_end - num_new_elements) - position;
+        thrust::detail::overlapped_copy(position, old_end - num_new_elements, old_end - copy_length);
+
+        // finally, copy the range to the insertion point
+        thrust::copy(first, last, position);
+      } // end if
+      else
+      {
+        ForwardIterator mid = first;
+        thrust::advance(mid, num_displaced_elements);
+
+        // construct copy new elements at the end of the vector
+        m_storage.uninitialized_copy(mid, last, end());
+
+        // extend the size
+        m_size += num_new_elements - num_displaced_elements;
+
+        // construct copy the displaced elements
+        m_storage.uninitialized_copy(position, old_end, end());
+
+        // extend the size
+        m_size += num_displaced_elements;
+
+        // copy to elements which already existed
+        thrust::copy(first, mid, position);
+      } // end else
+    } // end if
+    else
+    {
+      const size_type old_size = size();
+
+      // compute the new capacity after the allocation
+      size_type new_capacity = old_size + thrust::max THRUST_PREVENT_MACRO_SUBSTITUTION (old_size, num_new_elements);
+
+      // allocate exponentially larger new storage
+      new_capacity = thrust::max THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, 2 * capacity());
+
+      // do not exceed maximum storage
+      new_capacity = thrust::min THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, max_size());
+
+      if(new_capacity > max_size())
+      {
+        throw std::length_error("insert(): insertion exceeds max_size().");
+      } // end if
+
+      storage_type new_storage(copy_allocator_t(), m_storage, new_capacity);
+
+      // record how many constructors we invoke in the try block below
+      iterator new_end = new_storage.begin();
+
+      try
+      {
+        // construct copy elements before the insertion to the beginning of the newly
+        // allocated storage
+        new_end = m_storage.uninitialized_copy(begin(), position, new_storage.begin());
+
+        // construct copy elements to insert
+        new_end = m_storage.uninitialized_copy(first, last, new_end);
+
+        // construct copy displaced elements from the old storage to the new storage
+        // remember [position, end()) refers to the old storage
+        new_end = m_storage.uninitialized_copy(position, end(), new_end);
+      } // end try
+      catch(...)
+      {
+        // something went wrong, so destroy & deallocate the new storage
+        m_storage.destroy(new_storage.begin(), new_end);
+        new_storage.deallocate();
+
+        // rethrow
+        throw;
+      } // end catch
+
+      // call destructors on the elements in the old storage
+      m_storage.destroy(begin(), end());
+
+      // record the vector's new state
+      m_storage.swap(new_storage);
+      m_size = old_size + num_new_elements;
+    } // end else
+  } // end if
+} // end vector_base::copy_insert()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::append(size_type n)
+{
+  if(n != 0)
+  {
+    if(capacity() - size() >= n)
+    {
+      // we've got room for all of them
+
+      // default construct new elements at the end of the vector
+      m_storage.default_construct_n(end(), n);
+
+      // extend the size
+      m_size += n;
+    } // end if
+    else
+    {
+      const size_type old_size = size();
+
+      // compute the new capacity after the allocation
+      size_type new_capacity = old_size + thrust::max THRUST_PREVENT_MACRO_SUBSTITUTION (old_size, n);
+
+      // allocate exponentially larger new storage
+      new_capacity = thrust::max THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, 2 * capacity());
+
+      // do not exceed maximum storage
+      new_capacity = thrust::min THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, max_size());
+
+      // create new storage
+      storage_type new_storage(copy_allocator_t(), m_storage, new_capacity);
+
+      // record how many constructors we invoke in the try block below
+      iterator new_end = new_storage.begin();
+
+      try
+      {
+        // construct copy all elements into the newly allocated storage
+        new_end = m_storage.uninitialized_copy(begin(), end(), new_storage.begin());
+
+        // construct new elements to insert
+        new_storage.default_construct_n(new_end, n);
+        new_end += n;
+      } // end try
+      catch(...)
+      {
+        // something went wrong, so destroy & deallocate the new storage
+        new_storage.destroy(new_storage.begin(), new_end);
+        new_storage.deallocate();
+
+        // rethrow
+        throw;
+      } // end catch
+
+      // call destructors on the elements in the old storage
+      m_storage.destroy(begin(), end());
+
+      // record the vector's new state
+      m_storage.swap(new_storage);
+      m_size    = old_size + n;
+    } // end else
+  } // end if
+} // end vector_base::append()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::fill_insert(iterator position, size_type n, const T &x)
+{
+  if(n != 0)
+  {
+    if(capacity() - size() >= n)
+    {
+      // we've got room for all of them
+      // how many existing elements will we displace?
+      const size_type num_displaced_elements = end() - position;
+      iterator old_end = end();
+
+      if(num_displaced_elements > n)
+      {
+        // construct copy n displaced elements to new elements
+        // following the insertion
+        m_storage.uninitialized_copy(end() - n, end(), end());
+
+        // extend the size
+        m_size += n;
+
+        // copy num_displaced_elements - n elements to existing elements
+        // this copy overlaps
+        const size_type copy_length = (old_end - n) - position;
+        thrust::detail::overlapped_copy(position, old_end - n, old_end - copy_length);
+
+        // finally, fill the range to the insertion point
+        thrust::fill_n(position, n, x);
+      } // end if
+      else
+      {
+        // construct new elements at the end of the vector
+        m_storage.uninitialized_fill_n(end(), n - num_displaced_elements, x);
+
+        // extend the size
+        m_size += n - num_displaced_elements;
+
+        // construct copy the displaced elements
+        m_storage.uninitialized_copy(position, old_end, end());
+
+        // extend the size
+        m_size += num_displaced_elements;
+
+        // fill to elements which already existed
+        thrust::fill(position, old_end, x);
+      } // end else
+    } // end if
+    else
+    {
+      const size_type old_size = size();
+
+      // compute the new capacity after the allocation
+      size_type new_capacity = old_size + thrust::max THRUST_PREVENT_MACRO_SUBSTITUTION (old_size, n);
+
+      // allocate exponentially larger new storage
+      new_capacity = thrust::max THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, 2 * capacity());
+
+      // do not exceed maximum storage
+      new_capacity = thrust::min THRUST_PREVENT_MACRO_SUBSTITUTION <size_type>(new_capacity, max_size());
+
+      if(new_capacity > max_size())
+      {
+        throw std::length_error("insert(): insertion exceeds max_size().");
+      } // end if
+
+      storage_type new_storage(copy_allocator_t(), m_storage, new_capacity);
+
+      // record how many constructors we invoke in the try block below
+      iterator new_end = new_storage.begin();
+
+      try
+      {
+        // construct copy elements before the insertion to the beginning of the newly
+        // allocated storage
+        new_end = m_storage.uninitialized_copy(begin(), position, new_storage.begin());
+
+        // construct new elements to insert
+        m_storage.uninitialized_fill_n(new_end, n, x);
+        new_end += n;
+
+        // construct copy displaced elements from the old storage to the new storage
+        // remember [position, end()) refers to the old storage
+        new_end = m_storage.uninitialized_copy(position, end(), new_end);
+      } // end try
+      catch(...)
+      {
+        // something went wrong, so destroy & deallocate the new storage
+        m_storage.destroy(new_storage.begin(), new_end);
+        new_storage.deallocate();
+
+        // rethrow
+        throw;
+      } // end catch
+
+      // call destructors on the elements in the old storage
+      m_storage.destroy(begin(), end());
+
+      // record the vector's new state
+      m_storage.swap(new_storage);
+      m_size    = old_size + n;
+    } // end else
+  } // end if
+} // end vector_base::fill_insert()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::range_assign(InputIterator first,
+                     InputIterator last)
+{
+  // dispatch on traversal
+  range_assign(first, last,
+    typename thrust::iterator_traversal<InputIterator>::type());
+} // end range_assign()
+
+template<typename T, typename Alloc>
+  template<typename InputIterator>
+    void vector_base<T,Alloc>
+      ::range_assign(InputIterator first,
+                     InputIterator last,
+                     thrust::incrementable_traversal_tag)
+{
+  iterator current(begin());
+
+  // assign to elements which already exist
+  for(; first != last && current != end(); ++current, ++first)
+  {
+    *current = *first;
+  } // end for
+
+  // either just the input was exhausted or both
+  // the input and vector elements were exhausted
+  if(first == last)
+  {
+    // if we exhausted the input, erase leftover elements
+    erase(current, end());
+  } // end if
+  else
+  {
+    // insert the rest of the input at the end of the vector
+    insert(end(), first, last);
+  } // end else
+} // end vector_base::range_assign()
+
+template<typename T, typename Alloc>
+  template<typename RandomAccessIterator>
+    void vector_base<T,Alloc>
+      ::range_assign(RandomAccessIterator first,
+                     RandomAccessIterator last,
+                     thrust::random_access_traversal_tag)
+{
+  const size_type n = thrust::distance(first, last);
+
+  if(n > capacity())
+  {
+    storage_type new_storage(copy_allocator_t(), m_storage);
+    allocate_and_copy(n, first, last, new_storage);
+
+    // call destructors on the elements in the old storage
+    m_storage.destroy(begin(), end());
+
+    // record the vector's new state
+    m_storage.swap(new_storage);
+    m_size = n;
+  } // end if
+  else if(size() >= n)
+  {
+    // we can already accomodate the new range
+    iterator new_end = thrust::copy(first, last, begin());
+
+    // destroy the elements we don't need
+    m_storage.destroy(new_end, end());
+
+    // update size
+    m_size = n;
+  } // end else if
+  else
+  {
+    // range fits inside allocated storage, but some elements
+    // have not been constructed yet
+
+    // XXX TODO we could possibly implement this with one call
+    // to transform rather than copy + uninitialized_copy
+
+    // copy to elements which already exist
+    RandomAccessIterator mid = first;
+    thrust::advance(mid, size());
+    thrust::copy(first, mid, begin());
+
+    // uninitialize_copy to elements which must be constructed
+    m_storage.uninitialized_copy(mid, last, end());
+
+    // update size
+    m_size = n;
+  } // end else
+} // end vector_base::assign()
+
+template<typename T, typename Alloc>
+  void vector_base<T,Alloc>
+    ::fill_assign(size_type n, const T &x)
+{
+  if(n > capacity())
+  {
+    // XXX we should also include a copy of the allocator:
+    // vector_base<T,Alloc> temp(n, x, get_allocator());
+    vector_base<T,Alloc> temp(n, x);
+    temp.swap(*this);
+  } // end if
+  else if(n > size())
+  {
+    // fill to existing elements
+    thrust::fill(begin(), end(), x);
+
+    // construct uninitialized elements
+    m_storage.uninitialized_fill_n(end(), n - size(), x);
+
+    // adjust size
+    m_size += (n - size());
+  } // end else if
+  else
+  {
+    // fill to existing elements
+    iterator new_end = thrust::fill_n(begin(), n, x);
+
+    // erase the elements after the fill
+    erase(new_end, end());
+  } // end else
+} // end vector_base::fill_assign()
+
+template<typename T, typename Alloc>
+  template<typename ForwardIterator>
+    void vector_base<T,Alloc>
+      ::allocate_and_copy(size_type requested_size,
+                          ForwardIterator first, ForwardIterator last,
+                          storage_type &new_storage)
+{
+  if(requested_size == 0)
+  {
+    new_storage.deallocate();
+    return;
+  } // end if
+
+  // allocate exponentially larger new storage
+  size_type allocated_size = thrust::max<size_type>(requested_size, 2 * capacity());
+
+  // do not exceed maximum storage
+  allocated_size = thrust::min<size_type>(allocated_size, max_size());
+
+  if(requested_size > allocated_size)
+  {
+    throw std::length_error("assignment exceeds max_size().");
+  } // end if
+
+  new_storage.allocate(allocated_size);
+
+  try
+  {
+    // construct the range to the newly allocated storage
+    m_storage.uninitialized_copy(first, last, new_storage.begin());
+  } // end try
+  catch(...)
+  {
+    // something went wrong, so destroy & deallocate the new storage
+    // XXX seems like this destroys too many elements -- should just be last - first instead of requested_size
+    iterator new_storage_end = new_storage.begin();
+    thrust::advance(new_storage_end, requested_size);
+    m_storage.destroy(new_storage.begin(), new_storage_end);
+    new_storage.deallocate();
+
+    // rethrow
+    throw;
+  } // end catch
+} // end vector_base::allocate_and_copy()
+
+
+} // end detail
+
+template<typename T, typename Alloc>
+  void swap(detail::vector_base<T,Alloc> &a,
+            detail::vector_base<T,Alloc> &b)
+{
+  a.swap(b);
+} // end swap()
+
+
+
+namespace detail
+{
+
+// iterator tags match
+template <typename InputIterator1, typename InputIterator2>
+bool vector_equal(InputIterator1 first1, InputIterator1 last1,
+                  InputIterator2 first2,
+                  thrust::detail::true_type)
+{
+  return thrust::equal(first1, last1, first2);
+}
+
+// iterator tags differ
+template <typename InputIterator1, typename InputIterator2>
+bool vector_equal(InputIterator1 first1, InputIterator1 last1,
+                  InputIterator2 first2,
+                  thrust::detail::false_type)
+{
+  typename thrust::iterator_difference<InputIterator1>::type n = thrust::distance(first1,last1);
+
+  typedef typename thrust::iterator_system<InputIterator1>::type FromSystem1;
+  typedef typename thrust::iterator_system<InputIterator2>::type FromSystem2;
+
+  // bring both ranges to the host system
+  // note that these copies are no-ops if the range is already convertible to the host system
+  FromSystem1 from_system1;
+  FromSystem2 from_system2;
+  thrust::host_system_tag to_system;
+  thrust::detail::move_to_system<InputIterator1, FromSystem1, thrust::host_system_tag> rng1(from_system1, to_system, first1, last1);
+  thrust::detail::move_to_system<InputIterator2, FromSystem2, thrust::host_system_tag> rng2(from_system2, to_system, first2, first2 + n);
+
+  return thrust::equal(rng1.begin(), rng1.end(), rng2.begin());
+}
+
+template <typename InputIterator1, typename InputIterator2>
+bool vector_equal(InputIterator1 first1, InputIterator1 last1,
+                  InputIterator2 first2)
+{
+  typedef typename thrust::iterator_system<InputIterator1>::type system1;
+  typedef typename thrust::iterator_system<InputIterator2>::type system2;
+
+  // dispatch on the sameness of the two systems
+  return vector_equal(first1, last1, first2,
+    thrust::detail::is_same<system1,system2>());
+}
+
+} // end namespace detail
+
+
+
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator==(const detail::vector_base<T1,Alloc1>& lhs,
+                const detail::vector_base<T2,Alloc2>& rhs)
+{
+    return lhs.size() == rhs.size() && detail::vector_equal(lhs.begin(), lhs.end(), rhs.begin());
+}
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator==(const detail::vector_base<T1,Alloc1>& lhs,
+                const std::vector<T2,Alloc2>&         rhs)
+{
+    return lhs.size() == rhs.size() && detail::vector_equal(lhs.begin(), lhs.end(), rhs.begin());
+}
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator==(const std::vector<T1,Alloc1>&         lhs,
+                const detail::vector_base<T2,Alloc2>& rhs)
+{
+    return lhs.size() == rhs.size() && detail::vector_equal(lhs.begin(), lhs.end(), rhs.begin());
+}
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator!=(const detail::vector_base<T1,Alloc1>& lhs,
+                const detail::vector_base<T2,Alloc2>& rhs)
+{
+    return !(lhs == rhs);
+}
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator!=(const detail::vector_base<T1,Alloc1>& lhs,
+                const std::vector<T2,Alloc2>&         rhs)
+{
+    return !(lhs == rhs);
+}
+
+template<typename T1, typename Alloc1,
+         typename T2, typename Alloc2>
+bool operator!=(const std::vector<T1,Alloc1>&         lhs,
+                const detail::vector_base<T2,Alloc2>& rhs)
+{
+    return !(lhs == rhs);
+}
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_allocator.h
new file mode 100644
index 000000000000..f64c3854f0ee
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_allocator.h
@@ -0,0 +1,140 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file
+ *  \brief An allocator which creates new elements in memory accessible by
+ *  devices.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+#include <thrust/mr/allocator.h>
+#include <thrust/mr/device_memory_resource.h>
+
+#include <limits>
+#include <stdexcept>
+
+THRUST_NAMESPACE_BEGIN
+
+/** \addtogroup allocators Allocators
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! Memory resource adaptor that turns any memory resource that returns a fancy
+ *      with the same tag as \p device_ptr, and adapts it to a resource that returns
+ *      a \p device_ptr.
+ */
+template<typename Upstream>
+class device_ptr_memory_resource final
+    : public thrust::mr::memory_resource<
+        device_ptr<void>
+    >
+{
+    typedef typename Upstream::pointer upstream_ptr;
+
+public:
+    /*! Initialize the adaptor with the global instance of the upstream resource. Obtains
+     *      the global instance by calling \p get_global_resource.
+     */
+    __host__
+    device_ptr_memory_resource() : m_upstream(mr::get_global_resource<Upstream>())
+    {
+    }
+
+    /*! Initialize the adaptor with an upstream resource.
+     *
+     *  \param upstream the upstream memory resource to adapt.
+     */
+    __host__
+    device_ptr_memory_resource(Upstream * upstream) : m_upstream(upstream)
+    {
+    }
+
+    THRUST_NODISCARD __host__
+    virtual pointer do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        return pointer(m_upstream->do_allocate(bytes, alignment).get());
+    }
+
+    __host__
+    virtual void do_deallocate(pointer p, std::size_t bytes, std::size_t alignment) override
+    {
+        m_upstream->do_deallocate(upstream_ptr(p.get()), bytes, alignment);
+    }
+
+private:
+    Upstream * m_upstream;
+};
+
+/*! \brief An allocator which creates new elements in memory accessible by
+ *         devices.
+ *
+ *  \see https://en.cppreference.com/w/cpp/named_req/Allocator
+ */
+template<typename T>
+class device_allocator
+    : public thrust::mr::stateless_resource_allocator<
+        T,
+        device_ptr_memory_resource<device_memory_resource>
+    >
+{
+    typedef thrust::mr::stateless_resource_allocator<
+        T,
+        device_ptr_memory_resource<device_memory_resource>
+    > base;
+
+public:
+    /*! The \p rebind metafunction provides the type of a \p device_allocator
+     *  instantiated with another type.
+     *
+     *  \tparam U the other type to use for instantiation.
+     */
+    template<typename U>
+    struct rebind
+    {
+        /*! The typedef \p other gives the type of the rebound \p device_allocator.
+         */
+        typedef device_allocator<U> other;
+    };
+
+    /*! Default constructor has no effect. */
+    __host__ __device__
+    device_allocator() {}
+
+    /*! Copy constructor has no effect. */
+    __host__ __device__
+    device_allocator(const device_allocator& other) : base(other) {}
+
+    /*! Constructor from other \p device_allocator has no effect. */
+    template<typename U>
+    __host__ __device__
+    device_allocator(const device_allocator<U>& other) : base(other) {}
+
+    device_allocator & operator=(const device_allocator &) = default;
+
+    /*! Destructor has no effect. */
+    __host__ __device__
+    ~device_allocator() {}
+};
+
+/*! \} // allocators
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_delete.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_delete.h
new file mode 100644
index 000000000000..0811936fbbe7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_delete.h
@@ -0,0 +1,53 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Deletes variables in device memory.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+/*! \p device_delete deletes a \p device_ptr allocated with
+ *  \p device_new.
+ *
+ *  \param ptr The \p device_ptr to delete, assumed to have
+ *         been allocated with \p device_new.
+ *  \param n The number of objects to destroy at \p ptr. Defaults to \c 1
+ *         similar to \p device_new.
+ *
+ *  \see device_ptr
+ *  \see device_new
+ */
+template<typename T>
+  inline void device_delete(thrust::device_ptr<T> ptr,
+                            const size_t n = 1);
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/device_delete.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_free.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_free.h
new file mode 100644
index 000000000000..1cd305045e42
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_free.h
@@ -0,0 +1,65 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief Deallocates storage allocated by \p device_malloc.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+/*! \p device_free deallocates memory allocated by the function \p device_malloc.
+ *
+ *  \param ptr A \p device_ptr pointing to memory to be deallocated.
+ *
+ *  The following code snippet demonstrates how to use \p device_free to
+ *  deallocate memory allocated by \p device_malloc.
+ *
+ *  \code
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_free.h>
+ *  ...
+ *  // allocate some integers with device_malloc
+ *  const int N = 100;
+ *  thrust::device_ptr<int> int_array = thrust::device_malloc<int>(N);
+ *
+ *  // manipulate integers
+ *  ...
+ *
+ *  // deallocate with device_free
+ *  thrust::device_free(int_array);
+ *  \endcode
+ *
+ *  \see device_ptr
+ *  \see device_malloc
+ */
+inline void device_free(thrust::device_ptr<void> ptr);
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/device_free.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_make_unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_make_unique.h
new file mode 100644
index 000000000000..cdb8c31d82ff
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_make_unique.h
@@ -0,0 +1,60 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file device_make_unique.h
+ *  \brief A factory function for creating `unique_ptr`s to device objects.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/allocate_unique.h>
+#include <thrust/device_new.h>
+#include <thrust/device_ptr.h>
+#include <thrust/device_allocator.h>
+#include <thrust/detail/type_deduction.h>
+
+THRUST_NAMESPACE_BEGIN
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T, typename... Args>
+__host__
+auto device_make_unique(Args&&... args)
+  THRUST_TRAILING_RETURN(decltype(
+    uninitialized_allocate_unique<T>(device_allocator<T>{})
+  ))
+{
+#if !defined(THRUST_DOXYGEN) // This causes Doxygen to choke for some reason.
+  // FIXME: This is crude - we construct an unnecessary T on the host for
+  // `device_new`. We need a proper dispatched `construct` algorithm to
+  // do this properly.
+  auto p = uninitialized_allocate_unique<T>(device_allocator<T>{});
+  device_new<T>(p.get(), T(THRUST_FWD(args)...));
+  return p;
+#endif
+}
+
+///////////////////////////////////////////////////////////////////////////////
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_malloc.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_malloc.h
new file mode 100644
index 000000000000..790ddbac3a7d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_malloc.h
@@ -0,0 +1,100 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Allocates storage in device memory.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+#include <cstddef> // for std::size_t
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+/*! This version of \p device_malloc allocates sequential device storage
+ *  for bytes.
+ *
+ *  \param n The number of bytes to allocate sequentially
+ *           in device memory.
+ *  \return A \p device_ptr to the newly allocated memory.
+ *
+ *  The following code snippet demonstrates how to use \p device_malloc to
+ *  allocate a range of device memory.
+ *
+ *  \code
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_free.h>
+ *  ...
+ *  // allocate some memory with device_malloc
+ *  const int N = 100;
+ *  thrust::device_ptr<void> void_ptr = thrust::device_malloc(N);
+ *
+ *  // manipulate memory
+ *  ...
+ *
+ *  // deallocate with device_free
+ *  thrust::device_free(void_ptr);
+ *  \endcode
+ *
+ *  \see device_ptr
+ *  \see device_free
+ */
+inline thrust::device_ptr<void> device_malloc(const std::size_t n);
+
+/*! This version of \p device_malloc allocates sequential device storage for
+ *  new objects of the given type.
+ *
+ *  \param n The number of objects of type T to allocate
+ *           sequentially in device memory.
+ *  \return A \p device_ptr to the newly allocated memory.
+ *
+ *  The following code snippet demonstrates how to use \p device_malloc to
+ *  allocate a range of device memory.
+ *
+ *  \code
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_free.h>
+ *  ...
+ *  // allocate some integers with device_malloc
+ *  const int N = 100;
+ *  thrust::device_ptr<int> int_array = thrust::device_malloc<int>(N);
+ *
+ *  // manipulate integers
+ *  ...
+ *
+ *  // deallocate with device_free
+ *  thrust::device_free(int_array);
+ *  \endcode
+ *
+ *  \see device_ptr
+ *  \see device_free
+ */
+template<typename T>
+  inline thrust::device_ptr<T> device_malloc(const std::size_t n);
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/device_malloc.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_malloc_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_malloc_allocator.h
new file mode 100644
index 000000000000..1b15045f2f8b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_malloc_allocator.h
@@ -0,0 +1,180 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief An allocator which allocates storage with \p device_malloc.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+#include <thrust/device_reference.h>
+#include <thrust/device_malloc.h>
+#include <thrust/device_free.h>
+#include <limits>
+#include <stdexcept>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declarations to WAR circular #includes
+template<typename> class device_ptr;
+template<typename T> device_ptr<T> device_malloc(const std::size_t n);
+
+/*! \addtogroup allocators Allocators 
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! \p device_malloc_allocator is a device memory allocator that employs the
+ *  \p device_malloc function for allocation.
+ *
+ *  \p device_malloc_allocator is deprecated in favor of <tt>thrust::mr</tt>
+ *      memory resource-based allocators.
+ *
+ *  \see device_malloc
+ *  \see device_ptr
+ *  \see device_allocator
+ *  \see https://en.cppreference.com/w/cpp/memory/allocator
+ */
+template<typename T>
+  class device_malloc_allocator
+{
+  public:
+    /*! Type of element allocated, \c T. */
+    typedef T                                 value_type;
+
+    /*! Pointer to allocation, \c device_ptr<T>. */
+    typedef device_ptr<T>                     pointer;
+
+    /*! \c const pointer to allocation, \c device_ptr<const T>. */
+    typedef device_ptr<const T>               const_pointer;
+
+    /*! Reference to allocated element, \c device_reference<T>. */
+    typedef device_reference<T>               reference;
+
+    /*! \c const reference to allocated element, \c device_reference<const T>. */
+    typedef device_reference<const T>         const_reference;
+
+    /*! Type of allocation size, \c std::size_t. */
+    typedef std::size_t                       size_type;
+
+    /*! Type of allocation difference, \c pointer::difference_type. */
+    typedef typename pointer::difference_type difference_type;
+
+    /*! The \p rebind metafunction provides the type of a \p device_malloc_allocator
+     *  instantiated with another type.
+     *
+     *  \tparam U The other type to use for instantiation.
+     */
+    template<typename U>
+      struct rebind
+    {
+      /*! The typedef \p other gives the type of the rebound \p device_malloc_allocator.
+       */
+      typedef device_malloc_allocator<U> other;
+    }; // end rebind
+
+    /*! No-argument constructor has no effect. */
+    __host__ __device__
+    inline device_malloc_allocator() {}
+
+    /*! No-argument destructor has no effect. */
+    __host__ __device__
+    inline ~device_malloc_allocator() {}
+
+    /*! Copy constructor has no effect. */
+    __host__ __device__
+    inline device_malloc_allocator(device_malloc_allocator const&) {}
+
+    /*! Constructor from other \p device_malloc_allocator has no effect. */
+    template<typename U>
+    __host__ __device__
+    inline device_malloc_allocator(device_malloc_allocator<U> const&) {}
+
+#if THRUST_CPP_DIALECT >= 2011
+    device_malloc_allocator & operator=(const device_malloc_allocator &) = default;
+#endif
+
+    /*! Returns the address of an allocated object.
+     *  \return <tt>&r</tt>.
+     */
+    __host__ __device__
+    inline pointer address(reference r) { return &r; }
+
+    /*! Returns the address an allocated object.
+     *  \return <tt>&r</tt>.
+     */
+    __host__ __device__
+    inline const_pointer address(const_reference r) { return &r; }
+
+    /*! Allocates storage for \p cnt objects.
+     *  \param cnt The number of objects to allocate.
+     *  \return A \p pointer to uninitialized storage for \p cnt objects.
+     *  \note Memory allocated by this function must be deallocated with \p deallocate.
+     */
+    __host__
+    inline pointer allocate(size_type cnt,
+                            const_pointer = const_pointer(static_cast<T*>(0)))
+    {
+      if(cnt > this->max_size())
+      {
+        throw std::bad_alloc();
+      } // end if
+
+      return pointer(device_malloc<T>(cnt));
+    } // end allocate()
+
+    /*! Deallocates storage for objects allocated with \p allocate.
+     *  \param p A \p pointer to the storage to deallocate.
+     *  \param cnt The size of the previous allocation.
+     *  \note Memory deallocated by this function must previously have been
+     *        allocated with \p allocate.
+     */
+    __host__
+    inline void deallocate(pointer p, size_type cnt)
+    {
+      // silence unused parameter warning while still leaving the parameter name for Doxygen
+      (void)(cnt);
+
+      device_free(p);
+    } // end deallocate()
+
+    /*! Returns the largest value \c n for which <tt>allocate(n)</tt> might succeed.
+     *  \return The largest value \c n for which <tt>allocate(n)</tt> might succeed.
+     */
+    inline size_type max_size() const
+    {
+      return (std::numeric_limits<size_type>::max)() / sizeof(T);
+    } // end max_size()
+
+    /*! Compares against another \p device_malloc_allocator for equality.
+     *  \return \c true
+     */
+    __host__ __device__
+    inline bool operator==(device_malloc_allocator const&) const { return true; }
+
+    /*! Compares against another \p device_malloc_allocator for inequality.
+     *  \return \c false
+     */
+    __host__ __device__
+    inline bool operator!=(device_malloc_allocator const &a) const {return !operator==(a); }
+}; // end device_malloc_allocator
+
+/*! \} // allocators
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_new.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_new.h
new file mode 100644
index 000000000000..c615e58f2d6a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_new.h
@@ -0,0 +1,86 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file device_new.h
+ *  \brief Constructs new elements in device memory
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include this for size_t
+#include <cstddef>
+#include <thrust/device_ptr.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*!
+ *  \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+/*! \p device_new implements the placement \c new operator for types
+ *  resident in device memory. \p device_new calls <tt>T</tt>'s null
+ *  constructor on a array of objects in device memory.
+ *  No memory is allocated by this function.
+ *
+ *  \param  p A \p device_ptr to a region of device memory into which
+ *          to construct one or many <tt>T</tt>s.
+ *  \param  n The number of objects to construct at \p p.
+ *  \return p, casted to <tt>T</tt>'s type.
+ *
+ *  \see device_ptr
+ */
+template <typename T>
+  device_ptr<T> device_new(device_ptr<void> p,
+                           const size_t n = 1);
+
+/*! \p device_new implements the placement new operator for types
+ *  resident in device memory. \p device_new calls <tt>T</tt>'s copy
+ *  constructor on a array of objects in device memory. No memory is
+ *  allocated by this function.
+ *
+ *  \param  p A \p device_ptr to a region of device memory into which to
+ *          construct one or many <tt>T</tt>s.
+ *  \param exemplar The value from which to copy.
+ *  \param  n The number of objects to construct at \p p.
+ *  \return p, casted to <tt>T</tt>'s type.
+ *
+ *  \see device_ptr
+ *  \see fill
+ */
+template <typename T>
+  device_ptr<T> device_new(device_ptr<void> p,
+                           const T &exemplar,
+                           const size_t n = 1);
+
+/*! \p device_new implements the new operator for types resident in device memory.
+ *  It allocates device memory large enough to hold \p n new objects of type \c T.
+ *
+ *  \param n The number of objects to allocate. Defaults to \c 1.
+ *  \return A \p device_ptr to the newly allocated region of device memory.
+ */
+template <typename T>
+  device_ptr<T> device_new(const size_t n = 1);
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/device_new.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_new_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_new_allocator.h
new file mode 100644
index 000000000000..c9c6b0e9507f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_new_allocator.h
@@ -0,0 +1,172 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief An allocator which allocates storage with \p device_new.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+#include <thrust/device_reference.h>
+#include <thrust/device_new.h>
+#include <thrust/device_delete.h>
+
+#include <cuda/std/cstdint>
+#include <cuda/std/limits>
+
+#include <stdexcept>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup allocators Allocators
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! \p device_new_allocator is a device memory allocator that employs the
+ *  \p device_new function for allocation.
+ *
+ *  \see device_new
+ *  \see device_ptr
+ *  \see https://en.cppreference.com/w/cpp/memory/allocator
+ */
+template<typename T>
+  class device_new_allocator
+{
+  public:
+    /*! Type of element allocated, \c T. */
+    typedef T                                 value_type;
+
+    /*! Pointer to allocation, \c device_ptr<T>. */
+    typedef device_ptr<T>                     pointer;
+
+    /*! \c const pointer to allocation, \c device_ptr<const T>. */
+    typedef device_ptr<const T>               const_pointer;
+
+    /*! Reference to allocated element, \c device_reference<T>. */
+    typedef device_reference<T>               reference;
+
+    /*! \c const reference to allocated element, \c device_reference<const T>. */
+    typedef device_reference<const T>         const_reference;
+
+    /*! Type of allocation size, \c ::cuda::std::size_t. */
+    typedef ::cuda::std::size_t                 size_type;
+
+    /*! Type of allocation difference, \c pointer::difference_type. */
+    typedef typename pointer::difference_type difference_type;
+
+    /*! The \p rebind metafunction provides the type of a \p device_new_allocator
+     *  instantiated with another type.
+     *
+     *  \tparam U The other type to use for instantiation.
+     */
+    template<typename U>
+      struct rebind
+    {
+      /*! The typedef \p other gives the type of the rebound \p device_new_allocator.
+       */
+      typedef device_new_allocator<U> other;
+    }; // end rebind
+
+    /*! No-argument constructor has no effect. */
+    __host__ __device__
+    inline device_new_allocator() {}
+
+    /*! No-argument destructor has no effect. */
+    __host__ __device__
+    inline ~device_new_allocator() {}
+
+    /*! Copy constructor has no effect. */
+    __host__ __device__
+    inline device_new_allocator(device_new_allocator const&) {}
+
+    /*! Constructor from other \p device_malloc_allocator has no effect. */
+    template<typename U>
+    __host__ __device__
+    inline device_new_allocator(device_new_allocator<U> const&) {}
+
+    /*! Returns the address of an allocated object.
+     *  \return <tt>&r</tt>.
+     */
+    __host__ __device__
+    inline pointer address(reference r) { return &r; }
+    
+    /*! Returns the address an allocated object.
+     *  \return <tt>&r</tt>.
+     */
+    __host__ __device__
+    inline const_pointer address(const_reference r) { return &r; }
+
+    /*! Allocates storage for \p cnt objects.
+     *  \param cnt The number of objects to allocate.
+     *  \return A \p pointer to uninitialized storage for \p cnt objects.
+     *  \note Memory allocated by this function must be deallocated with \p deallocate.
+     */
+    __host__
+    inline pointer allocate(size_type cnt,
+                            const_pointer = const_pointer(static_cast<T*>(0)))
+    {
+      if(cnt > this->max_size())
+      {
+        throw std::bad_alloc();
+      } // end if
+
+      // use "::operator new" rather than keyword new
+      return pointer(device_new<T>(cnt));
+    } // end allocate()
+
+    /*! Deallocates storage for objects allocated with \p allocate.
+     *  \param p A \p pointer to the storage to deallocate.
+     *  \param cnt The size of the previous allocation.
+     *  \note Memory deallocated by this function must previously have been
+     *        allocated with \p allocate.
+     */
+    __host__
+    inline void deallocate(pointer p, size_type cnt)
+    {
+      // use "::operator delete" rather than keyword delete
+      (void)cnt;
+      device_delete(p);
+    } // end deallocate()
+
+    /*! Returns the largest value \c n for which <tt>allocate(n)</tt> might succeed.
+     *  \return The largest value \c n for which <tt>allocate(n)</tt> might succeed.
+     */
+    __host__ __device__
+    inline size_type max_size() const
+    {
+      return ::cuda::std::numeric_limits<size_type>::max THRUST_PREVENT_MACRO_SUBSTITUTION () / sizeof(T);
+    } // end max_size()
+
+    /*! Compares against another \p device_malloc_allocator for equality.
+     *  \return \c true
+     */
+    __host__ __device__
+    inline bool operator==(device_new_allocator const&) { return true; }
+
+    /*! Compares against another \p device_malloc_allocator for inequality.
+     *  \return \c false
+     */
+    __host__ __device__
+    inline bool operator!=(device_new_allocator const &a) {return !operator==(a); }
+}; // end device_new_allocator
+
+/*! \} // allocators
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_ptr.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_ptr.h
new file mode 100644
index 000000000000..5ef4aa464fcd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_ptr.h
@@ -0,0 +1,211 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief A pointer to an object which resides in memory associated with the
+ *  \c device system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/memory.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+template <typename T> class device_reference;
+
+/*! \brief \c device_ptr is a pointer-like object which points to an object that
+ *  resides in memory associated with the \ref device system.
+ *
+ *  \c device_ptr has pointer semantics: it may be dereferenced safely from
+ *  anywhere, including the \ref host, and may be manipulated with pointer
+ *  arithmetic.
+ *
+ *  \c device_ptr can be created with \ref device_new, \ref device_malloc,
+ *  \ref device_malloc_allocator, \ref device_allocator, or
+ *  \ref device_pointer_cast, or by explicitly calling its constructor with a
+ *  raw pointer.
+ *
+ *  The raw pointer contained in a \c device_ptr may be obtained via \c get
+ *  member function or the \ref raw_pointer_cast free function.
+ *
+ *  \ref algorithms operating on \c device_ptr types will automatically be
+ *  dispatched to the \ref device system.
+ *
+ *  \note \c device_ptr is not a smart pointer; it is the programmer's
+ *  responsibility to deallocate memory pointed to by \c device_ptr.
+ *
+ *  \see device_new
+ *  \see device_malloc
+ *  \see device_malloc_allocator
+ *  \see device_allocator
+ *  \see device_pointer_cast
+ *  \see raw_pointer_cast
+ */
+template <typename T>
+class device_ptr
+  : public thrust::pointer<
+      T,
+      thrust::device_system_tag,
+      thrust::device_reference<T>,
+      thrust::device_ptr<T>
+    >
+{
+  private:
+    using super_t = thrust::pointer<
+      T,
+      thrust::device_system_tag,
+      thrust::device_reference<T>,
+      thrust::device_ptr<T>
+    >;
+
+  public:
+    /*! \brief Construct a null \c device_ptr.
+     *
+     *  \post <tt>get() == nullptr</tt>.
+     */
+    __host__ __device__
+    device_ptr() : super_t() {}
+
+    /*! \brief Construct a null \c device_ptr.
+     *
+     *  \param ptr A null pointer.
+     *
+     *  \post <tt>get() == nullptr</tt>.
+     */
+    __host__ __device__
+    device_ptr(std::nullptr_t) : super_t(nullptr) {}
+
+    /*! \brief Construct a \c device_ptr from a raw pointer which is
+     *  convertible to \c T*.
+     *
+     *  \tparam U   A type whose pointer is convertible to \c T*.
+     *  \param  ptr A raw pointer to a \c U in device memory to construct from.
+     *
+     *  \pre <tt>std::is_convertible_v<U*, T*> == true</tt>.
+     *
+     *  \pre \c ptr points to a location in device memory.
+     *
+     *  \post <tt>get() == nullptr</tt>.
+     */
+    template <typename U>
+    __host__ __device__
+    explicit device_ptr(U* ptr) : super_t(ptr) {}
+
+    /*! \brief Copy construct a \c device_ptr from another \c device_ptr whose
+     *  pointer type is convertible to \c T*.
+     *
+     *  \tparam U     A type whose pointer is convertible to \c T*.
+     *  \param  other A \c device_ptr to a \c U to construct from.
+     *
+     *  \pre <tt>std::is_convertible_v<U*, T*> == true</tt>.
+     *
+     *  \post <tt>get() == other.get()</tt>.
+     */
+    template <typename U>
+    __host__ __device__
+    device_ptr(device_ptr<U> const& other) : super_t(other) {}
+
+    /*! \brief Set this \c device_ptr to point to the same object as another
+     *  \c device_ptr whose pointer type is convertible to \c T*.
+     *
+     *  \tparam U     A type whose pointer is convertible to \c T*.
+     *  \param  other A \c device_ptr to a \c U to assign from.
+     *
+     *  \pre <tt>std::is_convertible_v<U*, T*> == true</tt>.
+     *
+     *  \post <tt>get() == other.get()</tt>.
+     *
+     *  \return \c *this.
+     */
+    template <typename U>
+    __host__ __device__
+    device_ptr &operator=(device_ptr<U> const& other)
+    {
+      super_t::operator=(other);
+      return *this;
+    }
+
+    /*! \brief Set this \c device_ptr to null.
+     *
+     *  \param ptr A null pointer.
+     *
+     *  \post <tt>get() == nullptr</tt>.
+     *
+     *  \return \c *this.
+     */
+    __host__ __device__
+    device_ptr& operator=(std::nullptr_t)
+    {
+      super_t::operator=(nullptr);
+      return *this;
+    }
+
+#if THRUST_DOXYGEN
+    /*! \brief Return the raw pointer that this \c device_ptr points to.
+     */
+    __host__ __device__
+    T* get() const;
+#endif
+};
+
+#if THRUST_DOXYGEN
+/*! Write the address that a \c device_ptr points to to an output stream.
+ *
+ *  \param os The output stream.
+ *  \param dp The \c device_ptr to output.
+ *
+ *  \return \c os.
+ */
+template <typename T, typename CharT, typename Traits>
+__host__ std::basic_ostream<CharT, Traits>&
+operator<<(std::basic_ostream<CharT, Traits>& os, device_ptr<T> const& dp);
+#endif
+
+/*! \brief Create a \c device_ptr from a raw pointer.
+ *
+ *  \tparam T   Any type.
+ *  \param  ptr A raw pointer to a \c T in device memory.
+ *
+ *  \pre \c ptr points to a location in device memory.
+ *
+ *  \return A \c device_ptr<T> pointing to \c ptr.
+ */
+template <typename T>
+__host__ __device__
+device_ptr<T> device_pointer_cast(T* ptr);
+
+/*! \brief Create a \c device_ptr from another \c device_ptr.
+ *
+ *  \tparam T    Any type.
+ *  \param  dptr A \c device_ptr to a \c T.
+ */
+template<typename T>
+__host__ __device__
+device_ptr<T> device_pointer_cast(device_ptr<T> const& dptr);
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/device_ptr.inl>
+#include <thrust/detail/raw_pointer_cast.h>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_reference.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_reference.h
new file mode 100644
index 000000000000..512ab4c60c36
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_reference.h
@@ -0,0 +1,987 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A reference to an object which resides in memory associated with the
+ *  device system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/device_ptr.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/reference.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+/*! \p device_reference acts as a reference-like object to an object stored in device memory.
+ *  \p device_reference is not intended to be used directly; rather, this type
+ *  is the result of deferencing a \p device_ptr. Similarly, taking the address of
+ *  a \p device_reference yields a \p device_ptr.
+ *
+ *  \p device_reference may often be used from host code in place of operations defined on
+ *  its associated \c value_type. For example, when \p device_reference refers to an
+ *  arithmetic type, arithmetic operations on it are legal:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *
+ *  int main(void)
+ *  {
+ *    thrust::device_vector<int> vec(1, 13);
+ *
+ *    thrust::device_reference<int> ref_to_thirteen = vec[0];
+ *
+ *    int x = ref_to_thirteen + 1;
+ *
+ *    // x is 14
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  Similarly, we can print the value of \c ref_to_thirteen in the above code by using an
+ *  \c iostream:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <iostream>
+ *
+ *  int main(void)
+ *  {
+ *    thrust::device_vector<int> vec(1, 13);
+ *
+ *    thrust::device_reference<int> ref_to_thirteen = vec[0];
+ *
+ *    std::cout << ref_to_thirteen << std::endl;
+ *
+ *    // 13 is printed
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  Of course, we needn't explicitly create a \p device_reference in the previous
+ *  example, because one is returned by \p device_vector's bracket operator. A more natural
+ *  way to print the value of a \p device_vector element might be:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <iostream>
+ *
+ *  int main(void)
+ *  {
+ *    thrust::device_vector<int> vec(1, 13);
+ *
+ *    std::cout << vec[0] << std::endl;
+ *
+ *    // 13 is printed
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  These kinds of operations should be used sparingly in performance-critical code, because
+ *  they imply a potentially expensive copy between host and device space.
+ *
+ *  Some operations which are possible with regular objects are impossible with their
+ *  corresponding \p device_reference objects due to the requirements of the C++ language. For
+ *  example, because the member access operator cannot be overloaded, member variables and functions
+ *  of a referent object cannot be directly accessed through its \p device_reference.
+ *
+ *  The following code, which generates a compiler error, illustrates:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct foo
+ *  {
+ *    int x;
+ *  };
+ *
+ *  int main(void)
+ *  {
+ *    thrust::device_vector<foo> foo_vec(1);
+ *
+ *    thrust::device_reference<foo> foo_ref = foo_vec[0];
+ *
+ *    foo_ref.x = 13; // ERROR: x cannot be accessed through foo_ref
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  Instead, a host space copy must be created to access \c foo's \c x member:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct foo
+ *  {
+ *    int x;
+ *  };
+ *
+ *  int main(void)
+ *  {
+ *    thrust::device_vector<foo> foo_vec(1);
+ *
+ *    // create a local host-side foo object
+ *    foo host_foo;
+ *    host_foo.x = 13;
+ *
+ *    thrust::device_reference<foo> foo_ref = foo_vec[0];
+ *
+ *    foo_ref = host_foo;
+ *
+ *    // foo_ref's x member is 13
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  Another common case where a \p device_reference cannot directly be used in place of
+ *  its referent object occurs when passing them as parameters to functions like \c printf
+ *  which have varargs parameters. Because varargs parameters must be Plain Old Data, a
+ *  \p device_reference to a POD type requires a cast when passed to \c printf:
+ *
+ *  \code
+ *  #include <stdio.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  int main(void)
+ *  {
+ *    thrust::device_vector<int> vec(1,13);
+ *
+ *    // vec[0] must be cast to int when passing to printf
+ *    printf("%d\n", (int) vec[0]);
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see device_ptr
+ *  \see device_vector
+ */
+template<typename T>
+  class device_reference
+    : public thrust::reference<
+               T,
+               thrust::device_ptr<T>,
+               thrust::device_reference<T>
+             >
+{
+  private:
+    typedef thrust::reference<
+      T,
+      thrust::device_ptr<T>,
+      thrust::device_reference<T>
+    > super_t;
+
+  public:
+    /*! The type of the value referenced by this type of \p device_reference.
+     */
+    typedef typename super_t::value_type value_type;
+
+    /*! The type of the expression <tt>&ref</tt>, where <tt>ref</tt> is a \p device_reference.
+     */
+    typedef typename super_t::pointer    pointer;
+
+    /*! This copy constructor accepts a const reference to another
+     *  \p device_reference. After this \p device_reference is constructed,
+     *  it shall refer to the same object as \p other.
+     *
+     *  \param other A \p device_reference to copy from.
+     *
+     *  The following code snippet demonstrates the semantics of this
+     *  copy constructor.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_reference<int> ref = v[0];
+     *
+     *  // ref equals the object at v[0]
+     *  assert(ref == v[0]);
+     *
+     *  // the address of ref equals the address of v[0]
+     *  assert(&ref == &v[0]);
+     *
+     *  // modifying v[0] modifies ref
+     *  v[0] = 13;
+     *  assert(ref == 13);
+     *  \endcode
+     *
+     *  \note This constructor is templated primarily to allow initialization of
+     *  <tt>device_reference<const T></tt> from <tt>device_reference<T></tt>.
+     */
+    template<typename OtherT>
+    __host__ __device__
+    device_reference(const device_reference<OtherT> &other,
+                     typename thrust::detail::enable_if_convertible<
+                       typename device_reference<OtherT>::pointer,
+                       pointer
+                     >::type * = 0)
+      : super_t(other)
+    {}
+
+    /*! This copy constructor initializes this \p device_reference
+     *  to refer to an object pointed to by the given \p device_ptr. After
+     *  this \p device_reference is constructed, it shall refer to the
+     *  object pointed to by \p ptr.
+     *
+     *  \param ptr A \p device_ptr to copy from.
+     *
+     *  The following code snippet demonstrates the semantic of this
+     *  copy constructor.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals the object pointed to by ptr
+     *  assert(ref == *ptr);
+     *
+     *  // the address of ref equals ptr
+     *  assert(&ref == ptr);
+     *
+     *  // modifying *ptr modifies ref
+     *  *ptr = 13;
+     *  assert(ref == 13);
+     *  \endcode
+     */
+    __host__ __device__
+    explicit device_reference(const pointer &ptr)
+      : super_t(ptr)
+    {}
+
+    /*! This assignment operator assigns the value of the object referenced by
+     *  the given \p device_reference to the object referenced by this
+     *  \p device_reference.
+     *
+     *  \param other The \p device_reference to assign from.
+     *  \return <tt>*this</tt>
+     */
+    template<typename OtherT>
+    __host__ __device__
+    device_reference &operator=(const device_reference<OtherT> &other)
+    {
+      return super_t::operator=(other);
+    }
+
+    /*! Assignment operator assigns the value of the given value to the
+     *  value referenced by this \p device_reference.
+     *
+     *  \param x The value to assign from.
+     *  \return <tt>*this</tt>
+     */
+    __host__ __device__
+    device_reference &operator=(const value_type &x)
+    {
+      return super_t::operator=(x);
+    }
+
+// declare these members for the purpose of Doxygenating them
+// they actually exist in a derived-from class
+#if 0
+    /*! Address-of operator returns a \p device_ptr pointing to the object
+     *  referenced by this \p device_reference. It does not return the
+     *  address of this \p device_reference.
+     *
+     *  \return A \p device_ptr pointing to the object this
+     *  \p device_reference references.
+     */
+    __host__ __device__
+    pointer operator&(void) const;
+
+    /*! Conversion operator converts this \p device_reference to T
+     *  by returning a copy of the object referenced by this
+     *  \p device_reference.
+     *
+     *  \return A copy of the object referenced by this \p device_reference.
+     */
+    __host__ __device__
+    operator value_type (void) const;
+
+    /*! swaps the value this \p device_reference references with another.
+     *  \p other The other \p device_reference with which to swap.
+     */
+    __host__ __device__
+    void swap(device_reference &other);
+
+    /*! Prefix increment operator increments the object referenced by this
+     *  \p device_reference.
+     *
+     *  \return <tt>*this</tt>
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's prefix increment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *
+     *  // increment ref
+     *  ++ref;
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *  \endcode
+     *
+     *  \note The increment executes as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator++(void);
+
+    /*! Postfix increment operator copies the object referenced by this
+     *  \p device_reference, increments the object referenced by this
+     *  \p device_reference, and returns the copy.
+     *
+     *  \return A copy of the object referenced by this \p device_reference
+     *          before being incremented.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's postfix increment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *
+     *  // increment ref
+     *  int x = ref++;
+     *
+     *  // x equals 0
+     *  assert(x == 0)
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *  \endcode
+     *
+     *  \note The increment executes as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    value_type operator++(int);
+
+    /*! Addition assignment operator add-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the add-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's addition assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *
+     *  // add-assign ref
+     *  ref += 5;
+     *
+     *  // ref equals 5
+     *  assert(ref == 5);
+     *
+     *  // the object pointed to by ptr equals 5
+     *  assert(*ptr == 5);
+     *
+     *  // v[0] equals 5
+     *  assert(v[0] == 5);
+     *  \endcode
+     *
+     *  \note The add-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator+=(const T &rhs);
+
+    /*! Prefix decrement operator decrements the object referenced by this
+     *  \p device_reference.
+     *
+     *  \return <tt>*this</tt>
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's prefix decrement operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *
+     *  // decrement ref
+     *  --ref;
+     *
+     *  // ref equals -1
+     *  assert(ref == -1);
+     *
+     *  // the object pointed to by ptr equals -1
+     *  assert(*ptr == -1);
+     *
+     *  // v[0] equals -1
+     *  assert(v[0] == -1);
+     *  \endcode
+     *
+     *  \note The decrement executes as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator--(void);
+
+    /*! Postfix decrement operator copies the object referenced by this
+     *  \p device_reference, decrements the object referenced by this
+     *  \p device_reference, and returns the copy.
+     *
+     *  \return A copy of the object referenced by this \p device_reference
+     *          before being decremented.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's postfix decrement operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *
+     *  // decrement ref
+     *  int x = ref--;
+     *
+     *  // x equals 0
+     *  assert(x == 0)
+     *
+     *  // ref equals -1
+     *  assert(ref == -1);
+     *
+     *  // the object pointed to by ptr equals -1
+     *  assert(*ptr == -1);
+     *
+     *  // v[0] equals -1
+     *  assert(v[0] == -1);
+     *  \endcode
+     *
+     *  \note The decrement executes as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    value_type operator--(int);
+
+    /*! Subtraction assignment operator subtract-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the subtraction-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's addition assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *
+     *  // subtract-assign ref
+     *  ref -= 5;
+     *
+     *  // ref equals -5
+     *  assert(ref == -5);
+     *
+     *  // the object pointed to by ptr equals -5
+     *  assert(*ptr == -5);
+     *
+     *  // v[0] equals -5
+     *  assert(v[0] == -5);
+     *  \endcode
+     *
+     *  \note The subtract-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator-=(const T &rhs);
+
+    /*! Multiplication assignment operator multiply-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the multiply-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's multiply assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,1);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *
+     *  // multiply-assign ref
+     *  ref *= 5;
+     *
+     *  // ref equals 5
+     *  assert(ref == 5);
+     *
+     *  // the object pointed to by ptr equals 5
+     *  assert(*ptr == 5);
+     *
+     *  // v[0] equals 5
+     *  assert(v[0] == 5);
+     *  \endcode
+     *
+     *  \note The multiply-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator*=(const T &rhs);
+
+    /*! Division assignment operator divide-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the divide-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's divide assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,5);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 5
+     *  assert(ref == 5);
+     *
+     *  // the object pointed to by ptr equals 5
+     *  assert(*ptr == 5);
+     *
+     *  // v[0] equals 5
+     *  assert(v[0] == 5);
+     *
+     *  // divide-assign ref
+     *  ref /= 5;
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *  \endcode
+     *
+     *  \note The divide-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator/=(const T &rhs);
+
+    /*! Modulation assignment operator modulus-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the divide-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's divide assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,5);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 5
+     *  assert(ref == 5);
+     *
+     *  // the object pointed to by ptr equals 5
+     *  assert(*ptr == 5);
+     *
+     *  // v[0] equals 5
+     *  assert(v[0] == 5);
+     *
+     *  // modulus-assign ref
+     *  ref %= 5;
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *  \endcode
+     *
+     *  \note The modulus-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator%=(const T &rhs);
+
+    /*! Bitwise left shift assignment operator left shift-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the left shift-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's left shift assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,1);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *
+     *  // left shift-assign ref
+     *  ref <<= 1;
+     *
+     *  // ref equals 2
+     *  assert(ref == 2);
+     *
+     *  // the object pointed to by ptr equals 2
+     *  assert(*ptr == 2);
+     *
+     *  // v[0] equals 2
+     *  assert(v[0] == 2);
+     *  \endcode
+     *
+     *  \note The left shift-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator<<=(const T &rhs);
+
+    /*! Bitwise right shift assignment operator right shift-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the right shift-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's right shift assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,2);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 2
+     *  assert(ref == 2);
+     *
+     *  // the object pointed to by ptr equals 2
+     *  assert(*ptr == 2);
+     *
+     *  // v[0] equals 2
+     *  assert(v[0] == 2);
+     *
+     *  // right shift-assign ref
+     *  ref >>= 1;
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *  \endcode
+     *
+     *  \note The right shift-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator>>=(const T &rhs);
+
+    /*! Bitwise AND assignment operator AND-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the AND-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's AND assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,1);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *
+     *  // right AND-assign ref
+     *  ref &= 0;
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *  \endcode
+     *
+     *  \note The AND-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator&=(const T &rhs);
+
+    /*! Bitwise OR assignment operator OR-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the OR-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's OR assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,0);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *
+     *  // right OR-assign ref
+     *  ref |= 1;
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *  \endcode
+     *
+     *  \note The OR-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator|=(const T &rhs);
+
+    /*! Bitwise XOR assignment operator XOR-assigns the object referenced by this
+     *  \p device_reference and returns this \p device_reference.
+     *
+     *  \param rhs The right hand side of the XOR-assignment.
+     *  \return <tt>*this</tt>.
+     *
+     *  The following code snippet demonstrates the semantics of
+     *  \p device_reference's XOR assignment operator.
+     *
+     *  \code
+     *  #include <thrust/device_vector.h>
+     *  #include <assert.h>
+     *  ...
+     *  thrust::device_vector<int> v(1,1);
+     *  thrust::device_ptr<int> ptr = &v[0];
+     *  thrust::device_reference<int> ref(ptr);
+     *
+     *  // ref equals 1
+     *  assert(ref == 1);
+     *
+     *  // the object pointed to by ptr equals 1
+     *  assert(*ptr == 1);
+     *
+     *  // v[0] equals 1
+     *  assert(v[0] == 1);
+     *
+     *  // right XOR-assign ref
+     *  ref ^= 1;
+     *
+     *  // ref equals 0
+     *  assert(ref == 0);
+     *
+     *  // the object pointed to by ptr equals 0
+     *  assert(*ptr == 0);
+     *
+     *  // v[0] equals 0
+     *  assert(v[0] == 0);
+     *  \endcode
+     *
+     *  \note The XOR-assignment executes as as if it were executed on the host.
+     *  This may change in a later version.
+     */
+    device_reference &operator^=(const T &rhs);
+#endif // end doxygen-only members
+}; // end device_reference
+
+/*! swaps the value of one \p device_reference with another.
+ *  \p x The first \p device_reference of interest.
+ *  \p y The second \p device_reference of interest.
+ */
+template<typename T>
+__host__ __device__
+void swap(device_reference<T>& x, device_reference<T>& y)
+{
+  x.swap(y);
+}
+
+// declare these methods for the purpose of Doxygenating them
+// they actually are defined for a derived-from class
+#if THRUST_DOXYGEN
+/*! Writes to an output stream the value of a \p device_reference.
+ *
+ *  \param os The output stream.
+ *  \param y The \p device_reference to output.
+ *  \return os.
+ */
+template<typename T, typename charT, typename traits>
+std::basic_ostream<charT, traits> &
+operator<<(std::basic_ostream<charT, traits> &os, const device_reference<T> &y);
+#endif
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/device_vector.h b/thirdparty/manifold/thirdparty/thrust/thrust/device_vector.h
new file mode 100644
index 000000000000..9b97e8d70335
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/device_vector.h
@@ -0,0 +1,511 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file
+ *  \brief A dynamically-sizable array of elements which resides in memory
+ *         accessible to devices.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/vector_base.h>
+#include <thrust/device_allocator.h>
+
+#include <initializer_list>
+#include <vector>
+#include <utility>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup containers Containers
+ *  \{
+ */
+
+/*! A \p device_vector is a container that supports random access to elements,
+ *  constant time removal of elements at the end, and linear time insertion
+ *  and removal of elements at the beginning or in the middle. The number of
+ *  elements in a \p device_vector may vary dynamically; memory management is
+ *  automatic. The memory associated with a \p device_vector resides in the
+ *  memory accessible to devices.
+ *
+ *  \see https://en.cppreference.com/w/cpp/container/vector
+ *  \see device_allocator
+ *  \see host_vector
+ *  \see universal_vector
+ */
+template<typename T, typename Alloc = thrust::device_allocator<T> >
+  class device_vector
+    : public detail::vector_base<T,Alloc>
+{
+  private:
+    typedef detail::vector_base<T,Alloc> Parent;
+
+  public:
+    /*! \cond
+     */
+    typedef typename Parent::size_type  size_type;
+    typedef typename Parent::value_type value_type;
+    /*! \endcond
+     */
+
+    /*! This constructor creates an empty \p device_vector.
+     */
+    device_vector(void)
+      :Parent() {}
+
+    /*! This constructor creates an empty \p device_vector.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    device_vector(const Alloc &alloc)
+      :Parent(alloc) {}
+
+    /*! The destructor erases the elements.
+     */
+    //  Define an empty destructor to explicitly specify
+    //  its execution space qualifier, as a workaround for nvcc warning
+    ~device_vector(void) {}
+
+    /*! This constructor creates a \p device_vector with the given
+     *  size.
+     *  \param n The number of elements to initially create.
+     */
+    explicit device_vector(size_type n)
+      :Parent(n) {}
+
+    /*! This constructor creates a \p device_vector with the given
+     *  size.
+     *  \param n The number of elements to initially create.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    explicit device_vector(size_type n, const Alloc &alloc)
+      :Parent(n,alloc) {}
+
+    /*! This constructor creates a \p device_vector with copies
+     *  of an exemplar element.
+     *  \param n The number of elements to initially create.
+     *  \param value An element to copy.
+     */
+    explicit device_vector(size_type n, const value_type &value)
+      :Parent(n,value) {}
+
+    /*! This constructor creates a \p device_vector with copies
+     *  of an exemplar element.
+     *  \param n The number of elements to initially create.
+     *  \param value An element to copy.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    explicit device_vector(size_type n, const value_type &value, const Alloc &alloc)
+      :Parent(n,value,alloc) {}
+
+    /*! Copy constructor copies from an exemplar \p device_vector.
+     *  \param v The \p device_vector to copy.
+     */
+    device_vector(const device_vector &v)
+      :Parent(v) {}
+
+    /*! Copy constructor copies from an exemplar \p device_vector.
+     *  \param v The \p device_vector to copy.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    device_vector(const device_vector &v, const Alloc &alloc)
+      :Parent(v,alloc) {}
+
+  #if THRUST_CPP_DIALECT >= 2011
+    /*! Move constructor moves from another \p device_vector.
+     *  \param v The device_vector to move.
+     */
+    device_vector(device_vector &&v)
+      :Parent(std::move(v)) {}
+
+    /*! Move constructor moves from another \p device_vector.
+     *  \param v The device_vector to move.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    device_vector(device_vector &&v, const Alloc &alloc)
+      :Parent(std::move(v), alloc) {}
+  #endif // THRUST_CPP_DIALECT >= 2011
+
+    /*! Copy assign operator copies another \p device_vector with the same type.
+     *  \param v The \p device_vector to copy.
+     */
+    device_vector &operator=(const device_vector &v)
+    { Parent::operator=(v); return *this; }
+
+  #if THRUST_CPP_DIALECT >= 2011
+    /*! Move assign operator moves from another \p device_vector.
+     *  \param v The device_vector to move.
+     */
+     device_vector &operator=(device_vector &&v)
+     { Parent::operator=(std::move(v)); return *this; }
+  #endif // THRUST_CPP_DIALECT >= 2011
+
+    /*! Copy constructor copies from an exemplar \p device_vector with different type.
+     *  \param v The \p device_vector to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    explicit device_vector(const device_vector<OtherT,OtherAlloc> &v)
+      :Parent(v) {}
+
+    /*! Assign operator copies from an exemplar \p device_vector with different type.
+     *  \param v The \p device_vector to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    device_vector &operator=(const device_vector<OtherT,OtherAlloc> &v)
+    { Parent::operator=(v); return *this; }
+
+    /*! Copy constructor copies from an exemplar \c std::vector.
+     *  \param v The <tt>std::vector</tt> to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    device_vector(const std::vector<OtherT,OtherAlloc> &v)
+      :Parent(v) {}
+
+    /*! Assign operator copies from an exemplar <tt>std::vector</tt>.
+     *  \param v The <tt>std::vector</tt> to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    device_vector &operator=(const std::vector<OtherT,OtherAlloc> &v)
+    { Parent::operator=(v); return *this;}
+
+    /*! Copy construct from a \p vector_base whose element type is convertible
+     *  to \c T.
+     *
+     *  \param v The \p vector_base to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    device_vector(const detail::vector_base<OtherT,OtherAlloc> &v)
+      :Parent(v) {}
+
+    /*! Assign a \p vector_base whose element type is convertible to \c T.
+     *  \param v The \p vector_base to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    device_vector &operator=(const detail::vector_base<OtherT,OtherAlloc> &v)
+    { Parent::operator=(v); return *this; }
+
+    /*! This constructor builds a \p device_vector from an intializer_list.
+     *  \param il The intializer_list.
+     */
+    device_vector(std::initializer_list<T> il)
+      :Parent(il) {}
+      
+    /*! This constructor builds a \p device_vector from an intializer_list.
+     *  \param il The intializer_list.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    device_vector(std::initializer_list<T> il, const Alloc &alloc)
+      :Parent(il, alloc) {}
+      
+    /*! Assign an \p intializer_list with a matching element type
+     *  \param il The intializer_list.
+     */
+    device_vector &operator=(std::initializer_list<T> il)
+    { Parent::operator=(il); return *this; }
+
+    /*! This constructor builds a \p device_vector from a range.
+     *  \param first The beginning of the range.
+     *  \param last The end of the range.
+     */
+    template<typename InputIterator>
+    device_vector(InputIterator first, InputIterator last)
+      :Parent(first,last) {}
+
+    /*! This constructor builds a \p device_vector from a range.
+     *  \param first The beginning of the range.
+     *  \param last The end of the range.
+     *  \param alloc The allocator to use by this device_vector.
+     */
+    template<typename InputIterator>
+    device_vector(InputIterator first, InputIterator last, const Alloc &alloc)
+      :Parent(first,last,alloc) {}
+
+// declare these members for the purpose of Doxygenating them
+// they actually exist in a derived-from class
+#if 0
+    /*! \brief Resizes this vector to the specified number of elements.
+     *  \param new_size Number of elements this vector should contain.
+     *  \param x Data with which new elements should be populated.
+     *  \throw std::length_error If n exceeds max_size().
+     *
+     *  This method will resize this vector to the specified number of
+     *  elements.  If the number is smaller than this vector's current
+     *  size this vector is truncated, otherwise this vector is
+     *  extended and new elements are populated with given data.
+     */
+    void resize(size_type new_size, const value_type &x = value_type());
+
+    /*! Returns the number of elements in this vector.
+     */
+    size_type size(void) const;
+
+    /*! Returns the size() of the largest possible vector.
+     *  \return The largest possible return value of size().
+     */
+    size_type max_size(void) const;
+
+    /*! \brief If n is less than or equal to capacity(), this call has no effect.
+     *         Otherwise, this method is a request for allocation of additional memory. If
+     *         the request is successful, then capacity() is greater than or equal to
+     *         n; otherwise, capacity() is unchanged. In either case, size() is unchanged.
+     *  \throw std::length_error If n exceeds max_size().
+     */
+    void reserve(size_type n);
+
+    /*! Returns the number of elements which have been reserved in this
+     *  vector.
+     */
+    size_type capacity(void) const;
+
+    /*! This method shrinks the capacity of this vector to exactly
+     *  fit its elements.
+     */
+    void shrink_to_fit(void);
+
+    /*! \brief Subscript access to the data contained in this vector_dev.
+     *  \param n The index of the element for which data should be accessed.
+     *  \return Read/write reference to data.
+     *
+     *  This operator allows for easy, array-style, data access.
+     *  Note that data access with this operator is unchecked and
+     *  out_of_range lookups are not defined.
+     */
+    reference operator[](size_type n);
+
+    /*! \brief Subscript read access to the data contained in this vector_dev.
+     *  \param n The index of the element for which data should be accessed.
+     *  \return Read reference to data.
+     *
+     *  This operator allows for easy, array-style, data access.
+     *  Note that data access with this operator is unchecked and
+     *  out_of_range lookups are not defined.
+     */
+    const_reference operator[](size_type n) const;
+
+    /*! This method returns an iterator pointing to the beginning of
+     *  this vector.
+     *  \return mStart
+     */
+    iterator begin(void);
+
+    /*! This method returns a const_iterator pointing to the beginning
+     *  of this vector.
+     *  \return mStart
+     */
+    const_iterator begin(void) const;
+
+    /*! This method returns a const_iterator pointing to the beginning
+     *  of this vector.
+     *  \return mStart
+     */
+    const_iterator cbegin(void) const;
+
+    /*! This method returns a reverse_iterator pointing to the beginning of
+     *  this vector's reversed sequence.
+     *  \return A reverse_iterator pointing to the beginning of this
+     *          vector's reversed sequence.
+     */
+    reverse_iterator rbegin(void);
+
+    /*! This method returns a const_reverse_iterator pointing to the beginning of
+     *  this vector's reversed sequence.
+     *  \return A const_reverse_iterator pointing to the beginning of this
+     *          vector's reversed sequence.
+     */
+    const_reverse_iterator rbegin(void) const;
+
+    /*! This method returns a const_reverse_iterator pointing to the beginning of
+     *  this vector's reversed sequence.
+     *  \return A const_reverse_iterator pointing to the beginning of this
+     *          vector's reversed sequence.
+     */
+    const_reverse_iterator crbegin(void) const;
+
+    /*! This method returns an iterator pointing to one element past the
+     *  last of this vector.
+     *  \return begin() + size().
+     */
+    iterator end(void);
+
+    /*! This method returns a const_iterator pointing to one element past the
+     *  last of this vector.
+     *  \return begin() + size().
+     */
+    const_iterator end(void) const;
+
+    /*! This method returns a const_iterator pointing to one element past the
+     *  last of this vector.
+     *  \return begin() + size().
+     */
+    const_iterator cend(void) const;
+
+    /*! This method returns a reverse_iterator pointing to one element past the
+     *  last of this vector's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    reverse_iterator rend(void);
+
+    /*! This method returns a const_reverse_iterator pointing to one element past the
+     *  last of this vector's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    const_reverse_iterator rend(void) const;
+
+    /*! This method returns a const_reverse_iterator pointing to one element past the
+     *  last of this vector's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    const_reverse_iterator crend(void) const;
+
+    /*! This method returns a const_reference referring to the first element of this
+     *  vector.
+     *  \return The first element of this vector.
+     */
+    const_reference front(void) const;
+
+    /*! This method returns a reference pointing to the first element of this
+     *  vector.
+     *  \return The first element of this vector.
+     */
+    reference front(void);
+
+    /*! This method returns a const reference pointing to the last element of
+     *  this vector.
+     *  \return The last element of this vector.
+     */
+    const_reference back(void) const;
+
+    /*! This method returns a reference referring to the last element of
+     *  this vector_dev.
+     *  \return The last element of this vector.
+     */
+    reference back(void);
+
+    /*! This method returns a pointer to this vector's first element.
+     *  \return A pointer to the first element of this vector.
+     */
+    pointer data(void);
+
+    /*! This method returns a const_pointer to this vector's first element.
+     *  \return a const_pointer to the first element of this vector.
+     */
+    const_pointer data(void) const;
+
+    /*! This method resizes this vector to 0.
+     */
+    void clear(void);
+
+    /*! This method returns true iff size() == 0.
+     *  \return true if size() == 0; false, otherwise.
+     */
+    bool empty(void) const;
+
+    /*! This method appends the given element to the end of this vector.
+     *  \param x The element to append.
+     */
+    void push_back(const value_type &x);
+
+    /*! This method erases the last element of this vector, invalidating
+     *  all iterators and references to it.
+     */
+    void pop_back(void);
+
+    /*! This method swaps the contents of this device_vector with another vector.
+     *  \param v The vector with which to swap.
+     */
+    void swap(device_vector &v);
+
+    /*! This method removes the element at position pos.
+     *  \param pos The position of the element of interest.
+     *  \return An iterator pointing to the new location of the element that followed the element
+     *          at position pos.
+     */
+    iterator erase(iterator pos);
+
+    /*! This method removes the range of elements [first,last) from this vector.
+     *  \param first The beginning of the range of elements to remove.
+     *  \param last The end of the range of elements to remove.
+     *  \return An iterator pointing to the new location of the element that followed the last
+     *          element in the sequence [first,last).
+     */
+    iterator erase(iterator first, iterator last);
+
+    /*! This method inserts a single copy of a given exemplar value at the
+     *  specified position in this vector.
+     *  \param position The insertion position.
+     *  \param x The exemplar element to copy & insert.
+     *  \return An iterator pointing to the newly inserted element.
+     */
+    iterator insert(iterator position, const T &x);
+
+    /*! This method inserts a copy of an exemplar value to a range at the
+     *  specified position in this vector.
+     *  \param position The insertion position
+     *  \param n The number of insertions to perform.
+     *  \param x The value to replicate and insert.
+     */
+    void insert(iterator position, size_type n, const T &x);
+
+    /*! This method inserts a copy of an input range at the specified position
+     *  in this vector.
+     *  \param position The insertion position.
+     *  \param first The beginning of the range to copy.
+     *  \param last  The end of the range to copy.
+     *
+     *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator>Input Iterator</a>,
+     *                        and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>.
+     */
+    template<typename InputIterator>
+    void insert(iterator position, InputIterator first, InputIterator last);
+
+    /*! This version of \p assign replicates a given exemplar
+     *  \p n times into this vector.
+     *  \param n The number of times to copy \p x.
+     *  \param x The exemplar element to replicate.
+     */
+    void assign(size_type n, const T &x);
+
+    /*! This version of \p assign makes this vector a copy of a given input range.
+     *  \param first The beginning of the range to copy.
+     *  \param last  The end of the range to copy.
+     *
+     *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>.
+     */
+    template<typename InputIterator>
+    void assign(InputIterator first, InputIterator last);
+
+    /*! This method returns a copy of this vector's allocator.
+     *  \return A copy of the alloctor used by this vector.
+     */
+    allocator_type get_allocator(void) const;
+#endif // end doxygen-only members
+};
+
+/*! Exchanges the values of two vectors.
+ *  \p x The first \p device_vector of interest.
+ *  \p y The second \p device_vector of interest.
+ */
+template<typename T, typename Alloc>
+  void swap(device_vector<T,Alloc> &a, device_vector<T,Alloc> &b)
+{
+  a.swap(b);
+}
+
+/*! \} // containres
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/distance.h b/thirdparty/manifold/thirdparty/thrust/thrust/distance.h
new file mode 100644
index 000000000000..890879115126
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/distance.h
@@ -0,0 +1,74 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file distance.h
+ *  \brief Computes the size of a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \p distance finds the distance between \p first and \p last, i.e. the
+ *  number of times that \p first must be incremented until it is equal to
+ *  \p last.
+ *
+ *  \param first The beginning of an input range of interest.
+ *  \param last The end of an input range of interest.
+ *  \return The distance between the beginning and end of the input range.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *
+ *  \pre If \c InputIterator meets the requirements of random access iterator, \p last shall be reachable from \p first or
+ *       \p first shall be reachable from \p last; otherwise, \p last shall be reachable from \p first.
+ *
+ *  The following code snippet demonstrates how to use \p distance to compute
+ *  the distance to one iterator from another.
+ *
+ *  \code
+ *  #include <thrust/distance.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> vec(13);
+ *  thrust::device_vector<int>::iterator iter1 = vec.begin();
+ *  thrust::device_vector<int>::iterator iter2 = iter1 + 7;
+ *
+ *  int d = thrust::distance(iter1, iter2);
+ *
+ *  // d is 7
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/distance
+ */
+template<typename InputIterator>
+inline __host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last);
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/distance.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/equal.h b/thirdparty/manifold/thirdparty/thrust/thrust/equal.h
new file mode 100644
index 000000000000..2f3518907895
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/equal.h
@@ -0,0 +1,235 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file equal.h
+ *  \brief Equality between ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup comparisons
+ *  \ingroup reductions
+ *  \{
+ */
+
+
+/*! \p equal returns \c true if the two ranges <tt>[first1, last1)</tt>
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> are identical when
+ *  compared element-by-element, and otherwise returns \c false.
+ *
+ *  This version of \p equal returns \c true if and only if for every
+ *  iterator \c i in <tt>[first1, last1)</tt>, <tt>*i == *(first2 + (i - first1))</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \return \c true, if the sequences are equal; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and \p InputIterator1's \c value_type can be compared for equality with \c InputIterator2's \c value_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and \p InputIterator2's \c value_type can be compared for equality with \c InputIterator1's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p equal to test
+ *  two ranges for equality using the \p thrust::host execution policy:
+ *
+ *  \code
+ *  #include <thrust/equal.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {3, 1, 4, 1, 5, 9, 3};
+ *  int A2[7] = {3, 1, 4, 2, 8, 5, 7};
+ *  ...
+ *  bool result = thrust::equal(thrust::host, A1, A1 + 7, A2);
+ *
+ *  // result == false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+bool equal(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2);
+
+
+/*! \p equal returns \c true if the two ranges <tt>[first1, last1)</tt>
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> are identical when
+ *  compared element-by-element, and otherwise returns \c false.
+ *
+ *  This version of \p equal returns \c true if and only if for every
+ *  iterator \c i in <tt>[first1, last1)</tt>, <tt>*i == *(first2 + (i - first1))</tt>.
+ *
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \return \c true, if the sequences are equal; \c false, otherwise.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and \p InputIterator1's \c value_type can be compared for equality with \c InputIterator2's \c value_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and \p InputIterator2's \c value_type can be compared for equality with \c InputIterator1's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p equal to test
+ *  two ranges for equality.
+ *
+ *  \code
+ *  #include <thrust/equal.h>
+ *  ...
+ *  int A1[7] = {3, 1, 4, 1, 5, 9, 3};
+ *  int A2[7] = {3, 1, 4, 2, 8, 5, 7};
+ *  ...
+ *  bool result = thrust::equal(A1, A1 + 7, A2);
+ *
+ *  // result == false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal
+ */
+template <typename InputIterator1, typename InputIterator2>
+bool equal(InputIterator1 first1, InputIterator1 last1,
+           InputIterator2 first2);
+
+
+/*! \p equal returns \c true if the two ranges <tt>[first1, last1)</tt>
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> are identical when
+ *  compared element-by-element, and otherwise returns \c false.
+ *
+ *  This version of \p equal returns \c true if and only if for every
+ *  iterator \c i in <tt>[first1, last1)</tt>,
+ *  <tt>binary_pred(*i, *(first2 + (i - first1)))</tt> is \c true.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param binary_pred Binary predicate used to test element equality.
+ *  \return \c true, if the sequences are equal; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator1's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p equal to compare the
+ *  elements in two ranges modulo 2 using the \p thrust::host execution policy.
+ *
+ *  \code
+ *  #include <thrust/equal.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  struct compare_modulo_two
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x, int y) const
+ *    {
+ *      return (x % 2) == (y % 2);
+ *    }
+ *  };
+ *  ...
+ *  int x[6] = {0, 2, 4, 6, 8, 10};
+ *  int y[6] = {1, 3, 5, 7, 9, 11};
+ *
+ *  bool result = thrust::equal(x, x + 6, y, compare_modulo_two());
+ *
+ *  // result is false
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+bool equal(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, BinaryPredicate binary_pred);
+
+
+/*! \p equal returns \c true if the two ranges <tt>[first1, last1)</tt>
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> are identical when
+ *  compared element-by-element, and otherwise returns \c false.
+ *
+ *  This version of \p equal returns \c true if and only if for every
+ *  iterator \c i in <tt>[first1, last1)</tt>,
+ *  <tt>binary_pred(*i, *(first2 + (i - first1)))</tt> is \c true.
+ *
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param binary_pred Binary predicate used to test element equality.
+ *  \return \c true, if the sequences are equal; \c false, otherwise.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator1's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p equal to compare the
+ *  elements in two ranges modulo 2.
+ *
+ *  \code
+ *  #include <thrust/equal.h>
+ *  
+ *  struct compare_modulo_two
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x, int y) const
+ *    {
+ *      return (x % 2) == (y % 2);
+ *    }
+ *  };
+ *  ...
+ *  int x[6] = {0, 2, 4, 6, 8, 10};
+ *  int y[6] = {1, 3, 5, 7, 9, 11};
+ *
+ *  bool result = thrust::equal(x, x + 5, y, compare_modulo_two());
+ *
+ *  // result is true
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/equal
+ */
+template <typename InputIterator1, typename InputIterator2, 
+          typename BinaryPredicate>
+bool equal(InputIterator1 first1, InputIterator1 last1,
+           InputIterator2 first2, BinaryPredicate binary_pred);
+
+
+/*! \} // end comparisons
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/equal.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/event.h b/thirdparty/manifold/thirdparty/thrust/thrust/event.h
new file mode 100644
index 000000000000..75578d96443c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/event.h
@@ -0,0 +1,26 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/event.h
+ *  \brief `thrust::event`, an asynchronous handle type.
+ */
+
+#pragma once
+
+#include <thrust/future.h>
+
+// TODO: Actually separate `<thrust/future.h>` into two headers.
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/execution_policy.h
new file mode 100644
index 000000000000..ecf14413f604
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/execution_policy.h
@@ -0,0 +1,392 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/execution_policy.h
+ *  \brief Thrust execution policies.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/execute_with_allocator.h>
+#include <thrust/detail/seq.h>
+
+//! \cond
+
+// #include the host system's execution_policy header
+#define __THRUST_HOST_SYSTEM_EXECUTION_POLICY_HEADER <__THRUST_HOST_SYSTEM_ROOT/execution_policy.h>
+#include __THRUST_HOST_SYSTEM_EXECUTION_POLICY_HEADER
+#undef __THRUST_HOST_SYSTEM_EXECUTION_POLICY_HEADER
+
+// #include the device system's execution_policy.h header
+#define __THRUST_DEVICE_SYSTEM_EXECUTION_POLICY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/execution_policy.h>
+#include __THRUST_DEVICE_SYSTEM_EXECUTION_POLICY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_EXECUTION_POLICY_HEADER
+
+//! \endcond
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \cond
+ */
+
+
+namespace detail
+{
+
+
+typedef thrust::system::__THRUST_HOST_SYSTEM_NAMESPACE::detail::par_t host_t;
+
+
+typedef thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::detail::par_t device_t;
+
+
+} // end detail
+
+
+/*! \endcond
+ */
+
+
+/*! \addtogroup execution_policies Parallel Execution Policies
+ *  \{
+ */
+
+
+// define execution_policy for the purpose of Doxygenating it
+// it is actually defined elsewhere
+#if 0
+/*! \p execution_policy is the base class for all Thrust parallel execution policies
+ *  like \p thrust::host, \p thrust::device, and each backend system's tag type.
+ *
+ *  Custom user-defined backends should derive a policy from this type in order to
+ *  interoperate with Thrust algorithm dispatch.
+ *
+ *  The following code snippet demonstrates how to derive a standalone custom execution policy
+ *  from \p thrust::execution_policy to implement a backend which only implements \p for_each:
+ *
+ *  \code
+ *  #include <thrust/execution_policy.h>
+ *  #include <iostream>
+ *
+ *  // define a type derived from thrust::execution_policy to distinguish our custom execution policy:
+ *  struct my_policy : thrust::execution_policy<my_policy> {};
+ *
+ *  // overload for_each on my_policy
+ *  template<typename Iterator, typename Function>
+ *  Iterator for_each(my_policy, Iterator first, Iterator last, Function f)
+ *  {
+ *    std::cout << "Hello, world from for_each(my_policy)!" << std::endl;
+ *
+ *    for(; first < last; ++first)
+ *    {
+ *      f(*first);
+ *    }
+ *
+ *    return first;
+ *  }
+ *
+ *  struct ignore_argument
+ *  {
+ *    void operator()(int) {}
+ *  };
+ *
+ *  int main()
+ *  {
+ *    int data[4];
+ *
+ *    // dispatch thrust::for_each using our custom policy:
+ *    my_policy exec;
+ *    thrust::for_each(exec, data, data + 4, ignore_argument());
+ *
+ *    // can't dispatch thrust::transform because no overload exists for my_policy:
+ *    //thrust::transform(exec, data, data, + 4, data, thrust::identity<int>()); // error!
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see host_execution_policy
+ *  \see device_execution_policy
+ */
+template<typename DerivedPolicy>
+struct execution_policy : thrust::detail::execution_policy_base<DerivedPolicy>
+{};
+#endif
+
+
+/*! \p host_execution_policy is the base class for all Thrust parallel execution policies
+ *  which are derived from Thrust's default host backend system configured with the \p THRUST_HOST_SYSTEM
+ *  macro.
+ *
+ *  Custom user-defined backends which wish to inherit the functionality of Thrust's host backend system
+ *  should derive a policy from this type in order to interoperate with Thrust algorithm dispatch.
+ *
+ *  The following code snippet demonstrates how to derive a standalone custom execution policy from
+ *  \p thrust::host_execution_policy to implement a backend which specializes \p for_each while inheriting
+ *  the behavior of every other algorithm from the host system:
+ *
+ *  \code
+ *  #include <thrust/execution_policy.h>
+ *  #include <iostream>
+ *
+ *  // define a type derived from thrust::host_execution_policy to distinguish our custom execution policy:
+ *  struct my_policy : thrust::host_execution_policy<my_policy> {};
+ *
+ *  // overload for_each on my_policy
+ *  template<typename Iterator, typename Function>
+ *  Iterator for_each(my_policy, Iterator first, Iterator last, Function f)
+ *  {
+ *    std::cout << "Hello, world from for_each(my_policy)!" << std::endl;
+ *
+ *    for(; first < last; ++first)
+ *    {
+ *      f(*first);
+ *    }
+ *
+ *    return first;
+ *  }
+ *
+ *  struct ignore_argument
+ *  {
+ *    void operator()(int) {}
+ *  };
+ *
+ *  int main()
+ *  {
+ *    int data[4];
+ *
+ *    // dispatch thrust::for_each using our custom policy:
+ *    my_policy exec;
+ *    thrust::for_each(exec, data, data + 4, ignore_argument());
+ *
+ *    // dispatch thrust::transform whose behavior our policy inherits
+ *    thrust::transform(exec, data, data, + 4, data, thrust::identity<int>());
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see execution_policy
+ *  \see device_execution_policy
+ */
+template<typename DerivedPolicy>
+  struct host_execution_policy
+    : thrust::system::__THRUST_HOST_SYSTEM_NAMESPACE::execution_policy<DerivedPolicy>
+{};
+
+
+/*! \p device_execution_policy is the base class for all Thrust parallel execution policies
+ *  which are derived from Thrust's default device backend system configured with the \p THRUST_DEVICE_SYSTEM
+ *  macro.
+ *
+ *  Custom user-defined backends which wish to inherit the functionality of Thrust's device backend system
+ *  should derive a policy from this type in order to interoperate with Thrust algorithm dispatch.
+ *
+ *  The following code snippet demonstrates how to derive a standalone custom execution policy from
+ *  \p thrust::device_execution_policy to implement a backend which specializes \p for_each while inheriting
+ *  the behavior of every other algorithm from the device system:
+ *
+ *  \code
+ *  #include <thrust/execution_policy.h>
+ *  #include <iostream>
+ *
+ *  // define a type derived from thrust::device_execution_policy to distinguish our custom execution policy:
+ *  struct my_policy : thrust::device_execution_policy<my_policy> {};
+ *
+ *  // overload for_each on my_policy
+ *  template<typename Iterator, typename Function>
+ *  Iterator for_each(my_policy, Iterator first, Iterator last, Function f)
+ *  {
+ *    std::cout << "Hello, world from for_each(my_policy)!" << std::endl;
+ *
+ *    for(; first < last; ++first)
+ *    {
+ *      f(*first);
+ *    }
+ *
+ *    return first;
+ *  }
+ *
+ *  struct ignore_argument
+ *  {
+ *    void operator()(int) {}
+ *  };
+ *
+ *  int main()
+ *  {
+ *    int data[4];
+ *
+ *    // dispatch thrust::for_each using our custom policy:
+ *    my_policy exec;
+ *    thrust::for_each(exec, data, data + 4, ignore_argument());
+ *
+ *    // dispatch thrust::transform whose behavior our policy inherits
+ *    thrust::transform(exec, data, data, + 4, data, thrust::identity<int>());
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see execution_policy
+ *  \see host_execution_policy
+ */
+template<typename DerivedPolicy>
+  struct device_execution_policy
+    : thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::execution_policy<DerivedPolicy>
+{};
+
+
+/*! \p thrust::host is the default parallel execution policy associated with Thrust's host backend system
+ *  configured by the \p THRUST_HOST_SYSTEM macro.
+ *
+ *  Instead of relying on implicit algorithm dispatch through iterator system tags, users may directly target
+ *  algorithm dispatch at Thrust's host system by providing \p thrust::host as an algorithm parameter.
+ *
+ *  Explicit dispatch can be useful in avoiding the introduction of data copies into containers such as
+ *  \p thrust::host_vector.
+ *
+ *  Note that even though \p thrust::host targets the host CPU, it is a parallel execution policy. That is,
+ *  the order that an algorithm invokes functors or dereferences iterators is not defined.
+ *
+ *  The type of \p thrust::host is implementation-defined.
+ *
+ *  The following code snippet demonstrates how to use \p thrust::host to explicitly dispatch an invocation
+ *  of \p thrust::for_each to the host backend system:
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdio>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  int vec[] = { 0, 1, 2 };
+ *
+ *  thrust::for_each(thrust::host, vec, vec + 3, printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ *
+ *  \see host_execution_policy
+ *  \see thrust::device
+ */
+static const detail::host_t host;
+
+
+/*! \p thrust::device is the default parallel execution policy associated with Thrust's device backend system
+ *  configured by the \p THRUST_DEVICE_SYSTEM macro.
+ *
+ *  Instead of relying on implicit algorithm dispatch through iterator system tags, users may directly target
+ *  algorithm dispatch at Thrust's device system by providing \p thrust::device as an algorithm parameter.
+ *
+ *  Explicit dispatch can be useful in avoiding the introduction of data copies into containers such as
+ *  \p thrust::device_vector or to avoid wrapping e.g. raw pointers allocated by the CUDA API with types
+ *  such as \p thrust::device_ptr.
+ *
+ *  The user must take care to guarantee that the iterators provided to an algorithm are compatible with
+ *  the device backend system. For example, raw pointers allocated by <tt>std::malloc</tt> typically
+ *  cannot be dereferenced by a GPU. For this reason, raw pointers allocated by host APIs should not be mixed
+ *  with a \p thrust::device algorithm invocation when the device backend is CUDA.
+ *
+ *  The type of \p thrust::device is implementation-defined.
+ *
+ *  The following code snippet demonstrates how to use \p thrust::device to explicitly dispatch an invocation
+ *  of \p thrust::for_each to the device backend system:
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdio>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  thrust::device_vector<int> vec(3);
+ *  vec[0] = 0; vec[1] = 1; vec[2] = 2;
+ *
+ *  thrust::for_each(thrust::device, vec.begin(), vec.end(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ *
+ *  \see host_execution_policy
+ *  \see thrust::device
+ */
+THRUST_INLINE_CONSTANT detail::device_t device;
+
+
+// define seq for the purpose of Doxygenating it
+// it is actually defined elsewhere
+#if 0
+/*! \p thrust::seq is an execution policy which requires an algorithm invocation to execute sequentially
+ *  in the current thread. It can not be configured by a compile-time macro.
+ *
+ *  The type of \p thrust::seq is implementation-defined.
+ *
+ *  The following code snippet demonstrates how to use \p thrust::seq to explicitly execute an invocation
+ *  of \p thrust::for_each sequentially:
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <vector>
+ *  #include <cstdio>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  std::vector<int> vec(3);
+ *  vec[0] = 0; vec[1] = 1; vec[2] = 2;
+ *
+ *  thrust::for_each(thrust::seq, vec.begin(), vec.end(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in sequential order
+ *  \endcode
+ *
+ *  \see thrust::host
+ *  \see thrust::device
+ */
+static const detail::seq_t seq;
+#endif
+
+
+/*! \}
+ */
+
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/extrema.h b/thirdparty/manifold/thirdparty/thrust/thrust/extrema.h
new file mode 100644
index 000000000000..ca419a0aa5bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/extrema.h
@@ -0,0 +1,801 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file extrema.h
+ *  \brief Functions for computing computing extremal values
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! This version of \p min returns the smaller of two values, given a comparison operation.
+ *  \param lhs The first value to compare.
+ *  \param rhs The second value to compare.
+ *  \param comp A comparison operation.
+ *  \return The smaller element.
+ *
+ *  \tparam T is convertible to \p BinaryPredicate's first argument type and to its second argument type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p min to compute the smaller of two
+ *  key-value objects.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value a = {13, 0};
+ *  key_value b = { 7, 1);
+ *
+ *  key_value smaller = thrust::min(a, b, compare_key_value());
+ *
+ *  // smaller is {7, 1}
+ *  \endcode
+ *
+ *  \note Returns the first argument when the arguments are equivalent.
+ *  \see max
+ */
+template<typename T, typename BinaryPredicate>
+__host__ __device__
+  T min THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs, BinaryPredicate comp);
+
+
+/*! This version of \p min returns the smaller of two values.
+ *  \param lhs The first value to compare.
+ *  \param rhs The second value to compare.
+ *  \return The smaller element.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p min to compute the smaller of two
+ *  integers.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  int a = 13;
+ *  int b = 7;
+ *
+ *  int smaller = thrust::min(a, b);
+ *
+ *  // smaller is 7
+ *  \endcode
+ *
+ *  \note Returns the first argument when the arguments are equivalent.
+ *  \see max
+ */
+template<typename T>
+__host__ __device__
+  T min THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs);
+
+
+/*! This version of \p max returns the larger of two values, given a comparison operation.
+ *  \param lhs The first value to compare.
+ *  \param rhs The second value to compare.
+ *  \param comp A comparison operation.
+ *  \return The larger element.
+ *
+ *  \tparam T is convertible to \p BinaryPredicate's first argument type and to its second argument type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p max to compute the larger of two
+ *  key-value objects.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value a = {13, 0};
+ *  key_value b = { 7, 1);
+ *
+ *  key_value larger = thrust::max(a, b, compare_key_value());
+ *
+ *  // larger is {13, 0}
+ *  \endcode
+ *
+ *  \note Returns the first argument when the arguments are equivalent.
+ *  \see min
+ */
+template<typename T, typename BinaryPredicate>
+__host__ __device__
+  T max THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs, BinaryPredicate comp);
+
+
+/*! This version of \p max returns the larger of two values.
+ *  \param lhs The first value to compare.
+ *  \param rhs The second value to compare.
+ *  \return The larger element.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p max to compute the larger of two
+ *  integers.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  int a = 13;
+ *  int b = 7;
+ *
+ *  int larger = thrust::min(a, b);
+ *
+ *  // larger is 13
+ *  \endcode
+ *
+ *  \note Returns the first argument when the arguments are equivalent.
+ *  \see min
+ */
+template<typename T>
+__host__ __device__
+  T max THRUST_PREVENT_MACRO_SUBSTITUTION (const T &lhs, const T &rhs);
+
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup extrema
+ *  \ingroup reductions
+ *  \{
+ */
+
+/*! \p min_element finds the smallest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value smaller
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p min_element differ in how they define whether one element is
+ *  less than another. This version compares objects using \c operator<. Specifically,
+ *  this version of \p min_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>*j < *i</tt> is
+ *  \c false.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return An iterator pointing to the smallest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \c ForwardIterator's \c value_type is a model of
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int *result = thrust::min_element(thrust::host, data, data + 6);
+ *
+ *  // result is data + 1
+ *  // *result is 0
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ */
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator min_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last);
+
+
+/*! \p min_element finds the smallest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value smaller
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p min_element differ in how they define whether one element is
+ *  less than another. This version compares objects using \c operator<. Specifically,
+ *  this version of \p min_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>*j < *i</tt> is
+ *  \c false.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return An iterator pointing to the smallest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \c ForwardIterator's \c value_type is a model of
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int *result = thrust::min_element(data, data + 6);
+ *
+ *  // result is data + 1
+ *  // *result is 0
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ */
+template <typename ForwardIterator>
+ForwardIterator min_element(ForwardIterator first, ForwardIterator last);
+
+
+/*! \p min_element finds the smallest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value smaller
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p min_element differ in how they define whether one element is
+ *  less than another. This version compares objects using a function object \p comp.
+ *  Specifically, this version of \p min_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>comp(*j, *i)</tt> is
+ *  \c false.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp A binary predicate used for comparison.
+ *  \return An iterator pointing to the smallest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \p comp's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p min_element to find the smallest element
+ *  of a collection of key-value pairs using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value data[4] = { {4,5}, {0,7}, {2,3}, {6,1} };
+ *
+ *  key_value *smallest = thrust::min_element(thrust::host, data, data + 4, compare_key_value());
+ *
+ *  // smallest == data + 1
+ *  // *smallest == {0,7}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator min_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last, BinaryPredicate comp);
+
+
+/*! \p min_element finds the smallest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value smaller
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p min_element differ in how they define whether one element is
+ *  less than another. This version compares objects using a function object \p comp.
+ *  Specifically, this version of \p min_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>comp(*j, *i)</tt> is
+ *  \c false.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp A binary predicate used for comparison.
+ *  \return An iterator pointing to the smallest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \p comp's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p min_element to find the smallest element
+ *  of a collection of key-value pairs.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value data[4] = { {4,5}, {0,7}, {2,3}, {6,1} };
+ *
+ *  key_value *smallest = thrust::min_element(data, data + 4, compare_key_value());
+ *
+ *  // smallest == data + 1
+ *  // *smallest == {0,7}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/min_element 
+ */
+template <typename ForwardIterator, typename BinaryPredicate>
+ForwardIterator min_element(ForwardIterator first, ForwardIterator last,
+                            BinaryPredicate comp);
+
+
+/*! \p max_element finds the largest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value larger
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p max_element differ in how they define whether one element is
+ *  greater than another. This version compares objects using \c operator<. Specifically,
+ *  this version of \p max_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>*i < *j</tt> is
+ *  \c false.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return An iterator pointing to the largest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam A Thrust backend system.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \c ForwardIterator's \c value_type is a model of
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int *result = thrust::max_element(thrust::host, data, data + 6);
+ *
+ *  // *result == 3
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ */
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator max_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last);
+
+
+/*! \p max_element finds the largest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value larger
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p max_element differ in how they define whether one element is
+ *  greater than another. This version compares objects using \c operator<. Specifically,
+ *  this version of \p max_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>*i < *j</tt> is
+ *  \c false.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return An iterator pointing to the largest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \c ForwardIterator's \c value_type is a model of
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int *result = thrust::max_element(data, data + 6);
+ *
+ *  // *result == 3
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ */
+template <typename ForwardIterator>
+ForwardIterator max_element(ForwardIterator first, ForwardIterator last);
+
+
+/*! \p max_element finds the largest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value larger
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p max_element differ in how they define whether one element is
+ *  less than another. This version compares objects using a function object \p comp.
+ *  Specifically, this version of \p max_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>comp(*i, *j)</tt> is
+ *  \c false.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp A binary predicate used for comparison.
+ *  \return An iterator pointing to the largest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \p comp's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p max_element to find the largest element
+ *  of a collection of key-value pairs using the \p thrust::host execution policy for parallelization.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value data[4] = { {4,5}, {0,7}, {2,3}, {6,1} };
+ *
+ *  key_value *largest = thrust::max_element(thrust::host, data, data + 4, compare_key_value());
+ *
+ *  // largest == data + 3
+ *  // *largest == {6,1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator max_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last, BinaryPredicate comp);
+
+
+/*! \p max_element finds the largest element in the range <tt>[first, last)</tt>.
+ *  It returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that no other iterator in <tt>[first, last)</tt> points to a value larger
+ *  than \c *i. The return value is \p last if and only if <tt>[first, last)</tt> is an
+ *  empty range.
+ *
+ *  The two versions of \p max_element differ in how they define whether one element is
+ *  less than another. This version compares objects using a function object \p comp.
+ *  Specifically, this version of \p max_element returns the first iterator \c i in <tt>[first, last)</tt>
+ *  such that, for every iterator \c j in <tt>[first, last)</tt>, <tt>comp(*i, *j)</tt> is
+ *  \c false.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp A binary predicate used for comparison.
+ *  \return An iterator pointing to the largest element of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \p comp's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p max_element to find the largest element
+ *  of a collection of key-value pairs.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value data[4] = { {4,5}, {0,7}, {2,3}, {6,1} };
+ *
+ *  key_value *largest = thrust::max_element(data, data + 4, compare_key_value());
+ *
+ *  // largest == data + 3
+ *  // *largest == {6,1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/max_element 
+ */
+template <typename ForwardIterator, typename BinaryPredicate>
+ForwardIterator max_element(ForwardIterator first, ForwardIterator last,
+                            BinaryPredicate comp);
+
+
+/*! \p minmax_element finds the smallest and largest elements in the range <tt>[first, last)</tt>.
+ *  It returns a pair of iterators <tt>(imin, imax)</tt> where \c imin is the same iterator
+ *  returned by \p min_element and \c imax is the same iterator returned by \p max_element.
+ *  This function is potentially more efficient than separate calls to \p min_element and \p max_element.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return A pair of iterator pointing to the smallest and largest elements of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \c ForwardIterator's \c value_type is a model of
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  thrust::pair<int *, int *> result = thrust::minmax_element(thrust::host, data, data + 6);
+ *
+ *  // result.first is data + 1
+ *  // result.second is data + 5
+ *  // *result.first is 0
+ *  // *result.second is 3
+ *  \endcode
+ *
+ *  \see min_element
+ *  \see max_element
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1840.pdf
+ */
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last);
+
+
+/*! \p minmax_element finds the smallest and largest elements in the range <tt>[first, last)</tt>.
+ *  It returns a pair of iterators <tt>(imin, imax)</tt> where \c imin is the same iterator
+ *  returned by \p min_element and \c imax is the same iterator returned by \p max_element.
+ *  This function is potentially more efficient than separate calls to \p min_element and \p max_element.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return A pair of iterator pointing to the smallest and largest elements of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \c ForwardIterator's \c value_type is a model of
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  thrust::pair<int *, int *> result = thrust::minmax_element(data, data + 6);
+ *
+ *  // result.first is data + 1
+ *  // result.second is data + 5
+ *  // *result.first is 0
+ *  // *result.second is 3
+ *  \endcode
+ *
+ *  \see min_element
+ *  \see max_element
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1840.pdf
+ */
+template <typename ForwardIterator>
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(ForwardIterator first, 
+                                                             ForwardIterator last);
+
+
+/*! \p minmax_element finds the smallest and largest elements in the range <tt>[first, last)</tt>.
+ *  It returns a pair of iterators <tt>(imin, imax)</tt> where \c imin is the same iterator
+ *  returned by \p min_element and \c imax is the same iterator returned by \p max_element.
+ *  This function is potentially more efficient than separate calls to \p min_element and \p max_element.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp A binary predicate used for comparison.
+ *  \return A pair of iterator pointing to the smallest and largest elements of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \p comp's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p minmax_element to find the smallest and largest elements
+ *  of a collection of key-value pairs using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/pair.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value data[4] = { {4,5}, {0,7}, {2,3}, {6,1} };
+ *
+ *  thrust::pair<key_value*,key_value*> extrema = thrust::minmax_element(thrust::host, data, data + 4, compare_key_value());
+ *
+ *  // extrema.first   == data + 1
+ *  // *extrema.first  == {0,7}
+ *  // extrema.second  == data + 3
+ *  // *extrema.second == {6,1}
+ *  \endcode
+ *
+ *  \see min_element
+ *  \see max_element
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1840.pdf
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, ForwardIterator first, ForwardIterator last, BinaryPredicate comp);
+
+
+/*! \p minmax_element finds the smallest and largest elements in the range <tt>[first, last)</tt>.
+ *  It returns a pair of iterators <tt>(imin, imax)</tt> where \c imin is the same iterator
+ *  returned by \p min_element and \c imax is the same iterator returned by \p max_element.
+ *  This function is potentially more efficient than separate calls to \p min_element and \p max_element.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp A binary predicate used for comparison.
+ *  \return A pair of iterator pointing to the smallest and largest elements of the range <tt>[first, last)</tt>,
+ *          if it is not an empty range; \p last, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \p comp's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p minmax_element to find the smallest and largest elements
+ *  of a collection of key-value pairs.
+ *
+ *  \code
+ *  #include <thrust/extrema.h>
+ *  #include <thrust/pair.h>
+ *
+ *  struct key_value
+ *  {
+ *    int key;
+ *    int value;
+ *  };
+ *
+ *  struct compare_key_value
+ *  {
+ *    __host__ __device__
+ *    bool operator()(key_value lhs, key_value rhs)
+ *    {
+ *      return lhs.key < rhs.key;
+ *    }
+ *  };
+ *
+ *  ...
+ *  key_value data[4] = { {4,5}, {0,7}, {2,3}, {6,1} };
+ *
+ *  thrust::pair<key_value*,key_value*> extrema = thrust::minmax_element(data, data + 4, compare_key_value());
+ *
+ *  // extrema.first   == data + 1
+ *  // *extrema.first  == {0,7}
+ *  // extrema.second  == data + 3
+ *  // *extrema.second == {6,1}
+ *  \endcode
+ *
+ *  \see min_element
+ *  \see max_element
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1840.pdf
+ */
+template <typename ForwardIterator, typename BinaryPredicate>
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(ForwardIterator first, 
+                                                             ForwardIterator last,
+                                                             BinaryPredicate comp);
+
+/*! \} // end extrema
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/extrema.inl>
+#include <thrust/detail/minmax.h>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/fill.h
new file mode 100644
index 000000000000..bd9e40268972
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/fill.h
@@ -0,0 +1,206 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file fill.h
+ *  \brief Fills a range with a constant value
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup transformations
+ *  \addtogroup filling
+ *  \ingroup transformations
+ *  \{
+ */
+
+
+/*! \p fill assigns the value \p value to every element in
+ *  the range <tt>[first, last)</tt>. That is, for every
+ *  iterator \c i in <tt>[first, last)</tt>, it performs
+ *  the assignment <tt>*i = value</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param value The value to be copied.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T's \c value_type is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p fill to set a thrust::device_vector's
+ *  elements to a given value using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/fill.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> v(4);
+ *  thrust::fill(thrust::device, v.begin(), v.end(), 137);
+ *
+ *  // v[0] == 137, v[1] == 137, v[2] == 137, v[3] == 137
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/fill
+ *  \see \c fill_n
+ *  \see \c uninitialized_fill
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void fill(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const T &value);
+
+
+/*! \p fill assigns the value \p value to every element in
+ *  the range <tt>[first, last)</tt>. That is, for every
+ *  iterator \c i in <tt>[first, last)</tt>, it performs
+ *  the assignment <tt>*i = value</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param value The value to be copied.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T's \c value_type is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p fill to set a thrust::device_vector's
+ *  elements to a given value.
+ *
+ *  \code
+ *  #include <thrust/fill.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> v(4);
+ *  thrust::fill(v.begin(), v.end(), 137);
+ *
+ *  // v[0] == 137, v[1] == 137, v[2] == 137, v[3] == 137
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/fill
+ *  \see \c fill_n
+ *  \see \c uninitialized_fill
+ */
+template<typename ForwardIterator, typename T>
+__host__ __device__
+  void fill(ForwardIterator first,
+            ForwardIterator last,
+            const T &value);
+
+
+/*! \p fill_n assigns the value \p value to every element in
+ *  the range <tt>[first, first+n)</tt>. That is, for every
+ *  iterator \c i in <tt>[first, first+n)</tt>, it performs
+ *  the assignment <tt>*i = value</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param n The size of the sequence.
+ *  \param value The value to be copied.
+ *  \return <tt>first + n</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T's \c value_type is convertible to a type in \p OutputIterator's set of \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p fill to set a thrust::device_vector's
+ *  elements to a given value using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/fill.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> v(4);
+ *  thrust::fill_n(thrust::device, v.begin(), v.size(), 137);
+ *
+ *  // v[0] == 137, v[1] == 137, v[2] == 137, v[3] == 137
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/fill_n
+ *  \see \c fill
+ *  \see \c uninitialized_fill_n
+ */
+template<typename DerivedPolicy, typename OutputIterator, typename Size, typename T>
+__host__ __device__
+  OutputIterator fill_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        OutputIterator first,
+                        Size n,
+                        const T &value);
+
+
+/*! \p fill_n assigns the value \p value to every element in
+ *  the range <tt>[first, first+n)</tt>. That is, for every
+ *  iterator \c i in <tt>[first, first+n)</tt>, it performs
+ *  the assignment <tt>*i = value</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param n The size of the sequence.
+ *  \param value The value to be copied.
+ *  \return <tt>first + n</tt>
+ *
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T's \c value_type is convertible to a type in \p OutputIterator's set of \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p fill to set a thrust::device_vector's
+ *  elements to a given value.
+ *
+ *  \code
+ *  #include <thrust/fill.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> v(4);
+ *  thrust::fill_n(v.begin(), v.size(), 137);
+ *
+ *  // v[0] == 137, v[1] == 137, v[2] == 137, v[3] == 137
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/fill_n
+ *  \see \c fill
+ *  \see \c uninitialized_fill_n
+ */
+template<typename OutputIterator, typename Size, typename T>
+__host__ __device__
+  OutputIterator fill_n(OutputIterator first,
+                        Size n,
+                        const T &value);
+
+
+/*! \} // end filling
+ *  \} // transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/fill.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/find.h b/thirdparty/manifold/thirdparty/thrust/thrust/find.h
new file mode 100644
index 000000000000..5ab9b0a2df8f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/find.h
@@ -0,0 +1,381 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file find.h
+ *  \brief Locating values in (unsorted) ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+/*! \addtogroup searching
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p find returns the first iterator \c i in the range 
+ *  <tt>[first, last)</tt> such that <tt>*i == value</tt>
+ *  or \c last if no such iterator exists.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first Beginning of the sequence to search.
+ *  \param last End of the sequence to search.
+ *  \param value The value to find.
+ *  \return The first iterator \c i such that <tt>*i == value</tt> or \c last.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and \p InputIterator's \c value_type is equality comparable to type \c T.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">EqualityComparable</a>. 
+ *
+ *  \code
+ *  #include <thrust/find.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> input(4);
+ *
+ *  input[0] = 0;
+ *  input[1] = 5;
+ *  input[2] = 3;
+ *  input[3] = 7;
+ *
+ *  thrust::device_vector<int>::iterator iter;
+ *
+ *  iter = thrust::find(thrust::device, input.begin(), input.end(), 3); // returns input.first() + 2
+ *  iter = thrust::find(thrust::device, input.begin(), input.end(), 5); // returns input.first() + 1
+ *  iter = thrust::find(thrust::device, input.begin(), input.end(), 9); // returns input.end()
+ *  \endcode
+ *
+ *  \see find_if
+ *  \see mismatch
+ */
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+InputIterator find(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   const T& value);
+
+
+/*! \p find returns the first iterator \c i in the range 
+ *  <tt>[first, last)</tt> such that <tt>*i == value</tt>
+ *  or \c last if no such iterator exists.
+ *
+ *  \param first Beginning of the sequence to search.
+ *  \param last End of the sequence to search.
+ *  \param value The value to find.
+ *  \return The first iterator \c i such that <tt>*i == value</tt> or \c last.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and \p InputIterator's \c value_type is equality comparable to type \c T.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">EqualityComparable</a>. 
+ *
+ *  \code
+ *  #include <thrust/find.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> input(4);
+ *
+ *  input[0] = 0;
+ *  input[1] = 5;
+ *  input[2] = 3;
+ *  input[3] = 7;
+ *
+ *  thrust::device_vector<int>::iterator iter;
+ *
+ *  iter = thrust::find(input.begin(), input.end(), 3); // returns input.first() + 2
+ *  iter = thrust::find(input.begin(), input.end(), 5); // returns input.first() + 1
+ *  iter = thrust::find(input.begin(), input.end(), 9); // returns input.end()
+ *  \endcode
+ *
+ *  \see find_if
+ *  \see mismatch
+ */
+template <typename InputIterator, typename T>
+InputIterator find(InputIterator first,
+                   InputIterator last,
+                   const T& value);
+
+
+/*! \p find_if returns the first iterator \c i in the range 
+ *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true
+ *  or \c last if no such iterator exists.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first Beginning of the sequence to search.
+ *  \param last End of the sequence to search.
+ *  \param pred A predicate used to test range elements.
+ *  \return The first iterator \c i such that <tt>pred(*i)</tt> is \c true, or \c last.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/find.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  struct greater_than_four
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 4;
+ *    }
+ *  };
+ *
+ *  struct greater_than_ten
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 10;
+ *    }
+ *  };
+ *
+ *  ...
+ *  thrust::device_vector<int> input(4);
+ *
+ *  input[0] = 0;
+ *  input[1] = 5;
+ *  input[2] = 3;
+ *  input[3] = 7;
+ *
+ *  thrust::device_vector<int>::iterator iter;
+ *
+ *  iter = thrust::find_if(thrust::device, input.begin(), input.end(), greater_than_four()); // returns input.first() + 1
+ *
+ *  iter = thrust::find_if(thrust::device, input.begin(), input.end(), greater_than_ten());  // returns input.end()
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if_not
+ *  \see mismatch
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred);
+
+
+/*! \p find_if returns the first iterator \c i in the range 
+ *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true
+ *  or \c last if no such iterator exists.
+ *
+ *  \param first Beginning of the sequence to search.
+ *  \param last End of the sequence to search.
+ *  \param pred A predicate used to test range elements.
+ *  \return The first iterator \c i such that <tt>pred(*i)</tt> is \c true, or \c last.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/find.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct greater_than_four
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 4;
+ *    }
+ *  };
+ *
+ *  struct greater_than_ten
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 10;
+ *    }
+ *  };
+ *
+ *  ...
+ *  thrust::device_vector<int> input(4);
+ *
+ *  input[0] = 0;
+ *  input[1] = 5;
+ *  input[2] = 3;
+ *  input[3] = 7;
+ *
+ *  thrust::device_vector<int>::iterator iter;
+ *
+ *  iter = thrust::find_if(input.begin(), input.end(), greater_than_four()); // returns input.first() + 1
+ *
+ *  iter = thrust::find_if(input.begin(), input.end(), greater_than_ten());  // returns input.end()
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if_not
+ *  \see mismatch
+ */
+template <typename InputIterator, typename Predicate>
+InputIterator find_if(InputIterator first,
+                      InputIterator last,
+                      Predicate pred);
+
+
+/*! \p find_if_not returns the first iterator \c i in the range 
+ *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c false
+ *  or \c last if no such iterator exists.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first Beginning of the sequence to search.
+ *  \param last End of the sequence to search.
+ *  \param pred A predicate used to test range elements.
+ *  \return The first iterator \c i such that <tt>pred(*i)</tt> is \c false, or \c last.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/find.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  struct greater_than_four
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 4;
+ *    }
+ *  };
+ *
+ *  struct greater_than_ten
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 10;
+ *    }
+ *  };
+ *
+ *  ...
+ *  thrust::device_vector<int> input(4);
+ *
+ *  input[0] = 0;
+ *  input[1] = 5;
+ *  input[2] = 3;
+ *  input[3] = 7;
+ *
+ *  thrust::device_vector<int>::iterator iter;
+ *
+ *  iter = thrust::find_if_not(thrust::device, input.begin(), input.end(), greater_than_four()); // returns input.first()
+ *
+ *  iter = thrust::find_if_not(thrust::device, input.begin(), input.end(), greater_than_ten());  // returns input.first()
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if
+ *  \see mismatch
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if_not(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          Predicate pred);
+
+
+/*! \p find_if_not returns the first iterator \c i in the range 
+ *  <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c false
+ *  or \c last if no such iterator exists.
+ *
+ *  \param first Beginning of the sequence to search.
+ *  \param last End of the sequence to search.
+ *  \param pred A predicate used to test range elements.
+ *  \return The first iterator \c i such that <tt>pred(*i)</tt> is \c false, or \c last.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/find.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct greater_than_four
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 4;
+ *    }
+ *  };
+ *
+ *  struct greater_than_ten
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x > 10;
+ *    }
+ *  };
+ *
+ *  ...
+ *  thrust::device_vector<int> input(4);
+ *
+ *  input[0] = 0;
+ *  input[1] = 5;
+ *  input[2] = 3;
+ *  input[3] = 7;
+ *
+ *  thrust::device_vector<int>::iterator iter;
+ *
+ *  iter = thrust::find_if_not(input.begin(), input.end(), greater_than_four()); // returns input.first()
+ *
+ *  iter = thrust::find_if_not(input.begin(), input.end(), greater_than_ten());  // returns input.first()
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if
+ *  \see mismatch
+ */
+template <typename InputIterator, typename Predicate>
+InputIterator find_if_not(InputIterator first,
+                          InputIterator last,
+                          Predicate pred);
+
+/*! \} // end searching
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/find.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/for_each.h
new file mode 100644
index 000000000000..7d05e3ea16bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/for_each.h
@@ -0,0 +1,278 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ * *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/for_each.h
+ *  \brief Applies a function to each element in a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup modifying
+ *  \ingroup transformations
+ *  \{
+ */
+
+
+/*! \p for_each applies the function object \p f to each element
+ *  in the range <tt>[first, last)</tt>; \p f's return value, if any,
+ *  is ignored. Unlike the C++ Standard Template Library function
+ *  <tt>std::for_each</tt>, this version offers no guarantee on
+ *  order of execution. For this reason, this version of \p for_each
+ *  does not return a copy of the function object.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param f The function object to apply to the range <tt>[first, last)</tt>.
+ *  \return last
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p UnaryFunction's \c argument_type.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>,
+ *          and \p UnaryFunction does not apply any non-constant operation through its argument.
+ *
+ *  The following code snippet demonstrates how to use \p for_each to print the elements
+ *  of a \p thrust::device_vector using the \p thrust::device parallelization policy:
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdio>
+ *  ...
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      // note that using printf in a __device__ function requires
+ *      // code compiled for a GPU with compute capability 2.0 or
+ *      // higher (nvcc --arch=sm_20)
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  thrust::device_vector<int> d_vec(3);
+ *  d_vec[0] = 0; d_vec[1] = 1; d_vec[2] = 2;
+ *
+ *  thrust::for_each(thrust::device, d_vec.begin(), d_vec.end(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ *
+ *  \see for_each_n
+ *  \see https://en.cppreference.com/w/cpp/algorithm/for_each
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename UnaryFunction>
+__host__ __device__
+InputIterator for_each(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator first,
+                       InputIterator last,
+                       UnaryFunction f);
+
+
+/*! \p for_each_n applies the function object \p f to each element
+ *  in the range <tt>[first, first + n)</tt>; \p f's return value, if any,
+ *  is ignored. Unlike the C++ Standard Template Library function
+ *  <tt>std::for_each</tt>, this version offers no guarantee on
+ *  order of execution.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param n The size of the input sequence.
+ *  \param f The function object to apply to the range <tt>[first, first + n)</tt>.
+ *  \return <tt>first + n</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p UnaryFunction's \c argument_type.
+ *  \tparam Size is an integral type.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>,
+ *          and \p UnaryFunction does not apply any non-constant operation through its argument.
+ *
+ *  The following code snippet demonstrates how to use \p for_each_n to print the elements
+ *  of a \p device_vector using the \p thrust::device parallelization policy.
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdio>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      // note that using printf in a __device__ function requires
+ *      // code compiled for a GPU with compute capability 2.0 or
+ *      // higher (nvcc --arch=sm_20)
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  thrust::device_vector<int> d_vec(3);
+ *  d_vec[0] = 0; d_vec[1] = 1; d_vec[2] = 2;
+ *
+ *  thrust::for_each_n(thrust::device, d_vec.begin(), d_vec.size(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ *
+ *  \see for_each
+ *  \see https://en.cppreference.com/w/cpp/algorithm/for_each
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename UnaryFunction>
+__host__ __device__
+InputIterator for_each_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator first,
+                         Size n,
+                         UnaryFunction f);
+
+/*! \p for_each applies the function object \p f to each element
+ *  in the range <tt>[first, last)</tt>; \p f's return value, if any,
+ *  is ignored. Unlike the C++ Standard Template Library function
+ *  <tt>std::for_each</tt>, this version offers no guarantee on
+ *  order of execution. For this reason, this version of \p for_each
+ *  does not return a copy of the function object.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param f The function object to apply to the range <tt>[first, last)</tt>.
+ *  \return last
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p UnaryFunction's \c argument_type.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>,
+ *          and \p UnaryFunction does not apply any non-constant operation through its argument.
+ *
+ *  The following code snippet demonstrates how to use \p for_each to print the elements
+ *  of a \p device_vector.
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <stdio.h>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      // note that using printf in a __device__ function requires
+ *      // code compiled for a GPU with compute capability 2.0 or
+ *      // higher (nvcc --arch=sm_20)
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  thrust::device_vector<int> d_vec(3);
+ *  d_vec[0] = 0; d_vec[1] = 1; d_vec[2] = 2;
+ *
+ *  thrust::for_each(d_vec.begin(), d_vec.end(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ *
+ *  \see for_each_n
+ *  \see https://en.cppreference.com/w/cpp/algorithm/for_each
+ */
+template<typename InputIterator,
+         typename UnaryFunction>
+InputIterator for_each(InputIterator first,
+                       InputIterator last,
+                       UnaryFunction f);
+
+
+/*! \p for_each_n applies the function object \p f to each element
+ *  in the range <tt>[first, first + n)</tt>; \p f's return value, if any,
+ *  is ignored. Unlike the C++ Standard Template Library function
+ *  <tt>std::for_each</tt>, this version offers no guarantee on
+ *  order of execution.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param n The size of the input sequence.
+ *  \param f The function object to apply to the range <tt>[first, first + n)</tt>.
+ *  \return <tt>first + n</tt>
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p UnaryFunction's \c argument_type.
+ *  \tparam Size is an integral type.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>,
+ *          and \p UnaryFunction does not apply any non-constant operation through its argument.
+ *
+ *  The following code snippet demonstrates how to use \p for_each_n to print the elements
+ *  of a \p device_vector.
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <stdio.h>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      // note that using printf in a __device__ function requires
+ *      // code compiled for a GPU with compute capability 2.0 or
+ *      // higher (nvcc --arch=sm_20)
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  thrust::device_vector<int> d_vec(3);
+ *  d_vec[0] = 0; d_vec[1] = 1; d_vec[2] = 2;
+ *
+ *  thrust::for_each_n(d_vec.begin(), d_vec.size(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ *
+ *  \see for_each
+ *  \see https://en.cppreference.com/w/cpp/algorithm/for_each
+ */
+template<typename InputIterator,
+         typename Size,
+         typename UnaryFunction>
+InputIterator for_each_n(InputIterator first,
+                         Size n,
+                         UnaryFunction f);
+
+/*! \} // end modifying
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/for_each.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/functional.h b/thirdparty/manifold/thirdparty/thrust/thrust/functional.h
new file mode 100644
index 000000000000..0608f4b3d454
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/functional.h
@@ -0,0 +1,1719 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file functional.h
+ *  \brief Function objects and tools for manipulating them
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <functional>
+#include <thrust/detail/functional/placeholder.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup function_objects Function Objects
+ */
+
+template<typename Operation> struct unary_traits;
+
+template<typename Operation> struct binary_traits;
+
+/*! \addtogroup function_object_adaptors Function Object Adaptors
+ *  \ingroup function_objects
+ *  \{
+ */
+
+/*! \p unary_function is an empty base class: it contains no member functions
+ *  or member variables, but only type information. The only reason it exists
+ *  is to make it more convenient to define types that are models of the
+ *  concept Adaptable Unary Function. Specifically, any model of Adaptable
+ *  Unary Function must define nested \c typedefs. Those \c typedefs are
+ *  provided by the base class \p unary_function.
+ *
+ *  The following code snippet demonstrates how to construct an
+ *  Adaptable Unary Function using \p unary_function.
+ *
+ *  \code
+ *  struct sine : public thrust::unary_function<float,float>
+ *  {
+ *    __host__ __device__
+ *    float operator()(float x) { return sinf(x); }
+ *  };
+ *  \endcode
+ *
+ *  \note Because C++11 language support makes the functionality of
+ *        \c unary_function obsolete, its use is optional if C++11 language
+ *        features are enabled.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/unary_function
+ *  \see binary_function
+ */
+template<typename Argument,
+         typename Result>
+struct unary_function
+{
+  /*! \typedef argument_type
+   *  \brief The type of the function object's argument.
+   */
+  typedef Argument argument_type;
+
+  /*! \typedef result_type;
+   *  \brief The type of the function object's result.
+   */
+  typedef Result   result_type;
+}; // end unary_function
+
+/*! \p binary_function is an empty base class: it contains no member functions
+ *  or member variables, but only type information. The only reason it exists
+ *  is to make it more convenient to define types that are models of the
+ *  concept Adaptable Binary Function. Specifically, any model of Adaptable
+ *  Binary Function must define nested \c typedefs. Those \c typedefs are
+ *  provided by the base class \p binary_function.
+ *
+ *  The following code snippet demonstrates how to construct an
+ *  Adaptable Binary Function using \p binary_function.
+ *
+ *  \code
+ *  struct exponentiate : public thrust::binary_function<float,float,float>
+ *  {
+ *    __host__ __device__
+ *    float operator()(float x, float y) { return powf(x,y); }
+ *  };
+ *  \endcode
+ *
+ *  \note Because C++11 language support makes the functionality of
+ *        \c binary_function obsolete, its use is optional if C++11 language
+ *        features are enabled.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/binary_function
+ *  \see unary_function
+ */
+template<typename Argument1,
+         typename Argument2,
+         typename Result>
+struct binary_function
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef Argument1 first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef Argument2 second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef Result    result_type;
+}; // end binary_function
+
+/*! \}
+ */
+
+
+/*! \addtogroup predefined_function_objects Predefined Function Objects
+ *  \ingroup function_objects
+ */
+
+/*! \addtogroup arithmetic_operations Arithmetic Operations
+ *  \ingroup predefined_function_objects
+ *  \{
+ */
+
+#define THRUST_UNARY_FUNCTOR_VOID_SPECIALIZATION(func, impl)                   \
+  template <>                                                                  \
+  struct func<void>                                                            \
+  {                                                                            \
+    using is_transparent = void;                                               \
+    __thrust_exec_check_disable__                                              \
+    template <typename T>                                                      \
+    __host__ __device__                                                        \
+    constexpr auto operator()(T&& x) const                                     \
+      noexcept(noexcept(impl)) THRUST_TRAILING_RETURN(decltype(impl))          \
+    {                                                                          \
+      return impl;                                                             \
+    }                                                                          \
+  }
+
+#define THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION(func, impl)                  \
+  template <>                                                                  \
+  struct func<void>                                                            \
+  {                                                                            \
+    using is_transparent = void;                                               \
+    __thrust_exec_check_disable__                                              \
+    template <typename T1, typename T2>                                        \
+    __host__ __device__                                                        \
+    constexpr auto operator()(T1&& t1, T2&& t2) const                          \
+      noexcept(noexcept(impl)) THRUST_TRAILING_RETURN(decltype(impl))          \
+    {                                                                          \
+      return impl;                                                             \
+    }                                                                          \
+  }
+
+#define THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(func, op)                 \
+  THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION(                                   \
+    func, THRUST_FWD(t1) op THRUST_FWD(t2))
+
+
+/*! \p plus is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>plus<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x+y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x+y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>plus</tt> to sum two
+ *  device_vectors of \c floats.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *  thrust::device_vector<float> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 75);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::plus<float>());
+ *  // V3 is now {76, 77, 78, ..., 1075}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/plus
+ *  \see binary_function
+ */
+template<typename T = void>
+struct plus
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs + rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs + rhs;
+  }
+}; // end plus
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(plus, +);
+
+/*! \p minus is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>minus<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x-y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x-y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>minus</tt> to subtract
+ *  a device_vector of \c floats from another.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *  thrust::device_vector<float> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 75);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::minus<float>());
+ *  // V3 is now {-74, -73, -72, ..., 925}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/minus
+ *  \see binary_function
+ */
+template<typename T = void>
+struct minus
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs - rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs - rhs;
+  }
+}; // end minus
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(minus, -);
+
+/*! \p multiplies is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>multiplies<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x*y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x*y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>multiplies</tt> to multiply
+ *  two device_vectors of \c floats.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *  thrust::device_vector<float> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 75);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::multiplies<float>());
+ *  // V3 is now {75, 150, 225, ..., 75000}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/multiplies
+ *  \see binary_function
+ */
+template<typename T = void>
+struct multiplies
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs * rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs * rhs;
+  }
+}; // end multiplies
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(multiplies, *);
+
+/*! \p divides is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>divides<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x/y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x/y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>divides</tt> to divide
+ *  one device_vectors of \c floats by another.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *  thrust::device_vector<float> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 75);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::divides<float>());
+ *  // V3 is now {1/75, 2/75, 3/75, ..., 1000/75}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/divides
+ *  \see binary_function
+ */
+template<typename T = void>
+struct divides
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs / rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs / rhs;
+  }
+}; // end divides
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(divides, /);
+
+/*! \p modulus is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>modulus<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x \% y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x \% y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>modulus</tt> to take
+ *  the modulus of one device_vectors of \c floats by another.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *  thrust::device_vector<float> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 75);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::modulus<int>());
+ *  // V3 is now {1%75, 2%75, 3%75, ..., 1000%75}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/modulus
+ *  \see binary_function
+ */
+template<typename T = void>
+struct modulus
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs % rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs % rhs;
+  }
+}; // end modulus
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(modulus, %);
+
+/*! \p negate is a function object. Specifically, it is an Adaptable Unary Function.
+ *  If \c f is an object of class <tt>negate<T></tt>, and \c x is an object
+ *  of class \c T, then <tt>f(x)</tt> returns <tt>-x</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x is an object of type \p T, then <tt>-x</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>negate</tt> to negate
+ *  the elements of a device_vector of \c floats.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(),
+ *                    thrust::negate<float>());
+ *  // V2 is now {-1, -2, -3, ..., -1000}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/negate
+ *  \see unary_function
+ */
+template<typename T = void>
+struct negate
+{
+  /*! \typedef argument_type
+   *  \brief The type of the function object's argument.
+   */
+  typedef T argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>-x</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &x) const
+  {
+    return -x;
+  }
+}; // end negate
+
+THRUST_UNARY_FUNCTOR_VOID_SPECIALIZATION(negate, -THRUST_FWD(x));
+
+/*! \p square is a function object. Specifically, it is an Adaptable Unary Function.
+ *  If \c f is an object of class <tt>square<T></tt>, and \c x is an object
+ *  of class \c T, then <tt>f(x)</tt> returns <tt>x*x</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x is an object of type \p T, then <tt>x*x</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>square</tt> to square
+ *  the elements of a device_vector of \c floats.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<float> V1(N);
+ *  thrust::device_vector<float> V2(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(),
+ *                    thrust::square<float>());
+ *  // V2 is now {1, 4, 9, ..., 1000000}
+ *  \endcode
+ *
+ *  \see unary_function
+ */
+template<typename T = void>
+struct square
+{
+  /*! \typedef argument_type
+   *  \brief The type of the function object's argument.
+   */
+  typedef T argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>x*x</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &x) const
+  {
+    return x*x;
+  }
+}; // end square
+
+THRUST_UNARY_FUNCTOR_VOID_SPECIALIZATION(square, x*x);
+
+/*! \}
+ */
+
+/*! \addtogroup comparison_operations Comparison Operations
+ *  \ingroup predefined_function_objects
+ *  \{
+ */
+
+/*! \p equal_to is a function object. Specifically, it is an Adaptable Binary
+ *  Predicate, which means it is a function object that tests the truth or falsehood
+ *  of some condition. If \c f is an object of class <tt>equal_to<T></tt> and \c x
+ *  and \c y are objects of class \c T, then <tt>f(x,y)</tt> returns \c true if
+ *  <tt>x == y</tt> and \c false otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/equal_to
+ *  \see binary_function
+ */
+template<typename T = void>
+struct equal_to
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs == rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs == rhs;
+  }
+}; // end equal_to
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(equal_to, ==);
+
+/*! \p not_equal_to is a function object. Specifically, it is an Adaptable Binary
+ *  Predicate, which means it is a function object that tests the truth or falsehood
+ *  of some condition. If \c f is an object of class <tt>not_equal_to<T></tt> and \c x
+ *  and \c y are objects of class \c T, then <tt>f(x,y)</tt> returns \c true if
+ *  <tt>x != y</tt> and \c false otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/not_equal_to
+ *  \see binary_function
+ */
+template<typename T = void>
+struct not_equal_to
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs != rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs != rhs;
+  }
+}; // end not_equal_to
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(not_equal_to, !=);
+
+/*! \p greater is a function object. Specifically, it is an Adaptable Binary
+ *  Predicate, which means it is a function object that tests the truth or falsehood
+ *  of some condition. If \c f is an object of class <tt>greater<T></tt> and \c x
+ *  and \c y are objects of class \c T, then <tt>f(x,y)</tt> returns \c true if
+ *  <tt>x > y</tt> and \c false otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/greater
+ *  \see binary_function
+ */
+template<typename T = void>
+struct greater
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs > rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs > rhs;
+  }
+}; // end greater
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(greater, >);
+
+/*! \p less is a function object. Specifically, it is an Adaptable Binary
+ *  Predicate, which means it is a function object that tests the truth or falsehood
+ *  of some condition. If \c f is an object of class <tt>less<T></tt> and \c x
+ *  and \c y are objects of class \c T, then <tt>f(x,y)</tt> returns \c true if
+ *  <tt>x < y</tt> and \c false otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/less
+ *  \see binary_function
+ */
+template<typename T = void>
+struct less
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs < rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs < rhs;
+  }
+}; // end less
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(less, <);
+
+/*! \p greater_equal is a function object. Specifically, it is an Adaptable Binary
+ *  Predicate, which means it is a function object that tests the truth or falsehood
+ *  of some condition. If \c f is an object of class <tt>greater_equal<T></tt> and \c x
+ *  and \c y are objects of class \c T, then <tt>f(x,y)</tt> returns \c true if
+ *  <tt>x >= y</tt> and \c false otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/greater_equal
+ *  \see binary_function
+ */
+template<typename T = void>
+struct greater_equal
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs >= rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs >= rhs;
+  }
+}; // end greater_equal
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(greater_equal, >=);
+
+/*! \p less_equal is a function object. Specifically, it is an Adaptable Binary
+ *  Predicate, which means it is a function object that tests the truth or falsehood
+ *  of some condition. If \c f is an object of class <tt>less_equal<T></tt> and \c x
+ *  and \c y are objects of class \c T, then <tt>f(x,y)</tt> returns \c true if
+ *  <tt>x <= y</tt> and \c false otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/less_equal
+ *  \see binary_function
+ */
+template<typename T = void>
+struct less_equal
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs <= rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs <= rhs;
+  }
+}; // end less_equal
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(less_equal, <=);
+
+/*! \}
+ */
+
+
+/*! \addtogroup logical_operations Logical Operations
+ *  \ingroup predefined_function_objects
+ *  \{
+ */
+
+/*! \p logical_and is a function object. Specifically, it is an Adaptable Binary Predicate,
+ *  which means it is a function object that tests the truth or falsehood of some condition.
+ *  If \c f is an object of class <tt>logical_and<T></tt> and \c x and \c y are objects of
+ *  class \c T (where \c T is convertible to \c bool) then <tt>f(x,y)</tt> returns \c true
+ *  if and only if both \c x and \c y are \c true.
+ *
+ *  \tparam T must be convertible to \c bool.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/logical_and
+ *  \see binary_function
+ */
+template<typename T = void>
+struct logical_and
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs && rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs && rhs;
+  }
+}; // end logical_and
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(logical_and, &&);
+
+/*! \p logical_or is a function object. Specifically, it is an Adaptable Binary Predicate,
+ *  which means it is a function object that tests the truth or falsehood of some condition.
+ *  If \c f is an object of class <tt>logical_or<T></tt> and \c x and \c y are objects of
+ *  class \c T (where \c T is convertible to \c bool) then <tt>f(x,y)</tt> returns \c true
+ *  if and only if either \c x or \c y are \c true.
+ *
+ *  \tparam T must be convertible to \c bool.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/logical_or
+ *  \see binary_function
+ */
+template<typename T = void>
+struct logical_or
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>lhs || rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs || rhs;
+  }
+}; // end logical_or
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(logical_or, ||);
+
+/*! \p logical_not is a function object. Specifically, it is an Adaptable Predicate,
+ *  which means it is a function object that tests the truth or falsehood of some condition.
+ *  If \c f is an object of class <tt>logical_not<T></tt> and \c x is an object of
+ *  class \c T (where \c T is convertible to \c bool) then <tt>f(x)</tt> returns \c true
+ *  if and only if \c x is \c false.
+ *
+ *  \tparam T must be convertible to \c bool.
+ *
+ *  The following code snippet demonstrates how to use \p logical_not to transform
+ *  a device_vector of \c bools into its logical complement.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  thrust::device_vector<bool> V;
+ *  ...
+ *  thrust::transform(V.begin(), V.end(), V.begin(), thrust::logical_not<bool>());
+ *  // The elements of V are now the logical complement of what they were prior
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/logical_not
+ *  \see unary_function
+ */
+template<typename T = void>
+struct logical_not
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef bool result_type;
+
+  /*! Function call operator. The return value is <tt>!x</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool operator()(const T &x) const
+  {
+    return !x;
+  }
+}; // end logical_not
+
+THRUST_UNARY_FUNCTOR_VOID_SPECIALIZATION(logical_not, !THRUST_FWD(x));
+
+/*! \}
+ */
+
+/*! \addtogroup bitwise_operations Bitwise Operations
+ *  \ingroup predefined_function_objects
+ *  \{
+ */
+
+/*! \p bit_and is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>bit_and<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x&y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x&y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>bit_and</tt> to take
+ *  the bitwise AND of one device_vector of \c ints by another.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<int> V1(N);
+ *  thrust::device_vector<int> V2(N);
+ *  thrust::device_vector<int> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 13);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::bit_and<int>());
+ *  // V3 is now {1&13, 2&13, 3&13, ..., 1000%13}
+ *  \endcode
+ *
+ *  \see binary_function
+ */
+template<typename T = void>
+struct bit_and
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs & rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs & rhs;
+  }
+}; // end bit_and
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(bit_and, &);
+
+/*! \p bit_or is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>bit_and<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x|y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x|y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>bit_or</tt> to take
+ *  the bitwise OR of one device_vector of \c ints by another.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<int> V1(N);
+ *  thrust::device_vector<int> V2(N);
+ *  thrust::device_vector<int> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 13);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::bit_or<int>());
+ *  // V3 is now {1|13, 2|13, 3|13, ..., 1000|13}
+ *  \endcode
+ *
+ *  \see binary_function
+ */
+template<typename T = void>
+struct bit_or
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs | rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs | rhs;
+  }
+}; // end bit_or
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(bit_or, |);
+
+/*! \p bit_xor is a function object. Specifically, it is an Adaptable Binary Function.
+ *  If \c f is an object of class <tt>bit_and<T></tt>, and \c x and \c y are objects
+ *  of class \c T, then <tt>f(x,y)</tt> returns <tt>x^y</tt>.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x and \c y are objects of type \p T, then <tt>x^y</tt> must be defined and must have a return type that is convertible to \c T.
+ *
+ *  The following code snippet demonstrates how to use <tt>bit_xor</tt> to take
+ *  the bitwise XOR of one device_vector of \c ints by another.
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/fill.h>
+ *  #include <thrust/transform.h>
+ *  ...
+ *  const int N = 1000;
+ *  thrust::device_vector<int> V1(N);
+ *  thrust::device_vector<int> V2(N);
+ *  thrust::device_vector<int> V3(N);
+ *
+ *  thrust::sequence(V1.begin(), V1.end(), 1);
+ *  thrust::fill(V2.begin(), V2.end(), 13);
+ *
+ *  thrust::transform(V1.begin(), V1.end(), V2.begin(), V3.begin(),
+ *                    thrust::bit_xor<int>());
+ *  // V3 is now {1^13, 2^13, 3^13, ..., 1000^13}
+ *  \endcode
+ *
+ *  \see binary_function
+ */
+template<typename T = void>
+struct bit_xor
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs ^ rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs ^ rhs;
+  }
+}; // end bit_xor
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP(bit_xor, ^);
+
+/*! \}
+ */
+
+/*! \addtogroup generalized_identity_operations Generalized Identity Operations
+ *  \ingroup predefined_function_objects
+ *  \{
+ */
+
+/*! \p identity is a Unary Function that represents the identity function: it takes
+ *  a single argument \c x, and returns \c x.
+ *
+ *  \tparam T No requirements on \p T.
+ *
+ *  The following code snippet demonstrates that \p identity returns its
+ *  argument.
+ *
+ *  \code
+ *  #include <thrust/functional.h>
+ *  #include <assert.h>
+ *  ...
+ *  int x = 137;
+ *  thrust::identity<int> id;
+ *  assert(x == id(x));
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/identity
+ *  \see unary_function
+ */
+template<typename T = void>
+struct identity
+{
+  /*! \typedef argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>x</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T &operator()(const T &x) const
+  {
+    return x;
+  }
+}; // end identity
+
+THRUST_UNARY_FUNCTOR_VOID_SPECIALIZATION(identity, THRUST_FWD(x));
+
+/*! \p maximum is a function object that takes two arguments and returns the greater
+ *  of the two. Specifically, it is an Adaptable Binary Function. If \c f is an
+ *  object of class <tt>maximum<T></tt> and \c x and \c y are objects of class \c T
+ *  <tt>f(x,y)</tt> returns \c x if <tt>x > y</tt> and \c y, otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  The following code snippet demonstrates that \p maximum returns its
+ *  greater argument.
+ *
+ *  \code
+ *  #include <thrust/functional.h>
+ *  #include <assert.h>
+ *  ...
+ *  int x =  137;
+ *  int y = -137;
+ *  thrust::maximum<int> mx;
+ *  assert(x == mx(x,y));
+ *  \endcode
+ *
+ *  \see minimum
+ *  \see min
+ *  \see binary_function
+ */
+template<typename T = void>
+struct maximum
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>rhs < lhs ? lhs : rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs < rhs ? rhs : lhs;
+  }
+}; // end maximum
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION(maximum,
+                                          t1 < t2 ? THRUST_FWD(t2)
+                                                  : THRUST_FWD(t1));
+
+/*! \p minimum is a function object that takes two arguments and returns the lesser
+ *  of the two. Specifically, it is an Adaptable Binary Function. If \c f is an
+ *  object of class <tt>minimum<T></tt> and \c x and \c y are objects of class \c T
+ *  <tt>f(x,y)</tt> returns \c x if <tt>x < y</tt> and \c y, otherwise.
+ *
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  The following code snippet demonstrates that \p minimum returns its
+ *  lesser argument.
+ *
+ *  \code
+ *  #include <thrust/functional.h>
+ *  #include <assert.h>
+ *  ...
+ *  int x =  137;
+ *  int y = -137;
+ *  thrust::minimum<int> mn;
+ *  assert(y == mn(x,y));
+ *  \endcode
+ *
+ *  \see maximum
+ *  \see max
+ *  \see binary_function
+ */
+template<typename T = void>
+struct minimum
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T result_type;
+
+  /*! Function call operator. The return value is <tt>lhs < rhs ? lhs : rhs</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr T operator()(const T &lhs, const T &rhs) const
+  {
+    return lhs < rhs ? lhs : rhs;
+  }
+}; // end minimum
+
+THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION(minimum,
+                                          t1 < t2 ? THRUST_FWD(t1)
+                                                  : THRUST_FWD(t2));
+
+/*! \p project1st is a function object that takes two arguments and returns
+ *  its first argument; the second argument is unused. It is essentially a
+ *  generalization of identity to the case of a Binary Function.
+ *
+ *  \code
+ *  #include <thrust/functional.h>
+ *  #include <assert.h>
+ *  ...
+ *  int x =  137;
+ *  int y = -137;
+ *  thrust::project1st<int> pj1;
+ *  assert(x == pj1(x,y));
+ *  \endcode
+ *
+ *  \see identity
+ *  \see project2nd
+ *  \see binary_function
+ */
+template<typename T1 = void, typename T2 = void>
+struct project1st
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T1 first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T2 second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T1 result_type;
+
+  /*! Function call operator. The return value is <tt>lhs</tt>.
+   */
+  __host__ __device__
+  constexpr const T1 &operator()(const T1 &lhs, const T2 & /*rhs*/) const
+  {
+    return lhs;
+  }
+}; // end project1st
+
+template <>
+struct project1st<void, void>
+{
+  using is_transparent = void;
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&& t1, T2&&) const
+    noexcept(noexcept(THRUST_FWD(t1)))
+    THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t1)))
+  {
+    return THRUST_FWD(t1);
+  }
+};
+
+/*! \p project2nd is a function object that takes two arguments and returns
+ *  its second argument; the first argument is unused. It is essentially a
+ *  generalization of identity to the case of a Binary Function.
+ *
+ *  \code
+ *  #include <thrust/functional.h>
+ *  #include <assert.h>
+ *  ...
+ *  int x =  137;
+ *  int y = -137;
+ *  thrust::project2nd<int> pj2;
+ *  assert(y == pj2(x,y));
+ *  \endcode
+ *
+ *  \see identity
+ *  \see project1st
+ *  \see binary_function
+ */
+template<typename T1 = void, typename T2 = void>
+struct project2nd
+{
+  /*! \typedef first_argument_type
+   *  \brief The type of the function object's first argument.
+   */
+  typedef T1 first_argument_type;
+
+  /*! \typedef second_argument_type
+   *  \brief The type of the function object's second argument.
+   */
+  typedef T2 second_argument_type;
+
+  /*! \typedef result_type
+   *  \brief The type of the function object's result;
+   */
+  typedef T2 result_type;
+
+  /*! Function call operator. The return value is <tt>rhs</tt>.
+   */
+  __host__ __device__
+  constexpr const T2 &operator()(const T1 &/*lhs*/, const T2 &rhs) const
+  {
+    return rhs;
+  }
+}; // end project2nd
+
+template <>
+struct project2nd<void, void>
+{
+  using is_transparent = void;
+  __thrust_exec_check_disable__
+  template <typename T1, typename T2>
+  __host__ __device__
+  constexpr auto operator()(T1&&, T2&& t2) const
+  noexcept(noexcept(THRUST_FWD(t2)))
+  THRUST_TRAILING_RETURN(decltype(THRUST_FWD(t2)))
+  {
+    return THRUST_FWD(t2);
+  }
+};
+
+/*! \}
+ */
+
+// odds and ends
+
+/*! \addtogroup function_object_adaptors
+ *  \{
+ */
+
+/*! \p unary_negate is a function object adaptor: it is an Adaptable Predicate
+ *  that represents the logical negation of some other Adaptable Predicate.
+ *  That is: if \c f is an object of class <tt>unary_negate<AdaptablePredicate></tt>,
+ *  then there exists an object \c pred of class \c AdaptablePredicate such
+ *  that <tt>f(x)</tt> always returns the same value as <tt>!pred(x)</tt>.
+ *  There is rarely any reason to construct a <tt>unary_negate</tt> directly;
+ *  it is almost always easier to use the helper function not1.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/unary_negate
+ *  \see not1
+ */
+template<typename Predicate>
+struct unary_negate
+    : public thrust::unary_function<typename Predicate::argument_type, bool>
+{
+  /*! Constructor takes a \p Predicate object to negate.
+   *  \param p The \p Predicate object to negate.
+   */
+  __host__ __device__
+  explicit unary_negate(Predicate p) : pred(p){}
+
+  /*! Function call operator. The return value is <tt>!pred(x)</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  bool operator()(const typename Predicate::argument_type& x) { return !pred(x); }
+
+  /*! \cond
+   */
+  Predicate pred;
+  /*! \endcond
+   */
+}; // end unary_negate
+
+/*! \p not1 is a helper function to simplify the creation of Adaptable Predicates:
+ *  it takes an Adaptable Predicate \p pred as an argument and returns a new Adaptable
+ *  Predicate that represents the negation of \p pred. That is: if \c pred is an object
+ *  of a type which models Adaptable Predicate, then the the type of the result
+ *  \c npred of <tt>not1(pred)</tt> is also a model of Adaptable Predicate and
+ *  <tt>npred(x)</tt> always returns the same value as <tt>!pred(x)</tt>.
+ *
+ *  \param pred The Adaptable Predicate to negate.
+ *  \return A new object, <tt>npred</tt> such that <tt>npred(x)</tt> always returns
+ *          the same value as <tt>!pred(x)</tt>.
+ *
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_negate">Adaptable Predicate</a>.
+ *
+ *  \see unary_negate
+ *  \see not2
+ */
+template<typename Predicate>
+  __host__ __device__
+  unary_negate<Predicate> not1(const Predicate &pred);
+
+/*! \p binary_negate is a function object adaptor: it is an Adaptable Binary
+ *  Predicate that represents the logical negation of some other Adaptable
+ *  Binary Predicate. That is: if \c f is an object of class <tt>binary_negate<AdaptablePredicate></tt>,
+ *  then there exists an object \c pred of class \c AdaptableBinaryPredicate
+ *  such that <tt>f(x,y)</tt> always returns the same value as <tt>!pred(x,y)</tt>.
+ *  There is rarely any reason to construct a <tt>binary_negate</tt> directly;
+ *  it is almost always easier to use the helper function not2.
+ *
+ *  \see https://en.cppreference.com/w/cpp/utility/functional/binary_negate
+ */
+template<typename Predicate>
+struct binary_negate
+    : public thrust::binary_function<typename Predicate::first_argument_type,
+                                     typename Predicate::second_argument_type,
+                                     bool>
+{
+  /*! Constructor takes a \p Predicate object to negate.
+   *  \param p The \p Predicate object to negate.
+   */
+  __host__ __device__
+  explicit binary_negate(Predicate p) : pred(p){}
+
+  /*! Function call operator. The return value is <tt>!pred(x,y)</tt>.
+   */
+  __thrust_exec_check_disable__
+  __host__ __device__
+  bool operator()(const typename Predicate::first_argument_type& x, const typename Predicate::second_argument_type& y)
+  {
+      return !pred(x,y);
+  }
+
+  /*! \cond
+   */
+  Predicate pred;
+  /*! \endcond
+   */
+}; // end binary_negate
+
+/*! \p not2 is a helper function to simplify the creation of Adaptable Binary Predicates:
+ *  it takes an Adaptable Binary Predicate \p pred as an argument and returns a new Adaptable
+ *  Binary Predicate that represents the negation of \p pred. That is: if \c pred is an object
+ *  of a type which models Adaptable Binary Predicate, then the the type of the result
+ *  \c npred of <tt>not2(pred)</tt> is also a model of Adaptable Binary Predicate and
+ *  <tt>npred(x,y)</tt> always returns the same value as <tt>!pred(x,y)</tt>.
+ *
+ *  \param pred The Adaptable Binary Predicate to negate.
+ *  \return A new object, <tt>npred</tt> such that <tt>npred(x,y)</tt> always returns
+ *          the same value as <tt>!pred(x,y)</tt>.
+ *
+ *  \tparam Binary Predicate is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/AdaptableBinaryPredicate">Adaptable Binary Predicate</a>.
+ *
+ *  \see binary_negate
+ *  \see not1
+ */
+template<typename BinaryPredicate>
+  __host__ __device__
+  binary_negate<BinaryPredicate> not2(const BinaryPredicate &pred);
+
+/*! \}
+ */
+
+
+/*! \addtogroup placeholder_objects Placeholder Objects
+ *  \ingroup function_objects
+ *  \{
+ */
+
+
+/*! \namespace thrust::placeholders
+ *  \brief Facilities for constructing simple functions inline.
+ *
+ *  Objects in the \p thrust::placeholders namespace may be used to create simple arithmetic functions inline
+ *  in an algorithm invocation. Combining placeholders such as \p _1 and \p _2 with arithmetic operations such as \c +
+ *  creates an unnamed function object which applies the operation to their arguments.
+ *
+ *  The type of placeholder objects is implementation-defined.
+ *
+ *  The following code snippet demonstrates how to use the placeholders \p _1 and \p _2 with \p thrust::transform
+ *  to implement the SAXPY computation:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *
+ *  int main()
+ *  {
+ *    thrust::device_vector<float> x(4), y(4);
+ *    x[0] = 1;
+ *    x[1] = 2;
+ *    x[2] = 3;
+ *    x[3] = 4;
+ *
+ *    y[0] = 1;
+ *    y[1] = 1;
+ *    y[2] = 1;
+ *    y[3] = 1;
+ *
+ *    float a = 2.0f;
+ *
+ *    using namespace thrust::placeholders;
+ *
+ *    thrust::transform(x.begin(), x.end(), y.begin(), y.begin(),
+ *      a * _1 + _2
+ *    );
+ *
+ *    // y is now {3, 5, 7, 9}
+ *  }
+ *  \endcode
+ */
+namespace placeholders
+{
+
+
+/*! \p thrust::placeholders::_1 is the placeholder for the first function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<0>::type _1;
+
+
+/*! \p thrust::placeholders::_2 is the placeholder for the second function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<1>::type _2;
+
+
+/*! \p thrust::placeholders::_3 is the placeholder for the third function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<2>::type _3;
+
+
+/*! \p thrust::placeholders::_4 is the placeholder for the fourth function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<3>::type _4;
+
+
+/*! \p thrust::placeholders::_5 is the placeholder for the fifth function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<4>::type _5;
+
+
+/*! \p thrust::placeholders::_6 is the placeholder for the sixth function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<5>::type _6;
+
+
+/*! \p thrust::placeholders::_7 is the placeholder for the seventh function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<6>::type _7;
+
+
+/*! \p thrust::placeholders::_8 is the placeholder for the eighth function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<7>::type _8;
+
+
+/*! \p thrust::placeholders::_9 is the placeholder for the ninth function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<8>::type _9;
+
+
+/*! \p thrust::placeholders::_10 is the placeholder for the tenth function parameter.
+ */
+THRUST_INLINE_CONSTANT thrust::detail::functional::placeholder<9>::type _10;
+
+
+} // end placeholders
+
+
+/*! \} // placeholder_objects
+ */
+
+#undef THRUST_UNARY_FUNCTOR_VOID_SPECIALIZATION
+#undef THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION
+#undef THRUST_BINARY_FUNCTOR_VOID_SPECIALIZATION_OP
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/functional.inl>
+#include <thrust/detail/functional/operators.h>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/future.h b/thirdparty/manifold/thirdparty/thrust/thrust/future.h
new file mode 100644
index 000000000000..d8fb7544b177
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/future.h
@@ -0,0 +1,176 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/future.h
+ *  \brief `thrust::future`, an asynchronous value type.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp14_required.h>
+
+#if THRUST_CPP_DIALECT >= 2014
+
+#include <thrust/execution_policy.h>
+#include <thrust/detail/static_assert.h>
+
+#include <utility>
+
+/*
+// #include the host system's pointer.h header.
+#define __THRUST_HOST_SYSTEM_POINTER_HEADER <__THRUST_HOST_SYSTEM_ROOT/pointer.h>
+  #include __THRUST_HOST_SYSTEM_POINTER_HEADER
+#undef __THRUST_HOST_SYSTEM_POINTER_HEADER
+*/
+
+// #include the device system's pointer.h header.
+#define __THRUST_DEVICE_SYSTEM_POINTER_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/pointer.h>
+  #include __THRUST_DEVICE_SYSTEM_POINTER_HEADER
+#undef __THRUST_DEVICE_SYSTEM_POINTER_HEADER
+
+/*
+// #include the host system's future.h header.
+#define __THRUST_HOST_SYSTEM_FUTURE_HEADER <__THRUST_HOST_SYSTEM_ROOT/future.h>
+  #include __THRUST_HOST_SYSTEM_FUTURE_HEADER
+#undef __THRUST_HOST_SYSTEM_FUTURE_HEADER
+*/
+
+// #include the device system's future.h header.
+#define __THRUST_DEVICE_SYSTEM_FUTURE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/future.h>
+  #include __THRUST_DEVICE_SYSTEM_FUTURE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_FUTURE_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+///////////////////////////////////////////////////////////////////////////////
+
+// `select_unique_(future|event)_type` is a hook for choosing the
+// `unique_eager_event`/`unique_eager_future` type for a system. `decltype` is
+// used to determine the return type of an ADL call to
+// `select_unique_eager_(future|event)_type(system)`; that return type should
+// be the correct event/future type for `system`. Overloads should only be
+// declared, not defined.
+
+namespace unimplemented
+{
+
+struct no_unique_eager_event_type_found {};
+
+inline __host__ 
+no_unique_eager_event_type_found
+unique_eager_event_type(...) noexcept;
+
+struct no_unique_eager_future_type_found {};
+
+template <typename T>
+__host__ 
+no_unique_eager_future_type_found
+unique_eager_future_type(...) noexcept;
+
+} // namespace unimplemented
+
+namespace unique_eager_event_type_detail
+{
+
+using unimplemented::unique_eager_event_type;
+
+template <typename System>
+using select = decltype(
+  unique_eager_event_type(std::declval<System>())
+);
+
+} // namespace unique_eager_event_type_detail
+
+namespace unique_eager_future_type_detail
+{
+
+using unimplemented::unique_eager_future_type;
+
+template <typename System, typename T>
+using select = decltype(
+  unique_eager_future_type<T>(std::declval<System>())
+);
+
+} // namespace unique_eager_future_type_detail
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename System>
+using unique_eager_event = unique_eager_event_type_detail::select<System>;
+
+template <typename System>
+using event = unique_eager_event<System>;
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename System, typename T>
+using unique_eager_future = unique_eager_future_type_detail::select<System, T>;
+
+template <typename System, typename T>
+using future = unique_eager_future<System, T>;
+
+/*
+///////////////////////////////////////////////////////////////////////////////
+
+using host_unique_eager_event = unique_eager_event_type_detail::select<
+  thrust::system::__THRUST_HOST_SYSTEM_NAMESPACE::tag
+>;
+using host_event = host_unique_eager_event;
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T>
+using host_unique_eager_future = unique_eager_future_type_detail::select<
+  thrust::system::__THRUST_HOST_SYSTEM_NAMESPACE::tag, T
+>;
+template <typename T>
+using host_future = host_unique_eager_future<T>;
+*/
+
+///////////////////////////////////////////////////////////////////////////////
+
+using device_unique_eager_event = unique_eager_event_type_detail::select<
+  thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::tag
+>;
+
+using device_event = device_unique_eager_event;
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T>
+using device_unique_eager_future = unique_eager_future_type_detail::select<
+  thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::tag, T
+>;
+
+template <typename T>
+using device_future = device_unique_eager_future<T>;
+
+///////////////////////////////////////////////////////////////////////////////
+
+struct new_stream_t final {};
+
+THRUST_INLINE_CONSTANT new_stream_t new_stream{};
+
+///////////////////////////////////////////////////////////////////////////////
+
+using thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::when_all;
+
+///////////////////////////////////////////////////////////////////////////////
+
+THRUST_NAMESPACE_END
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/gather.h b/thirdparty/manifold/thirdparty/thrust/thrust/gather.h
new file mode 100644
index 000000000000..41acc22a34e8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/gather.h
@@ -0,0 +1,445 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file gather.h
+ *  \brief Irregular copying from a source range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup gathering
+ *  \ingroup copying
+ *  \{
+ */
+
+
+/*! \p gather copies elements from a source array into a destination range according 
+ *  to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>, the
+ *  value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
+ *  \p RandomAccessIterator must permit random access.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param map_first Beginning of the range of gather locations.
+ *  \param map_last End of the range of gather locations.
+ *  \param input_first Beginning of the source range.
+ *  \param result Beginning of the destination range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a> and \c RandomAccessIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The range <tt>[map_first, map_last)</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The input data shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *
+ *  \remark \p gather is the inverse of thrust::scatter.
+ *
+ *  The following code snippet demonstrates how to use \p gather to reorder
+ *  a range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/gather.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  // mark even indices with a 1; odd indices with a 0
+ *  int values[10] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 0};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // gather all even indices into the first half of the range
+ *  // and odd indices to the last half of the range
+ *  int map[10]   = {0, 2, 4, 6, 8, 1, 3, 5, 7, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10);
+ *  thrust::gather(thrust::device,
+ *                 d_map.begin(), d_map.end(),
+ *                 d_values.begin(),
+ *                 d_output.begin());
+ *  // d_output is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  \endcode
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                        InputIterator                                               map_first,
+                        InputIterator                                               map_last,
+                        RandomAccessIterator                                        input_first,
+                        OutputIterator                                              result);
+
+
+/*! \p gather copies elements from a source array into a destination range according 
+ *  to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>, the
+ *  value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
+ *  \p RandomAccessIterator must permit random access.
+ *
+ *  \param map_first Beginning of the range of gather locations.
+ *  \param map_last End of the range of gather locations.
+ *  \param input_first Beginning of the source range.
+ *  \param result Beginning of the destination range.
+ *
+ *  \tparam InputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a> and \c RandomAccessIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The range <tt>[map_first, map_last)</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The input data shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *
+ *  \remark \p gather is the inverse of thrust::scatter.
+ *
+ *  The following code snippet demonstrates how to use \p gather to reorder
+ *  a range.
+ *
+ *  \code
+ *  #include <thrust/gather.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  // mark even indices with a 1; odd indices with a 0
+ *  int values[10] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 0};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // gather all even indices into the first half of the range
+ *  // and odd indices to the last half of the range
+ *  int map[10]   = {0, 2, 4, 6, 8, 1, 3, 5, 7, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10);
+ *  thrust::gather(d_map.begin(), d_map.end(),
+ *                 d_values.begin(),
+ *                 d_output.begin());
+ *  // d_output is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  \endcode
+ */
+template<typename InputIterator,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+  OutputIterator gather(InputIterator        map_first,
+                        InputIterator        map_last,
+                        RandomAccessIterator input_first,
+                        OutputIterator       result);
+
+
+/*! \p gather_if conditionally copies elements from a source array into a destination 
+ *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>,
+ *  such that the value of <tt>\*(stencil + (i - map_first))</tt> is \c true, the value
+ *  <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
+ *  \p RandomAccessIterator must permit random access.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param map_first Beginning of the range of gather locations.
+ *  \param map_last End of the range of gather locations.
+ *  \param stencil Beginning of the range of predicate values.
+ *  \param input_first Beginning of the source range.
+ *  \param result Beginning of the destination range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c bool.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a> and \c RandomAccessIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The range <tt>[map_first, map_last)</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The range <tt>[stencil, stencil + (map_last - map_first))</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The input data shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *
+ *  \remark \p gather_if is the inverse of \p scatter_if.
+ *
+ *  The following code snippet demonstrates how to use \p gather_if to gather selected values from
+ *  an input range using the \p thrust::device execution policy:
+ *
+ *  \code
+ *  #include <thrust/gather.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *
+ *  int values[10] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // select elements at even-indexed locations
+ *  int stencil[10] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 0};
+ *  thrust::device_vector<int> d_stencil(stencil, stencil + 10);
+ *
+ *  // map all even indices into the first half of the range
+ *  // and odd indices to the last half of the range
+ *  int map[10]   = {0, 2, 4, 6, 8, 1, 3, 5, 7, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10, 7);
+ *  thrust::gather_if(thrust::device,
+ *                    d_map.begin(), d_map.end(),
+ *                    d_stencil.begin(),
+ *                    d_values.begin(),
+ *                    d_output.begin());
+ *  // d_output is now {0, 7, 4, 7, 8, 7, 3, 7, 7, 7}
+ *  \endcode
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1                                              map_first,
+                           InputIterator1                                              map_last,
+                           InputIterator2                                              stencil,
+                           RandomAccessIterator                                        input_first,
+                           OutputIterator                                              result);
+
+
+/*! \p gather_if conditionally copies elements from a source array into a destination 
+ *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>,
+ *  such that the value of <tt>\*(stencil + (i - map_first))</tt> is \c true, the value
+ *  <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
+ *  \p RandomAccessIterator must permit random access.
+ *
+ *  \param map_first Beginning of the range of gather locations.
+ *  \param map_last End of the range of gather locations.
+ *  \param stencil Beginning of the range of predicate values.
+ *  \param input_first Beginning of the source range.
+ *  \param result Beginning of the destination range.
+ *
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c bool.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a> and \c RandomAccessIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The range <tt>[map_first, map_last)</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The range <tt>[stencil, stencil + (map_last - map_first))</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The input data shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *
+ *  \remark \p gather_if is the inverse of \p scatter_if.
+ *
+ *  The following code snippet demonstrates how to use \p gather_if to gather selected values from
+ *  an input range.
+ *
+ *  \code
+ *  #include <thrust/gather.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *
+ *  int values[10] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // select elements at even-indexed locations
+ *  int stencil[10] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 0};
+ *  thrust::device_vector<int> d_stencil(stencil, stencil + 10);
+ *
+ *  // map all even indices into the first half of the range
+ *  // and odd indices to the last half of the range
+ *  int map[10]   = {0, 2, 4, 6, 8, 1, 3, 5, 7, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10, 7);
+ *  thrust::gather_if(d_map.begin(), d_map.end(),
+ *                    d_stencil.begin(),
+ *                    d_values.begin(),
+ *                    d_output.begin());
+ *  // d_output is now {0, 7, 4, 7, 8, 7, 3, 7, 7, 7}
+ *  \endcode
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+  OutputIterator gather_if(InputIterator1       map_first,
+                           InputIterator1       map_last,
+                           InputIterator2       stencil,
+                           RandomAccessIterator input_first,
+                           OutputIterator       result);
+
+
+/*! \p gather_if conditionally copies elements from a source array into a destination 
+ *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>
+ *  such that the value of <tt>pred(\*(stencil + (i - map_first)))</tt> is \c true,
+ *  the value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
+ *  \p RandomAccessIterator must permit random access.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param map_first Beginning of the range of gather locations.
+ *  \param map_last End of the range of gather locations.
+ *  \param stencil Beginning of the range of predicate values.
+ *  \param input_first Beginning of the source range.
+ *  \param result Beginning of the destination range.
+ *  \param pred Predicate to apply to the stencil values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c Predicate's \c argument_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a> and \c RandomAccessIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[map_first, map_last)</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The range <tt>[stencil, stencil + (map_last - map_first))</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The input data shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *
+ *  \remark \p gather_if is the inverse of \p scatter_if.
+ *
+ *  The following code snippet demonstrates how to use \p gather_if to gather selected values from
+ *  an input range based on an arbitrary selection function using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/gather.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *
+ *  int values[10] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // we will select an element when our stencil is even
+ *  int stencil[10] = {0, 3, 4, 1, 4, 1, 2, 7, 8, 9};
+ *  thrust::device_vector<int> d_stencil(stencil, stencil + 10);
+ *
+ *  // map all even indices into the first half of the range
+ *  // and odd indices to the last half of the range
+ *  int map[10]   = {0, 2, 4, 6, 8, 1, 3, 5, 7, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10, 7);
+ *  thrust::gather_if(thrust::device,
+ *                    d_map.begin(), d_map.end(),
+ *                    d_stencil.begin(),
+ *                    d_values.begin(),
+ *                    d_output.begin(),
+ *                    is_even());
+ *  // d_output is now {0, 7, 4, 7, 8, 7, 3, 7, 7, 7}
+ *  \endcode
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator gather_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1                                              map_first,
+                           InputIterator1                                              map_last,
+                           InputIterator2                                              stencil,
+                           RandomAccessIterator                                        input_first,
+                           OutputIterator                                              result,
+                           Predicate                                                   pred);
+
+
+/*! \p gather_if conditionally copies elements from a source array into a destination 
+ *  range according to a map. For each input iterator \c i in the range <tt>[map_first, map_last)</tt>
+ *  such that the value of <tt>pred(\*(stencil + (i - map_first)))</tt> is \c true,
+ *  the value <tt>input_first[\*i]</tt> is assigned to <tt>*(result + (i - map_first))</tt>.
+ *  \p RandomAccessIterator must permit random access.
+ *
+ *  \param map_first Beginning of the range of gather locations.
+ *  \param map_last End of the range of gather locations.
+ *  \param stencil Beginning of the range of predicate values.
+ *  \param input_first Beginning of the source range.
+ *  \param result Beginning of the destination range.
+ *  \param pred Predicate to apply to the stencil values.
+ *
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c Predicate's \c argument_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a> and \c RandomAccessIterator's \c value_type must be convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[map_first, map_last)</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The range <tt>[stencil, stencil + (map_last - map_first))</tt> shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *  \pre The input data shall not overlap the range <tt>[result, result + (map_last - map_first))</tt>.
+ *
+ *  \remark \p gather_if is the inverse of \p scatter_if.
+ *
+ *  The following code snippet demonstrates how to use \p gather_if to gather selected values from
+ *  an input range based on an arbitrary selection function.
+ *
+ *  \code
+ *  #include <thrust/gather.h>
+ *  #include <thrust/device_vector.h>
+ *  
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *
+ *  int values[10] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // we will select an element when our stencil is even
+ *  int stencil[10] = {0, 3, 4, 1, 4, 1, 2, 7, 8, 9};
+ *  thrust::device_vector<int> d_stencil(stencil, stencil + 10);
+ *
+ *  // map all even indices into the first half of the range
+ *  // and odd indices to the last half of the range
+ *  int map[10]   = {0, 2, 4, 6, 8, 1, 3, 5, 7, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10, 7);
+ *  thrust::gather_if(d_map.begin(), d_map.end(),
+ *                    d_stencil.begin(),
+ *                    d_values.begin(),
+ *                    d_output.begin(),
+ *                    is_even());
+ *  // d_output is now {0, 7, 4, 7, 8, 7, 3, 7, 7, 7}
+ *  \endcode
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator gather_if(InputIterator1       map_first,
+                           InputIterator1       map_last,
+                           InputIterator2       stencil,
+                           RandomAccessIterator input_first,
+                           OutputIterator       result,
+                           Predicate            pred);
+
+/*! \} // gathering
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/gather.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/generate.h b/thirdparty/manifold/thirdparty/thrust/thrust/generate.h
new file mode 100644
index 000000000000..d47295344607
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/generate.h
@@ -0,0 +1,211 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file generate.h
+ *  \brief Fills a range with values "generated" from a function of no arguments
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup transformations
+ *  \{
+ */
+
+
+/*! \p generate assigns the result of invoking \p gen, a function object that takes no arguments,
+ *  to each element in the range <tt>[first,last)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element in the range of interest.
+ *  \param last The last element in the range of interest.
+ *  \param gen A function argument, taking no parameters, used to generate values to assign to
+ *             elements in the range <tt>[first,last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam Generator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional">Generator</a>,
+ *          and \p Generator's \c result_type is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to fill a \c host_vector with random numbers,
+ *  using the standard C library function \c rand using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/generate.h>
+ *  #include <thrust/host_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdlib>
+ *  ...
+ *  thrust::host_vector<int> v(10);
+ *  srand(13);
+ *  thrust::generate(thrust::host, v.begin(), v.end(), rand);
+ *
+ *  // the elements of v are now pseudo-random numbers
+ *  \endcode
+ *
+ *  \see generate_n
+ *  \see https://en.cppreference.com/w/cpp/algorithm/generate
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Generator>
+__host__ __device__
+  void generate(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                Generator gen);
+
+
+/*! \p generate assigns the result of invoking \p gen, a function object that takes no arguments,
+ *  to each element in the range <tt>[first,last)</tt>.
+ *
+ *  \param first The first element in the range of interest.
+ *  \param last The last element in the range of interest.
+ *  \param gen A function argument, taking no parameters, used to generate values to assign to
+ *             elements in the range <tt>[first,last)</tt>.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam Generator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional">Generator</a>,
+ *          and \p Generator's \c result_type is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to fill a \c host_vector with random numbers,
+ *  using the standard C library function \c rand.
+ *
+ *  \code
+ *  #include <thrust/generate.h>
+ *  #include <thrust/host_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdlib>
+ *  ...
+ *  thrust::host_vector<int> v(10);
+ *  srand(13);
+ *  thrust::generate(v.begin(), v.end(), rand);
+ *
+ *  // the elements of v are now pseudo-random numbers
+ *  \endcode
+ *
+ *  \see generate_n
+ *  \see https://en.cppreference.com/w/cpp/algorithm/generate
+ */
+template<typename ForwardIterator,
+         typename Generator>
+  void generate(ForwardIterator first,
+                ForwardIterator last,
+                Generator gen);
+
+
+/*! \p generate_n assigns the result of invoking \p gen, a function object that takes no arguments,
+ *  to each element in the range <tt>[first,first + n)</tt>. The return value is <tt>first + n</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element in the range of interest.
+ *  \param n The size of the range of interest.
+ *  \param gen A function argument, taking no parameters, used to generate values to assign to
+ *             elements in the range <tt>[first,first + n)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Size is an integral type (either signed or unsigned).
+ *  \tparam Generator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional">Generator</a>,
+ *          and \p Generator's \c result_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *
+ *  The following code snippet demonstrates how to fill a \c host_vector with random numbers,
+ *  using the standard C library function \c rand using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/generate.h>
+ *  #include <thrust/host_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  #include <cstdlib>
+ *  ...
+ *  thrust::host_vector<int> v(10);
+ *  srand(13);
+ *  thrust::generate_n(thrust::host, v.begin(), 10, rand);
+ *
+ *  // the elements of v are now pseudo-random numbers
+ *  \endcode
+ *
+ *  \see generate
+ *  \see https://en.cppreference.com/w/cpp/algorithm/generate
+ */
+template<typename DerivedPolicy,
+         typename OutputIterator,
+         typename Size,
+         typename Generator>
+__host__ __device__
+  OutputIterator generate_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            OutputIterator first,
+                            Size n,
+                            Generator gen);
+
+
+/*! \p generate_n assigns the result of invoking \p gen, a function object that takes no arguments,
+ *  to each element in the range <tt>[first,first + n)</tt>. The return value is <tt>first + n</tt>.
+ *
+ *  \param first The first element in the range of interest.
+ *  \param n The size of the range of interest.
+ *  \param gen A function argument, taking no parameters, used to generate values to assign to
+ *             elements in the range <tt>[first,first + n)</tt>.
+ *
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Size is an integral type (either signed or unsigned).
+ *  \tparam Generator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional">Generator</a>,
+ *          and \p Generator's \c result_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *
+ *  The following code snippet demonstrates how to fill a \c host_vector with random numbers,
+ *  using the standard C library function \c rand.
+ *
+ *  \code
+ *  #include <thrust/generate.h>
+ *  #include <thrust/host_vector.h>
+ *  #include <stdlib.h>
+ *  ...
+ *  thrust::host_vector<int> v(10);
+ *  srand(13);
+ *  thrust::generate_n(v.begin(), 10, rand);
+ *
+ *  // the elements of v are now pseudo-random numbers
+ *  \endcode
+ *
+ *  \see generate
+ *  \see https://en.cppreference.com/w/cpp/algorithm/generate
+ */
+template<typename OutputIterator,
+         typename Size,
+         typename Generator>
+  OutputIterator generate_n(OutputIterator first,
+                            Size n,
+                            Generator gen);
+
+
+/*! \} // end transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/generate.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/host_vector.h b/thirdparty/manifold/thirdparty/thrust/thrust/host_vector.h
new file mode 100644
index 000000000000..bb925ea9ce85
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/host_vector.h
@@ -0,0 +1,534 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file host_vector.h
+ *  \brief A dynamically-sizable array of elements which resides in memory
+ *         accessible to hosts.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/memory_wrapper.h>
+#include <thrust/detail/vector_base.h>
+
+#include <initializer_list>
+#include <vector>
+#include <utility>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup container_classes Container Classes
+ *  \addtogroup host_containers Host Containers
+ *  \ingroup container_classes
+ *  \{
+ */
+
+/*! A \p host_vector is a container that supports random access to elements,
+ *  constant time removal of elements at the end, and linear time insertion
+ *  and removal of elements at the beginning or in the middle. The number of
+ *  elements in a \p host_vector may vary dynamically; memory management is
+ *  automatic. The memory associated with a \p host_vector resides in memory
+ *  accessible to hosts.
+ *
+ *  \see https://en.cppreference.com/w/cpp/container/vector
+ *  \see device_vector
+ *  \see universal_vector
+ */
+template<typename T, typename Alloc = std::allocator<T> >
+  class host_vector
+    : public detail::vector_base<T,Alloc>
+{
+  private:
+    typedef detail::vector_base<T,Alloc> Parent;
+
+  public:
+    /*! \cond
+     */
+    typedef typename Parent::size_type  size_type;
+    typedef typename Parent::value_type value_type;
+    /*! \endcond
+     */
+
+    /*! This constructor creates an empty \p host_vector.
+     */
+    __host__
+    host_vector(void)
+      :Parent() {}
+
+    /*! This constructor creates an empty \p host_vector.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+    __host__
+    host_vector(const Alloc &alloc)
+      :Parent(alloc) {}
+
+    /*! The destructor erases the elements.
+     */
+    //  Define an empty destructor to explicitly specify
+    //  its execution space qualifier, as a workaround for nvcc warning
+    __host__
+    ~host_vector(void) {}
+
+    /*! This constructor creates a \p host_vector with the given
+     *  size.
+     *  \param n The number of elements to initially create.
+     */
+    __host__
+    explicit host_vector(size_type n)
+      :Parent(n) {}
+
+    /*! This constructor creates a \p host_vector with the given
+     *  size.
+     *  \param n The number of elements to initially create.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+    __host__
+    explicit host_vector(size_type n, const Alloc &alloc)
+      :Parent(n,alloc) {}
+
+    /*! This constructor creates a \p host_vector with copies
+     *  of an exemplar element.
+     *  \param n The number of elements to initially create.
+     *  \param value An element to copy.
+     */
+    __host__
+    explicit host_vector(size_type n, const value_type &value)
+      :Parent(n,value) {}
+
+    /*! This constructor creates a \p host_vector with copies
+     *  of an exemplar element.
+     *  \param n The number of elements to initially create.
+     *  \param value An element to copy.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+    __host__
+    explicit host_vector(size_type n, const value_type &value, const Alloc &alloc)
+      :Parent(n,value,alloc) {}
+
+    /*! Copy constructor copies from an exemplar \p host_vector.
+     *  \param v The \p host_vector to copy.
+     */
+    __host__
+    host_vector(const host_vector &v)
+      :Parent(v) {}
+
+    /*! Copy constructor copies from an exemplar \p host_vector.
+     *  \param v The \p host_vector to copy.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+    __host__
+    host_vector(const host_vector &v, const Alloc &alloc)
+      :Parent(v,alloc) {}
+
+  #if THRUST_CPP_DIALECT >= 2011
+    /*! Move constructor moves from another host_vector.
+     *  \param v The host_vector to move.
+     */
+     __host__
+    host_vector(host_vector &&v)
+      :Parent(std::move(v)) {}
+
+    /*! Move constructor moves from another host_vector.
+     *  \param v The host_vector to move.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+     __host__
+    host_vector(host_vector &&v, const Alloc &alloc)
+      :Parent(std::move(v),alloc) {}
+  #endif
+
+  /*! Assign operator copies from an exemplar \p host_vector.
+   *  \param v The \p host_vector to copy.
+   */
+  __host__
+  host_vector &operator=(const host_vector &v)
+  { Parent::operator=(v); return *this; }
+
+  #if THRUST_CPP_DIALECT >= 2011
+    /*! Move assign operator moves from another host_vector.
+     *  \param v The host_vector to move.
+     */
+     __host__
+     host_vector &operator=(host_vector &&v)
+     { Parent::operator=(std::move(v)); return *this; }
+  #endif
+
+    /*! Copy constructor copies from an exemplar \p host_vector with different type.
+     *  \param v The \p host_vector to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    __host__
+    host_vector(const host_vector<OtherT,OtherAlloc> &v)
+      :Parent(v) {}
+
+    /*! Assign operator copies from an exemplar \p host_vector with different type.
+     *  \param v The \p host_vector to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    __host__
+    host_vector &operator=(const host_vector<OtherT,OtherAlloc> &v)
+    { Parent::operator=(v); return *this; }
+
+    /*! Copy constructor copies from an exemplar <tt>std::vector</tt>.
+     *  \param v The <tt>std::vector</tt> to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    __host__
+    host_vector(const std::vector<OtherT,OtherAlloc> &v)
+      :Parent(v) {}
+
+    /*! Assign operator copies from an exemplar <tt>std::vector</tt>.
+     *  \param v The <tt>std::vector</tt> to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    __host__
+    host_vector &operator=(const std::vector<OtherT,OtherAlloc> &v)
+    { Parent::operator=(v); return *this;}
+
+    /*! Copy construct from a \p vector_base whose element type is convertible
+     *  to \c T.
+     *
+     *  \param v The \p vector_base to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    __host__
+    host_vector(const detail::vector_base<OtherT,OtherAlloc> &v)
+      :Parent(v) {}
+
+    /*! Assign a \p vector_base whose element type is convertible to \c T.
+     *
+     *  \param v The \p vector_base to copy.
+     */
+    template<typename OtherT, typename OtherAlloc>
+    __host__
+    host_vector &operator=(const detail::vector_base<OtherT,OtherAlloc> &v)
+    { Parent::operator=(v); return *this; }
+    
+    /*! This constructor builds a \p host_vector from an intializer_list.
+     *  \param il The intializer_list.
+     */
+    host_vector(std::initializer_list<T> il)
+      :Parent(il) {}
+      
+    /*! This constructor builds a \p host_vector from an intializer_list.
+     *  \param il The intializer_list.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+    host_vector(std::initializer_list<T> il, const Alloc &alloc)
+      :Parent(il, alloc) {}
+      
+    /*! Assign an \p intializer_list with a matching element type
+     *  \param il The intializer_list.
+     */
+    host_vector &operator=(std::initializer_list<T> il)
+    { Parent::operator=(il); return *this; }
+
+    /*! This constructor builds a \p host_vector from a range.
+     *  \param first The beginning of the range.
+     *  \param last The end of the range.
+     */
+    template<typename InputIterator>
+    __host__
+    host_vector(InputIterator first, InputIterator last)
+      :Parent(first, last) {}
+
+    /*! This constructor builds a \p host_vector from a range.
+     *  \param first The beginning of the range.
+     *  \param last The end of the range.
+     *  \param alloc The allocator to use by this host_vector.
+     */
+    template<typename InputIterator>
+    __host__
+    host_vector(InputIterator first, InputIterator last, const Alloc &alloc)
+      :Parent(first, last, alloc) {}
+
+// declare these members for the purpose of Doxygenating them
+// they actually exist in a derived-from class
+#if 0
+    /*! \brief Resizes this vector to the specified number of elements.
+     *  \param new_size Number of elements this vector should contain.
+     *  \param x Data with which new elements should be populated.
+     *  \throw std::length_error If n exceeds max_size().
+     *
+     *  This method will resize this vector to the specified number of
+     *  elements.  If the number is smaller than this vector's current
+     *  size this vector is truncated, otherwise this vector is
+     *  extended and new elements are populated with given data.
+     */
+    void resize(size_type new_size, const value_type &x = value_type());
+
+    /*! Returns the number of elements in this vector.
+     */
+    size_type size(void) const;
+
+    /*! Returns the size() of the largest possible vector.
+     *  \return The largest possible return value of size().
+     */
+    size_type max_size(void) const;
+
+    /*! \brief If n is less than or equal to capacity(), this call has no effect.
+     *         Otherwise, this method is a request for allocation of additional memory. If
+     *         the request is successful, then capacity() is greater than or equal to
+     *         n; otherwise, capacity() is unchanged. In either case, size() is unchanged.
+     *  \throw std::length_error If n exceeds max_size().
+     */
+    void reserve(size_type n);
+
+    /*! Returns the number of elements which have been reserved in this
+     *  vector.
+     */
+    size_type capacity(void) const;
+
+    /*! This method shrinks the capacity of this vector to exactly
+     *  fit its elements.
+     */
+    void shrink_to_fit(void);
+
+    /*! \brief Subscript access to the data contained in this vector_dev.
+     *  \param n The index of the element for which data should be accessed.
+     *  \return Read/write reference to data.
+     *
+     *  This operator allows for easy, array-style, data access.
+     *  Note that data access with this operator is unchecked and
+     *  out_of_range lookups are not defined.
+     */
+    reference operator[](size_type n);
+
+    /*! \brief Subscript read access to the data contained in this vector_dev.
+     *  \param n The index of the element for which data should be accessed.
+     *  \return Read reference to data.
+     *
+     *  This operator allows for easy, array-style, data access.
+     *  Note that data access with this operator is unchecked and
+     *  out_of_range lookups are not defined.
+     */
+    const_reference operator[](size_type n) const;
+
+    /*! This method returns an iterator pointing to the beginning of
+     *  this vector.
+     *  \return mStart
+     */
+    iterator begin(void);
+
+    /*! This method returns a const_iterator pointing to the beginning
+     *  of this vector.
+     *  \return mStart
+     */
+    const_iterator begin(void) const;
+
+    /*! This method returns a const_iterator pointing to the beginning
+     *  of this vector.
+     *  \return mStart
+     */
+    const_iterator cbegin(void) const;
+
+    /*! This method returns a reverse_iterator pointing to the beginning of
+     *  this vector's reversed sequence.
+     *  \return A reverse_iterator pointing to the beginning of this
+     *          vector's reversed sequence.
+     */
+    reverse_iterator rbegin(void);
+
+    /*! This method returns a const_reverse_iterator pointing to the beginning of
+     *  this vector's reversed sequence.
+     *  \return A const_reverse_iterator pointing to the beginning of this
+     *          vector's reversed sequence.
+     */
+    const_reverse_iterator rbegin(void) const;
+
+    /*! This method returns a const_reverse_iterator pointing to the beginning of
+     *  this vector's reversed sequence.
+     *  \return A const_reverse_iterator pointing to the beginning of this
+     *          vector's reversed sequence.
+     */
+    const_reverse_iterator crbegin(void) const;
+
+    /*! This method returns an iterator pointing to one element past the
+     *  last of this vector.
+     *  \return begin() + size().
+     */
+    iterator end(void);
+
+    /*! This method returns a const_iterator pointing to one element past the
+     *  last of this vector.
+     *  \return begin() + size().
+     */
+    const_iterator end(void) const;
+
+    /*! This method returns a const_iterator pointing to one element past the
+     *  last of this vector.
+     *  \return begin() + size().
+     */
+    const_iterator cend(void) const;
+
+    /*! This method returns a reverse_iterator pointing to one element past the
+     *  last of this vector's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    reverse_iterator rend(void);
+
+    /*! This method returns a const_reverse_iterator pointing to one element past the
+     *  last of this vector's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    const_reverse_iterator rend(void) const;
+
+    /*! This method returns a const_reverse_iterator pointing to one element past the
+     *  last of this vector's reversed sequence.
+     *  \return rbegin() + size().
+     */
+    const_reverse_iterator crend(void) const;
+
+    /*! This method returns a const_reference referring to the first element of this
+     *  vector.
+     *  \return The first element of this vector.
+     */
+    const_reference front(void) const;
+
+    /*! This method returns a reference pointing to the first element of this
+     *  vector.
+     *  \return The first element of this vector.
+     */
+    reference front(void);
+
+    /*! This method returns a const reference pointing to the last element of
+     *  this vector.
+     *  \return The last element of this vector.
+     */
+    const_reference back(void) const;
+
+    /*! This method returns a reference referring to the last element of
+     *  this vector_dev.
+     *  \return The last element of this vector.
+     */
+    reference back(void);
+
+    /*! This method returns a pointer to this vector's first element.
+     *  \return A pointer to the first element of this vector.
+     */
+    pointer data(void);
+
+    /*! This method returns a const_pointer to this vector's first element.
+     *  \return a const_pointer to the first element of this vector.
+     */
+    const_pointer data(void) const;
+
+    /*! This method resizes this vector to 0.
+     */
+    void clear(void);
+
+    /*! This method returns true iff size() == 0.
+     *  \return true if size() == 0; false, otherwise.
+     */
+    bool empty(void) const;
+
+    /*! This method appends the given element to the end of this vector.
+     *  \param x The element to append.
+     */
+    void push_back(const value_type &x);
+
+    /*! This method erases the last element of this vector, invalidating
+     *  all iterators and references to it.
+     */
+    void pop_back(void);
+
+    /*! This method swaps the contents of this host_vector with another vector.
+     *  \param v The vector with which to swap.
+     */
+    void swap(host_vector &v);
+
+    /*! This method removes the element at position pos.
+     *  \param pos The position of the element of interest.
+     *  \return An iterator pointing to the new location of the element that followed the element
+     *          at position pos.
+     */
+    iterator erase(iterator pos);
+
+    /*! This method removes the range of elements [first,last) from this vector.
+     *  \param first The beginning of the range of elements to remove.
+     *  \param last The end of the range of elements to remove.
+     *  \return An iterator pointing to the new location of the element that followed the last
+     *          element in the sequence [first,last).
+     */
+    iterator erase(iterator first, iterator last);
+
+    /*! This method inserts a single copy of a given exemplar value at the
+     *  specified position in this vector.
+     *  \param position The insertion position.
+     *  \param x The exemplar element to copy & insert.
+     *  \return An iterator pointing to the newly inserted element.
+     */
+    iterator insert(iterator position, const T &x);
+
+    /*! This method inserts a copy of an exemplar value to a range at the
+     *  specified position in this vector.
+     *  \param position The insertion position
+     *  \param n The number of insertions to perform.
+     *  \param x The value to replicate and insert.
+     */
+    void insert(iterator position, size_type n, const T &x);
+
+    /*! This method inserts a copy of an input range at the specified position
+     *  in this vector.
+     *  \param position The insertion position.
+     *  \param first The beginning of the range to copy.
+     *  \param last  The end of the range to copy.
+     *
+     *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator>Input Iterator</a>,
+     *                        and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>.
+     */
+    template<typename InputIterator>
+    void insert(iterator position, InputIterator first, InputIterator last);
+
+    /*! This version of \p assign replicates a given exemplar
+     *  \p n times into this vector.
+     *  \param n The number of times to copy \p x.
+     *  \param x The exemplar element to replicate.
+     */
+    void assign(size_type n, const T &x);
+
+    /*! This version of \p assign makes this vector a copy of a given input range.
+     *  \param first The beginning of the range to copy.
+     *  \param last  The end of the range to copy.
+     *
+     *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/named_req/InputIterator">Input Iterator</a>.
+     */
+    template<typename InputIterator>
+    void assign(InputIterator first, InputIterator last);
+
+    /*! This method returns a copy of this vector's allocator.
+     *  \return A copy of the alloctor used by this vector.
+     */
+    allocator_type get_allocator(void) const;
+#endif // end doxygen-only members
+};
+
+/*! Exchanges the values of two vectors.
+ *  \p x The first \p host_vector of interest.
+ *  \p y The second \p host_vector of interest.
+ */
+template<typename T, typename Alloc>
+  void swap(host_vector<T,Alloc> &a, host_vector<T,Alloc> &b)
+{
+  a.swap(b);
+}
+
+/*! \}
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/inner_product.h b/thirdparty/manifold/thirdparty/thrust/thrust/inner_product.h
new file mode 100644
index 000000000000..80068cf0c834
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/inner_product.h
@@ -0,0 +1,262 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file inner_product.h
+ *  \brief Mathematical inner product between ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup transformed_reductions Transformed Reductions
+ *  \ingroup reductions
+ *  \{
+ */
+
+
+/*! \p inner_product calculates an inner product of the ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, first2 + (last1 - first1))</tt>.
+ *
+ *  Specifically, this version of \p inner_product computes the sum
+ *  <tt>init + (*first1 * *first2) + (*(first1+1) * *(first2+1)) + ... </tt>
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1 The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param init Initial value of the result.
+ *  \return The inner product of sequences <tt>[first1, last1)</tt>
+ *          and <tt>[first2, last2)</tt> plus \p init.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputType is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x is an object of type \p OutputType, and \c y is an object of \p InputIterator1's \c value_type,
+ *          and \c z is an object of \p InputIterator2's \c value_type, then <tt>x + y * z</tt> is defined
+ *          and is convertible to \p OutputType.
+ *
+ *  The following code demonstrates how to use \p inner_product to
+ *  compute the dot product of two vectors using the \p thrust::host execution policy for parallelization.
+ *
+ *  \code
+ *  #include <thrust/inner_product.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  float vec1[3] = {1.0f, 2.0f, 5.0f};
+ *  float vec2[3] = {4.0f, 1.0f, 5.0f};
+ *
+ *  float result = thrust::inner_product(thrust::host, vec1, vec1 + 3, vec2, 0.0f);
+ *
+ *  // result == 31.0f
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/inner_product
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputType>
+__host__ __device__
+OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init);
+
+
+/*! \p inner_product calculates an inner product of the ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, first2 + (last1 - first1))</tt>.
+ *
+ *  Specifically, this version of \p inner_product computes the sum
+ *  <tt>init + (*first1 * *first2) + (*(first1+1) * *(first2+1)) + ... </tt>
+ *
+ *  Unlike the C++ Standard Template Library function <tt>std::inner_product</tt>,
+ *  this version offers no guarantee on order of execution.
+ *
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1 The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param init Initial value of the result.
+ *  \return The inner product of sequences <tt>[first1, last1)</tt>
+ *          and <tt>[first2, last2)</tt> plus \p init.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputType is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and if \c x is an object of type \p OutputType, and \c y is an object of \p InputIterator1's \c value_type,
+ *          and \c z is an object of \p InputIterator2's \c value_type, then <tt>x + y * z</tt> is defined
+ *          and is convertible to \p OutputType.
+ *
+ *  The following code demonstrates how to use \p inner_product to
+ *  compute the dot product of two vectors.
+ *
+ *  \code
+ *  #include <thrust/inner_product.h>
+ *  ...
+ *  float vec1[3] = {1.0f, 2.0f, 5.0f};
+ *  float vec2[3] = {4.0f, 1.0f, 5.0f};
+ *
+ *  float result = thrust::inner_product(vec1, vec1 + 3, vec2, 0.0f);
+ *
+ *  // result == 31.0f
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/inner_product
+ */
+template<typename InputIterator1, typename InputIterator2, typename OutputType>
+OutputType inner_product(InputIterator1 first1, InputIterator1 last1,
+                         InputIterator2 first2, OutputType init);
+
+
+/*! \p inner_product calculates an inner product of the ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, first2 + (last1 - first1))</tt>.
+ *
+ *  This version of \p inner_product is identical to the first, except that is uses
+ *  two user-supplied function objects instead of \c operator+ and \c operator*.
+ *
+ *  Specifically, this version of \p inner_product computes the sum
+ *  <tt>binary_op1( init, binary_op2(*first1, *first2) ), ... </tt>
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1 The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param init Initial value of the result.
+ *  \param binary_op1 Generalized addition operation.
+ *  \param binary_op2 Generalized multiplication operation.
+ *  \return The inner product of sequences <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator1's \c value_type is convertible to \p BinaryFunction2's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *          and \p InputIterator2's \c value_type is convertible to \p BinaryFunction2's \c second_argument_type.
+ *  \tparam OutputType is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p OutputType is convertible to \p BinaryFunction1's \c first_argument_type.
+ *  \tparam BinaryFunction1 is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction1's \c return_type is convertible to \p OutputType.
+ *  \tparam BinaryFunction2 is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction2's \c return_type is convertible to \p BinaryFunction1's \c second_argument_type.
+ * 
+ *  \code
+ *  #include <thrust/inner_product.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  float vec1[3] = {1.0f, 2.0f, 5.0f};
+ *  float vec2[3] = {4.0f, 1.0f, 5.0f};
+ *
+ *  float init = 0.0f;
+ *  thrust::plus<float>       binary_op1;
+ *  thrust::multiplies<float> binary_op2;
+ *
+ *  float result = thrust::inner_product(thrust::host, vec1, vec1 + 3, vec2, init, binary_op1, binary_op2);
+ *
+ *  // result == 31.0f
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/inner_product
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputType,
+         typename BinaryFunction1,
+         typename BinaryFunction2>
+__host__ __device__
+OutputType inner_product(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init, 
+                         BinaryFunction1 binary_op1,
+                         BinaryFunction2 binary_op2);
+
+
+/*! \p inner_product calculates an inner product of the ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, first2 + (last1 - first1))</tt>.
+ *
+ *  This version of \p inner_product is identical to the first, except that is uses
+ *  two user-supplied function objects instead of \c operator+ and \c operator*.
+ *
+ *  Specifically, this version of \p inner_product computes the sum
+ *  <tt>binary_op1( init, binary_op2(*first1, *first2) ), ... </tt>
+ *
+ *  Unlike the C++ Standard Template Library function <tt>std::inner_product</tt>,
+ *  this version offers no guarantee on order of execution.
+ *
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1 The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param init Initial value of the result.
+ *  \param binary_op1 Generalized addition operation.
+ *  \param binary_op2 Generalized multiplication operation.
+ *  \return The inner product of sequences <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator1's \c value_type is convertible to \p BinaryFunction2's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *          and \p InputIterator2's \c value_type is convertible to \p BinaryFunction2's \c second_argument_type.
+ *  \tparam OutputType is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p OutputType is convertible to \p BinaryFunction1's \c first_argument_type.
+ *  \tparam BinaryFunction1 is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction1's \c return_type is convertible to \p OutputType.
+ *  \tparam BinaryFunction2 is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction2's \c return_type is convertible to \p BinaryFunction1's \c second_argument_type.
+ * 
+ *  \code
+ *  #include <thrust/inner_product.h>
+ *  ...
+ *  float vec1[3] = {1.0f, 2.0f, 5.0f};
+ *  float vec2[3] = {4.0f, 1.0f, 5.0f};
+ *
+ *  float init = 0.0f;
+ *  thrust::plus<float>       binary_op1;
+ *  thrust::multiplies<float> binary_op2;
+ *
+ *  float result = thrust::inner_product(vec1, vec1 + 3, vec2, init, binary_op1, binary_op2);
+ *
+ *  // result == 31.0f
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/inner_product
+ */
+template<typename InputIterator1, typename InputIterator2, typename OutputType,
+         typename BinaryFunction1, typename BinaryFunction2>
+OutputType inner_product(InputIterator1 first1, InputIterator1 last1,
+                         InputIterator2 first2, OutputType init, 
+                         BinaryFunction1 binary_op1, BinaryFunction2 binary_op2);
+
+
+/*! \} // end transformed_reductions
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/inner_product.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/constant_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/constant_iterator.h
new file mode 100644
index 000000000000..c6eec28e70e4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/constant_iterator.h
@@ -0,0 +1,250 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/constant_iterator.h
+ *  \brief An iterator which returns a constant value when
+ *         dereferenced
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/constant_iterator_base.h>
+#include <thrust/iterator/iterator_facade.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p constant_iterator is an iterator which represents a pointer into a range
+ *  of constant values. This iterator is useful for creating a range filled with the same
+ *  value without explicitly storing it in memory. Using \p constant_iterator saves both
+ *  memory capacity and bandwidth.
+ *
+ *  The following code snippet demonstrates how to create a \p constant_iterator whose
+ *  \c value_type is \c int and whose value is \c 10.
+ *
+ *  \code
+ *  #include <thrust/iterator/constant_iterator.h>
+ *
+ *  thrust::constant_iterator<int> iter(10);
+ *
+ *  *iter;    // returns 10
+ *  iter[0];  // returns 10
+ *  iter[1];  // returns 10
+ *  iter[13]; // returns 10
+ *
+ *  // and so on...
+ *  \endcode
+ *
+ *  This next example demonstrates how to use a \p constant_iterator with the
+ *  \p thrust::transform function to increment all elements of a sequence by the
+ *  same value. We will create a temporary \p constant_iterator with the function
+ *  \p make_constant_iterator function in order to avoid explicitly specifying
+ *  its type:
+ *
+ *  \code
+ *  #include <thrust/iterator/constant_iterator.h>
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  int main()
+ *  {
+ *    thrust::device_vector<int> data(4);
+ *    data[0] = 3;
+ *    data[1] = 7;
+ *    data[2] = 2;
+ *    data[3] = 5;
+ *    
+ *    // add 10 to all values in data
+ *    thrust::transform(data.begin(), data.end(),
+ *                      thrust::make_constant_iterator(10),
+ *                      data.begin(),
+ *                      thrust::plus<int>());
+ *    
+ *    // data is now [13, 17, 12, 15]
+ *    
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see make_constant_iterator
+ */
+template<typename Value,
+         typename Incrementable = use_default,
+         typename System = use_default>
+  class constant_iterator
+    : public detail::constant_iterator_base<Value, Incrementable, System>::type
+{
+    /*! \cond
+     */
+    friend class thrust::iterator_core_access;
+    typedef typename detail::constant_iterator_base<Value, Incrementable, System>::type          super_t;
+    typedef typename detail::constant_iterator_base<Value, Incrementable, System>::incrementable incrementable;
+    typedef typename detail::constant_iterator_base<Value, Incrementable, System>::base_iterator base_iterator;
+
+  public:
+    typedef typename super_t::reference  reference;
+    typedef typename super_t::value_type value_type;
+
+    /*! \endcond
+     */
+
+    /*! Null constructor initializes this \p constant_iterator's constant using its
+     *  null constructor.
+     */
+    __host__ __device__
+    constant_iterator()
+      : super_t(), m_value() {}
+
+    /*! Copy constructor copies the value of another \p constant_iterator into this
+     *  \p constant_iterator.
+     *
+     *  \p rhs The constant_iterator to copy.
+     */
+    __host__ __device__
+    constant_iterator(constant_iterator const &rhs)
+      : super_t(rhs.base()), m_value(rhs.m_value) {}
+
+    /*! Copy constructor copies the value of another \p constant_iterator with related
+     *  System type.
+     *
+     *  \param rhs The \p constant_iterator to copy.
+     */
+    template<typename OtherSystem>
+    __host__ __device__
+    constant_iterator(constant_iterator<Value,Incrementable,OtherSystem> const &rhs,
+                      typename thrust::detail::enable_if_convertible<
+                        typename thrust::iterator_system<constant_iterator<Value,Incrementable,OtherSystem> >::type,
+                        typename thrust::iterator_system<super_t>::type
+                      >::type * = 0)
+      : super_t(rhs.base()), m_value(rhs.value()) {}
+
+    /*! This constructor receives a value to use as the constant value of this
+     *  \p constant_iterator and an index specifying the location of this
+     *  \p constant_iterator in a sequence.
+     *  
+     *  \p v The value of this \p constant_iterator's constant value.
+     *  \p i The index of this \p constant_iterator in a sequence. Defaults to the
+     *       value returned by \c Incrementable's null constructor. For example,
+     *       when <tt>Incrementable == int</tt>, \c 0.
+     */
+    __host__ __device__
+    constant_iterator(value_type const& v, incrementable const &i = incrementable())
+      : super_t(base_iterator(i)), m_value(v) {}
+
+    /*! This constructor is templated to allow construction from a value type and
+     *  incrementable type related this this \p constant_iterator's respective types.
+     *
+     *  \p v The value of this \p constant_iterator's constant value.
+     *  \p i The index of this \p constant_iterator in a sequence. Defaults to the
+     *       value returned by \c Incrementable's null constructor. For example,
+     *       when <tt>Incrementable == int</tt>, \c 0.
+     */
+    template<typename OtherValue, typename OtherIncrementable>
+    __host__ __device__
+    constant_iterator(OtherValue const& v, OtherIncrementable const& i = incrementable())
+      : super_t(base_iterator(i)), m_value(v) {}
+
+    /*! This method returns the value of this \p constant_iterator's constant value.
+     *  \return A \c const reference to this \p constant_iterator's constant value.
+     */
+    __host__ __device__
+    Value const& value() const
+    { return m_value; }
+
+    /*! \cond
+     */
+
+  protected:
+    __host__ __device__
+    Value const& value_reference() const
+    { return m_value; }
+
+    __host__ __device__
+    Value & value_reference()
+    { return m_value; }
+  
+  private: // Core iterator interface
+    __host__ __device__
+    reference dereference() const
+    {
+      return m_value;
+    }
+
+  private:
+    Value m_value;
+
+    /*! \endcond
+     */
+}; // end constant_iterator
+
+
+/*! This version of \p make_constant_iterator creates a \p constant_iterator
+ *  from values given for both value and index. The type of \p constant_iterator
+ *  may be inferred by the compiler from the types of its parameters.
+ *
+ *  \param x The value of the returned \p constant_iterator's constant value.
+ *  \param i The index of the returned \p constant_iterator within a sequence.
+ *           The type of this parameter defaults to \c int. In the default case,
+ *           the value of this parameter is \c 0.
+ *
+ *  \return A new \p constant_iterator with constant value & index as given
+ *          by \p x & \p i.
+ *
+ *  \see constant_iterator
+ */
+template<typename ValueT, typename IndexT>
+inline __host__ __device__
+constant_iterator<ValueT, IndexT> make_constant_iterator(ValueT x, IndexT i = int())
+{
+  return constant_iterator<ValueT, IndexT>(x, i);
+} // end make_constant_iterator()
+
+
+/*! This version of \p make_constant_iterator creates a \p constant_iterator
+ *  using only a parameter for the desired constant value. The value of the
+ *  returned \p constant_iterator's index is set to \c 0.
+ *
+ *  \param x The value of the returned \p constant_iterator's constant value.
+ *  \return A new \p constant_iterator with constant value equal to \p x and
+ *          index equal to \c 0.
+ *  \see constant_iterator
+ */
+template<typename V>
+inline __host__ __device__
+constant_iterator<V> make_constant_iterator(V x)
+{
+  return constant_iterator<V>(x, 0);
+} // end make_constant_iterator()
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/counting_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/counting_iterator.h
new file mode 100644
index 000000000000..f66cb97ef82a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/counting_iterator.h
@@ -0,0 +1,246 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/counting_iterator.h
+ *  \brief An iterator which returns an increasing incrementable value
+ *         when dereferenced
+ */
+
+/*
+ * Copyright David Abrahams 2003.
+ *
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/iterator/iterator_categories.h>
+
+// #include the details first
+#include <thrust/iterator/detail/counting_iterator.inl>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p counting_iterator is an iterator which represents a pointer into a range
+ *  of sequentially changing values. This iterator is useful for creating a range
+ *  filled with a sequence without explicitly storing it in memory. Using
+ *  \p counting_iterator saves memory capacity and bandwidth.
+ *
+ *  The following code snippet demonstrates how to create a \p counting_iterator whose
+ *  \c value_type is \c int and which sequentially increments by \c 1.
+ *
+ *  \code
+ *  #include <thrust/iterator/counting_iterator.h>
+ *  ...
+ *  // create iterators
+ *  thrust::counting_iterator<int> first(10);
+ *  thrust::counting_iterator<int> last = first + 3;
+ *
+ *  first[0]   // returns 10
+ *  first[1]   // returns 11
+ *  first[100] // returns 110
+ *
+ *  // sum of [first, last)
+ *  thrust::reduce(first, last);   // returns 33 (i.e. 10 + 11 + 12)
+ *
+ *  // initialize vector to [0,1,2,..]
+ *  thrust::counting_iterator<int> iter(0);
+ *  thrust::device_vector<int> vec(500);
+ *  thrust::copy(iter, iter + vec.size(), vec.begin());
+ *  \endcode
+ *
+ *  This next example demonstrates how to use a \p counting_iterator with the
+ *  \p thrust::copy_if function to compute the indices of the non-zero elements
+ *  of a \p device_vector. In this example, we use the \p make_counting_iterator
+ *  function to avoid specifying the type of the \p counting_iterator.
+ *
+ *  \code
+ *  #include <thrust/iterator/counting_iterator.h>
+ *  #include <thrust/copy.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  int main()
+ *  {
+ *   // this example computes indices for all the nonzero values in a sequence
+ *
+ *   // sequence of zero and nonzero values
+ *   thrust::device_vector<int> stencil(8);
+ *   stencil[0] = 0;
+ *   stencil[1] = 1;
+ *   stencil[2] = 1;
+ *   stencil[3] = 0;
+ *   stencil[4] = 0;
+ *   stencil[5] = 1;
+ *   stencil[6] = 0;
+ *   stencil[7] = 1;
+ *
+ *   // storage for the nonzero indices
+ *   thrust::device_vector<int> indices(8);
+ *
+ *   // compute indices of nonzero elements
+ *   typedef thrust::device_vector<int>::iterator IndexIterator;
+ *
+ *   // use make_counting_iterator to define the sequence [0, 8)
+ *   IndexIterator indices_end = thrust::copy_if(thrust::make_counting_iterator(0),
+ *                                               thrust::make_counting_iterator(8),
+ *                                               stencil.begin(),
+ *                                               indices.begin(),
+ *                                               thrust::identity<int>());
+ *   // indices now contains [1,2,5,7]
+ *
+ *   return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see make_counting_iterator
+ */
+template<typename Incrementable,
+         typename System = use_default,
+         typename Traversal = use_default,
+         typename Difference = use_default>
+  class counting_iterator
+    : public detail::counting_iterator_base<Incrementable, System, Traversal, Difference>::type
+{
+    /*! \cond
+     */
+    typedef typename detail::counting_iterator_base<Incrementable, System, Traversal, Difference>::type super_t;
+
+    friend class thrust::iterator_core_access;
+
+  public:
+    typedef typename super_t::reference       reference;
+    typedef typename super_t::difference_type difference_type;
+
+    /*! \endcond
+     */
+
+    /*! Default constructor initializes this \p counting_iterator's counter to
+     * `Incrementable{}`.
+     */
+    __host__ __device__
+    counting_iterator() : super_t(Incrementable{}) {}
+
+    /*! Copy constructor copies the value of another \p counting_iterator into a
+     *  new \p counting_iterator.
+     *
+     *  \p rhs The \p counting_iterator to copy.
+     */
+    __host__ __device__
+    counting_iterator(counting_iterator const &rhs):super_t(rhs.base()){}
+
+    /*! Copy constructor copies the value of another counting_iterator
+     *  with related System type.
+     *
+     *  \param rhs The \p counting_iterator to copy.
+     */
+    template<typename OtherSystem>
+    __host__ __device__
+    counting_iterator(counting_iterator<Incrementable, OtherSystem, Traversal, Difference> const &rhs,
+                      typename thrust::detail::enable_if_convertible<
+                        typename thrust::iterator_system<counting_iterator<Incrementable,OtherSystem,Traversal,Difference> >::type,
+                        typename thrust::iterator_system<super_t>::type
+                      >::type * = 0)
+      : super_t(rhs.base()){}
+
+    /*! This \c explicit constructor copies the value of an \c Incrementable
+     *  into a new \p counting_iterator's \c Incrementable counter.
+     *
+     *  \param x The initial value of the new \p counting_iterator's \c Incrementable
+     *         counter.
+     */
+    __host__ __device__
+    explicit counting_iterator(Incrementable x):super_t(x){}
+
+#if THRUST_CPP_DIALECT >= 2011
+    counting_iterator & operator=(const counting_iterator &) = default;
+#endif
+
+    /*! \cond
+     */
+  private:
+    __host__ __device__
+    reference dereference() const
+    {
+      return this->base_reference();
+    }
+
+    // note that we implement equal specially for floating point counting_iterator
+    template <typename OtherIncrementable, typename OtherSystem, typename OtherTraversal, typename OtherDifference>
+    __host__ __device__
+    bool equal(counting_iterator<OtherIncrementable, OtherSystem, OtherTraversal, OtherDifference> const& y) const
+    {
+      typedef thrust::detail::counting_iterator_equal<difference_type,Incrementable,OtherIncrementable> e;
+      return e::equal(this->base(), y.base());
+    }
+
+    template <class OtherIncrementable>
+    __host__ __device__
+    difference_type
+    distance_to(counting_iterator<OtherIncrementable, System, Traversal, Difference> const& y) const
+    {
+      typedef typename
+      thrust::detail::eval_if<
+        thrust::detail::is_numeric<Incrementable>::value,
+        thrust::detail::identity_<thrust::detail::number_distance<difference_type, Incrementable, OtherIncrementable> >,
+        thrust::detail::identity_<thrust::detail::iterator_distance<difference_type, Incrementable, OtherIncrementable> >
+      >::type d;
+
+      return d::distance(this->base(), y.base());
+    }
+
+    /*! \endcond
+     */
+}; // end counting_iterator
+
+
+/*! \p make_counting_iterator creates a \p counting_iterator
+ *  using an initial value for its \c Incrementable counter.
+ *
+ *  \param x The initial value of the new \p counting_iterator's counter.
+ *  \return A new \p counting_iterator whose counter has been initialized to \p x.
+ */
+template <typename Incrementable>
+inline __host__ __device__
+counting_iterator<Incrementable> make_counting_iterator(Incrementable x)
+{
+  return counting_iterator<Incrementable>(x);
+}
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_assign.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_assign.h
new file mode 100644
index 000000000000..87192215c3a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_assign.h
@@ -0,0 +1,54 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+// a type which may be assigned any other type
+struct any_assign
+{
+  inline __host__ __device__ any_assign()
+  {}
+
+  template<typename T>
+  inline __host__ __device__ any_assign(T)
+  {}
+
+  template<typename T>
+  inline __host__ __device__
+  any_assign &operator=(T)
+  {
+    if(0)
+    {
+      // trick the compiler into silencing "warning: this expression has no effect"
+      int *x = 0;
+      *x = 13;
+    } // end if
+
+    return *this;
+  }
+};
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_system_tag.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_system_tag.h
new file mode 100644
index 000000000000..2c5ce6448be3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/any_system_tag.h
@@ -0,0 +1,33 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+struct any_system_tag
+  : thrust::execution_policy<any_system_tag>
+{
+  // allow any_system_tag to convert to any type at all
+  // XXX make this safer using enable_if<is_tag<T>> upon c++11
+  template<typename T> operator T () const {return T();}
+};
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/constant_iterator_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/constant_iterator_base.h
new file mode 100644
index 000000000000..56bb7a5d01f5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/constant_iterator_base.h
@@ -0,0 +1,71 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/iterator/iterator_adaptor.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declaration of constant_iterator
+template<typename,typename,typename> class constant_iterator;
+
+namespace detail
+{
+
+template<typename Value,
+         typename Incrementable,
+         typename System>
+  struct constant_iterator_base
+{
+  typedef Value              value_type;
+
+  // the reference type is the same as the value_type.
+  // we wish to avoid returning a reference to the internal state
+  // of the constant_iterator, which is prone to subtle bugs.
+  // consider the temporary iterator created in the expression
+  // *(iter + i)
+  typedef value_type         reference;
+
+  // the incrementable type is int unless otherwise specified
+  typedef typename thrust::detail::ia_dflt_help<
+    Incrementable,
+    thrust::detail::identity_<thrust::detail::intmax_t>
+  >::type incrementable;
+
+  typedef typename thrust::counting_iterator<
+    incrementable,
+    System,
+    thrust::random_access_traversal_tag
+  > base_iterator;
+
+  typedef typename thrust::iterator_adaptor<
+    constant_iterator<Value, Incrementable, System>,
+    base_iterator,
+    value_type, // XXX we may need to pass const value_type here as boost counting_iterator does
+    typename thrust::iterator_system<base_iterator>::type,
+    typename thrust::iterator_traversal<base_iterator>::type,
+    reference
+  > type;
+}; // end constant_iterator_base
+
+} // end detail
+  
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/counting_iterator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/counting_iterator.inl
new file mode 100644
index 000000000000..ee4a9df15565
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/counting_iterator.inl
@@ -0,0 +1,142 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/numeric_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <cstddef>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declaration of counting_iterator
+template <typename Incrementable, typename System, typename Traversal, typename Difference>
+  class counting_iterator;
+
+namespace detail
+{
+
+template <typename Incrementable, typename System, typename Traversal, typename Difference>
+  struct counting_iterator_base
+{
+  typedef typename thrust::detail::eval_if<
+    // use any_system_tag if we are given use_default
+    thrust::detail::is_same<System,use_default>::value,
+    thrust::detail::identity_<thrust::any_system_tag>,
+    thrust::detail::identity_<System>
+  >::type system;
+
+  typedef typename thrust::detail::ia_dflt_help<
+      Traversal,
+      thrust::detail::eval_if<
+          thrust::detail::is_numeric<Incrementable>::value,
+          thrust::detail::identity_<random_access_traversal_tag>,
+          thrust::iterator_traversal<Incrementable>
+      >
+  >::type traversal;
+
+  // unlike Boost, we explicitly use std::ptrdiff_t as the difference type
+  // for floating point counting_iterators
+  typedef typename thrust::detail::ia_dflt_help<
+    Difference,
+    thrust::detail::eval_if<
+      thrust::detail::is_numeric<Incrementable>::value,
+        thrust::detail::eval_if<
+          thrust::detail::is_integral<Incrementable>::value,
+          thrust::detail::numeric_difference<Incrementable>,
+          thrust::detail::identity_<std::ptrdiff_t>
+        >,
+      thrust::iterator_difference<Incrementable>
+    >
+  >::type difference;
+
+  // our implementation departs from Boost's in that counting_iterator::dereference
+  // returns a copy of its counter, rather than a reference to it. returning a reference
+  // to the internal state of an iterator causes subtle bugs (consider the temporary
+  // iterator created in the expression *(iter + i)) and has no compelling use case
+  typedef thrust::iterator_adaptor<
+    counting_iterator<Incrementable, System, Traversal, Difference>, // self
+    Incrementable,                                                  // Base
+    Incrementable,                                                  // XXX we may need to pass const here as Boost does
+    system,
+    traversal,
+    Incrementable,
+    difference
+  > type;
+}; // end counting_iterator_base
+
+
+template<typename Difference, typename Incrementable1, typename Incrementable2>
+  struct iterator_distance
+{
+  __host__ __device__
+  static Difference distance(Incrementable1 x, Incrementable2 y)
+  {
+    return y - x;
+  }
+};
+
+
+template<typename Difference, typename Incrementable1, typename Incrementable2>
+  struct number_distance
+{
+  __host__ __device__
+  static Difference distance(Incrementable1 x, Incrementable2 y)
+  {
+      return static_cast<Difference>(numeric_distance(x,y));
+  }
+};
+
+
+template<typename Difference, typename Incrementable1, typename Incrementable2, typename Enable = void>
+  struct counting_iterator_equal
+{
+  __host__ __device__
+  static bool equal(Incrementable1 x, Incrementable2 y)
+  {
+    return x == y;
+  }
+};
+
+
+// specialization for floating point equality
+template<typename Difference, typename Incrementable1, typename Incrementable2>
+  struct counting_iterator_equal<
+    Difference,
+    Incrementable1,
+    Incrementable2,
+    typename thrust::detail::enable_if<
+      thrust::detail::is_floating_point<Incrementable1>::value ||
+      thrust::detail::is_floating_point<Incrementable2>::value
+    >::type
+  >
+{
+  __host__ __device__
+  static bool equal(Incrementable1 x, Incrementable2 y)
+  {
+    typedef number_distance<Difference,Incrementable1,Incrementable2> d;
+    return d::distance(x,y) == 0;
+  }
+};
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/device_system_tag.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/device_system_tag.h
new file mode 100644
index 000000000000..b86109d216b7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/device_system_tag.h
@@ -0,0 +1,30 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include the device system's execution_policy header
+#define __THRUST_DEVICE_SYSTEM_TAG_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/execution_policy.h>
+#include __THRUST_DEVICE_SYSTEM_TAG_HEADER
+#undef __THRUST_DEVICE_SYSTEM_TAG_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+typedef thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::tag device_system_tag;
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/discard_iterator_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/discard_iterator_base.h
new file mode 100644
index 000000000000..38f77b378120
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/discard_iterator_base.h
@@ -0,0 +1,64 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/detail/any_assign.h>
+#include <cstddef> // for std::ptrdiff_t
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declaration of discard_iterator
+template<typename> class discard_iterator;
+
+namespace detail
+{
+
+
+template<typename System>
+  struct discard_iterator_base
+{
+  // XXX value_type should actually be void
+  //     but this interferes with zip_iterator<discard_iterator>
+  typedef any_assign         value_type;
+  typedef any_assign&        reference;
+  typedef std::ptrdiff_t     incrementable;
+
+  typedef typename thrust::counting_iterator<
+    incrementable,
+    System,
+    thrust::random_access_traversal_tag
+  > base_iterator;
+
+  typedef typename thrust::iterator_adaptor<
+    discard_iterator<System>,
+    base_iterator,
+    value_type,
+    typename thrust::iterator_system<base_iterator>::type,
+    typename thrust::iterator_traversal<base_iterator>::type,
+    reference
+  > type;
+}; // end discard_iterator_base
+
+
+} // end detail
+  
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/distance_from_result.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/distance_from_result.h
new file mode 100644
index 000000000000..fe140344d49c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/distance_from_result.h
@@ -0,0 +1,41 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// since both arguments are known to be specializations of iterator_facade,
+// it's legal to access IteratorFacade2::difference_type
+template<typename IteratorFacade1, typename IteratorFacade2>
+  struct distance_from_result
+    : eval_if<
+        is_convertible<IteratorFacade2,IteratorFacade1>::value,
+        identity_<typename IteratorFacade1::difference_type>,
+        identity_<typename IteratorFacade2::difference_type>
+      >
+{};
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/host_system_tag.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/host_system_tag.h
new file mode 100644
index 000000000000..58478f8d9266
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/host_system_tag.h
@@ -0,0 +1,30 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include the host system's execution_policy header
+#define __THRUST_HOST_SYSTEM_TAG_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/execution_policy.h>
+#include __THRUST_HOST_SYSTEM_TAG_HEADER
+#undef __THRUST_HOST_SYSTEM_TAG_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+typedef thrust::system::__THRUST_HOST_SYSTEM_NAMESPACE::tag host_system_tag;
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/is_iterator_category.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/is_iterator_category.h
new file mode 100644
index 000000000000..e520452a3ca8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/is_iterator_category.h
@@ -0,0 +1,59 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_categories.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+template <typename T>
+  struct is_host_iterator_category
+    : thrust::detail::or_<
+        thrust::detail::is_convertible<T, thrust::input_host_iterator_tag>,
+        thrust::detail::is_convertible<T, thrust::output_host_iterator_tag>
+      >
+{
+}; // end is_host_iterator_category
+
+template <typename T>
+  struct is_device_iterator_category
+    : thrust::detail::or_<
+        thrust::detail::is_convertible<T, thrust::input_device_iterator_tag>,
+        thrust::detail::is_convertible<T, thrust::output_device_iterator_tag>
+      >
+{
+}; // end is_device_iterator_category
+
+
+template <typename T>
+  struct is_iterator_category
+    : thrust::detail::or_<
+        is_host_iterator_category<T>,
+        is_device_iterator_category<T>
+      >
+{
+}; // end is_iterator_category
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_adaptor_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_adaptor_base.h
new file mode 100644
index 000000000000..1173e414ca55
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_adaptor_base.h
@@ -0,0 +1,112 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/use_default.h>
+#include <thrust/iterator/iterator_facade.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+// forward declaration of iterator_adaptor for iterator_adaptor_base below
+template<typename Derived,
+         typename Base,
+         typename Value,
+         typename System,
+         typename Traversal,
+         typename Reference,
+         typename Difference
+>
+class iterator_adaptor;
+
+
+namespace detail
+{
+
+// If T is use_default, return the result of invoking
+// DefaultNullaryFn, otherwise return T.
+// XXX rename to dflt_help
+template <class T, class DefaultNullaryFn>
+struct ia_dflt_help
+  : thrust::detail::eval_if<
+        thrust::detail::is_same<T, thrust::use_default>::value
+      , DefaultNullaryFn
+      , thrust::detail::identity_<T>
+    >
+{
+}; // end ia_dflt_help
+
+
+// A metafunction which computes an iterator_adaptor's base class,
+// a specialization of iterator_facade.
+template<typename Derived,
+         typename Base,
+         typename Value,
+         typename System,
+         typename Traversal,
+         typename Reference,
+         typename Difference
+>
+  struct iterator_adaptor_base
+{
+  typedef typename ia_dflt_help<
+    Value,
+    iterator_value<Base>
+  >::type value;
+
+  typedef typename ia_dflt_help<
+    System,
+    thrust::iterator_system<Base>
+  >::type system;
+
+  typedef typename ia_dflt_help<
+    Traversal,
+    thrust::iterator_traversal<Base>
+  >::type traversal;
+
+  typedef typename ia_dflt_help<
+    Reference,
+    thrust::detail::eval_if<
+      thrust::detail::is_same<Value,use_default>::value,
+      thrust::iterator_reference<Base>,
+      thrust::detail::add_reference<Value>
+    >
+  >::type reference;
+
+  typedef typename ia_dflt_help<
+    Difference,
+    iterator_difference<Base>
+  >::type difference;
+
+  typedef thrust::iterator_facade<
+    Derived,
+    value,
+    system,
+    traversal,
+    reference,
+    difference
+  > type;
+}; // end iterator_adaptor_base
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_system.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_system.h
new file mode 100644
index 000000000000..e6103b539ae5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_system.h
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_categories.h>
+#include <thrust/iterator/detail/iterator_traversal_tags.h>
+#include <thrust/iterator/detail/host_system_tag.h>
+#include <thrust/iterator/detail/device_system_tag.h>
+#include <thrust/iterator/detail/any_system_tag.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// forward declaration
+template <typename> struct is_iterator_system;
+
+template <typename> struct device_iterator_category_to_backend_system;
+
+// XXX this should work entirely differently
+// we should just specialize this metafunction for iterator_category_with_system_and_traversal
+template<typename Category>
+  struct iterator_category_to_system
+    // convertible to host iterator?
+    : eval_if<
+        or_<
+          is_convertible<Category, thrust::input_host_iterator_tag>,
+          is_convertible<Category, thrust::output_host_iterator_tag>
+        >::value,
+
+        detail::identity_<thrust::host_system_tag>,
+        
+        // convertible to device iterator?
+        eval_if<
+          or_<
+            is_convertible<Category, thrust::input_device_iterator_tag>,
+            is_convertible<Category, thrust::output_device_iterator_tag>
+          >::value,
+
+          detail::identity_<thrust::device_system_tag>,
+
+          // unknown system
+          detail::identity_<void>
+        > // if device
+      > // if host
+{
+}; // end iterator_category_to_system
+
+
+template<typename CategoryOrTraversal>
+  struct iterator_category_or_traversal_to_system
+    : eval_if<
+        is_iterator_system<CategoryOrTraversal>::value,
+        detail::identity_<CategoryOrTraversal>,
+        iterator_category_to_system<CategoryOrTraversal>
+      >
+{
+}; // end iterator_category_or_traversal_to_system
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_traversal.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_traversal.h
new file mode 100644
index 000000000000..d8c736c50a4a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_to_traversal.h
@@ -0,0 +1,130 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_categories.h>
+#include <thrust/iterator/detail/iterator_traversal_tags.h>
+#include <thrust/iterator/detail/iterator_category_to_system.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+// forward declarations
+template <typename> struct is_iterator_system;
+template <typename> struct is_iterator_traversal;
+
+template <typename Category>
+  struct host_system_category_to_traversal
+    : eval_if<
+        is_convertible<Category, random_access_host_iterator_tag>::value,
+        detail::identity_<random_access_traversal_tag>,
+        eval_if<
+          is_convertible<Category, bidirectional_host_iterator_tag>::value,
+          detail::identity_<bidirectional_traversal_tag>,
+          eval_if<
+            is_convertible<Category, forward_host_iterator_tag>::value,
+            detail::identity_<forward_traversal_tag>,
+            eval_if<
+              is_convertible<Category, input_host_iterator_tag>::value,
+              detail::identity_<single_pass_traversal_tag>,
+              eval_if<
+                is_convertible<Category, output_host_iterator_tag>::value,
+                detail::identity_<incrementable_traversal_tag>,
+                detail::identity_<void>
+              >
+            >
+          >
+        >
+      >
+{
+}; // end host_system_category_to_traversal
+
+
+
+template <typename Category>
+  struct device_system_category_to_traversal
+    : eval_if<
+        is_convertible<Category, random_access_device_iterator_tag>::value,
+        detail::identity_<random_access_traversal_tag>,
+        eval_if<
+          is_convertible<Category, bidirectional_device_iterator_tag>::value,
+          detail::identity_<bidirectional_traversal_tag>,
+          eval_if<
+            is_convertible<Category, forward_device_iterator_tag>::value,
+            detail::identity_<forward_traversal_tag>,
+            eval_if<
+              is_convertible<Category, input_device_iterator_tag>::value,
+              detail::identity_<single_pass_traversal_tag>,
+              eval_if<
+                is_convertible<Category, output_device_iterator_tag>::value,
+                detail::identity_<incrementable_traversal_tag>,
+                detail::identity_<void>
+              >
+            >
+          >
+        >
+      >
+{
+}; // end device_system_category_to_traversal
+
+
+template<typename Category>
+  struct category_to_traversal
+      // check for host system
+    : eval_if<
+        or_<
+          is_convertible<Category, thrust::input_host_iterator_tag>,
+          is_convertible<Category, thrust::output_host_iterator_tag>
+        >::value,
+
+        host_system_category_to_traversal<Category>,
+
+        // check for device system
+        eval_if<
+          or_<
+            is_convertible<Category, thrust::input_device_iterator_tag>,
+            is_convertible<Category, thrust::output_device_iterator_tag>
+          >::value,
+
+          device_system_category_to_traversal<Category>,
+
+          // unknown category
+          detail::identity_<void>
+        >
+      >
+{};
+
+
+template <typename CategoryOrTraversal>
+  struct iterator_category_to_traversal
+    : eval_if<
+        is_iterator_traversal<CategoryOrTraversal>::value,
+        detail::identity_<CategoryOrTraversal>,
+        category_to_traversal<CategoryOrTraversal>
+      >
+{
+}; // end iterator_category_to_traversal
+
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h
new file mode 100644
index 000000000000..cdd8a6d366a0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_category_with_system_and_traversal.h
@@ -0,0 +1,56 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename Category, typename System, typename Traversal>
+  struct iterator_category_with_system_and_traversal
+    : Category
+{
+}; // end iterator_category_with_system_and_traversal
+
+
+// specialize iterator_category_to_system for iterator_category_with_system_and_traversal
+template<typename Category> struct iterator_category_to_system;
+
+template<typename Category, typename System, typename Traversal>
+  struct iterator_category_to_system<iterator_category_with_system_and_traversal<Category,System,Traversal> >
+{
+  typedef System type;
+}; // end iterator_category_to_system
+
+
+// specialize iterator_category_to_traversal for iterator_category_with_system_and_traversal
+template<typename Category> struct iterator_category_to_traversal;
+
+template<typename Category, typename System, typename Traversal>
+  struct iterator_category_to_traversal<iterator_category_with_system_and_traversal<Category,System,Traversal> >
+{
+  typedef Traversal type;
+}; // end iterator_category_to_traversal
+
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_facade_category.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_facade_category.h
new file mode 100644
index 000000000000..81b518002baa
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_facade_category.h
@@ -0,0 +1,252 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/host_system_tag.h>
+#include <thrust/iterator/detail/device_system_tag.h>
+#include <thrust/iterator/detail/any_system_tag.h>
+#include <thrust/iterator/iterator_categories.h>
+#include <thrust/iterator/detail/iterator_traversal_tags.h>
+#include <thrust/iterator/detail/is_iterator_category.h>
+#include <thrust/iterator/detail/iterator_category_with_system_and_traversal.h>
+#include <thrust/iterator/detail/iterator_category_to_traversal.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+
+// adapted from http://www.boost.org/doc/libs/1_37_0/libs/iterator/doc/iterator_facade.html#iterator-category
+//
+// in our implementation, R need not be a reference type to result in a category
+// derived from forward_XXX_iterator_tag
+//
+// iterator-category(T,V,R) :=
+//   if(T is convertible to input_host_iterator_tag
+//      || T is convertible to output_host_iterator_tag
+//      || T is convertible to input_device_iterator_tag
+//      || T is convertible to output_device_iterator_tag
+//   )
+//     return T
+//
+//   else if (T is not convertible to incrementable_traversal_tag)
+//     the program is ill-formed
+//
+//   else return a type X satisfying the following two constraints:
+//
+//     1. X is convertible to X1, and not to any more-derived
+//        type, where X1 is defined by:
+//
+//        if (T is convertible to forward_traversal_tag)
+//        {
+//          if (T is convertible to random_access_traversal_tag)
+//            X1 = random_access_host_iterator_tag
+//          else if (T is convertible to bidirectional_traversal_tag)
+//            X1 = bidirectional_host_iterator_tag
+//          else
+//            X1 = forward_host_iterator_tag
+//        }
+//        else
+//        {
+//          if (T is convertible to single_pass_traversal_tag
+//              && R is convertible to V)
+//            X1 = input_host_iterator_tag
+//          else
+//            X1 = T
+//        }
+//
+//     2. category-to-traversal(X) is convertible to the most
+//        derived traversal tag type to which X is also convertible,
+//        and not to any more-derived traversal tag type.
+
+
+template<typename System, typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_default_category;
+
+
+// Thrust's implementation of iterator_facade_default_category is slightly
+// different from Boost's equivalent.
+// Thrust does not check is_convertible<Reference, ValueParam> because Reference
+// may not be a complete type at this point, and implementations of is_convertible
+// typically require that both types be complete.
+// Instead, it simply assumes that if is_convertible<Traversal, single_pass_traversal_tag>,
+// then the category is input_iterator_tag
+
+
+// this is the function for standard system iterators
+template<typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_default_category_std :
+    thrust::detail::eval_if<
+      thrust::detail::is_convertible<Traversal, thrust::forward_traversal_tag>::value,
+      thrust::detail::eval_if<
+        thrust::detail::is_convertible<Traversal, thrust::random_access_traversal_tag>::value,
+        thrust::detail::identity_<std::random_access_iterator_tag>,
+        thrust::detail::eval_if<
+          thrust::detail::is_convertible<Traversal, thrust::bidirectional_traversal_tag>::value,
+          thrust::detail::identity_<std::bidirectional_iterator_tag>,
+          thrust::detail::identity_<std::forward_iterator_tag>
+        >
+      >,
+      thrust::detail::eval_if< // XXX note we differ from Boost here
+        thrust::detail::is_convertible<Traversal, thrust::single_pass_traversal_tag>::value,
+        thrust::detail::identity_<std::input_iterator_tag>,
+        thrust::detail::identity_<Traversal>
+      >
+    >
+{
+}; // end iterator_facade_default_category_std
+
+
+// this is the function for host system iterators
+template<typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_default_category_host :
+    thrust::detail::eval_if<
+      thrust::detail::is_convertible<Traversal, thrust::forward_traversal_tag>::value,
+      thrust::detail::eval_if<
+        thrust::detail::is_convertible<Traversal, thrust::random_access_traversal_tag>::value,
+        thrust::detail::identity_<thrust::random_access_host_iterator_tag>,
+        thrust::detail::eval_if<
+          thrust::detail::is_convertible<Traversal, thrust::bidirectional_traversal_tag>::value,
+          thrust::detail::identity_<thrust::bidirectional_host_iterator_tag>,
+          thrust::detail::identity_<thrust::forward_host_iterator_tag>
+        >
+      >,
+      thrust::detail::eval_if< // XXX note we differ from Boost here
+        thrust::detail::is_convertible<Traversal, thrust::single_pass_traversal_tag>::value,
+        thrust::detail::identity_<thrust::input_host_iterator_tag>,
+        thrust::detail::identity_<Traversal>
+      >
+    >
+{
+}; // end iterator_facade_default_category_host
+
+
+// this is the function for device system iterators
+template<typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_default_category_device :
+    thrust::detail::eval_if<
+      thrust::detail::is_convertible<Traversal, thrust::forward_traversal_tag>::value,
+      thrust::detail::eval_if<
+        thrust::detail::is_convertible<Traversal, thrust::random_access_traversal_tag>::value,
+        thrust::detail::identity_<thrust::random_access_device_iterator_tag>,
+        thrust::detail::eval_if<
+          thrust::detail::is_convertible<Traversal, thrust::bidirectional_traversal_tag>::value,
+          thrust::detail::identity_<thrust::bidirectional_device_iterator_tag>,
+          thrust::detail::identity_<thrust::forward_device_iterator_tag>
+        >
+      >,
+      thrust::detail::eval_if<
+        thrust::detail::is_convertible<Traversal, thrust::single_pass_traversal_tag>::value, // XXX note we differ from Boost here
+        thrust::detail::identity_<thrust::input_device_iterator_tag>,
+        thrust::detail::identity_<Traversal>
+      >
+    >
+{
+}; // end iterator_facade_default_category_device
+
+
+// this is the function for any system iterators
+template<typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_default_category_any
+{
+  typedef thrust::detail::iterator_category_with_system_and_traversal<
+    typename iterator_facade_default_category_std<Traversal, ValueParam, Reference>::type,
+    thrust::any_system_tag,
+    Traversal
+  > type;
+}; // end iterator_facade_default_category_any
+
+
+template<typename System, typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_default_category
+      // check for any system
+    : thrust::detail::eval_if<
+        thrust::detail::is_convertible<System, thrust::any_system_tag>::value,
+        iterator_facade_default_category_any<Traversal, ValueParam, Reference>,
+
+        // check for host system
+        thrust::detail::eval_if<
+          thrust::detail::is_convertible<System, thrust::host_system_tag>::value,
+          iterator_facade_default_category_host<Traversal, ValueParam, Reference>,
+
+          // check for device system
+          thrust::detail::eval_if<
+            thrust::detail::is_convertible<System, thrust::device_system_tag>::value,
+            iterator_facade_default_category_device<Traversal, ValueParam, Reference>,
+
+            // if we don't recognize the system, get a standard iterator category
+            // and combine it with System & Traversal
+            thrust::detail::identity_<
+              thrust::detail::iterator_category_with_system_and_traversal<
+                typename iterator_facade_default_category_std<Traversal, ValueParam, Reference>::type,
+                System,
+                Traversal
+              >
+            >
+          >
+        >
+      >
+{};
+
+
+template<typename System, typename Traversal, typename ValueParam, typename Reference>
+  struct iterator_facade_category_impl
+{
+  typedef typename iterator_facade_default_category<
+    System,Traversal,ValueParam,Reference
+  >::type category;
+
+  // we must be able to deduce both Traversal & System from category
+  // otherwise, munge them all together
+  typedef typename thrust::detail::eval_if<
+    thrust::detail::and_<
+      thrust::detail::is_same<
+        Traversal,
+        typename thrust::detail::iterator_category_to_traversal<category>::type
+      >,
+      thrust::detail::is_same<
+        System,
+        typename thrust::detail::iterator_category_to_system<category>::type
+      >
+    >::value,
+    thrust::detail::identity_<category>,
+    thrust::detail::identity_<thrust::detail::iterator_category_with_system_and_traversal<category,System,Traversal> >
+  >::type type;
+}; // end iterator_facade_category_impl
+
+
+template<typename CategoryOrSystem,
+         typename CategoryOrTraversal,
+         typename ValueParam,
+         typename Reference>
+  struct iterator_facade_category
+{
+  typedef typename
+  thrust::detail::eval_if<
+    thrust::detail::is_iterator_category<CategoryOrTraversal>::value,
+    thrust::detail::identity_<CategoryOrTraversal>, // categories are fine as-is
+    iterator_facade_category_impl<CategoryOrSystem, CategoryOrTraversal, ValueParam, Reference>
+  >::type type;
+}; // end iterator_facade_category
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traits.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traits.inl
new file mode 100644
index 000000000000..544c24f0b426
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traits.inl
@@ -0,0 +1,136 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_categories.h>
+#include <thrust/iterator/detail/iterator_category_to_traversal.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/type_traits/void_t.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename Iterator>
+  struct iterator_value
+{
+  typedef typename thrust::iterator_traits<Iterator>::value_type type;
+}; // end iterator_value
+
+template <typename Iterator>
+using iterator_value_t = typename iterator_value<Iterator>::type;
+
+template<typename Iterator>
+  struct iterator_pointer
+{
+  typedef typename thrust::iterator_traits<Iterator>::pointer type;
+}; // end iterator_pointer
+
+template <typename Iterator>
+using iterator_pointer_t = typename iterator_pointer<Iterator>::type;
+
+template<typename Iterator>
+  struct iterator_reference
+{
+  typedef typename iterator_traits<Iterator>::reference type;
+}; // end iterator_reference
+
+template <typename Iterator>
+using iterator_reference_t = typename iterator_reference<Iterator>::type;
+
+template<typename Iterator>
+  struct iterator_difference
+{
+  typedef typename thrust::iterator_traits<Iterator>::difference_type type;
+}; // end iterator_difference
+
+template <typename Iterator>
+using iterator_difference_t = typename iterator_difference<Iterator>::type;
+
+namespace detail
+{
+
+template <typename Iterator, typename = void>
+struct iterator_system_impl {};
+
+template <typename Iterator>
+struct iterator_system_impl<
+  Iterator
+, typename voider<
+    typename iterator_traits<Iterator>::iterator_category
+  >::type
+>
+  : detail::iterator_category_to_system<
+      typename iterator_traits<Iterator>::iterator_category
+    >
+{};
+
+} // namespace detail
+
+template <typename Iterator>
+struct iterator_system : detail::iterator_system_impl<Iterator> {};
+
+// specialize iterator_system for void *, which has no category
+template<>
+  struct iterator_system<void *>
+{
+  typedef thrust::iterator_system<int*>::type type;
+}; // end iterator_system<void*>
+
+template<>
+  struct iterator_system<const void *>
+{
+  typedef thrust::iterator_system<const int*>::type type;
+}; // end iterator_system<void*>
+
+template <typename Iterator>
+using iterator_system_t = typename iterator_system<Iterator>::type;
+
+template <typename Iterator>
+  struct iterator_traversal
+    : detail::iterator_category_to_traversal<
+        typename thrust::iterator_traits<Iterator>::iterator_category
+      >
+{
+}; // end iterator_traversal
+
+namespace detail
+{
+
+template <typename T>
+  struct is_iterator_traversal
+    : thrust::detail::is_convertible<T, incrementable_traversal_tag>
+{
+}; // end is_iterator_traversal
+
+
+template<typename T>
+  struct is_iterator_system
+    : detail::or_<
+        detail::is_convertible<T, any_system_tag>,
+        detail::or_<
+          detail::is_convertible<T, host_system_tag>,
+          detail::is_convertible<T, device_system_tag>
+        >
+      >
+{
+}; // end is_iterator_system
+
+
+} // end namespace detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traversal_tags.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traversal_tags.h
new file mode 100644
index 000000000000..1fbc8a1e49c0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/iterator_traversal_tags.h
@@ -0,0 +1,42 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// define Boost's traversal tags
+struct no_traversal_tag {};
+
+struct incrementable_traversal_tag
+  : no_traversal_tag {};
+
+struct single_pass_traversal_tag
+  : incrementable_traversal_tag {};
+
+struct forward_traversal_tag
+  : single_pass_traversal_tag {};
+
+struct bidirectional_traversal_tag
+  : forward_traversal_tag {};
+
+struct random_access_traversal_tag
+  : bidirectional_traversal_tag {};
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/join_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/join_iterator.h
new file mode 100644
index 000000000000..83f143dc00a6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/join_iterator.h
@@ -0,0 +1,133 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/detail/type_traits.h>
+
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename Difference,
+         typename Reference>
+class join_iterator;
+
+
+namespace join_iterator_detail
+{
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename Difference,
+         typename Reference>
+struct join_iterator_base
+{
+  typedef typename thrust::detail::remove_reference<Reference>::type value_type;
+
+  typedef typename thrust::iterator_system<RandomAccessIterator1>::type  system1;
+  typedef typename thrust::iterator_system<RandomAccessIterator2>::type  system2;
+  typedef typename thrust::detail::minimum_system<system1,system2>::type system;
+
+  typedef thrust::iterator_adaptor<
+    join_iterator<RandomAccessIterator1,RandomAccessIterator2,Difference,Reference>,
+    thrust::counting_iterator<Difference>,
+    value_type,
+    system,
+    thrust::random_access_traversal_tag,
+    Reference,
+    Difference
+  > type;
+}; // end join_iterator_base
+
+
+} // end join_iterator_detail
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename Difference = typename thrust::iterator_difference<RandomAccessIterator1>::type,
+         typename Reference  = typename thrust::iterator_value<RandomAccessIterator1>::type>
+class join_iterator
+  : public join_iterator_detail::join_iterator_base<RandomAccessIterator1, RandomAccessIterator2, Difference, Reference>::type
+{
+  private:
+    typedef typename join_iterator_detail::join_iterator_base<RandomAccessIterator1, RandomAccessIterator2, Difference, Reference>::type super_t;
+    typedef typename super_t::difference_type size_type;
+
+  public:
+    inline __host__ __device__
+    join_iterator(RandomAccessIterator1 first1, size_type n, RandomAccessIterator2 first2)
+      : super_t(thrust::counting_iterator<size_type>(0)),
+        m_n1(n),
+        m_iter1(first1),
+        m_iter2(first2 - m_n1)
+    {}
+
+
+    inline __host__ __device__
+    join_iterator(const join_iterator &other)
+      : super_t(other),
+        m_n1(other.m_n1),
+        m_iter1(other.m_iter1),
+        m_iter2(other.m_iter2)
+    {}
+
+
+  private:
+    friend class thrust::iterator_core_access;
+
+    // MSVC 2013 and 2015 incorrectly warning about returning a reference to
+    // a local/temporary here.
+    // See goo.gl/LELTNp
+    THRUST_DISABLE_MSVC_WARNING_BEGIN(4172)
+
+    __host__ __device__
+    typename super_t::reference dereference() const
+    {
+      size_type i = *super_t::base();
+      return (i < m_n1) ? m_iter1[i] : static_cast<typename super_t::reference>(m_iter2[i]);
+    } // end dereference()
+
+    THRUST_DISABLE_MSVC_WARNING_END(4172)
+
+
+    size_type m_n1;
+    RandomAccessIterator1 m_iter1;
+    RandomAccessIterator2 m_iter2;
+}; // end join_iterator
+
+
+template<typename RandomAccessIterator1, typename Size, typename RandomAccessIterator2>
+__host__ __device__
+join_iterator<RandomAccessIterator1,RandomAccessIterator2,Size> make_join_iterator(RandomAccessIterator1 first1, Size n1, RandomAccessIterator2 first2)
+{
+  return join_iterator<RandomAccessIterator1,RandomAccessIterator2,Size>(first1, n1, first2);
+} // end make_join_iterator()
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_category.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_category.h
new file mode 100644
index 000000000000..01e7e82c5358
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_category.h
@@ -0,0 +1,53 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits/minimum_type.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{ 
+
+template<typename T1,
+         typename T2  = minimum_type_detail::any_conversion,
+         typename T3  = minimum_type_detail::any_conversion,
+         typename T4  = minimum_type_detail::any_conversion,
+         typename T5  = minimum_type_detail::any_conversion,
+         typename T6  = minimum_type_detail::any_conversion,
+         typename T7  = minimum_type_detail::any_conversion,
+         typename T8  = minimum_type_detail::any_conversion,
+         typename T9  = minimum_type_detail::any_conversion,
+         typename T10 = minimum_type_detail::any_conversion,
+         typename T11 = minimum_type_detail::any_conversion,
+         typename T12 = minimum_type_detail::any_conversion,
+         typename T13 = minimum_type_detail::any_conversion,
+         typename T14 = minimum_type_detail::any_conversion,
+         typename T15 = minimum_type_detail::any_conversion,
+         typename T16 = minimum_type_detail::any_conversion>
+  struct minimum_category
+    : minimum_type<T1,T2,T3,T4,T5,T6,T7,T8,T9,T10,T11,T12,T13,T14,T15,T16>
+{
+}; // end minimum_category
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_system.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_system.h
new file mode 100644
index 000000000000..dcb29ccd254b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/minimum_system.h
@@ -0,0 +1,81 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/is_metafunction_defined.h>
+#include <thrust/detail/type_traits/minimum_type.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{ 
+
+
+template<typename T1,
+         typename T2  = void,
+         typename T3  = void,
+         typename T4  = void,
+         typename T5  = void,
+         typename T6  = void,
+         typename T7  = void,
+         typename T8  = void,
+         typename T9  = void,
+         typename T10 = void,
+         typename T11 = void,
+         typename T12 = void,
+         typename T13 = void,
+         typename T14 = void,
+         typename T15 = void,
+         typename T16 = void>
+  struct unrelated_systems {};
+
+
+// if a minimum system exists for these arguments, return it
+// otherwise, collect the arguments and report them as unrelated
+template<typename T1,
+         typename T2  = minimum_type_detail::any_conversion,
+         typename T3  = minimum_type_detail::any_conversion,
+         typename T4  = minimum_type_detail::any_conversion,
+         typename T5  = minimum_type_detail::any_conversion,
+         typename T6  = minimum_type_detail::any_conversion,
+         typename T7  = minimum_type_detail::any_conversion,
+         typename T8  = minimum_type_detail::any_conversion,
+         typename T9  = minimum_type_detail::any_conversion,
+         typename T10 = minimum_type_detail::any_conversion,
+         typename T11 = minimum_type_detail::any_conversion,
+         typename T12 = minimum_type_detail::any_conversion,
+         typename T13 = minimum_type_detail::any_conversion,
+         typename T14 = minimum_type_detail::any_conversion,
+         typename T15 = minimum_type_detail::any_conversion,
+         typename T16 = minimum_type_detail::any_conversion>
+  struct minimum_system
+    : thrust::detail::eval_if<
+        is_metafunction_defined<
+          minimum_type<T1,T2,T3,T4,T5,T6,T7,T8,T9,T10,T11,T12,T13,T14,T15,T16>
+        >::value,
+        minimum_type<T1,T2,T3,T4,T5,T6,T7,T8,T9,T10,T11,T12,T13,T14,T15,T16>,
+        thrust::detail::identity_<
+          unrelated_systems<T1,T2,T3,T4,T5,T6,T7,T8,T9,T10,T11,T12,T13,T14,T15,T16>
+        >
+      >
+{}; // end minimum_system
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/normal_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/normal_iterator.h
new file mode 100644
index 000000000000..eb5d336044fb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/normal_iterator.h
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file normal_iterator.h
+ *  \brief Defines the interface to an iterator class
+ *         which adapts a pointer type.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/type_traits/is_contiguous_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+template<typename Pointer>
+  class normal_iterator
+    : public iterator_adaptor<
+        normal_iterator<Pointer>,
+        Pointer
+      >
+{
+  typedef iterator_adaptor<normal_iterator<Pointer>, Pointer> super_t;
+
+  public:
+    __host__ __device__
+    normal_iterator() {}
+
+    __host__ __device__
+    normal_iterator(Pointer p)
+      : super_t(p) {}
+    
+    template<typename OtherPointer>
+    __host__ __device__
+    normal_iterator(const normal_iterator<OtherPointer> &other,
+                    typename thrust::detail::enable_if_convertible<
+                      OtherPointer,
+                      Pointer
+                    >::type * = 0)
+      : super_t(other.base()) {}
+
+}; // end normal_iterator
+
+
+template<typename Pointer>
+  inline __host__ __device__ normal_iterator<Pointer> make_normal_iterator(Pointer ptr)
+{
+  return normal_iterator<Pointer>(ptr);
+}
+
+} // end detail
+
+template <typename T>
+struct proclaim_contiguous_iterator<
+  thrust::detail::normal_iterator<T>
+> : true_type {};
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/permutation_iterator_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/permutation_iterator_base.h
new file mode 100644
index 000000000000..d586cabb7802
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/permutation_iterator_base.h
@@ -0,0 +1,54 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template<typename,typename> class permutation_iterator;
+
+
+namespace detail
+{
+
+template<typename ElementIterator,
+         typename IndexIterator>
+  struct permutation_iterator_base
+{
+  typedef typename thrust::iterator_system<ElementIterator>::type System1;
+  typedef typename thrust::iterator_system<IndexIterator>::type System2;
+
+  typedef thrust::iterator_adaptor<
+    permutation_iterator<ElementIterator,IndexIterator>,
+    IndexIterator,
+    typename thrust::iterator_value<ElementIterator>::type,
+    typename detail::minimum_system<System1,System2>::type,
+    thrust::use_default,
+    typename thrust::iterator_reference<ElementIterator>::type
+  > type;
+}; // end permutation_iterator_base
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/retag.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/retag.h
new file mode 100644
index 000000000000..d277d8b6f2b8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/retag.h
@@ -0,0 +1,147 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/tagged_iterator.h>
+#include <thrust/detail/pointer.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+// we can retag an iterator if FromTag converts to ToTag
+// or vice versa
+template<typename FromTag, typename ToTag>
+  struct is_retaggable
+    : integral_constant<
+        bool,
+        (is_convertible<FromTag,ToTag>::value || is_convertible<ToTag,FromTag>::value)
+      >
+{};
+
+
+template<typename FromTag, typename ToTag, typename Result>
+  struct enable_if_retaggable
+    : enable_if<
+        is_retaggable<FromTag,ToTag>::value,
+        Result
+      >
+{}; // end enable_if_retaggable
+
+
+} // end detail
+
+
+template<typename Tag, typename Iterator>
+__host__ __device__
+  thrust::detail::tagged_iterator<Iterator,Tag>
+    reinterpret_tag(Iterator iter)
+{
+  return thrust::detail::tagged_iterator<Iterator,Tag>(iter);
+} // end reinterpret_tag()
+
+
+// specialization for raw pointer
+template<typename Tag, typename T>
+__host__ __device__
+  thrust::pointer<T,Tag>
+    reinterpret_tag(T *ptr)
+{
+  return thrust::pointer<T,Tag>(ptr);
+} // end reinterpret_tag()
+
+
+// specialization for thrust::pointer
+template<typename Tag, typename T, typename OtherTag, typename Reference, typename Derived>
+__host__ __device__
+  thrust::pointer<T,Tag>
+    reinterpret_tag(thrust::pointer<T,OtherTag,Reference,Derived> ptr)
+{
+  return reinterpret_tag<Tag>(ptr.get());
+} // end reinterpret_tag()
+
+
+// avoid deeply-nested tagged_iterator
+template<typename Tag, typename BaseIterator, typename OtherTag>
+__host__ __device__
+  thrust::detail::tagged_iterator<BaseIterator,Tag>
+    reinterpret_tag(thrust::detail::tagged_iterator<BaseIterator,OtherTag> iter)
+{
+  return reinterpret_tag<Tag>(iter.base());
+} // end reinterpret_tag()
+
+
+template<typename Tag, typename Iterator>
+__host__ __device__
+  typename thrust::detail::enable_if_retaggable<
+    typename thrust::iterator_system<Iterator>::type,
+    Tag,
+    thrust::detail::tagged_iterator<Iterator,Tag>
+  >::type
+    retag(Iterator iter)
+{
+  return reinterpret_tag<Tag>(iter);
+} // end retag()
+
+
+// specialization for raw pointer
+template<typename Tag, typename T>
+__host__ __device__
+  typename thrust::detail::enable_if_retaggable<
+    typename thrust::iterator_system<T*>::type,
+    Tag,
+    thrust::pointer<T,Tag>
+  >::type
+    retag(T *ptr)
+{
+  return reinterpret_tag<Tag>(ptr);
+} // end retag()
+
+
+// specialization for thrust::pointer
+template<typename Tag, typename T, typename OtherTag>
+__host__ __device__
+  typename thrust::detail::enable_if_retaggable<
+    OtherTag,
+    Tag,
+    thrust::pointer<T,Tag>
+  >::type
+    retag(thrust::pointer<T,OtherTag> ptr)
+{
+  return reinterpret_tag<Tag>(ptr);
+} // end retag()
+
+
+// avoid deeply-nested tagged_iterator
+template<typename Tag, typename BaseIterator, typename OtherTag>
+__host__ __device__
+  typename thrust::detail::enable_if_retaggable<
+    OtherTag,
+    Tag,
+    thrust::detail::tagged_iterator<BaseIterator,Tag>
+  >::type
+    retag(thrust::detail::tagged_iterator<BaseIterator,OtherTag> iter)
+{
+  return reinterpret_tag<Tag>(iter);
+} // end retag()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator.inl
new file mode 100644
index 000000000000..9182ac3e8967
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator.inl
@@ -0,0 +1,118 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/reverse_iterator.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace detail
+{
+
+__thrust_exec_check_disable__
+template<typename Iterator>
+  __host__ __device__
+  Iterator prior(Iterator x)
+{
+  return --x;
+} // end prior()
+
+} // end detail
+
+template<typename BidirectionalIterator>
+  __host__ __device__
+  reverse_iterator<BidirectionalIterator>
+    ::reverse_iterator(BidirectionalIterator x)
+      :super_t(x)
+{
+} // end reverse_iterator::reverse_iterator()
+
+template<typename BidirectionalIterator>
+  template<typename OtherBidirectionalIterator>
+    __host__ __device__
+    reverse_iterator<BidirectionalIterator>
+      ::reverse_iterator(reverse_iterator<OtherBidirectionalIterator> const &r
+// XXX msvc screws this up
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+                     , typename thrust::detail::enable_if<
+                         thrust::detail::is_convertible<
+                           OtherBidirectionalIterator,
+                           BidirectionalIterator
+                         >::value
+                       >::type *
+#endif // MSVC
+                     )
+        :super_t(r.base())
+{
+} // end reverse_iterator::reverse_iterator()
+
+template<typename BidirectionalIterator>
+  __host__ __device__
+  typename reverse_iterator<BidirectionalIterator>::super_t::reference
+    reverse_iterator<BidirectionalIterator>
+      ::dereference() const
+{
+  return *thrust::detail::prior(this->base());
+} // end reverse_iterator::increment()
+
+template<typename BidirectionalIterator>
+  __host__ __device__
+  void reverse_iterator<BidirectionalIterator>
+    ::increment()
+{
+  --this->base_reference();
+} // end reverse_iterator::increment()
+
+template<typename BidirectionalIterator>
+  __host__ __device__
+  void reverse_iterator<BidirectionalIterator>
+    ::decrement()
+{
+  ++this->base_reference();
+} // end reverse_iterator::decrement()
+
+template<typename BidirectionalIterator>
+  __host__ __device__
+  void reverse_iterator<BidirectionalIterator>
+    ::advance(typename super_t::difference_type n)
+{
+  this->base_reference() += -n;
+} // end reverse_iterator::advance()
+
+template<typename BidirectionalIterator>
+  template<typename OtherBidirectionalIterator>
+    __host__ __device__
+    typename reverse_iterator<BidirectionalIterator>::super_t::difference_type
+      reverse_iterator<BidirectionalIterator>
+        ::distance_to(reverse_iterator<OtherBidirectionalIterator> const &y) const
+{
+  return this->base_reference() - y.base();
+} // end reverse_iterator::distance_to()
+
+template<typename BidirectionalIterator>
+__host__ __device__
+reverse_iterator<BidirectionalIterator> make_reverse_iterator(BidirectionalIterator x)
+{
+  return reverse_iterator<BidirectionalIterator>(x);
+} // end make_reverse_iterator()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator_base.h
new file mode 100644
index 000000000000..de3bafde979f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/reverse_iterator_base.h
@@ -0,0 +1,43 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename> class reverse_iterator;
+
+namespace detail
+{
+
+template<typename BidirectionalIterator>
+  struct reverse_iterator_base
+{
+  typedef thrust::iterator_adaptor<
+    thrust::reverse_iterator<BidirectionalIterator>,
+    BidirectionalIterator
+  > type;
+}; // end reverse_iterator_base
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tagged_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tagged_iterator.h
new file mode 100644
index 000000000000..24cbbb7361d2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tagged_iterator.h
@@ -0,0 +1,85 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/use_default.h>
+#include <thrust/type_traits/is_contiguous_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+template <typename,typename> class tagged_iterator;
+
+template<typename Iterator, typename Tag>
+  struct tagged_iterator_base
+{
+  typedef thrust::iterator_adaptor<
+    tagged_iterator<Iterator,Tag>,
+    Iterator,
+    typename thrust::iterator_value<Iterator>::type,
+    Tag,
+    typename thrust::iterator_traversal<Iterator>::type,
+    typename thrust::iterator_reference<Iterator>::type,
+    typename thrust::iterator_difference<Iterator>::type
+  > type;
+}; // end tagged_iterator_base
+
+template<typename Iterator, typename Tag>
+  class tagged_iterator
+    : public tagged_iterator_base<Iterator,Tag>::type
+{
+  private:
+    typedef typename tagged_iterator_base<Iterator,Tag>::type super_t;
+
+  public:
+    __host__ __device__
+    tagged_iterator() {}
+
+    __host__ __device__
+    explicit tagged_iterator(Iterator x)
+      : super_t(x) {}
+}; // end tagged_iterator
+
+/*! \p make_tagged_iterator creates a \p tagged_iterator
+ *  from a \c Iterator with system tag \c Tag.
+ *
+ *  \tparam Tag Any system tag.
+ *  \tparam Iterator Any iterator type.
+ *  \param iter The iterator of interest.
+ *  \return An iterator whose system tag is \p Tag and whose behavior is otherwise
+ *          equivalent to \p iter.
+ */
+template <typename Tag, typename Iterator>
+inline auto make_tagged_iterator(Iterator iter) -> tagged_iterator<Iterator, Tag>
+{
+  return tagged_iterator<Iterator, Tag>(iter);
+}
+
+} // end detail
+
+// tagged_iterator is trivial if its base iterator is.
+template <typename BaseIterator, typename Tag>
+struct proclaim_contiguous_iterator<
+  detail::tagged_iterator<BaseIterator, Tag>
+> : is_contiguous_iterator<BaseIterator> {};
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_input_output_iterator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_input_output_iterator.inl
new file mode 100644
index 000000000000..b4792f724df3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_input_output_iterator.inl
@@ -0,0 +1,107 @@
+/*
+ *  Copyright 2020-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename InputFunction, typename OutputFunction, typename Iterator>
+  class transform_input_output_iterator;
+
+namespace detail
+{
+
+// Proxy reference that invokes InputFunction when reading from and
+// OutputFunction when writing to the dereferenced iterator
+template <typename InputFunction, typename OutputFunction, typename Iterator>
+  class transform_input_output_iterator_proxy
+{
+  using iterator_value_type = typename thrust::iterator_value<Iterator>::type;
+
+  using Value = invoke_result_t<InputFunction, iterator_value_type>;
+
+  public:
+    __host__ __device__
+    transform_input_output_iterator_proxy(const Iterator& io, InputFunction input_function, OutputFunction output_function)
+      : io(io), input_function(input_function), output_function(output_function)
+    {
+    }
+
+    transform_input_output_iterator_proxy(const transform_input_output_iterator_proxy&) = default;
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    operator Value const() const
+    {
+      return input_function(*io);
+    }
+
+    __thrust_exec_check_disable__
+    template <typename T>
+    __host__ __device__
+    transform_input_output_iterator_proxy operator=(const T& x)
+    {
+      *io = output_function(x);
+      return *this;
+    }
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    transform_input_output_iterator_proxy operator=(const transform_input_output_iterator_proxy& x)
+    {
+      *io = output_function(x);
+      return *this;
+    }
+
+  private:
+    Iterator io;
+    InputFunction input_function;
+    OutputFunction output_function;
+};
+
+// Compute the iterator_adaptor instantiation to be used for transform_input_output_iterator
+template <typename InputFunction, typename OutputFunction, typename Iterator>
+struct transform_input_output_iterator_base
+{
+private:
+  using iterator_value_type = typename thrust::iterator_value<Iterator>::type;
+
+public:
+    typedef thrust::iterator_adaptor
+    <
+        transform_input_output_iterator<InputFunction, OutputFunction, Iterator>
+      , Iterator
+      , detail::invoke_result_t<InputFunction, iterator_value_type>
+      , thrust::use_default
+      , thrust::use_default
+      , transform_input_output_iterator_proxy<InputFunction, OutputFunction, Iterator>
+    > type;
+};
+
+// Register transform_input_output_iterator_proxy with 'is_proxy_reference' from
+// type_traits to enable its use with algorithms.
+template <typename InputFunction, typename OutputFunction, typename Iterator>
+struct is_proxy_reference<
+    transform_input_output_iterator_proxy<InputFunction, OutputFunction, Iterator> >
+    : public thrust::detail::true_type {};
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_iterator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_iterator.inl
new file mode 100644
index 000000000000..6930a1b08361
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_iterator.inl
@@ -0,0 +1,75 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/iterator/iterator_adaptor.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/result_of_adaptable_function.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <class UnaryFunction, class Iterator, class Reference, class Value>
+  class transform_iterator;
+
+namespace detail
+{
+
+// Compute the iterator_adaptor instantiation to be used for transform_iterator
+template <class UnaryFunc, class Iterator, class Reference, class Value>
+struct transform_iterator_base
+{
+ private:
+    // By default, dereferencing the iterator yields the same as the function.
+    typedef typename thrust::detail::ia_dflt_help<
+      Reference,
+      thrust::detail::result_of_adaptable_function<UnaryFunc(typename thrust::iterator_value<Iterator>::type)>
+    >::type reference;
+
+    // To get the default for Value: remove any reference on the
+    // result type, but retain any constness to signal
+    // non-writability.  Note that if we adopt Thomas' suggestion
+    // to key non-writability *only* on the Reference argument,
+    // we'd need to strip constness here as well.
+    typedef typename thrust::detail::ia_dflt_help<
+      Value,
+      thrust::detail::remove_reference<reference>
+    >::type cv_value_type;
+
+ public:
+    typedef thrust::iterator_adaptor
+    <
+        transform_iterator<UnaryFunc, Iterator, Reference, Value>
+      , Iterator
+      , cv_value_type
+      , thrust::use_default   // Leave the system alone
+        //, thrust::use_default   // Leave the traversal alone
+        // use the Iterator's category to let any system iterators remain random access even though
+        // transform_iterator's reference type may not be a reference
+        // XXX figure out why only iterators whose reference types are true references are random access
+        , typename thrust::iterator_traits<Iterator>::iterator_category
+      , reference
+    > type;
+};
+
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_output_iterator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_output_iterator.inl
new file mode 100644
index 000000000000..d5033f105894
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/transform_output_iterator.inl
@@ -0,0 +1,81 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/transform_output_iterator.h>
+#include <thrust/iterator/iterator_adaptor.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename UnaryFunction, typename OutputIterator>
+  class transform_output_iterator;
+
+namespace detail
+{
+
+// Proxy reference that uses Unary Function to transform the rhs of assigment
+// operator before writing the result to OutputIterator
+template <typename UnaryFunction, typename OutputIterator>
+  class transform_output_iterator_proxy
+{
+  public:
+    __host__ __device__
+    transform_output_iterator_proxy(const OutputIterator& out, UnaryFunction fun) : out(out), fun(fun)
+    {
+    }
+
+    __thrust_exec_check_disable__
+    template <typename T>
+    __host__ __device__
+    transform_output_iterator_proxy operator=(const T& x)
+    {
+      *out = fun(x);
+      return *this;
+    }
+
+  private:
+    OutputIterator out;
+    UnaryFunction fun;
+};
+
+// Compute the iterator_adaptor instantiation to be used for transform_output_iterator
+template <typename UnaryFunction, typename OutputIterator>
+struct transform_output_iterator_base
+{
+    typedef thrust::iterator_adaptor
+    <
+        transform_output_iterator<UnaryFunction, OutputIterator>
+      , OutputIterator
+      , thrust::use_default
+      , thrust::use_default
+      , thrust::use_default
+      , transform_output_iterator_proxy<UnaryFunction, OutputIterator>
+    > type;
+};
+
+// Register transform_output_iterator_proxy with 'is_proxy_reference' from
+// type_traits to enable its use with algorithms.
+template <class UnaryFunction, class OutputIterator>
+struct is_proxy_reference<
+    transform_output_iterator_proxy<UnaryFunction, OutputIterator> >
+    : public thrust::detail::true_type {};
+
+} // end detail
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tuple_of_iterator_references.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tuple_of_iterator_references.h
new file mode 100644
index 000000000000..78c5e8a28f49
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/tuple_of_iterator_references.h
@@ -0,0 +1,146 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/tuple.h>
+#include <thrust/pair.h>
+#include <thrust/detail/reference_forward_declaration.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+  
+template<
+  typename... Ts
+>
+  class tuple_of_iterator_references
+    : public thrust::tuple<Ts...>
+{
+  private:
+    typedef thrust::tuple<Ts...> super_t;
+
+  public:
+    // allow implicit construction from tuple<refs>
+    inline __host__ __device__
+    tuple_of_iterator_references(const super_t &other)
+      : super_t(other)
+    {}
+
+    // allow assignment from tuples
+    // XXX might be worthwhile to guard this with an enable_if is_assignable
+    __thrust_exec_check_disable__
+    template<typename... Us>
+    inline __host__ __device__
+    tuple_of_iterator_references &operator=(const thrust::tuple<Us...> &other)
+    {
+      super_t::operator=(other);
+      return *this;
+    }
+
+    // allow assignment from pairs
+    // XXX might be worthwhile to guard this with an enable_if is_assignable
+    __thrust_exec_check_disable__
+    template<typename U1, typename U2>
+    inline __host__ __device__
+    tuple_of_iterator_references &operator=(const thrust::pair<U1,U2> &other)
+    {
+      super_t::operator=(other);
+      return *this;
+    }
+
+    // allow assignment from reference<tuple>
+    // XXX perhaps we should generalize to reference<T>
+    //     we could captures reference<pair> this way
+    __thrust_exec_check_disable__
+    template<typename Pointer, typename Derived,
+             typename... Us>
+    inline __host__ __device__
+// XXX gcc-4.2 crashes on is_assignable
+//    typename thrust::detail::enable_if<
+//      thrust::detail::is_assignable<
+//        super_t,
+//        const thrust::tuple<Us...>
+//      >::value,
+//      tuple_of_iterator_references &
+//    >::type
+    tuple_of_iterator_references &
+    operator=(const thrust::reference<thrust::tuple<Us...>, Pointer, Derived> &other)
+    {
+      typedef thrust::tuple<Us...> tuple_type;
+
+      // XXX perhaps this could be accelerated
+      tuple_type other_tuple = other;
+      super_t::operator=(other_tuple);
+      return *this;
+    }
+
+
+    // duplicate thrust::tuple's constructors
+    inline __host__ __device__
+    tuple_of_iterator_references() {}
+
+    inline __host__ __device__
+    tuple_of_iterator_references(typename access_traits<Ts>::parameter_type... ts)
+      : super_t(ts...)
+    {}
+};
+
+
+// this overload of swap() permits swapping tuple_of_iterator_references returned as temporaries from
+// iterator dereferences
+template<
+  typename... Ts,
+  typename... Us
+>
+inline __host__ __device__
+void swap(tuple_of_iterator_references<Ts...> x,
+          tuple_of_iterator_references<Us...> y)
+{
+  x.swap(y);
+}
+
+
+} // end detail
+
+// define tuple_size, tuple_element, etc.
+template<class... Ts>
+struct tuple_size<detail::tuple_of_iterator_references<Ts...>>
+  : std::integral_constant<size_t, sizeof...(Ts)>
+{};
+
+template<size_t i>
+struct tuple_element<i, detail::tuple_of_iterator_references<>> {};
+
+
+template<class T, class... Ts>
+struct tuple_element<0, detail::tuple_of_iterator_references<T,Ts...>>
+{
+  using type = T;
+};
+
+
+template<size_t i, class T, class... Ts>
+struct tuple_element<i, detail::tuple_of_iterator_references<T,Ts...>>
+{
+  using type = typename tuple_element<i - 1, detail::tuple_of_iterator_references<Ts...>>::type;
+};
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/universal_categories.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/universal_categories.h
new file mode 100644
index 000000000000..d2abd7f55bd7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/universal_categories.h
@@ -0,0 +1,86 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_categories.h>
+
+// XXX eliminate this file
+
+THRUST_NAMESPACE_BEGIN
+
+// define these types without inheritance to avoid ambiguous conversion to base classes
+
+struct input_universal_iterator_tag
+{
+  operator input_host_iterator_tag () {return input_host_iterator_tag();}
+
+  operator input_device_iterator_tag () {return input_device_iterator_tag();}
+};
+
+struct output_universal_iterator_tag
+{
+  operator output_host_iterator_tag () {return output_host_iterator_tag();}
+
+  operator output_device_iterator_tag () {return output_device_iterator_tag();}
+};
+
+struct forward_universal_iterator_tag
+  : input_universal_iterator_tag
+{
+  operator forward_host_iterator_tag () {return forward_host_iterator_tag();};
+
+  operator forward_device_iterator_tag () {return forward_device_iterator_tag();};
+};
+
+struct bidirectional_universal_iterator_tag
+  : forward_universal_iterator_tag
+{
+  operator bidirectional_host_iterator_tag () {return bidirectional_host_iterator_tag();};
+
+  operator bidirectional_device_iterator_tag () {return bidirectional_device_iterator_tag();};
+};
+
+
+namespace detail
+{
+
+// create this struct to control conversion precedence in random_access_universal_iterator_tag
+template<typename T>
+struct one_degree_of_separation
+  : T
+{
+};
+
+} // end detail
+
+
+struct random_access_universal_iterator_tag
+{
+  // these conversions are all P0
+  operator random_access_host_iterator_tag () {return random_access_host_iterator_tag();};
+
+  operator random_access_device_iterator_tag () {return random_access_device_iterator_tag();};
+
+  // bidirectional_universal_iterator_tag is P1
+  operator detail::one_degree_of_separation<bidirectional_universal_iterator_tag> () {return detail::one_degree_of_separation<bidirectional_universal_iterator_tag>();}
+
+};
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator.inl b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator.inl
new file mode 100644
index 000000000000..a2bc98afec2e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator.inl
@@ -0,0 +1,152 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/tuple_transform.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+template<typename IteratorTuple>
+__host__ __device__
+  zip_iterator<IteratorTuple>
+    ::zip_iterator()
+{
+} // end zip_iterator::zip_iterator()
+
+
+template<typename IteratorTuple>
+__host__ __device__
+  zip_iterator<IteratorTuple>
+    ::zip_iterator(IteratorTuple iterator_tuple)
+      :m_iterator_tuple(iterator_tuple)
+{
+} // end zip_iterator::zip_iterator()
+
+
+template<typename IteratorTuple>
+  template<typename OtherIteratorTuple>
+  __host__ __device__
+    zip_iterator<IteratorTuple>
+      ::zip_iterator(const zip_iterator<OtherIteratorTuple> &other,
+                     typename thrust::detail::enable_if_convertible<
+                       OtherIteratorTuple,
+                       IteratorTuple
+                     >::type *)
+        :m_iterator_tuple(other.get_iterator_tuple())
+{
+} // end zip_iterator::zip_iterator()
+
+
+template<typename IteratorTuple>
+__host__ __device__
+const IteratorTuple &zip_iterator<IteratorTuple>
+  ::get_iterator_tuple() const
+{
+  return m_iterator_tuple;
+} // end zip_iterator::get_iterator_tuple()
+
+
+template<typename IteratorTuple>
+  typename zip_iterator<IteratorTuple>::super_t::reference
+  __host__ __device__
+    zip_iterator<IteratorTuple>
+      ::dereference() const
+{
+  using namespace detail::tuple_impl_specific;
+
+  return thrust::detail::tuple_host_device_transform<
+    detail::dereference_iterator::template apply
+  >(get_iterator_tuple(), detail::dereference_iterator());
+} // end zip_iterator::dereference()
+
+
+__thrust_exec_check_disable__
+template<typename IteratorTuple>
+  template<typename OtherIteratorTuple>
+  __host__ __device__
+    bool zip_iterator<IteratorTuple>
+      ::equal(const zip_iterator<OtherIteratorTuple> &other) const
+{
+  return get<0>(get_iterator_tuple()) == get<0>(other.get_iterator_tuple());
+} // end zip_iterator::equal()
+
+
+template<typename IteratorTuple>
+__host__ __device__
+  void zip_iterator<IteratorTuple>
+    ::advance(typename super_t::difference_type n)
+{
+  using namespace detail::tuple_impl_specific;
+  tuple_for_each(m_iterator_tuple,
+                 detail::advance_iterator<typename super_t::difference_type>(n));
+} // end zip_iterator::advance()
+
+
+template<typename IteratorTuple>
+__host__ __device__
+  void zip_iterator<IteratorTuple>
+    ::increment()
+{
+  using namespace detail::tuple_impl_specific;
+  tuple_for_each(m_iterator_tuple, detail::increment_iterator());
+} // end zip_iterator::increment()
+
+
+template<typename IteratorTuple>
+__host__ __device__
+  void zip_iterator<IteratorTuple>
+    ::decrement()
+{
+  using namespace detail::tuple_impl_specific;
+  tuple_for_each(m_iterator_tuple, detail::decrement_iterator());
+} // end zip_iterator::decrement()
+
+
+__thrust_exec_check_disable__
+template<typename IteratorTuple>
+  template <typename OtherIteratorTuple>
+  __host__ __device__
+    typename zip_iterator<IteratorTuple>::super_t::difference_type
+      zip_iterator<IteratorTuple>
+        ::distance_to(const zip_iterator<OtherIteratorTuple> &other) const
+{
+  return get<0>(other.get_iterator_tuple()) - get<0>(get_iterator_tuple());
+} // end zip_iterator::distance_to()
+
+
+template<typename... Iterators>
+__host__ __device__
+  zip_iterator<thrust::tuple<Iterators...>> make_zip_iterator(thrust::tuple<Iterators...> t)
+{
+  return zip_iterator<thrust::tuple<Iterators...>>(t);
+} // end make_zip_iterator()
+
+
+template<typename... Iterators>
+__host__ __device__
+  zip_iterator<thrust::tuple<Iterators...>> make_zip_iterator(Iterators... its)
+{
+  return make_zip_iterator(thrust::make_tuple(its...));
+} // end make_zip_iterator()
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator_base.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator_base.h
new file mode 100644
index 000000000000..030153b65eae
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/detail/zip_iterator_base.h
@@ -0,0 +1,353 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/advance.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/iterator/iterator_categories.h>
+#include <thrust/iterator/detail/minimum_category.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/type_traits/integer_sequence.h>
+#include <thrust/tuple.h>
+#include <thrust/detail/tuple_meta_transform.h>
+#include <thrust/detail/tuple_transform.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/tuple_of_iterator_references.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declare zip_iterator for zip_iterator_base
+template<typename IteratorTuple> class zip_iterator;
+
+namespace detail
+{
+
+
+// Functors to be used with tuple algorithms
+//
+template<typename DiffType>
+class advance_iterator
+{
+public:
+  inline __host__ __device__
+  advance_iterator(DiffType step) : m_step(step) {}
+
+  __thrust_exec_check_disable__
+  template<typename Iterator>
+  inline __host__ __device__
+  void operator()(Iterator& it) const
+  { thrust::advance(it, m_step); }
+
+private:
+  DiffType m_step;
+}; // end advance_iterator
+
+
+struct increment_iterator
+{
+  __thrust_exec_check_disable__
+  template<typename Iterator>
+  inline __host__ __device__
+  void operator()(Iterator& it)
+  { ++it; }
+}; // end increment_iterator
+
+
+struct decrement_iterator
+{
+  __thrust_exec_check_disable__
+  template<typename Iterator>
+  inline __host__ __device__
+  void operator()(Iterator& it)
+  { --it; }
+}; // end decrement_iterator
+
+
+struct dereference_iterator
+{
+  template<typename Iterator>
+  struct apply
+  { 
+    typedef typename
+      iterator_traits<Iterator>::reference
+    type;
+  }; // end apply
+
+  // XXX silence warnings of the form "calling a __host__ function from a __host__ __device__ function is not allowed
+  __thrust_exec_check_disable__
+  template<typename Iterator>
+  __host__ __device__
+    typename apply<Iterator>::type operator()(Iterator const& it)
+  {
+    return *it;
+  }
+}; // end dereference_iterator
+
+
+// The namespace tuple_impl_specific provides two meta-
+// algorithms and two algorithms for tuples.
+namespace tuple_impl_specific
+{
+
+// define apply1 for tuple_meta_transform_impl
+template<typename UnaryMetaFunctionClass, class Arg>
+  struct apply1
+    : UnaryMetaFunctionClass::template apply<Arg>
+{
+}; // end apply1
+
+
+// define apply2 for tuple_meta_accumulate_impl
+template<typename UnaryMetaFunctionClass, class Arg1, class Arg2>
+  struct apply2
+    : UnaryMetaFunctionClass::template apply<Arg1,Arg2>
+{
+}; // end apply2
+
+
+// Meta-accumulate algorithm for tuples. Note: The template 
+// parameter StartType corresponds to the initial value in 
+// ordinary accumulation.
+//
+template<class Tuple, class BinaryMetaFun, class StartType>
+  struct tuple_meta_accumulate;
+
+template<
+    class BinaryMetaFun
+  , typename StartType
+>
+  struct tuple_meta_accumulate<thrust::tuple<>,BinaryMetaFun,StartType>
+{
+   typedef typename thrust::detail::identity_<StartType>::type type;
+};
+
+
+template<
+    class BinaryMetaFun
+  , typename StartType
+  , typename    T
+  , typename... Ts
+>
+  struct tuple_meta_accumulate<thrust::tuple<T,Ts...>,BinaryMetaFun,StartType>
+{
+   typedef typename apply2<
+       BinaryMetaFun
+     , T
+     , typename tuple_meta_accumulate<
+           thrust::tuple<Ts...>
+         , BinaryMetaFun
+         , StartType 
+       >::type
+   >::type type;
+};
+
+
+template<typename Fun>
+inline __host__ __device__
+Fun tuple_for_each_helper(Fun f)
+{
+  return f;
+}
+
+template<typename Fun, typename T, typename... Ts>
+inline __host__ __device__
+Fun tuple_for_each_helper(Fun f, T& t, Ts&... ts)
+{
+  f(t);
+  return tuple_for_each_helper(f, ts...);
+}
+
+// for_each algorithm for tuples.
+
+template<typename Fun, typename... Ts, size_t... Is>
+inline __host__ __device__
+Fun tuple_for_each(thrust::tuple<Ts...>& t, Fun f, thrust::index_sequence<Is...>)
+{
+  return tuple_for_each_helper(f, thrust::get<Is>(t)...);
+} // end tuple_for_each()
+
+// for_each algorithm for tuples.
+template<typename Fun, typename... Ts>
+inline __host__ __device__
+Fun tuple_for_each(thrust::tuple<Ts...>& t, Fun f)
+{ 
+  return tuple_for_each(t, f, thrust::make_index_sequence<thrust::tuple_size<thrust::tuple<Ts...>>::value>{});
+}
+
+
+} // end tuple_impl_specific
+
+
+// Metafunction to obtain the type of the tuple whose element types
+// are the value_types of an iterator tupel.
+//
+template<typename IteratorTuple>
+  struct tuple_of_value_types
+    : tuple_meta_transform<
+          IteratorTuple,
+          iterator_value
+        >
+{
+}; // end tuple_of_value_types
+
+
+struct minimum_category_lambda
+{
+  template<typename T1, typename T2>
+    struct apply : minimum_category<T1,T2>
+  {};
+};
+
+
+
+// Metafunction to obtain the minimal traversal tag in a tuple
+// of iterators.
+//
+template<typename IteratorTuple>
+struct minimum_traversal_category_in_iterator_tuple
+{
+  typedef typename tuple_meta_transform<
+      IteratorTuple
+    , thrust::iterator_traversal
+  >::type tuple_of_traversal_tags;
+      
+  typedef typename tuple_impl_specific::tuple_meta_accumulate<
+      tuple_of_traversal_tags
+    , minimum_category_lambda
+    , thrust::random_access_traversal_tag
+  >::type type;
+};
+
+
+struct minimum_system_lambda
+{
+  template<typename T1, typename T2>
+    struct apply : minimum_system<T1,T2>
+  {};
+};
+
+
+
+// Metafunction to obtain the minimal system tag in a tuple
+// of iterators.
+template<typename IteratorTuple>
+struct minimum_system_in_iterator_tuple
+{
+  typedef typename thrust::detail::tuple_meta_transform<
+    IteratorTuple,
+    thrust::iterator_system
+  >::type tuple_of_system_tags;
+
+  typedef typename tuple_impl_specific::tuple_meta_accumulate<
+    tuple_of_system_tags,
+    minimum_system_lambda,
+    thrust::any_system_tag
+  >::type type;
+};
+
+namespace zip_iterator_base_ns
+{
+
+
+template<typename Tuple, typename IndexSequence>
+  struct tuple_of_iterator_references_helper;
+
+
+template<typename Tuple, size_t... Is>
+  struct tuple_of_iterator_references_helper<Tuple, thrust::index_sequence<Is...>>
+{
+  typedef thrust::detail::tuple_of_iterator_references<
+    typename thrust::tuple_element<Is,Tuple>::type...
+  > type;
+};
+
+
+template<typename IteratorTuple>
+  struct tuple_of_iterator_references
+{
+  // get a thrust::tuple of the iterators' references
+  typedef typename tuple_meta_transform<
+    IteratorTuple,
+    iterator_reference
+  >::type tuple_of_references;
+
+  // map thrust::tuple<T...> to tuple_of_iterator_references<T...>
+  typedef typename tuple_of_iterator_references_helper<
+    tuple_of_references,
+    thrust::make_index_sequence<thrust::tuple_size<tuple_of_references>::value>
+  >::type type;
+};
+
+
+} // end zip_iterator_base_ns
+
+///////////////////////////////////////////////////////////////////
+//
+// Class zip_iterator_base
+//
+// Builds and exposes the iterator facade type from which the zip 
+// iterator will be derived.
+//
+template<typename IteratorTuple>
+  struct zip_iterator_base
+{
+ //private:
+    // reference type is the type of the tuple obtained from the
+    // iterators' reference types.
+    typedef typename zip_iterator_base_ns::tuple_of_iterator_references<IteratorTuple>::type reference;
+
+    // Boost's Value type is the same as reference type.
+    //typedef reference value_type;
+    typedef typename tuple_of_value_types<IteratorTuple>::type value_type;
+
+    // Difference type is the first iterator's difference type
+    typedef typename thrust::iterator_traits<
+      typename thrust::tuple_element<0, IteratorTuple>::type
+    >::difference_type difference_type;
+
+    // Iterator system is the minimum system tag in the
+    // iterator tuple
+    typedef typename
+    minimum_system_in_iterator_tuple<IteratorTuple>::type system;
+
+    // Traversal category is the minimum traversal category in the
+    // iterator tuple
+    typedef typename
+    minimum_traversal_category_in_iterator_tuple<IteratorTuple>::type traversal_category;
+  
+ public:
+  
+    // The iterator facade type from which the zip iterator will
+    // be derived.
+    typedef thrust::iterator_facade<
+        zip_iterator<IteratorTuple>,
+        value_type,  
+        system,
+        traversal_category,
+        reference,
+        difference_type
+    > type;
+}; // end zip_iterator_base
+
+} // end detail
+
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/discard_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/discard_iterator.h
new file mode 100644
index 000000000000..eb5156edac4a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/discard_iterator.h
@@ -0,0 +1,174 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/discard_iterator.h
+ *  \brief An iterator which "discards" (ignores) values assigned to it upon dereference
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/discard_iterator_base.h>
+#include <thrust/iterator/iterator_facade.h>
+
+THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_BEGIN
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p discard_iterator is an iterator which represents a special kind of pointer that
+ *  ignores values written to it upon dereference. This iterator is useful for ignoring
+ *  the output of certain algorithms without wasting memory capacity or bandwidth.
+ *  \p discard_iterator may also be used to count the size of an algorithm's output which
+ *  may not be known a priori.
+ *
+ *  The following code snippet demonstrates how to use \p discard_iterator to ignore
+ *  ignore one of the output ranges of reduce_by_key
+ *
+ *  \code
+ *  #include <thrust/iterator/discard_iterator.h>
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  int main()
+ *  {
+ *    thrust::device_vector<int> keys(7), values(7);
+ *
+ *    keys[0] = 1;
+ *    keys[1] = 3;
+ *    keys[2] = 3;
+ *    keys[3] = 3;
+ *    keys[4] = 2;
+ *    keys[5] = 2;
+ *    keys[6] = 1;
+ *
+ *    values[0] = 9;
+ *    values[1] = 8;
+ *    values[2] = 7;
+ *    values[3] = 6;
+ *    values[4] = 5;
+ *    values[5] = 4;
+ *    values[6] = 3;
+ *
+ *    thrust::device_vector<int> result(4);
+ *
+ *    // we are only interested in the reduced values
+ *    // use discard_iterator to ignore the output keys
+ *    thrust::reduce_by_key(keys.begin(), keys.end(),
+ *                          values.begin(),
+ *                          thrust::make_discard_iterator(),
+ *                          result.begin());
+ *
+ *    // result is now [9, 21, 9, 3]
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see make_discard_iterator
+ */
+template<typename System = use_default>
+  class discard_iterator
+    : public detail::discard_iterator_base<System>::type
+{
+    /*! \cond
+     */
+    friend class thrust::iterator_core_access;
+    typedef typename detail::discard_iterator_base<System>::type          super_t;
+    typedef typename detail::discard_iterator_base<System>::incrementable incrementable;
+    typedef typename detail::discard_iterator_base<System>::base_iterator base_iterator;
+
+  public:
+    typedef typename super_t::reference  reference;
+    typedef typename super_t::value_type value_type;
+
+    /*! \endcond
+     */
+
+    /*! Copy constructor copies from a source discard_iterator.
+     *
+     *  \p rhs The discard_iterator to copy.
+     */
+    __host__ __device__
+    discard_iterator(discard_iterator const &rhs)
+      : super_t(rhs.base()) {}
+
+#if THRUST_CPP_DIALECT >= 2011
+    discard_iterator & operator=(const discard_iterator &) = default;
+#endif
+
+    /*! This constructor receives an optional index specifying the position of this
+     *  \p discard_iterator in a range.
+     *
+     *  \p i The index of this \p discard_iterator in a range. Defaults to the
+     *       value returned by \c Incrementable's null constructor. For example,
+     *       when <tt>Incrementable == int</tt>, \c 0.
+     */
+    __host__ __device__
+    discard_iterator(incrementable const &i = incrementable())
+      : super_t(base_iterator(i)) {}
+
+    /*! \cond
+     */
+
+  private: // Core iterator interface
+    __host__ __device__
+    reference dereference() const
+    {
+      return m_element;
+    }
+
+    mutable value_type m_element;
+
+    /*! \endcond
+     */
+}; // end constant_iterator
+
+
+/*! \p make_discard_iterator creates a \p discard_iterator from an optional index parameter.
+ *
+ *  \param i The index of the returned \p discard_iterator within a range.
+ *           In the default case, the value of this parameter is \c 0.
+ *
+ *  \return A new \p discard_iterator with index as given by \p i.
+ *
+ *  \see constant_iterator
+ */
+inline __host__ __device__
+discard_iterator<> make_discard_iterator(discard_iterator<>::difference_type i = discard_iterator<>::difference_type(0))
+{
+  return discard_iterator<>(i);
+} // end make_discard_iterator()
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
+THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_adaptor.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_adaptor.h
new file mode 100644
index 000000000000..67d4866b9a55
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_adaptor.h
@@ -0,0 +1,242 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/iterator_adaptor.h
+ *  \brief An iterator which adapts a base iterator
+ */
+
+/*
+ * (C) Copyright David Abrahams 2002.
+ * (C) Copyright Jeremy Siek    2002.
+ * (C) Copyright Thomas Witt    2002.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/detail/use_default.h>
+#include <thrust/iterator/detail/iterator_adaptor_base.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p iterator_adaptor is an iterator which adapts an existing type of iterator to create a new type of
+ *  iterator. Most of Thrust's fancy iterators are defined via inheritance from \p iterator_adaptor.
+ *  While composition of these existing Thrust iterators is often sufficient for expressing the desired
+ *  functionality, it is occasionally more straightforward to derive from \p iterator_adaptor directly.
+ *
+ *  To see how to use \p iterator_adaptor to create a novel iterator type, let's examine how to use it to
+ *  define \p repeat_iterator, a fancy iterator which repeats elements from another range a given number of time:
+ *
+ *  \code
+ *  #include <thrust/iterator/iterator_adaptor.h>
+ *
+ *  // derive repeat_iterator from iterator_adaptor
+ *  template<typename Iterator>
+ *    class repeat_iterator
+ *      : public thrust::iterator_adaptor<
+ *          repeat_iterator<Iterator>, // the first template parameter is the name of the iterator we're creating
+ *          Iterator                   // the second template parameter is the name of the iterator we're adapting
+ *                                     // we can use the default for the additional template parameters
+ *        >
+ *  {
+ *    public:
+ *      // shorthand for the name of the iterator_adaptor we're deriving from
+ *      typedef thrust::iterator_adaptor<
+ *        repeat_iterator<Iterator>,
+ *        Iterator
+ *      > super_t;
+ *
+ *      __host__ __device__
+ *      repeat_iterator(const Iterator &x, int n) : super_t(x), begin(x), n(n) {}
+ *
+ *      // befriend thrust::iterator_core_access to allow it access to the private interface below
+ *      friend class thrust::iterator_core_access;
+ *
+ *    private:
+ *      // repeat each element of the adapted range n times
+ *      unsigned int n;
+ *
+ *      // used to keep track of where we began
+ *      const Iterator begin;
+ *
+ *      // it is private because only thrust::iterator_core_access needs access to it
+ *      __host__ __device__
+ *      typename super_t::reference dereference() const
+ *      {
+ *        return *(begin + (this->base() - begin) / n);
+ *      }
+ *  };
+ *  \endcode
+ *
+ *  Except for the first two, \p iterator_adaptor's template parameters are optional. When omitted, or when the
+ *  user specifies \p thrust::use_default in its place, \p iterator_adaptor will use a default type inferred from \p Base.
+ *
+ *  \p iterator_adaptor's functionality is derived from and generally equivalent to \p boost::iterator_adaptor.
+ *  The exception is Thrust's addition of the template parameter \p System, which is necessary to allow Thrust
+ *  to dispatch an algorithm to one of several parallel backend systems.
+ *
+ *  \p iterator_adaptor is a powerful tool for creating custom iterators directly. However, the large set of iterator semantics which must be satisfied
+ *  for algorithm compatibility can make \p iterator_adaptor difficult to use correctly. Unless you require the full expressivity of \p iterator_adaptor,
+ *  consider building a custom iterator through composition of existing higher-level fancy iterators instead. 
+ *
+ *  Interested users may refer to <tt>boost::iterator_adaptor</tt>'s documentation for further usage examples.
+ */
+template<typename Derived,
+         typename Base,
+         typename Value      = use_default,
+         typename System     = use_default,
+         typename Traversal  = use_default,
+         typename Reference  = use_default,
+         typename Difference = use_default>
+  class iterator_adaptor:
+    public detail::iterator_adaptor_base<
+      Derived, Base, Value, System, Traversal, Reference, Difference
+    >::type
+{
+  /*! \cond
+   */
+
+    friend class thrust::iterator_core_access;
+
+  protected:
+    typedef typename detail::iterator_adaptor_base<
+        Derived, Base, Value, System, Traversal, Reference, Difference
+    >::type super_t;
+
+  /*! \endcond
+   */
+  
+  public:
+    /*! \p iterator_adaptor's default constructor does nothing.
+     */
+    __host__ __device__
+    iterator_adaptor(){}
+
+    /*! This constructor copies from a given instance of the \p Base iterator.
+     */
+    __thrust_exec_check_disable__
+    __host__ __device__
+    explicit iterator_adaptor(Base const& iter)
+      : m_iterator(iter)
+    {}
+
+    /*! The type of iterator this \p iterator_adaptor's \p adapts.
+     */
+    typedef Base       base_type;
+                                                                                              
+    /*! \cond
+     */
+    typedef typename super_t::reference reference;
+                                                                                              
+    typedef typename super_t::difference_type difference_type;
+    /*! \endcond
+     */
+
+    /*! \return A \p const reference to the \p Base iterator this \p iterator_adaptor adapts.
+     */
+    __host__ __device__
+    Base const& base() const
+    { return m_iterator; }
+
+  protected:
+    /*! \return A \p const reference to the \p Base iterator this \p iterator_adaptor adapts.
+     */
+    __host__ __device__
+    Base const& base_reference() const
+    { return m_iterator; }
+
+    /*! \return A mutable reference to the \p Base iterator this \p iterator_adaptor adapts.
+     */
+    __host__ __device__
+    Base& base_reference()
+    { return m_iterator; }
+
+    /*! \cond
+     */
+  private: // Core iterator interface for iterator_facade
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    typename iterator_adaptor::reference dereference() const
+    { return *m_iterator; }
+
+    __thrust_exec_check_disable__
+    template<typename OtherDerived, typename OtherIterator, typename V, typename S, typename T, typename R, typename D>
+    __host__ __device__
+    bool equal(iterator_adaptor<OtherDerived, OtherIterator, V, S, T, R, D> const& x) const
+    { return m_iterator == x.base(); }
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    void advance(typename iterator_adaptor::difference_type n)
+    {
+      // XXX statically assert on random_access_traversal_tag
+
+      // counting_iterator will pick eg. diff_t=int64 when base=int32.
+      // Explicitly cast to avoid static conversion warnings.
+      m_iterator = static_cast<base_type>(m_iterator + n);
+    }
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    void increment()
+    { ++m_iterator; }
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    void decrement()
+    {
+      // XXX statically assert on bidirectional_traversal_tag
+      --m_iterator;
+    }
+
+    __thrust_exec_check_disable__
+    template<typename OtherDerived, typename OtherIterator, typename V, typename S, typename T, typename R, typename D>
+    __host__ __device__
+    typename iterator_adaptor::difference_type distance_to(iterator_adaptor<OtherDerived, OtherIterator, V, S, T, R, D> const& y) const
+    { return y.base() - m_iterator; }
+
+  private:
+    Base m_iterator;
+
+    /*! \endcond
+     */
+}; // end iterator_adaptor
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_categories.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_categories.h
new file mode 100644
index 000000000000..9a6f3f4ae524
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_categories.h
@@ -0,0 +1,223 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/iterator_categories.h
+ *  \brief Types for reasoning about the categories of iterators
+ */
+
+/*
+ * (C) Copyright Jeremy Siek 2002.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/iterator_category_with_system_and_traversal.h>
+#include <thrust/iterator/detail/iterator_traversal_tags.h>
+#include <thrust/iterator/detail/device_system_tag.h>
+
+// #include this for stl's iterator tags
+#include <iterator>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \addtogroup iterator_tags Iterator Tags
+ *  \ingroup iterators
+ *  \addtogroup iterator_tag_classes Iterator Tag Classes
+ *  \ingroup iterator_tags
+ *  \{
+ */
+
+/*! \p input_device_iterator_tag is an empty class: it has no member functions,
+ *  member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Input Device Iterator concept within the C++ type
+ *  system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags  iterator_traits,
+ *  output_device_iterator_tag, forward_device_iterator_tag,
+ *  bidirectional_device_iterator_tag, random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+struct input_device_iterator_tag
+  : thrust::detail::iterator_category_with_system_and_traversal<
+      std::input_iterator_tag,
+      thrust::device_system_tag,
+      thrust::single_pass_traversal_tag
+    >
+{};
+
+/*! \p output_device_iterator_tag is an empty class: it has no member functions,
+ *  member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Output Device Iterator concept within the C++ type
+ *  system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags  iterator_traits,
+ *  input_device_iterator_tag, forward_device_iterator_tag,
+ *  bidirectional_device_iterator_tag, random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+struct output_device_iterator_tag
+  : thrust::detail::iterator_category_with_system_and_traversal<
+      std::output_iterator_tag,
+      thrust::device_system_tag,
+      thrust::single_pass_traversal_tag
+    >
+{};
+
+/*! \p forward_device_iterator_tag is an empty class: it has no member functions,
+ *  member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Forward Device Iterator concept within the C++ type
+ *  system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags  iterator_traits,
+ *  input_device_iterator_tag, output_device_iterator_tag,
+ *  bidirectional_device_iterator_tag, random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+struct forward_device_iterator_tag
+  : thrust::detail::iterator_category_with_system_and_traversal<
+      std::forward_iterator_tag,
+      thrust::device_system_tag,
+      thrust::forward_traversal_tag
+    >
+{};
+
+/*! \p bidirectional_device_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Bidirectional Device Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+struct bidirectional_device_iterator_tag
+  : thrust::detail::iterator_category_with_system_and_traversal<
+      std::bidirectional_iterator_tag,
+      thrust::device_system_tag,
+      thrust::bidirectional_traversal_tag
+    >
+{};
+
+/*! \p random_access_device_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Random Access Device Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+struct random_access_device_iterator_tag
+  : thrust::detail::iterator_category_with_system_and_traversal<
+      std::random_access_iterator_tag,
+      thrust::device_system_tag,
+      thrust::random_access_traversal_tag
+    >
+{};
+
+/*! \p input_host_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Input Host Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
+ *  random_access_device_iterator_tag,
+ *  output_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+typedef std::input_iterator_tag input_host_iterator_tag;
+
+/*! \p output_host_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Output Host Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
+ *  random_access_device_iterator_tag,
+ *  input_host_iterator_tag, forward_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+typedef std::output_iterator_tag output_host_iterator_tag;
+
+/*! \p forward_host_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Forward Host Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
+ *  random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag,
+ *  bidirectional_host_iterator_tag, random_access_host_iterator_tag
+ */
+typedef std::forward_iterator_tag forward_host_iterator_tag;
+
+/*! \p bidirectional_host_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Forward Host Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
+ *  random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag,
+ *  forward_host_iterator_tag, random_access_host_iterator_tag
+ */
+typedef std::bidirectional_iterator_tag bidirectional_host_iterator_tag;
+
+/*! \p random_access_host_iterator_tag is an empty class: it has no member
+ *  functions, member variables, or nested types. It is used solely as a "tag": a
+ *  representation of the Forward Host Iterator concept within the C++
+ *  type system.
+ *
+ *  \see https://en.cppreference.com/w/cpp/iterator/iterator_tags 
+ *  iterator_traits, input_device_iterator_tag, output_device_iterator_tag,
+ *  forward_device_iterator_tag, bidirectional_device_iterator_tag,
+ *  random_access_device_iterator_tag,
+ *  input_host_iterator_tag, output_host_iterator_tag,
+ *  forward_host_iterator_tag, bidirectional_host_iterator_tag
+ */
+typedef std::random_access_iterator_tag random_access_host_iterator_tag;
+
+/*! \} // end iterator_tag_classes
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/iterator/detail/universal_categories.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_facade.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_facade.h
new file mode 100644
index 000000000000..f6920c5c85c0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_facade.h
@@ -0,0 +1,542 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/iterator/iterator_facade.h
+ *  \brief A class which exposes a public interface for iterators
+ */
+
+/*
+ * (C) Copyright David Abrahams 2002.
+ * (C) Copyright Jeremy Siek    2002.
+ * (C) Copyright Thomas Witt    2002.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/iterator_facade_category.h>
+#include <thrust/iterator/detail/distance_from_result.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+
+// This forward declaration is required for the friend declaration
+// in iterator_core_access
+template<typename Derived, typename Value, typename System, typename Traversal, typename Reference, typename Difference> class iterator_facade;
+
+
+/*! \p iterator_core_access is the class which user iterator types derived from \p thrust::iterator_adaptor
+ *  or \p thrust::iterator_facade must befriend to allow it to access their private interface.
+ */
+class iterator_core_access
+{
+    /*! \cond
+     */
+
+    // declare our friends
+    template<typename Derived, typename Value, typename System, typename Traversal, typename Reference, typename Difference> friend class iterator_facade;
+
+    // iterator comparisons are our friends
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend bool
+    operator ==(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+                iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend bool
+    operator !=(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+                iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend bool
+    operator <(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+               iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend bool
+    operator >(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+               iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend bool
+    operator <=(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+                iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend bool
+    operator >=(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+                iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    // iterator difference is our friend
+    template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+              typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+    inline __host__ __device__
+    friend
+      typename thrust::detail::distance_from_result<
+        iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1>,
+        iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2>
+      >::type
+    operator-(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+              iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs);
+
+    template<typename Facade>
+    __host__ __device__
+    static typename Facade::reference dereference(Facade const& f)
+    {
+      return f.dereference();
+    }
+
+    template<typename Facade>
+    __host__ __device__
+    static void increment(Facade& f)
+    {
+      f.increment();
+    }
+
+    template<typename Facade>
+    __host__ __device__
+    static void decrement(Facade& f)
+    {
+      f.decrement();
+    }
+
+    template <class Facade1, class Facade2>
+    __host__ __device__
+    static bool equal(Facade1 const& f1, Facade2 const& f2)
+    {
+      return f1.equal(f2);
+    }
+
+    // XXX TODO: Investigate whether we need both of these cases
+    //template <class Facade1, class Facade2>
+    //__host__ __device__
+    //static bool equal(Facade1 const& f1, Facade2 const& f2, mpl::true_)
+    //{
+    //  return f1.equal(f2);
+    //}
+
+    //template <class Facade1, class Facade2>
+    //__host__ __device__
+    //static bool equal(Facade1 const& f1, Facade2 const& f2, mpl::false_)
+    //{
+    //  return f2.equal(f1);
+    //}
+
+    template <class Facade>
+    __host__ __device__
+    static void advance(Facade& f, typename Facade::difference_type n)
+    {
+      f.advance(n);
+    }
+
+    // Facade2 is convertible to Facade1,
+    // so return Facade1's difference_type
+    template <class Facade1, class Facade2>
+    __host__ __device__
+    static typename Facade1::difference_type
+      distance_from(Facade1 const& f1, Facade2 const& f2, thrust::detail::true_type)
+    {
+      return -f1.distance_to(f2);
+    }
+
+    // Facade2 is not convertible to Facade1,
+    // so return Facade2's difference_type
+    template <class Facade1, class Facade2>
+    __host__ __device__
+    static typename Facade2::difference_type
+      distance_from(Facade1 const& f1, Facade2 const& f2, thrust::detail::false_type)
+    {
+      return f2.distance_to(f1);
+    }
+    
+    template <class Facade1, class Facade2>
+    __host__ __device__
+    static typename thrust::detail::distance_from_result<Facade1,Facade2>::type
+      distance_from(Facade1 const& f1, Facade2 const& f2)
+    {
+      // dispatch the implementation of this method upon whether or not
+      // Facade2 is convertible to Facade1
+      return distance_from(f1, f2,
+        typename thrust::detail::is_convertible<Facade2,Facade1>::type());
+    }
+
+    //
+    // Curiously Recurring Template interface.
+    //
+    template <typename Derived, typename Value, typename System, typename Traversal, typename Reference, typename Difference>
+    __host__ __device__
+    static Derived& derived(iterator_facade<Derived,Value,System,Traversal,Reference,Difference>& facade)
+    {
+      return *static_cast<Derived*>(&facade);
+    }
+
+    template <typename Derived, typename Value, typename System, typename Traversal, typename Reference, typename Difference>
+    __host__ __device__
+    static Derived const& derived(iterator_facade<Derived,Value,System,Traversal,Reference,Difference> const& facade)
+    {
+      return *static_cast<Derived const*>(&facade);
+    }
+
+    /*! \endcond
+     */
+}; // end iterator_core_access
+
+
+/*! \p iterator_facade is a template which allows the programmer to define a novel iterator with a standards-conforming interface
+ *  which Thrust can use to reason about algorithm acceleration opportunities.
+ *
+ *  Because most of a standard iterator's interface is defined in terms of a small set of core primitives, \p iterator_facade
+ *  defines the non-primitive portion mechanically. In principle a novel iterator could explicitly provide the entire interface in
+ *  an ad hoc fashion but doing so might be tedious and prone to subtle errors.
+ *
+ *  Often \p iterator_facade is too primitive a tool to use for defining novel iterators. In these cases, \p iterator_adaptor
+ *  or a specific fancy iterator should be used instead.
+ *
+ *  \p iterator_facade's functionality is derived from and generally equivalent to \p boost::iterator_facade.
+ *  The exception is Thrust's addition of the template parameter \p System, which is necessary to allow Thrust
+ *  to dispatch an algorithm to one of several parallel backend systems. An additional exception is Thrust's omission
+ *  of the \c operator-> member function.
+ *
+ *  Interested users may refer to <tt>boost::iterator_facade</tt>'s documentation for usage examples.
+ *
+ *  \note \p iterator_facade's arithmetic operator free functions exist with the usual meanings but are omitted here for brevity.
+ */
+template<typename Derived,
+         typename Value,
+         typename System,
+         typename Traversal,
+         typename Reference,
+         typename Difference = std::ptrdiff_t>
+  class iterator_facade
+{
+  private:
+    /*! \cond
+     */
+
+    //
+    // Curiously Recurring Template interface.
+    //
+    __host__ __device__
+    Derived& derived()
+    {
+      return *static_cast<Derived*>(this);
+    }
+
+    __host__ __device__
+    Derived const& derived() const
+    {
+      return *static_cast<Derived const*>(this);
+    }
+    /*! \endcond
+     */
+
+  public:
+    /*! The type of element pointed to by \p iterator_facade.
+     */
+    typedef typename thrust::detail::remove_const<Value>::type value_type;
+
+    /*! The return type of \p iterator_facade::operator*().
+     */
+    typedef Reference                                          reference;
+
+    /*! The return type of \p iterator_facade's non-existent \c operator->()
+     *  member function. Unlike \c boost::iterator_facade, \p iterator_facade
+     *  disallows access to the \p value_type's members through expressions of the
+     *  form <tt>iter->member</tt>. \p pointer is defined to \c void to indicate
+     *  that these expressions are not allowed. This limitation may be relaxed in a
+     *  future version of Thrust.
+     */
+    typedef void                                               pointer;
+
+    /*! The type of expressions of the form <tt>x - y</tt> where <tt>x</tt> and <tt>y</tt>
+     *  are of type \p iterator_facade.
+     */
+    typedef Difference                                         difference_type;
+
+    /*! The type of iterator category of \p iterator_facade.
+     */
+    typedef typename thrust::detail::iterator_facade_category<
+      System, Traversal, Value, Reference
+    >::type                                                    iterator_category;
+
+    /*! \p operator*() dereferences this \p iterator_facade.
+     *  \return A reference to the element pointed to by this \p iterator_facade.
+     */
+    __host__ __device__
+    reference operator*() const
+    {
+      return iterator_core_access::dereference(this->derived());
+    }
+
+    // XXX unimplemented for now, consider implementing it later
+    //pointer operator->() const
+    //{
+    //  return;
+    //}
+
+    // XXX investigate whether or not we need to go to the lengths
+    //     boost does to determine the return type
+
+    /*! \p operator[] performs indexed dereference.
+     *  \return A reference to the element \p n distance away from this \p iterator_facade.
+     */
+    __host__ __device__
+    reference operator[](difference_type n) const
+    {
+      return *(this->derived() + n);
+    }
+
+    /*! \p operator++ pre-increments this \p iterator_facade to refer to the element in the next position.
+     *  \return <tt>*this</tt>
+     */
+    __host__ __device__
+    Derived& operator++()
+    {
+      iterator_core_access::increment(this->derived());
+      return this->derived();
+    }
+
+    /*! \p operator++ post-increments this \p iterator_facade and returns a new \p iterator_facade referring to the element in the next position.
+     *  \return A copy of <tt>*this</tt> before increment.
+     */
+    __host__ __device__
+    Derived  operator++(int)
+    {
+      Derived tmp(this->derived());
+      ++*this;
+      return tmp;
+    }
+
+    /*! \p operator-- pre-decrements this \p iterator_facade to refer to the element in the previous position.
+     *  \return <tt>*this</tt>
+     */
+    __host__ __device__
+    Derived& operator--()
+    {
+      iterator_core_access::decrement(this->derived());
+      return this->derived();
+    }
+
+    /*! \p operator-- post-decrements this \p iterator_facade and returns a new \p iterator_facade referring to the element in the previous position.
+     *  \return A copy of <tt>*this</tt> before decrement.
+     */
+    __host__ __device__
+    Derived  operator--(int)
+    {
+      Derived tmp(this->derived());
+      --*this;
+      return tmp;
+    }
+
+    /*! \p operator+= increments this \p iterator_facade to refer to an element a given distance after its current position.
+     *  \param n The quantity to increment.
+     *  \return <tt>*this</tt>
+     */
+    __host__ __device__
+    Derived& operator+=(difference_type n)
+    {
+      iterator_core_access::advance(this->derived(), n);
+      return this->derived();
+    }
+
+    /*! \p operator-= decrements this \p iterator_facade to refer to an element a given distance before its current postition.
+     *  \param n The quantity to decrement.
+     *  \return <tt>*this</tt>
+     */
+    __host__ __device__
+    Derived& operator-=(difference_type n)
+    {
+      iterator_core_access::advance(this->derived(), -n);
+      return this->derived();
+    }
+
+    /*! \p operator- subtracts a given quantity from this \p iterator_facade and returns a new \p iterator_facade referring to the element at the given position before this \p iterator_facade.
+     *  \param n The quantity to decrement
+     *  \return An \p iterator_facade pointing \p n elements before this \p iterator_facade.
+     */
+    __host__ __device__
+    Derived  operator-(difference_type n) const
+    {
+      Derived result(this->derived());
+      return result -= n;
+    }
+}; // end iterator_facade
+
+/*! \cond
+ */
+
+// Comparison operators
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+// XXX it might be nice to implement this at some point
+//typename enable_if_interoperable<Dr1,Dr2,bool>::type // exposition
+bool
+operator ==(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+            iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return iterator_core_access
+    ::equal(*static_cast<Derived1 const*>(&lhs),
+            *static_cast<Derived2 const*>(&rhs));
+}
+
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+// XXX it might be nice to implement this at some point
+//typename enable_if_interoperable<Dr1,Dr2,bool>::type // exposition
+bool
+operator !=(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+            iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return !iterator_core_access
+    ::equal(*static_cast<Derived1 const*>(&lhs),
+            *static_cast<Derived2 const*>(&rhs));
+}
+
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+// XXX it might be nice to implement this at some point
+//typename enable_if_interoperable<Dr1,Dr2,bool>::type // exposition
+bool
+operator <(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+           iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return 0 > iterator_core_access
+    ::distance_from(*static_cast<Derived1 const*>(&lhs),
+                    *static_cast<Derived2 const*>(&rhs));
+}
+
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+// XXX it might be nice to implement this at some point
+//typename enable_if_interoperable<Dr1,Dr2,bool>::type // exposition
+bool
+operator >(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+           iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return 0 < iterator_core_access
+    ::distance_from(*static_cast<Derived1 const*>(&lhs),
+                    *static_cast<Derived2 const*>(&rhs));
+}
+
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+// XXX it might be nice to implement this at some point
+//typename enable_if_interoperable<Dr1,Dr2,bool>::type // exposition
+bool
+operator <=(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+            iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return 0 >= iterator_core_access
+    ::distance_from(*static_cast<Derived1 const*>(&lhs),
+                    *static_cast<Derived2 const*>(&rhs));
+}
+
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+// XXX it might be nice to implement this at some point
+//typename enable_if_interoperable<Dr1,Dr2,bool>::type // exposition
+bool
+operator >=(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+            iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return 0 <= iterator_core_access
+    ::distance_from(*static_cast<Derived1 const*>(&lhs),
+                    *static_cast<Derived2 const*>(&rhs));
+}
+
+// Iterator difference
+template <typename Derived1, typename Value1, typename System1, typename Traversal1, typename Reference1, typename Difference1,
+          typename Derived2, typename Value2, typename System2, typename Traversal2, typename Reference2, typename Difference2>
+inline __host__ __device__
+
+// divine the type this operator returns
+typename thrust::detail::distance_from_result<
+  iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1>,
+  iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2>
+>::type
+
+operator-(iterator_facade<Derived1,Value1,System1,Traversal1,Reference1,Difference1> const& lhs,
+          iterator_facade<Derived2,Value2,System2,Traversal2,Reference2,Difference2> const& rhs)
+{
+  return iterator_core_access
+    ::distance_from(*static_cast<Derived1 const*>(&lhs),
+                    *static_cast<Derived2 const*>(&rhs));
+}
+
+// Iterator addition
+template <typename Derived, typename Value, typename System, typename Traversal, typename Reference, typename Difference>
+inline __host__ __device__
+Derived operator+ (iterator_facade<Derived,Value,System,Traversal,Reference,Difference> const& i,
+                   typename Derived::difference_type n)
+{
+  Derived tmp(static_cast<Derived const&>(i));
+  return tmp += n;
+}
+
+template <typename Derived, typename Value, typename System, typename Traversal, typename Reference, typename Difference>
+inline __host__ __device__
+Derived operator+ (typename Derived::difference_type n,
+                   iterator_facade<Derived,Value,System,Traversal,Reference,Difference> const& i)
+{
+  Derived tmp(static_cast<Derived const&>(i));
+  return tmp += n;
+}
+
+/*! \endcond
+ */
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_traits.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_traits.h
new file mode 100644
index 000000000000..b2f4b175a3fa
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/iterator_traits.h
@@ -0,0 +1,65 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/iterator_traits.h
+ *  \brief Traits and metafunctions for reasoning about the traits of iterators
+ */
+
+/*
+ * (C) Copyright David Abrahams 2003.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/type_traits/void_t.h>
+
+#include <iterator>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \p iterator_traits is a type trait class that provides a uniform
+ *  interface for querying the properties of iterators at compile-time.
+ */
+template <typename T>
+struct iterator_traits : std::iterator_traits<T> {};
+
+template<typename Iterator> struct iterator_value;
+
+template<typename Iterator> struct iterator_pointer;
+
+template<typename Iterator> struct iterator_reference;
+
+template<typename Iterator> struct iterator_difference;
+
+template<typename Iterator> struct iterator_traversal;
+
+template<typename Iterator> struct iterator_system;
+
+THRUST_NAMESPACE_END
+
+#include <thrust/iterator/detail/iterator_traversal_tags.h>
+#include <thrust/iterator/detail/host_system_tag.h>
+#include <thrust/iterator/detail/device_system_tag.h>
+#include <thrust/iterator/detail/any_system_tag.h>
+#include <thrust/iterator/detail/iterator_traits.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/permutation_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/permutation_iterator.h
new file mode 100644
index 000000000000..be5010e543bb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/permutation_iterator.h
@@ -0,0 +1,216 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/iterator/permutation_iterator.h
+ *  \brief An iterator which performs a gather or scatter operation when dereferenced
+ */
+
+/*
+ * (C) Copyright Toon Knapen    2001.
+ * (C) Copyright David Abrahams 2003.
+ * (C) Copyright Roland Richter 2003.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/permutation_iterator_base.h>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p permutation_iterator is an iterator which represents a pointer into a
+ *  reordered view of a given range. \p permutation_iterator is an imprecise name;
+ *  the reordered view need not be a strict permutation. This iterator is useful
+ *  for fusing a scatter or gather operation with other algorithms.
+ *
+ *  This iterator takes two arguments:
+ *
+ *    - an iterator to the range \c V on which the "permutation" will be applied
+ *    - the reindexing scheme that defines how the elements of \c V will be permuted.
+ *
+ *  Note that \p permutation_iterator is not limited to strict permutations of the
+ *  given range \c V. The distance between begin and end of the reindexing iterators
+ *  is allowed to be smaller compared to the size of the range \c V, in which case
+ *  the \p permutation_iterator only provides a "permutation" of a subrange of \c V.
+ *  The indices neither need to be unique. In this same context, it must be noted
+ *  that the past-the-end \p permutation_iterator is completely defined by means of
+ *  the past-the-end iterator to the indices.
+ *
+ *  The following code snippet demonstrates how to create a \p permutation_iterator
+ *  which represents a reordering of the contents of a \p device_vector.
+ *
+ *  \code
+ *  #include <thrust/iterator/permutation_iterator.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<float> values(8);
+ *  values[0] = 10.0f;
+ *  values[1] = 20.0f;
+ *  values[2] = 30.0f;
+ *  values[3] = 40.0f;
+ *  values[4] = 50.0f;
+ *  values[5] = 60.0f;
+ *  values[6] = 70.0f;
+ *  values[7] = 80.0f;
+ *
+ *  thrust::device_vector<int> indices(4);
+ *  indices[0] = 2;
+ *  indices[1] = 6;
+ *  indices[2] = 1;
+ *  indices[3] = 3;
+ *
+ *  typedef thrust::device_vector<float>::iterator ElementIterator;
+ *  typedef thrust::device_vector<int>::iterator   IndexIterator;
+ *
+ *  thrust::permutation_iterator<ElementIterator,IndexIterator> iter(values.begin(), indices.begin());
+ *
+ *  *iter;   // returns 30.0f;
+ *  iter[0]; // returns 30.0f;
+ *  iter[1]; // returns 70.0f;
+ *  iter[2]; // returns 20.0f;
+ *  iter[3]; // returns 40.0f;
+ *
+ *  // iter[4] is an out-of-bounds error
+ *
+ *  *iter   = -1.0f; // sets values[2] to -1.0f;
+ *  iter[0] = -1.0f; // sets values[2] to -1.0f;
+ *  iter[1] = -1.0f; // sets values[6] to -1.0f;
+ *  iter[2] = -1.0f; // sets values[1] to -1.0f;
+ *  iter[3] = -1.0f; // sets values[3] to -1.0f;
+ *
+ *  // values is now {10, -1, -1, -1, 50, 60, -1, 80}
+ *  \endcode
+ *
+ *  \see make_permutation_iterator
+ */
+template <typename ElementIterator,
+          typename IndexIterator>
+  class permutation_iterator
+    : public thrust::detail::permutation_iterator_base<
+        ElementIterator,
+        IndexIterator
+      >::type
+{
+  /*! \cond
+   */
+  private:
+    typedef typename detail::permutation_iterator_base<ElementIterator,IndexIterator>::type super_t;
+
+    friend class thrust::iterator_core_access;
+  /*! \endcond
+   */
+
+  public:
+    /*! Null constructor calls the null constructor of this \p permutation_iterator's
+     *  element iterator.
+     */
+    __host__ __device__
+    permutation_iterator()
+      : m_element_iterator() {}
+
+    /*! Constructor accepts an \c ElementIterator into a range of values and an
+     *  \c IndexIterator into a range of indices defining the indexing scheme on the
+     *  values.
+     *
+     *  \param x An \c ElementIterator pointing this \p permutation_iterator's range of values.
+     *  \param y An \c IndexIterator pointing to an indexing scheme to use on \p x.
+     */
+    __host__ __device__
+    explicit permutation_iterator(ElementIterator x, IndexIterator y)
+      : super_t(y), m_element_iterator(x) {}
+
+    /*! Copy constructor accepts a related \p permutation_iterator.
+     *  \param r A compatible \p permutation_iterator to copy from.
+     */
+    template<typename OtherElementIterator, typename OtherIndexIterator>
+    __host__ __device__
+    permutation_iterator(permutation_iterator<OtherElementIterator,OtherIndexIterator> const &r
+    // XXX remove these guards when we have static_assert
+    , typename detail::enable_if_convertible<OtherElementIterator, ElementIterator>::type* = 0
+    , typename detail::enable_if_convertible<OtherIndexIterator, IndexIterator>::type* = 0
+    )
+      : super_t(r.base()), m_element_iterator(r.m_element_iterator)
+    {}
+
+  /*! \cond
+   */
+  private:
+    // MSVC 2013 and 2015 incorrectly warning about returning a reference to
+    // a local/temporary here.
+    // See goo.gl/LELTNp
+    THRUST_DISABLE_MSVC_WARNING_BEGIN(4172)
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    typename super_t::reference dereference() const
+    {
+      return *(m_element_iterator + *this->base());
+    }
+
+    THRUST_DISABLE_MSVC_WARNING_END(4172)
+
+    // make friends for the copy constructor
+    template<typename,typename> friend class permutation_iterator;
+
+    ElementIterator m_element_iterator;
+  /*! \endcond
+   */
+}; // end permutation_iterator
+
+
+/*! \p make_permutation_iterator creates a \p permutation_iterator
+ *  from an \c ElementIterator pointing to a range of elements to "permute"
+ *  and an \c IndexIterator pointing to a range of indices defining an indexing
+ *  scheme on the values.
+ *
+ *  \param e An \c ElementIterator pointing to a range of values.
+ *  \param i An \c IndexIterator pointing to an indexing scheme to use on \p e.
+ *  \return A new \p permutation_iterator which permutes the range \p e by \p i.
+ *  \see permutation_iterator
+ */
+template<typename ElementIterator, typename IndexIterator>
+__host__ __device__
+permutation_iterator<ElementIterator,IndexIterator> make_permutation_iterator(ElementIterator e, IndexIterator i)
+{
+  return permutation_iterator<ElementIterator,IndexIterator>(e,i);
+}
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/retag.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/retag.h
new file mode 100644
index 000000000000..1eb770ae3213
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/retag.h
@@ -0,0 +1,69 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/iterator/retag.h
+ *  \brief Functionality for altering an iterator's associated system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/retag.h>
+
+THRUST_NAMESPACE_BEGIN
+
+
+/*! \ingroup iterator_tags
+ *  \{
+ */
+
+#if 0
+/*! \p reinterpret_tag returns a copy of an iterator and changes the type of the result's system tag.
+ *  \tparam Tag Any system tag.
+ *  \tparam Iterator Any iterator type.
+ *  \param iter The iterator of interest.
+ *  \return An iterator of unspecified type whose system tag is \p Tag and whose behavior is otherwise
+ *          equivalent to \p iter.
+ *  \note Unlike \p retag, \p reinterpret_tag does not enforce that the converted-to system tag be
+ *        related to the converted-from system tag.
+ *  \see retag
+ */
+template<typename Tag, typename Iterator>
+__host__ __device__
+unspecified_iterator_type reinterpret_tag(Iterator iter);
+
+/*! \p retag returns a copy of an iterator and changes the type of the result's system tag.
+ *  \tparam Tag \p Tag shall be convertible to <tt>thrust::iterator_system<Iterator>::type</tt>,
+ *              or <tt>thrust::iterator_system<Iterator>::type</tt> is a base type of \p Tag.
+ *  \tparam Iterator Any iterator type.
+ *  \param iter The iterator of interest.
+ *  \return An iterator of unspecified type whose system tag is \p Tag and whose behavior is
+ *          otherwise equivalent to \p iter.
+ *  \note Unlike \p reinterpret_tag, \p retag enforces that the converted-to system tag be
+ *        related to the converted-from system tag.
+ *  \see reinterpret_tag
+ */
+template<typename Tag, typename Iterator>
+__host__ __device__
+unspecified_iterator_type retag(Iterator iter);
+#endif
+
+/*! \} // iterator_tags
+ */
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/reverse_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/reverse_iterator.h
new file mode 100644
index 000000000000..fe8bbe0cf832
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/reverse_iterator.h
@@ -0,0 +1,237 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/reverse_iterator.h
+ *  \brief An iterator adaptor which adapts another iterator to traverse backwards
+ */
+
+/*
+ * (C) Copyright David Abrahams 2002.
+ * (C) Copyright Jeremy Siek    2002.
+ * (C) Copyright Thomas Witt    2002.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/reverse_iterator_base.h>
+#include <thrust/iterator/iterator_facade.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p reverse_iterator is an iterator which represents a pointer into a
+ *  reversed view of a given range. In this way, \p reverse_iterator allows
+ *  backwards iteration through a bidirectional input range.
+ *
+ *  It is important to note that although \p reverse_iterator is constructed
+ *  from a given iterator, it points to the element preceding it. In this way,
+ *  the past-the-end \p reverse_iterator of a given range points to the element
+ *  preceding the first element of the input range. By the same token, the first
+ *  \p reverse_iterator of a given range is constructed from a past-the-end iterator
+ *  of the original range yet points to the last element of the input.
+ *
+ *  The following code snippet demonstrates how to create a \p reverse_iterator
+ *  which represents a reversed view of the contents of a \p device_vector.
+ *
+ *  \code
+ *  #include <thrust/iterator/reverse_iterator.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<float> v(4);
+ *  v[0] = 0.0f;
+ *  v[1] = 1.0f;
+ *  v[2] = 2.0f;
+ *  v[3] = 3.0f;
+ *
+ *  typedef thrust::device_vector<float>::iterator Iterator;
+ *
+ *  // note that we point the iterator to the *end* of the device_vector
+ *  thrust::reverse_iterator<Iterator> iter(values.end());
+ *
+ *  *iter;   // returns 3.0f;
+ *  iter[0]; // returns 3.0f;
+ *  iter[1]; // returns 2.0f;
+ *  iter[2]; // returns 1.0f;
+ *  iter[3]; // returns 0.0f;
+ *
+ *  // iter[4] is an out-of-bounds error
+ *  \endcode
+ *
+ *  Since reversing a range is a common operation, containers like \p device_vector
+ *  have nested typedefs for declaration shorthand and methods for constructing
+ *  reverse_iterators. The following code snippet is equivalent to the previous:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<float> v(4);
+ *  v[0] = 0.0f;
+ *  v[1] = 1.0f;
+ *  v[2] = 2.0f;
+ *  v[3] = 3.0f;
+ *
+ *  // we use the nested type reverse_iterator to refer to a reversed view of
+ *  // a device_vector and the method rbegin() to create a reverse_iterator pointing
+ *  // to the beginning of the reversed device_vector
+ *  thrust::device_iterator<float>::reverse_iterator iter = values.rbegin();
+ *
+ *  *iter;   // returns 3.0f;
+ *  iter[0]; // returns 3.0f;
+ *  iter[1]; // returns 2.0f;
+ *  iter[2]; // returns 1.0f;
+ *  iter[3]; // returns 0.0f;
+ *
+ *  // iter[4] is an out-of-bounds error
+ *
+ *  // similarly, rend() points to the end of the reversed sequence:
+ *  assert(values.rend() == (iter + 4));
+ *  \endcode
+ *
+ *  Finally, the following code snippet demonstrates how to use reverse_iterator to
+ *  perform a reversed prefix sum operation on the contents of a device_vector:
+ *
+ *  \code
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/scan.h>
+ *  ...
+ *  thrust::device_vector<int> v(5);
+ *  v[0] = 0;
+ *  v[1] = 1;
+ *  v[2] = 2;
+ *  v[3] = 3;
+ *  v[4] = 4;
+ *
+ *  thrust::device_vector<int> result(5);
+ *
+ *  // exclusive scan v into result in reverse
+ *  thrust::exclusive_scan(v.rbegin(), v.rend(), result.begin());
+ *
+ *  // result is now {0, 4, 7, 9, 10}
+ *  \endcode
+ *
+ *  \see make_reverse_iterator
+ */
+template<typename BidirectionalIterator>
+  class reverse_iterator
+    : public detail::reverse_iterator_base<BidirectionalIterator>::type
+{
+  /*! \cond
+   */
+  private:
+    typedef typename thrust::detail::reverse_iterator_base<
+      BidirectionalIterator
+    >::type super_t;
+
+    friend class thrust::iterator_core_access;
+  /*! \endcond
+   */
+
+  public:
+    /*! Default constructor does nothing.
+     */
+    __host__ __device__
+    reverse_iterator() {}
+
+    /*! \p Constructor accepts a \c BidirectionalIterator pointing to a range
+     *  for this \p reverse_iterator to reverse.
+     *
+     *  \param x A \c BidirectionalIterator pointing to a range to reverse.
+     */
+    __host__ __device__
+    explicit reverse_iterator(BidirectionalIterator x);
+
+    /*! \p Copy constructor allows construction from a related compatible
+     *  \p reverse_iterator.
+     *
+     *  \param r A \p reverse_iterator to copy from.
+     */
+    template<typename OtherBidirectionalIterator>
+    __host__ __device__
+    reverse_iterator(reverse_iterator<OtherBidirectionalIterator> const &r
+// XXX msvc screws this up
+// XXX remove these guards when we have static_assert
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+                     , typename thrust::detail::enable_if<
+                         thrust::detail::is_convertible<
+                           OtherBidirectionalIterator,
+                           BidirectionalIterator
+                         >::value
+                       >::type * = 0
+#endif // MSVC
+                     );
+
+  /*! \cond
+   */
+  private:
+    __thrust_exec_check_disable__
+    __host__ __device__
+    typename super_t::reference dereference() const;
+
+    __host__ __device__
+    void increment();
+
+    __host__ __device__
+    void decrement();
+
+    __host__ __device__
+    void advance(typename super_t::difference_type n);
+
+    template<typename OtherBidirectionalIterator>
+    __host__ __device__
+    typename super_t::difference_type
+    distance_to(reverse_iterator<OtherBidirectionalIterator> const &y) const;
+  /*! \endcond
+   */
+}; // end reverse_iterator
+
+
+/*! \p make_reverse_iterator creates a \p reverse_iterator
+ *  from a \c BidirectionalIterator pointing to a range of elements to reverse.
+ *  
+ *  \param x A \c BidirectionalIterator pointing to a range to reverse.
+ *  \return A new \p reverse_iterator which reverses the range \p x.
+ */
+template<typename BidirectionalIterator>
+__host__ __device__
+reverse_iterator<BidirectionalIterator> make_reverse_iterator(BidirectionalIterator x);
+
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/iterator/detail/reverse_iterator.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_input_output_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_input_output_iterator.h
new file mode 100644
index 000000000000..a5f725dc5371
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_input_output_iterator.h
@@ -0,0 +1,164 @@
+/*
+ *  Copyright 2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/iterator/transform_input_output_iterator.h
+ *  \brief An iterator which adapts another iterator by applying transform
+ *         functions when reading and writing dereferenced values.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/transform_input_output_iterator.inl>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p transform_input_output_iterator is a special kind of iterator which applies
+ * transform functions when reading from or writing to dereferenced values.
+ * This iterator is useful for algorithms that operate on a type that needs to
+ * be serialized/deserialized from values in another iterator, avoiding the
+ * need to materialize intermediate results in memory. This also enables the
+ * transform functions to be fused with the operations that read and write to
+ * the `transform_input_output_iterator`.
+ *
+ * The following code snippet demonstrates how to create a
+ * \p transform_input_output_iterator which performs different transformations when
+ * reading from and writing to the iterator.
+ *
+ * \code
+ * #include <thrust/iterator/transform_input_output_iterator.h>
+ * #include <thrust/device_vector.h>
+ *
+ *  int main()
+ *  {
+ *    const size_t size = 4;
+ *    thrust::device_vector<float> v(size);
+ *
+ *    // Write 1.0f, 2.0f, 3.0f, 4.0f to vector
+ *    thrust::sequence(v.begin(), v.end(), 1);
+ *
+ *    // Iterator that returns negated values and writes squared values
+ *    auto iter = thrust::make_transform_input_output_iterator(v.begin(),
+ *        thrust::negate<float>{}, thrust::square<float>{});
+ * 
+ *    // Iterator negates values when reading
+ *    std::cout << iter[0] << " ";  // -1.0f;
+ *    std::cout << iter[1] << " ";  // -2.0f;
+ *    std::cout << iter[2] << " ";  // -3.0f;
+ *    std::cout << iter[3] << "\n"; // -4.0f;
+ *
+ *    // Write 1.0f, 2.0f, 3.0f, 4.0f to iterator
+ *    thrust::sequence(iter, iter + size, 1);
+ *
+ *    // Values were squared before writing to vector
+ *    std::cout << v[0] << " ";  // 1.0f;
+ *    std::cout << v[1] << " ";  // 4.0f;
+ *    std::cout << v[2] << " ";  // 9.0f;
+ *    std::cout << v[3] << "\n"; // 16.0f;
+ *
+ *  }
+ * \endcode
+ *
+ * \see make_transform_input_output_iterator
+ */
+
+template <typename InputFunction, typename OutputFunction, typename Iterator>
+  class transform_input_output_iterator
+    : public detail::transform_input_output_iterator_base<InputFunction, OutputFunction, Iterator>::type
+{
+
+  /*! \cond
+   */
+
+  public:
+
+    typedef typename
+    detail::transform_input_output_iterator_base<InputFunction, OutputFunction, Iterator>::type
+    super_t;
+
+    friend class thrust::iterator_core_access;
+  /*! \endcond
+   */
+
+  transform_input_output_iterator() = default;
+
+  /*! This constructor takes as argument a \c Iterator an \c InputFunction and an
+   * \c OutputFunction and copies them to a new \p transform_input_output_iterator
+   *
+   * \param io An \c Iterator pointing to where the input to \c InputFunction
+   *           will be read from and the result of \c OutputFunction will be written to
+   * \param input_function An \c InputFunction to be executed on values read from the iterator
+   * \param output_function An \c OutputFunction to be executed on values written to the iterator
+   */
+    __host__ __device__
+    transform_input_output_iterator(Iterator const& io, InputFunction input_function, OutputFunction output_function)
+      : super_t(io), input_function(input_function), output_function(output_function)
+    {
+    }
+
+    /*! \cond
+     */
+  private:
+
+    __host__ __device__
+    typename super_t::reference dereference() const
+    {
+      return detail::transform_input_output_iterator_proxy<
+        InputFunction, OutputFunction, Iterator
+      >(this->base_reference(), input_function, output_function);
+    }
+
+    InputFunction input_function;
+    OutputFunction output_function;
+
+    /*! \endcond
+     */
+}; // end transform_input_output_iterator
+
+/*! \p make_transform_input_output_iterator creates a \p transform_input_output_iterator from
+ *  an \c Iterator a \c InputFunction and a \c OutputFunction
+ *
+ * \param io An \c Iterator pointing to where the input to \c InputFunction
+ *           will be read from and the result of \c OutputFunction will be written to
+ * \param input_function An \c InputFunction to be executed on values read from the iterator
+ * \param output_function An \c OutputFunction to be executed on values written to the iterator
+ *  \see transform_input_output_iterator
+ */
+template <typename InputFunction, typename OutputFunction, typename Iterator>
+transform_input_output_iterator<InputFunction, OutputFunction, Iterator>
+__host__ __device__
+make_transform_input_output_iterator(Iterator io, InputFunction input_function, OutputFunction output_function)
+{
+    return transform_input_output_iterator<InputFunction, OutputFunction, Iterator>(io, input_function, output_function);
+} // end make_transform_input_output_iterator
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_iterator.h
new file mode 100644
index 000000000000..5afb5f37b0fa
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_iterator.h
@@ -0,0 +1,355 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/transform_iterator.h
+ *  \brief An iterator which adapts another iterator by applying a function to the result of its dereference
+ */
+
+/*
+ * (C) Copyright David Abrahams 2002.
+ * (C) Copyright Jeremy Siek    2002.
+ * (C) Copyright Thomas Witt    2002.
+ *
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include the details first
+#include <thrust/iterator/detail/transform_iterator.inl>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p transform_iterator is an iterator which represents a pointer into a range
+ *  of values after transformation by a function. This iterator is useful for
+ *  creating a range filled with the result of applying an operation to another range
+ *  without either explicitly storing it in memory, or explicitly executing the transformation.
+ *  Using \p transform_iterator facilitates kernel fusion by deferring the execution
+ *  of a transformation until the value is needed while saving both memory capacity
+ *  and bandwidth.
+ *
+ *  The following code snippet demonstrates how to create a \p transform_iterator
+ *  which represents the result of \c sqrtf applied to the contents of a \p device_vector.
+ *
+ *  \code
+ *  #include <thrust/iterator/transform_iterator.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  // note: functor inherits from unary_function
+ *  struct square_root : public thrust::unary_function<float,float>
+ *  {
+ *    __host__ __device__
+ *    float operator()(float x) const
+ *    {
+ *      return sqrtf(x);
+ *    }
+ *  };
+ *
+ *  int main()
+ *  {
+ *    thrust::device_vector<float> v(4);
+ *    v[0] = 1.0f;
+ *    v[1] = 4.0f;
+ *    v[2] = 9.0f;
+ *    v[3] = 16.0f;
+ *
+ *    typedef thrust::device_vector<float>::iterator FloatIterator;
+ *
+ *    thrust::transform_iterator<square_root, FloatIterator> iter(v.begin(), square_root());
+ *
+ *    *iter;   // returns 1.0f
+ *    iter[0]; // returns 1.0f;
+ *    iter[1]; // returns 2.0f;
+ *    iter[2]; // returns 3.0f;
+ *    iter[3]; // returns 4.0f;
+ *
+ *    // iter[4] is an out-of-bounds error
+ *  }
+ *  \endcode
+ *
+ *  This next example demonstrates how to use a \p transform_iterator with the
+ *  \p thrust::reduce function to compute the sum of squares of a sequence.
+ *  We will create temporary \p transform_iterators with the
+ *  \p make_transform_iterator function in order to avoid explicitly specifying their type:
+ *
+ *  \code
+ *  #include <thrust/iterator/transform_iterator.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/reduce.h>
+ *  #include <iostream>
+ *
+ *  // note: functor inherits from unary_function
+ *  struct square : public thrust::unary_function<float,float>
+ *  {
+ *    __host__ __device__
+ *    float operator()(float x) const
+ *    {
+ *      return x * x;
+ *    }
+ *  };
+ *
+ *  int main()
+ *  {
+ *    // initialize a device array
+ *    thrust::device_vector<float> v(4);
+ *    v[0] = 1.0f;
+ *    v[1] = 2.0f;
+ *    v[2] = 3.0f;
+ *    v[3] = 4.0f;
+ *
+ *    float sum_of_squares =
+ *     thrust::reduce(thrust::make_transform_iterator(v.begin(), square()),
+ *                    thrust::make_transform_iterator(v.end(),   square()));
+ *
+ *    std::cout << "sum of squares: " << sum_of_squares << std::endl;
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  Note that in the previous two examples the transform functor (namely \c square_root
+ *  and \c square) inherits from \c thrust::unary_function.  Inheriting from
+ *  \c thrust::unary_function ensures that a functor is a valid \c AdaptableUnaryFunction
+ *  and provides all the necessary \c typedef declarations.  The \p transform_iterator
+ *  can also be applied to a \c UnaryFunction that does not inherit from
+ *  \c thrust::unary_function using an optional template argument.  The following example
+ *  illustrates how to use the third template argument to specify the \c result_type of
+ *  the function.
+ *
+ *  \code
+ *  #include <thrust/iterator/transform_iterator.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  // note: functor *does not* inherit from unary_function
+ *  struct square_root
+ *  {
+ *    __host__ __device__
+ *    float operator()(float x) const
+ *    {
+ *      return sqrtf(x);
+ *    }
+ *  };
+ *
+ *  int main()
+ *  {
+ *    thrust::device_vector<float> v(4);
+ *    v[0] = 1.0f;
+ *    v[1] = 4.0f;
+ *    v[2] = 9.0f;
+ *    v[3] = 16.0f;
+ *
+ *    typedef thrust::device_vector<float>::iterator FloatIterator;
+ *
+ *    // note: float result_type is specified explicitly
+ *    thrust::transform_iterator<square_root, FloatIterator, float> iter(v.begin(), square_root());
+ *
+ *    *iter;   // returns 1.0f
+ *    iter[0]; // returns 1.0f;
+ *    iter[1]; // returns 2.0f;
+ *    iter[2]; // returns 3.0f;
+ *    iter[3]; // returns 4.0f;
+ *
+ *    // iter[4] is an out-of-bounds error
+ *  }
+ *  \endcode
+ *
+ *  \see make_transform_iterator
+ */
+template <class AdaptableUnaryFunction, class Iterator, class Reference = use_default, class Value = use_default>
+  class transform_iterator
+    : public detail::transform_iterator_base<AdaptableUnaryFunction, Iterator, Reference, Value>::type
+{
+  /*! \cond
+   */
+  public:
+    typedef typename
+    detail::transform_iterator_base<AdaptableUnaryFunction, Iterator, Reference, Value>::type
+    super_t;
+
+    friend class thrust::iterator_core_access;
+  /*! \endcond
+   */
+
+  public:
+    /*! Null constructor does nothing.
+     */
+    __host__ __device__
+    transform_iterator() {}
+
+#if THRUST_CPP_DIALECT >= 2011
+    transform_iterator(transform_iterator const&) = default;
+#endif
+
+    /*! This constructor takes as arguments an \c Iterator and an \c AdaptableUnaryFunction
+     *  and copies them to a new \p transform_iterator.
+     *
+     *  \param x An \c Iterator pointing to the input to this \p transform_iterator's \c AdaptableUnaryFunction.
+     *  \param f An \c AdaptableUnaryFunction used to transform the objects pointed to by \p x.
+     */
+    __host__ __device__
+    transform_iterator(Iterator const& x, AdaptableUnaryFunction f)
+      : super_t(x), m_f(f) {
+    }
+
+    /*! This explicit constructor copies the value of a given \c Iterator and creates
+     *  this \p transform_iterator's \c AdaptableUnaryFunction using its null constructor.
+     *
+     *  \param x An \c Iterator to copy.
+     */
+    __host__ __device__
+    explicit transform_iterator(Iterator const& x)
+      : super_t(x) { }
+
+    /*! This copy constructor creates a new \p transform_iterator from another
+     *  \p transform_iterator.
+     *
+     *  \param other The \p transform_iterator to copy.
+     */
+    template<typename OtherAdaptableUnaryFunction,
+             typename OtherIterator,
+             typename OtherReference,
+             typename OtherValue>
+    __host__ __device__
+    transform_iterator(const transform_iterator<OtherAdaptableUnaryFunction, OtherIterator, OtherReference, OtherValue> &other,
+                       typename thrust::detail::enable_if_convertible<OtherIterator, Iterator>::type* = 0,
+                       typename thrust::detail::enable_if_convertible<OtherAdaptableUnaryFunction, AdaptableUnaryFunction>::type* = 0)
+      : super_t(other.base()), m_f(other.functor()) {}
+
+    /*! Copy assignment operator copies from another \p transform_iterator.
+     *  \p other The other \p transform_iterator to copy
+     *  \return <tt>*this</tt>
+     *
+     *  \note If the type of this \p transform_iterator's functor is not copy assignable
+     *        (for example, if it is a lambda) it is not an error to call this function.
+     *        In this case, however, the functor will not be modified.
+     *
+     *        In any case, this \p transform_iterator's underlying iterator will be copy assigned.
+     */
+    __host__ __device__
+    transform_iterator &operator=(const transform_iterator &other)
+    {
+      return do_assign(other,
+      // XXX gcc 4.2.1 crashes on is_copy_assignable; just assume the functor is assignable as a WAR
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC) && (THRUST_GCC_VERSION <= 40201)
+          thrust::detail::true_type()
+#else
+          typename thrust::detail::is_copy_assignable<AdaptableUnaryFunction>::type()
+#endif // THRUST_HOST_COMPILER
+      );
+    }
+
+    /*! This method returns a copy of this \p transform_iterator's \c AdaptableUnaryFunction.
+     *  \return A copy of this \p transform_iterator's \c AdaptableUnaryFunction.
+     */
+    __host__ __device__
+    AdaptableUnaryFunction functor() const
+      { return m_f; }
+
+    /*! \cond
+     */
+  private:
+    __host__ __device__
+    transform_iterator &do_assign(const transform_iterator &other, thrust::detail::true_type)
+    {
+      super_t::operator=(other);
+
+      // do assign to m_f
+      m_f = other.functor();
+
+      return *this;
+    }
+
+    __host__ __device__
+    transform_iterator &do_assign(const transform_iterator &other, thrust::detail::false_type)
+    {
+      super_t::operator=(other);
+
+      // don't assign to m_f
+
+      return *this;
+    }
+
+    // MSVC 2013 and 2015 incorrectly warning about returning a reference to
+    // a local/temporary here.
+    // See goo.gl/LELTNp
+    THRUST_DISABLE_MSVC_WARNING_BEGIN(4172)
+
+    __thrust_exec_check_disable__
+    __host__ __device__
+    typename super_t::reference dereference() const
+    {
+      // Create a temporary to allow iterators with wrapped references to
+      // convert to their value type before calling m_f. Note that this
+      // disallows non-constant operations through m_f.
+      typename thrust::iterator_value<Iterator>::type const& x = *this->base();
+      return m_f(x);
+    }
+
+    THRUST_DISABLE_MSVC_WARNING_END(4172)
+
+    // tag this as mutable per Dave Abrahams in this thread:
+    // http://lists.boost.org/Archives/boost/2004/05/65332.php
+    mutable AdaptableUnaryFunction m_f;
+
+    /*! \endcond
+     */
+}; // end transform_iterator
+
+
+/*! \p make_transform_iterator creates a \p transform_iterator
+ *  from an \c Iterator and \c AdaptableUnaryFunction.
+ *
+ *  \param it The \c Iterator pointing to the input range of the
+ *            newly created \p transform_iterator.
+ *  \param fun The \c AdaptableUnaryFunction used to transform the range pointed
+ *             to by \p it in the newly created \p transform_iterator.
+ *  \return A new \p transform_iterator which transforms the range at
+ *          \p it by \p fun.
+ *  \see transform_iterator
+ */
+template <class AdaptableUnaryFunction, class Iterator>
+inline __host__ __device__
+transform_iterator<AdaptableUnaryFunction, Iterator>
+make_transform_iterator(Iterator it, AdaptableUnaryFunction fun)
+{
+  return transform_iterator<AdaptableUnaryFunction, Iterator>(it, fun);
+} // end make_transform_iterator
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_output_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_output_iterator.h
new file mode 100644
index 000000000000..3ac4b85729e5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/transform_output_iterator.h
@@ -0,0 +1,164 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/iterator/transform_output_iterator.h
+ *  \brief An output iterator which adapts another output iterator by applying a
+ *         function to the result of its dereference before writing it.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/transform_output_iterator.inl>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p transform_output_iterator is a special kind of output iterator which
+ * transforms a value written upon dereference. This iterator is useful
+ * for transforming an output from algorithms without explicitly storing the
+ * intermediate result in the memory and applying subsequent transformation, 
+ * thereby avoiding wasting memory capacity and bandwidth.
+ * Using \p transform_iterator facilitates kernel fusion by deferring execution
+ * of transformation until the value is written while saving both memory
+ * capacity and bandwidth.
+ *
+ * The following code snippet demonstrated how to create a
+ * \p transform_output_iterator which applies \c sqrtf to the assigning value.
+ *
+ * \code
+ * #include <thrust/iterator/transform_output_iterator.h>
+ * #include <thrust/device_vector.h>
+ *
+ * // note: functor inherits form unary function
+ *  // note: functor inherits from unary_function
+ *  struct square_root : public thrust::unary_function<float,float>
+ *  {
+ *    __host__ __device__
+ *    float operator()(float x) const
+ *    {
+ *      return sqrtf(x);
+ *    }
+ *  };
+ *  
+ *  int main()
+ *  {
+ *    thrust::device_vector<float> v(4);
+ *
+ *    typedef thrust::device_vector<float>::iterator FloatIterator;
+ *    thrust::transform_output_iterator<square_root, FloatIterator> iter(v.begin(), square_root());
+ *
+ *    iter[0] =  1.0f;    // stores sqrtf( 1.0f) 
+ *    iter[1] =  4.0f;    // stores sqrtf( 4.0f)
+ *    iter[2] =  9.0f;    // stores sqrtf( 9.0f)
+ *    iter[3] = 16.0f;    // stores sqrtf(16.0f)
+ *    // iter[4] is an out-of-bounds error
+ *                                                                                           
+ *    v[0]; // returns 1.0f;
+ *    v[1]; // returns 2.0f;
+ *    v[2]; // returns 3.0f;
+ *    v[3]; // returns 4.0f;
+ *                                                                                           
+ *  }
+ *  \endcode
+ *
+ *  \see make_transform_output_iterator
+ */
+
+template <typename UnaryFunction, typename OutputIterator>
+  class transform_output_iterator
+    : public detail::transform_output_iterator_base<UnaryFunction, OutputIterator>::type
+{
+
+  /*! \cond
+   */
+
+  public:
+
+    typedef typename
+    detail::transform_output_iterator_base<UnaryFunction, OutputIterator>::type
+    super_t;
+
+    friend class thrust::iterator_core_access;
+  /*! \endcond
+   */
+
+  transform_output_iterator() = default;
+
+  /*! This constructor takes as argument an \c OutputIterator and an \c
+   * UnaryFunction and copies them to a new \p transform_output_iterator
+   *
+   * \param out An \c OutputIterator pointing to the output range whereto the result of 
+   *            \p transform_output_iterator's \c UnaryFunction will be written.
+   * \param fun An \c UnaryFunction used to transform the objects assigned to
+   *            this \p transform_output_iterator.
+   */
+    __host__ __device__
+    transform_output_iterator(OutputIterator const& out, UnaryFunction fun) : super_t(out), fun(fun)
+    {
+    }
+
+    /*! \cond
+     */
+  private:
+
+    __host__ __device__
+    typename super_t::reference dereference() const
+    {
+      return detail::transform_output_iterator_proxy<
+        UnaryFunction, OutputIterator
+      >(this->base_reference(), fun);
+    }
+
+    UnaryFunction fun;
+
+    /*! \endcond
+     */
+}; // end transform_output_iterator
+
+/*! \p make_transform_output_iterator creates a \p transform_output_iterator from
+ *  an \c OutputIterator and \c UnaryFunction.
+ *
+ *  \param out The \c OutputIterator pointing to the output range of the newly
+ *            created \p transform_output_iterator
+ *  \param fun The \c UnaryFunction transform the object before assigning it to
+ *            \c out by the newly created \p transform_output_iterator
+ *  \see transform_output_iterator
+ */
+template <typename UnaryFunction, typename OutputIterator>
+transform_output_iterator<UnaryFunction, OutputIterator>
+__host__ __device__
+make_transform_output_iterator(OutputIterator out, UnaryFunction fun)
+{
+    return transform_output_iterator<UnaryFunction, OutputIterator>(out, fun);
+} // end make_transform_output_iterator
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/iterator/zip_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/zip_iterator.h
new file mode 100644
index 000000000000..c2dd5ddc456f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/iterator/zip_iterator.h
@@ -0,0 +1,258 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/iterator/zip_iterator.h
+ *  \brief An iterator which returns a tuple of the result of dereferencing
+ *         a tuple of iterators when dereferenced
+ */
+
+/*
+ * Copyright David Abrahams and Thomas Becker 2000-2006.
+ * 
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/detail/zip_iterator_base.h>
+#include <thrust/iterator/iterator_facade.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup iterators
+ *  \{
+ */
+
+/*! \addtogroup fancyiterator Fancy Iterators
+ *  \ingroup iterators
+ *  \{
+ */
+
+/*! \p zip_iterator is an iterator which represents a pointer into a range
+ *  of \p tuples whose elements are themselves taken from a \p tuple of input
+ *  iterators. This iterator is useful for creating a virtual array of structures
+ *  while achieving the same performance and bandwidth as the structure of arrays
+ *  idiom. \p zip_iterator also facilitates kernel fusion by providing a convenient
+ *  means of amortizing the execution of the same operation over multiple ranges.
+ *
+ *  The following code snippet demonstrates how to create a \p zip_iterator
+ *  which represents the result of "zipping" multiple ranges together.
+ *  
+ *  \code
+ *  #include <thrust/iterator/zip_iterator.h>
+ *  #include <thrust/tuple.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> int_v(3);
+ *  int_v[0] = 0; int_v[1] = 1; int_v[2] = 2;
+ *
+ *  thrust::device_vector<float> float_v(3);
+ *  float_v[0] = 0.0f; float_v[1] = 1.0f; float_v[2] = 2.0f;
+ *
+ *  thrust::device_vector<char> char_v(3);
+ *  char_v[0] = 'a'; char_v[1] = 'b'; char_v[2] = 'c';
+ *
+ *  // typedef these iterators for shorthand
+ *  typedef thrust::device_vector<int>::iterator   IntIterator;
+ *  typedef thrust::device_vector<float>::iterator FloatIterator;
+ *  typedef thrust::device_vector<char>::iterator  CharIterator;
+ *
+ *  // typedef a tuple of these iterators
+ *  typedef thrust::tuple<IntIterator, FloatIterator, CharIterator> IteratorTuple;
+ *
+ *  // typedef the zip_iterator of this tuple
+ *  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+ *
+ *  // finally, create the zip_iterator
+ *  ZipIterator iter(thrust::make_tuple(int_v.begin(), float_v.begin(), char_v.begin()));
+ *
+ *  *iter;   // returns (0, 0.0f, 'a')
+ *  iter[0]; // returns (0, 0.0f, 'a')
+ *  iter[1]; // returns (1, 1.0f, 'b')
+ *  iter[2]; // returns (2, 2.0f, 'c')
+ *
+ *  thrust::get<0>(iter[2]); // returns 2
+ *  thrust::get<1>(iter[0]); // returns 0.0f
+ *  thrust::get<2>(iter[1]); // returns 'b'
+ *
+ *  // iter[3] is an out-of-bounds error
+ *  \endcode
+ *
+ *  Defining the type of a \p zip_iterator can be complex. The next code example demonstrates
+ *  how to use the \p make_zip_iterator function with the \p make_tuple function to avoid
+ *  explicitly specifying the type of the \p zip_iterator. This example shows how to use
+ *  \p zip_iterator to copy multiple ranges with a single call to \p thrust::copy.
+ *
+ *  \code
+ *  #include <thrust/zip_iterator.h>
+ *  #include <thrust/tuple.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  int main()
+ *  {
+ *    thrust::device_vector<int> int_in(3), int_out(3);
+ *    int_in[0] = 0;
+ *    int_in[1] = 1;
+ *    int_in[2] = 2;
+ *
+ *    thrust::device_vector<float> float_in(3), float_out(3);
+ *    float_in[0] =  0.0f;
+ *    float_in[1] = 10.0f;
+ *    float_in[2] = 20.0f;
+ *
+ *    thrust::copy(thrust::make_zip_iterator(thrust::make_tuple(int_in.begin(), float_in.begin())),
+ *                 thrust::make_zip_iterator(thrust::make_tuple(int_in.end(),   float_in.end())),
+ *                 thrust::make_zip_iterator(thrust::make_tuple(int_out.begin(),float_out.begin())));
+ *
+ *    // int_out is now [0, 1, 2]
+ *    // float_out is now [0.0f, 10.0f, 20.0f]
+ *
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see make_zip_iterator
+ *  \see make_tuple
+ *  \see tuple
+ *  \see get
+ */
+template <typename IteratorTuple>
+  class zip_iterator
+    : public detail::zip_iterator_base<IteratorTuple>::type
+{
+  public:
+    /*! Null constructor does nothing.
+     */
+    inline __host__ __device__
+    zip_iterator();
+
+    /*! This constructor creates a new \p zip_iterator from a
+     *  \p tuple of iterators.
+     *  
+     *  \param iterator_tuple The \p tuple of iterators to copy from.
+     */
+    inline __host__ __device__
+    zip_iterator(IteratorTuple iterator_tuple);
+
+    /*! This copy constructor creates a new \p zip_iterator from another
+     *  \p zip_iterator.
+     *
+     *  \param other The \p zip_iterator to copy.
+     */
+    template<typename OtherIteratorTuple>
+    inline __host__ __device__
+    zip_iterator(const zip_iterator<OtherIteratorTuple> &other,
+                 typename thrust::detail::enable_if_convertible<
+                   OtherIteratorTuple,
+                   IteratorTuple
+                 >::type * = 0);
+
+    /*! This method returns a \c const reference to this \p zip_iterator's
+     *  \p tuple of iterators.
+     *
+     *  \return A \c const reference to this \p zip_iterator's \p tuple
+     *          of iterators.
+     */
+    inline __host__ __device__
+    const IteratorTuple &get_iterator_tuple() const;
+
+    /*! \cond
+     */
+  private:
+    typedef typename
+    detail::zip_iterator_base<IteratorTuple>::type super_t;
+
+    friend class thrust::iterator_core_access;
+
+    // Dereferencing returns a tuple built from the dereferenced
+    // iterators in the iterator tuple.
+    __host__ __device__
+    typename super_t::reference dereference() const;
+
+    // Two zip_iterators are equal if the two first iterators of the
+    // tuple are equal. Note this differs from Boost's implementation, which
+    // considers the entire tuple.
+    template<typename OtherIteratorTuple>
+    inline __host__ __device__
+    bool equal(const zip_iterator<OtherIteratorTuple> &other) const;
+
+    // Advancing a zip_iterator means to advance all iterators in the tuple
+    inline __host__ __device__
+    void advance(typename super_t::difference_type n);
+
+    // Incrementing a zip iterator means to increment all iterators in the tuple
+    inline __host__ __device__
+    void increment();
+
+    // Decrementing a zip iterator means to decrement all iterators in the tuple
+    inline __host__ __device__
+    void decrement();
+
+    // Distance is calculated using the first iterator in the tuple.
+    template<typename OtherIteratorTuple>
+    inline __host__ __device__
+      typename super_t::difference_type
+        distance_to(const zip_iterator<OtherIteratorTuple> &other) const;
+
+    // The iterator tuple.
+    IteratorTuple m_iterator_tuple;
+
+    /*! \endcond
+     */
+}; // end zip_iterator
+
+/*! \p make_zip_iterator creates a \p zip_iterator from a \p tuple
+ *  of iterators.
+ *
+ *  \param t The \p tuple of iterators to copy.
+ *  \return A newly created \p zip_iterator which zips the iterators encapsulated in \p t.
+ *
+ *  \see zip_iterator
+ */
+template<typename... Iterators>
+inline __host__ __device__
+zip_iterator<thrust::tuple<Iterators...>> make_zip_iterator(thrust::tuple<Iterators...> t);
+
+
+/*! \p make_zip_iterator creates a \p zip_iterator from
+ *  iterators.
+ *
+ *  \param its The iterators to copy.
+ *  \return A newly created \p zip_iterator which zips the iterators.
+ *
+ *  \see zip_iterator
+ */
+template<typename... Iterators>
+inline __host__ __device__
+zip_iterator<thrust::tuple<Iterators...>> make_zip_iterator(Iterators... its);
+
+
+/*! \} // end fancyiterators
+ */
+
+/*! \} // end iterators
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/iterator/detail/zip_iterator.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/limits.h b/thirdparty/manifold/thirdparty/thrust/thrust/limits.h
new file mode 100644
index 000000000000..52f38b1fc9e1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/limits.h
@@ -0,0 +1,18 @@
+// Copyright (c) 2018 NVIDIA Corporation
+// Author: Bryce Adelstein Lelbach <brycelelbach@gmail.com>
+//
+// Distributed under the Boost Software License v1.0 (boost.org/LICENSE_1_0.txt)
+
+#pragma once
+
+#include <limits>
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+template <typename T>
+struct numeric_limits : std::numeric_limits<T> {};
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/logical.h b/thirdparty/manifold/thirdparty/thrust/thrust/logical.h
new file mode 100644
index 000000000000..5a8dbbecf6e1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/logical.h
@@ -0,0 +1,275 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file logical.h
+ *  \brief Logical operations on ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup logical
+ *  \ingroup reductions
+ *  \{
+ */
+
+
+/*! \p all_of determines whether all elements in a range satify a predicate.
+ *  Specifically, \p all_of returns \c true if <tt>pred(*i)</tt> is \c true
+ *  for every iterator \c i in the range <tt>[first, last)</tt> and 
+ *  \c false otherwise.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param pred A predicate used to test range elements.
+ *  \return \c true, if all elements satisfy the predicate; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/logical.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  bool A[3] = {true, true, false};
+ *
+ *  thrust::all_of(thrust::host, A, A + 2, thrust::identity<bool>()); // returns true
+ *  thrust::all_of(thrust::host, A, A + 3, thrust::identity<bool>()); // returns false
+ *
+ *  // empty range
+ *  thrust::all_of(thrust::host, A, A, thrust::identity<bool>()); // returns false
+ *  
+ *  \endcode
+ *
+ *  \see any_of
+ *  \see none_of
+ *  \see transform_reduce
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool all_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \p all_of determines whether all elements in a range satify a predicate.
+ * Specifically, \p all_of returns \c true if <tt>pred(*i)</tt> is \c true
+ * for every iterator \c i in the range <tt>[first, last)</tt> and 
+ * \c false otherwise.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param pred A predicate used to test range elements.
+ *  \return \c true, if all elements satisfy the predicate; \c false, otherwise.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/logical.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  bool A[3] = {true, true, false};
+ *
+ *  thrust::all_of(A, A + 2, thrust::identity<bool>()); // returns true
+ *  thrust::all_of(A, A + 3, thrust::identity<bool>()); // returns false
+ *
+ *  // empty range
+ *  thrust::all_of(A, A, thrust::identity<bool>()); // returns false
+ *  
+ *  \endcode
+ *
+ *  \see any_of
+ *  \see none_of
+ *  \see transform_reduce
+ */
+template<typename InputIterator, typename Predicate>
+bool all_of(InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \p any_of determines whether any element in a range satifies a predicate.
+ *  Specifically, \p any_of returns \c true if <tt>pred(*i)</tt> is \c true
+ *  for any iterator \c i in the range <tt>[first, last)</tt> and 
+ *  \c false otherwise.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param pred A predicate used to test range elements.
+ *  \return \c true, if any element satisfies the predicate; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/logical.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  bool A[3] = {true, true, false};
+ *
+ *  thrust::any_of(thrust::host, A, A + 2, thrust::identity<bool>()); // returns true
+ *  thrust::any_of(thrust::host, A, A + 3, thrust::identity<bool>()); // returns true
+ *
+ *  thrust::any_of(thrust::host, A + 2, A + 3, thrust::identity<bool>()); // returns false
+ *
+ *  // empty range
+ *  thrust::any_of(thrust::host, A, A, thrust::identity<bool>()); // returns false
+ *  \endcode
+ *
+ *  \see all_of
+ *  \see none_of
+ *  \see transform_reduce
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool any_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
+   
+
+/*! \p any_of determines whether any element in a range satifies a predicate.
+ * Specifically, \p any_of returns \c true if <tt>pred(*i)</tt> is \c true
+ * for any iterator \c i in the range <tt>[first, last)</tt> and 
+ * \c false otherwise.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param pred A predicate used to test range elements.
+ *  \return \c true, if any element satisfies the predicate; \c false, otherwise.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/logical.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  bool A[3] = {true, true, false};
+ *
+ *  thrust::any_of(A, A + 2, thrust::identity<bool>()); // returns true
+ *  thrust::any_of(A, A + 3, thrust::identity<bool>()); // returns true
+ *
+ *  thrust::any_of(A + 2, A + 3, thrust::identity<bool>()); // returns false
+ *
+ *  // empty range
+ *  thrust::any_of(A, A, thrust::identity<bool>()); // returns false
+ *  \endcode
+ *
+ *  \see all_of
+ *  \see none_of
+ *  \see transform_reduce
+ */
+template<typename InputIterator, typename Predicate>
+bool any_of(InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \p none_of determines whether no element in a range satifies a predicate.
+ *  Specifically, \p none_of returns \c true if there is no iterator \c i in 
+ *  the range <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true,
+ *  and \c false otherwise.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param pred A predicate used to test range elements.
+ *  \return \c true, if no element satisfies the predicate; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/logical.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  bool A[3] = {true, true, false};
+ *
+ *  thrust::none_of(thrust::host, A, A + 2, thrust::identity<bool>()); // returns false
+ *  thrust::none_of(thrust::host, A, A + 3, thrust::identity<bool>()); // returns false
+ *
+ *  thrust::none_of(thrust::host, A + 2, A + 3, thrust::identity<bool>()); // returns true
+ *
+ *  // empty range
+ *  thrust::none_of(thrust::host, A, A, thrust::identity<bool>()); // returns true
+ *  \endcode
+ *
+ *  \see all_of
+ *  \see any_of
+ *  \see transform_reduce
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool none_of(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \p none_of determines whether no element in a range satifies a predicate.
+ *  Specifically, \p none_of returns \c true if there is no iterator \c i in 
+ *  the range <tt>[first, last)</tt> such that <tt>pred(*i)</tt> is \c true,
+ *  and \c false otherwise.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param pred A predicate used to test range elements.
+ *  \return \c true, if no element satisfies the predicate; \c false, otherwise.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \code
+ *  #include <thrust/logical.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  bool A[3] = {true, true, false};
+ *
+ *  thrust::none_of(A, A + 2, thrust::identity<bool>()); // returns false
+ *  thrust::none_of(A, A + 3, thrust::identity<bool>()); // returns false
+ *
+ *  thrust::none_of(A + 2, A + 3, thrust::identity<bool>()); // returns true
+ *
+ *  // empty range
+ *  thrust::none_of(A, A, thrust::identity<bool>()); // returns true
+ *  \endcode
+ *
+ *  \see all_of
+ *  \see any_of
+ *  \see transform_reduce
+ */
+template<typename InputIterator, typename Predicate>
+bool none_of(InputIterator first, InputIterator last, Predicate pred);
+
+
+/*! \} // end logical
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/logical.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/memory.h b/thirdparty/manifold/thirdparty/thrust/thrust/memory.h
new file mode 100644
index 000000000000..819ac251301b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/memory.h
@@ -0,0 +1,396 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/memory.h
+ *  \brief Abstractions for Thrust's memory model.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/detail/pointer.h>
+#include <thrust/detail/reference.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/raw_reference_cast.h>
+#include <thrust/detail/malloc_and_free.h>
+#include <thrust/detail/temporary_buffer.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \defgroup memory_management Memory Management
+ *
+ *  All Thrust functionalities related to memory allocation and deallocation.
+ *
+ */
+
+/** \addtogroup memory_management Memory Management
+ *  \{
+ */
+
+// define pointer for the purpose of Doxygenating it
+// it is actually defined elsewhere
+#if 0
+/*! \p pointer stores a pointer to an object allocated in memory. Like \p device_ptr, this
+ *  type ensures type safety when dispatching standard algorithms on ranges resident in memory.
+ *
+ *  \p pointer generalizes \p device_ptr by relaxing the backend system associated with the \p pointer.
+ *  Instead of the backend system specified by \p THRUST_DEVICE_SYSTEM, \p pointer's
+ *  system is given by its second template parameter, \p Tag. For the purpose of Thrust dispatch,
+ *  <tt>device_ptr<Element></tt> and <tt>pointer<Element,device_system_tag></tt> are considered equivalent.
+ *
+ *  The raw pointer encapsulated by a \p pointer may be obtained through its <tt>get</tt> member function
+ *  or the \p raw_pointer_cast free function.
+ *
+ *  \tparam Element specifies the type of the pointed-to object.
+ *
+ *  \tparam Tag specifies the system with which this \p pointer is associated. This may be any Thrust
+ *          backend system, or a user-defined tag.
+ *
+ *  \tparam Reference allows the client to specify the reference type returned upon derereference.
+ *          By default, this type is <tt>reference<Element,pointer></tt>.
+ *
+ *  \tparam Derived allows the client to specify the name of the derived type when \p pointer is used as
+ *          a base class. This is useful to ensure that arithmetic on values of the derived type return
+ *          values of the derived type as a result. By default, this type is <tt>pointer<Element,Tag,Reference></tt>.
+ *
+ *  \note \p pointer is not a smart pointer; it is the client's responsibility to deallocate memory
+ *        pointer to by \p pointer.
+ *
+ *  \see device_ptr
+ *  \see reference
+ *  \see raw_pointer_cast
+ */
+template<typename Element, typename Tag, typename Reference = thrust::use_default, typename Derived = thrust::use_default>
+  class pointer
+{
+  public:
+    /*! The type of the raw pointer
+     */
+    typedef typename super_t::base_type raw_pointer;
+
+    /*! \p pointer's default constructor initializes its encapsulated pointer to \c 0
+     */
+    __host__ __device__
+    pointer();
+
+    /*! This constructor allows construction of a <tt>pointer<const T, ...></tt> from a <tt>T*</tt>.
+     *
+     *  \param ptr A raw pointer to copy from, presumed to point to a location in \p Tag's memory.
+     *  \tparam OtherElement \p OtherElement shall be convertible to \p Element.
+     */
+    template<typename OtherElement>
+    __host__ __device__
+    explicit pointer(OtherElement *ptr);
+
+    /*! This contructor allows initialization from another pointer-like object.
+     *
+     *  \param other The \p OtherPointer to copy.
+     *
+     *  \tparam OtherPointer The tag associated with \p OtherPointer shall be convertible to \p Tag,
+     *                       and its element type shall be convertible to \p Element.
+     */
+    template<typename OtherPointer>
+    __host__ __device__
+    pointer(const OtherPointer &other,
+            typename thrust::detail::enable_if_pointer_is_convertible<
+              OtherPointer,
+              pointer<Element,Tag,Reference,Derived>
+            >::type * = 0);
+
+    /*! Assignment operator allows assigning from another pointer-like object whose element type
+     *  is convertible to \c Element.
+     *
+     *  \param other The other pointer-like object to assign from.
+     *  \return <tt>*this</tt>
+     *
+     *  \tparam OtherPointer The tag associated with \p OtherPointer shall be convertible to \p Tag,
+     *                       and its element type shall be convertible to \p Element.
+     */
+    template<typename OtherPointer>
+    __host__ __device__
+    typename thrust::detail::enable_if_pointer_is_convertible<
+      OtherPointer,
+      pointer,
+      derived_type &
+    >::type
+    operator=(const OtherPointer &other);
+
+    /*! \p get returns this \p pointer's encapsulated raw pointer.
+     *  \return This \p pointer's raw pointer.
+     */
+    __host__ __device__
+    Element *get() const;
+};
+#endif
+
+/*! This version of \p malloc allocates untyped uninitialized storage associated with a given system.
+ *
+ *  \param system The Thrust system with which to associate the storage.
+ *  \param n The number of bytes of storage to allocate.
+ *  \return If allocation succeeds, a pointer to the allocated storage; a null pointer otherwise.
+ *          The pointer must be deallocated with \p thrust::free.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *
+ *  \pre \p DerivedPolicy must be publically derived from <tt>thrust::execution_policy<DerivedPolicy></tt>.
+ *
+ *  The following code snippet demonstrates how to use \p malloc to allocate a range of memory
+ *  associated with Thrust's device system.
+ *
+ *  \code
+ *  #include <thrust/memory.h>
+ *  ...
+ *  // allocate some memory with thrust::malloc
+ *  const int N = 100;
+ *  thrust::device_system_tag device_sys;
+ *  thrust::pointer<void,thrust::device_space_tag> void_ptr = thrust::malloc(device_sys, N);
+ *
+ *  // manipulate memory
+ *  ...
+ *
+ *  // deallocate void_ptr with thrust::free
+ *  thrust::free(device_sys, void_ptr);
+ *  \endcode
+ *
+ *  \see free
+ *  \see device_malloc
+ */
+template<typename DerivedPolicy>
+__host__ __device__
+pointer<void,DerivedPolicy> malloc(const thrust::detail::execution_policy_base<DerivedPolicy> &system, std::size_t n);
+
+
+/*! This version of \p malloc allocates typed uninitialized storage associated with a given system.
+ *
+ *  \param system The Thrust system with which to associate the storage.
+ *  \param n The number of elements of type \c T which the storage should accomodate.
+ *  \return If allocation succeeds, a pointer to an allocation large enough to accomodate \c n
+ *          elements of type \c T; a null pointer otherwise.
+ *          The pointer must be deallocated with \p thrust::free.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *
+ *  \pre \p DerivedPolicy must be publically derived from <tt>thrust::execution_policy<DerivedPolicy></tt>.
+ *
+ *  The following code snippet demonstrates how to use \p malloc to allocate a range of memory
+ *  to accomodate integers associated with Thrust's device system.
+ *
+ *  \code
+ *  #include <thrust/memory.h>
+ *  ...
+ *  // allocate storage for 100 ints with thrust::malloc
+ *  const int N = 100;
+ *  thrust::device_system_tag device_sys;
+ *  thrust::pointer<int,thrust::device_system_tag> ptr = thrust::malloc<int>(device_sys, N);
+ *
+ *  // manipulate memory
+ *  ...
+ *
+ *  // deallocate ptr with thrust::free
+ *  thrust::free(device_sys, ptr);
+ *  \endcode
+ *
+ *  \see free
+ *  \see device_malloc
+ */
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+pointer<T,DerivedPolicy> malloc(const thrust::detail::execution_policy_base<DerivedPolicy> &system, std::size_t n);
+
+
+/*! \p get_temporary_buffer returns a pointer to storage associated with a given Thrust system sufficient to store up to
+ *  \p n objects of type \c T. If not enough storage is available to accomodate \p n objects, an implementation may return
+ *  a smaller buffer. The number of objects the returned buffer can accomodate is also returned.
+ *
+ *  Thrust uses \p get_temporary_buffer internally when allocating temporary storage required by algorithm implementations.
+ *
+ *  The storage allocated with \p get_temporary_buffer must be returned to the system with \p return_temporary_buffer.
+ *
+ *  \param system The Thrust system with which to associate the storage.
+ *  \param n The requested number of objects of type \c T the storage should accomodate.
+ *  \return A pair \c p such that <tt>p.first</tt> is a pointer to the allocated storage and <tt>p.second</tt> is the number of
+ *          contiguous objects of type \c T that the storage can accomodate. If no storage can be allocated, <tt>p.first</tt> if
+ *          no storage can be obtained. The storage must be returned to the system using \p return_temporary_buffer.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *
+ *  \pre \p DerivedPolicy must be publically derived from <tt>thrust::execution_policy<DerivedPolicy></tt>.
+ *
+ *  The following code snippet demonstrates how to use \p get_temporary_buffer to allocate a range of memory
+ *  to accomodate integers associated with Thrust's device system.
+ *
+ *  \code
+ *  #include <thrust/memory.h>
+ *  ...
+ *  // allocate storage for 100 ints with thrust::get_temporary_buffer
+ *  const int N = 100;
+ *
+ *  typedef thrust::pair<
+ *    thrust::pointer<int,thrust::device_system_tag>,
+ *    std::ptrdiff_t
+ *  > ptr_and_size_t;
+ *
+ *  thrust::device_system_tag device_sys;
+ *  ptr_and_size_t ptr_and_size = thrust::get_temporary_buffer<int>(device_sys, N);
+ *
+ *  // manipulate up to 100 ints
+ *  for(int i = 0; i < ptr_and_size.second; ++i)
+ *  {
+ *    *ptr_and_size.first = i;
+ *  }
+ *
+ *  // deallocate storage with thrust::return_temporary_buffer
+ *  thrust::return_temporary_buffer(device_sys, ptr_and_size.first);
+ *  \endcode
+ *
+ *  \see malloc
+ *  \see return_temporary_buffer
+ */
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+thrust::pair<thrust::pointer<T,DerivedPolicy>, typename thrust::pointer<T,DerivedPolicy>::difference_type>
+get_temporary_buffer(const thrust::detail::execution_policy_base<DerivedPolicy> &system, typename thrust::pointer<T,DerivedPolicy>::difference_type n);
+
+/*! \p free deallocates the storage previously allocated by \p thrust::malloc.
+ *
+ *  \param system The Thrust system with which the storage is associated.
+ *  \param ptr A pointer previously returned by \p thrust::malloc. If \p ptr is null, \p free
+ *         does nothing.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *
+ *  \pre \p ptr shall have been returned by a previous call to <tt>thrust::malloc(system, n)</tt> or <tt>thrust::malloc<T>(system, n)</tt> for some type \c T.
+ *
+ *  The following code snippet demonstrates how to use \p free to deallocate a range of memory
+ *  previously allocated with \p thrust::malloc.
+ *
+ *  \code
+ *  #include <thrust/memory.h>
+ *  ...
+ *  // allocate storage for 100 ints with thrust::malloc
+ *  const int N = 100;
+ *  thrust::device_system_tag device_sys;
+ *  thrust::pointer<int,thrust::device_system_tag> ptr = thrust::malloc<int>(device_sys, N);
+ *
+ *  // mainpulate memory
+ *  ...
+ *
+ *  // deallocate ptr with thrust::free
+ *  thrust::free(device_sys, ptr);
+ *  \endcode
+ */
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+void free(const thrust::detail::execution_policy_base<DerivedPolicy> &system, Pointer ptr);
+
+
+/*! \p return_temporary_buffer deallocates storage associated with a given Thrust system previously allocated by \p get_temporary_buffer.
+ *
+ *  Thrust uses \p return_temporary_buffer internally when deallocating temporary storage required by algorithm implementations.
+ *
+ *  \param system The Thrust system with which the storage is associated.
+ *  \param p A pointer previously returned by \p thrust::get_temporary_buffer. If \p ptr is null, \p return_temporary_buffer does nothing.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *
+ *  \pre \p p shall have been previously allocated by \p thrust::get_temporary_buffer.
+ *
+ *  The following code snippet demonstrates how to use \p return_temporary_buffer to deallocate a range of memory
+ *  previously allocated by \p get_temporary_buffer.
+ *
+ *  \code
+ *  #include <thrust/memory.h>
+ *  ...
+ *  // allocate storage for 100 ints with thrust::get_temporary_buffer
+ *  const int N = 100;
+ *
+ *  typedef thrust::pair<
+ *    thrust::pointer<int,thrust::device_system_tag>,
+ *    std::ptrdiff_t
+ *  > ptr_and_size_t;
+ *
+ *  thrust::device_system_tag device_sys;
+ *  ptr_and_size_t ptr_and_size = thrust::get_temporary_buffer<int>(device_sys, N);
+ *
+ *  // manipulate up to 100 ints
+ *  for(int i = 0; i < ptr_and_size.second; ++i)
+ *  {
+ *    *ptr_and_size.first = i;
+ *  }
+ *
+ *  // deallocate storage with thrust::return_temporary_buffer
+ *  thrust::return_temporary_buffer(device_sys, ptr_and_size.first);
+ *  \endcode
+ *
+ *  \see free
+ *  \see get_temporary_buffer
+ */
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+void return_temporary_buffer(const thrust::detail::execution_policy_base<DerivedPolicy> &system, Pointer p, std::ptrdiff_t n);
+
+
+/*! \p raw_pointer_cast creates a "raw" pointer from a pointer-like type,
+ *  simply returning the wrapped pointer, should it exist.
+ *
+ *  \param ptr The pointer of interest.
+ *  \return <tt>ptr.get()</tt>, if the expression is well formed; <tt>ptr</tt>, otherwise.
+ *  \see raw_reference_cast
+ */
+template<typename Pointer>
+__host__ __device__
+typename thrust::detail::pointer_traits<Pointer>::raw_pointer
+  raw_pointer_cast(Pointer ptr);
+
+
+/*! \p raw_reference_cast creates a "raw" reference from a wrapped reference type,
+ *  simply returning the underlying reference, should it exist.
+ *
+ *  If the argument is not a reference wrapper, the result is a reference to the argument.
+ *
+ *  \param ref The reference of interest.
+ *  \return <tt>*thrust::raw_pointer_cast(&ref)</tt>.
+ *  \note There are two versions of \p raw_reference_cast. One for <tt>const</tt> references,
+ *        and one for non-<tt>const</tt>.
+ *  \see raw_pointer_cast
+ */
+template<typename T>
+__host__ __device__
+typename detail::raw_reference<T>::type
+  raw_reference_cast(T &ref);
+
+
+/*! \p raw_reference_cast creates a "raw" reference from a wrapped reference type,
+ *  simply returning the underlying reference, should it exist.
+ *
+ *  If the argument is not a reference wrapper, the result is a reference to the argument.
+ *
+ *  \param ref The reference of interest.
+ *  \return <tt>*thrust::raw_pointer_cast(&ref)</tt>.
+ *  \note There are two versions of \p raw_reference_cast. One for <tt>const</tt> references,
+ *        and one for non-<tt>const</tt>.
+ *  \see raw_pointer_cast
+ */
+template<typename T>
+__host__ __device__
+typename detail::raw_reference<const T>::type
+  raw_reference_cast(const T &ref);
+
+/*! \} // memory_management
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/merge.h b/thirdparty/manifold/thirdparty/thrust/thrust/merge.h
new file mode 100644
index 000000000000..724f4c167f16
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/merge.h
@@ -0,0 +1,677 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file merge.h
+ *  \brief Merging sorted ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup merging Merging
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p merge combines two sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>
+ *  into a single sorted range. That is, it copies from <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt> into <tt>[result, result + (last1 - first1) + (last2 - first2))</tt>
+ *  such that the resulting range is in ascending order. \p merge is stable, meaning both that the
+ *  relative order of elements within each input range is preserved, and that for equivalent elements
+ *  in both input ranges the element from the first range precedes the element from the second. The
+ *  return value is <tt>result + (last1 - first1) + (last2 - first2)</tt>.
+ *
+ *  This version of \p merge compares elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the merged output.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge to compute the merger of two sorted sets of integers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[6] = {1, 3, 5, 7, 9, 11};
+ *  int A2[7] = {1, 1, 2, 3, 5,  8, 13};
+ *
+ *  int result[13];
+ *
+ *  int *result_end =
+ *    thrust::merge(thrust::host,
+ *                  A1, A1 + 6,
+ *                  A2, A2 + 7,
+ *                  result);
+ *  // result = {1, 1, 1, 2, 3, 3, 5, 5, 7, 8, 9, 11, 13}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/merge
+ *  \see \p set_union
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator merge(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result);
+
+
+/*! \p merge combines two sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>
+ *  into a single sorted range. That is, it copies from <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt> into <tt>[result, result + (last1 - first1) + (last2 - first2))</tt>
+ *  such that the resulting range is in ascending order. \p merge is stable, meaning both that the
+ *  relative order of elements within each input range is preserved, and that for equivalent elements
+ *  in both input ranges the element from the first range precedes the element from the second. The
+ *  return value is <tt>result + (last1 - first1) + (last2 - first2)</tt>.
+ *
+ *  This version of \p merge compares elements using \c operator<.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the merged output.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge to compute the merger of two sorted sets of integers.
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  ...
+ *  int A1[6] = {1, 3, 5, 7, 9, 11};
+ *  int A2[7] = {1, 1, 2, 3, 5,  8, 13};
+ *
+ *  int result[13];
+ *
+ *  int *result_end = thrust::merge(A1, A1 + 6, A2, A2 + 7, result);
+ *  // result = {1, 1, 1, 2, 3, 3, 5, 5, 7, 8, 9, 11, 13}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/merge
+ *  \see \p set_union
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator merge(InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result);
+
+
+/*! \p merge combines two sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>
+ *  into a single sorted range. That is, it copies from <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt> into <tt>[result, result + (last1 - first1) + (last2 - first2))</tt>
+ *  such that the resulting range is in ascending order. \p merge is stable, meaning both that the
+ *  relative order of elements within each input range is preserved, and that for equivalent elements
+ *  in both input ranges the element from the first range precedes the element from the second. The
+ *  return value is <tt>result + (last1 - first1) + (last2 - first2)</tt>.
+ *
+ *  This version of \p merge compares elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the merged output.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge to compute the merger of two sets of integers sorted in
+ *  descending order using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[6] = {11, 9, 7, 5, 3, 1};
+ *  int A2[7] = {13, 8, 5, 3, 2, 1, 1};
+ *
+ *  int result[13];
+ *
+ *  int *result_end = thrust::merge(thrust::host,
+ *                                  A1, A1 + 6,
+ *                                  A2, A2 + 7,
+ *                                  result,
+ *                                  thrust::greater<int>());
+ *  // result = {13, 11, 9, 8, 7, 5, 5, 3, 3, 2, 1, 1, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/merge
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+  OutputIterator merge(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result,
+                       StrictWeakCompare comp);
+
+
+/*! \p merge combines two sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>
+ *  into a single sorted range. That is, it copies from <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt> into <tt>[result, result + (last1 - first1) + (last2 - first2))</tt>
+ *  such that the resulting range is in ascending order. \p merge is stable, meaning both that the
+ *  relative order of elements within each input range is preserved, and that for equivalent elements
+ *  in both input ranges the element from the first range precedes the element from the second. The
+ *  return value is <tt>result + (last1 - first1) + (last2 - first2)</tt>.
+ *
+ *  This version of \p merge compares elements using a function object \p comp.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the merged output.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge to compute the merger of two sets of integers sorted in
+ *  descending order.
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A1[6] = {11, 9, 7, 5, 3, 1};
+ *  int A2[7] = {13, 8, 5, 3, 2, 1, 1};
+ *
+ *  int result[13];
+ *
+ *  int *result_end = thrust::merge(A1, A1 + 6, A2, A2 + 7, result, thrust::greater<int>());
+ *  // result = {13, 11, 9, 8, 7, 5, 5, 3, 3, 2, 1, 1, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/merge
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+  OutputIterator merge(InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result,
+                       StrictWeakCompare comp);
+
+
+/*! \p merge_by_key performs a key-value merge. That is, \p merge_by_key copies elements from
+ *  <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> into a single range,
+ *  <tt>[keys_result, keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending key order.
+ *
+ *  At the same time, \p merge_by_key copies elements from the two associated ranges <tt>[values_first1 + (keys_last1 - keys_first1))</tt>
+ *  and <tt>[values_first2 + (keys_last2 - keys_first2))</tt> into a single range,
+ *  <tt>[values_result, values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending order implied by each input element's associated key.
+ *
+ *  \p merge_by_key is stable, meaning both that the relative order of elements within each input range is
+ *  preserved, and that for equivalent elements in all input key ranges the element from the first range
+ *  precedes the element from the second.
+ *
+ *  The return value is is <tt>(keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>
+ *  and <tt>(values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the merged output range of keys.
+ *  \param values_result The beginning of the merged output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge_by_key to compute the merger of two sets of integers sorted in
+ *  ascending order using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {1, 3, 5, 7, 9, 11};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[7] = {1, 1, 2, 3, 5, 8, 13};
+ *  int B_vals[7] = {1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int keys_result[13];
+ *  int vals_result[13];
+ *
+ *  thrust::pair<int*,int*> end =
+ *    thrust::merge_by_key(thrust::host,
+ *                         A_keys, A_keys + 6,
+ *                         B_keys, B_keys + 7,
+ *                         A_vals, B_vals,
+ *                         keys_result, vals_result);
+ *
+ *  // keys_result = {1, 1, 1, 2, 3, 3, 5, 5, 7, 8, 9, 11, 13}
+ *  // vals_result = {0, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0,  0,  1}
+ *  \endcode
+ *
+ *  \see merge
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result);
+
+
+/*! \p merge_by_key performs a key-value merge. That is, \p merge_by_key copies elements from
+ *  <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> into a single range,
+ *  <tt>[keys_result, keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending key order.
+ *
+ *  At the same time, \p merge_by_key copies elements from the two associated ranges <tt>[values_first1 + (keys_last1 - keys_first1))</tt>
+ *  and <tt>[values_first2 + (keys_last2 - keys_first2))</tt> into a single range,
+ *  <tt>[values_result, values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending order implied by each input element's associated key.
+ *
+ *  \p merge_by_key is stable, meaning both that the relative order of elements within each input range is
+ *  preserved, and that for equivalent elements in all input key ranges the element from the first range
+ *  precedes the element from the second.
+ *
+ *  The return value is is <tt>(keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>
+ *  and <tt>(values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the merged output range of keys.
+ *  \param values_result The beginning of the merged output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge_by_key to compute the merger of two sets of integers sorted in
+ *  ascending order.
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A_keys[6] = {1, 3, 5, 7, 9, 11};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[7] = {1, 1, 2, 3, 5, 8, 13};
+ *  int B_vals[7] = {1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int keys_result[13];
+ *  int vals_result[13];
+ *
+ *  thrust::pair<int*,int*> end = thrust::merge_by_key(A_keys, A_keys + 6, B_keys, B_keys + 7, A_vals, B_vals, keys_result, vals_result);
+ *
+ *  // keys_result = {1, 1, 1, 2, 3, 3, 5, 5, 7, 8, 9, 11, 13}
+ *  // vals_result = {0, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0,  0,  1}
+ *  \endcode
+ *
+ *  \see merge
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result);
+
+
+/*! \p merge_by_key performs a key-value merge. That is, \p merge_by_key copies elements from
+ *  <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> into a single range,
+ *  <tt>[keys_result, keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending key order.
+ *
+ *  At the same time, \p merge_by_key copies elements from the two associated ranges <tt>[values_first1 + (keys_last1 - keys_first1))</tt>
+ *  and <tt>[values_first2 + (keys_last2 - keys_first2))</tt> into a single range,
+ *  <tt>[values_result, values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending order implied by each input element's associated key.
+ *
+ *  \p merge_by_key is stable, meaning both that the relative order of elements within each input range is
+ *  preserved, and that for equivalent elements in all input key ranges the element from the first range
+ *  precedes the element from the second.
+ *
+ *  The return value is is <tt>(keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>
+ *  and <tt>(values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>.
+ *
+ *  This version of \p merge_by_key compares key elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized using \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the merged output range of keys.
+ *  \param values_result The beginning of the merged output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator1's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator1's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge_by_key to compute the merger of two sets of integers sorted in
+ *  descending order using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {11, 9, 7, 5, 3, 1};
+ *  int A_vals[6] = { 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[7] = {13, 8, 5, 3, 2, 1, 1};
+ *  int B_vals[7] = { 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int keys_result[13];
+ *  int vals_result[13];
+ *
+ *  thrust::pair<int*,int*> end =
+ *    thrust::merge_by_key(thrust::host,
+ *                         A_keys, A_keys + 6,
+ *                         B_keys, B_keys + 7,
+ *                         A_vals, B_vals,
+ *                         keys_result, vals_result,
+ *                         thrust::greater<int>());
+ *
+ *  // keys_result = {13, 11, 9, 8, 7, 5, 5, 3, 3, 2, 1, 1, 1}
+ *  // vals_result = { 1,  0, 0, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1}
+ *  \endcode
+ *
+ *  \see merge
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2, typename Compare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result,
+                 Compare comp);
+
+
+/*! \p merge_by_key performs a key-value merge. That is, \p merge_by_key copies elements from
+ *  <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> into a single range,
+ *  <tt>[keys_result, keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending key order.
+ *
+ *  At the same time, \p merge_by_key copies elements from the two associated ranges <tt>[values_first1 + (keys_last1 - keys_first1))</tt>
+ *  and <tt>[values_first2 + (keys_last2 - keys_first2))</tt> into a single range,
+ *  <tt>[values_result, values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt> such that
+ *  the resulting range is in ascending order implied by each input element's associated key.
+ *
+ *  \p merge_by_key is stable, meaning both that the relative order of elements within each input range is
+ *  preserved, and that for equivalent elements in all input key ranges the element from the first range
+ *  precedes the element from the second.
+ *
+ *  The return value is is <tt>(keys_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>
+ *  and <tt>(values_result + (keys_last1 - keys_first1) + (keys_last2 - keys_first2))</tt>.
+ *
+ *  This version of \p merge_by_key compares key elements using a function object \p comp.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the merged output range of keys.
+ *  \param values_result The beginning of the merged output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator1's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator1's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use
+ *  \p merge_by_key to compute the merger of two sets of integers sorted in
+ *  descending order.
+ *
+ *  \code
+ *  #include <thrust/merge.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A_keys[6] = {11, 9, 7, 5, 3, 1};
+ *  int A_vals[6] = { 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[7] = {13, 8, 5, 3, 2, 1, 1};
+ *  int B_vals[7] = { 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int keys_result[13];
+ *  int vals_result[13];
+ *
+ *  thrust::pair<int*,int*> end = thrust::merge_by_key(A_keys, A_keys + 6, B_keys, B_keys + 7, A_vals, B_vals, keys_result, vals_result, thrust::greater<int>());
+ *
+ *  // keys_result = {13, 11, 9, 8, 7, 5, 5, 3, 3, 2, 1, 1, 1}
+ *  // vals_result = { 1,  0, 0, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1}
+ *  \endcode
+ *
+ *  \see merge
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2, typename StrictWeakCompare>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result,
+                 StrictWeakCompare comp);
+
+
+/*! \} // merging
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/merge.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mismatch.h b/thirdparty/manifold/thirdparty/thrust/thrust/mismatch.h
new file mode 100644
index 000000000000..bbdf2923a09e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mismatch.h
@@ -0,0 +1,257 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file mismatch.h
+ *  \brief Search for differences between ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+/*! \addtogroup searching
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ *  \p mismatch use different tests for whether elements differ.
+ *
+ *  This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
+ *  such that <tt>*i == *(first2 + (i - first1))</tt> is \c false. The return value is a
+ *  \c pair whose first element is \c i and whose second element is <tt>*(first2 + (i - first1))</tt>.
+ *  If no such iterator \c i exists, the return value is a \c pair whose first element
+ *  is \c last1 and whose second element is <tt>*(first2 + (last1 - first1))</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \return The first position where the sequences differ.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and \p InputIterator1's \c value_type is equality comparable to \p InputIterator2's \c value_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *
+ *  \code
+ *  #include <thrust/mismatch.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> vec1(4);
+ *  thrust::device_vector<int> vec2(4);
+ *
+ *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[1] = 5;  vec2[1] = 5;
+ *  vec1[2] = 3;  vec2[2] = 8;
+ *  vec1[3] = 7;  vec2[3] = 7;
+ *
+ *  typedef thrust::device_vector<int>::iterator Iterator;
+ *  thrust::pair<Iterator,Iterator> result;
+ *
+ *  result = thrust::mismatch(thrust::device, vec1.begin(), vec1.end(), vec2.begin());
+ *
+ *  // result.first  is vec1.begin() + 2
+ *  // result.second is vec2.begin() + 2
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                                      InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2);
+
+
+/*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
+ * and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ * \p mismatch use different tests for whether elements differ.
+ *
+ * This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
+ * such that <tt>*i == *(first2 + (i - first1))</tt> is \c false. The return value is a
+ * \c pair whose first element is \c i and whose second element is <tt>*(first2 + (i - first1))</tt>.
+ * If no such iterator \c i exists, the return value is a \c pair whose first element
+ * is \c last1 and whose second element is <tt>*(first2 + (last1 - first1))</tt>.
+ *
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \return The first position where the sequences differ.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and \p InputIterator1's \c value_type is equality comparable to \p InputIterator2's \c value_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *
+ *  \code
+ *  #include <thrust/mismatch.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> vec1(4);
+ *  thrust::device_vector<int> vec2(4);
+ *
+ *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[1] = 5;  vec2[1] = 5;
+ *  vec1[2] = 3;  vec2[2] = 8;
+ *  vec1[3] = 7;  vec2[3] = 7;
+ *
+ *  typedef thrust::device_vector<int>::iterator Iterator;
+ *  thrust::pair<Iterator,Iterator> result;
+ *
+ *  result = thrust::mismatch(vec1.begin(), vec1.end(), vec2.begin());
+ *
+ *  // result.first  is vec1.begin() + 2
+ *  // result.second is vec2.begin() + 2
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if
+ */
+template <typename InputIterator1, typename InputIterator2>
+thrust::pair<InputIterator1, InputIterator2> mismatch(InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2);
+
+
+/*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
+ *  and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ *  \p mismatch use different tests for whether elements differ.
+ *
+ *  This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
+ *  such that <tt>pred(\*i, \*(first2 + (i - first1))</tt> is \c false. The return value is a
+ *  \c pair whose first element is \c i and whose second element is <tt>*(first2 + (i - first1))</tt>.
+ *  If no such iterator \c i exists, the return value is a \c pair whose first element is
+ *  \c last1 and whose second element is <tt>*(first2 + (last1 - first1))</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param pred   The binary predicate to compare elements.
+ *  \return The first position where the sequences differ.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Input Iterator</a>.
+ *
+ *  \code
+ *  #include <thrust/mismatch.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> vec1(4);
+ *  thrust::device_vector<int> vec2(4);
+ *
+ *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[1] = 5;  vec2[1] = 5;
+ *  vec1[2] = 3;  vec2[2] = 8;
+ *  vec1[3] = 7;  vec2[3] = 7;
+ *
+ *  typedef thrust::device_vector<int>::iterator Iterator;
+ *  thrust::pair<Iterator,Iterator> result;
+ *
+ *  result = thrust::mismatch(thrust::device, vec1.begin(), vec1.end(), vec2.begin(), thrust::equal_to<int>());
+ *
+ *  // result.first  is vec1.begin() + 2
+ *  // result.second is vec2.begin() + 2
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+thrust::pair<InputIterator1, InputIterator2> mismatch(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                                      InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2,
+                                                      BinaryPredicate pred);
+
+
+/*! \p mismatch finds the first position where the two ranges <tt>[first1, last1)</tt>
+ * and <tt>[first2, first2 + (last1 - first1))</tt> differ. The two versions of 
+ * \p mismatch use different tests for whether elements differ.
+ *
+ * This version of \p mismatch finds the first iterator \c i in <tt>[first1, last1)</tt>
+ * such that <tt>pred(\*i, \*(first2 + (i - first1))</tt> is \c false. The return value is a
+ * \c pair whose first element is \c i and whose second element is <tt>*(first2 + (i - first1))</tt>.
+ * If no such iterator \c i exists, the return value is a \c pair whose first element is
+ * \c last1 and whose second element is <tt>*(first2 + (last1 - first1))</tt>.
+ *
+ *  \param first1 The beginning of the first sequence.
+ *  \param last1  The end of the first sequence.
+ *  \param first2 The beginning of the second sequence.
+ *  \param pred   The binary predicate to compare elements.
+ *  \return The first position where the sequences differ.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Input Iterator</a>.
+ *
+ *  \code
+ *  #include <thrust/mismatch.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> vec1(4);
+ *  thrust::device_vector<int> vec2(4);
+ *
+ *  vec1[0] = 0;  vec2[0] = 0; 
+ *  vec1[1] = 5;  vec2[1] = 5;
+ *  vec1[2] = 3;  vec2[2] = 8;
+ *  vec1[3] = 7;  vec2[3] = 7;
+ *
+ *  typedef thrust::device_vector<int>::iterator Iterator;
+ *  thrust::pair<Iterator,Iterator> result;
+ *
+ *  result = thrust::mismatch(vec1.begin(), vec1.end(), vec2.begin(), thrust::equal_to<int>());
+ *
+ *  // result.first  is vec1.begin() + 2
+ *  // result.second is vec2.begin() + 2
+ *  \endcode
+ *
+ *  \see find
+ *  \see find_if
+ */
+template <typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+thrust::pair<InputIterator1, InputIterator2> mismatch(InputIterator1 first1,
+                                                      InputIterator1 last1,
+                                                      InputIterator2 first2,
+                                                      BinaryPredicate pred);
+
+/*! \} // end searching
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/mismatch.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/allocator.h
new file mode 100644
index 000000000000..67adbe87c347
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/allocator.h
@@ -0,0 +1,253 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief Allocator types usable with \ref Memory Resources.
+ */
+
+#pragma once
+
+#include <limits>
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/config/exec_check_disable.h>
+#include <thrust/detail/config/memory_resource.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+#include <thrust/mr/validator.h>
+#include <thrust/mr/polymorphic_adaptor.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/*! \addtogroup allocators Allocators
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! An \p mr::allocator is a template that fulfills the C++ requirements for Allocators,
+ *  allowing to use the NPA-based memory resources where an Allocator is required. Unlike
+ *  memory resources, but like other allocators, \p mr::allocator is typed and bound to
+ *  allocate object of a specific type, however it can be freely rebound to other types.
+ *
+ *  \tparam T the type that will be allocated by this allocator.
+ *  \tparam MR the upstream memory resource to use for memory allocation. Must derive from
+ *      \p thrust::mr::memory_resource and must be \p final (in C++11 and beyond).
+ */
+template<typename T, class MR>
+class allocator : private validator<MR>
+{
+public:
+    /*! The pointer to void type of this allocator. */
+    typedef typename MR::pointer void_pointer;
+
+    /*! The value type allocated by this allocator. Equivalent to \p T. */
+    typedef T value_type;
+    /*! The pointer type allocated by this allocator. Equivaled to the pointer type of \p MR rebound to \p T. */
+    typedef typename thrust::detail::pointer_traits<void_pointer>::template rebind<T>::other pointer;
+    /*! The pointer to const type. Equivalent to a pointer type of \p MR rebound to <tt>const T</tt>. */
+    typedef typename thrust::detail::pointer_traits<void_pointer>::template rebind<const T>::other const_pointer;
+    /*! The reference to the type allocated by this allocator. Supports smart references. */
+    typedef typename thrust::detail::pointer_traits<pointer>::reference reference;
+    /*! The const reference to the type allocated by this allocator. Supports smart references. */
+    typedef typename thrust::detail::pointer_traits<const_pointer>::reference const_reference;
+    /*! The size type of this allocator. Always \p std::size_t. */
+    typedef std::size_t size_type;
+    /*! The difference type between pointers allocated by this allocator. */
+    typedef typename thrust::detail::pointer_traits<pointer>::difference_type difference_type;
+
+    /*! Specifies that the allocator shall be propagated on container copy assignment. */
+    typedef detail::true_type propagate_on_container_copy_assignment;
+    /*! Specifies that the allocator shall be propagated on container move assignment. */
+    typedef detail::true_type propagate_on_container_move_assignment;
+    /*! Specifies that the allocator shall be propagated on container swap. */
+    typedef detail::true_type propagate_on_container_swap;
+
+    /*! The \p rebind metafunction provides the type of an \p allocator instantiated with another type.
+     *
+     *  \tparam U the other type to use for instantiation.
+     */
+    template<typename U>
+    struct rebind
+    {
+        /*! The typedef \p other gives the type of the rebound \p allocator.
+         */
+        typedef allocator<U, MR> other;
+    };
+
+    /*! Calculates the maximum number of elements allocated by this allocator.
+     *
+     *  \return the maximum value of \p std::size_t, divided by the size of \p T.
+     */
+    __thrust_exec_check_disable__
+    __host__ __device__
+    size_type max_size() const
+    {
+        return (std::numeric_limits<size_type>::max)() / sizeof(T);
+    }
+
+    /*! Constructor.
+     *
+     *  \param resource the resource to be used to allocate raw memory.
+     */
+    __host__ __device__
+    allocator(MR * resource) : mem_res(resource)
+    {
+    }
+
+    /*! Copy constructor. Copies the resource pointer. */
+    template<typename U>
+    __host__ __device__
+    allocator(const allocator<U, MR> & other) : mem_res(other.resource())
+    {
+    }
+
+    /*! Allocates objects of type \p T.
+     *
+     *  \param n number of elements to allocate
+     *  \return a pointer to the newly allocated storage.
+     */
+    THRUST_NODISCARD
+    __host__
+    pointer allocate(size_type n)
+    {
+        return static_cast<pointer>(mem_res->do_allocate(n * sizeof(T), THRUST_ALIGNOF(T)));
+    }
+
+    /*! Deallocates objects of type \p T.
+     *
+     *  \param p pointer returned by a previous call to \p allocate
+     *  \param n number of elements, passed as an argument to the \p allocate call that produced \p p
+     */
+    __host__
+    void deallocate(pointer p, size_type n)
+    {
+        return mem_res->do_deallocate(p, n * sizeof(T), THRUST_ALIGNOF(T));
+    }
+
+    /*! Extracts the memory resource used by this allocator.
+     *
+     *  \return the memory resource used by this allocator.
+     */
+    __host__ __device__
+    MR * resource() const
+    {
+        return mem_res;
+    }
+
+private:
+    MR * mem_res;
+};
+
+/*! Compares the allocators for equality by comparing the underlying memory resources. */
+template<typename T, typename MR>
+__host__ __device__
+bool operator==(const allocator<T, MR> & lhs, const allocator<T, MR> & rhs) noexcept
+{
+    return *lhs.resource() == *rhs.resource();
+}
+
+/*! Compares the allocators for inequality by comparing the underlying memory resources. */
+template<typename T, typename MR>
+__host__ __device__
+bool operator!=(const allocator<T, MR> & lhs, const allocator<T, MR> & rhs) noexcept
+{
+    return !(lhs == rhs);
+}
+
+#if THRUST_CPP_DIALECT >= 2011
+
+template<typename T, typename Pointer>
+using polymorphic_allocator = allocator<T, polymorphic_adaptor_resource<Pointer> >;
+
+#else // C++11
+
+template<typename T, typename Pointer>
+class polymorphic_allocator : public allocator<T, polymorphic_adaptor_resource<Pointer> >
+{
+    typedef allocator<T, polymorphic_adaptor_resource<Pointer> > base;
+
+public:
+    /*! Initializes the base class with the parameter \p resource.
+     */
+    polymorphic_allocator(polymorphic_adaptor_resource<Pointer>  * resource) : base(resource)
+    {
+    }
+};
+
+#endif // C++11
+
+/*! A helper allocator class that uses global instances of a given upstream memory resource. Requires the memory resource
+ *      to be default constructible.
+ *
+ *  \tparam T the type that will be allocated by this allocator.
+ *  \tparam Upstream the upstream memory resource to use for memory allocation. Must derive from
+ *      \p thrust::mr::memory_resource and must be \p final (in C++11 and beyond).
+ */
+template<typename T, typename Upstream>
+class stateless_resource_allocator : public thrust::mr::allocator<T, Upstream>
+{
+    typedef thrust::mr::allocator<T, Upstream> base;
+
+public:
+    /*! The \p rebind metafunction provides the type of an \p stateless_resource_allocator instantiated with another type.
+     *
+     *  \tparam U the other type to use for instantiation.
+     */
+    template<typename U>
+    struct rebind
+    {
+        /*! The typedef \p other gives the type of the rebound \p stateless_resource_allocator.
+         */
+        typedef stateless_resource_allocator<U, Upstream> other;
+    };
+
+    /*! Default constructor. Uses \p get_global_resource to get the global instance of \p Upstream and initializes the
+     *      \p allocator base subobject with that resource.
+     */
+    __thrust_exec_check_disable__
+    __host__ __device__
+    stateless_resource_allocator() : base(get_global_resource<Upstream>())
+    {
+    }
+
+    /*! Copy constructor. Copies the memory resource pointer. */
+    __host__ __device__
+    stateless_resource_allocator(const stateless_resource_allocator & other)
+        : base(other) {}
+
+    /*! Conversion constructor from an allocator of a different type. Copies the memory resource pointer. */
+    template<typename U>
+    __host__ __device__
+    stateless_resource_allocator(const stateless_resource_allocator<U, Upstream> & other)
+        : base(other) {}
+
+#if THRUST_CPP_DIALECT >= 2011
+    stateless_resource_allocator & operator=(const stateless_resource_allocator &) = default;
+#endif
+
+    /*! Destructor. */
+    __host__ __device__
+    ~stateless_resource_allocator() {}
+};
+
+/*! \} // allocators
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/device_memory_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/device_memory_resource.h
new file mode 100644
index 000000000000..3a671142aaf9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/device_memory_resource.h
@@ -0,0 +1,38 @@
+/*
+ *  Copyright 2018-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include the device system's memory_resource header
+#define __THRUST_DEVICE_SYSTEM_MEMORY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/memory_resource.h>
+#include __THRUST_DEVICE_SYSTEM_MEMORY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_MEMORY_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+
+typedef thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::memory_resource
+    device_memory_resource;
+typedef thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::universal_memory_resource
+    universal_memory_resource;
+typedef thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::universal_host_pinned_memory_resource
+    universal_host_pinned_memory_resource;
+
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_pool.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_pool.h
new file mode 100644
index 000000000000..b00a8644ccad
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_pool.h
@@ -0,0 +1,489 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A caching and pooling memory resource adaptor which uses separate upstream resources for memory allocation
+ *      and bookkeeping.
+ */
+
+#pragma once
+
+#include <thrust/detail/algorithm_wrapper.h>
+#include <thrust/detail/config.h>
+
+#include <thrust/host_vector.h>
+#include <thrust/binary_search.h>
+#include <thrust/detail/seq.h>
+
+#include <thrust/mr/memory_resource.h>
+#include <thrust/mr/allocator.h>
+#include <thrust/mr/pool_options.h>
+
+#include <cassert>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/** \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! A memory resource adaptor allowing for pooling and caching allocations from \p Upstream, using \p Bookkeeper for
+ *      management of that cached and pooled memory, allowing to cache portions of memory inaccessible from the host.
+ *
+ *  On a typical memory resource, calls to \p allocate and \p deallocate actually allocate and deallocate memory. Pooling
+ *      memory resources only allocate and deallocate memory from an external resource (the upstream memory resource) when
+ *      there's no suitable memory currently cached; otherwise, they use memory they have acquired beforehand, to make
+ *      memory allocation faster and more efficient.
+ *
+ *  The disjoint version of the pool resources uses a separate upstream memory resource, \p Bookkeeper, to allocate memory
+ *      necessary to manage the cached memory. There may be many reasons to do that; the canonical one is that \p Upstream
+ *      allocates memory that is inaccessible to the code of the pool resource, which means that it cannot embed the necessary
+ *      information in memory obtained from \p Upstream; for instance, \p Upstream can be a CUDA non-managed memory
+ *      resource, or a CUDA managed memory resource whose memory we would prefer to not migrate back and forth between
+ *      host and device when executing bookkeeping code.
+ *
+ *  This is not the only case where it makes sense to use a disjoint pool resource, though. In a multi-core environment
+ *      it may be beneficial to avoid stealing cache lines from other cores by writing over bookkeeping information
+ *      embedded in an allocated block of memory. In such a case, one can imagine wanting to use a disjoint pool where
+ *      both the upstream and the bookkeeper are of the same type, to allocate memory consistently, but separately for
+ *      those two purposes.
+ *
+ *  \tparam Upstream the type of memory resources that will be used for allocating memory blocks to be handed off to the user
+ *  \tparam Bookkeeper the type of memory resources that will be used for allocating bookkeeping memory
+ */
+template<typename Upstream, typename Bookkeeper>
+class disjoint_unsynchronized_pool_resource final
+    : public memory_resource<typename Upstream::pointer>,
+        private validator2<Upstream, Bookkeeper>
+{
+public:
+    /*! Get the default options for a disjoint pool. These are meant to be a sensible set of values for many use cases,
+     *      and as such, may be tuned in the future. This function is exposed so that creating a set of options that are
+     *      just a slight departure from the defaults is easy.
+     */
+    static pool_options get_default_options()
+    {
+        pool_options ret;
+
+        ret.min_blocks_per_chunk = 16;
+        ret.min_bytes_per_chunk = 1024;
+        ret.max_blocks_per_chunk = static_cast<std::size_t>(1) << 20;
+        ret.max_bytes_per_chunk = static_cast<std::size_t>(1) << 30;
+
+        ret.smallest_block_size = THRUST_MR_DEFAULT_ALIGNMENT;
+        ret.largest_block_size = static_cast<std::size_t>(1) << 20;
+
+        ret.alignment = THRUST_MR_DEFAULT_ALIGNMENT;
+
+        ret.cache_oversized = true;
+
+        ret.cached_size_cutoff_factor = 16;
+        ret.cached_alignment_cutoff_factor = 16;
+
+        return ret;
+    }
+
+    /*! Constructor.
+     *
+     *  \param upstream the upstream memory resource for allocations
+     *  \param bookkeeper the upstream memory resource for bookkeeping
+     *  \param options pool options to use
+     */
+    disjoint_unsynchronized_pool_resource(Upstream * upstream, Bookkeeper * bookkeeper,
+        pool_options options = get_default_options())
+        : m_upstream(upstream),
+        m_bookkeeper(bookkeeper),
+        m_options(options),
+        m_smallest_block_log2(detail::log2_ri(m_options.smallest_block_size)),
+        m_pools(m_bookkeeper),
+        m_allocated(m_bookkeeper),
+        m_cached_oversized(m_bookkeeper),
+        m_oversized(m_bookkeeper)
+    {
+        assert(m_options.validate());
+
+        pointer_vector free(m_bookkeeper);
+        pool p(free);
+        m_pools.resize(detail::log2_ri(m_options.largest_block_size) - m_smallest_block_log2 + 1, p);
+    }
+
+    // TODO: C++11: use delegating constructors
+
+    /*! Constructor. Upstream and bookkeeping resources are obtained by calling \p get_global_resource for their types.
+     *
+     *  \param options pool options to use
+     */
+    disjoint_unsynchronized_pool_resource(pool_options options = get_default_options())
+        : m_upstream(get_global_resource<Upstream>()),
+        m_bookkeeper(get_global_resource<Bookkeeper>()),
+        m_options(options),
+        m_smallest_block_log2(detail::log2_ri(m_options.smallest_block_size)),
+        m_pools(m_bookkeeper),
+        m_allocated(m_bookkeeper),
+        m_cached_oversized(m_bookkeeper),
+        m_oversized(m_bookkeeper)
+    {
+        assert(m_options.validate());
+
+        pointer_vector free(m_bookkeeper);
+        pool p(free);
+        m_pools.resize(detail::log2_ri(m_options.largest_block_size) - m_smallest_block_log2 + 1, p);
+    }
+
+    /*! Destructor. Releases all held memory to upstream.
+     */
+    ~disjoint_unsynchronized_pool_resource()
+    {
+        release();
+    }
+
+private:
+    typedef typename Upstream::pointer void_ptr;
+    typedef typename thrust::detail::pointer_traits<void_ptr>::template rebind<char>::other char_ptr;
+
+    struct chunk_descriptor
+    {
+        std::size_t size;
+        void_ptr pointer;
+    };
+
+    typedef thrust::host_vector<
+        chunk_descriptor,
+        allocator<chunk_descriptor, Bookkeeper>
+    > chunk_vector;
+
+    struct oversized_block_descriptor
+    {
+        std::size_t size;
+        std::size_t alignment;
+        void_ptr pointer;
+
+        __host__ __device__
+        bool operator==(const oversized_block_descriptor & other) const
+        {
+            return size == other.size && alignment == other.alignment && pointer == other.pointer;
+        }
+
+        __host__ __device__
+        bool operator<(const oversized_block_descriptor & other) const
+        {
+            return size < other.size || (size == other.size && alignment < other.alignment);
+        }
+    };
+
+    struct equal_pointers
+    {
+    public:
+        __host__ __device__
+        equal_pointers(void_ptr p) : p(p)
+        {
+        }
+
+        __host__ __device__
+        bool operator()(const oversized_block_descriptor & desc) const
+        {
+            return desc.pointer == p;
+        }
+
+    private:
+        void_ptr p;
+    };
+
+    struct matching_alignment
+    {
+    public:
+        __host__ __device__
+        matching_alignment(std::size_t requested) : requested(requested)
+        {
+        }
+
+        __host__ __device__
+        bool operator()(const oversized_block_descriptor & desc) const
+        {
+            return desc.alignment >= requested;
+        }
+
+    private:
+        std::size_t requested;
+    };
+
+    typedef thrust::host_vector<
+        oversized_block_descriptor,
+        allocator<oversized_block_descriptor, Bookkeeper>
+    > oversized_block_vector;
+
+    typedef thrust::host_vector<
+        void_ptr,
+        allocator<void_ptr, Bookkeeper>
+    > pointer_vector;
+
+    struct pool
+    {
+        __host__
+        pool(const pointer_vector & free)
+            : free_blocks(free),
+            previous_allocated_count(0)
+        {
+        }
+
+        __host__
+        pool(const pool & other)
+            : free_blocks(other.free_blocks),
+            previous_allocated_count(other.previous_allocated_count)
+        {
+        }
+
+#if THRUST_CPP_DIALECT >= 2011
+        pool & operator=(const pool &) = default;
+#endif
+
+        __host__
+        ~pool() {}
+
+        pointer_vector free_blocks;
+        std::size_t previous_allocated_count;
+    };
+
+    typedef thrust::host_vector<
+        pool,
+        allocator<pool, Bookkeeper>
+    > pool_vector;
+
+    Upstream * m_upstream;
+    Bookkeeper * m_bookkeeper;
+
+    pool_options m_options;
+    std::size_t m_smallest_block_log2;
+
+    // buckets containing free lists for each pooled size
+    pool_vector m_pools;
+    // list of all allocations from upstream for the above
+    chunk_vector m_allocated;
+    // list of all cached oversized/overaligned blocks that have been returned to the pool to cache
+    oversized_block_vector m_cached_oversized;
+    // list of all oversized/overaligned allocations from upstream
+    oversized_block_vector m_oversized;
+
+public:
+    /*! Releases all held memory to upstream.
+     */
+    void release()
+    {
+        // reset the buckets
+        for (std::size_t i = 0; i < m_pools.size(); ++i)
+        {
+            m_pools[i].free_blocks.clear();
+            m_pools[i].previous_allocated_count = 0;
+        }
+
+        // deallocate memory allocated for the buckets
+        for (std::size_t i = 0; i < m_allocated.size(); ++i)
+        {
+            m_upstream->do_deallocate(
+                m_allocated[i].pointer,
+                m_allocated[i].size,
+                m_options.alignment);
+        }
+
+        // deallocate cached oversized/overaligned memory
+        for (std::size_t i = 0; i < m_oversized.size(); ++i)
+        {
+            m_upstream->do_deallocate(
+                m_oversized[i].pointer,
+                m_oversized[i].size,
+                m_oversized[i].alignment);
+        }
+
+        m_allocated.clear();
+        m_oversized.clear();
+        m_cached_oversized.clear();
+    }
+
+    THRUST_NODISCARD virtual void_ptr do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        bytes = (std::max)(bytes, m_options.smallest_block_size);
+        assert(detail::is_power_of_2(alignment));
+
+        // an oversized and/or overaligned allocation requested; needs to be allocated separately
+        if (bytes > m_options.largest_block_size || alignment > m_options.alignment)
+        {
+            oversized_block_descriptor oversized;
+            oversized.size = bytes;
+            oversized.alignment = alignment;
+
+            if (m_options.cache_oversized && !m_cached_oversized.empty())
+            {
+                typename oversized_block_vector::iterator it = thrust::lower_bound(
+                    thrust::seq,
+                    m_cached_oversized.begin(),
+                    m_cached_oversized.end(),
+                    oversized);
+
+                // if the size is bigger than the requested size by a factor
+                // bigger than or equal to the specified cutoff for size,
+                // allocate a new block
+                if (it != m_cached_oversized.end())
+                {
+                    std::size_t size_factor = (*it).size / bytes;
+                    if (size_factor >= m_options.cached_size_cutoff_factor)
+                    {
+                        it = m_cached_oversized.end();
+                    }
+                }
+
+                if (it != m_cached_oversized.end() && (*it).alignment < alignment)
+                {
+                    it = find_if(it + 1, m_cached_oversized.end(), matching_alignment(alignment));
+                }
+
+                // if the alignment is bigger than the requested one by a factor
+                // bigger than or equal to the specified cutoff for alignment,
+                // allocate a new block
+                if (it != m_cached_oversized.end())
+                {
+                    std::size_t alignment_factor = (*it).alignment / alignment;
+                    if (alignment_factor >= m_options.cached_alignment_cutoff_factor)
+                    {
+                        it = m_cached_oversized.end();
+                    }
+                }
+
+                if (it != m_cached_oversized.end())
+                {
+                    oversized.pointer = (*it).pointer;
+                    m_cached_oversized.erase(it);
+                    return oversized.pointer;
+                }
+            }
+
+            // no fitting cached block found; allocate a new one that's just up to the specs
+            oversized.pointer = m_upstream->do_allocate(bytes, alignment);
+            m_oversized.push_back(oversized);
+
+            return oversized.pointer;
+        }
+
+        // the request is NOT for oversized and/or overaligned memory
+        // allocate a block from an appropriate bucket
+        std::size_t bytes_log2 = thrust::detail::log2_ri(bytes);
+        std::size_t bucket_idx = bytes_log2 - m_smallest_block_log2;
+        pool & bucket = m_pools[bucket_idx];
+
+        // if the free list of the bucket has no elements, allocate a new chunk
+        // and split it into blocks pushed to the free list
+        if (bucket.free_blocks.empty())
+        {
+            std::size_t bucket_size = static_cast<std::size_t>(1) << bytes_log2;
+
+            std::size_t n = bucket.previous_allocated_count;
+            if (n == 0)
+            {
+                n = m_options.min_blocks_per_chunk;
+                if (n < (m_options.min_bytes_per_chunk >> bytes_log2))
+                {
+                    n = m_options.min_bytes_per_chunk >> bytes_log2;
+                }
+            }
+            else
+            {
+                n = n * 3 / 2;
+                if (n > (m_options.max_bytes_per_chunk >> bytes_log2))
+                {
+                    n = m_options.max_bytes_per_chunk >> bytes_log2;
+                }
+                if (n > m_options.max_blocks_per_chunk)
+                {
+                    n = m_options.max_blocks_per_chunk;
+                }
+            }
+
+            bytes = n << bytes_log2;
+
+            assert(n >= m_options.min_blocks_per_chunk);
+            assert(n <= m_options.max_blocks_per_chunk);
+            assert(bytes >= m_options.min_bytes_per_chunk);
+            assert(bytes <= m_options.max_bytes_per_chunk);
+
+            chunk_descriptor allocated;
+            allocated.size = bytes;
+            allocated.pointer = m_upstream->do_allocate(bytes, m_options.alignment);
+            m_allocated.push_back(allocated);
+            bucket.previous_allocated_count = n;
+
+            for (std::size_t i = 0; i < n; ++i)
+            {
+                bucket.free_blocks.push_back(
+                    static_cast<void_ptr>(
+                        static_cast<char_ptr>(allocated.pointer) + i * bucket_size
+                    )
+                );
+            }
+        }
+
+        // allocate a block from the front of the bucket's free list
+        void_ptr ret = bucket.free_blocks.back();
+        bucket.free_blocks.pop_back();
+        return ret;
+    }
+
+    virtual void do_deallocate(void_ptr p, std::size_t n, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        n = (std::max)(n, m_options.smallest_block_size);
+        assert(detail::is_power_of_2(alignment));
+
+        // verify that the pointer is at least as aligned as claimed
+        assert(reinterpret_cast<detail::intmax_t>(detail::pointer_traits<void_ptr>::get(p)) % alignment == 0);
+
+        // the deallocated block is oversized and/or overaligned
+        if (n > m_options.largest_block_size || alignment > m_options.alignment)
+        {
+            typename oversized_block_vector::iterator it = find_if(m_oversized.begin(), m_oversized.end(), equal_pointers(p));
+            assert(it != m_oversized.end());
+
+            oversized_block_descriptor oversized = *it;
+
+            if (m_options.cache_oversized)
+            {
+                typename oversized_block_vector::iterator position = lower_bound(m_cached_oversized.begin(), m_cached_oversized.end(), oversized);
+                m_cached_oversized.insert(position, oversized);
+                return;
+            }
+
+            m_oversized.erase(it);
+
+            m_upstream->do_deallocate(p, oversized.size, oversized.alignment);
+
+            return;
+        }
+
+        // push the block to the front of the appropriate bucket's free list
+        std::size_t n_log2 = thrust::detail::log2_ri(n);
+        std::size_t bucket_idx = n_log2 - m_smallest_block_log2;
+        pool & bucket = m_pools[bucket_idx];
+
+        bucket.free_blocks.push_back(p);
+    }
+};
+
+/*! \} // memory_resource
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_sync_pool.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_sync_pool.h
new file mode 100644
index 000000000000..ed81ae4cb845
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_sync_pool.h
@@ -0,0 +1,117 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A mutex-synchronized version of \p disjoint_unsynchronized_pool_resource.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <mutex>
+
+#include <thrust/mr/disjoint_pool.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/*! \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! A mutex-synchronized version of \p disjoint_unsynchronized_pool_resource. Uses \p std::mutex, and therefore requires C++11.
+ *
+ *  \tparam Upstream the type of memory resources that will be used for allocating memory blocks to be handed off to the user
+ *  \tparam Bookkeeper the type of memory resources that will be used for allocating bookkeeping memory
+ */
+template<typename Upstream, typename Bookkeeper>
+struct disjoint_synchronized_pool_resource : public memory_resource<typename Upstream::pointer>
+{
+    typedef disjoint_unsynchronized_pool_resource<Upstream, Bookkeeper> unsync_pool;
+    typedef std::lock_guard<std::mutex> lock_t;
+
+    typedef typename Upstream::pointer void_ptr;
+
+public:
+    /*! Get the default options for a disjoint pool. These are meant to be a sensible set of values for many use cases,
+     *      and as such, may be tuned in the future. This function is exposed so that creating a set of options that are
+     *      just a slight departure from the defaults is easy.
+     */
+    static pool_options get_default_options()
+    {
+        return unsync_pool::get_default_options();
+    }
+
+    /*! Constructor.
+     *
+     *  \param upstream the upstream memory resource for allocations
+     *  \param bookkeeper the upstream memory resource for bookkeeping
+     *  \param options pool options to use
+     */
+    disjoint_synchronized_pool_resource(Upstream * upstream, Bookkeeper * bookkeeper,
+        pool_options options = get_default_options())
+        : upstream_pool(upstream, bookkeeper, options)
+    {
+    }
+
+    /*! Constructor. Upstream and bookkeeping resources are obtained by calling \p get_global_resource for their types.
+     *
+     *  \param options pool options to use
+     */
+    disjoint_synchronized_pool_resource(pool_options options = get_default_options())
+        : upstream_pool(get_global_resource<Upstream>(), get_global_resource<Bookkeeper>(), options)
+    {
+    }
+
+    /*! Releases all held memory to upstream.
+     */
+    void release()
+    {
+        lock_t lock(mtx);
+        upstream_pool.release();
+    }
+
+    THRUST_NODISCARD virtual void_ptr do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        lock_t lock(mtx);
+        return upstream_pool.do_allocate(bytes, alignment);
+    }
+
+    virtual void do_deallocate(void_ptr p, std::size_t n, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        lock_t lock(mtx);
+        upstream_pool.do_deallocate(p, n, alignment);
+    }
+
+private:
+    std::mutex mtx;
+    unsync_pool upstream_pool;
+};
+
+/*! \} // memory_resources
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_tls_pool.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_tls_pool.h
new file mode 100644
index 000000000000..9fc7917cae10
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/disjoint_tls_pool.h
@@ -0,0 +1,69 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file disjoint_tls_pool.h
+ *  \brief A function wrapping a thread local instance of a \p disjoint_unsynchronized_pool_resource.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/mr/disjoint_pool.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/*! \addtogroup memory_management Memory Management
+ *  \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_resources
+ *  \{
+ */
+
+/*! Potentially constructs, if not yet created, and then returns the address of a thread-local
+ *      \p disjoint_unsynchronized_pool_resource,
+ *
+ *  \tparam Upstream the first template argument to the pool template
+ *  \tparam Bookkeeper the second template argument to the pool template
+ *  \param upstream the first argument to the constructor, if invoked
+ *  \param bookkeeper the second argument to the constructor, if invoked
+ */
+template<typename Upstream, typename Bookkeeper>
+__host__
+thrust::mr::disjoint_unsynchronized_pool_resource<Upstream, Bookkeeper> & tls_disjoint_pool(
+    Upstream * upstream = NULL,
+    Bookkeeper * bookkeeper = NULL)
+{
+    static thread_local auto adaptor = [&]{
+        assert(upstream && bookkeeper);
+        return thrust::mr::disjoint_unsynchronized_pool_resource<Upstream, Bookkeeper>(upstream, bookkeeper);
+    }();
+
+    return adaptor;
+}
+
+/*! \}
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/fancy_pointer_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/fancy_pointer_resource.h
new file mode 100644
index 000000000000..b88107564096
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/fancy_pointer_resource.h
@@ -0,0 +1,61 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+#include <thrust/mr/memory_resource.h>
+#include <thrust/mr/validator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+template<typename Upstream, typename Pointer>
+class fancy_pointer_resource final : public memory_resource<Pointer>, private validator<Upstream>
+{
+public:
+    fancy_pointer_resource() : m_upstream(get_global_resource<Upstream>())
+    {
+    }
+
+    fancy_pointer_resource(Upstream * upstream) : m_upstream(upstream)
+    {
+    }
+
+    THRUST_NODISCARD
+    virtual Pointer do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        return static_cast<Pointer>(m_upstream->do_allocate(bytes, alignment));
+    }
+
+    virtual void do_deallocate(Pointer p, std::size_t bytes, std::size_t alignment) override
+    {
+        return m_upstream->do_deallocate(
+            static_cast<typename Upstream::pointer>(
+                thrust::detail::pointer_traits<Pointer>::get(p)),
+            bytes, alignment);
+    }
+
+private:
+    Upstream * m_upstream;
+};
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/host_memory_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/host_memory_resource.h
new file mode 100644
index 000000000000..9359a97a7de7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/host_memory_resource.h
@@ -0,0 +1,32 @@
+/*
+ *  Copyright 2018-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include the host system's memory_resource header
+#define __THRUST_HOST_SYSTEM_MEMORY_HEADER <__THRUST_HOST_SYSTEM_ROOT/memory_resource.h>
+#include __THRUST_HOST_SYSTEM_MEMORY_HEADER
+#undef __THRUST_HOST_SYSTEM_MEMORY_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+typedef thrust::system::__THRUST_HOST_SYSTEM_NAMESPACE::memory_resource
+    host_memory_resource;
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/memory_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/memory_resource.h
new file mode 100644
index 000000000000..6af2f167c646
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/memory_resource.h
@@ -0,0 +1,217 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief A base class for the memory resource system, similar to
+ *  std::memory_resource, and related utilities.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/config/memory_resource.h>
+#ifdef THRUST_MR_STD_MR_HEADER
+#  include THRUST_MR_STD_MR_HEADER
+#endif
+
+THRUST_NAMESPACE_BEGIN
+/*! \brief \p thrust::mr is the namespace containing system agnostic types and functions for \p memory_resource related functionalities.
+ */
+namespace mr
+{
+
+/** \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! \p memory_resource is the base class for all other memory resources.
+ *
+ *  \tparam Pointer the pointer type that is allocated and deallocated by the memory resource
+ *      derived from this base class. If this is <tt>void *</tt>, this class derives from
+ *      <tt>std::pmr::memory_resource</tt>.
+ */
+template<typename Pointer = void *>
+class memory_resource
+{
+public:
+    /*! Alias for the template parameter.
+     */
+    typedef Pointer pointer;
+
+    /*! Virtual destructor, defaulted when possible.
+     */
+    virtual ~memory_resource() = default;
+
+    /*! Allocates memory of size at least \p bytes and alignment at least \p alignment.
+     *
+     *  \param bytes size, in bytes, that is requested from this allocation
+     *  \param alignment alignment that is requested from this allocation
+     *  \throws thrust::bad_alloc when no memory with requested size and alignment can be allocated.
+     *  \return A pointer to void to the newly allocated memory.
+     */
+    THRUST_NODISCARD
+    pointer allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT)
+    {
+        return do_allocate(bytes, alignment);
+    }
+
+    /*! Deallocates memory pointed to by \p p.
+     *
+     *  \param p pointer to be deallocated
+     *  \param bytes the size of the allocation. This must be equivalent to the value of \p bytes that
+     *      was passed to the allocation function that returned \p p.
+     *  \param alignment the alignment of the allocation. This must be equivalent to the value of \p alignment
+     *      that was passed to the allocation function that returned \p p.
+     */
+    void deallocate(pointer p, std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT)
+    {
+        do_deallocate(p, bytes, alignment);
+    }
+
+    /*! Compares this resource to the other one. The default implementation uses identity comparison,
+     *      which is often the right thing to do and doesn't require RTTI involvement.
+     *
+     *  \param other the other resource to compare this resource to
+     *  \return whether the two resources are equivalent.
+     */
+    __host__ __device__
+    bool is_equal(const memory_resource & other) const noexcept
+    {
+        return do_is_equal(other);
+    }
+
+    /*! Allocates memory of size at least \p bytes and alignment at least \p alignment.
+     *
+     *  \param bytes size, in bytes, that is requested from this allocation
+     *  \param alignment alignment that is requested from this allocation
+     *  \throws thrust::bad_alloc when no memory with requested size and alignment can be allocated.
+     *  \return A pointer to void to the newly allocated memory.
+     */
+    virtual pointer do_allocate(std::size_t bytes, std::size_t alignment) = 0;
+
+    /*! Deallocates memory pointed to by \p p.
+     *
+     *  \param p pointer to be deallocated
+     *  \param bytes the size of the allocation. This must be equivalent to the value of \p bytes that
+     *      was passed to the allocation function that returned \p p.
+     *  \param alignment the size of the allocation. This must be equivalent to the value of \p alignment
+     *      that was passed to the allocation function that returned \p p.
+     */
+    virtual void do_deallocate(pointer p, std::size_t bytes, std::size_t alignment) = 0;
+
+    /*! Compares this resource to the other one. The default implementation uses identity comparison,
+     *      which is often the right thing to do and doesn't require RTTI involvement.
+     *
+     *  \param other the other resource to compare this resource to
+     *  \return whether the two resources are equivalent.
+     */
+    __host__ __device__
+    virtual bool do_is_equal(const memory_resource & other) const noexcept
+    {
+        return this == &other;
+    }
+};
+
+template<>
+class memory_resource<void *>
+#ifdef THRUST_STD_MR_NS
+    : THRUST_STD_MR_NS::memory_resource
+#endif
+{
+public:
+    typedef void * pointer;
+
+    virtual ~memory_resource() = default;
+
+    THRUST_NODISCARD
+    pointer allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT)
+    {
+        return do_allocate(bytes, alignment);
+    }
+
+    void deallocate(pointer p, std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT)
+    {
+        do_deallocate(p, bytes, alignment);
+    }
+
+    __host__ __device__
+    bool is_equal(const memory_resource & other) const noexcept
+    {
+        return do_is_equal(other);
+    }
+
+    virtual pointer do_allocate(std::size_t bytes, std::size_t alignment) = 0;
+    virtual void do_deallocate(pointer p, std::size_t bytes, std::size_t alignment) = 0;
+    __host__ __device__
+    virtual bool do_is_equal(const memory_resource & other) const noexcept
+    {
+        return this == &other;
+    }
+
+#ifdef THRUST_STD_MR_NS
+    // the above do_is_equal is a different function than the one from the standard memory resource
+    // can't implement this reasonably without RTTI though; it's reasonable to assume false otherwise
+
+    virtual bool do_is_equal(const THRUST_STD_MR_NS::memory_resource & other) const noexcept override
+    {
+#  ifdef THRUST_HAS_DYNAMIC_CAST
+        auto mr_resource = dynamic_cast<memory_resource<> *>(&other);
+        return mr_resource && do_is_equal(*mr_resource);
+#  else
+        return this == &other;
+#  endif
+    }
+#endif
+};
+
+/*! Compares the memory resources for equality, first by identity, then by \p is_equal.
+ */
+template<typename Pointer>
+__host__ __device__
+bool operator==(const memory_resource<Pointer> & lhs, const memory_resource<Pointer> & rhs) noexcept
+{
+    return &lhs == &rhs || rhs.is_equal(rhs);
+}
+
+/*! Compares the memory resources for inequality, first by identity, then by \p is_equal.
+ */
+template<typename Pointer>
+__host__ __device__
+bool operator!=(const memory_resource<Pointer> & lhs, const memory_resource<Pointer> & rhs) noexcept
+{
+    return !(lhs == rhs);
+}
+
+/*! Returns a global instance of \p MR, created as a function local static variable.
+ *
+ *  \tparam MR type of a memory resource to get an instance from. Must be \p DefaultConstructible.
+ *  \return a pointer to a global instance of \p MR.
+ */
+template<typename MR>
+__host__
+MR * get_global_resource()
+{
+    static MR resource;
+    return &resource;
+}
+
+/*! \} // memory_resource
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/new.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/new.h
new file mode 100644
index 000000000000..644e25169ed2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/new.h
@@ -0,0 +1,89 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Global operator new-based memory resource.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/mr/memory_resource.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/** \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! A memory resource that uses global operators new and delete to allocate and deallocate memory. Uses alignment-enabled
+ *      overloads when available, otherwise uses regular overloads and implements alignment requirements by itself.
+ */
+class new_delete_resource final : public memory_resource<>
+{
+public:
+    void * do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+#if defined(__cpp_aligned_new)
+        return ::operator new(bytes, std::align_val_t(alignment));
+#else
+        // allocate memory for bytes, plus potential alignment correction,
+        // plus store of the correction offset
+        void * p = ::operator new(bytes + alignment + sizeof(std::size_t));
+        std::size_t ptr_int = reinterpret_cast<std::size_t>(p);
+        // calculate the offset, i.e. how many bytes of correction was necessary
+        // to get an aligned pointer
+        std::size_t offset = (ptr_int % alignment) ? (alignment - ptr_int % alignment) : 0;
+        // calculate the return pointer
+        char * ptr = static_cast<char *>(p) + offset;
+        // store the offset right after the actually returned value
+        std::size_t * offset_store = reinterpret_cast<std::size_t *>(ptr + bytes);
+        *offset_store = offset;
+        return static_cast<void *>(ptr);
+#endif
+    }
+
+    void do_deallocate(void * p, std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+#if defined(__cpp_aligned_new)
+# if defined(__cpp_sized_deallocation)
+        ::operator delete(p, bytes, std::align_val_t(alignment));
+# else
+        (void)bytes;
+        ::operator delete(p, std::align_val_t(alignment));
+# endif
+#else
+        (void)alignment;
+        char * ptr = static_cast<char *>(p);
+        // calculate where the offset is stored
+        std::size_t * offset = reinterpret_cast<std::size_t *>(ptr + bytes);
+        // calculate the original pointer
+        p = static_cast<void *>(ptr - *offset);
+        ::operator delete(p);
+#endif
+    }
+};
+
+/*! \} // memory_resources
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/polymorphic_adaptor.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/polymorphic_adaptor.h
new file mode 100644
index 000000000000..0562a8f82698
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/polymorphic_adaptor.h
@@ -0,0 +1,57 @@
+/*
+ *  Copyright 2018-2019 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/mr/memory_resource.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+template<typename Pointer = void *>
+class polymorphic_adaptor_resource final : public memory_resource<Pointer>
+{
+public:
+    polymorphic_adaptor_resource(memory_resource<Pointer> * t) : upstream_resource(t)
+    {
+    }
+
+    virtual Pointer do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        return upstream_resource->allocate(bytes, alignment);
+    }
+
+    virtual void do_deallocate(Pointer p, std::size_t bytes, std::size_t alignment) override
+    {
+        return upstream_resource->deallocate(p, bytes, alignment);
+    }
+
+    __host__ __device__
+    virtual bool do_is_equal(const memory_resource<Pointer> & other) const noexcept override
+    {
+        return upstream_resource->is_equal(other);
+    }
+
+private:
+    memory_resource<Pointer> * upstream_resource;
+};
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/pool.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/pool.h
new file mode 100644
index 000000000000..6259a23f1726
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/pool.h
@@ -0,0 +1,507 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A caching and pooling memory resource adaptor which uses a single
+ *  upstream resource for memory allocation, and embeds bookkeeping information
+ *  in allocated blocks.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/algorithm_wrapper.h>
+
+#include <thrust/host_vector.h>
+
+#include <thrust/mr/memory_resource.h>
+#include <thrust/mr/allocator.h>
+#include <thrust/mr/pool_options.h>
+
+#include <cassert>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/** \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! A memory resource adaptor allowing for pooling and caching allocations from \p Upstream, using memory allocated
+ *      from it for both blocks then allocated to the user and for internal bookkeeping of the cached memory.
+ *
+ *  On a typical memory resource, calls to \p allocate and \p deallocate actually allocate and deallocate memory. Pooling
+ *      memory resources only allocate and deallocate memory from an external resource (the upstream memory resource) when
+ *      there's no suitable memory currently cached; otherwise, they use memory they have acquired beforehand, to make
+ *      memory allocation faster and more efficient.
+ *
+ *  The non-disjoint version of the pool resource uses a single upstream memory resource. Every allocation is larger than
+ *      strictly necessary to fulfill the end-user's request, because it needs to account for the memory overhead of tracking
+ *      the memory blocks and chunks inside those same memory regions. Nevertheless, this version should be more memory-efficient
+ *      than the \p disjoint_unsynchronized_pool_resource, because it doesn't need to allocate additional blocks of memory
+ *      from a separate resource, which in turn would necessitate the bookkeeping overhead in the upstream resource.
+ *
+ *  This version requires that memory allocated from Upstream is accessible from device. It supports smart references,
+ *      meaning that the non-managed CUDA resource, returning a device-tagged pointer, will work, but will be much less
+ *      efficient than the disjoint version, which wouldn't need to touch device memory at all, and therefore wouldn't need
+ *      to transfer it back and forth between the host and the device whenever an allocation or a deallocation happens.
+ *
+ *  \tparam Upstream the type of memory resources that will be used for allocating memory blocks
+ */
+template<typename Upstream>
+class unsynchronized_pool_resource final
+    : public memory_resource<typename Upstream::pointer>,
+        private validator<Upstream>
+{
+public:
+    /*! Get the default options for a pool. These are meant to be a sensible set of values for many use cases,
+     *      and as such, may be tuned in the future. This function is exposed so that creating a set of options that are
+     *      just a slight departure from the defaults is easy.
+     */
+    static pool_options get_default_options()
+    {
+        pool_options ret;
+
+        ret.min_blocks_per_chunk = 16;
+        ret.min_bytes_per_chunk = 1024;
+        ret.max_blocks_per_chunk = static_cast<std::size_t>(1) << 20;
+        ret.max_bytes_per_chunk = static_cast<std::size_t>(1) << 30;
+
+        ret.smallest_block_size = THRUST_MR_DEFAULT_ALIGNMENT;
+        ret.largest_block_size = static_cast<std::size_t>(1) << 20;
+
+        ret.alignment = THRUST_MR_DEFAULT_ALIGNMENT;
+
+        ret.cache_oversized = true;
+
+        ret.cached_size_cutoff_factor = 16;
+        ret.cached_alignment_cutoff_factor = 16;
+
+        return ret;
+    }
+
+    /*! Constructor.
+     *
+     *  \param upstream the upstream memory resource for allocations
+     *  \param options pool options to use
+     */
+    unsynchronized_pool_resource(Upstream * upstream, pool_options options = get_default_options())
+        : m_upstream(upstream),
+        m_options(options),
+        m_smallest_block_log2(detail::log2_ri(m_options.smallest_block_size)),
+        m_pools(upstream),
+        m_allocated(),
+        m_oversized(),
+        m_cached_oversized()
+    {
+        assert(m_options.validate());
+
+        pool p = { block_descriptor_ptr(), 0 };
+        m_pools.resize(detail::log2_ri(m_options.largest_block_size) - m_smallest_block_log2 + 1, p);
+    }
+
+    // TODO: C++11: use delegating constructors
+
+    /*! Constructor. The upstream resource is obtained by calling \p get_global_resource<Upstream>.
+     *
+     *  \param options pool options to use
+     */
+    unsynchronized_pool_resource(pool_options options = get_default_options())
+        : m_upstream(get_global_resource<Upstream>()),
+        m_options(options),
+        m_smallest_block_log2(detail::log2_ri(m_options.smallest_block_size)),
+        m_pools(get_global_resource<Upstream>()),
+        m_allocated(),
+        m_oversized(),
+        m_cached_oversized()
+    {
+        assert(m_options.validate());
+
+        pool p = { block_descriptor_ptr(), 0 };
+        m_pools.resize(detail::log2_ri(m_options.largest_block_size) - m_smallest_block_log2 + 1, p);
+    }
+
+    /*! Destructor. Releases all held memory to upstream.
+     */
+    ~unsynchronized_pool_resource()
+    {
+        release();
+    }
+
+private:
+    typedef typename Upstream::pointer void_ptr;
+    typedef typename thrust::detail::pointer_traits<void_ptr>::template rebind<char>::other char_ptr;
+
+    struct block_descriptor;
+    struct chunk_descriptor;
+    struct oversized_block_descriptor;
+
+    typedef typename thrust::detail::pointer_traits<void_ptr>::template rebind<block_descriptor>::other block_descriptor_ptr;
+    typedef typename thrust::detail::pointer_traits<void_ptr>::template rebind<chunk_descriptor>::other chunk_descriptor_ptr;
+    typedef typename thrust::detail::pointer_traits<void_ptr>::template rebind<oversized_block_descriptor>::other oversized_block_descriptor_ptr;
+
+    struct block_descriptor
+    {
+        block_descriptor_ptr next;
+    };
+
+    struct chunk_descriptor
+    {
+        std::size_t size;
+        chunk_descriptor_ptr next;
+    };
+
+    // this was originally a forward list, but I made it a doubly linked list
+    // because that way deallocation when not caching is faster and doesn't require
+    // traversal of a linked list (it's still a forward list for the cached list,
+    // because allocation from that list already traverses)
+    //
+    // TODO: investigate whether it's better to have this be a doubly-linked list
+    // with fast do_deallocate when !m_options.cache_oversized, or to have this be
+    // a forward list and require traversal in do_deallocate
+    //
+    // I assume that it is better this way, but the additional pointer could
+    // potentially hurt? these are supposed to be oversized and/or overaligned,
+    // so they are kinda memory intensive already
+    struct oversized_block_descriptor
+    {
+        std::size_t size;
+        std::size_t alignment;
+        oversized_block_descriptor_ptr prev;
+        oversized_block_descriptor_ptr next;
+        oversized_block_descriptor_ptr next_cached;
+    };
+
+    struct pool
+    {
+        block_descriptor_ptr free_list;
+        std::size_t previous_allocated_count;
+    };
+
+    typedef thrust::host_vector<
+        pool,
+        allocator<pool, Upstream>
+    > pool_vector;
+
+    Upstream * m_upstream;
+
+    pool_options m_options;
+    std::size_t m_smallest_block_log2;
+
+    pool_vector m_pools;
+    chunk_descriptor_ptr m_allocated;
+    oversized_block_descriptor_ptr m_oversized;
+    oversized_block_descriptor_ptr m_cached_oversized;
+
+public:
+    /*! Releases all held memory to upstream.
+     */
+    void release()
+    {
+        // reset the buckets
+        for (std::size_t i = 0; i < m_pools.size(); ++i)
+        {
+            thrust::raw_reference_cast(m_pools[i]).free_list = block_descriptor_ptr();
+            thrust::raw_reference_cast(m_pools[i]).previous_allocated_count = 0;
+        }
+
+        // deallocate memory allocated for the buckets
+        while (detail::pointer_traits<chunk_descriptor_ptr>::get(m_allocated))
+        {
+            chunk_descriptor_ptr alloc = m_allocated;
+            m_allocated = thrust::raw_reference_cast(*m_allocated).next;
+
+            void_ptr p = static_cast<void_ptr>(
+                static_cast<char_ptr>(
+                    static_cast<void_ptr>(alloc)
+                ) - thrust::raw_reference_cast(*alloc).size
+            );
+            m_upstream->do_deallocate(p, thrust::raw_reference_cast(*alloc).size + sizeof(chunk_descriptor), m_options.alignment);
+        }
+
+        // deallocate cached oversized/overaligned memory
+        while (detail::pointer_traits<oversized_block_descriptor_ptr>::get(m_oversized))
+        {
+            oversized_block_descriptor_ptr alloc = m_oversized;
+            m_oversized = thrust::raw_reference_cast(*m_oversized).next;
+
+            void_ptr p = static_cast<void_ptr>(
+                static_cast<char_ptr>(
+                    static_cast<void_ptr>(alloc)
+                ) - thrust::raw_reference_cast(*alloc).size
+            );
+            m_upstream->do_deallocate(p, thrust::raw_reference_cast(*alloc).size + sizeof(oversized_block_descriptor), thrust::raw_reference_cast(*alloc).alignment);
+        }
+
+        m_cached_oversized = oversized_block_descriptor_ptr();
+    }
+
+    THRUST_NODISCARD virtual void_ptr do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        bytes = (std::max)(bytes, m_options.smallest_block_size);
+        assert(detail::is_power_of_2(alignment));
+
+        // an oversized and/or overaligned allocation requested; needs to be allocated separately
+        if (bytes > m_options.largest_block_size || alignment > m_options.alignment)
+        {
+            if (m_options.cache_oversized)
+            {
+                oversized_block_descriptor_ptr ptr = m_cached_oversized;
+                oversized_block_descriptor_ptr * previous = &m_cached_oversized;
+                while (detail::pointer_traits<oversized_block_descriptor_ptr>::get(ptr))
+                {
+                    oversized_block_descriptor desc = *ptr;
+                    bool is_good = desc.size >= bytes && desc.alignment >= alignment;
+
+                    // if the size is bigger than the requested size by a factor
+                    // bigger than or equal to the specified cutoff for size,
+                    // allocate a new block
+                    if (is_good)
+                    {
+                        std::size_t size_factor = desc.size / bytes;
+                        if (size_factor >= m_options.cached_size_cutoff_factor)
+                        {
+                            is_good = false;
+                        }
+                    }
+
+                    // if the alignment is bigger than the requested one by a factor
+                    // bigger than or equal to the specified cutoff for alignment,
+                    // allocate a new block
+                    if (is_good)
+                    {
+                        std::size_t alignment_factor = desc.alignment / alignment;
+                        if (alignment_factor >= m_options.cached_alignment_cutoff_factor)
+                        {
+                            is_good = false;
+                        }
+                    }
+
+                    if (is_good)
+                    {
+                        if (previous != &m_cached_oversized)
+                        {
+                            oversized_block_descriptor previous_desc = **previous;
+                            previous_desc.next_cached = desc.next_cached;
+                            **previous = previous_desc;
+                        }
+                        else
+                        {
+                            m_cached_oversized = desc.next_cached;
+                        }
+
+                        desc.next_cached = oversized_block_descriptor_ptr();
+                        *ptr = desc;
+
+                        return static_cast<void_ptr>(
+                            static_cast<char_ptr>(
+                                static_cast<void_ptr>(ptr)
+                            ) - desc.size
+                        );
+                    }
+
+                    previous = &thrust::raw_reference_cast(*ptr).next_cached;
+                    ptr = *previous;
+                }
+            }
+
+            // no fitting cached block found; allocate a new one that's just up to the specs
+            void_ptr allocated = m_upstream->do_allocate(bytes + sizeof(oversized_block_descriptor), alignment);
+            oversized_block_descriptor_ptr block = static_cast<oversized_block_descriptor_ptr>(
+                static_cast<void_ptr>(
+                    static_cast<char_ptr>(allocated) + bytes
+                )
+            );
+
+            oversized_block_descriptor desc;
+            desc.size = bytes;
+            desc.alignment = alignment;
+            desc.prev = oversized_block_descriptor_ptr();
+            desc.next = m_oversized;
+            desc.next_cached = oversized_block_descriptor_ptr();
+            *block = desc;
+            m_oversized = block;
+
+            if (detail::pointer_traits<oversized_block_descriptor_ptr>::get(desc.next))
+            {
+                oversized_block_descriptor next = *desc.next;
+                next.prev = block;
+                *desc.next = next;
+            }
+
+            return allocated;
+        }
+
+        // the request is NOT for oversized and/or overaligned memory
+        // allocate a block from an appropriate bucket
+        std::size_t bytes_log2 = thrust::detail::log2_ri(bytes);
+        std::size_t bucket_idx = bytes_log2 - m_smallest_block_log2;
+        pool & bucket = thrust::raw_reference_cast(m_pools[bucket_idx]);
+
+        bytes = static_cast<std::size_t>(1) << bytes_log2;
+
+        // if the free list of the bucket has no elements, allocate a new chunk
+        // and split it into blocks pushed to the free list
+        if (!detail::pointer_traits<block_descriptor_ptr>::get(bucket.free_list))
+        {
+            std::size_t n = bucket.previous_allocated_count;
+            if (n == 0)
+            {
+                n = m_options.min_blocks_per_chunk;
+                if (n < (m_options.min_bytes_per_chunk >> bytes_log2))
+                {
+                    n = m_options.min_bytes_per_chunk >> bytes_log2;
+                }
+            }
+            else
+            {
+                n = n * 3 / 2;
+                if (n > (m_options.max_bytes_per_chunk >> bytes_log2))
+                {
+                    n = m_options.max_bytes_per_chunk >> bytes_log2;
+                }
+                if (n > m_options.max_blocks_per_chunk)
+                {
+                    n = m_options.max_blocks_per_chunk;
+                }
+            }
+
+            std::size_t descriptor_size = (std::max)(sizeof(block_descriptor), m_options.alignment);
+            std::size_t block_size = bytes + descriptor_size;
+            block_size += m_options.alignment - block_size % m_options.alignment;
+            std::size_t chunk_size = block_size * n;
+
+            void_ptr allocated = m_upstream->do_allocate(chunk_size + sizeof(chunk_descriptor), m_options.alignment);
+            chunk_descriptor_ptr chunk = static_cast<chunk_descriptor_ptr>(
+                static_cast<void_ptr>(
+                    static_cast<char_ptr>(allocated) + chunk_size
+                )
+            );
+
+            chunk_descriptor chunk_desc;
+            chunk_desc.size = chunk_size;
+            chunk_desc.next = m_allocated;
+            *chunk = chunk_desc;
+            m_allocated = chunk;
+
+            for (std::size_t i = 0; i < n; ++i)
+            {
+                block_descriptor_ptr block = static_cast<block_descriptor_ptr>(
+                    static_cast<void_ptr>(
+                        static_cast<char_ptr>(allocated) + block_size * i + bytes
+                    )
+                );
+
+                block_descriptor block_desc;
+                block_desc.next = bucket.free_list;
+                *block = block_desc;
+                bucket.free_list = block;
+            }
+        }
+
+        // allocate a block from the front of the bucket's free list
+        block_descriptor_ptr block = bucket.free_list;
+        bucket.free_list = thrust::raw_reference_cast(*block).next;
+        return static_cast<void_ptr>(
+            static_cast<char_ptr>(
+                static_cast<void_ptr>(block)
+            ) - bytes
+        );
+    }
+
+    virtual void do_deallocate(void_ptr p, std::size_t n, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        n = (std::max)(n, m_options.smallest_block_size);
+        assert(detail::is_power_of_2(alignment));
+
+        // verify that the pointer is at least as aligned as claimed
+        assert(reinterpret_cast<detail::intmax_t>(detail::pointer_traits<void_ptr>::get(p)) % alignment == 0);
+
+        // the deallocated block is oversized and/or overaligned
+        if (n > m_options.largest_block_size || alignment > m_options.alignment)
+        {
+            oversized_block_descriptor_ptr block = static_cast<oversized_block_descriptor_ptr>(
+                static_cast<void_ptr>(
+                    static_cast<char_ptr>(p) + n
+                )
+            );
+
+            oversized_block_descriptor desc = *block;
+
+            if (m_options.cache_oversized)
+            {
+                desc.next_cached = m_cached_oversized;
+                *block = desc;
+                m_cached_oversized = block;
+
+                return;
+            }
+
+            if (!detail::pointer_traits<oversized_block_descriptor_ptr>::get(desc.prev))
+            {
+                assert(m_oversized == block);
+                m_oversized = desc.next;
+            }
+            else
+            {
+                oversized_block_descriptor prev = *desc.prev;
+                assert(prev.next == block);
+                prev.next = desc.next;
+                *desc.prev = prev;
+            }
+
+            if (detail::pointer_traits<oversized_block_descriptor_ptr>::get(desc.next))
+            {
+                oversized_block_descriptor next = *desc.next;
+                assert(next.prev == block);
+                next.prev = desc.prev;
+                *desc.next = next;
+            }
+
+            m_upstream->do_deallocate(p, desc.size + sizeof(oversized_block_descriptor), desc.alignment);
+
+            return;
+        }
+
+        // push the block to the front of the appropriate bucket's free list
+        std::size_t n_log2 = thrust::detail::log2_ri(n);
+        std::size_t bucket_idx = n_log2 - m_smallest_block_log2;
+        pool & bucket = thrust::raw_reference_cast(m_pools[bucket_idx]);
+
+        n = static_cast<std::size_t>(1) << n_log2;
+
+        block_descriptor_ptr block = static_cast<block_descriptor_ptr>(
+            static_cast<void_ptr>(
+                static_cast<char_ptr>(p) + n
+            )
+        );
+
+        block_descriptor desc;
+        desc.next = bucket.free_list;
+        *block = desc;
+        bucket.free_list = block;
+    }
+};
+
+/*! \} // memory_resources
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/pool_options.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/pool_options.h
new file mode 100644
index 000000000000..13a8fe67400e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/pool_options.h
@@ -0,0 +1,128 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A type used by the pooling resource adaptors to fine-tune their
+ *  behavior.
+ */
+
+#pragma once
+
+#include <cstddef>
+
+#include <thrust/detail/integer_math.h>
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/config/memory_resource.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/*! \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! A type used for configuring pooling resource adaptors, to fine-tune their behavior and parameters.
+ */
+struct pool_options
+{
+    /*! The minimal number of blocks, i.e. pieces of memory handed off to the user from a pool of a given size, in a single
+     *      chunk allocated from upstream.
+     */
+    std::size_t min_blocks_per_chunk;
+    /*! The minimal number of bytes in a single chunk allocated from upstream.
+     */
+    std::size_t min_bytes_per_chunk;
+    /*! The maximal number of blocks, i.e. pieces of memory handed off to the user from a pool of a given size, in a single
+     *      chunk allocated from upstream.
+     */
+    std::size_t max_blocks_per_chunk;
+    /*! The maximal number of bytes in a single chunk allocated from upstream.
+     */
+    std::size_t max_bytes_per_chunk;
+
+    /*! The size of blocks in the smallest pool covered by the pool resource. All allocation requests below this size will
+     *      be rounded up to this size.
+     */
+    std::size_t smallest_block_size;
+    /*! The size of blocks in the largest pool covered by the pool resource. All allocation requests above this size will
+     *      be considered oversized, allocated directly from upstream (and not from a pool), and cached only of \p cache_oversized
+     *      is true.
+     */
+    std::size_t largest_block_size;
+
+    /*! The alignment of all blocks in internal pools of the pool resource. All allocation requests above this alignment
+     *      will be considered oversized, allocated directly from upstream (and not from a pool), and cached only of
+     *      \p cache_oversized is true.
+     */
+    std::size_t alignment;
+
+    /*! Decides whether oversized and overaligned blocks are cached for later use, or immediately return it to the upstream
+     *      resource.
+     */
+    bool cache_oversized;
+
+    /*! The size factor at which a cached allocation is considered too ridiculously oversized to use to fulfill an allocation
+     *      request. For instance: the user requests an allocation of size 1024 bytes. A block of size 32 * 1024 bytes is
+     *      cached. If \p cached_size_cutoff_factor is 32 or less, this block will be considered too big for that allocation
+     *      request.
+     */
+    std::size_t cached_size_cutoff_factor;
+    /*! The alignment factor at which a cached allocation is considered too ridiculously overaligned to use to fulfill an
+     *      allocation request. For instance: the user requests an allocation aligned to 32 bytes. A block aligned to 1024 bytes
+     *      is cached. If \p cached_size_cutoff_factor is 32 or less, this block will be considered too overaligned for that
+     *      allocation request.
+     */
+    std::size_t cached_alignment_cutoff_factor;
+
+    /*! Checks if the options are self-consistent.
+     *
+     *  /returns true if the options are self-consitent, false otherwise.
+     */
+    bool validate() const
+    {
+        if (!detail::is_power_of_2(smallest_block_size)) return false;
+        if (!detail::is_power_of_2(largest_block_size)) return false;
+        if (!detail::is_power_of_2(alignment)) return false;
+
+        if (max_bytes_per_chunk == 0 || max_blocks_per_chunk == 0) return false;
+        if (smallest_block_size == 0 || largest_block_size == 0) return false;
+
+        if (min_blocks_per_chunk > max_blocks_per_chunk) return false;
+        if (min_bytes_per_chunk > max_bytes_per_chunk) return false;
+
+        if (smallest_block_size > largest_block_size) return false;
+
+        if (min_blocks_per_chunk * smallest_block_size > max_bytes_per_chunk) return false;
+        if (min_blocks_per_chunk * largest_block_size > max_bytes_per_chunk) return false;
+
+        if (max_blocks_per_chunk * largest_block_size < min_bytes_per_chunk) return false;
+        if (max_blocks_per_chunk * smallest_block_size < min_bytes_per_chunk) return false;
+
+        if (alignment > smallest_block_size) return false;
+
+        return true;
+    }
+};
+
+/*! \} // memory_resources
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/sync_pool.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/sync_pool.h
new file mode 100644
index 000000000000..46c0e8441370
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/sync_pool.h
@@ -0,0 +1,114 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file 
+ *  \brief A mutex-synchronized version of \p unsynchronized_pool_resource.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <mutex>
+
+#include <thrust/mr/pool.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/*! \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! A mutex-synchronized version of \p unsynchronized_pool_resource. Uses \p std::mutex, and therefore requires C++11.
+ *
+ *  \tparam Upstream the type of memory resources that will be used for allocating memory
+ */
+template<typename Upstream>
+struct synchronized_pool_resource : public memory_resource<typename Upstream::pointer>
+{
+    typedef unsynchronized_pool_resource<Upstream> unsync_pool;
+    typedef std::lock_guard<std::mutex> lock_t;
+
+    typedef typename Upstream::pointer void_ptr;
+
+public:
+    /*! Get the default options for a pool. These are meant to be a sensible set of values for many use cases,
+     *      and as such, may be tuned in the future. This function is exposed so that creating a set of options that are
+     *      just a slight departure from the defaults is easy.
+     */
+    static pool_options get_default_options()
+    {
+        return unsync_pool::get_default_options();
+    }
+
+    /*! Constructor.
+     *
+     *  \param upstream the upstream memory resource for allocations
+     *  \param options pool options to use
+     */
+    synchronized_pool_resource(Upstream * upstream, pool_options options = get_default_options())
+        : upstream_pool(upstream, options)
+    {
+    }
+
+    /*! Constructor. The upstream resource is obtained by calling \p get_global_resource<Upstream>.
+     *
+     *  \param options pool options to use
+     */
+    synchronized_pool_resource(pool_options options = get_default_options())
+        : upstream_pool(get_global_resource<Upstream>(), options)
+    {
+    }
+
+    /*! Releases all held memory to upstream.
+     */
+    void release()
+    {
+        lock_t lock(mtx);
+        upstream_pool.release();
+    }
+
+    THRUST_NODISCARD virtual void_ptr do_allocate(std::size_t bytes, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        lock_t lock(mtx);
+        return upstream_pool.do_allocate(bytes, alignment);
+    }
+
+    virtual void do_deallocate(void_ptr p, std::size_t n, std::size_t alignment = THRUST_MR_DEFAULT_ALIGNMENT) override
+    {
+        lock_t lock(mtx);
+        upstream_pool.do_deallocate(p, n, alignment);
+    }
+
+private:
+    std::mutex mtx;
+    unsync_pool upstream_pool;
+};
+
+/*! \} // memory_resources
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/tls_pool.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/tls_pool.h
new file mode 100644
index 000000000000..8ee8127a3d7a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/tls_pool.h
@@ -0,0 +1,64 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file tls_pool.h
+ *  \brief A function wrapping a thread local instance of a \p unsynchronized_pool_resource.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/mr/pool.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+/*! \addtogroup memory_management Memory Management
+ *  \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_resources
+ *  \{
+ */
+
+/*! Potentially constructs, if not yet created, and then returns the address of a thread-local \p unsynchronized_pool_resource,
+ *
+ *  \tparam Upstream the template argument to the pool template
+ *  \param upstream the argument to the constructor, if invoked
+ */
+template<typename Upstream, typename Bookkeeper>
+__host__
+thrust::mr::unsynchronized_pool_resource<Upstream> & tls_pool(Upstream * upstream = NULL)
+{
+    static thread_local auto adaptor = [&]{
+        assert(upstream);
+        return thrust::mr::unsynchronized_pool_resource<Upstream>(upstream);
+    }();
+
+    return adaptor;
+}
+
+/*! \}
+ */
+
+} // end mr
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/universal_memory_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/universal_memory_resource.h
new file mode 100644
index 000000000000..b7f1ebd6f468
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/universal_memory_resource.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2018-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/mr/device_memory_resource.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/mr/validator.h b/thirdparty/manifold/thirdparty/thrust/thrust/mr/validator.h
new file mode 100644
index 000000000000..10e964821a5b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/mr/validator.h
@@ -0,0 +1,51 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/config/memory_resource.h>
+#include <thrust/mr/memory_resource.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace mr
+{
+
+template<typename MR>
+struct validator
+{
+#if THRUST_CPP_DIALECT >= 2011
+  static_assert(
+    std::is_base_of<memory_resource<typename MR::pointer>, MR>::value,
+    "a type used as a memory resource must derive from memory_resource"
+  );
+#endif
+};
+
+template<typename T, typename U>
+struct validator2 : private validator<T>, private validator<U>
+{
+};
+
+template<typename T>
+struct validator2<T, T> : private validator<T>
+{
+};
+
+} // end mr
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/optional.h b/thirdparty/manifold/thirdparty/thrust/thrust/optional.h
new file mode 100644
index 000000000000..5850b6ea04ee
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/optional.h
@@ -0,0 +1,2876 @@
+///
+// optional - An implementation of std::optional with extensions
+// Written in 2017 by Sy Brand (@TartanLlama)
+//
+// To the extent possible under law, the author(s) have dedicated all
+// copyright and related and neighboring rights to this software to the
+// public domain worldwide. This software is distributed without any warranty.
+//
+// You should have received a copy of the CC0 Public Domain Dedication
+// along with this software. If not, see
+// <http://creativecommons.org/publicdomain/zero/1.0/>.
+///
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+#include <thrust/detail/type_traits.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/addressof.h>
+#include <thrust/swap.h>
+
+#define THRUST_OPTIONAL_VERSION_MAJOR 0
+#define THRUST_OPTIONAL_VERSION_MINOR 2
+
+#include <exception>
+#include <functional>
+#include <new>
+#include <type_traits>
+#include <utility>
+
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC && _MSC_VER == 1900)
+#define THRUST_OPTIONAL_MSVC2015
+#endif
+
+#if (defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ <= 9 &&              \
+     !defined(__clang__))
+#define THRUST_OPTIONAL_GCC49
+#endif
+
+#if (defined(__GNUC__) && __GNUC__ == 5 && __GNUC_MINOR__ <= 4 &&              \
+     !defined(__clang__))
+#define THRUST_OPTIONAL_GCC54
+#endif
+
+#if (defined(__GNUC__) && __GNUC__ == 5 && __GNUC_MINOR__ <= 5 &&              \
+     !defined(__clang__))
+#define THRUST_OPTIONAL_GCC55
+#endif
+
+#if (defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ <= 9 &&              \
+     !defined(__clang__))
+// GCC < 5 doesn't support overloading on const&& for member functions
+#define THRUST_OPTIONAL_NO_CONSTRR
+
+// GCC < 5 doesn't support some standard C++11 type traits
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_CONSTRUCTIBLE(T)                                     \
+  std::has_trivial_copy_constructor<T>::value
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_ASSIGNABLE(T) std::has_trivial_copy_assign<T>::value
+
+// GCC < 5 doesn't provide a way to emulate std::is_trivially_move_*,
+// so don't enable any optimizations that rely on them:
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_CONSTRUCTIBLE(T) false
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_ASSIGNABLE(T) false
+
+// This one will be different for GCC 5.7 if it's ever supported
+#define THRUST_OPTIONAL_IS_TRIVIALLY_DESTRUCTIBLE(T) std::is_trivially_destructible<T>::value
+
+// GCC 5 < v < 8 has a bug in is_trivially_copy_constructible which breaks std::vector
+// for non-copyable types
+#elif (defined(__GNUC__) && __GNUC__ < 8 &&                                                \
+     !defined(__clang__))
+#ifndef THRUST_GCC_LESS_8_TRIVIALLY_COPY_CONSTRUCTIBLE_MUTEX
+#define THRUST_GCC_LESS_8_TRIVIALLY_COPY_CONSTRUCTIBLE_MUTEX
+THRUST_NAMESPACE_BEGIN
+  namespace detail {
+      template<class T>
+      struct is_trivially_copy_constructible : std::is_trivially_copy_constructible<T>{};
+#ifdef _GLIBCXX_VECTOR
+      template<class T, class A>
+      struct is_trivially_copy_constructible<std::vector<T,A>>
+          : std::is_trivially_copy_constructible<T>{};
+#endif
+  }
+THRUST_NAMESPACE_END
+#endif
+
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_CONSTRUCTIBLE(T)                                     \
+    thrust::detail::is_trivially_copy_constructible<T>::value
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_ASSIGNABLE(T)                                        \
+  std::is_trivially_copy_assignable<T>::value
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_CONSTRUCTIBLE(T)                                     \
+  std::is_trivially_move_constructible<T>::value
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_ASSIGNABLE(T)                                        \
+  std::is_trivially_move_assignable<T>::value
+#define THRUST_OPTIONAL_IS_TRIVIALLY_DESTRUCTIBLE(T) std::is_trivially_destructible<T>::value
+#else
+
+// To support clang + old libstdc++ without type traits, check for equivalent
+// clang built-ins and use them if present. See note above
+// is_trivially_copyable_impl in
+// thrust/type_traits/is_trivially_relocatable.h for more details.
+
+#ifndef __has_feature
+#define __has_feature(x) 0
+#endif
+
+#if defined(__GLIBCXX__) && __has_feature(is_trivially_constructible)
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_CONSTRUCTIBLE(T) \
+  __is_trivially_constructible(T, T const&)
+#else
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_CONSTRUCTIBLE(T) \
+  std::is_trivially_copy_constructible<T>::value
+#endif
+
+#if defined(__GLIBCXX__) && __has_feature(is_trivially_assignable)
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_ASSIGNABLE(T) \
+  __is_trivially_assignable(T, T const&)
+#else
+#define THRUST_OPTIONAL_IS_TRIVIALLY_COPY_ASSIGNABLE(T) \
+  std::is_trivially_copy_assignable<T>::value
+#endif
+
+#if defined(__GLIBCXX__) && __has_feature(is_trivially_constructible)
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_CONSTRUCTIBLE(T) \
+  __is_trivially_constructible(T, T&&)
+#else
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_CONSTRUCTIBLE(T) \
+  std::is_trivially_move_constructible<T>::value
+#endif
+
+#if defined(__GLIBCXX__) && __has_feature(is_trivially_assignable)
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_ASSIGNABLE(T) \
+  __is_trivially_assignable(T, T&&)
+#else
+#define THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_ASSIGNABLE(T) \
+  std::is_trivially_move_assignable<T>::value
+#endif
+
+#if defined(__GLIBCXX__) && __has_feature(is_trivially_destructible)
+#define THRUST_OPTIONAL_IS_TRIVIALLY_DESTRUCTIBLE(T) \
+  __is_trivially_destructible(T)
+#else
+#define THRUST_OPTIONAL_IS_TRIVIALLY_DESTRUCTIBLE(T) \
+  std::is_trivially_destructible<T>::value
+#endif
+
+#endif
+
+#if THRUST_CPP_DIALECT > 2011
+#define THRUST_OPTIONAL_CPP14
+#endif
+
+// constexpr implies const in C++11, not C++14
+#if (THRUST_CPP_DIALECT == 2011 || defined(THRUST_OPTIONAL_MSVC2015) ||                \
+     defined(THRUST_OPTIONAL_GCC49))
+/// \exclude
+#define THRUST_OPTIONAL_CPP11_CONSTEXPR
+#else
+/// \exclude
+#define THRUST_OPTIONAL_CPP11_CONSTEXPR constexpr
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+#ifndef THRUST_MONOSTATE_INPLACE_MUTEX
+#define THRUST_MONOSTATE_INPLACE_MUTEX
+/// \brief Used to represent an optional with no data; essentially a bool
+class monostate {};
+
+/// \brief A tag type to tell optional to construct its value in-place
+struct in_place_t {
+  explicit in_place_t() = default;
+};
+/// \brief A tag to tell optional to construct its value in-place
+static constexpr in_place_t in_place{};
+#endif
+
+template <class T> class optional;
+
+/// \exclude
+namespace detail {
+#ifndef THRUST_TRAITS_MUTEX
+#define THRUST_TRAITS_MUTEX
+// C++14-style aliases for brevity
+template <class T> using remove_const_t = typename std::remove_const<T>::type;
+template <class T>
+using remove_reference_t = typename std::remove_reference<T>::type;
+template <class T> using decay_t = typename std::decay<T>::type;
+template <bool E, class T = void>
+using enable_if_t = typename std::enable_if<E, T>::type;
+template <bool B, class T, class F>
+using conditional_t = typename std::conditional<B, T, F>::type;
+
+// std::conjunction from C++17
+template <class...> struct conjunction : std::true_type {};
+template <class B> struct conjunction<B> : B {};
+template <class B, class... Bs>
+struct conjunction<B, Bs...>
+    : std::conditional<bool(B::value), conjunction<Bs...>, B>::type {};
+
+#if defined(_LIBCPP_VERSION) && THRUST_CPP_DIALECT == 2011
+#define THRUST_OPTIONAL_LIBCXX_MEM_FN_WORKAROUND
+#endif
+
+// In C++11 mode, there's an issue in libc++'s std::mem_fn
+// which results in a hard-error when using it in a noexcept expression
+// in some cases. This is a check to workaround the common failing case.
+#ifdef THRUST_OPTIONAL_LIBCXX_MEM_FN_WORKAROUND
+template <class T> struct is_pointer_to_non_const_member_func : std::false_type{};
+template <class T, class Ret, class... Args>
+struct is_pointer_to_non_const_member_func<Ret (T::*) (Args...)> : std::true_type{};
+template <class T, class Ret, class... Args>
+struct is_pointer_to_non_const_member_func<Ret (T::*) (Args...)&> : std::true_type{};
+template <class T, class Ret, class... Args>
+struct is_pointer_to_non_const_member_func<Ret (T::*) (Args...)&&> : std::true_type{};
+template <class T, class Ret, class... Args>
+struct is_pointer_to_non_const_member_func<Ret (T::*) (Args...) volatile> : std::true_type{};
+template <class T, class Ret, class... Args>
+struct is_pointer_to_non_const_member_func<Ret (T::*) (Args...) volatile&> : std::true_type{};
+template <class T, class Ret, class... Args>
+struct is_pointer_to_non_const_member_func<Ret (T::*) (Args...) volatile&&> : std::true_type{};
+
+template <class T> struct is_const_or_const_ref : std::false_type{};
+template <class T> struct is_const_or_const_ref<T const&> : std::true_type{};
+template <class T> struct is_const_or_const_ref<T const> : std::true_type{};
+#endif
+
+// std::invoke from C++17
+// https://stackoverflow.com/questions/38288042/c11-14-invoke-workaround
+__thrust_exec_check_disable__
+template <typename Fn, typename... Args,
+#ifdef THRUST_OPTIONAL_LIBCXX_MEM_FN_WORKAROUND
+          typename = enable_if_t<!(is_pointer_to_non_const_member_func<Fn>::value
+                                 && is_const_or_const_ref<Args...>::value)>,
+#endif
+          typename = enable_if_t<std::is_member_pointer<decay_t<Fn>>::value>,
+          int = 0>
+__host__ __device__
+constexpr auto invoke(Fn &&f, Args &&... args)
+  noexcept(noexcept(std::mem_fn(f)(std::forward<Args>(args)...)))
+  THRUST_TRAILING_RETURN(decltype(std::mem_fn(f)(std::forward<Args>(args)...)))
+{
+  return std::mem_fn(f)(std::forward<Args>(args)...);
+}
+
+__thrust_exec_check_disable__
+template <typename Fn, typename... Args,
+          typename = enable_if_t<!std::is_member_pointer<decay_t<Fn>>::value>>
+__host__ __device__
+constexpr auto invoke(Fn &&f, Args &&... args)
+  noexcept(noexcept(std::forward<Fn>(f)(std::forward<Args>(args)...)))
+  THRUST_TRAILING_RETURN(decltype(std::forward<Fn>(f)(std::forward<Args>(args)...)))
+{
+  return std::forward<Fn>(f)(std::forward<Args>(args)...);
+}
+#endif
+
+// std::void_t from C++17
+template <class...> struct voider { using type = void; };
+template <class... Ts> using void_t = typename voider<Ts...>::type;
+
+// Trait for checking if a type is a thrust::optional
+template <class T> struct is_optional_impl : std::false_type {};
+template <class T> struct is_optional_impl<optional<T>> : std::true_type {};
+template <class T> using is_optional = is_optional_impl<decay_t<T>>;
+
+// Change void to thrust::monostate
+template <class U>
+using fixup_void = conditional_t<std::is_void<U>::value, monostate, U>;
+
+template <class F, class U, class = invoke_result_t<F, U>>
+using get_map_return = optional<fixup_void<invoke_result_t<F, U>>>;
+
+// Check if invoking F for some Us returns void
+template <class F, class = void, class... U> struct returns_void_impl;
+template <class F, class... U>
+struct returns_void_impl<F, void_t<invoke_result_t<F, U...>>, U...>
+    : std::is_void<invoke_result_t<F, U...>> {};
+template <class F, class... U>
+using returns_void = returns_void_impl<F, void, U...>;
+
+template <class T, class... U>
+using enable_if_ret_void = enable_if_t<returns_void<T &&, U...>::value>;
+
+template <class T, class... U>
+using disable_if_ret_void = enable_if_t<!returns_void<T &&, U...>::value>;
+
+template <class T, class U>
+using enable_forward_value =
+    detail::enable_if_t<std::is_constructible<T, U &&>::value &&
+                        !std::is_same<detail::decay_t<U>, in_place_t>::value &&
+                        !std::is_same<optional<T>, detail::decay_t<U>>::value>;
+
+template <class T, class U, class Other>
+using enable_from_other = detail::enable_if_t<
+    std::is_constructible<T, Other>::value &&
+    !std::is_constructible<T, optional<U> &>::value &&
+    !std::is_constructible<T, optional<U> &&>::value &&
+    !std::is_constructible<T, const optional<U> &>::value &&
+    !std::is_constructible<T, const optional<U> &&>::value &&
+    !std::is_convertible<optional<U> &, T>::value &&
+    !std::is_convertible<optional<U> &&, T>::value &&
+    !std::is_convertible<const optional<U> &, T>::value &&
+    !std::is_convertible<const optional<U> &&, T>::value>;
+
+template <class T, class U>
+using enable_assign_forward = detail::enable_if_t<
+    !std::is_same<optional<T>, detail::decay_t<U>>::value &&
+    !detail::conjunction<std::is_scalar<T>,
+                         std::is_same<T, detail::decay_t<U>>>::value &&
+    std::is_constructible<T, U>::value && std::is_assignable<T &, U>::value>;
+
+template <class T, class U, class Other>
+using enable_assign_from_other = detail::enable_if_t<
+    std::is_constructible<T, Other>::value &&
+    std::is_assignable<T &, Other>::value &&
+    !std::is_constructible<T, optional<U> &>::value &&
+    !std::is_constructible<T, optional<U> &&>::value &&
+    !std::is_constructible<T, const optional<U> &>::value &&
+    !std::is_constructible<T, const optional<U> &&>::value &&
+    !std::is_convertible<optional<U> &, T>::value &&
+    !std::is_convertible<optional<U> &&, T>::value &&
+    !std::is_convertible<const optional<U> &, T>::value &&
+    !std::is_convertible<const optional<U> &&, T>::value &&
+    !std::is_assignable<T &, optional<U> &>::value &&
+    !std::is_assignable<T &, optional<U> &&>::value &&
+    !std::is_assignable<T &, const optional<U> &>::value &&
+    !std::is_assignable<T &, const optional<U> &&>::value>;
+
+#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC
+// TODO make a version which works with MSVC
+template <class T, class U = T> struct is_swappable : std::true_type {};
+
+template <class T, class U = T> struct is_nothrow_swappable : std::true_type {};
+#else
+// https://stackoverflow.com/questions/26744589/what-is-a-proper-way-to-implement-is-swappable-to-test-for-the-swappable-concept
+namespace swap_adl_tests {
+// if swap ADL finds this then it would call std::swap otherwise (same
+// signature)
+struct tag {};
+
+template <class T> tag swap(T &, T &);
+template <class T, std::size_t N> tag swap(T (&a)[N], T (&b)[N]);
+
+// helper functions to test if an unqualified swap is possible, and if it
+// becomes std::swap
+template <class, class> std::false_type can_swap(...) noexcept(false);
+template <class T, class U,
+          class = decltype(swap(std::declval<T &>(), std::declval<U &>()))>
+std::true_type can_swap(int) noexcept(noexcept(swap(std::declval<T &>(),
+                                                    std::declval<U &>())));
+
+template <class, class> std::false_type uses_std(...);
+template <class T, class U>
+std::is_same<decltype(swap(std::declval<T &>(), std::declval<U &>())), tag>
+uses_std(int);
+
+template <class T>
+struct is_std_swap_noexcept
+    : std::integral_constant<bool,
+                             std::is_nothrow_move_constructible<T>::value &&
+                                 std::is_nothrow_move_assignable<T>::value> {};
+
+template <class T, std::size_t N>
+struct is_std_swap_noexcept<T[N]> : is_std_swap_noexcept<T> {};
+
+template <class T, class U>
+struct is_adl_swap_noexcept
+    : std::integral_constant<bool, noexcept(can_swap<T, U>(0))> {};
+} // namespace swap_adl_tests
+
+template <class T, class U = T>
+struct is_swappable
+    : std::integral_constant<
+          bool,
+          decltype(detail::swap_adl_tests::can_swap<T, U>(0))::value &&
+              (!decltype(detail::swap_adl_tests::uses_std<T, U>(0))::value ||
+               (std::is_move_assignable<T>::value &&
+                std::is_move_constructible<T>::value))> {};
+
+template <class T, std::size_t N>
+struct is_swappable<T[N], T[N]>
+    : std::integral_constant<
+          bool,
+          decltype(detail::swap_adl_tests::can_swap<T[N], T[N]>(0))::value &&
+              (!decltype(
+                   detail::swap_adl_tests::uses_std<T[N], T[N]>(0))::value ||
+               is_swappable<T, T>::value)> {};
+
+template <class T, class U = T>
+struct is_nothrow_swappable
+    : std::integral_constant<
+          bool,
+          is_swappable<T, U>::value &&
+              ((decltype(detail::swap_adl_tests::uses_std<T, U>(0))::value
+                    &&detail::swap_adl_tests::is_std_swap_noexcept<T>::value) ||
+               (!decltype(detail::swap_adl_tests::uses_std<T, U>(0))::value &&
+                    detail::swap_adl_tests::is_adl_swap_noexcept<T,
+                                                                 U>::value))> {
+};
+#endif
+
+// The storage base manages the actual storage, and correctly propagates
+// trivial destruction from T. This case is for when T is not trivially
+// destructible.
+template <class T, bool = ::std::is_trivially_destructible<T>::value>
+struct optional_storage_base {
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional_storage_base() noexcept
+      : m_dummy(), m_has_value(false) {}
+
+  __thrust_exec_check_disable__
+  template <class... U>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional_storage_base(in_place_t, U &&... u)
+      : m_value(std::forward<U>(u)...), m_has_value(true) {}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  ~optional_storage_base() {
+    if (m_has_value) {
+      m_value.~T();
+      m_has_value = false;
+    }
+  }
+
+  struct dummy {};
+  union {
+    dummy m_dummy;
+    T m_value;
+  };
+
+  bool m_has_value;
+};
+
+// This case is for when T is trivially destructible.
+template <class T> struct optional_storage_base<T, true> {
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional_storage_base() noexcept
+      : m_dummy(), m_has_value(false) {}
+
+  __thrust_exec_check_disable__
+  template <class... U>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional_storage_base(in_place_t, U &&... u)
+      : m_value(std::forward<U>(u)...), m_has_value(true) {}
+
+  // No destructor, so this class is trivially destructible
+
+  struct dummy {};
+  union {
+    dummy m_dummy;
+    T m_value;
+  };
+
+  bool m_has_value = false;
+};
+
+// This base class provides some handy member functions which can be used in
+// further derived classes
+template <class T> struct optional_operations_base : optional_storage_base<T> {
+  using optional_storage_base<T>::optional_storage_base;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  void hard_reset() noexcept {
+    get().~T();
+    this->m_has_value = false;
+  }
+
+  __thrust_exec_check_disable__
+  template <class... Args>
+  __host__ __device__
+  void construct(Args &&... args) noexcept {
+    new (thrust::addressof(this->m_value)) T(std::forward<Args>(args)...);
+    this->m_has_value = true;
+  }
+
+  __thrust_exec_check_disable__
+  template <class Opt>
+  __host__ __device__
+  void assign(Opt &&rhs) {
+    if (this->has_value()) {
+      if (rhs.has_value()) {
+        this->m_value = std::forward<Opt>(rhs).get();
+      } else {
+        this->m_value.~T();
+        this->m_has_value = false;
+      }
+    }
+
+    if (rhs.has_value()) {
+      construct(std::forward<Opt>(rhs).get());
+    }
+  }
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  bool has_value() const { return this->m_has_value; }
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &get() & { return this->m_value; }
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR const T &get() const & { return this->m_value; }
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &&get() && { return std::move(this->m_value); }
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T &&get() const && { return std::move(this->m_value); }
+#endif
+};
+
+// This class manages conditionally having a trivial copy constructor
+// This specialization is for when T is trivially copy constructible
+template <class T, bool = THRUST_OPTIONAL_IS_TRIVIALLY_COPY_CONSTRUCTIBLE(T)>
+struct optional_copy_base : optional_operations_base<T> {
+  using optional_operations_base<T>::optional_operations_base;
+};
+
+// This specialization is for when T is not trivially copy constructible
+template <class T>
+struct optional_copy_base<T, false> : optional_operations_base<T> {
+  using optional_operations_base<T>::optional_operations_base;
+
+  __thrust_exec_check_disable__
+  optional_copy_base() = default;
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional_copy_base(const optional_copy_base &rhs) {
+    if (rhs.has_value()) {
+      this->construct(rhs.get());
+    } else {
+      this->m_has_value = false;
+    }
+  }
+
+  __thrust_exec_check_disable__
+  optional_copy_base(optional_copy_base &&rhs) = default;
+  __thrust_exec_check_disable__
+  optional_copy_base &operator=(const optional_copy_base &rhs) = default;
+  __thrust_exec_check_disable__
+  optional_copy_base &operator=(optional_copy_base &&rhs) = default;
+};
+
+template <class T, bool = THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_CONSTRUCTIBLE(T)>
+struct optional_move_base : optional_copy_base<T> {
+  using optional_copy_base<T>::optional_copy_base;
+};
+template <class T> struct optional_move_base<T, false> : optional_copy_base<T> {
+  using optional_copy_base<T>::optional_copy_base;
+
+  __thrust_exec_check_disable__
+  optional_move_base() = default;
+  __thrust_exec_check_disable__
+  optional_move_base(const optional_move_base &rhs) = default;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional_move_base(optional_move_base &&rhs) noexcept(
+      std::is_nothrow_move_constructible<T>::value) {
+    if (rhs.has_value()) {
+      this->construct(std::move(rhs.get()));
+    } else {
+      this->m_has_value = false;
+    }
+  }
+  __thrust_exec_check_disable__
+  optional_move_base &operator=(const optional_move_base &rhs) = default;
+  __thrust_exec_check_disable__
+  optional_move_base &operator=(optional_move_base &&rhs) = default;
+};
+
+// This class manages conditionally having a trivial copy assignment operator
+template <class T, bool = THRUST_OPTIONAL_IS_TRIVIALLY_COPY_ASSIGNABLE(T) &&
+                          THRUST_OPTIONAL_IS_TRIVIALLY_COPY_CONSTRUCTIBLE(T) &&
+                          THRUST_OPTIONAL_IS_TRIVIALLY_DESTRUCTIBLE(T)>
+struct optional_copy_assign_base : optional_move_base<T> {
+  using optional_move_base<T>::optional_move_base;
+};
+
+template <class T>
+struct optional_copy_assign_base<T, false> : optional_move_base<T> {
+  using optional_move_base<T>::optional_move_base;
+
+  __thrust_exec_check_disable__
+  optional_copy_assign_base() = default;
+  __thrust_exec_check_disable__
+  optional_copy_assign_base(const optional_copy_assign_base &rhs) = default;
+
+  __thrust_exec_check_disable__
+  optional_copy_assign_base(optional_copy_assign_base &&rhs) = default;
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional_copy_assign_base &operator=(const optional_copy_assign_base &rhs) {
+    this->assign(rhs);
+    return *this;
+  }
+  __thrust_exec_check_disable__
+  optional_copy_assign_base &
+  operator=(optional_copy_assign_base &&rhs) = default;
+};
+
+template <class T,
+          bool = THRUST_OPTIONAL_IS_TRIVIALLY_DESTRUCTIBLE(T) &&
+                 THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_CONSTRUCTIBLE(T) &&
+                 THRUST_OPTIONAL_IS_TRIVIALLY_MOVE_ASSIGNABLE(T)>
+struct optional_move_assign_base : optional_copy_assign_base<T> {
+  using optional_copy_assign_base<T>::optional_copy_assign_base;
+};
+
+template <class T>
+struct optional_move_assign_base<T, false> : optional_copy_assign_base<T> {
+  using optional_copy_assign_base<T>::optional_copy_assign_base;
+
+  __thrust_exec_check_disable__
+  optional_move_assign_base() = default;
+  __thrust_exec_check_disable__
+  optional_move_assign_base(const optional_move_assign_base &rhs) = default;
+
+  __thrust_exec_check_disable__
+  optional_move_assign_base(optional_move_assign_base &&rhs) = default;
+
+  __thrust_exec_check_disable__
+  optional_move_assign_base &
+  operator=(const optional_move_assign_base &rhs) = default;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional_move_assign_base &
+  operator=(optional_move_assign_base &&rhs) noexcept(
+      std::is_nothrow_move_constructible<T>::value
+          &&std::is_nothrow_move_assignable<T>::value) {
+    this->assign(std::move(rhs));
+    return *this;
+  }
+};
+
+// optional_delete_ctor_base will conditionally delete copy and move
+// constructors depending on whether T is copy/move constructible
+template <class T, bool EnableCopy = std::is_copy_constructible<T>::value,
+          bool EnableMove = std::is_move_constructible<T>::value>
+struct optional_delete_ctor_base {
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(const optional_delete_ctor_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(optional_delete_ctor_base &&) noexcept = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(const optional_delete_ctor_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(optional_delete_ctor_base &&) noexcept = default;
+};
+
+template <class T> struct optional_delete_ctor_base<T, true, false> {
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(const optional_delete_ctor_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(optional_delete_ctor_base &&) noexcept = delete;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(const optional_delete_ctor_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(optional_delete_ctor_base &&) noexcept = default;
+};
+
+template <class T> struct optional_delete_ctor_base<T, false, true> {
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(const optional_delete_ctor_base &) = delete;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(optional_delete_ctor_base &&) noexcept = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(const optional_delete_ctor_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(optional_delete_ctor_base &&) noexcept = default;
+};
+
+template <class T> struct optional_delete_ctor_base<T, false, false> {
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(const optional_delete_ctor_base &) = delete;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base(optional_delete_ctor_base &&) noexcept = delete;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(const optional_delete_ctor_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_ctor_base &
+  operator=(optional_delete_ctor_base &&) noexcept = default;
+};
+
+// optional_delete_assign_base will conditionally delete copy and move
+// constructors depending on whether T is copy/move constructible + assignable
+template <class T,
+          bool EnableCopy = (std::is_copy_constructible<T>::value &&
+                             std::is_copy_assignable<T>::value),
+          bool EnableMove = (std::is_move_constructible<T>::value &&
+                             std::is_move_assignable<T>::value)>
+struct optional_delete_assign_base {
+  __thrust_exec_check_disable__
+  optional_delete_assign_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(const optional_delete_assign_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(optional_delete_assign_base &&) noexcept =
+      default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(const optional_delete_assign_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(optional_delete_assign_base &&) noexcept = default;
+};
+
+template <class T> struct optional_delete_assign_base<T, true, false> {
+  __thrust_exec_check_disable__
+  optional_delete_assign_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(const optional_delete_assign_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(optional_delete_assign_base &&) noexcept =
+      default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(const optional_delete_assign_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(optional_delete_assign_base &&) noexcept = delete;
+};
+
+template <class T> struct optional_delete_assign_base<T, false, true> {
+  __thrust_exec_check_disable__
+  optional_delete_assign_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(const optional_delete_assign_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(optional_delete_assign_base &&) noexcept =
+      default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(const optional_delete_assign_base &) = delete;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(optional_delete_assign_base &&) noexcept = default;
+};
+
+template <class T> struct optional_delete_assign_base<T, false, false> {
+  __thrust_exec_check_disable__
+  optional_delete_assign_base() = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(const optional_delete_assign_base &) = default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base(optional_delete_assign_base &&) noexcept =
+      default;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(const optional_delete_assign_base &) = delete;
+  __thrust_exec_check_disable__
+  optional_delete_assign_base &
+  operator=(optional_delete_assign_base &&) noexcept = delete;
+};
+
+} // namespace detail
+
+/// \brief A tag type to represent an empty optional
+struct nullopt_t {
+  struct do_not_use {};
+  __host__ __device__
+  constexpr explicit nullopt_t(do_not_use, do_not_use) noexcept {}
+};
+/// \brief Represents an empty optional
+/// \synopsis static constexpr nullopt_t nullopt;
+///
+/// *Examples*:
+/// ```
+/// thrust::optional<int> a = thrust::nullopt;
+/// void foo (thrust::optional<int>);
+/// foo(thrust::nullopt); //pass an empty optional
+/// ```
+static constexpr nullopt_t nullopt{nullopt_t::do_not_use{},
+                                   nullopt_t::do_not_use{}};
+
+class bad_optional_access : public std::exception {
+public:
+  bad_optional_access() = default;
+  __host__
+  const char *what() const noexcept { return "Optional has no value"; }
+};
+
+/// An optional object is an object that contains the storage for another
+/// object and manages the lifetime of this contained object, if any. The
+/// contained object may be initialized after the optional object has been
+/// initialized, and may be destroyed before the optional object has been
+/// destroyed. The initialization state of the contained object is tracked by
+/// the optional object.
+template <class T>
+class optional : private detail::optional_move_assign_base<T>,
+                 private detail::optional_delete_ctor_base<T>,
+                 private detail::optional_delete_assign_base<T> {
+  using base = detail::optional_move_assign_base<T>;
+
+  static_assert(!std::is_same<T, in_place_t>::value,
+                "instantiation of optional with in_place_t is ill-formed");
+  static_assert(!std::is_same<detail::decay_t<T>, nullopt_t>::value,
+                "instantiation of optional with nullopt_t is ill-formed");
+
+public:
+// The different versions for C++14 and 11 are needed because deduced return
+// types are not SFINAE-safe. This provides better support for things like
+// generic lambdas. C.f.
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2017/p0826r0
+#if defined(THRUST_OPTIONAL_CPP14) && !defined(THRUST_OPTIONAL_GCC49) &&               \
+    !defined(THRUST_OPTIONAL_GCC54) && !defined(THRUST_OPTIONAL_GCC55)
+  /// \group and_then
+  /// Carries out some operation which returns an optional on the stored
+  /// object if there is one. \requires `std::invoke(std::forward<F>(f),
+  /// value())` returns a `std::optional<U>` for some `U`. \return Let `U` be
+  /// the result of `std::invoke(std::forward<F>(f), value())`. Returns a
+  /// `std::optional<U>`. The return value is empty if `*this` is empty,
+  /// otherwise the return value of `std::invoke(std::forward<F>(f), value())`
+  /// is returned.
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto and_then(F &&f) & {
+    using result = detail::invoke_result_t<F, T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto and_then(F &&f) && {
+    using result = detail::invoke_result_t<F, T &&>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto and_then(F &&f) const & {
+    using result = detail::invoke_result_t<F, const T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto and_then(F &&f) const && {
+    using result = detail::invoke_result_t<F, const T &&>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : result(nullopt);
+  }
+#endif
+#else
+  /// \group and_then
+  /// Carries out some operation which returns an optional on the stored
+  /// object if there is one. \requires `std::invoke(std::forward<F>(f),
+  /// value())` returns a `std::optional<U>` for some `U`.
+  /// \return Let `U` be the result of `std::invoke(std::forward<F>(f),
+  /// value())`. Returns a `std::optional<U>`. The return value is empty if
+  /// `*this` is empty, otherwise the return value of
+  /// `std::invoke(std::forward<F>(f), value())` is returned.
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR detail::invoke_result_t<F, T &> and_then(F &&f) & {
+    using result = detail::invoke_result_t<F, T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR detail::invoke_result_t<F, T &&> and_then(F &&f) && {
+    using result = detail::invoke_result_t<F, T &&>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr detail::invoke_result_t<F, const T &> and_then(F &&f) const & {
+    using result = detail::invoke_result_t<F, const T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr detail::invoke_result_t<F, const T &&> and_then(F &&f) const && {
+    using result = detail::invoke_result_t<F, const T &&>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : result(nullopt);
+  }
+#endif
+#endif
+
+#if defined(THRUST_OPTIONAL_CPP14) && !defined(THRUST_OPTIONAL_GCC49) &&               \
+    !defined(THRUST_OPTIONAL_GCC54) && !defined(THRUST_OPTIONAL_GCC55)
+  /// \brief Carries out some operation on the stored object if there is one.
+  /// \return Let `U` be the result of `std::invoke(std::forward<F>(f),
+  /// value())`. Returns a `std::optional<U>`. The return value is empty if
+  /// `*this` is empty, otherwise an `optional<U>` is constructed from the
+  /// return value of `std::invoke(std::forward<F>(f), value())` and is
+  /// returned.
+  ///
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto map(F &&f) & {
+    return optional_map_impl(*this, std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto map(F &&f) && {
+    return optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) const&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto map(F &&f) const & {
+    return optional_map_impl(*this, std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) const&&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto map(F &&f) const && {
+    return optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+#else
+  /// \brief Carries out some operation on the stored object if there is one.
+  /// \return Let `U` be the result of `std::invoke(std::forward<F>(f),
+  /// value())`. Returns a `std::optional<U>`. The return value is empty if
+  /// `*this` is empty, otherwise an `optional<U>` is constructed from the
+  /// return value of `std::invoke(std::forward<F>(f), value())` and is
+  /// returned.
+  ///
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR decltype(optional_map_impl(std::declval<optional &>(),
+                                             std::declval<F &&>()))
+  map(F &&f) & {
+    return optional_map_impl(*this, std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR decltype(optional_map_impl(std::declval<optional &&>(),
+                                             std::declval<F &&>()))
+  map(F &&f) && {
+    return optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) const&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr decltype(optional_map_impl(std::declval<const optional &>(),
+                              std::declval<F &&>()))
+  map(F &&f) const & {
+    return optional_map_impl(*this, std::forward<F>(f));
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) const&&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr decltype(optional_map_impl(std::declval<const optional &&>(),
+                              std::declval<F &&>()))
+  map(F &&f) const && {
+    return optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+#endif
+#endif
+
+  /// \brief Calls `f` if the optional is empty
+  /// \requires `std::invoke_result_t<F>` must be void or convertible to
+  /// `optional<T>`.
+  /// \effects If `*this` has a value, returns `*this`.
+  /// Otherwise, if `f` returns `void`, calls `std::forward<F>(f)` and returns
+  /// `std::nullopt`. Otherwise, returns `std::forward<F>(f)()`.
+  ///
+  /// \group or_else
+  /// \synopsis template <class F> optional<T> or_else (F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) & {
+    if (has_value())
+      return *this;
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) & {
+    return has_value() ? *this : std::forward<F>(f)();
+  }
+
+  /// \group or_else
+  /// \synopsis template <class F> optional<T> or_else (F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) && {
+    if (has_value())
+      return std::move(*this);
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) && {
+    return has_value() ? std::move(*this) : std::forward<F>(f)();
+  }
+
+  /// \group or_else
+  /// \synopsis template <class F> optional<T> or_else (F &&f) const &;
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) const & {
+    if (has_value())
+      return *this;
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) const & {
+    return has_value() ? *this : std::forward<F>(f)();
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) const && {
+    if (has_value())
+      return std::move(*this);
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) const && {
+    return has_value() ? std::move(*this) : std::forward<F>(f)();
+  }
+#endif
+
+  /// \brief Maps the stored value with `f` if there is one, otherwise returns
+  /// `u`.
+  ///
+  /// \details If there is a value stored, then `f` is called with `**this`
+  /// and the value is returned. Otherwise `u` is returned.
+  ///
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u);
+  }
+
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u);
+  }
+
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) const & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) const && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u);
+  }
+#endif
+
+  /// \brief Maps the stored value with `f` if there is one, otherwise calls
+  /// `u` and returns the result.
+  ///
+  /// \details If there is a value stored, then `f` is
+  /// called with `**this` and the value is returned. Otherwise
+  /// `std::forward<U>(u)()` is returned.
+  ///
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u) &;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u)();
+  }
+
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u)
+  /// &&;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u)();
+  }
+
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u)
+  /// const &;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) const & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u)();
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u)
+  /// const &&;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) const && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u)();
+  }
+#endif
+
+  /// \return `u` if `*this` has a value, otherwise an empty optional.
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  constexpr optional<typename std::decay<U>::type> conjunction(U &&u) const {
+    using result = optional<detail::decay_t<U>>;
+    return has_value() ? result{u} : result{nullopt};
+  }
+
+  /// \return `rhs` if `*this` is empty, otherwise the current value.
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(const optional &rhs) & {
+    return has_value() ? *this : rhs;
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(const optional &rhs) const & {
+    return has_value() ? *this : rhs;
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(const optional &rhs) && {
+    return has_value() ? std::move(*this) : rhs;
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(const optional &rhs) const && {
+    return has_value() ? std::move(*this) : rhs;
+  }
+#endif
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(optional &&rhs) & {
+    return has_value() ? *this : std::move(rhs);
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(optional &&rhs) const & {
+    return has_value() ? *this : std::move(rhs);
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(optional &&rhs) && {
+    return has_value() ? std::move(*this) : std::move(rhs);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(optional &&rhs) const && {
+    return has_value() ? std::move(*this) : std::move(rhs);
+  }
+#endif
+
+  /// Takes the value out of the optional, leaving it empty
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() & {
+    optional ret = *this;
+    reset();
+    return ret;
+  }
+
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() const & {
+    optional ret = *this;
+    reset();
+    return ret;
+  }
+
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() && {
+    optional ret = std::move(*this);
+    reset();
+    return ret;
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() const && {
+    optional ret = std::move(*this);
+    reset();
+    return ret;
+  }
+#endif
+
+  using value_type = T;
+
+  /// Constructs an optional that does not contain a value.
+  /// \group ctor_empty
+  __thrust_exec_check_disable__
+  constexpr optional() noexcept = default;
+
+  /// \group ctor_empty
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional(nullopt_t) noexcept {}
+
+  /// Copy constructor
+  ///
+  /// If `rhs` contains a value, the stored value is direct-initialized with
+  /// it. Otherwise, the constructed optional is empty.
+  __thrust_exec_check_disable__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional(const optional &rhs) = default;
+
+  /// Move constructor
+  ///
+  /// If `rhs` contains a value, the stored value is direct-initialized with
+  /// it. Otherwise, the constructed optional is empty.
+  __thrust_exec_check_disable__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional(optional &&rhs) = default;
+
+  /// Constructs the stored value in-place using the given arguments.
+  /// \group in_place
+  /// \synopsis template <class... Args> constexpr explicit optional(in_place_t, Args&&... args);
+  __thrust_exec_check_disable__
+  template <class... Args>
+  __host__ __device__
+  constexpr explicit optional(
+      detail::enable_if_t<std::is_constructible<T, Args...>::value, in_place_t>,
+      Args &&... args)
+      : base(in_place, std::forward<Args>(args)...) {}
+
+  /// \group in_place
+  /// \synopsis template <class U, class... Args>\nconstexpr explicit optional(in_place_t, std::initializer_list<U>&, Args&&... args);
+  __thrust_exec_check_disable__
+  template <class U, class... Args>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR explicit optional(
+      detail::enable_if_t<std::is_constructible<T, std::initializer_list<U> &,
+                                                Args &&...>::value,
+                          in_place_t>,
+      std::initializer_list<U> il, Args &&... args) {
+    this->construct(il, std::forward<Args>(args)...);
+  }
+
+  /// Constructs the stored value with `u`.
+  /// \synopsis template <class U=T> constexpr optional(U &&u);
+  __thrust_exec_check_disable__
+  template <
+      class U = T,
+      detail::enable_if_t<std::is_convertible<U &&, T>::value> * = nullptr,
+      detail::enable_forward_value<T, U> * = nullptr>
+  __host__ __device__
+  constexpr optional(U &&u) : base(in_place, std::forward<U>(u)) {}
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <
+      class U = T,
+      detail::enable_if_t<!std::is_convertible<U &&, T>::value> * = nullptr,
+      detail::enable_forward_value<T, U> * = nullptr>
+  __host__ __device__
+  constexpr explicit optional(U &&u) : base(in_place, std::forward<U>(u)) {}
+
+  /// Converting copy constructor.
+  /// \synopsis template <class U> optional(const optional<U> &rhs);
+  __thrust_exec_check_disable__
+  template <
+      class U, detail::enable_from_other<T, U, const U &> * = nullptr,
+      detail::enable_if_t<std::is_convertible<const U &, T>::value> * = nullptr>
+  __host__ __device__
+  optional(const optional<U> &rhs) {
+    this->construct(*rhs);
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class U, detail::enable_from_other<T, U, const U &> * = nullptr,
+            detail::enable_if_t<!std::is_convertible<const U &, T>::value> * =
+                nullptr>
+  __host__ __device__
+  explicit optional(const optional<U> &rhs) {
+    this->construct(*rhs);
+  }
+
+  /// Converting move constructor.
+  /// \synopsis template <class U> optional(optional<U> &&rhs);
+  __thrust_exec_check_disable__
+  template <
+      class U, detail::enable_from_other<T, U, U &&> * = nullptr,
+      detail::enable_if_t<std::is_convertible<U &&, T>::value> * = nullptr>
+  __host__ __device__
+  optional(optional<U> &&rhs) {
+    this->construct(std::move(*rhs));
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <
+      class U, detail::enable_from_other<T, U, U &&> * = nullptr,
+      detail::enable_if_t<!std::is_convertible<U &&, T>::value> * = nullptr>
+  __host__ __device__
+  explicit optional(optional<U> &&rhs) {
+    this->construct(std::move(*rhs));
+  }
+
+  /// Destroys the stored value if there is one.
+  __thrust_exec_check_disable__
+  ~optional() = default;
+
+  /// Assignment to empty.
+  ///
+  /// Destroys the current value if there is one.
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional &operator=(nullopt_t) noexcept {
+    if (has_value()) {
+      this->m_value.~T();
+      this->m_has_value = false;
+    }
+
+    return *this;
+  }
+
+  /// Copy assignment.
+  ///
+  /// Copies the value from `rhs` if there is one. Otherwise resets the stored
+  /// value in `*this`.
+  __thrust_exec_check_disable__
+  optional &operator=(const optional &rhs) = default;
+
+  /// Move assignment.
+  ///
+  /// Moves the value from `rhs` if there is one. Otherwise resets the stored
+  /// value in `*this`.
+  __thrust_exec_check_disable__
+  optional &operator=(optional &&rhs) = default;
+
+  /// Assigns the stored value from `u`, destroying the old value if there was
+  /// one.
+  /// \synopsis optional &operator=(U &&u);
+  __thrust_exec_check_disable__
+  template <class U = T, detail::enable_assign_forward<T, U> * = nullptr>
+  __host__ __device__
+  optional &operator=(U &&u) {
+    if (has_value()) {
+      this->m_value = std::forward<U>(u);
+    } else {
+      this->construct(std::forward<U>(u));
+    }
+
+    return *this;
+  }
+
+  /// Converting copy assignment operator.
+  ///
+  /// Copies the value from `rhs` if there is one. Otherwise resets the stored
+  /// value in `*this`.
+  /// \synopsis optional &operator=(const optional<U> & rhs);
+  __thrust_exec_check_disable__
+  template <class U,
+            detail::enable_assign_from_other<T, U, const U &> * = nullptr>
+  __host__ __device__
+  optional &operator=(const optional<U> &rhs) {
+    if (has_value()) {
+      if (rhs.has_value()) {
+        this->m_value = *rhs;
+      } else {
+        this->hard_reset();
+      }
+    }
+
+    if (rhs.has_value()) {
+      this->construct(*rhs);
+    }
+
+    return *this;
+  }
+
+  // TODO check exception guarantee
+  /// Converting move assignment operator.
+  ///
+  /// Moves the value from `rhs` if there is one. Otherwise resets the stored
+  /// value in `*this`.
+  /// \synopsis optional &operator=(optional<U> && rhs);
+  __thrust_exec_check_disable__
+  template <class U, detail::enable_assign_from_other<T, U, U> * = nullptr>
+  __host__ __device__
+  optional &operator=(optional<U> &&rhs) {
+    if (has_value()) {
+      if (rhs.has_value()) {
+        this->m_value = std::move(*rhs);
+      } else {
+        this->hard_reset();
+      }
+    }
+
+    if (rhs.has_value()) {
+      this->construct(std::move(*rhs));
+    }
+
+    return *this;
+  }
+
+  /// Constructs the value in-place, destroying the current one if there is
+  /// one.
+  /// \group emplace
+  __thrust_exec_check_disable__
+  template <class... Args>
+  __host__ __device__
+  T &emplace(Args &&... args) {
+    static_assert(std::is_constructible<T, Args &&...>::value,
+                  "T must be constructible with Args");
+
+    *this = nullopt;
+    this->construct(std::forward<Args>(args)...);
+    return this->m_value;
+  }
+
+  /// \group emplace
+  /// \synopsis template <class U, class... Args>\nT& emplace(std::initializer_list<U> il, Args &&... args);
+  __thrust_exec_check_disable__
+  template <class U, class... Args>
+  __host__ __device__
+  detail::enable_if_t<
+      std::is_constructible<T, std::initializer_list<U> &, Args &&...>::value,
+      T &>
+  emplace(std::initializer_list<U> il, Args &&... args) {
+    *this = nullopt;
+    this->construct(il, std::forward<Args>(args)...);
+    return this->m_value;
+  }
+
+  /// Swaps this optional with the other.
+  ///
+  /// If neither optionals have a value, nothing happens.
+  /// If both have a value, the values are swapped.
+  /// If one has a value, it is moved to the other and the movee is left
+  /// valueless.
+  __thrust_exec_check_disable__
+  __host__ __device__
+  void
+  swap(optional &rhs) noexcept(std::is_nothrow_move_constructible<T>::value
+                                   &&detail::is_nothrow_swappable<T>::value) {
+    if (has_value()) {
+      if (rhs.has_value()) {
+        using thrust::swap;
+        swap(**this, *rhs);
+      } else {
+        new (addressof(rhs.m_value)) T(std::move(this->m_value));
+        this->m_value.T::~T();
+      }
+    } else if (rhs.has_value()) {
+      new (addressof(this->m_value)) T(std::move(rhs.m_value));
+      rhs.m_value.T::~T();
+    }
+  }
+
+  /// \return a pointer to the stored value
+  /// \requires a value is stored
+  /// \group pointer
+  /// \synopsis constexpr const T *operator->() const;
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T *operator->() const {
+    return addressof(this->m_value);
+  }
+
+  /// \group pointer
+  /// \synopsis constexpr T *operator->();
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T *operator->() {
+    return addressof(this->m_value);
+  }
+
+  /// \return the stored value
+  /// \requires a value is stored
+  /// \group deref
+  /// \synopsis constexpr T &operator*();
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &operator*() & { return this->m_value; }
+
+  /// \group deref
+  /// \synopsis constexpr const T &operator*() const;
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T &operator*() const & { return this->m_value; }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &&operator*() && {
+    return std::move(this->m_value);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \exclude
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T &&operator*() const && { return std::move(this->m_value); }
+#endif
+
+  /// \return whether or not the optional has a value
+  /// \group has_value
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool has_value() const noexcept { return this->m_has_value; }
+
+  /// \group has_value
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr explicit operator bool() const noexcept {
+    return this->m_has_value;
+  }
+
+  /// \return the contained value if there is one, otherwise throws
+  /// [bad_optional_access]
+  /// \group value
+  /// \synopsis constexpr T &value();
+  __host__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &value() & {
+    if (has_value())
+      return this->m_value;
+    throw bad_optional_access();
+  }
+  /// \group value
+  /// \synopsis constexpr const T &value() const;
+  __host__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR const T &value() const & {
+    if (has_value())
+      return this->m_value;
+    throw bad_optional_access();
+  }
+  /// \exclude
+  __host__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &&value() && {
+    if (has_value())
+      return std::move(this->m_value);
+    throw bad_optional_access();
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \exclude
+  __host__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR const T &&value() const && {
+    if (has_value())
+      return std::move(this->m_value);
+    throw bad_optional_access();
+  }
+#endif
+
+  /// \return the stored value if there is one, otherwise returns `u`
+  /// \group value_or
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  constexpr T value_or(U &&u) const & {
+    static_assert(std::is_copy_constructible<T>::value &&
+                      std::is_convertible<U &&, T>::value,
+                  "T must be copy constructible and convertible from U");
+    return has_value() ? **this : static_cast<T>(std::forward<U>(u));
+  }
+
+  /// \group value_or
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T value_or(U &&u) && {
+    static_assert(std::is_move_constructible<T>::value &&
+                      std::is_convertible<U &&, T>::value,
+                  "T must be move constructible and convertible from U");
+    return has_value() ? **this : static_cast<T>(std::forward<U>(u));
+  }
+
+  /// Destroys the stored value if one exists, making the optional empty
+  __thrust_exec_check_disable__
+  __host__ __device__
+  void reset() noexcept {
+    if (has_value()) {
+      this->m_value.~T();
+      this->m_has_value = false;
+    }
+  }
+};
+
+/// \group relop
+/// \brief Compares two optional objects
+/// \details If both optionals contain a value, they are compared with `T`s
+/// relational operators. Otherwise `lhs` and `rhs` are equal only if they are
+/// both empty, and `lhs` is less than `rhs` only if `rhs` is empty and `lhs`
+/// is not.
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator==(const optional<T> &lhs,
+                                 const optional<U> &rhs) {
+  return lhs.has_value() == rhs.has_value() &&
+         (!lhs.has_value() || *lhs == *rhs);
+}
+/// \group relop
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator!=(const optional<T> &lhs,
+                                 const optional<U> &rhs) {
+  return lhs.has_value() != rhs.has_value() ||
+         (lhs.has_value() && *lhs != *rhs);
+}
+/// \group relop
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator<(const optional<T> &lhs,
+                                const optional<U> &rhs) {
+  return rhs.has_value() && (!lhs.has_value() || *lhs < *rhs);
+}
+/// \group relop
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator>(const optional<T> &lhs,
+                                const optional<U> &rhs) {
+  return lhs.has_value() && (!rhs.has_value() || *lhs > *rhs);
+}
+/// \group relop
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator<=(const optional<T> &lhs,
+                                 const optional<U> &rhs) {
+  return !lhs.has_value() || (rhs.has_value() && *lhs <= *rhs);
+}
+/// \group relop
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator>=(const optional<T> &lhs,
+                                 const optional<U> &rhs) {
+  return !rhs.has_value() || (lhs.has_value() && *lhs >= *rhs);
+}
+
+/// \group relop_nullopt
+/// \brief Compares an optional to a `nullopt`
+/// \details Equivalent to comparing the optional to an empty optional
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator==(const optional<T> &lhs, nullopt_t) noexcept {
+  return !lhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator==(nullopt_t, const optional<T> &rhs) noexcept {
+  return !rhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator!=(const optional<T> &lhs, nullopt_t) noexcept {
+  return lhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator!=(nullopt_t, const optional<T> &rhs) noexcept {
+  return rhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator<(const optional<T> &, nullopt_t) noexcept {
+  return false;
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator<(nullopt_t, const optional<T> &rhs) noexcept {
+  return rhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator<=(const optional<T> &lhs, nullopt_t) noexcept {
+  return !lhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator<=(nullopt_t, const optional<T> &) noexcept {
+  return true;
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator>(const optional<T> &lhs, nullopt_t) noexcept {
+  return lhs.has_value();
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator>(nullopt_t, const optional<T> &) noexcept {
+  return false;
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator>=(const optional<T> &, nullopt_t) noexcept {
+  return true;
+}
+/// \group relop_nullopt
+__thrust_exec_check_disable__
+template <class T>
+__host__ __device__
+inline constexpr bool operator>=(nullopt_t, const optional<T> &rhs) noexcept {
+  return !rhs.has_value();
+}
+
+/// \group relop_t
+/// \brief Compares the optional with a value.
+/// \details If the optional has a value, it is compared with the other value
+/// using `T`s relational operators. Otherwise, the optional is considered
+/// less than the value.
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator==(const optional<T> &lhs, const U &rhs) {
+  return lhs.has_value() ? *lhs == rhs : false;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator==(const U &lhs, const optional<T> &rhs) {
+  return rhs.has_value() ? lhs == *rhs : false;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator!=(const optional<T> &lhs, const U &rhs) {
+  return lhs.has_value() ? *lhs != rhs : true;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator!=(const U &lhs, const optional<T> &rhs) {
+  return rhs.has_value() ? lhs != *rhs : true;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator<(const optional<T> &lhs, const U &rhs) {
+  return lhs.has_value() ? *lhs < rhs : true;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator<(const U &lhs, const optional<T> &rhs) {
+  return rhs.has_value() ? lhs < *rhs : false;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator<=(const optional<T> &lhs, const U &rhs) {
+  return lhs.has_value() ? *lhs <= rhs : true;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator<=(const U &lhs, const optional<T> &rhs) {
+  return rhs.has_value() ? lhs <= *rhs : false;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator>(const optional<T> &lhs, const U &rhs) {
+  return lhs.has_value() ? *lhs > rhs : false;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator>(const U &lhs, const optional<T> &rhs) {
+  return rhs.has_value() ? lhs > *rhs : true;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator>=(const optional<T> &lhs, const U &rhs) {
+  return lhs.has_value() ? *lhs >= rhs : false;
+}
+/// \group relop_t
+__thrust_exec_check_disable__
+template <class T, class U>
+__host__ __device__
+inline constexpr bool operator>=(const U &lhs, const optional<T> &rhs) {
+  return rhs.has_value() ? lhs >= *rhs : true;
+}
+
+/// \synopsis template <class T>\nvoid swap(optional<T> &lhs, optional<T> &rhs);
+__thrust_exec_check_disable__
+template <class T,
+          detail::enable_if_t<std::is_move_constructible<T>::value> * = nullptr,
+          detail::enable_if_t<detail::is_swappable<T>::value> * = nullptr>
+__host__ __device__
+void swap(optional<T> &lhs,
+          optional<T> &rhs) noexcept(noexcept(lhs.swap(rhs))) {
+  return lhs.swap(rhs);
+}
+
+namespace detail {
+struct i_am_secret {};
+} // namespace detail
+
+__thrust_exec_check_disable__
+template <class T = detail::i_am_secret, class U,
+          class Ret =
+              detail::conditional_t<std::is_same<T, detail::i_am_secret>::value,
+                                    detail::decay_t<U>, T>>
+__host__ __device__
+inline constexpr optional<Ret> make_optional(U &&v) {
+  return optional<Ret>(std::forward<U>(v));
+}
+
+__thrust_exec_check_disable__
+template <class T, class... Args>
+__host__ __device__
+inline constexpr optional<T> make_optional(Args &&... args) {
+  return optional<T>(in_place, std::forward<Args>(args)...);
+}
+__thrust_exec_check_disable__
+template <class T, class U, class... Args>
+__host__ __device__
+inline constexpr optional<T> make_optional(std::initializer_list<U> il,
+                                           Args &&... args) {
+  return optional<T>(in_place, il, std::forward<Args>(args)...);
+}
+
+#if THRUST_CPP_DIALECT >= 2017
+template <class T> optional(T)->optional<T>;
+#endif
+
+// Doxygen chokes on the trailing return types used below.
+#if !defined(THRUST_DOXYGEN)
+/// \exclude
+namespace detail {
+#ifdef THRUST_OPTIONAL_CPP14
+__thrust_exec_check_disable__
+template <class Opt, class F,
+          class Ret = decltype(detail::invoke(std::declval<F>(),
+                                              *std::declval<Opt>())),
+          detail::enable_if_t<!std::is_void<Ret>::value> * = nullptr>
+__host__ __device__
+constexpr auto optional_map_impl(Opt &&opt, F &&f) {
+  return opt.has_value()
+             ? detail::invoke(std::forward<F>(f), *std::forward<Opt>(opt))
+             : optional<Ret>(nullopt);
+}
+
+__thrust_exec_check_disable__
+template <class Opt, class F,
+          class Ret = decltype(detail::invoke(std::declval<F>(),
+                                              *std::declval<Opt>())),
+          detail::enable_if_t<std::is_void<Ret>::value> * = nullptr>
+__host__ __device__
+auto optional_map_impl(Opt &&opt, F &&f) {
+  if (opt.has_value()) {
+    detail::invoke(std::forward<F>(f), *std::forward<Opt>(opt));
+    return make_optional(monostate{});
+  }
+
+  return optional<monostate>(nullopt);
+}
+#else
+__thrust_exec_check_disable__
+template <class Opt, class F,
+          class Ret = decltype(detail::invoke(std::declval<F>(),
+                                              *std::declval<Opt>())),
+          detail::enable_if_t<!std::is_void<Ret>::value> * = nullptr>
+__host__ __device__
+constexpr optional<Ret> optional_map_impl(Opt &&opt, F &&f) {
+  return opt.has_value()
+             ? detail::invoke(std::forward<F>(f), *std::forward<Opt>(opt))
+             : optional<Ret>(nullopt);
+}
+
+__thrust_exec_check_disable__
+template <class Opt, class F,
+          class Ret = decltype(detail::invoke(std::declval<F>(),
+                                              *std::declval<Opt>())),
+          detail::enable_if_t<std::is_void<Ret>::value> * = nullptr>
+__host__ __device__
+auto optional_map_impl(Opt &&opt, F &&f) -> optional<monostate>
+{
+  if (opt.has_value()) {
+    detail::invoke(std::forward<F>(f), *std::forward<Opt>(opt));
+    return monostate{};
+  }
+
+  return nullopt;
+}
+#endif
+} // namespace detail
+#endif // !defined(THRUST_DOXYGEN)
+
+/// Specialization for when `T` is a reference. `optional<T&>` acts similarly
+/// to a `T*`, but provides more operations and shows intent more clearly.
+///
+/// *Examples*:
+///
+/// ```
+/// int i = 42;
+/// thrust::optional<int&> o = i;
+/// *o == 42; //true
+/// i = 12;
+/// *o = 12; //true
+/// &*o == &i; //true
+/// ```
+///
+/// Assignment has rebind semantics rather than assign-through semantics:
+///
+/// ```
+/// int j = 8;
+/// o = j;
+///
+/// &*o == &j; //true
+/// ```
+template <class T> class optional<T &> {
+public:
+// The different versions for C++14 and 11 are needed because deduced return
+// types are not SFINAE-safe. This provides better support for things like
+// generic lambdas. C.f.
+// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2017/p0826r0
+#if defined(THRUST_OPTIONAL_CPP14) && !defined(THRUST_OPTIONAL_GCC49) &&               \
+    !defined(THRUST_OPTIONAL_GCC54) && !defined(THRUST_OPTIONAL_GCC55)
+  /// \group and_then
+  /// Carries out some operation which returns an optional on the stored
+  /// object if there is one. \requires `std::invoke(std::forward<F>(f),
+  /// value())` returns a `std::optional<U>` for some `U`. \return Let `U` be
+  /// the result of `std::invoke(std::forward<F>(f), value())`. Returns a
+  /// `std::optional<U>`. The return value is empty if `*this` is empty,
+  /// otherwise the return value of `std::invoke(std::forward<F>(f), value())`
+  /// is returned.
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto and_then(F &&f) & {
+    using result = detail::invoke_result_t<F, T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto and_then(F &&f) && {
+    using result = detail::invoke_result_t<F, T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto and_then(F &&f) const & {
+    using result = detail::invoke_result_t<F, const T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto and_then(F &&f) const && {
+    using result = detail::invoke_result_t<F, const T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+#endif
+#else
+  /// \group and_then
+  /// Carries out some operation which returns an optional on the stored
+  /// object if there is one. \requires `std::invoke(std::forward<F>(f),
+  /// value())` returns a `std::optional<U>` for some `U`. \return Let `U` be
+  /// the result of `std::invoke(std::forward<F>(f), value())`. Returns a
+  /// `std::optional<U>`. The return value is empty if `*this` is empty,
+  /// otherwise the return value of `std::invoke(std::forward<F>(f), value())`
+  /// is returned.
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR detail::invoke_result_t<F, T &> and_then(F &&f) & {
+    using result = detail::invoke_result_t<F, T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR detail::invoke_result_t<F, T &> and_then(F &&f) && {
+    using result = detail::invoke_result_t<F, T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr detail::invoke_result_t<F, const T &> and_then(F &&f) const & {
+    using result = detail::invoke_result_t<F, const T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group and_then
+  /// \synopsis template <class F>\nconstexpr auto and_then(F &&f) const &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr detail::invoke_result_t<F, const T &> and_then(F &&f) const && {
+    using result = detail::invoke_result_t<F, const T &>;
+    static_assert(detail::is_optional<result>::value,
+                  "F must return an optional");
+
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : result(nullopt);
+  }
+#endif
+#endif
+
+#if defined(THRUST_OPTIONAL_CPP14) && !defined(THRUST_OPTIONAL_GCC49) &&               \
+    !defined(THRUST_OPTIONAL_GCC54) && !defined(THRUST_OPTIONAL_GCC55)
+  /// \brief Carries out some operation on the stored object if there is one.
+  /// \return Let `U` be the result of `std::invoke(std::forward<F>(f),
+  /// value())`. Returns a `std::optional<U>`. The return value is empty if
+  /// `*this` is empty, otherwise an `optional<U>` is constructed from the
+  /// return value of `std::invoke(std::forward<F>(f), value())` and is
+  /// returned.
+  ///
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto map(F &&f) & {
+    return detail::optional_map_impl(*this, std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR auto map(F &&f) && {
+    return detail::optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) const&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto map(F &&f) const & {
+    return detail::optional_map_impl(*this, std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> constexpr auto map(F &&f) const&&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr auto map(F &&f) const && {
+    return detail::optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+#else
+  /// \brief Carries out some operation on the stored object if there is one.
+  /// \return Let `U` be the result of `std::invoke(std::forward<F>(f),
+  /// value())`. Returns a `std::optional<U>`. The return value is empty if
+  /// `*this` is empty, otherwise an `optional<U>` is constructed from the
+  /// return value of `std::invoke(std::forward<F>(f), value())` and is
+  /// returned.
+  ///
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR decltype(detail::optional_map_impl(std::declval<optional &>(),
+                                                     std::declval<F &&>()))
+  map(F &&f) & {
+    return detail::optional_map_impl(*this, std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR decltype(detail::optional_map_impl(std::declval<optional &&>(),
+                                                     std::declval<F &&>()))
+  map(F &&f) && {
+    return detail::optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) const&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr decltype(detail::optional_map_impl(std::declval<const optional &>(),
+                                      std::declval<F &&>()))
+  map(F &&f) const & {
+    return detail::optional_map_impl(*this, std::forward<F>(f));
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group map
+  /// \synopsis template <class F> auto map(F &&f) const&&;
+  __thrust_exec_check_disable__
+  template <class F>
+  __host__ __device__
+  constexpr decltype(detail::optional_map_impl(std::declval<const optional &&>(),
+                                      std::declval<F &&>()))
+  map(F &&f) const && {
+    return detail::optional_map_impl(std::move(*this), std::forward<F>(f));
+  }
+#endif
+#endif
+
+  /// \brief Calls `f` if the optional is empty
+  /// \requires `std::invoke_result_t<F>` must be void or convertible to
+  /// `optional<T>`. \effects If `*this` has a value, returns `*this`.
+  /// Otherwise, if `f` returns `void`, calls `std::forward<F>(f)` and returns
+  /// `std::nullopt`. Otherwise, returns `std::forward<F>(f)()`.
+  ///
+  /// \group or_else
+  /// \synopsis template <class F> optional<T> or_else (F &&f) &;
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T>
+  THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) & {
+    if (has_value())
+      return *this;
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T>
+  THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) & {
+    return has_value() ? *this : std::forward<F>(f)();
+  }
+
+  /// \group or_else
+  /// \synopsis template <class F> optional<T> or_else (F &&f) &&;
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) && {
+    if (has_value())
+      return std::move(*this);
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) && {
+    return has_value() ? std::move(*this) : std::forward<F>(f)();
+  }
+
+  /// \group or_else
+  /// \synopsis template <class F> optional<T> or_else (F &&f) const &;
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) const & {
+    if (has_value())
+      return *this;
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> THRUST_OPTIONAL_CPP11_CONSTEXPR or_else(F &&f) const & {
+    return has_value() ? *this : std::forward<F>(f)();
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::enable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) const && {
+    if (has_value())
+      return std::move(*this);
+
+    std::forward<F>(f)();
+    return nullopt;
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class F, detail::disable_if_ret_void<F> * = nullptr>
+  __host__ __device__
+  optional<T> or_else(F &&f) const && {
+    return has_value() ? std::move(*this) : std::forward<F>(f)();
+  }
+#endif
+
+  /// \brief Maps the stored value with `f` if there is one, otherwise returns
+  /// `u`.
+  ///
+  /// \details If there is a value stored, then `f` is called with `**this`
+  /// and the value is returned. Otherwise `u` is returned.
+  ///
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u);
+  }
+
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u);
+  }
+
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) const & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group map_or
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  U map_or(F &&f, U &&u) const && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u);
+  }
+#endif
+
+  /// \brief Maps the stored value with `f` if there is one, otherwise calls
+  /// `u` and returns the result.
+  ///
+  /// \details If there is a value stored, then `f` is
+  /// called with `**this` and the value is returned. Otherwise
+  /// `std::forward<U>(u)()` is returned.
+  ///
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u) &;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u)();
+  }
+
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u)
+  /// &&;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u)();
+  }
+
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u)
+  /// const &;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) const & {
+    return has_value() ? detail::invoke(std::forward<F>(f), **this)
+                       : std::forward<U>(u)();
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group map_or_else
+  /// \synopsis template <class F, class U>\nauto map_or_else(F &&f, U &&u)
+  /// const &&;
+  __thrust_exec_check_disable__
+  template <class F, class U>
+  __host__ __device__
+  detail::invoke_result_t<U> map_or_else(F &&f, U &&u) const && {
+    return has_value() ? detail::invoke(std::forward<F>(f), std::move(**this))
+                       : std::forward<U>(u)();
+  }
+#endif
+
+  /// \return `u` if `*this` has a value, otherwise an empty optional.
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  constexpr optional<typename std::decay<U>::type> conjunction(U &&u) const {
+    using result = optional<detail::decay_t<U>>;
+    return has_value() ? result{u} : result{nullopt};
+  }
+
+  /// \return `rhs` if `*this` is empty, otherwise the current value.
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(const optional &rhs) & {
+    return has_value() ? *this : rhs;
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(const optional &rhs) const & {
+    return has_value() ? *this : rhs;
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(const optional &rhs) && {
+    return has_value() ? std::move(*this) : rhs;
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(const optional &rhs) const && {
+    return has_value() ? std::move(*this) : rhs;
+  }
+#endif
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(optional &&rhs) & {
+    return has_value() ? *this : std::move(rhs);
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(optional &&rhs) const & {
+    return has_value() ? *this : std::move(rhs);
+  }
+
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional disjunction(optional &&rhs) && {
+    return has_value() ? std::move(*this) : std::move(rhs);
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group disjunction
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional disjunction(optional &&rhs) const && {
+    return has_value() ? std::move(*this) : std::move(rhs);
+  }
+#endif
+
+  /// Takes the value out of the optional, leaving it empty
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() & {
+    optional ret = *this;
+    reset();
+    return ret;
+  }
+
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() const & {
+    optional ret = *this;
+    reset();
+    return ret;
+  }
+
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() && {
+    optional ret = std::move(*this);
+    reset();
+    return ret;
+  }
+
+#ifndef THRUST_OPTIONAL_NO_CONSTRR
+  /// \group take
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional take() const && {
+    optional ret = std::move(*this);
+    reset();
+    return ret;
+  }
+#endif
+
+  using value_type = T &;
+
+  /// Constructs an optional that does not contain a value.
+  /// \group ctor_empty
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional() noexcept : m_value(nullptr) {}
+
+  /// \group ctor_empty
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr optional(nullopt_t) noexcept : m_value(nullptr) {}
+
+  /// Copy constructor
+  ///
+  /// If `rhs` contains a value, the stored value is direct-initialized with
+  /// it. Otherwise, the constructed optional is empty.
+  __thrust_exec_check_disable__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional(const optional &rhs) noexcept = default;
+
+  /// Move constructor
+  ///
+  /// If `rhs` contains a value, the stored value is direct-initialized with
+  /// it. Otherwise, the constructed optional is empty.
+  __thrust_exec_check_disable__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR optional(optional &&rhs) = default;
+
+  /// Constructs the stored value with `u`.
+  /// \synopsis template <class U=T> constexpr optional(U &&u);
+  __thrust_exec_check_disable__
+  template <class U = T,
+            detail::enable_if_t<!detail::is_optional<detail::decay_t<U>>::value>
+                * = nullptr>
+  __host__ __device__
+  constexpr optional(U &&u) : m_value(addressof(u)) {
+    static_assert(std::is_lvalue_reference<U>::value, "U must be an lvalue");
+  }
+
+  /// \exclude
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  constexpr explicit optional(const optional<U> &rhs) : optional(*rhs) {}
+
+  /// No-op
+  __thrust_exec_check_disable__
+  ~optional() = default;
+
+  /// Assignment to empty.
+  ///
+  /// Destroys the current value if there is one.
+  __thrust_exec_check_disable__
+  __host__ __device__
+  optional &operator=(nullopt_t) noexcept {
+    m_value = nullptr;
+    return *this;
+  }
+
+  /// Copy assignment.
+  ///
+  /// Rebinds this optional to the referee of `rhs` if there is one. Otherwise
+  /// resets the stored value in `*this`.
+  __thrust_exec_check_disable__
+  optional &operator=(const optional &rhs) = default;
+
+  /// Rebinds this optional to `u`.
+  ///
+  /// \requires `U` must be an lvalue reference.
+  /// \synopsis optional &operator=(U &&u);
+  __thrust_exec_check_disable__
+  template <class U = T,
+            detail::enable_if_t<!detail::is_optional<detail::decay_t<U>>::value>
+                * = nullptr>
+  __host__ __device__
+  optional &operator=(U &&u) {
+    static_assert(std::is_lvalue_reference<U>::value, "U must be an lvalue");
+    m_value = addressof(u);
+    return *this;
+  }
+
+  /// Converting copy assignment operator.
+  ///
+  /// Rebinds this optional to the referee of `rhs` if there is one. Otherwise
+  /// resets the stored value in `*this`.
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  optional &operator=(const optional<U> &rhs) {
+    m_value = addressof(rhs.value());
+    return *this;
+  }
+
+  /// Constructs the value in-place, destroying the current one if there is
+  /// one.
+  ///
+  /// \group emplace
+  __thrust_exec_check_disable__
+  template <class... Args>
+  __host__ __device__
+  T &emplace(Args &&... args) noexcept {
+    static_assert(std::is_constructible<T, Args &&...>::value,
+                  "T must be constructible with Args");
+
+    *this = nullopt;
+    this->construct(std::forward<Args>(args)...);
+  }
+
+  /// Swaps this optional with the other.
+  ///
+  /// If neither optionals have a value, nothing happens.
+  /// If both have a value, the values are swapped.
+  /// If one has a value, it is moved to the other and the movee is left
+  /// valueless.
+  __thrust_exec_check_disable__
+  __host__ __device__
+  void swap(optional &rhs) noexcept { std::swap(m_value, rhs.m_value); }
+
+  /// \return a pointer to the stored value
+  /// \requires a value is stored
+  /// \group pointer
+  /// \synopsis constexpr const T *operator->() const;
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T *operator->() const { return m_value; }
+
+  /// \group pointer
+  /// \synopsis constexpr T *operator->();
+  __thrust_exec_check_disable__
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T *operator->() { return m_value; }
+
+  /// \return the stored value
+  /// \requires a value is stored
+  /// \group deref
+  /// \synopsis constexpr T &operator*();
+  __thrust_exec_check_disable__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &operator*() { return *m_value; }
+
+  /// \group deref
+  /// \synopsis constexpr const T &operator*() const;
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr const T &operator*() const { return *m_value; }
+
+  /// \return whether or not the optional has a value
+  /// \group has_value
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr bool has_value() const noexcept { return m_value != nullptr; }
+
+  /// \group has_value
+  __thrust_exec_check_disable__
+  __host__ __device__
+  constexpr explicit operator bool() const noexcept {
+    return m_value != nullptr;
+  }
+
+  /// \return the contained value if there is one, otherwise throws
+  /// [bad_optional_access]
+  /// \group value
+  /// synopsis constexpr T &value();
+  __host__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T &value() {
+    if (has_value())
+      return *m_value;
+    throw bad_optional_access();
+  }
+  /// \group value
+  /// \synopsis constexpr const T &value() const;
+  __host__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR const T &value() const {
+    if (has_value())
+      return *m_value;
+    throw bad_optional_access();
+  }
+
+  /// \return the stored value if there is one, otherwise returns `u`
+  /// \group value_or
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  constexpr T value_or(U &&u) const & {
+    static_assert(std::is_copy_constructible<T>::value &&
+                      std::is_convertible<U &&, T>::value,
+                  "T must be copy constructible and convertible from U");
+    return has_value() ? **this : static_cast<T>(std::forward<U>(u));
+  }
+
+  /// \group value_or
+  __thrust_exec_check_disable__
+  template <class U>
+  __host__ __device__
+  THRUST_OPTIONAL_CPP11_CONSTEXPR T value_or(U &&u) && {
+    static_assert(std::is_move_constructible<T>::value &&
+                      std::is_convertible<U &&, T>::value,
+                  "T must be move constructible and convertible from U");
+    return has_value() ? **this : static_cast<T>(std::forward<U>(u));
+  }
+
+  /// Destroys the stored value if one exists, making the optional empty
+  __thrust_exec_check_disable__
+  void reset() noexcept { m_value = nullptr; }
+
+private:
+  T *m_value;
+};
+
+THRUST_NAMESPACE_END
+
+namespace std {
+// TODO SFINAE
+template <class T> struct hash<THRUST_NS_QUALIFIER::optional<T>> {
+  __thrust_exec_check_disable__
+  __host__ __device__
+  ::std::size_t operator()(const THRUST_NS_QUALIFIER::optional<T> &o) const {
+    if (!o.has_value())
+      return 0;
+
+    return std::hash<THRUST_NS_QUALIFIER::detail::remove_const_t<T>>()(*o);
+  }
+};
+} // namespace std
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/pair.h b/thirdparty/manifold/thirdparty/thrust/thrust/pair.h
new file mode 100644
index 000000000000..eb2138aaf2ca
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/pair.h
@@ -0,0 +1,281 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file pair.h
+ *  \brief A type encapsulating a heterogeneous pair of elements
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <utility>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup pair
+ *  \{
+ */
+
+/*! \p pair is a generic data structure encapsulating a heterogeneous
+ *  pair of values.
+ *
+ *  \tparam T1 The type of \p pair's first object type.  There are no
+ *          requirements on the type of \p T1. <tt>T1</tt>'s type is
+ *          provided by <tt>pair::first_type</tt>.
+ *
+ *  \tparam T2 The type of \p pair's second object type.  There are no
+ *          requirements on the type of \p T2. <tt>T2</tt>'s type is
+ *          provided by <tt>pair::second_type</tt>.
+ */
+template <typename T1, typename T2>
+  struct pair
+{
+  /*! \p first_type is the type of \p pair's first object type.
+   */
+  typedef T1 first_type;
+
+  /*! \p second_type is the type of \p pair's second object type.
+   */
+  typedef T2 second_type;
+
+  /*! The \p pair's first object.
+   */
+  first_type first;
+
+  /*! The \p pair's second object.
+   */
+  second_type second;
+
+  /*! \p pair's default constructor constructs \p first
+   *  and \p second using \c first_type & \c second_type's
+   *  default constructors, respectively.
+   */
+  __host__ __device__ pair(void);
+
+  /*! This constructor accepts two objects to copy into this \p pair.
+   *
+   *  \param x The object to copy into \p first.
+   *  \param y The object to copy into \p second.
+   */
+  inline __host__ __device__
+  pair(const T1 &x, const T2 &y);
+
+  /*! This copy constructor copies from a \p pair whose types are
+   *  convertible to this \p pair's \c first_type and \c second_type,
+   *  respectively.
+   *
+   *  \param p The \p pair to copy from.
+   *
+   *  \tparam U1 is convertible to \c first_type.
+   *  \tparam U2 is convertible to \c second_type.
+   */
+  template <typename U1, typename U2>
+  inline __host__ __device__
+  pair(const pair<U1,U2> &p);
+
+  /*! This copy constructor copies from a <tt>std::pair</tt> whose types are
+   *  convertible to this \p pair's \c first_type and \c second_type,
+   *  respectively.
+   *
+   *  \param p The <tt>std::pair</tt> to copy from.
+   *
+   *  \tparam U1 is convertible to \c first_type.
+   *  \tparam U2 is convertible to \c second_type.
+   */
+  template <typename U1, typename U2>
+  inline __host__ __device__
+  pair(const std::pair<U1,U2> &p);
+
+  /*! \p swap swaps the elements of two <tt>pair</tt>s.
+   *  
+   *  \param p The other <tt>pair</tt> with which to swap.
+   */
+  inline __host__ __device__
+  void swap(pair &p);
+}; // end pair
+
+
+/*! This operator tests two \p pairs for equality.
+ *
+ *  \param x The first \p pair to compare.
+ *  \param y The second \p pair to compare.
+ *  \return \c true if and only if <tt>x.first == y.first && x.second == y.second</tt>.
+ *  
+ *  \tparam T1 is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam T2 is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator==(const pair<T1,T2> &x, const pair<T1,T2> &y);
+
+
+/*! This operator tests two pairs for ascending ordering.
+ *
+ *  \param x The first \p pair to compare.
+ *  \param y The second \p pair to compare.
+ *  \return \c true if and only if <tt>x.first < y.first || (!(y.first < x.first) && x.second < y.second)</tt>.
+ *
+ *  \tparam T1 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *  \tparam T2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator<(const pair<T1,T2> &x, const pair<T1,T2> &y);
+
+
+/*! This operator tests two pairs for inequality.
+ *
+ *  \param x The first \p pair to compare.
+ *  \param y The second \p pair to compare.
+ *  \return \c true if and only if <tt>!(x == y)</tt>.
+ *
+ *  \tparam T1 is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam T2 is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator!=(const pair<T1,T2> &x, const pair<T1,T2> &y);
+
+
+/*! This operator tests two pairs for descending ordering.
+ *
+ *  \param x The first \p pair to compare.
+ *  \param y The second \p pair to compare.
+ *  \return \c true if and only if <tt>y < x</tt>.
+ *
+ *  \tparam T1 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *  \tparam T2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator>(const pair<T1,T2> &x, const pair<T1,T2> &y);
+
+
+/*! This operator tests two pairs for ascending ordering or equivalence.
+ *
+ *  \param x The first \p pair to compare.
+ *  \param y The second \p pair to compare.
+ *  \return \c true if and only if <tt>!(y < x)</tt>.
+ *
+ *  \tparam T1 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *  \tparam T2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator<=(const pair<T1,T2> &x, const pair<T1,T2> &y);
+
+
+/*! This operator tests two pairs for descending ordering or equivalence.
+ *
+ *  \param x The first \p pair to compare.
+ *  \param y The second \p pair to compare.
+ *  \return \c true if and only if <tt>!(x < y)</tt>.
+ *
+ *  \tparam T1 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *  \tparam T2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    bool operator>=(const pair<T1,T2> &x, const pair<T1,T2> &y);
+
+
+/*! \p swap swaps the contents of two <tt>pair</tt>s.
+ *
+ *  \param x The first \p pair to swap.
+ *  \param y The second \p pair to swap.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    void swap(pair<T1,T2> &x, pair<T1,T2> &y);
+
+
+/*! This convenience function creates a \p pair from two objects.
+ *
+ *  \param x The first object to copy from.
+ *  \param y The second object to copy from.
+ *  \return A newly-constructed \p pair copied from \p a and \p b.
+ *
+ *  \tparam T1 There are no requirements on the type of \p T1.
+ *  \tparam T2 There are no requirements on the type of \p T2.
+ */
+template <typename T1, typename T2>
+  inline __host__ __device__
+    pair<T1,T2> make_pair(T1 x, T2 y);
+
+
+/*! This convenience metafunction is included for compatibility with
+ *  \p tuple. It returns either the type of a \p pair's
+ *  \c first_type or \c second_type in its nested type, \c type.
+ *
+ *  \tparam N This parameter selects the member of interest.
+ *  \tparam T A \c pair type of interest.
+ */
+template<size_t N, class T> struct tuple_element;
+
+
+/*! This convenience metafunction is included for compatibility with
+ *  \p tuple. It returns \c 2, the number of elements of a \p pair,
+ *  in its nested data member, \c value.
+ *
+ *  \tparam Pair A \c pair type of interest.
+ */
+template<typename Pair> struct tuple_size;
+
+
+/*! This convenience function returns a reference to either the first or
+ *  second member of a \p pair.
+ *
+ *  \param p The \p pair of interest.
+ *  \return \c p.first or \c p.second, depending on the template
+ *          parameter.
+ *
+ *  \tparam N This parameter selects the member of interest.
+ */
+// XXX comment out these prototypes as a WAR to a problem on MSVC 2005
+//template<unsigned int N, typename T1, typename T2>
+//  inline __host__ __device__
+//    typename tuple_element<N, pair<T1,T2> >::type &
+//      get(pair<T1,T2> &p);
+
+
+/*! This convenience function returns a const reference to either the
+ *  first or second member of a \p pair.
+ *
+ *  \param p The \p pair of interest.
+ *  \return \c p.first or \c p.second, depending on the template
+ *          parameter.
+ *
+ *  \tparam i This parameter selects the member of interest.
+ */
+// XXX comment out these prototypes as a WAR to a problem on MSVC 2005
+//template<int N, typename T1, typename T2>
+//  inline __host__ __device__
+//    const typename tuple_element<N, pair<T1,T2> >::type &
+//      get(const pair<T1,T2> &p);
+
+/*! \} // pair
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/pair.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/partition.h b/thirdparty/manifold/thirdparty/thrust/thrust/partition.h
new file mode 100644
index 000000000000..90768f246766
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/partition.h
@@ -0,0 +1,1436 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file partition.h
+ *  \brief Reorganizes a range based on a predicate
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reordering
+ *  \ingroup algorithms
+ *
+ *  \addtogroup partitioning
+ *  \ingroup reordering
+ *  \{
+ */
+
+
+/*! \p partition reorders the elements <tt>[first, last)</tt> based on the function
+ *  object \p pred, such that all of the elements that satisfy \p pred precede the
+ *  elements that fail to satisfy it. The postcondition is that, for some iterator
+ *  \c middle in the range <tt>[first, last)</tt>, <tt>pred(*i)</tt> is \c true for every
+ *  iterator \c i in the range <tt>[first,middle)</tt> and \c false for every iterator
+ *  \c i in the range <tt>[middle, last)</tt>. The return value of \p partition is
+ *  \c middle.
+ *
+ *  Note that the relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition, does guarantee to preserve the relative order.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements which do not satisfy \p pred.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p partition to reorder a
+ *  sequence so that even numbers precede odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::partition(thrust::host,
+ *                    A, A + N,
+ *                    is_even());
+ *  // A is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partition
+ *  \see \p stable_partition
+ *  \see \p partition_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred);
+
+
+/*! \p partition reorders the elements <tt>[first, last)</tt> based on the function
+ *  object \p pred, such that all of the elements that satisfy \p pred precede the
+ *  elements that fail to satisfy it. The postcondition is that, for some iterator
+ *  \c middle in the range <tt>[first, last)</tt>, <tt>pred(*i)</tt> is \c true for every
+ *  iterator \c i in the range <tt>[first,middle)</tt> and \c false for every iterator
+ *  \c i in the range <tt>[middle, last)</tt>. The return value of \p partition is
+ *  \c middle.
+ *
+ *  Note that the relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition, does guarantee to preserve the relative order.
+ *
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements which do not satisfy \p pred.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p partition to reorder a
+ *  sequence so that even numbers precede odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::partition(A, A + N,
+ *                     is_even());
+ *  // A is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partition
+ *  \see \p stable_partition
+ *  \see \p partition_copy
+ */
+template<typename ForwardIterator,
+         typename Predicate>
+  ForwardIterator partition(ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred);
+
+
+/*! \p partition reorders the elements <tt>[first, last)</tt> based on the function
+ *  object \p pred applied to a stencil range <tt>[stencil, stencil + (last - first))</tt>,
+ *  such that all of the elements whose corresponding stencil element satisfies \p pred precede all of the elements whose
+ *  corresponding stencil element fails to satisfy it. The postcondition is that, for some iterator
+ *  \c middle in the range <tt>[first, last)</tt>, <tt>pred(*stencil_i)</tt> is \c true for every iterator
+ *  \c stencil_i in the range <tt>[stencil,stencil + (middle - first))</tt> and \c false for every iterator \c stencil_i
+ *  in the range <tt>[stencil + (middle - first), stencil + (last - first))</tt>.
+ *  The return value of \p stable_partition is \c middle.
+ *
+ *  Note that the relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition, does guarantee to preserve the relative order.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements whose stencil elements do not satisfy \p pred.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[stencil, stencil + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p partition to reorder a
+ *  sequence so that even numbers precede odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int S[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::partition(thrust::host, A, A + N, S, is_even());
+ *  // A is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  // S is unmodified
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partition
+ *  \see \p stable_partition
+ *  \see \p partition_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred);
+
+
+/*! \p partition reorders the elements <tt>[first, last)</tt> based on the function
+ *  object \p pred applied to a stencil range <tt>[stencil, stencil + (last - first))</tt>,
+ *  such that all of the elements whose corresponding stencil element satisfies \p pred precede all of the elements whose
+ *  corresponding stencil element fails to satisfy it. The postcondition is that, for some iterator
+ *  \c middle in the range <tt>[first, last)</tt>, <tt>pred(*stencil_i)</tt> is \c true for every iterator
+ *  \c stencil_i in the range <tt>[stencil,stencil + (middle - first))</tt> and \c false for every iterator \c stencil_i
+ *  in the range <tt>[stencil + (middle - first), stencil + (last - first))</tt>.
+ *  The return value of \p stable_partition is \c middle.
+ *
+ *  Note that the relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition, does guarantee to preserve the relative order.
+ *
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements whose stencil elements do not satisfy \p pred.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The ranges <tt>[first,last)</tt> and <tt>[stencil, stencil + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p partition to reorder a
+ *  sequence so that even numbers precede odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int S[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::partition(A, A + N, S, is_even());
+ *  // A is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  // S is unmodified
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partition
+ *  \see \p stable_partition
+ *  \see \p partition_copy
+ */
+template<typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+  ForwardIterator partition(ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred);
+
+
+/*! \p partition_copy differs from \p partition only in that the reordered
+ *  sequence is written to difference output sequences, rather than in place.
+ *
+ *  \p partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred. All of the elements that satisfy \p pred are copied
+ *  to the range beginning at \p out_true and all the elements that fail to satisfy it
+ *  are copied to the range beginning at \p out_false.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type and \p InputIterator's \c value_type
+ *          is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input range shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p partition_copy to separate a
+ *  sequence into two output sequences of even and odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::partition_copy(thrust::host, A, A + N, evens, odds, is_even());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \note The relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition_copy, does guarantee to preserve the relative order.
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p stable_partition_copy
+ *  \see \p partition
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred);
+
+
+/*! \p partition_copy differs from \p partition only in that the reordered
+ *  sequence is written to difference output sequences, rather than in place.
+ *
+ *  \p partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred. All of the elements that satisfy \p pred are copied
+ *  to the range beginning at \p out_true and all the elements that fail to satisfy it
+ *  are copied to the range beginning at \p out_false.
+ *
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type and \p InputIterator's \c value_type
+ *          is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input range shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p partition_copy to separate a
+ *  sequence into two output sequences of even and odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::partition_copy(A, A + N, evens, odds, is_even());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \note The relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition_copy, does guarantee to preserve the relative order.
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p stable_partition_copy
+ *  \see \p partition
+ */
+template<typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(InputIterator first,
+                   InputIterator last,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred);
+
+
+/*! \p partition_copy differs from \p partition only in that the reordered
+ *  sequence is written to difference output sequences, rather than in place.
+ *
+ *  \p partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred which is applied to a range of stencil elements. All of the elements
+ *  whose corresponding stencil element satisfies \p pred are copied to the range beginning at \p out_true
+ *  and all the elements whose stencil element fails to satisfy it are copied to the range beginning
+ *  at \p out_false.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p partition_copy to separate a
+ *  sequence into two output sequences of even and odd numbers using the \p thrust::host execution
+ *  policy for parallelization.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int S[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::stable_partition_copy(thrust::host, A, A + N, S, evens, odds, thrust::identity<int>());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // S remains {0, 1, 0, 1, 0, 1, 0, 1, 0,  1}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \note The relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition_copy, does guarantee to preserve the relative order.
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p stable_partition_copy
+ *  \see \p partition
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   InputIterator1 first,
+                   InputIterator1 last,
+                   InputIterator2 stencil,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred);
+
+
+/*! \p partition_copy differs from \p partition only in that the reordered
+ *  sequence is written to difference output sequences, rather than in place.
+ *
+ *  \p partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred which is applied to a range of stencil elements. All of the elements
+ *  whose corresponding stencil element satisfies \p pred are copied to the range beginning at \p out_true
+ *  and all the elements whose stencil element fails to satisfy it are copied to the range beginning
+ *  at \p out_false.
+ *
+ *  \param first The beginning of the sequence to reorder.
+ *  \param last The end of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p partition_copy to separate a
+ *  sequence into two output sequences of even and odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int S[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::stable_partition_copy(A, A + N, S, evens, odds, thrust::identity<int>());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // S remains {0, 1, 0, 1, 0, 1, 0, 1, 0,  1}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \note The relative order of elements in the two reordered sequences is not
+ *  necessarily the same as it was in the original sequence. A different algorithm,
+ *  \p stable_partition_copy, does guarantee to preserve the relative order.
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p stable_partition_copy
+ *  \see \p partition
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(InputIterator1 first,
+                   InputIterator1 last,
+                   InputIterator2 stencil,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred);
+
+
+/*! \p stable_partition is much like \p partition : it reorders the elements in the
+ *  range <tt>[first, last)</tt> based on the function object \p pred, such that all of
+ *  the elements that satisfy \p pred precede all of the elements that fail to satisfy
+ *  it. The postcondition is that, for some iterator \p middle in the range
+ *  <tt>[first, last)</tt>, <tt>pred(*i)</tt> is \c true for every iterator \c i in the
+ *  range <tt>[first,middle)</tt> and \c false for every iterator \c i in the range
+ *  <tt>[middle, last)</tt>. The return value of \p stable_partition is \c middle.
+ *
+ *  \p stable_partition differs from \p partition in that \p stable_partition is
+ *  guaranteed to preserve relative order. That is, if \c x and \c y are elements in
+ *  <tt>[first, last)</tt>, and \c stencil_x and \c stencil_y are the stencil elements
+ *  in corresponding positions within <tt>[stencil, stencil + (last - first))</tt>,
+ *  and <tt>pred(stencil_x) == pred(stencil_y)</tt>, and if \c x precedes
+ *  \c y, then it will still be true after \p stable_partition that \c x precedes \c y.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements which do not satisfy pred.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition to reorder a
+ *  sequence so that even numbers precede odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::stable_partition(thrust::host,
+ *                           A, A + N,
+ *                           is_even());
+ *  // A is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_partition
+ *  \see \p partition
+ *  \see \p stable_partition_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred);
+
+
+/*! \p stable_partition is much like \p partition : it reorders the elements in the
+ *  range <tt>[first, last)</tt> based on the function object \p pred, such that all of
+ *  the elements that satisfy \p pred precede all of the elements that fail to satisfy
+ *  it. The postcondition is that, for some iterator \p middle in the range
+ *  <tt>[first, last)</tt>, <tt>pred(*i)</tt> is \c true for every iterator \c i in the
+ *  range <tt>[first,middle)</tt> and \c false for every iterator \c i in the range
+ *  <tt>[middle, last)</tt>. The return value of \p stable_partition is \c middle.
+ *
+ *  \p stable_partition differs from \p partition in that \p stable_partition is
+ *  guaranteed to preserve relative order. That is, if \c x and \c y are elements in
+ *  <tt>[first, last)</tt>, and \c stencil_x and \c stencil_y are the stencil elements
+ *  in corresponding positions within <tt>[stencil, stencil + (last - first))</tt>,
+ *  and <tt>pred(stencil_x) == pred(stencil_y)</tt>, and if \c x precedes
+ *  \c y, then it will still be true after \p stable_partition that \c x precedes \c y.
+ *
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements which do not satisfy pred.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition to reorder a
+ *  sequence so that even numbers precede odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::stable_partition(A, A + N,
+ *                            is_even());
+ *  // A is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_partition
+ *  \see \p partition
+ *  \see \p stable_partition_copy
+ */
+template<typename ForwardIterator,
+         typename Predicate>
+  ForwardIterator stable_partition(ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred);
+
+
+/*! \p stable_partition is much like \p partition: it reorders the elements in the
+ *  range <tt>[first, last)</tt> based on the function object \p pred applied to a stencil
+ *  range <tt>[stencil, stencil + (last - first))</tt>, such that all of
+ *  the elements whose corresponding stencil element satisfies \p pred precede all of the elements whose
+ *  corresponding stencil element fails to satisfy it. The postcondition is that, for some iterator
+ *  \c middle in the range <tt>[first, last)</tt>, <tt>pred(*stencil_i)</tt> is \c true for every iterator
+ *  \c stencil_i in the range <tt>[stencil,stencil + (middle - first))</tt> and \c false for every iterator \c stencil_i
+ *  in the range <tt>[stencil + (middle - first), stencil + (last - first))</tt>.
+ *  The return value of \p stable_partition is \c middle.
+ *
+ *  \p stable_partition differs from \p partition in that \p stable_partition is
+ *  guaranteed to preserve relative order. That is, if \c x and \c y are elements in
+ *  <tt>[first, last)</tt>, such that <tt>pred(x) == pred(y)</tt>, and if \c x precedes
+ *  \c y, then it will still be true after \p stable_partition that \c x precedes \c y.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements whose stencil elements do not satisfy \p pred.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap with the range <tt>[stencil, stencil + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition to reorder a
+ *  sequence so that even numbers precede odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int S[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::stable_partition(thrust::host, A, A + N, S, is_even());
+ *  // A is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  // S is unmodified
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_partition
+ *  \see \p partition
+ *  \see \p stable_partition_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred);
+
+
+/*! \p stable_partition is much like \p partition: it reorders the elements in the
+ *  range <tt>[first, last)</tt> based on the function object \p pred applied to a stencil
+ *  range <tt>[stencil, stencil + (last - first))</tt>, such that all of
+ *  the elements whose corresponding stencil element satisfies \p pred precede all of the elements whose
+ *  corresponding stencil element fails to satisfy it. The postcondition is that, for some iterator
+ *  \c middle in the range <tt>[first, last)</tt>, <tt>pred(*stencil_i)</tt> is \c true for every iterator
+ *  \c stencil_i in the range <tt>[stencil,stencil + (middle - first))</tt> and \c false for every iterator \c stencil_i
+ *  in the range <tt>[stencil + (middle - first), stencil + (last - first))</tt>.
+ *  The return value of \p stable_partition is \c middle.
+ *
+ *  \p stable_partition differs from \p partition in that \p stable_partition is
+ *  guaranteed to preserve relative order. That is, if \c x and \c y are elements in
+ *  <tt>[first, last)</tt>, such that <tt>pred(x) == pred(y)</tt>, and if \c x precedes
+ *  \c y, then it will still be true after \p stable_partition that \c x precedes \c y.
+ *
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return An iterator referring to the first element of the second partition, that is,
+ *          the sequence of the elements whose stencil elements do not satisfy \p pred.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap with the range <tt>[stencil, stencil + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition to reorder a
+ *  sequence so that even numbers precede odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int S[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::stable_partition(A, A + N, S, is_even());
+ *  // A is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  // S is unmodified
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_partition
+ *  \see \p partition
+ *  \see \p stable_partition_copy
+ */
+template<typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+  ForwardIterator stable_partition(ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred);
+
+
+/*! \p stable_partition_copy differs from \p stable_partition only in that the reordered
+ *  sequence is written to different output sequences, rather than in place.
+ *
+ *  \p stable_partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred. All of the elements that satisfy \p pred are copied
+ *  to the range beginning at \p out_true and all the elements that fail to satisfy it
+ *  are copied to the range beginning at \p out_false.
+ *
+ *  \p stable_partition_copy differs from \p partition_copy in that
+ *  \p stable_partition_copy is guaranteed to preserve relative order. That is, if
+ *  \c x and \c y are elements in <tt>[first, last)</tt>, such that
+ *  <tt>pred(x) == pred(y)</tt>, and if \c x precedes \c y, then it will still be true
+ *  after \p stable_partition_copy that \c x precedes \c y in the output.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type and \p InputIterator's \c value_type
+ *          is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition_copy to
+ *  reorder a sequence so that even numbers precede odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::stable_partition_copy(thrust::host, A, A + N, evens, odds, is_even());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p partition_copy
+ *  \see \p stable_partition
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred);
+
+
+/*! \p stable_partition_copy differs from \p stable_partition only in that the reordered
+ *  sequence is written to different output sequences, rather than in place.
+ *
+ *  \p stable_partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred. All of the elements that satisfy \p pred are copied
+ *  to the range beginning at \p out_true and all the elements that fail to satisfy it
+ *  are copied to the range beginning at \p out_false.
+ *
+ *  \p stable_partition_copy differs from \p partition_copy in that
+ *  \p stable_partition_copy is guaranteed to preserve relative order. That is, if
+ *  \c x and \c y are elements in <tt>[first, last)</tt>, such that
+ *  <tt>pred(x) == pred(y)</tt>, and if \c x precedes \c y, then it will still be true
+ *  after \p stable_partition_copy that \c x precedes \c y in the output.
+ *
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type and \p InputIterator's \c value_type
+ *          is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition_copy to
+ *  reorder a sequence so that even numbers precede odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::stable_partition_copy(A, A + N, evens, odds, is_even());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p partition_copy
+ *  \see \p stable_partition
+ */
+template<typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred);
+
+
+/*! \p stable_partition_copy differs from \p stable_partition only in that the reordered
+ *  sequence is written to different output sequences, rather than in place.
+ *
+ *  \p stable_partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred which is applied to a range of stencil elements. All of the elements
+ *  whose corresponding stencil element satisfies \p pred are copied to the range beginning at \p out_true
+ *  and all the elements whose stencil element fails to satisfy it are copied to the range beginning
+ *  at \p out_false.
+ *
+ *  \p stable_partition_copy differs from \p partition_copy in that
+ *  \p stable_partition_copy is guaranteed to preserve relative order. That is, if
+ *  \c x and \c y are elements in <tt>[first, last)</tt>, such that
+ *  <tt>pred(x) == pred(y)</tt>, and if \c x precedes \c y, then it will still be true
+ *  after \p stable_partition_copy that \c x precedes \c y in the output.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition_copy to
+ *  reorder a sequence so that even numbers precede odd numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int S[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::stable_partition_copy(thrust::host, A, A + N, S, evens, odds, thrust::identity<int>());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // S remains {0, 1, 0, 1, 0, 1, 0, 1, 0,  1}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p partition_copy
+ *  \see \p stable_partition
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred);
+
+
+/*! \p stable_partition_copy differs from \p stable_partition only in that the reordered
+ *  sequence is written to different output sequences, rather than in place.
+ *
+ *  \p stable_partition_copy copies the elements <tt>[first, last)</tt> based on the
+ *  function object \p pred which is applied to a range of stencil elements. All of the elements
+ *  whose corresponding stencil element satisfies \p pred are copied to the range beginning at \p out_true
+ *  and all the elements whose stencil element fails to satisfy it are copied to the range beginning
+ *  at \p out_false.
+ *
+ *  \p stable_partition_copy differs from \p partition_copy in that
+ *  \p stable_partition_copy is guaranteed to preserve relative order. That is, if
+ *  \c x and \c y are elements in <tt>[first, last)</tt>, such that
+ *  <tt>pred(x) == pred(y)</tt>, and if \c x precedes \c y, then it will still be true
+ *  after \p stable_partition_copy that \c x precedes \c y in the output.
+ *
+ *  \param first The first element of the sequence to reorder.
+ *  \param last One position past the last element of the sequence to reorder.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param out_true The destination of the resulting sequence of elements which satisfy \p pred.
+ *  \param out_false The destination of the resulting sequence of elements which fail to satisfy \p pred.
+ *  \param pred A function object which decides to which partition each element of the
+ *              sequence <tt>[first, last)</tt> belongs.
+ *  \return A \p pair p such that <tt>p.first</tt> is the end of the output range beginning
+ *          at \p out_true and <tt>p.second</tt> is the end of the output range beginning at
+ *          \p out_false.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p OutputIterator1 and \p OutputIterator2's \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap with either output range.
+ *
+ *  The following code snippet demonstrates how to use \p stable_partition_copy to
+ *  reorder a sequence so that even numbers precede odd numbers.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int S[] = {0, 1, 0, 1, 0, 1, 0, 1, 0,  1};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  int *evens = result;
+ *  int *odds  = result + 5;
+ *  thrust::stable_partition_copy(A, A + N, S, evens, odds, thrust::identity<int>());
+ *  // A remains {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  // S remains {0, 1, 0, 1, 0, 1, 0, 1, 0,  1}
+ *  // result is now {2, 4, 6, 8, 10, 1, 3, 5, 7, 9}
+ *  // evens points to {2, 4, 6, 8, 10}
+ *  // odds points to {1, 3, 5, 7, 9}
+ *  \endcode
+ *
+ *  \see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2569.pdf
+ *  \see \p partition_copy
+ *  \see \p stable_partition
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred);
+
+
+/*! \} // end stream_compaction
+ */
+
+/*! \} // end reordering
+ */
+
+/*! \addtogroup searching
+ *  \{
+ */
+
+
+/*! \p partition_point returns an iterator pointing to the end of the true
+ *  partition of a partitioned range. \p partition_point requires the input range
+ *  <tt>[first,last)</tt> to be a partition; that is, all elements which satisfy
+ *  <tt>pred</tt> shall appear before those that do not.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range to consider.
+ *  \param last The end of the range to consider.
+ *  \param pred A function object which decides to which partition each element of the
+ *              range <tt>[first, last)</tt> belongs.
+ *  \return An iterator \c mid such that <tt>all_of(first, mid, pred)</tt>
+ *          and <tt>none_of(mid, last, pred)</tt> are both true.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall be partitioned by \p pred.
+ *
+ *  \note Though similar, \p partition_point is not redundant with \p find_if_not.
+ *        \p partition_point's precondition provides an opportunity for a
+ *        faster implemention.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *
+ *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
+ *  int * B = thrust::partition_point(thrust::host, A, A + 10, is_even());
+ *  // B - A is 5
+ *  // [A, B) contains only even values
+ *  \endcode
+ *
+ *  \see \p partition
+ *  \see \p find_if_not
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename Predicate>
+__host__ __device__
+  ForwardIterator partition_point(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Predicate pred);
+
+
+/*! \p partition_point returns an iterator pointing to the end of the true
+ *  partition of a partitioned range. \p partition_point requires the input range
+ *  <tt>[first,last)</tt> to be a partition; that is, all elements which satisfy
+ *  <tt>pred</tt> shall appear before those that do not.
+ *  \param first The beginning of the range to consider.
+ *  \param last The end of the range to consider.
+ *  \param pred A function object which decides to which partition each element of the
+ *              range <tt>[first, last)</tt> belongs.
+ *  \return An iterator \c mid such that <tt>all_of(first, mid, pred)</tt>
+ *          and <tt>none_of(mid, last, pred)</tt> are both true.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall be partitioned by \p pred.
+ *
+ *  \note Though similar, \p partition_point is not redundant with \p find_if_not.
+ *        \p partition_point's precondition provides an opportunity for a
+ *        faster implemention.
+ *
+ *  \code
+ *  #include <thrust/partition.h>
+ *
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *
+ *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
+ *  int * B = thrust::partition_point(A, A + 10, is_even());
+ *  // B - A is 5
+ *  // [A, B) contains only even values
+ *  \endcode
+ *
+ *  \see \p partition
+ *  \see \p find_if_not
+ */
+template<typename ForwardIterator, typename Predicate>
+  ForwardIterator partition_point(ForwardIterator first,
+                                  ForwardIterator last,
+                                  Predicate pred);
+
+/*! \} // searching
+ */
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup predicates
+ *  \{
+ */
+
+
+/*! \p is_partitioned returns \c true if the given range 
+ *  is partitioned with respect to a predicate, and \c false otherwise.
+ *
+ *  Specifically, \p is_partitioned returns \c true if <tt>[first, last)</tt>
+ *  is empty of if <tt>[first, last)</tt> is partitioned by \p pred, i.e. if
+ *  all elements that satisfy \p pred appear before those that do not.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range to consider.
+ *  \param last The end of the range to consider.
+ *  \param pred A function object which decides to which partition each element of the
+ *         range <tt>[first, last)</tt> belongs.
+ *  \return \c true if the range <tt>[first, last)</tt> is partitioned with respect
+ *          to \p pred, or if <tt>[first, last)</tt> is empty. \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  
+ *  \code
+ *  #include <thrust/partition.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *
+ *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
+ *  int B[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *
+ *  thrust::is_partitioned(thrust::host, A, A + 10, is_even()); // returns true
+ *  thrust::is_partitioned(thrust::host, B, B + 10, is_even()); // returns false
+ *  \endcode
+ *
+ *  \see \p partition
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+  bool is_partitioned(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred);
+
+
+/*! \p is_partitioned returns \c true if the given range 
+ *  is partitioned with respect to a predicate, and \c false otherwise.
+ *
+ *  Specifically, \p is_partitioned returns \c true if <tt>[first, last)</tt>
+ *  is empty of if <tt>[first, last)</tt> is partitioned by \p pred, i.e. if
+ *  all elements that satisfy \p pred appear before those that do not.
+ *
+ *  \param first The beginning of the range to consider.
+ *  \param last The end of the range to consider.
+ *  \param pred A function object which decides to which partition each element of the
+ *         range <tt>[first, last)</tt> belongs.
+ *  \return \c true if the range <tt>[first, last)</tt> is partitioned with respect
+ *          to \p pred, or if <tt>[first, last)</tt> is empty. \c false, otherwise.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  
+ *  \code
+ *  #include <thrust/partition.h>
+ *
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int &x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *
+ *  int A[] = {2, 4, 6, 8, 10, 1, 3, 5, 7, 9};
+ *  int B[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *
+ *  thrust::is_partitioned(A, A + 10, is_even()); // returns true
+ *  thrust::is_partitioned(B, B + 10, is_even()); // returns false
+ *  \endcode
+ *
+ *  \see \p partition
+ */
+template<typename InputIterator, typename Predicate>
+  bool is_partitioned(InputIterator first,
+                      InputIterator last,
+                      Predicate pred);
+
+
+/*! \} // end predicates
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/partition.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/per_device_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/per_device_resource.h
new file mode 100644
index 000000000000..a6d620f85ec0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/per_device_resource.h
@@ -0,0 +1,102 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/system/detail/generic/per_device_resource.h>
+#include <thrust/system/detail/adl/per_device_resource.h>
+#include <thrust/mr/allocator.h>
+
+#include <thrust/detail/execution_policy.h>
+#include <thrust/mr/allocator.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! Returns a global instance of \p MR for the current device of the provided system.
+ *
+ *  \tparam MR type of a memory resource to get an instance from. Must be \p DefaultConstructible.
+ *  \param system execution policy for which the resource is requested.
+ *  \return a pointer to a global instance of \p MR for the current device.
+ */
+template<typename MR, typename DerivedPolicy>
+__host__
+MR * get_per_device_resource(const thrust::detail::execution_policy_base<DerivedPolicy> & system)
+{
+    using thrust::system::detail::generic::get_per_device_resource;
+
+    return get_per_device_resource<MR>(
+        thrust::detail::derived_cast(
+            thrust::detail::strip_const(system)));
+}
+
+/*! A helper allocator class that uses global per device instances of a given upstream memory resource. Requires the memory
+ *      resource to be default constructible.
+ *
+ *  \tparam T the type that will be allocated by this allocator.
+ *  \tparam MR the upstream memory resource to use for memory allocation. Must derive from
+ *      \p thrust::mr::memory_resource and must be \p final.
+ *  \tparam ExecutionPolicy the execution policy of the system to be used to retrieve the resource for the current device.
+ */
+template<typename T, typename Upstream, typename ExecutionPolicy>
+class per_device_allocator : public thrust::mr::allocator<T, Upstream>
+{
+    typedef thrust::mr::allocator<T, Upstream> base;
+
+public:
+    /*! The \p rebind metafunction provides the type of an \p per_device_allocator instantiated with another type.
+     *
+     *  \tparam U the other type to use for instantiation.
+     */
+    template<typename U>
+    struct rebind
+    {
+        /*! The typedef \p other gives the type of the rebound \p per_device_allocator.
+         */
+        typedef per_device_allocator<U, Upstream, ExecutionPolicy> other;
+    };
+
+    /*! Default constructor. Uses \p get_global_resource to get the global instance of \p Upstream and initializes the
+     *      \p allocator base subobject with that resource.
+     */
+    __host__
+    per_device_allocator() : base(get_per_device_resource<Upstream>(ExecutionPolicy()))
+    {
+    }
+
+    /*! Copy constructor. Copies the memory resource pointer. */
+    __host__ __device__
+    per_device_allocator(const per_device_allocator & other)
+        : base(other) {}
+
+    /*! Conversion constructor from an allocator of a different type. Copies the memory resource pointer. */
+    template<typename U>
+    __host__ __device__
+    per_device_allocator(const per_device_allocator<U, Upstream, ExecutionPolicy> & other)
+        : base(other) {}
+
+    /*! Destructor. */
+    __host__ __device__
+    ~per_device_allocator() {}
+};
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/random.h b/thirdparty/manifold/thirdparty/thrust/thrust/random.h
new file mode 100644
index 000000000000..7463620b7bce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/random.h
@@ -0,0 +1,117 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file random.h
+ *  \brief Pseudo-random number generators.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cstdint.h>
+
+// RNGs
+#include <thrust/random/discard_block_engine.h>
+#include <thrust/random/linear_congruential_engine.h>
+#include <thrust/random/linear_feedback_shift_engine.h>
+#include <thrust/random/subtract_with_carry_engine.h>
+#include <thrust/random/xor_combine_engine.h>
+
+// distributions
+#include <thrust/random/uniform_int_distribution.h>
+#include <thrust/random/uniform_real_distribution.h>
+#include <thrust/random/normal_distribution.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup random Random Number Generation
+ *  \{
+ */
+
+
+/*! \namespace thrust::random
+ *  \brief \p thrust::random is the namespace which contains random number engine class templates,
+ *  random number engine adaptor class templates, engines with predefined parameters,
+ *  and random number distribution class templates. They are provided in a separate namespace
+ *  for import convenience but are also aliased in the top-level \p thrust namespace for
+ *  easy access.
+ */
+namespace random
+{
+
+/*! \addtogroup predefined_random Random Number Engines with Predefined Parameters
+ *  \ingroup random
+ *  \{
+ */
+
+/*! \typedef ranlux24
+ *  \brief A random number engine with predefined parameters which implements the
+ *         RANLUX level-3 random number generation algorithm.
+ *  \note The 10000th consecutive invocation of a default-constructed object of type \p ranlux24
+ *        shall produce the value \c 9901578 .
+ */
+typedef discard_block_engine<ranlux24_base, 223, 23> ranlux24;
+
+
+/*! \typedef ranlux48
+ *  \brief A random number engine with predefined parameters which implements the
+ *         RANLUX level-4 random number generation algorithm.
+ *  \note The 10000th consecutive invocation of a default-constructed object of type \p ranlux48
+ *        shall produce the value \c 88229545517833 .
+ */
+typedef discard_block_engine<ranlux48_base, 389, 11> ranlux48;
+
+
+/*! \typedef taus88
+ *  \brief A random number engine with predefined parameters which implements
+ *         L'Ecuyer's 1996 three-component Tausworthe random number generator.
+ *
+ *  \note The 10000th consecutive invocation of a default-constructed object of type \p taus88
+ *        shall produce the value \c 3535848941 .
+ */
+typedef xor_combine_engine<
+  linear_feedback_shift_engine<thrust::detail::uint32_t, 32u, 31u, 13u, 12u>,
+  0,
+  xor_combine_engine<
+    linear_feedback_shift_engine<thrust::detail::uint32_t, 32u, 29u,  2u,  4u>, 0,
+    linear_feedback_shift_engine<thrust::detail::uint32_t, 32u, 28u,  3u, 17u>, 0
+  >,
+  0
+> taus88;
+
+/*! \typedef default_random_engine
+ *  \brief An implementation-defined "default" random number engine.
+ *  \note \p default_random_engine is currently an alias for \p minstd_rand, and may change
+ *        in a future version.
+ */
+typedef minstd_rand default_random_engine;
+
+/*! \} // end predefined_random
+ */
+
+} // end random
+
+
+/*! \} // end random
+ */
+
+// import names into thrust::
+using random::ranlux24;
+using random::ranlux48;
+using random::taus88;
+using random::default_random_engine;
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/reduce.h
new file mode 100644
index 000000000000..c7b378f7297e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/reduce.h
@@ -0,0 +1,781 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/reduce.h
+ *  \brief Functions for reducing a range to a single value
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reductions
+ *  \{
+ */
+
+
+/*! \p reduce is a generalization of summation: it computes the sum (or some
+ *  other binary operation) of all the elements in the range <tt>[first,
+ *  last)</tt>. This version of \p reduce uses \c 0 as the initial value of the
+ *  reduction. \p reduce is similar to the C++ Standard Template Library's
+ *  <tt>std::accumulate</tt>. The primary difference between the two functions
+ *  is that <tt>std::accumulate</tt> guarantees the order of summation, while
+ *  \p reduce requires associativity of the binary operation to parallelize
+ *  the reduction.
+ *
+ *  Note that \p reduce also assumes that the binary reduction operator (in this
+ *  case operator+) is commutative.  If the reduction operator is not commutative
+ *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  \p inclusive_scan (which does not require commutativity) and select the
+ *  last element of the output array.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \return The result of the reduction.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and if \c x and \c y are objects of \p InputIterator's \c value_type,
+ *          then <tt>x + y</tt> is defined and is convertible to \p InputIterator's
+ *          \c value_type. If \c T is \c InputIterator's \c value_type, then
+ *          <tt>T(0)</tt> is defined.
+ *
+ *  The following code snippet demonstrates how to use \p reduce to compute
+ *  the sum of a sequence of integers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int result = thrust::reduce(thrust::host, data, data + 6);
+ *
+ *  // result == 9
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/accumulate
+ */
+template<typename DerivedPolicy, typename InputIterator>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::value_type
+    reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator first, InputIterator last);
+
+
+/*! \p reduce is a generalization of summation: it computes the sum (or some
+ *  other binary operation) of all the elements in the range <tt>[first,
+ *  last)</tt>. This version of \p reduce uses \c 0 as the initial value of the
+ *  reduction. \p reduce is similar to the C++ Standard Template Library's
+ *  <tt>std::accumulate</tt>. The primary difference between the two functions
+ *  is that <tt>std::accumulate</tt> guarantees the order of summation, while
+ *  \p reduce requires associativity of the binary operation to parallelize
+ *  the reduction.
+ *
+ *  Note that \p reduce also assumes that the binary reduction operator (in this
+ *  case operator+) is commutative.  If the reduction operator is not commutative
+ *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  \p inclusive_scan (which does not require commutativity) and select the
+ *  last element of the output array.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \return The result of the reduction.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and if \c x and \c y are objects of \p InputIterator's \c value_type,
+ *          then <tt>x + y</tt> is defined and is convertible to \p InputIterator's
+ *          \c value_type. If \c T is \c InputIterator's \c value_type, then
+ *          <tt>T(0)</tt> is defined.
+ *
+ *  The following code snippet demonstrates how to use \p reduce to compute
+ *  the sum of a sequence of integers.
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int result = thrust::reduce(data, data + 6);
+ *
+ *  // result == 9
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/accumulate
+ */
+template<typename InputIterator> typename
+  thrust::iterator_traits<InputIterator>::value_type reduce(InputIterator first, InputIterator last);
+
+
+/*! \p reduce is a generalization of summation: it computes the sum (or some
+ *  other binary operation) of all the elements in the range <tt>[first,
+ *  last)</tt>. This version of \p reduce uses \p init as the initial value of the
+ *  reduction. \p reduce is similar to the C++ Standard Template Library's
+ *  <tt>std::accumulate</tt>. The primary difference between the two functions
+ *  is that <tt>std::accumulate</tt> guarantees the order of summation, while
+ *  \p reduce requires associativity of the binary operation to parallelize
+ *  the reduction.
+ *
+ *  Note that \p reduce also assumes that the binary reduction operator (in this
+ *  case operator+) is commutative.  If the reduction operator is not commutative
+ *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  \p inclusive_scan (which does not require commutativity) and select the
+ *  last element of the output array.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param init The initial value.
+ *  \return The result of the reduction.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and if \c x and \c y are objects of \p InputIterator's \c value_type,
+ *          then <tt>x + y</tt> is defined and is convertible to \p T.
+ *  \tparam T is convertible to \p InputIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p reduce to compute
+ *  the sum of a sequence of integers including an intialization value using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int result = thrust::reduce(thrust::host, data, data + 6, 1);
+ *
+ *  // result == 10
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/accumulate
+ */
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+  T reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+           InputIterator first,
+           InputIterator last,
+           T init);
+
+
+/*! \p reduce is a generalization of summation: it computes the sum (or some
+ *  other binary operation) of all the elements in the range <tt>[first,
+ *  last)</tt>. This version of \p reduce uses \p init as the initial value of the
+ *  reduction. \p reduce is similar to the C++ Standard Template Library's
+ *  <tt>std::accumulate</tt>. The primary difference between the two functions
+ *  is that <tt>std::accumulate</tt> guarantees the order of summation, while
+ *  \p reduce requires associativity of the binary operation to parallelize
+ *  the reduction.
+ *
+ *  Note that \p reduce also assumes that the binary reduction operator (in this
+ *  case operator+) is commutative.  If the reduction operator is not commutative
+ *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  \p inclusive_scan (which does not require commutativity) and select the
+ *  last element of the output array.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param init The initial value.
+ *  \return The result of the reduction.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and if \c x and \c y are objects of \p InputIterator's \c value_type,
+ *          then <tt>x + y</tt> is defined and is convertible to \p T.
+ *  \tparam T is convertible to \p InputIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p reduce to compute
+ *  the sum of a sequence of integers including an intialization value.
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int result = thrust::reduce(data, data + 6, 1);
+ *
+ *  // result == 10
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/accumulate
+ */
+template<typename InputIterator, typename T>
+  T reduce(InputIterator first,
+           InputIterator last,
+           T init);
+
+
+/*! \p reduce is a generalization of summation: it computes the sum (or some
+ *  other binary operation) of all the elements in the range <tt>[first,
+ *  last)</tt>. This version of \p reduce uses \p init as the initial value of the
+ *  reduction and \p binary_op as the binary function used for summation. \p reduce
+ *  is similar to the C++ Standard Template Library's <tt>std::accumulate</tt>.
+ *  The primary difference between the two functions is that <tt>std::accumulate</tt>
+ *  guarantees the order of summation, while \p reduce requires associativity of
+ *  \p binary_op to parallelize the reduction.
+ *
+ *  Note that \p reduce also assumes that the binary reduction operator (in this
+ *  case \p binary_op) is commutative.  If the reduction operator is not commutative
+ *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  \p inclusive_scan (which does not require commutativity) and select the
+ *  last element of the output array.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param init The initial value.
+ *  \param binary_op The binary function used to 'sum' values.
+ *  \return The result of the reduction.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and \c InputIterator's \c value_type is convertible to \c T.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and is convertible to \p BinaryFunction's \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction's \c result_type is convertible to \p OutputType.
+ *
+ *  The following code snippet demonstrates how to use \p reduce to
+ *  compute the maximum value of a sequence of integers using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int result = thrust::reduce(thrust::host,
+ *                              data, data + 6,
+ *                              -1,
+ *                              thrust::maximum<int>());
+ *  // result == 3
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/accumulate
+ *  \see transform_reduce
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename T,
+         typename BinaryFunction>
+__host__ __device__
+  T reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+           InputIterator first,
+           InputIterator last,
+           T init,
+           BinaryFunction binary_op);
+
+
+/*! \p reduce is a generalization of summation: it computes the sum (or some
+ *  other binary operation) of all the elements in the range <tt>[first,
+ *  last)</tt>. This version of \p reduce uses \p init as the initial value of the
+ *  reduction and \p binary_op as the binary function used for summation. \p reduce
+ *  is similar to the C++ Standard Template Library's <tt>std::accumulate</tt>.
+ *  The primary difference between the two functions is that <tt>std::accumulate</tt>
+ *  guarantees the order of summation, while \p reduce requires associativity of
+ *  \p binary_op to parallelize the reduction.
+ *
+ *  Note that \p reduce also assumes that the binary reduction operator (in this
+ *  case \p binary_op) is commutative.  If the reduction operator is not commutative
+ *  then \p thrust::reduce should not be used.  Instead, one could use 
+ *  \p inclusive_scan (which does not require commutativity) and select the
+ *  last element of the output array.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param init The initial value.
+ *  \param binary_op The binary function used to 'sum' values.
+ *  \return The result of the reduction.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *          and \c InputIterator's \c value_type is convertible to \c T.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and is convertible to \p BinaryFunction's \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction's \c result_type is convertible to \p OutputType.
+ *
+ *  The following code snippet demonstrates how to use \p reduce to
+ *  compute the maximum value of a sequence of integers.
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *  int result = thrust::reduce(data, data + 6,
+ *                              -1,
+ *                              thrust::maximum<int>());
+ *  // result == 3
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/accumulate
+ *  \see transform_reduce
+ */
+template<typename InputIterator,
+         typename T,
+         typename BinaryFunction>
+  T reduce(InputIterator first,
+           InputIterator last,
+           T init,
+           BinaryFunction binary_op);
+
+
+/*! \p reduce_by_key is a generalization of \p reduce to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p reduce_by_key copies the first element of the group to the
+ *  \c keys_output. The corresponding values in the range are reduced using the
+ *  \c plus and the result copied to \c values_output. 
+ *
+ *  This version of \p reduce_by_key uses the function object \c equal_to
+ *  to test for equality and \c plus to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_output The beginning of the output key range.
+ *  \param values_output The beginning of the output value range.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_output, keys_output_last)</tt> and <tt>[values_output, values_output_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p reduce_by_key to
+ *  compact a sequence of key/value pairs and sum values with equal keys using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  new_end = thrust::reduce_by_key(thrust::host, A, A + N, B, C, D);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
+ *  \endcode
+ *  
+ *  \see reduce
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see unique_by_key_copy
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                InputIterator1 keys_first, 
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output);
+
+
+/*! \p reduce_by_key is a generalization of \p reduce to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p reduce_by_key copies the first element of the group to the
+ *  \c keys_output. The corresponding values in the range are reduced using the
+ *  \c plus and the result copied to \c values_output. 
+ *
+ *  This version of \p reduce_by_key uses the function object \c equal_to
+ *  to test for equality and \c plus to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_output The beginning of the output key range.
+ *  \param values_output The beginning of the output value range.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_output, keys_output_last)</tt> and <tt>[values_output, values_output_last)</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p reduce_by_key to
+ *  compact a sequence of key/value pairs and sum values with equal keys.
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  new_end = thrust::reduce_by_key(A, A + N, B, C, D);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
+ *  \endcode
+ *  
+ *  \see reduce
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see unique_by_key_copy
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(InputIterator1 keys_first, 
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output);
+
+
+/*! \p reduce_by_key is a generalization of \p reduce to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p reduce_by_key copies the first element of the group to the
+ *  \c keys_output. The corresponding values in the range are reduced using the
+ *  \c plus and the result copied to \c values_output. 
+ *
+ *  This version of \p reduce_by_key uses the function object \c binary_pred
+ *  to test for equality and \c plus to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_output The beginning of the output key range.
+ *  \param values_output The beginning of the output value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_output, keys_output_last)</tt> and <tt>[values_output, values_output_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p reduce_by_key to
+ *  compact a sequence of key/value pairs and sum values with equal keys using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  new_end = thrust::reduce_by_key(thrust::host, A, A + N, B, C, D, binary_pred);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
+ *  \endcode
+ *  
+ *  \see reduce
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see unique_by_key_copy
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                InputIterator1 keys_first, 
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred);
+
+
+/*! \p reduce_by_key is a generalization of \p reduce to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p reduce_by_key copies the first element of the group to the
+ *  \c keys_output. The corresponding values in the range are reduced using the
+ *  \c plus and the result copied to \c values_output. 
+ *
+ *  This version of \p reduce_by_key uses the function object \c binary_pred
+ *  to test for equality and \c plus to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_output The beginning of the output key range.
+ *  \param values_output The beginning of the output value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_output, keys_output_last)</tt> and <tt>[values_output, values_output_last)</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p reduce_by_key to
+ *  compact a sequence of key/value pairs and sum values with equal keys.
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  new_end = thrust::reduce_by_key(A, A + N, B, C, D, binary_pred);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
+ *  \endcode
+ *  
+ *  \see reduce
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see unique_by_key_copy
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(InputIterator1 keys_first, 
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred);
+
+
+/*! \p reduce_by_key is a generalization of \p reduce to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p reduce_by_key copies the first element of the group to the
+ *  \c keys_output. The corresponding values in the range are reduced using the
+ *  \c BinaryFunction \c binary_op and the result copied to \c values_output. 
+ *  Specifically, if consecutive key iterators \c i and \c (i + 1) are 
+ *  such that <tt>binary_pred(*i, *(i+1))</tt> is \c true, then the corresponding
+ *  values are reduced to a single value with \c binary_op.
+ *
+ *  This version of \p reduce_by_key uses the function object \c binary_pred
+ *  to test for equality and \c binary_op to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_output The beginning of the output key range.
+ *  \param values_output The beginning of the output value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \param binary_op The binary function used to accumulate values.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_output, keys_output_last)</tt> and <tt>[values_output, values_output_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *          and \c BinaryFunction's \c result_type is convertible to \c OutputIterator2's \c value_type.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p reduce_by_key to
+ *  compact a sequence of key/value pairs and sum values with equal keys using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  thrust::plus<int> binary_op;
+ *  new_end = thrust::reduce_by_key(thrust::host, A, A + N, B, C, D, binary_pred, binary_op);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
+ *  \endcode
+ *  
+ *  \see reduce
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see unique_by_key_copy
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                InputIterator1 keys_first, 
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred,
+                BinaryFunction binary_op);
+
+
+/*! \p reduce_by_key is a generalization of \p reduce to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p reduce_by_key copies the first element of the group to the
+ *  \c keys_output. The corresponding values in the range are reduced using the
+ *  \c BinaryFunction \c binary_op and the result copied to \c values_output. 
+ *  Specifically, if consecutive key iterators \c i and \c (i + 1) are 
+ *  such that <tt>binary_pred(*i, *(i+1))</tt> is \c true, then the corresponding
+ *  values are reduced to a single value with \c binary_op.
+ *
+ *  This version of \p reduce_by_key uses the function object \c binary_pred
+ *  to test for equality and \c binary_op to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_output The beginning of the output key range.
+ *  \param values_output The beginning of the output value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \param binary_op The binary function used to accumulate values.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_output, keys_output_last)</tt> and <tt>[values_output, values_output_last)</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *          and \c BinaryFunction's \c result_type is convertible to \c OutputIterator2's \c value_type.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p reduce_by_key to
+ *  compact a sequence of key/value pairs and sum values with equal keys.
+ *
+ *  \code
+ *  #include <thrust/reduce.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  thrust::plus<int> binary_op;
+ *  new_end = thrust::reduce_by_key(A, A + N, B, C, D, binary_pred, binary_op);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 21, 9, 3} and new_end.second - D is 4.
+ *  \endcode
+ *  
+ *  \see reduce
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see unique_by_key_copy
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  reduce_by_key(InputIterator1 keys_first, 
+                InputIterator1 keys_last,
+                InputIterator2 values_first,
+                OutputIterator1 keys_output,
+                OutputIterator2 values_output,
+                BinaryPredicate binary_pred,
+                BinaryFunction binary_op);
+
+
+/*! \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/reduce.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/remove.h b/thirdparty/manifold/thirdparty/thrust/thrust/remove.h
new file mode 100644
index 000000000000..a57fcf211cf3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/remove.h
@@ -0,0 +1,802 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file remove.h
+ *  \brief Functions for removing elements from a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup stream_compaction Stream Compaction
+ *  \ingroup reordering
+ *  \{
+ *
+ */
+
+
+/*! \p remove removes from the range <tt>[first, last)</tt> all elements that are
+ *  equal to \p value. That is, \p remove returns an iterator \p new_last such
+ *  that the range <tt>[first, new_last)</tt> contains no elements equal to
+ *  \p value. The iterators in the range <tt>[new_first,last)</tt> are all still
+ *  dereferenceable, but the elements that they point to are unspecified. \p remove
+ *  is stable, meaning that the relative order of elements that are not equal to
+ *  \p value is unchanged.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param value The value to remove from the range <tt>[first, last)</tt>.
+ *         Elements which are equal to value are removed from the sequence.
+ *  \return A \p ForwardIterator pointing to the end of the resulting range of
+ *          elements which are not equal to \p value.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and objects of type \p T can be compared for equality with objects of \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p remove to remove a number
+ *  of interest from a range using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {3, 1, 4, 1, 5, 9};
+ *  int *new_end = thrust::remove(A, A + N, 1);
+ *  // The first four values of A are now {3, 4, 5, 9}
+ *  // Values beyond new_end are unspecified
+ *  \endcode
+ *
+ *  \note The meaning of "removal" is somewhat subtle. \p remove does not destroy any
+ *  iterators, and does not change the distance between \p first and \p last.
+ *  (There's no way that it could do anything of the sort.) So, for example, if
+ *  \c V is a device_vector, <tt>remove(V.begin(), V.end(), 0)</tt> does not
+ *  change <tt>V.size()</tt>: \c V will contain just as many elements as it did
+ *  before. \p remove returns an iterator that points to the end of the resulting
+ *  range after elements have been removed from it; it follows that the elements
+ *  after that iterator are of no interest, and may be discarded. If you are
+ *  removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a>, you may
+ *  simply erase them. That is, a reasonable way of removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a> is
+ *  <tt>S.erase(remove(S.begin(), S.end(), x), S.end())</tt>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove
+ *  \see remove_if
+ *  \see remove_copy
+ *  \see remove_copy_if
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  ForwardIterator remove(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last,
+                         const T &value);
+
+
+/*! \p remove removes from the range <tt>[first, last)</tt> all elements that are
+ *  equal to \p value. That is, \p remove returns an iterator \p new_last such
+ *  that the range <tt>[first, new_last)</tt> contains no elements equal to
+ *  \p value. The iterators in the range <tt>[new_first,last)</tt> are all still
+ *  dereferenceable, but the elements that they point to are unspecified. \p remove
+ *  is stable, meaning that the relative order of elements that are not equal to
+ *  \p value is unchanged.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param value The value to remove from the range <tt>[first, last)</tt>.
+ *         Elements which are equal to value are removed from the sequence.
+ *  \return A \p ForwardIterator pointing to the end of the resulting range of
+ *          elements which are not equal to \p value.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and objects of type \p T can be compared for equality with objects of \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p remove to remove a number
+ *  of interest from a range.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {3, 1, 4, 1, 5, 9};
+ *  int *new_end = thrust::remove(A, A + N, 1);
+ *  // The first four values of A are now {3, 4, 5, 9}
+ *  // Values beyond new_end are unspecified
+ *  \endcode
+ *
+ *  \note The meaning of "removal" is somewhat subtle. \p remove does not destroy any
+ *  iterators, and does not change the distance between \p first and \p last.
+ *  (There's no way that it could do anything of the sort.) So, for example, if
+ *  \c V is a device_vector, <tt>remove(V.begin(), V.end(), 0)</tt> does not
+ *  change <tt>V.size()</tt>: \c V will contain just as many elements as it did
+ *  before. \p remove returns an iterator that points to the end of the resulting
+ *  range after elements have been removed from it; it follows that the elements
+ *  after that iterator are of no interest, and may be discarded. If you are
+ *  removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a>, you may
+ *  simply erase them. That is, a reasonable way of removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a> is
+ *  <tt>S.erase(remove(S.begin(), S.end(), x), S.end())</tt>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove
+ *  \see remove_if
+ *  \see remove_copy
+ *  \see remove_copy_if
+ */
+template<typename ForwardIterator,
+         typename T>
+  ForwardIterator remove(ForwardIterator first,
+                         ForwardIterator last,
+                         const T &value);
+
+
+/*! \p remove_copy copies elements that are not equal to \p value from the range
+ *  <tt>[first, last)</tt> to a range beginning at \p result. The return value is
+ *  the end of the resulting range. This operation is stable, meaning that the
+ *  relative order of the elements that are copied is the same as in
+ *  the range <tt>[first, last)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param result The resulting range is copied to the sequence beginning at this
+ *                location.
+ *  \param value The value to omit from the copied range.
+ *  \return An OutputIterator pointing to the end of the resulting range of elements
+ *          which are not equal to \p value.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and objects of type \p T can be compared for equality with objects of \p InputIterator's \c value_type.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_copy to copy
+ *  a sequence of numbers to an output range while omitting a value of interest using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[N-2];
+ *  thrust::remove_copy(thrust::host, V, V + N, result, 0);
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-2, -1, 1, 2}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove_copy
+ *  \see remove
+ *  \see remove_if
+ *  \see remove_copy_if
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator remove_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator result,
+                             const T &value);
+
+
+/*! \p remove_copy copies elements that are not equal to \p value from the range
+ *  <tt>[first, last)</tt> to a range beginning at \p result. The return value is
+ *  the end of the resulting range. This operation is stable, meaning that the
+ *  relative order of the elements that are copied is the same as in
+ *  the range <tt>[first, last)</tt>.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param result The resulting range is copied to the sequence beginning at this
+ *                location.
+ *  \param value The value to omit from the copied range.
+ *  \return An OutputIterator pointing to the end of the resulting range of elements
+ *          which are not equal to \p value.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          and objects of type \p T can be compared for equality with objects of \p InputIterator's \c value_type.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_copy to copy
+ *  a sequence of numbers to an output range while omitting a value of interest.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[N-2];
+ *  thrust::remove_copy(V, V + N, result, 0);
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-2, -1, 1, 2}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove_copy
+ *  \see remove
+ *  \see remove_if
+ *  \see remove_copy_if
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename T>
+  OutputIterator remove_copy(InputIterator first,
+                             InputIterator last,
+                             OutputIterator result,
+                             const T &value);
+
+
+/*! \p remove_if removes from the range <tt>[first, last)</tt> every element \p x
+ *  such that <tt>pred(x)</tt> is \c true. That is, \p remove_if returns an
+ *  iterator \c new_last such that the range <tt>[first,new_last)</tt> contains
+ *  no elements for which \p pred is \c true. The iterators in the range
+ *  <tt>[new_last,last)</tt> are all still dereferenceable, but the elements that
+ *  they point to are unspecified. \p remove_if is stable, meaning that the
+ *  relative order of elements that are not removed is unchanged.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param pred A predicate to evaluate for each element of the range
+ *              <tt>[first,last)</tt>. Elements for which \p pred evaluates to
+ *              \c true are removed from the sequence.
+ *  \return A ForwardIterator pointing to the end of the resulting range of
+ *          elements for which \p pred evaluated to \c true.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_if to remove
+ *  all even numbers from an array of integers using the \p thrust::host execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  int *new_end = thrust::remove_if(thrust::host, A, A + N, is_even());
+ *  // The first three values of A are now {1, 5, 7}
+ *  // Values beyond new_end are unspecified
+ *  \endcode
+ *
+ *  \note The meaning of "removal" is somewhat subtle. \p remove_if does not
+ *  destroy any iterators, and does not change the distance between \p first and
+ *  \p last. (There's no way that it could do anything of the sort.) So, for
+ *  example, if \c V is a device_vector,
+ *  <tt>remove_if(V.begin(), V.end(), pred)</tt> does not change
+ *  <tt>V.size()</tt>: \c V will contain just as many elements as it did before.
+ *  \p remove_if returns an iterator that points to the end of the resulting
+ *  range after elements have been removed from it; it follows that the elements
+ *  after that iterator are of no interest, and may be discarded. If you are
+ *  removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a>, you may
+ *  simply erase them. That is, a reasonable way of removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a> is
+ *  <tt>S.erase(remove_if(S.begin(), S.end(), pred), S.end())</tt>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_copy_if
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred);
+
+
+/*! \p remove_if removes from the range <tt>[first, last)</tt> every element \p x
+ *  such that <tt>pred(x)</tt> is \c true. That is, \p remove_if returns an
+ *  iterator \c new_last such that the range <tt>[first,new_last)</tt> contains
+ *  no elements for which \p pred is \c true. The iterators in the range
+ *  <tt>[new_last,last)</tt> are all still dereferenceable, but the elements that
+ *  they point to are unspecified. \p remove_if is stable, meaning that the
+ *  relative order of elements that are not removed is unchanged.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param pred A predicate to evaluate for each element of the range
+ *              <tt>[first,last)</tt>. Elements for which \p pred evaluates to
+ *              \c true are removed from the sequence.
+ *  \return A ForwardIterator pointing to the end of the resulting range of
+ *          elements for which \p pred evaluated to \c true.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_if to remove
+ *  all even numbers from an array of integers.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  int *new_end = thrust::remove_if(A, A + N, is_even());
+ *  // The first three values of A are now {1, 5, 7}
+ *  // Values beyond new_end are unspecified
+ *  \endcode
+ *
+ *  \note The meaning of "removal" is somewhat subtle. \p remove_if does not
+ *  destroy any iterators, and does not change the distance between \p first and
+ *  \p last. (There's no way that it could do anything of the sort.) So, for
+ *  example, if \c V is a device_vector,
+ *  <tt>remove_if(V.begin(), V.end(), pred)</tt> does not change
+ *  <tt>V.size()</tt>: \c V will contain just as many elements as it did before.
+ *  \p remove_if returns an iterator that points to the end of the resulting
+ *  range after elements have been removed from it; it follows that the elements
+ *  after that iterator are of no interest, and may be discarded. If you are
+ *  removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a>, you may
+ *  simply erase them. That is, a reasonable way of removing elements from a
+ *  <a href="https://en.cppreference.com/w/cpp/container">Sequence</a> is
+ *  <tt>S.erase(remove_if(S.begin(), S.end(), pred), S.end())</tt>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_copy_if
+ */
+template<typename ForwardIterator,
+         typename Predicate>
+  ForwardIterator remove_if(ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred);
+
+
+/*! \p remove_copy_if copies elements from the range <tt>[first,last)</tt> to a
+ *  range beginning at \p result, except that elements for which \p pred is
+ *  \c true are not copied. The return value is the end of the resulting range.
+ *  This operation is stable, meaning that the relative order of the elements that
+ *  are copied is the same as the range <tt>[first,last)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param result The resulting range is copied to the sequence beginning at this
+ *                location.
+ *  \param pred A predicate to evaluate for each element of the range <tt>[first,last)</tt>.
+ *              Elements for which \p pred evaluates to \c false are not copied
+ *              to the resulting sequence.
+ *  \return An OutputIterator pointing to the end of the resulting range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_copy_if to copy
+ *  a sequence of numbers to an output range while omitting even numbers using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[2];
+ *  thrust::remove_copy_if(thrust::host, V, V + N, result, is_even());
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-1, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove_copy
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_if
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred);
+
+
+/*! \p remove_copy_if copies elements from the range <tt>[first,last)</tt> to a
+ *  range beginning at \p result, except that elements for which \p pred is
+ *  \c true are not copied. The return value is the end of the resulting range.
+ *  This operation is stable, meaning that the relative order of the elements that
+ *  are copied is the same as the range <tt>[first,last)</tt>.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param result The resulting range is copied to the sequence beginning at this
+ *                location.
+ *  \param pred A predicate to evaluate for each element of the range <tt>[first,last)</tt>.
+ *              Elements for which \p pred evaluates to \c false are not copied
+ *              to the resulting sequence.
+ *  \return An OutputIterator pointing to the end of the resulting range.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_copy_if to copy
+ *  a sequence of numbers to an output range while omitting even numbers.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  ...
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(const int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int result[2];
+ *  thrust::remove_copy_if(V, V + N, result, is_even());
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-1, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove_copy
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_if
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator remove_copy_if(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred);
+
+
+/*! \p remove_if removes from the range <tt>[first, last)</tt> every element \p x
+ *  such that <tt>pred(x)</tt> is \c true. That is, \p remove_if returns an
+ *  iterator \c new_last such that the range <tt>[first, new_last)</tt> contains
+ *  no elements for which \p pred of the corresponding stencil value is \c true. 
+ *  The iterators in the range <tt>[new_last,last)</tt> are all still dereferenceable,
+ *  but the elements that they point to are unspecified. \p remove_if is stable,
+ *  meaning that the relative order of elements that are not removed is unchanged.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred A predicate to evaluate for each element of the range
+ *              <tt>[stencil, stencil + (last - first))</tt>. Elements for which \p pred evaluates to
+ *              \c true are removed from the sequence <tt>[first, last)</tt>
+ *  \return A ForwardIterator pointing to the end of the resulting range of
+ *          elements for which \p pred evaluated to \c true.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *  \pre The range <tt>[stencil, stencil + (last - first))</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_if to remove
+ *  specific elements from an array of integers using the \p thrust::host execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  int S[N] = {0, 1, 1, 1, 0, 0};
+ *
+ *  int *new_end = thrust::remove_if(thrust::host, A, A + N, S, thrust::identity<int>());
+ *  // The first three values of A are now {1, 5, 7}
+ *  // Values beyond new_end are unspecified
+ *  \endcode
+ *
+ *  \note The range <tt>[first, last)</tt> is not permitted to overlap with the range <tt>[stencil, stencil + (last - first))</tt>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_copy_if
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred);
+
+
+/*! \p remove_if removes from the range <tt>[first, last)</tt> every element \p x
+ *  such that <tt>pred(x)</tt> is \c true. That is, \p remove_if returns an
+ *  iterator \c new_last such that the range <tt>[first, new_last)</tt> contains
+ *  no elements for which \p pred of the corresponding stencil value is \c true. 
+ *  The iterators in the range <tt>[new_last,last)</tt> are all still dereferenceable,
+ *  but the elements that they point to are unspecified. \p remove_if is stable,
+ *  meaning that the relative order of elements that are not removed is unchanged.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred A predicate to evaluate for each element of the range
+ *              <tt>[stencil, stencil + (last - first))</tt>. Elements for which \p pred evaluates to
+ *              \c true are removed from the sequence <tt>[first, last)</tt>
+ *  \return A ForwardIterator pointing to the end of the resulting range of
+ *          elements for which \p pred evaluated to \c true.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *  \pre The range <tt>[stencil, stencil + (last - first))</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_if to remove
+ *  specific elements from an array of integers.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  int S[N] = {0, 1, 1, 1, 0, 0};
+ *
+ *  int *new_end = thrust::remove_if(A, A + N, S, thrust::identity<int>());
+ *  // The first three values of A are now {1, 5, 7}
+ *  // Values beyond new_end are unspecified
+ *  \endcode
+ *
+ *  \note The range <tt>[first, last)</tt> is not permitted to overlap with the range <tt>[stencil, stencil + (last - first))</tt>.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_copy_if
+ */
+template<typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+  ForwardIterator remove_if(ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred);
+
+
+/*! \p remove_copy_if copies elements from the range <tt>[first,last)</tt> to a
+ *  range beginning at \p result, except that elements for which \p pred of the 
+ *  corresponding stencil value is \c true are not copied. The return value is 
+ *  the end of the resulting range.  This operation is stable, meaning that the
+ *  relative order of the elements that are copied is the same as the 
+ *  range <tt>[first,last)</tt>.
+ *
+ *  The algorithm's execution policy is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The resulting range is copied to the sequence beginning at this
+ *                location.
+ *  \param pred A predicate to evaluate for each element of the range <tt>[first,last)</tt>.
+ *              Elements for which \p pred evaluates to \c false are not copied
+ *              to the resulting sequence.
+ *  \return An OutputIterator pointing to the end of the resulting range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[stencil, stencil + (last - first))</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_copy_if to copy
+ *  a sequence of numbers to an output range while omitting specific elements using the \p thrust::host
+ *  execution policy for parallelization.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int S[N] = { 1, 1,  0, 1, 0, 1};
+ *  int result[2];
+ *  thrust::remove_copy_if(thrust::host, V, V + N, S, result, thrust::identity<int>());
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-1, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove_copy
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_if
+ *  \see copy_if
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred);
+
+
+/*! \p remove_copy_if copies elements from the range <tt>[first,last)</tt> to a
+ *  range beginning at \p result, except that elements for which \p pred of the 
+ *  corresponding stencil value is \c true are not copied. The return value is 
+ *  the end of the resulting range.  This operation is stable, meaning that the
+ *  relative order of the elements that are copied is the same as the 
+ *  range <tt>[first,last)</tt>.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The resulting range is copied to the sequence beginning at this
+ *                location.
+ *  \param pred A predicate to evaluate for each element of the range <tt>[first,last)</tt>.
+ *              Elements for which \p pred evaluates to \c false are not copied
+ *              to the resulting sequence.
+ *  \return An OutputIterator pointing to the end of the resulting range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertible to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The range <tt>[stencil, stencil + (last - first))</tt> shall not overlap the range <tt>[result, result + (last - first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p remove_copy_if to copy
+ *  a sequence of numbers to an output range while omitting specific elements.
+ *
+ *  \code
+ *  #include <thrust/remove.h>
+ *  ...
+ *  const int N = 6;
+ *  int V[N] = {-2, 0, -1, 0, 1, 2};
+ *  int S[N] = { 1, 1,  0, 1, 0, 1};
+ *  int result[2];
+ *  thrust::remove_copy_if(V, V + N, S, result, thrust::identity<int>());
+ *  // V remains {-2, 0, -1, 0, 1, 2}
+ *  // result is now {-1, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/remove_copy
+ *  \see remove
+ *  \see remove_copy
+ *  \see remove_if
+ *  \see copy_if
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+  OutputIterator remove_copy_if(InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred);
+
+
+/*! \} // end stream_compaction
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/remove.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/replace.h b/thirdparty/manifold/thirdparty/thrust/thrust/replace.h
new file mode 100644
index 000000000000..a5c0320c423a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/replace.h
@@ -0,0 +1,819 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file replace.h
+ *  \brief Functions for replacing elements in a range with a particular value
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup transformations
+ *  \addtogroup replacing
+ *  \ingroup transformations
+ *  \{
+ */
+
+
+/*! \p replace replaces every element in the range [first, last) equal to \p old_value
+ *  with \p new_value. That is: for every iterator \c i, if <tt>*i == old_value</tt>
+ *  then it performs the <tt>assignment *i = new_value</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence of interest.
+ *  \param last The end of the sequence of interest.
+ *  \param old_value The value to replace.
+ *  \param new_value The new value to replace \p old_value.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable>Assignable">Assignable</a>,
+ *          \p T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">EqualityComparable</a>,
+ *          objects of \p T may be compared for equality with objects of
+ *          \p ForwardIterator's \c value_type,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p replace to replace
+ *  a value of interest in a \c device_vector with another using the \p thrust::device
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] = 1;
+ *  A[1] = 2;
+ *  A[2] = 3;
+ *  A[3] = 1;
+ *
+ *  thrust::replace(thrust::device, A.begin(), A.end(), 1, 99);
+ *
+ *  // A contains [99, 2, 3, 99]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace
+ *  \see \c replace_if
+ *  \see \c replace_copy
+ *  \see \c replace_copy_if
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void replace(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+               ForwardIterator first, ForwardIterator last,
+               const T &old_value,
+               const T &new_value);
+
+
+/*! \p replace replaces every element in the range [first, last) equal to \p old_value
+ *  with \p new_value. That is: for every iterator \c i, if <tt>*i == old_value</tt>
+ *  then it performs the <tt>assignment *i = new_value</tt>.
+ *
+ *  \param first The beginning of the sequence of interest.
+ *  \param last The end of the sequence of interest.
+ *  \param old_value The value to replace.
+ *  \param new_value The new value to replace \p old_value.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable>Assignable">Assignable</a>,
+ *          \p T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">EqualityComparable</a>,
+ *          objects of \p T may be compared for equality with objects of
+ *          \p ForwardIterator's \c value_type,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p replace to replace
+ *  a value of interest in a \c device_vector with another.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] = 1;
+ *  A[1] = 2;
+ *  A[2] = 3;
+ *  A[3] = 1;
+ *
+ *  thrust::replace(A.begin(), A.end(), 1, 99);
+ *
+ *  // A contains [99, 2, 3, 99]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace
+ *  \see \c replace_if
+ *  \see \c replace_copy
+ *  \see \c replace_copy_if
+ */
+template<typename ForwardIterator, typename T>
+  void replace(ForwardIterator first, ForwardIterator last, const T &old_value,
+               const T &new_value);
+
+
+/*! \p replace_if replaces every element in the range <tt>[first, last)</tt> for which
+ *  \p pred returns \c true with \p new_value. That is: for every iterator \c i, if
+ *  <tt>pred(*i)</tt> is \c true then it performs the assignment <tt>*i = new_value</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence of interest.
+ *  \param last The end of the sequence of interest.
+ *  \param pred The predicate to test on every value of the range <tt>[first,last)</tt>.
+ *  \param new_value The new value to replace elements which <tt>pred(*i)</tt> evaluates
+ *         to \c true.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p replace_if to replace
+ *  a \c device_vector's negative elements with \c 0 using the \p thrust::device execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  1;
+ *  A[1] = -3;
+ *  A[2] =  2;
+ *  A[3] = -1;
+ *
+ *  is_less_than_zero pred;
+ *
+ *  thrust::replace_if(thrust::device, A.begin(), A.end(), pred, 0);
+ *
+ *  // A contains [1, 0, 2, 0]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace
+ *  \see \c replace
+ *  \see \c replace_copy
+ *  \see \c replace_copy_if
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  ForwardIterator first, ForwardIterator last,
+                  Predicate pred,
+                  const T &new_value);
+
+
+/*! \p replace_if replaces every element in the range <tt>[first, last)</tt> for which
+ *  \p pred returns \c true with \p new_value. That is: for every iterator \c i, if
+ *  <tt>pred(*i)</tt> is \c true then it performs the assignment <tt>*i = new_value</tt>.
+ *
+ *  \param first The beginning of the sequence of interest.
+ *  \param last The end of the sequence of interest.
+ *  \param pred The predicate to test on every value of the range <tt>[first,last)</tt>.
+ *  \param new_value The new value to replace elements which <tt>pred(*i)</tt> evaluates
+ *         to \c true.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p replace_if to replace
+ *  a \c device_vector's negative elements with \c 0.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  1;
+ *  A[1] = -3;
+ *  A[2] =  2;
+ *  A[3] = -1;
+ *
+ *  is_less_than_zero pred;
+ *
+ *  thrust::replace_if(A.begin(), A.end(), pred, 0);
+ *
+ *  // A contains [1, 0, 2, 0]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace
+ *  \see \c replace
+ *  \see \c replace_copy
+ *  \see \c replace_copy_if
+ */
+template<typename ForwardIterator, typename Predicate, typename T>
+  void replace_if(ForwardIterator first, ForwardIterator last,
+                  Predicate pred,
+                  const T &new_value);
+
+
+/*! \p replace_if replaces every element in the range <tt>[first, last)</tt> for which
+ *  <tt>pred(*s)</tt> returns \c true with \p new_value. That is: for every iterator
+ *  \c i in the range <tt>[first, last)</tt>, and \c s in the range <tt>[stencil, stencil + (last - first))</tt>,
+ *  if <tt>pred(*s)</tt> is \c true then it performs the assignment <tt>*i = new_value</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence of interest.
+ *  \param last The end of the sequence of interest.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred The predicate to test on every value of the range <tt>[first,last)</tt>.
+ *  \param new_value The new value to replace elements which <tt>pred(*i)</tt> evaluates
+ *         to \c true.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p replace_if to replace
+ *  a \c device_vector's element with \c 0 when its corresponding stencil element is less than zero
+ *  using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  10;
+ *  A[1] =  20;
+ *  A[2] =  30;
+ *  A[3] =  40;
+ *
+ *  thrust::device_vector<int> S(4);
+ *  S[0] = -1;
+ *  S[1] =  0;
+ *  S[2] = -1;
+ *  S[3] =  0;
+ *
+ *  is_less_than_zero pred;
+ *  thrust::replace_if(thrust::device, A.begin(), A.end(), S.begin(), pred, 0);
+ *
+ *  // A contains [0, 20, 0, 40]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace
+ *  \see \c replace
+ *  \see \c replace_copy
+ *  \see \c replace_copy_if
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  ForwardIterator first, ForwardIterator last,
+                  InputIterator stencil,
+                  Predicate pred,
+                  const T &new_value);
+
+
+/*! \p replace_if replaces every element in the range <tt>[first, last)</tt> for which
+ *  <tt>pred(*s)</tt> returns \c true with \p new_value. That is: for every iterator
+ *  \c i in the range <tt>[first, last)</tt>, and \c s in the range <tt>[stencil, stencil + (last - first))</tt>,
+ *  if <tt>pred(*s)</tt> is \c true then it performs the assignment <tt>*i = new_value</tt>.
+ *
+ *  \param first The beginning of the sequence of interest.
+ *  \param last The end of the sequence of interest.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param pred The predicate to test on every value of the range <tt>[first,last)</tt>.
+ *  \param new_value The new value to replace elements which <tt>pred(*i)</tt> evaluates
+ *         to \c true.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p replace_if to replace
+ *  a \c device_vector's element with \c 0 when its corresponding stencil element is less than zero.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  10;
+ *  A[1] =  20;
+ *  A[2] =  30;
+ *  A[3] =  40;
+ *
+ *  thrust::device_vector<int> S(4);
+ *  S[0] = -1;
+ *  S[1] =  0;
+ *  S[2] = -1;
+ *  S[3] =  0;
+ *
+ *  is_less_than_zero pred;
+ *  thrust::replace_if(A.begin(), A.end(), S.begin(), pred, 0);
+ *
+ *  // A contains [0, 20, 0, 40]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace
+ *  \see \c replace
+ *  \see \c replace_copy
+ *  \see \c replace_copy_if
+ */
+template<typename ForwardIterator, typename InputIterator, typename Predicate, typename T>
+  void replace_if(ForwardIterator first, ForwardIterator last,
+                  InputIterator stencil,
+                  Predicate pred,
+                  const T &new_value);
+
+
+/*! \p replace_copy copies elements from the range <tt>[first, last)</tt> to the range
+ *  <tt>[result, result + (last-first))</tt>, except that any element equal to \p old_value
+ *  is not copied; \p new_value is copied instead.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>, \p replace_copy
+ *  performs the assignment <tt>*(result+n) = new_value</tt> if <tt>*(first+n) == old_value</tt>,
+ *  and <tt>*(result+n) = *(first+n)</tt> otherwise.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to copy from.
+ *  \param last The end of the sequence to copy from.
+ *  \param result The beginning of the sequence to copy to.
+ *  \param old_value The value to replace.
+ *  \param new_value The replacement value for which <tt>*i == old_value</tt> evaluates to \c true.
+ *  \return <tt>result + (last-first)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          \p T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          \p T may be compared for equality with \p InputIterator's \c value_type,
+ *          and \p T is convertible to \p OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> A(4);
+ *  A[0] = 1;
+ *  A[1] = 2;
+ *  A[2] = 3;
+ *  A[3] = 1;
+ *
+ *  thrust::device_vector<int> B(4);
+ *
+ *  thrust::replace_copy(thrust::device, A.begin(), A.end(), B.begin(), 1, 99);
+ *
+ *  // B contains [99, 2, 3, 99]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace_copy
+ *  \see \c copy
+ *  \see \c replace
+ *  \see \c replace_if
+ *  \see \c replace_copy_if
+ */
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename T>
+__host__ __device__
+  OutputIterator replace_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              InputIterator first, InputIterator last,
+                              OutputIterator result,
+                              const T &old_value,
+                              const T &new_value);
+
+
+/*! \p replace_copy copies elements from the range <tt>[first, last)</tt> to the range
+ *  <tt>[result, result + (last-first))</tt>, except that any element equal to \p old_value
+ *  is not copied; \p new_value is copied instead.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>, \p replace_copy
+ *  performs the assignment <tt>*(result+n) = new_value</tt> if <tt>*(first+n) == old_value</tt>,
+ *  and <tt>*(result+n) = *(first+n)</tt> otherwise.
+ *
+ *  \param first The beginning of the sequence to copy from.
+ *  \param last The end of the sequence to copy from.
+ *  \param result The beginning of the sequence to copy to.
+ *  \param old_value The value to replace.
+ *  \param new_value The replacement value for which <tt>*i == old_value</tt> evaluates to \c true.
+ *  \return <tt>result + (last-first)</tt>
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          \p T is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>,
+ *          \p T may be compared for equality with \p InputIterator's \c value_type,
+ *          and \p T is convertible to \p OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> A(4);
+ *  A[0] = 1;
+ *  A[1] = 2;
+ *  A[2] = 3;
+ *  A[3] = 1;
+ *
+ *  thrust::device_vector<int> B(4);
+ *
+ *  thrust::replace_copy(A.begin(), A.end(), B.begin(), 1, 99);
+ *
+ *  // B contains [99, 2, 3, 99]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace_copy
+ *  \see \c copy
+ *  \see \c replace
+ *  \see \c replace_if
+ *  \see \c replace_copy_if
+ */
+template<typename InputIterator, typename OutputIterator, typename T>
+  OutputIterator replace_copy(InputIterator first, InputIterator last,
+                              OutputIterator result, const T &old_value,
+                              const T &new_value);
+
+
+/*! \p replace_copy_if copies elements from the range <tt>[first, last)</tt> to the range
+ *  <tt>[result, result + (last-first))</tt>, except that any element for which \p pred
+ *  is \c true is not copied; \p new_value is copied instead.
+ *
+ *  More precisely, for every integer \c n such that 0 <= n < last-first,
+ *  \p replace_copy_if performs the assignment <tt>*(result+n) = new_value</tt> if
+ *  <tt>pred(*(first+n))</tt>, and <tt>*(result+n) = *(first+n)</tt> otherwise.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to copy from.
+ *  \param last The end of the sequence to copy from.
+ *  \param result The beginning of the sequence to copy to.
+ *  \param pred The predicate to test on every value of the range <tt>[first,last)</tt>.
+ *  \param new_value The replacement value to assign <tt>pred(*i)</tt> evaluates to \c true.
+ *  \return <tt>result + (last-first)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  1;
+ *  A[1] = -3;
+ *  A[2] =  2;
+ *  A[3] = -1;
+ 
+ *  thrust::device_vector<int> B(4);
+ *  is_less_than_zero pred;
+ *
+ *  thrust::replace_copy_if(thrust::device, A.begin(), A.end(), B.begin(), pred, 0);
+ *
+ *  // B contains [1, 0, 2, 0]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace_copy
+ *  \see \c replace
+ *  \see \c replace_if
+ *  \see \c replace_copy
+ */
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                 InputIterator first, InputIterator last,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value);
+
+
+/*! \p replace_copy_if copies elements from the range <tt>[first, last)</tt> to the range
+ *  <tt>[result, result + (last-first))</tt>, except that any element for which \p pred
+ *  is \c true is not copied; \p new_value is copied instead.
+ *
+ *  More precisely, for every integer \c n such that 0 <= n < last-first,
+ *  \p replace_copy_if performs the assignment <tt>*(result+n) = new_value</tt> if
+ *  <tt>pred(*(first+n))</tt>, and <tt>*(result+n) = *(first+n)</tt> otherwise.
+ *
+ *  \param first The beginning of the sequence to copy from.
+ *  \param last The end of the sequence to copy from.
+ *  \param result The beginning of the sequence to copy to.
+ *  \param pred The predicate to test on every value of the range <tt>[first,last)</tt>.
+ *  \param new_value The replacement value to assign <tt>pred(*i)</tt> evaluates to \c true.
+ *  \return <tt>result + (last-first)</tt>
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  1;
+ *  A[1] = -3;
+ *  A[2] =  2;
+ *  A[3] = -1;
+ 
+ *  thrust::device_vector<int> B(4);
+ *  is_less_than_zero pred;
+ *
+ *  thrust::replace_copy_if(A.begin(), A.end(), B.begin(), pred, 0);
+ *
+ *  // B contains [1, 0, 2, 0]
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/replace_copy
+ *  \see \c replace
+ *  \see \c replace_if
+ *  \see \c replace_copy
+ */
+template<typename InputIterator, typename OutputIterator, typename Predicate, typename T>
+  OutputIterator replace_copy_if(InputIterator first, InputIterator last,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value);
+
+
+/*! This version of \p replace_copy_if copies elements from the range <tt>[first, last)</tt> to the range
+ *  <tt>[result, result + (last-first))</tt>, except that any element whose corresponding stencil
+ *  element causes \p pred to be \c true is not copied; \p new_value is copied instead.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>,
+ *  \p replace_copy_if performs the assignment <tt>*(result+n) = new_value</tt> if
+ *  <tt>pred(*(stencil+n))</tt>, and <tt>*(result+n) = *(first+n)</tt> otherwise.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to copy from.
+ *  \param last The end of the sequence to copy from.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the sequence to copy to.
+ *  \param pred The predicate to test on every value of the range <tt>[stencil, stencil + (last - first))</tt>.
+ *  \param new_value The replacement value to assign when <tt>pred(*s)</tt> evaluates to \c true. 
+ *  \return <tt>result + (last-first)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *  \pre \p stencil may equal \p result, but the ranges <tt>[stencil, stencil + (last - first))</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  10;
+ *  A[1] =  20;
+ *  A[2] =  30;
+ *  A[3] =  40;
+ *
+ *  thrust::device_vector<int> S(4);
+ *  S[0] = -1;
+ *  S[1] =  0;
+ *  S[2] = -1;
+ *  S[3] =  0;
+ *
+ *  thrust::device_vector<int> B(4);
+ *  is_less_than_zero pred;
+ *
+ *  thrust::replace_if(thrust::device, A.begin(), A.end(), S.begin(), B.begin(), pred, 0);
+ *
+ *  // B contains [0, 20, 0, 40]
+ *  \endcode
+ *
+ *  \see \c replace_copy
+ *  \see \c replace_if
+ */
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                 InputIterator1 first, InputIterator1 last,
+                                 InputIterator2 stencil,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value);
+
+
+/*! This version of \p replace_copy_if copies elements from the range <tt>[first, last)</tt> to the range
+ *  <tt>[result, result + (last-first))</tt>, except that any element whose corresponding stencil
+ *  element causes \p pred to be \c true is not copied; \p new_value is copied instead.
+ *
+ *  More precisely, for every integer \c n such that <tt>0 <= n < last-first</tt>,
+ *  \p replace_copy_if performs the assignment <tt>*(result+n) = new_value</tt> if
+ *  <tt>pred(*(stencil+n))</tt>, and <tt>*(result+n) = *(first+n)</tt> otherwise.
+ *
+ *  \param first The beginning of the sequence to copy from.
+ *  \param last The end of the sequence to copy from.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the sequence to copy to.
+ *  \param pred The predicate to test on every value of the range <tt>[stencil, stencil + (last - first))</tt>.
+ *  \param new_value The replacement value to assign when <tt>pred(*s)</tt> evaluates to \c true. 
+ *  \return <tt>result + (last-first)</tt>
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \p InputIterator2's \c value_type is convertible to \p Predicate's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the ranges <tt>[first, last)</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *  \pre \p stencil may equal \p result, but the ranges <tt>[stencil, stencil + (last - first))</tt> and <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  \code
+ *  #include <thrust/replace.h>
+ *  #include <thrust/device_vector.h>
+ *
+ *  struct is_less_than_zero
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x < 0;
+ *    }
+ *  };
+ *  
+ *  ...
+ *  
+ *  thrust::device_vector<int> A(4);
+ *  A[0] =  10;
+ *  A[1] =  20;
+ *  A[2] =  30;
+ *  A[3] =  40;
+ *
+ *  thrust::device_vector<int> S(4);
+ *  S[0] = -1;
+ *  S[1] =  0;
+ *  S[2] = -1;
+ *  S[3] =  0;
+ *
+ *  thrust::device_vector<int> B(4);
+ *  is_less_than_zero pred;
+ *
+ *  thrust::replace_if(A.begin(), A.end(), S.begin(), B.begin(), pred, 0);
+ *
+ *  // B contains [0, 20, 0, 40]
+ *  \endcode
+ *
+ *  \see \c replace_copy
+ *  \see \c replace_if
+ */
+template<typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate, typename T>
+  OutputIterator replace_copy_if(InputIterator1 first, InputIterator1 last,
+                                 InputIterator2 stencil,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value);
+
+
+/*! \} // end replacing
+ *  \} // transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/replace.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/reverse.h b/thirdparty/manifold/thirdparty/thrust/thrust/reverse.h
new file mode 100644
index 000000000000..056be200adca
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/reverse.h
@@ -0,0 +1,211 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file reverse.h
+ *  \brief Reverses the order of a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reordering
+ *  \ingroup algorithms
+ */
+
+
+/*! \p reverse reverses a range. That is: for every <tt>i</tt> such that
+ *  <tt>0 <= i <= (last - first) / 2</tt>, it exchanges <tt>*(first + i)</tt>
+ *  and <tt>*(last - (i + 1))</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range to reverse.
+ *  \param last The end of the range to reverse.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam BidirectionalIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/bidirectional_iterator">Bidirectional Iterator</a> and
+ *          \p BidirectionalIterator is mutable.
+ *
+ *  The following code snippet demonstrates how to use \p reverse to reverse a
+ *  \p device_vector of integers using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/reverse.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int data[N] = {0, 1, 2, 3, 4, 5};
+ *  thrust::device_vector<int> v(data, data + N);
+ *  thrust::reverse(thrust::device, v.begin(), v.end());
+ *  // v is now {5, 4, 3, 2, 1, 0}
+ *  \endcode
+ *  
+ *  \see https://en.cppreference.com/w/cpp/algorithm/reverse
+ *  \see \p reverse_copy
+ *  \see \p reverse_iterator
+ */
+template<typename DerivedPolicy, typename BidirectionalIterator>
+__host__ __device__
+  void reverse(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+               BidirectionalIterator first,
+               BidirectionalIterator last);
+
+
+/*! \p reverse reverses a range. That is: for every <tt>i</tt> such that
+ *  <tt>0 <= i <= (last - first) / 2</tt>, it exchanges <tt>*(first + i)</tt>
+ *  and <tt>*(last - (i + 1))</tt>.
+ *
+ *  \param first The beginning of the range to reverse.
+ *  \param last The end of the range to reverse.
+ *
+ *  \tparam BidirectionalIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/bidirectional_iterator">Bidirectional Iterator</a> and
+ *          \p BidirectionalIterator is mutable.
+ *
+ *  The following code snippet demonstrates how to use \p reverse to reverse a
+ *  \p device_vector of integers.
+ *
+ *  \code
+ *  #include <thrust/reverse.h>
+ *  ...
+ *  const int N = 6;
+ *  int data[N] = {0, 1, 2, 3, 4, 5};
+ *  thrust::device_vector<int> v(data, data + N);
+ *  thrust::reverse(v.begin(), v.end());
+ *  // v is now {5, 4, 3, 2, 1, 0}
+ *  \endcode
+ *  
+ *  \see https://en.cppreference.com/w/cpp/algorithm/reverse
+ *  \see \p reverse_copy
+ *  \see \p reverse_iterator
+ */
+template<typename BidirectionalIterator>
+  void reverse(BidirectionalIterator first,
+               BidirectionalIterator last);
+
+
+/*! \p reverse_copy differs from \p reverse only in that the reversed range
+ *  is written to a different output range, rather than inplace.
+ *
+ *  \p reverse_copy copies elements from the range <tt>[first, last)</tt> to the
+ *  range <tt>[result, result + (last - first))</tt> such that the copy is a 
+ *  reverse of the original range. Specifically: for every <tt>i</tt> such that
+ *  <tt>0 <= i < (last - first)</tt>, \p reverse_copy performs the assignment
+ *  <tt>*(result + (last - first) - i) = *(first + i)</tt>.
+ *
+ *  The return value is <tt>result + (last - first))</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range to reverse.
+ *  \param last The end of the range to reverse.
+ *  \param result The beginning of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam BidirectionalIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/bidirectional_iterator">Bidirectional Iterator</a>,
+ *          and \p BidirectionalIterator's \p value_type is convertible to \p OutputIterator's \p value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p reverse_copy to reverse
+ *  an input \p device_vector of integers to an output \p device_vector using the \p thrust::device
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/reverse.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int data[N] = {0, 1, 2, 3, 4, 5};
+ *  thrust::device_vector<int> input(data, data + N);
+ *  thrust::device_vector<int> output(N);
+ *  thrust::reverse_copy(thrust::device, v.begin(), v.end(), output.begin());
+ *  // input is still {0, 1, 2, 3, 4, 5}
+ *  // output is now  {5, 4, 3, 2, 1, 0}
+ *  \endcode
+ *  
+ *  \see https://en.cppreference.com/w/cpp/algorithm/reverse_copy
+ *  \see \p reverse
+ *  \see \p reverse_iterator
+ */
+template<typename DerivedPolicy, typename BidirectionalIterator, typename OutputIterator>
+__host__ __device__
+  OutputIterator reverse_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              BidirectionalIterator first,
+                              BidirectionalIterator last,
+                              OutputIterator result);
+
+
+/*! \p reverse_copy differs from \p reverse only in that the reversed range
+ *  is written to a different output range, rather than inplace.
+ *
+ *  \p reverse_copy copies elements from the range <tt>[first, last)</tt> to the
+ *  range <tt>[result, result + (last - first))</tt> such that the copy is a 
+ *  reverse of the original range. Specifically: for every <tt>i</tt> such that
+ *  <tt>0 <= i < (last - first)</tt>, \p reverse_copy performs the assignment
+ *  <tt>*(result + (last - first) - i) = *(first + i)</tt>.
+ *
+ *  The return value is <tt>result + (last - first))</tt>.
+ *
+ *  \param first The beginning of the range to reverse.
+ *  \param last The end of the range to reverse.
+ *  \param result The beginning of the output range.
+ *
+ *  \tparam BidirectionalIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/bidirectional_iterator">Bidirectional Iterator</a>,
+ *          and \p BidirectionalIterator's \p value_type is convertible to \p OutputIterator's \p value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p reverse_copy to reverse
+ *  an input \p device_vector of integers to an output \p device_vector.
+ *
+ *  \code
+ *  #include <thrust/reverse.h>
+ *  ...
+ *  const int N = 6;
+ *  int data[N] = {0, 1, 2, 3, 4, 5};
+ *  thrust::device_vector<int> input(data, data + N);
+ *  thrust::device_vector<int> output(N);
+ *  thrust::reverse_copy(v.begin(), v.end(), output.begin());
+ *  // input is still {0, 1, 2, 3, 4, 5}
+ *  // output is now  {5, 4, 3, 2, 1, 0}
+ *  \endcode
+ *  
+ *  \see https://en.cppreference.com/w/cpp/algorithm/reverse_copy
+ *  \see \p reverse
+ *  \see \p reverse_iterator
+ */
+template<typename BidirectionalIterator, typename OutputIterator>
+  OutputIterator reverse_copy(BidirectionalIterator first,
+                              BidirectionalIterator last,
+                              OutputIterator result);
+
+
+/*! \} // end reordering
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/reverse.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/scan.h
new file mode 100644
index 000000000000..9b381422323f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/scan.h
@@ -0,0 +1,1656 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file scan.h
+ *  \brief Functions for computing prefix sums
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+
+/*! \addtogroup prefixsums Prefix Sums
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p inclusive_scan computes an inclusive prefix sum operation. The
+ *  term 'inclusive' means that each result includes the corresponding
+ *  input operand in the partial sum. More precisely, <tt>*first</tt> is 
+ *  assigned to <tt>*result</tt> and the sum of <tt>*first</tt> and 
+ *  <tt>*(first + 1)</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
+ *  This version of \p inclusive_scan assumes plus as the associative operator.  
+ *  When the input and output sequences are the same, the scan is performed 
+ *  in-place.
+ *
+ *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
+ *  difference between the two functions is that \c std::partial_sum guarantees
+ *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  the binary operation to parallelize the prefix sum.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's
+ *                         \c value_type, then <tt>x + y</tt> is defined. If \c T is
+ *                         \c OutputIterator's \c value_type, then <tt>T(0)</tt> is
+ *                         defined.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan to compute an in-place
+ *  prefix sum using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::inclusive_scan(thrust::host, data, data + 6, data); // in-place scan
+ *
+ *  // data is now {1, 1, 3, 5, 6, 9}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result);
+
+
+/*! \p inclusive_scan computes an inclusive prefix sum operation. The
+ *  term 'inclusive' means that each result includes the corresponding
+ *  input operand in the partial sum. More precisely, <tt>*first</tt> is 
+ *  assigned to <tt>*result</tt> and the sum of <tt>*first</tt> and 
+ *  <tt>*(first + 1)</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
+ *  This version of \p inclusive_scan assumes plus as the associative operator.  
+ *  When the input and output sequences are the same, the scan is performed 
+ *  in-place.
+ *
+ *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
+ *  difference between the two functions is that \c std::partial_sum guarantees
+ *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  the binary operation to parallelize the prefix sum.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's
+ *                         \c value_type, then <tt>x + y</tt> is defined. If \c T is
+ *                         \c OutputIterator's \c value_type, then <tt>T(0)</tt> is
+ *                         defined.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::inclusive_scan(data, data + 6, data); // in-place scan
+ *
+ *  // data is now {1, 1, 3, 5, 6, 9}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ *
+ */
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator inclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result);
+
+
+/*! \p inclusive_scan computes an inclusive prefix sum operation. The
+ *  term 'inclusive' means that each result includes the corresponding
+ *  input operand in the partial sum.  When the input and output sequences 
+ *  are the same, the scan is performed in-place.
+ *
+ *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
+ *  difference between the two functions is that \c std::partial_sum guarantees
+ *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  the binary operation to parallelize the prefix sum.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>
+ *                         and \c OutputIterator's \c value_type is convertible to
+ *                         both \c AssociativeOperator's \c first_argument_type and
+ *                         \c second_argument_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan to compute an in-place
+ *  prefix sum using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ * 
+ *  thrust::maximum<int> binary_op;
+ *
+ *  thrust::inclusive_scan(thrust::host, data, data + 10, data, binary_op); // in-place scan
+ *
+ *  // data is now {-5, 0, 2, 2, 2, 4, 4, 4, 4, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator inclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                AssociativeOperator binary_op);
+
+
+/*! \p inclusive_scan computes an inclusive prefix sum operation. The
+ *  term 'inclusive' means that each result includes the corresponding
+ *  input operand in the partial sum.  When the input and output sequences 
+ *  are the same, the scan is performed in-place.
+ *    
+ *  \p inclusive_scan is similar to \c std::partial_sum in the STL.  The primary
+ *  difference between the two functions is that \c std::partial_sum guarantees
+ *  a serial summation order, while \p inclusive_scan requires associativity of 
+ *  the binary operation to parallelize the prefix sum.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>
+ *                         and \c OutputIterator's \c value_type is convertible to
+ *                         both \c AssociativeOperator's \c first_argument_type and
+ *                         \c second_argument_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan
+ *
+ *  \code
+ *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ * 
+ *  thrust::maximum<int> binary_op;
+ *
+ *  thrust::inclusive_scan(data, data + 10, data, binary_op); // in-place scan
+ *
+ *  // data is now {-5, 0, 2, 2, 2, 4, 4, 4, 4, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename AssociativeOperator>
+  OutputIterator inclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                AssociativeOperator binary_op);
+
+
+/*! \p exclusive_scan computes an exclusive prefix sum operation. The
+ *  term 'exclusive' means that each result does not include the 
+ *  corresponding input operand in the partial sum.  More precisely,
+ *  <tt>0</tt> is assigned to <tt>*result</tt> and the sum of 
+ *  <tt>0</tt> and <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>,
+ *  and so on. This version of \p exclusive_scan assumes plus as the 
+ *  associative operator and \c 0 as the initial value.  When the input and 
+ *  output sequences are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's
+ *                         \c value_type, then <tt>x + y</tt> is defined. If \c T is
+ *                         \c OutputIterator's \c value_type, then <tt>T(0)</tt> is
+ *                         defined.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan to compute an in-place
+ *  prefix sum using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::exclusive_scan(thrust::host, data, data + 6, data); // in-place scan
+ *
+ *  // data is now {0, 1, 1, 3, 5, 6}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result);
+
+
+/*! \p exclusive_scan computes an exclusive prefix sum operation. The
+ *  term 'exclusive' means that each result does not include the 
+ *  corresponding input operand in the partial sum.  More precisely,
+ *  <tt>0</tt> is assigned to <tt>*result</tt> and the sum of 
+ *  <tt>0</tt> and <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>,
+ *  and so on. This version of \p exclusive_scan assumes plus as the 
+ *  associative operator and \c 0 as the initial value.  When the input and 
+ *  output sequences are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's
+ *                         \c value_type, then <tt>x + y</tt> is defined. If \c T is
+ *                         \c OutputIterator's \c value_type, then <tt>T(0)</tt> is
+ *                         defined.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::exclusive_scan(data, data + 6, data); // in-place scan
+ *
+ *  // data is now {0, 1, 1, 3, 5, 6}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename InputIterator,
+         typename OutputIterator>
+  OutputIterator exclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result);
+
+
+/*! \p exclusive_scan computes an exclusive prefix sum operation. The
+ *  term 'exclusive' means that each result does not include the 
+ *  corresponding input operand in the partial sum.  More precisely,
+ *  \p init is assigned to <tt>*result</tt> and the sum of \p init and 
+ *  <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
+ *  This version of \p exclusive_scan assumes plus as the associative 
+ *  operator but requires an initial value \p init.  When the input and 
+ *  output sequences are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param init The initial value.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's
+ *                         \c value_type, then <tt>x + y</tt> is defined.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan to compute an in-place
+ *  prefix sum using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::exclusive_scan(thrust::host, data, data + 6, data, 4); // in-place scan
+ *
+ *  // data is now {4, 5, 5, 7, 9, 10}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init);
+
+
+/*! \p exclusive_scan computes an exclusive prefix sum operation. The
+ *  term 'exclusive' means that each result does not include the 
+ *  corresponding input operand in the partial sum.  More precisely,
+ *  \p init is assigned to <tt>*result</tt> and the sum of \p init and 
+ *  <tt>*first</tt> is assigned to <tt>*(result + 1)</tt>, and so on. 
+ *  This version of \p exclusive_scan assumes plus as the associative 
+ *  operator but requires an initial value \p init.  When the input and 
+ *  output sequences are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param init The initial value.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's
+ *                         \c value_type, then <tt>x + y</tt> is defined.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::exclusive_scan(data, data + 6, data, 4); // in-place scan
+ *
+ *  // data is now {4, 5, 5, 7, 9, 10}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename T>
+  OutputIterator exclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init);
+
+
+/*! \p exclusive_scan computes an exclusive prefix sum operation. The
+ *  term 'exclusive' means that each result does not include the 
+ *  corresponding input operand in the partial sum.  More precisely,
+ *  \p init is assigned to <tt>\*result</tt> and the value
+ *  <tt>binary_op(init, \*first)</tt> is assigned to <tt>\*(result + 1)</tt>,
+ *  and so on. This version of the function requires both an associative 
+ *  operator and an initial value \p init.  When the input and output
+ *  sequences are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param init The initial value.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>
+ *                         and \c OutputIterator's \c value_type is convertible to
+ *                         both \c AssociativeOperator's \c first_argument_type and
+ *                         \c second_argument_type.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan to compute an in-place
+ *  prefix sum using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ * 
+ *  thrust::maximum<int> binary_op;
+ *
+ *  thrust::exclusive_scan(thrust::host, data, data + 10, data, 1, binary_op); // in-place scan
+ *
+ *  // data is now {1, 1, 1, 2, 2, 2, 4, 4, 4, 4 }
+ *  \endcode
+ *  
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init,
+                                AssociativeOperator binary_op);
+
+
+/*! \p exclusive_scan computes an exclusive prefix sum operation. The
+ *  term 'exclusive' means that each result does not include the 
+ *  corresponding input operand in the partial sum.  More precisely,
+ *  \p init is assigned to <tt>\*result</tt> and the value
+ *  <tt>binary_op(init, \*first)</tt> is assigned to <tt>\*(result + 1)</tt>,
+ *  and so on. This version of the function requires both an associative 
+ *  operator and an initial value \p init.  When the input and output
+ *  sequences are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param init The initial value.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to
+ *                        \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>
+ *                         and \c OutputIterator's \c value_type is convertible to
+ *                         both \c AssociativeOperator's \c first_argument_type and
+ *                         \c second_argument_type.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ * 
+ *  thrust::maximum<int> binary_op;
+ *
+ *  thrust::exclusive_scan(data, data + 10, data, 1, binary_op); // in-place scan
+ *
+ *  // data is now {1, 1, 1, 2, 2, 2, 4, 4, 4, 4 }
+ *  \endcode
+ *  
+ *  \see https://en.cppreference.com/w/cpp/algorithm/partial_sum
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename T,
+         typename AssociativeOperator>
+  OutputIterator exclusive_scan(InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init,
+                                AssociativeOperator binary_op);
+
+
+/*! \addtogroup segmentedprefixsums Segmented Prefix Sums
+ *  \ingroup prefixsums
+ *  \{
+ */
+
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'inclusive' means that each result includes 
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate inclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p inclusive_scan_by_key assumes \c equal_to as the binary
+ *  predicate used to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
+ *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to
+ *  different segments otherwise.
+ *
+ *  This version of \p inclusive_scan_by_key assumes \c plus as the associative
+ *  operator used to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan_by_key using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *
+ *  thrust::inclusive_scan_by_key(thrust::host, keys, keys + 10, data, data); // in-place scan
+ *
+ *  // data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};
+ *  \endcode
+ *
+ *  \see inclusive_scan
+ *  \see exclusive_scan_by_key
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result);
+ 
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'inclusive' means that each result includes 
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate inclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p inclusive_scan_by_key assumes \c equal_to as the binary
+ *  predicate used to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
+ *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to
+ *  different segments otherwise.
+ *
+ *  This version of \p inclusive_scan_by_key assumes \c plus as the associative
+ *  operator used to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan_by_key
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  
+ *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *
+ *  thrust::inclusive_scan_by_key(keys, keys + 10, data, data); // in-place scan
+ *
+ *  // data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};
+ *  \endcode
+ *
+ *  \see inclusive_scan
+ *  \see exclusive_scan_by_key
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator inclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result);
+
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'inclusive' means that each result includes 
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate inclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  different segments otherwise.
+ *
+ *  This version of \p inclusive_scan_by_key assumes \c plus as the associative
+ *  operator used to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec. 
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param binary_pred  The binary predicate used to determine equality of keys.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan_by_key using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *
+ *  thrust::inclusive_scan_by_key(thrust::host, keys, keys + 10, data, data, binary_pred); // in-place scan
+ *
+ *  // data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};
+ *  \endcode
+ *
+ *  \see inclusive_scan
+ *  \see exclusive_scan_by_key
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred);
+
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'inclusive' means that each result includes 
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate inclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  different segments otherwise.
+ *
+ *  This version of \p inclusive_scan_by_key assumes \c plus as the associative
+ *  operator used to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param binary_pred  The binary predicate used to determine equality of keys.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan_by_key
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *
+ *  thrust::inclusive_scan_by_key(keys, keys + 10, data, data, binary_pred); // in-place scan
+ *
+ *  // data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};
+ *  \endcode
+ *
+ *  \see inclusive_scan
+ *  \see exclusive_scan_by_key
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate>
+  OutputIterator inclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred);
+
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'inclusive' means that each result includes 
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate inclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  different segments otherwise.
+ *
+ *  This version of \p inclusive_scan_by_key uses the associative operator 
+ *  \c binary_op to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param binary_pred  The binary predicate used to determine equality of keys.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan_by_key using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *  thrust::plus<int>     binary_op;
+ *
+ *  thrust::inclusive_scan_by_key(thrust::host, keys, keys + 10, data, data, binary_pred, binary_op); // in-place scan
+ *
+ *  // data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};
+ *  \endcode
+ *
+ *  \see inclusive_scan
+ *  \see exclusive_scan_by_key
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op);
+
+
+/*! \p inclusive_scan_by_key computes an inclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'inclusive' means that each result includes 
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate inclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p inclusive_scan_by_key uses the binary predicate 
+ *  \c pred to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1)</tt>
+ *  belong to the same segment if <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to 
+ *  different segments otherwise.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  This version of \p inclusive_scan_by_key uses the associative operator 
+ *  \c binary_op to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param binary_pred  The binary predicate used to determine equality of keys.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p inclusive_scan_by_key
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *  thrust::plus<int>     binary_op;
+ *
+ *  thrust::inclusive_scan_by_key(keys, keys + 10, data, data, binary_pred, binary_op); // in-place scan
+ *
+ *  // data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};
+ *  \endcode
+ *
+ *  \see inclusive_scan
+ *  \see exclusive_scan_by_key
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+  OutputIterator inclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive segmented prefix 
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c 0 to
+ *  initialize the exclusive scan operation.
+ *
+ *  This version of \p exclusive_scan_by_key assumes \c plus as the associative
+ *  operator used to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ * 
+ *  This version of \p exclusive_scan_by_key assumes \c equal_to as the binary
+ *  predicate used to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1</tt>
+ *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to 
+ *  different segments otherwise.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  Refer to the most general form of \p exclusive_scan_by_key for additional details.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  thrust::exclusive_scan_by_key(thrust::host, key, key + 10, vals, vals); // in-place scan
+ *
+ *  // vals is now {0, 1, 2, 0, 1, 0, 0, 1, 2, 3};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive segmented prefix 
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c 0 to
+ *  initialize the exclusive scan operation.
+ *
+ *  This version of \p exclusive_scan_by_key assumes \c plus as the associative
+ *  operator used to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ * 
+ *  This version of \p exclusive_scan_by_key assumes \c equal_to as the binary
+ *  predicate used to compare adjacent keys.  Specifically, consecutive iterators
+ *  <tt>i</tt> and <tt>i+1</tt> in the range <tt>[first1, last1</tt>
+ *  belong to the same segment if <tt>*i == *(i+1)</tt>, and belong to 
+ *  different segments otherwise.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  Refer to the most general form of \p exclusive_scan_by_key for additional details.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key.
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  thrust::exclusive_scan_by_key(key, key + 10, vals, vals); // in-place scan
+ *
+ *  // vals is now {0, 1, 2, 0, 1, 0, 0, 1, 2, 3};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'exclusive' means that each result does not include
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate exclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c init to
+ *  initialize the exclusive scan operation.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param init The initial of the exclusive sum value.
+ *  \return The end of the output sequence.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key using the \p
+ *  thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int init = 5;
+ *
+ *  thrust::exclusive_scan_by_key(thrust::host, key, key + 10, vals, vals, init); // in-place scan
+ *
+ *  // vals is now {5, 6, 7, 5, 6, 5, 5, 6, 7, 8};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *  \see inclusive_scan_by_key
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'exclusive' means that each result does not include
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate exclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c init to
+ *  initialize the exclusive scan operation.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param init The initial of the exclusive sum value.
+ *  \return The end of the output sequence.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int init = 5;
+ *
+ *  thrust::exclusive_scan_by_key(key, key + 10, vals, vals, init); // in-place scan
+ *
+ *  // vals is now {5, 6, 7, 5, 6, 5, 5, 6, 7, 8};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *  \see inclusive_scan_by_key
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'exclusive' means that each result does not include
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate exclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c init to
+ *  initialize the exclusive scan operation.
+ *
+ *  This version of \p exclusive_scan_by_key uses the binary predicate \c binary_pred
+ *  to compare adjacent keys.  Specifically, consecutive iterators <tt>i</tt> and
+ *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if
+ *  <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to different segments otherwise.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param init The initial of the exclusive sum value.
+ *  \param binary_pred The binary predicate used to determine equality of keys.
+ *  \return The end of the output sequence.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int init = 5;
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *
+ *  thrust::exclusive_scan_by_key(thrust::host, key, key + 10, vals, vals, init, binary_pred); // in-place scan
+ *
+ *  // vals is now {5, 6, 7, 5, 6, 5, 5, 6, 7, 8};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *  \see inclusive_scan_by_key
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'exclusive' means that each result does not include
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate exclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c init to
+ *  initialize the exclusive scan operation.
+ *
+ *  This version of \p exclusive_scan_by_key uses the binary predicate \c binary_pred
+ *  to compare adjacent keys.  Specifically, consecutive iterators <tt>i</tt> and
+ *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if
+ *  <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to different segments otherwise.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param init The initial of the exclusive sum value.
+ *  \param binary_pred The binary predicate used to determine equality of keys.
+ *  \return The end of the output sequence.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int init = 5;
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *
+ *  thrust::exclusive_scan_by_key(key, key + 10, vals, vals, init, binary_pred); // in-place scan
+ *
+ *  // vals is now {5, 6, 7, 5, 6, 5, 5, 6, 7, 8};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *  \see inclusive_scan_by_key
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'exclusive' means that each result does not include
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate exclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c init to
+ *  initialize the exclusive scan operation.
+ *
+ *  This version of \p exclusive_scan_by_key uses the binary predicate \c binary_pred
+ *  to compare adjacent keys.  Specifically, consecutive iterators <tt>i</tt> and
+ *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if 
+ *  <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to different segments otherwise.
+ *
+ *  This version of \p exclusive_scan_by_key uses the associative operator 
+ *  \c binary_op to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param init The initial of the exclusive sum value.
+ *  \param binary_pred The binary predicate used to determine equality of keys.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                         and \c AssociativeOperator's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int init = 5;
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *  thrust::plus<int>     binary_op;
+ *
+ *  thrust::exclusive_scan_by_key(thrust::host, key, key + 10, vals, vals, init, binary_pred, binary_op); // in-place scan
+ *
+ *  // vals is now {5, 6, 7, 5, 6, 5, 5, 6, 7, 8};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *  \see inclusive_scan_by_key
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op);
+
+
+/*! \p exclusive_scan_by_key computes an exclusive key-value or 'segmented' prefix 
+ *  sum operation. The term 'exclusive' means that each result does not include
+ *  the corresponding input operand in the partial sum. The term 'segmented'
+ *  means that the partial sums are broken into distinct segments.  In other
+ *  words, within each segment a separate exclusive scan operation is computed.
+ *  Refer to the code sample below for example usage.
+ *
+ *  This version of \p exclusive_scan_by_key uses the value \c init to
+ *  initialize the exclusive scan operation.
+ *
+ *  This version of \p exclusive_scan_by_key uses the binary predicate \c binary_pred
+ *  to compare adjacent keys.  Specifically, consecutive iterators <tt>i</tt> and
+ *  <tt>i+1</tt> in the range <tt>[first1, last1)</tt> belong to the same segment if 
+ *  <tt>binary_pred(*i, *(i+1))</tt> is true, and belong to different segments otherwise.
+ *
+ *  This version of \p exclusive_scan_by_key uses the associative operator 
+ *  \c binary_op to perform the prefix sum. When the input and output sequences
+ *  are the same, the scan is performed in-place.
+ *
+ *  Results are not deterministic for pseudo-associative operators (e.g.,
+ *  addition of floating-point types). Results for pseudo-associative
+ *  operators may vary from run to run.
+ *
+ *  \param first1 The beginning of the key sequence.
+ *  \param last1 The end of the key sequence.
+ *  \param first2 The beginning of the input value sequence.
+ *  \param result The beginning of the output value sequence.
+ *  \param init The initial of the exclusive sum value.
+ *  \param binary_pred The binary predicate used to determine equality of keys.
+ *  \param binary_op The associatve operator used to 'sum' values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>,
+ *                         and if \c x and \c y are objects of \c OutputIterator's \c value_type, then 
+ *                         <tt>binary_op(x,y)</tt> is defined.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                         and \c AssociativeOperator's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first1 may equal \p result but the range <tt>[first1, last1)</tt> and the range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *  \pre \p first2 may equal \p result but the range <tt>[first2, first2 + (last1 - first1)</tt> and range <tt>[result, result + (last1 - first1))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p exclusive_scan_by_key
+ *
+ *  \code
+ *  #include <thrust/scan.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};
+ *  int vals[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+ *
+ *  int init = 5;
+ *
+ *  thrust::equal_to<int> binary_pred;
+ *  thrust::plus<int>     binary_op;
+ *
+ *  thrust::exclusive_scan_by_key(key, key + 10, vals, vals, init, binary_pred, binary_op); // in-place scan
+ *
+ *  // vals is now {5, 6, 7, 5, 6, 5, 5, 6, 7, 8};
+ *  \endcode
+ *
+ *  \see exclusive_scan
+ *  \see inclusive_scan_by_key
+ *
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+  OutputIterator exclusive_scan_by_key(InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op);
+
+
+/*! \} // end segmentedprefixsums
+ */
+
+
+/*! \} // end prefix sums
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/scan.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/scatter.h b/thirdparty/manifold/thirdparty/thrust/thrust/scatter.h
new file mode 100644
index 000000000000..b8b0bd84f52a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/scatter.h
@@ -0,0 +1,419 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file scatter.h
+ *  \brief Irregular copying to a destination range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup scattering
+ *  \ingroup copying
+ *  \{
+ */
+
+
+/*! \p scatter copies elements from a source range into an output array
+ *  according to a map. For each iterator \c i in the range [\p first, \p last),
+ *  the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>. The 
+ *  output iterator must permit random access. If the same index 
+ *  appears more than once in the range <tt>[map, map + (last - first))</tt>,
+ *  the result is undefined.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first Beginning of the sequence of values to scatter.
+ *  \param last End of the sequence of values to scatter.
+ *  \param map  Beginning of the sequence of output indices.
+ *  \param result Destination of the source elements.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c value_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a>.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[first,last)` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[map,map + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The expression `result[*i]` shall be valid for all iterators in the range `[map,map + (last - first))`.
+ *
+ *  The following code snippet demonstrates how to use \p scatter to
+ *  reorder a range using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/scatter.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  // mark even indices with a 1; odd indices with a 0
+ *  int values[10] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 0};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // scatter all even indices into the first half of the
+ *  // range, and odd indices vice versa
+ *  int map[10]   = {0, 5, 1, 6, 2, 7, 3, 8, 4, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10);
+ *  thrust::scatter(thrust::device,
+ *                  d_values.begin(), d_values.end(),
+ *                  d_map.begin(), d_output.begin());
+ *  // d_output is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  \endcode
+ *
+ *  \note \p scatter is the inverse of thrust::gather.
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+               InputIterator1 first,
+               InputIterator1 last,
+               InputIterator2 map,
+               RandomAccessIterator result);
+
+
+/*! \p scatter copies elements from a source range into an output array
+ *  according to a map. For each iterator \c i in the range [\p first, \p last),
+ *  the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>. The 
+ *  output iterator must permit random access. If the same index 
+ *  appears more than once in the range <tt>[map, map + (last - first))</tt>,
+ *  the result is undefined.
+ *
+ *  \param first Beginning of the sequence of values to scatter.
+ *  \param last End of the sequence of values to scatter.
+ *  \param map  Beginning of the sequence of output indices.
+ *  \param result Destination of the source elements.
+ *
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c value_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a>.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[first,last)` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[map,map + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The expression `result[*i]` shall be valid for all iterators in the range `[map,map + (last - first))`.
+ *
+ *  The following code snippet demonstrates how to use \p scatter to
+ *  reorder a range.
+ *
+ *  \code
+ *  #include <thrust/scatter.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  // mark even indices with a 1; odd indices with a 0
+ *  int values[10] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 0};
+ *  thrust::device_vector<int> d_values(values, values + 10);
+ *
+ *  // scatter all even indices into the first half of the
+ *  // range, and odd indices vice versa
+ *  int map[10]   = {0, 5, 1, 6, 2, 7, 3, 8, 4, 9};
+ *  thrust::device_vector<int> d_map(map, map + 10);
+ *
+ *  thrust::device_vector<int> d_output(10);
+ *  thrust::scatter(d_values.begin(), d_values.end(),
+ *                  d_map.begin(), d_output.begin());
+ *  // d_output is now {1, 1, 1, 1, 1, 0, 0, 0, 0, 0}
+ *  \endcode
+ *
+ *  \note \p scatter is the inverse of thrust::gather.
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator>
+  void scatter(InputIterator1 first,
+               InputIterator1 last,
+               InputIterator2 map,
+               RandomAccessIterator result);
+
+
+/*! \p scatter_if conditionally copies elements from a source range into an 
+ *  output array according to a map. For each iterator \c i in the 
+ *  range <tt>[first, last)</tt> such that <tt>*(stencil + (i - first))</tt> is
+ *  true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
+ *  The output iterator must permit random access. If the same index 
+ *  appears more than once in the range <tt>[map, map + (last - first))</tt>
+ *  the result is undefined.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first Beginning of the sequence of values to scatter.
+ *  \param last End of the sequence of values to scatter.
+ *  \param map Beginning of the sequence of output indices.
+ *  \param stencil Beginning of the sequence of predicate values.
+ *  \param output Beginning of the destination range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c value_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator3 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator3's \c value_type must be convertible to \c bool.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a>.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[first,last)` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[map,map + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[stencil,stencil + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The expression `result[*i]` shall be valid for all iterators `i` in the range `[map,map + (last - first))` for which the following condition holds: `*(stencil + i) != false`.
+ *
+ *  \code
+ *  #include <thrust/scatter.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int V[8] = {10, 20, 30, 40, 50, 60, 70, 80};
+ *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
+ *  int S[8] = {1, 0, 1, 0, 1, 0, 1, 0};
+ *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
+ * 
+ *  thrust::scatter_if(thrust::host, V, V + 8, M, S, D);
+ * 
+ *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
+ *  \endcode
+ *
+ *  \note \p scatter_if is the inverse of thrust::gather_if.
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output);
+
+
+/*! \p scatter_if conditionally copies elements from a source range into an 
+ *  output array according to a map. For each iterator \c i in the 
+ *  range <tt>[first, last)</tt> such that <tt>*(stencil + (i - first))</tt> is
+ *  true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
+ *  The output iterator must permit random access. If the same index 
+ *  appears more than once in the range <tt>[map, map + (last - first))</tt>
+ *  the result is undefined.
+ *
+ *  \param first Beginning of the sequence of values to scatter.
+ *  \param last End of the sequence of values to scatter.
+ *  \param map Beginning of the sequence of output indices.
+ *  \param stencil Beginning of the sequence of predicate values.
+ *  \param output Beginning of the destination range.
+ *
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c value_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator3 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator3's \c value_type must be convertible to \c bool.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a>.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[first,last)` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[map,map + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[stencil,stencil + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The expression `result[*i]` shall be valid for all iterators `i` in the range `[map,map + (last - first))` for which the following condition holds: `*(stencil + i) != false`.
+ *
+ *  \code
+ *  #include <thrust/scatter.h>
+ *  ...
+ *  int V[8] = {10, 20, 30, 40, 50, 60, 70, 80};
+ *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
+ *  int S[8] = {1, 0, 1, 0, 1, 0, 1, 0};
+ *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
+ * 
+ *  thrust::scatter_if(V, V + 8, M, S, D);
+ * 
+ *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
+ *  \endcode
+ *
+ *  \note \p scatter_if is the inverse of thrust::gather_if.
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator>
+  void scatter_if(InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output);
+
+
+/*! \p scatter_if conditionally copies elements from a source range into an 
+ *  output array according to a map. For each iterator \c i in the 
+ *  range <tt>[first, last)</tt> such that <tt>pred(*(stencil + (i - first)))</tt> is
+ *  \c true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
+ *  The output iterator must permit random access. If the same index 
+ *  appears more than once in the range <tt>[map, map + (last - first))</tt>
+ *  the result is undefined.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first Beginning of the sequence of values to scatter.
+ *  \param last End of the sequence of values to scatter.
+ *  \param map Beginning of the sequence of output indices.
+ *  \param stencil Beginning of the sequence of predicate values.
+ *  \param output Beginning of the destination range.
+ *  \param pred Predicate to apply to the stencil values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c value_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator3 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator3's \c value_type must be convertible to \c Predicate's \c argument_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a>.
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[first,last)` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[map,map + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[stencil,stencil + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The expression `result[*i]` shall be valid for all iterators `i` in the range `[map,map + (last - first))` for which the following condition holds: `pred(*(stencil + i)) != false`.
+ *
+ *  \code
+ *  #include <thrust/scatter.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *
+ *  ...
+ *
+ *  int V[8] = {10, 20, 30, 40, 50, 60, 70, 80};
+ *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
+ *  int S[8] = {2, 1, 2, 1, 2, 1, 2, 1};
+ *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
+ * 
+ *  is_even pred;
+ *  thrust::scatter_if(thrust::host, V, V + 8, M, S, D, pred);
+ * 
+ *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
+ *  \endcode
+ *  
+ *  \note \p scatter_if is the inverse of thrust::gather_if.
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator,
+         typename Predicate>
+__host__ __device__
+  void scatter_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output,
+                  Predicate pred);
+                  
+
+/*! \p scatter_if conditionally copies elements from a source range into an 
+ *  output array according to a map. For each iterator \c i in the 
+ *  range <tt>[first, last)</tt> such that <tt>pred(*(stencil + (i - first)))</tt> is
+ *  \c true, the value \c *i is assigned to <tt>output[*(map + (i - first))]</tt>.
+ *  The output iterator must permit random access. If the same index 
+ *  appears more than once in the range <tt>[map, map + (last - first))</tt>
+ *  the result is undefined.
+ *
+ *  \param first Beginning of the sequence of values to scatter.
+ *  \param last End of the sequence of values to scatter.
+ *  \param map Beginning of the sequence of output indices.
+ *  \param stencil Beginning of the sequence of predicate values.
+ *  \param output Beginning of the destination range.
+ *  \param pred Predicate to apply to the stencil values.
+ *
+ *  \tparam InputIterator1 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator1's \c value_type must be convertible to \c RandomAccessIterator's \c value_type.
+ *  \tparam InputIterator2 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator2's \c value_type must be convertible to \c RandomAccessIterator's \c difference_type.
+ *  \tparam InputIterator3 must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a> and \c InputIterator3's \c value_type must be convertible to \c Predicate's \c argument_type.
+ *  \tparam RandomAccessIterator must be a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access iterator</a>.
+ *  \tparam Predicate must be a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[first,last)` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[map,map + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The iterator `result + i` shall not refer to any element referenced by any iterator `j` in the range `[stencil,stencil + (last - first))` for all iterators `i` in the range `[map,map + (last - first))`.
+ *
+ *  \pre The expression `result[*i]` shall be valid for all iterators `i` in the range `[map,map + (last - first))` for which the following condition holds: `pred(*(stencil + i)) != false`.
+ *
+ *  \code
+ *  #include <thrust/scatter.h>
+ *
+ *  struct is_even
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return (x % 2) == 0;
+ *    }
+ *  };
+ *
+ *  ...
+ *
+ *  int V[8] = {10, 20, 30, 40, 50, 60, 70, 80};
+ *  int M[8] = {0, 5, 1, 6, 2, 7, 3, 4};
+ *  int S[8] = {2, 1, 2, 1, 2, 1, 2, 1};
+ *  int D[8] = {0, 0, 0, 0, 0, 0, 0, 0};
+ * 
+ *  is_even pred;
+ *  thrust::scatter_if(V, V + 8, M, S, D, pred);
+ * 
+ *  // D contains [10, 30, 50, 70, 0, 0, 0, 0];
+ *  \endcode
+ *  
+ *  \note \p scatter_if is the inverse of thrust::gather_if.
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator,
+         typename Predicate>
+  void scatter_if(InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output,
+                  Predicate pred);
+
+
+/*! \} // end scattering
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/scatter.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/sequence.h
new file mode 100644
index 000000000000..fb3959e3cabe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/sequence.h
@@ -0,0 +1,293 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file sequence.h
+ *  \brief Fills a range with a sequence of numbers
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup transformations
+ *  \{
+ */
+
+
+/*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  \p sequence performs the assignment <tt>*i =  (i - first)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *
+ *  The following code snippet demonstrates how to use \p sequence to fill a range
+ *  with a sequence of numbers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::sequence(thrust::host, A, A + 10);
+ *  // A is now {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}
+ *  \endcode
+ *
+ *  \note Unlike the similar C++ STL function \c std::iota, \p sequence offers no
+ *        guarantee on order of execution.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/iota
+ */
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  void sequence(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last);
+
+
+/*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  \p sequence performs the assignment <tt>*i =  (i - first)</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *
+ *  The following code snippet demonstrates how to use \p sequence to fill a range
+ *  with a sequence of numbers.
+ *
+ *  \code
+ *  #include <thrust/sequence.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::sequence(A, A + 10);
+ *  // A is now {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}
+ *  \endcode
+ *
+ *  \note Unlike the similar C++ STL function \c std::iota, \p sequence offers no
+ *        guarantee on order of execution.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/iota
+ */
+template<typename ForwardIterator>
+  void sequence(ForwardIterator first,
+                ForwardIterator last);
+
+
+/*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  \p sequence performs the assignment <tt>*i =  init + (i - first)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param init The first value of the sequence of numbers.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p sequence to fill a range
+ *  with a sequence of numbers starting from the value 1 using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::sequence(thrust::host, A, A + 10, 1);
+ *  // A is now {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  \endcode
+ *
+ *  \note Unlike the similar C++ STL function \c std::iota, \p sequence offers no
+ *        guarantee on order of execution.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/iota
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init);
+
+
+/*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  \p sequence performs the assignment <tt>*i =  init + (i - first)</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param init The first value of the sequence of numbers.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p sequence to fill a range
+ *  with a sequence of numbers starting from the value 1.
+ *
+ *  \code
+ *  #include <thrust/sequence.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::sequence(A, A + 10, 1);
+ *  // A is now {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+ *  \endcode
+ *
+ *  \note Unlike the similar C++ STL function \c std::iota, \p sequence offers no
+ *        guarantee on order of execution.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/iota
+ */
+template<typename ForwardIterator, typename T>
+  void sequence(ForwardIterator first,
+                ForwardIterator last,
+                T init);
+
+
+/*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  \p sequence performs the assignment <tt>*i =  init + step * (i - first)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param init The first value of the sequence of numbers
+ *  \param step The difference between consecutive elements.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p sequence to fill a range
+ *  with a sequence of numbers starting from the value 1 with a step size of 3 using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sequence.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::sequence(thrust::host, A, A + 10, 1, 3);
+ *  // A is now {1, 4, 7, 10, 13, 16, 19, 22, 25, 28}
+ *  \endcode
+ *
+ *  \note Unlike the similar C++ STL function \c std::iota, \p sequence offers no
+ *        guarantee on order of execution.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/iota
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init,
+                T step);
+
+
+/*! \p sequence fills the range <tt>[first, last)</tt> with a sequence of numbers.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, this version of 
+ *  \p sequence performs the assignment <tt>*i =  init + step * (i - first)</tt>.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param init The first value of the sequence of numbers
+ *  \param step The difference between consecutive elements.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *  \tparam T is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and \p T is convertible to \p ForwardIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p sequence to fill a range
+ *  with a sequence of numbers starting from the value 1 with a step size of 3.
+ *
+ *  \code
+ *  #include <thrust/sequence.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::sequence(A, A + 10, 1, 3);
+ *  // A is now {1, 4, 7, 10, 13, 16, 19, 22, 25, 28}
+ *  \endcode
+ *
+ *  \note Unlike the similar C++ STL function \c std::iota, \p sequence offers no
+ *        guarantee on order of execution.
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/iota
+ */
+template<typename ForwardIterator, typename T>
+  void sequence(ForwardIterator first,
+                ForwardIterator last,
+                T init,
+                T step);
+
+
+/*! \} // end transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/sequence.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/set_operations.h b/thirdparty/manifold/thirdparty/thrust/thrust/set_operations.h
new file mode 100644
index 000000000000..65a48d1b6612
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/set_operations.h
@@ -0,0 +1,2959 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file set_operations.h
+ *  \brief Set theoretic operations for sorted ranges
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup set_operations Set Operations
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p set_difference constructs a sorted range that is the set difference of the sorted
+ *  ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_difference performs the "difference" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt> and not contained in <tt>[first2, last1)</tt>. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[first1, last1)</tt> range shall be copied to the output range.
+ *
+ *  This version of \p set_difference compares elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference to compute the
+ *  set difference of two sets of integers sorted in ascending order using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {0, 1, 3, 4, 5, 6, 9};
+ *  int A2[5] = {1, 3, 5, 7, 9};
+ *
+ *  int result[3];
+ *
+ *  int *result_end = thrust::set_difference(thrust::host, A1, A1 + 7, A2, A2 + 5, result);
+ *  // result is now {0, 4, 6}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_difference
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator set_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator1                                              first1,
+                                InputIterator1                                              last1,
+                                InputIterator2                                              first2,
+                                InputIterator2                                              last2,
+                                OutputIterator                                              result);
+
+
+/*! \p set_difference constructs a sorted range that is the set difference of the sorted
+ *  ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_difference performs the "difference" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt> and not contained in <tt>[first2, last1)</tt>. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[first1, last1)</tt> range shall be copied to the output range.
+ *
+ *  This version of \p set_difference compares elements using \c operator<.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference to compute the
+ *  set difference of two sets of integers sorted in ascending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A1[7] = {0, 1, 3, 4, 5, 6, 9};
+ *  int A2[5] = {1, 3, 5, 7, 9};
+ *
+ *  int result[3];
+ *
+ *  int *result_end = thrust::set_difference(A1, A1 + 7, A2, A2 + 5, result);
+ *  // result is now {0, 4, 6}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_difference
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_difference(InputIterator1 first1,
+                                InputIterator1 last1,
+                                InputIterator2 first2,
+                                InputIterator2 last2,
+                                OutputIterator result);
+
+
+/*! \p set_difference constructs a sorted range that is the set difference of the sorted
+ *  ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_difference performs the "difference" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt> and not contained in <tt>[first2, last1)</tt>. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[first1, last1)</tt> range shall be copied to the output range.
+ *
+ *  This version of \p set_difference compares elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference to compute the
+ *  set difference of two sets of integers sorted in descending order using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {9, 6, 5, 4, 3, 1, 0};
+ *  int A2[5] = {9, 7, 5, 3, 1};
+ *
+ *  int result[3];
+ *
+ *  int *result_end = thrust::set_difference(thrust::host, A1, A1 + 7, A2, A2 + 5, result, thrust::greater<int>());
+ *  // result is now {6, 4, 0}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_difference
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+  OutputIterator set_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                InputIterator1                                              first1,
+                                InputIterator1                                              last1,
+                                InputIterator2                                              first2,
+                                InputIterator2                                              last2,
+                                OutputIterator                                              result,
+                                StrictWeakCompare                                           comp);
+
+
+/*! \p set_difference constructs a sorted range that is the set difference of the sorted
+ *  ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_difference performs the "difference" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt> and not contained in <tt>[first2, last1)</tt>. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[first1, last1)</tt> range shall be copied to the output range.
+ *
+ *  This version of \p set_difference compares elements using a function object \p comp.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference to compute the
+ *  set difference of two sets of integers sorted in descending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A1[7] = {9, 6, 5, 4, 3, 1, 0};
+ *  int A2[5] = {9, 7, 5, 3, 1};
+ *
+ *  int result[3];
+ *
+ *  int *result_end = thrust::set_difference(A1, A1 + 7, A2, A2 + 5, result, thrust::greater<int>());
+ *  // result is now {6, 4, 0}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_difference
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+  OutputIterator set_difference(InputIterator1 first1,
+                                InputIterator1 last1,
+                                InputIterator2 first2,
+                                InputIterator2 last2,
+                                OutputIterator result,
+                                StrictWeakCompare comp);
+
+
+/*! \p set_intersection constructs a sorted range that is the
+ *  intersection of sorted ranges <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt>. The return value is the end of the
+ *  output range.
+ *
+ *  In the simplest case, \p set_intersection performs the
+ *  "intersection" operation from set theory: the output range
+ *  contains a copy of every element that is contained in both
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The
+ *  general case is more complicated, because the input ranges may
+ *  contain duplicate elements. The generalization is that if a value
+ *  appears \c m times in <tt>[first1, last1)</tt> and \c n times in
+ *  <tt>[first2, last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the output range.
+ *  \p set_intersection is stable, meaning that both elements are
+ *  copied from the first range rather than the second, and that the
+ *  relative order of elements in the output range is the same as in
+ *  the first input range.
+ *
+ *  This version of \p set_intersection compares objects using
+ *  \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection to compute the
+ *  set intersection of two sets of integers sorted in ascending order using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[6] = {1, 3, 5, 7, 9, 11};
+ *  int A2[7] = {1, 1, 2, 3, 5,  8, 13};
+ *
+ *  int result[7];
+ *
+ *  int *result_end = thrust::set_intersection(thrust::host, A1, A1 + 6, A2, A2 + 7, result);
+ *  // result is now {1, 3, 5}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_intersection
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator set_intersection(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  InputIterator1                                              first1,
+                                  InputIterator1                                              last1,
+                                  InputIterator2                                              first2,
+                                  InputIterator2                                              last2,
+                                  OutputIterator                                              result);
+
+
+/*! \p set_intersection constructs a sorted range that is the
+ *  intersection of sorted ranges <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt>. The return value is the end of the
+ *  output range.
+ *
+ *  In the simplest case, \p set_intersection performs the
+ *  "intersection" operation from set theory: the output range
+ *  contains a copy of every element that is contained in both
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The
+ *  general case is more complicated, because the input ranges may
+ *  contain duplicate elements. The generalization is that if a value
+ *  appears \c m times in <tt>[first1, last1)</tt> and \c n times in
+ *  <tt>[first2, last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the output range.
+ *  \p set_intersection is stable, meaning that both elements are
+ *  copied from the first range rather than the second, and that the
+ *  relative order of elements in the output range is the same as in
+ *  the first input range.
+ *
+ *  This version of \p set_intersection compares objects using
+ *  \c operator<.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection to compute the
+ *  set intersection of two sets of integers sorted in ascending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A1[6] = {1, 3, 5, 7, 9, 11};
+ *  int A2[7] = {1, 1, 2, 3, 5,  8, 13};
+ *
+ *  int result[7];
+ *
+ *  int *result_end = thrust::set_intersection(A1, A1 + 6, A2, A2 + 7, result);
+ *  // result is now {1, 3, 5}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_intersection
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_intersection(InputIterator1 first1,
+                                  InputIterator1 last1,
+                                  InputIterator2 first2,
+                                  InputIterator2 last2,
+                                  OutputIterator result);
+
+
+/*! \p set_intersection constructs a sorted range that is the
+ *  intersection of sorted ranges <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt>. The return value is the end of the
+ *  output range.
+ *
+ *  In the simplest case, \p set_intersection performs the
+ *  "intersection" operation from set theory: the output range
+ *  contains a copy of every element that is contained in both
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The
+ *  general case is more complicated, because the input ranges may
+ *  contain duplicate elements. The generalization is that if a value
+ *  appears \c m times in <tt>[first1, last1)</tt> and \c n times in
+ *  <tt>[first2, last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the output range.
+ *  \p set_intersection is stable, meaning that both elements are
+ *  copied from the first range rather than the second, and that the
+ *  relative order of elements in the output range is the same as in
+ *  the first input range.
+ *
+ *  This version of \p set_intersection compares elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection to compute
+ *  the set intersection of sets of integers sorted in descending order using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[6] = {11, 9, 7, 5, 3, 1};
+ *  int A2[7] = {13, 8, 5, 3, 2,  1, 1};
+ *
+ *  int result[3];
+ *
+ *  int *result_end = thrust::set_intersection(thrust::host, A1, A1 + 6, A2, A2 + 7, result, thrust::greater<int>());
+ *  // result is now {5, 3, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_intersection
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+  OutputIterator set_intersection(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  InputIterator1                                              first1,
+                                  InputIterator1                                              last1,
+                                  InputIterator2                                              first2,
+                                  InputIterator2                                              last2,
+                                  OutputIterator                                              result,
+                                  StrictWeakCompare                                           comp);
+
+
+/*! \p set_intersection constructs a sorted range that is the
+ *  intersection of sorted ranges <tt>[first1, last1)</tt> and
+ *  <tt>[first2, last2)</tt>. The return value is the end of the
+ *  output range.
+ *
+ *  In the simplest case, \p set_intersection performs the
+ *  "intersection" operation from set theory: the output range
+ *  contains a copy of every element that is contained in both
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The
+ *  general case is more complicated, because the input ranges may
+ *  contain duplicate elements. The generalization is that if a value
+ *  appears \c m times in <tt>[first1, last1)</tt> and \c n times in
+ *  <tt>[first2, last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the output range.
+ *  \p set_intersection is stable, meaning that both elements are
+ *  copied from the first range rather than the second, and that the
+ *  relative order of elements in the output range is the same as in
+ *  the first input range.
+ *
+ *  This version of \p set_intersection compares elements using a function object \p comp.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection to compute
+ *  the set intersection of sets of integers sorted in descending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A1[6] = {11, 9, 7, 5, 3, 1};
+ *  int A2[7] = {13, 8, 5, 3, 2,  1, 1};
+ *
+ *  int result[3];
+ *
+ *  int *result_end = thrust::set_intersection(A1, A1 + 6, A2, A2 + 7, result, thrust::greater<int>());
+ *  // result is now {5, 3, 1}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_intersection
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+  OutputIterator set_intersection(InputIterator1 first1,
+                                  InputIterator1 last1,
+                                  InputIterator2 first2,
+                                  InputIterator2 last2,
+                                  OutputIterator result,
+                                  StrictWeakCompare comp);
+
+
+/*! \p set_symmetric_difference constructs a sorted range that is the set symmetric
+ *  difference of the sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>.
+ *  The return value is the end of the output range.
+ *
+ *  In the simplest case, \p set_symmetric_difference performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[first1, last1)</tt> but not <tt>[first2, last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[first2, last2)</tt> but not <tt>[first1, last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[first2, last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[first1, last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[first2, last2)</tt> if <tt>m < n</tt>.
+ *
+ *  This version of \p set_union compares elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference to compute
+ *  the symmetric difference of two sets of integers sorted in ascending order using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {0, 1, 2, 2, 4, 6, 7};
+ *  int A2[5] = {1, 1, 2, 5, 8};
+ *
+ *  int result[6];
+ *
+ *  int *result_end = thrust::set_symmetric_difference(thrust::host, A1, A1 + 7, A2, A2 + 5, result);
+ *  // result = {0, 4, 5, 6, 7, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_symmetric_difference
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_difference
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator set_symmetric_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                          InputIterator1                                              first1,
+                                          InputIterator1                                              last1,
+                                          InputIterator2                                              first2,
+                                          InputIterator2                                              last2,
+                                          OutputIterator                                              result);
+
+
+/*! \p set_symmetric_difference constructs a sorted range that is the set symmetric
+ *  difference of the sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>.
+ *  The return value is the end of the output range.
+ *
+ *  In the simplest case, \p set_symmetric_difference performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[first1, last1)</tt> but not <tt>[first2, last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[first2, last2)</tt> but not <tt>[first1, last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[first2, last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[first1, last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[first2, last2)</tt> if <tt>m < n</tt>.
+ *
+ *  This version of \p set_union compares elements using \c operator<.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference to compute
+ *  the symmetric difference of two sets of integers sorted in ascending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A1[7] = {0, 1, 2, 2, 4, 6, 7};
+ *  int A2[5] = {1, 1, 2, 5, 8};
+ *
+ *  int result[6];
+ *
+ *  int *result_end = thrust::set_symmetric_difference(A1, A1 + 7, A2, A2 + 5, result);
+ *  // result = {0, 4, 5, 6, 7, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_symmetric_difference
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_difference
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_symmetric_difference(InputIterator1 first1,
+                                          InputIterator1 last1,
+                                          InputIterator2 first2,
+                                          InputIterator2 last2,
+                                          OutputIterator result);
+
+
+/*! \p set_symmetric_difference constructs a sorted range that is the set symmetric
+ *  difference of the sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>.
+ *  The return value is the end of the output range.
+ *
+ *  In the simplest case, \p set_symmetric_difference performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[first1, last1)</tt> but not <tt>[first2, last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[first2, last2)</tt> but not <tt>[first1, last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[first2, last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[first1, last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[first2, last2)</tt> if <tt>m < n</tt>.
+ *
+ *  This version of \p set_union compares elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference to compute
+ *  the symmetric difference of two sets of integers sorted in descending order using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {7, 6, 4, 2, 2, 1, 0};
+ *  int A2[5] = {8, 5, 2, 1, 1};
+ *
+ *  int result[6];
+ *
+ *  int *result_end = thrust::set_symmetric_difference(thrust::host, A1, A1 + 7, A2, A2 + 5, result);
+ *  // result = {8, 7, 6, 5, 4, 0}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_symmetric_difference
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_difference
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+  OutputIterator set_symmetric_difference(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                          InputIterator1                                              first1,
+                                          InputIterator1                                              last1,
+                                          InputIterator2                                              first2,
+                                          InputIterator2                                              last2,
+                                          OutputIterator                                              result,
+                                          StrictWeakCompare                                           comp);
+
+
+/*! \p set_symmetric_difference constructs a sorted range that is the set symmetric
+ *  difference of the sorted ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>.
+ *  The return value is the end of the output range.
+ *
+ *  In the simplest case, \p set_symmetric_difference performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[first1, last1)</tt> but not <tt>[first2, last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[first2, last2)</tt> but not <tt>[first1, last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[first2, last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[first1, last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[first2, last2)</tt> if <tt>m < n</tt>.
+ *
+ *  This version of \p set_union compares elements using a function object \p comp.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference to compute
+ *  the symmetric difference of two sets of integers sorted in descending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A1[7] = {7, 6, 4, 2, 2, 1, 0};
+ *  int A2[5] = {8, 5, 2, 1, 1};
+ *
+ *  int result[6];
+ *
+ *  int *result_end = thrust::set_symmetric_difference(A1, A1 + 7, A2, A2 + 5, result);
+ *  // result = {8, 7, 6, 5, 4, 0}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_symmetric_difference
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_difference
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+  OutputIterator set_symmetric_difference(InputIterator1 first1,
+                                          InputIterator1 last1,
+                                          InputIterator2 first2,
+                                          InputIterator2 last2,
+                                          OutputIterator result,
+                                          StrictWeakCompare comp);
+
+
+/*! \p set_union constructs a sorted range that is the union of the sorted ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_union performs the "union" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt>, <tt>[first2, last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  This version of \p set_union compares elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_union to compute the union of
+ *  two sets of integers sorted in ascending order using the \p thrust::host execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {0, 2, 4, 6, 8, 10, 12};
+ *  int A2[5] = {1, 3, 5, 7, 9};
+ *
+ *  int result[11];
+ *
+ *  int *result_end = thrust::set_union(thrust::host, A1, A1 + 7, A2, A2 + 5, result);
+ *  // result = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_union
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator set_union(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1                                              first1,
+                           InputIterator1                                              last1,
+                           InputIterator2                                              first2,
+                           InputIterator2                                              last2,
+                           OutputIterator                                              result);
+
+
+/*! \p set_union constructs a sorted range that is the union of the sorted ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_union performs the "union" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt>, <tt>[first2, last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  This version of \p set_union compares elements using \c operator<.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_union to compute the union of
+ *  two sets of integers sorted in ascending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A1[7] = {0, 2, 4, 6, 8, 10, 12};
+ *  int A2[5] = {1, 3, 5, 7, 9};
+ *
+ *  int result[11];
+ *
+ *  int *result_end = thrust::set_union(A1, A1 + 7, A2, A2 + 5, result);
+ *  // result = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_union
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+  OutputIterator set_union(InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           InputIterator2 last2,
+                           OutputIterator result);
+
+
+/*! \p set_union constructs a sorted range that is the union of the sorted ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_union performs the "union" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt>, <tt>[first2, last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  This version of \p set_union compares elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_union to compute the union of
+ *  two sets of integers sorted in ascending order using the \p thrust::host execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A1[7] = {12, 10, 8, 6, 4, 2, 0};
+ *  int A2[5] = {9, 7, 5, 3, 1};
+ *
+ *  int result[11];
+ *
+ *  int *result_end = thrust::set_union(thrust::host, A1, A1 + 7, A2, A2 + 5, result, thrust::greater<int>());
+ *  // result = {12, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_union
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+__host__ __device__
+  OutputIterator set_union(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1                                              first1,
+                           InputIterator1                                              last1,
+                           InputIterator2                                              first2,
+                           InputIterator2                                              last2,
+                           OutputIterator                                              result,
+                           StrictWeakCompare                                           comp);
+
+
+/*! \p set_union constructs a sorted range that is the union of the sorted ranges
+ *  <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt>. The return value is the
+ *  end of the output range.
+ *
+ *  In the simplest case, \p set_union performs the "union" operation from set
+ *  theory: the output range contains a copy of every element that is contained in
+ *  <tt>[first1, last1)</tt>, <tt>[first2, last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[first1, last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[first2, last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  This version of \p set_union compares elements using a function object \p comp.
+ *
+ *  \param first1 The beginning of the first input range.
+ *  \param last1 The end of the first input range.
+ *  \param first2 The beginning of the second input range.
+ *  \param last2 The end of the second input range.
+ *  \param result The beginning of the output range.
+ *  \param comp Comparison operator.
+ *  \return The end of the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1's \c value_type is convertable to \p StrictWeakCompare's \c first_argument_type.
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2's \c value_type is convertable to \p StrictWeakCompare's \c second_argument_type.
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[first1, last1)</tt> and <tt>[first2, last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting range shall not overlap with either input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_union to compute the union of
+ *  two sets of integers sorted in ascending order.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A1[7] = {12, 10, 8, 6, 4, 2, 0};
+ *  int A2[5] = {9, 7, 5, 3, 1};
+ *
+ *  int result[11];
+ *
+ *  int *result_end = thrust::set_union(A1, A1 + 7, A2, A2 + 5, result, thrust::greater<int>());
+ *  // result = {12, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/set_union
+ *  \see \p merge
+ *  \see \p includes
+ *  \see \p set_union
+ *  \see \p set_intersection
+ *  \see \p set_symmetric_difference
+ *  \see \p sort
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakCompare>
+  OutputIterator set_union(InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           InputIterator2 last2,
+                           OutputIterator result,
+                           StrictWeakCompare comp);
+
+
+/*! \p set_difference_by_key performs a key-value difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_difference_by_key performs the "difference" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt> and not contained in <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[keys_first1, keys_last1)</tt> range shall be copied to the output range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_difference_by_key compares key elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference_by_key to compute the
+ *  set difference of two sets of integers sorted in ascending order with their values using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {0, 1, 3, 4, 5, 6, 9};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {1, 3, 5, 7, 9};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[3];
+ *  int vals_result[3];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_difference_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {0, 4, 6}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator1                                              keys_first1,
+                          InputIterator1                                              keys_last1,
+                          InputIterator2                                              keys_first2,
+                          InputIterator2                                              keys_last2,
+                          InputIterator3                                              values_first1,
+                          InputIterator4                                              values_first2,
+                          OutputIterator1                                             keys_result,
+                          OutputIterator2                                             values_result);
+
+
+/*! \p set_difference_by_key performs a key-value difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_difference_by_key performs the "difference" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt> and not contained in <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[keys_first1, keys_last1)</tt> range shall be copied to the output range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_difference_by_key compares key elements using \c operator<.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference_by_key to compute the
+ *  set difference of two sets of integers sorted in ascending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A_keys[6] = {0, 1, 3, 4, 5, 6, 9};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {1, 3, 5, 7, 9};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[3];
+ *  int vals_result[3];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_difference_by_key(A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {0, 4, 6}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_difference_by_key(InputIterator1                             keys_first1,
+                          InputIterator1                             keys_last1,
+                          InputIterator2                             keys_first2,
+                          InputIterator2                             keys_last2,
+                          InputIterator3                             values_first1,
+                          InputIterator4                             values_first2,
+                          OutputIterator1                            keys_result,
+                          OutputIterator2                            values_result);
+
+
+/*! \p set_difference_by_key performs a key-value difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_difference_by_key performs the "difference" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt> and not contained in <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[keys_first1, keys_last1)</tt> range shall be copied to the output range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_difference_by_key compares key elements using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference_by_key to compute the
+ *  set difference of two sets of integers sorted in descending order with their values using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {9, 6, 5, 4, 3, 1, 0};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {9, 7, 5, 3, 1};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[3];
+ *  int vals_result[3];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_difference_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result, thrust::greater<int>());
+ *  // keys_result is now {0, 4, 6}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          InputIterator1                                              keys_first1,
+                          InputIterator1                                              keys_last1,
+                          InputIterator2                                              keys_first2,
+                          InputIterator2                                              keys_last2,
+                          InputIterator3                                              values_first1,
+                          InputIterator4                                              values_first2,
+                          OutputIterator1                                             keys_result,
+                          OutputIterator2                                             values_result,
+                          StrictWeakCompare                                           comp);
+
+
+/*! \p set_difference_by_key performs a key-value difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_difference_by_key performs the "difference" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt> and not contained in <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, the last <tt>max(m-n,0)</tt> elements from
+ *  <tt>[keys_first1, keys_last1)</tt> range shall be copied to the output range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_difference_by_key compares key elements using a function object \p comp.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_difference_by_key to compute the
+ *  set difference of two sets of integers sorted in descending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A_keys[6] = {9, 6, 5, 4, 3, 1, 0};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {9, 7, 5, 3, 1};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[3];
+ *  int vals_result[3];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_difference_by_key(A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result, thrust::greater<int>());
+ *  // keys_result is now {0, 4, 6}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_difference_by_key(InputIterator1                             keys_first1,
+                          InputIterator1                             keys_last1,
+                          InputIterator2                             keys_first2,
+                          InputIterator2                             keys_last2,
+                          InputIterator3                             values_first1,
+                          InputIterator4                             values_first2,
+                          OutputIterator1                            keys_result,
+                          OutputIterator2                            values_result,
+                          StrictWeakCompare                          comp);
+
+
+/*! \p set_intersection_by_key performs a key-value intersection operation from set theory.
+ *  \p set_intersection_by_key constructs a sorted range that is the intersection of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_intersection_by_key performs the "intersection" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in both
+ *  <tt>[keys_first1, keys_last1)</tt> <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if an element appears \c m times in <tt>[keys_first1, keys_last1)</tt>
+ *  and \c n times in <tt>[keys_first2, keys_last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the keys output range.
+ *  \p set_intersection_by_key is stable, meaning both that elements are copied from the first
+ *  input range rather than the second, and that the relative order of elements in the output range
+ *  is the same as the first input range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> to the keys output range,
+ *  the corresponding value element is copied from <tt>[values_first1, values_last1)</tt> to the values
+ *  output range.
+ *
+ *  This version of \p set_intersection_by_key compares objects using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \note Unlike the other key-value set operations, \p set_intersection_by_key is unique in that it has no
+ *        \c values_first2 parameter because elements from the second input range are never copied to the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection_by_key to compute the
+ *  set intersection of two sets of integers sorted in ascending order with their values using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {1, 3, 5, 7, 9, 11};
+ *  int A_vals[6] = {0, 0, 0, 0, 0,  0};
+ *  
+ *  int B_keys[7] = {1, 1, 2, 3, 5,  8, 13};
+ *
+ *  int keys_result[7];
+ *  int vals_result[7];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_intersection_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 7, A_vals, keys_result, vals_result);
+ *
+ *  // keys_result is now {1, 3, 5}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_intersection_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            InputIterator1                                              keys_first1,
+                            InputIterator1                                              keys_last1,
+                            InputIterator2                                              keys_first2,
+                            InputIterator2                                              keys_last2,
+                            InputIterator3                                              values_first1,
+                            OutputIterator1                                             keys_result,
+                            OutputIterator2                                             values_result);
+
+
+/*! \p set_intersection_by_key performs a key-value intersection operation from set theory.
+ *  \p set_intersection_by_key constructs a sorted range that is the intersection of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_intersection_by_key performs the "intersection" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in both
+ *  <tt>[keys_first1, keys_last1)</tt> <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if an element appears \c m times in <tt>[keys_first1, keys_last1)</tt>
+ *  and \c n times in <tt>[keys_first2, keys_last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the keys output range.
+ *  \p set_intersection_by_key is stable, meaning both that elements are copied from the first
+ *  input range rather than the second, and that the relative order of elements in the output range
+ *  is the same as the first input range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> to the keys output range,
+ *  the corresponding value element is copied from <tt>[values_first1, values_last1)</tt> to the values
+ *  output range.
+ *
+ *  This version of \p set_intersection_by_key compares objects using \c operator<.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \note Unlike the other key-value set operations, \p set_intersection_by_key is unique in that it has no
+ *        \c values_first2 parameter because elements from the second input range are never copied to the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection_by_key to compute the
+ *  set intersection of two sets of integers sorted in ascending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A_keys[6] = {1, 3, 5, 7, 9, 11};
+ *  int A_vals[6] = {0, 0, 0, 0, 0,  0};
+ *  
+ *  int B_keys[7] = {1, 1, 2, 3, 5,  8, 13};
+ *
+ *  int keys_result[7];
+ *  int vals_result[7];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_intersection_by_key(A_keys, A_keys + 6, B_keys, B_keys + 7, A_vals, keys_result, vals_result);
+ *
+ *  // keys_result is now {1, 3, 5}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_intersection_by_key(InputIterator1                             keys_first1,
+                            InputIterator1                             keys_last1,
+                            InputIterator2                             keys_first2,
+                            InputIterator2                             keys_last2,
+                            InputIterator3                             values_first1,
+                            OutputIterator1                            keys_result,
+                            OutputIterator2                            values_result);
+
+
+/*! \p set_intersection_by_key performs a key-value intersection operation from set theory.
+ *  \p set_intersection_by_key constructs a sorted range that is the intersection of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_intersection_by_key performs the "intersection" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in both
+ *  <tt>[keys_first1, keys_last1)</tt> <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if an element appears \c m times in <tt>[keys_first1, keys_last1)</tt>
+ *  and \c n times in <tt>[keys_first2, keys_last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the keys output range.
+ *  \p set_intersection_by_key is stable, meaning both that elements are copied from the first
+ *  input range rather than the second, and that the relative order of elements in the output range
+ *  is the same as the first input range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> to the keys output range,
+ *  the corresponding value element is copied from <tt>[values_first1, values_last1)</tt> to the values
+ *  output range.
+ *
+ *  This version of \p set_intersection_by_key compares objects using a function object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \note Unlike the other key-value set operations, \p set_intersection_by_key is unique in that it has no
+ *        \c values_first2 parameter because elements from the second input range are never copied to the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection_by_key to compute the
+ *  set intersection of two sets of integers sorted in descending order with their values using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {11, 9, 7, 5, 3, 1};
+ *  int A_vals[6] = { 0, 0, 0, 0, 0, 0};
+ *  
+ *  int B_keys[7] = {13, 8, 5, 3, 2, 1, 1};
+ *
+ *  int keys_result[7];
+ *  int vals_result[7];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_intersection_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 7, A_vals, keys_result, vals_result, thrust::greater<int>());
+ *
+ *  // keys_result is now {5, 3, 1}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_intersection_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                            InputIterator1                                              keys_first1,
+                            InputIterator1                                              keys_last1,
+                            InputIterator2                                              keys_first2,
+                            InputIterator2                                              keys_last2,
+                            InputIterator3                                              values_first1,
+                            OutputIterator1                                             keys_result,
+                            OutputIterator2                                             values_result,
+                            StrictWeakCompare                                           comp);
+
+
+/*! \p set_intersection_by_key performs a key-value intersection operation from set theory.
+ *  \p set_intersection_by_key constructs a sorted range that is the intersection of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_intersection_by_key performs the "intersection" operation from set
+ *  theory: the keys output range contains a copy of every element that is contained in both
+ *  <tt>[keys_first1, keys_last1)</tt> <tt>[keys_first2, keys_last2)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if an element appears \c m times in <tt>[keys_first1, keys_last1)</tt>
+ *  and \c n times in <tt>[keys_first2, keys_last2)</tt> (where \c m may be zero), then it
+ *  appears <tt>min(m,n)</tt> times in the keys output range.
+ *  \p set_intersection_by_key is stable, meaning both that elements are copied from the first
+ *  input range rather than the second, and that the relative order of elements in the output range
+ *  is the same as the first input range.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> to the keys output range,
+ *  the corresponding value element is copied from <tt>[values_first1, values_last1)</tt> to the values
+ *  output range.
+ *
+ *  This version of \p set_intersection_by_key compares objects using a function object \p comp.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \note Unlike the other key-value set operations, \p set_intersection_by_key is unique in that it has no
+ *        \c values_first2 parameter because elements from the second input range are never copied to the output range.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_intersection_by_key to compute the
+ *  set intersection of two sets of integers sorted in descending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A_keys[6] = {11, 9, 7, 5, 3, 1};
+ *  int A_vals[6] = { 0, 0, 0, 0, 0, 0};
+ *  
+ *  int B_keys[7] = {13, 8, 5, 3, 2, 1, 1};
+ *
+ *  int keys_result[7];
+ *  int vals_result[7];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_intersection_by_key(A_keys, A_keys + 6, B_keys, B_keys + 7, A_vals, keys_result, vals_result, thrust::greater<int>());
+ *
+ *  // keys_result is now {5, 3, 1}
+ *  // vals_result is now {0, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_intersection_by_key(InputIterator1                             keys_first1,
+                            InputIterator1                             keys_last1,
+                            InputIterator2                             keys_first2,
+                            InputIterator2                             keys_last2,
+                            InputIterator3                             values_first1,
+                            OutputIterator1                            keys_result,
+                            OutputIterator2                            values_result,
+                            StrictWeakCompare                          comp);
+
+
+/*! \p set_symmetric_difference_by_key performs a key-value symmetric difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the symmetric difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_symmetric_difference_by_key performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[keys_first1, keys_last1)</tt> but not <tt>[keys_first2, keys_last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[keys_first2, keys_last2)</tt> but not <tt>[keys_first1, keys_last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[keys_first2, keys_last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[keys_first1, keys_last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[keys_first2, keys_last2)</tt> if <tt>m < n</tt>.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_symmetric_difference_by_key compares key elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in ascending order with their values using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {0, 1, 2, 2, 4, 6, 7};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {1, 1, 2, 5, 8};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[6];
+ *  int vals_result[6];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {0, 4, 5, 6, 7, 8}
+ *  // vals_result is now {0, 0, 1, 0, 0, 1}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_symmetric_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                    InputIterator1                                              keys_first1,
+                                    InputIterator1                                              keys_last1,
+                                    InputIterator2                                              keys_first2,
+                                    InputIterator2                                              keys_last2,
+                                    InputIterator3                                              values_first1,
+                                    InputIterator4                                              values_first2,
+                                    OutputIterator1                                             keys_result,
+                                    OutputIterator2                                             values_result);
+
+
+/*! \p set_symmetric_difference_by_key performs a key-value symmetric difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the symmetric difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_symmetric_difference_by_key performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[keys_first1, keys_last1)</tt> but not <tt>[keys_first2, keys_last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[keys_first2, keys_last2)</tt> but not <tt>[keys_first1, keys_last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[keys_first2, keys_last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[keys_first1, keys_last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[keys_first2, keys_last2)</tt> if <tt>m < n</tt>.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_symmetric_difference_by_key compares key elements using \c operator<.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in ascending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A_keys[6] = {0, 1, 2, 2, 4, 6, 7};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {1, 1, 2, 5, 8};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[6];
+ *  int vals_result[6];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {0, 4, 5, 6, 7, 8}
+ *  // vals_result is now {0, 0, 1, 0, 0, 1}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_symmetric_difference_by_key(InputIterator1                             keys_first1,
+                                    InputIterator1                             keys_last1,
+                                    InputIterator2                             keys_first2,
+                                    InputIterator2                             keys_last2,
+                                    InputIterator3                             values_first1,
+                                    InputIterator4                             values_first2,
+                                    OutputIterator1                            keys_result,
+                                    OutputIterator2                            values_result);
+
+
+/*! \p set_symmetric_difference_by_key performs a key-value symmetric difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the symmetric difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_symmetric_difference_by_key performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[keys_first1, keys_last1)</tt> but not <tt>[keys_first2, keys_last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[keys_first2, keys_last2)</tt> but not <tt>[keys_first1, keys_last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[keys_first2, keys_last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[keys_first1, keys_last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[keys_first2, keys_last2)</tt> if <tt>m < n</tt>.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_symmetric_difference_by_key compares key elements using a function object \c comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in descending order with their values using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {7, 6, 4, 2, 2, 1, 0};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {8, 5, 2, 1, 1};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[6];
+ *  int vals_result[6];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {8, 7, 6, 5, 4, 0}
+ *  // vals_result is now {1, 0, 0, 1, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_symmetric_difference_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                    InputIterator1                                              keys_first1,
+                                    InputIterator1                                              keys_last1,
+                                    InputIterator2                                              keys_first2,
+                                    InputIterator2                                              keys_last2,
+                                    InputIterator3                                              values_first1,
+                                    InputIterator4                                              values_first2,
+                                    OutputIterator1                                             keys_result,
+                                    OutputIterator2                                             values_result,
+                                    StrictWeakCompare                                           comp);
+
+
+/*! \p set_symmetric_difference_by_key performs a key-value symmetric difference operation from set theory.
+ *  \p set_difference_by_key constructs a sorted range that is the symmetric difference of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_symmetric_difference_by_key performs a set theoretic calculation:
+ *  it constructs the union of the two sets A - B and B - A, where A and B are the two
+ *  input ranges. That is, the output range contains a copy of every element that is
+ *  contained in <tt>[keys_first1, keys_last1)</tt> but not <tt>[keys_first2, keys_last1)</tt>, and a copy of
+ *  every element that is contained in <tt>[keys_first2, keys_last2)</tt> but not <tt>[keys_first1, keys_last1)</tt>.
+ *  The general case is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements that are
+ *  equivalent to each other and <tt>[keys_first2, keys_last1)</tt> contains \c n elements that are
+ *  equivalent to them, then <tt>|m - n|</tt> of those elements shall be copied to the output
+ *  range: the last <tt>m - n</tt> elements from <tt>[keys_first1, keys_last1)</tt> if <tt>m > n</tt>, and
+ *  the last <tt>n - m</tt> of these elements from <tt>[keys_first2, keys_last2)</tt> if <tt>m < n</tt>.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_symmetric_difference_by_key compares key elements using a function object \c comp.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in descending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A_keys[6] = {7, 6, 4, 2, 2, 1, 0};
+ *  int A_vals[6] = {0, 0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {8, 5, 2, 1, 1};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[6];
+ *  int vals_result[6];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {8, 7, 6, 5, 4, 0}
+ *  // vals_result is now {1, 0, 0, 1, 0, 0}
+ *  \endcode
+ *
+ *  \see \p set_union_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_symmetric_difference_by_key(InputIterator1                             keys_first1,
+                                    InputIterator1                             keys_last1,
+                                    InputIterator2                             keys_first2,
+                                    InputIterator2                             keys_last2,
+                                    InputIterator3                             values_first1,
+                                    InputIterator4                             values_first2,
+                                    OutputIterator1                            keys_result,
+                                    OutputIterator2                            values_result,
+                                    StrictWeakCompare                          comp);
+
+
+/*! \p set_union_by_key performs a key-value union operation from set theory.
+ *  \p set_union_by_key constructs a sorted range that is the union of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_union_by_key performs the "union" operation from set theory:
+ *  the output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt>, <tt>[keys_first2, keys_last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_union_by_key compares key elements using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in ascending order with their values using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {0, 2, 4, 6, 8, 10, 12};
+ *  int A_vals[6] = {0, 0, 0, 0, 0,  0,  0};
+ *
+ *  int B_keys[5] = {1, 3, 5, 7, 9};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[11];
+ *  int vals_result[11];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12}
+ *  // vals_result is now {0, 1, 0, 1, 0, 1, 0, 1, 0, 1,  0,  0}
+ *  \endcode
+ *
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_union_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                     InputIterator1                                              keys_first1,
+                     InputIterator1                                              keys_last1,
+                     InputIterator2                                              keys_first2,
+                     InputIterator2                                              keys_last2,
+                     InputIterator3                                              values_first1,
+                     InputIterator4                                              values_first2,
+                     OutputIterator1                                             keys_result,
+                     OutputIterator2                                             values_result);
+
+
+/*! \p set_union_by_key performs a key-value union operation from set theory.
+ *  \p set_union_by_key constructs a sorted range that is the union of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_union_by_key performs the "union" operation from set theory:
+ *  the output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt>, <tt>[keys_first2, keys_last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_union_by_key compares key elements using \c operator<.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to <tt>operator<</tt>.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in ascending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  ...
+ *  int A_keys[6] = {0, 2, 4, 6, 8, 10, 12};
+ *  int A_vals[6] = {0, 0, 0, 0, 0,  0,  0};
+ *
+ *  int B_keys[5] = {1, 3, 5, 7, 9};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[11];
+ *  int vals_result[11];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result);
+ *  // keys_result is now {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12}
+ *  // vals_result is now {0, 1, 0, 1, 0, 1, 0, 1, 0, 1,  0,  0}
+ *  \endcode
+ *
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_union_by_key(InputIterator1                             keys_first1,
+                     InputIterator1                             keys_last1,
+                     InputIterator2                             keys_first2,
+                     InputIterator2                             keys_last2,
+                     InputIterator3                             values_first1,
+                     InputIterator4                             values_first2,
+                     OutputIterator1                            keys_result,
+                     OutputIterator2                            values_result);
+
+
+/*! \p set_union_by_key performs a key-value union operation from set theory.
+ *  \p set_union_by_key constructs a sorted range that is the union of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_union_by_key performs the "union" operation from set theory:
+ *  the output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt>, <tt>[keys_first2, keys_last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_union_by_key compares key elements using a function object \c comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in descending order with their values using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  int A_keys[6] = {12, 10, 8, 6, 4, 2, 0};
+ *  int A_vals[6] = { 0,  0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {9, 7, 5, 3, 1};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[11];
+ *  int vals_result[11];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(thrust::host, A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result, thrust::greater<int>());
+ *  // keys_result is now {12, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0}
+ *  // vals_result is now { 0,  1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0}
+ *  \endcode
+ *
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_union_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                     InputIterator1                                              keys_first1,
+                     InputIterator1                                              keys_last1,
+                     InputIterator2                                              keys_first2,
+                     InputIterator2                                              keys_last2,
+                     InputIterator3                                              values_first1,
+                     InputIterator4                                              values_first2,
+                     OutputIterator1                                             keys_result,
+                     OutputIterator2                                             values_result,
+                     StrictWeakCompare                                           comp);
+
+
+/*! \p set_union_by_key performs a key-value union operation from set theory.
+ *  \p set_union_by_key constructs a sorted range that is the union of the sorted
+ *  ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt>. Associated
+ *  with each element from the input and output key ranges is a value element. The associated input
+ *  value ranges need not be sorted.
+ *
+ *  In the simplest case, \p set_union_by_key performs the "union" operation from set theory:
+ *  the output range contains a copy of every element that is contained in
+ *  <tt>[keys_first1, keys_last1)</tt>, <tt>[keys_first2, keys_last1)</tt>, or both. The general case
+ *  is more complicated, because the input ranges may contain duplicate elements.
+ *  The generalization is that if <tt>[keys_first1, keys_last1)</tt> contains \c m elements
+ *  that are equivalent to each other and if <tt>[keys_first2, keys_last2)</tt> contains \c n
+ *  elements that are equivalent to them, then all \c m elements from the first
+ *  range shall be copied to the output range, in order, and then <tt>max(n - m, 0)</tt>
+ *  elements from the second range shall be copied to the output, in order.
+ *
+ *  Each time a key element is copied from <tt>[keys_first1, keys_last1)</tt> or
+ *  <tt>[keys_first2, keys_last2)</tt> is copied to the keys output range, the
+ *  corresponding value element is copied from the corresponding values input range (beginning at
+ *  \p values_first1 or \p values_first2) to the values output range.
+ *
+ *  This version of \p set_union_by_key compares key elements using a function object \c comp.
+ *
+ *  \param keys_first1 The beginning of the first input range of keys.
+ *  \param keys_last1 The end of the first input range of keys.
+ *  \param keys_first2 The beginning of the second input range of keys.
+ *  \param keys_last2 The end of the second input range of keys.
+ *  \param values_first1 The beginning of the first input range of values.
+ *  \param values_first2 The beginning of the first input range of values.
+ *  \param keys_result The beginning of the output range of keys.
+ *  \param values_result The beginning of the output range of values.
+ *  \param comp Comparison operator.
+ *  \return A \p pair \c p such that <tt>p.first</tt> is the end of the output range of keys,
+ *          and such that <tt>p.second</tt> is the end of the output range of values.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator1 and \p InputIterator2 have the same \c value_type,
+ *          \p InputIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator1's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator1's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          \p InputIterator2 and \p InputIterator1 have the same \c value_type,
+ *          \p InputIterator2's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          the ordering on \p InputIterator2's \c value_type is a strict weak ordering, as defined in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements,
+ *          and \p InputIterator2's \c value_type is convertable to a type in \p OutputIterator's set of \c value_types.
+ *  \tparam InputIterator3 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator3's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam InputIterator4 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator4's \c value_type is convertible to a type in \p OutputIterator2's set of \c value_types.
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam StrictWeakCompare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The ranges <tt>[keys_first1, keys_last1)</tt> and <tt>[keys_first2, keys_last2)</tt> shall be sorted with respect to \p comp.
+ *  \pre The resulting ranges shall not overlap with any input range.
+ *
+ *  The following code snippet demonstrates how to use \p set_symmetric_difference_by_key to compute the
+ *  symmetric difference of two sets of integers sorted in descending order with their values.
+ *
+ *  \code
+ *  #include <thrust/set_operations.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  int A_keys[6] = {12, 10, 8, 6, 4, 2, 0};
+ *  int A_vals[6] = { 0,  0, 0, 0, 0, 0, 0};
+ *
+ *  int B_keys[5] = {9, 7, 5, 3, 1};
+ *  int B_vals[5] = {1, 1, 1, 1, 1};
+ *
+ *  int keys_result[11];
+ *  int vals_result[11];
+ *
+ *  thrust::pair<int*,int*> end = thrust::set_symmetric_difference_by_key(A_keys, A_keys + 6, B_keys, B_keys + 5, A_vals, B_vals, keys_result, vals_result, thrust::greater<int>());
+ *  // keys_result is now {12, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0}
+ *  // vals_result is now { 0,  1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0}
+ *  \endcode
+ *
+ *  \see \p set_symmetric_difference_by_key
+ *  \see \p set_intersection_by_key
+ *  \see \p set_difference_by_key
+ *  \see \p sort_by_key
+ *  \see \p is_sorted
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakCompare>
+  thrust::pair<OutputIterator1,OutputIterator2>
+    set_union_by_key(InputIterator1                             keys_first1,
+                     InputIterator1                             keys_last1,
+                     InputIterator2                             keys_first2,
+                     InputIterator2                             keys_last2,
+                     InputIterator3                             values_first1,
+                     InputIterator4                             values_first2,
+                     OutputIterator1                            keys_result,
+                     OutputIterator2                            values_result,
+                     StrictWeakCompare                          comp);
+
+
+/*! \} // end set_operations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/set_operations.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/shuffle.h b/thirdparty/manifold/thirdparty/thrust/thrust/shuffle.h
new file mode 100644
index 000000000000..d95327e29ce4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/shuffle.h
@@ -0,0 +1,179 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file shuffle.h
+ *  \brief Reorders range by a uniform random permutation
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reordering
+*  \ingroup algorithms
+*
+*  \addtogroup shuffling
+*  \ingroup reordering
+*  \{
+*/
+
+
+/*! \p shuffle reorders the elements <tt>[first, last)</tt> by a uniform pseudorandom permutation, defined by
+ *  random engine \p g.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to shuffle.
+ *  \param last The end of the sequence to shuffle.
+ *  \param g A UniformRandomBitGenerator
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomIterator is a random access iterator
+ *  \tparam URBG is a uniform random bit generator
+ *
+ *  The following code snippet demonstrates how to use \p shuffle to create a random permutation
+ *  using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/shuffle.h>
+ *  #include <thrust/random.h>
+ *  #include <thrust/execution_policy.h>
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::default_random_engine g;
+ *  thrust::shuffle(thrust::host, A, A + N, g);
+ *  // A is now {6, 5, 8, 7, 2, 1, 4, 3, 10, 9}
+ *  \endcode
+ *
+ *  \see \p shuffle_copy
+ */
+template <typename DerivedPolicy, typename RandomIterator, typename URBG>
+__host__ __device__ void shuffle(
+    const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
+    RandomIterator first, RandomIterator last, URBG&& g);
+
+/*! \p shuffle reorders the elements <tt>[first, last)</tt> by a uniform pseudorandom permutation, defined by
+ *  random engine \p g.
+ *
+ *  \param first The beginning of the sequence to shuffle.
+ *  \param last The end of the sequence to shuffle.
+ *  \param g A UniformRandomBitGenerator
+ *
+ *  \tparam RandomIterator is a random access iterator
+ *  \tparam URBG is a uniform random bit generator
+ *
+ *  The following code snippet demonstrates how to use \p shuffle to create a random permutation.
+ *
+ *  \code
+ *  #include <thrust/shuffle.h>
+ *  #include <thrust/random.h>
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::default_random_engine g;
+ *  thrust::shuffle(A, A + N, g);
+ *  // A is now {6, 5, 8, 7, 2, 1, 4, 3, 10, 9}
+ *  \endcode
+ *
+ *  \see \p shuffle_copy
+ */
+template <typename RandomIterator, typename URBG>
+__host__ __device__ void shuffle(RandomIterator first, RandomIterator last,
+                                 URBG&& g);
+
+/*! shuffle_copy differs from shuffle only in that the reordered sequence is written to different output sequences, rather than in place.
+ *  \p shuffle_copy reorders the elements <tt>[first, last)</tt> by a uniform pseudorandom permutation, defined by
+ *  random engine \p g.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence to shuffle.
+ *  \param last The end of the sequence to shuffle.
+ *  \param result Destination of shuffled sequence
+ *  \param g A UniformRandomBitGenerator
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomIterator is a random access iterator
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam URBG is a uniform random bit generator
+ *
+ *  The following code snippet demonstrates how to use \p shuffle_copy to create a random permutation.
+ *
+ *  \code
+ *  #include <thrust/shuffle.h>
+ *  #include <thrust/random.h>
+ *  #include <thrust/execution_policy.h>
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::default_random_engine g;
+ *  thrust::shuffle_copy(thrust::host, A, A + N, result, g);
+ *  // result is now {6, 5, 8, 7, 2, 1, 4, 3, 10, 9}
+ *  \endcode
+ *
+ *  \see \p shuffle
+ */
+template <typename DerivedPolicy, typename RandomIterator,
+          typename OutputIterator, typename URBG>
+__host__ __device__ void shuffle_copy(
+    const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
+    RandomIterator first, RandomIterator last, OutputIterator result, URBG&& g);
+
+/*! shuffle_copy differs from shuffle only in that the reordered sequence is written to different output sequences, rather than in place.
+ *\p shuffle_copy reorders the elements <tt>[first, last)</tt> by a uniform pseudorandom permutation, defined by
+ *  random engine \p g.
+ *
+ *  \param first The beginning of the sequence to shuffle.
+ *  \param last The end of the sequence to shuffle.
+ *  \param result Destination of shuffled sequence
+ *  \param g A UniformRandomBitGenerator
+ *
+ *  \tparam RandomIterator is a random access iterator
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam URBG is a uniform random bit generator
+ *
+ *  The following code snippet demonstrates how to use \p shuffle_copy to create a random permutation.
+ *
+ *  \code
+ *  #include <thrust/shuffle.h>
+ *  #include <thrust/random.h>
+ *  int A[] = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
+ *  int result[10];
+ *  const int N = sizeof(A)/sizeof(int);
+ *  thrust::default_random_engine g;
+ *  thrust::shuffle_copy(A, A + N, result, g);
+ *  // result is now {6, 5, 8, 7, 2, 1, 4, 3, 10, 9}
+ *  \endcode
+ *
+ *  \see \p shuffle
+ */
+template <typename RandomIterator, typename OutputIterator, typename URBG>
+__host__ __device__ void shuffle_copy(RandomIterator first, RandomIterator last,
+                                      OutputIterator result, URBG&& g);
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/shuffle.inl>
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/sort.h
new file mode 100644
index 000000000000..5cf9d6217ab0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/sort.h
@@ -0,0 +1,1358 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/sort.h
+ *  \brief Functions for reorganizing ranges into sorted order
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup sorting
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! \p sort sorts the elements in <tt>[first, last)</tt> into
+ *  ascending order, meaning that if \c i and \c j are any two valid
+ *  iterators in <tt>[first, last)</tt> such that \c i precedes \c j,
+ *  then \c *j is not less than \c *i. Note: \c sort is not guaranteed
+ *  to be stable. That is, suppose that \c *i and \c *j are equivalent:
+ *  neither one is less than the other. It is not guaranteed that the
+ *  relative order of these two elements will be preserved by \p sort.
+ *
+ *  This version of \p sort compares objects using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *
+ *  The following code snippet demonstrates how to use \p sort to sort
+ *  a sequence of integers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::sort(thrust::host, A, A + N);
+ *  // A is now {1, 2, 4, 5, 7, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort
+ *  \see \p sort_by_key
+ */
+template<typename DerivedPolicy, typename RandomAccessIterator>
+__host__ __device__
+  void sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last);
+
+
+/*! \p sort sorts the elements in <tt>[first, last)</tt> into
+ *  ascending order, meaning that if \c i and \c j are any two valid
+ *  iterators in <tt>[first, last)</tt> such that \c i precedes \c j,
+ *  then \c *j is not less than \c *i. Note: \c sort is not guaranteed
+ *  to be stable. That is, suppose that \c *i and \c *j are equivalent:
+ *  neither one is less than the other. It is not guaranteed that the
+ *  relative order of these two elements will be preserved by \p sort.
+ *
+ *  This version of \p sort compares objects using \c operator<.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *
+ *  The following code snippet demonstrates how to use \p sort to sort
+ *  a sequence of integers.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::sort(A, A + N);
+ *  // A is now {1, 2, 4, 5, 7, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort
+ *  \see \p sort_by_key
+ */
+template<typename RandomAccessIterator>
+  void sort(RandomAccessIterator first,
+            RandomAccessIterator last);
+
+
+/*! \p sort sorts the elements in <tt>[first, last)</tt> into
+ *  ascending order, meaning that if \c i and \c j are any two valid
+ *  iterators in <tt>[first, last)</tt> such that \c i precedes \c j,
+ *  then \c *j is not less than \c *i. Note: \c sort is not guaranteed
+ *  to be stable. That is, suppose that \c *i and \c *j are equivalent:
+ *  neither one is less than the other. It is not guaranteed that the
+ *  relative order of these two elements will be preserved by \p sort.
+ *
+ *  This version of \p sort compares objects using a function object
+ *  \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param comp  Comparison operator.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code demonstrates how to sort integers in descending order
+ *  using the greater<int> comparison operator using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::sort(thrust::host, A, A + N, thrust::greater<int>());
+ *  // A is now {8, 7, 5, 4, 2, 1};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort
+ *  \see \p sort_by_key
+ */
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last,
+            StrictWeakOrdering comp);
+
+
+/*! \p sort sorts the elements in <tt>[first, last)</tt> into
+ *  ascending order, meaning that if \c i and \c j are any two valid
+ *  iterators in <tt>[first, last)</tt> such that \c i precedes \c j,
+ *  then \c *j is not less than \c *i. Note: \c sort is not guaranteed
+ *  to be stable. That is, suppose that \c *i and \c *j are equivalent:
+ *  neither one is less than the other. It is not guaranteed that the
+ *  relative order of these two elements will be preserved by \p sort.
+ *
+ *  This version of \p sort compares objects using a function object
+ *  \p comp.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param comp  Comparison operator.
+ *
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code demonstrates how to sort integers in descending order
+ *  using the greater<int> comparison operator.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::sort(A, A + N, thrust::greater<int>());
+ *  // A is now {8, 7, 5, 4, 2, 1};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort
+ *  \see \p sort_by_key
+ */
+template<typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort(RandomAccessIterator first,
+            RandomAccessIterator last,
+            StrictWeakOrdering comp);
+
+
+/*! \p stable_sort is much like \c sort: it sorts the elements in
+ *  <tt>[first, last)</tt> into ascending order, meaning that if \c i
+ *  and \c j are any two valid iterators in <tt>[first, last)</tt> such
+ *  that \c i precedes \c j, then \c *j is not less than \c *i.
+ *
+ *  As the name suggests, \p stable_sort is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[first, last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort compares objects using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *
+ *  The following code snippet demonstrates how to use \p sort to sort
+ *  a sequence of integers using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::stable_sort(thrust::host, A, A + N);
+ *  // A is now {1, 2, 4, 5, 7, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_sort
+ *  \see \p sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename DerivedPolicy, typename RandomAccessIterator>
+__host__ __device__
+  void stable_sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last);
+
+
+/*! \p stable_sort is much like \c sort: it sorts the elements in
+ *  <tt>[first, last)</tt> into ascending order, meaning that if \c i
+ *  and \c j are any two valid iterators in <tt>[first, last)</tt> such
+ *  that \c i precedes \c j, then \c *j is not less than \c *i.
+ *
+ *  As the name suggests, \p stable_sort is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[first, last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort compares objects using \c operator<.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *
+ *  The following code snippet demonstrates how to use \p sort to sort
+ *  a sequence of integers.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::stable_sort(A, A + N);
+ *  // A is now {1, 2, 4, 5, 7, 8}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_sort
+ *  \see \p sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename RandomAccessIterator>
+  void stable_sort(RandomAccessIterator first,
+                   RandomAccessIterator last);
+
+
+/*! \p stable_sort is much like \c sort: it sorts the elements in
+ *  <tt>[first, last)</tt> into ascending order, meaning that if \c i
+ *  and \c j are any two valid iterators in <tt>[first, last)</tt> such
+ *  that \c i precedes \c j, then \c *j is not less than \c *i.
+ *
+ *  As the name suggests, \p stable_sort is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[first, last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort compares objects using a function object
+ *  \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param comp Comparison operator.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code demonstrates how to sort integers in descending order
+ *  using the greater<int> comparison operator using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::sort(A, A + N, thrust::greater<int>());
+ *  // A is now {8, 7, 5, 4, 2, 1};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_sort
+ *  \see \p sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last,
+                   StrictWeakOrdering comp);
+
+
+/*! \p stable_sort is much like \c sort: it sorts the elements in
+ *  <tt>[first, last)</tt> into ascending order, meaning that if \c i
+ *  and \c j are any two valid iterators in <tt>[first, last)</tt> such
+ *  that \c i precedes \c j, then \c *j is not less than \c *i.
+ *
+ *  As the name suggests, \p stable_sort is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[first, last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort compares objects using a function object
+ *  \p comp.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param comp Comparison operator.
+ *
+ *  \tparam RandomAccessIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator is mutable,
+ *          and \p RandomAccessIterator's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code demonstrates how to sort integers in descending order
+ *  using the greater<int> comparison operator.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  const int N = 6;
+ *  int A[N] = {1, 4, 2, 8, 5, 7};
+ *  thrust::sort(A, A + N, thrust::greater<int>());
+ *  // A is now {8, 7, 5, 4, 2, 1};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/stable_sort
+ *  \see \p sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+  void stable_sort(RandomAccessIterator first,
+                   RandomAccessIterator last,
+                   StrictWeakOrdering comp);
+
+
+///////////////
+// Key Value //
+///////////////
+
+
+/*! \p sort_by_key performs a key-value sort. That is, \p sort_by_key sorts the
+ *  elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  Note: \c sort_by_key is not guaranteed to be stable. That is, suppose that
+ *  \c *i and \c *j are equivalent: neither one is less than the other. It is not
+ *  guaranteed that the relative order of these two keys or the relative
+ *  order of their corresponding values will be preserved by \p sort_by_key.
+ *
+ *  This version of \p sort_by_key compares key objects using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator1's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p sort_by_key to sort
+ *  an array of character values using integers as sorting keys using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::sort_by_key(thrust::host, keys, keys + N, values);
+ *  // keys is now   {  1,   2,   4,   5,   7,   8}
+ *  // values is now {'a', 'c', 'b', 'e', 'f', 'd'}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort_by_key
+ *  \see \p sort
+ */
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first);
+
+
+/*! \p sort_by_key performs a key-value sort. That is, \p sort_by_key sorts the
+ *  elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  Note: \c sort_by_key is not guaranteed to be stable. That is, suppose that
+ *  \c *i and \c *j are equivalent: neither one is less than the other. It is not
+ *  guaranteed that the relative order of these two keys or the relative
+ *  order of their corresponding values will be preserved by \p sort_by_key.
+ *
+ *  This version of \p sort_by_key compares key objects using \c operator<.
+ *
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator1's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p sort_by_key to sort
+ *  an array of character values using integers as sorting keys.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::sort_by_key(keys, keys + N, values);
+ *  // keys is now   {  1,   2,   4,   5,   7,   8}
+ *  // values is now {'a', 'c', 'b', 'e', 'f', 'd'}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort_by_key
+ *  \see \p sort
+ */
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  void sort_by_key(RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first);
+
+
+/*! \p sort_by_key performs a key-value sort. That is, \p sort_by_key sorts the
+ *  elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  Note: \c sort_by_key is not guaranteed to be stable. That is, suppose that
+ *  \c *i and \c *j are equivalent: neither one is less than the other. It is not
+ *  guaranteed that the relative order of these two keys or the relative
+ *  order of their corresponding values will be preserved by \p sort_by_key.
+ *
+ *  This version of \p sort_by_key compares key objects using a function object
+ *  \c comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *  \param comp Comparison operator.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p sort_by_key to sort
+ *  an array of character values using integers as sorting keys using the \p thrust::host execution policy
+ *  for parallelization.The keys are sorted in descending order using the <tt>greater<int></tt> comparison operator.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::sort_by_key(thrust::host, keys, keys + N, values, thrust::greater<int>());
+ *  // keys is now   {  8,   7,   5,   4,   2,   1}
+ *  // values is now {'d', 'f', 'e', 'b', 'c', 'a'}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort_by_key
+ *  \see \p sort
+ */
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first,
+                   StrictWeakOrdering comp);
+
+
+/*! \p sort_by_key performs a key-value sort. That is, \p sort_by_key sorts the
+ *  elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  Note: \c sort_by_key is not guaranteed to be stable. That is, suppose that
+ *  \c *i and \c *j are equivalent: neither one is less than the other. It is not
+ *  guaranteed that the relative order of these two keys or the relative
+ *  order of their corresponding values will be preserved by \p sort_by_key.
+ *
+ *  This version of \p sort_by_key compares key objects using a function object
+ *  \c comp.
+ *
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *  \param comp Comparison operator.
+ *
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p sort_by_key to sort
+ *  an array of character values using integers as sorting keys.  The keys
+ *  are sorted in descending order using the greater<int> comparison operator.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::sort_by_key(keys, keys + N, values, thrust::greater<int>());
+ *  // keys is now   {  8,   7,   5,   4,   2,   1}
+ *  // values is now {'d', 'f', 'e', 'b', 'c', 'a'}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p stable_sort_by_key
+ *  \see \p sort
+ */
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+  void sort_by_key(RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first,
+                   StrictWeakOrdering comp);
+
+
+/*! \p stable_sort_by_key performs a key-value sort. That is, \p stable_sort_by_key
+ *  sorts the elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  As the name suggests, \p stable_sort_by_key is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[keys_first, keys_last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort_by_key is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort_by_key compares key objects using \c operator<.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator1's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p stable_sort_by_key to sort
+ *  an array of characters using integers as sorting keys using the \p thrust::host execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::stable_sort_by_key(thrust::host, keys, keys + N, values);
+ *  // keys is now   {  1,   2,   4,   5,   7,   8}
+ *  // values is now {'a', 'c', 'b', 'e', 'f', 'd'}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ */
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void stable_sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first);
+
+
+/*! \p stable_sort_by_key performs a key-value sort. That is, \p stable_sort_by_key
+ *  sorts the elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  As the name suggests, \p stable_sort_by_key is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[keys_first, keys_last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort_by_key is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort_by_key compares key objects using \c operator<.
+ *
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering relation on \p RandomAccessIterator1's \c value_type is a <em>strict weak ordering</em>, as defined in the
+ *          <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p stable_sort_by_key to sort
+ *  an array of characters using integers as sorting keys.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::stable_sort_by_key(keys, keys + N, values);
+ *  // keys is now   {  1,   2,   4,   5,   7,   8}
+ *  // values is now {'a', 'c', 'b', 'e', 'f', 'd'}
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ */
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  void stable_sort_by_key(RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first);
+
+
+/*! \p stable_sort_by_key performs a key-value sort. That is, \p stable_sort_by_key
+ *  sorts the elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  As the name suggests, \p stable_sort_by_key is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[keys_first, keys_last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort_by_key is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort_by_key compares key objects using the function
+ *  object \p comp.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *  \param comp Comparison operator.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p sort_by_key to sort
+ *  an array of character values using integers as sorting keys using the \p thrust::host execution policy for
+ *  parallelization. The keys are sorted in descending order using the <tt>greater<int></tt> comparison operator.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::stable_sort_by_key(thrust::host, keys, keys + N, values, thrust::greater<int>());
+ *  // keys is now   {  8,   7,   5,   4,   2,   1}
+ *  // values is now {'d', 'f', 'e', 'b', 'c', 'a'}
+ *  \endcode
+ *
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ */
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first,
+                          StrictWeakOrdering comp);
+
+
+/*! \p stable_sort_by_key performs a key-value sort. That is, \p stable_sort_by_key
+ *  sorts the elements in <tt>[keys_first, keys_last)</tt> and <tt>[values_first,
+ *  values_first + (keys_last - keys_first))</tt> into ascending key order,
+ *  meaning that if \c i and \c j are any two valid iterators in <tt>[keys_first,
+ *  keys_last)</tt> such that \c i precedes \c j, and \c p and \c q are iterators
+ *  in <tt>[values_first, values_first + (keys_last - keys_first))</tt>
+ *  corresponding to \c i and \c j respectively, then \c *j is not less than
+ *  \c *i.
+ *
+ *  As the name suggests, \p stable_sort_by_key is stable: it preserves the
+ *  relative ordering of equivalent elements. That is, if \c x and \c y
+ *  are elements in <tt>[keys_first, keys_last)</tt> such that \c x precedes \c y,
+ *  and if the two elements are equivalent (neither <tt>x < y</tt> nor
+ *  <tt>y < x</tt>) then a postcondition of \p stable_sort_by_key is that \c x
+ *  still precedes \c y.
+ *
+ *  This version of \p stable_sort_by_key compares key objects using the function
+ *  object \p comp.
+ *
+ *  \param keys_first The beginning of the key sequence.
+ *  \param keys_last The end of the key sequence.
+ *  \param values_first The beginning of the value sequence.
+ *  \param comp Comparison operator.
+ *
+ *  \tparam RandomAccessIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/random_access_iterator">Random Access Iterator</a>,
+ *          \p RandomAccessIterator1 is mutable,
+ *          and \p RandomAccessIterator1's \c value_type is convertible to \p StrictWeakOrdering's
+ *          \c first_argument_type and \c second_argument_type.
+ *  \tparam RandomAccessIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator">Random Access Iterator</a>,
+ *          and \p RandomAccessIterator2 is mutable.
+ *  \tparam StrictWeakOrdering is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  \pre The range <tt>[keys_first, keys_last))</tt> shall not overlap the range <tt>[values_first, values_first + (keys_last - keys_first))</tt>.
+ *
+ *  The following code snippet demonstrates how to use \p sort_by_key to sort
+ *  an array of character values using integers as sorting keys.  The keys
+ *  are sorted in descending order using the greater<int> comparison operator.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  ...
+ *  const int N = 6;
+ *  int    keys[N] = {  1,   4,   2,   8,   5,   7};
+ *  char values[N] = {'a', 'b', 'c', 'd', 'e', 'f'};
+ *  thrust::stable_sort_by_key(keys, keys + N, values, thrust::greater<int>());
+ *  // keys is now   {  8,   7,   5,   4,   2,   1}
+ *  // values is now {'d', 'f', 'e', 'b', 'c', 'a'}
+ *  \endcode
+ *
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ */
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+  void stable_sort_by_key(RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first,
+                          StrictWeakOrdering comp);
+
+
+/*! \} // end sorting
+ */
+
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup predicates
+ *  \{
+ */
+
+
+/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is
+ *  sorted in ascending order, and \c false otherwise.
+ *
+ *  Specifically, this version of \p is_sorted returns \c false if for
+ *  some iterator \c i in the range <tt>[first, last - 1)</tt> the
+ *  expression <tt>*(i + 1) < *i</tt> is \c true.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return \c true, if the sequence is sorted; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering on objects of \p ForwardIterator's \c value_type is a <em>strict weak ordering</em>, as defined
+ *          in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *
+ *
+ *  The following code demonstrates how to use \p is_sorted to test whether the
+ *  contents of a \c device_vector are stored in ascending order using the \p thrust::device execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> v(6);
+ *  v[0] = 1;
+ *  v[1] = 4;
+ *  v[2] = 2;
+ *  v[3] = 8;
+ *  v[4] = 5;
+ *  v[5] = 7;
+ *
+ *  bool result = thrust::is_sorted(thrust::device, v.begin(), v.end());
+ *
+ *  // result == false
+ *
+ *  thrust::sort(v.begin(), v.end());
+ *  result = thrust::is_sorted(thrust::device, v.begin(), v.end());
+ *
+ *  // result == true
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/is_sorted
+ *  \see is_sorted_until
+ *  \see \c sort
+ *  \see \c stable_sort
+ *  \see \c less<T>
+ */
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  bool is_sorted(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last);
+
+
+/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is
+ *  sorted in ascending order, and \c false otherwise.
+ *
+ *  Specifically, this version of \p is_sorted returns \c false if for
+ *  some iterator \c i in the range <tt>[first, last - 1)</tt> the
+ *  expression <tt>*(i + 1) < *i</tt> is \c true.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \return \c true, if the sequence is sorted; \c false, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>,
+ *          and the ordering on objects of \p ForwardIterator's \c value_type is a <em>strict weak ordering</em>, as defined
+ *          in the <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a> requirements.
+ *
+ *
+ *  The following code demonstrates how to use \p is_sorted to test whether the
+ *  contents of a \c device_vector are stored in ascending order.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/sort.h>
+ *  ...
+ *  thrust::device_vector<int> v(6);
+ *  v[0] = 1;
+ *  v[1] = 4;
+ *  v[2] = 2;
+ *  v[3] = 8;
+ *  v[4] = 5;
+ *  v[5] = 7;
+ *
+ *  bool result = thrust::is_sorted(v.begin(), v.end());
+ *
+ *  // result == false
+ *
+ *  thrust::sort(v.begin(), v.end());
+ *  result = thrust::is_sorted(v.begin(), v.end());
+ *
+ *  // result == true
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/is_sorted
+ *  \see is_sorted_until
+ *  \see \c sort
+ *  \see \c stable_sort
+ *  \see \c less<T>
+ */
+template<typename ForwardIterator>
+  bool is_sorted(ForwardIterator first,
+                 ForwardIterator last);
+
+
+/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is sorted in ascending 
+ *  order accoring to a user-defined comparison operation, and \c false otherwise.
+ *
+ *  Specifically, this version of \p is_sorted returns \c false if for some iterator \c i in
+ *  the range <tt>[first, last - 1)</tt> the expression <tt>comp(*(i + 1), *i)</tt> is \c true.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp  Comparison operator.
+ *  \return \c true, if the sequence is sorted according to comp; \c false, otherwise.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \c StrictWeakOrdering's \c first_argument_type
+ *          and \c second_argument_type.
+ *  \tparam Compare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p is_sorted to test whether the
+ *  contents of a \c device_vector are stored in descending order using the \p thrust::device execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> v(6);
+ *  v[0] = 1;
+ *  v[1] = 4;
+ *  v[2] = 2;
+ *  v[3] = 8;
+ *  v[4] = 5;
+ *  v[5] = 7;
+ *
+ *  thrust::greater<int> comp;
+ *  bool result = thrust::is_sorted(thrust::device, v.begin(), v.end(), comp);
+ *
+ *  // result == false
+ *
+ *  thrust::sort(v.begin(), v.end(), comp);
+ *  result = thrust::is_sorted(thrust::device, v.begin(), v.end(), comp);
+ *
+ *  // result == true
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/is_sorted
+ *  \see \c sort
+ *  \see \c stable_sort
+ *  \see \c less<T>
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename Compare>
+__host__ __device__
+  bool is_sorted(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 Compare comp);
+
+
+/*! \p is_sorted returns \c true if the range <tt>[first, last)</tt> is sorted in ascending 
+ *  order accoring to a user-defined comparison operation, and \c false otherwise.
+ *
+ *  Specifically, this version of \p is_sorted returns \c false if for some iterator \c i in
+ *  the range <tt>[first, last - 1)</tt> the expression <tt>comp(*(i + 1), *i)</tt> is \c true.
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last  The end of the sequence.
+ *  \param comp  Comparison operator.
+ *  \return \c true, if the sequence is sorted according to comp; \c false, otherwise.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to both \c StrictWeakOrdering's \c first_argument_type
+ *          and \c second_argument_type.
+ *  \tparam Compare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p is_sorted to test whether the
+ *  contents of a \c device_vector are stored in descending order.
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> v(6);
+ *  v[0] = 1;
+ *  v[1] = 4;
+ *  v[2] = 2;
+ *  v[3] = 8;
+ *  v[4] = 5;
+ *  v[5] = 7;
+ *
+ *  thrust::greater<int> comp;
+ *  bool result = thrust::is_sorted(v.begin(), v.end(), comp);
+ *
+ *  // result == false
+ *
+ *  thrust::sort(v.begin(), v.end(), comp);
+ *  result = thrust::is_sorted(v.begin(), v.end(), comp);
+ *
+ *  // result == true
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/is_sorted
+ *  \see \c sort
+ *  \see \c stable_sort
+ *  \see \c less<T>
+ */
+template<typename ForwardIterator, typename Compare>
+  bool is_sorted(ForwardIterator first,
+                 ForwardIterator last,
+                 Compare comp);
+
+
+/*! This version of \p is_sorted_until returns the last iterator \c i in <tt>[first,last]</tt> for
+ *  which the range <tt>[first,last)</tt> is sorted using \c operator<. If <tt>distance(first,last) < 2</tt>,
+ *  \p is_sorted_until simply returns \p last.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \return The last iterator in the input range for which it is sorted.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a> and
+ *          \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p is_sorted_until to find the first position
+ *  in an array where the data becomes unsorted using the \p thrust::host execution policy for
+ *  parallelization:
+ *  
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  ...
+ *   
+ *  int A[8] = {0, 1, 2, 3, 0, 1, 2, 3};
+ *  
+ *  int * B = thrust::is_sorted_until(thrust::host, A, A + 8);
+ *  
+ *  // B - A is 4
+ *  // [A, B) is sorted
+ *  \endcode
+ *
+ *  \see \p is_sorted
+ *  \see \p sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator is_sorted_until(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last);
+
+
+/*! This version of \p is_sorted_until returns the last iterator \c i in <tt>[first,last]</tt> for
+ *  which the range <tt>[first,last)</tt> is sorted using \c operator<. If <tt>distance(first,last) < 2</tt>,
+ *  \p is_sorted_until simply returns \p last.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \return The last iterator in the input range for which it is sorted.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a> and
+ *          \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/named_req/LessThanComparable">LessThan Comparable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p is_sorted_until to find the first position
+ *  in an array where the data becomes unsorted:
+ *  
+ *  \code
+ *  #include <thrust/sort.h>
+ *
+ *  ...
+ *   
+ *  int A[8] = {0, 1, 2, 3, 0, 1, 2, 3};
+ *  
+ *  int * B = thrust::is_sorted_until(A, A + 8);
+ *  
+ *  // B - A is 4
+ *  // [A, B) is sorted
+ *  \endcode
+ *
+ *  \see \p is_sorted
+ *  \see \p sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename ForwardIterator>
+  ForwardIterator is_sorted_until(ForwardIterator first,
+                                  ForwardIterator last);
+
+
+/*! This version of \p is_sorted_until returns the last iterator \c i in <tt>[first,last]</tt> for
+ *  which the range <tt>[first,last)</tt> is sorted using the function object \c comp. If <tt>distance(first,last) < 2</tt>,
+ *  \p is_sorted_until simply returns \p last.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization:
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param comp The function object to use for comparison.
+ *  \return The last iterator in the input range for which it is sorted.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a> and
+ *          \p ForwardIterator's \c value_type is convertible to \p Compare's \c argument_type.
+ *  \tparam Compare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p is_sorted_until to find the first position
+ *  in an array where the data becomes unsorted in descending order using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  ...
+ *   
+ *  int A[8] = {3, 2, 1, 0, 3, 2, 1, 0};
+ *  
+ *  thrust::greater<int> comp;
+ *  int * B = thrust::is_sorted_until(thrust::host, A, A + 8, comp);
+ *  
+ *  // B - A is 4
+ *  // [A, B) is sorted in descending order
+ *  \endcode
+ *
+ *  \see \p is_sorted
+ *  \see \p sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename Compare>
+__host__ __device__
+  ForwardIterator is_sorted_until(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Compare comp);
+
+
+/*! This version of \p is_sorted_until returns the last iterator \c i in <tt>[first,last]</tt> for
+ *  which the range <tt>[first,last)</tt> is sorted using the function object \c comp. If <tt>distance(first,last) < 2</tt>,
+ *  \p is_sorted_until simply returns \p last.
+ *
+ *  \param first The beginning of the range of interest.
+ *  \param last The end of the range of interest.
+ *  \param comp The function object to use for comparison.
+ *  \return The last iterator in the input range for which it is sorted.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a> and
+ *          \p ForwardIterator's \c value_type is convertible to \p Compare's \c argument_type.
+ *  \tparam Compare is a model of <a href="https://en.cppreference.com/w/cpp/concepts/strict_weak_order">Strict Weak Ordering</a>.
+ *
+ *  The following code snippet demonstrates how to use \p is_sorted_until to find the first position
+ *  in an array where the data becomes unsorted in descending order:
+ *
+ *  \code
+ *  #include <thrust/sort.h>
+ *  #include <thrust/functional.h>
+ *
+ *  ...
+ *   
+ *  int A[8] = {3, 2, 1, 0, 3, 2, 1, 0};
+ *  
+ *  thrust::greater<int> comp;
+ *  int * B = thrust::is_sorted_until(A, A + 8, comp);
+ *  
+ *  // B - A is 4
+ *  // [A, B) is sorted in descending order
+ *  \endcode
+ *
+ *  \see \p is_sorted
+ *  \see \p sort
+ *  \see \p sort_by_key
+ *  \see \p stable_sort
+ *  \see \p stable_sort_by_key
+ */
+template<typename ForwardIterator, typename Compare>
+  ForwardIterator is_sorted_until(ForwardIterator first,
+                                  ForwardIterator last,
+                                  Compare comp);
+
+
+/*! \} // end predicates
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/sort.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/swap.h b/thirdparty/manifold/thirdparty/thrust/thrust/swap.h
new file mode 100644
index 000000000000..d8a8be73c311
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/swap.h
@@ -0,0 +1,184 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file swap.h
+ *  \brief Functions for swapping the value of elements
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup swap
+ *  \{
+ */
+
+/*! \p swap assigns the contents of \c a to \c b and the
+ *  contents of \c b to \c a. This is used as a primitive operation
+ *  by many other algorithms.
+ *  
+ *  \param a The first value of interest. After completion,
+ *           the value of b will be returned here.
+ *  \param b The second value of interest. After completion,
+ *           the value of a will be returned here.
+ *
+ *  \tparam Assignable is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p swap to
+ *  swap the contents of two variables.
+ *
+ *  \code
+ *  #include <thrust/swap.h>
+ *  ...
+ *  int x = 1;
+ *  int y = 2;
+ *  thrust::swap(x,h);
+ *
+ *  // x == 2, y == 1
+ *  \endcode
+ */
+template<typename Assignable1, typename Assignable2>
+__host__ __device__ 
+inline void swap(Assignable1 &a, Assignable2 &b);
+
+/*! \} // swap
+ */
+
+/*! \} // utility
+ */
+
+
+/*! \addtogroup copying
+ *  \{
+ */
+
+
+/*! \p swap_ranges swaps each of the elements in the range <tt>[first1, last1)</tt>
+ *  with the corresponding element in the range <tt>[first2, first2 + (last1 - first1))</tt>.
+ *  That is, for each integer \c n such that <tt>0 <= n < (last1 - first1)</tt>, it swaps
+ *  <tt>*(first1 + n)</tt> and <tt>*(first2 + n)</tt>. The return value is
+ *  <tt>first2 + (last1 - first1)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first sequence to swap.
+ *  \param last1 One position past the last element of the first sequence to swap.
+ *  \param first2 The beginning of the second sequence to swap.
+ *  \return An iterator pointing to one position past the last element of the second
+ *          sequence to swap.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator1's \c value_type must be convertible to \p ForwardIterator2's \c value_type.
+ *  \tparam ForwardIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator2's \c value_type must be convertible to \p ForwardIterator1's \c value_type.
+ *
+ *  \pre \p first1 may equal \p first2, but the range <tt>[first1, last1)</tt> shall not overlap the range <tt>[first2, first2 + (last1 - first1))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p swap_ranges to
+ *  swap the contents of two \c thrust::device_vectors using the \p thrust::device execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/swap.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  thrust::device_vector<int> v1(2), v2(2);
+ *  v1[0] = 1;
+ *  v1[1] = 2;
+ *  v2[0] = 3;
+ *  v2[1] = 4;
+ *
+ *  thrust::swap_ranges(thrust::device, v1.begin(), v1.end(), v2.begin());
+ *
+ *  // v1[0] == 3, v1[1] == 4, v2[0] == 1, v2[1] == 2
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/swap_ranges
+ *  \see \c swap
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  ForwardIterator2 swap_ranges(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               ForwardIterator1 first1,
+                               ForwardIterator1 last1,
+                               ForwardIterator2 first2);
+
+
+/*! \p swap_ranges swaps each of the elements in the range <tt>[first1, last1)</tt>
+ *  with the corresponding element in the range <tt>[first2, first2 + (last1 - first1))</tt>.
+ *  That is, for each integer \c n such that <tt>0 <= n < (last1 - first1)</tt>, it swaps
+ *  <tt>*(first1 + n)</tt> and <tt>*(first2 + n)</tt>. The return value is
+ *  <tt>first2 + (last1 - first1)</tt>.
+ *
+ *  \param first1 The beginning of the first sequence to swap.
+ *  \param last1 One position past the last element of the first sequence to swap.
+ *  \param first2 The beginning of the second sequence to swap.
+ *  \return An iterator pointing to one position past the last element of the second
+ *          sequence to swap.
+ *
+ *  \tparam ForwardIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator1's \c value_type must be convertible to \p ForwardIterator2's \c value_type.
+ *  \tparam ForwardIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator2's \c value_type must be convertible to \p ForwardIterator1's \c value_type.
+ *
+ *  \pre \p first1 may equal \p first2, but the range <tt>[first1, last1)</tt> shall not overlap the range <tt>[first2, first2 + (last1 - first1))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p swap_ranges to
+ *  swap the contents of two \c thrust::device_vectors.
+ *
+ *  \code
+ *  #include <thrust/swap.h>
+ *  #include <thrust/device_vector.h>
+ *  ...
+ *  thrust::device_vector<int> v1(2), v2(2);
+ *  v1[0] = 1;
+ *  v1[1] = 2;
+ *  v2[0] = 3;
+ *  v2[1] = 4;
+ *
+ *  thrust::swap_ranges(v1.begin(), v1.end(), v2.begin());
+ *
+ *  // v1[0] == 3, v1[1] == 4, v2[0] == 1, v2[1] == 2
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/swap_ranges
+ *  \see \c swap
+ */
+template<typename ForwardIterator1,
+         typename ForwardIterator2>
+  ForwardIterator2 swap_ranges(ForwardIterator1 first1,
+                               ForwardIterator1 last1,
+                               ForwardIterator2 first2);
+
+
+/*! \} // copying
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/swap.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/adjacent_difference.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/adjacent_difference.h
new file mode 100644
index 000000000000..b82242c7c079
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/adjacent_difference.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits adjacent_difference
+#include <thrust/system/detail/sequential/adjacent_difference.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/assign_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/assign_value.h
new file mode 100644
index 000000000000..d5f14bd1636d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/assign_value.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits assign_value
+#include <thrust/system/detail/sequential/assign_value.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/binary_search.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/binary_search.h
new file mode 100644
index 000000000000..a2c33f32aad3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/binary_search.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/system/cpp/detail/execution_policy.h>
+
+// this system inherits the binary search algorithms
+#include <thrust/system/detail/sequential/binary_search.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy.h
new file mode 100644
index 000000000000..d2f5b1bd45cd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits copy
+#include <thrust/system/detail/sequential/copy.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy_if.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy_if.h
new file mode 100644
index 000000000000..10869e2a9021
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/copy_if.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits copy_if
+#include <thrust/system/detail/sequential/copy_if.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/count.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/count.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/count.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/equal.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/equal.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/equal.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/execution_policy.h
new file mode 100644
index 000000000000..1a8193bf3acf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/execution_policy.h
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+// put the canonical tag in the same ns as the backend's entry points
+namespace cpp
+{
+namespace detail
+{
+
+// this awkward sequence of definitions arise
+// from the desire both for tag to derive
+// from execution_policy and for execution_policy
+// to convert to tag (when execution_policy is not
+// an ancestor of tag)
+
+// forward declaration of tag
+struct tag;
+
+// forward declaration of execution_policy
+template<typename> struct execution_policy;
+
+// specialize execution_policy for tag
+template<>
+  struct execution_policy<tag>
+    : thrust::system::detail::sequential::execution_policy<tag>
+{};
+
+// tag's definition comes before the
+// generic definition of execution_policy
+struct tag : execution_policy<tag> {};
+
+// allow conversion to tag when it is not a successor
+template<typename Derived>
+  struct execution_policy
+    : thrust::system::detail::sequential::execution_policy<Derived>
+{
+  typedef tag tag_type; 
+  operator tag() const { return tag(); }
+};
+
+} // end detail
+
+// alias execution_policy and tag here
+using thrust::system::cpp::detail::execution_policy;
+using thrust::system::cpp::detail::tag;
+
+} // end cpp
+} // end system
+
+// alias items at top-level
+namespace cpp
+{
+
+using thrust::system::cpp::execution_policy;
+using thrust::system::cpp::tag;
+
+} // end cpp
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/extrema.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/extrema.h
new file mode 100644
index 000000000000..5fbb8c55c287
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/extrema.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits extrema algorithms
+#include <thrust/system/detail/sequential/extrema.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/fill.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/fill.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/find.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/find.h
new file mode 100644
index 000000000000..29c0dafc8cee
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/find.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits find
+#include <thrust/system/detail/sequential/find.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/for_each.h
new file mode 100644
index 000000000000..9ef45dfe62e4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/for_each.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits for_each
+#include <thrust/system/detail/sequential/for_each.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/gather.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/gather.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/gather.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/generate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/generate.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/generate.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/get_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/get_value.h
new file mode 100644
index 000000000000..915001d37f4d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/get_value.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits get_value
+#include <thrust/system/detail/sequential/get_value.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/inner_product.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/inner_product.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/inner_product.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/iter_swap.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/iter_swap.h
new file mode 100644
index 000000000000..7a982becb394
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/iter_swap.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits iter_swap
+#include <thrust/system/detail/sequential/iter_swap.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/logical.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/logical.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/logical.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/malloc_and_free.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/malloc_and_free.h
new file mode 100644
index 000000000000..b1ad7a7341bf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/malloc_and_free.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits malloc & free
+#include <thrust/system/detail/sequential/malloc_and_free.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/memory.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/memory.inl
new file mode 100644
index 000000000000..650aa1cb5016
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/memory.inl
@@ -0,0 +1,52 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/cpp/memory.h>
+#include <thrust/system/cpp/detail/malloc_and_free.h>
+#include <limits>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace cpp
+{
+
+pointer<void> malloc(std::size_t n)
+{
+  tag t;
+  return pointer<void>(thrust::system::detail::sequential::malloc(t, n));
+} // end malloc()
+
+template<typename T>
+pointer<T> malloc(std::size_t n)
+{
+  pointer<void> raw_ptr = thrust::system::cpp::malloc(sizeof(T) * n);
+  return pointer<T>(reinterpret_cast<T*>(raw_ptr.get()));
+} // end malloc()
+
+void free(pointer<void> ptr)
+{
+  tag t;
+  return thrust::system::detail::sequential::free(t, ptr);
+} // end free()
+
+} // end cpp
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/merge.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/merge.h
new file mode 100644
index 000000000000..20e17f2d40dd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/merge.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits merge
+#include <thrust/system/detail/sequential/merge.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/mismatch.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/mismatch.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/mismatch.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/par.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/par.h
new file mode 100644
index 000000000000..c56921327f5e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/par.h
@@ -0,0 +1,61 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/allocator_aware_execution_policy.h>
+#include <thrust/system/cpp/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace cpp
+{
+namespace detail
+{
+
+
+struct par_t : thrust::system::cpp::detail::execution_policy<par_t>,
+  thrust::detail::allocator_aware_execution_policy<
+    thrust::system::cpp::detail::execution_policy>
+{
+  __host__ __device__
+  constexpr par_t() : thrust::system::cpp::detail::execution_policy<par_t>() {}
+};
+
+
+} // end detail
+
+
+THRUST_INLINE_CONSTANT detail::par_t par;
+
+
+} // end cpp
+} // end system
+
+
+// alias par here
+namespace cpp
+{
+
+
+using thrust::system::cpp::par;
+
+
+} // end cpp
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/partition.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/partition.h
new file mode 100644
index 000000000000..50d48e222aa5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/partition.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits partition
+#include <thrust/system/detail/sequential/partition.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/per_device_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/per_device_resource.h
new file mode 100644
index 000000000000..1b8d61f92169
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/per_device_resource.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special per device resource functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce.h
new file mode 100644
index 000000000000..e09652cd9fd6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits reduce
+#include <thrust/system/detail/sequential/reduce.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce_by_key.h
new file mode 100644
index 000000000000..a2c7744249bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reduce_by_key.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits reduce_by_key
+#include <thrust/system/detail/sequential/reduce_by_key.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/remove.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/remove.h
new file mode 100644
index 000000000000..a529f625d6a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/remove.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits remove
+#include <thrust/system/detail/sequential/remove.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/replace.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/replace.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/replace.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reverse.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reverse.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/reverse.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan.h
new file mode 100644
index 000000000000..4d38e6484373
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits scan
+#include <thrust/system/detail/sequential/scan.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan_by_key.h
new file mode 100644
index 000000000000..1e72e7c38f86
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scan_by_key.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits the scan_by_key algorithms
+#include <thrust/system/detail/sequential/scan_by_key.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scatter.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scatter.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/scatter.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sequence.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sequence.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/set_operations.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/set_operations.h
new file mode 100644
index 000000000000..bbde20114c2c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/set_operations.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits the set operations
+#include <thrust/system/detail/sequential/set_operations.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sort.h
new file mode 100644
index 000000000000..ae38b3ba8c78
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/sort.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits sort
+#include <thrust/system/detail/sequential/sort.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/swap_ranges.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/swap_ranges.h
new file mode 100644
index 000000000000..35cd5462207a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/swap_ranges.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// cpp has no special swap_ranges
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/tabulate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/tabulate.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/tabulate.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/temporary_buffer.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/temporary_buffer.h
new file mode 100644
index 000000000000..2adfaf2810c6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/temporary_buffer.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special temporary buffer functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform.h
new file mode 100644
index 000000000000..895164ce5afb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// cpp has no special transform
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_reduce.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_reduce.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_scan.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/transform_scan.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_copy.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_copy.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_fill.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/uninitialized_fill.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique.h
new file mode 100644
index 000000000000..2ff23e9d3db0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits unique
+#include <thrust/system/detail/sequential/unique.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique_by_key.h
new file mode 100644
index 000000000000..1d40011787cb
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/unique_by_key.h
@@ -0,0 +1,23 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system inherits unique_by_key
+#include <thrust/system/detail/sequential/unique_by_key.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/vector.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/vector.inl
new file mode 100644
index 000000000000..02980c62a429
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/detail/vector.inl
@@ -0,0 +1,146 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/cpp/vector.h>
+#include <utility>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace cpp
+{
+
+template<typename T, typename Allocator>
+  vector<T,Allocator>
+    ::vector()
+      : super_t()
+{}
+
+template<typename T, typename Allocator>
+  vector<T,Allocator>
+    ::vector(size_type n)
+      : super_t(n)
+{}
+
+template<typename T, typename Allocator>
+  vector<T,Allocator>
+    ::vector(size_type n, const value_type &value)
+      : super_t(n,value)
+{}
+
+template<typename T, typename Allocator>
+  vector<T,Allocator>
+    ::vector(const vector &x)
+      : super_t(x)
+{}
+
+#if THRUST_CPP_DIALECT >= 2011
+  template<typename T, typename Allocator>
+    vector<T,Allocator>
+      ::vector(vector &&x)
+        : super_t(std::move(x))
+  {}
+#endif
+
+template<typename T, typename Allocator>
+  template<typename OtherT, typename OtherAllocator>
+    vector<T,Allocator>
+      ::vector(const thrust::detail::vector_base<OtherT,OtherAllocator> &x)
+        : super_t(x)
+{}
+
+template<typename T, typename Allocator>
+  template<typename OtherT, typename OtherAllocator>
+    vector<T,Allocator>
+      ::vector(const std::vector<OtherT,OtherAllocator> &x)
+        : super_t(x)
+{}
+
+template<typename T, typename Allocator>
+  template<typename InputIterator>
+    vector<T,Allocator>
+      ::vector(InputIterator first, InputIterator last)
+        : super_t(first,last)
+{}
+
+template<typename T, typename Allocator>
+  vector<T,Allocator> &
+    vector<T,Allocator>
+      ::operator=(const vector &x)
+{
+  super_t::operator=(x);
+  return *this;
+}
+
+#if THRUST_CPP_DIALECT >= 2011
+  template<typename T, typename Allocator>
+    vector<T,Allocator> &
+      vector<T,Allocator>
+        ::operator=(vector &&x)
+  {
+    super_t::operator=(std::move(x));
+    return *this;
+  }
+#endif
+  
+  template<typename T, typename Allocator>
+    vector<T,Allocator>
+      ::vector(std::initializer_list<T> il)
+        : super_t(il)
+  {}
+  
+  template<typename T, typename Allocator>
+    vector<T,Allocator>
+      ::vector(std::initializer_list<T> il, const Allocator& alloc)
+        : super_t(il, alloc)
+  {}
+
+  template<typename T, typename Allocator>
+    vector<T,Allocator> &
+      vector<T,Allocator>
+        ::operator=(std::initializer_list<T> il)
+  {
+    super_t::operator=(il);
+    return *this;
+  }
+
+template<typename T, typename Allocator>
+  template<typename OtherT, typename OtherAllocator>
+    vector<T,Allocator> &
+      vector<T,Allocator>
+        ::operator=(const std::vector<OtherT,OtherAllocator> &x)
+{
+  super_t::operator=(x);
+  return *this;
+}
+
+template<typename T, typename Allocator>
+  template<typename OtherT, typename OtherAllocator>
+    vector<T,Allocator> &
+      vector<T,Allocator>
+        ::operator=(const thrust::detail::vector_base<OtherT,OtherAllocator> &x)
+{
+  super_t::operator=(x);
+  return *this;
+}
+      
+} // end cpp
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/execution_policy.h
new file mode 100644
index 000000000000..0d8a9a36790b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/execution_policy.h
@@ -0,0 +1,156 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/system/cpp/execution_policy.h
+ *  \brief Execution policies for Thrust's Standard C++ system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// get the execution policies definitions first
+#include <thrust/system/cpp/detail/execution_policy.h>
+
+// get the definition of par
+#include <thrust/system/cpp/detail/par.h>
+
+// now get all the algorithm definitions
+
+#include <thrust/system/cpp/detail/adjacent_difference.h>
+#include <thrust/system/cpp/detail/assign_value.h>
+#include <thrust/system/cpp/detail/binary_search.h>
+#include <thrust/system/cpp/detail/copy.h>
+#include <thrust/system/cpp/detail/copy_if.h>
+#include <thrust/system/cpp/detail/count.h>
+#include <thrust/system/cpp/detail/equal.h>
+#include <thrust/system/cpp/detail/extrema.h>
+#include <thrust/system/cpp/detail/fill.h>
+#include <thrust/system/cpp/detail/find.h>
+#include <thrust/system/cpp/detail/for_each.h>
+#include <thrust/system/cpp/detail/gather.h>
+#include <thrust/system/cpp/detail/generate.h>
+#include <thrust/system/cpp/detail/get_value.h>
+#include <thrust/system/cpp/detail/inner_product.h>
+#include <thrust/system/cpp/detail/iter_swap.h>
+#include <thrust/system/cpp/detail/logical.h>
+#include <thrust/system/cpp/detail/malloc_and_free.h>
+#include <thrust/system/cpp/detail/merge.h>
+#include <thrust/system/cpp/detail/mismatch.h>
+#include <thrust/system/cpp/detail/partition.h>
+#include <thrust/system/cpp/detail/reduce.h>
+#include <thrust/system/cpp/detail/reduce_by_key.h>
+#include <thrust/system/cpp/detail/remove.h>
+#include <thrust/system/cpp/detail/replace.h>
+#include <thrust/system/cpp/detail/reverse.h>
+#include <thrust/system/cpp/detail/scan.h>
+#include <thrust/system/cpp/detail/scan_by_key.h>
+#include <thrust/system/cpp/detail/scatter.h>
+#include <thrust/system/cpp/detail/sequence.h>
+#include <thrust/system/cpp/detail/set_operations.h>
+#include <thrust/system/cpp/detail/sort.h>
+#include <thrust/system/cpp/detail/swap_ranges.h>
+#include <thrust/system/cpp/detail/tabulate.h>
+#include <thrust/system/cpp/detail/transform.h>
+#include <thrust/system/cpp/detail/transform_reduce.h>
+#include <thrust/system/cpp/detail/transform_scan.h>
+#include <thrust/system/cpp/detail/uninitialized_copy.h>
+#include <thrust/system/cpp/detail/uninitialized_fill.h>
+#include <thrust/system/cpp/detail/unique.h>
+#include <thrust/system/cpp/detail/unique_by_key.h>
+
+
+// define these entities here for the purpose of Doxygenating them
+// they are actually defined elsewhere
+#if 0
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace cpp
+{
+
+
+/*! \addtogroup execution_policies
+ *  \{
+ */
+
+
+/*! \p thrust::system::cpp::execution_policy is the base class for all Thrust parallel execution
+ *  policies which are derived from Thrust's standard C++ backend system.
+ */
+template<typename DerivedPolicy>
+struct execution_policy : thrust::execution_policy<DerivedPolicy>
+{};
+
+
+/*! \p thrust::system::cpp::tag is a type representing Thrust's standard C++ backend system in C++'s type system.
+ *  Iterators "tagged" with a type which is convertible to \p cpp::tag assert that they may be
+ *  "dispatched" to algorithm implementations in the \p cpp system.
+ */
+struct tag : thrust::system::cpp::execution_policy<tag> { unspecified };
+
+
+/*!
+ *  \p thrust::system::cpp::par is the parallel execution policy associated with Thrust's standard
+ *  C++ backend system.
+ *
+ *  Instead of relying on implicit algorithm dispatch through iterator system tags, users may
+ *  directly target Thrust's C++ backend system by providing \p thrust::cpp::par as an algorithm
+ *  parameter.
+ *
+ *  Explicit dispatch can be useful in avoiding the introduction of data copies into containers such
+ *  as \p thrust::cpp::vector.
+ *
+ *  The type of \p thrust::cpp::par is implementation-defined.
+ *
+ *  The following code snippet demonstrates how to use \p thrust::cpp::par to explicitly dispatch an
+ *  invocation of \p thrust::for_each to the standard C++ backend system:
+ *
+ *  \code
+ *  #include <thrust/for_each.h>
+ *  #include <thrust/system/cpp/execution_policy.h>
+ *  #include <cstdio>
+ *
+ *  struct printf_functor
+ *  {
+ *    __host__ __device__
+ *    void operator()(int x)
+ *    {
+ *      printf("%d\n", x);
+ *    }
+ *  };
+ *  ...
+ *  int vec[3];
+ *  vec[0] = 0; vec[1] = 1; vec[2] = 2;
+ *
+ *  thrust::for_each(thrust::cpp::par, vec.begin(), vec.end(), printf_functor());
+ *
+ *  // 0 1 2 is printed to standard output in some unspecified order
+ *  \endcode
+ */
+static const unspecified par;
+
+
+/*! \}
+ */
+
+
+} // end cpp
+} // end system
+THRUST_NAMESPACE_END
+#endif
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory.h
new file mode 100644
index 000000000000..a18abeb8e4f5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory.h
@@ -0,0 +1,100 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/system/cpp/memory.h
+ *  \brief Managing memory associated with Thrust's Standard C++ system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/cpp/memory_resource.h>
+#include <thrust/memory.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/mr/allocator.h>
+#include <ostream>
+
+THRUST_NAMESPACE_BEGIN
+namespace system { namespace cpp
+{
+
+/*! Allocates an area of memory available to Thrust's <tt>cpp</tt> system.
+ *  \param n Number of bytes to allocate.
+ *  \return A <tt>cpp::pointer<void></tt> pointing to the beginning of the newly
+ *          allocated memory. A null <tt>cpp::pointer<void></tt> is returned if
+ *          an error occurs.
+ *  \note The <tt>cpp::pointer<void></tt> returned by this function must be
+ *        deallocated with \p cpp::free.
+ *  \see cpp::free
+ *  \see std::malloc
+ */
+inline pointer<void> malloc(std::size_t n);
+
+/*! Allocates a typed area of memory available to Thrust's <tt>cpp</tt> system.
+ *  \param n Number of elements to allocate.
+ *  \return A <tt>cpp::pointer<T></tt> pointing to the beginning of the newly
+ *          allocated elements. A null <tt>cpp::pointer<T></tt> is returned if
+ *          an error occurs.
+ *  \note The <tt>cpp::pointer<T></tt> returned by this function must be
+ *        deallocated with \p cpp::free.
+ *  \see cpp::free
+ *  \see std::malloc
+ */
+template<typename T>
+inline pointer<T> malloc(std::size_t n);
+
+/*! Deallocates an area of memory previously allocated by <tt>cpp::malloc</tt>.
+ *  \param ptr A <tt>cpp::pointer<void></tt> pointing to the beginning of an area
+ *         of memory previously allocated with <tt>cpp::malloc</tt>.
+ *  \see cpp::malloc
+ *  \see std::free
+ */
+inline void free(pointer<void> ptr);
+
+/*! \p cpp::allocator is the default allocator used by the \p cpp system's
+ *  containers such as <tt>cpp::vector</tt> if no user-specified allocator is
+ *  provided. \p cpp::allocator allocates (deallocates) storage with \p
+ *  cpp::malloc (\p cpp::free).
+ */
+template<typename T>
+using allocator = thrust::mr::stateless_resource_allocator<
+  T, thrust::system::cpp::memory_resource
+>;
+
+/*! \p cpp::universal_allocator allocates memory that can be used by the \p cpp
+ *  system and host systems.
+ */
+template<typename T>
+using universal_allocator = thrust::mr::stateless_resource_allocator<
+  T, thrust::system::cpp::universal_memory_resource
+>;
+
+}} // namespace system::cpp
+
+/*! \namespace thrust::cpp
+ *  \brief \p thrust::cpp is a top-level alias for thrust::system::cpp.
+ */
+namespace cpp
+{
+using thrust::system::cpp::malloc;
+using thrust::system::cpp::free;
+using thrust::system::cpp::allocator;
+} // namespace cpp
+
+THRUST_NAMESPACE_END
+
+#include <thrust/system/cpp/detail/memory.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory_resource.h
new file mode 100644
index 000000000000..04b4e3cf8dab
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/memory_resource.h
@@ -0,0 +1,71 @@
+/*
+ *  Copyright 2018-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file cpp/memory_resource.h
+ *  \brief Memory resources for the Standard C++ system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/mr/new.h>
+#include <thrust/mr/fancy_pointer_resource.h>
+
+#include <thrust/system/cpp/pointer.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system { namespace cpp
+{
+
+//! \cond
+namespace detail
+{
+    typedef thrust::mr::fancy_pointer_resource<
+        thrust::mr::new_delete_resource,
+        thrust::cpp::pointer<void>
+    > native_resource;
+
+    typedef thrust::mr::fancy_pointer_resource<
+        thrust::mr::new_delete_resource,
+        thrust::cpp::universal_pointer<void>
+    > universal_native_resource;
+} // namespace detail
+//! \endcond
+
+/*! \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management
+ *  \{
+ */
+
+/*! The memory resource for the Standard C++ system. Uses \p
+ *  mr::new_delete_resource and tags it with \p cpp::pointer.
+ */
+typedef detail::native_resource memory_resource;
+/*! The unified memory resource for the Standard C++ system. Uses
+ *  \p mr::new_delete_resource and tags it with \p cpp::universal_pointer.
+ */
+typedef detail::universal_native_resource universal_memory_resource;
+/*! An alias for \p cpp::universal_memory_resource. */
+typedef detail::native_resource universal_host_pinned_memory_resource;
+
+/*! \} // memory_resources
+ */
+
+
+}} // namespace system::cpp
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/pointer.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/pointer.h
new file mode 100644
index 000000000000..f204fa37509e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/pointer.h
@@ -0,0 +1,117 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/system/cpp/memory.h
+ *  \brief Managing memory associated with Thrust's TBB system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <type_traits>
+#include <thrust/system/cpp/detail/execution_policy.h>
+#include <thrust/detail/pointer.h>
+#include <thrust/detail/reference.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system { namespace cpp
+{
+
+/*! \p cpp::pointer stores a pointer to an object allocated in memory accessible
+ *  by the \p cpp system. This type provides type safety when dispatching
+ *  algorithms on ranges resident in \p cpp memory.
+ *
+ *  \p cpp::pointer has pointer semantics: it may be dereferenced and
+ *  manipulated with pointer arithmetic.
+ *
+ *  \p cpp::pointer can be created with the function \p cpp::malloc, or by
+ *  explicitly calling its constructor with a raw pointer.
+ *
+ *  The raw pointer encapsulated by a \p cpp::pointer may be obtained by eiter its
+ *  <tt>get</tt> member function or the \p raw_pointer_cast function.
+ *
+ *  \note \p cpp::pointer is not a "smart" pointer; it is the programmer's
+ *        responsibility to deallocate memory pointed to by \p cpp::pointer.
+ *
+ *  \tparam T specifies the type of the pointee.
+ *
+ *  \see cpp::malloc
+ *  \see cpp::free
+ *  \see raw_pointer_cast
+ */
+template <typename T>
+using pointer = thrust::pointer<
+  T,
+  thrust::system::cpp::tag,
+  thrust::tagged_reference<T, thrust::system::cpp::tag>
+>;
+
+/*! \p cpp::universal_pointer stores a pointer to an object allocated in memory
+ * accessible by the \p cpp system and host systems.
+ *
+ *  \p cpp::universal_pointer has pointer semantics: it may be dereferenced and
+ *  manipulated with pointer arithmetic.
+ *
+ *  \p cpp::universal_pointer can be created with \p cpp::universal_allocator
+ *  or by explicitly calling its constructor with a raw pointer.
+ *
+ *  The raw pointer encapsulated by a \p cpp::universal_pointer may be obtained
+ *  by eiter its <tt>get</tt> member function or the \p raw_pointer_cast
+ *  function.
+ *
+ *  \note \p cpp::universal_pointer is not a "smart" pointer; it is the
+ *        programmer's responsibility to deallocate memory pointed to by
+ *        \p cpp::universal_pointer.
+ *
+ *  \tparam T specifies the type of the pointee.
+ *
+ *  \see cpp::universal_allocator
+ *  \see raw_pointer_cast
+ */
+template <typename T>
+using universal_pointer = thrust::pointer<
+  T,
+  thrust::system::cpp::tag,
+  typename std::add_lvalue_reference<T>::type
+>;
+
+/*! \p reference is a wrapped reference to an object stored in memory available
+ *  to the \p cpp system. \p reference is the type of the result of
+ *  dereferencing a \p cpp::pointer.
+ *
+ *  \tparam T Specifies the type of the referenced object.
+ */
+template <typename T>
+using reference = thrust::reference<T, thrust::system::cpp::tag>;
+
+}} // namespace system::cpp
+
+/*! \addtogroup system_backends Systems
+ *  \ingroup system
+ *  \{
+ */
+
+/*! \namespace thrust::cpp
+ *  \brief \p thrust::cpp is a top-level alias for \p thrust::system::cpp. */
+namespace cpp
+{
+using thrust::system::cpp::pointer;
+using thrust::system::cpp::universal_pointer;
+using thrust::system::cpp::reference;
+} // namespace cpp
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/vector.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/vector.h
new file mode 100644
index 000000000000..2a418dbc30d4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/cpp/vector.h
@@ -0,0 +1,82 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/system/cpp/vector.h
+ *  \brief A dynamically-sizable array of elements which reside in memory available to
+ *         Thrust's standard C++ system.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/cpp/memory.h>
+#include <thrust/detail/vector_base.h>
+#include <vector>
+
+THRUST_NAMESPACE_BEGIN
+namespace system { namespace cpp
+{
+
+/*! \p cpp::vector is a container that supports random access to elements,
+ *  constant time removal of elements at the end, and linear time insertion
+ *  and removal of elements at the beginning or in the middle. The number of
+ *  elements in a \p cpp::vector may vary dynamically; memory management is
+ *  automatic. The elements contained in a \p cpp::vector reside in memory
+ *  accessible by the \p cpp system.
+ *
+ *  \tparam T The element type of the \p cpp::vector.
+ *  \tparam Allocator The allocator type of the \p cpp::vector.
+ *          Defaults to \p cpp::allocator.
+ *
+ *  \see https://en.cppreference.com/w/cpp/container/vector
+ *  \see host_vector For the documentation of the complete interface which is
+ *                   shared by \p cpp::vector.
+ *  \see device_vector
+ *  \see universal_vector
+ */
+template <typename T, typename Allocator = thrust::system::cpp::allocator<T>>
+using vector = thrust::detail::vector_base<T, Allocator>;
+
+/*! \p cpp::universal_vector is a container that supports random access to
+ *  elements, constant time removal of elements at the end, and linear time
+ *  insertion and removal of elements at the beginning or in the middle. The
+ *  number of elements in a \p cpp::universal_vector may vary dynamically;
+ *  memory management is automatic. The elements contained in a
+ *  \p cpp::universal_vector reside in memory accessible by the \p cpp system
+ *  and host systems.
+ *
+ *  \tparam T The element type of the \p cpp::universal_vector.
+ *  \tparam Allocator The allocator type of the \p cpp::universal_vector.
+ *          Defaults to \p cpp::universal_allocator.
+ *
+ *  \see https://en.cppreference.com/w/cpp/container/vector
+ *  \see host_vector For the documentation of the complete interface which is
+ *                   shared by \p cpp::universal_vector
+ *  \see device_vector
+ *  \see universal_vector
+ */
+template <typename T, typename Allocator = thrust::system::cpp::universal_allocator<T>>
+using universal_vector = thrust::detail::vector_base<T, Allocator>;
+
+}} // namespace system::cpp
+
+namespace cpp
+{
+using thrust::system::cpp::vector;
+using thrust::system::cpp::universal_vector;
+}
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/adjacent_difference.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/adjacent_difference.h
new file mode 100644
index 000000000000..c6f6c72820a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/adjacent_difference.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the adjacent_difference.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch adjacent_difference
+
+#include <thrust/system/detail/sequential/adjacent_difference.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/adjacent_difference.h>
+#include <thrust/system/cuda/detail/adjacent_difference.h>
+#include <thrust/system/omp/detail/adjacent_difference.h>
+#include <thrust/system/tbb/detail/adjacent_difference.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_ADJACENT_DIFFERENCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/adjacent_difference.h>
+#include __THRUST_HOST_SYSTEM_ADJACENT_DIFFERENCE_HEADER
+#undef __THRUST_HOST_SYSTEM_ADJACENT_DIFFERENCE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ADJACENT_DIFFERENCE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/adjacent_difference.h>
+#include __THRUST_DEVICE_SYSTEM_ADJACENT_DIFFERENCE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ADJACENT_DIFFERENCE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/assign_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/assign_value.h
new file mode 100644
index 000000000000..d38934affd1a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/assign_value.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the assign_value.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch assign_value
+
+#include <thrust/system/detail/sequential/assign_value.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/assign_value.h>
+#include <thrust/system/cuda/detail/assign_value.h>
+#include <thrust/system/omp/detail/assign_value.h>
+#include <thrust/system/tbb/detail/assign_value.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_ASSIGN_VALUE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/assign_value.h>
+#include __THRUST_HOST_SYSTEM_ASSIGN_VALUE_HEADER
+#undef __THRUST_HOST_SYSTEM_ASSIGN_VALUE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASSIGN_VALUE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/assign_value.h>
+#include __THRUST_DEVICE_SYSTEM_ASSIGN_VALUE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASSIGN_VALUE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/copy.h
new file mode 100644
index 000000000000..72debb3b6671
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/copy.h
@@ -0,0 +1,34 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// The purpose of this header is to #include the async/copy.h header of the
+// sequential, host, and device systems. It should be #included in any code
+// which uses ADL to dispatch async copy.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+//#include <thrust/system/detail/sequential/async/copy.h>
+
+//#define __THRUST_HOST_SYSTEM_ASYNC_COPY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/copy.h>
+//#include __THRUST_HOST_SYSTEM_ASYNC_COPY_HEADER
+//#undef __THRUST_HOST_SYSTEM_ASYNC_COPY_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASYNC_COPY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/async/copy.h>
+#include __THRUST_DEVICE_SYSTEM_ASYNC_COPY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASYNC_COPY_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/for_each.h
new file mode 100644
index 000000000000..08347f659279
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/for_each.h
@@ -0,0 +1,34 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// The purpose of this header is to #include the async/for_each.h header of the
+// sequential, host, and device systems. It should be #included in any code
+// which uses ADL to dispatch async for_each.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+//#include <thrust/system/detail/sequential/async/for_each.h>
+
+//#define __THRUST_HOST_SYSTEM_ASYNC_FOR_EACH_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/for_each.h>
+//#include __THRUST_HOST_SYSTEM_ASYNC_FOR_EACH_HEADER
+//#undef __THRUST_HOST_SYSTEM_ASYNC_FOR_EACH_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASYNC_FOR_EACH_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/async/for_each.h>
+#include __THRUST_DEVICE_SYSTEM_ASYNC_FOR_EACH_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASYNC_FOR_EACH_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/reduce.h
new file mode 100644
index 000000000000..f13ab02fdca0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/reduce.h
@@ -0,0 +1,34 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// The purpose of this header is to #include the async/reduce.h header of the
+// sequential, host, and device systems. It should be #included in any code
+// which uses ADL to dispatch async reduce.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+//#include <thrust/system/detail/sequential/async/reduce.h>
+
+//#define __THRUST_HOST_SYSTEM_ASYNC_REDUCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/reduce.h>
+//#include __THRUST_HOST_SYSTEM_ASYNC_REDUCE_HEADER
+//#undef __THRUST_HOST_SYSTEM_ASYNC_REDUCE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASYNC_REDUCE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/async/reduce.h>
+#include __THRUST_DEVICE_SYSTEM_ASYNC_REDUCE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASYNC_REDUCE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/scan.h
new file mode 100644
index 000000000000..a2a90618b487
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/scan.h
@@ -0,0 +1,34 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// The purpose of this header is to #include the async/scan.h header of the
+// sequential, host, and device systems. It should be #included in any code
+// which uses ADL to dispatch async scans.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+//#include <thrust/system/detail/sequential/async/scan.h>
+
+//#define __THRUST_HOST_SYSTEM_ASYNC_SCAN_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/scan.h>
+//#include __THRUST_HOST_SYSTEM_ASYNC_SCAN_HEADER
+//#undef __THRUST_HOST_SYSTEM_ASYNC_SCAN_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASYNC_SCAN_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/async/scan.h>
+#include __THRUST_DEVICE_SYSTEM_ASYNC_SCAN_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASYNC_SCAN_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/sort.h
new file mode 100644
index 000000000000..c3a83ad404f8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/sort.h
@@ -0,0 +1,34 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// The purpose of this header is to #include the async/sort.h header of the
+// sequential, host, and device systems. It should be #included in any code
+// which uses ADL to dispatch async sort.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+//#include <thrust/system/detail/sequential/async/sort.h>
+
+//#define __THRUST_HOST_SYSTEM_ASYNC_SORT_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/sort.h>
+//#include __THRUST_HOST_SYSTEM_ASYNC_SORT_HEADER
+//#undef __THRUST_HOST_SYSTEM_ASYNC_SORT_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASYNC_SORT_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/async/sort.h>
+#include __THRUST_DEVICE_SYSTEM_ASYNC_SORT_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASYNC_SORT_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/transform.h
new file mode 100644
index 000000000000..abb2163ead06
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/async/transform.h
@@ -0,0 +1,34 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a transform of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+// The purpose of this header is to #include the async/transform.h header of the
+// sequential, host, and device systems. It should be #included in any code
+// which uses ADL to dispatch async transform.
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+//#include <thrust/system/detail/sequential/async/transform.h>
+
+//#define __THRUST_HOST_SYSTEM_ASYNC_TRANSFORM_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/async/transform.h>
+//#include __THRUST_HOST_SYSTEM_ASYNC_TRANSFORM_HEADER
+//#undef __THRUST_HOST_SYSTEM_ASYNC_TRANSFORM_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ASYNC_TRANSFORM_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/async/transform.h>
+#include __THRUST_DEVICE_SYSTEM_ASYNC_TRANSFORM_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ASYNC_TRANSFORM_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/binary_search.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/binary_search.h
new file mode 100644
index 000000000000..2f9ac06df9bc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/binary_search.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the binary_search.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch binary_search
+
+#include <thrust/system/detail/sequential/binary_search.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/binary_search.h>
+#include <thrust/system/cuda/detail/binary_search.h>
+#include <thrust/system/omp/detail/binary_search.h>
+#include <thrust/system/tbb/detail/binary_search.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_BINARY_SEARCH_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/binary_search.h>
+#include __THRUST_HOST_SYSTEM_BINARY_SEARCH_HEADER
+#undef __THRUST_HOST_SYSTEM_BINARY_SEARCH_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_BINARY_SEARCH_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/binary_search.h>
+#include __THRUST_DEVICE_SYSTEM_BINARY_SEARCH_HEADER
+#undef __THRUST_DEVICE_SYSTEM_BINARY_SEARCH_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy.h
new file mode 100644
index 000000000000..0035b83efee0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the copy.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch copy
+
+#include <thrust/system/detail/sequential/copy.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/copy.h>
+#include <thrust/system/cuda/detail/copy.h>
+#include <thrust/system/omp/detail/copy.h>
+#include <thrust/system/tbb/detail/copy.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_COPY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/copy.h>
+#include __THRUST_HOST_SYSTEM_COPY_HEADER
+#undef __THRUST_HOST_SYSTEM_COPY_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_COPY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/copy.h>
+#include __THRUST_DEVICE_SYSTEM_COPY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_COPY_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy_if.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy_if.h
new file mode 100644
index 000000000000..31adaf8e1a73
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/copy_if.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy_if.h of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the copy_if.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch copy_if
+
+#include <thrust/system/detail/sequential/copy_if.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/copy_if.h>
+#include <thrust/system/cuda/detail/copy_if.h>
+#include <thrust/system/omp/detail/copy_if.h>
+#include <thrust/system/tbb/detail/copy_if.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_COPY_IF_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/copy_if.h>
+#include __THRUST_HOST_SYSTEM_COPY_IF_HEADER
+#undef __THRUST_HOST_SYSTEM_COPY_IF_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_COPY_IF_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/copy_if.h>
+#include __THRUST_DEVICE_SYSTEM_COPY_IF_HEADER
+#undef __THRUST_DEVICE_SYSTEM_COPY_IF_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/count.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/count.h
new file mode 100644
index 000000000000..5d6f1f748ffe
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/count.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a count of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the count.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch count
+
+#include <thrust/system/detail/sequential/count.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/count.h>
+#include <thrust/system/cuda/detail/count.h>
+#include <thrust/system/omp/detail/count.h>
+#include <thrust/system/tbb/detail/count.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_COUNT_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/count.h>
+#include __THRUST_HOST_SYSTEM_COUNT_HEADER
+#undef __THRUST_HOST_SYSTEM_COUNT_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_COUNT_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/count.h>
+#include __THRUST_DEVICE_SYSTEM_COUNT_HEADER
+#undef __THRUST_DEVICE_SYSTEM_COUNT_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/equal.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/equal.h
new file mode 100644
index 000000000000..6b02e33b857e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/equal.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a equal of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the equal.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch equal
+
+#include <thrust/system/detail/sequential/equal.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/equal.h>
+#include <thrust/system/cuda/detail/equal.h>
+#include <thrust/system/omp/detail/equal.h>
+#include <thrust/system/tbb/detail/equal.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_EQUAL_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/equal.h>
+#include __THRUST_HOST_SYSTEM_EQUAL_HEADER
+#undef __THRUST_HOST_SYSTEM_EQUAL_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_EQUAL_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/equal.h>
+#include __THRUST_DEVICE_SYSTEM_EQUAL_HEADER
+#undef __THRUST_DEVICE_SYSTEM_EQUAL_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/extrema.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/extrema.h
new file mode 100644
index 000000000000..62fb39be922d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/extrema.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a extrema of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the extrema.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch extrema
+
+#include <thrust/system/detail/sequential/extrema.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/extrema.h>
+#include <thrust/system/cuda/detail/extrema.h>
+#include <thrust/system/omp/detail/extrema.h>
+#include <thrust/system/tbb/detail/extrema.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_EXTREMA_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/extrema.h>
+#include __THRUST_HOST_SYSTEM_EXTREMA_HEADER
+#undef __THRUST_HOST_SYSTEM_EXTREMA_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_EXTREMA_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/extrema.h>
+#include __THRUST_DEVICE_SYSTEM_EXTREMA_HEADER
+#undef __THRUST_DEVICE_SYSTEM_EXTREMA_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/fill.h
new file mode 100644
index 000000000000..f76a81b4f347
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/fill.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the fill.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch fill
+
+#include <thrust/system/detail/sequential/fill.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/fill.h>
+#include <thrust/system/cuda/detail/fill.h>
+#include <thrust/system/omp/detail/fill.h>
+#include <thrust/system/tbb/detail/fill.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_FILL_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/fill.h>
+#include __THRUST_HOST_SYSTEM_FILL_HEADER
+#undef __THRUST_HOST_SYSTEM_FILL_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_FILL_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/fill.h>
+#include __THRUST_DEVICE_SYSTEM_FILL_HEADER
+#undef __THRUST_DEVICE_SYSTEM_FILL_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/find.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/find.h
new file mode 100644
index 000000000000..8d85e09a3ab7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/find.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the find.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch find
+
+#include <thrust/system/detail/sequential/find.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/find.h>
+#include <thrust/system/cuda/detail/find.h>
+#include <thrust/system/omp/detail/find.h>
+#include <thrust/system/tbb/detail/find.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_FIND_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/find.h>
+#include __THRUST_HOST_SYSTEM_FIND_HEADER
+#undef __THRUST_HOST_SYSTEM_FIND_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_FIND_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/find.h>
+#include __THRUST_DEVICE_SYSTEM_FIND_HEADER
+#undef __THRUST_DEVICE_SYSTEM_FIND_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/for_each.h
new file mode 100644
index 000000000000..8509edca3630
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/for_each.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the for_each.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch for_each
+
+#include <thrust/system/detail/sequential/for_each.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/for_each.h>
+#include <thrust/system/cuda/detail/for_each.h>
+#include <thrust/system/omp/detail/for_each.h>
+#include <thrust/system/tbb/detail/for_each.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_FOR_EACH_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/for_each.h>
+#include __THRUST_HOST_SYSTEM_FOR_EACH_HEADER
+#undef __THRUST_HOST_SYSTEM_FOR_EACH_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_FOR_EACH_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/for_each.h>
+#include __THRUST_DEVICE_SYSTEM_FOR_EACH_HEADER
+#undef __THRUST_DEVICE_SYSTEM_FOR_EACH_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/gather.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/gather.h
new file mode 100644
index 000000000000..242da3c90957
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/gather.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the gather.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch gather
+
+#include <thrust/system/detail/sequential/gather.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/gather.h>
+#include <thrust/system/cuda/detail/gather.h>
+#include <thrust/system/omp/detail/gather.h>
+#include <thrust/system/tbb/detail/gather.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_GATHER_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/gather.h>
+#include __THRUST_HOST_SYSTEM_GATHER_HEADER
+#undef __THRUST_HOST_SYSTEM_GATHER_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_GATHER_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/gather.h>
+#include __THRUST_DEVICE_SYSTEM_GATHER_HEADER
+#undef __THRUST_DEVICE_SYSTEM_GATHER_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/generate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/generate.h
new file mode 100644
index 000000000000..5b1d7b4bac5f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/generate.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the generate.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch generate
+
+#include <thrust/system/detail/sequential/generate.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/generate.h>
+#include <thrust/system/cuda/detail/generate.h>
+#include <thrust/system/omp/detail/generate.h>
+#include <thrust/system/tbb/detail/generate.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_GENERATE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/generate.h>
+#include __THRUST_HOST_SYSTEM_GENERATE_HEADER
+#undef __THRUST_HOST_SYSTEM_GENERATE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_GENERATE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/generate.h>
+#include __THRUST_DEVICE_SYSTEM_GENERATE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_GENERATE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/get_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/get_value.h
new file mode 100644
index 000000000000..306eb423eb4b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/get_value.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the get_value.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch get_value
+
+#include <thrust/system/detail/sequential/get_value.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/get_value.h>
+#include <thrust/system/cuda/detail/get_value.h>
+#include <thrust/system/omp/detail/get_value.h>
+#include <thrust/system/tbb/detail/get_value.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_GET_VALUE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/get_value.h>
+#include __THRUST_HOST_SYSTEM_GET_VALUE_HEADER
+#undef __THRUST_HOST_SYSTEM_GET_VALUE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_GET_VALUE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/get_value.h>
+#include __THRUST_DEVICE_SYSTEM_GET_VALUE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_GET_VALUE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/inner_product.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/inner_product.h
new file mode 100644
index 000000000000..9423b1bdbf19
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/inner_product.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the inner_product.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch inner_product
+
+#include <thrust/system/detail/sequential/inner_product.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/inner_product.h>
+#include <thrust/system/cuda/detail/inner_product.h>
+#include <thrust/system/omp/detail/inner_product.h>
+#include <thrust/system/tbb/detail/inner_product.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_INNER_PRODUCT_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/inner_product.h>
+#include __THRUST_HOST_SYSTEM_INNER_PRODUCT_HEADER
+#undef __THRUST_HOST_SYSTEM_INNER_PRODUCT_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_INNER_PRODUCT_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/inner_product.h>
+#include __THRUST_DEVICE_SYSTEM_INNER_PRODUCT_HEADER
+#undef __THRUST_DEVICE_SYSTEM_INNER_PRODUCT_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/iter_swap.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/iter_swap.h
new file mode 100644
index 000000000000..d9da52a6274c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/iter_swap.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the iter_swap.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch iter_swap
+
+#include <thrust/system/detail/sequential/iter_swap.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/iter_swap.h>
+#include <thrust/system/cuda/detail/iter_swap.h>
+#include <thrust/system/omp/detail/iter_swap.h>
+#include <thrust/system/tbb/detail/iter_swap.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_ITER_SWAP_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/iter_swap.h>
+#include __THRUST_HOST_SYSTEM_ITER_SWAP_HEADER
+#undef __THRUST_HOST_SYSTEM_ITER_SWAP_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_ITER_SWAP_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/iter_swap.h>
+#include __THRUST_DEVICE_SYSTEM_ITER_SWAP_HEADER
+#undef __THRUST_DEVICE_SYSTEM_ITER_SWAP_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/logical.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/logical.h
new file mode 100644
index 000000000000..bdaad4d293a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/logical.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the logical.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch logical
+
+#include <thrust/system/detail/sequential/logical.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/logical.h>
+#include <thrust/system/cuda/detail/logical.h>
+#include <thrust/system/omp/detail/logical.h>
+#include <thrust/system/tbb/detail/logical.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_LOGICAL_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/logical.h>
+#include __THRUST_HOST_SYSTEM_LOGICAL_HEADER
+#undef __THRUST_HOST_SYSTEM_LOGICAL_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_LOGICAL_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/logical.h>
+#include __THRUST_DEVICE_SYSTEM_LOGICAL_HEADER
+#undef __THRUST_DEVICE_SYSTEM_LOGICAL_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/malloc_and_free.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/malloc_and_free.h
new file mode 100644
index 000000000000..c36db0270ddf
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/malloc_and_free.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the malloc_and_free.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch malloc_and_free
+
+#include <thrust/system/detail/sequential/malloc_and_free.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/malloc_and_free.h>
+#include <thrust/system/cuda/detail/malloc_and_free.h>
+#include <thrust/system/omp/detail/malloc_and_free.h>
+#include <thrust/system/tbb/detail/malloc_and_free.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_MALLOC_AND_FREE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/malloc_and_free.h>
+#include __THRUST_HOST_SYSTEM_MALLOC_AND_FREE_HEADER
+#undef __THRUST_HOST_SYSTEM_MALLOC_AND_FREE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_MALLOC_AND_FREE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/malloc_and_free.h>
+#include __THRUST_DEVICE_SYSTEM_MALLOC_AND_FREE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_MALLOC_AND_FREE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/merge.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/merge.h
new file mode 100644
index 000000000000..7abca9bcf3f1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/merge.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the merge.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch merge
+
+#include <thrust/system/detail/sequential/merge.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/merge.h>
+#include <thrust/system/cuda/detail/merge.h>
+#include <thrust/system/omp/detail/merge.h>
+#include <thrust/system/tbb/detail/merge.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_MERGE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/merge.h>
+#include __THRUST_HOST_SYSTEM_MERGE_HEADER
+#undef __THRUST_HOST_SYSTEM_MERGE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_MERGE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/merge.h>
+#include __THRUST_DEVICE_SYSTEM_MERGE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_MERGE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/mismatch.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/mismatch.h
new file mode 100644
index 000000000000..74feb8269288
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/mismatch.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the mismatch.h header
+// of the host and device systems. It should be #included in any
+// code which uses adl to dispatch mismatch
+
+#include <thrust/system/detail/sequential/mismatch.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/mismatch.h>
+#include <thrust/system/cuda/detail/mismatch.h>
+#include <thrust/system/omp/detail/mismatch.h>
+#include <thrust/system/tbb/detail/mismatch.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_MISMATCH_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/mismatch.h>
+#include __THRUST_HOST_SYSTEM_MISMATCH_HEADER
+#undef __THRUST_HOST_SYSTEM_MISMATCH_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_MISMATCH_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/mismatch.h>
+#include __THRUST_DEVICE_SYSTEM_MISMATCH_HEADER
+#undef __THRUST_DEVICE_SYSTEM_MISMATCH_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/partition.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/partition.h
new file mode 100644
index 000000000000..a45f845a5c6e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/partition.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the partition.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch partition
+
+#include <thrust/system/detail/sequential/partition.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/partition.h>
+#include <thrust/system/cuda/detail/partition.h>
+#include <thrust/system/omp/detail/partition.h>
+#include <thrust/system/tbb/detail/partition.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_PARTITION_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/partition.h>
+#include __THRUST_HOST_SYSTEM_PARTITION_HEADER
+#undef __THRUST_HOST_SYSTEM_PARTITION_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_PARTITION_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/partition.h>
+#include __THRUST_DEVICE_SYSTEM_PARTITION_HEADER
+#undef __THRUST_DEVICE_SYSTEM_PARTITION_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/per_device_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/per_device_resource.h
new file mode 100644
index 000000000000..721f49e03fd4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/per_device_resource.h
@@ -0,0 +1,41 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the per_device_resource.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch per_device_resource
+
+#include <thrust/system/detail/sequential/per_device_resource.h>
+
+#if 0
+#include <thrust/system/cpp/detail/per_device_resource.h>
+#include <thrust/system/cuda/detail/per_device_resource.h>
+#include <thrust/system/omp/detail/per_device_resource.h>
+#include <thrust/system/tbb/detail/per_device_resource.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_PER_DEVICE_RESOURCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/per_device_resource.h>
+#include __THRUST_HOST_SYSTEM_PER_DEVICE_RESOURCE_HEADER
+#undef __THRUST_HOST_SYSTEM_PER_DEVICE_RESOURCE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_PER_DEVICE_RESOURCE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/per_device_resource.h>
+#include __THRUST_DEVICE_SYSTEM_PER_DEVICE_RESOURCE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_PER_DEVICE_RESOURCE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce.h
new file mode 100644
index 000000000000..8a9673b3f957
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the reduce.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch reduce
+
+#include <thrust/system/detail/sequential/reduce.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/reduce.h>
+#include <thrust/system/cuda/detail/reduce.h>
+#include <thrust/system/omp/detail/reduce.h>
+#include <thrust/system/tbb/detail/reduce.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_REDUCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/reduce.h>
+#include __THRUST_HOST_SYSTEM_REDUCE_HEADER
+#undef __THRUST_HOST_SYSTEM_REDUCE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_REDUCE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/reduce.h>
+#include __THRUST_DEVICE_SYSTEM_REDUCE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_REDUCE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce_by_key.h
new file mode 100644
index 000000000000..0605f9befaf9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reduce_by_key.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the reduce_by_key.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch reduce_by_key
+
+#include <thrust/system/detail/sequential/reduce_by_key.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/reduce_by_key.h>
+#include <thrust/system/cuda/detail/reduce_by_key.h>
+#include <thrust/system/omp/detail/reduce_by_key.h>
+#include <thrust/system/tbb/detail/reduce_by_key.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_REDUCE_BY_KEY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/reduce_by_key.h>
+#include __THRUST_HOST_SYSTEM_REDUCE_BY_KEY_HEADER
+#undef __THRUST_HOST_SYSTEM_REDUCE_BY_KEY_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_REDUCE_BY_KEY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/reduce_by_key.h>
+#include __THRUST_DEVICE_SYSTEM_REDUCE_BY_KEY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_REDUCE_BY_KEY_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/remove.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/remove.h
new file mode 100644
index 000000000000..c281379d54ed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/remove.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the remove.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch remove
+
+#include <thrust/system/detail/sequential/remove.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/remove.h>
+#include <thrust/system/cuda/detail/remove.h>
+#include <thrust/system/omp/detail/remove.h>
+#include <thrust/system/tbb/detail/remove.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_REMOVE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/remove.h>
+#include __THRUST_HOST_SYSTEM_REMOVE_HEADER
+#undef __THRUST_HOST_SYSTEM_REMOVE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_REMOVE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/remove.h>
+#include __THRUST_DEVICE_SYSTEM_REMOVE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_REMOVE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/replace.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/replace.h
new file mode 100644
index 000000000000..d8fb5746f1ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/replace.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the replace.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch replace
+
+#include <thrust/system/detail/sequential/replace.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/replace.h>
+#include <thrust/system/cuda/detail/replace.h>
+#include <thrust/system/omp/detail/replace.h>
+#include <thrust/system/tbb/detail/replace.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_REPLACE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/replace.h>
+#include __THRUST_HOST_SYSTEM_REPLACE_HEADER
+#undef __THRUST_HOST_SYSTEM_REPLACE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_REPLACE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/replace.h>
+#include __THRUST_DEVICE_SYSTEM_REPLACE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_REPLACE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reverse.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reverse.h
new file mode 100644
index 000000000000..f6bd8947ee4b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/reverse.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the reverse.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch reverse
+
+#include <thrust/system/detail/sequential/reverse.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/reverse.h>
+#include <thrust/system/cuda/detail/reverse.h>
+#include <thrust/system/omp/detail/reverse.h>
+#include <thrust/system/tbb/detail/reverse.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_REVERSE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/reverse.h>
+#include __THRUST_HOST_SYSTEM_REVERSE_HEADER
+#undef __THRUST_HOST_SYSTEM_REVERSE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_REVERSE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/reverse.h>
+#include __THRUST_DEVICE_SYSTEM_REVERSE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_REVERSE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan.h
new file mode 100644
index 000000000000..a24910410589
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the scan.h header
+// of the host and device systems. It should be #included in any
+// code which uses adl to dispatch scan
+
+#include <thrust/system/detail/sequential/scan.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/scan.h>
+#include <thrust/system/cuda/detail/scan.h>
+#include <thrust/system/omp/detail/scan.h>
+#include <thrust/system/tbb/detail/scan.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SCAN_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/scan.h>
+#include __THRUST_HOST_SYSTEM_SCAN_HEADER
+#undef __THRUST_HOST_SYSTEM_SCAN_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SCAN_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/scan.h>
+#include __THRUST_DEVICE_SYSTEM_SCAN_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SCAN_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan_by_key.h
new file mode 100644
index 000000000000..94f73503cac5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scan_by_key.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the scan_by_key.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch scan_by_key
+
+#include <thrust/system/detail/sequential/scan_by_key.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/scan_by_key.h>
+#include <thrust/system/cuda/detail/scan_by_key.h>
+#include <thrust/system/omp/detail/scan_by_key.h>
+#include <thrust/system/tbb/detail/scan_by_key.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SCAN_BY_KEY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/scan_by_key.h>
+#include __THRUST_HOST_SYSTEM_SCAN_BY_KEY_HEADER
+#undef __THRUST_HOST_SYSTEM_SCAN_BY_KEY_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SCAN_BY_KEY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/scan_by_key.h>
+#include __THRUST_DEVICE_SYSTEM_SCAN_BY_KEY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SCAN_BY_KEY_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scatter.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scatter.h
new file mode 100644
index 000000000000..d9f42b28b13c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/scatter.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the scatter.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch scatter
+
+#include <thrust/system/detail/sequential/scatter.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/scatter.h>
+#include <thrust/system/cuda/detail/scatter.h>
+#include <thrust/system/omp/detail/scatter.h>
+#include <thrust/system/tbb/detail/scatter.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SCATTER_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/scatter.h>
+#include __THRUST_HOST_SYSTEM_SCATTER_HEADER
+#undef __THRUST_HOST_SYSTEM_SCATTER_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SCATTER_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/scatter.h>
+#include __THRUST_DEVICE_SYSTEM_SCATTER_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SCATTER_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sequence.h
new file mode 100644
index 000000000000..d3c2a20f47e6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sequence.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the sequence.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch sequence
+
+#include <thrust/system/detail/sequential/sequence.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/sequence.h>
+#include <thrust/system/cuda/detail/sequence.h>
+#include <thrust/system/omp/detail/sequence.h>
+#include <thrust/system/tbb/detail/sequence.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SEQUENCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/sequence.h>
+#include __THRUST_HOST_SYSTEM_SEQUENCE_HEADER
+#undef __THRUST_HOST_SYSTEM_SEQUENCE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SEQUENCE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/sequence.h>
+#include __THRUST_DEVICE_SYSTEM_SEQUENCE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SEQUENCE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/set_operations.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/set_operations.h
new file mode 100644
index 000000000000..7d09355e12ad
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/set_operations.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the set_operations.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch set_operations
+
+#include <thrust/system/detail/sequential/set_operations.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/set_operations.h>
+#include <thrust/system/cuda/detail/set_operations.h>
+#include <thrust/system/omp/detail/set_operations.h>
+#include <thrust/system/tbb/detail/set_operations.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SET_OPERATIONS_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/set_operations.h>
+#include __THRUST_HOST_SYSTEM_SET_OPERATIONS_HEADER
+#undef __THRUST_HOST_SYSTEM_SET_OPERATIONS_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SET_OPERATIONS_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/set_operations.h>
+#include __THRUST_DEVICE_SYSTEM_SET_OPERATIONS_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SET_OPERATIONS_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sort.h
new file mode 100644
index 000000000000..1f6118c90bf6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/sort.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the sort.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch sort
+
+#include <thrust/system/detail/sequential/sort.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/sort.h>
+#include <thrust/system/cuda/detail/sort.h>
+#include <thrust/system/omp/detail/sort.h>
+#include <thrust/system/tbb/detail/sort.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SORT_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/sort.h>
+#include __THRUST_HOST_SYSTEM_SORT_HEADER
+#undef __THRUST_HOST_SYSTEM_SORT_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SORT_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/sort.h>
+#include __THRUST_DEVICE_SYSTEM_SORT_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SORT_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/swap_ranges.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/swap_ranges.h
new file mode 100644
index 000000000000..1ca3719d96c1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/swap_ranges.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the swap_ranges.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch swap_ranges
+
+#include <thrust/system/detail/sequential/swap_ranges.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/swap_ranges.h>
+#include <thrust/system/cuda/detail/swap_ranges.h>
+#include <thrust/system/omp/detail/swap_ranges.h>
+#include <thrust/system/tbb/detail/swap_ranges.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_SWAP_RANGES_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/swap_ranges.h>
+#include __THRUST_HOST_SYSTEM_SWAP_RANGES_HEADER
+#undef __THRUST_HOST_SYSTEM_SWAP_RANGES_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_SWAP_RANGES_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/swap_ranges.h>
+#include __THRUST_DEVICE_SYSTEM_SWAP_RANGES_HEADER
+#undef __THRUST_DEVICE_SYSTEM_SWAP_RANGES_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/tabulate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/tabulate.h
new file mode 100644
index 000000000000..6ae2b22a5cbd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/tabulate.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the tabulate.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch tabulate
+
+#include <thrust/system/detail/sequential/tabulate.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/tabulate.h>
+#include <thrust/system/cuda/detail/tabulate.h>
+#include <thrust/system/omp/detail/tabulate.h>
+#include <thrust/system/tbb/detail/tabulate.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_TABULATE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/tabulate.h>
+#include __THRUST_HOST_SYSTEM_TABULATE_HEADER
+#undef __THRUST_HOST_SYSTEM_TABULATE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_TABULATE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/tabulate.h>
+#include __THRUST_DEVICE_SYSTEM_TABULATE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_TABULATE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/temporary_buffer.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/temporary_buffer.h
new file mode 100644
index 000000000000..0cada5ee4b10
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/temporary_buffer.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the temporary_buffer.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch get_temporary_buffer or return_temporary_buffer
+
+#include <thrust/system/detail/sequential/temporary_buffer.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/temporary_buffer.h>
+#include <thrust/system/cuda/detail/temporary_buffer.h>
+#include <thrust/system/omp/detail/temporary_buffer.h>
+#include <thrust/system/tbb/detail/temporary_buffer.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_TEMPORARY_BUFFER_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/temporary_buffer.h>
+#include __THRUST_HOST_SYSTEM_TEMPORARY_BUFFER_HEADER
+#undef __THRUST_HOST_SYSTEM_TEMPORARY_BUFFER_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_TEMPORARY_BUFFER_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/temporary_buffer.h>
+#include __THRUST_DEVICE_SYSTEM_TEMPORARY_BUFFER_HEADER
+#undef __THRUST_DEVICE_SYSTEM_TEMPORARY_BUFFER_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform.h
new file mode 100644
index 000000000000..b70333093fd4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the transform.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch transform
+
+#include <thrust/system/detail/sequential/transform.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/transform.h>
+#include <thrust/system/cuda/detail/transform.h>
+#include <thrust/system/omp/detail/transform.h>
+#include <thrust/system/tbb/detail/transform.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_TRANSFORM_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/transform.h>
+#include __THRUST_HOST_SYSTEM_TRANSFORM_HEADER
+#undef __THRUST_HOST_SYSTEM_TRANSFORM_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_TRANSFORM_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/transform.h>
+#include __THRUST_DEVICE_SYSTEM_TRANSFORM_HEADER
+#undef __THRUST_DEVICE_SYSTEM_TRANSFORM_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_reduce.h
new file mode 100644
index 000000000000..e3f9494dfa6e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_reduce.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the transform_reduce.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch transform_reduce
+
+#include <thrust/system/detail/sequential/transform_reduce.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/transform_reduce.h>
+#include <thrust/system/cuda/detail/transform_reduce.h>
+#include <thrust/system/omp/detail/transform_reduce.h>
+#include <thrust/system/tbb/detail/transform_reduce.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_TRANSFORM_REDUCE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/transform_reduce.h>
+#include __THRUST_HOST_SYSTEM_TRANSFORM_REDUCE_HEADER
+#undef __THRUST_HOST_SYSTEM_TRANSFORM_REDUCE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_TRANSFORM_REDUCE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/transform_reduce.h>
+#include __THRUST_DEVICE_SYSTEM_TRANSFORM_REDUCE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_TRANSFORM_REDUCE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_scan.h
new file mode 100644
index 000000000000..3a05c7eeed9a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/transform_scan.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the transform_scan.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch transform_scan
+
+#include <thrust/system/detail/sequential/transform_scan.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/transform_scan.h>
+#include <thrust/system/cuda/detail/transform_scan.h>
+#include <thrust/system/omp/detail/transform_scan.h>
+#include <thrust/system/tbb/detail/transform_scan.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_TRANSFORM_SCAN_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/transform_scan.h>
+#include __THRUST_HOST_SYSTEM_TRANSFORM_SCAN_HEADER
+#undef __THRUST_HOST_SYSTEM_TRANSFORM_SCAN_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_TRANSFORM_SCAN_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/transform_scan.h>
+#include __THRUST_DEVICE_SYSTEM_TRANSFORM_SCAN_HEADER
+#undef __THRUST_DEVICE_SYSTEM_TRANSFORM_SCAN_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_copy.h
new file mode 100644
index 000000000000..a13b18aa8d73
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_copy.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the uninitialized_copy.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch uninitialized_copy
+
+#include <thrust/system/detail/sequential/uninitialized_copy.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/uninitialized_copy.h>
+#include <thrust/system/cuda/detail/uninitialized_copy.h>
+#include <thrust/system/omp/detail/uninitialized_copy.h>
+#include <thrust/system/tbb/detail/uninitialized_copy.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_UNINITIALIZED_COPY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/uninitialized_copy.h>
+#include __THRUST_HOST_SYSTEM_UNINITIALIZED_COPY_HEADER
+#undef __THRUST_HOST_SYSTEM_UNINITIALIZED_COPY_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_UNINITIALIZED_COPY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/uninitialized_copy.h>
+#include __THRUST_DEVICE_SYSTEM_UNINITIALIZED_COPY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_UNINITIALIZED_COPY_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_fill.h
new file mode 100644
index 000000000000..98b57836e124
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/uninitialized_fill.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the uninitialized_fill.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch uninitialized_fill
+
+#include <thrust/system/detail/sequential/uninitialized_fill.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/uninitialized_fill.h>
+#include <thrust/system/cuda/detail/uninitialized_fill.h>
+#include <thrust/system/omp/detail/uninitialized_fill.h>
+#include <thrust/system/tbb/detail/uninitialized_fill.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_UNINITIALIZED_FILL_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/uninitialized_fill.h>
+#include __THRUST_HOST_SYSTEM_UNINITIALIZED_FILL_HEADER
+#undef __THRUST_HOST_SYSTEM_UNINITIALIZED_FILL_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_UNINITIALIZED_FILL_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/uninitialized_fill.h>
+#include __THRUST_DEVICE_SYSTEM_UNINITIALIZED_FILL_HEADER
+#undef __THRUST_DEVICE_SYSTEM_UNINITIALIZED_FILL_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique.h
new file mode 100644
index 000000000000..4082f5299269
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the unique.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch unique
+
+#include <thrust/system/detail/sequential/unique.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/unique.h>
+#include <thrust/system/cuda/detail/unique.h>
+#include <thrust/system/omp/detail/unique.h>
+#include <thrust/system/tbb/detail/unique.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_UNIQUE_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/unique.h>
+#include __THRUST_HOST_SYSTEM_UNIQUE_HEADER
+#undef __THRUST_HOST_SYSTEM_UNIQUE_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_UNIQUE_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/unique.h>
+#include __THRUST_DEVICE_SYSTEM_UNIQUE_HEADER
+#undef __THRUST_DEVICE_SYSTEM_UNIQUE_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique_by_key.h
new file mode 100644
index 000000000000..dcf9acd42cd7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/adl/unique_by_key.h
@@ -0,0 +1,44 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a fill of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// the purpose of this header is to #include the unique_by_key.h header
+// of the sequential, host, and device systems. It should be #included in any
+// code which uses adl to dispatch unique_by_key
+
+#include <thrust/system/detail/sequential/unique_by_key.h>
+
+// SCons can't see through the #defines below to figure out what this header
+// includes, so we fake it out by specifying all possible files we might end up
+// including inside an #if 0.
+#if 0
+#include <thrust/system/cpp/detail/unique_by_key.h>
+#include <thrust/system/cuda/detail/unique_by_key.h>
+#include <thrust/system/omp/detail/unique_by_key.h>
+#include <thrust/system/tbb/detail/unique_by_key.h>
+#endif
+
+#define __THRUST_HOST_SYSTEM_UNIQUE_BY_KEY_HEADER <__THRUST_HOST_SYSTEM_ROOT/detail/unique_by_key.h>
+#include __THRUST_HOST_SYSTEM_UNIQUE_BY_KEY_HEADER
+#undef __THRUST_HOST_SYSTEM_UNIQUE_BY_KEY_HEADER
+
+#define __THRUST_DEVICE_SYSTEM_UNIQUE_BY_KEY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/detail/unique_by_key.h>
+#include __THRUST_DEVICE_SYSTEM_UNIQUE_BY_KEY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_UNIQUE_BY_KEY_HEADER
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/bad_alloc.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/bad_alloc.h
new file mode 100644
index 000000000000..ae5dd5994ef5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/bad_alloc.h
@@ -0,0 +1,58 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <new>
+#include <string>
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+
+// define our own bad_alloc so we can set its .what()
+class bad_alloc
+  : public std::bad_alloc
+{
+  public:
+    inline bad_alloc(const std::string &w)
+      : std::bad_alloc(), m_what()
+    {
+      m_what = std::bad_alloc::what();
+      m_what += ": ";
+      m_what += w;
+    } // end bad_alloc()
+
+    inline virtual ~bad_alloc(void) noexcept {};
+
+    inline virtual const char *what(void) const noexcept
+    {
+      return m_what.c_str();
+    } // end what()
+
+  private:
+    std::string m_what;
+}; // end bad_alloc
+  
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/errno.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/errno.h
new file mode 100644
index 000000000000..69cb2bd98062
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/errno.h
@@ -0,0 +1,119 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// The rationale for the existence of these apparently redundant definitions is
+// to provide them portably and to avoid bringing in system headers which might
+// pollute the global namespace. These identifiers are in lowercase to avoid
+// colliding with the real macros in errno.h.
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+namespace detail
+{
+
+static const int eafnosupport    = 9901;
+static const int eaddrinuse      = 9902;
+static const int eaddrnotavail   = 9903;
+static const int eisconn         = 9904;
+static const int ebadmsg         = 9905;
+static const int econnaborted    = 9906;
+static const int ealready        = 9907;
+static const int econnrefused    = 9908;
+static const int econnreset      = 9909;
+static const int edestaddrreq    = 9910;
+static const int ehostunreach    = 9911;
+static const int eidrm           = 9912;
+static const int emsgsize        = 9913;
+static const int enetdown        = 9914;
+static const int enetreset       = 9915;
+static const int enetunreach     = 9916;
+static const int enobufs         = 9917;
+static const int enolink         = 9918;
+static const int enodata         = 9919;
+static const int enomsg          = 9920;
+static const int enoprotoopt     = 9921;
+static const int enosr           = 9922;
+static const int enotsock        = 9923;
+static const int enostr          = 9924;
+static const int enotconn        = 9925;
+static const int enotsup         = 9926;
+static const int ecanceled       = 9927;
+static const int einprogress     = 9928;
+static const int eopnotsupp      = 9929;
+static const int ewouldblock     = 9930;
+static const int eownerdead      = 9931;
+static const int eproto          = 9932;
+static const int eprotonosupport = 9933;
+static const int enotrecoverable = 9934;
+static const int etime           = 9935;
+static const int etxtbsy         = 9936;
+static const int etimedout       = 9938;
+static const int eloop           = 9939;
+static const int eoverflow       = 9940;
+static const int eprototype      = 9941;
+static const int enosys          = 9942;
+static const int einval          = 9943;
+static const int erange          = 9944;
+static const int eilseq          = 9945;
+static const int e2big           = 9946;
+static const int edom            = 9947;
+static const int efault          = 9948;
+static const int ebadf           = 9949;
+static const int epipe           = 9950;
+static const int exdev           = 9951;
+static const int ebusy           = 9952;
+static const int enotempty       = 9953;
+static const int enoexec         = 9954;
+static const int eexist          = 9955;
+static const int efbig           = 9956;
+static const int enametoolong    = 9957;
+static const int enotty          = 9958;
+static const int eintr           = 9959;
+static const int espipe          = 9960;
+static const int eio             = 9961;
+static const int eisdir          = 9962;
+static const int echild          = 9963;
+static const int enolck          = 9964;
+static const int enospc          = 9965;
+static const int enxio           = 9966;
+static const int enodev          = 9967;
+static const int enoent          = 9968;
+static const int esrch           = 9969;
+static const int enotdir         = 9970;
+static const int enomem          = 9971;
+static const int eperm           = 9972;
+static const int eacces          = 9973;
+static const int erofs           = 9974;
+static const int edeadlk         = 9975;
+static const int eagain          = 9976;
+static const int enfile          = 9977;
+static const int emfile          = 9978;
+static const int emlink          = 9979;
+
+} // end detail
+
+} // end system
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_category.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_category.inl
new file mode 100644
index 000000000000..45fd15a3fd53
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_category.inl
@@ -0,0 +1,237 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/system/error_code.h>
+#include <thrust/system/detail/errno.h>
+#include <thrust/functional.h>
+#include <cstring>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+error_category
+  ::~error_category(void)
+{
+  ;
+} // end error_category::~error_category()
+
+
+error_condition error_category
+  ::default_error_condition(int ev) const
+{
+  return error_condition(ev, *this);
+} // end error_category::default_error_condition()
+
+
+bool error_category
+  ::equivalent(int code, const error_condition &condition) const
+{
+  return default_error_condition(code) == condition;
+} // end error_condition::equivalent()
+
+
+bool error_category
+  ::equivalent(const error_code &code, int condition) const
+{
+  bool result = (this->operator==(code.category())) && (code.value() == condition);
+  return result;
+} // end error_code::equivalent()
+
+
+bool error_category
+  ::operator==(const error_category &rhs) const
+{
+  return this == &rhs;
+} // end error_category::operator==()
+
+
+bool error_category
+  ::operator!=(const error_category &rhs) const
+{
+  return !this->operator==(rhs);
+} // end error_category::operator!=()
+
+
+bool error_category
+  ::operator<(const error_category &rhs) const
+{
+  return thrust::less<const error_category*>()(this,&rhs);
+} // end error_category::operator<()
+
+
+namespace detail
+{
+
+
+class generic_error_category
+  : public error_category
+{
+  public:
+    inline generic_error_category(void) {}
+
+    inline virtual const char *name(void) const
+    {
+      return "generic";
+    }
+
+    inline virtual std::string message(int ev) const
+    {
+      static const std::string unknown_err("Unknown error");
+
+      // XXX strerror is not thread-safe:
+      //     prefer strerror_r (which is not provided on windows)
+      THRUST_DISABLE_MSVC_WARNING_BEGIN(4996)
+      const char *c_str = std::strerror(ev);
+      THRUST_DISABLE_MSVC_WARNING_END(4996)
+      return c_str ? std::string(c_str) : unknown_err;
+    }
+}; // end generic_category_result
+
+
+class system_error_category
+  : public error_category
+{
+  public:
+    inline system_error_category(void) {}
+
+    inline virtual const char *name(void) const
+    {
+      return "system";
+    }
+
+    inline virtual std::string message(int ev) const
+    {
+      return generic_category().message(ev);
+    }
+
+    inline virtual error_condition default_error_condition(int ev) const
+    {
+      using namespace errc;
+
+      switch(ev)
+      {
+        case eafnosupport:    return make_error_condition(address_family_not_supported);
+        case eaddrinuse:      return make_error_condition(address_in_use);
+        case eaddrnotavail:   return make_error_condition(address_not_available);
+        case eisconn:         return make_error_condition(already_connected);
+        case e2big:           return make_error_condition(argument_list_too_long);
+        case edom:            return make_error_condition(argument_out_of_domain);
+        case efault:          return make_error_condition(bad_address);
+        case ebadf:           return make_error_condition(bad_file_descriptor);
+        case ebadmsg:         return make_error_condition(bad_message);
+        case epipe:           return make_error_condition(broken_pipe);
+        case econnaborted:    return make_error_condition(connection_aborted);
+        case ealready:        return make_error_condition(connection_already_in_progress);
+        case econnrefused:    return make_error_condition(connection_refused);
+        case econnreset:      return make_error_condition(connection_reset);
+        case exdev:           return make_error_condition(cross_device_link);
+        case edestaddrreq:    return make_error_condition(destination_address_required);
+        case ebusy:           return make_error_condition(device_or_resource_busy);
+        case enotempty:       return make_error_condition(directory_not_empty);
+        case enoexec:         return make_error_condition(executable_format_error);
+        case eexist:          return make_error_condition(file_exists);
+        case efbig:           return make_error_condition(file_too_large);
+        case enametoolong:    return make_error_condition(filename_too_long);
+        case enosys:          return make_error_condition(function_not_supported);
+        case ehostunreach:    return make_error_condition(host_unreachable);
+        case eidrm:           return make_error_condition(identifier_removed);
+        case eilseq:          return make_error_condition(illegal_byte_sequence);
+        case enotty:          return make_error_condition(inappropriate_io_control_operation);
+        case eintr:           return make_error_condition(interrupted);
+        case einval:          return make_error_condition(invalid_argument);
+        case espipe:          return make_error_condition(invalid_seek);
+        case eio:             return make_error_condition(io_error);
+        case eisdir:          return make_error_condition(is_a_directory);
+        case emsgsize:        return make_error_condition(message_size);
+        case enetdown:        return make_error_condition(network_down);
+        case enetreset:       return make_error_condition(network_reset);
+        case enetunreach:     return make_error_condition(network_unreachable);
+        case enobufs:         return make_error_condition(no_buffer_space);
+        case echild:          return make_error_condition(no_child_process);
+        case enolink:         return make_error_condition(no_link);
+        case enolck:          return make_error_condition(no_lock_available);
+        case enodata:         return make_error_condition(no_message_available);
+        case enomsg:          return make_error_condition(no_message);
+        case enoprotoopt:     return make_error_condition(no_protocol_option);
+        case enospc:          return make_error_condition(no_space_on_device);
+        case enosr:           return make_error_condition(no_stream_resources);
+        case enxio:           return make_error_condition(no_such_device_or_address);
+        case enodev:          return make_error_condition(no_such_device);
+        case enoent:          return make_error_condition(no_such_file_or_directory);
+        case esrch:           return make_error_condition(no_such_process);
+        case enotdir:         return make_error_condition(not_a_directory);
+        case enotsock:        return make_error_condition(not_a_socket);
+        case enostr:          return make_error_condition(not_a_stream);
+        case enotconn:        return make_error_condition(not_connected);
+        case enomem:          return make_error_condition(not_enough_memory);
+        case enotsup:         return make_error_condition(not_supported);
+        case ecanceled:       return make_error_condition(operation_canceled);
+        case einprogress:     return make_error_condition(operation_in_progress);
+        case eperm:           return make_error_condition(operation_not_permitted);
+        case eopnotsupp:      return make_error_condition(operation_not_supported);
+        case ewouldblock:     return make_error_condition(operation_would_block);
+        case eownerdead:      return make_error_condition(owner_dead);
+        case eacces:          return make_error_condition(permission_denied);
+        case eproto:          return make_error_condition(protocol_error);
+        case eprotonosupport: return make_error_condition(protocol_not_supported);
+        case erofs:           return make_error_condition(read_only_file_system);
+        case edeadlk:         return make_error_condition(resource_deadlock_would_occur);
+        case eagain:          return make_error_condition(resource_unavailable_try_again);
+        case erange:          return make_error_condition(result_out_of_range);
+        case enotrecoverable: return make_error_condition(state_not_recoverable);
+        case etime:           return make_error_condition(stream_timeout);
+        case etxtbsy:         return make_error_condition(text_file_busy);
+        case etimedout:       return make_error_condition(timed_out);
+        case enfile:          return make_error_condition(too_many_files_open_in_system);
+        case emfile:          return make_error_condition(too_many_files_open);
+        case emlink:          return make_error_condition(too_many_links);
+        case eloop:           return make_error_condition(too_many_symbolic_link_levels);
+        case eoverflow:       return make_error_condition(value_too_large);
+        case eprototype:      return make_error_condition(wrong_protocol_type);
+        default:              return error_condition(ev,system_category());
+      }
+    }
+}; // end system_category_result
+
+
+} // end detail
+
+
+const error_category &generic_category(void)
+{
+  static const detail::generic_error_category result;
+  return result;
+}
+
+
+const error_category &system_category(void)
+{
+  static const detail::system_error_category result;
+  return result;
+}
+
+
+} // end system
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_code.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_code.inl
new file mode 100644
index 000000000000..2b819c0485e9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_code.inl
@@ -0,0 +1,198 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/system/error_code.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+error_code
+  ::error_code(void)
+    :m_val(0),m_cat(&system_category())
+{
+  ;
+} // end error_code::error_code()
+
+
+error_code
+  ::error_code(int val, const error_category &cat)
+    :m_val(val),m_cat(&cat)
+{
+  ;
+} // end error_code::error_code()
+
+
+template <typename ErrorCodeEnum>
+  error_code
+    ::error_code(ErrorCodeEnum e
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+                 , typename thrust::detail::enable_if<is_error_code_enum<ErrorCodeEnum>::value>::type *
+#endif // THRUST_HOST_COMPILER_MSVC
+                )
+{
+  *this = make_error_code(e);
+} // end error_code::error_code()
+
+
+void error_code
+  ::assign(int val, const error_category &cat)
+{
+  m_val = val;
+  m_cat = &cat;
+} // end error_code::assign()
+
+
+template <typename ErrorCodeEnum>
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+  typename thrust::detail::enable_if<is_error_code_enum<ErrorCodeEnum>::value, error_code>::type &
+#else
+  error_code &
+#endif // THRUST_HOST_COMPILER_MSVC
+    error_code
+      ::operator=(ErrorCodeEnum e)
+{
+  *this = make_error_code(e);
+  return *this;
+} // end error_code::operator=()
+
+
+void error_code
+  ::clear(void)
+{
+  m_val = 0;
+  m_cat = &system_category();
+} // end error_code::clear()
+
+
+int error_code
+  ::value(void) const
+{
+  return m_val;
+} // end error_code::value()
+
+
+const error_category &error_code
+  ::category(void) const
+{
+  return *m_cat;
+} // end error_code::category()
+
+
+error_condition error_code
+  ::default_error_condition(void) const
+{
+  return category().default_error_condition(value());
+} // end error_code::default_error_condition()
+
+
+std::string error_code
+  ::message(void) const
+{
+  return category().message(value());
+} // end error_code::message()
+
+
+error_code
+  ::operator bool (void) const
+{
+  return value() != 0;
+} // end error_code::operator bool ()
+
+
+error_code make_error_code(errc::errc_t e)
+{
+  return error_code(static_cast<int>(e), generic_category());
+} // end make_error_code()
+
+
+bool operator<(const error_code &lhs, const error_code &rhs)
+{
+  bool result = lhs.category().operator<(rhs.category());
+  result = result || lhs.category().operator==(rhs.category());
+  result = result || lhs.value() < rhs.value();
+  return result;
+} // end operator==()
+
+
+template<typename charT, typename traits>
+  std::basic_ostream<charT,traits>&
+    operator<<(std::basic_ostream<charT,traits> &os, const error_code &ec)
+{
+  return os << ec.category().name() << ':' << ec.value();
+} // end operator<<()
+
+
+bool operator==(const error_code &lhs, const error_code &rhs)
+{
+  return lhs.category().operator==(rhs.category()) && lhs.value() == rhs.value();
+} // end operator==()
+
+
+bool operator==(const error_code &lhs, const error_condition &rhs)
+{
+  return lhs.category().equivalent(lhs.value(), rhs) || rhs.category().equivalent(lhs,rhs.value());
+} // end operator==()
+
+
+bool operator==(const error_condition &lhs, const error_code &rhs)
+{
+  return rhs.category().equivalent(lhs.value(), lhs) || lhs.category().equivalent(rhs, lhs.value());
+} // end operator==()
+
+
+bool operator==(const error_condition &lhs, const error_condition &rhs)
+{
+  return lhs.category().operator==(rhs.category()) && lhs.value() == rhs.value();
+} // end operator==()
+
+
+bool operator!=(const error_code &lhs, const error_code &rhs)
+{
+  return !(lhs == rhs);
+} // end operator!=()
+
+
+bool operator!=(const error_code &lhs, const error_condition &rhs)
+{
+  return !(lhs == rhs);
+} // end operator!=()
+
+
+bool operator!=(const error_condition &lhs, const error_code &rhs)
+{
+  return !(lhs == rhs);
+} // end operator!=()
+
+
+bool operator!=(const error_condition &lhs, const error_condition &rhs)
+{
+  return !(lhs == rhs);
+} // end operator!=()
+
+
+} // end system
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_condition.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_condition.inl
new file mode 100644
index 000000000000..0daf1f2938b9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/error_condition.inl
@@ -0,0 +1,134 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/system/detail/error_condition.inl>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+error_condition
+  ::error_condition(void)
+    :m_val(0),m_cat(&generic_category())
+{
+  ;
+} // end error_condition::error_condition()
+
+
+error_condition
+  ::error_condition(int val, const error_category &cat)
+    :m_val(val),m_cat(&cat)
+{
+  ;
+} // end error_condition::error_condition()
+
+
+template<typename ErrorConditionEnum>
+  error_condition
+    ::error_condition(ErrorConditionEnum e
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+                      , typename thrust::detail::enable_if<is_error_condition_enum<ErrorConditionEnum>::value>::type *
+#endif // THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+                     )
+{
+  *this = make_error_condition(e);
+} // end error_condition::error_condition()
+
+
+void error_condition
+  ::assign(int val, const error_category &cat)
+{
+  m_val = val;
+  m_cat = &cat;
+} // end error_category::assign()
+
+
+template<typename ErrorConditionEnum>
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+  typename thrust::detail::enable_if<is_error_condition_enum<ErrorConditionEnum>::value, error_condition>::type &
+#else
+  error_condition &
+#endif // THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+    error_condition
+      ::operator=(ErrorConditionEnum e)
+{
+  *this = make_error_condition(e);
+  return *this;
+} // end error_condition::operator=()
+
+
+void error_condition
+  ::clear(void)
+{
+  m_val = 0;
+  m_cat = &generic_category();
+} // end error_condition::clear()
+
+
+int error_condition
+  ::value(void) const
+{
+  return m_val;
+} // end error_condition::value()
+
+
+const error_category &error_condition
+  ::category(void) const
+{
+  return *m_cat;
+} // end error_condition::category()
+
+
+std::string error_condition
+  ::message(void) const
+{
+  return category().message(value());
+} // end error_condition::message()
+
+
+error_condition
+  ::operator bool (void) const
+{
+  return value() != 0;
+} // end error_condition::operator bool ()
+
+
+error_condition make_error_condition(errc::errc_t e)
+{
+  return error_condition(static_cast<int>(e), generic_category());
+} // end make_error_condition()
+
+
+bool operator<(const error_condition &lhs,
+               const error_condition &rhs)
+{
+  return lhs.category().operator<(rhs.category()) || (lhs.category().operator==(rhs.category()) && (lhs.value() < rhs.value()));
+} // end operator<()
+
+
+} // end system
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.h
new file mode 100644
index 000000000000..43592e15b033
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.h
@@ -0,0 +1,57 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file adjacent_difference.h
+ *  \brief Generic implementation of adjacent_difference.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator adjacent_difference(thrust::execution_policy<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result);
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename BinaryFunction>
+__host__ __device__
+OutputIterator adjacent_difference(thrust::execution_policy<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/adjacent_difference.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.inl
new file mode 100644
index 000000000000..504129328884
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/adjacent_difference.inl
@@ -0,0 +1,82 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/adjacent_difference.h>
+#include <thrust/adjacent_difference.h>
+#include <thrust/functional.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/transform.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator adjacent_difference(thrust::execution_policy<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result)
+{
+  typedef typename thrust::iterator_traits<InputIterator>::value_type InputType;
+  thrust::minus<InputType> binary_op;
+
+  return thrust::adjacent_difference(exec, first, last, result, binary_op);
+} // end adjacent_difference()
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename BinaryFunction>
+__host__ __device__
+OutputIterator adjacent_difference(thrust::execution_policy<DerivedPolicy> &exec,
+                                   InputIterator first, InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op)
+{
+  typedef typename thrust::iterator_traits<InputIterator>::value_type InputType;
+
+  if(first == last)
+  {
+    // empty range, nothing to do
+    return result;
+  }
+  else
+  {
+    // an in-place operation is requested, copy the input and call the entry point
+    // XXX a special-purpose kernel would be faster here since
+    // only block boundaries need to be copied
+    thrust::detail::temporary_array<InputType, DerivedPolicy> input_copy(exec, first, last);
+
+    *result = *first;
+    thrust::transform(exec, input_copy.begin() + 1, input_copy.end(), input_copy.begin(), result + 1, binary_op);
+  }
+
+  return result + (last - first);
+}
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.h
new file mode 100644
index 000000000000..4d6562e00fac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.h
@@ -0,0 +1,40 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename InputIterator, typename Distance>
+__host__ __device__
+void advance(InputIterator& i, Distance n);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/advance.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.inl
new file mode 100644
index 000000000000..21555ebb03f4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/advance.inl
@@ -0,0 +1,68 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/advance.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+__thrust_exec_check_disable__
+template<typename InputIterator, typename Distance>
+__host__ __device__
+void advance(InputIterator& i, Distance n, thrust::incrementable_traversal_tag)
+{
+  while(n)
+  {
+    ++i;
+    --n;
+  } // end while
+} // end advance()
+
+__thrust_exec_check_disable__
+template<typename InputIterator, typename Distance>
+__host__ __device__
+void advance(InputIterator& i, Distance n, thrust::random_access_traversal_tag)
+{
+  i += n;
+} // end advance()
+
+} // end detail
+
+template<typename InputIterator, typename Distance>
+__host__ __device__
+void advance(InputIterator& i, Distance n)
+{
+  // dispatch on iterator traversal
+  thrust::system::detail::generic::detail::advance(i, n,
+    typename thrust::iterator_traversal<InputIterator>::type());
+} // end advance()
+
+} // end namespace detail
+} // end namespace generic
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.h
new file mode 100644
index 000000000000..6603f6c302f8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.h
@@ -0,0 +1,173 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file binary_search.h
+ *  \brief Generic implementations of binary search functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec, 
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value);
+
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value, 
+                            StrictWeakOrdering comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value);
+
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec, 
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value, 
+                            StrictWeakOrdering comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+bool binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                   ForwardIterator begin,
+                   ForwardIterator end,
+                   const T& value);
+
+template <typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+bool binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                   ForwardIterator begin,
+                   ForwardIterator end,
+                   const T& value, 
+                   StrictWeakOrdering comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin, 
+                           ForwardIterator end,
+                           InputIterator values_begin, 
+                           InputIterator values_end,
+                           OutputIterator output);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin, 
+                           ForwardIterator end,
+                           InputIterator values_begin, 
+                           InputIterator values_end,
+                           OutputIterator output,
+                           StrictWeakOrdering comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin, 
+                           ForwardIterator end,
+                           InputIterator values_begin, 
+                           InputIterator values_end,
+                           OutputIterator output);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin, 
+                           ForwardIterator end,
+                           InputIterator values_begin, 
+                           InputIterator values_end,
+                           OutputIterator output,
+                           StrictWeakOrdering comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                             ForwardIterator begin, 
+                             ForwardIterator end,
+                             InputIterator values_begin, 
+                             InputIterator values_end,
+                             OutputIterator output);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                             ForwardIterator begin, 
+                             ForwardIterator end,
+                             InputIterator values_begin, 
+                             InputIterator values_end,
+                             OutputIterator output,
+                             StrictWeakOrdering comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator>
+equal_range(thrust::execution_policy<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable &value);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable, typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator>
+equal_range(thrust::execution_policy<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable &value,
+            StrictWeakOrdering comp);
+
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/binary_search.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.inl
new file mode 100644
index 000000000000..bc60bb8e5021
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/binary_search.inl
@@ -0,0 +1,396 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/distance.h>
+#include <thrust/binary_search.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/binary_search.h>
+
+#include <thrust/for_each.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/generic/scalar/binary_search.h>
+#include <thrust/system/detail/generic/select_system.h>
+
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+// XXX WAR circular #inclusion with this forward declaration
+template<typename,typename> class temporary_array;
+
+
+} // end detail
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+// short names to avoid nvcc bug
+struct lbf
+{
+  template<typename RandomAccessIterator, typename T, typename StrictWeakOrdering>
+  __host__ __device__
+  typename thrust::iterator_traits<RandomAccessIterator>::difference_type
+    operator()(RandomAccessIterator begin, RandomAccessIterator end, const T& value, StrictWeakOrdering comp)
+  {
+    return thrust::system::detail::generic::scalar::lower_bound(begin, end, value, comp) - begin;
+  }
+};
+
+
+struct ubf
+{
+  template<typename RandomAccessIterator, typename T, typename StrictWeakOrdering>
+  __host__ __device__
+  typename thrust::iterator_traits<RandomAccessIterator>::difference_type
+    operator()(RandomAccessIterator begin, RandomAccessIterator end, const T& value, StrictWeakOrdering comp)
+  {
+    return thrust::system::detail::generic::scalar::upper_bound(begin, end, value, comp) - begin;
+  }
+};
+
+
+struct bsf
+{
+  template<typename RandomAccessIterator, typename T, typename StrictWeakOrdering>
+  __host__ __device__
+  bool operator()(RandomAccessIterator begin, RandomAccessIterator end, const T& value, StrictWeakOrdering comp)
+  {
+    RandomAccessIterator iter = thrust::system::detail::generic::scalar::lower_bound(begin, end, value, comp);
+
+    thrust::detail::wrapped_function<StrictWeakOrdering,bool> wrapped_comp(comp);
+
+    return iter != end && !wrapped_comp(value, *iter);
+  }
+};
+
+
+template<typename ForwardIterator, typename StrictWeakOrdering, typename BinarySearchFunction>
+struct binary_search_functor
+{
+  ForwardIterator begin;
+  ForwardIterator end;
+  StrictWeakOrdering comp;
+  BinarySearchFunction func;
+
+  __host__ __device__
+  binary_search_functor(ForwardIterator begin, ForwardIterator end, StrictWeakOrdering comp, BinarySearchFunction func)
+    : begin(begin), end(end), comp(comp), func(func) {}
+
+  template<typename Tuple>
+  __host__ __device__
+  void operator()(Tuple t)
+  {
+    thrust::get<1>(t) = func(begin, end, thrust::get<0>(t), comp);
+  }
+}; // binary_search_functor
+
+
+// Vector Implementation
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering, typename BinarySearchFunction>
+__host__ __device__
+OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                             ForwardIterator begin,
+                             ForwardIterator end,
+                             InputIterator values_begin,
+                             InputIterator values_end,
+                             OutputIterator output,
+                             StrictWeakOrdering comp,
+                             BinarySearchFunction func)
+{
+  thrust::for_each(exec,
+                   thrust::make_zip_iterator(thrust::make_tuple(values_begin, output)),
+                   thrust::make_zip_iterator(thrust::make_tuple(values_end, output + thrust::distance(values_begin, values_end))),
+                   detail::binary_search_functor<ForwardIterator, StrictWeakOrdering, BinarySearchFunction>(begin, end, comp, func));
+
+  return output + thrust::distance(values_begin, values_end);
+}
+
+
+
+// Scalar Implementation
+template<typename OutputType, typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering, typename BinarySearchFunction>
+__host__ __device__
+OutputType binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                         ForwardIterator begin,
+                         ForwardIterator end,
+                         const T& value,
+                         StrictWeakOrdering comp,
+                         BinarySearchFunction func)
+{
+  // use the vectorized path to implement the scalar version
+
+  // allocate device buffers for value and output
+  thrust::detail::temporary_array<T,DerivedPolicy>          d_value(exec,1);
+  thrust::detail::temporary_array<OutputType,DerivedPolicy> d_output(exec,1);
+
+  { // copy value to device
+    typedef typename thrust::iterator_system<const T*>::type value_in_system_t;
+    value_in_system_t value_in_system;
+    using thrust::system::detail::generic::select_system;
+    thrust::copy_n(select_system(thrust::detail::derived_cast(thrust::detail::strip_const(value_in_system)),
+                                 thrust::detail::derived_cast(thrust::detail::strip_const(exec))),
+                   &value, 1, d_value.begin());
+  }
+
+  // perform the query
+  thrust::system::detail::generic::detail::binary_search(exec, begin, end, d_value.begin(), d_value.end(), d_output.begin(), comp, func);
+
+  OutputType output;
+  { // copy result to host and return
+    typedef typename thrust::iterator_system<OutputType*>::type result_out_system_t;
+    result_out_system_t result_out_system;
+    using thrust::system::detail::generic::select_system;
+    thrust::copy_n(select_system(thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
+                                 thrust::detail::derived_cast(thrust::detail::strip_const(result_out_system))),
+                   d_output.begin(), 1, &output);
+  }
+
+  return output;
+}
+
+
+// this functor differs from thrust::less<T>
+// because it allows the types of lhs & rhs to differ
+// which is required by the binary search functions
+// XXX use C++14 thrust::less<> when it's ready
+struct binary_search_less
+{
+  template<typename T1, typename T2>
+  __host__ __device__
+  bool operator()(const T1& lhs, const T2& rhs) const
+  {
+    return lhs < rhs;
+  }
+};
+
+
+} // end namespace detail
+
+
+//////////////////////
+// Scalar Functions //
+//////////////////////
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value)
+{
+  namespace p = thrust::placeholders;
+  return thrust::lower_bound(exec, begin, end, value, detail::binary_search_less());
+}
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value,
+                            StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::difference_type difference_type;
+
+  return begin + detail::binary_search<difference_type>(exec, begin, end, value, comp, detail::lbf());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value)
+{
+  namespace p = thrust::placeholders;
+  return thrust::upper_bound(exec, begin, end, value, detail::binary_search_less());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator begin,
+                            ForwardIterator end,
+                            const T& value,
+                            StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::difference_type difference_type;
+
+  return begin + detail::binary_search<difference_type>(exec, begin, end, value, comp, detail::ubf());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+bool binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                   ForwardIterator begin,
+                   ForwardIterator end,
+                   const T& value)
+{
+  return thrust::binary_search(exec, begin, end, value, detail::binary_search_less());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+bool binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                   ForwardIterator begin,
+                   ForwardIterator end,
+                   const T& value,
+                   StrictWeakOrdering comp)
+{
+  return detail::binary_search<bool>(exec, begin, end, value, comp, detail::bsf());
+}
+
+
+//////////////////////
+// Vector Functions //
+//////////////////////
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin,
+                           ForwardIterator end,
+                           InputIterator values_begin,
+                           InputIterator values_end,
+                           OutputIterator output)
+{
+  namespace p = thrust::placeholders;
+  return thrust::lower_bound(exec, begin, end, values_begin, values_end, output, detail::binary_search_less());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator lower_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin,
+                           ForwardIterator end,
+                           InputIterator values_begin,
+                           InputIterator values_end,
+                           OutputIterator output,
+                           StrictWeakOrdering comp)
+{
+  return detail::binary_search(exec, begin, end, values_begin, values_end, output, comp, detail::lbf());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin,
+                           ForwardIterator end,
+                           InputIterator values_begin,
+                           InputIterator values_end,
+                           OutputIterator output)
+{
+  namespace p = thrust::placeholders;
+  return thrust::upper_bound(exec, begin, end, values_begin, values_end, output, detail::binary_search_less());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator upper_bound(thrust::execution_policy<DerivedPolicy> &exec,
+                           ForwardIterator begin,
+                           ForwardIterator end,
+                           InputIterator values_begin,
+                           InputIterator values_end,
+                           OutputIterator output,
+                           StrictWeakOrdering comp)
+{
+  return detail::binary_search(exec, begin, end, values_begin, values_end, output, comp, detail::ubf());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator>
+__host__ __device__
+OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                             ForwardIterator begin,
+                             ForwardIterator end,
+                             InputIterator values_begin,
+                             InputIterator values_end,
+                             OutputIterator output)
+{
+  namespace p = thrust::placeholders;
+  return thrust::binary_search(exec, begin, end, values_begin, values_end, output, detail::binary_search_less());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename OutputIterator, typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator binary_search(thrust::execution_policy<DerivedPolicy> &exec,
+                             ForwardIterator begin,
+                             ForwardIterator end,
+                             InputIterator values_begin,
+                             InputIterator values_end,
+                             OutputIterator output,
+                             StrictWeakOrdering comp)
+{
+  return detail::binary_search(exec, begin, end, values_begin, values_end, output, comp, detail::bsf());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename LessThanComparable>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator>
+equal_range(thrust::execution_policy<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const LessThanComparable &value)
+{
+  return thrust::equal_range(exec, first, last, value, detail::binary_search_less());
+}
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T, typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator>
+equal_range(thrust::execution_policy<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const T &value,
+            StrictWeakOrdering comp)
+{
+  ForwardIterator lb = thrust::lower_bound(exec, first, last, value, comp);
+  ForwardIterator ub = thrust::upper_bound(exec, first, last, value, comp);
+  return thrust::make_pair(lb, ub);
+}
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.h
new file mode 100644
index 000000000000..36ac71899de0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.h
@@ -0,0 +1,58 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(thrust::execution_policy<DerivedPolicy> &exec,
+                      InputIterator  first,
+                      InputIterator  last,
+                      OutputIterator result);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(thrust::execution_policy<DerivedPolicy> &exec,
+                        InputIterator  first,
+                        Size           n,
+                        OutputIterator result);
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/copy.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.inl
new file mode 100644
index 000000000000..34d66baa6b2c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy.inl
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/copy.h>
+#include <thrust/functional.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/transform.h>
+#include <thrust/for_each.h>
+#include <thrust/tuple.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/iterator/detail/minimum_system.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(thrust::execution_policy<DerivedPolicy> &exec,
+                      InputIterator                            first,
+                      InputIterator                            last,
+                      OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator>::type T;
+  return thrust::transform(exec, first, last, result, thrust::identity<T>());
+} // end copy()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(thrust::execution_policy<DerivedPolicy> &exec,
+                        InputIterator                            first,
+                        Size                                     n,
+                        OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator>::type value_type;
+  typedef thrust::identity<value_type>                         xfrm_type;
+
+  typedef thrust::detail::unary_transform_functor<xfrm_type> functor_type;
+
+  typedef thrust::tuple<InputIterator,OutputIterator> iterator_tuple;
+  typedef thrust::zip_iterator<iterator_tuple>        zip_iter;
+
+  zip_iter zipped = thrust::make_zip_iterator(thrust::make_tuple(first,result));
+
+  return thrust::get<1>(thrust::for_each_n(exec, zipped, n, functor_type(xfrm_type())).get_iterator_tuple());
+} // end copy_n()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.h
new file mode 100644
index 000000000000..6a13edfdaf30
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.h
@@ -0,0 +1,63 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                         InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+   OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator result,
+                          Predicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/copy_if.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.inl
new file mode 100644
index 000000000000..5a6edd72e186
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/copy_if.inl
@@ -0,0 +1,160 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/copy_if.h>
+#include <thrust/detail/copy_if.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/functional.h>
+#include <thrust/distance.h>
+#include <thrust/transform.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/integer_traits.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/scan.h>
+#include <thrust/scatter.h>
+#include <limits>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+template<typename IndexType,
+         typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                       InputIterator1 first,
+                       InputIterator1 last,
+                       InputIterator2 stencil,
+                       OutputIterator result,
+                       Predicate pred)
+{
+  THRUST_DISABLE_MSVC_POSSIBLE_LOSS_OF_DATA_WARNING(IndexType n = thrust::distance(first, last));
+  
+  // compute {0,1} predicates
+  thrust::detail::temporary_array<IndexType, DerivedPolicy> predicates(exec, n);
+  thrust::transform(exec,
+                    stencil,
+                    stencil + n,
+                    predicates.begin(),
+                    thrust::detail::predicate_to_integral<Predicate,IndexType>(pred));
+  
+  // scan {0,1} predicates
+  thrust::detail::temporary_array<IndexType, DerivedPolicy> scatter_indices(exec, n);
+  thrust::exclusive_scan(exec,
+                         predicates.begin(),
+                         predicates.end(),
+                         scatter_indices.begin(),
+                         static_cast<IndexType>(0),
+                         thrust::plus<IndexType>());
+  
+  // scatter the true elements
+  thrust::scatter_if(exec,
+                     first,
+                     last,
+                     scatter_indices.begin(),
+                     predicates.begin(),
+                     result,
+                     thrust::identity<IndexType>());
+  
+  // find the end of the new sequence
+  IndexType output_size = scatter_indices[n - 1] + predicates[n - 1];
+  
+  return result + output_size;
+}
+
+
+} // end namespace detail
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                         InputIterator first,
+                         InputIterator last,
+                         OutputIterator result,
+                         Predicate pred)
+{
+  // XXX it's potentially expensive to send [first,last) twice
+  //     we should probably specialize this case for POD
+  //     since we can safely keep the input in a temporary instead
+  //     of doing two loads
+  return thrust::copy_if(exec, first, last, first, result, pred);
+} // end copy_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+   OutputIterator copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator result,
+                          Predicate pred)
+{
+  typedef typename thrust::iterator_traits<InputIterator1>::difference_type difference_type;
+  
+  // empty sequence
+  if(first == last)
+    return result;
+  
+  difference_type n = thrust::distance(first, last);
+  
+  // create an unsigned version of n (we know n is positive from the comparison above)
+  // to avoid a warning in the compare below
+  typename thrust::detail::make_unsigned<difference_type>::type unsigned_n(n);
+  
+  // use 32-bit indices when possible (almost always)
+  if(sizeof(difference_type) > sizeof(unsigned int) && unsigned_n > thrust::detail::integer_traits<unsigned int>::const_max)
+  {
+    result = detail::copy_if<difference_type>(exec, first, last, stencil, result, pred);
+  } // end if
+  else
+  {
+    result = detail::copy_if<unsigned int>(exec, first, last, stencil, result, pred);
+  } // end else
+
+  return result;
+} // end copy_if()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.h
new file mode 100644
index 000000000000..295d36e6b9df
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.h
@@ -0,0 +1,50 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template <typename DerivedPolicy, typename InputIterator, typename EqualityComparable>
+__host__ __device__
+typename thrust::iterator_traits<InputIterator>::difference_type
+count(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last, const EqualityComparable& value);
+
+
+template <typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+typename thrust::iterator_traits<InputIterator>::difference_type
+count_if(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/count.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.inl
new file mode 100644
index 000000000000..dafc1c1df2a1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/count.inl
@@ -0,0 +1,82 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/count.h>
+#include <thrust/transform_reduce.h>
+#include <thrust/detail/internal_functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template <typename InputType, typename Predicate, typename CountType>
+struct count_if_transform
+{
+  __host__ __device__
+  count_if_transform(Predicate _pred) : pred(_pred){}
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  CountType operator()(const InputType& val)
+  {
+    if(pred(val))
+      return 1;
+    else
+      return 0;
+  } // end operator()
+
+  Predicate pred;
+}; // end count_if_transform
+
+
+template <typename DerivedPolicy, typename InputIterator, typename EqualityComparable>
+__host__ __device__
+typename thrust::iterator_traits<InputIterator>::difference_type
+count(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last, const EqualityComparable& value)
+{
+  using thrust::placeholders::_1;
+
+  return thrust::count_if(exec, first, last, _1 == value);
+} // end count()
+
+
+template <typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+typename thrust::iterator_traits<InputIterator>::difference_type
+count_if(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  typedef typename thrust::iterator_traits<InputIterator>::value_type InputType;
+  typedef typename thrust::iterator_traits<InputIterator>::difference_type CountType;
+
+  thrust::system::detail::generic::count_if_transform<InputType, Predicate, CountType> unary_op(pred);
+  thrust::plus<CountType> binary_op;
+  return thrust::transform_reduce(exec, first, last, unary_op, CountType(0), binary_op);
+} // end count_if()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.h
new file mode 100644
index 000000000000..4627376b575c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.h
@@ -0,0 +1,42 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename InputIterator>
+inline __host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/distance.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.inl
new file mode 100644
index 000000000000..46bad7ba74ea
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/distance.inl
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/distance.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator>
+inline __host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last, thrust::incrementable_traversal_tag)
+{
+  typename thrust::iterator_traits<InputIterator>::difference_type result(0);
+
+  while(first != last)
+  {
+    ++first;
+    ++result;
+  } // end while
+
+  return result;
+} // end advance()
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator>
+inline __host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last, thrust::random_access_traversal_tag)
+{
+  return last - first;
+} // end distance()
+
+
+} // end detail
+
+__thrust_exec_check_disable__
+template<typename InputIterator>
+inline __host__ __device__
+  typename thrust::iterator_traits<InputIterator>::difference_type
+    distance(InputIterator first, InputIterator last)
+{
+  // dispatch on iterator traversal
+  return thrust::system::detail::generic::detail::distance(first, last,
+    typename thrust::iterator_traversal<InputIterator>::type());
+} // end advance()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.h
new file mode 100644
index 000000000000..4afd88d001ff
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.h
@@ -0,0 +1,47 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+bool equal(thrust::execution_policy<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2);
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+bool equal(thrust::execution_policy<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, BinaryPredicate binary_pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/equal.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.inl
new file mode 100644
index 000000000000..c023070cdcb0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/equal.inl
@@ -0,0 +1,59 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/equal.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/mismatch.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+bool equal(thrust::execution_policy<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2)
+{
+  typedef typename thrust::iterator_traits<InputIterator1>::value_type InputType1;
+
+  return thrust::equal(exec, first1, last1, first2, thrust::detail::equal_to<InputType1>());
+}
+
+
+// the == below could be a __host__ function in the case of std::vector::iterator::operator==
+// we make this exception for equal and use __thrust_exec_check_disable__ because it is used in vector's implementation
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+bool equal(thrust::execution_policy<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, BinaryPredicate binary_pred)
+{
+  return thrust::mismatch(exec, first1, last1, first2, binary_pred).first == last1;
+}
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.h
new file mode 100644
index 000000000000..e3b447958e0a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.h
@@ -0,0 +1,88 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file extrema.h
+ *  \brief Generic device implementations of extrema functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/pair.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template <typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator max_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator max_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            BinaryPredicate comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator min_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator min_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            BinaryPredicate comp);
+
+
+template <typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
+                                                             ForwardIterator first, 
+                                                             ForwardIterator last);
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
+                                                             ForwardIterator first, 
+                                                             ForwardIterator last,
+                                                             BinaryPredicate comp);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/extrema.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.inl
new file mode 100644
index 000000000000..744d137de1d8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/extrema.inl
@@ -0,0 +1,262 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file distance.h
+ *  \brief Device implementations for distance.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/get_iterator_value.h>
+#include <thrust/extrema.h>
+#include <thrust/functional.h>
+#include <thrust/pair.h>
+#include <thrust/reduce.h>
+#include <thrust/transform_reduce.h>
+
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/iterator/zip_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+//////////////
+// Functors //
+//////////////
+//
+
+// return the smaller/larger element making sure to prefer the 
+// first occurance of the minimum/maximum element
+template <typename InputType, typename IndexType, typename BinaryPredicate>
+struct min_element_reduction
+{
+  BinaryPredicate comp;
+
+  __host__ __device__ 
+  min_element_reduction(BinaryPredicate comp) : comp(comp){}
+
+  __host__ __device__ 
+  thrust::tuple<InputType, IndexType>
+  operator()(const thrust::tuple<InputType, IndexType>& lhs, 
+             const thrust::tuple<InputType, IndexType>& rhs )
+  {
+    if(comp(thrust::get<0>(lhs), thrust::get<0>(rhs)))
+      return lhs;
+    if(comp(thrust::get<0>(rhs), thrust::get<0>(lhs)))
+      return rhs;
+
+    // values are equivalent, prefer value with smaller index
+    if(thrust::get<1>(lhs) < thrust::get<1>(rhs))
+      return lhs;
+    else
+      return rhs;
+  } // end operator()()
+}; // end min_element_reduction
+
+
+template <typename InputType, typename IndexType, typename BinaryPredicate>
+struct max_element_reduction
+{
+  BinaryPredicate comp;
+
+  __host__ __device__ 
+  max_element_reduction(BinaryPredicate comp) : comp(comp){}
+
+  __host__ __device__ 
+  thrust::tuple<InputType, IndexType>
+  operator()(const thrust::tuple<InputType, IndexType>& lhs, 
+             const thrust::tuple<InputType, IndexType>& rhs )
+  {
+    if(comp(thrust::get<0>(lhs), thrust::get<0>(rhs)))
+      return rhs;
+    if(comp(thrust::get<0>(rhs), thrust::get<0>(lhs)))
+      return lhs;
+
+    // values are equivalent, prefer value with smaller index
+    if(thrust::get<1>(lhs) < thrust::get<1>(rhs))
+      return lhs;
+    else
+      return rhs;
+  } // end operator()()
+}; // end max_element_reduction
+
+
+// return the smaller & larger element making sure to prefer the 
+// first occurance of the minimum/maximum element
+template <typename InputType, typename IndexType, typename BinaryPredicate>
+struct minmax_element_reduction
+{
+  BinaryPredicate comp;
+
+  __host__ __device__
+  minmax_element_reduction(BinaryPredicate comp) : comp(comp){}
+
+  __host__ __device__ 
+  thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >
+  operator()(const thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >& lhs, 
+             const thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >& rhs )
+  {
+
+    return thrust::make_tuple(min_element_reduction<InputType, IndexType, BinaryPredicate>(comp)(thrust::get<0>(lhs), thrust::get<0>(rhs)),
+                              max_element_reduction<InputType, IndexType, BinaryPredicate>(comp)(thrust::get<1>(lhs), thrust::get<1>(rhs)));
+  } // end operator()()
+}; // end minmax_element_reduction
+
+
+template <typename InputType, typename IndexType>
+struct duplicate_tuple
+{
+  __host__ __device__ 
+  thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> >
+  operator()(const thrust::tuple<InputType,IndexType>& t)
+  {
+    return thrust::make_tuple(t, t);
+  }
+}; // end duplicate_tuple
+
+
+} // end namespace detail
+
+
+template <typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator min_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last)
+{
+  typedef typename thrust::iterator_value<ForwardIterator>::type value_type;
+
+  return thrust::min_element(exec, first, last, thrust::less<value_type>());
+} // end min_element()
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator min_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            BinaryPredicate comp)
+{
+  if (first == last)
+    return last;
+
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type      InputType;
+  typedef typename thrust::iterator_traits<ForwardIterator>::difference_type IndexType;
+
+  thrust::tuple<InputType, IndexType> result =
+    thrust::reduce
+      (exec,
+       thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))),
+       thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))) + (last - first),
+       thrust::tuple<InputType, IndexType>(thrust::detail::get_iterator_value(derived_cast(exec), first), 0),
+       detail::min_element_reduction<InputType, IndexType, BinaryPredicate>(comp));
+
+  return first + thrust::get<1>(result);
+} // end min_element()
+
+
+template <typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+ForwardIterator max_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last)
+{
+  typedef typename thrust::iterator_value<ForwardIterator>::type value_type;
+
+  return thrust::max_element(exec, first, last, thrust::less<value_type>());
+} // end max_element()
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+ForwardIterator max_element(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            BinaryPredicate comp)
+{
+  if (first == last)
+    return last;
+
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type      InputType;
+  typedef typename thrust::iterator_traits<ForwardIterator>::difference_type IndexType;
+
+  thrust::tuple<InputType, IndexType> result =
+    thrust::reduce
+      (exec,
+       thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))),
+       thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))) + (last - first),
+       thrust::tuple<InputType, IndexType>(thrust::detail::get_iterator_value(derived_cast(exec),first), 0),
+       detail::max_element_reduction<InputType, IndexType, BinaryPredicate>(comp));
+
+  return first + thrust::get<1>(result);
+} // end max_element()
+
+
+template <typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
+                                                             ForwardIterator first, 
+                                                             ForwardIterator last)
+{
+  typedef typename thrust::iterator_value<ForwardIterator>::type value_type;
+
+  return thrust::minmax_element(exec, first, last, thrust::less<value_type>());
+} // end minmax_element()
+
+
+template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(thrust::execution_policy<DerivedPolicy> &exec,
+                                                             ForwardIterator first, 
+                                                             ForwardIterator last,
+                                                             BinaryPredicate comp)
+{
+  if (first == last)
+    return thrust::make_pair(last, last);
+
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type      InputType;
+  typedef typename thrust::iterator_traits<ForwardIterator>::difference_type IndexType;
+
+  thrust::tuple< thrust::tuple<InputType,IndexType>, thrust::tuple<InputType,IndexType> > result = 
+    thrust::transform_reduce
+      (exec,
+       thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))),
+       thrust::make_zip_iterator(thrust::make_tuple(first, thrust::counting_iterator<IndexType>(0))) + (last - first),
+       detail::duplicate_tuple<InputType, IndexType>(),
+       detail::duplicate_tuple<InputType, IndexType>()(
+         thrust::tuple<InputType, IndexType>(thrust::detail::get_iterator_value(derived_cast(exec),first), 0)),
+       detail::minmax_element_reduction<InputType, IndexType, BinaryPredicate>(comp));
+
+  return thrust::make_pair(first + thrust::get<1>(thrust::get<0>(result)), first + thrust::get<1>(thrust::get<1>(result)));
+} // end minmax_element()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/fill.h
new file mode 100644
index 000000000000..5a881359b56b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/fill.h
@@ -0,0 +1,61 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/internal_functional.h>
+#include <thrust/generate.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename OutputIterator, typename Size, typename T>
+__host__ __device__
+  OutputIterator fill_n(thrust::execution_policy<DerivedPolicy> &exec,
+                        OutputIterator first,
+                        Size n,
+                        const T &value)
+{
+  // XXX consider using the placeholder expression _1 = value
+  return thrust::generate_n(exec, first, n, thrust::detail::fill_functor<T>(value));
+}
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void fill(thrust::execution_policy<DerivedPolicy> &exec,
+            ForwardIterator first,
+            ForwardIterator last,
+            const T &value)
+{
+  // XXX consider using the placeholder expression _1 = value
+  thrust::generate(exec, first, last, thrust::detail::fill_functor<T>(value));
+}
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.h
new file mode 100644
index 000000000000..6db441d02f72
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.h
@@ -0,0 +1,62 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+InputIterator find(thrust::execution_policy<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   const T& value);
+
+
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if(thrust::execution_policy<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred);
+
+
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if_not(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          Predicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/find.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.inl
new file mode 100644
index 000000000000..8bd619561e92
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/find.inl
@@ -0,0 +1,151 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/find.h>
+#include <thrust/reduce.h>
+
+#include <thrust/tuple.h>
+#include <thrust/detail/minmax.h>
+#include <thrust/iterator/counting_iterator.h>
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/internal_functional.h>
+
+
+// Contributed by Erich Elsen
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+InputIterator find(thrust::execution_policy<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   const T& value)
+{
+  using thrust::placeholders::_1;
+
+  return thrust::find_if(exec, first, last, _1 == value);
+} // end find()
+
+
+template<typename TupleType>
+struct find_if_functor
+{
+  __host__ __device__
+  TupleType operator()(const TupleType& lhs, const TupleType& rhs) const
+  {
+    // select the smallest index among true results
+    if(thrust::get<0>(lhs) && thrust::get<0>(rhs))
+    {
+      return TupleType(true, (thrust::min)(thrust::get<1>(lhs), thrust::get<1>(rhs)));
+    }
+    else if(thrust::get<0>(lhs))
+    {
+      return lhs;
+    }
+    else
+    {
+      return rhs;
+    }
+  }
+};
+
+
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if(thrust::execution_policy<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  typedef typename thrust::iterator_traits<InputIterator>::difference_type difference_type;
+  typedef typename thrust::tuple<bool,difference_type> result_type;
+
+  // empty sequence
+  if(first == last) return last;
+
+  const difference_type n = thrust::distance(first, last);
+
+  // this implementation breaks up the sequence into separate intervals
+  // in an attempt to early-out as soon as a value is found
+
+  // TODO incorporate sizeof(InputType) into interval_threshold and round to multiple of 32
+  const difference_type interval_threshold = 1 << 20;
+  const difference_type interval_size = (thrust::min)(interval_threshold, n);
+
+  // force transform_iterator output to bool
+  typedef thrust::transform_iterator<Predicate, InputIterator, bool> XfrmIterator;
+  typedef thrust::tuple<XfrmIterator, thrust::counting_iterator<difference_type> > IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  IteratorTuple iter_tuple = thrust::make_tuple(XfrmIterator(first, pred),
+                                                thrust::counting_iterator<difference_type>(0));
+
+  ZipIterator begin = thrust::make_zip_iterator(iter_tuple);
+  ZipIterator end   = begin + n;
+
+  for(ZipIterator interval_begin = begin; interval_begin < end; interval_begin += interval_size)
+  {
+    ZipIterator interval_end = interval_begin + interval_size;
+    if(end < interval_end)
+    {
+      interval_end = end;
+    } // end if
+
+    result_type result = thrust::reduce(exec,
+                                        interval_begin, interval_end,
+                                        result_type(false,interval_end - begin),
+                                        find_if_functor<result_type>());
+
+    // see if we found something
+    if(thrust::get<0>(result))
+    {
+      return first + thrust::get<1>(result);
+    }
+  }
+
+  //nothing was found if we reach here...
+  return first + n;
+}
+
+
+template<typename DerivedPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+InputIterator find_if_not(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          Predicate pred)
+{
+  return thrust::find_if(exec, first, last, thrust::detail::not1(pred));
+} // end find()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/for_each.h
new file mode 100644
index 000000000000..3c2ec12cd21f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/for_each.h
@@ -0,0 +1,77 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file for_each.h
+ *  \brief Generic implementation of for_each & for_each_n.
+ *         It is an error to call these functions; they have no implementation.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/detail/static_assert.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename UnaryFunction>
+__host__ __device__
+InputIterator for_each(thrust::execution_policy<DerivedPolicy> &,
+                       InputIterator first,
+                       InputIterator ,
+                       UnaryFunction )
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator, false>::value)
+  , "unimplemented for this system"
+  );
+  return first;
+} // end for_each()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename UnaryFunction>
+__host__ __device__
+InputIterator for_each_n(thrust::execution_policy<DerivedPolicy> &,
+                         InputIterator first,
+                         Size ,
+                         UnaryFunction )
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator, false>::value)
+  , "unimplemented for this system"
+  );
+  return first;
+} // end for_each_n()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.h
new file mode 100644
index 000000000000..5b6b41831135
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.h
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather(thrust::execution_policy<DerivedPolicy> &exec,
+                        InputIterator                            map_first,
+                        InputIterator                            map_last,
+                        RandomAccessIterator                     input_first,
+                        OutputIterator                           result);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather_if(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1                           map_first,
+                           InputIterator1                           map_last,
+                           InputIterator2                           stencil,
+                           RandomAccessIterator                     input_first,
+                           OutputIterator                           result);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator gather_if(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1                           map_first,
+                           InputIterator1                           map_last,
+                           InputIterator2                           stencil,
+                           RandomAccessIterator                     input_first,
+                           OutputIterator                           result,
+                           Predicate                                pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/gather.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.inl
new file mode 100644
index 000000000000..7ab550edf144
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/gather.inl
@@ -0,0 +1,108 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/gather.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/functional.h>
+#include <thrust/transform.h>
+#include <thrust/iterator/permutation_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather(thrust::execution_policy<DerivedPolicy> &exec,
+                        InputIterator                            map_first,
+                        InputIterator                            map_last,
+                        RandomAccessIterator                     input_first,
+                        OutputIterator                           result)
+{
+  return thrust::transform(exec,
+                           thrust::make_permutation_iterator(input_first, map_first),
+                           thrust::make_permutation_iterator(input_first, map_last),
+                           result,
+                           thrust::identity<typename thrust::iterator_value<RandomAccessIterator>::type>());
+} // end gather()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator gather_if(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1                           map_first,
+                           InputIterator1                           map_last,
+                           InputIterator2                           stencil,
+                           RandomAccessIterator                     input_first,
+                           OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator2>::type StencilType;
+  return thrust::gather_if(exec,
+                           map_first,
+                           map_last,
+                           stencil,
+                           input_first,
+                           result,
+                           thrust::identity<StencilType>());
+} // end gather_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator gather_if(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1                           map_first,
+                           InputIterator1                           map_last,
+                           InputIterator2                           stencil,
+                           RandomAccessIterator                     input_first,
+                           OutputIterator                           result,
+                           Predicate                                pred)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type InputType;
+  return thrust::transform_if(exec,
+                              thrust::make_permutation_iterator(input_first, map_first),
+                              thrust::make_permutation_iterator(input_first, map_last),
+                              stencil,
+                              result,
+                              thrust::identity<InputType>(),
+                              pred);
+} // end gather_if()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.h
new file mode 100644
index 000000000000..a9846c5bed6b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.h
@@ -0,0 +1,56 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename Generator>
+__host__ __device__
+  void generate(thrust::execution_policy<ExecutionPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                Generator gen);
+
+template<typename ExecutionPolicy,
+         typename OutputIterator,
+         typename Size,
+         typename Generator>
+__host__ __device__
+  OutputIterator generate_n(thrust::execution_policy<ExecutionPolicy> &exec,
+                            OutputIterator first,
+                            Size n,
+                            Generator gen);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/generate.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.inl
new file mode 100644
index 000000000000..869e0f32bf4d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/generate.inl
@@ -0,0 +1,100 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/generate.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/for_each.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename Generator>
+__host__ __device__
+  void generate(thrust::execution_policy<ExecutionPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                Generator gen)
+{
+  // this static assert is necessary due to a workaround in generate_functor
+  // it takes a const reference to accept temporaries from proxy iterators
+  // and then const_casts the constness away
+  //
+  // this had the weird side effect of allowing generate (and fill, and whatever
+  // else is implemented in terms of generate) to fill through const iterators.
+  // this might become unnecessary once Thrust is C++11-and-above only, since the
+  // other solution is to take an rvalue reference in a second overload of
+  // operator() of the function object, but until we support pre-11, this is a
+  // nice solution that validates the const_cast and doesn't take away any
+  // functionality.
+  THRUST_STATIC_ASSERT_MSG(
+    !thrust::detail::is_const<
+      typename thrust::detail::remove_reference<
+        typename thrust::iterator_traits<ForwardIterator>::reference
+      >::type
+    >::value
+  , "generating to `const` iterators is not allowed"
+  );
+  thrust::for_each(exec, first, last, typename thrust::detail::generate_functor<ExecutionPolicy,Generator>::type(gen));
+} // end generate()
+
+template<typename ExecutionPolicy,
+         typename OutputIterator,
+         typename Size,
+         typename Generator>
+__host__ __device__
+  OutputIterator generate_n(thrust::execution_policy<ExecutionPolicy> &exec,
+                            OutputIterator first,
+                            Size n,
+                            Generator gen)
+{
+  // this static assert is necessary due to a workaround in generate_functor
+  // it takes a const reference to accept temporaries from proxy iterators
+  // and then const_casts the constness away
+  //
+  // this had the weird side effect of allowing generate (and fill, and whatever
+  // else is implemented in terms of generate) to fill through const iterators.
+  // this might become unnecessary once Thrust is C++11-and-above only, since the
+  // other solution is to take an rvalue reference in a second overload of
+  // operator() of the function object, but until we support pre-11, this is a
+  // nice solution that validates the const_cast and doesn't take away any
+  // functionality.
+  THRUST_STATIC_ASSERT_MSG(
+    !thrust::detail::is_const<
+      typename thrust::detail::remove_reference<
+        typename thrust::iterator_traits<OutputIterator>::reference
+      >::type
+    >::value
+  , "generating to `const` iterators is not allowed"
+  );
+  return thrust::for_each_n(exec, first, n, typename thrust::detail::generate_functor<ExecutionPolicy,Generator>::type(gen));
+} // end generate()
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.h
new file mode 100644
index 000000000000..62d10d31f588
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.h
@@ -0,0 +1,58 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputType>
+__host__ __device__
+  OutputType inner_product(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputType init);
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputType, typename BinaryFunction1, typename BinaryFunction2>
+__host__ __device__
+OutputType inner_product(thrust::execution_policy<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init, 
+                         BinaryFunction1 binary_op1,
+                         BinaryFunction2 binary_op2);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/inner_product.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.inl
new file mode 100644
index 000000000000..5055ec10f665
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/inner_product.inl
@@ -0,0 +1,73 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/inner_product.h>
+#include <thrust/functional.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/transform_reduce.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputType>
+__host__ __device__
+OutputType inner_product(thrust::execution_policy<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init)
+{
+  thrust::plus<OutputType>       binary_op1;
+  thrust::multiplies<OutputType> binary_op2;
+  return thrust::inner_product(exec, first1, last1, first2, init, binary_op1, binary_op2);
+} // end inner_product()
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputType, typename BinaryFunction1, typename BinaryFunction2>
+__host__ __device__
+OutputType inner_product(thrust::execution_policy<DerivedPolicy> &exec,
+                         InputIterator1 first1,
+                         InputIterator1 last1,
+                         InputIterator2 first2,
+                         OutputType init,
+                         BinaryFunction1 binary_op1,
+                         BinaryFunction2 binary_op2)
+{
+  typedef thrust::zip_iterator<thrust::tuple<InputIterator1,InputIterator2> > ZipIter;
+
+  ZipIter first = thrust::make_zip_iterator(thrust::make_tuple(first1,first2));
+
+  // only the first iterator in the tuple is relevant for the purposes of last
+  ZipIter last  = thrust::make_zip_iterator(thrust::make_tuple(last1, first2));
+
+  return thrust::transform_reduce(exec, first, last, thrust::detail::zipped_binary_op<OutputType,BinaryFunction2>(binary_op2), init, binary_op1);
+} // end inner_product()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/logical.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/logical.h
new file mode 100644
index 000000000000..e261154e2487
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/logical.h
@@ -0,0 +1,62 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/find.h>
+#include <thrust/logical.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool all_of(thrust::execution_policy<ExecutionPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  return thrust::find_if(exec, first, last, thrust::detail::not1(pred)) == last;
+}
+
+
+template<typename ExecutionPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool any_of(thrust::execution_policy<ExecutionPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  return thrust::find_if(exec, first, last, pred) != last;
+}
+
+
+template<typename ExecutionPolicy, typename InputIterator, typename Predicate>
+__host__ __device__
+bool none_of(thrust::execution_policy<ExecutionPolicy> &exec, InputIterator first, InputIterator last, Predicate pred)
+{
+  return !thrust::any_of(exec, first, last, pred);
+}
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.h
new file mode 100644
index 000000000000..675cc73023c2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.h
@@ -0,0 +1,70 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file generic/memory.h
+ *  \brief Generic implementation of memory functions.
+ *         Calling some of these is an error. They have no implementation.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/pointer.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename DerivedPolicy, typename Size>
+__host__ __device__
+void malloc(thrust::execution_policy<DerivedPolicy> &, Size);
+
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+thrust::pointer<T,DerivedPolicy> malloc(thrust::execution_policy<DerivedPolicy> &s, std::size_t n);
+
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+void free(thrust::execution_policy<DerivedPolicy> &, Pointer);
+
+template<typename Pointer1, typename Pointer2>
+__host__ __device__
+void assign_value(tag, Pointer1, Pointer2);
+
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+void get_value(thrust::execution_policy<DerivedPolicy> &, Pointer);
+
+template<typename DerivedPolicy, typename Pointer1, typename Pointer2>
+__host__ __device__
+void iter_swap(thrust::execution_policy<DerivedPolicy>&, Pointer1, Pointer2);
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/memory.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.inl
new file mode 100644
index 000000000000..b857290985ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/memory.inl
@@ -0,0 +1,105 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/system/detail/generic/memory.h>
+#include <thrust/system/detail/adl/malloc_and_free.h>
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/malloc_and_free.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename Size>
+__host__ __device__
+  void malloc(thrust::execution_policy<DerivedPolicy> &, Size)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<Size, false>::value)
+  , "unimplemented for this system"
+  );
+}
+
+
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+  thrust::pointer<T,DerivedPolicy>
+    malloc(thrust::execution_policy<DerivedPolicy> &exec, std::size_t n)
+{
+  thrust::pointer<void,DerivedPolicy> void_ptr = thrust::malloc(exec, sizeof(T) * n);
+
+  return pointer<T,DerivedPolicy>(static_cast<T*>(void_ptr.get()));
+} // end malloc()
+
+
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  void free(thrust::execution_policy<DerivedPolicy> &, Pointer)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<Pointer, false>::value)
+  , "unimplemented for this system"
+  );
+}
+
+
+template<typename DerivedPolicy, typename Pointer1, typename Pointer2>
+__host__ __device__
+void assign_value(thrust::execution_policy<DerivedPolicy> &, Pointer1, Pointer2)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<Pointer1, false>::value)
+  , "unimplemented for this system"
+  );
+}
+
+
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+void get_value(thrust::execution_policy<DerivedPolicy> &, Pointer)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<Pointer, false>::value)
+  , "unimplemented for this system"
+  );
+}
+
+
+template<typename DerivedPolicy, typename Pointer1, typename Pointer2>
+__host__ __device__
+void iter_swap(thrust::execution_policy<DerivedPolicy> &, Pointer1, Pointer2)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<Pointer1, false>::value)
+  , "unimplemented for this system"
+  );
+}
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.h
new file mode 100644
index 000000000000..6e82464074e1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.h
@@ -0,0 +1,90 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+// XXX calling this function is an error; there is no implementation
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  OutputIterator merge(thrust::execution_policy<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result,
+                       StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator merge(thrust::execution_policy<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result);
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2, typename Compare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result,
+                 Compare comp);
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/merge.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.inl
new file mode 100644
index 000000000000..03b77e6235b5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/merge.inl
@@ -0,0 +1,130 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/static_assert.h>
+#include <thrust/system/detail/generic/merge.h>
+#include <thrust/merge.h>
+#include <thrust/functional.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/internal_functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  OutputIterator merge(thrust::execution_policy<DerivedPolicy> &,
+                       InputIterator1,
+                       InputIterator1,
+                       InputIterator2,
+                       InputIterator2,
+                       OutputIterator result,
+                       StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator1, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end merge()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator merge(thrust::execution_policy<DerivedPolicy> &exec,
+                       InputIterator1 first1,
+                       InputIterator1 last1,
+                       InputIterator2 first2,
+                       InputIterator2 last2,
+                       OutputIterator result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::merge(exec,first1,last1,first2,last2,result,thrust::less<value_type>());
+} // end merge()
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2, typename Compare>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result,
+                 Compare comp)
+{
+  typedef thrust::tuple<InputIterator1, InputIterator3>   iterator_tuple1;
+  typedef thrust::tuple<InputIterator2, InputIterator4>   iterator_tuple2;
+  typedef thrust::tuple<OutputIterator1, OutputIterator2> iterator_tuple3;
+
+  typedef thrust::zip_iterator<iterator_tuple1> zip_iterator1;
+  typedef thrust::zip_iterator<iterator_tuple2> zip_iterator2;
+  typedef thrust::zip_iterator<iterator_tuple3> zip_iterator3;
+
+  zip_iterator1 zipped_first1 = thrust::make_zip_iterator(thrust::make_tuple(keys_first1, values_first1));
+  zip_iterator1 zipped_last1  = thrust::make_zip_iterator(thrust::make_tuple(keys_last1, values_first1));
+
+  zip_iterator2 zipped_first2 = thrust::make_zip_iterator(thrust::make_tuple(keys_first2, values_first2));
+  zip_iterator2 zipped_last2  = thrust::make_zip_iterator(thrust::make_tuple(keys_last2, values_first2));
+
+  zip_iterator3 zipped_result = thrust::make_zip_iterator(thrust::make_tuple(keys_result, values_result));
+
+  thrust::detail::compare_first<Compare> comp_first(comp);
+
+  iterator_tuple3 result = thrust::merge(exec, zipped_first1, zipped_last1, zipped_first2, zipped_last2, zipped_result, comp_first).get_iterator_tuple();
+
+  return thrust::make_pair(thrust::get<0>(result), thrust::get<1>(result));
+} // end merge_by_key()
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename InputIterator3, typename InputIterator4, typename OutputIterator1, typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    merge_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                 InputIterator1 keys_first1, InputIterator1 keys_last1,
+                 InputIterator2 keys_first2, InputIterator2 keys_last2,
+                 InputIterator3 values_first1, InputIterator4 values_first2,
+                 OutputIterator1 keys_result,
+                 OutputIterator2 values_result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::merge_by_key(exec, keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, thrust::less<value_type>());
+} // end merge_by_key()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.h
new file mode 100644
index 000000000000..4a71cd344142
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.h
@@ -0,0 +1,57 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+  thrust::pair<InputIterator1, InputIterator2>
+    mismatch(thrust::execution_policy<DerivedPolicy> &exec,
+             InputIterator1 first1,
+             InputIterator1 last1,
+             InputIterator2 first2);
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<InputIterator1, InputIterator2>
+    mismatch(thrust::execution_policy<DerivedPolicy> &exec,
+             InputIterator1 first1,
+             InputIterator1 last1,
+             InputIterator2 first2,
+             BinaryPredicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/mismatch.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.inl
new file mode 100644
index 000000000000..f6b9674a1cc6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/mismatch.inl
@@ -0,0 +1,75 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/mismatch.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/find.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2>
+__host__ __device__
+  thrust::pair<InputIterator1, InputIterator2>
+    mismatch(thrust::execution_policy<DerivedPolicy> &exec,
+             InputIterator1 first1,
+             InputIterator1 last1,
+             InputIterator2 first2)
+{
+  using namespace thrust::placeholders;
+
+  return thrust::mismatch(exec, first1, last1, first2, _1 == _2);
+} // end mismatch()
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<InputIterator1, InputIterator2>
+    mismatch(thrust::execution_policy<DerivedPolicy> &exec,
+             InputIterator1 first1,
+             InputIterator1 last1,
+             InputIterator2 first2,
+             BinaryPredicate pred)
+{
+  // Contributed by Erich Elsen
+  typedef thrust::tuple<InputIterator1,InputIterator2> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple>          ZipIterator;
+
+  ZipIterator zipped_first = thrust::make_zip_iterator(thrust::make_tuple(first1,first2));
+  ZipIterator zipped_last  = thrust::make_zip_iterator(thrust::make_tuple(last1, first2));
+
+  ZipIterator result = thrust::find_if_not(exec, zipped_first, zipped_last, thrust::detail::tuple_binary_predicate<BinaryPredicate>(pred));
+
+  return thrust::make_pair(thrust::get<0>(result.get_iterator_tuple()),
+                           thrust::get<1>(result.get_iterator_tuple()));
+} // end mismatch()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.h
new file mode 100644
index 000000000000..113d6ecbc3d3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.h
@@ -0,0 +1,169 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file partition.h
+ *  \brief Generic implementations of partition functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(thrust::execution_policy<ExecutionPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred);
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(thrust::execution_policy<ExecutionPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(thrust::execution_policy<ExecutionPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(thrust::execution_policy<ExecutionPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                   InputIterator1 first,
+                   InputIterator1 last,
+                   InputIterator2 stencil,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition_point(thrust::execution_policy<ExecutionPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Predicate pred);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  bool is_partitioned(thrust::execution_policy<ExecutionPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/partition.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.inl
new file mode 100644
index 000000000000..ab56fdd57512
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/partition.inl
@@ -0,0 +1,249 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/partition.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/pair.h>
+
+#include <thrust/remove.h>
+#include <thrust/count.h>
+#include <thrust/advance.h>
+#include <thrust/partition.h>
+#include <thrust/sort.h>
+#include <thrust/iterator/transform_iterator.h>
+
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/temporary_array.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(thrust::execution_policy<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type InputType;
+
+  // copy input to temp buffer
+  thrust::detail::temporary_array<InputType,DerivedPolicy> temp(exec, first, last);
+
+  // count the size of the true partition
+  typename thrust::iterator_difference<ForwardIterator>::type num_true = thrust::count_if(exec, first,last,pred);
+
+  // point to the beginning of the false partition
+  ForwardIterator out_false = first;
+  thrust::advance(out_false, num_true);
+
+  return thrust::stable_partition_copy(exec, temp.begin(), temp.end(), first, out_false, pred).first;
+} // end stable_partition()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(thrust::execution_policy<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type InputType;
+
+  // copy input to temp buffer
+  thrust::detail::temporary_array<InputType,DerivedPolicy> temp(exec, first, last);
+
+  // count the size of the true partition
+  InputIterator stencil_last = stencil;
+  thrust::advance(stencil_last, temp.size());
+  typename thrust::iterator_difference<InputIterator>::type num_true = thrust::count_if(exec, stencil, stencil_last, pred);
+
+  // point to the beginning of the false partition
+  ForwardIterator out_false = first;
+  thrust::advance(out_false, num_true);
+
+  return thrust::stable_partition_copy(exec, temp.begin(), temp.end(), stencil, first, out_false, pred).first;
+} // end stable_partition()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  thrust::detail::unary_negate<Predicate> not_pred(pred);
+
+  // remove_copy_if the true partition to out_true
+  OutputIterator1 end_of_true_partition = thrust::remove_copy_if(exec, first, last, out_true, not_pred);
+
+  // remove_copy_if the false partition to out_false
+  OutputIterator2 end_of_false_partition = thrust::remove_copy_if(exec, first, last, out_false, pred);
+
+  return thrust::make_pair(end_of_true_partition, end_of_false_partition);
+} // end stable_partition_copy()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  thrust::detail::unary_negate<Predicate> not_pred(pred);
+
+  // remove_copy_if the true partition to out_true
+  OutputIterator1 end_of_true_partition = thrust::remove_copy_if(exec, first, last, stencil, out_true, not_pred);
+
+  // remove_copy_if the false partition to out_false
+  OutputIterator2 end_of_false_partition = thrust::remove_copy_if(exec, first, last, stencil, out_false, pred);
+
+  return thrust::make_pair(end_of_true_partition, end_of_false_partition);
+} // end stable_partition_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  return thrust::stable_partition(exec, first, last, pred);
+} // end partition()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  return thrust::stable_partition(exec, first, last, stencil, pred);
+} // end partition()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                   InputIterator first,
+                   InputIterator last,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred)
+{
+  return thrust::stable_partition_copy(exec,first,last,out_true,out_false,pred);
+} // end partition_copy()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    partition_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                   InputIterator1 first,
+                   InputIterator1 last,
+                   InputIterator2 stencil,
+                   OutputIterator1 out_true,
+                   OutputIterator2 out_false,
+                   Predicate pred)
+{
+  return thrust::stable_partition_copy(exec,first,last,stencil,out_true,out_false,pred);
+} // end partition_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition_point(thrust::execution_policy<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Predicate pred)
+{
+  return thrust::find_if_not(exec, first, last, pred);
+} // end partition_point()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  bool is_partitioned(thrust::execution_policy<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  return thrust::is_sorted(exec,
+                           thrust::make_transform_iterator(first, thrust::detail::not1(pred)),
+                           thrust::make_transform_iterator(last,  thrust::detail::not1(pred)));
+} // end is_partitioned()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/per_device_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/per_device_resource.h
new file mode 100644
index 000000000000..606f91f36581
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/per_device_resource.h
@@ -0,0 +1,46 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/mr/memory_resource.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename MR, typename DerivedPolicy>
+__host__
+MR * get_per_device_resource(thrust::detail::execution_policy_base<DerivedPolicy>&)
+{
+    return mr::get_global_resource<MR>();
+}
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.h
new file mode 100644
index 000000000000..f28b11a87bb7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.h
@@ -0,0 +1,58 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::value_type
+    reduce(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last);
+
+
+template<typename DerivedPolicy, typename InputIterator, typename T>
+__host__ __device__
+  T reduce(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last, T init);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename T,
+         typename BinaryFunction>
+__host__ __device__
+  T reduce(thrust::execution_policy<DerivedPolicy> &exec, InputIterator first, InputIterator last, T init, BinaryFunction binary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/reduce.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.inl
new file mode 100644
index 000000000000..d673d0cf882c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce.inl
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/reduce.h>
+#include <thrust/system/detail/generic/reduce.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/functional.h>
+#include <thrust/detail/static_assert.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy, typename InputIterator>
+__host__ __device__
+  typename thrust::iterator_traits<InputIterator>::value_type
+    reduce(thrust::execution_policy<ExecutionPolicy> &exec, InputIterator first, InputIterator last)
+{
+  typedef typename thrust::iterator_value<InputIterator>::type InputType;
+
+  // use InputType(0) as init by default
+  return thrust::reduce(exec, first, last, InputType(0));
+} // end reduce()
+
+
+template<typename ExecutionPolicy, typename InputIterator, typename T>
+__host__ __device__
+  T reduce(thrust::execution_policy<ExecutionPolicy> &exec, InputIterator first, InputIterator last, T init)
+{
+  // use plus<T> by default
+  return thrust::reduce(exec, first, last, init, thrust::plus<T>());
+} // end reduce()
+
+
+template<typename ExecutionPolicy,
+         typename RandomAccessIterator,
+         typename OutputType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputType reduce(thrust::execution_policy<ExecutionPolicy> &,
+                    RandomAccessIterator,
+                    RandomAccessIterator,
+                    OutputType,
+                    BinaryFunction)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<RandomAccessIterator, false>::value)
+  , "unimplemented for this system"
+  );
+  return OutputType();
+} // end reduce()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.h
new file mode 100644
index 000000000000..8ba47e11f0a9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.h
@@ -0,0 +1,88 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 keys_first, 
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output);
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 keys_first, 
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output,
+                  BinaryPredicate binary_pred);
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 keys_first, 
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output,
+                  BinaryPredicate binary_pred,
+                  BinaryFunction binary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/reduce_by_key.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.inl
new file mode 100644
index 000000000000..2ea73fedaa19
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reduce_by_key.inl
@@ -0,0 +1,193 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/iterator/is_output_iterator.h>
+#include <thrust/detail/type_traits/function_traits.h>
+#include <thrust/transform.h>
+#include <thrust/scatter.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <limits>
+
+#include <thrust/detail/internal_functional.h>
+#include <thrust/scan.h>
+#include <thrust/detail/temporary_array.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+template <typename ValueType, typename TailFlagType, typename AssociativeOperator>
+struct reduce_by_key_functor
+{
+  AssociativeOperator binary_op;
+
+  typedef typename thrust::tuple<ValueType, TailFlagType> result_type;
+
+  __host__ __device__
+  reduce_by_key_functor(AssociativeOperator _binary_op) : binary_op(_binary_op) {}
+
+  __host__ __device__
+  result_type operator()(result_type a, result_type b)
+  {
+    return result_type(thrust::get<1>(b) ? thrust::get<0>(b) : binary_op(thrust::get<0>(a), thrust::get<0>(b)),
+                       thrust::get<1>(a) | thrust::get<1>(b));
+  }
+};
+
+
+} // end namespace detail
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                  InputIterator1 keys_first,
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output,
+                  BinaryPredicate binary_pred,
+                  BinaryFunction binary_op)
+{
+    typedef typename thrust::iterator_traits<InputIterator1>::difference_type difference_type;
+
+    typedef unsigned int FlagType;  // TODO use difference_type
+
+    // Use the input iterator's value type per https://wg21.link/P0571
+    using ValueType = typename thrust::iterator_value<InputIterator2>::type;
+
+    if (keys_first == keys_last)
+        return thrust::make_pair(keys_output, values_output);
+
+    // input size
+    difference_type n = keys_last - keys_first;
+
+    InputIterator2 values_last = values_first + n;
+
+    // compute head flags
+    thrust::detail::temporary_array<FlagType,ExecutionPolicy> head_flags(exec, n);
+    thrust::transform(exec, keys_first, keys_last - 1, keys_first + 1, head_flags.begin() + 1, thrust::detail::not2(binary_pred));
+    head_flags[0] = 1;
+
+    // compute tail flags
+    thrust::detail::temporary_array<FlagType,ExecutionPolicy> tail_flags(exec, n); //COPY INSTEAD OF TRANSFORM
+    thrust::transform(exec, keys_first, keys_last - 1, keys_first + 1, tail_flags.begin(), thrust::detail::not2(binary_pred));
+    tail_flags[n-1] = 1;
+
+    // scan the values by flag
+    thrust::detail::temporary_array<ValueType,ExecutionPolicy> scanned_values(exec, n);
+    thrust::detail::temporary_array<FlagType,ExecutionPolicy>  scanned_tail_flags(exec, n);
+
+    thrust::inclusive_scan
+        (exec,
+         thrust::make_zip_iterator(thrust::make_tuple(values_first,           head_flags.begin())),
+         thrust::make_zip_iterator(thrust::make_tuple(values_last,            head_flags.end())),
+         thrust::make_zip_iterator(thrust::make_tuple(scanned_values.begin(), scanned_tail_flags.begin())),
+         detail::reduce_by_key_functor<ValueType, FlagType, BinaryFunction>(binary_op));
+
+    thrust::exclusive_scan(exec, tail_flags.begin(), tail_flags.end(), scanned_tail_flags.begin(), FlagType(0), thrust::plus<FlagType>());
+
+    // number of unique keys
+    FlagType N = scanned_tail_flags[n - 1] + 1;
+
+    // scatter the keys and accumulated values
+    thrust::scatter_if(exec, keys_first,            keys_last,             scanned_tail_flags.begin(), head_flags.begin(), keys_output);
+    thrust::scatter_if(exec, scanned_values.begin(), scanned_values.end(), scanned_tail_flags.begin(), tail_flags.begin(), values_output);
+
+    return thrust::make_pair(keys_output + N, values_output + N);
+} // end reduce_by_key()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                  InputIterator1 keys_first,
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type KeyType;
+
+  // use equal_to<KeyType> as default BinaryPredicate
+  return thrust::reduce_by_key(exec, keys_first, keys_last, values_first, keys_output, values_output, thrust::equal_to<KeyType>());
+} // end reduce_by_key()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                  InputIterator1 keys_first,
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output,
+                  BinaryPredicate binary_pred)
+{
+  typedef typename thrust::detail::eval_if<
+    thrust::detail::is_output_iterator<OutputIterator2>::value,
+    thrust::iterator_value<InputIterator2>,
+    thrust::iterator_value<OutputIterator2>
+  >::type T;
+
+  // use plus<T> as default BinaryFunction
+  return thrust::reduce_by_key(exec,
+                               keys_first, keys_last,
+                               values_first,
+                               keys_output,
+                               values_output,
+                               binary_pred,
+                               thrust::plus<T>());
+} // end reduce_by_key()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.h
new file mode 100644
index 000000000000..37354ef805cc
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.h
@@ -0,0 +1,112 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file remove.h
+ *  \brief Generic implementations of remove functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  ForwardIterator remove(thrust::execution_policy<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last,
+                         const T &value);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator remove_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator result,
+                             const T &value);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/remove.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.inl
new file mode 100644
index 000000000000..e51a3caeee1b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/remove.inl
@@ -0,0 +1,146 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/remove.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/copy_if.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/remove.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  ForwardIterator remove(thrust::execution_policy<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last,
+                         const T &value)
+{
+  thrust::detail::equal_to_value<T> pred(value);
+
+  // XXX consider using a placeholder here
+  return thrust::remove_if(exec, first, last, pred);
+} // end remove()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator remove_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator result,
+                             const T &value)
+{
+  thrust::detail::equal_to_value<T> pred(value);
+
+  // XXX consider using a placeholder here
+  return thrust::remove_copy_if(exec, first, last, result, pred);
+} // end remove_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type InputType;
+
+  // create temporary storage for an intermediate result
+  thrust::detail::temporary_array<InputType,DerivedPolicy> temp(exec, first, last);
+
+  // remove into temp
+  return thrust::remove_copy_if(exec, temp.begin(), temp.end(), temp.begin(), first, pred);
+} // end remove_if()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(thrust::execution_policy<DerivedPolicy> &exec,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type InputType;
+
+  // create temporary storage for an intermediate result
+  thrust::detail::temporary_array<InputType,DerivedPolicy> temp(exec, first, last);
+
+  // remove into temp
+  return thrust::remove_copy_if(exec, temp.begin(), temp.end(), stencil, first, pred);
+} // end remove_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  return thrust::remove_copy_if(exec, first, last, first, result, pred);
+} // end remove_copy_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  return thrust::copy_if(exec, first, last, stencil, result, thrust::detail::not1(pred));
+} // end remove_copy_if()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.h
new file mode 100644
index 000000000000..0821d6c07db9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.h
@@ -0,0 +1,97 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                 InputIterator first,
+                                 InputIterator last,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value);
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                 InputIterator1 first,
+                                 InputIterator1 last,
+                                 InputIterator2 stencil,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value);
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename T>
+__host__ __device__
+  OutputIterator replace_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                              InputIterator first,
+                              InputIterator last,
+                              OutputIterator result,
+                              const T &old_value,
+                              const T &new_value);
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  ForwardIterator first,
+                  ForwardIterator last,
+                  Predicate pred,
+                  const T &new_value);
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  ForwardIterator first,
+                  ForwardIterator last,
+                  InputIterator stencil,
+                  Predicate pred,
+                  const T &new_value);
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void replace(thrust::execution_policy<DerivedPolicy> &exec,
+               ForwardIterator first,
+               ForwardIterator last,
+               const T &old_value,
+               const T &new_value);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/replace.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.inl
new file mode 100644
index 000000000000..ed845dd458c9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/replace.inl
@@ -0,0 +1,179 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/functional.h>
+#include <thrust/system/detail/generic/replace.h>
+#include <thrust/transform.h>
+#include <thrust/replace.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+// this functor receives x, and returns a new_value if predicate(x) is true; otherwise,
+// it returns x
+template<typename Predicate, typename NewType, typename OutputType>
+  struct new_value_if
+{
+  __host__ __device__
+  new_value_if(Predicate p, NewType nv):pred(p),new_value(nv){}
+
+  template<typename InputType>
+  __host__ __device__
+  OutputType operator()(const InputType &x) const
+  {
+    return pred(x) ? new_value : x;
+  } // end operator()()
+
+  // this version of operator()() works like the previous but
+  // feeds its second argument to pred
+  template<typename InputType, typename PredicateArgumentType>
+  __host__ __device__
+  OutputType operator()(const InputType &x, const PredicateArgumentType &y)
+  {
+    return pred(y) ? new_value : x;
+  } // end operator()()
+
+  Predicate pred;
+  NewType new_value;
+}; // end new_value_if
+
+
+// this unary functor ignores its argument and returns a constant
+template<typename T>
+  struct constant_unary
+{
+  __host__ __device__
+  constant_unary(T _c):c(_c){}
+
+  template<typename U>
+  __host__ __device__
+  T operator()(U &)
+  {
+    return c;
+  } // end operator()()
+
+  T c;
+}; // end constant_unary
+
+
+} // end detail
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                 InputIterator first,
+                                 InputIterator last,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value)
+{
+  typedef typename thrust::iterator_traits<OutputIterator>::value_type OutputType;
+
+  detail::new_value_if<Predicate,T,OutputType> op(pred,new_value);
+  return thrust::transform(exec, first, last, result, op);
+} // end replace_copy_if()
+
+
+template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputIterator, typename Predicate, typename T>
+__host__ __device__
+  OutputIterator replace_copy_if(thrust::execution_policy<DerivedPolicy> &exec,
+                                 InputIterator1 first,
+                                 InputIterator1 last,
+                                 InputIterator2 stencil,
+                                 OutputIterator result,
+                                 Predicate pred,
+                                 const T &new_value)
+{
+  typedef typename thrust::iterator_traits<OutputIterator>::value_type OutputType;
+
+  detail::new_value_if<Predicate,T,OutputType> op(pred,new_value);
+  return thrust::transform(exec, first, last, stencil, result, op);
+} // end replace_copy_if()
+
+
+template<typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename T>
+__host__ __device__
+  OutputIterator replace_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                              InputIterator first,
+                              InputIterator last,
+                              OutputIterator result,
+                              const T &old_value,
+                              const T &new_value)
+{
+  using thrust::placeholders::_1;
+
+  return thrust::replace_copy_if(exec, first, last, result, _1 == old_value, new_value);
+} // end replace_copy()
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  ForwardIterator first,
+                  ForwardIterator last,
+                  Predicate pred,
+                  const T &new_value)
+{
+  detail::constant_unary<T> f(new_value);
+  thrust::transform_if(exec, first, last, first, first, f, pred);
+} // end replace_if()
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename InputIterator, typename Predicate, typename T>
+__host__ __device__
+  void replace_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  ForwardIterator first,
+                  ForwardIterator last,
+                  InputIterator stencil,
+                  Predicate pred,
+                  const T &new_value)
+{
+  detail::constant_unary<T> f(new_value);
+  thrust::transform_if(exec, first, last, stencil, first, f, pred);
+} // end replace_if()
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void replace(thrust::execution_policy<DerivedPolicy> &exec,
+               ForwardIterator first,
+               ForwardIterator last,
+               const T &old_value,
+               const T &new_value)
+{
+  using thrust::placeholders::_1;
+
+  return thrust::replace_if(exec, first, last, _1 == old_value, new_value);
+} // end replace()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.h
new file mode 100644
index 000000000000..65c77ae75024
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.h
@@ -0,0 +1,55 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename BidirectionalIterator>
+__host__ __device__
+  void reverse(thrust::execution_policy<DerivedPolicy> &exec,
+               BidirectionalIterator first,
+               BidirectionalIterator last);
+
+
+template<typename DerivedPolicy,
+         typename BidirectionalIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator reverse_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                              BidirectionalIterator first,
+                              BidirectionalIterator last,
+                              OutputIterator result);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/reverse.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.inl
new file mode 100644
index 000000000000..1ce6db38b988
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/reverse.inl
@@ -0,0 +1,76 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/reverse.h>
+#include <thrust/advance.h>
+#include <thrust/distance.h>
+#include <thrust/detail/copy.h>
+#include <thrust/swap.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/reverse_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy, typename BidirectionalIterator>
+__host__ __device__
+  void reverse(thrust::execution_policy<ExecutionPolicy> &exec,
+               BidirectionalIterator first,
+               BidirectionalIterator last)
+{
+  typedef typename thrust::iterator_difference<BidirectionalIterator>::type difference_type;
+
+  // find the midpoint of [first,last)
+  difference_type N = thrust::distance(first, last);
+  BidirectionalIterator mid(first);
+  thrust::advance(mid, N / 2);
+
+  // swap elements of [first,mid) with [last - 1, mid)
+  thrust::swap_ranges(exec, first, mid, thrust::make_reverse_iterator(last));
+} // end reverse()
+
+
+template<typename ExecutionPolicy,
+         typename BidirectionalIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator reverse_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                              BidirectionalIterator first,
+                              BidirectionalIterator last,
+                              OutputIterator result)
+{
+  return thrust::copy(exec,
+                      thrust::make_reverse_iterator(last),
+                      thrust::make_reverse_iterator(first),
+                      result);
+} // end reverse_copy()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.h
new file mode 100644
index 000000000000..3e019c2233e2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.h
@@ -0,0 +1,84 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+namespace detail
+{
+
+namespace generic
+{
+
+namespace scalar
+{
+
+template<typename RandomAccessIterator, typename Size, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator lower_bound_n(RandomAccessIterator first,
+                                   Size n,
+                                   const T &val,
+                                   BinaryPredicate comp);
+
+template<typename RandomAccessIterator, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator lower_bound(RandomAccessIterator first, RandomAccessIterator last,
+                                 const T &val,
+                                 BinaryPredicate comp);
+
+template<typename RandomAccessIterator, typename Size, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator upper_bound_n(RandomAccessIterator first,
+                                   Size n,
+                                   const T &val,
+                                   BinaryPredicate comp);
+
+template<typename RandomAccessIterator, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator upper_bound(RandomAccessIterator first, RandomAccessIterator last,
+                                 const T &val,
+                                 BinaryPredicate comp);
+
+template<typename RandomAccessIterator, typename T, typename BinaryPredicate>
+__host__ __device__
+  pair<RandomAccessIterator,RandomAccessIterator>
+    equal_range(RandomAccessIterator first, RandomAccessIterator last,
+                const T &val,
+                BinaryPredicate comp);
+
+template<typename RandomAccessIterator, typename T, typename Compare>
+__host__ __device__
+bool binary_search(RandomAccessIterator first, RandomAccessIterator last, const T &value, Compare comp);
+
+} // end scalar
+
+} // end generic
+
+} // end detail
+
+} // end system
+
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/scalar/binary_search.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.inl
new file mode 100644
index 000000000000..61c71fba4df3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scalar/binary_search.inl
@@ -0,0 +1,157 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/pair.h>
+#include <thrust/detail/function.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+namespace detail
+{
+
+namespace generic
+{
+
+namespace scalar
+{
+
+template<typename RandomAccessIterator, typename Size, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator lower_bound_n(RandomAccessIterator first,
+                                   Size n,
+                                   const T &val,
+                                   BinaryPredicate comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    BinaryPredicate,
+    bool
+  > wrapped_comp(comp);
+
+  Size start = 0, i;
+  while(start < n)
+  {
+    i = start + (n - start) / 2;  // Overflow-safe variant of (a+b)/2
+    if(wrapped_comp(first[i], val))
+    {
+      start = i + 1;
+    }
+    else
+    {
+      n = i;
+    }
+  } // end while
+
+  return first + start;
+}
+
+// XXX generalize these upon implementation of scalar::distance & scalar::advance
+
+template<typename RandomAccessIterator, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator lower_bound(RandomAccessIterator first, RandomAccessIterator last,
+                                 const T &val,
+                                 BinaryPredicate comp)
+{
+  typename thrust::iterator_difference<RandomAccessIterator>::type n = last - first;
+  return lower_bound_n(first, n, val, comp);
+}
+
+template<typename RandomAccessIterator, typename Size, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator upper_bound_n(RandomAccessIterator first,
+                                   Size n,
+                                   const T &val,
+                                   BinaryPredicate comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    BinaryPredicate,
+    bool
+  > wrapped_comp(comp);
+
+  Size start = 0, i;
+  while(start < n)
+  {
+    i = start + (n - start) / 2;  // Overflow-safe variant of (a+b)/2
+    if(wrapped_comp(val, first[i]))
+    {
+      n = i;
+    }
+    else
+    {
+      start = i + 1;
+    }
+  } // end while
+
+  return first + start;
+}
+
+template<typename RandomAccessIterator, typename T, typename BinaryPredicate>
+__host__ __device__
+RandomAccessIterator upper_bound(RandomAccessIterator first, RandomAccessIterator last,
+                                 const T &val,
+                                 BinaryPredicate comp)
+{
+  typename thrust::iterator_difference<RandomAccessIterator>::type n = last - first;
+  return upper_bound_n(first, n, val, comp);
+}
+
+template<typename RandomAccessIterator, typename T, typename BinaryPredicate>
+__host__ __device__
+  pair<RandomAccessIterator,RandomAccessIterator>
+    equal_range(RandomAccessIterator first, RandomAccessIterator last,
+                const T &val,
+                BinaryPredicate comp)
+{
+  RandomAccessIterator lb = thrust::system::detail::generic::scalar::lower_bound(first, last, val, comp);
+  return thrust::make_pair(lb, thrust::system::detail::generic::scalar::upper_bound(lb, last, val, comp));
+}
+
+
+template<typename RandomAccessIterator, typename T, typename Compare>
+__host__ __device__
+bool binary_search(RandomAccessIterator first, RandomAccessIterator last, const T &value, Compare comp)
+{
+  RandomAccessIterator iter = thrust::system::detail::generic::scalar::lower_bound(first, last, value, comp);
+
+  // wrap comp
+  thrust::detail::wrapped_function<
+    Compare,
+    bool
+  > wrapped_comp(comp);
+
+  return iter != last && !wrapped_comp(value,*iter);
+}
+
+} // end scalar
+
+} // end generic
+
+} // end detail
+
+} // end system
+
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/scalar/binary_search.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.h
new file mode 100644
index 000000000000..476441ab6f64
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.h
@@ -0,0 +1,98 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result);
+
+
+// XXX it is an error to call this function; it has no implementation 
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator inclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                BinaryFunction binary_op);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init);
+
+
+// XXX it is an error to call this function; it has no implementation 
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init,
+                                BinaryFunction binary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/scan.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.inl
new file mode 100644
index 000000000000..45a2aadd087b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan.inl
@@ -0,0 +1,127 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/static_assert.h>
+#include <thrust/system/detail/generic/scan.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/scan.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/iterator/is_output_iterator.h>
+#include <thrust/functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result)
+{
+  // assume plus as the associative operator
+  return thrust::inclusive_scan(exec, first, last, result, thrust::plus<>());
+} // end inclusive_scan()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result)
+{
+  // Use the input iterator's value type per https://wg21.link/P0571
+  using ValueType = typename thrust::iterator_value<InputIterator>::type;
+  return thrust::exclusive_scan(exec, first, last, result, ValueType{});
+} // end exclusive_scan()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                T init)
+{
+  // assume plus as the associative operator
+  return thrust::exclusive_scan(exec, first, last, result, init, thrust::plus<>());
+} // end exclusive_scan()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator inclusive_scan(thrust::execution_policy<ExecutionPolicy> &,
+                                InputIterator,
+                                InputIterator,
+                                OutputIterator result,
+                                BinaryFunction)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end inclusive_scan
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename T,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator exclusive_scan(thrust::execution_policy<ExecutionPolicy> &,
+                                InputIterator,
+                                InputIterator,
+                                OutputIterator result,
+                                T,
+                                BinaryFunction)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end exclusive_scan()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.h
new file mode 100644
index 000000000000..9e38ac933923
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.h
@@ -0,0 +1,143 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file scan_by_key.h
+ *  \brief Generic implementations of key-value scans.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/scan_by_key.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.inl
new file mode 100644
index 000000000000..0e3100224a4d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scan_by_key.inl
@@ -0,0 +1,245 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cstdint.h>
+#include <thrust/system/detail/generic/scan_by_key.h>
+#include <thrust/functional.h>
+#include <thrust/transform.h>
+#include <thrust/replace.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/scan.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+template <typename OutputType, typename HeadFlagType, typename AssociativeOperator>
+struct segmented_scan_functor
+{
+  AssociativeOperator binary_op;
+
+  typedef typename thrust::tuple<OutputType, HeadFlagType> result_type;
+
+  __host__ __device__
+  segmented_scan_functor(AssociativeOperator _binary_op) : binary_op(_binary_op) {}
+
+  __host__ __device__
+  result_type operator()(result_type a, result_type b)
+  {
+    return result_type(thrust::get<1>(b) ? thrust::get<0>(b) : binary_op(thrust::get<0>(a), thrust::get<0>(b)),
+                       thrust::get<1>(a) | thrust::get<1>(b));
+  }
+};
+
+
+} // end namespace detail
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result)
+{
+  return thrust::inclusive_scan_by_key(exec, first1, last1, first2, result, thrust::equal_to<>());
+}
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred)
+{
+  return thrust::inclusive_scan_by_key(exec, first1, last1, first2, result, binary_pred, thrust::plus<>());
+}
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op)
+{
+  using OutputType = typename thrust::iterator_traits<InputIterator2>::value_type;
+  using HeadFlagType = thrust::detail::uint8_t;
+
+  const size_t n = last1 - first1;
+
+  if(n != 0)
+  {
+    // compute head flags
+    thrust::detail::temporary_array<HeadFlagType,DerivedPolicy> flags(exec, n);
+    flags[0] = 1; thrust::transform(exec, first1, last1 - 1, first1 + 1, flags.begin() + 1, thrust::detail::not2(binary_pred));
+
+    // scan key-flag tuples,
+    // For additional details refer to Section 2 of the following paper
+    //    S. Sengupta, M. Harris, and M. Garland. "Efficient parallel scan algorithms for GPUs"
+    //    NVIDIA Technical Report NVR-2008-003, December 2008
+    //    http://mgarland.org/files/papers/nvr-2008-003.pdf
+    thrust::inclusive_scan(exec,
+                           thrust::make_zip_iterator(thrust::make_tuple(first2, flags.begin())),
+                           thrust::make_zip_iterator(thrust::make_tuple(first2, flags.begin())) + n,
+                           thrust::make_zip_iterator(thrust::make_tuple(result, flags.begin())),
+                           detail::segmented_scan_functor<OutputType, HeadFlagType, AssociativeOperator>(binary_op));
+  }
+
+  return result + n;
+}
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result)
+{
+  typedef typename thrust::iterator_traits<InputIterator2>::value_type InitType;
+  return thrust::exclusive_scan_by_key(exec, first1, last1, first2, result, InitType{});
+}
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init)
+{
+  return thrust::exclusive_scan_by_key(exec, first1, last1, first2, result, init, thrust::equal_to<>());
+}
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred)
+{
+  return thrust::exclusive_scan_by_key(exec, first1, last1, first2, result, init, binary_pred, thrust::plus<>());
+}
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       AssociativeOperator binary_op)
+{
+  using OutputType = T;
+  using HeadFlagType = thrust::detail::uint8_t;
+
+  const size_t n = last1 - first1;
+
+  if(n != 0)
+  {
+    InputIterator2 last2 = first2 + n;
+
+    // compute head flags
+    thrust::detail::temporary_array<HeadFlagType,DerivedPolicy> flags(exec, n);
+    flags[0] = 1; thrust::transform(exec, first1, last1 - 1, first1 + 1, flags.begin() + 1, thrust::detail::not2(binary_pred));
+
+    // shift input one to the right and initialize segments with init
+    thrust::detail::temporary_array<OutputType,DerivedPolicy> temp(exec, n);
+    thrust::replace_copy_if(exec, first2, last2 - 1, flags.begin() + 1, temp.begin() + 1, thrust::negate<HeadFlagType>(), init);
+    temp[0] = init;
+
+    // scan key-flag tuples,
+    // For additional details refer to Section 2 of the following paper
+    //    S. Sengupta, M. Harris, and M. Garland. "Efficient parallel scan algorithms for GPUs"
+    //    NVIDIA Technical Report NVR-2008-003, December 2008
+    //    http://mgarland.org/files/papers/nvr-2008-003.pdf
+    thrust::inclusive_scan(exec,
+                           thrust::make_zip_iterator(thrust::make_tuple(temp.begin(), flags.begin())),
+                           thrust::make_zip_iterator(thrust::make_tuple(temp.begin(), flags.begin())) + n,
+                           thrust::make_zip_iterator(thrust::make_tuple(result,       flags.begin())),
+                           detail::segmented_scan_functor<OutputType, HeadFlagType, AssociativeOperator>(binary_op));
+  }
+
+  return result + n;
+}
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.h
new file mode 100644
index 000000000000..6bb7949efa46
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.h
@@ -0,0 +1,80 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter(thrust::execution_policy<DerivedPolicy> &exec,
+               InputIterator1 first,
+               InputIterator1 last,
+               InputIterator2 map,
+               RandomAccessIterator output);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator,
+         typename Predicate>
+__host__ __device__
+  void scatter_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output,
+                  Predicate pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/scatter.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.inl
new file mode 100644
index 000000000000..5b479870897d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/scatter.inl
@@ -0,0 +1,97 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/scatter.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/functional.h>
+#include <thrust/transform.h>
+#include <thrust/iterator/permutation_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter(thrust::execution_policy<DerivedPolicy> &exec,
+               InputIterator1 first,
+               InputIterator1 last,
+               InputIterator2 map,
+               RandomAccessIterator output)
+{
+  thrust::transform(exec,
+                    first,
+                    last,
+                    thrust::make_permutation_iterator(output, map),
+                    thrust::identity<typename thrust::iterator_value<InputIterator1>::type>());
+} // end scatter()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator>
+__host__ __device__
+  void scatter_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output)
+{
+  // default predicate is identity
+  typedef typename thrust::iterator_value<InputIterator3>::type StencilType;
+  thrust::scatter_if(exec, first, last, map, stencil, output, thrust::identity<StencilType>());
+} // end scatter_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename RandomAccessIterator,
+         typename Predicate>
+__host__ __device__
+  void scatter_if(thrust::execution_policy<DerivedPolicy> &exec,
+                  InputIterator1 first,
+                  InputIterator1 last,
+                  InputIterator2 map,
+                  InputIterator3 stencil,
+                  RandomAccessIterator output,
+                  Predicate pred)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type InputType;
+  thrust::transform_if(exec, first, last, stencil, thrust::make_permutation_iterator(output, map), thrust::identity<InputType>(), pred);
+} // end scatter_if()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.h
new file mode 100644
index 000000000000..7619b80e5ad1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.h
@@ -0,0 +1,124 @@
+
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/iterator/detail/device_system_tag.h>
+#include <thrust/iterator/detail/any_system_tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename Tag>
+  struct select_system1_exists;
+
+template<typename Tag1, typename Tag2>
+  struct select_system2_exists;
+
+template<typename Tag1, typename Tag2, typename Tag3>
+  struct select_system3_exists;
+
+template<typename Tag1, typename Tag2, typename Tag3, typename Tag4>
+  struct select_system4_exists;
+
+template<typename Tag1, typename Tag2, typename Tag3, typename Tag4, typename Tag5>
+  struct select_system5_exists;
+
+template<typename Tag1, typename Tag2, typename Tag3, typename Tag4, typename Tag5, typename Tag6>
+  struct select_system6_exists;
+
+template<typename System>
+__host__ __device__
+  typename thrust::detail::disable_if<
+    select_system1_exists<System>::value,
+    System &
+  >::type
+    select_system(thrust::execution_policy<System> &system);
+
+template<typename System1, typename System2>
+__host__ __device__
+  typename thrust::detail::enable_if_defined<
+    thrust::detail::minimum_system<System1,System2>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2);
+
+template<typename System1, typename System2, typename System3>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system3_exists<System1,System2,System3>::value,
+    thrust::detail::minimum_system<System1,System2,System3>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3);
+
+template<typename System1, typename System2, typename System3, typename System4>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system4_exists<System1,System2,System3,System4>::value,
+    thrust::detail::minimum_system<System1,System2,System3,System4>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3,
+                   thrust::execution_policy<System4> &system4);
+
+template<typename System1, typename System2, typename System3, typename System4, typename System5>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system5_exists<System1,System2,System3,System4,System5>::value,
+    thrust::detail::minimum_system<System1,System2,System3,System4,System5>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3,
+                   thrust::execution_policy<System4> &system4,
+                   thrust::execution_policy<System5> &system5);
+
+template<typename System1, typename System2, typename System3, typename System4, typename System5, typename System6>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system6_exists<System1,System2,System3,System4,System5,System6>::value,
+    thrust::detail::minimum_system<System1,System2,System3,System4,System5,System6>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3,
+                   thrust::execution_policy<System4> &system4,
+                   thrust::execution_policy<System5> &system5,
+                   thrust::execution_policy<System6> &system6);
+
+// Map a single any_system_tag to device_system_tag.
+inline __host__ __device__
+thrust::device_system_tag select_system(thrust::any_system_tag);
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/select_system.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.inl
new file mode 100644
index 000000000000..b69d17c45075
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system.inl
@@ -0,0 +1,178 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/system/detail/generic/select_system_exists.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace select_system_detail
+{
+
+
+// min_system case 1: both systems have the same type, just return the first one
+template<typename System>
+__host__ __device__
+System &min_system(thrust::execution_policy<System> &system1,
+                   thrust::execution_policy<System> &)
+{
+  return thrust::detail::derived_cast(system1);
+} // end min_system()
+
+
+// min_system case 2: systems have differing type and the first type is considered the minimum
+template<typename System1, typename System2>
+__host__ __device__
+typename thrust::detail::enable_if<
+  thrust::detail::is_same<
+    System1,
+    typename thrust::detail::minimum_system<System1,System2>::type
+  >::value,
+  System1 &
+>::type
+  min_system(thrust::execution_policy<System1> &system1, thrust::execution_policy<System2> &)
+{
+  return thrust::detail::derived_cast(system1);
+} // end min_system()
+
+
+// min_system case 3: systems have differing type and the second type is considered the minimum
+template<typename System1, typename System2>
+__host__ __device__
+typename thrust::detail::enable_if<
+  thrust::detail::is_same<
+    System2,
+    typename thrust::detail::minimum_system<System1,System2>::type
+  >::value,
+    System2 &
+  >::type
+    min_system(thrust::execution_policy<System1> &, thrust::execution_policy<System2> &system2)
+{
+  return thrust::detail::derived_cast(system2);
+} // end min_system()
+
+
+} // end select_system_detail
+
+
+template<typename System>
+__host__ __device__
+  typename thrust::detail::disable_if<
+    select_system1_exists<System>::value,
+    System &
+  >::type
+    select_system(thrust::execution_policy<System> &system)
+{
+  return thrust::detail::derived_cast(system);
+} // end select_system()
+
+
+template<typename System1, typename System2>
+__host__ __device__
+  typename thrust::detail::enable_if_defined<
+    thrust::detail::minimum_system<System1,System2>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2)
+{
+  return select_system_detail::min_system(system1,system2);
+} // end select_system()
+
+
+template<typename System1, typename System2, typename System3>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system3_exists<System1,System2,System3>::value,
+    thrust::detail::minimum_system<System1,System2,System3>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3)
+{
+  return select_system(select_system(system1,system2), system3);
+} // end select_system()
+
+
+template<typename System1, typename System2, typename System3, typename System4>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system4_exists<System1,System2,System3,System4>::value,
+    thrust::detail::minimum_system<System1,System2,System3,System4>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3,
+                   thrust::execution_policy<System4> &system4)
+{
+  return select_system(select_system(system1,system2,system3), system4);
+} // end select_system()
+
+
+template<typename System1, typename System2, typename System3, typename System4, typename System5>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system5_exists<System1,System2,System3,System4,System5>::value,
+    thrust::detail::minimum_system<System1,System2,System3,System4,System5>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3,
+                   thrust::execution_policy<System4> &system4,
+                   thrust::execution_policy<System5> &system5)
+{
+  return select_system(select_system(system1,system2,system3,system4), system5);
+} // end select_system()
+
+
+template<typename System1, typename System2, typename System3, typename System4, typename System5, typename System6>
+__host__ __device__
+  typename thrust::detail::lazy_disable_if<
+    select_system6_exists<System1,System2,System3,System4,System5,System6>::value,
+    thrust::detail::minimum_system<System1,System2,System3,System4,System5,System6>
+  >::type
+    &select_system(thrust::execution_policy<System1> &system1,
+                   thrust::execution_policy<System2> &system2,
+                   thrust::execution_policy<System3> &system3,
+                   thrust::execution_policy<System4> &system4,
+                   thrust::execution_policy<System5> &system5,
+                   thrust::execution_policy<System6> &system6)
+{
+  return select_system(select_system(system1,system2,system3,system4,system5), system6);
+} // end select_system()
+
+
+// map a single any_system_tag to device_system_tag
+inline __host__ __device__
+thrust::device_system_tag select_system(thrust::any_system_tag)
+{
+  return thrust::device_system_tag();
+} // end select_system()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system_exists.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system_exists.h
new file mode 100644
index 000000000000..29d05781d9ec
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/select_system_exists.h
@@ -0,0 +1,167 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file generic/type_traits.h
+ *  \brief Introspection for free functions defined in generic.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+// forward declaration of any_system_tag for any_conversion below
+struct any_system_tag;
+
+namespace system
+{
+namespace detail
+{
+
+// we must define these traits outside of generic's namespace
+namespace generic_type_traits_ns
+{
+
+typedef char yes;
+typedef char (&no)[2];
+
+struct any_conversion
+{
+  template<typename T> any_conversion(const T &);
+
+  // add this extra constructor to disambiguate conversion from any_system_tag
+  any_conversion(const any_system_tag &);
+};
+
+namespace select_system_exists_ns
+{
+  no select_system(const any_conversion &);
+  no select_system(const any_conversion &, const any_conversion &);
+  no select_system(const any_conversion &, const any_conversion &, const any_conversion &);
+  no select_system(const any_conversion &, const any_conversion &, const any_conversion &, const any_conversion &);
+  no select_system(const any_conversion &, const any_conversion &, const any_conversion &, const any_conversion &, const any_conversion &);
+  no select_system(const any_conversion &, const any_conversion &, const any_conversion &, const any_conversion &, const any_conversion &, const any_conversion &);
+
+  template<typename T> yes check(const T &);
+
+  no check(no);
+
+  template<typename Tag>
+    struct select_system1_exists
+  {
+    static Tag &tag;
+
+    static const bool value = sizeof(check(select_system(tag))) == sizeof(yes);
+  };
+
+  template<typename Tag1, typename Tag2>
+    struct select_system2_exists
+  {
+    static Tag1 &tag1;
+    static Tag2 &tag2;
+
+    static const bool value = sizeof(check(select_system(tag1,tag2))) == sizeof(yes);
+  };
+
+  template<typename Tag1, typename Tag2, typename Tag3>
+    struct select_system3_exists
+  {
+    static Tag1 &tag1;
+    static Tag2 &tag2;
+    static Tag3 &tag3;
+
+    static const bool value = sizeof(check(select_system(tag1,tag2,tag3))) == sizeof(yes);
+  };
+
+  template<typename Tag1, typename Tag2, typename Tag3, typename Tag4>
+    struct select_system4_exists
+  {
+    static Tag1 &tag1;
+    static Tag2 &tag2;
+    static Tag3 &tag3;
+    static Tag4 &tag4;
+
+    static const bool value = sizeof(check(select_system(tag1,tag2,tag3,tag4))) == sizeof(yes);
+  };
+
+  template<typename Tag1, typename Tag2, typename Tag3, typename Tag4, typename Tag5>
+    struct select_system5_exists
+  {
+    static Tag1 &tag1;
+    static Tag2 &tag2;
+    static Tag3 &tag3;
+    static Tag4 &tag4;
+    static Tag5 &tag5;
+
+    static const bool value = sizeof(check(select_system(tag1,tag2,tag3,tag4,tag5))) == sizeof(yes);
+  };
+
+  template<typename Tag1, typename Tag2, typename Tag3, typename Tag4, typename Tag5, typename Tag6>
+    struct select_system6_exists
+  {
+    static Tag1 &tag1;
+    static Tag2 &tag2;
+    static Tag3 &tag3;
+    static Tag4 &tag4;
+    static Tag5 &tag5;
+    static Tag6 &tag6;
+
+    static const bool value = sizeof(check(select_system(tag1,tag2,tag3,tag4,tag5,tag6))) == sizeof(yes);
+  };
+} // end select_system_exists_ns
+
+} // end generic_type_traits_ns
+
+namespace generic
+{
+
+template<typename Tag>
+  struct select_system1_exists
+    : generic_type_traits_ns::select_system_exists_ns::select_system1_exists<Tag>
+{};
+
+template<typename Tag1, typename Tag2>
+  struct select_system2_exists
+    : generic_type_traits_ns::select_system_exists_ns::select_system2_exists<Tag1,Tag2>
+{};
+
+template<typename Tag1, typename Tag2, typename Tag3>
+  struct select_system3_exists
+    : generic_type_traits_ns::select_system_exists_ns::select_system3_exists<Tag1,Tag2,Tag3>
+{};
+
+template<typename Tag1, typename Tag2, typename Tag3, typename Tag4>
+  struct select_system4_exists
+    : generic_type_traits_ns::select_system_exists_ns::select_system4_exists<Tag1,Tag2,Tag3,Tag4>
+{};
+
+template<typename Tag1, typename Tag2, typename Tag3, typename Tag4, typename Tag5>
+  struct select_system5_exists
+    : generic_type_traits_ns::select_system_exists_ns::select_system5_exists<Tag1,Tag2,Tag3,Tag4,Tag5>
+{};
+
+template<typename Tag1, typename Tag2, typename Tag3, typename Tag4, typename Tag5, typename Tag6>
+  struct select_system6_exists
+    : generic_type_traits_ns::select_system_exists_ns::select_system6_exists<Tag1,Tag2,Tag3,Tag4,Tag5,Tag6>
+{};
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.h
new file mode 100644
index 000000000000..26bf17bb8ba2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.h
@@ -0,0 +1,63 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+  void sequence(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last);
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init);
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init,
+                T step);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/sequence.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.inl
new file mode 100644
index 000000000000..0e11dd75d77e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sequence.inl
@@ -0,0 +1,106 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/sequence.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/tabulate.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  void sequence(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type T;
+
+  thrust::sequence(exec, first, last, T(0));
+} // end sequence()
+
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init)
+{
+  thrust::sequence(exec, first, last, init, T(1));
+} // end sequence()
+
+namespace detail
+{
+template <typename T, typename = void>
+struct compute_sequence_value
+{
+  T init;
+  T step;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  T operator()(std::size_t i) const
+  {
+    return init + step * i;
+  }
+};
+template <typename T>
+struct compute_sequence_value<T, typename std::enable_if<std::is_arithmetic<T>::value>::type>
+{
+  T init;
+  T step;
+
+  __thrust_exec_check_disable__
+  __host__ __device__
+  T operator()(std::size_t i) const
+  {
+    return init + step * static_cast<T>(i);
+  }
+};
+}
+
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void sequence(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                T init,
+                T step)
+{
+
+  thrust::tabulate(exec,
+                   first,
+                   last,
+                   detail::compute_sequence_value<T>{std::move(init),
+                                                     std::move(step)});
+} // end sequence()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.h
new file mode 100644
index 000000000000..37665d78dd9c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.h
@@ -0,0 +1,318 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_difference(thrust::execution_policy<ExecutionPolicy> &exec,
+                              InputIterator1                             first1,
+                              InputIterator1                             last1,
+                              InputIterator2                             first2,
+                              InputIterator2                             last2,
+                              OutputIterator                             result);
+
+
+// XXX it is an error to call this function; it has no implementation
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_difference(thrust::execution_policy<ExecutionPolicy> &exec,
+                              InputIterator1                             first1,
+                              InputIterator1                             last1,
+                              InputIterator2                             first2,
+                              InputIterator2                             last2,
+                              OutputIterator                             result,
+                              StrictWeakOrdering                         comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_difference_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                        InputIterator1                             keys_first1,
+                        InputIterator1                             keys_last1,
+                        InputIterator2                             keys_first2,
+                        InputIterator2                             keys_last2,
+                        InputIterator3                             values_first1,
+                        InputIterator4                             values_first2,
+                        OutputIterator1                            keys_result,
+                        OutputIterator2                            values_result);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_difference_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                        InputIterator1                             keys_first1,
+                        InputIterator1                             keys_last1,
+                        InputIterator2                             keys_first2,
+                        InputIterator2                             keys_last2,
+                        InputIterator3                             values_first1,
+                        InputIterator4                             values_first2,
+                        OutputIterator1                            keys_result,
+                        OutputIterator2                            values_result,
+                        StrictWeakOrdering                         comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_intersection(thrust::execution_policy<ExecutionPolicy> &system,
+                                InputIterator1                             first1,
+                                InputIterator1                             last1,
+                                InputIterator2                             first2,
+                                InputIterator2                             last2,
+                                OutputIterator                             result);
+
+
+// XXX it is an error to call this function; it has no implementation
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_intersection(thrust::execution_policy<StrictWeakOrdering> &system,
+                                InputIterator1                                first1,
+                                InputIterator1                                last1,
+                                InputIterator2                                first2,
+                                InputIterator2                                last2,
+                                OutputIterator                                result,
+                                StrictWeakOrdering                            comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_intersection_by_key(thrust::execution_policy<ExecutionPolicy> &system,
+                          InputIterator1                             keys_first1,
+                          InputIterator1                             keys_last1,
+                          InputIterator2                             keys_first2,
+                          InputIterator2                             keys_last2,
+                          InputIterator3                             values_first1,
+                          OutputIterator1                            keys_result,
+                          OutputIterator2                            values_result);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_intersection_by_key(thrust::execution_policy<ExecutionPolicy> &system,
+                          InputIterator1                             keys_first1,
+                          InputIterator1                             keys_last1,
+                          InputIterator2                             keys_first2,
+                          InputIterator2                             keys_last2,
+                          InputIterator3                             values_first1,
+                          OutputIterator1                            keys_result,
+                          OutputIterator2                            values_result,
+                          StrictWeakOrdering                         comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_symmetric_difference(thrust::execution_policy<ExecutionPolicy> &system,
+                                        InputIterator1                             first1,
+                                        InputIterator1                             last1,
+                                        InputIterator2                             first2,
+                                        InputIterator2                             last2,
+                                        OutputIterator                             result);
+
+
+// XXX it is an error to call this function; it has no implementation
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_symmetric_difference(thrust::execution_policy<ExecutionPolicy> &system,
+                                        InputIterator1                             first1,
+                                        InputIterator1                             last1,
+                                        InputIterator2                             first2,
+                                        InputIterator2                             last2,
+                                        OutputIterator                             result,
+                                        StrictWeakOrdering                         comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_symmetric_difference_by_key(thrust::execution_policy<ExecutionPolicy> &system,
+                                  InputIterator1                             keys_first1,
+                                  InputIterator1                             keys_last1,
+                                  InputIterator2                             keys_first2,
+                                  InputIterator2                             keys_last2,
+                                  InputIterator3                             values_first1,
+                                  InputIterator4                             values_first2,
+                                  OutputIterator1                            keys_result,
+                                  OutputIterator2                            values_result);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_symmetric_difference_by_key(thrust::execution_policy<ExecutionPolicy> &system,
+                                  InputIterator1                             keys_first1,
+                                  InputIterator1                             keys_last1,
+                                  InputIterator2                             keys_first2,
+                                  InputIterator2                             keys_last2,
+                                  InputIterator3                             values_first1,
+                                  InputIterator4                             values_first2,
+                                  OutputIterator1                            keys_result,
+                                  OutputIterator2                            values_result,
+                                  StrictWeakOrdering                         comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_union(thrust::execution_policy<ExecutionPolicy> &system,
+                         InputIterator1                             first1,
+                         InputIterator1                             last1,
+                         InputIterator2                             first2,
+                         InputIterator2                             last2,
+                         OutputIterator                             result);
+
+
+// XXX it is an error to call this function; it has no implementation
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_union(thrust::execution_policy<ExecutionPolicy> &system,
+                         InputIterator1                             first1,
+                         InputIterator1                             last1,
+                         InputIterator2                             first2,
+                         InputIterator2                             last2,
+                         OutputIterator                             result,
+                         StrictWeakOrdering                         comp);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_union_by_key(thrust::execution_policy<ExecutionPolicy> &system,
+                   InputIterator1                             keys_first1,
+                   InputIterator1                             keys_last1,
+                   InputIterator2                             keys_first2,
+                   InputIterator2                             keys_last2,
+                   InputIterator3                             values_first1,
+                   InputIterator4                             values_first2,
+                   OutputIterator1                            keys_result,
+                   OutputIterator2                            values_result);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_union_by_key(thrust::execution_policy<ExecutionPolicy> &system,
+                   InputIterator1                             keys_first1,
+                   InputIterator1                             keys_last1,
+                   InputIterator2                             keys_first2,
+                   InputIterator2                             keys_last2,
+                   InputIterator3                             values_first1,
+                   InputIterator4                             values_first2,
+                   OutputIterator1                            keys_result,
+                   OutputIterator2                            values_result,
+                   StrictWeakOrdering                         comp);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/set_operations.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.inl
new file mode 100644
index 000000000000..4363be5c0d8a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/set_operations.inl
@@ -0,0 +1,476 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/static_assert.h>
+#include <thrust/system/detail/generic/set_operations.h>
+#include <thrust/functional.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/constant_iterator.h>
+#include <thrust/iterator/zip_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_difference(thrust::execution_policy<DerivedPolicy> &exec,
+                              InputIterator1                           first1,
+                              InputIterator1                           last1,
+                              InputIterator2                           first2,
+                              InputIterator2                           last2,
+                              OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_difference(exec, first1, last1, first2, last2, result, thrust::less<value_type>());
+} // end set_difference()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_difference_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                        InputIterator1                           keys_first1,
+                        InputIterator1                           keys_last1,
+                        InputIterator2                           keys_first2,
+                        InputIterator2                           keys_last2,
+                        InputIterator3                           values_first1,
+                        InputIterator4                           values_first2,
+                        OutputIterator1                          keys_result,
+                        OutputIterator2                          values_result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_difference_by_key(exec, keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, thrust::less<value_type>());
+} // end set_difference_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_difference_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                        InputIterator1                           keys_first1,
+                        InputIterator1                           keys_last1,
+                        InputIterator2                           keys_first2,
+                        InputIterator2                           keys_last2,
+                        InputIterator3                           values_first1,
+                        InputIterator4                           values_first2,
+                        OutputIterator1                          keys_result,
+                        OutputIterator2                          values_result,
+                        StrictWeakOrdering                       comp)
+{
+  typedef thrust::tuple<InputIterator1, InputIterator3>   iterator_tuple1;
+  typedef thrust::tuple<InputIterator2, InputIterator4>   iterator_tuple2;
+  typedef thrust::tuple<OutputIterator1, OutputIterator2> iterator_tuple3;
+
+  typedef thrust::zip_iterator<iterator_tuple1> zip_iterator1;
+  typedef thrust::zip_iterator<iterator_tuple2> zip_iterator2;
+  typedef thrust::zip_iterator<iterator_tuple3> zip_iterator3;
+
+  zip_iterator1 zipped_first1 = thrust::make_zip_iterator(thrust::make_tuple(keys_first1, values_first1));
+  zip_iterator1 zipped_last1  = thrust::make_zip_iterator(thrust::make_tuple(keys_last1, values_first1));
+
+  zip_iterator2 zipped_first2 = thrust::make_zip_iterator(thrust::make_tuple(keys_first2, values_first2));
+  zip_iterator2 zipped_last2  = thrust::make_zip_iterator(thrust::make_tuple(keys_last2, values_first2));
+
+  zip_iterator3 zipped_result = thrust::make_zip_iterator(thrust::make_tuple(keys_result, values_result));
+
+  thrust::detail::compare_first<StrictWeakOrdering> comp_first(comp);
+
+  iterator_tuple3 result = thrust::set_difference(exec, zipped_first1, zipped_last1, zipped_first2, zipped_last2, zipped_result, comp_first).get_iterator_tuple();
+
+  return thrust::make_pair(thrust::get<0>(result), thrust::get<1>(result));
+} // end set_difference_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_intersection(thrust::execution_policy<DerivedPolicy> &exec,
+                                InputIterator1                           first1,
+                                InputIterator1                           last1,
+                                InputIterator2                           first2,
+                                InputIterator2                           last2,
+                                OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_intersection(exec, first1, last1, first2, last2, result, thrust::less<value_type>());
+} // end set_intersection()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_intersection_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator1                           keys_first1,
+                          InputIterator1                           keys_last1,
+                          InputIterator2                           keys_first2,
+                          InputIterator2                           keys_last2,
+                          InputIterator3                           values_first1,
+                          OutputIterator1                          keys_result,
+                          OutputIterator2                          values_result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_intersection_by_key(exec, keys_first1, keys_last1, keys_first2, keys_last2, values_first1, keys_result, values_result, thrust::less<value_type>());
+} // end set_intersection_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_intersection_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                          InputIterator1                           keys_first1,
+                          InputIterator1                           keys_last1,
+                          InputIterator2                           keys_first2,
+                          InputIterator2                           keys_last2,
+                          InputIterator3                           values_first1,
+                          OutputIterator1                          keys_result,
+                          OutputIterator2                          values_result,
+                          StrictWeakOrdering                       comp)
+{
+  typedef typename thrust::iterator_value<InputIterator3>::type value_type1;
+  typedef thrust::constant_iterator<value_type1>                constant_iterator;
+
+  typedef thrust::tuple<InputIterator1, InputIterator3>     iterator_tuple1;
+  typedef thrust::tuple<InputIterator2, constant_iterator>  iterator_tuple2;
+  typedef thrust::tuple<OutputIterator1, OutputIterator2>   iterator_tuple3;
+
+  typedef thrust::zip_iterator<iterator_tuple1> zip_iterator1;
+  typedef thrust::zip_iterator<iterator_tuple2> zip_iterator2;
+  typedef thrust::zip_iterator<iterator_tuple3> zip_iterator3;
+
+  // fabricate a values_first2 by repeating a default-constructed value_type1
+  // XXX assumes value_type1 is default-constructible
+  constant_iterator values_first2 = thrust::make_constant_iterator(value_type1());
+
+  zip_iterator1 zipped_first1 = thrust::make_zip_iterator(thrust::make_tuple(keys_first1, values_first1));
+  zip_iterator1 zipped_last1  = thrust::make_zip_iterator(thrust::make_tuple(keys_last1, values_first1));
+
+  zip_iterator2 zipped_first2 = thrust::make_zip_iterator(thrust::make_tuple(keys_first2, values_first2));
+  zip_iterator2 zipped_last2  = thrust::make_zip_iterator(thrust::make_tuple(keys_last2, values_first2));
+
+  zip_iterator3 zipped_result = thrust::make_zip_iterator(thrust::make_tuple(keys_result, values_result));
+
+  thrust::detail::compare_first<StrictWeakOrdering> comp_first(comp);
+
+  iterator_tuple3 result = thrust::set_intersection(exec, zipped_first1, zipped_last1, zipped_first2, zipped_last2, zipped_result, comp_first).get_iterator_tuple();
+
+  return thrust::make_pair(thrust::get<0>(result), thrust::get<1>(result));
+} // end set_intersection_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_symmetric_difference(thrust::execution_policy<DerivedPolicy> &exec,
+                                        InputIterator1                           first1,
+                                        InputIterator1                           last1,
+                                        InputIterator2                           first2,
+                                        InputIterator2                           last2,
+                                        OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_symmetric_difference(exec, first1, last1, first2, last2, result, thrust::less<value_type>());
+} // end set_symmetric_difference()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_symmetric_difference_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                  InputIterator1                           keys_first1,
+                                  InputIterator1                           keys_last1,
+                                  InputIterator2                           keys_first2,
+                                  InputIterator2                           keys_last2,
+                                  InputIterator3                           values_first1,
+                                  InputIterator4                           values_first2,
+                                  OutputIterator1                          keys_result,
+                                  OutputIterator2                          values_result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_symmetric_difference_by_key(exec, keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, thrust::less<value_type>());
+} // end set_symmetric_difference_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_symmetric_difference_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                                  InputIterator1                           keys_first1,
+                                  InputIterator1                           keys_last1,
+                                  InputIterator2                           keys_first2,
+                                  InputIterator2                           keys_last2,
+                                  InputIterator3                           values_first1,
+                                  InputIterator4                           values_first2,
+                                  OutputIterator1                          keys_result,
+                                  OutputIterator2                          values_result,
+                                  StrictWeakOrdering                       comp)
+{
+  typedef thrust::tuple<InputIterator1, InputIterator3>   iterator_tuple1;
+  typedef thrust::tuple<InputIterator2, InputIterator4>   iterator_tuple2;
+  typedef thrust::tuple<OutputIterator1, OutputIterator2> iterator_tuple3;
+
+  typedef thrust::zip_iterator<iterator_tuple1> zip_iterator1;
+  typedef thrust::zip_iterator<iterator_tuple2> zip_iterator2;
+  typedef thrust::zip_iterator<iterator_tuple3> zip_iterator3;
+
+  zip_iterator1 zipped_first1 = thrust::make_zip_iterator(thrust::make_tuple(keys_first1, values_first1));
+  zip_iterator1 zipped_last1  = thrust::make_zip_iterator(thrust::make_tuple(keys_last1, values_first1));
+
+  zip_iterator2 zipped_first2 = thrust::make_zip_iterator(thrust::make_tuple(keys_first2, values_first2));
+  zip_iterator2 zipped_last2  = thrust::make_zip_iterator(thrust::make_tuple(keys_last2, values_first2));
+
+  zip_iterator3 zipped_result = thrust::make_zip_iterator(thrust::make_tuple(keys_result, values_result));
+
+  thrust::detail::compare_first<StrictWeakOrdering> comp_first(comp);
+
+  iterator_tuple3 result = thrust::set_symmetric_difference(exec, zipped_first1, zipped_last1, zipped_first2, zipped_last2, zipped_result, comp_first).get_iterator_tuple();
+
+  return thrust::make_pair(thrust::get<0>(result), thrust::get<1>(result));
+} // end set_symmetric_difference_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator set_union(thrust::execution_policy<DerivedPolicy> &exec,
+                         InputIterator1                           first1,
+                         InputIterator1                           last1,
+                         InputIterator2                           first2,
+                         InputIterator2                           last2,
+                         OutputIterator                           result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_union(exec, first1, last1, first2, last2, result, thrust::less<value_type>());
+} // end set_union()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_union_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                   InputIterator1                           keys_first1,
+                   InputIterator1                           keys_last1,
+                   InputIterator2                           keys_first2,
+                   InputIterator2                           keys_last2,
+                   InputIterator3                           values_first1,
+                   InputIterator4                           values_first2,
+                   OutputIterator1                          keys_result,
+                   OutputIterator2                          values_result)
+{
+  typedef typename thrust::iterator_value<InputIterator1>::type value_type;
+  return thrust::set_union_by_key(exec, keys_first1, keys_last1, keys_first2, keys_last2, values_first1, values_first2, keys_result, values_result, thrust::less<value_type>());
+} // end set_union_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  set_union_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                   InputIterator1                           keys_first1,
+                   InputIterator1                           keys_last1,
+                   InputIterator2                           keys_first2,
+                   InputIterator2                           keys_last2,
+                   InputIterator3                           values_first1,
+                   InputIterator4                           values_first2,
+                   OutputIterator1                          keys_result,
+                   OutputIterator2                          values_result,
+                   StrictWeakOrdering                       comp)
+{
+  typedef thrust::tuple<InputIterator1, InputIterator3>   iterator_tuple1;
+  typedef thrust::tuple<InputIterator2, InputIterator4>   iterator_tuple2;
+  typedef thrust::tuple<OutputIterator1, OutputIterator2> iterator_tuple3;
+
+  typedef thrust::zip_iterator<iterator_tuple1> zip_iterator1;
+  typedef thrust::zip_iterator<iterator_tuple2> zip_iterator2;
+  typedef thrust::zip_iterator<iterator_tuple3> zip_iterator3;
+
+  zip_iterator1 zipped_first1 = thrust::make_zip_iterator(thrust::make_tuple(keys_first1, values_first1));
+  zip_iterator1 zipped_last1  = thrust::make_zip_iterator(thrust::make_tuple(keys_last1, values_first1));
+
+  zip_iterator2 zipped_first2 = thrust::make_zip_iterator(thrust::make_tuple(keys_first2, values_first2));
+  zip_iterator2 zipped_last2  = thrust::make_zip_iterator(thrust::make_tuple(keys_last2, values_first2));
+
+  zip_iterator3 zipped_result = thrust::make_zip_iterator(thrust::make_tuple(keys_result, values_result));
+
+  thrust::detail::compare_first<StrictWeakOrdering> comp_first(comp);
+
+  iterator_tuple3 result = thrust::set_union(exec, zipped_first1, zipped_last1, zipped_first2, zipped_last2, zipped_result, comp_first).get_iterator_tuple();
+
+  return thrust::make_pair(thrust::get<0>(result), thrust::get<1>(result));
+} // end set_union_by_key()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_difference(thrust::execution_policy<DerivedPolicy> &,
+                              InputIterator1,
+                              InputIterator1,
+                              InputIterator2,
+                              InputIterator2,
+                              OutputIterator  result,
+                              StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator1, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end set_difference()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_intersection(thrust::execution_policy<DerivedPolicy> &,
+                                InputIterator1,
+                                InputIterator1,
+                                InputIterator2,
+                                InputIterator2,
+                                OutputIterator result,
+                                StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator1, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end set_intersection()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_symmetric_difference(thrust::execution_policy<DerivedPolicy> &,
+                                        InputIterator1,
+                                        InputIterator1,
+                                        InputIterator2,
+                                        InputIterator2,
+                                        OutputIterator result,
+                                        StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator1, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end set_symmetric_difference()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator set_union(thrust::execution_policy<DerivedPolicy> &,
+                         InputIterator1,
+                         InputIterator1,
+                         InputIterator2,
+                         InputIterator2,
+                         OutputIterator result,
+                         StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<InputIterator1, false>::value)
+  , "unimplemented for this system"
+  );
+  return result;
+} // end set_union()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.h
new file mode 100644
index 000000000000..8f8e21afdf90
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.h
@@ -0,0 +1,54 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file shuffle.h
+ *  \brief Generic implementations of shuffle functions.
+ */
+
+#pragma once
+
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system {
+namespace detail {
+namespace generic {
+
+template <typename ExecutionPolicy, typename RandomIterator, typename URBG>
+__host__ __device__ void shuffle(
+    thrust::execution_policy<ExecutionPolicy>& exec, RandomIterator first,
+    RandomIterator last, URBG&& g);
+
+template <typename ExecutionPolicy, typename RandomIterator,
+          typename OutputIterator, typename URBG>
+__host__ __device__ void shuffle_copy(
+    thrust::execution_policy<ExecutionPolicy>& exec, RandomIterator first,
+    RandomIterator last, OutputIterator result, URBG&& g);
+
+}  // end namespace generic
+}  // end namespace detail
+}  // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/shuffle.inl>
+
+#endif
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.inl
new file mode 100644
index 000000000000..baece51bee69
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/shuffle.inl
@@ -0,0 +1,190 @@
+/*
+ *  Copyright 2008-20120 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/iterator/discard_iterator.h>
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/iterator/transform_output_iterator.h>
+#include <thrust/random.h>
+#include <thrust/scan.h>
+#include <thrust/system/detail/generic/shuffle.h>
+
+#include <cstdint>
+
+THRUST_NAMESPACE_BEGIN
+namespace system {
+namespace detail {
+namespace generic {
+
+// An implementation of a Feistel cipher for operating on 64 bit keys
+class feistel_bijection {
+  struct round_state {
+    std::uint32_t left;
+    std::uint32_t right;
+  };
+
+ public:
+  template <class URBG>
+  __host__ __device__ feistel_bijection(std::uint64_t m, URBG&& g) {
+    std::uint64_t total_bits = get_cipher_bits(m);
+    // Half bits rounded down
+    left_side_bits = total_bits / 2;
+    left_side_mask = (1ull << left_side_bits) - 1;
+    // Half the bits rounded up
+    right_side_bits = total_bits - left_side_bits;
+    right_side_mask = (1ull << right_side_bits) - 1;
+
+    for (std::uint32_t i = 0; i < num_rounds; i++) {
+      key[i] = g();
+    }
+  }
+
+  __host__ __device__ std::uint64_t nearest_power_of_two() const {
+    return 1ull << (left_side_bits + right_side_bits);
+  }
+
+  __host__ __device__ std::uint64_t operator()(const std::uint64_t val) const {
+    std::uint32_t state[2] = { static_cast<std::uint32_t>( val >> right_side_bits ), static_cast<std::uint32_t>( val & right_side_mask ) };
+    for( std::uint32_t i = 0; i < num_rounds; i++ )
+    {
+        std::uint32_t hi, lo;
+        constexpr std::uint64_t M0 = UINT64_C( 0xD2B74407B1CE6E93 );
+        mulhilo( M0, state[0], hi, lo );
+        lo = ( lo << ( right_side_bits - left_side_bits ) ) | state[1] >> left_side_bits;
+        state[0] = ( ( hi ^ key[i] ) ^ state[1] ) & left_side_mask;
+        state[1] = lo & right_side_mask;
+    }
+    // Combine the left and right sides together to get result
+    return static_cast<std::uint64_t>(state[0] << right_side_bits) | static_cast<std::uint64_t>(state[1]);
+  }
+
+ private:
+   // Perform 64 bit multiplication and save result in two 32 bit int
+   static __host__ __device__ void mulhilo( std::uint64_t a, std::uint64_t b, std::uint32_t& hi, std::uint32_t& lo )
+   {
+       std::uint64_t product = a * b;
+       hi = static_cast<std::uint32_t>( product >> 32 );
+       lo = static_cast<std::uint32_t>( product );
+   }
+
+  // Find the nearest power of two
+  static __host__ __device__ std::uint64_t get_cipher_bits(std::uint64_t m) {
+    if (m <= 16) return 4;
+    std::uint64_t i = 0;
+    m--;
+    while (m != 0) {
+      i++;
+      m >>= 1;
+    }
+    return i;
+  }
+
+  static constexpr std::uint32_t num_rounds = 24;
+  std::uint64_t right_side_bits;
+  std::uint64_t left_side_bits;
+  std::uint64_t right_side_mask;
+  std::uint64_t left_side_mask;
+  std::uint32_t key[num_rounds];
+};
+
+struct key_flag_tuple {
+  std::uint64_t key;
+  std::uint64_t flag;
+};
+
+// scan only flags
+struct key_flag_scan_op {
+  __host__ __device__ key_flag_tuple operator()(const key_flag_tuple& a,
+                                                const key_flag_tuple& b) {
+    return {b.key, a.flag + b.flag};
+  }
+};
+
+struct construct_key_flag_op {
+  std::uint64_t m;
+  feistel_bijection bijection;
+  __host__ __device__ construct_key_flag_op(std::uint64_t m,
+                                            feistel_bijection bijection)
+      : m(m), bijection(bijection) {}
+  __host__ __device__ key_flag_tuple operator()(std::uint64_t idx) {
+    auto gather_key = bijection(idx);
+    return key_flag_tuple{gather_key, (gather_key < m) ? 1ull : 0ull};
+  }
+};
+
+template <typename InputIterT, typename OutputIterT>
+struct write_output_op {
+  std::uint64_t m;
+  InputIterT in;
+  OutputIterT out;
+  // flag contains inclusive scan of valid keys
+  // perform gather using valid keys
+  __thrust_exec_check_disable__
+  __host__ __device__ std::size_t operator()(key_flag_tuple x) {
+    if (x.key < m) {
+      // -1 because inclusive scan
+      out[x.flag - 1] = in[x.key];
+    }
+    return 0;  // Discarded
+  }
+};
+
+template <typename ExecutionPolicy, typename RandomIterator, typename URBG>
+__host__ __device__ void shuffle(
+    thrust::execution_policy<ExecutionPolicy>& exec, RandomIterator first,
+    RandomIterator last, URBG&& g) {
+  using InputType = typename thrust::iterator_value_t<RandomIterator>;
+
+  // copy input to temp buffer
+  thrust::detail::temporary_array<InputType, ExecutionPolicy> temp(exec, first,
+                                                                   last);
+  thrust::shuffle_copy(exec, temp.begin(), temp.end(), first, g);
+}
+
+template <typename ExecutionPolicy, typename RandomIterator,
+          typename OutputIterator, typename URBG>
+__host__ __device__ void shuffle_copy(
+    thrust::execution_policy<ExecutionPolicy>& exec, RandomIterator first,
+    RandomIterator last, OutputIterator result, URBG&& g) {
+  // m is the length of the input
+  // we have an available bijection of length n via a feistel cipher
+  std::size_t m = last - first;
+  feistel_bijection bijection(m, g);
+  std::uint64_t n = bijection.nearest_power_of_two();
+
+  // perform stream compaction over length n bijection to get length m
+  // pseudorandom bijection over the original input
+  thrust::counting_iterator<std::uint64_t> indices(0);
+  thrust::transform_iterator<construct_key_flag_op, decltype(indices),
+                             key_flag_tuple>
+      key_flag_it(indices, construct_key_flag_op(m, bijection));
+  write_output_op<RandomIterator, decltype(result)> write_functor{m, first,
+                                                                  result};
+  auto gather_output_it = thrust::make_transform_output_iterator(
+      thrust::discard_iterator<std::size_t>(), write_functor);
+  // the feistel_bijection outputs a stream of permuted indices in range [0,n)
+  // flag each value < m and compact it, so we have a set of permuted indices in
+  // range [0,m) each thread gathers an input element according to its
+  // pseudorandom permuted index
+  thrust::inclusive_scan(exec, key_flag_it, key_flag_it + n, gather_output_it,
+                         key_flag_scan_op());
+}
+
+}  // end namespace generic
+}  // end namespace detail
+}  // end namespace system
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.h
new file mode 100644
index 000000000000..cd8d455624c8
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.h
@@ -0,0 +1,153 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+  void sort(thrust::execution_policy<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort(thrust::execution_policy<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last,
+            StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first,
+                   StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+  void stable_sort(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last);
+
+
+// XXX it is an error to call this function; it has no implementation
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last,
+                   StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void stable_sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first);
+
+
+// XXX it is an error to call this function; it has no implementation
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first,
+                          StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  bool is_sorted(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Compare>
+__host__ __device__
+  bool is_sorted(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 Compare comp);
+
+
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator is_sorted_until(thrust::execution_policy<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Compare>
+__host__ __device__
+  ForwardIterator is_sorted_until(thrust::execution_policy<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Compare comp);
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/sort.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.inl
new file mode 100644
index 000000000000..632cab435ea9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/sort.inl
@@ -0,0 +1,219 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/sort.h>
+#include <thrust/functional.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/distance.h>
+#include <thrust/functional.h>
+#include <thrust/find.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/tuple.h>
+#include <thrust/detail/internal_functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+  void sort(thrust::execution_policy<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type; 
+  thrust::sort(exec, first, last, thrust::less<value_type>());
+} // end sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort(thrust::execution_policy<DerivedPolicy> &exec,
+            RandomAccessIterator first,
+            RandomAccessIterator last,
+            StrictWeakOrdering comp)
+{
+  // implement with stable_sort
+  thrust::stable_sort(exec, first, last, comp);
+} // end sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type value_type;
+  thrust::sort_by_key(exec, keys_first, keys_last, values_first, thrust::less<value_type>());
+} // end sort_by_key()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator1 keys_first,
+                   RandomAccessIterator1 keys_last,
+                   RandomAccessIterator2 values_first,
+                   StrictWeakOrdering comp)
+{
+  // implement with stable_sort_by_key
+  thrust::stable_sort_by_key(exec, keys_first, keys_last, values_first, comp);
+} // end sort_by_key()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+  void stable_sort(thrust::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator last)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type;
+  thrust::stable_sort(exec, first, last, thrust::less<value_type>());
+} // end stable_sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+  void stable_sort_by_key(thrust::execution_policy<DerivedPolicy> &exec,
+                          RandomAccessIterator1 keys_first,
+                          RandomAccessIterator1 keys_last,
+                          RandomAccessIterator2 values_first)
+{
+  typedef typename iterator_value<RandomAccessIterator1>::type value_type;
+  thrust::stable_sort_by_key(exec, keys_first, keys_last, values_first, thrust::less<value_type>());
+} // end stable_sort_by_key()
+
+
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  bool is_sorted(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last)
+{
+  return thrust::is_sorted_until(exec, first, last) == last;
+} // end is_sorted()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Compare>
+__host__ __device__
+  bool is_sorted(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 Compare comp)
+{
+  return thrust::is_sorted_until(exec, first, last, comp) == last;
+} // end is_sorted()
+
+
+template<typename DerivedPolicy, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator is_sorted_until(thrust::execution_policy<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last)
+{
+  typedef typename thrust::iterator_value<ForwardIterator>::type InputType;
+
+  return thrust::is_sorted_until(exec, first, last, thrust::less<InputType>());
+} // end is_sorted_until()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Compare>
+__host__ __device__
+  ForwardIterator is_sorted_until(thrust::execution_policy<DerivedPolicy> &exec,
+                                  ForwardIterator first,
+                                  ForwardIterator last,
+                                  Compare comp)
+{
+  if(thrust::distance(first,last) < 2) return last;
+
+  typedef thrust::tuple<ForwardIterator,ForwardIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple>            ZipIterator;
+
+  ForwardIterator first_plus_one = first;
+  thrust::advance(first_plus_one, 1);
+
+  ZipIterator zipped_first = thrust::make_zip_iterator(thrust::make_tuple(first_plus_one, first));
+  ZipIterator zipped_last  = thrust::make_zip_iterator(thrust::make_tuple(last, first));
+
+  return thrust::get<0>(thrust::find_if(exec, zipped_first, zipped_last, thrust::detail::tuple_binary_predicate<Compare>(comp)).get_iterator_tuple());
+} // end is_sorted_until()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort(thrust::execution_policy<DerivedPolicy> &,
+                   RandomAccessIterator,
+                   RandomAccessIterator,
+                   StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<RandomAccessIterator, false>::value)
+  , "unimplemented for this system"
+  );
+} // end stable_sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+  void stable_sort_by_key(thrust::execution_policy<DerivedPolicy> &,
+                          RandomAccessIterator1,
+                          RandomAccessIterator1,
+                          RandomAccessIterator2,
+                          StrictWeakOrdering)
+{
+  THRUST_STATIC_ASSERT_MSG(
+    (thrust::detail::depend_on_instantiation<RandomAccessIterator1, false>::value)
+  , "unimplemented for this system"
+  );
+} // end stable_sort_by_key()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.h
new file mode 100644
index 000000000000..edb5acf31fb2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.h
@@ -0,0 +1,46 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  ForwardIterator2 swap_ranges(thrust::execution_policy<DerivedPolicy> &exec,
+                               ForwardIterator1 first1,
+                               ForwardIterator1 last1,
+                               ForwardIterator2 first2);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/swap_ranges.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.inl
new file mode 100644
index 000000000000..ea42df35bcde
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/swap_ranges.inl
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/swap_ranges.h>
+#include <thrust/tuple.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/for_each.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+
+// XXX define this here rather than in internal_functional.h
+// to avoid circular dependence between swap.h & internal_functional.h
+struct swap_pair_elements
+{
+  template <typename Tuple>
+  __host__ __device__
+  void operator()(Tuple t)
+  {
+    // use unqualified swap to allow ADL to catch any user-defined swap
+    using thrust::swap;
+    swap(thrust::get<0>(t), thrust::get<1>(t));
+  }
+}; // end swap_pair_elements
+
+
+} // end detail
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  ForwardIterator2 swap_ranges(thrust::execution_policy<DerivedPolicy> &exec,
+                               ForwardIterator1 first1,
+                               ForwardIterator1 last1,
+                               ForwardIterator2 first2)
+{
+  typedef thrust::tuple<ForwardIterator1,ForwardIterator2> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple>              ZipIterator;
+
+  ZipIterator result = thrust::for_each(exec,
+                                        thrust::make_zip_iterator(thrust::make_tuple(first1, first2)),
+                                        thrust::make_zip_iterator(thrust::make_tuple(last1,  first2)),
+                                        detail::swap_pair_elements());
+  return thrust::get<1>(result.get_iterator_tuple());
+} // end swap_ranges()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.h
new file mode 100644
index 000000000000..041093e82e62
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.h
@@ -0,0 +1,48 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename UnaryOperation>
+__host__ __device__
+  void tabulate(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                UnaryOperation unary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/tabulate.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.inl
new file mode 100644
index 000000000000..0fd2121c149b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tabulate.inl
@@ -0,0 +1,61 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tabulate.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/transform.h>
+#include <thrust/distance.h>
+#include <thrust/iterator/counting_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename UnaryOperation>
+__host__ __device__
+  void tabulate(thrust::execution_policy<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                UnaryOperation unary_op)
+{
+  typedef typename iterator_difference<ForwardIterator>::type difference_type;
+
+  // by default, counting_iterator uses a 64b difference_type on 32b platforms to avoid overflowing its counter.
+  // this causes problems when a zip_iterator is created in transform's implementation -- ForwardIterator is
+  // incremented by a 64b difference_type and some compilers warn
+  // to avoid this, specify the counting_iterator's difference_type to be the same as ForwardIterator's.
+  thrust::counting_iterator<difference_type, thrust::use_default, thrust::use_default, difference_type> iter(0);
+
+  thrust::transform(exec, iter, iter + thrust::distance(first, last), first, unary_op);
+} // end tabulate()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tag.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tag.h
new file mode 100644
index 000000000000..48f0947974f7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/tag.h
@@ -0,0 +1,47 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file generic/tag.h
+ *  \brief Implementation of the generic backend's tag.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+// tag exists only to make the generic entry points the least priority match
+// during ADL. tag should not be derived from and is constructible from anything
+struct tag
+{
+  template<typename T>
+  __host__ __device__ inline
+  tag(const T &) {}
+};
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.h
new file mode 100644
index 000000000000..6b7e01ff24bd
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.h
@@ -0,0 +1,57 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/pair.h>
+#include <thrust/detail/pointer.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+  thrust::pair<thrust::pointer<T,DerivedPolicy>, typename thrust::pointer<T,DerivedPolicy>::difference_type>
+    get_temporary_buffer(thrust::execution_policy<DerivedPolicy> &exec, typename thrust::pointer<T,DerivedPolicy>::difference_type n);
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  void return_temporary_buffer(thrust::execution_policy<DerivedPolicy> &exec, Pointer p, std::ptrdiff_t n);
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  void return_temporary_buffer(thrust::execution_policy<DerivedPolicy> &exec, Pointer p);
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/temporary_buffer.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.inl
new file mode 100644
index 000000000000..254c48cb9156
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/temporary_buffer.inl
@@ -0,0 +1,86 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/temporary_buffer.h>
+#include <thrust/detail/pointer.h>
+#include <thrust/detail/malloc_and_free.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename T, typename DerivedPolicy>
+__host__ __device__
+  thrust::pair<thrust::pointer<T,DerivedPolicy>, typename thrust::pointer<T,DerivedPolicy>::difference_type>
+    get_temporary_buffer(thrust::execution_policy<DerivedPolicy> &exec, typename thrust::pointer<T,DerivedPolicy>::difference_type n)
+{
+  thrust::pointer<T,DerivedPolicy> ptr = thrust::malloc<T>(exec, n);
+
+  // check for a failed malloc
+  if(!ptr.get())
+  {
+    n = 0;
+  } // end if
+
+  return thrust::make_pair(ptr, n);
+} // end get_temporary_buffer()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  void return_temporary_buffer(thrust::execution_policy<DerivedPolicy> &exec, Pointer p, std::ptrdiff_t)
+{
+  // If we are here, no user customization of the three-argument signature with
+  // a size parameter of `return_temporary_buffer` was found. There may be an
+  // old two-argument signature `return_temporary_buffer` though, so we make
+  // another ADL call to try and find one.
+  //
+  // The interface layer downcast and then did ADL dispatch - there were no
+  // matches for DerivedPolicy (aka no one customized the three-argument
+  // signature), so this overload got found an implicit upcast to
+  // `execution_policy<DerivedPolicy>` was done. Now, we're looking for a
+  // customization of the two-argument signature so we need to downcast again.
+  return_temporary_buffer(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), p);
+} // end return_temporary_buffer()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  void return_temporary_buffer(thrust::execution_policy<DerivedPolicy> &exec, Pointer p)
+{
+  // If we are here, no user customization of either the old two-argument
+  // signature or the new three-argument signature with a size parameter of
+  // `return_temporary_buffer` was found.
+  thrust::free(exec, p);
+} // end return_temporary_buffer()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.h
new file mode 100644
index 000000000000..30e032696bc2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.h
@@ -0,0 +1,105 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction>
+__host__ __device__
+  OutputIterator transform(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator result,
+                           UnaryFunction op);
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator transform(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputIterator result,
+                           BinaryFunction op);
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(thrust::execution_policy<DerivedPolicy> &exec,
+                               InputIterator first,
+                               InputIterator last,
+                               ForwardIterator result,
+                               UnaryFunction unary_op,
+                               Predicate pred);
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(thrust::execution_policy<DerivedPolicy> &exec,
+                               InputIterator1 first,
+                               InputIterator1 last,
+                               InputIterator2 stencil,
+                               ForwardIterator result,
+                               UnaryFunction unary_op,
+                               Predicate pred);
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename ForwardIterator,
+         typename BinaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(thrust::execution_policy<DerivedPolicy> &exec,
+                               InputIterator1 first1,
+                               InputIterator1 last1,
+                               InputIterator2 first2,
+                               InputIterator3 stencil,
+                               ForwardIterator result,
+                               BinaryFunction binary_op,
+                               Predicate pred);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/transform.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.inl
new file mode 100644
index 000000000000..122c42580b0b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform.inl
@@ -0,0 +1,191 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/transform.h>
+#include <thrust/for_each.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/tuple.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/internal_functional.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction>
+__host__ __device__
+  OutputIterator transform(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator result,
+                           UnaryFunction op)
+{
+  typedef thrust::detail::unary_transform_functor<UnaryFunction> UnaryTransformFunctor;
+
+  // make an iterator tuple
+  typedef thrust::tuple<InputIterator,OutputIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator zipped_result =
+    thrust::for_each(exec,
+                     thrust::make_zip_iterator(thrust::make_tuple(first,result)),
+                     thrust::make_zip_iterator(thrust::make_tuple(last,result)),
+                     UnaryTransformFunctor(op));
+
+  return thrust::get<1>(zipped_result.get_iterator_tuple());
+} // end transform()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator transform(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputIterator result,
+                           BinaryFunction op)
+{
+  // given the minimal system, determine the binary transform functor we need
+  typedef thrust::detail::binary_transform_functor<BinaryFunction> BinaryTransformFunctor;
+
+  // make an iterator tuple
+  typedef thrust::tuple<InputIterator1,InputIterator2,OutputIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator zipped_result =
+    thrust::for_each(exec,
+                     thrust::make_zip_iterator(thrust::make_tuple(first1,first2,result)),
+                     thrust::make_zip_iterator(thrust::make_tuple(last1,first2,result)),
+                     BinaryTransformFunctor(op));
+
+  return thrust::get<2>(zipped_result.get_iterator_tuple());
+} // end transform()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(thrust::execution_policy<DerivedPolicy> &exec,
+                               InputIterator first,
+                               InputIterator last,
+                               ForwardIterator result,
+                               UnaryFunction unary_op,
+                               Predicate pred)
+{
+  typedef thrust::detail::unary_transform_if_functor<UnaryFunction,Predicate> UnaryTransformIfFunctor;
+
+  // make an iterator tuple
+  typedef thrust::tuple<InputIterator,ForwardIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator zipped_result =
+    thrust::for_each(exec,
+                     thrust::make_zip_iterator(thrust::make_tuple(first,result)),
+                     thrust::make_zip_iterator(thrust::make_tuple(last,result)),
+                     UnaryTransformIfFunctor(unary_op,pred));
+
+  return thrust::get<1>(zipped_result.get_iterator_tuple());
+} // end transform_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(thrust::execution_policy<DerivedPolicy> &exec,
+                               InputIterator1 first,
+                               InputIterator1 last,
+                               InputIterator2 stencil,
+                               ForwardIterator result,
+                               UnaryFunction unary_op,
+                               Predicate pred)
+{
+  typedef thrust::detail::unary_transform_if_with_stencil_functor<UnaryFunction,Predicate> UnaryTransformIfFunctor;
+
+  // make an iterator tuple
+  typedef thrust::tuple<InputIterator1,InputIterator2,ForwardIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator zipped_result =
+    thrust::for_each(exec,
+                     thrust::make_zip_iterator(thrust::make_tuple(first,stencil,result)),
+                     thrust::make_zip_iterator(thrust::make_tuple(last,stencil,result)),
+                     UnaryTransformIfFunctor(unary_op,pred));
+
+  return thrust::get<2>(zipped_result.get_iterator_tuple());
+} // end transform_if()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename ForwardIterator,
+         typename BinaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(thrust::execution_policy<DerivedPolicy> &exec,
+                               InputIterator1 first1,
+                               InputIterator1 last1,
+                               InputIterator2 first2,
+                               InputIterator3 stencil,
+                               ForwardIterator result,
+                               BinaryFunction binary_op,
+                               Predicate pred)
+{
+  typedef thrust::detail::binary_transform_if_functor<BinaryFunction,Predicate> BinaryTransformIfFunctor;
+
+  // make an iterator tuple
+  typedef thrust::tuple<InputIterator1,InputIterator2,InputIterator3,ForwardIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator zipped_result =
+    thrust::for_each(exec,
+                     thrust::make_zip_iterator(thrust::make_tuple(first1,first2,stencil,result)),
+                     thrust::make_zip_iterator(thrust::make_tuple(last1,first2,stencil,result)),
+                     BinaryTransformIfFunctor(binary_op,pred));
+
+  return thrust::get<3>(zipped_result.get_iterator_tuple());
+} // end transform_if()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.h
new file mode 100644
index 000000000000..af510296e1ce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.h
@@ -0,0 +1,52 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator, 
+         typename UnaryFunction, 
+         typename OutputType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputType transform_reduce(thrust::execution_policy<ExecutionPolicy> &exec,
+                              InputIterator first,
+                              InputIterator last,
+                              UnaryFunction unary_op,
+                              OutputType init,
+                              BinaryFunction binary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/transform_reduce.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.inl
new file mode 100644
index 000000000000..539c3b22c047
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_reduce.inl
@@ -0,0 +1,57 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/transform_reduce.h>
+#include <thrust/reduce.h>
+#include <thrust/iterator/transform_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename UnaryFunction,
+         typename OutputType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputType transform_reduce(thrust::execution_policy<DerivedPolicy> &exec,
+                              InputIterator first,
+                              InputIterator last,
+                              UnaryFunction unary_op,
+                              OutputType init,
+                              BinaryFunction binary_op)
+{
+  thrust::transform_iterator<UnaryFunction, InputIterator, OutputType> xfrm_first(first, unary_op);
+  thrust::transform_iterator<UnaryFunction, InputIterator, OutputType> xfrm_last(last, unary_op);
+
+  return thrust::reduce(exec, xfrm_first, xfrm_last, init, binary_op);
+} // end transform_reduce()
+
+
+} // end generic
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.h
new file mode 100644
index 000000000000..05054c965a01
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.h
@@ -0,0 +1,67 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator transform_inclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          BinaryFunction binary_op);
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename T,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator transform_exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          T init,
+                                          AssociativeOperator binary_op);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/transform_scan.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.inl
new file mode 100644
index 000000000000..505bdbfab122
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/transform_scan.inl
@@ -0,0 +1,91 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/transform_scan.h>
+#include <thrust/scan.h>
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/function_traits.h>
+#include <thrust/detail/type_traits/iterator/is_output_iterator.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator transform_inclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          BinaryFunction binary_op)
+{
+  // Use the input iterator's value type per https://wg21.link/P0571
+  using InputType = typename thrust::iterator_value<InputIterator>::type;
+  using ResultType = thrust::detail::invoke_result_t<UnaryFunction, InputType>;
+  using ValueType = typename std::remove_reference<ResultType>::type;
+
+  thrust::transform_iterator<UnaryFunction, InputIterator, ValueType> _first(first, unary_op);
+  thrust::transform_iterator<UnaryFunction, InputIterator, ValueType> _last(last, unary_op);
+
+  return thrust::inclusive_scan(exec, _first, _last, result, binary_op);
+} // end transform_inclusive_scan()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename InitialValueType,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator transform_exclusive_scan(thrust::execution_policy<ExecutionPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          InitialValueType init,
+                                          AssociativeOperator binary_op)
+{
+  // Use the initial value type per https://wg21.link/P0571
+  using ValueType = typename std::remove_reference<InitialValueType>::type;
+
+  thrust::transform_iterator<UnaryFunction, InputIterator, ValueType> _first(first, unary_op);
+  thrust::transform_iterator<UnaryFunction, InputIterator, ValueType> _last(last, unary_op);
+
+  return thrust::exclusive_scan(exec, _first, _last, result, init, binary_op);
+} // end transform_exclusive_scan()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.h
new file mode 100644
index 000000000000..bac5bcf96092
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.h
@@ -0,0 +1,56 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                                     InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result);
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename Size,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy_n(thrust::execution_policy<ExecutionPolicy> &exec,
+                                       InputIterator first,
+                                       Size n,
+                                       ForwardIterator result);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/uninitialized_copy.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.inl
new file mode 100644
index 000000000000..679d1f6bae8c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_copy.inl
@@ -0,0 +1,194 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/uninitialized_copy.h>
+#include <thrust/copy.h>
+#include <thrust/for_each.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+template<typename InputType,
+         typename OutputType>
+  struct uninitialized_copy_functor
+{
+  template<typename Tuple>
+  __host__ __device__
+  void operator()(Tuple t)
+  {
+    const InputType &in = thrust::get<0>(t);
+    OutputType &out = thrust::get<1>(t);
+
+    ::new(static_cast<void*>(&out)) OutputType(in);
+  } // end operator()()
+}; // end uninitialized_copy_functor
+
+
+// non-trivial copy constructor path
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                                     InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result,
+                                     thrust::detail::false_type) // has_trivial_copy_constructor
+{
+  // zip up the iterators
+  typedef thrust::tuple<InputIterator,ForwardIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator begin = thrust::make_zip_iterator(thrust::make_tuple(first,result));
+  ZipIterator end = begin;
+
+  // get a zip_iterator pointing to the end
+  const typename thrust::iterator_difference<InputIterator>::type n = thrust::distance(first,last);
+  thrust::advance(end, n);
+
+  // create a functor
+  typedef typename iterator_traits<InputIterator>::value_type InputType;
+  typedef typename iterator_traits<ForwardIterator>::value_type OutputType;
+
+  detail::uninitialized_copy_functor<InputType, OutputType> f;
+
+  // do the for_each
+  thrust::for_each(exec, begin, end, f);
+
+  // return the end of the output range
+  return thrust::get<1>(end.get_iterator_tuple());
+} // end uninitialized_copy()
+
+
+// trivial copy constructor path
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                                     InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result,
+                                     thrust::detail::true_type) // has_trivial_copy_constructor
+{
+  return thrust::copy(exec, first, last, result);
+} // end uninitialized_copy()
+
+
+// non-trivial copy constructor path
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename Size,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy_n(thrust::execution_policy<ExecutionPolicy> &exec,
+                                       InputIterator first,
+                                       Size n,
+                                       ForwardIterator result,
+                                       thrust::detail::false_type) // has_trivial_copy_constructor
+{
+  // zip up the iterators
+  typedef thrust::tuple<InputIterator,ForwardIterator> IteratorTuple;
+  typedef thrust::zip_iterator<IteratorTuple> ZipIterator;
+
+  ZipIterator zipped_first = thrust::make_zip_iterator(thrust::make_tuple(first,result));
+
+  // create a functor
+  typedef typename iterator_traits<InputIterator>::value_type   InputType;
+  typedef typename iterator_traits<ForwardIterator>::value_type OutputType;
+
+  detail::uninitialized_copy_functor<InputType, OutputType> f;
+
+  // do the for_each_n
+  ZipIterator zipped_last = thrust::for_each_n(exec, zipped_first, n, f);
+
+  // return the end of the output range
+  return thrust::get<1>(zipped_last.get_iterator_tuple());
+} // end uninitialized_copy_n()
+
+
+// trivial copy constructor path
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename Size,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy_n(thrust::execution_policy<ExecutionPolicy> &exec,
+                                       InputIterator first,
+                                       Size n,
+                                       ForwardIterator result,
+                                       thrust::detail::true_type) // has_trivial_copy_constructor
+{
+  return thrust::copy_n(exec, first, n, result);
+} // end uninitialized_copy_n()
+
+
+} // end detail
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                                     InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result)
+{
+  typedef typename iterator_traits<ForwardIterator>::value_type ResultType;
+
+  typedef typename thrust::detail::has_trivial_copy_constructor<ResultType>::type ResultTypeHasTrivialCopyConstructor;
+
+  return thrust::system::detail::generic::detail::uninitialized_copy(exec, first, last, result, ResultTypeHasTrivialCopyConstructor());
+} // end uninitialized_copy()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator,
+         typename Size,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy_n(thrust::execution_policy<ExecutionPolicy> &exec,
+                                       InputIterator first,
+                                       Size n,
+                                       ForwardIterator result)
+{
+  typedef typename iterator_traits<ForwardIterator>::value_type ResultType;
+
+  typedef typename thrust::detail::has_trivial_copy_constructor<ResultType>::type ResultTypeHasTrivialCopyConstructor;
+
+  return thrust::system::detail::generic::detail::uninitialized_copy_n(exec, first, n, result, ResultTypeHasTrivialCopyConstructor());
+} // end uninitialized_copy_n()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.h
new file mode 100644
index 000000000000..4f5404508c0a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.h
@@ -0,0 +1,56 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  void uninitialized_fill(thrust::execution_policy<DerivedPolicy> &exec,
+                          ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x);
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Size,
+         typename T>
+__host__ __device__
+  ForwardIterator uninitialized_fill_n(thrust::execution_policy<DerivedPolicy> &exec,
+                                       ForwardIterator first,
+                                       Size n,
+                                       const T &x);
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/uninitialized_fill.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.inl
new file mode 100644
index 000000000000..06241494548c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/uninitialized_fill.inl
@@ -0,0 +1,135 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/uninitialized_fill.h>
+#include <thrust/fill.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+namespace detail
+{
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  void uninitialized_fill(thrust::execution_policy<DerivedPolicy> &exec,
+                          ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x,
+                          thrust::detail::true_type) // has_trivial_copy_constructor
+{
+  thrust::fill(exec, first, last, x);
+} // end uninitialized_fill()
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  void uninitialized_fill(thrust::execution_policy<DerivedPolicy> &exec,
+                          ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x,
+                          thrust::detail::false_type) // has_trivial_copy_constructor
+{
+  typedef typename iterator_traits<ForwardIterator>::value_type ValueType;
+
+  thrust::for_each(exec, first, last, thrust::detail::uninitialized_fill_functor<ValueType>(x));
+} // end uninitialized_fill()
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Size,
+         typename T>
+__host__ __device__
+  ForwardIterator uninitialized_fill_n(thrust::execution_policy<DerivedPolicy> &exec,
+                                       ForwardIterator first,
+                                       Size n,
+                                       const T &x,
+                                       thrust::detail::true_type) // has_trivial_copy_constructor
+{
+  return thrust::fill_n(exec, first, n, x);
+} // end uninitialized_fill()
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Size,
+         typename T>
+__host__ __device__
+  ForwardIterator uninitialized_fill_n(thrust::execution_policy<DerivedPolicy> &exec,
+                                       ForwardIterator first,
+                                       Size n,
+                                       const T &x,
+                                       thrust::detail::false_type) // has_trivial_copy_constructor
+{
+  typedef typename iterator_traits<ForwardIterator>::value_type ValueType;
+
+  return thrust::for_each_n(exec, first, n, thrust::detail::uninitialized_fill_functor<ValueType>(x));
+} // end uninitialized_fill()
+
+} // end detail
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T>
+__host__ __device__
+  void uninitialized_fill(thrust::execution_policy<DerivedPolicy> &exec,
+                          ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x)
+{
+  typedef typename iterator_traits<ForwardIterator>::value_type ValueType;
+
+  typedef thrust::detail::has_trivial_copy_constructor<ValueType> ValueTypeHasTrivialCopyConstructor;
+
+  thrust::system::detail::generic::detail::uninitialized_fill(exec, first, last, x,
+    ValueTypeHasTrivialCopyConstructor());
+} // end uninitialized_fill()
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Size,
+         typename T>
+__host__ __device__
+  ForwardIterator uninitialized_fill_n(thrust::execution_policy<DerivedPolicy> &exec,
+                                       ForwardIterator first,
+                                       Size n,
+                                       const T &x)
+{
+  typedef typename iterator_traits<ForwardIterator>::value_type ValueType;
+
+  typedef thrust::detail::has_trivial_copy_constructor<ValueType> ValueTypeHasTrivialCopyConstructor;
+
+  return thrust::system::detail::generic::detail::uninitialized_fill_n(exec, first, n, x,
+    ValueTypeHasTrivialCopyConstructor());
+} // end uninitialized_fill()
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.h
new file mode 100644
index 000000000000..ce3bff884b3d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.h
@@ -0,0 +1,97 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+ForwardIterator unique(thrust::execution_policy<DerivedPolicy> &exec,
+                       ForwardIterator first,
+                       ForwardIterator last);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+ForwardIterator unique(thrust::execution_policy<DerivedPolicy> &exec,
+                       ForwardIterator first,
+                       ForwardIterator last,
+                       BinaryPredicate binary_pred);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator unique_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator output);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+OutputIterator unique_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator output,
+                           BinaryPredicate binary_pred);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last);
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/unique.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.inl
new file mode 100644
index 000000000000..bb66e3585d28
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique.inl
@@ -0,0 +1,139 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/unique.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/transform.h>
+#include <thrust/unique.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/copy_if.h>
+#include <thrust/detail/count.h>
+#include <thrust/distance.h>
+#include <thrust/functional.h>
+#include <thrust/detail/range/head_flags.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+  ForwardIterator unique(thrust::execution_policy<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type InputType;
+
+  return thrust::unique(exec, first, last, thrust::equal_to<InputType>());
+} // end unique()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  ForwardIterator unique(thrust::execution_policy<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last,
+                         BinaryPredicate binary_pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type InputType;
+
+  thrust::detail::temporary_array<InputType,DerivedPolicy> input(exec, first, last);
+
+  return thrust::unique_copy(exec, input.begin(), input.end(), first, binary_pred);
+} // end unique()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator unique_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator output)
+{
+  typedef typename thrust::iterator_value<InputIterator>::type value_type;
+  return thrust::unique_copy(exec, first,last,output,thrust::equal_to<value_type>());
+} // end unique_copy()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator unique_copy(thrust::execution_policy<DerivedPolicy> &exec,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator output,
+                             BinaryPredicate binary_pred)
+{
+  thrust::detail::head_flags<InputIterator, BinaryPredicate> stencil(first, last, binary_pred);
+
+  using namespace thrust::placeholders;
+
+  return thrust::copy_if(exec, first, last, stencil.begin(), output, _1);
+} // end unique_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred)
+{
+  thrust::detail::head_flags<ForwardIterator, BinaryPredicate> stencil(first, last, binary_pred);
+  
+  using namespace thrust::placeholders;
+  
+  return thrust::count_if(exec, stencil.begin(), stencil.end(), _1);
+} // end unique_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(thrust::execution_policy<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last)
+{
+  typedef typename thrust::iterator_value<ForwardIterator>::type value_type;
+  return thrust::unique_count(exec, first, last, thrust::equal_to<value_type>());
+} // end unique_copy()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.h
new file mode 100644
index 000000000000..0ea9e7cc8452
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.h
@@ -0,0 +1,94 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/tag.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+    unique_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_last,
+                  ForwardIterator2 values_first);
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+    unique_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_last,
+                  ForwardIterator2 values_first,
+                  BinaryPredicate binary_pred);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                       InputIterator1 keys_first, 
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_output,
+                       OutputIterator2 values_output);
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                       InputIterator1 keys_first, 
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_output,
+                       OutputIterator2 values_output,
+                       BinaryPredicate binary_pred);
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/generic/unique_by_key.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.inl
new file mode 100644
index 000000000000..ffcf1dd0c741
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/generic/unique_by_key.inl
@@ -0,0 +1,139 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/generic/unique_by_key.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/detail/minimum_system.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/transform.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/internal_functional.h>
+#include <thrust/detail/copy_if.h>
+#include <thrust/unique.h>
+#include <thrust/detail/range/head_flags.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace generic
+{
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+thrust::pair<ForwardIterator1,ForwardIterator2>
+unique_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+              ForwardIterator1 keys_first,
+              ForwardIterator1 keys_last,
+              ForwardIterator2 values_first)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator1>::value_type KeyType;
+  return thrust::unique_by_key(exec, keys_first, keys_last, values_first, thrust::equal_to<KeyType>());
+} // end unique_by_key()
+
+
+template<typename ExecutionPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+thrust::pair<ForwardIterator1,ForwardIterator2>
+unique_by_key(thrust::execution_policy<ExecutionPolicy> &exec,
+              ForwardIterator1 keys_first,
+              ForwardIterator1 keys_last,
+              ForwardIterator2 values_first,
+              BinaryPredicate binary_pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator1>::value_type InputType1;
+  typedef typename thrust::iterator_traits<ForwardIterator2>::value_type InputType2;
+
+  ForwardIterator2 values_last = values_first + (keys_last - keys_first);
+
+  thrust::detail::temporary_array<InputType1,ExecutionPolicy> keys(exec, keys_first, keys_last);
+  thrust::detail::temporary_array<InputType2,ExecutionPolicy> vals(exec, values_first, values_last);
+
+  return thrust::unique_by_key_copy(exec, keys.begin(), keys.end(), vals.begin(), keys_first, values_first, binary_pred);
+} // end unique_by_key()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+unique_by_key_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                   InputIterator1 keys_first,
+                   InputIterator1 keys_last,
+                   InputIterator2 values_first,
+                   OutputIterator1 keys_output,
+                   OutputIterator2 values_output)
+{
+  typedef typename thrust::iterator_traits<InputIterator1>::value_type KeyType;
+  return thrust::unique_by_key_copy(exec, keys_first, keys_last, values_first, keys_output, values_output, thrust::equal_to<KeyType>());
+} // end unique_by_key_copy()
+
+
+template<typename ExecutionPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+unique_by_key_copy(thrust::execution_policy<ExecutionPolicy> &exec,
+                   InputIterator1 keys_first,
+                   InputIterator1 keys_last,
+                   InputIterator2 values_first,
+                   OutputIterator1 keys_output,
+                   OutputIterator2 values_output,
+                   BinaryPredicate binary_pred)
+{
+  typedef typename thrust::iterator_traits<InputIterator1>::difference_type difference_type;
+
+  difference_type n = thrust::distance(keys_first, keys_last);
+
+  thrust::detail::head_flags<InputIterator1, BinaryPredicate> stencil(keys_first, keys_last, binary_pred);
+
+  using namespace thrust::placeholders;
+  thrust::zip_iterator< thrust::tuple<OutputIterator1, OutputIterator2> > result =
+    thrust::copy_if(exec,
+                    thrust::make_zip_iterator(thrust::make_tuple(keys_first, values_first)),
+                    thrust::make_zip_iterator(thrust::make_tuple(keys_first, values_first)) + n,
+                    stencil.begin(),
+                    thrust::make_zip_iterator(thrust::make_tuple(keys_output, values_output)),
+                    _1);
+
+  difference_type output_size = result - thrust::make_zip_iterator(thrust::make_tuple(keys_output, values_output));
+
+  return thrust::make_pair(keys_output + output_size, values_output + output_size);
+} // end unique_by_key_copy()
+
+
+} // end namespace generic
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/internal/decompose.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/internal/decompose.h
new file mode 100644
index 000000000000..58af7c55144c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/internal/decompose.h
@@ -0,0 +1,113 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace internal
+{
+
+  template <typename IndexType>
+    class index_range
+    {
+      public:
+        typedef IndexType index_type;
+
+        __host__ __device__
+          index_range(index_type begin, index_type end) : m_begin(begin), m_end(end) {}
+
+        __host__ __device__
+          index_type begin(void) const { return m_begin; }
+
+        __host__ __device__
+          index_type end(void)   const { return m_end; }
+
+        __host__ __device__
+          index_type size(void)  const { return m_end - m_begin; }
+
+      private:
+        index_type m_begin;
+        index_type m_end;
+    };
+
+  template <typename IndexType>
+    class uniform_decomposition
+    {
+      public:
+        typedef IndexType               index_type;
+        typedef index_range<index_type> range_type;
+
+        __host__ __device__
+        uniform_decomposition(index_type N, index_type granularity, index_type max_intervals)
+          : m_N(N),
+	    m_intervals((N + granularity - 1) / granularity),
+	    m_threshold(0),
+	    m_small_interval(granularity),
+	    m_large_interval(0)
+        {
+	  if(m_intervals > max_intervals)
+          {
+	    m_small_interval = granularity * (m_intervals / max_intervals);
+	    m_large_interval = m_small_interval + granularity;
+	    m_threshold      = m_intervals % max_intervals;
+	    m_intervals      = max_intervals;
+	  }
+        }
+
+        __host__ __device__
+          index_range<index_type> operator[](const index_type& i) const
+          {
+            if (i < m_threshold)
+            {
+              index_type begin = m_large_interval * i;
+              index_type end   = begin + m_large_interval;
+              return range_type(begin, end);
+            }
+            else
+            {
+              index_type begin = m_large_interval * m_threshold + m_small_interval * (i - m_threshold);
+              index_type end   = (begin + m_small_interval < m_N) ? begin + m_small_interval : m_N;
+              return range_type(begin, end);
+            }
+          }
+
+        __host__ __device__
+          index_type size(void) const
+          {
+            return m_intervals;
+          }
+
+      private:
+
+        index_type m_N;
+        index_type m_intervals;
+        index_type m_threshold;
+        index_type m_small_interval;
+        index_type m_large_interval;
+    };
+
+
+} // end namespace internal
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/adjacent_difference.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/adjacent_difference.h
new file mode 100644
index 000000000000..4a9dad82c93d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/adjacent_difference.h
@@ -0,0 +1,73 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file adjacent_difference.h
+ *  \brief Sequential implementation of adjacent_difference.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+OutputIterator adjacent_difference(sequential::execution_policy<DerivedPolicy> &,
+                                   InputIterator first,
+                                   InputIterator last,
+                                   OutputIterator result,
+                                   BinaryFunction binary_op)
+{
+  typedef typename thrust::iterator_traits<InputIterator>::value_type InputType;
+
+  if(first == last)
+    return result;
+
+  InputType curr = *first;
+
+  *result = curr;
+
+  while(++first != last)
+  {
+    InputType next = *first;
+    *(++result) = binary_op(next, curr);
+    curr = next;
+  }
+
+  return ++result;
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/assign_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/assign_value.h
new file mode 100644
index 000000000000..0eb145d135a2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/assign_value.h
@@ -0,0 +1,42 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/detail/raw_pointer_cast.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+template<typename DerivedPolicy, typename Pointer1, typename Pointer2>
+__host__ __device__
+  void assign_value(sequential::execution_policy<DerivedPolicy> &, Pointer1 dst, Pointer2 src)
+{
+  *thrust::raw_pointer_cast(dst) = *thrust::raw_pointer_cast(src);
+} // end assign_value()
+
+} // end sequential
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/binary_search.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/binary_search.h
new file mode 100644
index 000000000000..2da5080f4f85
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/binary_search.h
@@ -0,0 +1,158 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file binary_search.h
+ *  \brief Sequential implementation of binary search algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/advance.h>
+#include <thrust/distance.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/function.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T,
+         typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator lower_bound(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const T& val,
+                            StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  typedef typename thrust::iterator_difference<ForwardIterator>::type difference_type;
+
+  difference_type len = thrust::distance(first, last);
+
+  while(len > 0)
+  {
+    difference_type half = len >> 1;
+    ForwardIterator middle = first;
+
+    thrust::advance(middle, half);
+
+    if(wrapped_comp(*middle, val))
+    {
+      first = middle;
+      ++first;
+      len = len - half - 1;
+    }
+    else
+    {
+      len = half;
+    }
+  }
+
+  return first;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T,
+         typename StrictWeakOrdering>
+__host__ __device__
+ForwardIterator upper_bound(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            const T& val, 
+                            StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  typedef typename thrust::iterator_difference<ForwardIterator>::type difference_type;
+
+  difference_type len = thrust::distance(first, last);
+
+  while(len > 0)
+  {
+    difference_type half = len >> 1;
+    ForwardIterator middle = first;
+
+    thrust::advance(middle, half);
+
+    if(wrapped_comp(val, *middle))
+    {
+      len = half;
+    }
+    else
+    {
+      first = middle;
+      ++first;
+      len = len - half - 1;
+    }
+  }
+
+  return first;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename T,
+         typename StrictWeakOrdering>
+__host__ __device__
+bool binary_search(sequential::execution_policy<DerivedPolicy> &exec,
+                   ForwardIterator first,
+                   ForwardIterator last,
+                   const T& val, 
+                   StrictWeakOrdering comp)
+{
+  ForwardIterator iter = sequential::lower_bound(exec, first, last, val, comp);
+
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  return iter != last && !wrapped_comp(val,*iter);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.h
new file mode 100644
index 000000000000..0dd2cdad5c36
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.h
@@ -0,0 +1,62 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file copy.h
+ *  \brief Sequential implementations of copy algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(sequential::execution_policy<DerivedPolicy> &exec,
+                      InputIterator first,
+                      InputIterator last,
+                      OutputIterator result);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(sequential::execution_policy<DerivedPolicy> &exec,
+                        InputIterator first,
+                        Size n,
+                        OutputIterator result);
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/sequential/copy.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.inl
new file mode 100644
index 000000000000..850f20f1eaa5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy.inl
@@ -0,0 +1,146 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/copy.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/system/detail/sequential/general_copy.h>
+#include <thrust/system/detail/sequential/trivial_copy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+#include <thrust/type_traits/is_trivially_relocatable.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+namespace copy_detail
+{
+
+
+// returns the raw pointer associated with a Pointer-like thing
+template<typename Pointer>
+__host__ __device__
+  typename thrust::detail::pointer_traits<Pointer>::raw_pointer
+    get(Pointer ptr)
+{
+  return thrust::detail::pointer_traits<Pointer>::get(ptr);
+}
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(InputIterator first,
+                      InputIterator last,
+                      OutputIterator result,
+                      thrust::detail::true_type)  // is_indirectly_trivially_relocatable_to
+{
+  typedef typename thrust::iterator_difference<InputIterator>::type Size;
+
+  const Size n = last - first;
+  thrust::system::detail::sequential::trivial_copy_n(get(&*first), n, get(&*result));
+  return result + n;
+} // end copy()
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(InputIterator first,
+                      InputIterator last,
+                      OutputIterator result,
+                      thrust::detail::false_type)  // is_indirectly_trivially_relocatable_to
+{
+  return thrust::system::detail::sequential::general_copy(first,last,result);
+} // end copy()
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(InputIterator first,
+                        Size n,
+                        OutputIterator result,
+                        thrust::detail::true_type)  // is_indirectly_trivially_relocatable_to
+{
+  thrust::system::detail::sequential::trivial_copy_n(get(&*first), n, get(&*result));
+  return result + n;
+} // end copy_n()
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(InputIterator first,
+                        Size n,
+                        OutputIterator result,
+                        thrust::detail::false_type)  // is_indirectly_trivially_relocatable_to
+{
+  return thrust::system::detail::sequential::general_copy_n(first,n,result);
+} // end copy_n()
+
+
+} // end namespace copy_detail
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy(sequential::execution_policy<DerivedPolicy> &,
+                      InputIterator first,
+                      InputIterator last,
+                      OutputIterator result)
+{
+  return thrust::system::detail::sequential::copy_detail::copy(first, last, result,
+    typename thrust::is_indirectly_trivially_relocatable_to<InputIterator,OutputIterator>::type());
+} // end copy()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator copy_n(sequential::execution_policy<DerivedPolicy> &,
+                        InputIterator first,
+                        Size n,
+                        OutputIterator result)
+{
+  return thrust::system::detail::sequential::copy_detail::copy_n(first, n, result,
+    typename thrust::is_indirectly_trivially_relocatable_to<InputIterator,OutputIterator>::type());
+} // end copy_n()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_backward.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_backward.h
new file mode 100644
index 000000000000..d127ac80db24
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_backward.h
@@ -0,0 +1,53 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename BidirectionalIterator1,
+         typename BidirectionalIterator2>
+__host__ __device__
+BidirectionalIterator2 copy_backward(BidirectionalIterator1 first, 
+                                     BidirectionalIterator1 last, 
+                                     BidirectionalIterator2 result)
+{
+  while (first != last)
+  {
+    --last;
+    --result;
+    *result = *last;
+  }
+
+  return result;
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_if.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_if.h
new file mode 100644
index 000000000000..3c00956deb31
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/copy_if.h
@@ -0,0 +1,72 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file copy_if.h
+ *  \brief Sequential implementation of copy_if.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator copy_if(sequential::execution_policy<DerivedPolicy> &,
+                         InputIterator1 first,
+                         InputIterator1 last,
+                         InputIterator2 stencil,
+                         OutputIterator result,
+                         Predicate pred)
+{
+  thrust::detail::wrapped_function<Predicate,bool> wrapped_pred(pred);
+
+  while(first != last)
+  {
+    if(wrapped_pred(*stencil))
+    {
+      *result = *first;
+      ++result;
+    } // end if
+
+    ++first;
+    ++stencil;
+  } // end while
+
+  return result;
+} // end copy_if()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/count.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/count.h
new file mode 100644
index 000000000000..ed3cadffe364
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/count.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special count functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/equal.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/equal.h
new file mode 100644
index 000000000000..9d31e70f6fda
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/equal.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special equal functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/execution_policy.h
new file mode 100644
index 000000000000..99d78fc27f0b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/execution_policy.h
@@ -0,0 +1,75 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+// this awkward sequence of definitions arises
+// from the desire both for tag to derive
+// from execution_policy and for execution_policy
+// to convert to tag (when execution_policy is not
+// an ancestor of tag)
+
+// forward declaration of tag
+struct tag;
+
+// forward declaration of execution_policy
+template<typename> struct execution_policy;
+
+// specialize execution_policy for tag
+template<>
+  struct execution_policy<tag>
+    : thrust::execution_policy<tag>
+{};
+
+// tag's definition comes before the generic definition of execution_policy
+struct tag : execution_policy<tag>
+{
+  __host__ __device__ constexpr tag() {}
+};
+
+// allow conversion to tag when it is not a successor
+template<typename Derived>
+  struct execution_policy
+    : thrust::execution_policy<Derived>
+{
+  // allow conversion to tag
+  inline operator tag () const
+  {
+    return tag();
+  }
+};
+
+
+THRUST_INLINE_CONSTANT tag seq;
+
+
+} // end sequential
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/extrema.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/extrema.h
new file mode 100644
index 000000000000..5e5c62da6946
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/extrema.h
@@ -0,0 +1,138 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file extrema.h
+ *  \brief Sequential implementations of extrema functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/pair.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+ForwardIterator min_element(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first, 
+                            ForwardIterator last,
+                            BinaryPredicate comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    BinaryPredicate,
+    bool
+  > wrapped_comp(comp);
+
+  ForwardIterator imin = first;
+
+  for(; first != last; ++first)
+  {
+    if(wrapped_comp(*first, *imin))
+    {
+      imin = first;
+    }
+  }
+
+  return imin;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+ForwardIterator max_element(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first, 
+                            ForwardIterator last,
+                            BinaryPredicate comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    BinaryPredicate,
+    bool
+  > wrapped_comp(comp);
+
+  ForwardIterator imax = first;
+
+  for(; first != last; ++first)
+  {
+    if(wrapped_comp(*imax, *first))
+    {
+      imax = first;
+    }
+  }
+
+  return imax;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+thrust::pair<ForwardIterator,ForwardIterator> minmax_element(sequential::execution_policy<DerivedPolicy> &,
+                                                             ForwardIterator first, 
+                                                             ForwardIterator last,
+                                                             BinaryPredicate comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    BinaryPredicate,
+    bool
+  > wrapped_comp(comp);
+  
+  ForwardIterator imin = first;
+  ForwardIterator imax = first;
+
+  for(; first != last; ++first)
+  {
+    if(wrapped_comp(*first, *imin))
+    {
+      imin = first;
+    }
+
+    if(wrapped_comp(*imax, *first))
+    {
+      imax = first;
+    }
+  }
+
+  return thrust::make_pair(imin, imax);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/fill.h
new file mode 100644
index 000000000000..20c636096c25
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/fill.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special fill functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/find.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/find.h
new file mode 100644
index 000000000000..54c238c71582
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/find.h
@@ -0,0 +1,70 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file find.h
+ *  \brief Sequential implementation of find_if. 
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+InputIterator find_if(execution_policy<DerivedPolicy> &,
+                      InputIterator first,
+                      InputIterator last,
+                      Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  while(first != last)
+  {
+    if (wrapped_pred(*first))
+      return first;
+
+    ++first;
+  }
+
+  // return first so zip_iterator works correctly
+  return first;
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/for_each.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/for_each.h
new file mode 100644
index 000000000000..7058c56f28ed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/for_each.h
@@ -0,0 +1,94 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file for_each.h
+ *  \brief Sequential implementations of for_each functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename UnaryFunction>
+__host__ __device__
+InputIterator for_each(sequential::execution_policy<DerivedPolicy> &,
+                       InputIterator first,
+                       InputIterator last,
+                       UnaryFunction f)
+{
+  // wrap f
+  thrust::detail::wrapped_function<
+    UnaryFunction,
+    void
+  > wrapped_f(f);
+
+  for(; first != last; ++first)
+  {
+    wrapped_f(*first);
+  }
+
+  return first;
+} // end for_each()
+
+
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename Size,
+         typename UnaryFunction>
+__host__ __device__
+InputIterator for_each_n(sequential::execution_policy<DerivedPolicy> &,
+                         InputIterator first,
+                         Size n,
+                         UnaryFunction f)
+{
+  // wrap f
+  thrust::detail::wrapped_function<
+    UnaryFunction,
+    void
+  > wrapped_f(f);
+
+  for(Size i = 0; i != n; i++)
+  {
+    // we can dereference an OutputIterator if f does not
+    // try to use the reference for anything besides assignment
+    wrapped_f(*first);
+    ++first;
+  }
+
+  return first;
+} // end for_each_n()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/gather.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/gather.h
new file mode 100644
index 000000000000..a5657585eae6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/gather.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special gather functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/general_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/general_copy.h
new file mode 100644
index 000000000000..6ea87bbacbed
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/general_copy.h
@@ -0,0 +1,146 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file general_copy.h
+ *  \brief Sequential copy algorithms for general iterators.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/raw_reference_cast.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+namespace general_copy_detail
+{
+
+
+template<typename T1, typename T2>
+struct lazy_is_assignable
+  : thrust::detail::is_assignable<
+      typename T1::type,
+      typename T2::type
+    >
+{};
+
+
+// sometimes OutputIterator's reference type is reported as void
+// in that case, just assume that we're able to assign to it OK
+template<typename InputIterator, typename OutputIterator>
+struct reference_is_assignable
+  : thrust::detail::eval_if<
+      thrust::detail::is_same<
+        typename thrust::iterator_reference<OutputIterator>::type, void
+      >::value,
+      thrust::detail::true_type,
+      lazy_is_assignable<
+        thrust::iterator_reference<OutputIterator>,
+        thrust::iterator_reference<InputIterator>
+      >
+    >::type
+{};
+
+
+// introduce an iterator assign helper to deal with assignments from
+// a wrapped reference
+
+__thrust_exec_check_disable__
+template<typename OutputIterator, typename InputIterator>
+inline __host__ __device__
+typename thrust::detail::enable_if<
+  reference_is_assignable<InputIterator,OutputIterator>::value
+>::type
+iter_assign(OutputIterator dst, InputIterator src)
+{
+  *dst = *src;
+}
+
+
+__thrust_exec_check_disable__
+template<typename OutputIterator, typename InputIterator>
+inline __host__ __device__
+typename thrust::detail::disable_if<
+  reference_is_assignable<InputIterator,OutputIterator>::value
+>::type
+iter_assign(OutputIterator dst, InputIterator src)
+{
+  typedef typename thrust::iterator_value<InputIterator>::type value_type;
+
+  // insert a temporary and hope for the best
+  *dst = static_cast<value_type>(*src);
+}
+
+
+} // end general_copy_detail
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator general_copy(InputIterator first,
+                              InputIterator last,
+                              OutputIterator result)
+{
+  for(; first != last; ++first, ++result)
+  {
+    // gcc 4.2 crashes while instantiating iter_assign
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC) && (THRUST_GCC_VERSION < 40300)
+    *result = *first;
+#else
+    general_copy_detail::iter_assign(result, first);
+#endif
+  }
+
+  return result;
+} // end general_copy()
+
+
+__thrust_exec_check_disable__
+template<typename InputIterator,
+         typename Size,
+         typename OutputIterator>
+__host__ __device__
+  OutputIterator general_copy_n(InputIterator first,
+                                Size n,
+                                OutputIterator result)
+{
+  for(; n > Size(0); ++first, ++result, --n)
+  {
+    // gcc 4.2 crashes while instantiating iter_assign
+#if (THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC) && (THRUST_GCC_VERSION < 40300)
+    *result = *first;
+#else
+    general_copy_detail::iter_assign(result, first);
+#endif
+  }
+
+  return result;
+} // end general_copy_n()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/generate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/generate.h
new file mode 100644
index 000000000000..ac38be51617d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/generate.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special generate functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/get_value.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/get_value.h
new file mode 100644
index 000000000000..90752d8671a3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/get_value.h
@@ -0,0 +1,45 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/detail/raw_pointer_cast.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy, typename Pointer>
+__host__ __device__
+  typename thrust::iterator_value<Pointer>::type
+    get_value(sequential::execution_policy<DerivedPolicy> &, Pointer ptr)
+{
+  return *thrust::raw_pointer_cast(ptr);
+} // end get_value()
+
+
+} // end sequential
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/inner_product.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/inner_product.h
new file mode 100644
index 000000000000..e6b3c0ae1747
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/inner_product.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special inner_product functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/insertion_sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/insertion_sort.h
new file mode 100644
index 000000000000..9acccd8e92ec
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/insertion_sort.h
@@ -0,0 +1,152 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/copy_backward.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void insertion_sort(RandomAccessIterator first,
+                    RandomAccessIterator last,
+                    StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type;
+
+  if(first == last) return;
+
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  for(RandomAccessIterator i = first + 1; i != last; ++i)
+  {
+    value_type tmp = *i;
+
+    if(wrapped_comp(tmp, *first))
+    {
+      // tmp is the smallest value encountered so far
+      sequential::copy_backward(first, i, i + 1);
+
+      *first = tmp;
+    }
+    else
+    {
+      // tmp is not the smallest value, can avoid checking for j == first
+      RandomAccessIterator j = i;
+      RandomAccessIterator k = i - 1;
+
+      while(wrapped_comp(tmp, *k))
+      {
+        *j = *k;
+        j = k;
+        --k;
+      }
+
+      *j = tmp;
+    }
+  }
+}
+
+
+__thrust_exec_check_disable__
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void insertion_sort_by_key(RandomAccessIterator1 first1,
+                           RandomAccessIterator1 last1,
+                           RandomAccessIterator2 first2,
+                           StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type value_type1;
+  typedef typename thrust::iterator_value<RandomAccessIterator2>::type value_type2;
+
+  if(first1 == last1) return;
+
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  RandomAccessIterator1 i1 = first1 + 1;
+  RandomAccessIterator2 i2 = first2 + 1;
+
+  for(; i1 != last1; ++i1, ++i2)
+  {
+    value_type1 tmp1 = *i1;
+    value_type2 tmp2 = *i2;
+
+    if(wrapped_comp(tmp1, *first1))
+    {
+      // tmp is the smallest value encountered so far
+      sequential::copy_backward(first1, i1, i1 + 1);
+      sequential::copy_backward(first2, i2, i2 + 1);
+
+      *first1 = tmp1;
+      *first2 = tmp2;
+    }
+    else
+    {
+      // tmp is not the smallest value, can avoid checking for j == first
+      RandomAccessIterator1 j1 = i1;
+      RandomAccessIterator1 k1 = i1 - 1;
+
+      RandomAccessIterator2 j2 = i2;
+      RandomAccessIterator2 k2 = i2 - 1;
+
+      while(wrapped_comp(tmp1, *k1))
+      {
+        *j1 = *k1;
+        *j2 = *k2;
+
+        j1 = k1;
+        j2 = k2;
+
+        --k1;
+        --k2;
+      }
+
+      *j1 = tmp1;
+      *j2 = tmp2;
+    }
+  }
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/iter_swap.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/iter_swap.h
new file mode 100644
index 000000000000..7a5c481fc03f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/iter_swap.h
@@ -0,0 +1,46 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/swap.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy, typename Pointer1, typename Pointer2>
+__host__ __device__
+  void iter_swap(sequential::execution_policy<DerivedPolicy> &, Pointer1 a, Pointer2 b)
+{
+  using thrust::swap;
+  swap(*thrust::raw_pointer_cast(a), *thrust::raw_pointer_cast(b));
+} // end iter_swap()
+
+
+} // end sequential
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/logical.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/logical.h
new file mode 100644
index 000000000000..ee4586273566
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/logical.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special logical functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/malloc_and_free.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/malloc_and_free.h
new file mode 100644
index 000000000000..b250140e083c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/malloc_and_free.h
@@ -0,0 +1,53 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <cstdlib> // for malloc & free
+#include <thrust/detail/raw_pointer_cast.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy>
+inline __host__ __device__
+void *malloc(execution_policy<DerivedPolicy> &, std::size_t n)
+{
+  return std::malloc(n);
+} // end mallc()
+
+
+template<typename DerivedPolicy, typename Pointer>
+inline __host__ __device__
+void free(sequential::execution_policy<DerivedPolicy> &, Pointer ptr)
+{
+  std::free(thrust::raw_pointer_cast(ptr));
+} // end mallc()
+
+
+} // end sequential
+} // end detail
+} // end system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.h
new file mode 100644
index 000000000000..a45e18004d4e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.h
@@ -0,0 +1,79 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file merge.h
+ *  \brief Sequential implementation of merge algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator merge(sequential::execution_policy<DerivedPolicy> &exec,
+                     InputIterator1 first1,
+                     InputIterator1 last1,
+                     InputIterator2 first2,
+                     InputIterator2 last2,
+                     OutputIterator result,
+                     StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  merge_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+               InputIterator1 keys_first1,
+               InputIterator1 keys_last1,
+               InputIterator2 keys_first2,
+               InputIterator2 keys_last2,
+               InputIterator3 values_first1,
+               InputIterator4 values_first2,
+               OutputIterator1 keys_result,
+               OutputIterator2 values_result,
+               StrictWeakOrdering comp);
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/sequential/merge.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.inl
new file mode 100644
index 000000000000..08d7c0b0d4e1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/merge.inl
@@ -0,0 +1,154 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/merge.h>
+#include <thrust/detail/copy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/function.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+OutputIterator merge(sequential::execution_policy<DerivedPolicy> &exec,
+                     InputIterator1 first1,
+                     InputIterator1 last1,
+                     InputIterator2 first2,
+                     InputIterator2 last2,
+                     OutputIterator result,
+                     StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  while(first1 != last1 && first2 != last2)
+  {
+    if(wrapped_comp(*first2, *first1))
+    {
+      *result = *first2;
+      ++first2;
+    } // end if
+    else
+    {
+      *result = *first1;
+      ++first1;
+    } // end else
+
+    ++result;
+  } // end while
+
+  return thrust::copy(exec, first2, last2, thrust::copy(exec, first1, last1, result));
+} // end merge()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename InputIterator4,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+thrust::pair<OutputIterator1,OutputIterator2>
+  merge_by_key(sequential::execution_policy<DerivedPolicy> &,
+               InputIterator1 keys_first1,
+               InputIterator1 keys_last1,
+               InputIterator2 keys_first2,
+               InputIterator2 keys_last2,
+               InputIterator3 values_first1,
+               InputIterator4 values_first2,
+               OutputIterator1 keys_result,
+               OutputIterator2 values_result,
+               StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  while(keys_first1 != keys_last1 && keys_first2 != keys_last2)
+  {
+    if(!wrapped_comp(*keys_first2, *keys_first1))
+    {
+      // *keys_first1 <= *keys_first2
+      *keys_result   = *keys_first1;
+      *values_result = *values_first1;
+      ++keys_first1;
+      ++values_first1;
+    }
+    else
+    {
+      // *keys_first1 > keys_first2
+      *keys_result   = *keys_first2;
+      *values_result = *values_first2;
+      ++keys_first2;
+      ++values_first2;
+    }
+
+    ++keys_result;
+    ++values_result;
+  }
+
+  while(keys_first1 != keys_last1)
+  {
+    *keys_result   = *keys_first1;
+    *values_result = *values_first1;
+    ++keys_first1;
+    ++values_first1;
+    ++keys_result;
+    ++values_result;
+  }
+
+  while(keys_first2 != keys_last2)
+  {
+    *keys_result   = *keys_first2;
+    *values_result = *values_first2;
+    ++keys_first2;
+    ++values_first2;
+    ++keys_result;
+    ++values_result;
+  }
+
+  return thrust::make_pair(keys_result, values_result);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/mismatch.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/mismatch.h
new file mode 100644
index 000000000000..e6094d261a0f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/mismatch.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special mismatch functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/partition.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/partition.h
new file mode 100644
index 000000000000..43d5b0e2378c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/partition.h
@@ -0,0 +1,340 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file partition.h
+ *  \brief Sequential implementations of partition functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/pair.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace detail
+{
+
+
+// XXX WAR an unfortunate circular #inclusion problem
+template<typename,typename> class temporary_array;
+
+
+} // end detail
+
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+void iter_swap(ForwardIterator1 iter1, ForwardIterator2 iter2)
+{
+  // XXX this isn't correct because it doesn't use thrust::swap
+  using namespace thrust::detail;
+
+  typedef typename thrust::iterator_value<ForwardIterator1>::type T;
+
+  T temp = *iter1;
+  *iter1 = *iter2;
+  *iter2 = temp;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  if(first == last)
+    return first;
+
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  while(wrapped_pred(*first))
+  {
+    if(++first == last)
+      return first;
+  }
+
+  ForwardIterator next = first;
+
+  while(++next != last)
+  {
+    if(wrapped_pred(*next))
+    {
+      // Fully qualify name to disambiguate overloads found via ADL.
+      THRUST_NS_QUALIFIER::system::detail::sequential::iter_swap(first, next);
+      ++first;
+    }
+  }
+
+  return first;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator partition(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil_first,
+                            Predicate pred)
+{
+  if(first == last)
+    return first;
+
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  while(wrapped_pred(*stencil_first))
+  {
+    ++stencil_first;
+    if(++first == last)
+    {
+      return first;
+    }
+  }
+
+  ForwardIterator next = first;
+
+  // advance stencil to next element as well
+  ++stencil_first;
+
+  while(++next != last)
+  {
+    if(wrapped_pred(*stencil_first))
+    {
+      // Fully qualify name to disambiguate overloads found via ADL.
+      THRUST_NS_QUALIFIER::system::detail::sequential::iter_swap(first, next);
+      ++first;
+    }
+
+    ++stencil_first;
+  }
+
+  return first;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(sequential::execution_policy<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  typedef typename thrust::iterator_value<ForwardIterator>::type T;
+
+  typedef thrust::detail::temporary_array<T,DerivedPolicy> TempRange;
+  typedef typename TempRange::iterator                     TempIterator;
+
+  TempRange temp(exec, first, last);
+
+  for(TempIterator iter = temp.begin(); iter != temp.end(); ++iter)
+  {
+    if(wrapped_pred(*iter))
+    {
+      *first = *iter;
+      ++first;
+    }
+  }
+
+  ForwardIterator middle = first;
+
+  for(TempIterator iter = temp.begin(); iter != temp.end(); ++iter)
+  {
+    if(!wrapped_pred(*iter))
+    {
+      *first = *iter;
+      ++first;
+    }
+  }
+
+  return middle;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator stable_partition(sequential::execution_policy<DerivedPolicy> &exec,
+                                   ForwardIterator first,
+                                   ForwardIterator last,
+                                   InputIterator stencil,
+                                   Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  typedef typename thrust::iterator_value<ForwardIterator>::type T;
+
+  typedef thrust::detail::temporary_array<T,DerivedPolicy> TempRange;
+  typedef typename TempRange::iterator                     TempIterator;
+
+  TempRange temp(exec, first, last);
+
+  InputIterator stencil_iter = stencil;
+  for(TempIterator iter = temp.begin(); iter != temp.end(); ++iter, ++stencil_iter)
+  {
+    if(wrapped_pred(*stencil_iter))
+    {
+      *first = *iter;
+      ++first;
+    }
+  }
+
+  ForwardIterator middle = first;
+  stencil_iter = stencil;
+
+  for(TempIterator iter = temp.begin(); iter != temp.end(); ++iter, ++stencil_iter)
+  {
+    if(!wrapped_pred(*stencil_iter))
+    {
+      *first = *iter;
+      ++first;
+    }
+  }
+
+  return middle;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(sequential::execution_policy<DerivedPolicy> &,
+                          InputIterator first,
+                          InputIterator last,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  for(; first != last; ++first)
+  {
+    if(wrapped_pred(*first))
+    {
+      *out_true = *first;
+      ++out_true;
+    } // end if
+    else
+    {
+      *out_false = *first;
+      ++out_false;
+    } // end else
+  }
+
+  return thrust::make_pair(out_true, out_false);
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename Predicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    stable_partition_copy(sequential::execution_policy<DerivedPolicy> &,
+                          InputIterator1 first,
+                          InputIterator1 last,
+                          InputIterator2 stencil,
+                          OutputIterator1 out_true,
+                          OutputIterator2 out_false,
+                          Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  for(; first != last; ++first, ++stencil)
+  {
+    if(wrapped_pred(*stencil))
+    {
+      *out_true = *first;
+      ++out_true;
+    } // end if
+    else
+    {
+      *out_false = *first;
+      ++out_false;
+    } // end else
+  }
+
+  return thrust::make_pair(out_true, out_false);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/per_device_resource.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/per_device_resource.h
new file mode 100644
index 000000000000..1b8d61f92169
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/per_device_resource.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special per device resource functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce.h
new file mode 100644
index 000000000000..a532f71b2466
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce.h
@@ -0,0 +1,72 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file reduce.h
+ *  \brief Sequential implementation of reduce algorithm.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator, 
+         typename OutputType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputType reduce(sequential::execution_policy<DerivedPolicy> &,
+                    InputIterator begin,
+                    InputIterator end,
+                    OutputType init,
+                    BinaryFunction binary_op)
+{
+  // wrap binary_op
+  thrust::detail::wrapped_function<
+    BinaryFunction,
+    OutputType
+  > wrapped_binary_op(binary_op);
+
+  // initialize the result
+  OutputType result = init;
+
+  while(begin != end)
+  {
+    result = wrapped_binary_op(result, *begin);
+    ++begin;
+  } // end while
+
+  return result;
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce_by_key.h
new file mode 100644
index 000000000000..ef17ac5b0eac
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reduce_by_key.h
@@ -0,0 +1,102 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/pair.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    reduce_by_key(sequential::execution_policy<DerivedPolicy> &,
+                  InputIterator1 keys_first, 
+                  InputIterator1 keys_last,
+                  InputIterator2 values_first,
+                  OutputIterator1 keys_output,
+                  OutputIterator2 values_output,
+                  BinaryPredicate binary_pred,
+                  BinaryFunction binary_op)
+{
+  typedef typename thrust::iterator_traits<InputIterator1>::value_type  InputKeyType;
+  typedef typename thrust::iterator_traits<InputIterator2>::value_type  InputValueType;
+
+  // Use the input iterator's value type per https://wg21.link/P0571
+  using TemporaryType = typename thrust::iterator_value<InputIterator2>::type;
+
+  if(keys_first != keys_last)
+  {
+    InputKeyType  temp_key   = *keys_first;
+    TemporaryType temp_value = *values_first;
+
+    for(++keys_first, ++values_first;
+        keys_first != keys_last;
+        ++keys_first, ++values_first)
+    {
+      InputKeyType    key  = *keys_first;
+      InputValueType value = *values_first;
+
+      if(binary_pred(temp_key, key))
+      {
+        temp_value = binary_op(temp_value, value);
+      }
+      else
+      {
+        *keys_output   = temp_key;
+        *values_output = temp_value;
+
+        ++keys_output;
+        ++values_output;
+
+        temp_key   = key;
+        temp_value = value;
+      }
+    }
+
+    *keys_output   = temp_key;
+    *values_output = temp_value;
+
+    ++keys_output;
+    ++values_output;
+  }
+
+  return thrust::make_pair(keys_output, values_output);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/remove.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/remove.h
new file mode 100644
index 000000000000..df564f15b03c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/remove.h
@@ -0,0 +1,201 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file remove.h
+ *  \brief Sequential implementations of remove functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  // advance iterators until wrapped_pred(*first) is true or we reach the end of input
+  while(first != last && !wrapped_pred(*first))
+    ++first;
+
+  if(first == last)
+    return first;
+
+  // result always trails first 
+  ForwardIterator result = first;
+
+  ++first;
+
+  while(first != last)
+  {
+    if(!wrapped_pred(*first))
+    {
+      *result = *first;
+      ++result;
+    }
+    ++first;
+  }
+
+  return result;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename InputIterator,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator remove_if(sequential::execution_policy<DerivedPolicy> &,
+                            ForwardIterator first,
+                            ForwardIterator last,
+                            InputIterator stencil,
+                            Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  // advance iterators until wrapped_pred(*stencil) is true or we reach the end of input
+  while(first != last && !wrapped_pred(*stencil))
+  {
+    ++first;
+    ++stencil;
+  }
+
+  if(first == last)
+    return first;
+
+  // result always trails first 
+  ForwardIterator result = first;
+
+  ++first;
+  ++stencil;
+
+  while(first != last)
+  {
+    if(!wrapped_pred(*stencil))
+    {
+      *result = *first;
+      ++result;
+    }
+    ++first;
+    ++stencil;
+  }
+
+  return result;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(sequential::execution_policy<DerivedPolicy> &,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  while (first != last)
+  {
+    if (!wrapped_pred(*first))
+    {
+      *result = *first;
+      ++result;
+    }
+
+    ++first;
+  }
+
+  return result;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename Predicate>
+__host__ __device__
+  OutputIterator remove_copy_if(sequential::execution_policy<DerivedPolicy> &,
+                                InputIterator1 first,
+                                InputIterator1 last,
+                                InputIterator2 stencil,
+                                OutputIterator result,
+                                Predicate pred)
+{
+  // wrap pred
+  thrust::detail::wrapped_function<
+    Predicate,
+    bool
+  > wrapped_pred(pred);
+
+  while (first != last)
+  {
+    if (!wrapped_pred(*stencil))
+    {
+      *result = *first;
+      ++result;
+    }
+
+    ++first;
+    ++stencil;
+  }
+
+  return result;
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/replace.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/replace.h
new file mode 100644
index 000000000000..f5c8e8385717
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/replace.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special replace functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reverse.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reverse.h
new file mode 100644
index 000000000000..f80974e8a8d7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/reverse.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special reverse functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan.h
new file mode 100644
index 000000000000..c5fce2475821
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan.h
@@ -0,0 +1,121 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file scan.h
+ *  \brief Sequential implementations of scan functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/function_traits.h>
+#include <thrust/detail/type_traits/iterator/is_output_iterator.h>
+#include <thrust/detail/function.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator inclusive_scan(sequential::execution_policy<DerivedPolicy> &,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                BinaryFunction binary_op)
+{
+  using namespace thrust::detail;
+
+  // Use the input iterator's value type per https://wg21.link/P0571
+  using ValueType = typename thrust::iterator_value<InputIterator>::type;
+
+  // wrap binary_op
+  thrust::detail::wrapped_function<
+    BinaryFunction,
+    ValueType
+  > wrapped_binary_op(binary_op);
+
+  if(first != last)
+  {
+    ValueType sum = *first;
+
+    *result = *first;
+
+    for(++first, ++result; first != last; ++first, ++result)
+      *result = sum = wrapped_binary_op(sum,*first);
+  }
+
+  return result;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename InitialValueType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator exclusive_scan(sequential::execution_policy<DerivedPolicy> &,
+                                InputIterator first,
+                                InputIterator last,
+                                OutputIterator result,
+                                InitialValueType init,
+                                BinaryFunction binary_op)
+{
+  using namespace thrust::detail;
+
+  // Use the initial value type per https://wg21.link/P0571
+  using ValueType = InitialValueType;
+
+  if(first != last)
+  {
+    ValueType tmp = *first;  // temporary value allows in-situ scan
+    ValueType sum = init;
+
+    *result = sum;
+    sum = binary_op(sum, tmp);
+
+    for(++first, ++result; first != last; ++first, ++result)
+    {
+      tmp = *first;
+      *result = sum;
+      sum = binary_op(sum, tmp);
+    }
+  }
+
+  return result;
+} 
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan_by_key.h
new file mode 100644
index 000000000000..c428c1050e8c
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scan_by_key.h
@@ -0,0 +1,149 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file scan_by_key.h
+ *  \brief Sequential implementation of scan_by_key functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/function.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator inclusive_scan_by_key(sequential::execution_policy<DerivedPolicy> &,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       BinaryPredicate binary_pred,
+                                       BinaryFunction binary_op)
+{
+  using KeyType = typename thrust::iterator_traits<InputIterator1>::value_type;
+  using ValueType = typename thrust::iterator_traits<InputIterator2>::value_type;
+
+  // wrap binary_op
+  thrust::detail::wrapped_function<
+    BinaryFunction,
+    ValueType
+  > wrapped_binary_op(binary_op);
+
+  if(first1 != last1)
+  {
+    KeyType   prev_key   = *first1;
+    ValueType prev_value = *first2;
+
+    *result = prev_value;
+
+    for(++first1, ++first2, ++result;
+        first1 != last1;
+        ++first1, ++first2, ++result)
+    {
+      KeyType key = *first1;
+
+      if(binary_pred(prev_key, key))
+        *result = prev_value = wrapped_binary_op(prev_value,*first2);
+      else
+        *result = prev_value = *first2;
+
+      prev_key = key;
+    }
+  }
+
+  return result;
+}
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename T,
+         typename BinaryPredicate,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator exclusive_scan_by_key(sequential::execution_policy<DerivedPolicy> &,
+                                       InputIterator1 first1,
+                                       InputIterator1 last1,
+                                       InputIterator2 first2,
+                                       OutputIterator result,
+                                       T init,
+                                       BinaryPredicate binary_pred,
+                                       BinaryFunction binary_op)
+{
+  using KeyType = typename thrust::iterator_traits<InputIterator1>::value_type;
+  using ValueType = T;
+
+  if(first1 != last1)
+  {
+    KeyType   temp_key   = *first1;
+    ValueType temp_value = *first2;
+
+    ValueType next = init;
+
+    // first one is init
+    *result = next;
+
+    next = binary_op(next, temp_value);
+
+    for(++first1, ++first2, ++result;
+        first1 != last1;
+        ++first1, ++first2, ++result)
+    {
+      KeyType key = *first1;
+
+      // use temp to permit in-place scans
+      temp_value = *first2;
+
+      if (!binary_pred(temp_key, key))
+        next = init;  // reset sum
+
+      *result = next;  
+      next = binary_op(next, temp_value);
+
+      temp_key = key;
+    }
+  }
+
+  return result;
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scatter.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scatter.h
new file mode 100644
index 000000000000..d6817b4cb26b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/scatter.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special scatter functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sequence.h
new file mode 100644
index 000000000000..0c090f96f348
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sequence.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special sequence functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/set_operations.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/set_operations.h
new file mode 100644
index 000000000000..678754b45dd6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/set_operations.h
@@ -0,0 +1,223 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file set_operations.h
+ *  \brief Sequential implementation of set operation functions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/detail/copy.h>
+#include <thrust/detail/function.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  OutputIterator set_difference(sequential::execution_policy<DerivedPolicy> &exec,
+                                InputIterator1 first1,
+                                InputIterator1 last1,
+                                InputIterator2 first2,
+                                InputIterator2 last2,
+                                OutputIterator result,
+                                StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  while(first1 != last1 && first2 != last2)
+  {
+    if(wrapped_comp(*first1,*first2))
+    {
+      *result = *first1;
+      ++first1;
+      ++result;
+    } // end if
+    else if(wrapped_comp(*first2,*first1))
+    {
+      ++first2;
+    } // end else if
+    else
+    {
+      ++first1;
+      ++first2;
+    } // end else
+  } // end while
+
+  return thrust::copy(exec, first1, last1, result);
+} // end set_difference()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  OutputIterator set_intersection(sequential::execution_policy<DerivedPolicy> &,
+                                  InputIterator1 first1,
+                                  InputIterator1 last1,
+                                  InputIterator2 first2,
+                                  InputIterator2 last2,
+                                  OutputIterator result,
+                                  StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  while(first1 != last1 && first2 != last2)
+  {
+    if(wrapped_comp(*first1,*first2))
+    {
+      ++first1;
+    } // end if
+    else if(wrapped_comp(*first2,*first1))
+    {
+      ++first2;
+    } // end else if
+    else
+    {
+      *result = *first1;
+      ++first1;
+      ++first2;
+      ++result;
+    } // end else
+  } // end while
+
+  return result;
+} // end set_intersection()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  OutputIterator set_symmetric_difference(sequential::execution_policy<DerivedPolicy> &exec,
+                                          InputIterator1 first1,
+                                          InputIterator1 last1,
+                                          InputIterator2 first2,
+                                          InputIterator2 last2,
+                                          OutputIterator result,
+                                          StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  while(first1 != last1 && first2 != last2)
+  {
+    if(wrapped_comp(*first1,*first2))
+    {
+      *result = *first1;
+      ++first1;
+      ++result;
+    } // end if
+    else if(wrapped_comp(*first2,*first1))
+    {
+      *result = *first2;
+      ++first2;
+      ++result;
+    } // end else if
+    else
+    {
+      ++first1;
+      ++first2;
+    } // end else
+  } // end while
+
+  return thrust::copy(exec, first2, last2, thrust::copy(exec, first1, last1, result));
+} // end set_symmetric_difference()
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+  OutputIterator set_union(sequential::execution_policy<DerivedPolicy> &exec,
+                           InputIterator1 first1,
+                           InputIterator1 last1,
+                           InputIterator2 first2,
+                           InputIterator2 last2,
+                           OutputIterator result,
+                           StrictWeakOrdering comp)
+{
+  // wrap comp
+  thrust::detail::wrapped_function<
+    StrictWeakOrdering,
+    bool
+  > wrapped_comp(comp);
+
+  while(first1 != last1 && first2 != last2)
+  {
+    if(wrapped_comp(*first1,*first2))
+    {
+      *result = *first1;
+      ++first1;
+    } // end if
+    else if(wrapped_comp(*first2,*first1))
+    {
+      *result = *first2;
+      ++first2;
+    } // end else if
+    else
+    {
+      *result = *first1;
+      ++first1;
+      ++first2;
+    } // end else
+
+    ++result;
+  } // end while
+
+  return thrust::copy(exec, first2, last2, thrust::copy(exec, first1, last1, result));
+} // end set_union()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.h
new file mode 100644
index 000000000000..34cc7a8ba895
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.h
@@ -0,0 +1,63 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file sort.h
+ *  \brief Sequential implementations of sort algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                 RandomAccessIterator first,
+                 RandomAccessIterator last,
+                 StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                        RandomAccessIterator1 first1,
+                        RandomAccessIterator1 last1,
+                        RandomAccessIterator2 first2,
+                        StrictWeakOrdering comp);
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/sequential/sort.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.inl
new file mode 100644
index 000000000000..241a860aff57
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/sort.inl
@@ -0,0 +1,208 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/reverse.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/system/detail/sequential/stable_merge_sort.h>
+#include <thrust/system/detail/sequential/stable_primitive_sort.h>
+
+#include <nv/target>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+namespace sort_detail
+{
+
+
+////////////////////
+// Primitive Sort //
+////////////////////
+
+
+template<typename KeyType, typename Compare>
+struct needs_reverse
+  : thrust::detail::integral_constant<
+      bool,
+      thrust::detail::is_same<Compare, typename thrust::greater<KeyType> >::value
+    >
+{};
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                 RandomAccessIterator first,
+                 RandomAccessIterator last,
+                 StrictWeakOrdering,
+                 thrust::detail::true_type)
+{
+  thrust::system::detail::sequential::stable_primitive_sort(exec, first, last);
+
+  // if comp is greater<T> then reverse the keys
+  typedef typename thrust::iterator_traits<RandomAccessIterator>::value_type KeyType;
+
+  if(needs_reverse<KeyType,StrictWeakOrdering>::value)
+  {
+    thrust::reverse(exec, first, last);
+  }
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                        RandomAccessIterator1 first1,
+                        RandomAccessIterator1 last1,
+                        RandomAccessIterator2 first2,
+                        StrictWeakOrdering,
+                        thrust::detail::true_type)
+{
+  // if comp is greater<T> then reverse the keys and values
+  typedef typename thrust::iterator_traits<RandomAccessIterator1>::value_type KeyType;
+
+  // note, we also have to reverse the (unordered) input to preserve stability
+  if(needs_reverse<KeyType,StrictWeakOrdering>::value)
+  {
+    thrust::reverse(exec, first1,  last1);
+    thrust::reverse(exec, first2, first2 + (last1 - first1));
+  }
+
+  thrust::system::detail::sequential::stable_primitive_sort_by_key(exec, first1, last1, first2);
+
+  if(needs_reverse<KeyType,StrictWeakOrdering>::value)
+  {
+    thrust::reverse(exec, first1,  last1);
+    thrust::reverse(exec, first2, first2 + (last1 - first1));
+  }
+}
+
+
+////////////////
+// Merge Sort //
+////////////////
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                 RandomAccessIterator first,
+                 RandomAccessIterator last,
+                 StrictWeakOrdering comp,
+                 thrust::detail::false_type)
+{
+  thrust::system::detail::sequential::stable_merge_sort(exec, first, last, comp);
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                        RandomAccessIterator1 first1,
+                        RandomAccessIterator1 last1,
+                        RandomAccessIterator2 first2,
+                        StrictWeakOrdering comp,
+                        thrust::detail::false_type)
+{
+  thrust::system::detail::sequential::stable_merge_sort_by_key(exec, first1, last1, first2, comp);
+}
+
+
+template<typename KeyType, typename Compare>
+struct use_primitive_sort
+  : thrust::detail::and_<
+      thrust::detail::is_arithmetic<KeyType>,
+      thrust::detail::or_<
+        thrust::detail::is_same<Compare, thrust::less<KeyType> >,
+        thrust::detail::is_same<Compare, thrust::greater<KeyType> >
+      >
+    >
+{};
+
+
+} // end namespace sort_detail
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                 RandomAccessIterator first,
+                 RandomAccessIterator last,
+                 StrictWeakOrdering comp)
+{
+
+  // the compilation time of stable_primitive_sort is too expensive to use within a single CUDA thread
+  NV_IF_TARGET(NV_IS_HOST, (
+    using KeyType = thrust::iterator_value_t<RandomAccessIterator>;
+    sort_detail::use_primitive_sort<KeyType, StrictWeakOrdering> use_primitive_sort;
+    sort_detail::stable_sort(exec, first, last, comp, use_primitive_sort);
+  ), ( // NV_IS_DEVICE:
+    thrust::detail::false_type use_primitive_sort;
+    sort_detail::stable_sort(exec, first, last, comp, use_primitive_sort);
+  ));
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                        RandomAccessIterator1 first1,
+                        RandomAccessIterator1 last1,
+                        RandomAccessIterator2 first2,
+                        StrictWeakOrdering comp)
+{
+
+  // the compilation time of stable_primitive_sort_by_key is too expensive to use within a single CUDA thread
+  NV_IF_TARGET(NV_IS_HOST, (
+    using KeyType = thrust::iterator_value_t<RandomAccessIterator1>;
+    sort_detail::use_primitive_sort<KeyType, StrictWeakOrdering> use_primitive_sort;
+    sort_detail::stable_sort_by_key(exec, first1, last1, first2, comp, use_primitive_sort);
+  ), ( // NV_IS_DEVICE:
+    thrust::detail::false_type use_primitive_sort;
+    sort_detail::stable_sort_by_key(exec, first1, last1, first2, comp, use_primitive_sort);
+  ));
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.h
new file mode 100644
index 000000000000..64aa2bf96cec
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.h
@@ -0,0 +1,59 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_merge_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                       RandomAccessIterator begin,
+                       RandomAccessIterator end,
+                       StrictWeakOrdering comp);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_merge_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                              RandomAccessIterator1 keys_begin,
+                              RandomAccessIterator1 keys_end,
+                              RandomAccessIterator2 values_begin,
+                              StrictWeakOrdering comp);
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/sequential/stable_merge_sort.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.inl
new file mode 100644
index 000000000000..02f384afba1b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_merge_sort.inl
@@ -0,0 +1,393 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/merge.h>
+#include <thrust/system/detail/sequential/insertion_sort.h>
+#include <thrust/detail/minmax.h>
+
+#include <nv/target>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+namespace stable_merge_sort_detail
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void inplace_merge(sequential::execution_policy<DerivedPolicy> &exec,
+                   RandomAccessIterator first,
+                   RandomAccessIterator middle,
+                   RandomAccessIterator last,
+                   StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type;
+
+  thrust::detail::temporary_array<value_type, DerivedPolicy> a(exec, first, middle);
+  thrust::detail::temporary_array<value_type, DerivedPolicy> b(exec, middle, last);
+
+  thrust::merge(exec, a.begin(), a.end(), b.begin(), b.end(), first, comp);
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void inplace_merge_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                          RandomAccessIterator1 first1,
+                          RandomAccessIterator1 middle1,
+                          RandomAccessIterator1 last1,
+                          RandomAccessIterator2 first2,
+                          StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type value_type1;
+  typedef typename thrust::iterator_value<RandomAccessIterator2>::type value_type2;
+
+  RandomAccessIterator2 middle2 = first2 + (middle1 - first1);
+  RandomAccessIterator2 last2   = first2 + (last1   - first1);
+
+  thrust::detail::temporary_array<value_type1, DerivedPolicy> lhs1(exec, first1, middle1);
+  thrust::detail::temporary_array<value_type1, DerivedPolicy> rhs1(exec, middle1, last1);
+  thrust::detail::temporary_array<value_type2, DerivedPolicy> lhs2(exec, first2, middle2);
+  thrust::detail::temporary_array<value_type2, DerivedPolicy> rhs2(exec, middle2, last2);
+
+  thrust::merge_by_key(exec,
+                       lhs1.begin(), lhs1.end(),
+                       rhs1.begin(), rhs1.end(),
+                       lhs2.begin(), rhs2.begin(),
+                       first1, first2,
+                       comp);
+}
+
+
+template<typename RandomAccessIterator,
+         typename Size,
+         typename StrictWeakOrdering>
+__host__ __device__
+void insertion_sort_each(RandomAccessIterator first,
+                         RandomAccessIterator last,
+                         Size partition_size,
+                         StrictWeakOrdering comp)
+{
+  if(partition_size > 1)
+  {
+    for(; first < last; first += partition_size)
+    {
+      RandomAccessIterator partition_last = (thrust::min)(last, first + partition_size);
+
+      thrust::system::detail::sequential::insertion_sort(first, partition_last, comp);
+    } // end for
+  } // end if
+} // end insertion_sort_each()
+
+
+template<typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename Size,
+         typename StrictWeakOrdering>
+__host__ __device__
+void insertion_sort_each_by_key(RandomAccessIterator1 keys_first,
+                                RandomAccessIterator1 keys_last,
+                                RandomAccessIterator2 values_first,
+                                Size partition_size,
+                                StrictWeakOrdering comp)
+{
+  if(partition_size > 1)
+  {
+    for(; keys_first < keys_last; keys_first += partition_size, values_first += partition_size)
+    {
+      RandomAccessIterator1 keys_partition_last = (thrust::min)(keys_last, keys_first + partition_size);
+
+      thrust::system::detail::sequential::insertion_sort_by_key(keys_first, keys_partition_last, values_first, comp);
+    } // end for
+  } // end if
+} // end insertion_sort_each()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename Size,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void merge_adjacent_partitions(sequential::execution_policy<DerivedPolicy> &exec,
+                               RandomAccessIterator1 first,
+                               RandomAccessIterator1 last,
+                               Size partition_size,
+                               RandomAccessIterator2 result,
+                               StrictWeakOrdering comp)
+{
+  for(; first < last; first += 2 * partition_size, result += 2 * partition_size)
+  {
+    RandomAccessIterator1 interval_middle = (thrust::min)(last, first + partition_size);
+    RandomAccessIterator1 interval_last   = (thrust::min)(last, interval_middle + partition_size);
+
+    thrust::merge(exec,
+                  first, interval_middle,
+                  interval_middle, interval_last,
+                  result,
+                  comp);
+  } // end for
+} // end merge_adjacent_partitions()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename Size,
+         typename RandomAccessIterator3,
+         typename RandomAccessIterator4,
+         typename StrictWeakOrdering>
+__host__ __device__
+void merge_adjacent_partitions_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                      RandomAccessIterator1 keys_first,
+                                      RandomAccessIterator1 keys_last,
+                                      RandomAccessIterator2 values_first,
+                                      Size partition_size,
+                                      RandomAccessIterator3 keys_result,
+                                      RandomAccessIterator4 values_result,
+                                      StrictWeakOrdering comp)
+{
+  Size stride = 2 * partition_size;
+
+  for(;
+      keys_first < keys_last;
+      keys_first += stride, values_first += stride, keys_result += stride, values_result += stride)
+  {
+    RandomAccessIterator1 keys_interval_middle = (thrust::min)(keys_last, keys_first + partition_size);
+    RandomAccessIterator1 keys_interval_last   = (thrust::min)(keys_last, keys_interval_middle + partition_size);
+
+    RandomAccessIterator2 values_first2 = values_first + (keys_interval_middle - keys_first);
+
+    thrust::merge_by_key(exec,
+                         keys_first, keys_interval_middle,
+                         keys_interval_middle, keys_interval_last,
+                         values_first,
+                         values_first2,
+                         keys_result,
+                         values_result,
+                         comp);
+  } // end for
+} // end merge_adjacent_partitions()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void iterative_stable_merge_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                                 RandomAccessIterator first,
+                                 RandomAccessIterator last,
+                                 StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type value_type;
+  typedef typename thrust::iterator_difference<RandomAccessIterator>::type difference_type;
+
+  difference_type n = last - first;
+
+  thrust::detail::temporary_array<value_type, DerivedPolicy> temp(exec, n);
+
+  // insertion sort each 32 element partition
+  difference_type partition_size = 32;
+  insertion_sort_each(first, last, partition_size, comp);
+
+  // ping indicates whether or not the latest data is in the source range [first, last)
+  bool ping = true;
+
+  // merge adjacent partitions until the partition size covers the entire range
+  for(;
+      partition_size < n;
+      partition_size *= 2, ping = !ping)
+  {
+    if(ping)
+    {
+      merge_adjacent_partitions(exec, first, last, partition_size, temp.begin(), comp);
+    } // end if
+    else
+    {
+      merge_adjacent_partitions(exec, temp.begin(), temp.end(), partition_size, first, comp);
+    } // end else
+  } // end for m
+
+  if(!ping)
+  {
+    thrust::copy(exec, temp.begin(), temp.end(), first);
+  } // end if
+} // end iterative_stable_merge_sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void iterative_stable_merge_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                        RandomAccessIterator1 keys_first,
+                                        RandomAccessIterator1 keys_last,
+                                        RandomAccessIterator2 values_first,
+                                        StrictWeakOrdering comp)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type      value_type1;
+  typedef typename thrust::iterator_value<RandomAccessIterator2>::type      value_type2;
+  typedef typename thrust::iterator_difference<RandomAccessIterator1>::type difference_type;
+
+  difference_type n = keys_last - keys_first;
+
+  thrust::detail::temporary_array<value_type1, DerivedPolicy> keys_temp(exec, n);
+  thrust::detail::temporary_array<value_type2, DerivedPolicy> values_temp(exec, n);
+
+  // insertion sort each 32 element partition
+  difference_type partition_size = 32;
+  insertion_sort_each_by_key(keys_first, keys_last, values_first, partition_size, comp);
+
+  // ping indicates whether or not the latest data is in the source range [first, last)
+  bool ping = true;
+
+  // merge adjacent partitions until the partition size covers the entire range
+  for(;
+      partition_size < n;
+      partition_size *= 2, ping = !ping)
+  {
+    if(ping)
+    {
+      merge_adjacent_partitions_by_key(exec, keys_first, keys_last, values_first, partition_size, keys_temp.begin(), values_temp.begin(), comp);
+    } // end if
+    else
+    {
+      merge_adjacent_partitions_by_key(exec, keys_temp.begin(), keys_temp.end(), values_temp.begin(), partition_size, keys_first, values_first, comp);
+    } // end else
+  } // end for m
+
+  if(!ping)
+  {
+    thrust::copy(exec, keys_temp.begin(), keys_temp.end(), keys_first);
+    thrust::copy(exec, values_temp.begin(), values_temp.end(), values_first);
+  } // end if
+} // end iterative_stable_merge_sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void recursive_stable_merge_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                                 RandomAccessIterator first,
+                                 RandomAccessIterator last,
+                                 StrictWeakOrdering comp)
+{
+  if(last - first <= 32)
+  {
+    thrust::system::detail::sequential::insertion_sort(first, last, comp);
+  } // end if
+  else
+  {
+    RandomAccessIterator middle = first + (last - first) / 2;
+
+    stable_merge_sort_detail::recursive_stable_merge_sort(exec, first, middle, comp);
+    stable_merge_sort_detail::recursive_stable_merge_sort(exec, middle,  last, comp);
+    stable_merge_sort_detail::inplace_merge(exec, first, middle, last, comp);
+  } // end else
+} // end recursive_stable_merge_sort()
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void recursive_stable_merge_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                        RandomAccessIterator1 first1,
+                                        RandomAccessIterator1 last1,
+                                        RandomAccessIterator2 first2,
+                                        StrictWeakOrdering comp)
+{
+  if(last1 - first1 <= 32)
+  {
+    thrust::system::detail::sequential::insertion_sort_by_key(first1, last1, first2, comp);
+  } // end if
+  else
+  {
+    RandomAccessIterator1 middle1 = first1 + (last1 - first1) / 2;
+    RandomAccessIterator2 middle2 = first2 + (last1 - first1) / 2;
+
+    stable_merge_sort_detail::recursive_stable_merge_sort_by_key(exec, first1, middle1, first2,  comp);
+    stable_merge_sort_detail::recursive_stable_merge_sort_by_key(exec, middle1,  last1, middle2, comp);
+    stable_merge_sort_detail::inplace_merge_by_key(exec, first1, middle1, last1, first2, comp);
+  } // end else
+} // end recursive_stable_merge_sort_by_key()
+
+
+} // end namespace stable_merge_sort_detail
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_merge_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                       RandomAccessIterator first,
+                       RandomAccessIterator last,
+                       StrictWeakOrdering comp)
+{
+  NV_IF_TARGET(NV_IS_DEVICE, (
+    // avoid recursion in CUDA threads
+    stable_merge_sort_detail::iterative_stable_merge_sort(exec, first, last, comp);
+  ), (
+    stable_merge_sort_detail::recursive_stable_merge_sort(exec, first, last, comp);
+  ));
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename StrictWeakOrdering>
+__host__ __device__
+void stable_merge_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                              RandomAccessIterator1 first1,
+                              RandomAccessIterator1 last1,
+                              RandomAccessIterator2 first2,
+                              StrictWeakOrdering comp)
+{
+  NV_IF_TARGET(NV_IS_DEVICE, (
+    // avoid recursion in CUDA threads
+    stable_merge_sort_detail::iterative_stable_merge_sort_by_key(exec, first1, last1, first2, comp);
+  ), (
+    stable_merge_sort_detail::recursive_stable_merge_sort_by_key(exec, first1, last1, first2, comp);
+  ));
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.h
new file mode 100644
index 000000000000..acbb81217bd9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.h
@@ -0,0 +1,55 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+void stable_primitive_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                           RandomAccessIterator first,
+                           RandomAccessIterator last);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+void stable_primitive_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                  RandomAccessIterator1 keys_first,
+                                  RandomAccessIterator1 keys_last,
+                                  RandomAccessIterator2 values_first);
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/sequential/stable_primitive_sort.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl
new file mode 100644
index 000000000000..9897d6798da2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_primitive_sort.inl
@@ -0,0 +1,160 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/stable_primitive_sort.h>
+#include <thrust/system/detail/sequential/stable_radix_sort.h>
+#include <thrust/functional.h>
+#include <thrust/system/detail/sequential/partition.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/type_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+namespace stable_primitive_sort_detail
+{
+
+
+template<typename Iterator>
+  struct enable_if_bool_sort
+    : thrust::detail::enable_if<
+        thrust::detail::is_same<
+          bool,
+          typename thrust::iterator_value<Iterator>::type
+        >::value
+      >
+{};
+
+
+template<typename Iterator>
+  struct disable_if_bool_sort
+    : thrust::detail::disable_if<
+        thrust::detail::is_same<
+          bool,
+          typename thrust::iterator_value<Iterator>::type
+        >::value
+      >
+{};
+
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+  typename enable_if_bool_sort<RandomAccessIterator>::type
+__host__ __device__
+    stable_primitive_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                          RandomAccessIterator first, RandomAccessIterator last)
+{
+  // use stable_partition if we're sorting bool
+  // stable_partition puts true values first, so we need to logical_not
+  sequential::stable_partition(exec, first, last, thrust::logical_not<bool>());
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+  typename disable_if_bool_sort<RandomAccessIterator>::type
+__host__ __device__
+    stable_primitive_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                          RandomAccessIterator first, RandomAccessIterator last)
+{
+  // call stable_radix_sort
+  sequential::stable_radix_sort(exec,first,last);
+}
+
+
+struct logical_not_first
+{
+  template<typename Tuple>
+  __host__ __device__
+  bool operator()(Tuple t)
+  {
+    return !thrust::get<0>(t);
+  }
+};
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  typename enable_if_bool_sort<RandomAccessIterator1>::type
+__host__ __device__
+    stable_primitive_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                 RandomAccessIterator1 keys_first, RandomAccessIterator1 keys_last,
+                                 RandomAccessIterator2 values_first)
+{
+  // use stable_partition if we're sorting bool
+  // stable_partition puts true values first, so we need to logical_not
+  sequential::stable_partition(exec,
+                               thrust::make_zip_iterator(thrust::make_tuple(keys_first, values_first)),
+                               thrust::make_zip_iterator(thrust::make_tuple(keys_last, values_first)),
+                               logical_not_first());
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+  typename disable_if_bool_sort<RandomAccessIterator1>::type
+__host__ __device__
+    stable_primitive_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                 RandomAccessIterator1 keys_first, RandomAccessIterator1 keys_last,
+                                 RandomAccessIterator2 values_first)
+{
+  // call stable_radix_sort_by_key
+  sequential::stable_radix_sort_by_key(exec, keys_first, keys_last, values_first);
+}
+
+
+} // end stable_primitive_sort_detail
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+void stable_primitive_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                           RandomAccessIterator first,
+                           RandomAccessIterator last)
+{
+  stable_primitive_sort_detail::stable_primitive_sort(exec, first,last);
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+void stable_primitive_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                                  RandomAccessIterator1 keys_first,
+                                  RandomAccessIterator1 keys_last,
+                                  RandomAccessIterator2 values_first)
+{
+  stable_primitive_sort_detail::stable_primitive_sort_by_key(exec, keys_first, keys_last, values_first);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.h
new file mode 100644
index 000000000000..1e9713a2cc9b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.h
@@ -0,0 +1,55 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+void stable_radix_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                       RandomAccessIterator begin,
+                       RandomAccessIterator end);
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+void stable_radix_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                              RandomAccessIterator1 keys_begin,
+                              RandomAccessIterator1 keys_end,
+                              RandomAccessIterator2 values_begin);
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/sequential/stable_radix_sort.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.inl
new file mode 100644
index 000000000000..83d95ebfd499
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/stable_radix_sort.inl
@@ -0,0 +1,597 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/copy.h>
+#include <thrust/functional.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/iterator/transform_iterator.h>
+#include <thrust/iterator/zip_iterator.h>
+#include <thrust/detail/temporary_array.h>
+#include <thrust/detail/cstdint.h>
+#include <thrust/scatter.h>
+
+#include <limits>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+namespace radix_sort_detail
+{
+
+
+template <typename T>
+struct RadixEncoder : public thrust::identity<T>
+{};
+
+
+template <>
+struct RadixEncoder<char> : public thrust::unary_function<char, unsigned char>
+{
+  __host__ __device__
+  unsigned char operator()(char x) const
+  {
+    if(std::numeric_limits<char>::is_signed)
+    {
+      return static_cast<unsigned char>(x) ^ static_cast<unsigned char>(1) << (8 * sizeof(unsigned char) - 1);
+    }
+    else
+    {
+      return x;
+    }
+  }
+};
+
+template <>
+struct RadixEncoder<signed char> : public thrust::unary_function<signed char, unsigned char>
+{
+  __host__ __device__
+  unsigned char operator()(signed char x) const
+  {
+    return static_cast<unsigned char>(x) ^ static_cast<unsigned char>(1) << (8 * sizeof(unsigned char) - 1);
+  }
+};
+
+template <>
+struct RadixEncoder<short> : public thrust::unary_function<short, unsigned short>
+{
+  __host__ __device__
+  unsigned short operator()(short x) const
+  {
+    return static_cast<unsigned short>(x) ^ static_cast<unsigned short>(1) << (8 * sizeof(unsigned short) - 1);
+  }
+};
+
+template <>
+struct RadixEncoder<int> : public thrust::unary_function<int, unsigned int>
+{
+  __host__ __device__
+  unsigned long operator()(long x) const
+  {
+    return x ^ static_cast<unsigned int>(1) << (8 * sizeof(unsigned int) - 1);
+  }
+};
+
+template <>
+struct RadixEncoder<long> : public thrust::unary_function<long, unsigned long>
+{
+  __host__ __device__
+  unsigned long operator()(long x) const
+  {
+    return x ^ static_cast<unsigned long>(1) << (8 * sizeof(unsigned long) - 1);
+  }
+};
+
+template <>
+struct RadixEncoder<long long> : public thrust::unary_function<long long, unsigned long long>
+{
+  __host__ __device__
+  unsigned long long operator()(long long x) const
+  {
+    return x ^ static_cast<unsigned long long>(1) << (8 * sizeof(unsigned long long) - 1);
+  }
+};
+
+// ideally we'd use uint32 here and uint64 below
+template <>
+struct RadixEncoder<float> : public thrust::unary_function<float, thrust::detail::uint32_t>
+{
+  __host__ __device__
+  thrust::detail::uint32_t operator()(float x) const
+  {
+    union { float f; thrust::detail::uint32_t i; } u;
+    u.f = x;
+    thrust::detail::uint32_t mask = -static_cast<thrust::detail::int32_t>(u.i >> 31) | (static_cast<thrust::detail::uint32_t>(1) << 31);
+    return u.i ^ mask;
+  }
+};
+
+template <>
+struct RadixEncoder<double> : public thrust::unary_function<double, thrust::detail::uint64_t>
+{
+  __host__ __device__
+  thrust::detail::uint64_t operator()(double x) const
+  {
+    union { double f; thrust::detail::uint64_t i; } u;
+    u.f = x;
+    thrust::detail::uint64_t mask = -static_cast<thrust::detail::int64_t>(u.i >> 63) | (static_cast<thrust::detail::uint64_t>(1) << 63);
+    return u.i ^ mask;
+  }
+};
+
+
+// this functor returns a key's to its histogram bucket count and post-increments the bucket
+template<unsigned int RadixBits, typename KeyType>
+  struct bucket_functor
+{
+  typedef RadixEncoder<KeyType> Encoder;
+  typedef typename Encoder::result_type EncodedType;
+  typedef size_t result_type;
+  static const EncodedType BitMask = static_cast<EncodedType>((1 << RadixBits) - 1);
+
+  Encoder encode;
+  EncodedType bit_shift;
+  size_t *histogram;
+
+  __host__ __device__
+  bucket_functor(EncodedType bit_shift, size_t *histogram)
+    : encode(),
+      bit_shift(bit_shift),
+      histogram(histogram)
+  {}
+
+  inline __host__ __device__
+  size_t operator()(KeyType key)
+  {
+    const EncodedType x = encode(key);
+
+    // note that we mutate the histogram here
+    return histogram[(x >> bit_shift) & BitMask]++;
+  }
+};
+
+
+template<unsigned int RadixBits,
+         typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename Integer>
+inline __host__ __device__
+void radix_shuffle_n(sequential::execution_policy<DerivedPolicy> &exec,
+                     RandomAccessIterator1 first,
+                     const size_t n,
+                     RandomAccessIterator2 result,
+                     Integer bit_shift,
+                     size_t *histogram)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type KeyType;
+
+  // note that we are going to mutate the histogram during this sequential scatter
+  thrust::scatter(exec,
+                  first, first + n,
+                  thrust::make_transform_iterator(first, bucket_functor<RadixBits,KeyType>(bit_shift, histogram)),
+                  result);
+}
+
+
+template<unsigned int RadixBits,
+         typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename RandomAccessIterator3,
+         typename RandomAccessIterator4,
+         typename Integer>
+__host__ __device__
+void radix_shuffle_n(sequential::execution_policy<DerivedPolicy> &exec,
+                     RandomAccessIterator1 keys_first,
+                     RandomAccessIterator2 values_first,
+                     const size_t n,
+                     RandomAccessIterator3 keys_result,
+                     RandomAccessIterator4 values_result,
+                     Integer bit_shift,
+                     size_t *histogram)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type KeyType;
+
+  // note that we are going to mutate the histogram during this sequential scatter
+  thrust::scatter(exec,
+                  thrust::make_zip_iterator(thrust::make_tuple(keys_first, values_first)),
+                  thrust::make_zip_iterator(thrust::make_tuple(keys_first + n, values_first + n)),
+                  thrust::make_transform_iterator(keys_first, bucket_functor<RadixBits,KeyType>(bit_shift, histogram)),
+                  thrust::make_zip_iterator(thrust::make_tuple(keys_result, values_result)));
+}
+
+
+template<unsigned int RadixBits,
+         bool HasValues,
+         typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename RandomAccessIterator3,
+         typename RandomAccessIterator4>
+__host__ __device__
+void radix_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                RandomAccessIterator1 keys1,
+                RandomAccessIterator2 keys2,
+                RandomAccessIterator3 vals1,
+                RandomAccessIterator4 vals2,
+                const size_t N)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type KeyType;
+
+  typedef RadixEncoder<KeyType> Encoder;
+  typedef typename Encoder::result_type EncodedType;
+
+  const unsigned int NumHistograms = (8 * sizeof(EncodedType) + (RadixBits - 1)) / RadixBits;
+  const unsigned int HistogramSize =  1 << RadixBits;
+
+  const EncodedType BitMask = static_cast<EncodedType>((1 << RadixBits) - 1);
+
+  Encoder encode;
+
+  // storage for histograms
+  size_t histograms[NumHistograms][HistogramSize] = {{0}};
+
+  // see which passes can be eliminated
+  bool skip_shuffle[NumHistograms] = {false};
+
+  // false if most recent data is stored in (keys1,vals1)
+  bool flip = false;
+
+  // compute histograms
+  for(size_t i = 0; i < N; i++)
+  {
+    const EncodedType x = encode(keys1[i]);
+
+    for(unsigned int j = 0; j < NumHistograms; j++)
+    {
+      const auto BitShift = static_cast<EncodedType>(RadixBits * j);
+      histograms[j][(x >> BitShift) & BitMask]++;
+    }
+  }
+
+  // scan histograms
+  for(unsigned int i = 0; i < NumHistograms; i++)
+  {
+    size_t sum = 0;
+
+    for(unsigned int j = 0; j < HistogramSize; j++)
+    {
+      size_t bin = histograms[i][j];
+
+      if(bin == N)
+        skip_shuffle[i] = true;
+
+      histograms[i][j] = sum;
+
+      sum = sum + bin;
+    }
+  }
+
+  // shuffle keys and (optionally) values
+  for(unsigned int i = 0; i < NumHistograms; i++)
+  {
+    const EncodedType BitShift = static_cast<EncodedType>(RadixBits * i);
+
+    if(!skip_shuffle[i])
+    {
+      if(flip)
+      {
+        if(HasValues)
+        {
+          radix_shuffle_n<RadixBits>(exec, keys2, vals2, N, keys1, vals1, BitShift, histograms[i]);
+        }
+        else
+        {
+          radix_shuffle_n<RadixBits>(exec, keys2, N, keys1, BitShift, histograms[i]);
+        }
+      }
+      else
+      {
+        if(HasValues)
+        {
+          radix_shuffle_n<RadixBits>(exec, keys1, vals1, N, keys2, vals2, BitShift, histograms[i]);
+        }
+        else
+        {
+          radix_shuffle_n<RadixBits>(exec, keys1, N, keys2, BitShift, histograms[i]);
+        }
+      }
+
+      flip = (flip) ? false : true;
+    }
+  }
+
+  // ensure final values are in (keys1,vals1)
+  if(flip)
+  {
+    thrust::copy(exec, keys2, keys2 + N, keys1);
+
+    if(HasValues)
+    {
+      thrust::copy(exec, vals2, vals2 + N, vals1);
+    }
+  }
+}
+
+
+// Select best radix sort parameters based on sizeof(T) and input size
+// These particular values were determined through empirical testing on a Core i7 950 CPU
+template <size_t KeySize>
+struct radix_sort_dispatcher
+{
+};
+
+template <>
+struct radix_sort_dispatcher<1>
+{
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  const size_t N)
+  {
+    radix_sort_detail::radix_sort<8,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+  }
+
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2,
+           typename RandomAccessIterator3,
+           typename RandomAccessIterator4>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  RandomAccessIterator3 vals1, RandomAccessIterator4 vals2,
+                  const size_t N)
+  {
+    radix_sort_detail::radix_sort<8,true>(exec, keys1, keys2, vals1, vals2, N);
+  }
+};
+
+
+template <>
+struct radix_sort_dispatcher<2>
+{
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  const size_t N)
+  {
+#ifdef __QNX__
+    // XXX war for nvbug 200193674
+    const bool condition = true;
+#else
+    const bool condition = N < (1 << 16);
+#endif
+    if (condition)
+    {
+      radix_sort_detail::radix_sort<8,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+    }
+    else
+    {
+      radix_sort_detail::radix_sort<16,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+    }
+  }
+
+
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2,
+           typename RandomAccessIterator3,
+           typename RandomAccessIterator4>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  RandomAccessIterator3 vals1, RandomAccessIterator4 vals2,
+                  const size_t N)
+  {
+#ifdef __QNX__
+    // XXX war for nvbug 200193674
+    const bool condition = true;
+#else
+    const bool condition = N < (1 << 15);
+#endif
+    if (condition)
+    {
+      radix_sort_detail::radix_sort<8,true>(exec, keys1, keys2, vals1, vals2, N);
+    }
+    else
+    {
+      radix_sort_detail::radix_sort<16,true>(exec, keys1, keys2, vals1, vals2, N);
+    }
+  }
+};
+
+
+template <>
+struct radix_sort_dispatcher<4>
+{
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  const size_t N)
+  {
+    if(N < (1 << 22))
+    {
+      radix_sort_detail::radix_sort<8,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+    }
+    else
+    {
+      radix_sort_detail::radix_sort<4,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+    }
+  }
+
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2,
+           typename RandomAccessIterator3,
+           typename RandomAccessIterator4>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  RandomAccessIterator3 vals1, RandomAccessIterator4 vals2,
+                  const size_t N)
+  {
+    if(N < (1 << 22))
+    {
+      radix_sort_detail::radix_sort<8,true>(exec, keys1, keys2, vals1, vals2, N);
+    }
+    else
+    {
+      radix_sort_detail::radix_sort<3,true>(exec, keys1, keys2, vals1, vals2, N);
+    }
+  }
+};
+
+
+template <>
+struct radix_sort_dispatcher<8>
+{
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  const size_t N)
+  {
+    if(N < (1 << 21))
+    {
+      radix_sort_detail::radix_sort<8,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+    }
+    else
+    {
+      radix_sort_detail::radix_sort<4,false>(exec, keys1, keys2, static_cast<int *>(0), static_cast<int *>(0), N);
+    }
+  }
+
+  template<typename DerivedPolicy,
+           typename RandomAccessIterator1,
+           typename RandomAccessIterator2,
+           typename RandomAccessIterator3,
+           typename RandomAccessIterator4>
+  __host__ __device__
+  void operator()(sequential::execution_policy<DerivedPolicy> &exec,
+                  RandomAccessIterator1 keys1, RandomAccessIterator2 keys2,
+                  RandomAccessIterator3 vals1, RandomAccessIterator4 vals2,
+                  const size_t N)
+  {
+    if(N < (1 << 21))
+    {
+      radix_sort_detail::radix_sort<8,true>(exec, keys1, keys2, vals1, vals2, N);
+    }
+    else
+    {
+      radix_sort_detail::radix_sort<3,true>(exec, keys1, keys2, vals1, vals2, N);
+    }
+  }
+};
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+void radix_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                RandomAccessIterator1 keys1,
+                RandomAccessIterator2 keys2,
+                const size_t N)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type KeyType;
+  radix_sort_dispatcher<sizeof(KeyType)>()(exec, keys1, keys2, N);
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2,
+         typename RandomAccessIterator3,
+         typename RandomAccessIterator4>
+__host__ __device__
+void radix_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                RandomAccessIterator1 keys1,
+                RandomAccessIterator2 keys2,
+                RandomAccessIterator3 vals1,
+                RandomAccessIterator4 vals2,
+                const size_t N)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type KeyType;
+  radix_sort_dispatcher<sizeof(KeyType)>()(exec, keys1, keys2, vals1, vals2, N);
+}
+
+
+} // namespace radix_sort_detail
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator>
+__host__ __device__
+void stable_radix_sort(sequential::execution_policy<DerivedPolicy> &exec,
+                       RandomAccessIterator first,
+                       RandomAccessIterator last)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator>::type KeyType;
+
+  size_t N = last - first;
+
+  thrust::detail::temporary_array<KeyType, DerivedPolicy> temp(exec, N);
+
+  radix_sort_detail::radix_sort(exec, first, temp.begin(), N);
+}
+
+
+template<typename DerivedPolicy,
+         typename RandomAccessIterator1,
+         typename RandomAccessIterator2>
+__host__ __device__
+void stable_radix_sort_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                              RandomAccessIterator1 first1,
+                              RandomAccessIterator1 last1,
+                              RandomAccessIterator2 first2)
+{
+  typedef typename thrust::iterator_value<RandomAccessIterator1>::type KeyType;
+  typedef typename thrust::iterator_value<RandomAccessIterator2>::type ValueType;
+
+  size_t N = last1 - first1;
+
+  thrust::detail::temporary_array<KeyType, DerivedPolicy>   temp1(exec, N);
+  thrust::detail::temporary_array<ValueType, DerivedPolicy> temp2(exec, N);
+
+  radix_sort_detail::radix_sort(exec, first1, temp1.begin(), first2, temp2.begin(), N);
+}
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/swap_ranges.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/swap_ranges.h
new file mode 100644
index 000000000000..497497d6a152
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/swap_ranges.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special swap_ranges functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/tabulate.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/tabulate.h
new file mode 100644
index 000000000000..926cdec15a0e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/tabulate.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special tabulate functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/temporary_buffer.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/temporary_buffer.h
new file mode 100644
index 000000000000..2adfaf2810c6
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/temporary_buffer.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special temporary buffer functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform.h
new file mode 100644
index 000000000000..30305c152879
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special transform functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_reduce.h
new file mode 100644
index 000000000000..8d2a1b3850de
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_reduce.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special transform_reduce functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_scan.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/transform_scan.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/trivial_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/trivial_copy.h
new file mode 100644
index 000000000000..ea55c8fd2019
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/trivial_copy.h
@@ -0,0 +1,61 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file trivial_copy.h
+ *  \brief Sequential copy algorithms for plain-old-data.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <cstring>
+#include <thrust/system/detail/sequential/general_copy.h>
+
+#include <nv/target>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+template<typename T>
+__host__ __device__
+  T *trivial_copy_n(const T *first,
+                    std::ptrdiff_t n,
+                    T *result)
+{
+  T* return_value = NULL;
+
+  NV_IF_TARGET(NV_IS_HOST, (
+    std::memmove(result, first, n * sizeof(T));
+    return_value = result + n;
+  ), ( // NV_IS_DEVICE:
+    return_value = thrust::system::detail::sequential::general_copy_n(first, n, result);
+  ));
+
+  return return_value;
+} // end trivial_copy_n()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_copy.h
new file mode 100644
index 000000000000..c6ae90664ad9
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_copy.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special version of this algorithm 
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_fill.h
new file mode 100644
index 000000000000..65e59fae5dce
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/uninitialized_fill.h
@@ -0,0 +1,22 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// this system has no special unintialized_fill functions
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique.h
new file mode 100644
index 000000000000..c4fe5268a337
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique.h
@@ -0,0 +1,130 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file unique.h
+ *  \brief Sequential implementations of unique algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  OutputIterator unique_copy(sequential::execution_policy<DerivedPolicy> &,
+                             InputIterator first,
+                             InputIterator last,
+                             OutputIterator output,
+                             BinaryPredicate binary_pred)
+{
+  typedef typename thrust::iterator_traits<InputIterator>::value_type T;
+
+  if(first != last)
+  {
+    T prev = *first;
+
+    for(++first; first != last; ++first)
+    {
+      T temp = *first;
+
+      if (!binary_pred(prev, temp))
+      {
+        *output = prev;
+
+        ++output;
+
+        prev = temp;
+      }
+    }
+
+    *output = prev;
+    ++output;
+  }
+
+  return output;
+} // end unique_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  ForwardIterator unique(sequential::execution_policy<DerivedPolicy> &exec,
+                         ForwardIterator first,
+                         ForwardIterator last,
+                         BinaryPredicate binary_pred)
+{
+  // sequential unique_copy permits in-situ operation
+  return sequential::unique_copy(exec, first, last, first, binary_pred);
+} // end unique()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(sequential::execution_policy<DerivedPolicy> &,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred)
+{
+  typedef typename thrust::iterator_traits<ForwardIterator>::value_type T;
+  typename thrust::iterator_traits<ForwardIterator>::difference_type count{};
+
+  if(first != last)
+  {
+    count++;
+    T prev = *first;
+
+    for(++first; first != last; ++first)
+    {
+      T temp = *first;
+
+      if (!binary_pred(prev, temp))
+      {
+        count++;
+        prev = temp;
+      }
+    }
+  }
+
+  return count;
+} // end unique()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique_by_key.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique_by_key.h
new file mode 100644
index 000000000000..d30cc7c71530
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/sequential/unique_by_key.h
@@ -0,0 +1,115 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file unique_by_key.h
+ *  \brief Sequential implementations of unique_by_key algorithms.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/system/detail/sequential/execution_policy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+namespace system
+{
+namespace detail
+{
+namespace sequential
+{
+
+
+__thrust_exec_check_disable__
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(sequential::execution_policy<DerivedPolicy> &,
+                       InputIterator1 keys_first, 
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_output,
+                       OutputIterator2 values_output,
+                       BinaryPredicate binary_pred)
+{
+  typedef typename thrust::iterator_traits<InputIterator1>::value_type  InputKeyType;
+  typedef typename thrust::iterator_traits<OutputIterator2>::value_type OutputValueType;
+
+  if(keys_first != keys_last)
+  {
+    InputKeyType    temp_key   = *keys_first;
+    OutputValueType temp_value = *values_first;
+
+    for(++keys_first, ++values_first;
+        keys_first != keys_last;
+        ++keys_first, ++values_first)
+    {
+      InputKeyType    key   = *keys_first;
+      OutputValueType value = *values_first;
+
+      if(!binary_pred(temp_key, key))
+      {
+        *keys_output   = temp_key;
+        *values_output = temp_value;
+
+        ++keys_output;
+        ++values_output;
+
+        temp_key   = key;
+        temp_value = value;
+      }
+    }
+
+    *keys_output   = temp_key;
+    *values_output = temp_value;
+
+    ++keys_output;
+    ++values_output;
+  }
+
+  return thrust::make_pair(keys_output, values_output);
+} // end unique_by_key_copy()
+
+
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+    unique_by_key(sequential::execution_policy<DerivedPolicy> &exec,
+                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_last,
+                  ForwardIterator2 values_first,
+                  BinaryPredicate binary_pred)
+{
+  // sequential unique_by_key_copy() permits in-situ operation
+  return sequential::unique_by_key_copy(exec, keys_first, keys_last, values_first, keys_first, values_first, binary_pred);
+} // end unique_by_key()
+
+
+} // end namespace sequential
+} // end namespace detail
+} // end namespace system
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/system_error.inl b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/system_error.inl
new file mode 100644
index 000000000000..075fe88e4639
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/detail/system_error.inl
@@ -0,0 +1,112 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/system/system_error.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+
+system_error
+  ::system_error(error_code ec, const std::string &what_arg)
+    : std::runtime_error(what_arg), m_error_code(ec)
+{
+
+} // end system_error::system_error()
+
+
+system_error
+  ::system_error(error_code ec, const char *what_arg)
+    : std::runtime_error(what_arg), m_error_code(ec)
+{
+  ;
+} // end system_error::system_error()
+
+
+system_error
+  ::system_error(error_code ec)
+    : std::runtime_error(""), m_error_code(ec)
+{
+  ;
+} // end system_error::system_error()
+
+
+system_error
+  ::system_error(int ev, const error_category &ecat, const std::string &what_arg)
+    : std::runtime_error(what_arg), m_error_code(ev,ecat)
+{
+  ;
+} // end system_error::system_error()
+
+
+system_error
+  ::system_error(int ev, const error_category &ecat, const char *what_arg)
+    : std::runtime_error(what_arg), m_error_code(ev,ecat)
+{
+  ;
+} // end system_error::system_error()
+
+
+system_error
+  ::system_error(int ev, const error_category &ecat)
+    : std::runtime_error(""), m_error_code(ev,ecat)
+{
+  ;
+} // end system_error::system_error()
+
+
+const error_code &system_error
+  ::code(void) const noexcept
+{
+  return m_error_code;
+} // end system_error::code()
+
+
+const char *system_error
+  ::what(void) const noexcept
+{
+  if(m_what.empty())
+  {
+    try
+    {
+      m_what = this->std::runtime_error::what();
+      if(m_error_code)
+      {
+        if(!m_what.empty()) m_what += ": ";
+        m_what += m_error_code.message();
+      }
+    }
+    catch(...)
+    {
+      return std::runtime_error::what();
+    }
+  }
+
+  return m_what.c_str();
+} // end system_error::what()
+
+
+} // end system
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/error_code.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/error_code.h
new file mode 100644
index 000000000000..d460a315b1d5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/error_code.h
@@ -0,0 +1,522 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file error_code.h
+ *  \brief An object used to hold error values, such as those originating from the
+ *         operating system or other low-level application program interfaces.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/system/detail/errno.h>
+#include <iostream>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+
+/*! \addtogroup system_diagnostics
+ *  \{
+ */
+
+class error_condition;
+class error_code;
+
+/*! A metafunction returning whether or not the parameter is an \p error_code enum.
+ */
+template<typename T> struct is_error_code_enum : public thrust::detail::false_type {};
+
+/*! A metafunction returning whether or not the parameter is an \p error_condition enum.
+ */
+template<typename T> struct is_error_condition_enum : public thrust::detail::false_type {};
+
+
+// XXX N3092 prefers enum class errc { ... }
+namespace errc
+{
+
+/*! An enum containing common error codes.
+ */
+enum errc_t
+{
+  address_family_not_supported       = detail::eafnosupport,
+  address_in_use                     = detail::eaddrinuse,
+  address_not_available              = detail::eaddrnotavail,
+  already_connected                  = detail::eisconn,
+  argument_list_too_long             = detail::e2big,
+  argument_out_of_domain             = detail::edom,
+  bad_address                        = detail::efault,
+  bad_file_descriptor                = detail::ebadf,
+  bad_message                        = detail::ebadmsg,
+  broken_pipe                        = detail::epipe,
+  connection_aborted                 = detail::econnaborted,
+  connection_already_in_progress     = detail::ealready,
+  connection_refused                 = detail::econnrefused,
+  connection_reset                   = detail::econnreset,
+  cross_device_link                  = detail::exdev,
+  destination_address_required       = detail::edestaddrreq,
+  device_or_resource_busy            = detail::ebusy,
+  directory_not_empty                = detail::enotempty,
+  executable_format_error            = detail::enoexec,
+  file_exists                        = detail::eexist,
+  file_too_large                     = detail::efbig,
+  filename_too_long                  = detail::enametoolong,
+  function_not_supported             = detail::enosys,
+  host_unreachable                   = detail::ehostunreach,
+  identifier_removed                 = detail::eidrm,
+  illegal_byte_sequence              = detail::eilseq,
+  inappropriate_io_control_operation = detail::enotty,
+  interrupted                        = detail::eintr,
+  invalid_argument                   = detail::einval,
+  invalid_seek                       = detail::espipe,
+  io_error                           = detail::eio,
+  is_a_directory                     = detail::eisdir,
+  message_size                       = detail::emsgsize,
+  network_down                       = detail::enetdown,
+  network_reset                      = detail::enetreset,
+  network_unreachable                = detail::enetunreach,
+  no_buffer_space                    = detail::enobufs,
+  no_child_process                   = detail::echild,
+  no_link                            = detail::enolink,
+  no_lock_available                  = detail::enolck,
+  no_message_available               = detail::enodata,
+  no_message                         = detail::enomsg,
+  no_protocol_option                 = detail::enoprotoopt,
+  no_space_on_device                 = detail::enospc,
+  no_stream_resources                = detail::enosr,
+  no_such_device_or_address          = detail::enxio,
+  no_such_device                     = detail::enodev,
+  no_such_file_or_directory          = detail::enoent,
+  no_such_process                    = detail::esrch,
+  not_a_directory                    = detail::enotdir,
+  not_a_socket                       = detail::enotsock,
+  not_a_stream                       = detail::enostr,
+  not_connected                      = detail::enotconn,
+  not_enough_memory                  = detail::enomem,
+  not_supported                      = detail::enotsup,
+  operation_canceled                 = detail::ecanceled,
+  operation_in_progress              = detail::einprogress,
+  operation_not_permitted            = detail::eperm,
+  operation_not_supported            = detail::eopnotsupp,
+  operation_would_block              = detail::ewouldblock,
+  owner_dead                         = detail::eownerdead,
+  permission_denied                  = detail::eacces,
+  protocol_error                     = detail::eproto,
+  protocol_not_supported             = detail::eprotonosupport,
+  read_only_file_system              = detail::erofs,
+  resource_deadlock_would_occur      = detail::edeadlk,
+  resource_unavailable_try_again     = detail::eagain,
+  result_out_of_range                = detail::erange,
+  state_not_recoverable              = detail::enotrecoverable,
+  stream_timeout                     = detail::etime,
+  text_file_busy                     = detail::etxtbsy,
+  timed_out                          = detail::etimedout,
+  too_many_files_open_in_system      = detail::enfile,
+  too_many_files_open                = detail::emfile,
+  too_many_links                     = detail::emlink,
+  too_many_symbolic_link_levels      = detail::eloop,
+  value_too_large                    = detail::eoverflow,
+  wrong_protocol_type                = detail::eprototype
+}; // end errc_t
+
+} // end namespace errc
+
+
+/*! Specialization of \p is_error_condition_enum for \p errc::errc_t
+ */
+template<> struct is_error_condition_enum<errc::errc_t> : public thrust::detail::true_type {};
+
+
+// [19.5.1.1] class error_category
+
+/*! \brief The class \p error_category serves as a base class for types used to identify the
+ *         source and encoding of a particular category of error code. Classes may be derived
+ *         from \p error_category to support categories of errors in addition to those defined
+ *         in the C++ International Standard.
+ */
+class error_category
+{
+  public:
+    /*! Destructor does nothing.
+     */
+    inline virtual ~error_category(void);
+
+    // XXX enable upon c++0x
+    // error_category(const error_category &) = delete;
+    // error_category &operator=(const error_category &) = delete;
+
+    /*! \return A string naming the error category.
+     */
+    inline virtual const char *name(void) const = 0;
+
+    /*! \return \p error_condition(ev, *this).
+     */
+    inline virtual error_condition default_error_condition(int ev) const;
+
+    /*! \return <tt>default_error_condition(code) == condition</tt>
+     */
+    inline virtual bool equivalent(int code, const error_condition &condition) const;
+
+    /*! \return <tt>*this == code.category() && code.value() == condition</tt>
+     */
+    inline virtual bool equivalent(const error_code &code, int condition) const;
+
+    /*! \return A string that describes the error condition denoted by \p ev.
+     */
+    virtual std::string message(int ev) const = 0;
+
+    /*! \return <tt>*this == &rhs</tt>
+     */
+    inline bool operator==(const error_category &rhs) const;
+
+    /*! \return <tt>!(*this == rhs)</tt>
+     */
+    inline bool operator!=(const error_category &rhs) const;
+
+    /*! \return <tt>less<const error_category*>()(this, &rhs)</tt>
+     *  \note \c less provides a total ordering for pointers.
+     */
+    inline bool operator<(const error_category &rhs) const;
+}; // end error_category
+
+
+// [19.5.1.5] error_category objects
+
+
+/*! \return A reference to an object of a type derived from class \p error_category.
+ *  \note The object's \p default_error_condition and \p equivalent virtual functions
+ *        shall behave as specified for the class \p error_category. The object's
+ *        \p name virtual function shall return a pointer to the string <tt>"generic"</tt>.
+ */
+inline const error_category &generic_category(void);
+
+
+/*! \return A reference to an object of a type derived from class \p error_category.
+ *  \note The object's \p equivalent virtual functions shall behave as specified for
+ *        class \p error_category. The object's \p name virtual function shall return
+ *        a pointer to the string <tt>"system"</tt>. The object's \p default_error_condition
+ *        virtual function shall behave as follows:
+ *
+ *        If the argument <tt>ev</tt> corresponds to a POSIX <tt>errno</tt> value
+ *        \c posv, the function shall return <tt>error_condition(ev,generic_category())</tt>.
+ *        Otherwise, the function shall return <tt>error_condition(ev,system_category())</tt>.
+ *        What constitutes correspondence for any given operating system is unspecified.
+ */
+inline const error_category &system_category(void);
+
+
+// [19.5.2] Class error_code
+
+
+/*! \brief The class \p error_code describes an object used to hold error code values, such as
+ *         those originating from the operating system or other low-level application program
+ *         interfaces.
+ */
+class error_code
+{
+  public:
+    // [19.5.2.2] constructors:
+
+    /*! Effects: Constructs an object of type \p error_code.
+     *  \post <tt>value() == 0</tt> and <tt>category() == &system_category()</tt>.
+     */
+    inline error_code(void);
+
+    /*! Effects: Constructs an object of type \p error_code.
+     *  \post <tt>value() == val</tt> and <tt>category() == &cat</tt>.
+     */
+    inline error_code(int val, const error_category &cat);
+
+    /*! Effects: Constructs an object of type \p error_code.
+     *  \post <tt>*this == make_error_code(e)</tt>.
+     */
+    template <typename ErrorCodeEnum>
+      error_code(ErrorCodeEnum e
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+        , typename thrust::detail::enable_if<is_error_code_enum<ErrorCodeEnum>::value>::type * = 0
+#endif // THRUST_HOST_COMPILER_MSVC
+        );
+
+    // [19.5.2.3] modifiers:
+
+    /*! \post <tt>value() == val</tt> and <tt>category() == &cat</tt>.
+     */
+    inline void assign(int val, const error_category &cat);
+
+    /*! \post <tt>*this == make_error_code(e)</tt>.
+     */
+    template <typename ErrorCodeEnum>
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+      typename thrust::detail::enable_if<is_error_code_enum<ErrorCodeEnum>::value, error_code>::type &
+#else
+      error_code &
+#endif // THRUST_HOST_COMPILER_MSVC
+        operator=(ErrorCodeEnum e);
+
+    /*! \post <tt>value() == 0</tt> and <tt>category() == system_category()</tt>.
+     */
+    inline void clear(void);
+
+    // [19.5.2.4] observers:
+
+    /*! \return An integral value of this \p error_code object.
+     */
+    inline int value(void) const;
+
+    /*! \return An \p error_category describing the category of this \p error_code object.
+     */
+    inline const error_category &category(void) const;
+
+    /*! \return <tt>category().default_error_condition()</tt>.
+     */
+    inline error_condition default_error_condition(void) const;
+
+    /*! \return <tt>category().message(value())</tt>.
+     */
+    inline std::string message(void) const;
+
+    // XXX replace the below upon c++0x
+    // inline explicit operator bool (void) const;
+
+    /*! \return <tt>value() != 0</tt>.
+     */
+    inline operator bool (void) const;
+
+    /*! \cond
+     */
+  private:
+    int m_val;
+    const error_category *m_cat;
+    /*! \endcond
+     */
+}; // end error_code
+
+
+// [19.5.2.5] Class error_code non-member functions
+
+
+// XXX replace errc::errc_t with errc upon c++0x
+/*! \return <tt>error_code(static_cast<int>(e), generic_category())</tt>
+ */
+inline error_code make_error_code(errc::errc_t e);
+
+
+/*! \return <tt>lhs.category() < rhs.category() || lhs.category() == rhs.category() && lhs.value() < rhs.value()</tt>.
+ */
+inline bool operator<(const error_code &lhs, const error_code &rhs);
+
+
+/*! Effects: <tt>os << ec.category().name() << ':' << ec.value()</tt>.
+ */
+template <typename charT, typename traits>
+  std::basic_ostream<charT,traits>&
+    operator<<(std::basic_ostream<charT,traits>& os, const error_code &ec);
+
+
+// [19.5.3] class error_condition
+
+
+/*! \brief The class \p error_condition describes an object used to hold values identifying
+ *  error conditions.
+ *
+ *  \note \p error_condition values are portable abstractions, while \p error_code values
+ *        are implementation specific.
+ */
+class error_condition
+{
+  public:
+    // [19.5.3.2] constructors
+
+    /*! Constructs an object of type \p error_condition.
+     *  \post <tt>value() == 0</tt>.
+     *  \post <tt>category() == generic_category()</tt>.
+     */
+    inline error_condition(void);
+
+    /*! Constructs an object of type \p error_condition.
+     *  \post <tt>value() == val</tt>.
+     *  \post <tt>category() == cat</tt>.
+     */
+    inline error_condition(int val, const error_category &cat);
+
+    /*! Constructs an object of type \p error_condition.
+     *  \post <tt>*this == make_error_condition(e)</tt>.
+     *  \note This constructor shall not participate in overload resolution unless
+     *        <tt>is_error_condition_enum<ErrorConditionEnum>::value</tt> is <tt>true</tt>.
+     */
+    template<typename ErrorConditionEnum>
+      error_condition(ErrorConditionEnum e
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+        , typename thrust::detail::enable_if<is_error_condition_enum<ErrorConditionEnum>::value>::type * = 0
+#endif // THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+                     );
+
+    // [19.5.3.3] modifiers
+
+    /*! Assigns to this \p error_code object from an error value and an \p error_category.
+     *  \param val The new value to return from <tt>value()</tt>.
+     *  \param cat The new \p error_category to return from <tt>category()</tt>.
+     *  \post <tt>value() == val</tt>.
+     *  \post <tt>category() == cat</tt>.
+     */
+    inline void assign(int val, const error_category &cat);
+
+    /*! Assigns to this \p error_code object from an error condition enumeration.
+     *  \return *this
+     *  \post <tt>*this == make_error_condition(e)</tt>.
+     *  \note This operator shall not participate in overload resolution unless
+     *        <tt>is_error_condition_enum<ErrorConditionEnum>::value</tt> is <tt>true</tt>.
+     */
+    template<typename ErrorConditionEnum>
+// XXX WAR msvc's problem with enable_if
+#if THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+      typename thrust::detail::enable_if<is_error_condition_enum<ErrorConditionEnum>::value, error_condition>::type &
+#else
+      error_condition &
+#endif // THRUST_HOST_COMPILER != THRUST_HOST_COMPILER_MSVC
+        operator=(ErrorConditionEnum e);
+
+    /*! Clears this \p error_code object.
+     *  \post <tt>value == 0</tt>
+     *  \post <tt>category() == generic_category()</tt>.
+     */
+    inline void clear(void);
+
+    // [19.5.3.4] observers
+
+    /*! \return The value encoded by this \p error_condition.
+     */
+    inline int value(void) const;
+
+    /*! \return A <tt>const</tt> reference to the \p error_category encoded by this \p error_condition.
+     */
+    inline const error_category &category(void) const;
+
+    /*! \return <tt>category().message(value())</tt>.
+     */
+    inline std::string message(void) const;
+
+    // XXX replace below with this upon c++0x
+    //explicit operator bool (void) const;
+    
+    /*! \return <tt>value() != 0</tt>.
+     */
+    inline operator bool (void) const;
+
+    /*! \cond
+     */
+
+  private:
+    int m_val;
+    const error_category *m_cat;
+
+    /*! \endcond
+     */
+}; // end error_condition
+
+
+
+// [19.5.3.5] Class error_condition non-member functions
+
+// XXX replace errc::errc_t with errc upon c++0x
+/*! \return <tt>error_condition(static_cast<int>(e), generic_category())</tt>.
+ */
+inline error_condition make_error_condition(errc::errc_t e);
+
+
+/*! \return <tt>lhs.category() < rhs.category() || lhs.category() == rhs.category() && lhs.value() < rhs.value()</tt>.
+ */
+inline bool operator<(const error_condition &lhs, const error_condition &rhs);
+
+
+// [19.5.4] Comparison operators
+
+
+/*! \return <tt>lhs.category() == rhs.category() && lhs.value() == rhs.value()</tt>.
+ */
+inline bool operator==(const error_code &lhs, const error_code &rhs);
+
+
+/*! \return <tt>lhs.category().equivalent(lhs.value(), rhs) || rhs.category().equivalent(lhs,rhs.value())</tt>.
+ */
+inline bool operator==(const error_code &lhs, const error_condition &rhs);
+
+
+/*! \return <tt>rhs.category().equivalent(lhs.value(), lhs) || lhs.category().equivalent(rhs, lhs.value())</tt>.
+ */
+inline bool operator==(const error_condition &lhs, const error_code &rhs);
+
+
+/*! \return <tt>lhs.category() == rhs.category() && lhs.value() == rhs.value()</tt>
+ */
+inline bool operator==(const error_condition &lhs, const error_condition &rhs);
+
+
+/*! \return <tt>!(lhs == rhs)</tt>
+ */
+inline bool operator!=(const error_code &lhs, const error_code &rhs);
+
+
+/*! \return <tt>!(lhs == rhs)</tt>
+ */
+inline bool operator!=(const error_code &lhs, const error_condition &rhs);
+
+
+/*! \return <tt>!(lhs == rhs)</tt>
+ */
+inline bool operator!=(const error_condition &lhs, const error_code &rhs);
+
+
+/*! \return <tt>!(lhs == rhs)</tt>
+ */
+inline bool operator!=(const error_condition &lhs, const error_condition &rhs);
+
+/*! \} // end system_diagnostics
+ */
+
+
+} // end system
+
+
+// import names into thrust::
+using system::error_category;
+using system::error_code;
+using system::error_condition;
+using system::is_error_code_enum;
+using system::is_error_condition_enum;
+using system::make_error_code;
+using system::make_error_condition;
+
+// XXX replace with using system::errc upon c++0x
+namespace errc = system::errc;
+
+using system::generic_category;
+using system::system_category;
+
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/error_category.inl>
+#include <thrust/system/detail/error_code.inl>
+#include <thrust/system/detail/error_condition.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system/system_error.h b/thirdparty/manifold/thirdparty/thrust/thrust/system/system_error.h
new file mode 100644
index 000000000000..fb31a2da805e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system/system_error.h
@@ -0,0 +1,178 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file system/system_error.h
+ *  \brief An exception object used to report error conditions that have an
+ *         associated error code
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <stdexcept>
+#include <string>
+
+#include <thrust/system/error_code.h>
+
+THRUST_NAMESPACE_BEGIN
+
+namespace system
+{
+
+// [19.5.5] Class system_error
+
+// [19.5.5.1] Class system_error overview
+
+/*! \addtogroup system_diagnostics System Diagnostics
+ *  \ingroup system
+ *  \{
+ */
+
+/*! \brief The class \p system_error describes an exception object used to report error
+ *  conditions that have an associated \p error_code. Such error conditions typically
+ *  originate from the operating system or other low-level application program interfaces.
+ *
+ *  Thrust uses \p system_error to report the error codes returned from device backends
+ *  such as the CUDA runtime.
+ *
+ *  The following code listing demonstrates how to catch a \p system_error to recover
+ *  from an error.
+ *
+ *  \code
+ *
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/system.h>
+ *  #include <thrust/sort.h>
+ *
+ *  void terminate_gracefully(void)
+ *  {
+ *    // application-specific termination code here
+ *    ...
+ *  }
+ *
+ *  int main(void)
+ *  {
+ *    try
+ *    {
+ *      thrust::device_vector<float> vec;
+ *      thrust::sort(vec.begin(), vec.end());
+ *    }
+ *    catch(thrust::system_error e)
+ *    {
+ *      std::cerr << "Error inside sort: " << e.what() << std::endl;
+ *      terminate_gracefully();
+ *    }
+ *
+ *    return 0;
+ *  }
+ *
+ *  \endcode
+ *
+ *  \note If an error represents an out-of-memory condition, implementations are encouraged
+ *  to throw an exception object of type \p std::bad_alloc rather than \p system_error.
+ */
+class system_error
+  : public std::runtime_error
+{
+  public:
+    // [19.5.5.2] Class system_error members
+    
+    /*! Constructs an object of class \p system_error.
+     *  \param ec The value returned by \p code().
+     *  \param what_arg A string to include in the result returned by \p what().
+     *  \post <tt>code() == ec</tt>.
+     *  \post <tt>std::string(what()).find(what_arg) != string::npos</tt>.
+     */
+    inline system_error(error_code ec, const std::string &what_arg);
+
+    /*! Constructs an object of class \p system_error.
+     *  \param ec The value returned by \p code().
+     *  \param what_arg A string to include in the result returned by \p what().
+     *  \post <tt>code() == ec</tt>.
+     *  \post <tt>std::string(what()).find(what_arg) != string::npos</tt>.
+     */
+    inline system_error(error_code ec, const char *what_arg);
+
+    /*! Constructs an object of class \p system_error.
+     *  \param ec The value returned by \p code().
+     *  \post <tt>code() == ec</tt>.
+     */
+    inline system_error(error_code ec);
+
+    /*! Constructs an object of class \p system_error.
+     *  \param ev The error value used to create an \p error_code.
+     *  \param ecat The \p error_category used to create an \p error_code.
+     *  \param what_arg A string to include in the result returned by \p what().
+     *  \post <tt>code() == error_code(ev, ecat)</tt>.
+     *  \post <tt>std::string(what()).find(what_arg) != string::npos</tt>.
+     */
+    inline system_error(int ev, const error_category &ecat, const std::string &what_arg);
+
+    /*! Constructs an object of class \p system_error.
+     *  \param ev The error value used to create an \p error_code.
+     *  \param ecat The \p error_category used to create an \p error_code.
+     *  \param what_arg A string to include in the result returned by \p what().
+     *  \post <tt>code() == error_code(ev, ecat)</tt>.
+     *  \post <tt>std::string(what()).find(what_arg) != string::npos</tt>.
+     */
+    inline system_error(int ev, const error_category &ecat, const char *what_arg);
+
+    /*! Constructs an object of class \p system_error.
+     *  \param ev The error value used to create an \p error_code.
+     *  \param ecat The \p error_category used to create an \p error_code.
+     *  \post <tt>code() == error_code(ev, ecat)</tt>.
+     */
+    inline system_error(int ev, const error_category &ecat);
+
+    /*! Destructor does not throw.
+     */
+    inline virtual ~system_error(void) noexcept {};
+    
+    /*! Returns an object encoding the error.
+     *  \return <tt>ec</tt> or <tt>error_code(ev, ecat)</tt>, from the
+     *          constructor, as appropriate.
+     */
+    inline const error_code &code(void) const noexcept;
+
+    /*! Returns a human-readable string indicating the nature of the error.
+     *  \return a string incorporating <tt>code().message()</tt> and the
+     *          arguments supplied in the constructor.
+     */
+    inline const char *what(void) const noexcept;
+
+    /*! \cond
+     */
+  private:
+    error_code          m_error_code;
+    mutable std::string m_what;
+
+    /*! \endcond
+     */
+}; // end system_error
+
+} // end system
+
+/*! \} // end system_diagnostics
+ */
+
+// import names into thrust::
+using system::system_error;
+
+THRUST_NAMESPACE_END
+
+#include <thrust/system/detail/system_error.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/system_error.h b/thirdparty/manifold/thirdparty/thrust/thrust/system_error.h
new file mode 100644
index 000000000000..6bf240e51234
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/system_error.h
@@ -0,0 +1,49 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file thrust/system_error.h
+ *  \brief System diagnostics
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup system
+ *  \{
+ */
+
+/*! \namespace thrust::system
+ *  \brief \p thrust::system is the namespace which contains specific Thrust
+ *         backend systems. It also contains functionality for reporting error
+ *         conditions originating from the operating system or other low-level
+ *         application program interfaces such as the CUDA runtime. They are
+ *         provided in a separate namespace for import convenience but are
+ *         also aliased in the top-level \p thrust namespace for easy access.
+ */
+namespace system
+{
+} // end system
+
+/*! \} // end system
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/system/error_code.h>
+#include <thrust/system/system_error.h>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/tabulate.h b/thirdparty/manifold/thirdparty/thrust/thrust/tabulate.h
new file mode 100644
index 000000000000..7cb7945508d4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/tabulate.h
@@ -0,0 +1,125 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file tabulate.h
+ *  \brief Fills a range with the tabulation of a function
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup transformations
+ *  \{
+ */
+
+
+/*! \p tabulate fills the range <tt>[first, last)</tt> with the value of a function applied to each
+ *     element's index.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, \p tabulate performs the assignment
+ *  <tt>*i = unary_op(i - first)</tt>.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the range.
+ *  \param last The end of the range.
+ *  \param unary_op The unary operation to apply.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *  \tparam UnaryOperation is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                         and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p tabulate to generate the first \c n non-positive integers
+ *  using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/tabulate.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::tabulate(thrust::host, A, A + 10, thrust::negate<int>());
+ *  // A is now {0, -1, -2, -3, -4, -5, -6, -7, -8, -9}
+ *  \endcode
+ *
+ *  \see thrust::fill
+ *  \see thrust::generate
+ *  \see thrust::sequence
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename UnaryOperation>
+__host__ __device__
+  void tabulate(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator first,
+                ForwardIterator last,
+                UnaryOperation unary_op);
+
+
+/*! \p tabulate fills the range <tt>[first, last)</tt> with the value of a function applied to each
+ *     element's index.
+ *
+ *  For each iterator \c i in the range <tt>[first, last)</tt>, \p tabulate performs the assignment
+ *  <tt>*i = unary_op(i - first)</tt>.
+ *
+ *  \param first The beginning of the range.
+ *  \param last The end of the range.
+ *  \param unary_op The unary operation to apply.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and if \c x and \c y are objects of \c ForwardIterator's \c value_type, then <tt>x + y</tt> is defined,
+ *          and if \c T is \p ForwardIterator's \c value_type, then <tt>T(0)</tt> is defined.
+ *  \tparam UnaryOperation is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                         and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  The following code snippet demonstrates how to use \p tabulate to generate the first \c n non-positive integers:
+ *
+ *  \code
+ *  #include <thrust/tabulate.h>
+ *  #include <thrust/functional.h>
+ *  ...
+ *  const int N = 10;
+ *  int A[N];
+ *  thrust::tabulate(A, A + 10, thrust::negate<int>());
+ *  // A is now {0, -1, -2, -3, -4, -5, -6, -7, -8, -9}
+ *  \endcode
+ *
+ *  \see thrust::fill
+ *  \see thrust::generate
+ *  \see thrust::sequence
+ */
+template<typename ForwardIterator, typename UnaryOperation>
+  void tabulate(ForwardIterator first,
+                ForwardIterator last,
+                UnaryOperation unary_op);
+
+
+/*! \} // end transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/tabulate.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/transform.h b/thirdparty/manifold/thirdparty/thrust/thrust/transform.h
new file mode 100644
index 000000000000..2d064c13b3e4
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/transform.h
@@ -0,0 +1,721 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file thrust/transform.h
+ *  \brief Transforms input ranges using a function object
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+/*! \addtogroup transformations
+ *  \ingroup algorithms
+ *  \{
+ */
+
+
+/*! This version of \p transform applies a unary function to each element
+ *  of an input sequence and stores the result in the corresponding 
+ *  position in an output sequence.  Specifically, for each iterator 
+ *  <tt>i</tt> in the range [\p first, \p last) the operation 
+ *  <tt>op(*i)</tt> is performed and the result is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  [\p result, \p result + (\p last - \p first) ).  The input and
+ *  output sequences may coincide, resulting in an in-place transformation.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The transformation operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to \c UnaryFunction's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                              and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform to negate a range in-place
+ *  using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ * 
+ *  thrust::negate<int> op;
+ *
+ *  thrust::transform(thrust::host, data, data + 10, data, op); // in-place transformation
+ *
+ *  // data is now {5, 0, -2, 3, -2, -4, 0, 1, -2, -8};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/transform
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction>
+__host__ __device__
+  OutputIterator transform(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first, InputIterator last,
+                           OutputIterator result,
+                           UnaryFunction op);
+
+	
+/*! This version of \p transform applies a unary function to each element
+ *  of an input sequence and stores the result in the corresponding 
+ *  position in an output sequence.  Specifically, for each iterator 
+ *  <tt>i</tt> in the range [\p first, \p last) the operation 
+ *  <tt>op(*i)</tt> is performed and the result is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  [\p result, \p result + (\p last - \p first) ).  The input and
+ *  output sequences may coincide, resulting in an in-place transformation.
+ *    
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator's \c value_type is convertible to \c UnaryFunction's \c argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                              and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int data[10] = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ * 
+ *  thrust::negate<int> op;
+ *
+ *  thrust::transform(data, data + 10, data, op); // in-place transformation
+ *
+ *  // data is now {5, 0, -2, 3, -2, -4, 0, 1, -2, -8};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/transform
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction>
+  OutputIterator transform(InputIterator first, InputIterator last,
+                           OutputIterator result,
+                           UnaryFunction op);
+
+
+/*! This version of \p transform applies a binary function to each pair
+ *  of elements from two input sequences and stores the result in the
+ *  corresponding position in an output sequence.  Specifically, for
+ *  each iterator <tt>i</tt> in the range [\p first1, \p last1) and 
+ *  <tt>j = first + (i - first1)</tt> in the range [\p first2, \p last2)
+ *  the operation <tt>op(*i,*j)</tt> is performed and the result is 
+ *  assigned to <tt>*o</tt>,  where <tt>o</tt> is the corresponding
+ *  output iterator in the range [\p result, \p result + (\p last - \p first) ).
+ *  The input and output sequences may coincide, resulting in an 
+ *  in-place transformation.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input sequence.
+ *  \param last1 The end of the first input sequence.
+ *  \param first2 The beginning of the second input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator1's \c value_type is convertible to \c BinaryFunction's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator2's \c value_type is convertible to \c BinaryFunction's \c second_argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c BinaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first1 may equal \p result, but the range <tt>[first1, last1)</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *  \pre \p first2 may equal \p result, but the range <tt>[first2, first2 + (last1 - first1))</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform to compute the sum of two
+ *  ranges using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int input1[6] = {-5,  0,  2,  3,  2,  4};
+ *  int input2[6] = { 3,  6, -2,  1,  2,  3};
+ *  int output[6];
+ * 
+ *  thrust::plus<int> op;
+ *
+ *  thrust::transform(thrust::host, input1, input1 + 6, input2, output, op);
+ *
+ *  // output is now {-2,  6,  0,  4,  4,  7};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/transform
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryFunction>
+__host__ __device__
+  OutputIterator transform(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator1 first1, InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputIterator result,
+                           BinaryFunction op);
+
+
+/*! This version of \p transform applies a binary function to each pair
+ *  of elements from two input sequences and stores the result in the
+ *  corresponding position in an output sequence.  Specifically, for
+ *  each iterator <tt>i</tt> in the range [\p first1, \p last1) and 
+ *  <tt>j = first + (i - first1)</tt> in the range [\p first2, \p last2)
+ *  the operation <tt>op(*i,*j)</tt> is performed and the result is 
+ *  assigned to <tt>*o</tt>,  where <tt>o</tt> is the corresponding
+ *  output iterator in the range [\p result, \p result + (\p last - \p first) ).
+ *  The input and output sequences may coincide, resulting in an 
+ *  in-place transformation.
+ *    
+ *  \param first1 The beginning of the first input sequence.
+ *  \param last1 The end of the first input sequence.
+ *  \param first2 The beginning of the second input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator1's \c value_type is convertible to \c BinaryFunction's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                        and \c InputIterator2's \c value_type is convertible to \c BinaryFunction's \c second_argument_type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c BinaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first1 may equal \p result, but the range <tt>[first1, last1)</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *  \pre \p first2 may equal \p result, but the range <tt>[first2, first2 + (last1 - first1))</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int input1[6] = {-5,  0,  2,  3,  2,  4};
+ *  int input2[6] = { 3,  6, -2,  1,  2,  3};
+ *  int output[6];
+ * 
+ *  thrust::plus<int> op;
+ *
+ *  thrust::transform(input1, input1 + 6, input2, output, op);
+ *
+ *  // output is now {-2,  6,  0,  4,  4,  7};
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/transform
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator,
+         typename BinaryFunction>
+  OutputIterator transform(InputIterator1 first1, InputIterator1 last1,
+                           InputIterator2 first2,
+                           OutputIterator result,
+                           BinaryFunction op);
+
+
+/*! This version of \p transform_if conditionally applies a unary function
+ *  to each element of an input sequence and stores the result in the corresponding 
+ *  position in an output sequence if the corresponding position in the input sequence
+ *  satifies a predicate. Otherwise, the corresponding position in the
+ *  output sequence is not modified.
+ *
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first, last)</tt> the
+ *  predicate <tt>pred(*i)</tt> is evaluated. If this predicate
+ *  evaluates to \c true, the result of <tt>op(*i)</tt> is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  <tt>[result, result + (last - first) )</tt>. Otherwise, <tt>op(*i)</tt> is
+ *  not evaluated and no assignment occurs. The input and output sequences may coincide,
+ *  resulting in an in-place transformation.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \param pred The predicate operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *                        and \c InputIterator's \c value_type is convertible to \c Predicate's \c argument_type,
+ *                        and \c InputIterator's \c value_type is convertible to \c UnaryFunction's \c argument_type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                        and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_if to negate the odd-valued
+ *  elements of a range using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ *
+ *  struct is_odd
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x % 2;
+ *    }
+ *  };
+ * 
+ *  thrust::negate<int> op;
+ *  thrust::identity<int> identity;
+ *
+ *  // negate odd elements
+ *  thrust::transform_if(thrust::host, data, data + 10, data, op, is_odd()); // in-place transformation
+ *
+ *  // data is now {5, 0, 2, 3, 2, 4, 0, 1, 2, 8};
+ *  \endcode
+ *
+ *  \see thrust::transform
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               InputIterator first, InputIterator last,
+                               ForwardIterator result,
+                               UnaryFunction op,
+                               Predicate pred);
+
+
+/*! This version of \p transform_if conditionally applies a unary function
+ *  to each element of an input sequence and stores the result in the corresponding 
+ *  position in an output sequence if the corresponding position in the input sequence
+ *  satifies a predicate. Otherwise, the corresponding position in the
+ *  output sequence is not modified.
+ *
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first, last)</tt> the
+ *  predicate <tt>pred(*i)</tt> is evaluated. If this predicate
+ *  evaluates to \c true, the result of <tt>op(*i)</tt> is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  <tt>[result, result + (last - first) )</tt>. Otherwise, <tt>op(*i)</tt> is
+ *  not evaluated and no assignment occurs. The input and output sequences may coincide,
+ *  resulting in an in-place transformation.
+ *    
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \param pred The predicate operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *                        and \c InputIterator's \c value_type is convertible to \c Predicate's \c argument_type,
+ *                        and \c InputIterator's \c value_type is convertible to \c UnaryFunction's \c argument_type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                        and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_if:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ *
+ *  struct is_odd
+ *  {
+ *    __host__ __device__
+ *    bool operator()(int x)
+ *    {
+ *      return x % 2;
+ *    }
+ *  };
+ * 
+ *  thrust::negate<int> op;
+ *  thrust::identity<int> identity;
+ *
+ *  // negate odd elements
+ *  thrust::transform_if(data, data + 10, data, op, is_odd()); // in-place transformation
+ *
+ *  // data is now {5, 0, 2, 3, 2, 4, 0, 1, 2, 8};
+ *  \endcode
+ *
+ *  \see thrust::transform
+ */
+template<typename InputIterator,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+  ForwardIterator transform_if(InputIterator first, InputIterator last,
+                               ForwardIterator result,
+                               UnaryFunction op,
+                               Predicate pred);
+
+
+/*! This version of \p transform_if conditionally applies a unary function
+ *  to each element of an input sequence and stores the result in the corresponding 
+ *  position in an output sequence if the corresponding position in a stencil sequence
+ *  satisfies a predicate. Otherwise, the corresponding position in the
+ *  output sequence is not modified.
+ *
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first, last)</tt> the
+ *  predicate <tt>pred(*s)</tt> is evaluated, where <tt>s</tt> is the corresponding input
+ *  iterator in the range <tt>[stencil, stencil + (last - first) )</tt>. If this predicate
+ *  evaluates to \c true, the result of <tt>op(*i)</tt> is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  <tt>[result, result + (last - first) )</tt>. Otherwise, <tt>op(*i)</tt> is
+ *  not evaluated and no assignment occurs. The input and output sequences may coincide,
+ *  resulting in an in-place transformation.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \param pred The predicate operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator1's \c value_type is convertible to \c UnaryFunction's \c argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c Predicate's \c argument_type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                        and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *  \pre \p stencil may equal \p result, but the range <tt>[stencil, stencil + (last - first))</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_if using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ *  int stencil[10] = { 1, 0, 1,  0, 1, 0, 1,  0, 1, 0};
+ * 
+ *  thrust::negate<int> op;
+ *  thrust::identity<int> identity;
+ *
+ *  thrust::transform_if(thrust::host, data, data + 10, stencil, data, op, identity); // in-place transformation
+ *
+ *  // data is now {5, 0, -2, -3, -2,  4, 0, -1, -2,  8};
+ *  \endcode
+ *
+ *  \see thrust::transform
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               InputIterator1 first, InputIterator1 last,
+                               InputIterator2 stencil,
+                               ForwardIterator result,
+                               UnaryFunction op,
+                               Predicate pred);
+
+
+/*! This version of \p transform_if conditionally applies a unary function
+ *  to each element of an input sequence and stores the result in the corresponding 
+ *  position in an output sequence if the corresponding position in a stencil sequence
+ *  satisfies a predicate. Otherwise, the corresponding position in the
+ *  output sequence is not modified.
+ *
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first, last)</tt> the
+ *  predicate <tt>pred(*s)</tt> is evaluated, where <tt>s</tt> is the corresponding input
+ *  iterator in the range <tt>[stencil, stencil + (last - first) )</tt>. If this predicate
+ *  evaluates to \c true, the result of <tt>op(*i)</tt> is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  <tt>[result, result + (last - first) )</tt>. Otherwise, <tt>op(*i)</tt> is
+ *  not evaluated and no assignment occurs. The input and output sequences may coincide,
+ *  resulting in an in-place transformation.
+ *    
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param op The tranformation operation.
+ *  \param pred The predicate operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator1's \c value_type is convertible to \c UnaryFunction's \c argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c Predicate's \c argument_type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                        and \c UnaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *  \pre \p stencil may equal \p result, but the range <tt>[stencil, stencil + (last - first))</tt> shall not overlap the range <tt>[result, result + (last - first))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_if:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int data[10]    = {-5, 0, 2, -3, 2, 4, 0, -1, 2, 8};
+ *  int stencil[10] = { 1, 0, 1,  0, 1, 0, 1,  0, 1, 0};
+ * 
+ *  thrust::negate<int> op;
+ *  thrust::identity<int> identity;
+ *
+ *  thrust::transform_if(data, data + 10, stencil, data, op, identity); // in-place transformation
+ *
+ *  // data is now {5, 0, -2, -3, -2,  4, 0, -1, -2,  8};
+ *  \endcode
+ *
+ *  \see thrust::transform
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename ForwardIterator,
+         typename UnaryFunction,
+         typename Predicate>
+  ForwardIterator transform_if(InputIterator1 first, InputIterator1 last,
+                               InputIterator2 stencil,
+                               ForwardIterator result,
+                               UnaryFunction op,
+                               Predicate pred);
+
+
+/*! This version of \p transform_if conditionally applies a binary function
+ *  to each pair of elements from two input sequences and stores the result in the corresponding 
+ *  position in an output sequence if the corresponding position in a stencil sequence
+ *  satifies a predicate. Otherwise, the corresponding position in the
+ *  output sequence is not modified.
+ *
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first1, last1)</tt> and 
+ *  <tt>j = first2 + (i - first1)</tt> in the range <tt>[first2, first2 + (last1 - first1) )</tt>,
+ *  the predicate <tt>pred(*s)</tt> is evaluated, where <tt>s</tt> is the corresponding input
+ *  iterator in the range <tt>[stencil, stencil + (last1 - first1) )</tt>. If this predicate
+ *  evaluates to \c true, the result of <tt>binary_op(*i,*j)</tt> is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  <tt>[result, result + (last1 - first1) )</tt>. Otherwise, <tt>binary_op(*i,*j)</tt> is
+ *  not evaluated and no assignment occurs. The input and output sequences may coincide,
+ *  resulting in an in-place transformation.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *    
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first1 The beginning of the first input sequence.
+ *  \param last1 The end of the first input sequence.
+ *  \param first2 The beginning of the second input sequence.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param binary_op The transformation operation.
+ *  \param pred The predicate operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator1's \c value_type is convertible to \c BinaryFunction's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c BinaryFunction's \c second_argument_type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                         and \c BinaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre \p first1 may equal \p result, but the range <tt>[first1, last1)</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *  \pre \p first2 may equal \p result, but the range <tt>[first2, first2 + (last1 - first1))</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *  \pre \p stencil may equal \p result, but the range <tt>[stencil, stencil + (last1 - first1))</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_if using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int input1[6]  = {-5,  0,  2,  3,  2,  4};
+ *  int input2[6]  = { 3,  6, -2,  1,  2,  3};
+ *  int stencil[8] = { 1,  0,  1,  0,  1,  0};
+ *  int output[6];
+ * 
+ *  thrust::plus<int> op;
+ *  thrust::identity<int> identity;
+ *
+ *  thrust::transform_if(thrust::host, input1, input1 + 6, input2, stencil, output, op, identity);
+ *
+ *  // output is now {-2,  0,  0,  3,  4,  4};
+ *  \endcode
+ *
+ *  \see thrust::transform
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename ForwardIterator,
+         typename BinaryFunction,
+         typename Predicate>
+__host__ __device__
+  ForwardIterator transform_if(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                               InputIterator1 first1, InputIterator1 last1,
+                               InputIterator2 first2,
+                               InputIterator3 stencil,
+                               ForwardIterator result,
+                               BinaryFunction binary_op,
+                               Predicate pred);
+
+
+/*! This version of \p transform_if conditionally applies a binary function
+ *  to each pair of elements from two input sequences and stores the result in the corresponding 
+ *  position in an output sequence if the corresponding position in a stencil sequence
+ *  satifies a predicate. Otherwise, the corresponding position in the
+ *  output sequence is not modified.
+ *
+ *  Specifically, for each iterator <tt>i</tt> in the range <tt>[first1, last1)</tt> and 
+ *  <tt>j = first2 + (i - first1)</tt> in the range <tt>[first2, first2 + (last1 - first1) )</tt>,
+ *  the predicate <tt>pred(*s)</tt> is evaluated, where <tt>s</tt> is the corresponding input
+ *  iterator in the range <tt>[stencil, stencil + (last1 - first1) )</tt>. If this predicate
+ *  evaluates to \c true, the result of <tt>binary_op(*i,*j)</tt> is assigned to <tt>*o</tt>,
+ *  where <tt>o</tt> is the corresponding output iterator in the range
+ *  <tt>[result, result + (last1 - first1) )</tt>. Otherwise, <tt>binary_op(*i,*j)</tt> is
+ *  not evaluated and no assignment occurs. The input and output sequences may coincide,
+ *  resulting in an in-place transformation.
+ *    
+ *  \param first1 The beginning of the first input sequence.
+ *  \param last1 The end of the first input sequence.
+ *  \param first2 The beginning of the second input sequence.
+ *  \param stencil The beginning of the stencil sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param binary_op The transformation operation.
+ *  \param pred The predicate operation.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator1's \c value_type is convertible to \c BinaryFunction's \c first_argument_type.
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                         and \c InputIterator2's \c value_type is convertible to \c BinaryFunction's \c second_argument_type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                         and \c BinaryFunction's \c result_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam Predicate is a model of <a href="https://en.cppreference.com/w/cpp/concepts/predicate">Predicate</a>.
+ *
+ *  \pre \p first1 may equal \p result, but the range <tt>[first1, last1)</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *  \pre \p first2 may equal \p result, but the range <tt>[first2, first2 + (last1 - first1))</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *  \pre \p stencil may equal \p result, but the range <tt>[stencil, stencil + (last1 - first1))</tt> shall not overlap the range <tt>[result, result + (last1 - first1))</tt> otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_if:
+ *
+ *  \code
+ *  #include <thrust/transform.h>
+ *  #include <thrust/functional.h>
+ *  
+ *  int input1[6]  = {-5,  0,  2,  3,  2,  4};
+ *  int input2[6]  = { 3,  6, -2,  1,  2,  3};
+ *  int stencil[8] = { 1,  0,  1,  0,  1,  0};
+ *  int output[6];
+ * 
+ *  thrust::plus<int> op;
+ *  thrust::identity<int> identity;
+ *
+ *  thrust::transform_if(input1, input1 + 6, input2, stencil, output, op, identity);
+ *
+ *  // output is now {-2,  0,  0,  3,  4,  4};
+ *  \endcode
+ *
+ *  \see thrust::transform
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename InputIterator3,
+         typename ForwardIterator,
+         typename BinaryFunction,
+         typename Predicate>
+  ForwardIterator transform_if(InputIterator1 first1, InputIterator1 last1,
+                               InputIterator2 first2,
+                               InputIterator3 stencil,
+                               ForwardIterator result,
+                               BinaryFunction binary_op,
+                               Predicate pred);
+
+
+/*! \} // end transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/transform.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/transform_reduce.h b/thirdparty/manifold/thirdparty/thrust/thrust/transform_reduce.h
new file mode 100644
index 000000000000..11d6b84c3eb1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/transform_reduce.h
@@ -0,0 +1,194 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file transform_reduce.h
+ *  \brief Fused transform / reduction
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup reductions
+ *  \{
+ *  \addtogroup transformed_reductions Transformed Reductions
+ *  \ingroup reductions
+ *  \{
+ */
+
+
+/*! \p transform_reduce fuses the \p transform and \p reduce operations.
+ *  \p transform_reduce is equivalent to performing a transformation defined by
+ *  \p unary_op into a temporary sequence and then performing \p reduce on the
+ *  transformed sequence. In most cases, fusing these two operations together is
+ *  more efficient, since fewer memory reads and writes are required.
+ *
+ *  \p transform_reduce performs a reduction on the transformation of the
+ *  sequence <tt>[first, last)</tt> according to \p unary_op. Specifically,
+ *  \p unary_op is applied to each element of the sequence and then the result
+ *  is reduced to a single value with \p binary_op using the initial value 
+ *  \p init.  Note that the transformation \p unary_op is not applied to 
+ *  the initial value \p init.  The order of reduction is not specified, 
+ *  so \p binary_op must be both commutative and associative. 
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param unary_op The function to apply to each element of the input sequence.
+ *  \param init The result is initialized to this value.
+ *  \param binary_op The reduction operation.
+ *  \return The result of the transformed reduction.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p UnaryFunction's \c argument_type.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>,
+ *          and \p UnaryFunction's \c result_type is convertible to \c OutputType.
+ *  \tparam OutputType is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and is convertible to \p BinaryFunction's \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction's \c result_type is convertible to \p OutputType.
+ *
+ *  The following code snippet demonstrates how to use \p transform_reduce
+ *  to compute the maximum value of the absolute value of the elements
+ *  of a range using the \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform_reduce.h>
+ *  #include <thrust/functional.h>
+ *  #include <thrust/execution_policy.h>
+ *
+ *  template<typename T>
+ *  struct absolute_value : public unary_function<T,T>
+ *  {
+ *    __host__ __device__ T operator()(const T &x) const
+ *    {
+ *      return x < T(0) ? -x : x;
+ *    }
+ *  };
+ *
+ *  ...
+ *
+ *  int data[6] = {-1, 0, -2, -2, 1, -3};
+ *  int result = thrust::transform_reduce(thrust::host,
+ *                                        data, data + 6,
+ *                                        absolute_value<int>(),
+ *                                        0,
+ *                                        thrust::maximum<int>());
+ *  // result == 3
+ *  \endcode
+ *
+ *  \see \c transform
+ *  \see \c reduce
+ */
+template<typename DerivedPolicy,
+         typename InputIterator, 
+         typename UnaryFunction, 
+         typename OutputType,
+         typename BinaryFunction>
+__host__ __device__
+  OutputType transform_reduce(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                              InputIterator first,
+                              InputIterator last,
+                              UnaryFunction unary_op,
+                              OutputType init,
+                              BinaryFunction binary_op);
+
+
+/*! \p transform_reduce fuses the \p transform and \p reduce operations.
+ *  \p transform_reduce is equivalent to performing a transformation defined by
+ *  \p unary_op into a temporary sequence and then performing \p reduce on the
+ *  transformed sequence. In most cases, fusing these two operations together is
+ *  more efficient, since fewer memory reads and writes are required.
+ *
+ *  \p transform_reduce performs a reduction on the transformation of the
+ *  sequence <tt>[first, last)</tt> according to \p unary_op. Specifically,
+ *  \p unary_op is applied to each element of the sequence and then the result
+ *  is reduced to a single value with \p binary_op using the initial value 
+ *  \p init.  Note that the transformation \p unary_op is not applied to 
+ *  the initial value \p init.  The order of reduction is not specified, 
+ *  so \p binary_op must be both commutative and associative. 
+ *
+ *  \param first The beginning of the sequence.
+ *  \param last The end of the sequence.
+ *  \param unary_op The function to apply to each element of the input sequence.
+ *  \param init The result is initialized to this value.
+ *  \param binary_op The reduction operation.
+ *  \return The result of the transformed reduction.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is convertible to \p UnaryFunction's \c argument_type.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>,
+ *          and \p UnaryFunction's \c result_type is convertible to \c OutputType.
+ *  \tparam OutputType is a model of <a href="https://en.cppreference.com/w/cpp/named_req/CopyAssignable">Assignable</a>,
+ *          and is convertible to \p BinaryFunction's \c first_argument_type and \c second_argument_type.
+ *  \tparam BinaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>,
+ *          and \p BinaryFunction's \c result_type is convertible to \p OutputType.
+ *
+ *  The following code snippet demonstrates how to use \p transform_reduce
+ *  to compute the maximum value of the absolute value of the elements
+ *  of a range.
+ *
+ *  \code
+ *  #include <thrust/transform_reduce.h>
+ *  #include <thrust/functional.h>
+ *
+ *  template<typename T>
+ *  struct absolute_value : public unary_function<T,T>
+ *  {
+ *    __host__ __device__ T operator()(const T &x) const
+ *    {
+ *      return x < T(0) ? -x : x;
+ *    }
+ *  };
+ *
+ *  ...
+ *
+ *  int data[6] = {-1, 0, -2, -2, 1, -3};
+ *  int result = thrust::transform_reduce(data, data + 6,
+ *                                        absolute_value<int>(),
+ *                                        0,
+ *                                        thrust::maximum<int>());
+ *  // result == 3
+ *  \endcode
+ *
+ *  \see \c transform
+ *  \see \c reduce
+ */
+template<typename InputIterator, 
+         typename UnaryFunction, 
+         typename OutputType,
+         typename BinaryFunction>
+  OutputType transform_reduce(InputIterator first,
+                              InputIterator last,
+                              UnaryFunction unary_op,
+                              OutputType init,
+                              BinaryFunction binary_op);
+
+
+/*! \} // end transformed_reductions
+ *  \} // end reductions
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/transform_reduce.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/transform_scan.h b/thirdparty/manifold/thirdparty/thrust/thrust/transform_scan.h
new file mode 100644
index 000000000000..6c0fe8116680
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/transform_scan.h
@@ -0,0 +1,320 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file transform_scan.h
+ *  \brief Fused transform / prefix-sum
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup algorithms
+ */
+
+/*! \addtogroup prefixsums Prefix Sums
+ *  \ingroup algorithms
+ *  \{
+ */
+	
+/*! \addtogroup transformed_prefixsums Transformed Prefix Sums
+ *  \ingroup prefixsums
+ *  \{
+ */
+
+
+/*! \p transform_inclusive_scan fuses the \p transform and \p inclusive_scan
+ *  operations.  \p transform_inclusive_scan is equivalent to performing a
+ *  tranformation defined by \p unary_op into a temporary sequence and then
+ *  performing an \p inclusive_scan on the tranformed sequence.  In most
+ *  cases, fusing these two operations together is more efficient, since
+ *  fewer memory reads and writes are required. In \p transform_inclusive_scan,
+ *  <tt>unary_op(\*first)</tt> is assigned to <tt>\*result</tt> and the result
+ *  of <tt>binary_op(unary_op(\*first), unary_op(\*(first + 1)))</tt> is
+ *  assigned to <tt>\*(result + 1)</tt>, and so on.  The transform scan
+ *  operation is permitted to be in-place.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param unary_op The function used to tranform the input sequence.
+ *  \param binary_op The associatve operator used to 'sum' transformed values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                               and \c InputIterator's \c value_type is convertible to \c unary_op's input type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                               and accepts inputs of \c InputIterator's \c value_type.  \c UnaryFunction's result_type
+ *                               is convertable to \c OutputIterator's \c value_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_inclusive_scan using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform_scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::negate<int> unary_op;
+ *  thrust::plus<int> binary_op;
+ *
+ *  thrust::transform_inclusive_scan(thrust::host, data, data + 6, data, unary_op, binary_op); // in-place scan
+ *
+ *  // data is now {-1, -1, -3, -5, -6, -9}
+ *  \endcode
+ *
+ *  \see \p transform
+ *  \see \p inclusive_scan
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator transform_inclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          AssociativeOperator binary_op);
+
+
+/*! \p transform_inclusive_scan fuses the \p transform and \p inclusive_scan
+ *  operations.  \p transform_inclusive_scan is equivalent to performing a
+ *  tranformation defined by \p unary_op into a temporary sequence and then
+ *  performing an \p inclusive_scan on the tranformed sequence.  In most
+ *  cases, fusing these two operations together is more efficient, since
+ *  fewer memory reads and writes are required. In \p transform_inclusive_scan,
+ *  <tt>unary_op(\*first)</tt> is assigned to <tt>\*result</tt> and the result
+ *  of <tt>binary_op(unary_op(\*first), unary_op(\*(first + 1)))</tt> is
+ *  assigned to <tt>\*(result + 1)</tt>, and so on.  The transform scan
+ *  operation is permitted to be in-place.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param unary_op The function used to tranform the input sequence.
+ *  \param binary_op The associatve operator used to 'sum' transformed values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                               and \c InputIterator's \c value_type is convertible to \c unary_op's input type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                               and accepts inputs of \c InputIterator's \c value_type.  \c UnaryFunction's result_type
+ *                               is convertable to \c OutputIterator's \c value_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_inclusive_scan
+ *
+ *  \code
+ *  #include <thrust/transform_scan.h>
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::negate<int> unary_op;
+ *  thrust::plus<int> binary_op;
+ *
+ *  thrust::transform_inclusive_scan(data, data + 6, data, unary_op, binary_op); // in-place scan
+ *
+ *  // data is now {-1, -1, -3, -5, -6, -9}
+ *  \endcode
+ *
+ *  \see \p transform
+ *  \see \p inclusive_scan
+ *
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename AssociativeOperator>
+  OutputIterator transform_inclusive_scan(InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          AssociativeOperator binary_op);
+
+
+/*! \p transform_exclusive_scan fuses the \p transform and \p exclusive_scan
+ *  operations.  \p transform_exclusive_scan is equivalent to performing a
+ *  tranformation defined by \p unary_op into a temporary sequence and then
+ *  performing an \p exclusive_scan on the tranformed sequence.  In most
+ *  cases, fusing these two operations together is more efficient, since
+ *  fewer memory reads and writes are required. In 
+ *  \p transform_exclusive_scan, \p init is assigned to <tt>\*result</tt> 
+ *  and the result of <tt>binary_op(init, unary_op(\*first))</tt> is assigned
+ *  to <tt>\*(result + 1)</tt>, and so on.  The transform scan operation is 
+ *  permitted to be in-place.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param unary_op The function used to tranform the input sequence.
+ *  \param init The initial value of the \p exclusive_scan
+ *  \param binary_op The associatve operator used to 'sum' transformed values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                               and \c InputIterator's \c value_type is convertible to \c unary_op's input type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                               and accepts inputs of \c InputIterator's \c value_type.  \c UnaryFunction's result_type
+ *                               is convertable to \c OutputIterator's \c value_type.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_exclusive_scan using the
+ *  \p thrust::host execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/transform_scan.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::negate<int> unary_op;
+ *  thrust::plus<int> binary_op;
+ *
+ *  thrust::transform_exclusive_scan(thrust::host, data, data + 6, data, unary_op, 4, binary_op); // in-place scan
+ *
+ *  // data is now {4, 3, 3, 1, -1, -2}
+ *  \endcode
+ *
+ *  \see \p transform
+ *  \see \p exclusive_scan
+ *
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename T,
+         typename AssociativeOperator>
+__host__ __device__
+  OutputIterator transform_exclusive_scan(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                          InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          T init,
+                                          AssociativeOperator binary_op);
+
+
+/*! \p transform_exclusive_scan fuses the \p transform and \p exclusive_scan
+ *  operations.  \p transform_exclusive_scan is equivalent to performing a
+ *  tranformation defined by \p unary_op into a temporary sequence and then
+ *  performing an \p exclusive_scan on the tranformed sequence.  In most
+ *  cases, fusing these two operations together is more efficient, since
+ *  fewer memory reads and writes are required. In 
+ *  \p transform_exclusive_scan, \p init is assigned to <tt>\*result</tt> 
+ *  and the result of <tt>binary_op(init, unary_op(\*first))</tt> is assigned
+ *  to <tt>\*(result + 1)</tt>, and so on.  The transform scan operation is 
+ *  permitted to be in-place.
+ *
+ *  \param first The beginning of the input sequence.
+ *  \param last The end of the input sequence.
+ *  \param result The beginning of the output sequence.
+ *  \param unary_op The function used to tranform the input sequence.
+ *  \param init The initial value of the \p exclusive_scan
+ *  \param binary_op The associatve operator used to 'sum' transformed values.
+ *  \return The end of the output sequence.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>
+ *                               and \c InputIterator's \c value_type is convertible to \c unary_op's input type.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a>.
+ *  \tparam UnaryFunction is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/unary_function">Unary Function</a>
+ *                               and accepts inputs of \c InputIterator's \c value_type.  \c UnaryFunction's result_type
+ *                               is convertable to \c OutputIterator's \c value_type.
+ *  \tparam T is convertible to \c OutputIterator's \c value_type.
+ *  \tparam AssociativeOperator is a model of <a href="https://en.cppreference.com/w/cpp/utility/functional/binary_function">Binary Function</a>
+ *                              and \c AssociativeOperator's \c result_type is
+ *                              convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p transform_exclusive_scan
+ *
+ *  \code
+ *  #include <thrust/transform_scan.h>
+ *  
+ *  int data[6] = {1, 0, 2, 2, 1, 3};
+ *
+ *  thrust::negate<int> unary_op;
+ *  thrust::plus<int> binary_op;
+ *
+ *  thrust::transform_exclusive_scan(data, data + 6, data, unary_op, 4, binary_op); // in-place scan
+ *
+ *  // data is now {4, 3, 3, 1, -1, -2}
+ *  \endcode
+ *
+ *  \see \p transform
+ *  \see \p exclusive_scan
+ *
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename UnaryFunction,
+         typename T,
+         typename AssociativeOperator>
+  OutputIterator transform_exclusive_scan(InputIterator first,
+                                          InputIterator last,
+                                          OutputIterator result,
+                                          UnaryFunction unary_op,
+                                          T init,
+                                          AssociativeOperator binary_op);
+
+
+/*! \} // end transformed_prefixsums
+ */
+
+
+/*! \} // end prefixsums
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/transform_scan.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/tuple.h b/thirdparty/manifold/thirdparty/thrust/thrust/tuple.h
new file mode 100644
index 000000000000..04f3154a3b3b
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/tuple.h
@@ -0,0 +1,575 @@
+/*
+ *  Copyright 2008-2018 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file tuple.h
+ *  \brief A type encapsulating a heterogeneous collection of elements.
+ */
+
+/*
+ * Copyright (C) 1999, 2000 Jaakko Järvi (jaakko.jarvi@cs.utu.fi)
+ *
+ * Distributed under the Boost Software License, Version 1.0.
+ * (See accompanying NOTICE file for the complete license)
+ *
+ * For more information, see http://www.boost.org
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/tuple.inl>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup tuple
+ *  \{
+ */
+
+/*! \cond
+ */
+
+struct null_type;
+
+/*! \endcond
+ */
+
+/*! This metafunction returns the type of a
+ *  \p tuple's <tt>N</tt>th element.
+ *
+ *  \tparam N This parameter selects the element of interest.
+ *  \tparam T A \c tuple type of interest.
+ *
+ *  \see pair
+ *  \see tuple
+ */
+template <size_t N, class T> struct tuple_element;
+
+/*! This metafunction returns the number of elements
+ *  of a \p tuple type of interest.
+ *
+ *  \tparam T A \c tuple type of interest.
+ *
+ *  \see pair
+ *  \see tuple
+ */
+template <class T> struct tuple_size;
+
+
+// get function for non-const cons-lists, returns a reference to the element
+
+/*! The \p get function returns a reference to a \p tuple element of
+ *  interest.
+ *
+ *  \param t A reference to a \p tuple of interest.
+ *  \return A reference to \p t's <tt>N</tt>th element.
+ *
+ *  \tparam N The index of the element of interest.
+ *
+ *  The following code snippet demonstrates how to use \p get to print
+ *  the value of a \p tuple element.
+ *
+ *  \code
+ *  #include <thrust/tuple.h>
+ *  #include <iostream>
+ *  ...
+ *  thrust::tuple<int, const char *> t(13, "thrust");
+ *
+ *  std::cout << "The 1st value of t is " << thrust::get<0>(t) << std::endl;
+ *  \endcode
+ *
+ *  \see pair
+ *  \see tuple
+ */
+template<int N, class HT, class TT>
+__host__ __device__
+inline typename access_traits<
+                  typename tuple_element<N, detail::cons<HT, TT> >::type
+                >::non_const_type
+get(detail::cons<HT, TT>& t);
+
+
+/*! The \p get function returns a \c const reference to a \p tuple element of
+ *  interest.
+ *
+ *  \param t A reference to a \p tuple of interest.
+ *  \return A \c const reference to \p t's <tt>N</tt>th element.
+ *
+ *  \tparam N The index of the element of interest.
+ *
+ *  The following code snippet demonstrates how to use \p get to print
+ *  the value of a \p tuple element.
+ *
+ *  \code
+ *  #include <thrust/tuple.h>
+ *  #include <iostream>
+ *  ...
+ *  thrust::tuple<int, const char *> t(13, "thrust");
+ *
+ *  std::cout << "The 1st value of t is " << thrust::get<0>(t) << std::endl;
+ *  \endcode
+ *
+ *  \see pair
+ *  \see tuple
+ */
+template<int N, class HT, class TT>
+__host__ __device__
+inline typename access_traits<
+                  typename tuple_element<N, detail::cons<HT, TT> >::type
+                >::const_type
+get(const detail::cons<HT, TT>& t);
+
+
+
+/*! \brief \p tuple is a class template that can be instantiated with up to ten
+ *  arguments. Each template argument specifies the type of element in the \p
+ *  tuple. Consequently, tuples are heterogeneous, fixed-size collections of
+ *  values. An instantiation of \p tuple with two arguments is similar to an
+ *  instantiation of \p pair with the same two arguments. Individual elements
+ *  of a \p tuple may be accessed with the \p get function.
+ *
+ *  \tparam TN The type of the <tt>N</tt> \c tuple element. Thrust's \p tuple
+ *          type currently supports up to ten elements.
+ *
+ *  The following code snippet demonstrates how to create a new \p tuple object
+ *  and inspect and modify the value of its elements.
+ *
+ *  \code
+ *  #include <thrust/tuple.h>
+ *  #include <iostream>
+ *  
+ *  int main() {
+ *    // Create a tuple containing an `int`, a `float`, and a string.
+ *    thrust::tuple<int, float, const char*> t(13, 0.1f, "thrust");
+ *
+ *    // Individual members are accessed with the free function `get`.
+ *    std::cout << "The first element's value is " << thrust::get<0>(t) << std::endl;
+ *
+ *    // ... or the member function `get`.
+ *    std::cout << "The second element's value is " << t.get<1>() << std::endl;
+ *
+ *    // We can also modify elements with the same function.
+ *    thrust::get<0>(t) += 10;
+ *  }
+ *  \endcode
+ *
+ *  \see pair
+ *  \see get
+ *  \see make_tuple
+ *  \see tuple_element
+ *  \see tuple_size
+ *  \see tie
+ */
+template <class T0, class T1, class T2, class T3, class T4,
+          class T5, class T6, class T7, class T8, class T9>
+  class tuple
+  /*! \cond
+   */
+    : public detail::map_tuple_to_cons<T0, T1, T2, T3, T4, T5, T6, T7, T8, T9>::type
+  /*! \endcond
+   */
+{
+  /*! \cond
+   */
+
+  private:
+  typedef typename detail::map_tuple_to_cons<T0, T1, T2, T3, T4, T5, T6, T7, T8, T9>::type inherited;
+
+  /*! \endcond
+   */
+
+  public:
+
+  /*! \p tuple's no-argument constructor initializes each element.
+   */
+  inline __host__ __device__
+  tuple(void) {}
+
+  /*! \p tuple's one-argument constructor copy constructs the first element from the given parameter
+   *     and intializes all other elements.
+   *  \param t0 The value to assign to this \p tuple's first element.
+   */
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0)
+    : inherited(t0,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  /*! \p tuple's one-argument constructor copy constructs the first two elements from the given parameters
+   *     and intializes all other elements.
+   *  \param t0 The value to assign to this \p tuple's first element.
+   *  \param t1 The value to assign to this \p tuple's second element.
+   *  \note \p tuple's constructor has ten variants of this form, the rest of which are ommitted here for brevity.
+   */
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1)
+    : inherited(t0, t1,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  /*! \cond
+   */
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2)
+    : inherited(t0, t1, t2,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3)
+    : inherited(t0, t1, t2, t3,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3,
+        typename access_traits<T4>::parameter_type t4)
+    : inherited(t0, t1, t2, t3, t4,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3,
+        typename access_traits<T4>::parameter_type t4,
+        typename access_traits<T5>::parameter_type t5)
+    : inherited(t0, t1, t2, t3, t4, t5,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3,
+        typename access_traits<T4>::parameter_type t4,
+        typename access_traits<T5>::parameter_type t5,
+        typename access_traits<T6>::parameter_type t6)
+    : inherited(t0, t1, t2, t3, t4, t5, t6,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3,
+        typename access_traits<T4>::parameter_type t4,
+        typename access_traits<T5>::parameter_type t5,
+        typename access_traits<T6>::parameter_type t6,
+        typename access_traits<T7>::parameter_type t7)
+    : inherited(t0, t1, t2, t3, t4, t5, t6, t7,
+                static_cast<const null_type&>(null_type()),
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3,
+        typename access_traits<T4>::parameter_type t4,
+        typename access_traits<T5>::parameter_type t5,
+        typename access_traits<T6>::parameter_type t6,
+        typename access_traits<T7>::parameter_type t7,
+        typename access_traits<T8>::parameter_type t8)
+    : inherited(t0, t1, t2, t3, t4, t5, t6, t7, t8,
+                static_cast<const null_type&>(null_type())) {}
+
+  inline __host__ __device__
+  tuple(typename access_traits<T0>::parameter_type t0,
+        typename access_traits<T1>::parameter_type t1,
+        typename access_traits<T2>::parameter_type t2,
+        typename access_traits<T3>::parameter_type t3,
+        typename access_traits<T4>::parameter_type t4,
+        typename access_traits<T5>::parameter_type t5,
+        typename access_traits<T6>::parameter_type t6,
+        typename access_traits<T7>::parameter_type t7,
+        typename access_traits<T8>::parameter_type t8,
+        typename access_traits<T9>::parameter_type t9)
+    : inherited(t0, t1, t2, t3, t4, t5, t6, t7, t8, t9) {}
+
+
+  template<class U1, class U2>
+  inline __host__ __device__
+  tuple(const detail::cons<U1, U2>& p) : inherited(p) {}
+
+  __thrust_exec_check_disable__
+  template <class U1, class U2>
+  inline __host__ __device__
+  tuple& operator=(const detail::cons<U1, U2>& k)
+  {
+    inherited::operator=(k);
+    return *this;
+  }
+
+  /*! \endcond
+   */
+
+  /*! This assignment operator allows assigning the first two elements of this \p tuple from a \p pair.
+   *  \param k A \p pair to assign from.
+   */
+  __thrust_exec_check_disable__
+  template <class U1, class U2>
+  __host__ __device__ inline
+  tuple& operator=(const thrust::pair<U1, U2>& k) {
+    //BOOST_STATIC_ASSERT(length<tuple>::value == 2);// check_length = 2
+    this->head = k.first;
+    this->tail.head = k.second;
+    return *this;
+  }
+
+  /*! \p swap swaps the elements of two <tt>tuple</tt>s.
+   *
+   *  \param t The other <tt>tuple</tt> with which to swap.
+   */
+  inline __host__ __device__
+  void swap(tuple &t)
+  {
+    inherited::swap(t);
+  }
+};
+
+/*! \cond
+ */
+
+template <>
+class tuple<null_type, null_type, null_type, null_type, null_type, null_type, null_type, null_type, null_type, null_type>  :
+  public null_type
+{
+public:
+  typedef null_type inherited;
+};
+
+/*! \endcond
+ */
+
+
+/*! This version of \p make_tuple creates a new \c tuple object from a
+ *  single object.
+ *
+ *  \param t0 The object to copy from.
+ *  \return A \p tuple object with a single member which is a copy of \p t0.
+ */
+template<class T0>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0>::type
+    make_tuple(const T0& t0);
+
+/*! This version of \p make_tuple creates a new \c tuple object from two
+ *  objects.
+ *
+ *  \param t0 The first object to copy from.
+ *  \param t1 The second object to copy from.
+ *  \return A \p tuple object with two members which are copies of \p t0
+ *          and \p t1.
+ *
+ *  \note \p make_tuple has ten variants, the rest of which are omitted here
+ *        for brevity.
+ */
+template<class T0, class T1>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1>::type
+    make_tuple(const T0& t0, const T1& t1);
+
+/*! This version of \p tie creates a new \c tuple whose single element is
+ *  a reference which refers to this function's argument.
+ *
+ *  \param t0 The object to reference.
+ *  \return A \p tuple object with one member which is a reference to \p t0.
+ */
+template<typename T0>
+__host__ __device__ inline
+tuple<T0&> tie(T0& t0);
+
+/*! This version of \p tie creates a new \c tuple of references object which
+ *  refers to this function's arguments.
+ *
+ *  \param t0 The first object to reference.
+ *  \param t1 The second object to reference.
+ *  \return A \p tuple object with two members which are references to \p t0
+ *          and \p t1.
+ *
+ *  \note \p tie has ten variants, the rest of which are omitted here for
+ *           brevity.
+ */
+template<typename T0, typename T1>
+__host__ __device__ inline
+tuple<T0&,T1&> tie(T0& t0, T1& t1);
+
+/*! \p swap swaps the contents of two <tt>tuple</tt>s.
+ *
+ *  \param x The first \p tuple to swap.
+ *  \param y The second \p tuple to swap.
+ */
+template<
+  typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9,
+  typename U0, typename U1, typename U2, typename U3, typename U4, typename U5, typename U6, typename U7, typename U8, typename U9
+>
+inline __host__ __device__
+void swap(tuple<T0,T1,T2,T3,T4,T5,T6,T7,T8,T9> &x,
+          tuple<U0,U1,U2,U3,U4,U5,U6,U7,U8,U9> &y);
+
+
+
+/*! \cond
+ */
+
+template<class T0, class T1, class T2>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2);
+
+template<class T0, class T1, class T2, class T3>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3);
+
+template<class T0, class T1, class T2, class T3, class T4>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4);
+
+template<class T0, class T1, class T2, class T3, class T4, class T5>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5);
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6);
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6, class T7>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6, T7>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6, const T7& t7);
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6, class T7, class T8>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6, T7, T8>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6, const T7& t7, const T8& t8);
+
+template<class T0, class T1, class T2, class T3, class T4, class T5, class T6, class T7, class T8, class T9>
+__host__ __device__ inline
+  typename detail::make_tuple_mapper<T0, T1, T2, T3, T4, T5, T6, T7, T8, T9>::type
+    make_tuple(const T0& t0, const T1& t1, const T2& t2, const T3& t3, const T4& t4, const T5& t5, const T6& t6, const T7& t7, const T8& t8, const T9& t9);
+
+template<typename T0, typename T1, typename T2>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&> tie(T0 &t0, T1 &t1, T2 &t2);
+
+template<typename T0, typename T1, typename T2, typename T3>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3);
+
+template<typename T0, typename T1, typename T2, typename T3, typename T4>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4);
+
+template<typename T0, typename T1, typename T2, typename T3, typename T4, typename T5>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5);
+
+template<typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6);
+
+template<typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6, T7 &t7);
+
+template<typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&,T8&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6, T7 &t7, T8 &t8);
+
+template<typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9>
+__host__ __device__ inline
+tuple<T0&,T1&,T2&,T3&,T4&,T5&,T6&,T7&,T8&,T9&> tie(T0 &t0, T1 &t1, T2 &t2, T3 &t3, T4 &t4, T5 &t5, T6 &t6, T7 &t7, T8 &t8, T9 &t9);
+
+
+__host__ __device__ inline
+bool operator==(const null_type&, const null_type&);
+
+__host__ __device__ inline
+bool operator>=(const null_type&, const null_type&);
+
+__host__ __device__ inline
+bool operator<=(const null_type&, const null_type&);
+
+__host__ __device__ inline
+bool operator!=(const null_type&, const null_type&);
+
+__host__ __device__ inline
+bool operator<(const null_type&, const null_type&);
+
+__host__ __device__ inline
+bool operator>(const null_type&, const null_type&);
+
+/*! \endcond
+ */
+
+/*! \} // tuple
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/integer_sequence.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/integer_sequence.h
new file mode 100644
index 000000000000..26ea542134a3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/integer_sequence.h
@@ -0,0 +1,381 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief C++14's
+ *  <a href="https://en.cppreference.com/w/cpp/utility/integer_sequence"><tt>std::index_sequence</tt></a>,
+ *  associated helper aliases, and some related extensions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <type_traits>
+#include <utility>
+#include <cstdint>
+#include <utility>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \brief A compile-time sequence of
+ *  <a href="https://en.cppreference.com/w/cpp/language/constant_expression#Integral_constant_expression"><i>integral constants</i></a>
+ *  of type \c T with values <tt>Is...</tt>.
+ *
+ *  \see <a href="https://en.cppreference.com/w/cpp/language/constant_expression#Integral_constant_expression"><i>integral constants</i></a>
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_index_sequence
+ *  \see make_reversed_index_sequence
+ *  \see integer_sequence_push_front
+ *  \see integer_sequence_push_back
+ *  \see <a href="https://en.cppreference.com/w/cpp/utility/integer_sequence"><tt>std::integer_sequence</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <typename T, T... Is>
+using integer_sequence = std::integer_sequence<T, Is...>;
+#else
+template <typename T, T... Is>
+struct integer_sequence
+{
+  using type = integer_sequence;
+  using value_type = T;
+  using size_type = std::size_t;
+
+  __host__ __device__
+  static constexpr size_type size() noexcept
+  {
+    return sizeof...(Is);
+  }
+};
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \brief A compile-time sequence of type
+ *  <a href="https://en.cppreference.com/w/cpp/types/size_t">std::size_t</a>
+ *  with values <tt>Is...</tt>.
+ *
+ *  \see integer_sequence
+ *  \see make_integer_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_index_sequence
+ *  \see make_reversed_index_sequence
+ *  \see integer_sequence_push_front
+ *  \see integer_sequence_push_back
+ *  \see <a href="https://en.cppreference.com/w/cpp/utility/integer_sequence"><tt>std::index_sequence</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <std::size_t... Is>
+using index_sequence = std::index_sequence<Is...>;
+#else
+template <std::size_t... Is>
+using index_sequence = integer_sequence<std::size_t, Is...>;
+#endif
+
+#if THRUST_CPP_DIALECT < 2014
+/*! \cond
+ */
+
+namespace detail
+{
+
+/*! \brief Create a new \c integer_sequence containing the elements of \c
+ * Sequence0 followed by the elements of \c Sequence1. \c Sequence0::size() is
+ * added to each element from \c Sequence1 in the new sequence.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_index_sequence
+ *  \see make_reversed_index_sequence
+ *  \see merge_and_renumber_reversed_integer_sequences_impl
+ */
+template <typename Sequence0, typename Sequence1>
+  struct merge_and_renumber_integer_sequences_impl;
+template <typename Sequence0, typename Sequence1>
+  using merge_and_renumber_integer_sequences =
+      typename merge_and_renumber_integer_sequences_impl<
+          Sequence0, Sequence1
+      >::type;
+
+template <typename T, std::size_t N>
+  struct make_integer_sequence_impl;
+
+} // namespace detail
+
+/*! \endcond
+ */
+#endif
+
+/*! \brief Create a new \c integer_sequence with elements
+ *  <tt>0, 1, 2, ..., N - 1</tt> of type \c T.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_index_sequence
+ *  \see make_reversed_index_sequence
+ *  \see <a href="https://en.cppreference.com/w/cpp/utility/integer_sequence"><tt>std::make_integer_sequence</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <typename T, std::size_t N>
+using make_integer_sequence = std::make_integer_sequence<T, N>;
+#else
+template <typename T, std::size_t N>
+using make_integer_sequence =
+  typename detail::make_integer_sequence_impl<T, N>::type;
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T, T... Is0, T... Is1>
+struct merge_and_renumber_integer_sequences_impl<
+  integer_sequence<T, Is0...>, integer_sequence<T, Is1...>
+>
+{
+  using type = integer_sequence<T, Is0..., (sizeof...(Is0) + Is1)...>;
+};
+
+template <typename T, std::size_t N>
+struct make_integer_sequence_impl
+{
+  using type = merge_and_renumber_integer_sequences<
+    make_integer_sequence<T, N / 2>
+  , make_integer_sequence<T, N - N / 2>
+  >;
+};
+
+template <typename T>
+struct make_integer_sequence_impl<T, 0>
+{
+  using type = integer_sequence<T>;
+};
+
+template <typename T>
+struct make_integer_sequence_impl<T, 1>
+{
+  using type = integer_sequence<T, 0>;
+};
+
+} // namespace detail
+
+/*! \endcond
+ */
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \brief Create a new \c integer_sequence with elements
+ *  <tt>0, 1, 2, ..., N - 1</tt> of type
+ *  <a href="https://en.cppreference.com/w/cpp/types/size_t">std::size_t</a>.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_reversed_index_sequence
+ *  \see <a href="https://en.cppreference.com/w/cpp/utility/integer_sequence"><tt>std::make_index_sequence</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <std::size_t N>
+using make_index_sequence = std::make_index_sequence<N>;
+#else
+template <std::size_t N>
+using make_index_sequence =
+  make_integer_sequence<std::size_t, N>;
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+/*! \brief Create a new \c integer_sequence containing the elements of \c
+ *  Sequence0 followed by the elements of \c Sequence1. \c Sequence1::size() is
+ *  added to each element from \c Sequence0 in the new sequence.
+ *
+ *  \see make_reversed_integer_sequence
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_index_sequence
+ *  \see make_reversed_index_sequence
+ *  \see merge_and_renumber_integer_sequences_impl
+ */
+template <typename Sequence0, typename Sequence1>
+  struct merge_and_renumber_reversed_integer_sequences_impl;
+template <typename Sequence0, typename Sequence1>
+  using merge_and_renumber_reversed_integer_sequences =
+      typename merge_and_renumber_reversed_integer_sequences_impl<
+          Sequence0, Sequence1
+      >::type;
+
+template <typename T, std::size_t N>
+struct make_reversed_integer_sequence_impl;
+
+template <typename T, T Value, typename Sequence>
+struct integer_sequence_push_front_impl;
+
+template <typename T, T Value, typename Sequence>
+struct integer_sequence_push_back_impl;
+
+template <typename T, T... Is0, T... Is1>
+struct merge_and_renumber_reversed_integer_sequences_impl<
+  integer_sequence<T, Is0...>, integer_sequence<T, Is1...>
+>
+{
+  using type = integer_sequence<T, (sizeof...(Is1) + Is0)..., Is1...>;
+};
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \brief Create a new \c integer_sequence with elements
+ *  <tt>N - 1, N - 2, N - 3, ..., 0</tt>.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_index_sequence
+ *  \see make_reversed_index_sequence
+ */
+template <typename T, std::size_t N>
+using make_reversed_integer_sequence =
+  typename detail::make_reversed_integer_sequence_impl<T, N>::type;
+
+/*! \brief Create a new \c index_sequence with elements
+ *  <tt>N - 1, N - 2, N - 3, ..., 0</tt>.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_reversed_integer_sequence
+ *  \see make_reversed_index_sequence
+ */
+template <std::size_t N>
+using make_reversed_index_sequence =
+  make_reversed_integer_sequence<std::size_t, N>;
+
+/*! \brief Add a new element to the front of an \c integer_sequence.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_index_sequence
+ */
+template <typename T, T Value, typename Sequence>
+using integer_sequence_push_front =
+  typename detail::integer_sequence_push_front_impl<T, Value, Sequence>::type;
+
+/*! \brief Add a new element to the back of an \c integer_sequence.
+ *
+ *  \see integer_sequence
+ *  \see index_sequence
+ *  \see make_integer_sequence
+ *  \see make_index_sequence
+ */
+template <typename T, T Value, typename Sequence>
+using integer_sequence_push_back =
+  typename detail::integer_sequence_push_back_impl<T, Value, Sequence>::type;
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T, std::size_t N>
+struct make_reversed_integer_sequence_impl
+{
+  using type = merge_and_renumber_reversed_integer_sequences<
+      make_reversed_integer_sequence<T, N / 2>
+    , make_reversed_integer_sequence<T, N - N / 2>
+  >;
+};
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T>
+struct make_reversed_integer_sequence_impl<T, 0>
+{
+  using type = integer_sequence<T>;
+};
+
+template <typename T>
+struct make_reversed_integer_sequence_impl<T, 1>
+{
+  using type = integer_sequence<T, 0>;
+};
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T, T I0, T... Is>
+struct integer_sequence_push_front_impl<T, I0, integer_sequence<T, Is...> >
+{
+  using type = integer_sequence<T, I0, Is...>;
+};
+
+///////////////////////////////////////////////////////////////////////////////
+
+template <typename T, T I0, T... Is>
+struct integer_sequence_push_back_impl<T, I0, integer_sequence<T, Is...> >
+{
+  using type = integer_sequence<T, Is..., I0>;
+};
+
+///////////////////////////////////////////////////////////////////////////////
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_contiguous_iterator.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_contiguous_iterator.h
new file mode 100644
index 000000000000..eaa088978c45
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_contiguous_iterator.h
@@ -0,0 +1,295 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief An extensible type trait for determining if an iterator satisifies the
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/ContiguousIterator">ContiguousIterator</a>
+ *  requirements (aka is pointer-like).
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/raw_pointer_cast.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+#include <iterator>
+#include <type_traits>
+#include <utility>
+
+#if THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_MSVC && _MSC_VER < 1916 // MSVC 2017 version 15.9
+  #include <vector>
+  #include <string>
+  #include <array>
+
+  #if THRUST_CPP_DIALECT >= 2017
+    #include <string_view>
+  #endif
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename Iterator>
+struct is_contiguous_iterator_impl;
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c Iterator satisfies
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/ContiguousIterator">ContiguousIterator</a>,
+ *  aka it points to elements that are contiguous in memory, and \c false_type
+ *  otherwise.
+ *
+ * \see is_contiguous_iterator_v
+ * \see proclaim_contiguous_iterator
+ * \see THRUST_PROCLAIM_CONTIGUOUS_ITERATOR
+ */
+template <typename Iterator>
+#if THRUST_CPP_DIALECT >= 2011
+using is_contiguous_iterator =
+#else
+struct is_contiguous_iterator :
+#endif
+  detail::is_contiguous_iterator_impl<Iterator>
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c Iterator satisfies
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/ContiguousIterator">ContiguousIterator</a>,
+ *  aka it points to elements that are contiguous in memory, and \c false
+ *  otherwise.
+ *
+ * \see is_contiguous_iterator
+ * \see proclaim_contiguous_iterator
+ * \see THRUST_PROCLAIM_CONTIGUOUS_ITERATOR
+ */
+template <typename Iterator>
+constexpr bool is_contiguous_iterator_v = is_contiguous_iterator<Iterator>::value;
+#endif
+
+/*! \brief Customization point that can be customized to indicate that an
+ *  iterator type \c Iterator satisfies
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/ContiguousIterator">ContiguousIterator</a>,
+ *  aka it points to elements that are contiguous in memory.
+ *
+ * \see is_contiguous_iterator
+ * \see THRUST_PROCLAIM_CONTIGUOUS_ITERATOR
+ */
+template <typename Iterator>
+struct proclaim_contiguous_iterator : false_type {};
+
+/*! \brief Declares that the iterator \c Iterator is
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/ContiguousIterator">ContiguousIterator</a>
+ *  by specializing \c proclaim_contiguous_iterator.
+ *
+ * \see is_contiguous_iterator
+ * \see proclaim_contiguous_iterator
+ */
+#define THRUST_PROCLAIM_CONTIGUOUS_ITERATOR(Iterator)                         \
+  THRUST_NAMESPACE_BEGIN                                                      \
+  template <>                                                                 \
+  struct proclaim_contiguous_iterator<Iterator>                               \
+      : THRUST_NS_QUALIFIER::true_type {};                                    \
+  THRUST_NAMESPACE_END                                                        \
+  /**/
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename Iterator>
+struct is_libcxx_wrap_iter : false_type {};
+
+#if defined(_LIBCPP_VERSION)
+template <typename Iterator>
+struct is_libcxx_wrap_iter<
+  _VSTD::__wrap_iter<Iterator>
+> : true_type {};
+#endif
+
+template <typename Iterator>
+struct is_libstdcxx_normal_iterator : false_type {};
+
+#if defined(__GLIBCXX__)
+template <typename Iterator, typename Container>
+struct is_libstdcxx_normal_iterator<
+  ::__gnu_cxx::__normal_iterator<Iterator, Container>
+> : true_type {};
+#endif
+
+#if   _MSC_VER >= 1916 // MSVC 2017 version 15.9.
+template <typename Iterator>
+struct is_msvc_contiguous_iterator
+  : is_pointer<::std::_Unwrapped_t<Iterator> > {};
+#elif _MSC_VER >= 1700 // MSVC 2012.
+template <typename Iterator>
+struct is_msvc_contiguous_iterator : false_type {};
+
+template <typename Vector>
+struct is_msvc_contiguous_iterator<
+  ::std::_Vector_const_iterator<Vector>
+> : true_type {};
+
+template <typename Vector>
+struct is_msvc_contiguous_iterator<
+  ::std::_Vector_iterator<Vector>
+> : true_type {};
+
+template <typename String>
+struct is_msvc_contiguous_iterator<
+  ::std::_String_const_iterator<String>
+> : true_type {};
+
+template <typename String>
+struct is_msvc_contiguous_iterator<
+  ::std::_String_iterator<String>
+> : true_type {};
+
+template <typename T, std::size_t N>
+struct is_msvc_contiguous_iterator<
+  ::std::_Array_const_iterator<T, N>
+> : true_type {};
+
+template <typename T, std::size_t N>
+struct is_msvc_contiguous_iterator<
+  ::std::_Array_iterator<T, N>
+> : true_type {};
+
+#if THRUST_CPP_DIALECT >= 2017
+template <typename Traits>
+struct is_msvc_contiguous_iterator<
+  ::std::_String_view_iterator<Traits>
+> : true_type {};
+#endif
+#else
+template <typename Iterator>
+struct is_msvc_contiguous_iterator : false_type {};
+#endif
+
+template <typename Iterator>
+struct is_contiguous_iterator_impl
+  : integral_constant<
+      bool
+    ,    is_pointer<Iterator>::value
+      || is_thrust_pointer<Iterator>::value
+      || is_libcxx_wrap_iter<Iterator>::value
+      || is_libstdcxx_normal_iterator<Iterator>::value
+      || is_msvc_contiguous_iterator<Iterator>::value
+      || proclaim_contiguous_iterator<Iterator>::value
+    >
+{};
+
+// Type traits for contiguous iterators:
+template <typename Iterator>
+struct contiguous_iterator_traits
+{
+  static_assert(thrust::is_contiguous_iterator<Iterator>::value,
+                "contiguous_iterator_traits requires a contiguous iterator.");
+
+  using raw_pointer = typename thrust::detail::pointer_traits<
+    decltype(&*std::declval<Iterator>())>::raw_pointer;
+};
+
+template <typename Iterator>
+using contiguous_iterator_raw_pointer_t =
+  typename contiguous_iterator_traits<Iterator>::raw_pointer;
+
+// Converts a contiguous iterator to a raw pointer:
+template <typename Iterator>
+__host__ __device__
+contiguous_iterator_raw_pointer_t<Iterator>
+contiguous_iterator_raw_pointer_cast(Iterator it)
+{
+  static_assert(thrust::is_contiguous_iterator<Iterator>::value,
+                "contiguous_iterator_raw_pointer_cast called with "
+                "non-contiguous iterator.");
+  return thrust::raw_pointer_cast(&*it);
+}
+
+// Implementation for non-contiguous iterators -- passthrough.
+template <typename Iterator,
+          bool IsContiguous = thrust::is_contiguous_iterator<Iterator>::value>
+struct try_unwrap_contiguous_iterator_impl
+{
+  using type = Iterator;
+
+  static __host__ __device__ type get(Iterator it) { return it; }
+};
+
+// Implementation for contiguous iterators -- unwraps to raw pointer.
+template <typename Iterator>
+struct try_unwrap_contiguous_iterator_impl<Iterator, true /*is_contiguous*/>
+{
+  using type = contiguous_iterator_raw_pointer_t<Iterator>;
+
+  static __host__ __device__ type get(Iterator it)
+  {
+    return contiguous_iterator_raw_pointer_cast(it);
+  }
+};
+
+template <typename Iterator>
+using try_unwrap_contiguous_iterator_return_t =
+  typename try_unwrap_contiguous_iterator_impl<Iterator>::type;
+
+// Casts to a raw pointer if iterator is marked as contiguous, otherwise returns
+// the input iterator.
+template <typename Iterator>
+__host__ __device__
+try_unwrap_contiguous_iterator_return_t<Iterator>
+try_unwrap_contiguous_iterator(Iterator it)
+{
+  return try_unwrap_contiguous_iterator_impl<Iterator>::get(it);
+}
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_execution_policy.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_execution_policy.h
new file mode 100644
index 000000000000..f83751ea2da7
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_execution_policy.h
@@ -0,0 +1,69 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief A type trait that determines if a type is an \a ExecutionPolicy.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c T is an \a ExecutionPolicy and \c false_type
+ *  otherwise.
+ */
+template <typename T>
+#if THRUST_CPP_DIALECT >= 2011
+using is_execution_policy =
+#else
+struct is_execution_policy :
+#endif
+  detail::is_base_of<detail::execution_policy_marker, T>
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c T is an
+ *  \a ExecutionPolicy and \c false otherwise.
+ */
+template <typename T>
+constexpr bool is_execution_policy_v = is_execution_policy<T>::value;
+#endif
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h
new file mode 100644
index 000000000000..ef5a19f691e5
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_less_or_greater_function_object.h
@@ -0,0 +1,208 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Type traits for determining if a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  is equivalent to either \c operator< or \c operator>.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/functional.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T>
+struct is_operator_less_function_object_impl;
+
+template <typename T>
+struct is_operator_greater_function_object_impl;
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  equivalent to \c operator<, and \c false_type otherwise.
+ *
+ *  \see is_operator_less_function_object_v
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_less_or_greater_function_object
+ *  \see is_operator_plus_function_object
+ */
+template <typename T>
+#if THRUST_CPP_DIALECT >= 2011
+using is_operator_less_function_object =
+#else
+struct is_operator_less_function_object :
+#endif
+  detail::is_operator_less_function_object_impl<T>
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  equivalent to \c operator<, and \c false otherwise.
+ *
+ *  \see is_operator_less_function_object
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_less_or_greater_function_object
+ *  \see is_operator_plus_function_object
+ */
+template <typename T>
+constexpr bool is_operator_less_function_object_v
+  = is_operator_less_function_object<T>::value;
+#endif
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  equivalent to \c operator>, and \c false_type otherwise.
+ *
+ *  \see is_operator_greater_function_object_v
+ *  \see is_operator_less_function_object
+ *  \see is_operator_less_or_greater_function_object
+ *  \see is_operator_plus_function_object
+ */
+template <typename T>
+#if THRUST_CPP_DIALECT >= 2011
+using is_operator_greater_function_object =
+#else
+struct is_operator_greater_function_object :
+#endif
+  detail::is_operator_greater_function_object_impl<T>
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  equivalent to \c operator>, and \c false otherwise.
+ *
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_less_function_object
+ *  \see is_operator_less_or_greater_function_object
+ *  \see is_operator_plus_function_object
+ */
+template <typename T>
+constexpr bool is_operator_greater_function_object_v
+  = is_operator_greater_function_object<T>::value;
+#endif
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  equivalent to \c operator< or \c operator>, and \c false_type otherwise.
+ *
+ *  \see is_operator_less_or_greater_function_object_v
+ *  \see is_operator_less_function_object
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_plus_function_object
+ */
+template <typename T>
+#if THRUST_CPP_DIALECT >= 2011
+using is_operator_less_or_greater_function_object =
+#else
+struct is_operator_less_or_greater_function_object :
+#endif
+  integral_constant<
+    bool
+  ,    detail::is_operator_less_function_object_impl<T>::value
+    || detail::is_operator_greater_function_object_impl<T>::value
+  >
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">BinaryPredicate</a>
+ *  equivalent to \c operator< or \c operator>, and \c false otherwise.
+ *
+ *  \see is_operator_less_or_greater_function_object
+ *  \see is_operator_less_function_object
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_plus_function_object
+ */
+template <typename T>
+constexpr bool is_operator_less_or_greater_function_object_v
+  = is_operator_less_or_greater_function_object<T>::value;
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T>
+struct is_operator_less_function_object_impl                   : false_type {};
+template <typename T>
+struct is_operator_less_function_object_impl<thrust::less<T> > : true_type {};
+template <typename T>
+struct is_operator_less_function_object_impl<std::less<T>    > : true_type {};
+
+template <typename T>
+struct is_operator_greater_function_object_impl                      : false_type {};
+template <typename T>
+struct is_operator_greater_function_object_impl<thrust::greater<T> > : true_type {};
+template <typename T>
+struct is_operator_greater_function_object_impl<std::greater<T>    > : true_type {};
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_plus_function_object.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_plus_function_object.h
new file mode 100644
index 000000000000..800847532cf3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_operator_plus_function_object.h
@@ -0,0 +1,118 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief Type traits for determining if a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/FunctionObject">FunctionObject</a>
+ *  is equivalent to \c operator+.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/functional.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/detail/type_traits/pointer_traits.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T>
+struct is_operator_plus_function_object_impl;
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/FunctionObject">FunctionObject</a>
+ *  equivalent to \c operator+, and \c false_type otherwise.
+ *
+ *  \see is_operator_plus_function_object_v
+ *  \see is_operator_less_function_object
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_less_or_greater_function_object
+ */
+template <typename T>
+#if THRUST_CPP_DIALECT >= 2011
+using is_operator_plus_function_object =
+#else
+struct is_operator_plus_function_object :
+#endif
+  detail::is_operator_plus_function_object_impl<T>
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c T is a
+ *  <a href="https://en.cppreference.com/w/cpp/named_req/FunctionObject">FunctionObject</a>
+ *  equivalent to \c operator<, and \c false otherwise.
+ *
+ *  \see is_operator_plus_function_object
+ *  \see is_operator_less_function_object
+ *  \see is_operator_greater_function_object
+ *  \see is_operator_less_or_greater_function_object
+ */
+template <typename T>
+constexpr bool is_operator_plus_function_object_v
+  = is_operator_plus_function_object<T>::value;
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T>
+struct is_operator_plus_function_object_impl                   : false_type {};
+template <typename T>
+struct is_operator_plus_function_object_impl<thrust::plus<T> > : true_type {};
+template <typename T>
+struct is_operator_plus_function_object_impl<std::plus<T>    > : true_type {};
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_trivially_relocatable.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_trivially_relocatable.h
new file mode 100644
index 000000000000..21d1f09d8d7f
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/is_trivially_relocatable.h
@@ -0,0 +1,366 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief <a href="https://wg21.link/P1144">P1144</a>'s proposed
+ *  \c std::is_trivially_relocatable, an extensible type trait indicating
+ *  whether a type can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/static_assert.h>
+#include <thrust/detail/type_traits.h>
+#include <thrust/type_traits/is_contiguous_iterator.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+  #include <type_traits>
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+template <typename T>
+struct is_trivially_relocatable_impl;
+
+} // namespace detail
+
+/*! \endcond
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that returns \c true_type if \c T is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  aka can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  and \c false_type otherwise.
+ *
+ * \see is_trivially_relocatable_v
+ * \see is_trivially_relocatable_to
+ * \see is_indirectly_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename T>
+#if THRUST_CPP_DIALECT >= 2011
+using is_trivially_relocatable =
+#else
+struct is_trivially_relocatable :
+#endif
+  detail::is_trivially_relocatable_impl<T>
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c T is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  aka can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  and \c false otherwise.
+ *
+ * \see is_trivially_relocatable
+ * \see is_trivially_relocatable_to
+ * \see is_indirectly_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename T>
+constexpr bool is_trivially_relocatable_v = is_trivially_relocatable<T>::value;
+#endif
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/BinaryTypeTrait"><i>BinaryTypeTrait</i></a>
+ *  that returns \c true_type if \c From is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  to \c To, aka can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  and \c false_type otherwise.
+ *
+ * \see is_trivially_relocatable_to_v
+ * \see is_trivially_relocatable
+ * \see is_indirectly_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename From, typename To>
+#if THRUST_CPP_DIALECT >= 2011
+using is_trivially_relocatable_to =
+#else
+struct is_trivially_relocatable_to :
+#endif
+  integral_constant<
+    bool
+  , detail::is_same<From, To>::value && is_trivially_relocatable<To>::value
+  >
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if \c From is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  to \c To, aka can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  and \c false otherwise.
+ *
+ * \see is_trivially_relocatable_to
+ * \see is_trivially_relocatable
+ * \see is_indirectly_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename From, typename To>
+constexpr bool is_trivially_relocatable_to_v
+  = is_trivially_relocatable_to<From, To>::value;
+#endif
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/BinaryTypeTrait"><i>BinaryTypeTrait</i></a>
+ *  that returns \c true_type if the element type of \c FromIterator is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  to the element type of \c ToIterator, aka can be bitwise copied with a
+ *  facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  and \c false_type otherwise.
+ *
+ * \see is_indirectly_trivially_relocatable_to_v
+ * \see is_trivially_relocatable
+ * \see is_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename FromIterator, typename ToIterator>
+#if THRUST_CPP_DIALECT >= 2011
+using is_indirectly_trivially_relocatable_to =
+#else
+struct is_indirectly_trivially_relocatable_to :
+#endif
+  integral_constant<
+    bool
+  ,    is_contiguous_iterator<FromIterator>::value
+    && is_contiguous_iterator<ToIterator>::value
+    && is_trivially_relocatable_to<
+         typename thrust::iterator_traits<FromIterator>::value_type,
+         typename thrust::iterator_traits<ToIterator>::value_type
+       >::value
+  >
+#if THRUST_CPP_DIALECT < 2011
+{}
+#endif
+;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> that is \c true if the element type of
+ *  \c FromIterator is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  to the element type of \c ToIterator, aka can be bitwise copied with a
+ *  facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  and \c false otherwise.
+ *
+ * \see is_indirectly_trivially_relocatable_to
+ * \see is_trivially_relocatable
+ * \see is_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename FromIterator, typename ToIterator>
+constexpr bool is_indirectly_trivially_relocate_to_v
+  = is_indirectly_trivially_relocatable_to<FromIterator, ToIterator>::value;
+#endif
+
+/*! \brief <a href="http://eel.is/c++draft/namespace.std#def:customization_point"><i>customization point</i></a>
+ *  that can be specialized customized to indicate that a type \c T is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  aka it can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>.
+ *
+ * \see is_indirectly_trivially_relocatable_to
+ * \see is_trivially_relocatable
+ * \see is_trivially_relocatable_to
+ * \see THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE
+ */
+template <typename T>
+struct proclaim_trivially_relocatable : false_type {};
+
+/*! \brief Declares that the type \c T is
+ *  <a href="https://wg21.link/P1144"><i>TriviallyRelocatable</i></a>,
+ *  aka it can be bitwise copied with a facility like
+ *  <a href="https://en.cppreference.com/w/cpp/string/byte/memcpy"><tt>std::memcpy</tt></a>,
+ *  by specializing \c proclaim_trivially_relocatable.
+ *
+ * \see is_indirectly_trivially_relocatable_to
+ * \see is_trivially_relocatable
+ * \see is_trivially_relocatable_to
+ * \see proclaim_trivially_relocatable
+ */
+#define THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(T)                              \
+  THRUST_NAMESPACE_BEGIN                                                      \
+  template <>                                                                 \
+  struct proclaim_trivially_relocatable<T> : THRUST_NS_QUALIFIER::true_type   \
+  {};                                                                         \
+  THRUST_NAMESPACE_END                                                        \
+  /**/
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \cond
+ */
+
+namespace detail
+{
+
+// There is no way to actually detect the libstdc++ version; __GLIBCXX__
+// is always set to the date of libstdc++ being packaged, not the release
+// day or version. This means that we can't detect the libstdc++ version,
+// except when compiling with GCC.
+//
+// Therefore, for the best approximation of is_trivially_copyable, we need to
+// handle three distinct cases:
+// 1) GCC above 5, or another C++11 compiler not using libstdc++: use the
+//      standard trait directly.
+// 2) A C++11 compiler using libstdc++ that provides the intrinsic: use the
+//      intrinsic.
+// 3) Any other case (essentially: compiling without C++11): has_trivial_assign.
+
+#ifndef __has_feature
+    #define __has_feature(x) 0
+#endif
+
+template <typename T>
+struct is_trivially_copyable_impl
+    : integral_constant<
+        bool,
+        #if THRUST_CPP_DIALECT >= 2011
+            #if defined(__GLIBCXX__) && __has_feature(is_trivially_copyable)
+                __is_trivially_copyable(T)
+            #elif THRUST_HOST_COMPILER == THRUST_HOST_COMPILER_GCC && THRUST_GCC_VERSION >= 50000
+                std::is_trivially_copyable<T>::value
+            #else
+                has_trivial_assign<T>::value
+            #endif
+        #else
+            has_trivial_assign<T>::value
+        #endif
+    >
+{
+};
+
+// https://wg21.link/P1144R0#wording-inheritance
+template <typename T>
+struct is_trivially_relocatable_impl
+    : integral_constant<
+        bool,
+        is_trivially_copyable_impl<T>::value
+            || proclaim_trivially_relocatable<T>::value
+    >
+{};
+
+template <typename T, std::size_t N>
+struct is_trivially_relocatable_impl<T[N]> : is_trivially_relocatable_impl<T> {};
+
+} // namespace detail
+
+THRUST_NAMESPACE_END
+
+#if THRUST_DEVICE_SYSTEM == THRUST_DEVICE_SYSTEM_CUDA
+
+#include <thrust/system/cuda/detail/guarded_cuda_runtime_api.h>
+
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(char1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(char2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(char3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(char4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uchar1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uchar2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uchar3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uchar4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(short1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(short2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(short3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(short4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ushort1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ushort2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ushort3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ushort4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(int1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(int2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(int3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(int4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uint1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uint2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uint3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(uint4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(long1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(long2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(long3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(long4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulong1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulong2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulong3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulong4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(longlong1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(longlong2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(longlong3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(longlong4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulonglong1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulonglong2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulonglong3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(ulonglong4)
+
+struct __half;
+struct __half2;
+
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(__half)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(__half2)
+
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(float1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(float2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(float3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(float4)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(double1)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(double2)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(double3)
+THRUST_PROCLAIM_TRIVIALLY_RELOCATABLE(double4)
+#endif
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/logical_metafunctions.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/logical_metafunctions.h
new file mode 100644
index 000000000000..914b477e8ef2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/logical_metafunctions.h
@@ -0,0 +1,287 @@
+/*
+ *  Copyright 2008-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief C++17's
+ *  <a href="https://en.cppreference.com/w/cpp/types/conjunction"><tt>std::conjunction</tt></a>,
+ *  <a href="https://en.cppreference.com/w/cpp/types/disjunction"><tt>std::disjunction</tt></a>,
+ *  and <a href="https://en.cppreference.com/w/cpp/types/negation"><tt>std::negation</tt></a>
+ *  metafunctions and related extensions.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011
+
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/types/integral_constant"><tt>std::integral_constant</tt></a>
+ *  whose value is <tt>(... && Ts::value)</tt>.
+ *
+ *  \see conjunction_v
+ *  \see conjunction_value
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/conjunction"><tt>std::conjunction</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2017
+template <typename... Ts>
+using conjunction = std::conjunction<Ts...>;
+#else // Older than C++17.
+template <typename... Ts>
+struct conjunction;
+
+/*! \cond
+ */
+
+template <>
+struct conjunction<> : std::true_type {};
+
+template <typename T>
+struct conjunction<T> : T {};
+
+template <typename T0, typename T1>
+struct conjunction<T0, T1> : std::conditional<T0::value, T1, T0>::type {};
+
+template<typename T0, typename T1, typename T2, typename... TN>
+struct conjunction<T0, T1, T2, TN...>
+  : std::conditional<T0::value, conjunction<T1, T2, TN...>, T0>::type {};
+
+/*! \endcond
+ */
+#endif
+
+/*! \brief <tt>constexpr bool</tt> whose value is <tt>(... && Ts::value)</tt>.
+ *
+ *  \see conjunction
+ *  \see conjunction_value
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/conjunction"><tt>std::conjunction</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <typename... Ts>
+constexpr bool conjunction_v = conjunction<Ts...>::value;
+#endif
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/types/integral_constant"><tt>std::integral_constant</tt></a>
+ *  whose value is <tt>(... || Ts::value)</tt>.
+ *
+ *  \see disjunction_v
+ *  \see disjunction_value
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/disjunction"><tt>std::disjunction</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2017
+template <typename... Ts>
+using disjunction = std::disjunction<Ts...>;
+#else // Older than C++17.
+template <typename... Ts>
+struct disjunction;
+
+/*! \cond
+ */
+
+template <>
+struct disjunction<> : std::false_type {};
+
+template <typename T>
+struct disjunction<T> : T {};
+
+template <typename T0, typename... TN>
+struct disjunction<T0, TN...>
+  : std::conditional<T0::value != false, T0, disjunction<TN...> >::type {};
+
+/*! \endcond
+ */
+#endif
+
+/*! \brief <tt>constexpr bool</tt> whose value is <tt>(... || Ts::value)</tt>.
+ *
+ *  \see disjunction
+ *  \see disjunction_value
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/disjunction"><tt>std::disjunction</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <typename... Ts>
+constexpr bool disjunction_v = disjunction<Ts...>::value;
+#endif
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/types/integral_constant"><tt>std::integral_constant</tt></a>
+ *  whose value is <tt>!Ts::value</tt>.
+ *
+ *  \see negation_v
+ *  \see negation_value
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/negation"><tt>std::negation</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2017
+template <typename T>
+using negation = std::negation<T>;
+#else // Older than C++17.
+template <typename T>
+struct negation;
+
+/*! \cond
+ */
+
+template <typename T>
+struct negation : std::integral_constant<bool, !T::value> {};
+
+/*! \endcond
+ */
+#endif
+
+/*! \brief <tt>constexpr bool</tt> whose value is <tt>!Ts::value</tt>.
+ *
+ *  \see negation
+ *  \see negation_value
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/negation"><tt>std::negation</tt></a>
+ */
+#if THRUST_CPP_DIALECT >= 2014
+template <typename T>
+constexpr bool negation_v = negation<T>::value;
+#endif
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/types/integral_constant"><tt>std::integral_constant</tt></a>
+ *  whose value is <tt>(... && Bs)</tt>.
+ *
+ *  \see conjunction_value_v
+ *  \see conjunction
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/conjunction"><tt>std::conjunction</tt></a>
+ */
+template <bool... Bs>
+struct conjunction_value;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> whose value is <tt>(... && Bs)</tt>.
+ *
+ *  \see conjunction_value
+ *  \see conjunction
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/conjunction"><tt>std::conjunction</tt></a>
+ */
+template <bool... Bs>
+constexpr bool conjunction_value_v = conjunction_value<Bs...>::value;
+#endif
+
+/*! \cond
+ */
+
+template <>
+struct conjunction_value<> : std::true_type {};
+
+template <bool B>
+struct conjunction_value<B> : std::integral_constant<bool, B> {};
+
+template <bool B, bool... Bs>
+struct conjunction_value<B, Bs...>
+  : std::integral_constant<bool, B && conjunction_value<Bs...>::value> {};
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/types/integral_constant"><tt>std::integral_constant</tt></a>
+ *  whose value is <tt>(... || Bs)</tt>.
+ *
+ *  \see disjunction_value_v
+ *  \see disjunction
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/disjunction"><tt>std::disjunction</tt></a>
+ */
+template <bool... Bs>
+struct disjunction_value;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> whose value is <tt>(... || Bs)</tt>.
+ *
+ *  \see disjunction_value
+ *  \see disjunction
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/disjunction"><tt>std::disjunction</tt></a>
+ */
+template <bool... Bs>
+constexpr bool disjunction_value_v = disjunction_value<Bs...>::value;
+#endif
+
+/*! \cond
+ */
+
+template <>
+struct disjunction_value<> : std::false_type {};
+
+template <bool B>
+struct disjunction_value<B> : std::integral_constant<bool, B> {};
+
+template <bool B, bool... Bs>
+struct disjunction_value<B, Bs...>
+  : std::integral_constant<bool, B || disjunction_value<Bs...>::value> {};
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/types/integral_constant"><tt>std::integral_constant</tt></a>
+ *  whose value is <tt>!Bs</tt>.
+ *
+ *  \see negation_value_v
+ *  \see negation
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/negation"><tt>std::negation</tt></a>
+ */
+template <bool B>
+struct negation_value;
+
+#if THRUST_CPP_DIALECT >= 2014
+/*! \brief <tt>constexpr bool</tt> whose value is <tt>!Ts::value</tt>.
+ *
+ *  \see negation_value
+ *  \see negation
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/negation"><tt>std::negation</tt></a>
+ */
+template <bool B>
+constexpr bool negation_value_v = negation_value<B>::value;
+#endif
+
+/*! \cond
+ */
+
+template <bool B>
+struct negation_value : std::integral_constant<bool, !B> {};
+
+/*! \endcond
+ */
+
+///////////////////////////////////////////////////////////////////////////////
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
+#endif // THRUST_CPP_DIALECT >= 2011
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/remove_cvref.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/remove_cvref.h
new file mode 100644
index 000000000000..1da2e0de35c3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/remove_cvref.h
@@ -0,0 +1,97 @@
+/*
+ *  Copyright 2018-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief C++20's
+ *  <a href="https://en.cppreference.com/w/cpp/types/remove_cvref">std::remove_cvref</a>.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#if  THRUST_CPP_DIALECT >= 2017
+#if __has_include(<version>)
+#  include <version>
+#endif
+#endif
+
+#include <type_traits>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+/*! \brief <a href="https://en.cppreference.com/w/cpp/named_req/UnaryTypeTrait"><i>UnaryTypeTrait</i></a>
+ *  that removes
+ *  <a href="https://en.cppreference.com/w/cpp/language/cv">const-volatile qualifiers</a>
+ *  and
+ *  <a href="https://en.cppreference.com/w/cpp/language/reference">references</a>
+ *  from \c T.
+ *  Equivalent to \c remove_cv_t<remove_reference_t<T>>.
+ *
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cvref">std::remove_cvref</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_cv</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_const</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_volatile</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_reference</a>
+ */
+#if defined(__cpp_lib_remove_cvref) && (__cpp_lib_remove_cvref >= 201711L)
+using std::remove_cvref;
+#else // Older than C++20.
+template <typename T>
+struct remove_cvref
+{
+  using type = typename std::remove_cv<
+    typename std::remove_reference<T>::type
+  >::type;
+};
+#endif
+
+/*! \brief Type alias that removes
+ *  <a href="https://en.cppreference.com/w/cpp/language/cv">const-volatile qualifiers</a>
+ *  and
+ *  <a href="https://en.cppreference.com/w/cpp/language/reference">references</a>
+ *  from \c T.
+ *  Equivalent to \c remove_cv_t<remove_reference_t<T>>.
+ *
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cvref">std::remove_cvref</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_cv</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_const</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_volatile</a>
+ *  \see <a href="https://en.cppreference.com/w/cpp/types/remove_cv">std::remove_reference</a>
+ */
+#if defined(__cpp_lib_remove_cvref) && (__cpp_lib_remove_cvref >= 201711L)
+using std::remove_cvref_t;
+#else // Older than C++20.
+template <typename T>
+using remove_cvref_t = typename remove_cvref<T>::type;
+#endif
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/void_t.h b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/void_t.h
new file mode 100644
index 000000000000..ed12d861dcb0
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/type_traits/void_t.h
@@ -0,0 +1,77 @@
+/*
+ *  Copyright 2018-2021 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief C++17's `void_t`.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+#if THRUST_CPP_DIALECT >= 2017
+#  include <type_traits>
+#endif
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup utility
+ *  \{
+ */
+
+/*! \addtogroup type_traits Type Traits
+ *  \{
+ */
+
+#if THRUST_CPP_DIALECT >= 2011
+
+template <typename...> struct voider { using type = void; };
+
+#if THRUST_CPP_DIALECT >= 2017
+using std::void_t;
+#else
+template <typename... Ts> using void_t = typename voider<Ts...>::type;
+#endif
+
+#else // Older than C++11.
+
+template <
+  typename = void
+, typename = void
+, typename = void
+, typename = void
+, typename = void
+, typename = void
+, typename = void
+, typename = void
+, typename = void
+, typename = void
+>
+struct voider
+{
+  typedef void type;
+};
+
+#endif
+
+/*! \} // type traits
+ */
+
+/*! \} // utility
+ */
+
+THRUST_NAMESPACE_END
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_copy.h b/thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_copy.h
new file mode 100644
index 000000000000..94c2763e3e46
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_copy.h
@@ -0,0 +1,299 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file uninitialized_copy.h
+ *  \brief Copy construction into a range of uninitialized elements from a source range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup copying
+ *  \{
+ */
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a constructor.
+ *  Occasionally, however, it is useful to separate those two operations.
+ *  If each iterator in the range <tt>[result, result + (last - first))</tt> points
+ *  to uninitialized memory, then \p uninitialized_copy creates a copy of
+ *  <tt>[first, last)</tt> in that range. That is, for each iterator \c i in
+ *  the input, \p uninitialized_copy creates a copy of \c *i in the location pointed
+ *  to by the corresponding iterator in the output range by \p ForwardIterator's
+ *  \c value_type's copy constructor with *i as its argument.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the input range to copy from.
+ *  \param last The last element of the input range to copy from.
+ *  \param result The first element of the output range to copy to.
+ *  \return An iterator pointing to the last element of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that takes
+ *          a single argument whose type is \p InputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_copy to initialize
+ *  a range of uninitialized memory using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/uninitialized_copy.h>
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_vector<Int> input(N, val);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_copy(thrust::device, input.begin(), input.end(), array);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_copy
+ *  \see \c copy
+ *  \see \c uninitialized_fill
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename DerivedPolicy, typename InputIterator, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                     InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result);
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a constructor.
+ *  Occasionally, however, it is useful to separate those two operations.
+ *  If each iterator in the range <tt>[result, result + (last - first))</tt> points
+ *  to uninitialized memory, then \p uninitialized_copy creates a copy of
+ *  <tt>[first, last)</tt> in that range. That is, for each iterator \c i in
+ *  the input, \p uninitialized_copy creates a copy of \c *i in the location pointed
+ *  to by the corresponding iterator in the output range by \p ForwardIterator's
+ *  \c value_type's copy constructor with *i as its argument.
+ *
+ *  \param first The first element of the input range to copy from.
+ *  \param last The last element of the input range to copy from.
+ *  \param result The first element of the output range to copy to.
+ *  \return An iterator pointing to the last element of the output range.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that takes
+ *          a single argument whose type is \p InputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_copy to initialize
+ *  a range of uninitialized memory.
+ *
+ *  \code
+ *  #include <thrust/uninitialized_copy.h>
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_vector.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_vector<Int> input(N, val);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_copy(input.begin(), input.end(), array);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_copy
+ *  \see \c copy
+ *  \see \c uninitialized_fill
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename InputIterator, typename ForwardIterator>
+  ForwardIterator uninitialized_copy(InputIterator first,
+                                     InputIterator last,
+                                     ForwardIterator result);
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a constructor.
+ *  Occasionally, however, it is useful to separate those two operations.
+ *  If each iterator in the range <tt>[result, result + n)</tt> points
+ *  to uninitialized memory, then \p uninitialized_copy_n creates a copy of
+ *  <tt>[first, first + n)</tt> in that range. That is, for each iterator \c i in
+ *  the input, \p uninitialized_copy_n creates a copy of \c *i in the location pointed
+ *  to by the corresponding iterator in the output range by \p InputIterator's
+ *  \c value_type's copy constructor with *i as its argument.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the input range to copy from.
+ *  \param n The number of elements to copy.
+ *  \param result The first element of the output range to copy to.
+ *  \return An iterator pointing to the last element of the output range.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Size is an integral type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that takes
+ *          a single argument whose type is \p InputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, first + n)</tt> and the range <tt>[result, result + n)</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_copy to initialize
+ *  a range of uninitialized memory using the \p thrust::device execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/uninitialized_copy.h>
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/execution_policy.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_vector<Int> input(N, val);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_copy_n(thrust::device, input.begin(), N, array);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_copy
+ *  \see \c uninitialized_copy
+ *  \see \c copy
+ *  \see \c uninitialized_fill
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename DerivedPolicy, typename InputIterator, typename Size, typename ForwardIterator>
+__host__ __device__
+  ForwardIterator uninitialized_copy_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       InputIterator first,
+                                       Size n,
+                                       ForwardIterator result);
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a constructor.
+ *  Occasionally, however, it is useful to separate those two operations.
+ *  If each iterator in the range <tt>[result, result + n)</tt> points
+ *  to uninitialized memory, then \p uninitialized_copy_n creates a copy of
+ *  <tt>[first, first + n)</tt> in that range. That is, for each iterator \c i in
+ *  the input, \p uninitialized_copy_n creates a copy of \c *i in the location pointed
+ *  to by the corresponding iterator in the output range by \p InputIterator's
+ *  \c value_type's copy constructor with *i as its argument.
+ *
+ *  \param first The first element of the input range to copy from.
+ *  \param n The number of elements to copy.
+ *  \param result The first element of the output range to copy to.
+ *  \return An iterator pointing to the last element of the output range.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>.
+ *  \tparam Size is an integral type.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that takes
+ *          a single argument whose type is \p InputIterator's \c value_type.
+ *
+ *  \pre \p first may equal \p result, but the range <tt>[first, first + n)</tt> and the range <tt>[result, result + n)</tt> shall not overlap otherwise.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_copy to initialize
+ *  a range of uninitialized memory.
+ *
+ *  \code
+ *  #include <thrust/uninitialized_copy.h>
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/device_vector.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_vector<Int> input(N, val);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_copy_n(input.begin(), N, array);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_copy
+ *  \see \c uninitialized_copy
+ *  \see \c copy
+ *  \see \c uninitialized_fill
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename InputIterator, typename Size, typename ForwardIterator>
+  ForwardIterator uninitialized_copy_n(InputIterator first,
+                                       Size n,
+                                       ForwardIterator result);
+
+
+/*! \} // copying
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/uninitialized_copy.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_fill.h b/thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_fill.h
new file mode 100644
index 000000000000..b46758a3c8c3
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/uninitialized_fill.h
@@ -0,0 +1,272 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file uninitialized_fill.h
+ *  \brief Copy construction into a range of uninitialized elements from a source value
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup filling
+ *  \ingroup transformations
+ *  \{
+ */
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a
+ *  constructor. Occasionally, however, it is useful to separate those two
+ *  operations. If each iterator in the range <tt>[first, last)</tt> points
+ *  to uninitialized memory, then \p uninitialized_fill creates copies of \c x
+ *  in that range. That is, for each iterator \c i in the range <tt>[first, last)</tt>,
+ *  \p uninitialized_fill creates a copy of \c x in the location pointed to \c i by
+ *  calling \p ForwardIterator's \c value_type's copy constructor.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *  
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the range of interest.
+ *  \param last The last element of the range of interest.
+ *  \param x The value to use as the exemplar of the copy constructor.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that
+ *          takes a single argument of type \p T.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_fill to initialize a range of
+ *  uninitialized memory using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/uninitialized_fill.h>
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/execution_policy.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_fill(thrust::device, array, array + N, val);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_fill
+ *  \see \c uninitialized_fill_n
+ *  \see \c fill
+ *  \see \c uninitialized_copy
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename T>
+__host__ __device__
+  void uninitialized_fill(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                          ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x);
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a
+ *  constructor. Occasionally, however, it is useful to separate those two
+ *  operations. If each iterator in the range <tt>[first, last)</tt> points
+ *  to uninitialized memory, then \p uninitialized_fill creates copies of \c x
+ *  in that range. That is, for each iterator \c i in the range <tt>[first, last)</tt>,
+ *  \p uninitialized_fill creates a copy of \c x in the location pointed to \c i by
+ *  calling \p ForwardIterator's \c value_type's copy constructor.
+ *  
+ *  \param first The first element of the range of interest.
+ *  \param last The last element of the range of interest.
+ *  \param x The value to use as the exemplar of the copy constructor.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that
+ *          takes a single argument of type \p T.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_fill to initialize a range of
+ *  uninitialized memory.
+ *
+ *  \code
+ *  #include <thrust/uninitialized_fill.h>
+ *  #include <thrust/device_malloc.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_fill(array, array + N, val);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_fill
+ *  \see \c uninitialized_fill_n
+ *  \see \c fill
+ *  \see \c uninitialized_copy
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename ForwardIterator, typename T>
+  void uninitialized_fill(ForwardIterator first,
+                          ForwardIterator last,
+                          const T &x);
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a
+ *  constructor. Occasionally, however, it is useful to separate those two
+ *  operations. If each iterator in the range <tt>[first, first+n)</tt> points
+ *  to uninitialized memory, then \p uninitialized_fill creates copies of \c x
+ *  in that range. That is, for each iterator \c i in the range <tt>[first, first+n)</tt>,
+ *  \p uninitialized_fill creates a copy of \c x in the location pointed to \c i by
+ *  calling \p ForwardIterator's \c value_type's copy constructor.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *  
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The first element of the range of interest.
+ *  \param n The size of the range of interest.
+ *  \param x The value to use as the exemplar of the copy constructor.
+ *  \return <tt>first+n</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that
+ *          takes a single argument of type \p T.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_fill to initialize a range of
+ *  uninitialized memory using the \p thrust::device execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/uninitialized_fill.h>
+ *  #include <thrust/device_malloc.h>
+ *  #include <thrust/execution_policy.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_fill_n(thrust::device, array, N, val);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_fill
+ *  \see \c uninitialized_fill
+ *  \see \c fill
+ *  \see \c uninitialized_copy_n
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename DerivedPolicy, typename ForwardIterator, typename Size, typename T>
+__host__ __device__
+  ForwardIterator uninitialized_fill_n(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                                       ForwardIterator first,
+                                       Size n,
+                                       const T &x);
+
+
+/*! In \c thrust, the function \c thrust::device_new allocates memory for
+ *  an object and then creates an object at that location by calling a
+ *  constructor. Occasionally, however, it is useful to separate those two
+ *  operations. If each iterator in the range <tt>[first, first+n)</tt> points
+ *  to uninitialized memory, then \p uninitialized_fill creates copies of \c x
+ *  in that range. That is, for each iterator \c i in the range <tt>[first, first+n)</tt>,
+ *  \p uninitialized_fill creates a copy of \c x in the location pointed to \c i by
+ *  calling \p ForwardIterator's \c value_type's copy constructor.
+ *  
+ *  \param first The first element of the range of interest.
+ *  \param n The size of the range of interest.
+ *  \param x The value to use as the exemplar of the copy constructor.
+ *  \return <tt>first+n</tt>
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          \p ForwardIterator is mutable, and \p ForwardIterator's \c value_type has a constructor that
+ *          takes a single argument of type \p T.
+ *
+ *  The following code snippet demonstrates how to use \p uninitialized_fill to initialize a range of
+ *  uninitialized memory.
+ *
+ *  \code
+ *  #include <thrust/uninitialized_fill.h>
+ *  #include <thrust/device_malloc.h>
+ *  
+ *  struct Int
+ *  {
+ *    __host__ __device__
+ *    Int(int x) : val(x) {}
+ *    int val;
+ *  };  
+ *  ...
+ *  const int N = 137;
+ *
+ *  Int val(46);
+ *  thrust::device_ptr<Int> array = thrust::device_malloc<Int>(N);
+ *  thrust::uninitialized_fill_n(array, N, val);
+ *
+ *  // Int x = array[i];
+ *  // x.val == 46 for all 0 <= i < N
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/memory/uninitialized_fill
+ *  \see \c uninitialized_fill
+ *  \see \c fill
+ *  \see \c uninitialized_copy_n
+ *  \see \c device_new
+ *  \see \c device_malloc
+ */
+template<typename ForwardIterator, typename Size, typename T>
+  ForwardIterator uninitialized_fill_n(ForwardIterator first,
+                                       Size n,
+                                       const T &x);
+
+/*! \} // end filling
+ *  \} // transformations
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/uninitialized_fill.inl>
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/unique.h b/thirdparty/manifold/thirdparty/thrust/thrust/unique.h
new file mode 100644
index 000000000000..234cd4935693
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/unique.h
@@ -0,0 +1,1140 @@
+/*
+ *  Copyright 2008-2013 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file unique.h
+ *  \brief Move unique elements to the front of a range
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/execution_policy.h>
+#include <thrust/iterator/iterator_traits.h>
+#include <thrust/pair.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup stream_compaction
+ *  \{
+ */
+
+
+/*! For each group of consecutive elements in the range <tt>[first, last)</tt>
+ *  with the same value, \p unique removes all but the first element of 
+ *  the group. The return value is an iterator \c new_last such that 
+ *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
+ *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
+ *  dereferenceable, but the elements that they point to are unspecified.
+ *  \p unique is stable, meaning that the relative order of elements that are
+ *  not removed is unchanged.
+ *
+ *  This version of \p unique uses \c operator== to test for equality.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \return The end of the unique range <tt>[first, new_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique to
+ *  compact a sequence of numbers to remove consecutive duplicates using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int *new_end = thrust::unique(thrust::host, A, A + N);
+ *  // The first four values of A are now {1, 3, 2, 1}
+ *  // Values beyond new_end are unspecified.
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique
+ *  \see unique_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+ForwardIterator unique(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       ForwardIterator first,
+                       ForwardIterator last);
+
+
+/*! For each group of consecutive elements in the range <tt>[first, last)</tt>
+ *  with the same value, \p unique removes all but the first element of 
+ *  the group. The return value is an iterator \c new_last such that 
+ *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
+ *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
+ *  dereferenceable, but the elements that they point to are unspecified.
+ *  \p unique is stable, meaning that the relative order of elements that are
+ *  not removed is unchanged.
+ *
+ *  This version of \p unique uses \c operator== to test for equality.
+ *
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \return The end of the unique range <tt>[first, new_last)</tt>.
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique to
+ *  compact a sequence of numbers to remove consecutive duplicates.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int *new_end = thrust::unique(A, A + N);
+ *  // The first four values of A are now {1, 3, 2, 1}
+ *  // Values beyond new_end are unspecified.
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique
+ *  \see unique_copy
+ */
+template<typename ForwardIterator>
+ForwardIterator unique(ForwardIterator first,
+                       ForwardIterator last);
+
+
+/*! For each group of consecutive elements in the range <tt>[first, last)</tt>
+ *  with the same value, \p unique removes all but the first element of 
+ *  the group. The return value is an iterator \c new_last such that 
+ *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
+ *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
+ *  dereferenceable, but the elements that they point to are unspecified.
+ *  \p unique is stable, meaning that the relative order of elements that are
+ *  not removed is unchanged.
+ *
+ *  This version of \p unique uses the function object \p binary_pred to test
+ *  for equality.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The end of the unique range <tt>[first, new_last)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type and to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique to
+ *  compact a sequence of numbers to remove consecutive duplicates using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int *new_end = thrust::unique(thrust::host, A, A + N, thrust::equal_to<int>());
+ *  // The first four values of A are now {1, 3, 2, 1}
+ *  // Values beyond new_end are unspecified.
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique
+ *  \see unique_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+ForwardIterator unique(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       ForwardIterator first,
+                       ForwardIterator last,
+                       BinaryPredicate binary_pred);
+
+
+/*! For each group of consecutive elements in the range <tt>[first, last)</tt>
+ *  with the same value, \p unique removes all but the first element of 
+ *  the group. The return value is an iterator \c new_last such that 
+ *  no two consecutive elements in the range <tt>[first, new_last)</tt> are
+ *  equal. The iterators in the range <tt>[new_last, last)</tt> are all still
+ *  dereferenceable, but the elements that they point to are unspecified.
+ *  \p unique is stable, meaning that the relative order of elements that are
+ *  not removed is unchanged.
+ *
+ *  This version of \p unique uses the function object \p binary_pred to test
+ *  for equality.
+ *
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The end of the unique range <tt>[first, new_last)</tt>
+ *
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator is mutable,
+ *          and \p ForwardIterator's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type and to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique to
+ *  compact a sequence of numbers to remove consecutive duplicates.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int *new_end = thrust::unique(A, A + N, thrust::equal_to<int>());
+ *  // The first four values of A are now {1, 3, 2, 1}
+ *  // Values beyond new_end are unspecified.
+ *  \endcode
+ *
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique
+ *  \see unique_copy
+ */
+template<typename ForwardIterator,
+         typename BinaryPredicate>
+ForwardIterator unique(ForwardIterator first,
+                       ForwardIterator last,
+                       BinaryPredicate binary_pred);
+
+
+/*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
+ * to a range beginning with \p result, except that in a consecutive group
+ * of duplicate elements only the first one is copied. The return value
+ * is the end of the range to which the elements are copied. 
+ *
+ * The reason there are two different versions of unique_copy is that there
+ * are two different definitions of what it means for a consecutive group of
+ * elements to be duplicates. In the first version, the test is simple
+ * equality: the elements in a range <tt>[f, l)</tt> are duplicates if,
+ * for every iterator \p i in the range, either <tt>i == f</tt> or else 
+ * <tt>*i == *(i-1)</tt>. In the second, the test is an arbitrary 
+ * \p BinaryPredicate \p binary_pred: the elements in <tt>[f, l)</tt> are
+ * duplicates if, for every iterator \p i in the range, either <tt>i == f</tt>
+ * or else <tt>binary_pred(*i, *(i-1))</tt> is \p true.
+ *
+ * This version of \p unique_copy uses \c operator== to test for equality.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the unique range <tt>[result, result_end)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The range <tt>[first,last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_copy to
+ *  compact a sequence of numbers to remove consecutive duplicates using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int B[N];
+ *  int *result_end = thrust::unique_copy(thrust::host, A, A + N, B);
+ *  // The first four values of B are now {1, 3, 2, 1} and (result_end - B) is 4
+ *  // Values beyond result_end are unspecified
+ *  \endcode
+ *
+ *  \see unique
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique_copy
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator>
+__host__ __device__
+OutputIterator unique_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator result);
+
+
+/*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
+ * to a range beginning with \p result, except that in a consecutive group
+ * of duplicate elements only the first one is copied. The return value
+ * is the end of the range to which the elements are copied. 
+ *
+ * The reason there are two different versions of unique_copy is that there
+ * are two different definitions of what it means for a consecutive group of
+ * elements to be duplicates. In the first version, the test is simple
+ * equality: the elements in a range <tt>[f, l)</tt> are duplicates if,
+ * for every iterator \p i in the range, either <tt>i == f</tt> or else 
+ * <tt>*i == *(i-1)</tt>. In the second, the test is an arbitrary 
+ * \p BinaryPredicate \p binary_pred: the elements in <tt>[f, l)</tt> are
+ * duplicates if, for every iterator \p i in the range, either <tt>i == f</tt>
+ * or else <tt>binary_pred(*i, *(i-1))</tt> is \p true.
+ *
+ * This version of \p unique_copy uses \c operator== to test for equality.
+ *
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \return The end of the unique range <tt>[result, result_end)</tt>.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *
+ *  \pre The range <tt>[first,last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_copy to
+ *  compact a sequence of numbers to remove consecutive duplicates.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int B[N];
+ *  int *result_end = thrust::unique_copy(A, A + N, B);
+ *  // The first four values of B are now {1, 3, 2, 1} and (result_end - B) is 4
+ *  // Values beyond result_end are unspecified
+ *  \endcode
+ *
+ *  \see unique
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique_copy
+ */
+template<typename InputIterator,
+         typename OutputIterator>
+OutputIterator unique_copy(InputIterator first,
+                           InputIterator last,
+                           OutputIterator result);
+
+
+/*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
+ * to a range beginning with \p result, except that in a consecutive group
+ * of duplicate elements only the first one is copied. The return value
+ * is the end of the range to which the elements are copied. 
+ *
+ * This version of \p unique_copy uses the function object \c binary_pred 
+ * to test for equality.
+ *
+ * The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The end of the unique range <tt>[result, result_end)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The range <tt>[first,last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_copy to
+ *  compact a sequence of numbers to remove consecutive duplicates using the \p thrust::host execution
+ *  policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int B[N];
+ *  int *result_end = thrust::unique_copy(thrust::host, A, A + N, B, thrust::equal_to<int>());
+ *  // The first four values of B are now {1, 3, 2, 1} and (result_end - B) is 4
+ *  // Values beyond result_end are unspecified.
+ *  \endcode
+ *
+ *  \see unique
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique_copy
+ */
+template<typename DerivedPolicy,
+         typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+__host__ __device__
+OutputIterator unique_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                           InputIterator first,
+                           InputIterator last,
+                           OutputIterator result,
+                           BinaryPredicate binary_pred);
+                       
+
+/*! \p unique_copy copies elements from the range <tt>[first, last)</tt>
+ * to a range beginning with \p result, except that in a consecutive group
+ * of duplicate elements only the first one is copied. The return value
+ * is the end of the range to which the elements are copied. 
+ *
+ * This version of \p unique_copy uses the function object \c binary_pred 
+ * to test for equality.
+ *
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param result The beginning of the output range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The end of the unique range <tt>[result, result_end)</tt>.
+ *
+ *  \tparam InputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *          and \p InputIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam OutputIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator's \c value_type is convertible to \c OutputIterator's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The range <tt>[first,last)</tt> and the range <tt>[result, result + (last - first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_copy to
+ *  compact a sequence of numbers to remove consecutive duplicates.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int B[N];
+ *  int *result_end = thrust::unique_copy(A, A + N, B, thrust::equal_to<int>());
+ *  // The first four values of B are now {1, 3, 2, 1} and (result_end - B) is 4
+ *  // Values beyond result_end are unspecified.
+ *  \endcode
+ *
+ *  \see unique
+ *  \see https://en.cppreference.com/w/cpp/algorithm/unique_copy
+ */
+template<typename InputIterator,
+         typename OutputIterator,
+         typename BinaryPredicate>
+OutputIterator unique_copy(InputIterator first,
+                           InputIterator last,
+                           OutputIterator result,
+                           BinaryPredicate binary_pred);
+
+
+/*! \p unique_by_key is a generalization of \p unique to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key removes all but the first element of 
+ *  the group.  Similarly, the corresponding values in the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  are also removed.
+ *
+ *  The return value is a \p pair of iterators <tt>(new_keys_last,new_values_last)</tt>
+ *  such that no two consecutive elements in the range <tt>[keys_first, new_keys_last)</tt>
+ *  are equal.
+ *
+ *  This version of \p unique_by_key uses \c operator== to test for equality and 
+ *  \c project1st to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the key range.
+ *  \param keys_last  The end of the key range.
+ *  \param values_first The beginning of the value range.
+ *  \return A pair of iterators at end of the ranges <tt>[key_first, keys_new_last)</tt> and <tt>[values_first, values_new_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator1 is mutable,
+ *          and \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam ForwardIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator2 is mutable.
+ *
+ *  \pre The range <tt>[keys_first, keys_last)</tt> and the range <tt>[values_first, values_first + (keys_last - keys_first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key to
+ *  compact a sequence of key/value pairs to remove consecutive duplicates using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  new_end = thrust::unique_by_key(thrust::host, A, A + N, B);
+ *
+ *  // The first four keys in A are now {1, 3, 2, 1} and new_end.first - A is 4.
+ *  // The first four values in B are now {9, 8, 5, 3} and new_end.second - B is 4.
+ *  \endcode
+ *
+ *  \see unique
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+  unique_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                ForwardIterator1 keys_first, 
+                ForwardIterator1 keys_last,
+                ForwardIterator2 values_first);
+
+
+/*! \p unique_by_key is a generalization of \p unique to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key removes all but the first element of 
+ *  the group.  Similarly, the corresponding values in the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  are also removed.
+ *
+ *  The return value is a \p pair of iterators <tt>(new_keys_last,new_values_last)</tt>
+ *  such that no two consecutive elements in the range <tt>[keys_first, new_keys_last)</tt>
+ *  are equal.
+ *
+ *  This version of \p unique_by_key uses \c operator== to test for equality and 
+ *  \c project1st to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the key range.
+ *  \param keys_last  The end of the key range.
+ *  \param values_first The beginning of the value range.
+ *  \return A pair of iterators at end of the ranges <tt>[key_first, keys_new_last)</tt> and <tt>[values_first, values_new_last)</tt>.
+ *
+ *  \tparam ForwardIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator1 is mutable,
+ *          and \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam ForwardIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator2 is mutable.
+ *
+ *  \pre The range <tt>[keys_first, keys_last)</tt> and the range <tt>[values_first, values_first + (keys_last - keys_first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key to
+ *  compact a sequence of key/value pairs to remove consecutive duplicates.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  new_end = thrust::unique_by_key(A, A + N, B);
+ *
+ *  // The first four keys in A are now {1, 3, 2, 1} and new_end.first - A is 4.
+ *  // The first four values in B are now {9, 8, 5, 3} and new_end.second - B is 4.
+ *  \endcode
+ *
+ *  \see unique
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key
+ */
+template<typename ForwardIterator1,
+         typename ForwardIterator2>
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+  unique_by_key(ForwardIterator1 keys_first, 
+                ForwardIterator1 keys_last,
+                ForwardIterator2 values_first);
+
+
+/*! \p unique_by_key is a generalization of \p unique to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key removes all but the first element of 
+ *  the group.  Similarly, the corresponding values in the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  are also removed.
+ *
+ *  This version of \p unique_by_key uses the function object \c binary_pred
+ *  to test for equality and \c project1st to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the key range.
+ *  \param keys_last  The end of the key range.
+ *  \param values_first The beginning of the value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The end of the unique range <tt>[first, new_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator1 is mutable,
+ *          and \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam ForwardIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator2 is mutable.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The range <tt>[keys_first, keys_last)</tt> and the range <tt>[values_first, values_first + (keys_last - keys_first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key to
+ *  compact a sequence of key/value pairs to remove consecutive duplicates using the \p thrust::host
+ *  execution policy for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  new_end = thrust::unique_by_key(thrust::host, keys, keys + N, values, binary_pred);
+ *
+ *  // The first four keys in A are now {1, 3, 2, 1} and new_end.first - A is 4.
+ *  // The first four values in B are now {9, 8, 5, 3} and new_end.second - B is 4.
+ *  \endcode
+ *
+ *  \see unique
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+    unique_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                  ForwardIterator1 keys_first, 
+                  ForwardIterator1 keys_last,
+                  ForwardIterator2 values_first,
+                  BinaryPredicate binary_pred);
+
+
+/*! \p unique_by_key is a generalization of \p unique to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key removes all but the first element of 
+ *  the group.  Similarly, the corresponding values in the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> 
+ *  are also removed.
+ *
+ *  This version of \p unique_by_key uses the function object \c binary_pred
+ *  to test for equality and \c project1st to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the key range.
+ *  \param keys_last  The end of the key range.
+ *  \param values_first The beginning of the value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The end of the unique range <tt>[first, new_last)</tt>.
+ *
+ *  \tparam ForwardIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator1 is mutable,
+ *          and \p ForwardIterator's \c value_type is a model of <a href="https://en.cppreference.com/w/cpp/concepts/equality_comparable">Equality Comparable</a>.
+ *  \tparam ForwardIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator2 is mutable.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The range <tt>[keys_first, keys_last)</tt> and the range <tt>[values_first, values_first + (keys_last - keys_first))</tt> shall not overlap.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key to
+ *  compact a sequence of key/value pairs to remove consecutive duplicates.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  new_end = thrust::unique_by_key(keys, keys + N, values, binary_pred);
+ *
+ *  // The first four keys in A are now {1, 3, 2, 1} and new_end.first - A is 4.
+ *  // The first four values in B are now {9, 8, 5, 3} and new_end.second - B is 4.
+ *  \endcode
+ *
+ *  \see unique
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key
+ */
+template<typename ForwardIterator1,
+         typename ForwardIterator2,
+         typename BinaryPredicate>
+  thrust::pair<ForwardIterator1,ForwardIterator2>
+  unique_by_key(ForwardIterator1 keys_first, 
+                ForwardIterator1 keys_last,
+                ForwardIterator2 values_first,
+                BinaryPredicate binary_pred);
+
+
+/*! \p unique_by_key_copy is a generalization of \p unique_copy to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key_copy copies the first element of the group to
+ *  a range beginning with \c keys_result and the corresponding values from the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> are copied to a range
+ *  beginning with \c values_result.
+ *
+ *  This version of \p unique_by_key_copy uses \c operator== to test for equality and
+ *  \c project1st to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_result The beginning of the output key range.
+ *  \param values_result The beginning of the output value range.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_result, keys_result_last)</tt> and <tt>[values_result, values_result_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key_copy to
+ *  compact a sequence of key/value pairs and with equal keys using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  new_end = thrust::unique_by_key_copy(thrust::host, A, A + N, B, C, D);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 8, 5, 3} and new_end.second - D is 4.
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see reduce_by_key
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator1 keys_first, 
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_result,
+                       OutputIterator2 values_result);
+
+
+/*! \p unique_by_key_copy is a generalization of \p unique_copy to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key_copy copies the first element of the group to
+ *  a range beginning with \c keys_result and the corresponding values from the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> are copied to a range
+ *  beginning with \c values_result.
+ *
+ *  This version of \p unique_by_key_copy uses \c operator== to test for equality and
+ *  \c project1st to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_result The beginning of the output key range.
+ *  \param values_result The beginning of the output value range.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_result, keys_result_last)</tt> and <tt>[values_result, values_result_last)</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key_copy to
+ *  compact a sequence of key/value pairs and with equal keys.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  new_end = thrust::unique_by_key_copy(A, A + N, B, C, D);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 8, 5, 3} and new_end.second - D is 4.
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see reduce_by_key
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  unique_by_key_copy(InputIterator1 keys_first, 
+                     InputIterator1 keys_last,
+                     InputIterator2 values_first,
+                     OutputIterator1 keys_result,
+                     OutputIterator2 values_result);
+
+
+/*! \p unique_by_key_copy is a generalization of \p unique_copy to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key_copy copies the first element of the group to
+ *  a range beginning with \c keys_result and the corresponding values from the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> are copied to a range
+ *  beginning with \c values_result.
+ *
+ *  This version of \p unique_by_key_copy uses the function object \c binary_pred
+ *  to test for equality and \c project1st to reduce values with equal keys.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_result The beginning of the output key range.
+ *  \param values_result The beginning of the output value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_result, keys_result_last)</tt> and <tt>[values_result, values_result_last)</tt>.
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key_copy to
+ *  compact a sequence of key/value pairs and with equal keys using the \p thrust::host execution policy for
+ *  parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  new_end = thrust::unique_by_key_copy(thrust::host, A, A + N, B, C, D, binary_pred);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 8, 5, 3} and new_end.second - D is 4.
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see reduce_by_key
+ */
+template<typename DerivedPolicy,
+         typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+__host__ __device__
+  thrust::pair<OutputIterator1,OutputIterator2>
+    unique_by_key_copy(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                       InputIterator1 keys_first, 
+                       InputIterator1 keys_last,
+                       InputIterator2 values_first,
+                       OutputIterator1 keys_result,
+                       OutputIterator2 values_result,
+                       BinaryPredicate binary_pred);
+
+
+/*! \p unique_by_key_copy is a generalization of \p unique_copy to key-value pairs.
+ *  For each group of consecutive keys in the range <tt>[keys_first, keys_last)</tt>
+ *  that are equal, \p unique_by_key_copy copies the first element of the group to
+ *  a range beginning with \c keys_result and the corresponding values from the range
+ *  <tt>[values_first, values_first + (keys_last - keys_first))</tt> are copied to a range
+ *  beginning with \c values_result.
+ *
+ *  This version of \p unique_by_key_copy uses the function object \c binary_pred
+ *  to test for equality and \c project1st to reduce values with equal keys.
+ *
+ *  \param keys_first The beginning of the input key range.
+ *  \param keys_last  The end of the input key range.
+ *  \param values_first The beginning of the input value range.
+ *  \param keys_result The beginning of the output key range.
+ *  \param values_result The beginning of the output value range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return A pair of iterators at end of the ranges <tt>[keys_result, keys_result_last)</tt> and <tt>[values_result, values_result_last)</tt>.
+ *
+ *  \tparam InputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam InputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/input_iterator">Input Iterator</a>,
+ *  \tparam OutputIterator1 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator1's \c value_type is convertible to \c OutputIterator1's \c value_type.
+ *  \tparam OutputIterator2 is a model of <a href="https://en.cppreference.com/w/cpp/iterator/output_iterator">Output Iterator</a> and
+ *          and \p InputIterator2's \c value_type is convertible to \c OutputIterator2's \c value_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  \pre The input ranges shall not overlap either output range.
+ *
+ *  The following code snippet demonstrates how to use \p unique_by_key_copy to
+ *  compact a sequence of key/value pairs and with equal keys.
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1}; // input keys
+ *  int B[N] = {9, 8, 7, 6, 5, 4, 3}; // input values
+ *  int C[N];                         // output keys
+ *  int D[N];                         // output values
+ *
+ *  thrust::pair<int*,int*> new_end;
+ *  thrust::equal_to<int> binary_pred;
+ *  new_end = thrust::unique_by_key_copy(A, A + N, B, C, D, binary_pred);
+ *
+ *  // The first four keys in C are now {1, 3, 2, 1} and new_end.first - C is 4.
+ *  // The first four values in D are now {9, 8, 5, 3} and new_end.second - D is 4.
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key
+ *  \see reduce_by_key
+ */
+template<typename InputIterator1,
+         typename InputIterator2,
+         typename OutputIterator1,
+         typename OutputIterator2,
+         typename BinaryPredicate>
+  thrust::pair<OutputIterator1,OutputIterator2>
+  unique_by_key_copy(InputIterator1 keys_first, 
+                     InputIterator1 keys_last,
+                     InputIterator2 values_first,
+                     OutputIterator1 keys_result,
+                     OutputIterator2 values_result,
+                     BinaryPredicate binary_pred);
+
+
+/*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
+ *  with the same value, 
+ *
+ *  This version of \p unique_count uses the function object \p binary_pred to test for equality.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The number of runs of equal elements in <tt>[first, new_last)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type and to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique_count to
+ *  determine a number of runs of equal elements using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int count = thrust::unique_count(thrust::host, A, A + N, thrust::equal_to<int>());
+ *  // count is now 4
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred);
+
+
+/*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
+ *  with the same value, 
+ *
+ *  This version of \p unique_count uses \c operator== to test for equality.
+ *
+ *  The algorithm's execution is parallelized as determined by \p exec.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The number of runs of equal elements in <tt>[first, new_last)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type and to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique_count to
+ *  determine the number of runs of equal elements using the \p thrust::host execution policy
+ *  for parallelization:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int count = thrust::unique_count(thrust::host, A, A + N);
+ *  // count is now 4
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key_copy
+ */
+template<typename DerivedPolicy,
+         typename ForwardIterator>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(const thrust::detail::execution_policy_base<DerivedPolicy> &exec,
+                 ForwardIterator first,
+                 ForwardIterator last);
+
+
+/*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
+ *  with the same value, 
+ *
+ *  This version of \p unique_count uses the function object \p binary_pred to test for equality.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The number of runs of equal elements in <tt>[first, new_last)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type and to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique_count to
+ *  determine the number of runs of equal elements:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int count = thrust::unique_count(A, A + N, thrust::equal_to<int>());
+ *  // count is now 4
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key_copy
+ */
+template<typename ForwardIterator,
+         typename BinaryPredicate>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(ForwardIterator first,
+                 ForwardIterator last,
+                 BinaryPredicate binary_pred);
+
+
+/*! \p unique_count counts runs of equal elements in the range <tt>[first, last)</tt>
+ *  with the same value, 
+ *
+ *  This version of \p unique_count uses \c operator== to test for equality.
+ *
+ *  \param exec The execution policy to use for parallelization.
+ *  \param first The beginning of the input range.
+ *  \param last  The end of the input range.
+ *  \param binary_pred  The binary predicate used to determine equality.
+ *  \return The number of runs of equal elements in <tt>[first, new_last)</tt>
+ *
+ *  \tparam DerivedPolicy The name of the derived execution policy.
+ *  \tparam ForwardIterator is a model of <a href="https://en.cppreference.com/w/cpp/iterator/forward_iterator">Forward Iterator</a>,
+ *          and \p ForwardIterator's \c value_type is convertible to \p BinaryPredicate's \c first_argument_type and to \p BinaryPredicate's \c second_argument_type.
+ *  \tparam BinaryPredicate is a model of <a href="https://en.cppreference.com/w/cpp/named_req/BinaryPredicate">Binary Predicate</a>.
+ *
+ *  The following code snippet demonstrates how to use \p unique_count to
+ *  determine the number of runs of equal elements:
+ *
+ *  \code
+ *  #include <thrust/unique.h>
+ *  #include <thrust/execution_policy.h>
+ *  ...
+ *  const int N = 7;
+ *  int A[N] = {1, 3, 3, 3, 2, 2, 1};
+ *  int count = thrust::unique_count(thrust::host, A, A + N);
+ *  // count is now 4
+ *  \endcode
+ *
+ *  \see unique_copy
+ *  \see unique_by_key_copy
+ *  \see reduce_by_key_copy
+ */
+template<typename ForwardIterator>
+__host__ __device__
+  typename thrust::iterator_traits<ForwardIterator>::difference_type
+    unique_count(ForwardIterator first,
+                 ForwardIterator last);
+
+
+/*! \} // end stream_compaction
+ */
+
+THRUST_NAMESPACE_END
+
+#include <thrust/detail/unique.inl>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/universal_allocator.h b/thirdparty/manifold/thirdparty/thrust/thrust/universal_allocator.h
new file mode 100644
index 000000000000..8d85cd20de0d
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/universal_allocator.h
@@ -0,0 +1,77 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file universal_allocator.h
+ *  \brief An allocator which creates new elements in memory accessible to both
+ *         hosts and devices.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+
+// #include the device system's vector header
+#define __THRUST_DEVICE_SYSTEM_MEMORY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/memory.h>
+#include __THRUST_DEVICE_SYSTEM_MEMORY_HEADER
+#undef __THRUST_DEVICE_SYSTEM_MEMORY_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+/** \addtogroup memory_resources Memory Resources
+ *  \ingroup memory_management_classes
+ *  \{
+ */
+
+/*! \brief An allocator which creates new elements in memory accessible by
+ *         both hosts and devices.
+ *
+ *  \see https://en.cppreference.com/w/cpp/named_req/Allocator
+ */
+using thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::universal_allocator;
+
+/*! \p universal_ptr stores a pointer to an object allocated in memory accessible
+ *  to both hosts and devices.
+ *
+ *  Algorithms dispatched with this type of pointer will be dispatched to
+ *  either host or device, depending on which backend you are using. Explicit
+ *  policies (\p thrust::device, etc) can be used to specify where an algorithm
+ *  should be run.
+ *
+ *  \p universal_ptr has pointer semantics: it may be dereferenced safely from
+ *  both hosts and devices and may be manipulated with pointer arithmetic.
+ *
+ *  \p universal_ptr can be created with \p universal_allocator or by explicitly
+ *  calling its constructor with a raw pointer.
+ *
+ *  The raw pointer encapsulated by a \p universal_ptr may be obtained by
+ *  either its <tt>get</tt> method or the \p raw_pointer_cast free function.
+ *
+ *  \note \p universal_ptr is not a smart pointer; it is the programmer's
+ *  responsibility to deallocate memory pointed to by \p universal_ptr.
+ *
+ *  \see host_ptr For the documentation of the complete interface which is
+ *                shared by \p universal_ptr.
+ *  \see raw_pointer_cast
+ */
+template <typename T>
+using universal_ptr =
+  thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::universal_pointer<T>;
+
+/*! \}
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/universal_ptr.h b/thirdparty/manifold/thirdparty/thrust/thrust/universal_ptr.h
new file mode 100644
index 000000000000..9d1de19d5f1a
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/universal_ptr.h
@@ -0,0 +1,26 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+
+/*! \file universal_ptr.h
+ *  \brief A pointer to a variable which resides memory accessible to both
+ *         hosts and devices.
+ */
+
+#pragma once
+
+#include <thrust/universal_allocator.h>
+
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/universal_vector.h b/thirdparty/manifold/thirdparty/thrust/thrust/universal_vector.h
new file mode 100644
index 000000000000..0ce38fd868c2
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/universal_vector.h
@@ -0,0 +1,55 @@
+/*
+ *  Copyright 2008-2020 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file
+ *  \brief A dynamically-sizable array of elements which resides in memory
+ *         accessible to both hosts and devices.
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/universal_allocator.h>
+
+// #include the device system's vector header
+#define __THRUST_DEVICE_SYSTEM_VECTOR_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/vector.h>
+#include __THRUST_DEVICE_SYSTEM_VECTOR_HEADER
+#undef __THRUST_DEVICE_SYSTEM_VECTOR_HEADER
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup containers Containers
+ *  \{
+ */
+
+/*! A \p universal_vector is a container that supports random access to elements,
+ *  constant time removal of elements at the end, and linear time insertion
+ *  and removal of elements at the beginning or in the middle. The number of
+ *  elements in a \p universal_vector may vary dynamically; memory management is
+ *  automatic. The memory associated with a \p universal_vector resides in memory
+ *  accessible to hosts and devices.
+ *
+ *  \see https://en.cppreference.com/w/cpp/container/vector
+ *  \see host_vector For the documentation of the complete interface which is
+ *                   shared by \p universal_vector.
+ *  \see device_vector
+ */
+using thrust::system::__THRUST_DEVICE_SYSTEM_NAMESPACE::universal_vector;
+
+/*! \} // containers
+ */
+
+THRUST_NAMESPACE_END
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/version.h b/thirdparty/manifold/thirdparty/thrust/thrust/version.h
new file mode 100644
index 000000000000..b92a6494ba6e
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/version.h
@@ -0,0 +1,75 @@
+/*
+ *  Copyright 2008-2022 NVIDIA Corporation
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ *  See the License for the specific language governing permissions and
+ *  limitations under the License.
+ */
+
+/*! \file version.h
+ *  \brief Compile-time macros encoding Thrust release version
+ *
+ *         <thrust/version.h> is the only Thrust header that is guaranteed to
+ *         change with every thrust release.
+ *
+ *         It is also the only header that does not cause THRUST_HOST_SYSTEM
+ *         and THRUST_DEVICE_SYSTEM to be defined. This way, a user may include
+ *         this header and inspect THRUST_VERSION before programatically defining
+ *         either of these macros herself.
+ */
+
+#pragma once
+
+//  This is the only Thrust header that is guaranteed to
+//  change with every Thrust release.
+//
+//  THRUST_VERSION % 100 is the sub-minor version
+//  THRUST_VERSION / 100 % 1000 is the minor version
+//  THRUST_VERSION / 100000 is the major version
+//
+//  Because this header does not #include <thrust/detail/config.h>,
+//  it is the only Thrust header that does not cause
+//  THRUST_HOST_SYSTEM and THRUST_DEVICE_SYSTEM to be defined.
+
+/*! \def THRUST_VERSION
+ *  \brief The preprocessor macro \p THRUST_VERSION encodes the version
+ *         number of the Thrust library.
+ *
+ *         <tt>THRUST_VERSION % 100</tt> is the sub-minor version.
+ *         <tt>THRUST_VERSION / 100 % 1000</tt> is the minor version.
+ *         <tt>THRUST_VERSION / 100000</tt> is the major version.
+ */
+#define THRUST_VERSION 200100
+
+/*! \def THRUST_MAJOR_VERSION
+ *  \brief The preprocessor macro \p THRUST_MAJOR_VERSION encodes the
+ *         major version number of the Thrust library.
+ */
+#define THRUST_MAJOR_VERSION     (THRUST_VERSION / 100000)
+
+/*! \def THRUST_MINOR_VERSION
+ *  \brief The preprocessor macro \p THRUST_MINOR_VERSION encodes the
+ *         minor version number of the Thrust library.
+ */
+#define THRUST_MINOR_VERSION     (THRUST_VERSION / 100 % 1000)
+
+/*! \def THRUST_SUBMINOR_VERSION
+ *  \brief The preprocessor macro \p THRUST_SUBMINOR_VERSION encodes the
+ *         sub-minor version number of the Thrust library.
+ */
+#define THRUST_SUBMINOR_VERSION  (THRUST_VERSION % 100)
+
+/*! \def THRUST_PATCH_NUMBER
+ *  \brief The preprocessor macro \p THRUST_PATCH_NUMBER encodes the
+ *         patch number of the Thrust library.
+ *         Legacy; will be 0 for all future releases.
+ */
+#define THRUST_PATCH_NUMBER 0
diff --git a/thirdparty/manifold/thirdparty/thrust/thrust/zip_function.h b/thirdparty/manifold/thirdparty/thrust/thrust/zip_function.h
new file mode 100644
index 000000000000..7653f9b7f3e1
--- /dev/null
+++ b/thirdparty/manifold/thirdparty/thrust/thrust/zip_function.h
@@ -0,0 +1,212 @@
+
+/*! \file thrust/zip_function.h
+ *  \brief Adaptor type that turns an N-ary function object into one that takes
+ *         a tuple of size N so it can easily be used with algorithms taking zip
+ *         iterators
+ */
+
+#pragma once
+
+#include <thrust/detail/config.h>
+#include <thrust/detail/cpp11_required.h>
+#include <thrust/detail/modern_gcc_required.h>
+
+#if THRUST_CPP_DIALECT >= 2011 && !defined(THRUST_LEGACY_GCC)
+
+#include <thrust/tuple.h>
+#include <thrust/type_traits/integer_sequence.h>
+#include <thrust/detail/type_deduction.h>
+
+THRUST_NAMESPACE_BEGIN
+
+/*! \addtogroup function_objects Function Objects
+ *  \{
+ */
+
+/*! \addtogroup function_object_adaptors Function Object Adaptors
+ *  \ingroup function_objects
+ *  \{
+ */
+
+namespace detail {
+namespace zip_detail {
+
+// Add workaround for decltype(auto) on C++11-only compilers:
+#if THRUST_CPP_DIALECT >= 2014
+
+__thrust_exec_check_disable__
+template <typename Function, typename Tuple, std::size_t... Is>
+__host__ __device__
+decltype(auto) apply_impl(Function&& func, Tuple&& args, index_sequence<Is...>)
+{
+  return func(thrust::get<Is>(THRUST_FWD(args))...);
+}
+
+template <typename Function, typename Tuple>
+__host__ __device__
+decltype(auto) apply(Function&& func, Tuple&& args)
+{
+  constexpr auto tuple_size = thrust::tuple_size<typename std::decay<Tuple>::type>::value;
+  return apply_impl(THRUST_FWD(func), THRUST_FWD(args), make_index_sequence<tuple_size>{});
+}
+
+#else // THRUST_CPP_DIALECT
+
+__thrust_exec_check_disable__
+template <typename Function, typename Tuple, std::size_t... Is>
+__host__ __device__
+auto apply_impl(Function&& func, Tuple&& args, index_sequence<Is...>)
+THRUST_DECLTYPE_RETURNS(func(thrust::get<Is>(THRUST_FWD(args))...))
+
+template <typename Function, typename Tuple>
+__host__ __device__
+auto apply(Function&& func, Tuple&& args)
+THRUST_DECLTYPE_RETURNS(
+    apply_impl(
+      THRUST_FWD(func),
+      THRUST_FWD(args),
+      make_index_sequence<
+        thrust::tuple_size<typename std::decay<Tuple>::type>::value>{})
+)
+
+#endif // THRUST_CPP_DIALECT
+
+} // namespace zip_detail
+} // namespace detail
+
+/*! \p zip_function is a function object that allows the easy use of N-ary
+ *  function objects with \p zip_iterators without redefining them to take a
+ *  \p tuple instead of N arguments.
+ *
+ *  This means that if a functor that takes 2 arguments which could be used with
+ *  the \p transform function and \p device_iterators can be extended to take 3
+ *  arguments and \p zip_iterators without rewriting the functor in terms of
+ *  \p tuple.
+ *
+ *  The \p make_zip_function convenience function is provided to avoid having
+ *  to explicitely define the type of the functor when creating a \p zip_function,
+ *  whic is especially helpful when using lambdas as the functor.
+ *
+ *  \code
+ *  #include <thrust/iterator/zip_iterator.h>
+ *  #include <thrust/device_vector.h>
+ *  #include <thrust/transform.h>
+ *  #include <thrust/zip_function.h>
+ *
+ *  struct SumTuple {
+ *    float operator()(Tuple tup) {
+ *      return std::get<0>(tup) + std::get<1>(tup) + std::get<2>(tup);
+ *    }
+ *  };
+ *  struct SumArgs {
+ *    float operator()(float a, float b, float c) {
+ *      return a + b + c;
+ *    }
+ *  };
+ *
+ *  int main() {
+ *    thrust::device_vector<float> A(3);
+ *    thrust::device_vector<float> B(3);
+ *    thrust::device_vector<float> C(3);
+ *    thrust::device_vector<float> D(3);
+ *    A[0] = 0.f; A[1] = 1.f; A[2] = 2.f;
+ *    B[0] = 1.f; B[1] = 2.f; B[2] = 3.f;
+ *    C[0] = 2.f; C[1] = 3.f; C[2] = 4.f;
+ *
+ *    // The following four invocations of transform are equivalent
+ *    // Transform with 3-tuple
+ *    thrust::transform(thrust::make_zip_iterator(thrust::make_tuple(A.begin(), B.begin(), C.begin())),
+ *                      thrust::make_zip_iterator(thrust::make_tuple(A.end(), B.end(), C.end())),
+ *                      D.begin(),
+ *                      SumTuple{});
+ *
+ *    // Transform with 3 parameters
+ *    thrust::zip_function<SumArgs> adapted{};
+ *    thrust::transform(thrust::make_zip_iterator(thrust::make_tuple(A.begin(), B.begin(), C.begin())),
+ *                      thrust::make_zip_iterator(thrust::make_tuple(A.end(), B.end(), C.end())),
+ *                      D.begin(),
+ *                      adapted);
+ *
+ *    // Transform with 3 parameters with convenience function
+ *    thrust::zip_function<SumArgs> adapted{};
+ *    thrust::transform(thrust::make_zip_iterator(thrust::make_tuple(A.begin(), B.begin(), C.begin())),
+ *                      thrust::make_zip_iterator(thrust::make_tuple(A.end(), B.end(), C.end())),
+ *                      D.begin(),
+ *                      thrust::make_zip_function(SumArgs{}));
+ *
+ *    // Transform with 3 parameters with convenience function and lambda
+ *    thrust::zip_function<SumArgs> adapted{};
+ *    thrust::transform(thrust::make_zip_iterator(thrust::make_tuple(A.begin(), B.begin(), C.begin())),
+ *                      thrust::make_zip_iterator(thrust::make_tuple(A.end(), B.end(), C.end())),
+ *                      D.begin(),
+ *                      thrust::make_zip_function([] (float a, float b, float c) {
+ *                                                  return a + b + c;
+ *                                                }));
+ *    return 0;
+ *  }
+ *  \endcode
+ *
+ *  \see make_zip_function
+ *  \see zip_iterator
+ */
+template <typename Function>
+class zip_function
+{
+  public:
+     __host__ __device__
+    zip_function(Function func) : func(std::move(func)) {}
+
+// Add workaround for decltype(auto) on C++11-only compilers:
+#if THRUST_CPP_DIALECT >= 2014
+
+    template <typename Tuple>
+    __host__ __device__
+    decltype(auto) operator()(Tuple&& args) const
+    {
+        return detail::zip_detail::apply(func, THRUST_FWD(args));
+    }
+
+#else // THRUST_CPP_DIALECT
+
+    // Can't just use THRUST_DECLTYPE_RETURNS here since we need to use
+    // std::declval for the signature components:
+    template <typename Tuple>
+    __host__ __device__
+    auto operator()(Tuple&& args) const
+    noexcept(noexcept(detail::zip_detail::apply(std::declval<Function>(), THRUST_FWD(args))))
+    THRUST_TRAILING_RETURN(decltype(detail::zip_detail::apply(std::declval<Function>(), THRUST_FWD(args))))
+    {
+        return detail::zip_detail::apply(func, THRUST_FWD(args));
+    }
+
+#endif // THRUST_CPP_DIALECT
+
+  private:
+    mutable Function func;
+};
+
+/*! \p make_zip_function creates a \p zip_function from a function object.
+ *
+ *  \param fun The N-ary function object.
+ *  \return A \p zip_function that takes a N-tuple.
+ *
+ *  \see zip_function
+ */
+template <typename Function>
+__host__ __device__
+zip_function<typename std::decay<Function>::type>
+make_zip_function(Function&& fun)
+{
+    using func_t = typename std::decay<Function>::type;
+    return zip_function<func_t>(THRUST_FWD(fun));
+}
+
+/*! \} // end function_object_adaptors
+ */
+
+/*! \} // end function_objects
+ */
+
+THRUST_NAMESPACE_END
+
+#endif